Diffusion in Silicon - 10 Years of Research

Diffusion in Silicon10 Years of Research

Editor:

D.J. Fisher

SCITEC PUBLICATIONS

Notes:Each item in this section of the volume begins with a graphical compilation of relevant diffusion data whichhave been reported during the past decade. The plotted data are also tabulated as indicated on the graph. In somecases, the tabulated data have been obtained by digitizing published graphs and the values may not correspondexactly with the author's unpublished raw data.

3N Bulk Diffusion - Quantitative DataThe migration of Ag from epitaxial layers and into (111) samples of Si,during annealing at temperatures of between 450 and 500C, was studiedby means of secondary ion mass spectrometric depth profiling. It wasfound that the diffusivities lay between 8 x 10 -16 and 1.6 x 10-15cm2/s(table N). These values were lower than were expected on the basis ofprevious data.T.C.Nason, G.R.Yang, K.H.Park, T.M.Lu: Journal of Applied Physics,1991, 70[3], 1392-6

[446-91/92-027]

Refers to table N

Indicates volume and page number inDDF where abstract first appeared

217

Ag

Figure 1: Diffusivity of Ag in Si

31 Bulk Diffusion - Quantitative DataTransition metals in amorphous samples exhibit a direct interstitial diffusion behaviorwhich is retarded by temporary trapping at defects that are intrinsic to the amorphousstructure. Diffusion was investigated here by means of Rutherford back-scatteringspectrometry. It was found that the data (table 1) could be fitted by using foreign-atominterstitial diffusion coefficients for crystalline Si; modified by the presence of traps inconcentrations of between 0.2 and 1at%, and with trapping enthalpies of about 0.9eV.The results could be expressed as:

D (cm2/s) = 0.16 exp[-1.67(eV)/kT]

1.0E-16

1.0E-15

1.0E-14

1.0E-13

1.0E-12

1.0E-11

12 13 14 15 16 17 18

table 1table 2

104/T(K)

D (cm2/s)

218

Ag Bulk Diffusion Ag

S.Coffa, J.M.Poate, D.C.Jacobson, W.Frank, W.Gustin: Physical Review B, 1992,45[15], 8355-8

[446-86/87-049]

Table 1Diffusivity of Ag in Amorphous Si

T (C) D (cm2/s)485 1.2 x 10-12

400 4.2 x 10-14

355 4.1 x 10-15

305 4.5 x 10-16

32 Bulk Diffusion - Quantitative DataThe migration of Ag from epitaxial layers and into (111) samples of Si, during annealingat temperatures of between 450 and 500C, was studied by means of secondary ion massspectrometric depth profiling. It was found that the diffusivities lay between 8 x 10-16 and1.6 x 10-15cm2/s (table 2). These values were lower than were expected on the basis ofprevious data.T.C.Nason, G.R.Yang, K.H.Park, T.M.Lu: Journal of Applied Physics, 1991, 70[3],1392-6

[446-91/92-027]

Bulk Diffusion - Quantitative DataConcentration versus depth profiles of Ag were measured by using neutron activationanalysis and serial sectioning. The Ag diffusion appeared to be very fast. In the bulk ofdislocation-free wafers, saturation was achieved after short periods of annealing. Fromthis, it was concluded that interstitial Ag was the predominant configuration in Si withoutdislocations. Equilibrium concentrations of Ag were determined for temperatures ofbetween 1287 and 1598K. The results were thermodynamically analyzed, taking accountof Ag-Si liquidus data. In dislocated Si, much higher Ag concentrations were found whichvaried irregularly with penetration depth. A comparison of the diffusion and solubility ofAg and Au in Si suggested that, in material with dislocations, substitutional Ag couldarise from Agi-Ags transitions. Finally, the Agi diffusivity was deduced to be given by:

D(m2/s) = 6 x 10-5 exp[-1.15(eV)/kT]F.Rollert, N.A.Stolwijk, H.Mehrer: Journal of Physics D, 1987, 20[9], 1148-55

[446-55/56-033]

Bulk Diffusion - Qualitative Observations - Amorphous FilmsAn investigation of the diffusion and solubility of Ag in hydrogenated amorphous filmswas carried out at temperatures ranging from 290 to 525C. This made it possible todetermine suitable conditions for diffusion doping. Such diffusion was used to preparehydrogenated amorphous films which were doped with this metal, and a study was made

219

Ag Surface Diffusion Ag

of their dark conductivity and photoconductivity. It was noted that diffusion doping withmetals gave rise to activation energies, for electrical conduction, which were as high as1.4eV. That is, the activation energy exceeded 50% of the band gap; atypical behavior foran intrinsic material.M.S.Ablova, G.S.Kulikov, S.K.Persheev, K.K.Khodzhaev: Fizika i TekhnikaPoluprovodnikov, 1990, 24[11], 1943-7 (Soviet Physics - Semiconductors, 1990, 24[11],1208-11)

[446-81/82-043]

Table 2Diffusion of Ag into (111)Si

Temperature (C) Surface Concentration(/cm3) D (cm2/s)450 6.5 x 1019 1.5 x 10-15

450 2.9 x 1019 8.0 x 10-16

500 4.0 x 1020 1.6 x 10-15

Bulk Diffusion - Qualitative Observations - Effect of Surface LayerThe behavior of Ag atoms at the interface between a deposited Si layer and a Si(111)surface was studied by using a new ion-scattering spectroscopic technique. During theroom temperature deposition of Si, the Ag layer was anchored at the interface betweenamorphous and crystalline Si. With increasing annealing temperature, the Ag atoms werefound to diffuse out over the adsorbed amorphous Si layer. The buried structure of thesubstrate was also reflected by the multiple scattering component of the ion-scatteringspectrum.K.Kawamoto, T.Mori, S.Kujime, K.Oura: Surface Science, 1996, 363[1-3], 156-60

[446-141/142-113]

Surface Diffusion - Quantitative DataMass transport on the (111) surface was studied by means of scanning electronmicroscopy and Auger analysis (with spatial resolution) under ultra-high vacuumconditions. The spreading of Ag deposits was investigated at temperatures ranging from350 to 450C; where no desorption occurred. In order to avoid electromigration, thesamples were heated by using a halogen lamp. When the first islands had formed(Stranski-Krastanov growth), Ag began to spread out of the initial deposit zone. The mainfeatures deduced from the concentration profiles were that, at temperatures above about400C, the profiles exhibited a rather constant concentration that ended in a very sharpfront (attributed to an unrolling carpet mechanism). At temperatures below about 400C,the corresponding profiles had 2 gradient zones. In both cases, a t¾ kinetic law was foundwhich suggested that Ag/Si mass transport might be controlled by the surface self-diffusion of Ag atoms, on 3-dimensional Ag islands, with an activation energy of about2.4eV/atom.N.Boutaoui, H.Roux, M.Tholomier: Surface Science, 1990, 239[3], 213-21

[446-78/79-047]

220

Ag Surface Diffusion Ag

Surface Diffusion - Qualitative Observations - Effect of StepsThe mass transport of Ag on stepped (111) surfaces was investigated by means ofscanning Auger microscopy. Highly anisotropic surface diffusion of Ag ultra-thin filmswas observed on 0º, 0.5º, 3º and 6º vicinal (111). Mass transport in a direction which wasparallel to the step edge was overwhelmingly greater than that perpendicular to the edge.Preferential mass transport towards the cathode, due to direct current heating, was alsoobserved. This increased with the resistivity of the (111) substrate. The anisotropic masstransport was related to the difference in the binding energies at the step edge sites and atthe terrace sites.N.J.Wu, A.Natori, H.Yasunaga: Surface Science, 1991, 242[1-3], 191-4

[446-81/82-043]

221

AlBulk Diffusion - Quantitative DataMeasurements were made of the concentration profiles of Al atoms which had beenintroduced into p-type material by irradiation with a Nd:YAG laser in the continuouswave or Q-switched modes. It was found that limiting concentrations of Al atoms in theSi substrate were attained during irradiation under continuous wave conditions. Theconcentrations which were measured under irradiation in the Q-switched mode were up to2 orders of magnitude higher than those which resulted from continuous wave operation.It was found that the volume diffusivity could be described by:

D(cm2/s) = 8.0 exp[-3.47(eV)/kT]It was also deduced that pipe diffusion occurred which could be described by:

D(cm2/s) = 140 exp[-3.01(eV)/kT]D.Demireva, B.Lämmel: Journal of Physics D, 1997, 30[14], 1972-5

[446-152-0221]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesA study was reported of the vapor-phase diffusion of Al into single crystals of n-type(111) float-zone material. The resultant diffusion profiles were measured by usingresistivity techniques. No specific diffusion data were presented, but qualitativeobservations were described. Thus, diffusion temperatures greater than 1150C yieldedlower surface concentrations. This was partly due to an increase in Al vapor oxidation. Atlower pressures and lower diffusion temperatures, the partial pressure increased to theextent that the surface concentration in the Si corresponded to the solid solubility of theimpurity. Under these conditions, the surface concentration was independent of the Alvapor pressure and was reproducible and uniform; even if the vapor pressure was not.S.Azimi-Nam: Journal of Materials Science Letters, 1987, 6[9], 1073-5

[446-61-084]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesDiffusion profiles of Al were produced under rapid thermal annealing conditions. Theeffect of P upon the diffusion behavior of Al was investigated by the pre-deposition of Pinto Al-diffused wafers, and vice versa. The dopant profiles were determined by meansof secondary-ion mass spectroscopy, electrochemical, capacitance-voltage and spreadingresistance methods. The profiles after Al pre-deposition at 1293K exhibited Al surface

222

Al Bulk Diffusion Al

concentrations which ranged up to the solid solubility limit of about 2 x 1025/m3. It wasshown that P had a marked effect upon the drive-in behavior of Al; leading to acceleratedAl diffusion ahead of the P profile (due to an enhanced Al diffusion which was caused bya supersaturation of self-interstitials) and to the up-hill migration of Al in the high-concentration regime. This was explained in terms of field-assisted diffusion. A strongretardation of Al diffusion, combined with concentrations which were well above thesolid-solubility limit, was observed during Al pre-deposition into P-diffused wafers.D.Nagel, C.Frohne, R.Sittig: Applied Physics A, 1995, 60[1], 61-5

[446-134/135-155]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe 2-dimensional diffusion of Al that was implanted through a mask was measured afterfurnace annealing or rapid thermal annealing. The chemical staining of cross-sectionedsamples was used to determine the vertical and lateral junction depths. The spreadingresistance method was also used to measure the vertical profiles, while a new procedurewas developed for the determination of the lateral diffusion profiles. Differences in theprofiles which were measured by using the staining and spreading resistance methodswere attributed to a spilling phenomenon that distorted the carrier concentration profile onbevelled samples.G.Galvagno, F.La Via, M.G.Saggio, A.La Mantia, E.Rimini: Journal of theElectrochemical Society, 1995, 142[5], 1585-90

[446-134/135-155]

Bulk Diffusion - Qualitative Observations - Doping EffectsThe electrical activity of implanted Al atoms was investigated. Precipitates of Al at thesurface did not contribute to the conductivity. Upon annealing at 1250C, the Alprecipitates disappeared. The low electrical activity of the Al atoms was related to theexistence of Al precipitates, which corresponded to O precipitates, and to the out-diffusion of Al. By using a new annealing method in which a capping film covered theimplanted surface, the out-diffusion of Al atoms from the implanted region into thecapping film and simultaneous back-diffusion into the Si substrate was observed. Such aback-diffusion mechanism increased the Al atom concentration and the electrical activity.M.Watanabe, O.Ishiwata, M.Nagano, H.Kirihata: Journal of the Electrochemical Society,1991, 138[11], 3427-31

[446-84/85-059]

Bulk Diffusion - Qualitative Observations - Doping EffectsDefect states which were introduced into float-zone material by heavy dopant diffusionwere investigated by using the electron beam-induced current method. By measuring theminority carrier diffusion length, using the first-order moment method, and by directlyimaging the defects, their electrical activity could be determined. The diffused sampleswere then dry-oxidized, so that changes in the electrical and morphological properties ofthe introduced defects could be monitored. Two sets of samples were investigated, one ofwhich had been diffused with B and another which had been diffused with B and Al.Significant improvements in the diffusion length were observed in the samples into which

223

Al Bulk Diffusion Al

Al had been co-diffused; thus providing evidence for an effect of Al upon the electricalactivity of bulk defective states.A.Castaldini, A.Cavallini, B.Fraboni, E.Giannotte: Journal of Applied Physics, 1992,72[12], 5622-7

[446-106/107-130]

Table 3Diffusion of Al on Si Surfaces [4º off (111)]

as a Function of Acceleration Voltage and Substrate Temperature

Voltage (kV) T (C) Direction Diffusion Distance (mm)0 50 [110] 0.0100 50 [110] 0.0143 50 [110] 0.0123 50 [110] 0.0260 200 [110] 0.0170 200 [110] 0.0403 200 [110] 0.0223 200 [110] 0.0350 400 [110] 0.0110 400 [110] 0.0213 400 [110] 0.0153 400 [110] 0.017

Bulk Diffusion - Qualitative Observations - Doping EffectsThe diffusion of 300keV Al ions, which had been implanted to doses ranging from 10

13 to

5 x 1015

/cm2, was investigated in capped and uncapped material. The Al-based

precipitates which formed when the Al concentration exceeded its solid solubility in Siwere electrically inactive. Out-diffusion from uncapped samples reduced the Al dosewhich diffused into the Si substrate. In capped samples, Al segregation in the SiO2 layeroccurred.A.Scandurra, G.Galvagno, V.Raineri, F.Frisina, A.Torrisi: Journal of the ElectrochemicalSociety, 1993, 140[7], 2057-62

[446-106/107-130]

Bulk Diffusion - Qualitative Observations - Effect of Electron IrradiationThe redistribution of Al and Si under the influence of irradiation at temperatures of up to60C, after the formation of a p-n structure by the thermal diffusion of Al, was analyzedby assuming the operation of an interstitial mechanism. It was deduced that the maincontribution to an increase in the reverse breakdown voltage, due to electron irradiation,arose from equalization of the inhomogeneity of the diffusant distribution at the p-njunction.

224

Al Surface Diffusion/Bulk Diffusion Ar

S.Makhkamov, J.V.Pakharukov, M.S.Yunusov: Fizika i Tekhnika Poluprovodnikov, 1989,23[9], 1686-9 (Soviet Physics - Semiconductors, 1989, 23[9], 1042-3)

[446-76/77-031]

33 Surface Diffusion - Quantitative DataScanning Auger electron microprobe line scan analysis was used to determine the surfacemobility of Al atoms on (111) Si substrates which were cut 4º off-axis in the [110]direction. Such substrates were believed to have clusters of single atomic layer stepswhich were oriented normal to the [110] direction. The Al films, which were about 6nmthick, were deposited from an ionized cluster beam source. The slot was oriented with itslong axis in the [112] direction (normal to the [110] direction). Auger line scans showedthat the surface mobility of the Al was anisotropic, and that the anisotropy changed as theacceleration voltage was increased from 0 to 3kV and as the substrate temperature wasincreased from 50 to 400C (table 3). The anisotropy was attributed to a higher probabilityfor Al atoms to be absorbed or reflected at steps in the ascending direction. In thedescending direction, the passage of Al atoms over steps was facilitated by the higherthermal energies which were provided by an increased substrate temperature or ionacceleration. An increased evaporation of Al at high temperatures resulted in a decreasein diffusion distance above 200C.L.L.Levenson, H.Usui, I.Yamada, T.Takagi, A.B.Swartzlander: Journal of VacuumScience and Technology A, 1989, 7[3], 1206-9

[446-74-045]

ArBulk Diffusion - Quantitative DataThermal desorption and Rutherford back-scattering spectrometric studies were made ofAr-implanted samples. Thermal desorption spectra were obtained for implantation doseswhich ranged from 1013 to 1017/cm2, at energies of between 20 and 60keV. The thermaldesorption measurements showed that Ar ions were released from specimens in 2 stages,between 800 and 1100K, thus revealing that the implanted Ar ions were present in 2different states. The temperatures of these stages were measured as functions of theimplantation dose and energy, and it was concluded that a solute Ar ion was released attemperatures of between 800 and 1000K, depending upon the implantation energy, andthat agglomerates formed at 1100K. The activation energy for Ar diffusion was estimatedto be equal to 1.47eV.R.Hanada, S.Saito, S.Nagata, S.Yamaguchi, T.Shinozuka, M.Fujioka: Materials ScienceForum, 1995, 196-201, 1375-80

[446-127/128-153]

225

As

Figure 2: Diffusivity of As in Si

Bulk Diffusion - Quantitative DataThe co-diffusion of As and B in monocrystalline samples was studied by means ofsecondary ion mass spectrometry and rapid thermal annealing. The migration of As aloneduring annealing at 1050 to 1100C could be described by:

D(cm2/s) = 35 exp[-4.00(eV)/kT]while the co-diffusion could be described by:

As: D(cm2/s) = 22.83 exp[-4.10(eV)/kT]B: D(cm2/s) = 0.9 exp[-3.43(eV)/kT]

1.0E-16

1.0E-15

1.0E-14

1.0E-13

1.0E-12

1.0E-11

1.0E-10

7 8 9

table 4table 5table 6

104/T(K)

D (cm2/s)

226

As Bulk Diffusion As

C.Gontrand, P.Ancey, H.Haddab, G.Chaussemy: Semiconductor Science andTechnology, 1992, 7[2], 181-7

[446-88/89-048]

Table 4Diffusivity of As in Si

T (C) D (cm2/s)1002 1.1 x 10-13

902 3.9 x 10-15

852 9.0 x 10-16

Bulk Diffusion - Quantitative DataAn extrinsic As background was used to provide a constant electron concentration for thediffusion of an Sb profile. Annealing was carried out at 850, 950 or 1050C, and thediffusion of Sb and As was measured under inert and oxidizing conditions by means ofsecondary ion mass spectrometry. It was found that the diffusion of As was enhanced,while the diffusion of Sb within the As layer was retarded. The data could be describedby:

DSb = 0.214 exp[-3.65/kT] + 15 (n/ni)exp[-4.08/kT]and

DAs = 8.0 exp[-4.05/kT] + 12.8 (n/ni)exp[-4.05/kT]where n/ni accounted for the concentration dependent diffusion which was proportional tothe donor concentration (n) over the intrinsic electron concentration (ni). It was concludedthat the results provided support for a dual, interstitial/vacancy, mode of dopant diffusion.They also provided evidence against the use of a vacancy-only diffusion model.E.A.Perozziello, P.B.Griffin, J.D.Plummer: Applied Physics Letters, 1992, 61[3], 303-5

[446-93/94-045]

Bulk Diffusion - Quantitative DataThe co-diffusion of As and B which had been implanted to doses of about 1016/cm2, usingenergies which gave the same projected range, was investigated at 900 and 1000C bymeasuring dopant and carrier profiles. A comparison of co-diffusion data, with the resultswhich were obtained by the separate diffusion of each element, revealed anomalouseffects which could be explained by positing the formation of neutral donor-acceptorpairs. These complexes were mobile, with diffusivities that could be described by:

D (cm2/s) = 17 exp[-4(eV)/kT]Such values were very close to the diffusivity of As in intrinsic Si. On the basis of theresults, a diffusion model was proposed which took pairing into account. A simulationwhich included this model permitted the prediction of the anomalous phenomena thatoccurred during the high-concentration co-diffusion of donors and acceptors. Theagreement with observed profiles was generally good.S.Solmi, S.Valmorri, R.Canteri: Journal of Applied Physics, 1995, 77[6], 2400-6

[446-121/122-084]

227


34 Bulk Diffusion - Quantitative DataMigration into an underlying Si substrate, from CoSi2 layers which had been implantedwith As ions, was studied by using a high-resolution carrier delineation technique. Thejunction shape was deeper near to the CoSi2 grain boundaries. By using 2-step annealing,or a thin silicide diffusion source, a laterally uniform junction was obtained with As-implanted CoSi2. The diffusion coefficients of As could be measured by using thistechnique. The activation energy for As diffusion (table 4) was 4.05eV.F.La Via, C.Spinella, E.Rimini: Semiconductor Science and Technology, 1995, 10[10],1362-7

[446-125/126-144]

35 Bulk Diffusion - Quantitative DataStudies were made of diffusion in material with high donor concentrations that wereproduced by P doping. It was found that, for donor concentrations which were belowabout 2 x 10

20/cm

3, the diffusivity (table 5) depended linearly upon the dopant

concentration. However, at higher dopant concentrations, the diffusivity increasedmarkedly with increasing donor concentration. This behavior was successfully modelledin terms of the vacancy-percolation model, and it was concluded that collectivephenomena played a significant role at high donor concentrations.A.N.Larsen, K.K.Larsen, P.E.Andersen, B.G.Svensson: Journal of Applied Physics, 1993,73[2], 691-8

[446-106/107-132]


T (C) D (cm2/s)1100 5.0 x 10-12

1075 3.7 x 10-12

1050 3.1 x 10-12

1025 2.1 x 10-12

1000 1.9 x 10-12

36 Bulk Diffusion - Quantitative DataThe concentration profiles of As which had been diffused, from polycrystalline material,into underlying monocrystalline material were analyzed by means of secondary ion massspectrometry. The co-diffusion of As and B was studied in an emitter and extrinsic baseconfiguration. The process was investigated by carrying out simulations. These indicatedthat diffusion of the dopant at the lowest fluence was slowed much more by the in-depthinhomogeneous grain growth that was induced by amorphization and annealing, than by abuilt-in electric field. It was assumed that the dopant, at the highest doses, saturated thegrain boundary traps. This was especially true of As. The diffusivity of As (table 6) couldbe described by:

228


D (cm2/s) = 7.85 x 10-8 exp[-0.9(eV)/kT]A.Merabet, C.Gontrand: Physica Status Solidi A, 1994, 145[1], 77-88

[446-117/118-193]


T (C) D (cm2/s)1150 5.1 x 10-11

1100 4.4 x 10-11

1050 2.6 x 10-11

1000 2.3 x 10-11

Bulk Diffusion - Qualitative Observations - Charged SolutesExperiments showed that ion pairing had a marked effect upon the diffusion of oppositelycharged impurities. An analysis of literature data was used to deduce the ion pairingcoefficients for n-type impurities with B and In. A coefficient with the value of 0.17/Nwas found to describe the pairing case of As-B, where N was the intrinsic electronconcentration. It was suggested that the paired ions occupied adjacent substitutional sites;with a small perturbation in the Coulomb binding which arose from elastic effects.N.E.B.Cowern. Applied Physics Letters, 1989, 54[8], 703-5

[446-64/65-176]

Bulk Diffusion - Qualitative Observations - Concentration DependenceAt fairly high concentrations, the effective As diffusion coefficient was known to increaselinearly as a function of concentration. Here, the markedly enhanced diffusion of As wasstudied by using asymptotic methods. The results of a previous paper were extended bycharacterizing an additional (low-concentration) region, by considering the effect of theinitial data, and by considering the case of constant surface concentration.J.R.King: IMI Journal of Applied Mathematics, 1987, 38[2], 87-95

[446-55/56-033]

Bulk Diffusion - Qualitative Observations - Concentration DependenceIt was pointed out that, over quite a wide range of concentrations, the diffusioncoefficient increased linearly with concentration. However, at very high concentrations,precipitation or clustering led to a plateau or to a decrease in the diffusivity.J.R.King: IMA Journal of Applied Mathematics, 1988, 40[3], 163-81

[446-78/79-048]

Bulk Diffusion - Qualitative Observations - Concentration DependenceThe co-diffusion of implanted As and P was investigated, after annealing at 900 or1000C, for various dopant concentrations. The results did not reveal any direct interactionbetween the dopants. All of the observed anomalous effects of co-diffusion, as compared

229


with diffusion of the elements by themselves, seemed to be explained by interactionsbetween the dopants and the defects which were produced by ion implantation. It wasobserved that the presence of a high concentration of As atoms made annealing of theimplantation damage more rapid, and strongly reduced P transient-enhanced diffusion.S.Solmi, P.Maccagnani, R.Canteri: Journal of Applied Physics, 1993, 74[8], 5005-12

[446-106/107-131]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe diffusion of this element from polycrystalline material and into monocrystallinematerial during rapid optical annealing was investigated. The samples were characterizedby using secondary ion mass spectrometry, transmission electron microscopy, and sheetresistance measurements. It was demonstrated that very shallow diffusion profiles inmonocrystalline material could be formed by rapid optical annealing. By annealing(1050C, 5s), a junction depth of about 35nm was obtained.H.J.Böhm, H.Wendt, H.Oppolzer, K.Masseli, R.Kassing: Journal of Applied Physics,1987, 62[7], 2784-8

[446-55/56-034]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesA secondary ion mass spectrometric analysis of diffusion across a TaSi2/Si interface andinto monocrystalline or polycrystalline Si was carried out, together with an electricalcharacterization of the resultant structures. At temperatures ranging from 900 to 1000C,the dopant readily diffused into Si, without drastic segregation effects, when appropriateinterface cleaning was used. In particular, very shallow diffusion regions were obtained inmonocrystalline material beneath the implanted TaSi2; even at the relatively longannealing times that were sometimes needed for processing.H.Gierisch, F.Neppl, E.Frenzel, P.Eichinger, K.Hieber: Journal of Vacuum Science andTechnology B, 1987, 5[2], 508-14

[446-55/56-034]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe migration of As from implantation-doped polycrystalline Si films and intomonocrystalline Si was investigated as a function of various process parameters. Theeffects of interface treatment before polycrystalline Si deposition, and of the Si grain size,were analyzed. Well-behaved diffusion sources were obtained only if epitaxialrealignment of the polycrystalline Si film to the substrate was avoided and if no diffusionbarrier was present at the polycrystalline/monocrystalline interface.V.Probst, H.J.Böhm, H.Schaber, H.Oppolzer, I.Weitzel: Journal of the ElectrochemicalSociety, 1988, 135[3], 671-6

[446-60-012]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe diffusion of As in Si-on-insulator structures, formed by O implantation, was studied.Although cross-sectional transmission electron microscopic analyses confirmed the highquality of the material, secondary ion mass spectroscopy and spreading resistancemeasurements revealed a pile-up phenomenon as well as an enhanced diffusivity of As in

230


the Si over-layer. An explanation for these 2 observations was proposed, and theMcNabb-Foster equations for diffusion with trapping were solved in order to simulatethese effects.N.Guillemot, P.Normand, D.Tsoukalas, P.Chenevier: Revue de Physique Appliquée,1988, 23[8], 1369-73

[446-61-084]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesNon-uniformities in polysilicon/silicon interfaces, and in the polysilicon structure, wereexpected to produce a non-uniform diffusion front when As was diffused in fromepitaxially aligned polysilicon. By using transmission electron microscopy, surprisinglyuniform As diffusion fronts were found in the underlying Si substrate when As wasimplanted before annealing at high temperatures (in order to induce epitaxial alignment).Thinned specimens were examined, using cross-sectional transmission electronmicroscopy, before and after immersion in a junction-delineating etch. The latterconsisted of 0.5%HF in HNO3. At As penetration depths which were greater than 50nm,uniform junctions were found; even when annealing caused a non-uniform break-up ofthe interfacial layer and an epitaxial alignment of the polysilicon with the (100) substrate.It was concluded that the use of high As concentrations and high-temperature rapidthermal annealing, in order to induce an epitaxial alignment of the polysilicon, remained aviable technique.J.L.Hoyt, E.F.Crabbé, R.F.W.Pease, J.F.Gibbons, A.F.Marshall: Journal of theElectrochemical Society, 1988, 135[7], 1773-9

[446-62/63-226]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe use of TiSi2 as a dopant diffusion source for As was studied. The TiSi2 layers weredoped via ion implantation. Diffusion was carried out by using furnace and rapid thermalprocessing. Secondary ion mass spectrometry, scanning electron microscopy, and X-raydiffraction furnished clear evidence for compound formation between Ti and the twodopant species. This led to low dopant concentrations at the silicide/Si interface and to avery poor efficiency of the diffusion source.V.Probst, H.Schaber, P.Lippens, L.Van den Hove, R.De Keersmaecker: Applied PhysicsLetters, 1988, 52[21], 1803-5

[446-62/63-226]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesShallow n+-p junctions were formed by using an in situ doped thin polycrystalline Silayer as a diffusion source. The As-doped films were applied by means of rapid thermalprocessing chemical vapor deposition. Dopant pile-up phenomena were observed at boththe polycrystalline Si/Si interface and at the surface. The dopant concentrations werehigher at lower deposition temperatures. The observed pile-up phenomena at thepolycrystalline Si/Si interface were temperature dependent and were due mainly to Assegregation at the grain boundary. The dopant distribution was due mainly to grainboundary diffusion and grain growth mechanisms. Extremely shallow n -p junctions were

231


obtained and laterally uniform delineated junctions were observed. The dopantconcentration in the Si substrate fell by 2 orders of magnitude within less than 50nm.T.Y.Hsieh, H.G.Chun, D.L.Kwong, D.B.Spratt. Applied Physics Letters, 1990, 56[18],1778-80

[446-76/77-031]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe migration of As in Si-on-insulator structures, which were formed by deep Oimplantation, was studied by using secondary ion mass spectrometry and spreadingresistance methods. An enhanced diffusivity, and a pile-up phenomenon, were observedin thin Si layers.P.Normand, D.Tsoukalas, N.Guillemot, P.Chenevier: Journal of Applied Physics, 1989,66[8], 3585-9

[446-74-046]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe creation of a buried collector layer, by As diffusion from oxidized As-implantedamorphous Si, and the effect of solid-phase epitaxial layer formation upon the quality ofthe As-diffused buried collector, were clarified. In order to break up the native oxide andto grow the solid-phase epitaxial layer at the interface, As pile-up during the oxidation ofAs-implanted chemical vapor deposited Si was necessary. The crystallinity of a buriedlayer which was formed by As diffusion from oxidized As-implanted amorphous Si, wasgreater than that which was formed from oxidized As-implanted polysilicon.Y.Tsunoda: Japanese Journal of Applied Physics, 1990, 29[10], L1926-8

[446-78/79-047]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesDopant redistribution, during rapid thermal annealing, from implantation-dopedpolycrystalline films into monocrystals was studied. Diffusivity and co-diffusivity datawere deduced from secondary ion mass spectrometry results.C.Gontrand, C.Dubois, A.Laugier: Journal of Physics - Condensed Matter, 1991, 3[18],3091-8

[446-84/85-059]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesComparative studies of As and B diffusion in polysilicon-on-monocrystal systems wereperformed by means of cross-sectional transmission electron microscopy and secondaryion mass spectrometry. The As and BF2 which were implanted into 300nm polysilicon,deposited via low-pressure chemical vapor deposition, were diffused into the underlyingSi substrate by using rapid thermal annealing or furnace annealing. The As diffusionprofiles were continuous across the polysilicon/monocrystal interface, except for a peak atthe interface. They exhibited a gradual increase towards the interface, within thepolysilicon layer, due to an inhomogeneous distribution of grain sizes in As-implantedpolysilicon. At low annealing temperatures, the B profiles in monocrystalline Si wereshallower than the As profiles. This then reduced the effect of grain boundaries withregard to high diffusivities. During high thermal budget annealing, B diffusion into the

232


substrate was greater than As diffusion because of higher B diffusivities inmonocrystalline Si. At high annealing temperatures, the native oxide at the interfacebroke up and caused the polysilicon layer to align epitaxially with the underlyingsubstrate. The elapsed time up to the break-up of the interfacial oxide depended upon thedopant species and the annealing temperature. The oxide break-up took longer in As-doped samples than in B-doped samples.K.Park, S.Batra, S.Banerjee, G.Lux, R.Manukonda: Journal of the ElectrochemicalSociety, 1991, 138[2], 545-9

[446-81/82-044]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesDepth profiles were determined by using spreading resistance, Auger electronmicroscopy, and secondary ion mass spectrometry methods. These profiles showed thatenhanced dopant diffusion and surface depletion resulted from rapid thermal treatments.It was found that Si interstitial injection from a N-supersaturated oxynitride interfacefacilitated the diffusion of As in the Si substrate. An appreciable amount of N was foundbelow the Si surface, and this suggested that N interstitials might play an important role inthe observed enhancement of the impurity diffusion.J.Bustillo, C.Chang, S.Haddad, A.Wang: Applied Physics Letters, 1991, 58[17], 1872-4

[446-86/87-049]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesMigration in amorphous and polycrystalline Si-on-monocrystal systems during rapidthermal annealing and furnace annealing was studied. It was found that changes inmicrostructure during annealing played a major role in determining diffusion profiles inthe substrate, as well as in the polycrystalline Si layer. In the case of As doping, drive-indiffusion resulted in a much larger grain microstructure in as-deposited amorphous Sithan in as-deposited polycrystalline Si. This led to the formation of shallower junctions inthe first case. At high annealing temperatures, the native interfacial oxide disintegrated,caused epitaxial realignment of the polycrystalline Si film, and led to enhanced diffusionin the substrate.K.Park, S.Batra, S.Banerjee, G.Lux, T.C.Smith: Journal of Applied Physics, 1991, 70[3],1397-404

[446-91/92-028]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe As ions were implanted into thin layers of WSi2 or CoSi2, on polycrystalline ormonocrystalline Si, and were out-diffused into the substrate during furnace heating orrapid thermal annealing. Shallow (< 100nm) junctions were obtained by using rapidthermal annealing. The interface concentrations were close to the solid solubilities of therespective dopant; that is, more than 3 x 1020/cm3. It was found that dopant redistributionwas determined by the lattice and grain-boundary diffusivities, solubility limits, layerinhomogeneities, dopant segregation at interfaces and grain boundaries. In addition, therewas probably a phase transformation of the dopants which were segregated at theSiO2/silicide interface.

233


V.Probst, H.Schaber, A.Mitwalsky, H.Kabza, L.Van den Hove, K.Maex. Journal ofApplied Physics, 1991, 70[2], 708-19

[446-93/94-045]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesAnomalous diffusion of lightly implanted As into a Si substrate was observed duringfurnace annealing in a N ambient. The anomalous behavior was reflected by theappearance of clear features in the near-surface and tail regions. In the near-surfaceregion, which extended to between 30 and 50nm from the interface, a large number of Asatoms moved towards the oxide/Si interface and occasionally accumulated in a narrowregion which was between 10 and 20nm from the interface. Retarded diffusion wasobserved in the tail region. The diffusivity in the tail region increased with increasingannealing time.N.Aoki, T.Kanemura, I.Mizushima: Applied Physics Letters, 1994, 64[23], 3133-5

[446-115/116-149]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe precipitation and dissolution of monoclinic SiAs in samples which had beenimplanted with 10

17 or 1.5 x 10

17As

+/cm

2, was monitored at 800, 900, and 1050C by

means of transmission electron microscopy and secondary neutral mass spectrometry. Itwas found that the saturation concentration of mobile As, which diffused afterequilibration with the monoclinic SiAs precipitates, could be described by: C(/cm2) = 1.3x 10

23exp[-0.42(eV)/kT]. This saturation concentration comprised various states of As

which were in equilibrium with the monoclinic phase at the annealing temperature: suchas electrically active and inactive mobile dopants. The present results indicated that theinactive As was in the form of clusters. The link between clustering, and the presence ofsmall particles which were detected by transmission electron microscopic observations,was considered.D.Nobili, S.Solmi, A.Parisini, M.Derdour, A.Armigliato, L.Moro: Physical Review B,1994, 49[4], 2477-83

[446-115/116-150]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe co-diffusion of As and B in polycrystalline and monocrystalline bi-layers, during theformation of shallow N+ emitters, was studied. It was found that rapid thermal annealingled to the redistribution, as measured by secondary ion mass spectrometry, of As and Bwhich were successively implanted into a 380nm low-pressure chemical vapor-depositedpolycrystalline layer.C.Gontrand, P.Sellitto, S.Tabikh, S.Latreche, A.Kaminski: Journal de Physique III, 1997,7[1], 47-58

[446-150/151-151]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe effect of lattice defects, produced by Si ion implantation, upon dopant diffusivity wasinvestigated after annealing specimens at temperatures of between 700 and 900C. Thenature and depth of residual implantation defects in undoped samples was determined byanalysing the rocking curves which were obtained by using triple-crystal X-ray diffraction

234


and transmission electron microscopy. As well as interstitial dislocation loops andclusters lying below the original amorphous/crystal interface, epitaxial re-growth of theamorphized Si left a vacancy-rich surface layer and a deeper region which was enrichedin interstitials. These regions corresponded to those for which Monte Carlo simulations ofdefect production predicted an excess of point defects. According to whether the dopantwas associated with vacancy or interstitial clusters, different anomalous diffusionbehaviors were observed. In the deep region where an excess of interstitials was present,only a small enhancement was exhibited by As. On the other hand, a slight enhancementin the case of As was observed in the surface layer. This was consistent with the differingcontributions which vacancies and interstitials made to the diffusion mechanisms of thedopants.M.Servidori, R.Angelucci, F.Cembali, P.Negrini, S.Solmi, P.Zaumseil, U.Winter: Journalof Applied Physics, 1987, 61[5], 1834-40

[446-60-013]

Bulk Diffusion - Qualitative Observations - Defect InteractionsA theory was developed for impurity diffusion under equilibrium and non-equilibriumconcentrations of point defects. The results of first-principles calculations of several keyquantities were combined with the theory and were compared with experimental data. Itwas found that vacancies and self-interstitials governed the equilibrium diffusion of As.Interstitials tended to predominate. It was also found that the direct-exchange mechanismplayed only a minor role.C.S.Nichols, C.G.Van de Walle, S.T.Pantelides: Physical Review Letters, 1989, 62[9],1049-52

[446-64/65-177]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe behavior of the dopant during infra-red heating was investigated by using spin-ondiffusion sources, repeated etching, and capacitance-voltage measurements. The diffusioncoefficient was calculated by fitting the concentration profiles to the complementary errorfunction. It was found that the diffusivity was enhanced during infra-red heating. This didnot appear to be due to ion-implantation damage or rapid heating. The enhanceddiffusivity was attributed to the generation of an excess of self-interstitials.Y.Ishikawa, K.Yamauchi, I.Nakamichi: Japanese Journal of Applied Physics, 1989,28[8], L1319-21

[446-72/73-041]

Bulk Diffusion - Qualitative Observations - Defect InteractionsA comprehensive investigation of dopant diffusion in the presence of equilibrium andnon-equilibrium concentrations of intrinsic point defects was presented. It was found that,under equilibrium conditions, vacancies and interstitials mediated the diffusion of thedopant. Relevant expressions were derived for the activation energies, of variousdiffusion and injection mechanisms, under non-equilibrium conditions (such as thoseproduced by the injection of excess point defects). For oxidation conditions, thecalculated values were in excellent agreement with available experimental data. Boththeory and experiment

235


suggested that the concerted exchange mechanism, which involved no point defects,played only a minor role in dopant diffusion.C.S.Nichols, C.G.Van de Walle, S.T.Pantelides: Physical Review B, 1989, 40[8], 5484-96

[446-72/73-041]

Bulk Diffusion - Qualitative Observations - Defect InteractionsIon implantation produces point defects, during annealing, which can significantlyenhance dopant diffusion. This effect was studied for application to implant diffusion atlow temperatures. Enhanced diffusion was detected below critical dopant concentrations.The latter concentrations depended only upon the temperature. This concentrationcorresponded to the As solid solubility.R.B.Fair: Journal of the Electrochemical Society, 1990, 137[2], 667-71

[446-74-046]

Bulk Diffusion - Qualitative Observations - Defect InteractionsIt was noted that the binding energy between a vacancy and an As atom in this materialwas large. Therefore, the pair diffusion model and a decrease in quasi-vacancy formationenergy were applicable to As diffusion. The binding energy also decreased withdecreasing quasi-vacancy formation energy. The Fermi level was deduced fromBoltzmann statistics.M.Yoshida, E.Arai: Japanese Journal of Applied Physics 1, 1995, 34[11], 5891-903

[446-134/135-156]

Bulk Diffusion - Qualitative Observations - Defect InteractionsAn experimental characteristic of As diffusion is that the coefficient is independent of thesurface concentration, but increases roughly linearly with the electrically active Asconcentration. From the point of view of the pair diffusion model, this implied that excessvacancies were not generated and that vacancies were in a thermal equilibrium state.Here, As diffusion was studied on the basis of the pair diffusion model, while takingaccount of a decrease in the quasi-vacancy formation energy. With decreasing quasi-vacancy formation energy, the binding energy between an As atom and a vacancy alsodecreased. The Fermi level which was deduced from Boltzmann statistics was used. Byusing published data for the self-diffusion coefficient of vacancies, and the As diffusioncoefficient, the experimentally observed As concentration profile could be explained.M.Yoshida, E.Arai: Japanese Journal of Applied Physics 1, 1996, 35[1A], 44-55

[446-136/137-126]

Bulk Diffusion - Qualitative Observations - Diffusion EnhancementThe effect of HCl upon the oxidation-enhanced diffusion of As at temperatures of 1000,1100, and 1150C was investigated by using secondary ion mass spectrometry. The dataobtained were used to determine the parameters of a model which had previously beendeveloped in order to describe the retardation of oxidation-enhanced diffusion. Goodagreement was found between theory and experiment.R.Subrahmanyan, H.Z.Massoud, R.B.Fair: Journal of Applied Physics, 1987, 61[10],4804-7

[446-61-084]

236


Bulk Diffusion - Qualitative Observations - Doping EffectsThe effect of the time and temperature of As diffusion from doped SiO2 films wasstudied. It was shown that, at temperatures ranging from 1370 to 1480K and timesranging from 0.5 to 8h, the SiO2 films were an essentially unlimited source of As.J.D.Nisnevich: Izvestiya Akademii Nauk SSSR - Neorganicheskie Materialy, 1990, 26[4],687-9 (Inorganic Materials, 1990, 26[4], 575-7)

[446-84/85-059]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe effects of implantation damage upon As diffusion during rapid thermal annealing andconventional heat treatment at 750 to 900C were investigated. A comparison of thedopant profiles and junction depths in damaged and undamaged regions demonstrated thatSi+ implantation under the typical conditions used for pre-amorphization led to only anegligible increase in observed As transmission reactions.A.Jacques, A.George, X.Baillin, J.J.Bacmann: Philosophical Magazine A, 1987, 55[2],165-81

[446-51/52-130]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationDiffusion experiments were performed in the presence of lattice defects which had beenproduced by Si ion implantation. The effects of transient enhanced diffusion were studiedby means of bevelling and staining measurements of implanted samples, and secondaryion mass spectroscopic determinations of dopant profiles. The annealing temperatures fordoped implanted specimens ranged from 700 to 1100C and this treatment was performedusing an electron beam. The low temperatures which were used permitted the kinetics ofanomalous diffusion to be monitored. It was ascertained that the enhanced diffusioncoefficient was almost constant during a period which decreased with increasingtemperature. It then tended gradually to the equilibrium value. This trend agreed with thatof lattice damage changes which were revealed by double crystal X-ray analyses of therocking curves of implanted samples. The secondary ion mass spectroscopy profilesindicated that only a fraction of the dopant which was located at residual implantationdamage was responsible for the anomalous diffusion.R.Angelucci, F.Cembali, P.Negrini, M.Servidori, S.Solmi: Journal of the ElectrochemicalSociety, 1987, 134[12], 3130-4

[446-60-013]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationIt was found that, despite implantation away from the crystal orientation, the profile of80keV As which had been implanted into (100) to a dose of 8 x 1015/cm2 had a longchannelling tail. Enhanced diffusion of the implanted As was detected, in the early stagesof annealing, by means of Hall effect and sheet resistivity measurements. Theenhancement factor decayed with annealing time, and the time constant of the decayshortened with increasing temperature. The activation energy for the decay of theenhancement factor was equal to 1.2eV at temperatures of between 900 and 1000C, andwas equal to 3.4eV at temperatures of between 1000 and 1100C. The maximumenhancement factor was 2.7, and occurred at the beginning of annealing.

237


Y.Sasaki, K.Itoh, S.Tanuma: Japanese Journal of Applied Physics, 1989, 28[8], 1421-5[446-72/73-042]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationRutherford back-scattering and channelling spectrometry were used to investigate changesin the crystal structure of surface layers, and dopant accumulation during the self-annealing As+ ion irradiation of monocrystals. It was found that irradiation with ionbeams of high current density (about 0.0002A/cm2) resulted in an anomalousredistribution of the impurities towards the bulk of the crystal. The As redistributionmechanism which was most likely to explain the observations was impurity diffusion viathe inverse Kirkendall effect.F.F.Komarov, A.P.Novikov, E.V.Kotov, E.A.Podlipko: Physica Status Solidi A, 1989,112[1], 323-6

[446-74-046]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe co-diffusion of As and P was studied by using a buried B layer. An analysis of thedopant profiles was performed by using secondary ion mass spectroscopy, spreadingresistance profiles, and junction staining. It was found that the profiles of As and P, whendiffused separately, were enhanced with respect to the co-diffused profiles. The buried Bprofile in the case of P alone was enhanced with respect to the B profile in the case of Asand P co-diffusion. The annealing of residual implantation damage was suggested to beresponsible for these effects. In addition, the suppression of the self-interstitialsupersaturation under co-diffusive conditions was a transient effect and was suggested tobe caused by an increase in the recombination of implantation-generated vacancies andself-interstitials due to the presence of a high concentration of As atoms.R.Deaton, U.Gösele, P.Smith: Journal of Applied Physics, 1990, 67[4], 1793-800

[446-74-046]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe anomalous distribution of As which was implanted under self-annealing conditions(with simultaneous damage recovery activated by beam heating) was investigated.Rutherford back-scattering/channelling, transmission electron microscopy, and carrierprofiling techniques were used to analyze the dopant profiles and the microstructure ofsamples which had been irradiated with 150keV As+ ions to a current density of about0.000207A/cm2 for between 2 and 6s. Two relevant effects were observed. The first oneconsisted of the formation of 2 dopant peaks which were electrically inactive and wereseparated by a depletion region at the position of the ion projected range. Whereas thedeeper peak disappeared with increasing irradiation time, the one which was located atthe maximum of nuclear energy loss tended to grow. Microstructural analysis suggestedthat both peaks occurred as a result of As segregation at lattice defects. In particular, theone which was located at the maximum of nuclear energy loss was a consequence of thesegregation of As atoms at voids which formed during irradiation at high temperatures.The second relevant effect was the formation of a deeply penetrating tail, in the Asprofile, which could not be explained by a simple thermal diffusion mechanism. Althougha contribution from channelling effects could not be ruled out, the effect seemed to be

238


related mainly to a diffusivity enhancement that was weakly temperature dependent andwas due to interaction of the dopant with radiation-induced defects.G.Lulli, P.G.Merli, R.Rizzoli, M.Berti, A.V.Drigo: Journal of Applied Physics, 1989,66[7], 2940-6

[446-76/77-032]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe effect of the annealing ambient upon dopant diffusion was studied, during low-temperature processing by implanting As into (100) samples (at room temperature)through a 14nm SiO2 layer. The implantation dose and energy were sufficient toamorphize the surface. After low-temperature furnace annealing, the ion-implanted Asexhibited a transient enhanced diffusion regime for both inert and oxidizing ambients. Ithad been expected that point defect generation during the annealing of implantationdamage would predominate during the transient enhanced diffusion process; regardless ofthe nature of the ambient.Y.Kim, H.Z.Massoud, R.B.Fair: Journal of the Electrochemical Society, 1990, 137[8],2599-603

[446-76/77-033]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe migration of impurity atoms in solid-phase epitaxially grown layers was studied bymeans of back-scattering spectrometry. The samples were amorphized by As implantationinto [100]- or [111]-oriented samples. The amorphized regions were then recrystallizedby furnace annealing. Asymmetrically enhanced diffusion was observed in the case of[111]-type samples. A striking result was that, in the case of [111]-type samples, therewas extensive smearing of the profiles while, in the case of [100]-type samples, this washardly detectable. It was recalled that the re-growth features of [100]-type and [111]-typeamorphized Si were different. In essential agreement with previous observations, it wasfound here (using cross-sectional electron microscopy) that the re-grown layer on [111]-type material was characterized by extensive multiple twinning. A 2-step annealingtreatment usually led to a superior crystalline perfection, as confirmed by the cross-sectional transmission electron microscopy of samples which had first been annealed at650C, and then at 1050C. Overall, the extent of diffusion was related to defects.R.Turan, T.G.Finstad: Philosophical Magazine A, 1991, 63[3], 519-25

[446-78/79-048]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe effects of low-dose Si implantation damage upon the diffusion of low concentrationsof As in wafer samples were investigated. The dopant was implanted to low doses, andwas then pre-annealed in order to remove any self-damage. No enhanced As diffusion inthe damaged region was detected.H.Park, M.E.Law: Applied Physics Letters, 1991, 58[7], 732-4

[446-81/82-044]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationA study was made of the effects of implantation and annealing conditions upon As-implanted (100) material. It was found that an amorphous layer was created by

239


implantation, and satisfactory recrystallization was achieved by annealing. Annealing attemperatures above 1000C caused appreciable spreading of the dopant profile, and piling-up of As near to the surface. Annealing at temperatures greater than 1100C resulted in Asout-diffusion from the pile-up zone. This out-diffusion was suppressed by the use of a90nm SiO2 cap. The diffusion of As was attributed to As4V- and As+Vo pairs.G.Chaussemy, C.Gontrand, S.N.Kumar, B.Canut, D.Barbier, A.Laugier: Physica StatusSolidi A, 1991, 124[1], 103-14

[446-88/89-049]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationSamples which had been implanted with 310keV or 38MeV As ions, and annealed at1200C, were studied by using electrochemical C-V measurements and secondary ionmass spectrometry. It was found that there was an optimum annealing temperature, atwhich the diffusion of implanted As was minimized. It was proposed that the tails of thedistributions could be described by the Pearson IV analytical expression. In the case of38MeV As ion implantation, the straggling deviation, skewness, and kurtosis were500nm, -1 and -10, respectively.N.Sieber, G.Otto, H.Syhre: Physica Status Solidi A, 1992, 132[1], 177-82

[446-91/92-027]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationAn experimental and theoretical study was made of the physical mechanisms which wereinvolved in the co-diffusion of As and B in polycrystalline/monocrystalline bilayersduring the formation of shallow N+ emitters. The redistribution, of successivelyimplanted As and B, which was induced by rapid thermal annealing was studied by meansof secondary ion mass spectroscopic measurements. A weak retardation effect was notedfor As in the presence of B, but this had no significant effect upon the dopant profile inthe monocrystalline emitter region.C.Gontrand, A.Merabet, B.Semmache, S.Krieger-Kaddour, C.Bergaud, M.Lemiti,D.Barbier, A.Laugier: Semiconductor Science and Technology, 1993, 8[2], 155-62

[446-099/100-095]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationIons of As and B were implanted using energies which had been chosen so that theirprojected ranges coincided, and the implanted material was annealed in Ar gas (950C, 0.5to 5h). It was found that the activation efficiency of the implanted As atoms decreasedwith an increase in the implantation dose of B ions, and that the diffusivity of As atomsdecreased.K.Yokota, Y.Okamoto, F.Miyashita, T.Hirao, M.Watanabe, K.Sekine, Y.Ando,K.Matsuda: Journal of Applied Physics, 1994, 75[11], 7247-51

[446-117/118-193]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationIt was noted that redistribution phenomena (pile-ups, push-back) of As during thermaloxidation depended upon the oxidation rate, the diffusivities of As in Si and SiO2, and thesegregation rates of As impurities at the interface between the oxide and the Si. The

240


diffusivity of As in the oxide was known to be negligible as compared with its diffusivityin Si and with the oxidation rate of Si. The As diffusivity in Si also depended upon the Asconcentration. Pile-ups at Si/SiO2 interfaces, as a result of a concentration dependence ofAs, had not been reported. In the case of Si samples which had been implanted with 1015

or 3 x 1015/cm2 of 100keV As ions, a pile-up of As was observed, using Rutherford back-scattering spectroscopy, during thermal oxidation at 1050C. In the case of samples whichhad been implanted to fluences of more than 3 x 1016/cm2, push-back phenomena wereobserved. These observations could be explained only in terms of an As diffusivity whichdepended upon the As concentration in the Si.S.S.Choi, M.J.Park, W.K.Chu: Thin Solid Films, 1995, 258[1-2], 336-40

[446-136/137-126]

Bulk Diffusion - Qualitative Observations - Effect of Surface LayerThe migration (due to heating with non-coherent light for some seconds) of As from asurface emission source, across a porous layer, and into monocrystalline Si was studied.Stimulated As diffusion under the porous Si layer was attributed to a heat-induced changein the structure of the porous material. Also, vacancy diffusion of As from the porous Silayer into the single crystal was thought to be enhanced by high mechanical stresses. Theeffect of these stresses upon As diffusion was so great that no difference between thestructures of porous layers on p-type and n-type monocrystals was observed.V.P.Bondarenko, V.E.Borisenko, V.A.Labunov: Fizika i Tekhnika Poluprovodnikov,1986, 20[5], 929-33. (Soviet Physics - Semiconductors, 1986, 20[5], 586-81)

[446-55/56-033]

Bulk Diffusion - Qualitative Observations - Effect of Surface LayerThe diffusion of As into p-type substrates, from As-implanted TiSi2 layers, wasinvestigated for various heat treatments at temperatures of between 900 and 1100C. Thedrive-in diffusion was carried out by using either rapid thermal annealing or traditionalfurnace annealing. Shallow (20 to 80nm deep) junctions were obtained when there was ahigh (1019 to 1020/cm3) dopant concentration at the silicide/Si interface. The amount ofdiffused As, as measured using Rutherford back-scattering spectrometry, increasedlinearly with the square root of the annealing time. A similar relationship was found forthe amount of electrically active As, as deduced from the use of Van der Pauw and anodicoxidation techniques. The 2 quantities were different, and the inactive dopantsprecipitated in the diffused layer, as revealed by transmission electron microscopy. It wassuggested that this behavior was associated with the high tensile stress that was created bythe silicide layer, and with its effect upon solid solubility and upon the clustering of Asatoms. The precipitates were easily dissolved by annealing in the absence of a TiSi2 layer.F.La Via, V.Privitera, S.Lombardo, C.Spinella, V.Raineri, E.Rimini, P.Baeri, G.Ferla:Journal of Applied Physics, 1991, 69[2], 726-31

[446-78/79-047]

Bulk Diffusion - Qualitative Observations - Effect of ThermomigrationThe effect of a temperature gradient upon As diffusion was investigated. Wafersimplanted with As were subjected to temperature gradients of the order of 100C/cm in arapid thermal annealing apparatus. No measurable differences were observed in the As

241


profiles upon changing the direction of the temperature gradient in low (1014/cm2) dose(111) and (100) material. There was also no measurable difference in the As profiles inhigh (6 x 1015/cm2) dose (100) material due to a temperature gradient. However, in high-dose (111) material, the As penetrated deeper on the cold side. The difference in Aspenetration into the cold and hot sides of the sample was of the order of 10 to 15nm for a10 to 90s anneal. The Soret effect contribution to isothermal diffusion was estimated bysolving the diffusion equation. The calculated results agreed well with measured resultsfor low and high-dose (100) material. However, they disagreed in the case of high-dose(111) material. It was suggested that a defect-enhanced Soret effect, which coupled the Asdiffusion flux to the heat flux in the highly disordered crystal, could cause the observedtemperature gradient dependence of the As penetration.A.Feygenson, J.N.Zemel: Thin Solid Films, 1988, 157[1], 49-57

[446-60-012]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsAn investigation was made of the suggestion that P diffusion at concentrations above thesolid solubility generated Si self-interstitials. Buried layers of As were created byimplanting 100keV As into [100]-type substrates, to 1014/cm2. After annealing (900C,0.5h) in N, an 0.008 to 0.01mm thick epilayer was grown over the buried layers. Maskingoxides were then created by liquid-phase chemical vapor deposition of 0.001mm SiO2

layers, or by thermal oxidation at 900C in steam. Spreading resistance profiles wererecorded and secondary ion mass spectrometry was used to characterize the chemical Pdoping densities. Plan-view and cross-sectional transmission electron microscopy wasused to look for SiP precipitates and defects in the P-diffused layers. It was found that P-related precipitates, 0.0001 to 0.0002mm in size, were observed in some samples by usingtransmission electron microscopy. In every case, their concentration was insufficient toaccount for the concentration of non-electrically active P. A measurably enhanceddiffusion of As in the buried layers occurred at temperatures ranging from 1000 to 1150C,with an activation energy of about 1.5eV. This enhanced diffusion occurred only underthe windows where P was diffusing. Stacking fault growth in buried As layers under theP-diffused region implied that Si self-interstitial supersaturation was produced in the P-doped layer. Self-interstitials fed stacking fault growth and enhanced As diffusion. Thenumber of self-interstitials generated could not be accounted for by the sparse number ofprecipitates which formed in the P-diffused region.J.C.C.Tsai, D.G.Schimmel, R.B.Fair, W.Maszara: Journal of the Electrochemical Society,1987, 134[6], 1508-18

[446-55/56-035]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesThe question of the diffusion mechanism which controlled the diffusion of As at veryhigh concentrations was examined in terms of a system of reaction-diffusion equations. Inparticular, the role of the solubility limit was investigated because it represented athreshold value at which some of the quasi-chemical reactions which were involvedabruptly changed their nature. The corresponding solutions made it possible to calculatethe concentration profiles for both electrically active and inactive impurities as a functionof time and pre-deposition. The formation of a plateau of electrically active dopants was

242


considered in detail, together with the distribution of electrically inactive precipitates inthe neighborhood of the surface.E.Antoncik: Applied Physics A, 1994, 58[2], 117-23

[446-115/116-150]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesAn analytical model was presented for the non-linear evolution of the diffusion profile ofimplanted As. Particular emphasis was placed on the study of the front region of thediffusing impurity profile, where the concentration fell steeply and previously givenapproximate investigations failed to apply. In particular, explicit expressions were foundfor the junction depth and the steepness of the profile at the junction.K.O.Jeppson, D.Anderson, G.Amaratunga, C.P.Please: Journal of the ElectrochemicalSociety, 1987, 134[9], 2316-9

[446-55/56-033]

Bulk Diffusion - Theoretical Analysis - Defect InteractionsA general expression for the intrinsic diffusivity of a substitutional impurity was obtainedin terms of the self-interstitial and vacancy concentrations. The vacancy mechanism, theexchange mechanism, and 2 processes which were associated with the motion ofinterstitial impurity atoms (kick-out, Frank-Turnbull) were taken into account. The latter2 contributions were not additive under non-equilibrium point defect conditions, andcould partially cancel in some cases. The general expression could be fitted to Asdiffusion data which were obtained under nitridation and oxidation conditions. Theresults indicated the possible presence of appreciable vacancy, push-out, and Frank-Turnbull terms. However, there was no significant exchange term.N.E.B.Cowern: Journal of Applied Physics, 1988, 64[9], 4484-90

[446-72/73-040]

Bulk Diffusion - Theoretical Analysis - Defect InteractionsResults for As diffusion in polycrystalline material, during annealing (900C, 0.25h), wereexplained by means of a new model which took account of the different diffusion rates ingrains and in grain boundaries. The model was also divided into early and late stages. Theformer indicated that grain growth was the main mechanism for As transfer from thegrains to the boundaries. The late-stage model suggested that grain growth could beignored, and analytical solutions for the As concentrations were derived.A.G.O'Neil, C.Hill, J.King, C.Please: Journal of Applied Physics, 1988, 64[1], 167-74

[446-61-084]

Bulk Diffusion - Theoretical Analysis - Defect InteractionsThe diffusion of As during rapid thermal annealing, in material which had been doped byion implantation, was simulated by means of finite difference calculations. The effect ofenhanced diffusion via radiation defects, and the effect of the temperature field, weremodelled by including thermal diffusion in the diffusion equation. A comparison of thepredicted results with experimental data revealed good agreement for diffusion activationenergies which were as low as 1.8eV. This suggested that migration occurred via

243


interstitials. The incidence of enhanced diffusion decreased for rapid thermal annealingtimes of more than 10s.L.N.Aleksandrov, T.V.Bondareva: Physica Status Solidi A, 1991, 125[2], K71-5

[446-88/89-049]

Bulk Diffusion - Theoretical Analysis - Defect InteractionsA simplified model was presented for the treatment of dopant diffusion in the presence ofnon-equilibrium point defect concentrations. The dopant flux was expressed in terms ofan effective diffusivity, and took account of the various couplings which arose from thepresence of defect gradients. The point defect concentrations were calculated by resolvingthe corresponding continuity equations. It was found that the model permitted the fast andaccurate simulated diffusion of As.D.Mathiot, S.Martin: Journal of Applied Physics, 1991, 70[6], 3071-80

[446-91/92-028]

Bulk Diffusion - Theoretical Analysis - Defect InteractionsThe effective diffusivity of substitutional impurities was considered with regard to pointdefects. It was noted that 4 basic reactions, which involved interstitial and vacancycontributions, had to be considered in order to derive a general formula for the effectivediffusion coefficient. The model was fitted to data on As diffusion under nitridationconditions, and was found to give very good agreement. On the basis of the results, it wasdeduced that the As diffused mainly via a vacancy mechanism; but with a non-negligiblepartial interstitial contribution. It was also observed that, under these conditions, the pointdefect concentrations were a complicated function of the observed P and Sb diffusivities.The general behavior of the dopant diffusivities as a function of point defectconcentration was also studied. It was concluded that the recombination of point defects,mediated by dopant pairs, was important over a wide range of point defect concentrations.E.Vandenbossche, B.Baccus: Journal of Applied Physics, 1992, 72[2], 447-53

[446-93/94-046]

Bulk Diffusion - Theoretical Analysis - Defect InteractionsThe enhanced diffusion of implanted As impurities during thermal annealing can beexplained in terms of a system of reaction-diffusion equations. It was shown that, for highdoses, the local solubility limit could appreciably affect the reactions between the defectswhich were involved, and thus markedly change the effective diffusion of As donors. Asimilar effect could be brought about by the presence of pre-doped donors or acceptors,which could also significantly accelerate or retard the effective diffusion of As implants.An explanation was here proposed for some precipitate clustering processes during rapidor slow cooling. It was found that some apparently contradictory experimental resultswere compatible with the present model.E.Antoncik: Applied Physics A, 1993, 56[4], 291-8

[446-099/100-095]

Bulk Diffusion - Theoretical Analysis - Defect InteractionsFirst-principles total-energy calculations of the atomic diffusion of group-V impuritiesrevealed an effect of pressure upon the activation energy for diffusion. In the case of the

244

As Bulk Diffusion/Grain Boundary Diffusion As

vacancy mechanism, the activation energy for As decreased with pressure. In the case ofthe interstitial mechanism, the formation energy of the interstitial impurity exhibited ageneral tendency to increase with pressure. By combining these results with experimentaldata, it was deduced that a vacancy mechanism made the predominant contribution to Asdiffusion. The microscopic origin of the pressure dependence was explained in terms ofthe local strains around defects. The negative pressure dependence which was common tothe vacancy-mediated diffusion of group-V impurities could be explained by the peculiarproperties of the isolated vacancy. These included a breathing distortion of surrounding Siatoms towards the vacancy site, giving rise to a tensile strain around the vacancy, andlattice distortions which originated from the vacancy and caused weak vacancy-impurityinteractions. The positive pressure dependence of interstitial-mediated diffusion wasclosely related to the atomic structures of the interstitial impurities, which producedcompressive strains in the surrounding Si-Si bonds.O.Sugino, A.Oshiyama: Physical Review B, 1992, 46[19], 12335-41

[446-106/107-131]

Bulk Diffusion - Theoretical Analysis - Defect InteractionsThe binding energy between a vacancy and an As atom is large, and the pair diffusionmodel (and decreased quasi-vacancy formation energy) are applicable to As diffusion. Itwas noted here that, with decreasing quasi-vacancy formation energy, the binding energyalso decreased.M.Yoshida, E.Arai: Japanese Journal of Applied Physics 1, 1995, 34[11], 5891-903

[446-136/137-126]

Bulk Diffusion - Theoretical Analysis - Effect of Ion ImplantationA model was developed in order to describe, in terms of an effective diffusivity, theimplantation-enhanced diffusivity of impurity atoms via a dual vacancy-interstitialmechanism. The model included the mechanisms of vacancy-interstitial pair generationby high-dose impurity implantation, which produced dopant diffusion enhancement, and aso-called transient activation mechanism for dopants in the initial stages of the annealingprocess.H.R.Soleimani: Journal of the Electrochemical Society, 1992, 139[11], 3275-84

[446-93/94-044]

Grain Boundary Diffusion - Qualitative Observations - Concentration ProfilesThe diffusion of donor elements in fine-grained and coarse-grained polycrystallinematerial was studied at temperatures ranging from 900 to 1150C. Tracers (74[73]As) wereused to determine the concentration/depth profiles via sectioning. By means ofautoradiography, the lateral distribution of the radiotracers over the sample surface wasrevealed.F.H.M.Spit, H.Bakker: Physica Status Solidi A, 1986, 97[1], 135-42

[446-48-063]

Grain Boundary Diffusion - Qualitative Observations - Pipe DiffusionThe laser crystallization mechanism of hydrogenated amorphous material, and thediffusion behavior of implanted impurities in crystallized films, were studied. A new

245

As Grain Boundary/Pipe/Surface Diffusion As

region of liquid-phase laser crystallization was found at low scan velocities. In order toinvestigate the diffusion mechanisms, various regions were implanted with 80keV As+

ions to a dose of 5 x 1015/cm2 and then subjected to rapid thermal annealing. Spreadingresistance measurements and secondary ion mass spectroscopy showed that diffusion ingrains and along grain boundaries occurred in all 3 regions. In addition, diffusion alongdefects occurred in the optical damage zone. This was the fastest of the 3 routes.Q.Gu, X.M.Bao: Physica Status Solidi A, 1988, 108[1], 81-7

[446-74-047]

Grain Boundary Diffusion - Qualitative Observations - Positron TrappingIt was shown that the positron beam technique was a sensitive indicator of the presence ofAs within the grain boundaries of polycrystalline material. Variable-energy positron beamand secondary-ion mass spectrometric studies were carried out on As+-implanted pre-amorphized samples as a function of dose and rapid thermal annealing temperature.Positron trapping was observed within negatively-charged grain boundaries of therecrystallized polycrystalline material; resulting in a 2% increase in the Doppler-broadening S-parameter value. The infusion of As+ ions into the grain boundariespassivated the charge and reduced their specific positron-trapping rate.D.W.Lawther, R.Khatri, P.J.Simpson, P.J.Schultz, I.Calder, L.Weaver: Applied SurfaceScience, 1995, 85, 265-70

[446-127/128-153]

Pipe Diffusion - Qualitative Observations - Diffusion EnhancementThe occurrence of As-enhanced diffusion along individual misfit dislocations in Si/SiGeheterostructures was detected and imaged by using scanning electron microscopic andelectron beam induced current methods. The formation of buried cylindrical, or conical,diodes surrounding misfit dislocations was observed. The diffusion enhancement was notuniform for each dislocation. The EBIC/SEM micrographs revealed a dark recombinationcontrast in the vicinity of the dislocation core, and a white generation signal within thespace-charge region of the surrounding n/p diode. On the basis of experimentallydetermined iso-concentration etching profiles, and a simple model for enhanced diffusion,the dislocation diffusion coefficient for As was estimated to be up to 6 orders ofmagnitude higher than that in the host crystal.N.Braga, A.Buczkowski, H.R.Kirk, G.A.Rozgonyi: Applied Physics Letters, 1994,64[11], 1410-2

[446-115/116-185]

Surface Diffusion - Qualitative Observations - Effect of StepsIt was pointed out that the lowest-energy configuration of a class of vicinal (100) surfaceshad equally spaced steps; each with a height of 2 atomic units. This resulted in a single-domain surface upon which the surface Si-Si dimers were aligned parallel to the stepedges. The interaction of As with vicinal (100) surfaces was crucial for GaAs growth onSi. The present scanning tunnelling microscopic, low-energy electron diffraction, and X-ray photo-emission results yielded a consistent picture of the interaction of As withvicinal (100). It was found that, depending upon the temporal order of As exposure andsurface heating, the directions of the As-As dimers could be reproducibly made to lie

246

As Surface Diffusion/Melt Diffusion As

perpendicular or parallel to the step edge. In the latter case, the step array on the Sisubstrate was completely rearranged. Any GaAs which was grown on a given type ofdimer arrangement was controllably oriented at 90º with respect to that on the other type.R.D.Bringans, D.K.Biegelsen, L.E.Swartz: Physical Review B, 1991, 44[7], 3054-63

[446-84/85-060]

Melt Diffusion - Quantitative DataDoping of molten material with As was carried out by using tertiarybutylarsine and anArF excimer laser. The As atoms diffused from a limited source at the surface, and thedepth profiles of the As concentration could be accurately described by a simple Gaussiandistribution. The diffusion coefficient of As in the Si melt was found experimentally to beabout 0.00014cm

2/s.

S.Chichibu, T.Nii, T.Akane, S.Matsumoto: Materials Science Forum, 1993, 117-118,243-8

[446-113/114-132]

247

Au

Figure 3: Diffusivity of Au in Si

37 Bulk Diffusion - Quantitative DataThe diffusion coefficient of Au, which had been implanted into chemical vapor depositedamorphous material, was measured at temperatures ranging from 400 to 800C by usingRutherford back-scattering spectrometry. Within this temperature range, the diffusion

1.0E-21

1.0E-20

1.0E-19

1.0E-18

1.0E-17

1.0E-16

1.0E-15

1.0E-14

1.0E-13

1.0E-12

1.0E-11

1.0E-10

1.0E-09

1.0E-08

1.0E-07

1.0E-06

1.0E-05

6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

table 7

table 8

table 9

table 10

table 11

table 12

table 13

table 14

table 15

table 16

104/T(K)

D (cm2/s)

248

Au Bulk Diffusion Au

coefficient varied between 10-15 and 10-10cm2/s (table 7), with an activation energy of1.5eV. The diffusion coefficient in this material correlated well with diffusion in ion-implanted amorphous material and with high-temperature diffusion in highly-dislocatedcrystals.L.Calcagno, S.U.Campisano, S.Coffa: Journal of Applied Physics, 1989, 66[4], 1874-6

[446-74-047]

Table 7Diffusivity of Au in Amorphous Si

T (C) D (cm2/s)800 4.5 x 10-11

695 5.8 x 10-12

595 1.0 x 10-12

495 5.4 x 10-14

395 2.2 x 10-15

38 Bulk Diffusion - Quantitative DataThe radiation-enhanced diffusion of implanted Au markers in amorphous Si wasmeasured at temperatures ranging from 77 to 693K (table 8). The samples werebombarded with 2.5MeV Ar ions, and the diffusion coefficients exhibited 3 well-definedranges. At temperatures below 400K, the diffusion was athermal and was due to ballisticmixing. At temperatures ranging from 400 to 700K, the diffusion was of Arrhenius type;with an activation energy of 0.37eV. It was considerably enhanced with respect to normalthermal diffusion. The defects which caused the enhanced diffusion arose from nuclearenergy loss processes. Thermal diffusion, with an activation energy of 1.42eV,predominated at temperatures greater than 750K.F.Priolo, J.M.Poate, D.C.Jacobson, J.Linnros, J.L.Batstone, S.U.Campisano: AppliedPhysics Letters, 1998, 52[15], 1213-5

[446-62/63-227]

Bulk Diffusion - Quantitative DataDetailed measurements were made of Au concentration profiles in [110]-type wafers of p-type material. The Au was introduced by means of ion implantation, and was diffused attemperatures of between 1073 and 1473K for times ranging from 60s to 100h. Theresistivity profiles were converted into Au concentration profiles by determining theentropy factor for the ionization of the Au donor level. It was suggested that the measuredprofiles and their time dependence could be explained in terms of the kick-out diffusionmechanism. It was found that the diffusivity was described by the expression:

D(cm2/s) = 0.021 exp[-1.7(eV)/kT]S.Coffa, L.Calcagno, S.U.Campisano, G.Calleri, G.Ferla: Journal of Applied Physics,1988, 64[11], 6291-5

[446-72/73-042]

249


Table 8Radiation-Enhanced Diffusivity of Au in Amorphous Si

Dose (Au/cm2) T (C) D (cm2/s)2 x 1015 407 1.4 x 10-14

2 x 1015 367 7.2 x 10-15

2 x 1015 312 3.4 x 10-15

2 x 1015 242 1.4 x 10-15

2 x 1015 157 3.4 x 10-16

5 x 1014 407 1.3 x 10-14

5 x 1014 367 6.4 x 10-15

5 x 1014 312 3.8 x 10-15

5 x 1014 242 1.6 x 10-15

39 Bulk Diffusion - Quantitative DataArtificial multi-layers of amorphous Si, and amorphous Si which contained 0.7at%Au,were prepared by ion beam sputtering. The repeat lengths were between 4.4 and 4.8nm.The change in the intensity of the first-order X-ray diffraction peak, that was due tocomposition modulation, was monitored as a function of annealing time. It was found thatthe diffusivity data at temperatures ranging from 200 to 260C (table 9) exhibited anArrhenius behavior, with an activation energy of about 1.3eV.E.Nygren, B.Park, L.M.Goldman, F.Spaepen: Applied Physics Letters, 1990, 56[21],2094-6

[446-76/77-033]

310 Bulk Diffusion - Quantitative DataThe migration of ion-implanted Au in a monocrystal was studied by using the spreadingresistance method. One-dimensional and 2-dimensional diffusion across the wafer andalong the wafer were studied by using limited or unlimited Au sources. The data (table10) could be described by:

D(cm2/s) = 0.28 exp[-1.6(eV)/kT]It was shown that, by ion-implanting Au, it was possible to produce unique concentrationprofiles via close control of the number of Au atoms in the diffusion source. Thispermitted both depth and surface profiles to be tailored. All of the measured profiles wereconsistent with a kick-out mechanism for Au diffusion.S.Coffa, L.Calcagno, S.U.Campisano, G.Ferla: Journal of Applied Physics, 1991, 69[3],1350-4

[446-78/79-049]

311 Bulk Diffusion - Quantitative DataThe ion-beam enhanced diffusion of Au in amorphous samples of undoped or B-dopedmaterial was investigated (table 11). It was found that the diffusion coefficients depended

250


linearly upon the ion flux and exhibited an Arrhenius-type temperature dependence, withan activation energy of 0.37eV, at temperatures ranging from 200 to 350C. The diffusivitywas enhanced, by a factor of 5, by B doping to a concentration of 1020/cm3. A similarenhancement of Au thermal diffusion occurred, giving an activation energy of 1.5eV. Onthe basis of the results, a model for ion-beam enhanced Au diffusion was proposed inwhich the high density of defects that was present in amorphous material acted as trapsfor fast-moving interstitial Au atoms. The effectiveness of the trapping process could bechanged by the high concentration of mobile defects which was generated by the incidentbeam and by a change, in the charge state of the traps, which was caused by the presenceof B.S.Coffa, D.C.Jacobson, J.M.Poate, F.Priolo: Applied Physics A, 1992, 54[6], 481-4

[446-93/94-047]

Table 9Diffusion of Au in Amorphous Si Multi-Layers

Repeat Length (nm) T (C) Diffusion Length (nm) D (cm2/s)4.8 200 0.23 6.0 x 10-20

4.8 200 0.32 2.5 x 10-20

4.7 220 0.34 3.9 x 10-19

4.7 220 0.44 6.3 x 10-20

4.7 220 0.46 3.0 x 10-21

4.7 240 0.39 1.7 x 10-18

4.7 240 0.52 1.8 x 10-19

4.5 250 0.37 2.3 x 10-18

4.5 250 0.42 4.5 x 10-19

4.4 260 0.35 4.0 x 10-18

4.4 260 0.39 2.8 x 10-19

4.4 260 0.50 4.6 x 10-20

4.6 260 0.31 3.2 x 10-18

4.6 260 0.43 5.9 x 10-19

4.6 260 0.49 5.7 x 10-20

4.6 260 0.51 1.7 x 10-20

312 Bulk Diffusion - Quantitative DataTransition metals in amorphous samples exhibit a direct interstitial diffusion behaviorwhich is retarded by temporary trapping at defects that are intrinsic to the amorphousstructure. Diffusion was investigated here by means of Rutherford back-scatteringspectrometry. The migration of Au tracer atoms was also studied by using neutronactivation and sputter-sectioning analyses. It was found that the data (table 12) could befitted by using foreign-atom interstitial diffusion coefficients for crystalline Si; modified

251


by the presence of traps in concentrations of between 0.2 and 1at%, and with trappingenthalpies of about 0.9eV.S.Coffa, J.M.Poate, D.C.Jacobson, W.Frank, W.Gustin: Physical Review B, 1992,45[15], 8355-8

[446-86/87-049]

Table 10Diffusivity of Au in Si

T (C) D (cm2/s)1200 1.1 x 10-6

1105 4.1 x 10-7

975 1.2 x 10-7

905 3.5 x 10-8

313 Bulk Diffusion - Quantitative DataThe diffusion of Au in dislocation-free material was investigated at 1473K by usingneutron activation analyses and mechanical sectioning. In most wafers, the diffusionprofiles were U-shaped, as in the case of previous studies performed at lowertemperatures. In these wafers, the Au concentrations in the profile centers, measured after0.75h diffusion anneals, decreased in inverse proportion to the wafer thickness, inaccordance with the so-called kick-out diffusion model. The latter was further confirmedby the good agreement of the contribution from self-interstitials to the self-diffusioncoefficient at 1473K, as calculated from the present Au diffusion data using directlymeasured values of the Si self-diffusion coefficient (table 13). In some Au diffusedwafers, W-shaped penetration profiles were found. Using spreading resistance,preferential etching, and X-ray topography studies, it was found that the Au concentrationpeaks which occurred within these specimens were located at faulted dislocation loopswhich formed during the in-diffusion of Au. An electron microscopic analysis showedthat the stacking faults were of extrinsic type. It was presumed that they were formed bythe agglomeration of supersaturated self-interstitials which were produced via the kick-out diffusion mechanism.J.Hauber, N.A.Stolwijk, L.Tapfer, H.Mehrer, W.Frank: Journal of Physics C, 1986,19[29], 5817-36

[446-49-028]

314 Bulk Diffusion - Quantitative DataRutherford back-scattering spectrometry and spreading resistance techniques were used todetermine concentration profiles for Au which had been implanted and diffused into n-type material. The diffusion annealing was carried out at 1243K for times ranging from 1to 20h. The resistivity profiles were converted into concentration profiles by solving thecharge balance equation. It was found that the diffusivity data for n-type materialcoincided with those for p-type material (table 14).

252


S.Coffa, L.Calcagno, G.Ferla, S.U.Campisano: Journal of Applied Physics, 1990, 68[4],1601-5

[446-86/87-049]

Table 11Thermal and Ion Beam-Induced Diffusivity of Au in B-Doped and Undoped Si

Treatment Dopant T (C) D (cm2/s)ion beam B 302 1.2 x 10-14

ion beam B 252 7.8 x 10-15

ion beam B 202 2.4 x 10-15

ion beam - 302 3.0 x 10-15

ion beam - 252 1.3 x 10-15

ion beam - 202 5.9 x 10-16

thermal B 507 2.1 x 10-13

thermal B 452 2.0 x 10-14

thermal B 402 3.0 x 10-15

thermal - 507 2.0 x 10-14

thermal - 452 2.0 x 10-15

thermal - 402 2.9 x 10-16

315 Bulk Diffusion - Quantitative DataDiffusion of ion-implanted elements in crystalline Si was investigated (table 15). Theimplantation was limited to photolithographically defined areas of the wafer, and aspreading resistance technique was used to measure the 3-dimensional concentrationprofiles of the metal atoms after high-temperature annealing. It was found that lateralspread under the mask was greater than vertical diffusion; especially on the side oppositeto the implanted diffusion source. All of the important features of the measured profilescould be explained as being a result of a kick-out diffusion mechanism. The peculiarshape of the concentration profiles was attributed to an interplay between the incomingflux of interstitial metal atoms and the outgoing flux of Si self-interstitials that weregenerated by the kick-out reaction. In spite of the high lateral diffusion it was noted that,by a suitable combination of implantation fluence and annealing temperature, it waspossible to limit this lateral spread to within about 200µ, while maintaining a high metalconcentration in the region under the implanted area.S.Coffa, V.Privitera, F.Frisina, F.Priolo: Journal of Applied Physics, 1993, 74[1], 195-200

[446-106/107-133]

316 Bulk Diffusion - Quantitative DataExperiments were carried out on 3 types of polycrystalline material, at temperaturesranging from 551 to 1265C, by using 195Au radiotracer and serial sectioning techniques(table 16). The 3 materials were termed Siemens (1015B/cm3, 30µ grain size, dislocationdensity of about 107/cm2), Polix (1016B/cm3, 5 to 25mm grain size, dislocation density of

253


between 105 and 107/cm2) and MRC (1016O2/cm3, 1018C/cm3, non-uniform grain size).Depending upon the temperature, material structure, and diffusion time, various types ofprofile were obtained. Some of these could not be explained, within the framework ofclassical solutions of Fick’s equations, in the presence of grain boundaries. However, allof the profiles could be successfully analyzed by using a new diffusion model which wasapplicable to host-impurity systems which involved strong segregation effects andnegligible diffusion along extended defects. At temperatures above about 1100C,effective diffusion coefficients that were slightly lower then those previously reported inthe literature were measured. Upon decreasing the temperature, the effective Audiffusivity began to depend upon the structure of the sample. This produced variousArrhenius plots which exhibited a downward curvature. A segregation enthalpy, for Au,of about 141kJ/mol was estimated from the observed diffusion behavior.C.Poisson, A.Rolland, J.Bernardini, N.A.Stolwijk: Journal of Applied Physics, 1996,80[11], 6179-87

[446-141/142-113]

Table 12Diffusivity of Au in Amorphous Si

T (C) D (cm2/s)250 3.1 x 10-19

280 2.4 x 10-18

345 2.8 x 10-17

350 2.1 x 10-16

375 2.2 x 10-16

385 1.2 x 10-15

405 6.1 x 10-16

430 3.1 x 10-15

455 4.9 x 10-15

500 3.5 x 10-14

550 1.9 x 10-13

Bulk Diffusion - Quantitative DataSamples with a deposited film of Au were irradiated with a neutron flux of 6 x 1018/cm2 at50C, and then annealed. The redistribution of Au was studied by using the neutronactivation method. It was found that some of the adsorbed Au atoms diffused into the bulkunder the influence of the irradiation. It was estimated that the radiation-enhanceddiffusion coefficient was between 10-13 and 10-12cm2/s.T.P.Svistelnikova, T.V.Moiseenkova, F.P.Korshunov, N.A.Sobolev, V.A.Kharchenko:Izvestiya Akademii Nauk SSSR - Neorganicheskie Materialy, 1991, 27[5], 1079-80(Inorganic Materials, 1991, 27[5], 902-4)

[446-88/89-049]

254


Table 13Diffusivity of Au in Si

T (C) D (cm2/s)1200 1.0 x 10-14

1100 7.2 x 10-16

1050 1.4 x 10-16

1005 3.0 x 10-17

900 6.2 x 10-19

800 8.4 x 10-21

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe P diffusion gettering of Au is a reversible process which exhibits marked temperatureand P concentration dependences. It was shown that Au diffused back and forth betweenthe highly-doped P layer and the bulk material when the annealing temperature wasvaried. Secondary ion mass spectroscopy was used to study Au in the gettering layer, anddeep-level transient spectroscopy was used to estimate the Au content in the bulk. Nointernal gettering or Au out-diffusion was detected when the Au concentration was belowthe solubility limit. The Au concentration profile after successful gettering followed the Pprofile, but almost all of the Au atoms were found in regions where the P concentrationexceeded about 3 x 10

19/cm

3. This was attributed to a large enhancement of Au solubility

which occurred when the P concentration was greater than 3 x 1019

/cm3. The simplest

explanation for the observed gettering was the assumed formation of Au-P pairs withinthe highly-doped P layer. However, quantitative agreement could not be obtained betweenthe present results and a simple segregation model. If the Au bulk solubility was modifiedby taking account of a supersaturation of Si self-interstitials, agreement between themodel and the experimental data was then obtained.E.O.Sveinbjörnsson, O.Engström, U.Södervall: Journal of Applied Physics, 1993, 73[11],7311-21

[446-106/107-132]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe diffusion profiles of Au in P-doped regions were measured. It was found that, due tothe gettering effect of P, the Au concentration was considerably lower than in controlwafers without P. A model was developed which took account of Au-P pairing and ofself-interstitial injection from the P regions, and was used to simulate the measured Auprofiles.H.Zimmermann: Materials Science Forum, 1994, 143-147, 1647-52

[446-113/114-040]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesA wafer of float-zone material was partially subjected to prolonged annealing at 960C.The annealed and unannealed portions of the wafer were then diffused with Au for

255


various lengths of time. The Au concentration was deduced from spreading resistancemeasurements. It was found that the crystal portion which was annealed before Audeposition exhibited a different Au accumulation, as a function of time, than did the as-received parts of the crystal. The Au concentration at the center of the wafer did not obeythe previously reported t½ behavior. It instead followed a curve that exhibited aninflection. The results were explained in terms of vacancy clusters that were retainedduring the original crystal growth.T.K.Monson, J.A.Van Vechten, Q.S.Zhang: Journal of the Electrochemical Society, 1995,142[6], 2077-80

[446-134/135-156]

Table 14Diffusivity of Au in n-Type and p-Type Si

Type T (C) D (cm2/s)p 1180 2.0 x 10-8

n 1105 1.0 x 10-8

p 1105 8.8 x 10-9

n 955 1.8 x 10-9

p 955 1.6 x 10-9

p 890 6.8 x 10-10

p 802 1.6 x 10-10

Bulk Diffusion - Qualitative Observations - Defect InteractionsExperiments were performed on the short-term diffusion of Au at temperatures rangingfrom 900 to 1100C. A complete set of parameters was determined from these experimentsby using the Arrhenius law. It was found that the short-term diffusion experiments couldnot be simulated without assuming barrier energies for both the Au point-defect reactionand the Frenkel pair reaction. Their values were EAu/i = 0.482eV, EAu/v = 0.971eV and Ei/v

= 0.30eV.K.Ghaderi, G.Hobler, M.Budil, L.Mader, H.J.Schulze: Journal of Applied Physics, 1995,77[3], 1320-2

[446-121/122-084]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe effect of a low-temperature (about 100C) plasma treatment upon the properties ofnear-surface layers was studied. It was found that a decrease in the concentration of deep-level centers, as well as the formation of new vacancy-related complexes at a depth of0.001 to 0.01mm, occurred. The changes in the properties were closely related to thegeneration of intrinsic point defects. Experimental data which confirmed the possibility ofenhanced Au diffusion were presented.S.V.Koveshnikov, E.B.Yakimov, N.A.Yarykin, V.A.Yunkin: Physica Status Solidi A,1989, 111[1], 81-8

[446-64/65-177]

256


Bulk Diffusion - Qualitative Observations - Defect InteractionsBy using a dissociative model, the diffusion of Au atoms into Si plates was investigated attemperatures below 1270K. It was found that, as the plate thickness was decreased, theinterstitial regime could change to the vacancy regime for diffusion transfer.A.V.Vaisleib, M.G.Goldiner. Physics Letters A, 1990, 146[7-8], 421-5

[446-76/77-033]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe migration of Au in amorphous material was studied by using Rutherford back-scattering spectrometry, neutron activation and serial sectioning techniques. It wasconcluded that, like transition metals, Au underwent direct interstitial diffusion whichwas retarded by temporary trapping at defects that were intrinsic to the amorphousstructure. Consequently, Au diffused more rapidly in amorphous samples than indislocation-free crystalline material; where it diffused via the kick-out mechanism. Thesame was true of Pt and Zn, while Cu and Pd diffused more slowly in amorphousmaterial, where they were unretarded interstitial diffusers.W.Frank, W.Gustin, S.Coffa, J.M.Poate, D.C.Jacobson: Materials Science Forum, 1992,83-87, 203-8

[446-93/94-046]

Table 15Diffusivity of Au in Dislocated and Dislocation-Free Si

Material T (C) D (cm2/s)dislocated 1215 3.0 x 10-6

dislocated 1090 7.0 x 10-7



dislocation-free 1080 2.1 x 10-8




Bulk Diffusion - Qualitative Observations - Defect InteractionsIt was found that the study of Au diffusion permitted the separate observation of Si self-interstitials and vacancies. The diffusion of Au in samples of float-zone material wasgoverned by the kick-out mechanism at temperatures above 800C. The use of numericalsimulation furnished a consistent set of parameters which described the diffusion of Au attemperatures ranging from 800 to 1100C. The generation or recombination of self-interstitials and vacancies was ineffective; at least below 850C.H.Zimmermann, H.Ryssel: Applied Physics A, 1992, 55[2], 121-34

[446-93/94-047]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe carrier concentration profiles in low-O (float-zone) lightly-doped p-type materialafter Au diffusion were found to be a function of the heat treatment which was given to

257


the crystal before Au deposition. The anomalous carrier concentration profiles wereexplained in terms of vacancy diffusion, and clusters of excess vacancies which wereretained during crystal growth.T.K.Monson, J.A.Van Vechten, Q.S.Zhang: Applied Physics Letters, 1995, 66[7], 854-6

[446-121/122-084]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe concentration, x, of substitutional atoms which transformed via the kick-outmechanism during Au diffusion was measured as a function of d, the density ofdislocations. Well-characterized dislocation distributions were produced in the float-zone(111) samples by using a special deformation procedure. Two domains, with d less than10

4/cm

2 or greater than 10

7/cm

2, were found which were such that x was independent of d.

In the intermediate range, x was proportional to d½; in agreement with theory. The sink

efficiency was equal to 0.3. This value was higher than previously reportedmeasurements. It was concluded that the sink efficiency was sensitive to the type ofdislocation which was involved. There was a wide transient zone, between the d

½

behavior and the constant value above 107/cm

2, whose cause was unclear.

B.Pichaud, G.Mariani: Journal de Physique III, 1992, 2[3], 295-302[446-099/100-096]

Table 16Diffusivity of 195Au in Si

Sample Type Profile Type T (C) D (cm2/s)MRC 0 897 8.11 x 10-9

MRC 0 949 2.95 x 10-8

Siemens 0 981 2.37 x 10-8

MRC 0 981 2.44 x 10-9

Siemens 0 1035 4.08 x 10-8

MRC 0 1035 1.19 x 10-8

Siemens 0 1060 7.43 x 10-8

Polix 0 1060 1.48 x 10-7

Siemens 0 1099 1.29 x 10-7

MRC 0 1117 1.95 x 10-7

MRC 0 1152 9.73 x 10-8

Siemens 0 1193 4.34 x 10-7

MRC 0 1265 3.82 x 10-7

MRC 1 665 2.72 x 10-13

Siemens 1 721 8.06 x 10-12

MRC 1 721 1.50 x 10-12

MRC 1 752 2.42 x 10-10

Siemens 1 778 1.89 x 10-9

MRC 1 806 3.79 x 10-11

Siemens 1 851 6.77 x 10-10

258


Bulk Diffusion - Qualitative Observations - Defect InteractionsThe Au concentration was measured as a function of the dislocation density, which wasintroduced by various deformation processes in float-zone or Czochralski material, inorder to check a theory of Au diffusion which involved a self-interstitial sink efficiencycoefficient. In this way, the value of the coefficient could be determined. It was noted thatthe value of the coefficient was very sensitive to the nature of the dislocations which wereinvolved and to the O concentration. On the other hand, transmission electronmicroscopic observations showed that the dislocations were sinks for self-interstitials.B.Pichaud, G.Mariani, W.J.Taylor, W.S.Yang: Solid State Phenomena, 1994, 35-36, 491-6

[446-111/112-069]

Bulk Diffusion - Qualitative Observations - Defect InteractionsIt was shown here that Au diffused back and forth, between a highly P-doped layer andthe bulk of the material, when the annealing temperature was varied. This behavior wasinvestigated by using secondary ion mass spectroscopy to study Au within the getteringlayer, and by using deep-level transient spectroscopy to estimate the Au content of thebulk. It was found that the gettered Au was located within the P profile, but was notgettered to the surface. No internal gettering or out-diffusion of Au was observed whenthe Au concentration was below the solubility limit. The results could not be described byusing a simple segregation model unless the Au solubility in the bulk was modified bytaking account of a possible supersaturation of Si self-interstitials.E.O.Sveinbjörnsson, O.Engström, U.Södervall: Materials Science Forum, 1994, 143-147,1641-6

[446-113/114-041]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe substitutional Au concentration which was introduced by a diffusion step attemperatures of between 850 and 1000C was measured, by means of deep-level transientspectroscopy, in float-zone and Czochralski material which contained various dislocationdensities. A comparison, for a given sample, of dislocated and non-dislocated regionspermitted the efficiency of dislocations as sinks for self-interstitials to be measured as afunction of the diffusion temperature. In the case of float-zone material, the aboveefficiency was found to be temperature-independent. In the case of Czochralski material,a marked temperature dependence of the efficiency was observed. This was attributed tothe release of dislocations from obstacles via a thermally stimulated climb mechanism.E.Yakimov, G.Mariani, B.Pichaud: Journal of Applied Physics, 1995, 78[3], 1495-9

[446-123/124-181]

Bulk Diffusion - Qualitative Observations - Defect InteractionsA study of Au diffusion was used to clarify the effect of P diffusion upon the defectstructure of 2-step annealed wafers of Czochralski material which contained O-relatedprecipitates. By means of deep level transient spectroscopy, and the use of publishedrelationships, it was established that the substitutional Au concentration was independentof O precipitation. On the other hand, P diffusion (via the injection of self-interstitialsinto

259


the bulk) shrank the O precipitates and increased the substitutional Au concentration tonear to the solubility limit.E.Yakimov, I.Périchaud: Applied Physics Letters, 1995, 67[14], 2054-6

[446-125/126-145]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe migration of some elements yields Si self-interstitial diffusivities that exceed thosewhich are obtained from dopant marker experiments by 6 orders of magnitude at 800C. Itwas noted that both types of experiment could be reconciled by assuming the existence ofa non-annihilating interstitial trap that was related to C. Selected metal diffusion datawere re-analyzed in this context. Non-annihilating immobile traps and a second-orderreaction which involved interstitial C and C-C pairs were considered. Quantitative pointdefect parameters were obtained at 1115C for an assumed trap concentration of 5 x1016/cm3. The estimated equilibrium concentration of Si self-interstitials varied inverselywith the trap concentration, while the product of the self-interstitial diffusivity and theself-interstitial concentration remained almost constant. It was concluded that agreement,or disagreement, of metal diffusivity results with the Si self-diffusion coefficient couldnot be used to exclude or prove the occurrence of trap-limited diffusion. The publishedvalues of point-defect parameters which had been obtained by neglecting traps weresuggested to represent lower bounds on the self-interstitial diffusivity, and upper boundson the equilibrium concentration of self-interstitials.H.J.Gossmann, P.A.Stolk, D.J.Eaglesham, C.S.Rafferty, J.M.Poate: Applied PhysicsLetters, 1995, 67[21], 3135-7

[446-127/128-154]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe diffusion and gettering of Au to defects which were introduced by H implantationwere studied in (100) wafers of Czochralski material that had first been implanted with 50and/or 100keV H ions. The damaged layers were then annealed at 850C in order to drivethe H out and leave a band of well-faceted cavities that were close to the originalprojected range. A known amount of Au was introduced, by ion implantation, into thenear-surface region. The Au profiles were determined by using Rutherford back-scatteringspectrometry and channelling, and the microstructure was studied by means of cross-sectional transmission electron microscopy. The gettering of Au to cavities, and themovement of Au from one cavity to another, were studied as a function of time andtemperature. The Au concentration at cavities was found to be strongly dependent uponthe annealing time. The movement of Au from one set of cavities to another was found tobe markedly affected by the interplay between Au diffusion and solubility, and theavailability of strongly reactive bonding sites at cavity walls.J.Wong-Leung, J.S.Williams, E.Nygren: Nuclear Instruments and Methods in PhysicsResearch B, 1995, 106[1-4], 424-8

[446-136/137-127]

Bulk Diffusion - Qualitative Observations - Defect InteractionsMonocrystals which contained extended defects (dislocations, stacking faults) or whichwere defect-free, were diffused with Au at 1150C. It was found that the diffusion of Au in

260


crystals without defects, and diffusion very near to the surface in crystals with defects,was limited by the diffusion of self-interstitials to the surface. The diffusion in crystalswith defects was limited by the diffusion of interstitial Au atoms from the surface. Incrystals with an extremely low concentration of extended defects, diffusion was limitedby the migration of self-interstitials to the defects.M.Morooka: Japanese Journal of Applied Physics 1, 1996, 35[5A], 2537-43

[446-136/137-127]

Bulk Diffusion - Qualitative Observations - Defect InteractionsRadiotracer 195Au atoms that were produced by the decay of implanted 195Hg ions wereused to study the diffusivity of Au in amorphous samples by means of serial Ar+-beamsectioning. The Au diffusivity data were explained in terms of a process in which thediffusing 195Au atoms were temporarily trapped by various types of vacancy-like defect.These traps could be saturated with Au or H. Their nature and concentration were foundto change as a result of annealing-induced structural relaxation of the amorphous samples.W.Frank, W.Gustin, M.Horz: Journal of Non-Crystalline Solids, 1996, 205-207[1], 208-11

[446-148/149-179]

Bulk Diffusion - Qualitative Observations - Diffusion EnhancementThe simultaneous diffusion of Au and B was studied at temperatures of between 800 and1200C by using spreading resistance and Rutherford back-scattering spectrometricmethods. Islands of Au formed on the Si surface at high temperatures. In general, Bdiffusion was enhanced and driven deeper with increasing Au content. The enhancementwas attributed to a form of Kirkendall effect.D.K.An, K.Madl, A.Barna, G.Battistig, J.Gyulai: Physica Status Solidi A, 1989, 116[2],561-9

[446-76/77-033]

Bulk Diffusion - Qualitative Observations - Doping EffectsSingle crystals of n-type material were diffused with Au at temperatures of 800 to 1150C.Deep-level transient spectroscopy revealed peaks at Ec - 0.16, Ec - 0.24, Ec - 0.30, and Ec -0.54eV. As the Au concentration and Au diffusion temperature were increased, therelative height of the Ec - 0.24eV peak increased in comparison with the most prominentpeak at Ec - 0.54eV. The concentration of the Ec - 0.30eV peak saturated at between 1014

and 4 x 1014/cm3.M.Morooka, H.Tomokage, M.Yoshida: Japanese Journal of Applied Physics, 1988, 27[9],1778-9

[446-62/63-227]

Bulk Diffusion - Qualitative Observations - Effects of Electron IrradiationRadiotracer, secondary ion mass spectrometry, and electrical methods were used to studythe radiation-stimulated diffusion of Au. It was found that acceleration of the diffusion byelectron irradiation was entirely due to the generation of excess vacancies and was notdue to ionization processes.

261


J.N.Kazarinov, V.V.Kozlovskii, V.N.Lomasov, M.V.Pitkevich: Fizika i TekhnikaPoluprovodnikov, 1986, 20[9], 1577-81. (Soviet Physics -Semiconductors, 1986, 20[9],989-92)

[446-51/52-129]

Bulk Diffusion - Qualitative Observations - Effect of Intermediate LayerInterlayer interaction and phase formation in thin multi-layer metallization systems(Au/Pd/Ti, Au/Pd/Cr/Ti) on (111)Si were investigated. It was found that the additionallayer (Cr) in the metallization system was no barrier to the interdiffusion of Au and Siduring annealing at 650K in air. In fact, the Cr film accelerated the diffusion of Au to thesupport.J.N.Makogon, S.I.Sidorenko, J.A.Bazarnyi, N.S.Voltovets, T.V.Litvinova,V.L.Tkachenko: Izvestiya Akademii Nauk SSSR - Neorganicheskie Materialy, 1990,26[12], 2461-6 (Inorganic Materials, 1990, 26[12], 2116-20)

[446-84/85-100]

Bulk Diffusion - Qualitative Observations - Effects of Ion ImplantationRutherford back-scattering spectrometry and cross-sectional transmission electronmicroscopy were used to study the implantation of MeV Au+ ions. The measured rangeand straggle values for the implanted samples were found to be consistently larger thanthe values which were predicted by simulations. The magnitude of the discrepancies wassuch that the differences could not be attributed to implantation effects alone. Theexperimental results showed that a single Gaussian Au profile was obtained in the case oflow-current low-energy implants. Low-power implants produced a single damage bandwhich consisted of simple point defects. High-current high-energy implants led to thecreation of more complex defect structures, such as dislocation networks. These arose asa result of dynamic beam recrystallization. Multiple layers of precipitation were observed,in material which had been implanted with MeV Au+ ions, when dynamicrecrystallization occurred. Precipitation occurred because the local Au concentrationexceeded the solid solubility during beam-induced recrystallization. Various mechanismsoperated so as to cause an anomalous Au migration which resulted in the formation ofmultiple precipitate layers. One mechanism caused the implanted Au to segregate into adensely defected region. When the concentration exceeded the local solid solubility, Auprecipitated out of the matrix. Another mechanism involved Au migration along thedislocations in a network. Here, the diffusing Au reached a dislocation node where itexceeded the local threshold for precipitation and therefore precipitated out. Theoccurrence of enhanced Au diffusion depended upon the degree of dynamicrecrystallization which occurred during implantation.T.L.Alford, N.D.Theodore: Journal of Applied Physics, 1994, 76[11], 7265-71

[446-119/120-222]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe diffusion of Au, out of a film which had been deposited onto a Si wafer, was studiedduring ion bombardment of the back of the wafer. The ions (100keV Ne+) produced adisplacement of the Au atoms from normal sites, and prevented the diffusion of Au awayfrom the surface and into the interior of the sample. A similar effect was found during the

262


excitation of an acoustic wave in the crystal. It was suggested that the phenomenon whichwas observed during bombardment arose from elastic waves which were created as theions were stopped.I.V.Antonova, S.S.Shaimeev: Fizika i Tekhnika Poluprovodnikov, 1995, 29[1], 3-8(Semiconductors, 1995, 29[1], 1-3)

[446-134/135-156]

Bulk Diffusion - Qualitative Observations - Effect of PrecipitatesThe effect of C and O precipitation upon the diffusivity of Au in monocrystallineCzochralski- and float-zone material was studied by using infra-red absorption andcharged-particle activation analysis methods. It was found that the diffusion of Au wasenhanced by the presence of C and/or O precipitates. In the case of Czochralski-typematerial, the enhancement was attributed to the precipitation of C and O. In the case offloat-zone material, the diffusion of Au was not enhanced; due to the lower Oprecipitation and absence of C precipitation.Y.Itoh, Y.Sugita, T.Nozaki: Japanese Journal of Applied Physics, 1989, 28[10], 1746-9

[446-72/73-042]

Bulk Diffusion - Qualitative Observations - Effect of PrecipitatesThe effect of O precipitation upon Au diffusion was investigated. It was found that anincrease in the concentration of precipitated O, and therefore a decrease in theconcentration of free O, was associated with a monotonic increase in the specificresistance of the samples. This indicated a more rapid entry of the Au.A.A.Vlasov, V.L.Kryukov, G.P.Furmanov, S.E.Cheshuina: Izvestiya Akademii NaukSSSR - Neorganicheskie Materialy, 1990, 26[12], 2649-50 (Inorganic Materials, 1990,26[12], 2280-1)

[446-84/85-060]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsThe Au was diffused into n-type crystals, which contained nuclei for point defects, at1150C. Hexagonal stacking faults were induced by the Au diffusion. Some of thesespecimens were heat-treated again at 1000C by continuous annealing. The size anddensity of the stacking faults depended upon the substitutional Au concentration, andcould be controlled by the Au diffusion. During in-diffusion, the size and areal density ofthe faults were proportional to the square root of the Au concentration. During annealing,the areal and volume densities of the faults were proportional to the Au concentration.The size during annealing, and the volume density during in-diffusion, were independentof the Au diffusion.M.Morooka, M.Takahashi, F.Hashimoto: Materials Science Forum, 1994, 143-147, 1523-8

[446-113/114-042]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsThe concentration profiles of the deep levels (Ec - 0.54eV, Ec - 0.24eV, Ec - 0.30eV, Ec -0.16eV) which were introduced by Au diffusion (1150C, 6h) were measured by usingdeep-level transient spectroscopic methods. The profile of Ec - 0.54eV agreed well with

263


that of Au atoms, as measured by means of neutron activation analysis. The shape of theprofile of Ec - 0.24eV was very similar to that of Ec - 0.54eV, and the concentration ratioof Ec - 0.24eV to Ec - 0.54eV was almost constant at 0.12. The existence of the Ec -0.16eV level depended upon the method which was used to prepare the Schottky diodefor deep-level transient spectroscopic measurements. The shape of the profile of the Ec -0.16eV level, when observed, was similar to that of the Ec - 0.54eV level. The profile ofEc - 0.30eV was quite different to those of other levels, and the concentration level wasalmost constant throughout the thickness.M.Takahashi, M.Morooka, F.Ueda, F.Hashimoto: Japanese Journal of Applied Physics 1,1994, 33[4A], 1713-6

[446-113/114-040]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsIt was found that regular hexagonal stacking faults were generated, on 4 equivalent (111)-type planes of some crystals, during Au diffusion at 1150C. The stacking faults formed inregions which were more than 0.1mm from the specimen surface, and their size andnumber increased with increasing substitutional Au concentration. Their numberdecreased with decreasing Au concentration. The number of point defects whichcontributed to the growth of stacking faults was almost equal to that of the substitutionalAu atoms which diffused into the material. Each stacking fault was of extrinsic type andcontained one atomic layer of point defects. Reaction between the nuclei of stackingfaults, and the point defects, was localized rather than being uniformly distributed. Thepoint defect capture radius of sink/sources was equal to about 0.065mm at 1150C. Thus,the specimen surface was the principal sink/source for point defects in regions whichwere closer to the surface than 0.065mm.M.Morooka, M.Takahashi, F.Hashimoto: Japanese Journal of Applied Physics 1, 1992,31[8], 2327-32

[446-93/94-048]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsDeep-level transient spectroscopy was used to study the energy levels which wereintroduced by Au diffusion in polycrystalline p-type Si. Diffusion annealing (900C, 6h)was carried out in an Ar flow after a Au layer had been deposited, by cathodic sputtering,onto one side of the sample. Diodes of Al-Si were created on the other side of thesamples, in grains, and on grain boundaries. The results showed that Au atoms reachedthe other side of the sample (40nm). They were in substitutional sites, and gave rise to thewell-known donor level (Ev + 0.35eV) in p-type Si. However, in the case of the diodeswhich covered the grain boundaries, an additional level was found at 0.43eV above thevalence band. The enhanced diffusion of Au atoms through the grains was explained interms of the kick-out mechanism. In the case of the grain boundaries, it was suggestedthat the additional level might result from the formation of Au-Fe or Au-defectcomplexes.M.Pasquinelli: Journal de Physique III, 1992, 2[3], 303-11

[446-099/100-096]

264


Bulk Diffusion - Theoretical Analysis - Defect InteractionsAccount was taken of substitutional Au, interstitial Au, vacancies, and self-interstitials inthe partial differential equations for Au diffusion involving dissociative and kick-outmechanisms. As it was time-consuming to solve the equations numerically, it was foundmore convenient to solve the approximate partial differential equation (for substitutionalAu) which was obtained after reactions between the 4 species had reached their localequilibrium states and the interstitial Au had reached its thermal equilibrium state undersuitable boundary and initial conditions. In the case of in-diffusion processes, theconditions for the approximate equation were different to those for the basic equation. Inthe case of annealing processes, the conditions were the same for the basic and for theapproximate equation.M.Morooka, M.Yoshida: Japanese Journal of Applied Physics, 1989, 28[3], 457-63

[446-70/71-122]

Bulk Diffusion - Theoretical Analysis - Defect InteractionsAnalytical techniques and computer simulation methods were used to study the out-diffusion characteristics of Au. The analysis was based upon the kick-out mechanism, andassumed the applicability of various boundary conditions. An estimate for the Aubackside gettering efficiency was obtained, and various process-limiting factors wereconsidered. In particular, out-diffusion simulation results were presented for Sii

recombination and generation in a given volume.H.Gdanitz, K.Schmalz: Physica Status Solidi A, 1990, 117[2], 395-402

[446-76/77-033]

Bulk Diffusion - Theoretical Analysis - Defect InteractionsMany analyses of the effects of surface treatments upon atomic diffusion have explainedthese effects mainly in terms of the behavior of Si self-interstitials. However, it wasshown here (by means of straightforward Monte Carlo simulations of the diffusion of Auinto Si according to the Seeger kick-out mechanism) that this mechanism was completelyincapable of explaining the 2-sided U-shaped profile of substitutional Au which resultedfrom the 1-sided in-diffusion of Au. It was shown that, if Au interstitials could displace Siat any appreciable rate, then the Au substitutional profile had to decrease monotonicallyfrom the source side to the far side. It was suggested that this provided strong evidencefor the concept that Si self-interstitials played no role in thermal processes in Si. Theresult was that the surface treatment effects which were often attributed to Si self-interstitials could be explained without invoking them.J.A.Van Vechten, U.Schmid, Q.S.Zhang: Journal of Electronic Materials, 1991, 20[6],431-40

[446-91/92-029]

Bulk Diffusion - Theoretical Analysis - Defect InteractionsThe diffusion of Au into monocrystals was analyzed in terms of the dissociative and kick-out mechanisms. Expressions were derived for the critical sample thicknesses andtemperatures at which one mechanism was replaced by the other.

265

Au Bulk Diffusion/Surface Diffusion Au

A.V.Vaisleib, M.G.Goldiner, O.J.Keloglu, I.N.Kotov: Journal of Applied Physics, 1991,70[11], 6809-14

[446-91/92-029]

Bulk Diffusion - Theoretical Analysis - Defect InteractionsThe mechanism of Au diffusion was studied by solving a set of equations which tookaccount of various couplings between impurities and point defects. It was concluded thatthe Frank-Turnbull mechanism did not permit a good description of the bulk profiles forAu diffusion from the 2 faces of thin samples. Only the kick-out mechanism furnished anaccurate description of the bulk behavior, and the contributions of the 2 interchangemechanisms had to be taken into account in order to describe the surface profiles. Theparameters which controlled the behavior of native point defects were determined byanalyzing Au and dopant diffusion in a consistent manner.D.Mathiot: Physical Review B, 1992, 45[23], 13345-55

[446-93/94-046]

Bulk Diffusion - Theoretical Analysis - Defect InteractionsMultiple-pathway diffusion with traps was analyzed within the kinetic approach, wherethe diffusion terms in the diffusion-kinetic equations were replaced by kinetic ones. Asthe solutions were linear combinations of exponents, both the components and the pre-exponential factors were taken into account. As an example, Au diffusion into a Si filmwas studied, and account was taken of both dissociative and kick-out mechanisms. Inorder to analyze thin-film diffusion, thin-film surfaces were considered to be substantialsources and/or sinks of point defects in addition to terms which described diffusionaltransport in crystals of infinite size. The various diffusion mechanisms and regimes, andthe conditions for transitions from one mechanism and/or regime to another, weredetermined. Expressions were obtained for the critical sample thickness, dislocationdensity, swirl and vacancy cluster densities when transitions in the diffusion transportmechanisms and regimes occurred. Quantitative criteria for the critical values oftemperature, dislocation density and thin-film thickness transitions of the Au-Si diffusiontransport mechanisms were obtained. It was found that, for fairly long times, the resultswhich were obtained using the kinetic approach coincided with those of the diffusion-kinetic approach. However, the set of pure kinetic equations was much simpler to solve. Itwas concluded that the kinetic approach was a valuable approximation to multiple-pathway diffusion with traps.M.G.Goldiner, A.V.Vaysleyb: Physical Review B, 1995, 52[14], 10060-8

[446-125/126-145]

Surface Diffusion - Quantitative DataThe movement of a patch of Au film on clean (111) surfaces, which occurred when adirect current was fed through the substrate, was investigated by using scanning Augermicroscopy. The Au layer exhibited a directional movement towards the anode, inaddition to isotropic spreading via surface thermal diffusion. In time, the movement of theover-layer exhibited 3 distinct processes successively. These were: fast isotropicspreading (supplemented by directional movement over an intermediate monolayer), theadvance of the foremost over-layer on the anode (at the expense of coverage on the

266

Au Surface Diffusion Au

cathode side), and a slow directional movement of the whole layer. The latter producedan oscillatory profile of the Auger peak height on the cathode side. The lateral flux ofatoms at the leading edge was 8.1 x 108 and 2.0 x 108/cm-s (for a direct current of 352mAat 596C), respectively, for the latter 2 processes. The corresponding activation energieswere 0.77 and 1.2eV.H.Yasunaga, E.Sasuga: Surface Science, 1990, 231[3], 263-70

[446-76/77-034]

Surface Diffusion - Qualitative Observations - AsymmetryThe intra-row diffusion of Au atoms on the (111)-(5 x 2)Au reconstructed surface wasstudied by means of high-temperature scanning tunnelling microscopy. It was concludedthat loosely bonded Au atoms, which were arranged with 2(n+1)a spacings at roomtemperature, diffused in the 2-fold direction at higher temperatures among adsorbed siteswhich were arranged at intervals of 2a. It was suggested that these diffusing Au atomswere responsible for the presence of a non-integral number of Au atoms in the (5 x 2) unitcell.T.Hasegawa, S.Hosoki: Physical Review B, 1996, 54[15], 10300-3

[446-140-008]

267

B

Figure 4: Diffusivity of B in Si

Bulk Diffusion - Quantitative DataThe co-diffusion of As and B in monocrystalline samples was studied by means ofsecondary ion mass spectrometry and rapid thermal annealing. The migration of B aloneduring annealing at 1050 to 1100C could be described by:

D(cm2/s) = 3 exp[-3.43(eV)/kT]while the co-diffusion could be described by:

As: D(cm2/s) = 22.83 exp[-4.10(eV)/kT]B: D(cm2/s) = 0.9 exp[-3.43(eV)/kT]

1.0E-18

1.0E-17

1.0E-16

1.0E-15

1.0E-14

1.0E-13

1.0E-12

1.0E-11

1.0E-10

7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

table 17table 18table 21table 22table 23table 25

104/T(K)

D (cm2/s)

268

B Bulk Diffusion B

C.Gontrand, P.Ancey, H.Haddab, G.Chaussemy: Semiconductor Science andTechnology, 1992, 7[2], 181-7

[446-88/89-048]

Bulk Diffusion - Quantitative DataThe co-diffusion of As and B which had been implanted to doses of about 1016/cm2, usingenergies which gave the same projected range, was investigated at 900 and 1000C bymeasuring dopant and carrier profiles. A comparison of co-diffusion data, with the resultswhich were obtained by the separate diffusion of each element, revealed anomalouseffects which could be explained by positing the formation of neutral donor-acceptorpairs. These complexes were mobile, with diffusivities that could be described by:

D (cm2/s) = 17 exp[-4(eV)/kT]Such values were very close to the diffusivity of As in intrinsic Si. On the basis of theseresults, a diffusion model was proposed which took pairing into account. A simulationwhich included this model permitted the prediction of the anomalous phenomena thatoccurred during the high-concentration co-diffusion of donors and acceptors. Theagreement with observed profiles was generally good.S.Solmi, S.Valmorri, R.Canteri: Journal of Applied Physics, 1995, 77[6], 2400-6

[446-121/122-084]

317 Bulk Diffusion - Quantitative DataMigration into an underlying Si substrate, from CoSi2 layers which had been implantedwith B ions, was studied by using a high-resolution carrier delineation technique. In theearly stages of diffusion, the junction shape followed the silicide/silicon interface. Byusing 2-step annealing, or a thin silicide diffusion source, a laterally uniform junction wasobtained with As-implanted CoSi2. The diffusion coefficients of B (table 17) could bemeasured by using this technique. The activation energy for B diffusion was 3.8eV.F.La Via, C.Spinella, E.Rimini: Semiconductor Science and Technology, 1995, 10[10],1362-7

[446-125/126-144]

Table 17Diffusivity of B in Si

T (C) D (cm2/s)1002 1.5 x 10-13

902 1.9 x 10-14

852 1.3 x 10-15

Bulk Diffusion - Quantitative DataThe diffusion of B from polycrystalline material into monocrystalline material wasmeasured for B concentrations ranging from 1020 to 7 x 1020/cm3 by using secondary ionmass spectrometry. The B concentration on the monocrystalline side of the interface wasindependent of the diffusion time when either B-implanted or in situ doped

269

B Bulk Diffusion B

polycrystalline material was used. The diffusivities found ranged from 4.2 x 10-14 to 9.0 x10-14cm2/s.B.Garben, W.A.Orr-Arienzo, R.F.Lever: Journal of the Electrochemical Society, 1986,133[10], 2152-6

[446-51/52-134]

Table 18Diffusion of B in Si Molecular Beam Epitaxial Layers

as a Function of the Growth Temperature

T (C) D (cm2/s)540 1.7 x 10-15

560 2.2 x 10-15

590 3.2 x 10-15

600 2.1 x 10-15

610 4.8 x 10-15

660 6.9 x 10-15

750 9.8 x 10-15

Bulk Diffusion - Quantitative DataProfiling of the B distribution was carried out by using nuclear reaction techniques. Noevidence was found for any measurable B diffusion at 285C in glow-discharge H-dopedamorphous material, and only an upper limit (of less than 7 x 10-20cm2/s) upon thediffusion coefficient could be obtained.H.P.Schölch, S.Kalbitzer, D.Fink, M.Behar: Materials Science and Engineering B, 1988,1, 135-7

[446-62/63-228]

Bulk Diffusion - Quantitative DataThe transient enhanced diffusion of B, which had been ion-implanted to a dose of1014/cm2, was studied by means of rapid thermal annealing and secondary ion massspectrometry. It was found that the fraction of B diffusivity which was due toimplantation damage decreased with time according to:

D = Doexp[-t/L]where Do was the diffusion coefficient at t = 0 and L was the lifetime of the point defectwhich caused the enhanced diffusion. When the implantation energy was 150keV, themeasured parameters were:

Do(cm2/s) = 1.4 x 10-7 exp[-1.1(eV)/kT]L(s) = 2.9 x 10-6 exp[-1.57(eV)/kT]

The point defect which led to the transient enhanced diffusivity was deduced to be avacancy. It was shown that final profile broadening due to the enhanced diffusion wassmaller at higher temperatures. Both Do and L increased with increasing implantationenergy.

270

B Bulk Diffusion B

M.Miyake, S.Aoyama: Journal of Applied Physics, 1988, 63[5], 1754-7[446-72/73-043]

Table 19Ambient Diffusion of B as a Function of B Dose, Si Dose, and Defect Concentration

B (/cm2) Si (/cm2) C (/cm3) D (cm2/s)6 x 1013 0 5 x 1018 4 x 10-13

6 x 1013 6 x 1013 1.8 x 1020 1 x 10-12

6 x 1013 1 x 1014 2.3 x 1020 4 x 10-12

6 x 1013 6 x 1014 - 4 x 10-12

6 x 1014 0 5 x 1019 2 x 10-12

6 x 1014 6 x 1014 - 3.5 x 10-12

3.7 x 1015 0 1 x 1020 4 x 10-12

3.7 x 1015 6 x 1013 2.5 x 1020 4 x 10-12

3.7 x 1015 1 x 1014 - 4 x 10-12

3.7 x 1015 6 x 1014 - 4 x 10-12

318 Bulk Diffusion - Quantitative DataThe co-evaporation of B2O3 during Si molecular beam epitaxial growth at temperatures ofbetween 540 and 800C was used to prepare superlattice structures with B concentrationswhich ranged from 3 x 1018 to 3 x 1020/cm3. The superlattices were then annealed andstudied by using secondary ion mass spectrometry, electrochemical profiling and cross-sectional transmission electron microscopy. It was found that an appreciableredistribution of B, at high B concentrations, occurred before annealing when the growthtemperature was greater than 700C. The diffusivity of B in the layers was found todepend markedly upon the growth temperature (table 18).T.E.Jackman, D.C.Houghton, J.A.Jackman, M.W.Denhoff, S.Kechang, J.McCaffrey,A.Rockett: Journal of Applied Physics, 1989, 66[5], 1984-92

[446-74-048]

319/20 Bulk Diffusion - Quantitative DataAnomalously high levels of ion-implanted B diffusivity (table 19) during pulse annealingwere attributed to the participation of interstitial Si atoms in the redistribution ofmigrating B atoms between interstitial and vacancy diffusion mechanisms. At B+ doses ofmore than 1015/cm2, the high values of B diffusivity were caused by the effect ofincompatibility elastic stresses (table 20) that prevented the transition of dopant atomsinto lattice sites.V.F.Stelmakh, J.R.Suprun-Belevich, A.R.Chelyadinski: Physica Status Solidi A, 1989,112[1], 381-4

[446-74-049]

321 Bulk Diffusion - Quantitative DataThe depth profiles of B in heterojunctions of amorphous hydrogenated Si and SiC weremeasured by means of nuclear reaction analysis. It was found that the B concentration in

271

B Bulk Diffusion B

the Si layer depended markedly upon the substrate temperature. It was concluded that thediffusivity (table 21) of B during hydrogenated amorphous Si film growth was very high.F.Zhang, D.He, Z.Song, G.Chen: Physica Status Solidi A, 1990, 118[1], K17-20

[446-76/77-059]

322 Bulk Diffusion - Quantitative DataThe migration of implanted B was investigated at temperatures of between 800 and1000C (table 22) by using furnace and rapid thermal annealing. The transient enhanceddiffusion which was caused by implantation damage in the early phases of annealing wasanalyzed, and it was found that the data could be described by:

D(cm2/s) = 0.022 exp[-2.5(eV)/kT]S.Solmi, F.Baruffaldi, R.Canteri: Journal of Applied Physics, 1991, 69[4], 2135-42

[446-78/79-051]

Table 20Ambient Diffusion of B as a Function of B Dose, Ge Dose, and Incompatibility Stress

B (/cm2) Ge (/cm2) S (N/m2) D (cm2/s)6 x 1014 0 4 x 107 2.5 x 10-13

6 x 1014 5 x 1015 - 8 x 10-14

6 x 1014 1 x 1016 - 5 x 10-14

1.3 x 1015 0 9 x 107 4 x 10-13

1.3 x 1015 5 x 1015 6 x 107 2.5 x 10-13

1.3 x 1015 1 x 1016 5 x 107 1.3 x 10-13

2 x 1015 0 1.5 x 108 5 x 10-13

2 x 1015 5 x 1015 - 4 x 10-13

2 x 1015 1 x 1016 1 x 107 2.5 x 10-13

323 Bulk Diffusion - Quantitative DataSamples were implanted with B through a surface oxide, and then annealed. This led toan enhanced B diffusivity. This enhancement was suppressed during an initial incubationperiod. An activation energy of 2eV was associated with the enhanced diffusion (table23), and this suggested that excess Si interstitials were involved. However, the processwhich led to the onset of enhanced diffusion was associated with an apparent activationenergy of 3.7eV. The use of 2-step annealing reduced the latter value to 2.6eV. This wasthe activation energy for the diffusion of interstitial O. Transmission electron microscopyrevealed that the coalescence of dislocations, as well as the growth of faulted loops,occurred rapidly after the incubation period. Precipitates which were situated along smalldislocation loops were observed after the incubation period. It was suggested that Oprecipitation, with emission of Si interstitials, predominated upon annealing beyond theincubation period and was responsible for the enhanced B diffusion. It was proposed thatthe enhanced diffusion was initially incubated by the trapping of O at dislocations. The

272

B Bulk Diffusion B

true onset of enhanced diffusion occurred when the dislocations were saturated and theoxide precipitation began at dislocations.D.Fan, R.J.Jaccodine: Journal of Applied Physics, 1990, 67[10], 6135-40

[446-78/79-052]

Bulk Diffusion - Quantitative DataThe migration of B was studied in material which had been pre-damaged with Si+ dosesof up to the threshold value for amorphization. In the case of material which had beenpre-implanted under amorphization conditions (40keV Si+, 5 x 1014/cm2), a B profile kinkappeared to arise from retarded diffusion after annealing at 800C. In samples which hadbeen implanted to 3 x 1014/cm2, an enhanced diffusion tail appeared in addition to thekink. The tail was associated with a diffusivity of 10-14cm2/s. Amorphous islands whichwere about 10nm in diameter were observed in samples which had been pre-damagedwith 40keV Si+ to a dose of 3.3 x 1014/cm2. It was suggested that the kink and tailoriginated in the vacancy and interstitial-rich regions, respectively.M.Kase, Y.Kikuchi, M.Kimura, H.Mori, T.Ogawa: Applied Physics Letters, 1991,59[11], 1335-7

[446-84/85-063]

Table 21Diffusivity of B in Si/SiC Heterojunctions

T (C) D (cm2/s)300 2.0 x 10-16

250 1.0 x 10-16

200 4.3 x 10-17

150 4.2 x 10-18

Bulk Diffusion - Quantitative DataLaser pulses were directed at n- epitaxial layers (0.006mm thick) on a substrate of p+ B-doped Si. It was found that local melting and re-freezing of the layer, and of a smallvolume of the substrate, occurred. In the melt, the occurrence of B diffusion from thesubstrate gave an almost uniform dopant concentration of 5 x 1018/cm3. On the basis ofthe results, a limited number of data (1.0 x 10-10 at 1420C and 1.1 x 10-15 at 900C) werededuced for B diffusion in the solid.K.M.Kim, S.N.Mei, M.J.Saccamango, S.F.Chu, R.J.Von Gutfeld, D.R.Vigliotti: AppliedPhysics Letters, 1992, 61[9], 1066-8

[446-93/94-048]

Bulk Diffusion - Quantitative DataThe migration of B in heavily-doped material was studied by using directly bondedsamples. In the case of directly bonded samples of heavily doped and lightly doped Si, apronounced tail was observed in the diffusion profile. This was attributed to the presence

273

B Bulk Diffusion B

of mobile interstitial B plus Si interstitial complexes which originated from heavily dopedmaterial. The presence of interstitial-type defects was confirmed by stacking faultshrinkage experiments. The diffusivity of [Bi-Sii] complexes was estimated to be one ortwo orders of magnitude higher than the intrinsic B diffusivity. In material which washeavily doped with B, the formation of [Bi-Sii] complexes could occur via interactionsbetween positively charged Si interstitials and ionized B during solidification. On thebasis of stacking fault shrinkage experiments, the activation energies of Si interstitialswere estimated to be between 2.11 and 2.5eV for P+ material, and between 3.45 and4.11eV for float-zone material.W.Wijaranakula: Journal of the Electrochemical Society, 1991, 138[4], 1131-7

[446-84/85-065]

Table 22Diffusivity of B in Si

T (C) D (cm2/s)1000 3.3 x 10-12

900 2.4 x 10-13

850 1.4 x 10-13

800 5.0 x 10-14

Bulk Diffusion - Quantitative DataThe diffusion of B under pure N, pure NH3, or their mixtures, was investigated in order todetermine the effect of the oxynitridation reaction upon diffusivity. An oxynitridation-enhanced diffusivity was explained in terms of a dual mechanism which involved bothvacancy and interstitial Si atoms. With a thin SiO2 layer on the Si wafer, and with a lowB concentration, the diffusion data could be described by:

D(cm2/s) = 0.105 exp [-3.22(eV)/kT] + (1.0 x 10

-6)exp[-1.71(eV)/kT]p

where p was the partial pressure of NH3.N.K.Chen, C.Lee: Journal of the Electrochemical Society, 1993, 140[8], 2390-4

[446-106/107-133]

324 Bulk Diffusion - Quantitative DataThe diffusion of B into single crystals, from a BF2-implanted polycrystalline filmdeposited on the surface, was accurately modelled. The effective diffusivities of B in themonocrystalline substrate were extracted by using Boltzmann-Matano analyses, and thephenomenological model for B diffusivity was implemented in a computer simulationprogram. The model was applied to a range of furnace annealing conditions (800 to 950C,0.5 to 6h) and BF2 doses ranging from 5 x 10

15 to 2 x 10

16/cm

2 at 70keV. It was found that

the diffusion parameters exhibited a complicated dependence upon the B composition(table 24).A.Sultan, M.Lobo, S.Bhattacharya, S.Banerjee, S.Batra, M.Manning, C.Dennison:Journal of Electronic Materials, 1993, 22[9], 1129-36

[446-109/110-044]

274

B Bulk Diffusion B

Table 23Enhanced Diffusivity of B in Si

T (C) D (cm2/s)1100 1.9 x 10-12

1100 1.4 x 10-12

1005 6.5 x 10-13

1005 4.8 x 10-13

980 4.7 x 10-13

955 2.8 x 10-13

885 7.9 x 10-14

885 7.0 x 10-14

Bulk Diffusion - Quantitative DataAnomalous enhanced tail diffusion of B in directly bonded samples of heavily- or lightly-doped material was observed after annealing at temperatures ranging from 1000 to 1200C.As in the case of the enhanced diffusion which was commonly observed in B-implantedmaterial during damage annealing, the enhanced diffusion in directly bonded samples wassuggested to be due to mobile interstitial B species. In heavily-doped Czochralski-typematerial, interstitial B species could be produced via elastic interactions between excessSi interstitials and substitutional B atoms during solidification. The enhanced diffusivity,as deduced by comparing the present results with those which had been obtained for B-implanted specimens, could be described by the Arrhenius expression:

D(cm2/s) = 0.0003 exp[-2.1(eV)/kT]

W.Wijaranakula: Japanese Journal of Applied Physics 1, 1993, 32[9A], 3872-8[446-113/114-042]

Bulk Diffusion - Quantitative DataA survey was made of isothermal diffusivity values for B in Si. It was noted that thestandard error in the evaluation of the logarithm of the pre-exponential was directlyproportional to the standard error in the evaluation of the activation energy. The constantin this proportionality was 8.3. Also, a statistical relationship existed between the aboveparameters, for B diffusing in Si, and the fact that this relationship was characterized by acorrelation factor of 0.99 supported its linearity. It was further noted that, by expressingthe pre-exponential factor in terms of the activation energy, it was possible to exclude itfrom the equation for the temperature dependence of the diffusion coefficient. Thisrevealed the existence of an unambiguous relationship between the isothermal diffusioncoefficient and the activation energy for B diffusion in Si. Experimental data for the lattersystem were shown to be satisfactorily described by the single equation,

ln[D] = -28 - [1/kT - 8.17]QD.Rakhimbaev, A.Avezmuradov, M.D.Rakhimbaeva: Neorganicheskie Materialy, 1994,30[4], 442-8 (Inorganic Materials, 1994, 30[4], 418-23)

[446-115/116-150]

275

B Bulk Diffusion B

Table 24Diffusion Parameters for B in Si, as a Function of the B Composition

Concentration Range (/cm3) Do(cm2/s) Q(eV)less than 5.0 x 1018 0.0012848 2.6950

5 x 1018

to 1.4 x 1019 0.0001393(10[1.9280E-19]C) 2.5222(10[5.7581E-21]C)1.4 x 10

19 to 3.0 x 1019 4.8828 x 10-373(C 19.376) 2.4735 x 10-10(C 0.5268)

3.0 x 1019

to 4.0 x 1019 1.2226 x 10-90(C 4.8746) 0.0069046(C 0.1444)4.0 x 10

19 to 1.0 x 1020 8.0643 x 10295(C -14.8190) 1.3329 x 107(C -0.3295)

greater than 1.0 x 1020 0.3362 3.4260

Bulk Diffusion - Quantitative DataThe damage which was produced by the direct-current plasma doping of B was studied.The damage density was measured by using 1.5MeV He+ Rutherford back-scatteringspectrometry and photo-acoustic displacement techniques. These results showed that lessdamage due to doping could be achieved by plasma doping as compared with thatproduced by ion implantation. Impurity diffusion also did not occur as rapidly here, as inthe case of B+ and BF2

+ implantation. The B diffusion coefficient could be as low as 4.9 x10-15cm2/s at 950C after plasma doping. This value was 5 and 40 times lower than in thecases of B+ and BF2

+ implantation, respectively. It was found that there was a correlationbetween the diffusion coefficient and the damage density.T.Hara, K.Shinada, S.Nakamura: Japanese Journal of Applied Physics 1, 1994, 33[10],5608-11

[446-117/118-194]

Bulk Diffusion - Quantitative DataThe effects of heavy Sb ion implantation and annealing (at 800 to 1000C) upon B delta-doped superlattices in Si layers that had been grown by molecular beam epitaxy wereanalyzed. Secondary ion mass spectroscopic measurements of these structures were usedto investigate the generation and diffusion of point defects. An enhanced diffusion of B,from delta-doped spikes in as-grown and Sb-implanted layers, was described theoreticallyby solving the diffusion equation; using a point defect model for various initial pointdefect distributions. In order to fit the experimental secondary ion mass spectrometryprofiles, the diffusion coefficient of positively charged B interstitials had to be changedfrom a default value of 0.68, to 0.45cm2/s. It was found that the molecular beam epitaxialgrowth process produced interstitials and vacancies with an almost constant average valueof about 5 x 1016/cm3. The Sb-implanted B modulation-doped superlattice permitted thedepth profile of the defect concentration to be deduced. The main features of the Bdiffusion profile in the superlattice could be fitted by assuming a constant overlap of 5 x1016/cm3 for interstitials and vacancies that were caused by the molecular beam epitaxialgrowth. The distribution which was deduced from damage calculations consisted of a flat

276

B Bulk Diffusion B

high-concentration region with an exponential decrease towards the level of the molecularbeam epitaxial layer.D.Krüger, U.Jagdhold, R.Kurps, H.P.Zeindl: Journal of Applied Physics, 1995, 78[8],5008-12

[446-125/126-146]

Table 25Diffusivity of B in Poly-Si Layers

Layer T (C) D (cm2/s)first 1150 2.7 x 10-11

first 1100 1.3 x 10-11

first 1050 6.3 x 10-12

first 1000 2.5 x 10-12

second 1150 7.2 x 10-12

second 1100 4.8 x 10-12

second 1050 2.6 x 10-12

second 1000 1.2 x 10-12

Bulk Diffusion - Quantitative DataSecondary ion mass spectroscopic measurements of B δ-doped superlattices were used toinvestigate the generation and diffusion of point defects. The effect of heavy Sb ionimplantation and annealing (800 to 1000C) was analyzed. The enhanced diffusion of Bfrom doping spikes in as grown and Sb-implanted layers was analyzed by solving thediffusion equation; using a computer model for various initial point defect distributions.In order to fit the secondary ion mass spectrometry profiles, the positively charged Binterstitial diffusion coefficient was changed, from a default value of 0.68, to 0.45cm2/s. Itwas found that the molecular beam epitaxial growth process produced interstitials andvacancies with an almost constant average value of about 5 x 1016/cm3. The use of an Sb-implanted B modulation-doped superlattice permitted the depth profile of the defectconcentration to be obtained.D.Krüger, R.Kurps, H.P.Zeindl, U.Jagdhold: Solid State Phenomena, 1996, 47-48, 313-8

[446-134/135-158]

Bulk Diffusion - Quantitative DataThe injection of interstitials during the annealing of non-amorphizing Si implants wasmonitored by using sharply defined B-doped marker layers that had been grown by meansof reduced-pressure chemical vapor deposition. The enhancement of B diffusivity whichwas measured during the initial annealing stage (700C, < 15s,) was at least an order ofmagnitude greater than the enhancement which occurred during subsequent annealing.The high supersaturation of interstitials during the early stages of annealing led to theimmobilization of B down to concentrations of about 1017/cm3. This was consistent with

277

B Bulk Diffusion B

interstitial-driven B clustering. It was concluded that the ultra-fast diffusion set a lowerlimit, on the B interstitial diffusivities (700C), of 2 x 10-13cm2/s.H.G.A.Huizing, C.C.G.Visser, N.E.B.Cowern, P.A.Stolk, R.C.M.De Kruif: AppliedPhysics Letters, 1996, 69[9], 1211-3

[446-138/139-104]

325 Bulk Diffusion - Quantitative DataThe concentration profiles of B which had been diffused, from polycrystalline material,into underlying monocrystalline material were analyzed by means of secondary ion massspectrometry. The co-diffusion of As and B was studied in an emitter and extrinsic baseconfiguration. The process was investigated by carrying out simulations. These indicatedthat diffusion of the dopant at the lowest fluence was slowed much more by the in-depthinhomogeneous grain growth that was induced by amorphization and annealing, than by abuilt-in electric field. It was assumed that the dopant, at the highest doses, saturated thegrain boundary traps. This was true of B. In a first poly-Si layer, the diffusivity (table 25)could be described by:

D (cm2/s) = 0.019 exp[-2.5(eV)/kT]In a second poly-Si layer, the diffusivity could be described by:

D (cm2/s) = 0.000032 exp[-1.86(eV)/kT]A.Merabet, C.Gontrand: Physica Status Solidi A, 1994, 145[1], 77-88

[446-117/118-193]

Bulk Diffusion - Qualitative Observations - Concentration DependenceIt was pointed out that, over quite a wide range of concentrations, the diffusioncoefficient of B increased linearly with concentration. However, at very highconcentrations, precipitation or clustering led to a plateau or to a decrease in thediffusivity.J.R.King: IMA Journal of Applied Mathematics, 1988, 40[3], 163-81

[446-78/79-048]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesA technique was proposed for the simultaneous diffusion of B and P on opposite sides ofa wafer. The solid dopant sources which were used in this method could be tailored so asto produce a wide range of B diffusion profiles, for a given thermal diffusion cycle. Theuniformity of the sheet resistance across a 49cm2 area was often greater than 95%. Aunique feature was that the resultant diffusion glass was extremely thin.T.Krygowski, A.Rohatgi: Journal of the Electrochemical Society, 1997, 144[1], 346-52

[446-152-0277]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesA secondary ion mass spectrometric analysis of diffusion across a TaSi2/Si interface andinto monocrystalline or polycrystalline Si was carried out, together with an electricalcharacterization of the resultant structures. At temperatures ranging from 900 to 1000C,the dopant readily diffused into Si, without drastic segregation effects, when appropriateinterface cleaning was used. In particular, very shallow diffusion regions were obtained inmonocrystalline material beneath the implanted TaSi2; even at the relatively longannealing times sometimes needed for processing.

278

B Bulk Diffusion B

H.Gierisch, F.Neppl, E.Frenzel, P.Eichinger, K.Hieber: Journal of Vacuum Science andTechnology B, 1987, 5[2], 508-14

[446-55/56-034]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe diffusion of this element from polycrystalline material and into monocrystallinematerial during rapid optical annealing was investigated. The samples were characterizedby using secondary ion mass spectrometry, transmission electron microscopy, and sheetresistance measurements. It was demonstrated that very shallow diffusion profiles inmonocrystalline material could be formed by rapid optical annealing.H.J.Böhm, H.Wendt, H.Oppolzer, K.Masseli, R.Kassing: Journal of Applied Physics,1987, 62[7], 2784-8

[446-55/56-034]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe use of TiSi2 as a dopant diffusion source for B was studied. The TiSi2 layers weredoped via ion implantation. Diffusion was carried out by using furnace and rapid thermalprocessing. Secondary ion mass spectrometry, scanning electron microscopy, and X-raydiffraction furnished clear evidence for compound formation between Ti and the twodopant species. This led to low dopant concentrations at the silicide/Si interface and to avery poor efficiency of the diffusion source.V.Probst, H.Schaber, P.Lippens, L.Van den Hove, R.De Keersmaecker: Applied PhysicsLetters, 1988, 52[21], 1803-5

[446-62/63-226]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesComparative studies of As and B diffusion in polysilicon-on-monocrystal systems wereperformed by means of cross-sectional transmission electron microscopy and secondaryion mass spectrometry. The As and BF2 which were implanted into 300nm polysilicon,deposited via low-pressure chemical vapor deposition, were diffused into the underlyingSi substrate by using rapid thermal annealing or furnace annealing. The B interdiffusiongave a discontinuous dopant profile at the interface, due to the accumulation of B-defectcomplexes that were created by the dissolution of defect clusters in polysilicon. At lowannealing temperatures, the B profiles in monocrystalline Si were shallower than the Asprofiles. This was because most of the B in polysilicon films was immobile duringannealing near to the peak region, and because of low B segregation at grain boundaries.This then reduced the effect of grain boundaries with regard to high diffusivities. Duringhigh thermal budget annealing, B diffusion into the substrate was greater than Asdiffusion because of higher B diffusivities in monocrystalline Si. At high annealingtemperatures, the native oxide at the interface broke up and caused the polysilicon layerto align epitaxially with the underlying substrate. The elapsed time up to the break-up ofthe interfacial oxide depended upon the dopant species and the annealing temperature.The oxide break-up took longer in As-doped samples than in B-doped samples.K.Park, S.Batra, S.Banerjee, G.Lux, R.Manukonda: Journal of the ElectrochemicalSociety, 1991, 138[2], 545-9

[446-81/82-044]

279

B Bulk Diffusion B

Bulk Diffusion - Qualitative Observations - Concentration ProfilesDopant redistribution, during rapid thermal annealing, from implantation-dopedpolycrystalline films and into monocrystals was studied. Diffusivity and co-diffusivitydata were deduced from secondary ion mass spectrometry results.C.Gontrand, C.Dubois, A.Laugier: Journal of Physics - Condensed Matter, 1991, 3[18],3091-8

[446-84/85-059]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesDepth profiles were determined by using spreading resistance, Auger electronmicroscopy, and secondary ion mass spectrometry methods. These profiles showed thatenhanced dopant diffusion and surface depletion resulted from rapid thermal treatments.It was found that Si interstitial injection from a N-supersaturated oxynitride interfacefacilitated the diffusion of B in the Si substrate. An appreciable amount of N was foundbelow the Si surface, and this suggested that N interstitials might play an important role inthe observed enhancement of the impurity diffusion.J.Bustillo, C.Chang, S.Haddad, A.Wang: Applied Physics Letters, 1991, 58[17], 1872-4

[446-86/87-049]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesMigration in amorphous and polycrystalline Si-on-monocrystal systems during rapidthermal annealing and furnace annealing was studied. It was found that changes inmicrostructure during annealing played a major role in determining diffusion profiles inthe substrate, as well as in the polycrystalline Si layer. In the case of B doping, there waslittle difference between the final microstructures and junction depths for as-depositedamorphous and as-deposited polycrystalline Si. At high annealing temperatures, the nativeinterfacial oxide disintegrated, caused epitaxial realignment of the polycrystalline Si film,and led to enhanced diffusion in the substrate.K.Park, S.Batra, S.Banerjee, G.Lux, T.C.Smith: Journal of Applied Physics, 1991, 70[3],1397-404

[446-91/92-028]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe B ions were implanted into thin layers of WSi2 or CoSi2, on polycrystalline ormonocrystalline Si, and were out-diffused into the substrate during furnace heating orrapid thermal annealing. Shallow (less than 100nm) junctions were obtained by usingrapid thermal annealing. The interface concentrations were close to the solid solubilitiesof the respective dopant; that is, more than 8 x 1019/cm3. During furnace annealing attemperatures above 800C, the migration of B from CoSi2 led to a decrease, in theinterface concentration, to less than 2 x 1019/cm3. This was attributed to the effect ofmarked B segregation, and probably to reactive losses at the SiO2/CoSi2 interface. It wasfound that the dopant redistribution was determined by the lattice and grain-boundarydiffusivities, the solubility limits, layer inhomogeneities, dopant segregation at interfacesand grain boundaries. In addition, there was probably a phase transformation of thedopants which were segregated at the SiO2/silicide interface.

280

B Bulk Diffusion B

V.Probst, H.Schaber, A.Mitwalsky, H.Kabza, L.Van den Hove, K.Maex. Journal ofApplied Physics, 1991, 70[2], 708-19

[446-93/94-045]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesAn experimental and theoretical study was made of the physical mechanisms which wereinvolved in the co-diffusion of As and B in polycrystalline/monocrystalline bilayersduring the formation of shallow N+ emitter. The redistribution, of successively implantedAs and B, which was induced by rapid thermal annealing was studied by means ofsecondary ion mass spectroscopic measurements. A strong retardation of diffusion wasobserved for B, and this was attributed essentially to grain growth in the polycrystallinelayer.C.Gontrand, A.Merabet, B.Semmache, S.Krieger-Kaddour, C.Bergaud, M.Lemiti,D.Barbier, A.Laugier: Semiconductor Science and Technology, 1993, 8[2], 155-62

[446-099/100-095]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe co-diffusion of As and B in polycrystalline and monocrystalline bi-layers, during theformation of shallow N+ emitters, was studied. It was found that rapid thermal annealingled to the redistribution, as measured by secondary ion mass spectrometry, of As and Bwhich were successively implanted into a 380nm low-pressure chemical vapor-depositedpolycrystalline layer.C.Gontrand, P.Sellitto, S.Tabikh, S.Latreche, A.Kaminski: Journal de Physique III, 1997,7[1], 47-58

[446-150/151-151]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesA modified neutron activation technique was described which was able to measure theprofiles of isotopic impurities with a short half-life. A deep acceptor impurity profile wasobtained by applying the method to Si diffused with B. It was found that concentrationsas low as 1013/cm3 could be measured with an accuracy of ±10%.R.M.Huang, R.S.Huang: Journal of the Electrochemical Society, 1986, 133[12], 2605-8

[446-51/52-134]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesSecondary ion mass spectrometry depth profiles were used to compare B diffusion fromvarious sources at temperatures ranging from 850 to 1050C. The sources included in situdoped and ion-implanted polycrystalline Si and a vapor source involving doped powder.The surface concentrations exhibited little source dependence. A Boltzmann-Matanoanalysis was used to show that the concentration dependence of the diffusivity upon thesource was very small. It was demonstrated that conventional models for B diffusioncould not explain the experimental data or the Boltzmann-Matano results; regardless ofthe source. A new model described the B diffusion profiles more accurately.W.A.Orr Arienzo, R.Glang, R.F.Lever, R.K.Lewis, F.F.Morehead: Journal of AppliedPhysics, 1988, 63[1], 116-20

[446-61-085]

281

B Bulk Diffusion B

Bulk Diffusion - Qualitative Observations - Concentration ProfilesAn experimental technique for the measurement of two-dimensional impurity diffusionprofiles was developed. Both the lateral and in-depth extents of dopant diffusion under amask edge were exaggerated by angle lapping, and the magnified junction contour wasdelineated by chemical staining. The two-dimensional shape of the junction wasreconstructed from the stained contour. A complete diffusion profile, consisting of severaliso-concentration contours, could be obtained by measuring the junction shape on a seriesof samples with increasing substrate resistivities; provided that the doping level in thesubstrate did not affect the diffusion of the impurity which was being studied. Data on thetwo-dimensional diffusion of B in Si at 1050C were presented.R.Subrahmanyan, H.Z.Massoud, R.B.Fair: Applied Physics Letters, 1988, 52[25], 2145-7

[446-62/63-228]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe migration of B in ion-implanted and annealed monocrystalline or amorphous materialwas compared in order to determine the effect of amorphization upon the initial transientB motion in single crystals. The B was implanted at 20keV to doses of 1015 or 3 x1015/cm2. The material was either pre-amorphized or post-amorphized, to a depth of320nm, by implanting Si ions at 3 different energies. The B profile always lay within theamorphized depth. The samples were furnace-annealed or rapidly thermally annealed at900 to 1100C, and were sometimes pre-annealed at 600C. An initial rapid diffusiontransient in the tail region of the B profile was observed in crystalline samples but wasabsent from amorphous samples. When significant motion occurred, the profiles of theamorphized samples could be fitted by a model which did not take account of transienteffects. It was suggested that excess interstitials caused transient diffusion in crystallinesamples. The source of the interstitials was attributed to the thermal dissolution of smallclusters which were formed by the implantation process. Since there was no enhanceddiffusion in the amorphous regions, it was proposed that either interstitial clusters werenot produced or that they did not survive the re-growth process in that region. It wasfurther concluded that interstitials which were generated by damage beyond theamorphous/crystalline boundary were prevented from entering the re-grown region bydislocation loops which formed at the boundary. These acted as sinks which interceptedinterstitials that were diffusing towards the surface.T.O.Sedgwick, A.E.Michel, V.R.Deline, S.A.Cohen, J.B.Lasky: Journal of AppliedPhysics, 1988, 63[5], 1452-63

[446-72/73-043]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesTwo-step rapid thermal diffusion of B, using a BN solid diffusion source, was described.During the first step, HBO2 glass was deposited onto the Si wafer from the diffusionsource by keeping the wafer temperature at 750C and the diffusion source temperature atabout 900C. During the second step, the B was diffused into the wafer (from the HBO2

glass) by annealing at 1000C or 1100C in N2. Extremely shallow junctions, with depths ofabout 20nm and sheet resistances of about 350O/square, could be achieved as well asrelatively deep junctions with depths of about 175nm and sheet resistances of about55O/square. If diffusion annealing was performed at 1100C, the junction depth and the

282

B Bulk Diffusion B

electrically active B concentration at the surface increased when the ambient gas waschanged from N2 and O2. The sheet resistance also decreased. No high-resistivity B-richlayer was formed at the surface when diffusion annealing was performed at 1100C in anO2 ambient.J.G.Kim, C.K.Kim: Journal of Electronic Materials, 1989, 18[5], 573-8

[446-72/73-044]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesA comparison was made of the B profiles which were determined by using spreadingresistance and secondary ion mass spectrometry techniques. It was established that thespreading resistance method could be used for the semi-quantitative measurement ofconcentrations under transistor preparation conditions. It was concluded that secondaryion mass spectrometry gave better reproducibility than did the spreading resistancetechnique. That is, the spreading resistance method was not an absolute one, whereas thesecondary ion mass spectrometry method was reproducible over differing instruments andperiods of at least 10 years. On the other hand, the spreading resistance method wassensitive right up to junctions whereas the secondary ion mass spectrometry techniquewas not.R.A.Clapper, D.G.Schimmel, J.C.C.Tsai, F.S.Jabara, F.A.Stevie, P.M.Kahora: Journal ofthe Electrochemical Society, 1990, 137[6], 1877-83

[446-76/77-034]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesEpilayers of B-doped material were grown by using molecular beam epitaxy with anelemental B source (up to 2 x 1020/cm3) in order to elucidate profile control at growthtemperatures which ranged from 450 to 900C. Precipitation and surface segregationeffects were observed at dopant levels of 2 x 1020/cm3 and growth temperatures of morethan 600C. At growth temperatures below 600C, excellent profile control was achieved(with complete electrical activation) at a concentration of 2 x 1020/cm3. These valuescorresponded to the optimum molecular beam epitaxial growth conditions for a range ofSi/SiGe heterostructures.C.P.Parry, S.M.Newstead, R.D.Barlow, P.Augustus, R.A.A.Kubiak, M.G.Dowsett,T.E.Whall, E.H.C.Parker: Applied Physics Letters, 1991, 58[5], 481-3

[446-81/82-047]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe interdependence of the diffusion behavior and grain microstructure in amorphouspolysilicon - on - monocrystal systems was studied with regard to the rapid thermal andfurnace annealing of BF2 implants. It was found that changes in microstructure duringannealing played a major role in determining the diffusion profiles in the substrate as wellas in the polysilicon layer. This led to the formation of shallower junctions in thesubstrate for the first case. There was little difference in the final microstructure andjunction depth between the 2 cases. The B junctions which formed in the substrate werefound to be very uniform laterally, in spite of expected doping inhomogeneities that werecaused by polysilicon grain boundaries, for both as-deposited amorphous Si diffusionsources and for as-deposited polysilicon diffusion sources.

283

B Bulk Diffusion B

K.Park, S.Batra, S.Banerjee, G.Lux: Journal of Electronic Materials, 1991, 20[3], 261-5[446-81/82-047]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe in-diffusion of B from borosilicate glass, using rapid thermal processing, was studied.It was found that the sheet resistance of diffused layers decreased more steeply than t-½ asa function of time, after short diffusion times and at low temperatures. This correspondedto a rapid increase in the surface B concentration in the initial stages of diffusion. Thiswas attributed to the presence of thin native oxides between the borosilicate glass and theSi. The time that was required for the surface B concentration to saturate was shorter forborosilicate glass with higher B concentrations. The diffusion of B was enhanced whenthe B concentration in the borosilicate glass was greater than 18at%. An enhanceddiffusivity was also observed when furnace annealing was used and the B concentrationin the glass was greater than 18at%. It did not occur when the B concentration was 7at%.Enhanced diffusion was not observed during rapid thermal annealing if the glass was notused.M.Miyake: Journal of the Electrochemical Society, 1991, 138[10], 3031-9

[446-84/85-062]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThree-dimensional B impurity profiles in substrates were measured, with a resolution of10nm, via the scanning tunnelling microscopy of cleaved surfaces which had beenprogressively etched with HF-HNO3-H2O mixtures. The optimum composition had avolume ratio of 1:100:25, and could be applied to dopant concentrations of between 1016

and 1020/cm2. The etching rate exhibited no dependence upon the crystallographicorientation. It was noted that this technique revealed the active B impurity content ratherthan the implanted B concentration.T.Takigami, M.Tanimoto: Applied Physics Letters, 1991, 58[20], 2288-90

[446-84/85-063]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesAnomalous threshold voltage shifts in various devices under high-temperature rapidthermal annealing and glass reflow conditions were studied. Transient enhanced B out-diffusion and P pile-ups, as well as interface states which were generated by thedegradation of thin gate oxides during rapid thermal annealing, were suggested to be themain source of this anomalous shift.Y.K.Fang, J.C.Hsieh, C.W.Chen, C.H.Koung, N.S.Tsai, J.Y.Lee, F.C.Tseng: AppliedPhysics Letters, 1992, 61[4], 447-9

[446-93/94-048]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe diffusion of δ-function spikes of B in thin films which had been grown by solid-phase epitaxy was studied during annealing in vacuum. The results were compared withthe effects of diffusion in films that had been grown by using low-temperature molecularbeam epitaxy. The diffusion temperatures ranged from 750 to 900C, and the 2-dimensional concentrations were between 7 x 10

13 and 1.6 x 10

14/cm

2. The diffusive

behavior of dopants in the solid-phase epitaxial films was qualitatively different to that in

284

B Bulk Diffusion B

films grown by low-temperature molecular beam epitaxy. This difference was attributedto the vacancy-like defects that were intrinsic to growth by the former method but not togrowth by the latter method. The dopant profiles widened appreciably during solid-phaseepitaxial re-growth; thus making the achievement of δ-function spikes impossible. Aftervacuum annealing, the diffusion coefficients for both n-type and p-type dopants werelower in solid-phase epitaxial films, than in films which had been grown by using theother method, by up to one order of magnitude. The diffused depth profile of the dopantin low-temperature molecular beam epitaxial films exhibited the characteristic deviationfrom a pure Gaussian that was expected, due to the concentration dependence ofdiffusion. That is, it had a flat top and steep shoulders. On the other hand, the dopantdepth profiles in the other material exhibited a central spike and relatively flat shouldersafter diffusion. The width of the central spike was, following an initial transient that wasimpossible to resolve, independent of the diffusion time and temperature. This indicatedthat the solid-phase epitaxial material was defective; with the defects acting as trapsduring diffusion.H.J.Gossmann, A.M.Vredenberg, C.S.Rafferty, H.S.Luftman, F.C.Unterwald,D.C.Jacobson, T.Boone, J.M.Poate: Journal of Applied Physics, 1993, 74[5], 3150-5

[446-106/107-135]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe distribution of B atoms in samples which had previously been implanted with F

+ and

Ne+ to doses of 1.2 x 1013

or 1.5 x 1015

/cm2 was investigated, before and after annealing,

by means of secondary ion mass spectrometry. After annealing at 900C, a strongdependence of the B redistribution rate upon F concentration was observed. It was foundthat the presence of F decreased B redistribution, while that of Ne (which was close to Fin the periodic table, but which was inert) did not. When post-implantation annealing wascarried out at 1000C, the B distribution did not depend upon the type of previouslyimplanted ion. At this temperature, F atoms left the crystal. The results showed that theredistribution of B which was implanted into Si depended upon the chemical activity ofthe impurity which was present in a crystal, and upon its concentration.L.J.Krasnobaev, N.M.Omelyanovskaya, V.V.Makarov: Journal of Applied Physics, 1993,74[10], 6020-2

[446-109/110-044]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe behavior of B and F which was introduced into an oxide/Si system was investigated.It was found that the introduction of F increased the sheet resistance of B layers in Siafter heat-treatment in both N and O atmospheres. The B depth profiles revealed that Fcaused B to migrate into the oxide from the Si. This behavior was explained by assumingthat F, incorporated into the oxide network together with B, lowered the viscosity of theoxide.Y.Sato, I.Kawashima: Journal of the Electrochemical Society, 1994, 141[5], 1381-6

[446-119/120-289]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe effect of lattice defects, produced by Si ion implantation, upon dopant diffusivity wasinvestigated after annealing specimens at temperatures of between 700 and 900C. The

285

B Bulk Diffusion B

nature and depth of residual implantation defects in undoped samples was determined byanalysing the rocking curves which were obtained by using triple-crystal X-ray diffractionand transmission electron microscopy. As well as interstitial dislocation loops andclusters lying below the original amorphous/crystal interface, epitaxial re-growth of theamorphized Si left a vacancy-rich surface layer and a deeper region which was enrichedin interstitials. These regions corresponded to those for which Monte Carlo simulations ofdefect production predicted an excess of point defects. According to whether the dopantwas associated with vacancy or interstitial clusters, different anomalous diffusionbehaviors were observed. In the deep region where an excess of interstitials was present,B underwent markedly enhanced diffusion. On the other hand, retarded diffusion of Bwas observed in the surface layer. This was consistent with the differing contributionswhich vacancies and interstitials made to the diffusion mechanisms of the dopants.M.Servidori, R.Angelucci, F.Cembali, P.Negrini, S.Solmi, P.Zaumseil, U.Winter: Journalof Applied Physics, 1987, 61[5], 1834-40

[446-60-013]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe behavior of the dopant during infra-red heating was investigated by using spin-ondiffusion sources, repeated etching, and capacitance-voltage measurements. The diffusioncoefficient was calculated by fitting the concentration profiles to the complementary errorfunction. It was found that the diffusivity was enhanced during infra-red heating. This didnot appear to be due to ion-implantation damage or rapid heating. The enhanceddiffusivity was attributed to the generation of an excess of self-interstitials.Y.Ishikawa, K.Yamauchi, I.Nakamichi: Japanese Journal of Applied Physics, 1989,28[8], L1319-21

[446-72/73-041]

Bulk Diffusion - Qualitative Observations - Defect InteractionsIon implantation produces point defects, during annealing, which can significantlyenhance dopant diffusion. This effect was studied for implant diffusion at lowtemperatures. Enhanced diffusion was detected below critical dopant concentrations. Thelatter concentrations depended only upon the temperature.R.B.Fair: Journal of the Electrochemical Society, 1990, 137[2], 667-71

[446-74-046]

Bulk Diffusion - Qualitative Observations - Defect InteractionsA study was made of the oxidation-retarded diffusion of B into bare samples duringO2/NF3 oxidation at 1100C for times of up to 36h. The results confirmed the large effectwhich small (ppm) additions of F had upon the point defect equilibrium at the growingSi/SiO2 interface. They also supported conclusions which were drawn from previousstudies of the shrinkage of oxidation-induced stacking faults during the same oxidationprocess.U.S.Kim, T.Kook, R.J.Jaccodine: Journal of the Electrochemical Society, 1988, 135[1],270-1

[446-61-085]

286

B Bulk Diffusion B

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe dependence of anomalous B diffusion upon the defect depth position was studiedafter furnace and electron beam annealing of samples which had been damaged by 28Siions implanted at various energies. This behavior was correlated with the excessvacancies and interstitials which were produced by bombardment of the surface regionand the bulk, respectively. The spatial separation of these point defects was indicated byanalyses of intensity profiles obtained by using double-crystal X-ray diffraction.S.Solmi, R.Angelucci, F.Cembali, M.Servidori, M.Anderle: Applied Physics Letters,1987, 51[5], 331-3

[446-55/56-036]

Bulk Diffusion - Qualitative Observations - Defect InteractionsTantalum silicide, when deposited directly onto monocrystalline Si substrates andannealed at 950C, led to the enhanced diffusion of B in buried layers. This indicated asubstantial vacancy component in B diffusion; at least up to 950C. The fact that theenhanced diffusion occurred in buried layers excluded a snow-plough mechanism. TheSi/Ta ratio in the sputter-deposited silicide was slightly less than 2. It was suggested thatfurther silicidation generated vacancies by removing Si atoms from the Si substrate.Enhanced diffusion was not detected when there was a 150nm intermediate layer ofpolycrystalline Si film between the silicide and the monocrystalline Si substrate. Thisindicated that polycrystalline Si was an effective sink for excess vacancies; perhaps morethan for excess interstitials.S.M.Hu: Applied Physics Letters, 1987, 51[5], 308-10

[446-55/56-037]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe behavior of B which had been implanted into pre-amorphized material was studiedby using secondary ion mass spectroscopy and transmission electron microscopy. A pre-amorphized surface layer was created by double Si ion implantation, and the as-implantedB profiles were entirely confined to the pre-amorphized layer. The results showed that Bdiffusion during rapid thermal annealing was anomalous in nature, and that the magnitudeof the anomalous diffusion depended upon the rapid thermal annealing temperature.Whereas rapid thermal annealing at 1150C led to enhanced B diffusion, as compared tothat in monocrystalline samples, reduced diffusion was observed in pre-amorphizedsamples which were annealed at 1000C. The results were explained in terms ofdifferences in defect evolution during rapid thermal annealing.Y.M.Kim, G.Q.Lo, D.L.Kwong, H.H.Tseng, R.Hance: Applied Physics Letters, 1989,55[22], 2316-8

[446-72/73-044]

Bulk Diffusion - Qualitative Observations - Defect InteractionsStudies of Sn+ pre-amorphized p+n junctions revealed the presence of some 1017/cm3 ofdeep-level donors after amorphous layer re-growth at temperatures below 800C. Removalof the defects by annealing at 800C permitted the leakage current which was associatedwith these defects to be distinguished from that which was due to dislocation loops at theamorphous/crystalline boundary. In both cases, the defects and leakage currents were

287

B Bulk Diffusion B

similar to those which resulted from Si+ pre-amorphization. For a given Sn+ pre-amorphization treatment, the leakage current which was measured in substrates doped to4 x 1013/cm2 was about 1000 times lower than in substrates which had been doped to1016/cm2. When compared with Si+ pre-amorphization, the use of Sn resulted in a greaterenhancement of B diffusion within the re-grown amorphous layer itself.J.R.Ayres, S.D.Brotherton, J.B.Clegg, A.Gill, J.P.Gowers: Semiconductor Science andTechnology, 1989, 4[5], 339-407

[446-72/73-044]

Bulk Diffusion - Qualitative Observations - Defect InteractionsA B-doped surface layer, with a thickness of 0.0003mm, was irradiated with 40keV N+

ions, 135keV Ar+ ions, 135keV Ar+ ions, or 135keV Xe+ ions, at temperatures of between600 and 1100C. The B and carrier profiles exhibited radiation-enhanced dopantredistributions. The diffusion enhancement was more effective for lighter ions and forhigher irradiation temperatures. This behavior was attributed to a lower probability for thetrapping of mobile point defects at extended stable lattice defects. A comparison of thecarrier and dopant profiles revealed the formation of immobile electrically inactive Bspecies by interaction with lattice defects at temperatures which were not too high. Acharacteristic double-peak structure was observed at higher irradiation temperatures. Thiswas explained by the diffusion of B in the form of unstable B-defect pairs. The B whichcorresponded to these peaks was located at normal lattice sites and was electrically active.G.A.Kachurin, I.E.Tyshchenko, E.Wieser, C.Weise: Physica Status Solidi A, 1989,109[1], 141-8

[446-74-049]

Bulk Diffusion - Qualitative Observations - Defect InteractionsAtoms of B were incorporated into (100) wafers by heating the substrates (800C, 0.5h) ina B2H6/H2 atmosphere and by subsequent rapid thermal annealing at temperatures above900C. The atomic and carrier concentration profiles of B-doped layers were examined bymeans of secondary ion mass spectrometry and differential Hall measurements,respectively. The results clearly showed that ultra-shallow p+ layers, 30nm-thick and witha surface carrier concentration of 7.26 x 1019/cm3, could be formed by B diffusion at800C and by subsequent rapid thermal annealing at 100C.T.Inada, A.Kuranouchi, H.Hirano, T.Nakamura, Y.Kiyota, T.Onai: Applied PhysicsLetters, 1991, 58[16], 1748-50

[446-81/82-045]

Bulk Diffusion - Qualitative Observations - Defect InteractionsMany substitutional impurities in crystalline solids diffuse via a hybrid mechanism whichinvolves a fast-migrating intermediate species. Non-equilibrium measurements of themigration frequency and the migration length for the impurity could provide data onatomic-scale diffusion mechanisms. The method was applied to the case of B incrystalline Si. It was found that the B diffused mainly via a migrating interstitial species,Bi, which was generated by a kick-out reaction.N.E.B.Cowern, G.F.A.Van de Walle, D.J.Gravesteijn, C.J.Vriezema: Physical ReviewLetters, 1991, 67[2], 212-5

[446-84/85-063]

288

B Bulk Diffusion B

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe annealing of B through-oxide implanted material can lead to incubated Oprecipitation-enhanced B diffusion. It was shown here that annealing of through-oxideimplants in an NF3-containing N2 ambient effectively reduced incubated enhanceddiffusion. The effect of F was also illustrated, for B plus F through-oxide implants withpure N2 annealing. By comparing B diffusion between B-plus-F and B-plus-Ne implants,it was deduced that F did not have a chemical effect upon the capture of point defectswhich caused enhanced diffusion. Instead, F was thought to be incorporated into Oprecipitates. This then affected point defect generation.D.Fan, J.M.Parks, R.J.Jaccodine: Applied Physics Letters, 1991, 59[10], 1212-4

[446-84/85-064]

Bulk Diffusion - Qualitative Observations - Defect InteractionsIt was demonstrated that, during the formation of Pd2Si, point defects were injected intothe Si substrate and led to an appreciable degree of dopant diffusion at temperatures aslow as 200C. This was the lowest temperature at which dopant diffusion had ever beenobserved. Buried marker layers exhibited asymmetrical diffusion, with migrationoccurring preferentially towards the silicide-forming interface. The results showed that alarge gradient in point defect concentration did not develop across the width of a markerlayer. This showed that it was not possible for a large point defect gradient to be createdby trapping as they diffused through the doped layers.M.Wittmer, P.Fahey, G.J.Scilla, S.S.Iyer, M.Tejwani: Physical Review Letters, 1991,66[5], 632-5

[446-84/85-065]

Bulk Diffusion - Qualitative Observations - Defect InteractionsDefect states which were introduced into float-zone material by heavy dopant diffusionwere investigated by using the electron beam-induced current method. By measuring theminority carrier diffusion length, using the first-order moment method, and by directlyimaging the defects, their electrical activity could be determined. The diffused sampleswere then dry-oxidized, so that changes in the electrical and morphological properties ofthe introduced defects could be monitored. Two sets of samples were investigated, one ofwhich had been diffused with B and another which had been diffused with B and Al.Significant improvements in the diffusion length were observed in the samples into whichAl had been co-diffused; thus providing evidence for an effect of Al upon the electricalactivity of bulk defective states.A.Castaldini, A.Cavallini, B.Fraboni, E.Giannotte: Journal of Applied Physics, 1992,72[12], 5622-7

[446-106/107-130]

Bulk Diffusion - Qualitative Observations - Defect InteractionsBy using the B decoration method, an investigation was made of the structure of avacancy-dislocation cluster in the disturbed layer of monocrystals. The B diffusioncoefficient was determined for various parts of a vacancy-dislocation cluster which wasformed from excess vacancies. The experimental results demonstrated that vacancies were

289

B Bulk Diffusion B

distributed in such clusters according to a Gaussian law. A non-linear diffusionmechanism was suggested to apply at low dopant concentrations.A.E.Alekseev, O.I.Gerasimov, A.P.Fedchuk: Fizika Tverdogo Tela, 1991, 33[7], 2153-8(Soviet Physics - Solid State, 1991, 33[7], 1212-5)

[446-86/87-053]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe non-equilibrium impurity diffusion of dopants in monocrystalline Si was carried outby means of the controlled surface injection of self-interstitials and vacancies. By varyingthe parameters of the surface oxide layer during the B diffusion process, it was possible toproduce quantum-sized profiles and p-n junctions with dimensions that could becontrolled within the 1 to 22nm range.N.T.Bagraev, L.E.Klyachkin, V.L.Sukhanov: Semiconductor Science and Technology,1991, 6[7], 577-81

[446-88/89-051]

Bulk Diffusion - Qualitative Observations - Defect InteractionsIons of B were implanted at a wafer temperature of 950C. It was found that the resultantB profile revealed the occurrence of marked up-hill diffusion at the surface, plus a veryhigh degree of diffusion enhancement. The results confirmed the effect of point defectgradients upon dopant migration. It was found that the experimental results agreed wellwith the predictions of pair diffusion theories.P.Pichler, R.Schork, T.Klauser, H.Ryssel: Applied Physics Letters, 1992, 60[8], 953-5

[446-88/89-051]

Bulk Diffusion - Qualitative Observations - Defect InteractionsA study was made of a low-temperature long-range interstitial B migration behaviorwhich was initiated by kick-out reactions with self-interstitials that had been created by Siimplantation. At 450C, migration path lengths of about 100nm were measured within atime frame of 0.25h. Observations of the numbers and spatial distribution of displaced Batoms, with regard to self-interstitial and vacancy distributions which had been deducedfrom ion transport theory, permitted the reaction kinetics to be estimated. The reactions:

self-interstitial + Bs ? Bi

vacancy + Bi ? Bs

self-interstitial + vacancy ? 0were essentially diffusion-limited and were associated with capture radii which were ofatomic dimensions.N.E.B.Cowern, G.F.A.Van de Walle, P.C.Zalm, D.J.Oostra: Physical Review Letters,1992, 69[1], 116-9

[446-93/94-049]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe effect of substitutional C upon interstitial-enhanced B diffusion was investigated. Thesubstitutional C was incorporated into B-doped superlattices by using molecular beamepitaxial growth under acetylene gas. Excess Si self-interstitials were generated by near-surface Si implantation (5 x 1018/cm2, 40keV) and diffusion at 790C. It was found that theinterstitial-enhanced diffusion of the B marker layers was entirely suppressed by a C

290

B Bulk Diffusion B

concentration of 2 x 1019/cm3. This demonstrated that substitutional C acted as a trap forinterstitials in crystalline Si. The uniform incorporation of 5 x 1018/cm2 of C significantlyreduced the transient enhanced diffusion of a typical B junction implant withoutperturbing its electrical activity.P.A.Stolk, D.J.Eaglesham, H.J.Gossmann, J.M.Poate: Applied Physics Letters, 1995,66[11], 1370-2

[446-121/122-085]

Bulk Diffusion - Qualitative Observations - Defect InteractionsDifferences in B diffusivity were used to characterize epitaxially grown layers. After anoxidation-enhanced diffusion of B spikes, a decrease in B diffusivity with increasingdepth was observed in epitaxial Si layers which had been grown by means of molecularbeam epitaxial deposition and fast gas switching vapor deposition. This differed from thebehavior which was observed in layers that had been grown by means of low-temperaturechemical vapor deposition. The reduced B diffusivity was suggested to be caused by asupersaturation of vacancy defects, which acted as interstitial traps and suppressed thediffusion of B.K.J.Van Oostrum, P.C.Zalm, W.B.De Boer, D.J.Gravesteijn, J.W.F.Maes: AppliedPhysics Letters, 1992, 61[13], 1513-5

[446-099/100-096]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe diffusion of B in polycrystal-on-monocrystal samples was studied by using secondaryion mass spectrometry. The extrapolated diffusion profiles in polycrystalline material andin the monocrystalline substrate revealed a discontinuity at their interface. Thisdiscontinuity was attributed to the blockage of B-defect complexes by an interfacial oxidebetween the polycrystalline and monocrystalline materials, as well as to the immobility ofthese defect complexes in monocrystalline Si. The B in the implant peak region above theB solid solubility limit was found to be immobile in monocrystalline Si during annealing,due to the formation of electrically inactive B-defect complexes. However, inpolycrystalline material the B in the peak region spread out more rapidly than it would inmonocrystalline material; due possibly to the diffusion of B-defect complexes along grainboundaries. The latter complexes were electrically inactive, according to spreadingresistance data. If the B concentration was decreased to below the solid solubility limit,either by decreasing the dose or by increasing the annealing temperature, no discontinuitywas observed (in the B profile) across the polycrystalline/monocrystalline interface.S.Batra, M.Manning, C.Dennison, A.Sultan, S.Bhattacharya, K.Park, S.Banerjee,M.Lobo, G.Lux, C.Kirschbaum, J.Norberg, T.Smith, B.Mulvaney: Journal of AppliedPhysics, 1993, 73[8], 3800-4

[446-106/107-136]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe behavior of B in degenerately doped Czochralski material was studied after extendedannealing at temperatures ranging from 750 to 900C in a N ambient. The results ofsecondary ion mass analyses revealed an accumulation of B atoms near to the Si surface.On the basis of the defect morphology, as observed using high-resolution electron

291

B Bulk Diffusion B

microscopy, it was concluded that degenerate Si was saturated with interstitial defectspecies during extended annealing. Point defect reaction and dynamic clustering modelssuggested that an accumulation of B near to the Si surface was due to a B out-diffusionwhich was driven by the difference in the chemical potential of B saturation between thebulk and surface regions.W.Wijaranakula: Applied Physics Letters, 1993, 62[23], 2974-6

[446-106/107-137]

Bulk Diffusion - Qualitative Observations - Defect InteractionsA test structure which consisted of a box-shaped B profile, capped by a lightly-doped Aslayer, was used to determine whether extended defects absorbed some of the interstitialswhich were injected during wet thermal oxidation. Reduced oxidation-enhanced diffusionof the B layer was observed in samples which contained extended defects. Secondary ionmass spectrometry measurements were combined with transmission electron microscopicmeasurements in order to calculate a product, of interstitial diffusivity and interstitialconcentration, which was in good agreement with values which had previously beendeduced from Au diffusion experiments. In addition, a lower bound on the ratio, of thenet numbers of Si atoms which were injected during oxidation to the number of Si atomsconsumed, was calculated. A one-dimensional model for the growth of extended defectswas studied, and simulations which were performed using the new model agreed withexperimental data. The growth of the extended defects was shown to be a reaction-limitedprocess.R.Y.S.Huang, R.W.Dutton: Journal of Applied Physics, 1993, 74[9], 5821-7

[446-111/112-070]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe diffusion of B in thin films which had been grown by means of low-temperaturemolecular beam epitaxy was investigated at temperatures ranging from 750 to 900C, andtimes of between 0.25 and 60h. The small spatial extent of the initial δ-function-likedopant profiles permitted the detection of very small diffusional displacements. Thedopant atoms were used as tracers of Si point defects (vacancies and self-interstitials).The diffusion of B was found to be retarded relative to the equilibrium values. A modelwas proposed which was based upon an initial supersaturation of vacancies. Thematching of this model to experimental data permitted the extraction of the vacancydiffusivity, the activation energy for vacancy formation, and the recombination lifetime ofinterstitials. The results showed that the interstitial and vacancy populations could not beconsidered to be independent at low temperatures, as had previously been suggested.H.J.Gossmann, C.S.Rafferty, A.M.Vredenberg, H.S.Luftman, F.C.Unterwald,D.J.Eaglesham, D.C.Jacobson, T.Boone, J.M.Poate: Applied Physics Letters, 1994, 64[3],312-4

[446-115/116-151]

Bulk Diffusion - Qualitative Observations - Defect InteractionsAn effective interstitial surface recombination velocity for the buried Si/SiO2 interface inSIMOX material was used to model accurately the oxidation-enhanced diffusion of B insingly- and multiply-implanted material. The effective recombination velocity at the

292

B Bulk Diffusion B

SIMOX interface was found to be higher than the value for a thermally grown SiO2interface. The enhancement of the effective recombination velocity depended upon thematerial formation conditions and was empirically related to the near-interface dislocationdensity. An increased surface interfacial area was considered to be the most likely causeof the increased effective recombination velocity.S.W.Crowder, P.B.Griffin, C.J.Hsieh, G.Y.Wei, J.D.Plummer, L.P.Allen: AppliedPhysics Letters, 1994, 64[24], 3264-6

[446-115/116-185]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe diffusion behavior of B was used to study point defect kinetics in Si-on-insulatorsamples. Marker layers of P were used to study oxidation-enhanced diffusion in bulk andbonded and etched-back samples under oxidizing conditions at 750, 800 and 850C. Aneffective interstitial recombination velocity for the buried Si/SiO2 interface in the etched-back material was extracted by comparing the experimentally obtained P profiles withsimulation results. The data could be modelled by assuming a time-independent interfacerecombination velocity. The same parameter set, but incorporating this deducedrecombination velocity, accurately reproduced the implant-enhanced diffusion of Bmarker layers at 750 and 800C in thin films. This implied that the recombination velocitywas independent of the interstitial supersaturation.S.W.Crowder, C.J.Hsieh, P.B.Griffin, J.D.Plummer: Journal of Applied Physics, 1994,76[5], 2756-64

[446-117/118-216]

Bulk Diffusion - Qualitative Observations - Defect InteractionsImplanted B and P exhibit transient enhanced diffusion during initial annealing, due to Siinterstitials which are emitted from regions of implantation damage. However, the sourceof these interstitials had not previously been identified. Here, quantitative transmissionelectron microscopy of extended defects was used to demonstrate that transient enhanceddiffusion was caused by the emission of interstitials from specific defects. The defectswere rod-like, ran along <110> directions, and consisted of interstitials which precipitatedonto 311 planes as a single monolayer of hexagonal Si. The evaporation of 311defects during annealing at 670 and 815C was correlated with the length of the diffusiontransient, and it was demonstrated that there was a link between the numbers ofinterstitials which were emitted by the defects and the flux of interstitials which led totransient enhanced diffusion. It was concluded that not only did 311 defects contributeto the interstitial flux, but that a contribution which could be attributed to 311 defectevaporation was sufficient to explain all of the observed transient. The 311 defectswere the source of the interstitials.D.J.Eaglesham, P.A.Stolk, H.J.Gossmann, J.M.Poate: Applied Physics Letters, 1994,65[18], 2305-7

[446-119/120-222]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe interaction of implantation-induced dislocation loops and interstitials was studied.Experiments which were carried under dry oxidation conditions consistently revealed a

293

B Bulk Diffusion B

significant reduction in oxidation-enhanced B diffusion in a buried layer, due to the veryefficient interstitial capture effect of dislocation loops; thus suggesting the occurrence ofdiffusion-limited dislocation loop growth. A simple analytical solution for interstitialsupersaturation, and an analysis of the data in terms of a time dependence of theoxidation-enhanced diffusion suppression, demonstrated that the interaction of dislocationloops and interstitials was not a reaction-limited, but a diffusion-limited, process.Simulations which incorporated a model for the interaction mechanism agreed withsecondary ion mass spectroscopic and with transmission electron spectroscopic data.H.Park, H.Robinson, K.S.Jones, M.E.Law: Applied Physics Letters, 1994, 65[4], 436-8

[446-119/120-223]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe effect of the electrical deactivation of As was studied. High concentrations of Aswere implanted and were laser melt annealed so as create box-like fully electrically activeAs layers, with no residual implantation damage. The wafers were then subjected to low-temperature thermal cycles while a buried B layer monitored point defects. A markedenhancement of B diffusion was observed which suggested that As deactivation releasedlarge numbers of interstitials. This was explained in terms of a process in which thevacancies that were required by deactivated As structures were created by a deactivation-assisted Frenkel pair generation process, thus injecting interstitials.P.M.Rousseau, P.B.Griffin, J.D.Plummer: Applied Physics Letters, 1994, 65[5], 578-80

[446-119/120-223]

Bulk Diffusion - Qualitative Observations - Defect InteractionsThe migration of B in heavily doped layers (extrinsic at the process temperature) wasstudied in order to obtain information on Si interstitial concentrations in heavily dopedlayers. The diffusion of the dopant in n-type and p-type iso-concentration structures wasmeasured for 2 surfaces. One surface was passivated with low-pressure chemical vapordeposited oxides and nitrides, and the other was a growing thermal oxide. The degree ofoxidation-enhanced diffusion was measured for both dopants and was found to be smallerthan for the same annealing treatments under lightly doped conditions (intrinsic at processtemperatures). It was found that samples with epitaxial layers exhibited greaterenhancements as compared with samples that did not. The results were explained in termsof charged interstitial drift in the electric field which was set up by ionized dopant atoms,and a resultant increase in the flux of interstitials away from the oxidizing interface.D.J.Roth, J.D.Plummer: Journal of the Electrochemical Society, 1994, 141[4], 1074-81

[446-119/120-223]

Bulk Diffusion - Qualitative Observations - Defect InteractionsA B-doped epilayer was used to investigate the interaction between end-of-rangedislocation loops (created by Ge+ implantation) and excess point defects which weregenerated by low doses (1014/cm2) of implanted B+. The B doping spike was grown in, bymeans of chemical vapor deposition, at a depth of 800nm below the surface. The intrinsicdiffusivity of the B in the doped epilayer was determined by annealing the as-grownlayer. The end-of-range (type-II) dislocation loops were created by using 2 overlappingroom-temperature Ge+ implantations, of 75 and 190keV, to a dose of 1015/cm2. During

294

B Bulk Diffusion B

annealing, the amorphous layer re-grew and a layer of type-II dislocation loops formed ata depth of about 230nm and with a density of about 8 x 1010/cm2. The enhanceddiffusivity in the buried B layer, due to type-II loop formation by Ge+ implantation, wasobserved to increase (within some 150 to 300s), from 1500 to 2500 times the intrinsicdiffusivity value, before falling back to intrinsic levels after 0.5h at 800C. Low-energy(8keV) implantation of B+, to a dose of 1014/cm2 and a range of 32nm, into materialwithout loops resulted in an average enhancement to 1540 times the B epilayer diffusivityafter 150s at 800C. The diffusivity fell back to intrinsic levels after 300s at 800C. Whenan annealed layer of loops was introduced (before, and deeper than, subsequent low-energy implanted B+), annealing of the implanted B+ produced no measurableenhancement of the buried B-layer diffusivity. It was deduced that the kinetics ofinteraction between the dislocation loop layer and the damage-induced interstitials weremainly diffusion-limited, and that the loops absorbed a significant fraction of theinterstitials which were produced by low-energy B+ implantation.J.K.Listebarger, H.G.Robinson, K.S.Jones, M.E.Law, D.D.Sieloff, J.A.Slinkman,T.O.Sedgwick: Journal of Applied Physics, 1995, 78[4], 2298-302

[446-123/124-181]

Bulk Diffusion - Qualitative Observations - Defect InteractionsA study was made of interactions, between point defects and pre-existing extendeddefects, via associated changes in the transient enhanced diffusion of B. The extendeddefects were produced by 40 to 300keV Si implantation to a dose of 2 x 1015/cm2,followed by rapid thermal annealing (1000C, 40s, N ambient). Ions of B were thenimplanted, using an energy of 40keV and a dose of 4 x 1014/cm2. The samples were againannealed by means of rapid thermal annealing (800 to 1100C). Transmission electronmicroscopy and secondary ion mass spectrometry were used to monitor changes indislocation loops and B profiles. These loops lay close to the B distribution, and it wasnoted that the transient enhanced diffusion of B during post-implantation annealing wasstrongly suppressed. This was attributed to the trapping, by loops, of interstitials whichwere associated with the implanted B.S.Pan, I.V.Mitchell: Materials Chemistry and Physics, 1996, 46[2-3], 252-8

[446-148/149-185]

Bulk Diffusion - Qualitative Observations - Effect of AlloyingAn investigation was made of B diffusion in Si and in strained SiGe in situ dopedepitaxial layers. During inert ambient annealing at 860C, B diffusion was observed to beslower in Si0.83Ge0.17 than in Si for B concentrations of between 5 x 10

16 and 2.5 x

1019

/cm3. Computer simulations of the measured profiles in annealed samples indicated

that the effective B diffusivity in Si0.83Ge0.17 was about an order of magnitude lower thanthat in Si. This disparity increased with increasing B concentration.P.Kuo, J.L.Hoyt, J.F.Gibbons, J.E.Turner, R.D.Jacowitz, T.I.Kamins: Applied PhysicsLetters, 1993, 62[6], 612-4

[446-106/107-134]Bulk Diffusion - Qualitative Observations - Effect of AmbientThe effect of the annealing ambient upon dopant diffusion was studied, during low-temperature processing by implanting BF2 into (100) samples (at room temperature)

295

B Bulk Diffusion B

through a 14nm SiO2 layer. The implantation dose and energy were sufficient toamorphize the surface. After low-temperature furnace annealing, the ion-implanted Bexhibited a transient enhanced diffusion regime for both inert and oxidizing ambients. Ithad been expected that point defect generation during the annealing of implantationdamage would predominate during the transient enhanced diffusion process; regardless ofthe nature of the ambient.Y.Kim, H.Z.Massoud, R.B.Fair: Journal of the Electrochemical Society, 1990, 137[8],2599-603

[446-76/77-033]

Bulk Diffusion - Qualitative Observations - Effect of AmbientThe out-diffusion of elements from bare wafers at 1200C, and especially its dependenceupon the annealing atmosphere, was studied by using spreading resistance and secondaryion mass spectroscopic methods. It was found that B out-diffusion occurred when thesample was annealed in H, but was completely suppressed in Ar; even when the dopantconcentration was as high as 3 x 1018/cm3 and the annealing time was as long as 2h.L.Zhong, Y.Kirino, Y.Matsushita, Y.Aiba, K.Hayashi, R.Takeda, H.Shirai, H.Saito,J.Matsushita, J.Yoshikawa: Applied Physics Letters, 1996, 68[9], 1229-31

[446-131/132-182]

Bulk Diffusion - Qualitative Observations - Effect of DefectsIt was noted that the transient diffusion of ion-implanted B was inhibited by the presenceof high C or B concentrations. This was due to the formation of interstitial clusters thatwere stabilized by impurity atoms. A comparison of experimental data and simulationresults suggested that the number of self-interstitials that was trapped per clusteredimpurity atom was equal to about 1.15 for C and to about 1 for B. These values wereconsistent with the operation of a volume compensation mechanism.N.E.B.Cowern, A.Cacciato, J.S.Custer, F.W.Saris, W.Vandervorst: Applied PhysicsLetters, 1996, 68[8], 1150-2

[446-134/135-157]

Bulk Diffusion - Qualitative Observations - Effect of DefectsPhenomena which were associated with the transient diffusion of implanted B werereviewed. It was recalled that the Si self-interstitial flux, which caused B interstitialcydiffusion, had been directly observed by using B diffusion marker layers that had beenfabricated by means of low-temperature crystal growth. Low-energy and low-dose Siimplantation into surface layers permitted the experimental separation of the source andthe flux of interstitials. The clustering of B was directly correlated with thesupersaturation of injected Si interstitials. The diffusivity of the Si interstitial wasdominated by the presence of trapping sites. Implantation produced rod-like defects whichconsisted of interstitial precipitates, on 311 planes, that were in the form of a singlemonolayer of hexagonal Si. The numbers of Si interstitials which evaporated from the311 defect were sufficient to explain all of the transient diffusion.P.A.Stolk, H.J.Gossmann, D.J.Eaglesham, J.M.Poate: Nuclear Instruments and Methodsin Physics Research B, 1995, 96[1-2], 187-95

[446-134/135-158]

296

B Bulk Diffusion B

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe contribution, to gettering, of interstitial injection that was induced by B diffusionunder non-oxidizing conditions was considered with regard to short-range (doped-layer)effects and long-range (bulk) effects. It was shown that a positive concentration gradientof Si interstitials within a highly doped layer was favorable to the gettering ofsubstitutional metallic species via a kick-out mechanism during B diffusion. Also, thecontribution of B diffusion to the gettering of 3d elements was significant only within ahighly P-doped layer (short-range effect). The predictions of the diffusion-inducedgettering model agreed with experimental evidence on the B diffusion-gettering of Au.F.Gaiseanu, W.Schröter: Journal of the Electrochemical Society, 1996, 143[1], 361-2

[446-134/135-160]

Bulk Diffusion - Qualitative Observations - Effect of DefectsAn experimental study was made of the interaction between C and interstitials incrystalline material. Superlattices of B-doped material, prepared by means of molecularbeam epitaxy, were used to detect the injection of excess Si self-interstitials from near-surface 40keV Si implants (5 x 1013/cm2) during annealing. It was found that interstitial-enhanced diffusion of the B marker layer was entirely suppressed when the C content wasincreased uniformly from 1018 to 2 x 1019/cm3. This demonstrated that C acted as a trapfor interstitials. The incorporation of a buried C layer led to a local reduction in theinterstitial concentration without fully obstructing the interstitial flow towards the bulk.The interstitial trapping was thought to involve the formation of mobile C-interstitialcomplexes which paired with substitutional C and became immobile.P.A.Stolk, H.J.Gossmann, D.J.Eaglesham, J.M.Poate: Materials Science and EngineeringB, 1996, 36[1-3], 275-81

[446-136/137-128]

Bulk Diffusion - Qualitative Observations - Effect of DopantExperimental results on the effects of the As dopant concentration upon B out-diffusionin n-polycrystalline/p-monocrystalline structures were considered. It was found that the Bdiffusivity was only 30 times higher in polycrystalline material than in monocrystallinematerial if the degree of As doping was high enough to cause enhanced grain growth. Thediffusivity increase was equal to about 130, if the polycrystalline material had a smallgrain size due to a low As dopant concentration. The B loss from the base region of anadvanced bipolar transistor doping profile, by out-diffusion into the emitterpolycrystalline Si, was of the order of 20% and had to be considered in order to ensureaccurate device modelling.J.N.Burghartz, C.L.Stanis, P.A.Ronsheim: Applied Physics Letters, 1995, 67[21], 3156-8

[446-127/128-154]

Bulk Diffusion - Qualitative Observations - Effect of IlluminationThe effect of in situ photo-excitation, during B ion implantation, upon subsequenttransient enhanced B diffusion was investigated. Photo-excitation, using a Hg arc-lamp,was performed during 35keV B+ implantation to a dose of 5 x 1014/cm2 at 177K. A reverseannealing effect was observed at temperatures ranging from 550 to 700C. Also, thetransient enhanced diffusion of B, as measured using secondary ion mass spectrometry

297

B Bulk Diffusion B

following annealing (800C, 0.5h), was suppressed. Both effects demonstrated that thecreation of self-interstitials during implantation was significantly reduced.J.Ravi, J.Erokhin, G.A.Rozgonyi, C.W.White: Applied Physics Letters, 1995, 67[15],2158-60

[446-125/126-146]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe effects of implantation damage upon B diffusion during rapid thermal annealing andconventional heat treatment at 750 to 900C were investigated. A comparison of thedopant profiles and junction depths in damaged and undamaged regions demonstrated thatSi+ implantation under the typical conditions used for pre-amorphization produced amarkedly enhanced diffusion of B atoms.A.Jacques, A.George, X.Baillin, J.J.Bacmann: Philosophical Magazine A, 1987, 55[2],165-81

[446-51/52-130]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationDiffusion experiments were performed in the presence of lattice defects which had beenproduced by Si ion implantation. The effects of transient enhanced diffusion were studiedby means of bevelling and staining measurements of implanted samples, and secondaryion mass spectroscopic determinations of dopant profiles. The annealing temperatures fordoped implanted specimens ranged from 700 to 1100C and this treatment was performedusing an electron beam. The low temperatures which were used permitted the kinetics ofanomalous diffusion to be monitored. It was ascertained that the enhanced diffusioncoefficient was almost constant during a period which decreased with increasingtemperature. It then tended gradually to the equilibrium value. This trend agreed with thatof lattice damage changes which were revealed by double crystal X-ray analyses of therocking curves of implanted samples. The secondary ion mass spectroscopy profilesindicated that only a fraction of the dopant which was located at residual implantationdamage was responsible for the anomalous diffusion.R.Angelucci, F.Cembali, P.Negrini, M.Servidori, S.Solmi: Journal of the ElectrochemicalSociety, 1987, 134[12], 3130-4

[446-60-013]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationIons of As and B were implanted using energies which had been chosen so that theirprojected ranges coincided, and the implanted material was annealed in Ar gas (950C, 0.5to 5h). The diffusivity of B atoms decreased with an increase in annealing time.K.Yokota, Y.Okamoto, F.Miyashita, T.Hirao, M.Watanabe, K.Sekine, Y.Ando,K.Matsuda: Journal of Applied Physics, 1994, 75[11], 7247-51

[446-117/118-193]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe enhanced diffusion tail of implanted B was studied using conventional furnaceannealing. Surface layer stripping was used to remove part of the sample before annealingin order to distinguish the effect of the defect-rich surface region from the tail region. The

298

B Bulk Diffusion B

B concentration profiles were obtained by using secondary ion mass spectrometry, andspreading resistance profiles were determined for comparison. The results showed thatimplantation-induced damage in the surface region was responsible for enhanced Bdiffusion. Channelled B in the tail of the implantation profile was activated duringannealing and had little effect upon the tail redistribution during annealing.D.Fan, J.Huang, R.J.Jaccodine, P.Kahora, F.Stevie: Applied Physics Letters, 1987,50[24], 1745-7

[446-55/56-035]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe effect of Si ion implantation upon the anomalous transient diffusion of ion-implantedB was investigated. It was found that Si ion fluences which were well below that whichwas required to amorphize the lattice substantially reduced the anomalous transientdiffusion of subsequently implanted B. However, the sheet resistance was increased byadditional Si implantation. The implantation of Si ions into activated B layers causedadditional anomalous diffusion at distances which were substantially beyond the range ofthe Si ions. Anomalous motion was also reduced in regions where the damage wasgreater. These effects could be explained in terms of the generation of point defectclusters which dissolved during annealing, and the annihilation of point defects in regionsof high damage by the formation of interstitial-type extended defects.A.E.Michel, W.Rausch, P.A.Ronsheim: Applied Physics Letters, 1987, 51[7], 487-9

[446-55/56-035]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe effect of the rate of defect generation upon the radiation-enhanced diffusion of pre-implanted B atoms, due to post-bombardment at 750C, was investigated by usingsecondary ion mass spectrometry depth profiling. The generation rate was varied bychanging the beam current density and the ion species. It was found that the excess Bdiffusivity increased in proportion to the generation rate for light ion irradiation.However, for production rates greater than about 0.2dpa/s, it again decreased. The rateeffects were explained in terms of a dynamic overlapping model.K.Holldack, H.Kerkow, M.Gericke: Physica Status Solidi A, 1987, 102[2], 633-8

[446-55/56-036]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe properties of B-doped Si layers obtained by bombardment with BF2

+ molecular ionswere studied by using Rutherford back-scattering spectrometry, transmission electronmicroscopy, secondary ion mass spectrometry, and incremental sheet resistance and sheetHall coefficient measurements involving anodic sectioning. The implantation step wasresponsible for the formation of interstitial aggregates at the amorphous/crystal interface.The size of these aggregates, and the total number of defects involved, depended upon theimplantation conditions. It was different for samples implanted at medium currents orhigh currents. During high (900 to 1000C) temperature diffusion in an inert atmosphere,the aggregates annealed out and caused interstitial over-saturation. As a result, diffusionenhancement occurred. This was modelled by simulating the creation of an excess ofinterstitials by heat treatment in a partially oxidizing atmosphere.

299

B Bulk Diffusion B

G.Queirolo, P.Caprara, L.Meda, C.Guareschi, M.Anderle, G.Ottaviani, A.Armigliato:Journal of the Electrochemical Society, 1987, 134[11], 2905-11

[446-60-014]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationA comparison was made of 2 shallow pre-amorphization techniques which involved Si+

and Ge+ implantation followed by B+ implantation and rapid thermal annealing. Theresultant effect upon B diffusion profiles and extended defects was studied by usingsecondary ion mass spectroscopy and cross-sectional transmission electron microscopy.Enhanced or retarded B diffusion during rapid thermal annealing was related to therelative depths of the original amorphous/crystalline interface and the as-implanted Bprofiles, to the type of ion used for pre-amorphization, and to the initial type and relativelocation of radiation-induced point defects. In general, Si+ self-implanted samplesexhibited less B profile broadening than did Ge+ implanted samples after rapid thermalannealing (1050C, 10s). The conditions which were necessary for the completeannihilation of end-of-range interstitial loops after Si+ self-amorphization were identified.A.C.Ajmera, G.A.Rozgonyi, R.B.Fair: Applied Physics Letters, 1988, 52[10], 813-5

[446-62/63-228]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe transient-enhanced diffusion of B was studied by comparing through-oxide and directB implants. Some of the directly implanted wafers were differentially anodized andetched to depths which were equal to the thickness of the thermal oxide. The spreadingresistance profiles which were measured after annealing revealed that transient-enhancedB diffusion could be prevented by implantation through the oxide. The results indicatedthat the presence of recoiled O in through-oxide implanted material could be important inreducing dopant-enhanced diffusion.D.Fan, R.J.Jaccodine: Applied Physics Letters, 1989, 54[7], 603-5

[446-64/65-177]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationIons of BF2

+ were implanted into (100) samples at room temperature, with an energy of40keV, through a 14nm-thick SiO2 layer. Profiling of B by secondary ion massspectrometry indicated that subsequent annealing (650-850C, 0.5-4h) in a conventionalfurnace led to a pronounced secondary peak in the B and F profiles; in addition to thenear-surface primary peak which was situated in the vicinity of the projected range of theimplanted species. This effect was also observed in samples which were rapidly thermallyannealed (900C, 15-60s). The depths of the secondary peaks in the B profilescorresponded to the depths of a damaged layer which was observed via cross-sectionaltransmission electron microscopy. Isochronal furnace annealing revealed that there wasno chemical interaction between B and F atoms during annealing. This was alsosupported by the observation that F atoms did not affect the B segregation coefficientduring the oxidation of implanted samples. End-of-range extended dislocations appearedto be responsible for the gettering of atoms during annealing.Y.Kim, H.Z.Massoud, R.B.Fair: Applied Physics Letters, 1988, 53[22], 2197-9

[446-64/65-178]

300

B Bulk Diffusion B

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationEnhanced B diffusion was habitually observed during the growth of ion-bombardedepitaxial layers via molecular beam epitaxy. Ion-bombardment methods were usuallyrequired in order to obtain high levels of n-type doping, and the damage caused by low-level ion bombardment was responsible for the enhanced B diffusion. The concentrationprofiles of as-grown and post-growth annealed samples revealed that the diffusion was atransient effect which occurred at growth temperatures of 600 to 700C. Simulation of thediffusion process demonstrated that almost all of the B participated in diffusion, and thatthe built-in electric field at the p-n junction led to a further smearing of the B profile.P.R.Pukite, S.S.Iyer, G.J.Scilla: Applied Physics Letters, 1989, 54[10], 916-8

[446-70/71-122]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationFloat-zone [100]-type wafers were implanted with 15 to 50MeV B ions to doses ofbetween 1013 and 1015/cm2. The implanted samples were furnace-annealed at temperaturesof between 800 and 1250C, and were analyzed by using spreading resistanceprofilometry. The range distributions for various implant energies were compared withpredictions which were based upon Bethe electronic energy loss, electronic straggling,and large-angle Rutherford back-scattering. Good agreement was found with regard to theprojected range and straggling behavior. However, some discrepancies were evident in thetails of the dopant distributions. It was found that B diffusion during oxidation wasenhanced by about 60%. This suggested that interstitials migrated over distances of atleast 0.1mm at these temperatures.A.La Ferla, A.Di Franco, E.Rimini, G.Ciavola, G.Ferla: Materials Science andEngineering B, 1989, 2, 69-73

[446-70/71-122]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe anomalous diffusion of implanted B, in material which had been damaged by self-implantation, was investigated by means of cross-sectional transmission electronmicroscopy and secondary ion mass spectroscopy. During rapid thermal annealing, bulgesin the B profile were observed at the edges of the defect band, and profile broadening wasaccompanied by shrinking of the band. The decay time essentially vanished due to theemission of point defects from the edges of the defect band. It was suggested that theanomalous B diffusion was related to the point defects which were produced byimplantation and which were emitted from the defect band.Q.Guo, X.Bao, J.Hong, Y.Yan, D.Feng: Applied Physics Letters, 1989, 54[15], 1433-5

[446-70/71-123]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe effects of extended defect evolution upon the anomalous diffusion of ion-implantedB during rapid thermal annealing were studied by means of transmission electronmicroscopy and secondary-ion mass spectroscopy. It was found that, for low-dose Bimplants (less than 1014/cm2), no extended defects were observable after rapid thermalannealing at 1000C, and the anomalous diffusion saturated within less than 10s. However,extended defects developed after making high-dose B implants (greater than 5 x 1014

301

B Bulk Diffusion B

/cm2). In this case, the anomalous diffusion persisted for a much longer time and wasdose-dependent. The evolution of extended defects was linked to the anomalous Bdiffusion behavior.Y.M.Kim, G.Q.Lo, D.L.Kwong, A.F.Tasch, S.Novak: Applied Physics Letters, 1990,56[13], 1254-6

[446-74-050]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe effect of implantation damage upon anomalous diffusion was studied by varying therelative positions of the B profile and of the damage distribution which was produced bySi pre-implantation using various energies. It was shown that the anomalous diffusion ofimplanted B was caused by the implantation damage rather than by the fast-diffusinginterstitial B. During annealing, the extended defects which acted as sinks for pointdefects retarded anomalous diffusion in, and near to, the defect band. However, duringprolonged annealing they began to dissolve and to emit point defects which then causedthe anomalous diffusion.X.M.Bao, Q.Guo, M.S.Hu, D.Feng: Journal of Applied Physics, 1989, 66[3], 1475-7

[446-74-051]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationOxygen-implanted film material was subsequently implanted with BF2 to low or highdoses and then annealed at various temperatures. Transmission electron microscopy,secondary ion mass spectrometry, and spreading resistance methods were used to analyzethe resultant structures. It was found that a pile-up phenomenon occurred close to theupper surface of the film, and anomalous B diffusion was detected after low-dose BF2

implants. These observations were explained in terms of the quality of the film. TheMcNabb-Foster model for diffusion plus trapping was solved in order to interpret theeffects. The redistribution of B after high-dose BF2 implants was the same for O-implanted films as for bulk Si.P.Normand, D.Tsoukalas, N.Guillemot, J.Stoemenos: Journal of the ElectrochemicalSociety, 1990, 137[7], 2306-13

[446-76/77-035]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe role which was played by extended defect evolution in anomalous B diffusion duringrapid thermal annealing was studied by means of secondary ion mass spectroscopy andtransmission electron microscopy. Pre-damaged wafers with low-dose Si implantation,pre-amorphized wafers with high-dose Si implantation, and monocrystalline wafers withno previous implantation were used. It was found that low-dose Si pre-implantationsignificantly reduced the channelled tail during subsequent B implantation, and resultedin more anomalous B diffusion and more rapid annealing of extended defects during rapidthermal annealing, as compared with crystalline Si samples. The diffusion of B which hadbeen implanted into pre-amorphized Si was found to be anomalous in nature and itsmagnitude depended upon the rapid thermal annealing temperature. The temperaturedependence was found to be due to the difference in the density of dislocation loopswhich formed during rapid thermal annealing at the original amorphous/crystalline

302

B Bulk Diffusion B

interface. These loops determined the effectiveness of the trapping Si interstitials whichdiffused, from the crystalline side of the original amorphous/crystalline interface, to theepitaxially re-grown region. The anomalous B diffusion in the crystalline Si samples wasfound to be a sensitive function of the implant dose. The diffusion displacementincreased, and the anomalous diffusion effect lasted longer, with increasing implant dose.An enhanced diffusion after longer periods of time, in samples with higher implant doses,was related to the formation and annealing of extended defects. At very high doses(greater than 2 x 1015/cm2), where the peak B concentration was above the solid solubility,excess diffusion in the high B concentration region was observed. The excess diffusionwas explained in terms of B segregation into interstitial dislocation loops in the earlystages of rapid thermal annealing, and to the subsequent annealing of these dislocationloops after the initially large anomalous tail diffusion.Y.M.Kim, G.Q.Lo, H.Kinoshita, D.L.Kwong, H.H.Tseng, R.Hance: Journal of theElectrochemical Society, 1991, 138[4], 1122-30

[446-81/82-049]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationSamples which contained various p-type and n-type dopants were implanted with spin-polarized 12B at temperatures of between 300 and 950K. By using β-radiation detectednuclear magnetic resonance, the fraction of 12B in normal substitutional sites wasdetermined. It was found that this fraction increased from about 20% at 300K to about100% at a temperature which depended systematically upon the bulk doping. It wasargued that the diffusive motion of interstitial B led to the presence of B in substitutionalsites.H.Metzner, G.Sulzer, W.Seelinger, B.Ittermann, H.P.Frank, B.Fischer, K.H.Ergezinger,R.Dippel, E.Diehl, H.J.Stöckmann, H.Ackermann: Physical Review B, 1990, 42[17],11419-22

[446-81/82-048]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe effects of implantation damage upon B diffusion were analyzed by using 29Siimplants at doses ranging from 1012 to 1014/cm2. These created controlled amounts ofdamage. Temperatures ranging from 800 to 1000C were used to anneal the implantdamage. For all annealing temperatures, the peak B concentration was well below theintrinsic electron concentration. Significantly enhanced B diffusion was observed, even atdoses as low as 1012/cm2. The greatest enhancement of B diffusion was found for thehighest dose and lowest annealing temperature. The kinetics of damage annealinggoverned the transient enhancement of the B profile.P.A.Packan, J.D.Plummer: Applied Physics Letters, 1990, 56[18], 1787-9

[446-76/77-035]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationReductions in the transient diffusion of 10keV B, which had been implanted along [100] toa dose of 1013/cm2, were observed after annealing (900C, 10s) when the samples wereirradiated with 1MeV 29Si ions to a dose of 5 x 1013/cm2 or more. It was found that a lowerSi dose did not affect the transient B tail-diffusion. Secondary defects formed, near to the

303

B Bulk Diffusion B

peak of the 1MeV Si damage distribution, at doses of 5 x 1013/cm2 and more. These actedas efficient sinks for interstitials from shallower depths, and thereby reduced transient taildiffusion.V.Raineri, R.J.Schreutelkamp, F.W.Saris, K.T.F.Janssen, R.E.Kaim: Applied PhysicsLetters, 1991, 58[9], 922-4

[446-81/82-046]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe anomalous diffusion of ion-implanted B was suppressed by using laser annealing. Inthe case of rapid thermal annealing, high-dose B implantation significantly enhancedanomalous B diffusion. This was attributed mainly to an increased density of interstitialclusters. The laser annealing method was superior in that it promoted dopant activation.M.H.Juang, F.S.Wan, H.W.Liu, K.L.Cheng, H.C.Cheng: Journal of Applied Physics,1992, 71[7], 3628-30

[446-86/87-050]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationIt was found that high-dose B implantation conditions were important with regard to thereduction of transiently enhanced B diffusion. The use of BF2

+ implantation led toexcellent dopant activation and to a reduction, in anomalous diffusion, due to theformation of an amorphous layer and to a scarcity of defects below theamorphous/crystalline interface. On the other hand, B+-implanted crystalline samplesexhibited a poor activation efficiency and markedly anomalous diffusion at hightemperatures. The latter was attributed to the effect of damage which was produced byhigh-dose implantation. Samples of B+-implanted Si+ pre-amorphized material alsoexhibited marked transiently enhanced diffusion, in spite of a good dopant activation.M.H.Juang, F.S.Wan, H.W.Liu, K.L.Cheng, H.C.Cheng: Journal of Applied Physics,1992, 71[6], 2611-4

[446-86/87-050]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationTransiently enhanced diffusivity of implanted B. in crystalline material, in Ge pre-amorphized material, and in epitaxially re-grown material, was studied by usingsecondary ion mass spectrometry. The total Ge dose which was used for amorphizationwas equal to 1.2 x 1015/cm2. It was found that the transiently enhanced diffusivity in re-grown material was equal to 1/3 of the diffusivity in crystalline material. The enhanceddiffusivity in pre-amorphized material was retarded to 70% of the value for crystallinematerial. It was deduced that a moderate Ge implantation dose was sufficient to reducethe depth of the B dopant profile during high-temperature furnace annealing.S.Peterström, B.G.Svensson: Journal of Applied Physics, 1992, 71[3], 1215-8

[446-86/87-051]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe migration of B in pre-amorphized material was studied as a function of the C ionimplantation dose. The 20keV B ions were implanted to a dose of 1015/cm2, afterimplantation with 60 to 90keV C ions to give a pre-amorphized depth of 230nm. It wasshown that transient enhanced diffusion occurred even in the pre-amorphized region

304

B Bulk Diffusion B

without C+ implantation. The diffusion coefficient was larger than normal, by about oneorder of magnitude, during annealing at 1000C for 15s. This enhancement was eliminatedby C+ implantation to a dose of about 1015/cm2. Such implantation also reduced thenumber of defects at the amorphous/crystal interface. It was suggested that the implantedC acted as a sink for excess interstitials.S.Nishikawa, A.Tanaka, T.Yamaji: Applied Physics Letters, 1992, 60[18], 2270-2

[446-86/87-052]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe temporal evolution of B diffusion after ion implantation, and annealing at 800 or900C, was studied by means of secondary ion mass spectrometry and spreading resistanceprofiling. The temporal evolution at 800C was monitored in both crystalline anddevitrified samples. The amorphized samples exhibited near-normal concentrationenhanced diffusion. The crystalline samples exhibited anomalous transient diffusion, witha rapidly diffusing low-concentration region and a static peak region above a criticalconcentration of 3.5 x 1018/cm3. The peak region above this critical concentration waselectrically inactive. The static inactive B was released over many hours, as comparedwith the transient diffusion enhancement; which returned nearly to normal within 0.5h.The temporal evolution of B diffusion at 900C was studied as a function of theimplantation dose. A critical concentration, for transient diffusion, of 8 x 1018/cm2 wasobserved at this temperature; regardless of the dose. The transient enhancement in thediffusing part of the B profile increased with increasing dose up to a value of about 5 x1014/cm3, and saturated at higher doses. The results were explained in terms of a non-equilibrium point-defect model for diffusion which involved intermediate defectformation.N.E.B.Cowern, K.T.F.Janssen, H.F.F.Jos: Journal of Applied Physics, 1990, 68[12],6191-8

[446-86/87-052]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationDislocation-free (100) wafers of n-type material were implanted with 110keV Si ions to adose of 2 x 1015/cm2. An anomalous diffusional behavior of B was observed, and wasattributed to the effect of interstitials which came from the damage tail.X.M.Bao, X.M.Hua: Physica Status Solidi A, 1991, 123[2], K89-93

[446-86/87-052]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe migration of implanted B in pre-amorphized material was monitored by usingsecondary ion mass spectrometry and cross-sectional transmission electron microscopy. Itwas found that 90keV Si implants created an amorphous layer that extended from thesurface to a depth of 186nm; as measured by means of Rutherford back-scatteringspectrometry. The as-implanted B profiles were entirely confined to the amorphous layer.The results indicated that B diffusion in the amorphous layer was suppressed during rapidthermal annealing at temperatures of between 1000 and 1150C.A.J.Walker: Journal of Applied Physics, 1992, 71[4], 2033-5

[446-88/89-050]

305

B Bulk Diffusion B

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationAn investigation was made of the diffusion of B which had been implanted to highconcentrations in pre-amorphized material. Rapid epitaxial re-growth of the amorphouslayer brought B atoms into substitutional positions, up to a threshold concentration ofabout 3.5 x 1020/cm3. This value was almost independent of the annealing temperature andthe implanted dose.S.Solmi, E.Landi, F.Baruffaldi: Journal of Applied Physics, 1990, 68[7], 3250-8

[446-86/87-053]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe effects of low-dose implantation, with Si+, Ne+ and F+, upon the transient enhanceddiffusion of B after annealing (900C, 0.5h) were investigated. The processing conditions,such as the dose (3.5 x 1013/cm2) and energy (30 to 60keV) were chosen so as to simulatethe lightly doped drain implant in an 0.35µ complementary metal-oxide semiconductordevice. An epitaxially grown B-doping superlattice was used to extract directly the depthprofiles of the average Si self-interstitial concentration after processing. In the case of Si+,the transient enhanced diffusion of B increased with the energy of the implanted ion.When Ne+ was implanted by using the same energy as that used for Si+, it caused moretransient enhanced diffusion. When Ne+ was implanted so as to have the same range asSi+, it caused slightly less transient enhanced diffusion. Implantation with F+ enhanced theB diffusivity considerably less than did Si or Ne implantation. These effects weremodelled by using simulations of defect diffusion in the presence of traps. A trapconcentration of 2.4 x 1016/cm3 gave good agreement in all situations except that of F+

implantation; where 6.6 x 1016/cm3 traps were necessary. It was suggested that this wascaused by additional traps, for Si interstitials, that were related to F+.H.H.Vuong, H.J.Gossmann, C.S.Rafferty, H.S.Luftman, F.C.Unterwald, D.C.Jacobson,R.E.Ahrens, T.Boone, P.M.Zeitzoff. Journal of Applied Physics, 1995, 77[7], 3056-60

[446-121/122-085]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationExcellent shallow p+n junctions were formed by implanting BF2

+ ions into thinpolycrystalline films and then annealing them. It was found that anomalous B diffusionoccurred when subsequent silicidation was carried out. Silicidation which was performedby using 30nm of Ti only slightly affected the junction profile. The junction wasconsiderably deepened by 60nm Ti silicidation; yielding a resultant depth of about 0.11µ.The large B redistribution was attributed to the effect of the point defects which wereintroduced by silicidation.M.H.Juang, C.T.Lin, S.T.Jan, H.C.Cheng: Applied Physics Letters, 1993, 63[9], 1267-9

[446-106/107-135]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe diffusion and activation of low-energy implanted B in F pre-amorphized material wasstudied during rapid thermal annealing. When compared with low-energy B or BF2

implantation into crystalline Si, low-energy B ion implantation into F pre-amorphized Sipermitted the formation of shallow junctions with a reduced junction depth and anincreased B activation. The use of F pre-amorphization suppressed transient enhanced

306

B Bulk Diffusion B

diffusion in the low B concentration region and resulted in a steep dopant profile whichwas necessary for shallow junction formation. Secondary ion mass spectroscopy andcross-sectional transmission electron micrography revealed the occurrence of Faccumulation near to the surface, and at end-of-range defects. The interaction of F withdefects was believed to reduce B diffusion in the low B concentration region. Low-energyB implantation into F pre-amorphized material, followed by rapid thermal annealing, wasconcluded to be a promising process for shallow junction formation.T.H.Huang, H.Kinoshita, D.L.Kwong: Applied Physics Letters, 1994, 65[14], 1829-31

[446-119/120-222]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationIn Si/SiGe/Si heterojunction bipolar transistor structures, very shallow As implantation ofthe emitter can cause anomalous B diffusion to occur in the base. It was found that, byusing PtSi contacts, the anomalous diffusion in the base was appreciably reduced.D.X.Xu, C.J.Peters, J.P.Noël, S.J.Rolfe, N.G.Tarr: Applied Physics Letters, 1994, 64[24],3270-2

[446-115/116-185]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationSamples of n-type material were implanted with B+ and As+ ions having energies such thattheir projected ranges (30nm) coincided. The As+ ions were implanted to doses ofbetween 5 x 1014 and 2 x 1015/cm2, and B+ ions were then implanted to a dose of 1015/cm2.It was found that material which had been implanted with As+ doses that were greaterthan 2 x 1015/cm2, and with B+ ions to 1015/cm2, was n-type after annealing at 950C. Avery shallow p-type layer was formed after annealing at 1000C. A reduction indiffusivity, and a decrease in the activation efficiency of the implanted B, increased withincreasing As+ dose.K.Yokota, T.Nakamura, F.Miyashita, K.Hirai, H.Takano, M.Kumagai, Y.Ando,K.Matsuda: Materials Science Forum, 1995, 196-201, 1637-42

[446-127/128-154]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe suppression of dislocation formation and of transient B diffusion, by C co-implantation, was studied by means of transmission electron microscopy, secondary ionmass spectrometry, photoluminescence spectroscopy and high-resolution X-raydiffraction techniques. It was shown that both effects were due to the formation of C-related damage which acted as traps for Si interstitials. Quantitative simulations indicatedthat this damage was probably caused by the co-precipitation of Si and C atoms as Si1.15Ccomplexes. The latter also degraded the electrical properties of the implanted layers. Theydissolved at annealing temperatures that were higher than 900C. When this occurred, theeffect of the C was reduced. Dislocations, transient B diffusion and recovery of theelectrical properties were then observed.A.Cacciato, J.G.E.Klappe, N.E.B.Cowern, W.Vandervost, L.P.Biró, J.S.Custer,F.W.Saris: Journal of Applied Physics, 1996, 79[5], 2314-25

[446-131/132-183]

307

B Bulk Diffusion B

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe effect of the implanted species and dose upon the transient enhanced diffusion of a Bmarker layer was studied. It was found that, for lower implantation doses, the transientenhanced diffusion was species-independent. However, at higher implantation doses, thedependence of transient enhanced diffusion upon the species type became very marked. Itwas found that, at these higher doses (which included amorphization effects), theimplantation of P caused greater transient enhanced diffusion than did implanted Si or As.This was explained on the basis of the fully-coupled diffusion mechanism for dopedimpurities, in which dopants diffused by temporarily pairing with point defects. Inaddition, both point defect clusters (and extended defects such as dislocations)significantly affected the evolution of both the implanted profile and of the buried markerlayer. By modelling these effects, the experimental results were simulated and aconsistent parameter set was found which fitted the data to a reasonable extent.H.S.Chao, S.W.Crowder, P.B.Griffin, J.D.Plummer: Journal of Applied Physics, 1996,79[5], 2352-63

[446-131/132-182]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationA B-doped superlattice which consisted of three B spikes, separated by 170nm ofundoped Si, was grown by means of molecular beam epitaxy and was used to studychanges in point defects following Si implantation. After molecular beam epitaxialgrowth, the wafer was implanted with 146keV or 292keV Si+ to a dose of 5 x 1015/cm2 at77K. This implantation produced amorphous layers with depths that coincided with thoseof the middle B peak, or just below the deepest B peak. The samples were then annealedat 800C in Ar. Secondary-ion mass spectrometry and transmission electron microscopywere used to monitor the diffusion of the B spikes and the evolution of the extendeddefects during annealing. At lower implantation energies, an enhancement of the Bdiffusivity by over 500 times was observed for both the surface B spike and for thedeepest B spike. At higher implantation energies, the results showed conclusively that theback-flow of interstitials into the re-grown region arose from end-of-range damage, justbelow the amorphous/crystalline interface. It was concluded that, for these implantationconditions, the end-of-range damage did not act as a barrier to the flow of interstitials tothe surface. In addition, it was noted that B in re-grown Si did not cluster; whereas the Bbelow the amorphous/crystalline interface did do so.K.S.Jones, R.G.Elliman, M.M.Petravic, P.Kringhøj: Applied Physics Letters, 1996,68[22], 3111-3

[446-134/135-160]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationA novel method for suppressing B penetration was reported. The co-implantation of Sband BF2 into metal-oxide semiconductors was found to diminish the B penetration whichwas enhanced by the presence of F. The B penetration in BF2-implanted devices was seento be significantly reduced with increasing dose of co-implanted Sb. Fourier transforminfra-red spectroscopic and X-ray photoelectron spectroscopic analyses suggested that thesuppression of B penetration was perhaps due to the formation of an Sb-F compound. Thelatter would reduce F-enhanced B diffusion.

308

B Bulk Diffusion B

W.T.Sun, C.S.Yang, M.C.Liaw, C.C.H.Hsu: Japanese Journal of Applied Physics 2,1996, 35[3B], L377-9

[446-136/137-127]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe local atomic environment of Sb dopants, in samples which had been implanted to 2 x1016 or 5 x 1016/cm2, was studied by using near-grazing incidence fluorescence extendedX-ray absorption fine structure techniques at various stages of the Sb deactivationprocess. Annealing was performed, at temperatures of between 900 and 1000C, for timesof 30s to 4h. A comparison of Sb and B co-diffusion data, with corresponding results forthe diffusion of Sb alone, revealed several anomalous effects that were due to dopantinteraction. A simulation program which took account of dopant precipitation and donor-acceptor pairing permitted the prediction of most of the anomalous phenomena thatoccurred in high-concentration co-diffusion experiments.C.Revenant-Brizard, J.R.Regnard, S.Solmi, A.Armigliato, S.Valmorri, C.Cellini,F.Romanato: Journal of Applied Physics, 1996, 79[12], 9037-42

[446-136/137-128]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe effect of N co-implantation in suppressing B penetration into p+-type polycrystallinematerial was investigated. The co-implantation of N with BF2

+ led to combination with Bso as to form B-N complexes. This resulted in a retardation of the B diffusion.T.S.Chao, M.C.Liaw, C.H.Chu, C.Y.Chang, C.H.Chien, C.P.Hao, T.F.Lei: AppliedPhysics Letters, 1996, 69[12], 1781-2

[446-138/139-102]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationIt was shown that ion beam defect engineering could be used for the reduction ofsecondary defects that were created by dopant ion implantation, for the gettering of Fatoms from B-doped regions in BF2-implanted Si, and for reducing B diffusion in BF2-implanted Si.W.Zhonglie, Z.Qingtai, W.Keming, S.Borong: Nuclear Instruments and Methods inPhysics Research B, 1996, 115[1-4], 421-9

[446-141/142-113]

Bulk Diffusion - Qualitative Observations - Effect of OxidationThe diffusion of B under extrinsic conditions was investigated in both oxidizing and inertatmospheres at temperatures of between 950 and 1100C. Oxidation-enhanced diffusionwas found. This result contrasted with the oxidation-retarded diffusion of As.Y.Ishikawa, I.Nakamichi, S.Matsumoto, T.Niimi: Japanese Journal of Applied Physics,1987, 26[9], 1602-3

[446-55/56-035]

Bulk Diffusion - Qualitative Observations - Effect of OxidationThe migration of B during surface oxidation in dry O was measured at 900 to 1100C, fortimes of between 600s and 4h. The enhancement of B diffusion decreased with increasing

309

B Bulk Diffusion B

temperature. An increase in the oxidation time at a given temperature led to a decrease inthe diffusion enhancement. The fraction of B diffusion which occurred via the interstitialmechanism was estimated to be 0.8 at 1100C.P.A.Packan, J.D.Plummer: Journal of Applied Physics, 1990, 68[8], 4327-9

[446-86/87-051]

Bulk Diffusion - Qualitative Observations - Effect of OxidationThe effect of HCl upon the oxidation-enhanced diffusion of B at temperatures of 1000,1100, and 1150C was investigated by using secondary ion mass spectrometry. The dataobtained were used to determine the parameters of a model which had previously beendeveloped in order to describe the retardation of oxidation-enhanced diffusion. Goodagreement was found between theory and experiment.R.Subrahmanyan, H.Z.Massoud, R.B.Fair: Journal of Applied Physics, 1987, 61[10],4804-7

[446-61-084]

Bulk Diffusion - Qualitative Observations - Effect of OxidationBuried B layers were grown epitaxially onto monocrystalline Si substrates and weresubjected to steam oxidation at 650 to 750C, under pressures of 1, 5, or 15atm. The layerswere approximately 200nm-thick and were capped with 400nm of undoped Si. The Bconcentration varied from 8 x 1017 to 4 x 1018/cm3. A resultant enhanced B diffusion wasmodelled by assuming that the oxidation maintained a supersaturation of interstitials, atthe surface, which was proportional to the square root of the oxidation rate. It was foundthat the use of fully-coupled dopant-defect diffusion equations was necessary in order tomodel the oxidation-enhanced diffusion accurately.R.F.Lever, P.B.Griffin, W.A.Rausch: Journal of Applied Physics, 1995, 78[5], 3115-20

[446-123/124-182]

Bulk Diffusion - Qualitative Observations - Effect of OxidationLayers of Si-B were used as a new B diffusion source for polysilicon/silicon systems. Thelayer was deposited onto polysilicon in an ultrahigh-vacuum chemical vapor depositionsystem at 550C. The characteristics of B diffusion in Si-B/polysilicon/silicon systemswere investigated by using secondary ion mass spectroscopy and cross-sectionaltransmission electron microscopy. In order to remove the Si-B layer, after the drive-instep, the Si-B layer was completely oxidized during the drive-in stage and was removedusing HF. It was found that the B profiles within the polysilicon were slightly dependentupon the oxidation of the Si-B layer. The polysilicon grain size, after using a Si-B layersource, was greater than that after using conventional BF2

+-implanted polysilicon sources.This was attributed to the effects of O gettering by the Si-B layer, and to secondary graingrowth during Si-B layer oxidation. The B diffusion profiles in the Si substrate, afterusing a Si-B layer source, also exhibited a shallower junction depth and a lowersensitivity to the thermal budget; as compared with a BF2

+-implanted polysilicon source.This was attributed to the effect of a smaller surface concentration in the Si substrate, inthe case of a Si-B layer source.

310

B Bulk Diffusion B

T.P.Chen, T.F.Lei, H.C.Lin, C.Y.Chang: Journal of the Electrochemical Society, 1995,142[2], 532-7

[446-134/135-157]

Bulk Diffusion - Qualitative Observations - Effect of RadiationRadiotracer, secondary ion mass spectrometry, and electrical methods were used to studythe radiation-stimulated diffusion of B. It was found that acceleration of the diffusion byelectron irradiation was entirely due to the generation of excess vacancies and was notdue to ionization processes.J.N.Kazarinov, V.V.Kozlovskii, V.N.Lomasov, M.V.Pitkevich: Fizika i TekhnikaPoluprovodnikov, 1986, 20[9], 1577-81. (Soviet Physics -Semiconductors, 1986, 20[9],989-92)

[446-51/52-129]

Bulk Diffusion - Qualitative Observations - Effect of StressAn investigation was made, of the interface of Si/Si directly bonded wafers, by using X-ray topographic, spreading resistance, and secondary ion mass spectroscopic methods.The residual stress which was induced at the interface of the bonded wafer was revealedas ring-like patterns by X-ray topography when the 2 wafers, flat and convex, werebonded. It appeared that the residual stress enhanced the diffusion of B atoms and causedthem to aggregate near to the bonded interface. Self-interstitials which were ejectedduring the formation of an amorphous SiOx (where x was between 0.2 and 0.3) layer atthe bonded interface led to an increased diffusion length of B on the substrate side (whichwas under tensile stress) than in the active layer. The resistivity increased in the regionunder compressive stress; in spite of the aggregation of B.S.Ishigami, Y.Kawai, H.Furuya, T.Shingyouji: Japanese Journal of Applied Physics 1,1993, 32[10], 4408-12

[446-115/116-151]

Bulk Diffusion - Qualitative Observations - Effect of StressSamples of in situ doped Si were subjected to inert-gas furnace annealing at 800C. Theeffect of biaxial tension was investigated by using relaxed SiGe layers as substratetemplates for epitaxial Si layers. The B diffusivity did not depend strongly upon thebiaxial strain.P.Kuo, J.L.Hoyt, J.F.Gibbons, J.E.Turner, D.Lefforge: Applied Physics Letters, 1995,66[5], 580-2

[446-121/122-084]

Bulk Diffusion - Qualitative Observations - Effect of StressThe effect of stresses, in Si3N4 films, upon B diffusion was studied. It was noted thatduring annealing in N, the degree of retardation of B diffusion became larger as thetemperature decreased and the thickness of the nitride film increased. On the other hand,B diffusion in Si which was covered with a very thin nitride film was enhanced duringlong diffusion times at 1014C. These results suggested that nitride films compressed theregion near to the substrate surface and altered the vacancy and interstitial concentrations.

311

B Bulk Diffusion B

K.Osada, Y.Zaitsu, S.Matsumoto, M.Yoshida, E.Arai, T.Abe: Journal of theElectrochemical Society, 1995, 142[1], 202-6

[446-134/135-160]

Bulk Diffusion - Qualitative Observations - Effect of Surface LayerThe simultaneous diffusion of Au and B was studied at temperatures of between 800 and1200C by using spreading resistance and Rutherford back-scattering spectrometricmethods. An enhanced B diffusivity was detected when a thick layer existed on the Sisurface. In general, B diffusion was enhanced and driven deeper with increasing Aucontent. The enhancement was attributed to a form of Kirkendall effect.D.K.An, K.Madl, A.Barna, G.Battistig, J.Gyulai: Physica Status Solidi A, 1989, 116[2],561-9

[446-76/77-033]

Bulk Diffusion - Qualitative Observations - Effect of Surface LayerThe migration of B from implantation-doped polycrystalline Si films and intomonocrystalline Si was investigated as a function of various process parameters. Theeffects of interface treatment before polycrystalline Si deposition, and of the Si grain size,were analyzed. Well-behaved diffusion sources were obtained only if epitaxialrealignment of the polycrystalline Si film to the substrate was avoided and if no diffusionbarrier was present at the polycrystalline/monocrystalline interface.V.Probst, H.J.Böhm, H.Schaber, H.Oppolzer, I.Weitzel: Journal of the ElectrochemicalSociety, 1988, 135[3], 671-6

[446-60-012]

Bulk Diffusion - Qualitative Observations - Effect of Surface LayerA comparison was made of B diffusivity, during oxidation, in pure Si and in Si which wascovered with a very thin layer of SiGe. The profiles were determined by using secondaryion mass spectrometry. It was found that a thin layer of the alloy suppressed oxidation-enhanced B diffusion. A mechanism was suggested which was based upon thesuppression of Si interstitials. This effect appeared to explain previously reported fasterrates of oxidation of bulk SiGe and of SiGe-covered Si samples.F.K.LeGoues, R.Rosenberg, B.S.Meyerson: Applied Physics Letters, 1989, 54[8], 751-3

[446-64/65-187]

Bulk Diffusion - Qualitative Observations - Effect of Surface LayerA detailed study was made of attempts to suppress the penetration of B into p-type metaloxide semiconductor field-effect transistors. Four types of polycrystalline Si gatestructure and 2 types of gate dielectric were considered for the suppression of Bpenetration. A stacked amorphous Si structure was found to be the most effective meansof retarding B penetration. The use of N2O led to a superior retardation of B diffusion. Itwas found that the use of a combination of stacked amorphous Si, and N2O, was the mosteffective method for preventing B penetration.T.S.Chao, C.H.Chu, C.F.Wang, K.J.Ho, T.F.Lei, C.L.Lee: Japanese Journal of AppliedPhysics 1, 1996, 35[12A], 6003-7

[446-148/149-185]

312

B Bulk Diffusion B

Bulk Diffusion - Qualitative Observations - Effect of Surface LayerThe effects of Si3N4 films upon the anomalous enhanced diffusion of B, and uponextended defect formation, were studied in float-zone material during post-implantationannealing. Under the films, the diffusion of B was retarded and stacking faults did notform after annealing. The results suggested that stress in the films decreased the self-interstitial concentration in Si near to the interface.Y.Zaitsu, K.Osada, T.Shimizu, S.Matsumoto, M.Yoshida, E.Arai, T.Abe: MaterialsScience Forum, 1995, 196-201, 1891-6

[446-127/128-155]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsThe diffusion of B was studied by means of plan-view and cross-sectional transmissionelectron microscopy. This showed that misfit dislocation half-loops were initiallygenerated at the surface. The Burgers vector of the dislocation half-loop was inclined withrespect to the surface, and the initial misfit dislocations were therefore not very efficientwith regard to strain relaxation. As diffusion proceeded, non-parallel dislocationsinteracted and gave rise to product segments that had in-plane Burgers vectors whichwere parallel to the surface. On the basis of the observations, a model was presented inorder to clarify the details of these interactions; and the formation or more efficient misfitdislocations from less-efficient inclined ones.X.J.Ning, P.Pirouz: Acta Materialia, 1996, 44[5], 2127-43

[446-134/135-159]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsThe misfit dislocations which were introduced into (001), (111) or (110) wafers by Bdiffusion were found to nucleate from the diffusion surface; as half-loops with Burgersvectors that were inclined with respect to the surface. The expansion of the half-loop onits glide plane left a 60° misfit dislocation in the diffusion front. The numerousdislocation reaction products which were the result of interactions among the half-loopsduring expansion were characterized by using cross-sectional and plan-view transmissionelectron microscopy. It was found that dislocations with a density of the order of 108/cm2

were distributed mainly in a region that was a certain depth below the surface; thusleaving the top region essentially dislocation free. The misfit dislocations that nucleatedfrom the surface were insufficient to relax all of the strain in the diffused layer.X.J.Ning: Philosophical Magazine A, 1997, 75[1], 115-35

[446-141/142-114]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionLow-energy and low-dose B+-implanted material was studied by using transmissionelectron microscopy and secondary ion mass spectrometry. Czochralski-type (100) waferswere implanted with 4keV B+ to a dose of 1014/cm2. Annealing was carried out attemperatures of between 700 and 800C, for times of between 15s and 8h, in a Natmosphere. The secondary ion mass spectrometry results revealed a transient enhanced Bdiffusion which saturated in less than 0.25h at all of the annealing temperatures that wereused. Transient enhanced diffusion resulted in an increase, in the junction depth, by atleast 60nm at a concentration of 1016/cm3. Transmission electron microscopic studies

313

B Bulk Diffusion B

showed that, even for the shortest times before transient enhanced diffusion wasobserved, 311 defects were not detected. It was concluded that there might be morethan one source of interstitials for transient enhanced diffusion.L.H.Zhang, K.S.Jones, P.H.Chi, D.S.Simons: Applied Physics Letters, 1995, 67[14],2025-7

[446-125/126-145]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe transient enhanced diffusion of B in pre-amorphized and re-grown material wasstudied by using secondary ion mass spectrometric and transmission electron microscopictechniques. A comparison was made of 4keV B implantation (1014/cm2) into crystallineand Ge+ pre-amorphized samples. Upon annealing, the B-implanted crystalline materialexhibited the familiar transient enhanced diffusion. In this case, the peak of the Bdistribution was relatively immobile and only the B in the tail exhibited transientenhanced diffusion. In the case of the other samples, the surface was first pre-amorphizedby implantation with 180keV Ge+ (1015/cm2). This produced an amorphous layer that was230nm deep and which was then implanted with B. After implantation, the tail of the Bdistribution extended to only 70nm. Upon annealing, transient enhanced diffusion of B inthe re-grown material was also observed, but the diffusion profile was very different. Inthis case, the peak exhibited no clustering and the entire profile diffused. The time whichwas required for the transient enhanced diffusion to decay was equal to about 0.25h at800C. The transmission electron microscopic results indicated that (311) defects in theend-of-range damage stopped dissolving within 600s to 1h at 800C. These resultsindicated that, under these pre-amorphization conditions, not only did the end-of-rangedefects not block the flow of interstitials into the re-grown material, the (311) defects inthe end-of-range damage acted as sources of interstitials. In addition, B did not appear tocluster in re-grown material.K.S.Jones, L.H.Zhang, V.Krishnamoorthy, M.Law, D.S.Simons, P.Chi, L.Rubin,R.G.Elliman: Applied Physics Letters, 1996, 68[19], 2672-4

[446-134/135-159]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionWafer samples were pre-amorphized by implanting Si+ or Ge+ ions at temperatures ofbetween 5 and 40C. The diffusion of 4keV B+ implants into the pre-amorphized materialwas used to monitor the flow of interstitials from the end-of-range region and towards thesurface. Transient enhanced diffusion in re-grown material was observed for all of theimplantation conditions. It was noted that increasing the temperature for Si+ implantationby as little as 15C could produce a marked decrease in the magnitude of the interstitialflux which passed from the end-of-range region and towards the surface. This sensitivityto the temperature appeared to be even greater for Ge+ implantation. As-implanted cross-sectional transmission electron micrographs revealed a measurable decrease in thethickness of the amorphous layer (up to 30nm) when the implantation temperature wasincreased from 5 to 40C, as a result of ion beam-induced epitaxial recrystallization. Uponannealing at 800C, two types of defect were observed in the end-of-range region. Thesewere 311 defects and dislocation loops. The 311 defects were unstable, and acomparison of secondary ion mass spectroscopic and transmission electron microscopic

314

B Bulk Diffusion B

data for annealed samples indicated that the dissolution of these 311 defects constitutedat least one source of interstitials for transient enhanced diffusion in re-grown material at800C. The end-of-range dislocation loops were stable under the annealing conditions(800C, 0.25h) which were used. There appeared to be an exponential dependence oftransient enhanced diffusion, in re-grown material, upon the density of end-of-rangedislocation loops.K.S.Jones, K.Moller, J.Chen, M.Puga-Lambers, B.Freer, J.Berstein, L.Rubin: Journal ofApplied Physics, 1997, 81[9], 6051-5

[446-150/151-151]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionA study was made of the transient enhanced diffusion which resulted from implantationdamage, and which led to more rapid dopant migration. Experiments were performed byusing B and dislocation markers to compare transient enhanced diffusion effects. Theresults showed that P diffusion was enhanced considerably more than B diffusion duringdamage annealing. Dislocation growth indicated that a number of interstitials that wasgreater than the damage dose was captured during annealing. It was found that the timethat was required in order to saturate dislocation growth agreed well with P diffusionsaturation, and was greater than B saturation.J.Xu, V.Krishnamoorthy, K.S.Jones, M.E.Law: Journal of Applied Physics, 1997, 81[1],107-11

[446-141/142-114]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionTransient enhanced B diffusion was observed during the annealing of amorphized wafers.This anomalous diffusion originated from interactions between end-of-range defects anddopants. Dopant trapping also occurred at dislocation loops. During annealing, thesedefects grew in size and reduced their density via the emission and capture of Siinterstitial atoms. This behavior could be explained in terms of Ostwald ripening. The Bdiffusivity enhancement that was experimentally observed arose from a largesupersaturation of Si interstitials in the defect-rich region and from a strong couplingbetween B atoms and Si interstitials. Both phenomena were transient.C.Bonafos, A.Martinez, M.M.Faye, C.Bergaud, D.Mathiot, A.Claverie: NuclearInstruments and Methods in Physics Research B, 1995, 106[1-4], 222-6

[446-136/137-128]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionIt was noted that implanted B dopants exhibited a transient enhanced diffusion, duringannealing, which was due to the excess interstitials which were generated byimplantation. Transmission electron microscopic measurements of the implantationdamage were combined with B diffusion experiments; using dopant marker structures thathad been grown by molecular beam epitaxy. It was found that the damage which resultedfrom non-amorphizing Si implants, for doses that ranged from 5 x 1012 to 1014/cm2,evolved into a distribution of 311 interstitial agglomerates during initial annealing at670 to 815C. The excess interstitial concentrations which were contained in these defectswere roughly equal to the implanted ion doses. This was confirmed by atomistic Monte

315

B Bulk Diffusion B

Carlo simulations of the implantation and annealing processes. The injection ofinterstitials from the damaged regions involved the dissolution of 311 defects, duringOstwald ripening, with an activation energy of 3.8eV. The excess interstitials drovesubstitutional B into electrically inactive metastable clusters of two or three B atoms, atconcentrations below the solid solubility. This explained the immobile B peak which wasusually observed during the transient enhanced diffusion of implanted B. Injectedinterstitials underwent retarded diffusion, in molecular beam epitaxially grown material,with an effective migration energy of about 3.5eV. This arose from trapping atsubstitutional C. It was suggested that the concept of trap-limited diffusion could help toexplain the enormous observed disparity among published values of interstitial diffusivity.The population of excess interstitials was strongly reduced by incorporating substitutionalC to levels of about 1019/cm3 before implantation.P.A.Stolk, H.J.Gossmann, D.J.Eaglesham, D.C.Jacobson, C.S.Rafferty, G.H.Gilmer,M.Jaraíz, J.M.Poate, H.S.Luftman, T.E.Haynes: Journal of Applied Physics, 1997, 81[9],6031-50

[446-150/151-152]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe transient enhanced diffusion of dopants during short annealing treatments wasinvestigated by using a buried B marker layer structure and various Si implantation dosesand energies. It was found that the diffusion behavior of the marker layer indicated thatdiffusivity enhancement was essentially independent of the implantation conditions, forshort annealing times, while the overall transient motion increased with increasingimplantation. The data were explained by using an interstitial clustering model that tookaccount of both cluster evaporation and cluster growth.H.S.Chao, P.B.Griffin, J.D.Plummer, C.S.Rafferty: Applied Physics Letters, 1996,69[14], 2113-5

[446-138/139-102]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe effect of P pre-doping upon the transient enhanced diffusion of implanted B wasstudied by means of secondary ion mass spectroscopy. The 40keV B ions were implantedinto (100) samples to a dose of 3 x 1014/cm2. These samples had already been uniformlypre-doped with P to concentrations that ranged from 3 x 1019 to 1020/cm3. The effective Bdiffusivity during transient enhanced diffusion, and the fraction of immobile B, werededuced from secondary ion mass spectrometric profiles. The results showed that both ofthem decreased with increasing P doping level, and saturated at a P doping level of 6 x1019/cm3 after low-temperature annealing.M.B.Huang, T.W.Simpson, I.V.Mitchell: Applied Physics Letters, 1997, 70[9], 1146-8

[446-148/149-185]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe transient enhanced diffusion of shallow molecular beam epitaxially grown markerlayers of B, resulting from deep MeV Si+ ion implantation to doses of about 1016/cm2, wasmeasured. It was expected that the near-surface regions of these implanted samples would

316

B Bulk Diffusion B

be vacancy-rich. The shallow B marker layers exhibited transient enhanced diffusion.This implied that an interstitial supersaturation was present.D.J.Eaglesham, T.E.Haynes, H.J.Gossmann, D.C.Jacobson, P.A.Stolk, J.M.Poate:Applied Physics Letters, 1997, 70[24], 3281-3

[446-152-0316]

Bulk Diffusion - Qualitative Observations - Transient EffectsThe anomalous diffusion of ion-implanted B was shown to be a transient effect with adecay time which decreased rapidly with increasing annealing temperature. The decaytime was approximately 0.75h at 800C and decreased to the order of seconds at 1000C.The anomalous displacement in the low concentration region was greater at lowtemperatures but a larger fraction of the B was redistributed at high temperatures. Sheetresistance measurements agreed with the idea that the moving fraction of B atoms waselectrically active and limited to the intrinsic carrier concentration at the annealingtemperature. The activation energy for the decay of the transient was greater than that forthe diffusion coefficient.A.E.Michel, W.Rausch, P.A.Ronsheim, R.H.Kastl: Applied Physics Letters, 1987, 50[7],416-8

[446-51/52-134]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesAn analysis was made of discrepancies concerning the diffusivity of B. In the case of aninert atmosphere and a BN source, an independence of the diffusivity with respect to theB concentration was found. Using a non-Fickian model based upon Si exodiffusion, ananalytical expression (ratio of 2 erfc terms) was obtained for the diffusion profile.E.Dominguez, M.Jaraiz: Journal of the Electrochemical Society, 1986, 133[9], 1895-900

[446-51/52-134]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesA diffusion potential was proposed for 2 different materials and their interface. Aninterfacial impurity flux model was derived by using this potential. The flux model wasrelated to measurable physical parameters; such as diffusion coefficients in the 2materials, and segregation coefficients. This model also included the transport segregationcoefficient, which had to be determined experimentally. This theory was applied to theprofiles of B which was diffused from polysilicon into Si. The concentration discontinuityat the polysilicon/Si interface appeared clearly in time. This was explained by the presentinterfacial impurity flux model. The segregation transport coefficient was deduced to be 5x 10-6cm/s.K.Suzuki, T.Fukano: Journal of the Electrochemical Society, 1991, 138[1], 230-3

[446-78/79-050]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesA model, for B diffusion in poly-Si structures, was presented. This took account of grainboundary movement, impurity segregation to grain boundaries, differing diffusivities inthe grain and in the grain boundary, the effect of interface-oxide thickness, and high-concentration effects upon the B profile. It was found that the model gave goodagreement

317

B Bulk Diffusion B

with experiment. A drawback was a general ignorance of data on temporal and spatialvariations in grain size.A.D.Sadovnikov: Solid-State Electronics, 1991, 34[9], 969-75

[446-84/85-061]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesThe effect of dopant-dopant interactions upon diffusion was modelled firstly at thequantum-chemical level. The relationship between atomic-scale results and macroscopicbehavior was then obtained by assuming that interactions between molecular orbitalscould be transmitted over hundreds of lattice spacings. Additional flux components weregenerated which modified the usual form of Fick's second law.S.Aronowitz: Journal of Applied Physics, 1991, 69[7], 3901-6

[446-86/87-051]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesA model was developed for the description of B diffusion in poly-Si/Si structures. It tookaccount of grain boundary motion, segregation of impurities to grain boundaries, differingdiffusivities in the grains and the grain boundaries, and the effect of the interfacial oxidelayer thickness upon the two-dimensional diffusion profile.A.D.Sadovnikov, A.V.Tchernyaev: Solid-State Electronics, 1992, 35[2], 193-200

[446-88/89-050]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesA macroscopic model for dopant diffusion, involving dopant-dopant interactions, wasextended so as to include 3-dimensional effects. The model was applied to dopantdiffusion in thin crystalline films under conditions where a second dopant was present inhigh concentrations and was distributed uniformly throughout the film. It was found thatthis resulted in an unexpected thickness dependence of the behavior in very thin films.S.Aronowitz: Journal of Applied Physics, 1991, 70[11], 6815-20

[446-91/92-029]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesA Monte Carlo study was made of B diffusion. The kick-out mechanism was consideredfor the case of a delta-function impurity profile under both inert and oxidation conditions.It was shown that the initial conditions played an important role in determining the meanmigration path length of the atoms. The kick-out mechanism was examined analyticallyfor the case of an initial delta-function interstitial impurity profile. The present atomic-level computations validated previous results on so-called intermittent B diffusion in Si.The pre-factor for the mean migration path-length was found to lie between 0.024 and0.035nm.M.M.De Souza, G.A.J.Amaratunga: Journal of Applied Physics, 1996, 79[5], 2418-25

[446-131/132-183]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesIn the case of B diffusion, the value of the sheet resistance that is observed on plain checkslices is almost always appreciably smaller than the sheet resistance that is deduced fromdiffused resistors. The actual value of the discrepancy depended upon the size of the

318

B Bulk Diffusion B

window and the surrounding masking oxide geometry. These facts were further importantfactors in the 2-dimensional simulation of diffusion processes. The results were presentedhere of a series of experiments which involved the use of multi-layered mask structuresthat consisted of a thermal oxide, Si3N4, and SiO2. On the basis of the results, it wassuggested that most of the observations could be explained with the help of a theoreticalmodel that was based upon the surface diffusion of B over Si and SiO2. The modelassumed a high solubility of B in the oxide, and a high surface diffusion coefficient.Numerical calculations demonstrated that the model could explain most experimentalobservations.S.A.Abbasi, F.Rahman: Journal of the Electrochemical Society, 1995, 142[11], 3928-32

[446-134/135-157]

Bulk Diffusion - Theoretical Analysis - Defect InteractionExperiments show that ion pairing has a marked effect upon the diffusion of oppositelycharged impurities. An analysis of literature data was used to deduce the ion pairingcoefficients for n-type impurities with B and In. A coefficient with the value of 0.17/Nwas found to describe the pairing case of B-P, where N was the intrinsic electronconcentration. It was suggested that the paired ions occupied adjacent substitutional sites;with a small perturbation in the Coulomb binding which arose from elastic effects.N.E.B.Cowern. Applied Physics Letters, 1989, 54[8], 703-5

[446-64/65-176]

Bulk Diffusion - Theoretical Analysis - Defect InteractionThe variation in the diffusivity as a function of oxidation time was studied under extrinsicconditions. A model was developed which satisfactorily explained the variations. Themodel assumed that recombination between self-interstitials and vacancies took place,thus reducing their numbers, and their contribution to B diffusion. It was concluded thatthe model could fit the results by using only the interstitialcy diffusion component as aparameter. The best fit was obtained when the latter was equal to 0.4.D.Tsoukalas: Journal of Applied Physics, 1991, 70[12], 7309-14

[446-91/92-029]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsThe ab initio pseudopotential method was used to study B diffusion and pairing incrystalline samples. The results showed that substitutional B attracted interstitial Si with abinding energy of 1.1eV. It was shown that B diffusion was significantly enhanced, in thepresence of the Si interstitial, due to a substantial lowering of the migrational barrier;arising probably from a kick-out mechanism. The resultant mobile B could also betrapped by another substitutional B with a binding energy of 1.8eV. This led to theformation of an immobile and electrically inactive 2-B pair along a <001> direction. Itwas also found that the pairing of these two B atoms involved the trapping of a Siinterstitial. On the other hand, two B pairs that did not trap a Si interstitial were found tobe energetically unfavorable. All of the results were consistent with experimental results.J.Zhu, T.Diaz de la Rubia, L.H.Yang, C.Mailhiot, G.H.Gilmer: Physical Review B, 1996,54[7], 4741-7

[446-141/142-114]

319

B Bulk Diffusion B

Bulk Diffusion - Theoretical Analysis - Effect of DefectsDopant and self-diffusion were known to be governed by both self-interstitials andvacancies; as normalized to their thermal equilibrium values. Since these point defectconcentrations were too low to be investigated directly, the diffusion mechanisms had tobe considered to be unknown. After substituting experimental results on B diffusion at1100C into a new equation that had been derived from the dopant diffusion equation, theequation was solved at the same time as the equation for oxidation-induced stacking faultsunder conditions of local equilibrium between interstitials and vacancies. In this way, theinterstitial and vacancy concentrations were obtained as functions of the diffusion time,and the fractional components of the interstitialcy mechanism for B diffusion, could bedetermined.T.Okino, R.Takaue, M.Onishi: Materials Science Forum, 1995, 196-201, 1631-6

[446-127/128-155]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA model was proposed for the point defect assisted transient diffusion and activation ofhigh-dose B implants. In order to model transient diffusion, a non-linear equilibriumclustering model for point defects was used. The activation of B was modelled as a Fermilevel-dependent transformation of inactive dopant clusters into substitutional atoms. Acomparison with experimental data showed that this approach could provide a descriptionof both rapid thermal annealing and of long-term furnace annealing; with excellentpredictability for both chemical and electrically active profiles.A.Höfler, T.Feudel, N.Strecker, W.Fichtner, K.H.Stegemann, H.Syhre, G.Dallmann:Journal of Applied Physics, 1995, 78[6], 3671-9

[446-125/126-146]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsDopant migration paths and activation energies for self-interstitial assisted migration wereinvestigated on the basis of first-principles and critical path calculations. The resultsrevealed that B migration could take various alternative routes. The present migrationmechanisms suggested that B diffusion via the interstitialcy mechanism could involve asmall pre-exponential factor, whereas dopant migration via an interstitial mechanismcould involve large pre-exponential factors, as reflected by the experimental data.K.Kato: Journal of Physics - Condensed Matter, 1994, 6[21], L287-92

[446-115/116-151]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA study was made of high-concentration B diffusion by using a precipitation model.Recent experimental data on the annealing of B which had been implanted into pre-amorphized Si gave the opportunity to analyze, with the aid of simulations, theprecipitation kinetics and the effect of dislocation loops which existed at theamorphous/crystalline interface. A non-equilibrium point defect diffusion model was usedwhich explicitly included equations that described the precipitation kinetics. The initialconditions took account of the high level of activation which was observed after solid-phase epitaxy. This affected both the sheet resistance and the doping profile shape at theend of the process. It was also shown that correct modelling of these diffused profiles

320

B Bulk Diffusion B

included the effect of dislocation loops, at the amorphous/crystalline interface, whichacted as sinks for interstitials. It was noted that more classical formulations of thediffusion equation did not require the modelling of such phenomena since an equilibriumconcentration of point defects was implicitly assumed.B.Baccus, E.Vandenbossche, M.Lannoo: Journal of Applied Physics, 1995, 77[11], 5630-41

[446-121/122-085]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsPoint-defect models were compared with experimental data on intrinsically dopedmaterial. Transient dopant diffusion due to low-dose Si implantation damage could bemodelled by using the same parameters as those used for oxidation-enhanced diffusion.This therefore provided an additional technique for monitoring point defect behavior.Consistent parameters were extracted for both experimental conditions, and were fitted toArrhenius relationships. The theory of dopant-defect pairing was found to be crucialwhen modelling implantation damage effects. The effective binding energies for B-defectand P-defect pairs were determined experimentally.H.Park, M.E.Law: Journal of Applied Physics, 1992, 72[8], 3431-9

[446-106/107-137]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsThe motion of a substitutional impurity via a fast-diffusing intermediate species wasconsidered. An analytical solution was given for the case of low impurity concentrations.At short times, only a few atoms migrated and the solution behaved exponentially. Thisexponential signature was observed in the diffusion of nm-scale B-doping profiles. Themigration frequency during oxidation-enhanced diffusion was consistent with diffusion-limited kick-out of an interstitial-type B species. This was the first direct experimentalevidence for dopant diffusion via an intermediate species.N.E.B.Cowern, K.T.F.Janssen, G.F.A.Van de Walle, D.J.Gravesteijn: Physical ReviewLetters, 1990, 65[19], 2434-7

[446-78/79-051]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsAn earlier model for tails in the diffusion profiles of high solute concentrations wasdeveloped. The quasi steady-state approach was extended so as to include both vacanciesand interstitials at equivalent levels. The I-V recombination was assumed to be close tolocal equilibrium and to occur via the reaction of defects with defect-impurity pairs. Themodel predicted characteristic details such as the plateau, kink, and tail which wereobserved in high solute concentration surface diffusion.F.F.Morehead, R.F.Lever: Journal of Applied Physics, 1989, 66[11], 5349-52

[446-74-054]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsNumerical simulations were made of the redistribution of B in Ar-implanted wafers. Theoriginal wafers had been uniformly doped with B. Redistribution of B atoms wasobserved for Ar doses of 1015/cm2, after annealing at 900 or 1000C. It was completely

321

B Bulk Diffusion B

absent for low Ar doses. It was suggested that the excess self-interstitials which weregenerated by recrystallization of the amorphized layer were responsible for theredistribution. The interaction of these excess self-interstitials with B atoms occurred viaa kick-out mechanism in which a large number of mobile B species were produced atinterstitial sites.V.C.Lo, M.X.Pan, S.P.Wong, Y.W.Lam: Modelling and Simulation in Materials Scienceand Engineering, 1996, 4[2], 179-91

[446-138/139-103]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA theory was developed for impurity diffusion under equilibrium and non-equilibriumconcentrations of point defects. The results of first-principles calculations of several keyquantities were combined with the theory and were compared with experimental data. Itwas found that vacancies and self-interstitials governed the equilibrium diffusion of B.Interstitials tended to predominate. It was also found that the direct-exchange mechanismplayed only a minor role.C.S.Nichols, C.G.Van de Walle, S.T.Pantelides: Physical Review Letters, 1989, 62[9],1049-52

[446-64/65-177]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA comprehensive investigation of dopant diffusion in the presence of equilibrium andnon-equilibrium concentrations of intrinsic point defects was presented. It was found that,under equilibrium conditions, vacancies and interstitials mediated the diffusion of thedopant. Relevant expressions were derived for the activation energies, of variousdiffusion and injection mechanisms, under non-equilibrium conditions (such as thoseproduced by the injection of excess point defects). For oxidation conditions, thecalculated values were in excellent agreement with available experimental data. Boththeory and experiment suggested that the concerted exchange mechanism, which involvedno point defects, played only a minor role in dopant diffusion.C.S.Nichols, C.G.Van de Walle, S.T.Pantelides: Physical Review B, 1989, 40[8], 5484-96

[446-72/73-041]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA simplified model was presented for the treatment of dopant diffusion in the presence ofnon-equilibrium point defect concentrations. The dopant flux was expressed in terms ofan effective diffusivity, and took account of the various couplings which arose from thepresence of defect gradients. The point defect concentrations were calculated by resolvingthe corresponding continuity equations. It was found that the model permitted the fast andaccurate simulated diffusion of B.D.Mathiot, S.Martin: Journal of Applied Physics, 1991, 70[6], 3071-80

[446-91/92-028]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsRecent experimental data on B diffusion as a function of substrate concentration, underinert or oxidizing ambients, were interpreted in terms of a dual vacancy and interstitiallymechanism. By also assuming that positively ionized vacancies and interstitials

322

B Bulk Diffusion B

dominated the contribution of the corresponding neutral species, a simple model wasdeveloped which explained well the results obtained when B diffused alone or when itdiffused in heavily P-doped Si. It was concluded that the interstitialcy component waslimited to between 0.21 and 0.43. This revealed an important vacancy contribution to Bdiffusion.D.Tsoukalas, P.Chenevier: Physica Status Solidi A, 1987, 100[2], 461-5

[446-55/56-035]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsTwo-dimensional distributions of implanted B, and implantation-induced displacementdamage near to a mask edge, were used to calculate the two-dimensional redistribution ofB which resulted from typical short-term annealing. The damage was removed, duringannealing, by the release of vacancies which enhanced the diffusion of B. The effect wasthat B preferentially redistributed further into the bulk.J.F.Marchiando, J.Albers: Journal of Applied Physics, 1987, 61[4], 1380-91

[446-60-013]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA new model for impurity diffusion via a point defect plus impurity pair mechanism wasdeveloped. A pair of coupled non-linear partial differential equations for the Si self-interstitial and the impurity was derived and solved numerically. The usual kink and tailof P and, to some extent, the B diffusion profiles arose naturally from the solution. Thecoupling between defect and impurity became smaller at high temperatures and at lowimpurity concentrations; in agreement with experimental observations.B.J.Mulvaney, W.B.Richardson: Applied Physics Letters, 1987, 51[18], 1439-41

[446-60-014]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsModelling of the diffusion of high concentrations of implanted B was carried out. Thenormally accepted model, which assumed that the diffusivity of B was controlled bysingly-charged donor vacancies, was shown to be inadequate for describing B profiles.Instead, a two-stream model (in which B movement was dominated by a dissociativeprocess involving both interstitial and substitutional diffusion) was found to be in goodagreement with the data.O.W.Holland: Applied Physics Letters, 1989, 54[9], 798-800

[446-70/71-122]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA simple method was proposed for the estimation of the fractional interstitialcycomponent, f, of B. It was based upon an interpretation of oxidation-enhanced diffusiondata under extrinsic conditions. The model satisfactorily explained the experimentalresults when the value of f was 0.25. This indicated the existence of an importantcontribution of vacancies to B diffusion.D.Tsoukalas, P.Chenevier: Journal of Applied Physics, 1989, 66[4], 1858-60

[446-74-048]

323

B Bulk Diffusion B

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA new model was proposed, for the diffusion of impurities in amorphous Si, which tookaccount of its structural and electronic properties. The model was based upon the many-body kinetic theory of thermally activated rate processes in solids. The low activationenergies which were commonly observed, as well as their dependence upon the impurityconcentration, were explained.J.L.Khait, R.Brener, R.Beserman: Physical Review B, 1988, 38[9], 6107-12

[446-62/63-228]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsAn earlier model for tails in the diffusion profiles of high concentrations of B wasdeveloped. The quasi steady-state approach was extended so as to include both vacanciesand interstitials at equivalent levels. The I-V recombination was assumed to be close tolocal equilibrium and to occur via the reaction of defects with defect-impurity pairs. Themodel predicted characteristic details such as the plateau, kink, and the tails which werefound in the case of B diffusion.F.F.Morehead, R.F.Lever: Journal of Applied Physics, 1989, 66[11], 5349-52

[446-74-054]

Bulk Diffusion - Theoretical Analysis - Effect of DopantIt was found that there was a striking agreement between theory and experiment withregard to the behavior of p-type dopants. Group-III species exhibited retarded diffusionwhen they were interstitial, and exhibited enhanced diffusion when they weresubstitutional. The theoretical and experimental results indicated that the presence oflarge quantities of Ge, when completely integrated into Si, altered the molecular orbitalstructure of the lattice to such a degree that large long-range effects were exerted upon p-type dopant species. It was concluded that interactions between p-type dopants, andinteractions with Ge, could be used to define and alter diffusion patterns.S.Aronowitz, C.Hart, S.Myers, P.Hale: Journal of the Electrochemical Society, 1991,138[6], 1802-6

[446-84/85-063]

Bulk Diffusion - Theoretical Analysis - Effect of Ion ImplantationThe temporal evolution of transient enhanced diffusion in crystalline material duringannealing after B implantation was modelled as a system of diffusion-reaction equationsfor the dopant species and for Si point defects. The concept of point defect impurity pairdiffusion under equilibrium conditions was used to describe the diffusion. The out-diffusion of implantation-induced Si self-interstitials, and a kick-out reaction (Bi ⇔ Bs +I) were assumed to be the leading mechanisms for B activation. In the case of low-doseB-ion implantation, the analysis began with a defect distribution of Gaussian form, with 1interstitial per implanted B atom. At higher B doses, the areal density of this interstitialdistribution was constant, but the depth position of its peak depended upon the B dose.Local equilibrium of the reactions between point defects and B species was assumed to beachieved before the onset of diffusion. The predicted B depth profiles were comparedwith published data. Doses of between 2 x 1014 and 5 x 1015/cm2 were analyzed, andannealing temperatures and times which ranged from 800 to 1000C and from to 10s to 8h

324

B Bulk Diffusion B

were used. Although the approach involved only simple assumptions, importantdeficiencies were found only in certain cases of annealing after high-dose B implantation.It was concluded that the trapping of free interstitials by extended defects becameimportant at low temperatures and at long annealing times. If the depth region with themaximum B concentration was close to amorphization in its as-implanted state, B over-activation which was beyond the present model was found. In all other cases, it waspossible to obtain a reasonable model for transient enhanced diffusion.H.U.Jäger: Journal of Applied Physics, 1995, 78[1], 176-86

[446-123/124-182]

Bulk Diffusion - Theoretical Analysis - Effect of Ion ImplantationA model was developed in order to describe, in terms of an effective diffusivity, theimplantation-enhanced diffusivity of impurity atoms via a dual vacancy-interstitialmechanism. The model included the mechanisms of vacancy-interstitial pair generationby high-dose impurity implantation, which produced dopant diffusion enhancement, and aso-called transient activation mechanism for dopants in the initial stages of the annealingprocess.H.R.Soleimani: Journal of the Electrochemical Society, 1992, 139[11], 3275-84

[446-93/94-044]

Bulk Diffusion - Theoretical Analysis - Effect of Ion ImplantationA computer study was made of the accelerated diffusion of impurities which wereintroduced by high-temperature implantation. The diffusion equations were solved byusing the finite difference method. It was assumed that the acceleration was due to thepresence of an excess of point defects, and was proportional to their change inconcentration with depth. The calculated profiles were compared with data on theimplantation of B (at 900C) to doses of between 3 x 1013 and 1016/cm2. Agreement withexperimental data was obtained at all doses only if the point defect diffusion length wasreduced upon approaching the ion deceleration region. This was due to the fact thatdeeply penetrating B atoms acted as sinks for point defects, and thus reduced theirdiffusion length.L.N.Aleksandrov, T.V.Bondareva, G.A.Kachurin, I.E.Tyschenko: Fizika i TekhnikaPoluprovodnikov, 1991, 25[2], 227-30 (Soviet Physics - Semiconductors, 1991, 25[2],137-9)

[446-84/85-064]

Bulk Diffusion - Theoretical Analysis - Effect of Ion ImplantationThe implantation of 49BF2

+ ions into evaporated amorphous films was investigated byusing energies which ranged from 20 to 140keV. The implanted profiles werecharacterized by means of secondary ion mass spectrometry, and quantitative data on thefirst 4 moments of the profiles were gathered. The Pearson IV distributions and profileswhich were calculated by using an Edgeworth expansion were fitted to the measureddistributions by using an iterative procedure which used the profile moments as fittingparameters. The optimum set of parameter values was extracted for all of the implantationenergies which were investigated, and both types of analytical function were found toprovide an excellent fit to the experimental profiles at low energies. However, the

325

B Bulk Diffusion B

Edgeworth expansion became less applicable above 100keV; where the stopping waspredominantly electronic. The effect, upon the secondary ion mass spectrometry profiles,of primary sputtering O2

+ ions was studied. In particular, a shallow distribution at 20keVwas investigated. It was found that an energy of less than 5.5keV was required in order tominimize the broadening of the B tails towards greater depths. Monte Carlo simulationswere performed by using 3 different versions of ion transport in matter, and variousestimates for the electronic stopping cross-section. A comparison of the simulations andthe secondary ion mass spectrometry profiles revealed large deviations from a velocity-proportional dependence of the electronic stopping cross-section at energies of less than100keV. It was concluded that an improved theoretical model was required for theelectronic interaction of low-velocity ions.J.Tirén, B.G.Svensson: Journal of the Electrochemical Society, 1991, 138[2], 571-6

[446-81/82-046]

Bulk Diffusion - Theoretical Analysis - Effect of Ion ImplantationA comprehensive model was presented for describing B implantation, diffusion andclustering. The model, as demonstrated by means of Monte Carlo simulations, couldexplain and predict the behavior of B under a wide range of implantation and annealingconditions by proposing that immobile precursors of B clusters were formed before theonset of transient enhanced diffusion. The model also included the usual mechanisms ofSi self-interstitial diffusion and B kick-out. The immobile B cluster precursors, such as aB atom plus two Si self-interstitials, formed during implantation or in the very earlystages of annealing when the Si interstitial supersaturation was very high. They then actedas nucleation centers for the formation of B-rich clusters during annealing. The B-richclusters constituted the electrically inactive B component, so that clustering markedlyaffected both the junction depth and the dopant level in high-dose implants.L.Pelaz, M.Jaraiz, G.H.Gilmer, H.J.Gossmann, C.S.Rafferty, D.J.Eaglesham, J.M.Poate:Applied Physics Letters, 1997, 70[17], 2285-7

[446-150/151-152]

Bulk Diffusion - Theoretical Analysis - Effect of StressA model which was based upon point defects was developed for treating the effect ofstress upon dopant diffusion. Variations in the binding energies and diffusivities ofdopant-defect pairs under hydrostatic pressure were modelled, and a pressure-dependentdopant diffusion equation was derived. New experiments were performed on B pile-upsnear to dislocation loops, and the results were compared with the model predictions.Qualitative agreement was found, and it was concluded that stresses could play asignificant role in certain device structures.H.Park, K.S.Jones, J.A.Slinkman, M.E.Law: Journal of Applied Physics, 1995, 78[6],3664-70

[446-125/126-146]

Bulk Diffusion - Theoretical Analysis - Effect upon DefectsCritical conditions, for the generation of misfit dislocations during B diffusion fromchemical sources, were deduced analytically. The critical thickness of a highly-doped Bregion was defined in terms of the maximum concentration gradient, and analytical

326

B Bulk Diffusion/Melt Diffusion B

solutions of the B diffusion equation; for which a suitable dependence of the B diffusioncoefficient upon concentration was deduced. The critical conditions were also explainedin terms of the B surface concentration, diffusion time and diffusion temperature.F.Gaiseanu: Solid State Phenomena, 1996, 47-48, 223-8

[446-134/135-159]

Bulk Diffusion - Theoretical Analysis - Enhanced DiffusionAn accurate simulation of the enhanced diffusion of B during rapid thermal annealing wasobtained by incorporating the effects of extended defect formation and annealing into amulti-zone semi-empirical model. The multi-zone model involved the division of theimplantation profile into 3 zones which defined regions of differing defects and diffusionenhancement. The model also included the initial enhanced diffusion and transientdiffusion effects which were associated with the dissolution of defect clusters and withthe annealing of extended defects, respectively. The saturation time for transientenhanced diffusion involved an exponential function of implantation dose so as to modelthe increase in point defect which was generated by higher doses. The model accuratelysimulated the B diffusion profile over wide range of doses, and also reproduced theimmobile peak of precipitated dopant which was produced by high-dose implantation.H.Kinoshita, G.Q.Lo, D.L.Kwong, S.Novak: Journal of the Electrochemical Society,1993, 140[1], 248-52

[446-106/107-136]

Bulk Diffusion - Theoretical Analysis - Transient EffectsThe transient diffusion behavior of B during rapid thermal annealing was simulated bymodifying a recently developed pair diffusion model. The B was assumed to reside oninterstitial sites after ion implantation, and to form B-interstitial pairs. Decay to formsubstitutional B, plus interstitials, began upon increasing the temperature. Implantationdamage was taken into account by the model, which also allowed for a temperaturedependence of the transient diffusion effect.M.Heinrich, M.Budil, H.W.Pötzl: Journal of Applied Physics, 1991, 69[12], 8133-8

[446-86/87-050]

Melt Diffusion - Quantitative DataLaser pulses were directed at n- epitaxial layers (0.006mm thick) on a substrate of p+ B-doped Si. It was found that local melting and re-freezing of the layer, and of a smallvolume of the substrate, occurred. In the melt, the occurrence of B diffusion from thesubstrate gave an almost uniform dopant concentration of 5 x 1018/cm3. On the basis ofthe results, a value (2.4 x 10-4 at 1420C) was deduced for B diffusion in the liquid.K.M.Kim, S.N.Mei, M.J.Saccamango, S.F.Chu, R.J.Von Gutfeld, D.R.Vigliotti: AppliedPhysics Letters, 1992, 61[9], 1066-8

[446-93/94-048]

327

BiBulk Diffusion - Quantitative DataThe Bi was diffused in from a spin-on source, and the specimens were annealed attemperatures of between 1050 and 1200C. The resultant Bi profiles were measured byusing sheet resistivity and etching techniques. The profiles could be described by thecomplementary error function, and the diffusivity within the above temperature rangecould be described by the expression:

D(cm2/s) = 0.0002 exp[-2.50(eV)/kT]Y.Ishikawa, K.Yazaki, I.Nakamichi: Japanese Journal of Applied Physics, 1989, 28[7],1272-3

[446-70/71-123]

Bulk Diffusion - Qualitative Observations - Effect of OxidationThe effect of thermal oxidation upon the diffusion of Bi was investigated. The diffusioncoefficients of Bi during drive-in were determined from the best-fitted parameters tonumerical simulations of the experiments. It was found that the diffusion of Bi wasretarded by the thermal oxidation of Si. The degree of retardation was less than that of Sb,and the oxidation time dependence of the retardation was more marked than that of Sb.These contradictions suggested that the diffusion mechanism of Bi might not involve asimple dual mechanism.Y.Ishikawa, I.Kobayashi, I.Nakamichi: Japanese Journal of Applied Physics, 1990,29[10], L1929-31

[446-78/79-052]

328

C

Figure 5: Diffusivity of C in Si (see table 26)

326 Bulk Diffusion - Quantitative DataFloat-zone material was irradiated with 2MeV electrons at 140K in order to displacesubstitutional carbon atoms, Cs, into interstitial sites, Ci. The concentrations of the 2species were determined from infra-red localized vibrational mode absorptionmeasurements. Isothermal annealing at temperatures of between 297 and 331K revealed aloss of neutral Ci atoms under first-order kinetics, and the formation of di-carbon centers.From the known concentration of Cs traps, values for the diffusion coefficient weredetermined. These data (table 26), together with previous electron paramagneticresonance reorientation data, gave:

D(cm2/s) = 0.44 exp[-0.87(eV)/kT]

1.0E-16

1.0E-15

1.0E-14

1.0E-13

29 30 31 32 33 34

29.8531.7532.7933.33

104/T(K)

D (cm2/s)

329

C Bulk Diffusion C

A.K.Tipping, R.C.Newman: Semiconductor Science and Technology, 1987, 2[5], 315-7[446-51/52-135]

Table 26Diffusivity of Interstitial C in Float-Zone Si

T (C) D (cm2/s)62 2.9 x 10-14

42 4.6 x 10-15

32 1.7 x 10-15

27 8.0 x 10-16

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe out-diffusion behavior of C from heat-treated Czochralski material was investigatedby using secondary ion mass spectroscopy. In C-doped material, the diffusion of C wasgreatly enhanced at 750C, but was significantly retarded at 1000C. The retarded diffusionwas tentatively attributed to the formation of slow-diffusing complexes such as Si-O-C.F.Shimura, T.Higuchi, R.S.Hockett: Applied Physics Letters, 1988, 53[1], 69-71

[446-62/63-229]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsFirst-principles total energy calculations were made of the ground-state properties andmigration paths of interstitial C. The ground state involved 3-fold coordinated C and Siatoms, and its geometry suggested mainly p and sp bonding for C, rather than the sp2

bonding that might have been expected. A consideration of the possible migration pathsrevealed that only 3 of them corresponded to small jumps which involved single bond-breaking. Of these, it was predicted that only 1 had a barrier of considerably lower energy(about 0.5eV) and involved an intermediate saddle-point configuration with C2 symmetry.R.B.Capaz, A.Dal Pino, J.D.Joannopoulos: Physical Review B, 1994, 50[11], 7439-42

[446-119/120-224]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsThe co-precipitation of O and C in Czochralski-type material was studied in terms of adiffusion-limited growth model. Interfacial energy increases which occurred upon Cincorporation into oxide precipitates, as well as changes in the O and C concentrations ofthe Si matrix as a function of annealing time, were taken into account. A comparison ofthe model predictions with available experimental data led to the conclusion that,regardless of the C content of the crystal, it was necessary to introduce sinks forprecipitation-induced excess Si self-interstitials in the matrix at high annealingtemperatures. At annealing temperatures below about 1000C, the enhancement effect of Cupon O precipitation resulted mainly from an increase in the precipitate density. Atransition in the C precipitation behavior, that was observed in C-rich Si crystals atannealing temperatures of about 800C, was related to a change in the availability of

330

C Bulk Diffusion C

effective Si self-interstitial sinks in the Si matrix at these temperatures. An enhancementof the C diffusivity in the presence of excess Si self-interstitials played an important rolein increasing the precipitate growth rate; especially at low temperatures, where noefficient Si self-interstitial sinks were available in the Si matrix.J.Y.Huh, U.Gösele, T.Y.Tan: Journal of Applied Physics, 1995, 78[10], 5926-35

[446-125/126-147]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsIt was noted that C was present at interstitial and substitutional sites in Si, and incomplexes. Both the C and the Si interstitial atoms were mobile, and became trapped atother defects (thus altering their properties) or displaced impurity atoms into interstitialsites. An approximate molecular orbital procedure, which incorporated Car-Parrinello likedynamics, was used to model the structure of the Si and C interstitial atoms and of the Csubstitutional atom. By displacing them along their possible migration paths, and byallowing them to relax back into their stable sites, the details of low-energy migrationroutes were calculated. The neutral C interstitial occupied a (001) split site, with the Cbeing slightly nearer the site than the Si. Negative and singly positively chargedinterstitials occupied the same sites but, when doubly positively charged, the hexagonalsites were more stable. The lowest-energy C migration route was to pass via thehexagonal site, with a saddle-point energy of 0.77eV. Both the Si and C interstitialsformed complexes, of which the metastable di-C center was the best understood. Similarcalculations for this defect showed that the metastable Cs-Ci form had an energy whichwas about 0.8eV above the Cs-Sii-Cs form. An energy barrier of 0.1eV inhibited its re-ordering.A.Mainwood: Materials Science Forum, 1995, 196-201, 1589-94

[446-127/128-155]

331

ClBulk Diffusion - Theoretical Analysis - Concentration ProfileThe kinetics of thermal desorption from an excimer laser-melted surface, where diffusioninto the bulk competed with desorption, were analyzed numerically and were comparedwith experimental data (Auger electron spectroscopy, time-of-flight mass spectrometry,secondary-ion mass spectrometry, transient reflectivity) for Cl on Si. The modelcalculations involved non-equilibrium thermal diffusion, phase transitions, segregation,and first-order and second-order desorption kinetics. Upon assuming that the pre-exponential factors of the desorption rates did not change for the liquid surface, ascompared with the solid one, activation desorption energies were found which were lowerby about 0.5eV for the liquid surface than for the Si(100):Cl solid surface. This differencewas of the same order of magnitude as the latent heat of melting. The segregationcoefficient of Cl at the liquid/solid interface was less than 0.02 at a recrystallization speedof about 6m/s. The calculations also provided information on the dynamics of desorptionand melting. The desorption rate from the liquid attained a value of about 1ML/ns, whereML was defined to be the density of atoms in a (100) plane.B.Dragnea, J.Boulmer, J.P.Budin, D.Débarre, B.Bourguignon: Physical Review B, 1997,55[20], 13904-15

[446-152-0331]

CoBulk Diffusion - Quantitative DataThe introduction of Co from an infinitely thin source into a Si single crystal wasinvestigated at temperatures ranging from 700 to 1000C. Radioactive tracer andsecondary ion mass spectroscopic techniques were used independently in order todetermine the diffusion profiles. It was found that the results could be described by anArrhenius expression in which the pre-exponential factor was between 6.8 x 10-10 and1.35 x 10-9cm2/s and the activation energy was 1.5eV. The values given by the expressionwere several orders of magnitude lower than those previously reported when using thickfilms as the diffusion source in anneals which were carried out at higher temperatures.

332

Co Bulk Diffusion Co

A.Appelbaum, D.L.Malm, S.P.Murarka: Journal of Vacuum Science and Technology B,1987, 5[4], 858-64

[446-55/56-037]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe solubility and diffusion of Co in extrinsic material at 700C was investigated by usingradiotracer measurements. It was found that the solubility was enhanced by B or Pdoping, and that this was accompanied by a decrease in the diffusivity which was a factorof 3 smaller than that in intrinsic material with 8 x 1019B/cm3 and 3500 times smaller inmaterial with 1020P/cm3. Mossbauer spectra from the latter sample showed that the Cooccurred as two species, Cos and CosP, and the effect of this upon P-diffusion getteringwas explained. In B-doped material, the diffusion coefficient decrease could be explainedby the reaction of Coi

+ and B, and a single donor level of Coi at Ev + 0.15eV was deduced.D.Gilles, W.Schröter: Materials Science Forum, 1986, 10-12, 169-74

[446-49-028]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe annealing of supersaturated Co in n-type dislocation-free (111) float-zone wafersamples with a P content of 1014/cm3 was studied by measuring the Co depth profile at1100C. The results suggested that the diffusion of electrically active Co was controlledmainly by the interstitial Co content and could be described approximately by means of asimple diffusion equation.K.Hashimoto, H.Nakashima, K.Hashimoto: Japanese Journal of Applied Physics, 1988,27[9], 1776-7

[446-62/63-229]

Bulk Diffusion - Qualitative Observation - Effect of DefectsThe nature of Co/(100)Si interfaces which had been formed at room temperature wasstudied by using X-ray absorption fine-structure, low-energy electron diffraction, andAuger electron spectroscopic methods. The evaporation of Co onto chemically etched andannealed (100)Si 2 x 1 surfaces led to a coverage of about half of a monolayer for theadsorption of Co at the 4-fold hollow sites of the locally non-reconstructed (100)Sisurface. At higher coverages, the Co atoms occupied interstitial and substitutional sites ofthe Si lattice. Annealing (550 to 650C, 300s) caused the diffusion of Co into the Si lattice,where the Co atoms occupied only adamanthane sites. After the evaporation of Co ontosputter-annealed 2 x 1 (100)Si surfaces, only the formation of a locally ordered CoSi2-likephase was observed at coverages of between 0.8 and 14 monolayers.H.L.Meyerheim, U.Döbler, A.Puschmann: Physical Review B, 1991, 44[11], 5738-44

[446-84/85-066]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationAn investigation was made of the annealing behavior of <111> Si which had beenimplanted with Co, to doses of between 2.5 x 10

16 and 2 x 10

17/cm

2, using energies of

between 30 and 250keV. The formation of silicide during post-annealing was found toproceed via 2 stages which were characterized by activation energies of 0.8 and 2.9eV.During the first stage, fast Co redistribution and pronounced nucleation and growth of

333

Co Bulk Diffusion Cr

CoSi2 precipitates at defects were observed. An investigation of the dose dependenceproved that the degree of Co redistribution during this stage depended upon the initial Coand defect concentrations. However, the processes which contributed to this fast Coredistribution required further investigation. The second stage was characterized byOstwald ripening. Multiple implantations of Co with various energies were used tomodify the Co concentration profile. In this way, it was possible to form Si/CoSi2/Si/CoSi2 layer systems, on Si substrates, with various layer thicknesses. All of the Si layers,and the deeper-lying CoSi2 layer, were A-type whereas the crystalline orientation of thenear-surface layer depended upon the formation process. If the dose of the secondimplantation was too high, CoSi nucleated during implantation. The CoSi precipitatesdissolved during annealing, and a CoSi2 surface layer of type-A was formed. AdditionalSi implantation offered the possibility of modifying the defect profile without changingthe Co distribution. During subsequent annealing, Co redistribution into the region of theadditional defects was observed. It was concluded that implantation defects acted as sinksfor metal atoms, and played an important role during the first stage of annealing.A.Witzmann, S.Schippel, A.Zentgraf, P.I.Gajduk: Journal of Applied Physics, 1993,73[11], 7250-60

[446-109/110-045]

CrBulk Diffusion - Quantitative DataA combination of rapid thermal annealing and deep-level transient spectroscopy was usedto measure Cr profiles in the near-surface region of Czochralski material after aconventional 3-step gettering cycle. By fitting an erfc curve to the results, it wasestimated that the Cr diffusivity at 850C was equal to 1.5 x 10-7cm2/s.J.Zhu, G.Chaussemy, D.Barbier: Applied Physics Letters, 1989, 54[7], 611-3

[446-64/65-178]

Bulk Diffusion - Quantitative DataThe behavior of type-3d transition metal impurities was investigated by using deep-leveltransient spectroscopic and Hall-effect measurements. Pairs of donors with B, acting asdonors, were detected. The diffusivity of Cr, at temperatures ranging from 300 to 673K,was described by:

D(cm2/s) = 0.00068 exp[-0.79(eV)/kT]

The electrically active components of Co were judged to be substitutional species. Theseactive components, which represented only a small fraction of the total Co content, hadan amphoteric nature in n-type and p-type samples.

334

Cr Bulk Diffusion Cs

H.Nakashima, T.Sadoh, H.Kitagawa, K.Hashimoto: Materials Science Forum, 1994, 143-147, 761-6

[446-113/114-044]

Bulk Diffusion - Quantitative DataEdge-defined film-fed material was investigated by using deep-level transientspectroscopic and surface photo-voltage techniques. An impurity energy level of CrB wasfound at 0.27eV above the valence band in samples which were contaminated with Cr.The Cr diffusion coefficient in this material was deduced to be equal to 2 x 10-17cm2/s atroom temperature on the basis of the association and dissociation of CrB pairs during a210C dissociation anneal. Most of the deep-level transient spectra could not be analyzedby using conventional methods, due to the appearance of abnormally broad peaks. Thespectra of as-grown material were modelled by using a Gaussian distribution of impurityenergy states. The simulated peaks agreed well with the experimental data; thusexplaining the origin of the deep-level impurities in this material.S.H.Park, D.K.Schroder: Journal of Applied Physics, 1995, 78[2], 801-10

[446-123/124-183]

Bulk Diffusion - Qualitative Observations - Effect of DefectsInteractions between group-VI elements (S, Se) and fast-diffusing Cr impurities werestudied. Infra-red absorption, electron spin resonance, and neutron activation studiesshowed that the group-VI elements effectively interacted with each other or with the fast-diffusing impurities. It was found that, for each pairing of a group-VI element with a fast-diffusing impurity, there was a certain annealing temperature at which they interactedmost efficiently. A correlation was established between this annealing temperature andthe thermodynamic Gibbs free energy (at 298K) of the corresponding compounds. Thiscorrelation suggested that the interaction process mainly involved the formation ofelectrically neutral chemically bonded complexes by substitutional group-VI elementcenters and interstitial fast-diffusant centers.M.K.Bakhadirkhanov, S.I.Askarov, N.Norkulov: Physica Status Solidi A, 1994, 142[2],339-46

[446-117/118-194]

CsSurface Diffusion - Quantitative DataThe diffusion of Cs on (100) was studied by using biased secondary electron imagingmethods which were capable of detecting Cs coverages of more than 0.005 of a

335

Cs Surface Diffusion Cs

monolayer. Unusual diffusion profiles, which were linear at low coverages and had edgeswhich expanded as t½, were obtained at coverages of less than half of a monolayer. Theresults were modelled by assuming a diffusion coefficient which was of the form, D1 +D2(A/kT)[θ(1-θ)], where θ was the coverage. This form was consistent with a diffusiontheory which took account of strongly repulsive Cs-Cs interactions, as deduced from adecrease in the adsorption energy, q, as a function of coverage (dq/dθ = 2eV/monolayer).The diffusivity at temperatures ranging from 60 to 90C was consistent with an adatomdiffusion energy of 0.47eV.R.H.Milne, M.Azim, R.Persaud, J.A.Venables: Physical Review Letters, 1994, 73[10],1396-9

[446-115/116-152]

Surface Diffusion - Qualitative ObservationsThe migration and thermal desorption of sub-monolayer coverages of Cs on (100) 2 x 1were studied by using biased secondary electron imaging. The results indicated theexistence of 3 different bonding regimes as a function of coverage. These were found atcoverages of less than 1/6 of a monolayer, between 1/6 and 1/2 of a monolayer, andbetween 1/2 and 1 monolayer. When the coverage was less than 1/6 of a monolayer, thediffusion could best be described in terms of strongly repulsive dipole interactions. Thecomplex diffusion profiles which were observed at coverages of between 1/6 and 1/2 of amonolayer were explained by incorporating a 2-phase model into coupled rate anddiffusion equations. The dilute phase represented all of the structures where individualadatoms were separated by at least 2 unoccupied sites, while the condensed phaserepresented structures in which adatoms were on neighboring sites. The model predictedthe existence of large activation energies for adatom transfer between the 2 phases, and adifference of 0.04eV between the energies which were required in order for an adatom tomove from the dilute phase to the condensed phase and vice versa.R.H.Milne, M.Azim, R.Persaud, J.A.Venables: Surface Science, 1995, 336, 63-75

[446-136/137-129]

336

Cu

Figure 6: Diffusivity of Cu in Si

327 Bulk Diffusion - Quantitative DataThe diffusivity of Cu impurity was studied in amorphous samples which had beenprepared by MeV Si implantation. The 0.0022mm-thick layers were first annealed at500C and then implanted with 200keV Cu ions, which restored a 300nm-thick surfacelayer to the as-implanted state. The Cu concentration profiles were measured by using4He back-scattering. After diffusion at temperatures of between 150 and 270C (table 27),solution partitioning was detected at sharp phase boundaries between the annealed andCu-implanted layers. The partition coefficient could be as high as 8. The diffusioncoefficient in annealed amorphous material was 2 to 5 times larger than that in as-

1.0E-16

1.0E-15

1.0E-14

1.0E-13

1.0E-12

1.0E-11

1.0E-10

1.0E-09

1.0E-08

1.0E-07

1.0E-06

1.0E-05

7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37

table 27table 28table 29table 30table 31

104/T(K)

D (cm2/s)

337

Cu Bulk Diffusion Cu

implanted amorphous material. The activation energies were 1.39 and 1.25eV,respectively. Defects played an important role.A.Polman, D.C.Jacobson, S.Coffa, J.M.Poate, S.Roorda, W.C.Sinke: Applied PhysicsLetters, 1990, 57[12], 1230-2

[446-76/77-035]

Table 27Diffusivity of Cu in Amorphous Si

T (C) D (cm2/s)270 2.4 x 10-12

240 5.8 x 10-13

220 1.0 x 10-13

200 2.8 x 10-14

170 2.8 x 10-15

150 6.0 x 10-16

328 Bulk Diffusion - Quantitative DataTransition metals in amorphous samples exhibit a direct interstitial diffusion behaviorwhich is retarded by temporary trapping at defects that are intrinsic to the amorphousstructure. Diffusion was investigated here by means of Rutherford back-scatteringspectrometry. It was found that the data (table 28) could be fitted by using foreign-atominterstitial diffusion coefficients for crystalline Si; modified by the presence of traps inconcentrations of between 0.2 and 1at%, and with trapping enthalpies of about 0.9eV.S.Coffa, J.M.Poate, D.C.Jacobson, W.Frank, W.Gustin: Physical Review B, 1992,45[15], 8355-8

[446-86/87-049]

Table 28Diffusivity of Cu in Amorphous Si

T (C) D (cm2/s)323 2.2 x 10-12

290 4.7 x 10-13

235 2.2 x 10-14

155 1.4 x 10-16

Bulk Diffusion - Quantitative DataThe presence of Cu atoms in p-type material was revealed via their characteristic electricfield gradients, which were measured at the radioactive acceptor, 111In/111Cd, via theperturbed ?-? angular correlation technique. The Cu formed pairs with acceptor atoms and

338


electrically passivated them. By using known Cu diffusion data and taking account of theeffect of ion-pairing, an activation energy of 0.15eV was deduced. This value was inaccord with the energy of 0.70eV which had been deduced for Coulomb-bound acceptor-Cu pairs.R.Keller, M.Deicher, W.Pfeiffer, H.Skudlik, D.Steiner, T.Wichert: Physical ReviewLetters, 1990, 65[16], 2023-6

[446-76/77-035]

329 Bulk Diffusion - Quantitative DataThe diffusion parameters of Cu were determined in profiled Cu samples, that had beenprepared by using the Stepanov method, at temperatures ranging from 900 to 1050C. Itwas found that the data (table 29) could be described by:

D(cm2/s) = 0.015 exp[-0.86(eV)/kT]K.P.Abdurakhmanov, M.B.Zaks, V.V.Kasatkin, G.S.Kulikov, S.K.Persheev,K.K.Khodzhaev: Fizika i Tekhnika Poluprovodnikov, 1989, 23[10], 1891-3 (SovietPhysics - Semiconductors, 1989, 23[10], 1170-1)

[446-76/77-036]

Table 29Diffusion of Cu in Si

T (C) D (cm2/s)900 2.9 x 10-6

950 4.3 x 10-6

1000 5.3 x 10-6

1050 7.6 x 10-6

Bulk Diffusion - Quantitative DataNear-surface Cu-rich layers were observed by means of the secondary ion massspectroscopy of intentionally contaminated wafers after thermal oxidation (950C, 2100s).The diffusivity of Cu, at 950C, which was deduced from secondary ion massspectroscopy profiles was 9 x 10

-14. This value was almost 9 orders of magnitude lower

than that expected for interstitial diffusants. It was comparable to those of group-III andgroup-V atoms; which were known to be substitutional diffusants. On the basis of theresults, a model was proposed in which the interstitial and substitutional atoms diffusedindependently, and the slow substitutional diffusion gave rise to a near-surface impurity-rich layer.L.Zhong, F.Shimura: Japanese Journal of Applied Physics 2, 1993, 32[8B], L1113-6

[446-106/107-137]

330 Bulk Diffusion - Quantitative DataThe transient ion drift in the depletion region of a Schottky barrier was used to determineion diffusivities at moderate temperatures. The pulsed reverse bias led to temperature-

339


dependent capacitance transients which were similar to deep-level carrier emissiontransients. A simple theoretical model, together with classical transient signal analysis,permitted the ion diffusion constant to be deduced. When the method was applied to thediffusion of Cu, data were obtained for the previously uninvestigated temperature rangeof 280 to 400K (table 30). These results agreed well with both low-temperature and high-temperature diffusion data and they could all be described by the expression:

D(cm2/s) = 0.0045 exp[-0.39(eV)/kT]

T.Heiser, A.Mesli: Applied Physics A, 1993, 57[4], 325-8[446-113/114-044]

Table 30Diffusivity of Cu in Si

T (C) D (cm2/s)132 7.3 x 10-8

112 2.9 x 10-8

100 2.1 x 10-8

90 1.3 x 10-8

81 1.1 x 10-8

70 6.0 x 10-9

38 2.8 x 10-9

9 6.1 x 10-10

1 4.1 x 10-10

331 Bulk Diffusion - Quantitative DataTransient ion drift in depletion regions of a Schottky barrier was used to investigatediffusion in B- or Al-doped material (table 31). It was shown that, within the studiedtemperature range, Cu-B pairing was negligible. Excellent agreement with publisheddiffusivity data was found for Cu ions, as described by the expression:

D (cm2/s) = 0.0045 exp[-0.39(eV)/kT]A.Zamouche, T.Heiser, A.Mesli: Applied Physics Letters, 1995, 66[5], 631-3

[446-125/126-148]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe diffusion of evaporated Cu in porous material, and the deposition of metal ions inaqueous solution through the porous network, was measured by monitoring the metalconcentration depth profile as a function of time by using Auger electron spectroscopy. Itwas observed that increasing the metal penetration from Cu evaporated samples wasrelated to a quenching of the photoluminescence; in agreement with previous ionquenching results. The diffusion coefficient was deduced from Auger concentration depthprofiles and was found to be 7 orders of magnitude lower than that expected for thediffusion of Cu in bulk crystalline Si at room temperature. It was concluded that the

340


deposition of ionic species could not be characterized as being a simple diffusion process.The observed deposition rates depended strongly upon the solute concentration.D.Andsager, J.M.Hetrick, J.Hilliard, M.H.Nayfeh: Journal of Applied Physics, 1995,77[9], 4399-402

[446-121/122-086]

Table 31Diffusivity of Cu in Si

T (C) D (cm2/s)117 4.5 x 10-8

102 2.4 x 10-8

87 1.5 x 10-8

82 1.0 x 10-8

72 7.0 x 10-9

72 1.7 x 10-8

67 3.6 x 10-9

62 5.0 x 10-9

57 1.0 x 10-8

52 2.8 x 10-9

47 6.4 x 10-9

42 3.0 x 10-9

37 1.9 x 10-9

37 4.0 x 10-9

27 2.3 x 10-9

17 1.4 x 10-9

9 4.8 x 10-10

Bulk Diffusion - Qualitative Observations - Concentration ProfilesAn investigation was made of the low-temperature out-diffusion of Cu from the bulk of p-type and n-type wafers after Cu contamination during annealing. It was shown that Cuimpurity in the bulk, after low-temperature out-diffusion, could be measured at thesurface by using total X-ray fluorescence and graphite-furnace atomic absorptionspectroscopy. A particular benefit of low-temperature annealing was the removal of Cucontamination from the bulk by surface cleaning. It was also noted that Cu contaminationin the bulk of p-type wafers out-diffused at room temperature, after removing the surfaceoxide. This did not happen in the case of n-type material.M.B.Shabani, T.Yoshimi, H.Abe: Journal of the Electrochemical Society, 1996, 143[6],2025-30

[446-148/149-185]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsTransmission electron microscopy was used to study monocrystals which containedstacking faults, and the effect of Cu diffusion was investigated. When no Cu was

341


introduced, hexagonal stacking faults with hexagonal Si oxide precipitates were observed.When Cu diffusion had occurred, stacking faults which had branches with CuSiprecipitates on the edges or inner parts of these branches were seen. The CuSi precipitatesgrew on the Frank dislocation loop, and interstitial Si atoms were released. The interstitialSi atoms condensed around the CuSi precipitates, and branches oriented in [110]-typedirections grew from the stacking faults.S.Aoki: Journal of the Japan Institute of Metals, 1991, 55[10], 1039-44

[446-86/87-053]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsA systematic experimental investigation was made of D-line luminescence, at photonenergies of 0.81 and 0.87eV, in heat-treated C-poor Czochralski-type crystalline wafersthat had been deliberately contaminated with Cu. The results revealed the presence ofplanar-type defects that were bounded by 111 planes and were composed of 60°-typedislocations and stacking faults. These led to the generation of D1 and D2 lines in heat-treated C-poor Czochralski-type crystals. It was noted that Cu diffusion into heat-treatedsamples markedly reduced the D2 line intensity, but enhanced the D1 line luminescence,up to a certain degree of Cu contamination.K.H.Cho, J.Y.Lee, E.K.Suh, K.W.Kim: Japanese Journal of Applied Physics 1, 1997,36[1A], 1-5

[446-148/149-186]

342

D, H

Figure 7: Diffusivity of Hydrogen Isotopes in Si

332 Bulk Diffusion - Quantitative DataExperimental data, which revealed the essential features of atomic H that had beenincorporated from various sources at low temperatures, were presented. Appropriateconditions for hydrogenation were chosen, and the H diffusivity (table 32) was found tobe described by:

D(cm2/s) = 0.00002 exp[-0.49(eV)/kT]

1.0E-20

1.0E-19

1.0E-18

1.0E-17

1.0E-16

1.0E-15

1.0E-14

1.0E-13

1.0E-12

1.0E-11

1.0E-10

1.0E-09

1.0E-08

1.0E-07

10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46

table 32table 33table 34table 35table 36table 37table 38table 39table 40table 41table 42table 43table 44table 46table 47table 48table 49table 50

104/T(K)

D (cm2/s)

343

D, H Bulk Diffusion D, H

S.V.Koveshnikov, S.V.Nosenko, E.B.Yakimov: Physica Status Solidi A, 1990, 120[2],391-5

[446-78/79-071]

Table 32Diffusivity of H in Si

T (C) D (cm2/s)280 8.1 x 10-10

200 1.3 x 10-10

145 2.6 x 10-11

95 4.3 x 10-12

Bulk Diffusion - Quantitative DataUsing coherent neutron scattering and secondary ion mass spectrometry techniques, itwas found that atomic D which was introduced into amorphous hydrogenated material(from a plasma) diffused fast interstitially and replaced bonded H atoms. The effectivediffusion coefficient was estimated to be about 10-14cm2/s. This was much larger than thatfor bond-breaking diffusion, which was equal to about 10-20cm2/s at 160C.B.Abeles, L.Yang, D.Leta, C.Majkrzak: Journal of Non-Crystalline Solids, 1987, 97-98,353-6

[446-61-085]

333 Bulk Diffusion - Quantitative DataThe diffusion kinetics were studied, at temperatures of between 220 and 270K (table 33),by analyzing the photo-induced dissociation of an etchant-generated H-C or D-Ccomplex. Under suitably strong illumination, the annihilation rate of the complex wasproportional to the P density. This indicated that the rate-determining step was thediffusion of H or D to P atoms. By invoking diffusion-controlled reaction theory, it wasdeduced that the diffusion behaviors were described by:

D: D(cm2/s) = 0.005 exp[-0.49(eV)/kT]H: D(cm2/s) = 0.07 exp[-0.54(eV)/kT]

Y.Kamiura, M.Yoneta, F.Hashimoto: Applied Physics Letters, 1991, 59[24], 3165-7[446-84/85-067]

334 Bulk Diffusion - Quantitative DataThe real-time detection of H motion and bonding was carried out by using capacitance-voltage profiling of various Schottky and metal-insulator capacitors during low-energy H-ion beam injection into the barrier metallization. Numerical modelling indicated that anappreciable fraction of the H interstitials were positively charged, and that bonding ofthese species with charged B acceptors involved the large cross-section which was to beexpected of a Coulomb capture process. The H diffusivity (table 34) at 300K was about

344


10-10cm2/s. This value was consistent with extrapolations of high-temperature diffusivitydata.C.H.Seager, R.A.Anderson: Applied Physics Letters, 1988, 53[13], 1181-3

[446-62/63-230]

Table 33Diffusivity of D and H in Si

Diffusant T (K) D (cm2/s)D 270 3.7 x 10-12

D 265 2.2 x 10-12

D 260 1.6 x 10-12

D 255 1.2 x 10-12

D 250 6.6 x 10-13

D 245 4.7 x 10-13

D 240 1.9 x 10-13

D 235 1.8 x 10-13

D 230 8.2 x 10-14

D 225 5.4 x 10-14

H 270 7.6 x 10-12

H 265 5.7 x 10-12

H 260 2.7 x 10-12

H 255 1.9 x 10-12

H 245 6.0 x 10-13

H 240 3.7 x 10-13

H 235 2.8 x 10-13

H 230 1.2 x 10-13

H 220 4.0 x 10-14

Bulk Diffusion - Quantitative DataExperiments were performed in which H was injected by using various techniques, andwas detected via the neutralization of B acceptor sites. Wet chemical etching injectedprotons to a depth of several microns within a few seconds. The results suggested that alower limit on the diffusivity of H at 300K was about 2 x 10-11cm2/s. This value was inreasonable agreement with published permeation data.C.H.Seager, R.A.Anderson, J.K.G.Panitz: Journal of Materials Research, 1987, 2[1], 96-106

[446-62/63-230]

Bulk Diffusion - Quantitative DataThe effect of heat treatment upon the minority carrier surface recombination velocity inH-passivated polycrystalline material was studied, at temperatures ranging from 350 to

345


500C, by using the electron beam induced current mode of a scanning electronmicroscope. The minority carrier trap center densities, which were deduced from minoritycarrier surface recombination velocity data, varied from 8 x 1012 to 1.2 x 1012/cm2. Afinite decrease in the minority carrier trap center density indicated that H atoms diffusedto the surface from the bulk of the hydrogenated samples. The activation energy for Hdiffusion in Si was found to be 0.53eV.R.Kumar, R.K.Kotnala, N.K.Arora, B.K.Das: Applied Physics Letters, 1988, 52[20],1670-1

[446-62/63-231]


T (C) D (cm2/s)57 5.3 x 10-10

57 3.1 x 10-10

52 2.5 x 10-10

42 2.6 x 10-10

22 1.3 x 10-10

10 8.1 x 10-11

4 6.2 x 10-11

0 3.8 x 10-11

0 3.0 x 10-11

Bulk Diffusion - Quantitative DataThe motion of H in the B-H complex was studied. An applied stress was used to producea preferential alignment of the B-H complex at temperatures which were sufficiently high(above about 60K) for the H to move within the complex. This alignment of thecomplexes was detected by comparing the optical absorption strengths of the H stretchingvibrations for light which was polarized parallel to and perpendicular to the stress axis.From the kinetics of the decay of the alignment after the stress was removed, it wasdeduced that the motion of H from bond-centered site to bond-centered site around the Batom was thermally activated with an activation energy of 0.19eV.M.Stavola, K.Bergman, S.J.Pearton, J.Lopata: Physical Review Letters, 1988, 61[24],2786-8

[446-64/65-179]

335 Bulk Diffusion - Quantitative DataSamples of (100)-oriented B-doped Czochralski material were diffused with H from a gelsource, at temperatures of between 80 and 100C (table 35). The resultant profiles weredetermined by means of C-U measurements at 1MHz and fitted by a Gaussian function.At 100C, the diffusion coefficient agreed with the expression,

D(cm2/s) = 0.000042 exp[-0.56(eV)/kT]

346


which had been determined previously. The coefficient which was measured at 80C wasan order of magnitude smaller.K.Schmalz, K.Tittel-Helmrich: Physica Status Solidi A, 1989, 113[1], K9-13

[446-76/77-038]


T (C) D (cm2/s)100 5.4 x 10-13

90 1.5 x 10-13

80 4.2 x 10-14

336 Bulk Diffusion - Quantitative DataThe migration of H in amorphous material which had been deposited at a temperature of77K was determined by monitoring the decay of the low-angle neutron scatteringintensity during the annealing of .../Si-H/Si-D/... multi-layers. Coefficients which rangedfrom 10-19 to 5 x 10-19cm2/s were found for annealing temperatures of between 250 and300C (table 36).M.Vergnat, S.Houssaini, C.Dufour, A.Bruson, G.Marchal, P.Mangin, R.Erwin,J.J.Rhyne, C.Vettier: Europhysics Letters, 1991, 14[5], 457-62

[446-78/79-054]

Table 36Diffusion of H in Si

T (C) D (cm2/s)250 9.33 x 10-20

270 1.87 x 10-19

280 3.60 x 10-19

300 5.61 x 10-19

337 Bulk Diffusion - Quantitative DataThe migration of H in un-doped amorphous hydrogenated material was measured byusing the elastic-recoil detection of variations in the dispersion parameter at temperaturesof between 350 and 470C (table 37). It was found that the dispersion parameter increasedwith increasing annealing temperature.X.M.Tang, J.Weber, Y.Baer, F.Finger: Physical Review B, 1990, 42[11], 7277-9

[446-78/79-054]

338 Bulk Diffusion - Quantitative DataA great deal of the H that was introduced into n-type material at about 150C formedcomplexes which were termed, H2*, in order to distinguish them from the stable and

347


almost immobile complex, H2, which formed in p-type material. Secondary ion massspectrometric measurements were made of the thermally-induced depth distribution of Dwhich was present mainly as 2H2*. The activation energy for migration (table 38) was0.81eV, but the data did not reveal whether the migration involved dissociation. Otherwork indicated that, for annealing times of up to 25h at 60C, there was no detectableredistribution or dissociation of 2H2*. The effect of the presence of H2* upon theevolution of neutralized donor profiles at these temperatures was negligible.N.M.Johnson, C.Herring: Physical Review B, 1991, 43[17], 14297-300

[446-84/85-070]

Table 37Diffusion of H in Si

T (C) D (cm2/s)470 1.6 x 10-13

450 5.9 x 10-14

425 8.7 x 10-15

400 3.2 x 10-15

375 2.9 x 10-15

350 1.3 x 10-15

339 Bulk Diffusion - Quantitative DataThe hydrogenation of MOS diode structures was studied by means of high-frequencycapacitance profiling. The results indicated that the irradiation of 40nm Al-gated diodeswith 0.8 to 1.4keV H ions soon established a time-independent near-surface Hconcentration which was directly proportional to the ion flux and the implantation depthand was inversely proportional to the H diffusivity. The results (table 39) indicated adiffusion activation energy of 0.58eV. The results of the modelling of low-field and high-field data suggested that the positive charge state was occupied for only 10% of the time.The time dependence of the H penetration indicated that H became immobilized at othersites in Si. The density of these secondary trapping sites was related to the dopantpopulation. It was concluded that additional H could become trapped near to already-passivated dopant atoms.C.H.Seager, R.A.Anderson, D.K.Brice: Journal of Applied Physics, 1990, 68[7], 3268-84

[446-86/87-055]

Bulk Diffusion - Quantitative DataThe locally restricted motion of H and D atoms which were trapped at Cd acceptors wasstudied by using perturbed angular correlation spectroscopy. Close Cd-H and Cd-D pairswere identified via their characteristic electric field gradients. The kinetics of thesecomplexes were deduced directly from the time spectra. At temperatures of between 270and 430K, clear evidence was found for the thermally activated motion of H or D aboutCd; with an activation energy of 0.21eV for both H and D. The isotopic mass-dependent

348


attempt frequencies were of the order of 109/s. The ratio of the frequency for H, ascompared with that for D, was 3.3.M.Gebhard, B.Vogt, W.Witthuhn: Physical Review Letters, 1991, 67[7], 947-50

[446-91/92-031]

Table 38Diffusivity of 2H2 in n-Type Si

Type T (C) D (cm2/s)float-zone 200 3.1 x 10-12

float-zone 200 2.1 x 10-12

float-zone 175 8.6 x 10-13

float-zone 175 6.6 x 10-13

float-zone 150 2.5 x 10-13

float-zone 150 1.4 x 10-13

float-zone 125 2.0 x 10-14

float-zone 125 1.5 x 10-14

Czochralski 200 1.7 x 10-12








340 Bulk Diffusion - Quantitative DataThe relaxation of stress-induced dichroism of the 9000nm O infra-red absorption bandwas investigated in samples of Czochralski material which had been annealed in a Hplasma at temperatures of between 225 and 350C. It was found that the in-diffusion of Hatoms enhanced the rate of O diffusion, so that dichroism disappeared gradually from theexternal surfaces. It was deduced that the H diffusivity (table 40) could be described by:

D(cm2/s) = 170 exp[-1.2(eV)/kT]R.C.Newman, J.H.Tucker, A.R.Brown, S.A.McQuaid: Journal of Applied Physics, 1991,70[6], 3061-70. See also: Materials Science Forum, 1992, 83-87, 87-92

[446-91/92-031], [446-93/94-055]

Bulk Diffusion - Quantitative DataA theoretical study was made of the diffusion barriers, Ti and Cu, in crystalline samples,and of the interactions between these elements and H. Calculations were performed byusing molecular cluster and Hartree-Fock methods. Use of the partial retention ofdiatomic differential overlap method predicted diffusion barriers of 3.29eV for Ti+,

349


2.25eV for Tio, and 0.24eV for Cu+. The latter method also predicted that substitutionalTio was a deep trap for interstitial H, with a gain in energy of 1.84eV (relative to atomic Hwhich was far from the cluster). First-principles Hartree-Fock calculations showed that aTi+ ion at a tetrahedral interstitial site also formed a bond with interstitial H. Thisinvolved a dissociation energy of 2.31eV. On the other hand, interstitial Cu+ did not forma bond with H.D.E.Woon, D.S.Marynick, S.K.Estreicher: Physical Review B, 1992, 45[23], 13383-9

[446-93/94-055]


T (C) D (cm2/s)48 1.4 x 10-10

28 7.3 x 10-11

10 8.8 x 10-12

Bulk Diffusion - Quantitative DataIt was demonstrated that H diffusion in hydrogenated amorphous material was trap-controlled. A 1.4eV barrier was measured for deep D emission to a transport level in D-doped amorphous material. It was shown that light-enhanced diffusion in this materialwas caused by the light-enhanced de-trapping of H, and not by heating of the sample.Estimates were obtained for the free-H diffusion coefficient (3 x 10

-8cm

2/s), the mean H

displacement between deep trapping events (25nm), and for other parameters that affectedthe measured H diffusion coefficient in this material.H.M.Branz, S.E.Asher, B.P.Nelson: Physical Review B, 1993, 47[12], 7061-6

[446-106/107-138]

Bulk Diffusion - Quantitative DataThe incorporation of H was studied under various electrochemical conditions, includinganodization in fluoride solutions (where porous Si formed). The results suggested thatthere was a large near-surface H concentration, while simulations showed that themaximum penetration depth was governed by the volume diffusion of H and by materialremoval. The diffusion coefficients were found to depend upon the electrochemicalconditions, and ranged from 10-13 to 10-11cm2/s.P.Allongue, C.H.De Villeneuve, L.Pinsard, M.C.Bernard: Applied Physics Letters, 1995,67[7], 941-3

[446-123/124-183]

341 Bulk Diffusion - Quantitative DataThe reactivation of passivated B acceptors in hydrogenated material was studied duringzero-bias annealing at temperatures ranging from 65 to 130C (table 41). It was found that,for long annealing times and high annealing temperatures, the reactivation process

350


exhibited second-order kinetics and was rate-limited by thermally activated H2 complexformation. At short annealing times and low annealing temperatures, the reactivation ratewas larger than that due to H2 complex formation alone. It was concluded that the fasterreactivation was caused by the diffusion of liberated H atoms into the bulk, as well as byH2 complex formation. The effective diffusion coefficient of H obeyed an Arrheniusrelationship, with an activation energy of 1.41eV.A.Majumdar, S.Balasubramanian, V.Venkataraman, N.Balasubramanian: Journal ofApplied Physics, 1997, 82[1], 192-5

[446-152-0350]


T (C) D (cm2/s)225 7.8 x 10-11

235 1.6 x 10-10

235 4.0 x 10-10

250 4.5 x 10-10

250 5.4 x 10-10

275 1.3 x 10-9

275 1.7 x 10-9

300 1.6 x 10-8

300 2.3 x 10-8

325 1.2 x 10-8

325 1.8 x 10-8

330 4.9 x 10-9

330 7.6 x 10-9

350 5.6 x 10-8

350 7.6 x 10-8

342 Bulk Diffusion - Quantitative DataThe migration of D was investigated by performing experiments in which the D sourcewas either a remote atomic D plasma or a deuterated amorphous layer (table 42).Enhanced diffusion, and a considerably lower activation energy, was observed in the caseof diffusion from a plasma. This behavior was attributed to the saturation of Dconfigurations with a high binding energy. These acted as deep D traps and controlledtransport during diffusion from a deuterated layer, but were used up by D atoms whichwere injected from the plasma. The diffusion from a plasma was then governed byhopping through states with a low binding energy. A density of states distribution wasdeduced for the configurations which controlled diffusion. This consisted partly of ashallow state in which the D was weakly bonded and could diffuse with an activationenergy of 0.5eV. There were also deep states, for which the binding energy was greater

351


than 1.2eV, and which had a total density of about 1022/cm3. In the case of glow-dischargehydrogenated amorphous material, the D was bonded to deep states and only 20 to 60%of these states were empty.P.V.Santos, W.B.Jackson: Physical Review B, 1992, 46[8], 4595-606

[446-93/94-050]


T (C) D (cm2/s)130 5.9 x 10-12

110 9.0 x 10-13

100 2.2 x 10-13

85 2.3 x 10-14

65 3.0 x 10-15

343 Bulk Diffusion - Quantitative DataThe migration of H in solid-state crystallized and low-pressure chemical-vapor depositedpolycrystalline material was investigated by performing D diffusion experiments (table43). The concentration profiles, of D which was introduced into the samples from aremote D plasma or from a deuterated amorphous Si layer, were measured as a functionof time and temperature. At high D concentrations, the diffusion was dispersive anddepended upon the exposure time. The dispersion was consistent with multiple trappingwithin a distribution of hopping barriers. The data could be explained in terms of a 2-level model that was used to explain diffusion in hydrogenated amorphous Si. The energydifference between the transport level, and the D chemical potential, was found to besome 1.2 to 1.3eV. The shallow levels for H trapping were about 0.5eV below thetransport level, while the deep levels were about 1.5 to 1.7eV below. The H chemicalpotential decreased as the temperature increased. At lower concentrations, the H chemicalpotential depended markedly upon the method which was used to prepare the material.This was suggested to be partly due to the dependence of the crystallite size upon thedeposition process. Clear evidence for D deep traps was found only in solid-statecrystallized material. The low-pressure chemical-vapor deposited material, with columnargrains that extended through the film thickness, displayed little evidence of deep trappingand exhibited enhanced D diffusion. Many of the concentration profiles in columnarchemical-vapor deposited material reflected a complex diffusion behavior. The latter wasattributed to spatial variations in trap density, complex formation and/or multipletransport paths. Many aspects of the present diffusion behavior were consistent withdiffusivity data for amorphous Si.N.H.Nickel, W.B.Jackson, J.Walker: Physical Review B, 1996, 53[12], 7750-61

[446-134/135-161]

344 Bulk Diffusion - Quantitative DataMigration in doped and compensated hydrogenated amorphous material was measuredusing secondary ion mass spectrometry profiling at temperatures ranging from 155 to

352


300C (table 44). It was found that doping reduced the activation energy and enhanced thediffusion coefficient by up to 3 orders of magnitude at 200C. There was a correlationbetween the diffusion coefficient and the dangling-bond density. An analysis of 3different diffusion models indicated that the breaking of weak Si-Si bonds by H could bean important process. The relationship between the diffusion results and the thermalequilibration of the electronic structure was considered.R.A.Street, C.C.Tsai, J.Kakalios, W.B.Jackson: Philosophical Magazine B, 1987, 56[3],305-20

[446-55/56-039]

Table 42Diffusivity of D in Si

Source T (C) D (cm2/s)plasma 350 3.4 x 10-15

plasma 275 8.7 x 10-16

plasma 225 3.1 x 10-16

plasma 200 1.8 x 10-16

plasma 200 1.7 x 10-16

plasma 175 9.4 x 10-17

plasma 175 8.3 x 10-17

plasma 175 7.5 x 10-17

layer 350 1.5 x 10-15

layer 300 2.2 x 10-16

layer 300 2.0 x 10-16

layer 250 3.1 x 10-17

345 Bulk Diffusion - Quantitative DataThe dependence of the diffusion coefficient upon the growth conditions of dopedhydrogenated amorphous material (table 45) was measured by using secondary ion massspectrometry. A markedly enhanced diffusivity (4.9 x 10-13 at 240C and 4.8 x 10-15cm2/s at240C) was found in material with a columnar microstructure, due to the preferentialmotion of H along the columns. Increasing the deposition temperature of non-columnarmaterial resulted in a higher diffusion coefficient and a lower H concentration. Nosignificant change in the diffusivity was found for different n-type dopants.R.A.Street, C.C.Tsai: Philosophical Magazine B, 1988, 57[5], 663-9

[446-60-015]

346 Bulk Diffusion - Quantitative DataThe elastic recoil detection method was used to measure the dispersion parameter for Hdiffusion in undoped hydrogenated amorphous material. It was found that decreasing thedeposition temperature of the films increased the value of the dispersion parameter. This

353


variation could also be clearly related to the increase in the initial concentration of weaklybound H. This was expected to reflect disorder in the microstructure. The results wereexpected to contribute to the elucidation of the dispersive character of H diffusion (table46) in various types of amorphous material.X.M.Tang, J.Weber, Y.Baer, F.Finger: Physical Review B, 1990, 41[11], 7945-7

[446-76/77-039]

Table 43Diffusivity of H in Polycrystalline Si

Material Source T (C) D (cm2/s)LPCVD plasma 452 9.0 x 10-13

LPCVD plasma 392 2.1 x 10-13




SSC plasma 452 3.8 x 10-13

SSC plasma 452 1.6 x 10-13

SSC plasma 452 1.4 x 10-13

SSC plasma 401 1.0 x 10-13

SSC plasma 350 7.0 x 10-13

SSC plasma 350 5.0 x 10-13

SSC plasma 350 3.8 x 10-13

SSC plasma 248 5.3 x 10-13

SSC plasma 248 3.5 x 10-13

SSC plasma 248 3.0 x 10-13

LPCVD layer 421 6.3 x 10-14



SSC layer 452 7.1 x 10-14

SSC layer 421 1.2 x 10-14

SSC layer 350 1.7 x 10-15

SSC layer 298 4.1 x 10-17

347 Bulk Diffusion - Quantitative DataDeuteration experiments were carried out on crystalline samples, and the resultantdiffusion depth profiles were measured by means of secondary ion mass spectrometry.Careful analysis of the latter data (table 47) permitted a reduction to be made in thenumber of fitting parameters by excluding H molecule formation. It was found that themobilities could be ranked in the increasing order: Ho - H- - H+. The dissociation energiesof BH, AlH, and PH complexes were also calculated and permitted an estimation to be

354


made of the corresponding bonding energies of the complexes. It was deduced thatcomplexes could be ordered in terms of increasing stability: PH - BH - AlH.R.Rizk, P.De Mierry, D.Ballutaud, M.Aucouturier, D.Mathiot: Physical Review B, 1991,44[12], 6141-51

[446-84/85-069]

Table 44Diffusivity of H in Doped Amorphous Si

B P T (C) D (cm2/s)0.01 - 272 9.7 x 10-14

0.01 - 242 3.0 x 10-14

0.01 - 201 1.7 x 10-15

0.001 - 301 1.4 x 10-13

0.001 - 273 5.3 x 10-14

0.001 - 255 1.2 x 10-14

0.001 - 252 1.1 x 10-14

0.001 - 229 3.6 x 10-15

0.001 - 201 9.1 x 10-16

0.001 - 177 1.5 x 10-16

0.001 - 156 3.0 x 10-17

0.0001 - 273 1.4 x 10-14

0.0001 - 242 1.9 x 10-15

0.0001 - 199 1.5 x 10-16

0.001 0.0001 272 2.5 x 10-15

0.001 0.0001 240 3.0 x 10-16

0.001 0.0001 200 3.2 x 10-17

- 0.001 300 2.8 x 10-15

- 0.001 272 1.1 x 10-15

- 0.001 251 3.3 x 10-16

- 0.001 242 2.2 x 10-16

- 0.001 227 8.8 x 10-17

- 0.001 201 2.0 x 10-17

- 0.0001 272 6.1 x 10-16

- 0.0001 240 1.4 x 10-16

- 0.0001 199 1.1 x 10-17

- 0.0001 199 6.6 x 10-18

- 0.00001 272 2.0 x 10-16

- 0.00001 240 3.7 x 10-17

348 Bulk Diffusion - Quantitative DataHigher H diffusivities were observed in some solar-cell materials than in Czochralski orfloat-zone wafers. Secondary ion mass spectroscopic profiles of H or D, which had been

355


implanted at low energies and at temperatures ranging from 100 to 300C, were comparedfor various types of Si substrate. The data (table 48) could be described by theexpressions:

polycrystalline: D (cm2/s) = 0.0001 exp[-0.50(eV)/kT]Czochralski: D (cm2/s) = 0.0001 exp[-0.58(eV)/kT]float-zone: D (cm2/s) = 0.0001 exp[-0.56(eV)/kT]

It was found that the presence of O seemed to lower the grain-boundary diffusivities. Thediffusivity in Czochralski material was lower than that in float-zone material. It wassuggested that the lower diffusivity in the former case was also related to O. A newtechnique, which exploited the H decoration of dislocations was used to verify directlythe large diffusion depths in some solar-cell material. The higher H diffusivity permittedreverse-side hydrogenation of solar cells to be carried out in less than 0.5h; with asignificant improvement in cell performance.B.L.Sopori, K.Jones, X.J.Deng: Applied Physics Letters, 1992, 61[21], 2560-2

[446-106/107-138]

Table 45Effect of P and As upon H Diffusion in Si

Dopant Level D (cm2/s)As 0.0001 4 x 10-17

As 0.001 8 x 10-17

As 0.01 4 x 10-16

P 0.0001 1.2 x 10-16

P 0.001 2 x 10-16

P 0.01 2 x 10-16

349 Bulk Diffusion - Quantitative DataA new type of experiment was described which offered additional opportunities for theisolation and measurement of the basic energetic and kinetic parameters whichdetermined the behavior of interstitial H in semiconductors. The technique involved therelease of monatomic H at a sharply defined time, and in a known spatial pattern, byexposing dopant-H complexes (in the depletion layer of a Schottky diode) to a pulse ofminority carriers. Time-resolved measurements were then made of the capacitancetransient that arose from H migration, charge-state changes, and complex re-formation.The new technique was demonstrated by measuring the diffusion coefficient of H speciesin Si (table 49). Useful constraints were presented which were related to the energies ofthe H donor and acceptor levels, and to rates of spontaneous charge changes among H+,Ho, and H-.N.M.Johnson, C.Herring: Physical Review B, 1992, 46[23], 15554-7

[446-106/107-140]

350 Bulk Diffusion - Quantitative DataDepth profiles in heavily doped p-type material were obtained by analyzing the infra-redreflectance spectra of H-passivated samples. From these profiles, H diffusion coefficients

356


were calculated for various temperatures (table 50) and dopant concentrations. The resultswere explained by assuming that H diffusion was limited by trapping at acceptor sites. Abinding energy of 0.6eV was found for B-H complexes; in agreement with previouscalculations.C.P.Herrero, M.Stutzmann, A.Breitschwerdt, P.V.Santos: Physical Review B, 1990,41[2], 1054-8

[446-76/77-037]

Table 46Diffusivity of H in Amorphous Hydrogenated Si

Tannealing(C) tannealing(h)

D (cm2/s)

50 0.5 5.9 x 10-15

100 0.5 8.0 x 10-15

150 0.5 1.1 x 10-14

200 0.5 1.3 x 10-14

250 0.5 2.0 x 10-14

50 1 4.1 x 10-15

100 1 6.0 x 10-15

150 1 8.8 x 10-15

200 1 9.9 x 10-15

250 1 1.8 x 10-14

50 1.5 3.2 x 10-15

100 1.5 4.5 x 10-15

150 1.5 7.6 x 10-15

200 1.5 8.6 x 10-15

250 1.5 1.5 x 10-14

Bulk Diffusion - Quantitative DataEvidence was presented which showed that the H emission rate during exposure to anexternal atomic D source had the same activation energy (0.6 to 0.8eV) as that fordiffusion under the same conditions. This result was similar to that in the layer diffusioncase. This result, as well as H exchange with the sample during deuteration (and resultswhich indicated a narrow H emission distribution), could be explained in terms of anexchange mechanism in which mobile D exchanged places with bound H; with nosignificant activation barrier. It was also shown, by an analysis of H evolution curves, thatthe H chemical potential was pinned as H was lost from the sample. This result wasconsistent with the negative-U model for H bonding and H clustering.W.Jackson: Journal of Non-Crystalline Solids, 1993, 164-166[1], 263-8

[446-113/114-046]

Bulk Diffusion - Qualitative ObservationsData on D effusion from doped n-type or p-type polycrystalline material, when exposedto D2 plasma, were considered. The dependence of the effusion spectra upon dopant level,

357


passivation temperature, and pre-annealing treatment, was explained in terms of theexistence of various D-bonding configurations in passivated material.M.Stutzmann, M.S.Brandt: Journal of Applied Physics. 1990, 68[3], 1406-9

[446-86/87-053]

Table 47Diffusivity of Various H Species in n-Type and p-Type Si

Species Type T (C) D (cm2/s)Ho p 185 5.0 x 10-12

Ho p 185 3.6 x 10-12

Ho p 185 4.0 x 10-12

Ho p 165 7.5 x 10-13

Ho p 150 3.6 x 10-13

Ho p 150 3.5 x 10-13

Ho p 145 2.5 x 10-13

Ho p 120 4.0 x 10-14

H+ p 185 3.0 x 10-11

H+ p 185 1.0 x 10-12

H+ p 185 1.3 x 10-14

H+ p 165 1.3 x 10-11

H+ p 150 8.0 x 10-12

H+ p 150 5.5 x 10-13

H+ p 145 6.0 x 10-13

H+ p 120 2.5 x 10-12

Ho n 120 4.0 x 10-14

Ho n 150 3.6 x 10-13

H- n 150 4.0 x 10-12

H- n 120 7.5 x 10-13

H- n 120 1.0 x 10-13

H- n 120 5.0 x 10-15

Bulk Diffusion - Qualitative Observations - Concentration DependenceThe incorporation and diffusion of H in undoped plasma-deposited hydrogenatedamorphous films were studied, as a function of the H concentration, by means ofsecondary ion mass spectrometry profiling. The results showed that, between 1 and10at%, a structure-related H solubility limited the incorporation of H. The observedconcentration dependence of the H diffusion coefficient could be explained in terms of aband model which involved a rather mobile H chemical potential for material with a lowH concentration.W.Beyer: Physica Status Solidi A, 1997, 159[1], 53-63

[446-150/151-153]

358


Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe D effusion spectra of crystalline B-doped (2.5 x 1017 to 1019/cm3) material werestudied. By means of chemical etching of plasma-damaged surface layers, and secondaryion mass spectroscopic profiling, it was shown that the effusion rate appeared to belimited by surface desorption. It was suggested that the various effusion peaks could beattributed to specific H configurations.D.Ballutaud, P.De Mierry, J.C.Pesant, R.Rizk, A.Boutry-Forveille, M.Aucouturier:Materials Science Forum, 1992, 83-87, 45-50

[446-93/94-050]

Table 48Diffusivity of H in Various Types of Si

Type T (C) D (cm2/s)polycrystalline 305 1.6 x 10-8

polycrystalline 245 1.1 x 10-8



float-zone 305 6.1 x 10-10

float-zone 245 3.1 x 10-9

float-zone 200 6.1 x 10-10

float-zone 155 1.5 x 10-10





Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe tracer diffusion of D in hydrogenated amorphous material was studied by annealingsandwich structures of hydrogenated and deuterated material. The profiles could be fittedclosely by analytical solutions which were based upon a trap-controlled diffusion model.It was found that, for annealing times which were shorter than the deep-H release time,the D profiles followed the as-grown profiles, except for the appearance of exponentialwings at low concentrations. The wing amplitude increased almost linearly with time. Inthe long term, the solutions were identical to those for the trap-free case, but with aneffective diffusion coefficient that could be calculated from features of the short-termtracer profiles. Various parameters of H diffusion were measured, including the H releasetime from deep traps and the mean displacement of free H before re-trapping.H.M.Branz, S.Asher, B.P.Nelson, M.Kemp: Journal of Non-Crystalline Solids, 1993,164-166[1], 269-72

[446-113/114-044]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesA study was made of the effect of the deposition temperature and annealing treatmentupon H depth profiles in thin films of sputter-deposited hydrogenated amorphous

359


material. The results showed that higher deposition temperatures (285 to 300C) led to alower degree of H incorporation, and resulted in very steep H distribution profiles. The Hcontent at the film/substrate interface of such films could be equal to about 2/3 of that atthe film surface. It was suggested that marked H out-diffusion from the bulk took placeduring film growth at those relatively high temperatures. When in situ annealing wascarried out just after deposition, and before the samples reached room temperature, the Hdistribution through the film thickness had a more homogeneous profile. The resultantfilms also appeared to be more dense when the samples were subjected to this treatmentin an atomic H atmosphere.R.Rüther, J.Livingstone, N.Dytlewski, D.Cohen: Journal of Applied Physics, 1996, 79[1],175-8

[446-131/132-183]

Table 49Diffusivity of 2H- in Si

T (C) D (cm2/s)47 1.6 x 10-11

37 6.5 x 10-12

27 2.9 x 10-12

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe depth profiles of H which had been implanted into crystalline Si in randomdirections, to various fluences, were measured by using the 15N technique and secondaryion mass spectrometry. It was found that, whereas H which had been implanted to afluence of 1015/cm2 exhibited limited mobility, no such mobility was observed afterimplantation at higher fluences. In these cases, ballistic computer models were found todescribe the depth distributions to within experimental and theoretical accuracy.Annealing at temperatures of up to 510K did not change the H distribution. High-fluenceH implantation into the oxide was also examined. There was some sign of radiation-enhanced diffusion during implantation. During subsequent thermal annealing, the Hdiffused; probably via a trapping/detrapping mechanism that was associated with anOH/H2 transformation of the H bonding.D.Fink, J.Krauser, D.Nagengast, T.A.Murphy, J.Erxmeier, L.Palmetshofer, D.Braeunig,A.Weidinger: Applied Physics A, 1995, 61[4], 381-8

[446-131/132-184]

Bulk Diffusion - Qualitative Observations - Effect of ChargeIt was pointed out that H could migrate as a negatively charged species, H-. The evidencefor this was the combined observation of a strong electric-field dependence of theremoval rate of PH complexes during bias-temperature stressing of Schottky barrierdiodes, and the resultant redistribution of neutralized donors. The detection of H-

established that, as well as a previously identified deep-donor level, there also existed anacceptor level for H.

360


in the Si band gap. The dissociation kinetics of PH yielded an activation energy of1.18eV; from which a binding energy of between 0.35 and 0.65eV was deduced.J.Zhu, N.M.Johnson, C.Herring: Physical Review B, 1990, 41[17], 12354-7

[446-76/77-037]

Table 50Diffusivity of H in Si Containing 1.5 x 1019/cm3 of B

T (C) D (cm2/s)210 7.6 x 10-12

180 3.3 x 10-12

150 1.5 x 10-12

120 3.0 x 10-13

90 3.2 x 10-14

Bulk Diffusion - Qualitative Observations - Effect of ChargeIt was demonstrated that a periodic exposure to zero bias during the in situhydrogenation of reverse-biased p-type Schottky barrier structures had marked effectsupon H penetration. The H influx could be slowed, or even stopped, by such treatments.On the other hand, similar pulsing techniques produced almost no change in thepenetration of n-type barriers during hydrogenation. The latter observation contrastedsharply with the expectation that charge conversion from H+ to H- would reverse the driftof the H species. It was suggested that these effects were caused by the charge conversionof relatively immobile H-related defects. In the case of p-type barriers, this resulted in aweakening or reversal of the near-surface electric field, thus effectively stopping the driftof H+ into the bulk.C.H.Seager, R.A.Anderson: Journal of Applied Physics, 1996, 80[1], 151-5

[446-136/137-129]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe migration and acceptor neutralization of H in monocrystalline material was studied.Spreading resistance measurements showed that surface, and sub-surface, disorderedregions directly inhibited H penetration. These effects were independent of the type ofdisorder and the hydrogenation technique. Secondary ion mass spectrometry profiles ofdeuterated samples confirmed that D migration through the disordered regions wassuppressed. It was found that annealing of these hydrogenated damage regions resulted inthe generation of an acceptor deactivation profile which persisted at temperatures of up to800C and for durations of up to 60s. This sustained deactivation effect resulted in achange, in the free carrier concentration, by as much as 4 orders of magnitude. The resultsstrongly suggested that H-soaked damage regions acted as a source of atomic H duringrapid thermal annealing.K.Srikanth, S.Ashok: Journal of Applied Physics, 1991, 70[9], 4779-83

[446-93/94-051]

361


Bulk Diffusion - Qualitative Observations - Effect of DefectsA review was presented of the authors' studies of the physiochemical behavior of P, C,and H in Si. The results covered the diffusion and segregation of impurities inmonocrystalline and polycrystalline material, the passivation of recombining defects byH, and H-dopant interaction. More detailed information was given on H diffusion. Theresults were discussed in terms of the existence of complex mechanisms of interactionbetween H and impurities or defects.A.Chari, P.De Mierry, A.Menikh, M.Aucouturier: Revue de Physique Appliquée, 1987,22[7], 655-62

[446-55/56-040]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe diffusion of H in molecular beam deposited amorphous material, or in glow dischargefilms, which had been grown at temperatures below 200C was studied. It was concludedthat the fast diffusion which was observed was due to the presence of the micro-voids andcolumnar structural defects which were typical of these materials.V.Petrova-Koch, H.P.Zeindl, J.Herion, W.Beyer: Journal of Non-Crystalline Solids,1987, 97-98, 807-10

[446-60-014]

Bulk Diffusion - Qualitative Observations - Effect of DefectsAn infra-red and secondary ion mass spectrometric study was made of diffusion inundoped, radio-frequency sputtered, multi-layered, hydrogenated or deuterated,amorphous material. The results indicated that the long-range motion of atomic H wassuppressed when the micro-void content exceeded a critical value that was related to theinitial SiH2 and SiH3 content. A marked power-law time dependence (exponent = 0.75) ofthe diffusion constant, when below the critical value of 7at%, was explained in terms ofH migration involving mobile intrinsic defects and an exponential distribution of traps forH atoms or defects.J.Shinar, R.Shinar, S.Mitra, J.Y.Kim: Physical Review Letters, 1989, 62[17], 2001-3

[446-70/71-125]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe diffusion of H in radio-frequency sputter-deposited undoped or B-doped amorphousmaterial was studied by means of secondary ion mass spectrometric profiling of annealedmulti-layers. It was found that long-range atomic H motion appeared to be suppressed inundoped amorphous hydrogenated material with an excess micro-void content. However,in amorphous hydrogenated multi-layers which exhibited diffusion, the diffusion constantexhibited a power-law time dependence:

D(t) = kDo(wt)-n

where n was between 0.2 and 0.9. These results were explained in terms of relaxationprocesses and the dynamics of floating and dangling bonds in amorphous hydrogenatedSi.R.Shinar, S.Mitra, X.L.Wu, J.Shinar: Journal of Non-Crystalline Solids, 1989, 114[1],220-2

[446-72/73-030]

362


Bulk Diffusion - Qualitative Observations - Effect of DefectsThe diffusion coefficient of the dominant (g = 2.0055) paramagnetic defect (D-center) inamorphous material was found to be less than half as high as that for H. High D-centerdensities were found to trap H. These results removed the possibility that the D-centerwas responsible for H motion in hydrogenated amorphous Si.W.B.Jackson, C.C.Tsai, R.Thompson: Physical Review Letters, 1990, 64[1], 56-9

[446-76/77-038]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe elastic recoil analysis technique was used to measure the H diffusion coefficient at400C in 5 undoped thin films of amorphous hydrogenated Si and crystalline Si. Thedeposition temperature of the films ranged from 50 to 250C, and the correspondingnominal H concentrations ranged from 18 to 11at%. After annealing the samples, the Hconcentration profiles were measured and permitted the diffusion coefficient to bedetermined. It was found that the higher the deposition temperature, the larger was thediffusion coefficient. It was suggested that this behavior could be attributed to variationsin the cluster concentration in the films.X.M.Tang, J.Weber, Y.Baer, F.Finger. Solid State Communications, 1990, 74[3], 171-4

[446-76/77-039]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe migration of H in B-doped radio-frequency sputter-deposited hydrogenatedamorphous material was deduced from the secondary ion mass spectrometry profiles of Din annealed a-Si:H/a-Si:(H,D)/a-Si:H multi-layers. The exponent in the diffusion constant:

D = Do(wt)-a

(where w was an attempt frequency and Do was a constant) was equal to 0.7, anddecreased markedly after annealing for about 50h at 180C or about 35h at 225C. It thenincreased to between 0.8 and 1. That is, diffusion was almost quenched by annealing forup to 1100h. This behavior was attributed to the effect of modifications (in the micro-voidsystem) that were related to structural relaxation of the network.S.Mitra, R.Shinar, J.Shinar: Physical Review B, 1990, 42[10], 6746-9

[446-78/79-054]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe behavior of B-H complexes, and the diffusion of H in B-doped material, wereanalyzed by using Raman scattering and infra-red reflection spectroscopy. It was foundthat, at temperatures below 200C, H diffusion was controlled by trapping at acceptorsites. This became negligible at higher temperatures. Changes in the zone-center opticalphonon of Si, and in the vibrational local modes of B, occurred after H passivation. The Hand B local modes were studied as a function of temperature and an external uniaxialstress. Analysis of H vibrational modes under stress revealed a non-trigonal symmetry ofthe B-H complexes at 100K, and a high mobility of H in these complexes. The resultswere found to be compatible with a bond-minimum site model for H at low temperatures.However, under stress and at high temperatures, off-bond H positions were suggested to

363


exist. The stability of B-H complexes was also analyzed, and a dissociation energy of0.6eV was deduced.C.P.Herrero, M.Stutzmann, A.Breitschwerdt: Physical Review B, 1991, 43[2], 1555-75

[446-81/82-049]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe behavior of H in P-doped material was investigated by monitoring shallow donor P,using electron spin resonance techniques. Appreciable broadening in motionally narrowedelectron spin resonance lines was first observed in samples which had been treated with Hplasma. It was found, from the donor-concentration dependence, that the broadening wascaused by Fermi-contact hyperfine interactions between H nuclear spins and donor orconduction electrons. The annealing of H-passivated material indicated that anintermediate H state with a P donor neighbor was formed via the dissociation of P-Hcomplexes at temperatures of between 150 and 350C. It was deduced that H diffused in n-type Si at temperatures greater than 350C.K.Murakami, H.Suhara, S.Fujita, K.Masuda: Physical Review B, 1991, 44[7], 3409-12

[446-84/85-070]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe density of dangling bond defects in plasma-deposited amorphous hydrogenatedmaterial was suggested to depend mainly upon the deposition temperature. This, in turn,was due to its effect upon the H diffusivity. It was found that the optimum temperaturewas equal to about 62% of the bulk-bonded H evolution temperature. Overall, it wasproposed that H diffusion in amorphous hydrogenated Si-based alloys occurred mainlyvia the release of H, from Si-H bonds, to interstitial H sites which were associated withthe remaining weak Si-Si bonds. The maximum activation energy for H diffusion wasexpected to be 3.1eV.J.Robertson: Applied Physics Letters, 1991, 59[26], 3425-7

[446-84/85-104]

Bulk Diffusion - Qualitative Observations - Effect of DefectsAn optically detected electron nuclear double resonance signal was detected inamorphous hydrogenated material which had been prepared under 2 different sets ofconditions. Differing changes in the line-shape were observed after light exposure of thetwo types of sample. A similar line-shape change was observed upon applying pressure. Itwas proposed that probable explanations for the line-shape change included light-inducedH migration, and an effect of dangling bonds upon spin diffusion.M.Kondo, K.Morigaki: Journal of Non-Crystalline Solids, 1993, 164-166[1], 227-30

[446-115/116-153]

Bulk Diffusion - Qualitative Observations - Effect of DefectsAn attempt was made to explain H multi-trapping and H2 molecule formation in P-dopedand B-doped material. The model which was developed assumed the existence of ametastable configuration (of the H-donor complex) which had a donor character, as wellas the ability of donor atoms to trap several H ions (by Coulombic attraction). This led tothe prediction of microscopic mechanisms, for H trapping in P-doped Si, which were

364


more efficient than those proposed for B-doped Si. It was concluded that this accountedfor the differing atomic H profiles and diffusivities which were observed in the 2materials. These microscopic mechanisms also provided useful guidance for theimprovement of phenomenological models for H diffusion. The existence of a metastablecomplex configuration which had donor characteristics agreed well with published resultson the annealing of hydrogenated Schottky diodes which were made from P-doped Si.A.A.Bonapasta: Physical Review B, 1992, 46[16], 10119-26

[446-93/94-054]

Bulk Diffusion - Qualitative Observations - Effect of DefectsExperiments demonstrated that H migration in hydrogenated amorphous material wascontrolled by the concentration of electronic carriers. Therefore, it was stronglysuppressed when the carriers were removed by an electric field. The few H atoms whichmigrated in a region that was depleted of carriers had a longer mean free path fordiffusion; due to a reduction in the concentration of trapping centers. The latter wereprobably Si dangling bonds.P.V.Santos, N.M.Johnson: Applied Physics Letters, 1993, 62[7], 720-2

[446-106/107-137]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe neutral vacancy (V) and V,Hn complexes, where n = 1 to 4, were studied invarious molecular clusters at the approximate ab initio Hartree-Fock level, using postHartree-Fock corrections in electron correlation. Calculations were made of theequilibrium configurations and electronic structures, dissociation energies, diffusion pathsand activation energies. The dissociation energies were compared with those, of other Htraps, which had been calculated by using the same level of theory. These traps includedsubstitutional C, interstitial Ti, or B,H pairs. The results predicted that the V,H1 pairshould be mobile above room temperature. A part of the barrier for the diffusion ofV,H1 was lower than that for V, due to a mechanism that was analogous to the H-enhanced diffusion of interstitial O in Si. Another part of this barrier was higher than thatfor V. The effects of this included the possibility of enhanced H diffusion inpolycrystalline, as compared with crystalline, Si and suggested a mechanism for thenucleation of platelets in the sub-surface region of plasma-exposed Si.M.A.Roberson, S.K.Estreicher: Physical Review B, 1994, 49[24], 17040-9

[446-115/116-153]

Bulk Diffusion - Qualitative Observations - Effect of DefectsEffusion experiments and secondary ion mass spectrometric profiling were carried out onpost-hydrogenated (or deuterated) micro nano-crystallized films that were obtained by thethermal annealing of amorphous sputtered layers. Analysis of the effusion spectra andsecondary ion mass spectrometry profiles revealed the existence of cavities that containedmolecular H, the presence of weakly-bonded H in small clusters, and of H that wastrapped at grain boundaries.L.Lusson, P.Elkaim, A.Correia, D.Ballutaud: Journal de Physique III, 1995, 5[8], 1173-84

[446-125/126-150]

365


Bulk Diffusion - Qualitative Observations - Effect of Electric FieldBy using the elastic recoil detection technique, measurements were made of H profiles ina structure which consisted of amorphous hydrogenated material on an amorphous Sisubstrate. The samples were annealed at various temperatures, with or without anelectrical bias. The results showed that H diffusion in amorphous Si was appreciablyenhanced by the imposition of an electrical bias. The existence of excess carriers whichwere introduced by electrical injection was considered to be responsible for theenhancement of H diffusion.Z.Song, F.Zhang, G.Yu, G.Kong, G.Chen: Journal of Non-Crystalline Solids, 1993, 164-166[1], 305-8

[446-113/114-045]

Bulk Diffusion - Qualitative Observations - Effect of Electric FieldDiffusion experiments were performed on amorphous hydrogenated p-i-n diodes whichwere subjected to a constant or pulsed reverse electric field. Under a reverse field, themean free path for the diffusion of interstitial H was enhanced by a factor of at least 4.The present results indicated that the enhancement was not due to the field-drift ofcharged H atoms, but rather to a reduction in the electronic carrier density under reversebias. The carriers controlled the capture rate of interstitial H into Si-H traps and thereforecontrolled the lifetime and the mean free path for diffusion via interstitials.P.V.Santos, N.M.Johnson, R.A.Street: Journal of Non-Crystalline Solids, 1993, 164-166[1], 277-80

[446-113/114-048]

Bulk Diffusion - Qualitative Observations - Effect of IlluminationMeasurements were made of the variation of the dispersion parameter of H diffusion inundoped amorphous hydrogenated material at 350C, using elastic-recoil detectionanalysis, as a function of illumination intensities of 8 or 15W/cm2. The dispersionparameter value in the illuminated case was compared with the dark value, and it wasfound that it increased with illumination intensity. This was thought to be the firstobservation of the effect.O.Greim, J.Weber, Y.Baer, Y.Ziegler: Solid State Communications, 1995, 93[9], 719-23

[446-121/122-087]

Bulk Diffusion - Qualitative Observations - Effect of IlluminationExperimental results were presented which demonstrated that H diffusion was enhancedby illumination, and that the enhancement was suppressed when photo-generated carrierswere removed from the diffusion region by using an electric field. The results suggestedthat there was a close relationship between H diffusion and metastability in amorphoushydrogenated material.P.V.Santos: Journal of Physics - Condensed Matter, 1993, 5[SA], A335-6

[446-109/110-045]

Bulk Diffusion - Qualitative Observations - Effect of IlluminationThe relationship between electronic carriers, metastable defects, and H motion inamorphous hydrogenated material was investigated by performing photoconductivity and

366


H diffusion experiments at temperatures above 200C. The photoconductivity decayedwith illumination time, due to defect formation. The intensity dependence of thephotoconductivity transients indicated that the defect formation mechanism was similar tothat which operated at lower temperatures. The photoconductivity data were used toanalyze the dependence of the diffusion coefficient upon continuous or pulsed radiation.P.V.Santos, M.S.Brandt, R.A.Street, M.Stutzmann: Journal of Non-Crystalline Solids,1993, 164-166[1], 273-6

[446-113/114-045]

Bulk Diffusion - Qualitative Observations - Effect of IlluminationHydrogenated amorphous films were annealed (220 to 270C, 24 to 48h) under intensevisible light (4 to 16W/cm2) or in the dark. After annealing, the H concentration profilewas measured by means of Rutherford back-scattering spectrometry and elastic recoildetection ion-beam analytical methods. A model was proposed which showed that, ingood agreement with the present results, the H diffusion coefficient was proportional tothe illuminating power and to the loosely bonded H concentration.O.Greim, J.Weber, Y.Baer, U.Kroll: Physical Review B, 1994, 50[15], 10644-8

[446-119/120-224]

Bulk Diffusion - Qualitative Observations - Effect of IlluminationEvidence for the participation of atomic H in the light-induced metastability ofhydrogenated amorphous material was observed by means of low-temperature infra-redtransmission spectroscopy and room-temperature infra-red phase-modulated ellipsometry.A band at 1730/cm, accompanied by an increase in the amplitude of the bending mode ofa new broad band at about 870/cm, was observed after 1h of intense illumination of anannealed amorphous hydrogenated sample. A shift of this band to lower frequencies wasobserved during prolonged light-soaking, and the band disappeared after about 4h ofillumination. At the same time, the density of the atomic H which was bonded to Sidecreased; as deduced from the area under the broad band at about 640/cm. The resultscould be interpreted in terms of a H diffusion model in which the H atom of a Si-H bondnext to a weak Si-Si bond diffused so as to form a 3-center bond at intermediateillumination times. It then diffused further so as to form a stable Si-H bond at longertimes.R.Darwich, P.Roca i Cabarrocas, S.Vallon, R.Ossikovski, P.Morin, K.Zellama:Philosophical Magazine B, 1995, 72[3], 363-72

[446-125/126-150]

Bulk Diffusion - Qualitative Observations - Effect of IlluminationThe dependences of the dispersion parameter and time-dependent diffusion parameterupon the annealing temperature were investigated by means of elastic-recoil detectionanalyses of undoped amorphous hydrogenated material. It was found that the dark valueof the dispersion parameter increased with the annealing temperature, as previouslyobserved, and that the illuminated dispersion parameter also increased. However, it wasobserved that the time-dependent diffusivity under illumination deviated from Arrheniusbehavior, whereas the time-dependent diffusivity that was measured in the dark exhibited

367


an Arrhenius behavior. The H diffusion under intense illumination was explained in termsof existing models.O.Greim, J.Weber, Y.Baer, Y.Ziegler: Solid State Communications, 1996, 97[2], 109-14

[446-134/135-161]

Bulk Diffusion - Qualitative Observations - Effect of IlluminationIt was found that high-intensity (9W/cm2) red-light soaking (120h) of amorphoushydrogenated material at 65C yielded no detectable H diffusion in tracer experiments. Anew upper bound on the light-induced diffusion coefficient was found at a temperaturethat was so low that thermal diffusion was negligible. The null result which was foundhere was incompatible with models in which H emission from Si-H bonds was assumed tobe proportional to both the light intensity and the metastable defect creation rate.However, this result was compatible with a model in which both H emission andmetastable defect creation were proportional to the product of the free electron and holedensities. This implied that fewer than 500 H emissions occurred per metastable defectthat was createdH.M.Branz, J.Bullock, S.Asher, H.Gleskova, S.Wagner: Journal of Non-CrystallineSolids, 1996, 198-200[1], 441-4

[446-138/139-107]

Bulk Diffusion - Qualitative Observations - Effect of IlluminationDirect observations were made of light-enhanced H migration, in hydrogenatedamorphous material, by means of secondary ion mass spectrometry. It was found that theenhancement of diffusion increased with the intensity of illumination in undopedmaterial, and was suppressed in doped and compensated material. The enhancement wasattributed to an increased release rate of H from Si-H bonds in the presence of photo-generated carriers.P.V.Santos, N.M.Johnson, R.A.Street: Physical Review Letters, 1991, 67[19], 2686-9

[446-88/89-052]

Bulk Diffusion - Qualitative Observations - Effect of Ion BombardmentAn in-line mass spectrometer, and Langmuir probes, were used to study the plasma-immersion H passivation of grain-boundary defects in polycrystalline thin-filmtransistors. The relative fluxes of H+ and H2

+, as well as the total ion-current density, weremeasured at the substrate location in an electron cyclotron resonance H discharge.Measurements were made within the range of operating conditions where passivationrates were known to vary widely. The resultant data exhibited a strong correlation of bothH+ flux and ion bombardment energy, with good transistor performance being obtained atoperating pressures below 0.001Torr. It was suggested that discharge operating conditionswhich promoted the dissociation of H2 to form H and H+ (which could diffuse morerapidly through the solid material than did H2) were important.E.S.Cielaszyk, K.H.R.Kirmse, R.A.Stewart, A.E.Wendt: Applied Physics Letters, 1995,67[21], 3099-101

[446-127/128-158]

368


Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe migration of atomic H in material which had been damaged by Ar implantation wasstudied. The Ar was implanted with 2 distinct doses which straggled the amorphizationthreshold. Atomic H was then introduced via low-energy ion implantation. Thedeactivation of dopant B atoms by atomic H was markedly reduced in wafers which hadbeen subjected to low-energy Ar-ion implantation. Trapping of H in the defect sites whichwere generated by Ar implantation, and perhaps the formation of molecular H in theimplanted region, hindered H permeation into the bulk.S.Ashok, K.Srikanth: Journal of Applied Physics, 1989, 66[3], 1491-4

[446-74-055]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe behavior of H ion-implanted crystalline material was investigated, at the ns time-scale and at temperatures up to the melting point, by using pulsed laser annealing. Laserreflectivity, scanning electron microscopy, and surface profilometry methods were used tocharacterize the implantation and annealing effects. Computer modelling of H release, asa function of laser energy, was used to interpret the data. Desorption of H which had beenimplanted at 1 or 2keV occurred at temperatures above 1000K, with no extensive surfacedeformation. This desorption could be fitted to de-trapping curves by using an activationenergy (of about 2eV) that slowly decreased for H/Si ratios which ranged from 0.04 to0.27. Contrary to expectation, no diffusion limitation was observed. When implanted at 5or 10keV, for H/Si ratios greater than 0.2, H was released at temperatures greater than550K (via blister rupture). It was found that, in spite of differences between the effects oflow and high implantation energies, a unified picture emerged which posited the existenceof a layer with a high-temperature H mobility that was greater than that of ordinaryatomic diffusion.R.Boivin, B.Terreault: Journal of Applied Physics, 1993, 73[4], 1943-51

[446-106/107-140]

Bulk Diffusion - Qualitative Observations - Effect of LayersHydrogenated amorphous material was prepared by evaporation onto low-temperaturesubstrates. It was typified by a low-density network and by the presence of (SiH2)n bonds.By monitoring the decay of the small-angle neutron-scattering intensity during theannealing of Si/Si:H/Si/Si:M/ ... (M = H28D72) multilayers, it was possible to followsimultaneously the densification of the Si network and the diffusion of H atoms.M.Vergnat, S.Houssaïni, G.Marchal, P.Mangin, C.Vettier: Physical Review B, 1993,47[12], 7584-7

[446-106/107-138]

Bulk Diffusion - Qualitative Observations - Effect of Pipe DiffusionThe evolution of H was studied by using gas effusion spectroscopic and infra-redspectroscopic studies of multi-layer films of amorphous hydrogenated Si and Si3N4. TheH effusion mechanisms in 3 temperature regimes were explained in terms of pipediffusion, and the diffusion was limited by the diffusion coefficients of the barrier andwell layers. The effective thickness of the interfacial region was estimated to be about2nm.

369


S.Nonomura, M.Ohta, H.Suzuki, T.Muraki, S.Kusakabe, S.Nitta: Journal of Non-Crystalline Solids, 1991, 137-138, 1139-42

[446-88/89-054]

Bulk Diffusion - Qualitative Observations - Effect of RelaxationIt was found that the stretched-exponential relaxation which was commonly observed indisordered systems could be explained by time-dependent atomic diffusion. Relaxationwas observed in the electronic properties of hydrogenated amorphous samples (a so-called H-glass) and reflected the equilibrium of localized electronic states. The relaxationwas attributed to the motion of bonded H which exhibited dispersive diffusion with acharacteristic power-law time dependence. A quantitative relationship between relaxationand diffusion was established.J.Kakalios, R.A.Street, W.B.Jackson: Physical Review Letters, 1987, 59[9], 1037-40

[446-55/56-039]

Bulk Diffusion - Qualitative Observations - Effect of TrappingSecondary ion mass spectrometry measurements confirmed that the rate of H diffusion inp-type material was higher than that in n-type material. Here, the difference was not aslarge as that observed in the case of ion-implanted specimens. This discrepancy wasattributed to the effect of H capture at the boundaries of implanted layers in the lattercase.E.M.Omelyanovskii, A.V.Pakhomov, A.J.Polyakov, O.M.Borodina: Fizika i TekhnikaPoluprovodnikov, 1989, 23[1], 178-80 (Soviet Physics - Semiconductors, 1989, 23[1],112-3)

[446-70/71-124]

Bulk Diffusion - Qualitative Observations - Effect of TrappingWave-functions and energy spectra were calculated for an isolated proton in a crystallinematerial. The ground-state wave-function was strongly localized at the minimum energypositions near to the bond centers. It was noted that so-called M-sites, which had almostthe minimum energy, had only a small probability of occupation. The lower-energyexcited states were also well-localized. Overall, it was deduced that quantum effects wereunimportant with regard to H diffusion through the host.C.Pennetta: Europhysics Letters, 1991, 14[7], 683-8

[446-86/87-054]

Bulk Diffusion - Qualitative Observations - Effect of TrappingThe diffusive motion of H was used to investigate H-trapping in H-depleted amorphoussamples, and to determine an approximate H-diffusion density of states. The diffusionprofiles revealed clear evidence of deep traps which were separate from shallow traps,and the results could be explained in terms of a simple division of the H states into deeptraps, shallow traps, and transport states. The concentration of deep traps was between 8 x1019 and 2 x 1020/cm3; of which about 30% could be identified with dangling bonds. Theenergy of the deep traps was at least 1.9eV below the transport states. The diffusion wasdispersive, with a power-law time dependence, and could be characterized by anexponential distribution of hopping barriers with a width of about 0.09eV. The shallowtraps were identified with clustered H pairs which determined the H chemical potential at

370


high H concentrations. It was suggested that the results were consistent with variouspossibilities. One extreme was the case where H was bonded mainly at void surfaces andthe transport energy was very different in amorphous material, as compared withcrystalline material. The other extreme was the case where H resided mainly in plateletstructures, and the transport energy was essentially the same as in crystalline material.The actual situation depended upon the deposition conditions.W.B.Jackson, C.C.Tsai: Physical Review B, 1992, 45[12], 6564-80

[446-86/87-055]

Bulk Diffusion - Qualitative Observations - Effect of TrappingIt was shown that the multiple trapping of H atoms at shallow acceptors during plasmahydrogenation could explain various features of the resultant H diffusion profile. In thecase of heavily-doped p-type material, a plateau in the H profile at the acceptorconcentration was preceded by a high concentration plateau near to the surface. It wasdemonstrated that the plateau could be explained by the successive aggregation of Hatoms at acceptors, and that the multiple trapping process appeared to terminate after theagglomeration of 8 to 12 H atoms. The reaction cross-section for multiple trapping wasmuch smaller than that for the first capture, and this behavior was consistent with theionic capture of H+ at acceptors, followed by neutral multiple capture. It was suggestedthat the formation of H dimers, which predominated in high-resistivity material, mightstill be important with regard to the high-concentration near-surface profiles which wereseen in heavily-doped samples.J.T.Borenstein, J.W.Corbett, S.J.Pearton: Materials Science Forum, 1992, 83-87, 51-6

[446-93/94-055]

Bulk Diffusion - Qualitative Observations - Effect of TrappingDiffusion in amorphous hydrogenated material, as studied by using implantation andsecondary ion mass spectroscopic techniques, was found to be trap-limited. It wassuggested that H probably diffused in a neutral charge state and was trapped by intrinsic(Si bonding) sites.W.Beyer, U.Zastrow: Journal of Non-Crystalline Solids, 1993, 164-166[1], 289-92

[446-113/114-047]

Bulk Diffusion - Qualitative Observations - Effect of TrappingThe relationship between electronic carriers and H migration in amorphous hydrogenatedmaterial was investigated by using secondary ion mass spectrometry to measure Ddiffusion profiles in the intrinsic layer of p-i-n photodiodes. The carrier concentration inthe i-layer was controlled by varying the temperature, the illumination intensity, or thebias which was applied to the devices. It was demonstrated that H migration wascontrolled by an electronic mechanism, and was enhanced when the carrier populationwas increased by illumination. It was suppressed when the population was reduced tobelow the thermal equilibrium value by applying a reverse bias to the diodes. This effectwas attributed to the dependence upon carrier density of the dissociation rate of H fromSi-H bonds and into the diffusion path, which consisted of interstitial sites. In addition,the migration length in the diffusion path increased under a reverse bias. The enhancedmigration was associated with a decrease in the effective density of traps for H in a

371


carrier-depleted layer. The trap density under these conditions was close to the dangling-bond density. This suggested that the migration length was governed by capture into thesedefects.P.V.Santos, N.M.Johnson, R.A.Street, M.Hack, R.Thompson, C.C.Tsai: Physical ReviewB, 1993, 47[16], 10244-60

[446-106/107-141]

Bulk Diffusion - Qualitative Observations - Effect of TrappingThe effects of the H concentration and of the thermal history upon H-trapping inhydrogenated amorphous material were investigated by analyzing D diffusion profiles.The unexpected results indicated that the number of traps increased with H concentration,while the trap depth increased upon annealing. The H which was equilibrated within thefilm was associated with traps, while rapidly introduced H was not. The results wereconsistent with a model of trapping and release from H clusters which nucleated and grewin response to added H. It was proposed that the effective diffusion coefficients for highconcentrations of H in amorphous, polycrystalline, or monocrystalline Si were determinedby these clusters.W.B.Jackson, P.V.Santos, C.C.Tsai: Physical Review B, 1993, 47[15], 9993-6

[446-106/107-141]

Bulk Diffusion - Theoretical Analysis - Effect of TunnellingQuantum transition-state theory, based upon the path-integral formalism, was used tostudy the jump rate of atomic H and D in crystalline Si. This technique provided a methodfor studying the effect of vibrational mode quantization and quantum tunnelling upon theimpurity jump rate. The atomic interactions were modelled by using potentials which hadbeen fitted to earlier ab initio pseudopotential calculations. The Si nuclei were treated asquantum particles, up to second-nearest neighbors of the impurity. The H jump rateobeyed an Arrhenius law, and could be described by classical transition-state theory attemperatures above 100K. At about 80K, a change in the slope of the Arrhenius plot wasobserved in the case of H. This was attributed to the onset of a diffusion regime that wascontrolled by phonon-assisted tunnelling of the impurity. In the case of D, no change inslope was observed at temperatures ranging down to 40K.C.P.Herrero: Physical Review B, 1997, 55[15], 9235-8

[446-150/151-153]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsLayers of Ga-implanted p+ material were exposed to atomic H from a plasma. It wasfound that H concentrations which were up to 7.5 times greater than the peak Gaconcentration (7 x 1019/cm3) segregated into the p+ layer during treatment at 200C. Theshape of the H concentration profile was similar to that of the Ga profile. Ion channellingexperiments showed that the H atoms did not occupy simple high-symmetry sites in thelattice. The use of electron microscopy revealed the presence of extended (111)-typestacking fault defects which were associated with the layer of high H content.A.D.Marwick, G.S.Oehrlein, M.Wittmer: Applied Physics Letters, 1991, 59[2], 198-200

[446-84/85-070]

372


Bulk Diffusion - Qualitative Observations - Enhanced DiffusionHydrogen was introduced, using a radio frequency plasma, into Czochralski material at275C. Most of the H was trapped near to the surface, where it formed Si-H bonds, while asmall fraction diffused into the Si. This fraction enhanced O-related thermal donorformation rates in a diffusion-like profile during subsequent furnace annealing attemperatures of between 350 and 400C. It was found that a H concentration that wasequal to only a few percent of the O concentration was sufficient to enhance the thermaldonor formation rate; thus indicating a H-catalyzed process. The maximumconcentrations of thermal donors after annealing at 400C exceeded those for retained H.A mechanism for H diffusion through O traps, and correlated H-promoted O diffusion,was proposed in order to explain the enhanced thermal donor formation rates.H.J.Stein, S.Hahn: Journal of Applied Physics, 1994, 75[7], 3477-84

[446-117/118-196]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionMeasurements were made of the dipolar spin lattice relaxation time of 1H in amorphoushydrogenated material in order monitor the local motion of H. The variation in relaxationtime with doping level revealed a trend which was similar to the changes in H diffusivitythat were measured by means of secondary ion mass spectroscopy. The temperaturevariation of the dipolar spin lattice relaxation time indicated that, if the local motion wasthermally activated, the activation energies were at least an order of magnitude lower thanthose measured using secondary ion mass spectrometry. The so-called diffusioncoefficients for local H motion that were deduced from the relaxation time measurementswere many orders of magnitude higher than the diffusion constants that were deducedfrom secondary ion mass spectrometric data. It was noted that the relaxation timemeasurements were performed at the ms scale, while the secondary ion massspectrometric measurements were performed over many hours. A phenomenologicalmodel was developed which provided a reasonable rationalization of these 2 types ofdata.P.Hari, P.C.Taylor, R.A.Street: Journal of Non-Crystalline Solids, 1996, 198-200[1], 52-5

[446-138/139-104]

Bulk Diffusion - Qualitative Observations - Positron AnnihilationPositron lifetime studies were made of films of hydrogenated amorphous materialprepared using the glow-discharge method. Films which had been deposited andthermally annealed at various temperatures were investigated in order to determinemicrostructural parameters such as the contents of H, micro-voids, and vacancies. Theappearance of a long-lifetime (more than 3ns) component in the lifetime spectrum,together with a narrow peak in the two-dimensional angular correlation of positron-annihilation radiation, confirmed the existence of large micro-voids in the films.Variations in the intensity of the long-lifetime component, as a function of the depositionand annealing temperatures, was studied in detail. This clearly showed that molecular Hexisted at high pressures in the micro-voids and effused out at high temperatures, to leavebehind empty micro-voids in the film. Two stages of H effusion at 275 and 600C wereclearly identified in films deposited at 25C. Films deposited at 300C exhibited only thehigh-temperature effusion stage and indicated that the low-temperature stage was relatedto trapped molecular H, while the high-temperature (600C) one was related to bonded H.

373


The positronium lifetime increased with annealing temperature, reflecting micro-voidgrowth, and was attributed to the onset of an agglomeration process in parallel with Heffusion. Information concerning the presence of quadri-vacancies and penta-vacancies inthe films and their response to heat-treatment was also deduced from the intensity andlifetime corresponding to the trapped positron state. An attempt was made to correlate thepositronium component with electron spin resonance results.V.G.Bhide, R.O.Dusane, S.V.Rajarshi, A.D.Shaligram, S.K.David: Journal of AppliedPhysics, 1987, 62[1], 108-16

[446-55/56-040]

Bulk Diffusion - Qualitative Observations - Time DependenceThe time-dependent diffusion constant, D = Do(ω t)-α, of H in radio-frequency sputter-deposited hydrogenated amorphous films which contained 3 to 5at% of Si-bonded H, wasfound to increase with time at temperatures of between 300 and 380C. That is, thedispersion parameter, α, was negative. As in previous studies, α generally decreased withtemperature up to a sample-dependent temperature, and then increased at highertemperatures.R.Shinar, J.Shinar, H.Jia, X.L.Wu: Physical Review B, 1993, 47[15], 9361-5

[446-106/107-141]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesThe diffusivity of H in crystalline material, deduced from profiling experiments involvingnuclear resonance retention and secondary ion mass spectroscopy, was found to be 3 to 9orders of magnitude smaller than the previously accepted value. Several facts which wereoften overlooked, when analyzing profiling measurements, were pointed out. A limitedflux model was used to explain the observed results. The predictions of the model weresupported by further experiments.B.Y.Tong, X.W.Wu, G.R.Yang, S.K.Wong: Canadian Journal of Physics, 1989, 67[4],379-83

[446-70/71-124]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesIt was shown that the exponential depth profiles which were sometimes observed duringH diffusion into semiconductors could be explained by including a term (in the diffusionequation) that described the multiple trapping of H at an impurity. It was shown that theeffective dimensionality of the random walk in the present case was infinite.D.A.Tulchinsky, J.W.Corbett, J.T.Borenstein, S.J.Pearton: Physical Review B, 1990,42[18], 11881-3

[446-81/82-010]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesIt was shown that depth profiles with plateaux, which were sometimes detected when Hwas diffused, could be explained by including terms (in the diffusion equation) whichdescribed the various manners in which H could interact with impurity atoms. It wassuggested that such plateau-type profiles could result from a loss of system stability, dueto local interactions, and from diffusional exchange between neighboring volumes due to

374


the appearance of long-range correlations. It was found that the behavior of the systemdepended upon the dimensionally of the random walk via which the H diffused. Changesin temperature and impurity concentration altered the parameters of the random walk, andthus changed the nature of the diffusion process.J.W.Corbett, I.V.Verner: Materials Science Forum, 1992, 83-87, 57-62

[446-93/94-054]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesA comprehensive kinetic model was developed for the diffusion of H atoms into B-dopedSi during low-temperature plasma hydrogenation. The model accounted for severalcommonly observed features of the secondary ion mass spectrometry profiles ofhydrogenated p-type samples with various resistivities. In high-resistivity material, the Hprofile was explained by invoking H dimer formation which, for short plasma exposuretimes, appeared to obey steady-state kinetics. The H profiles in more heavily doped p-type samples, which were often assumed to be controlled by a single pairing reactionbetween H+ ions and dopant sites, were shown to be consistent with the trapping ofseveral H atoms at each acceptor site.J.T.Borenstein, J.W.Corbett, S.J.Pearton: Journal of Applied Physics, 1993, 73[6], 2751-4

[446-106/107-139]

Bulk Diffusion - Theoretical Analysis - Effect of ChargeThe total energy surfaces for a positively charged particle, diffusing in a perfect Sicrystal, were calculated. Negligible lattice distortion was assumed. Local densityfunctional theory and ab initio norm-conserving pseudopotentials were used within alinear response scheme. It was concluded that H+ diffusion involved many different sites;including the bond center. Various migration paths, with entirely different energybarriers, were possible.C.Pennetta: Solid State Communications, 1989, 69[3], 305-9

[446-64/65-178]

Bulk Diffusion - Theoretical Analysis - Effect of ChargeThe behavior of H in crystalline material was examined by using theoretical techniqueswhich were based upon the pseudopotential density functional method and a super-cellgeometry. Stable sites, migration paths, and barriers for different charge states wereexplored and were displayed as total-energy surfaces which provided an immediateinsight into these properties. The bond-center site was a global minimum for the neutraland positive charge states. In the negative charge state, the tetrahedral interstitial site waspreferred. The positive charge state was energetically favorable in p-type material andprovided a mechanism for the passivation of shallow acceptors. Electrons from the Hatoms annihilated the free holes, and the formation of H-acceptor pairs followedcompensation. Also considered were molecule formation and H-induced damage. Anumber of different mechanisms for defect formation were examined, and H-assistedvacancy formation was found to be an exothermic process.C.G.Van de Walle, P.J.H.Denteneer, Y.Bar-Yam, S.T.Pantelides: Physical Review B,1989, 39[15], 10791-808

[446-70/71-125]

375


Bulk Diffusion - Theoretical Analysis - Effect of ChargeA model was proposed for the behavior of H at moderate temperatures. It was assumedthat H had both a donor and an acceptor level in the band gap, and could thus exist in 3charge states: Ho, H+, H-. Its motion was slowed by the formation of H2 molecules as wellas by interactions with dopant atoms. Good simulations were obtained for both n-type andp-type material with various dopant levels. The diffusivity of neutral H was in agreementwith extrapolations of high-temperature data.D.Mathiot: Physical Review B, 1989, 40[8], 5867-70

[446-72/73-045]Bulk Diffusion - Theoretical Analysis - Effect of ChargeA first tight-binding molecular dynamics simulation was made of H diffusion in Si attemperatures of between 800 and 1800K. It was shown that the diffusivity deviated fromthe high-temperature Arrhenius plot at temperatures below 1200K, and the diffusioncoefficient was calculated in a region for which no experimental data were available. Thediffusion mechanism and path were observed during very long simulations. It wasdemonstrated that H diffused via jumps; avoiding low valence charge density regions.Observations were reported of jumps between non nearest-neighbor bond-center sites.G.Panzarini, L.Colombo: Physical Review Letters, 1994, 73[12], 1636-9

[446-115/116-152]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsThe diffusion mechanisms in amorphous material were investigated within the context offirst-principles results for Si-H interactions. Specific definitions were provided for thevarious quantities which entered into the description of H diffusion in amorphous Si, andvalues which were deduced from first-principles theory were linked to these quantities.The H chemical potential, µH, was associated with the formation energy of Si-H bondswith respect to bulk Si. The Si-H bond energy level, ESi-H, was located at a distance,below µH, which corresponded to the formation energy of a dangling bond. It was arguedthat H atoms could move readily between the µH and ESi-H levels in amorphous Si; thusmaintaining the equilibrium picture that underpinned the description of diffusion. Theresultant framework was consistent with experimental observations, and avoided the needfor more convoluted explanations: such as diffusion along interconnected void surfaces. Itwas concluded that, although explicit first-principles calculations of amorphous networkswere still required in order to obtain the specific energy distributions, the basiccorrelation between observed quantities and calculated values had now beenaccomplished.C.G.Van de Walle, R.A.Street: Physical Review B, 1995, 51[16], 10615-8

[446-123/124-184]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsThe jump-rate constants of atomic H and D in B-doped crystalline material werecalculated by using quantum transition-state theory which was based upon the path-integral centroid formalism. A break in the slope of the Arrhenius plot for the jump rate ofH was predicted to occur about 60K. This reflected a cross-over, from thermally activatedquasi-classical motion over a barrier, to thermally assisted quantum tunnelling; in goodagreement with experimental results. In the case of D, no deviation from Arrhenius

376


behavior was found at temperatures down to 30K. It was shown that the defect complexwhich underwent quantum tunnelling consisted of H, B and the nearest Si atoms.J.C.Noya, C.P.Herrero, R.Ramírez: Physical Review Letters, 1997, 79[1], 111-4

[446-152-0376]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsAn open-shell version of the modified intermediate neglected diffusional overlapmolecular orbital method was used to deduce the equilibrium positions of H, H+ and H2 ina 32-atom cyclic Si cluster. Two shells of Si neighbors around the defect were allowed torelax and to reconstruct, while the potential surface for H was mapped. The absoluteminimum, for both H and H+ was found at the bond-centered interstitial site. The H or H+

broke the lattice bond only if lattice relaxation was allowed. The energy of the H atoms,relative to the lattice, was 0.1eV lower at this point than its energy in a H2 moleculelocated at a tetrahedral interstitial site. An additional local minimum for atomic H wasfound at a distance of 0.042nm from the tetrahedral interstitial site, in the [111] directiontowards the next Si atom (the anti-bonding site). This site was 0.92eV higher in energythan the bond-centered interstitial site. Saddle-point calculations showed that theactivation energy for H diffusion was about 0.8eV from the bond-centered interstitial siteand was about 0.4eV for diffusion between anti-bonding sites. No gap levels arose withthe H atom on the bond-centered interstitial site. A donor level emerged above thevalence band edge when the H atom moved from one bond-centered interstitial site toanother. This could be ionized in p-type material. The ionized system exhibited the samediffusion path for H.P.Deak, L.C.Snyder, J.W.Corbett: Physical Review B, 1988, 37[12], 6887-92

[446-62/63-232]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA new metastable diatomic H complex was suggested for crystalline material. Some abinitio calculations showed that the complex was about 1.2eV more stable than were 2separated neutral H atoms in their most energetically favorable configurations. A newdiffusion mechanism was proposed which was based upon metastable and molecular H-complex formation and dissociation.K.J.Chang, D.J.Chadi: Physical Review Letters, 1989, 62[8], 937-40

[446-64/65-178]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsThe H effusion spectra of oxidized samples were investigated, and the characteristicactivation energies which were associated with the rupture of H bonds were estimated. Adislocation-enhanced H solubility was found for deformed crystals, and macroscopic(mm) diffusion depths could be achieved. Hydrogenation was carried out by exposingcrystals to H2 at 800C. As well as the desorption of H which was bound close to the Sisurface, rupture of the H-H bonds of molecules which were stored in deformed crystalswas observed. The storage of H2 molecules required the presence of dislocations ordeformation-induced point defects, and the H-H binding energy ranged from 2.7 to 3.8eV.It was suggested that this energy variation probably arose from the differing local strainsaround dislocations. Atomic H was found to dominate the effusion kinetics.

377


C.Kisielowski-Kemmerich, W.Beyer: Journal of Applied Physics, 1989, 66[2], 552-8[446-74-052]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsThe diffusion rate of the predominant paramagnetic defect in hydrogenated amorphousmaterial was estimated at temperatures of between 360 and 450C. The diffusion rate forthis defect was less than the rate for H. The high defect density of the material tended toretard H diffusion. The results did not support models which assumed a high defectmobility or the occurrence of H motion via paramagnetic defect interactions.W.Jackson, C.C.Tsai, R.Thompson: Journal of Non-Crystalline Solids, 1989, 114[2],396-8

[446-72/73-045]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsThe nature of H bonding and diffusion in crystalline material was investigated by using afirst-principles self-consistent pseudopotential method. The relative energies of interstitialatomic H, diatomic H complexes, and shallow dopant-H complexes were considered. Amechanism for H diffusion was presented which involved a new metastable diatomiccomplex with a much lower activation barrier for H diffusion than that for molecular H. Itwas suggested that metastable diatomic-complex formation was very likely to occur atlow temperatures and at high H concentrations; especially in n-type material. Diffusionvia an ionized H form was suggested to be more likely to occur in p-type material.K.J.Chang, D.J.Chadi: Physical Review B, 1989, 40[17], 11644-53

[446-74-052]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsThe H effusion spectra of oxidized samples were investigated, and the characteristicactivation energies which were associated with the rupture of H bonds were estimated. Adislocation-enhanced H solubility was found for deformed crystals, and macroscopic(mm) diffusion depths could be achieved. Hydrogenation was carried out by exposingcrystals to H2 at 800C. As well as the desorption of H which was bound close to the Sisurface, rupture of the H-H bonds of molecules which were stored in deformed crystalswas observed. The storage of H2 molecules required the presence of dislocations ordeformation-induced point defects, and the H-H binding energy ranged from 2.7 to 3.8eV.It was suggested that this energy variation probably arose from the differing local strainsaround dislocations. Atomic H was found to dominate the effusion kinetics.C.Kisielowski-Kemmerich, W.Beyer: Journal of Applied Physics, 1989, 66[2], 552-8

[446-74-052]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsIn order to understand the H diffusion mechanism in amorphous films which had beenprepared by chemical vapor deposition, changes in H depth profiles were studied as afunction of the plasma exposure time. It was found that the post-hydrogenation behaviorcould be explained by a model which involved the fast diffusion of atomic H through

378


weakly bound sites such as interstitials, capture by reactive sites such as weak Si-Si bondsand dangling bonds, and exchange between weakly bound and bonded H atoms.M.Nakamura, Y.Misawa: Journal of Applied Physics, 1990, 68[3], 1005-8

[446-86/87-054]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsExtensive molecular dynamics simulations of amorphous hydrogenated material werecarried out. The super-cells contained some 70 atoms, and just 1 defect, so as to minimizedefect-defect interactions. Simulations of super-cell samples, which originally contained 1bond-centered H atom in an otherwise defect-free volume, exhibited bond-centered tobond-centered diffusion; as in the case of crystalline material. A localized motion ofdefects and H atoms was also observed, and reflected experimental observations.P.A.Fedders: Journal of Non-Crystalline Solids, 1996, 198-200[1], 56-9

[446-138/139-105]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA model was developed, for H diffusion and passivation, which explained experimentalobservations of solar cell passivation, such as variations in the degree of passivation indiffering substrates, passivation due to annealing after Al alloying, and the effects ofplasma-enhanced chemical vapor deposition nitridation. Two important features of themodel were a new H diffusion mechanism that involved H-vacancy complex formation,and surface damage which produced a high H solubility at the surface and the dissociationof molecular H at low temperatures. The theoretical analysis was based upon staticpotential energy surfaces at the ab initio Hartree-Fock level.B.L.Sopori, X.Deng, J.P.Benner, A.Rohatgi, P.Sana, S.K.Estreicher, Y.K.Park,M.A.Roberson: Solar Energy Materials and Solar Cells, 1996, 41-42, 159-69

[446-138/139-105]

Bulk Diffusion - Theoretical Analysis - Effect of DynamicsA first-principles molecular dynamics simulation of high-temperature proton diffusion incrystalline material was performed. This was the first time that dynamic effects had beenexplicitly included in simulations of this material. It was found that the diffusionproceeded via a jump-like mechanism. Because of dynamic effects, the diffusion path wassubstantially different to that deduced from static total-energy calculations. Thecalculated diffusion coefficient, and its temperature dependence, were in good agreementwith available experimental data. It was suggested that scattering experiments coulddistinguish various diffusion paths.F.Buda, G.L.Chiarotti, R.Car, M.Parrinello: Physical Review Letters, 1989, 63[3], 294-7

[446-70/71-125]

Bulk Diffusion - Theoretical Analysis - Effect of TrappingA review was presented of recent measurements of trap-dominated H diffusion indisordered Si. The H transport could be described by a model with 3 levels: a transportlevel, shallow traps and deep traps. At low concentrations, diffusion was dominated bydeep traps which were well separated in energy from shallow traps. At high Hconcentrations, the trap density ranged from 8 x 1019 to 1021/cm3, and increased

379


approximately as the square root of the H concentration. Moreover, H diffusion dependedupon the carrier concentration and doping level as well as upon the H concentration. Itwas concluded that muon spectroscopy should be able to provide previously unavailableinformation concerning transport levels, shallow traps and the effects of carriers upon Hmotion in shallow traps.W.B.Jackson: Philosophical Transactions of the Royal Society A, 1995, 350[1693], 237-48

[446-119/120-224]

Bulk Diffusion - Theoretical Analysis - Effect of TrappingIn a medium which contains a high density of traps for the diffusant, the penetrationspeed is controlled by the time that is required to fill the traps. The predictions of thelimited flux model were verified for D transport in crystalline material that was highlydoped with B. The results also implied that the maximum amount of mobile D incrystalline material was less than 1016/cm3; the detection limit of secondary ion massspectroscopy.X.W.Wu, B.Y.Tong: Philosophical Magazine Letters, 1990, 61[3], 147-52

[446-76/77-036]

Bulk Diffusion - Theoretical Analysis - Effect of TrappingA model for trap-controlled diffusion was proposed in which diffusing atoms could becaptured either by trapping at an empty trap or by exchange with chemically identicaltrapped atoms of differing isotope number. The equations which described tracerdiffusion experiments were solved, and the predictions of the effects of dominantexchange or of dominant trapping were compared. A comparison with data on amorphoushydrogenated material showed that exchange was the predominant capture mechanism inD tracer diffusion measurements. This explained the relatively short distance (some20nm) which D travelled before trapping. It also reconciled the smoothness of the Dtracer diffusion profile, after long times, with the existence of deeply bound (by morethan 2.0eV) H in amorphous hydrogenated material.M.Kemp, H.M.Branz: Physical Review B, 1995, 52[19], 13946-54

[446-127/128-156]

Bulk Diffusion - Theoretical Analysis - Effect upon DefectsThe equilibration rate of frozen defects in P-doped hydrogenated amorphous material wasdescribed by the stretched-exponential time dependence, exp[-(t/K)B], where B wasbetween zero and unity. It had been proposed that the process which eliminated frozendefects was diffusion between interstitial sites; with a diffusivity that was dispersive intime due to the distribution of H trapping sites on weak Si-Si bonds. Here, this proposalwas criticised. The alternative suggestion was based upon the assumption that the relevantprocess was the jumping of H from the Si-H bond and into the interstitial site.Y.Osaka: Japanese Journal of Applied Physics, 1991, 30[11A], 2705-8

[446-84/85-069]

Bulk Diffusion - Theoretical Analysis - Effect upon DefectsThe behavior of H in crystalline material has been the object of intense experimentalinterest, but understanding has been limited by the absence of reliable theoretical

380


calculations. Advanced techniques were here used to investigate the properties of H withregard to stable configurations, migration paths, and cooperative interactions. Thecalculations were based upon local density functional theory, and involved ab initiopseudopotentials in a super-cell geometry. The results were used to supplement thecurrent understanding of the observed phenomena. A novel mechanism for H-induceddamage was proposed.C.G.Van de Walle, Y.Bar-Yam, S.T.Pantelides: Physical Review Letters, 1988, 60[26],2761-4

[446-62/63-232]

Bulk Diffusion - Theoretical Analysis - Enhanced DiffusionNuclear magnetic resonance 1H dipolar echo measurements were performed on B-doped,P-doped, compensated and intrinsic samples of amorphous hydrogenated material. It wasfound that, at room temperature, the dipolar spin lattice relaxation time was related to themacroscopic diffusion constants which were determined by means of secondary ion massspectroscopy measurements. The local jump frequencies which were deduced from thesevalues of the spin lattice relaxation time were orders of magnitude higher than thosevalues which were deduced from macroscopic diffusion results. Also, the associatedactivation energies were much lower (about 0.2eV). At 150 and 100K, the dipolar echo insamples which contained 0.001P exhibited both a narrow line and a broad line. The fullwidths at half-maximum of the narrow line and the broad line were the same as those ofthe narrow and broad components of the free induction decay, respectively. Because thenarrow line exhibited a spin lattice relaxation time with a slightly different temperaturedependence to that of the spin lattice relaxation time for the broad line, the local motionwas slightly different for H atoms in the clustered and dilute phases.P.Hari, P.C.Taylor, R.A.Street: Journal of Non-Crystalline Solids, 1993, 164-166[1], 313-6

[446-113/114-045]

Bulk Diffusion - Theoretical Analysis - Enhanced DiffusionA 2-channel model was proposed which accounted for the 2 activation energies whichhad been observed for H diffusion in crystalline material. A unified model for H andmuonium properties (static or thermally activated) was obtained in terms of a stable stateat the near bond center site of a metastable state at a tetrahedral site and in terms ofenergy barrier arguments. The results were obtained via a cluster approach to a Hartree-Fock analysis of the H potential energy surface.A.A.Bonapasta, A.Lapiccirella, N.Tomassini, M.Capizzi: Europhysics Letters, 1988, 7[2],145-9

[446-64/65-178]

Bulk Diffusion - Theoretical Analysis - Temperature DependencePrevious calculations for diffusion coefficients in solids had relied upon empiricalpotentials and/or dynamical simulations, both of which had considerable limitations. Apractical approach was presented here that was based upon rate theory and whichpermitted the calculation of temperature-dependent diffusion coefficients from static first-principles calculations. The results for H in Si were in excellent agreement with recent

381

D, H Grain Boundary/Surface Diffusion D, H

first-principles dynamical calculations for high temperatures, and with experimental data.They also clarified the nature of diffusion pathways and anharmonic effects.P.E.Blöchl, C.G.Van de Walle, S.T.Pantelides: Physical Review Letters, 1990, 64[12],1401-4

[446-76/77-038]

Grain Boundary Diffusion - Qualitative Observations - Effect of TrappingGrain boundaries in polycrystalline thin films of undoped material were shown to act asefficient H traps rather than as easy paths for diffusion. A comparison of H diffusion inpolycrystalline and monocrystalline material demonstrated that diffusion in the formerwas appreciably suppressed as compared to the latter. It was concluded that these resultshad important implications with regard to the hydrogenation of thin polycrystalline films.W.B.Jackson, N.M.Johnson, C.C.Tsai, I.W.Wu, A.Chiang, D.Smith: Applied PhysicsLetters, 1992, 61[14], 1670-2

[446-93/94-056]

Surface Diffusion - Quantitative DataThe migration of atomic H upon the (111) 7 x 7 surface was investigated by monitoringoptical second-harmonic diffraction from a sub-monolayer grating of adsorbed H. Withthe aid of sub-micron grating periods, it was possible to measure surface diffusivities ofless than 10-14cm2/s. A thermally activated diffusion process was identified which couldbe described by:

D(cm2/s) = 0.001 exp[-1.5(eV)/kT]G.A.Reider, U.Höfer, T.F.Heinz: Physical Review Letters, 1991, 66[15], 1994-7

[446-84/85-071]

Surface Diffusion - Quantitative DataAn analytical Si/H potential was developed and was fitted to the results of first-principleselectronic-structure calculations of H adatom adsorption and diffusion on embedded Siclusters. The latter were intended to model (100) 2 x 1. On the basis of this interactionpotential, the rate constants were calculated for a H adatom that hopped from one site toanother, parallel or perpendicular to the Si dimer rows. Monte Carlo simulations wereused to extract exact classical transition-state theory rate constants. The diffusionconstants for H adatoms which moved parallel to, or perpendicular to, the surface dimerrows were both found to exhibit an Arrhenius temperature dependence at temperaturesranging from 700 to 900K; with pre-exponential factors and activation energies of4.0cm

2/s and 38.1kcal/mol (parallel to dimer rows) or 0.48cm

2/s and 62.8kcal/mol

(perpendicular to dimer rows). The results confirmed a previous suggestion thatanisotropic diffusion of H adatoms on the (100) 2 x 1 surface would occur preferentiallyalong the edges of Si dimer rows. However, these predicted H adatom diffusion rateswere orders of magnitude higher (along the dimer rows) or lower (across the dimer rows)than the measured rates of H2 desorption from (100) 2 x 1-H. It was concluded that Hatoms might not be involved in the rate-limiting step for H desorption from Si(100).C.J.Wu, I.V.Ionova, E.A.Carter: Physical Review B, 1994, 49[19], 13488-500

[446-115/116-153]

382

D, H Surface Diffusion D, H

Surface Diffusion - Quantitative DataScanning tunnelling microscopy was used to image low coverages of H on (001) surfacesat temperatures of between 300 and 700K. It was found that individual H atoms becamemobile at about 570K. There was no apparent movement at temperatures below 500Kand, at temperatures above 640K, the H was moving as fast as (or faster) than thescanning tip. The observed hopping rate along the dimer rows was consistent with anactivation energy of 1.68eV; assuming an attempt frequency of 1013/s. Motion across thedimer rows was rarely observed, even at the higher temperatures. The diffusion barrier formotion along the dimer rows was estimated by using tight-binding and density-functionaltheory in the generalized gradient approximation. The calculated barrier was deduced tobe equal to 1.65eV on the basis of tight binding considerations, and equal to 1.51eVaccording to the generalized gradient approximation. The experimental and theoreticalresults were in good agreement, and showed that thermal diffusion could not operate attemperatures below 500K. It was therefor suggested that a mobile precursor mechanismmight be responsible for H pairing.J.H.G.Owen, D.R.Bowler, C.M.Goringe, K.Miki, G.A.D.Briggs: Physical Review B,1996, 54[19], 14153-7

[446-141/142-117]

Surface Diffusion - Theoretical Analysis - Effect of CoverageThe Si(100) surface is known to be reconstructed due to the formation of dimer rows.Recent temperature-programmed experiments on H desorption from this surface hadindicated that adsorbate-adsorbate lateral interactions in dimers were attractive and verystrong (-6kcal/mol). By using the lattice-gas model, a study was made here of the effect ofthis interaction upon the coverage dependence of the chemical diffusion coefficients forjumps of H atoms along, and perpendicular, to dimer rows. The adsorbate-adsorbatelateral interaction in the activated state was also taken into account. It was found that, forreasonable values of the parameters, both diffusion coefficients decreased with increasingcoverage. The anisotropy in the coverage dependence of the diffusion coefficients wasexpected to be rather weak.V.P.Zhdanov: Physical Review B, 1993, 48[19], 14325-30

[446-106/107-140]

Surface Diffusion - Theoretical Analysis - Effect of DefectsFirst-principles total-energy calculations were made of H adatom diffusion on a (100) 2 x1 surface. The transition states for H diffusion pathways were established by mapping thepotential energy for a H atom which was migrating between the dangling bonds of a (100)2 x 1 surface, The latter was modelled by using embedded finite Si clusters. It was foundthat the diffusion barriers were high (2 to 3eV) and wide (0.3 to 0.4nm); thus suggestingthat H diffusion on (100) occurred mainly via a classical hopping mechanism, rather thantunnelling. Moreover, the diffusion of H was predicted to be anisotropic, beingpreferentially directed parallel to Si-dimer rows; with an activation energy of 2.0eV.Higher activation energies (2.5 and 2.7eV) were predicted (for diffusion perpendicular todimer rows) for the cases of hopping from one dangling bond to a neighboring danglingbond which was on the same dimer or on an adjacent dimer, respectively. The mechanismfor H-atom diffusion along dimer rows was markedly different to that which had been

383

D, H Surface Diffusion D, H

proposed for Si adatom diffusion on (100). That is, H atoms were predicted to diffusealong the edges of the dimer rows, rather than down the middle.C.J.Wu, E.A.Carter: Physical Review B, 1992, 46[8], 4651-8

[446-93/94-056]

Surface Diffusion - Theoretical Analysis - Effect of DefectsDensity functional calculations were made of the potential energy surface for the bindingand diffusion of a H atom on a (111) adatom structure with 2 adatom and 1 rest-atomdangling bonds per unit cell. These were used to model the (111)-(7 x 7) surface. It wasfound that H binding was stronger at rest-atom, than at adatom, sites by about 0.2eV; ingood agreement with desorption data. This result, together with a detailed analysis of Hdiffusion paths and barriers, indicated that R → A → R jumps provided the mechanismfor H diffusion at low coverages. The calculated barrier for such jumps agreed well withexperimental data.A.Vittadini, A.Selloni: Physical Review Letters, 1995, 75[26], 4756-9

[446-127/128-158]

384

ErBulk Diffusion - Qualitative Observations - Concentration ProfilesPhotoluminescence at room temperature and at 77K was monitored in porous materialwhich had been doped with Er from a spun-on silica gel film. The incorporation of Er intoSiO2 at the surface of porous Si, and rapid thermal processing at temperatures above1223K, were found to be necessary for producing Er-related luminescence from porousSi. It was found that there was no Er diffusion, into monocrystalline Si, from spun-onfilms. The depth-dependent Er concentration in the bulk of porous Si was determined bymeans of secondary-neutral- and secondary-ion mass spectrometry depth profiling.Laterally resolved Er distributions in porous Si were deduced from energy-dispersive X-ray analyses.A.M.Dorofeev, N.V.Gaponenko, V.P.Bondarenko, E.E.Bachilo, N.M.Kazuchits,A.A.Leshok, G.N.Troyanova, N.N.Vorosov, V.E.Borisenko, H.Gnaser, W.Bock,P.Becker, H.Oechsner: Journal of Applied Physics, 1995, 77[6], 2679-83

[446-121/122-086]

Bulk Diffusion - Qualitative Observations - Effect of TrappingThe segregation and trapping of Er during the solid-phase crystallization of amorphous Si,on crystalline Si, was studied for Er concentrations ranging from 1016 to 5 x 1020/cm3.Amorphous surface layers were prepared on (100) Si by the implantation of 250keV Erion, followed by recrystallization at 600C. The samples were analyzed by using high-resolution Rutherford back-scattering spectrometry with 2MeV He+ or 100keV H+. It wasfound that the segregation coefficient depended strongly upon the Er concentration. At Erinterface areal densities of less than 6 x 1013/cm2, almost complete segregation to thesurface was observed; with a segregation coefficient of 0.01. At higher Er densities,segregation and trapping in the crystal were observed; with a segregation coefficient of0.20. The results were consistent with a model which assumed that defects in amorphousSi, near to the interface, acted as a trap for the Er.A.Polman, J.S.Custer, P.M.Zagwijn, A.M.Molenbroek, P.F.A.Alkemade: Journal ofApplied Physics, 1997, 81[1], 150-3

[446-141/142-115]

Bulk Diffusion - Qualitative Observations - Effect of TrappingIt was noted that the solid phase epitaxy of Er-implanted amorphous material resulted inthe segregation and trapping of Er, and led to the incorporation of up to 2 x 1020/cm3 into

385

Er Bulk Diffusion Er

monocrystalline material. Segregation occurred in spite of an extremely low Er diffusivity(less than 10-17cm2/s) in bulk amorphous material. A narrow segregation spike (with ameasured width of about 3nm) suggested that kinetic trapping was responsible for non-equilibrium concentrations of Er. The dependence of trapping upon temperature,concentration, and impurities instead indicated that thermodynamics controlled thesegregation. It was proposed that Er, by analogy with transition metals, diffusedinterstitially in amorphous Si, but was strongly bound at trapping centers. The bindingenthalpy of these trapping sites caused the amorphous phase to be energetically favorablefor Er so that, at low concentrations, the Er was almost completely segregated. However,when the concentration of Er in the segregation spike exceeded the amorphous trap centerconcentration, more Er was trapped in the crystal. A similar segregation and trappingbehavior was exhibited by Pr.J.S.Custer, A.Polman, H.M.Van Pinxteren: Journal of Applied Physics, 1994, 75[6],2809-17

[446-117/118-195]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsThe diffusion of Er at 1100 to 1250C led to the formation of shallow acceptors at Ev +0.045eV. The acceptor concentrations depended upon the heat-treatment conditions andupon the type and concentration of intrinsic point defects. The addition of vacanciesincreased the acceptor concentration and the presence of interstitials decreased theconcentration. These centers consisted of Er atoms in substitutional sites.O.V.Aleksandrov, V.V.Emtsev, D.S.Poloskin, N.A.Sobolev, E.I.Shek: Fizika i TekhnikaPoluprovodnikov, 1994, 28[11], 2045-8

[446-121/122-087]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsBy means of Hall-effect measurements, the characteristics of centers which were createdby Er diffusion and implantation were determined. It was shown that Er diffusion at 1100to 1250C led to the formation of shallow acceptor centers at about Ev + 0.045eV. Theconcentration of these acceptors was governed largely by the type and concentration ofthe intrinsic point defects which predominated during annealing at high temperatures. Thedata indicated that the acceptor centers could be substitutional Er atoms. An analysis ofelectrical data on float-zone material after Er implantation indicated the existence ofdonor states at Ec - 0.03eV, Ec - 0.05eV and Ec - 0.09eV. After implanting Er intoCzochralski material, and co-implanting Er and O into float-zone material, 3 types ofdonor state appeared at energies ranging from Ec - 0.03eV to Ec - 0.04eV, and at Ec -0.06eV and Ec - 0.09eV. Variations in the annealing temperature had a marked effectupon the concentrations of these donors. The donor states were thought to correspond toEr-related complexes whose structures were similar to those of O-related thermal donorsin Czochralski material.V.V.Emtsev, O.V.Alexandrov, D.S.Poloskin, E.I.Shek, N.A.Sobolev: Materials ScienceForum, 1995, 196-201, 615-20

[446-127/128-156]

386

FBulk Diffusion - Qualitative Observations - Effect of DefectsPlasma-etched samples exhibit deep in-diffusion. It was suggested here that a plasma-induced porosity was responsible for the diffusion of the etchant species in the substrate.It was found that in-diffusion occurred, regardless of the steric properties of the diffusingparticles and of the plasma-ion energies. The results for F-plasma interaction with Si wereconsidered at length.P.Brault: Journal of Physics - Condensed Matter, 1991, 3[36], 7073-8

[446-84/85-067]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe diffusion of ion-implanted F was studied by using secondary ion mass spectroscopicand thermal desorption spectroscopic techniques. Within the sub-amorphization doserange which was used, the F exhibited an anomalous out-diffusion behavior which wascharacterized by the depletion of F in Si substrates at temperatures above 550C. Therewas a complete suppression of diffusion deeper into the bulk. The F species whichmigrated to the surface reacted with the native oxide, and Si, to form the volatileoxyfluoride or fluoride, which then evaporated from the surface. There was clearevidence that the formation of the oxyfluoride was closely related to the thermallyactivated anomalous migration of F. Although the driving force for anomalous Fmigration was not identified, it appeared that the electric field was not a principal cause.S.P.Jeng, T.P.Ma, R.Canteri, M.Anderle, G.W.Rubloff: Applied Physics Letters, 1992,61[11], 1310-2

[446-93/94-051]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationIons of BF2

+ were implanted into (100) samples at room temperature, with an energy of40keV, through a 14nm-thick SiO2 layer. Profiling by secondary ion mass spectrometryindicated that subsequent annealing (650-850C, 0.5-4h) in a conventional furnace led to apronounced secondary peak in the B and F profiles; in addition to the near-surfaceprimary peak which was situated in the vicinity of the projected range of the implantedspecies. This effect was also observed in implanted samples which were rapidly thermallyannealed (900C, 15-60s). The depths of the secondary peaks in the F profilescorresponded to the depths of a damaged layer which was observed via cross-sectionaltransmission electron microscopy. Isochronal furnace annealing revealed that there was

387

F Bulk Diffusion/Surface Diffusion F

no chemical interaction between B and F atoms during annealing. This was alsosupported by the observation that F atoms did not affect the B segregation coefficientduring the oxidation of implanted samples. End-of-range extended dislocations appearedto be responsible for the gettering of atoms during annealing.Y.Kim, H.Z.Massoud, R.B.Fair: Applied Physics Letters, 1988, 53[22], 2197-9

[446-64/65-178]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsThe interfacial reactions of thin Co films with BF2

+-implanted (001) samples were studiedby using cross-sectional and plan-view transmission electron microscopy, and sheetresistance measurements. Implantation-amorphized samples were found to encourage theformation of CoSi2 at 400C; and its lateral growth at higher temperatures. The residualinterstitial defects in implanted samples were completely annihilated by CoSi2 formationat temperatures of between 800 and 900C. The elimination of all of the interstitial defectswas attributed to the injection of a high density of the vacancies which were producedduring silicide formation. Two discrete layers of F bubbles were observed in the silicideswhen they were annealed at temperatures of between 400 and 800C. When they wereannealed at temperatures of between 800 and 900C, large bubbles were often distributednear to CoSi2 grain boundaries. It was concluded that F atoms diffused rapidly attemperatures as low as 400C, and that appreciable amounts of F were present afterannealing at temperatures of between 400 and 900C.W.Lur, L.J.Chen: Journal of Applied Physics, 1988, 64[7], 3505-11

[446-72/73-045]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsThe presence of F in amorphous films was reported to have a beneficial effect upon theirelectronic properties. Consequently, the diffusion parameters of this material werestudied. A computer model was developed for amorphous Si, and molecular dynamicsmethods were used to simulate the process of F diffusion in amorphous material. Themolecular dynamics were generated from first principles by using Si-Si and Si-Fpotentials. A new simulation scheme was proposed that generated amorphous Si samplesdirectly from the crystalline phase. The model allowed for dangling bonds, vacancies andmicro-voids. Such models could be used to determine bond strength distributions anddiffusion mechanisms as functions of temperature and concentration.A.Silverman, J.Adler, R.Weil: Thin Solid Films, 1990, 193/194, 571-6

[446-78/79-053]

Surface Diffusion - Qualitative Observations - Effect of DefectsThe open-volume defects which were introduced into (100) crystals during F implantationwere investigated by means of variable-energy positron beam depth profiling. Thebehavior of the implantation-induced lattice defects during high-temperature annealing,and their role in the surface-oriented diffusion of F impurities, was considered. Thedefects became mobile, and recovered at temperatures below 550C. This was well beforethe onset of F diffusion, as detected by using secondary ion mass spectroscopic profiling.This suggested that, after irradiation and annealing, the F occupied substitutional sites towhich positrons were insensitive. The anomalous F diffusion which was seen in

388

F Bulk Diffusion/Surface Diffusion F

secondary ion mass spectrometry data was explained in terms of a 2-step diffusionmechanism in which the diffusion kinetics were governed by the dissociation ofsubstitutional F into an interstitial F and a vacancy; followed by rapid diffusion of theinterstitial F and of the vacancy through the crystal to the surface.C.Szeles, B.Nielsen, P.Asoka-Kumar, K.G.Lynn, M.Anderle, T.P.Ma, G.W.Rubloff:Journal of Applied Physics, 1994, 76[6], 3403-9

[446-117/118-195]

389

Fe

Figure 8: Diffusivity of Fe in Si. Omitting table 51, the data can be described overall bythe expression, D (cm2/s) = 0.0020 exp[-0.71(eV)/kT]

Bulk Diffusion - Quantitative DataThe behavior of type-3d transition metal impurities was investigated by using deep-leveltransient spectroscopic and Hall-effect measurements. Electrically active componentswere attributed to interstitial species. Pairs of the donors with B, acting as donors, werealso detected. The diffusivity of Fe, at temperatures ranging from 273 to 1343K, wasdescribed by:

D(cm2/s) = 0.0011 exp[-0.66(eV)/kT]

1.0E-16

1.0E-15

1.0E-14

1.0E-13

1.0E-12

1.0E-11

1.0E-10

1.0E-09

1.0E-08

1.0E-07

1.0E-06

1.0E-05

7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37

table 51table 52table 53table 54

104/T(K)

D (cm2/s)

390

Fe Bulk Diffusion Fe

H.Nakashima, T.Sadoh, H.Kitagawa, K.Hashimoto: Materials Science Forum, 1994, 143-147, 761-6

[446-113/114-044]

Table 51Diffusion of Fe in Amorphous Hydrogenated Si

T (C) D (cm2/s)350 10-16

400 10-15

450 10-14

Bulk Diffusion - Quantitative DataThe migration of Fe during low-temperature annealing was studied by usingphotocapacitance techniques. The effect of an electric field upon the underlying defectreactions was also investigated. The resultant depth profiles revealed the occurrence ofFei out-diffusion, but no precipitation in the bulk, at temperatures of up to 470K. Thekinetics in the presence of an electric field were explained in terms of carrier emission-limited Fe drift. The Fei diffusion data at zero field were described by:

D(cm2/s) = 0.01 exp[-0.84(eV)/kT]The data for a field of -10V were described by:

D(cm2/s) = 0.0014 exp[-0.69(eV)/kT]It was concluded that the out-diffusion and drift data reflected the operation of a charge-state dependent diffusion mechanism.T.Heiser, A.Mesli: Physical Review Letters, 1992, 68[7], 978-81

[446-88/89-053]

Bulk Diffusion - Quantitative DataMigration was studied by using deep-level transient spectroscopic techniques, or bydetermining the diffusion profiles of a deep level within depletion regions. The profilescould be accurately described by solutions of Fick's equation. Near to room temperature,the diffusivity of Fe was studied by monitoring pair reactions with substitutional B. It wasfound that the results, at temperatures ranging from 273 to 1343K, could be described by:

D(cm2/s) = 0.0011 exp[-0.66(eV)/kT]H.Nakashima, K.Hashimoto: Materials Science Forum, 1992, 83-87, 227-32

[446-93/94-052]

351 Bulk Diffusion - Quantitative DataRadioactive tracer techniques and X-ray photoelectron spectroscopy were used to studydiffusion in films of hydrogenated amorphous material at temperatures of between 200and 500C. The films were P-doped and had various defect structures. The migration rateand the diffusion coefficient were found to depend upon the defect structure. When Fediffusion was measured at temperatures of between 350 and 450C (table 51), thecoefficients decreased with increasing P content.

391


V.K.Kudoyarova, G.S.Kulikov, E.I.Terukov, K.K.Khodzaev: Journal of Non-CrystallineSolids, 1987, 90, 211-4

[446-55/56-038]

352 Bulk Diffusion - Quantitative DataThe pairing reaction of interstitial Fe and substitutional B atoms, in Fe-diffused B-dopedp-type material, was studied by using deep-level transient spectroscopy. Measurementswere made as a function of storage time at temperatures of 0, 27, 42, 57 and 72C (table52). The diffusivity of Fe at temperatures of between 0 and 72C was found to bedescribed by the expression:

D(cm2/s) = 0.33 exp[-0.81(eV)/kT]H.Nakashima, T.Isobe, Y.Yamamoto, K.Hashimoto: Japanese Journal of Applied Physics,1988, 27[8], 1542-3

[446-62/63-230]

Table 52Diffusion of Fe in Si

T (C) D (cm2/s)0 3.2 x 10-16

27 9.3 x 10-15

42 4.4 x 10-14

57 1.3 x 10-13

72 4.6 x 10-13

353 Bulk Diffusion - Quantitative DataProfiles in samples which had been diffused with Fe at temperatures of between 800 and1070C were determined by means of deep-level transient spectroscopy. It was found thatthe diffusivity of interstitial Fe within the above temperature range (table 53) wasdescribed by the expression:

D(cm2/s) = 0.00095 exp[-0.65(eV)/kT]T.Isobe, H.Nakashima, K.Hashimoto: Japanese Journal of Applied Physics, 1989, 28[7],1282-3

[446-70/71-124]

354 Bulk Diffusion - Quantitative DataThe effect of the charge state of Fe upon its diffusivity in p-type material was studied byusing capacitance-voltage and deep-level transient spectroscopic methods. It was foundthat the migration enthalpies of both positive and neutral Fe were equal to 0.92 and0.56eV, respectively (table 54). A disagreement between the predicted and observedresults was explained in terms of the diffusion path, elastic strain and Coulombicinteraction for various charge states.S.V.Koveshnikov, G.A.Rozgonyi: Applied Physics Letters, 1995, 66[7], 860-2

[446-123/124-183]

392


Bulk Diffusion - Quantitative DataIt was shown that Fe-Al and Fe-B pairs underwent excitation-induced dissociation underillumination at about 150K, where the thermally activated motion of Fe atoms was almostimpossible. The Fe-doping was achieved by annealing specimens, which had been pre-doped with Al or B, in Fe vapor at 1050C; followed by quenching. At first, Fe-acceptorpairs were generated by annealing the specimens at 80C. The illumination effect upon theconcentrations of the first- and second-nearest neighbor pairs of Fe-acceptors was thenstudied at various temperatures. Illumination with light at about 150K reduced theconcentration of the first-nearest neighbor pair while increasing that of the second-nearestneighbor pair. The activation energy for Fe atom motion from the first-nearest to thesecond-nearest neighbor of an Al atom was estimated to be equal to 0.11eV underillumination.S.Sakauchi, M.Suezawa, K.Sumino: Materials Science Forum, 1995, 196-201, 1345-50

[446-127/128-157]


T (C) D (cm2/s)800 8.0 x 10-7

900 1.7 x 10-6

1000 3.0 x 10-6

Bulk Diffusion - Quantitative DataThe precipitation of Feo in n-type material, and the generation of Fe-acceptor pairs(controlled by the diffusion of Fe+ in p-type material), were investigated by using electronspin resonance methods. Isochronal annealing data showed that the temperature rangewithin which Fe+ diffusion was active was lower than that for Feo. Annealing experimentswhich had been performed at various temperatures showed that the activation energies forthe diffusion of Feo and Fe+ were equal to 0.80 and 0.68eV, respectively.H.Takahashi, M.Suezawa, K.Sumino: Physical Review B, 1992, 46[3], 1882-5

[446-93/94-051]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe migration of interstitial Fe was observed at the atomic scale for the first time.Coulomb-excited 57Fe nuclei were implanted into high-purity n-type material, andMossbauer spectra were recorded at temperatures of between 300 and 850K. Diffusionalbroadening of one of the spectral components was attributed to interstitial Fe. The isomershift of the interstitial Fe was determined, and the proposal that a single mechanismgoverned Fe diffusion at temperatures of between 300 and 1500K was confirmed.P.Schwalbach, S.Laubach, M.Hartick, E.Kankeleit, B.Keck, M.Menningen, R.Sielemann:Physical Review Letters, 1990, 64[11], 1274-7

[446-74-051]

393


Bulk Diffusion - Qualitative Observations - Effect of DefectsThe recombination-enhanced migration of interstitial Fe in the vicinity of substitutional Bwas investigated by using space charge techniques, combined with minority carrierinjection at temperatures below 200K. The temperature dependence of the creation ratesof metastable pairs revealed a small negative activation energy. This implied that therecombination-enhanced defect reaction for each defect was athermal and that therecombination energy was large relative to the thermal energy which was required inorder to surmount the reaction barrier. Four electron traps and a hole trap were observedas structurally metastable Fei-Bs pairs after injection. The creation and annihilation ofthese pairs by injection was analyzed in terms of the theory of recombination-enhanceddefect reactions. The migration rates for traps, E2, E5 and H2, were much lower thanthose for E3, E4, and H2*. The attribution of the E2, E5 and H2 traps to Td sites couldexplain the migration behavior quite well; in addition to the thermal and electricalproperties. The traps, E3, E4, and H2, seemed to be attributable to unstable sites such ashexagonal sites, or to lattice strain-related sites in the vicinity of Bs.H.Nakashima, T.Sadoh, T.Tsurushima: Materials Science Forum, 1995, 196-201, 1351-6

[446-127/128-157]


Species T (C) D (cm2/s)Feo 130 4.08 x 10-11

Feo 122 3.16 x 10-11

Feo 108 2.02 x 10-11

Feo 100 1.07 x 10-11

Feo 90 8.28 x 10-12

Fe+ 130 3.16 x 10-12

Fe+ 122 2.30 x 10-12

Fe+ 108 1.0 x 10-12

Fe+ 100 4.9 x 10-13

Fe+ 90 2.78 x 10-13

Bulk Diffusion - Qualitative Observations - Effect of DefectsA study was made of recombination-enhanced Fe atom jumps between the first- andsecond-nearest neighbor sites of Fe-Al and Fe-B pairs. Specimens were first annealed at80C in order to generate Fe acceptor pairs after Fe doping. The concentrations of the first-and second-nearest neighbor Fe-acceptor pairs were determined by performing electronspin resonance measurements after annealing at about 150K under optical excitation. Itwas found that the concentration of the first-nearest neighbor pairs decreased, while thatof the second-nearest neighbor pairs increased, during annealing. The activation energyfor the above changes was equal to about 0.11eV in the case of Fe-Al pairs. This wasmuch smaller than that (0.8eV) for thermal annealing alone. In the case of Fe-B pairs, the

394


electron spin resonance signals from the second-nearest-neighbor pairs could be detecteddue to annealing at about 160K under optical excitation.S.Sakauchi, M.Suezawa, K.Sumino, H.Nakashima: Journal of Applied Physics, 1996,80[11], 6198-203

[446-141/142-115]

Bulk Diffusion - Qualitative Observations - Effect of DefectsInteractions between group-VI elements (S, Se) and fast-diffusing Fe impurities werestudied. Infra-red absorption, electron spin resonance, and neutron activation studiesshowed that the group-VI elements effectively interacted with each other or with the fast-diffusing impurities. It was found that, for each pairing of a group-VI element with a fast-diffusing impurity, there was a certain annealing temperature at which they interactedmost efficiently. A definite correlation was established between this annealingtemperature and the thermodynamic Gibbs free energy (at 298K) of the correspondingcompounds. This correlation suggested that the interaction process mainly involved theformation of electrically neutral chemically bonded complexes by substitutional group-VIelement centers and interstitial fast-diffusant centers.M.K.Bakhadirkhanov, S.I.Askarov, N.Norkulov: Physica Status Solidi A, 1994, 142[2],339-46

[446-117/118-194]

Bulk Diffusion - Qualitative Observations - Effect of Electric FieldThe irreversible loss of interstitial Fe during the low-temperature annealing of n-typematerial was investigated by using capacitance techniques and strong electric fields. Byanalyzing isothermal annealing kinetics and Fei depth profiles, it was shown that thebehavior of Fei was governed mainly by simple diffusion processes, in which the drivingforce was either the concentration gradient near to the sample surface, or was the electricfield which was present in the depletion region.T.Heiser, A.Mesli, N.Amroun: Materials Science Forum, 1992, 83-87, 173-8

[446-93/94-052]

Bulk Diffusion - Qualitative Observations - Effect of DopingAn investigation of the diffusion and solubility of Fe in hydrogenated amorphous filmswas carried out at temperatures ranging from 290 to 525C. This made it possible todetermine suitable conditions for diffusion doping. Such diffusion was used to preparehydrogenated amorphous films which were doped with this metal, and a study was madeof their dark conductivity and photoconductivity. It was noted that diffusion doping withmetals gave rise to activation energies, for electrical conduction, which were as high as1.4eV. That is, the activation energy exceeded 50% of the band gap; atypical behavior foran intrinsic material. Cyclic heating of the doped films, at temperatures ranging from 20to 200C, resulted in a gradual recovery of the electrical properties which they had had inthe original state. This recovery resembled the precipitation of a supersaturated solidsolution in crystals.

395

Fe Bulk Diffusion Ga

M.S.Ablova, G.S.Kulikov, S.K.Persheev, K.K.Khodzhaev: Fizika i TekhnikaPoluprovodnikov, 1990, 24[11], 1943-7 (Soviet Physics - Semiconductors, 1990, 24[11],1208-11)

[446-81/82-043]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsThe behavior of (100) surface defects which were produced by Fe contamination werestudied by using transmission electron microscopy. After annealing (1150C, 1h) and heat-treatment (850C, 2h), Fe-containing precipitates were observed close to the interface withSiO2 which formed during annealing. The results demonstrated that Fe atoms diffusedinto the Si substrate during annealing at 1150C, and precipitated at Si/SiO2 interfaces. TheFe atoms which were left on the surface formed inclusions in the surface SiO2 layer.During additional thermal oxidation at 1000C, oxidation-induced stacking faults formed.Unlike those in Cu- or Ni-contaminated material, the latter were not decorated.S.Sadamitsu, A.Sasaki, M.Hourai, S.Sumita, N.Fujino: Japanese Journal of AppliedPhysics, 1991, 30[8], 1591-6

[446-84/85-067]

GaBulk Diffusion - Qualitative Observations - Concentration ProfilesDelta-doped layers were produced by depositing 0.39 of a monolayer of Ga onto the(001) surface. The dopant atoms were then buried by means of solid-phase epitaxy. It wasfound, using Rutherford back-scattering spectrometry, that some 30% of the Ga atomswere located at substitutional sites; in a peak that was less than 1nm wide. The otheratoms were at the surface. The surface atoms desorbed upon annealing at 985K, while theburied atoms were only slightly redistributed. The profile completely degraded attemperatures which were slightly higher than 1100K.P.M.Zagwijn, Y.N.Erokhin, W.F.J.Slijkerman, J.F.Van der Veen, G.F.A.Van der Walle,D.J.Gravesteijn, A.A.Van Gorkum: Applied Physics Letters, 1991, 59[12], 1461-3

[446-84/85-068]

Bulk Diffusion - Qualitative Observations - Effect of Surface LayerThe electrical activity of implanted Ga atoms was investigated during annealing, with orwithout a SiO2/SixNy/SiO2 capping film. A low electrical activity of the Ga atoms, in theabsence of a capping film, was attributed to Ga out-diffusion. In order to prevent suchout-diffusion, it was necessary to put a capping film on the Si surface during annealing.By using a capping film, an electrical activity of 75% could be obtained after annealing

396

Ga Bulk Diffusion Ga

(1250C, 16h). It was suggested that Ga atoms which out-diffused from the Si substrate,and into the capping film, acted as a diffusion source, which then re-diffused Ga into thesubstrate.M.Watanabe, O.Ishiwata, M.Nagano, H.Kirihata: Journal of the Electrochemical Society,1992, 139[6], 1748-51

[446-93/94-052]

Bulk Diffusion - Qualitative Observations - Transient EffectsThe redistribution of Ga in amorphous material by means of medium-energy ionscattering was studied at temperatures ranging from 560 to 830K. During the first 10s ofannealing, the diffusivity exhibited a transient behavior that was attributed to a change inthe relaxation state of the amorphous matrix. At temperatures of between 560 and 830K,the diffusivity during relaxation was enhanced by 7 to 2 orders of magnitude; ascompared with the value for bulk amorphous material.P.M.Zagwijn, W.J.Huisman, A.Polman, E.Vlieg, A.H.Reader, D.J.Gravesteijn: Journal ofApplied Physics, 1994, 76[10], 5719-23

[446-117/118-195]

Bulk Diffusion - Theoretical Analysis - Effect of DopantA macroscopic model for dopant diffusion, involving dopant-dopant interactions, wasextended so as to include 3-dimensional effects. The model was applied to dopantdiffusion in thin crystalline films under conditions where a second dopant was present inhigh concentrations and was distributed uniformly throughout the film. It was found thatthis resulted in an unexpected thickness dependence of the behavior in very thin films.S.Aronowitz: Journal of Applied Physics, 1991, 70[11], 6815-20

[446-91/92-029]

Bulk Diffusion - Theoretical Analysis - Effect of DopantIt was found that there was a striking agreement between theory and experiment withregard to the behavior of p-type dopants. Group-III species exhibited retarded diffusionwhen they were interstitial, and exhibited enhanced diffusion when they weresubstitutional. The theoretical and experimental results indicated that the presence oflarge quantities of Ge, when completely integrated into Si, altered the molecular orbitalstructure of the lattice to such a degree that large long-range effects were exerted upon p-type dopant species. It was concluded that interactions between p-type dopants, andinteractions with Ge, could be used to define and alter diffusion patterns.S.Aronowitz, C.Hart, S.Myers, P.Hale: Journal of the Electrochemical Society, 1991,138[6], 1802-6

[446-84/85-063]

397

Ge

Figure 9: Surface Diffusivity of Ge on (111) Si (see table 55)

Bulk Diffusion - Quantitative DataThe diffusion of Ge impurity in B-doped (about 1016 or 1018/cm3) material wasinvestigated at temperatures of between 1000 and 1200C by using radiotracer andsectioning techniques. At low doping levels, it was found that:

D(cm2/s) = 1.38 x 105 exp[-5.39(eV)/kT]The doping effect led to an enhancement of Ge diffusion. The results were analyzed byassuming that the mechanism involved singly charged and neutral vacancies. Below1050C, diffusion coefficients which were higher than expected were measured in bothslightly and heavily doped material.

1.0E-11

1.0E-10

1.0E-09

1.0E-08

9 10 11

9.099.269.529.751010.2110.42

104/T(K)

D (cm2/s)

398

Ge Bulk Diffusion Ge

A.L.Bouchetout, N.Tabet, C.Monty: Materials Science Forum, 1986, 10-12, 127-32[446-49-028]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe mechanism which led to a concentration dependence of the effective distributioncoefficient of Ge was considered. It was shown that the dependence of the diffusioncoefficient, of impurity atoms in a melt, upon their concentration could lead to aconcentration dependence of the effective distribution coefficient.N.I.Gorbacheva: Izvestiya Akademii Nauk SSSR - Neorganicheskie Materialy, 1991,27[10], 2036-40 (Inorganic Materials, 1991, 27[10], 1728-32)

[446-93/94-053]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe migration of impurity atoms in solid-phase epitaxially grown layers was studied bymeans of back-scattering spectrometry. The samples were amorphized by Ge implantationinto [100]- or [111]-oriented samples. The amorphized regions were then recrystallizedby furnace annealing. Asymmetrically enhanced diffusion was observed in the case of[111]-type samples; for both As and Ge implantation. A striking result was that, in thecase of [111]-type samples, there was extensive smearing of the profiles while, in the caseof [100]-type samples, this was hardly detectable. It was recalled that it was well knownthat the re-growth features of [100]-type and [111]-type amorphized Si were different. Inessential agreement with previous observations, it was found here (using cross-sectionalelectron microscopy) that the re-grown layer on [111]-type material was characterized byextensive multiple twinning. A 2-step annealing treatment usually led to a superiorcrystalline perfection, as confirmed by the cross-sectional transmission electronmicroscopy of samples which had first been annealed at 650C, and then at 1050C. Thediffusion of Ge was then also reduced. Overall, the extent of diffusion was related todefects.R.Turan, T.G.Finstad: Philosophical Magazine A, 1991, 63[3], 519-25

[446-78/79-048]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe first experimental identification of the diffusion mechanism of Ge was presented. Byusing thermal nitridation reactions to create an excess of either self-interstitials orvacancies, it was established that (under equilibrium conditions at 1050C) Ge diffusionoccurred via both substitutional-interstitial interchange and vacancy mechanisms; withcomparable contributions arising from each of them. If previous conjectures, that Gediffusion in Si was similar to Si self-diffusion, were correct the present findings supportedthe idea that Si self-diffusion took place via both interstitial and vacancy mechanisms.P.Fahey, S.S.Iyer, G.J.Scilla: Applied Physics Letters, 1989, 54[9], 843-5

[446-70/71-124]

Bulk Diffusion - Qualitative Observations - Effect of DopantStudies were made of diffusion in material with high donor concentrations that wereproduced by P doping. It was found that, for donor concentrations which were belowabout 2 x 1020/cm3, the diffusivity depended linearly upon the dopant concentration.However, at higher dopant concentrations, the diffusivity increased markedly with

399

Ge Bulk Diffusion/Surface Diffusion Ge

increasing donor concentration. This behavior was successfully modelled in terms of thevacancy-percolation model, and it was concluded that collective phenomena played asignificant role at high donor concentrations.A.N.Larsen, K.K.Larsen, P.E.Andersen, B.G.Svensson: Journal of Applied Physics, 1993,73[2], 691-8

[446-106/107-132]

Bulk Diffusion - Qualitative Observations - Effect of DopantAnomalous diffusion effects were observed when Ge was diffused with P, B, or As ininert or oxidizing ambients. The Ge, which was an unchanged column-IV dopant with acovalent radius that was close to that of Si, appeared to alter the point defect populationin the Si lattice. During oxidation, the usual oxidation-enhanced diffusion was notobserved when Ge was present in the lattice. The anomalous tail diffusion of highconcentrations of P was also reduced when Ge was present.J.R.Pfiester, P.B.Griffin: Applied Physics Letters, 1988, 52[6], 471-3

[446-60-014]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe diffusion of Ge, after the solid-phase epitaxial growth of material which had beenamorphized by Ge2+ implantation, was measured using back-scattering spectrometry.Asymmetrically enhanced diffusion was observed in [111]-type oriented samples whichwere annealed at 1050C. A high concentration of twins, which was detected using cross-sectional transmission electron microscopy, appeared to be related to the enhancement.There was a large difference in Ge diffusivity and defect structure between [100]-typeand [111]-type oriented samples.R.Turan, B.Hugsted, O.M.Lonsjo, T.G.Finstad: Journal of Applied Physics, 1989, 66[3],1155-8

[446-74-052]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesA theoretical study was made of a self-limitation of the surface segregation of Ge inepitaxial Si overlayers. This was attributed to the surface bond geometry on (100)Siduring molecular beam epitaxy. It was found that Ge surface segregation was markedlylimiting when the Ge concentration exceeded 0.01 of a monolayer. As a result of this self-limitation, the segregation profiles of Ge in Si overlayers decayed exponentially in thegrowth direction. There was a kink in the profile at about 3 x 1020/cm3; in good agreementwith experimental observations. The kinetic barrier to Ge surface segregation wasestimated to be 1.63eV.S.Fukatsu, K.Fujita, H.Yaguchi, Y.Shiraki, R.Ito: Applied Physics Letters, 1991, 59[17],2103-5

[446-84/85-068]

Surface Diffusion - Theoretical Analysis - AnisotropyThe adsorption and diffusion of Ge adatoms upon fully-relaxed (100) 2 x 1 surfaces wasstudied by using molecular dynamics methods, Tersoff's potential for the description ofGe-Si interactions, simple transition-state theory, and lattice-gas simulators. It was found

400

Ge Surface Diffusion Ge

that there were 6 local minima for adsorption at the surface, and the associated activationenergies were determined. The resultant diffusivity could be described by:

D(cm2/s) = 4.3 x 10-4 exp[-0.73(eV)/kT]It was also found that the adatom diffusion was anisotropic in nature, and that thedirection of easy diffusion was perpendicular to the dimers of the original surface; that is,parallel to the dimer rows. A comparison with Si adatom diffusion showed that Geadatom diffusion was less anisotropic, and that Ge adatoms diffused between 2 and 3times more slowly than did Si adatoms (on a given surface). The diffusion coefficients forGe and Si adatom migration perpendicular to the dimer rows were found to be describedby:

Ge: D(cm2/s) = 2.8 x 10-3 exp[-1.17(eV)/kT]Si: D(cm2/s) = 4.8 x 10-3 exp[-1.20(eV)/kT]

D.Srivastava, B.J.Garrison: Physical Review B, 1992, 46[3], 1472-9[446-93/94-053]

Table 55Diffusivity of Ge on the (111) Surface of Si

T (C) D (cm2/s)827 1.9 x 10-9

807 9.9 x 10-10

777 6.8 x 10-10

753 3.5 x 10-10

727 1.6 x 10-10

707 5.8 x 10-11

687 4.6 x 10-11

355 Surface Diffusion - Quantitative DataThe diffusion of Ge on (111)Si at high temperatures was studied experimentally by meansof second-harmonic microscopy, and was computed using molecular dynamicssimulations and a Stillinger-Weber potential. The experimental results (table 55) could bedescribed by:

D (cm2/s) = 600 exp[-2.48(eV)/kT]The simulations yielded essentially the same values; thus confirming the validity of theStillinger-Weber potential for diffusional studies. A previously developed semi-empiricalcorrelation also gave similar results. The simulations also furnished estimates of thecorresponding parameters for intrinsic diffusion, and for the enthalpy and entropy of Geadatom-vacancy pair formation on Si. The simulations also provided evidence of minorcontributions, to intrinsic diffusion, which arose from atom exchange, as well as theeffect of complex high-temperature islanding phenomena at ps time-scales.C.E.Allen, R.Ditchfield, E.G.Seebauer: Physical Review B, 1997, 55[19], 13304-13

[446-152-0400]

401

Ge Surface Diffusion Ge

Surface Diffusion - Quantitative DataIt was pointed out that quantitative surface diffusion coefficients for clustering systemswere of great importance for thin film growth applications. Here, the first determinationsof the activation energy for surface diffusion on Stranski-Krastanov layers in atechnologically important hetero-system were reported. The findings were based uponpreviously published independent measurements of the activation energy for clustergrowth and for cluster formation from a free adatom concentration. For a Stranski-Krastanov layer of Ge on Si(100), an equivalent coverage of 3.08 monolayers wasdeduced, together with an activation energy (for surface diffusion) of 0.84eV.M.Zinke-Allmang, S.Stoyanov: Japanese Journal of Applied Physics, 1990, 29[10],L1884-7

[446-78/79-053]

Surface Diffusion - Qualitative Observations - Effect of AnisotropyThe growth of Ge on the (001) surfaces of Si was studied by means of scanningtunnelling microscopy. By monitoring the spatial distribution of two-dimensional Geislands after sub-molecular deposition, anisotropies were found in the surface diffusivityand lateral sticking coefficient of Ge adatoms at various types of step. It was found thatthe fast direction for Ge surface diffusion lay along the substrate dimer rows. The 2 typesof monatomic step had very different lateral sticking coefficients for Ge adatoms. Ametastable so-called hut cluster phase was found to provide a kinetic pathway for thetransition between the 2-dimensional Ge layers and so-called macroscopic 3-dimensionalclusters. The hut clusters had (105)-type facet planes, and nucleated preferentially at[100]-type steps. The nucleation barrier for the huts was lower than that for themacroscopic clusters, and therefore provided an easier path, for accommodating Geadatoms from the gas phase, than did the direct formation of macroscopic clusters.Y.W.Mo, M.G.Lagally: Journal of Crystal Growth, 1991, 111[1-4], 876-81

[446-88/89-054]

Surface Diffusion - Qualitative Observations - Effect of AnisotropyScanning tunnelling microscopy was used to study the anisotropy of surface diffusion onthe (001) plane. This was done by analyzing the denuded zones around steps in the spatialdistribution of 2-dimensional islands which formed after sub-monolayer deposition. Itwas found that diffusion was some 1000 times faster along the dimer row direction thanperpendicular to it. It was also found that SB steps were good sinks for Ge adatoms,whereas SA steps were not. The SB steps appeared to be symmetrical sinks for adatomscoming from either the down terraces or the up terraces.Y.W.Mo, M.G.Lagally: Surface Science, 1991, 248[3], 313-20

[446-84/85-068]

Surface Diffusion - Qualitative Observations - Effect of Electron IrradiationThe surface of a Si substrate was partially covered with an over-layer of Ge. The surfaceof the Ge sheet was then irradiated with 7MeV electrons at 40C in a vacuum to a fluenceof 1018/cm2. The surface diffusion of Ge atoms on the Si substrate was confirmed by usingatomic force microscopy and secondary-ion mass spectrometry.T.Wada, H.Fujimoto, H.Masuda: Materials Science Forum, 1995, 196-201, 1625-30

[446-127/128-157]

402

Ge Surface Diffusion/Interdiffusion Ge

Surface Diffusion - Theoretical Analysis - Effect of DefectsThe binding sites for the adsorption of a single Ge atom on the (100) Si surface wereidentified by using first-principles total-energy calculations. It was found that thetheoretical diffusion barriers were in excellent agreement with experimental data. Byusing a large super-cell, a controversy concerning the binding geometry and migrationpath of the adatom was resolved, and its effect upon the buckling of Si dimers could beinvestigated. It was shown that the adatom caused a buckling defect which was frequentlydetected by using scanning tunnelling microscopy, and that the study of a single adatomwas therefore experimentally possible.V.Milman, D.E.Jesson, S.J.Pennycook, M.C.Payne, M.H.Lee, I.Stich: Physical ReviewB, 1994, 50[4], 2663-6

[446-115/116-152]

Table 56Diffusion in Si/Ge Films

T (K) Si/Ge D (cm2/s)622 50/30 6.00 x 10-22

672 60/20 1.32 x 10-20

682 60/20 4.60 x 10-20

690 60/20 4.62 x 10-20

708 60/20 9.87 x 10-20

719 60/20 2.50 x 10-19

Melt Diffusion - Quantitative DataPulsed-laser melting experiments were performed on GexSi1-x alloys, where x was lessthan 0.10, which involved re-growth velocities that ranged from 0.25 to 3.9m/s. Ananalysis of post-solidification Ge concentration profiles, together with time-resolved melt-depth measurements, permitted the determination of the liquid-phase diffusivity of Ge inSi and of the dependence of the Ge partition coefficient upon interface velocity. Adiffusivity of 0.00025cm2/s was deduced. The partition coefficient versus interfacevelocity data were analyzed by using various models: including the dilute and non-dilutecontinuous growth models. Extrapolation to zero velocity, by using these models,furnished an equilibrium partition coefficient of about 0.45 and a diffusive speed of2.5m/s.D.P.Brunco, M.O.Thompson, D.E.Hoglund, M.J.Aziz, H.J.Gossmann: Journal of AppliedPhysics, 1995, 78[3], 1575-85

[446-123/124-183]

356 Interdiffusion - Quantitative DataMulti-layered amorphous Si/amorphous Ge films with a periodicity of 8 to 10nm wereobtained by using ultra-high vacuum evaporation techniques. The interdiffusioncoefficient (table 56) for the system was determined by measuring the intensity, of the

403

He Bulk Diffusion Hg

neutron (000) forward scattering satellites arising from the modulation, as a function ofannealing temperature and time. The temperature dependence of the interdiffusioncoefficient at temperatures of between 620 and 710K could be described by theexpression:

D(cm2/s) = 0.00634 exp[-2.35(eV)/kT]C.Janot, A.Bruson, G.Marchal: Journal de Physique, 1986, 47[10], 1751-6

[446-48-067]

HeBulk Diffusion - Qualitative Observations and DislocationsThe conditions which were necessary for driving the reconstruction of a dislocation coreby means of He atoms diffusing along a dislocation pipe were examined. Using a simplemodel, it was shown that, unlike the case of Ge (in which such reconstruction occurred),the chances of driving such a reconstruction in Si by He were rather small. Attention hadto be focussed on H, which was known to diffuse effectively into deformed Si.B.Pohoryles: Philosophical Magazine Letters, 1988, 58[1], 1-5

[446-61-078]

HgBulk Diffusion - Quantitative DataWafers were amorphized by implanting 2 x 1015 or 3 x 1017cm2 of Ar+ ions, and were thenimplanted with 400keV Hg+ ions to a dose of 4 x 1015/cm2. The diffusion and evaporationof the implanted Hg, and the recrystallization of the Hg-implanted amorphous material,were studied at temperatures ranging from 700 to 1000C by using MeV He ionRutherford back-scattering/channelling techniques. It was found that, at 700C, thermaldiffusion predominated and there was no loss of implanted Hg in amorphous materialwhich had been implanted to 2 x 1015/cm2. The diffusion coefficient was deduced to be7.8 x 10-15cm2/s. At temperatures above 800C, evaporation predominated. At 1000C, theimplanted Hg disappeared. Recrystallization was also observed. In the case of sampleswhich had been implanted with 3 x 1017 Ar+/cm2, when annealed at 700C, 87% of theimplanted Hg was lost and, at 800C, the Hg entirely disappeared. The remaining Arsegregated towards the surface and the amorphous/crystalline interfaces.K.M.Wang, S.Ma, B.R.Shi, H.Y.Zhai, X.D.Liu, J.T.Liu, X.J.Liu: Solid StateCommunications, 1995, 93[2], 155-8

[446-121/122-087]

404

InBulk Diffusion - Quantitative DataLayers (2nm) of In were incorporated into amorphous material by using a laser quenchingtechnique. The samples were then analyzed by using Rutherford back-scatteringspectrometry and channelling methods. It was found that the diffusivity of was equal to4.8 x 10-17cm2/s at 220C, and to 8.9 x 10-17cm2/s at 300C. Impurity diffusion in theamorphous material was found to be rapid at high concentrations, and very slow at lowconcentrations.M.A.Harith, S.U.Campisano, J.M.Poate: Semiconductor Science and Technology, 1988,3[9], 829-31

[446-72/73-046]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesA modified neutron activation technique was described which was able to measure theprofiles of isotopic impurities with a short half-life. A deep acceptor impurity profile wasobtained by applying the method to Si diffused with In. It was found that concentrationsas low as 1013/cm3 could be measured with an accuracy of ±10%.R.M.Huang, R.S.Huang: Journal of the Electrochemical Society, 1986, 133[12], 2605-8

[446-51/52-134]

Bulk Diffusion - Theoretical Analysis - Effect of ChargeExperiments show that ion pairing has a marked effect upon the diffusion of oppositelycharged impurities. An analysis of literature data was used to deduce the ion pairingcoefficients for n-type impurities with B and In. It was suggested that the paired ionsoccupied adjacent substitutional sites; with a small perturbation in the Coulomb bindingwhich arose from elastic effects.N.E.B.Cowern. Applied Physics Letters, 1989, 54[8], 703-5

[446-64/65-176]

Bulk Diffusion - Qualitative Observations - Effect of OxidationSystematic diffusion experiments were performed in dry oxidizing ambients, attemperatures ranging from 800 to 1050C, by using wafers which were implanted with In.Secondary-ion mass spectrometry was used to analyze the dopant distributions before andafter heat treatment. The oxidation-enhanced diffusion parameter and the segregationcoefficient at the Si/SiO2 interface (ratio of the In concentration in the Si to that in the

405

In Surface Diffusion In

oxide) were estimated as a function of temperature by using secondary ion massspectrometry depth profiles. It was observed that the segregation coefficient of In at theSi/SiO2 interface was much less than unity (as in the case of B). However, unlike the caseof B, the segregation coefficient of In at the Si/SiO2 interface decreased with increasingtemperature. The validity of the deduced parameters was verified by comparing simulatedand measured secondary ion mass spectrometry profiles for an In-implanted metal-oxidesemiconductor field-effect-transistor.I.C.Kizilyalli, T.L.Rich, F.A.Stevie, C.S.Rafferty: Journal of Applied Physics, 1996,80[9], 4944-7

[446-141/142-117]

Surface Diffusion - Quantitative DataThe perturbed ?-? angular correlation spectroscopic method was used to study thediffusion and desorption of isolated 111In atoms which were adsorbed on the 7 x 7 (111)surface. Two different adsorption sites were found. The activation energy for migration tonormal adsorption sites was deduced to be 0.72eV. The relative populations of the 2adsorption sites changed at about 500K; suggesting an activation energy of 1.61eV for theprocess. The binding energy of the In atoms was estimated to be 1.93eV. The desorptionbehavior depended strongly upon the In concentration in the low-coverage regime.G.Krausch, T.Detzel, R.Fink, B.Luckscheiter, R.Platzer, U.Wöhrmann, G.Schatz:Physical Review Letters, 1992, 68[3], 377-80

[446-86/87-056]

Surface Diffusion - Quantitative DataSecond harmonic microscopy was used to study In diffusion on (111) surfaces. It wasfound that, for homologous temperatures that were near to 0.5 and coverages rangingfrom 0 to 0.7, the In diffusivity could be described by:

D (cm2/s) = 3000 exp[-42(kcal/mol)/RT]The values of the Arrhenius parameters, which were quite large, were explained semi-quantitatively in terms of an adatom-vacancy model which had been developed for similarsystems. The present work offered considerable evidence for the effects of adatom-vacancy ionization.C.E.Allen, R.Ditchfield, E.G.Seebauer: Journal of Vacuum Science and Technology A,1996, 14[1], 22-9

[446-141/142-118]

406

KBulk Diffusion - Theoretical AnalysisThe free energy for diffusion in crystalline Si was calculated by combining athermodynamic integration method with ab initio molecular dynamics simulations. Thediffusion entropy was found to be negative.V.Milman, M.C.Payne, V.Heine, R.J.Needs, J.S.Lin, M.H.Lee: Physical Review Letters,1993, 70[19], 2928-31

[446-106/107-142]

LiBulk Diffusion - Qualitative Observations - Effect of DefectsThe lattice sites of implanted 8Li were studied by using the α-emission channellingtechnique. Tetrahedral interstitial sites were found to be the occupied (60 to 80%) latticesites following room-temperature implantation. Depending upon the dopingcharacteristics of the sample, up to 55% of the Li was found on bond-center sites afterimplantation at temperatures of between 425 and 475K. The change in lattice site, fromtetrahedral to bond-centered, was attributed to the onset of interstitial Li diffusion and toits capture by additional defects. The latter were expected to be vacancy-type defectswhich arose from implantation. The trapping of Li at these defects inhibited its long-rangediffusion.U.Wahl, S.G.Jahn, M.Restle, H.Quintel, H.Hofsäss, Isolde: Materials Science Forum,1995, 196-201, 115-20

[446-127/128-144]

Bulk Diffusion - Qualitative Observations - Effect of TrappingDiffusion in amorphous hydrogenated material, as studied by using implantation andsecondary ion mass spectroscopic techniques, was found to be trap-limited. No time orconcentration dependence of Li diffusion was observed. It was suggested that Li diffusedlargely in a positive charge state and was trapped mainly by extrinsic doping-relatednegatively charged ions or O.

407

Li Bulk Diffusion Li

W.Beyer, U.Zastrow: Journal of Non-Crystalline Solids, 1993, 164-166[1], 289-92[446-113/114-047]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsBy using deep-level transient spectroscopy, combined with secondary-ion massspectroscopy and capacitance-voltage profiling, it was demonstrated that Li diffusion intoAu-doped n-type samples at temperatures of between 200 and 300C resulted in theformation of two Li-Au complexes. One of these complexes appeared to be electricallypassive, and was observed indirectly as Au acceptor passivation. It was noted that nearlyall of the passivated Au acceptors were reactivated after annealing (400C, 0.5h) sampleswith comparable Au and Li concentrations; which were of the order of 1014/cm3. Theprocess could be reversed again by further heat treatment at lower temperatures. Thepassivation-reactivation cycle could be repeated as long as there was enough Li in thecrystal. This reaction could be described by a mass-action law involving negativelycharged Au atoms and positively charged Li (Au- + Li+) with a free binding energy ofabout 0.87eV. The other Au-Li complex had a deep level (L1), within the Si band-gap,with an activation energy of 0.41eV. The L1 signal was strongest after annealing attemperatures of between 250 and 300C, but was weaker at lower temperatures where theelectrically passive Au-Li complex was favored. On the basis of the dissociation kineticsof L1 during reverse bias annealing, it was deduced that the complex consisted of one Auatom and one or more Li atoms. By using deep-level transient spectroscopy depthprofiling, it was also observed that the injection of H into the surface region by wetchemical etching resulted in deactivation of the L1 trap.E.O.Sveinbjörnsson, S.Kristjansson, H.P.Gislason: Journal of Applied Physics, 1995,77[7], 3146-54

[446-121/122-088]

Bulk Diffusion - Theoretical AnalysisThe free energy for diffusion in crystalline Si was calculated by combining athermodynamic integration method with ab initio molecular dynamics simulations. Thediffusion entropy was found to be negative.V.Milman, M.C.Payne, V.Heine, R.J.Needs, J.S.Lin, M.H.Lee: Physical Review Letters,1993, 70[19], 2928-31

[446-106/107-142]

Bulk Diffusion - Theoretical AnalysisTransient ion drift in the depletion regions of a Schottky barrier was used to investigatediffusion in B- or Al-doped material. In order to test the validity of a model that was usedto extract the diffusion coefficients, the method was applied to the case of Li diffusion.Excellent agreement with published diffusivity data was found for Li ions.A.Zamouche, T.Heiser, A.Mesli: Applied Physics Letters, 1995, 66[5], 631-3

[446-125/126-148]

408

Mn

Figure 10: Diffusivity of Mn in Si (see table 57)

357 Bulk Diffusion - Quantitative DataDeep levels which were related to Mn, in n-type and p-type material, were studied bymeans of deep-level transient spectroscopy and Hall effect measurements. Two electrontraps, at Ec - 0.12eV and Ec - 0.41eV, and a hole trap, at Ev + 0.32eV, were found in theMn-doped material. The energy levels of these traps corresponded to transitions between 4charge states (Mn-, Mno, Mn+, Mn2+) of interstitial Mn. An additional donor-type electrontrap, at Ec - 0.51eV, was detected in n-type samples, and the trap was attributed tosubstitutional Mn. An electron trap at Ec - 0.50eV was observed in n+p junction sampleswhich had been diffused with Mn in B-doped p-type material. This trap was attributed to aMn-B complex which was formed via a pairing reaction between interstitial Mn and

1.0E-16

1.0E-15

1.0E-14

1.0E-13

1.0E-12

27 28 29 30 31 32 33 34 35

34.8433.1131.5530.0328.7427.55

104/T(K)

D (cm2/s)

409

Mn Bulk Diffusion Mn

substitutional B. By studying this pairing reaction, it was found that the diffusivity ofinterstitial Mn at temperatures of between 14 and 90C (table 57) could be described by:

D(cm2/s) = 0.0024 exp[-0.72(eV)/kT]H.Nakashima, K.Hashimoto: Journal of Applied Physics, 1991, 69[3], 1440-5

[446-78/79-055]

Table 57Diffusivity of Mn in Si

T (C) D (cm2/s)14 5.1 x 10-16

29 2.8 x 10-15

44 7.4 x 10-15

60 3.2 x 10-14

75 8.3 x 10-14

90 2.6 x 10-13

Bulk Diffusion - Quantitative DataMigration was studied by using deep-level transient spectroscopic techniques, or bydetermining the diffusion profiles of a deep level within depletion regions. The profilescould be accurately described by solutions of Fick's equation. Near to room temperature,the diffusivity was studied by monitoring pair reactions with substitutional B. It wasfound that the results, at temperatures ranging from 287 to 363K, could be described by:

D(cm2/s) = 0.0024 exp[-0.72(eV)/kT]H.Nakashima, K.Hashimoto: Materials Science Forum, 1992, 83-87, 227-32. See also:Materials Science Forum, 1994, 143-147, 761-6

[446-93/94-052], [446-113/114-044]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesA radiotracer technique was used to investigate how the diffusion of Mn from a layer on aSi surface affected the distribution of 63Ni in material which had been doped with Niusing a diffusion technique (1200C, 2h). The results revealed a sharp decrease (by anorder of magnitude) in the Ni concentration within the interior; due to the repeatedannealing of samples with a Mn layer at the surface. This effect was attributed to agettering that was induced by the Mn layer.G.S.Kulikov, J.A.Chichikalyuk, S.A.Yusupova: Fizika i Tekhnika Poluprovodnikov,1995, 29[3], 469-73 (Semiconductors, 1995, 29[3], 242-4)

[446-134/135-161]

Bulk Diffusion - Qualitative Observations - Effect of DefectsInteractions between group-VI elements (S, Se) and fast-diffusing Mn impurities werestudied. Infra-red absorption, electron spin resonance, and neutron activation studiesshowed that, in bulk Si, the group-VI elements effectively interacted with each other orwith the fast-diffusing impurities. It was found that, for each pairing of a group-VI

410

Mn Bulk Diffusion Mo

element with a fast-diffusing impurity, there was a certain annealing temperature at whichthey interacted most efficiently. A definite correlation was established between thisannealing temperature and the thermodynamic Gibbs free energy (at 298K) of thecorresponding compounds. This correlation suggested that the interaction process mainlyinvolved the formation of electrically neutral chemically bonded complexes bysubstitutional group-VI element centers and interstitial fast-diffusant centers.M.K.Bakhadirkhanov, S.I.Askarov, N.Norkulov: Physica Status Solidi A, 1994, 142[2],339-46

[446-117/118-194]

MoBulk Diffusion - Quantitative DataInterdiffusion and structural changes, which occurred upon annealing sputter-depositedMo/Si and Mo(N)/Si(N) multilayer thin films, were investigated at temperatures rangingfrom 674 to 1027K. The use of X-ray diffractometry showed that, in the as-depositedMo/Si multi-layers, the Mo was body-centered cubic; with the (110) planes parallel to thesubstrate. The Si was amorphous. In the case of as-deposited Mo(N)/Si(N) multi-layers,both the Mo and Si nitrides were amorphous. The interdiffusivities were deduced fromthe decay rate of satellite peak intensifies around (000). The activation energies for theinterdiffusion in Mo/Si and Mo(N)/Si(N) multi-layers were 105 and 351kJ/mol,respectively. A sharp decrease in the satellite peak intensity upon annealing was observedin Mo/Si multi-layer films. This was attributed to interdiffusion and structural relaxation.On the other hand, a marked increase in the satellite intensity was found for the nitridemulti-layer films. This was explained by crystallization into β-Mo2N and α-Si3N4. Themodulation wavelength decreased by 8 to 12% after annealing. A decrease in thethickness of annealed Mo/Si multilayer films was revealed by depth profiling.H.Nakajima, H.Fujimori, M.Koiwa: Journal of Applied Physics, 1988, 63[4], 1046-51

[446-62/63-233]

411

N

Figure 11: Diffusivity of N in Si (see table 58)

358 Bulk Diffusion - Quantitative DataThe out-diffusion profiles of N from float-zone material were measured by usingsecondary ion mass spectrometry. The diffusion coefficient was determined attemperatures ranging from 800 to 1200C. The total amount of N out-diffusion agreed withthe change in infra-red absorption by heat treatment. The diffusivity data (table 58) couldbe described by:

D(cm2/s) = 2700 exp[-2.9(eV)/kT]These values were 5 orders of magnitude larger than previously reported ones. It wassuggested that the former results applied to N-N pair-like molecules, while the latter

1.0E-10

1.0E-09

1.0E-08

1.0E-07

1.0E-06

6 7 8 9 10

9.327.867.867.867.287.286.79

104/T(K)

D (cm2/s)

412

N Bulk Diffusion N

applied to substitutional N atoms. It was concluded that the recognition of the existenceof the 2 types of N would permit a clarification of the effects of N upon the strengtheningof crystals and upon the suppression of swirls and D-defect generation.T.Itoh, T.Abe: Applied Physics Letters, 1988, 53[1], 39-41

[446-62/63-233]

Table 58Diffusivity of Paired N in Si

T (C) D (cm2/s)800 1 x 10-10

1000 2 x 10-8

1000 3 x 10-8

1100 2 x 10-7

1100 1 x 10-7

1200 3 x 10-7

Bulk Diffusion - Quantitative DataComplexes of O and N were formed in Czochralski-type material upon annealing in a Natmosphere. By fitting the depth profile of the defect to a diffusion equation, it wasdeduced that the diffusion coefficient of N was about 2 x 10-6cm2/s at 1270C. This valuewas some 105 times larger than those previously reported.A.Hara, T.Fukuda, T.Miyabo, I.Hirai: Applied Physics Letters, 1989, 54[7], 626-8

[446-64/65-179]

Bulk Diffusion - Quantitative DataAnalytical solutions were presented for Fick's one-dimensional diffusion equation in asemi-infinite medium with an exponentially decaying initial impurity concentrationprofile and various boundary conditions. The properties of the solution for a constantsurface concentration were discussed in more detail. The theoretical results were appliedto the diffusion of N which had been incorporated by laser melting. The diffusioncoefficient of N in Si near to its melting point was estimated to be of the order of 10

-6

cm2/s. This result confirmed previously reported high diffusion coefficients for N in Si.

G.J.Willems, H.E.Maes: Journal of Applied Physics, 1993, 73[7], 3256-60[446-099/100-097]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesA secondary ion mass spectroscopic study was made of the redistribution of in situdoped or implanted N in polycrystalline material, and of N segregation at Si/SiO2

interfaces during heat treatment at 700 to 1000C. When N-doped samples were subjectedto heat treatment at temperatures above 800C, the N diffused to the Si/SiO2 interface andsurface and piled up there. Some of the N was immobile when the concentration wasabove a threshold value. This immobile N became mobile and diffused during annealing.The threshold concentration for N diffusion depended upon the grain size. There was a

413

N Bulk Diffusion Na

limit on how much N segregated to the interface. This limit did not depend upon theinitial N content, but only upon the annealing temperature. A comparison of data forpolycrystalline films, with data on bulk material, suggested that the redistribution of N inthe former was limited by the transformation process via which immobile N becamemobile.S.Nakayama, T.Sakai: Journal of Applied Physics, 1996, 79[8], 4024-8

[446-134/135-162]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationA secondary ion mass spectrometry study of the co-implantation of N, C and O into float-zone material, followed by rapid thermal annealing (10s) at various temperatures, wasused to investigate the anomalous diffusion behavior of N. The results could be onlypartially explained by a model which involved paired N atom diffusion. The complexityof N diffusion in ion-implanted samples, with and without co-implants (and theexpectation that post-annealing N could be in many different forms), suggested thatstudies which used N implantation in fundamental studies of N-related defects could bemisleading.R.S.Hockett: Applied Physics Letters, 1989, 54[18], 1793-5

[446-70/71-126]

NaBulk Diffusion - Qualitative Observations - Concentration ProfilesMobile ion contamination of p+ and n+ polysilicon gates was studied by using thetriangular voltage sweep method and secondary ion mass spectroscopy. The mobile iondensities for p+ polysilicon were much higher than those for n+ polysilicon. Most of theNa+ ions in the gate oxide, which predominated over mobile ions, diffused laterally fromthe gate edge in the case of p+ polysilicon; rather than through the polysilicon. In the caseof n+ polysilicon, the laterally diffusing Na+ ions were gettered more effectively than werethose which penetrated vertically through the polysilicon.H.Tanaka, I.Aikawa, T.Ajioka: Journal of the Electrochemical Society, 1990, 137[2],644-7

[446-74-053]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationAn extensive review was presented of the subject of Na implantation. The topics whichwere covered included the solubility and diffusion of Na, radiation damage effects uponthe diffusion of implanted Na atoms (radiation damage proper, defects produced by

414

Na Bulk Diffusion Na

bombardment with other ions, residual defects), electrical characteristics of doped layers(electrical characteristics at room temperature, low-temperature Hall-effect studies, effectof radiation defects upon the electrical activity of Na), specific features of defectannealing, interaction of Na atoms with point defects, and Na-ion implantation.V.M.Korol: Physica Status Solidi A, 1988, 110[1], 9-34

[446-74-053]

Bulk Diffusion - Theoretical AnalysisThe free energy for diffusion in crystalline Si was calculated by combining athermodynamic integration method with ab initio molecular dynamics simulations. Thediffusion entropy was found to be negative.V.Milman, M.C.Payne, V.Heine, R.J.Needs, J.S.Lin, M.H.Lee: Physical Review Letters,1993, 70[19], 2928-31

[446-106/107-142]

415

Ni

Figure 12: Diffusivity of Ni in Si (see table 59)

Bulk Diffusion - Quantitative DataNear-surface Ni-rich layers were observed by means of the secondary ion massspectroscopy of intentionally contaminated wafers after thermal oxidation (950C, 2100s).The diffusivity of Ni, at 950C, which was deduced from secondary ion mass spectroscopyprofiles was 3 x 10-14 cm2/s. This value was almost 9 orders of magnitude lower than thatexpected for interstitial diffusants. It was comparable to those of group-III and group-Vatoms; which were known to be substitutional diffusants. On the basis of the results, amodel was proposed in which the interstitial and substitutional atoms diffused

1.0E-15

1.0E-14

1.0E-13

1.0E-12

1.0E-11

14 15 16 17 18 19

14.1214.9715.9217.4518.42

104/T(K)

D (cm2/s)

416

Ni Bulk Diffusion Ni

independently, and the slow substitutional diffusion gave rise to a near-surface impurity-rich layer.L.Zhong, F.Shimura: Japanese Journal of Applied Physics 2, 1993, 32[8B], L1113-6

[446-106/107-137]

Table 59Diffusivity of Ni in Amorphous Si

T (C) D (cm2/s)435 1.2 x 10-12

395 4.2 x 10-13

355 1.1 x 10-13

300 1.2 x 10-14

270 2.4 x 10-15

Bulk Diffusion - Quantitative DataThe diffusion of Ni in P-doped Czochralski monocrystals was studied by using a 63Niradiotracer, autoradiographic, and sectioning techniques. It was found that the data fortemperatures of between 220 and 540C, were described by:

D(cm2/s) = 0.0006 exp[-0.76(eV)/kT]The data supported an interstitial diffusion mechanism, with no participation of native Sidefects. The solubility of Ni at 800C was of the order of 1016/cm3. At low temperatures,the solubility was unclear due to experimental limitations. An analysis of previous resultsfor Ni diffusion revealed a scatter of 10 orders of magnitude. It was suggested that studieswhich involved low diffusion coefficients and large activation energies were dubious, dueto problems which were associated with surface conditions and with the limits of Nidetection. The latter arose from the rapid diffusion and low solubility of Ni.F.H.M.Spit, D.Gupta, K.N.Tu: Physical Review B, 1989, 39[2], 1255-60

[446-64/65-179]

359 Bulk Diffusion - Quantitative DataMeasurements were made of the diffusion coefficient of Ni in unrelaxed amorphousmaterial, at temperatures between 270 and 435C, by means of secondary ion massspectrometry. The data (table 59) could be described by:

D(cm2/s) = 0.003 exp[-1.30(eV)/kT]These values were some 6 to 8 orders of magnitude lower than that for the interstitialdiffusion of Ni in crystalline Si. The diffusion process in amorphous material could bedescribed by a model which invoked trap-retarded interstitial migration. The differencebetween the diffusion coefficients in amorphous and crystalline material was attributedmainly to the presence of intrinsic traps in the amorphous phase; with a binding enthalpyof about 0.83eV. Moreover, the experimental data indicated a lower attempt frequency fortrapped Ni atoms than for free interstitial Ni.A.J.Kuznetsov, B.G.Svensson: Applied Physics Letters, 1995, 66[17], 2229-30

[446-123/124-184]

417

Ni Bulk Diffusion/Surface Diffusion Ni

Bulk Diffusion - Qualitative Observations - Concentration ProfilesSamples with oxide layers which were between 120 and 400nm in thickness werediffused with Ni, through the oxide, at temperatures of between 1420 and 1520K. The Nidistribution was determined by using a 4-probe resistivity technique. It was found that thedistributions were anomalous in nature. The impurity was more heterogeneouslydistributed in the sub-surface region, and the concentration of electrically active Ni atomsincreased with decreasing oxide layer thickness. The degree of heterogeneity of theimpurity distribution in the sub-surface region, and the extent of this region, decreasedupon increasing the thickness of the oxide layer.F.M.Talipov, M.K.Bakhadyrkhanov, U.B.Soltamov: Izvestiya Akademii Nauk SSSR -Neorganicheskie Materialy, 1989, 25[6], 1037-8 (Inorganic Materials, 1989, 25[6], 872-3)

[446-76/77-039]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesAn investigation was made of diffusion in wafers at 1200C. The wafers of Czochralski-type material had a Ni film on one surface and were mechanically damaged on the othersurface. They were annealed in various ambients. It was found that, when the wafers wereannealed in vacuum, a Ni-rich liquid phase formed at the opposite surface. The quantityof liquid phase increased with increasing annealing time. After cooling, the solidifiedmaterial exhibited various morphologies, including dendrites and terraces. Thethermodynamic driving force for these was suggested to be the Si surface free energy.J.Li, W.S.Yang, T.Y.Tan, S.Chevacharoenkul, R.Chapman: Journal of Applied Physics,1992, 71[1], 196-203

[446-86/87-057]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsTwo mechanisms (dissociative, kick-out) have been proposed for the site exchange ofimpurities. Here, numerical and approximate solutions to the basic equations whichdescribed these mechanisms (for Ni in Si) were considered. The object was to decide howthese mechanisms might be distinguished by performing an in-diffusion experiment. Itwas found that the 2 mechanisms could be indistinguishable under certain conditions, ifvacancies and self-interstitials coexisted.H.Kitagawa, M.Yoshida: Japanese Journal of Applied Physics 1, 1992, 31[9A], 2859-63

[446-93/94-056]

Surface Diffusion - Qualitative Observations - Concentration ProfilesA direct comparison was made of the rates of bulk versus surface transport for Ni, in andon Si(111), by depositing a laterally confined dot of Ni onto one side of a doubly-polished and ultra-high vacuum-cleaned Si wafer and measuring the lateral Auger profileon the reverse side following annealing and quenching. It was found that the Ni reachedthe far side of the wafer, at temperatures as low as 550C, via bulk diffusion with nomeasurable contribution from surface paths; which were short-circuited by numerous fastbulk paths.M.Y.Lee, P.A.Bennett: Physical Review Letters, 1995, 75[24], 4460-3

[446-127/128-156]

418

Ni Surface Diffusion Ni

Surface Diffusion - Theoretical AnalysisThe molecular dynamics method was used to simulate the diffusion of adatoms on the(111) and (001) surfaces. A technique was proposed in which interatomic interactions inmulti-component systems could be obtained from a knowledge of the interactions withineach component. It was found that Ni atoms were more mobile than Si adatoms.P.Ashu, C.C.Matthai, T.H.Shen: Surface Science, 1991, 251-252, 955-9

[446-84/85-072]

Interdiffusion - Quantitative DataInterdiffusion in amorphous Ni-Si multi-layers was investigated by using an in situ X-raydiffraction technique. It was found that the temperature-dependent interdiffusivity,obtained by monitoring the decay of the first-order modulation peak as a function ofannealing time, could be described by:

D(m2/s) = 2.3 x 10

-17 exp[-0.61(eV)/kT]at temperatures of between 423 and 613K. It was suggested that a retarded interstitialdiffusion mechanism accounted for diffusion in the amorphous multi-layer films.W.H.Wang, H.Y.Bai, W.K.Wang: Journal of Applied Physics, 1993, 74[4], 2471-4

[446-113/114-182]

419

O

Figure 13: Diffusivity of O in Si

Bulk Diffusion - Quantitative DataThe relaxation of stress-induced dichroism of the 0.009mm band was measured in orderto determine the O diffusion coefficient at temperatures of between 250 and 400C.Enhanced diffusion, with an activation energy of 1.85eV, occurred as a transient behaviorin certain samples and was attributed to O-vacancy interactions. It was shown that A-centers, [O-V], annealed in two stages. The first involved an activation energy of 1.8eV.There was no evidence for an enhanced rate of long-range mass transport.

1.0E-241.0E-231.0E-221.0E-211.0E-201.0E-191.0E-181.0E-171.0E-161.0E-151.0E-141.0E-131.0E-121.0E-111.0E-101.0E-091.0E-08

6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

table 60table 61table 62table 63table 65table 66

104/T(K)

D (cm2/s)

420

O Bulk Diffusion O

A.K.Tipping, R.C.Newman, D.C.Newton, J.H.Tucker: Materials Science Forum, 1986,10-12, 887-92

[446-49-029]

Bulk Diffusion - Quantitative DataFrom O out-diffusion profiles for Czochralski-type material at 500C, measured usingsecondary ion mass spectrometry, O diffusivities of between 2.5 x 10-14 and 4.0 x 10-14

cm2/s were deduced. These values were higher, by nearly four orders of magnitude, thanthe normal diffusivity. The diffusion of implanted 18O in float-zone material at 400 to525C yielded secondary ion mass spectrometry profiles which consisted of an exponentialdecay of the O concentration. Exponential profiles of 18O above the background(1014/cm3), which were as deep as 0.004 to 0.016mm, also revealed direct evidence forenhanced long-range O diffusion. The enhanced diffusion was explained in terms of afast-diffusing species which was in dynamic equilibrium with interstitial O.S.T.Lee, P.Fellinger: Materials Science Forum, 1986, 10-12, 1021-6

[446-49-029]

Bulk Diffusion - Quantitative DataA theoretical calculation was made of the diffusivity of O in crystalline material. Thiswas based upon constrained-path energy minimization and jump-rate theory by using anempirical interatomic potential which had been newly developed for modellinginteractions between Si and O atoms. The calculations predicted that an O atom jumped,on (110) planes, from one bond-center site to another. The saddle-point configuration wasfarther away from the starting configuration than was the midpoint of the path. The Odiffusivity was predicted to be given by:

D(cm2/s) = 0.025 exp[-2.43(eV)/kT]and was in excellent agreement with experiment.Z.Jiang, R.A.Brown: Physical Review Letters, 1995, 74[11], 2046-9

[446-121/122-089]

Bulk Diffusion - Quantitative DataSecondary ion mass spectrometry was used to determine out-diffusion profiles from Sb-doped (6 x 1015 to 3 x 1018/cm3) Czochralski material. These profiles were then used toestimate the diffusivity of 16O. Its value at 1100C was 1.1 x 10-10cm2/s. It was found thatthe out-diffusion behavior was not affected by the Sb content.W.Wijaranakula, J.H.Matlock, H.Mollenkopf: Applied Physics Letters, 1988, 53[12],1068-70

[446-62/63-233]

360 Bulk Diffusion - Quantitative DataThe diffusivity of O at temperatures of between 500 and 650C was measured directly forthe first time by using microprobe techniques. It was found that the O diffusivity withinthis temperature range was enhanced, relative to the normal diffusivity of interstitial O.The degree of enhancement increased with decreasing temperature, and the enhancementfactor reached a value of more than 100 at temperatures below 550C. The diffusivity attemperatures of between 550 and 750C varied by up to an order of magnitude between

421

O Bulk Diffusion O

wafers (table 60). The O diffusivity generally decreased with increasing annealing time,but was not proportional to the O concentration.S.T.Lee, P.Fellinger, S.Chen: Journal of Applied Physics, 1988, 63[6], 1924-7

[446-72/73-047]

Table 60Diffusivity of O in Si Wafers

Orientation Resistivity (Ω cm) T (C) D (cm2/s)(111) 10-20 500 4.3 x 10-14

(111) 10-20 550 1.6 x 10-14

(111) 10-20 600 2.0 x 10-14

(111) 10-20 650 2.3 x 10-14

(111) 10-20 700 4.3 x 10-14

(111) 10-20 750 1.2 x 10-13

(111) 10-20 800 1.2 x 10-13

(111) 10-20 900 1.5 x 10-12

(111) 10-20 1000 1.1 x 10-11

(100) 3 650 1.4 x 10-14

(100) 3 700 6.9 x 10-14

(100) 3 750 1.3 x 10-13

(100) 3 800 2.1 x 10-13

(100) 3 900 1.7 x 10-12

(100) 3 1000 1.2 x 10-11

(100) 1-2 500 4.2 x 10-14

(100) 1-2 550 9.3 x 10-15

(100) 1-2 600 8.2 x 10-15

(100) 1-2 650 1.4 x 10-14

(100) 1-2 700 4.1 x 10-14

(100) 1-2 750 8.5 x 10-14

(100) 1-2 900 1.8 x 10-12

(100) 1-2 1000 1.1 x 10-11

(100) 8-16 500 2.3 x 10-14

(100) 8-16 550 4.5 x 10-15

(100) 8-16 600 6.2 x 10-15

(100) 8-16 650 5.6 x 10-15

(100) 8-16 700 1.5 x 10-14

(100) 8-16 750 3.8 x 10-14

(100) 8-16 900 1.8 x 10-12

(100) 8-16 1000 1.1 x 10-11

(100) 9 550 1.4 x 10-14

continued

422

O Bulk Diffusion O

Table 60 (continued)Diffusivity of O in Si Wafers

Orientation Resistivity (Ω cm) T (C) D (cm2/s)(100) 9 600 2.3 x 10-14

(100) 9 650 1.1 x 10-14

(100) 9 700 4.4 x 10-14

(100) 9 750 1.2 x 10-13

(100) 9 800 2.0 x 10-13

(100) 9 800 2.1 x 10-12

(100) 9 1000 1.2 x 10-11

(100) 10-20 600 5.6 x 10-15

(100) 10-20 650 6.7 x 10-15

(100) 10-20 700 2.2 x 10-14

(100) 10-20 750 5.4 x 10-14

(100) 10-20 800 2.1 x 10-13

(100) 10-20 900 1.4 x 10-12

(100) 10-20 1000 1.2 x 10-11

(100) 16 550 2.6 x 10-14

(100) 16 600 3.1 x 10-14

(100) 16 650 3.3 x 10-14

(100) 16 700 4.1 x 10-14

(100) 16 750 1.3 x 10-13

(100) 16 800 1.4 x 10-13

(100) 16 1000 1.2 x 10-11

361 Bulk Diffusion - Quantitative DataThe diffusivity of O in heavily Sb-doped Czochralski-type material was measured, attemperatures ranging from 950 to 1100C, by means of secondary ion mass spectrometry.It was found that the diffusion coefficient exhibited no dependence upon the Sbconcentration. The results (table 61) indicated a diffusion activation energy of 2.68eV.M.Pagani: Journal of Applied Physics, 1990, 68[7], 3726-8

[446-86/87-057]

Bulk Diffusion - Quantitative DataThe migration of O in C-doped material, during extended isochronal annealing attemperatures ranging from 460 to 850C, was studied. It was found that, at temperaturesbelow 690C, enhanced O diffusion occurred. The diffusivity at the highest temperaturewas equal to about 7.3 x 10-13cm2/s. This behavior had previously been attributed to theeffect of mobile C and O complexes. It was suggested here that these complexes couldalso act as an activated state in an enhanced diffusion process.

423

O Bulk Diffusion O

W.Wijaranakula: Journal of Applied Physics, 1990, 68[12], 6538-40[446-86/87-057]

Table 61Diffusion of O in Sb-Doped Si

T (C) D (cm2/s)950 1.2 x 10-11

950 1.3 x 10-11

950 8.7 x 10-12

1025 2.5 x 10-11

1100 1.3 x 10-10

1100 1.1 x 10-10

1100 8.2 x 10-11

362 Bulk Diffusion - Quantitative DataSamples of B-doped (about 1017/cm3) Czochralski material were heated to temperaturesranging from 800 to 1300C in H, and were then quenched. The concentration of [H-B]pairs was then measured by means of infra-red localized vibrational mode spectroscopy.It was concluded that the solubility of atomic H was greater than:

S(/cm3) = 5.6 x 1018 exp[-0.95(eV)/kT]within the above temperature range. Undoped Czochralski material was also annealedunder similar conditions. This showed that the diffusion of O (table 62) could bedescribed by:

D(cm2/s) = 0.00071 exp[-2.0(eV)/kT]S.A.McQuaid, R.C.Newman, J.H.Tucker, E.C.Lightowlers, R.A.A.Kubiak, M.Golding:Applied Physics Letters, 1991, 58[25], 2933-5

[446-84/85-073]

363 Bulk Diffusion - Quantitative DataThe out-diffusion of O, from Czochralski wafers which had been annealed at 1000 or1200C in a H ambient, was studied by means of secondary ion mass spectroscopy. Theexpression,

D(cm2/s) = 141 exp[-3.1(eV)/kT]

was deduced by fitting the O secondary ion mass spectroscopy profile, and thediffusivities (table 63) were significantly higher than expected. This H enhancementeffect was found at temperatures which were much greater than those (below 500C)which were reported in the literature. The enhancement was attributed to directinteraction between in-diffused H and interstitial O atoms. The O diffusivity which wasdeduced from H solubility and diffusivity data was in reasonable agreement with theexperimental results.L.Zhong, F.Shimura: Journal of Applied Physics, 1993, 73[2], 707-10

[446-106/107-142]

424

O Bulk Diffusion O

364 Bulk Diffusion - Quantitative DataInterstitial O profiles across epitaxial/Czochralski interfaces were studied by using highspatial resolution Fourier transform infra-red spectroscopy. Transmission measurementsof a transverse wafer cross-section revealed an O contamination of the epilayer whichwas due to solid-state out-diffusion from the substrate during epilayer deposition. The Odiffusivity values which were deduced from the experiments (table 64) suggested that themechanism was hardly affected by the interface. The O contamination was strictly relatedto the type of dopant present in the substrate and not to that present in the epilayer. Suchcontamination of the epilayer, which was significant in n-type substrate samples, wassuggested to explain the structural defects which were often observed in epitaxial layers.M.Geddo, B.Pivac, A.Sassella, A.Stella, A.Borghesi, A.Maierna: Journal of AppliedPhysics, 1992, 72[9], 4313-20

[446-106/107-143]

Table 62Diffusivity of O in Czochralski-Type Si

T (C) D (cm2/s)340 2.7 x 10-20

340 2.2 x 10-20

325 7.0 x 10-21

300 1.4 x 10-21

365 Bulk Diffusion - Quantitative DataThe relaxation of stress-induced dichroism of the 9000nm O infra-red absorption bandwas investigated in samples of Czochralski material which had been annealed in a Hplasma at temperatures of between 225 and 350C. It was found that the in-diffusion of Hatoms enhanced the rate of O diffusion (table 65), so that dichroism disappeared graduallyfrom the external surfaces. Other measurements indicated that O diffusion jumps werecatalysed by collisions with diffusing H atoms. Increased rates of thermal donorformation were attributed to enhanced long-range O diffusion. It was concluded that Hatom concentrations which were as low as 108/cm3 could significantly enhance Odiffusivity.R.C.Newman, J.H.Tucker, A.R.Brown, S.A.McQuaid: Journal of Applied Physics, 1991,70[6], 3061-70

[446-91/92-031]

366 Bulk Diffusion - Quantitative DataDirect measurements were made of the diffusivity of interstitial O in Sb-dopedCzochralski material at temperatures ranging from 750 to 1150C (table 66). Usingsecondary ion mass spectroscopy of the out-diffusion profiles, it was shown that thediffusivity was the same as that for lightly B-doped crystals which were heated underidentical conditions over the temperature range studied.A.S.Oates, W.Lin: Applied Physics Letters, 1988, 53[26], 2659-61

[446-64/65-179]

425

O Bulk Diffusion O

Bulk Diffusion - Qualitative Observations - Concentration DependenceWafers with a range of initial O and C concentrations were annealed at 450, 475, or500C, for up to 500h. The interstitial O concentration was monitored every 20h. It wasfound that the rate of O loss scaled as the fifth power of the O concentration in waferswhich were annealed at temperatures below 500C.M.P.Guse, R.Kleinhenz: Journal of Applied Physics, 1992, 72[10], 4615-8

[446-106/107-143]

Table 63Diffusivity of O in Si

T (C) D (cm2/s)1195 5.0 x 10-9

1000 1.2 x 10-10

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe solid-state out-diffusion of O from a substrate to an epitaxial layer was investigatedby using micro-Fourier transform infra-red measurements in a transverse wafer cross-sectional configuration. The interstitial O concentrations, which were obtained byanalysing the 1107/cm absorption band, indicated that an O content (clearly detectable viathe use of infra-red techniques) was present in an epitaxial layer near to the interface. Itwas suggested that this was the first evidence for O out-diffusion from the substrate andinto the epitaxial layer.M.Geddo, B.Pivac, A.Borghesi, A.Stella, M.Pedrotti: Applied Physics Letters, 1990,57[15], 1511-3

[446-76/77-040]

Table 64Diffusivity of O in Epitaxial/Czochralski Si Layers

Epilayer Substrate D (cm2/s)n-type n-type 0.022p-type n-type 0.009p-type p-type 0.009n-type p-type 0.006

Bulk Diffusion - Qualitative Observations - Effect of ChargeA model which was based upon the space-charge assisted diffusion of O- was proposed inorder to describe rapid thermal oxidation kinetics in a dry O ambient. The space-chargeregion within the growing oxide layer was assumed to have a constant planar chargedensity which resulted in a characteristic length: the extent of the space-charge region.The diffusion of O-, and therefore the oxide growth rate, was expected to be driven by 2components. These were the concentration gradient and the space-charge drift. Adiscrepancy in the calculated values of the diffusion activation energies for O- wasexplained by comparing the strength of the space-charge regions. The latter strength

426

O Bulk Diffusion O

depended upon the range of photon energies of the lamps which were used for rapidthermal treatment.A.Kazor: Journal of Applied Physics, 1995, 77[4], 1477-81

[446-121/122-088]

Table 65Diffusivity of O in Si

H-Treatment T (C) D (cm2/s)furnace 385 6.1 x 10-21

furnace 345 3.5 x 10-22

furnace 325 1.2 x 10-22

furnace 325 8.4 x 10-23

plasma 350 2.0 x 10-20

plasma 350 1.2 x 10-20

plasma 325 3.6 x 10-21

plasma 325 2.9 x 10-21

plasma 300 3.0 x 10-21

plasma 300 2.4 x 10-21

plasma 275 3.7 x 10-22

plasma 275 2.4 x 10-22

plasma 250 1.4 x 10-22

plasma 250 1.2 x 10-22

plasma 240 6.3 x 10-23

plasma 240 4.8 x 10-23

plasma 225 1.8 x 10-23

plasma 225 1.1 x 10-23

Bulk Diffusion - Qualitative Observations - Effect of DefectsAs-grown Czochralski-type material, with various O contents, was heated to temperaturesranging from 350 to 500C. The O loss during annealing at temperatures of up to 400Cwas shown to obey second-order kinetics, and measurement indicated values of the Odiffusivity which were larger than normal by a factor of about 3; assuming that thecapture radius for dimer formation was 0.5nm. Variations in the rate of [Oi] loss duringmore extended annealing could be explained if the O diffusion was initially enhanced, buttended to its normal value as annealing progressed. A much greater initial enhancementwas deduced from similar measurements of samples which had been hydrogenated byheat treatment in H2 gas (1300C, 1.5h), followed by rapid quenching to room temperature.The enhancements were consistent with values which had been deduced frommeasurements of the relaxation of stress-induced dichroism. At temperatures greater than450C, the measured rates of [Oi] loss were less than the expected rate of Oi-Oi interactionand tended to vary with increasingly high powers of [Oi]. Modelling of the clustering

427

O Bulk Diffusion O

process demonstrated that the reductions could be explained if the O dimers were presentin a quasi-equilibrium concentration throughout annealing. The establishment of thisequilibrium appeared to require that O dimers should diffuse much more rapidly thanisolated Oi atoms. The kinetics of O loss over the entire range of temperatures could thenbe explained if dimer clustering led mainly to increases in the concentration ofagglomerates which contained large numbers of O atoms (more than 8). It was thereforepossible to account for thermal donor formation on the basis of the formation of Oclusters of various sizes, although the possibility that self-interstitials were involved inthermal donor formation was not excluded.S.A.McQuaid, M.J.Binns, C.A.Londos, J.H.Tucker, A.R.Brown, R.C.Newman: Journalof Applied Physics, 1995, 77[4], 1427-42

[446-121/122-089]

Table 66Diffusivity of O in Doped Si

Type T (C) D (cm2/s)p 1150 2.4 x 10-10

p 1050 6.5 x 10-11

p 950 7.3 x 10-12

p 950 6.4 x 10-12

p 850 1.4 x 10-12

p 850 7.5 x 10-13

p 750 3.3 x 10-14

p 750 2.9 x 10-14

n 1150 4.3 x 10-10

n 1050 6.5 x 10-11

n 950 1.4 x 10-11

n 950 1.1 x 10-11

n 850 1.7 x 10-12

n 750 9.6 x 10-14

Bulk Diffusion - Qualitative Observations - Effect of DefectsIt had previously been proposed that the extremely high rate of single diffusion jumps ofO atoms during the electron irradiation of crystals at temperatures greater than 300C wasdue to the sequential trapping and de-trapping of vacancies. To support this assumption, itwas demonstrated that thermal dissociation of up to 75% of the O-vacancy centersoccurred. This was done by monitoring their destruction in irradiated and annealedcrystals. An important new step was the calibration of the strength of the infra-red 830/cmabsorption band via a novel procedure which gave the concentration of O-vacancycomplexes in the crystals.A.S.Oates, R.C.Newman: Applied Physics Letters, 1986, 49[5], 262-4

[446-49-029]

428

O Bulk Diffusion O

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe precipitation of O was studied at temperatures of between 500 and 600C by usinginfra-red absorption measurements of the 0.0195mm band. The data were fitted to Ham'stheoretical models for the diffusion-limited growth of randomly distributed particles.Values of the equilibrium O concentration, Cs, were obtained directly. Assuming the usualdiffusion coefficient for interstitial O, the number density of the precipitates and theirspherical radii were determined. These results were linked to previous data obtained athigher and lower temperatures, and yielded a self-consistent model. It was found that thevalue of Cs decreased to 8 x 1015 at 650C, but then increased again to 2 x 1017/cm3 at450C. This previously unreported behavior was attributed to the importance of the surfaceenergy to very small agglomerates. The results implied that, if the thermal donors andcoesite precipitates which were observed at low temperatures arose from O aggregation,they could not form unless an O complex, such as a di-oxygen species, had a much largerdiffusion coefficient than an isolated interstitial O atom.S.Messoloras, R.C.Newman, R.J.Stewart, J.H.Tucker: Semiconductor Science andTechnology, 1987, 2[1], 14-9

[446-51/52-136]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe effect of Ge upon the diffusivity of O at 620 to 670K was studied, in dislocation-free[111]-type monocrystals, by using Hall-effect methods. It was found that, in comparisonwith un-doped monocrystals, the activation energy for O diffusion in Ge-doped materialwas much closer to the activation energy for O diffusion via the interstitial mechanism.D.N.Korlyakov: Izvestiya Akademii Nauk SSSR - Neorganicheskie Materialy, 1991,27[7], 1333-6 (Inorganic Materials, 1991, 27[7], 1129-32)

[446-88/89-050]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe infra-red absorption coefficient of the 0.009mm band was measured by using Fouriertransform infra-red spectroscopy at temperatures ranging from 77 to 775K. It was foundthat the center of the 0.009mm band shifted to longer wavelengths with increasingtemperature. The interstitial O concentration was almost constant at temperatures below600K, but decreased rapidly at temperatures above 600K. The results confirmed that therewere 2 types of O configuration: a bonded Si2O configuration with a binding energy of0.8 to 1.0eV at 77 to 600K, and a Si2O configuration which coexisted with a quasi-freeinterstitial O state at temperatures above 600K. The lattice potential barrier, whichimpeded the migration of quasi-free interstitial O atoms in the lattice, was estimated to bebetween 1.5 and 1.6eV. The anomalous diffusivity of O could be quite well explained interms of these configurations.C.S.Chen, F.Q.Zeng, Y.X.Huang, H.J.Ye, C.M.Hu, D.K.Schroder: Applied Physics A,1992, 55[4], 317-23

[446-93/94-057]

Bulk Diffusion - Qualitative Observations - Effect of DefectsStudies of the diffusion behavior of dopants and neutral impurities have revealed thatvacancies and self-interstitials coexist in Si at high temperatures. On the basis of this

429

O Bulk Diffusion O

conclusion, the effect of these point defects upon SiO2 precipitate growth wasinvestigated at temperatures ranging from 400 to 1200C. It was assumed that these pointdefects played an essential role in providing strain relief during SiO2 precipitate growth.The manner in which the point defect concentrations changed in the vicinity of theprecipitate was modelled, and it was shown that these local variations affected the flux ofthe O atoms via an O/point-defect reaction at the precipitate surface. In the 2 extremecases which were studied (vacancies only and interstitials only), the model predicted an Oflux which was lower than that limited by O diffusion alone. For the vacancy-only case,this difference amounted to some orders of magnitude, indicating that a V-only modelwas not capable of fitting experimental data. However, the interstitials-only modelpredicted a flux which was only 30% lower than that limited by O diffusion attemperatures as low as 400C. It was concluded that, while vacancies could contribute tosome extent, self-interstitials played the predominant role in providing strain relief duringthe precipitate growth.W.J.Taylor, T.Y.Tan, U.M.Gösele: Materials Science Forum, 1992, 83-87, 1451-6

[446-099/100-098]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe out-diffusion behavior of O from heat-treated Czochralski material was investigatedby using secondary ion mass spectroscopy. The results showed that O diffusion wasgreatly retarded by O precipitation. This strongly supported a vacancy-dominateddiffusion mechanism for O. In C-doped material, the diffusion of O was greatly enhancedat 750C, but was significantly retarded at 1000C. The enhanced diffusion was attributedto the formation of fast-diffusing O-C complexes.F.Shimura, T.Higuchi, R.S.Hockett: Applied Physics Letters, 1988, 53[1], 69-71

[446-62/63-229]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsThe diffusion, at temperatures below 500C, of interstitial O in Ge-doped crystals wasmonitored via the pressure-induced dichroism of an infra-red band at 1106/cm. Thedopant concentration ranged from 1018 to 2 x 1020/cm3. The results were compared withthe formation kinetics of thermal donors. This showed that the Ge dopant had an effect,mediated by a local stress field, upon the diffusion of O. That is, there was an increasedOi diffusion component. There was also an effect upon the generation and annihilation ofintrinsic point defects such as vacancies and interstitials (Sii). In particular, Sii generationwas hindered. On the other hand, the annihilation of these defects in the presence of Gewas promoted. The results were attributed to a slowing down, of the formation of thermaldonors, with increasing Ge concentration.V.M.Babich, N.P.Baran, K.I.Zotov, V.L.Kiritsa, V.B.Kovalchuk: Fizika i TekhnikaPoluprovodnikov, 1995, 29[1], 58-64 (Semiconductors, 1995, 29[1], 30-3)

[446-134/135-162]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsThe behavior of O in degenerately doped Czochralski material was studied after extendedannealing at temperatures ranging from 750 to 900C in a N ambient. Out-diffusion of Oappeared to be significantly retarded. On the basis of the defect morphology, as observed

430

O Bulk Diffusion O

using high-resolution electron microscopy, it was concluded that degenerate Si wassaturated with interstitial defect species during extended annealing. The retardation of Odiffusion was suggested to be caused by a strong pairing reaction between B and O atoms.W.Wijaranakula: Applied Physics Letters, 1993, 62[23], 2974-6

[446-106/107-137]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe out-diffusion of O from Czochralski material was studied under atmospheres of O, N,and Ar/H(10%). The enhancement of O diffusion by H was confirmed. Also, heattreatment in a halogen lamp furnace led to a 100-fold higher O diffusion coefficient in aneutral atmosphere, and to a 1000-fold enhancement in an hydrogenated atmosphere. Inthe latter case, a level of 6.5 x 10

17/cm

3 was observed in the O profile at the sample

surface. Electron-hole pair generation under intense ultra-violet irradiation, perhapscombined with the presence of H, was suggested to be responsible for anomalously highO diffusion during rapid thermal processing.C.Maddalon-Vinante, D.Barbier, H.Erramli, G.Blondiaux: Journal of Applied Physics,1993, 74[10], 6115-9

[446-109/110-045]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionCzochralski-type samples which contained a range of grown-in O concentrations wereannealed at temperatures of up to 500C, and the rates of Oi loss due to solution andthermal donor formation were monitored. The results could be explained in terms of serialO aggregation at a rate which was controlled by dimer formation, but only if O2 dimersdiffused more rapidly than Oi atoms and if the dimers and other small clusters dissociatedeasily at higher temperatures. A comparison of the deduced Oi diffusion coefficients, withthose deduced from the relaxation of stress-induced dichroism, implied enhancementfactors of 3 to 10 at 350C. This was attributed to the presence of grown-in H; consistentwith the much greater effect of sample pre-annealing in H2 gas at 1300C. The Oi

diffusivity could also be enhanced by the trapping and subsequent dissociation of rapidlydiffusing dimers. It was suggested that thermal donors could be associated with O clustersthat contained up to 10 Oi atoms.R.C.Newman, M.J.Binns, C.A.Londos, S.A.McQuaid, J.H.Tucker: Solid StatePhenomena, 1996, 47-48, 247-58

[446-134/135-162]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe properties of isolated O, C, N and H were considered with respect to O precipitationabove and below 500C. Thermal donors formed at the lower temperatures. In order togenerate O clusters that contained up to some 10 atoms at 450C, it was necessary to takethe dissociation of smaller O clusters into account and to assume the rapid diffusion of O2

dimers. It was demonstrated that as-grown Czochralski crystals could contain H whichenhanced the diffusion of single O atoms. These impurities were also incorporated intocertain O clusters which acted as shallow donors. It was suggested that dimer interactionwith Oi atoms led to an enhancement of the O diffusivity.R.C.Newman: Materials Science and Engineering B, 1996, 36[1-3], 1-12

[446-136/137-130]

431

O Bulk Diffusion O

Bulk Diffusion - Qualitative Observations - PositronsThe annihilation of grown-in defects in Czochralski-type material was investigated,during O out-diffusion, by making positron measurements using a variable-energy beam.It was found that the diffusion length of the positrons became greater as the annealingtime was increased, although the line-shape parameter was constant. The data ondiffusion lengths, based upon the so-called semi-trapping phenomenon (and including theformation of positron Rydberg states between positrons and defects), was more sensitiveto the presence of minute defects. An increase in diffusion length indicated that thegrown-in defects were annihilated. The concentration of annihilated defects, related to Oclusters, decreased to one tenth of that in as-grown crystals. The grown-in defectsannihilated at temperatures of the order of 1150C.T.Kitano, S.Saito, S.Tanigawa: Applied Physics Letters, 1994, 65[19], 2434-6

[446-119/120-224]

Bulk Diffusion - Theoretical Analysis - Activation EnergyIt was noted that the migration of O in this material corresponded to an apparently simplejump between neighboring bridge sites. However, theoretical calculations had producedconflicting results and had failed to provide a satisfactory explanation for the observedactivation energy of 2.5eV. Comprehensive first-principles calculations were presentedhere which demonstrated that the apparently simple O jump was actually a complexprocess which involved coupled barriers and could be properly described quantitatively interms of an energy hyper-surface with a saddle ridge and an activation energy of about2.5eV. Earlier calculations had corresponded to various points or lines on this hyper-surface.M.Ramamoorthy, S.T.Pantelides: Physical Review Letters, 1996, 76[2], 267-70

[446-131/132-184]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesAn empirical solution was developed for treating O distributions in thin epitaxial layersand substrate wafers which had been heavily doped with B and Sb. The results ofsecondary ion mass spectroscopic analyses indicated that heavy doping with B or Sb hadno effect upon O out-diffusion during epitaxial deposition. In contrast to slow-diffusingspecies such as B and Sb, the plane at which the O concentration was equal to 50% of thebulk O concentration did not remain at the epitaxial layer/substrate interface, but was afunction of the effective diffusion length of O atoms for the overall epitaxy process.W.Wijaranakula: Journal of Applied Physics, 1993, 73[2], 1004-6

[446-106/107-142]

Bulk Diffusion - Theoretical Analysis - Effect upon DefectsAn analysis was made of O migration near to a 60º dislocation, and of the consequentretardation of dislocation motion. Quantitative predictions were based upon a solution ofthe macroscopic equations for the transport of O in the elastic stress field that was createdby the dislocation. The link, between the microscopic dynamics of interstitial O within thediamond lattice and macroscopic transport, was expressed in terms of a constitutive modelfor the dependence of the drift velocity and diffusivity of O upon the elastic interaction ofO atoms and dislocations, and upon the temperature. The transport equation was solved

432

O Bulk Diffusion/Surface Diffusion OH

numerically by assuming that the dislocation core was fully saturated with O. The dragforce on the gliding dislocation, caused by the surrounding O, was deduced on the basisof linear elasticity theory.D.Maroudas, R.A.Brown: Journal of Applied Physics, 1991, 69[7], 3865-77

[446-86/87-057]

Bulk Diffusion - Theoretical Analysis - Enhanced DiffusionInteractions between interstitial H atoms and interstitial O atoms in crystalline materialwere analyzed by applying a parameter-free Hartree-Fock method to clusters whichcontained 5, 26 or 35 host atoms. It was found that no configurations with significant O-Hbonding were energetically favorable, and that the activation energy for the diffusion ofinterstitial O was considerably lower when H was present than when it was not. It wasdeduced that H enhanced the diffusivity of interstitial O by several orders of magnitude.S.K.Estreicher: Physical Review B, 1990, 41[14], 9886-91

[446-76/77-040]

OHSurface Diffusion - Theoretical AnalysisA local density functional investigation was made of the adsorption geometry and thesurface diffusion activation energies of hydroxyl (OH) radicals that resulted fromdissociative water adsorption on Si(100)-2 x 1 surfaces. In a similar manner to atomic H,the OH preferred to bind to a single surface Si atom. Due to both dative interactions withsurface dangling bonds, and to adsorbate-adsorbate H bond-like interactions, the O-Hbonds tended to be oriented perpendicularly to the dimer direction. This was in agreementwith electron stimulated desorption ion angular distribution data. The energetics of OHdiffusion on clean or saturated surfaces were similar to that of H, but with slightly lowerbarriers. In particular, the intra-dimer barrier was found to be about 0.2eV lower. Thisimplied that room-temperature intra-dimer adsorbate oscillations should occur some 1000times faster for OH. The absolute value of this barrier (0.9eV) was in accord withscanning tunnelling microscopic observations.A.Vittadini, A.Selloni, M.Casarin: Physical Review B, 1995, 52[8], 5885-9

[446-125/126-153]

433

P

Figure 14: Diffusivity of P in Si. The best least-squares regression line through thesepoints gives D (cm2/s) = 0.79 exp[-3.29(eV)/kT]

Bulk Diffusion - Quantitative DataThe migration of P under ambients of pure N2, pure NH3, or their mixtures, wasinvestigated in order to determine the effect of oxynitridation reactions upon diffusivity.In the presence of a thin SiO2 layer on the Si wafer, and a low P concentration, thediffusivity of P could be described by:

D(cm2/s) = 0.145 exp[-3.26(eV)/kT] + 1.718 x 10-6 exp[-1.72(eV)/kT]PNH3

1.0E-17

1.0E-16

1.0E-15

1.0E-14

1.0E-13

1.0E-12

1.0E-11

1.0E-10

6 7 8 9 10


104/T(K)

D (cm2/s)

434

P Bulk Diffusion P

The ratio of the interstitial concentration under oxynitridation conditions, to that underinert conditions, could be described by:

R = 1 + 0.00001183 exp[1.54(eV)/kT]PNH3

By applying the present results to published data, the fraction of B diffusion whichoccurred via an interstitial mechanism was estimated to be 0.88.N.K.Chen, C.Lee: Journal of the Electrochemical Society, 1995, 142[6], 2051-4

[446-134/135-158]

367 Bulk Diffusion - Quantitative DataThe diffusivity of P in profiled samples of Stepanov-prepared material was studied attemperatures of between 1015 and 1210C by using radiotracer, neutron activation andserial sectioning methods. It was found that the results (table 67) could be described bythe expression:

D(cm2/s) = 1.9 exp[-3.3(eV)/kT]K.P.Abdurakhmanov, M.B.Zaks, V.V.Kasatkin, G.S.Kulikov, S.K.Persheev,K.K.Khodzhaev: Fizika i Tekhnika Poluprovodnikov, 1988, 22[11], 2088-90 (SovietPhysics - Semiconductors, 1988, 22[11], 1324-5)

[446-70/71-126]

Table 67Diffusivity of P in Profiled Si

T (C) D (cm2/s)1210 9.0 x 10-12

1165 3.1 x 10-12

1105 1.4 x 10-12

1015 2.0 x 10-13

368 Bulk Diffusion - Quantitative DataThe P diffusivity was measured, in samples with almost uniform As or B backgrounddoping, at temperatures of 915, 1015 and 1105C (table 68). It was found that diffusion vianeutral and singly negatively charged point defects was sufficient to account for theexperimental data, and that there was no need to include diffusion via doubly negativelycharged point defects. A greatly reduced diffusivity was observed in B-doped samples,and this was consistent with the formation of immobile donor-acceptor pairs.F.Wittel, S.Dunham: Applied Physics Letters, 1995, 66[11], 1415-7

[446-121/122-090]

369 Bulk Diffusion - Quantitative DataDiffusion in monocrystalline Si and in a Si-TaSi2 eutectic alloy was investigated, using aP concentration of about 10

19/cm

3, at temperatures ranging from 767 to 1227C (table 69).

No clear difference could be found between P diffusion in the eutectic structure and in theSi. A non-linear Arrhenius plot was obtained, thus revealing P diffusion-enhancement atlower temperatures. The linear part could be described by:

435

P Bulk Diffusion P

D(cm2/s) = 5.7 exp[-3.75(eV)/kT]

J.Pelleg, B.M.Ditchek: Journal of Applied Physics, 1993, 73[2], 699-706[446-106/107-146]

Table 68Diffusivity of P in Si with Various Dopants

Dopant Concentration (/cm3) T (C) Diffusivity (cm2/s)B 5.0 x 1019 1105 5.4 x 10-14

B 5.0 x 1019 1015 3.1 x 10-15

B 5.0 x 1019 915 1.1 x 10-16

B 2.9 x 1019 1105 7.3 x 10-14

B 2.9 x 1019 915 1.9 x 10-16

B 1.6 x 1019 1015 9.2 x 10-15

B 1.6 x 1019 915 5.2 x 10-16

- - 1105 1.1 x 10-13

- - 1015 1.8 x 10-14

- - 915 2.1 x 10-15

As 3.1 x 1019 1015 3.3 x 10-14

As 3.1 x 1019 915 5.2 x 10-15

As 5.0 x 1019 1105 3.3 x 10-13

As 5.0 x 1019 1015 4.9 x 10-14

As 5.0 x 1019 915 7.4 x 10-15

As 1.1 x 1020 1105 5.4 x 10-13

As 1.1 x 1020 1015 7.1 x 10-14

As 1.1 x 1020 915 1.2 x 10-14

As 1.8 x 1020 1105 8.4 x 10-13

As 1.8 x 1020 1015 1.3 x 10-13

As 1.8 x 1020 915 1.5 x 10-14

370 Bulk Diffusion - Quantitative DataThe diffusion of donor elements in fine-grained and coarse-grained polycrystallinematerial was studied at temperatures ranging from 900 to 1150C (table 70). Tracers (32P)were used to determine the concentration/depth profiles via sectioning. By means ofautoradiography, the lateral distribution of the radiotracers over the sample surface wasrevealed.F.H.M.Spit, H.Bakker: Physica Status Solidi A, 1986, 97[1], 135-42

[446-48-063]

371 Bulk Diffusion - Quantitative DataThe supersaturation of interstitials, during oxidation in pyrogenic steam at 900 or 1000C,was deduced by measuring the enhancement of P diffusion (table 71). At 900C, the

436

P Bulk Diffusion P

supersaturation during steam oxidation was considerably lower than that observed duringdry oxidation at the same growth rate. Also, the interstitial supersaturation varied as thesquare root of the oxidation rate. At 1000C, the interstitial supersaturation was similar tothat observed in dry O2. The interstitial concentration exhibited only a ¼-powerdependence upon the oxidation rate. It was found that the annealing (in an inert ambient),of oxide which had been deposited at low temperatures, resulted in an enhanced Pdiffusivity.N.Jeng, S.T.Dunham: Journal of Applied Physics, 1992, 72[5], 2049-53

[446-91/92-033]

Table 69Diffusivity of P in Si

T (C) D (cm2/s)1227 1.45 x 10-12

1176 4.02 x 10-13

1123 1.09 x 10-13

1079 4.49 x 10-14

987 6.40 x 10-15

911 8.58 x 10-16

767 8.17 x 10-17

Bulk Diffusion - Qualitative Observations - Concentration ProfilesA secondary ion mass spectrometric analysis of diffusion across a TaSi2/Si interface andinto monocrystalline or polycrystalline Si was carried out, together with an electricalcharacterization of the resultant structures. At temperatures ranging from 900 to 1000C,the dopant readily diffused into Si, without drastic segregation effects, when appropriateinterface cleaning was used. In particular, very shallow diffusion regions were obtained inmonocrystalline material beneath the implanted TaSi2; even at the relatively longannealing times sometimes needed for processing.H.Gierisch, F.Neppl, E.Frenzel, P.Eichinger, K.Hieber: Journal of Vacuum Science andTechnology B, 1987, 5[2], 508-14

[446-55/56-034]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe co-diffusion of As and P was studied by using a buried B layer. An analysis of thedopant profiles was performed by using secondary ion mass spectroscopy, spreadingresistance profiles, and junction staining. It was found that the profiles of As and P, whendiffused separately, were enhanced with respect to the co-diffused profiles. The buried Bprofile in the case of P alone was enhanced with respect to the B profile in the case of Asand P co-diffusion. The annealing of residual implantation damage was suggested to beresponsible for these effects. In addition, the suppression of the self-interstitialsupersaturation under co-diffusive conditions was a transient effect and was suggested to

437

P Bulk Diffusion P

be caused by an increase in the recombination of implantation-generated vacancies andself-interstitials due to the presence of a high concentration of As atoms.R.Deaton, U.Gösele, P.Smith: Journal of Applied Physics, 1990, 67[4], 1793-800

[446-74-046]

Table 70Diffusivity of P in Si

T (C) D (cm2/s)1000 3.9 x 10-14

950 1.1 x 10-14

900 3.0 x 10-15

Bulk Diffusion - Qualitative Observations - Concentration ProfilesDepth profiles were determined by using spreading resistance, Auger electronmicroscopy, and secondary ion mass spectrometry methods. These profiles showed thatenhanced dopant diffusion and surface depletion resulted from rapid thermal treatments.It was found that Si interstitial injection from a N-supersaturated oxynitride interfacefacilitated the diffusion of P in the Si substrate. An appreciable amount of N was foundbelow the Si surface, and this suggested that N interstitials might play an important role inthe observed enhancement of the impurity diffusion.J.Bustillo, C.Chang, S.Haddad, A.Wang: Applied Physics Letters, 1991, 58[17], 1872-4

[446-86/87-049]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesMigration in amorphous and polycrystalline Si-on-monocrystal systems during rapidthermal annealing and furnace annealing was studied. It was found that changes inmicrostructure during annealing played a major role in determining diffusion profiles inthe substrate, as well as in the polycrystalline Si layer. In the case of P doping, drive-indiffusion resulted in a much larger grain microstructure in as-deposited amorphous Sithan in as-deposited polycrystalline Si. This led to the formation of shallower junctions inthe first case. At high annealing temperatures, the native interfacial oxide disintegrated,caused epitaxial realignment of the polycrystalline Si film, and led to enhanced diffusionin the substrate.K.Park, S.Batra, S.Banerjee, G.Lux, T.C.Smith: Journal of Applied Physics, 1991, 70[3],1397-404

[446-91/92-028]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe interdependence of the diffusion behavior and grain microstructure in amorphouspolysilicon - on - monocrystal systems was studied with regard to the rapid thermal andfurnace annealing of P implants. It was found that changes in microstructure duringannealing played a major role in determining the diffusion profiles in the substrate as wellas in the polysilicon layer. Drive-in diffusion resulted in a much larger grain

438

P Bulk Diffusion P

microstructure for as-deposited amorphous Si than for as-deposited polysilicon. This ledto the formation of shallower junctions in the substrate for the first case. The P junctionswhich formed in the substrate were found to be very uniform laterally, in spite ofexpected doping inhomogeneities that were caused by polysilicon grain boundaries, forboth as-deposited amorphous Si diffusion sources and for as-deposited polysilicondiffusion sources.K.Park, S.Batra, S.Banerjee, G.Lux: Journal of Electronic Materials, 1991, 20[3], 261-5

[446-81/82-047]

Table 71Intrinsic Diffusivity of P in Si

T (C) D (cm2/s)1150 3.3 x 10-13

1000 9.7 x 10-15

900 5.8 x 10-16

Bulk Diffusion - Qualitative Observations - Concentration ProfilesA technique was proposed for the simultaneous diffusion of B and P on opposite sides ofa wafer. The solid dopant sources which were used in this method could be tailored so asto produce a wide range of P diffusion profiles, for a given thermal diffusion cycle. Theuniformity of the sheet resistance across a 49cm2 area was often greater than 95%. Aunique feature was that the resultant diffusion glass was extremely thin.T.Krygowski, A.Rohatgi: Journal of the Electrochemical Society, 1997, 144[1], 346-52

[446-152-0438]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe migration of P in spike-doped chemical vapor deposited layers, and segregation atpoly-Si/Si-substrate interfaces, were studied by means of secondary-ion massspectroscopic and spreading resistance measurements. The time dependence of the Psegregation indicated a diffusion-limited behavior; apart from a short initial period. Theeffect of a thin interfacial oxide upon the segregation behavior was studied by analyzingthe binding energy of P atoms; segregated at the interface. This energy was compatiblewith values for elemental P; thus indicating that no chemical interaction with residualinterfacial oxides took place. It was found that there were homogeneous charge carrierprofiles across the poly-Si layer, shallow junctions (less than 50nm within the Sisubstrate) and steep concentration profiles with a slope of 8nm/decade at lowtemperatures.D.Krüger, P.Gaworzewski, R.Kurps, J.Schlote: Semiconductor Science and Technology,1995, 10[3], 326-31

[446-121/122-089]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesShallower carrier concentration profiles were observed in 50keV P+-implanted Si(100)after annealing (1000C, 1h) if a buried amorphous layer was formed by further irradiation

439

P Bulk Diffusion P

with 1MeV Si+ ions before annealing. The secondary defects which were formed in theMeV Si+-damaged region acted as gettering sites for the collection of the interstitials,from shallower depths, which were responsible for the transient diffusion of P. Therefore,transient P diffusion was reduced and the carrier concentration profiles becomeshallower.Q.Zhao, Z.Wang, T.Xu, P.Zhu, J.Zhou: Applied Physics Letters, 1993, 62[24], 3183-5

[446-106/107-145]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe P doping of polycrystalline Si by the direct diffusion of PH3 at 685 and 785C,followed by rapid thermal annealing, was investigated. Contrary to the behavior at 785C,the dopant profile was not uniform at 685C, but became so after rapid thermal annealing(1050C, 10s); as revealed by secondary ion mass spectroscopic data. In the absence ofany protective layer, rapid thermal annealing in N caused the out-diffusion of about 50%of the P from a sample which was diffused at 785C; but it activated almost 100% of theremaining P atoms (as indicated by Hall and secondary ion mass spectrometrymeasurements). On the other hand, rapid thermal annealing in O resulted in 95%activation of the incorporated P from the same sample, together with the oxidation ofabout 7nm of polycrystalline Si. In the case of samples diffused at 685C, rapid thermalannealing in N did not cause any out-diffusion, but it activated less than 80% of theincorporated P atoms. This difference in behavior was attributed to the tendency toproduce equilibrium concentrations of P in the crystallites, and at the grain boundaries, atthe rapid thermal annealing temperature. The highest active P concentration which wasachieved was 2.0 x 10

20/cm

3 at 785C, followed by rapid thermal annealing in O to yield a

resistivity which could be as low as 0.0009Ω cm.S.Lourdudoss, S.L.Zhang: Applied Physics Letters, 1994, 64[25], 3461-3

[446-115/116-154]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesAuger electron spectroscopic depth profiles of the dopant in the surface layers of a waferwere determined down to depths which were not usually achieved by using normal ionetching techniques. It was found that there were substantial deviations of theexperimentally determined curve from the expected complementary error function. Theshape of the diffusion profile, and the presence of an inflection, could be satisfactorilyexplained by an acceleration of the diffusion by an electric field and by a change in thediffusion coefficient during the dissociation of (PV)- complexes.B.M.Kostishko, A.M.Orlov, T.G.Levkina: Neorganicheskie Materialy, 1994, 30[6], 853-5(Inorganic Materials, 1994, 30[6], 794-5)

[446-117/118-197]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe incorporation of P from a spun-on dopant layer at 400C was considered. Annealingexperiments were performed on dopant films which had been deposited onto (100)substrates by using the spin-on technique. Conventional heating, in a normal atmosphere attemperatures of up to 400C, was used to study dopant incorporation. After removing thespun-on dopant film, part of the Si substrate was annealed at higher temperatures. The Pconcentration profiles in low-temperature annealed samples revealed diffusion depths of

440

P Bulk Diffusion P

60 to 80nm (extrapolated to a substrate dopant level of 1016/cm3), and surfaceconcentrations of between 1010 and 1020/cm2.U.Mohr, R.Leihkauf, K.Jacob: Applied Physics A, 1997, 64[1], 77-81

[446-150/151-153]

Bulk Diffusion - Qualitative Observations - Effect of AmbientThe effect of the annealing ambient upon dopant diffusion was studied, during low-temperature processing by implanting P into (100) samples (at room temperature) througha 14nm SiO2 layer. The implantation dose and energy were sufficient to amorphize thesurface. After low-temperature furnace annealing, the ion-implanted P exhibited atransient enhanced diffusion regime for both inert and oxidizing ambients. It had beenexpected that point defect generation during the annealing of implantation damage wouldpredominate during the transient enhanced diffusion process; regardless of the nature ofthe ambient. Deeper P junctions were found, for implants which had been annealed in anoxidizing ambient, when surface oxidation had consumed more than 5nm of the Si. It wassuggested that stress in the surface layer played an important role in the diffusion of high-dose P implants. Oxidation tended to consume this highly stressed surface layer, whichcould suppress P diffusion in the tail region.Y.Kim, H.Z.Massoud, R.B.Fair: Journal of the Electrochemical Society, 1990, 137[8],2599-603

[446-76/77-033]

Bulk Diffusion - Qualitative Observations - Effect of AmbientA solid planar diffusion source which contained pyrophosphate (SiP2O7) was evaluatedfor P doping at reduced pressures. A reduction in pressure generally resulted inimprovements in doping uniformity. At the edges of 3in wafers which were diffused at1000C, the maximum deviations in sheet resistance were reduced from about 15% toabout 5% when the total pressure was reduced from 1atm to 5torr. This improvement waseven more significant for diffusion at 900C, where the maximum deviation was reducedfrom about 100% to about 10% by the same reduction in total pressure. The use of aclosed wafer boat configuration, combined with a reduced pressure, resulted in furtherimprovements in doping uniformity; as reflected by a deviation in sheet resistance ofabout 1% for diffusion at 900C. The use of low pressures led to a decreased dopingperformance of the source, as reflected by thinner doped oxide formation and higher sheetresistance values for diffused Si wafers. The rate of evaporation of P2O5 from the sourceincreased by almost an order of magnitude when the pressure was reduced from 1atm to0.5torr, thus leading to a reduced lifetime of the source.J.R.Flemish, R.E.Tressler, J.Ruzyllo: Journal of the Electrochemical Society, 1991,138[1], 233-8

[446-78/79-057]

Bulk Diffusion - Qualitative Observations - Effect of DefectsA comprehensive investigation of dopant diffusion in the presence of equilibrium and non-equilibrium concentrations of intrinsic point defects was presented. It was found that,under equilibrium conditions, vacancies and interstitials mediated the diffusion of thedopant. Relevant expressions were derived for the activation energies, of various diffusion

441

P Bulk Diffusion P

and injection mechanisms, under non-equilibrium conditions (such as those produced bythe injection of excess point defects). For oxidation conditions, the calculated values werein excellent agreement with available experimental data. Both theory and experimentsuggested that the concerted exchange mechanism, which involved no point defects,played only a minor role in dopant diffusion.C.S.Nichols, C.G.Van de Walle, S.T.Pantelides: Physical Review B, 1989, 40[8], 5484-96

[446-72/73-041]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe co-diffusion of implanted As and P was investigated, after annealing at 900 or1000C, for various dopant concentrations. The results did not reveal any direct interactionbetween the dopants. All of the observed anomalous effects of co-diffusion, as comparedwith diffusion of the elements by themselves, seemed to be explained by interactionsbetween the dopants and the defects which were produced by ion implantation. It wasobserved that the presence of a high concentration of As atoms made annealing of theimplantation damage more rapid, and strongly reduced P transient-enhanced diffusion.S.Solmi, P.Maccagnani, R.Canteri: Journal of Applied Physics, 1993, 74[8], 5005-12

[446-106/107-131]

Bulk Diffusion - Qualitative Observations - Effect of DefectsA study of Au diffusion was used to clarify the effect of P diffusion upon the defectstructure of 2-step annealed wafers of Czochralski material which contained O-relatedprecipitates. By means of deep level transient spectroscopy, and the use of publishedrelationships, it was established that the substitutional Au concentration was independentof O precipitation. On the other hand, P diffusion (via the injection of self-interstitialsinto the bulk) shrank the O precipitates and increased the substitutional Au concentrationto near to the solubility limit.E.Yakimov, I.Périchaud: Applied Physics Letters, 1995, 67[14], 2054-6

[446-125/126-145]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe non-equilibrium impurity diffusion of dopants in monocrystalline Si was carried outby means of the controlled surface injection of self-interstitials and vacancies. By varyingthe parameters of the surface oxide layer during the P diffusion process, it was possible toproduce quantum-sized profiles and p-n junctions with dimensions that could becontrolled within the 1 to 22nm range.N.T.Bagraev, L.E.Klyachkin, V.L.Sukhanov: Semiconductor Science and Technology,1991, 6[7], 577-81

[446-88/89-051]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe migration of P in heavily doped layers (extrinsic at the process temperature) wasstudied in order to obtain information on Si interstitial concentrations in heavily dopedlayers. The diffusion of the dopant in n-type and p-type iso-concentration structures wasmeasured for 2 surfaces. One surface was passivated with low-pressure chemical vapordeposited oxides and nitrides, and the other was a growing thermal oxide. The degree of

442

P Bulk Diffusion P

oxidation-enhanced diffusion was measured for both dopants and was found to be smallerthan for the same annealing treatments under lightly doped conditions (intrinsic at processtemperatures). The P data were analyzed, and effective Si interstitial energy levels formodelling the diffusion of heavily doped layers were obtained. The diffusion of P wasalso studied in iso-concentration structures, with and without a lightly doped epitaxiallayer at the surface, in order to determine the effect of doping at the Si/SiO2 interfaceupon diffusivity enhancement. It was found that samples with epitaxial layers exhibitedgreater enhancements as compared with samples that did not. The results were explainedin terms of charged interstitial drift in the electric field which was set up by ionizeddopant atoms, and a resultant increase in the flux of interstitials away from the oxidizinginterface.D.J.Roth, J.D.Plummer: Journal of the Electrochemical Society, 1994, 141[4], 1074-81

[446-119/120-223]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe effect of precipitation upon P diffusion was investigated at temperatures of 850 or1000C in heavily implanted (4 x 1016P+/cm2) and laser-annealed specimens. By varyingthe amount of precipitated P, while keeping the dopant concentration constant, it wasshown that precipitation did not generate interstitials in sufficient amounts to affect Pdiffusion.R.Angelucci, S.Solmi, Q.Zini: Physica Status Solidi A, 1986, 96[2], 541-6

[446-55/56-041]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe effect of point defects upon dopant diffusion was considered. Their influence wasanalyzed by comparing experimental data with a simulation. The experiments concernedthe oxidation-enhanced diffusion of P in float-zone material at 1100C. The simulationwas carried out by using a physical model without simplifying assumptions, and it wasshown that such assumptions were not admissible. A reasonable set of parameters wasdeduced from the analysis. Since each parameter represented a physical effect,information about the importance of the bulk and surface recombination of point defects,and about the equilibrium concentration values and diffusion coefficients of diffusingspecies, could be obtained. It was found that the effect of the surface played a decisiverole in the distribution of point defects.T.Brabec, E.Guerrero, M.Budil, H.W.Poetzl: Zeitschrift für Physik B, 1987, 67[4], 415-20

[446-55/56-043]

Bulk Diffusion - Qualitative Observations - Effect of DefectsSamples, with or without interstitial dislocation loops, were implanted with P to doseswhich were below the amorphization threshold. The loops were produced by Si+

implantation to high doses, followed by annealing (900C, 0.5h). Triple-crystal X-raydiffraction, junction depth measurements, and secondary ion mass spectrometry wereused to study implantation defects and P distributions. Annealing was carried out in afurnace at temperatures of between 600 and 900C, or by electron-beam treatment (1000C,10s). It was found that the presence of loops markedly reduced anomalous P diffusion.

443

P Bulk Diffusion P

This was attributed to the effect of absorption, by the loops, of excess interstitials arisingfrom dissolution of the clusters which were produced by P implants.M.Servidori, S.Solmi, P.Zaumseil, U.Winter, M.Anderle: Journal of Applied Physics,1989, 65[1], 98-104

[446-72/73-048]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe behavior of Co was studied with respect to P doping via in-diffusion or via SiO2

growth. It was found that both techniques led to moderate gettering. However,cooperation of the 2 processes led to strong gettering of Co. The gettering behavior wastentatively attributed to a coupling of local currents of Si self-interstitials, which weregenerated during P diffusion gettering, and of 3d impurities on substitutional sites.W.Schröter, R.Kühnapfel: Applied Physics Letters, 1990, 56[22], 2207-9

[446-76/77-040]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe external gettering effect, due to P, was used to improve the electrical properties ofpolycrystalline wafers. After annealing (900C, 4h), it was found that the effectivediffusion lengths of minority carriers attained or exceeded the thickness of the wafers.After annealing at 850C for 4h, the improvements were less marked, and hydrogenationwas required in order to obtain the same increase in effective diffusion length. Secondaryion mass spectrometric analyses indicated that the gettered impurities were mainly Fe,Cu, and Ni. Some limited regions of the wafers were only slightly improved. It wasfound, after chemical etching, that these regions contained a high density of sub-grain.boundaries. A mechanism for the gettering effect was proposed which took account ofdissolved impurities in the grains and of impurities which segregated to dislocations. Itwas suggested that the additivity of the hydrogenation effect might be due to theneutralization of recombination centers which were related to O atoms that weresegregated at dislocations.I.Perichaud, S.Martinuzzi: Journal de Physique III, 1992, 2[3], 313-24

[446-099/100-099]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe diffusivity of P in iso-concentration backgrounds under inert conditions wasinvestigated. Low doses of P were implanted into wafers that were doped with As and B.These samples were annealed to remove any damage, and secondary ion massspectroscopy measurements were made of the as-implanted samples. The wafers werethen annealed at 900 and 1100C for 1h and 0.5h, respectively. The wafer profiles weremeasured by using secondary ion mass spectroscopy. The P diffusivity was measured for5 different dopant concentrations. The results showed that a double-negative interstitialcomponent of P was needed in order to model the data adequately. Also, heavily B-dopedsamples exhibited significantly retarded diffusivity at 1100C. It was suggested that thisreflected P-B pairing or significant strain effects upon diffusion.J.P.John, M.E.Law: Applied Physics Letters, 1993, 62[12], 1388-9. See also: Journal ofthe Electrochemical Society, 1993, 140[5], 1489-91

[446-106/107-144]

444

P Bulk Diffusion P

Bulk Diffusion - Qualitative Observations - Effect of DefectsExternal gettering by P diffusion was applied to interstitial O rich wafers which containedextended crystallographic defects such as dislocations, twins and grain boundaries. It wasfound that P diffusion sharply increased the minority carrier diffusion length when the Ocontent was lower than 6 x 10

17/cm

3, but was less effective (or impaired the diffusion

length) when the O content was of the order of 1018/cm3. In the latter case, light beaminduced current mappings showed that, when extended crystallographic defects werepresent, the carrier length increased around the defects and decreased in homogeneousregions. It was suggested that the O precipitates which formed in O-rich samples inducedan internal gettering which contributed to the cleaning of extended defects (in the samemanner as P diffusion) but degraded the carrier length in the grains.S.Martinuzzi, I.Perichaud: Materials Science Forum, 1994, 143-147, 1629-34

[446-113/114-048]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe diffusion of ion-implanted P in the presence of various concentrations of radiationdefects, during rapid thermal annealing and furnace annealing, was investigated by usingelectrical methods. The diffusion of P in layers which had been previously doped with Gewas also studied. It was established that part of the inserted P was captured, during heattreatment, by surplus vacancies. These traps governed the displacement of the Pdistribution towards the surface. The remainder of the impurity, in the form of P-Sicomplexes, accelerated P diffusion. The temperature independence, over a widetemperature range, of the effective coefficient for the diffusion of implanted P duringrapid thermal annealing was attributed to the closeness of the activation energy fordiffusion of the P-Si complexes and of the activation energy which was required to annealit.A.R.Chelyadinskii, H.I.H.Taher: Physica Status Solidi A, 1994, 142[2], 331-8

[446-117/118-197]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe effects of implantation damage upon P diffusion during rapid thermal annealing andconventional heat treatment at 750 to 900C were investigated. A comparison of thedopant profiles and junction depths in damaged and undamaged regions demonstrated thatSi+ implantation under the typical conditions used for pre-amorphization produced amarkedly enhanced diffusion of P atoms.A.Jacques, A.George, X.Baillin, J.J.Bacmann: Philosophical Magazine A, 1987, 55[2],165-81

[446-51/52-130]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationIon implantation produces point defects, during annealing, which can significantlyenhance dopant diffusion. This effect was studied for implant diffusion at lowtemperatures. Enhanced diffusion was detected below critical dopant concentrations. Thelatter concentrations depended only upon the temperature. Enhanced diffusion occurredbelow the well-known kink concentration.R.B.Fair: Journal of the Electrochemical Society, 1990, 137[2], 667-71

[446-74-046]

445

P Bulk Diffusion P

Bulk Diffusion - Qualitative Observations - Effect of OxidationThe diffusion of P in a Si-on-insulator structure, formed by heavy doses of O-implantation, was compared with diffusion in normal bulk wafers. It was found that thenon-equilibrium effects upon diffusion during direct nitridation or nitridation of a SiO2

layer were mitigated in the present structures. Under oxidizing conditions, there were nolarge differences in the diffusion behavior of bulk and Si-on-insulator samples. It was alsodemonstrated that the spatial extent of the low-concentration tail region of P diffusionfrom POCl3 was reduced in the Si-on-insulator structure.P.Fahey, S.Solmi: Journal of Applied Physics, 1986, 60[12], 4329-32

[446-51/52-136]

Bulk Diffusion - Qualitative Observations - Effect of PressureAn investigation was made of P diffusion out of an ion-implanted layer at 1000C underAr pressures of up to 2kbar. The implanted P dose was 0.003C/cm2. An increase in thedepth of P atom penetration, compared with annealing in the absence of pressure, wasobserved. The effect depended non-linearly upon the duration of the pressure treatment,and was essentially absent during the initial stages of annealing. A considerable increasein the annealing time also reduced the acceleration. The observed effect was explained bya 2-stream diffusion model which assumed that the pressure increased the concentrationof interstitial P atoms in the diffusion zone by altering the defect structure of an implantedlayer.A.S.Vasin, V.I.Okulich, V.A.Panteleev: Fizika i Tekhnika Poluprovodnikov, 1989, 23[3],483-7 (Soviet Physics - Semiconductors, 1989, 23[3], 302-4)

[446-72/73-047]

Bulk Diffusion - Qualitative Observations - Effect of Proton IrradiationSecondary ion mass spectrometry was used to study the diffusion of P which had beenimplanted into the Si, and stimulated by proton irradiation. The latter irradiation wascarried out at 700C by using an energy of between 50 and 100keV and doses of between1016 and 1017/cm2. A maximum which was observed in the impurity distribution, afterradiation-stimulated diffusion and at depths beyond the range of the implanted ions, wasattributed to the accumulation of impurities at radiation defects of interstitial type whichhad been created during the implantation.V.V.Kozlovskii, V.N.Lomasov: Fizika i Tekhnika Poluprovodnikov, 1987, 21[2], 360-2.(Soviet Physics - Semiconductors, 1987, 21[2], 220)

[446-55/56-042]

Bulk Diffusion - Qualitative Observations - Effect of Surface LayerThe effects of O precipitation and surface films upon P diffusion in Czochralski materialat 1100C were studied. In the case of a high precipitation rate, P diffusion under bothtypes of film was enhanced because of a supersaturation of Si interstitials that was causedby O precipitation. A greater enhancement of P diffusion under Si3N4, than under SiO2

covered with Si3N4 was attributed to the lower recombination velocity of interstitials atthe Si3N4/Si interface. The diffusion within a denuded zone resembled that in float-zoneSi until interstitials, which were generated under that zone, arrived at the interface.

446

P Bulk Diffusion P

S.T.Ahn, H.W.Kennel, J.D.Plummer, W.A.Tiller, Z.U.Rek, S.R.Stock: Applied PhysicsLetters, 1988, 53[1], 34-6

[446-62/63-234]

Bulk Diffusion - Qualitative Observations - Effect of Surface LayerThe effect of selective masking, with WO3 electron resist, upon P diffusion and oxidationat 1000C was described. Although the diffusion coefficient of P in the WO3 layer wasabout twice as large as that in a SiO2 layer, the WO3 resist was found to be useful formasking P diffusion at temperatures of about 1000C. On the other hand, the maskingeffect of the WO3 layer upon the oxidation of a Si substrate was insufficient since thediffusion coefficient of O2 molecules in the WO3 layer was about 8 times as large as thatin the SiO2 layer.M.Baba, S.Abe, T.Ikeda: Japanese Journal of Applied Physics, 1987, 26[9], 1561-4

[446-55/56-041]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsScanning infra-red microscopy, mapping, Fourier transform infra-red spectroscopy, andminority carrier diffusion length measurements were combined in order to studyprecipitates in annealed wafers of Czochralski material and to evaluate theirrecombination strength; with or without metallic contamination. The effect of P diffusionwas also investigated. After 2-step annealing (750C, 16h; 900C, 24 or 96h), scanninginfra-red microscopy revealed the presence of precipitates, while light beam-inducedcurrent mapping indicated a ring-like distribution of recombination centers and minoritycarrier diffusion length collapses to 2µ. The presence of Cu contamination did notsignificantly modify these observations when precipitates formed. The occurrence of Pdiffusion near to the surface shrank the precipitates which were revealed by MIRB, butdid not suppress the ring-like patterns which were seen in light beam-induced currentmaps, and slightly increased the carrier diffusion length. The results suggested that therecombination strength of precipitates did not depend upon metallic decoration, but wasmore likely to depend upon interfacial states between the precipitates and the host crystal,or upon associated extended defects.C.Veve, N.Gay, M.Stemmer, S.Martinuzzi: Journal de Physique III, 1995, 5[9], 1353-63

[446-125/126-150]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsThe effect of lattice defects, produced by Si ion implantation, upon dopant diffusivity wasinvestigated after annealing specimens at temperatures of between 700 and 900C. Thenature and depth of residual implantation defects in undoped samples was determined byanalysing the rocking curves which were obtained by using triple-crystal X-ray diffractionand transmission electron microscopy. As well as interstitial dislocation loops andclusters lying below the original amorphous/crystal interface, epitaxial re-growth of theamorphized Si left a vacancy-rich surface layer and a deeper region which was enrichedin interstitials. These regions corresponded to those for which Monte Carlo simulations ofdefect production predicted an excess of point defects. According to whether the dopantwas associated with vacancy or interstitial clusters, different anomalous diffusionbehaviors were observed. In the deep region where an excess of interstitials was present,

447

P Bulk Diffusion P

P underwent markedly enhanced diffusion. This trend was consistent with the differingcontributions which vacancies and interstitials made to the diffusion mechanisms of thedopants.M.Servidori, R.Angelucci, F.Cembali, P.Negrini, S.Solmi, P.Zaumseil, U.Winter: Journalof Applied Physics, 1987, 61[5], 1834-40

[446-60-013]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsBy using deep-level transient spectroscopy, it was found that the substitutional Au atomconcentration in precipitate-containing O-rich monocrystals was about an order ofmagnitude lower than that in O-poor material (under given diffusion conditions). Thediffusion of P near to the surfaces increased this concentration, by 2 orders of magnitude,up to the solubility limit. This was suggested to be due to the injection of self-interstitialsinto the bulk, and to the gettering of contaminants; thus increasing their efficiency assinks for self interstitials, and transforming interstitial Au atoms into substitutional ones.E.Yakimov, I.Périchaud, S.Martinuzzi: Solid State Phenomena, 1996, 47-48, 275-80

[446-134/135-163]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsThe contribution, to gettering, of interstitial injection that was induced by P diffusionunder non-oxidizing conditions was considered with regard to short-range (doped-layer)effects and long-range (bulk) effects. It was shown that a positive concentration gradientof Si interstitials within a highly doped layer was favorable to the gettering ofsubstitutional metallic species via a kick-out mechanism during P diffusion. Also, thecontribution of P diffusion to the gettering of 3d elements was significant only within ahighly P-doped layer (short-range effect). The predictions of the diffusion-inducedgettering model agreed with experimental evidence on the P diffusion-gettering of Co andon the P diffusion-gettering of Au.F.Gaiseanu, W.Schröter: Journal of the Electrochemical Society, 1996, 143[1], 361-2

[446-134/135-160]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsThe use of C or O as a diffusion-suppressing agent for P was suggested. In order to studythis complex phenomenon, an experimental study was made of the effects of low-dose Si,C and O implantation damage upon the diffusion of lightly-doped P layers. The effects ofSi and C co-implantation upon the diffusion of P were also studied. Finally, a lightlydoped drain structure was annealed in the presence of implanted C. It was found that Cwas the most effective diffusion-suppressing agent among the present 3 species. Theresults of the second experiment suggested that C strongly affected the interstitial profileand thus the final P profile.S.Chaudhry, M.E.Law: Journal of the Electrochemical Society, 1994, 141[12], 3516-21

[446-119/120-225]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsDefect generation during P diffusion at concentration levels below the solid solubility wasinvestigated using ion-implanted samples. Diffusion anneals were carried out in low-O

448

P Bulk Diffusion P

ambients or in N with a Si3N4 cap. Buried layer markers of As and Sb were used to studypoint defect generation due to the diffusion. The defects were revealed using a Schimmeletch. The results showed that stacking faults were formed below the P-diffused region ina low-O ambient at temperatures ranging from 1050 to 1150C. Enhancement of Asdiffusion and retardation of Sb diffusion were observed at long diffusion times. Diffusionin N with a Si3N4 cap caused few stacking faults and neither enhancement of As diffusionnor retardation of Sb diffusion for the same temperatures and times. At P concentrationsabove the solid solubility diffusion in N, with a Si3N4 film cap caused an enhancement ofAs diffusion. Stacking faults were also observed.J.C.C.Tsai, D.G.Schimmel, R.E.Ahrens, R.B.Fair: Journal of the Electrochemical Society,1987, 134[9], 2348-56

[446-55/56-042]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionDiffusion experiments were performed in the presence of lattice defects which had beenproduced by Si ion implantation. The effects of transient enhanced diffusion were studiedby means of bevelling and staining measurements of implanted samples, and secondaryion mass spectroscopic determinations of dopant profiles. The annealing temperatures fordoped implanted specimens ranged from 700 to 1100C and this treatment was performedusing an electron beam. The low temperatures which were used permitted the kinetics ofanomalous diffusion to be monitored. It was ascertained that the enhanced diffusioncoefficient was almost constant during a period which decreased with increasingtemperature. It then tended gradually to the equilibrium value. This trend agreed with thatof lattice damage changes which were revealed by double crystal X-ray analyses of therocking curves of implanted samples. The secondary ion mass spectroscopy profilesindicated that only a fraction of the dopant which was located at residual implantationdamage was responsible for the anomalous diffusion.R.Angelucci, F.Cembali, P.Negrini, M.Servidori, S.Solmi: Journal of the ElectrochemicalSociety, 1987, 134[12], 3130-4

[446-60-013]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe behavior of the dopant during infra-red heating was investigated by using spin-ondiffusion sources, repeated etching, and capacitance-voltage measurements. The diffusioncoefficient was calculated by fitting the concentration profiles to the complementary errorfunction. It was found that the diffusivity was enhanced during infra-red heating. This didnot appear to be due to ion-implantation damage or rapid heating. The enhanceddiffusivity was attributed to the generation of an excess of self-interstitials.Y.Ishikawa, K.Yamauchi, I.Nakamichi: Japanese Journal of Applied Physics, 1989,28[8], L1319-21

[446-72/73-041]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe effects of low-dose Si implantation damage upon the diffusion of low concentrationsof P in wafer samples were investigated. The dopant was implanted to low doses, and wasthen pre-annealed in order to remove any self-damage. Enhanced P diffusion was

449

P Bulk Diffusion P

observed by directly comparing the dopant profiles in damaged and undamaged regions.Monitoring of the effective P diffusivity at various annealing temperatures and timesrevealed that the enhanced diffusion was a transient process with a time constant that waslarger at lower temperatures. This enhancement was larger and of longer duration, thelower the annealing temperature. It was noted that As diffusion, in such damaged regions,did not exhibit any enhancement; thus implying that the defects which were induced by Siimplantation had differing interaction mechanisms for each type of dopant.H.Park, M.E.Law: Applied Physics Letters, 1991, 58[7], 732-4

[446-81/82-044]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe migration of P during surface oxidation in dry O was measured at 900 to 1100C, fortimes of between 600s and 4h. During 0.5h oxidation at 900C, the P diffusivity was 15times greater than it would have been in an inert ambient. The enhancement of P diffusiondecreased with increasing temperature. An increase in the oxidation time at a giventemperature led to a decrease in the diffusion enhancement.P.A.Packan, J.D.Plummer: Journal of Applied Physics, 1990, 68[8], 4327-9

[446-86/87-051]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe diffusion behavior of P was used to study point defect kinetics in Si-on-insulatorsamples. Marker layers were used to study oxidation-enhanced diffusion in bulk andbonded and etched-back samples under oxidizing conditions at 750, 800 and 850C. Aneffective interstitial recombination velocity for the buried Si/SiO2 interface in the etched-back material was extracted by comparing the experimentally obtained profiles withsimulation results. The data could be modelled by assuming a time-independent interfacerecombination velocity. The same parameter set, but incorporating this deducedrecombination velocity, accurately reproduced the implant-enhanced diffusion of Bmarker layers at 750 and 800C in thin films. This implied that the recombination velocitywas independent of the interstitial supersaturation.S.W.Crowder, C.J.Hsieh, P.B.Griffin, J.D.Plummer: Journal of Applied Physics, 1994,76[5], 2756-64

[446-117/118-216]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionPhosphorus junctions, having depths ranging from 80nm to 0.005mm, were produced inwafers by means of thermal pre-deposition and drive-in heating. The wafers were thenimplanted with Si ions in order to create lattice damage both below and above theamorphization threshold. Subsequent heat-treatment led to recovery of the damage whichwas produced by implantation. The resultant excess interstitials favored enhanced dopantdiffusion. Anomalous diffusivity was associated with non-amorphizing implants,regardless of the depth of the junction with respect to that of the damaged region. On theother hand, negligible or strong anomalous diffusivity was found after amorphizingimplants, depending upon whether the initial P junction was shallower or deeper than theamorphous/crystalline interface. The asymmetric behavior of the P diffusivity, across theinterface between amorphous and crystalline Si, was confirmed by using secondary ion

450

P Bulk Diffusion P

mass spectrometry to profile an ion-implanted uniformly doped wafer. The enhanceddiffusion was closely related to changes in the implant damage during furnace heating.P.Negrini, M.Servidori, S.Solmi: Philosophical Magazine A, 1990, 61[4], 553-61

[446-76/77-042]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionA study was made of the transient enhanced diffusion which resulted from implantationdamage, and which led to more rapid dopant migration. Experiments were performed hereby using P and dislocation markers to compare transient enhanced diffusion effects. Theresults showed that P diffusion was enhanced considerably more than B diffusion duringdamage annealing. Dislocation growth indicated that a number of interstitials that wasgreater than the damage dose was captured during annealing. It was found that the timethat was required in order to saturate dislocation growth agreed well with P diffusionsaturation, and was greater than the B saturation.J.Xu, V.Krishnamoorthy, K.S.Jones, M.E.Law: Journal of Applied Physics, 1997, 81[1],107-11

[446-141/142-114]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe transient-enhanced diffusion of P was investigated at doses and energies which werebelow the threshold for amorphization of the substrate. It was found that transient-enhanced diffusion occurred during both furnace annealing and rapid thermal annealing,and could increase the junction depth by 100 to 200nm, as compared to that predicted bystandard diffusion models. The magnitude of the transient was shown to depend upon theimplantation dose and energy, the annealing time and temperature, and the backgrounddoping level. A simple model for damage annealing was proposed which quantitativelydescribed the point defect enhancement of dopant diffusion and thus permitted each ofthese dependences to be understood. These results also provided a new means forcharacterizing the diffusion and charge states of the point defects themselves.M.D.Giles: Journal of the Electrochemical Society, 1991, 138[4], 1160-5

[446-81/82-052]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe migration of P was studied in [100]-type material at 1150C, under both oxidizing andnon-oxidizing conditions, in ambients which contained dry O that was diluted with N. Itwas found that enhanced diffusion occurred even at very low rates, as compared with thatin [111]-type material.S.T.Dunham, N.Jeng: Applied Physics Letters, 1991, 59[16], 2016-8

[446-84/85-073]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionAn investigation was made of the transient enhanced diffusion of P followingimplantation with Si or Ar ions to low doses. Both led to the up-hill diffusion of P, due todefect gradients, but the resultant profiles were quite different because of differences inthe initial defect distributions. The results supported the assumption of an interstitial pair

451

P Bulk Diffusion P

diffusion mechanism for P, and illustrated the importance of bulk recombination indetermining defect distributions for Ar damage annealing.M.D.Giles: Applied Physics Letters, 1993, 62[16], 1940-2

[446-099/100-098]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionTransient enhanced P diffusion, following implantation with Si or Ar ions to low doses,was investigated. Both treatments led to up-hill P diffusion, due to defect gradients, butthe resultant profiles were quite different because of differences in the initial defectdistributions. It was concluded that these experiments supported an interstitial pairdiffusion mechanism for P, and revealed the importance of bulk recombination indetermining defect distributions for Ar damage annealing.M.D.Giles: Applied Physics Letters, 1993, 62[16], 1940-2

[446-106/107-146]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe enhanced diffusion of P buried layers, as a function of depth in Si substrates, wasmeasured during interstitial injection by thermal oxidation of the wafer surface. Theresults were used to calculate the effective interstitial diffusivity in epitaxial Si at 1100C.The experimental results were explained by using a model which incorporated the bulkrecombination of interstitials with vacancies, as well as a rapid interstitial diffusivity. Themodel effectively accounted for the experimental data and, by comparing the data withmodel simulations, ratios of the equilibrium concentrations of vacancies and interstitialswere obtained.A.M.Agarwal, S.T.Dunham: Applied Physics Letters, 1993, 63[6], 800-2

[446-106/107-147]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe out-diffusion of P during the glow discharge deposition of amorphous hydrogenatedmaterial was studied. Secondary ion mass spectrometric measurements, of P profiles inn+/i/SiO2 and i/n+/SiO2 double-layer structures, showed that thermally assisted Pdiffusion was negligibly small at 300C. The P diffusion was significantly enhanced afterpassivation with H plasma at the same temperature. It was suggested that the release of Pby atomic H, to form PHx radicals, was the rate-limiting step for enhanced diffusion.J.S.Chou, J.H.Wei, S.C.Lee: Applied Physics Letters, 1993, 63[22], 3060-2

[446-115/116-154]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe transient diffusion of ion-implanted P under non-oxidizing conditions was studied forP doses of 7 x 1013 and 1015/cm2 at 20keV. Annealing (rapid thermal, or furnace) wascarried out in N2 at 950 or 1050C. A kink was observed in secondary ion massspectroscopy profiles after both types of annealing. Significant enhanced diffusion wasobserved for concentrations which were below the kink. It also appeared that P atoms inthe surface region tended to pile up at the SiO2/Si interface. The P diffusion was modelledby using a published empirical model for B and As diffusion. A comparison of varioussimulations showed that P concentrations above the kink (P pile-up) were immobile. The

452

P Bulk Diffusion P

P pile-up appeared to be active in high-dose samples, but inactive in low-dose samples.This behavior was attributed to the formation of some type of P-vacancy complex in thevacancy-rich region at the surface.H.R.Soleimani: Journal of the Electrochemical Society, 1994, 141[8], 2182-8

[446-119/120-225]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionProximity rapid thermal diffusion was investigated, using P spun-on dopant sources, withregard to the effect of the spun-on thickness, the diffusion temperature, the separationbetween wafers, and the nature of the ambient. Sheet resistance measurements, dopantdistributions (obtained by secondary ion mass spectroscopy), and electron microprobeanalyses revealed fast P out-diffusion from the dopant source. A simplified model for Pdiffusion within the spun-on source, and for evaporation, confirmed the occurrence of fasttransport and limited evaporation. An analysis of the process kinetics suggested that, athigh rapid thermal diffusion temperatures, doping of the processed wafer was controlledby gas-phase transport. Surface reactions controlled the process at low diffusiontemperatures.P.B.Grabiec, W.Zagozdzon-Wosik, G.Lux: Journal of Applied Physics, 1995, 78[1], 204-11

[446-123/124-185]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionIn order to investigate the role which is played by H in the anomalous diffusion of P innon mass-separation type plasma doping, a study was made of the effects of pre-hydrogenation upon P diffusion in crystalline samples. Secondary-ion mass spectrometricmeasurements indicated that pre-hydrogenated samples with a -200V radio-frequencybias exhibited a much greater P penetration depth than did samples without pre-hydrogenation, after PH3 plasma treatment, thus indicating that the diffusion of P wasenhanced by pre-hydrogenation. The use of Fourier transformation infra-red spectroscopyrevealed that the Si-H bonds increased after the plasma treatment. On the basis of theseresults, the mechanism of enhanced diffusion was explained in terms of negativelycharged mono-hydrogen and metastable di-hydrides.H.Kakinuma, M.Mohri: Japanese Journal of Applied Physics 2, 1995, 34[10B], L1325-8

[446-134/135-163]

Bulk Diffusion - Theoretical Analysis - Activation EnergyA new model was proposed, for the diffusion of impurities in amorphous Si, which tookaccount of its structural and electronic properties. The model was based upon the many-body kinetic theory of thermally activated rate processes in solids. The low activationenergies which were commonly observed, as well as their dependence upon the impurityconcentration, were explained.J.L.Khait, R.Brener, R.Beserman: Physical Review B, 1988, 38[9], 6107-12

[446-62/63-228]

453

P Bulk Diffusion P

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesThe effect of dopant-dopant interactions upon diffusion was modelled firstly at thequantum-chemical level. The relationship between atomic-scale results and macroscopicbehavior was then obtained by assuming that interactions between molecular orbitalscould be transmitted over hundreds of lattice spacings. Additional flux components weregenerated which modified the usual form of Fick's second law.S.Aronowitz: Journal of Applied Physics, 1991, 69[7], 3901-6

[446-86/87-051]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesA simplified 5-species non-equilibrium kinetic model for P diffusion was presented. Theresultant system of evolution equations was of simple reaction-diffusion form, withconstant diffusivities. Upon using first-order thermodynamic estimates for the reactionrates, the P profile after 10min of pre-deposition exhibited the expected tail. Only whenthe bimolecular generation-recombination rate was significantly increased did a kinkplateau result. This suggested that recombination might be the predominant factor whichproduced the known non-linearity.W.B.Richardson, B.J.Mulvaney: Applied Physics Letters, 1988, 53[20], 1917-9

[446-64/65-180]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesA non-equilibrium kinetic model for P diffusion was developed. The concentrations ofcharged defects and defect-impurity pairs were explicitly determined by solving a systemof quasi-linear equations. Each of these was of reaction-diffusion type, with a constantdiffusivity. This formulation reproduced a number of previously published model results.The calculated profiles for a 600s pre-deposition exhibited both a tail and a well-definedkink plateau. The latter was a direct result of the kinetic model; under the assumption ofstrong bimolecular recombination.W.B.Richardson, B.J.Mulvaney: Journal of Applied Physics, 1989, 65[6], 2243-7

[446-72/73-050]

Bulk Diffusion - Theoretical Analysis - Concentration ProfilesTwo-dimensional P diffusion was studied after furnace annealing or rapid thermalannealing. It was found that there was a considerable disagreement between measured andsimulated profiles in the lateral direction; especially in regions which were directly undera mask edge. These differences were attributed to inaccuracies in the initial implantsimulation, to the absence of damage annealing modelling in the simulation, and to theeffect of stress (in the vicinity of the window edge) upon diffusion. When vacancygeneration (resulting from reaction of the oxide with the Si substrate) was included in thesimulation, better agreement with experimental profiles was found.R.Subrahmanyan, H.Z.Massoud, R.B.Fair: Journal of the Electrochemical Society, 1990,137[5], 1573-81

[446-76/77-041]

Bulk Diffusion - Theoretical Analysis - Effect of ChargeExperiments show that ion pairing has a marked effect upon the diffusion of oppositelycharged impurities. An analysis of literature data was used to deduce the ion pairing

454

P Bulk Diffusion P

coefficients for n-type impurities with B and In. A coefficient with the value of 0.17/Nwas found to describe the pairing case of B-P, where N was the intrinsic electronconcentration. In the case of In-P, the coefficient was about an order of magnitudesmaller. It was suggested that the paired ions occupied adjacent substitutional sites; with asmall perturbation in the Coulomb binding which arose from elastic effects.N.E.B.Cowern: Applied Physics Letters, 1989, 54[8], 703-5

[446-64/65-176]

Bulk Diffusion - Theoretical Analysis - Effect of ChargeThe question of the diffusion mechanism which controlled the diffusion of P into Si atvery high concentrations was examined in terms of a system of reaction-diffusionequations. In particular, the role of the solubility limit was investigated because itrepresented a threshold value at which some of the quasi-chemical reactions which wereinvolved abruptly changed their nature. The corresponding solutions made it possible tocalculate the concentration profiles for both electrically active and inactive impurities as afunction of time and pre-deposition. The formation of a plateau of electrically activedopants was considered in detail, together with the distribution of electrically inactiveprecipitates in the neighborhood of the surface.E.Antoncik: Applied Physics A, 1994, 58[2], 117-23

[446-115/116-150]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsIt was noted that the binding energy between a vacancy and a P atom was large in Si.Therefore, the pair diffusion model and a decrease in quasi-vacancy formation energywith increasing P concentration could be used in the study of P diffusion in Si. Theanomalous diffusion of P consisted of 2 parts: a tail and a plateau. The tail was explainedby using the pair diffusion model, in that the tail was attributed to the presence of excessvacancies. The generation of such vacancies was caused by the dissociation of E centers,and was a characteristic of the pair diffusion model. The plateau was explainedmathematically, using the pair diffusion model, by taking account of the decrease inquasi-vacancy formation energy. The main point which was made here was that, with adecrease in quasi-vacancy formation energy, the binding energy also decreased and thepair diffusion model became identical to the ordinary diffusion model; in which excessvacancies were not generated. It was emphasized that a large value of the effective Pdiffusion coefficient at the plateau was not anomalous, but normal.M.Yoshida: Materials Science Forum, 1995, 196-201, 1595-600

[446-127/128-159]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA theory was developed for impurity diffusion under equilibrium and non-equilibriumconcentrations of point defects. The results of first-principles calculations of several keyquantities were combined with the theory and were compared with experimental data. Itwas found that vacancies and self-interstitials governed the equilibrium diffusion of P.Interstitials tended to predominate. It was also found that the direct-exchange mechanismplayed only a minor role.

455

P Bulk Diffusion P

C.S.Nichols, C.G.Van de Walle, S.T.Pantelides: Physical Review Letters, 1989, 62[9],1049-52

[446-64/65-177]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA simplified model was presented for the treatment of dopant diffusion in the presence ofnon-equilibrium point defect concentrations. The dopant flux was expressed in terms ofan effective diffusivity, and took account of the various couplings which arose from thepresence of defect gradients. The point defect concentrations were calculated by resolvingthe corresponding continuity equations. It was found that the model permitted the fast andaccurate simulated diffusion of P.D.Mathiot, S.Martin: Journal of Applied Physics, 1991, 70[6], 3071-80

[446-91/92-028]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsIt was noted that the binding energy between a vacancy and a P atom in this material waslarge. Therefore, the pair diffusion model and a decrease in quasi-vacancy formationenergy were applicable to P diffusion. With decreasing quasi-vacancy formation energy,the binding energy also decreased. The Fermi level was deduced from Boltzmannstatistics. An anomalous diffusion of P was indicated by the presence of 2 components; atail and a plateau. The tail was attributed to the existence of excess vacancies, which werea characteristic of the pair diffusion model. The plateau was explained by the pairdiffusion model and a decrease in quasi-vacancy formation energy. It was emphasizedthat a large value of the effective P diffusion coefficient at the plateau was normal.M.Yoshida, E.Arai: Japanese Journal of Applied Physics 1, 1995, 34[11], 5891-903. Seealso: Japanese Journal of Applied Physics 1, 1995, 34[11], 5891-903

[446-134/135-156], [446-136/137-126]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsPoint-defect models were compared with experimental data on intrinsically dopedmaterial. Transient dopant diffusion due to low-dose Si implantation damage could bemodelled by using the same parameters as those used for oxidation-enhanced diffusion.This therefore provided an additional technique for monitoring point defect behavior.Consistent parameters were extracted for both experimental conditions, and were fitted toArrhenius relationships. The theory of dopant-defect pairing was found to be crucialwhen modelling implantation damage effects. The effective binding energies for B-defectand P-defect pairs were determined experimentally.H.Park, M.E.Law: Journal of Applied Physics, 1992, 72[8], 3431-9

[446-106/107-137]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsAn effective interstitial surface recombination velocity for the buried Si/SiO2 interface inSIMOX material was used to model accurately the oxidation-enhanced diffusion of P insingly- and multiply-implanted material. The effective recombination velocity at theSIMOX interface was found to be higher than the value for a thermally grown SiO2

interface. The enhancement of the effective recombination velocity depended upon the

456

P Bulk Diffusion P

material formation conditions and was empirically related to the near-interface dislocationdensity. An increased surface interfacial area was considered to be the most likely causeof the increased effective recombination velocity.S.W.Crowder, P.B.Griffin, C.J.Hsieh, G.Y.Wei, J.D.Plummer, L.P.Allen: AppliedPhysics Letters, 1994, 64[24], 3264-6

[446-115/116-185]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsDopant and self-diffusion were known to be governed by both self-interstitials andvacancies; as normalized to their thermal equilibrium values. Since these point defectconcentrations were too low to be investigated directly, the diffusion mechanisms had tobe considered to be unknown. After substituting experimental results on P diffusion at1100C into a new equation that had been derived from the dopant diffusion equation, theequation was solved at the same time as the equation for oxidation-induced stacking faultsunder conditions of local equilibrium between interstitials and vacancies. In this way, theinterstitial and vacancy concentrations were obtained as functions of the diffusion time,and the fractional components of the interstitialcy mechanism for P diffusion, could bedetermined.T.Okino, R.Takaue, M.Onishi: Materials Science Forum, 1995, 196-201, 1631-6

[446-127/128-155]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsEquations for oxidation-enhanced and oxidation-retarded diffusion, and for oxidation-induced stacking faults, were solved simultaneously by using experimental results whichhad been obtained at 1100C. A simple relationship between the concentrations of self-interstitials and vacancies was assumed in order to obtain the solutions. It was concludedthat the product of the concentrations of self-interstitials and vacancies was almost equalto the value at thermal equilibrium, and that the fractional component of the interstitialcymechanism for P-diffusion was larger than 0.5. This showed that the growth of oxidation-induced stacking faults was caused mainly by vacancy undersaturation, and thatoxidation-enhanced diffusion of P occurred.M.Yoshida, S.Matsumoto, Y.Ishikawa: Japanese Journal of Applied Physics, 1986, 25[7],1031-5

[446-48-064]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA model for P diffusion during rapid thermal processing was developed. It was basedupon observations of an enhanced diffusivity in the presence of a relatively highconcentration of interstitial dopant. The effect of an amorphous to crystallinetransformation, upon the initial point defect concentration, was considered for high-doseimplantation. A pair of coupled non-linear partial differential equations for vacancy andself-interstitial assisted P diffusion was solved numerically. From implantation profiledata, it was estimated that the P diffusion behavior could be described by the expression:

D(cm2/s) = 1.41 x 10-5 exp[-1.8(eV)/kT]It was concluded that the so-called anomalous diffusion of P during rapid thermalannealing was due to the presence of a proportion of interstitial dopant which could

457

P Bulk Diffusion P

diffuse relatively quickly, via a self-interstitial assisted mechanism. The above proportiondepended upon the implantation conditions and the crystallographic transformationswhich the lattice underwent during annealing.L.Nanu, A.G.R.Evans: Semiconductor Science and Technology, 1989, 4[9], 711-4

[446-72/73-051]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA previous model for P diffusion was extended so as to include 4 additional reactionsbetween substitutional P, Si self-interstitials, vacancies, and P-defect pairs. All of thereaction rates used in the model were based upon physically plausible kinetic estimates.Numerical solution of the resultant system of 8 coupled partial differential equationsyielded profiles which were in good agreement with experiment.B.J.Mulvaney, W.B.Richardson: Journal of Applied Physics, 1990, 67[6], 3197-9

[446-74-054]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA previously published point defect impurity pair diffusion model was modified in orderto simulate the coupled diffusion of P and self-interstitials in P-implanted material. Theassumed existence of implantation-produced, but empirically determined initial interstitialdistributions of Gaussian type, permitted the net effect of transient enhanced diffusion tobe simulated. In this way, an improved model for P diffusion could be developed for awide range of ion implantation and annealing conditions.H.U.Jäger, T.Feudel, S.Ulbricht: Physica Status Solidi A, 1989, 116[2], 571-81

[446-76/77-042]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsModels were developed for the purpose of describing the coupled diffusion of dopantsand point defects. Firstly, a general model was found for the occurrence of P diffusion viadopant/defect pairs. This assumed that there was local equilibrium of electronic processesbut not of chemical processes. It was concluded that dopant/defect pairing reactions werenear to local equilibrium, whereas defect recombination reactions were not. A simplifiedmodel, which was based upon the assumption that the pairing reactions were near toequilibrium, was then used to simulate P profiles. By including the diffusion of P vianegatively-charged P/vacancy pairs, as well as diffusion via P/interstitial pairs (whichpredominated in intrinsic materials), it was possible to match experimentally determinedP diffusion profiles to surface concentrations which ranged from intrinsic levels to thesolid solubility, at 900 or 1000C.S.T.Dunham: Journal of the Electrochemical Society, 1992, 139[9], 2628-35

[446-93/94-058]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA general model was described, for dopant diffusion via dopant-defect pairs whichcomprised all possible charge states of the species. Local equilibrium was assumed forelectronic processes but not for chemical processes. By using this method, all of thebackward reaction rate coefficients could be replaced by the forward reaction ratecoefficients, plus 2 new parameters. By introducing the intrinsic diffusion coefficient, oneof the 2 new parameters could also be replaced. There was therefore only 1 constant

458

P Bulk Diffusion P

parameter, instead of all of the backward reaction rate coefficients. In addition, the modelautomatically yielded correct results in the intrinsic case. Several simulations of Pdiffusion were presented, and the results were compared with published experimentaldata. Very good agreement between simulation and experiment was obtained. A completeset of parameters was derived for P diffusion in Si at temperatures ranging from 900 to1200C.K.Ghaderi, G.Hobler: Journal of the Electrochemical Society, 1995, 142[5], 1654-8

[446-134/135-163]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA quantitative model was developed for the build-up of P in the surface region duringthermal annealing. The model took account of the migration of mobile diffusioncomponents (vacancies, E-centers) towards the surface, under the influence of the surfacepotential, and of complex formation at high impurity concentrations. The model provideda quantitative explanation for an experimentally observed dependence of the relativeheight of the surface concentration peak upon the volume impurity concentration.O.V.Aleksandrov, N.N.Afonin: Fizika i Tekhnika Poluprovodnikov, 1996, 30[9], 1570-7(Semiconductors, 1996, 30[9], 823-6)

[446-141/142-119]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA quantitative model of P-diffusion gettering was presented which combined the effectsof segregation and self-interstitial injection upon the distribution of dissolved metallicimpurities. The model described metal diffusion in the bulk and in highly P-doped layers,and made it possible to include P diffusion models. By analyzing an approximate solutionfor the quasi steady-state metal distribution it was shown that, for impurities such as Auand Pt, self-interstitial injection enhanced the gettering efficiency as compared with puresegregation. The results were applied to the P-diffusion gettering of Au, and it wasdemonstrated that all of the relevant features of reported Au distributions could beexplained. In the case of 3d metals, which dissolved predominantly at interstitial sites inintrinsic Si, the model permitted the inclusion of the formation of precipitates whichresulted from self-interstitial injection.E.Spiecker, M.Seibt, W.Schröter: Physical Review B, 1997, 55[15], 9577-83

[446-150/151-154]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsThe P diffusion gettering of Au is a reversible process which exhibits marked temperatureand P concentration dependences. The Au concentration profile after successful getteringfollowed the P profile, but almost all of the Au atoms were found in regions where the Pconcentration exceeded about 3 x 10

19/cm

3. This was attributed to a large enhancement of

Au solubility which occurred when the P concentration was greater than 3 x 1019

/cm3. The

simplest explanation for the observed gettering was the assumed formation of Au-P pairswithin the highly-doped P layer. However, quantitative agreement could not be obtainedbetween the present results and a simple segregation model. If the Au bulk solubility wasmodified by taking account of a supersaturation of Si self-interstitials, agreement betweenthe model and the experimental data was then obtained.

459

P Bulk Diffusion P

E.O.Sveinbjörnsson, O.Engström, U.Södervall: Journal of Applied Physics, 1993, 73[11],7311-21

[446-106/107-132]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsIt was shown that Au diffused back and forth, between a highly P-doped layer and thebulk of the material, when the annealing temperature was varied. It was found that thegettered Au was located within the P profile, but was not gettered to the surface. Nointernal gettering or out-diffusion of Au was observed when the Au concentration wasbelow the solubility limit. The results could not be described by using a simplesegregation model unless the Au solubility in the bulk was modified by taking account ofa possible supersaturation of Si self-interstitials.E.O.Sveinbjörnsson, O.Engström, U.Södervall: Materials Science Forum, 1994, 143-147,1641-6

[446-113/114-041]

Bulk Diffusion - Theoretical Analysis - Effect of OxidationThe oxidation-enhanced diffusion of P was investigated over a wide range of oxidationrates, as controlled by varying the initial oxide thickness and the O partial pressure in theambient. Oxidation-retarded diffusion during high-temperature oxidation of (100)-oriented samples was here observed for the first time at very low oxidation rates. Over thepresent range of experimental data, the conventional oxidation-enhanced diffusionformula, Da/DA* = 1 + ARn, was found to be unsuitable for describing the oxidation-enhanced diffusion effect over a wide range of oxidation rates. A new model for theoxidation rate dependence of oxidation-enhanced diffusion was proposed. This tookaccount of the behavior of Si self-interstitials at the Si/SiO2 interface. There, a strongcorrelation was noted between the self-interstitial concentration and the oxidationreaction kinetics.Y.Shibata, S.Hashimoto, K.Taniguchi, C.Hamaguchi: Journal of the ElectrochemicalSociety, 1992, 139[1], 231-6

[446-86/87-080]

Bulk Diffusion - Theoretical Analysis - Effect of OxidationThe oxidation-enhanced or oxidation-retarded diffusion of substitutional impurities isusually expressed in terms of equations which involve self-interstitial (I) and vacancy (V)concentrations. By using the asymptotic forms of the solutions to these equations, anapproximate relationship between I and V was obtained. Identification of thisapproximate relationship with that for local equilibrium between I and V gave a simplemethod for obtaining mathematically self-consistent solutions to the oxidation-enhancedand oxidation-retarded diffusion equations. In this way, I and V were obtained asfunctions of the diffusion time. The fractional component of the interstitialcy mechanismfor the oxidation-enhanced diffusion of P was obtained. The value of d1

P was 0.93 at1100C.T.Okino: Japanese Journal of Applied Physics 1, 1992, 31[4], 965-9

[446-88/89-055]

460

P Bulk Diffusion P

Bulk Diffusion - Theoretical Analysis - Effect of OxidationOxidation-enhanced and oxidation-retarded diffusion effects were incorporated intodifferential equations for self-interstitials and vacancies. An extended relationship for thelocal equilibrium between interstitials and vacancies, VIm = n, was derived fromasymptotic forms of the solutions to the oxidation-enhanced and oxidation-retardeddiffusion equations. It was found that local equilibrium between interstitials andvacancies did not occur within the present range of diffusion times. Interstitial andvacancy diffusivities of 2.4 x 10

-9 and 2.1 x 10

-10cm

2/s, respectively, at 1100C were

deduced from the results. The corresponding thermal equilibrium concentrations were 3.3x 10

16 and 2.2 x 10

17/cm

3.

T.Okino, M.Onishi: Japanese Journal of Applied Physics 1, 1994, 33[6A], 3362-7[446-113/114-048]

Bulk Diffusion - Theoretical Analysis - Effect of PressureFirst-principles total-energy calculations of the atomic diffusion of group-V impuritiesrevealed an effect of pressure upon the activation energy for diffusion. In the case of thevacancy mechanism, the activation energy for P decreased with pressure. In the case ofthe interstitial mechanism, the formation energy of the interstitial impurity exhibited ageneral tendency to increase with pressure. The microscopic origin of the pressuredependence was explained in terms of the local strains around defects. The negativepressure dependence which was common to the vacancy-mediated diffusion of group-Vimpurities could be explained by the peculiar properties of the isolated vacancy. Theseincluded a breathing distortion of surrounding Si atoms towards the vacancy site, givingrise to a tensile strain around the vacancy, and lattice distortions which originated fromthe vacancy and caused weak vacancy-impurity interactions. The positive pressuredependence of interstitial-mediated diffusion was closely related to the atomic structuresof the interstitial impurities, which produced compressive strains in the surrounding Si-Sibonds.O.Sugino, A.Oshiyama: Physical Review B, 1992, 46[19], 12335-41

[446-106/107-131]

Bulk Diffusion - Theoretical Analysis - Effect upon DefectsThe contribution of diffusion-induced structural changes to the anisotropy of P diffusioninto monocrystalline material was analyzed. A difference in the dislocation densities on(100)-type and (111)-type planes, due to P diffusion on these planes, affected thediffusion process and led to anisotropy.L.N.Larikov, E.I.Bogdanov, E.A.Maximenko: Crystal Research and Technology, 1990,25[1], 51-4

[446-74-054]

Bulk Diffusion - Theoretical Analysis - Enhanced DiffusionA computer study was made of the accelerated diffusion of impurities which wereintroduced by high-temperature implantation. The diffusion equations were solved byusing the finite difference method. It was assumed that the acceleration was due to thepresence of an excess of point defects, and was proportional to their change inconcentration with depth. The calculated profiles were compared with data on the

461

P Bulk/Grain Boundary Diffusion P

implantation of P (at 900C) to doses of between 3 x 1013 and 1016/cm2. It was found thatthe dopant profiles which were observed after implantation agreed well with the results ofcalculations which assumed that the diffusion length of the point defects was 500nm.L.N.Aleksandrov, T.V.Bondareva, G.A.Kachurin, I.E.Tyschenko: Fizika i TekhnikaPoluprovodnikov, 1991, 25[2], 227-30 (Soviet Physics - Semiconductors, 1991, 25[2],137-9)

[446-84/85-064]

Table 72Grain Boundary Diffusion of P in Polycrystalline Si

T (C) Dd (cm3/s)566 4.96 x 10-25

639 6.00 x 10-24

693 1.26 x 10-23

793 5.66 x 10-22

980 8.26 x 10-20

Bulk Diffusion - Theoretical Analysis - Enhanced DiffusionThe transient enhanced diffusion which occurred during the annealing of ion implantationdamage was analyzed as a function of temperature; especially with regard to a markedenhancement which resulted at low temperatures. A non-equilibrium point defect modelwas used to investigate this behavior. A simulation was performed of publishedmeasurements on P diffusion after Si implantation to low doses. This revealed that thediffusivity enhancement could be as high as 10000 during the first few minutes ofannealing at about 800C. However, at high temperatures the enhancement factor was lessthan 10 for a few seconds. The effect of this behavior was reflected by the dopant profileof a p-n-p bipolar transistor. It was found that the base Gummel number dependedmarkedly upon the annealing conditions which were used just after P implantation.B.Baccus: Solid-State Electronics, 1992, 35[8], 1045-9

[446-93/94-057]

372 Grain Boundary Diffusion - Quantitative DataThe migration of radioactive P in polycrystalline material was studied at temperaturesranging from 566 to 980C. The diffusion profiles were determined by using anodicoxidation sectioning and radiotracer detection. The grain boundary diffusivities wereobtained by using the LeClaire method. It was found that the results (table 72), assuminga boundary width of 5nm, were described by the expression:

D(cm2/s) = 0.0048 exp[-2.65(eV)/kT]M.R.Murti, K.V.Reddy: Semiconductor Science and Technology, 1989, 4[8], 622-5

[446-72/73-052]

Grain Boundary Diffusion - Quantitative DataSamples of polycrystalline material were studied at temperatures of between 750 and1050C by using a groove and stain technique. It was shown that the Fisher model for

462

P Grain Boundary Diffusion P

grain boundary diffusion accurately described the data when an infinite source wasassumed. It was found that the bulk diffusion data were described by:

D(cm2/s) = 1.6 x 1015 exp[-2.09(eV)/kT]while the grain boundary diffusivity was described by:

D(cm2/s) = 4.0 x 10-5 exp[-1.4(eV)/kT]A grain boundary width of 0.5nm was assumed in the latter case.P.H.Holloway, T.Abrantes: Journal of Vacuum Science and Technology A, 1989, 7[3],1573-8

[446-74-054]

373 Grain Boundary Diffusion - Quantitative DataThe diffusion of donor elements in fine-grained and coarse-grained polycrystallinematerial was studied at temperatures ranging from 900 to 1150C. Tracers (32P) were usedto determine the concentration/depth profiles via sectioning. By means ofautoradiography, the lateral distribution of the radiotracers over the sample surface wasrevealed. The grain boundary diffusivity (table 73) could be described by:

D (cm2/s) = 120 exp[-2.87(eV)/kT]F.H.M.Spit, H.Bakker: Physica Status Solidi A, 1986, 97[1], 135-42

[446-48-063]

Table 73Grain Boundary Diffusion of P in Si

T (C) D (cm2/s)1150 6.9 x 10-9

1100 3.9 x 10-9

1050 1.4 x 10-9

1000 4.7 x 10-10

950 2.2 x 10-10

900 4.0 x 10-11

Grain Boundary Diffusion - Qualitative Observations - Effect of DopingIt was demonstrated that the P diffusion doping of polycrystalline films that weredeposited on SiO2 resulted in the movement of underlying Na, Li, K, and B impuritiesinto the diffusion glass. It was suggested that the B and alkali element transport was dueto the preferential segregation of impurities into the P glass via grain boundary diffusion.This behavior contrasted with that of Si films which had been doped in situ with P, andthen recrystallized by annealing. In this case, no diffusion glass was formed and little orno impurity removal was detected. It was concluded that P-diffused polycrystallinematerial was a potentially better gate electrode for metal oxide semiconductor devices.C.W.Pearce, J.L.Moore, F.A.Stevie: Journal of the Electrochemical Society, 1993,140[5], 1409-13

[446-106/107-144]

463

P Grain Boundary Diffusion/Surface Diffusion Pb

Grain Boundary Diffusion - Qualitative Observations - Enhanced DiffusionTwo different electron beam-induced current methods were used to examine the diffusionof P at the grain boundaries. These were: imaging of the displacement of the pn junctionat grain boundaries, and bright grain boundary contrast at low beam voltages. Solar cellson polycrystalline Si were used. Enhanced diffusion at grain boundaries was observed in2 of 5 cells when using either of the above methods. One of the cells also exhibitedenhanced diffusion at dislocations. It was found that there was a correlation between Pdiffusivity and an enhanced recombination activity at the grain boundaries.K.Schimpf, J.Palm, H.Alexander: Crystal Research and Technology, 1994, 29[8], 1123-9

[446-117/118-197]

PbSurface Diffusion - Quantitative DataThe very early stages of Pb deposition on (111)-(7 x 7) surfaces were investigated bymeans of scanning tunnelling microscopy. The combination of variable-temperaturescanning with unusually long periods of imaging time permitted the observation thatsingle Pb atoms were highly mobile within each half-(7 x 7) unit cell. Individual jumps ofsingle atoms between different half-cells had to be resolved, as well as the formation ofatom pairs. An activation energy of 0.64eV was deduced for the diffusion of single atomsbetween different half-cells. The decrease in the density of single atoms (and aconcomitant change in the density of larger clusters) was consistent with the results ofcontinuous observation. This demonstrated that only a very small effect upon thedynamics of single Pb atoms could be attributed to the scanning motions of the tip.J.M.Gómez-Rodríguez, J.J.Sáenz, A.M.Baró, J.Y.Veuillen, R.C.Cinti: Physical ReviewLetters, 1996, 76[5], 799-802

[446-134/135-165]

Surface Diffusion - Qualitative ObservationsScanning tunnelling microscopy was used to study the temperature and coveragedependences of the structure of Pb on the (111) 7 x 7 surface. It was found that, at lowcoverages, the Pb atoms favored faulted sites. The ratio of the numbers of Pb atoms onfaulted and unfaulted sites increased during annealing. An energy difference of 0.05eVwas estimated to be associated with a Pb atom on these 2 types of site. Mobility of Pbatoms on (111) was observed, at temperatures which were as low as 260C, at coveragesof 0.1 or 1 monolayer.D.Tang, H.E.Elsayed-Ali, J.Wendelken, J.Xu: Physical Review B, 1995, 52[3], 1481-4

[446-123/124-185]

464

Pd

Figure 15: Diffusivity of Pd in Si (see table 74). The least-squares regression line throughthese points gives D (cm2/s) = 3.13 x 10-4 exp[1.10(eV)/kT]

Bulk Diffusion - Quantitative DataRadioactive tracer techniques and X-ray photoelectron spectroscopy were used to studydiffusion in films of hydrogenated amorphous material at temperatures of between 200and 500C. The films were P-doped and had various defect structures. The migration rateand the diffusion coefficient were found to depend upon the defect structure. Thediffusivity increased from 1.4 x 10-17 to 6.8 x 10-16cm2/s upon increasing the temperaturefrom 200 to 250C.

1.0E-16

1.0E-15

1.0E-14

1.0E-13

1.0E-12

16 17 18 19 20 21 22

16.1817.319.1221.14

104/T(K)

D (cm2/s)

465

Pd Bulk Diffusion Pd


[446-55/56-038]


[446-86/87-049]

Table 74Diffusivity of Pd in Amorphous Si

T (C) D (cm2/s)345 3.3 x 10-13

305 7.6 x 10-14

250 8.2 x 10-15

200 5.9 x 10-16

Bulk Diffusion - Qualitative Observations - Concentration ProfilesIt had previously been concluded that asymmetrical diffusion during Pd2Si formationcould not be easily explained within the framework of existing theories. Here, aninvestigation was made of whether observed asymmetries in profile shape aftersilicidation could be caused by the secondary ion mass spectrometric technique.However, it was not possible to show that the asymmetries were due only to secondaryion mass spectrometric measurements. High-resolution Rutherford back-scatteringspectrometry was used to confirm that dopant diffusion really did occur due tosilicidation.M.Wittmer, P.Fahey, J.Cotte, S.S.Iyer, G.J.Scilla: Physical Review B, 1992, 45[19],11383-6

[446-88/89-055]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe in-diffusion and annealing kinetics of Pd were investigated by means of the deep-level transient spectroscopy of p+nn+ structures. The substitutional Pd concentration wasmonitored via measurements of the electrically active centers. A study of the in-diffusionprocess, at temperatures ranging from 880 to 1050C, showed that Pd diffused mainly viathe so-called kick-out mechanism. Annealing was performed at 550C on wafers whichwere homogeneously supersaturated with Pd. A consequent decrease in the electricalactivity of Pd in the bulk was observed. The experimental results suggested that, during

466

Pd Bulk Diffusion Pr

annealing, the substitutional Pd changed its configuration via interactions with vacanciesaccording to the dissociative reaction.J.Vicente, L.Enríquez, E.Rubio, L.Bailón, J.Barbolla: Journal of the ElectrochemicalSociety, 1993, 140[3], 868-70

[446-099/100-099]

PrBulk Diffusion - Quantitative DataThe first investigations of the diffusion of Pr in Si were reported. It was found that, attemperatures ranging from 1100 to 1250C, the Pr diffusivity increased from about 10-13 toabout 1.5 x 10-12cm2/s. The temperature dependence of the diffusion coefficient could bedescribed by:

D (cm2/s) = 0.005 exp[-3.3(eV)/kT]D.E.Nazyrov, G.S.Kulikov, R.S.Malkovich: Technical Physics Letters, 1997, 23[1], 68-9

[446-150/151-154]

467

Pt

Figure 16: Diffusivity of Pt in Si. The least-squares regression line through these pointsgives D (cm2/s) = 3.39 x 104 exp[-2.95(eV)/kT]

Bulk Diffusion - Quantitative DataRadioactive tracer techniques and X-ray photoelectron spectroscopy were used to studydiffusion in films of hydrogenated amorphous material at temperatures of between 200and 500C. The films were P-doped and had various defect structures. The migration rateand the diffusion coefficient were found to depend upon the defect structure. Thediffusivity at 200C was 10-16 cm2/s.

1.0E-16

1.0E-15

1.0E-14

1.0E-13

1.0E-12

1.0E-11

1.0E-10

1.0E-09

1.0E-08

1.0E-07

1.0E-06

7 8 9 10 11 12 13 14

table 75table 76

104/T(K)

D (cm2/s)

468

Pt Bulk Diffusion Pt


[446-55/56-038]

Bulk Diffusion - Quantitative DataLayers (2nm) of Pt were incorporated into amorphous material by using a laser quenchingtechnique. The samples were then analyzed by using Rutherford back-scatteringspectrometry and channelling methods. It was found that the diffusivity was equal to 1.8 x10-16cm2/s at 300C. Impurity diffusion in the amorphous material was found to be rapid athigh concentrations, and very slow at low concentrations.M.A.Harith, S.U.Campisano, J.M.Poate: Semiconductor Science and Technology, 1988,3[9], 829-31

[446-72/73-046]

Bulk Diffusion - Quantitative DataA method was described for the measurement of equilibrium concentrations of vacanciesat 800C. The concentration was established during long-term annealing in an inert andclean ambient. Assuming the predominant operation of the Frank-Turnbull mechanismduring Pt diffusion at 800C, the equilibrium vacancy concentration could be deducedfrom the resultant diffusion profiles. A concentration of 4.1 x 1013/cm3 at 800C wasdeduced from the equations of the Frank-Turnbull mechanism. It was estimated that thevacancy diffusion coefficient was equal to 1.3 x 10-10cm2/s.H.Zimmermann: Applied Physics Letters, 1991, 59[24], 3133-5

[446-84/85-073]

Bulk Diffusion - Quantitative DataAnalytical expressions which permitted a direct determination of the equilibriumconcentrations of self-interstitials and vacancies, on the basis of measured impuritydiffusion profiles, were deduced from the equations for kick-out and Frank-Turnbulldiffusion mechanisms. In order to deduce the equilibrium point-defect concentrations, itwas necessary only to measure the diffusion profiles. As a test, the equilibrium vacancyconcentration in Si was determined by means of Pt diffusion. A value of about 4.4 x 1013

/cm3 at 770C was deduced for the equilibrium vacancy concentration. The vacancydiffusion coefficient was estimated to be equal to 2.2 x 10-11cm2/s. The rate of generationof vacancies by internal sources appeared to be lower than 10-5/s at 770C.H.Zimmermann, H.Ryssel: Physical Review B, 1991, 44[16], 9064-7

[446-84/85-074]

375 Bulk Diffusion - Quantitative DataThe migration of Pt was investigated by using spreading resistance techniques, andsubstrates with various dislocation densities. The results for essentially perfect non-dislocated material provided evidence for the kick-out mechanism, due to the appearanceof characteristic features in the concentration-depth profiles and to the diffusion-inducedformation of stacking faults. In the case of solar-grade Si, the incorporation ofsubstitutional Pt into the bulk was governed by the annihilation of self-interstitials atgrown-in dislocations with a density of about 109/m2. In this material, the efficiency of

469


dislocations as self-interstitial sinks appeared to be reduced and to depend upon thepenetration depth. Measurements of plastically deformed material with a dislocationdensity of 1011 to 1013/m2 yielded, for the first time, diffusion profiles that were governedby the transport properties of interstitial Pt. The temperature dependence of the overall Ptdiffusivity (table 75) could be described by:

D(m2/s) = 0.00021 exp[-1.79(eV)/kT]W.Lerch, N.A.Stolwijk, H.Mehrer, C.Poisson: Semiconductor Science and Technology,1995, 10[9], 1257-63

[446-125/126-151]

Table 75Diffusivity of Pt in Si

T (C) D (cm2/s)1120 5.6 x 10-7

1085 6.0 x 10-7

1050 4.4 x 10-7

1000 1.7 x 10-7

950 8.1 x 10-8


[446-86/87-049]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesDiffusion of ion-implanted elements in crystalline Si was investigated. The implantationwas limited to photolithographically defined areas of the wafer, and a spreadingresistance technique was used to measure the 3-dimensional concentration profiles of themetal atoms after high-temperature annealing. It was found that lateral spread under themask was greater than vertical diffusion; especially on the side opposite to the implanteddiffusion source. All of the important features of the measured profiles could be explainedas being a result of a kick-out diffusion mechanism. The peculiar shape of theconcentration profiles was attributed to an interplay between the incoming flux ofinterstitial metal atoms and the outgoing flux of Si self-interstitials that were generated bythe kick-out reaction. In spite of the high lateral diffusion it was noted that, by a suitablecombination of implantation fluence and annealing temperature, it was possible to limit

470


this lateral spread to within about 200µ, while maintaining a high metal concentration inthe region under the implanted area.S.Coffa, V.Privitera, F.Frisina, F.Priolo: Journal of Applied Physics, 1993, 74[1], 195-200

[446-106/107-133]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesA 2-step diffusion method was proposed for leveling the depth profiles of Pt-inducedtraps. The Pt was first diffused at high temperatures in a N ambient, and was then re-diffused at a lower temperature. The results showed that this method caused the trapconcentration near to the surface of a wafer to decrease, while that deep within the waferincreased. Therefore, the depth profile of the trap concentration could be evened outthroughout the entire wafer.B.Deng, C.Shu, H.Kuwano: Semiconductor Science and Technology, 1996, 11[4], 535-7

[446-134/135-165]

Bulk Diffusion - Qualitative Observations - Effect of DefectsIt was found that the study of Pt diffusion permitted the separate observation of Si self-interstitials and vacancies. The diffusion of Pt could be described by the kick-outmechanism at temperatures above about 900C. At temperatures below about 850C, thedissociative mechanism governed Pt diffusion. The use of numerical simulation furnisheda consistent set of parameters which described the diffusion of Pt at temperatures rangingfrom 700 to 950C. The generation or recombination of self-interstitials and vacancies wasineffective; at least below 850C. The concentration of substitutional Pt was determined bythe initial concentration of vacancies at temperatures below 850C. It was concluded thatPt diffusion experiments, when performed at temperatures below 850C, could be used tomeasure vacancy distributions in Si.H.Zimmermann, H.Ryssel: Applied Physics A, 1992, 55[2], 121-34

[446-93/94-047]

Table 76Diffusivity of Pt in Amorphous Si

T (C) D (cm2/s)600 2.4 x 10-13

560 4.2 x 10-14

502 5.2 x 10-15

455 3.5 x 10-16

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe diffusion of Pt into n-type or p-type material was carried out at temperatures rangingfrom 850 to 1000C; for times ranging from 1 to 50h. Three deep levels which wereassociated with Pt were detected using deep-level transient spectroscopy. These were 2acceptor levels at Ec - 0.23 and Ec - 0.52eV, and a donor level of Ev + 0.36eV. The Ec -

471


0.23 and Ev + 0.36eV levels were produced by Pt which occupied substitutional sites. TheEc - 0.52eV level was not characterized, but was probably associated with an interstitialPt-O or other defect complex. Diffusion profiles of the substitutional Pt showed that Ptdiffused into Si mainly via the so-called kick-out mechanism.Y.K.Kwon, T.Ishikawa, H.Kuwano: Journal of Applied Physics, 1987, 61[3], 1055-8

[446-60-015]

Bulk Diffusion - Qualitative Observations - Effect of DefectsDeep-level transient spectroscopy was used to determine Pt concentrations at variousdepths. Inverse U-shaped Pt profiles were found after diffusion at 700C. A higher Ptconcentration was found in the bulk than at the surface. Such profiles could be explainedby invoking the Frank-Turnbull mechanism, if it were assumed that the initial vacancyconcentration was higher than the equilibrium concentration. It was suggested that theobservation of inverse U-shaped profiles implied that the diffusion process in the bulkwas governed by the initial vacancy concentration, and not by vacancy equilibrium, attemperatures below about 850C.H.Zimmermann, H.Ryssel: Applied Physics Letters, 1991, 59[10], 1209-11

[446-84/85-073]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe diffusion of Pt into lightly damaged regions of n-type material was investigated byusing deep-level transient spectroscopy and capacitance-voltage profiling of Pt-silicideSchottky diodes. The diodes had been implanted through the junction (with O, F, or Clions), and had then been annealed at 700C. The distribution of the in-diffused Pt wasdeduced by monitoring electron emissions from the acceptor level of Pt which occupied adistorted substitutional configuration. At a typical implantation dose of about 1011/cm2, anincreased accumulation (by about 2 orders of magnitude) of Pt was observed; ascompared with diffusion into non-implanted material. It was found that the in-diffused Ptwas distributed almost in parallel with the vacancy distribution which was generatedduring implantation. This apparent decoration of the primary damage profile occurredwithout the simultaneous introduction of other electrically active defects at a comparablescale. It was concluded that the residual damage which was present during annealing wassufficient to promote the accumulation to saturation of substitutional Pt in the region ofthe primary implantation damage. It was suggested that this so-called guided in-diffusioneffect might be exploited for device processing.K.B.Nielsen, B.Holm: Materials Science Forum, 1995, 196-201, 1985-90

[446-127/128-159]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe diffusion of Pt at low temperatures is a convenient method for characterizing vacancyprofiles in Si. It was found that the results of experiments on float-zone and Czochralski-type samples, at temperatures ranging from 680 to 842C, disagreed with the predictions ofpublished models. The parameters which governed the diffusion of point defects and Pt inSi were determined for the above temperature range.M.Jacob, P.Pichler, H.Ryssel, R.Falster: Journal of Applied Physics, 1997, 82[1], 182-91

[446-152-0471]

472


Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe diffusion of Pt into slightly-damaged regions of ion-implanted material wasinvestigated by using deep-level transient spectroscopy and capacitance-voltage profiling.The samples consisted of PtSi/Si n-type Schottky diodes which had been implanted withO, F or Cl to a depth that was intermediate between the zero-bias and reverse-biasdepletion boundaries, and had then been annealed at 700C. The distribution of in-diffusedPt was found by monitoring electron emissions from an acceptor level which hadpreviously been attributed to Pt in a distorted substitutional configuration. The Pt wasfound to be distributed almost in parallel with the vacancy distribution that was generatedduring implantation. For a typical implantation dose of about 1011/cm2, there was anenhancement of the Pt accumulation by about 2 orders of magnitude; as compared withdiffusion into non-implanted material. This apparent decoration of the primary damageprofile occurred without the simultaneous introduction of other electrically active defectson a comparable scale. It was deduced that the residual damage which was present duringannealing was sufficient to promote the accumulation to saturation of substitutional Pt inthe region of primary implantation damage. This quite remarkable property of guided in-diffusion, a type of gettering phenomenon, was expected to have potential applications indevice engineering.B.Holm, K.B.Nielsen: Journal of Applied Physics, 1995, 78[10], 5970-4

[446-125/126-152]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsThe migration of some elements yields Si self-interstitial diffusivities, that exceed thosewhich are obtained from dopant marker experiments by 6 orders of magnitude at 800C.Both types of experiment could be reconciled by assuming the existence of a non-annihilating interstitial trap that was related to C. Selected metal diffusion data were re-analyzed in this context. Non-annihilating immobile traps and a second-order reactionwhich involved interstitial C and C-C pairs were considered. Quantitative point defectparameters were obtained at 1115C for an assumed trap concentration of 5 x 1016/cm3.The estimated equilibrium concentration of Si self-interstitials varied inversely with thetrap concentration, while the product of the self-interstitial diffusivity and the self-interstitial concentration remained almost constant. It was concluded that agreement, ordisagreement, of metal diffusivity results with the Si self-diffusion coefficient could notbe used to exclude or prove the occurrence of trap-limited diffusion. The published valuesof point-defect parameters which had been obtained by neglecting traps were suggested torepresent lower bounds on the self-interstitial diffusivity, and upper bounds on theequilibrium concentration of self-interstitials.H.J.Gossmann, P.A.Stolk, D.J.Eaglesham, C.S.Rafferty, J.M.Poate: Applied PhysicsLetters, 1995, 67[21], 3135-7

[446-127/128-154]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsThe diffusion of Pt can be explained in terms of the kick-out and dissociativemechanisms. The resultant set of 4 coupled differential equations was here solvednumerically without making any simplifications. It was found that the concentration ofsubstitutional Pt was strongly affected by the initial vacancy concentration. A complete

473

Pt Bulk Diffusion S

set of parameters was suggested which described the diffusion of Pt at temperaturesranging from 700 to 950C. The dissociative mechanism predominated at temperaturesbelow 850C, while the kick-out mechanism predominated at temperatures above 900C.The reaction constant for the dissociative mechanism was estimated to be 1.6 x 10-14cm3/s,whereas the reaction constant for the kick-out mechanism was of the order of 0.00017/s at700C. The use of Pt diffusion at temperatures below 850C was suggested to be a meansfor measuring vacancy distributions in Si.H.Zimmermann, H.Ryssel: Journal of the Electrochemical Society, 1992, 139[1], 256-62

[446-86/87-058]

S

Figure 17: Diffusivity of 35S in Si (see table 77)

1.0E-09

1.0E-08

1.0E-07

1.0E-06

5 6 7 8

5.986.126.296.446.66.746.846.987.257.53

104/T(K)

D (cm2/s)

474

S Bulk Diffusion S

377 Bulk Diffusion - Quantitative DataThe diffusion of

35S into Si was investigated by using the closed-ampoule technique.

Penetration profiles of erfc-type were determined by means of mechanical sectioning. Itwas found that the diffusion coefficients at temperatures ranging from 1328 to 1671K(table 77) obeyed the Arrhenius law:

D(m2/s) = 4.7 x 10

-6 exp[-1.80(eV)/kT]The high diffusivity could be reconciled with the preferred incorporation of S on latticesites by supposing substitutional-interstitial exchange. Under this assumption, the long-range transport appeared to be controlled by a minority of interstitial S atoms.F.Rollert, N.A.Stolwijk, H.Mehrer: Applied Physics Letters, 1993, 63[4], 506-8

[446-106/107-147]

Table 77Diffusion of 35S in Si as a Function of Temperature

T (K) D (cm2/s)1671 1.8 x 10-7

1633 1.3 x 10-7

1590 9.2 x 10-8

1554 6.8 x 10-8

1516 4.7 x 10-8

1484 4.0 x 10-8

1463 3.1 x 10-8

1432 2.2 x 10-8

1380 1.2 x 10-8

1328 7.1 x 10-9

Bulk Diffusion - Qualitative Observations and Point DefectsInteractions between this group-VI element and fast-diffusing impurities (Cr, Fe, Mn)were studied. Infra-red absorption, electron spin resonance, and neutron activation studiesshowed that the group-VI element effectively interacted with each other or with the fast-diffusing impurities. It was found that, for each pairing of a group-VI element with a fast-diffusing impurity, there was a certain annealing temperature at which they interactedmost efficiently. A definite correlation was established between this annealingtemperature and the thermodynamic Gibbs free energy (at 298K) of the correspondingcompounds. This correlation suggested that the interaction process mainly involved theformation of electrically neutral chemically bonded complexes by substitutional group-VIelement centers and interstitial fast-diffusant centers.M.K.Bakhadirkhanov, S.I.Askarov, N.Norkulov: Physica Status Solidi A, 1994, 142[2],339-46

[446-117/118-194]

475

Sb

Figure 18: Diffusivity of Sb in Si (see table 77)

Bulk Diffusion - Quantitative DataThe effect of surface oxidation upon Sb diffusion was investigated. An extrinsic Asbackground was used to provide a constant electron concentration for the diffusion of theSb profile. Annealing was carried out at 850, 950 or 1050C, and the diffusion of Sb andAs was measured under inert and oxidizing conditions by means of secondary ion massspectrometry. It was found that the diffusion of As was enhanced, while the diffusion ofSb within the As layer was retarded. The data could be described by:

DSb = 0.214 exp[-3.65/kT] + 15 (n/ni)exp[-4.08/kT]and

DAs = 8.0 exp[-4.05/kT] + 12.8 (n/ni)exp[-4.05/kT]

1.0E-18

1.0E-17

1.0E-16

1.0E-15

1.0E-14

1.0E-13

1.0E-12

1.0E-11

1.0E-10

1.0E-09

1.0E-08

1.0E-07

7 8 9


104/T(K)

D (cm2/s)

476

Sb Bulk Diffusion Sb

where n/ni accounted for the concentration dependent diffusion which was proportional tothe donor concentration (n) over the intrinsic electron concentration (ni). It was concludedthat the results provided support for a dual, interstitial/vacancy, mode of dopant diffusion.They also provided evidence against the use of a vacancy-only diffusion model.E.A.Perozziello, P.B.Griffin, J.D.Plummer: Applied Physics Letters, 1992, 61[3], 303-5

[446-93/94-045]

Bulk Diffusion - Quantitative DataThe redistribution of atomic-layer doped Sb was investigated during post-growthannealing by using secondary ion mass spectrometry depth profiling. Kink developmentand SiO2/Si interface segregation were observed. The diffusivity was concentration-dependent over Sb contents of about 5 x 1017/cm3. Two types of Sb diffusion appeared toinvolve activated processes at temperatures of between 710 and 850C. The associatedactivation energies ranged from 0.4 to 0.6eV. These values were much lower than that(4eV) for bulk diffusion; which included the enthalpies of vacancy formation andmigration.S.Fukatsu, S.Kubo, Y.Shiraki, R.Ito: Applied Physics Letters, 1991, 58[11], 1152-4

[446-84/85-074]

378 Bulk Diffusion - Quantitative DataThe diffusion of donor elements in fine-grained and coarse-grained polycrystallinematerial was studied at temperatures ranging from 900 to 1150C (table 78). Tracers(125Sb) were used to determine the concentration/depth profiles via sectioning. By meansof autoradiography, the lateral distribution of the radiotracers over the sample surface wasrevealed.F.H.M.Spit, H.Bakker: Physica Status Solidi A, 1986, 97[1], 135-42

[446-48-063]

Table 78Diffusivity of Sb in Si

T (C) D (cm2/s)1150 2.3 x 10-13

1100 7.9 x 10-14

1050 2.1 x 10-14

379 Bulk Diffusion - Quantitative DataThe effect of P diffusion, in high concentrations, upon the diffusion of an Sb marker layerwas investigated (table 79). The marker layer was separated from the surface by a 4µlayer of epitaxially grown Si. In order to reduce the effects of implantation enhancementand P precipitation upon the diffusion of the marker layer, the P was implanted into apoly-Si layer that had been deposited onto monocrystalline substrate. It was found that thediffusion of the Sb marker layer was already reduced by the epilayer. On the basis of the

477


diffusion coefficients, upper limits were placed on the fractional diffusivity of Sb.Contrary to the results of previous investigations, the diffusion of the Sb marker layer wasfound to be enhanced below the regions into which P had been implanted. A comparisonof the behavior of float-zone and Czochralski samples showed that the enhanced diffusionof Sb could be explained only by the injection of self-interstitials from the P-dopedregion. Since the poly-Si layer recrystallized, this self-interstitial injection could be theresult of P which diffused mainly via self-interstitials, P precipitation, or both. By usingBoltzmann-Matano analyses and Sb diffusivity data, the fractional diffusivity of P viaself-interstitials was estimated to be lower than 0.71 at 950C.P.Pichler, H.Ryssel, R.Ploss, C.Bonafos, A.Claverie: Journal of Applied Physics, 1995,78[3], 1623-9

[446-123/124-186]

Table 79Effect of P (3 x 1016/cm2) upon the Diffusion of Sb

T (C) Material P D(cm2/s)950 float-zone no 1.1 x 10-16

950 float-zone yes 1.8 x 10-16

950 Czochralski no 1.2 x 10-16

950 Czochralski yes 2.1 x 10-16

1050 float-zone no 4.5 x 10-15

1050 float-zone yes 5.7 x 10-15

1050 Czochralski no 2.8 x 10-15

1050 Czochralski yes 3.6 x 10-15

380 Bulk Diffusion - Quantitative DataStudies were made of diffusion in material with high donor concentrations that wereproduced by P doping. It was found that, for donor concentrations which were belowabout 2 x 10

20/cm

3, the diffusivity (table 80) depended linearly upon the dopant

concentration. However, at higher dopant concentrations, the diffusivity increasedmarkedly with increasing donor concentration. This behavior was successfully modelledin terms of the vacancy-percolation model, and it was concluded that collectivephenomena played a significant role at high donor concentrations.A.N.Larsen, K.K.Larsen, P.E.Andersen, B.G.Svensson: Journal of Applied Physics, 1993,73[2], 691-8

[446-106/107-132]

381 Bulk Diffusion - Quantitative DataThe diffusion of Sb in relaxed SiGe layers which had been prepared by means ofmolecular beam epitaxy, with x-values of between 0 (table 81) and 0.5, was studied as afunction of composition. The diffusivity of Sb was found to increase with the Ge content,while the activation energy for diffusion decreased with increasing Ge content. However,

478


the measured activation energies were significantly higher than the activation energiesthat were predicted by extrapolating between the activation energies for pure Si and pureGe.A.N.Larsen, P.Kringhøj: Applied Physics Letters, 1996, 68[19], 2684-6

[446-134/135-200]


T (C) D (cm2/s)1100 6.0 x 10-12

1075 4.0 x 10-12

1050 3.5 x 10-12

1025 2.4 x 10-12

1000 1.6 x 10-12

Bulk Diffusion - Quantitative DataThe depth-dependent nitridation-enhanced diffusion of Sb was investigated by theannealing (810 to 910C, 0.25 to 2h) of (100) doping superlattice structures in NH3. Thesemulti-layered doping superlattices consisted of six 10nm-wide Sb doping spikes that werespaced 100nm apart. The nitridation-enhanced diffusion of Sb, which was attributed tovacancy injection, indicated vacancy supersaturation values of 3 to 5. On the basis of thespatial decay of Sb nitridation-enhanced diffusion, lower bounds on vacancy diffusivitieswere deduced. They were equal to 7.9 x 10-14, 1.2 x 10-12 and 2.1 x 10-11cm2/s at 810, 860and 910C, respectively. Evidence was found for a trap-limited vacancy diffusivity.T.K.Mogi, M.O.Thompson, H.J.Gossmann, J.M.Poate, H.S.Luftman: Applied PhysicsLetters, 1996, 69[9], 1273-5

[446-138/139-108]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe diffusion of δ-function spikes of Sb in thin films which had been grown by solid-phase epitaxy was studied during annealing in vacuum. The results were compared withthe effects of diffusion in films that had been grown by using low-temperature molecularbeam epitaxy. The diffusion temperatures ranged from 750 to 900C, and the 2-dimensional concentrations were between 7 x 1013 and 1.6 x 1014/cm2. The diffusivebehavior of dopants in the solid-phase epitaxial films was qualitatively different to that infilms grown by low-temperature molecular beam epitaxy. This difference was attributedto the vacancy-like defects that were intrinsic to growth by the former method but not togrowth by the latter method. The dopant profiles widened appreciably during solid-phaseepitaxial re-growth; thus making the achievement of δ-function spikes impossible. Aftervacuum annealing, the diffusion coefficients for both n-type and p-type dopants werelower in solid-phase epitaxial films, than in films which had been grown by using theother method, by up to one order of magnitude. The diffused depth profile of the dopant

479


in low-temperature molecular beam epitaxial films exhibited the characteristic deviationfrom a pure Gaussian that was expected, due to the concentration dependence ofdiffusion. That is, it had a flat top and steep shoulders. On the other hand, the dopantdepth profiles in the other material exhibited a central spike and relatively flat shouldersafter diffusion. The width of the central spike was, following an initial transient that wasimpossible to resolve, independent of the diffusion time and temperature. This indicatedthat the solid-phase epitaxial material was defective; with the defects acting as trapsduring diffusion.H.J.Gossmann, A.M.Vredenberg, C.S.Rafferty, H.S.Luftman, F.C.Unterwald,D.C.Jacobson, T.Boone, J.M.Poate: Journal of Applied Physics, 1993, 74[5], 3150-5

[446-106/107-135]


T (C) D (cm2/s)1032 3.3 x 10-15

1027 2.8 x 10-15

982 6.4 x 10-16

972 5.9 x 10-16

897 5.1 x 10-17

850 9.8 x 10-18

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe diffusion of Sb in thin films which had been grown by means of low-temperaturemolecular beam epitaxy was investigated at temperatures ranging from 750 to 900C, andtimes of between 0.25 and 60h. The small spatial extent of the initial δ-function-likedopant profiles permitted the detection of very small diffusional displacements. Thedopant atoms were used as tracers of Si point defects (vacancies and self-interstitials).The diffusion of Sb was found to be enhanced relative to the equilibrium values. A modelwas proposed which was based upon an initial supersaturation of vacancies. Thematching of this model to experimental data permitted the extraction of the vacancydiffusivity, the activation energy for vacancy formation, and the recombination lifetime ofinterstitials. The results showed that the interstitial and vacancy populations could not beconsidered to be independent at low temperatures, as had previously been suggested.H.J.Gossmann, C.S.Rafferty, A.M.Vredenberg, H.S.Luftman, F.C.Unterwald,D.J.Eaglesham, D.C.Jacobson, T.Boone, J.M.Poate: Applied Physics Letters, 1994, 64[3],312-4

[446-115/116-151]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe redistribution, during post-growth annealing, of d-doped Sb with various arealconcentrations was investigated by means of secondary ion mass spectrometry. The

480


diffusion profiles depended upon the initial dopant density, and non-Gaussian profileswere obtained; except in the case of the least-doped sample. The doped layers were stableat up to 550C, regardless of the dopant density. In the case of treatment at temperaturesthat were higher than 700C, the upper limit of the dopant density was found to be 0.01 ofa monolayer. It was also found that differences in crystalline quality did not affect thediffusion of Sb between recrystallized and molecular beam epitaxial layers.S.J.Fukatsu, S.Kubo, Y.Shiraki, R.Ito: Journal of Crystal Growth, 1991, 111[1-4], 843-6

[446-91/92-033]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesThe segregation of Sb at moving crystalline/amorphous (001) interfaces, during thepreparation of d-doped layers, was studied. The use of X-ray reflectivity measurementsrevealed a broadening of the d-doping profile, due to segregation during thecrystallization of amorphous Si. On the basis of ion back-scattering and channelling data,it was deduced that the bulk diffusion coefficients were too low to explain the observedsegregation behavior. The broadening was attributed to enhanced diffusion at theamorphous/crystalline interface. The interfacial diffusion coefficient was at least 2 ordersof magnitude higher than the diffusivity in bulk amorphous material.W.F.J.Slijkerman, P.M.Zagwijn, J.F.Van der Veen, G.F.A.Van de Walle, D.J.Gravesteijn:Journal of Applied Physics, 1991, 70[4], 2111-6

[446-91/92-033]

Bulk Diffusion - Qualitative Observations - Concentration ProfilesDelta-doped layers were made by depositing Sb onto monocrystalline Si, and thendepositing amorphous Si, before performing a final solid-phase epitaxy treatment at620C. After post-annealing at temperatures of between 625 and 725C, Sb precipitateswith a diameter of several nm were observed in the δ-plane by using transmission electronmicroscopy. By using channelling Rutherford back-scattering spectrometry, the increasein the precipitate fraction with time was deduced from the minimum-yield signal. Theresults were explained by using a model which was based on Sb nuclei that grew via thelateral diffusion of Sb atoms, in the δ-plane, followed by incorporation into the nucleus.The generation of nuclei appeared to involve 2 parallel processes. One was the rapidsimultaneous generation of a limited number of nuclei at low-energy sites in the δ-plane,with subsequent diffusion-controlled growth, and the other was the slow continuousgeneration of a larger number of nuclei at random sites in the δ-plane, with subsequentincorporation-controlled growth. The diffusion of Sb at the extremely high concentrationsused here was very fast and concentration-dependent. This could be explained by theMathiot-Pfister model for vacancy-percolation diffusion. The activation energy for theincorporation of Sb atoms into liquid precipitates appeared to be considerably lower thanthat for incorporation into solid ones.C.Van Opdorp, L.J.Van Ijzendoorn, C.W.Fredriksz, D.J.Gravesteijn: Journal of AppliedPhysics, 1992, 72[9], 4047-62

[446-106/107-148]

481


Bulk Diffusion - Qualitative Observations - Concentration ProfilesThermal diffusion of Sb from Sb-doped silicate glass and into porous Si layers wasstudied. The profiles of diffused Sb in the porous layers were fitted to complementaryerror function curves in order to estimate the diffusion coefficients. It was found that thediffusion coefficients for Sb in porous Si were higher, and involved a lower activationenergy, than those for Sb in hydrogenated amorphous Si within the temperature range thatwas investigated.K.Nishimura, Y.Nagao, N.Ikeda: Japanese Journal of Applied Physics 2, 1996, 35[9B],L1145-7

[446-141/142-119]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe effect of lattice defects, produced by Si ion implantation, upon dopant diffusivity wasinvestigated after annealing specimens at temperatures of between 700 and 900C. Thenature and depth of residual implantation defects in undoped samples was determined byanalysing the rocking curves which were obtained by using triple-crystal X-ray diffractionand transmission electron microscopy. As well as interstitial dislocation loops andclusters lying below the original amorphous/crystal interface, epitaxial re-growth of theamorphized Si left a vacancy-rich surface layer and a deeper region which was enrichedin interstitials. These regions corresponded to those for which Monte Carlo simulations ofdefect production predicted an excess of point defects. According to whether the dopantwas associated with vacancy or interstitial clusters, different anomalous diffusionbehaviors were observed. In the deep region where an excess of interstitials was present,only a small enhancement was exhibited by Sb. On the other hand, a slight enhancementin the case of Sb was observed in the surface layer. This was consistent with the differingaccepted contributions which vacancies and interstitials made to the diffusionmechanisms of the dopants.M.Servidori, R.Angelucci, F.Cembali, P.Negrini, S.Solmi, P.Zaumseil, U.Winter: Journalof Applied Physics, 1987, 61[5], 1834-40

[446-60-013]

Bulk Diffusion - Qualitative Observations - Effect of DefectsInitially homogeneously distributed Sb atoms exhibited marked redistribution effects afterimplantation. The results confirmed the effect of point defect gradients upon dopantmigration. It was found that the experimental results agreed well with the predictions ofpair diffusion theories.P.Pichler, R.Schork, T.Klauser, H.Ryssel: Applied Physics Letters, 1992, 60[8], 953-5

[446-88/89-051]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe transient enhanced diffusion of shallow molecular beam epitaxially grown markerlayers of Sb, resulting from deep MeV Si+ ion implantation to doses of about 1016/cm2,was measured. It was expected that the near-surface regions of these implanted sampleswould be vacancy-rich, and transient enhanced diffusion of Sb (a typical vacancydiffuser) was observed. The marked enhancements implied the existence of anappreciable vacancy supersaturation. Double implantation of high-dose MeV-implanted

482


samples, followed by shallow (40keV) Si implantation and annealing, produced a greatlyreduced number of 311 defects; as compared with the 40keV implantation of virginmaterial. This again was considered to be consistent with a vacancy-rich region near tothe surface of MeV-implanted samples.D.J.Eaglesham, T.E.Haynes, H.J.Gossmann, D.C.Jacobson, P.A.Stolk, J.M.Poate:Applied Physics Letters, 1997, 70[24], 3281-3

[446-152-0482]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe oxidation-retarded diffusion of Sb during dry oxidation was investigated attemperatures of 1000 and 1100C. Float-zone [100]-type wafers with a pattern of Si3N4

and a free surface were used to obtain domains, on the same wafer, having normalintrinsic and retarded diffusion. In addition, numerical simulations were performed inorder to determine quantitatively the degree of oxidation-retarded diffusion. The resultswere compared with published values. It was thought to be remarkable that therecombination of self-interstitials and vacancies at 1000C was sufficiently fast to cause asignificant retardation in the diffusion of Sb after only 500min of oxidation.E.Guerrero, W.Jüngling, H.Pötzl, U.Gösele, L.Mader, M.Grasserbauer, G.Stingeder:Journal of the Electrochemical Society, 1986, 133[10], 2181-5

[446-55/56-043]

Bulk Diffusion - Qualitative Observations - Effect of DefectsAn enhanced diffusivity of Sb at very high donor concentrations was found to be relatedto the appearance of a new impurity complex. On the basis of these observations, and thestrong bonding between a vacancy and a donor impurity, a new diffusion mechanism wasproposed in which an Sb-vacancy-donor complex was the mobile species.A.N.Larsen, G.Weyer: Materials Science Forum, 1992, 83-87, 273-8

[446-93/94-058]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe diffusion of Sb, and its dependence upon the Fermi-level position and the structure oflattice defects, was investigated by implanting high (2 x 1016 or 5 x 1016/cm2) doses of Sbions. The donor concentration was strongly increased by subsequent pulsed-laserannealing. For comparison, laser annealing of samples which were co-implanted with thesame doses of Sb and B was carried out in order to obtain strong electrical compensation.The diffusion heat treatment of Sb- and Sb+B-implanted wafers was performed at arelatively low temperature (600C, 1h). Contrary to the extrapolated predictions ofprevious work, a significant shift in the dopant concentration profiles was observed, aswell as the formation of Sb precipitates, Sb-vacancies and/or Sb-B pairing. In order toexplain the diffusivity data, a diffusion coefficient had to be assumed which wasindependent of the dopant concentration and was 7 orders of magnitude higher than thatpreviously determined. This suggested that the increased Sb diffusivity was not duemainly to the Fermi-level position. The huge increase in the Sb diffusivity was attributedto the defects (twins, dislocations, rod-like defects, precipitates, dopant complexes) whichwere revealed by extended X-ray-absorption fine-structure, Rutherford back-scatteringspectrometry or channelling, and transmission electron microscopic techniques. An

483


anomalously high tensile strain in the samples indicated a marked incorporation ofvacancies. These defects were also held to be responsible for a backwards diffusion andan out-diffusion of the dopant which occurred during thermal annealing of most of thesamples.A.Armigliato, F.Romanato, A.Drigo, A.Carnera, C.Brizard, J.R.Regnard, J.L.Allain:Physical Review B, 1995, 52[3], 1859-73

[446-123/124-186]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe diffusion of Sb at concentrations which were close to its solid solubility was studiedby using iso-concentration experiments. The samples were grown, by using molecularbeam epitaxy, so as to have a constant 121Sb background doping and a 123Sb spike withinthis background. Diffusion was then monitored as a function of the Sb backgroundconcentration, at 872 and 1019C, by means of secondary ion mass spectrometry,differential Hall resistometry and transmission electron microscopy. Interstitial-typedislocation loops and Sb precipitates were observed at concentrations which exceeded thesolid solubility. At these concentrations, the diffusivity decreased with increasing Sbbackground concentration. At concentrations which were below the solid solubility andthe intrinsic carrier concentration, at a diffusion temperature of 1019C, the diffusivityincreased with increasing Sb background dopant content. This behavior was explained interms of mobile Sb2V complexes.A.N.Larsen, P.Kringhøj, J.L.Hansen, S.J.Shiryaev: Journal of Applied Physics, 1997,81[5], 2173-8

[446-148/149-187]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe effects of implantation damage upon Sb diffusion during rapid thermal annealing andconventional heat treatment at 750 to 900C were investigated. A comparison of thedopant profiles and junction depths in damaged and undamaged regions demonstrated thatSi+ implantation under the typical conditions used for pre-amorphization produced only anegligible increase in observed Sb transmission reactions.A.Jacques, A.George, X.Baillin, J.J.Bacmann: Philosophical Magazine A, 1987, 55[2],165-81

[446-51/52-130]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe local atomic environment of Sb dopants, in samples which had been implanted to 2 x1016 or 5 x 1016/cm2, was studied by using near-grazing incidence fluorescence extendedX-ray absorption fine structure techniques at various stages of the Sb deactivationprocess. Annealing was performed, at temperatures of between 900 and 1000C, for timesof 30s to 4h. The occurrence of Sb out-diffusion, and a high volume of Sb precipitates,were revealed; especially in the case of samples which had been implanted only with Sb.A comparison of Sb and B co-diffusion data, with corresponding results for the diffusionof Sb alone, revealed several anomalous effects that were due to dopant interaction. Asimulation program which took account of dopant precipitation and donor-acceptor

484


pairing permitted the prediction of most of the anomalous phenomena that occurred inhigh-concentration co-diffusion experiments.C.Revenant-Brizard, J.R.Regnard, S.Solmi, A.Armigliato, S.Valmorri, C.Cellini,F.Romanato: Journal of Applied Physics, 1996, 79[12], 9037-42

[446-136/137-128]

Bulk Diffusion - Qualitative Observations - Effect of Ion ImplantationThe diffusion of Sb in a Si-on-insulator structure, formed by heavy doses of O-implantation, was compared with diffusion in normal bulk wafers. It was found that thenon-equilibrium effects upon diffusion during direct nitridation or nitridation of a SiO2

layer were mitigated in the present structure. Under oxidizing conditions, there were nolarge differences in the diffusion behavior of bulk and Si-on-insulator samples.P.Fahey, S.Solmi: Journal of Applied Physics, 1986, 60[12], 4329-32

[446-51/52-136]

Bulk Diffusion - Qualitative Observations - Effect of StressThe effect of strain upon vacancy-mediated diffusion was investigated by comparingdiffusivities in tensile-strained and relaxed Si and in compressively strained and relaxedSi0.91Ge0.09. It was found that the diffusivity was enhanced by a compressive strain andwas retarded by a tensile strain. A convincing degree of agreement was found between thepresent trends, and the predictions of total-energy calculations. It was possible todetermine changes in activation energy per unit strain for compressive and tensile strains.It was also demonstrated that there was an appreciable enhancement, due to chemicaleffects, in Si0.91Ge0.09. However, contrary to interstitialcy-mediated diffusion, thischemical effect could not alone account for the enhancement. This implied that strain hada greater effect upon vacancy-mediated diffusion than upon interstitialcy-mediateddiffusion.P.Kringhøj, A.N.Larsen, S.J.Shirayev: Physical Review Letters, 1996, 76[18], 3372-4

[446-134/135-165]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsIt was demonstrated that, during the formation of Pd2Si, point defects were injected intothe Si substrate and led to an appreciable degree of dopant diffusion at temperatures aslow as 200C. This was the lowest temperature at which dopant diffusion had ever beenobserved. Buried marker layers exhibited asymmetrical diffusion, with migrationoccurring preferentially towards the silicide-forming interface. The results showed that alarge gradient in point defect concentration did not develop across the width of a markerlayer. This showed that it was not possible for a large point defect gradient to be createdby trapping as they diffused through the doped layers.M.Wittmer, P.Fahey, G.J.Scilla, S.S.Iyer, M.Tejwani: Physical Review Letters, 1991,66[5], 632-5

[446-84/85-065]

Bulk Diffusion - Qualitative Observations - Effect upon DefectsAn investigation was made of the suggestion that P diffusion at concentrations above thesolid solubility generated Si self-interstitials. Buried layers of Sb were created byimplanting 150keV Sb into [100]-type substrates, to 5 x 1013/cm2. After annealing (900C,

485


0.5h) in N, an 0.008 to 0.01mm thick epilayer was grown over the buried layers. Maskingoxides were then created by liquid-phase chemical vapor deposition of 0.001mm SiO2

layers, or by thermal oxidation at 900C in steam. Spreading resistance profiles wererecorded and secondary ion mass spectrometry was used to characterize the chemical Pdoping densities. Plan-view and cross-sectional transmission electron microscopy wasused to look for SiP precipitates and defects in the P-diffused layers. It was found that P-related precipitates, 0.0001 to 0.0002mm in size, were observed in some samples by usingtransmission electron microscopy. In every case, their concentration was insufficient toaccount for the concentration of non-electrically active P. Measurably retarded diffusionof Sb in the buried layers occurred at temperatures ranging from 1100 to 1200C, with anactivation energy of about 6.6eV. Stacking fault growth in buried Sb layers under the P-diffused region implied that Si self-interstitial supersaturation was produced in the P-doped layer. Self-interstitials fed stacking fault growth and retarded Sb diffusion. Thenumber of self-interstitials generated could not be accounted for by the sparse number ofprecipitates which formed in the P-diffused region.J.C.C.Tsai, D.G.Schimmel, R.B.Fair, W.Maszara: Journal of the Electrochemical Society,1987, 134[6], 1508-18

[446-55/56-035]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionDiffusion experiments were performed in the presence of lattice defects which had beenproduced by Si ion implantation. The effects of transient enhanced diffusion were studiedby means of bevelling and staining measurements of implanted samples, and secondaryion mass spectroscopic determinations of dopant profiles. The annealing temperatures fordoped implanted specimens ranged from 700 to 1100C and this treatment was performedusing an electron beam. The low temperatures which were used permitted the kinetics ofanomalous diffusion to be monitored. It was ascertained that the enhanced diffusioncoefficient was almost constant during a period which decreased with increasingtemperature. It then tended gradually to the equilibrium value. This trend agreed with thatof lattice damage changes which were revealed by double crystal X-ray analyses of therocking curves of implanted samples. The secondary ion mass spectroscopy profilesindicated that only a fraction of the dopant which was located at residual implantationdamage was responsible for the anomalous diffusion.R.Angelucci, F.Cembali, P.Negrini, M.Servidori, S.Solmi: Journal of the ElectrochemicalSociety, 1987, 134[12], 3130-4

[446-60-013]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionTantalum silicide, when deposited directly onto monocrystalline Si substrates andannealed at 950C, led to the enhanced diffusion of Sb in buried layers. The effect wasattributed to vacancy supersaturation since it was known that Sb diffused almost entirelyvia a vacancy mechanism. The occurrence of Sb-enhanced diffusion contrasted with apreviously reported nitridation effect upon diffusion. The fact that the enhanced diffusionoccurred in buried layers excluded a snow-plough mechanism. The Si/Ta ratio in thesputter-deposited silicide was slightly less than 2. It was suggested that further silicidationgenerated vacancies by removing Si atoms from the Si substrate. Enhanced diffusion was

486


not detected when there was a 150nm intermediate layer of polycrystalline Si filmbetween the silicide and the monocrystalline Si substrate. This indicated thatpolycrystalline Si was an effective sink for excess vacancies; perhaps more than forexcess interstitials.S.M.Hu: Applied Physics Letters, 1987, 51[5], 308-10

[446-55/56-037]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe effect of a thermally grown silicon nitride film during annealing in Ar at 1100C wasstudied. An enhanced diffusivity of Sb was detected, but this behavior disappeared whenthe nitride was removed before annealing in Ar. It was suggested that the enhanceddiffusivity was not directly related to the growth of the nitride, but was instead due tostresses which were present in the film.S.T.Ahn, H.W.Kennel, J.D.Plummer, W.A.Tiller: Applied Physics Letters, 1988, 53[17],1593-5

[446-62/63-235]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe diffusion of ion-implanted Sb in single crystals was studied, as a function of the Pdonor concentration, during rapid thermal annealing at 1000 or 1050C. Depth profileswere determined by using Rutherford back-scattering spectroscopy and Mossbauertechniques. When the extrinsic/intrinsic carrier concentration ratio was greater than 20,extremely large diffusion coefficients were found. The value of the diffusion coefficientwas approximately proportional to the 4th power of the extrinsic/intrinsic carrierconcentration ratio. The onset of such a dependence was associated with the appearanceof a new defect complex which contained Sb.P.E.Andersen, A.N.Larsen, P.Tidemand-Petersson, G.Weyer: Applied Physics Letters,1988, 53[9], 755-7

[446-62/63-236]

Bulk Diffusion - Qualitative Observations - Enhanced DiffusionThe effects of Co and Ti thin-film reactions with Si, upon the diffusion of buried Sb-doped layers, were investigated. Analysis of Sb profiles by means of secondary ion massspectrometry showed that greatly enhanced non-uniform Sb diffusion occurred during thereaction of various thicknesses (30 to 300nm) of Co and Ti via rapid thermal annealing. Asimple non-equilibrium intrinsic diffusion model was used to estimate the time-averagedexcess vacancy concentrations. Concentrations which were about 107 times theequilibrium values were shown to exist during CoSi2 formation via the reaction of a 30nmCo film in Ar at 700C for 600s. Diffusion enhancement at large distances from thesilicide edge was observed by using bevelling and etching techniques.J.W.Honeycutt, G.A.Rozgonyi: Applied Physics Letters, 1991, 58[12], 1302-4

[446-81/82-053]

Bulk Diffusion - Theoretical Analysis - Activation EnergyA new model was proposed, for the diffusion of impurities in amorphous Si, which tookaccount of its structural and electronic properties. The model was based upon the many-body kinetic theory of thermally activated rate processes in solids. The low activation

487


energies which were commonly observed, as well as their dependence upon the impurityconcentration, were explained.J.L.Khait, R.Brener, R.Beserman: Physical Review B, 1988, 38[9], 6107-12

[446-62/63-228]

Bulk Diffusion - Theoretical Analysis - Effect of ChargeExperiments show that ion pairing has a marked effect upon the diffusion of oppositelycharged impurities. An analysis of literature data was used to deduce the ion pairingcoefficients for n-type impurities with B and In. A coefficient with the value of 0.17/Nwas found to describe the pairing case of B-Sb, where N was the intrinsic electronconcentration. In the case of In-Sb, the coefficient was about an order of magnitudesmaller. It was suggested that the paired ions occupied adjacent substitutional sites; with asmall perturbation in the Coulomb binding which arose from elastic effects.N.E.B.Cowern: Applied Physics Letters, 1989, 54[8], 703-5

[446-64/65-176]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA comprehensive investigation of dopant diffusion in the presence of equilibrium andnon-equilibrium concentrations of intrinsic point defects was presented. It was found that,under equilibrium conditions, vacancies and interstitials mediated the diffusion of thedopant. The interstitial component had a large activation energy. Relevant expressionswere derived for the activation energies, of various diffusion and injection mechanisms,under non-equilibrium conditions (such as those produced by the injection of excess pointdefects). For oxidation conditions, the calculated values were in excellent agreement withavailable experimental data. Both theory and experiment suggested that the concertedexchange mechanism, which involved no point defects, played only a minor role indopant diffusion.C.S.Nichols, C.G.Van de Walle, S.T.Pantelides: Physical Review B, 1989, 40[8], 5484-96

[446-72/73-041]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsThe mechanisms of electrically active impurity diffusion in heavily doped material duringrapid thermal annealing were considered. Monte Carlo computer simulations of impurityatom migration in the lattice were carried out with regard to collective phenomena. On thebasis of a comparison with experimental data on the rapid thermal annealing of implantedSb at temperatures ranging from 1000 to 1200C, conclusions were drawn concerning thenature of the interaction between impurity atoms and defects. Good agreement betweencalculated and experimental results was obtained by modifying the model with regard tothe non-equilibrium vacancy distribution in the lattice.S.A.Fedotov: Physica Status Solidi B, 1994, 186[2], 375-82

[446-119/120-225]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsDopant and self-diffusion were known to be governed by both self-interstitials andvacancies; as normalized to their thermal equilibrium values. Since these point defectconcentrations were too low to be investigated directly, the diffusion mechanisms had tobe considered to be unknown. After substituting experimental results on Sb diffusion at

488


1100C into a new equation that had been derived from the dopant diffusion equation, theequation was solved at the same time as the equation for oxidation-induced stacking faultsunder conditions of local equilibrium between interstitials and vacancies. In this way, theinterstitial and vacancy concentrations were obtained as functions of the diffusion time,and the fractional components of the interstitialcy mechanism for Sb diffusion, could bedetermined. It was found that Sb diffusion was governed mainly by a vacancymechanism.T.Okino, R.Takaue, M.Onishi: Materials Science Forum, 1995, 196-201, 1631-6

[446-127/128-155]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsThe effect of point defects upon dopant diffusion was considered. Their influence wasanalyzed by comparing experimental data with a simulation. The experiments concernedthe oxidation-retarded diffusion of Sb in float-zone material at 1100C. The simulationwas carried out by using a physical model without simplifying assumptions, and it wasshown that such assumptions were not admissible. A reasonable set of parameters wasdeduced from the analysis. Since each parameter represented a physical effect,information about the importance of the bulk and surface recombination of point defects,and about the equilibrium concentration values and diffusion coefficients of diffusingspecies, could be obtained. It was found that the effect of the surface played a decisiverole in the distribution of point defects.T.Brabec, E.Guerrero, M.Budil, H.W.Poetzl: Zeitschrift für Physik B, 1987, 67[4], 415-20

[446-55/56-043]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsA thermodynamic formalism was developed for the predominant point defect mechanismof self-diffusion and impurity diffusion in Si. It was used to provide a rigorous basis forthe point-defect based interpretation of diffusion results for biaxially strained epitaxiallayers in the present system. It was found that a certain combination of the hydrostaticand biaxial stress dependences of the diffusivity was equal to ±1 times the atomicvolume; depending upon whether the predominant mechanism involved vacancies orinterstitials. Experimental results for Sb diffusion in biaxially strained Si-Ge films, andfirst-principles calculations of the activation volume for Sb diffusion via a vacancymechanism, were shown to be quantitative agreement (with no free parameters).M.J.Aziz: Applied Physics Letters, 1997, 70[21], 2810-2

[446-152-0488]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsEquations for oxidation-enhanced and oxidation-retarded diffusion, and for oxidation-induced stacking faults were solved simultaneously by using experimental results whichhad been obtained at 1100C. A simple relationship between the concentrations of self-interstitials and vacancies was assumed in order to obtain the solutions. It was concludedthat the product of the concentrations of self-interstitials and vacancies was almost equalto the value at thermal equilibrium, and that the fractional component of the interstitialcymechanism for Sb diffusion was smaller than 0.5. This showed that the growth of

489


oxidation-induced stacking faults was caused mainly by vacancy undersaturation, and thatoxidation-retarded diffusion of Sb occurred.M.Yoshida, S.Matsumoto, Y.Ishikawa: Japanese Journal of Applied Physics, 1986, 25[7],1031-5

[446-48-064]

Bulk Diffusion - Theoretical Analysis - Effect of OxidationThe oxidation-enhanced or oxidation-retarded diffusion of substitutional impurities isusually expressed in terms of equations which involve self-interstitial (I) and vacancy (V)concentrations. By using the asymptotic forms of the solutions to these equations, anapproximate relationship between I and V was obtained. Identification of thisapproximate relationship with that for local equilibrium between I and V gave a simplemethod for obtaining mathematically self-consistent solutions to the oxidation-enhancedand oxidation-retarded diffusion equations. In this way, I and V were obtained asfunctions of the diffusion time. The fractional component of the interstitialcy mechanismfor the oxidation-retarded diffusion of Sb was obtained. The value of d1

Sb was 0.029 at1100C.T.Okino: Japanese Journal of Applied Physics 1, 1992, 31[4], 965-9

[446-88/89-055]

Bulk Diffusion - Theoretical Analysis - Effect of OxidationOxidation-enhanced and oxidation-retarded diffusion effects were incorporated intodifferential equations for self-interstitials and vacancies. An extended relationship for thelocal equilibrium between interstitials and vacancies, VIm = n, was derived fromasymptotic forms of the solutions to the oxidation-enhanced and oxidation-retardeddiffusion equations. It was found that local equilibrium between interstitials andvacancies did not occur within the present range of diffusion times. Interstitial andvacancy diffusivities of 2.4 x 10

-9 and 2.1 x 10

-10cm

2/s, respectively, at 1100C were

deduced from the results. The corresponding thermal equilibrium concentrations were 3.3x 10

16 and 2.2 x 10

17/cm

3.

T.Okino, M.Onishi: Japanese Journal of Applied Physics 1, 1994, 33[6A], 3362-7[446-113/114-048]

Bulk Diffusion - Theoretical Analysis - Effect of PressureFirst-principles total-energy calculations of the atomic diffusion of group-V impuritiesrevealed an effect of pressure upon the activation energy for diffusion. In the case of thevacancy mechanism, the activation energy for Sb decreased with pressure. In the case ofthe interstitial mechanism, the formation energy of the interstitial impurity exhibited ageneral tendency to increase with pressure. The microscopic origin of the pressuredependence was explained in terms of the local strains around defects. The negativepressure dependence which was common to the vacancy-mediated diffusion of group-Vimpurities could be explained by the peculiar properties of the isolated vacancy. Theseincluded a breathing distortion of surrounding Si atoms towards the vacancy site, givingrise to a tensile strain around the vacancy, and lattice distortions which originated fromthe vacancy and caused weak vacancy-impurity interactions. The positive pressuredependence of interstitial-mediated diffusion was closely related to the atomic structures

490


of the interstitial impurities, which produced compressive strains in the surrounding Si-Sibonds.O.Sugino, A.Oshiyama: Physical Review B, 1992, 46[19], 12335-41

[446-106/107-131]

Table 82Grain Boundary Diffusion of Sb in Si

T (C) D (cm2/s)1150 1.7 x 10-8

1100 5.9 x 10-9

1050 3.3 x 10-9

1000 1.1 x 10-9

930 1.6 x 10-10

Bulk Diffusion - Theoretical Analysis - Enhanced DiffusionA simple model was developed in order to explain the enhanced diffusion of Sb whichhad been implanted at high temperatures. It was found that the concentration profilescould be predicted by using a system of reaction-diffusion equations which described thediffusion and decay of the impurity-vacancy pairs which were created duringimplantation. Good agreement with experiment was obtained.E.Antoncik: Radiation Effects and Defects in Solids, 1993, 127[1], 75-82

[446-113/114-048]

Bulk Diffusion - Theoretical Analysis - Percolation ModelThe physical origin of the diffusion enhancement which had often been observed at highdopant levels, during the annealing of dopant-implanted material, was considered. It wasshown here that the experimental results could be accounted for in terms of a percolationmodel which had previously been developed for high-concentration P diffusion. Aquantitative treatment was presented for the case of the diffusion of implanted Sb inheavily doped n-type Si.D.Mathiot, J.C.Pfister: Journal of Applied Physics, 1989, 66[2], 970-2

[446-74-055]

382 Grain Boundary Diffusion - Quantitative DataThe diffusion of donor elements in fine-grained and coarse-grained polycrystallinematerial was studied at temperatures ranging from 900 to 1150C (table 82). Tracers(125Sb) were used to determine the concentration/depth profiles via sectioning. By meansof autoradiography, the lateral distribution of the radiotracers over the sample surface wasrevealed. The grain boundary diffusivity could be described by:

D (cm2/s) = 380 exp[-2.9(eV)/kT]F.H.M.Spit, H.Bakker: Physica Status Solidi A, 1986, 97[1], 135-42

[446-48-063]

491

Sb Surface Diffusion/Bulk Diffusion Sc

Surface Diffusion - Quantitative DataA scanning tunnelling microscopy method for studying surface diffusion was based uponmeasurements of the displacement distribution of adsorbates via so-called image-anneal-image cycles. This permitted direct observation of the diffusion process, while avoidingpotential scanning tunnelling microscope tip effects. The method was used to study theanisotropic diffusion of Sb dimers on Si(001). The energy barrier (1.2eV) and theprefactor (0.0001cm

2/s) for the faster diffusion across the substrate dimer rows were

measured. It was noted that the diffusion which was observed by using the so-calledimage-while-hot method appeared to be almost isotropic, and much faster. It was shownthat this discrepancy was due to an influence of the tip, which interfered in the lattermethod.Y.W.Mo: Physical Review Letters, 1993, 71[18], 2923-6

[446-106/107-147]

Sc

Figure 19: Diffusivity of Sc in Si (see table 83)

1.0E-13

1.0E-12

1.0E-11

6 7 8

6.576.797.037.28

104/T(K)

D (cm2/s)

492

Sc Bulk Diffusion Se

383 Bulk Diffusion - Quantitative DataA source, in the form of the 46Sc-labelled chloride, was deposited onto Si plate samples toa depth of about 0.35mm. Diffusion annealing was then carried out in air (1100 to 1250C,5 to 30h). The resultant Sc profile was determined by using etching and ?-ray counting.The diffusion coefficient was deduced by fitting the profile to the complementary errorfunction. It was found that, with increasing temperature, the diffusivity of Sc increasedfrom about 1.4 x 10-13 to about 1.9 x 10-12cm2/s. Overall, the results (table 83) could bedescribed by the expression:

D(cm2/s) = 0.08 exp[-3.2(eV)/kT]G.K.Azimov, S.Zainabidinov, D.E.Nazyrov: Fizika i Tekhnika Poluprovodnikov, 1989,23[3], 556-7 (Soviet Physics - Semiconductors, 1989, 23[3], 347)

[446-72/73-053]

Table 83Diffusivity of Sc in Si

T (C) D (cm2/s)1250 1.8 x 10-12

1200 1.2 x 10-12

1150 4.5 x 10-13

1100 1.4 x 10-13

SeBulk Diffusion - Qualitative Observations and Point DefectsInteractions between this group-VI element and fast-diffusing impurities (Cr, Fe, Mn)were studied. Infra-red absorption, electron spin resonance, and neutron activation studiesshowed that the group-VI element effectively interacted with each other or with the fast-diffusing impurities. It was found that, for each pairing of a group-VI element with a fast-diffusing impurity, there was a certain annealing temperature at which they interactedmost efficiently. A definite correlation was established between this annealingtemperature and the thermodynamic Gibbs free energy (at 298K) of the correspondingcompounds. This correlation suggested that the interaction process mainly involved theformation of electrically neutral chemically bonded complexes by substitutional group-VIelement centers and interstitial fast-diffusant centers.M.K.Bakhadirkhanov, S.I.Askarov, N.Norkulov: Physica Status Solidi A, 1994, 142[2],339-46

[446-117/118-194]

493

Si

Figure 20: Surface Self-Diffusivity of Si

Bulk Diffusion - Quantitative DataDetailed measurements were made of Au concentration profiles in [110]-type wafers of p-type material. The Au was introduced by means of ion implantation, and was diffused attemperatures of between 1073 and 1473K for times ranging from 60s to 100h. It wassuggested that the measured profiles and their time dependence could be explained interms of the kick-out diffusion mechanism. It was found that the diffusivity of Si self-interstitials was described by:

D(cm2/s) = 0.006 exp[-3.3(eV)/kT]

1.0E-13

1.0E-12

1.0E-11

1.0E-10

1.0E-09

1.0E-08

1.0E-07

8 9 10 11 12 13

table 85table 86

104/T(K)

D (cm2/s)

494

Si Bulk Diffusion Si

S.Coffa, L.Calcagno, S.U.Campisano, G.Calleri, G.Ferla: Journal of Applied Physics,1988, 64[11], 6291-5

[446-72/73-042]

384 Bulk Diffusion - Quantitative DataSpecial film structures were grown, by means of low-temperature molecular beamepitaxy, in order to investigate the properties of self-interstitials in Si. It consisted of asuperlattice which was made up of B spikes which were separated from each other by100nm of Si. After dry oxidation, the width of each spike was directly proportional to theinterstitial concentration at that depth. The superlattice as a whole thus furnished a depthprofile of the time-averaged interstitial concentration, and permitted direct determinationsto be made of the diffusion coefficients of interstitials. The abrupt dopant concentrationtransitions which were achievable in low-temperature molecular beam epitaxial filmspermitted this type of investigation to be carried out at temperatures of between 750 and900C. A value of 1.4 x 10

-13cm

2/s was found at 800C. Overall, the results (table 84) could

be described by:D(cm

2/s) = 100 exp[-3.1(eV)/kT]

H.J.Gossmann, C.S.Rafferty, H.S.Luftman, F.C.Unterwald, T.Boone, J.M.Poate: AppliedPhysics Letters, 1993, 63[5], 639-41

[446-106/107-176]

Table 84Diffusivity of Self-Interstitials in Si

T (C) D (cm2/s)900 4.1 x 10-12

850 8.2 x 10-13

800 1.4 x 10-13

750 5.0 x 10-14

Bulk Diffusion - Quantitative DataThe injection of interstitials during the annealing of non-amorphizing Si implants wasmonitored by using sharply defined B-doped marker layers that had been grown by meansof reduced-pressure chemical vapor deposition. The enhancement of B diffusivity whichwas measured during the initial annealing stage (up to 15s, 700C) was at least an order ofmagnitude greater than the enhancement which occurred during subsequent annealing. Itwas concluded that the ultra-fast diffusion set a lower limit, on the Si interstitialdiffusivities (700C), of 2 x 10-10cm2/s.H.G.A.Huizing, C.C.G.Visser, N.E.B.Cowern, P.A.Stolk, R.C.M.De Kruif: AppliedPhysics Letters, 1996, 69[9], 1211-3

[446-138/139-104]

Bulk Diffusion - Quantitative DataA simple method for the estimation of the diffusivity of Si interstitials was described. Attemperatures of between 460 and 500C, O thermal donors were used to monitor Si

495


interstitials. The estimated diffusivity of Si interstitials at O-donor formation temperatureswas fitted to published data on oxidation-enhanced and retarded diffusion experimentswhich had been performed at temperatures above 950C. At temperatures of between 460and 1200C, the diffusivity of Si interstitials could be described by:

D(cm2/s) = 0.355 exp[-1.86(eV)/kT]W.Wijaranakula: Journal of Applied Physics, 1990, 67[12], 7624-7

[446-78/79-057]

Bulk Diffusion - Qualitative Observations - Effect of DefectsIt was pointed out that, when an intrinsic defect (vacancy, self-interstitial) mediated self-diffusion, the activation enthalpy was equal to the sum of the enthalpies for defectformation and defect migration. However, if the migration enthalpy was small and themigration entropy was large then, at some critical temperature, the saddle point becamethe stable configuration and a totally different path could become dominant. Therefore, itwas possible that paths of higher activation enthalpy could predominate at hightemperatures and lead to a curvature of the Arrhenius plot. These effects were very likelyto occur in crystalline Si, and might account for the large entropy observed in high-temperature self-diffusion.S.T.Pantelides: Physical Review B, 1987, 36[6], 3462-4

[446-55/56-044]

Bulk Diffusion - Theoretical Analysis - Activation EnergyThe free energy of the concerted exchange mechanism for self-diffusion was estimated byusing the thermodynamic integration method and Monte Carlo simulations, with aninteratomic potential that had been fitted to local density approximation calculations.Anharmonicity and relaxation were taken fully into account in the calculations, since thephase space was extensively covered by the Monte Carlo simulations. The resultsindicated that the concerted exchange mechanism could make a significant contribution tothe self-diffusion coefficient in Si.A.Antonelli, S.Ismail-Beigi, E.Kaxiras, K.C.Pandey: Physical Review B, 1996, 53[3],1310-4

[446-131/132-185]

Bulk Diffusion - Theoretical Analysis - Activation EnergyA new mechanism for the diffusion of substitutional atoms in semiconductors waspresented. It did not involve vacancies or interstitials. The activation barrier for self-diffusion in Si via this mechanism was calculated to be 4.3eV. This compared well withthe measured activation barrier. The new mechanism was able to account for severalambiguous experimental observations of diffusion in semiconductors. It was suggestedthat this mechanism contributed significantly to diffusion in semiconductors.K.Pandey: Materials Science Forum, 1986, 10-12, 121-6

[446-49-029]

Bulk Diffusion - Theoretical Analysis - Activation EntropyThe entropy of Si self-diffusion during the concerted exchange process was calculated.The Vineyard transition-rate approach, and an accurate first-principles quantum-

496


mechanical energy surface, were used. The high-temperature value of the entropy wasestimated to be 3.3k. This was 2 or 3 times lower than the experimentally determinedvalue. This relatively high entropy of the concerted exchange mechanism was attributedto the existence of multiple equivalent exchange paths, and to a combination of a stiffmode at the equilibrium configuration with a soft mode at the saddle-point configurations.K.C.Pandey, E.Kaxiras: Physical Review Letters, 1991, 66[7], 915-8

[446-84/85-074]

Bulk Diffusion - Theoretical Analysis - Activation EntropyThe concerted exchange mechanism is currently the only viable candidate, for atomicexchange, that does not involve intrinsic lattice defects. In order to evaluate itscontribution to Si self-diffusion, the entropy which was associated with concertedexchange was calculated. Full relaxation of the energy surface was included via aninteratomic potential. The predominant contribution of vibrational modes was taken intoaccount via the harmonic approximation. This gave an entropy of 6.3k; which agreed withthe observed range of experimental results. It was concluded that the concerted exchangemechanism was a significant contributor to self-diffusion in Si.E.Kaxiras, K.C.Pandey: Physical Review B, 1993, 47[3], 1659-62

[446-106/107-148]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsDopant and self-diffusion were known to be governed by both self-interstitials andvacancies; as normalized to their thermal equilibrium values. Since these point defectconcentrations were too low to be investigated directly, the diffusion mechanisms had tobe considered to be unknown. A new equation that had been derived from the dopantdiffusion equation was solved at the same time as the equation for oxidation-inducedstacking faults under conditions of local equilibrium between interstitials and vacancies.In this way, the interstitial and vacancy concentrations were obtained as functions of thediffusion time, and the fractional components of the interstitialcy mechanism for self-diffusion could be determined. It was found that self-diffusion was governed mainly by avacancy mechanism.T.Okino, R.Takaue, M.Onishi: Materials Science Forum, 1995, 196-201, 1631-6

[446-127/128-155]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsThe first parameter-free calculations of self-diffusion constants in Si were reported. Theconstants were calculated for defect-mediated mechanisms by using the local-densityapproximation, together with ab initio molecular dynamics simulations. The diffusionconstant for the concerted exchange mechanism was obtained from earlier results. Withinthe range of experimental values, the self-interstitial mechanism predominated over thecontributions of other mechanisms.P.E.Blöchl, E.Smargiassi, R.Car, D.B.Laks, W.Andreoni, S.T.Pantelides: PhysicalReview Letters, 1993, 70[16], 2435-8

[446-099/100-099]

497


Bulk Diffusion - Theoretical Analysis - Effect of DopantSelf-interstitial diffusivities can be deduced from the diffusive behavior of metals such asAu in an inert annealing ambient, or from the diffusion of dopant markers such as B underoxidizing conditions. Each type of experiment yields fairly consistent results. However,the interstitial diffusivities which are obtained by using these 2 methods differ greatly.Moreover, the marker layer experiments depend upon the assumption that the presence ofthe dopant does not disturb the diffusion of the interstitials. The validity of thatassumption was explored here. A model for interstitial diffusion in the presence of B wasdeveloped, and 2 extreme cases of the B-interstitial interaction strength were considered.The predictions of the model were compared with experimental data on oxidation-enhanced diffusion in B-doped superlattices. On the basis of this comparison, it wasconcluded that the trapping of interstitials by B atoms in the markers could not beresponsible for the differing values of the Si interstitial diffusivity which were reported inthe literature. It was also shown that the presence of the dopant did not perturb thebehavior of Si self-interstitials in a doped superlattice.H.J.Gossmann, G.H.Gilmer, C.S.Rafferty, F.C.Unterwald, T.Boone, J.M.Poate,H.S.Luftman, W.Frank: Journal of Applied Physics, 1995, 77[5], 1948-51

[446-121/122-091]

Bulk Diffusion - Theoretical Analysis - Effect of PressureThe effect of hydrostatic pressures upon the energy of self-diffusion was investigated byusing parameter-free total energy calculations. It was found that the vacancy, interstitial,and concerted exchange mechanisms (which had similar activation energies) exhibiteddiffering pressure dependences. It was also shown that, unlike the (111) surface, thenearest neighbors of the Si vacancy relaxed inwards rather than outwards.A.Antonelli, J.Bernholc: Physical Review B, 1989, 40[15], 10643-6

[446-72/73-053]

Bulk Diffusion - Theoretical Analysis - Effect of TrappingPublished values of the diffusivity of the Si interstitial or interstitialcy range over severalorders of magnitude indicate activation energies of between 1 and 4eV. A model wasproposed for the effect of bulk trapping effects upon the interstitialcy diffusivity. Thisprovided a consistent explanation for the observed discrepancies. It reconciled the effectsof different materials (float-zone, Czochralski, and epitaxial) and processes (diffusion,gettering) upon the apparent value of the interstitialcy diffusivity. New experimentalresults which directly indicated substantial bulk effects in various types of Si supportedthe validity of the model.P.B.Griffin, S.T.Ahn, W.A.Tiller, J.D.Plummer: Applied Physics Letters, 1987, 51[2],115-7

[446-55/56-043]

Bulk Diffusion - Theoretical Analysis - Effect upon DefectsIt was noted that both C and Si interstitial atoms were mobile, and became trapped atother defects (thus altering their properties) or displaced impurity atoms into interstitialsites. An approximate molecular orbital procedure, which incorporated Car-Parrinello likedynamics, was used to model the structure of Si and C interstitials. By displacing them

498

Si Bulk/Grain Boundary Diffusion Si

along their possible migration paths, and by allowing them to relax back into their stablesites, the details of low-energy migration routes were calculated. It was found that theneutrally charged Si interstitial occupied a (110) split site, but with the axis displaced(along the z-direction) away from the nearest pair of Si neighbors. Negative and singlypositively charged interstitials occupied the same sites but, when doubly positivelycharged, the hexagonal sites were more stable. The migration path of the neutral Siinterstitial involved a complex route, with an activation energy of 0.2eV.A.Mainwood: Materials Science Forum, 1995, 196-201, 1589-94

[446-127/128-155]

Bulk Diffusion - Theoretical Analysis - Effect upon DefectsThe possible role which was played by Pandey's concerted exchange mechanism in Sidiffusion and deformation was considered with regard to atomistic computer calculationsby using a form of interatomic total energy potential which had been devised andmodified by Tersoff. It was concluded that, if the concerted exchange was viable for self-diffusion, then a similar mechanism should exist for the creation of edge dipoles in high-temperature deformed Si, and in Si near to its melting point. It was shown that ametastable state which was created during concerted exchange was the smallestconceivable faulted edge dipole.M.I.Heggie: Philosophical Magazine Letters, 1988, 58[2], 75-80

[446-61-085]

Bulk Diffusion - Theoretical Analysis - Potential FunctionIt was shown that existing classical potentials were not suitable for calculating the energyof realistic atomic processes. A new potential was presented which was especially suitedfor the simulation of processes in a diamond cubic lattice rather than in the high-energybulk structures of Si. The potential was based upon a very large quantum mechanical database. It consisted of 2-body and 3-body terms with short-range separable forms, andaccurately reproduced the energy surface for atomic exchanges in Si. It was thus ideallysuited to the molecular dynamics simulation of atomic processes in this material.E.Kaxiras, K.C.Pandey: Physical Review B, 1988, 38[17], 12736-9

[446-64/65-180]

Grain Boundary Diffusion - Qualitative Observations - Effect of ElectromigrationRapid migration of Si atoms along the grain boundaries of electron beam annealedpolycrystalline material (in a Si-on-insulator island structure with a Si3N4 and SiO2 cap)was found. An appreciable amount of mass transport was observed in the structure attemperatures which were high enough to melt the grain boundaries but not the bulkmaterial. The results supported the suggestion that the electromigration of Si4+ ions, inmolten Si at the grain boundaries, was responsible for the rapid migration of Si atoms.Y.Hayafuji, T.Yanada, A.Shibata, S.Kawado: Journal of Applied Physics, 1989, 66[8],3579-84

[446-74-055]

499

Si Surface Diffusion Si

Surface Diffusion - Quantitative DataThe migration of Si on the (001) surface was investigated by analyzing the numberdensity of islands which was formed during deposition. By comparing the data with thepredictions of various models, it was deduced that diffusion in the fast direction (alongsurface dimer rows) was described by:

D(cm2/s) = 0.001 exp[-0.67(eV)/kT]The associated model involved the assumption of anisotropic bonding and a 1:1000diffusion anisotropy.Y.W.Mo, J.Kleiner, M.B.Webb, M.G.Lagally: Physical Review Letters, 1991, 66[15],1998-2001

[446-84/85-075]

Surface Diffusion - Quantitative DataThe development of periodic atomic step arrays, associated with etched grating structureson Si(001) during annealing, was monitored by using optical and scanning tunnellingmicroscopic methods. It was found that the grating amplitudes decayed exponentiallywith time at temperatures ranging from 800 to 1100C; with characteristic decay constantsthat scaled approximately as the inverse fourth power of the grating period. This indicatedthe predominance of surface diffusion as the mass transport mechanism. The activationenergy for Si surface self-diffusion was about 2.3eV, and the pre-exponential factor wasabout 0.1m2/s. The experimental data were consistent with an adatom transfer process.The details of the atomic step morphologies of the grating structures were described, andinteractions of the atomic steps during decay were related to curvature-dependent drivingforces for mass transfer.M.E.Keeffe, C.C.Umbach, J.M.Blakely: Journal of the Physics and Chemistry of Solids,1994, 55[10], 965-73

[446-119/120-226]

Surface Diffusion - Quantitative DataThe diffusion of Si dimers on the (001) surface, at temperatures of between ambient and128C, was measured by using a novel atom-tracking technique that could resolveindividual diffusion events by using lateral positioning feedback to lock the scanningtunnelling microscope probe into position above selected atoms to within a small fractionof a nanometre. When locked, the scanning tunnelling microscope tracked the atoms asthey migrated over the crystal surface. By tracking individual atoms directly, the ability tomeasure dynamic events was increased by a factor of about 1000 with respect toconventional scanning tunnelling microscopic imaging techniques. The results could bedescribed by an attempt frequency of about 1013/s and an activation energy of 0.94eV.B.S.Swartzentruber: Physical Review Letters, 1996, 76[3], 459-62

[446-134/135-165]

385 Surface Diffusion - Quantitative DataThe migration of Si adsorbates on a clean (001) surface was investigated by means ofreflection electron microscopy. It was found that, when the sample was heated by usingdirect current, denuded zones with no observable Si islands were created at the terraceedges of the surface. The diffusivities, parallel and perpendicular to the surface dimer

500


(table 85), were deduced from the denuded zone widths at temperatures ranging from 500to 850C.T.Doi, M.Ichikawa, S.Hosoki, K.Ninomiya: Japanese Journal of Applied Physics 1, 1996,35[5A], 2770-3

[446-136/137-131]

Table 85Diffusivity of Si on (001) Si Surfaces

T (K) Terrace D (cm2/s)773 1 x 2 4.7 x 10-13

823 1 x 2 1.8 x 10-12

873 1 x 2 6.1 x 10-12

923 2 x 1 3.2 x 10-12

923 1 x 2 1.7 x 10-11

973 2 x 1 7.6 x 10-12

1023 2 x 1 1.5 x 10-11

1023 1 x 2 8.6 x 10-11

1073 2 x 1 6.8 x 10-11

1073 1 x 2 3.6 x 10-10

1123 2 x 1 2.7 x 10-10

Surface Diffusion - Quantitative DataThe kinetics of vacancy diffusion on (111) surfaces was studied by means of scanningreflection electron microscopy. Two types of layer-by-layer etching were observed during500eV Ar-ion bombardment at high substrate temperatures. One was step-retreat (reversalof step-flow growth), and the other was 2-dimensional vacancy island nucleation. Theresults showed that vacancies, which were created by low-energy ion impact, diffused onthe surfaces and were annihilated at step edges. The vacancy diffusion kinetics on thesurface were also examined by means of scanning reflection electron microscopy. Anactivation energy of 3.0eV was deduced from the vacancy diffusion length, as estimatedfrom the width of denuded zones. The latter were created on both sides of an atomic step,by heating, after the introduction of vacancies by ion bombardment at room temperature.The results indicated that the vacancy diffusion kinetics were dominated by mono-vacancy formation and diffusion. These processes required thermal excitation in order toovercome the potential barrier to the surface diffusion of adatoms, and to overcome thelateral binding energy and thus release adatoms from the step edges.H.Watanabe, M.Ichikawa: Physical Review B, 1996, 54[8], 5574-80

[446-138/139-110]

386 Surface Diffusion - Quantitative DataIt was noted that, when a (001) substrate was heated by passing a direct current through it,electromigration of the Si atoms occurred. The activation energies for Si migration (table

501


86) on the surface were investigated here by means of reflection electron microscopy.During growth of the 2 x 1 terrace, Si atoms were released from Sb substrate steps in thestep-up direction. The activation energy was deduced to be about 1.60eV. During growthof the 1 x 2 terrace, Si atoms were released from Sa substrate steps, in the step-downdirection, and the activation energy was equal to about 1.52eV. The front edge of the 2 x1 terrace was the Sa step that was parallel to the dimer and the front edge of the 1 x 2terrace was the Sb step that was perpendicular to it. In the case of electromigration, thedifference in activation energies was attributed to a difference in the release energies fromsteps. The release energy of Si atoms from the Sb step in the step-up direction was about0.08eV higher than the energy for their release from the Sa step in the step-downdirection. The Si atoms were easily released from steps in the step-down direction on thesurface. There was also a difference in the frequency factors. The frequency factor for Sb

steps in the step-up direction was about twice as high as the frequency factor for Sa stepsin the step-down direction.T.Doi, M.Ichikawa, S.Hosoki, K.Ninomiya: Physical Review B, 1996, 53[24], 16609-14

[446-136/137-131]

Table 86Diffusion of Si Over Si Surface Terraces

Terrace T (C) D (cm2/s)2 x 1 600 2.0 x 10-10

2 x 1 650 7.6 x 10-10

2 x 1 700 2.3 x 10-9

2 x 1 750 5.1 x 10-9

2 x 1 800 1.2 x 10-8

2 x 1 850 2.6 x 10-8

2 x 1 900 5.3 x 10-8

1 x 2 600 3.2 x 10-10

1 x 2 650 9.0 x 10-10

1 x 2 700 2.6 x 10-9

1 x 2 750 6.5 x 10-9

1 x 2 800 1.5 x 10-8

1 x 2 850 3.0 x 10-8

1 x 2 900 5.6 x 10-8

Surface Diffusion - Qualitative Observations - AnisotropyScanning tunnelling microscopy was used to study the anisotropy of surface diffusion onthe (001) plane. This was done by analyzing the denuded zones around steps in the spatialdistribution of 2-dimensional islands which formed after sub-monolayer deposition. Itwas found that diffusion was some 1000 times faster along the dimer row direction thanperpendicular to it. It was also found that SB steps were good sinks for Si adatoms,

502


whereas SA steps were not. The SB steps appeared to be symmetrical sinks for adatomscoming from either the down terraces or the up terraces.Y.W.Mo, M.G.Lagally: Surface Science, 1991, 248[3], 313-20

[446-84/85-068]

Surface Diffusion - Qualitative Observations - AnisotropyThe diffusion of Si on (001) surfaces was studied by using scanning tunnellingmicroscopy. An anisotropy of island shapes during epitaxial growth was attributed mainlyto an anisotropic accommodation coefficient. The diffusional anisotropy was small. Anordered so-called diluted-dimer structure was detected at low coverages and temperatures.Y.W.Mo, R.Kariotis, B.S.Swartzentruber, M.B.Webb, M.G.Lagally: Journal of VacuumScience and Technology A, 1990, 8[1], 201-6

[446-74-055]

Surface Diffusion - Qualitative Observations - AnisotropyThe discovery of a novel diffusion pathway for Si dimers on the (001)Si surface wasreported. As in the case of a small molecule, it was noted that the configuration of thedimer could play a central role in diffusion. Scanning tunnelling microscopy, attemperatures near to 450K, revealed changes in the configuration of the dimers duringdiffusion. These changes provided a pathway for diffusion across the substrate dimerrows; unlike the usual diffusion along dimer rows.B.Borovsky, M.Krueger, E.Ganz: Physical Review Letters, 1997, 78[22], 4229-32

[446-152-0502]

Surface Diffusion - Qualitative Observations - Effect of ElectromigrationThe migration of Si micro-clusters on the (001) surface, at about 1500K, was observed bymeans of electron microscopy. Such clusters always had an associated region whichmelted first during heating, because of its lower crystallinity as compared with the Sisubstrate. The Si atoms in this melted zone were struck by electrons moving towards theanode, and were pushed - by the electron wind - in the same direction as the electrons.The Si micro-cluster, floating in the melted zone, also migrated towards the anode side;about 10 times as fast as the rate at which the Si atoms migrated in the opposite direction.T.Doi, M.Ichikawa, S.Hosoki, K.Ninomiya: Applied Physics Letters, 1996, 68[11], 1493-4

[446-134/135-167]

Surface Diffusion - Qualitative Observations - Effect of ElectromigrationThe driving force on Si adsorbates at a (001) surface was investigated by means ofreflection electron microscopy. When the sample was heated by using a direct current, themigration length of the adsorbates was proportional to the square root of the heating time,up to a critical time. This indicated that the dominant force on the Si adsorbates changedfrom thermal diffusion to electromigration at the critical time.T.Doi, M.Ichikawa, S.Hosoki: Applied Physics Letters, 1996, 69[4], 532-4

[446-136/137-131]

503


Surface Diffusion - Qualitative Observations - Effect of TrappingMonomers of Si were observed in empty-state scanning tunnelling microscopic images attemperatures between ambient and 115C. The monomers were trapped at the ends of re-bonded SB-type dimer rows. When the monomers escaped thermally from the traps, theyrapidly diffused along the substrate dimer row until they either found another unoccupiedtrap, or returned to their original trap. It was deduced that the binding activation barrier atisolated traps was about 1.0eV. It was noted that a slightly lower barrier existed for thehopping of monomers between the ends of neighboring dimer rows. This processfacilitated diffusion along segments of SB-type steps.B.S.Swartzentruber: Physical Review B, 1997, 55[3], 1322-5

[446-148/149-187]

Surface Diffusion - Theoretical Analysis - Activation EnergyBinding sites for the adsorption of a single Si atom upon the reconstructed (100) surfacewere identified by performing first-principles total-energy calculations. Several saddle-points were identified, for the migration of the adatom, by mapping the total energy as afunction of its position on the surface. An activation energy of 0.6eV was found fordiffusion parallel to dimer rows on the surface. For diffusion perpendicular to the rows,the activation energy was 1eV.G.Brocks, P.J.Kelly, R.Car: Physical Review Letters, 1991, 66[13], 1729-32

[446-84/85-075]

Surface Diffusion - Theoretical Analysis - Activation EnergyA simulation was performed, using a density functional method, of adatom diffusion on aflat (001) c(4 x 2) surface and around a single type of surface step (SA). The resultsindicated that a moderate additional energy barrier of 0.2eV had to be surmounted inorder to cross the SA step; as compared with the energy for diffusion on a flat surface. Thedimer-top lattice site on the lower terrace adjacent to the step edge was stabilized by0.15eV with respect to the flat surface value, although the most stable binding sites nearto the step were unaffected. This behavior was explained in terms of the disruption ofdimer tilt near to the step. The results suggested that adatoms were more likely to stop onlattice sites at the SA step edge than on lattice sites on the open surface. This could affectthe relative dimer formation rate near to the step, with respect to the behavior on the flatsurface; even in the absence of a clear change in binding energy. The effect of the SA stepterrace edge upon adatom behavior was very short-ranged and weak. This was consideredto be consistent with the relatively small strain field and lack of change in dangling-bonddensity which was associated with the step edge.J.Wang, D.A.Drabold, A.Rockett: Applied Physics Letters, 1995, 66[15], 1954-6

[446-121/122-092]

Surface Diffusion - Theoretical Analysis - Activation EnergyThe initial growth of Si on the (001) surface was investigated by studying the stability ofvarious Si ad-clusters and by calculating energy barriers for some fundamental processesby using a first-principles method. A perpendicular Si ad-dimer on top of a substratedimer row was easily formed and was the most stable of dimers; in accord withexperiment. The dimer could be a diffusing unit at high temperatures, but could not be a

504


nucleus for dimer row growth at lower temperatures. On the other hand, a quasi-stableparallel dimer in a trough attracted monomers and could be the nucleus of a diluted-dimerrow.T.Yamasaki, T.Uda, K.Terakura: Physical Review Letters, 1996, 76[16], 2949-52

[446-134/135-166]

Surface Diffusion - Theoretical Analysis - Activation EnergyInstabilities that were caused by Schwoebel barriers during growth, and their effects upongrowth or sublimation by step flows, were investigated by using the Stillinger-Weberpotential; especially with regard to how this step-edge barrier arose in the case of the 2high-symmetry steps on 1 x 1 reconstructed (111). Relative to a barrier of 0.97eV on thesurface, additional (Schwoebel) barriers of 0.61 and 0.16eV were found for adatommigration over the [211] and [112] steps, respectively. The adatom potential energy wasfound to be strongly related to that which was derived from the local geometry of atomson the adatom-free surface or step edges. This correlation preserved a strictcorrespondence between the barrier-determining features of the spatial variation of theadatom potential energy, and the same value as derived from the local geometry of theSi(111) surface and the [211] step. It was suggested that the Schwoebel barrier on the[211] step was a feature that would appear in more searching ab initio or tight-bindingcalculations. By using a diffusion equation for the adatom concentration, the relevance ofthe barrier to the electromigration of steps was analyzed. Experimental data on (111) wereused to place an upper bound on the Schwoebel barrier, and a lower bound on theelectromigration force.S.Kodiyalam, K.E.Khor, S.Das Sarma: Physical Review B, 1996, 53[15], 9913-22

[446-134/135-167]

Surface Diffusion - Theoretical Analysis - AnisotropyTotal energy calculations were made of reconstructed (001) surfaces with adatoms onvarious adsorption sites. The results were interpreted in terms of adatom/substrateinteractions in order to clarify the microscopic mechanisms of Si adatom surfacediffusion. The spatial variations in the adsorption energy suggested that there was amarked anisotropy in the direction of adatom diffusion on (001). The most probabledirection of diffusion was suggested to be along dimer rows. The surface dimer broke upwhen the adatom moved along the lateral projection of the dimer bond.T.Miyazaki, H.Hiramoto, M.Okazaki: Japanese Journal of Applied Physics, 1990, 29[7],L1165-8

[446-78/79-058]

Surface Diffusion - Theoretical Analysis - AnisotropyThe Stillinger-Weber potential was used to model the migration of a Si atom which wasadsorbed at a 2 x 1 (100) surface. The adsorption sites and the barriers between them weredetermined by means of energy minimization methods, and were used to estimate the rateof site-to-site hopping. A stochastic kinetic simulation which used these rates determinedhow the adsorbed atom migrated on the surface. It was found that the mobility was highlyanisotropic, and that the motion was quasi one-dimensional. The particle diffused along thedimer row and then became trapped on sites which were located between the dimers.

505


The activation energy for diffusion was a complicated function of the energy barrierswhich separated the surface sites. The pre-exponential factor could have unusual values.Y.T.Lu, Z.Zhang, H.Metiu: Surface Science, 1991, 257[1-3], 199-209

[446-88/89-055]

Surface Diffusion - Theoretical Analysis - AnisotropyA modified form of the Stillinger-Weber potential was used to determine the adsorptionsites of a Si atom on the fully-relaxed 2 x 1 (001) surface. The barriers between sites werealso found, and these were used to estimate the hopping rates from site to site. It wasconcluded that the direction of easy diffusion was parallel to, and to one side of, thedimer rows. The effective activation energy for the resultant quasi one-dimensionalmotion was about 0.33eV. The results also explained the appearance of regular islandgrowth at high temperatures, and so-called low-temperature diluted-dimer growth.C.P.Toh, C.K.Ong: Physical Review B, 1992, 45[19], 11120-5

[446-88/89-056]

Surface Diffusion - Theoretical Analysis - AnisotropyThe various binding sites and saddle points for an adatom on the (100) 2 x 1 symmetricalreconstructed surface were obtained by using a self-consistent semi-empirical method.This was done by minimizing the total energy, at each selected position on the surface,with respect to the adatom height and to the positions of its nearest-neighbor atoms. Thediffusion of an adatom on the surface was found to be anisotropic. The preferred motionoccurred along a zig-zag path, which was parallel to the dimer rows, with an activationenergy of about 0.63eV. This value was consistent with published theoretical calculationsand experimental estimates.C.K.Ong: Journal of the Physics and Chemistry of Solids, 1993, 54[2], 183-5

[446-099/100-100]

Surface Diffusion - Theoretical Analysis - AnisotropyThe migration of Si adatoms over the flat reconstructed (001) surface was studied byusing molecular dynamics techniques and the empirical Stillinger-Weber potential. Thepotential-energy surface which was seen by a single adatom over the surface was mappedout. The binding of an adatom to the (001) surface was found to be quite strong. Its valueof 3.0eV made up a sizable fraction of the energy of a bulk Si atom (4.33eV). Inagreement with previous studies, it was found that the diffusion of single adatoms overthe surface was anisotropic; with the fast diffusion direction lying along the dimer rows.Direct measurement of a diffusion coefficient was carried out by studying the motion ofindividual adatoms over periods of 300ps. The motion of individual adatoms conformedquite well to the underlying energy surface. However, frequent exchange events betweenadatoms and atoms near to the surface were observed. These furnished an additionalcontribution to the rate of diffusion. In addition, the potential energy surfaces which wereseen by a Si adatom over (001) surfaces that were under compressive and tensile strainswere mapped out. Here, the binding was found to increase or decrease depending uponwhether the surface was compressed or expanded. Little change was observed in theanisotropy of the diffusion constant. A study was also made of diffusion over a (001)surface in which the compression was either along, or perpendicular to, the dimer axis.

506


Compression of the surface perpendicular to the dimer axis decreased the diffusionalanisotropy, while compression of the surface parallel to the dimer axis increased thediffusional anisotropy.C.Roland, G.H.Gilmer: Physical Review B, 1992, 46[20], 13428-36

[446-106/107-149]

Surface Diffusion - Theoretical Analysis - AnisotropyThe migration of Si atoms on (100) was studied by analyzing, using a Monte Carlomethod, the islands which formed during deposition and by comparing the simulationswith experimental data for temperatures ranging from 350 to 500K. It was shown that thetemperature dependence of the experimental island number density could be reproducedby making a minimal number of assumptions. This model also reproduced manyexperimentally observed phenomena very well; such as the preferred formation of chainswith an odd number of dimers, the preferential ending of dimer chains over the dimerrows of the substrate, and the formation of various domains.J.L.Iguain, H.O.Mártin, C.M.Aldao: Physical Review B, 1996, 54[12], 8751-5

[446-138/139-103]

Surface Diffusion - Theoretical Analysis - AnisotropyThe molecular dynamics method was used to simulate the diffusion of adatoms on the(111) and (001) surfaces. A technique was proposed in which interatomic interactions inmulti-component systems could be obtained from a knowledge of the interactions withineach component. It was found that Ni atoms were more mobile than Si adatoms. On the(001) 2 x 1 reconstructed surface, Si adatoms preferred to move across dimers with a verylow diffusion barrier.P.Ashu, C.C.Matthai, T.H.Shen: Surface Science, 1991, 251-252, 955-9

[446-84/85-072]

Surface Diffusion - Theoretical Analysis - Effect of ElectromigrationIt was noted that electromigration had a marked effect upon the morphology ofsemiconductor surfaces and produced, for example, step-bunching instabilities. In order toexplain these phenomena from a microscopic point of view, self-diffusion on (111) Siwas studied by means of first-principles calculations. A mechanism was proposed for adiffusion bias that was produced by external electric fields. A competing wind-effectarose from an enhanced surface electron density; due to incomplete melting. The abovecompetition led to 2 transitions, in the surface kinetics, with increasing temperature. Thiswas in agreement with experimental observations.D.Kandel, E.Kaxiras: Physical Review Letters, 1996, 76[7], 1114-7

[446-131/132-185]

Surface Diffusion - Theoretical Analysis - Effect of StrainA computer simulation study was made of the effect of strain upon the diffusion of singleSi adatoms on an (001)(2 x 1) surface. A Stillinger-Weber potential was used to simulatethe interatomic interactions. Five layers of atoms were included in the simulation, withthe top 3 layers of the Si surface being free to reconstruct (leading to surface dimers),while the remaining 2 layers were fixed in bulk positions. The potential energy map andenergy

507


barriers were determined for 0, 1 and 1.5% strain; with the strain being appliedperpendicular to the dimer rows. The results indicated that diffusion enhancement alongthe dimer rows could be produced by strain.H.Spjut, D.A.Faux: Surface Science, 1994, 306[1-2], 233-9

[446-119/120-226]

Surface Diffusion - Theoretical Analysis - Potential FunctionA modified form of the Tersoff empirical interatomic potential was proposed in order toimprove the simulation of adatom behavior on surfaces. The modified form of thepotential was consistent with local density approximation calculations of the surfaceelectronic band structure of Si(001) 2 x 1. It was demonstrated that the addition of ascreened Morse potential tail to the bulk Tersoff interaction, when tetrahedralcoordination was disrupted, significantly improved the results. The surface structure wascalculated and was shown to yield substantial differences with respect to the originalpotential form. In particular, anomalous abrupt variations in adatom bonding energy wereeliminated and the probability of successful deposition of an adatom on a lattice site wasincreased.J.Wang, A.Rockett: Physical Review B, 1991, 43[15], 12571-9

[446-81/82-053]

Melt Diffusion - Quantitative DataThe diffusion coefficient in a melt was calculated for temperatures ranging from 1550 to1900K by using molecular dynamics simulations which were based upon a modifiedStillinger-Weber interatomic potential. The temperature dependence of the calculateddiffusion coefficient could be described by:

D(cm2/s) = 0.00038 exp[-0.27(eV)/kT]The viscosity of the melt, which was related to the diffusion constant, was also estimatedby using the Eyring relationship. It was found that the calculated viscosity agreed wellwith experimental data which were obtained by using an oscillating cup method; except attemperatures ranging from 1550 to 1750K.K.Kakimoto: Journal of Applied Physics, 1995, 77[8], 4122-4

[446-121/122-114]

508

Sn

Figure 21: Diffusivity of Sn in Si (see table 87)

Bulk Diffusion - Quantitative DataRadioactive tracer techniques and X-ray photoelectron spectroscopy were used to studydiffusion in films of hydrogenated amorphous material at temperatures of between 200and 500C. The films were P-doped and had various defect structures. The migration rateand the diffusion coefficient were found to depend upon the defect structure. Thus, thediffusion coefficient at 300C in P-doped material increased from 10-16 to 10-15cm2/s withincreasing defect density.V.K.Kudoyarova, G.S.Kulikov, E.I.Terukov, K.K.Khodzaev: Journal of Non-CrystallineSolids, 1987, 90, 211-4

[446-55/56-038]

1.0E-16

1.0E-15

1.0E-14

1.0E-13

6 7 8

6.816.87.597.597.877.87

104/T(K)

D (cm2/s)

509

Sn Bulk Diffusion Sn

Bulk Diffusion - Quantitative DataThe diffusion of ion-implanted Sn in single crystals was studied, as a function of the Pdonor concentration, during rapid thermal annealing at 1000 or 1050C. Depth profileswere determined by using Rutherford back-scattering spectroscopy and Mossbauertechniques. When the extrinsic/intrinsic carrier concentration ratio was greater than 20,extremely large diffusion coefficients were found. The value of the diffusion coefficientwas approximately proportional to the 4th power of the extrinsic/intrinsic carrierconcentration ratio. The onset of such a dependence was associated with the appearanceof a new defect complex which contained Sn.P.E.Andersen, A.N.Larsen, P.Tidemand-Petersson, G.Weyer: Applied Physics Letters,1988, 53[9], 755-7

[446-62/63-236]

Table 87Diffusivity of Sn in Si

T (C) D (cm2/s)1195 7.8 x 10-14

1097 6.3 x 10-15

1045 1.2 x 10-15

1045 1.0 x 10-15

998 1.9 x 10-16

998 1.7 x 10-16

Bulk Diffusion - Quantitative DataThe distribution of Sn which had been implanted into Si, Si0.79Ge0.21, and Ge0.53Si0.47 wasdetermined by means of secondary ion mass spectrometry. By comparing the Sndistribution before and after high-temperature annealing, the diffusion coefficient of Snwas obtained as a function of temperature. It was found that the diffusion coefficientsexhibited an Arrhenius behavior in all 3 materials; with activation energies for diffusionof 4.91, 4.61, and 3.88eV, respectively. At a given temperature, the diffusion coefficientof Sn increased almost exponentially with increasing Ge content. Although the diffusioncoefficient for Sn in Si and Ge was higher than the corresponding value for self-diffusion,the activation energies were similar to those for Si and Ge self-diffusion. It was suggestedthat the diffusion mechanism for Sn was similar to that for self-diffusion in Si and Ge,and in SiGe alloys.P.Kringhøj, R.G.Elliman: Applied Physics Letters, 1994, 65[3], 324-6

[446-119/120-210]

387 Bulk Diffusion - Quantitative DataThe distribution of implanted Sn was determined by means of secondary ion massspectrometry. By comparing the distribution before and after high-temperature annealing,the diffusion coefficient was obtained as a function of temperature. It was found that the

510


diffusion coefficients (table 87) exhibited an Arrhenius behavior; with an activationenergy for diffusion of 4.91eV. Although the diffusion coefficient was higher than thecorresponding value for self-diffusion, the activation energies were similar to those for Siand self-diffusion. It was suggested that the diffusion mechanism for Sn was similar tothat for self-diffusion.P.Kringhøj, R.G.Elliman: Applied Physics Letters, 1994, 65[3], 324-6

[446-119/120-210]

Bulk Diffusion - Qualitative Observations - Effect of DefectsThe effect of surface nitridation upon Sn diffusion at 1100C was investigated. Thenitridation, which was associated with vacancy generation, was found to enhance Sndiffusion. This was suggested to occur because the large Sn atoms preferred to diffuse viavacancies. On the basis of published values of the vacancy supersaturation which wasgenerated by the nitridation of bare Si, it was estimated that 33% of the Sn diffused viavacancies. However, the actual percentage might be higher because the high Snconcentration could affect nitridation-induced vacancy generation. The effect of oxidationwas unclear.B.P.R.Marioton, U.Gösele: Japanese Journal of Applied Physics, 1989, 28[7], 1274-5

[446-70/71-126]

Bulk Diffusion - Qualitative Observations - Effect of DopantStudies were made of diffusion in material with high donor concentrations that wereproduced by P doping. It was found that, for donor concentrations which were belowabout 2 x 10

20/cm

3, the diffusivity depended linearly upon the dopant concentration.

However, at higher dopant concentrations, the diffusivity increased markedly withincreasing donor concentration. This behavior was successfully modelled in terms of thevacancy-percolation model, and it was concluded that collective phenomena played asignificant role at high donor concentrations.A.N.Larsen, K.K.Larsen, P.E.Andersen, B.G.Svensson: Journal of Applied Physics, 1993,73[2], 691-8

[446-106/107-132]

Bulk Diffusion - Qualitative Observations - Effect of IlluminationAn investigation of the diffusion and solubility of Sn in hydrogenated amorphous filmswas carried out at temperatures ranging from 290 to 525C. This made it possible todetermine suitable conditions for diffusion doping. Such diffusion was used to preparehydrogenated amorphous films which were doped with this metal, and a study was madeof their dark conductivity and photoconductivity. It was noted that diffusion doping withmetals gave rise to activation energies, for electrical conduction, which were as high as1.4eV. That is, the activation energy exceeded 50% of the band gap; atypical behavior foran intrinsic material. Cyclic heating of the doped films, at temperatures ranging from 20to 200C, resulted in a gradual recovery of the electrical properties which they had had inthe original state. This recovery resembled the precipitation of a supersaturated solidsolution in crystals.

511


M.S.Ablova, G.S.Kulikov, S.K.Persheev, K.K.Khodzhaev: Fizika i TekhnikaPoluprovodnikov, 1990, 24[11], 1943-7 (Soviet Physics - Semiconductors, 1990, 24[11],1208-11)

[446-81/82-043]

Bulk Diffusion - Qualitative Observations - Effect of Surface LayerThe diffusion of Sn into hydrogenated amorphous material, deposited onto a SnO2/glasssubstrate, was found to be significantly reduced by coating a very thin hydrogenatedamorphous Si3N4 layer onto SnO2. It was shown that an 0.3nm-thick hydrogenatedamorphous Si3N4 over-layer, which was produced by plasma chemical vapor deposition orphotochemical vapor deposition, markedly suppressed reduction reactions on the SnO2

surface during hydrogenated amorphous Si growth. An 0.5nm-thick photochemical vapordeposited nitride layer could completely eliminate Sn diffusion into photochemical vapordeposited hydrogenated amorphous Si.Y.Shiratsuki, K.Kawabata, M.Sagawa, S.Miyazaki, M.Hirose: Journal of Non-CrystallineSolids, 1989, 115[1-3], 81-3

[446-72/73-072]

TeBulk Diffusion - Qualitative ObservationsThe migration of Te was studied by using a radiotracer technique which involved 121Tewhich was produced and on-line implanted. Spreading resistance measurements were alsomade after diffusion from the vapor phase. Penetration profiles of Gaussian type and oferfc form were found, respectively, corresponding to the actual boundary conditions. Thediffusion coefficients deduced from the two experiments exhibited excellent agreement. Apronounced curvature was observed in the Arrhenius diagram. Another striking featurewas the occurrence of diffusion profiles which had an exponential dependence upon thepenetration depth.N.A.Stolwijk, F.Rollert, D.Grünebaum, H.Mehrer, G.Weyer: Materials Science Forum,1986, 10-12, 133-8

[446-49-030]

512

Ti

Figure 22: Diffusivity of Ti in Si. The least-squares regression line through these pointsgives D (cm2/s) = 0.23 exp[-2.15(eV)/kT]

Bulk Diffusion - Quantitative DataThe behavior of type-3d transition metal impurities was investigated by using deep-leveltransient spectroscopic and Hall-effect measurements. Electrically active componentswere attributed to interstitial species, and the deep-level transient spectroscopic datarevealed double donors and an acceptor. The diffusivity of Ti, at temperatures rangingfrom 873 to 1423K, was described by:

D(cm2/s) = 0.12 exp[-2.05(eV)/kT]

1.0E-13

1.0E-12

1.0E-11

1.0E-10

1.0E-09

1.0E-08

6 7 8 9 10 11 12

table 88table 89

104/T(K)

D (cm2/s)

513

Ti Bulk Diffusion Ti

H.Nakashima, T.Sadoh, H.Kitagawa, K.Hashimoto: Materials Science Forum, 1994, 143-147, 761-6. See also: Materials Science Forum, 1992, 83-87, 227-32

[446-113/114-044], [446-93/94-052]

388 Bulk Diffusion - Quantitative DataThe diffusion profiles were determined at temperatures of between 950 and 1200C byusing etching techniques and deep-level transient spectroscopy. The experimentalconditions were particularly chosen so as to avoid O and N contamination. The resultantdiffusion coefficients ranged from 5 x 10-10 to 10-8cm2/s (table 88). These values weresome two orders of magnitude greater than those which had previously been reported.The associated activation energy was 1.79eV.S.Hocine, D.Mathiot: Applied Physics Letters, 1988, 53[14], 1269-71

[446-62/63-236]

Table 88Diffusion of Ti in Si

T (C) D (cm2/s)950 5.2 x 10-10

1000 1.3 x 10-9

1050 2.2 x 10-9

1100 3.5 x 10-9

1150 7.3 x 10-9

1200 8.8 x 10-9

389 Bulk Diffusion - Quantitative DataDeep impurity levels and Ti solubilities were determined by means of deep-level transientspectroscopy. It was found that the Ti formed multi-levels; with an acceptor level of Ec -0.09eV, a donor level of Ec - 0.27eV, and a double-donor level of Ev + 0.28eV. Thesolubilities, as a function of reciprocal temperature, were well-represented by straightlines with an associated activation energy of 3.8eV. The deep-level transient spectroscopydata indicated that the diffusivity (table 89) could be described by:

D(cm2/s) = 0.12 exp[-2.05(eV)/kT]S.Kuge, H.Nakashima: Japanese Journal of Applied Physics, 1991, 30[11A], 2659-63

[446-84/85-076]

Surface Diffusion - Quantitative DataScanning tunnelling microscopy was used to provide atomic-scale views of the reactionand diffusion of Ti on (001) at the monatomic adsorption stage. Two monatomicadsorption structures, below and above 440K, were found. The characteristic adsorptionbelow 440K involved a Ti atom at the pedestal site on a dimer row. The high-temperatureadsorption structure above 440K involved adsorption at a dimer vacancy that was inducedby a dimer ejection process on the structural conversion path. The high-temperatureadsorption structure exhibited a 1-dimensional hopping motion along dimer rows.

514

Ti Bulk Diffusion Ti

Measurements of the hopping rate yielded a diffusion activation energy and pre-factor of1.8eV and 1014/s, respectively. The growth features and diffusion mechanism wereinterpreted in terms of Ti-Si bond formation.K.Ishiyama, Y.Taga, A.Ichimiya: Physical Review B, 1995, 51[4], 2380-6

[446-121/122-093]

Table 89Diffusivity of Ti in Si

T (C) D (cm2/s)1150 6.6 x 10-9

1100 4.6 x 10-9

1050 1.8 x 10-9

800 2.4 x 10-11

750 9.0 x 10-13

700 2.4 x 10-12

600 2.3 x 10-13

515

Tl

Figure 23: Diffusivity of Tl in Si (see table 90)

390 Bulk Diffusion - Quantitative DataMonocrystalline wafer samples were annealed at temperatures ranging from 1070 to1300C, and the resultant diffusion profiles were determined by using a 4-point sheetconductivity technique. It was found that the results (table 90) could be described by theexpression:

D(cm2/s) = 15 exp[-3.75(eV)/kT]R.Sellmann, J.Mimkes: Physica Status Solidi A, 1989, 112[1], K5-7

[446-76/77-042]

1.0E-14

1.0E-13

1.0E-12

1.0E-11

1.0E-10

6 7 8

7.467.357.237.126.986.876.786.76.626.566.466.456.36

104/T(K)

D (cm2/s)

516

Tl Bulk Diffusion Tl

Table 90Diffusivity of Tl in Si

T (C) D (cm2/s)1067 6.5 x 10-14

1087 1.8 x 10-13

1110 4.8 x 10-13

1132 8.0 x 10-13

1160 1.2 x 10-12

1182 1.8 x 10-12

1202 2.1 x 10-12

1220 2.4 x 10-12

1237 5.7 x 10-12

1252 8.6 x 10-12

1275 5.7 x 10-12

1277 7.4 x 10-12

1300 2.0 x 10-11

517

V

Figure 24: Diffusivity of V in Si (see table 91)

Bulk Diffusion - Quantitative DataThe behavior of type-3d transition metal impurities was investigated by using deep-leveltransient spectroscopic and Hall-effect measurements. The diffusivity, at temperaturesranging from 873 to 1473K, was described by:

D(cm2/s) = 0.009 exp[-1.55(eV)/kT]

H.Nakashima, T.Sadoh, H.Kitagawa, K.Hashimoto: Materials Science Forum, 1994, 143-147, 761-6. See also: Materials Science Forum, 1992, 83-87, 227-32

[446-113/114-044], [446-93/94-052]

1.0E-11

1.0E-10

1.0E-09

1.0E-08

1.0E-07

6 7 8 9 10 11 12

6.797.037.287.567.868.189.3210.2811.46

104/T(K)

D (cm2/s)

518

V Bulk Diffusion V

Bulk Diffusion - Quantitative DataMonocrystalline samples were diffused with 48V (half-life = 16.1d) and annealed (1100 to1250C, 5 to 25h). The results demonstrated that the diffusivity of V increased from 3.4 x10-11 at 1100C to 4.4 x 10-10cm2/s at 1250C. Overall, the diffusivity obeyed theexpression:

D(cm2/s) = 0.61 exp[-2.8(eV)/kT]G.K.Azimov, S.Z.Zainabidinov, J.I.Kozlov: Fizika i Tekhnika Poluprovodnikov, 1989,23[10], 1890-1 (Soviet Physics - Semiconductors, 1989, 23[10], 1169-70)

[446-76/77-042]

Table 91Diffusivity of V in Si

T (C) D (cm2/s)1200 4.7 x 10-8

1150 2.4 x 10-8

1100 2.1 x 10-8

1050 1.4 x 10-8

1000 6.2 x 10-9

950 3.4 x 10-9

800 4.7 x 10-10

700 8.0 x 10-11

600 1.1 x 10-11

391 Bulk Diffusion - Quantitative DataThe concentration profiles of V, at annealing temperatures of between 600 and 1200C(table 91), were measured by means of deep-level transient spectroscopy. On the basis ofthe data, it was found that the diffusivity could be described by:

D(cm2/s) = 0.0090 exp[-1.55(eV)/kT]T.Sadoh, H.Nakashima: Applied Physics Letters, 1991, 58[15], 1653-5

[446-84/85-076]

519

WBulk Diffusion - Qualitative Observations and Point DefectsDeep-level transient spectroscopy was used to study the deep levels which wereintroduced, into the band gap, by W. The results indicated that W created 3 defectcenters: with levels at Ev + 0.22eV, Ev + 0.33eV and Ec - 0.59eV. The shape of theconcentration profiles indicated that W did not diffuse via a simple mechanism.S.Boughaba, D.Mathiot: Journal of Applied Physics, 1991, 69[1], 278-83

[446-78/79-058]

YbBulk Diffusion - Quantitative DataMonocrystalline wafers of n-type material were diffused with Yb from a layer source byannealing (1220 to 1370K, 1 to 5h). The resultant concentration profiles were determinedby means of neutron activation analysis. It was found that the data could be described by:

D(cm2/s) = 2.8 x 10-5 exp[-0.95(eV)/kT]M.K.Bakhadyrkhanov, F.M.Talipov, N.V.Sultanova, U.S.Dzhurabekov, S.S.Shasaidov,A.S.Lyutovich, A.A.Kasymov: Izvestiya Akademii Nauk SSSR - NeorganicheskieMaterialy, 1990, 26[3], 458-61 (Inorganic Materials, 1990, 26[3], 385-8)

[446-84/85-077]

Bulk Diffusion - Qualitative Observations and Ion ImplantationThe dependence of diffusion upon rapid thermal annealing was studied in Yb-dopedmaterial. During annealing at temperatures of between 800 and 1000C, the Yb segregatedat crystal/amorphous interfaces and at the free surfaces. The efficiency of opticalactivation also increased with annealing temperature. The amorphous layer re-grew, andno photoluminescence was observed, after annealing at 1200C.T.B.Xu, P.R.Zhu, D.Q.Li, T.Q.Ren, H.L.Sun, S.K.Wan: Physics Letters A, 1994, 189[5],423-7

[446-115/116-154]

520

Zn

Figure 25: Diffusivity of W in Si


[446-86/87-049]

1.0E-16

1.0E-15

1.0E-14

1.0E-13

1.0E-12

1.0E-11

1.0E-10

1.0E-09

1.0E-08

1.0E-07

1.0E-06

6 7 8 9 10 11 12 13 14

table 92table 93

104/T(K)

D (cm2/s)

521

Zn Bulk Diffusion Zn

393 Bulk Diffusion - Quantitative DataThe diffusion of Zn into dislocation-rich monocrystals was investigated by means ofspreading resistance analysis (table 93). In order to carry out short-term isothermalannealing, a special technique was used which exploited the volatility of Zn at hightemperatures. This permitted the progressive incorporation of Zn to be monitored in detailin the early stages of diffusion. By assuming the operation of the kick-out mechanism,effective diffusivities and interstitial-substitutional exchange rates were deduced from thetemporal development of penetration profiles at temperature ranging from 1143 to 1481K.Deviations from the theoretical predictions were frequently observed, and could beaccounted for by a deep trapping of Zn which was probably caused by dislocations. Thefitting of computer-simulated profiles to the data furnished volume averages of the trapconcentrations and trapping rates.H.Bracht, N.A.Stolwijk, I.Yonenaga, H.Mehrer: Physica Status Solidi A, 1993, 137[2],499-514

[446-113/114-050]

Table 92Diffusivity of Zn in Amorphous Si

T (C) D (cm2/s)625 4.7 x 10-14

560 1.2 x 10-14

502 2.5 x 10-15

455 5.9 x 10-16

Bulk Diffusion - Qualitative Observations - Concentration ProfilesDiffusion of ion-implanted elements in crystalline Si was investigated. The implantationwas limited to photolithographically defined areas of the wafer, and a spreadingresistance technique was used to measure the 3-dimensional concentration profiles of themetal atoms after high-temperature annealing. It was found that lateral spread under themask was greater than vertical diffusion; especially on the side opposite to the implanteddiffusion source. All of the important features of the measured profiles could be explainedas being a result of a kick-out diffusion mechanism. The peculiar shape of theconcentration profiles was attributed to an interplay between the incoming flux ofinterstitial metal atoms and the outgoing flux of Si self-interstitials that were generated bythe kick-out reaction. In spite of the high lateral diffusion it was noted that, by a suitablecombination of implantation fluence and annealing temperature, it was possible to limitthis lateral spread to within about 200µ, while maintaining a high metal concentration inthe region under the implanted area.S.Coffa, V.Privitera, F.Frisina, F.Priolo: Journal of Applied Physics, 1993, 74[1], 195-200

[446-106/107-133]

Bulk Diffusion - Qualitative Observations - Effect of DefectsPenetration profiles of Zn were recorded, by using the spreading resistance technique,after in-diffusion from the vapor phase at 1262K. The conversion of resistance data to

522


substitutional Zn concentrations was based upon an acceptor level, at Ev + 0.24eV, whichwas detected by Hall effect measurements of Zn-diffused samples. The diffusion behaviorwas found to be affected by the condition of the surface. In dislocation-free crystals withdamaged surfaces, the profile shape and the Zn incorporation rate indicated a self-interstitial limited diffusivity that depended upon C-2. Together with the much largerdiffusion constant in highly dislocated samples, this provided strong evidence for thesuggestion that Zn migrated via the kick-out mechanism.M.Perret, N.A.Stolwijk, L.Cohausz: Journal of Physics - Condensed Matter, 1989, 1[36],6347-61

[446-70/71-127]

Table 93Penetration Rate of Zn into Dislocation-Rich Si

T (C) D (cm2/s)870 4.0 x 10-9

942 1.6 x 10-8

1021 4.4 x 10-8

1115 1.3 x 10-7

1208 3.1 x 10-7

Bulk Diffusion - Qualitative Observations - Effect of DefectsA study was made of self-diffusion and foreign-atom diffusion; especially with regard toZn, whose diffusion behavior was affected by intrinsic point defects such as Si self-interstitials and vacancies. The diffusion of Zn in dislocation-free material was studied attemperatures of between 1208 and 870C. A special method was used to perform iso-thermal anneals which were as short as a few seconds. It was found that the concentrationversus depth profiles that were measured by using spreading resistance techniques couldbe completely described by simultaneous diffusion via the kick-out and dissociativemechanisms. The evolution of the Zn diffusion with time could be divided into short-,intermediate- and long-term diffusion regimes. The profiles which were associated withthe long-term regime were suitable for the extraction of the transport capacities ofintrinsic defects, as given by the product of the thermal equilibrium concentration and thediffusion coefficient. The Zn profiles for intermediate diffusion times were shown to besensitive to the prevailing thermal equilibrium concentrations of Si self-interstitials.H.Bracht, N.A.Stolwijk, H.Mehrer: Physical Review B, 1995, 52[23], 16542-60

[446-134/135-167]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsThe migration of some elements yielded Si self-interstitial diffusivities, that exceededthose which were obtained from dopant marker experiments by 6 orders of magnitude at800C. Both types of experiment could be reconciled by assuming the existence of a non-annihilating interstitial trap that was related to C. Selected metal diffusion data were re-

523


analyzed in this context. Non-annihilating immobile traps and a second-order reactionwhich involved interstitial C and C-C pairs were considered. Quantitative point defectparameters were obtained at 1115C for an assumed trap concentration of 5 x 1016/cm3.Excellent fits to Zn concentration versus depth profiles were obtained in all cases,regardless of the trap concentration. The estimated equilibrium concentration of Si self-interstitials varied inversely with the trap concentration, while the product of the self-interstitial diffusivity and the self-interstitial concentration remained almost constant. Itwas concluded that agreement, or disagreement, of metal diffusivity results with the Siself-diffusion coefficient could not be used to exclude or prove the occurrence of trap-limited diffusion. The published values of point-defect parameters which had beenobtained by neglecting traps were suggested to represent lower bounds on the self-interstitial diffusivity, and upper bounds on the equilibrium concentration of self-interstitials. The temporal evolution of the Zn profile in the presence of traps wasconsidered, and the proper inclusion of traps in the analysis of metal diffusion data led toa decrease in the estimated activation energy for self-interstitial diffusion.H.J.Gossmann, P.A.Stolk, D.J.Eaglesham, C.S.Rafferty, J.M.Poate: Applied PhysicsLetters, 1995, 67[21], 3135-7

[446-127/128-154]

Bulk Diffusion - Theoretical Analysis - Effect of DefectsProfiles, after diffusion at temperatures of between 870 and 1208C, were analyzed inorder to obtain information concerning the equilibrium concentrations of the intrinsicpoint defects which were involved in diffusion. An analysis of Zn diffusion data yieldedthe fractions of Zn atoms which were located on interstitial or substitutional sites, and themass-action constants of the underlying diffusion-reaction processes. The enthalpydifference between interstitial and substitutional Zn, and the reaction enthalpies fordiffusion processes, were deduced from the temperature dependences of these quantities.Total-energy calculations for Zn on substitutional and tetrahedral interstitial sites werealso carried out.H.Bracht, H.Overhof: Physica Status Solidi A, 1996, 158[1], 47-55

[446-150/151-151]

Bulk Diffusion - Qualitative Observations - Effect of IlluminationPhonon spectroscopy was used to analyze low-lying photon scattering states from defectsthat were introduced by the diffusion of Zn into thick Si wafers. It was concluded that Znwas an effective recombination center. Thus, illumination of the samples, although itproduced changes in the depth of various absorption lines, did not lead to appreciablechanges in the resistivity; as compared with the analogous situation for single acceptors inSi.J.Staiger, P.Gross, K.Lassmann, H.Bracht, N.A.Stolwijk: Materials Science Forum, 1994,143-147, 675-80

[446-113/114-050]

524

MiscellaneousIncluded under this heading are some studies which were not made of pure silicon or oflightly doped silicon, and studies whose primary purpose was not the determination ofelemental diffusivities themselves, but those of various types of point defect. While notfitting exactly into the general theme of the present volumes, the results of such studiesare nevertheless invaluable for evaluating proposed mechanisms of solute-affected ordefect-mediated diffusion.

SiGeOne of the most noticeable developments during the past decade has been the increasinginterest in alloying the already two most investigated semiconductors in order to benefitfrom an increased flexibility of control of the electronic properties, while profiting fromall of the accumulated information on the handling of the separate materials. Thecoverage of these alloys here is not as complete as for Si itself, because the subject ofsemiconductor alloys and compounds will be covered in future volumes.

B DiffusionThe thermal oxidation of Si was used to characterize the interaction of Si interstitials witha Si1-xGex layer, and with B in Si1-xGex. The diffusion of B in Si marker layers under Si1-x

Gex layers was monitored as excess Si interstitials were injected into the bulk viaoxidation of the Si capping layer. It was found that the Si1-xGex layer did not affect theoxidation-enhanced diffusion of B in the Si marker layer. Therefore, Si1-xGex (with x lessthan 0.30) did not appear to be a strong sink for Si interstitials. In addition, theenhancement of the measured B diffusivity in Si1-xGex (table 94), which arose from Sithermal oxidation, was similar to that in Si. It was concluded that, as in the case of Si, themechanism of B diffusion in Si1-xGex (for x-values of less than 0.18) involved mainlyinterstitials.P.Kuo, J.L.Hoyt, J.F.Gibbons, J.E.Turner, D.Lefforge: Applied Physics Letters, 1995,67[5], 706-8

[446-125/126-147]

525

SiGe

Table 94Intrinsic Diffusivity of B in Si1-xGex

x D (cm2/s)0 5.4 x 10-17

0.097 1.7 x 10-17

0.18 1.2 x 10-17

B DiffusionThe diffusion of B in SiGe strained layers on Si was studied as a function of annealingtemperature and Ge content, and was shown to be characterized by a lower diffusivitywhen compared with unstrained Si. The influence of the Ge content upon dopantdiffusion was also measured; thus demonstrating that the diffusivity of the B atoms wasreduced upon increasing the Ge fraction in the strained layer. The reduced B diffusivity inthe strained SiGe, relative to the dopant diffusivity in unstrained Si, was attributed to achange in the charged point-defect concentration which was caused by band-gapnarrowing. Good agreement was found between the measured and simulated diffusivitiesby using the known band-gap for the strained layers.N.Moriya, L.C.Feldman, H.S.Luftman, C.A.King, J.Bevk, B.Freer: Physical ReviewLetters, 1993, 71[6], 883-6

[446-106/107-202]

B DiffusionSecondary ion mass spectrometric studies of Si-rich layers revealed an decrease in Bdiffusion, as a function of compressive strain; thus indicating a linear dependence of theactivation energy upon strain. It was suggested that the effect arose from structuralrelaxation of the lattice around the defects which mediated the diffusion. This relaxationwas inward for a vacancy and outward for an interstitial.N.E.B.Cowern, P.C.Zalm, P.Van der Sluis, D.J.Gravesteijn, W.B.De Boer: PhysicalReview Letters, 1994, 72[16], 2585-8

[446-117/118-216]

B DiffusionThe investigation of B diffusion in strained buried Se 0.8Ge0.2 layers revealed a fractionalinterstitial component of the B diffusivity which was approximately equal to theequivalent value for Si. When combined with computer-simulated B profiles, the resultsplaced an absolute lower bound of about 0.8 upon the fractional interstitial component forthe present material. The method also provided a unique method for measuring thesegregation coefficient. Thus, oxidation-enhanced diffusion was used instead of extendedinert annealing in order to diffuse the dopant rapidly to equilibrium levels across theinterface; thus permitting the segregation coefficient to be measured more quickly.T.T.Fang, W.T.C.Fang, P.B.Griffin, J.D.Plummer: Applied Physics Letters, 1996, 68[6],791-3

[446-131/132-198]

526

SiGe

B DiffusionAn investigation was made of B diffusion in Si and in strained SiGe in situ dopedepitaxial layers. During inert ambient annealing at 860C, B diffusion was observed to beslower in Si0.83Ge0.17 than in Si for B concentrations of between 5 x 10

16 and 2.5 x

1019

/cm3. Computer simulations of the measured profiles in annealed samples indicated

that the effective B diffusivity in Si0.83Ge0.17 was about an order of magnitude lower thanthat in Si. This disparity increased with increasing B concentration.P.Kuo, J.L.Hoyt, J.F.Gibbons, J.E.Turner, R.D.Jacowitz, T.I.Kamins: Applied PhysicsLetters, 1993, 62[6], 612-4

[446-106/107-134]

B DiffusionA study was made of the diffusion of B in Si0.7Ge0.3, at temperatures of between 800 and1050C. It was found that, when the dopant concentration was lower than the intrinsiccarrier concentration at the diffusion temperature, the diffusion had a simple Fickiancharacter. The corresponding intrinsic diffusivities had an associated activation energy of1.79eV for B. The high-concentration B profiles were asymmetrical, with a pronouncedshoulder at the near-surface side. These results suggested that the diffusion of this dopantatom in SiGe involved complex couplings with lattice point defects, as for the case ofdiffusion in pure Si.D.Mathiot, J.C.Dupuy: Materials Science Forum, 1992, 83-87, 1303-8

[446-099/100-113]

Ge DiffusionSecondary ion mass spectrometric studies of Si-rich layers revealed an exponentialincrease in Ge diffusion as a function of compressive strain; thus indicating a lineardependence of the activation energy upon strain. It was suggested that the effect arosefrom structural relaxation of the lattice around the defects which mediated the diffusion.This relaxation was inward for a vacancy and outward for an interstitial.N.E.B.Cowern, P.C.Zalm, P.Van der Sluis, D.J.Gravesteijn, W.B.De Boer: PhysicalReview Letters, 1994, 72[16], 2585-8

[446-117/118-216]

Ge DiffusionA method was presented which permitted the simultaneous in situ investigation of therelaxation and diffusion behavior, of SiGe layers on Si substrates, by using a conventionalX-ray powder diffractometer with a high-temperature attachment. This method permittedthe direct determination of the time and temperature dependences of the relaxation and ofthe maximum Ge content. The diffusivity of Ge in Si was monitored by means of X-raydiffraction and secondary ion mass spectroscopic measurements, and a non-lineardependence of the effective diffusion coefficient upon the Ge content was deduced bysolving the diffusion equation. It was found that the Ge diffusivity was constant at about 8x 10-17 (1006C) or 5 x 10-17cm2/s (993C), for Ge contents of between 1020 and 1021/cm3, and

527

SiGe

then increased sharply to values of about 5 x 10-16 (1006C) or 4 x 10-16cm2/s (993C) at aGe content of 1022/cm3.P.Zaumseil, U.Jagdhold, D.Krüger: Journal of Applied Physics, 1994, 76[4], 2191-6

[446-117/118-218]

Ge DiffusionThe thermal stability of strained SiGe layers that had been grown, via molecular beamepitaxy, onto (100)Si was measured by using Rutherford back-scattering spectrometry,secondary ion mass spectrometry and high-resolution X-ray diffractometry. Diffusionexperiments were carried out on 50nm-thick Si1-xGex layers (where x was equal to 0.07,0.16 or 0.33) that had been annealed at temperatures of between 775 and 1010C forvarious times. The diffusion of Ge was deduced from broadening of the Rutherford back-scattering spectrometry and secondary ion mass spectrometry Ge profiles, while strainrelaxation was deduced from angular shifts of the (400) reflection in high-resolution X-ray diffractometry. The diffusion coefficients which were measured in this way werestrongly dependent upon the local Ge concentration in the film. In the tails of the profiles,the diffusion coefficient was comparable with that for Ge in bulk Si. In the center of thefilm, enhanced diffusion was noted. Both the initial Ge fraction in the as-grown film, andthe presence of misfit dislocations, had only a minor effect upon the diffusion behavior. Itwas concluded that safe thermal processing of such structures was possible for severalhours at up to 850C.G.F.A.Van de Walle, L.J.Van Ijzendoorn, A.A.Van Gorkum, R.A.Van den Heuvel,A.M.L.Theunissen, D.J.Gravesteijn: Thin Solid Films, 1989, 183, 183-90

[446-76/77-062]

Table 95Diffusivity of P in Si0.7Ge0.3

T (C) D (cm2/s)1050 3.1 x 10-13

1000 1.3 x 10-13

950 5.0 x 10-14

900 5.2 x 10-14

850 2.3 x 10-14

830 1.7 x 10-14

800 6.7 x 10-15

P DiffusionA study was made of the migration of P in Si0.7Ge0.3 at temperatures of between 800 and1050C (table 95). When the P concentration was lower than the intrinsic carrierconcentration at the diffusion temperature, the diffusivity exhibited a simple Fick's lawbehavior. The corresponding intrinsic diffusivity had an associated activation energy of1.62eV. At high concentrations, the diffusion behavior of P was similar to that which was

528

SiGe

observed in pure Si; in that the concentration profiles had kinks and tails. It was suggestedthat the diffusion mechanism was similar in both materials.D.Mathiot, J.C.Dupuy: Applied Physics Letters, 1991, 59[1], 93-5. See also: MaterialsScience Forum, 1992, 83-87, 1303-8

[446-099/100-113], [446-84/85-105]

Sb DiffusionA preliminary study was made of the effect of the Ge content upon Sb diffusion in Si-richSi1-xGex strained-layer heterostructures. It was found that the behavior of Sb wascharacterized by a higher diffusivity in the alloy than in Si. It was therefore deduced thatthere was a higher vacancy mobility and/or vacancy concentration in the alloy than in Si.It was proposed that, because B diffusion was reduced as the Ge content increased, itfollowed that B diffusion in this region did not occur mainly via vacancies. In the range oflow Ge contents, the presence of Ge suppressed B diffusion whereas, at higher Gecontents, at least one report had shown that B diffusion became faster. By assuming thatthe vacancy concentration or mobility was greater at higher Ge contents, it could bededuced that B diffusion was dominated by an interstitial-based mechanism in the low-Gerange and by a vacancy mechanism in the high-Ge range. It was noted that such strainedSi1-xGex layers were used to avoid dislocation short-circuit diffusion, but it remained anopen question as to whether the strain itself affected Sb diffusion in the layers. The effectof compressive strain upon the vacancy parameters was unclear, but the reported effectsof compressive strain upon interdiffusion did not indicate that strain could explain thepresent results. It was suggested that Sb could be used as a monitor, for dopant diffusionvia vacancies, for comparison with other dopants.A.D.N.Paine, M.Morooka, A.F.W.Willoughby, J.M.Bonar, P.Phillips, M.G.Dowsett,G.Cooke: Materials Science Forum, 1995, 196-201, 345-8

[446-127/128-195]

Sb DiffusionThe diffusion of Sb in relaxed Si1-xGex layers which had been prepared by means ofmolecular beam epitaxy, with x-values of between 0 and 0.5, was studied as a function ofcomposition. The diffusivity of Sb was found to increase with the Ge content, while theactivation energy for diffusion decreased with increasing Ge content. However, themeasured activation energies were significantly higher than the activation energies thatwere predicted by extrapolating between the activation energies for pure Si and pure Ge.A.N.Larsen, P.Kringhøj: Applied Physics Letters, 1996, 68[19], 2684-6

[446-134/135-200]

529

Point DefectsBulk - Quantitative DataAn investigation was made of the room-temperature diffusion and trapping ofimplantation-generated point defects in crystalline samples. This was done by monitoringtheir interaction with dopants, native contaminants such as C or O, and other defects.Spreading resistance measurements showed that some 10-7 to 10-6 of the defects whichwere generated at the surface, by 40keV Si implantation, were injected into the bulk.These defects underwent trap-limited diffusion and produced dopant deactivation and/orpartial annihilation of pre-existing µ-deep defect markers which had been produced byMeV He implantation. It was found that, in highly pure epitaxial layers, these effectsextended to several microns from the surface, thus demonstrating the long-rangemigration of point defects at room temperature. More detailed analysis identified Si self-interstitials, injected into the bulk, as being the major cause of dopant deactivation andpartial annealing of the vacancy-type defects (di-vacancies, P-vacancy, O-vacancy) whichwere generated by implantation. A lower limit of about 6 x 10-11cm2/s was deduced for theroom temperature diffusivity of Si self-interstitials.V.Privitera, S.Coffa, F.Priolo, L.K.Kyllesbech, S.Libertino, A.Carnera: NuclearInstruments and Methods in Physics Research B, 1996, 120[1-4], 9-13

[446-150/151-157]

Bulk - Quantitative DataSelf-interstitial and vacancy concentrations in Si had previously been obtained, as afunction of the diffusion time, by analyzing experimental data on the oxidation-enhancedor retarded diffusion of P and Sb. By using these results, and approximate solutions towell-known diffusion equations for self-interstitials and vacancies, self-interstitial andvacancy diffusivities of 2.3 x 10

-9 and 2.3 x 10

-10cm

2/s were determined at 1100C. The

thermal equilibrium concentrations of self-interstitials and vacancies were deduced to be3.4 x 10

16 and 2.1 x 10

17/cm

3, respectively.

Okino, T.: Japanese Journal of Applied Physics 2, 1993, 32[6B], L856-8[446-101/102-036]

Bulk - Quantitative DataTransient enhanced diffusion and electrical activation, after the non-amorphizing Siimplantation of lightly B-doped multi-layers, were found to have 2 distinct time-scales;each of which was related to a different class of interstitial defect. At 700C, ultra-fasttransient enhanced diffusion occurred within the first 15s, with a B diffusivityenhancement of more than 2 x 105. Immobile clustered B was present at lowconcentrations, after the ultra-fast transient, and persisted for between 100 and 1000s.The later phase of transient enhanced diffusion exhibited a near-constant diffusivityenhancement of about 104. This was consistent with interstitial injection that wascontrolled by dissolving 113 interstitial clusters. The relative contributions of ultra-fastand normal transient enhanced diffusion regimes, to the overall diffusive broadening ofthe B profile, depended upon the proportion of interstitials that escaped capture by 113clusters that were growing within the implantation-damaged region during annealing. Thepresent results explained an ultra-fast transient enhanced diffusion which had been

530

Point Defects

observed after medium-dose B implantation. In that case, there were enough B atoms totrap a large proportion of interstitials in Si-B clusters, and the remaining interstitialscontributed to transient enhanced diffusion without passing through an intermediate113 defect stage. The data on ultra-fast transient enhanced diffusion pulses placedlower limits of 2 x 10-10 and 2 x 10-13cm2/s on the diffusivities of the Si interstitial and Binterstitial, respectively, at 700C.N.E.B.Cowern, H.G.A.Huizing, P.A.Stolk, C.C.G.Visser, R.C.M.De Kruif,L.K.Kyllesbech, V.Privitera, L.K.Nanver, W.Crans: Nuclear Instruments and Methods inPhysics Research B, 1996, 120[1-4], 14-8

[446-150/151-159]

Bulk - Quantitative DataCompensated material containing B and a group-V donor was irradiated with 2MeVelectrons at 110K. Localized vibrational mode lines at 730 and 757/cm were attributed tointerstitial 11B and 10B atoms, respectively. These defects annealed out under second-orderkinetics at temperatures of between 200 and 250K, leading to a diffusion behavior of theform:

D(cm2/s) = 0.04 exp[-0.58(eV)/kT]Centers involving 2 equivalent B atoms were formed and gave localized vibrational modelines at 903, 912, 928, 599, 613, and 624/cm; corresponding to a defect with axialsymmetry. This center annealed out at the slightly higher temperatures of 250 to 300K, togive the previously reported Q center which had one localized vibrational mode line perisotope. At higher temperatures, this defect annealed out, with the formation of nearest-neighbor (B-donor) substitutional pairs.A.K.Tipping, R.C.Newman: Semiconductor Science and Technology, 1987, 2[7], 389-8

[446-51/52-151]

Bulk - Quantitative DataA deactivation of acceptor In after mechanochemical polishing of p-type material wasshown to result from the formation of a complex which involved In ions and a positivelycharged very fast-moving unknown defect. At temperatures of between 220 and 280K, thedissociation frequency of the complex and the diffusion coefficient of the unknown defectwere thermally activated. In particular, the diffusion behavior was described by theexpression:

D(cm2/s) = 50000 exp[-0.665(eV)/kT]T.Zundel, J.Weber, B.Benson, P.O.Hahn, A.Schnegg, H.Prigge: Applied Physics Letters,1988, 53[15], 1426-8

[446-62/63-243]

Bulk - Qualitative DataThe motion and clustering of vacancies and interstitials were simulated by usingmolecular dynamics methods. The diffusion coefficients of isolated defects wereestimated from the atomic displacements in simulations which were performed over awide range of temperatures. The results yielded an apparent migration energy of 0.43eVfor vacancies and 0.9eV for interstitials. The diffusion coefficients were between 10-6 and10-5cm2/s at 800C, and agreed approximately with first-principles calculations, but were

531

Point Defects/Comments

many orders of magnitude higher than the most direct experimental values. Simulations ofhigh concentrations of defects showed that like defects aggregated into stable clusters,and that individual defects were bound to these clusters with energies ranging from 0.6 to1.3eV. The defect clusters had mobilities which could be appreciably different to those ofindividual defects. The di-interstitial had a much smaller diffusion energy (about 0.2eV),whereas the tri-interstitial had a mobility which was so small that it was difficult tomeasure accurately in molecular dynamics simulations.G.H.Gilmer, T.Diaz de la Rubia, D.M.Stock, M.Jaraiz: Nuclear Instruments and Methodsin Physics Research B, 1995, 102[1-4], 247-55

[446-134/135-192]

Surface - Quantitative DataThe formation of vacancies and their diffusion on the 7 x 7 (111) surfaces, in the initialstages of oxidation, were studied by using in situ ultra-high vacuum reflection electronmicroscopy. During O exposure experiments at low pressures, vacancies were formed onthe surfaces, due to the formation and sublimation of SiO, and coalesced into hollows inthe central parts of the terraces between successive surface atomic steps when the Sisurface temperature was greater than 500C. The distribution of hollows on the surfaceterraces, and their growth kinetics, were studied. Analysis showed that the activationenergy for surface diffusion of the vacancies was about 1.4eV. The reaction rate requiredto form 2 vacancies by the impingement of an O molecule on the surface was estimated tobe about 0.23 at 630C.N.Shimizu, Y.Tanishiro, K.Takayanagi, K.Yagi: Surface Science, 1987, 191[1-2], 28-44

[446-55/56-044]

CommentsThe previous review of the present type (Diffusion and Defect Data, volume 47, 1986)included a summary of all of the known data on diffusion in silicon, up to that date. Theaveraged values of the activation energies were as given in table 96, where they aresupplemented by, and compared with, the averaged activation energies from the presentcollection (table 97).

It is noticeable that the consistency of the values for the most important, andtherefore most extensively studied, elements is high. The fact that dependable data nowseem to be available, on the most technologically important solutes, may explain thedownward trend, in recent years, in the frequency of 'classical' studies of diffusion in Si(and indeed in other hosts). Much of the current experimental effort on silicon now seemsto be preferentially directed towards highly sophisticated methodologies in which themigration of individual atoms can be monitored by using scanning tunnelling electronmicroscopy, field ion microscopy, and similar techniques.

532

Comments

This is quite logical, as the technological trend is also towards the 'micro-manipulation' of atoms in order to fulfil the theoretical promise of 'quantum-dots' andother 'low-dimensional' structures. It is obviously advantageous to determine diffusionparameters under conditions which are as close as possible to those of the applications forwhich the information will be needed. Indeed, one might argue that the time when onehad to make do with a 'general purpose' coefficient, culled from a reference book, haslong gone. One can instead hope to find data which were obtained under the exactconditions which are relevant to a projected application, or even to have access to acomputer code which will simulate the situation in question.

Table 96Activation Energies (eV) for Diffusion in Si

Diffusant pre-1986 1987-1997Ag 2.32 1.41Al 3.47 3.24As 3.79 3.25Au 1.14 1.65B 3.44 2.26Bi 4.25 2.50C 2.93 0.87Cd 3.70 -Co 2.79 -Cr 1.00 0.79Cu 0.72 0.61Fe 0.73 0.72Ga 3.26 -Ge 4.79 5.39H 1.07 0.80In 3.77 -Ir 1.30 -K 0.79 -Li 0.67 -Mn 0.97 0.72N 3.63 2.9Na 1.00 -Ni 1.39 1.03O 2.50 2.51P 2.77 2.99

Pm 0.64 -Pr - 3.3Pt 2.19 1.79S 2.02 1.80

533

Comments

Table 96 (continued)Activation Energies (eV) for Diffusion in Si

Diffusant pre-1986 1987-1997Sb 3.76 3.65Sc - 3.2Se 1.19 -Si 4.57 1.65Sn 3.90 -Te 3.34 -Ti 1.50 2.05Tl 3.79 3.75V - 2.18Yb - 0.95

For there has also been an accelerating upward trend in the tendency to carry outcomputer simulations in preference to experimental studies. This can alarm someexperimental purists, especially when the conclusions of computer modelling papers oftenstate, perhaps carelessly, that experimental data has been checked against the results ofthe simulation; rather than vice versa.

A wide scatter of activation energy data in the case of certain problematicelements, such as carbon, may reflect experimental difficulties. In other cases, anapparent scatter (as in the case of silicon self-diffusion) may indicate that the diffusantcan migrate via several mechanisms. In summarizing the data on activation energiesbelow, no attempt has been made to distinguish between the figures for amorphous andcrystalline samples, as the reader has immediate access here to all of the relevant abstractson the subject and can thereby judge the significance of the reported value.

Table 97Summary of Diffusion Parameters

Diffusant Path T (C) Do (cm2/s) E (eV) PageAg bulk 305 to 485 0.16 1.67 217Agi bulk 1014 to 1325 0.6 1.15 218Al bulk - 8.0 3.47 221Al pipe - 140 3.01 221As bulk 1050 to 1100 35 4.00 225As bulk 1050 to 1100 22.83 4.10 225a

As bulk 850 to 1050 8.0 4.05 226b

As bulk 1000 to 115 7.85 x 10-8 0.9 228Au bulk 800 to 1200 0.021 1.7 248

534

Comments

Table 97 (continued)Summary of Diffusion Parameters

Diffusant Path T (C) Do (cm2/s) E (eV) PageAu bulk 905 to 1200 0.28 1.6 249B bulk 1050 to 1100 0.9 3.43 225c

B bulk 800 to 1000 0.022 2.5 271B bulk - 0.105 3.22 273d

B bulk 1000 to 1200 0.0003 2.1 274B bulk 1000 to 1150 0.019 2.5 277B bulk 1000 to 1150 0.000032 1.86 277Bi bulk 1050 to 1200 0.0002 2.50 327C bulk 24 to 58 0.44 0.87 328Cr bulk 27 to 400 0.00068 0.79 333Cu bulk 900 to 1050 0.015 0.86 338Cu bulk 7 to 127 0.0045 0.39 339D bulk -53 to -3 0.005 0.49 343Fe bulk 0 to 1070 0.0011 0.66 389Fe bulk 0 to 72 0.33 0.81 391Fei bulk <470K 0.01 0.84 390Fei bulk <470K 0.0014 0.69 390e

Fei bulk 800 to 1070 0.00095 0.65 391Ge bulk 1000 to 1200 1.38 x 105 5.39 397Ge surface 687 to 827 600 2.48 400H bulk 95 to 280 0.00002 0.49 342H bulk - 0.07 0.54 343H bulk 80 to 100 0.000042 0.56 345H bulk 225 to 350 170 1.2 348H bulk 100 to 300 0.0001 0.50 355H surface - 0.001 1.5 381

Mn bulk 14 to 90 0.0024 0.72 409N bulk 800 to 1200 2700 2.9 411Ni bulk 220 to 540 0.0006 0.76 416Ni bulk 270 to 435 0.003 1.30 416O bulk - 0.025 2.43 420O bulk 800 to 1300 0.00071 2.0 423O bulk 1000 to 1200 141 3.1 423P bulk - 0.145 3.26 433f

P bulk 1015 to 1210 1.9 3.3 434P bulk 767 to 1227 5.7 3.75 434P grain boundary 566 to 980 0.0048 2.65 461

535

Comments

Table 97 (continued)Summary of Diffusion Parameters

Diffusant Path T (C) Do (cm2/s) E (eV) PageP bulk 750 to 1050 1.6 x 1015 2.09 462P grain boundary 750 to 1050 4.0 x 10-5 1.4 462P grain boundary 900 to 1150 120 2.87 462Pr bulk 1100 to 1250 0.005 3.3 466Pt bulk 950 to 1120 2.1 1.79 469S bulk 1055 to 1398 0.047 1.80 474Sb bulk 850 to 1050 0.214 3.65 226g

Sb grain boundary 900 to 1150 380 2.9 490Sc bulk 1100 to 1250 0.08 3.2 492Si surface - 0.001 0.67 499Si melt 1277 to 1627 0.00038 0.27 507Sii bulk 800 to 1200 0.006 3.3 493Sii bulk 750 to 900 100 3.1 494Sii bulk 460 to 500 0.355 1.86 495Ti bulk 600 to 1150 0.12 2.05 512Tl bulk 1067 to 1300 15 3.75 515V bulk 600 to 1200 0.009 1.55 517V bulk 1100 to 1250 0.61 2.8 518Yb bulk 947 to 1097 2.8 x 10-5 0.95 519

aco-diffused with B, bdepends upon donor concentration, cco-diffused with As, ddepends upon partial pressure ofNH3, eunder influence of electric field, fdepends upon NH3 pressure, gdepends upon donor concentration

It is always interesting to try to find correlations between data, especially before anacceptable theoretical explanation becomes available. One possibly useful, and perhapstheoretically important correlation has been discussed for many years and has acquiredseveral different names during its history. It has been called the theta effect, thecompensation law and, for the present purpose, will be referred to as the Meyer-Neldelrule.

As is well known, this relates the logarithm of the pre-exponential factor to a linearfunction of the activation energy. When the present data are plotted in this way, the result(figure 26) indeed indicates that there could well exist such a relationship; although thedegree of scatter is unpleasing. One can nevertheless hope to detect obviously anomalousoutlying points by using such a plot.

Figure 26 also incidentally illustrates another trend which was indicated by adiagram in volume 47 of Diffusion and Defect Data (p2). The averaged activationenergies for each element were there plotted on a periodic table. This suggested that therewas a general trend towards higher energies in moving from left to right (with thetransition elements constituting an exception, as usual).

536

Comments

Figure 26: A Meyer-Neldel Plot of Diffusion Parameters from the Present Abstracts.The best linear least-squares fit to these points yields,

ln [Do] = 11.77E - 26.24

The existence of this trend is also borne out by figure 26, where there is a markedtendency for elements from the right hand side of the periodic table to occupy the upperhalf of the plot, while the opposite is true of elements from the left hand side.

0

1

2

3

4

5

6

-16 -6 4 14 24 34

AgAlAsAuBBiCCrCuDFeGeHMnNNiOPPrPtSSbScSiTiTlVYb

E (eV)

ln[Do (cm2/s)]

537

Comments

Figure 27: Activation Energy for Diffusion as a Functionof the Outer Electron Structure of the Diffusant

(21 means 2 electrons in s-level, 1 electron in p-level, etc.The activation energies are averages of all available data.

The horizontal scale is arbitrary)

The reason for this tendency is not as clear as one might assume. The apparentlyobvious correlation with atomic mass or radius produces plots (not shown) which are 'allscatter'; as does an attempted correlation with electronegativity. Nevertheless, themagnitude of the activation energy does appear to be quite strongly linked to theelectronic structure of the diffusant. It is unclear what manifestation of the electronicstructure finally exerts its effect at the atomic scale.

0

1

2

3

4

1 2 3 4 5 6

1221222324

E (eV)

538

539

Author Index

Abbasi, S.A. [318]Abdurakhmanov, K.P. [338, 434]Abe, H. [340]Abe, S. [446]Abe, T. [25, 311, 312, 412]Abeles, B. [343]Ablova, M.S. [219, 395, 511]Abrantes, T. [462]Ackermann, H. [302]Adler, J. [387]Afonin, N.N. [458]Ahn, S.T. [446, 486, 497]Ahrens, R.E. [305, 448]Aiba, Y. [295]Aikawa, I. [413]Ajioka, T. [413]Ajmera, A.C. [299]Albers, J. [322]Aldao, C.M. [506]Aleksandrov, L.N. [243, 324, 461]Aleksandrov, O.V. [385, 458]Alekseev, A.E. [289]Alexander, H. [463]Alexandrov, O.V. [385]Alford, T.L. [261]Alkemade, P.F.A. [384]Allain, J.L. [483]Allen, C.E. [400, 405]Allen, L.P. [292, 456]Allongue, P. [349]Amaratunga, G. [242]Amaratunga, G.A.J. [317]Amroun, N. [394]An, D.K. [260, 311]Ancey, P. [226, 268]Anderle, M. [286, 299, 386, 388, 443]

Andersen, P.E. [227, 399, 477, 486, 509, 510]

Anderson, D. [242]Anderson, R.A. [344, 347, 360]Ando, Y. [239, 297, 306]Andsager, D. [340]Angelucci, R. [286, 234, 236, 285, 297,

442, 447, 448, 481, 485]Antoncik, E. [183, 242, 243, 454, 490]Antonelli, A. [495, 497]Antonova, I.V. [262]Aoki, N. [233]Aoki, S. [341]Aoyama, S. [270]Appelbaum, A. [332]Arai, E. [25, 235, 244, 311, 312, 455]Armigliato, A. [233, 299, 308, 483, 484]Aronowitz, S. [317, 323, 396, 453]Arora, N.K. [345]Asher, S. [358, 367]Ashok, S. [360, 368]Ashu, P. [418, 506]Askarov, S.I. [334, 394, 410, 474, 492]Asoka, P. [388]Aucouturier, M. [354, 358, 361]Augustus, P. [282]Avezmuradov, A. [274]Ayres, J.R. [287]Azim, M. [335]Azimi-Nam, S. [221]Azimov, G.K. [492, 518]Aziz, M.J. [1, 402, 488]Baba, M. [446]Babich, V.M. [429]Baccus, B. [243, 320, 461]Bachilo, E.E. [384]

Author Index

540

Bacmann, J.J. [236, 297, 444, 483]Baer, Y. [346, 353, 362, 365, 366, 367]Baeri, P. [240]Bagraev, N.T. [289, 441]Baillin, X. [236, 297, 444, 483]Bakhadyrkhanov, M.K. [334, 394, 410,

417, 474, 492, 519]Bakker, H. [244, 435, 462, 476, 490]Balasubramanian, N. [350]Balasubramanian, S. [350]Ballutaud, D. [354, 358, 364]Banerjee, S. [232, 273, 278, 279, 283,

290, 437, 438]Bao, X. [300]Bao, X.M. [245, 301, 304]Bar, Y. [374, 380]Baran, N.P. [429]Barbier, D. [239, 280, 333, 430]Barlow, R.D. [282]Barna, A. [260, 311]Baró, A.M. [463]Baruffaldi, F. [271, 305]Batra, S. [232, 273, 278, 279, 283,

290, 437, 438]Batstone, J.L. [248]Battistig, G. [260, 311]Bazarnyi, J.A. [261]Becker, P. [384]Behar, M. [269]Benner, J.P. [378]Bennett, P.A. [417]Benson, B. [532]Bergaud, C. [239, 280, 314]Bergman, K. [345]Bernard, M.C. [349]Bernardini, J. [253]Bernholc, J. [497]Berstein, J. [314]Berti, M. [238]Beserman, R. [323, 452, 487]Bevk, J. [527]Beyer, W. [357, 361, 370, 377, 407]Bhide, V.G. [373]Biegelsen, D.K. [246]Binns, M.J. [427, 430]Biró, L.P. [306]Blakely, J.M. [499]Blöchl, P.E. [381, 496]Bock, W. [384]Bogdanov, E.I. [460]

Böhm, H.J. [229, 278, 311]Bonafos, C. [314, 477]Bonapasta, A.A. [364, 380]Bonar, J.M. [530]Bondarenko, V.P. [240, 384]Bondareva, T.V. [243, 324, 461]Boone, T. [284, 291, 305,

479, 494, 497]Borenstein, J.T. [370, 373, 374]Borghesi, A. [424, 425]Borisenko, V.E. [240, 384]Borodina, O.M. [369]Borong, S. [308]Borovsky, B. [502]Bouchetout, A.L. [398]Boughaba, S. [519]Boulmer, J. [331]Bourguignon, B. [331]Boutaoui, N. [219]Boutry-Forveille, A. [358]Bowler, D.R. [382]Brabec, T. [442, 488]Braeunig, D. [359]Braga, N. [245]Brandt, M.S. [357, 366]Branz, H.M. [349, 358, 367, 379]Brault, P. [386]Breitschwerdt, A. [356, 363]Brener, R. [323, 452, 487]Brice, D.K. [347]Briggs, G.A.D. [382]Bringans, R.D. [246]Brizard, C. [483]Brocks, G. [503]Brotherton, S.D. [287]Brown, A.R. [348, 424, 427]Brown, R.A. [420, 432]Brunco, D.P. [402]Bruson, A. [346, 403]Buczkowski, A. [245]Buda, F. [378]Budil, M. [255, 326, 442, 488]Budin, J.P. [331]Bullock, J. [367]Burghartz, J.N. [296]Bustillo, J. [232, 279, 437]Cacciato, A. [295, 306]Calcagno, L. [248, 249, 252, 494]Calder, I. [245]Calleri, G. [248, 494]

Author Index

541

Campisano, S.U. [248, 249, 252, 404, 468, 494]

Canteri, R. [226, 229, 268, 271, 386, 441]Canut, B. [239]Capaz, R.B. [329]Capizzi, M. [380]Caprara, P. [299]Car, R. [378, 496, 503]Carnera, A. [483, 531]Carter, E.A. [381, 383]Casarin, M. [432]Cellini, C. [308, 484]Cembali, F. [234, 236, 285, 286,

297, 447, 448, 481, 485]Chadi, D.J. [376, 377]Chang, C. [232, 279, 437]Chang, C.Y. [308, 310]Chang, K.J. [376, 377]Chao, H.S. [307, 315]Chao, T.S. [308, 311]Chapman, R. [417]Chari, A. [361]Chaudhry, S. [447]Chaussemy, G. [226, 239, 268, 333]Chelyadinski, A.R. [270, 444]Chen, C.S. [428]Chen, C.W. [283]Chen, G. [271, 365]Chen, J. [314]Chen, L.J. [387]Chen, N.K. [205, 273, 434]Chen, S. [421]Chen, T.P. [310]Chenevier, P. [230, 231, 322]Cheng, H.C. [303, 305]Cheng, K.L. [303]Cheshuina, S.E. [262]Chevacharoenkul, S. [417]Chi, P. [313]Chi, P.H. [313]Chiang, A. [381]Chiarotti, G.L. [378]Chichikalyuk, J.A. [409]Chien, C.H. [308]Cho, K.H. [341]Choi, S.S. [240]Chou, J.S. [451]Chu, C.H. [308, 311]Chu, S.F. [272, 326]Chu, W.K. [240]

Chun, H.G. [231]Ciavola, G. [300]Cielaszyk, E.S. [367]Cinti, R.C. [463]Clapper, R.A. [282]Claverie, A. [11, 314, 477]Clegg, J.B. [287]Coffa, S. [137, 218, 248, 249, 250, 251,

252, 256, 337, 465, 469, 470, 494, 520, 521, 531]

Cohausz, L. [522]Cohen, D. [359]Cohen, S.A. [281]Colombo, L. [375]Cooke, G. [530]Corbel, C. [97]Corbett, J.W. [370, 373, 374, 376]Correia, A. [364]Cotte, J. [465]Cowern, N.E.B. [228, 242, 277, 287,

289, 295, 304, 306, 318, 320, 404, 454, 487, 494, 527, 528, 531,532]

Crabbé, E.F. [230]Crans, W. [532]Crowder, S.W. [292, 307, 449, 456]Custer, J.S. [295, 306, 384, 385]Dal Pino, A. [329]Dallmann, G. [319]Darwich, R. [366]Das Sarma, S. [504]Das, B.K. [345]David, S.K. [373]De, M.M. [317]De, W.B. [290, 525, 526]Deak, P. [376]Deaton, R. [237, 437]Débarre, D. [331]De Boer, W.B. [527, 528]Deicher, M. [338]De Keersmaecker, R. [230, 237, 278,

281, 437]De Kruif, R.C.M. [277, 494, 532]Deline, V.R. [281]De Mierry, P. [354, 358, 361, 376]Demireva, D. [221]Deng, B. [470]Deng, J. [355]Deng, X. [378]Denhoff, M.W. [270]Denteneer, P.J.H. [374]

Author Index

542

Derdour, M. [233]De Souza, M.M. [69]Detzel, T. [405]De Villeneuve, C.H. [349]Di, A. [300]Diaz de la Rubia, T. [318, 533]Diehl, E. [302]Dippel, R. [302]Ditchfield, R. [400, 405]Döbler, U. [332]Doi, T. [81, 500, 501, 502]Dominguez, E. [316]Dorofeev, A.M. [384]Dowsett, M.G. [282, 530]Drabold, D.A. [503]Dragnea, B. [331]Drigo, A. [483]Drigo, A.V. [238]Dubois, C. [231, 279]Dufour, C. [346]Dunham, S. [434]Dunham, S.T. [436, 450, 451, 457]Dupuy, J.C. [526, 530]Dusane, R.O. [373]Dytlewski, N. [359]Dzhurabekov, U.S. [519]Eaglesham, D.J. [259, 290, 291, 292,

295, 296, 315, 316, 325, 472, 479, 482, 523]

Eichinger, P. [229, 278, 436]Elkaim, P. [364]Elliman, R.G. [307, 313, 509, 510]Elsayed, H.E. [463]Emtsev, V.V. [385]Ergezinger, K.H. [302]Erokhin, J. [297]Erokhin, Y.N. [395]Erwin, R. [346]Erxmeier, J. [359]Estreicher, S.K. [349, 364, 378, 432]Evans, A.G.R. [457]Fahey, P. [288, 398, 445, 465, 484]Fair, R.B. [235, 238, 241, 281, 285,

295, 299, 309, 387, 440, 444, 448, 453, 485]

Falster, R. [471]Fan, D. [272, 288, 298, 299]Fang, T.T. [527]Fang, W.T.C. [525]Fang, Y.K. [283]

Faux, D.A. [507]Faye, M.M. [314]Fedchuk, A.P. [289]Fedders, P.A. [378]Fedotov, S.A. [487]Feldman, L.C. [527]Fellinger, P. [420, 421]Feng, D. [300, 301]Ferla, G. [240, 248, 249, 252, 300, 494]Feudel, T. [319, 457]Feygenson, A. [241]Fichtner, W. [319]Finger, F. [346, 353, 362]Fink, D. [269, 359]Fink, R. [405]Finstad, T.G. [238, 398, 399]Fischer, B. [302]Flemish, J.R. [440]Frank, H.P. [302]Frank, W. [218, 251, 256, 260, 337,

465, 469, 497, 520]Freer, B. [314, 527]Frenzel, E. [229, 278, 436]Frohne, C. [222]Fujimori, H. [410]Fujimoto, H. [401]Fujino, N. [395]Fujioka, M. [224]Fujita, K. [399]Fujita, S. [363]Fukano, T. [316]Fukatsu, S. [399, 476]Fukatsu, S.J. [480]Fukuda, T. [412]Furmanov, G.P. [262]Furuya, H. [310]Gaiseanu, F. [296, 326, 447]Galvagno, G. [222, 223]Ganz, E. [502]Gaponenko, N.V. [384]Garben, B. [269]Garrison, B.J. [400]Gaworzewski, P. [438]Gay, N. [446]Gdanitz, H. [264]Gebhard, M. [348]Geddo, M. [424, 425]George, A. [236, 297, 444, 483]Gerasimov, O.I. [289]Gericke, M. [298]

Author Index

543

Ghaderi, K. [255, 458]Gibbons, J.F. [230, 294, 310, 526, 528]Gierisch, H. [229, 278, 436]Giles, M.D. [450, 451]Gill, A. [287]Gilles, D. [332]Gilmer, G.H. [315, 318, 325, 497, 506, 533]Gislason, H.P. [407]Glang, R. [280]Gleskova, H. [367]Gnaser, H. [384]Goldiner, M.G. [256, 265]Golding, M. [423]Goldman, L.M. [249]Gómez-Rodríguez, J.M. [463]Gontrand, C. [226, 228, 231, 233,

239, 268, 277, 279, 280]Gorbacheva, N.I. [398]Goringe, C.M. [382]Gösele, U. [237, 330, 437, 482, 510]Gossmann, H.J. [259, 284, 290, 291, 292,

295, 296, 305, 315, 316, 325, 402, 472, 478, 479, 482, 494, 497, 523]

Gowers, J.P. [287]Grabiec, P.B. [452]Grasserbauer, M. [482]Gravesteijn, D.J. [287, 290, 320, 395,

396, 480, 527, 528, 529]Greim, O. [365, 366, 367]Griffin, P.B. [226, 292, 293, 307, 309,

315, 399, 449, 456, 476, 497, 527]Grünebaum, D. [511]Gu, Q. [245, 300, 301]Guareschi, C. [299]Guerrero, E. [442, 482, 488]Guillemot, N. [230, 231, 301]Guo, Q. [300, 301]Gupta, D. [416]Gust, W. [218, 251, 256, 260, 337,

465, 469, 520]Gustin, W. [218, 251, 256, 260,

337, 465, 469, 520]Gutfeld, 272, 326]Gyulai, J. [260, 311]Haddab, H. [226, 268]Haddad, S. [232, 279, 437]Hahn, P.O. [532]Hahn, S. [372]Hale, P. [323, 396]Hallén, A. [193]

Hamaguchi, C. [459]Hanada, R. [224]Hance, R. [286, 302]Hansen, J.L. [483]Hao, C.P. [308]Hara, A. [412]Hara, T. [275]Hari, P. [372, 380]Harith, M.A. [404, 468]Hart, C. [323, 396]Hartick, M. [392]Hasegawa, T. [266]Hashimoto, F. [262, 263, 343]Hashimoto, K. [332, 334, 390, 391,

409, 513, 517]Hashimoto, S. [459]Hauber, J. [251]Hautojärvi, P. [97]Hayafuji, Y. [498]Hayashi, K. [295]Haynes, T.E. [315, 316, 482]He, D. [271]Heggie, M.I. [498]Heinrich, M. [326]Heinz, T.F. [381]Heiser, T. [339, 390, 394, 407]Herion, J. [361]Herrero, C.P. [45, 356, 363, 371, 376]Herring, C. [347, 355, 360]Hetrick, J.M. [340]Hieber, K. [229, 278, 436]Higuchi, T. [329, 429]Hill, C. [242]Hilliard, J. [340]Hirai, I. [412]Hirai, K. [306]Hiramoto, H. [504]Hirano, H. [287]Hirao, T. [239, 297]Hirose, M. [511]Ho, K.J. [311]Hobler, G. [255, 458]Hocine, S. [513]Hockett, R.S. [329, 413, 429]Höfer, U. [381]Höfler, A. [319]Hofsäss, H. [406]Hoglund, D.E. [402]Holland, O.W. [322]Holldack, K. [298]

Author Index

544

Holloway, P.H. [462]Holm, B. [471, 472]Honeycutt, J.W. [486]Hong, J. [300]Horz, M. [260]Hosoki, S. [81, 266, 500, 501, 502]Houghton, D.C. [270]Hourai, M. [395]Houssaini, S. [346]Hoyt, J.L. [230, 294, 310, 526, 528]Hsieh, C.J. [292, 449, 456]Hsieh, J.C. [283]Hsieh, T.Y. [231]Hsu, C.C.H. [308]Hu, C.M. [428]Hu, M.S. [301]Hu, S.M. [286, 486]Hua, X.M. [304]Huang, J. [298]Huang, M.B. [315]Huang, R.M. [280, 404]Huang, R.S. [280, 404]Huang, T.H. [306]Huang, Y.X. [428]Hugsted, B. [399]Huh, J.Y. [330]Huisman, W.J. [396]Huizing, H.G.A. [277, 494, 532]Ichikawa, M. [81, 500, 501, 502]Ichimiya, A. [514]Iguain, J.L. [506]Ikeda, N. [481]Ikeda, T. [446]Inada, T. [287]Ionova, I.V. [381]Ishigami, S. [310]Ishikawa, T. [471]Ishikawa, Y. [234, 285, 308, 327,

448, 456, 489]Ishiwata, O. [222, 396]Ishiyama, K. [514]Ismail-Beigi, S. [495]Isobe, T. [391]Isolde [406]Ito, R. [399, 476, 480]Itoh, K. [237]Itoh, T. [412]Itoh, Y. [262]Ittermann, B. [302]Iyer, S.S. [288, 300, 398, 465, 484]

Jabara, F.S. [282]Jaccodine, R.J. [272, 285, 288, 298, 299]Jackman, J.A. [270]Jackman, T.E. [270]Jackson, W. [356, 377]Jackson, W.B. [351, 352, 362, 369,

370, 371, 379, 381]Jacob, K. [440]Jacob, M. [471]Jacobson, D.C. [218, 248, 250, 251,

256, 284, 291, 305, 315, 316, 337, 465, 469, 479, 482, 520]

Jacques, A. [236, 297, 444, 483]Jacowitz, R.D. [528]Jagdhold, U. [276, 529]Jäger, H.U. [324, 457]Jahn, S.G. [406]Janot, C. [403]Janssen, K.T.F. [303, 304, 320]Jaraíz, M. [315, 316, 325, 533]Jeng, N. [436, 450]Jeng, S.P. [386]Jeppson, K.O. [242]Jesson, D.E. [402]Jiang, Z. [420]Joannopoulos, J.D. [329]Johnson, N.M. [347, 355, 360, 364, 365,

367, 371, 381]Jones, K.S. [293, 294, 307, 313,

314, 325, 450]Jos, H.F.F. [304]Juang, M.H. [303, 305]Jüngling, W. [482]Kabza, H. [233, 280]Kachurin, G.A. [287, 324, 461]Kahora, P. [298]Kahora, P.M. [282]Kaim, R.E. [303]Kakalios, J. [352, 369]Kakimoto, K. [507]Kakinuma, H. [452]Kalbitzer, S. [269]Kamins, T.I. [528]Kaminski, A. [233, 280]Kamiura, Y. [343]Kandel, D. [506]Kanemura, T. [233]Kankeleit, E. [392]Kariotis, R. [502]Kasatkin, V.V. [338, 434]

Author Index

545

Kase, M. [272]Kassing, R. [229, 278]Kastl, R.H. [316]Kasymov, A.A. [519]Kato, K. [319]Kawabata, K. [511]Kawado, S. [498]Kawai, Y. [310]Kawamoto, K. [219]Kawashima, I. [284]Kaxiras, E. [495, 496, 498, 506]Kazarinov, J.N. [261, 310]Kazor, A. [426]Kazuchits, N.M. [384]Kechang, S. [270]Keck, B. [392]Keeffe, M.E. [499]Keller, R. [338]Kelly, P.J. [503]Keloglu, O.J. [265]Keming, W. [308]Kemp, M. [358, 379]Kennel, H.W. [446, 486]Kerkow, H. [298]Keskitalo, N. [193]Khait, J.L. [323, 452, 487]Kharchenko, V.A. [253]Khatri, R. [245]Khodzaev, K.K. [391, 465, 468, 508]Khodzhaev, K.K. [219, 338, 395, 434, 511]Khor, K.E. [504]Kikuchi, Y. [272]Kim, C.K. [282]Kim, J.G. [282]Kim, J.Y. [361]Kim, K.M. [272, 326]Kim, K.W. [341]Kim, U.S. [285]Kim, Y. [238, 295, 299, 361, 387, 440]Kim, Y.M. [286, 301, 302]Kimura, M. [272]King, C.A. [527]King, J. [242]King, J.R. [228, 277]Kinoshita, H. [302, 306, 326]Kirihata, H. [222, 396]Kirino, Y. [295]Kiritsa, V.L. [429]Kirk, H.R. [245]Kirmse, K.H.R. [367]

Kisielowski-Kemmerich, C. [377]Kitagawa, H. [334, 390, 417, 513, 517]Kitano, T. [431]Kiyota, Y. [287]Kizilyalli, I.C. [405]Klappe, J.G.E. [306]Klauser, T. [289, 481]Kleiner, J. [499]Klyachkin, L.E. [289, 441]Kobayashi, I. [327]Kodiyalam, S. [504]Koiwa, M. [410]Komarov, F.F. [237]Kondo, M. [363]Kook, T. [285]Korlyakov, D.N. [428]Korol, V.M. [414]Korshunov, F.P. [253]Kostishko, B.M. [439]Kotnala, R.K. [345]Kotov, E.V. [237]Kotov, I.N. [265]Koung, C.H. [283]Kovalchuk, V.B. [429]Koveshnikov, S.V. [255, 343, 391]Kozlov, J.I. [518]Kozlovskii, V.V. [261, 310, 445]Krausch, G. [405]Krauser, J. [359]Kringhøj, P. [307, 478, 483, 484,

509, 510, 530]Krishnamoorthy, V. [313, 314, 450]Kristjansson, S. [407]Kroll, U. [366]Krueger, M. [502]Krüger, D. [276, 438, 529]Krygowski, T. [277, 438]Kryukov, V.L. [262]Kubiak, R.A.A. [282, 423]Kubo, S. [476, 480]Kudoyarova, V.K. [391, 465, 468, 508]Kuge, S. [513]Kühnapfel, R. [443]Kujime, S. [219]Kulikov, G.S. [219, 338, 391, 395,

409, 434, 465, 466, 468, 508, 511]Kumagai, M. [306]Kumar, R. [345]Kumar, S.N. [239]Kuo, P. [294, 310, 526, 528]

Author Index

546

Kuranouchi, A. [287]Kurps, R. [276, 438]Kusakabe, S. [369]Kuwano, H. [470, 471]Kuznetsov, A.J. [416]Kwon, Y.K. [471]Kwong, D.L. [231, 286, 301, 302, 306, 326]Kyllesbech, L.K. [531, 532]La Mantia, A. [222, 231, 239, 240,

279, 280, 300, 380]La, F. [222, 227, 240, 268]Labunov, V.A. [240]Lagally, M.G. [401, 499, 502]Lam, Y.W. [321]Lämmel, B. [221]Landi, E. [305]Lannoo, M. [320]Lapiccirella, A. [380]Larikov, L.N. [460]Larsen, A.N. [227, 399, 477, 478, 482,

483, 484, 486, 509, 510, 530]Lasky, J.B. [281]Lassmann, K. [523]Latreche, S. [233, 280]Laubach, S. [392]Laugier, A. [231, 239, 279, 280]Law, M. [313]Law, M.E. [238, 293, 294, 314, 320,

325, 443, 447, 449, 450, 455]Lawther, D.W. [245]Lee, C. [205, 273, 434]Lee, C.L. [311]Lee, J.Y. [283, 341]Lee, M.H. [402, 406, 407, 414]Lee, M.Y. [417]Lee, S.C. [451]Lee, S.T. [420, 421]Lefforge, D. [310, 526]LeGoues, F.K. [311]Lei, T.F. [308, 310, 311]Leihkauf, R. [440]Lerch, W. [469]Leshok, A.A. [384]Leta, D. [343]Levenson, L.L. [224]Lever, R.F. [269, 280, 309, 320, 323]Levkina, T.G. [439]Lewis, R.K. [280]Li, D.Q. [519]Li, J. [248, 359, 403, 417]

Liaw, M.C. [308]Libertino, S. [137, 531]Lightowlers, E.C. [423]Lin, H.C. [310]Lin, W. [424]Linnros, J. [248]Lippens, P. [230, 278]Listebarger, J.K. [294]Litvinova, T.V. [261]Liu, H.W. [303]Liu, J.T. [403]Liu, X.D. [403]Liu, X.J. [403]Livingstone, J. [359]Lo, G.Q. [286, 301, 302, 326]Lo, V.C. [321]Lomasov, V.N. [261, 310, 445]Lombardo, S. [240]Londos, C.A. [427, 430]Lonsjo, O.M. [399]Lopata, J. [345]Lourdudoss, S. [439]Lu, T.M. [218]Lu, Y.T. [505]Luckscheiter, B. [405]Luftman, H.S. [284, 291, 305, 315, 478,

479, 494, 497, 527]Lulli, G. [238]Lur, W. [387]Lusson, L. [364]Lux, G. [232, 278, 279, 283, 290,

437, 438, 452]Lynn, K.G. [388]Lyutovich, A.S. [519]Ma, S. [224, 243, 246, 308, 311,

312, 321, 403, 443, 444, 446,447, 455, 456, 489]

Ma, T.P. [386, 388]Mader, L. [255, 482]Madl, K. [260, 311]Maex, K. [233, 280]Mailhiot, C. [318]Mainwood, A. [330, 498]Majkrzak, C. [343]Majumdar, A. [350]Makhkamov, S. [224]Mäkinen, J. [97]Makogon, J.N. [261]Malkovich, R.S. [466]Malm, D.L. [332]

Author Index

547

Mangin, P. [346, 368]Mannino, G. [137]Manukonda, R. [232, 278]Marchal, G. [346, 368, 403]Marchiando, J.F. [322]Mariani, G. [257, 258]Marioton, B.P.R. [510]Maroudas, D. [432]Marshall, A.F. [230]Mártin, H.O. [506]Martin, S. [243, 321, 443,

444, 446, 447, 455]Martinez, A. [11, 314]Martinuzzi, S. [443, 444, 446, 447]Marwick, A.D. [371]Marynick, D.S. [349]Masseli, K. [229, 278]Massoud, H.Z. [235, 238, 281,

295, 299, 309,387, 440, 453]

Masuda, H. [401]Masuda, K. [363]Maszara, W. [241, 485]Mathiot, D. [11, 243, 265, 314, 321,

354, 375, 455, 490, 513, 519, 526, 530]

Matlock, J.H. [420]Matsuda, K. [239, 297, 306]Matsumoto, S. [25, 246, 308, 311,

312, 456, 489]Matsushita, J. [295]Matsushita, Y. [295]Matthai, C.C. [418, 506]Maximenko, E.A. [460]McCaffrey, J. [270]McQuaid, S.A. [348, 423, 424, 427, 430]Meda, L. [299]Mehrer, H. [218, 251, 469, 474,

511, 521, 522]Mei, S.N. [272, 326]Menikh, A. [361]Menningen, M. [392]Merabet, A. [228, 239, 277, 280]Merli, P.G. [238]Mesli, A. [339, 390, 394, 407]Messoloras, S. [428]Metiu, H. [505]Metzner, H. [302]Meyerheim, H.L. [332]Meyerson, B.S. [311]

Michel, A.E. [281, 298, 316]Miki, K. [382]Milman, V. [402, 406, 407, 414]Milne, R.H. [335]Mimkes, J. [515]Misawa, Y. [378]Mitchell, I.V. [294, 315]Mitra, S. [361, 362]Mitwalsky, A. [233, 280]Miyabo, T. [412]Miyake, M. [270, 283]Miyashita, F. [239, 297, 306]Miyazaki, S. [511]Miyazaki, T. [504]Mizushima, I. [233]Mo, Y.W. [401, 491, 499, 502]Mogi, T.K. [478]Mohr, U. [440]Mohri, M. [452]Moiseenkova, T.V. [253]Molenbroek, A.M. [384]Mollenkopf, H. [420]Moller, K. [314]Monson, T.K. [255, 257]Monty, C. [398]Morehead, F.F. [280, 320, 323]Mori, H. [272]Mori, T. [219]Morigaki, K. [363]Moriya, N. [527]Morin, P. [366]Moro, L. [233]Morooka, M. [260, 262, 263, 264, 530]Mulvaney, B.J. [322, 453, 457]Murakami, K. [363]Muraki, T. [369]Murarka, S.P. [332]Murphy, T.A. [359]Murti, M.R. [461]Myers, S. [323, 396]Nagano, M. [222, 396]Nagao, Y. [481]Nagata, S. [224]Nagel, D. [222]Nagengast, D. [359]Nakajima, H. [410]Nakamichi, I. [234, 285, 308, 327, 448]Nakamura, M. [378]Nakamura, S. [275]Nakamura, T. [287, 306]

Author Index

548

Nakashima, H. [332, 334, 390, 391, 393, 394, 409, 513, 517, 518]

Nakayama, S. [413]Nanu, L. [457]Nanver, L.K. [532]Nason, T.C. [218]Natori, A. [220]Nayfeh, M.H. [340]Nazyrov, D.E. [466, 492]Negrini, P. [234, 236, 285, 297, 447,

448, 450, 481, 485]Neppl, F. [229, 278, 436]Newman, R.C. [329, 348, 420, 423,

424, 427, 428, 430, 532]Newstead, S.M. [282]Newton, D.C. [420]Nichols, C.S. [234, 235, 321, 441, 455, 487]Nickel, N.H. [351]Nielsen, B. [388]Nielsen, K.B. [471, 472]Niimi, T. [308]Ning, X.J. [312]Ninomiya, K. [81, 500, 501, 502]Nishikawa, S. [304]Nishimura, K. [481]Nisnevich, J.D. [236]Nitta, S. [369]Nobili, D. [233]Noël, J.P. [306]Nonomura, S. [369]Norkulov, N. [334, 394, 410, 474, 492]Normand, P. [230, 231, 301]Nosenko, S.V. [343]Novak, S. [301, 326]Novikov, A.P. [237]Noya, J.C. [376]Nozaki, T. [262]Nygren, E. [249, 259]Oates, A.S. [424, 427]Oechsner, H. [384]Oehrlein, G.S. [371]Ogawa, T. [272]Ohta, M. [369]Okamoto, Y. [239, 297]Okazaki, M. [504]Okino, T. [319, 456, 459, 460,

488, 489, 496, 531]Okulich, V.I. [445]Omelyanovskii, E.M. [369]Onai, T. [287]

O'Neil, A.G. [242]Ong, C.K. [505]Onishi, M. [319, 456, 460, 488, 489, 496]Oostra, D.J. [289]Oppolzer, H. [229, 278, 311]Orlov, A.M. [439]Orr Arienzo, W.A. [269, 280]Osada, K. [25, 311, 312Osaka, Y. [379]Oshiyama, A. [244, 460, 490]Ossikovski, R. [366]Ottaviani, G. [299]Otto, G. [239]Oura, K. [219]Overhof, H. [523]Owen, J.H.G. [382]Packan, P.A. [302, 309, 449]Pagani, M. [422]Paine, A.D.N. [530]Pakharukov, J.V. [224]Pakhomov, A.V. [369]Palm, J. [463]Palmetshofer, L. [359]Pan, M.X. [321]Pan, S. [294]Pandey, K. [495]Pandey, K.C. [495, 496, 498]Panitz, J.K.G. [344]Panteleev, V.A. [445]Pantelides, S.T. [234, 235, 321, 374,

380, 381, 431, 441, 455, 487, 495, 496]Panzarini, G. [375]Parisini, A. [233]Park, B. [249]Park, H. [238, 293, 320, 325, 449, 455]Park, K. [232, 278, 279, 283,

290, 378, 437, 438]Park, K.H. [218]Park, M.J. [240]Park, S.H. [334]Park, Y.K. [378]Parker, E.H.C. [282]Parks, J.M. [288]Parrinello, M. [378]Parry, C.P. [282]Payne, M.C. [402, 406, 407, 414]Pearton, S.J. [345, 370, 373, 374]Pease, R.F.W. [230]Pedrotti, M. [425]Pelaz, L. [325]

Author Index

549

Pennetta, C. [369, 374]Pennycook, S.J. [402]Périchaud, I. [259, 441, 443, 444, 447]Perozziello, E.A. [226, 476]Perret, M. [522]Persaud, R. [335]Persheev, S.K. [219, 338, 395, 434, 511]Pesant, J.C. [358]Peters, C.J. [306]Peterström, S. [303]Petravic, M.M. [307]Petrova, V. [361]Pfeiffer, W. [338]Pfiester, J.R. [399]Pfister, J.C. [490]Phillips, P. [530]Pichaud, B. [257, 258]Pichler, P. [289, 471, 477, 481]Pinsard, L. [349]Pirouz, P. [312]Pitkevich, M.V. [261, 310]Pivac, B. [424, 425]Platzer, R. [405]Please, C. [242]Please, C.P. [242]Ploss, R. [477]Plummer, J.D. [226, 292, 293, 302,

307, 309, 315, 442, 446, 449, 456, 476, 486, 497, 527]

Poate, J.M. [218, 248, 250, 251, 256, 259, 284, 290, 291, 292, 295, 296, 315, 316, 325, 337, 404, 465, 468,

469, 472, 478, 479, 482, 494, 497, 520, 523

Podlipko, E.A. [237]Poetzl, H.W. [442, 488]Pohoryles, B. [403]Poisson, C. [253, 469]Polman, A. [337, 384, 385, 396]Poloskin, D.S. [385]Polyakov, A.J. [369]Pötzl, H. [482]Pötzl, H.W. [326]Prigge, H. [532]Priolo, F. [137, 248, 250, 252,

470, 521, 531]Privitera, V. [137, 240, 252, 470,

521, 531, 532]Probst, V. [229, 230, 233, 278, 280, 311]Puga-Lambers, M. [314]

Pukite, P.R. [300]Puschmann, A. [332]Qingtai, Z. [308]Queirolo, G. [299]Quintel, H. [406]Rafferty, C.S. [259, 284, 291, 305,

315, 325, 405, 472, 479, 494, 497, 523]Rahman, F. [318]Raineri, V. [223, 240, 303]Rajarshi, S.V. [373]Rakhimbaev, D. [274]Rakhimbaeva, M.D. [274]Ramamoorthy, M. [431]Ramírez, R. [376]Rausch, W. [298, 316]Rausch, W.A. [309]Ravi, J. [297]Reader, A.H. [396]Reddy, K.V. [461]Regnard, J.R. [308, 483, 484]Reider, G.A. [381]Rek, Z.U. [446]Ren, T.Q. [519]Restle, M. [406]Revenant-Brizard, C. [308, 484]Rhyne, J.J. [346]Rich, T.L. [405]Richardson, W.B. [322, 453, 457]Rimini, E. [137, 222, 227, 240, 268, 300]Rizk, R. [354, 358]Rizzoli, R. [238]Roberson, M.A. [364, 378]Robertson, J. [363]Robinson, H. [293]Robinson, H.G. [294]Roca i Cabarrocas, P. [366]Rockett, A. [270, 503, 507]Rohatgi, A. [277, 378, 438]Roland, C. [506]Rolfe, S.J. [306]Rolland, A. [253]Rollert, F. [218, 474, 511]Romanato, F. [308, 483, 484]Ronsheim, P.A. [296, 298, 316]Roorda, S. [337]Rosenberg, R. [311]Roth, D.J. [293, 442]Rousseau, P.M. [293]Roux, H. [219]Rozgonyi, G.A. [245, 297, 299, 391, 486]

Author Index

550

Rubin, L. [313, 314]Rubloff, G.W. [386, 388]Rüther, R. [359]Ruzyllo, J. [440]Ryssel, H. [256, 289, 468, 470, 471,

473, 477, 481]Saarinen, K. [97]Saccamango, M.J. [272, 326]Sadamitsu, S. [395]Sadoh, T. [334, 390, 393, 513, 517, 518]Sadovnikov, A.D. [317]Sáenz, J.J. [463]Sagawa, M. [511]Saggio, M.G. [222]Saito, H. [295]Saito, S. [224, 431]Sakai, T. [413]Sakauchi, S. [392, 394]Sana, P. [378]Sankara Narayanan, E.M. [69]Santos, P.V. [351, 356, 364, 365,

366, 367, 371]Saris, F.W. [295, 303, 306]Sasaki, A. [395]Sasaki, Y. [237]Sasuga, E. [266]Sato, Y. [284]Schaber, H. [229, 230, 233, 278, 280, 311]Schatz, G. [405]Schimmel, D.G. [241, 282, 448, 485]Schimpf, K. [463]Schlote, J. [438]Schmalz, K. [264, 346]Schmid, U. [264]Schnegg, A. [532]Schölch, H.P. [269]Schork, R. [289, 481]Schreutelkamp, R.J. [303]Schroder, D.K. [334, 428]Schröter, W. [296, 332, 443, 447, 458]Schultz, P.J. [245]Schulze, H.J. [255]Schwalbach, P. [392]Scilla, G.J. [288, 300, 398, 465, 484]Seager, C.H. [344, 347, 360]Sedgwick, T.O. [281, 294]Seebauer, E.G. [400, 405]Seelinger, W. [302]Seibt, M. [458]Sekine, K. [239, 297]

Sellitto, P. [233, 280]Sellmann, R. [515]Selloni, A. [383, 432]Servidori, M. [234, 236, 285, 286,

297, 443, 447, 448, 450, 481, 485]Shabani, M.B. [340]Shaimeev, S.S. [262]Shaligram, A.D. [373]Shasaidov, S.S. [519]Shek, E.I. [385]Shen, T.H. [418, 506]Shi, B.R. [403]Shibata, A. [498]Shibata, Y. [459]Shimizu, N. [533]Shimizu, T. [25, 312]Shimura, F. [329, 338, 416, 423, 429]Shinada, K. [275]Shinar, J. [361, 362, 373]Shinar, R. [361, 362, 373]Shingyouji, T. [310]Shinozuka, T. [224]Shirai, H. [295]Shiraki, Y. [399, 476, 480]Shiratsuki, Y. [511]Shirayev, S.J. [484, 483]Shu, C. [470]Sidorenko, S.I. [261]Sieber, N. [239]Sielemann, R. [392]Sieloff, D.D. [294]Silverman, A. [387]Simons, D.S. [313]Simpson, P.J. [245]Simpson, T.W. [315]Sinke, W.C. [337]Sittig, R. [222]Skudlik, H. [338]Slijkerman, W.F.J. [395, 480]Slinkman, J.A. [294, 325]Smith, D. [381]Smith, P. [237, 437]Smith, T.C. [232, 279, 437]Snyder, L.C. [376]Sobolev, N.A. [253, 385]Soleimani, H.R. [244, 324, 452]Solmi, S. [226, 229, 233, 234, 236,

268, 271, 285, 286, 297, 305, 308, 441, 442, 443, 445, 447, 448, 450,

481, 484, 485]

Author Index

551

Soltamov, U.B. [417]Song, Z. [271, 365]Sopori, B.L. [355, 378]Spaepen, F. [249]Spiecker, E. [458]Spinella, C. [227, 240, 268]Spit, F.H.M. [244, 416, 435, 462, 476, 490]Spjut, H. [507]Spratt, D.B. [231]Srikanth, K. [360, 368]Srivastava, D. [400]Stanis, C.L. [296]Stavola, M. [345]Stegemann, K.H. [319]Stein, H.J. [372]Steiner, D. [338]Stella, A. [424, 425]Stelmakh, V.F. [270]Stemmer, M. [446]Stevie, F. [298]Stevie, F.A. [282, 405, 462]Stewart, R.A. [367]Stewart, R.J. [428]Stich, I. [402]Stingeder, G. [482]Stock, D.M. [533]Stock, S.R. [446]Stöckmann, H.J. [302]Stoemenos, J. [301]Stolk, P.A. [259, 277, 290, 292, 295,

296, 315, 316, 472, 482, 494, 523, 532]Stolwijk, N.A. [218, 251, 253, 469,

474, 511, 521, 522, 523]Stoyanov, S. [401]Strecker, N. [319]Street, R.A. [352, 365, 366, 367, 369,

371, 372, 375, 380]Stutzmann, M. [356, 357, 363, 366]Subrahmanyan, R. [235, 281, 309, 453]Suezawa, M. [392, 394]Sugita, Y. [262]Suh, E.K. [341]Suhara, H. [363]Sukhanov, V.L. [289, 441]Sultanova, N.V. [519]Sulzer, G. [302]Sumino, K. [392, 394]Sumita, S. [395]Sun, H.L. [519]Sun, W.T. [308]

Suprun-Belevich, J.R. [270]Suzuki, H. [369]Suzuki, K. [316]Sveinbjörnsson, E.O. [254, 258, 407, 459]Svensson, B.G. [193, 227, 303, 325, 399,

416, 477, 510]Svistelnikova, T.P. [253]Swartz, L.E. [246]Swartzentruber, B.S. [499, 502, 503]Swartzlander, A.B. [224]Syhre, H. [239, 319]Szeles, C. [388]Tabet, N. [398]Tabikh, S. [233, 280]Taga, Y. [514]Taher, H.I.H. [444]Takagi, T. [224]Takahashi, H. [392]Takahashi, M. [262, 263]Takano, H. [306]Takaue, R. [319, 456, 488, 496]Takayanagi, K. [533]Takeda, R. [295]Takigami, T. [283]Talipov, F.M. [417, 519]Tan, T.Y. [330, 417, 429]Tanaka, A. [304]Tanaka, H. [413]Tang, D. [463]Tang, X.M. [346, 353, 362]Tanigawa, S. [25, 431]Taniguchi, K. [459]Tanimoto, M. [283]Tanishiro, Y. [533]Tanuma, S. [237]Tapfer, L. [251]Tarr, N.G. [306]Tasch, A.F. [301]Taylor, P.C. [372, 380]Tchernyaev, A.V. [317]Tejwani, M. [288, 484]Terakura, K. [504]Terukov, E.I. [391, 465, 468, 508]Theodore, N.D. [261]Theunissen, A.M.L. [529]Tholomier, M. [219]Thompson, M.O. [402, 478]Thompson, R. [362, 371, 377]Tidemand-Petersson, P. [486, 509]Tiller, W.A. [446, 486, 497]

Author Index

552

Tipping, A.K. [329, 420, 532]Tirén, J. [325]Tittel-Helmrich, K. [346]Tkachenko, V.L. [261]Toh, C.P. [505]Tomassini, N. [380]Tomokage, H. [260]Tong, B.Y. [373, 379]Tressler, R.E. [440]Troyanova, G.N. [384]Tsai, C.C. [282, 352, 362, 370, 371,

377, 381]Tsai, J.C.C. [241, 282, 448, 485]Tsai, N.S. [283]Tseng, F.C. [283]Tseng, H.H. [286, 302]Tsoukalas, D. [230, 231, 301, 318, 322]Tsunoda, Y. [231]Tsurushima, T. [393]Tu, K.N. [416]Tucker, J.H. [348, 420, 423, 424,

427, 428, 430]Tulchinsky, D.A. [373]Turan, R. [238, 398, 399]Turner, J.E. [294, 310, 526, 528]Tyschenko, I.E. [287, 324, 461]Uda, T. [504]Ulbricht, S. [457]Umbach, C.C. [499]Unterwald, F.C. [284, 291, 305,

479, 494, 497]Usui, H. [224]Vaisleib, A.V. [256, 265]Vallon, S. [366]Valmorri, S. [226, 268, 308, 484]Van de Walle, C.G. [234, 235, 287,

289, 320, 321, 374, 375, 380, 381, 395, 441, 455, 480, 487, 529]

Van den Heuvel, R.A. [529]Van der Sluis, P. [527, 528]Van der Veen, J.F. [395, 480]Van Gorkum, A.A. [395, 529]Van Ijzendoorn, L.J. [480, 529]Van Pinxteren, H.M. [385]Van Vechten, J.A. [255, 257, 264]Van, A. [320]Van, G.F.A. [287, 289, 320, 395, 480, 527]Van, L. [230, 233, 278, 280]Vandenbossche, E. [243, 320]Vandervorst, W. [295, 306]

Vasin, A.S. [445]Vaysleyb, A.V. [265]Venables, J.A. [335]Venkataraman, V. [350]Vergnat, M. [346, 368]Verner, I.V. [374]Vettier, C. [346, 368]Veuillen, J.Y. [463]Veve, C. [446]Vigliotti, D.R. [272, 326]Visser, C.C.G. [277, 494, 532]Vittadini, A. [383, 432]Vlasov, A.A. [262]Vlieg, E. [396]Vogt, B. [348]Voltovets, N.S. [261]Von, R.J. [272, 326]Vorosov, N.N. [384]Vredenberg, A.M. [284, 291, 479]Vriezema, C.J. [287]Vuong, H.H. [305]Wada, T. [401]Wagner, S. [367]Wahl, U. [406]Walker, A.J. [304]Walker, J. [351]Wan, F.S. [303]Wan, S.K. [519]Wang, A. [232, 279, 437]Wang, C.F. [311]Wang, J. [503, 507]Wang, K.M. [403]Wang, W.H. [418]Wang, W.K. [418]Watanabe, H. [500]Watanabe, M. [222, 239, 297, 396]Weaver, L. [245]Webb, M.B. [499, 502, 540]Weber, J. [346, 353, 362, 365,

366, 367, 532]Wei, G.Y. [292, 456]Wei, J.H. [451]Weidinger, A. [359]Weil, R. [387]Weise, C. [287]Weitzel, I. [229, 311]Wendelken, J. [463]Wendt, A.E. [367]Wendt, H. [229, 278]Weyer, G. [482, 486, 509, 511]

Author Index

553

Whall, T.E. [282]White, C.W. [297]Wichert, T. [338]Wieser, E. [287]Wijaranakula, W. [273, 274, 291, 420,

423, 430, 431, 495]Williams, J.S. [259]Willoughby, A.F.W. [530]Winter, U. [234, 285, 443,

447, 481]Wittel, F. [434]Witthuhn, W. [348]Wittmer, M. [288, 371, 465, 484]Wöhrmann, U. [405]Wong, S.K. [373]Wong, S.P. [321]Wong-Leung, J. [259]Woon, D.E. [349]Wu, C.J. [381, 383]Wu, I.W. [381]Wu, N.J. [220]Wu, X.L. [361, 373]Wu, X.W. [373, 379]Xu, D.X. [306]Xu, J. [314, 450, 463]Xu, T.B. [519]Yagi, K. [533]Yaguchi, H. [399]Yakimov, E. [258, 259, 441, 447]Yakimov, E.B. [255, 343]Yamada, I. [224]Yamagishi, H. [159]Yamaguchi, S. [224]Yamaji, T. [304]Yamamoto, Y. [391]Yamasaki, T. [504]Yamauchi, K. [234, 285, 448]Yan, Y. [300]Yanada, T. [498]Yang, C.S. [308]Yang, G.R. [218, 373]Yang, L. [343]Yang, L.H. [318]Yang, W.S. [258, 417]Yarykin, N.A. [255]Yasunaga, H. [220, 266]Yazaki, K. [327]Ye, H.J. [428]Yokota, K. [239, 297, 306]Yoneta, M. [343]

Yoshida, M. [235, 244, 260, 264, 311, 312, 417, 454, 455, 456, 489]

Yoshikawa, J. [295]Yoshimi, T. [340]Yu, K.C. [205]Yunkin, V.A. [255]Yunusov, M.S. [224]Yusupova, S.A. [409]Zagozdzon-Wosik, W. [452]Zagwijn, P.M. [384, 395, 396, 480]Zainabidinov, S. [492]Zainabidinov, S.Z. [518]Zaitsu, Y. [25, 311, 312]Zaks, M.B. [338, 434]Zalm, P.C. [289, 290, 527, 528]Zamouche, A. [339, 407]Zaumseil, P. [234, 285, 443, 447, 481, 529]Zeindl, H.P. [276, 361]Zeitzoff, P.M. [305]Zellama, K. [366]Zemel, J.N. [241]Zeng, F.Q. [428]Zhai, H.Y. [403]Zhang, F. [271, 365]Zhang, L.H. [313]Zhang, Q.S. [255, 257, 264]Zhang, S.L. [439]Zhang, Z. [505]Zhong, L. [295, 338, 416, 423]Zhonglie, W. [308]Zhu, J. [318, 333, 360]Zhu, P.R. [519]Ziegler, Y. [365, 367]Zimmermann, H. [111, 254, 256, 468,

470, 471, 473]Zini, Q. [442]Zinke-Allmang, M. [401]Zotov, K.I. [429]Zundel, T. [532]

Diffusion in Silicon - 10 Years of Research

Documents