Page 1
HAL Id: jpa-00229590https://hal.archives-ouvertes.fr/jpa-00229590
Submitted on 1 Jan 1989
HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.
CHEMICAL VAPOR DEPOSITION OFREFRACTORY METALS DISILICIDES : A REVIEW
R. Madar, C. Bernard
To cite this version:R. Madar, C. Bernard. CHEMICAL VAPOR DEPOSITION OF REFRACTORY METALSDISILICIDES : A REVIEW. Journal de Physique Colloques, 1989, 50 (C5), pp.C5-479-C5-497.�10.1051/jphyscol:1989559�. �jpa-00229590�
Page 2
JOURNAL DE PHYSIQUE Colloque C5, supplement au n05, Tome 50, mai 1989
CHEMICAL VAPOR DEPOSITION OF REFRACTORY METALS DISILICIDES : A REVIEW
R . MADAR and C . BERNARD*
I N P G , ENSPG, CNRS uRA-1109, Domaine universitaire, B P . 46, F-38402 S t M a r t i n dlH&res, France " I N P G , ENSEEG, CNRS URA-29, Domaine universitaire, B P . 75, F-38402 S t Martin d1H&res, France
RESUME
Les siliciures m6talliques jouent un r61e de plus en plus important en technologie circuits
intEi&s silicium de type VLSI. Parmi les diffkrentes techniques susceptibles dttre utili&s pour obtenir de
manitre reproductibie, des couches minces de ces m a ~ ~ u x de bonne qualitd, le d6p6t chimique en phase
vapeur (CVD) ap-t actuellement comme la plus performante. Le but de cette publication est de passer en
revue les diff6rents proct!d6s, matkriaux et rksultats obtenus jusqu' pdsent par cette mCthode de d6Nt en
technologie VLSI. De plus, les proc&ks rkents susceptibles de rkpondre aux exigences de la future
technologie ULSI seront examinks.
ABSTRACT
The importance of metal siliciats films in current VLSI technology is obvious. Among the
different techniques which can be used to obtain high quality layers of these materials in a reproducible
manner, the CVD emerges as the more promising solution. The purpose of this paper is to review the
various chemical vapor deposition processes, materials and results reported so far for refractory metals
disilicides in today's VLSI technology. In addition, potentially new CVD processes that could meet the future
reeds of the ULSI technology will be explored.
INTRODUCTION
The applications for metal silicides in the field of VLSI technology have increased
dramatically in the recent years. Among the different phases which have been studied, refractory metal
Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphyscol:1989559
Page 3
C5-480 JOURNAL DE PHYSIQUE
disilicides such as WSi2, TaSi2, MoSi2 and TiSi2 have been selected for use as gates and interconnexions
due to their relatively low resistivity and high thermal stability (1-6). These silicides are compatible with
most IC wafer processing, offering good dry etch, adhesion, oxidation and contact properties (7-10).
Substituting polycrystalline silicon by low resistivity refractory metal disilicides in gate metallization can
overcome limitations in size reduction which stem for the high resistance of fine-line, doped ply-Si. At
present, the disilicides films are deposited on poly-Si to shunt the poly-Si gate electrode. To simplify
metallization schemes, this multilayer structure can be replaced by pure silicide film (1 1).
Several physical techniques have been used to form thin films of these silicides : evaporation
of the metal followed by annealing, coevaporation of the metal and silicon, sputtering from the stoichiomerric
compound, cosputtering of metal and silicon. However, by using these techniques, several problems have
been encountered (8-10) such as poor step coverage and irradiation damage caused by electron beams or
sputtering. By comparison, Chemical Vapor Deposition with its advantage of high purity materials, good
conformal coverage, uniform film thickness, possibility of selective deposition, high throughput and low cost
processing is very attractive for the deposition of these materials.
The intent of this paper is to give a general overview of the work already done in the field of
refractory metals disilicides CVD with indications of the present status of current research programs,
especially in the field of selective deposition.
Two aspects of the CVD of metal silicides have been deliberately ignored in this review :
- The formation by CVD at high temperature of thick silicides films for protective coatings
against corrosion since the range of temperatures generally used does not correspond to what can be tolerated
in IC technology.
- The OMCVD of MSi2 phases by pyrolisis of volatile molecular silicon-metal compounds
(12-13) since most of these organic compounds are not available commercially, limiting for the moment the
utilization of this promising technique.
Among the different works which have been reported for the CVD of these materials, the
originality of our approach lies in the systematic use of thermodynamic equilibrium calculations before any
experiment to predict the general trends of our process. Some of these calculations published in an elsewhere
(14,15) are also d e d in this paper together with the experimental results recently obtained.
Page 4
Several different CVD methods are used to deposit metal silicides films. These methods
cover a wide range of processing temperatures and pressures, energy sources to obtain the chemical reactions
involved and reactor designs.
FCK the definition of a CVD process for metal silicides, two parameters have to be fixed :
- The choice of the gaseous metal compound and silicon-bearing molecules which induces
the chemical reactions involved in the process.
- The choice of the chemical vapor deposition system among the three basic types:
atmospheric pressure (APCVD) low or very low pressure. (LFC!VD or VLPCVD) and enhanced LPCVD
methods (Piasma PECVD or Laser LICVD).
As concern the first parameter, we will consider only the deposition of refractory metal
disilicides films based on the reaction of a silicon bearing gas commercially available (e.g., silane,
chlorosilanes) with a gaseous metal compound. For proper compositional control of a CVD film, the thermal
stability (or the stability with respect to hydrogen) of the gaseous metal compound and the silicon-bearing
molecules should be simular (16). According to this crude selection criterion, the most attractive metal
precursors are the metal halides MoC15, MoF6, TaC15, TiCl4, WCk and WFg.
For practical purposes, only those which are easy to vaporize, handle and supply can be
considered for IC manufacturing. Thus, among these metal halides we must consider first Tic14 and WQ
which are Iiquid compounds with vapor pressures at room temperature sufficient to flow directly into a low
pressure CVD reactor. Their flow rates can be easily monitored using mass flow controllers and they can be
supplied with high purity at a reasonnable cost. As a result, CVD reactors and deposition processes for
titanium silicide (17, 18) and tungsten silicide films (7. 19.20) are now commercially available.
Various investigations have demonstrated the feasability of TaSi2 (21.23) and MoSi2 (22,
24) CVD using TaC15 and MoC15 solid sources as the respective metal precursors. However, these reactive
pentachlorides pose problems of corrosion, contamination and reproducibility which have impeded so far their
implementation in device manufacturing. One possible solution that we have studied recently is the
production of the gaseous chlorides in situ by passing chlorine through a pure metal bed (25). The results
obtained by this method in the case of the APCVD of TaSi2 (25) show that at least in this case, the in situ
chlorination process can be well controlled and may produce reproducible amounts of tantalum chlorides
TaC14. TaC15. A similar work is in progress for the CVD of WSi2.
Page 5
JOURNAL DE PHYSIQUE
As concern the second m e t e r , the choice of the chemical vapor deposition technique, it
depends on the nature of the silicide which has to be deposited.
For example, high quality films of TiSi2 have been obtained by APCVD, LPCVD and also
PECVD while in the case of WSi2, starting from WQ and SiH4, a LPCVD technique is needed in order to
avoid gas phase nucleation problems.
PECVD systems have been used successfully for the production of high quality, smooth and
reproducible MSi2 films but the as-deposited layers are amorphous and an annealing step is needed to obtain
the required transport properties. Moreover, PECVD system cannot be used for a selective process.
Thus, it appears difficult to define at first the best system for the CVD of these materials.
However, the actual trends are directed to the use of LPCVD, cold walls, single wafer, systems to avoid
depletion effect during the deposition process.
THERh4ODYNAMIC EQUILIBRIUM CALCULATIONS
As the fist step of CVD experimentation, equilibrium calcutations based on a
thermodynamic analysis of gas systems can provide an invaluable aid, guiding the selection of suitable gas
mixture, temperature and other experimental conditions for the deposition of a given material.
There are two common approaches that can be used for equilibrium calculations. The fist
one is to consider only some chemical reactions and calculate the concentration of the different species.at
equilibrium. Most of the thermodynamic calculations performed in the case of the CVD of refractory metals
disilicides are limited to this approach. Some of the reactions which have been used are the following.
Page 6
The validity of this kind of calculation depends of the choice of the chemical reactions
amsidered.
The second method which is generally p r e f d is a minimization of the Gibbs energy of the
whole system, using the technique already described by many authors and previouly performed successfully in
numerous systems (30-32). With this method, all the conceivable species which can appear during the
chemical reactions involved are taken into account. For a given set of reactants, total pressure and reaction
temperature, the composition of both gaseous and condensed phases at equilibrium can be calculated.
Enws may arise from the use of incorrect thermochemid data or the omission of important
species. So the first step in this method in to list the chemical compounds or species which are likely to be
formed and be present in the system at equilibrium. Then, a judicious selection of the data reported in the
literature and an estimation of unkown data must be done.
The thermodynamic data of the gaseous species which have been used in our calculations
come from the "Scientif~c Group Thermodata Europen (SGTE) and classical references sources (33.36). For
the condensed phases of metal-silicon systems, the selection of the most suitable thermodynamic data was
based on the consistency of calculated and experimental phase diagrams. More often, this thermodynamic
procedure shows that the scarce data available for the different solid phases of the M-Si (M = W, Mo, Ta, Ti)
systems are not self consistent. A critical assessement of the phase diagrams leads to a new set of
thermodynamic data required to start the equilibrium calculations (14,lS). Selected thermodynamic data used
for these calculations are given in Table 1.
Table 1 : Selected thermodynamic data for the four silicides of interest in IC technology
Page 7
JOURNAL DE PHYSIQUE
The thennodynamic equilibrium calculations performed in the different systems investigated
were used to represent the equilibrium conditions during the deposition process in the form of "CVD phase
diagrams". This kind of diagram shows computed phase boundaries for the condensed phases deposited at
equilibrium as function of the experimental CVD parameters chosen. These boundaries separate ranges of
input conditions under which specific condensed phases are computed to be equilibrium products. Example of
these diagrams are given for different metal-silicon system in figure 1 together with the conditions chosen for
these calculations.
Similar mults have been obtained for some of these systems by other groups (28,23) using
different programs based also on the technique of minimization of the Gibbs free energy of the systems.
As examples of the information which may be obtained from these calculations, let us
consider the results for metal chloride-silane-hydrogen systems with similar input conditions as shown in
figure 1 for W, Ta and Ti. The three CVD phase diagrams present numerous domains and the principal
evidence which comes out at first sight is the various extents of pure disilicide deposition domains. The TiSi2
case is quite promising for CVD experiments, while the WSi2 and TaSi2 ones are much more restricted.
An extensive investigation of the thermodynamic equilibrium calculations in these systems
have been reported in an elsewhere (15): We will recall the main conclusions :
- Among the four disilicides WSi2,MeSi2,TaSi2, TiSi2, the latter seems to be the most
suitable for a CVD process.
- The metal chlorides are themiodynamically more appropriate than the corresponding
fluorides as metal sources.
- Replacing silane by dichlorosilane extends in all cases the pure MoSi2 deposition domain.
RESULTS
We will describe successively the reported results obtained so far for the four silicides of
interest in IC technology.
The reported work on the CVD of blanket TiSi2 include APCVD (37,38) LPCVD (37.39,
43) PECVD (37.44) and LICVD (45,47).
Page 8
Ta, SI
Ta Si + Ta
Fig. 1 : Computed equilibrium "CVD phase
diagrams" in the systems M-Si-H-C1-Ar
(M = Ta (a), Ti (b), W(c)) for the following
experimental parameters : total pressure 1 atm,
deposition temperature 7m0C, argon partial
pressure : 0.9
Page 9
JOURNAL DE PHYSIQUE
A general overview comes from the original work of Kemwr et al (37) who have
investigated the synthesis of TiSi2 by APCVD. LPCVD and PECVD. The APCVD of TiSi2 (37) performed
in a temperature range of 800-1000°C and TiC14/SiH2C12 ratios of 1.5 to 4 resulted in rough films with
poor adhesion which was attributed to gas phase nucleation during the deposition process. To improve these
results, they investigated the LPCVD of this compound in a temperature range of 600-800°C, a pressure range
of 0.1-5 torr and a TiCk/SiH4 or TiCldSi H2C12 ratios of 0.1 - 10.
During these experiments, they were the first to point out the selective character of this
process and concluded that the deposition of TiSi2 is inhibited by the presence of a native oxide layer. So in
order to obtain a TiSi2 Film, it is nece*ssary to deposit fvst a polysilicon layer by SiH4 pyrolisis leading to a
polycide structure. There is presently a general agreement on this procedure which has been used by almost alI
groups working on this subject.
They also indicated that the surface roughness of their films was unacceptable and suggested
that it must be possible to obtained better results by an improvement of the vacuum system.
Effectively, soon after, better results were obtained by Reif et al (26, 41, 43) using a
LPCVD cold wall reactor by a careful optimization of the deposition conditions. Smooth, reproducible low
resistivity titanium disilicide films were obtained at a pressure of 67 mtorr, a temperature of 730°C and a
SiHfliC14 gas flow rate ratio of 20/2. They also demonstrated that the structure of the deposited films was
extremely sensitive to the reactive gases flow rate ratios.
LPCVD of TiSi2 has also been reported by Bouteville at a1 (39). using direct reaction of
Tic14 with the silicon substrate in the presence of hydrogen in the temperature range 700-1O0O0C and a total
'pressure of 0.75 torr.
However, the influence of the total pressure parameter is still unclear, since we have
obtained results as good as those just I'nentio~ed by APCVD in a cold wall reactor working in a temperature
range of 600-900°C and a TiC14fSQ ratio of 0.5 to 1 (38). This experimental work was based on the results
of thermodynamic equilibrium calculations in the Ti - C1 - Si - H - Ar system as mentionned previously.
High quality polycrystalline films with good surface smoothness were obtained by sequential deposition of
amorphous silicon and titanium disilicide at a low growth rate (2 pm/h) which was achieved by decreasing
both the deposition temperature ( 6 600°C) and the partial pressures of the reactive gases. The resistivity of
the as deposited films is in the range 15-25 JLQ x cm even for this low deposition temperature. Thus no
supplementary annealing step was needed
Page 10
Another appoach for the synthesis of TiSi2 films is the use of the PECVD technique. The
fitst experiment was made also by Kemper et al(37) in a parallel plate experimental reactor at a frequency of
300 kHz. The deposition occurs in a temperature range of 300-350°C and TiClqlSiHq gas flow rate ratio of 1
to 2. The as deposited films were reported to be smooth but amorphous and highly resistive. An annealing
step at 750°C for one hour was needed to reduce the resistivity to 20 pi2 x cm.
Following these first attempts, a PECVD technique was also used by Rosler et al (44) to
deposit TiSi2 films at a temperature of 450°C, a frequency of 50 kHz and SiHq/TiQ ratio between 2.75 and
5.00. An annealing step at 600 -650°C was also required to crystallize the as deposited amorphous films and
to lower the resistivity to its usual value 15-20 pC2 x cm. A complete characterization of TiSi2 films
obtained by this technique has been made by Mcrgan et al(48).
This process has been improved recently by Hara et al (49. 50) using a horizontal type
hot-wall plasma CVD reactor at 450°C, a pressure of 1.95 tom and an rf frequency of 50 kHz. By varying the
TiClq/SiHq gas flow rate ratio from 0.23 to 0.09, they were able to change the composition of the films
from 1.1 to 2.0. They have also demonstrated the importance of the annealing process for the stability of the
TiSi21Si interface.
To complete this review of the work done in the field of the CVD of TiSi2lSi polycide
structure, we must briefly mention the results of the two LICVD investigations reported so far even if they
are not especially good.
Gupta et al(45,46) used an excimer laser to initiate a gas phase reaction process through the
absorption of the laser energy by the Tic14 molecule, leading to the deposition of titanium silicide films at a
substrate temperature above 350°C. The mostly amorphous as deposited films required to be annealed at a
temperature of 650-700°C to reduce their resistivity.
A somewhat different LICVD pyrolitic process was reported by West et al(47) using a CO;?
laser. The deposition was carried out at a pressure of around 6 torr with a TiC14/SiH4 partial pressure ratio
varying fn>m 0.04 to 0.4. The microstructure and the transport properties ot the films were studied as function
of the substrate temperature in the range 400-50O0C. An annealing step at 800OC was also required to obtain a
low resistivity.
The reported works on TaSi2 include the exclusive use of tantalum chlorides and
Page 11
C5-488 JOURNAL DE PHYSIQUE
silicon-bearing molecules in the three basic systems : APCVD (22, 25). LPCVD (40, 23, 51, 55), and
PECVD (56).
The fmt work on the CVD of this material for IC manufacturing comes from Lehrer et a1
(5 1,52) using a multi-wafer LPCVD hot-wall reactor in a temperature range of 500-650°C, a pressure of 0.3
torr and various TaC15/Sq gas flow rate ratios. A polycide structure was produced by sequentially depositing
ply-Si and Ta5Si3 in a single process. The Ta5Sig was then converted to TaSi2 by heat treatment either
outside (51) ot inside the reactor (52).
Similar processes and results have been described by Williams et a1 (53) and Widmer et a1
(54). However, using the same kind of LPCVD system, Bouteville et al(40) reported the direct deposition of
TaSi2 on silicon substrate starting from the same chemical species and a similar temperature range. No
satisfactory explanation of this descre~ancy has been propcmd so far.
Using now a cold-wall single wafer LPCVD system, a TaC15 solid source, silane and
hydrogen, Wieczorek el a1 (28) demonstrated that TaSi2 fdms could be formed directly in a temperature range
of 630 to 750°C at a deposition rate of 70 nm.mn-* with a minimum resistivity of 60 x cm, after
annealing for one hour at 900°C in argon.
A similar process has been described by Reynolds (23) using a mixture of S a . TaC15, H2.
HCI and Ar at pressure between 100 mtorr and 1 torr over the temperature range 620-700OC. The
compositions of the films could be controlled by adjusting the deposition temperature, total pressure and the
partial pressures of reactant gases. A thermodynamic analysis of the deposition process was performed but the
results have not been published so far.TaSi2 films with good adhesion were deposited on Si and SiO;! and as
deposited resistivity values as low as 55 pi2 x cm were achieved at a deposition temperature of 700°C.
A different approach has been used by Hieber et al(56) to deposit tantalum silicide films at
low temperature in a PECVD system. An inductively generated plasma was applied to a mixture of TaCl5,
SiH2C12 and H2 at an overall pressure of about 2 mbars and a substrate temperature above 400°C. Up to a
substrate temperature of 540°C the films were amorphous after deposition whereas above 580°C crystalline
TaSi2 has been produced with a resistivity of about 70 pS2 x cm. A further drop of the resistivity to about 55
pQ x cm was obtained aftefanneding these films 1 hour at 900°C in argon.
Starting from an equilibrium thermodynamic calculations of the CVD diagrams for the
system Ta - C1- H - Ar - Si, we have recently defined the experimental conditions leading to TaSi2 system by
APCVD (25). Based on these calculations, high quality TaSi2 films have been deposited on both Si and Si%
Page 12
in cold wall restor working at atmospheric pressure, using sitane and tantalum chlorides produced in situ by
the reaction of chlorine on tantalum maintained at 550°C (figure 2). Experiments were carried out at
temperaNres between 600 and 850°C. Below 680°C, no deposition occurred. Gas phase compositions were
chosen in order to obtain films with different Sfla ratios from 0.7 (TaSi2 + Ta5Si3) to 2.6 (TaSi2 + Si)
according to the calculated phase diagrams. Deposited films of TaSi2 are 200-2000 nm thick, have a surface
roughness of below 15 nm, a plane substrate-silicide interface and a resistivity lower than 70 p.Q x cm after
annealing at 900°C for 2 hours.
Fig. 2 : Schematic =presentation of the APCVD
experimental set up for TaSi2 : 1-quartz tubes,
2 graphite susceptor, 3 rf heating, 4 resistance
heating, 5 Ni-Cr thermocouples, 6 tantalum bed,
7 gas inlet, 8 gas mixing, 9 gas outlet,
10 gas products neutralization, 11 vacuum
A r . S i H q
A r
w e n disilicide : WS12
Films of tungsten disilicide have been obtained by APCVD (29, 57, 58) LPCVD (7, 19,
20. 59-64) and PECVD (65) using almost exclusively a mixture of W ' - SiH4 and H2 with or without a
neutral carrier gas. Only one study of the CVD of WSi2 starting from WCl6 has been reported so far by
Lehrer et a1 (57). Bilayer films of Si and WSi2 were obtained at 600°C by sequential deposition of Si and W
using S I H ~ / N ~ and WCb/H2 mixtures at aunospheric pressure. During the W deposition, a crystalline WSi,
layer grows on the Si layer with grain size in the order of 10 nm. After anneal, the grain size is 30-50 nm and
the elecmcal resistivity as low as 65 pi2 x cm.
Page 13
JOURNAL DE PHYSIQUE
However, the fmt reported work on the APCVD of WSi2 by Lo et al(29) makes already
used of the commercially available and easy to handle tunsgten hexatluoride WFg.
Just after Lehrer's work, was reported a study of the deposition of WSi2 by PECVD (65)
using also a mixture of WFg and SS-Q in a rf parallel plate reactor at 13.56 Mhz, a substrate temperature of
230°C and a pressure of 0.5 to 0.7 Tom. The atomic ratio of W/Si in the film depended on the W i H 4 gas
flow rate ratios. Most of the deposited films were amorphous and a high temperature annealing step was used
to lower their resistivity.
Soon after Brors et al (I 7,19,20) reported the deposition of WSi2 in a cold wall multiwafer
LPCVD system. Deposition were done on Si and SiO;! in a temperature range of 300 to 450°C and a pressure
range of 0.1 to 0.3 ton. In these ranges, the deposition rate was found to be independent of temperature and
pressure, independent of SiH4 flow rate ; however, a linear function of WF6 flow rate. Any change in the
W W i H 4 gas flow rate ratio causes a corresponding change in the Si to W ratio in the deposited films.
This can be understood easily, looking to the CVD phase diagram calculated recently for this
system (19, and represented in figure 3. The WSi2 domain is reduced to a single line ; so, any deviation of
the WFdSiH4 gas flow rate ratio from the ideal composition corresonding to WSi2 must produce mixed
phases films, going from W5Si3 + WSi2 to WSi2 + Si.
Fig. 3 : Computed equilibrium "CVD
phase diagram" in the system
W-Si-H-F-Ar for the following
experimental parameters : total
presslm 1 atm, deposition
t e m p e ~ 720°C, argon partial
pressure : 0.9
Page 14
Despite this problem, this LPCVD process has been considered as a reference in the field of
CVD of WSi2 and most of the works reported after concerned almost exclusively the charadterization of films
obtained by this technique (64,66-68).
There have been only few reported studies of the CVD of MoSi2 for IC manufacturing, all
using a LPCVD system.
The first report comes from Inoue et al(24) using a LPCVD hot wall multi wafer reactor at
a pressure of 0.6-2 Torr and in a temperature range of 520-600°C. Silane was used as the silicon source and
MoCl5 as the metal source. MoSi2 films were deposited at low temperature (670°C) onto oxidized Si
substrates. An annealing step at temperatures between 700-1000°C was needed to reduce the resistivity of the
asdeposited films to its usual value (60-100 pi2 x cm ).
The other reports on the LPCVD of MoSi2 (27,62,69,70) deal with the reaction of MoF6
with silane in the presence of hydrogen in cold wall reactors. The LPCVD of molybdenum silicides was
observed at temperature as low as 150°C, some 200°C lower than LPCVD WSix. The reported results are
similar to those obtained previously for WSi, made by the reaction of WFg and SiH4 with at least one main
difference : the composition of the deposited films depends only on the deposition temperature for a wide range
of MoFd S1Hq gas flow rate ratio.
We have not performed yet thermodynamic equilibrium calculations in the system Mo - Si -
F - H - Ar. But for obvious reasons, we suspect that the CVD phase diagram for this system must be very
similar to the one calculated for W - Si - F - H - Ar system. The reported difference in the sensitivity of the
deposited film composition to the reactive gases flow rate ratio may be attributed to some kinetic effect which
must be studied to get a better understanding of these processes.
FUTURE DIRECTIONS IN CVD OF METAL SILICIDES
Today VLSI technology is moving into the ULSI era in which the feature size of practical
devices must be well into the submicron range. Like the CVD of dielectrics (71), the CVD of refractory metal
silicides must make some accomodations as well.
The results reported so far in this paper concerned the deposition of "blanket silicide" mostly
for polycide structure in conjunction with a predeposited ply-silicon layer. As regards to the materials
Page 15
JOURNAL DE PHYSIQUE
requirements for the new generation of ULSI devices, it must be extremely important to use silicides also as
contacts for the m e and drain regions in addition to the gate. One solution which is investigated is the
salicidation process in which the metal deposited by a PVD technique is reacted onto Si/Si% patterned wafer
to form the metal silicide only on the bare silicon surface. This has been achieved for example with TiSi2
(72). However, this method requires in this case a titanium and titanium nitride strip and an annealing step
after the TiSi2 film is formed. Moreover the salicide process implies the consumption of the underlying
silicon and thus is not compatible with the fabrication of shallow junctions.
This salicide process may be replaced in the near future by selective CVD of the metal
silicide. Several attempts have been made in order to set up a selective deposition process of TiSi2 based on
the silicon reduction of Tic4 in the presence of hydrogen (21). But there was still a consumption of silicon
from the substrate which may cause reliability problems. One solution to this problem which is presently
investigated is the use of a reactive mixture containing a silicon bearing molecule.
Two selective deposition processes of TiSi2 have been reported recently, using both a
mixture of Tic14 and SiQ. The f i t one reported by Ilderem et al(73.74) makes use of a thin polysilicon
layer deposited prior to the silicide in a cold-wall very low pressure CVD reactor. It is shown that selective
deposition is possible by controlling the polysilicon and the titanium silicide deposition times. But even at a
deposition temperature as low as 730°C. 40 % of the silicon in the titanium silicide film originates from the
consumption of the underlying silicon substrate. The process is based on the etching and/or consumption of
the thin polysilicon layer due to the deposition chemistry which has not been explained further yet.
In fact, all the results obtained so far in the selective deposition of this material by CVD
may be explained by the presence of a native oxide layer on the surface of the silicon substrate which prevents
the nucleation of TiSi2. Based on this assumption, three solutions may be considered for a selective
deposition of this material :
- A ciassical etching step prior to deposition, by reacting the substrate with a gas mixture
of H2/HCl at 1 100°C.
- The selective deposition of silicon prior to the deposition of TiSi2 in a LPCVD reactor,
leading to a selective polycide structure.
- Or the solution we have investigated, namely the use of an appropriate choice of the
reaction gas (TiClq, SiHq, Hz, Ar) for an in situ etching of the native oxide at the beginning of the
deposition process (74,75).
Page 16
Indeed thermodynamnc equilibrium calculations perfomed in this system show that a slight
excess of silane in comparison with the composition leading to the formation of TiSi2 converts SiO2 into
volatil SiO with a significant yield above 900°C at atmospheric pressure (75). Starting from these
calculations, selective deposition of I'iSi2 has been obtained recently at temperature as low as 700°C, in
conformation with the design down to 1 pm size patterns (74). An example of selective deposition on
panerned wafer is given in figure 4.
Fig. 4 : SEM micrograph of selective
titanium disiilicide &posited on
patterned silicon wafer. TiSi2
line width : 4 pn.
Successful selective deposition of metal silicide was also obtained by Hieber et al in the case
of TaSi2 (76,77) according to the reaction chemistry :
This method requires adjusting the substrate deposition temperature and composition of the
reaction gas to values at which silicide nucleation in regions of the substrate other than silicon regions is
suppressed during deposition from the gaseous phase due to the presence of the hydrogen halide. The
deposition ternperahire were between 600OC and 800°C and the deposition pressures ranged from 0.5 mbar to 1
mbar. No consumption of underlying silicon was detected by TEM examinations. This process may be of
great potential in the metallization of shallow junctions in submicron IC structures, once solved the problem
of the surface roughness of the deposited films.
Page 17
JOURNAL DE PHYSIQUE
The authors are grateful to J.F. Million-BW, C. Vahlas, E. Mastromatteo. E. Blanquet. J.
Torres and J. PaIIeau for their helpful contribution to this work and to "the Centre National d'Etude des
Tt5lhtnmunications" and the "Groupement Circuits Int6gn5s Silicium" for their financial s u p p a
REFERENCES
/I/ R.H. Dennard, F.H. Gaensslen, M. Yu. V.L. Rideout, E. Bassous and A.R. Leblanc, IEEE. J. Sol. Stat.
Circuits. SC-9, (1976) p 256
/Ll S.P. Murarka, J. Vac. Sci. Technol., 17, (1980) p 769
Dl L. Mohammadi. Sol. Stat. Technol. 1, (1981) p 65
/4/ J.D. Meindl, K.N. Ramakumar, L. Gerzerbg and K.C. Saraswat, in "Int. Sol. Stat. Circuits Conf." L.
Winner, Edit, IEEE, New York (1981) p 36
/5/ F.M. D'Heurle, Proc. ECS 1982, VLSI Sci. Tech., (1982) p 194
/6/ S.P. Murarka, "Silicides for VLSI applications", A d . Press, New York (1983)
/7/ K.C. Saraswat, D.L. Brors, J.A. Fair, K.A. Monnig and R. Beyers, IEEE Trans. Elec. Dev., ED-30,
(1983) p 1497
/8/ B.L. Crowder and S. Zirinsky, IEEE Trans. Elect. Dev., ED-26, (1979) p 369
P/ A.K. Sinha, J. Vac. Sci. Techn., 19, (1981) p 778
/lo/ T.P. Chow and AJ. Steckl, IEEE Trans. Elec. Dev., ED-30, (1983) p 1480
/I I/ Y. Pauleau. Solid State Technol., vol. 30 (4) (1987) p 155
1121 B.J. Aylett and RM. Colquhoun. J. Chem. Soc. Dalton Trans., (1977) p 2058
1131 B J. Ayleu and A.A. Tannahill, Vacuum., 35, (1985) p 435
1141 C. Vahlas, E. Blanquet, P.Y. Chevalier, Calphad-XVII. University of California (Berkeley), July 10-15,
1988
/15/ C. Bernard. C. V*, J.F. Million-Brodaz, R. Madar, m. loth Int. Conf. on CVD., G.W. Cullen
Edit, Electrochem. Soc., Pennignton NJ (1987) p 700
1161 W.A. Bryant, G.H. Meier, Proc. sth Int. Conf. on CVD., J.M. Blocher Jr, HE. Hintermann and L.H.
Hall Eds, Electrochem. Soc., Pennington NJ (1975) p 161
1171 D.G. Hemmes, J. Vac. Sci. Technol., B, vol. 4, (1986) p 1332
/18/ R.S. Rosler, G.M. Engle, J. Vac. Sci. Technol., B, vol. 2, (1984) p 733
1191 DL. Bmrs, J.A. Fair, K.A. Monnig, K.C. Saraswat, Solid. Stat. Technol., vol. 26, (1983) p 183
/20/ DL. Brors, J.A. Fair, K.A. Monnig, K.C. Saraswat, Proc. 9th Int. Conf. on CVD, Mc D. Robinson et
a1 Eds, Electrochem. Soc., Pennington NJ (1984) p 275
Page 18
Dl/ A. Bouteville, A. Royer, J.C. Remy, J. Electrochem. Soc., vol. 134, (1987) p 2080
/22/ D.E.R. Kehr, k. 6th Int. Conf. on CVD, L.F. Donaghey, P. Rai-Choudhury, R.N. Tauber
Eds, Electrochem. Soc., Pennington NJ (1977) p 51 1
/23/ GJ. Reynolds, Proc. Symposium on Multilevel metallization, Interconnexion and contact technologies,
L.B. Rothmann, T. Hemdon Eds, Elecwchem. Soc., P h g t o n NJ (1987) p 39
I%/ S. Inoue, N. Toyokura, T. Nakamura, M. Maeda, M. Takagi, J. Electrochem. Soc., 130, 7, (1983) p
1603
/25/ E. Blanquet et al, To appear in Thin Solid Films (1989)
/26/ P. Tedrow, V. Ilderem and R. Reif. Appl. Phys. Lett., (1985) p 1
/27/ PJ . Gaczi, Proc. Symposium on Multilevel metallization, Interconnexion and contact technologies,
L.B. Rothmann, T. Hemdon Eds, Electrochem. Soc., Pennington NJ (1987) p 32
/28/ C. WiecuKek, Thin Solid Films., 126, (1985) p 227
/29/ J.S. Lo, R.W. Haskell, J.G. Byrne and A. Sosin, Proc. 4th Int. Conf. on CVD., G.F. Wakefield and
J.M. Blocher, Eds., the Electrochem. Soc., Pennington NJ (1974) p 74
/30/ P. Vay, PhD Thesis (Grenoble, FRANCE) (1971)
131/ G. Eriksson, Acta Chem. Scand., E, (1971) p 2651
/32/ C. Bernard. Ptoc. 8th Int. Conf. on CVD. J.M. Blocher Jr, G.E. Vuillard and G. Whal Eds. The
Electrochem. Soc. Pennington, NJ (1981) p 3
/33/ Janaf, Thermochemical Tables, 2nd Edit. @ow. Chem. Co, Midland MI) (1971) and Suppl. 1975,1978
f34/ 0. Kubaschewski, 0. Kubaschewski-Von Goldbeck, P. Rogl. M.F. Granzen, Atomic Energy
Review-Titanium : Physico-Chemical properties of its compounds and alloys, Special Issue no 9 (1983)
/35/ R. Multgren, Selected values of the thermodynamic properties of binary alloys. American Society of
Metals (1973)
/36/ I. Barin and 0. Knacke, Thermochemical properties of Inorganic substances (Springer, Berlin 1973 and
1977)
f371 MJ.H. Kemper, S.W. Koo and F Huizinga, Roc. of the 5th symposium on Plasma Processing., G.S.
Mathad, G.C. Schwarb and G. Smolinsky Eds., 85-1 (1985) p 285
/38/ J.F. Million-Brodaz, C. Vahlas, C. Bernard, J. Torres and R. Madar, Proc. of the 6th European Conf. on
CVD, R. Porat Edt., (1987) p 280
/39/ A. Bouteville, A. Royer and J.C. Remy, J. Electrochem. Soc., 134 (8). (1987) p 2080
1401 A. Bouteville, A. Royer and J.C. Remy, Proc. of the 6th European Conf. on CVD, R. Porat Edt. (1987)
P 264
1411 R. Reif, P. Tedmw and V. Ilderem, US Patent no 4,666,530 M a y 1987
/42/ V. Ilderem. J. Lee and R. Reif, Ext. Abstracts, Electrochem. Soc. Fall Meeting (Hawdi, October 1987)
87-2, p 1467
Page 19
C5-496 JOURNAL DE PHYSIQUE
1431 V. Uderem and R. Reif, J. Electrochem. Soc., 135.10, (1988) p 2590
14-41 R.S. Rosler and G.M. Engle, J. Vac. Sci. Technol. B., 2 (4). (1984) p 733
1451 A. Gupta, G.A. West and K.W. Beeson, Ext, Abstracts, Electrochem. Soc. Spring Meeting (Toronto,
CANADA, 1985) 85-1, p 390
1461 A. Gupta, G.A. West and K.W. Beeson, J. Appl. Phys., 58 (9), (1985) p 3573
1471 G.A. West, A. Gupta and K.W. Beeson, Appl. Phys. Lett., 47 (5), (1985) p 476
1481 AE. Morgan, W.T. Stacy, J.M. De Biasi and T.Y.J. Chen, J. Vac. Sci. Technol., B4 (3). (1986) p 723
1491 T. Hara, Y. Ishizawa. H.M. Wu, D.G. Hemmes, R.S. Rosler, Proc. loth Int. Conf. on CVD, G.W.
Cullen Edit, Electrochem. Soc., Pennigton NJ (1987) p 867
1501 T. Hara, Y. Ishizawa, H.M. Wu, D.G. Hemmes, R.S. Rosler, Thin Solid Films, 157 (1988) p 135
/51/ W.I. Lehrer, J.M. Pierce, E. Goo, S. Justi. Proc. lSt Int. Conf. VLSI Science and Technol.,
Electrochem. Soc., Pemington NJ (1982) p 258
1521 W.I. Lehrer and J.M. Pierce. Ext. Abstracts 164th Meeting of the Electrochem. Soc., Abstract no 284,
83-2, (1983) p 440
1531 D.S. Williams, E. Coleman, J.M. Brown, Thin Solid Films., 133 , (1986) p 2637
1541 AE. Widmer, R. Fehlmann, Thin Solid Films., 138, (1986) p 131
/55/ K. Hieber and F. Neppl, Thin Solid FiIms., 140, (1986) p 131
1.561 K. Hieber. M. Stoh, C. Wieczorek, Proc. 9th Int. Conf. on CVD. Mc D. Robinson et al. Eds,
Electrochem Soc., (1984) p 205
/57/ W.I. Lehrer and J.M. Pierce, "Semiconductor Silicon 1981". Proc. 4th Int. Symp. on Silicon Materials
Science and Technology., H.R. Huff, R.J. Kiegler and Y. Takeishi Eds., the Electrochem. Soc.,
Pennington NJ (1981) p 588
/58/ K.C. Saraswat and F. Mohammadi, Ext. Abstracts of the Electrochem. Soc. Spring Meeting., Saint
Louis, Missouri (1980) p 419
1591 S. Sachdev and T.T. Doan, Workshop on Refractory Metal Silicides for VLSI 11, U.C. Berkeley, May
14-17 (1984)
1601 K. Monnig, Workshop on Refractory Metal Silicides for VLSI 11, U.C. Berkeley, May 14-17 (1984)
1611 K.Y. Ahn and S. Basavaiah, Thin Solid Films., 118, (1984) p 163
1621 S. Tokuhara, A. Takamatsu. S. Moribe, H. Sakai, T. Yoshimi, Ext. Abstracts of the 170th Meeting of
the Electrochem. Soc., San Diego, CA, oct. 19-24 (1986) vol. 86-2, p 398
I631 Y. Shioya, K. Ikegami, M. Maeda, K. Yanagida, J. Appl. Phys., 61 (2). (1987) p 561
I641 M. Fukumoto and T. Ohzone, Appl. Phys. Lett., 50 (14), (1987) p 894
I651 K. Akitmoto and K. Watanabe, Appl. Phys. Lett., 39 (5), (1981) p 445
1661 Y. Shioya, T. Itoh, I. Kobayashi and M. Maeda, J. Electrochem. Soc., 133 no 7, (1986) p 1475
1671 M. Kottke. F. Pintchovski. TR. White and PJ . Tobin, J. Appl. Phys., 60 (8). (1986) p 2835
Page 20
/68/ T. Hara, S. Enomoto and T. Jinbo, Jap. Journal of Applied Physics., 23 (7), (1984) p 2455
1691 G.A. West and K.W. Beeson, Proc of the loth Int. Conf. on CVD, the Electrochem. Soc., Pennington
NJ (1987) p 68
PO/ PJ . Gaczi, Ext. Abstracts of the 170th Meeting of the Electrochem. S.oc., San Diego, CA,
~ t . 19-24 (1986) VO~. 86-2, p 506
/71/ J.M. BIum, Proc. of the 10th Int. Conf. on CVD, The Electrochem. Soc., Pennington NJ (1987) p 476
/12/ ME. Alperin et Al, IEEE Journal of Solid-State Circuits, vol. 5C-20, no 1, (1985) p 61
f73/ V. Ilderem and R. Reif, Appl. Phys. Lett., 53 (8). (1988) p 687
p4/ R. Madat. E. Mastromatteo, C. Vahlas, C. Bernard, J. Palleau, J. Torres, Proc. of the MRS Soc. Fa11
Meeting., Boston USA (1988)
P5/ R. Madar, J.F. Million-Brodaz, C. Bernard, J. Torres, French Patent no 87115885 November 1987
f761 T.P.H.F. Wendling. C. Wieczorek and K. Hieber, Pm. of the 2nd European Workshop on Refractory
Metals and Silicides., Aussois (FRANCE), (1987) p 255
P71 M. Stolz, K. Hieber, C. Wieczorek, Thin Solid. Films., LOO, (1983) p 209