Conditions required for achieving superhardness of ≥45GPa in nc-TiN/a-Si3N4 nanocomposites

Materials Science and Engineering A 384 (2004) 102–116

Conditions required for achieving superhardness of≥45 GPain nc-TiN/a-Si3N4 nanocomposites

Jan Procházka, Pavla Karvánková, Maritza G.J. Veprek-Heijman, Stan Veprek∗

Institute for Chemistry of Inorganic Materials, Technical University Munich, Lichtenbergstr. 4, D-85747 Garching b., Munich, Germany

Received 27 November 2003; received in revised form 19 May 2004

Abstract

Based on our generic concept for the design of superhard (HVickers ≥ 40 GPa) nanocomposites with a high thermal stability and oxidationresistance we discuss the role of nitrogen pressure and substrate temperature during the deposition and the detrimental effect of impurities onthe formation of superhard nanocomposites with a high thermal stability. It is shown that inappropriate choice of the deposition parametersor impurities in the coatings are the possible reason of the poor reproducibility of our results by some authors. In order to differentiatebetween the superhard nanocomposites in which the superhardness originates from a stable nanostructure and ordinary coatings in which thehardness enhancement is due to energetic ion bombardment during their deposition we discuss the different behavior of such coatings and ournc-TiN/a-Si3N4 superhard nanocomposites upon annealing. It is further shown that hydrogen and oxygen impurities degrade the hardness. Ifthe oxygen content in the coatings amounts to about 0.5 at.% or more, the hardness remains limited to 35 GPa or less.© 2004 Published by Elsevier B.V.

Keywords: Superhard nanocomposites; TiN/Si3N4; Reactive sputtering; Plasma CVD; Impurities; Self-hardening

1. Introduction

Since the publication of our generic concept for their de-sign[1], the superhard nanocomposites with hardness of 40to ≥100 GPa attracted large attention of the scientific com-munity resulting in a fast growing number of publications(e.g. [2–6]). There exists presently some confusion[4,5,7]between the stable nanocomposites in which the superhard-ness originates from a stable nanostructure that is formed bya strong thermodynamically driven (spinodal) segregationwhich yields a sharp and strong interface[1,2,6,8,9]and or-dinary coatings in which the hardness enhancement is dueto energetic ion bombardment during their deposition[10].To the latter class belong hard transition metal nitrides (e.g.in [11,12]hardness of 100 GPa and 60–80 GPa was reportedfor (AlTiV)N and TiN coatings, respectively), carbides (e.g.[18]), borides (e.g.[15]) and the so called Me(1)N/Me(2)“nanocomposites” consisting of a hard transition metal ni-tride (Me(1)= Ti, Zr, Cr, etc.) and a ductile metal (Me(2)= Cu, Ni, etc.) which are deposited by means of unbalanced

∗ Corresponding author. Tel.:+49 89 289 136 24;fax: +49 89 289 136 26.

E-mail address: [email protected] (S. Veprek).

magnetron reactive sputtering at a low pressure of the or-der of≤10−3 mbar (see e.g.[7,19]). Under such conditions,the growing film is bombarded by energetic primary ionsthat are reflected from the target even in the absence of asubstrate bias voltage. It was shown that the latter coatingssoften and the hardness decreases to the ordinary bulk valueupon annealing to a temperature of≥450◦C [2,6,15] thusmaking them of a limited use. Also in the so called “super-hard nanocomposites” Me(1)N/Me(2) the hardness enhance-ment is solely due to the ion bombardment and any effect ofa nanocomposite hardening is absent[19,20]. The decreaseof the hardness and the concomitant decrease of the residualcompressive stress upon annealing can be used as a sim-ple check which shows if the hardness enhancement is dueto the complex, synergistic effect of the energetic ion bom-bardment[10] or to the formation of a stable nanocompositestructure[1]. In the latter case, the hardness remains stableupon annealing up to≥1100◦C. This high stability may besomewhat reduced to about 900–950◦C caused by the dif-fusion of impurities from the substrate, such as chromiumfrom steel[21], cobalt from cemented carbide[22] or molyb-denum (see below).

As explained earlier[1] and further extended later on[2,6,8,9,23,24], the high thermal stability of the superhard

0921-5093/$ – see front matter © 2004 Published by Elsevier B.V.doi:10.1016/j.msea.2004.05.046

J. Prochazka et al. / Materials Science and Engineering A 384 (2004) 102–116 103

nanocomposites is achieved when the thermodynamicallydriven (spinodal) segregation is completed during the depo-sition. Because this segregation is thermodynamically drivenand rate controlled by diffusion, a sufficiently high activity ofnitrogen is needed to bring the system into the region of thephase diagram with fully segregated, immiscible phases[1].A temperature of about 500–600◦C is necessary in order tofacilitate a sufficiently fast diffusion which assures that thesegregation is completed during the deposition (for detailssee[1,6,9]). If these conditions are not met, the effect of thenanocomposite hardening is relatively small, and such coat-ings may show an increase of hardness (“self-hardening”)upon annealing in nitrogen to 600–800◦C [8].

Here we shall focus on the binary nc-TiN/a-Si3N4nanocomposites as an example par excellence that has beenstudied by us as well as by many other workers in muchdetail. Considering the thermodynamics of other binarysystems consisting of a hard, stable transition metal nitride(ZrN, VN, W2N, CrxN, etc.) and a non-metallic, covalentone (Si3N4, BN, AlN) one can easily see that the conclu-sions that will be drawn for the nc-TiN/a-Si3N4 systemapply also for the other nitride systems and also for carbideswith excess of carbon, borides and others as discussed in[2].

Fig. 1 shows a simplified phase diagram of the Ti–Si–Nsystem at a relatively high temperature of 900–1100◦C (formore details see[25–29]). The exact separation between theregions of stability of different phases is to some extent sub-ject to a definition as to how large a fraction of a given mixedphase can be tolerated in a given coating system, or, in otherterms, how much Si is allowed to be dissolved in substoi-chiometric TiN1−δ and Ti dissolved in Si3N4−δ, particularlywithin the interface. Such a partially mixed interface willnot be sharp and strong and, therefore, the mechanical prop-erties of the nanocomposites will not reach their optimum.Furthermore, the nitrogen activity at which the transition be-tween the phases occurs depends on temperature: With de-

Fig. 1. A simplified phase diagram of the Ti–Si–N system at a temperatureof 900–1100◦C.

creasing temperature it shifts towards a lower value of nitro-gen activity. The thermodynamic values are subject to someuncertainties because the Gibbs energies of formation arenot precisely known, particularly for the less well-definedphases, such as Ti5Si3Ny. As a rough approximation, thenitrogen partial pressure (in units of atmosphere) can be setequal to its activity in order to perform a simple thermody-namic estimate of the conditions needed for the phase seg-regation as it was done in our first paper on that subject[1].

The important conclusion to be drawn from the phase di-agram (Fig. 1) is the fact that the segregation of stoichio-metric TiN and Si3N4 phases can occur only when the ni-trogen activity is larger than about 10−6, i.e. at nitrogenpressure above 10−6 atm (ca. 0.001 mbar). This condition ismet in plasma CVD that operates at a pressure of 0.1 toseveral mbar[1,8,9,24,30]. In the case of PVD techniques(sputtering, vacuum arc evaporation, ion beam assisted de-position) that operate at a nitrogen partial pressure in therange of≤0.001 mbar, the low activity may be insufficientto drive the segregation to a completion. Consequently, thebinary nanocomposites do not reach the upper values ofthe hardness of≥50 GPa, the maximum of the hardness –when plotted versus the Si-content – is less sharp and, uponpost-annealing in nitrogen, such nanocomposites may show“self-hardening”[8,19,23,24,31,32].

Fig. 2a and bshow the values of hardness reported bydifferent research groups versus the Si-content for coatingsdeposited by plasma CVD at a pressure of a few mbar anda temperature of 450–600◦C but a different plasma density(Fig. 2a) and by reactive sputtering at low pressure of ni-trogen and, in some cases, a lower deposition temperature(Fig. 2b). (Here we assume that all reported results werethe correctly measured load-invariant values of hardness asdone in our reports from the very beginning[1,2].) For pos-sible artefacts of the measurement of hardness on superhardcoatings by means of the load–depth indentation techniquesee[2,42,43]. Fig. 2cshows the correlation of the hardnesswith residual compressive stress in the “Ti–Si–N” coatingsas function of Si-content reported by Vaz[41], andFig. 2dshows for comparison a similar correlation for TiNx as func-tion of the N2/Ar flow rate (that changes the stoichiometry)as reported by Musil and co-workers[11].

The results shown inFig. 2a–draise several questions:

1. What is the origin of the different values of the maximumachievable hardness and its dependence on the Si-contentas found by different groups,

2. why, in the coatings reported by Meng et al. (Fig. 2b),does the hardness reach only 32 GPa for pure TiN (cor-rect bulk value about 20–22 GPa) and it decreases withincreasing Si-content showing no enhancement (maxi-mum) at any particular value of Si-content, and

3. whether the correlation between the hardness and resid-ual compressive stress in the “Ti–Si–N” coatings of Vaz(Fig. 2c) may be evidence that the hardness enhancement

104 J. Prochazka et al. / Materials Science and Engineering A 384 (2004) 102–116

Fig. 2. (a) Hardness of “Ti–Si–N” coatings deposited by means of plasma CVD from TiCl4 and SiCl4 [33,34] and TiCl4 and SiH4 [1,35,36] dilutedwith N2 and H2 vs. silicon content; (b) hardness of “Ti–Si–N” coatings vs. Si-content for coatings deposited by magnetron sputtering (see[37–40] andtext); (c) hardness and biaxial compressive stress of “Ti–Si–N” coatings deposited by magnetron sputtering by Vaz et al.[41] and (d) a hardness of TiNxcoatings deposited by reactive magnetron sputtering and the concomitant stress vs. the Ar:N2 ratio of the flow rates; notice that stoichiometric TiN wasformed at the flow ratio corresponding to the maximum hardness and stress[11].

in these coatings is due to energetic ion bombardment asin pure TiNx (Fig. 2d) and not to the formation of nanos-tructure.

In this paper, we shall try to answer these questions.

2. Experimental

Fig. 3 shows the apparatus that was used in the presentwork for the deposition of the coatings by means of reac-tive magnetron sputtering. In order to minimize the impu-rities that can deteriorate the properties of the coatings, theapparatus was built of ultra-high vacuum (UHV) compati-ble parts and turbo-molecular pumps that provided a back-ground pressure (at room temperature) of about<10−8 mbar.However, after long-term deposition when the inner wallswere coated, new substrates introduced and the substrateholder was heated to 500–630◦C prior to the deposition, thebackground pressure could increase up to≤1 × 10−5 mbar.

Therefore, the apparatus had to be out-gassed over night. Adirect current (dc) power supply (0–2.4 kV, 5 A) protectedwith special electronic circuits against arcing[44] was usedto maintain a high power density of up to 18 W/cm2 at thetarget. This enabled us to achieve high deposition rates of≥1.5 nm/s (≥5.4�m/h) even with pure nitrogen as both thesputtering medium and as reactive gas. A high depositionrate together with the low background pressure minimize theincorporation of impurities into the growing film. All gasesused were of high purity (99.9999 for Ar and N2 and evenbetter for SiH4 and SiH4/H2).

X-ray diffraction measurements were done by meansof Siemens D 5000 diffractometer using Cu K� radiation(Ni-filter) and a secondary monochromator that provided aresolution of 0.05◦. The crystallite size,d, was determinedfrom the integrated width of the Bragg reflections, that werefitted with the Pearson VII function, using the Scherrerformula with integral half width whend ≤ 8 nm and by theWarren–Averbach Fourier transform integral deconvolutionmethod ford > 8 nm. We have shown that this procedure


Fig. 3. Schematics of the apparatus used for the deposition of the coatingswith a cross-section of the reactor 1. The Ti target of a purity of 99.995%in the unbalanced magnetron UBM had a diameter of 200 mm. Thetemperature of the substrate holder S (diameter of 100 mm) was controlledby means of Joule heatingH and temperature controller up to about800◦C. The gases were introduced via mass flow controllers MFC througha shower head and kept at the chosen value by means of a computer(not shown) that obtained the signal from quadrupole mass spectrometerQMS that was located in a separately pumped chamber. TP and DSP areturbomolecular and “rotary” pumps, respectively; PI, PE and CAP arePirani, Penning and capacitive (MKS Baratron) pressure meters. Chamber4 is an oven for thermal decomposition of silane in the outlet.

yields correct values in agreement with the high-resolutiontransmission electron microscopy (HR-TEM)[45]. The el-emental composition was determined by means of energydispersive analysis of X-rays (EDX) and verified for a num-ber of coatings by the more accurate but time-consumingelastic recoil detection spectroscopy (ERD)[46]. Because ofits large dynamic range, ERD also provided the concentra-tion of light impurities, such as hydrogen and oxygen. Thenature and fraction of phases (e.g. Si3N4 and TiSi2) weredetermined by carefully calibrated X-ray photoelectronspectroscopy (XPS). The biaxial stress in the coatings wasdetermined from the bending of a steel strip substrate afterthe deposition of the coatings using the Stoney’s formula.In addition, it was compared with the value of stress ob-tained by the sin2Ψ method[47] that yielded a lower value.The load-independent hardness of the coatings was mea-sured by means of an automated indentometer, Fischerscope100, typically at an applied load between 30 and 100 mNwhere the indentation depth was >0.3�m and the valuesof hardness were load-invariant[1,2,42,43]. Correction forthe finite radius of the indenter tip and calibration using Si(H = 10.3 GPa) and sapphire (H = 21 GPa) was done priorto and after each series of measurements. The thicknessof the coatings deposited on stainless steel was between 6and 10�m because the high hardness of≥40 GPa cannotbe correctly measured on thinner coatings (for details see[42,43]). All other techniques used for the characterizationof the coatings (e.g. scanning electron microscopy, etc.)were of the standard type.

3. Results

For brevity we shall present only typical examples ofthe most important results from a large series of experi-ments[44]. Depositions at 550◦C without Si yielded TiNcoatings with the typical columnar structure. The hard-nessH and biaxial compressive stressσbiax varied fromH = 16 GPa andσbiax= −0.5 to 36 GPa and−3.5 GPa,respectively, when the substrate bias was changed from0 to −300 V. Simultaneously with the increasing biasthe columnar morphology became denser but remainedclearly visible by SEM. Further data can be found in[44].

3.1. Combined reactive sputtering of Ti and plasma CVDof Si3N4

The use of silane as source of silicon appeared to be thesimplest way how to vary in a controllable manner the sil-icon content of the deposited films. The following depo-sition conditions were found optimum and therefore usedin this series: Gas flow of 17 sccm (standard cubic cen-timeter per minute) and partial pressure of Ar and N2 of0.01 mbar, substrate temperature of about 500◦C and bias−200 V. In the first series, silane diluted with hydrogen (seeinserts inFig. 4a) was used at a total flow varied between 0and 16 sccm. The Ti-target was soldered to the water-cooledcathode made of stainless steel. The deposition rate was be-tween 1 and 1.3 nm/s.

With increasing Si-content from 0 to about 12 at.% theresidual, compressive biaxial stress increased from about−1.5 to −4.5 GPa. Hardness of coatings deposited on thesteel strips used for the stress measurement remained, how-ever, independent of the Si-content (or more exactly, withcoverage of TiN nanocrystals by Si3N4) as shown inFig. 4a(average value of five measurements at each load of 30, 50and 70 mN). The crystallite size of 7± 2 nm was almost in-dependent of Si-content. Only Bragg reflections correspond-ing to the fcc lattice of stoichiometric TiN were found inthe XRD pattern. The Si 2p line in the XPS spectra showed,besides of a minor signal from oxide contamination, only asignal at a binding energy of 101.7± 0.2 eV correspond-ing to Si-atom fourfold coordinated to nitrogen as in Si3N4(seeFig. 8a and also[1]). Fig. 4b shows an example ofa depth-dependence of the concentrations of the main ele-ments (Ti, Si, N), and oxygen and hydrogen impurities de-termined by ERD. Although the nc-TiN/a-Si3N4 nanostruc-ture was formed (XRD and XPS data), there is only a slightenhancement of the hardness to about 30 GPa as comparedto pure TiN (H = 22–25 GPa, see below) but no maximumat the coverage of about 1 monolayer (ML; seeFig. 4a).Because both nitrogen pressure and deposition temperaturewere sufficiently high in order to complete the phase seg-regation to stoichiometric TiN and Si3N4, we attributed theabsence of a maximum to the effect of about 10 at.% ofhydrogen impurities because the oxygen impurity of about


Fig. 4. (a) Hardness of nc-TiN/a-Si3N4 coatings deposited by reactivemagnetron sputtering of Ti and plasma CVD of Si from SiH4/H2 mixturevs. the coverage of TiN nanocrystals with Si3N4. (b) Depth profile of theconcentration of Ti, Si, N and O- and H-impurities as determined by ERD.

0.3 at.% should allow us to reach hardness of≥35 GPa (seebelow and[22]).

Because the very high power density at the target andthe resulting need of a more efficient cooling, all the fol-lowing depositions were done with the Ti-target fixed tothe body of the magnetron via O-rings and directly cooledwith water. In order to reduce the hydrogen content, puresilane at a flow rate between 0 and 2.25 sccm was used inthis series with a nitrogen pressure of 0.02 mbar and flowof 22–25 sccm. All other parameters (substrate temperatureof about 500◦C, power density at the Ti-target of about15 W/cm2) were kept constant. As in the previous series, thesilicon was chemically bonded as Si3N4 with a small ox-ide contamination seen in the XPS spectrum (seeFig. 8b).The hydrogen impurity content decreased to about 2 at.%but that of oxygen increased to 2 at.% as compared with thedata from the previous series (ERD analysis). When plottedversus Si-content, the hardness showed a clear maximum atSi-content of about 6± 1 at.% resembling the dependencereported by us for superhard nc-TiN/a-Si3N4 nanocompos-ites deposited by plasma CVD[1] and similar but less clearlypronounced by Diserens et al. for sputtered “Ti–Si–N” coat-

Fig. 5. Hardness of nc-TiN/a-Si3N4 coatings deposited by reactive mag-netron sputtering of Ti and plasma CVD of Si from pure silane vs. thecoverage of the TiN nanocrystals with Si3N4.

ings [38]. When plotted against the coverage of the TiNnanocrystals with Si3N4, the hardness reaches a maximumat coverage of about 1 monolayer (seeFig. 5) in agreementwith our earlier data for nc-MeN/a-Si3N4 (Me = Ti, W, V)[9] and nc-TiN/a-BN[30]. The biaxial compressive stressdetermined from the bending of the steel strip substratesof about−1.75± 0.2 GPa is relatively small and indepen-dent of the Si-content. This indicates that defects, whichmay be induced by the energetic ion bombardment withinthe near-surface region of the growing film, are annealedout in course of the continuous growth when the bulk ofthe film is not accessible to the ions. The crystallite size of5.5 ± 1 nm is also almost constant up to the Si-content of15 at.% and decreases afterwards. This makes it possible toobserve the dilatation of the lattice constant that increaseswith decreasing crystallite size as seen inFig. 6. This di-latation agrees fairly well with that found and explained fornanocrystalline silicon nc-Si[17] and for nc-MeN/a-Si3N4nanocomposites (Me= Ti, W, V) deposited by plasma CVD

Fig. 6. Lattice dilatation increases with decreasing crystallite size (see[1]for nc-TiN/a-Si3N4 and nc-W2N/a-Si3N4 [35,48,50]for nc-Si).


Fig. 7. Development of the morphology of the nc-TiN/a-Si3N4 coatings with increasing content of Si3N4. (Only the Si-content in at.% is indicated, butthe chemical bonding of Si as “Si3N4” was verified by XPS.)

under low energy ion bombardment[35] when one accountsfor the higher compressive stress in the present coatings thatdecreases that dilatation (see also nc-Si films deposited un-der negative substrate bias in[48–50]).

Fig. 7 shows the development of the morphology ofthe coatings. Although some densification is seen with in-creasing content of Si3N4, the morphology remains colum-nar whereas in the case of plasma CVD coatings, suchas nc-MeN/a-Si3N4 (Me = Ti, W, V) [2,35] as well asnc-TiN/a-BN [21,51], the columnar structure completelyvanished at the optimum composition corresponding toabout 1 monolayer coverage and the maximum hardness(see also below). We attribute this difference inFig. 7 toeither the impurities or to the still relatively low depositiontemperature (notice, that the nitrogen pressure of 0.02 mbaris fairly high) or a combined effect of both. Because ourmain goal was to find conditions for the optimum depositionof the superhard nanocomposites by means of sputtering,we did not investigate this problem in further detail.

3.2. Deposition by reactive magnetron sputtering ofTi and Si

In order to control the Si-content of the coatings withoutthe use of silane, silicon strips cut of a wafer were mounted

on the Ti-target by means of screws made of titanium. Theirarea was varied in different series of deposition between 0and 16% of the total area of the Ti-target which resulted ina silicon content between 0 and 14 at.%. An argon/nitrogenmixture was used in the first series and pure nitrogen in thesecond one of the studies to be reported here.

3.2.1. Deposition using an Ar and N2 mixtureIn this series, an argon-to-nitrogen ratio of 1:1, total pres-

sure of 0.002 mbar, total flow rate of about 14 sccm, lowsubstrate bias voltage of−25 V and deposition temperatureof 550◦C were used. The partial pressure of nitrogen wasexpected to be high enough to be just within the stabilityrange of the coexisting segregated TiN and Si3N4 phasesat this relatively low deposition temperature, since it wasclose to the pressure when both Si and Ti targets reachedthe “poisoned” sputtering regime at the power density onthe target of≥15 W/cm2. At the same time, we expectedto reach the maximum deposition rate with the given powersupply. A high deposition rate favors lower oxygen contentin the coatings[52], because the oxygen impurity flux fromthe walls of the reactor is dependent on temperature of thatwall during the deposition, which is almost constant in ourcase. The deposition rate increased from 1.77 nm/s for pureTiN to about 1.9 nm/s for nanocomposites with Si-content


Fig. 8. (a) Examples of XPS spectra of the Si2p lines from coatings deposited by sputtering of Ti and plasma CVD of Si from SiH4/H2 mixture(Si-content in sample no. 100300: 11.4 at.%, no. 080200: 3.7 at.%); (b) similar conditions as in (a) but Si CVD from SiH4 only (Si-content in sample no.290101: 22.7 at.%, no. 241100: 4.7 at.%); (c) magnetron sputtering from Ti and Si target using the Ar/N2 mixture (Si-content 1604/1: 9.6 at.%, 1505/2:6.1 at.%); (d) magnetron sputtering of Ti and Si with pure nitrogen (Si-content in sample no. 150103/3: 3.2 at.%, no. 160103/1: 15.9 at.%).

of about 14 at.% due to the higher sputter yield of Si as com-pared to Ti. Oxygen and hydrogen impurities were of about≤0.3 at.% (ERD). The oxide component in the XPS spectraof the Si2p region (Fig. 8c) is due to surface contaminationupon the exposure of the coatings to air that could not becompletely removed by sputtering because of a finite albeitsmall surface roughness. The main XPS signal in the Si2pregion is due to silicon fourfold coordinated to nitrogen asin Si3N4 (seeFig. 8c). The minor contribution of the signalfrom TiSi2 is almost within the accuracy of the fitting ofthe measured curve by Gaussian peaks (notice that the ex-perimental XPS peaks are asymmetrically broadened due toelectron shake-ups at the Fermi level and inelastically scat-tered photoelectrons of the given signal) but, as it will be

seen in the next section, these signals are likely to be indeedfrom TiSi2. The biaxial compressive stress of about−3.1±0.4 GPa determined from the bending of the stainless steelstrips and about−1.3± 0.2 GPa determined from the sin2ψ

method from the XRD patterns was essentially independentof the composition. The crystallite size decreased from about9 nm for pure TiN to about 3 nm for nc-TiN/a-Si3N4 coatingswith Si-content of 14 at.%. The lattice dilatation increasedwith decreasing size in a similar manner as shown inFig. 6.

Fig. 9 shows that by analogy with the results inFig. 5the hardness reaches a maximum at coverage of about 1.2monolayer but it is significantly higher in this series. Theslightly higher coverage than 1 is due to the fact that the crys-tallites are not cubic and, therefore, some additional Si3N4


Fig. 9. Hardness of nc-TiN/a-Si3N4 coatings deposited by means ofreactive magnetron sputtering in an Ar:N2 = 1:1 mixture at a total pressureof 0.002 mbar and temperature 550◦C.

is needed to fill e.g. the triple junctions[9]. The maximumhardness reaches 38 GPa and the columnar morphology al-most vanishes at the optimum Si-content (Fig. 10).

Because the properties of these coatings are already sim-ilar to those of coatings obtained by plasma CVD[1,35] itis appropriate to study their thermal stability. For these rea-

Fig. 10. Development of the morphology of nc-TiN/a-Si3N4 coatings deposited by reactive magnetron sputtering in Ar:N2 = 1:1 mixture at a totalpressure of 0.002 mbar and 550◦C with increasing Si-content.

sons the coatings deposited on molybdenum substrates wereisochronally annealed in pure nitrogen for 30 min at eachtemperature.Fig. 11shows the high thermal stability of coat-ings with the surface coverage of about 1.4 monolayer (ML,Fig. 11a) and a slightly lower one for a coating with a lowercoverage of about 0.8 ML (Fig. 11b). One notices that in thecoatings with the surface coverage of 0.8 ML the coarseningof the nanostructure and decrease of the hardness commenceonly at about 1100◦C, and it is almost absent in the coatingwith coverage of 1.4 ML, in agreement with the data fromthe plasma CVD coatings[8,23]. This stability is much bet-ter than that of coatings where the superhardness was due tothe energetic ion bombardment during their deposition (e.g.the Me(1)N/Me(2) “nanocomposites”[20]). The compres-sive biaxial stress in the optimum coatings from this seriesdecreased upon annealing to≥800◦C whereas the hardnessremained constant up to more than 1100◦C. This provesthat the hardness enhancement in these coatings is due com-pletely to the stable nanostructure and not to energetic ionbombardment during their deposition.

It is not quite clear if the onset of coarsening and a slightdecrease of the hardness after annealing to 1200◦C repre-sent the upper limit to the stability of these coatings be-cause above 1000◦C molybdenum from the substrate dif-fuses into the coatings (seeFig. 12). It is worth mentioning


Fig. 11. Hardness and crystallite size of nc-TiN/a-Si3N4 coatings depositedby reactive magnetron sputtering in Ar:N2 = 1:1 mixture at a total pressureof 0.002 mbar and temperature of 550◦C measured at room temperatureafter the isochronal annealing to a given temperature for half hour in purenitrogen. (a) Surface coverage of about 1.4 ML and (b) surface coverageabout 0.8 ML.

that the rate of the Mo-diffusion into these coatings is signif-icantly slower than that into pure TiN[44]. This illustratesthe strong protecting effect of the dense Si3N4 within thegrain boundaries that is also responsible for the strongly en-hanced oxidation resistance of nc-TiN/a-Si3N4 as comparedto TiN (see[1,44]).

3.2.2. Deposition using pure N2The results of the foregoing section have shown that the

partial pressure of nitrogen of about 0.001 mbar was prob-ably sufficient to ensure the thermodynamically driven seg-regation of the TiN and Si3N4 phases, but the small amountof TiSi2 seen in XPS (Fig. 8c) suggests that either a highernitrogen pressure, or a lower deposition rate, or a higherdeposition temperature that enhances the kinetics of phasesegregation[1] may be more favorable in order to assure theformation of stoichiometric Si3N4 as the only Si-containingphase. However, because a lower deposition rate can resultin higher impurity content, pure nitrogen at a pressure of

Fig. 12. Concentration of molybdenum that diffused from the substrate(the interface is at 0) into the coatings shown inFig. 11aafter annealingin 1 atm N2 for half hour at the temperature indicated.

0.002 mbar and a relatively large flow rate of 11 sccm, to-gether with a high deposition rate was chosen in this series.Moreover, in order to assure that the kinetics of the forma-tion will be sufficiently fast we increased the deposition tem-perature to approximately 630◦C. Under these conditionsthe substrate bias could be decreased to 0 V. The high powerdensity on the Ti and Si-target of 18.6 W/cm2 provided ahigh deposition rate of≥1.6 nm/s that assured high purityof the deposited coatings. In the first series to be reportedhere, the oxygen impurity content determined by ERD wasabout≤0.3 at.% (ERD).

The crystallite size decreased from about 12 nm for pureTiN to 3 nm for nc-TiN/a-Si3N4 with silicon content of≥6 at.% in agreement with our old data for the nanocom-posites prepared by plasma CVD[1]. The biaxial compres-sive stress determined from the bending of the stainlesssteel strips increased almost monotonously from about−2.5 GPa for pure TiN (hardness 25 GPa) to about−5 GPafor Si-content of 16 at.%. The lattice dilatation increasedwith decreasing crystallite size below 10 nm to about 1%for the smallest crystallite size in agreement with the data ofFig. 6and our earlier results for nc-Si and nc-MeN/a-Si3N4(Me = Ti, W, V) deposited by plasma CVD. The XPSspectra of the Si2p region showed only the peak associ-ated with Si fourfold coordinated to nitrogen i.e. the Si3N4phase (seeFig. 8d). Obviously, with the somewhat highernitrogen pressure of 0.002 mbar used in this series, the seg-regation into stoichiometric TiN and Si3N4 with a sharp andstrong interface is completed (cf.Fig. 8c with d). Fig. 13shows the hardness versus the silicon content (Fig. 13a)and versus the surface coverage of the TiN nanocrystalswith Si3N4 (Fig. 13b). The hardness of coatings withthe optimum composition reaches 45 GPa. The somewhathigher value of the coverage of about 1.7 is due probablyto an irregular shape of the nanocrystals (TEM work inprogress).


Fig. 13. Hardness of nc-TiN/a-Si3N4 nanocomposites deposited by reactivemagnetron sputtering in pure nitrogen vs. the total Si-content (a) and vs.the surface coverage of the TiN nanocrystals with Si3N4 (b). Open circlesin (b) show the compressive biaxial stress. The oxygen impurity contentin these coatings was about≤0.3 at.%.

In Fig. 13bwe show also the biaxial compressive stressdetermined from the bending of the stainless steel strips.It is important to notice that the biaxial compressive stressdoes not correlate with the hardness. In fact, the compres-sive stress slightly increases whereas the hardness decreasesfrom maximum value of 45 GPa when the coverage increasesabove 1.7 ML. This unambiguously shows that the hardnessenhancement is due to the formation of stable nanostruc-ture and not to the energetic ion bombardment (comparewith Fig. 2d). In order to definitively prove this conclusionwe performed extensive series of annealing of all the coat-ings with a surface coverage between 1.0 and 2.7 depositedin this series that have a hardness of≥40 GPa. The resultswere essentially the same as found for the previous series(seeFig. 11a), i.e. the hardness did not change up to about1100◦C where molybdenum diffusion from the substrateinto the coating was observed.

These findings are finally supported by the developmentof the morphology of the coatings from the columnar mor-phology of TiN towards the fine-grain, isotropic one ofnc-TiN/a-Si3N4 coatings with the optimum surface cover-

age and superhardness of 40–45 GPa as shown inFig. 14. Acomparison withFigs. 7 and 10shown above and with thosepublished for nc-MeN/a-Si3N4 (Me = Ti, W, V) coatingsdeposited by plasma CVD in intense glow discharge[35]reveals that such development of the morphology is also atypical sign of the formation of the stable nanostructure witha sharp and strong interface that results in the superhardnessas already discussed in our earlier papers[1,2,35].

We also investigated the oxidation resistance (0.5 h in air)of the coatings deposited in this series in comparison withnc-TiN/a-Si3N4 deposited by plasma CVD[1] and with avariety of (TiAl)N coatings[54]. The high oxidation resis-tance of the nc-TiN/a-Si3N4 coatings deposited in course ofthe present work by reactive sputtering compares very wellwith that of nc-TiN/a-Si3N4 coatings deposited by plasmaCVD [1] and with the best (TiAl)N ones[21,36,54]. Thisfinding further supports the concept of the formation of adense nanostructure with sharp and strong interface. Thisnanostructure not only hinders the grain boundary slidingand thus provides the nanocomposites with a high hardness,but it also hinders the diffusion of oxygen along the grainboundaries and therefore enhances their stability against ox-idation [1,2].

3.3. Degradation of the hardness by oxygen impurities

The results presented in the foregoing section clearlyshow that a nitrogen pressure of 0.002 mbar and a highdeposition temperature of≥600◦C are sufficient to assurethe formation of the stable nanostructure during the de-position by reactive sputtering. Nevertheless, the questionarises why is the maximum achievable hardness still lim-ited to about 45 GPa whereas in the binary nc-TiN/a-Si3N4coatings deposited by plasma CVD we reported hardnessof ≥50 GPa[1,9]. (Notice that the much higher hardnessof 80–≥100 GPa reported by us[24,46] was found in theternary nc-TiN/a-Si3N4/a- and nc-TiSi2 system with oxy-gen impurity below 0.07 at.%.) As already mentioned, wereported recently that in nc-TiN/a-Si3N4 coatings preparedby plasma CVD, oxygen impurities of about 0.1 at.%, re-duce the maximum achievable hardness to about 45 GPa and,when the impurity content increases above about≥0.4 at.%,the hardness decreases to about≤35 GPa[22,55]. Therefore,we repeated the present deposition experiments by mag-netron sputtering under the same conditions as in the fore-going section but using an extended out-gassing of the ap-paratus in order to reduce the background pressure prior tothe deposition (when the whole substrate holder was heatedto >540◦C) to about≤1 × 10−6 mbar which resulted in adecrease of the oxygen impurity content to≤0.2 at.%.

Fig. 15ashows clearly that, compared with the results inFig. 13, the hardness increases when oxygen impurity con-tent decreases. The hardness reaches values close to 50 GPaor even more as in nanocomposites prepared by plasma CVDthat had a similar low oxygen impurity content[1]. Becausethe extended out-gassing is very much time-consuming we


Fig. 14. Development of the morphology of nc-TiN/a-Si3N4 coatings deposited by reactive magnetron sputtering in pure N2 at a total pressure of0.002 mbar and 630◦C with increasing Si-content.

have to modify the apparatus in order to further improve thebackground pressure prior to the deposition when the sub-strate holder is at the high temperature. Nevertheless, thecomparison of the results shown inFig. 15awith those inFigs. 5, 9 and 13clearly substantiates the finding that a lowoxygen impurity content of≤0.05 at.% is necessary in or-der to reach a hardness of≥50 GPa in the nc-TiN/a-Si3N4nanocomposites.

Let us emphasize that the degradation of the hardness byimpurities and the solution how to avoid it was reported byus already long time ago[23]. Also the high purity of [O]<0.07 at.%, [Cl]< 0.5 at.% and [H]< 0.1 at.% of our coat-ings whose hardness was 50–105 GPa was reported 4 yearsago [46]. Interestingly, these papers are not quoted in therecent reports where the workers complain about the irre-producibility of our results and speculate, that our measure-ments of the hardness were done incorrectly (see e.g.[56]and reply in[42,57]).

Fig. 15bis a summary of all our data (available at the dateof the submission of this paper) for the stoichiometric andfully segregated binary nc-TiN/a-Si3N4 coatings depositedby both plasma CVD[22,55] and reactive sputtering. Thefigure shows the dependence of the maximum hardness (thatis achieved in a given series of depositions with variableSi-content for coatings with the optimum composition) ver-

sus the oxygen impurity content in the coatings. One canclearly see that, regardless of the deposition technique used,the oxygen impurities control critically the hardness in thissystem. Only if oxygen impurity content is<0.07 at.% fornanocomposites deposited by plasma CVD and≤0.2 at.%for those deposited by sputtering, the hardness of≥50 GPacan be achieved.

4. Discussion

Based on the results presented above we shall now try toanswer the questions put forward in the introduction turningour attention back toFig. 2.

4.1. Coatings deposited by means of plasma CVD

All three groups that used P CVD were working with aglow discharge operating in a mixture of H2 and N2 withTiCl4 as a source of Ti and either SiCl4 [33,34] or SiH4[1,36] as a source of silicon. The important difference wasthe plasma density used: Veprek et al. used a high fre-quency (HF) discharge operating at a relatively high pres-sure and high power density, Li et al. used a weaker directcurrent (dc) glow discharge with the substrates mounted at


Fig. 15. (a) Hardness of nc-TiN/a-Si3N4 coatings deposited with purenitrogen as inFig. 13abut with long-term out-gassing of the apparatusprior to the deposition. (b) Hardness of stoichiometric nc-TiN/a-Si3N4

coatings with the optimum coverage of TiN nanocrystals with Si3N4

deposited by plasma CVD (full circles) and by reactive sputtering (crosses)vs. oxygen impurity content.

the cathode and Lee and Kim used an “inductively coupled”HF discharge at a relatively low power that, based on ourlong-term experience with glow discharges, provided theweakest plasma of these three papers. Veprek et al. re-ported almost the same dependence of the hardness on theSi-content as shown inFig. 2a also for nc-W2N/a-Si3N4[35] and nc-VN/a-Si3N4 [36]. For convenience of the readerwe included inFig. 2athe dependence for nc-VN/a-Si3N4.When deposited under energetic ion bombardment at thecathode of an intense abnormal glow discharge, the crystal-lite size of TiN was smaller due to the ion bombardmentthan in the HF discharge and the correlation between the

hardness and the Si-content was somewhat different. How-ever, when a complete analysis of the coatings was done itwas shown that the maximum hardness of the stoichiomet-ric nc-TiN/a-Si3N4 fully segregated coatings deposited inintense HF or dc discharge is obtained when there is about1 monolayer of Si3N4 between the TiN nanocrystals[9].Let us emphasize that a low compressive stress of≤0.5 GPaand a high thermal stability up to≥1100◦C [8,23,24] thatwere found in our nanocomposites exclude any hardness en-hancement due to energetic ion bombardment during theirdeposition[10].

X-ray photoelectron spectroscopy has shown that ourcoatings prepared in both HF and dc intense dischargescontained silicon only as Si3N4 whereas TiSi2 was found, inaddition to Si3N4, in the coatings prepared by Li et al.[33]and Lee and Kim[34] and in our deposition experimentsdone later on when a low current density was used[9,24,46].Considering the phase diagram inFig. 1 it is clear that ahigh plasma density together with a high nitrogen pressureand temperature are necessary to obtain the fully segre-gated binary system of stoichiometric TiN and Si3N4 withthe Si3N4 tissue at percolating threshold acting as an effi-cient strong “glue” between the TiN nanocrystals (see e.g.[1,6,9,46]). Because also in the ternary nc-TiN/a-Si3N4/a-and nc-TiSi2 nanocomposite the maximum hardness of 80to ≥100 GPa requires the nanocrystals to be covered withabout 1 monolayer of Si3N4 (see Fig. 3 in[9]), the differ-ences in the Si-content at which the maximum hardnesswas achieved by different groups inFig. 2a (and also inFig. 2b) are evidently due to a different content of the TiSi2phase and possibly also to a different crystallite size and/orshape. The absence of the TiSi2 phase in the nanocompos-ites deposited by our group in the intense discharge is thereason why the Si-content corresponding to the maximumhardness is lowest as compared with the coatings of theother groups shown inFig. 2a and it is very similar fornc-TiN/-, nc-W2N/- and nc-VN/a-Si3N4 nanocomposites.

The somewhat higher value of the maximum hardnessreported by Li et al. is due probably to their ternary naturebecause only in the ternary nc-TiN/a-Si3N4/a- and nc-TiSi2,that were deposited in a weaker discharge than the binaryones and had oxygen impurity content of<0.07at.%, weachieved the hardness of 80 to≥100 GPa[9,24,46]. Thelower value of the maximum hardness found by Lee andKim may be due to impurities.

4.2. Coatings deposited by means of reactive sputtering

In the work of Patscheider and co-workers[37,38] andKim et al. [39] the maximum hardness of the coatings, pre-pared by PVD at a much lower nitrogen pressure, appears atdifferent silicon content. Moreover, there is only one coat-ing in the figures from these groups with hardness reaching40 GPa. This can be due to a relatively low nitrogen pres-sure and deposition temperature used by these researchersor by oxygen impurities. The incomplete data regarding the


deposition conditions and impurity content as reported inthese papers do not allow us to draw unambiguous conclu-sions regarding the exact reason for this lower hardness. Theelongated shape of the TiN nanocrystals in the coatings ofPatscheider[32] clearly indicates that they were preparedunder conditions remote from those needed for the forma-tion of the fully segregated, stable nanostructure.

The results of Meng et al. (seeFig. 2b), who did not obtainany hardness enhancement in Ti–Si–N coatings depositedby magnetron sputtering in combination with an inductivelycoupled discharge, are not surprising when considering thevery low nitrogen pressure of 0.198 mTorr (0.00015 mbar= 1.5 × 10−7 atm) and a too low deposition temperatureof about 250◦C used by these authors[40]. Furthermore,because the thickness of the coatings prepared by these au-thors was<0.5�m, the deposition rate was most probablylow which resulted in the high oxygen impurity content ofabout 2 at.%, as reported recently by Meng et al.[58].

This fully agrees with our results when we introduced alarge amount of impurities: No hardness enhancement wasfound in our experiments described above, when the impu-rity content was comparably high as shown inFig. 4a. Ex-trapolation of the data inFig. 15bof the present paper tooxygen content of 2 at.%, as found in the coatings of Menget al., indeed shows that that hardness should be roughlyabout 32 GPa in agreement with the data of Meng et al.[40].A similar value of the hardness of about 32 GPa and lackof any hardness enhancement with increasing silicon con-tent was found also for nc-TiN/a-Si3N4 coatings depositedby plasma CVD and having 1–1.5 at.% of oxygen impurities(seeFig. 4 in [55]).

The necessity to use a sufficiently high pressure of ni-trogen and high deposition temperature of 500–550◦C wasexplained on the basis of thermodynamic (seeFig. 1 in [1])and kinetic considerations regarding the driving force forthe segregation and the need of a sufficiently fast kineticsalready in our first paper[1], and emphasized in many ofour later publications. Later on, this was confirmed for thesputtering also by Vaz et al.[59–63] and for plasma CVDby Lee and Kim[34]. These authors found a decrease of thehardness in that system to about 24 GPa when the substratetemperature was only 200◦C (see Fig. 5a in[59]) or even400◦C (see Fig. 3 in[34]).

Thus, the low values of the maximum achieved hardnessreported in many recent papers, and the absence of any hard-ness enhancement reported by Meng et al.[40] is due (atleast partially) to the fact that the formation of the stablenanostructure and the concomitant hardness enhancementcannot be achieved when the thermodynamic and kineticconditions, that are necessary for the thermodynamicallydriven (spinodal) phase segregation to occur[1], are not met.

Last but not least, impurities are likely to limit the achiev-able hardness (as reported by us years ago[23]) in manyrecently published papers because of the relatively low de-position rates and high background pressure. Notice thatunless special precaution is met, in a typical sputter sys-

tem with deposition rate of≤1�m/h and background pres-sure of 2× 10−6 to 6 × 10−6 mbar [37,38], the flux ofimpurity ions from background gas will results in an im-purity level of several at.%[52] or more. Moreover, in themajority of the published papers it is not clear if the re-ported “base pressure” was measured with substrate on thedeposition temperature. For example, in the experiments ofPatscheider and co-workers[37,38], the reported base pres-sure was measured at room temperature and it increased toabout 10−5 mbar when the substrate was heated to the rela-tively modest deposition temperature of 300◦C [64].

Hu et al. [53] reported recently on the preparation of“Ti–Si–N” coatings by means of magnetron sputtering at ni-trogen partial pressure of 2× 10−4 mbar (i.e. 2× 10−7 atm,cf. Fig. 1). Unfortunately, they report neither the depositionrate nor the impurity content. The maximum hardness ofonly 36 GPa was found for only one coating deposited atroom temperature. An increase of the temperature to 400◦Cresulted in decrease of the hardness to 29 GPa which stronglysuggests that the hardness of 36 GPa obtained at room tem-perature was not due to the formation of a nanocompositebut, more likely, it resulted from energetic ion bombardmentduring the deposition at the low temperature. Obviously, theauthors have chosen a too low pressure of nitrogen and im-purities may have contributed to these disappointing resultsas well.

Let us now turn our attention to the correlation of the hard-ness and compressive stress as found by Vaz for “Ti–Si–N”coatings and by Musil and co-workers for TiNx (Fig. 2cand d). As shown inFig. 2c, the maximum of the hard-ness in the coatings deposited by Vaz et al. at 300◦C corre-lates with the maximum of biaxial compressive stress[41]in a very similar way as in the TiNx coatings of Musil andco-workers[11]. The hardness enhancement of up to 60 GPain TiN seen inFig. 2dis surely due to the energetic ion bom-bardment. Obviously, the surface of the “Ti–Si–N” coatingsreported by Vaz was bombarded by energetic ions duringtheir deposition and the temperature was too low to allowrelaxation of the induced defects. Therefore, there may besome contribution of the “hardening by ion bombardment”in the “Ti–Si–N” coatings as it is present in TiNx (Fig. 2d).The question is which fraction of the hardness enhancementin the “Ti–Si–N” coatings is due to this effect and howmuch is the contribution by the formation of nanocompos-ite. The answer to this question provided annealing exper-iments: Whereas the hardness of TiN, HfB2, the so called“nanocomposites” ZrN/Cu, ZrN/Ni, CrxN/Ni and others de-creases to the usual bulk value of about≤20 GPa uponannealing to 400–600◦C, the hardness of the “Ti–Si–N”coatings reported by Vaz remains nearly constant[65,66]in agreement with our earlier data on nc-TiN/a-Si3N4 andmore recent data on the nc-(Al1−xTix)N/a-Si3N4 coatingsprepared by industrial-scale PVD[8,23,24,31]. These find-ings show clearly that the hardness enhancement in thework of Vaz et al. was to a dominant extent due to the for-mation of a nanostructure which, however, was far from


the optimum because of insufficient deposition conditionsused.

A hardness increase upon annealing was recently reportedalso by Patscheider for one of his “Ti–Si–N” coating, butthe other coating (of total two reported in his review paper[32]) lost the hardness upon annealing already above about600◦C which, together with the decrease of the compres-sive stress reported in[32] indicates that there may havebeen a significant contribution of the hardness enhancementby energetic ion bombardment in his work. Because of thelack of data regarding the deposition conditions and of in-complete characterization of his coatings, it is impossibleto elucidate the reasons for such poor reproducibility of theannealing behavior of the coatings reported by Patscheider.Let us point out that among many tens of coatings inves-tigated upon annealing in our laboratory, only the ZrN/Ni,Cr2N/Ni “nanocomposites” prepared by Musil et al.[20] andone “Ti–Si–N” coating prepared by another laboratory (thatcontained a large fraction of TiSi2 and oxygen impurities)showed a similar behavior.

This is one of many examples that show why it is difficultto fully analyze the majority of recent papers because of thelack of the relevant date therein. Nevertheless, we hope thatthe experimental results presented in this paper together withearlier ones quoted from the literature convincingly showthat in order to reproduce the high hardness and its thermalstability in the binary nanocomposites nc-MenN/a-Si3N4(Me = Ti, W, V, (Al 1−xTix), etc.) and nc-TiN/a-BN asufficiently high nitrogen activity (partial pressure of>0.001 mbar), a high temperature of 550–630◦C and highplasma density are needed as explained in[1] and furtherelaborated in our later papers. Although the detrimentaleffect of impurities on the hardness of our nc-TiN/a-Si3N4nanocomposites was recognized and reported by us al-ready 5 years ago[23], little attention was paid to it sofar.

5. Summary and conclusions

Superhard binary nc-TiN/a-Si3N4 nanocomposites with astable nanostructure and hardness of≥45 GPa can be pre-pared by reactive magnetron sputtering when the nitrogenpressure of≥0.002 mbar, that corresponds to the regimeof fully segregated immiscible TiN and Si3N4 phases, anda sufficiently high deposition temperature of 550–630◦C,that assures a sufficiently fast diffusion which is needed tocomplete the phase segregation and the formation of strongand sharp interface during the deposition, are used. Impu-rities, such as hydrogen and oxygen (and chlorine in thecase of coatings deposited by plasma CVD), that may beincorporated into the coatings either from the process orfrom the residual gas during the deposition, reduce the max-imum achievable hardness to 40–45 GPa already at a levelof 0.1–0.2 at.%. Whenever the oxygen impurity content ex-ceeds 0.5 at.%, the hardness of the coatings remains below

35 GPa. At even higher impurity content, no hardness en-hancement is found when Si is added to TiN.

The lack of reproducibility of our results reported by someworkers (e.g.[40,53]) can be plausibly explained either bya too low nitrogen pressure and deposition temperature orby impurities, or by a combination of all.

The correlation between the hardness and biaxial com-pressive stress as found by Vaz[41] (seeFig. 2c) for his“Ti–Si–N” and by Musil and co-workers[11] for TiN(Fig. 2d) does not mean that the hardness enhancement inthe coatings of Vaz et al. is due to energetic ion bombard-ment during their deposition for the following reasons: Thehardness of these “Ti–Si–N” coatings remains almost con-stant upon annealing whereas the hardness enhancementdue to the ion bombardment in TiN and other similar coat-ings (HfB2, ZrN/Ni, CrxN/Ni, etc.) vanishes upon annealingto 400–600◦C.

The high thermal stability of up to 1100◦C found in ourcoatings is the evidence that, in this case, the superhardnessis due to the stable nanostructure. This stable nanostructurealso provides the coatings with a high oxidation resistancethat is comparable with or better than the state-of-the-artindustrial (Ti1−xAlx)N and (TiAlCrY)N coatings[1].

Acknowledgements

We would like to thank Drs. G. Dollinger and A.Bergmaier for the ERD analyses. The work has beensupported in part by the European Commission GrowthProgramme in the frame of the 6th RTD Framework Pro-gramme under the Project “MACHERENA” and by theNATO SfP Project No. 972379 “Protection Coatings”.

References

[1] S. Veprek, S. Reiprich, Thin Solid Films 268 (1995) 64.[2] S. Veprek, J. Vac. Sci. Technol. A 17 (1999) 2401 (review paper).[3] R. Hauert, J. Patscheider, Adv. Eng. Mater. 2 (2000) 247.[4] S. Zhang, D. Sun, Y. Fu, J. Mater. Sci. Technol. 18 (2002) 485.[5] S. Zhang, D. Sun, Y. Fu, H. Du, Surf. Coat. Technol. 167 (2003) 113.[6] S. Veprek, A.S. Argon, J. Vac. Sci. Technol. B 20 (2002) 650.[7] J. Musil, Surf. Coat. Technol. 125 (2000) 322.[8] H. Männling, D.S. Patil, K. Moto, M. Jilek, S. Veprek, Surf. Coat.

Technol. 146–147 (2001) 263.[9] A. Niederhofer, T. Bolom, P. Nesladek, K. Moto, C. Eggs, D.S. Patil,

S. Veprek, Surf. Coat. Technol. 146–147 (2001) 183.[10] Such a hardness enhancement, that may reach 60–80 and 100 GPa in

ordinary coatings, such as TiN and (AlTiV)N, respectively[11,12] isusually attributed to the compressive stress induced in the coatingsdue to the energetic ion bombardment. However, the mechanism ofthat enhancement is more complex because a biaxial stress ofσbiax

≈ −5 to −8 GPa< 0 can cause only an enhancement of the flowstress (hardness) by that valueHbiax = H0 − σbiax (see[13,14] andearlier references therein) and, therefore it cannot account for a highhardness of 50–100 GPa that is frequently reported in the literaturefor coatings, such as HfB2, TiN, (TiAlV)N and others whose bulkhardness amounts to 20–25 GPa (see[2,6,15]and references therein).


This can be easily understood because the biaxial stress acts onlyagainst the shear stress that develops under the indenter[16]. Theeffect of the energetic ion bombardment includes, besides of thatstress, a decrease of the crystallite size, densification of the grainboundaries and the formation of point and extended defects[17] (e.g.dislocation networks) all of them resulting in a much more complexsynergistic effect of hardening.

[11] V. Valvoda, R. Kuzel, R. Cerny, J. Musil, Thin Solid Films 156(1988) 53.

[12] J. Musil, S. Kadlec, J. Vyskocil, V. Valvoda, Thin Solid Films 167(1988) 107.

[13] T.Y. Tsui, W.C. Oliver, G.M.O. Pharr, J. Mater. Res. 11 (1996) 752.[14] A. Bolshakov, W.C. Oliver, G.M. Pharr, J. Mater. Res. 11 (1996) 760.[15] W. Herr, E. Broszeit, Surf. Coat. Technol. 97 (1997) 335.[16] G. Tabor, The Hardness of Metals, Clarendon Press, Oxford, 1951.[17] S. Veprek, F.-A. Sarott, Z. Iqbal, Phys. Rev. B 36 (1987) 3344.[18] O. Knotek, A. Barimani, Thin Solid Films 174 (1989) 51.[19] S. Veprek, P. Karvankova, J. Prochazka, H. Männling, MRS

Symp. Proc. 697 (2002) 27.[20] P. Karvankova, H. Männling, Ch. Eggs, S. Veprek, Surf. Coat. Tech-

nol. 146–147 (2001) 280.[21] P. Karvankova, Ph.D. Thesis, Technical University Munich, 2003 (in

English).[22] H.-D. Männling, Ph.D. Thesis, Technical University Munich, 2003

(in German).[23] S. Veprek, O. Nesladek, A. Niederhofer, F. Glatz, M. Jilek, M. Sima,

Surf. Coat. Technol. 108–109 (1998) 138.[24] A. Niederhofer, P. Nesladek, H. Männling, K. Moto, S. Veprek, M.

Jilek, Surf. Coat. Technol. 120–121 (1999) 173.[25] L.E. Toth, Transition Metal Carbides and Nitrides, Academic Press,

New York, 1971.[26] G.V. Samsonov, Nitrides, Naukova Dumka, Kiev, 1969 (in Russian).[27] P. Rogl, in: J.C. Schuster (Eds.), Phase Diagrams of Ternary Boron

Nitride and Silicon Nitride, ASM International, Materials Park, OH,1992.

[28] S. Sambasivan, W.T. Petuskey, J. Mater. Res. 9 (1994) 2362.[29] M. Paulasto, J.K. Kivilahti, F.J.J. van Loo, J. Appl. Phys. 77 (1995)

4412.[30] P. Karvankova, M.G.J. Veprek-Heijman, O. Zindulka, A. Bergmeier,

S. Veprek, Surf. Coat. Technol. 163–164 (2003) 149.[31] P. Holubar, M. Jilek, M. Sima, Surf. Coat. Technol. 133–134 (2000)

145.[32] J. Patscheider, MRS Bull. 28 (3) (2003) 180.[33] S. Li, Y. Shi, H. Peng, Plasma Chem. Plasma Process. 12 (1992) 287.[34] E.-A. Lee, K.H. Kim, Thin Solid Films 420–421 (2002) 371.[35] S. Veprek, M. Haussmann, S. Reiprich, J. Vac. Sci. Technol. A 14

(1996) 46.[36] S. Veprek, M. Haussmann, S. Reiprich, L. Shizhi, J. Dian, Surf.

Coat. Technol. 86–87 (1996) 394.[37] M. Diserens, J. Patscheider, F. Lévy, Surf. Coat. Technol. 108–109

(1998) 241.[38] M. Diserens, J. Patscheider, F. Lévy, Surf. Coat. Technol. 120–121

(1999) 158.[39] S.H. Kim, J.K. Kim, K.H. Kim, Thin Solid Films 420–421 (2002)

360.[40] W.J. Meng, X.D. Zhang, B. Shi, R.C. Tittsworth, L.E. Rehn, P.M.

Baldo, J. Mater. Res. 17 (2002) 2628.

[41] F. Vaz, Ph.D. Thesis, University of Minho, Portugal, 2000 (in Por-tuguese; we could not find this figure in any of the publications ofVaz et al.).

[42] S. Veprek, S. Mukherjee, P. Karvankova, J.L. He, K. Moto, J. Proc-hazka, A.S. Argon, J. Vac. Sci. Technol. A 21 (2003) 532.

[43] S. Veprek, S. Mukherjee, P. Karvankova, H.-D. Männling, J.L. He,K. Moto, S. Argon, Thin Solid Films 436 (2003) 220.

[44] J. Prochazka, Ph.D. Thesis, Technical University Munich, 2003 (inGerman).

[45] S. Christiansen, M. Albrecht, H.P. Strunk, S. Veprek, J. Vac. Sci.Technol. A 14 (1996) 46.

[46] S. Veprek, A. Niederhofer, K. Moto, T. Bolom, H.-D. Männling, P.Nesladek, G. Dollinger, A. Bergmaier, Surf. Coat. Technol. 133–134(2000) 152.

[47] I.C. Noyan, J.B. Cohen, Residual Stress, Springer, New York, 1987.[48] S. Veprek, Z. Iqbal, H.R. Oswald, F.-A. Sarott, J.J. Wagner, A.P.

Webb, Solid State Commun. 39 (1981) 509.[49] S. Veprek, Z. Iqbal, F.-A. Sarott, Phil. Mag. 45 (1982) 137.[50] S. Veprek, in: P. Pinard, S. Kalbitzer (Eds.), Proc. Europ. Mater.

Res. Society, Strasbourg, France, 1983, Les éditions de physique,Les Ullis, France, 1984, p. 425.

[51] P. Karvankova, D. Azinovic, M.G.J. Veprek-Heijman, S. Veprek,Surf. Coat. Technol., 2003, submitted for publication.

[52] Notice that at the “base pressure of 10−6 mbar” reported in the workof Hu et al. [53] (and in many others), the flux of the impuritymolecules from the background gas corresponds to about 0.5 mono-layer per second. Because of the reactive sticking coefficient of O2

and H2O on clean metallic surfaces is close to 1, oxygen impuri-ties of the order of≥1 at.% are expected to be incorporated in thegrowing coatings provided the oxygen and water fraction was about20% as in air (underestimate) and deposition rate was in the rangeof 0.5–1 nm/s (overestimate).

[53] X. Hu, Z. Han, G. Li, M. Gu, J. Vac. Sci. Technol. A 20 (2002) 1921.[54] W.-D. Münz, J. Vac. Sci. Technol. A 4 (1986) 2717.[55] S. Veprek, H.-D. Männling, A. Niederhofer, D. Ma, S. Mukherjee,

J. Vac. Sci. Technol. B 22 (2004) L5.[56] J. Musil, H. Zeman, F. Kunc, J. Vlcek, Mater. Sci. Eng. A 340

(2003) 281.[57] S. Veprek, S. Mukherjee, H.-D. Männling, J.L. He, Mater. Sci. Eng.

A 340 (2003) 292.[58] J.W. Meng, et al., Invited Paper E3/F1-1-1, Presented at the Inter-

national Conference on Metallurgical Coatings and Thin Films, SanDiego, USA, April 19–23, 2004, Unpublished.

[59] F. Vaz, L. Rebouta, P. Godeau, J. Pacaud, H. Garem, J.P. Riviere,A. Cavaleiro, E. Alves, Surf. Coat. Technol. 133–134 (2000) 307.

[60] F. Vaz, L. Rebouta, S. Ramos, M.F. da Silva, J.C. Soares, Surf. Coat.Technol. 108–109 (1998) 236.

[61] F. Vaz, L. Rebouta, B. Almeida, P. Godeau, J. Pacaud, J.P. Riviere,J. Bessa e Sousa, Surf. Coat. Technol. 120–121 (1999) 166.

[62] F. Vaz, L. Rebouta, R.M.C. da Silva, M.F. da Silva, J.C. Soares,Vacuum 52 (1999) 209.

[63] F. Vaz, L. Rebouta, M. Andritschky, M.F. da Silva, J.C. Soares, J.Mater. Process. Technol. 92–93 (1999) 169.

[64] J. Patscheider, April 2004, Private communication.[65] F. Vaz, 2003, Private communication.[66] J. Prochazka, F. Vaz, L. Rebouta, S. Carvalho, M.G.J. Veprek-

Heijman, S. Veprek, Surf. Coat. Technol., submitted for publication.

Conditions required for achieving superhardness of ≥45GPa in nc-TiN/a-Si3N4 nanocomposites

Documents