Barnett-4薄膜

Structure,stability,and mechanical properties of epitaxialWÕNbN superlatticesA.Madan a)and S.A.BarnettDepartment of Materials Science and Engineering and Advanced Coating Technology Group,Northwestern University,Evanston,Illinois60208A.Misra,H.Kung,and M.NastasiLos Alamos National Laboratory,Los Alamos,New Mexico͑Received3November2000;accepted19February2001͒Epitaxial W/NbN superlattices with modulation wavelengths⌳ranging from1.3to25nm were grown on MgO͑001͒substrates by dc reactive magnetron sputtering in Ar/N2mixtures.The epitaxial relationship between the layers is given by W͑001͒ʈNbN͑001͒and W͓110͔ʈNbN͓100͔.X-ray diffraction and Rutherford backscattering resultsfitted using simulations showed that the superlattices had well-defined planar layers with interface widths ofϷ0.2nm.Nanoindentation measurements showed superlattice hardnesses as high as33GPa compared to8for W and20for NbN.The superlattices showed little change in x-ray superlattice reflections or nanoindentation hardness after vacuum annealing up to the highest temperature tested,1000°C for6h.Thus,the layers remained intact during annealing,allowing the superlattice hardness enhancement to be retained.©2001American Vacuum Society.͓DOI:10.1116/1.1365133͔I.INTRODUCTIONThinfilm superlattices containing an elemental metal anda compound layer are a type of nanocomposite material thatcan exhibit substantial hardness increases,oftenϾ100%,over rule-of-mixtures values.1Hardness enhancements havebeen seen for polycrystalline Ni/TiN,2Pt/TiN,3Fe/TiC,4W/TiC,5Mo/NbN,6,7and W/NbN.6The latter four superlat-tices consist of a body centered cubic͑bcc͒metal͑M͒andB1-structure group-VIA compound͑C͒,with an epitaxialorientation of(001)Mʈ(001)C and͓110͔Mʈ͓100͔C.With this 45°rotation of their unit cubes about the͑001͒direction,there is a good lattice match.High quality epitaxial superlat-tices of Mo/NbN6and W/NbN6,8have been grown onMgO͑001͒over a wide range of periods.For W/TiC multi-layers,there is a transition from epitaxial to polycrystallinewith increasing modulation wavelength.5Metal/nitride nano-structures have the potential to be stable at elevated tempera-tures because of their low-energy coherent interfaces andtheir limited solubility.In a recent article,9it was shown thatMo/NbN superlattices were stable up to1h at1000°C.Forlonger anneals,the ternary compound MoNbN formed,re-sulting in a loss of x-ray diffraction superlattice reflections.Thefilms retained their high hardnesses͑Ͼ30GPa͒evenafter the loss of the superlattice reflections,apparently due tothe formation of a stable MoNbN/NbN nanostructure.Thehigh-temperature stability of the W/NbN system has notbeen studied;it may be more stable than Mo/NbN since thereis no ternary phase reported.In this article,we report a detailed study of epitaxialW/NbN superlattices.In addition to x-ray diffraction andnanoindentation measurements performed before and afterhigh-temperature annealing,more complete structural char-acterization of the as-depositedfilms is provided than in prior articles.6Section II describes the experimental proce-dure for deposition and characterization of the superlattices.The results of x-ray diffraction͑XRD͒scans and correspond-ing computer simulations are detailed in Sec.III.Rutherfordbackscattering͑RBS͒results are detailed in Sec.IV.Anneal-ing results and nanoindentation measurements are presentedin Sec.V.II.EXPERIMENTAL PROCEDUREW/NbN superlattices with modulation wavelengths⌳from1.3to24nm and l W/⌳ϳ0.5͑where l W is the W layerthickness and⌳is the period͒were deposited using an ultra-high vacuum reactive dc magnetron sputtering system thathas been described in detail elsewhere.10Films were depos-ited from two water-cooled magnetron sputter sources with5-cm-diameter,99.95%purity W and Nb targets in Ar–N2͑puritiesϾ99.999%͒mixtures at a total pressure of10 mTorr.The substrate temperature ofϷ800Ϯ50°C was suf-ficiently large that minimal nitrogen was incorporated intothe individual layers of W during growth,6even with nitro-gen partial pressures͑4–8mTorr͒high enough to form stoi-chiometric NbN.This results from the relatively low heat ofWN formation compared to that of NbN.7The sputteringcurrents and voltages were0.3A and440–540V for W and0.6A and440–560V for Nb.As the targets eroded,thecurrents were kept constant and the voltages of the gunsdecreased in the range indicated.Under these conditions,thedeposition rates for W and NbN were2.7and1.8Å/s,re-spectively.Totalfilm thicknesses wereϷ1␮m.The sub-strates were polished MgO͑001͒.A retarding-field analyzer mounted in the growth chamber was used for in situ Auger electron spectroscopy͑AES͒and low-energy electron dif-fraction͑LEED͒measurements.Both high-and low-angle XRD␪/2␪scans were made with a double-crystal diffractometer in theϩ/Ϫgeometry.a͒Electronic mail:a-madan@952952 J.Vac.Sci.Technol.A19…3…,MayÕJun20010734-2101Õ2001Õ19…3…Õ952Õ6Õ$18.00©2001American Vacuum SocietyThe diffractometer was equipped with a LiF͑002͒focusing monochromator.Since the intensity of the satellite peaks var-ied over5–6orders of magnitude,the power of the x-ray tube was adjusted,with a maximum of1.2kW used for the weakest peaks.Keeping an overlap of at least one peak,the different parts of the spectrum were then combined.The azi-muthal scans were done on a four circle diffractometer which allows rotation about␹,␸,and␪͑the symbols have their usual meanings͒.11The scans were done keeping␹and␪fixed and varying␸at intervals of0.2°.The low-angle XRD patterns were simulated using Paratt’s formalism.12The superlattice period⌳was deter-mined by matching the peak positions.The layer thickness ratio l W/⌳͑where l W is the W layer thickness͒was then adjusted so that the relative peak intensities matched.The high-angle XRD patterns were simulated using a kinematical calculation assuming a trapezoidal compositional modula-tion.The high-angle simulation is similar to other such calculations13,14and is described elsewhere.15Layer thick-nesses were taken from the low-angle simulation results.The lattice spacings of the individual constituents were adjusted so that the relative intensities between the most-intense peaks in the W͑002͒and the NbN͑002͒regions matched. Peak widths were accounted for by adjusting the following three factors.First,instrumental peak broadening was ac-counted for byfitting the MgO substrate peak.Second,crys-talline defects were included via a Gaussian distribution of interplanar spacings with width␴d around the average.Fi-nally,a Gaussian distribution of the modulation wavelengthwith width␴⌳was used to account for variations in layer thicknesses.Interface widths were then adjusted to obtain a bestfit to the peak intensities.RBS was performed with2MeV Heϩions to determine the chemical composition of the superlattices and ion chan-neling was performed with5.6MeV Heϩϩions to check the epitaxial quality of thefilm.Both normal and60°angle of incidence were used at2MeV but only normal incidence was needed for the channeling experiment.The layered structure was better resolved at60°tilt at small bilayer peri-ods.The experimental data was compared with simulations performed using the software RUMP.16The annealing experiments were performed in a vacuum furnace with a maximum temperature of1000°C and a base pressure of1.0ϫ10Ϫ8Torr obtained using turbomolecular and ion pumps.Typical test pressures during the anneals were5.0ϫ10Ϫ8Torr.After the furnace was brought to the annealing temperature,the sample was inserted via a load lock into the preheated furnace.After annealing,the sample was removed from the furnace and allowed to cool under vacuum.Nanoindentation measurements on as-deposited and an-nealedfilms were performed at room temperature,using a UMIS-2000ultramicroindentation system.The triangular Berkovich diamond indenter tip was calibrated using fused silica.17Each sample was indented ten times,with a maxi-mum load of8mN,which yielded typical maximum depths ofϳ80nm for the superlattices.The depth resolution of the instrument isϳ0.1nm.A typical experiment would include a load segment,a hold segment for100s,and an unload segment.The data shown are averages over the ten indents. III.AS-DEPOSITED FILMSA.Monolithicfilm structure and compositionPreliminary experiments were carried out to determine the optimum conditions for depositing epitaxial Wfilms on MgO in an Ar/N2atmosphere.The substrate temperature was varied from650to800°C.Good quality epitaxial growth was achieved atϷ800°C,as verified by spotty in situ LEED patterns.XRD scans confirmed the͑002͒orientation,with no peaks corresponding to WN.At lower temperatures,the W films showed an additional W͑011͒reflection.A N2partial pressureу4mTorr yielded stoichiometric NbN.Figure1shows the in situ AES spectra taken after deposition of monolithic NbN͑a͒and W͑b͒films in an Ar/N2mixture with a nitrogen partial pressure ofϳ6mTorr. Comparison of the NbN spectrum with prior results10indi-cates that thefilm was stoichiometric.The W spectrum shows a N peak barely above the background.This very low N content is not surprising given that the substrate tempera-ture of800°C is higher than where N desorbs from W–N films,Ϸ700°C.18These results show that essentially metallic W can be grown in the same Ar–N2mixtures used to deposit stoichiometricNbN.F IG.1.In situ Auger spectra from͑a͒a NbNfilm and͑b͒a Wfilm deposited in an Ar/N2atmosphere.The NbN is stoichiometric and W spectra shows little or no residual nitrogen.JVST A-Vacuum,Surfaces,and FilmsB.Superlattice x-ray analysis1.Pole figure scansThe epitaxial orientation of W/NbN superlattices on MgO ͑001͒was observed by aligning the diffractometer to observe the MgO ͕220͖,NbN ͕220͖,or W ͕110͖out-of-plane reflections,and rotating the sample azimuthally.Figure 2shows four peaks from MgO ͕220͖separated by 90°in ␸,corresponding to the angles where those planes were aligned with the diffractometer.The observed differences in intensity among each family of reflections occurs because of imperfect substrate orientation in the diffractometer.The NbN ͕220͖peak ␸values matched those of MgO,showing that the ori-entation was the same as the substrate.The four peaks from the W ͕110͖are rotated by 45°with respect to the NbN ͕220͖peaks.This shows that the epitaxial orientation is (001)W ʈ͑001͒NbN and ͓110͔W ʈ͓100͔NbN ,i.e.,there is a 45degree rotation of the unit cubes about the ͑001͒normal asshown schematically in Fig.3.The result is similar to that obtained for Mo/NbN superlattices.2.Low-angle XRDFigure 4depicts typical low-angle XRD scans for W/NbN superlattices with different ⌳and l W /⌳ϳ0.5.About 18re-flections are seen for the ⌳ϭ16.4nm superlattice.The strong superlattice reflections result from the large difference in x-ray scattering factors for the layers,and also indicate that the superlattices had relatively abrupt flat interfaces.The relative peak intensities depend on the layer thickness ratio.For l W /⌳ϳ0.5,every alternate peak is weak as seen in Fig.4.The low-angle patterns were used to accurately determine ⌳.The peaks of different orders n in the low angle XRD pattern occur at values of ␪n given by 19sin 2␪n Х͑␭/2⌳͒2n 2ϩ2␦,͑1͒where ␦is the correction to the average refractive index and ␭is the x-ray wavelength.A plot of sin 2␪vs n 2is a straight line and ⌳can be evaluated from ⌳ϭ␭2ͱslope.3.High-angle XRDFigure 5shows typical high-angle XRD patterns about the ͑002͒reflections for W/NbN superlattices (l W /⌳ϳ0.45)for different ⌳values.The expected Bragg peak positions for the ͑002͒reflections for W,NbN,and MgO are labeled.Mul-tiple strong superlattice reflections were observed in all cases.Similar XRD patterns with a large number of strong reflections were seen in another study.8The envelope func-tions have two maxima near the W and NbN Bragg-peak positions.As ⌳increases,thesuperlattice reflection spacingF IG .3.Schematic drawing showing the epitaxial relationship observed be-tween BCC Wand rocksalt NbN.F IG .4.Low-angle XRD patterns from W/NbN superlattices with superlat-tice periods ⌳ϭ3.48,6.95,13.1,and 16.4nm.The 3.48nm period super-lattice is 34.8nm thick whereas the other superlattices have a total thickness of 1␮m.Simulated ͑–͒low-angle XRD patterns for the ⌳ϭ3.48nm W/NbN superlatticeare also shown.F IG .2.Pole figure XRD scans of ͑a ͒NbN ͑220͒,͑b ͒W ͑110͒,and ͑c ͒MgO ͑220͒reflections from a ⌳ϭ3.8nm W/NbN superlattice.J.Vac.Sci.Technol.A,Vol.19,No.3,May ÕJun 2001decreases and the envelope function becomes stronger near the W ͑002͒and NbN ͑002͒positions and weaker between.The transition from an envelope function with a central maximum to one with two maxima occurs at relatively small ⌳due to the large ratio of out-of-plane lattice spacings a NbN ͓002͔/a W ͓002͔ϭ1.4,compared, e.g.,with NbN/TiN (a NbN ͓002͔/a TiN ͓002͔ϭ1.036).15Superlattices with a large ra-tio of out-of-plane spacings,such as Pt/Fe (a Pt ͓002͔/a Fe ͓002͔ϭ1.46),20Au/Fe (a Au ͓002͔/a Fe ͓002͔ϭ1.42),21and Au/Cr (a Au ͓002͔/a Cr ͓002͔ϭ1.42),22also have broad envelope func-tions with two maxima and hence also show superlattice re-flections over a large range in 2␪.4.XRD simulationsA best fit to an experimental low-angle XRD pattern is shown in Fig.4.The modulation wavelength ⌳ϭ3.48nm and the layer thickness ratio l W /⌳ϭ0.56obtained from this low-angle XRD simulation agree well with the growth rate calibrations and the ⌳obtained using Eq.͑1͒.The best-fit simulation of the high-angle XRD pattern from the ⌳ϭ3.86nm superlattice is shown in Fig.5.Interface widths obtained from both high-and low-angle simulations were Ͻ0.2nm,similar to values obtained previously for epitaxial Mo/NbN superlattices.6These relatively abrupt interfaces are not sur-prising given that the layer materials are immiscible in both these cases.In contrast,larger interface widths of Ϸ1nm are observed for miscible-layer TiN/NbN 14superlattices grown under similar conditions.The simulation lattice spacings in-dicate that the W was strained,whereas the NbN was not strained.This is similar to what was seen for Mo/NbN su-perlattices and is the net effect of the coherency and thermal-mismatch strains.6Cross-sectional transmission electron mi-croscopy images of a W/NbN superlattice from another study showed continuous well-defined layers with verticallines propagating vertically down through the image inter-rupting the lateral coherence of the layers.8This was attrib-uted to dislocation lines relieving the relative strain between the W and NbN.The lattice spacing fluctuations ␴d ϳ0.002nm indicate some crystalline imperfections,and are similar to the values obtained for TiN/NbN 15and Mo/NbN 6superlattices grown under similar conditions.Fluctuations in the modulation wavelength were ␴⌳ϳ0.2nm,and indicated some interface roughness.C.Rutherford backscattering …RBS …resultsThe epitaxial quality of the W and NbN layers on MgO was revealed in the ion-channeling experiments on a ⌳ϭ17nm superlattice,using 5.6MeV He ϩϩions,shown in Fig.6.To obtain the backscattering yield for a random orientation,the sample stage was wobbled Ϯ4°during the scan ͑solid lines in Fig.6͒.The RBS yield when the sample was not wobbled,but rather tilted with the incoming ions aligned along the easy ͓100͔channeling direction,is also plotted on Fig.6.The significant drop in the backscattering yield in the channeled sample,as compared to random,indicates that the superlattice was epitaxial.At the surface W peak ͑channel ϳ495in Fig.6͒,the minimum channeling yield,␹min ͑ratio of the backscattering yield from channeled to random spectra,23was calculated as 46%.The theoretical ␹min value for W is ϳ1.2%.23This higher value of channeling yield as compared to the theoretical estimate for a perfect crystal is attributed to the mosaic spread ͑full width at half maximum ϭ1.92°for W peaks in the XRD ␾scans ͒in these epitaxial films.Typically,for angular misorientations Ͼϳ1°the back-scattering yield will be above the theoretical minimum.23Due to the large number of very thin superlattice layers,spectra taken with a normal incidence beam did notclearlyF IG .5.High-angle XRD spectra around the 002reflections from W/NbN superlattices with periods ⌳ϭ3.86,6.95,and 13.1nm.The NbN ͑002͒and W ͑002͒Bragg angles,41.06°and 58.27°,respectively,are indicated.Simu-lated ͑᭹͒XRD patterns for the 3.86nm W/NbN superlattice are also shown.The parameters obtained from the simulations are l W /⌳ϭ0.53and a w ϭ0.322nm,a NbN ϭ0.439nm,␴⌳ϭ0.1nm,and ␴d ϭ0.002nm.F IG parison of the backscattering yields from a W/NbN sample that was wobbled ͑random orientation,solid lines ͒and a sample that was ori-ented for ion channeling ͑dotted lines ͒.The low backscattering yield in the channeled sample is consistent with epitaxial alignment of the superlattice.JVST A -Vacuum,Surfaces,and Filmsresolve the individual layers.However,when the sample was tilted by ϳ60°,as in the data shown in Fig.7,some of the bilayers near the surface can be separated.The resolution deteriorates with increasing depth due to straggling,i.e.,the spread in energy of the incident He ϩions increases with increasing depth.This allowed the compositions of the indi-vidual layers and the separation of elements across the inter-faces to be determined via simulations.The RBS spectrum from a superlattice with ⌳ϭ17nm,obtained using 2MeV He ϩions,is shown in Fig.7.Only the near-surface region,where distinct peaks from the layers are observed,is shown in Fig.7.Four clear oscillations in the yield are seen just below the W edge,corresponding to the four W layers nearest the surface.Oscillations are also seen below the Nb edge,but the oscillation amplitude is reduced due to interference between scattering from the W and Nb layers.The simulations results are superimposed on the ex-perimental data in Fig.7.The best fit was obtained for the following compositional profile:͑i ͒stoichiometric NbN lay-ers,͑ii ͒pure W layers ͑note,however,that a nitrogen con-tamination of Ͻ2at.%could not be clearly resolved by com-paring the experimental and simulated patterns ͒,͑iii ͒no interdiffusion of Nb and W across the interfaces,and ͑iv ͒equal numbers of atoms/cm 2͑within ϳ2%͒of the W and NbN layers,with each containing 55ϫ1015atoms/cm -ing the atomic density of W of 6.32ϫ1022atoms/cm 3the layer thickness obtained is ϳ8.7nm.This agrees well with the XRD data in Fig.4,where the simulation yielded a W layer thickness of 8.5nm.RBS experiments on samples with only one or two bilay-ers might allow more accurate determination of the compo-sitions.Nevertheless,the data presented here ͑Figs.6and 7͒conclusively show that the superlattices were epitaxial ͑con-sistent with XRD ͒,and that the W and Nb are well separated across the interface with no detectable N in the W layers and stoichiometric NbN layers.D.Hardness resultsFigure 8shows the typical loading–unloading sequences for W,NbN,and an as-deposited W/NbN superlattice ͑⌳ϳ3.2nm ͒.As expected,the W film showed the largest displacement depth and the largest area between the loading and unloading curves,indicating a relatively low hardness ͑8GPa ͒and substantial plastic deformation.The NbN film dis-placement depth was considerably smaller,and the hardness obtained was larger,Ϸpared to the monolithic films,the superlattices showed lower indent depths and much smaller areas between the loading and unloading curves,as shown in Fig.8.This indicates less plastic deformation and substantially higher hardness.The superlattice hardness was always larger than the rule-of-mixtures value of 14GPa,de-creasing from Ϸ33GPa at small ⌳͑2–3nm ͒to Ϸ22GPa at ⌳ϭ24nm.The hardness values for as-deposited films agreed well with a prior report.6It is to be noted that the creep values ͑displacement at maximum load ͒are 6.8nm for the W,14.5nm for the NbN,and 0.3nm for the as-deposited superlattice.The creep value is lower for the superlattice because of the interfaces hindering dislocation motion.Fig-ure 8also shows the loading–unloading curve for the W/NbN superlattice after annealing at 1000°C for 6h,which is discussed in the following section.IV.ANNEALING RESULTSFigure 9compares the ͑a ͒high-angle and ͑b ͒low-angle XRD scans from a W/NbN superlattice with ⌳ϭ3.2nm be-fore and after annealing at 1000°C for 6h.There is little change in the XRD peaks in the low-angle pattern.Every second peak remains lower in intensity for the annealed films indicating that the superlattice retained its original layer thickness ratio of Ϸ0.5.This is different from what was seen for Mo/NbN superlattices under the same annealing condi-tions,where the low-angle XRD peak intensitiesgraduallyF IG .7.Experimental RBS data compared with the simulation ͑dotted lines ͒assuming discrete W and NbN layers of equalthicknesses.F IG .8.Typical nanoindenter loading–unloading displacement curves for W ͑ϫ͒,NbN ͑᭿͒,and an as-deposited ⌳ϳ3.2nm W/NbN superlattice ͑–͒.The result for the superlattice after annealing at 1000°C for 6h is also shown ͑•••͒.J.Vac.Sci.Technol.A,Vol.19,No.3,May ÕJun 2001decreased with increasing annealing time,indicating degra-dation of the layered structure.7There is also no evidence of any interfacial compound for the annealed W/NbN superlat-tices,unlike Mo/NbN where reflections from the ternary compound MoNbN were observed as the annealed time in-creased.Note that the phase diagram for the W–Nb–N sys-tem does not show the existence of the ternary WNbN.24W has a substantially higher melting point͑3422°C͒than Mo ͑2623°C͒,and hence it is possible that the ternary compound would form at a higher temperature.The loading–unloading curve for the superlattice with ⌳ϭ3.2nm,after annealing at1000°C for6h,is shown in Fig.8.The curves look similar to those for the as-deposited film,and HϷ33GPa was obtained,the same as for the as-depositedfilm within experimental error.This is similar to the behavior observed for Mo/NbN,where the hardness re-mained stable or increased slightly upon annealing.9The W/NbN superlattices were much more stable than the mis-cible TiN/NbN superlattices,where the layers interdiffused rapidly and the hardness decreased at temperatures у800°C.25This is presumably due to the immiscibility of W and NbN.V.SUMMARY AND CONCLUSIONSEpitaxial W/NbN superlattices have been grown on MgO ͑001͒substrates in an Ar/N2atmosphere.A45°rotation of the W lattice with respect to the NbN lattice and the MgO substrate has been verified via polefigure scans.The super-lattices have a well-defined layered structure,as confirmed by low and high XRD scans along with RBS.W/NbN super-lattices retain their structure and hardness even after the samples have been annealed at1000°C for6h.ACKNOWLEDGMENTSTwo authors͑A.M.and S.B.͒would like to acknowledge thefinancial support of National Science Foundation Grant No.DMR-9442873.Three authors͑A.M.,M.N.,and H.K.͒acknowledge support at Los Alamos National Laboratory from DOE/OBES͑Dept.of Energy/Office of Basic Energy Science͒.1S.A.Barnett,in Physics of Thin Films,edited by M.Francombe and J.A. Vossen͑Academic,New York,1993͒,p.1.2X.Chu,M.S.Wong,W.D.Sproul,S.L.Rohde,and S.A.Barnett,J. Vac.Sci.Technol.A10,1604͑1992͒.3J.L.He,W.Z.Li,and H.D.Li,Jpn.J.Appl.Phys.Lett.37,679͑1997͒. 4C.H.Liu,W.Z.Li,and H.D.Li,Nucl.Instrum.Methods Phys.Res.B 95,323͑1995͒.5T.D.Moustakas,J.Scanlon,J.Y.Koo,H.W.Deckman,A.Ozekcin,R. Friedman,and J.A.McHenry,Mater.Sci.Eng.,B6,179͑1990͒.6A.Madan,Y.-Y.Wang,S.A.Barnett,C.Engstrom,H.Ljungcrantz,L. Hultman,and M.Grimsditch,J.Appl.Phys.84,776͑1998͒.7A.Madan,X.Chu,and S.A.Barnett,Appl.Phys.Lett.68,2198͑1996͒. 8R.M.Osgood III,J.E.Pearson,C.H.Sowers,and S.D.Bader,J.Appl. Phys.84,940͑1998͒.9C.Engstrom,A.Madan,J.Birch,M.Nastasi,L.Hultman,and S.A. Barnett,J.Mater.Res.15,554͑2000͒.10P.B.Mirkarimi,M.Shinn,and S.A.Barnett,J.Vac.Sci.Technol.A10, 618͑1992͒.11B.D.Cullity,Elements of X-ray Diffraction͑Addison-Wesley,Reading, MA,1978͒.12L.G.Paratt,Phys.Rev.A95,359͑1954͒.13M.B.Stearns,Phys.Rev.B38,8109͑1988͒.14G.Gladyszewski,Thin Solid Films170,99͑1989͒.15A.Madan,P.Yashar,M.Shinn,and S.A.Barnett,Thin Solid Films302, 147͑1997͒.16L.R.Doolittle,Ph.D.thesis,Cornell University,Ithaca,NY,1987.17W.C.Oliver and G.M.Pharr,J.Mater.Res.7,1564͑1992͒.18K.Y.Ahn,S.B.Brodsky,C.Y.Ting,and J.Kim,J.Vac.Sci.Technol. A4,3111͑1986͒.19B.K.Aggarwal,X-ray Spectroscopy͑Springer,Berlin,1979͒.20M.Sakurai,J.Appl.Phys.76,7272͑1994͒.21N.Nakayama,T.Okuyama,and T.Shinjo,J.Phys.:Condens.Matter5, 1173͑1993͒.22P.Bisanti,M.B.Brodsky,G.P.Felcher,M.Grimsditch,and L.R.Sill, Phys.Rev.B35,7813͑1987͒.23L.C.Feldman,J.W.Mayer,and S.T.Picraux,Materials Analysis by Ion Channeling͑Academic,New York,1982͒.24P.Villars,A.Prince,and H.Okamoto,Handbook of Ternary Alloy Phase Diagrams͑ASM International,Materials Park,Ohio,1995͒.25C.Engstro¨m,L.Hultman,A.Madan,S.A.Barnett,M.Nastasi,and C. Lavoie,unpublished.F IG.9.͑a͒X-ray diffraction patternsaround the002Bragg reflections forthe W/NbN superlattices with period ⌳ϭ3.2nm,both as-deposited and af-ter annealing at1000°C for6h.͑b͒X-ray reflectivity scans from W/NbNsuperlattices with period⌳ϭ3.2nm,both as-deposited and after annealingat1000°C for6h.JVST A-Vacuum,Surfaces,and Films。