Defects in nanocrystalline Nb films: Effect of sputtering temperature

Applied Surface Science 252 (2006) 3245–3251www.elsevier.com/locate/apsuscDefects in nanocrystalline Nb films: Effectof sputtering temperatureJ. Čížek a, *, O. Melikhova a , I. Procházka a , G. Brauer b , W. Anwand b ,A. Mücklich b , R. Kirchheim c , A. Pundt ca Department of Low-Temperature Physics, Faculty of Mathematics and Physics,Charles University, V Holešovičkách 2, CZ-180 00, Prague 8, Czech Republicb Institut für Ionenstrahlphysik und Materialforschung, Forschungszentrum Rossendorf, Dresden, Germanyc Institut für Materialphysik, Universität Göttingen, GermanyAvailable online 7 October 2005AbstractThin niobium (Nb) films (thickness 350–400 nm) were prepared on (1 0 0)Si substrate in a UHV chamber using the cathodebeam sputtering. The sputtering temperature T s was varied from 40 up to 500 8C and the influence of the sputtering temperatureon the microstructure of thin Nb films was investigated. Defect studies of the thin Nb films sputtered at various temperatureswere performed by slow positron implantation spectroscopy (SPIS) with measurement of the Doppler broadening of theannihilation line. SPIS was combined with transmission electron microscopy (TEM) and X-ray diffraction (XRD). We havefound that the films sputtered at T s =408C exhibit elongated, column-like nanocrystalline grains. No significant increase ofgrain size with T s (up to 500 8C) was observed by TEM. The thin Nb films sputtered at T s =408C contain a high density ofdefects. It is demonstrated by shortened positron diffusion length and a high value of the S parameter for Nb layer compared tothe well-annealed (defect-free) bulk Nb reference sample. A drastic decrease of defect density was found in the films sputtered atT s 300 8C. It is reflected by a significant increase of the positron diffusion length and a decrease of the S parameter for the Nblayer. The defect density in the Nb layer is, however, still substantially higher than in the well-annealed reference bulk Nbsample. Moreover, there is a layer at the interface between the Nb film and the substrate with very high density of defectscomparable to that in the films sputtered at T s < 300 8C. All the Nb films studied exhibit a strong (1 1 0) texture. The filmssputtered at T s < 300 8C are characterized by a compressive macroscopic in-plane stress due to lattice mismatch between thefilm and the substrate. Relaxation of the in-plane stress was observed in the films sputtered at T s 300 8C. The width of the XRDprofiles of the films sputtered at T s 300 8C is significantly smaller compared to the films sputtered at lower temperatures. Thisis most probably due to a lower defect density which results in reduced microstrains in the films sputtered at higher temperatures.# 2005 Elsevier B.V. All rights reserved.Keywords: Niobium films; Cathode beam sputtering; Slow positron implantation spectroscopy; X-ray diffraction* Corresponding author. Tel.: +420 2 2191 2788; fax: +420 2 2191 2567.E-mail address: jcizek@mbox.troja.mff.cuni.cz (J. Čížek).0169-4332/$ – see front matter # 2005 Elsevier B.V. All rights reserved.doi:10.1016/j.apsusc.2005.08.083

3246J. Čížek et al. / Applied Surface Science 252 (2006) 3245–32511. IntroductionThe growth mode of thin films prepared bysputtering is influenced by the sputtering rate (i.e.the number of atoms which hit the substrate per unit oftime) and by the sputtering temperature T s (thetemperature of substrate). Higher T s leads to anenhanced mobility of the sputtered atoms in thesubstrate plane and long-range diffusion processesstart to play an important role in the growth of the film.The effect of T s is clearly visible on niobium (Nb)films [1]: polycrystalline Nb films are produced bysputtering at room temperature, while a high sputteringtemperature T s = 850 8C results in a formation ofepitaxial films. One can expect that the sputteringtemperature influences not only the grain size but alsothe density and type of lattice defects in thin films.Thus, it is highly desirable to perform defect studies ofthin films sputtered at various T s . The characterizationof thin films prepared by sputtering is usuallyperformed by X-ray diffraction (XRD). The shapeof the diffraction profile is influenced by grain size aswell as microstrains which could originate, e.g. fromdislocations. On the other hand, point defects (e.g.vacancies) can only hardly be detected by XRD. Forthese reasons positron annihilation spectroscopy(PAS) is very useful for defect studies of thin Nbfilms because it exhibits a high sensitivity to openvolumepoint defects like vacancies, vacancy clusters,etc. [2]. Hence, XRD and PAS represent in certain waycomplementary techniques and their combination ininvestigations of thin films is very efficient.In the present work we investigated the microstructureof thin Nb films prepared at various T s . Defectstudies of the films were performed by slow positronimplantation spectroscopy (SPIS) with measurement ofthe Doppler broadening of the annihilation line. TheSPIS investigations were combined with XRD studiesand direct observation of the microstructure bytransmission electron microscopy (TEM). The purposeof this work was to clarify the influence of T s on themicrostructure of thin Nb films.2. Experimental detailsThin Nb films were prepared on polished (1 0 0)Sisubstrates in an UHV chamber using cathode beamsputtering at sputtering temperatures T s = 40, 100,200, 300, 400 and 500 8C. The desired T s wasstabilized during sputtering within 1 8C. In addition,one Nb film was sputtered at T s =408C andsubsequently annealed at 500 8C for 1 h in UHV(10 10 mbar). The surface of all samples was coveredwith a 20 nm thick Pd cap in order to preventoxidation. The thickness of the films was measured byTEM and found to lie in the range 350–400 nm forvarious films. The SPIS studies were performed at amagnetically guided positron beam ‘‘SPONSOR’’ [3]with positron energies adjustable from 0.03 to 36 keV.Energy spectra of annihilation gamma rays weremeasured by a Ge detector with an energy resolutionof 1.09 0.01 keV at 511 keV. The texture measurementswere carried out on a four-axis Philips X’pertMPD diffractometer using Co Ka radiation. XRDmeasurements of lattice constants were performed atHasylab (DESY) using synchrotron radiation withwavelength l = 1.13 Å. The inter-planar distance wasmeasured in the out-of-plane direction (i.e. in thedirection perpendicular to the film surface, whichcorresponds to c =08) and in the directions tilted byc =458 and 608 with respect to normal to the surface.In such a way, we obtained information about stress inthe studied films. Diffraction profiles were fitted by thePearson VII function. TEM studies were performedwith a Philips CM300SuperTWIN microscope operatingat 300 kV. Thin foils for cross sectional TEMwere produced by conventional preparation usingGatan precision ion polishing system (PIPS).3. Results and discussionA bright field TEM image of the film sputtered atT s =408C is shown in Fig. 1a. The film exhibitselongated ‘‘column-like’’ grains. The width ofcolumns does not exceed 100 nm. Typically it liesaround 50 nm. A high resolution image of a column isshown in Fig. 1b. The columns are divided horizontally(i.e. in direction perpendicular to film-growingdirection) into two ‘‘generations’’ of sub-columns.The ‘‘first generation’’ sub-columns are situated closeto the Si substrate, while the ‘‘second generation’’ subcolumnsare situated close to the film surface. A brightfield TEM image of the film sputtered at T s = 500 8Cisshown in Fig. 2a. From comparison of Figs. 1a and 2a

J. Čížek et al. / Applied Surface Science 252 (2006) 3245–3251 3247Fig. 1. (a) A bright field TEM image of the film sputtered at T s =408C and (b) a high resolution image of a single column-like grain in the film.it is clear that there is basically no differencedetectable by TEM between the film sputtered at40 8C and at 500 8C. A high resolution TEM image ofthe film sputtered at 500 8C is shown in Fig. 2b.A strong 1 1 0 texture was found in all the studiedfilms. Most of grains are oriented so that the {1 1 0}planes are situated parallel with the film surface.However, the lateral orientation of grains is random.The texture becomes better developed at higher T s .In order to illustrate the effect of T s on the X-raydiffraction pattern, the XRD profiles corresponding to(1 1 0)Nb reflection for selected films sputtered atvarious T s are shown in Fig. 3a and b. The profilesshown in Fig. 3a were measured in the directionc =08, i.e. the scattering vector is perpendicular tothe film surface. The XRD profiles plotted in Fig. 3bwere measured on tilted sample at position c =608,i.e. the angle included by the scattering vector and thenormal to the film surface is 608. Thus, the interplanardistance determined from position of thereflections in Fig. 3a and b, respectively, correspondto the distance between the {1 1 0} planes in the outof-planedirection and in the direction tilted by 608with respect to normal to the film surface. Thedistance d 110 between the {1 1 0} planes in bulk Nbis indicated in Fig. 3a and b by a dashed line. One cansee from Fig. 3a that d 110 in the direction c =08 ishigher than that for bulk Nb in all the films studied.On the other hand, d 110 in the direction c =608 isslightly smaller than in bulk Nb. It indicates anexistence of compressive in-plane stresses in the filmscaused by a lattice mismatch between the film and theSi substrate. The magnitude of the compressive stressdecreases with increasing T s , which is reflected by ashift of the reflections towards the positions for bulkNb. One can see from Fig. 3a that there is a significantFig. 2. (a) A bright field TEM image of the film sputtered at T s = 500 8C and (b) a high resolution image of a single column-like grain in the film.

3248J. Čížek et al. / Applied Surface Science 252 (2006) 3245–3251Fig. 3. XRD profiles for selected films measured at position (a) c =08 and (b) c =608. Thicker solid line, T s =408C; dotted line, T s = 200 8C;dashed line, T s = 300 8C; thin solid line, T s = 500 8C; dash-dotted line, T s =408C with subsequent annealing at 500 8C for 1 h. The d 110 interplanardistance in the bulk Nb is indicated by the vertical-dashed lines.change of the position of the (1 1 0) reflection forT s 300 8C. Thus, starting from T s = 300 8C thecompressive stresses in the films are partially relaxed.Not only the position but also the width of thereflections depends strongly on T s . The broadening ofXRD profiles is caused either by small grain size or bymicrostrains induced by local elastic field of certaindefects, e.g. misfit dislocations. From Fig. 3a itisclear that the sample prepared at T s =5008C exhibitsthe narrowest XRD profile. On the other hand, thehighest width of the XRD profile was found on thefilm sputtered at T s =408C. As no difference in grainsize between these two samples has been observed byTEM, we conclude that the narrowing of XRD profilefor the film sputtered at T s =5008C is due to adecrease of defect density. The increase of T s from 40to 200 8C leads to a decrease of width of XRD profilescaused by a decrease of defect density. However, therelaxation of compressive stresses at T s =3008Cwhich leads to a shift of the peak position isaccompanied by an increase of the width of thereflection, see Fig. 3a. It could be due to introductionof misfit dislocations which accommodate thestresses caused by the lattice mismatch. Furtherincrease of T s from 300 to 500 8C leads again to asignificant narrowing of XRD profiles due to decreaseof defect concentration in the film. Note that the filmsputtered at T s =408C and subjected to subsequentannealing at 500 8C is characterized by relaxation ofthe stresses, i.e. a shift of the peak position towardsthat for bulk Nb, however, the width of the profileremains large. It indicates that defect density in thisfilm is comparable with the film sputtered atT s =408C.The reflection measured at c =08 for the filmsputtered at T s =408C is asymmetric. It can be wellseen in Fig. 4a where it is plotted in a finer scale. Theshape of the profile indicates that it is a superpositionFig. 4. (a) XRD profile for the film sputtered at T s =408C, measured in position c =08 and (b) XRD profile for the film sputtered at T s = 300 8C,measured in position c =08. The profile is a superposition of several reflections shown by the dotted lines.

J. Čížek et al. / Applied Surface Science 252 (2006) 3245–3251 3249of several reflections, namely: (i) the (1 1 0) reflectionfrom the ‘‘first generation’’ sub-columns in Nb layer,(ii) the (1 1 0) reflection from the ‘‘second generation’’sub-columns in Nb layer and (iii) a weak andbroad (1 1 1) reflection from the Pd cap. Note that it isnot possible to fit the XRD profile properly without theassumption of different distances d 110 in the ‘‘firstgeneration’’ and in the ‘‘second generation’’ subcolumns.The same result was obtained on a thicker(1.1 mm) Nb film sputtered under the same conditions(see ref. [4] for more detailed discussion). Theasymmetry of XRD profiles disappears at T s 300 8C8C and the profiles are then well-fitted by a single(1 1 0) reflection from the Nb layer and the reflectionfrom Pd cap. As an example, the XRD profile for thefilm sputtered at T s = 300 8C is shown in Fig. 3b.Additionally, the inter-planar distance d 110(Nb(1 1 0) reflection) in the direction c =608 andthe distance d 200 (Nb(2 0 0) reflection) in thedirection c =458 were measured. Both the d 110and d 200 distances for the two generations of subcolumnslie in these directions too close to each otherto separate their reflections in XRD spectrum.The lattice constants a calculated from the interplanardistance in the directions c =08,458 and 608 areplotted in Fig. 5a as a function of T s . The lattice constantfor bulk Nb [5] is shown in the figure by a dashed line.Below T s = 300 8C, the inter-planar distance in the two‘‘generations’’ of sub-columns differs in the directionc =08. The ‘‘first generation’’ sub-columns exhibit ahigher d 110 in this direction than the ‘‘secondgeneration’’ sub-columns. On the other hand, aboveT s = 300 8C, there is no significant difference betweend 110 in both ‘‘generations’’ of sub-columns and thereis only a single value of a for c =08. One can see inFig. 5a that a in the direction c =08 lies significantlyabove that for bulk Nb. A progressive decrease of atakes place from T s = 40–200 8C, while at higherT s > 300 8C it becomes approximately constant. Thevalues of a in the direction c =608 are lower than in thebulk Nb. It testifies the existence of compressivestresses in the in-plane direction in all the films studied.Only moderate changes of a with T s can be seen in thedirections c =458 and 608. The main feature is adecrease of a at T s = 300 8C. Annealing of the film at500 8C after sputtering (T s =408C) leads to asubstantial decrease of a, see Fig. 5a, which becomeseven lower than that for the film sputtered atT s = 500 8C. Thus, annealing of the as-sputtered filmsseems to be an effective way how to relax thecompressive stresses.The dependence of the S parameter on positronenergy E for selected Nb films is plotted in Fig. 5b. Alocal minimum of S at E = 1 keV is due to positronannihilations in the Pd cup. It was confirmed bymeasurement of a reference Pd film. Subsequentincrease of S is caused by an increasing fraction ofpositrons annihilating inside the Nb layer. In theenergy range from 5 to 7 keV, the contribution ofpositron annihilations inside the Nb layer is maximaland S exhibits a plateau-like behavior. At higherenergies, positrons start to penetrate into the SiFig. 5. (a) The lattice constant a determined from inter-planar distances in various directions as a function of T s . Full circles, direction c =08,‘‘first generation’’ sub-columns; open circles, direction c =08, ‘‘second generation’’ sub-columns; full squares, direction c =08, both‘‘generations’’ of sub-columns (T s 300 8C); full triangles, direction c =458; open triangles, direction c =458, the gray symbols correspondsto the film sputtered at T s =408C and subjected to subsequent annealing at 500 8C; gray circle, direction c =458; gray triangle oriented-up,c =458; gray triangle oriented-down, c =608 and (b) dependencies of the S parameter on positron energy E for selected films. The solid linesshow the fit performed by VEPFIT.

3250J. Čížek et al. / Applied Surface Science 252 (2006) 3245–3251Fig. 6. Results of fit of the S(E) curves (a) dependence of fitted S on T s ; full circles, S for the Nb layer; open circles, S for the layer with misfitdislocations situated close to the interface between the films and the substrate; full triangle, S for the Nb layer for the film sputtered at T s =408Cand subjected to subsequent annealing at 500 8C; open triangle, S for the layer with misfit dislocations for the film sputtered at T s =408C andsubjected to subsequent annealing at 500 8C and (b) dependence of positron diffusion length on T s ; full circles, Nb layer; open circles, the layerwith misfit dislocations, the triangles denote the results for the film sputtered at T s =408C and subjected to subsequent annealing at 500 8C; fulltriangle, Nb layer; open triangle, the layer with misfit dislocations.substrate which leads to further increase of S. As onecan see in Fig. 5b, there is not much differencebetween the S(E) curves for T s = 40 and 200 8C as wellas for the film sputtered at T s =408C and subjected toadditional annealing at 500 8C. On the other hand, theshape of the S(E) curves has changed remarkably forthe film sputtered at T s = 300 and 500 8C. The S(E)curves were fitted by the VEPFIT software package[6] (see the solid lines in Fig. 5b). For T s < 300 8C, theS(E) curves were well-fitted using a three-layer model:Pd cap, Nb layer and Si substrate. However, thisapproach failed for the films sputtered at T s > 300 8C.In order to create a model which fits satisfactorily theS(E) curves for the films sputtered at highertemperatures it was necessary to assume the existenceof an additional layer with a high density of defectssituated close to the interface between the film and Sisubstrate. Such a layer could contain predominantlymisfit dislocations which accumulate the latticemismatch between the film and the substrate. Thedependence of the fitted S parameter for Nb layer on T sis plotted in Fig. 6a. For T s 300 8C, the S parameterfor the defected layer situated close to the filminterface with substrate is plotted in Fig. 6a as well.The fitted positron diffusion length L for the Nb layer(for T s 300 8C also for the layer with defects) isplotted in Fig. 6b as a function of T s . The S parameterlies well above that for the reference defect-free bulkNb at all T s . It indicates that a significant fraction ofpositrons annihilate from the trapped state at defects.Clearly, there is a pronounced change of S as well as Lat T s =3008C. The XRD studies have revealed thatrelaxation of the compressive stresses occurs atT s = 300 8C. This process leads probably to anintroduction of misfit dislocations and to the formationof the defect layer because the lattice mismatchbetween the film and the substrate is no morecompensated by the elastic strains. Thus, a layer withhigh density of misfit dislocations is formed close tothe interface between the film and the substrate, whilethe defect density in the rest of the film is reduced asindicated by a decrease of S and increase of L. One cansee in Fig. 6b that the positron diffusion length in thedefected layer is comparable with that in the filmssputtered at T s < 300 8C and it does not change withT s . The positron diffusion length for the remaining Nblayer exhibits a moderate increase also at T s > 300 8C,accompanied by a decrease of corresponding Sparameter. It indicates a further reduction of thedefect density in the films.4. ConclusionsThe influence of the sputtering temperature T s(varied from 40 to 500 8C) on the microstructure ofthin Nb films was studied. The films are characterizedby column-like grains (width < 100 nm) and compressivein-plane stresses. It was found that the grainsize is basically independent on T s . A partialrelaxation of the stresses occurs at T s = 300 8C andis demonstrated by a shift of XRD reflections. SPIS

J. Čížek et al. / Applied Surface Science 252 (2006) 3245–3251 3251studies have shown that the stress relaxation isaccompanied by an introduction of misfit dislocationsin a layer situated close to the interface between thefilm and the substrate, while the defect density in theremaining part of the film decreases.AcknowledgementsThis work was supported by The Czech ScienceFoundation (project no. 202/05/0074), The Ministry ofEducation, Youth and Sports of The Czech Republic(projects nos. 1K03025 and MS 0021620834) and TheAlexander von Humboldt Foundation.References[1] U. Laudahn, A. Pundt, M. Bicker, U.V. Hülsen, U. Geyer, T.Wagner, R. Kirchheim, J. Alloys Compd. 293–295 (1999)490.[2] P. Hautojärvi, C. Corbel, in: A. Dupasquier, A.P. Mills, Jr.(Eds.), Positron Spectroscopy of Solids, IOS, Amsterdam, 1995,p. 491.[3] W. Anwand, H.-R. Kissener, G. Brauer, Acta Phys. Pol. A 88(1995) 7.[4] J. Čížek, I. Procházka, G. Brauer, W. Anwand, A. Mücklich, R.Kirchheim, A. Pundt, this volume.[5] ICDD (International Centrum for Diffraction Data) powderdiffraction pattern database PDF-2.[6] A. van Ween, H. Schut, M. Clement, J. de Nijs, A. Kruseman,M. Ijpma, Appl. Surf. Sci. 85 (1995) 216.