SCANNING VOL. 22, 258–262 (2000) Received: January 10, 2000© FAMS, Inc. Accepted with revision: May 26, 2000

High-Resolution Scanning Electron Microscopy Study of SputteredNanolaminated Ti/TiN Multilayers

C. LE PAVEN-THIVET, C. SANT, F. GRILLON,* P. HOUDY

Laboratoire Multicouches Nanométriques, Département Sciences des Matériaux, Université d’Evry Val d’Essonne; *U.M.R.7633, Ecole Nationale Supérieure des Mines de Paris / Armines, Centre des Matériaux - BP 87, Evry cedex, France

Summary: This work presents the morphologic and struc-tural study of nanolaminated Ti/TiN multilayers usinghigh-resolution scanning electron microscopy (HR-SEM),coupled to x-ray reflectometry (XRR). The multilayershave been deposited by reactive rf-sputtering on siliconsubstrates. For large period thickness (λ=40 nm, 10 peri-ods), in XRR, the low number of interfaces makes theinterference less structured. An experimental pattern withbroad and weakly intense Braggs peaks is obtained, but isdifficult to simulate. On the other hand, HR-SEM obser-vation of cross sections gives excellent pictures of the mul-tilayer, so that precise measurements of the thickness canbe achieved: a 42 nm thick period is observed, formedwith 17 nm of Ti and with 25 nm of TiN. For small(Ti+TiN) period thickness (λ=2.5 nm, 120 periods), theXRR pattern exhibits intense and narrow Bragg peaks: thenumber of interfaces is sufficient to structure the interfer-ence and an intense signal is obtained. The best fit of sim-ulation is obtained for a 2.6 nm thin period, made of 0.9 nmof Ti and 1.7 nm of TiN. No laminated structure has beenobserved by cross-section HR-SEM observation becauseits resolution (around 2 nm at 10 kV) is larger than the layerthickness in a period. High-resolution SEM and XRR arethus two complementary techniques for the routine char-acterization of multilayers.

Key words: high-resolution scanning electron microscopy,nanometric, Ti/TiN, multilayer

PACS: 61.16.Bg, 68.55.-a

Introduction

Thin film and multilayers thickness measurement canbe achieved employing various methods, such asbackscattered electron (BSE) spectroscopy (Schlichtinget al. 1999), ellipsometry (Azza and Bashara 1987),Rutherford backscattering spectroscopy (RBS) (Perrière1987), x-ray reflectometry (XRR) (Boher et al. 1990),wavelength dispersive spectrometry (WDS) (Maurice etal. 1980), or direct observation by scanning electronmicroscopy (SEM). Nevertheless, in the nanometricrange, this measure is not trivial and problems can occurwhen one wants to observe layers in nanolaminated struc-tures. Obviously, transmission electron microscopy(TEM) is the reliable tool for observation and structuralstudy of these structures, but the great difficulty of thethinning step of samples is the reason it is not used in rou-tine characterization. Only a few papers on multilayersmention SEM characterization; in addition, most of themreport on microlaminated rather than nanolaminatedcomposites. Ebrahimi et al. (1988) report on electrode-posited Cu/Ag multilayered composites, with (Cu+Ag)period thickness λ ranges from 119 to 320 nm; multi-layers are observed in SEM mid-section fracture sur-faces. Zhang and Xue (1997) deal with electrodepositedNi/Cu multilayer film with λ=100 nm; multilayers areobserved by SEM cross-section observations after pol-ishing and etching steps. Ding et al. (1994) show cross-section SEM observations of sputtered Al/Al2O3 andTi/TiN multilayers, with λ ranging from 80 to 520 nm.In addition to TEM analysis, Chung et al. (1998) reporton laser-ablated NbAl3/Al multilayer thin films, withλ=80 nm.

We have used high-resolution-SEM (HR-SEM) for sys-tematic characterization of thin films and nanometricAl/Al2O3 multilayers (Le Paven-Thivet et al. 1998). In thispaper, we present a morphologic and structural study ofsputtered Ti/TiN multilayers, with period thicknessesλ=40 and λ=2.5 nm, using HR-SEM, coupled to XRR.Ti/TiN multilayers present high hardness and tribologicalproperties which make them suitable for hard and wear-resistant coating applications (Bull and Jones 1996, Ben-daia et al. 2000).

Address for reprints:

Claire Le Paven-ThivetL.M.N., Département Sciences des MatériauxUniversité d’Evry Val d’Essonneboulevard F. Mitterrand91025 Evry cedex, Francee-mail: claire.lepaven@chimie.univ-evry.fr

Experimental

Deposition of Multilayers

Ti and TiN monolithic films and Ti/TiN multilayershave been grown at ambient temperature by reactive rf-sputtering on <100> silicon wafers with a metallic titaniumtarget (Sant et al. 2000). The total pressure is 0.7 Pa, witha target to substrate distance of 7 cm. For the TiN growth,the ratio of partial pressures Pargon / Pnitrogen is about 6 andis controlled by a mass spectrometer through two massflows. The power density is 2 W/cm2, and the depositionrates of Ti and TiN are 0.06 and 0.02 nm.s-1, respectively.

In this paper, we report on two multilayers: the first onewith a (Ti+TiN) period thickness λ=40 nm and the secondone with λ=2.5 nm. In a period, Ti layer and TiN layer areexpected to have the same thickness; that is, 20 nm in theformer multilayer and 1.25 nm in the later multilayer. Totalthicknesses are 400 nm (10 periods) and 300 nm (120 peri-ods), respectively.

Microstructure Characterization

Observations have been made with a thermally assistedfield emission gun SEM. The Zeiss DSM 982 Gemini(Carl Zeiss, Inc., Thornwood, N.Y., USA) has a coaxial in-lens secondary detector, with which a resolution in nanome-ter range is obtainable even at low accelerating voltages.This detector was used in cross-section observations, withan accelerating voltage of 5 and 10 kV. The magnificationof the SEM is calibrated using an array for the working dis-tance and high voltage used; the accuracy of the measuresis < 1%. The error on the measured thickness is estimatedaround 1 nm.

For film surface observation, an external Everhart-Thornley detector with an accelerating voltage of 2 kV wasused to analyse the surface of samples.

No preparation of samples was made, except a rapid dustremoval by a CO2 gas blower. For cross-section observa-tion, since films are deposited on single-crystalline siliconsubstrates, samples were cleaved just before observation.

X-ray reflectometry experiments were conducted on aD8 Bruker ω−2θ goniometer, mounted on a conventionalx-ray generator (λCukα1

=0.154056 nm) and equipped byhalf Eulerian circle and Göbel mirrors. The experimentalpattern is fitted with a simulated pattern; the parameters arethickness, roughness, and density of each layer.

Results and Discussion

Titanium and Titanium Nitride Films

The surface morphology of monolithic Ti and TiN films,100 nm thick, has been characterized by HR-SEM obser-vation. In both cases, smooth surfaces have been observed,even if some diffused grains are observed at high magni-

C. Le Paven-Thivet et al.: HR-SEM study of nanolaminated Ti/TiN multilayers 259

fication, as seen in Figure 1a for the Ti film and Figure 1bfor the TiN film. These grains were not observed by highmagnification cross-section observation, so that the rough-ness of the films was estimated to be very low. This is inaccordance with the roughness values calculated from theXRR patterns: 1.8 nm for the Ti film and 0.7 nm for the TiNfilm. Smooth surfaces, associated with very low solid statediffusion processes, is a strong requirement to obtain a goodstratification of layers.

λ=40 nm Ti/TiN Multilayer

The surface morphology of this multilayer, characterizedby HR-SEM, is similar to that of Ti and TiN films. The mul-tilayer thus reproduces the morphology of the individuallayers.

An HR-SEM cross-section view of this multilayer,obtained with the in-lens detector, is shown in Figure 2a.This picture shows a topographical step structure with a

FIG. 1 High-resolution scanning electron microscopy pictures of thesurface morphology of (a) a Ti film and (b) a TiN film.

(a)

(b)

260 Scanning Vol. 22, 3 (2000)

pronounced edge contrast. When cutting the sample, thecleavage occurs perpendicular to the surface, along somespecific crystallographic directions of the silicon substrate.We think that the Ti/TiN multilayer is cut in the same man-ner. Even if it is cut at a certain angle; that is, the cross-sec-tion of the sample is not always perpendicular to the elec-tron beam, the angle may be low in comparison with thenanometric distances found in the sample; that is, the erroron the measure is negligible.

Our interpretation of this picture is that bright lines cor-respond to the edges of the steps and that a step is a(Ti+TiN) period. The dark zone under the multilayer is thesilicon substrate. Thus, periods in the multilayer are clearlyevidenced: 10 periods are counted, in agreement with thedeposition. The measured total film thickness is 420 nm.

A high magnification of this view is presented in Figure2b. From inside to outside, the multilayer begins with a Tilayer and terminates with a TiN layer. In that case, thebrightest lines; that is, on the top of the steps, correspond

to the interface between Ti (up) and TiN (down); this iscalled Ti / TiN interface (Ti on TiN). The measured meanperiod thickness is 42 nm.

The contrast of TiN on Ti ( TiN/Ti interface) is more dif-fuse. Features in several periods, as seen in Figure 2b, canlead to an estimation of the thickness of the layers: 17 nmfor Ti and 25 nm for TiN. The difference with the expectedvalues could be caused by a deviation of the deposition ratesor a local inhomogeneity of thicknesses. The measuredthicknesses are listed in Table I.

The experimental XRR pattern of the λ=40 nm Ti/TiNmultilayer is shown in Figure 3. The appearance of Braggpeaks is the signature of the existence of a periodic struc-ture in the sample. Here, the observed Bragg peaks arebroad and weakly intense. The simulation of the experi-mental pattern was not easy to achieve and, in fact, yieldedno satisfactory fit. Good fits are obtained, but only for a partof the experimental pattern; that is, either for the first threepeaks or for the latest peaks. No reliable simulated patterncan be presented here.

λ=2.5 nm Ti/TiN Multilayer

High-resolution SEM cross-section pictures are pre-sented in Figure 4a and b. No laminated structure can beobserved. This is well explained by the fact that the HR-SEM resolution, around 2 nm at 10 kV, is larger than thelayer thickness (1.25 nm). The measured total film thick-ness of 300 nm is equal to the expected value for 120(Ti+TiN) periods of λ=2.5 nm.

X-ray reflectometry was performed: the experimentalpattern shows two intense Bragg peaks (Fig. 5). The fit ofthis pattern was conducted with 120 periods of titanium,with a roughness of 0.5 nm and a theoretical density of 4.5,and with titanium nitride, with a roughness of 0.5 nm anda theoretical density of 5.4. The substrate was silicon, witha roughness of 0.1 nm and a theoretical density of 2.32. Thebest fit was obtained for a period made of 0.9 nm Ti and

FIG. 2 High-resolution scanning electron microscopy cross-sec-tion pictures of a λ=40 nm period thickness Ti/TiN multilayer: (a)global view of the multilayer, (b) high magnification of the multilayer.

FIG. 3 Experimental x-ray reflectometry pattern of a λ=40 nmperiod thickness Ti/TiN multilayer.

(a)

(b)

Inte

nsity

(a.

u.)

0 2 4 6 82 Theta (deg.)

C. Le Paven-Thivet et al.: HR-SEM study of nanolaminated Ti/TiN multilayers 261

1.7 nm TiN. As seen in Figure 3, these values allow a verygood fit with the two Bragg peaks. The simulated thick-nesses are listed in Table I.

Discussion

Smooth surfaces of the Ti and TiN layers have been evi-denced by HR-SEM surface observation (Fig 1). Well-defined Ti/TiN interfaces; that is, periods, can be achievedin Ti/TiN multilayers.

For a large period thickness, for example λ=40 nm, peri-ods and layers are clearly evidenced by HR-SEM (Fig 2).The measuring of the different thicknesses is possible. Onthe other hand, the XRR pattern exhibits weak Bragg peaks(Fig. 3) and the simulated fit is more difficult to achieve.

For small period thicknesses, for example λ=2.5 nm,cross-section HR-SEM does not allow the observation of

the layers (Fig. 4) because the resolution of the SEM islarger than their thickness. On the other hand, XRR patternsexhibit intense and narrow Bragg peaks, suitable for a reli-able fit (Fig. 5).

In XRR patterns, on the hypothesis that the interfaces arewell defined in morphology and chemical composition,achieved in Ti/TiN composite, the more interfaces there arein the multilayer (i.e., λ is small), the more intense and nar-row are the Bragg peaks. On the other hand, if the numberof interfaces is small (λ is large), the system of interferenceis less constructive and Bragg peaks become broad andweakly intense.

High-resolution SEM and XRR are thus two comple-mentary characterization techniques for the study of mul-tilayers with thicknesses ranging from the nanometer to themicrometer scale. Since no special preparation of samplesis needed for HR-SEM and XRR, it makes them verysuitable for systematic and rapid characterization of mor-phology and structure of films. Nevertheless, one has tokeep in mind that HR-SEM is a local investigation of thesample, when XRR is an average measure, so that smalldifferences in thickness may occur between these twotechniques.

Concerning HR-SEM, clear cross-section observationshave been obtained, but only with the in-lens detector. Thesignal obtained with this detector comes from the impactof the incident electron beam with the first two or threeatomic strata; that is, the so-called secondary electron I sig-nal (SE I). The consequences are as follows: (1) theobserved surface is the extreme surface; and (2) the reso-lution is very high, close to the diameter of the incidentbeam; that is, around 2 nm at 10 kV. Using the externalEverhart-Thornley detector, with the loss of resolution,the layers are very difficult to observe.

In the λ=40 nm period thickness Ti/TiN multilayer, dif-ferent layers are seen. They are attributed to Ti and TiN, but

FIG. 4 High-resolution scanning electron microscopy cross-sec-tion pictures of a λ=2.5 nm period thickness Ti/TiN multilayer: (a)global view of the multilayer, (b) high magnification of the multilayer.

FIG. 5 Experimental and simulated x-ray reflectometry patterns ofa λ=2.5 nm period thickness Ti/TiN multilayer.

(a)

(b)

Inte

nsity

(a.

u.)

0 2 4 6 82 Theta (deg.)

Experimental (black line)Simulation (grey line)

262 Scanning Vol. 22, 3 (2000)

chemical differentiation is not obtained because the col-lection of the BSEs is totally eliminated with the in-lensdetector. However, a small part of the electrons detected areSEs generated by BSEs, in a way that this signal containsa part of the chemical information. Nevertheless, we thinkthat the observed contrast is mainly a topographical contrast.

The specific collection of BSEs with a classical BSEdetector has given no results since resolution of this signal ispoorer than the SEs signal and periods are no longer observed.Quantitative chemical analysis by WDS is possible for tita-nium and nitrogen, even if the Ti L-line is very close to theN K-line. However, the resolution of the x-ray signal, aroundone micron, even if operating on the surface of the sampleswith different accelerating voltages, prevents any reliable andsystematic analysis of these nanolaminated films.

For HR-SEM surface observation, the external Ever-hart-Thornley detector was preferred to the in-lens detec-tor: the latter gives information on the top surface whichis prejudicial to our analysis because part of the contrast ofthe relief is lost. Better topographic contrast is obtainedwith the external detector. Anyway, low accelerating volt-age of 2 kV was used to minimize the signal coming fromthe inner sample.

Conclusion

Ti/TiN nanolaminated structures have been deposited byreactive rf-sputtering. The films were systematically char-acterised by HR-SEM and XRR.

For large (Ti+TiN) period thicknesses (λ=40 nm, 10periods), in XRR, the low number of interfaces makes theinterference weakly structured and the Bragg peaks are notintense. Thickness determinations from simulations aredifficult to obtain. On the other hand, HR-SEM observa-tion gives excellent pictures of the multilayer, and precisemeasurements of the thickness can be achieved.

For small period thicknesses (λ=2.5 nm, 120 periods),the XRR pattern exhibits intense and narrow Bragg peaks:the number of interfaces is sufficient to structure the inter-ference and an intense signal is obtained. Thickness valuesfrom the simulation are thus considered to be suitable. Nolaminated structure has been observed by cross-sectionobservation by HR-SEM because of the limitation of theresolution.

High-resolution SEM and XRR are thus two comple-mentary techniques for the characterization of morphologyand structure of multilayers. The HR-SEM is a reliable toolin routine characterization, versatile, and easy to use forobservation and measurement of films and multilayerswith thicknesses in the nanometric range.

Acknowledgment

The authors would like to thank G. Renou for technicalassistance.

References

Azzam RMA, Bashara NM: Ellipsometry and Polarized Light. North-Holland Personal Library, Elsevier Science Publishers B.V.(1987)

Bendaia M, Aubert P, Labdi S, Sant C, Sadi FA, Houdy P: Nano-indentation investigation of Ti/TiN multilayers films. J ApplPhys 87, 7753–7757 (2000)

Boher P, Houdy P, Schiller C: Grazing x-ray reflection analysis ofnanometric scale structures. J Appl Phys 68, 6133–6142 (1990)

Bull SJ, Jones AM: Multilayer coatings for improved performance.Surf Coat Technol 78, 173–184 (1996)

Chung H, Jilavi MH, Duffey TP, Mazumder J, Kriven WM: Nanocrys-talline NbAl3-Al multilayer thin films deposited by excimerlaser ablation. Scripta Materialia 38, 429–435 (1998)

Ding Y, Farhat Z, Northwood DO, Alpas AT: Mechanical propertiesand tribological behaviour of nanolayered Al/Al2O3 and Ti/TiNcomposites. Surf Coat Technol 68, 459–467 (1994)

Ebrahimi F, Zhai Q, Kong D: Mechanical properties of Cu/Ag mul-tilayered composites. Mater Sci Eng A 255, 20–32 (1998)

Le Paven-Thivet C, Malibert C, Houdy P, Albouy PA: RF-sputteringdeposition of al/Al2O3 multilayers. Thin Solid Films 336,373–376 (1998)

Maurice F, Merry L, Tixier R: Microanalysis Scanning ElectronMicroscopy. Les Editions de Physique (1980)

Perrière J: Rutherford backscattering spectrometry. Vacuum 37,429–432 (1987)

Sant C, Bendaia M, Aubert P, Labdi S, Houdy P: Interface effect ontribiological properties of titanium-titanium nitride structures.Surf Coat Technol 127, 167–173 (2000)

Schlichting F, Berger D, Niedrig H: Thickness determination ofultra-thin films using backscattered electron spectra of a newtoroidal electrostatic spectrometer. Scanning 21, 197–203 (1999)

Zhang W, Xue Q: Tribiological properties of nickel-copper multilayerfilm on beryllium bronze substrate. Thin Solid Films 305,292–296 (1997)

TABLE I Measured high-resolution scanning electron microscopy cross-section and simulated x-ray reflectometry thickness of the λ=40nm and λ=2.5 nm period thickness Ti/TiN multilayers

Period TiN thickness Ti thickness Roughness thickness Roughness

(nm) (nm) (nm) Density (nm) (nm) Density

Ti / TiN HR-SEM 42 17 / / 25 / /10 periods

Ti / TiN XRR 2.6 0.9 0.5 4.5 1.7 0.5 5.4120 periods

Abbreviations: HR-SEM=high-resolution scanning electron microscopy. XRR=x-ray reflectrometry.