Anda di halaman 1dari 10

Dense fully 111-textured TiN diffusion barriers: Enhanced lifetime through microstructure control during layer growth

J.-S. Chun, I. Petrov, and J. E. Greene Citation: J. Appl. Phys. 86, 3633 (1999); doi: 10.1063/1.371271 View online: http://dx.doi.org/10.1063/1.371271 View Table of Contents: http://jap.aip.org/resource/1/JAPIAU/v86/i7 Published by the American Institute of Physics.

Related Articles
Evaluation and modeling of lanthanum diffusion in TiN/La2O3/HfSiON/SiO2/Si high-k stacks Appl. Phys. Lett. 101, 182901 (2012) Influence of catalytic gold and silver metal nanoparticles on structural, optical, and vibrational properties of silicon nanowires synthesized by metal-assisted chemical etching J. Appl. Phys. 112, 073509 (2012) Chemical and structural investigations of the interactions of Cu with MnSiO3 diffusion barrier layers J. Appl. Phys. 112, 064507 (2012) Thermal tweezers for nano-manipulation and trapping of interacting atoms or nanoparticles on crystalline surfaces J. Chem. Phys. 137, 114701 (2012) Thermal conductance at atomically clean and disordered silicon/aluminum interfaces: A molecular dynamics simulation study J. Appl. Phys. 112, 054305 (2012)

Additional information on J. Appl. Phys.


Journal Homepage: http://jap.aip.org/ Journal Information: http://jap.aip.org/about/about_the_journal Top downloads: http://jap.aip.org/features/most_downloaded Information for Authors: http://jap.aip.org/authors

Downloaded 18 Dec 2012 to 129.108.36.70. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

JOURNAL OF APPLIED PHYSICS

VOLUME 86, NUMBER 7

1 OCTOBER 1999

Dense fully 111-textured TiN diffusion barriers: Enhanced lifetime through microstructure control during layer growth
J.-S. Chun, I. Petrov, and J. E. Greenea)
Materials Science Department, Materials Research Laboratory, and Coordinated Science Laboratory, University of Illinois, 1101 West Springeld Avenue, Urbana, Illinois 61801

Received 16 April 1999; accepted for publication 24 June 1999 Low-temperature deposition of TiN by reactive evaporation or sputter deposition onto amorphous substrates leads to highly underdense layers which develop mixed 111/002 orientations through competitive growth. In contrast, we demonstrate here the growth of low-temperature 450 C fully dense polycrystalline TiN layers with complete 111 texture. This was achieved by reactive magnetron sputter deposition using a combination of: 1 highly oriented 25-nm-thick 0002 Ti underlayers to provide 111 TiN orientation through texture inheritance local epitaxy and 2 high ux (J N /J Ti 14), low-energy (E N 20 eV), N2 ion irradiation in a magnetically unbalanced 2 2 mode to provide enhanced adatom diffusion leading to densication during TiN deposition. The Ti underlayers were also grown in a magnetically unbalanced mode, in this case with an incident Ar /Ti ux ratio of 2 and E Ar 11 eV. All TiN lms were slightly overstoichiometric with a N/Ti ratio of 1.02 0.03. In order to assess the diffusion-barrier properties of dense 111-textured TiN, Al overlayers were deposited without breaking vacuum at 100 C. Al/TiN bilayers were then annealed at a constant ramp rate of 3 C s 1 to 650 C s 1 and the interfacial reaction between Al and TiN was monitored by in situ synchrotron x-ray diffraction measurements. As a reference point, we nd that interfacial Al3Ti formation is observed at 450 C in Al/TiN bilayers in which the TiN layer is deposited directly on SiO2 in a conventional magnetically balanced mode and, hence, is underdense with a mixed 111/002 orientation. However, the onset temperature for interfacial reaction was increased to 610 C in bilayers with fully dense TiN exhibiting complete 111 preferred orientation. 1999 American Institute of Physics. S0021-8979 99 04119-5

I. INTRODUCTION

Al-based interconnects with TiN diffusion barriers are widely used in multilevel metallization architectures incorporated in submicron ultralarge-scale integrated ULSI devices. As the dimensions of integrated circuits continue to decrease while their packing density increases, interconnection reliability issues such as electromigration resistance and diffusion barrier lifetime become increasingly important. It is well known15 that electromigration lifetimes are enhanced in Al metal lines having a large average grain size, a small grain-size distribution, and strong 111 texture. Shibata et al.6 found that the dominant factor for Al linewidths below 0.5 m is the degree of 111 preferred orientation. Texture control can also be important for controlling the stability of heterointerfaces during high temperature processing. Both the rate and extent of reaction at metal/diffusion barrier interfaces have been shown to depend upon lm texture.7,8 In all previous studies718 of polycrystalline Al/ TiN interfacial reactions, there were no attempts made to control TiN lm texture or, except for Refs. 7 and 9, porosity. In fact, the TiN layers are often uncharacterized. Under typical industrial deposition conditions, sputter-deposited TiN is underdense with a mixed 111/002 texture.4,7 Al/TiN interfacial reactions, the formation of Al3Ti, and barrier faila

ure occur during isothermal annealing for 30 min at temperatures below 400 C.19 Air-exposure leading to oxide formation prior to Al deposition has been shown to increase the thermal stability of the barrier layer.19 However, the results are not reproducible. Moreover, air-exposure reduces the degree of Al 111 texture and increases the contact resistance. Greene et al.20,21 has shown that the preferred orientation of polycrystalline TiN lms grown by ultrahigh-vacuum UHV reactive-magnetron sputter deposition on amorphous SiO2 at 350 C in pure N2 discharges can be controllably varied from essentially complete 111 to purely 002 by varying the incident ion/metal ux ratio J i /J Ti from 1 to 5 while the ion energy E N is maintained constant at 20 eV. The incident ions are predominantly N2 96.3% ,22 corresponding to 10 eV per incident N, well below the lattice displacement threshold. During low-temperature deposition in the absence of signicant ion irradiation (J i /J Ti 1), TiN initially exhibits a mixed texturepredominantly 111, 002, and 022and then slowly evolves to a strong 111 texture containing a network of both inter- and intracolumnar porosity due to the combination of low adatom mobilities and competitive columnar growth. Increasing J i /J Ti 5 results in enhanced adatom mobilities giving rise to a dense microstructure. In addition, the low-surface-energy 002 texture now dominates even at submonolayer coverage and there is
3633 1999 American Institute of Physics
2

Electronic mail: jegreene@uiuc.edu

0021-8979/99/86(7)/3633/9/$15.00

Downloaded 18 Dec 2012 to 129.108.36.70. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

3634

J. Appl. Phys., Vol. 86, No. 7, 1 October 1999

Chun, Petrov, and Greene

no longer any indication of competitive growth as 002 columns grow essentially straight up.20,21 Obtaining a dense TiN layer with complete 111 texture at low growth temperatures, as required for diffusion barrier and many hard-coating applications, is a challenging growth kinetics problem. Moving toward higher growth temperatures favors the development of the 002 texture,23 rather than 111, since 002 is the low-energy surface for TiN.24 In addition, the use of ion irradiation to promote densication of 111-textured TiN leads to either 002 preferred orientation as noted above for low ion energies with high ion-to-neutral ratios20,21 or very high in-plane compressive stress and mixed texture with higher energies and low ion-to-neutral ratios.23 In this article, we show, for the rst time, that fully dense completely 111-textured TiN layers can be achieved at low growth temperatures using a combination of 1 highly oriented 25-nm-thick 0002 Ti underlayers to provide 111 orientation through texture inheritance local epitaxy and 2 high-ux (J N /J Ti 14), low-energy (E N 20 eV), N2 ion 2 2 irradiation to provide enhanced adatom diffusion leading to densication. Both the Ti and TiN layers were grown in a magnetically unbalanced mode. While interfacial Al3Ti formation was observed at 450 C in Al/TiN bilayers with underdense TiN having a mixed texture, the onset temperature was increased to 610 C for bilayers with fully dense 111 TiN. The Al overlayers are dense in both cases, but exhibit much stronger 111 texture i.e., texture inheritance on dense 111 TiN.
II. EXPERIMENTAL PROCEDURE

All lms were grown in a load-locked multichamber UHV stainless-steel dc magnetron sputter deposition system which has been described in detail elsewhere.25 The pressure in the sample introduction chamber was reduced to less than 5 10 8 Torr 7 10 6 Pa) prior to initiating substrate exchange into the deposition chamber which has a base pressure of 5 10 10 Torr 7 10 8 Pa). For the present experiments, an additional magnetron-sputtering source with separate water cooling lines and shutter was added in the viewport ange opposing the original source. The Ti target 99.999% pure was operated in a magnetically unbalanced mode for the growth of both Ti and TiN lms while the Al target 99.999% pure was sputtered in a magnetically balanced mode. Target-to-substrate separations were 6.5 cm for TiN and 10 cm for Al deposition. A pair of external Helmholtz coils with Fe pole pieces were utilized to create a uniform axial magnetic eld B ext in the region between the target and the substrate. The positive and negative signs refer to geometries in which B ext aids and opposes, respectively, the eld of the outer pole of the magnetron. B ext has a strong effect on the ion ux incident at the substrate, with only minor effects on the target atom ux.25 The substrates used in these experiments were thermally oxidized 1 1 cm2 Si 001 wafers with an SiO2 thickness of 0.6 m. They were cleaned with successive rinses in ultrasonic baths of trichloroethane, acetone, ethanol, and deion-

ized water and blown dry with dry N2 . The wafers were mounted on resistively heated Ta platens using Mo clips and inserted into the sample introduction chamber for transport to the growth chamber where they were thermally degassed at 800 C for 1 h. The substrate temperature T s was then adjusted to 450 C as the Ti target was sputter etched for 5 min while shielding the substrate and Al target. Reported T s values, which were calibrated using a 0.25-mm-diam chromelalumel thermocouple bonded to the surface of a sacricial substrate using Ag paste, include the contribution from plasma heating. Plasma characteristics in the vicinity of the substrate were determined as a function of B ext from electrical measurements, following the procedures described in Refs. 25 and 26, obtained using both disk-shaped and cylindrical probes. Most of the ions incident at the substrate experience the full substrate sheath potential V s (V b V p ), where V p is the plasma potential and V b is substrate bias, since the mean-free path for charge exchange collisions,27 8 mm, is more than an order of magnitude larger than the sheath width,28 estimated from the ChildLangmuir equation. In the present experiments, V b is equal to the oating potential V f and the ion energy E i at the substrate is e( V f V p ). Two sets of TiN layers were grown with T s 450 C at the relatively high sputtering pressure of 20 mTorr N 2 (99.999%), chosen to suppress kinetic energy transfer to the growing lm by particles backscattered from the target. In sample set i, TiN layers, 160 nm thick, were deposited on SiO2 with J N /J Ti 1 and E N 20 eV (V p 3 V and V f 2 2 17 V). The ion current density at the substrate, corresponding to an ion ux j N 0.12 mA cm 2 2 14 2 1 J N 7.5 10 cm s ), was obtained using an external 2 axial magnetic eld B ext of 20 G. These conditions resulted in a deposition rate R of 6.0 nm min 1. The second set of TiN layers, series ii, were grown on 25-nm-thick Ti underlayers deposited on SiO2 at T s 80 C in 20 mTorr Ar 99.999% with an incident Ar /Ti ux ratio of 2 (J Ar 6.1 1015 cm 2 s 1 obtained with B ext 200 G), and 12.5 V and V f an Ar ion energy E Ar 11 eV (V p 23.5 V). The TiN overlayers, 140 nm thick, were then deposited without breaking vacuum using J N /J Ti 14 2 (J N 1.6 1016 cm 2 s 1 achieved with B ext 150 G) and 2 11 V and V f 34 V). The deposition E N 23 eV (V p 2 rates R of Ti and TiN were 30 and 5.4 nm min 1, respectively. Calculations carried out using the TRIM90 computer code,29 combined with Monte Carlo gas transport simulations based upon energy-dependent scattering cross sections,30 show that for the TiN growth conditions described above, a majority of both the sputter-ejected species and the energetic neutrals backscattered from the target are thermalized during gas-phase transport and contribute little to the total kinetic energy transfer to the substrate and growing lm. The ion ux to the substrate is due, as determined by double-modulated mass spectroscopy measurements,22 primarily to N2 (96.3%) with the remainder being N (3.5%) and Ti (0.2%).

Downloaded 18 Dec 2012 to 129.108.36.70. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

J. Appl. Phys., Vol. 86, No. 7, 1 October 1999

Chun, Petrov, and Greene

3635

Immediately following TiN deposition, for both series i and ii, T s was decreased to 100 C and 160-nm-thick Al overlayers were grown in 5 mTorr Ar discharges without breaking vacuum. The Al deposition rate was 71.1 nm min 1 with an incident Ar /Al ux ratio of 0.8 and E Ar 10 eV. The microstructure and microchemistry of as-deposited and annealed samples were determined using a combination of Rutherford backscattering spectroscopy RBS , x-ray diffraction XRD , and plan-view and cross-sectional transmission electron microscopy TEM and XTEM . The RBS probe beam consisted of 2 MeV He ions incident at an angle of 22.5 relative to the sample surface normal with the detector set at a 150 scattering angle. Backscattered spectra were analyzed using the RUMP simulation program.31 The uncertainty in determining the N/Ti ratio was less than 0.03. TiN exhibits a wide single-phase eld with a N/Ti ratio extending from 0.6 to 1.2.32 In the present experiments, all TiN lms, irrespective of J i /J Ti , were found to be slightly overstoichiometric with N/Ti 1.02 0.03, as expected for lowtemperature growth in pure N2 discharges. XRD 2 and -rocking curves were obtained using Cu K radiation in a system equipped with a double-crystal spectrometer to provide a resolution of 0.01 2 in the powder diffraction mode 45 kV, 20 mA, 1 divergent slit . The TEM and XTEM analyses were carried out using Philips CM12 and Hitachi 9000 electron microscopes operated at 120 and 300 kV, respectively. Sample preparation for TEM/ XTEM examination followed the procedure outlined in Ref. 21. The surface morphologies of as-deposited TiN lms were investigated using atomic force microscopy AFM . The measurements were carried out in a Digital Nanoscope II microscope operated in air in the contact mode with oxidesharpened Si3N4 tips having radii of 540 nm. In situ x-ray diffraction XRD measurements during sample annealing were performed at the Brookhaven National Synchrotron Light Source, beamline X-20C. Asdeposited samples were loaded in an annealing chamber aligned to the beamline and equipped with an x-ray transparent Be window. The chamber was evacuated to 5 10 6 Torr, backlled with puried He, re-evacuated, and then backlled with 1 atm of He. As-deposited samples were annealed using linear temperature ramps from 100 to 650 C at 3 C s 1. A high-intensity monochromator employing bandpass multilayer lters provided an energy resolution of 1.5% at 6.9 keV 0.1797 nm with a typical intensity at the sample of 3 1012 photons s 1 . The incident x rays illuminated a sample area 1 5 mm2. Diffraction scans were collected using a position-sensitive detector with time resolution of ms.
III. EXPERIMENTAL RESULTS AND DISCUSSION

FIG. 1. a Bright-eld XTEM micrograph and b corresponding -2 XRD scan from an as-deposited series-i TiN layer grown on SiO2 at 450 C with J N /J Ti 1 and E N 20 eV.
2 2

Figure 1 a is an XTEM micrograph from a TiN layer grown directly on SiO2 with J N /J Ti 1 and E N 20 eV, 2 2 conditions corresponding to TiN sample series i. The micrograph is typical of TiN lms grown under weak ion irradiation conditions at T s 500 C (T s /T m 0.2). The layer is underdense due to limited adatom mobilities leading to

atomic shadowing which, in turn, results in a columnar microstructure with both inter- and intracolumnar voids.20,21 Careful analyses, using nanodiffraction and highresolution XTEM, of the region near the lm/substrate interface show that the layer initially nucleates with islands leading to grains which are predominantly broad-based 002 and narrow 111. However, the 111 grains eventually overgrow the 002 via a competitive growth mechanism. That is, diffusion on the low-energy 002 surface24in which there is only one adatom backbondis fast, leading to islands spreading out relatively rapidly and accounting for the broad-based 002 columns, compared to the 111 surface where each adatom has three backbonds. Thus, the steady-state adatom density on 002 islands is low as adatoms quickly move to the edges, while 111 grains grow upwards rapidly and intercept ever larger fractions of the incoming ux. While the majority of the 002 grains are overgrown early in the growth process, some survive even in lms with thicknesses up to 1 m.21 The open intercolumnar boundaries observed in lowtemperature 111-oriented transition-metal nitride layers e.g., TiN,20,21 Ti1 x Alx N, 33 and ScN34 is simply a consequence of the effect of the atomic shadowing process during texture evolution via competitive growth combined with the distribution of particle incidence angles. Figure 1 a also shows that the sample surface is rough and faceted, as commonly observed for lms with voided column boundaries. The layer thickness determined from the micrograph is 160 nm, in good agreement with the deposition rate calibration. Figure 1 b is a -2 XRD scan from the series i layer shown in Fig. 1 a . The pattern contains an intense 111 peak and a small 002 peak, both at positions corresponding to B1-NaCl-structure TiN. The integrated 111 to 002 peak intensity ratio I 111 /I 002 is 8.8, a factor of 14 higher than that

Downloaded 18 Dec 2012 to 129.108.36.70. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

3636

J. Appl. Phys., Vol. 86, No. 7, 1 October 1999

Chun, Petrov, and Greene

FIG. 2. Schematic view of the atomic arrangement at the interface between Ti 0002 and TiN 111 layers.

obtained from randomly oriented TiN powder. Thus, the dominant texture at this lm thickness is 111 in agreement with the XTEM results discussed above. The 111 interplanar spacing d 111 is 0.244 nm which is close to the reported bulk value of 0.2449 nm and indicates that the layer is nearly fully relaxed. The full width at half maximum FWHM intensity of the 111 -rocking curve, not shown, is 23.1. The is consistent with kinetically limited comlarge value of petitive texture evolution. Low-energy, high-ux, N2 irradiation during growth densies the TiN microstructure leaving no residual ioninduced defects detectable by XRD and TEM.20,21 However, the layers also exhibit complete 002 preferred orientation. Thus, in an attempt to obtain TiN layers which are both dense and have complete 111 texture, we have carried out experiments combining intense N2 ion irradiation with the use of a dense 0002 oriented hcp Ti underlayer as a crystallographic template sample series ii . The hcp Ti basal plane and the fcc TiN 111 plane the NaCl crystal structure is fcc with a two-atom basis have the same two-dimensional symmetry. Moreover, Ti underlayers are often used in multilevel interconnections to provide enhanced adhesion between the diffusion barrier and SiO2 35 and lower contact resistance with Si.36 Figure 2 is a schematic view of the atomic arrangement

FIG. 3. a -2 XRD scan of an as-deposited Ti layer grown on SiO2 at 80 C with an incident Ar /Ti ux ratio of 2 and E Ar 11 eV. b An 0002 -rocking curve and c a bright-eld plan-view TEM micrograph from the same sample.

at the interface between 0002 Ti and 111 TiN layers. The mist along equivalent close-packed directions in TiN 111 and Ti 0002 is approximately 1.6% with x TiN 0.29995 nm and x Ti 0.29503 nm. 111 is a polar direction in TiN with alternate 111 planes consisting of all Ti and all N atoms. Preferred orientation in metal lms, as opposed to transition-metal nitride layers, typically corresponds to the low-energy surface even at low growth temperatures37 due to

Downloaded 18 Dec 2012 to 129.108.36.70. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

J. Appl. Phys., Vol. 86, No. 7, 1 October 1999

Chun, Petrov, and Greene

3637

FIG. 4. a Bright-eld XTEM micrograph from an as-deposited series-ii TiN/Ti bilayer structure. The TiN layer was grown at 450 C with J N /J Ti 14 and E N 20 eV. b A -2 XRD scan and c a 111 2 2 -rocking curve from the same sample.

FIG. 5. High-resolution XTEM micrograph of the TiN/Ti interface from the sample corresponding to Fig. 4.

much higher adatom mobilities on metal surfaces. In the present experiments, Ti underlayers grown on SiO2 at 80 C (T s /T m 18%) were found to be 0002. However, to enhance the degree of orientation, we employed low-energy Ar irradiation through the use of magnetically unbalanced sputter deposition to provide, at the growth surface, an ion-to-metal ux ratio J Ar /J Ti 2 with E Ar 11 eV. Figure 3 a is a -2 XRD scan of an as-deposited Ti layer. The only peaks observed are 000l (l 2,4,...) whereas for Ti powder patterns the strongest peak is 1011 2 40.17 with an intensity ratio I 101 1 /I 0002 3.4. We obtain an 0002 interplanar spacing d 0002 of 0.2342 nm in agreement with the reported bulk value, 0.2341.38 The -rocking curve

shown in Fig. 3 b has a FWHM value 3.3. A brighteld plan-view TEM micrograph of the same sample is shown in Fig. 3 c . The absence of Moire fringes is indicative of a columnar structure. However, the individual grains appear fully dense with no evidence of either inter- or intragrain porosity. The average Ti grain size obtained from planview TEM images using image analysis software39 is 26 21 nm. Immediately following growth of the oriented Ti underlayers, TiN was deposited at P N2 20 mTorr and T s 450 C with J N /J Ti 14 and E N 20 eV. Figure 4 a is 2 2 a typical bright-eld XTEM micrograph from an asdeposited TiN/Ti bilayer structure. TiN and Ti layer thicknesses determined from the micrographs are 160 and 5 nm,

Downloaded 18 Dec 2012 to 129.108.36.70. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

3638

J. Appl. Phys., Vol. 86, No. 7, 1 October 1999

Chun, Petrov, and Greene

FIG. 7. AFM images of the samples shown in a Fig. 6 a and b Fig. 6 b .

FIG. 6. Bright-eld plan-view TEM micrographs of a series-i and b series-ii TiN/Ti bilayers.

respectively. The decrease in the Ti thickness from 25 to 5 nm indicates that 80% of the underlayer has reacted with the incident N2 ux, due to the high N2 /Ti ratio used in these experiments, to form TiN during the initial stages of nitride deposition. There are two important points to note in this micrograph. The rst is that this TiN layer, in contrast to the sample in Fig. 1 a , is dense with no detectable inter- or intracolumnar porosity. The lm appears fully dense even in high-resolution micrographs. The second point is that Fig. 4 a reveals no indication of competitive 111/002 grain growth. In fact, all TiN columns are 111 and grow commen-

surately with the underlying transformed TiN grains as will be shown below using high-resolution XTEM HR-XTEM . Figure 4 b is a -2 XRD scan from the type-ii bilayer sample corresponding to Fig. 4 a . Consistent with the XTEM results, and in contrast to XRD scans from the type-i TiN layer grown with no intentional ion irradiation and without the oriented Ti underlayer, only the 111 peak is obtained with no evidence of a residual 002 feature. The d 111 lattice spacing is 0.2442 nm in reasonable agreement with the bulk value indicating that the layer is relaxed. XRD pole gures not shown obtained with 2 set to the 111 Bragg angle for TiN conrm that the lm texture is azimuthally symmetric i.e., it exhibits ber texture . No other diffraction peaks were detected in either pole gures or -2 scans. The lack of Ti reections is due to the fact that most of the underlayer, as shown in the XTEM micrograph in Fig. 4 a , was transformed to TiN. The 111 -rocking curve in Fig. 4 c is extremely narrow, 1.9, compared to the FWHM obtained from the type-i layer, 23.1. Thus, the mosaicity is greatly reduced giving rise to a factor of 16 increase in the in-plane coherence length. A HR-XTEM bright-eld image of the TiN/Ti interface in Fig. 4 a is presented in Fig. 5. It shows clearly, together with high and lower-magnication images obtained from different regions of the same sample, that cubic 111 TiN grains grow with a local epitaxial relationship to hexagonal 0002 Ti grains: TiN111 Ti0002 . The TiN/Ti interface is locally coherent with TiN and Ti exhibiting equal in-plane grain sizes. The micrograph also shows that the Ti/SiO2 interface remains abrupt. Figures 6 a and 6 b are plan-view TEM micrographs showing a direct comparison of the microstructure of type-i

Downloaded 18 Dec 2012 to 129.108.36.70. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

J. Appl. Phys., Vol. 86, No. 7, 1 October 1999

Chun, Petrov, and Greene

3639

FIG. 8. Bright-eld XTEM micrographs from a series-i Al/TiN and b series-ii Al/TiN bilayers.

TiN lms grown by conventional reactive magnetron sputtering and type-ii TiN layers grown under intense lowenergy N2 irradiation and with an oriented Ti underlayer. Both TiN lms were deposited at T s 450 C with E N 20 eV. However, the series-i lm in Fig. 6 a grown 2 with J N /J Ti 1 exhibits an underdense structure consistent 2 with the XTEM micrograph in Fig. 1 a with a mixed texture which is predominantly 111. The intercolumn voids have widths of up to 2 nm. In contrast, the series-ii layer in Fig. 6 b is fully dense with a complete 111 texture. Average grain sizes d measured near the top of the underdense and dense TiN lms are 20 14 and 43 30 nm, respectively. Since the average grain size of the as-deposited Ti underlayer is 26 21 nm, d in the 111 TiN overlayer grains increases with layer thickness. This can be also seen in the XTEM micrograph in Fig. 4 a . Figures 7 a and 7 b are AFM images of the TiN lms corresponding to Figs. 6 a and 6 b . The 160-nm-thick underdense series-i TiN layer in Fig. 7 a is extremely rough with a root-mean-square surface width w 8.5 nm and an in-plane surface coherence length 78 nm. The same thickness dense series-ii 111-oriented TiN layer Fig. 7 b

FIG. 9. -2 XRD scans from as-deposited a series i and b series-ii Al/TiN bilayers. c Al 111 -rocking curve from the sample corresponding to b .

grown on the 0002 Ti template is much smoother with w 3.3 nm and 49 nm. In order to examine the effects of texture, microstructure, and surface roughness on the diffusion barrier properties of TiN, 160-nm-thick Al overlayers grown with J Ar /J Al 0.8 and E Ar 10 eV were deposited at P Ar 5 mTorr and T s 100 C on both series-i and series-ii TiN layers without breaking vacuum. Figures 8 a and 8 b are typical XTEM micrographs from Al/TiN diffusionbarrier test structures. Both Al overlayers were fully dense with average grain sizes, obtained from plan-view images, of 210 nm. They exhibited quite different degrees of preferred orientation, however, and the surface of the Al lm grown on the type-i TiN layer is much rougher with an in-

Downloaded 18 Dec 2012 to 129.108.36.70. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

3640

J. Appl. Phys., Vol. 86, No. 7, 1 October 1999

Chun, Petrov, and Greene

FIG. 10. XRD diffracted intensity contour maps plotted as a function of temperature during thermal ramping of a series i and b series-ii Al/TiN bilayers from 100 to 650 C at 3 C s 1.

plane wavelength of the order of the grain size. -2 XRD scans from the Al/TiN bilayers in Figs. 8 a and 8 b are shown in Figs. 9 a and 9 b . The patterns demonstrate that the Al lm grown on TiN with mixed 111/002 texture also exhibits a mixed orientation Fig. 9 a while the Al layer grown on the TiN lm with a complete 111 preferred orientation also has a pure 111 texture note the scale change by a factor of 102 in the inset of Fig. 9 b . Both Al layers are relaxed with a 111 interplanar spacing equal to the bulk value of 0.2338 nm.40 In addition, the 111 Al and TiN integrated peak intensities obtained from bilayer ii are two orders of magnitude higher than those of bilayer i at equal layer thicknesses. for Al layers grown on both series i and ii TiN are 13.5 and 0.82, respectively. The above results show that the stronger 111 orientation in the TiN lm in bilayer-ii results in a corresponding enhancement in the 111 texture of the Al overlayer. This occurs through texture inheritance as metal islands nucleate and, at least initially, grow with a local epitaxial relationship with underlying TiN grains. However, the nal Al grain size is more than ve times larger than that of TiN indicating the presence of recrystallization which is known to occur at temperatures T r as low as T r /T m 0.24 in high purity Al.41 While this further reduces grain misorientation in both Al overlayers, the mosaicity of the layers grown on purely 111oriented TiN is still more than an order of magnitude lower, as indicated by the measurements, than for Al grown on mixed 111/002-oriented TiN. Figures 10 a and 10 b are typical contour plots of the logarithm of the diffracted synchrotron x-ray intensity as a function of the diffraction angle 2 and temperature T a during thermal ramping of underdense series-i TiN and dense 111 series-ii TiN samples to 650 C at a rate of 3 C s 1. The intensity is indicated by gray scale. White corresponds to an

intensity of 1.9 104 in Fig. 10 a and 9.2 104 in Fig. 10 b while black is 102 counts s 1 . The angular range of the position sensitive detector was chosen based upon ex situ XRD results following isothermal anneals showing, in addition to the strong TiN and Al 111 peaks, the emergence of the 112 tetragonal-structure Al3Ti peak also observed by TEM and XTEM at 39.132 . 112 was the only XRD peak observed, indicating that the product phase is highly textured. Grains of the intermetallic compound Al3Ti, which has the DO22 structure with lattice constants a 0 b 0 0.385 37 nm and c 0.858 39 nm, 42 are aligned with their 112 plane perpendicular to the Al lm growth direction. 112 is the closest-packed Al3Ti plane and exhibits an inplane mist with Al 111 of only 1%. Figure 10 a shows that for type-i samples, the intensity of the 111 Al peak at 45.2 0.1797 nm 2 decreases with increasing annealing temperature T a and is completely lost at T a 620 C. This is well below the Al melting point 660 C and indicates that the Al layer has been completely consumed. Al3Ti starts to form at 450 C and reaches its maximum intensity at the same temperature at which the Al intensity disappears. In contrast, Al3Ti formation in type-ii samples is suppressed up to 610 C and signicant unreacted Al still remains even at 650 C as shown in Fig. 10 b . We attribute the enhanced interfacial stability of series ii, compared to series i, Al/TiN samples to the densication of TiN grain boundaries with high-ux (J N /J Ti 14), low2 energy (E N 20 eV), N2 ion irradiation during TiN depo2 sition, the larger TiN grain size by a factor of two , and the atter surface. It has been reported previously8,9 that TiN barrier failure is initiated at the Al/TiN interface by the formation of Al3Ti in the Al layer. In type-i samples, massive diffusion of Al into the open TiN inter- and intracolumnar voids is expected to occur rapidly at relatively low annealing temperatures leading to an enormous increase in the actual area of the Al/TiN interface.

IV. CONCLUSION

We have demonstrated, for the rst time, the lowtemperature growth of fully dense polycrystalline TiN with complete 111 preferred orientation. This was achieved using a combination of highly oriented thin 0002 Ti underlayers to provide orientation through texture inheritance local epitaxy and high-ux, low-energy, N2 ion irradiation 20 eV) to provide enhanced adatom (J N /J Ti 14, E N 2 2 diffusion leading to densication. The preferred orientation of Al overlayers grown on dense fully 111-oriented TiN was also greatly enhanced due to texture inheritance, thus reducing the FWHM of the 111 -rocking curve from 13.5 for Al layers grown on TiN deposited by conventional reactive magnetron sputter deposition to 0.82. This corresponds to an enhancement in the average mosaic coherence length by more than a factor of 15. Finally, we show that the thermal stability of Al/TiN bilayers to interfacial breakdown is increased from 450 to 610 C through the use of fully dense TiN layers with complete 111 texture.

Downloaded 18 Dec 2012 to 129.108.36.70. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

J. Appl. Phys., Vol. 86, No. 7, 1 October 1999

Chun, Petrov, and Greene


20

3641

ACKNOWLEDGMENTS

The authors gratefully acknowledge the nancial support of the NSF/DARPA VIP Program and the Department of Energy under Contract No. DEAC0276ER01198 during the course of this research. We also appreciate the use of the Brookhaven National Synchrotron Light Source beamline X-20C together with the expert technical assistance of Dr. C. Lavoie and Dr. C. Cabral, Jr. of the IBM T.J. Watson Research Center, and the facilities in the Center for Microanalysis, which is partially supported by DOE, at the University of Illinois.
M. Kageyama, K. Hashimoto, and H. Onoda, in Proceedings of the 29th International Reliability Physics Symposium, 1991 unpublished , p. 97. 2 S. Vaidya and A. K. Sinha, Thin Solid Films 75, 253 1981 . 3 M. Tsukada and S. Ohfuji, J. Vac. Sci. Technol. B 11, 326 1983 . 4 K. Hashimoto and H. Onoda, Appl. Phys. Lett. 10, 120 1989 . 5 H. Onoda, M. Kageyama, and K. Hashimoto, J. Appl. Phys. 32, 4479 1993 . 6 H. Shibata, M. Murota, and K. Hashimoto, Jpn. J. Appl. Phys., Part 1 32, 4479 1993 . 7 Q. Z. Hong, S. P. Jeng, R. H. Havemann, H. L. Tsai, and H. Y. Liu, J. Appl. Phys. 78, 7419 1995 . 8 J.-S. Chun, J. R. A. Carlsson, D. B. Bergstrom, I. Petrov, J. E. Greene, C. Lavoie, C. Cabral, Jr., and L. Hultman unpublished . 9 L. Hultman, S. Benhenda, G. Radnoczi, J.-E. Sundgren, J. E. Greene, and I. Petrov, Thin Solid Films 215, 152 1992 . 10 M. Wittmer, J. Appl. Phys. 53, 1007 1982 . 11 H. Norstrom, T. Donchev, M. Ostling, and C. S. Petersson, Phys. Scr. 28, 33 1983 . 12 C. Y. Ting, J. Vac. Sci. Technol. 21, 14 1982 . 13 N. Kumar, K. Pourrezaei, B. Lee, and E. C. Douglas, Thin Solid Films 164, 417 1998 . 14 H. P. Kattelus, J. L. Tandon, C. Sala, and M. A. Nicolet, J. Vac. Sci. Technol. A 4, 1850 1986 . 15 R. Beyer, R. Sinclair, and M. E. Thomas, J. Vac. Sci. Technol. B 2, 781 1984 . 16 R. C. Elwanger and J. M. Towner, Thin Solid Films 161, 289 1988 . 17 K.-H. Bather and H. Schreibe, Thin Solid Films 200, 93 1991 . 18 A. Kohlhase, M. Mandle, and W. Pamler, J. Appl. Phys. 65, 2464 1989 . 19 W. Sinke, G. P. A. Frijlink, and F. W. Saris, Appl. Phys. Lett. 47, 471 1985 .
1

J. E. Greene, J.-E. Sundgren, L. Hultman, I. Petrov, and D. B. Bergstrom, Appl. Phys. Lett. 67, 2928 1996 . 21 L. Hultman, J.-E. Sundgren, J. E. Greene, D. B. Bergstrom, and I. Petrov, J. Appl. Phys. 78, 5395 1995 . 22 I. Petrov, A. M. Myers, J. E. Greene, and J. Abelson, J. Vac. Sci. Technol. A 12, 2846 1994 . 23 I. Petrov, L. Hultman, U. Helmersson, J.-E. Sundgren, and J. E. Greene, Thin Solid Films 169, 299 1989 . 24 L. Hultman, J.-E. Sundgren, and J. E. Greene, J. Appl. Phys. 66, 536 1989 . 25 I. Petrov, F. Adibi, J. E. Greene, W. D. Sproul, and W.-D. Munz, J. Vac. Sci. Technol. A 10, 3283 1992 . 26 J. A. Thornton, J. Vac. Sci. Technol. 15, 188 1978 . 27 R. F. Stebbing, B. R. Turner, and A. C. H. Smith, J. Chem. Phys. 38, 2277 1963 . 28 B. Chapman, Glow Discharge Processes Wiley, New York, 1980 , p. 108. 29 J. A. Biersack and L. G. Haggmark, Nucl. Instrum. Methods 174, 257 1980 . 30 A. V. Phelps, J. Phys. Chem. Ref. Data 20, 557 1991 . 31 R. L. Doolittle, Nucl. Instrum. Methods Phys. Res. B 15, 344 1985 . 32 J.-E. Sundgren, B.-O. Johansson, A. Rockett, S. A. Barnett, and J. E. Greene, in Physics and Chemistry of Protective Coatings, edited by J. E. Greene, W. D. Sproul, and J. A. Thornton American Institute of Physics, New York, 1986 , Ser. 149, p. 95. 33 F. Adibi, I. Petrov, J. E. Greene, L. Hultman, and J.-E. Sundgren, J. Appl. Phys. 73, 8580 1993 . 34 D. Gall, I. Petrov, L. D. Madsen, J.-E. Sundgren, and J. E. Greene, J. Vac. Sci. Technol. A 16, 2411 1998 . 35 C. Y. Ting and M. Wittmer, Thin Solid Films 96, 327 1982 . 36 R. W. Bower, Appl. Phys. Lett. 23, 99 1973 . 37 J. E. Greene, in Handbook of Crystal Growth, Volume 1: Fundamentals, edited by D. T. J. Hurle Elsevier, Amsterdam, 1993 , p. 639. 38 Inorganic Index to Powder Diffraction File Joint Committee on Powder Diffraction Standards, Pennsylvania, 1996 . 39 IMAGE software V. 1.55, from the National Institute of Health, obtainable from public program site with access address zippy.nimh.nih.gov under the directory/pub/nih-image. 40 Inorganic Index to Powder Diffraction File Joint Committee on Powder Diffraction Standards, Pennsylvania, 1996 . 41 E. C. W. Perryman, ASM Seminar, Creep and Recovery, 1957 American Society for Metals, Cleveland, Ohio, 1957 , p. 111. 42 Inorganic Index to Powder Diffraction File Joint Committee on Powder Diffraction Standards, Pennsylvania, 1996 .

Downloaded 18 Dec 2012 to 129.108.36.70. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

Anda mungkin juga menyukai