Hydrogen diffusion and stability in polycrystalline CVD undoped diamond

Hydrogen diffusion and stability in polycrystalline CVD undoped diamond

Diamond and Related Materials 10 Ž2001. 405᎐410 Hydrogen diffusion and stability in polycrystalline CVD undoped diamond D. Ballutaud a,U , F. Jomarda...

117KB Sizes 4 Downloads 209 Views

Diamond and Related Materials 10 Ž2001. 405᎐410

Hydrogen diffusion and stability in polycrystalline CVD undoped diamond D. Ballutaud a,U , F. Jomarda , B. Theys a , C. Mer b, D. Tromsonb, P. Bergonzo b a

CNRSr LPSC, 1 Place Aristide Briand, Meudon, France b CEAr LETI, Gif-sur-Y¨ ette, France

Abstract Diffusion profiles and effusion experiments performed on post-hydrogenated Ždeuterated. CVD diamond layers Žgrain size 2 and 0.2 ␮m. are reported in order to study the configurations and stability of hydrogen bonding in polycrystalline undoped CVD diamond. Deuterium is used as a tracer to improve the hydrogen detection limit. The diamond layers are first annealed at 1200⬚C in order to out-diffuse hydrogen present in the as-grown sample. Then the samples are exposed either to a radiofrequency plasma or a microwave plasma and the deuterium diffusion profiles are analyzed by secondary ion mass spectrometry. For r.f. and microwave plasma, the diffusion profiles are explained in term of trapping on plasma-induced defects near the surface andror on inter- and intra-granular defects. The mean free paths of deuterium and capture radius of traps are calculated by fitting the deuterium diffusion profiles and depend on the grain sizes. Some CVD diamond layers are deposited using a gas mixture ŽCH 4 q D 2 . and a deuterium concentration of 3 = 10 19 cmy3 , originating from the vector gas, is found in these as-grown samples. The stabilities of deuterium bonding in as-grown and post-deuterated samples are compared. 䊚 2001 Elsevier Science B.V. All rights reserved. Keywords: CVD polycrystalline diamond; Diffusion; Hydrogen; Thermal stability

1. Introduction Hydrogen is known to play an important role in the chemical vapor deposition ŽCVD. diamond growth processes w1x. Contrary to polycrystalline silicon where sp 2 hybridization does not exist, hydrogen is thought to maintain sp 3 hybridization of carbon atoms at the growth surface. The formation of thin-films of diamond by CVD from CH 4rH 2 mixture precursors containing more than 95% of molecular hydrogen has been widely demonstrated and a large number of applications are developed including optical and mechanical coatings and high performances electronic devices as detectors w2x. However, the diamond polycrystalline films are U

Corresponding author. Tel.: q33-1-45-07-51-96; fax: q33-1-4507-58-41. E-mail address: [email protected] ŽD. Ballutaud..

highly defective and present a high density of grain boundaries or dislocations and incorporated hydrogen in semiconductors is known to terminate dangling bonds and passivate both shallow and deep levels w3᎐5x. Furthermore, the hydrogen present at the diamond surface andror the sub-surface region induces a superficial highly p-type conductive layer w6x, the physical origin of which is not clearly understood. In this paper, we deal with new data on the diffusion and thermal stability properties of hydrogen in CVD polycrystalline diamond, using deuterium as tracer to improve the detection limit of secondary ion mass spectrometry ŽSIMS. and effusion measurements. Deuterium is introduced in the CVD diamond either during the layer deposition by replacing the hydrogen vector gas by deuterium, or by post-deuteration in a microwave or a radiofrequency Žr.f.. plasma. Preliminary deuterium effusion results on deuterated samples

0925-9635r01r$ - see front matter 䊚 2001 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 5 - 9 6 3 5 Ž 0 0 . 0 0 5 9 0 - 2

406

D. Ballutaud et al. r Diamond and Related Materials 10 (2001) 405᎐410

are also presented. The deuterium trapping at grain boundaries and the origin of the effusion peaks will be discussed. 2. Experimental The polycrystalline CVD diamond samples used throughout this study were grown by CVD using a 0.5% CH 4rH 2 mixture Ž950 W, 60 mbars, 780⬚C. on a monocrystalline silicon substrate w7x. In these conditions, it has been observed in our previous work w7x that the grain size is about a tenth of the layer thickness. Some experiments described in this paper were performed on two kinds of undoped polycrystalline diamond: Ža. 2-␮m grain size samples Žlayer thickness 20 ␮m.; and Žb. 0.2-␮m grain size samples Žlayer thickness 2 ␮m.. These diamond layers have been exposed either to a deuterium r.f. plasma Ž1 W cmy2 , 1 mbar, 500⬚C. or to a microwave deuterium plasma Ž950 W, 60 mbars, 800⬚C. after being submitted to an isothermal annealing at 1200⬚C during 2 h in ultra high vacuum Ž- 5 = 10y1 0 mbar. in order to outgas the hydrogen introduced into the layer during the deposition process. Some samples were deposited with deuterium instead of hydrogen as vector gas, on a polycrystalline diamond layer of 20-␮m thickness as substrate. The deuterium diffusion profiles were analyzed with a CAMECA IMS 4f secondary ion mass spectrometer with a Csq primary ion beam. Quantifications of deuterium and hydrogen were achieved by using undoped CVD diamond standards implanted with a known dose of deuterium or hydrogen. The effusion spectra of deuterium were measured by the ion current of a quadrupole mass spectrometer coupled to an evacuated quartz tube Ž10y9 ᎐10y10 mbar. which contained the deuterated samples. The desorption rates were recorded during the heating of the sample inside the quartz tube with a linear temperature ramp of 20 Krmin. During the effusion measurements the quartz tube was submitted to a constant pumping speed by adjustment of the ionic pump output valve. The low basic pressure allowed to detect a low rate of outgassed deuterium and to record a mass 4 ŽD 2 . spectrometer signal directly proportional to the effusion rate w8x. 3. Results and discussion 3.1. Hydrogen content in as-grown CVD polycrystalline diamond The hydrogen and deuterium concentrations have been measured by SIMS in a sample deposited with deuterium on a polycrystalline diamond layer of 20-␮m thickness as substrate with the conditions mentioned

Fig. 1. Deuterium concentration profile in a CVD polycrystalline layer Ž2-␮m grain size.. The arrow shows the limit between the substrate layer deposited with hydrogen and the upper layer deposited with deuterium ŽCH 4 0.5%, D 2 ..

above ŽFig. 1.. The arrow on the figure indicates the interface between the deeper layers deposited with hydrogen and the top layers deposited with deuterium. The deuterium accumulation at the interface may be due to the regulation of the gas flows. Some interdiffusion of hydrogen and deuterium occurs near the interface. The hydrogen concentration is under the SIMS detection limit Ž6 = 10 18 cmy3 . in the top deuterated layer, while the deuterium concentration is approximately 3 = 10 19 at. cmy3 . The hydrogen concentration in the substrate bulk is close to this value. It leads to conclude that hydrogen in the diamond layer would come mainly from the vector gas H 2 ŽD 2 ., or that the methane or adsorbed methyl, undergo deuterium exchange on a rapid timescale compared to the deposition kinetics. Assuming that all the hydrogen atoms are bonded on grain boundaries and that grains are comparable to cubes, this result leads to a deuterium density of 3 = 10 14 at. per cm2 of grain boundary Žto be compared with polycrystalline silicon, where it is found to be approximately 5 = 10 12 at. cmy2 . w5x. Considering this high surface concentration value, it may be supposed that deuterium is also bonded in disordered regions near grain boundaries and on intragrain defects. Higher hydrogen concentrations Ž) 3 = 10 20 cmy3 . have been found in diamond homoepitaxial layers in previous works w9x. These results would confirm the presence of intragrain hydrogen the nature of which is not yet explained, considering these high concentrations. Further experiments are being performed on monocrystalline homoepitaxial CVD diamond layers in order to clarify the hydrogen location w10x. 3.2. Hydrogen diffusion from a plasma Generally, because the fact that hydrogen can be

D. Ballutaud et al. r Diamond and Related Materials 10 (2001) 405᎐410

407

atomic, molecular or trapped, the hydrogen diffusion profile does not follow the erfc function: w Hx Ž x,t . s w H 0 x erfc xr2 D H t

ž

'

/

where wHx is the hydrogen concentration, D H the hydrogen diffusion coefficient, x the diffusion depth and t the diffusion time. The presence of traps for hydrogen will reduce the apparent diffusivity, leading to an effective coefficient w11x: Deff s D H Žw Hx r w H T x. ,

Ž1.

where wHxrwH T x is the fraction of free hydrogen. This fraction is constant and can be calculated as a function of the equilibrium constants of the reaction of trapping and detrapping. Assuming no dissociation of the hydrogen complex, the diffusion for the free hydrogen is fickian and has an erfc solution provided one uses Deff . In this case, it may be observed experimentally, an exponential fall off of the depth profile: w Hx s w H 0 x exp Ž y␣ x .

Ž2.

x being the diffusion depth and the slope ␣ of that exponential varying with the trapping concentration w11,12x: ␣ s 4 ␲ Rw nT x

'

Ž3.

where w n T x is the total initial trap concentration and R is the capture radius of this trap for hydrogen. Then d, the mean free path between trapping events is given by 1r␣ w12x. In the case of an unsaturable trap, if the steady state is assumed, the depth profile is only exponential w11x: w Hx s w H 0 x exp Ž y␣ x .

Ž4.

with 1rds ␣ s 4 ␲ R w n T x

'

In the case of trapping by ‘molecule’ formation Žwhatever may be the structure of this ‘molecule’ which may be a kind of pair. the concentration of hydrogen may be expressed in the steady-state regime by: w Hx s w H 0 x r Ž 1 q ␤ x . 2 with ␤ s ␲ ⭈ 2r3 ⭈ R w H 0 x

'

Ž5.

where wH 0 x is the surface hydrogen concentration, provided that there is no dissociation of the wH.Hx complex w11x.

Fig. 2. Deuterium concentration profile in a CVD polycrystalline layer Ž2-␮m grain size. after exposure to microwave plasma. The solid line represents the experimental profile, the dotted line represents a fit to an exponential decay and the dashed line is a fit to an erfc function.

3.3. Interpretation of the experimental results Fig. 2 shows the experimental deuterium concentration profiles obtained in 2-␮m grain size CVD polycrystalline diamond after an exposure to a microwave plasma Ž800⬚C, 1 h. Žsolid curve.. Scanning electron microscopy analysis leads to conclude that the grain size is a tenth of the diamond layer thickness. As the total deuterium profile depth is less than 350 nm, the grain size may be considered as constant within this limit. At depth of f 140 nm the profile exhibits a kink and at lower concentration the profile follows an exponential decay. The dotted line represents a least-square fit to an exponential decay with a characteristic length of ds 1r␣ s 95 nm. The concentration of traps is derived from the deuterium depth profiles, where the decay of the deuterium concentration changes from an erfc to an exponential decay with depth w12x. If it is assumed that this change occurs at the kink on the curve Žshown by an arrow on the curve in Fig. 2., the concentration of traps is found to be approximately 8 = 10 17 cmy3 . Then d is the mean free path between trapping events and is given by 1r␣, with: ␣ s 4 ␲ Rw nT x

'

Ž see Eq. Ž 3 .. .

From the observed value of ds 95 nm, the capture radius computes to R s 0.011 nm. Fitting the high-concentration region of the deuterium diffusion profile with an erfc function Ždotted line in Fig. 2. would yield an effective diffusion coefficient Deff s 4 = 10y1 4 cm2 sy1 . Fig. 3 shows the experimental deuterium concentration profiles obtained in 2-␮m grain size CVD polycrystalline diamond after an exposure to a r.f. plasma Ž500⬚C, 2 h. Žsolid curve.. The deuterium depth profile

408

D. Ballutaud et al. r Diamond and Related Materials 10 (2001) 405᎐410

presents an enhanced deuterium concentration in the surface region. This sub-surface accumulation may be due to deuterium trapping on plasma-induced defects or free deuterium diffusion Ždotted line in Fig. 3.. On the other hand, this diffusion profile cannot be described on the basis of wD᎐Dx molecule formation Žsee Eq. Ž4.. w11x. ŽThe dashed line in Fig. 3 represents an attempt of fit with a capture radius of 0.1 nm.. The high concentration region appears to be distinct from the region at lower concentrations which has been fitted finally to an exponential decay Ždotted-dashed line on Fig. 3. using the same characteristic length of ds 1r␣ s 95 nm as in Fig. 2. Fitting the high-concentration region of the deuterium diffusion profile with an erfc function Ždotted line in Fig. 2. would yield the effective diffusion coefficient Deff s 1 = 10y1 5 cm2 sy1 Žat 500⬚C., of the same order, but lower than the value Deff s 4 = 10y1 4 cm 2 sy1 found at 800⬚C in the case of diffusion from a microwave plasma as mentioned above. In both cases ᎏ r.f. and microwave plasma ᎏ the analysis of the deuterium diffusion profiles is performed on the basis of a model of saturable traps with the same mean free path ds 95 nm. In order to investigate the role of grain sizes, deuteration has been performed on 0.2-␮m grain size diamond layers. As the total deuterium profile depth is less than 250 nm, the grain size may be considered as constant to within approximately 0.02 ␮m. Deuterium diffusion profiles, obtained from a r.f. plama Ž500⬚C, 3 h., are presented on Fig. 4, respectively in normal conditions Žcurve 1. and in ‘remote’ plasma by mounting the sample on the side out of the plasma in order to attenuate the flux of ionized deuterium into the sample Žcurve 2.. The enhanced deuterium concentration observed in the surface region is decreased by operating in remote plasma Žcurve 2.. In this case, the deu-

Fig. 3. Deuterium concentration profile in a CVD polycrystalline layer Ž2-␮m grain size. after exposure to r.f. plasma. The solid line is the experimental profile, the dotted line represents a fit to an erfc function, the dashed line represents a fit assuming molecule formation and the dotted-dashed line a fit to an exponential decay.

Fig. 4. Deuterium concentration profile in a CVD polycrystalline layer Ž0.2-␮m grain size., curve 1: after exposure to r.f. plasma; curve 2: after exposure to remote r.f. plasma; and curve 3: after exposure to microwave plasma. The dotted lines represent the fits to an exponential decay.

terium diffusion profile may be fitted assuming an unsaturable trap: w Dx s w D 0 x exp Ž y␣ x .

Ž see Eq. Ž 4 ..

with ␣ s 4 ␲ R w nT x and x the diffusion depth. Fitting the depth profile to an exponential decay, it is found a mean free path between trapping events ds 1r␣ s 42 nm. The deuterium profile obtained in the same layer after an exposure to microwave plasma Ž800⬚C, 2 h. has been also reported on Fig. 4 Žcurve 3.. The use of microwave plasma instead of r.f. plasma causes a decrease of the deuterium surface concentration. The dotted line on curve 3 represents the fit to an exponential decay, with a decay length of 40 nm, similar to the value obtained for remote r.f. plasma ŽFig. 4 curve 2.. This result ᎏ mean free path independant of temperature and time ᎏ suggests that the low concentration region represents deuterium trapping onto deep traps. The same d value found for r.f. plasma, remote r.f. plasma and microwave plasma ŽFig. 4., seems to be governed by the polycrystalline structure, the respective shifts of the deuterium diffusion profiles following the y-axis Žcurves 1, 2 and 3 of Fig. 4. being due to the deuterium surface concentrations levels induced, respectively, by the three different plasmas that has been used for the diffusion experiments. The mean free path d between trapping events is roughly twice as large in the 2-␮m grain size layer than in the 0.2-␮m grain size one. In both cases, d is small compared to the grain size. Normally, the mean free path would be approximately equal to the grain size if the traps were associated with grain boundaries. The smaller mean free paths suggest that the traps are also intragranular point defects. This results have to be compared with those obtained in monocrystalline CVD

'

D. Ballutaud et al. r Diamond and Related Materials 10 (2001) 405᎐410

409

mal annealing will be achieved in order to confirm the origin of the different effusion peaks.

4. Conclusion

Fig. 5. Deuterium effusion spectra of a CVD polycrystalline layer Ž2-␮m grain size., curve 1: after exposure to a r.f. plasma; curve 2: after exposure to microwave plasma; and curve 3: sample deuterated during the deposition process.

epitaxial diamond layers and bring up the question of hydrogen nature and location in diamond, where the presence of hydrogen platelets could be assumed as in monocrystalline silicon w13x.

3.4. Effusion results

Fig. 5 represents the results of effusion experiments performed on the 2-␮m grain size samples. Comparing with the two samples exposed to deuterium plasma Žcurves 1 and 2., the deuterium is more stable when it is introduced in the layer during the deposition process Žcurve 3.. The deuterium effusion in this case becomes effective at 920⬚C, while the two samples exposed to plasma exhibit broad peaks centered at lower temperatures. The sample exposed to r.f. plasma Žcurve 1. shows two peaks centered at 750 and 860⬚C, respectively. For the sample exposed to the microwave plasma Žcurve 2., the effusion peak is centered at 860⬚C, with a shoulder at 750⬚C. From the comparison of the deuterium diffusion profiles and the effusion results and referring to previous results obtained in monocrystalline CVD diamond epilayers by Chevallier et al. w14,15x, one can try to attribute the effusion peaks to different deuterium locations in the polycrystalline diamond layers. The peak at 750⬚C is due to the deuterium subsurface accumulation, important in case of r.f. plasma ŽFig. 3., the peak at 860⬚C being related to deuterium adsorbed at the surface of the sample w14x. Deuterium coming from detrapping from bulk intragranular and intergranular defects would effuse at approximately 920⬚C, contributing then to the peak at 860⬚C. Some complementary effusion experiments after sample etching or ther-

From the shape of the deuterium concentration profiles, it appears that deuterium diffusion is governed mainly by trapping on defects in CVD polycrystalline diamond layers, deuterium-induced defects in the subsurface region andror defects due to the polycrystalline nature of the layers. In the 2-␮m grain size layers, the deuterium diffusion profiles may be described on the basis of a model of saturable traps for both r.f. and microwave plasma post-deuterations, while in the 0.2-␮m layers the diffusion profiles are described on the basis of a model of unsaturable traps. The deuterium mean free path values, calculated by fitting the diffusion profiles, are smaller than the grain sizes and show that the traps are not only associated with grain boundaries but are located also within the grains. The same mean free path is found with r.f. or microwave plasma. It decreases as the grain size decreases. The hydrogen content in as-grown 20-␮m-thick CVD polycrystalline diamond layers with grain size of 2 ␮m has been measured using deuterium as vector gas as a tracer and is found to be approximately 3 = 10 19 at. cmy3 . The deuterium bonding in as-grown samples is more stable than deuterium introduced in the diamond layer by plasma post-deuteration. In the plasma exposed samples, the two effusion peaks at 750 and 860⬚C have been attributed, respectively, to the deuterium accumulation in the sub-surface region Ž750⬚C. and deuterium adsorbed at the diamond surface or trapped on intragranular and intergranular defects Ž860⬚C.. References w1x J. Kuppers, Surf. Sci. Rep. 22 Ž1995. 249. ¨ w2x Applications of Diamond Films and Related Materials, edited by Y. Tzeng, M. Yoshikawa, M. Murakawa and A Feldman, Amsterdam: Elsevier, 1991. w3x R. Rizk, P. de Mierry, D. Ballutaud, M. Aucouturier, D. Mathiot, Phys. Rev. B 44 Ž12. Ž1991. 6141. w4x R. Rizk, P. de Mierry, D. Ballutaud, M. Aucouturier, D. Mathiot, Phys. B 170 Ž1990. 129. w5x D. Ballutaud, M. Aucouturier, F. Babonneau, Appl. Phys. Lett. 49 Ž23. Ž1986. 1620. w6x K. Hayashi, S. Hamanaka, H. Watanabe, T. Sekigushi, H. Okushi, K. Kajimura, J. Appl. Phys. 81 Ž1997. 744. w7x C. Jany, F. Foulon, P. Bergonzo, B. Guizard, C. Borel, A. Brambilla, S. Haan, A. Tardieu, A. Gicquel, Proceeding of the fifth international symposium on diamond materials, Paris, September, Electrochem. Soc. Proc. 97᎐32 Ž1998. Ž1997. 208᎐216. w8x D. Ballutaud, P. de Mierry, J.-C. Pesant, R. Rizk, A. BoutryForveille, M. Aucouturier, Mater. Sci. Forum 83᎐87 Ž1992. 45.

410

D. Ballutaud et al. r Diamond and Related Materials 10 (2001) 405᎐410

w9x R. Samlenski, J. Schmalzlin, R. Brenn, C. Wild, W. Muller-Se¨ bert, P. Koidl, Diam. Relat. Mater. 4 Ž1995. 503. w10x J. Chevallier et al., private communication. w11x Hydrogen in Crystalline Semiconductors, S.J. Pearton, J.W. Corbett, M. Stavola ŽEds.., Springer Series in Materials Science, Springer-Verlag, Berlin, 1992, p. 200. w12x N.H. Nickel, W.B. Jackson, J. Walker, Phys. Rev. B 53 Ž12. Ž1996. 7750᎐7761.

w13x H.H. Lamb, S.G. Bedge, Z. Wan, Mater. Res. Proc. 513 Ž1998. 381. w14x J. Chevallier, D. Ballutaud, B. Theys, F. Jomard, A. Deneuville, E. Gheeraert, F. Pruvost, Phys. Stat. Sol. 174 Ž1999. 73. w15x D. Ballutaud, F. Jomard, J. Le Duigou, B. Theys, J. Chevallier, A. Deneuville, F. Pruvost, Diam. Relat. Mater. 9 Ž2000. 1171.