Lattice strain of hydrogen-implanted silicon: Correlation between X-ray scattering analysis and ab-initio simulations F. Rieutord, F. Mazen, S. Reboh, J. D. Penot, L. Bilteanu, J. P. Crocombette, V. Vales, V. Holy, and L. Capello Citation: Journal of Applied Physics 113, 153511 (2013); doi: 10.1063/1.4800538 View online: http://dx.doi.org/10.1063/1.4800538 View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/113/15?ver=pdfcov Published by the AIP Publishing

[This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 132.166.68.9 On: Fri, 29 Nov 2013 11:11:16

JOURNAL OF APPLIED PHYSICS 113, 153511 (2013)

Lattice strain of hydrogen-implanted silicon: Correlation between X-ray scattering analysis and ab-initio simulations F. Rieutord,1,a) F. Mazen,2 S. Reboh,2 J. D. Penot,3 L. Bilteanu,4 J. P. Crocombette,4 V. Vales,5 V. Holy,5 and L. Capello6 1

CEA-Grenoble, INAC-SP2M, 38054 Grenoble, France CEA-LETI, MINATEC Campus, 17 rue des Martyrs, 38054 GRENOBLE Cedex 9, France 3 CEA-INES, 50 avenue du lac L eman, 73375 Le Bourget du Lac, France 4 CEA-Saclay, SRMP, 91191 Gif-sur-Yvette Cedex, France 5 Charles University, Prague, Czech Republic 6 SOITEC, Parc Technologique des Fontaines, Bernin, 38926 Crolles, France 2

(Received 19 November 2012; accepted 22 March 2013; published online 18 April 2013) Hydrogen implanted silicon has been studied using high resolution X-ray scattering. Strain induced by implantation has been measured as a function of implantation dose. The dependence of strain with implanted dose shows different regimes starting from linear to quadratic and saturation. The observed strain is consistent with ab-initio and elasticity calculations. Strain rate changes can be associated to the predominant location of hydrogen in bond center location. C 2013 AIP Publishing LLC [http://dx.doi.org/10.1063/1.4800538] V I. INTRODUCTION

Hydrogen implantation is a widespread process in the microelectronics industry for a variety of applications aiming at modifying the electrical or mechanical properties of semiconductor materials. Regarding electrical properties, applications are mostly based on the ability of hydrogen to passivate a large variety of defects (including dopant and surface states) in a wide variety of semiconductor crystalline states.1 For mechanical properties, hydrogen implantation is a key step of the so-called SmartCutTM technology,2 aiming at creating a weakened layer at a controlled distance beneath a surface, for further detachment and transfer of thin films of the implanted material. For this application (that we consider here), the chemical state of hydrogen and the relation to the strain level induced is of special importance. In the sole case of silicon, several studies have been issued on this process that have demonstrated a wide range of association of hydrogen with various types of defects using, e.g., infra-red (IR) spectroscopy.4 The mechanical effect of ion implantation has been also studied with the effect of crystal orientation,6 post implant annealing, and mechanical strength parameters of the implanted material.3 Unexpected results have been obtained such as an increase of strain as a function of material strength, for similar implant conditions.3 In this paper, we report on the strain behavior of silicon upon implantation dose, and the data are compared to ab-initio calculations results. II. EXPERIMENTS

[100] P-type silicon wafers (300 mm) were used in this study. Implantation energy was set to 60 keV resulting in 500 nm implantation depth using an Applied Materials Vista implanter. The temperature of the wafer has been maintained close to ambient conditions (25  C) keeping the implant current below 5 mA. The measure of strain induced by a)

Electronic mail: [email protected].

0021-8979/2013/113(15)/153511/4/$30.00

implantation can be readily performed using standard high resolution X-ray scattering. Fig. 1 shows a typical X-ray scattering curve from which the main parameters of the strain profile can be extracted. The width and amplitude of maximum strain can be obtained from the interference structure visible on the left-hand side of the 004 Bragg peak (Fig. 1). The interference structure is due to slabs of the crystal with expanded lattice parameters and the scattered intensity can be readily reproduced summing the amplitudes reflected from each lattice plane, within a standard kinematical approximation.6 The observation of narrow fringes originating from the distance between crystal surface and the strain profile also allows the accurate determination of the depth of the implant profile. The strain profile was described satisfactorily using a pseudoGaussian profile with 2 different half widths to account for the asymmetry of the implant profile (55 þ 45 nm). The typical width and distance from silicon/silicon oxide interface extracted from the X-ray data are 100 nm and 360 nm, respectively. Implant is performed through 145 nm thick oxide. The strain versus dose curve was recorded for implantation doses ranging from 1015 to 1.2  1017 Hþ/cm2 (Fig. 2). Three regimes can be identified on the curve. A first regime where strain is linear with dose is visible for fluences below 2  1016 Hþ/cm2. This regime is followed by a regime where strain increases faster with dose (2–8  1016 Hþ/cm2) and a saturation at elevated doses (above 8  1016 Hþ/cm2). The (“supra linear”) increase in the strain to dose rate has been observed in a number of instances8 and has been often attributed to radiation damage build-up in materials. III. ANALYSIS

We shall first concentrate on the low dose regime. The linear behavior below 1016 has a slope of 2.6  1019cm2/[Hþ] for the relation between (Da/a)max and total implanted dose, which corresponds to a proportion of cv ¼ 2.6  1024 cm3/ [Hþ] between the [Hþ]-concentration and relative strain, using a mean width profile of 100 nm. Note that strain profile and

113, 153511-1

C 2013 AIP Publishing LLC V

[This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 132.166.68.9 On: Fri, 29 Nov 2013 11:11:16

153511-2

Rieutord et al.

J. Appl. Phys. 113, 153511 (2013)

FIG. 1. High resolution X-ray scattering intensity around 004 Bragg peak. The interference structure on the left-hand side of the peak allows the determination of the strain profile (Hþ implant dose 3  1016 cm2).

hydrogen concentration profile (as measured, e.g., by secondary ion mass spectroscopy) are similar. A. Ab-initio calculations

The relaxation of the silicon atom network upon introduction of hydrogen atoms was obtained from ab-initio density functional theory (DFT) calculations aiming at calculating the energy landscape of various types of hydrogen defects in

silicon. Calculations were carried out with the SIESTA code using a 216 Si atom box in which 1 or 2 hydrogen atom(s) were introduced at different site locations and for different charge states. Calculations were performed using the generalized gradient approximation (GGA) of the exchangecorrelation functionals in PBE (Perdew-Burke-Ernzerhof) implementation, using a mesh cut-off of 150 Ry.11 These calculations provided the relaxed volume for constant pressure boundary conditions. They have been performed for a variety of locations and charge states of the H atom (Fermi level positions). We have reproduced Table I the results for the most energetically favorable defects. B. Elasticity analysis

The “free” volume expansion x per hydrogen atom has been obtained from the X-ray measured linear expansions along the normal to the surface ez using elasticity: The volume expansion is just three times the free linear expansion, with the restriction of taking into account a Poisson ratio effect (no strain is observed in plane due to the presence of the bulk silicon “substrate” hence a modification of the ez expansion) Da DV ð1 þ Þ ¼ ; a V 3ð1  Þ

(1)

where  is the Poisson ratio. The experimental relation between Da/a and concentration reads FIG. 2. Variation of the maximum strain as a function of the implanted dose showing the three different regimes. Solid line: low dose regime, slope ¼ 2.6  1019 cm2/Hþ. Dotted line: medium dose regime, slope ¼ 6  1019 cm2/Hþ.

Da=a ¼ cv N=V: Since DV ¼ Nx, one gets finally from Eq. (1)

[This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 132.166.68.9 On: Fri, 29 Nov 2013 11:11:16

153511-3

Rieutord et al.

J. Appl. Phys. 113, 153511 (2013)

TABLE I. Calculated expansion volumes (per specie) for different lowenergy species including one or two hydrogen atoms. Calculations are per˚ 3) assuming constant pressure formed on a 216-atom box (V ¼ 4410 A boundary conditions. BC ¼ bond center, AB ¼ anti-bonding, T ¼ tetrahedral, and Hex ¼ hexagonal refer to different sites for atom or molecule locations in the Si structure. The species in bold are those consistent with the lowdose slope, in italic those consistent with the medium dose slope. Species

˚ 3) DV (A

H0AB H0BC HþAB HþBC HAB HBC H20BCAB H20Hex H20T H21BCAB H2BCAB H2þBCBC Vacancy Interstitial

3,2 5,6 3,9 11,3 5,0 11.8 10.8 4.7 1.8 12.8 4.8 30.4 21.4 5.3

x ¼ cv

3ð1  Þ : 1þ

IV. DISCUSSION

Incorporation of one hydrogen atom in the 216 Si atom ˚ 3 volume) corresponds typically to the average box (4410 A concentration obtained when implanting 0.23  1016 Hþ/cm2 over a 100 nm layer. According to our strain measurements and Eq. (1), the corresponding volume increase per Hþ would ˚ 3. This value matches several values of be typically of 5 A expansions calculated for a variety of H species which have been observed using IR spectroscopy4 (H0BC, H2þBCAB, H20BCAB). All these configurations share a bond center location for one hydrogen, possibly associated to a hydrogen in anti-bonding position forming the H2* complex. These configurations are reported to be the most energetically favorable configurations for hydrogen in silicon. Note that, from the volume expansions only, the presence of interstitial defects cannot be excluded. Our calculations show that the expansion associated to one interstitial (consistent with literature12) per hydrogen could match the observations but it is very unlikely that one implanted hydrogen atom gives exactly one interstitial defect with no vacancy, located at the same final depth. A series of other observations can be connected to the volume measurements.13 Infra-red and Raman spectroscopy measurements show a strong increase with fluence of a broad absorption band around 2000 cm1, with a maximum close to the Si-HBC band position.4 The large width and intensity of the absorption around 2000 cm1 on as-implanted samples is reminiscent of features observed for the O-H stretching band when it establishes hydrogen bonds with neighboring molecules(-O-H…O).5 In the implanted Si lattice, it is not surprising that Si-H bands have large width centered around the exact position of pure Si-H-Si HBC line.

Note that the level of expansion of the lattice spacing is inconsistent with any kind of direct mechanical action (e.g., of the hydrogen gas) on the silicon and the stress calculated from strain under this assumption using elasticity is several times higher than the one directly measured from curvature experiments10 in these dose rates. The effect of hydrogen should rather be viewed as a chemical alloy formation rather than an external stressor on a silicon matrix. Also, when annealing the implanted layer at small thermal budgets, an increase of the maximum lattice strain is observed from the X-ray experiments. Again this is inconsistent with stress relaxation normally observed in such case. Finally, X-ray experiments have been performed at low temperatures (down to 4 K) to test a possible direct mechanical pressure effect from molecular hydrogen gas within cavities. No changes of strain have been observed on temperature, indicating the strain profile originates from direct action of individual hydrogen atoms on the lattice structure of silicon, at least when no annealing is performed. At higher doses, the strain behavior evolves in a faster (by a factor of 2.5) linear manner before saturating, giving some kind of sigmoidal shape to the strain dose-curve. The maximum slope would correspond to a volume change of ˚ 3/atom. The surprising behavior of an the order of 11 A increase in the strain to dose rate at higher doses has been observed in a number of instances.8,14 and has been so far attributed to radiation damage build-up in these materials although specific literature studies seem to indicate that such damage create rather a sublinear dependence to fluence.15,16 The contribution of point defects to the strain does not seem to be dominant here also, as the deformation profile is measured to coincide with the hydrogen distribution profile.6,7,17 The expansion due to point defect creation (e.g., Frenkel pairs), located in the upper profile18 is not visible on the X-ray profile for silicon in this dose range. Finally, the order of magnitude or the sign (contraction or expansion) is not consistent with a volume change induced by silicon atom displacement: a vacancy produces a strong contraction of the lattice. Diffuse X-ray scattering studies by Ehrhart and coworkers19,20 have directly measured the strain field around point defects after different types of irradiations. They have shown that point defects are mobile at 300 K leading to possible recombination and annihilation of close pair defects. This may explain their small contributions to initial strain. The exact concentration of such defects is difficult to estimate: associations of point defects with hydrogen are clearly visible on IR spectra, especially after a slight post implant anneal but may not be statistically significant at room temperature (the surface area of narrow peaks is small compared to the wide central component attributed to Si-H-Si4). Ab initio calculations show that the effect of one or two charged Hþ hydrogen in BC site inducing higher volume expansion may be consistent with the observed slope increase. As this positively charged species is known to be most stable for p-type silicon (due to lower acceptor level compared to donor level and negative-U character for H in silicon9), the correspondence of its expansion effect (HþBC and H2þBCBC) with measured value is expected.

[This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 132.166.68.9 On: Fri, 29 Nov 2013 11:11:16

153511-4

Rieutord et al.

Finally, a saturation is observed at higher doses. In this case, it is difficult to really make sure the deformation saturates as the X-rays probe the periodicity of lattice planes at a given level of strain only. Strains and implantation-induced damage also reduces the intensity reflected by individual planes according to some Debye-Waller factors that represent, in a similar way to thermal motions, the phasing out of different part of planes due to atomic displacement with respect to the mean plane level. At high doses, this effect is so large that the crystal planes do not reflect the incident wave and do not contribute to the reflected intensity any longer. Stress measurements indicate also a saturation at high dose.10 Ab-initio calculations performed including several H atoms in the box also show a tendency to reduce the total volume expansion for interacting atoms compared to the sum of individual atoms volume expansion. V. CONCLUSION

In summary, comparison of the measured expansion effects of hydrogen incorporation in silicon with ab-initio calculations shows that the observed expansions of the material upon implantation are consistent with predictions of volume change associated to the formation of the most stable and simple silicon hydrogen complexes. ACKNOWLEDGMENTS

We thank the RTRA-Fondation Nanoscience for the funding of the part-time chair of V. Holy. The BM32 beamline is funded by CEA and CNRS. Soitec is acknowledged for financial support.

J. Appl. Phys. 113, 153511 (2013) 1

Hydrogen in Semiconductors, edited by J. Pankove and N. Johnson, Semiconductors and Semimetals Vol. 34 (Academic Press, 1991). 2 M. Bruel, Electron. Lett. 31, 1201 (1995). 3 C. Ventosa-Moulet, S. L. Hayashi, and M. S. Goorsky, ECS Trans. 33(4), 249 (2010). 4 M. K. Weldon, V. E. Marsico, Y. J. Chabal, A. Agarwal, D. J. Eaglesham, J. Sapjeta, W. L. Brown, D. C. Jacobson, Y. Caudano, S. B. Christman, and E. E. Chaban, J. Vac. Sci. Technol. B 15, 1065–1073 (1997). 5 Y. Marechal, The Hydrogen Bond and the Water Molecule (Elsevier, 2007). 6 N. Sousbie, L. Capello, J. Eymery, F. Rieutord, and C. Lagahe, J. Appl. Phys. 99, 103509 (2006). 7 L. Capello, F. Rieutord, A. Tauzin, and F. Mazen, J. Appl. Phys. 102, 026106 (2007). 8 G. F. Cerofolini, L. Meda, C. Volpones, R. Dierckx, G. Mercurio, M. Anderle, R. Canter, F. Cembali, R. Fabbri, and M. Servidori, Nucl. Instrum. Methods Phys. Res. B 39, 26 (1989). 9 N. M. Johnson, C. Herring, and C. Van de Walle, Phys. Rev. Lett. 73, 130 (1994). 10 T. H€ ochbauer, A. Misra, M. Nastasi, and J. W. Mayer, J. Appl. Phys. 92, 2335 (2002). 11 L. Bilteanu, Ph.D. dissertation, Univ. Paris XI Orsay, France, 2010; J. M. Soler, E. Artacho, J. D. Gale, A. Garcıa, J. Junquera, P. Ordej on, and D. Sanchez-Portal, J. Phys.: Condens. Matter 14, 2745–2779 (2002). 12 S. A. Centoni, B. Sadigh, G. H. Gilmer, T. J. Lenosky, T. D. de la Rubia, and C. B. Musgrave, Phys. Rev. B 72, 195206 (2005). 13 B. Terreault, Phys. Status Solidi 204, 2129 (2007). 14 A. Debelle, L. Thome, D. Dompoint, A. Boulle, F. Garrido, J. Jagielski, and D. Chaussende, J. Phys. D: Appl. Phys. 43, 455408 (2010). 15 P. K. Giri, V. Raineri, G. Franzo, and E. Rimini, Phys. Rev. B 65, 012110 (2001). 16 E. Sealy, R. C. Barklie, G. Lulli, R. Nipoti, R. Balboni, S. Milita, and M. Servidori, Nucl. Instrum. Methods Phys. Res. B 96, 215 (1995). 17 S. Reboh, F. Schaurich, A. Declemy, J. F. Barbot, M. F. Beaufort, N. Cherkashin, and P. F. P. Fichtner, J. Appl. Phys. 108, 023502 (2010). 18 J. Keinonen, M. Hautala, E. Rauhala, M. Erola, J. Lahtinen, H. Huomo, A. Vehanen, and P. Hautojarvi, Phys. Rev. B 36, 1344 (1987). 19 K. Nordlund, P. Partyka, R. S. Averback, I. K. Robinson, and P. Ehrhart, J. Appl. Phys. 88, 2278 (2000). 20 P. Partyka, Y. Zhong, K. Nordlund, R. S. Averback, I. M. Robinson, and P. Ehrhart, Phys. Rev. B 64, 235207 (2001).

[This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 132.166.68.9 On: Fri, 29 Nov 2013 11:11:16