- Email: [email protected]
Influence of the Germanium content on the amorphization of silicon–germanium alloys during ion implantation
Materials Science in Semiconductor Processing 16 (2013) 1655–1658
Contents lists available at SciVerse ScienceDirect
Materials Science in Semiconductor Processing journal homepage: www.elsevier.com/locate/mssp
Influence of the Germanium content on the amorphization of silicon–germanium alloys during ion implantation A. Belafhaili a, L. Laânab a,n, F. Cristiano b, N. Cherkashin c, A. Claverie c a b c
LCS, Faculté des Sciences, Université Mohammed V, Rabat, Morocco LAAS/CNRS, 7 Av de Col Roche , Toulouse Cedex, 31077, France CEMES/CNRS and Université de Toulouse, BP 4347, F-31055 Toulouse Cedex, France
a r t i c l e i n f o
abstract
Keywords: Ion implantation SiGe alloy Amorphization Simulation Single defects
We have studied by transmission electron microscopy the amorphization of silicon– germanium (SiGe) alloys by Ge+ implantation. We show that when implanted with the same amorphization dose, the resulting amorphous layers get narrower when the Ge content increases. The experimental results can be simulated using the critical damage energy density model assuming that the amorphization threshold rises linearly with the Ge content from 3 eV/at for pure Si to 5 eV/at for pure Ge. These results and simulations are needed to optimize the fabrication of highly doped regions in SiGe alloys. & 2013 Elsevier Ltd. All rights reserved.
1. Introduction Silicon–germanium (SiGe) alloys are being integrated by microelectronics industries as channel materials for optimized p-type Metal-Oxide-Semiconductor (PMOS) devices owing to the much better hole mobility they offer as compared to Si. As in Si, highly p-doped source and drain regions can be fabricated following three process steps: (1) preamorphization of the SiGe crystal by Ge implantation, (2) low energy B+ implantation and (3) Solid Phase Epitaxial Regrowth (SPER) of the amorphous layer at low temperature i.e., typically in the 500–600 1C range [1]. Moreover, future planar MOS devices below the current nodes (typically below 22 nm) will be built on Silicon-OnInsulator (SOI) wafers, the traditional thin Si top layer being possibly replaced by a SiGe alloy (SGOI wafers). The successful implementation of the same sequence of process steps to form highly B-doped regions in SOI and SGOI wafers requires that the top crystalline region is not fully amorphized by the Ge implantation so that some thin crystalline layer remains close to the box to provide a seed
n
Corresponding author. Tel.: +212662096186; fax: +212537774261. E-mail addresses: [email protected], [email protected] (L. Laânab).
1369-8001/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.mssp.2013.04.014
to the defect-free SPER of the amorphous layer during annealing. An additional advantage of this approach is that the End-Of-Range (EOR) defects, which inevitably form below the original crystalline/amorphous interface during annealing [2] and which affect dopant diffusion and degrade the electrical characteristics of the device, will form in smaller densities and will dissolve faster than in the bulk because of the capability of the Si/SiO2 interface to recombine Si interstitials [3]. Thus, the optimal use of the pre-amorphization technique requires that models exist to predict the depth at which the Si and SiGe layers are amorphized by any given Ge implantation. At the moment, such a model exists for pure Si and Ge and it has been integrated into modern process simulators. But the effect of adding Ge to Si, i.e. forming a SiGe alloy, onto the efficiency of the damage generation and amorphization processes is, at best, controversial. On one hand, T. E. Haynes and O. W. Holland [4] reported that, for (Si+, 80–90KeV) implantations at room temperature, the amount of ion-induced damage increases with the Ge content at a “larger rate than expected from calculations”. They suggested that increasing the Ge fraction progressively reduces the mobility of primary defects within the collision cascades. A. N. Larsen et al.
1656
A. Belafhaili et al. / Materials Science in Semiconductor Processing 16 (2013) 1655–1658
[5,6] studied the damage produced by implanting relaxed n-type Si1−xGex alloys at room temperature with 2 MeV Si+ as a function of the Ge content. The ion doses were ranging from 1010 to 2 1015 cm−2. An enhanced level of damage and a strong decrease in the critical dose for the formation of a buried amorphous layer were observed when increasing the Ge content. More recently, G. Impellizzeri et al. [7] reported a higher damage rate in Ge than in Si, when investigating the implantation of germanium at different energies and doses. They attributed this effect to both the higher stopping power of Ge atoms and a reduced mobility of the defects within the collision cascades. Similarly, R. Kögler et al. [8] reported that, for a given set of implantation conditions, the amorphization threshold decreases and the extension of the buried amorphous layer increases when the Ge content increases. This time, the authors suggested that during annealing the recombination of vacancies and interstitials is relatively inefficient and that the widely used “+1 model” [9] describing the ion-induced damage in Si is not valid for SiGe alloys. On the other hand, R. Wittmann et al. [10] reported that while the implantation profiles in Ge are shallower than in Si, due to the larger nuclear and electronic stopping powers of Ge atoms, damage production is less efficient, because the displacement energy of atoms is larger (30 eV) and that the energy transfers from the ions to the primary recoils are smaller than in Si. Finally, our own works [11–13] have shown that, for a large variety of implantation conditions, the width of the generated amorphous layers decreases, i.e. the “amorphization threshold” increases when switching from silicon to germanium. Actually, this confusion arises because the different techniques used to characterize the “damage” created by ion implantation are sensitive to different types of disorder and that, consequently, the terminology used to qualify the observed “damage” differs from one author to the next. The goal of this work is to clarify the situation by firstly providing robust experimental results about the amorphization efficiency of Ge implantation into SiGe alloys of various Ge contents. These data, obtained by Transmission Electron Microscopy (TEM), will be tested against the Critical Damage Energy Density (CDED) model already used to predict the amorphization kinetics of both Si and Ge for a large variety of experimental, industry relevant, conditions.
Refs. [14,15]. The pre-amophization was performed by a Ge+ 35 keV implantation with a dose of 1.1015 cm−2 and at room temperature. Samples were then prepared by mechanical polishing and subsequent argon ion thinning until electron transparency and ready for cross-sectional imaging. Low beam densities were used for ion thinning while keeping the sample at low temperature to prevent the regrowth of the amorphous layers which are a possible artifact for Ge-rich layers.
3. Results and discussions Fig. 1 presents a set of bright-field XTEM images obtained from the five samples of increasing Ge content. In all samples, the Ge+ implantation has resulted in the formation of a continuous amorphous layer extending from the surface and toward a depth which decreases from 57 nm to 47 nm (+/−3 nm) when the Ge content increases from 0 to 1. The amorphous nature of the upper layer can be easily demonstrated by electron diffraction. At this point, several would claim that the “amorphization efficiency” decreases when increasing the Ge content in SiGe layers. We prefer to remain factual and state that the width of the amorphous layers generated by some given Ge implantation decreases when increasing the Ge content. Fig. 2 shows the damage energy distributions, i.e. the energy transferred by the incident ion and the recoiling atoms to the target through nuclear collisions at each depth, obtained for a 35 keV Ge+ implantation into Si1 −xGex alloys of different compositions. These profiles have been obtained by Monte Carlo simulation of the slowing down process of the ions using displacement energies of 15 eV and 30 eV for Si and Ge, respectively [16]. This damage energy is proportional to the number of Frenkel's pairs generated in the crystal by the implantation. It is noticeable that the maximum of this damage energy shifts toward the surface while its amplitude increases with increase in the Ge content in the target. This characteristic results from both the increased stopping power of the
2. Experimental details By using an Epi Centura RP-CVD tool from Applied Materials, relaxed Si1−xGex alloy layers (about 1 μm thick) of various Ge contents (x¼0; 0.2; 0.35; 0.5 and 1) were grown by Chemical Vapor Deposition (CVD), via a compositionally graded buffer layer, in which the Ge is incorporated into the sample at an increasing rate of 10% Ge/μm. The virtual substrates were deposited on 200 mm slightly p-type doped Si(001) substrates. Dichlorosilane and germane diluted at 2% in hydrogen were used as the silicon and the germanium gaseous precursors. The growth pressure was fixed to 20 Torr for all the samples. Details on the growth kinetics in RP-CVD in the temperature range used to grow the SiGe virtual substrates can be found in
Fig. 1. Set of XTEM images showing the amorphous layers created by a 35 keV 1.1015 cm−2 Ge+ implantation at RT in Si1−xGex alloys of increasing Ge content. The upper dotted line delineates the surface, and the lower one is to help visualize the shift of the c/a interface.
A. Belafhaili et al. / Materials Science in Semiconductor Processing 16 (2013) 1655–1658
Fig. 2. Damage energy distributions obtained for a 35 keV 1.1015 cm−2 Ge+ implantation into Si1−xGex alloys of increasing Ge content. Note the shift towards the surface and the increase of damage when increasing the Ge content.
target and the larger nuclear energy transfers when increasing the Ge content of the target. Again, this characteristic alone leads to state that “damage generation” increases when increasing the Ge content, as it was experimentally observed by A. N. Larsen et al. [5,6] in the case of a damaged crystal and/or the formation of a buried amorphous layer (not extending up to the surface). Here we focus on the ability of the CDED model to describe the observations shown in Fig. 1. This model is used to predict Si amorphization and is integrated into modern process simulators. Recently, it has been shown that it could also be used to predict the formation and extension of amorphous layers created by a large variety of ions of different energies and doses in Ge [11,12]. This model states that the amorphization process can be seen as a first order transition: a region of a crystal that contains a high enough, critical, density of defects will spontaneously relax into the amorphous state [17,18]. Since the number of Frenkel pairs initially created by one “average” ion is proportional to the “damage energy” transferred to the crystal, to this critical defect density corresponds a damage energy density Edc. This parameter, also named “amorphization threshold”, is an intrinsic characteristic of the crystalline target and is thus thought to vary depending on the Ge content of the alloy. Following the CDED model, the dose D(z) (cm−2) needed to amorphize a SiGe crystal at a depth of z obeys the relation D(z) x Ed(z) ¼Edc, where Ed(z) is the damage energy received by the target at the depth, z, and expressed in eV/ion/nm. We have reported in Fig. 3 the depth-positions of the c/a interface we have measured by TEM and shown in Fig. 1. The widths of the experimental bars correspond to the measured roughness of these interfaces. We have also reported the prediction of the CDED model based on our simulations (Fig. 2) and assuming different Edc values for the c-a transition to occur. Each Edc value has been chosen so that the curve obtained from simulation, which separates the crystalline from the amorphous regions, passes very close to the experimental point which represents the position of the crystal to amorphous interface.
1657
Fig. 3. Comparison between the depth positions of amorphous layers created by self-implantation measured by XTEM and predicted by the CDED model assuming different Edc values for the different SiGe alloys. The bars centered on the experimental results represent the interfacial roughness (about 3 nm).
Fig. 4. Variation of the critical damage energy density values as a function of the Ge content in the SiGe alloys. The quite large vertical bars only reflect the measured roughness of the interfaces and must not be seen as true error bars.
An excellent fit is obtained for Edc ¼ 5 eV/atom in the case of germanium, for example. It is clear that the model well describes the experimental results assuming Edc values ranging from 3 to 5 eV/at depending on the Ge content in the alloy. Finally, we have reported in Fig. 4 the variation of the Edc values needed to describe our results as a function of the composition of the SiGe alloys. It evidences a linear increase of the Edc values from 3 to 5 eV/at as the composition of the SiGe alloys varies from x ¼0 (pure Si) to x ¼1 (pure Ge). These extreme values are identical to those found in silicon by L. Laânab et al. [19] for Ge+ implantations over the 15–150 keV energy range and by Koffel et al. in germanium for a very large set of ions, does and energies [11,12,20]. Moreover, in both cases, the CDED model was able to describe the progression of the amorphous layers for increasing doses.
1658
A. Belafhaili et al. / Materials Science in Semiconductor Processing 16 (2013) 1655–1658
Actually, the fact that Edc increases when passing from Si to Ge does not straightforwardly mean that Ge is more resistant to amorphization than Si. Stating that Edc equals 3 eV/at. in Si means that 1/5 of the target atoms have been displaced from their crystalline sites (0.2 dpa) since the calculations have been carried out using a displacement energy of 15 eV for Si. For Ge, 5 eV/at. corresponds to 1/6 dpa. This shows that in Si, Ge and SiGe alloys, the crystalline to amorphous phase transition occurs when the crystals contain more or less the same concentration of defects. 4. Conclusion We have studied by TEM the influence of the Ge content on the amorphization kinetic of SiGe alloys, induced by a Ge+ implantation. Our results show that the CDED model, fed by accurate simulations of the damage energy deposition, is able to predict the widths of the amorphous layers created by Ge+ implantation in SiGe alloys whatever their compositions. The critical damage energy density needed for the crystalline to amorphous transition to occur in Si1−xGex alloys follows the Vegard's law, i.e. is a linear function of the Ge content. Revisiting the previous statements found in literature which have sustained some confusion over the subject, we can conclude that, for one given set of implantation conditions, when increasing the Ge content in SiGe alloys, 1. the damage zone gets narrower (stopping increase); 2. the amplitude of the damage increases locally at its maximum (nuclear stopping increase); 3. the amorphization threshold (Edc) increases; 4. as a consequence of the previous points, the amorphous layers get narrower and shallower. Finally, these data can be used and injected into process simulators aimed at optimizing the amorphization conditions required for the fabrication of highly doped regions on SGOI wafers. Acknowledgment This work has been financially supported by the PHC “Volubilis” program (Action Intégrée no MA/09/212).
The authors would like to thank J.M. Hartmann for the growth and providing of the SiGe samples. References [1] E.J.H. Collart, S.B. Felch, H. Graoui, D. Kirkwood, S. Tallavarjula, J.A. Van den Berg, J. Hamilton, N.E.B. Cowern, K.J. Kirkby, Materials Science and Engineering B 114–115 (2004) 118–129. [2] A. Claverie, B. Colombeau, B. De Mauduit, C. Bonafos, X. Hebras, G. Ben Assayag, F. Cristiano, Applied Physics A-Materials Science and Processing 76 (2003) 1025–1033. [3] E.M. Bazizi, P.F. Fazzini, A. Pakfar, C. Tavernier, B. Vandelle, H. Kheyrandish, S. Paul, W. Lerch, F. Cristiano, Journal of Applied Physics 107 (2010) 074503. [4] T.E. Haynes, O.W. Holland, Applied Physics Letter 61 (1992) 61. [5] A. Nylandsted Larsen, C. O’Raifeartaigh, R.C. Barklie, B. Holm, F. Priolo, G. Franzo, G. Lulli, M. Bianconi, R. Nipoti, J.K.N. Lindner, A. Mesli, J.J. Grob, F. Cristiano, P.L.F. Hemment, Journal of Applied Physics 81 (5) (1997) 2208–2218. [6] C. O’Raifeartaigh, R.C. Barklie, A. Nylandsted Larsen, F. Priolo, G. Franz, G. Lulli, M. Bianconi, J.K.N. Lindner, F. Cristiano, P.L. F. Hemment, NIMB B120 (1996) 165–168. [7] G. Impellizzeri, S. Mirabella, M.G. Grimaldi, Applied Physics A 103 (2011) 323–328. [8] R. Kögler, A. Mucklich, W. Skorupa, A. Peeva, A. Yu. Kuznetsov, J.S. Christensen, B.G. Svensson, Journal of Applied Physics 101 (3) (2007)033508-1–033508-5. [9] M. Giles, Journal of Electrochemical Society 138 (1991) 1160. [10] R. Wittmann, S. Selberherr, Solid-State Electronics 51 (2007) 982–988. [11] S. Koffel, P. Scheiblin, A. Claverie, G. Benassayag, Journal of Applied Physics 105 (2009) 013528. [12] A. Claverie, S. Koffel, N. Cherkashin, G. Benassayag, P. Scheiblin, Thin Solid Films 518 (2010) 2307–2313. [13] L. Laânab, A. Roumili, M.M. Faye, N. Gessinn, A. Claverie, Materials Research Society Symposium Procedings 279 (1993) 529. [14] Y. Bogumilowicz, J.M. Hartmann, G. Rolland, T. Billon, Journal of Crystal Growth 274 (2005) 28–37. [15] Y. Bogumilowicz, J.M. Hartmann, F. Laugier, G. Rolland, T. Billon, N. Cherkashinc, A. Claverie, Journal of Crystal Growth 283 (2005) 346–355. [16] R. Wittmann, S. Uppal, A. Hossinger, J. Cervenka, ECS Selberherr, Transactions 3 (7) (2006) 667–676. [17] C. Vieu, A. Claverie, J. Fauré, J. Beauvillain, NIMB B36 (1989) 137. [18] J. Narayan, D. Fathy, O.S. Oen, O.W. Holland, Journal of Vacuum Science Technology A 2 (1984) 1303. [19] L. Laânab, Ph.D. Thesis, University of Paul Sabatier, Toulouse, 1993. [20] S. Koffel, A. Claverie, G. BenAssayag, P. Scheiblin, Materials Science in Semiconductor Processing 9 (4–5) (2006) 664–667.
Contents lists available at SciVerse ScienceDirect
Materials Science in Semiconductor Processing journal homepage: www.elsevier.com/locate/mssp
Influence of the Germanium content on the amorphization of silicon–germanium alloys during ion implantation A. Belafhaili a, L. Laânab a,n, F. Cristiano b, N. Cherkashin c, A. Claverie c a b c
LCS, Faculté des Sciences, Université Mohammed V, Rabat, Morocco LAAS/CNRS, 7 Av de Col Roche , Toulouse Cedex, 31077, France CEMES/CNRS and Université de Toulouse, BP 4347, F-31055 Toulouse Cedex, France
a r t i c l e i n f o
abstract
Keywords: Ion implantation SiGe alloy Amorphization Simulation Single defects
We have studied by transmission electron microscopy the amorphization of silicon– germanium (SiGe) alloys by Ge+ implantation. We show that when implanted with the same amorphization dose, the resulting amorphous layers get narrower when the Ge content increases. The experimental results can be simulated using the critical damage energy density model assuming that the amorphization threshold rises linearly with the Ge content from 3 eV/at for pure Si to 5 eV/at for pure Ge. These results and simulations are needed to optimize the fabrication of highly doped regions in SiGe alloys. & 2013 Elsevier Ltd. All rights reserved.
1. Introduction Silicon–germanium (SiGe) alloys are being integrated by microelectronics industries as channel materials for optimized p-type Metal-Oxide-Semiconductor (PMOS) devices owing to the much better hole mobility they offer as compared to Si. As in Si, highly p-doped source and drain regions can be fabricated following three process steps: (1) preamorphization of the SiGe crystal by Ge implantation, (2) low energy B+ implantation and (3) Solid Phase Epitaxial Regrowth (SPER) of the amorphous layer at low temperature i.e., typically in the 500–600 1C range [1]. Moreover, future planar MOS devices below the current nodes (typically below 22 nm) will be built on Silicon-OnInsulator (SOI) wafers, the traditional thin Si top layer being possibly replaced by a SiGe alloy (SGOI wafers). The successful implementation of the same sequence of process steps to form highly B-doped regions in SOI and SGOI wafers requires that the top crystalline region is not fully amorphized by the Ge implantation so that some thin crystalline layer remains close to the box to provide a seed
n
Corresponding author. Tel.: +212662096186; fax: +212537774261. E-mail addresses: [email protected], [email protected] (L. Laânab).
1369-8001/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.mssp.2013.04.014
to the defect-free SPER of the amorphous layer during annealing. An additional advantage of this approach is that the End-Of-Range (EOR) defects, which inevitably form below the original crystalline/amorphous interface during annealing [2] and which affect dopant diffusion and degrade the electrical characteristics of the device, will form in smaller densities and will dissolve faster than in the bulk because of the capability of the Si/SiO2 interface to recombine Si interstitials [3]. Thus, the optimal use of the pre-amorphization technique requires that models exist to predict the depth at which the Si and SiGe layers are amorphized by any given Ge implantation. At the moment, such a model exists for pure Si and Ge and it has been integrated into modern process simulators. But the effect of adding Ge to Si, i.e. forming a SiGe alloy, onto the efficiency of the damage generation and amorphization processes is, at best, controversial. On one hand, T. E. Haynes and O. W. Holland [4] reported that, for (Si+, 80–90KeV) implantations at room temperature, the amount of ion-induced damage increases with the Ge content at a “larger rate than expected from calculations”. They suggested that increasing the Ge fraction progressively reduces the mobility of primary defects within the collision cascades. A. N. Larsen et al.
1656
A. Belafhaili et al. / Materials Science in Semiconductor Processing 16 (2013) 1655–1658
[5,6] studied the damage produced by implanting relaxed n-type Si1−xGex alloys at room temperature with 2 MeV Si+ as a function of the Ge content. The ion doses were ranging from 1010 to 2 1015 cm−2. An enhanced level of damage and a strong decrease in the critical dose for the formation of a buried amorphous layer were observed when increasing the Ge content. More recently, G. Impellizzeri et al. [7] reported a higher damage rate in Ge than in Si, when investigating the implantation of germanium at different energies and doses. They attributed this effect to both the higher stopping power of Ge atoms and a reduced mobility of the defects within the collision cascades. Similarly, R. Kögler et al. [8] reported that, for a given set of implantation conditions, the amorphization threshold decreases and the extension of the buried amorphous layer increases when the Ge content increases. This time, the authors suggested that during annealing the recombination of vacancies and interstitials is relatively inefficient and that the widely used “+1 model” [9] describing the ion-induced damage in Si is not valid for SiGe alloys. On the other hand, R. Wittmann et al. [10] reported that while the implantation profiles in Ge are shallower than in Si, due to the larger nuclear and electronic stopping powers of Ge atoms, damage production is less efficient, because the displacement energy of atoms is larger (30 eV) and that the energy transfers from the ions to the primary recoils are smaller than in Si. Finally, our own works [11–13] have shown that, for a large variety of implantation conditions, the width of the generated amorphous layers decreases, i.e. the “amorphization threshold” increases when switching from silicon to germanium. Actually, this confusion arises because the different techniques used to characterize the “damage” created by ion implantation are sensitive to different types of disorder and that, consequently, the terminology used to qualify the observed “damage” differs from one author to the next. The goal of this work is to clarify the situation by firstly providing robust experimental results about the amorphization efficiency of Ge implantation into SiGe alloys of various Ge contents. These data, obtained by Transmission Electron Microscopy (TEM), will be tested against the Critical Damage Energy Density (CDED) model already used to predict the amorphization kinetics of both Si and Ge for a large variety of experimental, industry relevant, conditions.
Refs. [14,15]. The pre-amophization was performed by a Ge+ 35 keV implantation with a dose of 1.1015 cm−2 and at room temperature. Samples were then prepared by mechanical polishing and subsequent argon ion thinning until electron transparency and ready for cross-sectional imaging. Low beam densities were used for ion thinning while keeping the sample at low temperature to prevent the regrowth of the amorphous layers which are a possible artifact for Ge-rich layers.
3. Results and discussions Fig. 1 presents a set of bright-field XTEM images obtained from the five samples of increasing Ge content. In all samples, the Ge+ implantation has resulted in the formation of a continuous amorphous layer extending from the surface and toward a depth which decreases from 57 nm to 47 nm (+/−3 nm) when the Ge content increases from 0 to 1. The amorphous nature of the upper layer can be easily demonstrated by electron diffraction. At this point, several would claim that the “amorphization efficiency” decreases when increasing the Ge content in SiGe layers. We prefer to remain factual and state that the width of the amorphous layers generated by some given Ge implantation decreases when increasing the Ge content. Fig. 2 shows the damage energy distributions, i.e. the energy transferred by the incident ion and the recoiling atoms to the target through nuclear collisions at each depth, obtained for a 35 keV Ge+ implantation into Si1 −xGex alloys of different compositions. These profiles have been obtained by Monte Carlo simulation of the slowing down process of the ions using displacement energies of 15 eV and 30 eV for Si and Ge, respectively [16]. This damage energy is proportional to the number of Frenkel's pairs generated in the crystal by the implantation. It is noticeable that the maximum of this damage energy shifts toward the surface while its amplitude increases with increase in the Ge content in the target. This characteristic results from both the increased stopping power of the
2. Experimental details By using an Epi Centura RP-CVD tool from Applied Materials, relaxed Si1−xGex alloy layers (about 1 μm thick) of various Ge contents (x¼0; 0.2; 0.35; 0.5 and 1) were grown by Chemical Vapor Deposition (CVD), via a compositionally graded buffer layer, in which the Ge is incorporated into the sample at an increasing rate of 10% Ge/μm. The virtual substrates were deposited on 200 mm slightly p-type doped Si(001) substrates. Dichlorosilane and germane diluted at 2% in hydrogen were used as the silicon and the germanium gaseous precursors. The growth pressure was fixed to 20 Torr for all the samples. Details on the growth kinetics in RP-CVD in the temperature range used to grow the SiGe virtual substrates can be found in
Fig. 1. Set of XTEM images showing the amorphous layers created by a 35 keV 1.1015 cm−2 Ge+ implantation at RT in Si1−xGex alloys of increasing Ge content. The upper dotted line delineates the surface, and the lower one is to help visualize the shift of the c/a interface.
A. Belafhaili et al. / Materials Science in Semiconductor Processing 16 (2013) 1655–1658
Fig. 2. Damage energy distributions obtained for a 35 keV 1.1015 cm−2 Ge+ implantation into Si1−xGex alloys of increasing Ge content. Note the shift towards the surface and the increase of damage when increasing the Ge content.
target and the larger nuclear energy transfers when increasing the Ge content of the target. Again, this characteristic alone leads to state that “damage generation” increases when increasing the Ge content, as it was experimentally observed by A. N. Larsen et al. [5,6] in the case of a damaged crystal and/or the formation of a buried amorphous layer (not extending up to the surface). Here we focus on the ability of the CDED model to describe the observations shown in Fig. 1. This model is used to predict Si amorphization and is integrated into modern process simulators. Recently, it has been shown that it could also be used to predict the formation and extension of amorphous layers created by a large variety of ions of different energies and doses in Ge [11,12]. This model states that the amorphization process can be seen as a first order transition: a region of a crystal that contains a high enough, critical, density of defects will spontaneously relax into the amorphous state [17,18]. Since the number of Frenkel pairs initially created by one “average” ion is proportional to the “damage energy” transferred to the crystal, to this critical defect density corresponds a damage energy density Edc. This parameter, also named “amorphization threshold”, is an intrinsic characteristic of the crystalline target and is thus thought to vary depending on the Ge content of the alloy. Following the CDED model, the dose D(z) (cm−2) needed to amorphize a SiGe crystal at a depth of z obeys the relation D(z) x Ed(z) ¼Edc, where Ed(z) is the damage energy received by the target at the depth, z, and expressed in eV/ion/nm. We have reported in Fig. 3 the depth-positions of the c/a interface we have measured by TEM and shown in Fig. 1. The widths of the experimental bars correspond to the measured roughness of these interfaces. We have also reported the prediction of the CDED model based on our simulations (Fig. 2) and assuming different Edc values for the c-a transition to occur. Each Edc value has been chosen so that the curve obtained from simulation, which separates the crystalline from the amorphous regions, passes very close to the experimental point which represents the position of the crystal to amorphous interface.
1657
Fig. 3. Comparison between the depth positions of amorphous layers created by self-implantation measured by XTEM and predicted by the CDED model assuming different Edc values for the different SiGe alloys. The bars centered on the experimental results represent the interfacial roughness (about 3 nm).
Fig. 4. Variation of the critical damage energy density values as a function of the Ge content in the SiGe alloys. The quite large vertical bars only reflect the measured roughness of the interfaces and must not be seen as true error bars.
An excellent fit is obtained for Edc ¼ 5 eV/atom in the case of germanium, for example. It is clear that the model well describes the experimental results assuming Edc values ranging from 3 to 5 eV/at depending on the Ge content in the alloy. Finally, we have reported in Fig. 4 the variation of the Edc values needed to describe our results as a function of the composition of the SiGe alloys. It evidences a linear increase of the Edc values from 3 to 5 eV/at as the composition of the SiGe alloys varies from x ¼0 (pure Si) to x ¼1 (pure Ge). These extreme values are identical to those found in silicon by L. Laânab et al. [19] for Ge+ implantations over the 15–150 keV energy range and by Koffel et al. in germanium for a very large set of ions, does and energies [11,12,20]. Moreover, in both cases, the CDED model was able to describe the progression of the amorphous layers for increasing doses.
1658
A. Belafhaili et al. / Materials Science in Semiconductor Processing 16 (2013) 1655–1658
Actually, the fact that Edc increases when passing from Si to Ge does not straightforwardly mean that Ge is more resistant to amorphization than Si. Stating that Edc equals 3 eV/at. in Si means that 1/5 of the target atoms have been displaced from their crystalline sites (0.2 dpa) since the calculations have been carried out using a displacement energy of 15 eV for Si. For Ge, 5 eV/at. corresponds to 1/6 dpa. This shows that in Si, Ge and SiGe alloys, the crystalline to amorphous phase transition occurs when the crystals contain more or less the same concentration of defects. 4. Conclusion We have studied by TEM the influence of the Ge content on the amorphization kinetic of SiGe alloys, induced by a Ge+ implantation. Our results show that the CDED model, fed by accurate simulations of the damage energy deposition, is able to predict the widths of the amorphous layers created by Ge+ implantation in SiGe alloys whatever their compositions. The critical damage energy density needed for the crystalline to amorphous transition to occur in Si1−xGex alloys follows the Vegard's law, i.e. is a linear function of the Ge content. Revisiting the previous statements found in literature which have sustained some confusion over the subject, we can conclude that, for one given set of implantation conditions, when increasing the Ge content in SiGe alloys, 1. the damage zone gets narrower (stopping increase); 2. the amplitude of the damage increases locally at its maximum (nuclear stopping increase); 3. the amorphization threshold (Edc) increases; 4. as a consequence of the previous points, the amorphous layers get narrower and shallower. Finally, these data can be used and injected into process simulators aimed at optimizing the amorphization conditions required for the fabrication of highly doped regions on SGOI wafers. Acknowledgment This work has been financially supported by the PHC “Volubilis” program (Action Intégrée no MA/09/212).
The authors would like to thank J.M. Hartmann for the growth and providing of the SiGe samples. References [1] E.J.H. Collart, S.B. Felch, H. Graoui, D. Kirkwood, S. Tallavarjula, J.A. Van den Berg, J. Hamilton, N.E.B. Cowern, K.J. Kirkby, Materials Science and Engineering B 114–115 (2004) 118–129. [2] A. Claverie, B. Colombeau, B. De Mauduit, C. Bonafos, X. Hebras, G. Ben Assayag, F. Cristiano, Applied Physics A-Materials Science and Processing 76 (2003) 1025–1033. [3] E.M. Bazizi, P.F. Fazzini, A. Pakfar, C. Tavernier, B. Vandelle, H. Kheyrandish, S. Paul, W. Lerch, F. Cristiano, Journal of Applied Physics 107 (2010) 074503. [4] T.E. Haynes, O.W. Holland, Applied Physics Letter 61 (1992) 61. [5] A. Nylandsted Larsen, C. O’Raifeartaigh, R.C. Barklie, B. Holm, F. Priolo, G. Franzo, G. Lulli, M. Bianconi, R. Nipoti, J.K.N. Lindner, A. Mesli, J.J. Grob, F. Cristiano, P.L.F. Hemment, Journal of Applied Physics 81 (5) (1997) 2208–2218. [6] C. O’Raifeartaigh, R.C. Barklie, A. Nylandsted Larsen, F. Priolo, G. Franz, G. Lulli, M. Bianconi, J.K.N. Lindner, F. Cristiano, P.L. F. Hemment, NIMB B120 (1996) 165–168. [7] G. Impellizzeri, S. Mirabella, M.G. Grimaldi, Applied Physics A 103 (2011) 323–328. [8] R. Kögler, A. Mucklich, W. Skorupa, A. Peeva, A. Yu. Kuznetsov, J.S. Christensen, B.G. Svensson, Journal of Applied Physics 101 (3) (2007)033508-1–033508-5. [9] M. Giles, Journal of Electrochemical Society 138 (1991) 1160. [10] R. Wittmann, S. Selberherr, Solid-State Electronics 51 (2007) 982–988. [11] S. Koffel, P. Scheiblin, A. Claverie, G. Benassayag, Journal of Applied Physics 105 (2009) 013528. [12] A. Claverie, S. Koffel, N. Cherkashin, G. Benassayag, P. Scheiblin, Thin Solid Films 518 (2010) 2307–2313. [13] L. Laânab, A. Roumili, M.M. Faye, N. Gessinn, A. Claverie, Materials Research Society Symposium Procedings 279 (1993) 529. [14] Y. Bogumilowicz, J.M. Hartmann, G. Rolland, T. Billon, Journal of Crystal Growth 274 (2005) 28–37. [15] Y. Bogumilowicz, J.M. Hartmann, F. Laugier, G. Rolland, T. Billon, N. Cherkashinc, A. Claverie, Journal of Crystal Growth 283 (2005) 346–355. [16] R. Wittmann, S. Uppal, A. Hossinger, J. Cervenka, ECS Selberherr, Transactions 3 (7) (2006) 667–676. [17] C. Vieu, A. Claverie, J. Fauré, J. Beauvillain, NIMB B36 (1989) 137. [18] J. Narayan, D. Fathy, O.S. Oen, O.W. Holland, Journal of Vacuum Science Technology A 2 (1984) 1303. [19] L. Laânab, Ph.D. Thesis, University of Paul Sabatier, Toulouse, 1993. [20] S. Koffel, A. Claverie, G. BenAssayag, P. Scheiblin, Materials Science in Semiconductor Processing 9 (4–5) (2006) 664–667.