Epitaxial growth of non-cubic silicon - Semantic Scholar
Microelectronics Journal 36 (2005) 506–509 www.elsevier.com/locate/mejo
Epitaxial growth of non-cubic silicon A. Fissela,*, C. Wanga, E. Bugielb, H.J. Ostenb b
a Information Technology Laboratory, University of Hannover, Schneiderberg 32, D-30167 Hannover, Germany Institute for Semiconductor Devices and Electronic Materials, University of Hannover, Appelstraße 11A, D-30167 Hannover, Germany
Available online 19 March 2005
Abstract We report about the formation of a Si twinning-superlattice (TSL). Si layers containing 1808 rotation twin boundaries arranged periodical along the [111] direction and only separated by 2.5 nm (corresponding to TSL) have been prepared by molecular beam epitaxy. The method consists of repeating of several growth and annealing circles on boron(B)-predeposited undoped Si substrates. It is shown that the amount of subsurface pffiffiffi pffiffiffi B on the surface and the growth mode influence the formation of twin boundaries. Only the nucleation of Si on the ð 3 ! 3ÞR308-Si(111) surface covered by atleast 1/3 ML B resulted in the formation of a 1808 rotation twin. The presented technology should also be suitable to prepare a new type of semiconductor heterostructures based on Si polytypes. q 2005 Elsevier Ltd. All rights reserved. Keywords: Molecular beam epitaxy; Silicon; Superlattice; Twinning
1. Introduction Changes in crystal symmetry can have a significant impact on the electronic properties, such as band structure. In recent years therefore, new types of heterostructures consisting of only one material constituent in different crystal structures (heteropolytypic structures), such as wurtzite/zinc-blende or lonsdaleite(2H)/diamond(3C) heterostructures, are under discussion, where problems due to different chemical constituents can be avoided [1–5]. Heteropolytypic structures are also under discussion for the most common semiconductor, silicon [1,2,6–8]. The use of 2H/3C, or at least non-cubic/cubic, Si heterostructures should have some advantages, such as absence of alloy scattering and strain-induced defects, combined with properties that can otherwise only be achieved by incorporating SiGe or SiGeC into Si technology. It offers therefore, the potential of increased flexibility in the design of future Sibased electronic devices. Moreover, the novel properties of such structures could open new fields of Si applications, such as resonant tunnelling or other heterojunction devices. Hexagonal silicon is also an indirect semiconductor. The energy gap reduces with the hexagonality from 1.2 eV * Corresponding author. Fax: C49 511 762 5051. E-mail address: [email protected] (A. Fissel).
0026-2692/$ - see front matter q 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.mejo.2005.02.064
(6H–Si) to about 1 eV(2H–Si) [6]. In hexagonal polytypes, the threefold-degenerate valence-band maximum of cubic Si splits into a twofold and a lower onefold splittoff band. Thereby, the crystal-field splitting increases with the hexagonality [7]. The polytype combination 3C/2H leads to type-I heterostructure, while the combination 3C/4H and 3C/6H are of type-II, with electrons localization in the cubic parts and holes localization in the hexagonal parts. The electronic heteropolytypic structure is governed by the valence-band discontinuity, varying between 135 meV for 3C/6H and 235 meV for 3C/2H [7].
2. Growth of heteropolytypic Si structures The preparation of closed-packed non-cubic Si, however, is still a challenge. The transformation of 3C–Si crystals into the 2H-structure under high pressure and heat treatment was already reported more than 40 years ago [9]. The generation of hexagonal Si was often observed during indentation experiments [10,11]. Small hexagonal Si crystals were prepared by laser ablation [12] and laser beam annealing of amorphous Si thin films [13]. But so far, hexagonal Si could not be obtained in an epitaxial growth process. Even more challenging is the growth of heteropolytypic Si structures, because it demands defined growth conditions for each of the polytypes forming the structure, considering also the high energy of stacking fault formation in Si [14]. A change
A. Fissel et al. / Microelectronics Journal 36 (2005) 506–509
of stacking order can therefore, only be achieved by modification of the growth process, and in particular the surface structure. In case of Si(111) homoepitaxial MBE growth the influence of 1/3 monolayer (ML) subsurface B on the orientation of epitaxial Si layers has been reported [15,16]. Based on that, first attempt to prepare Si-based TSL has been reported based on B-mediated epitaxial growth [17]. However, in these experiments, segregation pffiffiffi pffiffiB-surface ffi has been used to induce ð 3 ! 3ÞR308 reconstructed Si(111) surfaces by high-temperature(HT) annealing (between 900 and 1250 8C) of highly B-doped Si(111) substrates. The growth experiments were performed on 0.38 misoriented Si(111) substrates leading to short terraces. Only after very HT-annealing (1250 8C) larger terraces were formed by step bunching. However, the superstructure periodicity was found to be not adjustable. Because of the inchoate way for obtaining the B surface coverage and the high temperatures used in the experiments, such a preparation is not suitable for any technological application in a Si-polytypes production process. Furthermore, the influence of other disturbing effects like B-clustering [18], oxygen diffusion or SiC formation along the twin formation can be not excluded. In this investigation, we report about the stacking conversion in molecular beam epitaxy (MBE) of Si/Si(111) leading to TSL by B deposition and mediumtemperature annealing. We demonstrate that the distance between the stacking faults can be controlled by the amount of Si MLs grown in each growth run. The structure formation strongly depends on the B surface coverage, the post-growth annealing temperature and the growth mode.
3. Experimental Si layers were grown by solid-source MBE on 4-inch n-type Si(111) (phosphor 1017 cmK3) wafer with a miscut of !0.18, prepared by standard wet-chemical RCA cleaning. Si was evaporated by electron beam evaporation and the flux was monitored by quadrupole based mass spectrometry. B was evaporated from high-temperature effusion cell at temperatures around 1800 8C, corresponding to a deposition rate of 0.05 ML/min. The growth and B-induced superstructure formation were investigated in situ by reflection high-energy electron reflection (RHEED). The layers were investigated ex situ by transmission electron microscopy (TEM) and m-Raman spectroscopy.
507
4. Results After annealing at 500 8C, the samples exhibit the known (7!7) surface reconstruction, (Fig. 1(a)). Exposing the surface pffiffiffi pto ffiffiffi B at 780 8C results in a gradual transition to a ð 3 ! 3ÞR308 surface superstructure (Fig. 1(b)), which already appears in the RHEED pattern at coverage around 0.1 ML. Here, B-atoms occupy the lowest-energy subsurface site (S5, Fig. 2) at TO750 8C [15]. That is a substitutional site in the second layer of the surface with a Si adatom directly above. After cooling down to 400 8C, 2.5 nm Si were grown with a rate of 0.1 nm/min on top of the reconstructed surface, resulting in a change of surface structure to (1!1) (Fig. 1(c)). p Then ffiffiffi pthe ffiffiffi samples were annealed at 780 8C and the ð 3 ! 3ÞR308 structure re-occurs indicating B-surface segregation (Fig. 1(d)). On top of this new surface, Si was grown again at 400 8C. The procedure was repeated six times. The next set of samples was grown on surfaces covered by 1/3 ML boron. At this coverage, the superstructure was full developed, as revealed by the maximum in 1/3 fractional order diffraction spot intensity. Further preparation was performed in the way described above. The last set of samples was prepared by stabilizing the 1/3 ML B during each annealing step by an excess of B deposition. The resulting layer structure was investigated by TEM. Layers grown on surfaces with initially B-coverage below 0.2 ML do not show any change in the stacking order. Stacking faults can also be introduced by surface contaminations, thus this results demonstrate that contamination effect can be neglected in our investigations. Fig. 3(a) shows a cross-sectional (X)TEM micrograph taken from a layer where initially 1/3 ML B was deposited and for the subsequent growth further B-surface coverage was realized only by annealing after each growth step. The micrograph shows a single stacking fault at the interface between the substrate and the layer. A repeated change in stacking direction was found partially only in some regions, what can be due to inhomogeneous B-segregation, as observed also in Si MBE [19]. That is also shown in the plan-view TEM image (Fig. 3(b)). The regions of homogeneous layer structure are in the range of some hundred nm. The selected area diffraction pattern (SAD) shown in the inset exhibits 6 bright and sharp spots in a 6-fold symmetry corresponding to {220} reflection of the Si[111] single-crystal direction. Some additional slight spots are visible in a distance of 1/3 to the main spots. These spots can be identified as 1/3{422}
pffiffiffi pffiffiffi Fig. 1. RHEED pattern of (a) initially clean (7!7) reconstructed Si(111) surface; (b) B-induced ð 3 ! 3ÞR308 surface structure after B deposition at 780 8C; (c) (1!1) surface structure after deposition of 4-Si double layers at 400 8C; (d) after annealing at 780 8C for of 10 min.
508
A. Fissel et al. / Microelectronics Journal 36 (2005) 506–509
clearly that stacking conversation is due to Si nucleation in twin position. Stabilizing the 1/3 ML B-surface coverage by an excess of B after each annealing step results in the formation of TSL in several regions with lateral dimensions in the order of 1 mm. Such TSL structure is shown in Fig. 4, where a XTEM micrograph of a TSL with eightfold periodicity (corresponding to the deposited of 2.5 nm Si) is presented together with a structural model of the TSL period. The structure presented in Fig. 4 has been investigated by m-Raman spectroscopy in comparison to untwined Si to get additional information about structural changes due to the artificially stacking of the Si bilayers. Two Raman active modes located at 502 and 520 cmK1 should be expected for hexagonal silicon [20]. However, only the signal at 520 cmK1 was found, indicating either that no change in the crystal structure had occurred or the number of lonsdaleite-like bonds is to small. pffiffiffi pffiffiffi Fig. 2. Illustration of adatom positions on the B-induced ð 3 ! 3ÞR308 Si(111) surface superstructure (a) top-view; (b) side-view showing B in the subsurface S5 site.
5. Summary and discussion
reflections, which are forbidden for the fcc structure. The appearance of these extra spots can be explained by stacking faults (rotation twins) occurring parallel to the Si(111) planes. Using off-oriented samples, no change in the stacking order were found under the same conditions. In this case step-flow dominates the growth process, indicating
MBE homoepitaxial growth of Si on subsurface pffiffiffi pffiffiffi B-induced ð 3 ! 3ÞR308 Si(111) surface have been investigated, which was prepared by B deposition. It has been shown that Si growth on such a surface containing 1/3 ML B introduces a rotation twin at the layer-substrate interface, indicating that the stacking fault energy becomes
Fig. 3. (a) XTEM micrograph of a layer grown at initially 0.3 ML B-coverage and subsequent B-coverage by boron surface segregation after each Si growth run; (b) XTEM micrograph with lower magnification; (c) plan-view image with SAD pattern obtained in [111] direction exhibiting 1/3{422} superstructure spots due to twinning.
Fig. 4. (a) High-resolution XTEM micrograph of TSL with eightfold periodicity; (b) crystal structure model of a TSL sequence.
A. Fissel et al. / Microelectronics Journal 36 (2005) 506–509
negative in the presence of 1/3 ML subsurface B likely induced by strain within the pSi-B surface layer. ffiffiffi palloyed ffiffiffi Repeating several times the ð 3 ! 3ÞR308 surface preparation by annealing and B deposition to stabilize the 1/3 ML coverage and subsequent Si growth with defined thickness results in the formation of a TSL in which twin boundaries appear periodical along the [111] growth direction. By such a two-step B-mediated epitaxial growth the preparation of close-packed non-cubic Si as well as noncubic/cubic Si heterostructures should be possible. However, already such TSL should have some new interesting properties as suggested recently [21]. It was predicted that both the conduction and the valence band split into minibands separated by energies in the range of 100 meV. Moreover, it has also been suggested that the stacking fault planes will act as potential barriers to the carrier diffusion rather than as carrier recombination centres. The barrier height of the stacking faults was estimated to be 55 meV from temperature dependent electron beam induced current contrast [22]. Therefore, TSL already would offer almost as much versatility in tailoring the electronic miniband structure as there exists in ordinary heterostructure-based superlattices. Such possibility for ‘band-structure engineering’ could even more be enhanced when combining with material composition and doping variations. Besides the investigations of TSL properties, the preparation of a really new structure type of Si will be a subject of further investigations. From recent RHEED and scanning tunnelling microscopy (STM) investigations it can be concluded [16], that the preparation of structures containing twins with distances of two bilayers is possible. That would results in the formation of a hexagonal 4H–Si polytype.
Acknowledgements The authors like to thank Prof. Karl-Ludwig Oehme and his group from the University of Jena for performing mRaman measurements and Michael Seibt from the University of Go¨ttingen for using the TEM equipment.
References [1] M. Murayama, T. Nakayama, Electronic structures of hetero-crystalline semiconductor superlattices, J. Phys. Soc. Jpn. 61 (1992) 2419– 2433; M. Murayama, T. Nakayama, Chemical trend of band offsets at wurtzite/zinc-blende heterocrystalline semiconductor interfaces, Phys. Rev. B 49 (1994) 4710–4724.
509
[2] Ch.-Y. Yeh, Z.W. Lu, S. Froyen, A. Zunger, Zinc-blende-wurtzite polytypism in semiconductors, Phys. Rev. B 46 (1992) 10086–10097. [3] F. Bechstedt, P. Ka¨ckell, Heterocrystalline structures: new types of superlattices? Phys. Rev. Lett. 75 (1995) 2180–2183. [4] S.-H. Ke, J. Zi, K.-M. Zhang, X.-D. Xie, Electronic structures and band offsets of heterocrystalline superlattice, Phys. Rev. B 54 (1996) 8789–8793. [5] A. Fissel, Artificially layered heteropolytypic structures based on SiC polytypes: molecular beam epitaxy, characterization and properties, Phys. Rep. 379 (2003) 149–255. [6] C. Persson, E. Janzen, Electronic band structure in hexagonal closepacked Si polytypes, J. Phys. Condens. Matter 10 (1998) 10549– 10555. [7] C. Raffy, J. Furthmu¨ller, F. Bechstedt, Properties of hexagonal polytypes of group-IV elements from first-principles calculations, Phys. Rev. B 66 (2002) 075201/1–075201/10. [8] S.Q. Wang, H.Q. Ye, First-principle study on the lonsdaleite phases of C, Si and Ge, J. Phys. Condens. Matter 15 (2003) L197–L202. [9] R.H. Wentorf, J.S. Kasper, New forms of silicon, Science 139 (1963) 338. [10] A. Kailer, Y.G. Gogotsi, K.G. Nickel, Phase transformation of silicon caused by contact loading, J. Appl. Phys. 81 (1997) 3057–3063. [11] V.G. Eremenko, V.I. Nikitenko, Electron-microscope investigation of microplastic deformation mechanisms of silicon by indentation, Phys. Status Solidi A 14 (1972) 317–330. [12] Y. Zhang, Z. Iqbal, S. Vijayalakshmi, H. Grebel, Stable hexagonalwurtzite silicon phase by laser ablation, Appl. Phys. Lett. 75 (1999) 2758–2761. [13] J.H. Kim, J.Y. Lee, Hexagonal silicon formation by pulsed laser beam annealing, Mater. Lett. 27 (1996) 275–279. [14] P. Ka¨ckell, J. Furthmu¨ller, F. Bechstedt, Stacking faults in group-IV crystals: an ab initio study, Phys. Rev. B 58 (1998) 1326–1330. [15] R.L. Headrick, B.E. Weir, J. Bevk, B.S. Freer, D.J. Eaglesham, L.C. Feldman, Influence of surface reconstruction on the orientation of homoepitaxial silicon films, Phys. Rev. Lett. 65 (1990) 1128–1131. [16] H. Hibino, T. Ogino, Formation of twinned two-bilayer-high islands during initial stages of Si growth on Si(111)O3!O3-B, Surf. Sci. 412/413 (1998) 132–140. [17] H. Hibino, K. Sumimoto, T. Ogino, Twinned epitaxial layers formed on Si(111)O3!O3-B, J. Vac. Sci. Technol. A 16 (1998) 1934–1937. [18] F. Thibaudau, T.P. Roge, Ph. Methiez, P. Dumas, F. Salvan, Cooperative segregation of boron at Si(111), Europhys. Lett. 25 (1994) 353–356. [19] D. Kru¨ger, G. Lippert, R. Kurps, H.J. Osten, Lateral inhomogeneous boron segregation during silicon thin film growth molecular beam epitaxy, J. Cryst. Growth 135 (1994) 246–252. [20] R.J. Kobliska, S.A. Solin, Raman spectrum of wurtzite silicon, Phys. Rev. B 8 (1973) 3799–3802. [21] Z. Ikonic, G.P. Srivastava, J.C. Inkson, Electronic properties of twin-boudaries and twinning superlattices in diamond-type and zinc-blende-type semiconductors, Phys. Rev. B 48 (1993) 17181– 17193. [22] T. Sekiguchi, B. Shen, T. Watanabe, K. Sumino, EBIC study on the electrical activity of stacking faults in silicon, Mater. Sci. Eng. B 42 (1996) 235–239.
Epitaxial growth of non-cubic silicon A. Fissela,*, C. Wanga, E. Bugielb, H.J. Ostenb b
a Information Technology Laboratory, University of Hannover, Schneiderberg 32, D-30167 Hannover, Germany Institute for Semiconductor Devices and Electronic Materials, University of Hannover, Appelstraße 11A, D-30167 Hannover, Germany
Available online 19 March 2005
Abstract We report about the formation of a Si twinning-superlattice (TSL). Si layers containing 1808 rotation twin boundaries arranged periodical along the [111] direction and only separated by 2.5 nm (corresponding to TSL) have been prepared by molecular beam epitaxy. The method consists of repeating of several growth and annealing circles on boron(B)-predeposited undoped Si substrates. It is shown that the amount of subsurface pffiffiffi pffiffiffi B on the surface and the growth mode influence the formation of twin boundaries. Only the nucleation of Si on the ð 3 ! 3ÞR308-Si(111) surface covered by atleast 1/3 ML B resulted in the formation of a 1808 rotation twin. The presented technology should also be suitable to prepare a new type of semiconductor heterostructures based on Si polytypes. q 2005 Elsevier Ltd. All rights reserved. Keywords: Molecular beam epitaxy; Silicon; Superlattice; Twinning
1. Introduction Changes in crystal symmetry can have a significant impact on the electronic properties, such as band structure. In recent years therefore, new types of heterostructures consisting of only one material constituent in different crystal structures (heteropolytypic structures), such as wurtzite/zinc-blende or lonsdaleite(2H)/diamond(3C) heterostructures, are under discussion, where problems due to different chemical constituents can be avoided [1–5]. Heteropolytypic structures are also under discussion for the most common semiconductor, silicon [1,2,6–8]. The use of 2H/3C, or at least non-cubic/cubic, Si heterostructures should have some advantages, such as absence of alloy scattering and strain-induced defects, combined with properties that can otherwise only be achieved by incorporating SiGe or SiGeC into Si technology. It offers therefore, the potential of increased flexibility in the design of future Sibased electronic devices. Moreover, the novel properties of such structures could open new fields of Si applications, such as resonant tunnelling or other heterojunction devices. Hexagonal silicon is also an indirect semiconductor. The energy gap reduces with the hexagonality from 1.2 eV * Corresponding author. Fax: C49 511 762 5051. E-mail address: [email protected] (A. Fissel).
0026-2692/$ - see front matter q 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.mejo.2005.02.064
(6H–Si) to about 1 eV(2H–Si) [6]. In hexagonal polytypes, the threefold-degenerate valence-band maximum of cubic Si splits into a twofold and a lower onefold splittoff band. Thereby, the crystal-field splitting increases with the hexagonality [7]. The polytype combination 3C/2H leads to type-I heterostructure, while the combination 3C/4H and 3C/6H are of type-II, with electrons localization in the cubic parts and holes localization in the hexagonal parts. The electronic heteropolytypic structure is governed by the valence-band discontinuity, varying between 135 meV for 3C/6H and 235 meV for 3C/2H [7].
2. Growth of heteropolytypic Si structures The preparation of closed-packed non-cubic Si, however, is still a challenge. The transformation of 3C–Si crystals into the 2H-structure under high pressure and heat treatment was already reported more than 40 years ago [9]. The generation of hexagonal Si was often observed during indentation experiments [10,11]. Small hexagonal Si crystals were prepared by laser ablation [12] and laser beam annealing of amorphous Si thin films [13]. But so far, hexagonal Si could not be obtained in an epitaxial growth process. Even more challenging is the growth of heteropolytypic Si structures, because it demands defined growth conditions for each of the polytypes forming the structure, considering also the high energy of stacking fault formation in Si [14]. A change
A. Fissel et al. / Microelectronics Journal 36 (2005) 506–509
of stacking order can therefore, only be achieved by modification of the growth process, and in particular the surface structure. In case of Si(111) homoepitaxial MBE growth the influence of 1/3 monolayer (ML) subsurface B on the orientation of epitaxial Si layers has been reported [15,16]. Based on that, first attempt to prepare Si-based TSL has been reported based on B-mediated epitaxial growth [17]. However, in these experiments, segregation pffiffiffi pffiffiB-surface ffi has been used to induce ð 3 ! 3ÞR308 reconstructed Si(111) surfaces by high-temperature(HT) annealing (between 900 and 1250 8C) of highly B-doped Si(111) substrates. The growth experiments were performed on 0.38 misoriented Si(111) substrates leading to short terraces. Only after very HT-annealing (1250 8C) larger terraces were formed by step bunching. However, the superstructure periodicity was found to be not adjustable. Because of the inchoate way for obtaining the B surface coverage and the high temperatures used in the experiments, such a preparation is not suitable for any technological application in a Si-polytypes production process. Furthermore, the influence of other disturbing effects like B-clustering [18], oxygen diffusion or SiC formation along the twin formation can be not excluded. In this investigation, we report about the stacking conversion in molecular beam epitaxy (MBE) of Si/Si(111) leading to TSL by B deposition and mediumtemperature annealing. We demonstrate that the distance between the stacking faults can be controlled by the amount of Si MLs grown in each growth run. The structure formation strongly depends on the B surface coverage, the post-growth annealing temperature and the growth mode.
3. Experimental Si layers were grown by solid-source MBE on 4-inch n-type Si(111) (phosphor 1017 cmK3) wafer with a miscut of !0.18, prepared by standard wet-chemical RCA cleaning. Si was evaporated by electron beam evaporation and the flux was monitored by quadrupole based mass spectrometry. B was evaporated from high-temperature effusion cell at temperatures around 1800 8C, corresponding to a deposition rate of 0.05 ML/min. The growth and B-induced superstructure formation were investigated in situ by reflection high-energy electron reflection (RHEED). The layers were investigated ex situ by transmission electron microscopy (TEM) and m-Raman spectroscopy.
507
4. Results After annealing at 500 8C, the samples exhibit the known (7!7) surface reconstruction, (Fig. 1(a)). Exposing the surface pffiffiffi pto ffiffiffi B at 780 8C results in a gradual transition to a ð 3 ! 3ÞR308 surface superstructure (Fig. 1(b)), which already appears in the RHEED pattern at coverage around 0.1 ML. Here, B-atoms occupy the lowest-energy subsurface site (S5, Fig. 2) at TO750 8C [15]. That is a substitutional site in the second layer of the surface with a Si adatom directly above. After cooling down to 400 8C, 2.5 nm Si were grown with a rate of 0.1 nm/min on top of the reconstructed surface, resulting in a change of surface structure to (1!1) (Fig. 1(c)). p Then ffiffiffi pthe ffiffiffi samples were annealed at 780 8C and the ð 3 ! 3ÞR308 structure re-occurs indicating B-surface segregation (Fig. 1(d)). On top of this new surface, Si was grown again at 400 8C. The procedure was repeated six times. The next set of samples was grown on surfaces covered by 1/3 ML boron. At this coverage, the superstructure was full developed, as revealed by the maximum in 1/3 fractional order diffraction spot intensity. Further preparation was performed in the way described above. The last set of samples was prepared by stabilizing the 1/3 ML B during each annealing step by an excess of B deposition. The resulting layer structure was investigated by TEM. Layers grown on surfaces with initially B-coverage below 0.2 ML do not show any change in the stacking order. Stacking faults can also be introduced by surface contaminations, thus this results demonstrate that contamination effect can be neglected in our investigations. Fig. 3(a) shows a cross-sectional (X)TEM micrograph taken from a layer where initially 1/3 ML B was deposited and for the subsequent growth further B-surface coverage was realized only by annealing after each growth step. The micrograph shows a single stacking fault at the interface between the substrate and the layer. A repeated change in stacking direction was found partially only in some regions, what can be due to inhomogeneous B-segregation, as observed also in Si MBE [19]. That is also shown in the plan-view TEM image (Fig. 3(b)). The regions of homogeneous layer structure are in the range of some hundred nm. The selected area diffraction pattern (SAD) shown in the inset exhibits 6 bright and sharp spots in a 6-fold symmetry corresponding to {220} reflection of the Si[111] single-crystal direction. Some additional slight spots are visible in a distance of 1/3 to the main spots. These spots can be identified as 1/3{422}
pffiffiffi pffiffiffi Fig. 1. RHEED pattern of (a) initially clean (7!7) reconstructed Si(111) surface; (b) B-induced ð 3 ! 3ÞR308 surface structure after B deposition at 780 8C; (c) (1!1) surface structure after deposition of 4-Si double layers at 400 8C; (d) after annealing at 780 8C for of 10 min.
508
A. Fissel et al. / Microelectronics Journal 36 (2005) 506–509
clearly that stacking conversation is due to Si nucleation in twin position. Stabilizing the 1/3 ML B-surface coverage by an excess of B after each annealing step results in the formation of TSL in several regions with lateral dimensions in the order of 1 mm. Such TSL structure is shown in Fig. 4, where a XTEM micrograph of a TSL with eightfold periodicity (corresponding to the deposited of 2.5 nm Si) is presented together with a structural model of the TSL period. The structure presented in Fig. 4 has been investigated by m-Raman spectroscopy in comparison to untwined Si to get additional information about structural changes due to the artificially stacking of the Si bilayers. Two Raman active modes located at 502 and 520 cmK1 should be expected for hexagonal silicon [20]. However, only the signal at 520 cmK1 was found, indicating either that no change in the crystal structure had occurred or the number of lonsdaleite-like bonds is to small. pffiffiffi pffiffiffi Fig. 2. Illustration of adatom positions on the B-induced ð 3 ! 3ÞR308 Si(111) surface superstructure (a) top-view; (b) side-view showing B in the subsurface S5 site.
5. Summary and discussion
reflections, which are forbidden for the fcc structure. The appearance of these extra spots can be explained by stacking faults (rotation twins) occurring parallel to the Si(111) planes. Using off-oriented samples, no change in the stacking order were found under the same conditions. In this case step-flow dominates the growth process, indicating
MBE homoepitaxial growth of Si on subsurface pffiffiffi pffiffiffi B-induced ð 3 ! 3ÞR308 Si(111) surface have been investigated, which was prepared by B deposition. It has been shown that Si growth on such a surface containing 1/3 ML B introduces a rotation twin at the layer-substrate interface, indicating that the stacking fault energy becomes
Fig. 3. (a) XTEM micrograph of a layer grown at initially 0.3 ML B-coverage and subsequent B-coverage by boron surface segregation after each Si growth run; (b) XTEM micrograph with lower magnification; (c) plan-view image with SAD pattern obtained in [111] direction exhibiting 1/3{422} superstructure spots due to twinning.
Fig. 4. (a) High-resolution XTEM micrograph of TSL with eightfold periodicity; (b) crystal structure model of a TSL sequence.
A. Fissel et al. / Microelectronics Journal 36 (2005) 506–509
negative in the presence of 1/3 ML subsurface B likely induced by strain within the pSi-B surface layer. ffiffiffi palloyed ffiffiffi Repeating several times the ð 3 ! 3ÞR308 surface preparation by annealing and B deposition to stabilize the 1/3 ML coverage and subsequent Si growth with defined thickness results in the formation of a TSL in which twin boundaries appear periodical along the [111] growth direction. By such a two-step B-mediated epitaxial growth the preparation of close-packed non-cubic Si as well as noncubic/cubic Si heterostructures should be possible. However, already such TSL should have some new interesting properties as suggested recently [21]. It was predicted that both the conduction and the valence band split into minibands separated by energies in the range of 100 meV. Moreover, it has also been suggested that the stacking fault planes will act as potential barriers to the carrier diffusion rather than as carrier recombination centres. The barrier height of the stacking faults was estimated to be 55 meV from temperature dependent electron beam induced current contrast [22]. Therefore, TSL already would offer almost as much versatility in tailoring the electronic miniband structure as there exists in ordinary heterostructure-based superlattices. Such possibility for ‘band-structure engineering’ could even more be enhanced when combining with material composition and doping variations. Besides the investigations of TSL properties, the preparation of a really new structure type of Si will be a subject of further investigations. From recent RHEED and scanning tunnelling microscopy (STM) investigations it can be concluded [16], that the preparation of structures containing twins with distances of two bilayers is possible. That would results in the formation of a hexagonal 4H–Si polytype.
Acknowledgements The authors like to thank Prof. Karl-Ludwig Oehme and his group from the University of Jena for performing mRaman measurements and Michael Seibt from the University of Go¨ttingen for using the TEM equipment.
References [1] M. Murayama, T. Nakayama, Electronic structures of hetero-crystalline semiconductor superlattices, J. Phys. Soc. Jpn. 61 (1992) 2419– 2433; M. Murayama, T. Nakayama, Chemical trend of band offsets at wurtzite/zinc-blende heterocrystalline semiconductor interfaces, Phys. Rev. B 49 (1994) 4710–4724.
509
[2] Ch.-Y. Yeh, Z.W. Lu, S. Froyen, A. Zunger, Zinc-blende-wurtzite polytypism in semiconductors, Phys. Rev. B 46 (1992) 10086–10097. [3] F. Bechstedt, P. Ka¨ckell, Heterocrystalline structures: new types of superlattices? Phys. Rev. Lett. 75 (1995) 2180–2183. [4] S.-H. Ke, J. Zi, K.-M. Zhang, X.-D. Xie, Electronic structures and band offsets of heterocrystalline superlattice, Phys. Rev. B 54 (1996) 8789–8793. [5] A. Fissel, Artificially layered heteropolytypic structures based on SiC polytypes: molecular beam epitaxy, characterization and properties, Phys. Rep. 379 (2003) 149–255. [6] C. Persson, E. Janzen, Electronic band structure in hexagonal closepacked Si polytypes, J. Phys. Condens. Matter 10 (1998) 10549– 10555. [7] C. Raffy, J. Furthmu¨ller, F. Bechstedt, Properties of hexagonal polytypes of group-IV elements from first-principles calculations, Phys. Rev. B 66 (2002) 075201/1–075201/10. [8] S.Q. Wang, H.Q. Ye, First-principle study on the lonsdaleite phases of C, Si and Ge, J. Phys. Condens. Matter 15 (2003) L197–L202. [9] R.H. Wentorf, J.S. Kasper, New forms of silicon, Science 139 (1963) 338. [10] A. Kailer, Y.G. Gogotsi, K.G. Nickel, Phase transformation of silicon caused by contact loading, J. Appl. Phys. 81 (1997) 3057–3063. [11] V.G. Eremenko, V.I. Nikitenko, Electron-microscope investigation of microplastic deformation mechanisms of silicon by indentation, Phys. Status Solidi A 14 (1972) 317–330. [12] Y. Zhang, Z. Iqbal, S. Vijayalakshmi, H. Grebel, Stable hexagonalwurtzite silicon phase by laser ablation, Appl. Phys. Lett. 75 (1999) 2758–2761. [13] J.H. Kim, J.Y. Lee, Hexagonal silicon formation by pulsed laser beam annealing, Mater. Lett. 27 (1996) 275–279. [14] P. Ka¨ckell, J. Furthmu¨ller, F. Bechstedt, Stacking faults in group-IV crystals: an ab initio study, Phys. Rev. B 58 (1998) 1326–1330. [15] R.L. Headrick, B.E. Weir, J. Bevk, B.S. Freer, D.J. Eaglesham, L.C. Feldman, Influence of surface reconstruction on the orientation of homoepitaxial silicon films, Phys. Rev. Lett. 65 (1990) 1128–1131. [16] H. Hibino, T. Ogino, Formation of twinned two-bilayer-high islands during initial stages of Si growth on Si(111)O3!O3-B, Surf. Sci. 412/413 (1998) 132–140. [17] H. Hibino, K. Sumimoto, T. Ogino, Twinned epitaxial layers formed on Si(111)O3!O3-B, J. Vac. Sci. Technol. A 16 (1998) 1934–1937. [18] F. Thibaudau, T.P. Roge, Ph. Methiez, P. Dumas, F. Salvan, Cooperative segregation of boron at Si(111), Europhys. Lett. 25 (1994) 353–356. [19] D. Kru¨ger, G. Lippert, R. Kurps, H.J. Osten, Lateral inhomogeneous boron segregation during silicon thin film growth molecular beam epitaxy, J. Cryst. Growth 135 (1994) 246–252. [20] R.J. Kobliska, S.A. Solin, Raman spectrum of wurtzite silicon, Phys. Rev. B 8 (1973) 3799–3802. [21] Z. Ikonic, G.P. Srivastava, J.C. Inkson, Electronic properties of twin-boudaries and twinning superlattices in diamond-type and zinc-blende-type semiconductors, Phys. Rev. B 48 (1993) 17181– 17193. [22] T. Sekiguchi, B. Shen, T. Watanabe, K. Sumino, EBIC study on the electrical activity of stacking faults in silicon, Mater. Sci. Eng. B 42 (1996) 235–239.
