ARTICLE IN PRESS Physica B 404 (2009) 4708–4711
Contents lists available at ScienceDirect
Physica B journal homepage: www.elsevier.com/locate/physb
Laser-induced melting and recrystallization of CVD grown polycrystalline Si/SiGe/Ge layers P.I. Gaiduk a,, S.L. Prakopyeu a, V.A. Zajkov a, G.D. Ivlev b, E.I. Gatskevich b a b
Belarusian State University, Minsk, Belarus Institute of Physics, NASB, Minsk, Belarus
a r t i c l e in f o
a b s t r a c t
Keywords: SiGe/Ge Laser recrystallization Microstructure TEM
SiGe/Ge layers were deposited by CVD on either Si or Si/SiO2 substrates and were subjected with pulsed laser annealing (LA). In situ measurements of time-resolved reflectivity revealed strong dependence of melting time on both energy density and type of the substrate. Depending on laser energy density and on type of substrate, initially microcrystalline and amorphous as-deposited layers were transformed to polycrystalline with different morphology and size of the grains. The results are discussed within the model of liquid-phase recrystallization. & 2009 Elsevier B.V. All rights reserved.
1. Introduction Silicon-based heterostructures (Si/SiGe) have attracted much attention in recent years, mainly because of continued progress in epitaxial-growth technology and the ability to fabricate devices compatible with silicon-based integrated circuit technology [1,2]. Thin layers of polycrystalline silicon–germanium (poly-SiGe) alloys are attractive for thin-film electronic devices (e.g. thin film transistors—TFT), photodiodes, infrared photodetectors and next generation of solar cells [3–5]. In the latter devices, one of the main advantages of alloying Ge with Si is the enhanced optical absorption of the alloy in comparison to pure Si. It is well known that crystalline silicon (c-Si) solar cells have insufficient optical absorption coefficient. Thus, effective absorption of the sunlight in a c-Si solar cell requires deposition of rather thick layer. The use of optimized light trapping structures can strongly reduce the necessary absorber thickness. One possible method to efficient c-Si based solar cells has been the use of microcrystalline-Si (mc-Si). Such thin film solar cells based on mcSi/Si tandem structures have achieved efficiencies up to 11% [6]. Major problems remaining for the use of mc-Si as the absorber material in thin-film Si-based solar cells are the low absorption coefficient as well as the low growth rates of high-quality mc-Si layers in PECVD. Here, the use of strongly absorbing crystalline SiGe and Ge thin films with a suitably adjusted Ge content may provide alternative solutions for the near future. Laser recrystallization of SiGe and Ge thin films seems to be a promising method for modification of electrical and optical properties of the layers. In particular, pulsed laser treatment of Corresponding author. Tel.: +375 17 278 97 00; fax; +375 17 278 04 17.
E-mail address:
[email protected] (P.I. Gaiduk). 0921-4526/$ - see front matter & 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.physb.2009.08.127
Fig. 1. RBS spectra obtained from as grown layer Si/SiGe/Ge/cup-Si (line) and after LA at 0.9 J/cm2 (triangles).
SiGe/Ge alloy layers has been proven to be useful for application in strain engineering of high mobility MOSFETs and HBTs [7–9]. Some studies of laser crystallization have been recently done to improve structural and optical properties of deposited polycrystalline SiGe/Ge layers [10,11], also for solar cell and photodetector applications.
ARTICLE IN PRESS P.I. Gaiduk et al. / Physica B 404 (2009) 4708–4711
2. Methodical In the present study, thin microcrystalline/amorphous SiGe/Ge layers were deposited by CVD on (0 0 1)-Si or on (0 0 1)-Si/ (540 nm)-SiO2 wafers at temperature 470 1C. A commercially available Izotron machine was used for deposition. Thirty nm layer of pure Si followed by 50 nm layer of SiGe were first grown. This SiGe layer was deposited in a compositionally grading manner: a concentration of Ge increases from 0 to 100 at% during
Fig. 2. Time-resolved reflectivity at 1064 nm (1,2) and at 532 nm (3) of Si/SiO2/ SiGe/Ge (1,3) and Si/SiGe/Ge (2) during LA with 0.9 J/cm2. Arrows indicate two steps of reflectivity raise.
4709
deposition of this SiGe layer. The aim of SiGe deposition in a composition grading manner was to avoid Ge island formation on Si and make sure smooth surface of deposited layers. Then, about 250 nm of pure Ge was grown. To prevent possible thermal
Fig. 3. A duration of melting versus laser energy density in different structures: 1,2—Si/SiO2/SiGe/Ge, 3–5—Si/SiGe/Ge. 1,3—surface was additionally covered with 2 nm cup Si layer. Sample 5 is MBE grown sample. 6—pure Si (no Ge on the surface).
Fig. 4. Bright-field TEM images of the structure of Si/SiGe/Ge samples after deposition and laser annealing at different energy density. (a) As grown; (b) 0.73 J/cm2; (c) 0.9 J/ cm2; and (d) 1.4 J/cm2.
ARTICLE IN PRESS 4710
P.I. Gaiduk et al. / Physica B 404 (2009) 4708–4711
Fig. 5. BF TEM images of the structure of Si/SiO2/SiGe/Ge samples after deposition and laser annealing at different energy density. (a) 0.58 J/cm2; (c) 0.73 J/cm2; and (d) 0.9 J/cm2. Cup Si layer was deposited on samples (a) and (c). Indicated by arrows in (b) are explosively crystallized grains.
decomposition of the layers due to Ge oxidation and evaporation, the top of the sample was finally covered with a cup Si layer (2– 20 nm). For comparison, at a number of samples this cup Si layer was not deposited; also some other samples of Si/SiGe/Ge were grown by molecular-beam epitaxy (MBE). The samples were then exposed with pulsed ruby laser (l ¼ 694 nm, a pulse length of 80 ns, an average energy density W ¼ 0.2–3 J/cm2). The spatial variation in the density of laser energy did not exceed 75% over the laser spot of 4.5 mm in diameter. The process of laser annealing (LA) was investigated in situ by time-resolved reflectivity measurements. Two beams of the probing light of a Nd glass laser (1064 and 532 nm) were focused into a center of LA spot at an incidence angle of 401. The experimental setup was described in detail elsewhere [12]. The as-annealed samples were finally investigated ex situ. Structural properties were studied by transmission electron microscopy (TEM). EM-125 device operating at 100 kV was used. Composition of the SiGe alloy layers was measured by using Rutherford backscattering spectrometry (RBS).
3. Results and discussion Fig. 1 demonstrates comparative RBS spectra, obtained from as grown and laser annealed samples. It can be concluded from the
spectra that LA at W ¼ 0.9 J/cm2 does not result in neither evaporation nor depth redistribution of Ge. On the other hand, exposition of deposited SiGe/Si layers with laser pulse of energy density 0.9 J/cm2 results in melting of surface layers for 100– 600 ns as it can be seen from time-resolved reflectivity (TRR) measurements (Figs. 2 and 3). Analysis of the TRR data allows concluding the following. First, the duration of melting of the surface layer strongly depends on substrate type. Compositionally graded SiGe/Ge grown on Si/SiO2 substrate are melted at lower energy density and melt keeps much longer as compared to Si substrate (Fig. 3). This evidently is due to considerably lower thermal conductivity of SiO2 substrate as compared to Si one (1.5 W/cm/K for Si against 0.014 W/cm/K for SiO2 [13,14]). Additional evidence of lower thermal conductivity is slower solidification rate in the case of Si/ SiO2/SiGe/Ge structure (compare back fronts of curves 1 and 2 in Fig. 2). Second, presence of thin Si cup layer on the surface increases the melting time in both Si/SiO2/SiGe/Ge and Si/SiGe/Ge structures. Probably, this fact reflects lower loss rate of the heat due to anti-reflectance ability of the capping layer. Finally, at probing wavelength 532 nm (Fig. 2, curve 3), the reflectivity rises in two steps, separated with a short (about 25 ns) plateau. Such feature manifests itself at a relatively high LA energy density and, apparently, is due to sequential melting of Ge and Si surface layers: Ge has a lower melting temperature than Si (970 and 1412 C, respectively). A good correlation of time-resolved reflectivity and structural investigations is documented in this study. Fig. 4 shows representative set of bright-field (BF) TEM images of microstructure of Si/SiGe/Ge samples after deposition and laser annealing with different energy density. It is found from electron diffraction patterns (not shown here) that the surface layer of as grown sample consist of mixture of amorphous and polycrystalline SiGe/Ge material. A grain size of polycrystalline material varies between 10 and 50 nm. Laser annealing with energy density of 0.73 J/cm2 results in complete annealing of amorphous phase and slight increase of average grain size (Fig. 4b). Further rise of energy density is accompanied by strong change of both size and morphology of the grains (Fig. 4c,d). In particular, the average grain size increases to 50–200 nm after LA at 0.9 J/cm2 and to about 0.5–1 mm after LA at 1.4 J/cm2 (here the size was measured in plan-view section). This structural evolution correlates well with the lifetime of melted surface as it can be seen from comparison of the data, presented in Figs. 3 and 4. It is obviously found that laser induced structural changes are significantly stronger in the samples, which use less thermally conductive Si/SiO2 substrate (Fig. 5). In particular, SiGe/Ge layers deposited on Si/SiO2 substrate and annealed at 0.73 J/cm2 have polycrystalline structure with grain morphology and size, similar those in layers, deposited on pure Si substrate and annealed at much higher (1.4 J/cm2) energy density (compare Figs. 4d and 5b). In addition to enlargement of the grains, a spherulit-like morphology of the structure (indicated by arrows in Fig. 5b) might be pointed out which is usually indicate lateral crystallization in conditions of low number of crystalline nucleolus and fast crystal growth. Such dissimilarity in structural morphology is clearly due to different thermal conductivity of Si/SiO2 and Si substrates. However, a certain role of different nucleation rate at SiO2 and at Si substrates should be taken into account. Deposition of additional cup Si layer on the surface of Si/SiO2/ SiGe/Ge graded structure strongly influence on laser induced recrystallization of polycrystalline layer. A good example follows from comparison of BF TEM images in Fig. 5b and c. Specifically, a visible contradiction is that increase of energy density from
ARTICLE IN PRESS P.I. Gaiduk et al. / Physica B 404 (2009) 4708–4711
0.73 J/cm2 (Fig. 5b) to 0.9 J/cm2 (Fig. 5c) results in lower average grain size. This contradiction might be disclosed in terms of higher density of nucleation centers for recrystallization in the case of surface cap Si layer. We believe that due to different melting temperatures of Ge and Si, melting of graded SiGe/Ge layer takes place first in between two solid layers of Si. After cooling, each of these Si layers initiate solidification fronts. It is well known [15,16] that laser resolidification is extremely fast and occurs from undercooled melt. Thus, resolidification is determining by both undercooling value and number of nucleus centers. A crucial evidence of nucleus deficiency in the case of uncapped layers is an existence of explosive crystallization: one typical example of spherulite-like area of explosively crystallized material is indicated by arrows in Fig. 5b.
substrate. The results are discussed within the model of liquidphase recrystallization.
Acknowledgments This work was supported by Grant nos. 20080834 and 20061216 of Belarusian Research Foundations.
References [1] [2] [3] [4] [5] [6]
4. Conclusion In situ measurements of time-resolved reflectivity (TRR) and ex situ TEM investigations were carried out to study pulsed laser recrystallization of SiGe/Ge layers, CVD deposited on Si or Si/SiO2 substrates. Strong dependence of melting time on both energy density and type of the substrate are found. According to TEM, microcrystalline/amorphous as-deposited layers recrystallize to polycrystalline state of different morphology and grain size which depends on both energy density of laser pulse and on type of
4711
[7] [8] [9]
[10] [11] [12] [13] [14] [15] [16]
E. Kasper, Appl. Surf. Sci. 254 (2008) 6158. I.Z. Mitrovic, et al., Solid-State Electron. 49 (2005) 1556. G. Isella, et al., Semicond. Sci. Technol. 22 (2007) S26. A. Yamada, M. Sakuraba, J. Murota, Thin Solid Films 508 (2006) 399. L. Colace, et al., IEEE Photonic Tech. Lett. 19 (2007) 1813. A.V. Shah, et al., in: Proceedings of the 17th European Photovoltaic Solar Energy Conference, Munich, 2001, p. 2823. G.H. Wang, et al., Appl. Phys. Lett. 91 (2007) 202105. C.N. Chle irigh, et al., J. Appl. Phys. 103 (2008) 104501. G. Lorito, V. Gonda, T.L.M. Scholtes, L.K. Nanver, in: 26th International Conference on Microelectronics, 11–14 May 2008, Nis, Ieee, Piscataway, NJ, 2008, p. 291. C. Eisele, et al., Thin Solid Films 427 (2003) 176. M. Weizman, et al., J. Non-Cryst. Solids 352 (2006) 1259. E.I. Gatskevich, G.D. Ivlev, A.M. Chaplanov, Quantum Electron. 25 (1995) 774. C.J. Glassbrenner, G.A. Slack, Phys. Rev. 134 (4A) (1964) A1058. /http://www.virginiasemi.com/pdf/generalpropertiesSi62002.pdfS. J.A. Yater, M.O. Thompson, Phys. Rev. Lett. 63 (1989) 2088. M.J. Aziz, C.W. White, Phys. Rev. Lett. 57 (1986) 2675.