Thin Solid Films 520 (2012) 1784–1788
Contents lists available at SciVerse ScienceDirect
Thin Solid Films journal homepage: www.elsevier.com/locate/tsf
Polycrystalline silicon ﬁlms on glass grown by amorphous–liquid–crystalline transition at temperatures below 330 °C Robert Heimburger ⁎, Nils Deßmann 1, Thomas Teubner, Hans-Peter Schramm, Torsten Boeck, Roberto Fornari Leibniz-Institut für Kristallzüchtung Max-Born-Straße 2, 12489, Berlin-Adlershof, Germany
a r t i c l e
i n f o
Article history: Received 23 November 2010 Received in revised form 22 August 2011 Accepted 23 August 2011 Available online 31 August 2011 Keywords: Metal-induced crystallization Silicon on glass Solution growth Amorphous–liquid–crystalline transition Physical vapor deposition Raman spectroscopy Scanning electron microscopy
a b s t r a c t An approach to deposit polycrystalline silicon layers on amorphous substrates is presented. It is shown that metastable amorphous silicon can be transformed into its more stable microcrystalline structure at a temperature below 330 °C via an intermediate liquid solution stage. In particular, the interaction of liquid indium nanodroplets in contact with amorphous silicon is shown to lead to the formation of circular polycrystalline domains. Crystallinity of these domains is analyzed by micro-Raman spectroscopy. The droplet size necessary for the onset of crystallization is related to the temperature of the ﬁlm. Full coverage of the substrate with microcrystalline silicon has been obtained at 320 °C within less than one hour. These ﬁlms might act as seeding layers for further enlargement by steady-state liquid phase epitaxy. © 2011 Elsevier B.V. All rights reserved.
2. Experimental details
The growth of closed and high quality polycrystalline silicon ﬁlms on glass still remains an unsolved problem. To overcome the inherent problem that glass provides no crystalline template for epitaxy, deposition techniques consist of two distinct steps. First, an arrangement of separated microcrystals or a polycrystalline seeding layer is grown, that is afterwards epitaxially enlarged. We report on a technique to grow such a microcrystalline seeding layer at low temperature. By introducing an intermediate liquid solution stage, our process takes advantage of the fact, that the nucleation barrier for the phase transition of amorphous silicon (a-Si) to microcrystalline silicon (μc-Si) is signiﬁcantly reduced by the presence of the liquid catalyst. This effect has also been utilized for growth of in-plane nanowires . In contrast, our work is focused on fabrication of a microcrystalline layer, which might be sufﬁcient for further epitaxial enlargement by means of steady-state solution growth [2,3]. Growth of μc-Si on glass has been achieved at temperatures below 330 °C, which comprises the potential of cost-efﬁcient application in the ﬁeld of thin ﬁlm solar cells.
Borosilicate glass substrates, each with a size of 4 × 4 cm 2, were cleaned in acetone with a subsequent ultrasonic treatment in an alkaline detergent (LM2). Afterwards, the samples were rinsed in deionized water and dried in nitrogen ﬂow. The cleaned substrates were coated with molybdenum (5 nm at 0.03 nm/s) and silicon (400 nm or 800 nm at 0.12 nm/s) by means of electron beam evaporation. The base pressure inside the deposition chamber did not exceed 2 × 10 − 4 Pa and the samples were not heated during deposition. Afterwards, the sample temperature was increased to the annealing temperature Θa within 30 min using radiant heating from four tungsten–halogen lamps. Then, deposition of about 25 nm indium from a resistively heated ceramic crucible was started. Indium deposition was carried out at a constant rate of 0.01 nm/s in all experiments. Before ex situ examination by scanning electron microscopy (SEM) and micro-Raman spectroscopy, indium was removed from the sample using aqua regia. The FEI Nova 600 NanoLab SEM was operated in secondary electron mode with its primary electron beam accelerated at 5–10 kV. The Raman spectra were recorded at room temperature using a LabRAM HR 800 spectrometer. Light of a He– Ne laser operating at 633 nm was focused through an objective lens (magniﬁcation: 100×, numerical aperture: 0.90, spot diameter approximately 1 μm) and used for excitation. To prevent laser-induced crystallization during the measurements, the beam intensity was attenuated to a power estimated below 0.5 mW. The intensity level
⁎ Corresponding author. Tel.: + 49 30 6392 3055; fax: + 49 30 6392 3003. E-mail address: [email protected]
(R. Heimburger). URL: http://www.ikz-berlin.de (R. Heimburger). 1 Present address: Brandenburgische Technische Universität Cottbus, KonradWachsmann-Allee 1, 03046 Cottbus, Germany. 0040-6090/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2011.08.084
R. Heimburger et al. / Thin Solid Films 520 (2012) 1784–1788
leading to crystallization was checked with separate a-Si samples. The crystallinity of our samples was studied in the vicinity of a domain boundary by stepwise movement of the sample (2 μm/step) from an outer part into the domain. Another sample has been breached in order to study its cross section by means of SEM. The inﬂuence of the annealing temperature on the size of indium nanodroplets necessary for the onset of the amorphous–liquid–crystalline (ALC) transition was studied between 266 and 328 °C. For that purpose, indium was deposited at a constant rate of 0.01 nm/s, but now both, indium deposition and sample heating, were stopped immediately after the ﬁrst μc-Si domains formed. SEM images of indium droplets in direct vicinity of these domains were analyzed by an image processing software . In order to ensure reproducibility, all images were processed in the same way. A Gaussian blur ﬁlter (standard deviation: 1.50), Huang white Auto Threshold detection, watershed separation and a particle analysis was utilized to determine the radius of each individual droplet . The size distribution of at least 4000 droplets was plotted as a histogram and their average size was determined by ﬁtting a Gaussian distribution curve. 3. Calculation To gain a closer insight into the process, we calculated the equilibrium solubility of a-Si and c-Si (crystalline silicon) in a liquid indium nanodroplet. At equilibrium, the chemical potential of silicon in the liquid solution and the crystalline solid is the same: l
μ Si ¼ μ Si :
is an entropy factor measured in kJ/(mol K) which arises due to the volume change during mixing  E
g ¼ ða−bT Þxð1−xÞ:
2σ : r
To obtain an expression for the chemical potential of silicon dissolved in the metallic nanodroplet μ lSi , we ﬁrst consider the molar Gibb's free energy of the liquid nano-alloy g l as  l
g ¼ xμ Si þ ð1−xÞμ In þ RT ðx ln x þ ð1−xÞ ln ð1−xÞÞ þ g þ g :
The variables are denoted as follows: x — molar fraction of silicon in the solution, T — temperature, g E — molar excess Gibb's free energy of mixing, g S — molar Gibb's free energy due to capillary pressure inl;0 side the droplet. μ l;0 Si and μ In are the chemical potentials of pure liquid silicon and pure liquid indium. Within the approximation of a quasi-regular solution , g E can be written in terms of two parameters: a denotes the caloric interaction parameter measured in kJ/mol and accounts for the heat effect and b
Thus, the surface term can be written in the form of Eq. (6). S
2σ VM r
σ denotes the speciﬁc surface energy and VM is the molar volume of the solution. As an approximation, the surface energy and the molar volume of pure indium are used in the calculation. The chemical potential of the silicon component in the solution μSil can be derived by applying the slope interception method  l
μ Si ¼ g þ ð1−xÞ
∂g l : ∂x
Using this relation, we obtain l
μ Si ¼ μ Si
Assuming a hemispheric shape of the droplets with radius r, the capillary pressure is given by the Young–Laplace equation
As indium is soluble in solid silicon only at dopant level, solid solubility can be neglected. Thus the chemical potential of silicon in the solid state can be identiﬁed with the chemical potential of the pure element.
μ Si ¼ μ Si þ RT ln x þ ða−bT Þð1−xÞ þ
2σ VM : r
Combining Eqs. (1), (2) and (8) and following the approach of Bennema et al. , we ﬁnally obtain an expression for the equilibrium solubility of silicon in a liquid nanodroplet as ln x ¼
Δl;s h R
! 1 ða−bT Þð1−xÞ2 1 2σ VM − : − T RT r RT
Here, Δ l, sh* and Tm, * are the molar heat of fusion and the melting temperature of silicon and the asterisk denotes a-Si or c-Si, respectively. Some conventions are necessary to relate discrete values for Δ l, sh∗ and Tm, * to amorphous silicon. a-Si and c-Si form a monotropic system with a-Si being the metastable form of silicon carrying an inherent driving force for crystallization, which is of the order of 0.11 eV/atom . The authors are aware of the fact that a-Si will transform into c-Si, whenever a proper kinetic path is provided. Nevertheless, by very fast heating using pulsed laser irradiation, Grimaldi et al. observed the melting of a-Si without crystallization . The molar heat of fusion and the melting temperature has been determined by Donovan et al. . Using data of Table 1 and Eq. (9), the maximum solubility of a-Si in indium can be evaluated.
Table 1 Physicochemical properties of the silicon–indium system used to calculate equilibrium solubilities (Θ — temperature). Variable
Δl;s hðcQSiÞ Δl;s hðaQSiÞ Tm ðcQSiÞ Tm ðaQSiÞ a b σ ρ VM
50.6 kJ/mol 37.2 kJ/mol 1685 K 1460 K 39.2 kJ/mol 0.0071 kJ/mol K 0.568 N/m–9.45 × 10− 5 N/(m K)[Θ–429.75 K] 7.05 × 103 kg/m3–0.776 kg/(m3K)[Θ–429.75 K] 0.1148 kg/mol[1/ρ]
Molar heat of fusion of c-Si Molar hear of fusion of a-Si Melting temperature of c-Si Melting temperature of a-Si Caloric interaction parameter Entropic interaction parameter Surface energy of Indium Mass density of indium Molar volume if indium
       
R. Heimburger et al. / Thin Solid Films 520 (2012) 1784–1788
Fig. 1. a) 45° tilt SEM image of an initial μc-Si cluster near breaking edge, b) schematic view of the ALC-mechanism.
4. Results Indium was found to form nanodroplets at the a-Si surface. Some of them selectively start to dissolve and crystallize the subjacent amorphous silicon layer. This crystallization process was found to proceed by a simultaneous in-plane movement of the solvent droplet. A schematic view of this process is shown in Fig. 1b. While a-Si (chemical potential μa‐Si ) is dissolved at the left margin of the droplet (frontside), μc-Si (chemical potential μμc‐Si ) is segregated at the right margin. As the droplet moves along the surface, it grows by capturing
smaller droplets located at the front side. This enables continuous growth of individual silicon grains with a lateral dimension of several micrometers. Their depth perpendicular to the surface did not exceed 0.4 μm, thus leaving a thin and presumably amorphous ﬁlm between μc-Si and the substrate which is seen in Fig. 1a. SEM micrographs in Fig. 2 show top views of the morphology produced by the crystallization process as found on one sample after etching of indium. Fig. 2a corresponds to the initial stage as illustrated in Fig. 1. While the direction of motion of the initial droplets can be assumed as being stochastic, it becomes constricted for droplets moving in direct vicinity of a
Fig. 2. SEM micrographs taken at one sample after removal of indium showing different phases of crystallization: a) initial μc-Si cluster embedded in amorphous matrix, b) development of a circular μc-Si domain with radial crystallite pattern, c) coalescence of neighboring domains, d) closed μc-Si layer.
R. Heimburger et al. / Thin Solid Films 520 (2012) 1784–1788
two Gaussian curves as shown in Fig. 3c. The FWHM (full width at half maximum) of the narrow contribution amounts to 5.7 cm − 1. Fig. 4 shows the average size of the indium droplets in direct vicinity of a just emerged μc-Si domain (like in Fig. 2a) as a function of sample temperature. At 266 °C, the initial droplet radius necessary for the onset of the ALC transition is equal to (39 ± 9). This value steadily decreases with increasing temperature to (23 ± 2) at 328 °C. The results of the solubility calculation using Eq. (9) and data compiled in Table 1 are shown in the same ﬁgure. For each temperature, one data point has been computed for a-Si and c-Si, each including the mean radius of the initial indium droplets. The capillar pressure (5) inside the droplets shifts the equilibrium solubility by more than 10% towards lower values and thus cannot be neglected in the calculation. In consequence of its higher chemical potential, the solubility of a-Si is about one order of magnitude higher than of c-Si.
ν Fig. 3. Raman spectrum recorded a) outside μc-Si domain, b) crossing of the domain boundary and c) inside the domain (stepwidth: 2 μm).
crystallized region leading to radial spreading. Proceeding crystallization leads to enlargement and coalescence of single domains (Fig. 2a and c). By maintaining indium deposition and radiant heating, full sample coverage with μc-Si as in Fig. 2d has been achieved at 320 °C within less than 1 h. The results of the Raman measurement after removal of indium are shown in Fig. 3. The spectra indicate a sharp change in crystallinity near the boundary of the examined μc-Si cluster. Outside the cluster, we found no evidence for crystalline phases. The broad feature around ν = 470 − 1 indicates the presence of amorphous silicon . When crossing the domain boundary, the longitudinal optical–transverse optical (LO–TO) phonon mode of crystalline silicon becomes visible at ν = 521 cm − 1. The peak exhibits slightly asymmetric broadening towards lower energies and can be ﬁtted by
Due to the higher chemical potential of a-Si, an indium droplet in contact with amorphous silicon necessarily gets supersaturated in relation to the crystalline phase, and a driving force for crystallization is built-up inside the solution. In its magnitude, the highest possible supersaturation is the same as if the solution would had been equilibrated with c-Si at a temperature about 80 °C higher and then cooled to the actual annealing temperature. This leads to the formation and growth of supercritical c-Si nuclei. After this initial nucleation event, new amorphous material is consumed at the front side of the active droplet. Thus supersaturation in the solution is maintained and the process proceeds continuously. The difference of latent heat during crystallization amounts to a;c
Δ h ¼ 13:4 kJ=mol
Release of latent heat may cause a local increase of ﬁlm temperature in the vicinity of a moving droplet and promote further crystallization. As the size of the initial droplets necessary for the onset of the ALC-transition was found to decrease with increasing sample temperature (see Fig. 4), the nucleation probability in a ﬁxed solvent volume increases with temperature. Therefore, the occurrence of new nucleation events is more likely in the vicinity of an already active region and round-shaped polycrystalline domains form as a result (see Fig. 2). The Raman measurements demonstrate the crystalline nature of the grown material. The FWHM of the peak is comparable to values that are obtained for silicon on glass deposited using the aluminuminduced layer exchange with a slight asymmetric broadening that could be attributed to the presence of smaller grains or defects [14,15]. 6. Conclusion A method for low-temperature growth of polycrystalline silicon layers on glass based on solvent-induced crystallization of amorphous silicon has been proposed. By lowering the activation barrier for nucleation, indium mediates crystal growth at temperatures below 330 °C. Full μc-Si coverage of the substrate has been observed within less than one hour at 320 °C. By this method both, process temperature and duration, can be signiﬁcantly decreased compared to wellestablished solid-phase recrystallization processes . Additionally, grains, which show good crystallinity in Raman measurements with their maximum size of several micrometers, are obtained. This opens up a way to fabricate a seeding layer for further epitaxial enlargement by means of steady-state solution growth or other growth techniques.
Θ Fig. 4. Average size of indium droplets at the start of the ALC-transition and equilibrium solubility of a-Si and c-Si in those droplets versus temperature.
References  L. Yu, P.-J. Alet, G. Picardi, P. Roca i Cabarrocas, Phys. Rev. Lett. 102 (2009) 125501.  T. Teubner, R. Heimburger, K. Böttcher, T. Boeck, R. Fornari, Cryst. Growth Des. 7 (8) (2008) 2484.
R. Heimburger et al. / Thin Solid Films 520 (2012) 1784–1788
 R. Heimburger, T. Teubner, N. Deßmann, H.-P. Schramm, T. Boeck, R. Fornari, J. Cryst. Growth 312 (9) (2010) 1632.  M. Abramoff, P. Magalhães, S. Ram, Biophoton. Int. 11 (7) (2004) 36.  Nuclei Watershed Separation, http://paciﬁc.mpi-cbg.de/wiki/index.php/Nuclei_ Watershed_Separation August 2010.  D. Hourlier-Bahloul, P. Perrot, C.R. Chim. 10 (2007) 658.  P. Bennema, J. van Eupen, B. van der Wolf, J. Los, H. Meekes, Int. J. Pharm. 351 (2008) 74.  J.H. Hildebrand, Nature 168 (1951) 868.  P.A. Rock, Chemical Thermodynamics, University Science Books, 1983.  C. Spinella, S. Lombardo, F. Priolo, J. Appl. Phys. 84 (10) (1998) 5383.  M. Grimaldi, P. Baeri, M. Malvezzi, C. Sirtori, Int. J. Thermophys. 13 (1992) 141.
 E. Donovan, F. Spaepen, J. Poate, D. Jacobson, Appl. Phys. Lett. 55 (1989) 1516.  O. Nast, T. Puzzer, L. Koschier, A. Sproul, S. Wenham, Appl. Phys. Lett. 73 (1998) 3214.  D. Dimova-Malinovska, V. Grigorov, M. Nikolaeva-Dimitrova, O. Angelov, N. Peev, Thin Solid Films 501 (1–2) (2006) 358.  T. Antesberger, C. Jaeger, M. Stutzmann, J. Non-Cryst. Solids 354 (19–25) (2008) 2324.  R. Buitrago, G. Risso, M. Cutrera, M. Battioni, L.D. Bernardez, J. Schmidt, R. Arce, R. Koropecki, Int. J. Hydrogen Energy 33 (13) (2008) 3522.  M. Alonso, E. Bauser, J. Appl. Phys. 62 (1987) 4445.  M. McClelland, J. Sze, Surf. Sci. 330 (1995) 313.