A new approach to thin film crystal silicon on glass: Biaxially-textured silicon on foreign template layers

Journal of Non-Crystalline Solids 352 (2006) 984–988 www.elsevier.com/locate/jnoncrysol

A new approach to thin film crystal silicon on glass: Biaxially-textured silicon on foreign template layers Charles W. Teplin *, David S. Ginley, Howard M. Branz National Renewable Energy Laboratory, NCPV, 1617 Cole Blvd., Golden, CO 80401, United States Available online 13 March 2006

Abstract We propose a new approach to growing photovoltaic-quality crystal silicon films on glass. Other approaches to film Si focus on increasing grain size in order to reduce the deleterious effects of grain boundaries. Instead, we propose aligning the silicon grains biaxially (both in and out of plane) so that (1) grain boundaries are ‘low-angle’ and have less effect on the electronic properties of the material and (2) subsequent epitaxial thickening is simplified. They key to our approach is the use of a foreign template layer that can be grown with biaxial texture directly on glass by a technique such as ion-beam-assisted deposition or inclined substrate deposition. After deposition of the template layer, silicon is then grown aligned to the template and subsequently thickened. Here, we outline this new approach to silicon on glass, describe initial experimental results and discuss challenges that must be overcome. Published by Elsevier B.V. PACS: 84.60.Jt; 81.10. h; 81.15. z; 68.55.Jk Keywords: Solar cells; Crystal growth; Nucleation; Photovoltaics; Chemical vapor deposition; Sputtering; Surfaces and interface

1. Introduction Crystalline silicon (c-Si) wafers are the foundation of today’s rapidly growing photovoltaic (PV) industry. Module sales have increased 30–40% per year for the last decade, with steadily increasing energy conversion efficiencies that now range from 14% to 20%. Silicon appears to be the near-ideal PV material – it is abundant, non-toxic, and benefits from an enormous knowledge base and industrial infrastructure. However, the high cost of silicon wafers and ribbons likely places a lower limit to c-Si module cost of $1.00 to 1.50/W of 1 sun electricity generation capacity [1]. This limitation has recently stimulated considerable research into Si-based thin film PV that would eliminate the need for silicon wafers altogether. Although amorphous and nanocrystalline Si thin films on glass have the potential for far lower manufacturing costs per unit *

Corresponding author. Tel.: +1 303 384 6440; fax: +1 303 384 6430. E-mail addresses: [email protected] (C.W. Teplin), [email protected] (H.M. Branz). 0022-3093/$ - see front matter Published by Elsevier B.V. doi:10.1016/j.jnoncrysol.2006.01.024

area [2] the low quality of these semiconductors make the achievement of module conversion efficiencies above 10–12% unlikely. The low sunlight energy conversion efficiencies of thin film silicon-based photovoltaics to-date are primarily due to low carrier mobility and short lifetimes. This has led to research into growing a thicker (2–10 lm) crystal silicon film of higher quality at low cost on inexpensive substrates. Effective light trapping technologies need to be applied in these structures to achieve high PV current densities. A c-Si film technology on a low-cost substrate would allow large-scale manufacturers to realize the cost advantages of thin films while achieving the high efficiency of c-Si PV. Crystal silicon films can be readily grown by depositing amorphous silicon (a-Si:H) on glass and subsequently annealing the films at 550–600 °C for 6–24 h. Unfortunately, the resulting material has small (micron-sized and smaller) randomly oriented grains that limit the minoritycarrier diffusion length, which needs to be about twice the film thickness to obtain good PV efficiency. Thus far, no scalable approach has been demonstrated for growing

C.W. Teplin et al. / Journal of Non-Crystalline Solids 352 (2006) 984–988

c-Si films with uniformly large (>4 lm) or ordered grains. While larger grains improve the electrical properties by simply reducing the number of grain boundaries, increasing the ordering of the grains results in low-angle grain boundaries, which do not degrade the electrical properties as strongly [3]. Such control will be necessary for c-Si film PV with efficiencies that approach that of c-Si wafer PV. Currently there are three potentially low-cost approaches that have been used to improve the electrical qualities of ‘thick’ c-Si films. The first and thus far, the most successful technique has been to employ films with the submicron grains formed by annealing a-Si:H films, but to passivate the grain boundaries with hydrogen after the c-Si has formed [4]. Though this technique has led to prototype modules about 8% in efficiency with c-Si layers about 1 lm thick [4], it is unlikely that this approach will lead to efficiencies comparable to those achieved using wafers. In a second approach, researchers have employed Al induced crystallization through a layer exchange process [5,6] to grow a Si seed layer on glass with grains up to about 10 lm. In this approach, an amorphous silicon layer is grown on top of a metallic aluminum layer. During annealing, it is found that the silicon and aluminum exchange positions and that the silicon has crystallized with relatively large grains. After removing the Al top layer, the thin c-Si seed layer is thickened with direct epitaxy or by solid-phase epitaxy (SPE) of evaporated a-Si. One advantage of this approach is that there is some preferential (1 0 0) orientation of the grains in the crystallized silicon layer (however, all orientations are observed and there is no in-plane ordering). This weak uniaxial texture facilitates subsequent thickening of the silicon layer by direct epitaxial growth, as low-temperature silicon epitaxy is easiest on (1 0 0) surfaces. A third approach is to spin a sparse array of nickel nanoparticles from solution onto a deposited a-Si:H layer [7]. After annealing, Niinduced crystallization causes silicon grains to grow to nearly the size of the spacing between Ni nanoparticles. Again, the resulting thin c-Si seed layer can be thickened with direct epitaxy. While both of these approaches result in crystalline layers with larger grains than directly annealed amorphous films, the resulting grains are not oriented. Even after H passivation, the minority-carrier diffusion length in these materials is limited by the deleterious high-angle grain boundaries and the resulting cell efficiencies have been low. Here, we propose the application of foreign template layers to control both the in-plane and out-of-plane orientation of thin crystal silicon films. The resulting low-angle grain boundaries should be less deleterious to film properties and should permit improved photovoltaic efficiencies. This new template approach relies on the initial deposition of a biaxially-textured film and the subsequent templated or heteroepitaxial growth of a c-Si seed layer. Further thickening of the Si seed layer by homoepitaxy or by Si SPE will also be simplified by the uniform out-of-plane grain orientation. In this paper, we outline the template approach to biaxially-textured c-Si films on glass, describe relevant experimental results and discuss the research chal-

985

lenges that will need to be overcome to fabricate films good enough to incorporate in high efficiency devices. 2. Foreign templates for biaxially-textured film c-Si on glass In recent years, the growth of biaxially-textured films on polycrystalline or amorphous substrates has become an important research topic [8]. Researchers are now able to grow crystalline superconductors with high critical currents on fine-grained substrates [8] by employing biaxially-textured template layers formed by ion-beam-assisted deposition (IBAD). A second approach to forming biaxial template layers is to use inclined substrate deposition (ISD) [9]. The essence of our proposed approach is to apply the process technology and basic approach developed for growing biaxially-textured superconductors on polycrystalline metal tapes to the challenge of growing biaxially-textured silicon on glass. While silicon has not been grown directly on glass with biaxial texture, IBAD and ISD have been used to directly deposit a number of other materials with biaxial texture on both amorphous oxides and polycrystalline metals. As shown in Fig. 1, we propose to use IBAD or ISD to grow a biaxially-textured template layer on glass and subsequently grow a silicon layer on this template. We anticipate that biaxially-textured thin silicon will be much easier to achieve on crystalline template layer than on the initial amorphous substrate. Subsequent thickening of the silicon layer will also be easier, as the silicon seed layer will be uniformly oriented and it will not be necessary to achieve epitaxial silicon growth on different crystalline orientations simultaneously. Recently, Findikoglu et al. [3] showed that c-Si with a total mosaic spread of 2–3° can be grown on template-coated polycrystalline Hastelloy

Fig. 1. Schematic of the proposed approach to crystal silicon film growth on glass. Initially, IBAD or ISD is used to grow a template layer with biaxial texture. Next, a silicon biaxially-textured seed layer film is grown heteroepitaxially on the template. If necessary, this silicon layer can be thickened. Both stages of silicon growth could be accomplished with either direct growth or solid-phase epitaxy crystallization.

986

C.W. Teplin et al. / Journal of Non-Crystalline Solids 352 (2006) 984–988

with a maximum processing temperature of 780 °C. These c-Si films have Hall mobilities that increase exponentially as the grain-angle disorder is decreased and approach that of single-crystal Si. The challenge will be to achieve similar results on glass at a lower temperature. 2.1. Foreign template layer growth Using normal deposition techniques, many materials will form a preferential out-of-plane orientation (i.e., uniaxial texture). For example, nanocrystalline silicon deposited by PECVD using high hydrogen dilution results in preferential formation of (2 2 0)-oriented grains [10]. The uniaxial texturing is due either to a preferential crystalline bonding orientation to the substrate, ballistic approach of deposition species at normal incidence and/or differences in the growth-surface energies of the competing crystallites. However, biaxial texturing is unlikely on a disordered substrate using normal incidence film deposition because there is no experimental parameter to break the axial in-plane symmetry. The most promising approaches to growing a biaxially-textured template on glass is to use IBAD or ISD. In the IBAD process, the in-plane symmetry is broken by a collimated beam of ions that is directed at the growing film surface at a selected angle, different from the surface normal. Under the right conditions, the interaction of the ion beam and the growing film results in biaxial texture. Another possible route to biaxially-textured film growth is inclined angle deposition, where the depositing atoms themselves are incident at non-normal incidence to the substrate [9]. A useful template material will be readily deposited with biaxial texture on glass and will be compatible with subsequent silicon epitaxy. Ideally, the template would also be conductive, so as to provide a back contact for the solar cell. Other considerations, such as diffusion of the template atoms into the crystalline silicon layer will also be important, especially if these impurities reduce carrier lifetime. We have identified a number of possible foreign template layers. A large knowledge base already exists for growing biaxially-textured thin films of oxides such as Ystabilized ZrO2, CeO2, and MgO [8] on amorphous or polycrystalline substrates and any of these cubic materials could be used as template layers for c-Si growth. Other classes of materials, particularly the silicides, have been demonstrated to be compatible with silicon heteroepitaxy. We believe that the biggest challenge with oxide templates will be the chemical stability of the silicon and oxide interface. With silicide templates, the problem is likely to be establishing biaxial texture in the silicide layer. Of the oxide candidates that we have identified (YSZ, CeO2, and MgO), CeO2 has the advantage of being nearly perfectly lattice˚ compared to matched to silicon (aceria = 5.41 A ˚ ac-Si = 5.43 A). We also note that contamination of the silicon layer is unlikely to be a problem with CeO2 because cerium is immobile in CeO2 [11] and it is unlikely that oxygen would diffuse past the CeO2/silicon interface. MgO is

exciting because it rapidly develops biaxial texture within about 10 nm of IBAD growth [12]. MgO could also be used to template growth of another material (e.g., c-Al2O3 [3]) with better chemistry for subsequent Si growth. Of the silicides, CoSi2 may be the best candidate template because the chemical compatibility of the Si/CoSi2 interface has already been established through the epitaxial growth of CoSi2 on Si(1 0 0) [13] and because Co is a relatively benign impurity in c-Si. 2.2. Growth of c-Si seed on foreign templates Epitaxial thin-film growth of silicon has been achieved at glass-compatible temperatures on single-crystal wafer substrates by many techniques [14–17], but heteroepitaxy or templated growth on a foreign substrate is more difficult. Heteroepitaxy of various materials on Si has been accomplished for a wide variety of materials. However, less research has been done on the reverse structure: epitaxial Si films on foreign crystalline substrates. This is likely due to the wide availability of high quality silicon wafers and the fact that the cost of wafers is not a limiting factor in the electronics industry. The greatest challenge is likely to be the high interfacial reactivity of the Si. Nonetheless, silicon epitaxy has been achieved on a number of foreign substrates [3,18]. However, these films were typically grown under ultra-high vacuum or at temperatures too high to be compatible with glass. A crucial challenge will be growth of epitaxial silicon seed layers at temperatures low enough to be compatible with inexpensive glasses (below 600 °C) and at reasonable pressures (10 7 Torr). Temperature and pressure restrictions are less problematic for solid-phase epitaxy (SPE), where silicon is initially deposited as an amorphous film and then crystallized during a subsequent annealing. A key advantage of solid-phase epitaxy is that it is a simple and scalable process. On Si substrates, the amorphous silicon precursor layer can be deposited very rapidly (up to 10 nm/s) [19] and films can be crystallized using batch annealing. However, we are unaware of any research on Si SPE on foreign materials. There are many problems to be overcome to achieve heteroepitaxy at low temperatures. These include surface cleaning without a high-temperature anneal to remove impurities, control of surface reconstruction of the foreign template, identification of templates that do not react strongly with Si to form new compounds at the heterointerface and compatibility of the crystal structures and surface chemistry at the hetero-interface. 2.3. Silicon film thickening The most promising deposition technique for homoepitaxial thickening from a biaxially-textured Si seed is ion-assisted deposition (IAD). IAD has been used to grow several microns of silicon on (1 0 0)-oriented silicon wafers ˚ /s in ultra-high at T < 450 °C at growth rates of 50 A vacuum [14]. Recently, this technique has been adapted

C.W. Teplin et al. / Journal of Non-Crystalline Solids 352 (2006) 984–988

for higher chamber base pressures more compatible with economical and large-scale processing [15] and applied on polycrystalline seed layers. The glass industry has found that IAD can be scaled in a cost effective way to coat large areas [20]. The techniques of hot-wire chemical vapor deposition (CVD) [16] and electron–cyclotron resonance CVD [17] can also be used to grow epitaxial films on (1 0 0)-oriented silicon wafers at low temperatures, but problems arise during epitaxial growth on other c-Si facets. Using SPE, several microns of epitaxial film have been grown on (1 0 0) silicon wafers during a single step [19]. Thus, SPE could be an effective film thickening technique if repeated amorphous silicon growth and annealing steps are employed. 3. Initial experiments at NREL To-date, work in our laboratory has focused on growth of biaxially-textured CeO2 films on glass using ion-beamassisted deposition and inclined substrate deposition by magnetron sputtering and growing heteroepitaxial silicon films by SPE on single-crystal CeO2 layers. The CeO2 films grown on glass using IBAD or directional sputtering have shown biaxial (1 0 0) texturing mixed with a small amount of uniaxial (1 1 1) texturing. It is likely that the films begin growth with a mixture of textures but that in-plane aligned (1 0 0) grains increase in size as the film is thickened due to preferential etching by the sputter source or ion beam. Thicker films or more rapid texture development are needed. We are studying Si film growth on single crystal (1 0 0)oriented pulse laser deposited CeO2 films in order to demonstrate the feasibility of preparing heteroepitaxial CeO2/ Si structures at glass-compatible temperatures. Amorphous silicon films were deposited onto the CeO2 using hot-wire (HW) CVD. Subsequently, the films were crystallized by annealing in a furnace for 8 h at 580 °C. X-ray diffraction and transmission electron microscope (TEM) measurements have shown that the films are polycrystalline. The TEM measurements also show that an amorphous SiOx forms about 3 nm thick at the CeO2/Si interface. We believe that this oxide forms by a rapid reaction of the Si and CeO2, at the interface. Thus, the crystalline order of the CeO2 template is disrupted and this prevents the desired heteroepitaxial nucleation of well-oriented silicon grains. Random homogeneous nucleation of polycrystalline Si results. We are currently exploring ways of better preparing the CeO2/Si interface in order to prevent or minimize the formation of this oxide. It is possible that better CeO2 surface cleaning/post-treatment or physical vapor deposition of silicon (as opposed to silane-based HWCVD with its reducing atomic H) may control formation of the amorphous SiO2 layer that we feel prevents epitaxy. Alternatively, it may be that silicon epitaxy is not possible on the (1 0 0) CeO2 interface. Currently, we are focusing on Si growth on other template materials, including CoSi2 and MgO.

987

4. Conclusions Here we propose a new approach to growing ordered crystalline silicon films on glass for low-cost solar cells. Our approach begins with growth of a biaxially-textured crystalline template layer, grown by ion-beam-assisted deposition or inclined substrate deposition. A biaxially-textured crystal silicon seed layer is then grown epitaxially on this layer. Photovoltaic solar cells will likely require epitaxial thickening to 2–10 lm and effective light trapping to reach efficiencies of 15% or higher. The low-angle grain boundaries will be less deleterious to solar cell performance, reducing the extent of passivation needed, and the uniform orientation of the silicon seed layer is expected to simplify thickening. Foreign template materials that can be grown biaxially-textured on glass are being surveyed for compatibility with subsequent Si growth. Acknowledgments We are grateful to colleagues for permitting us to briefly summarize their as yet unpublished experimental work on this project, some of it preliminary. Maikel Van Hest grew the textured CeO2; John Perkins assisted us with X-ray crystallography; Eugene Iwaniczko grew a-Si:H on CeO2; Paul Stradins, Qi Wang and David Young contributed to the thermal recrysallization; Kim Jones and Mowafak AlJassim provided transmission electron microscopy images. We also acknowledge the US Department of Energy for financial support under Contract DE-AC36-99GO10337. References [1] The US Photovoltaic Industry Roadmap. Available from: . [2] M.S. Keshner, R.R. Arya, Study of potential cost reductions resulting from super-large-scale manufacturing of PV modules. Available from: . [3] A.T. Findikoglu, W. Choi, V. Matias, T.G. Holesinger, Q.X. Jia, D.E. Peterson, Adv. Mater. 17 (2005) 1527. [4] A. Turner, M.J. Keevers, U. Schubert, P.A. Basore, M.A. Green, in: 10th EU PVSEC, Barcelona, 2005. [5] W. Fuhs, S. Gall, B. Rau, et al., Sol. Energy (UK) 77 (6) (2004) 961. [6] A.G. Aberle, A. Straub, P.I. Widenborg, A.B. Sproul, Y. Huang, P. Campbell, Progr. Photovoltaics 13 (2005) 37. [7] C.E. Richardson, M.S. Mason, H.A. Atwater, in: Materials Research Symposium Proceedings, Vol. 808, 2004, p. 227. [8] P.N. Arendt, S.R. Foltyn, MRS Bull. 29 (2004) 543. [9] A. Mancini, G. Celentano, F. Fabbri, V. Galluzzi, T. Petrisor, A. Rufoloni, E. Varesi, A. Vannozzi, R. Rogai, V. Boffa, U. Gambardella, Int. J. Mod. Phys. B 17 (2003) 886. [10] E. Vallat-Sauvain, U. Kroll, J. Meier, A. Shah, J. Pohl, J. Appl. Phys. 87 (2000) 3137. [11] C.L. Perkins, M.A. Henderson, C.H.F. Peden, G.S. Herman, J. Vac. Sci. Technol. A 19 (2001) 1942. [12] C.P. Wang, K.B. Do, M.R. Beasley, T.H. Geballe, R.H. Hammond, Appl. Phys. Lett. 71 (1997) 2955. [13] K. Ishida, Y. Miura, K. Hirose, S. Harada, T. Narusawa, Appl. Phys. Lett. 82 (2003) 1842. [14] M. Nerding, L. Oberbeck, T.A. Wagner, R.B. Bergmann, H.P. Strunk, J. Appl. Phys. 93 (2003) 2570.

988

C.W. Teplin et al. / Journal of Non-Crystalline Solids 352 (2006) 984–988

[15] A. Straub, N.P. Harder, Y.D. Huang, A.G. Aberle, J. Cryst. Growth 268 (2004) 41. [16] C.W. Teplin, Q. Wang, E. Iwaniczko, K.M. Jones, M. Al-Jassim, R.C. Reedy, H.M. Branz, J. Cryst. Growth 287 (2006) 414. [17] B. Rau, I. Sieber, B. Selle, S. Brehme, U. Knipper, S. Gall, W. Fuhs, Thin Solid Films 451&452 (2004) 644.

[18] C.G. Kim, J. Vac. Sci. Technol. B 18 (2000) 2650. [19] P. Stradins, D. Young, H.M. Branz, M. Page, Q. Wang, Mater. Res. Soc. Symp. Proc. A16.1.1 (2005). [20] Private communication with Guardian Industries.