THIN-FILM CRYSTALLINE SILICON SOLAR CELLS ON CERAMIC SUBSTRATES C.J.J. Tool, J.A.M. van Roosmalen, S.E.A. Schiermeier, R.C. Huiberts, G.M. Christie, W.C. Sinke Netherlands Energy Research Foundation ECN PO Box 1, 1755 ZG Petten, The Netherlands phone: +31 224 564135 fax: +31 224 563214 email: [email protected]
ABSTRACT: For a successful development of thin-film crystalline silicon solar cells, both the availability of cheap substrates and a suitable deposition technique are required. To ease the introduction of thin-film silicon solar cells in commercial process lines, substrates which can withstand temperatures up to 900°C are preferred. Because of additional demands such as low-cost, non-toxicity and high strength, ceramics are good candidates to serve as substrate materials. To facilitate nucleation, and to match the silicon thermal expansion coefficient, the production of silicon based ceramics by the potentially low-cost tape casting method is being investigated. To increase the density and strength of the substrates, Al and Sialon were used as additives. To investigate nucleation and layer growth on the ceramic substrates, Liquid Phase Epitaxy has been chosen as the deposition technique. Although we can grow easily from aluminium melts, we focus on (doped) indium and tin melts. The results indicate that layer growth on the substrates is possible, but no continuous layers have been realised yet. Keywords: Si-Films - 1: Substrate - 2: LPE - 3
For a further cost reduction of silicon solar cells, thinfilm silicon solar cells on cheap ceramic substrates represent a good candidate . The reduced use of expensive silicon material and the larger processing area will result in a lower price per watt-peak. For a successful development of this technique, the availability of cheap substrate and a suitable deposition process are required. At ECN both aspects are subject of investigated. The emphasis of this paper will be on the demands and preparation of cheap ceramic substrates. 2
THIN-FILM Si SOLAR CELLS at ECN
2.1 Substrate The deposition of crystalline silicon films on low-cost substrates is the subject of many investigations (e.g. [2-6]. In most of these investigations glass is used as the substrate material. Due to the glass substrate, the temperatures at which the films can be processed into solar cells are limited. Although the development of high temperature glass is part of some investigations , the outcome with respect to efficiency and costs remains uncertain. Completely low-temperature process routes for thinfilm silicon solar cells may be possible, but high temperature processes are common in silicon solar cell processing nowadays. To ease the introduction of thin-film silicon solar cells, it is probably beneficial that the silicon films on the cheap substrates can be processed in commercially available production processes. So a substrate which can withstand temperatures up to 900°C is preferred. Also, to use standard interconnection schemes (back and front contacting), a conductive substrate is preferred. Besides, on a conducting substrate interconnection schemes specially defined to take full advantage of thin film Si solar cells can be used [8, 9].
The substrate must also be non-toxic and have a high strength. Because of these combined demands ceramics are a suitable candidate material. Because of the high processing temperature and the temperature cycling during processing, as well as during long-term outdoor use, the thermal expansion coefficient must meet that of silicon. Also, to enforce Si-deposition and layer growth, Si nucleation sites in the ceramic substrate are beneficial. Therefore, we consider siliconbased ceramics most appropriate. Additions, e.g. Al, SiO2, Al2O3 or Sialon are used to increase the density and the strength of the sintered substrates. Within the Solid Oxide Fuel Cell Group at ECN a lot of experience exists on producing thin ceramic sheets by the tape-casting method . Therefore, and because tape casting is in principle a cheap process, this technique is being used to develop the thin ceramic sheets. To make ceramics by tape-casting, the powder or powder mixture is blended with water, a binder and special additives. This slip is spread out on a glass bed using the equipment shown in Fig. 1. E
D C B A
Figure 1: tape-casting equipment; A: glass bed, B: carrier film, C: slip, D: doctor blades, E: precision height adjustment After drying under ambient conditions, the flexible green tape is cut to the desired sizes and sintered in an
inert atmosphere. The sintering temperature used varied between 1300 to 1600°C. 2.2 Silicon deposition Liquid phase epitaxy is a relatively low temperature process. Because the deposition can be done under nearequilibrium conditions, very good quality silicon layers can in principle be grown. Although growth is performed in the presence of a major impurity (the melt), the composition of the melt can be tuned to dope the layer during growth (e.g. add InAs to an indium melt to grow ntype layers from the p-type indium melt). Some melts can even be used to accumulate impurities . In that case less pure silicon can be used as feedstock material to grow higher purity silicon layers. In this investigation initially an aluminium melt has been used. Due to the low oxygen potential of aluminium, layers can be grown easily. However, the dopant level of the resulting layer (≈ 1019 at. Al / cc ) is too high for the active layer. Therefore, additional experiments have been performed using either Sn (not electrically active in Si) and In (dopant concentration ≈ 1016 at. In / cc ). Because of the simplicity of the equipment, in all experiments growth is forced by temperature variation of the melt. The general procedure was to first saturate the melt using a silicon wafer. Then the saturated melt was brought into contact with the substrate. Growth is forced by a controlled (slow) cooling of the melt. In between saturation and growth the melt is cooled down to room temperature to change the substrate. Immediately prior to the experiment the saturation wafer or the substrate was given an HF dip to remove native oxides. At ECN horizontal and vertical growth systems are used. In the horizontal system, the melt is brought in contact with the substrate by tipping the (graphite) boat. The melt volume to substrate surface area ratio is small, and the melt is refreshed after each growth experiment. This system can easily be used to investigate the influence of the melt composition on the growth. In the vertical system the substrate is brought into contact with the melt by lowering the substrate with the aid of a pull-rod. In this system the melt volume to substrate surface area is much higher, giving the opportunity to grow layers at a relatively high rate. The dipping system has been described elsewhere , the tipping system is shown in fig. 2.
B A AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA C AAAA AAAA AAAA AAAA AAAAAAAA AAAAAAAA AAAAAAAA AAAAAAAA AAAAB AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAA AAAAAAAAAAAAAAAAAAAA AAAA
Figure 2: Tipping boat system; A: quarts tube, B: graphite boat, C: melt, D: substrate In both systems, growth is a batch process, growing one layer at a time. However, the aim is to develop a cheap production process, so (semi) continuous growth is preferred over batch wise growth. Therefore, the development of more advanced growth techniques is a part of the continuation of this investigation.
Within the scope of a nationally funded project, three different deposition techniques (hot wire deposition , CVD  and LPE) are compared. 3
3.1 Si / Al substrate To sinter silicon into nearly dense sheets, a sintering temperature close to its melting point has to be used. To enhance sintering by a liquid phase formation during sintering and to create a highly conductive substrate, 3 w-% aluminium was added to the tape. This increased the sinterability of silicon, but it limited the sintering temperature to about 1300°C. Above this temperature the amount of liquid phase became to high, and the substrate melted partially. Sintering at 1300°C resulted in substrates with a porosity of about 50 % and a strength of about 10 MPa, measured using a standard ring-on-ring test  (mc-Si wafers showed a strength of about 80 MPa under the same test conditions). Fig. 3 clearly shows the high porosity of the Si/Al substrate.
Figure 3: top view of a 50% porous Si/Al substrate sintered at 1300°C. Specimens with sizes up to 70 × 70 mm2 can be prepared from this Si/Al ceramic reproducibly. To prove the nucleatability on Si/Al substrate, growth experiments have been performed on this substrate from an aluminium melt using a cooling rate of 10°C/hr from 905 to 903°C using the vertical system. Despite the large pores in the substrate, a nearly continuous layer has been grown (see fig. 4). In this figure the 500 µm thick substrate is in the middle, with a ≈ 50 µm thick Si layer on each side. Due to the presence of some irregular silicon crystals, aluminium melt adhered to the substrate when it was removed from the melt. Due to the presence of aluminium in the substrate this melt could not be removed chemically. Additional growth experiments from indium and tin melts resulted in large faceted crystals, with some coalescence of the crystals. Due to the presence of the large pores in the substrate, no continuous layer growth has been observed. Fig. 5 shows a cross section of a sample grown from an indium melt at a cooling rate of 50°C per hour from 960 to 800°C. This figure reveals that some (small) pores are overgrown, but the large pores are not closed by the grown layer.
Figure 4: cross section of growth on Si/Al from Al.
Figure 6: top view of Si growth on plasma sprayed silicon from an indium melt.
Figure 5: cross section of growth from indium melt. To minimise the influence of the porosity of the substrate on the silicon growth, several methods are possible: • increasing the sintering temperature to force further sintering of the substrate. However, increasing the sintering temperature did not result in a decrease of the porosity, but in melting of the substrate due to the presence of too much liquid phase. • increasing the aluminium content to increase the liquid phase which enables sintering also resulted in partial melting of the substrate. • closing the pores by applying a special coating on the substrate. It was found that it is relatively easy to plasma spray a silicon coating of about 50 µm thick on the porous Si/Al substrates, resulting in a nearly closed surface. On plasma sprayed silicon layers, growth has been performed from an indium melt at a cooling rate of 30°C per hour from 960 to 920°C using the vertical system. In fig. 6 a top view of the resulting layer is shown. This photograph clearly reveals the large faceted silicon crystals. Although there is some coalescence of the crystals, no closed layer has been obtained yet.
3.2 Si/Sialon substrate Another approach to remove the pores from the Si/Al substrate is by adding a sinterable powder to the slip. As a first experiment, silicon was mixed with sialon powder (50 / 50 w-%). Sialon powder has the advantage over other sinterable ceramic powders such as mullite, that it is a conducting material. Sintering a tape of 50 / 50 w-% Si / Sialon at 1500°C resulted in a 75% dense substrate. The strength of this ceramic is 84 MPa, measured with the ring-on-ring test. The substrate is much stronger compared to the Si/Al substrates, and comparable in strength to mc-Si wafers. Although the porosity is still relatively high, a cross section revealed the pores are much smaller than the pores in the Si/Al substrate (a few microns vs. several tens of microns). In fig. 7 a top view of the Si/Sialon substrate is given. This photograph clearly shows that the porosity is reduced compared to the Si/Al substrate (compare fig. 3). Specimens with sizes up to 30 × 30 mm2 can be prepared reproducibly from this Si/Sialon ceramic.
Figure 7: top view of 75% dense Si/Sialon substrate sintered at 1600°C. Growth on this Si/Sialon substrate from an indium + 1% aluminium melt at a cooling rate of 50°C per hour
from 960 to 860°C using the vertical growth system gave a result comparable to the growth on plasma sprayed silicon. The sample contained large faceted crystals with some coalescence. The layers grown on Si/Sialon contained some more open areas compared to the layers grown on plasma sprayer silicon. Fig. 7 also reveals the presence of “globes” or “bubbles” on the surface of the Si/Sialon substrate. These “bubbles” have not been analysed yet, however it seems that these are silicon bubbles secreted during the sintering process. An increase of the sintering temperature results in an increase of the “bubbles”. Sintering at 1600°C for 4 hours resulted in a partly closed layer of this secretion. We have not yet succeeded in controlling this secretion, but it may open a route in creating a closed silicon BSF layer in situ during the ceramic production on which growth of the active silicon layer is possibly relatively easy.
  
Despite the presence of large pores, silicon layer growth has been shown on the Si/Al substrate from an aluminium melt. To grow layers on this substrate from a melt which results in a lower doping level (e.g. tin or indium), the porosity or the pore size must be decreased. Proof is not extensive, but the results indicate that the coverage of the substrate by the grown silicon layer is influenced by the amount of silicon available at the substrate surface. The highest coverage of the substrate is obtained on the plasma sprayed surface, even in the absence of a reducing element such as aluminium. This substrate also has the highest silicon content on the surface (nearly 100%). The lowest coverage is found on the Al/Si surface. Due to the high porosity, this substrate contains the lowest silicon content on the surface (< 50% due to 50% pores and some Al). The silicon surface content of the Si/Sialon substrate normally would be lower (50% Si; 75% dense therefore 37.5% silicon), the surface silicon content of this substrate is increased by the silicon secretion. This observation supports the initial choice to use silicon based ceramics with a high silicon content. REFERENCES  
J.A.M. van Roosmalen, Thin-Film Solar Cells, report nr. ECN-C--95-107 (1995). T.H. Wang, T.F. Ciszek, Y.S. Tsuo, J. Alleman, X. Wu, C.R. Schwerdtfenger, R.W. Burrows, AIP Conf. Proc. 306, (1994), 92. R.L. Wallace, B. Jagannathan, X. Gu, N. Sridhar, K. Etemadi, D.D.L. Chung, W.A. Anderson, AIP Conf. Proc. 353, (1995), 511. J.E. Cotter, A.M. Barnett, D.H. Ford, R.B. Hall, A.E. Ingram, J.A. Rand, T.R. Ruffins, K.P. Shreve, C.J. Thomas, Progress Photovoltaics: Res. & Appl. 3, (1995), 351. I. Silier, A. Gutjahr, F. Banhart, M. Konuma, E. Bauser, V. Schöllkopf, H. Frey, Mater. Lett. 28, (1996), 87. T. Vermeulen, J. Poortmans, K. Said, O. Evrard, W. Laurey, M. Caymax, J. Nijs, R. Mertens, C. Vinckier, 25th IEEE Photovoltaic Specialists Conference (1996), IEEE Ed (1996) 653.
R.B. Bergmann, R. Brendel, M. Wolf, P. Lölgen, J.H. Werner, J. Krinke, H.P. Strunk, 25th IEEE Photovoltaic Specialist Conference (1996), IEEE Ed (1996), 365. J.A.M. van Roosmalen, Fotovoltaisch zonnepaneel met in serie geschakelde zonnecellen, Dutch patent application nr 1.005.926 (1997). J.A.M. van Roosmalen, M.W. Brieko, A.R. Burgers, J. Daey Ouwens, A.C. Tip, C.J.J. Tool, A.W. Weeber, W.C. Sinke, A new design for thin-film crystalline silicon solar cell modules. Abstract submitted to the 26th IEEE Photovoltaic Specialists Conference. J.P.P. Huijsmans, F.P.F. van Berkel, J.P. de Jong, Fuel Cell Seminar, San Diego, California (1994), 499. T.F. Ciszek, T.H. Wang, R.W. Burrows, X. Wu, J. Cryst. Growth 128, (1993), 314. J.L. Murray, A.J. Mc.Alister, Bull. Alloy Phase Diagrams 5, (1984), 74. A.W. Blakers, J.H. Werner, E. Bauser, H.J. Queisser, Appl. Phys. Lett. 20, (1992), 2752. J.A.M. van Roosmalen, C.J.J. Tool, R.C. Huiberts, R.J.G. Beenen, J.P.P. Huijsmans, W.C. Sinke, 25th IEEE PV Specialists Conference (1996), IEEE Ed (1996) 657. J.K. Rath, M. Galetto, C.H.M. van der Werf, K.F. Feenstra, H. Meiling, M.W.M. van Cleef, R.E.I. Schropp 9th Int. Photovoltaic Science and Engineering Conference Ridumi Printing Inc. Ed (1996), 227. A.J.M.M. van Zutphen, M. Zeman, J.W. Metselaar, presented at 14th EC Photovoltaic Solar Energy Conference (1997) P5A.18. P.P.A.C. Pex, M. Jong, Mechanical testing of thin electrolyte materials for solid oxide fuel cells using the concentric ring test, report nr. ECN-2918-GR10.
ACKNOWLEDGEMENT This work has been carried out with financial support of the European Commission within the Joule HIFI project, the Netherlands Agency for Energy and the Environment (Novem) within the NOZ-PV programme, and ECN within the Energy Generation in the Natural Environment Programme “Engine”.