process a high potential for manufacturing of thin film silicon solar cells based on ... module of 0.447 m2 active area with 7.1 % initial efficiency has been ...
THIN FILM SILICON SOLAR CELL UP-SCALING BY LARGE-AREA PECVD KAI SYSTEMS J. Meier1, U. Kroll1, J. Spitznagel1, S. Benagli1, T. Roschek2, G. Pfanner2, Ch. Ellert2, G. Androutsopoulos, A. Hügli1, G. Büchel2, D. Plesa2, A. Büchel2, A. Shah3 1 Unaxis SPTec, Puits-Godets 12a, CH-2000 Neuchâtel, Switzerland 2 Unaxis Balzers AG, 9496 Balzers, Liechtenstein 3 Institut de Microtechnique (IMT), A.L.-Breguet 2, CH-2000 Neuchâtel ABSTRACT: UNAXIS KAI PECVD reactors developed for AM LCD technology have been demonstrated to process a high potential for manufacturing of thin film silicon solar cells based on amorphous and microcrystalline silicon. The single-chamber KAI reactors could fabricate already cells and modules. First microcrystalline silicon and micromorph tandem test cells of 5.5 % respectively 9.16 % were obtained. Up-scaling to sub-modules (491.5 cm2) by laser patterning resulted in initial aperture efficiencies of 8.8 % for a-Si:H p-i-n cells. A first prototype module of 0.447 m2 active area with 7.1 % initial efficiency has been achieved for amorphous silicon. Up-scaling of LP-CVD ZnO to areas of 1.4 m2 is ongoing for the next generation of a-Si:H and µc-Si:H based solar cell modules. All important module fabrication steps are developed in-house. Keywords: PECVD, a-Si, Micro Crystalline Si, ZnO, Manufacturing and Processing 1
INTRODUCTION
1.1 Up-scaling of thin film silicon cell technology – synergy with display industry Without doubt thin film solar cells have the potential for a substantial cost reduction of photovoltaic (PV) modules compared to crystalline wafer-based silicon technology [1, 2]. However, the main reason why thin film solar cells did not reach the commercial breakthrough lays in the problem of up-scaling of laboratory research results to large-area mass products of high performance. Up-scaling of thin film solar cells to areas of 1 m2 is a much more difficult task compared to the crystalline wafer technology, as there is no state-of-theart production equipment available. Conversely, thin film solar modules are involved with a higher risk of uncertainty and investment with respect to the development of production equipment compared to crystalline silicon where small well-known production units can be added to increase the production volume. In case of thin film solar cells the absorber needs to be prepared by a large-area deposition process with the appropriate properties for a high efficiency device, whereas todays conventional crystalline silicon solar cell technology profits mostly from the well-defined wafer as absorber. The Institut de Microtechnique (IMT) Neuchâtel has focused its research on the development of future thin film silicon solar cells. In 1987 IMT introduced the VHFGD (Very High Frequency Glow Discharge) technique for the improvement of the deposition rate of silicon films [3], and it has pioneered microcrystalline silicon as a new thin film absorber material [4]. IMT’s concept of combining amorphous silicon high bandgap and microcrystalline silicon low bandgap cells, the so-called “micromorph” concept, is considered one of the most promising thin film concepts [5-9]. In order to bring thin film solar cells to market, synergies with other types of mass production technologies have to be considered and adapted. In case of amorphous silicon solar cells one way is the synergy with the strong growing market of flat displays. In this industry segment the need of larger and larger deposition systems for the fabrication of growing demand of flat panel screens based on the amorphous silicon thin film
transistor (TFT) is driven by the market itself. The development in the display technology involves investments in larger and larger production equipments. In Fig. 1 Unaxis’ most recent reactor generation of 5 m2 size for the depositon of amorphous silicon is shown. Due to the coming market of flat panel TV screens larger and larger mother glasses are required. It is interesting to compare both markets the one of the display technology and the one of PV: Figure 2 reveals, similar to PV, a the strong growth in flat panel display industry, however, in comparison to PV and especially to
Figure 1: a-Si: film deposited on glass of 5 m2 area of Unaxis’ most recent generation of KAI reactor. x1000 m2
30000 25000 20000 15000 10000
5000 0 1998 1999 2000 2001 2002 2003 2004 2005
Large
small/medium
Solar cells: c-Si: ~7 km2 Thin film:~0.6km2
Figure 2: Strong market growth of AM-LCD technology in comparison with PV. Note, the thin film solar part is still very small compared to display thin film technology (source: Nikkei Microdevices FPD 2002 yearbook and display search Feb. 2003).
the thin film solar cells, thin film technology of display is much larger. Todays total worldwide produced thin film solars and even the total volume of all types of solar cells represent only a fraction of the yearly fabricated display area volume. In consequence production equipment of the flat panel industry is more developed than in thin film solar cells, and, hence, should be used as driving force for the development of new thin film silicon solar cell production tool equipment. Therefore, the compatibility of UNAXIS KAI PECVD systems for thin film solar cells combined with demonstrated performance of efficient high volume, high yield manufacturing can potentially fill this void of the past. 1.2 KAI PECVD Systems for thin film Si solar cells Unaxis display is a major manufacturer of production equipment for the deposition of amorphous silicon TFT’s on glass for flat panel displays. To react on the fast growing display market over the last ten years, different generations of large-area KAI production equipment have been developed at Unaxis (see result of most recent in Fig. 1). These KAI manufacturing systems as schematically shown in Fig. 3 are well approved in display production, they run 24h a day, 360 days a year. Unaxis has been investigating the potential of the KAI PECVD reactors for the thin film silicon solar cell concepts as developed at IMT Neuchâtel. In 2003 Unaxis decided to enter into the field of solar, in order to develop and adapt the present KAI PECVD technology to the fabrication of thin film silicon solar cells. Hereby, during autumn 2003 a R&D laboratory was created in Neuchâtel, whereas in Trübbach large-area developments have started. The goal of our business is not to produce thin fim solar cells, our intention is the production of manufacturing equipment for thin film silicon solar cell manufacturing. Highly automated KAI production systems developed for display technology once adapted for solar will have a significant impact on the cost reduction of PV modules (see principal in Fig. 3). The parallel processing concept of KAI systems, where up to 20 reactors operate in parallel in two process chambers assure the productivity needed for economical deposition of thin film silicon devices. In a KAI 1200 system, for example shown in Fig. 3, in each of the 2 process chambers, ten reactors are stacked vertically and operate in parallel. Therefore, 20 sheets of glass with a size of 1.4m2 size are processed simultaneously, resulting to a projected annual production capacity of 20 MWp for a-Si:H thin film solar cell production. In this paper we report on progress in preparation of amorphous and microcrystalline silicon based solar cells within our KAI reactors. Furthermore, results in upscaling of present cell performance to large-area modules by laser-scribing will be given. 2
EXPERIMENTAL
2.1 PECVD KAI reactor The heart of the KAI system is the plasma box® PECVD reactor [10]. The plasma box® combines isothermal heating in the reactor with differential pumping, which minimizes process contamination and make it fully compatible with fluorine-based selfcleaning processes, a necessity for high yield, low
maintenance manufacturing. Large-area deposition has already been demonstrated in this reactor concept up to 5 m2. In order to improve the deposition rate we adapted our reactors to an excitation frequency of 40 MHz. Our depositions run in two R&D reactors, one in a KAI-1200 (substrate size 1250x1100 mm) and one in a KAI-M (520x410 mm). Homogeneous depositions can be achieved in both systems.
Figure 3: Princip of a KAI-1200 production equipment system allowing for an yearly production volume of 20 MW/a of amorphous silicon solar cells. 2.2 Modified LP-CVD ZnO process In order to render the full efficiency potential of thin film silicon solar cells management of light plays a fundamental role. It has been shown by the work of IMT that low-pressure chemical vapor deposition (LP-CVD) of ZnO has excellent light-trapping properties in amorphous and microcrystalline silicon solar cells [7-9]. IMT has developed a “modified LP-CVD process” for the fabrication of ZnO. High Haze and good electrical properties could already be obtained. Amorphous and microcrystalline cells were deposited on ZnO and on various SnO2 coated glass substrates like Asahi U-type and AFG (American Float Glass). Due to the low process temperature of our deposition process this type of ZnO was successfully applied as back contact as well. 2.3 Laser-scribing One of the main advantages of thin film solar technology is the concept of series interconnection of the cell segments to large area modules. In amorphous silicon technology the series interconnection is realized by laser patterning. Unaxis demonstrated laser patterning on substrate areas of up to 1250x1100 mm. In order to focus on high performance and cost reduction of thin film silicon modules special attention was given to a high position accuracy of all three scribes and throughput. Meanwhile, all three patterns have reached the stage of scribe speeds above 20 m/min on our 1250x1100 mm substrate areas, both for commercial available SnO2 and LP-CVD ZnO. Modules of different type and dimension have been fabricated.
2.4 Light-trapping: White paint back reflector Light management has to be optimized on both sides of the thin film silicon cell absorber, i.e. on the front and on the back. Whereas in laboratory research most groups apply for highly reflective silver back contacts, mass production of a-Si:H based modules uses aluminum reflectors [11, 12]. Reasons for that are reduced costs and better properties of aluminum in passivating shunts. Thanks to the simple, “cheap” and low temperature ZnO LP-CVD process we followed a different option, namely using ZnO not only as front contact, but as well as back contact. Hereby, the thickness of the back ZnO is chosen in the way that it functions as electrical contact. Thus, the reflective properties of a standard metallic Ag or Al reflector can be replaced by a dielectric high resistive reflector, e.g. a white paint. This white reflector concept in combination with a highly conductive TCO divides the function of the back contact in two parts; one has to fulfill the electrical conductivity and the other the high reflectance of the back contact. The concept is shown in Fig. 4, and has been applied on our different types of cells and on all our modules. In case of amorphous silicon solar cells a similar option has been applied by the former Siemens group more than 10 years ago [13].
s
u
n
l
i
g
h
successfully applied in modules, have been developed to over 20 m/min for all three scribes.
Figure 5: ZnO covered glass substrate (1.4 m2) prepared at Unaxis.
t SnO2 or ZnO
glass
TCO Si film cell
Figure 6: Series interconnection achieved in Unaxis modules of 1.25x1.1 m size. Note the regularity of the scribe lines and the small scribe width of only 140 µm.
ZnO back reflector
Ag or Al
back reflector
white paint
Figure 4: Concept of the white back reflector in combination with LP-CVD ZnO as back contact in comparison with conventional ZnO/Ag or Al or ITO/Ag or Al contacts. 3
RESULTS AND DISCUSSION
3.1 ZnO prepared by a modified LP-CVD process We have already achieved ZnO layers with good homogeneity and conductivity within our modified LPCVD process. Sheet resistivity of 7.5 Ωsq have been obtained at excellent transmission and haze. Sofar we are able to upscale the film properties homogenously over larger areas. Figure 5 shows a 1.4 m2 glass with a typical ZnO deposition we apply in our cell technology. For more details of ZnO film properties we refer on reference [14]. 3.2 Laser-scribing Laser patterning of all three scribes has been studied for different TCO and for amorphous and micromorph tandem solar cells. We investigated in both small area and 1250x1100 mm large area the three patterns for the series interconnection. An excellent positioning of the three scribe lines over 1.25 m length could be achieved. Figure 6 reveals that all three pattern of the series interconnection lie within a range of only 140 µm. Each line has been optimized to function in a large-area module. Furthermore, the scribe velocities, as
Figure 7: Amorphous silicon modules prepared at Unaxis: On the right hand a 1250x1100 mm KAI 1200 module on SnO2 (AFG), on the left a 900x 550 mm module based on ZnO as front TCO. Note the dark appearance of ZnO due to better light-trapping [7]. The high precision of the laser system allows reducing dead area losses down to basic limitations of the individual scribe width, and, thus, simply improves the efficiency of the modules by a better structuration. The sofar developed high scribes speeds are fully compatible with mass production requirements. In Fig. 7 amorphous silicon modules are shown with complete patterning for the series interconnection. 3.3 White back reflector Our back reflector has been optically characterized and compared with conventional contacts used in amorphous silicon cells. There is a significant difference in the reflectance properties of between our white reflector and these sputtered ITO/Al or ZnO/Al contacts. Considering the diffused part of the reflectance LP-CVD
LP-CVD ZnO & white reflector
80
3
2.5
2
R = 7.5 A/s
1.5
R = 6.6 A/s
1
60
R = 5.0 A/s 0.5
R = 4.1 A/s
40
R = 2.3 A/s 0
400
600
800
1000
1200
Wavelength
Figure 8: Diffuse reflectance of a white reflector in comparison with standard sputtered ZnO or ITO / metallic contacts deposited on AF 45 Schott glass. The reflectance is characterized from the glass side by an Ulbricht sphere. An additional advantage of this concept may be in the reduced sensitivity to short circuits at the interconnects to the cell segments as ZnO has a lower specific conductivity than a metallic film. Furthermore, we observe clearly an improvement in the solar cell device compared to sputtered ZnO/ or ITO/Al contacts. Due to the simplicity of this white reflector combined with the modified LP-CVD ZnO technology and the stated advantages of this solution we consider this type of reflector very attractive in mass production. From the view of manufacturing and investment cost the white reflector is certainly a very attractive option. 3.4 Test cell devices and materiel Amorphous silicon single-junction p-i-n cells Amorphous silicon single-junction p-i-n test cells have been developed in both KAI-reactors (KAI-M and KAI-1200). Further details can be found in references [15, 16]. In the KAI-M reactor initial cell efficiencies of around 10 % could be achieved for both type of TCO’s (ZnO and Asahi SnO2) for a cell thickness of 0.3 µm. In the 1.4 m2 large-area reactor high FF in the range of 70 to 72 % could be obtained at a deposition rate of around 3 A/sec. Microcrystalline silicon films and solar cells First of all we investigated the microcrystalline silicon growth regime in our KAI-M reactor. Our preliminary studies reveal microcrystalline deposition sofar up to 7.5 A/sec. The Raman measurements of this series are given in Fig. 9. Further detailed investigations are needed to explore the full potential of the microcrystalline deposition regime. In order to fabricate first microcrystalline p-i-n solar cells p-doped µc-Si:H window layers have been developed carefully. Entirely microcrystalline cells have been prepared and optimized. Our sofar best cell is given
400
500
600
700
800
900
1000
-1
Wavenumbers [cm ]
Figure 9: Raman study (λ = 514 nm) of microcrystalline silicon growth in a KAI reactor. Deposition rates of over 7 A/sec could be obtained. 20 M270504
Current density [mA/cm2]
0
300
15
µc-Si:H p-i-n 10 V = 440 mV oc
FF = 69.2 % 2
Jsc = 18.21 mA/cm
5
η = 5.5 % 0
0
0.1
0.2
0.3
0.4
0.5
Voltage [Volts]
Figure 10: AM1.5 I-V characteristics of best microcrystalline p-i-n solar cell deposited in a KAI reactor. The cell has a thickness of 0.8 µm. Micromorph (a-Si:H/µc-Si:H) tandem cells Amorphous silicon top cells were combined with microcrystalline bottom cells to form the micromorph tandem cell. In case of the bottom cells a cell thickness of about 1.5–1.6 µm was deposited. Figure 11 shows the AM1.5 I-V characteristics of preliminary tandem cells 12
ZnO Current density [mA/cm ]
sputtered contacts: ZnO / Al ITO / Al Ag only
20
2
Diffuse reflection
100
in Fig. 10. Starting from low Voc and FF values we were able to increase these steadily to 440 mV respectively 69.2 %. Further improvement can be achieved by further detailed study of all deposition conditions. The cell in Fig. 10 is deposited at around 3 A/sec on LP-CVD ZnO as front TCO.
Intensity [a.u.]
ZnO/white paint reflectors show a much higher diffuse contribution than flat conventional back contacts. Fig. 8 shows the properties of the diffuse reflectance of the different reflectors in function of the wavelength. Each reflector was deposited on an AF45 Schott glass as in case of the backside of a solar cell and then characterized by the spectrometer from the glass side. The sputtered ITO/Al or ZnO/Ag contacts are developed for amorphous and microcrystalline silicon solar cells, e.g. the thickness of the sputtered ITO or ZnO is about 80 nm.
η = 9.16 %
10
Asahi SnO
8
2
η = 8.35 %
6 4 2 0
0
0.2
0.4
0.6
0.8
1
1.2
Voltage [Volts]
Figure 11: First micromorph (a-Si:H/µc-Si:H) tandem solar cells prepared in a KAI reactor on in-house ZnO and Asahi U-type SnO2 under AM1.5 illumination. Note, as the bottom cell has a thickness of only 1.5-1.6 µm the Jsc is bottom limited by light-trapping of the TCO [7].
Amorphous silicon single-junction modules The laser scribing parameters were first optimized on small area substrates of 10x10 cm for a-Si:H p-i-n solar cells and then on larger Asahi substrates. Our 30x22 cm module is shown in Fig. 13. The high fill factor reveals an excellent series interconnection for the laser patterning. Initial aperture module efficiencies of remarkable 8.8 % could be obtained. Further optimization with respect to the segment width and LPCVD ZnO and cell deposition will allow increasing the module efficiency. In order to check our large-area cell deposition and laser pattern technology we fabricated a-Si:H p-i-n modules of the substrate size of 900x550 mm. The deposition was done in a KAI-1200 reactor. To take the I-V characteristics the measurement was performed under outdoor conditions. A calibrated reference device determined the sun intensity. In order to check the quality of our laser patterning and possible shunt formation we took the I-V measurement at lower sun intensity (in the morning). The high FF and high Voc (all 86 segments are functional) at these conditions reveal a very good series interconnection and low shunts in the module. Under these conditions (6. June 04, 10 a.m at Neuchâtel site) a remarkable initial aperture efficiency of 7.1 % has been
0.35 2
outdoor 97.7 mW/cm 0.3
a-Si:H p-i-n module
0.25
Current [A]
3.5 Modules Semitransparent modules Our special ZnO back contact allows the realization of special semitransparent modules: As this back contact is transparent and conductive one is able to compose different semitransparent silicon modules based on aSi:H or uc-SiH single-junction solar cells and adapted front TCO’s. By haze and absorber thickness one has the option to design different appearance of semitransparent PV modules. Fig. 12 shows an example of a semitransparent a-Si:H module.
measured for our first a-Si:H prototype module. Applying LP-CVD ZnO technology will enhance the module efficiency further up due to a much better lighttrapping.
0.2
0.15
V = 19.8 V (22x 0.90V) oc
FF = 67.0 %
0.1
2
η = 8.8 %
(active area 491.5 cm )
0.05 0
0
5
10
15
20
Voltage [Volts]
Figure 13: I-V characteristics of an a-Si:H sub-module. Due to the lack of a sun simulator the measurement was performed under outdoor conditions and calibrated by a reference device. The a-Si:H cell is 0.3 µm thick. 0.8
a-Si:H p-i-n module active area 0.447 m2
0.7
2
95.6 mW/cm : η= 7.1 %
0.6
Current [A]
deposited on different front TCO. For both front TCOs already a good interface leading to a high FF of over 70 % could be achieved. The lower short-circuit-current density in case of Asahi is due to the reduced lighttrapping in the bottom cell. Further optimization in each doped layer and the Voc of the microcrystalline bottom cell will enhance the efficiency of our micromorph tandems to over 10 % stable efficiency.
0.5 0.4 0.3 low intensity: 0.2
2
56.8 mW/cm : η = 7.06 %
0.1 0
0
10
20
30
40
50
60
70
Voltage [Volts]
Figure 14: First large-area a-Si:H module fully fabricated at Unaxis. AFG SnO2 substrates were used, the laser series interconnection is of the quality as shown in Fig. 6. 3.3 Micromorph (a-Si:H /µc-Si:H) tandem modules Micromorph tandem cell mini modules on the substrate size of 10x10 cm were tested as well. Figure 15 shows our sofar best result of a module with 10 Segments. Already an initial efficiency of 7.9 % could be obtained. Further improvements will lead to efficiencies over 10 %, especially when LP-CVD ZnO will be applied. 64 56
Current [mA]
48
Micromorph a-Si:H/µc-Si:H tandem module
40 32 24
Voc = 12.24 V (10 Segm.)
16
FF = 69.0 %
η = 7.9 % (active area 63.36 cm 2)
8
Figure 12: Semi-transparent module based on an amorphous single-junction cell and LP-CVD ZnO as back contact. Note the absorber can be replaced by microcrystalline silicon resulting in a different color due to the band gap. The transmission can be adapted by the absorber thickness, the choice of front TCO (Haze) and in addition to the amount of scribed area.
0
0
2
4
6
8
10
12
Voltage [Volts]
Figure 15: I-V characteristics of first micromorph aSi:H/µc-Si:H tandem module under AM1.5 illumination prepared in a UNAXIS KAI single-chamber reactor on Asahi SnO2.
4. MANUFACTURING ISSUES OF THIN FILM SILICON MODULES A very critical aspect is process yield because of its big impact on direct cost. Moreover, high yield is important from a logistical point of view in production planning. In AM-LCD yield levels of above 95% are commonly reached and the remaining yield loss is mainly due to process particles, causing defects of individual TFT’s. The impact of this type of yield loss will be far less severe in thin film silicon solar cell production and it can be assumed that yield levels of above 95% are attainable for PECVD with KAI production technology. Material cost estimation for the thin film device (TCO, Si absorber and back contact) is promisingly low. They amount to roughly 1/3 €/Wp and can be reduced further with process optimizations and increased production volumes. Larger production volumes combined with larger module sizes and cell efficiency optimizations will allow additional cost reductions. The high productivity of proven AM-LCD machines such as the KAI PECVD systems reduces both costs and risks for a thin-film production line. Using today’s generation of KAI machines, one arrives already at similar investment costs compared to the crystalline technology for wafer, cell and module manufacturing. Further reductions are expected here with process improvements and with each new generation of AMLCD machines. Moreover, thin film Si PV modules are ideally suited for building integrated use, because of their uniformly dark appearance and their fairly large size. Buildingintegrated use of thin film silicon PV modules can offset a large percentage of the balance–of-system cost (BOS) in the future because of savings for the material that will be substituted. 5
CONCLUSIONS
In the past high investment-related cost in production machinery were considered an obstacle for the growth of thin film silicon PV. Moreover, typically prototype systems were built that only reached the targeted uptime performance after an extended ramp-up phase. In order to realize the cost reduction potential of thin film silicon solar cells the synergies to the AM-LCD market give UNAXIS immediate access to an advanced production technology. The high productivity of UNAXIS KAI PECVD machines has the strength to minimize cost and risk of a thin film production line. Already reasonable cell and module performance could be demonstrated for silicon solar cells deposited in KAI single-chamber reactors. Thus, first microcrystalline silicon and micromorph tandem cells of 5.5 % respectively 9.16 % efficiency could be prepared. With our laser patterning system we were able to upscale amorphous silicon p-i-n single-junction modules. A sub-module of 491.5 cm2 active area showed an initial efficiency of 8.8 % for Asahi substrates. On AFG SnO2 we achieved with our first prototype module on 0.447 m2 area an aperture efficiency of 7.1 %. In case of micromorph tandem cell modules we fabricated a first mini module (of 63.3 cm2) of 7.9 %. We were able to scale up our “modified” LPCVD process of ZnO to areas of 1.4 m2 for the next
generation of thin film silicon modules. In order to improve light-trapping the concept of LP-CVD ZnO as conductive back contact in combination with a white resisitive back reflector has been successfully applied on all our modules. Unaxis manages all process steps from glass cleaning to the fabrication of large-area modules. Further detailed work will improve our cell and module performance. If high module performance at high throughput is demonstrated on a production level as well, Unaxis KAI PECVD systems can have a pronounced impact on cost reduction of future PV modules. In conclusion, direct manufacturing cost of close to 1 €/Wp should be achievable with KAI production systems in near future. Each of critical cost contribution clearly bears for further improvement in the near term. Therefore, thin film silicon technology can pave the path to economical largescale application of PV. 6. ACKNOWLEDGEMENTS The authors thank the IMT PV research group for scientific and the technical support. Furthermore, we wish to thank Dr. Hollenstein and his group at CRPP Lausanne for helpful support in developing the large-area deposition. 7. REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16]
J.M. Woodcock, H. Schade, H. Maurus, B. Dimmler, J. Springer, A. Ricaud, Proc. 14th EPVSEC 1997, p. 857. M.A. Green, Proc. 3rd World Conference on Photovoltaic Energy Conversion (Osaka May 2003), paper OPL-02. H. Curtins, N. Wyrsch, A. Shah, Electron. Lett. 23 (1987), p. 228. J. Meier et al., Appl. Phys. Lett. 65 (1994) 860. J. Meier et al., Proc. of the 1st WCPEC (Hawaii, USA, 1994) p. 409. J. Meier et al., Proc. of the Mat. Res. Soc. Symp. 420 (1996) 3. J. Meier et al., Solar Energy Mat. & Solar Cells 74 (2002), p. 457. see Proc. 3rd World Conference on Photovoltaic Energy Conversion (Osaka May 2003), session S2. J. Meier et al., Thin Solid Films 451-452 (2004), p. 518. E. Turlot and F. Leblanc, Nikkei Microdevices, April 1999, p. 151. S. Guha, Proc. of 13th Sunshine Workshop on Thin Film Solar Cells; Technical Digest, Tokyo, Japan (2000), p. 25. D.E. Carlson et al., Proc. NREL/SNL Photovoltaics Programm Review, AIP Conference 394 (1997), p. 479. R. van den Berg, H. Calwer, P. Marklstorfer, R. Meckes, F.W. Schulze, K.-D. Ufert, H. Vogt, Solar Energy Materials and Solar Cells 31 (1993), p. 253. S. Faÿ, L. Feitknecht, R. Schlüchter, U. Kroll, E. Vallat-Sauvain, A. Shah, to be publ. in Proc. 14th PVSEC (Thailand January 2004). U. Kroll et al., Thin Solid Films 451-452 (2004), p. 525. U. Kroll et al., this conference, to be publ. in Proc. of 19th Europ. PV SEC (2004).