Light Trapping in Thin Film Silicon nip Solar Cells

2 downloads 0 Views 1MB Size Report
Moreover, these cells are shown to be totally stable under light-soaking tests. .... on top facilitated a proper charge carrier collection. ... bottom cell and with a similar triple cell with a nc-Si:H middle cell. Here, we .... density of μc-Si:H n-i-p cells and the weighted ... integrating sphere. .... The dark arrows point to the cavities,.
Mater. Res. Soc. Symp. Proc. Vol. 1101 © 2008 Materials Research Society

1101-KK07-03

Light Trapping in Thin Film Silicon n-i-p Solar Cells - Gains and Losses Ruud E.I. Schropp, Hongbo Li, Jatin K. Rath, and Ronald H. Franken Faculty of Science, Debye Institute of Nanomaterials Science, Nanophotonics - Physics of Devices, Utrecht University, P.O. Box 80.000, Utrecht, 3508 TA, Netherlands ABSTRACT Thin film silicon solar cell technology frequently makes use of rough or textured surfaces in order to enhance light absorption within the thin absorber layers by scattering and total internal reflection (“light trapping”). The rough morphology of the optically functional internal surfaces both in superstrate and substrate cells however, not only has a beneficial effect on light scattering properties, but on the other hand may also have deleterious effects on the microscopic structure of the deposited layers, in particular if these layers are nanocrystalline. The narrow valleys in the surface morphology may lead to structural defects, such as cavities and pinholes. By adjusting the morphology, these defects can be avoided. However, even when structural defects in layers directly deposited on rough interfaces are avoided, the obtained optically defined maximum current density is still much lower than expected. For instance, in n-i-p structures the rough interface (the textured back reflector consisting of nanostructured Ag coated with ZnO) is located at the back of the cell, where only long wavelength light is present. The natively textured Ag film is sputtered at elevated temperature and optimized for diffusely reflecting this long wavelength light. From experiments we infer that the nanostructured metallic surface also gives rise to plasmon absorption in the red and near IR, and that this leads to a parasitic absorption, i.e. at least part of the absorbed energy is not re-emitted to the active layers.

INTRODUCTION Recently we have developed proto-Si/proto-SiGe/nc-Si:H triple junction n-i-p solar cells in which the top and bottom cell i-layers are deposited by Hot-Wire CVD. Firstly, a significant current enhancement is obtained by using textured Ag/ZnO back contacts developed in house instead of plain stainless steel. We studied the correlation between the integrated current density in the long wavelength range (650-1000 nm) with the back reflector surface roughness and clarified that the rms roughness from 2D AFM images correlates well with the long wavelength response of the cell when weighted with a Power Spectral Density function. For single junction 2-μm thick nc-Si:H n-i-p cells we improved the short circuit current density from the value of 15.2 mA/cm2 for plain stainless steel to 23.4 mA/cm2 for stainless steel coated with a textured Ag/ZnO back reflector. Secondly, we optimized the nc-Si:H n-type doped layer on this rough back reflector, the n/i interface, and in addition we used a profiling scheme for the H2/SiH4 ratio during i-layer deposition. The H2 dilution during growth was stepwise increased in order to prevent a transition to amorphous growth. The efficiency that was reached for a single junction nc-Si:H n-i-p cell was 8.6%, which is the highest reported value for hot-wire deposited cells of this kind, whereas the deposition rate of 2.1 Å/s is about twice as high as in record cells with HWCVD nc-Si:H so far. Moreover, these cells are shown to be totally stable under light-soaking tests. Combining the above techniques, a rather thin triple junction cell (total silicon thickness

2.5 μm) has been obtained with an initial efficiency of 10.9%. Light-soaking tests show that these triple cells degrade by less than 3.5% [1]. Hot-wire chemical vapor deposition (HWCVD) has become a viable method for the preparation of high-quality silicon and silicon alloy materials for application in thin film transistors and solar cells. At Utrecht University, we have developed HWCVD intrinsic protocrystalline silicon (proto-Si:H), which is characterized by an enhanced medium range structural order and a higher stability against light-soaking [2] compared to amorphous silicon, and nanocrystalline silicon (nc-Si:H), which is characterized by a low density of states [3] at a crystalline volume fraction of ~40% as determined by Raman spectroscopy. These materials were successfully applied in thin film solar cells on plain stainless steel [4,5]. To enhance the efficiency, multibandgap structures are required [6]. Here, we will discuss the differences in cell design between proto-Si/nc-Si/nc-Si triple junction cells and protoSi/proto-SiGe/nc-Si:H triple junction cells. The design is guided by the principle that the nonnanocrystalline cells have to be stable, and therefore these cells have to comprise absorber layers that are protocrystalline (more resistant against light-induced defect formation [7]) and thin (~150-200 nm). This then determines the thickness of the other absorber layers. A schematic picture of the triple junction cell structure is shown in Fig. 1.

Fig. 1. Schematic cross section of a triple junction thin film solar cell deposited onto a stainless steel substrate with a textured back reflector. The different subcell absorber materials are μc-Si:H, proto-SiGe, and protoSi (from bottom to top). On top of the silicon layers is an indium tinoxide/gold contact.

It will become clear that in the former case (with two nc-Si:H cells), the intrinsic absorber layers have to be made extremely thick in order to match the currents generated in the stacked subcells. Further we investigated the effects of a Ag/ZnO natively textured back reflector on the nc-Si:H layer structure and the light management within the cell, and found that there is a risk of considerable losses both due to structural defects and parasitic plasmon resonances in the rough Ag contact. Although the (optical) gain is much higher than the losses, it will be shown that the large potential of scattering back reflectors is not yet exploited fully.

PREPARATION OF THE LAYERS AND CELLS The silicon layers of the n-i-p structured solar cells were deposited in the PASTA multi-chamber ultra-high vacuum system. Details of the cell structures can be found in previous publications [5,8,9]. Doped layers and intrinsic proto-SiGe:H [9,10] were prepared using 13.56 MHz PECVD, whereas HWCVD was applied to fabricate intrinsic proto-Si:H [4] and nc-Si:H [6,11]. For the hot-wire deposition, two straight Ta filaments with a diameter of 0.5 mm were used, through which a current of 10.5 A was passed, yielding a wire temperature of 1850 ºC (vacuum calibration). The calibrated substrate temperature was 250 ºC. Proto-Si:H was deposited from undiluted SiH4, whereas H2-diluted SiH4 was used for nc-Si:H deposition (H2 dilution (H2flow/total gas flow) of around 0.95). The respective deposition rates were 10 Å/s and 2.1 Å/s. The nc-Si:H is a so-called mixed phase or transition material, consisting of nanocrystallites in an a-Si:H matrix [4]. The proto-SiGe:H was optimized on textured Asahi U-type SnO2:F substrates conformally coated with Ag and ZnO to provide a constant-quality textured back contact. The source gases for proto-SiGe:H were SiH4, GeH4, and H2. Typical band-gap values of a-SiGe:H are 1.5-1.6 eV [12]. The GeH4/(SiH4+GeH4) flow ratio was 0.45 and the H2/(SiH4+GeH4) ratio was 45. We used an exponential grading profile for the GeH4 flow, based on the work described in [13]. Two types of substrates were used: a Ag/ZnO TBR made on stainless steel (SS) foil in our laboratory [14,15] using our in-house Ag/ZnO magnetron sputtering tool SALSA, where the texture of the Ag surface is obtained using elevated substrate temperatures, and a SS/Ag/ZnO substrate provided by United Solar Ovonic LLC Corporation, for comparison. Indium-tin-oxide served as an anti-reflecting TCO top window; an evaporated gold grid on top facilitated a proper charge carrier collection. Both the metal oxide layers and the textured Ag of the TBR were deposited by rf magnetron sputtering in our SALSA system [14,15]. The active area of the solar cells was 0.13 cm2. RESULTS The efficiency that was reached for proto-Si/proto-SiGe/nc-Si:H triple junction n-i-p solar cells is 10.9%. In table I, the parameters are compared with a nc-Si:H single junction bottom cell and with a similar triple cell with a nc-Si:H middle cell. Here, we consider the significant consequences of choosing either proto-SiGe:H or nc-Si:H as the absorber material in the middle cell. In order to obtain sufficient current in triple junction cells of the type proto-Si/ nc-Si:H/ nc-Si:H, the design leads to the use of very thick absorber layers, even when enhanced scattering back reflection is applied. For instance, if we keep the top cell thickness at 165 nm, which is needed in order to generate ~ 8 mA/cm2, the middle cell and the bottom cell have to be made as thick as ~ 2.4 μm and ~ 3.7 μm, respectively. The total thickness of over 6 μm of ncSi:H material is very large compared to the micromorph tandem concept and leads to long deposition times at present-day deposition rates and high cost if implemented in production. In fact, if the deposition rate is not increased significantly, the thickness is prohibitively large. In other designs, one or two intermediate layers (IL) are used at the respective tunnel-recombination junctions to partially reflect light back into the top cell(s) so that these cell(s) can be kept thinner (and thus more stable) [16]. Unfortunately, unless the ILs are spectrally selective, the use of such ILs makes it necessary to even further increase the thickness of the bottom cells in order to match the (enhanced) current from the top cell.

However, replacing the nc-Si:H middle cell by a proto-SiGe:H cell, then not only the middle cell can be made an order of magnitude thinner, but also the bottom nc-Si:H cell can be made considerably thinner, since the SiGe:H middle cell does not absorb within exactly the same spectral region. A second important advantage is that a higher Voc can be obtained for the triple cell, because proto-SiGe:H has a higher band gap than nc-Si:H. Thirdly, in principle the achievable conversion efficiency is higher, because all three band gaps are different and less photon energy is lost as heat. We compared the performance of the two types of triple cells discussed above, as obtained within the limited time that was available for optimization. The cells are HWCVD proto-Si/proto-SiGe/ HWCVD nc-Si:H and HWCVD proto-Si/HWCVD nc-Si:H/ HWCVD ncSi:H triple n-i-p/n i p/n-i-p cells on textured Ag/ZnO made in house. For the moment we use PECVD for the deposition of the middle cell, since this process for proto-SiGe:H was readily available in our lab. The SiGe:H middle cell can also very well be made by HWCVD, as is demonstrated, for instance, by the excellent single junction cell results obtained by NREL [17]. Although the two types of triple cells were designed to give the same Jsc, the middle cell of the a-Si:H/ nc-Si:H/ nc-Si:H cell became slightly too thick, so that the bottom cell did not receive sufficient light. Therefore the bottom cell became the current limiting cell, and since its FF decreased as a result of its large thickness, the FF of the triple cell was reduced to 0.62. Due to the reduced Jsc and FF of the current limiting cell, the efficiency of the a Si:H/ nc-Si:H/ nc-Si:H is lower than for the a-Si:H/ a-SiGe:H/ nc-Si:H cell. Table I shows the comparison between the triple junction cell results where the middle cell is either nc-Si:H or proto-SiGe:H. Fig. 2 shows the schematic structure of two triple cells side by side. They were designed to deliver the same Jsc. On the left is the proto Si:H/ proto-SiGe:H/ nc-Si:H cell and on the right the proto Si:H/ nc-Si:H/ nc-Si:H cell. The thicknesses of the respective absorber layers are 180 nm/ 250 nm/ 2000 nm for the left hand cell and 180 nm/ 2400 nm /3700 nm for the right hand cell. This drawing conveys the important features of the design: the i-layers are drawn to scale and their relative thickness can be compared, while the other layers and the substrate are not to scale. This shows the thickness ratios that are needed in the two respective triple junction designs. It should be noted that if dielectric mirrors were used at the tunnel-recombination junctions in order to allow the use of a thinner (more stable) top cell, the bottom cells will have to be made even thicker to match its photogenerated current [18]. Moreover, a certain amount of light will be lost due to incomplete absorption of the (long-wavelength) light reflected into the top cell. Table I Performance of a n-i-p HWCVD nc-Si:H single junction cell, and triple n-i-p/n i p/n-i-p HWCVD proto-Si/proto-SiGe/ HWCVD nc-Si:H and HWCVD proto-Si/HWCVD ncSi:H/ HWCVD nc-Si:H solar cells on textured Ag/ZnO made in house.

Type of cell Single junction nc-Si:H n-i-p Triple junction proto-Si/ protoSiGe:H/ nc-Si:H Triple junction proto-Si/ nc-Si:H/ nc-Si:H

Voc (V)

FF

Jsc (mA/cm2)

Efficiency (%)

0.55

0.67

23.6

8.6

1.98

0.66

8.35

10.9

1.89

0.62

8.15

9.6

ITO a-Si:H

nc-Si:H

ITO a-Si:H a-SiGe:H

nc-Si:H

ZnO Ag SS Fig. 2. Side-by-side comparison of the schematic structure of two triple cells, designed to deliver the same Jsc. Left: an a Si:H/ a-SiGe:H/ nc-Si:H cell. Right an a Si:H/ nc-Si:H/ ncSi:H cell. The thicknesses of the respective absorber layers are 180 nm/ 250 nm/ 2000 nm for the left hand cell and 180 nm/ 2400 nm /3700 nm for the right hand cell.

Higher initial efficiency can of course be obtained by using thicker active layers for all three cells. However, this is at the risk of reduced stability of the performance. In addition, there is an additional economical advantage since the deposition time can be kept shorter. One of the improvements is expected to come from optimization of the middle cell (which is still made with PECVD). At present the middle cell is a limiting factor for the triple cell performance [12]. The roughness (morphology) at the back electrode/silicon interface is the key to efficient light trapping in n-i-p type thin film silicon solar cells. The texture of the Ag surface is obtained by magnetron sputtering at elevated temperatures. Larger crystals are obtained – and thus increasing roughness - as the substrate temperature during sputtering is increased. The rms roughness varies from 34 nm to well over 150 nm by varying the temperature from 225 °C to 425 °C.

Normalized total reflection intensity (%)

2

Corrected int. ECE; 650-1000 nm (mA/cm )

We studied the correlation between the integrated current density in the long 12 wavelength range (650-1000 nm) and the back 11 10 reflector surface roughness, for nc-Si:H n-i-p 9 cells with an i-layer thickness of 1.5 μm. It 8 became clear that the rms roughness from 2D 7 AFM images correlates well with the long 6 wavelength response of the cell when 5 weighted with a Power Spectral Density Glass + Ag 4 Asahi + Ag (PSD) function. The nature of this PSD 3 Glass + Ag:AlO function is that it gives a larger weight to the 2 lateral features with dimensions similar to the 0 20 40 60 effective wavelength to be scattered [19]. The Weighted rms roughness (nm) rms roughness σ was weighted with a Fig. 3. The correlation between the plasmon weighing factor PSD(λscat)/PSDtot, where corrected long-wavelength generated current PSD(λscat) is the contribution of the features density of μc-Si:H n-i-p cells and the weighted with sizes in the region 350 nm – 1400 nm. rms roughness for a large variety of back The correlation between surface morphology reflector morphologies. and generated current improves further if surface plasmon absorption is taken into account. The broad surface plasmon absorption of the rough Ag was calculated from total reflection measurements in air [19], which will be elucidated in the next paragraph. The result of weighing and correcting the relation between the External Collection Efficiency and the rms roughness is shown in Fig. 3. A limiting factor for further increasing the photogenerated current, 1.0 and thus the cell efficiency, by light scattering from a metallic back reflector 0.8 with increasing substrate roughness originates in the free carrier absorption 0.6 Ag, σ = 4 nm [20]. In Fig. 4 we show the total Ag, σ = 67 nm SP: 350 nm Ag, σ = 83 nm integrated reflection measurements of a 0.4 Ag, σ = 92 nm number of back reflectors with different Ag, σ = 111 nm surface roughness. Surface plasmons 0.2 BP: 320 nm (SP) can be localized (this occurs in metallic particles 10-200 nm in size) or 0.0 300 400 500 600 700 800 900 1000 propagating (this is associated with Wavelength (nm) smooth thin films). The propagating plasmons are often called surface Fig. 4 Total reflection measurements of various plasmon polaritons (SPP). In our types of back reflectors, showing indirectly the experiments, only the smoothest Ag absorption due to the surface plasmons (SP), as layers (rms roughness 4 nm) show just found by total reflection measurements with an the SPP absorption at 3.8 eV [21] and do integrating sphere. The red shifted absorption of the not show considerable localized SP localized SPs is dependent on the dimensions of the absorption. The textured Ag layers (rms morphology of the back contacts. > 30 nm) have a broad SP resonance x

Noramlized total reflection intensity (%)

extending from 350 nm to very long wavelengths, well into the spectral range of operation of the bottom cell. The red shift may be due to dipolar interactions of localized surface plasmons at neighboring Ag protrusions. The intensity of the SP absorption is dependent on the morphology of the surface of the back contact. In addition, the presence of the intermediate refractive index interlayer of ZnO:Al between the Ag surface and nc-Si:H i-layer may results in a weaker or stronger plasmon coupling, depending on its thickness [22]. Using AlOx inhibitor particle formation during sputtering, the lateral feature size L can be kept short while enhancing the (vertical) rms roughness [19]. In Fig. 5, the reflection from various such Ag:AlOx layers is shown. It is clear that either higher rms roughness, or smaller lateral feature size, lead to increased SP absorption and a red shift of the SP peak position. This is consistent with the enhanced interaction between adjacent Ag protrusions.

Ag, σ = 67 nm, L =1700 nm Ag:AlOx, σ = 61 nm, L = 1250 nm Ag:AlOx, σ = 76 nm, L = 830 nm Ag:AlOx, σ = 133 nm, L = 1600 nm

1.0 0.8

0.6

0.6

Energy shift + Increased Absorption

0.4

0.4

Increased Absorption

0.2

0.2

SP: 360 nm

0.0

0.0 300

400

320

600 800 Wavelenght (nm)

340

1000

360

380

1200

Fig. 5. Total reflection measurement of three Ag:AlOx layers with different roughness σ and feature size L. It is seen that either higher rms roughness, or smaller lateral feature size, lead to increased SP absorption and a red shift of the SP peak position.

We conclude that when the surface lateral feature size is in the range where maximum scattering occurs, the surface plasmon absorption can be strongly enhanced, which counteracts the beneficial scattering at SS/Ag/ZnO interfaces. The third reason for the importance of controlling this morphology of the surface of the textured back reflector, in particular when the first deposited cell is a nanocrystalline cell, is the risk of development of large structural defects in the nc-Si:H i-layer. To study this, we deposited nc-Si:H n-i-p cells on a variety of surfaces and found that the silicon structure becomes more defective when the rms roughness of the substrate increases. Fig. 6 shows the typical structural defects observed from the cross-sectional TEM images of a nc-Si:H n-i-p solar cell deposited on

a very rough Ag/ZnO surface. It can clearly be seen that structural defects are formed in the silicon layers near the edges of micro-valleys of the substrate surface. Structural defects of this kind may create shunting paths that largely deteriorate the cell performance [23]. In addition, electronic defects may be created in the bulk of the i-layers, reducing the solar cell fill factor and Voc. At higher deposition temperatures for the Ag, the surface shows large crystal facets with steep angles, leading to canyon-like valleys and sharp V-shaped valleys. The sputtering conditions for the ~ 100 nm thick ZnO cover layer on top of Ag were kept constant and it was shown that ZnO layer with this relatively small thickness did not introduce significant alteration of the Ag surface morphology [24]. As the nanocrystalline Si growth evolves anisotropically (almost perpendicular to the local surfaces), the shadowing effect of the large grains leads to cavities during growth of the nc-Si:H, while in the V-shaped valleys long void-rich regions are formed . For optimum yield, while still taking benefit from the internal light trapping effect, we have found an empirical guideline that the opening angles at the textured surface should be larger than 110 degrees [25]. Glass ~30nm nc-Si:H n-layer

~20nm nc-Si:H p-layer

Ag/ZnO layers

~ Ф100nm

ITO Ag/ZnO layers ITO nc-Si:H i-layer

a

b

Fig. 6. Cross sectional TEM pictures of a nc-Si:H n-i-p solar cell deposited on a rough Ag/ZnO coating. A Corning glass substrate is used instead of stainless steel to facilitate preparation of the cross section of the sample for TEM. The dark arrows point to the cavities, which are not completely filled with silicon.

CONCLUSIONS We conclude that the most economically viable thin film silicon triple junction consists of a nanocrystalline silicon bottom cell, a silicon-germanium middle cell, and an amorphous (preferably protocrystalline) silicon top cell. HWCVD shows to be a reliable technique for highquality thin film silicon deposition. Triple cells with a total silicon thickness of only 2.5 μm show an efficiency of 10.9%, and the light-induced degradation stays within only 3.5% relative. Further improvement of the triple cell would mainly come from optimization of the middle cell.

The roughness (morphology) at the back electrode/silicon interface is the key to efficient light trapping in n-i-p type thin film silicon solar cells and thus to enhancement of the photocurrent. The current enhancement in nc-Si:H single junction cells in practice can be as large as 50%. Purely optically, the net gain could be much higher, be it that further enhancement is hampered by two loss factors: 1) enhanced red-shifted plasmon absorption at the textured Ag surface due to electronic resonances in nanosized protrusions (particles), and 2) the defect evolution in nanocrystalline silicon layers due to acute angles and narrow valleys, which affects fill factor and Voc, and ultimately, also yield. Modifications of the dielectric overlayer and further control of the morphology of the rough Ag surface may lead a significant further increase of the net gain in performance.

ACKNOWLEDGMENTS We thank United Solar Ovonic LLC for providing stainless steel substrates with textured Ag/ZnO coatings. The research was financially supported by SenterNovem. REFERENCES 1. R.E.I. Schropp, H. Li, R.H.J. Franken, J.K. Rath, C.H.M. van der Werf, J.A. Schüttauf, and R.L. Stolk, Mater. Res. Soc. Symp. Proc. 989 (2007) 0989-A15-03. 2. R.E.I. Schropp, M.K. van Veen, C.H.M. van der Werf, D.L Williamson, A.H. Mahan, Proceedings of the 19th European Photovoltaic Solar Energy Conference, Paris (France), June 2004, p. 1526. 3. J.J.H. Strengers, F.A. Rubinelli, J.K. Rath, R.E.I. Schropp, Thin Solid Films 501 (2006) 291 4. M.K. van Veen and R.E.I. Schropp, Appl. Phys. Lett. 82 (2003) 287. 5. R.L. Stolk, H. Li, R.H. Franken, J.J.H. Strengers, C.H.M. van der Werf, J.K. Rath, R.E.I. Schropp, J. Non-Cryst. Solids 352 (2006) 1933. 6. B. Yan, G. Yue, J.M. Owens, J. Yang, S. Guha, 4th WCPEC, May 2006, Waikoloa Village, Hawaii (USA). 7. A. Gordijn, R. Jimenez Zambrano, J.K. Rath, R.E.I. Schropp, IEEE Trans. Elec. Dev. 49 (2002) 949. 8. R.L. Stolk, J.J.H. Strengers, H. Li, R.H. Franken, C.H.M. van der Werf, J.K. Rath, R.E.I. Schropp, Proc. 20th PVSEC, June 2005, Barcelona (Spain), p. 1655. 9. H. Li, R.L. Stolk, C.H.M. van der Werf, R.H. Franken, J.K. Rath, R.E.I. Schropp, J. NonCryst. Solids 352 (2006) 1941. 10. A. Gordijn,. R.J. Zambrano, J.K. Rath, R.E.I. Schropp, IEEE Transactions on Electron Devices 49 (2002) 949. 11. M.K. van Veen, Tandem Solar Cells Deposited Using Hot Wire Chemical Vapor Deposition, Ph.D. thesis, Utrecht University, the Netherlands, 2003. 12. H. Li, R.L. Stolk, C.H.M. van der Werf, R.H. Franken, J.K. Rath, R.E.I. Schropp, J. NonCryst. Solids 352 (2006), 1941-1944 13. R. Jimenez Zambrano, J.K. Rath, R.E.I. Schropp, J. Non-Cryst. Solids 299-302 (2002) 1131-1135

14. R.H. Franken, R.L. Stolk, H. Li, C.H.M. van der Werf, J.K. Rath, R.E.I. Schropp, Proceedings of the 21st European Photovoltaic Solar Energy Conference, Dresden (Germany), September 2006, p. 1565. 15. R.H. Franken, R.L. Stolk, H. Li, C.H.M. van der Werf, J.K. Rath, R.E.I. Schropp, Proceedings of the 21st European Photovoltaic Solar Energy Conference, Dresden (Germany), September 2006, p. 1744. 16. D. Fischer, S. Dubail, J. A. Anna Selvan, N. Pellaton Vaucher, R. Platz, Ch. Hof, U. Kroll, J. Meier, P. Torres, H. Keppner, N. Wyrsch, M. Goetz, A. Shah, and K.-D. Ufert, Proc. of the 25th IEEE PVSC, Washington D. C., USA _IEEE, New York, 1996, p1053. 17. Y. Xu, A.H. Mahan, L.M. Gedvillas, R.C. Reedy, H. Branz, Thin Solid Films 501 (2006) 198. 18. P. Buehlmann, J. Bailat, D. Dominé, A. Billet, F. Meillaud, A. Feltrin, and C. Ballif, Appl. Phys. Lett. 91 (2007) 143505. 19. R.H. Franken, R.L. Stolk, H. Li, C.H.M. van der Werf, J.K. Rath, R.E.I. Schropp, J. Appl. Phys. 102 (2007) 014503. 20. G. R. Aizin, N. J. M. Horing, L. G. Mourokh, and V. M. Kovalev, J. Appl. Phys. 96 (2004) 4225. 21. S.R. Barman, C. Biswas, K. Horn, Surf. Sci. 566-569 (2004) p538. 22. R.W. Collins, D. Sainju, L.R. Dahal, J. Li, J.A. Stoke, N.J. Pordraza, X. Deng, this symposium, this conference. 23. H. Li, R.H. Franken, R.L Stolk, J.K. Rath, R.E.I. Schropp, Solid State Phenomena Vols. 131-133 (2008) 27. 24. R. H. Franken, Transparent conducting oxide contacts and textured metal back reflectors for thin film silicon solar cells, Ph.D. thesis, Utrecht University, the Netherlands, 2006. 25. Hongbo Li, Single and multijunction silicon based thin film solar cells on flexible substrates with absorber layers made by hot-wire CVD, Ph.D. Thesis, Utrecht University, the Netherlands, 2007.