Towards monocrystalline silicon thin films grown on glass by liquid phase crystallization

Towards monocrystalline silicon thin films grown on glass by liquid phase crystallization

Solar Energy Materials & Solar Cells 140 (2015) 86–91 Contents lists available at ScienceDirect Solar Energy Materials & Solar Cells journal homepag...

2MB Sizes 1 Downloads 36 Views

Solar Energy Materials & Solar Cells 140 (2015) 86–91

Contents lists available at ScienceDirect

Solar Energy Materials & Solar Cells journal homepage: www.elsevier.com/locate/solmat

Towards monocrystalline silicon thin films grown on glass by liquid phase crystallization S. Kühnapfel n, S. Gall, B. Rech, D. Amkreutz Helmholtz-Zentrum Berlin für Materialien und Energie-Institut für Silizium-Photovoltaik, Kekuléstr. 5, 12489 Berlin, Germany

art ic l e i nf o

a b s t r a c t

Article history: Received 1 February 2015 Received in revised form 24 March 2015 Accepted 26 March 2015

Liquid phase crystallization of silicon is a promising technology to grow crystalline silicon thin films on glass. It has already been demonstrated that open circuit voltages of up to 656 mV and efficiencies of up to 11.8% can be achieved by this technique. Nevertheless further improvements are required to become competitive with wafer based silicon solar cells. A possibility to improve the quality is to enlarge the grain size and to control the crystallographic orientation of the resulting layers. While a preferential {100} surface orientation can be triggered utilizing suitable crystallization parameters, the in-plane orientation still remains random. Also the grain size stays within the same range. By introducing a local monocrystalline seed at the beginning of the crystallization process, we are able to control both the surface and the in-plane orientation of the silicon films. In addition the grain size is significantly increased in scanning direction and only limited by the substrate size of 5 cm. This high morphological quality is accompanied by an improved electrical quality confirmed by photoluminescence imaging and Hall measurements. This is a big step towards the final goal to directly grow monocrystalline silicon thin films on glass substrates. & 2015 Elsevier B.V. All rights reserved.

Keywords: Zone melting recrystallization Liquid phase crystallization Lateral laser epitaxy Preferential orientation

1. Introduction Up to now, the photovoltaic market is dominated by waferbased silicon solar cells using either multi- or monocrystalline silicon. However, the production process is cost and energy intensive and suffers from high material losses. While the thickness of the wafer is gradually reduced the thin-film technology, as a bottom-up approach, is working to achieve competitive electronic properties without these issues. The most promising technique to prepare crystalline silicon thin-films directly on a glass substrate with a thickness of 1–40 mm is based on liquid phase crystallization (LPC) using line-shaped energy sources such as an electron beam [1] or a continuous wave laser beam [2]. Process improvements [3], selection of suitable intermediate layers [4,5] and a change of the dopant type [6] were able to boost the material quality from initial 545 [1] to 656 mV [6]. These encouraging Voc results have been obtained on randomly oriented LPC films. Recently, we reported on the formation of a preferential crystallographic {100} orientation of the surface by the application of specific process parameters during laser crystallization [7]. In addition it was shown that the number of electrically active grain boundaries is reduced. In the present work we show a new route n

Corresponding author. Tel.: þ 49 30 8062 41396. E-mail address: [email protected] (S. Kühnapfel).

http://dx.doi.org/10.1016/j.solmat.2015.03.030 0927-0248/& 2015 Elsevier B.V. All rights reserved.

to further improve the crystallographic properties of the LPC silicon films in order to directly grow monocrystalline silicon on a large-area glass substrate. This would allow for a new high quality crystalline silicon technology which would be of high interest not only for silicon solar cells but also for other silicon based devices (e.g. thin-film transistors).

2. Sample preparation and characterization For the experiments, schematically shown in Fig. 1, two different types of samples were prepared. The structure of the standard samples we use in our experiments is schematically shown in Fig. 1(a). The related preparation sequence is as follows: (i) the glass substrate (Cornings EAGLE XGs) with an initial sample size of 5  5 cm2 is cleaned with a heated alkaline cleaning agent in an ultrasonic bath and subsequently dried with nitrogen, (ii) a reactively sputtered SiO2 layer of 200 nm thickness is deposited as a diffusion barrier to prevent contamination of the silicon during subsequent process steps [2,8], and (iii) a p-type 10 mm thick silicon film is deposited at a substrate temperature of 600 °C using high-rate (600 nm/min) electron-beam evaporation of silicon [9] and co-evaporation of boron from an effusion cell. Prior to the crystallization the samples are slowly pre-heated to Tsub. ¼ 700 °C in order to reduce thermal stress. The LPC process is carried out scanning a line-shaped continuous wave laser beam

S. Kühnapfel et al. / Solar Energy Materials & Solar Cells 140 (2015) 86–91

87

Fig. 1. Schematic drawing of a sample crystallized from left to right (a) without a seed layer, (b) with a buried seed layer and (c) with a seed stamp. In (d) a second scan perpendicular to the grain growth is performed.

with a center wavelength of 808 nm across the sample. The lineshaped laser beam features a Gaussian profile with a width of 0.177 mm (FWHM) in scan direction and a top-hat profile with a length of 31 mm perpendicular to the scanning direction. For all experiments described in this paper a scanning velocity of vscan ¼ 3 mm/s was used. The optical intensity was varied between Iopt. ¼ 2.0–2.9 kW/cm2. The Iopt. used in the specific experiments are mentioned in Section 3. In addition to the standard samples [see Fig. 1(a)] we also prepared samples with a buried silicon seed layer [see Fig. 1(b)]. For these samples we used Silicon-on-Glass (SiOG) substrates from Corning [10,11] as starting material. The SiOG-substrates consist of a 200 nm p-type monocrystalline silicon film on Cornings EAGLE XGs glass. The monocrystalline silicon film features a {100} surface orientation and a resistivity of 10–15 Ω cm. Except for a small stripe, which later on acts as a local buried seed layer, the silicon film of the SiOG-substrate was removed by wet chemical etching using HNO3, H2O and HF [50:20:1]. The remaining preparation sequence corresponds with one exception to the preparation of the standard sample sequence described above. In order to deposit the 10 mm thick electron-beam evaporated silicon directly on the 200 nm thin silicon seed layer, the silicon seed layer was masked during the deposition of the SiO2 layer. The resulting sample structure is shown in Fig. 1(b). The crystallization of the buried seed layer samples always starts on a region with the seed layer underneath and progresses towards the silicon on the SiO2 barrier [see Fig. 1(b)]. Both Figs. 1(c) and 1(d) depict specific configurations for the crystallization process which are described in Section 3. The characterization of the resulting crystalline silicon films was carried out using optical microscopy, electron backscatter diffraction (EBSD), scanning electron microscopy (SEM), photoluminescence (PL) imaging, ultraviolet–visible (UV–vis) spectroscopy and Hall measurements. For EBSD we used a cold field emission gun at an acceleration voltage of 25 kV and an emission current of 20 mA. All measured samples had a size of 850  850 mm2 and a lateral resolution of 6 μm. The inter-band PL images were taken at room temperature using a peltier cooled silicon detector. For excitation a wavelength of λ ¼650 nm at an intensity of approximately 1 sun was utilized. Both, excitation and detection were performed through the glass substrate, with an 820 nm low pass filter (blocking factor5) for the LED array and a Si3N4 coated (λ/4 at 1000 nm) GaAs wafer as a 880 nm high pass filter for the detector.

3. Results and discussion 3.1. Without seed layer Fig. 2 depicts the EBSD mapping of the surface normal [Fig. 2(a)] and the scanning direction [Fig. 2(b)] of a standard sample crystallized at a rather high intensity of Iopt. ¼2.9 kW/cm2 with a setup shown in Fig. 1(a). Crystallization was performed from left to right.

Fig. 2. EBSD map and the corresponding inverse pole figures of a sample crystallized at Tsub. ¼ 700 °C, vscan ¼ 3 mm/s and Iopt. ¼2.9 kW/cm2. (a) The surface normal and (b) the scanning direction.

Fig. 3. EBSD map and the corresponding inverse pole figures of a sample crystallized at Tsub. ¼ 700 °C, vscan ¼ 3 mm/s and Iopt. ¼2.0 kW/cm2. (a) The surface normal and (b) the scanning direction.

The corresponding inverse pole figure (IPF) of each mapping is displayed right next to it. As can be seen, the crystallization starts with a small grained matrix which serves as the seedling for the solidifying silicon melt and evolves upon laser crystallized distance covered to large elongated grains. Even though an increased fraction of {100} oriented grains towards the surface normal and a {101} in scanning direction are observed a large fraction of grains are randomly distributed as can be seen in the IPF's of Fig. 2. As a result a high amount of large angle grain boundaries are created. To this end various studies on the grain size and the nature of the grain boundaries have already been made showing their influence on the electrical properties of the resulting silicon layers. With exception of the Σ3 twin boundaries most high angle boundaries in particular the higher orders of Σ3 have a negative impact on the electrical properties [12–15]. It is often observed that Σ3n boundaries dissociate into lower order boundaries due to a reduction in surface energy [16], which eventually leads into a splitting and faceting. This is

88

S. Kühnapfel et al. / Solar Energy Materials & Solar Cells 140 (2015) 86–91

regarded to be an origin of highly dislocated regions, which can easily be decorated by other impurities [17,18]. It has already been shown that moderate scanning velocities and low temperature differences ΔT (Tmelt  Tsub.) lead to an increasing fraction of {100} oriented grains for the surface normal [7] if the sample setup shown in Fig. 1(a) is used. Fig. 3 exemplarily depicts such a layer. This sample has a well pronounced {100} surface texture [Fig. 3(a)]. However the scanning direction still comprises silicon grains which can randomly distribute within {100} and {101} direction as can be seen in the corresponding IPF of Fig. 3(b). Comparing the grain orientation at the start of the crystallization with that of a sample crystallized at higher Iopt., one can clearly see that the area of small grained silicon already consists of a high fraction of {100} oriented grains. This indicates that the starting region has a major impact on the texture formation. Thus we tried to directly influence the growth behavior of our multicrystalline silicon layer using different seed layer approaches.

Fig. 5. EBSD map and the corresponding inverse pole figures of a sample crystallized at Tsub. ¼ 700 °C, vscan ¼ 3 mm/s and Iopt. ¼ 2.0 kW/cm2 using a seed layer at the beginning of the crystallization process. (a) The surface normal and (b) the scanning direction. The region with the seed layer underneath is marked by the black dashed rectangle.

3.2. Multiple scans One method was established using multiple scans. Since a single scan LPC process as shown in Fig. 1(a) leads to elongated grains (see Figs. 2 and 3) one can rotate the sample by 90° after the first crystallization, performing a second scan perpendicular to the first scan [Fig. 1(d)]. The resulting layer is exemplarily shown in Fig. 4. The first scan was performed from bottom to top (left part of Fig. 4) using the adjusted crystallization parameters to gain a high fraction of {100} oriented grains for the surface normal. As can be seen the grains grow towards the scanning direction. The second scan was performed from left to right with the starting position indicated by the black dotted line. As can be seen the texture of the surface normal is maintained while the grain size and the amount of high angle boundaries is reduced. It is obvious that one could also think of even more scans. However additional scans result in a higher demand on the diffusion barrier. Therefore the first scan should only comprise a thin stripe to avoid unnecessary strain on the sample while the second one covers the complete sample area. 3.3. Buried seed layer and seed stamp We also tried to directly influence the growth behavior of our multicrystalline silicon layers by using buried monocrystalline seed layers [see Fig. 1(b)] at the start of the crystallization process. However in order to relinquish on a complicated real time power control of the laser, it is essential that the specific heat capacity and the conductance of the seed layer are negligibly small as compared to the layer to crystallize. Otherwise two scenarios occur. Either the crystallization of the absorber layer is not initiated due to the high specific heat capacity of the seed layer itself or

Fig. 4. EBSD map and the corresponding inverse pole figures of a sample crystallized at Tsub. ¼700 °C, vscan ¼ 3 mm/s and Iopt. ¼2.0 kW/cm2 using multiple scans. (a) The surface normal and (b) the scanning direction.

Fig. 6. EBSD map and the corresponding inverse pole figures of a sample crystallized at Tsub. ¼ 700 °C, vscan ¼ 3 mm/s and Iopt. ¼ 2.0 kW/cm2 using a seed layer at the beginning of the crystallization process. (a) The surface normal and (b) the scanning direction. The region with the seed layer underneath is marked by the black dashed rectangle.

for high Iopt. the crystallization initiates but dewetting occurs as soon as the laser has passed the area with the underlying seed layer. Upon crystallization on the seed layer a closed epitaxial layer is grown as can be seen in the black rectangles of Fig. 5. The monocrystalline starting region partially disintegrates into individual grains after the laser has passed the seed. The reason for the formation of grain boundaries is assumed to be due to stress within the layer stack upon crystallization and impurities e.g. particles on the samples during crystallization or pinholes in the as-deposited silicon thin film. Nevertheless, the silicon grains maintain the crystal information of the seed layer as they grow across the substrate resulting in a lateral epitaxial growth. Thus a preferential texture for the surface normal and in scanning direction can be achieved and even be controlled. Since we are using a {100}-Wafer as a seed we can vary the orientation of the scanning direction from {101} up to {100} simply by either etching the seed stripe in an angle of 45° or rotating the laser source by that angle. This is exemplarily shown in Fig. 6 for a sample that has been etched. Again the silicon is taking over the crystal orientation of the seed layer underneath (black rectangle) and grows laterally across the substrate resulting in a {100} texture of the surface normal and in scanning direction. By being able to control the grain orientation of the surface normal as well as for the scanning direction a strong reduction of grain boundary angles can be observed. The material exhibits more than 90% Σ1 boundaries and only some Σ3 twin boundaries which are known to have no major electrical impact. Higher orders of the Σ3 and other CSL boundaries have not been detected.

S. Kühnapfel et al. / Solar Energy Materials & Solar Cells 140 (2015) 86–91

89

Instead of a buried seed layer [Fig. 1(b)] we also used a seed stamp [Fig. 1(c)]. As seed stamp, the above described SiOG-substrates were utilized. However, the experimental setup has to be adjusted. In this case the samples have to be flipped upside down and crystallization has to be performed through the glass substrate with the sample pressed onto the seed stamp during the crystallization. Just as the samples with a buried seed layer a preferential texture formation was observed (not shown here). Details on this process will be published separately. Using this experimental arrangement one could also think of using commercially available wafers instead of thin SiOG seeds, which are pre-heated separately in order to overcome the specific heat capacity limitation mentioned before.

4. Application In order to be applicable e.g. as solar cell material, the preferential orientation has to be sustained on large scale. Hence microscope mappings of the surface have been made to investigate the grain evolution. To visualize individual grains and their different orientation the surface was etched for 300 s with a KOH isopropyl alcohol solution. Due to the anisotropic etching an orientation dependent material removal of up to 2 mm can be observed whereby upright pyramids are created at former {100}surfaces, while orientations differentiating from the {100}-orientation result in tilted structures. Thus the light scattering of the resulting layer strongly depends on the orientation of the grains. In Fig. 7(a) an optical micrograph of a sample such as the one depicted in Fig. 2 is shown. As can be seen, the resulting layer exhibits randomly oriented grains with a length of several centimeters and a width of a few millimeters. Fig. 7(b) depicts a sample crystallized with a decreased optical intensity of 2.0 kW/cm2 (see Fig. 3). The grain size is slightly increased compared to the sample crystallized at higher energies and a preferential texture begins to

Fig. 7. Optical microscope image of KOH/IPA etched sample (a) without a seed layer crystallized at Iopt. ¼ 2.9 kW/cm2, (b) at 2.0 kW/cm2 and (c) with a buried seed layer crystallized at Iopt. ¼ 2.0 kW/cm2. In (d) the photoluminescence image of the third sample is depicted.

Fig. 8. Absorption of a (black) planar sample, a random pyramid textured sample (blue) without a preferential texture and (red) with a preferential surface texture as indicated by the SEM inlets. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

emerge, recognizable by the similar scattering behavior (color) of the grains. In contrast to this Fig. 7(c) depicts a sample crystallized at Iopt. ¼2.0 kW/cm2 using a thin buried seed layer schematically depicted in Fig. 1(b). Using this approach the size of the grains is significantly increased and the texture of the seed sustained upon laser crystallized distance covered resulting in large grains extending across the whole substrate area. Fig. 7(d) depicts the inter-band PL image of the sample in Fig. 7 (c). The image is scaled from 0 to 5000 counts. Except the influence from a few grain boundaries, a homogenous lifetime distribution of the material is observed with peak values of approximately 5000 counts. In contrast, the luminescence of the bare glass [Fig. 7(d) left black area] features peak values of 1000 counts which is 4–5 times lower than the luminescence of the silicon. Up to now inter-band PL measurements of multicrystalline silicon films grown on glass were very challenging [19,20] because both, silicon and glass luminescence were within the same magnitude. Thus the high silicon signal with regard to the glass signal in Fig. 7(d) demonstrates the high silicon quality. Fig. 8 depicts the absorption spectra measured of a planar silicon layer (black curve), one with strongly tilted pyramids due to a random surface texture (blue curve) and a layer with a preferential {100} grain orientation of the surface normal resulting in upright pyramids after etching (red curve). Measurements have been performed in substrate (from silicon side) configuration without anti-reflective coatings, textured light trapping foils and/or back reflectors. It can be seen that due to reduced reflective losses a uniform increase in absorption occurs between 300 and 700 nm. The additional increase in absorption within the near infrared spectrum results because of an enhancement of the optical path length in our thin silicon films. This can particularly be seen for the absorber with a preferential surface texture of {100}. Integrating the photon flux of the AM 1.5 g spectrum over the wavelength region absorbed by silicon (EG ¼1.12 eV at 300 K) leads to a maximum current density of  45 mA that could be used for electric conversion. The maximum current densities of those three layers are  19,  23 and  34 mA/cm2 for planar, tilted and upright structures respectively. Those values would increase even further using antireflective coatings, textured light trapping foils and/or back reflectors. Hence one advantage of a preferential surface texture of {100} is the possibility to reduce reflective losses and to enhance the optical path length of our silicon thin films by implementing lighttrapping concepts based on random pyramids. Moreover the resulting {111} facets are known to form a well passivated a-Si:H/cSi interface [21], which is beneficial for the preparation of an a-Si:H/ c-Si heterojunction solar cell. In addition to improved optical properties an additional preferential grain orientation of the scanning direction leads to a

90

S. Kühnapfel et al. / Solar Energy Materials & Solar Cells 140 (2015) 86–91

Table 1 Comparison of the calculated mhole of c-Si with measured hall results of samples crystallized at Tsub. ¼700 °C, vscan ¼3 mm/s and Iopt. ¼ 2.0 kW/cm2 with and without seed layers.

c-Si LPC w/o seed LPC w/ seed

NA (cm  3)

mhole (cm2/V s)

%c-Si (%)

1.0  1017 0.9  1017 1.0  1017

317 215 249

– 66 79

Eq. (1) Fig. 7(b) Fig. 7(c)

significant reduction of large angle grain boundaries and their density, which eventually should lead to improved electronic properties of the material [as can be seen in Fig. 7]. Hence Hall measurements have been performed on layers crystallized at Tsub. ¼700 °C, vscan ¼ 3 mm/s and Iopt. ¼2.0 kW/cm2 with and without seed layer. The sample size for all Hall structures was 5  5 mm2. The measurements were performed at 300 K using the Van der Pauw method. In order to compare our measured hole mobility values (mhole) at a given charge carrier density (NA) with the theoretical values for monocrystalline silicon wafers we used the empirical relationship [22] described below:

μ hole = μmin +

μmax − μmin 1+

NA α N0

( )

(1)

As model parameters we used the following values: mmin ¼ 44.5 cm2/V s, mmax ¼470 cm2/V s, N0 ¼2.23  1017 cm  3 and α ¼ 0.719 [23]. The best results of each layer and the calculated c-Si reference are shown in Table 1. As can be seen samples without the seed layer approach, which have been processed with adjusted crystallization parameters, reached a maximum mhole of 66%. However, the results strongly depend on the number and type of grain boundaries within the hall structures, as they result in a strongly differing sheet resistance along and perpendicular to the growth direction of the grains. Hence values down to  160 cm2/V s (or even not measurable using the Van der Pauw method) are observed. Samples with the seed layer, however, have a very homogeneous mobility distribution across the substrate area with less than 5% overall deviation and reach a maximum value of almost 80% of the theoretically possible limit for monocrystalline silicon. Whether this increase results from a reduction of grain boundaries, an improved crystal quality or both is subject to future investigations.

5. Summary We have investigated and implemented different seed layer approaches for our laser assisted liquid phase crystallization process. The resulting silicon layers sustain the crystal information of the seed layer as they grow laterally across the substrate exhibiting a preferential orientation of the surface normal and in scanning direction. Hence the LPC layers can easily be tuned to meet their respective requirements. For the fabrication of thin film solar cells that would be a {100} orientation of the surface normal in order to implement pyramid textures and a {100} orientation in scanning direction because of its high growth velocity and low defect density. Moreover the grains size is strongly increased for samples with seed layers, with single grains almost covering the whole substrate area. This consequently leads to a very homogeneous and more than 13% increased mobility across the entire sample and apart from a few electrical active grain boundaries to a very homogenous lifetime distribution measured using PL imaging.

Acknowledgments The authors would like to thank Dr. J.G. Couillard from Corning for providing SiOG-substrates, M. Reiche for the deposition of the silicon layers and C. Klimm for EBSD and SEM measurements. Financial support is acknowledged from the Bundesministerium für Umwelt Naturschutz und Reaktorsicherheit (BMU) under Contract no. 0325446A.

References [1] D. Amkreutz, J. Müller, M. Schmidt, T. Hänel, T.F. Schulze, Electron-beam crystallized large grained silicon solar cell on glass substrate, Prog. Photovolt. Res. Appl. 19 (2011) 937–945. http://dx.doi.org/10.1002/pip.1098. [2] J. Dore, R. Evans, B.D. Eggleston, S. Varlamov, M.A. Green, Intermediate layers for thin-film polycrystalline silicon solar cells on glass formed by diode laser crystallization, MRS Online Proc. Libr. 1426 (2012) 63–68. http://dx.doi.org/ 10.1557/opl.2012.866. [3] D. Amkreutz, J. Haschke, T. Häring, F. Ruske, B. Rech, Conversion efficiency and process stability improvement of electron beam crystallized thin film silicon solar cells on glass, Sol. Energy Mater. Sol. Cells 123 (2014) 13–16. http://dx. doi.org/10.1016/j.solmat.2013.12.021. [4] J. Dore, R. Evans, U. Schubert, B.D. Eggleston, D. Ong, K. Kim, et al., Thin-film polycrystalline silicon solar cells formed by diode laser crystallisation, Prog. Photovolt. Res. Appl. 21 (2013) 1377–1383. http://dx.doi.org/10.1002/pip.2282. [5] J. Dore, D. Ong, S. Varlamov, R. Egan, M.A. Green, Progress in laser-crystallized thin-film polycrystalline silicon solar cells: intermediate layers, light trapping, and metallization, IEEE J. Photovolt. 4 (2014) 33–39. http://dx.doi.org/10.1109/ JPHOTOV.2013.2280016. [6] J. Haschke, D. Amkreutz, L. Korte, F. Ruske, B. Rech, Towards wafer quality crystalline silicon thin-film solar cells on glass, Sol. Energy Mater. Sol. Cells 128 (2014) 190–197. http://dx.doi.org/10.1016/j.solmat.2014.04.035. [7] S. Kühnapfel, N.H. Nickel, S. Gall, M. Klaus, C. Genzel, B. Rech, Preferential {100} grain orientation in 10μm-thick laser crystallized multicrystalline silicon on glass, Thin Solid Films 576 (2015) 68–74. http://dx.doi.org/10.1016/j. tsf.2015.01.006. [8] C. Becker, D. Amkreutz, T. Sontheimer, V. Preidel, D. Lockau, J. Haschke, et al., Polycrystalline silicon thin-film solar cells: status and perspectives, Sol. Energy Mater. Sol. Cells 119 (2013) 112–123. http://dx.doi.org/10.1016/j. solmat.2013.05.043. [9] T. Sontheimer, P. Dogan, C. Becker, S. Gall, B. Rech, U. Schubert, et al., 2009. 6.7% Efficient poly-Si thin film mini-modules by high-rate electron-beam evaporation, in: Proceedings of the 24th European Photovoltaic Solar Energy Conference, 2009, pp. 2478–2481. doi: 10.4229/24thEUPVSEC2009-3AV.1.1. [10] D.F. Dawson-Elli, C.A.K. Williams, J.G. Couillard, J.S. Cites, R.G. Manley, G. Fenger, et al., Demonstration of high performance TFTs on silicon-on-glass (SiOG) substrate, ECS Trans. 8 (2007) 223–228. http://dx.doi.org/10.1149/ 1.2767312. [11] J.S. Cites, J.G. Couillard, K.P. Gadkaree, Silicon-on-glass (SiOG) substrate technology: process and materials properties, IEEE Int. 2009 (2009) 1–2. http://dx. doi.org/10.1109/SOI.2009.5318764. [12] M. Kohyama, R. Yamamoto, M. Doyama, Structures and energies of symmetrical (011) tilt grain boundaries in Silicon, Phys. Status Solidi B 137 (1986) 11–20. http://dx.doi.org/10.1002/pssb.2221370102. [13] J. Chen, T. Sekiguchi, D. Yang, F. Yin, K. Kido, S. Tsurekawa, Electron-beaminduced current study of grain boundaries in multicrystalline silicon, J. Appl. Phys. 96 (2004) 5490–5495. http://dx.doi.org/10.1063/1.1797548. [14] W. Seifert, D. Amkreutz, T. Arguirov, H.M. Krause, M. Schmidt, Analysis of electron-beam crystallized large grained Si films on glass substrate by EBIC, EBSD and PL, Solid State Phenom. 178–179 (2011) 116–121, www.scientific. net/SSP.178-179.116. [15] S. Tsurekawa, K. Kido, T. Watanabe, Measurements of potential barrier height of grain boundaries in polycrystalline silicon by Kelvin probe force microscopy, Philos. Mag. Lett. 85 (2005) 4–49. http://dx.doi.org/10.1080/ 09500830500153859. [16] A. Garg, W.A.T. Clark, J.P. Hirth, Dissociated and faceted large-angle coincidentsite-lattice boundaries in silicon, Philos. Mag. A 59 (1989) 479–499. http://dx. doi.org/10.1080/01418618908229780. [17] Z.-J. Wang, S. Tsurekawa, K. Ikeda, T. Sekiguchi, T. Watanabe, Relationship between electrical activity and grain boundary structural configuration in polycrystalline silicon, Interface Sci. 7 (1999) 197–205. http://dx.doi.org/ 10.1023/A:1008796005240. [18] T. Ervik, M. Kivambe, G. Stokkan, B. Ryningen, O. Lohne, Dislocation formation at ∑ ¼27a boundaries in multicrystalline silicon for solar cells, in: Proceedings of the 26th European Photovoltaic Solar Energy Conference, 2011, pp. 1895– 1899. doi: 10.4229/26thEUPVSEC2011-2BV.4.31.in. [19] S. Schönau, J. Rappich, M. Weizman, D. Amkreutz, B. Rech, Photoluminescence study of polycrystalline silicon thin films prepared by liquid and solid phase crystallization, Phys. Status Solidi A 210 (2013) 1652–1656. http://dx.doi.org/ 10.1002/pssa.201228809.

S. Kühnapfel et al. / Solar Energy Materials & Solar Cells 140 (2015) 86–91 [20] A. Teal, J. Dore, S. Varlamov, Photoluminescence imaging of thin film silicon on glass, Sol. Energy Mater. Sol. Cells 130 (2014) 1–5. http://dx.doi.org/10.1016/j. solmat.2014.06.024. [21] H. Angermann, J. Rappich, L. Korte, I. Sieber, E. Conrad, M. Schmidt, et al., Wetchemical passivation of atomically flat and structured silicon substrates for solar cell application, Appl. Surf. Sci. 254 (2008) 3615–3625. http://dx.doi.org/ 10.1016/j.apsusc.2007.10.099.

91

[22] D.M. Caughey, R.E. Thomas, Carrier mobilities in silicon empirically related to doping and field, Proc. IEEE 55 (1967) 2192–2193. http://dx.doi.org/10.1109/ PROC.1967.6123. [23] B.V.V. Zeghbroeck, Principles of Semiconductor Devices and Heterojunctions, Auflage: 1, Prentice Hall, Upper Saddle River, N.J.; London, 2010.