**Crystalline Silicon Thin Film Solar Cells**

Fritz Falk and Gudrun Andrä

*Institute of Photonic Technology Germany* 

#### **1. Introduction**

136 Solar Cells – Thin-Film Technologies

Ashok, S.; Sharma, P.P.; Fonash, S.J. (1980) IEEE Transactions on Electron Devices 27, (1980)

Bruk, L.; Fedorov, V.; Sherban, D.; Simashkevich, A.; Usatii, I.; Bobeico, E.; Morvillo, P.

Baik, D.G.; Cho, S.M. (1999). "Application of sol-gel derived films for ZnO/n-Si junction

Chaabouni, F.; Abaab, M. Rezig, B. (2006). "Characterization of n-ZnO/p-Si films grown by

Cheknane, Ali (2009) "Analytical modelling and experimental studies of SIS tunnel solar

Chen, X. D.; Ling, C. C.; Fung, S.; Beling, C. D. (2006). "Current transport studies of ZnO/p-

Canfield, L.R.; Kerner, J.; Korde, R. (1990). "Silicon photodiodes optimized for the EUV and

Granqvist, C.G. (2007) "Transparent conductors as solar energy materials: A panoramic

Granqvist, Claes G. (2007) "Transparent conductors as solar energy materials: A panoramic review", Solar Energy Materials & Solar Cells 91 (2007) pp.1529–1598 He, B.; Ma, Z.Q. et al., (2009). "Realization and Characterization of ITO/AZO/SiO2/p-Si SIS Heterojunction", Superlattices and Microstructures 46, (2009) pp.664-671 He, B.; Ma, Z.Q. et al., (2010). "Investigation of ultraviolet response enhanced PV cell with

Koida, T.; Fujiwara, H.; Kondo, M. (2009) "High-mobility hydrogen-doped In2O3

Kittidachachan, P.; Markvart, T.; Ensell, G.J.; Greef, R.; Bagnall, D.M. (2005). "An analysis of

Mridha, S.; Basak, Durga. (2007). "Ultraviolet and visible photoresponse properties of n-ZnO/p-Si heterojunction", Journal of Applied Physics, Vol 101, (2007) pp.083102 Ramamoorthy, K.; Kumar, K.; Chandramohan, R.; Sankaranarayanan, K. (2006) "Review on

Song, Dengyuan; Aberle, Armin G.; Xia, James (2002). "Optimisation of ZnO:Al films by

Singh, R.; Rajkanan, K.; Brodie, D.E.; Morgan, J.H. (1980) IEEE Transactions on Electron

Wenas, Wilson W.; Riyadi, S. (2006). "Carrier transport in high-efficiency ZnO/SiO2/Si solar cells", Solar Energy Materials and Solar Cells 90,(2006) pp.3261-3267

Materials Science and Engineering B 159–160 (2009) pp.282–285

solar cells, Thin Solid Films 354 (1999) pp.227-231

cells", J. Phys. D: Appl. Phys. 42 (2009) pp.115302-115307

Applied Physics Letters, vol 88, (2006) pp.132104-132107

review", Sol. Ener. Mater. Sol. Cells 91, (2007) pp.1529-1598

Energy Materials & Solar Cells 93 (2009) pp.851–854

SPIE vol 1344 , (1990) pp. 372—377

(2010) pp.1028-1037

1103-1106

(2006) pp.1–15

195, (2002) pp.291-296

Devices 27, (1980) pp. 656-662

(2009) "Isotype bifacial silicon solar cells obtained by ITO spray pyrolysis",

magnetron sputtering", Superlattices and Microstructures, vol 39, (2006) pp.171-178

Si heterostructures grown by plasma immersion ion implantation and deposition",

soft xray regions", EUV.X-Ray.andGamma-Ray Instrumentation for Astronomy.

silicon-based SINP configuration", SCIENCE CHINA Technological Sciences 53,

transparent conductive oxide for a-Si:H/c-Si heterojunction solar cells", Solar

a "dead layer" in the emitter of n+pp+ solar cells", Photovoltaic Specialists Conference, Conference Record of the Thirty-first IEEE, vol 3-7, (Jan., 2005) pp.

material properties of IZO thin films useful as epi-n-TCOs in opto-electronic (SIS solar cells, polymeric LEDs) devices", Materials Science and Engineering B 126

change of sputter gas pressure for solar cell application", Applied Surface Science

**10. References** 

pp. 725-730

In the last few years the marked share of thin film solar cells increased appreciably to 16.8% (in 2009). The main part of that increase refers to CdTe modules (9.1%) followed by silicon thin film cells, that is amorphous silicon (a-Si) cells or tandem cells consisting of a-Si and nanocrystalline silicon (µc-Si). For a review on thin film solar cells in general see (Green, 2007) and on a-Si/µc-Si cells see (Beaucarne, 2007). The a-Si cells suffer from a low efficiency. In the lab the highest efficiency up to now is 10.1% on 1 cm² (Green et al., 2011), whereas in the industrial production modules reach about 7%. In order to achieve the required electronic quality of hydrogenated amorphous silicon (a-Si:H), low deposition rate (max. 50 nm/min) PECVD (plasma enhanced chemical vapour deposition) is used for deposition which makes production more expensive as compared to CdTe modules. This is even worse for the layer system in a-Si/µc-Si tandem cells for which the more than 1 µm thick nanocrystalline µc-Si layer is deposited by PECVD, too, however with much lower deposition rates in the 10 nm/min range. Cells consisting just of µc-Si reached 10.1% efficiency (Green et al., 2011), just as a-Si-cells, whereas tandem cells arrived at 11.9%, both for lab cells, whereas in production the results are below 10%. The low deposition rate combined with the limited efficiency, make these cells not too competitive compared to CdTe cells, which, at lower cost, reach 11% in industrial production, or to CIGS (Copperindium-gallium-diselenide) cells with similar efficiencies.

As an alternative, polycrystalline (grains in the µm range) or multicrystalline (grains >10 µm) silicon thin film solar cells receive growing interest (Beaucarne et al., 2006). The present paper reviews the status of these cells, and on the other hand gives details of laser based preparation methods, on which the authors have been working for many years.

Both types, poly- and multicrystalline silicon thin film cells, are prepared by depositing amorphous silicon followed by some crystallization process. One main advantage of the crystallization process is that the electronic quality of the virgin a-Si is not important. Therefore high rate deposition processes such as electron beam evaporation or sputtering can be used which are much less expensive as compared to low rate PECVD. In case of sputtering doped thin films can be deposited by using doped sputtering targets, whereas in electron beam evaporation the dopands are coevaporated from additional sources. So, in these deposition processes the use of toxic or hazardous gases such as silane, phosphine or diborane is avoided, reducing the abatement cost.

Polycrystalline silicon layers for solar cells can be prepared in a single crystallization step. The layer system containing the doping profile is deposited in the amorphous state and is

Crystalline Silicon Thin Film Solar Cells 139

thermal annealing step at 900°C follows. The silicon layers are passivated by a hydrogen plasma treatment. Finally rather demanding structuring and contacting processes follow. In production, modules 1x1.4 m² in size reached about 7% efficiency. In the lab 10.4% efficiency were achieved on 92 cm² minimodules (Keevers et al., 2007). The production was stopped, probably because of the high cost PECVD deposition, which was used because the method was the only one available for silicon deposition in the m² range. In the lab, high rate electron beam evaporation was tested as an alternative which delivered minimodules with the efficiency of 6.7%, similar to that of the industrially produced modules (Egan et al., 2009;

The grain size originating in the furnace anneal is dictated by the interplay of crystal nucleation within the amorphous matrix and growth of the nuclei (see Sect. 5). One can influence both processes by the temperature of the annealing step. Practically, however, there is not much choice. At lower temperature the annealing time required for complete crystallization would reach unrealistic high values so that this is not possible in production. Higher temperatures are not endured by the glass substrate for the time span needed for crystallization. Even at 600°C 18 h are required for crystallization and high temperature

As an alternative for the furnace crystallization pulsed excimer laser crystallization via the melt is a process industrially used in flat panel display production. For this application, however, rather thin films (<100 nm) are required and the resulting grain size is rather small, typically below 1 µm. In the context of solar cell preparation requiring films thicker

As mentioned in the last paragraph, grains larger than about 1 µm cannot be prepared by direct deposition of crystalline silicon, nor by solid phase crystallization of a-Si nor via melting a-Si by short laser pulses. Large grains can be produced from the melt only if the melt is cooled below the equilibrium melting point slowly so that the melt stays long enough in a region of low nucleation rate and there is time enough for the few nucleating crystallites to grow to large size. Low cooling rate means low heat flow into the substrate following from a low temperature gradient in the substrate. This can be achieved if the melting time of the silicon layer is larger than in excimer laser crystallization, i.e. much larger than 100 ns. To reach longer melting times the energy for melting has to be delivered on a longer time scale. For energy delivery scanned electron beams or scanned laser beams have been used. However, the longer melting time has the consequence, that dopand profiles, introduced into the virgin a-Si for emitter, absorber, and back surface field, get intermixed due to diffusion. Typical diffusion constants in liquid silicon are in the 10-4 to 10-3 cm²/s range (Kodera, 1963) so that dopands will intermix over a distance of 1 µm within 10 to 100 µs. Nevertheless a one-step crystallization procedure for a solar cell layer system has been done by electron beam melting, discussed in Sect. 3.2. Alternatively a two-step procedure has been used. In a first step a thin seed layer is crystallized to large grains from a-Si by laser irradiation. In a second step the seed is thickened epitaxially. Seed and epitaxial layer can be differently doped so that the seed can act as the emitter and the epitaxial layer as the absorber of the solar cell. Alternatively, the seed may act as a highly doped back surface field layer with the epitaxial layer acting as a moderately doped absorber. The

resistant borosilicate glass has to be used instead of a much cheaper soda lime glass.

than 1 µm this method has been mentioned only rarely (Kuo, 2009).

emitter is generated on top in a third preparation step.

**3. Multicrystalline silicon thin film solar cells: grains > 10 µm** 

Sontheimer et al., 2009).

**3.1 Basic considerations** 

crystallized in a furnace to result in grains about 1 µm in size. This process had been industrialized by the company CSG and is described in Sect. 2. In the lab CSG reached 10.4% efficiency on 90 cm² minimodules (Keevers et al., 2007). Alternatively pulsed excimer laser melting and solidification can been used, which is a standard process in flat panel display production (Sect. 2).

Preparation of multicrystalline silicon thin film solar cells with grains exceeding 10 µm in size is under investigation. This topic is extensively dealt with in Sect. 3. Usually a two-step preparation scheme is used. In a first step a multicrystalline thin seed layer with the desired crystal structure is prepared (Sect. 3.3), which in a second step is epitaxially thickened (Sect. 3.4). For both, seed layer preparation and epitaxial thickening, different processes have been tested. There are, however, attempts to crystallize the complete layer stack of a thin film solar cell in one electron beam melting step (sect. 3.2). The idea for the multicrystalline cells is that in the large grains recombination is reduced, if the crystal quality is high enough, so that the efficiency should exceed that of cells with µm sized grains. Particularly, if the ratio of grain size to layer thickness is large (e.g. 50), such as in multicrystalline wafer cells, a similar efficiency potential is expected. This would require 100 µm large grains for 2 µm thick silicon layers. The preparation methods for large grained multicrystalline silicon layers divide in low and high temperature processes. The high temperature processes are rather straight forward for producing large grains (Beaucarne et al., 2004). However, temperature resistant substrates are required which are expensive. Much more demanding are preparation methods working at temperatures endured by low cost substrates such as glass. One such method is diode laser crystallization. The epitaxial thickening processes, as well, divide in high and low temperature processes with the same drawbacks and advantages. Several methods are presented in Sect. 3. The result of seed layer preparation as well as epitaxial thickening, via melt or in the solid state, depends on temperature history and is explained by the kinetics of phase transformation. The basic notions of this theory as far as they are important for silicon thin film solar cell preparation, are summarized in Sect. 5. Postcrystallization treatments such as rapid thermal annealing and hydrogen passivation are explained in Sect. 4.

Even single crystalline silicon thin film cells have been prepared by a transfer process starting from a silicon wafer from which a layer is detached and epitaxially thickened to several 10 µm thickness (Reuter et al., 2009; Brendel, 2001; Brendel et al., 2003; Werner et al., 2009) to reach an efficiency of 17%. This type of cells, which are much thicker than the polyand multicrystalline silicon thin film cells, and which cannot be prepared in the typical sizes of thin film technology such as > 1 m², is not the topic of this paper.
