Thin-film microcrystalline silicon solar cells illuminated through the n layer were studied and compared with classical p-layer illuminated cells. To investigate the corresponding charge carrier extraction properties, variation of the intrinsic absorber layer thickness was carried out. It was found that the J -V characteristic and the quantum efficiency of the n-and p-side illuminated cells are almost identical in the thickness range investigated, up to 7 m. No differences in the collection of

more »

... hotogenerated electrons or holes are observed. Hence, the illumination side of c-Si:H single junction solar cells of conventional thickness may be randomly chosen without adverse effect on their performance. Microcrystalline silicon (c-Si:H) prepared by plasma enhanced chemical vapor deposition ͑PECVD͒ has received considerable attention since its first successful application as an absorber layer in thin-film solar cells 1-3 and much improvement in the solar cell efficiencies has been obtained since. However, there is no detailed knowledge of the transport behavior, in particular with relation to the performance of the material in solar cells. As an example, deposition conditions near the transition to the amorphous growth regime yield the highest solar cell efficiency. 4, 5 In this context, the question arises as to what extent solar cells with c-Si:H absorber layers are limited in performance by the features of the amorphous silicon (a-Si:H) device physics. As is widely known, solar cells with a-Si:H absorber layers require illumination through the p side for optimum stabilized performance. The origin of this is the complex relationship between the carrier mobility, which is much lower for holes than for electrons, and the charging of ambipolar defects, causing electric field distortions, especially in the degraded state, where the defect density is high. This has been studied in great detail over the last two decades ͑see Refs. 6 and 7 for early experimental and theoretical work and Ref. 8, and references cited therein͒. Whether effects similar to those of collection asymmetry in thick degraded a-Si:H cells also apply to c-Si:H devices is the topic of the present study. The solar cells were prepared by PECVD in a multichamber system at a plasma excitation frequency of 95 MHz. 5,9 Layer deposition sequences of p -i -n and n -i -p were applied to glass substrates coated with textured ZnO ͑superstrate design͒ for illumination from the p-layer and n-layer sides, respectively. The deposition conditions were independently optimized for both cell types, resulting in a silane concentration (SC)(ϭ͓SiH 4 ͔/͓SiH 4 ϩH 2 ͔) of 5.5% and a substrate temperature (T S ) of 250°C for the n -i -p cells, while for the p -i -n cells a SC of 5.0% and a T S of 200°C were used. From previous studies 4,5,9 we know that both conditions resulted in high quality materials having very similar carrier transport properties when employed in solar cells. For example, the fill factor ͑FF͒, short-circuit current density ( j sc ) and quantum efficiencies were identical. However, the use of higher SC values had an ͑presently not understood͒ influence on the solar cell device performance, as characterized by the shift of dark J -V curves and a corresponding higher opencircuit voltage V oc . 4, 5, 9 To indicate the illumination side of the solar cells in the present study, an arrow is used to complement the deposition sequence designations of the p -i -n and n -i -p cases. Therefore, p-side illuminated p -i -n cells are indicated by →p -i -n, while n-side illuminated n -i -p cells are indicated by →n -i -p. It should be pointed out that, despite the reverse deposition sequence, both structures employed the same device preparation conditions, i.e., both were deposited on ZnO/glass substrates and were illuminated through glass. Figure 1 shows a comparison of the quantum efficiency ͑QE͒ curves measured for 1 m thick c-Si:H→p -i -n and →n -i -p cells. These data were obtained from spectral response measurements without any bias illumination. It was found that identical QE curves were obtained using a differential spectral response method in which bias illumination was used. These results suggest that the collection behavior a͒ Also at