1. Introduction

The increasing demand for Si-based solar cells has led to a greater emphasis on the development of thin-film Si cells due to the need to make efficient use of Si resources. However, thin-film Si solar cells suffer from relatively poor light absorption characteristics in the wavelength range of 600-1000 nm, which is due to the small absorption coefficient of Si. To address this issue, techniques for managing and trapping light inside the cells and thus enhancing absorption are crucially important [1–24]. To enhance the optical absorption in thin-film silicon solar cells, randomly textured structures have been widely used to scatter light [1–3,8,9,12]. However, there is an upper limit on the absorption referred to as the Lambertian-limit [14]. In addition, it is difficulat to regulate defect formation in random structures; particularly structures involving rough surfaces or V-shaped valleys induce the formation of cracks/voids in photovoltaic layer, which degrade the cell peformances [8,15–17]. Furthermore, the application of regularly arranged structures has been also studied for improving absorption efficiency [4–7,10,11,13]. In particular, the honey-comb textures with a period of micrometers, which are favorable to control the formation of unintentional cracks formed in microcrystalline silicon (μc-Si) layer, have shown the highest performances of micro-crystalline silicon cells using the 1.8-μm-thick photovoltaic layer [13].

Recently, photonic crystals (PCs), which have sub-micrometer scaled periodic patterns, have been investigated for more efficient light trapping in thin film layers [18–27]. In our previous works [24,25], we have discussed the guidelines for enhancing optical absorption using large-area resonant modes formed by incorporating photonic crystals. We analytically showed that introduction of the PC structures breaks through the absorption limit in randomly textured structures, or the Lambertian limit, and experimentally demonstrated ultra-thin microcrystalline silicon solar cells having a ~500-nm thick photovoltaic layer, which is ~1/4 the thickness of typical microcrystalline silicon cells. Incorporation of the photonic crystal yielded a 1.31-fold enhancement of the short-circuit current density compared to a cell with no photonic crystal. This led to a 1.25-fold enhancement of the efficiency, from 5.6% to 6.8%. Moreover, it is noteworthy that, as a result of the calculations of absorption by considering the experimental structure, we have predicted the efficiency would be more than 9% even in an ultra-thin photovoltaic layer with ~500nm thickness. Nevertheless, our previous experimental results (V_oc = 0.505 V, FF = 0.692, and J_sc = 19.6 mA/cm²) correspond to lower values of V_oc and FF than typically found in μc-Si solar cells, as well as a lower J_sc than theoretically predicted. Therefore, to increase the conversion efficiency it is necessary to consider how each of these parameters can be improved. Here we address this issue from the viewpoint of the quality of the intrinsic μc-Si grown on photonic-crystal structures with lattice constants of several hundred nanometers. We focus on the crystalline fraction (X_c) and on the growth directions of the crystalline grains. By addressing both of these issues during fabrication of the cell, we achieve an efficiency of 8.7% in an ultra-thin (~500 nm), homo-junction μc-Si solar cell. We also discuss the possibility of further improving the efficiency by performing calculations based on the fabricated structure where we consider parasitic losses due to the electrodes.

2. Concept of the enhancement of optical absorption by photonic crystals

We at first explain our concept for broadband light absorption enhancement using large-area resonant modes formed by incorporating photonic crystals, which possess wavelength-scale periodic structures as shown in Fig. 1(a). We have proposed that the in-plane band-edge resonant effect of the Γ-point resonant mode, whose group velocity is zero, can be utilized in two-dimensional (2D) photonic crystals [28,29]. Since the incident light from the normal direction can be coupled and trapped to such resonant modes, an enhancement of absorption is expected. We have revealed that a number of resonant modes can be created in the wavelength range of interest (600-1000 nm) by utilizing multiple transverse modes in the cell-thickness (vertical) direction and also the band-folding effect in super-lattice structures [24–26].

 

Fig. 1 (a) Schematic image of a photonic-crystal solar cell. (b) Photonic band diagram, assuming a 500-nm-thick Si layer and a lattice constant of 600 nm which corresponds to a 2 × 2 super-lattice photonic crystal with a fundamental lattice constant of ~300 nm.

Download Full Size | PDF

In Fig. 1(b), we show a photonic band diagram to explain the formation of multiple band edges by introducing photonic crystals. We assume a 500-nm-thick Si layer and a lattice constant of 600 nm, taking our previous fabrication into account [24]. In the calculation, we considered the multiple transverse modes in the thickness direction and the band-folding effect by the periodic structure, but did not incorporate an actual photonic-crystal solar cell structure for simplicity. The black line indicates the dispersion of the fundamental mode and the colored lines show those of higher-order modes confined in the vertical direction. We find that some transverse modes in the vertical direction appear, and they provide a number of resonant modes at the Γ point. Here, in consideration of the numbers and frequencies of the Γ-point modes, the employment of a lattice constant of 600 nm corresponds to the usage of 2 × 2 super-lattice photonic crystal with a fundamental lattice constant of ~300 nm. When using the lattice constant of ~300 nm, we obtain the lowest-frequency Γ-point mode at around 1000 nm [25], which is the longest wavelength of the target region (600-1000 nm). However, in this case, although several other resonant modes are created at wavelengths shorter than 1000 nm, the density of the modes remains at a lower value. In contrast, when applying the larger period (~600 nm) by using the concept of a super-lattice, we can create a much higher density of resonant modes due to the band-folding effect as shown in Fig. 1(b). There is a possibility to further increase the number of resonant modes by using a 4 × 4 super-lattice [25]. We have revealed that absorption enhancement using such photonic crystals, which provide a large number of resonant modes, exceeds that obtained by using Lambertian textures [25].

3. Analysis of previously reported photonic-crystal solr cell

To experimentally obtain high efficiency in ultra-thin solar cells using photonic crystals, it is important to investigate the photovoltaic performance, including not only absorption current (short circuit current J_sc) but also open-circuit voltage V_oc and fill factor FF. Our previously reported cell [21] showed lower values of V_oc (0.505V) and FF (0.692) than typical values in μc-Si solar cells (for example, V_oc ~0.52V and FF ~0.72 [11]), as well as a lower J_sc (19.6 mA/cm²) than our theoretical predictions. The crystallinity or the quality of the μc-Si layer generally affects all of these parameters, and thus we examine our previously reported cell in view of the crystalline fraction of the μc-Si layer, which has a great impact on the photovoltaic performance; the optimum crystalline fraction in μc-Si solar cells is reported to be 50-70% [1].

We precede this examination of the crystalline fraction with an explanation of the fabrication procedure used in previous work. We first created a photonic-crystal structure with a lattice constant of 600 nm on a SiO₂ layer using electron-beam (EB) lithography and reactive ion etching (RIE) techniques. Silver (Ag) and Ga-doped ZnO (ZnO:Ga) were deposited on the photonic-crystal structure both as a back reflector and a back electrode. The deposition of n-i-p μc-Si photovoltaic layers was then performed using a plasma-enhanced chemical-vapor deposition (PE-CVD) method operating under very high frequency (60 MHz). Finally, top Sn-doped In₂O₃ (ITO) and Ag contacts were deposited. It is important that the photonic crystal is formed from tapered rods in order to suppress the formation of defects (voids or cracks) in the μc-Si layer deposited on the photonic-crystal structure [24]. The intrinsic μc-Si layer was deposited using a constant SiH₄ concentration ([SiH₄]/([SiH₄] + [H₂]) = 5.25%) and a total flow rate of ~194 sccm. The deposition rate of the intrinsic μc-Si layer was ~0.14 nm/sec. We did not apply in situ controls of the deposition conditions during the deposition of intrinsic μc-Si layer, which is often required for high-rate depostions [30,31].

We then evaluate the crystalline fraction of our previously fabricated solar cell using Raman spectroscopy [32], where an argon (Ar⁺) ion laser (488 nm) was used for excitation. The measured crystalline fraction of the completed cell was ~46%, which is in the range suitable for solar-cell applications. However, this value only represents the crystalline fraction near the upper surface of the cell, because the penetration depth of the Ar⁺ ion laser in μc-Si is limited to ~100 nm.

To investigate the crystalline fraction over the entire thickness, we deposited μc-Si layers of different thicknesses on the photonic crystals and performed Raman spectroscopy. For all samples, we first deposited a 100 nm layer of Ag, 50 nm of ZnO:Ga, and 30 nm of n-doped μc-Si with a crystalline fraction of ~50% on the photonic crystal, which is almost identical to the structure of the fabricated solar cell in our previous work. The measured crystalline fraction as a function of thickness of the i-layers is plotted in Fig. 2. The crystalline fraction at the initial stage of the growth (lower part of the cell) was only ~30%, whereas the crystalline fraction reached ~46% near the top surface.

 

Fig. 2 Crystalline fraction with respect to the thickness of intrinsic μc-Si layer, where the i-layer was deposited by using a constant SiH₄ concentration.

Download Full Size | PDF

The results shown in Fig. 2 suggest that, in our previous cell, the lower part, which is more amorphous than the upper part, is not suitable for solar cells. Since our μc-Si layer is relatively thin (~500 nm) compared to thicknesses of ~2 μm in conventional cells, the influence of such an amorphous-like region on the photo-electric characteristics might be significant. Hence, we next investigate the adjustment of the crystalline fraction over the entire thickness to improve cell performance.

4. Improvement of crystallinity of μc-Si layer on photonic crystal

Following the analysis above, we investigated to obtain a uniform crystalline fraction over the entire thickness of the μc-Si layer on photonic crystals by adjusting the SiH₄ concentration during the growth. We were able to increase the crystalline fraction of the lower part by decreasing the SiH₄ concentration during the initial stages of growth, as shown in Fig. 3(a). The total flow rate of ([SiH₄] + [H₂]) was kept constant at ~194 sccm. Using these conditions, we prepared samples with different i-layer thicknesses. The measured crystalline fraction as a function of thickness is shown in Fig. 3(b). A crystalline fraction of ~50% could be achieved over the entire thickness by tuning the [SiH₄]/([SiH₄] + [H₂]) flow ratio, particularly during the initial stages of growth.

 

Fig. 3 Tuning of crystalline fraction by modifying the SiH₄ concentration during the growth. (a), (b) Change of the SiH₄ concentration and the crystalline fraction with respect to the thickness.

Download Full Size | PDF

We also obtained transmission electron microscope (TEM) images of μc-Si layers deposited on the photonic crystals using constant and adjusted SiH₄ concentrations. Figures 4(a) and 4(b) shows dark field TEM images of the samples grown using constant and adjusted SiH₄ concentrations, respectively. The corresponding diffraction patterns of the samples taken using an aperture diameter of 200 nm on the area above the photonic-crystal rods are shown in Figs. 4(c) and 4(d); [111] and [220] diffraction rings are apparent. The white area in the dark field images corresponds to the crystalline grains grown in the [111] and [220] directions, since the electron beam diffracted by such crystalline components is collected in the detector in the dark-field TEM measurement. The black area also corresponds to the crystalline component where diffracted electron beam is not completely collected due to the rotation of the crystalline grain. The residual gray area corresponds to an amorphous matrix (see Ref [33].). We estimated the crystalline fraction [crystal / (crystal + amorphous) ratio] by analyzing these dark-field TEM images in Fig. 4; we extracted white and black regions using image processing and estimated their areas. For the sample in which we adjusted the crystalline fraction by Raman spectroscopy [Fig. 4(b)], when we extracted 150-nm-thick top area of the μc-Si layer and estimated the crystalline fraction, we obtained ~45% of the crystalline fraction from the portion of the crystallized area to the total area. When analyzing the crystalline fraction near the bottom part in the same manner, we found it to be ~42%. In contrast, the analysis of Fig. 4(a), where we did not adjusted the crystalline fraction in the entire film, revealed that the crystalline fractions at the top and bottom regions are ~44% and ~33%, respectively. Those results are consistent with the crystalline fraction obtained from Raman spectroscopy shown in Fig. 2 and Fig. 3(b). Thus, the uniformity of the crystallinity by adjusting the flow ratio of SiH₄ and H₂ was also confirmed from the TEM image. We also found from Figs. 4(c) and 4(d) that the (220/111) diffraction intensity ratio of the sample with a uniform crystalline fraction over the entire thickness was 1.3 times larger than that of the sample prepared under a constant SiH₄ concentration. It is known that [220] preferential growth of μc-Si is better for solar cells [34], thus our results imply that the growth conditions used to obtain a uniform crystalline fraction might be better for solar cell applications.

 

Fig. 4 TEM observation of the μc-Si layers deposited under constant and adjusted SiH₄ concentrations. (a), (b) Dark field images of the samples with constant and adjusted SiH₄ concentrations, respectively. (c), (d) Diffraction patterns of the samples shown in (a) and (b), respectively.

Download Full Size | PDF

Based on the above investigations, we fabricated solar cells and measured their performance. The basic fabrication procedure was identical to that previously used. We varied the [SiH₄]/([SiH₄] + [H₂]) ratio as shown in Fig. 3, but the ratio was changed smoothly rather than in step-like fashion during the initial stages of growth. We also tuned the conditions used to deposit the p-doped μc-Si layer and the annealing conditions after cell fabrication. Figure 5(a) shows a scanning electron microscope (SEM) image of a fabricated sample with a lattice constant of 700 nm. The measured J-V curves under AM1.5G irradiation at 25°C and the external quantum efficiency (EQE) spectra are shown in Figs. 5(b) and 5(c), respectively. By way of reference, we also showed results of previous photonic crystal solar cells [24]. As can be seen in Fig. 5(b), we improved V_oc, FF, and J_sc compared with our previous reports due to the improvement of the crystallinity of the μc-Si i-layer; we obtained J_sc = 20.4 mA/cm², V_oc = 0.541 V, and FF = 0.732. As a result, the efficiency was improved to 8.06%. Here, in the EQE spectra shown in Fig. 5(c), we find that the response in the long wavelength (>~600 nm) is changed compared to the previous cell, and the response at the short wavelength (<~500 nm) is improved. At long wavelengths, we assume that the slight change of the bottom and top shapes of the structures affected the absorption characteristics; particularly, we find in the SEM image of Fig. 5(a) that there remains a flat region between the hemispherical patterns of the photonic crystal at the top face of the cell, which was not seen in the previous cell [24]. This change was due to the reduction of the rod size at the bottom of the cell. As a result of the formation of remaining flat area on the top (and also the bottom) of the cell, Fabry-Perot interference affected the absorption characteristics, which is clearly seen in the cell without a photonic crystal. We have calculated the change of absorption spectrum and confirmed that the change of the EQE spectrum is reasonable when considering the change of surface shape. A detailed discussion on the influence of the surface and bottom shapes will be reported elsewhere. As for the improvement of the EQE at short wavelengths, we assume that the tuning of the deposition condition of the p-doped layer (particularly, the thickness was thinned from 12.5 nm to 10 nm) suppressed this layer’s parasitic absorption loss.

 

Fig. 5 Fabrication of solar cell under the growth condition using adjusted SiH₄ concentration. (a) SEM image, (b) J-V characteristics under AM1.5G irradiation, (c) EQE spectrum.

Download Full Size | PDF

5. Improvement of photonic-crystal surface shape for high quality μc-Si growth

For further improvement of the conversion efficiency, we also investigated the influence of the surface shape of the photonic-crystal structure prepared in the first step of fabrication on the cell performance. Careful observation of the TEM image in Fig. 4(b) shows that the μc-Si grains tend to grow perpendicular to both the surfaces of the flat bottom area and of the tapered rods. Consequently, grains growing from the flat area collide with grains growing at a different angle from the tapered-rod surfaces. This growth mode adds to the concentration of defects even during the initial stages of μc-Si deposition, and might degrade the cell performance.

To suppress this collision of μc-Si grains, we investigated the employment of a photonic-crystal surface structure without sharp changes in topology between the flat bottom region and the tapered rods. We created photonic crystals with rounded surface geometry by utilizing an isotropic RIE mechanism. This time, we utilized Si as a substrate and used SF₆ gas for etching. Chemical etching can be preferable with SF₆ plasma, and thus the isotropic etching becomes more predominant than the anisotropic (or vertical) etching. We tuned gas pressure and RF power to fabricate rounded surfaces. Figure 6(a) shows an example of a photonic-crystal structure with a well-defined, rounded surface, where the surface was coated by Ag and ZnO:Ga. We next deposited μc-Si layers using the conditions for a uniform crystalline fraction of ~50%. SEM and TEM images of this structure are shown in Figs. 6(b) and 6(c), respectively. For comparing the influence of the bottom shape of the photonic-crystal structure, in Figs. 7(a) and 7(b), we show guidelines (dotted arrows) indicating the growth direction of μc-Si grains shown in Fig. 4(b) for the rod-type photonic crystal, and in Fig. 6(c) for the rounded-type photonic crystal, respectively. It is apparent that the μc-Si grains grew perpendicular to the rounded surface, and the collision of grains could be suppressed compared to the case of tapered rods.

 

Fig. 6 Tuning of the photonic-crystal surface topology for high-quality μc-Si growth. (a) SEM image of a sample with rounded surface after the deposition of Ag and ZnO:Ga. (b), (c) SEM and TEM images after the deposition of μc-Si and ITO layers, respectively.

Download Full Size | PDF

 

Fig. 7 Comparison of the growth results of microcrystalline silicon on (a) rod-type and (b) rounded photonic crystals.

Download Full Size | PDF

Figures 8(a)-8(c) show a SEM image, the J-V characteristics, and the EQE spectrum of a cell containing a photonic crystal with a rounded surface and for which the device isolation method was tuned. As for the device isolation here, we first deposited an ITO layer aligned to the photonic crystal area, then we etched μc-Si existing around the photonic-crystal area covered with ITO by RIE etching using ITO as an etching mask [4]. The RIE etching conditions were adjusted to curtail damage to both the ITO and μc-Si silicon layers. The lattice constant of the photonic crystal was 600 nm. By way of reference, results are also shown for a cell without a photonic crystal but prepared using the same fabrication conditions. The cell performance was further improved compared to the case of Fig. 5: V_oc = 0.555 V, FF = 0.758, and J_sc = 20.6 mA/cm². We expect that the tuning of the bottom shape of the photonic crystals suppressed the formation of defects in the μc-Si layer and thereby improved cell performance, particularly V_oc and FF. Our J_sc of 20.6 mA/cm² is higher than values of less than ~20 mA/cm² reported for other thin μc-Si single junction cells with 500~600-nm thickness [10,12]. This suggests that the employment of photonic crystals is highly beneficial for increasing light absorption in ultra-thin μc-Si layers. As a result of the improvements of each parameter, a high efficiency of 8.70% was achieved for this ~500 nm-thick, ultra-thin μc-Si solar cell, which represents 1.42-fold enhancement compared to the cell without a photonic crystal. Here, in the EQE spectrum shown in Fig. 8(c), we found that the influence of the Fabry-Perot interference seen in Fig. 5(c) was suppressed, leading to a flattened spectrum. This is due to the avoidance of the remaining flat region on top of the cell, as can be seen in Fig. 8(a), by adequately tuning the shape of the rounded bottom-face shape.

 

Fig. 8 Fabrication of solar cell on the rounded photonic-crystal structure. (a) SEM image, (b) J-V characteristics, (c) EQE spectrum.

Download Full Size | PDF

6. Discussion on further improvement

Finally, we discuss the possibility of further improving the efficiency of photonic-crystal solar cells with an ultra-thin photovoltaic layer. With respect to V_oc and FF, the values that we obtained above are now favorable considering that we used homo-junction-type μc-Si layers (that is, all the p/i/n layers are μc-Si) as a result of the improvement of the quality of the intrinsic μc-Si layer, although it may be possible to increase them by employing hetero-junction structures using wide-gap materials. Hence we here discuss the possibility of increasing J_sc to improve the efficiency by analyzing absorption characteristics taking into account the fabricated structure [Fig. 8(a)].

Figure 9 shows the calculation results of the absorption in the fabricated structure, in which the absorption in each layer (n/i/p μc-Si, Ag, ITO, and ZnO:Ga) is color-coded. We used the rigorous coupled wave analysis (RCWA) method and incorporated the influences of all the stacked layers of the fabricated cell for actual analysis, including the wavelength dependence of the absorption characteristics of μc-Si [35], ITO, ZnO:Ga [36], and Ag. From Fig. 9, when we consider that only the absorption of the i-layer is responsible for J_sc, the expected J_sc becomes 20.9 mA/cm². By comparing the experimental EQE and simulated absorption spectrum [Fig. 8(c) and Fig. 9, respectively], we find that the experimental EQE corresponds very closely with the calculated absorption spectrum of the i-layer. Hence, to increase J_sc more, it is necessary to suppress the absorption in the layers other than intrinsic μc-Si layers. From Fig. 9, we find that in the wavelength range longer than ~500 nm, the absorptions in n-doped layer, Ag, and ZnO:Ga are not negligible even though they are placed at the bottom of the cell. An influence of the back electrodes is thought to be considerable because our cells have only ~500nm thickness, which is ~1/4 that of conventional μc-Si cells. Such unwanted absorptions are also seen in thin amorphous-silicon solar cells with the thickness of around 200-300 nm [37,38]. In our detailed calculations, we found that it is possible to reduce Ag loss by carefully designing the layered structure; a detailed discussion will be reported elsewhere. In addition, absorption in the doped layers can be reduced by using wide-gap materials such as nc-SiO_x. We expect from our calculations that J_sc could exceed at least 22 mA/cm² by only tuning the thicknesses of stacked layers, even for cells with intrinsic μc-Si layers of only 500 nm thickness.

 

Fig. 9 Analysis of absorption characteristics considering the whole fabricated solar-cell structure. Portion of absorption in each layer is color coded. White area represents loss due to reflection.

Download Full Size | PDF

7. Conclusion

We have investigated the improvement of the conversion efficiency of photonic-crystal μc-Si solar cells with an ultra-thin (~500-nm thick) intrinsic layer, where light absorption is strongly enhanced by the multiple resonant modes in the photonic crystal. We have studied the quality of the μc-Si intrinsic layer deposited on the photonic-crystal structure from the viewpoint of the crystalline fraction. We have shown that a uniform crystalline fraction can be obtained across the entire thickness by modifying the SiH₄ concentration during growth, and have demonstrated an improvement of efficiency when an intrinsic μc-Si layer with a constant crystalline fraction of ~50% is used. We have also addressed the influence of the photonic-crystal surface topology, and suggested that the use of a rounded surface is beneficial for suppressing the collision of μc-Si grains growing in different directions. These factors have allowed us to achieve an experimental efficiency of 8.7% in an ultra-thin ~500-nm thick μc-Si solar cell, whose thickness is only 1/4 that of typical μc-Si cells. We have analyzed our fabricated cells with calculations that accounted for parasitic losses and discussed further possible improvements of J_sc based on the suppression of the parasitic losses of the doped layers and conductive layers.