The effect of trapping defects on CIGS solar-cell performance

Abstract

The relationship between basic solar-cell parameters and carrier-trapping states is explored through current-voltage, quantum efficiency, and temperature-dependent capacitance measurements, including admittance spectroscopy (AS) and drive-level capacitance profiling (DLCP). The study focuses on several categories of CuIn1-xGax(Se1-ySy)2 (CIGS) solar-cell devices. Some devices were produced by evaporation, and some by selenization of metal films. Within each of these two groups of devices, some devices have a standard CdS buffer layer separating the CIGS absorber from the ZnO window layer, while others have had the CIGS absorber treated in a Cd-containing bath, but no CdS layer deposited between the absorber and the ZnO window layer. The latter device preparation technique (Cd PE treatment) tended to result in devices with a weaker junction. Although there are numerous ways in which the devices studied can be categorized, three sets of general experimental trends were noted. One set of comparisons was made between all evaporated and all selenized devices. Another set of trends was observed when comparing CdS devices to their Cd PE counterparts. The final set of trends was observed when comparing all selenized devices to each other. Within these studies, all evaporated CuIn1-xGax(Se1-ySy)2 devices had y = 0, and all selenized devices had y ≈ 0.3. The typical maximum value of x in all devices studied is 0.3, although x varies differently with depth in the evaporated than in the selenized devices. Comparisons between evaporated and selenized devices yielded several strong trends. The evaporated devices tend to have higher open-circuit voltages (VOC) when considering comparable junctions, lower defect activation energies (Ea), and fewer defect states detected near the interface than the selenized devices. Slightly weaker trends observed include a tendency of the evaporated devices to have a lower diode quality factor than selenized devices made with a comparable junction. Admittance spectroscopy results imply that the overall defect density (Nt) of the evaporated devices is lower. Within the group of selenized devices, the observed trends were somewhat surprising. A higher trap density deduced from DLCP appears to modestly correlate with a higher VOC. This correlation suggests that the majority of the detected traps do not participate in recombination that limits the performance of the devices. When the total response of the DLCP measurements is evaluated, i.e. the approximate trap density plus the free carriers, the devices with larger DLCP responses were the devices with larger VOC. Thus, the detected traps may be affecting a device much like non-frequency dependent shallow acceptor levels. Barring other differences, the larger the density of shallow acceptor states in a device, the larger the open-circuit voltage of the device. This must now be put in perspective with the aforementioned observations comparing evaporated and selenized devices. Based on VOC, it does not appear that the best selenized devices have the lowest trap density. Thus, although the selenized devices do have a higher trap density than the evaporated devices, the overall number density of shallow traps may not be the reason the selenized devices do not perform as well as the evaporated devices, rather merely the presence of a measurable density of defect states near the interface may be the reason for the performance difference. Although the best devices (evaporated) seem to have an overall trap density that is lower than the other devices, it does not appear likely that the density of traps within the detection range is what is limiting the performance of the selenized devices. However, the location of the traps (near the interface vs. in the bulk) does appear to be related to device performance. Thus, based on this work, the CIGS research community should use the DLCP technique supplemented with AS for evaluating more precise free carrier densities for the best devices and DLCP data analyzed using multiple analysis techniques for evaluating trap states near the interface, but should perhaps focus on other techniques that are more likely to probe deeper into the band gap to search for recombination centers in the bulk in all well-behaved devices.