GaAs. These factors are favourable for growing good quality epitaxial layers of GaAs on Ge. Also Ge is stronger than GaAs and the higher strength can be exploited to form larger area, thinner cells. In the long term, Ge substrates will cost less than GaAs substrates, especially for larger area slices. Previous work at Jet Propulsion Laboratory, Lincoln Laboratories and Spire Corporation had shown that good quality GaAs cells could be epitaxially grown on Ge substrates. The AF contract extended this previous work and led to GaAs/Ge cells, up to 4x4 cm2 in area and 3-4 mils thick, with efficiency at about 17%. The layer growth and processing of those cells were similar to the methods used for GaAs/GaAs cells. Interim Re-evaluation of GaAs/Ge Cells As the GaAs/Ge cell work progressed, the MOCVD growth conditions which appeared to give the best quality GaAs layers on Ge substrates led to cells which had higher Voc than GaAs cells, and usually lower curve fill factors (CFF). The enhanced Voc was generated at the GaAs/Ge interface, and operated in cascade (series) with the GaAs PN junction. At the time, it was believed that the additional voltage (up to 200 mV) would lead to efficiencies greater than those obtained for GaAs/GaAs cells, expecially if the CFF values could be increased. However, further tests of these cells consisting of a GaAs cell grown on active-Ge substrates showed some possible problems. The temperature coefficient of Voc for the cascade cells was higher, up to twice the values typical of GaAs cells, and this offset a major advantage of GaAs cells. During measurements of the temperature coefficients, an older two-light solar simulator was used. The spectral output for this simulator was closely matched to the true AMO spectrum, and the enhanced Voc cells showed ‘kinked' I-V curves (Fig. 1). The kinks were attributed to insufficient current generation and collection at the GaAs/Ge interface, the generation resulting from absorption of the near infrared region of the AMO spectrum which is transmitted through the GaAs layers (i.e. wavelengths above 0.9 /zm). If the test spectrum was richer in nearinfrared wavelengths than the true AMO spectrum, no kinks were observed. In Fig. 1, the Hoffman simulator was close to AMO, the XT-10 simulator was red-rich. When the cascaded cells operated under the true AMO spectrum, near maximum power conditions, because the current generated near the GaAs/Ge interface was lower than the current collected at the GaAs PN junction, the GaAs/Ge interface was driven into reverse bias operation. The kink resulted from superposition of the reverse I-V characteristic of the interface and the forward I-V characteristic of the GaAs PN junction. The tests with different simulators were confirmed by testing at high altitudes by NASA-Lewis aeroplane flights, and it was concluded that the high efficiencies obtained for the active-Ge structures were mainly caused by red-rich test conditions. The red- rich simulators had good spectral match for the wavelengths used to test silicon or GaAs cells, which only extend to 1.2 /zm or 0.9 /zm respectively. On revising theoretical estimates of the maximum current available for collection at the GaAs/Ge interface (after passing through the GaAs layers), it was concluded that even for ideal conditions, current matching for the interface and the PN junction would be difficult to obtain under the true AMO spectrum. Tests were made to shift the anti-reflective coating reflectance minima, to reduce the reflectance of the near infrared wavelengths, but they did not lead to current
RkJQdWJsaXNoZXIy MTU5NjU0Mg==