The standard EL setup consists in general of a power supply, a contacting unit and a Si CCD camera which is covered by a light-tight black box to ensure dark light condition. In our setup, the EL setup was equipped with an InGaAs CCD as shown schematically in Figure 1(a) (ELsub). The InGaAs camera is a XEVA-FPA-1.7-640 from the company XenICs which is thermoelectric cooled to 263 K using the Peltier effect. The camera has a spectral band from 900 nm to 1700 nm with a lateral resolution of 640 × 512 pixel and a pixel pitch of 20 µm. In order to filter out the band-to-band luminescence at around 1.1 eV a longpass filter with a cut-on wavelength of 1450 nm was added between the lenses and the InGaAs CCD (green ring in Figure 1(a)). The solar-cell was electrically contacted on the rear side by an aluminum chuck and on the front side grid. A four point measurement was used to ensure the applied voltage and

measure correctly the current. For our measurements we used a current of I = 5 A and voltage of U ~ 0.6 V. The integration times were around 10 s to 30 s depending on the signal strength. Regions with a high defect luminescence appear bright (see Figure 2(b)). The lateral resolution of the measurements is basically limited by the lower lateral resolution of the InGaAs CCD compared to common Si CCDs and is approximately at around 500 µm.

For investigations on a micrometer scale the InGaAs CCD camera was mounted onto an optical microscope; a Zeiss Axio Scope.A1 (µ-ELsub). The setup shown in Figure 1(b) is quite similar to the setup for measuring the prebreakdown positions of a solar-cell under reverse bias (micro reverse-biased electroluminescence—µ-ReBEL) as shown in [14]. However, a longpass filter was added to the optical path of the microscope to filter out the band-to-band luminescence of the forward biased solar-cell. The solar-cell was electrically contacted on the scanning stage by and aluminum chuck (back contact) and needles or solder wires (front contact). The forward bias conditions of the solar-cell, hence the current of I = 5 A and voltage of U ~ 0.6 V, were taken according to EL measurements in Section 3.1. In µ-ELsub usually integration times over 60 s were used. All measurements were performed in reflected light bright field condition and under dark room conditions. However, the different spectrum of about 0.8 eV compared to standard EL at about 1.1 eV leads to a slightly shifted focal plane which was optimized under light. The lateral resolution d is actually limited by the lateral resolution of the microscope which is theoretically given with the wavelength λ = 1550 nm and an aperture angle of α to d = 0.61 × λ/sin α ~ 1 µm. However, it will be shown later that the resolution is actually limited by the defect luminescence itself.

In Figure 2(a), an EL image of a mc-Si solar cell at a current I = 5 A, voltage U ~ 0.6 V and 1 s Integration time is shown. Recombination active defects can be localized by their decreased EL signal, hence by a decreased intensity of the band-to-band luminescence at ~1.1 eV. The same solar-cell was investigated with the same parameters of the power supply (I and U) by the ELsub setup, but with an integration time of about 60 s (see Figure 2(b)). The bright regions are region emitting a high intensity of the defect luminescence between 0.72 - 0.85 eV. Note, the band-to-band luminescence is filtered out. Here, only so called type-B defects (marked with an orange rectangle) appear with a strongly increased intensity of defect luminescence.

Three prominent regions were selected for spectrally resolved EL measurements at room temperature shown in Figure 2(c). The inset in the upper left shows the whole measured spectrum with a high EL intensity at 1.1 eV caused by the band-to-band transition [5]. The dark blue marked region between 0.7 and 1 eV is shown in Figure 2(c). As mentioned type-B defects (orange line in Figure 2(c)) exhibit a high intensity between 0.75 and 0.9 eV compared to the rest of the solar-cell in good agreement to the ELsub imaging in Figure 1(b). Regions with lower or no defect luminescence in Figure 2(b) are showing lower or no defect luminescence in Figure 2(c) respectively. In general, the increased ELsub intensity shows in all investigated positions a broad peak with a maximum at around 0.8 eV.

In order to achieve a clear one to one correlation of single defects and positions with increased defect luminescence, we have applied an InGaAs CCD detector to an optical microscope (see Section 3.2). In Figure 3, two different regions taken out of regions having an increased intensity of defect luminescence are shown. In the EBIC image in Figures 3(a) and (b), line defects with a decreased EBIC signal indicating an increased recombination activity can be seen. These line defects are small angle grain boundaries (not shown here). Exactly at these grain boundaries an increased intensity of the defect luminescence can be seen by µ-ELsub as bright contrast in Figures 3(c) and (d), hence the grain boundaries can be clearly classified as type-B defects [10]. Few prominent

grain boundaries are marked by orange lines. The defect luminescence seems to occur homogeneous distributed along the grain boundary rather than spot-like localized, which is in good agreement to results of Schmid et al. [15] who have scanned laterally the so called D1 luminescence at 0.8 eV. Nevertheless, the defect luminescence occur a bit blurred around the defect structures limiting the possible lateral resolution strongly. Even under high magnification no localized defect luminescence can be observed, but further measurements will give more knowledge, i.e. varying voltage and temperature, increasing magnification and integration time.