Mapping solar cell parameters using hyperspectral imaging

Research into photovoltaic energy technologies seeks to reduce costs and increase efficiencies. To meet these aims, solar cells made from copper indium gallium diselenide: Cu(In,Ga)Se₂ (CIGS) are good candidates. Industry and laboratory tests show these cells to be highly efficient,¹ and they can be fabricated at low cost. However, there remain questions about their effectiveness in manipulating light and charge carriers because of their structure. Photovoltaic parameters in CIGS fluctuate on scales of below 1 micrometer to 1cm, which reduces efficiency. Mapping methods would help to understand and improve cell mechanisms.

To measure CIGS cell characteristics with spatial resolution, we can use photoluminescence (PL) or electroluminescence (EL). In both cases, we apply an external excitation source (light for PL, electrical power for EL) to the cell sample. The cell emits luminescence in response, which we can record with an imaging system such as a camera or a confocal microscope. We use the generalized Planck's law,² which describes the radiation emitted by a hypothetical body at thermal equilibrium to describe the device emissions and to determine optoelectronic properties, including local quasi-Fermi level (qFl) splitting. The qFls describes the concentrations of holes and electrons. The difference between these quantities is assumed to be the upper limit of the cell's voltage.

To perform such experiments, we can use standard imaging, although this does not provide spectral information for the emission, which is necessary to derive qFl splitting. Alternatively, we can use confocal microscopy, but it is a time-consuming scanning technique that needs high excitation power and does not provide absolute values of qFl splitting.

Instead, we developed a hyperspectral imager which records spectrally resolved images. Its ability to directly acquire maps reduces by two or three orders of magnitude the acquisition time taken using a confocal microscope. We also demonstrated the set-up's absolute calibration,³ measuring light in photons per energy interval, units of time, and surface area instead of relative values. We therefore calculated the qFl splitting in electron volts.

We demonstrated our set-up on high-efficiency gallium arsenide (GaAs) solar cells by mapping saturation currents without electrical contact.⁴ To do this, we derived a short circuit current/open circuit voltage characteristic curve, I_CC(V_OC), which was fitted in a two-diode model. The results were in good agreement with standard electrical measurements. We determined the sheet resistance of the cells' window layer (the top strata, which admits incoming light) from the spatial decay of the EL signal.³ We applied the reciprocal relationship between solar cells and LEDs (based on the p-n junction structure of both devices),⁵ and from the EL images we determined in absolute values the cells' external quantum efficiency (EQE)—the ratio of charge carriers collected by the cell to the incident photons.

We then used the hyperspectral imager to characterize CIGS samples in PL and EL. Figure 1 shows the emission from a 35μm diameter microcell at 1070nm, and Figure 2 shows two spectra extracted from points a and b. We observed qFl lateral variations in the order of 30 milli-electron volts, in agreement with previous results. In a second experiment, we excited locally the center of the sample and imaged the extent of the luminescence revealing the carrier transport. This kind of localized excitation is used in confocal microscopy, and this experiment may help with interpreting those results.

Figure 1. Photoluminescence of a 35μm-diameter copper indium gallium diselenide (CIGS) microcell at 1070nm. Points a and b are the extracted spectra displayed in Figure 2.

Figure 2. CIGS photoluminescence from two distinct locations: (a) and (b). We monitored variations in shape and intensity.

We also recorded and interpreted the EL of CIGS cells with regard to the LED/solar cell reciprocity relationship, and we noted strong variations in spectra and intensity. The EQE shapes deduced from the spectral variations do not exhibit significant fluctuations. However, intensity variations are stronger, and can be explained as non-uniformities of the EQE or voltage. We will conduct further experiments to investigate these factors.

Our next step is to relate the local properties of the cell to the global characteristics, such as its efficiency. Eventually, our method may be used as a diagnostic tool at each step of a solar cell's fabrication.