Spatially Resolved Determination of Metallization-Induced Recombination Losses Using Photoluminescence Imaging

Abstract

Metallization induced recombination losses are one dominant loss mechanism for current industrial solar cells. A precise determination of these losses is important for contacting technology optimization, as well as precise solar cell modeling. Usually, for state-of-the-art approaches to determine j <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0,met</sub> , it is assumed that the samples itself exhibit spatially uniform properties (e.g., carrier lifetime or sheet resistance) or that the used reference samples are identical to the metallized samples. Finally, in most cases, only one global j <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0,met</sub> -value for the entire sample is given, neglecting possible spatial inhomogeneities. In this article, we mostly eliminate the necessity for the assumptions of perfect sample homogeneity by means of an interpolation scheme of the photoluminescence (PL) signal. Thereby, we can predict the PL signal of a virtually nonmetallized test field with a relative standard deviation of about σ ≈ 0.7%. Additionally, we determine j <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0,met</sub> for specific test fields at different positions on the sample and correlate the results to the local emitter sheet resistance R <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">sh</sub> , the local peak firing temperature of the sample during the fast firing process T <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">peak</sub> , and the test field finger spacing d. For our samples, a reduction of d from d = 1000 μm to d = 200 μm leads to a reduction of j <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0,met</sub> by up to 18%. This strong effect is physically unexpected and so far not considered by the state-of-the-art approach, frequently performed in the photovoltaic community. Further, we perform a sensitivity and error analysis which reveals that we are able to determine j <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0,met</sub> within an estimated accuracy between 15% and 18%.