This paper presents a combination of numerical and experimental methods used to characterize defect clusters in multicrystalline silicon solar cells.

Multicrystalline silicon wafers used for solar cells exhibit defect clusters--localized crystal defects in and near grains of some specific orientations. Defect clusters are also dominant sites for impurity precipitation, and they remain ungettered and unpassivated through the solar cell processing. This paper describes characteristics of defect clusters, and shows, through theory and experiment, that defect clusters typically lower cell efficiency by 3 to 4 absolute percentage points. To recover this efficiency loss, it is necessary to getter precipitated impurities.

Our analyses show that defect clusters can lower the efficiency of multicrystalline silicon (mc-Si) solar cells by 2 to 4 absolute percentage points. This large loss can be recovered if impurities precipitated at the defect cluster sites can be gettered. We describe a new technique for gettering precipitated impurities.

About 30% of the total market share of industrial manufacture of silicon solar cells is taken by single crystalline Czochralski (CZ) grown wafers. The efficiency of solar cells fabricated on boron-doped Czochralski silicon degrades due to the formation of metastable defects when excess electrons are created by illumination or minority carrier injection during forward bias. The recombination path can be removed by annealing the cell at about 200° C but recombination returns on exposure to light. Several mono-crystalline and multi-crystalline solar cells have been characterized by methods such as laser beam induced current (LBIC), Four-Probe electrical resistivity etc. to better understand the light induced degradation (LID) effect in silicon solar cells. All the measurements are performed as a function of light soaking time. Annealed states are produced by exposing the cells/wafer to temperature above 200° C for 30 minutes and light soaked state was produced by exposure to 1000 W/m2 light using AM1.5 solar simulator for 72 hours. Dark I-V data are analyzed by a software developed at NREL. This study shows that LID, typically, has two components- a bulk component that arises from boron-oxygen defects and a surface component that appears to be due to the SiNx:H-Si interface. With the analysis of dark saturation current (J02), it is seen that the surface LID increases with an increase in the q/2kT component. Results show that cell performance due to bulk effect is fully recovered upon annealing where as surface LID does not recover fully. This statement is also verified by the study of mc- silicon solar cells. Multi-crystalline silicon solar cell has very low oxygen content and, therefore, recombination sites will not be able to form. This shows that there is no bulk degradation in mc- Si solar cells but they exhibit surface degradation. The results suggest that a typical Cz-silicon solar cell with an initial efficiency of - 18% could suffer a reduction in efficiency to - 17.5% after the formation of a metastable defect, out of which - 0.4% comes from a bulk effect and - 0.1 % is linked to a surface effect.