Impurities in silicon and their impact on solar cell performance

Abstract

Photovoltaic conversion of solar energy is a rapidly growing technology. More than 80% of global solar cell production is currently based on silicon. The aim of this thesis is to understand the complex relation between impurity content of silicon starting material (“feedstock”) and the resulting solar cell device performance, taking into account the effects of processing and device architecture. Among the most important differences from the groundbreaking research performed in the ‘70s by Westinghouse Corporation are that the current data refers to state-of-the-art industrial solar cell processes which are very far from those used more than 3 decades ago , and that the impact of impurities is investigated on directionally-solidified multicrystalline silicon ingots and Float-Zone (FZ) ingots, as opposed to mainly Czochralski (Cz) ingots. In the current work the direct relationship between impurities in the feedstock and the final solar cell performance is reported and analyzed. This provides the most straightforward and meaningful interpretation, especially aimed at specifying feedstock impurity tolerances.

The transition metals iron, chromium, titanium, copper and nickel have been considered in this study. The first four cause a reduction of the minority carrier diffusion length in the silicon wafer bulk. Nickel does not have an impact on the bulk diffusion length, but is found to strongly affect recombination in the diffused emitter. Copper has the peculiarity to impact both.

There is a difference in the crystal structure for ingots contaminated with high levels (40-50 ppm wt) of impurities compared to reference ingots. At the bottom and at the top of the contaminated ingots the density of crystal defects is enhanced showing an influence of metal impurity concentration on formation of defects during crystal nucleation and growth. This is also reflected in the solar cell efficiencies.

A model based on the Scheil distribution of impurities has been derived to fit the observed effects along the ingot. The model fits the experimental data very well and has been successfully validated on ingots with different concentration levels of impurities. It has been used to classify the harmfulness of each impurity and to estimate the impact of a certain amount of impurities in the feedstock on the final solar cell performance.

In the past several years discussion has taken place on the specifications for solar grade (SoG) silicon feedstock. Many studies report that solar cell performances comparable to those obtained using high-purity material can also be obtained for feedstock with higher concentration levels of metal impurities. The physicochemical properties of the specific impurities, in particular diffusivity and segregation coefficient, are decisive for their impact. This leads to the definition of classes of metal impurities: impurities with high diffusivity in silicon can be better tolerated than slowly diffusing ones, since they can be more easily manipulated during cell manufacturing by the gettering processes.

More studies are required to reveal the relative impact of other elements than the ones studied here and especially to understand the interactions between impurities and between impurities and crystallographic defects and their impact on future advanced solar cell architecture.