Optimizing Multicrystalline Silicon for Low-Cost Photovoltaics

The most commonly employed material for solar cell fabrication is multicrystalline silicon (mc-Si) with a global market share of about 45%, corresponding to worldwide installations totaling about 45GW (as of end of 2012). The rapid growth within the photovoltaic (PV) market over the last decade has generated strong competition among various PV material candidates.  The most important factors are material abundance, raw material cost, wafer and solar cell manufacturing costs, and cell efficiency. While solar cells produced on single-crystal silicon substrates are about 10-15% more efficient that mc-Si cells, the expense associated with single-crystal growth is substantially higher.  As a result, there is currently much effort aimed at understanding how defects and impurities in mc-Si reduce cell efficiency with the aim of reducing the deficit to single-crystal material.

 

The aim of our work is to use atomistic simulations to study the thermodynamics, transport properties, and interactions of point and extended defects that are commonly encountered in mc-Si. The simplest of these are the native point defects, namely self-interstitials and vacancies, which can cluster into precipitates. Examples of extended defects in mc-Si include dislocations and grain boundaries that form during the crystallization process, and which may interact with point defect clusters, impurity precipitates such as SiC and SiO2 and Si3N4, and transition metal atoms. Most of the impurities are introduced during crystallization from the crucible, its coating and the furnace heater.

 

Large-scale molecular dynamics simulations are employed to study the crystallization process, as well as defect transport and evolution within the solid.  Shown in the figure below, from left to right are example simulation snapshots that show (1) a silicon crystal that has been rapidly solidified and in which grain boundaries are visible (highlighted with arrows and denoted by cyan atoms), (2) a crystalline silicon simulation cell containing two grain boundaries (cyan atoms) and a mobile point defect, and (3) a close-up view of the so-called Σ3(111) twin which is one of the most common (and lowest energy) grain boundaries in mc-Si crystals. In addition to direct molecular dynamics, we are also developing additional computational tools based on the Monte Carlo method to investigate, for example, the nucleation and growth behavior of various precipitates related to efficiency limitations in mc-Si.

Peter-Optimizing-Si-PV
This project is supported by the SolarWinS Consortium based in Germany (www.solarwins.de).