Dislocation Density Reduction in Multicrystalline Silicon through Cyclic Annealing

Multicrystalline silicon solar cells are an important renewable energy technology that have the potential to provide the world with much of its energy. While they are relatively inexpensive, their efficiency is limited by material defects, and in particular by dislocations. Reducing dislocation densities in multicrystalline silicon solar cells could greatly increase their efficiency while only marginally increasing their manufacturing cost, making solar energy much more affordable. Previous studies have shown that applying stress during high temperature annealing can reduce dislocation densities in multicrystalline silicon. One way to apply stress to blocks of silicon is through cyclic annealing. In this work, small blocks of multicrystalline silicon were subjected to thermal cycling at high temperatures. The stress levels induced by the thermal cycling were modeled using finite element analysis (FEA) on Abaqus CAE and compared to the dislocation density reductions observed in the lab. As too low of stress will have no effect on dislocation density reduction and too high of stress will cause dislocations to multiply, it is important to find the proper intermediate stress level for dislocation density reduction. By comparing the dislocation density reductions observed in the lab to the stress levels predicted by the FEA modeling, this intermediate stress level is determined. Thesis Supervisor: Tonio Buonassisi Title: Assistant Professor of Mechanical Engineering Acknowledgements I would like to thank my adviser, Prof. Tonio Buonassisi, for his insights, guidance, inspiration, and kindness throughout my graduate studies. I have learned a great deal from him about photovoltaics, materials, and how to do good science. I would also like to thank Sergio Castellanos, Doug Powell, Hyunjoo Choi, Mariana Bertoni, and Allie Fecych, my past and present colleagues on the dislocation density reduction project. They have each been invaluable to my studies and research. David Fenning was also a great resource for enlightening discussions. Credit also goes to Doug Powell for creating the original Abaqus model, which I modified to get the results in this thesis. I would also like to thank my colleagues at MIT, both inside and outside of the Buonassisi group, for the great conversations, help in the lab, and fond memories: Sarah Bernardis, Rupak Chakraborty, Katy Hartman, Eric Johlin, Yun Seog Lee, Sin Cheng Siah, Joe Sullivan, David Needleman, Yaron Segal, Christie Simmons, Mark Winkler, Andre Augusto, Bonna Newman, Steve Hudelson, Sebastian Castro, Vidya Ganapati, Sabine Langkau, Keith Richtman, Jim Serdy, Alison Greenlee, and Erik Verlage. Not only are they great scientists, but they are wonderful people. Finally, I would like to thank my family and Andrew for their love and support. Cliches aside, I couldn't have done it without you.