Computational frameworks to enable accelerated development of defect-tolerant photovoltaic materials

Abstract

Widespread adoption of carbon-free energy technologies, including and especially photovoltaics (PV), is vital to address the issue of climate change. Absent sweeping policy action, this will only happen if these technologies become the most economically viable energy source. While PV has become cheaper in recent years, technoeconomic modeling suggests that current PV technologies cannot come down in price enough to be deployed at the scale required by future energy-mix scenarios that avert catastrophic climate change. This issue can be addressed by developing new technologies - in particular, ones that rely on materials that can be manufactured at drastically cheaper costs than present-day ones. To perform well in a PV device, silicon (the active material in approximately 90% of devices on the market today) must be free of detrimental metallic impurities at levels of parts per billion; the process to achieve this purity requires expensive equipment and large energy expenditures.

In contrast, hybrid halide perovskites (a new class of PV materials developed in the past decade) are extremely defect-tolerant. Synthesized using solution-based methods at ambient temperatures and pressures, they contain orders of magnitude more defects and yet achieve power conversion efficiencies comparable to silicon-based devices. Unfortunately, these materials suffer from lack of long-term stability as well as concerns surrounding toxicity since all high-performing variants to date contain lead. This work centers on accelerating the process of discovering other defect-tolerant materials that would share the remarkable optoelectronic properties of the perovskites without suffering these drawbacks, and focuses on two particular areas. The first is aimed at understanding the atomic-scale physics enabling the defect-tolerant behavior of the perovskites in order to formulate screening/design criteria for new materials.

The primary reason that perovskites perform so well even in the presence of defects is that the energy states due to the most abundant defects are all shallow in nature, i.e., close in energy to the band edges, and hence contribute very little to nonradiative recombination current losses. I identified several novel mechanisms for this behavior that have strong explanatory power for systems that have been examined in detail; this improved understanding also promises to aid in prediction of future compounds. To complement the theoretical/computational identification of new candidate materials, the second thrust of this thesis is accelerating their experimental characterization.

I have developed open-source software that enables the use of high-throughput experimental measurements (e.g., photocurrent as a function of voltage, temperature, and light intensity), in concert with device simulation run on high-performance computers and Bayesian parameter estimation, to construct probability distributions over unknown input parameters of those device simulations. This enables extraction of multiple parameters (in realistic, device-relevant contexts) from a single set of inexpensive, automatable measurements. This approach has the potential to supplant traditional direct characterization methods, which can be time-consuming and subject to confounding factors such as different sample preparation requirements. Taken together, these two primary thrusts can dramatically accelerate PV materials discovery.

First, we reduce the search space of materials by defining better selection criteria, focusing limited experimental bandwidth on only the most promising candidate compounds. Second, once a material has been synthesized, it can be characterized and optimized rapidly through the Bayesian inference technique. While I have focused primarily on PV materials, many aspects of this work could be applicable in a broader array of energy materials studies such as batteries or thermoelectrics. If we can speed up the process of discovering and developing new materials systems, then we can speed up the adoption of the resulting energy technologies, thereby lowering costs, reducing emissions, and improving lives.