There is increasing demand for renewable energy, as highlighted by the UK government's aim of reducing carbon emissions by 80% before 2050. Solar power is the most promising renewable technology due to the enormous amount of energy the sun can provide. Most commercially available solar panels - based on crystalline silicon - are relatively efficient but expensive to manufacture. Accordingly, there is significant interest in alternative photovoltaic absorbers that are just as efficient but have lower materials and processing costs.

One route to finding novel solar absorbers is using quantum-mechanical computations. Indeed, many of the properties that determine photovoltaic performance - such as the strength of visible light absorption - can be calculated relatively easily. Many studies have taken advantage of this by searching for new solar absorbers based solely on electronic and optical properties. Unfortunately, this approach generally gives rise to many false positives where materials are predicted as efficient but perform poorly in practice. These shortcomings often result when the behaviour of crystal imperfections is not considered. These imperfections, called point-defects, play a crucial role in photovoltaic devices by limiting the maximum obtainable voltage and current. However, predicting the effects of defects on photovoltaic performance has so far proved tricky and has only been achieved for a select few systems.

By gaining an understanding of the fundamental factors that control defect formation we can design new materials that are resistant to their effects. Materials in which defects do not significantly affect photovoltaic performance are called "defect tolerant". Due to the difficulty of calculating the impact of defects, the structural and chemical properties that give rise to defect tolerance are not well understood. However, recent advances in computational workflow software means it is now possible to automate the calculation of complex properties. This project will develop an automatic computational workflow to determine whether a material is defect tolerant. By applying the workflow to many hundreds of materials and analysing the trends, we can extract the structure-property relationships that give rise to defect tolerance. We can also use this information to develop machine learning models for predicting the impact of defects without needing to perform any calculations. As many other applications also rely on the formation of point-defects - such as thermoelectrics and quantum computers - our calculated data will be of broad interest to the scientific community. We will therefore make the results available as an online database of computed defect properties.

An advanced understanding of the factors that govern defect tolerance will enable the rational design of the next generation of photovoltaic materials. Photovoltaics with reduced cost will facilitate the adoption of solar power and pave the way for a revolution in clean energy.

Potential Impact:
This project aims to uncover the design rules that enable highly efficient photovoltaic devices. In particular, the behaviour of defects at typical device operating conductions - i.e., under illumination - is severally understudied. An understanding of these fundamental factors may enable the next generation of breakthrough materials that are both cheaper and more efficient. This work will benefit photovoltaics researchers by directing the focus of future research towards more optimal materials for solar applications. Advances in this area may also benefit researchers of more established solar technologies, such as crystalline silicon devices. These benefits will be felt both in the UK (through the SUPERGEN Supersolar and Superfuel networks) and abroad, due to the international nature and global impact of photovoltaics research. In addition, defect properties are fundamentally important for many other emerging applications, including thermoelectrics, batteries, and quantum computing. A deeper understanding of the factors that control defect formation will therefore enormously benefit researchers across numerous disciplines and may lead to advances outside the field of photovoltaics.

According to a recent study by the Energy Watch Group, solar power could provide 70% of the world's total energy by 2050. In the UK alone, the installed capacity of photovoltaics is expected to almost triple in the next ten years, as predicted by the 2019 Solar Commission. Accordingly, there are increasing scientific and industrial efforts to achieve the cost-competitiveness needed for utility scale solar power generation. Greater use of renewable technologies will minimise the environmental impact of energy use and alleviate the UK's reliance on energy imports, thereby benefiting many consumers across the UK.

Solar power is an enormous and rapidly expanding industry - worth £40 billion in 2018 and expected to rise to over £200 billion by 2026. Silicon based solar panels currently dominate the market despite their relatively high processing costs. The UK is home to several world-leading companies attempting to find alternative photovoltaic materials with increased cost-competitiveness, including Oxford PV the largest global developer of commercial perovskite devices. As part of this project, I will release open-source machine learning models to predict defect tolerance that can be employed by industry for rapid prototyping of novel materials. Furthermore, my results may enable the optimisation of the synthesis conditions of existing materials. Contributing to a vibrant research and development process will allow the UK to maintain a competitive advantage over its international competitors.