Multiscale Modeling of Heteroepitaxial Interfaces for Scalable Thin-Film Solar Cell Applications

Interfaces occur in many aspects of science from heterogeneous catalysis to device fabrication. For instance, thin-film photovoltaic devices consist of a multilayer architecture, for which charge-carrier transport across the interfaces plays a crucial role in minimizing the associated recombination losses and achieving high solar conversion efficiencies. To achieve insight-driven engineering and optimization of thin-film solar cells, a high-level characterization that gives a local, electronic, and chemical picture of the interface properties is needed. However, owing to their narrow widths and their often non-crystalline structures, interfaces are difficult to resolve or access by purely experimental means. Atomistic modeling and simulation are therefore ideally suited to complement experiments and supply the missing information on interface structure, properties, and their evolution with time. Nevertheless, interfaces extend in many cases well beyond the size limit of first-principles quantum mechanical methods (i.e., accompanied by huge computational cost), therefore, there is an urgent need for the development of more efficient multiscale methods, that can operate at larger time and size scales. Two main challenges of simulating realistic interface models level remain: (i) the simulation cell must be sufficiently large to accommodate the incommensurate nature of the system and include misfit-induced structural modifications and (ii) the final structure must not be influenced artificially by the starting structure. With the emergence of machine learning interatomic potentials (MLIPs), we are now at the threshold of achieving efficient atomistic simulations with DFT-level accuracy, which can greatly enhance the capability of multiscale modeling of realistic surface and interface systems. This project aims to develop and implement a multi-scale modeling environment (integrating advanced first-principles DFT calculations, ab initio molecular dynamics (MD), and classical MD simulations accelerated by high-accuracy machine learning interatomic potentials) to simulate large-scale realistic interfaces. We will establish a universal set of reliable and transferable MLIPs for large-scale simulated amorphization and recrystallization molecular dynamics that will evolve the complex microstructural features observed experimentally at perovskite/CTL interfaces, including interfacial reconstructions, and defects such as dislocations, grain boundaries, vacancies, and interstitials. We will also establish rational molecular design principles and structure-property relationships needed to provide experimental guidance for selecting passivation ligands to improve stability and performance. We will perform a microkinetic simulation of temperature-programmed desorption (TPD) to predict the desorption temperatures of selected passivation molecules at perovskite surfaces. The detailed atomic level information will be experimentally tested and used to enhance the perovskite solar cell technology. The precise understanding of the microscopic structure of interfaces and the mechanisms of interfacial phenomena obtained from the proposed work, will not only help to identify specific bottlenecks to the performance and stability of photovoltaic devices but enable an insight-driven optimization of interface properties to achieve more efficient and stable solar cells with obvious socio-economic and environmental benefits. The tools and methodologies developed in this work would have much wider applications than the specific case of solar cells. This project will also have a broader impact in training and inspiring a new generation of young scientists to pursue emerging technologies such as quantum computing, artificial intelligence (AI), and machine learning (ML), which are poised to lead STEM-proficient nations to innovation-based competitiveness.