Simulation

• Drift-diffusion device simulation in SCAPS-1D and comparable tools
• Transfer-matrix optical modelling of layer stacks
• Where density functional theory helps in absorber screening, and its cost
• Architectures compared: silicon, thin film, perovskite and tandem stacks

Data

• Materials Project, NOMAD and experimental databases: coverage and bias
• Composition and structure descriptors, including Magpie and SOAP representations
• Dataset curation, duplicate handling and the leakage introduced by random splits

Prediction

• Regression models for bandgap, efficiency and stability endpoints
• Graph neural networks on crystal structures, including CGCNN-style models
• Uncertainty quantification and why a point prediction is not enough for screening
• Active learning to place the next experiment where it is most informative

Inverse Design

• Bayesian optimisation over composition and process space
• Evolutionary and generative approaches to candidate proposal
• Multi-objective optimisation across efficiency, stability and cost
• Constraining the search to what can actually be synthesised

Closing the Loop

• High-throughput screening and self-driving laboratory workflows
• Degradation modelling and accelerated ageing under ISOS protocols
• Reconciling predicted with measured performance and diagnosing the gap
• Techno-economic assessment: efficiency gains that do not survive contact with cost