Home / Research / Pillar 01

Pillar 01

AI-Enhanced Characterisation

Multimodal characterisation experiments are often difficult to interpret, and many of the fundamental properties governing energy materials are hard to measure directly. We develop hybrid approaches that combine advanced optoelectronic characterisation, physics-based modelling and Scientific AI to infer hidden material properties, reveal degradation mechanisms and predict long-term device behaviour from rich multimodal data. By transforming complex observations into quantitative physical understanding, we aim to transform how energy materials and devices are characterised and understood.

PHYSICAL INFERENCE

Seeing beyond what experiments can directly measure.

Understanding the fundamental physics governing next-generation energy materials requires extracting meaningful information from increasingly complex experimental data. At EMHL, we combine advanced optoelectronic characterisation, spectroscopy, imaging and physics-based simulations to investigate light–matter interactions, charge transport, ionic migration, degradation mechanisms and other fundamental processes that determine device performance and stability.

Artificial intelligence serves as a powerful scientific tool that complements these experimental and computational approaches. By integrating multimodal measurements with machine learning, we develop predictive models capable of inferring latent material properties, forecasting long-term device behaviour and uncovering physical information that would otherwise remain experimentally inaccessible. Together, these approaches enable a more quantitative understanding of energy materials while accelerating the development of next-generation photovoltaic technologies.

Core questions

  1. 01

    How do charge transport, recombination and ionic migration determine the stability and performance of photovoltaic devices?

  2. 02

    Can hidden material properties be inferred directly from multimodal optoelectronic measurements?

  3. 03

    Which early experimental signatures predict long-term degradation?

  4. 04

    How can AI and physics-based models improve the interpretation of complex experimental data?

Hybrid Scientific Workflow

⌁

Advanced Experimental Characterisation

  • Steady-state and time-resolved optoelectronic characterisation
  • Photoluminescence (PL) and electroluminescence (EL) spectroscopy
  • Hyperspectral and multimodal imaging
  • Device electrical characterisation
  • Scanning electron microscopy (SEM), cathodoluminescence and complementary imaging techniques
◇

Physics-Based Modelling

  • Drift–diffusion simulations of photovoltaic devices
  • Charge transport and recombination modelling
  • DFT simulations (in collaboration with computational groups)
  • Generation of high-fidelity synthetic datasets for AI
✣

Scientific AI

  • Interpretable machine learning
  • Deep learning for image and spectroscopy analysis
  • Prediction of latent material and device properties
  • Early prediction of long-term degradation
  • Multimodal data fusion and scientific inference

Selected projects

  1. 01

    Predicting Long-Term Stability of Perovskite Solar Cells from Early Fingerprints

  2. 02

    AI-Enhanced Multimodal Inference of Material and Device Properties from Imaging Techniques

  3. 03

    Inferring Latent Material Properties from Electrical Measurements in Perovskite Solar Cells

Collaborations

Interested in collaborating on AI-enhanced characterisation?

Get in touch →