Machine Learning Applications in Perovskite Materials

Machine learning has emerged as a transformative approach in the design, optimisation and stability analysis of perovskite materials. By leveraging statistical models and high-throughput experimentation, researchers can now navigate vast compositional and process‐parameter spaces with unprecedented speed and precision. Data‐driven screening workflows enable the rapid identification of novel, lead‐free hybrid organic–inorganic perovskites with tailored bandgaps and enhanced thermal stability. Concurrently, autonomous material and device acceleration platforms integrate robotic synthesis, in situ characterisation and intelligent experimental planning to refine thin-film deposition, capping‐layer selection and ageing protocols. Large, curated device databases facilitate big-data analyses that yield unified stability metrics, while predictive models uncover underlying structure–property relationships and guide cation engineering strategies. Collectively, these advances are driving perovskite photovoltaics towards higher efficiency, reproducibility and operational lifetime, laying the groundwork for scalable fabrication and commercial deployment.

Research from Nature Portfolio

Recent studies have applied machine learning to assemble and interrogate extensive perovskite device datasets, deriving a single stability indicator that normalises disparate stress‐test conditions and enables direct comparison across over seven thousand devices. This approach has highlighted the most effective compositional and processing routes for enhanced long‐term performance. In parallel, high-throughput robotic experimentation coupled with predictive modelling uncovered a temperature‐dependent stability reversal in multi-cation perovskites, revealing that cation ratios optimal at elevated ageing temperatures may be deleterious at operating conditions below 100 °C, and proposing precise MA–Cs/Rb compositions to maximise device lifetime. Foundational work has also demonstrated rapid in silico screening of thousands of candidate hybrid organic–inorganic perovskites, combining machine learning with density functional theory to identify lead-free compounds with appropriate bandgaps and environmental resilience.

Machine Learning Applications in Perovskite Materials publication trend

The graph below shows the total number of articles in machine learning applications in perovskite materials across all publications each year (not limited to Nature Index journals).

Technical terms

Machine learning: Algorithms enabling data-driven prediction and optimisation of material properties.

Bayesian optimisation: A sequential design strategy using probabilistic models to identify optimal experimental parameters.

High-throughput experimentation: Automated synthesis and characterisation of large numbers of samples to accelerate data acquisition.

Device acceleration platform: Integrated robotic system for automated fabrication and testing of devices under varied conditions.

Bandgap: Energy difference between valence and conduction bands determining light absorption in semiconductors.

Cation engineering: Tuning of positively charged ions within the perovskite lattice to modify stability and performance.

Photoluminescence: Emission of light from a material following photon absorption, used to assess film quality.

References

1. Precise control of process parameters for >23% efficiency perovskite solar cells in ambient air using an automated device acceleration platform. Energy & Environmental Science (2024).
2. Optimizing Perovskite Thin‐Film Parameter Spaces with Machine Learning‐Guided Robotic Platform for High‐Performance Perovskite Solar Cells. Advanced Energy Materials (2023).
3. Toward Self-Driven Autonomous Material and Device Acceleration Platforms (AMADAP) for Emerging Photovoltaics Technologies. Accounts of Chemical Research (2024).
4. Accelerated discovery of stable lead-free hybrid organic-inorganic perovskites via machine learning. Nature Communications (2018).
5. Discovery of temperature-induced stability reversal in perovskites using high-throughput robotic learning. Nature Communications (2021).
6. Big data driven perovskite solar cell stability analysis. Nature Communications (2022).
7. Machine learning for perovskite materials design and discovery. npj Computational Materials (2021).