Data-driven optimization and machine learning analysis of compatible molecules for halide perovskite material

Shaojun Wang,Yiru Huang,Wenguang Hu,Lei Zhang

DOI: https://doi.org/10.1038/s41524-024-01297-4

IF: 12.256

2024-05-30

npj Computational Materials

Abstract:Optoelectronic stability of halide perovskite material in hostile conditions such as water is rather limited, preventing them from further industrial deployment. Here, we optimize and perform machine learning analysis on CH 3 NH 3 PbI 3 materials with additives, solvents and post-treatment molecules using combined experimental and data-driven methods. A champion system consisting of a compatible tertiary molecular combination 'calcein+PbBr 2 + DMSO' active at diverse surfaces is identified, delivering a large aqueous photoelectrochemical (PEC) photocurrent of 10 -5 A/cm 2 and an improved aqueous stability of 92.5%. Subsequently, machine interpretation is provided to decouple the multi-molecule contributions with the assistance of genetic programming (GP) and extra-trees (ET) machine learning models, highlighting the intricate molecular features for the target outputs. The post-hoc density functional theory (DFT) calculation suggests the presence of multiple hydrogen bond and anion··π surface interactions to stabilize the interfacial structures. The present 'PEC + GP + ET + DFT' approach is suggested to be an effective approach to design and comprehensively evaluate molecule-modified materials.

materials science, multidisciplinary,chemistry, physical

What problem does this paper attempt to address?

This paper attempts to solve the problem of poor stability of halide perovskite materials in the water environment, which hinders their further industrial applications. Specifically, halide perovskite materials (such as CH₃NH₃PbI₃) are prone to degradation in humid or aqueous environments, resulting in a rapid decline in their photoelectric properties. This instability limits their applications in practical devices, especially in optoelectronic devices such as solar cells and photodetectors.

To solve this problem, the author adopted data - driven optimization and machine - learning analysis methods to study the modification of CH₃NH₃PbI₃ materials through additives, solvents, and post - treatment molecules to improve their stability in the water environment and photoelectrochemical properties. The main objectives of the study are:

1. **Improve water stability**: Enhance the stability of perovskite materials in the water environment by introducing compatible organic molecule combinations, such as "calcein + PbBr₂ + DMSO".
2. **Optimize photoelectrochemical properties**: Find molecular combinations that can significantly increase the photocurrent of materials in aqueous solutions, for example, achieving a photocurrent density as high as 10⁻⁵ A/cm².
3. **Reveal the mechanism of molecular contribution**: Use genetic programming (GP) and extra - tree (ET) machine - learning models to analyze the specific contributions of different molecules to material properties, and verify the interaction mechanisms between these molecules through DFT calculations.

Through these methods, the author hopes to design perovskite materials with excellent water stability and photoelectric properties, thereby promoting their deployment in practical applications.

### Main achievements
- **Discovery of effective molecular combinations**: The combination of "calcein + PbBr₂ + DMSO" significantly improves the water stability and photocurrent output of the material, achieving a 92.5% improvement in water stability and a photocurrent density of 10⁻⁵ A/cm².
- **Machine - learning explanation**: Through genetic algorithms and extra - tree models, the influence of multi - molecule systems on photoelectrochemical properties was analyzed, revealing the importance of the synergistic effect between molecules.
- **DFT calculation verification**: Through density functional theory (DFT) calculations, the existence of intermolecular forces such as hydrogen bonds and anion - π interactions was confirmed, explaining the mechanism of interface structure stability.

In summary, this study provides a comprehensive method combining experiments, machine - learning, and theoretical calculations, aiming to design and evaluate molecular - modified perovskite materials that can maintain good performance in harsh water environments.