Work

Machine learning-assisted optimization of eco-friendly Sb₂S₃ thin-film solar cells: a sustainable SCAPS-1D and SHAP-based design framework

Year: 2026

Type: article

Abstract: The development of earth-abundant and non-toxic photovoltaics is imperative for a sustainable energy future. Antimony sulfide (Sb₂S₃) is a promising absorber material, yet its optimization has traditionally relied on empirical, resource-intensive methods. This study presents an interpretable machine learning (ML) framework that integrates SCAPS-1D device simulations with XGBoost and SHAP analysis to efficiently optimize and elucidate the physics ofTiO₂/Sb₂S₃/CuI solar cells. From a systematically generated dataset of 3,621 configurations, the XGBoost model achieved exceptional predictive accuracy for power conversion efficiency (PCE, R² = 0.997). SHAP analysis provided transformative physical insights, quantitatively identifying absorber defect density as the paramount performance driverand establishing a clear hierarchy over other parameters. Crucially, it revealed non-linear interactions, such as a critical defect density threshold (~10¹⁶ cm⁻³) and the conditional benefit of transport layer doping, which is only effective when coupled with a high-quality absorber. These ML-derived design rules guided a final optimization that significantly enhanced device performance. This work demonstrates a synergistic simulation–ML approach that not only accelerates the discovery of high-performance photovoltaic designs but also provides a deeply interpretable, resource-efficient framework for the sustainable development of next-generation solar cells.

Source: Energy and AI

DOI: 10.1016/j.egyai.2026.100743

Authors: Ajwa Ikram, Usama Ghulam Mustafa, Adham Hashibon, Mingqing Wang, Hafeez Anwar

Institutions: University of Agriculture Faisalabad, University College London

Language: English

Topic: Chalcogenide Semiconductor Thin Films

Subfield: Electrical and Electronic Engineering

Field: Engineering

Domain: Physical Sciences

SDG: Affordable and clean energy

Open Access status: gold

Estimated APC paid: $3,630