Predicting Optical Bandgaps in C60 and Functionalized Derivatives from Limited Data for Renewable Energy Applications

Tunçel M., Kurban H., Serpedin E., KURBAN M.

19th IEEE International Conference on Compatibility, Power Electronics and Power Engineering, CPE-POWERENG 2025, Antalya, Türkiye, 20 - 22 Mayıs 2025, (Tam Metin Bildiri) identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/cpe-powereng63314.2025.11027245
  • Basıldığı Şehir: Antalya
  • Basıldığı Ülke: Türkiye
  • Anahtar Kelimeler: AI for Science, Density Functional Theory, Gaussian Process Regression, Optimal Bandgap Prediction, Small Data Learning
  • Ankara Üniversitesi Adresli: Evet

Özet

Accurate optical bandgap prediction is essential for advancing renewable energy materials, but data scarcity and complex non-linear dependencies limit traditional approaches, especially for C60-based materials in organic photovoltaics. We propose a novel data-efficient machine learning framework that integrates Gaussian Process Regression (GPR) with Density Functional Theory (DFT) simulations to predict absorbance spectra and compute optical bandgaps from limited data. Our approach leverages Bayesian inference for uncertainty quantification, significantly reducing reliance on computationally expensive DFT calculations while maintaining accuracy. We demonstrate that our method generalizes across C60 and functionalized derivatives, enabling scalable and cost-effective bandgap prediction with minimal training samples.