← Back to Research

Undergraduate Research in Physics — Spring 2026

This page highlights undergraduate research completed during Spring 2026 at SUNY New Paltz and presented at the APS Global Physics Summit in Denver, Colorado.

Band Gap Prediction in h-BN Using mBJ Data and Physically Informed Descriptors

Michael Buccino and Kendra Scheele

Department of Physics and Astronomy, SUNY New Paltz

This project combines density functional theory and machine learning to predict the band gap of monolayer hexagonal boron nitride (h-BN). Standard PAW-DFT reproduces the optimized lattice constant well but underestimates the band gap. A random-forest model trained on mBJ reference data and physically informed composition descriptors improves the prediction while avoiding the computational cost of higher-level band-gap corrections.

Using 246 monolayer materials, the model achieved a mean absolute error of 0.336 eV and an R² score of 0.938. For held-out monolayer h-BN, the predicted mBJ band gap was 6.049 eV, close to the 5.917 eV reference value.

Poster (PDF)

Michael Buccino and Kendra Scheele presenting their h-BN band-gap project at the APS Global Physics Summit, Denver, March 2026.

Band Gap Prediction in h-BN

Using mBJ Data and Physically Informed Descriptors

Michael Buccino · Kendra Scheele · Greis J. Kim-Reyes

DFT foundation
Optimized h-BN lattice constant and calculated PAW-DFT band structure.

ML correction
Random-forest prediction trained on mBJ data and physical descriptors.

Key result
6.049 eV predicted versus 5.917 eV mBJ reference.

APS Global Physics Summit · Denver, Colorado · March 2026

Project highlights

• Standard PAW-DFT band gap for monolayer h-BN: approximately 4.6 eV.
• Optimized lattice constant: approximately 4.7 Bohr (2.49 Å), close to the accepted value.
• Machine-learning dataset: 246 monolayer materials with mBJ band-gap references.
• Model performance: MAE = 0.336 eV and R² = 0.938.
• Held-out h-BN prediction: 6.049 eV compared with the 5.917 eV mBJ reference.
• Presented at the APS Global Physics Summit, Denver, Colorado, March 2026.

Acknowledgments

This work was supported by the Academic Year Undergraduate Research Experience (AYURE) program at SUNY New Paltz and by the Research, Scholarship, and Creative Activities Office through the Student Opportunity Grant.

← Back to Research