Electron-vibrational renormalization in fullerenes through ab initio and machine learning methods.

Area of Science:

• Computational Materials Science
• Quantum Chemistry
• Condensed Matter Physics

Background:

• Nuclear vibrations influence electronic properties in various carbon materials.
• The impact of nuclear vibrations on fullerene electronic structures remains underexplored.
• Fullerenes are of significant theoretical and technological interest.

Purpose of the Study:

• To investigate the renormalization of electronic eigenvalues and the HOMO-LUMO gap in fullerenes due to nuclear vibrations.
• To analyze the effect of zero-point motion on a large set of fullerenes and their derivatives.
• To explore the potential of machine learning for predicting these vibrational effects.

Main Methods:

• Density-functional theory (DFT) calculations.
• Frozen-phonon method to incorporate nuclear vibrations.
• Machine learning models for classification and regression.

Main Results:

• Nuclear vibrations cause non-negligible HOMO-LUMO gap renormalization (above 0.1 eV) in fullerenes relevant to photovoltaics.
• The strength of this renormalization increases with the size of the electronic gap.
• Machine learning models can approximate renormalization predictions using ground-state calculation outputs.

Conclusions:

• Zero-point motion significantly affects fullerene electronic properties, particularly the HOMO-LUMO gap.
• These effects are crucial for fullerene applications in areas like photovoltaics.
• Computational efficiency can be enhanced by using machine learning for predicting vibrational effects.