Machine Learning Approach Based on a Range-Corrected Deep Potential Model for Efficient Vibrational Frequency...

Area of Science:

• Computational chemistry
• Machine learning applications in spectroscopy

Background:

• Accurate vibrational spectrum simulations using high-accuracy methods are computationally expensive.
• Existing methods face challenges in balancing accuracy and computational efficiency.

Purpose of the Study:

• To introduce a machine learning approach, the range-corrected deep potential (DPRc) model, for accelerating vibrational spectrum simulations.
• To enhance the computational efficiency of calculating vibrational frequencies.

Main Methods:

• Implemented a machine learning approach based on the DPRc model, dividing systems into "probe" and "solvent" regions.
• Trained and tested the model on formic acid C═O and MeCN C≡N stretching vibrational frequency shifts in water.
• Investigated the impact of region division, one-body correction, cut range, and training data size.

Main Results:

• The DPRc model with a single-molecule "probe region" achieved stable accuracy.
• The method demonstrated a roughly 10-fold increase in speed compared to regular deep potential.
• Training time was reduced by approximately four times.

Conclusions:

• The DPRc model offers an efficient and accurate approach for vibrational spectrum simulations.
• The method is easily applicable and extendable to various spectral calculations.
• This machine learning strategy significantly improves computational efficiency in spectroscopy.