Abstract:Raman spectroscopy is an important characterization tool with diverse applications in many areas of research. We propose a machine learning method for predicting polarizabilities with the goal of providing Raman spectra from molecular dynamics trajectories at reduced computational cost. A linear-response model is used as a first step and symmetry-adapted machine learning is employed for the higher-order contributions as a second step. We investigate the performance of the approach for several systems including molecules and extended solids. The method can reduce training set sizes required for accurate dielectric properties and Raman spectra in comparison to a single-step machine learning approach.

What problem does this paper attempt to address?

The paper aims to address the problem of accurately predicting the dielectric properties of materials (specifically the polarizability tensor) and their Raman spectra at a lower computational cost. To solve this problem, the authors propose a method that combines the Linear Response Model (LRM) with Machine Learning (ML), referred to as the "Δ-ML" method. Specifically, the main contributions of the paper can be summarized as follows: 1. **Proposing the Δ-ML method**: This method first uses a simple physical model (Linear Response Model) to obtain initial estimates, and then employs machine learning techniques to further correct these estimates, thereby improving prediction accuracy. This approach effectively reduces the need for a large training set and enhances prediction accuracy. 2. **Addressing the limitations of traditional methods**: Traditional methods based on molecular dynamics simulations can capture complex phenomena such as anharmonic effects but are computationally very time-consuming. While Density Functional Perturbation Theory (DFPT) is accurate, it has a high computational cost. Therefore, the Δ-ML method proposed in the paper aims to combine the advantages of both, reducing computational cost while maintaining high prediction accuracy. 3. **Validating the effectiveness of the method**: The authors demonstrate the superiority of the Δ-ML method over the Direct-ML method in predicting polarizability tensors and Raman spectra by testing it on various systems, including small molecules and extended solids. Experimental results show that the Δ-ML method can achieve similar predictive performance with significantly reduced training data requirements. In summary, the main goal of the paper is to develop an efficient and accurate method for predicting the dielectric properties and Raman spectra of materials, especially when considering temperature effects and anharmonic effects. The Δ-ML method effectively addresses this issue by combining simple physical models with advanced machine learning techniques.

Delta Machine Learning for Predicting Dielectric Properties and Raman Spectra

Transferability of Machine Learning Models for Predicting Raman Spectra

A Bond-Based Machine Learning Model for Molecular Polarizabilities and A Priori Raman Spectra

Polarizability Models for Simulations of Finite Temperature Raman Spectra from Machine Learning Molecular Dynamics

Learning Dipole Moments and Polarizabilities

Raman spectra of amino acids and peptides from machine learning polarizabilities

A chemical bond-based machine learning model for dipole moment: Application to dielectric properties of liquid methanol and ethanol

Recent Progresses in Machine Learning Assisted Raman Spectroscopy

Machine Learning Molecular Dynamics for the Simulation of Infrared Spectra

Machine Learning Isotropic g Values of Radical Polymers

Spectroscopy Learning: A Machine Learning Method For Study Diatomic Vibrational Spectra Including Dissociation Behavior

Accurate Machine Learning Prediction of Protein Circular Dichroism Spectra with Embedded Density Descriptors

Accurate prediction of dielectric properties and bandgaps in materials with a machine learning approach

A Machine Learning Vibrational Spectroscopy Protocol for Spectrum Prediction and Spectrum-Based Structure Recognition

Machine Learning for Vibrational Spectroscopic Maps

Deep Spatial Learning with Molecular Vibration

Applying machine learning techniques to predict the properties of energetic materials

Machine Learning in QM/MM Molecular Dynamics Simulations of Condensed-Phase Systems

Spectroscopic constants from atomic properties: a machine learning approach

Machine Learning for Mie-Tronics