Abstract:Vibrational spectroscopy provides direct information on molecular environment and motions but, its interpretation is often hampered by band broadening. Over the past decade, two-dimensional (2D) vibrational spectroscopy has emerged as a promising technique to overcome a number of difficulties associated with linear spectroscopy and provided significantly detailed information on the structure and dynamics of complex molecules in condensed phases. This Account reviews recently developed computational methods used to simulate 1D and 2D vibrational spectra. The central quantity to calculate in computational spectroscopy is the spectroscopic response function, which is the product of many contributing factors such as vibrational transition energies, transition moments, and their modulations by fluctuating local environment around a solute. Accurate calculations of such linear and nonlinear responses thus require a concerted effort employing a wide range of methods including electronic structure calculation (ESC) and molecular dynamics (MD) simulation. The electronic structure calculation can provide fundamental quantities such as normal-mode frequencies and transition multipole strengths. However, since the treatable system size is limited with this method, classical MD simulation has also been used to account for the dynamics of the solvent environment. To achieve chemical accuracy, these two results are combined to generate time series of fluctuating transition frequencies and transition moments with the distributed multipole analysis, and this particular approach has been known as the hybrid ESC/MD method. For coupled multichromophore systems, vibrational properties of each chromophore such as a peptide are individually calculated by electronic structure methods and the Hessian matrix reconstruction scheme was used to obtain local mode frequencies and couplings of constituting anharmonic oscillators. The spectra thus obtained, especially for biomolecules including polypeptides and proteins, have proven to be reliable and in good agreement with experimental spectra. An alternative to the hybrid method has also been developed, where the classical limit of the vibrational response function was considered. Its main attraction is the capability to obtain the spectra directly from a set of MD trajectories. A novel development along this direction has been achieved by using quantum mechanical/molecular mechanical (QM/MM) force fields for the accurate description of vibrational anharmonicity and chromophore polarization effects. The latter aspects are critical in the 2D case because classical force fields employing harmonic intramolecular potential cannot produce reliable 2D signal. We anticipate that the computational methods presented here will continue to evolve along with experimental advancements and will be of use to further elucidate ultrafast dynamics of chemical and biological systems.

Efficient and Accurate Simulations of Vibrational and Electronic Spectra with Symmetry-Preserving Neural Network Models for Tensorial Properties

Tensorial properties via the neuroevolution potential framework: Fast simulation of infrared and Raman spectra

Accelerating Molecular Vibrational Spectra Simulations with a Physically Informed Deep Learning Model

Transferability of Machine Learning Models for Predicting Raman Spectra

Accurate and Affordable Simulation of Molecular Infrared Spectra with AIQM Models

Machine Learning Molecular Dynamics for the Simulation of Infrared Spectra

Neural canonical transformations for vibrational spectra of molecules

A deep learning model for predicting selected organic molecular spectra.

Machine Learning Assisted Interpretation of 2D Solid-State Nuclear Magnetic Resonance Spectra.

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

Performance metrics for tensorial learning: prediction of Li4Ti5O12 nuclear magnetic resonance observables at experimental accuracy

A neural network protocol for electronic excitations of N-methylacetamide.

Resolution-vs.-Accuracy Dilemma in Machine Learning Modeling of Electronic Excitation Spectra

Raman spectra of amino acids and peptides from machine learning polarizabilities

Neural Network-Based Tensor Model for Nematic Liquid Crystals with Accurate Microscopic Information

Raman Spectrum and Polarizability of Liquid Water from Deep Neural Networks

Computational vibrational spectroscopy of peptides and proteins in one and two dimensions

Neural Network Approach for Predicting Infrared Spectra from 3D Molecular Structure