Solvent fluctuations and nuclear quantum effects modulate the molecular hyperpolarizability of water

Abstract

Second-Harmonic Scatteringh (SHS) experiments provide a unique approach to probe noncentrosymmetric environments in aqueous media, from bulk solutions to interfaces, living cells and tissue. A central assumption made in analyzing SHS experiments is that the each molecule scatters light according to a constant molecular hyperpolarizability tensor β (2) . Here, we investigate the dependence of the molecular hyperpolarizability of water on its environment and internal geometric distortions, in order to test the hypothesis of constant β (2) . We use quantum chemistry calculations of the hyperpolarizability of a molecule embedded in point-charge environments obtained from simulations of bulk water. We demonstrate that both the heterogeneity of the solvent configurations and the quantum mechanical fluctuations of the molecular geometry introduce large variations in the non-linear optical response of water. This finding has the potential to change the way SHS experiments are interpreted: in particular, isotopic differences between H 2 O and D 2 O could explain recent SHG scattering observations. Finally, we show that a simple machine-learning framework can predict accurately the fluctuations of the molecular hyperpolarizability. This model accounts for the microscopic inhomogeneity of the solvent and represents a first step towards quantitative modelling of SHS experiments.

Using Gaussian process regression to simulate the vibrational Raman spectra of molecular crystals

“…Most efforts have been concentrated into predicting total energies and forces from atomic coordinates [5][6][7][8][9][10][11][12][13], which are most often the largest cost in a first-principles simulation. More recently, machine-learning models have been also applied to the prediction of response properties of molecules [14][15][16][17][18]. When dealing with the response of a material to an applied field, the cost of a first-principles calculation is often larger than that of force evaluation.…”

“…Here we investigate frameworks to obtain accurate predictions of the dielectric response properties for a number of consecutive molecular dynamics (MD) configurations, that are necessary to converge simulated vibrational Raman spectra of molecules and molecular crystals. We compare different flavors of Gaussian process regression (GPR), which is a method that has already been proven to be efficient in predicting dielectric response properties [14,15,17]. In particular, we compare standard GPR schemes with symmetry-adapted (SA)-GPR [30], which is advantageous when describing tensorial quantities.…”

Communication: Mean-field theory of water-water correlations in electrolyte solutions

“…As such it is clearly only a first step in a complete description of the experimental data, which should also include a re-evaluation of the molecular hyperpolarizability tensor, 30 particularly when probed by femtosecond laser pulses. 31 See supplementary material for more detailed derivations of the formulas used in the main text, as well as a list of the numerical values of physical constants used.…”

Atomic-Scale Representation and Statistical Learning of Tensorial Properties

“…where we have defined the best alignment operator aŝ R X →X =R X →R T X → . This strategy has been successfully used in the learning of electronic multipoles of organic molecules [10] as well as for predicting optical response functions of water molecules in their liquid environments [12]. For the latter example, a representation of the best-alignment structural comparison is reported in Fig.…”

“…The purpose of a statistical learning model is the prediction of regression targets by means of simple and easily accessible input parameters [1]. In chemistry, physics and materials science, regression targets are usually scalars or tensors, including electronic energies [2][3][4][5], quantummechanical forces [6][7][8], electronic multipoles [9][10][11], response functions [12][13][14][15] and scalar fields like the electron density [16][17][18]. For ground-state properties, the regression input usually consists of all the information connected with the atomic structure at a given point of the Born-Oppenheimer surface, e.g.…”