Integration of Quantum Chemistry, Statistical Mechanics, and Artificial Intelligence for Computational Spectroscopy: The UV–Vis Spectrum of TEMPO Radical in Different Solvents

Abstract

The ongoing integration of quantum chemistry, statistical mechanics, and artificial intelligence is paving the route toward more effective and accurate strategies for the investigation of the spectroscopic properties of medium-to-large size chromophores in condensed phases. In this context we are developing a novel workflow aimed at improving the generality, reliability, and ease of use of the available computational tools. In this paper we report our latest developments with specific reference to unsupervised atomistic simulations employing non periodic boundary conditions (NPBC) followed by clustering of the trajectories employing optimized feature spaces. Next accurate variational computations are performed for a representative point of each cluster, whereas intracluster fluctuations are taken into account by a cheap yet reliable perturbative approach. A number of methodological improvements have been introduced including, e.g., more realistic reaction field effects at the outer boundary of the simulation sphere, automatic definition of the feature space by continuous perception of solute–solvent interactions, full account of polarization and charge transfer in the first solvation shell, and inclusion of vibronic contributions. After its validation, this new approach has been applied to the challenging case of solvatochromic effects on the UV–vis spectra of a prototypical nitroxide radical (TEMPO) in different solvents. The reliability, effectiveness, and robustness of the new platform is demonstrated by the remarkable agreement with experiment of the results obtained through an unsupervised approach characterized by a strongly reduced computational cost as compared to that of conventional quantum mechanics and molecular mechanics models without any accuracy reduction.

Direct Force Constants for Spectroscopic Analysis from Quantum Mechanical Information: SmartField Force Field

“…We further refine the PMM results by adding the shift between 11 and 0 QM solvent molecules computed at the centroids of the clusters. As was previously reported, 22 this general approach offers a fast but accurate way to compute the optical properties of condensed-phase systems. In this context, this addition leads to an improvement of both absorption and emission of L a state, 267.8 and 305.3 nm respectively, which make them closer to the experimental data.…”

“…In contrast, a relatively smaller difference between the wavelengths of L a state of clusters is observed due to the ability of PMM of averaging the contribution of electrostatic potentials of all frames present in single cluster. 22 This is, for instance, particularly evident for second cluster of excitedstate trajectory, where an increase of 5.2 nm is shown with respect to the QM/MM result on the bare solute. We further refine the PMM results by adding the shift between 11 and 0 QM solvent molecules computed at the centroids of the clusters.…”

“…the computation of UV-vis spectra is a crucial step for improving the calculated absorption. 22 As result, we compute the electronic transition of L a state for both the bare solute and clusters including some solvent molecules close to the center of mass of indole which are treated at QM level. The QM/MM computations on bare solute show that the weighted average absorption is 257.0 nm, whereas the emission is 289.36 nm, resulting in a shift of about 32 nm with respect to the absorption.…”

“…Next, we compare the QM/MM results to the use an inexpensive perturbative approach as implemented in the latest developments of PMM. 22 These are shown in Table 5. The PMM results show that an maximum absorption occurs at 257.7 nm which differ by less than 1.0 nm from the QM/MM result on the bare solute.…”

“…It should be stressed that torsional profiles are expected to be fundamental when carrying out MD simulations with parameterized FF, since their values drive the conformational changes, and therefore, also the optical properties of the snapshots extracted from the dynamics. In order to apply the procedure previously developed in our work, 22 we took into account the distances between the tryptophan and first and second neighbouring atoms of each solvent molecule which best represent their interaction: O T rp -A 1 , O T rp -A 2 and HN T rp -A 1 HN T rp -A 2 distances with A 1 and A 2 being atoms of solvent. In addition to these, the values of the dihedrals in Figure 8 are also added to build the inital data set.…”