Electron dynamics mediate the water-carbon π bond

Abstract

The intermolecular interaction between a water molecule and the electrons in aromatic π systems--the water-π bond--lies at the heart of many chemical processes, yet its properties remain challenging to measure experimentally and model computationally. Infrared spectroscopy of pyrene anions hydrated by a single water molecule reveals vibrational and electronic motions that are often hidden in condensed phase measurements. Results from new machine-learning approaches to potentials and dipole moments show that the electron dynamics of the aromatic π cloud quench signals from some of water's vibrations and amplify others. The observed interplay between electronic and vibrational motions has general implications for modeling intermolecular interactions between water and aromatic systems in clusters, solutions, and at interfaces.

Figures (4)

• Figure 1: Structure and energy level diagram for the vibrations of the anionic pyrene monohydrate cluster in its deuterated forms. A typical configuration of the pyrene monohydrate anion cluster and the directions of the symmetric (sym) and antisymmetric (asym) transition dipole moments (A) along with relative vibrational energies for the water and pyrene moieties (B). Both H$_2$O and D$_2$O have symmetric and antisymmetric stretching vibrations, depicted in dark blue and red, respectively. Deuteration in HOD introduces an isotope effect that breaks the local symmetry of the water molecule and separates vibrations into two local modes, $\ket{\mathrm{OH}}$ and $\ket{\mathrm{OD}}$, depicted in teal. The OH resonances occur above $\approx$ 3500 cm$^{-1}$ and are vibrations of the water molecule. The transitions at lower frequencies overlap the spectral region of the CH stretching fundamentals, $\ket{\omega_s^{CH}}$, and the CC stretching/CH bending combination bands ($\ket{\omega_B^{CH}}+\ket{\omega_s^{CC}}$ of the pyrene moiety Heinrich2, shown in purple. This region of the spectrum is thus more complicated to interpret than the resonances from the OH vibrations. More details appear in the Supplementary Materials.
• Figure 2: Free energy landscapes of the water molecule moving across the surface of the pyrene anion highlight the differences that the empirical model and the machine-learned one predict for intermolecular structure. The empirical model assigns points charges to atoms of Pyr$^-$ according to the Merz-Singh-Kollman ESP technique (A) and corresponding free energy derived probability density at 75 K for the oxygen of the water molecule to lie in the plane of the Pyr$^-$ molecule (white frame) (B). The electron density from electronic structure (C) is more diffuse and contoured than the skeletal charge distribution in (A), which can impact interactions at short distances. The free energy surface from the machine-learned potential is markedly different than that in (B), showing four new basins at the points of the butterfly shape, marked with a red asterisk in the top right quadrant. All electronic structure calculations, including those that generate the training data for the machine-learned potential use density functional theory with the $\omega$B97-XD/def2-TZVPP method and basis set.
• Figure 3: Spectra of the OH and OD stretching regions for Pyr$^-\cdot$H$_2$O and Pyr$^-\cdot$D$_2$O compared to predictions from the empirical and machine-learned models highlight the relative roles of electronic and vibrational dynamics. (Top left panels) Experimental IR spectra of Pyr$^-\cdot$H$_2$O$\cdot$Ar$_2$ (A, taken from ref. Lemessurier1) and Pyr$^-\cdot$D$_2$O$\cdot$Ar$_2$ (D). (Middle left panels) Nuclear vibrational spectra (NVS) calculated at 75 K with the empirical potential (red) the ML potential (teal) show nuclear dynamics but neglect electron dynamics. Both potentials show an intense antisymmetric stretching peak that does not appear in the experimental data (repeated in gray). (Bottom) The IR spectra calculated from ML dynamics using the fluctuations of the total dipole moment, inferred from a machine-learned method, (blue) match the experimental data much better and show the suppression of the intensity of the antisymmetric stretching fundamental relative to the symmetric one in both normal and heavy water (C, F). The electrons in the $\pi$ orbitals of the Pyr$^-$ molecule follow the nuclear vibrations, forming an image radiation dipole that quenches the signal from the antisymmetric vibration, lying mostly in the pyrene plane, while amplifying the symmetric vibration, which lies mostly normal to the pyrene plane (G,H).
• Figure 4: Spectra of the OH and OD stretching regions for Pyr$^-\cdot$HOD from experiment and computation. (Top) The experimental IR spectra of Pyr$^-\cdot$HOD$\cdot$Ar$_2$ in the OD stretching (A) and OH stretching regions (D). Some of the vibrational signatures in the OD stretching region originate from vibrational levels of the pyrene anion (see Supplementary Materials). (Middle) Nuclear vibrational spectra (NVS) calculated at 75 K (B, E) with the empirical potential (red) and ML potential (teal) neglect electron dynamics and have a large peak for the blue component of each doublet. These transitions are not found in the experiment (gray). (Bottom) The IR spectra calculated from ML dynamics at 75 K with an ML-predicted dipole (blue) match the experimental data more closely (C, F). The doublet structure emerges from different average orientations of the OH and OD groups (G,H). The tilted orientation of the H$_2$O is exaggerated here for illustration.