Relating terahertz spectra to the structural dynamics deep in supercritical water

Recently, supercritical water gained a lot of attention in the scientific community because numerous applications are envisaged in both, science and industry. One major advantage of supercritical water is the tunable density, which allows one to specifically select ideal solvation conditions. Notably, macroscopic and microscopic properties of supercritical water differ dramatically from those of ambient liquid water which is why a deep understanding of solvation in supercritical water is still lacking. Moreover, it fulfills most of the criteria of a "green" solvent because it is environmentally friendly, nontoxic and easily accessible. In the case of ambient liquid water, the unique solvation properties are largely attributed to the H-bond network. Here, THz spectroscopy emerged as a key tool because this H-bond network can directly be probed by the H-bond network mode which resides around 200 cm${}^{-1}$ in the THz regime of the IR spectrum. It is thus suggestive to employ THz spectroscopy to assess how the H-bond network changes compared to ambient conditions and thereby analyze solvation properties of supercritical water. Clearly, before turning to actual solutions it is key to fully understand the microscopic properties of the pure solvent first. The goal of this dissertation therefore is to investigate how microscopic properties of the water molecules, e.g. the H-bond network, cluster size distributions or molecular dipole moments, change in the supercritical phase with respect to the ambient liquid phase. Here a clear focus lays on the computation and detailed decomposition of THz spectra which is an experimentally measurable observable and provides a link to the mentioned microscopic properties. As such, this dissertation provides a first step towards the understanding of solvation in supercritical water and how it differs qualitatively from water at ambient conditions. To ensure that the performed ab initio simulations are truly supercritical, the critical point of the employed density functional (RPBE-D3) needs to be known. This issue has been addressed in the early stage of this dissertation by estimating the liquid-vapor coexistence curve of water by ab initio Gibbs Ensemble Monte Carlo simulations. Thereby the critical point is extrapolated and it turns out that the experimental critical point is overestimated by about 60 K corresponding to roughly 10 %. However, by choosing a temperature well above the critical temperature as predicted by the simulations, I am sure that the performed computations truly mimic the supercritical phase of water. Compared to ambient liquid water it turns out that supercritical water features a much diffuser structure which manifests in much looser radial and joint distribution functions. Most importantly, also the definition of H-bonds is thus more arbitrary than in ambient liquid water. Hence, all derived properties, for example cluster size distributions or the average number of H-bonds per molecule, quantitatively strongly depend on the chosen criterion. Absolute values must therefore be taken with care but they can still be compared quantitatively if the same H-bond criterion is employed consistently. Doing so, a clear trend is found that the average number of H-bonds per molecule and the average cluster size gradually increase with density. The much looser definition of H-bonds correlates with the fact that H-bonds are less stable in supercritical water than in ambient liquid water. Indeed, the lifetime of H-bonds is much shorter causing water clusters to frequently break and reform. It therefore seems that H-bonds in general play a much less important role in supercritical water simply because the kinetic energy is that large that the potential barrier of the H-bond can frequently be passed. Even at high densities where one large interconnected H-bond network exists (so-called percolating), the lifetime of an individual H-bond is still about one order of magnitude shorter than at ambient conditions. This understanding of a H-bond is in stark contrast to the one at ambient conditions where H-bonds are rather strong and the H-bond network is comparatively rigid. Independent on the chosen H-bond criterion, the average number of H-bonds depends linearly on the average molecular dipole moment which, in turn, does intrinsically not depend on any H-bond criterion. Therefore, the molecular dipole moment is a measure of the extent of H-bonding in supercritical water. This is reasonable as the molecular dipole moment largely depends on collective effects among the neighboring water molecules. Consequently, the amount of H-bonding can quantitatively be compared between two states by comparing the respective molecular dipole moments. Interestingly, the found linear relation only holds for isothermal state points but fails to capture the temperature difference between ambient liquid water and supercritical water at 1.0 kg L${}^{-1}$. To capture this temperature effect, additional simulations are currently running starting at ambient conditions and systematically increasing the temperature while keeping the density constant, see below. While investigating the IR spectrum it is found that the librational band totally dominates the low-frequency regime up to 1000 cm${}^{-1}$. Although the libration band systematically blue-shifts with increasing density, it remains significantly red-shifted compared to ambient liquid water at all investigated supercritical densities. Most importantly, the location of the band is sensitive to the local environment of the individual molecule and generally shifts to higher frequencies if more H-bonds are locally present. Hence, the libration band is certainly also responsive to the presence of different solutes which could be a matter of future research. In contrast, the exact location of the H-bond band is less clear yet because the distinct spectral decomposition techniques yield partially contradicting results. While it is certain that the H-bond band still exists in supercritical water, it is completely masked by the dominating libration band. The issue might be resolved by monitoring the change of the H-bond band starting at ambient conditions while systematically increasing the temperature while keeping the density constant. Currently, simulations are running at 1.0 kg L${}^{-1}$ and 350, 400, 450, 500, 550, 600 and 650K to deduce the location of the H-bond band from its location at ambient conditions. Simultaneously, the intermolecular H-bond band at the low-density end of supercritical water might be deduced from the H-bond band in an isolated water dimer at the same temperature. At very low densities, supercritical water mostly consists of monomers, followed by a smaller amount of dimers. Clearly, only dimers can contribute intermolecular H-bond vibrations to the IR spectrum. Certainly it is not possible to simulate the water dimer in vacuum above the critical temperature as the H-bond between the two water molecule will quickly break and the two molecules depart. Obviously, the simulation is worthless as soon as the water molecules depart and thus it is currently carried out at lower temperatures. All computed properties, be it structural, dynamical, electronic or spectral properties, smoothly change with the macroscopic density and therefore no indication for a discontinuous change is found. Still it is undeniable that supercritical water behaves "more liquid-like" at very high densities and "more gas-like" at very low densities in a loose sense, for instance in the spirit of the Fisher-Widom criterion. However, there is no indication that there is a unique separation line between liquid- and gas-like regimes. This fully agrees with the fact that the Widom-Lines which have been claimed to be such a unique separation line are intrinsically not unique. As such I can conclude that microscopic and macroscopic properties change in a perfectly continuous manner in supercritical water. Having said that this dissertation provides the first step towards the understanding of solvation processes in supercritical water by studying pure supercritical water first, it is natural to investigate solutions in the future. Recall that at low densities, supercritical water can dissolve nonpolar compounds whereas at high densities it prefers polar compounds. An interesting test case might be glycine or valine, being both relatively small and thereby computationally affordable and whose THz spectrum in ambient liquid water has recently been assessed in detail. Here, one would expect that the zwitterionic form is preferred at high densities whereas the neutral form is preferred at low densities. In addition, the highly hydrophobic side chain is expected to feature significant different solvation patterns depending on the water density. Especially interesting would be the change between these two extreme cases by gradually varying the density. Still, one would first have to test if these two molecules are stable in supercritical water, having the highly enhanced corrosivity in mind. Nevertheless also other compounds, e.g. simple ions, whose THz spectra are well-known in aqueous solutions at ambient conditions might also be interesting as further projects.