Multiscale simulations on the structure and dynamics of ionic liquids
Final Report Abstract
Reliable and transferable classical force fields for ionic liquids (ILs) are rare. For this reason, the main objective of this project was to establish a multi-scale modelling scheme, which combines information from high-level post-Hartree–Fock (pHF) electronic structure calculations for small length scales (few ions) and static systems, density functional theory (DFT) for intermediate length scales (up to 30 ion pairs) and short to intermediate time dynamics (tens of picoseconds) and finally classical force fields for larger length scales (hundreds of ion pairs) and longer time dynamics (hundreds of nanoseconds). One research line was thus to establish a method to optimize and generate an improved set of force field (FF) parameters. This was achieved by first comparing structures, relative energies, harmonic vibrational force fields and electrostatic moments for few ions as obtained on the pHF, DFT or classical force field level. Subsequently, electronic and geometric structure information of the bulk gathered on the DFT level were mapped to the classical scale. Apart from the test system dimethylimidazolium chloride [MMIM][Cl], which was used to develop corresponding methods, a broad range of imidazolium based cations was investigated combined with the anions [SCN]¯, [Cl]¯, and [DCA]¯. These studies were based on Car-Parrinello molecular dynamics (CPMD) simulation snapshots (on non-hybrid DFT level) of tens of ion pairs and allowed to gain insight into electronic structure of ionic liquids, especially in mechanisms of partial charge distribution as benchmarked against pHF calculations and the net–charge reduction, which has been identified as a model for implicit polarization and charge transfer. These features are not only characteristic for an IL, but it has also been shown, that they occur already on a very local scale. Indeed, viscous systems as IL’s need rather large samples and long time trajectories in order to properly describe density fluctuations and molecular diffusion; the computational demand for DFT calculations heavily limits the accessible time scales. Nevertheless in this project it was possible to simulate samples with 30 ion pairs for even 44.0 ps. The resulting picture was rather interesting: it could be shown that on a time scale, much beyond the one needed for the relaxation of local electronic degrees of freedom, IL’s have very localized structural properties, which means that only the immediate environment influences the molecular behaviour, however the local behaviour is subject of large fluctuations. This conclusion was then supported by using such essential features to model a classical system and simulate a much larger sample for a time scale of the orders of hudreds of nanoseconds; classical simulations were then able to reproduce dynamical properties properly. This implies that the basic input and the idea of locality and fluctuation of the DFT calculations is at least consistent with a more macroscopical picture of IL’s. A further test was performed by calculating the power spectra using the data of the DFT calculations. Despite power spectra being essentially a theoretical tool, they can be related to experimental spectra and assist their interpretation. Consistency between theory and experiments was found and it emerged that the picture of locality and fluctuation is indeed the essential feature of IL’s. This locality is also the reason for the good performance of partial charges that were derived from small ion pair clusters and employed for development of improved classical force fields. It was subsequently shown that the predicted charge transfer and polarization is in agreement with NMR experiments and measurements of the refractive index. From the computational side, consistency of the dipole moments given by the partial charges or by a Wannier analysis of the CPMD results was achieved. Though the width of dipole moment distribution could not be completely reproduced by the static partial charges, the increase in the dipole moment for an increasing number of interacting ion pairs (IPs) coincides very well. This shows that the proposed set of partial charges is well suitable for the bulk phase of an IL. Finally the short–range (SR) interactions of the classical FF were adapted in respect to the proposed set of partial charges. An algorithm was implemented that optimizes SR parameters based on an iterative adaption to match static properties given by the experimental mass density and the radial distribution functions derived from CPMD simulations. The routine has been successfully tested for [MMIM][Cl] and it has been shown that the implicit consideration of polarization and charge transfer models the dynamics already well, such that one can rely on static properties during the SR parameter optimization. This alleviates the tuning process, because the amount of simulation time required to sample a highly viscous IL decreases and, apart from the experimental mass density, only computational results are required. The latter aspect is of great importance for ILs, because a broad spectrum of experimental knowledge is not a given. Thus, during this work, the technique and computational framework to optimize FF parameters was established and a large data set was accumulated. Therefore we did not only reach our main target, but also provided a solid basis for future work, the development of a generic and transferable force fields for ILs. Moreover, based on CPMD results we could comment on the importance of hydrogen bonding in IL’s and its specific role in depolarization of water when is used as a solute in IL’s. All these results, beside the interesting insights on their own, were precious data for designing a multiscale modeling strategy for IL’s.
