From Covalent Systems to Bulk Phases: Addressing Structural Complexity with Computational NMR

ConspectusThe free induction decay (FID) signal acquired in a typical NMR experiment contains information about the chemical shifts, δ, and the spin–spin coupling constants, J, of the system investigated. These two parameters, particularly the chemical shifts, are very sensitive to both intramolecular and intermolecular perturbations. As such, they are also very good probes of the structure of both the molecule and the hosting solvent/matrix. This sensitivity is exploited in natural product studies to deduce the molecular structure of newly isolated compounds from the analysis of their NMR spectra. However, for complex carbon skeletons, the interpretation of the NMR data is far from trivial; the structural information is too deeply buried within the overlapping high-order multiplets and close resonances. In these cases, it is useful to compare the experimental NMR data of the unknown substance with the ones predicted by density functional theory (DFT) based methods for hypothetical molecules. Ideally, one will discard all putative structures resulting in a disagreement with the experiments and will keep the only one exhibiting an agreement within the benchmarked accuracy of the level of theory used. Two other sources of complexity, besides the topological complexity of natural substances, may strongly affect the interpretation of the NMR spectra. One, still related with covalent compounds, is the presence of heavy atoms which brings in relativistic effects in the NMR. Even for a simple molecular structure, they turn the interpretation of the NMR spectrum into a very difficult task since empirical rules often do not allow a full elucidation of the structure; thus, such effects can be accounted for only with relativistic versions of DFT. The other source of complexity is the presence of strong noncovalent interactions of the NMR probe molecule with its environment. In these cases, the full dynamics of the solute and solvent system has to be taken into account and the structure that is responsible for the observed NMR is in fact the average bulk structure of the solute–solvent system. Then, molecular dynamics (MD) simulations have to be coupled with the DFT-NMR calculations in order to predict the NMR properties. In turn, the comparison between the calculated and experimental data can shed light on the force field (FF) parameters used in the MD simulation. Therefore, computational NMR can be used to shed light on both covalent and noncovalent structural problems: in one case, the exploration of a discrete structural space will allow one to select the correct structure of an unknown compound among several hypothesis; in the other one, it will enable the fine-tuning of classical FF parameters over a continuum range of possibilities.