Inferring Multicomponent Properties Through Thermal Lens

Experimental detection of molecular diffusion in multi-component mixtures is of prime importance for much of the physical sciences. Light matter interactions are always associated with residual heating in the sample. We demonstrate that using non-linear optical processes, we can infer diffusive properties of molecules in complexes. In turn, with this information coupled with a suitable machine learning model and additional (known) molecular data, we are able to form correlational models for phenomenological properties. As a concrete example, we show how the elusive human interpretation of "scent" can be computed through this novel theoretical model with our experimental data.