Machine Learning and Single Molecule Magnets

"""Single-molecule magnets (SMMs) are complexes of metal ions which couple
ferromagnetically or antiferromagnetically, exhibiting a large anisotropy in their magnetic
response. They are of current interest in molecular spintronics, magnetic cooling and
information storage, but finding an SMM with the desired properties (i.e., magnetic
response and its anisotropy) currently requires experimental screening, which takes a
large amount of time. Screening candidate SMMs with quantum-mechanical calculations is
also possible, but decreases the screening time from weeks to days, meaning that a
substantial outlay is still required.
Machine-learning (ML) methods are becoming an extremely popular method to circumvent
time-intensive quantum-mechanical calculations, with modern ML approaches able to
predict the results of solving the Schrödinger equation with high accuracy. The student will
develop the methodology to predict the magnetic properties of spin-crossover complexes
using ML, choosing the best ML method and descriptors to give a model that requires no
specialist knowledge or computationally intensive calculations but nonetheless gives highly
accurate results. In addition, they will carry out experiments to characterise potential single
molecule magnets, allowing a feedback between theory and experiment, with each strand
of the project guiding the other."""