msc project: machine-learning models for molecular simulations
TRANSCRIPT
Free energy surface
BackgroundPhotoactive molecules have countless applications in energy conversion andstorage, including, e.g., triplet-triplet annihilation up-conversion, singletfission, and molecular solar thermal storage. The functionality of thesematerials is based on their structure being sensitive to light. The absorptionof a photon can, e.g., trigger a change in the structure of an individualmolecule or change the interaction between.GoalThe goal is to build models for the prediction of molecular properties suchas energy, forces, and absorption strength as a function of the atomicconfiguration with an accuracy that is on-par with the underlying first-principles calculations but at a fraction of the computational cost.Scope and techniques• Development of a framework for model construction,
validation, and sampling using Python• Linear models based on atomic scale descriptors• Kernel methods (sGDML)
MSc project: Machine-learningmodels for molecular simulations
DOI: 10.1063/1.5078687
Potential energy surface
Kernel method (sGDML) Photoactive molecules
Strongly collaborative environmentCollaboration with experimentPotential PhD project
For more information:https://materialsmodeling.org
ContactProf. Paul Erhart [email protected]