This team is using exascale-ready open-source software to exploit advanced, high-level quantum chemical methods to study transition metal systems with non-trivial electronic structure features, then use the data generated to train a machine-learning model that circumvents known limitations of the methods.
An under-appreciated reality of present-day digital electronics is that the electricity consumed for each digital operation is unsustainable. The power per operation of the next generation of microelectronic devices must be drastically smaller. Achieving that requires big advances that bridge current knowledge gaps. A plausible route forward is quantum technologies that manipulate electron spin rather than charge, as in current systems (for example, single-molecule spintronic devices). A key fundamental science challenge to that route is the limitation of current electronic structure methods to model transition metal physics and chemistry in diverse environments accurately. This is in stark contrast with other areas of chemistry and materials science for which contemporary electronic structure methods have provided invaluable predictive insight.
This project will use exascale-ready open-source software to exploit advanced, high-level quantum chemical methods to study transition metal systems with non-trivial electronic structure features, then use the data generated to train a machine-learning model that circumvents known limitations of the methods. The generated model will have high value for exploring the phenomena critical to emerging next-generation low-power, high-storage density technologies such as molecular-spin-based qubits for quantum information systems and molecular-complex spintronic components. We also expect the model to be valuable for exploring catalysts that promote nonnatural transformations within living organisms with potential applications in bioremediation and carbon uptake. A collaboration with the University of Florida will test the generated model on qubits and spin-crossover.
This research is funded by the Center for Scalable Predictive methods for Excitations and Correlated phenomena (PNNL, FWP 70942) and the Center for Molecular Magnetic Quantum Materials (University of Florida, DE-SC0019330). A portion of the research is funded under the Laboratory Directed Research and Development Program at Pacific Northwest National Laboratory, a multiprogram national laboratory operated by Battelle for the U.S. Department of Energy.
