Data-Driven Discovery of Point Defects for Quantum Technology

Optically addressable, spin-carrying point defects in solids represent a versatile platform for a range of quantum technologies. However, only a modest number of useful spin-defects are currently known and their properties are often suboptimal. I will present our data-driven strategy to the in-silico design of novel point defects for quantum technology applications. Using the GPAW electronic structure code, we perform high-throughput density functional theory (DFT) calculations to screen thousands of point defects in bulk and two-dimensional host materials. The elementary physical properties comprising the key design parameters are the magnetic states and sub-levels of the defect, the optical excitation energy (zero phonon line), and the photoluminescence (PL) line shape. We use the Python workflow framework TaskBlaster to compose, manage, and execute the computational workflows, which can be complex and involve dynamic branching and inter-dependent tasks. By integrating the DFT calculations with machine learning technologies, we achieve an order of magnitude speed-up of the otherwise costly calculation of PL spectra. We have created a FAIR online database with all the results from our high-throughput calculations. The Quantum Point Defects (QPOD) database currently holds data for more than 10.000 point defects across a range of different host materials, and should be useful for identifying novel defects for various applications and/or for training of machine learning models.