Scientific Achievement
An autonomous workflow to efficiently design upconverting nanoparticles (UCNPs) with enhanced UV-violet emission.
Significance and Impact
This is the first demonstration of using machine learning to accelerate the discovery of new UCNP heterostructures. This approach can be expanded to structures with higher complexity, accelerating the discovery of novel UCNPs while domain-specific knowledge is being developed.
Research Details
- The team used Bayesian Optimization (BO) to optimize the emission of lanthanide-doped multishell UCNPs using high-throughput kinetic Monte Carlo (kMC) simulations.
- They achieved 10- & 110-fold increase in the UV-V intensity of doubly & triply doped UCNPs after 22 & 40 iterations, respectively.
- This workflow learned commonly known UCNP design rules 500x faster than human researchers.
