Integrated Computational and Experimental Structure Determination for Nanoparticles

Atomic details of nanostructures are important to materials performance for catalysis, solar energy, optoelectronics, sensing and many other fields. However, solving the three-dimensional (3D) structure of nano-scale materials at the atomic level is challenging, especially for predicting metastable, out-ofequilibrium systems. Scanning transmission electron microscopy (STEM) provides structural images of materials at atomic resolution, but a single image provides only a two-dimensional (2D) projection of the structure, and three-dimensional tomographic imaging at atomic resolution with single-atom sensitivity remains challenging. Experimentally driven structural refinement approaches typically rely on minimizing the error between forward simulation from atomic models and the experiment data. Such optimizations are difficult with limited data and rely on knowing good initial guesses for the structure. They also typically make no direct use of information about the energy of the potential structures. Purely computational techniques, such as genetic algorithms (GAs) [1], have proven to be extremely effective at predicting the ground state structures of a wide range of complex structures, including clusters, crystals, and grain boundaries. However metastable configurations are often neglected by methods designed to find the global minimum of the system energy.