"New ways to determine and describe ferroelectric structures from diffraction and scattering"
The importance of structure-property relationships in materials research cannot be overstated. At the crystallographic level, the structure relates directly to nearly all physical properties ranging from thermal expansion and elastic compliance to dielectric polarization, piezoelectricity, and pyroelectricity. There is great complexity in the ferroelectric crystal structures investigated today with examples including disordered atomic displacements, chemical disorder, octahedral tilting, and incommensurate modulations. A growing emphasis on in situ characterization, e.g. crystallographic study during application of electric fields or mechanical stress, further increases the challenges changes in determining and describing the structural features of interest. This talk will first demonstrate the use of advanced in situ X-ray and neutron scattering methods (including diffraction and pair distribution functions from total scattering) to discern the underlying mechanics and physics at play in electro-active materials such as dielectrics and piezoelectrics. The interesting local structure changes in a variety of compounds and solid solutions will be described. I will also introduce the development of in situ PDF for determining how local structures of ferroelectrics respond to an electric field, providing critical insight to understanding of dynamic structure-property relationships. The second part of this talk will include an introduction to an alternative statistical framework for analysis of diffraction data, that of Bayesian statistics in conjunction with a Markov Chain Monte Carlo (MCMC) algorithm. This analysis approach is applied to modeling doublets from ferroelastic degenerate reflections and quantifying the extent of domain wall motion in ferroelectrics on a probability basis. We have also applied these approaches to full-pattern profile fitting. The parameters in the new models represent structure using probability distributions, treating solutions probabilistically with improved uncertainty quantification.