Connecting Atomic Scale Chemistry and Structure to Relaxor Ferroelectric Properties Using In Situ Scanning Transmission Electron Microscopy

Relaxor ferroelectrics are a class of technologically important functional oxides characterized by diffuse dielectric and piezoelectric properties, distinguishing them from traditional ferroelectrics. This behavior is most commonly attributed to the existence of polar nanoregions — nanoscale domains that are embedded in an otherwise non-polar matrix. Despite it's general acceptance, this “plum pudding” model fails to explain relaxor behavior in a variety of materials, including polymer-based systems. To formulate models of local polar behavior, the structural and chemical complexity in relaxor systems have largely been probed with diffuse scattering using X-rays, neutrons, and electrons. These techniques, however, provide a measure of the average global and local structure across a relatively large volume of the material. Consequently, the origin of relaxor properties continues to be debated although these materials have been studied extensively.<br/><br/>In this talk, I will discuss how aberration corrected scanning transmission electron microscopy (STEM) combined with in situ heating+biasing can be used to directly separate nanoscale structural and chemical inhomogeneities at the atomic scale in relaxor ferroelectric materials, complementing diffraction studies. These techniques are applied to the relaxor ferroelectric system — Pb(Mg_1/3Nb_2/3)O3-PbTiO3 (PMN-PT). Through a simultaneous acquisition images that are sensitive to chemistry (angle annular dark-field STEM) and light elements (integrated differential phase contrast STEM), we directly connect nanoscale chemical order regions, distorted oxygen octahedra, and local polarization as a function of temperature and/or applied biasing. First, we find that contrary to the model of a binary distribution of chemically ordered regions within a disordered matrix, the degree of chemical order smoothly varies within ordered domains and approaches a minimum at anti-phase boundaries. Second, regions of correlated oxygen octahedral titling are found to be anti-correlated with regions of maximal chemical order. Comparing with the projected polarization, we observe that the regions of greatest variation in polarization correspond to the regions of maximum chemical order and maximum octahedral distortion. Based on these results, we show that both structural and chemical inhomogeneities act as a barrier for polarization rotation and thus frustrate long range polar order. By heating the sample and/or applying a bias during atomic resoltuion STEM imaging, we will directly show how changes in the atomic structure can provide insight into the regions where polarization rotation first occurs, and the connection with local structural/chemical inhomogeneity. Finally, I will discuss these results in the context of proposed models of relaxor ferroelectricity.