AN INFORMATICS APPROACH TO INTERFACE CHARACTERIZATION: Establishing a 'Materials by Design' Paradigm

摘要

It is of course well established that interfaces play a major role in controlling the properties of materials. The relationships between specific interface chemistries and/or structure and final properties is generally studied (experimentally and theoretically) from material to material. Yet there is a need to develop a more generalized approach which permits one to explore vast arrays of data on structure-property relationships based on interface characteristics. Such an approach can permit one search for correlations between structure and property across vastly different length scales. This type of "data mining" or informatics approach is a well established tool in the chemical and biochemical sciences in searching for structure-property relationships in a large combinatorial designs of molecular chemistries. In this paper we describe some of mathematical tools in analyzing the statistics of grain boundary crystallography from grain specific measurements. The issues governing data clustering and pattern recognition are common to both the experimental characterization of grain boundary crystallography as well as the computational aspects of characterizing preferred orientations. It is pointed out that the cohesive use of statistical methodologies both in the experimental as well as the computational analysis of texture data has essentially taken an informatics approach that is being used in the biological crystallography community , although the physical basis of the interpretation is of course different. In this paper the collective use of Hough transforms, quarternion projections and fuzzy clustering techniques when collectively used , all serve to gather and assess large quantities of data in an efficient , rapid yet robust manner. The paper concludes by suggesting how these techniques can be integrated to data sets on a vast library of materials properties along with other descriptors of interfaces to develop a data mining strategy for seeking structure-property correlations across length and time scales. This in turn can help to establish a new "Materials by Design" paradigm for materials research.