Materials Development Workstation

Materials design tool

The combinatorial experimentation and materials informatics approaches outlined here are only elements of a problem-solving strategy for the materials engineer.  At least as important, the materials engineer must have capability for controlling how the elements are exercised and how a learning pathway to a reasonable solution is pursued.  One can envision a materials design tool or environment that presents such power to the materials engineer, allowing choices and exploration of different elements along the way.  For example, one might want the engineer to be able to choose which materials synthesis process is used, which characterization methods are necessary, what protocols of annotation are employed, what priority is given to different legacy data sources, which approach is taken to model development, whether physics-based models should be included, etc.

Furthermore, the materials engineer should have primary and direct control over three other important functions – visualization and analysis, multivariate optimization, and combinatorial design-of-experiments generation.

Visualization

Visualization of combinatorial data, modeling relationships, and database structure and annotation is essential for understanding and exploiting this new paradigm, since combinatorial experimentation promises major increases in data volumes, and informatics provides multiple tools with which to explore and infer from the data.  In addition, the visualization environment must also be a management environment that not only enables the user to review and data, but also to choose how to manipulate and analyze it.  In short, visualization must be an active control as well as a powerful observational environment.

The multivariate complexity of materials science and engineering places a special premium on visualization.  Complex materials systems are multicomponent, involve a number of key synthesis/processing parameters, form multiple forms of crystal structure and microstructure, and are chosen for the performance of multiple parameters.  Keeping track of this complexity requires powerful visualization and management tools that are easily used in a software environment.

Multivariate optimization

Perhaps most important for materials engineer, and an essential feature of a materials development workplace, is the ability to tailor the choice of materials systems and processes to achieve the desired materials properties.  This might be a simple task, if only one performance metric for the material and few independent design parameters were available, but precisely the reverse is the norm for materials engineering, as described in the discussion of multivariate complexity above. The challenges here are several.  For one, a value or utility function must be defined that sensibly weights the various performance metrics of interest (e.g., electrical resistivity vs. electromigration resistance, dielectric constant vs. switching lifetime, or magnetic permeability vs. hysteresis).  And in general this involves more than two metrics, some of which may be linear or nonlinear, and others which may  be highly nonlinear or even discontinuous.

Ultimately, the challenge of multivariate optimization benefits from more complete knowledge of materials system behavior and performance as afforded by combinatorial experimentation,  but it places a major responsibility on the informatics side of this paradigm, and on developing powerful, user-friendly tools to carry out complex exercises of analysis and optimization.

Combinatorial design-of-experiments generation

Optimizing materials and the corresponding processes to make them is typically an iterative exercise, and even with the efficiency obtained through combinatorial experiments one would like to minimize the cost and time associated with experimentation.  To that end, efficient experiments are needed, for which the traditional solution is to plan those experiments within the context of structured design-of-experiments (DOE).

Within an integrated framework of combinatorial experimentation and materials informatics, however, new modes for experimental design can be envisioned.  Since combinatorial libraries produce intentional gradients of independent parameters, the database analyses and models developed from this data using informatics techniques provide trend information, i.e. how various parameters vary as a function of other parameters, or more simply a library of relevant partial derivatives.  With the large array size of combinatorial libraries from a single experiment, such dependencies are known far better than would be the case for traditional experimental methods and DOE’s. Furthermore, the modeling facilitated by the availability of large datasets and parametric trends, along with informatics approaches, represents another avenue by which significant knowledge may be extracted. 

In principle, then, the expanded databases of combinatorial experimentation and the materials informatics strategies applied to extract value from these databases promises a rapid development of new knowledge.  Since iterative sequences of experiments are expected in general, there is an opportunity to design experiments which will more efficiently optimize materials and extract fundamental understanding about them.  The arrival of combinatorial-informatics strategies for materials discovery and development offers a major incentive to construct and demonstrate new methodologies for design-of-experiments that truly take advantage of  both combinatorial experimentation and materials informatics.