Optimization-Based Material Discovery

Identification of new materials with desired properties is the holy grail in materials science; recent thrusts in this area, such as the Materials Genome Initiative, aim to speed up the process of discovering materials customized for a specific application. In the context of heterogeneous catalysis, this problem is one of identifying a material that can be used as a catalyst to maximize rate, yield, or selectivity of a desired product in a reaction system. Such problems are computationally and experimentally challenging due to the large space of potential compounds that need to be explored.

The first step in materials discovery is to determine the target material characteristics that will optimize its performance objective. For catalysis, this means identifying the values describing kinetics thermochemistry in the reaction system for which the overall predicted yield/rate/selectivity (or any other user-defined performance criteria) is optimized. Once the targets are identified, computational chemistry (specifically density functional theory, or DFT) calculations and/or experimental analyes can be used to identify the materials with these target characterisitcs.

We formulate this step as a nonlinear optimization problem subject to physico-chemical, thermodynamic, and material constraints. Such problems tend to involve stiff systems of equations and the number of decision variables increase fast with the size of the reaction network or the complexity of the catalyst. We leverage recent advances in global optimization and develop novel reformulation strategies to solve such problems quickly and robustly. In addition, we are also developing methodologies to identify a few “key” characteristic parameters that determine the performance of the material, thereby reducing the dimensionality of the problem.

Our group is collaborating with experts in the area of heterogeneous catalysis to develop a three pronged approach comprising of: (i) density functional theory (DFT) calculations to obtain microscopic properties of the reaction system, (ii) experimental studies to test/measure the overall performance of the catalyst, and (iii) optimization to link microscopic parameters to optimal values of macroscopic observables (Figure). New software tools are being developed for catalyst discovery and we are currently pursuing applications in the domain of CO2 upgrading and methane processing.

Workflow diagram

A workflow diagram for the proposed integrated computational-experimental approach for identifying new catalysts.

References:

Rangarajan S, Maravelias CT, Mavrikakis M. Sequential Optimization-Based Framework for Robust Modeling and Design of Heterogeneous Catalytic Systems. Journal of Physical Chemistry C, 121, 25847-25863, 2017.

Nason T, Grabow L, Mavrikakis M, Biegler L, Maravelias CT. Advanced Solution Methods for Microkinetic Models of Catalytic Reactions: a Methanol Synthesis Case Study. AIChE J., 60(4), 1336-1346, 2014.