430759 Towards Accurate and Fast Discovery of Compound Materials As Catalysts: Lessons Learned from Oxides

Sunday, November 8, 2015
Exhibit Hall 1 (Salt Palace Convention Center)
Zhongnan Xu, Chemical Engineering, Carnegie Mellon University, Pittsburgh, PA

Non-metal, compound catalytic materials are used in mature, well-established industrial applications and but also lie at the forefront of state of the art technologies. These materials are typically discovered through costly trial and error. Density functional theory (DFT) calculations can elucidate key mechanistic details of catalytic processes and structure-property relationships of both known and potential catalytic materials. For metal catalysts, work using these principles have produced a number of successful examples of high-throughput discovery of novel, inexpensive, and active materials. Unfortunately, our ability and understanding on non-metal compound catalysts pales in comparison. Transition metal oxides (TMOs) constitute a large family of compound, catalytic materials. In contrast to metals, oxides can exist in multiple oxidation states and crystal structures, hindering our ability to make comprehensive models for predicting their activity. Standard implementations of DFT also fail to capture the behavior of correlated d-electrons present in TMOs. These failures limit our ability to capture accurate energetic data, which is required for a kinetic and thermodynamic analysis of catalytic processes.

Currently in my PhD with my advisor Prof. John Kitchin, we have addressed both of these difficulties. The hallmark method of our work is the coupling of structural perturbation calculations and identified structure motifs across phase space. We first demonstrate how these calculations can be used to discover subtle electronic structure relationships that describe catalytic effect of changing composition and crystal structure. Using well known structure-property relationships of metals as a foundation, we have gained insight to the critical features of the electronic structure that are responsible for adsorption across the transition metal series of oxides.1 Similarly, we identify the strain and ligand effect present in mixed metal oxides, using structural perturbations to relate calculated electronic and adsorption features to known chemical properties.This work establishes one of the first models that can predict the activity of mixed metal oxides from known chemical data. Our most recent work explores the potential activity of polymorph oxides, where we realize the unique chemical properties offered by polymorphic structures and relate these properties to the both the crystal and electronic structure.

Secondly, we use structural perturbations coupled with a self-consistently calculated Hubbard U to construct a DFT+U(V) method for more accurate predictions of relative stability of oxide materials.3 We find that our method is both computationally faster and more accurate than the current state-of-the-art electronic structure methods. We also show how using the linear response U leads to more accurate predictions of the activity of late TMOs for the oxygen evolution reaction (OER), which is the current bottleneck towards the electrochemical production of the alternative fuel H2.4 We are currently extending our method to other transition metal compounds (sulfides, nitrides, etc) to ascertain their need for DFT+U.

My future work aims to create predictive models that relate the structure and composition of an oxide to its activity. These models will be validated with a wide variety of known compounds and will take advantage of the large databases of crystal structures already constructed in the materials science community (materialsproject.org, OQPMD, etc). This work will include development of algorithms of the construction of the surface facets as well as atomistic-thermodynamic models to predict the structure and surface segregation of mixed metal oxides. Using structure-property relationships determined in my PhD as a foundation, machine learning techniques will be applied to pinpoint chemical properties key for determining their activity/stability. Combined with the aforementioned established materials databases and structure-property relationship insight, high-throughput evaluations of catalytic activity can be realized. Finally, the methods and ideas established in my PhD will also be applied to other compounds such as nitrides, phosphides, carbides, etc.

Relevant Publications

(1) Xu, Z., & Kitchin, J. R. (2014). Catal. Commun., 52, 60–64.

(2) Xu, Z., & Kitchin, J. R. (2015). J. Chem. Phys., 142(10), 104703

(3) Xu, Z., Joshi, Y. V, Raman, S., & Kitchin, J. R. (2015). J. Chem. Phys., 142(14), 144701

(4) Xu, Z., Rossmeisl, J., & Kitchin, J. R. (2015). J. Phys. Chem. C., 119(9), 4827–4833

Extended Abstract: File Not Uploaded