(689g) On the Bottom-up Design of Bimetallic Catalytic Nanoparticles with Atomic Level Resolution

Choksi, T. S., Stanford University
Streibel, V., Stanford University
Roling, L., Stanford University
Abild-Pedersen, F., SLAC National Accelerator Laboratory
The advent of increasingly sophisticated physics-based [1]-[5] and data-driven workflows [6]-[8] has transformed designing bimetallic alloy catalysts from an Edisonian endeavor into one guided by descriptor-based design principles. Recent advances notwithstanding, prevailing catalyst design paradigms remain limited by two key features. First, these paradigms use idealized models based on extended surfaces to describe nanoparticles encompassing diverse shapes, compositions, and sizes in their working state. These assumptions create a materials gap between model structures and catalytic architectures under reaction conditions. Second, it remains a formidable endeavor to precisely engineer morphological and compositional features of active site motifs while ensuring the stability of these motifs under working conditions. We address these knowledge gaps by introducing a generalized property ⇒ structure workflow[2]-[5] that screens binding energies of catalytic descriptors with atomic level resolution while concurrently determining stabilities of active site motifs on-the-fly. Site stabilities (BEsite) are predicted using our coordination-based alloy stability model[2]-[3] while binding strengths of catalytic descriptors (BEads) are determined through a new-class of site-specific scaling relations[4]-[5]. Our workflow which uses a remarkably limited parameter space, nonetheless, predicts BEads and BEsite with accuracies of 0.10-0.15 eV for bimetallic nanoparticles of wide-ranging shapes (cubo-octahedral, decahedral, icosahedral, and octahedral), sizes (larger than 1.6 nm), compositions (binary pairings of 30 transition and post-transition metals), bulk crystal structures (fcc, bcc, hcp, and tetragonal), and under applied strain (± 3%).

Our coordination-based alloy stability model is constructed by partitioning energies of metal atoms in terms of bond-associated parameters (αiZ (r)), which represent contributions of each bond formed.[2]-[3] For fcc, hcp, and tetragonal crystal systems, we only include the first coordination shell, while for bcc crystal systems, both the first and second coordination shells are included. Compositional variations are treated through a mean-field approximation. This deconstruction of active sites in alloys is inspired by the “near-sightedness” of itinerant d-electrons, which efficiently screens structural and compositional perturbations beyond the first coordination shell of an adsorption site. Finite-size effects prevalent in nanoparticles are accounted for using readily transferable quadratic functional forms for αiZ(r) parameters (MAE of 0.03 eV in the training set). Despite fitting αiZ(r) parameters to a limited set of energies derived from calculations on simple slabs, αiZ(r) parameters successfully predict both BEsite and relative energies of single atom swaps for nanoparticles having wide-ranging shapes, sizes, and compositions with errors generally within 0.15 eV.[3]

In addition to reflecting the thermodynamic stability of an adsorption site, BEsite also functions as a powerful descriptor for predicting BEads (CO*, OH*, CH3*, CH*, NO*) through a new class of site-specific scaling relations.[4]-[5] These scaling relations correlate BEsite to adsorption strengths of metal adsorbate complexes (BEsite-ads). We strategically construct a family of scaling lines such that each correlation within this family corresponds to a given site-identity (e.g. Au) and site-coordination (e.g. coordination number of 9). Points along a correlation represent variations in chemical environment, applied strain, and bulk crystal structures. Slopes and intercepts of site-specific scaling are discussed in the context of the d-band theory. By unifying diverse structural and compositional features of nanoparticles across a single continuous descriptor space (BEsite), we expeditiously predict adsorption properties of catalytic descriptors on bimetallic nanoalloys with atomic level specificity within MAEs of 0.15 eV.[5] These site-specific scaling relations further reveal that diverse structural and compositional features of nanoalloys can be compressed into a single robust descriptor for adsorption, reinforcing the local nature of metal-adsorbate bonds. Finally, by propagating site-specific scaling relations through a microkinetic model for NO reduction, we interrogate the effect of structural and compositional features on catalytic turnovers with site-specific precision. We envision that our property ⇒ structure workflow opens up new avenues to design the next generation of computationally engineered bimetallic alloy catalysts that are stable, active, and selective.


[1] Saravanan K., Kitchin J. K., von Lillenfeld O. A., Keith J. A., J. Phys. Chem. Lett. 2017, 8, 5002.

[2] Roling L. T., Lin L., Abild-Pedersen, F., J. Phys. Chem. C, 2017, 121, 23002.

[3] Roling L. T., Choksi T. S., Abild-Pedersen, F., Nanoscale, 2019, 11, 4438.

[4] Roling L. T., Abild-Pedersen, F., ChemCatChem. 2018, 10, 1643.

[5] Choksi T. S., Roling L. T., Streibel V., Abild-Pedersen, F., J. Phys. Chem. Lett, 2019, 10, 1852.

[6] Ma X., Li Z., Achenie L. E. K., Xin H., J. Phys. Chem. Lett, 2015, 6, 3528.

[7] Jinnouchi R., Asahi R., J. Phys. Chem. Lett, 2017, 8, 4279.

[8] O’Connor N. J., Jonayat A. S. M., Janik M. J., Senftle T. P., Nat. Catal. 2018, 1, 531.