Research Highlights

My work so far has encompassed utilizing a combination of data-science and physics driven methods to build atomistical models for complex heterogeneous, electric as well as thermal, catalytic systems. Key achievements and foundational research pillars are highlighted below:

a. Data science and machine learning (ML) driven methods to model catalytic surfaces encompassing complex adsorbate configurations: Complex surface chemistry resulting from strong binding and multidentate nature of adsorbate, as well as presence of alloys and defects on the catalyst surface is ubiquitous in heterogeneous catalytic reactions. Algorithms were developed to capture these complexities. Figure 1 shows (a) ML and graph theory-based active learning framework to identify the most stable adsorbate configurations from a large phase space of possible configurations,^[1] (b) most stable atomic configurations identified for OH* on a stepped Pt(221) and NO* on Pt₃Sn(111) surface utilizing framework in Figure 1a, and (c) the phase diagram for NO* on Pt₃Sn(111), as a function of temperature and pressure of NO(g), to understand experimentally relevant configurations at reaction conditions.^[1,2]

Figure 1: Data science and machine learning driven methods to estimate relevant surface adsorbate configuration for complex chemistries. (a) Machine learning based active learning algorithm to effectively navigate the large phase space of possible atomic configurations. (b) Relevant configurations identified for OH* on Pt(221) and NO* on Pt₃Sn(111) surface. (c) Phase diagram predicting the most stable coverage of NO* on Pt₃Sn(111) at experimentally relevant conditions.

b. Designing realistic atomistic models for next generation catalysts: Next generation of catalytic materials encompass complexities arising from the co-existence of multiple phases and components. Capturing these effects is essential to fundamentally understand the working of such materials under realistic conditions. Figure 2 summarizes key results, derived utilizing a physics driven algorithmic approach, for two important cases relevant to beyond Li-ion batteries, such as Li-O₂, ^[3] and Strong Metal Support Interaction (SMSI) based catalysts. ^[4] Shown in Figure 2 are solid-solid interface models constructed for understanding (a) formation of LiO_x species on La₂NiO₄(001) surface and the corresponding experiments to confirm the prediction of different LiO_x phases as a function of applied voltage; (b) formation of WO_xislands on different defected Pt surfaces and their stability as a function of defect density.

Figure 2: Designing realistic solid-solid interface models for next generation of catalytic materials. (a) Theoretically predicted phase diagram of LiO_x phase stability as a function of discharge voltage. In line with theoretical predictions, experiments show selectivity decrease for the Li₂O₂formation at high potentials. (b) Theoretically predicted stable structure for Tungsten Oxide (WO_x) formation on defected Pt(221) edge. In line with experiments, our models predict WO_x predominantly stabilizing on Pt terraces.

c. Understanding electro-catalytic reactions for next-generation nitrogen and carbon-based feedstock: Environmentally important electro-catalytic reactions pertaining to complex carbon and nitrogen-based feedstock, such as nitrate, ethanol amongst others, are poorly understood compared to traditional feedstock derived from H₂ and O₂. Studies combining data-driven and physics-based approaches were utilized to understand the intricacies of such systems. Figure 3 shows studies to understand (a) NO electro-reduction on Pt-Sn alloys, and the importance of Sn and Pt ensembles to alter the selectivity of the reaction; ^[5]and (b) ethanol electro-reduction on Pt(100) surface, ^[6] and the selective destabilization of multi-dentate intermediates, essential for C—C bond breaking, at high potentials due to the presence of oxygenated intermediates. Both the results were corroborated by experimental findings present in literature and provide fundamental insights necessary to identify promising stable and selective catalysts for such chemistries.

Figure 3: Understanding the surface chemistry for complex electrocatalytic reactions of (a) NO and (b) ethanol. (a) Presence of Sn and high-coverages on NO on the surface enable the selectivity of NO reduction to (NH₃)⁺OH on Pt-Sn alloy. (b) Presence of oxygenated intermediates (OH*) at high potentials, selectivity destabilizes multi-dentate intermediates, essential for C—C bond breaking in ethanol electro-oxidation reaction.

References: († equal contribution, * corresponding authors)

1. Ghanekar P.†; Deshpande S.†*; Greeley, J.* Adsorbate chemical environment-based machine learning framework for heterogeneous catalysis. Nature Communications 2022, 13, 5788.

2. Deshpande, S.†; Maxson, T.†; Greeley, J.* Graph Theory Approach to Determine Configurations of Multidentate and High Coverage Adsorbates for Heterogeneous Catalysis. Nature partner journals Computational Materials 2020, 6, 79.

3. Samira, S.†; Deshpande, S.†; Roberts, C. A.; Nacy, A. M.; Kubal, J.; Matesić, K.; Oesterling, O.; Greeley, J.*; Nikolla, E.* Nonprecious Metal Catalysts for Tuning Discharge Product Distribution at Solid–Solid Interfaces of Aprotic Li–O2 Batteries. Chemistry of Materials 2019, 31 (18), 7300–7310.

4. Denny S. †; Deshpande S. †; Lin Z. †; Porter W.; Rykov S.; Batchu S.; Caratzoulas S.; Vlachos D.*; Chen J.* Ring-opening of tetrahydrofurfuryl alcohol over WOx-modified Pt(111) surfaces. (Submitted)

5. Deshpande, S.; Greeley, J.* First-Principles Analysis of Coverage, Ensemble, and Solvation Effects on Selectivity Trends in NO Electroreduction on Pt3Sn Alloys. ACS Catalysis 2020, 10 (16), 9320–9327.

6. Deshpande S.; Greeley, J.* Elucidating Selectivity Determining Elementary Steps for Ethanol Electrooxidation on Pt(100) via Combined Molecular Dynamics and DFT Analysis. (Submitted)