Suman Bhasker Ranganath
Postdoctoral Scholar, Photon Science, SLAC
Stanford Advisors
-
Thomas Jaramillo, Postdoctoral Faculty Sponsor
-
Johannes Voss, Postdoctoral Research Mentor
Current Research and Scholarly Interests
Development of machine-learning models from high-throughput catalysis simulations.
All Publications
-
Optimizing Prediction of Chemical Bonds in Interfacial Dynamics through Local Uncertainty Estimates with Neural Network Ensembles.
Journal of chemical information and modeling
2026
Abstract
We present a framework for data-efficient training of machine-learning interatomic potentials for interfacial chemistry, especially heterogeneous catalytic systems. We establish strategies for density functional theory training data generation consisting of procedurally generated bulk, surface, and gas-phase atomic geometries, as well as moderately randomized structures. We show how ensembles of neural network machine-learning interatomic potentials trained on different splits of these training structures yield reliable uncertainty estimates at the atomic node energy level. Our models can thus identify which atomic sites and chemical bonds in a system lead to uncertainties in the predicted potential energy surface. Using hydrogen interacting with platinum as a test case, we find that the atomic uncertainty estimates identify both unphysical bonding scenarios and physically relevant interactions that are underrepresented in the original training data, such as surface diffusion, bond breaking, and bond formation. Building on these insights, we propose local uncertainty-informed strategies that flag outliers via statistical correlations, thereby improving active learning efficiency and enhancing the reliability of neural network-based potentials for extended-scale reactive dynamics.
View details for DOI 10.1021/acs.jcim.5c02083
View details for PubMedID 41591878
-
A research database for experimental electrocatalysis: Advancing data sharing and reusability.
The Journal of chemical physics
2025; 163 (12)
Abstract
The availability of high-fidelity catalysis data is essential for training machine learning models to advance catalyst discovery. Furthermore, the sharing of data is crucial to ensure the comparability of scientific results. In electrocatalysis, where complex experimental conditions and measurement uncertainties pose unique challenges, structured data collection and sharing are critical to improving reproducibility and enabling robust model development. Addressing these challenges requires standardized approaches to data collection, metadata inclusion, and accessibility. To support this effort, we have developed an extensive data infrastructure that curates and organizes multimodal data from electrocatalysis experiments, making them openly available through the catalysis-hub.org platform. Our datasets, comprising 241 experimental entries, provide detailed information on reaction conditions, material properties, and performance metrics, ensuring transparency and interoperability. By structuring electrocatalysis data in web-based as well as machine-readable formats, we aim to bridge the gap between experimental and computational research, allowing for improved benchmarking and predictive modeling. This work highlights the importance of well-structured, accessible data in overcoming reproducibility challenges and advancing machine learning applications in catalysis. The framework we present lays the foundation for future data-driven research in electrocatalysis and offers a scalable model for other experimental disciplines.
View details for DOI 10.1063/5.0280821
View details for PubMedID 41025581
-
Hydrolysis of Acetamide on Low-Index CeO2 Surfaces: Ceria as a Deamidation and General De-esterification Catalyst
ACS CATALYSIS
2022; 12 (16): 10222-10234
Abstract
Using DFT calculations and acetamide as the main example, we show that ceria is a potential catalyst for the hydrolysis of amide and similar bonds. The overall reaction is endergonic in the gas phase, yielding acetic acid and ammonia, but is slightly exergonic in the aqueous phase, which facilitates ionization of the products (CH3COO- and NH4 +). Neighboring Ce and O sites on the CeO2(111), (110), and (100) facets are conducive to the formation of an activated metastable tetrahedral intermediate (TI) complex, followed by C-N bond scission. With van der Waals and solvation effects taken into account, the overall reaction energetics is found to be most favorable on the (111) facet as desorption of acetic acid is much more uphill energetically on (110) and (100). We further suggest that the Ce-O-Ce sites on ceria surfaces can activate X(=Y)-Z type bonds in amides, amidines, and carboxylate and phosphate esters, among many others that we term "generalized esters". A Brønsted-Evans-Polanyi relationship is identified correlating the stability of the transition and final states of the X-Z generalized ester bond scission. A simple descriptor (ΣΔχ) based on the electronegativity of the atoms that constitute the bond (X, Y, Z) versus those of the catalytic site (O, Ce, Ce) captures the trend in the stability of the transition state of generalized ester bond scission and suggests a direction for modifying ceria for targeting specific organic substrates.
View details for DOI 10.1021/acscatal.2c02514
View details for Web of Science ID 000844147600001
View details for PubMedID 36033367
View details for PubMedCentralID PMC9397537
-
Elucidating the Mechanism of Ambient-Temperature Aldol Condensation of Acetaldehyde on Ceria
ACS CATALYSIS
2021; 11 (14): 8621-8634
Abstract
Using in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) and density functional theory (DFT) calculations, we conclusively demonstrate that acetaldehyde (AcH) undergoes aldol condensation when flown over ceria octahedral nanoparticles, and the reaction is desorption-limited at ambient temperature. trans-Crotonaldehyde (CrH) is the predominant product whose coverage builds up on the catalyst with time on stream. The proposed mechanism on CeO2(111) proceeds via AcH enolization (i.e., α C-H bond scission), C-C coupling, and further enolization and dehydroxylation of the aldol adduct, 3-hydroxybutanal, to yield trans-CrH. The mechanism with its DFT-calculated parameters is consistent with reactivity at ambient temperature and with the kinetic behavior of the aldol condensation of AcH reported on other oxides. The slightly less stable cis-CrH can be produced by the same mechanism depending on how the enolate and AcH are positioned with respect to each other in C-C coupling. All vibrational modes in DRIFTS are identified with AcH or trans-CrH, except for a feature at 1620 cm-1 that is more intense relative to the other bands on the partially reduced ceria sample than on the oxidized sample. It is identified to be the C=C stretch mode of CH3CHOHCHCHO adsorbed on an oxygen vacancy. It constitutes a deep energy minimum, rendering oxygen vacancies an inactive site for CrH formation under given conditions.
View details for DOI 10.1021/acscatal.1c01216
View details for Web of Science ID 000674927200023
View details for PubMedID 34306815
View details for PubMedCentralID PMC8294007
-
Theoretical analysis of the adsorption of phosphoric acid and model phosphate monoesters on CeO2 (111)
SURFACE SCIENCE
2021; 705
View details for DOI 10.1016/j.susc.2020.121776
View details for Web of Science ID 000618759800004
-
Adsorption structure of adenine on cerium oxide
APPLIED SURFACE SCIENCE
2020; 530
View details for DOI 10.1016/j.apsusc.2020.147257
View details for Web of Science ID 000562341900003
-
Computational insights into the molecular mechanisms for chromium passivation of stainless-steel surfaces
MATERIALS TODAY CHEMISTRY
2020; 17
View details for DOI 10.1016/j.mtchem.2020.100298
View details for Web of Science ID 000572167800012
-
Role of Metal-Lithium Oxide Interfaces in the Extra Lithium Capacity of Metal Oxide Lithium-Ion Battery Anode Materials
JOURNAL OF THE ELECTROCHEMICAL SOCIETY
2016; 163 (10): A2172-A2178
View details for DOI 10.1149/2.0281610jes
View details for Web of Science ID 000389150900007