Student: Dinushika Kotudura

Topic:
Atomic-Level Engineering of MXenes Enables Selective CO₂-to-Methane Conversion

Abstract:
Two-dimensional MXenes have emerged as promising catalyst platforms due to their tunable surface chemistry, high electrical conductivity, and structural versatility. These exceptional material properties make MXenes attractive supports for atomic-level catalyst design to address selectivity challenges in electrochemical CO₂ reduction. In particular, achieving selective methane formation while suppressing the competing hydrogen evolution reaction remains difficult with conventional catalysts. In this talk, I will present how material-level engineering of Mo₂CTx MXenes through the incorporation of Cu single atoms enables selective CO₂ reduction to methane at low applied potentials. The MXene surface provides a conductive and chemically tunable platform that stabilizes atomically dispersed Cu sites and modifies the local electronic structure. This electronic modulation strengthens CO adsorption and stabilizes reaction intermediates, promoting stepwise hydrogenation pathways that favor methane formation while suppressing hydrogen evolution.

Student: Akash Surendran

Topic:
My adventures through chemical space (so far)

Abstract:
As chemical libraries expand into the realm of millions to billions of compounds, navigating these vast “galaxies” of data has become a critical bottleneck in drug discovery. In this talk, I will introduce a suite of unsupervised-learning-based cheminformatics tools designed to address this challenge at multiple scales. I will begin by addressing the hyperparameter sensitivities of traditional nonlinear dimensionality reduction (NLDR) tools and optimizing their performance for small-library visualization. To bridge the gap to larger datasets, I will present NAMI (N-ary Mapping Interface), a linear, hyperparameter-free framework that enables the faithful visualization of million-sized libraries. Finally, to move beyond just pure visualization, I introduce a framework that leverages hierarchical, similarity-based clustering to calculate local activity landscapes. This approach enables the automated detection of activity cliffs, regions where subtle structural modifications trigger dramatic potency shifts, as well as the identification of “smooth clusters” for reliable SAR modeling. Together, these tools provide a scalable, intuitive framework for navigating the expanse of large chemical libraries. By focusing on the structural-activity landscape, this workflow facilitates superior data curation for training the next generation of chemistry-aware property prediction models.

University of Florida Calendar of Events

Insights into GatorEvals

Askew and Haskell Scholar Info Session

The H-1B Insider Guide for Professionals (Lunch & Learn)