Computational Materials Scientist Heterogeneous Catalysis | Atomistic Modeling | Physics-Informed ML
I build computational frameworks that translate electronic-structure calculations into predictive catalytic performance.
📍 Portland, OR
📧 ankit05puri@gmail.com
🔗 LinkedIn
🔗 [Portfolio] (predict-catalysis-lab.lovable.app)
DFT Surface Modeling
↓
Adsorption Energetics
↓
Microkinetic Simulation
↓
Catalytic Performance Maps
↓
Physics-Informed ML SurrogatePt(111) Surface and Adsorption Energetics
• Constructed periodic Pt(111) slab models using ASE with vacuum separation and constrained bulk layers
• Performed surface and adsorption geometry optimizations using GPAW (PBE, plane-wave basis)
• Investigated adsorption configurations for CO* and O* across surface sites
• Computed adsorption energetics to identify preferred binding geometries and site stability
• Implemented reproducible Python workflows for surface construction, adsorbate placement, and geometry relaxation
• Prepared co-adsorption configurations and reaction pathway setup for CO oxidation studies
🔗 Repository: cat-adsorption-dft
Developed a physics-based microkinetic framework translating surface reaction energetics into catalytic performance.
• Implemented a mean-field microkinetic model for heterogeneous CO oxidation on Pt(111)
• Constructed a reaction network including CO adsorption, O₂ dissociation, surface reaction, and CO₂ desorption
• Solved stiff ODE systems (BDF integration) to simulate surface coverage evolution and steady-state catalytic flux
• Generated TOF maps across temperature and CO partial pressure revealing oxygen-activated, balanced, and CO-poisoned regimes
• Extracted apparent activation energies from Arrhenius analysis of steady-state catalytic rates
• Performed barrier perturbation and degree-of-rate-control (DRC) analysis to identify regime-dependent rate-controlling steps
• Visualized kinetic regime maps showing how catalytic behavior shifts across operating conditions
🔗 Repository: cat-microkinetics
Developed a machine learning surrogate model to accelerate evaluation of catalytic performance.
• Generated structured datasets from microkinetic simulations across temperature and CO partial pressure
• Engineered physics-informed features including inverse temperature (1/T) and log partial pressure
• Trained regression models to predict steady-state turnover frequency (TOF) in log space
• Achieved high predictive accuracy within trained kinetic regimes
• Evaluated generalization under CO pressure holdout conditions
• Demonstrated how surrogate models approximate learned kinetic manifolds but degrade across regime transitions
🔗 Repository: cat-ml-surrogate
• Electronic Structure & Surface Science: Periodic DFT (GPAW), surface slab modeling, adsorption energetics, surface site analysis
• Catalysis & Reaction Modeling: Microkinetic modeling, Arrhenius kinetics, surface coverage dynamics, degree-of-rate-control (DRC)
• Machine Learning for Physical Systems: Physics-informed ML, regression models, surrogate modeling, regime generalization analysis
• Scientific Programming: Python, NumPy, SciPy, Matplotlib, reproducible simulation workflows
Developing scalable computational frameworks that connect electronic structure calculations, microkinetic modeling, and machine learning to enable predictive catalyst discovery.