JR Schmidt

EDUCATION

• B.S. 2001, Hope College
• Ph.D. 2006, University of Wisconsin-Madison
• Postdoctoral Associate at Yale University

VIEW PUBLICATIONS

RESEARCH DESCRIPTION

The Schmidt group builds theory and computational tools to explain — and ultimately to predict — how chemistry happens at complex interfaces. The work spans electronic structure theory, statistical mechanics, kinetics, and machine learning, often in close partnership with experiment.

Two questions organize the group’s current research. First, how do molecules actually react at electrified, solvated surfaces — and can we model those interfaces accurately enough to guide catalyst and electrolyte design? Second, how does ordered matter emerge from disordered precursors, whether a metal–organic framework nucleating from solution or an amorphous oxide crystallizing against a substrate?

Computational electrocatalysis for sustainable chemistry

Electrification offers a route to fuels and commodity chemicals without fossil carbon, but it demands atomic-level control over which bond reacts and how. The group uses density functional theory and first-principles microkinetic modeling to dissect these selectivity problems.

In the reductive upgrading of lignocellulosic biomass, the central difficulty is that hydrogenation of a C=O bond (giving an alcohol) competes with the hydrogenolysis that actually removes oxygen and raises energy density. Our collaborative work traced this competition to two distinct mechanisms — proton-coupled electron transfer versus hydrogen atom transfer from adsorbed H* — and showed how it can be steered by the nature of the electrode/electrocatalyst, by pH, or even by solution additives.

A parallel effort targets the two-electron oxygen reduction reaction, which yields hydrogen peroxide rather than water. Combined theory and experiment identified CoS₂, CoSe₂, NiSe₂, and layered pentagonal PdSe₂ as active, selective, and stable catalysts under realistic operating conditions, and microkinetic models now rationalize overall H₂O₂ selectivity across the metal dichalcogenide family.

More generally, accurate electrocatalysis modeling is limited less by the catalyst than by the interface. Continuum solvation models, ubiquitous in computational catalysis, turn out to fail qualitatively for organic adsorbates at metal/water interfaces — getting both adsorption free energies and conformational preferences wrong — because they omit the competition between adsorbate and water for surface sites. We are thus also pursuing new, computationally efficient approaches to interfacial solvation that go beyond conventional continuum models.

Nucleation and crystallization of metal–organic frameworks (MOFs)

Predicting which structure a material adopts — and how to make it — remains far harder than computing the properties of a known structure. This is particularly true of nanoporous metal–organic framework (MOF) materials, whose porosity makes them attractive targets for applications that include gas storage, separations, and catalysis. Synthesis discovery and scale-up in this area remain largely empirical, and the group has mapped what is actually known about MOF nucleation and growth while developing Monte Carlo, graph-theoretic, and kinetic methods — including hybrid GCMC/MD approaches for weak electrolytes — to connect synthetic parameters such as temperature, solvent, and supersaturation to crystallization outcome and polymorph selection.

We also work extensively to characterize spontaneous and water-induced point defects in MOFs, developing and applying novel QM/MM and QM/QM methods. A major focus has been zeolitic imidazolate frameworks (ZIFs), where our collaborative work explained the striking acid-gas stability of ZIF-71 as a kinetic effect arising from how SO₂ distributes relative to reactive sites.

First-principles force fields

Underlying much of this work is a long-running interest in force fields derived entirely from electronic structure theory. Using symmetry-adapted perturbation theory, the group developed transferable, physically motivated force fields — each term mapped to a distinct physical contribution — for CO₂, N₂, ionic liquids, polymers, and MOFs, along with improved functional forms for short-range repulsion and atomic-level anisotropy that go beyond the Born–Mayer and isotropic assumptions of standard force fields.

Machine learning and chemical AI

Most recently, the group has begun applying deep learning where first-principles methods are too expensive or insufficiently accurate. Graph neural network models built to reflect the ligand structure of heteroleptic Ir(III) complexes predict full phosphorescence emission spectra more accurately than DFT or correlated wavefunction methods, are robust to noisy experimental training data, and support design-of-experiment strategies that reduce the synthetic burden of high-throughput screening.

Software for computational chemistry education

Schmidt co-develops WebMO, a browser-based interface for computational chemistry used by students and researchers worldwide. Now past its 25th year, WebMO makes electronic structure calculations accessible without command-line expertise and has become a standard tool in undergraduate and graduate curricula.