Jaehyun Kim
Research

High-flux electrochemical interfaces

I connect atomistic computation to industrial-relevant devices — machine-learning catalyst design, synthesis, operando mechanism, and membrane-electrode-assembly validation, run as one closed loop.

Focus catalyst–ionomer–membrane interfaces Approach computation + experiment Group SNU · Prof. Jungwon Park
At a glance
13.1 A cm⁻²
noble-metal-free AEM electrolysis @ 2.0 V (flagship)
3,976
single-atom configurations ML-screened
~1,000×
faster screening than DFT (EquiformerV2)
6
reaction systems — water splitting, CO₂, NH₃ oxidation & more
The closed loop

One workflow, prediction to device

01 ML screening→ 02 DFT · MLIP · MD→ 03 Synthesis→ 04 Operando→ 05 MEA device

Machine-learned potentials and DFT rank thousands of candidates; the winners are electrodeposited, watched under operando synchrotron and Raman, and validated in zero-gap membrane-electrode assemblies — atoms to industrial current density, one loop.

Directions
Flagship study · Nature Communications 2026

ML-guided tungsten single atoms for noble-metal-free water electrolysis

The clearest example of how I work: I fine-tuned EquiformerV2 on custom DFT data, screened the full configuration space, identified W₁-NiFeOOH, synthesized it by scalable cyclic electrodeposition, resolved its mechanism with operando spectroscopy, and validated it in an AEM electrolyzer — prediction, mechanism, and device connected in one loop.

13.1 A cm⁻²
AEM electrolysis at 2.0 V, practical conditions
3,976
single-atom oxyhydroxide configs screened
W₁-NiFeOOH
noble-metal-free catalyst identified
1st
first author · Nat. Commun. 17, 2344
Machine learning-guided catalyst screening (Nature Communications 2026, Fig. 1)
Fig. 1 — ML-guided catalyst screening: EquiformerV2 + DFT, OER activity volcano, and raw-material cost analysis. Nature Communications 17, 2344 (2026).
Beyond water splitting

The same theory-to-device approach carries across CO₂ electroreduction, glycerol and oxygenated-feed upgrading, acidic enol electrooxidation, and electrocatalytic ammonia oxidation. In my current postdoctoral work I am developing a multiscale computational framework — coupling DFT, machine-learned interatomic potentials, and molecular dynamics with explicit solvent — to model ion transport within anion-exchange membranes and ionomer layers, validated through MEA fabrication and testing.

Next: integrating direct CO₂ capture with electrochemical conversion — treating capture and conversion as one system rather than two separate processes.