Computational materials scientist
GiyeokLee
Postdoctoral Research Associate · School of Physics, The University of Sydney
I combine first-principles surface modelling, uncertainty-aware machine-learning interatomic potentials, electronic-structure calculations, and molecular dynamics to connect atomic structure with material properties.
NARA MOVE
xᵢ′ = xᵢ + β₀e−γr²(xⱼ−xᵢ) + αε
LLUMYS LCB
ELCB = EMLIP − κΣᵢuᵢ, uᵢ = max(asl̂ᵢ+bs, 0)
GEN000
EMLIP--
Σuᵢ--
ELCB--
Active-learning cycle
32 DFT LABELS / MLIP v01
01PredictEMLIP + atom-wise uᵢ
02AcquireELCB < min tracker
03Freeze + DFT0 / 3 in batch
04Retrainmerge + recalibrate uᵢ
MLIP MD LOCKED
32 / 38 DFT LABELS
AVAILABLE AFTER TWO ACTIVE-LEARNING BATCHES
MLIP deployment · precomputed MD
CTP / NVT TRAJECTORY
Au₁₁ / 300 K / LANGEVIN
WAITING FOR DFT LABELS
T -- K / SET 300
EPOT -- eV
t -- ps
FRAME 000 / 300
02 / RESEARCH
Research across structure, potentials, and properties
Surface structure modelling, uncertainty-aware interatomic potentials, and electronic and finite-temperature property analysis.
aiATSURFACE DFTGLOBAL STRUCTURE SEARCHSIMULATED STM + XPSUNCERTAINTY-AWARE MLIPDOS + WORK FUNCTIONMOLECULAR DYNAMICS
aiATSURFACE DFTGLOBAL STRUCTURE SEARCHSIMULATED STM + XPSUNCERTAINTY-AWARE MLIPDOS + WORK FUNCTIONMOLECULAR DYNAMICS
04 / SOFTWARE
Software
Open-source packages for atomic structure search, machine-learning potentials, STM analysis, and structure visualization.
05 / CONTACT
Contact
For research, software, or collaboration enquiries, email is the most direct route.