Research
Research from the SKKU-STEM Lab — methods at the intersection of scanning transmission electron microscopy, electron spectroscopy, and machine learning, applied to energy materials.
“A shuttle guiding us to atomic world.”
Research Themes
Automated defect, particle, and structure analysis
Machine learning and deep learning push past the limits of manual electron-microscopy analysis. From defect quantification in 2D TMDs and Pt nanoparticle morphology (Attention U-Net) to crystallography, tomographic reconstruction, and PEMFC electrode characterization, we automate and quantify the full EM workflow.
Picoscale lattice analysis with hybrid deep learning
From 4D-STEM diffraction patterns we extract picometer-precision lattice strain, crystallographic phase, and domain maps. Applied to perovskite solar cells, HfO2-based ferroelectrics, and 2D transition-metal dichalcogenides, the approach is paired with deep-learning crystallography to resolve structures in sub-10 nm polycrystalline thin films.
Multiscale reconstruction of porous battery & fuel-cell electrodes
Segmented and deep-learning-accelerated electron tomography quantifies the three-dimensional morphology and degradation mechanisms of Li-ion cathodes and PEMFC carbon supports. Sparse-section imaging combined with DL reconstruction reduces data requirements by more than 95%, making routine 3D characterization practical.
STEM-EELS/EDX chemical, valence, and oxygen-vacancy mapping
Atomic-resolution STEM combined with EELS/EDX spectroscopy directly visualizes oxygen vacancies, cation distributions, and valence states in energy oxides — SOFC cathodes, Li-ion electrodes, solid electrolytes, catalysts, and thermoelectrics. By quantifying the defect chemistry that governs functionality at the single-atom level, we trace the microscopic origins of structure–property relationships.
Mechanism of structure-related phenomena
Atomic-resolution electron microscopy reveals the monoatomic step-level flatness and grain-boundary-free epitaxial growth mechanisms of Cu(111) and Ag thin films. We resolve long-standing mysteries of thin-film growth — from self-oxidation resistance on curved copper surfaces to ultraflat growth across a 13% lattice mismatch.
2D materials & van der Waals heterostructures
We use STEM to directly observe atomic defects, interlayer coupling, and patterning effects in 2D transition-metal dichalcogenides (MoS2, WSe2, ReS2) and van der Waals heterostructures. Topics span chalcogen-vacancy healing by organic molecules, anisotropic lattice control, twisted-bilayer coupling, and electrical-switching device applications — atomic dynamics in two-dimensional planes.
Research Highlights
STEM revealed a feasible mechanism for a breakthrough in metal thin-film epitaxy, enabling wafer-scale growth of atomically flat, grain-boundary-free Ag films on Cu buffers with a 13% lattice mismatch.
Reports on Progress in Physics · 89(2), 028002
Machine learning-driven electron spectroscopic imaging (ESI) approach enables nanoscale visualization and statistical analysis of structural degradation in PEMFC electrodes.
Applied Catalysis B: Environment and Energy · 382, 125911
Deep learning-empowered electron tomography visualizes the 3D structure of porous carbon supports in PEMFC with only a few image frames, reducing data requirements by more than 95%.
Journal of Energy Chemistry · 104, 795-806
The hidden role of the oxygen vacancy (VO) in facilitating Li-ion transport in lithium lanthanum titanate solid electrolytes was unveiled. The VO is directionally interconnected to form a 2D network parallel to the c-plane.
ACS Energy Letters · 9(11), 5606-5615
4D STEM-based crystallographic domain mapping unfolded the mystery of aging-induced efficiency improvement of perovskite solar cells, revealing that aging can induce partial lattice strain relaxation.
ACS Energy Letters · 9(7), 3618-3627
Multimodal 4D-STEM (PA)CBED/EDX/EELS at multiple length scales revealed the complicated interplay of structure, cation defects, and oxygen vacancies in La-doped HfO2 films.
ACS Nano · 17(19), 19076-19086
Pico-scale precision electron microscopy revealed that porous copper nanostructures self-regulate giant oxidation resistance by constructing a curved surface that generates a series of monoatomic steps.
Advanced Materials · 35(42), 2210564
Proposes a rapid, automated, and reliable analytical method for the morphological distribution of Pt-based electrocatalyst nanoparticles using deep learning processing.
Materials Today Energy · 36, 101348
Electron microscopy describes in detail the initial growth of copper thin films required for mono-atom step-level flat surfaces (MSFSs). Deposition by atomic sputtering epitaxy leads to the coherent merging of trillions of islands.
Nature Communications · 14, 685
Computation-aided STEM-EELS showed that the distributions of surface oxygen vacancies and reduced Ti valencies enclosing a stoichiometric core were responsible for enhanced photocatalytic activity.
Applied Catalysis B-Environmental · 322, 122140
Advanced segmented electron tomography visualizes the 3D structure of carbon supports in proton exchange membrane fuel cells (PEM-FCs) and explains structural degradation phenomena.
Journal of Energy Chemistry · 74, 359-367
A multimodal investigation based on microscopy and spectroscopy showed that ferroelectric properties improve when films of hafnium oxide are bombarded with a beam of helium ions.
Science · 376(6594), 731-738
Deep learning crystallographic analysis unequivocally addresses structure problems for sub-10 nm polycrystalline hafnium zirconium oxide thin films.
Small · 18(18), 2107620
Comprehensive atomic-resolution microscopy revealed that wafer-scale Cu(111) single-crystal thin films free of multilayer step edges show a semi-permanent oxidation resistance.
Nature · 603(7901), 434-438
Electron microscopy and spectroscopy combined with theoretical modeling can directly resolve the site-specific doping phenomena of transition metals in SnO2 nanoparticles.
Applied Catalysis B-Environmental · 305, 121083
A deep learning-assisted quantification algorithm reduces the heavy load of data processing for researchers, which has hindered the pace of design and development of two-dimensional transition metal dichalcogenides.
Advanced Science · 8, 2101099
STEM-based statistical lattice strain analysis provided the first demonstration of successful control of the lattice structure of anisotropic ReS2, resulting in isotropic responses.
ACS Nano · 15, 13770-13780
Successfully mapped copper dopants added at 0.2 at.% into a multi-component thermoelectric alloy (Bi2Te2.7Se0.3) system using advanced energy dispersive X-ray spectroscopy (EDX) spectrum imaging.
Materials Today Physics · 17, 100347
STEM-EDX chemical mapping provides an atomic-level picture of what truly occurs with cation vacancies at an oxide interface — Sr and Ti vacancies in SrTiO3 film and Nb ions diffused from a Nb:SrTiO3 substrate.
Materials Today Physics · 16, 100302
Atomic-scale STEM observations combined with site-specific electron energy loss spectroscopy revealed that two transition metal ions substitute for Zn, in different valence states (Cr3+ and Co2+).
Journal of Materials Chemistry A · 8, 25345-25354
STEM-EELS/EDX-based work revealed that surface oxygen vacancies in anatase-type Fe@TiO2 nanoparticles for superior photocatalytic activities can be readily controlled by simple pH treatment.
Applied Surface Science · 507, 144916
Demonstrates how the geometrical parameters of active cathode materials are inextricably linked to their battery performance, based on multimodal/multiscale characterizations.
ACS Applied Materials & Interfaces · 11, 4017
Investigates the hitherto unsolved conundrum of how organic chemicals can heal chalcogen vacancies in MoS2 monolayer at the atomic scale, observing chalcogen vacancy healing by an organic molecule.
Nano Letters · 18(7), 4523
Provides the first demonstration of octahedra-derived multiferroic properties that can be stabilized in a thin film form without the help of complex chemical modifications.
Advanced Functional Materials · 19, 1800839
Using atom-resolved STEM-EDX/EELS, this study reveals that the electrochemical properties of Li(NixCoyMnz)O2 (NCM) materials are decisively determined by the interplay of combined cation disordering.
Journal of Materials Chemistry A · 6, 16111
Atomic-resolution STEM-EDX chemical mapping was conducted for the direct visualization of atomic-scale interstitial and antisite defects in complete single-phase Ti1-xHfxNiSn1-ySby half-Heusler (HH) thermoelectric alloys.
Advanced Materials · 29, 1702091
By a combination of microscopy and spectroscopy techniques, we directly observed for the first time the field effect in a ferroelectric material that exhibits switchable electrical polarization at the ferroelectric–electrode interface.
Nature Materials · 13, 1019
Local oxygen stoichiometry in functional oxides has been a long-standing challenge. The method developed in this work quantifies oxygen vacancy distribution and homogeneity, which directly control the operation of solid-oxide fuel cells.
Nature Materials · 11, 888
