Research by Problem
This page maps our work to concrete scientific bottlenecks in electron microscopy and computational imaging.
1) Low-dose, high-resolution 3D imaging
Problem
Atomic-scale 3D imaging often requires long acquisition times and high electron dose, which can damage beam-sensitive materials.
Why this is hard
Signal is weak, data are noisy, and conventional reconstructions struggle to preserve fine structure at practical dose levels.
Our approach
We combine physics-based inverse methods with ptychographic and 4D-STEM reconstruction pipelines to recover more information per electron.
Representative results - Atomic-resolution 3D volume from 4D-STEM tomography (Nature Communications, 2023) - Atomic-resolution phase-contrast volume beyond depth-of-focus limits (Physica Scripta, 2024)
2) Large-volume 3D reconstruction at practical throughput
Problem
Many materials questions require statistically meaningful volumes, not only small proof-of-concept reconstructions.
Why this is hard
Scaling to larger fields of view and many tilt conditions quickly increases compute cost and reconstruction complexity.
Our approach
We design scalable reconstruction workflows and efficient compute strategies for high-throughput, high-fidelity 3D imaging.
Representative results - Sub-Ångström end-to-end reconstruction pipeline (Physica Scripta, 2024) - Fast electron microscopy simulations (>100x) supporting rapid method development (Microscopy and Microanalysis, 2021)
3) Robust reconstruction under model mismatch and experimental complexity
Problem
Real experiments deviate from idealized assumptions (aberrations, partial coherence, drift, sample complexity), reducing reconstruction reliability.
Why this is hard
Small modeling errors can accumulate and create artifacts or bias quantitative interpretation.
Our approach
We integrate physical priors, uncertainty-aware optimization, and automation to stabilize reconstructions in realistic conditions.
Representative results - Automated end-to-end ptychographic tomography workflow (Microscopy and Microanalysis, 2025)
4) Multi-modal integration for richer materials insight
Problem
Single-mode imaging often cannot jointly resolve structure, chemistry, and functional context.
Why this is hard
Combining modalities requires careful alignment, calibration, and reconstruction across heterogeneous signals.
Our approach
We develop multimodal computational frameworks to fuse complementary signals in 3D at high resolution.
Status - Ongoing work; updates will be added as results are released.
Collaboration
If your project matches one of the problems above, please see Opportunities or Contact.