research
acBF/Direct Ptychography
Tilt-corrected brightfield STEM was pioneered by the Kourkoutis group at Cornell as a very efficient method for reconstructing large-field-of-view images from defocused 4D-STEM data, and proved very useful for biological imaging. When I arrived, I began to consider if we could derive its contrast transfer function. This turned out to be a very fruitful question: not only could we derive the CTF1, in doing so we also arrived at expressions for how and where information is encoded by phase contrast in 4D-STEM data, and a way to turn tcBF into a direct ptychography method called acBF. acBF is capable of correcting the deleterious effects of higher order aberrations, which turn out to be a huge limitation in tcBF but can actually yield smoother information transfer when appropriately corrected.
For very low dose data or very weakly scattering samples, acBF yields images that match the quality of multislice ptychography but take only about 1 second to reconstruct. GPU-accelerated reconstructions can now be done in the py4DSTEM GUI using the fast-acbf plugin developed with Chia-Hao Lee.
acBF also enables high-quality 3D reconstructions by leveraging the parallax effect.
MEP
Multislice Electron Ptychography reconstructs the 3D scattering potential of a sample from a 4D-STEM measurement, correctly accounting for multiple scattering and other nonlinear effects, yielding a lateral resolution ultimately limited only by thermal vibrations and an axial resolution of a few nanometers. You can find ample details in the resources from the weeklong course we ran in 2025 (here) and in Chen et al, Science 2021.
MEP is now sufficiently developed to apply to many materials science challenges. The volume rendered2 below is from a gate-all-around transistor at the early stages of development: several defects such as a “mousebite” of missing channel as well as a stacking fault, are resolved by the multislice reconstruction. Multislice ptychography is a powerful tool for investigating nanometer-scale devices (of many different types).
Dynamical Diffraction Refinement
Simple approaches for mapping polarization with nanobeam diffraction break down surprisingly quickly due to sample thickness and tilt effects. A differentiable Bloch-wave diffraction model can be used to show this breakdown and to robustly reconstruct polarization in complicated samples over a large field of view.
This approach can also be used to map crystal orientation with extreme precision and simultaneously measure sample thickness.
Strain Mapping
Strain mapping with 4D-STEM suffers from dynamical diffraction artifacts: these can be largely eliminated by adding patterning to the incoming probe using a custom-made aperture (Zeltmann et al, Ultramicroscopy 2019). For complex heterostructures, better strain maps can be produced by simultaneous model-based fitting of the diffraction pattern (including the weak Moiré peaks), as implemented in py4DSTEM and applied in Bhusal et al, ACS Nano 2026.
-
When only defocus is present the CTF is \( \sin \chi \) times an approximately triangular envelope reaching to double the aperture size; other aberrations do very complicated things. ↩
-
This 3D viewer was made by the sentient GPU in my workstation ↩