[HELP] Is non-circular atoms normal in ptychography?

[PhysXu]

Hi,

I am trying a reconstruction of a ML WS2 sample under 60 kV, the preliminary result looks OK for me.
After 1000 cycles, I can see the separate S-S due to the slight tilt of this sample, as shown below:

However, I can see the atoms in the reconstructed image are not so round, which is a bit strange to me.
I do not know whether it is normal and how to improve/solve it.

Additionally, it is not a clean enough sample and there is some surface carbon contamination.
Does this affect the reconstructed result?

I attached the result in the zip file and the yml file.
https://drive.google.com/drive/folders/1ZH04djLjLUw90_mHDD2D4H3pPM3I20mg?usp=drive_link

Any suggestions will be appreciated very much.

[chiahao3]

Hi @PhysXu ,

I'd say this is a particular successful reconstruction already! If we backtrack a couple years, people would not expect 60kV can ever revel all these details. I did similar studies on my ML WSe2 dataset back in 2023 (single slice ptycho), although the results were only included in my PhD thesis, and I used it to reconstruct the ML surface corrugation.

My colleague Dr. Jeffrey Huang did an amazing extension to retrieve 3D structure of tBL-WSe2 using tilted MEP with single projection as shown here: https://arxiv.org/abs/2509.07140

---

Regarding atom roundness, there're a couple factors in play:

1. Fundamentally, atomic potentials are not necessariy spherically symmetric and can change due to bonding. Jim Zuo had some early work on extracting this with quantitative diffraction. I'm not sure if we'll ever reach the required precision to do it in real space though.
2. On top of it, there's phonon vibration that can introduce elipticity to the perceived atom shape. Both 2021 Zhen paper and our recent moire phonon paper show that MEP can capture this to some extent if done carefully.
3. Ptychography aims to retrieve object from difffraction patterns by decoupling probe, although this is generally imperfect due to noise and algorithm factors. Therefore, there're almost always some residual probe intensities around the retrieved atom in ptychography. This gives similar probe-dependence just like ADF-STEM, where atom shape is essentially the probe shape. Ptychography improves it, but not perfectly.
4. Sample tilt also play a role and is essentially a projection artifact. For the S2 (chalcogen) column, small tilts would make the atomic column look elliptical until it's separated enough to be resolved by ptychography. Same thing happens when your do MEP on thicker sample, that's why it's better to correct for tilt first, or to post-register your slices so they don't blur your final projected image.
5. Lastly, contamination simply puts additional intensity on top of your image so they can surely affect your atom shape, although usually the contamination runs in much lower spatial frequency than the lattice structure, that's why you can still see sharp atoms on contaminated region. However, they do interfere the reconstruction process, especially when running single slice reconstruction.

---

Back to your original question, is non-circular atoms normal?
I'd say it's completely normal in ptychography, and it's mostly due to dataset quality (tilt, contamination) and reconstruction convergence (i.e., whether the probe is perfectly decoupled). You can surely improve the dataset quality and reconstruction part, although it takes significant amount of work to distinguish whether it's the reconstruction, or it's a physical feature of the system (i.e., phonon vibration) that causes the non-circularity.

Personally I think this particular reconstruction is pretty decent, as the W atoms seem mostly circular, and you can already spot the local-corrugation-induced S2 splitting, and some S vacancies. Whether to invest more time in the dataset acquisition or reconstruction really depends on your goal.

I hope these helps! By the way, the Google Drive link seems to link to an empty folder, although your screenshot is clear enough so it's totally fine.