Stemdiff

STEMDIFF :: 4D-STEM dataset to 2D-diffractogram

• The STEMDIFF package converts... <br> ... a 4D-STEM dataset from a SEM microscope (huge and complex) <br> ... to a 2D-powder diffraction pattern (simple and easy to work with).
• If you use EDIFF in your research, please cite our recent paper:
  • Microscopy and Microanalysis 31, 2025, ozaf045. <br> https://doi.org/10.1093/mam/ozaf045

Principle

<img src="https://mirekslouf.github.io/stemdiff/docs/assets/principle.pptx.png" alt="STEMDIFF principle" width="500"/>

Installation

• Requirement: Python with sci-modules = numpy, matplotlib, scipy, pandas
• pip install scikit-image = 3rd party package for advanced image processing
• pip install tqdm = to show progress meter during long summations
• pip install idiff = to improve diffractograms (remove noise, background ...)
• pip install stemdiff = STEMDIFF package itself (uses all packages above)

Quick start

• Worked example shows the STEMDIFF package in action.
• Help on GitHub with complete package documentation and additional examples.

Documentation, help and examples

• PyPI repository - the stable version to install.
• GitHub repository - the current version under development.
• GitHub Pages - the more user-friendly version of GitHub website.

Versions of STEMDIFF

• Version 1.0 = Matlab: just a simple summation of 4D-dataset
• Version 2.0 = like v.1.0 + post-processing in Jupyter
• Version 3.0 = Python scripts: summation + S-filtering
• Version 4.0 = Python package: summation + S-filtering + deconvolution
  • summation = summation of all 2D-diffractograms
  • S-filtering = sum only diffractograms with strong diffractions = high S
  • deconvolution = reduce the primary beam spread effect ⇒ better resolution
• Version 4.2 = like v.4.0 + a few important improvements, such as:
  • sum just the central region with the strongest diffractions ⇒ higher speed
  • 3 centering types: (0) geometry, (1) center of 1st, (2) individual centers
  • improved definition of summation and better documentation
• Version 5.0 = complete rewrite of v.4.2
  • conversion 2D-diffractogram → 1D-profile moved to package EDIFF
  • better filtering (including estimated number of diffractions)
  • more detectors + more types of deconvolution (beta; to finish in v.6.0)
• Version 5.1 = (beta) support for parallel processing
• Version 5.2 = (beta) improvement of diff.patterns in sister package idiff

Acknowledgement

The development was co-funded by TACR, program NCK, project TN02000020.