PANDA: Perovskite Assembly and Nano Domain Analysis

PANDA is a Python-based workflow for the automated analysis of 4D Scanning Transmission Electron Microscopy (4D-STEM) data. It was specifically developed for analyzing Perovskite Quantum Dots and Superlattices (e.g., CsPbBr3), focusing on particle detection, structural clustering, and internal strain mapping.

Overview

This repository contains Jupyter Notebooks that perform end-to-end analysis of 4D-STEM datasets. The code allows researchers to move from raw datacubes to quantified strain maps and domain classification.

The workflow includes:

1. py4DSTEM + skimage Blob Analysis: Automatically identifies nanoparticle locations in spatial images using Laplacian of Gaussian (LoG) blob detection from scikit-image, integrated with py4DSTEM data structures.
2. Diffraction Feature Extraction: Extracts and processes diffraction patterns from the 4D datacube for every detected particle.
3. Unsupervised Clustering: Classifies particles based on crystallographic orientation and structural similarity using PCA (Principal Component Analysis), K-Means clustering, and UMAP.
4. Virtual Imaging: Generates Bright Field (BF) and Dark Field (DF) images from the 4D data.
5. Automated Crystal Orientation Mapping (ACOM): Matches experimental diffraction patterns to simulated crystal structures to determine orientation at every pixel.
6. Strain & Rotation Analysis: Quantifies relative lattice rotation and strain components ($\epsilon_{xx}, \epsilon_{yy}, \epsilon_{xy}$) to identify internal twist domains within superlattices.

Dependencies

This code relies heavily on the py4DSTEM open-source toolkit for processing 4D-STEM data and scikit-image for image analysis.

To run the notebooks, you will need the following Python packages:

• py4DSTEM (Core analysis library)
• scikit-image (Image processing and blob detection)
• numpy & matplotlib (Data manipulation and plotting)
• scikit-learn (PCA and K-Means clustering)
• umap-learn (Dimensionality reduction for visualization)
• h5py (Handling HDF5 data files)
• pandas (Data organization)
• scipy (Statistical fitting)

Data Availability

This code was originally written to analyze experimental data for lead halide perovskite superlattices.

Note: The raw experimental datasets (.h5 and .prz files) referenced in the notebooks are not included in this repository due to size constraints and proprietary restrictions.

The notebooks are provided as a comprehensive reference for the analysis workflow. To use this code, you will need to adapt the file paths to point to your own 4D-STEM datasets formatted compatible with py4DSTEM.

Usage

1. Clone this repository.
2. Ensure your environment has the required dependencies installed.
3. Open PANDA.ipynb in Jupyter Notebook or Jupyter Lab.
4. Update the file_path and dirpath variables to point to your local data.