This repository contains the implementation of a novel machine learning workflow for linking and cross-reconstruction of characterization data from complementary structural characterization techniques called PairVAE on and open-access block copolymer morphology characterization dataset (complementary characterizations of Small Angle X-ray Scattering (SAXS) and Scanning Electron Microscopy (SEM) of template-directed assembly of block copolymer morphologies) from Doerk. et al. Science Advances. 2023 Jan 13;9(2): eadd3687 as a proof-of-concept.
The paired SEM-SAXS dataset can be downloaded here
If you use the codes in this repository, please cite the following manuscript: Shizhao Lu and Arthi Jayaraman JACS Au 3 (9), 2510-2521 DOI: 10.1021/jacsau.3c00275
We acknowledge financial support from the U.S. Department of Energy, Office of Science (Grant DE-SC 0023264) for this work and the collaborative discussions with Quentin Michaudel and Ryan Hayward. We are also grateful for the useful technical discussions with Todd Emrick and Jessica Schiffman at the University of Massachusetts, Amherst, that were facilitated via an ongoing collaboration supported by the NSF (DMREF Grant 1921871). We are truly grateful to Gregory S. Doerk, Kevin Yager, and their coauthors for making their block copolymer characterization data openly available, which enabled us to train the models used in PairVAE. The authors also thank Gregory S. Doerk, Jonathan A. Malen, Austin M. Evans, Brent S. Sumerlin, Pramod Reddy, Vasudevan Venkateshwaran, and Rama K. Vasudevan for their questions, comments, and discussion when we presented this method to them for their critique and feedback.