For streamlit implementaion see link
We investigate the intrinsic dimension across the many-body localization transition as a means of identification. We obtain intrinsic dimension via 2NN, a neighbour graphs algorithm which asses distribution of neighbour-distance-ratios. Using scaling collapse we attempt to determine the critical parameter value.
See paperFigs.ipynb for code along (this can be run on a laptop, but we recommend that you limit yourself to small system sizes and few realizations to start).
We observe that eigenvectors become dominated by a single configuration as disorder is increased beyond a critical region, see the figure below.
If you have comments or questions please send them our way!