A Complex Network Analysis on The Eigenvalue Spectra of Random Spin Systems

Qiaomu Xue,Wenjia Rao

DOI: https://doi.org/10.1016/j.physa.2024.129572

2024-02-06

Abstract:Recent works have established a novel viewpoint that treats the eigenvalue spectra of disordered quantum systems as time-series, and corresponding algorithms such as singular-value-decomposition has proven its advantage in studying subtle physical quantities like Thouless energy and non-ergodic extended regime. On the other hand, algorithms from complex networks have long been known as a powerful tool to study highly nonlinear time-series. In this work, we combine these two ideas together. Using the particular algorithm called visibility graph (VG) that transforms the eigenvalue spectra of a random spin system into complex networks, it's shown the degree distribution of the resulting network is capable of signaturing the eigenvalue evolution during the thermal to many-body localization transition, and the networks in the thermal phase have a small-world structure. We further show these results are robust even when the eigenvalues are incomplete with missing levels, which reveals the advantage of the VG algorithm.

Disordered Systems and Neural Networks

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to explore how to use complex network analysis methods, especially the Visibility Graph (VG) algorithm, to study the evolution characteristics of the eigenvalue spectrum in the process of thermalization to many - body localization (MBL) transition in disordered quantum systems. Specifically, by converting the eigenvalue spectra of random spin systems into complex networks, the paper studies the topological properties of these networks, such as degree distribution, clustering coefficient and shortest path length, to reveal the differences between different phases and explore the effectiveness of these methods in dealing with incomplete eigenvalue spectrum data.

### Main research questions:
1. **Evolution of the eigenvalue spectrum**: Study how the eigenvalue spectrum changes during the transition from the thermalized phase to the many - body localized phase, especially in terms of eigenvalue correlations.
2. **Topological properties of complex networks**: Use the Visibility Graph algorithm to convert the eigenvalue spectrum into a complex network and study the topological properties such as degree distribution, clustering coefficient and shortest path length of these networks.
3. **Small - world structure**: Explore whether the networks in the thermalized phase and the transition region have a small - world structure and whether this structure exists in the many - body localized phase.
4. **Processing of incomplete data**: Evaluate the robustness and effectiveness of the Visibility Graph algorithm in dealing with incomplete eigenvalue spectrum data.

### Research background:
- **Thermalized phase and many - body localized phase**: The thermalized phase is characterized by the correlation of the eigenvalue spectrum, following the Wigner - Dyson distribution; the many - body localized phase is characterized by the non - correlation of the eigenvalue spectrum, following the Poisson distribution.
- **Limitations of traditional methods**: Traditional random matrix theory (RMT) methods mainly rely on statistics such as eigenvalue spacing distribution, but these methods have limitations in dealing with incomplete data.
- **Advantages of the Visibility Graph algorithm**: The Visibility Graph algorithm can regard the eigenvalue spectrum as a time series and convert it into a complex network, thus providing a new perspective to study the evolution of the eigenvalue spectrum.

### Main findings:
1. **Change in degree distribution**: The degree distribution of the eigenvalue spectrum in the thermalized phase is significantly smaller than that in the many - body localized phase, which reflects the change in eigenvalue correlations.
2. **Small - world structure**: The networks in the thermalized phase and the transition region have a small - world structure, while the many - body localized phase does not have this structure.
3. **Robustness**: The Visibility Graph algorithm shows good robustness in dealing with incomplete eigenvalue spectrum data and can maintain the topological properties of the eigenvalue spectrum to a certain extent.

### Conclusion:
By converting the eigenvalue spectrum into a complex network, the paper successfully reveals the differences between the thermalized phase and the many - body localized phase, and shows the advantages of the Visibility Graph algorithm in dealing with incomplete data. These results provide new tools and methods for further research on phase transitions in disordered quantum systems.