Spectral Detection of Connectivity in Experimental Networks Under Chaotic Regimes

The inference of links out of time series in complex systems of interacting nodes is a fundamental issue in experimental science. Connectivity detection becomes challenging when the system dynamics lacks well-defined spectral features, as is most notably the case in the presence of chaos. By leveraging an analog electronic platform implementing a network of Burridge–Knopoff oscillators, it was recently shown that chaos induces an exponential decay of connectivity as a function of the topological distance between oscillators, a phenomenon interpreted in terms of a “topological unpredictability”. Here we extend these findings by addressing, on the same experimental system, the spectral properties of the dynamics. Interestingly, connectivity assessed in the frequency domain also exhibits an exponential decay with topological distance. Our results show that chaotic regimes can hinder the detection of links in real-world systems, with significant implications in many research fields, from neuroscience to Earth science.