What this paper is about: the authors show that a geometric object called the higher Berry curvature can be used, in a practical and manifestly quantized way, to compute a higher topological invariant known as the second Chern number. They test this idea on a standard four-dimensional lattice model (a lattice version of the 4D Dirac or Chern insulator) and find exact agreement between the phase diagram produced by the higher invariant and the known analytic second Chern number for that model.

What the researchers did: they rewrote the four-dimensional lattice model as a family of infinite, translationally invariant one-dimensional chains. The family is labeled by the three remaining crystal momenta, so the problem becomes a set of one-dimensional problems that vary over a three‑dimensional parameter space (the three-torus). For each parameter value they find the ground state with an infinite matrix product state (iMPS) using the infinite density matrix renormalization group (iDMRG). They then compute a discrete version of the higher (three-form) Berry curvature across a triangulation of the parameter space and sum it to obtain the Dixmier–Douady–Kapustin–Spodyneiko (DDKS) number, the higher analogue of the usual Chern number.

How the method works at a high level: ordinary Berry curvature measures how a single quantum state changes when you slowly change parameters. The higher Berry curvature is a higher-dimensional generalization that applies to families of extended states. In the iMPS approach the authors use the dominant eigenvectors of mixed transfer matrices between nearby iMPS tensors to define the needed transition data. This construction is a discrete, lattice-friendly generalization of the Fukui–Hatsugai–Suzuki algorithm for the ordinary Chern number. By summing contributions over the triangulated parameter space the method returns integer values by construction.