Gradient-based search of quantum phases: discovering unconventional fractional Chern insulators

André Grossi Fonseca, Yan Wang, Sachin Vaidya, Patrick J. Ledwith, Ashvin Vishwanath, Marin SoljačićarXiv 2025 · DOI · arXiv

Abstract

The discovery and understanding of new quantum phases has time and again transformed both fundamental physics and technology, yet progress often relies on slow, intuition-based theoretical considerations or experimental serendipity. Here, we introduce a general gradient-based framework for targeted phase discovery. We define a differentiable function, dubbed "target-phase loss function", which encodes fingerprints of a quantum state, thereby recasting phase search as a tractable optimization problem in Hamiltonian space. The method is broadly applicable to a wide range of symmetry-broken and topological orders and can be interfaced with most many-body numerical solvers. As a demonstration, we apply it to spinless fermions on the kagome lattice using exact diagonalization and discover two distinctive fractional Chern insulators (FCIs): (i) at filling $\nu= 1/3$, a "non-ideal" Abelian FCI whose band geometry lies far beyond the Landau-level mimicry paradigm and all recent generalizations; and (ii) at $\nu= 1/2$, a non-Abelian FCI stabilized purely by finite-range two-body interactions. These results provide the first explicit realization of such types of FCIs and establish a versatile paradigm for systematic quantum-phase discovery.

Introduction

Identifying novel phases of matter is a central challenge in condensed matter physics. Traditionally, this process has been driven by theoretical intuition—often focusing on "ideal" limits—or accidental experimental discovery. However, as researchers explore increasingly complex systems with high-dimensional parameter spaces (e.g., varying hopping, interaction ranges, and lattice geometries), traditional grid-based or brute-force searches become computationally unfeasible. There is a pressing need for systematic, automated methods to explore these vast landscapes, especially in regimes where physical intuition fails.

Methodology: The Target-Phase Loss Function

The researchers propose a novel framework that transforms phase discovery into a gradient-based optimization problem. At the heart of this method is the Target-Phase Loss Function ($\mathcal{L}$). Instead of minimizing energy alone, this function is designed to penalize deviations from the "fingerprints" of a specific target phase.

The gradient-based framework for targeted phase discovery, illustrating the target-phase loss function and the optimization workflow.

Key features of the loss function include:

  • Spectral Signatures: For topological phases, the function penalizes the energy gap between the target ground-state manifold and excited states.
  • Spectral Flow: By accounting for how energy levels evolve under flux insertion, the method ensures the discovered phase is robustly topological.
  • Automatic Differentiation: By using differentiable many-body solvers, the algorithm calculates gradients of the loss with respect to Hamiltonian parameters, allowing for efficient navigation of the parameter space using optimizers like Adam.

The workflow begins with a random initialization of parameters, followed by gradient descent to find candidate regions, and a final refinement step to maximize the many-body gap.

Results: Unconventional Fractional Chern Insulators

The framework was applied to a model of spinless fermions on a kagome lattice, leading to the discovery of two exotic fractional Chern insulators (FCIs) that challenge long-standing theoretical assumptions.

Discovery of a Non-Ideal Abelian FCI

Conventional wisdom suggested that FCIs require "ideal" band geometry—flat bands with uniform Berry curvature—to mimic the physics of Landau levels. The optimization search identified a stable Abelian FCI at filling $\nu=1/3$ that exists far from this ideal regime.

Discovery and characterization of a non-ideal Abelian FCI at filling ν=1/3, showing the lattice, Berry curvature, and spectral properties.

As shown in Figure 4, this state possesses significant Berry curvature fluctuations and high trace violation ($T \approx 3.3$), yet it remains a robustly gapped topological insulator. This finding proves that the "ideal" regime is only a small subset of the parameter space where these exotic phases can thrive.

Non-Abelian FCI from Two-Body Interactions

The search also uncovered a non-Abelian Moore-Read state at $\nu=1/2$. Remarkably, this state is stabilized solely by finite-range two-body interactions. This discovery is significant because it was previously thought that more complex three-body or infinite-range interactions were necessary to realize such non-Abelian states in lattice systems.

Conclusion

This gradient-based framework represents a significant step toward automated quantum discovery. By treating phase identification as a targeted optimization problem, the researchers achieved a 300-fold increase in efficiency over traditional methods. While currently demonstrated with Exact Diagonalization, the modularity of the approach allows for future integration with more scalable numerical methods, opening new frontiers for the systematic exploration of quantum matter.

Summary updated

Abstract

The discovery and understanding of new quantum phases has time and again transformed both fundamental physics and technology, yet progress often relies on slow, intuition-based theoretical considerations or experimental serendipity. Here, we introduce a general gradient-based framework for targeted phase discovery. We define a differentiable function, dubbed "target-phase loss function", which encodes fingerprints of a quantum state, thereby recasting phase search as a tractable optimization problem in Hamiltonian space. The method is broadly applicable to a wide range of symmetry-broken and topological orders and can be interfaced with most many-body numerical solvers. As a demonstration, we apply it to spinless fermions on the kagome lattice using exact diagonalization and discover two distinctive fractional Chern insulators (FCIs): (i) at filling $ν= 1/3$, a "non-ideal" Abelian FCI whose band geometry lies far beyond the Landau-level mimicry paradigm and all recent generalizations; and (ii) at $ν= 1/2$, a non-Abelian FCI stabilized purely by finite-range two-body interactions. These results provide the first explicit realization of such types of FCIs and establish a versatile paradigm for systematic quantum-phase discovery.
BibTeX
@misc{fonseca2025gradient,
  author = {André Grossi Fonseca and Yan Wang and Sachin Vaidya and Patrick J. Ledwith and Ashvin Vishwanath and Marin Soljačić},
  title = {{Gradient-based search of quantum phases: discovering unconventional fractional Chern insulators}},
  year = {2025},
  eprint = {2509.10438},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2509.10438},
  howpublished = {arXiv},
}
Figure