In a research study published on arXiv, physicists at the Massachusetts Institute of Technology (MIT)—Pierre-Antoine Graham, Ahmed Abouelkomsan, and Liang Fu—have demonstrated the first complete, non-perturbative solution of the non-Abelian fractional quantum Hall (FQH) problem across all Landau levels.
Detailed in the paper (“Non-Abelian quantum Hall states at strong Landau level mixing: finding anyons with first-principles AI“), the team deployed a self-attention neural network wave function (MagNet) to solve the continuous real-space Hamiltonian for spin-polarized electrons at filling factor ν = 3/2, identifying the anti-Pfaffian as the ground-state topological order under realistic inter-level mixing.
| [ Summary of Neural Network VMC Benchmark Results (ν = 3/2 FQH State) ] | ||
|---|---|---|
| Parameter / Metric | NN-VMC Result | Physical Significance |
| • Variational Energy (κ = 1.0) | • 5.75 ℏωc (NN) vs 6.08 ℏωc (ED) | • Substantially lower energy achieved by capturing inter-level mixing. |
| • Ground-State Degeneracy | • 6-fold (N = 18) / 2-fold (N = 21) | • Matches theoretical non-Abelian torus topology for even/odd particles. |
| • Anyon Quantum Dimension (da) | • dσ = √2 (ΔSn = log√2) | • First-principles proof of Ising-type non-Abelian anyons. |
| • Candidate State Overlap | • Anti-Pfaffian > Pfaffian >> PH-Pfaffian | • Anti-Pfaffian favored across all mixing scales κ ∈ [0.5, 5.0]. |
| • Phase Boundary Transition | • κ = 20.0 | • Non-Abelian liquid melts into a triangular Wigner crystal. |
Overcoming the 30-Year Landau Level Mixing Bottleneck
Non-Abelian anyons in even-denominator FQH states (such as ν = 5/2 or ν = 3/2) host non-local degenerate ground states suitable for fault-tolerant topological quantum computation. However, identifying whether the ground state is a Moore–Read Pfaffian, an anti-Pfaffian, or a particle-hole (PH) Pfaffian has remained contentious. Conventional numerical exact diagonalization (ED) truncates calculations to a single Landau level, failing when Coulomb interactions create strong Landau level mixing (κ = U / ℏωc ≥ 1).
To solve this without approximations, the MIT team developed a momentum-resolved extension of their MagNet neural network ansatz:
- Continuous Real-Space Architecture: The network parametrizes many-body electron orbitals using multi-head self-attention and multilayer perceptrons, enforcing exact magnetic boundary conditions on a torus while summing center-of-mass magnetic translations to isolate specific momentum sectors.
- First-Principles Anyon Extraction: Minimally Entangled State (MES) analysis over noncontractible torus cuts revealed that two of the six degenerate ground states exhibit an excess Rényi entanglement entropy of ΔSn = log√2, proving the emergence of non-Abelian Ising anyons (dσ = √2) without inputting prior topological assumptions.
- Anti-Pfaffian Dominance: Overlap calculations across mixing parameters up to κ = 5 confirmed that breaking particle-hole symmetry through Landau level mixing consistently favors the anti-Pfaffian topological order over the Pfaffian and PH-Pfaffian.
- Landau Level Depletion & Wigner Crystallization: The model revealed that strong mixing depletes electron occupations in both the lowest (n = 0) and first-excited (n = 1) Landau levels, transferring weight into n ≥ 2 levels. At extreme mixing (κ = 20), the pair correlation function g(r) develops crystalline peaks, tracking the transition from a non-Abelian Hall liquid to a Wigner crystal.
The findings provide direct theoretical guidance for ongoing non-Abelian experiments in monolayer WSe2, ZnO, and wide GaAs quantum wells where strong Landau level mixing dominates.
What This Means (In Plain English)
Topological quantum computers aim to store and process information using “non-Abelian anyons”—exotic quantum quasiparticles whose braided movements naturally shield quantum information from environmental noise and decoherence. For over three decades, theoretical physicists predicted these particles should exist in two-dimensional electron liquids under strong magnetic fields. However, proving exactly which quantum state nature prefers remained an open question because classical supercomputers could not handle the mathematical complexity of electrons jumping between energy levels (a phenomenon known as Landau level mixing). By using a specialized self-attention neural network, the MIT team solved these complex continuous equations from first principles without pre-programmed physics assumptions. Their AI model proved that realistic energy-level mixing favors a specific topological state called the anti-Pfaffian, confirming the presence of Ising-type non-Abelian anyons and providing a clear computational roadmap for building noise-resistant topological qubits in materials like monolayer WSe2 and semiconductor quantum wells.
Review Ahmed Abouelkomsan’s summary on LinkedIn here and inspect the technical manuscript on arXiv here.
October 17, 2026