Simulating fermionic fractional Chern insulators with infinite projected entangled-pair states

The Gist

Imagine you are trying to understand a complex dance performed by a crowd of invisible particles. These particles, called fermions, have a very strict rule: they hate being in the same spot as each other, and they move in a way that creates a swirling, chiral (handed) pattern. When these particles arrange themselves in a specific way on a grid, they form a state of matter called a Fractional Chern Insulator (FCI). This state is exotic because it has "fractional" properties (like carrying a fraction of an electron's charge) and behaves like a fluid that never gets stuck, even without a magnetic field.

The problem is that simulating this dance on a computer is incredibly hard. Traditional methods are like trying to watch the dance by looking at only a tiny, 3x3 square of the floor. You miss the big picture and get confused by the edges of your view.

The New Approach: An Infinite Floor
The authors of this paper developed a new way to simulate this dance using a tool called iPEPS (infinite projected entangled-pair states). Think of iPEPS not as a snapshot of a small room, but as a blueprint for an infinite floor. Because it's built for an infinite grid, it doesn't suffer from the "edge effects" that confuse other methods. It allows them to see the true, infinite dance of the particles.

The Challenge: The "No-Go" Wall
There is a known rule in physics (a "no-go theorem") that says you can't perfectly describe these swirling, chiral dances using a simple, finite blueprint. It's like trying to draw a perfect circle using only straight lines; you can get close, but you'll always have little jagged edges.

To get around this, the team used a clever trick involving mathematical compression. They built their blueprint with a "bond dimension" (let's call this the detail level or D).

• Low Detail (D=4 to 6): The blueprint was too blurry. The particles looked like they were forming strange, clumpy patterns that didn't match the real physics.
• High Detail (D=7 and above): Once they increased the detail level to 7, the blueprint suddenly snapped into focus. The particles started behaving exactly as they should in an FCI. The authors found that D=7 is the "critical threshold"; below it, the simulation is wrong, but above it, the simulation is faithful to reality.

How They Checked Their Work
To make sure their infinite blueprint was correct, they looked at three specific "signatures" of the dance:

1. The Green's Function (The "Echo"): They checked how a particle's influence spreads out. In a healthy FCI, this influence should die out quickly (like a shout in a quiet room) but leave a tiny, faint "gossamer" tail. This tail is actually a sign that their simulation is hitting the "no-go" wall, but it's so small that it doesn't ruin the main picture.
2. The Pair-Correlation (The "Personal Space"): They measured how likely it is to find two particles near each other. In this state, particles keep a specific distance apart (a "correlation hole"), much like people at a party who instinctively avoid standing too close. Their simulation matched the math for a famous theoretical state called the "Laughlin state" almost perfectly.
3. The Entanglement Spectrum (The "Edge Fingerprint"): This is the most important test. Even though they simulated an infinite floor, they could mathematically "cut" it to look at the edge. The energy levels at this edge act like a fingerprint. They found a specific pattern of numbers (1, 1, 2, 3, 5) appearing in the data. This specific sequence is the unique signature of the Fractional Chern Insulator, proving they successfully simulated the exotic phase.

The Secret Sauce: Compression
Simulating these infinite floors requires massive computer power. To handle the math for the "edge fingerprint," the authors invented a compression scheme. Imagine trying to solve a puzzle with a million pieces. Instead of looking at all of them at once, they found a way to group the pieces into smaller, manageable chunks without losing the picture. This allowed them to run simulations on a "cylinder" of the lattice that was wide enough to see the true pattern, something previous methods couldn't do easily.

The Bottom Line
This paper is a breakthrough because it successfully used a powerful new simulation tool (iPEPS) to model a very difficult type of quantum matter (fermionic Fractional Chern Insulators) for the first time. They proved that if you give the simulation enough "detail" (a bond dimension of at least 7), it can accurately reproduce the complex, swirling behavior of these particles, matching both theoretical predictions and other numerical methods. This opens the door for scientists to study these exotic materials with much greater precision than before.

Technical Summary

Technical Summary: Simulating Fermionic Fractional Chern Insulators with Infinite Projected Entangled-Pair States

Problem Statement
Fractional Chern insulators (FCIs) are lattice analogues of fractional quantum Hall (FQH) states that emerge in fractionally filled narrow Chern bands without external magnetic fields. While recent experimental realizations in 2D moiré materials have renewed interest in these phases, simulating the microscopic Hamiltonians that host FCIs remains a significant numerical challenge. Existing methods, such as exact diagonalization (ED) and infinite density matrix renormalization group (iDMRG), are restricted to small clusters or finite-width cylinders, making them susceptible to finite-size effects. Furthermore, while infinite projected entangled-pair states (iPEPS) have successfully simulated bosonic chiral topological orders (e.g., chiral spin liquids), they have not previously been applied to fermionic topological order. A specific theoretical hurdle is the "no-go theorem," which forbids an exact finite-correlation-length iPEPS representation of gapped chiral phases, often leading to slow convergence in standard algorithms.

Methodology
The authors extend the iPEPS framework to fermionic systems by integrating several algorithmic advances to optimize a $U(1)$-symmetric fermionic iPEPS for a spinless-fermion Haldane model on a honeycomb lattice with density-density interactions.

1. Model and Lattice Construction: The study focuses on a Hamiltonian with nearest-neighbor (NN), second-NN, and third-NN hopping, featuring complex phases to induce a Chern band. The honeycomb lattice is coarse-grained into a Bravais lattice with a local physical Hilbert space of dimension $d=4$ (representing empty, A-site occupied, B-site occupied, and doubly occupied states).
2. Fermionic Encoding and Symmetry: To handle fermionic statistics efficiently, the authors utilize $U(1)$-symmetric tensors. This approach enforces exact particle number conservation per unit cell, avoiding the need for chemical potential tuning. For the target filling $\nu=1/3$, a $3 \times 3$ unit cell is constructed using three inequivalent tensors ($A_1, A_2, A_3$), with an auxiliary leg on $A_1$ to inject a fermion, ensuring exactly three fermions per unit cell.
3. Optimization via Fixed-Point AD: Standard imaginary-time evolution is replaced by gradient-based optimization using automatic differentiation (AD). A critical innovation is the use of the fixed-point AD method. Standard backpropagation through the Corner Transfer Matrix Renormalization Group (CTMRG) algorithm is memory-prohibitive for chiral states because CTMRG requires a large number of iterations ($O(10^3)$) to converge due to long correlation lengths. The fixed-point approach differentiates the converged fixed point directly, eliminating the need to store intermediate CTMRG steps and bypassing the memory bottleneck.
4. Entanglement Spectrum (ES) Compression: To calculate the momentum-resolved edge entanglement spectrum for large unit cells, the authors introduce a compression scheme. The matrix-product operator (MPO) representation of the reduced density matrix is coarse-grained from a large unit cell (dimension $D^M$) to a single site with a truncated dimension $D_c$ using isometries derived from Higher-Order Singular Value Decomposition (HOSVD). This allows the calculation of the ES on wider cylinders ($W=5$) that would otherwise be computationally infeasible.

Key Contributions and Results
The authors successfully optimized iPEPS states with bond dimensions up to $D=9$ and identified a critical bond dimension required to faithfully represent the FCI phase.

• Critical Bond Dimension ($D_{min} = 7$): By analyzing ground-state energy density and charge inhomogeneity, the authors found that bond dimensions $D < 7$ fail to capture the FCI phase, instead favoring states with charge modulations. Only for $D \ge 7$ does the energy drop below the first excitation gap (compared to ED results), and the charge distribution becomes homogeneous.
• Bulk Observables:
  • Green's Function: The equal-time single-particle Green's function exhibits rapid exponential decay (correlation length $\xi_{bulk}/a \approx 0.63$) followed by a long-range "gossamer" tail, consistent with the no-go theorem for finite-$D$ chiral states.
  • Pair-Correlation Function: The pair-correlation function $g(x)$ shows a correlation hole at short distances and rapid convergence to unity, matching the qualitative features of the $\nu=1/3$ Laughlin state. The lattice data aligns quantitatively with continuum Laughlin and integer quantum Hall states using a single magnetic length parameter.
• Edge Entanglement Spectrum: Using the compression scheme, the authors computed the momentum-resolved ES on a cylinder of circumference $W=5$. The low-lying spectrum exhibits the characteristic level counting $(1, 1, 2, 3, 5)$, which corresponds to the chiral $U(1)$ boson edge theory of the $\nu=1/3$ Laughlin state. This counting is robust across different environment bond dimensions ($\chi$) and compression cutoffs ($D_c$).

Significance and Claims
The paper claims to demonstrate that iPEPS can simulate chiral topological order in fermionic systems, specifically realizing an FCI at filling $\nu=1/3$ in the thermodynamic limit. The work establishes that:

1. iPEPS is a viable tool for fermionic topological order when combined with $U(1)$ symmetry and fixed-point AD optimization.
2. A specific bond dimension threshold ($D \ge 7$) is necessary to overcome the variational competition with charge-ordered states and accurately capture the chiral topological order.
3. The introduced compression scheme enables the calculation of entanglement spectra for large unit cells, providing a robust fingerprint of the FCI phase without requiring open boundaries.

The authors position this work as adding iPEPS to the limited set of numerical methods capable of simulating chiral fermionic topological order, offering a pathway to quantitatively reliable predictions for experimental realizations in moiré materials. They note that while the method is currently limited by the slow convergence of CTMRG in the forward pass, the framework is extensible to other filling fractions, lattice geometries, and potentially to the study of neutral and charged excitations using iPEPS excitation ansatzes.