Fermion sign problem and the structure of Lee-Yang zeros. II. Finite temperature results for a model system without interactions

Using an analytically solvable noninteracting one-dimensional particle-on-a-ring model, this paper investigates how Lee-Yang zeros evolve with temperature to explain the failure of standard analytic continuation methods at low temperatures and proposes a novel fitting strategy that combines high-temperature extrapolation with temperature-dependent modeling to overcome the fermion sign problem.

The Gist

Imagine you are trying to predict the behavior of a crowd of invisible, ghostly dancers (fermions) in a room. In the world of quantum physics, these dancers have a very strict rule: no two can ever occupy the exact same spot at the same time. This rule makes them incredibly difficult to simulate on a computer because their mathematical "signs" keep flipping between positive and negative, canceling each other out like noise in a radio signal. This is known as the Fermion Sign Problem.

To solve this, scientists usually try to simulate the dancers when they are "friendly" (bosons, who can share spots) or "neutral" (distinguishable particles), and then mathematically stretch or "extrapolate" those results to figure out what the strict fermions are doing.

This paper acts as a guidebook for understanding why this stretching trick often fails when the room gets cold, and offers a new way to make the trick work.

The Map of "Zero" Points (Lee-Yang Zeros)

The authors use a special mathematical map to track invisible "zero points" (called Lee-Yang zeros). Think of these zeros as landmines on a bridge.

• The Bridge: The bridge represents the path from "friendly" particles to "strict" fermions.
• The Landmines: If you try to walk across the bridge and step on a landmine (a zero point), your calculation explodes or becomes nonsense.

At Absolute Zero (0 Kelvin):
The landmines are lined up perfectly on the bridge, blocking the path. You cannot walk from the start to the strict fermion side without hitting a mine. This explains why, at very low temperatures, standard computer simulations fail.

As the Room Warms Up (Finite Temperature):
As the temperature rises, the landmines start to move. They drift off the bridge and into the "ocean" of imaginary numbers.

• Low Temperature: The most dangerous landmine (the one closest to the strict fermion side) stays right on the bridge. It's like a guard standing in your way. Even if you try to walk around it with a fancy high-tech map (high-order fitting), you still can't get past it. This is why previous methods failed at low temperatures.
• High Temperature: Eventually, as it gets warmer, all the landmines move far enough into the ocean that the bridge is clear. Now, you can safely walk from the friendly side to the strict side.

The Parity Puzzle (Even vs. Odd Dancers)

The paper also noticed a funny quirk based on whether there is an even or odd number of dancers:

• Even Number: The landmines behave like a pair of dancers holding hands; they merge and then jump off the bridge together.
• Odd Number: One landmine stays on the bridge a bit longer, waiting for a partner before they both jump off.
  This difference changes the shape of the "bridge" slightly, but the main rule remains: Cold = Blocked Bridge; Warm = Clear Bridge.

The New Strategy: The "Two-Step" Dance

Since the bridge is blocked at low temperatures, the authors propose a clever workaround, like taking a detour:

1. Step 1: The High-Temperature Run: Wait until the room is warm enough that the landmines have moved off the bridge. Now, safely walk across the bridge to get a reliable snapshot of the strict fermions' behavior.
2. Step 2: The Temperature Slide: Once you have that reliable snapshot from the warm room, don't try to walk back across the blocked bridge to get the cold data. Instead, use the warm data to draw a smooth curve (a mathematical fit) that slides down the temperature scale.

Think of it like this: If you want to know how a car engine behaves in the freezing cold, but the engine freezes up if you try to test it directly, you first test it in a warm garage where it runs perfectly. Then, you use that perfect data to mathematically predict how it would behave in the cold, without ever actually trying to start the frozen engine.

The Bottom Line

The paper proves that the reason old methods failed at low temperatures was that they were trying to cross a bridge full of landmines. By understanding exactly where those landmines move as the temperature changes, the authors show that we can bypass the problem entirely. We can get accurate data by starting in the "safe zone" (high temperature) and sliding down to the cold, rather than trying to force our way through the blocked path.

This provides a clear, solvable example of how to handle difficult quantum simulations, offering a potential new path for understanding more complex, real-world systems in the future.

Technical Summary

Technical Summary: Fermion Sign Problem and the Structure of Lee-Yang Zeros. II. Finite Temperature Results for a Model System without Interactions

Problem Statement
The fermion sign problem (FSP) remains a fundamental obstacle in simulating fermionic systems at finite temperatures. Previous work established that at absolute zero ($T=0$), the Lee-Yang (LY) zeros of the exchange parity parameter $\xi$ are universally distributed at $z_k = -1/k$ for $k=1, \dots, N-1$, independent of interactions. This distribution implies that the fermionic point ($\xi = -1$) is a singularity, obstructing direct analytic continuation from the bosonic/distinguishable regime ($\xi \in [0, 1]$). However, the behavior of these zeros at finite temperatures ($T > 0$) and how their evolution impacts the validity of extrapolation schemes used in state-of-the-art simulations (such as direct polynomial fitting or constant-energy contour methods) had not been analytically characterized. Specifically, it was unclear why high-order fitting schemes often fail at low $T$ but succeed at higher $T$, and how the analytic structure of the partition function reshapes as temperature increases.

Methodology
To isolate finite-temperature effects from interaction effects, the authors employ an exactly solvable, non-interacting model: indistinguishable particles on a one-dimensional ring.

1. Analytic Construction: The partition function $Z(N, \beta, \xi)$ is derived analytically as a polynomial in $\xi$ using the permutation group expansion. The coefficients are determined by the single-particle energy eigenvalues and the cycle structures of permutations.
2. Zero Tracking: The trajectories of the LY zeros ($z_i$) in the complex $\xi$-plane are tracked as a function of temperature $T$ for systems with $N=4$ and $N=5$.
3. Thermodynamic Analysis: The authors analyze the free energy $F(T, \xi)$ and the sign factor $\langle s \rangle_B$ in relation to the positions of these zeros, utilizing an electrostatic analogy where zeros act as charged lines.
4. Extrapolation Testing: The validity of two extrapolation approaches is tested against the exact solution:
   • Direct polynomial extrapolation of $\xi$ from $[0, 1]$ to $-1$.
   • Implicit extrapolation via constant-energy contours (fitting $\xi$ as a function of $T$ for fixed energy).
5. Structural Decomposition: The partition function is decomposed to isolate the $\xi$-independent remainder $\phi(\beta)$, which is identified as the fermionic partition function $Z_F$.

Key Contributions and Results

• Finite-T Evolution of LY Zeros:

  • At low $T$, the zero originating from $z_1 = -1$ remains extremely close to the real axis (specifically near $\xi = -1$), governed by the relation $z_1 \approx -1 + e^{-\beta E_0}$.
  • As $T$ increases, zeros move away from the real axis. For even $N$, parity constraints force the $z_1$ zero to move to the left of $-1$ before pairing with $z_2$ to form a complex conjugate pair. For odd $N$, $z_1$ remains on the real axis until it meets another zero, after which they leave the real axis as a conjugate pair.
  • At sufficiently high $T$, all zeros move off the real axis, restoring analyticity along the real $\xi$ axis between $-1$ and $1$.

• Mechanism of Extrapolation Failure:

  • The failure of both direct and implicit (constant-energy contour) extrapolation schemes at low $T$ is directly attributed to the proximity of LY zeros to the real axis.
  • When zeros lie on or near the real axis, the free energy exhibits sharp divergences or strong distortions, making the function non-analytic or highly sensitive to noise.
  • Even high-order polynomial fits (e.g., sixth-order) fail to recover the exact fermionic energy at low $T$ because the continuation path is obstructed by these zeros. The error persists regardless of the fitting order until the zeros move sufficiently far from the real axis at higher temperatures.

• Sign Factor and Reweighting:

  • The sign factor $\langle s \rangle_B$ is dominated by the zero $z_1$ closest to $\xi = -1$. As $T \to 0$, $z_1 \to -1$, causing $\langle s \rangle_B \to 0$ exponentially. This explains the convergence failure of reweighting methods in the low-$T$ limit, as the fermionic signal becomes exponentially small compared to the bosonic background.

• Partition Function Structure and New Strategy:

  • The authors demonstrate that dividing the partition function polynomial by $(\xi - z_1)$ (approximated by $\xi + 1$ at low $T$) yields a remainder $\phi(\beta)$ that is exactly the fermionic partition function $Z_F$.
  • At low $T$, $Z_F$ is exponentially small, making it difficult to extract via standard extrapolation from $\xi \in [0, 1]$ due to the overwhelming contribution from the remaining polynomial terms.

• Proposed Two-Step Strategy for Low-T Properties:
  Based on the structural analysis, the paper proposes a specific route to obtain low-$T$ fermionic properties:

  1. High-T Sampling: Sample sign-problem-free data in the region $\xi \in [0, 1]$ at moderate to high temperatures where LY zeros are far from the real axis.
  2. High-T Extrapolation: Extrapolate this data to $\xi = -1$ to obtain reliable estimates of $Z_F$ (or derived quantities) at these higher temperatures.
  3. Temperature Fitting: Fit the high-$T$ fermionic data as a function of temperature (e.g., fitting $\ln Z_F$ or $\ln \phi$) and continue this fit toward lower temperatures.
  • Results show that this temperature-based continuation accurately recovers exact low-$T$ thermodynamic properties (energy, heat capacity) even when direct $\xi$-extrapolation fails.

Significance
The paper provides a rigorous, analytically solvable benchmark that diagnoses the limitations of analytic continuation methods for the fermion sign problem. It clarifies that the breakdown of extrapolation at low temperatures is a fundamental consequence of the Lee-Yang zero structure, not merely a numerical artifact. By identifying the $\xi$-independent remainder $\phi(\beta)$ as the fermionic partition function, the authors suggest a viable alternative strategy: shifting the continuation burden from the complex $\xi$ plane (where zeros obstruct the path at low $T$) to the temperature axis (where the fermionic properties, once isolated at high $T$, can be smoothly continued). This offers a potential pathway for treating more realistic interacting fermionic systems, provided the high-temperature input data can be reliably obtained.