[Paper Review] Numerical Evidence for the Haldane Conjecture

B. Allés, Alessandro Papa|ArXiv.org|Nov 10, 2008

Spectral Theory in Mathematical Physics2 references3 citations

TL;DR

This study provides numerical evidence supporting the Haldane conjecture by simulating the 2D O(3) nonlinear sigma model with an imaginary θ term, enabling Monte Carlo calculations that avoid the sign problem. The correlation length diverges at θ = 3.10(5), consistent with the predicted critical point at θ = π, confirming the conjecture via analytic continuation from imaginary to real θ values using a novel fast cluster algorithm.

ABSTRACT

The Haldane conjecture, when applied to the Heisenberg O(3) model with a θterm in two dimensions, states that the correlation length ξdiverges when θapproaches π. To verify this conjecture we have numerically simulated the model at imaginary θand then analytically continued the results to real θ. We have obtained that the value where the model should become critical is θ=3.10(5) in agreement with the expectation.

Motivation & Objective

• To test the Haldane conjecture, which predicts a diverging correlation length at θ = π in the 2D O(3) nonlinear sigma model with a θ term.
• To overcome the sign problem in Monte Carlo simulations at real θ by performing simulations at imaginary θ values.
• To develop and apply a new fast cluster algorithm for simulating the O(3) model with a nontrivial topological charge operator at imaginary θ.
• To analytically continue correlation length data from imaginary θ to real θ and determine the critical θ value where divergence occurs.
• To verify the robustness of the result across different topological charge density operators, including one requiring renormalization and one geometrically defined.

Proposed method

• The model is formulated on a square lattice with periodic boundary conditions, using a 3-component unit vector field φ(x) and a lattice action that includes a θ term via the topological charge Q.
• The partition function is defined with a Boltzmann weight exp(−S), where S = A − iθQ, with A representing the spin-spin interaction and Q the total winding number.
• Simulations are performed at imaginary θ (denoted as ϑ = −iθ) to avoid the complex action problem and allow importance sampling.
• A new fast cluster algorithm is developed specifically for the nontrivial topological charge operator Q^(1), enabling efficient sampling at nonzero imaginary θ.
• The correlation length ξ is extracted from correlation functions using plateaux in the effective mass, and extrapolated to real θ via functional fits.
• Analytic continuation from imaginary θ to real θ is performed using functional forms without theoretical prejudice, with consistency checked across different fitting functions.

Experimental results

Research questions

• RQ1 Does the correlation length of the 2D O(3) nonlinear sigma model diverge at θ = π, as predicted by the Haldane conjecture?
• RQ2 Can numerical evidence for criticality at θ = π be obtained despite the sign problem at real θ?
• RQ3 Is the critical θ value robust under different definitions of the topological charge density operator, including one requiring renormalization?
• RQ4 Does the analytic continuation from imaginary θ to real θ yield consistent and reliable results for the critical point?
• RQ5 What is the range of imaginary θ values that allows stable and precise extraction of the correlation length?

Key findings

• The correlation length diverges at θ = 3.10(5), providing strong numerical support for the Haldane conjecture.
• The critical value θ = 3.10(5) is consistent with the predicted θ = π, within statistical errors.
• The result is robust across different topological charge density operators: one requiring renormalization (Q^(1)) and one geometrically defined (Q^(2)), both yielding the same critical θ.
• The analytic continuation from imaginary θ to real θ is reliable and insensitive to the choice of functional form used in the extrapolation.
• Simulations at large imaginary θ values (ϑ ≥ 15) show reduced precision in correlation length extraction due to shrinking plateaux, indicating a practical limit on the range of usable data.
• The analytic continuation paths remain far from phase transition lines in the θ–β plane, avoiding nonanalytic regions and ensuring the validity of the extrapolation.