arXiv:2609.19798
Variational Solution of Non-Hermitian Quantum Field Theory in Two Dimensions
Paul Romatschke
Abstract
This work describes a variational technique to solve non-Hermitian quantum systems, in particular those with negative coupling quartic self-interaction. Differences of applying variational methods from standard quantum mechanics to non-Hermitian quantum mechanics -- including potential pitfalls -- are highlighted. A successful implementation of the variational approach for the non-Hermitian case is presented. The technique is used to analyze scalar quantum field theory in d=2 with negative quartic self-interaction in the Hamiltonian formulation in lattice discretization.
AI Summary
The paper studies whether Hamiltonian basis truncation can compute nonperturbative spectra and observables of $\mathcal{PT}$-symmetric scalar field theory with negative quartic interaction in $1+1$ dimensions, where conventional Monte Carlo sampling encounters a sign problem. The construction uses contour-normalized, $\mathcal{PT}$-symmetric polynomial basis functions multiplied by a cubic exponential, then diagonalizes the truncated non-Hermitian Hamiltonian. Because ordinary variational energy minimization can fail, the basis parameter is selected by minimizing the change in the ground-state energy between successive truncations. For the massless wrong-sign quartic oscillator, eight basis states give $E_0/g^{1/3}\simeq0.930546$, with reported relative error $10^{-6}$. Massive-oscillator spectra are also compared with an isospectral Hermitian formulation. For the spatially discretized field theory, calculations mainly at $ga^2=0.05$ find two regimes distinguished by an imaginary field expectation value, separated by a finite-volume crossover. The correlation-length peak grows and narrows between three and seven sites. Low-lying eigenvalues become real as the basis is enlarged; no spontaneous $\mathcal{PT}$-symmetry breaking is observed for the parameters examined. Ground-state energies and observables are compared with saddle-point approximations. Finite-temperature pressure and entropy show no clear transition, while the field expectation value changes with temperature. The stated main limitation is exponential basis growth: approximately $10^3$, $10^4$, and $10^5$ states are required for three, five, and seven sites. These volumes do not permit a large-volume extrapolation, and finite-temperature conclusions remain inconclusive. A permutation-symmetrized basis is proposed as a route to larger systems.
Error found: Minimizing changes between successive truncations is presented as a ground-state selection principle without a justification establishing convergence or accuracy. The contour-defined Hamiltonian matrix requires a generalized eigenproblem with the indefinite overlap metric, which is absent from the displayed spectral equations even though an implicitly corrected implementation remains possible. The stated identity between a bare-mass second derivative of the ground-state energy and the inverse squared mass gap fails even for a free single-site oscillator. The proposed permutation-symmetric basis does not generally span the zero-momentum sector of a nearest-neighbor lattice Hamiltonian. The appendix's finite-temperature entropy formula equals T² times the derivative of its stated free pressure rather than the correct entropy.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 26 pages, 11 figures