arXiv:2610.23681
Impact of Irrep Truncations on Quantum Simulations of QCD
Patrick Draper, Drishti Gupta
Abstract
Quantum simulations of lattice gauge theories with continuous gauge groups require a choice of basis (how states of the gauge theory map to the qubit computational basis) and truncation scheme (which of the infinitely many gauge theory basis states should be omitted in the map to the finite dimensional qubit Hilbert space.) The most common choices are adapted to an electric basis for the gauge theory, where the basis states for each link are associated with irreducible representations of the gauge group. Electric bases and truncations are generally advantageous for preserving gauge symmetries and simulating the strong coupling regime, but their effectiveness in the weak coupling regimes of physical interest are less well understood. In this work, we compute relative probabilities $p(R')/p(R)$ for finding a link in the irreps $R,R'$ in the ground state of SU(3) lattice gauge theory using Euclidean Monte Carlo methods and weak-coupling lattice perturbation theory. The observables that compute these relative probabilities are constructed by introducing seams into an anisotropic lattice, and by a similar construction we are also able to measure relative probabilities of different site singlets (ways that electric flux braids together into a gauge singlet) at each vertex. With these methods we can quantify, for example, the importance of higher electric flux states in the weak coupling ground state. These results directly inform what truncations are needed to accurately capture non-abelian gauge dynamics near the continuum limit.
AI Summary
Quantum simulations of $SU(3)$ lattice gauge theory require finite-dimensional link Hilbert spaces, but the electric-basis truncations needed at physically relevant couplings are not well quantified. This paper studies the ground-state populations of link irreducible representations and vertex singlets in pure-gauge Kogut–Susskind theory, to inform truncation choices. A Euclidean Monte Carlo construction introduces doubled links, or “seams,” allowing electric-basis projectors to be measured through characters and representation matrices. It yields probabilities relative to the trivial representation or trivial vertex configuration. Measurements use Wilson and heat-kernel temporal actions on $4^3\times L_t$ lattices, with $L_t=24\xi$, and extrapolate to the Hamiltonian limit $\xi\to\infty$. The respective leading extrapolation corrections scale as $1/\xi$ and $1/\xi^2$. At $\beta=6$ ($g=1$), two-index and several three-index representations have appreciable populations, while the four-index $27$ is strongly suppressed. These results motivate a candidate cutoff $p+q\leq3$ at this coupling. Higher representations become more populated near $\beta\simeq7$. Only a few dominant vertex singlets, constructed from fundamental and antifundamental representations, are resolved. Leading-order lattice perturbation theory gives $\langle E^2\rangle\simeq6.37/g^2$, versus Monte Carlo values around $2.5$–$2.7$ at $g=1$; tadpole improvement reduces but does not remove the discrepancy. Limitations include small spatial volume, noise-limited higher-representation and singlet measurements, truncated normalization sums, and sampling difficulties at large anisotropy. Dynamical fermions are absent, and the proposed cutoff is not validated through truncated-Hamiltonian spectra or real-time observables.
Central claim may not hold: With eight simultaneously open link seams, the described character estimator measures an irrep probability ratio conditioned on the other seven links being trivial, not the claimed unconditional single-link ratio, unless an unverified factorization holds. The displayed site-singlet kernel has the wrong group-matrix ordering relative to the explicitly defined basis overlaps and projector, so its derivation does not establish the reported singlet probabilities. Selected local vacuum probabilities on a small spatial lattice do not establish that the recommended p+q≤3 truncation faithfully reproduces low-energy continuum observables. The perturbative correlator calculation truncates spatial momenta, not irreducible representations, so its continuum agreement cannot validate a finite representation cutoff.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 38 pages, 7 figures, 2 tables