Lattice and Hamiltonian Field Theory

Lattice and Hamiltonian field theory builds regulated many-body systems whose controlled limits answer continuum QFT questions. The field spans Euclidean Monte Carlo, transfer-matrix and Hamiltonian formulations, Hamiltonian truncation, tensor networks, and analog or digital quantum simulation. Its unifying standard is not discreteness but certification: the target theory, regulator removal, observable matching, and full error budget must all be explicit.

Evidence cutoff. 11 August 2026.

Required background. Lattice regulators and continuum targets explains universality and target specification; lines of constant physics and continuum extrapolation supplies the regulator-removal logic.

Helpful background. Hamiltonian continuum validation connects spectra and real time to Euclidean checks, tensor-network error certification separates bond and contraction errors, quantum-simulation resource certification connects encodings to physical precision, and correlated evidence and triangulation prevents double counting.

Lattice and Hamiltonian regulators share a burden of proof

Program	Best-suited targets	Dominant choices	Errors that must converge
Euclidean importance sampling	equilibrium spectra, matrix elements, thermodynamics	action, lattice spacing, volume, boundary conditions, estimator	autocorrelation, finite volume, discretization, renormalization, scale setting
Hamiltonian lattice methods	spectra and real-time evolution	gauge-invariant basis, spatial regulator, temporal treatment	volume, spacing, basis size, time step
Hamiltonian truncation	low-energy data near a solvable theory or CFT	reference basis, energy cutoff, counterterms	omitted-state and induced-operator errors
tensor networks	low-entanglement states and one-dimensional dynamics	ansatz, bond dimension, contraction, symmetry encoding	variational, entanglement, contraction, finite-size, continuum errors
quantum simulation	real-time or finite-density observables inaccessible to sampling	encoding, state preparation, circuit/analog protocol, mitigation	Hilbert-space truncation, Trotter/algorithmic error, noise, measurement, continuum

Wilson’s gauge-invariant lattice construction made confinement and the regulator-to-continuum problem concrete Wilson 1974, foundational lattice-gauge construction. Euclidean lattice QCD has precision domains because several lattice spacings, near-physical masses, finite-volume control, nonperturbative renormalization, and blinded or independent analyses can be combined. The FLAG synthesis makes its quality criteria explicit and does not average calculations that fail them FLAG 2024, current synthesis. That mature culture should be exported cautiously: a single finite-spacing result in a new formulation is a benchmark, not a continuum prediction.

Validation is a sequence of limiting operations

For a Euclidean observable, a schematic target is

$$O_{\mathrm{phys}}=\lim_{a\to0}\lim_{L\to\infty} Z_O(a,\mu)\,O_{\mathrm{lat}}(a,L,m_i,\ldots),$$

with quark-mass tuning and excited-state limits inserted as needed. The order and joint modeling of limits matter. Continuum fits need enough lever arm to distinguish plausible cutoff terms; covariance estimated from finite ensembles can be ill-conditioned; topology can freeze at fine spacing; and model averaging does not rescue a set of uniformly misspecified ansätze.

Hamiltonian and tensor-network calculations replace sampling errors with different ceilings. Increasing bond dimension at one lattice spacing does not test the continuum. Increasing a local Hilbert-space cutoff without renormalizing induced operators can converge to the wrong regulated theory. The few-qubit Schwinger-model experiment is an instructive implementation benchmark Martinez et al. 2016, quantum-simulation demonstration, but reproducing a small exact-diagonalization result demonstrates implementation control only in that regime, not quantum advantage or continuum reach. Reviews of lattice gauge theories on quantum platforms Bañuls et al. 2020, quantum-technologies review and of tensor networks beyond one spatial dimension Magnifico et al. 2025, tensor-network review show how gauge encoding, contraction, and truncation costs delimit extrapolation.

Useful common benchmarks include the Schwinger model, Ising and $O(N)$ critical systems, pure-gauge spectra, finite-volume scattering levels, and thermodynamic equations of state. A benchmark is strongest when the same continuum observable is reached with formulations whose leading systematics differ. Tensor-network studies should extrapolate bond dimension, volume, and lattice spacing separately; stochastic studies should recover integrated autocorrelation times by several windowing or blocking choices. Sign-problem methods should be compared with exact small-volume results and a sign-free overlap regime while reporting cost growth, and finite-volume amplitude extractions should recover a synthetic phase shift across several volumes and irreducible representations before application to new data.

Structural obstructions

Importance sampling fails when the Euclidean weight is not a useful nonnegative probability measure; reweighting then typically suffers exponentially poor signal with volume. Euclidean correlators encode real-time spectra through an ill-posed inverse transform, so fine temporal data and a reconstruction prior do not automatically yield controlled spectral detail. Fermion doubling constrains local lattice discretizations, although formulations evade different assumptions at different costs. Tensor-network efficiency depends on entanglement structure, and generic real-time entanglement growth limits reachable times. Fault-tolerant quantum algorithms may change asymptotic access but still require state preparation, observable extraction, resource accounting, and a regulator-to-continuum program.

Entering the field

Begin with a benchmark whose continuum value is known or independently precise. Write down every limit before generating data, predefine stability tests, and retain enough raw metadata to recompute autocorrelations and covariance. The computational field theory pathway supplies preparation; reproduce and validate a result supplies the evidence workflow.

This guide does not catalog hardware platforms or software packages, and it does not treat a method’s access to real time or density as equivalent to controlled precision there. Nuclear many-body methods and generic quantum-computing algorithms appear only when they bear directly on a field-theory regulator.

Evidence scope and related assessments

The finite search used INSPIRE, arXiv, official FLAG and PDG resources, journal/DOI records, and citation chaining, with explicit searches for sign, continuum, and reproducibility failures. Sources public through 11 August 2026 were eligible. Reassess when a new formulation demonstrates a continuum observable beyond existing reach, a benchmark fails independent reproduction, or hardware/systematic scaling changes a resource conclusion.

Continue to real-time continuum dynamics, finite-density QCD, Euclidean lattice inference, or Hamiltonian, tensor-network, and quantum-simulation methods.

References

• Y. Aoki et al. (Flavour Lattice Averaging Group), “FLAG Review 2024,” Physical Review D 113 (2026) 014508. DOI.
• M. C. Bañuls et al., “Simulating Lattice Gauge Theories within Quantum Technologies,” European Physical Journal D 74 (2020) 165. arXiv.
• G. Magnifico et al., “Tensor Networks for Lattice Gauge Theories beyond One Dimension,” Communications Physics 8 (2025) 322. DOI.
• E. A. Martinez et al., “Real-Time Dynamics of Lattice Gauge Theories with a Few-Qubit Quantum Computer,” Nature 534 (2016) 516–519. DOI.
• K. G. Wilson, “Confinement of Quarks,” Physical Review D 10 (1974) 2445–2459. DOI.