Lattice and Hamiltonian QFT

Lattice and Hamiltonian methods replace a formal continuum field theory by finite, computable regulators. Their strength is that symmetries, measures, Hilbert spaces, algorithms, and observables can be tested nonperturbatively. Their discipline is equally important: a result belongs to the target QFT only after the regulated definition, parameter tuning, operator matching, uncertainty model, and all relevant limits have been demonstrated. The compact gauge construction of Wilson 1974, pp. 2445–2459 and the cutoff-expansion framework of Symanzik 1983, pp. 187–204 exemplify these finite-definition and continuum-control requirements.

This volume connects Euclidean spacetime lattices, finite-volume amplitudes, equal-time and light-front Hamiltonians, basis truncation, tensor networks, and quantum simulation. These formulations do not share the same finite cutoff. They become comparable only through matched renormalized observables and a common continuum target.

Helpful background. Regulated bosonic field integrals prepare the Euclidean measure; Hamiltonian field theory prepares the equal-time routes; and probabilistic convergence and limit theorems prepare sampling and inference. These are route-specific aids, not entrance requirements for reading this overview.

One target, several finite formulations

Every route begins by declaring a target QFT and observable. A regulator then defines a finite problem, but the map back to the target includes tuning, matching, independent cutoff variations, and validation. The solid arrows in the figure are required logical steps; the dashed stops mark common ways a finite calculation can fail to support its advertised claim.

Different finite formulations can support one continuum statement only after their conventions, renormalized observables, regulator axes, and limit orders are matched. Internal numerical convergence is necessary but does not alone establish the target QFT. The diagram is schematic and not to scale.

The volume is organized into four recurring passes:

1. Define: target, finite variables, measure or Hilbert space, action or Hamiltonian, state, observables, boundaries, and conventions.
2. Compute: construct transfer, gauge, fermion, sampling, variational, tensor, or quantum-evolution machinery and verify its exact identities.
3. Infer: isolate states, renormalize operators, propagate covariance, and remove volume, spacing, basis, bond, time, or encoding cutoffs.
4. Certify: compare with analytic fixtures, independent algorithms, alternative regulators, Ward identities, and held-out observables; then state the strongest conclusion actually supported.

The thirteen chapters

Chapter	Central question	Result you should be able to defend
1. Lattice Regulators and Continuum Targets	What finite data and ordered limits define a lattice QFT?	A regulator specification tested by dispersion, symmetry, positivity, and scaling
2. Lattice Observables and Continuum Inference	How do correlators become spectra, matrix elements, and continuum observables?	A covariance-preserving chain from operator basis through renormalization and extrapolation
3. Lattice Gauge Theory	Which gauge statement is exact at finite spacing, and which needs continuum evidence?	Correctly oriented links and actions plus distinct loop, topology, flow, and gauge-fixed tests
4. Lattice Fermions and Chirality	How do formulations trade doubling, chirality, locality, flavor, and cost?	A formulation choice that keeps anomaly, determinant, index, and chiral-gauge claims separate
5. Sampling Algorithms for Lattice Fields	Does a transition kernel sample the intended measure?	Invariance, reversibility or balance, ergodicity, acceptance, and exact-reference tests
6. Statistical Inference and Error Budgets	What uncertainty follows from a realized correlated analysis?	Defensible equilibration, autocorrelation, covariance, fit, bias, and reproducibility evidence
7. Finite-Volume Spectra and Amplitudes	When can discrete levels determine an infinite-volume amplitude?	A branch-specific quantization analysis with range, channel, irrep, and analytic assumptions explicit
8. Finite Density and Sign Problems	Is a complex-measure method correct, overlapping, and scalable?	Quantified severity and cross-method reliability without a universal-cure claim
9. Hamiltonian Lattice Field Theory	Which physical Hilbert space and regulator limits define real-time QFT?	A constrained Hamiltonian calculation cross-validated against Euclidean data
10. Light-Front QFT	Which advantages survive zero-mode, current, and counterterm control?	A regulated light-front bound-state problem with every cutoff and restoration test visible
11. Hamiltonian Truncation and Variational Methods	What do omitted states induce, and how is convergence certified?	Renormalized truncated operators, cutoff sequences, residuals, and held-out benchmarks
12. Tensor-Network QFT	Which tensor geometry represents the state, and which errors remain independent?	Separate bond, contraction, optimization, local-dimension, volume, time, and continuum controls
13. Quantum Simulation of QFT	What evidence connects an encoded experiment or algorithm to a QFT observable?	A target-to-measurement chain with verification, resources, and a calibrated claim ceiling

The sidebar order reveals dependencies but is not compulsory. A Euclidean spectroscopy calculation may stop after Chapters 1–6; a finite-volume scattering problem adds Chapter 7; a real-time calculation can enter at Chapter 9 once its foundational regulator is clear.

The target–regulator–observable–limit table

Use this table before starting any substantial calculation.

Part	Information that must be fixed	Independent check	Stop condition
Target QFT	Dimension, fields, interactions, state or ensemble, global structure	Recover a known symmetry, pole, Ward identity, or weak-coupling limit	Only a named numerical model is given
Finite regulator	Geometry, variables, measure or Hilbert representation, constraints, boundaries	Exact small system, free kernel, or algebraic identity	A cutoff or boundary sector is hidden
Bare definition	Action or Hamiltonian, domains, all parameters and normalization	Hermiticity/positivity, dimensions, exact invariance	Stability or physical sector is unknown
Tuning	Independent renormalized conditions along the cutoff sequence	Closure with an unused ratio or observable	Bare parameters are held fixed without justification
Observable map	Operator basis, state isolation, subtraction, mixing, scale and scheme	Alternative operator, Ward identity, or matching route	Bare finite-cutoff number is labeled physical
Estimator	Sampling or evolution algorithm, covariance, solver and stopping errors	Exact fixture, negative control, refinement	Correctness is inferred from plausible output
Limits	Spacing, volume, mass, local dimension, basis, bond, time, noise, harmonic resolution	Vary axes separately and reverse an iterated order where relevant	Several axes move on one unidentified diagonal
Final evidence	Model stability, held-out checks, alternative regulator or formulation	Reproduce an observable not used in tuning	Claimed universality rests on one sequence

At finite cutoff, the regulated theory may be exactly defined and exactly solved. The continuum interpretation remains an inference until the last rows are closed.

Choose a route by the physical question

A scalar observable from lattice to continuum. Begin with the regulator specification, derive a free dispersion and correlator, then use operator bases, scale setting, sampling, and continuum extrapolation. Stop if a zero mode, tuning condition, autocorrelation scale, or operator normalization is missing.

A gauge-theory spectrum or matrix element. Construct links and plaquettes, choose a fermion formulation, generate a validated ensemble, and carry the operator through the full inference and error-budget chain. Loop, center, topology, and flow diagnostics are not interchangeable.

An amplitude from a finite box. Start with finite volume as a controlled deformation, identify the elastic, coupled-channel, current, or three-body branch, and use its exact symmetry and analyticity conditions. Stop before applying a two-body short-range formula to massless exchange, omitted channels, or an unresolved three-body threshold.

A finite-density result. Define the chemical-potential deformation and compute an average phase or another quantitative severity measure. Choose reweighting, expansion, canonical, density-of-states, dual, complex-Langevin, or contour methods only after their correctness and overlap requirements are stated. Close with at least two methods whose dominant assumptions are genuinely different.

A real-time observable. Enter through regulated Hamiltonian field theory, construct the physical sector, prepare the state, refine time evolution, and separate the finite observation window from the spatial and local-Hilbert cutoffs. Continue to Hamiltonian truncation, tensor networks, or quantum simulation according to the computational representation.

A cross-formulation comparison. Fix one dimensionless observable, match scale and operator conventions, make each formulation reproduce its own finite-regulator reference, and share only the continuum intercept in the final fit. Equal-time, Euclidean, light-front, tensor, and encoded finite systems must not be identified before their distinct limits are removed.

Readiness checks

Capability	Inspectable task	Sufficient answer	Where to repair
Regulator versus target	Let $a\to0$ with fixed site count and find $L=Na$	You identify a shrinking box, not an infinite-volume continuum limit	Lattice geometry and limits
Euclidean inference	Explain why one effective-mass plateau need not be the ground state	You name overlap, gap, covariance, and basis/window tests	Operator bases and excited states
Gauge covariance	Transform every factor around one plaquette	Intermediate site matrices cancel; the product is conjugated at its base point	Links and plaquettes
Markov statistics	Convert $N$ correlated samples into an effective sample size	You use the observable’s integrated autocorrelation time and state the resampling unit	Autocorrelation times
Physical Hilbert space	Count states of a small gauge graph two ways	Projector rank and independent-cycle counting agree	Physical gauge Hilbert spaces
Multiple regulators	Compare two values obtained at unequal finite cutoffs	You match the target observable and extrapolate each regulator’s own errors before comparing	Hamiltonian–Euclidean validation

Conventions that travel with results

The site-wide conventions use natural units and the (+---) Lorentzian metric. Euclidean pages state their Wick rotation and time boundary conditions; Hamiltonian pages state continuous-time normalization. Every lattice result additionally records spacings, extents, boundaries, field normalization, Fourier factors, lattice momentum, and zero-mode treatment.

Gauge pages carry the group and global form, representation, generator trace, link and plaquette orientation, boundary charge, and determinant or Pfaffian phase. Hamiltonian and light-front pages carry the local Hilbert or basis cutoff, physical-sector definition, longitudinal compactification and zero modes, time step and window. Tensor and quantum pages add bond, contraction, optimization, encoding, padding, preparation, measurement, mitigation, and resource cutoffs.

When translating between conventions, transform the complete package—fields, measures, couplings, generators, states, and observables—and close the translation with an invariant benchmark. “Up to conventions” is not evidence.

Evidence and stopping rules

A reliable conclusion names its level:

• finite-regulator identity: exact algebra, enumeration, diagonalization, or analytic formula for the stated finite problem;
• controlled regulator result: numerical or asymptotic calculation with the relevant algorithmic and cutoff errors varied;
• continuum-trending result: several matched spacings or cutoffs with a justified extrapolation model, still subject to declared limitations;
• continuum observable: regulator, volume, operator, and uncertainty conditions closed to the stated precision;
• current capability claim: a dated statement that belongs with changing evidence, not with a stable theoretical definition.

These labels are not interchangeable. A passing static build, a visually smooth plot, or an exact finite-system benchmark cannot establish scientific review, rights clearance, universality, quantum advantage, or a continuum theorem.

Exercises

1. Design a minimal Cartesian convergence study for an observable depending on spatial spacing $a$, volume $L$, and local Hilbert dimension $d$.

Solution

Choose at least two or three $d$ values at several fixed $(a,L)$ pairs to bound or extrapolate the local cutoff; repeat several volumes at fixed physical spacing; then use several spacings along matched renormalized conditions. Include at least one crossed point to reveal compensation that a single diagonal sequence $d(a,L)$ could hide.

2. Two methods agree because both use the same analytic continuation ansatz. Is this independent cross-validation?

Solution

Not for errors controlled by that ansatz. The comparison can test implementation differences, but the shared assumption remains one correlated source. Add a method anchored in a different representation or an exact/sign-free regime.

References

• Symanzik, K. (1983). Continuum limit and improved action in lattice theories. I. Principles and $\phi^4$ theory. Nuclear Physics B, 226, 187–204. DOI.
• Wilson, K. G. (1974). Confinement of quarks. Physical Review D, 10, 2445–2459. DOI.

Further reading

• Gattringer, C., and Lang, C. B. (2010). Quantum Chromodynamics on the Lattice: An Introductory Presentation. Springer. DOI.
• Kogut, J. B. (1979). An introduction to lattice gauge theory and spin systems. Reviews of Modern Physics, 51, 659–713. DOI.
• Montvay, I., and Münster, G. (1994). Quantum Fields on a Lattice. Cambridge University Press. DOI.