Evidence, Validation, and Nonperturbative Status

This chapter is the volume’s comparison layer. It helps a reader turn a broad statement such as “several nonperturbative methods find a mass gap” into a set of testable propositions: which observable each method computes, how the outputs are translated, which assumptions are shared, what evidence class each implication earns, and what theorem, calculation, or dated assessment would strengthen the result. It does not replace the scientific chapters or convert agreement into proof.

Helpful background. What “nonperturbative” means in a declared problem supplies the minimum language of observable, regime, control, order of limits, and falsifier. Readers may also enter from any chapter whose result they want to compare.

Enter this chapter

Choose an entry by the question you already have.

• If a paper or calculation uses words such as exact, controlled, evidence, or open, begin with Claim Status Across Nonperturbative Methods.
• If two results share a name but arise from different correlators, regulators, ensembles, or kinematics, begin with Translating Observables Across Analytic and Regulated Methods.
• If several methods agree and you need to know how independent that agreement is, begin with Correlated Evidence, Independence, and Triangulation.
• If the question is what has actually been constructed or proved, and which limit remains open, begin with Rigorous Status, Construction, and Open Problems.

The chapter owns stable inference rules and representative comparisons. Method-specific derivations remain in their scientific chapters, regulated extraction belongs to Lattice and Hamiltonian QFT, theorem proofs belong to Mathematical QFT, and time-sensitive assessments belong to Research.

Check your preparation

No single prerequisite blocks the chapter. Use these tasks to select a repair route.

Can you write the target as a proposition? Name the theory, state, observable, property, and order of limits. If yes, enter the claim-status page. If not, repair the distinction between method and evidence in What Does Nonperturbative Mean?.

Can you distinguish a pole, screening rate, and finite-volume level? Explain which correlator and limit defines each. If yes, enter the translation page. If not, review spectral decomposition of two-point functions.

Can you name a shared systematic error? Give an example in which two calculations use the same scale, ansatz, ensemble, or matching coefficient. If yes, enter the correlated-evidence page. If not, compare the control fields in the claim-and-control table.

Can you separate a regulated model from a continuum construction? State one extra estimate needed to remove a cutoff or take infinite volume. If yes, enter the rigorous-status page. If not, use existence, construction, reconstruction, and continuum claims as a theorem-language repair.

Choose a route

Reader goal	Suggested route	Capability at the end
First systematic encounter	Claim status → observable translation → correlated evidence → rigorous status	Decompose a composite claim and identify its weakest implication
Compare two mass determinations	Observable translation → correlated evidence → claim status	Decide whether the masses are commensurate and how much their agreement adds
Review a confinement or mass-gap claim	Claim status → corresponding Chapter 7 comparison → rigorous status	Separate diagnostic, controlled mechanism, numerical evidence, and open theorem
Validate a functional calculation	Chapter 9 validation → observable translation → correlated evidence	Keep closure, branch, continuation, and shared ansatz uncertainties distinct
Assess an exact-model statement	Chapter 11 exact-data chain → claim status → rigorous status	Distinguish algebraic exactness, completeness, locality, and construction
Re-enter current research	Stable page for definitions and inference → Nonperturbative Gauge Dynamics research guide	Carry a precise observable and upgrade question into a dated assessment

The arrows in this table are suggested reading order, not new hard prerequisites. A focused reader may use one leaf and return to the originating scientific chapter.

The conceptual dependency

Every comparison proceeds through four questions in order.

1. What proposition is being supported? Break a broad conclusion into definition, method output, translation, limiting statement, and inference.
2. Are the observables the same? Match operator, state, normalization, scheme, regulator, continuation, volume, and kinematics.
3. Which failure modes are shared? Draw the dependency graph and separate calibration, common identities, method-specific systematics, and held-out checks.
4. What is the terminal status? State the exact hypotheses and conclusion of any theorem or construction, and name the first unproved or uncontrolled edge.

Changing the order causes predictable errors. Combining central values before observable translation gives precision about the wrong target. Assigning evidence labels before exposing shared assumptions can count one systematic effect several times. Asking whether something is “proved” before defining the object produces a status answer with no stable proposition.

Chapter guide

1. Claim Status Across Nonperturbative Methods defines proposition-level classes—identity, theorem, construction, controlled approximation, regulated evidence, phenomenological inference, conjecture, and open problem. Its worked example decomposes the four-dimensional Yang–Mills mass-gap claim and its exact-status comparison is the chapter’s common reference.
2. Translating Observables Across Analytic and Regulated Methods gives the map from each method’s native output to one renormalized observable. It compares pole masses, Euclidean decay rates, screening lengths, transfer-matrix gaps, and finite-volume levels, with a free-scalar round trip and a regulated-spectrum workflow.
3. Correlated Evidence, Independence, and Triangulation identifies shared theory inputs, calibration data, truncations, matching, and continuations. It derives the effect of correlation on a combined estimate and designs common falsifiers and genuinely discriminating held-out checks.
4. Rigorous Status, Construction, and Open Problems follows the chain from finite regulator through uniform estimates, continuum and volume limits, reconstruction, and a physical theorem. Constructive scalar models, factorizing S matrices, compact three-dimensional U(1), and four-dimensional Yang–Mills show why results must remain attached to their dimensions and hypotheses.

A mass-gap thread across the chapter

Keep one target fixed: a positive gap above the vacuum in the gauge-invariant infinite-volume spectrum of a declared continuum theory.

On the first leaf, decompose this into a definition, regulated calculation, continuum and volume translation, and construction theorem. On the second, distinguish an infinite-volume pole or spectral edge from a Euclidean screening rate and a box energy. On the third, ask whether analytic, functional, and regulated determinations share renormalization, scale setting, ansätze, or calibration data. On the fourth, specify which models have constructions or controlled theorems and which continuum statement remains open.

Three invariant checks travel with the thread:

• the operator is gauge invariant and has the declared quantum numbers;
• the continuum and infinite-volume limits are explicit and their order is tested; and
• the conclusion never becomes stronger than its weakest translation or existence step.

The thread stops before a changing literature judgment. Current numerical precision, disputed mechanisms, and new proof claims require dated sources and independent review in Research.

Convention and uncertainty bridge

The chapter inherits the site’s (+---) metric, spectral, Euclidean, and $+i0$ conventions. Its additional notation is minimal:

• $D_i$ denotes the native output of method $i$;
• $R_i$ denotes renormalization and regulator removal;
• $C_i$ denotes continuation, matching, and remaining physical limits;
• $O$ is the common renormalized target; and
• a graph arrow means logical or data dependence, never an unstated evidence weight.

Uncertainty components retain their origin. Sampling error, discretization, finite volume, truncation, scale setting, analytic continuation, and model discrepancy are not collapsed merely because a final covariance exists. Shared components appear once in the dependency graph and as off-diagonal covariance when they can be quantified.

For a theorem, quote primitive objects, hypotheses, topology of convergence, and conclusion. For a computational result, quote regulator, estimator, data or ensemble, extrapolations, and systematic variations. “Up to conventions” and “within errors” are incomplete unless the translation and error components are visible.

Synthesis: one conclusion, four gates

A cross-method conclusion is ready for reuse only after four gates are passed.

Proposition gate. The theory, state, observable, property, regime, limits, and falsifier are explicit.

Translation gate. Each method reaches that observable through a reproducible normalization, matching, continuation, and limit map.

Independence gate. Shared ancestors and calibration data are exposed; the synthesis includes at least one check with an independent failure opportunity.

Status gate. Every use of exact, controlled, constructed, evidence, or open modifies a specific proposition under stated hypotheses.

Passing these gates does not guarantee truth. It produces a bounded claim whose failure and upgrade paths are intelligible. That is the appropriate endpoint of a stable educational chapter.

Review the chapter

1. Retrieval: Define a controlled approximation, regulated numerical result, constructive result, and open problem without ranking them on one scale.
2. Translation: Give conditions under which a zero-temperature Euclidean decay exponent equals a stable pole mass, then remove one condition and describe the failure.
3. Correlation: For two equal-variance estimates with correlation $\rho$, derive the variance of their mean and interpret $\rho=1$.
4. Application: Decompose one confinement, mass-gap, transseries, or exact-S-matrix statement into propositions and label each separately.
5. Adversarial check: Identify a datum used in calibration and replace it with a held-out observable that can distinguish two methods.
6. Theorem boundary: Explain why a positive finite-volume transfer-matrix gap is not the four-dimensional Yang–Mills mass-gap theorem.
7. Exit criterion: Name the one result—new matching calculation, limit study, independent benchmark, construction, or proof—that would most change the status of your chosen claim.

One model answer for item 6

A finite-volume regulated Hamiltonian has a discrete spectrum, and its first excitation can remain positive for purely kinematic reasons. The Yang–Mills problem asks for a nontrivial continuum theory on $\mathbb R^4$ and a positive gap in its physical infinite-volume spectrum. Regulator removal, Hilbert-space construction, model identification, and a gap bound uniform in volume are additional propositions.

References

• Glimm, James, and Arthur Jaffe. Quantum Physics: A Functional Integral Point of View. 2nd ed. New York: Springer, 1987. DOI.
• Lyons, Louis, Duncan Gibaut, and Peter Clifford. “How to Combine Correlated Estimates of a Single Physical Quantity.” Nuclear Instruments and Methods in Physics Research Section A 270, no. 1 (1988): 110–117. DOI.
• Osterwalder, Konrad, and Robert Schrader. “Axioms for Euclidean Green’s Functions.” Communications in Mathematical Physics 31, no. 2 (1973): 83–112. DOI.