Existence, Construction, Reconstruction, and Continuum Claims

Existence, construction, reconstruction, removal of a regulator, and passage to infinite volume are different mathematical achievements. A theorem can establish any one of them without establishing the others. The distinction matters most in QFT because the primitive finite-cutoff object is often elementary, while the desired continuum field is a random distribution or an unbounded-operator theory obtained only after several noncommuting limits and a reconstruction theorem.

Required background. Theorem-First Claim Records fixes objects and quantifiers, while Limits, Completeness, and Modes of Convergence supplies the topological language. Helpful background. Regulators, Cutoffs, and Continuum Limits gives the physical interpretation, and Interacting Fields, Asymptotic Observables, and Effective Descriptions separates exact fields from perturbative and effective descriptions.

Six claims that should not share one verb

Let $X_{a,L}$ denote regulated data depending on ultraviolet scale $a$ and volume scale $L$. The following statements have distinct quantifiers and codomains.

Abstract existence. There is an object $X\in\mathcal X$ satisfying $P(X)$. A compactness argument may prove this without providing a computable formula or a canonical choice.

Explicit construction. A specified procedure produces $X$. Typically one gives approximants $X_n$, proves they are defined, and proves $X_n\to X$ in a named topology. Construction implies existence; existence need not imply a useful construction.

Ultraviolet removal. At fixed physical volume, a family $X_{a,L}$ converges as $a\downarrow0$ after counterterms and observable normalizations have been fixed:

$$X_{a,L}\longrightarrow X_L \quad\text{in }\mathcal T_{\mathrm{UV}}.$$

This is not the infinite-volume limit.

Thermodynamic limit. A family $X_L$ converges as $L\to\infty$ in a topology $\mathcal T_{\mathrm{IR}}$. Boundary-condition independence or uniqueness of the limiting phase is an additional conclusion, not part of bare subsequential existence.

Continuum scaling limit. Lattice observables are rescaled and tuned while $a\downarrow0$, possibly with $L=L(a)\to\infty$. The result specifies the renormalization trajectory, field-strength normalization, test objects, and joint or iterated limit. It is stronger than saying that a formal continuum Lagrangian can be written.

Reconstruction. Starting from data $D$ that already exist and satisfy hypotheses $H$, a theorem constructs an object $R(D)$ in another framework and proves properties $Q(R(D))$. Reconstruction proves

$$D\in\mathcal D_H\Longrightarrow R(D)\in\mathcal Q,$$

not that some $D\in\mathcal D_H$ exists. In Osterwalder–Schrader reconstruction, the input is a complete compatible sequence of Euclidean Green functions satisfying symmetry, covariance, reflection positivity, regularity or growth, and clustering conditions. The theorem constructs Wightman data; it does not manufacture the Euclidean sequence.

The implication structure and its nonconverses

A typical constructive chain is

$$\begin{gathered} \text{finite-cutoff measures and uniform estimates} \Longrightarrow \text{tightness},\\ \text{tightness}+\text{identification of every cluster point} \Longrightarrow \text{continuum Schwinger data},\\ \text{Schwinger data}+\text{OS hypotheses} \Longrightarrow \text{relativistic Wightman theory}. \end{gathered}$$

Every arrow has a converse boundary.

• Tightness gives subsequences, not uniqueness of the limit.
• Convergence of each fixed $n$-point function does not by itself give a compatible hierarchy or a probability measure.
• A measure determines moments only when those moments exist; moments determine the measure only under a determinacy condition.
• Wightman reconstruction from a positive hierarchy is uniqueness up to unitary equivalence of the reconstructed cyclic realization, not uniqueness of a Lagrangian or of all field coordinatizations.
• Existence of a continuum limit does not imply that it is interacting, has a mass gap, or is the only continuum limit of the microscopic family.
• A perturbative construction in $\mathbb C[[\lambda]]$ proves coefficientwise algebraic statements, not convergence at fixed nonzero $\lambda$.

The corrected Osterwalder–Schrader paper makes the logical direction explicit: its new conditions are sufficient for analytic continuation to a relativistic field theory, and the proof begins with an already given sequence of Euclidean functions Osterwalder and Schrader 1975, §§ III–IV, pp. 285–289.

Compatibility is an existence obligation

Suppose $S_2$ is a tempered, Euclidean-invariant, reflection-positive two-point distribution. This does not yet define an interacting field. One needs distributions $S_n$ for every $n\geq0$ with a common set of axioms. At minimum, the family must have consistent permutation symmetry, covariance, positivity, and restriction behavior. If the $S_n$ are to be moments of a measure $\mu$ on a distribution space, then for test functions $f_1,\ldots,f_n$,

$$S_n(f_1\otimes\cdots\otimes f_n) =\int\Phi(f_1)\cdots\Phi(f_n)\,d\mu(\Phi)$$

must be jointly realizable. Choosing each $S_n$ independently can violate positivity of polynomial expectations or projective consistency. A convergent sequence of two-point functions therefore proves only convergence of covariances unless Gaussianity or an independently constructed higher hierarchy is supplied.

Construction of the P(φ)₂ model

The two-dimensional polynomial scalar field shows how the claims fit together. Let $\mu_C$ be the massive Gaussian free-field measure on $\mathcal S'(\mathbb R^2)$ with covariance $C=(-\Delta+m^2)^{-1}$. For a bounded region $\Lambda$ and a polynomial $P$ bounded below, introduce a mollified field $\phi_\kappa$ and the Wick-ordered interaction

$$V_{\kappa,\Lambda}(\phi) =\int_\Lambda :P(\phi_\kappa(x)):_C\,d^2x,$$

then define

$$d\mu_{\kappa,\Lambda} =Z_{\kappa,\Lambda}^{-1}e^{-V_{\kappa,\Lambda}}d\mu_C.$$

This formula by itself is only a regulated finite-region definition. The constructive work separates four further steps.

Ultraviolet construction at finite volume

Wick ordering subtracts the divergent Gaussian self-contractions. One proves that $V_{\kappa,\Lambda}$ converges in suitable $L^p(\mu_C)$ spaces and that the normalized densities have the integrability needed to define a cutoff-free finite-volume measure $\mu_\Lambda$. The theorem is not ordinary pointwise convergence of $\phi_\kappa(x)$—the field remains a distribution.

Infinite-volume control

Uniform correlation inequalities, stability estimates, cluster expansions, or statistical-mechanical compactness are then used as $\Lambda\nearrow\mathbb R^2$. Boundary conditions and phase selection must be stated. The classical statistical-mechanics treatment develops this passage, including correlation inequalities and infinite-volume states, in Guerra, Rosen, and Simon 1975, Part I, Chapters I–III, pp. 111–189 and Part II, Chapters IV–VII, pp. 191–259.

Schwinger-function bounds and OS properties

From the limiting measure one defines the complete hierarchy of moments. Euclidean covariance, reflection positivity, regularity bounds, and clustering are separate propositions. Reflection positivity is inherited only because the approximation and limiting procedure preserve the relevant positive-time quadratic form; weak convergence alone would not suffice without control of the test algebra and integrability.

Relativistic reconstruction and particle structure

Once the input hierarchy meets the reconstruction theorem’s hypotheses, one obtains a Hilbert space, vacuum, positive-energy representation, and operator-valued fields. For small-coupling $P(\phi)_2$, the subsequent Wightman and spectral analysis proves more: the Wightman axioms and isolated vacuum and one-particle mass eigenvalues Glimm, Jaffe, and Spencer 1974, pp. 585–632. Those particle conclusions are not contained in the finite-volume measure or in reconstruction alone.

The broader construction status is organized on Rigorous Status, Construction, and Open Problems. The lesson here is the logical factorization: ultraviolet removal, infinite volume, OS verification, reconstruction, and spectral analysis are individually named obligations.

An adversarial two-point sequence

Let $C_n\to C$ in $\mathcal S'(\mathbb R^d\times\mathbb R^d)$, with every $C_n$ a positive covariance. Define only $S_{2,n}=C_n$. It is tempting to announce a limiting QFT because the propagator converges.

The claim fails at the next arity. No $S_{4,n}$ has been supplied, so there is no compatible sequence of moments and no stated measure. One can complete the same $C_n$ in at least two conceptually different ways: as a Gaussian hierarchy with $S_4$ given by Wick pairings, or, when a positive non-Gaussian measure with that covariance exists, with a nonzero connected four-point function. The two-point limit does not choose between them.

The strongest surviving statement is:

The covariance distributions converge to $C$ in the specified topology.

To strengthen it to a Gaussian field, declare Gaussianity and construct the characteristic functional $Z(f)=\exp[-C(f,f)/2]$, then verify continuity and positivity. To strengthen it to an interacting QFT, construct and control the entire hierarchy or measure and prove the reconstruction hypotheses.

Independent checks

Gaussian reconstruction check. For a claimed centered Gaussian limit, compute the four-point function independently from the characteristic functional and confirm

$$S_4(f_1,f_2,f_3,f_4) =\sum_{\text{three pairings}}C(f_i,f_j)C(f_k,f_l).$$

Any nonzero fourth cumulant contradicts the Gaussian claim.

Order-of-limits check. Bound the finite-volume remainder uniformly in the ultraviolet parameter before interchanging $\kappa\to\infty$ with $\Lambda\nearrow\mathbb R^2$. If the bound deteriorates with $\kappa$, the interchange is not proved.

Axiom-survival check. Apply reflection positivity at every regulator and pass to the limit using the exact convergence mode. Positivity of unrelated pointwise kernels is not a substitute.

Nontriviality check. Evaluate a connected four-point functional at separated test functions. Vanishing is consistent with a generalized free field; nonvanishing requires a proof with its sign and normalization controlled.

Common pitfalls

Calling a subsequential limit “the limit.” Compactness supplies at least one cluster point. Full convergence needs uniqueness of every cluster point, often through correlation identities or boundary-condition-independent estimates.

Treating reconstruction as regulator removal. Reconstruction changes mathematical presentation after the Euclidean data exist. It does not prove that a lattice or mollifier family converges to those data.

Equating nonzero with interacting. A Gaussian field with nonzero covariance is a nonzero continuum QFT. Interaction requires a criterion such as a nonvanishing higher truncated correlation, nontrivial scattering, or another precisely defined structure.

Exercises

1. Classify three statements. Classify: (a) “there is a subsequence $\mu_{n_k}$ converging weakly”; (b) “the limiting Schwinger functions satisfy OS hypotheses, hence determine a Wightman theory”; (c) “the coefficients of the local $S$-matrix are defined to every order in $\lambda$.”

Solution

(a) is subsequential existence obtained from compactness, not uniqueness. (b) is reconstruction from already existing Euclidean data. (c) is a formal perturbative construction in a power-series ring; without summability or convergence it is not a nonperturbative theory at fixed $\lambda$.

2. Find the missing theorem. Assume $S_{n,a,L}\to S_n$ for every $n$ when $L\to\infty$ first and $a\downarrow0$ second. What is needed to claim a joint limit along arbitrary $L(a)\to\infty$?

Solution

One needs uniform control of the finite-volume error as $a\downarrow0$, or an equivalent joint compactness and uniqueness theorem. Iterated convergence permits the scale at which $L$ is “large enough” to depend uncontrollably on $a$; an arbitrary path can therefore miss the iterated limit.

3. Reconstruction is conditional. Write the logical form of OS reconstruction and explain why it cannot prove existence of $P(\phi)_2$ by itself.

Solution

Its form is: for every complete Schwinger hierarchy $S$ in the stated regularity class, if $S$ satisfies the Euclidean axioms and growth condition, then there exists a Wightman realization $R(S)$, unique in the stated cyclic sense. The theorem contains no existential premise asserting that the interacting $P(\phi)_2$ hierarchy exists. Constructive estimates must first produce that input and verify its hypotheses.

References

• Glimm, James, Arthur Jaffe, and Thomas Spencer. “The Wightman Axioms and Particle Structure in the $\mathscr P(\phi)_2$ Quantum Field Model.” Annals of Mathematics 100, no. 3 (1974): 585–632. DOI.
• Guerra, Francesco, Lon Rosen, and Barry Simon. “The $P(\phi)_2$ Euclidean Quantum Field Theory as Classical Statistical Mechanics. Part I.” Annals of Mathematics 101, no. 1 (1975): 111–189. DOI.
• Guerra, Francesco, Lon Rosen, and Barry Simon. “The $P(\phi)_2$ Euclidean Quantum Field Theory as Classical Statistical Mechanics. Part II.” Annals of Mathematics 101, no. 2 (1975): 191–259. DOI.
• Osterwalder, Konrad, and Robert Schrader. “Axioms for Euclidean Green’s Functions II.” Communications in Mathematical Physics 42 (1975): 281–305. DOI. Open PDF.