Home·The theorem·The proof in one reading·Architecture·Units·Manuscript·Glossary·FAQ·Work items·Data·Method·Provenance·Changes
← VII.6 Scaling variants and fixed-field estimates (notes)Part VIII·Unit outline

VIII.1. Assembly of the main theorem — detailed notes

Contents

1. What this unit proves 2

2. What it uses 8

3. Strategy 10

4. Steps 11

4.1 The run and the ultraviolet endpoint 11

4.2 The data at one extraction order 12

4.3 The internal headline 15

4.4 Translation to the public statement 17

4.5 The headline theorems 19

4.6 Endpoints filed here and not on the path of the main theorem 20

5. Where the difficulty lies 20

6. Notation 21

7. What remains unexplained 21

The full account of unit VIII.1: every proposition with its proof sketch, checked against the Lean in two rounds and edited. The condensed outline and the map of the Lean are on the unit page.

Unit page · All units · Overview

Part: VIII, Assembly of the main theorem. Lean: the top-level files Solution.lean, Construction.lean and ConcreteSU2.lean; the bridges WilsonBridge.lean, HierarchyBridge.lean, TraceBridge.lean and ContactBridge.lean; the public definitions PublicTheory, PublicWilson, PublicView, ChallengeSpec and ContactSpec; the folders YM/Nonperturbative, YM/Independent, YM/Spec, YM/Org and YM/Pert/Parity; the files YM/R2I/Campaign/Accepted.lean and YM/RG17/Joint.lean; and parts of YM/FullVolume, YM/EveryPhase, YM/TraceExtension, YM/OPE, YM/Contacts and YM/Pert. Size: 12,060 lines in 1,674 declarations.

The construction chooses one Wilson family for each gauge group and block-spin parameter \(b > 0\). At each requested extraction order, it assembles the continuum table, phases, fields, reconstruction and OPE on one refinement. Fixed-volume ultraviolet convergence removes the mesh subsequence. The bridges then give the public theorem in statement.md.

1. What this unit proves

Conventions. The notation follows statement.md:

Internal objects:

Ordered limit. \(\text{Ord}_{\sigma} (\mathcal{C}, S)\) means the following. For every \(n\), \(v \in \mathcal{T}^n\) and \(t \in C^{\infty}_c (\text{Conf}_n)\):

Final object. A final object \(\mathfrak{F}\) on \((\mathcal{C}, \sigma, S)\) consists of five parts:

  1. Chosen renormalization data and step witnesses whose trajectory is exactly that of \(\mathcal{C}\) (D6).

  2. A seed: (D3), \(\text{Ord}_{\sigma} (\mathcal{C}, S)\), point limits, the Euclidean hierarchy axioms for \(S\), cell hypotheses and corrected-field widths.

  3. A field algebra: derivatives, rotations and a stress tensor.

  4. A phase selection with its certificate.

  5. Analytic continuations of the mixed and selected correlations, and a Wightman theory realizing the selected one.

Proposition VIII.1.1 (main theorem, public form). Let \((m, G, \mathfrak{g})\) be a gauge group with \(\mathfrak{g}\) simple. Then there is a Wilson family \(\mathcal{W}\) for \(G\) satisfying (Reg1)–(Reg4), with the following property. For every strictly increasing \(p : \mathbb{N} \to \mathbb{N}\) there are a strictly increasing \(q\) and a hierarchy over \(\mathcal{W}\) satisfying (P1)–(P4) along \(p \circ q\):

Status: theorem. Lean: YangMills.jointExistence, statement JointExistence, with WilsonRegularity, Hierarchy, Properties, Converges, Reconstructed, OperatorProductExpansion and TraceTest.

Proposition VIII.1.2 (detailed construction theorem). Fix a gauge group \((m, G, \mathfrak{g})\) with \(\mathfrak{g}\) simple and a real \(b > 0\). Then there is an internal family \(\mathcal{C}\) such that:

Moreover, for every extraction order \(\sigma\) there are:

such that the following hold:

  1. The public hierarchy \(T (\mathfrak{F}, Q)\) over \(\mathcal{W} (\mathcal{C})\) has (P1)–(P4) along \((\sigma \circ \tau)^{\text{v}}\), and every correlation \(S^s (a ; \cdot)\) extends to a distribution on \((\mathbb{R}^4)^n\) (D1, D2).

  2. \(\text{Ord}_{(\text{id}, (\sigma \circ \tau)^{\text{v}})} (\mathcal{C}, S)\) holds: the ordered limit with the full mesh sequence (D4).

  3. Positional ultraviolet convergence at every fixed volume holds along the full mesh sequence (D5). See VIII.1.6.

  4. Chosen renormalization data and step witnesses generate exactly the trajectory of \(\mathcal{C}\) (D6).

  5. The bundle (Core) holds: OPE services, reconstruction in every phase, the trace tests, the OPE foundation, and integrated OPE remainder bounds for integrable complex densities against positive measures, with bounded \(L^1\) mass.

  6. The internal Wightman axioms hold for the selected phase and for every phase \(\xi \in X\), and the public (W0)–(W3) hold for every phase. The \(\nu\)-average of the phase Wightman distributions is the boundary distribution of the mixed continuation (D7).

  7. The separated OPE and the nested OPE over finite trees hold (D8, D10).

  8. The seven headline trace clauses hold, together with the unique high-energy spectral law contained in (Core) (D9; VIII.1.n and VIII.1.o).

  9. The contact devices use the reader's prescription and give a full contact resolution, real bases of the OPE strata and contact thresholds (D11)–(D13).

  10. The leading scaling comparison with free Maxwell fields along \(\Sigma\) and the normalization at degree four hold (D14, D15).

The family \(\mathcal{C}\) depends only on \((G, b)\). Items 3 and 4 are properties of \(\mathcal{C}\) alone, although they are stated inside the quantifier over \(\sigma\). Status: theorem. Lean: YangMills.constructed. Internal form: Nonperturbative.existence, statement Headline.WilsonExistence.

Proposition VIII.1.3 (continuum theorem for the funded run). Fix a gauge group with \(\mathfrak{g}\) simple and \(b > 0\). Let \(\mathcal{C}_{G, b}\) be the family of the run \(\textsf{R}_{G, b}\) chosen in VIII.1.a. Its penalty is \(b\). For every extraction order \(\sigma\) there are \(\tau, S, \mathfrak{F}, \mathfrak{L}, Q, \textsf{M}\), with \(\mathfrak{F}\) on \((\mathcal{C}_{G, b}, \sigma \circ \tau, S)\), satisfying:

Status: theorem. Lean: continuum_with_position_tests. The form without \(\Sigma\) is continuum.

Proposition VIII.1.4 (a funded run supplies everything). Let \(\textsf{R}\) be a run with campaign \(T\) and pillars \(W\), let \(\Delta\) be a choice of local step estimates, and let \(\mathfrak{g}\) be simple. Assume \(\textsf{R}\) is funded (Fd): see §2, item 1. Then the assembly package (Goal) holds:

  1. (Fd).

  2. The ultraviolet package (UV\(_{\infty}\)) of VIII.1.5.

  3. The logarithmic OPE package.

  4. The dense-coverage property of the sharp Gram family.

  5. For every \(\sigma\), every table \(S\) and every final object \(\mathfrak{F}\) on \((\mathcal{C}_{\textsf{R}}, \sigma, S)\): the trace result of VIII.1.h.

  6. For every \(\sigma\) there are \(\tau\), numbers \((B_m)_{m \in \mathbb{N}}\) and data \(S, \mathfrak{F}\) (on \(\sigma \circ \tau\)), \(\Sigma, \mathfrak{L}, Q\) such that:

Status: theorem. Lean: R2IGoals.fromFunded, statement Broad.

Proposition VIII.1.5 (ultraviolet endpoint). Let \(\mathfrak{g}\) be simple, let \(\textsf{R}\) be a run with campaign \(T\) and pillars \(W\), and assume the stationary-point and amplification condition, the delay condition, and the weight-and-tableau data \((\vartheta, v)\) of (Fd). Then the ultraviolet package (UV\(_{\infty}\)) holds:

  1. (D3) holds for \(\mathcal{C}_{\textsf{R}}\), and the weight data exist.

  2. Sampler. Every \(\sigma\) has a refinement \(\tau\) carrying a table \(S\) and a final object on \((\mathcal{C}_{\textsf{R}}, \sigma \circ \tau, S)\), and \(\text{Ord}_{(\text{id}, (\sigma \circ \tau)^{\text{v}})}\) holds.

  3. The trajectory at every \((p, N)\) competes in, and is issued by, the fixed tuning recursion.

  4. Tuning and convergence. The parameter \(s \in [- 1, 1]\) specifies the target coupling \(\bar{\beta} + ws\) at level \(0\), where \(w > 0\) is the fixed tuning width. For every \((p, N, s)\) there is a successful trajectory, and its bare coupling at level \(N\) is unique within the prescribed recursion. These bare couplings are values of a function analytic on the open unit disc in \(s\). At each fixed \((m, p)\), the couplings and activities of successful trajectories converge as \(N \to \infty\) whenever their level-zero targets converge. For complex targets in the disc, convergence of the analytic couplings and activities, with every parameter derivative, is uniform on compact subsets; activities are measured in the fixed weighted norm at \((m, p)\).

  5. For every fixed \((m, p)\) there is \(c_{m, p} > 0\) such that

    \(\displaystyle e^{- \text{scalar}_{N, p, m}} Z_{N, p} \geqslant c_{m, p} \qquad (N \geqslant m),\)

    where \(Z_{N, p}\) is the bare Wilson partition function relative to probability Haar measure and \(\text{scalar}_{N, p, m}\) is the field-independent scalar in the level-\(m\) action. Its scalar density \(v_{N, p, m}\) tends to \(- \infty\) as \(N \to \infty\).

  6. The volume-uniform bounds hold, as stated in VIII.1.c, step 1.

  7. (D5) holds along the full mesh sequence.

  8. For every fixed \(p\) there is a family of successful trajectories with target \(\bar{\beta} + \lambda ws\), \(s \in [- 1, 1]\), for some \(\lambda > 0\), whose central member is the chosen family. They obey the same step relations and insertion normalization. For each token list, each compact \(K \subset \text{NoHit}_p\) and each finite spatial derivative order \(h\), there are an open neighbourhood \(\Omega\) of \(K\) in \(\text{NoHit}_p\) and a smaller positive parameter width, independent of \(N\). In the resulting unit-disc coordinate, the limiting correlation is holomorphic in the parameter and \(C^h\) in position on \(\Omega\). All fixed coordinate differences of order at most \(h\), their interpolants and their admissible smearings converge to the corresponding derivatives, uniformly on compact parameter and position sets, with every parameter derivative.

Status: theorem. It is assembled here; its ingredients are in III.12–III.15. Lean: FullVolume.Final.jewelOf17 and Jewel.

Proposition VIII.1.6 (the ultraviolet limit along the full spacing sequence). For \(p \in \mathbb{N}\), let \(\text{NoHit}_p\) be the set of configurations no two of whose points are congruent modulo \(L_p \mathbb{Z}^4\).

(a) For the family of a run satisfying VIII.1.5's hypotheses, the following holds for every \(p\), \(n\) and \(v \in \mathcal{T}^n\). There is \(f_{p, v} \in C^{\infty} (\text{NoHit}_p)\) such that:

The cutoff thresholds may depend on \(p, n, v\), the compact set, the tests and the difference stencils.

(b) Suppose a final object exists on \((\mathcal{C}, \sigma, S)\) and (a) holds for \(\mathcal{C}\). Then \(\text{Ord}_{(\text{id}, \sigma^{\text{v}})} (\mathcal{C}, S)\) holds.

Status: theorems. Lean: (a) Campaign.realUV, with the definition PositionsUV. (b) FullVolume.Final.from_spec.

Proposition VIII.1.7 (public bridges). Let \(\mathfrak{F}\) be a final object on \((\mathcal{C}, \sigma, S)\) and \(Q\) a reader, and let \(T (\mathfrak{F}, Q)\) be the public hierarchy built from them (VIII.1.p).

  1. The public Wilson energy, Wilson measure, lattice moments and smeared moments of \(\mathcal{W} (\mathcal{C})\) equal the internal ones. \(\text{Ord}_{(\text{id}, s)} (\mathcal{C}, S)\) implies (P1) along \(s\).

  2. If (D3) holds and \(\mathfrak{g}\) is simple, then \(\mathcal{W} (\mathcal{C})\) satisfies (Reg1)–(Reg4), with \(b_{\ast} = 2 b_0\).

  3. The reconstruction in every phase implies (R1)–(R15).

  4. The internal separated OPE implies (O1)–(O2).

  5. If \(\mathfrak{g}\) is simple, the seven internal headline trace clauses imply (T1)–(T4), with \(3 \dim \mathfrak{g}/ \pi^4\) identified with the internal amplitude.

  6. The full contact resolution implies (D2).

Status: theorems. Lean: wilsonFamily and fromWilson (1); regular (2); publicReconstruction (3); publicOPE (4); publicTrace (5); contactDistribution (6).

Proposition VIII.1.8 (exclusion of Wick descriptions in momentum space). Let \(\mathfrak{F}\) be a final object. Assume \(s^8 (\log 1 / s)^2 G_{\Theta} (s) \to A > 0\) as \(s \downarrow 0\). Let \(W\) be any finite list of complex coefficients \(c_{\alpha}\) and products of mass-shell modes, in the class defining Wick profiles in statement.md. There are Schwartz functions \(f, g\) on \(\mathbb{R}^4\) for which

\(\displaystyle v_{f, g} (p) := \overline{\mathcal{F}_+ f (p)} \hspace{0.17em} \mathcal{F}_+ g (p)\)

has compact support and

\(\displaystyle \mathcal{W}^{\xi_{\ast}}_2 (\Theta^{\ast}, \Theta) (\bar{f} \otimes g) \ne \sum_{\alpha} c_{\alpha} \hspace{0.17em} \text{read}_{\alpha} (v_{f, g}) .\)

Here \(\mathcal{F}_+\) is the unnormalized positive-sign Fourier transform used in the spectral representation. The four-dimensional test \(v_{f, g}\) is the transform of the joint test restricted to relative momenta \((- p, p)\). Status: theorem. It enters the trace result of VII.3. The public (T4) follows directly from the Euclidean exclusion (VIII.1.r). Lean: TraceExtension.Wick.polynomial_excludes.

Proposition VIII.1.9 (\(SU (2)\)). \(\mathfrak{su} (2) \subset M_2 (\mathbb{C})\) is simple. Hence VIII.1.1 holds for \((2, SU (2), \mathfrak{su} (2))\). Status: theorem. Lean: su2Simple (from Examples.SU2.simple) and su2Theory.

Proposition VIII.1.10 (definitions filed here and used elsewhere). Several definitions used by III.10–III.15, V.1–V.3, VI.1–VI.2 and VII.1–VII.4 are filed in this unit:

Status: definitions, and theorems described in the units that use them.

2. What it uses

From other units.

  1. III.14 and III.2 (selection of the run). For simple \(\mathfrak{g}\) and \(b > 0\) there are a run \(\textsf{R}\), a campaign \(T\) and pillars \(W\) such that the penalty is \(b\) and \(\textsf{R}\) is funded for the fixed local-estimate choice \(\Delta\). A run records levels, schedule of constants, step budgets, the trajectory and its tuning, the reference levels and the transfer data of the labels.

  2. III.12–III.15 (ultraviolet bounds and convergence). Used for VIII.1.5 and (D5):

  3. I.1, II.1–II.2 (setting and construction record). The internal family, (D3) including the running law: for every \(\varepsilon > 0\), \(| \beta_{N, p} (m) - \hat{\beta}_m | \leqslant \varepsilon \log (1 / a_m)\) for all sufficiently large \(m\) and all \(N \geqslant m\) and all \(p\), where \(\hat{\beta}_m = \bar{\beta} + 2 b_0 \log (1 / a_m)\); the band condition \(\beta_{N, p} (m) \in [\hat{\beta}_m / 2, 2 \hat{\beta}_m]\); tokens, local sections and their stencils; tables.

  4. IV.1 (infinite volume). The seed: a subsequential table \(S\) with \(\text{Ord}_{\sigma \circ \tau}\), point limits, the Euclidean hierarchy axioms, cell hypotheses and corrected-field widths. It is produced inside the run's extraction property and used in VIII.1.g.

  5. V.1–V.3 (phases and reconstruction).

  6. VI.1–VI.2 (symmetry). The field algebra (derivatives, rotations, conserved stress tensor) and its local laws: the local-law package built by Bid.station from work items A01–A17, and Org.Tab.

  7. VII.1 (free Maxwell comparison). The free-Maxwell data maxwell_supplies and FreeStress.fire; the leading short-distance comparison; the degree-four normalization (credible, square_four).

  8. VII.2–VII.3 (trace).

  9. VII.4 (OPE and contacts).

  10. VII.5 (asymptotic freedom). The value of the slope, \(2 b_0 = 11 C_{\mathfrak{g}} / (24 \pi^2)\), and \(C_{\mathfrak{g}} > 0\) (casimir_pos).

Standard mathematics used.

The inputs rest on further standard mathematics:

Imported constants and assumptions. The final existence theorem has no additional analytic hypotheses. The run, local-law packages, final object and (Core) are supplied by the preceding construction results.

Parameters and auxiliary choices:

Constants:

3. Strategy

Choose one run \(\textsf{R}_{G, b}\) before any extraction order, observable or test. Given \(\sigma\), first extract a refinement \(\sigma \circ g\) on which the couplings at every fixed level have ordered limits. Next extract \((\sigma \circ g) \circ w\), carrying a continuum table and a final object. The later reconstruction, trace, OPE and contact results attach to that object. Passing to further subsequences preserves the earlier limits, so \(\tau = g \circ w\) serves every conclusion.

At each fixed volume, the renormalization-step estimates make the lattice correlations Cauchy on a dense union of grids. Equicontinuity gives convergence on separated configurations; discrete Taylor estimates recover all derivatives. Dominated convergence gives the smeared limits. Uniqueness of limits identifies these full-sequence limits with the rows of the extracted table, removing the mesh subsequence.

Specialize the detailed theorem to \(b = 1\) and \(\sigma = (\text{id}, p)\). The volume part of \(\tau\) is the required \(q\). The public hierarchy is read from the final object and its OPE reader. The bridges identify the Wilson measures and moments, reconstruction axioms, OPE estimates and trace tests. The plaquette identity, equality of Wick profiles and evaluation of the trace amplitude are the required computations.

4. Steps

The internal statements use a fixed gauge group \((m, G, \mathfrak{g})\). When a run is specified, its campaign and pillars are those appearing in (Fd) and VIII.1.5.

4.1 The run and the ultraviolet endpoint

Lemma VIII.1.a (the chosen run). Fix a gauge group with \(\mathfrak{g}\) simple and \(b > 0\). There are a run \(\textsf{R}_{G, b}\), a campaign and pillars with penalty \(b\) satisfying (Fd). Write \(\mathcal{C}_{G, b}\) for its family. Proof sketch. Apply the run-selection theorem of III.2 with the fixed dynamics, local estimates and component prescription of III.14. The required schedule predicate is identically true, so it is satisfied after any delay extension. Choose one witness before specifying an extraction order.

Lean: constructionRun (III.14) and fundedRun. (proof read; selection theorem statement only)

Lemma VIII.1.b (ultraviolet endpoint). This is VIII.1.5. Proof sketch. The convergence theorem of III.13–III.15 supplies the analytic families of couplings and activities. Its uniqueness clause identifies successful trajectories with the same level-zero target, and its analytic clause gives the bare-coupling interpolation. Removing the field-independent scalar from the partition function gives the positive lower bound; decay of the bare partition function then forces the scalar density to \(- \infty\). The volume-uniform bounds and VIII.1.c give (D5). The parameter-family construction gives item 8. Finally, refine to a final object and apply VIII.1.d to obtain the sampler with the full mesh sequence. The full funding hypothesis supplies the three assumptions of VIII.1.5. Lean: clients; jewelOf17, wins, bareTrial and Funded.jewel. Further sources: YM.PhysicalNext.Repeat.Legato.Campaign.Winter, YM.PhysicalNext.Repeat.Legato.Campaign.offers, YM.FullVolume.stripped, YM.FullVolume.saved, YM.FullVolume.field_independent. (assembly, analytic convergence, tuning, costs and parameter-family proofs read; their estimates are inputs from III.12–III.15)

Lemma VIII.1.c (convergence along all spacings at fixed volume). This is VIII.1.6(a). Proof sketch. Fix \(p, n, v\).

  1. The volume-uniform bounds control every fixed list of lattice differences on separated configurations. The bound and its cutoff threshold may depend on \(n, v\), the differences and the separation, but are uniform in \(p\).

  2. The union of the configuration grids is dense. Grid nesting makes each point of this union a grid point at all sufficiently fine spacings. At these configurations, the renormalization-step comparison from III.12 gives a Cauchy sequence along all \(N\). Equicontinuity extends convergence locally uniformly to \(\text{NoHit}_p\).

  3. A discrete Taylor estimate compares each first difference with a difference quotient over a fixed small physical displacement. The function values already converge; first let \(N \to \infty\), then let the displacement shrink. This proves that the first differences are Cauchy. Iterating gives limits of all differences, and the Taylor estimate identifies them as the derivatives of a smooth function \(f_{p, v}\). Multilinear interpolation has the same limits.

  4. The lattice sums are integrals of piecewise constant functions on a fundamental cell. A uniform bound on the test support permits dominated convergence, giving the smeared limits.

Lean: realUV, purchased, positioned and FixedBounds. Further sources: YM.FullVolume.Eves.dense, YM.FullVolume.Eves.lateAnch, YM.FullVolume.from_previous, YM.FullVolume.honors, YM.FullVolume.augments, YM.FullVolume.pudding. (proof read, including dense-grid convergence, Taylor recovery and smearing; the renormalization-step estimate is an input)

Lemma VIII.1.d (removing the mesh subsequence). This is VIII.1.6(b). Proof sketch. The seed of \(\mathfrak{F}\) gives, for each admissible \(p\), \(C_{\sigma^{\text{m}} (j), p} (v ; t) \to R_p\), and also \(R_{\sigma^{\text{v}} (k)} \to \int tS (v ; \cdot)\). By VIII.1.c, \(C_{N, p} (v ; t) \to \int tf_{p, v}\) along all \(N\). Since \(\sigma^{\text{m}}\) is cofinal, uniqueness of limits gives \(R_p = \int tf_{p, v}\). Thus the same rows \(R_p\) are limits along all \(N\), and the volume limit is unchanged. Lean: from_spec.

Lemma VIII.1.e (OPE and logarithmic packages). Let \(\textsf{R}\) be a run with campaign \(T\) and pillars \(W\). Assume \(\mathfrak{g}\) simple and (Fd). Then the family carries the OPE package and its logarithmic extension. Proof sketch. The pillars, polynomial schedule compatibility, halo condition and weight data supply the ordinary package. The logarithmic extension adds the logarithmic exit estimate, its budget, the fixed support exponent, and divergence of the composed schedule \(m \mapsto Q (J (m))\). The last property follows from \(J (m) \to \infty\) and \(Q (k) \to \infty\).

Lean: Funded.logPackage and opePackage17. (assembly proofs read)

4.2 The data at one extraction order

Lemma VIII.1.f (limits of the couplings at fixed level). Assume (D3) holds for \(\mathcal{C}\), and let \(\sigma\) be an extraction order. Then there are \(g\) and \((B_m)_m\) such that, for every \(m\), \(\beta_{N, p} (m)\) has the ordered limit \(B_m\) along \(\sigma \circ g\). Moreover \(B_m / \log (1 / a_m) \to 2 b_0\). Proof sketch. Set \(c_N (p, m) := \beta_{N, p} (m)\) for \(m \leqslant N\) and \(c_N (p, m) := \hat{\beta}_m\) for \(m > N\). By the band condition of (D3), \(c_N \in \prod_{(p, m)} [\hat{\beta}_m / 2, 2 \hat{\beta}_m]\), a compact metrizable space.

  1. Extract \(g^{\text{m}}\) with \(c_{\sigma^{\text{m}} (g^{\text{m}} (j))} \to a\) in the product.

  2. Extract \(g^{\text{v}}\) with \(a (\sigma^{\text{v}} (g^{\text{v}} (k)), \cdot) \to B\) in \(\prod_m [\hat{\beta}_m / 2, 2 \hat{\beta}_m]\).

  3. For fixed \(m\) the truncation is eventually inactive.

  4. The running law of (D3) gives \(|B_m - \hat{\beta}_m | \leqslant \varepsilon \log (1 / a_m)\) for large \(m\), hence the slope.

The slope is not used by the main theorem. Lean: Running.extraction and Running.discrete.

Lemma VIII.1.g (a final object on a refinement). Let \(\mathfrak{g}\) be simple, let \(\textsf{R}\) be a run with the OPE package, and let \(\sigma\) be an extraction order. Then there are a refinement \(w\), a table \(S\), a final object \(\mathfrak{F}\) on \((\mathcal{C}_{\textsf{R}}, \sigma \circ w, S)\) and scaling data \(\Sigma\), with OPE services and (D14). Proof sketch.

  1. The run's extraction property supplies a table and seed on a refinement, together with the trajectory witnesses (D6).

  2. Add the free Maxwell data, with the translation normalization in which \(\kappa^{- 1} J\) is the free Maxwell stress tensor.

  3. The weak comparison estimates give the infrared data, scaling data \(\Sigma\) and the first comparison.

  4. The symmetry results of VI.1–VI.2 produce the derivatives, rotations and stress tensor with their local laws.

  5. The phase results B01–B18 produce the selection and its certificate. The reconstruction theorem then supplies the mixed and selected continuations and the selected Wightman theory. These data form \(\mathfrak{F}\) with the prescribed field algebra.

  6. The leading mixed comparison, the one-dimensional space of conserved tensors, and the limits of derivatives and rotations give (D14). Lean: Pert.withFinal; Run.extractible17. (assembly and final-object proofs read; analytic and symmetry inputs as stated)

Lemma VIII.1.h (trace result for every final object). Assume (Fd) and \(\mathfrak{g}\) simple. For every extraction order \(\sigma\), every table \(S\) and every final object \(\mathfrak{F}\) on \((\mathcal{C}_{\textsf{R}}, \sigma, S)\):

Proof sketch. The phase-comparison condition in (Fd) supplies the trace comparison data at \(\sigma\). The trace asymptotic, spectral law and exclusions of VII.2–VII.3 then apply to this final object. Centering and scalar covariance follow from the stress-tensor laws. The stress-comparison theorem first constructs one admissible normalized scheme, then proves the comparison for every scheme satisfying those same conditions.

Lean: Trace.Target.Result, assembled, arenaT, crest_ready and Funded.gallery. (assembly proofs read; the trace estimates are inputs from VII.2–VII.3)

Lemma VIII.1.i (Wick exclusion in momentum space). This is VIII.1.8. Proof sketch.

  1. Let \(\mu\) be the trace spectral measure. It is tempered and supported in the forward cone. For a joint Schwartz test \(\varphi\), the selected two-point function is the integral of its relative-momentum transform against \(\mu\), and

    \(\displaystyle G_{\Theta} (s) = \int e^{- sp^4} \hspace{0.17em} \mu (dp) \qquad (s > 0) .\)
  2. Suppose the Wick identity held for every allowed pair \((f, g)\). Then \(\int v \hspace{0.17em} d \mu = \sum_{\alpha} c_{\alpha} \text{read}_{\alpha} (v)\) for every compactly supported Schwartz \(v\). To see this, choose \(u\) with \(|\mathcal{F}_+ u| = 1\) on a ball containing \(\text{supp} \hspace{0.17em} v\), and use Fourier surjectivity to choose \(w\) with \(\mathcal{F}_+ w = (\mathcal{F}_+ u) v\). Thus \(\overline{\mathcal{F}_+ u} \mathcal{F}_+ w = v\).

  3. For each \(s > 0\), approximate \(e^{- sp^4}\) by compactly supported smooth cutoffs bounded in absolute value by \(e^{- sp^4}\). Exponential integrability for \(\mu\) and for every mass-shell term permits dominated convergence on both sides. Hence \(G_{\Theta} (s) = \sum_{\alpha} c_{\alpha} K_{\alpha} (s)\) for every \(s > 0\).

  4. The resulting Wick profile contradicts VIII.1.r(c).

Lean: polynomial_excludes, identifies_euclid and traceSpectrum_exists. Further sources: YM.TraceExtension.productTest, YM.TraceExtension.waveTest_product.

Lemma VIII.1.j (reconstruction in every phase; input from V.3). For every final object \(\mathfrak{F}\) there are reconstruction data \(\mathfrak{L}\) with the following properties. The phasewise assertions hold for every \(\xi \in X\).

  1. For every \(\xi \in X\), a Wightman theory \(\mathfrak{W}_{\xi}\) realizes the continuation of that phase, with the same labels. At \(\xi_{\ast}\) it is the selected theory of \(\mathfrak{F}\).

  2. The comparison map from \(\mathfrak{W}_{\xi}\) to its Osterwalder–Schrader space is onto.

  3. The Euclidean ground state is unique, and the zero-energy subspace is \(\mathbb{C} \Omega_{\xi}\).

  4. For every finite sum \(A\) of field words there is \(M_A \geqslant 0\), independent of \(\xi\), such that

    \(\displaystyle \mu_{A \Omega_{\xi}} (\{0 < p^4 \leqslant u\}) \leqslant M_A u^2 \quad (0 < u \leqslant 1),\)
    \(\displaystyle 0 \leqslant \langle A \Omega_{\xi}, (e^{- sH_{\xi}} - P_{0, \xi}) A \Omega_{\xi} \rangle \leqslant M_A s^{- 2} \quad (s \geqslant 1) .\)
  5. The bounded weak commutant, defined using the unbounded adjoints on the field-word domain, is \(\mathbb{C} \hspace{0.17em} \text{id}\).

  6. The stress tensor continues to every phase, and each Wightman distribution evaluated on a fixed joint Schwartz test is continuous in \(\xi\).

  7. \(\int S^{\xi} \hspace{0.17em} d \nu = S\) on separated words, and \(\int \mathcal{W}^{\xi} \hspace{0.17em} d \nu =\mathcal{W}^{\text{mix}}\) on every joint Schwartz test.

Proof sketch (V.1, V.3). The phases are characters of the commutative \(C^{\ast}\)-algebra generated by time-slab word operators compressed to the mixed Osterwalder–Schrader zero-energy subspace. Its vacuum measure is \(\nu\). Translate a separated word into a slab to obtain the Euclidean barycentre identity. Apply the reconstruction, low-energy, heat-decay, irreducibility and stress-continuation results of V.3 to every character, retaining the original selected theory at \(\xi_{\ast}\). Their continuity and boundary barycentre identities give the last two assertions.

Lean: all_characters and Completed. (statement and top-level proof read)

Lemma VIII.1.k (the OPE reader and its estimates). Let \(\mathfrak{F}\) be a final object on \((\mathcal{C}, \sigma, S)\) with OPE services for that family, order and table. Choose a coherent prescription \(P\) for the field derivatives, adapted to the symmetric laws of the table and to the transport injection. Let \(Q\) be the reader formed by \(P\), its coherence, the Euclidean kernel and the transport data (injection, orientation, local laws). Then:

Proof sketch. Extend the finite-degree prescriptions compatibly and take their common restriction in each degree. The OPE services give the coefficient and remainder estimates, which also supply the nested expansions and the all-shapes expansion. Dense coverage supplies the test portfolios used to construct the sharp Gram allocation.

Lean: Prescription.chosen (VII.4), Orbits.produce, Nesting.uniform and attains_mantle17. (assembly proofs read; OPE estimates as stated)

Lemma VIII.1.l (assembly of the full package). This is VIII.1.4. Proof sketch. The proof has three parts.

Lean: fromFunded and Order.retained.

4.3 The internal headline

Lemma VIII.1.m (contact devices). Let \(\mathfrak{F}\) carry a reader \(Q\) and OPE services. Then there are devices \(\textsf{M}\) with the following properties:

Proof sketch. Use the reader’s coherent prescription, the rotations, the Euclidean hierarchy and the phase transport laws, together with a multiple-contact detector in each degree. The contact theorem of VII.4 gives the resolution from the OPE services. Real conjugation permits a real basis in each stratum, containing the unit in degree zero. Successively choose each positive threshold below one eighth of its predecessor, \(2^{- n}\) and \(\text{clearance}_{\textsf{M}} (n) / (2 (n + 1))\). Lean: contacts_on and MultipleLight. Further sources: YM.Contacts.Solstice.Thresholds.

Lemma VIII.1.n (continuum theorem). This is VIII.1.3. Proof sketch. Take \((T, W)\) and (Fd) from the specification of \(\textsf{R}_{G, b}\), and apply the refinement clause of VIII.1.4 at \(\sigma\). The coupling limits, the all-shapes OPE and the sharp Gram allocations are omitted. Contacts come from VIII.1.m. (Core) comprises:

The internal trace-test bundle has eight clauses:

Finally, the degree-four determining tests of VII.1 convert (D14) into (D15). Lean: continuum_with_position_tests, Core, Core.of_receipts, ManuscriptTraceTests.of_result and Pert.Position.four_on. (assembly and degree-four normalization proofs read)

Lemma VIII.1.o (internal headline). For a gauge group with simple \(\mathfrak{g}\) and \(b > 0\), the internal detailed existence statement holds with \(\mathcal{C}=\mathcal{C}_{G, b}\). Proof sketch. Item by item:

Lean: Nonperturbative.existence, Reconstruction, WightmanAxioms, SeparatedOPE, NestedOPE, TraceExclusions and Core.point.

4.4 Translation to the public statement

Lemma VIII.1.p (the public Wilson family and hierarchy). Let \(\mathfrak{F}\) be a final object on \((\mathcal{C}, \sigma, S)\) and \(Q\) a reader. The public Wilson family \(\mathcal{W} (\mathcal{C})\) is defined by:

The public Wilson energy equals the internal one. The proof matches the plaquette \(U (y, \mu) U (y + e_{\mu}, \nu) U (y + e_{\nu}, \mu)^{- 1} U (y, \nu)^{- 1}\) with the internal face value. Hence the measures, the lattice moments and the smeared moments agree.

The hierarchy \(T (\mathfrak{F}, Q)\) has the following components:

Lean: energy_eq, law_eq, moment_eq, smear_eq and hierarchy.

Lemma VIII.1.q (regularity and the Wilson limit). Let \(\mathfrak{F}\) be a final object on \((\mathcal{C}, \sigma, S)\) and \(Q\) a reader.

(a) If (D3) holds and \(\mathfrak{g}\) is simple, then (Reg1)–(Reg4) hold:

(b) If \(\text{Ord}_{(\text{id}, s)} (\mathcal{C}, S)\) holds, then (P1) holds along \(s\). The smeared moments agree by VIII.1.p. The mixed correlation of \(\pi (v)\) equals \(S (v ; \cdot)\) on \(\text{Conf}_n\), and \(t = 0\) off \(\text{Conf}_n\). Lean: regular, zeroUp_bound, fromWilson and wilsonLimit.

Lemma VIII.1.r (reconstruction, OPE and trace clauses). Let \(\mathfrak{F}\) be a final object and \(Q\) a reader for it. The conclusions concern \(T (\mathfrak{F}, Q)\).

(a) The reconstruction in every phase implies (R1)–(R15). Clause by clause:

(b) The internal separated OPE implies (O1)–(O2), after enumerating \(I \sqcup J\) once.

(c) Assume \(\mathfrak{g}\) simple and the seven internal headline trace clauses. Then (T1)–(T4) hold. Clause (T1) is the centered clause. For (T2) and (T3), the internal amplitude is \(12 (d_2 / b_0)^2 \dim \mathfrak{g}/ \pi^4 = 3 \dim \mathfrak{g}/ \pi^4\), and \(G_{\Theta}\) is by definition the internal Euclidean two-point function. For (T4), public and internal Wick profiles coincide; then apply the Euclidean Wick exclusion. The real part of a finite Wick profile has, for every \(k \in \mathbb{Z}\), an expansion modulo \(O (r^k)\) in the span of \(r^a (\log r)^j\) with \(a \in \mathbb{Z}\) and \(j \in \mathbb{N}\). No function with such an expansion modulo \(O (1)\) satisfies \(r^8 (\log 1 / r)^2 f (r) \to A \ne 0\), since that would need the power \((\log r)^{- 2}\). Thus (T3), together with \(A_{\mathfrak{g}} > 0\) from (T2), implies (T4). Lean: publicReconstruction, view_axioms, publicMoment, publicOPE, publicTrace and euclidean_excludes. (proofs read, including the power-log exclusion)

Lemma VIII.1.s (contact distributions). Let \(\mathfrak{F}\) be a final object with contact devices \(\textsf{M}\) satisfying the full contact resolution. For every state \(s\), every \(k\) and every \(a \in \mathbb{F}^k\), there is a continuous linear functional \(\mathcal{D}\) on \(C^{\infty}_c ((\mathbb{R}^4)^k)\) with \(\mathcal{D} (f) = \int fS^s (a ; \cdot)\) whenever \(\text{supp} f \subset \text{Conf}_k\), using the usual test-function topology. No uniqueness is asserted. Proof sketch. Apply the distribution clause of the resolution with:

The contact theorem supplies \(\mathcal{D}\). On a separated test, the subtraction formula reduces to the original field word. Reading that word gives the correlation; off the support of the test both integrands vanish. Lean: contactDistribution.

4.5 The headline theorems

Lemma VIII.1.t (detailed theorem). This is VIII.1.2. Proof sketch. Apply VIII.1.o to obtain \(\mathcal{C}\), the penalty and (D3). Regularity follows from VIII.1.q(a). For each \(\sigma\), unpack the data and conclude as follows:

Lean: constructed.

Lemma VIII.1.u (public theorem). This is VIII.1.1. Proof sketch. Apply VIII.1.t with \(b = 1\). Given \(p\), let \(\sigma = (\text{id}, p)\), and let \(\tau\) be the refinement produced. Take \(q := \tau^{\text{v}}\) and the hierarchy \(T (\mathfrak{F}, Q)\). Then \((\sigma \circ \tau)^{\text{v}} = p \circ q\), and the mesh part of \(\tau\) plays no role because (P1) is proved along the full mesh sequence (VIII.1.d). Lean: jointExistence and Order.refine.

Lemma VIII.1.v (\(SU (2)\)). This is VIII.1.9. Proof sketch. The matrices

\(\displaystyle A = \left( \begin{array}{cc} i & 0\\ 0 & - i \end{array} \right), \qquad B = \left( \begin{array}{cc} 0 & 1\\ - 1 & 0 \end{array} \right), \qquad C = \left( \begin{array}{cc} 0 & i\\ i & 0 \end{array} \right)\)

span \(\mathfrak{su} (2)\) over \(\mathbb{R}\) and satisfy \([A, B] = 2 C\), \([B, C] = 2 A\), \([C, A] = 2 B\). If an ideal contains \(aA + bB + cC \ne 0\), a suitable operator \(1 + \tfrac{1}{4} (\text{ad} \hspace{0.17em} A)^2\), or its cyclic counterpart, isolates a nonzero multiple of one basis vector. The bracket identities then put all three in the ideal. Apply VIII.1.u. Lean: su2 and su2Theory. (proof read, including simplicity)

4.6 Endpoints filed here and not on the path of the main theorem

These separate results and alternative existence theorems are not inputs to the assembly above:

The volume-uniform ultraviolet bounds belong mathematically to III.12–III.15; the supporting reconstruction and symmetry lemmas belong to V.2–V.3 and VI.1.

5. Where the difficulty lies

The assembly preserves the order of choices. A family fixed before \(\sigma\) must support the coupling limits, continuum table and all later structures on one refinement. In particular, the fixed-volume ultraviolet convergence must concern this same family. Uniqueness of row limits then removes the mesh subsequence without replacing the final object or its point-limit data.

The momentum-space Wick exclusion requires a further argument: every compactly supported relative-momentum test must be realizable as \(\overline{\mathcal{F}_+ u} \mathcal{F}_+ w\). Fourier surjectivity and an auxiliary transform of modulus one on the test support give that realization. Exponential integrability then permits passage to Euclidean Laplace transforms.

The public theorem retains less than the detailed construction. Its Wilson limit concerns the mixed state, while its trace test concerns the selected phase. The Euclidean Wick exclusion follows from (T2)–(T3); the detailed theorem also gives momentum-space and Maxwell relative-locality countertests. The low-energy and heat bounds are polynomial and imply no mass gap. Coupling limits, all-shapes OPE and sharp Gram allocations occur in (Goal) but are omitted from VIII.1.3. Neither independence of the volume subsequence nor agreement between different penalties or run choices is asserted.

6. Notation

object code name proposed notation
gauge group GaugeGroup, IsSimple \((m, G, \mathfrak{g})\), \(\mathfrak{g}\) simple
public Wilson family wilsonFamily \(\mathcal{W} (\mathcal{C}) = (L_p, D_N, \beta_{N, p}, h, \mathcal{T}, I_{N, p})\)
regularity WilsonRegularity (Reg1)–(Reg4)
public hierarchy hierarchy \(T (\mathfrak{F}, Q)\)
extraction order, refinement Extraction.Order, refine \(\sigma = (\sigma^{\text{m}}, \sigma^{\text{v}})\), \(\sigma \circ \tau\)
ordered limit Ordered \(\text{Ord}_{\sigma} (\mathcal{C}, S)\)
internal family, construction predicate Family, Makes \(\mathcal{C}\), (D3)
run, funding conditions Run, fundedRun, Funded \(\textsf{R}_{G, b}\), (Fd)
ultraviolet endpoint Jewel (UV\(_{\infty}\))
positional convergence PositionsUV (D5), \(f_{p, v}\)
full assembly package Broad (Goal)
table Table \(S\), read as \(S^s\)
final object Spec.Final \(\mathfrak{F}\) on \((\mathcal{C}, \sigma, S)\)
reconstruction in every phase Almanac, Completed \(\mathfrak{L}= (\mathfrak{W}_{\xi})_{\xi \in X}\)
OPE reader Orbits.Reader \(Q\); \((\text{Ch}_j, O_{j, \kappa}, c_{j, \kappa})\)
contact devices Contacts.Devices \(\textsf{M}\)
scaling sequence Climb.Scaling \(\Sigma = (r_k, \xi_k, U_k)\)
combined continuum properties Independent.Core (Core)
first comparison, degree-four normalization FirstExtension, FourMeters (D14), (D15)
centered trace, its two-point function TraceExtension.center, euclidean \(\Theta\), \(G_{\Theta}\)
trace amplitude Trace.Verse.award \(A_{\mathfrak{g}} = 3 \dim \mathfrak{g}/ \pi^4\)
one-loop constants Pert.Asymptotics.b0, Trace.Over.d2 \(b_0 = 11 C_{\mathfrak{g}} / (48 \pi^2)\), \(b_{\ast} = 2 b_0\), \(d_2 = - b_0 / 2\)
reference coupling Calendar.hatBeta \(\hat{\beta}_m = \bar{\beta} + 2 b_0 \log (1 / a_m)\)
penalty of the block-spin kernel KernelChoices.penalty \(b\)

7. What remains unexplained

  1. Numerical run conditions. The thirteen funding conditions in §2 are specified by their roles. The exact inequalities for the frame and polynomial schedules, delays, pacing, thermal estimates and weight tableau must be supplied from III.15. The numerical proofs are inputs.

  2. Renormalization comparison. VIII.1.c uses the fixed-volume Cauchy estimate from III.12. The underlying renormalization-step estimate is an input; the dense-grid, Taylor and smearing arguments are given in VIII.1.c.

  3. Normalized trace schemes. VIII.1.h retains the existence and universal quantifiers of the stress-comparison theorem. The alignment, leading-derivative and scalar normalization conditions require the definitions and notation of VII.3. No comparison for an arbitrary renormalization prescription follows.

  4. Upstream reconstruction. The phase selection and selected reconstruction are produced inside the final-object construction. The every-phase theorem is an input from V.3; its low-energy, decay, irreducibility and boundary arguments remain inputs from V.3.

← VII.6 Scaling variants and fixed-field estimates (notes)Part VIII·Unit outline