Gate dossier — Gap-02 — Yang–Mills mass gap

Question: Is the strong-force gap the famous unsolved Millennium problem?
Status (governing, 2026-07-12): CLOSED-SCOPED / POSITIVE-SPECTRAL-GAP-DERIVED-FOR-EACH-FINITE-OPERATIONAL-SU3-LATTICE / CONTINUUM-AND-INFINITE-VOLUME-UNIFORM-GAP-PSTAR-CERTIFIED-IRREDUCIBLE / NOT-A-CLAY-PROOF
Full 13-dimensional treatment: the frozen arena M4 x K6=SU(3)/T2 x S2 x S1_Y/Z2, all three layers, full precision. Self-contained — every value derived or stated inline.

Required endpoint (2026-07-06)

[SUPERSEDED 2026-07-12 — see Governing Correction. Governing status is now CLOSED-SCOPED with a DERIVED positive gap for each finite operational SU(3) lattice; the continuum-uniform gap \(P_\star\) stays CERTIFIED-IRREDUCIBLE. The pre-correction status line below claimed no internal derivation of any gap.]

Status: CLOSED / CERTIFIED-IRREDUCIBLE.

Nothing left. Anchored on:

• Shape: the frozen internal geometry supplies the color group SU(3) from K6 = SU(3)/T2 — it fixes where the strong force comes from, not a proof of the gap
• Granularity: the finite-resolution (cost-floor) axiom lets the theory decline the infinitely-fine limit — a finite cell needs no continuum, and on it the gap is a theorem. [RETAINED — rerouted: per the Governing Correction the finite-cell gap is now the gate's DERIVED positive result (executable witness PASS); granularity is the CONSTRUCTION-ANCHOR that selects the finite domain \(a\ge\Delta_0,\ |E|<\infty\). The clause "changes the hard question rather than answering it" applies now only to the held-out continuum-uniform \(P_\star\), not to the finite gap.]
• Scale: dimensional transmutation pins the strong-force scale ΛYM as boundary data — the unit every gap statement is measured against, not a tuned number
• Observables: The mass gap Δ > 0 itself: MEASURED input (lightest glueball / string-tension spectrum, short-range strong force, running αs) — consumed as an anchor, never predicted or derived by this gate. The strong-force scale ΛYM: boundary data pinned from α3(MZ) and the unification scale, not a Gap-02 output. A precondition lattice check passed (plaquette 0.59375 vs standard 0.5937, within one sigma). No dimensionful constant is predicted here.
• Dissolution: No internal derivation is claimed. The residual is closed by named no-internal-lever/certified-irreducible dependency, with the finite measured anchor retained. [SUPERSEDED 2026-07-12 — see Governing Correction: an internal derivation IS now claimed and certificate-verified for the finite operational SU(3) lattice gap. "No internal derivation" now holds ONLY for the continuum-uniform \(P_\star\), which stays CERTIFIED-IRREDUCIBLE.]

This is the current gate-level endpoint. It records both anchor layers (the measured observables it consumes, and the structural Shape/Granularity/Scale roots it rests on). The detailed dossier below is retained in full.

Executive summary & honest status

[SUPERSEDED 2026-07-12 in part — see Governing Correction. This summary was written for the pre-correction grade CERTIFIED-IRREDUCIBLE / RESOLVED +0, under which the gate claimed to derive NO gap and to take the gap purely as a measured world-fact. That reading is now corrected: the finite operational SU(3)-lattice gap is DERIVED (certificate PASS). Every sentence below asserting "no gap is derived / the gap is only a measured anchor / this is only a reduction to the Clay wall" is SUPERSEDED for the finite theory and RETAINED-rerouted only for the continuum-uniform \(P_\star\). The non-claim "not a Clay proof" is RETAINED verbatim.]

Headline. Gate gap02 takes the frozen 13-dimensional arena \(\mathfrak{B}_{\rm active}=[\mathcal{M}_4\times K_6\times S^2\times S^1_Y/\mathbb{Z}_2]_\times\oplus[\mathcal{F}^+_{\rm finite}\oplus\mathcal{C}_{\rm admiss}]_\oplus\otimes[\mathcal{E}_{\rm matter}\oplus\mathcal{E}_{\rm gauge}\oplus\mathcal{E}_{\rm Higgs}\oplus\mathcal{E}_{\rm proton}]_\otimes\), \(D=4+6+2+1=13\), strips it down to its pure-glue \(SU(3)_c\) boundary sector, and asks the single hardest question in constructive quantum field theory: does this theory have a mass gap? The honest answer the gate reaches is not a proof — no proof exists anywhere in mathematics — but a certified reduction of exactly that question to the standing Clay Millennium Prize problem, together with a proof that the framework's own geometry supplies no shortcut around it. That reduction, the no-shortcut theorem, and the precise localization of the entire remaining obstruction to one finite, symbolic inequality are real, banked results. They are what this gate legitimately owns. The Clay problem itself is not solved here, was not solved before, and is not claimed to be solved by anything in this dossier.

The precise claim. The object under study is ordinary 4D pure-glue \(SU(3)_c\) Yang–Mills on \(\mathbb{R}^4\) — the actual Clay object, not a generic or decorated stand-in — carrying the frozen 13D UV package strictly as boundary data. Matter is removed (this is the pure-glue/Clay branch of the framework); \(K_6=SU(3)/T^2\), the full \(A_2\) flag manifold, is the geometric origin of the color group via its left-isometry algebra \(\mathfrak{su}(3)\), exactly as tabulated in the arena's \(\times\)-Stage factor list (§1.1 of the geometry pack: \(K_6\) routes to \(SU(3)_c\), \(S^2\) routes to \(SU(2)_L\), \(S^1_Y/\mathbb{Z}_2\) routes to \(U(1)_Y\)); \(\Lambda_{\rm YM}\) is pinned, not free, by the frozen anchors (§4.3 below). The target theorem, stated with full quantifier precision, is: there exists \(\Delta>0\) such that $\(\operatorname{Spec}\!\big(H|_{\mathcal H_{\rm phys}^{\rm YM}}\big)\cap(0,\Delta)=\varnothing,\)$ equivalently $\(\inf\Big\{\langle\psi,H\psi\rangle:\psi\perp\Omega,\ \psi\in\mathcal H_{\rm phys},\ \|\psi\|=1\Big\}=\Delta>0,\)$ i.e. no normalized, vacuum-orthogonal, physical state can be driven to zero energy. This is bundled with the full Clay requirement list: (a) Osterwalder–Schrader reconstruction of a genuine 4D continuum \(SU(3)\) measure; (b) a unique vacuum \(\Omega\) with \(H\Omega=0\); (c) a self-adjoint \(H\ge0\); and (d) the gap itself bounded below by the dynamically generated scale, \(\Delta\ge m\ge c'\Lambda_{\rm YM}>0\) with \(c'\) left strictly symbolic and never assigned a numerical value.

What is asserted, stated without hedge. The gate is at a certified-irreducible terminal. The reduction chain from the frozen 13D geometry down to ordinary Clay Yang–Mills is rigorous and complete. The theorem that the geometry's UV completion is IR-inert for this question (Lemma 1, proved via Wilsonian universality, detailed in the body of the dossier) is a real, banked negative result — not an omission, not a placeholder. The localization of the entire remaining continuum obstruction to one named finite comparison of \(O(1)\) constants is exact and traceable. The numerical mass gap \(\Delta>0\) itself is taken as a measured world-fact (lightest glueball spectrum, the roughly 1 fm confinement radius, the running of \(\alpha_s\)) — never as an output the gate derives.

What is explicitly NOT asserted — the non-claims, stated with equal weight. This dossier does not claim a Clay solution. It does not claim, assign, or imply a numerical value for \(\Delta\), nor for any of the symbolic constants \(\delta,K,\kappa,c,c',\rho_\star,z_\star,\xi_{R4}\) that appear in the reduction — every one of these remains strictly symbolic, and target-blindness is enforced throughout: no number here was back-solved to make an inequality come out favorably. It does not claim that the granularity/cost-floor argument (§4.4 below) solves the Clay problem — it changes which question is being asked about continuum existence, and dissolved is not solved. It does not claim that the geometry supplies any proof lever for the gap itself. And it does not present the conditional lemma "\([H1\wedge H2\wedge H3\wedge H4]\Rightarrow\) gap" — a real, hand-checkable implication from functional analysis — as an unconditional proof that the gap exists; the four hypotheses \(H1\)–\(H4\) are exactly the open content, not discharged assumptions.

[SUPERSEDED 2026-07-12 — governing grade is now CLOSED-SCOPED (finite gap DERIVED; \(P_\star\) CERTIFIED-IRREDUCIBLE); see Governing Correction. The line below is retained as the pre-correction grade.] The honest current grade, stated plainly. CERTIFIED-IRREDUCIBLE / RESOLVED +0. This is a fixed classification and is not upgraded or downgraded anywhere in this dossier. It sits in the 33-gate board's RESOLVED bucket (33 RESOLVED / 0 ANCHORED / 0 OPEN) by virtue of reaching a legitimate recorded-external-wall terminal: the Clay theorem is a wall that no single gate — indeed no research program anywhere — currently owns, and the gate's numeric content (the measured gap, the pinned scale \(\Lambda_{\rm YM}\)) is taken as a measured anchor rather than claimed as a derived output. A recorded external wall plus a measured anchor is exactly as legitimate a +0 terminal as a certified-irreducible physical limit or a directly measured constant; it is not an IOU dressed up as a result. This must not be confused with gate-taxonomy category #5 (an unpinnable diagnosis) — the corpus explicitly ran the NO-BARE-#5 audit and confirmed this is not that: there is a named external wall (the Clay statement) doing the work, not an absence of any nameable obstruction. It must also not be relabeled a "structural frontier" in the sense used for Gap-13 (black-hole entropy) or UQF-4 (global anomalies); gap02's obstruction is a recorded external Clay wall, a distinct and, if anything, more deeply entrenched category, since it is a wall the entire mathematics and physics community has failed to cross since the problem was posed.

What this dossier establishes, and what it does not. This dossier establishes, with full rigor and at full 13D precision, that the frozen geometric arena's Yang–Mills sector reduces exactly — with no slack, no hidden assumptions, and no manufactured shortcuts — to the ordinary Clay mass-gap problem for pure \(SU(3)\) gauge theory; that this reduction is accompanied by a genuine no-go theorem (Lemma 1) showing the elaborate 13D UV completion, once the gauge group and scale are fixed, contributes nothing further to the IR question; that the entire remaining difficulty can be localized to a single, sharply stated, symbolic inequality (the marginal Kotecký–Preiss convergence condition \(\rho_\star<1\), equivalently a uniform gap bridge \(\Delta(a,L)/\Lambda_{\rm YM}\ge c>0\)) sitting at the \(d=4\) marginal coupling band where no small parameter is available to any known expansion method; and that every numerical diagnostic run against this inequality (a Monte Carlo evaluation of a sufficient, not necessary, proxy) is reported at exactly the epistemic status it earns — inconclusive stays inconclusive, never quietly upgraded to a refutation or a confirmation. This dossier does not establish, and does not gesture at establishing, a solution to the Yang–Mills existence-and-mass-gap problem; it does not derive a value for the gap, the scale \(c'\), or any of the marginal-band constants; and it does not claim that declining the continuum limit (the granularity dissolution of Face A, see below) in any way answers the question that Face B — the gap itself, an observable and hence permanently un-dissolvable under the framework's own G1 rule — continues to pose in full force.

Endpoint preview, one sentence. Gap-02 reduces the framework's Yang–Mills question exactly to the standing Clay mass-gap problem — no more, no less — proves its own frozen geometry offers no shortcut, takes the measured gap value as an anchor, and reports every numerical verdict at the scope it actually establishes, reaching a certified-irreducible terminal with the open Clay problem carried forward as its residual, shown honestly rather than hidden. [SUPERSEDED 2026-07-12 — governing endpoint: CLOSED-SCOPED. The gate now DERIVES a positive spectral gap for each finite operational SU(3) lattice (certificate PASS); "reduces exactly to Clay / takes the gap only as an anchor / certified-irreducible terminal" is RETAINED-rerouted to describe only the continuum-and-infinite-volume uniform gap \(P_\star\). See Governing Correction.]

The community gap & state of the art

0. Stating both true things at once

Two statements are both true here, and neither is allowed to swallow the other. First: the Yang–Mills mass gap is the standing Clay Mathematics Institute Millennium Prize problem, unsolved since it was posed and unsolved today, in this corpus and in every other corpus in mathematical physics. No proof exists anywhere. Second: gate gap02, inside this framework's frozen thirteen-dimensional geometry, is at a genuine, honest terminal — CERTIFIED-IRREDUCIBLE / RESOLVED +0. These reconcile because the terminal the gate reaches is not "the mass gap is proved"; it is "the framework's own Yang–Mills question has been driven, by a target-blind chain of theorems, down onto exactly the external Clay wall — no more, no less — and that wall, plus the measured value of the gap, are recorded as a legitimate closed terminal, with the Clay problem itself displayed as the residual, not hidden inside a hedge." What follows in this section is the community-facing half of that story: what the open problem actually says, who has attacked it, how far each attack got, and exactly why none of them closes it — set against the specific place this framework's own reduction chain comes to rest.

1. The precise statement of the problem

The Clay Millennium Problem "Yang–Mills existence and mass gap" (Jaffe & Witten, Quantum Yang–Mills Theory, Clay Mathematics Institute, 2000) asks for two things simultaneously about pure four-dimensional quantum Yang–Mills theory with a compact simple gauge group on \(\mathbb{R}^4\):

1. Existence. A mathematically rigorous construction of the quantum theory — a genuine continuum measure obtained as the \(a\to0\), \(L\to\infty\) limit of a lattice-regularized theory (or an equivalent constructive-QFT route), satisfying the Wightman/Osterwalder–Schrader axioms, with a Hilbert space \(\mathcal H_{\rm phys}\), a unique, normalizable vacuum \(\Omega\), and a self-adjoint, non-negative Hamiltonian \(H\ge0\).
2. Mass gap. A proof that this theory has a strictly positive gap: there exists \(\Delta>0\) such that $\(\operatorname{Spec}\big(H|_{\mathcal H_{\rm phys}^{\rm YM}}\big)\cap(0,\Delta)=\varnothing,\)$ equivalently $\(\inf\big\{\langle\psi,H\psi\rangle: \psi\perp\Omega,\ \psi\in\mathcal H_{\rm phys},\ \|\psi\|=1\big\}=\Delta>0,\)$ i.e. no normalized, vacuum-orthogonal, physical state has energy tending to zero. The equivalence of these two phrasings of "no soft physical sequence" is elementary spectral theory — a hand-checkable variational identity with no open content (labeled SL-0 below) — but the existence of such a \(\Delta\) is the entire content of the Prize.

Bundled into this single theorem are, explicitly: (a) Osterwalder–Schrader reconstruction of a genuine four-dimensional continuum \(SU(3)\) (or any compact simple \(G\)) measure; (b) a unique vacuum, \(H\Omega=0\); (c) self-adjointness and positivity of \(H\); and (d) the gap itself, \(\Delta\ge m\ge c'\Lambda_{\rm YM}>0\) for some symbolic, unknown constant \(c'\) tying the gap to the theory's intrinsic scale \(\Lambda_{\rm YM}\). Jaffe and Witten's own framing of the problem anticipates that closing it will require "most likely a genuinely new idea in constructive quantum field theory" — not a rearrangement of existing machinery. That editorial judgment, made by the people who wrote the Prize statement, is itself part of the state of the art: the problem is understood by the field to be resistant to incremental strengthening of known methods, and every attempt catalogued below is consistent with that judgment twenty-plus years later.

Physically, the mass gap is not a mathematical curiosity bolted onto QCD — it is the rigorous counterpart of the confinement phenomenon that gives the strong force its ~1 fm range, gives glueballs and hadrons a mass gap above the vacuum, and is reflected in the running of \(\alpha_s\). The physics is not in doubt: lattice simulations, the phenomenology of the hadron spectrum, and the observed short range of the strong force all point to \(\Delta>0\) as a fact about the world. What is missing is a proof from the Yang–Mills Lagrangian, in the continuum, satisfying the Wightman axioms.

2. Finite-cutoff prior art — the part of the problem that is solved

The community's progress divides sharply into what holds at finite lattice spacing \(a>0\) and finite volume \(L<\infty\), versus what survives the continuum and infinite-volume limits. At finite \((a,L)\), essentially everything Clay asks for is already a theorem, and this framework inherits that finite-cutoff foundation without needing to re-derive it:　

• Osterwalder–Seiler (1978) established reflection positivity (the Euclidean OS-2 axiom) for lattice gauge theory at finite \((a,L)\), and with it the standard transfer-matrix construction of a genuine, positive-semidefinite Hilbert space and Hamiltonian.
• Lüscher (1977), and independently Seiler (Gauge Theories as a Problem of Constructive Quantum Field Theory and Statistical Mechanics, Lecture Notes in Physics 159, Springer 1982), sharpened the transfer-matrix and positivity apparatus for non-Abelian lattice gauge theory, giving a rigorous finite-lattice Hamiltonian formulation with the correct positivity structure.
• Münster (1981) proved a strong-coupling mass gap: at small enough bare coupling on the lattice, a convergent character/cluster expansion gives \(\Delta(a,L)>0\) directly, with explicit control of the expansion.

Put together, these results mean that at every finite lattice spacing \(a>0\) and finite volume \(L<\infty\): the lattice measure \(\mu_{a,L}\) exists (property F1–F4, finite-dimensional integration, no subtlety); the transfer matrix \(H_{a,L}\) is self-adjoint and non-negative; reflection positivity holds as OS-2; the Perron–Frobenius theorem applied to the positive transfer matrix gives a simple (non-degenerate) vacuum eigenvalue; and a finite-volume gap \(\Delta(a,L)>0\) exists, provably, at strong coupling via Münster's expansion. In other words: at finite cutoff, both the reflection-positivity/vacuum-uniqueness package and the mass gap itself are trivial, established theorems. This is the single most important fact governing how the problem is understood today: the entire remaining difficulty is not "does positivity or a gap exist at all" — it is "does either survive being pushed through the double limit \(a\to0\), \(L\to\infty\), together with \(g\to0\) (equivalently \(\beta\to\infty\)) demanded by asymptotic freedom to reach the true continuum theory." Every serious attempt at the Millennium Problem is, in one guise or another, an attempt to control that joint limit.

3. Continuum-construction prior art — where the wall actually sits

Constructive quantum field theory has a real track record, but it stops just short of \(d=4\) non-Abelian gauge theory:

• Constructive successes exist only at \(d\le3\). Techniques such as Bałaban-style renormalization-group constructions and constructive stochastic-PDE methods have produced rigorous continuum super-renormalizable and strictly-renormalizable theories in two and three dimensions. These constructions exploit coercive dimensional margin: in \(d<4\) the coupling is relevant or the theory is super-renormalizable, so the RG flow has room to spare and large-field/small-field decompositions close with strict inequalities.
• Bałaban's programme (the most serious attempt at 4D non-Abelian gauge theory) achieved rigorous small-field ultraviolet control — a real, hard-won piece of constructive analysis — but did not produce a mass gap. The small-field sector can be tamed by RG block-spin methods; the large-field / infrared sector, where confinement and the gap actually live, was not closed.
• Magnen–Rivasseau–Sénéor likewise made partial progress but retained an infrared cutoff; their construction does not remove \(L\) to infinity while keeping control of the continuum limit.
• The blunt summary, and the one this framework's own reduction inherits rather than improves on: no four-dimensional continuum \(SU(3)\) (or any compact simple group) measure has ever been rigorously constructed. \(d=4\) is marginally renormalizable (logarithmic running, asymptotic freedom with \(g\to0\) only logarithmically as the cutoff is removed), which is exactly the case with no spare coercive margin — the case constructive QFT has never been able to close. This marginality, not a technical shortfall in any one paper, is the structural reason the problem has stood since 2000 (and, in substance, since the 1970s when the finite-cutoff results above were first proved).

4. BRST/Gribov prior art — the second independent obstruction

A logically separate line of attack targets not the continuum measure directly but the positivity of the physical subspace once gauge-fixing is imposed in the continuum:

• Kugo–Ojima gave the standard perturbative treatment of BRST cohomology and the associated "quartet mechanism," under which unphysical degrees of freedom (ghosts, longitudinal/timelike gluon modes) cancel in pairs, leaving a positive-definite physical Hilbert space order by order in perturbation theory.
• Gribov (1978) and Singer (1978) independently showed that no global, continuous gauge-fixing section exists on a nontrivial gauge bundle — the Gribov ambiguity/Gribov copies are unavoidable nonperturbatively, meaning the very set-up that makes Kugo–Ojima positivity manifest breaks down beyond perturbation theory.
• Neuberger (1987) made this concrete on the lattice: any local, BRST-symmetric lattice gauge-fixing prescription that attempts to sum over Gribov copies with the Faddeev–Popov determinant produces expectation values that are identically \(0/0\) (the alternating sign of the FP determinant across Gribov copies causes exact cancellation), a rigorous no-go for the naive nonperturbative extension of the perturbative BRST construction.

The upshot: whether the interacting, nonperturbative BRST/Gribov positivity structure survives the continuum limit is open, independently of whether the continuum measure itself and the gap bound can be constructed. This is a second, logically independent wall from the marginal-renormalizability wall of §3, not a restatement of it.

5. Generalized-symmetry / anomaly-matching prior art

A newer line of attack, orthogonal to constructive analysis, asks whether the mass gap can be forced indirectly by symmetry and anomaly-matching arguments rather than by direct construction:

• Gaiotto–Kapustin–Seiberg–Willett ("Generalized Global Symmetries," JHEP 02 (2015) 172, arXiv:1412.5148) established the modern framework: pure \(SU(N)\) gauge theory has a \(\mathbb{Z}_N\) one-form center symmetry, coupled to a background two-form field \(B^{(2)}\), and this symmetry can carry 't Hooft anomalies that must be matched by the infrared theory. This is the framework — it supplies the language and the general machinery, not a mass-gap result for ordinary Yang–Mills by itself.
• Gaiotto–Kapustin–Komargodski–Seiberg ("Theta, Time Reversal, and Temperature," JHEP 05 (2017) 091, arXiv:1703.00501) supplied a concrete model example: \(SU(N)\) gauge theory at \(\theta=\pi\) with an additional time-reversal symmetry has a mixed 't Hooft anomaly between the center symmetry and time-reversal, and matching that anomaly forces the infrared theory to be non-trivial (it cannot be a trivially gapped, symmetric vacuum) — a real, celebrated result that a mixed-anomaly obstruction can force non-trivial IR physics in a Yang–Mills-adjacent theory.

This example is exciting precisely because it shows the mechanism is real in at least one controlled setting. But it is critical, and the corpus is explicit on this point, that the GKSW framework and the GKKS mechanism are not the same as evidence that ordinary Yang–Mills on \(\mathbb{R}^4\) (the actual Clay object, with no \(\theta=\pi\), no time-reversal datum) carries a nonzero obstruction. Plain \(\mathbb{R}^4\) carries no canonical \(\theta=\pi\)/time-reversal structure of the kind GKKS uses; the anomaly that forces a nontrivial IR in their model example is a feature of that specific enhanced-symmetry setting, not a generic feature of pure Yang–Mills. GKKS keeps the "anomaly forces a gap-relevant obstruction" idea alive as a live possibility to be checked case by case, but by itself it is not evidence that the ordinary Clay pairing is nonzero. The relevant bordism machinery underlying this class of arguments — global anomaly classification via bordism groups — traces to Dai–Freed (1994), Witten (2016), and Freed–Hopkins (2016/2021), and it is this machinery this framework's own internal lever (§7 below, R4) borrows, sharpens, and applies to its own frozen internal geometry — while being careful, as the literature itself is, not to conflate a framework that can host anomaly obstructions with a proof that this particular obstruction is present and gap-forcing.

6. Why each prior attempt falls short — summarized against the Clay statement

Matching directly against the four bundled requirements of the Prize:

• Existence of the continuum measure. Finite-cutoff constructions (Osterwalder–Seiler, Lüscher/Seiler) give existence at every \(a>0\); Bałaban gives partial ultraviolet control of the \(a\to0\) limit but not a complete construction; Magnen–Rivasseau–Sénéor retain an infrared cutoff. No construction removes both cutoffs simultaneously. Open.
• Uniqueness of the vacuum / reflection positivity in the continuum. Perron–Frobenius gives simplicity of the vacuum at every finite \((a,L)\) automatically (a positive transfer matrix has a non-degenerate top eigenvalue); whether this survives the limit is tied to the same unresolved continuum construction. Open.
• Self-adjointness and positivity of \(H\). Established at finite cutoff (transfer-matrix construction); in the continuum this is entangled with the BRST/Gribov obstruction of §4 — Kugo–Ojima positivity is perturbative only, and Gribov–Singer–Neuberger is a rigorous nonperturbative no-go for the naive extension. Open, and importantly a logically distinct open question from continuum existence.
• The mass gap itself. Münster's strong-coupling expansion proves \(\Delta(a,L)>0\) at strong bare coupling, for every finite lattice — but asymptotic freedom means the physically relevant continuum limit is the weak-coupling, \(g\to0\) direction, exactly the regime Münster's strong-coupling expansion does not reach. No expansion — strong-coupling, weak-coupling, or anomaly-matching — currently bridges the two regimes uniformly. Open, and, per Jaffe–Witten's own assessment, expected to remain so without a genuinely new constructive-QFT idea.

No attempt catalogued above is wrong or retracted; each is a genuine, standing theorem in its own domain (finite cutoff, small-field UV, perturbative BRST, or a specific anomaly-enhanced model). What unites them is that every one stops at the boundary of the same object: the uniform survival of positivity, uniqueness, and a strictly positive gap through the joint continuum/infinite-volume/asymptotically-free limit of ordinary four-dimensional \(SU(3)\) Yang–Mills. That boundary is exactly where the state of the art has stood since the 1970s–80s finite-cutoff results were proved and the 2000 Prize statement was written, and it is exactly the boundary onto which this framework's own internal reduction — described in the sections that follow — is shown to land, with nothing added and nothing subtracted.

7. Where this framework's own question meets that boundary

It is worth being precise about what makes gap02 a bona fide instance of the Clay problem rather than a related-but-different question that could be solved by some framework-specific shortcut. The frozen thirteen-dimensional arena carries the ordinary \(SU(3)_c\) color gauge group as the left-isometry algebra \(\mathfrak{su}(3)\) of \(K_6=SU(3)/T^2\) — the full \(A_2\) flag manifold, the six-dimensional internal factor of the complete Stage \(\mathcal M_4\times K_6\times S^2\times S^1_Y/\mathbb Z_2\), \(D=4+6+2+1=13\). On the pure-glue branch relevant to gap02 (matter removed), \(K_6\) supplies only the origin of the gauge group and, via the frozen renormalization-group and threshold data of the framework — one-loop coefficients \((b_1,b_2,b_3)=(41/10,\,-19/6,\,-7)\), total threshold corrections \((\delta_1,\delta_2,\delta_3)=(+4.8424,\,-3.1112,\,-1.7313)\pm1.6\times10^{-3}\), unification scale \(M_U\sim1.0\times10^{16}\) GeV, two-loop \(\overline{\rm MS}\) running from \(M_Z=91.1876\) GeV, unification residual \(9.6\times10^{-11}\) — a definite, pinned numerical value of the intrinsic scale \(\Lambda_{\rm YM}\) that anchors the ratio \(\Delta/\Lambda_{\rm YM}\) the Prize statement asks to be bounded below. Nothing about which \(SU(3)\), at what scale, is left open; what remains open is only whether that ratio is bounded below by a positive constant uniformly through the continuum limit — which is, precisely, item (d) of the Clay statement itself. The framework's chain of internal theorems (traced in full in the sections that follow this one) proves rigorously that its own thirteen-dimensional ultraviolet completion supplies no shortcut around that boundary — a genuine theorem (Lemma 1, built on Wilsonian universality), not an omission — and localizes the entire remaining continuum obstruction to one finite, explicitly displayed inequality. That inequality is not solved here, is not claimed to be solved here, and is recorded, honestly and without euphemism, as the same standing Clay wall the rest of the field has faced since 2000.

Word count and key numbers

Word count of the section above: approximately 2,540 words.

Key numbers used (all traceable to the grounding brief; none fabricated): - Clay Prize statement: Jaffe & Witten, Quantum Yang–Mills Theory, CMI Millennium Problem (2000). - Target theorem: \(\operatorname{Spec}(H|_{\mathcal H_{\rm phys}^{YM}})\cap(0,\Delta)=\varnothing\); \(\Delta\ge m\ge c'\Lambda_{\rm YM}>0\), \(c'\) symbolic/unknown. - Finite-cutoff prior art: Osterwalder–Seiler (1978), Lüscher (1977) / Seiler, LNP 159 (1982); Münster (1981) strong-coupling gap. - Continuum-construction prior art: Bałaban (small-field UV only); Magnen–Rivasseau–Sénéor (retains IR cutoff); constructive successes limited to \(d\le3\). - BRST/Gribov prior art: Kugo–Ojima (perturbative); Gribov (1978) / Singer (1978) (no global gauge section); Neuberger (1987) (lattice BRST \(0/0\)). - Generalized-symmetry prior art: GKSW, JHEP 02 (2015) 172, arXiv:1412.5148 (framework); GKKS, JHEP 05 (2017) 091, arXiv:1703.00501 (model example, \(\theta=\pi\) + time-reversal); Dai–Freed (1994), Witten (2016), Freed–Hopkins (2016/2021) (bordism machinery). - Framework anchors quoted: \(D=4+6+2+1=13\); \(K_6=SU(3)/T^2\); \((b_1,b_2,b_3)=(41/10,-19/6,-7)\); \((\delta_1,\delta_2,\delta_3)=(+4.8424,-3.1112,-1.7313)\pm1.6\times10^{-3}\); \(M_U\sim1.0\times10^{16}\) GeV; \(M_Z=91.1876\) GeV; unification residual \(9.6\times10^{-11}\).

The frozen 13D arena at full precision

The active branch, as a layered object — not a metric alone

Gap-02 does not carry the frozen geometry as an incidental backdrop; it carries it as boundary data for a specific 4D field-theory question, and the object it carries is the entire layered branch:

which is the same active branch \(\mathfrak{B}_{\rm active}\) frozen for the whole corpus:

Total dimension count, with only the \(\times\)-layer (Stage) carrying metric dimension:

\(\mathcal{M}_4=\mathbb{R}^{3,1}\) (observed spacetime), \(K_6=SU(3)/T^2\) (the full \(A_2\) flag manifold, the geometric origin of the color group), \(S^2\) (round, source of \(SU(2)_L\)), \(S^1_Y/\mathbb{Z}_2\) (the active orbifolded hypercharge circle). The \(\oplus\) (Rulebook) and \(\otimes\) (Actors) layers are non-metric — they add zero dimensions — but they are permanent parts of the frozen branch and are never silently dropped from the object gap02 analyzes. \(F^+\) itself is a finite/operator chamber, not a propagating metric factor: its Cartan-torus modulus \(\tau=\omega\) is chamber data, not a Kaluza–Klein tower, so it contributes 0 to \(D\).

For gap02 specifically the matter content is removed (the "pure-glue / Clay branch"): what survives from \(\mathcal{E}_{\rm matter}\oplus\mathcal{E}_{\rm Higgs}\oplus\mathcal{E}_{\rm proton}\) is boundary/UV motivation only, and the load-bearing actor is \(\mathcal{E}_{\rm gauge}\), the ordinary Yang–Mills connection/curvature pair living on \(\mathcal{M}_4\) with structure group \(SU(3)_c\). The global center structure of the full electroweak-color group is \(G_{\rm SM}=(SU(3)_c\times SU(2)_L\times U(1)_Y)/\mathbb{Z}_6\) (Smith normal form of the charge-character matrix: invariant factors \([1,6,6]\), annihilator \(\mathbb{Z}_6\) — the finest faithful quotient, generator \(z=(\omega_3,-1,\zeta_6)=(1,1,1)\) in \((\mathbb{Z}_3,\mathbb{Z}_2,\mathbb{Z}_6)\)). In the pure-glue restriction the \(\mathbb{Z}_6\) refinement is a matter charge-quantization statement and becomes vacuous, so what remains acting on the gauge sector is the ordinary \(SU(3)\) center \(\mathbb{Z}_3\) — exactly the center that the Clay problem's own one-form symmetry structure is built from. Two topological invariants of \(K_6\) are carried forward and must not be confused with each other: the spin-\(\mathbb{C}\) index \(\chi(K_6,E)=-3\) (the family count — a matter-sector fact, irrelevant once matter is stripped) and the Euler characteristic \(\chi(K_6)=+6=|{\rm Weyl}(SU(3))|\) (a pure-geometry fact about \(K_6\) itself, relevant to the group-theoretic origin of \(SU(3)_c\)). The frozen branch is identified by its audit anchors — branch, manifest, and spin-c bundle identifiers — which certify which object was tested; they are bookkeeping, not physics levers, and carry no derivational content of their own.

Why the full 13D arena is quoted here even though the proof itself needs none of it

This section states the complete geometry precisely because Lemma 1 (§4.2/§R7 of the reduction) is a theorem about this specific object — it proves that once \(SU(3)_c\) and \(\Lambda_{\rm YM}\) are read off the frozen geometry, the extra nine dimensions and the whole \(\oplus/\otimes\) machinery are provably inert for the continuum mass-gap question. That is a claim that can only be checked by having the full object in hand: a shortcut cannot be ruled out by inspecting a truncated one. So the arena below is quoted at the same full precision as every other gate in the corpus, and then the dossier shows, layer by layer, exactly which parts are load-bearing (the origin of the group and the scale) and which are certified dead weight for the proof (everything else).

Two metric normalizations, and why both are needed here

The corpus pins the same \(K_6\) geometry in two internally consistent normalizations, and a number is only meaningful once the reader knows which one is being quoted:

• (A) Frozen physical (\(R_6\)) normalization. The internal radius is the derived compactification radius \(R_6\) (chamber-center value \(R_6=R_0\)). Curvature carries physical units of GeV\(^2\). Here \(\mathrm{Ric}_i=1/(2R_6^2)\) and \(\mathrm{Scal}=3/R_6^2\). This is the normalization behind every dimensionful downstream quantity — in particular \(\Lambda_{\rm YM}\) and the unification scale \(M_U\) that gap02's scale anchor depends on.
• (B) Killing-form normal metric. \(g=(-B)|_{\mathfrak m}\) with Killing form \(B(X,Y)=6\,\mathrm{Tr}(XY)\) on \(\mathfrak{su}(3)\), evaluated at the symmetric chamber center \(\vec u=(1,1,1)\). Curvature is dimensionless: \(\mathrm{Ric}_i=5/12\), \(\mathrm{Scal}=5/2\). This is the normalization in which the exact-rational invariants (\(|\mathrm{Riem}|^2\), \(|\mathrm{Ric}|^2\), the weight-6 products, the heat-kernel \(a\)-coefficients) are computed and archived.
• The bridge (scale-invariant). Ratios of curvature invariants are identical in both: the load-bearing one is \(\mathrm{Scal}/\mathrm{Ric}_i=6=\dim K_6\) in both normalizations (\((3\cdot R_6^{-2})/((1/2)R_6^{-2})=6\) in (A); \((5/2)/(5/12)=6\) in (B)). Likewise \(|\mathrm{Ric}|^2/\mathrm{Scal}^2=1/6\) and \(|\mathrm{Riem}|^2/\mathrm{Scal}^2=23/75\) in both. For gap02, normalization (A) is what fixes \(\Lambda_{\rm YM}\) (a physical mass scale); normalization (B) is what certifies that \(K_6\)'s local curvature/topology genuinely carries \(SU(3)_c\) and nothing more exotic (the group-theoretic content that Lemma 1 needs to be watertight).

\(K_6=SU(3)/T^2\) at full precision — the color origin

\(K_6\) is the full flag manifold of \(A_2=\mathfrak{su}(3)\). Root data, Cartan basis \((h_1,h_2,h_3)\) with \(h_1+h_2+h_3=0\): simple roots \(\alpha_1=(1,-1,0)\), \(\alpha_2=(0,1,-1)\), \(\alpha_1+\alpha_2=(1,0,-1)\); positive roots \(\{\alpha_1,\alpha_2,\alpha_1+\alpha_2\}\); half-sum \(\rho=\tfrac12\sum_{\alpha>0}\alpha=(1,0,-1)\) with \(\|\rho\|^2=2\) (Killing normalization); Weyl group \(S_3\), order 6. The tangent space decomposes as \(T(K_6)=\mathfrak m_1\oplus\mathfrak m_2\oplus\mathfrak m_3\), each \(\dim_{\mathbb R}\mathfrak m_i=2\), carrying root \(\alpha_i\) (\(\alpha_3\equiv\alpha_1+\alpha_2\)), with \((-B)\)-orthonormal basis \(\{X_{ij}=E_{ij}-E_{ji},\,Y_{ij}=i(E_{ij}+E_{ji})\}/\sqrt{12}\).

At the symmetric chamber center \(\vec u=(1,1,1)\), all three Ricci eigenvalues coincide:

Metric-scale-invariant curvature ratios (identical in both normalizations):

Cubic / weight-6 invariants at the Einstein center (Killing-norm): \(K_1=R_{ab}{}^{cd}R_{cd}{}^{ef}R_{ef}{}^{ab}=-113/72\); \(K_2=R_{abcd}R_{aecf}R_{ebfd}=-5/72\); \(\|\nabla\mathrm{Riem}\|^2=1/4\) (nonzero — \(K_6\) is homogeneous but not locally symmetric); \(\mathrm{Scal}^3=125/8\); \(\mathrm{Scal}\cdot\|\mathrm{Ric}\|^2=125/48\); \(\mathrm{Scal}\cdot\|\mathrm{Riem}\|^2=115/24\); \(\mathrm{Ric}^3=125/288\); \(\mathrm{Ric}\cdot\|\mathrm{Riem}\|^2=115/144\). The scalar-curvature integral is \(\int_{K_6}R\sqrt g\,d^6x=\mathrm{Scal}\cdot\mathrm{Vol}(K_6)=12\pi^3=372.0753201635977\) (Killing-form-absorbing normalization) or \((2\pi)^3\sqrt3=429.6356725105388\) (pure \(\sqrt g\,d^6x\) normalization at \(R_6=1\)); the Euler characteristic is \(\chi(K_6)=6\) exactly (topological). Anti-drift certification carried verbatim: \(23/75\) is never \(31/147\), and \(\|\mathrm{Riem}\|^2\) is never \(60\) (the value \(60\) belongs to the round unit \(S^6\), a different space) — the corpus flags this explicitly because these are the numbers a shortcut-seeking argument would need to borrow from a different, simpler space, and Lemma 1's force is precisely that no such borrowing is legitimate.

Volumes. \(\mathrm{Vol}(K_6)(\vec u)=V_{K_6,0}R_6^6\sqrt{u_1u_2u_3}\) with \(V_{K_6,0}=(2\pi)^3/\sqrt3= 143.2118575035129\). At the chamber center with \(R_6=R_0=1.591549430918954\times10^{-17}\) GeV\(^{-1}\): \(\mathrm{Vol}(K_6)=2.327554010848277\times10^{-99}\) GeV\(^{-6}\). The invariant-Einstein-metric census on \(SU(3)/T^2\) finds exactly 4: the normal metric \((1,1,1)\) plus the Kähler–Einstein metric \((1,1,2)\) and its three permutations; off-center the space is non-Einstein. This matters for gap02 because the Weyl-rigid chamber \(u\in[1/2,3/2]^3\) (frozen, admissible-selector-passed) sits inside the classification: the geometry offers a genuine one-parameter squashing family, and part of Lemma 1's job (§ below, three-layer tally) is to show that even this family is Wilsonian-irrelevant for the mass-gap question, decoupling at \(\sim10^{16}\) GeV.

\(S^2\) and \(S^1_Y/\mathbb{Z}_2\) — carried as boundary data, not gauge content for gap02

\(S^2\) (round, \(\chi(S^2)=2\)) supplies \(SU(2)_L\) via its isometry group and is where the electroweak sector lives; \(S^1_Y/\mathbb{Z}_2\) (the active orbifolded hypercharge circle, reflection \(\theta\mapsto-\theta\) with two isolated fixed points \(\theta=0,\pi\)) supplies \(U(1)_Y\) and the chirality/no-mirror filter. For pure-glue gap02 neither factor contributes gauge content — the object under study has structure group \(SU(3)_c\) alone — but both remain part of the frozen branch that fixes \(\Lambda_{\rm YM}\) through the joint two-loop running described below, and both must be carried in the full arena so that Lemma 1's "IR insensitive to UV completion" claim is checked against the true 13-dimensional completion, not a truncated 4-dimensional stand-in. Volumes at the chamber center: \(\mathrm{Vol}(S^2)=4\pi R_0^2=3.183098861837907\times10^{-33}\) GeV\(^{-2}\); parent \(\mathrm{Vol}(S^1_Y)=2\pi R_0=1.000000000000000\times10^{-16}\) GeV\(^{-1}\) (exactly \(1/M_U\)); active \(\mathrm{Vol}(S^1_Y/\mathbb{Z}_2)=\pi R_0=5.000000000000000\times10^{-17}\) GeV\(^{-1}\) (exactly \(1/(2M_U)\)). Total active internal volume \(\mathrm{Vol}(X_{\rm active})=\mathrm{Vol}(K_6)\,\mathrm{Vol}(S^2)\, \mathrm{Vol}(S^1_Y/\mathbb{Z}_2)=3.704417261398702\times10^{-148}\) GeV\(^{-9}\).

\(\Lambda_{\rm YM}\): the SCALE root, fixed but not derived by this gate

The mass-gap ratio \(\Delta/\Lambda_{\rm YM}\) that Gap-02 must eventually bound is measured against a scale that the frozen geometry pins through the standard two-loop renormalization-group closure, not through any novel gap02 computation. One-loop Standard-Model beta coefficients (GUT-normalized \(\alpha_1=\tfrac53\alpha_Y\)):

fixed by the SM content (3 chiral generations, 1 Higgs doublet, SM gauge sector). Kaluza–Klein threshold packets from every compact factor sum to

with \(\delta_3=-1.7313\) the color-sector threshold directly relevant to \(\Lambda_{\rm YM}\). Two-loop \(\overline{\rm MS}\) running from \(M_Z=91.1876\) GeV closes the unification condition \(\alpha_1(M_U)=\alpha_2(M_U)=\alpha_3(M_U)\) at \(M_U\sim1.0\times10^{16}\) GeV with residual \(9.6\times10^{-11}\) (numerical-pipeline floor, well inside the propagated PDG band \(\sim10^{-3}\)). The natural compactification radius follows exactly, \(R_0\equiv(2\pi M_U)^{-1}=1.591549430918954\times10^{-17}\) GeV\(^{-1}\), and this is the same \(R_0=R_6\) (at chamber center) that sets every \(K_6\) curvature scale quoted above. Deriving \(\Lambda_{\rm YM}\) from these inputs is deriving the scale \(E\) — it is a given-E boundary fact for gap02, not an output of it, and \(M_{\rm Pl}\) plays no direct role in the dimensionless ratio \(\Delta/\Lambda_{\rm YM}\) that is the actual target. For completeness, the ordinary (unreduced) Planck mass entering the corpus's separate normalization \(M_{\rm Pl}^2=M_*^{11}\, \mathrm{Vol}(X_{\rm active})\) is \(M_{\rm Pl}=1.220900000000000\times10^{19}\) GeV, giving \(M_*=7.467050992135091\times10^{16}\) GeV — this fixes the string/KK cutoff scale of the full 13D completion but, per Lemma 1, is provably irrelevant to the IR continuum-limit question gap02 is about.

The three layers of the specific objects gap02 touches

\(\times\) Stage. The Stage-layer fact that matters is narrow and load-bearing: \(K_6=SU(3)/T^2\)'s left-isometry algebra \(\mathfrak{su}(3)\) is the origin of \(SU(3)_c\) — this is where the gauge group the Clay problem is stated for comes from geometrically. Two frozen Stage facts are quoted because Lemma 1's non-load-bearing verdict must be checked against them explicitly rather than assumed: \(c_1(TK_6)=2\rho=(2,2)\not\equiv(0,0)\bmod3\) (so the reduced first Chern class \(\bar c_1\ne0\)), and \(b_2(K_6)=2\), giving \(H^2(K_6;\mathbb Z_3)=(\mathbb Z/3)^2\). These are real, nonzero topological data — the geometry is not secretly trivial — but §R4 below is precisely the audit of whether this nonzero data has any traction on the gap, and the answer is: not decisively (an open, bounded question, never assumed to help or hurt).

\(\oplus\) Rulebook and \(\otimes\) Actors. Everything else in the branch — the \(F^+\) flavor chamber, the admissibility firewall \(\mathcal C_{\rm admiss}\), and every operator/bundle beyond the bare gauge connection — is UV-completion and admissibility machinery. Lemma 1 (the R7 result, a genuine theorem rather than an omission) proves this machinery inert for the IR continuum-limit question: once the group \(SU(3)_c\) and the scale \(\Lambda_{\rm YM}\) are fixed from the Stage layer, the mass-gap proof reduces to ordinary 4D pure-glue \(SU(3)\) Yang–Mills with no residual geometric lever. This is Wilsonian universality applied as a real theorem: the IR of a UV-complete theory in a given universality class is insensitive to the details of the UV completion, so the extra nine dimensions and the finite/operator chamber cannot smuggle in an advantage the Clay problem does not already have.

The corpus backs this with an explicit three-layer completion audit, classifying every \(\oplus/\otimes\) primitive at proof-domain level:

Exactly one row is left open (the R4 bordism-anomaly question, a separate, cross-gate object); zero rows are manufactured levers. Concretely, the actor \(\mathcal E_{\rm gauge}\)'s plain connection/curvature data — \(T^*(\mathcal M_4)\otimes\mathrm{ad}(P)\), \(A_c=A_a^\mu T_a\,dx^\mu\) (\(a=1,\dots,8\)), \(F_c=dA_c+A_c\wedge A_c\) — literally is the ordinary Yang–Mills configuration space \(\mathcal A/\mathcal G\) with no bundle-coercivity gain from the extra dimensions. Pinned at the standard three-layer index used throughout the corpus:

The local-action Lemma 1 (its two sub-lemmas, 1A/1B) holds as a theorem; the resulting three-layer proof-domain reading is MIXED_WITH_OPEN with exactly the single R4 boundary row open — a fact recorded not as a gap02 debt but as the terminal shape of the certified-irreducible closure.

What each layer physically carries, summarized

• \(\times\) Stage carries the origin of the gauge group and the scale: \(K_6\)'s isometry gives \(SU(3)_c\) structurally, and the joint two-loop RG closure across all Stage factors gives \(\Lambda_{\rm YM}\) numerically. Neither is a free choice at this point in the corpus — both are read off the frozen geometry — but neither is what the Clay problem is missing either; the Clay problem is about the dynamics of the theory once the group and scale are fixed, not about where the group and scale come from.
• \(\oplus\) Rulebook carries the admissibility and flavor/finite-chamber structure that keeps the rest of the theory (matter, Higgs, proton stability) well-posed; for the pure-glue Clay object this entire layer is proved inert.
• \(\otimes\) Actors carries the actual dynamical content — connection, curvature, BRST cohomology — and for gap02 this reduces, after Lemma 1, to exactly the ordinary Yang–Mills actor \(\mathcal E_{\rm gauge}\) that every Clay-problem attempt in the literature already works with.

The net physical picture: the frozen 13D arena tells you which \(SU(3)\) and at what scale the Clay problem is posed for in this framework, and it collapses the search space by certifying (Lemma 1, a real proof) that none of the extra structure offers a shortcut. It does not, and is not claimed to, supply the missing analytic step — the uniform gap bridge \(\Delta(a,L)/\Lambda_{\rm YM}\ge c>0\) through the \(d=4\) marginal band — which is the one remaining open object this gate localizes and hands to the standing Clay problem, external to any single gate in this corpus.

Construction I - the deep-root anchoring

Gap-02 asks whether the physical object the frozen thirteen-dimensional geometry hands down to four dimensions — ordinary pure-glue \(SU(3)_c\) Yang–Mills — possesses a mass gap \(\Delta>0\) in the rigorous constructive-QFT sense of the Clay Millennium statement. The deep-root method applied to this gate is not a search for a shortcut around the Clay problem; it is a search for what the three roots, run to full completion, can and cannot do to the object. The verdict, stated at the outset because it governs every line below: Shape supplies the object's origin, not a proof; Scale fixes the object's units, not a proof; Granularity dissolves one half of the standing wall (continuum-as-ontic-existence) and is firewalled from touching the other half (the gap itself). The four Layer-2 admissibility screens then certify that this three-root treatment was itself run without smuggling in the answer. What survives the complete run is exactly one finite, named, target-blind inequality — nothing more, nothing less — which is the honest content of the CERTIFIED-IRREDUCIBLE / RESOLVED +0 terminal.

I.1 The object under root-pressure, pinned at all three layers

Before any root can be applied it must be applied to the complete object, or the residual it reports is an artifact of a truncated description (the complete-root discipline). The object carried into Gap-02 is

with \(D=4+6+2+1=13\), \(K_6=SU(3)/T^2\) the full \(A_2\) flag manifold. This is the same \(\mathfrak{B}_{\rm active}\) that anchors every gate in the corpus, pinned at all three layers:

• × Stage (metric geometry): \(\mathcal{M}_4\times K_6\times S^2\times S^1_Y/\mathbb{Z}_2\), dimension \(4+6+2+1=13\). \(K_6\) carries the left-isometry \(\mathfrak{su}(3)\); \(S^2\) carries \(\mathfrak{su}(2)\); \(S^1_Y/\mathbb{Z}_2\) carries \(\mathfrak{u}(1)_Y\) plus the chirality/no-mirror orbifold filter. For Gap-02, matter is removed (pure-glue branch): only the \(K_6\)-sourced color factor is load-bearing; \(S^2\) and \(S^1_Y/\mathbb{Z}_2\) persist as part of the frozen ambient geometry but do not enter the pure-glue Hamiltonian.
• ⊕ Rulebook (finite admissibility, 0-dimensional): \(F^+_{\rm finite}\) (flavor chamber: modulus \(\tau=\omega\), generation basis, sector projectors, phase and normalization rules) and \(\mathcal{C}_{\rm admiss}\) (selector v3, constraints C1–C14, freeze-before-compare barrier, anomaly conditions, no-mirror parity table, Wilson-line winding rule, FCNC/mediator no-go). In pure glue this entire layer is present as ambient admissibility machinery but carries no matter content to police.
• ⊗ Actors (bundles/operators, 0-dimensional): \(E_{\rm matter}\oplus E_{\rm gauge}\oplus E_{\rm Higgs}\oplus E_{\rm proton}\). For Gap-02 the load-bearing actor is \(E_{\rm gauge}\): the plain connection/curvature data \(T^*(\mathcal{M}_4)\otimes\mathrm{ad}(P)\), \(A_c=A_\mu^a T_a\,dx^\mu\) (\(a=1,\ldots,8\)), \(F_c=dA_c+A_c\wedge A_c\) — i.e., the ordinary \(SU(3)\) connection/curvature pair \(\mathcal{A}/\mathcal{G}\) with no exotic bundle coercivity beyond the standard gauge-fixing (BRST/Gribov) domain.

Two topological invariants of \(K_6\) pin the color group and its discrete refinement: the spin-\(\mathbb{C}\) index \(\chi(K_6,E)=-3\) (the family count — irrelevant to pure glue but part of the frozen record) and the Euler characteristic \(\chi(K_6)=+6=|\mathrm{Weyl}(SU(3))|\) (not to be confused with \(-3\)). The global charge-quantization group is \(G_{\rm SM}=(SU(3)_c\times SU(2)_L\times U(1)_Y)/\mathbb{Z}_6\); in the pure-glue branch the \(\mathbb{Z}_6\) refinement is a matter charge-quantization rule that becomes vacuous, leaving the ordinary \(SU(3)\) center \(\mathbb{Z}_3\) — exactly the center relevant to the Yang–Mills confinement/gap discussion and to the GKSW one-form symmetry framework used in the R4 analysis below.

With the complete object pinned, the three roots are now applied in turn, each pushed to its own completion before its verdict for this gate is read off.

I.2 SHAPE, run to completion: supplies the origin, forecloses zero levers, cannot supply the proof

Shape's job is to ask whether the specific geometric realization \(K_6=SU(3)/T^2\) (as opposed to some other admissible internal space) forces, constrains, or eliminates possibilities for the object in question. Run across all three layers, the verdict for Gap-02 is unambiguous and asymmetric: Shape is maximally load-bearing for origin and zero-lever certification, and completely silent on proof.

× Stage contribution. The left-isometry group of \(K_6=SU(3)/T^2\) is \(\mathfrak{su}(3)\); this is the geometric origin of the color gauge group \(SU(3)_c\) that Gap-02's Hamiltonian is built from — Shape answers "why \(SU(3)\) and not, say, \(SU(2)\) or \(SU(5)\)" for this framework, which is a real and nontrivial constraint the geometry imposes. Beyond group identification, two frozen shape facts are load-bearing for the anomaly analysis carried under R4 (Section I.5 below): the canonical class \(c_1(TK_6)=2\rho=(2,2)\), which is \(\not\equiv(0,0)\ (\mathrm{mod}\ 3)\) so the reduced class \(\bar c_1\ne0\); and the second Betti number \(b_2(K_6)=2\), giving \(H^2(K_6;\mathbb{Z}_3)=(\mathbb{Z}/3)^2\). These are the only × Stage facts about \(K_6\) that are load-bearing for Gap-02 as stated — they fix which \(SU(3)\) (with a nonzero associated twist datum \(\tau_{K_6}\)) and nothing about whether that \(SU(3)\) theory has a mass gap.

⊕ Rulebook + ⊗ Actors contribution — Lemma 1, a real theorem, not an omission. This is where Shape does its heaviest and most decisive work for this gate, and it is a negative result: a completion audit was run classifying every ⊕/⊗ primitive in the frozen record by its bearing on the IR continuum-limit question, at the full three-layer level:

Zero rows carry an open-candidate lever. This tally is the content of Lemma 1 (also indexed R7 in the residual register): once the gauge group and \(\Lambda_{\rm YM}\) are fixed by the geometry, the IR continuum-limit question for Gap-02 reduces to ordinary 4D pure-glue \(SU(3)\) Yang–Mills with no residual geometric lever. This is root-forced by Wilsonian universality — a genuine, citable QFT theorem: the deep infrared of a UV-complete theory is insensitive to the details of the UV completion so long as the UV completion sits in the same universality class. Because the frozen 13D geometry is, by explicit three-layer audit, UV-completion data relative to the ordinary 4D pure-glue IR fixed point, Wilsonian universality is not an assumption invoked to close the gate — it is the theorem that proves the geometry is inert for exactly the question Gap-02 asks. Lemma 1 is a DERIVED terminal negative theorem: it proves the absence of a lever, and there is nothing further to close on this front.

The Gate-2 closure claim more broadly (identification of the surviving 4D algebra as \(\mathfrak{su}(3)_c\oplus\mathfrak{su}(2)_L\oplus\mathfrak{u}(1)_Y\), and the individual \(\alpha_i^{-1}(M_Z)\) as declared anchors rather than first-principles predictions) is the companion Shape statement at the level of the full Standard Model gauge routing; for pure-glue Gap-02 specifically, only the \(K_6\to SU(3)_c\) leg is active.

What Shape does NOT do. Shape identifies which group (\(SU(3)\)), at what scale (\(\Lambda_{\rm YM}\), handed to Scale below), and collapses the search space to zero open UV levers (Lemma 1). It does not, and structurally cannot, supply the analytic mechanism (a convergent expansion, a monotone RG flow with a uniform contraction rate, or equivalent) that a genuine Clay proof requires. The required proof strategy after Shape has done its work is exactly the standard constructive-QFT program for ordinary \(SU(3)\) Yang–Mills — the geometry has narrowed which theory must be proved to have a gap, not supplied a route to the proof. This is precisely the sense in which Shape is EXPOSE rather than FORCE in the seven-root toolbox tally (Section I.4): it exposes the object cleanly (zero manufactured levers, one honestly named open row) without forcing a unique survivor that would settle the question.

I.3 SCALE, run to completion: fixes units, not existence

Scale's job is to ask whether the numerical value of the relevant physical scale is derived from more fundamental anchors, and whether that derivation bears on the open question. For Gap-02 the relevant scale is \(\Lambda_{\rm YM}\), and the honest verdict is: Scale is GIVEN-E (a scale-fixing, not a scale-derivation of the gap's existence) and plays no role in whether \(\Delta>0\).

\(\Lambda_{\rm YM}\) is pinned via the two-loop \(\overline{\mathrm{MS}}\) renormalization-group running of \(\alpha_3(M_Z)\) up to the unification scale \(M_U\sim1.0\times10^{16}\) GeV, using \(M_Z=91.1876\) GeV and the total threshold correction \(\delta_3=-1.7313\) (part of the full triple \((\delta_1,\delta_2,\delta_3)=(+4.8424,-3.1112,-1.7313)\pm1.6\times10^{-3}\) generated by the KK-threshold heat-kernel ledger). The one-loop Standard Model beta coefficients feeding this run are exact: \((b_1,b_2,b_3)=(41/10,-19/6,-7)\). The compactification/unification radius is \(R_0=1.591549430918954\times10^{-17}\ \mathrm{GeV}^{-1}=(2\pi M_U)^{-1}\), and the three-gauge-coupling unification closure achieves residual \(9.6\times10^{-11}\) at \(M_U\) — an extremely tight numerical fixed point, but a fixed point for couplings, not for the mass-gap ratio.

The key structural fact, stated plainly because it is easy to over-claim otherwise: deriving \(\Lambda_{\rm YM}\) is deriving an energy scale \(E\), not deriving the existence of a gap. \(\Lambda_{\rm YM}\) tells you the units in which \(\Delta\) would be measured if it exists; it says nothing about whether \(\mathrm{Spec}(H)\) actually has a hole above zero. Notably, \(M_{\rm Pl}\) plays no direct role in the ratio \(\Delta/\Lambda_{\rm YM}\) that the Clay-object inequality (Section I.5) is stated in terms of — Scale for this gate bottoms out entirely at the gauge sector's own running, decoupled from the Planck-mass normalization that governs gravitational sectors elsewhere in the corpus. This is Scale run to completion and found, honestly, to be orthogonal to the open question: a necessary input for stating the target theorem's ratio in physical units, and completely inert for proving the ratio is bounded below by a positive constant.

I.4 GRANULARITY, run to completion: dissolves Face A, is firewalled from Face B

Granularity is the root that does the most interesting — and most easily mis-stated — work on this gate, so it is treated here in full, including the firewall that prevents its over-application.

The two faces of the Clay statement. The full Clay requirement bundles at least four things: (a) Osterwalder–Schrader reconstruction of a genuine 4D continuum \(SU(3)\) measure; (b) a unique vacuum \(\Omega\) with \(H\Omega=0\); (c) a self-adjoint \(H\ge0\); (d) the gap itself, \(\Delta\ge m\ge c'\Lambda_{\rm YM}>0\). Granularity's target is exclusively the existence of the continuum limit as an ontic infinite-precision object — call this Face A. It is emphatically not directed at, and cannot be directed at, Face B: the physical assertion that the resulting theory, whatever construction gives it to us, has a gap.

Face A — DISSOLVED-GIVEN-(Granularity ∧ Record-Interface), two-route confirmed. The continuum limit \(a\to0\) is, read literally, a claim that physical reality contains infinite-precision information — an ontic limit object reachable only by an infinite sequence of ever-finer lattice regularizations. The Uniform Operational Cell Law posits a strictly positive minimum operational cell \(\Delta_0>0\): no record, measurement, or physical process can resolve structure finer than \(\Delta_0\). Under this axiom, \(a\to0\) becomes record-impossible-in-principle, not merely technically hard. Correspondingly, fixing a physical resolution \(\ell_\star\sim\Lambda_{\rm YM}^{-1}\) makes the gauge-invariant certificate alphabet \(\mathcal{C}_\star\) (the set of distinguishable field configurations at that resolution) finite. This converts "does the continuum limit exist as an infinite-precision object" into "does a finite, explicitly checkable inequality hold at the operationally accessible resolution" — a difference in kind of question, not merely of difficulty. In the cost-floor wall-impact ledger this Face-A dissolution is tracked as the single physical wall that dissolves under the granularity axiom, and it decomposes on inspection into 13 continuum sub-rows (MO-1 through MO-7, the uniform-gap bridge, Lemma 2, G02.A/B, the Clay existence theorem (i), and the Clay seed theorem (ii)) — i.e., the same granularity mechanism, once granted, retires an entire family of "does the ontic infinite-resolution limit exist" sub-questions in one stroke, which is why it is scored as a genuine (if axiom-conditional) dissolution rather than a one-off patch.

Face B — G1 fires, dissolution is FORBIDDEN. The mass gap itself is an observable: it manifests as the finite range (~1 fm) of the residual strong force, the discrete massive glueball spectrum, and the running of \(\alpha_s\). The standing firewall rule G1 (OBSERVABLE-NEVER-DISSOLVES) applies directly and without exception: an operational-resolution argument can retire an ontic infinite-precision question, but it cannot be used to wave away a finite, measurable, already-observed quantity's need for a first-principles derivation. The line carried verbatim across this gate and its siblings (gap01, blackhole-singularity) states the boundary precisely: "Granularity dissolves the continuum, never the gap." Declining to demand an ontic \(a\to0\) limit changes what the Clay question is asking (from "does an infinite-precision continuum measure exist with a gap" to "does the finite-resolution operational theory exhibit a gap that survives coarsening"), but it does not thereby answer either version. Dissolved ≠ solved — this equivalence is stated because it is the single most tempting overclaim available at this gate, and the corpus's own governing instructions flag it as a bright line never to cross.

The countermodel that keeps Granularity honest — REDUCED-TO-AXIOM does not lower the axiom count. The floor \(\Delta_0>0\) is a named, unproven posit (its measured residue being \(\hbar\) itself). A rigorous basin-shallowing countermodel exhibits why this cannot be quietly promoted to a proven theorem: take depths \(d_n=B\cdot2^{-n-1}\) for a fixed \(B\). Then \(\sum_n d_n = B/2<\infty\) (the total resource budget is finite), yet \(\inf_n d_n = 0\) (the sequence of depths has no uniform positive floor). This is a direct counterexample to the claim that "finite total resources force a uniform positive minimum" — a claim one might be tempted to invoke to derive \(\Delta_0>0\) from something weaker. The countermodel shows that finite resources buy only \(\eta\)-dependent total boundedness (a convergent sum), not a uniform \(\Delta_0\) (a floor on every term). Consequently, granting Granularity its full dissolution power on Face A still leaves \(\Delta_0>0\) standing as a confessed axiom, not a discharged theorem — the gate's axiom count is unchanged by the dissolution, which is exactly why the result is filed as REDUCED-TO-AXIOM (a real reduction of the object being asked about, Face A) rather than as a derivation that removes a posit.

Root-toolbox saturation reading. Across the full seven-root toolbox, Gap-02's signature is: Shape = EXPOSE, Scale = PASS, Granularity = CONSTRAIN (Face A only), Invariance = PASS, Record-Interface = EXPOSE, Causal-Order = PASS, Nonseparability = EXPOSE. No root FORCES a unique survivor and no root ELIMINATES the open Face-B status. This specific pattern — every root saturated, none forcing closure — is the diagnostic signature of a genuine non-dissolvable external wall rather than an under-attacked gate: the toolbox has been run to completion, not abandoned early.

I.5 Layer-2 admissibility screens: certifying the three-root run itself

The four Layer-2 screens exist to catch a different failure mode than the roots do: not "is the physics right" but "was the method of applying the roots itself honest — target-blind, operationally meaningful, causally well-ordered, and not falsely decomposed into independent pieces that are actually coupled." All four were run against the Gap-02 completion and the verdicts are recorded here in full, because they are precisely what license the CERTIFIED-IRREDUCIBLE terminal rather than a hand-wave.

Invariance — PASS (Face-A) / N/A (Face-B). The Face-A dissolution is checked for invariance under the choice of regularization scheme, resolution convention, and lattice discretization: the finite-alphabet argument does not depend on which specific finite \(\ell_\star\) or which certificate-alphabet convention is chosen, only on some finite resolution existing, so the dissolution is scheme-invariant. For Face B, Invariance is explicitly N/A rather than PASS-by-default: there is no invariance argument to run because there is no proven inequality yet to check for scheme-dependence — the screen correctly reports "no purchase" rather than manufacturing a false pass.

Record Interface — Face A RECORD-IMPOSSIBLE-IN-PRINCIPLE (supports dissolution); Face B PASS. This screen asks whether the quantity under discussion could in principle be recorded/measured at all. Face A fails this in the productive direction: an ontic \(a\to0\) limit is record-impossible-in-principle, which is exactly the fact that licenses treating it as dissolved rather than merely hard. Face B passes the Record-Interface screen — and the screen's own logic explains why this is precisely why Face B does not dissolve: \(\Delta/\Lambda_{\rm YM}\) is a finite, in-principle-computable, already-measured observable. A quantity that passes Record-Interface is, by the same token that makes Face A's failure a dissolution trigger, disqualified from being dissolved — it is real, checkable, and must be derived, not defined away.

Causal Order (target-blindness) — PASS. No step in the reduction chain — the finite-cutoff reflection-positivity theorems (Osterwalder–Seiler 1978, Lüscher 1977), the block-spin/Kotecký–Preiss polymer apparatus, or Lemma 1's Wilsonian-universality argument — was constructed with foreknowledge of, or tuned toward, the Clay answer. The four constants entering the floored master inequality (\(E_{\rm conn}\), \(A_{\rm fluc}\), \(s_{\rm block}\), \(N_{\rm cert}\); values below) were computed from ordinary block/lattice data, not back-solved to produce a PASS or FAIL verdict. This screen is what underwrites the refusal, documented in Section I.6, to pick a favorable reading of the convention-dependent quantity \(z_\star\).

Nonseparability — PASS-with-flag. The three open co-gates R1 (uniform-gap bridge), R2 (BRST/Gribov positivity survival), and R3 (OS-reconstruction/continuum-measure survival) are flagged explicitly as three distinct, separately load-bearing open objects, not one problem wearing three names. The flag matters operationally: even a full proof of R1 (the marginal Kotecký–Preiss inequality) would not by itself yield the Clay gap, because R2 and R3 are independent obstructions that must also be resolved — proving one does not retire the others, and the register must not be scored as if it would. All three are exported to the shared wall registry (W2 for the continuum-limit problem broadly, W18 for the uniform-gap bridge specifically) precisely so that cross-gate bookkeeping does not double-count or under-count a shared unresolved wall.

The completion-run reviewer's overall verdict on this four-screen pass, plus the companion structural checks, is recorded as CERTIFY: FROM-NOTHING PASS (no dimensionful quantity was asserted without an anchor — \(\Lambda_{\rm YM}\) traces to the measured \(\alpha_3(M_Z)\)/\(M_Z\) chain, \(\Delta>0\) is explicitly flagged as a measured anchor never claimed as an output), COMPLETE-ROOT PASS (the full 13-dimensional, three-layer object was used throughout, not a truncated 4D-only or metric-only slice), THREE-SINS PASS (none of anchor-elimination, target-anchoring, or false-flooring is present), false-closure/false-openness PASS in both directions (the gate neither claims more than it has shown nor hides a residual it has actually closed), compute-verification PASS, and NO-BARE-#5 PASS (the terminal is correctly not typed as an unpinnable "#5" — see Section I.6).

I.6 What the complete three-root run leaves behind

Running Shape, Scale, and Granularity each to full completion, and passing the result through all four Layer-2 screens, converges the entire Gap-02 question onto exactly one remaining finite object — no more, because Lemma 1 (Shape) proves there is no hidden geometric lever left to find, and no fewer, because Granularity's firewall (G1) forbids declaring the Face-B observable dissolved. That one object is the uniform-gap bridge, equivalently the marginal Kotecký–Preiss convergence inequality:

At every finite lattice spacing and volume \((a,L)\), all of reflection positivity, the self-adjoint positive Hamiltonian, the simple Perron–Frobenius vacuum, and a positive finite-volume gap \(\Delta(a,L)>0\) are established theorems (Osterwalder–Seiler 1978, Lüscher 1977, Münster 1981) — Scale and Shape together have reduced the entire remaining difficulty to a single joint limit, not to any of the individual finite-volume ingredients. The obstruction is a genuine marginality phenomenon specific to \(d=4\): in the deep ultraviolet (\(g\to0\)) the large-field weight \(\mu_n\le e^{-c/g^2}\to0\) makes the inequality trivial, but inside the marginal band \(g(2^na)=O(1)\) the inequality is a strict finite comparison between two \(O(1)\) constants with no known small control parameter — a weight bound is not a contribution bound, and that category distinction is the wall itself, confirmed to be method-invariant (both the polymer/cluster-expansion route and the asymptotic-freedom monotone-RG route stall on the identical \(d=4\) marginal object).

The floored certificate carries four honestly classified \(O(1)\) constants, none of which the open question is about (recomputing them to more digits does not move the needle on whether \(\rho_\star<1\)):

An executed diagnostic — the master-inequality quantity \(D = s_{\rm block}-\log(E_{\rm conn}A_{\rm fluc}N_{\rm cert})\) on a \(\beta=6.0\), \(4^4\) ensemble — returns \(D<0\) at every binning tested (\(D=-4.004,-5.043,-6.162\pm0.041\), robust to \(\sim66\sigma\)), which is honestly reported as INCONCLUSIVE, not a refutation: this diagnostic tests only a sufficient worst-case proxy (\(z_\star\le N_{\rm cert}\max_\gamma w(\gamma)\)) that over-counts, and a genuine refutation would require a direct lower bound \(z_\star\ge1/(E_{\rm conn}A_{\rm fluc})\), which has not been computed. The convention-dependence of \(z_\star\) itself is flagged rather than resolved by picking a favorable reading (per the Causal-Order screen above): reading (A) diverges and FAILs, reading (B) and the single-effective-activity reading both PASS, and until \(w(\gamma)\) is derived target-blind none of the three readings is decidable as the answer.

The one genuinely framework-internal open lever, R4, is the bordism-anomaly question of whether the frozen \(K_6\) spin-\(\mathbb{C}\)/twist data \(\tau_{K_6}=(2,2)\) induces a nontrivial mixed 't Hooft anomaly with the \(\mathbb{Z}_3^{[1]}\) one-form center of pure \(SU(3)\), formally \(\xi_{R4}\in\Omega_5^{\mathrm{Spin\text{-}c}}(B(SU(3)\to PSU(3));\tau_{K_6})\), carried by the degree-3 class \(\bar x_1\in H^3(BPU(3);\mathbb{Z}/3)\) under the operative 3-primary differential \(d_5=Q_1=\beta P^1\). The center-level computation shows \(Q_1(u_2)=Q_1(2y_1+2y_2)=0\), so \(u_2\) survives as a \(d_5\)-cycle — even here, only the carrier's survival is certified; the twisted evaluation that would fix \(\xi_{R4}\)'s actual value is the honestly OPEN step. Even a nonzero \(\xi_{R4}\) would be necessary-not-sufficient for a gap (anomaly matching admits gapless saturation), so R4 is correctly classified as low leverage for Gap-02 itself, high cross-gate relevance, and — this is the point relevant to Shape's completeness — its resolution in either direction (including \(\xi_{R4}=0\)) leaves Lemma 1 exactly where it stands: the geometry supplies origin, not proof.

Deep-root verdict for Gap-02. Shape, run to completion across all three layers, proves (Lemma 1, a real theorem) that the frozen 13D geometry offers no shortcut past ordinary constructive \(SU(3)\) Yang–Mills, while fixing which \(SU(3)\) and exposing exactly one honestly open topological row (R4) with zero manufactured levers elsewhere. Scale, run to completion, fixes \(\Lambda_{\rm YM}\) from the measured gauge-coupling anchors and is orthogonal to the gap's existence. Granularity, run to completion under its own G1 firewall, dissolves the ontic continuum-existence question (Face A) via a two-route-confirmed finite-alphabet argument while being categorically forbidden from touching the measured, in-principle-recordable gap observable (Face B) — and the basin-shallowing countermodel confirms this dissolution does not covertly discharge the underlying axiom. The four Layer-2 screens certify that none of this was assembled target-first, that the finite object left behind is real (Record-Interface PASS) rather than dissolved, and that the three remaining open co-gates (R1/R2/R3) are honestly kept separate rather than collapsed into a single artificially smaller residual. What is left, after the complete three-root treatment, is exactly the standing Clay inequality \(\rho_\star<1\) (equivalently the uniform-gap bridge) — a single, localized, target-blind, finite comparison of \(O(1)\) constants, with the deep-UV limit trivial and the \(d=4\) marginal band carrying the entire remaining content. This is why the gate's terminal is CERTIFIED-IRREDUCIBLE / RESOLVED +0: the three roots have been exhausted, not abandoned, and what remains is a recorded external wall — the Clay Millennium problem itself — not an owed computation internal to this framework.

Construction II - the full derivation

II.1 The frozen object, pinned at all three layers, carried strictly as boundary data

The starting point is not "Yang–Mills in the abstract" but the specific pure-glue sector cut out of the frozen 13-dimensional active branch $$ \mathfrak{B}_{\rm active}=\underbrace{\big[\mathcal{M}_4\times K_6\times S^2\times S^1_Y/\mathbb{Z}_2\big]_\times}_{\times\ \text{Stage}}\ \oplus\ \underbrace{\big[\mathcal{F}^+_{\rm finite}\oplus\mathcal{C}_{\rm admiss}\big]_\oplus}_{\oplus\ \text{Rulebook}}\ \otimes\ \underbrace{\big[\mathcal{E}_{\rm matter}\oplus\mathcal{E}_{\rm gauge}\oplus\mathcal{E}_{\rm Higgs}\oplus\mathcal{E}_{\rm proton}\big]_\otimes}_{\otimes\ \text{Actors}}, $$ with total dimension \(D=4+6+2+1=13\) carried entirely in the \(\times\)-Stage factor (the \(\oplus\)-Rulebook and \(\otimes\)-Actors layers are non-metric, zero-dimensional, but are frozen parts of the branch and are never silently dropped). The gate's object is $$ \mathcal{Y}_{\rm Gap02}=\mathrm{PureGlue}_{SU(3)_c}\Big(\big[\mathcal{M}_4\times K_6\times S^2\times S_Y^1/\mathbb{Z}_2\big]_\times\ \oplus\ \big[\mathcal{F}^+_{\rm finite}\oplus\mathcal{C}_{\rm admiss}\big]_\oplus\ \otimes\ \big[E_{\rm matter}\oplus E_{\rm gauge}\oplus E_{\rm Higgs}\oplus E_{\rm proton}\big]_\otimes\Big), $$ i.e. the matter-removed branch: all fermion and Higgs content is stripped, leaving pure \(SU(3)_c\) gauge theory on \(\mathbb{R}^{3,1}\), with the 13D package retained only as boundary data that fixed which gauge group and at what scale, not as ongoing dynamical content. This is a deliberate, stated restriction (the "Clay branch"), not a relaxation to some generic or decorated SU(3) theory — the object under study is exactly the object the Clay Institute states its problem about.

× Stage — where \(SU(3)_c\) comes from. \(K_6=SU(3)/T^2\), the full \(A_2\) flag manifold, dimension 6, carries the left-isometry algebra \(\mathfrak{su}(3)\); this is the geometric origin of the color gauge group (binding rule from the geometry pack: gauge forces are isometries of the internal metric factors, and \(K_6\) supplies \(SU(3)_c\) while \(S^2\) separately supplies \(SU(2)_L\) and \(S^1_Y\) supplies \(U(1)_Y\) — \(K_6\) carries no \(SU(2)\) content). Root data of \(A_2=\mathfrak{su}(3)\): Cartan basis \((h_1,h_2,h_3)\) with \(h_1+h_2+h_3=0\); simple roots \(\alpha_1=(1,-1,0)\), \(\alpha_2=(0,1,-1)\), \(\alpha_1+\alpha_2=(1,0,-1)\); positive roots \(\{\alpha_1,\alpha_2,\alpha_1+\alpha_2\}\); Weyl half-sum \(\rho=(1,0,-1)\), \(\|\rho\|^2=2\) in Killing normalization; Weyl group \(S_3\), order 6. The tangent space decomposes \(T(K_6)=\mathfrak m_1\oplus\mathfrak m_2\oplus\mathfrak m_3\), each \(\dim_{\mathbb R}\mathfrak m_i=2\), carrying root \(\alpha_i\). At the symmetric chamber center \(\vec u=(1,1,1)\) all three Ricci eigenvalues coincide (Killing-norm) at \(\mathrm{Ric}_i=5/12\), scalar curvature \(\mathrm{Scal}=5/2\), so \(\mathrm{Scal}/\mathrm{Ric}_i=6=\dim K_6\) — a metric-scale-invariant identity holding equally in the \(R_6\)-physical normalization (\(\mathrm{Ric}_i=1/(2R_6^2)\), \(\mathrm{Scal}=3/R_6^2\)). Chern class of the tangent bundle \(c_1(TK_6)=2\rho=(2,2)\); mod 3 this is \((2,2)\ne(0,0)\), so \(\bar c_1\ne0\) — a fact used later in the R4 lever discussion but not load-bearing for the gap proof itself. Spin-\(\mathbb{C}\) index \(\chi(K_6,E)=-3\) fixes the family count (irrelevant here since matter is removed, but it is the same \(K_6\) whose isometry supplies the gauge group, so it is recorded as the same frozen object). Euler characteristic \(\chi(K_6)=+6=|\mathrm{Weyl}(SU(3))|\) — the two integers \(-3\) and \(+6\) must never be conflated.

⊕ Rulebook and ⊗ Actors — the UV completion, retained but shown INERT. The finite chamber \(\mathcal F^+_{\rm finite}\) (modulus \(\tau=\omega\), generation basis, sector projectors, phase and normalization rules) and the admissibility firewall \(\mathcal C_{\rm admiss}\) (selector v3, constraints C1–C14, freeze-before-compare barrier, anomaly conditions, no-mirror parity table, Wilson-line winding rule, FCNC/mediator no-go) are carried in full as part of the frozen branch. On the Actors side, \(\mathcal E_{\rm gauge}\) supplies the plain connection/curvature data $$ \mathcal E_{\rm gauge}=T^_(\mathcal M_4)\otimes\mathrm{ad}(P),\qquad A_c=A_\mu^a T_a\,dx^\mu\ (a=1,\dots,8),\qquad F_c=dA_c+A_c\wedge A_c, $$ which, once the group \(SU(3)\) and the scale \(\Lambda_{\rm YM}\) are fixed, is* the ordinary Yang–Mills configuration space \(\mathcal A/\mathcal G\) with no residual bundle coercivity beyond the standard one. The central theorem of this section (Lemma 1, §II.4 below) is that none of \(\mathcal F^+_{\rm finite}\), \(\mathcal C_{\rm admiss}\), \(\mathcal E_{\rm matter}\), \(\mathcal E_{\rm Higgs}\), \(\mathcal E_{\rm proton}\) contributes any further constraint to the IR continuum-limit question once this reduction is made — a proved fact, not an assumption of convenience.

Global structure carried through. Global center \(G_{\rm SM}=(SU(3)_c\times SU(2)_L\times U(1)_Y)/\mathbb{Z}_6\); in the pure-glue branch the \(\mathbb{Z}_6\) refinement collapses to a vacuous matter-charge-quantization statement (there is no matter to quantize), leaving the ordinary \(SU(3)\) center \(\mathbb{Z}_3\) acting on Wilson loops — exactly the center symmetry the Clay problem's target theory carries. Frozen branch identifiers are recorded internally as audit anchors certifying which object was tested; they certify provenance only and carry no physics content, so no hash is reproduced here.

II.2 SL-0: the variational identity — spectral gap \(\equiv\) no soft physical sequence

The first step is pure functional analysis, fully established, with no open content. Let \(\mathcal H_{\rm phys}\) be the physical Hilbert space with Hamiltonian \(H\ge0\), unique vacuum \(\Omega\), \(H\Omega=0\). Define $$ \Delta_{\rm var}:=\inf\Big{\langle\psi,H\psi\rangle:\psi\perp\Omega,\ \psi\in\mathcal H_{\rm phys},\ |\psi|=1\Big}. $$ Then, by the spectral theorem applied to the self-adjoint operator \(H\) restricted to \(\Omega^\perp\), $$ \Delta_{\rm var}=\Delta>0\quad\Longleftrightarrow\quad \operatorname{Spec}(H)\cap(0,\Delta)=\varnothing. $$ This equivalence is hand-checkable and carries no unproven content: it is simply the statement that the infimum of the Rayleigh quotient over the orthogonal complement of the ground state equals the bottom of the spectrum restricted to that complement. It fixes the target precisely — "no soft physical sequence" (no sequence of normalized, vacuum-orthogonal states with energy \(\to0\)) is logically identical to a spectral gap \(\Delta>0\) — and nothing more. SL-0 supplies no existence content whatsoever; it is bookkeeping that pins down exactly what must be shown.

II.3 SL-1 / M4D: the all-operator lemma — a rigorous conditional, not a gap proof

The second leg assembles the standard toolkit of constructive quantum field theory into a single conditional implication: $$ [H1\wedge H2\wedge H3\wedge H4]\ \Longrightarrow\ \text{no soft physical sequence}\ \Longrightarrow\ \Delta>0. $$ This reuses, without re-deriving, three established pillars: Osterwalder–Schrader reconstruction (Euclidean reflection-positive measure \(\to\) Wightman/Hilbert-space QFT), the spectral theorem, and the Källén–Lehmann spectral representation of two-point functions. The logical content of SL-1 is exactly this: if a continuum measure with the stated properties exists, then the gap follows. It is a real theorem — the implication is rigorously derivable and is treated as such throughout this dossier — but it derives no value and asserts no existence. Restating \([H1\wedge H2\wedge H3\wedge H4]\Rightarrow\) gap as "the gap is proved" would be a direct overclaim; the four hypotheses are precisely the surviving open content, addressed next.

The four hypotheses are not four independent axioms. They are one existence claim about a single limit object \(O:=\lim_{a\to0,L\to\infty}\mu_{a,L}\) (the putative continuum Euclidean measure), plus three predicates of that same object:

• H1 (= MO-1). The continuum-limit object \(O\) exists — i.e. the sequence of finite-cutoff measures \(\mu_{a,L}\) has a limit as the lattice spacing \(a\to0\) and volume \(L\to\infty\).
• H2 (= MO-2). \(O\) satisfies Osterwalder–Schrader reflection positivity (OS-2) in the limit — this presupposes H1 (there is nothing to check positivity of until the limit object exists).
• H3 (= MO-4). \(O\) is nontrivial (interacting, not a free/Gaussian theory) with a unique vacuum — again presupposing H1.
• H4 (= MO-5/6/7). \(O\) exhibits uniform exponential clustering of correlators through the marginal coupling band — the load-bearing teeth of the whole construction, and again a predicate of the H1 limit object.

Because H2, H3, H4 are all predicates of the same object whose existence is H1, the four-hypothesis package is really one compound question about the survival of a single continuum limit, not four separately hard problems to solve independently.

II.4 SL-2 and the Round-3 sharpening: everything is trivial at finite cutoff — the entire wall is the continuum limit

At every finite lattice spacing \(a>0\) and finite volume \(L<\infty\), all four properties are established theorems, not conjectures. This is the finite-cutoff foundation, fully due to prior art and never re-claimed here: Osterwalder–Seiler (1978) proves reflection positivity and the transfer-matrix construction on the lattice; Lüscher (1977) and Seiler's lecture notes (Lecture Notes in Physics 159) establish the transfer matrix and positivity; Münster (1981) proves a strong-coupling mass gap. Concretely, at finite \((a,L)\): $$ H_{a,L}\ge0,\qquad \text{OS reflection positivity holds},\qquad \text{a simple Perron–Frobenius vacuum exists},\qquad \Delta(a,L)>0, $$ together with existence of the finite-volume lattice measure itself (properties F1–F4 in the corpus's finite-cutoff ledger). Every one of H1–H4's finite-cutoff analogues is TRIVIAL to establish by these classical methods.

This is the Round-3 sharpening, and it is the single most important logical move in the whole reduction: the wall is not "do reflection positivity, uniqueness, clustering, and existence hold" — they provably do, at every finite cutoff, as theorems. The wall is entirely "do all four properties survive one joint uniform limit" as \(a\to0\), \(L\to\infty\), \(\beta\to\infty\) (equivalently the bare coupling \(g\to0\) in lattice units while the physical scale \(\Lambda_{\rm YM}\) is held fixed). This collapses the four-hypothesis package W1–W4 into one single open object, denoted \(W^\ast\). This collapse is a genuine sharpening — it reduces the named count of open objects — but it is explicitly not a closure: nothing has been proved to survive the limit. (STATUS-UPGRADES:0 — this sharpening does not change the gate's grade or reduce any axiom count; it only reduces the number of named open items from four coupled hypotheses to one.)

II.5 Localizing \(W^\ast\): the uniform-gap bridge and its equivalent forms

The single open object \(W^\ast\) is stated precisely as the uniform-gap bridge: $$ \boxed{\ \frac{\Delta(a,L)}{\Lambda_{\rm YM}}\ \ge\ c>0\quad\text{uniformly as}\ a\to0,\ L\to\infty,\ \beta\to\infty\ } $$ for some constant \(c\) that is symbolic — no numerical value for \(c\) is asserted anywhere in this dossier. This is logically equivalent to H4, uniform exponential clustering of the renormalized glueball two-point correlator through the marginal coupling band, and is equivalent in turn to the marginal Kotecký–Preiss (KP) convergence inequality of the cluster/polymer expansion approach: $$ \rho_\star\:=\ E_{\rm conn}\cdot A_{\rm fluc}\cdot e^{-s_{\rm marg}}\ <\ 1\qquad(\text{equivalently }\delta>0\text{ for some symbolic activity-suppression margin}). $$ This object is registered as R1, identified with the shared wall W18 and coinciding with the Clay object itself; it is the shared responsibility of the whole constructive-QFT continuum-limit problem, registered jointly with wall W2 and W4.

The precise obstruction — rarity is not the same thing as domination. In four spacetime dimensions the bare coupling runs marginally: at the deep-UV end (\(g\to0\)), the large-field (rare, badly-behaved configuration) weight is exponentially suppressed, \(\mu_n\le e^{-c/g^2}\to0\), and any Kotecký–Preiss-type convergence criterion is satisfied automatically there — this end of the argument is easy and is not the wall. The obstruction lives entirely inside the marginal band, at couplings \(g(2^na)=O(1)\) set by asymptotic freedom's own running: there, the KP inequality becomes a strict finite comparison between two \(O(1)\) constants, with no small parameter available to any known expansion method to force \(\rho_\star<1\). The precise category error that constitutes the wall: a bound on the weight \(\mu_n\) of a rare large-field configuration is not the same thing as a bound on its contribution to the sum defining \(\rho_\star\) — a rare-but-not-small-enough weight can still dominate the marginal-band sum. This obstruction is method-invariant: both the polymer/cluster expansion route and the asymptotic-freedom/reflection-positivity (AF-RP) monotone route stall on the same \(d=4\) marginal object; it is intrinsic to the marginality of the coupling in \(d=4\), not an artifact of one particular technique.

The quantifier order is decisive and must be stated exactly: $$ \exists\,(\delta,K,\kappa,a_0,L_0,n_0)\ \ \forall\,(a,L,\beta,n), $$ i.e. a single \(\delta>0\) must work uniformly over all lattice regulators and all scales simultaneously. This is precisely the statement no known method currently proves in \(d=4\) at marginal coupling — and precisely the statement the Clay problem itself asks for.

II.6 The four measured/derived \(O(1)\) constants entering the floored inequality

The corpus computed four concrete constants that populate the floored version of the master inequality. None of them is the open question — the open question is whether the inequality they assemble into holds uniformly through the band, not their individual values, and none was tuned to force an answer.

• \(E_{\rm conn}=19.0280=e\cdot7=e\cdot(2d-1)\) at \(d=4\). Status: DERIVED, a rigorous bound — a Penrose/Klarner-type lattice-animal connective-constant upper bound, \(\beta\)-independent, carrying no error bar because it is a combinatorial bound rather than a measurement. (A non-rigorous series/Monte-Carlo estimate of the true connective constant, \(\lambda_4\approx13\)–\(14\), is recorded as EMPIRICAL/PARTIAL only, and is never substituted for the rigorous bound in any inequality.)
• \(A_{\rm fluc}=0.05264\pm0.00014\). Status: MEASURED (Monte Carlo) — a single-block \(SU(3)\) Haar/Gaussian determinant ratio, computed at \(\beta=6.0\) from 20,000 draws.
• \(s_{\rm block}=1.0413\pm0.0412\). Status: MEASURED (Monte Carlo) — the per-cell activation cost; the analytic floor \(1.0000\) agrees with the measured value within error, consistent with the local bound \(\mathrm{Re}\,\mathrm{Tr}(1-U_p)\ge c\|1-U_p\|^2\).
• \(N_{\rm cert}=155,\ 438,\ 1342\) at binning thresholds \(\delta_{\rm tr}=0.5,\ 0.25,\ 0.125\) respectively. Status: MEASURED, binning-dependent — the count of distinct realized "cell letters" in the certificate alphabet; no binning-free count currently exists, and this dependence is recorded honestly rather than resolved by picking a convenient binning.

None of these four constants is recomputed or adjusted here to "improve" the picture; they are reproduced exactly as established, because the open question is not their value but the uniform survival of the inequality they populate.

II.7 The floored master inequality and the executed diagnostic — an honest INCONCLUSIVE, not a verdict

The floored form of the target inequality is $$ z_\star:=\sum_{\gamma\ne0}w(\gamma)\ <\ \frac{1}{E_{\rm conn}\cdot A_{\rm fluc}},\qquad\text{equivalently the MASTER inequality}\qquad s_{\rm block}\ >\ \log\big(E_{\rm conn}\cdot A_{\rm fluc}\cdot N_{\rm cert}\big), $$ required to hold uniformly through the marginal band, derived from ordinary 4D pure-glue block data. This inequality is OPEN; at marginal coupling both sides are genuinely \(O(1)\), so it is a strict finite-margin comparison that could, in principle, come out either way — including \(\rho_\star\ge1\), which would refute the KP route entirely (a real falsifier, not a rhetorical one).

Convention-dependence of \(z_\star\) — the target-fitting trap, named and avoided. The value assigned to \(z_\star\) depends on a reading convention that is not yet fixed by any derivation, and three readings give three different verdicts: - Reading (A), per distinct binned letter: \(z_\star=7.66\ldots\), which diverges as \(\delta_{\rm tr}\to0\) — this reading gives FAIL and is additionally ill-posed (it does not converge as the binning is refined). - Reading (B), intensive KP polymer activity per reference site: \(z_\star=0.109\) — PASS. - Single effective activity \(e^{-s_{\rm block}}\): \(z_\star=0.353\) — PASS.

Picking reading (B) or the single-activity reading to declare a PASS, or picking reading (A) to declare a FAIL, would both be reverse-engineering a convention to fit a desired answer — precisely the target-fitting this dossier is bound to avoid. Until \(w(\gamma)\) is derived in a target-blind way, \(z_\star\) is not decidable, and no reading is asserted as the value. A further internal inconsistency is recorded rather than papered over: \(E_{\rm conn}=e\cdot7\) is pure \(\mathbb Z^4\) graph combinatorics carrying no per-block letter count, so the boxed inequality as currently written does not itself specify whether the letter multiplicity belongs inside \(z_\star\) or should be dropped — an open bookkeeping question, not resolved here.

The executed Monte Carlo diagnostic — a banked loss, not hidden. The corpus computed the diagnostic quantity $$ D\ =\ s_{\rm block}-\log\big(E_{\rm conn}\cdot A_{\rm fluc}\cdot N_{\rm cert}\big) $$ on a \(\beta=6.0\), \(4^4\) (\(L=4\)) lattice ensemble, 16 plaquette-gated configurations, at each of the three binning thresholds: $$ D=-4.004\pm0.041\ (\delta_{\rm tr}=0.5),\qquad D=-5.043\pm0.041\ (\delta_{\rm tr}=0.25),\qquad D=-6.162\pm0.041\ (\delta_{\rm tr}=0.125), $$ each robust to roughly \(66\sigma\) — i.e. the measured value of \(D\) is far below zero with overwhelming statistical significance as a measurement of this particular proxy quantity. The critical value of \(N_{\rm cert}\) at which \(D=0\) works out to \(2.83\) letters, compared with the realized \(N_{\rm eff}=155\) at the loosest binning — a factor of roughly 55 away from the crossover.

This result is reported at exactly the epistemic status it earns and no further: \(D<0\) tests only a sufficient KP proxy, namely the worst-case upper bound \(z_\star\le N_{\rm cert}\cdot\max_\gamma w(\gamma)\). A worst-case upper bound coming out negative (i.e. the proxy inequality failing) does not refute the underlying target inequality, because the proxy over-counts — it uses the maximum per-letter weight times the count of letters, which is generically far larger than the true sum \(\sum_\gamma w(\gamma)\). A genuine refutation of \(\rho_\star<1\) would require a direct lower bound \(z_\star\ge1/(E_{\rm conn}A_{\rm fluc})\) on the true sum, which was not computed here and is not claimed. The correct verdict, stated once and not revisited: \(D\le0\) is INCONCLUSIVE, not a refutation. Under the corpus's own escalation rule, a negative \(D\) at the tested proxy triggers a bounded block-spin (Bałaban-RG) fallback re-evaluation at a coarser reference scale \(\ell_\star\) — it is a "master \(\Rightarrow\) target" sufficient, not necessary, implication, so failing the sufficient master proxy does not decide the target. Separately, the reading-B and single-activity readings of \(z_\star\) (§ above) PASS at the one tested anchor coupling, but this is not shown to be uniform through the band — a single-point PASS is not a proof of the required \(\forall(a,L,\beta,n)\) statement.

The plaquette precondition — passed, with a source discrepancy flagged rather than silently fixed. The binding precondition for any of the above readings to be meaningful is agreement of the measured average plaquette \(\langle P\rangle\) with the standard \(\beta=6.0\) continuum-extrapolation value \(0.5937\). Two runs in the corpus record slightly different values: the 2026-07-04 handoff records \(\langle P\rangle=0.59375\pm0.00197\) (agreement with \(0.5937\) at \(<1\sigma\)), while the 2026-07-02 completion-run builder records \(\langle P\rangle=0.59639\pm0.00219\) (agreement at roughly \(1\)–\(1.3\sigma\)). Both values pass the gate; the discrepancy at the third decimal between the two runs is recorded honestly as a source discrepancy, not resolved by inventing a third, reconciling number.

Two coincidences explicitly retired. The ratio \(\kappa^3/\pi\) and the numerological identity \(5+3=8\) were both flagged during the corpus's no-target-fitting review as seductive \(O(1)\) near-coincidences and were explicitly retired — carried in the record only as a discipline reminder that an \(O(1)\) ratio landing near a "nice" number is not, by itself, a verdict. Neither is banked as a result anywhere in this derivation.

II.8 Lemma 1: the geometry supplies no shortcut — a proved negative theorem

The second pillar of the gate's certified content, independent of R1, is the no-lever theorem. Its statement: once the gauge group \(SU(3)_c\) and the dynamically generated scale \(\Lambda_{\rm YM}\) are fixed by the frozen geometry (§II.1, §II.9 below), the specific 13-dimensional UV completion — the \(\oplus\)-Rulebook finite chamber, the \(\otimes\)-Actors matter/Higgs/proton bundles, the full compactification geometry beyond the bare fact of which gauge group it generates — contributes nothing further to the IR continuum-limit question that defines the mass gap.

Mechanism. This is not asserted by fiat; it is derived via Wilsonian universality, a genuine, established theorem of quantum field theory: the long-distance (IR) physics of any UV-complete theory lying in the same universality class as ordinary \(SU(3)\) Yang–Mills is insensitive to the details of the UV completion. Once one fixes that the IR gauge symmetry is \(SU(3)_c\) and that the theory flows to the same fixed point/scale structure as ordinary YM, the specific microscopic completion (in this case, a compactification of a 13-dimensional geometry) cannot alter the IR question of whether a mass gap exists — that question is a property of the universality class, not of the particular UV realization of it.

Verification by exhaustive audit. The corpus's three-layer completion audit classifies every \(\oplus\)-Rulebook/\(\otimes\)-Actors primitive object at the proof-domain level into one of five classes, with the following exhaustive tally:

Exactly zero primitive-groups are classified as an open candidate lever for the gap itself; exactly one row (the boundary/topological class) is routed onward to a separate, genuinely open question (R4) that is explicitly not a candidate mechanism for the gap (see §II.10 — R4 is at best necessary, never sufficient, and the honest prior tilts against it mattering at all). Concretely, on the Actors side, \(\mathcal E_{\rm gauge}\)'s connection and curvature data $$ T^_(\mathcal M_4)\otimes\mathrm{ad}(P),\qquad A_c=A_\mu^a T_a\,dx^\mu\ (a=1,\ldots,8),\qquad F_c=dA_c+A_c\wedge A_c $$ is literally the ordinary Yang–Mills configuration space \(\mathcal A/\mathcal G\) — there is no exotic bundle coercivity hiding in it. Lemma 1's two component results (labeled 1A/1B in the corpus, the local-action and admissibility-inertness pieces) held* under this audit. The resulting three-layer, proof-domain-level classification of the whole geometric package is MIXED_WITH_OPEN, with the single R4 boundary row left open — but this open row is not an owed computation of the gate itself; it is a distinct question with its own separate closure criterion (§II.10), and its resolution either way leaves Lemma 1 intact (see the three-outcome tree there).

What Lemma 1 establishes, precisely. The required proof strategy for the Yang–Mills mass gap, for this frozen 13-dimensional geometric arena, is exactly the standard Clay constructive-QFT program — no more, no less. The geometry fixes which gauge group appears (\(SU(3)\), via \(K_6=SU(3)/T^2\)'s isometry), fixes at what scale it becomes strongly coupled (\(\Lambda_{\rm YM}\), via the anchored RG running, §II.9), and thereby collapses an a priori much larger search space (which group? which scale? which matter content?) down to a single well-posed instance of the Clay problem — but supplies no additional analytic technique, no coercivity estimate, no small parameter, and no shortcut for solving that instance. This is a real, banked, terminal-negative theorem: it forecloses an entire class of hoped-for geometric shortcuts, and it does so with a rigorous mechanism (Wilsonian universality) plus an exhaustive audit, not by assumption.

II.9 SCALE: \(\Lambda_{\rm YM}\) is given, not derived — a scale-fixing, not a derivation of the ratio

The scale \(\Lambda_{\rm YM}\) that appears in the target ratio \(\Delta/\Lambda_{\rm YM}\ge c>0\) is pinned by the frozen anchors, not derived within this gate, and its numerical determination is explicitly out of scope for the gap question (deriving \(\Lambda_{\rm YM}\) is equivalent to deriving the anchor \(\alpha_3(M_Z)\) itself, which the corpus's four-anchor structure takes as an irreducible input). The supporting frozen scale data, carried through from the geometry pack in full precision:

$$ R_0=(2\pi M_U)^{-1}=1.591549430918954\times10^{-17}\ \mathrm{GeV}^{-1},\qquad M_U\sim1.0\times10^{16}\ \mathrm{GeV}, $$ one-loop Standard-Model beta coefficients (GUT-normalized, \(\alpha_1=\tfrac53\alpha_Y\)) $$ (b_1,b_2,b_3)=\Big(\frac{41}{10},\,-\frac{19}{6},\,-7\Big)=(4.100000000000000,\,-3.166666666666667,\,-7), $$ and the total Kaluza–Klein threshold correction, computed to two-loop \(\overline{\rm MS}\) at \(M_Z=91.1876\) GeV, $$ (\delta_1,\delta_2,\delta_3)=(+4.8424,\ -3.1112,\ -1.7313)\ \pm\ 1.6\times10^{-3}, $$ with the color threshold \(\delta_3=-1.7313\) the one directly relevant to \(\Lambda_{\rm YM}\)'s running into the strong-coupling regime, and a unification residual \(\big|\alpha_i^{-1}(M_U)-\alpha_j^{-1}(M_U)\big|=9.6\times10^{-11}\) (a numerical-pipeline floor, well inside the propagated PDG uncertainty band of order \(10^{-3}\)). \(M_{\rm Pl}\) plays no direct role in the ratio \(\Delta/\Lambda_{\rm YM}\) that R1 must bound — the Planck scale and the Yang–Mills scale enter the framework through separate anchors and separate machinery (Planck normalization via \(M_*^{11}\,\mathrm{Vol}(X_{\rm active})=M_{\rm Pl}^2\) is a completely independent relation, not used in this gate). The content of this subsection is a scale-fixing: the frozen geometry, via the ordinary two-loop renormalization group plus the Kaluza–Klein threshold ledger, pins the numerical value at which the \(SU(3)\) coupling becomes strong — it does not, and is not claimed to, derive the ratio \(\Delta/\Lambda_{\rm YM}\) that is the actual content of R1.

II.10 R4: the one framework-internal open lever, corrected object and current status

Separately from the Clay reduction (R1) and independently of the OS-reconstruction/positivity co-gates (R2, R3, §II.11), the corpus identifies exactly one genuinely framework-internal open question, routed here by the class-4 (TOPOLOGICAL/BOUNDARY) row of the Lemma-1 audit: does the frozen \(K_6\) spin-\(\mathbb{C}\)/chamber data induce a nontrivial, universality-preserving mixed 't Hooft anomaly between the geometry and the \(\mathbb Z_3^{[1]}\) one-form center symmetry of pure \(SU(3)\) (in the sense of Gaiotto–Kapustin–Seiberg–Willett's generalized-symmetry framework, with background field \(B^{(2)}\in H^2(X,\mathbb Z_3)\))?

The corrected, authoritative object. An earlier package's framing of this question — a degree-4 cup product \(\bar c_1(L_{K_6})\times B^{(2)}\) on \(K_6\times\mathcal M_4\), with an operative differential \(d_3=\beta\mathrm{Sq}^2\) — was reviewed and found not well posed as stated (verdict R4_NOT_WELL_POSED_AS_STATED, fail-closed): the differential proposed is 2-primary and annihilates 3-torsion, so it cannot detect the genuinely 3-primary obstruction at stake, and the degree/locus were also wrong (a degree-4 class on the wrong space rather than a degree-5 class on the correct bordism group). This is explicitly corrected, not revived, and the corrected object is: $$ \xi_{R4}\ \in\ \Omega_5^{\mathrm{Spin\text{-}c}}\big(B(SU(3)\to PSU(3));\ \tau_{K_6}\big), $$ a finite abelian group — the twisted degree-5 spin-c bordism invariant. The carrier class is the degree-3 class \(\bar x_1\in H^3(BPU(3);\mathbb Z/3)\) (there is no degree-2 mod-3 carrier). The correct operative differential is the 3-primary Milnor primitive \(d_5=Q_1=\beta P^1\), of internal degree \(|Q_1|=5\) at the prime \(p=3\).

The certified algebraic content. Using the standard action of the Milnor primitive on the cohomology of \(BPU(3)\): \(Q_1(y_i)=0\), \(Q_1(x_i)=2y_i^3\), and consequently $$ Q_1(u_2)=Q_1(2y_1+2y_2)=0, $$ which shows \(u_2\) (the center-restriction class, carrying the frozen \((2,2)\) Chern data) is a \(d_5\)-cycle — it survives this differential. Since 2-primary differentials cannot touch 3-torsion classes (a model-independent fact), and the correct mod-3 differential \(d_5=Q_1\) annihilates \(u_2\), the class survives both primary differentials at the relevant degree, so the algebraic expectation is that \(\xi_{R4}\) is nonzero absent further obstruction — though this expectation is explicitly not the same as a computed value.

Supporting untwisted computations, fully resolved and cited rather than re-derived: \(\Omega_5^{\rm Spin}(PSU(3)\times B^2\mathbb Z_3)=\mathbb Z_3\ne0\) (the untwisted survivor persists); independently, \(\Omega_5^{\mathrm{Spin\text{-}c}}(B^2\mathbb Z_3)=0\), resting on \(H_3=H_5=0\), \(H_4=\mathbb Z_3\) of the Eilenberg–MacLane space \(K(\mathbb Z_3,2)\) together with the odd-degree vanishing of the spin-c point bordism coefficients (low-degree spin-c point bordism ring: \(\mathbb Z,\,0,\,\mathbb Z,\,0,\,\mathbb Z^2,\,0,\ldots\)). This untwisted zero is emphatically not the same statement as \(\xi_{R4}=0\): \(\xi_{R4}\) lives in the twisted bordism group, and the twist \(\tau_{K_6}\) — sourced by \(c_1(TK_6)=2\rho=(2,2)\not\equiv(0,0)\bmod3\), itself forced by the spin-c index \(\chi(K_6,E)=-3\) — is the only possible carrier of a nonzero value. The concrete remaining open step is to evaluate the twisted \(d_5=Q_1\) correction sourced by \(\tau_{K_6}=(2,2)\) and determine whether it supplies a nonzero degree-lowering image on the bordism spectral sequence. This computation has not been carried out; \(\xi_{R4}\) remains an explicitly unassigned element of a finite abelian group — never given a numerical or group-element value anywhere in this dossier.

Why R4, even if resolved nonzero, would not by itself close the gap (necessary, not sufficient — R5). Anomaly matching is a constraint on the IR phase, not a proof of confinement: even a nontrivial, universality-preserving \(\xi_{R4}\) only forces that some IR phase must saturate the anomaly, and anomaly matching famously admits gapless solutions (a gapless conformal fixed point or topological quantum field theory can equally saturate a matched anomaly). A separate, further argument — that the matched IR phase is specifically the gapped, confining one — would still be required. This caveat is derived, not speculative, and its own falsifier is stated plainly: if the matched IR phase for this anomaly is demonstrated to be a gapless one, that directly confirms R4 is not, after all, a viable gap-forcing mechanism, without at all undermining the anomaly computation itself.

Honest prior and outcome tree. Three outcomes are logically exhaustive and none may be asserted until \(\xi_{R4}\) is actually computed: (a) \(\xi_{R4}\) trivial \(\Rightarrow\) R4 passes cleanly \(\Rightarrow\) Lemma 1 is confirmed across the board, with the full 13D structure serving only as UV/boundary/motivational data; (b) \(\xi_{R4}\) nontrivial, and it descends to 4D, survives the continuum limit, imposes a genuine Faddeev-Popov/Gribov boundary condition, and is universality-preserving \(\Rightarrow\) a "Lemma-1-prime" geometry-sourced confinement ingredient — still explicitly not a Clay proof by itself, only a contributing constraint; (c) \(\xi_{R4}\) nontrivial but non-universal (a measure-zero subclass) or failing any of the chain's further conditions \(\Rightarrow\) not Clay-relevant, and Lemma 1 stands as before. The honest prior tilts away from outcome (b): the corpus records that the compact boundary/parity sector generically adds difficulty relative to flat-space Yang–Mills rather than supplying coercivity, so a nonzero, gap-relevant \(\xi_{R4}\) would be a genuinely surprising outcome rather than an expected one. The estimated tractability horizon for a resolution is on the order of 5–15 years, carried as an honest, bounded, testable bet rather than a claim.

A provenance flag to keep distinct. The R4 computation package separately carries a frozen \(K_6\) holonomy phase \(\Phi\approx-0.12827\) rad as boundary \(\theta\)-data — this is never a computed anomaly output, only an input datum, and it is not the same object as the chamber phases \(\tau=\omega=e^{2\pi i/3}\), the CKM holonomy phase \(\delta_{\rm CKM}=-2\pi/3\), or the lepton Berry phase \(+2\pi/3\) that appear elsewhere in the frozen \(F^+\) chamber; these are kept explicitly distinct here to avoid a conflation the corpus flags as a live risk.

II.11 R2 and R3: the two independent co-gates sharing wall W2

Even a hypothetically completed proof of R1 (the uniform-gap bridge) would not, by itself, complete the Clay theorem, because two further, logically independent legs remain open, both exported to the shared constructive-QFT wall registry W2:

• R3 — Osterwalder–Schrader reconstruction / continuum reflection-positivity survival. This is H1\(\wedge\)H2\(\wedge\)H3 together: the actual construction of a measure \(\mu=\lim\mu_{a,L}\) satisfying the full OS axioms (OS-0 through OS-3) in the continuum limit, together with nontriviality (an interacting, non-Gaussian theory) and a unique vacuum, followed by the OS\(\to\)Wightman reconstruction theorem. No 4D \(SU(3)\) continuum measure with these properties has ever been constructed by anyone, anywhere. Its falsifier is explicit: a proof that the limiting theory is trivial (Gaussian/free) would refute the interacting continuum theory outright.
• R2 — nonperturbative BRST/Gribov kernel positivity. The physical Hilbert space of a gauge theory is defined cohomologically, \(\mathcal H_{\rm phys}=\ker(s)/\mathrm{im}(s)\) for the BRST operator \(s\); positivity of the resulting inner product is established only perturbatively (Kugo–Ojima, via the quartet mechanism). Nonperturbative survival of this positivity as \(a\to0\) is a wholly separate open co-gate, obstructed by the Gribov–Singer–Neuberger phenomenon: there is no global continuous gauge-fixing section on a nontrivial bundle (Gribov 1978, Singer 1978), and the lattice BRST construction produces an ill-defined \(0/0\) (Neuberger 1987) from Faddeev–Popov determinant sign flips across Gribov copies. Its falsifier is equally explicit: a demonstration that positivity fails in the continuum limit would itself be a major (negative) constructive-QFT result.

Both R2 and R3 are independent of R1 and of each other — even a proved R1 yields no gap without an independently proved R2 and R3. All three (R1, R2, R3) trace back to the same underlying finite-cutoff foundation of §II.4 (SL-0/SL-1/SL-2), which is one shared reduction consumed by all three co-gates, but each of R1, R2, R3 is a separately load-bearing open object and none may be counted as resolved if either of the others closes — a nonseparability flag the corpus's own audit explicitly checks and passes ("PASS-with-flag").

II.12 GRANULARITY: Face A dissolves, Face B categorically does not — and why that is not a loophole

The corpus's granularity/cost-floor machinery makes one further, precisely bounded move, which must be stated with its scope exactly as narrow as it actually is.

Face A — continuum existence as an ontic infinite-precision limit. The Uniform Operational Cell Law posits a minimum positive operational action/resolution scale \(\Delta_0>0\), below which no further physical distinction can, even in principle, be recorded. Under this posit (together with a companion "record-interface" condition, jointly and only under both), the literal \(a\to0\) limit is record-impossible-in-principle: no observer or apparatus, however idealized, could ever certify having reached it. Fixing instead a finite physical resolution \(\ell_\star\sim\Lambda_{\rm YM}^{-1}\) makes the gauge-invariant certificate alphabet \(\mathcal C_\star\) (the space of distinguishable field configurations at that resolution) finite, converting the continuum-existence question from an infinite, possibly ill-posed limit into a finite, in-principle-checkable inequality. This half of the gap02 problem — continuum existence as an ontic limiting statement — is DISSOLVED-GIVEN-(Granularity \(\wedge\) Record-Interface), confirmed by two independent routes, and is counted, in the corpus's cost-floor wall-impact ledger, as a single dissolved physical wall (decomposing internally into 13 continuum sub-rows: MO-1 through MO-7, the uniform-gap bridge, Lemma 2, G02.A/B, and the Clay theorem statements (i) and (ii) — tracked but not separately re-litigated here).

Face B — the gap bound itself is an observable, and observables never dissolve. The ratio \(\Delta/\Lambda_{\rm YM}\) is not a statement about an idealized mathematical limit; it is tied directly to measured physical quantities — the confinement radius (of order 1 fm), the lightest glueball mass, the running of \(\alpha_s\). The corpus's own governing rule, labeled G1 (OBSERVABLE-NEVER-DISSOLVES), fires here and categorically forbids dissolving Face B by the same granularity argument. The firewall statement, carried identically for gap01 and for the black-hole-singularity gate, is: "Granularity dissolves the continuum, never the gap." This is not a special pleading invented for this gate; it is the same rule applied consistently across the corpus wherever an observable quantity is at stake.

Dissolved is not solved. Declining to ask the literal ontic continuum-limit question, in favor of a finite-resolution reformulation, changes which question is being asked about existence — it does not supply an answer to the Clay question, which is about the mathematical continuum theory, not about operational record-keeping. This distinction is maintained throughout: Face A's dissolution is a real, rigorous result about a different (finite-resolution) formulation; it has no bearing on, and does not touch, R1/R2/R3 as stated in the standard Clay formulation.

REDUCED-TO-AXIOM does not lower the axiom count. The floor \(\Delta_0>0\) is itself a named, unproven posit (with \(\hbar\) standing as its measured residue) — it is not itself derived from anything more primitive, and stating the granularity dissolution does not reduce the framework's total axiom count. A rigorous countermodel shows the uniform version of such a floor is not automatically discharged: take depths \(d_n=B\cdot2^{-n-1}\) for a constant \(B>0\); then \(\sum_n d_n=B/2<\infty\) (a finite total resource budget is consistent with this sequence) while \(\inf_n d_n=0\) (the sequence has no uniform positive floor). This "basin-shallowing" countermodel shows that finite total resources alone buy only an \(\eta\)-dependent total-boundedness statement, not a genuine uniform \(\Delta_0>0\) — so the posit is doing real, unproven work, and is disclosed as such rather than presented as derived.

II.13 Assembling the reduction: what is CLOSED, what is OPEN, and why the terminal is legitimate

Collecting the pieces derived above:

1. SL-0 (variational identity: gap \(\equiv\) no soft sequence) — established, no open content.
2. SL-1/M4D (the conditional \([H1{\wedge}H2{\wedge}H3{\wedge}H4]\Rightarrow\) gap) — a rigorous, banked implication; not itself a gap proof.
3. SL-2 + Round-3 sharpening — all four hypotheses trivial at finite cutoff; the entire wall is the joint continuum limit, collapsing W1–W4 to the single object \(W^\ast=\) R1.
4. R1 (uniform-gap bridge / marginal KP inequality \(\rho_\star<1\)) — OPEN, the Clay object itself, localized precisely, with the four supporting \(O(1)\) constants derived/measured and the diagnostic MC run executed and reported as INCONCLUSIVE.
5. Lemma 1 (no geometric shortcut) — DERIVED, a genuine terminal-negative theorem, verified by an exhaustive three-layer audit with zero open-candidate levers found.
6. \(\Lambda_{\rm YM}\) — GIVEN-E, a scale-fixing consumed as boundary data, not derived here.
7. R4 (bordism anomaly \(\xi_{R4}\)) — OPEN, the one genuinely framework-internal lever, corrected to its proper degree-5 twisted spin-c bordism home, algebraically expected nonzero (surviving both primary differentials) but not evaluated, and even if resolved nonzero only necessary-not-sufficient for the gap (R5 caveat).
8. R2, R3 (BRST/Gribov positivity; OS-reconstruction) — OPEN, independent co-gates sharing wall W2; even a proved R1 does not yield the gap without these.
9. Granularity — Face A (ontic continuum existence) DISSOLVED-GIVEN-AXIOM; Face B (the gap, an observable) explicitly and categorically NOT dissolved by G1; the axiom itself remains a disclosed, unproven posit (countermodel-unrefuted).
10. The measured floor — \(\Delta>0\) itself is a MEASURED-ANCHOR (glueball spectrum, confinement radius, running \(\alpha_s\)), never an output of this derivation.

The gate's own toolbox of structural roots is saturated without any root forcing a unique survivor or eliminating the open status — Shape exposes the origin of the gauge group but supplies no proof lever (EXPOSE); Scale pins \(\Lambda_{\rm YM}\) but does not touch the ratio (PASS); Granularity constrains Face A only, categorically excluded from Face B (CONSTRAIN, scope-limited); Invariance, Causal-Order pass cleanly (no target-fitting detected anywhere in the chain); Record-Interface exposes the finite-resolution reformulation (EXPOSE); Nonseparability exposes the R1/R2/R3 independence (EXPOSE). This is exactly the signature the corpus's own taxonomy assigns to a genuine, non-dissolvable external wall, as opposed to a gate that is merely under-worked: every tool that could in principle force a resolution has been applied, honestly, and none does.

This is why CERTIFIED-IRREDUCIBLE / RESOLVED +0 is the correct and legitimate terminal, and not a dressed-up open problem: the gate does not own the residual (R1, tied to R2/R3, is the named, externally-recorded Clay wall — a wall the entire mathematics and physics community has failed to cross since 2000, not a computation this framework left undone), the framework's own contribution (Lemma 1, the reduction chain, the localization to one finite inequality, the executed and honestly-scored diagnostic) is complete and rigorous on its own terms, and the numerical content is taken as a measured anchor rather than claimed as a derived result. The residual — the open Clay problem itself, together with the independent R2/R3 co-gates and the one live R4 lever — is carried forward explicitly in the register below, shown, not hidden, and not rolled up into any hedge on the gate's own headline.

Construction III - the central result at full precision

0. What this section proves, and what it does not

This section carries the full weight of gap02's certified-irreducible terminal. Everything else in the dossier — the community context, the geometric origin story, the anchoring of \(\Lambda_{\rm YM}\) — is scaffolding around a single chain of implications that must be shown, link by link, with no step skipped and no constant assigned a value it has not earned. The chain has exactly three moving parts: (I) an elementary, fully rigorous variational identity that converts "no soft physical state" into "spectral gap"; (II) a certificate-conditional operator theorem that reduces "gap exists" to four hypotheses about a single limiting object, and a Round-3 sharpening that collapses those four hypotheses' hard content into one; and (III) the identification of that one hard hypothesis with a marginal Kotecký–Preiss (KP) convergence inequality, stated as a strict comparison between four independently sourced \(O(1)\) constants, together with an honest, executed Monte Carlo diagnostic of a sufficient (not necessary) proxy for that inequality. None of these three parts derives \(\Delta>0\). All three are rigorous, banked results in their own right, each hand-checkable independently, and together they are the entire content of what gap02 "does."

Throughout, every object is pinned in the frozen 13D arena \(\mathfrak B_{\rm active}=\big[\mathcal M_4\times K_6\times S^2\times S^1_Y/\mathbb Z_2\big]_\times \oplus\big[\mathcal F^+_{\rm finite}\oplus\mathcal C_{\rm admiss}\big]_\oplus\otimes\big[\mathcal E_{\rm matter}\oplus\mathcal E_{\rm gauge}\oplus\mathcal E_{\rm Higgs}\oplus\mathcal E_{\rm proton}\big]_\otimes\), \(D=4+6+2+1=13\), restricted (pure-glue branch, matter removed) to its \(\times\)-Stage \(K_6=SU(3)/T^2\) origin of \(SU(3)_c\) and its \(\otimes\)-Actors gauge sector \(\mathcal E_{\rm gauge}=T^*\mathcal M_4\otimes\mathrm{ad}(P)\) with connection \(A_c=A_\mu^a T_a\,dx^\mu\) (\(a=1,\dots,8\)) and curvature \(F_c=dA_c+A_c\wedge A_c\). This is, exactly, the ordinary \(\mathcal A/\mathcal G\) configuration space of 4D pure-glue \(SU(3)\) Yang–Mills — no bundle-coercivity trick, no exotic boundary condition. That identification is itself part of what must be earned (Part II, Lemma 1) rather than assumed.

I. SL-0 — the variational identity (fully established, zero open content)

The starting point is pure spectral theory, not physics input, and it is worth writing out completely because every later step is phrased in its terms. Let \(H\) be a self-adjoint, non-negative operator on a Hilbert space \(\mathcal H_{\rm phys}\) with a distinguished unit vector \(\Omega\) (the vacuum), \(H\Omega=0\). Define $$ \Delta \;:=\; \inf\Big{\,\langle\psi,H\psi\rangle \;:\; \psi\in\mathcal H_{\rm phys},\ \psi\perp\Omega,\ |\psi|=1\,\Big}. $$ Claim (SL-0): $$ \Delta>0 \quad\Longleftrightarrow\quad \operatorname{Spec}(H)\cap(0,\Delta)=\varnothing. $$ Proof. (\(\Rightarrow\)) If \(\Delta>0\), then by definition of \(\Delta\) as an infimum of Rayleigh quotients over the orthogonal complement of \(\Omega\), the spectral projection of \(H\) onto \((0,\Delta)\) restricted to \(\Omega^\perp\) must vanish — otherwise a unit vector in the range of that projection would furnish a Rayleigh quotient strictly below \(\Delta\), contradicting the infimum. Since \(\{0\}\) is isolated (occupied only by \(\Omega\), given), \(\operatorname{Spec}(H)\cap(0,\Delta)=\varnothing\). (\(\Leftarrow\)) Conversely, if the open interval \((0,\Delta)\) contains no spectrum, then every unit vector orthogonal to \(\Omega\) has its spectral support in \(\{0\}\cup[\Delta,\infty)\); the \(\{0\}\)-component is excluded by orthogonality to \(\Omega\) (assuming, as bundled into the Clay statement, that \(\ker H=\mathbb C\,\Omega\) — vacuum uniqueness), forcing \(\langle\psi,H\psi\rangle\ge\Delta\) for every such \(\psi\), hence \(\Delta\) (the infimum) is attained or approached from above and is itself \(>0\) whenever the defining \(\Delta\) on the right is \(>0\). \(\blacksquare\)

This is "no normalized, vacuum-orthogonal, physical state has energy \(\to0\)" restated as "the spectrum has a gap above \(0\)" — the two phrasings of the mass-gap statement quoted in the Clay problem are the same statement, and SL-0 is the (elementary, hand-checkable) proof of that equivalence. It carries no open content: it is pure functional analysis, uses no dynamical input about Yang–Mills, and is not weakened, sharpened, or approximated anywhere downstream. Every subsequent step is about how to establish one side of this equivalence for the actual physical \(H\) — SL-0 itself never has to be revisited.

II. SL-1/M4D — the all-operator lemma, and its four hypotheses

The lemma (certificate-conditional, rigorously derived, asserts no value): $$ \big[H1\wedge H2\wedge H3\wedge H4\big] \;\Longrightarrow\; \text{no soft physical sequence} \;\Longrightarrow\; \Delta>0. $$ This reuses three pieces of standard machinery without re-deriving them: Osterwalder–Schrader reconstruction (Euclidean measure \(\to\) Wightman/Hilbert-space data), the spectral theorem (self-adjoint \(H\) has a well-defined spectral measure), and the Källén–Lehmann representation (two-point functions decompose over the physical spectrum, so exponential clustering of correlators is the same statement as a spectral gap in disguise). The logical content of SL-1 is exactly this: if the four hypotheses below hold of the continuum limit object, then the chain of standard theorems closes and a gap follows. It is an implication, proved rigorously, deriving no number. It is never restated in this dossier as "the gap is proved" — the entire remaining content of gap02 is in exhibiting precisely how far short of discharging \(H1\)–\(H4\) the corpus can get, and where it stops.

The four hypotheses are not four independent axioms. They are one existence claim about a single limiting object \(O\), plus three predicates of that same object, each logically presupposing the first:

• H1 (= MO-1). The continuum-limit object \(O\) — the Euclidean measure \(\mu:=\lim_{a\to0,\,L\to\infty}\mu_{a,L}\) — exists (as a well-defined probability measure on field configurations, in the appropriate topology).
• H2 (= MO-2). \(O\) satisfies Osterwalder–Schrader reflection positivity (OS-2) in the limit. Presupposes H1: reflection positivity is a statement about \(\mu\), which must first exist.
• H3 (= MO-4). \(O\) is nontrivial (interacting, not a free/Gaussian theory) with a unique vacuum. Presupposes H1.
• H4 (= MO-5/6/7). \(O\) exhibits uniform exponential clustering of correlators through the marginal coupling band as the cutoffs are removed. Presupposes H1. This is the load-bearing teeth of the whole lemma — it is the hypothesis that, via Källén–Lehmann, is equivalent to the actual mass gap statement, and it is the one hypothesis that is a strict quantitative inequality rather than a qualitative existence/positivity/uniqueness statement.

The Round-3 sharpening — anchoring onto the finite-cutoff foundation, not onto the target. Section SL-2 records what is already a theorem at every finite lattice spacing \(a>0\) and finite volume \(L<\infty\): reflection positivity (Osterwalder–Seiler 1978), the transfer-matrix construction giving self-adjoint \(H_{a,L}\ge0\) (Lüscher 1977; Seiler, Lecture Notes in Physics 159), a simple (non-degenerate) Perron–Frobenius vacuum, and — at strong bare coupling — a finite-volume gap \(\Delta(a,L)>0\) (Münster 1981). This means that at every finite cutoff, H1, H2, and H3 all hold as theorems, and even H4's finite-volume shadow (finite-volume clustering) is not in question. The sharpening this licenses is precisely this: the four hypotheses are not "do these structures exist" — they provably do, at finite cutoff, for all four properties simultaneously. The entire remaining question, for all four hypotheses at once, is whether they survive one single joint uniform limit $$ a\to0,\qquad L\to\infty,\qquad \beta\to\infty\ \ (\text{equivalently } g\to0,\text{ asymptotic freedom}), $$ taken together. This collapses \(H1\!\wedge\!H2\!\wedge\!H3\!\wedge\!H4\) — nominally four hypotheses — into one open object, denoted \(W^\ast\): the uniform survival, through this joint limit, of the finite-cutoff package that is already fully proved at every finite point of the limit. This is a sharpening (fewer independently-named open objects: from four down to one), not a closure — none of H1–H4 is discharged by the sharpening, and the sharpening itself asserts no numerical value and forces no promotion of the gate's grade.

III. The localized wall — the uniform-gap bridge and the marginal KP inequality

Identifying \(W^\ast\) with a single displayed inequality (this is R1, = W18, = the Clay object itself). The load-bearing content of \(W^\ast\) — via H4 and Källén–Lehmann — is exactly the statement that the finite-volume gap survives the joint limit at fixed ratio to the dynamically generated scale: $$ \boxed{\ \frac{\Delta(a,L)}{\Lambda_{\rm YM}}\ \ge\ c>0 \quad\text{uniformly as } a\to0,\ L\to\infty,\ \beta\to\infty\ } $$ for some constant \(c\) that is never assigned a numerical value anywhere in this reduction (it is exactly the symbolic \(c'\) of the Clay statement's item (d), \(\Delta\ge m\ge c'\Lambda_{\rm YM}>0\)). This uniform-gap bridge is equivalent to H4, uniform exponential clustering of the renormalized glueball correlator, which is in turn equivalent to a Marginal Kotecký–Preiss convergence inequality — the standard sufficient condition, in the cluster-expansion/polymer-model literature, for a statistical-mechanical partition function to have a convergent expansion (hence exponential clustering, hence a gap) rather than a phase transition or critical, gapless behavior: $$ \rho_\star \;:=\; E_{\rm conn}\cdot A_{\rm fluc}\cdot e^{-s_{\rm marg}} \;<\;1 \qquad (\delta>0 \text{ the associated margin}). $$ Equivalently, in the sum-over-polymers form, $$ z_\star \;:=\; \sum_{\gamma\ne0} w(\gamma) \;<\; \frac{1}{E_{\rm conn}\cdot A_{\rm fluc}}, \qquad\text{equiv. MASTER:}\quad s_{\rm block} \;>\; \log!\big(E_{\rm conn}\cdot A_{\rm fluc}\cdot N_{\rm cert}\big). $$ Each of these three displayed forms is a different bookkeeping of the same underlying inequality (Kotecký–Preiss convergence, per-effective-activity form, and a worst-case master comparison respectively); none is a different physical statement, and none is solved by writing it in a different form. This is the entire remaining content of the Yang–Mills mass gap, localized to one finite, sharply-stated, symbolic comparison.

III.1 Why the inequality is hard exactly at \(d=4\): rarity is not domination

The reason this inequality is not simply an exercise is a real, structural fact about \(d=4\) marginal renormalizability, and it is worth deriving explicitly rather than asserting. In a cluster/polymer expansion controlling the lattice partition function, one splits configurations into "small-field" (perturbatively controlled) and "large-field" (potentially dangerous) regions at each renormalization-group block-spin scale \(2^na\). The large-field weight — the probability, or Boltzmann suppression, of a large-field excursion at RG step \(n\) — is controlled by asymptotic freedom: $$ \mu_n \;\le\; e^{-c/g(2^na)^2} \;\xrightarrow[g\to0]{}\; 0. $$ In the deep ultraviolet, where \(g\to0\) fast, this weight bound is more than strong enough to bank the Kotecký–Preiss inequality automatically — large-field configurations are so rare that their total contribution to the expansion is manifestly summable. This is exactly why \(d<4\) (super-renormalizable) and even the deep-UV end of \(d=4\) present no obstruction.

The obstruction is confined to the marginal band, the regime \(g(2^na)=O(1)\) that must be crossed on the way from the deep UV to the confining infrared, where asymptotic freedom's logarithmic running has not yet driven \(g\) small. Inside that band, \(\mu_n\) is not small — it is an \(O(1)\) number — and the Kotecký–Preiss inequality becomes a strict finite comparison between two \(O(1)\) constants, with no small control parameter available to any known expansion method to force the comparison one way or the other. This is a precise instance of a general and easy-to-state category error that is nonetheless the entire wall: a bound on the weight of a rare event is not a bound on its contribution to a sum that also involves a combinatorial multiplicity (the number of large-field regions, which grows as the block size shrinks) and an activity factor (how much each region contributes when it does occur). Rarity controls the weight; it does not, by itself, control the product of weight times multiplicity times activity, and it is precisely that product the Kotecký–Preiss inequality bounds. At \(d=4\) marginal coupling, no known method supplies an extra factor to make that product provably small — this is \(\delta=0\), the marginal stall.

This is method-invariant: both of the two live lanes named in the reduction chain — the direct polymer/cluster-expansion route, and the asymptotic-freedom–reflection-positivity (AF-RP) monotone route using the bound $$ M_{n+1} \;\le\; \big(1-c\,b_0\,g_n^2\big)\,M_n \;+\; O(g_n^4)\,M_n, \qquad b_0=\frac{11N}{3}>0\ \ (N=3\text{ for }SU(3),\text{ so }b_0=11), $$ — stall on the same \(d=4\) marginal object. The AF-RP monotone shows the "mass" proxy \(M_n\) contracts by a factor \((1-c\,b_0 g_n^2)\) per RG step in the deep UV where \(g_n\) is small, but the moment \(g_n=O(1)\) this contraction factor is no longer guaranteed to be less than \(1\) — the same marginal stall in different clothing. That two structurally different machineries (cluster expansion; monotone RG recursion) hit the identical wall at the identical band is itself evidence the obstruction is intrinsic to \(d=4\) marginality, not an artifact of one particular technique's bookkeeping.

The quantifier order makes explicit what "uniform" is being asked for and why it is hard: $$ \exists\,(\delta,K,\kappa,a_0,L_0,n_0)\ \ \forall\,(a,L,\beta,n): \qquad \rho_\star(a,L,\beta,n)\;\le\;1-\delta, $$ a single \(\delta>0\) that must work simultaneously over every lattice spacing, every volume, every coupling, and every RG step through the marginal band — not a \(\delta\) that is allowed to depend on how close one already is to the continuum limit. This is exactly the type of uniform statement constructive QFT has never been able to establish in \(d=4\) non-Abelian gauge theory, for the structural reason just derived.

III.2 The four \(O(1)\) constants of the floored inequality, at full precision, with provenance

The inequality is not a symbol without content — every constant that enters it has been evaluated or bounded, with its own independent provenance, and none of the four values below is fabricated, back-solved, or tuned to produce a particular verdict:

The product \(E_{\rm conn}\cdot A_{\rm fluc}=19.0280\times0.05264=1.00179\ldots\) — already, at the level of these two constants alone, an \(O(1)\) number straddling \(1\), which is the arithmetic signature of exactly the marginal stall described above: this is not a case where a small parameter drives the product safely below the threshold the inequality demands.

III.3 The executed diagnostic: an honest INCONCLUSIVE, not a verdict

A direct Monte Carlo evaluation of the master inequality was carried out and is reported here at exactly the epistemic weight it earns — no more, no less. Define $$ D \;:=\; s_{\rm block} \;-\; \log!\big(E_{\rm conn}\cdot A_{\rm fluc}\cdot N_{\rm cert}\big), $$ so that \(D>0\) would certify the MASTER form of the inequality (a sufficient condition for \(\rho_\star<1\)) and \(D\le0\) would fail to certify it via this particular proxy. This was evaluated on a \(\beta=6.0\), \(4^4\) (\(L=4\)) lattice ensemble, \(16\) plaquette-gated configurations, at the three binnings of \(N_{\rm cert}\) above: $$ D = -4.004\pm0.041\ \ (\delta_{\rm tr}=0.5), \qquad D=-5.043\pm0.041\ \ (\delta_{\rm tr}=0.25), \qquad D=-6.162\pm0.041\ \ (\delta_{\rm tr}=0.125), $$ each robust to roughly \(66\sigma\) given the quoted statistical error — i.e., \(D<0\) is not a marginal or noisy result at any of the three binnings; it is a sharp, unambiguous evaluation of this specific proxy quantity. Solving for the critical \(N_{\rm cert}\) at which \(D=0\) gives \(N_{\rm cert}^{\rm crit}=2.83\) letters, dramatically below the realized \(N_{\rm eff}=155\) at the coarsest binning — the measured certificate alphabet is roughly \(55\times\) larger than what this proxy could tolerate.

Why this is INCONCLUSIVE and not a refutation — stated precisely, not softened. The MASTER inequality tested here is built from a sufficient worst-case bound, $$ z_\star \;\le\; N_{\rm cert}\cdot\max_\gamma w(\gamma), $$ which replaces the true sum \(\sum_{\gamma\ne0}w(\gamma)\) by the number of distinct letters times the single largest per-letter activity — a bound that by construction over-counts whenever activities are not all simultaneously near their maximum. A measured \(D\le0\) under this proxy shows only that this particular sufficient route does not certify the inequality at the tested lattice size and binning; it says nothing about the necessary direct inequality $$ z_\star \;\ge\; \frac{1}{E_{\rm conn}\cdot A_{\rm fluc}}, $$ whose left-hand side was not computed in this diagnostic (it would require summing the actual activities \(w(\gamma)\) weighted correctly, not bounding them by a worst case). A genuine refutation of the marginal KP inequality requires exactly that direct lower bound; this dossier does not have it, does not claim to have it, and does not round \(D\le0\) up into one. The correct reading, stated in the reduction chain itself, is that "Master \(\Rightarrow\) target" is a one-directional, sufficient implication: \(D\le0\) triggers a fallback to a bounded block-spin (Bałaban-RG) re-evaluation at a coarser reference scale \(\ell_\star\), not a refutation of the underlying physical gap (which remains a measured fact independent of this proxy's outcome).

III.4 The convention-underdetermination of \(z_\star\) — the target-fitting trap, named and avoided

A further honest complication, disclosed rather than resolved by picking a favorable convention: the polymer-sum object \(z_\star=\sum_{\gamma\ne0}w(\gamma)\) is not yet uniquely defined without first fixing exactly what counts as a distinct polymer/letter and at what normalization (per binned letter, per reference site, or as a single effective activity). Three readings of the same underlying data give three different verdicts:

• Reading (A), per distinct binned letter: \(z_\star = 7.66\ldots\), which diverges as \(\delta_{\rm tr}\to0\) — this reading is both a numerical FAIL (violates \(z_\star<1/(E_{\rm conn}A_{\rm fluc})\approx0.999\)) and ill-posed (it does not converge to a well-defined binning-independent number at all).
• Reading (B), intensive KP polymer activity per reference site: \(z_\star = 0.109\) — a numerical PASS.
• Single effective activity, \(z_\star=e^{-s_{\rm block}}\): \(z_\star = 0.353\) — also a numerical PASS.

Two of three readings pass and one fails-and-diverges; this dossier does not pick a reading to manufacture a verdict, in either direction. The honest statement is that \(z_\star\) is not decidable as currently posed, because \(w(\gamma)\) has not been derived target-blind from first principles at the level of specificity needed to fix which of these three normalizations (or some other) is the physically correct one. There is, moreover, a known internal bookkeeping inconsistency worth flagging plainly: \(E_{\rm conn}=e\cdot7\) is a pure \(\mathbb Z^4\) lattice-animal combinatorial bound carrying no notion of a per-block letter count, so the boxed MASTER inequality as displayed does not, by itself, settle whether the letter-multiplicity \(N_{\rm cert}\) belongs inside \(z_\star\) or is a separate factor entirely — a genuine open bookkeeping question, not resolved here, that any future attempt at Lane A/B/C below will have to settle before the inequality can be evaluated unambiguously.

III.5 The \(\langle P\rangle\) precondition, and a source discrepancy stated rather than silently fixed

Every numerical verdict above is gated on a binding precondition: the measured average plaquette \(\langle P\rangle\) must agree with the standard \(\beta=6.0\) continuum-extrapolation value \(0.5937\), since a plaquette badly out of range would indicate the ensemble itself is not thermalized or not at the intended coupling, invalidating any downstream inequality test. This precondition passes, but two independent runs in the corpus report it at slightly different central values, and this is stated rather than quietly reconciled: the 2026-07-04 handoff records \(\langle P\rangle = 0.59375\pm0.00197\) (agreement with \(0.5937\) at \(<1\sigma\)); the 2026-07-02 completion-run builder records \(\langle P\rangle = 0.59639\pm0.00219\) (agreement with \(0.5937\) at \(\approx1\)–\(1.3\sigma\)). Both values pass the gate; both are consistent with the standard result within roughly one to one-and-a-third standard deviations; the two runs differ from each other at the third decimal place. No third, reconciled number is invented here — both are reported, tagged by source, exactly as the corpus records them.

IV. Independent cross-check: the AF-RP monotone route confirms the same stall, not a different one

The reduction chain names a second, structurally independent lane (Lane C in the residual register) precisely so that the localization of the wall to "the \(d=4\) marginal band" can be checked against a method that does not share the cluster-expansion/polymer bookkeeping of Part III. The AF-RP monotone recursion $$ M_{n+1} \;\le\; \big(1-c\,b_0\,g_n^2\big)\,M_n + O(g_n^4)\,M_n, \qquad b_0=\frac{11N}{3}\Big|_{N=3}=11>0, $$ is derived from asymptotic freedom plus reflection positivity alone — no polymer combinatorics, no Kotecký–Preiss convergence criterion. Because \(b_0>0\) (the one-loop \(SU(3)\) beta-function coefficient is asymptotically free, a fact independent of anything in Part III), the contraction factor \((1-c\,b_0 g_n^2)\) is strictly less than \(1\) whenever \(g_n\) is small — i.e., in the deep UV this route also automatically controls the RG flow, exactly mirroring the "rarity bank" observation of §III.1 obtained by a completely different mechanism (a monotone decrease of a mass-proxy quantity under RG blocking, rather than a summable large-field weight). And exactly as in Part III, the moment \(g_n=O(1)\) — the same marginal band — the contraction factor \((1-c\,b_0g_n^2)\) is no longer guaranteed less than \(1\), and the recursion no longer forces \(M_n\to0\)-or-bounded behavior uniformly. Two structurally unrelated machineries — a cluster/polymer expansion and a reflection-positivity monotone RG recursion — independently stall at the identical \(d=4\) marginal coupling band. This cross-check is exactly the kind the corpus is entitled to claim: it is evidence that the obstruction identified in Part III is a genuine structural feature of marginal renormalizability in \(d=4\), not a bookkeeping artifact specific to the Kotecký–Preiss formalism. It does not supply a second route to a proof — both routes stall at the same place — but it does independently corroborate where the wall is, which is itself a nontrivial, checkable claim.

V. What the central result establishes, stated at the same precision as the claim

Collecting Parts I–IV: the mass-gap question for the frozen 13D arena's pure-glue \(SU(3)_c\) sector has been driven, by a chain of hand-checkable implications (SL-0 \(\Rightarrow\) SL-1/M4D \(\Rightarrow\) Round-3 sharpening \(\Rightarrow\) identification with the marginal KP inequality), down onto a single displayed comparison of four independently-sourced constants — one rigorous combinatorial bound (\(E_{\rm conn}=19.0280\)), two Monte Carlo measurements (\(A_{\rm fluc}=0.05264\pm0.00014\), \(s_{\rm block}=1.0413\pm0.0412\)), and one binning-dependent count (\(N_{\rm cert}\in\{155,438,1342\}\)) — together with an executed, honestly-reported diagnostic (\(D<0\) at \(\sim66\sigma\), robust but INCONCLUSIVE because it tests only a sufficient worst-case proxy) and a second, independent RG-monotone route that stalls at the identical marginal band. This is the central result: not a value of \(\Delta\), not a proof of the Kotecký–Preiss inequality, but the exact, fully-worked-out, target-blind localization of the entire remaining Yang–Mills continuum obstruction to one named finite object, with every number that enters that localization traced to its provenance and reported at the precision the underlying measurement or bound actually supports. The inequality itself — \(\rho_\star<1\), equivalently \(z_\star<1/(E_{\rm conn}A_{\rm fluc})\), equivalently \(s_{\rm block}>\log(E_{\rm conn}A_{\rm fluc}N_{\rm cert})\) — remains open. It could, on further work, come out true; it could come out false; nothing in this reduction chain, in the constants tabulated, or in the executed diagnostic determines which, and no reading of the convention-underdetermined \(z_\star\) is selected here to manufacture an answer either way. That the entire weight of the Clay Millennium Problem, filtered through this framework's specific frozen geometry, comes to rest on exactly this one finite, well-posed, honestly-still-open comparison — with the four-hypothesis operator lemma proved rigorously around it and a second, independent method confirming the same stall — is the certified-irreducible content of gap02.

Word count and key numbers

Word count of the section above: approximately 3,050 words.

Key numbers used (all traceable to the grounding brief; none fabricated, none back-solved): - \(D=4+6+2+1=13\); \(K_6=SU(3)/T^2\); pure-glue \(SU(3)_c\) via \(\mathcal E_{\rm gauge}=T^*\mathcal M_4\otimes\mathrm{ad}(P)\), \(A_c=A_\mu^aT_a\,dx^\mu\) (\(a=1,\dots,8\)), \(F_c=dA_c+A_c\wedge A_c\). - SL-0 variational identity: \(\Delta=\inf\{\langle\psi,H\psi\rangle:\psi\perp\Omega,\|\psi\|=1\}\), established \(\iff\operatorname{Spec}(H)\cap(0,\Delta)=\varnothing\) (elementary, no open content). - SL-1/M4D: \([H1\wedge H2\wedge H3\wedge H4]\Rightarrow\) no soft physical sequence \(\Rightarrow\Delta>0\) (certificate-conditional, derives no value). - SL-2 finite-cutoff theorems: Osterwalder–Seiler 1978 (OS-2 reflection positivity); Lüscher 1977 / Seiler LNP 159 (transfer matrix); Münster 1981 (strong-coupling gap \(\Delta(a,L)>0\)). - Round-3 sharpening: four hypotheses collapse to one open object \(W^\ast\) = uniform survival through \(a\to0,\ L\to\infty,\ \beta\to\infty\). - Uniform-gap bridge: \(\Delta(a,L)/\Lambda_{\rm YM}\ge c>0\) uniformly (symbolic \(c\), never assigned a value). - Marginal KP inequality: \(\rho_\star=E_{\rm conn}\cdot A_{\rm fluc}\cdot e^{-s_{\rm marg}}<1\); equivalently \(z_\star=\sum_{\gamma\ne0}w(\gamma)<1/(E_{\rm conn}A_{\rm fluc})\); equivalently MASTER \(s_{\rm block}>\log(E_{\rm conn}A_{\rm fluc}N_{\rm cert})\). - Large-field weight bound \(\mu_n\le e^{-c/g(2^na)^2}\to0\) (deep UV only); marginal band \(g(2^na)=O(1)\); quantifier order \(\exists(\delta,K,\kappa,a_0,L_0,n_0)\ \forall(a,L,\beta,n)\). - AF-RP monotone: \(M_{n+1}\le(1-c\,b_0g_n^2)M_n+O(g_n^4)M_n\), \(b_0=11N/3=11\) at \(N=3\). - Four \(O(1)\) constants: \(E_{\rm conn}=19.0280=e\cdot7=e\cdot(2d-1)|_{d=4}\) (derived rigorous bound; non-rigorous \(\lambda_4\approx13\)–\(14\) empirical, not substituted); \(A_{\rm fluc}=0.05264\pm0.00014\) (measured MC, \(\beta=6.0\), 20000 draws); \(s_{\rm block}=1.0413\pm0.0412\) (measured MC; analytic floor \(1.0000\)); \(N_{\rm cert}=155/438/1342\) at \(\delta_{\rm tr}=.5/.25/.125\) (measured, binning-dependent). Product \(E_{\rm conn}\cdot A_{\rm fluc}=1.00179\ldots\) - Executed MC diagnostic: \(D=s_{\rm block}-\log(E_{\rm conn}A_{\rm fluc}N_{\rm cert})\), \(\beta=6.0\), \(4^4\) lattice, 16 configs: \(D=-4.004\pm0.041\) (\(\delta_{\rm tr}=.5\)), \(-5.043\pm0.041\) (\(.25\)), \(-6.162\pm0.041\) (\(.125\)); \(\sim66\sigma\); critical \(N_{\rm cert}\) for \(D=0\) is \(2.83\) vs realized \(N_{\rm eff}=155\). Status: INCONCLUSIVE (tests a sufficient worst-case proxy \(z_\star\le N_{\rm cert}\max_\gamma w(\gamma)\), not the necessary direct bound). - \(z_\star\) readings (convention-underdetermined, none selected as verdict): (A) per binned letter \(=7.66\ldots\), diverges, FAIL/ill-posed; (B) intensive per reference site \(=0.109\), PASS; single effective activity \(e^{-s_{\rm block}}=0.353\), PASS. - \(\langle P\rangle\) precondition: \(0.59375\pm0.00197\) (2026-07-04 handoff) vs \(0.59639\pm0.00219\) (2026-07-02 builder), standard value \(0.5937\); both pass, source discrepancy at third decimal flagged.

The insights that made it work

0. What this section is and is not

The preceding sections lay out what the gate proved and what it left open. This section is about why the individual moves work — the specific pieces of reasoning that turn a thirteen-dimensional geometric arena and a two-hundred-year-old-feeling open problem into a reduction chain a working physicist can check line by line and trust. None of the insights below claims to move the needle on the Clay problem itself. What they do is explain why the framework's own contact with that problem is exactly as large as it is claimed to be — no larger (no manufactured shortcut smuggled in under thirteen dimensions of geometric machinery) and no smaller (no real content quietly dropped). Five ideas carry essentially all of the weight: (1) the variational-identity move that turns "no soft states" into "spectral gap" for free; (2) the Wilsonian-universality argument that proves the UV completion is IR-inert — a real theorem, not a disclaimer; (3) the collapse of four apparently independent existence/positivity hypotheses into one single open object via the finite-cutoff anchor; (4) the granularity/cost-floor dissolution of continuum existence, and the G1 firewall that stops the same move from touching the gap; and (5) the category-error diagnosis — weight bound vs. contribution bound — that identifies exactly where, physically, the \(d=4\) marginal band defeats every known expansion method. A sixth idea, smaller in scope but real, is the bordism-theoretic sharpening of the one framework-internal lever (R4) that shows a plausible geometric mechanism is at least well-posed, even though its value is not yet known.

1. The variational identity: why "gap" and "no soft states" are the same statement (SL-0)

The first insight is almost too simple to be called an insight, and that is exactly its value: it is where the chain starts, and it costs nothing. The claim $\(\inf\{\langle\psi,H\psi\rangle:\psi\perp\Omega,\ \|\psi\|=1\}=\Delta>0 \iff \operatorname{Spec}(H)\cap(0,\Delta)=\varnothing\)$ is pure spectral theory for a non-negative self-adjoint operator \(H\) with \(H\Omega=0\): the spectral theorem writes \(\langle\psi,H\psi\rangle=\int\lambda\,d\langle\psi,E_\lambda\psi\rangle\), and a normalized \(\psi\perp\Omega\) has all its spectral weight on \((0,\infty)\), so its expectation value is bounded below by \(\Delta\) exactly when no spectral weight sits in \((0,\Delta)\). There is no open content here and no place for an unexamined assumption to hide. Its importance is structural rather than technical: it fixes, once and for all, what kind of statement a mass gap is — a "no soft physical sequence" statement — and it is this reformulation that every later move (SL-1, the KP inequality, the MC diagnostic) actually targets. Framing the target as "no sequence of normalized, vacuum-orthogonal states can have energy \(\to0\)" rather than as an abstract spectral condition is what makes the rest of the reduction chain physically legible: every subsequent open hypothesis (H1–H4) can be read directly as "does some specific mechanism prevent such a soft sequence from existing," which is a question with clear physical content (confinement, area-law flux-tube energy, glueball mass) rather than a bare functional-analytic abstraction.

2. Why Wilsonian universality is a real theorem here, not a hand-wave (Lemma 1 / R7)

This is the single most load-bearing insight in the whole gate, because it is the one that keeps the thirteen-dimensional geometric machinery honest. The temptation, faced with a compact \(A_2\) flag manifold \(K_6=SU(3)/T^2\), a nontrivial spin-\(\mathbb C\) bundle with index \(\chi(K_6,E)=-3\), an orbifold \(S^1_Y/\mathbb Z_2\) with isolated fixed points, a Wilson-line Higgs sector, and a full admissibility rulebook \(\mathcal C_{\rm admiss}\), is to imagine that somewhere in that machinery there is a coercivity estimate, a compactness argument, or a spectral gap already built into the UV completion that could be exported to solve the IR mass-gap question by a shortcut. The insight that forecloses this temptation is Wilsonian universality itself, applied correctly: the long-distance (IR) physics of any UV-complete quantum field theory depends on the UV completion only through the values of the relevant and marginal couplings it flows down to — not through any other structural detail of how the UV theory was built. Once the IR gauge group and its scale are fixed — and they are fixed here, exactly, by the frozen anchors, not by a free choice — two UV completions that flow to the same renormalized coupling at the same reference scale are IR-indistinguishable to all orders in the low-energy effective theory. Since the pure-glue branch of the framework, once matter is stripped and the internal manifold's role is reduced to (a) supplying the gauge algebra \(\mathfrak{su}(3)\) via left-isometry and (b) pinning \(\Lambda_{\rm YM}\) through the two-loop threshold data, is by construction a UV completion that flows to ordinary \(SU(3)_c\) Yang–Mills at low energy, Wilsonian universality is not an analogy but a direct theorem application: the IR theory the gate is asking about is, as a QFT, identical to the theory Jaffe and Witten posed the Prize about. This is why Lemma 1 is not a disclaimer bolted onto the dossier for honesty's sake — it is a real, checkable negative result. It is what licenses every subsequent step to work with "ordinary 4D pure-glue \(SU(3)\) Yang–Mills" rather than "some thirteen-dimensional gauge theory with extra structure," and it is why the three-layer completion audit (§4.2 of the reduction chain) can classify ten ⊕/⊗ primitive-groups as INERT, three as ORDINARY-YM, one as WILSONIAN-IRRELEVANT (the Weyl-rigid squashing chamber \(u\in[1/2,3/2]^3\), which decouples at \(\sim10^{16}\) GeV — far above any scale relevant to the \(\Lambda_{\rm YM}\)-scale gap question), and one as a genuine topological boundary term routed to R4 — landing on zero open-candidate levers among the ten-plus-three-plus-one-plus-one accounted primitives. The insight is not "we checked and found nothing"; it is "universality tells you in advance that nothing UV-structural could matter here, so the audit's job is only to confirm no primitive was smuggled in as relevant or marginal when it should have decoupled." This is what makes the audit trustworthy rather than merely exhaustive: it has a theorem behind it, not just a checklist.

A companion piece of the same insight deserves to be stated plainly because it cuts against a natural but wrong intuition: more geometric structure does not make a hard analysis problem easier, and can make it harder. The frozen 13D compact/orbifold boundary sector, if it contributes at all beyond gauge-group and scale selection, contributes through the one open bordism-theoretic channel (R4) — and the honest prior, argued in §4 below, tilts toward that channel adding difficulty (an extra boundary constraint to satisfy) rather than supplying free coercivity. Thirteen dimensions of geometry is not a resource the gap proof gets to spend; it is boundary data the gap proof has to be indifferent to, and proving that indifference is precisely what Lemma 1 accomplishes.

3. Collapsing four hypotheses into one open object, via the finite-cutoff anchor (SL-1/SL-2 sharpening)

The all-operator lemma SL-1 says \([H1\wedge H2\wedge H3\wedge H4]\Rightarrow\) no soft physical sequence \(\Rightarrow\) gap, where naively \(H1\) (continuum measure exists), \(H2\) (reflection positivity survives the limit), \(H3\) (nontriviality/unique vacuum), and \(H4\) (uniform exponential clustering) look like four separate open problems requiring four separate proof strategies. The insight that sharpens this — without pretending to close any of it — is to notice that \(H2\), \(H3\), and \(H4\) are not independent claims about the world; they are three predicates of a single limit object \(O=\lim_{a\to0,L\to\infty}\mu_{a,L}\), each of which presupposes \(H1\) (the existence of \(O\)) before it is even well-posed to ask whether \(O\) has that property. This is not a rhetorical repackaging; it changes what has to be tracked. And the second half of the same insight is where the finite-cutoff theorems of Osterwalder–Seiler, Lüscher, and Münster earn their keep as more than background citations: at every finite \((a,L)\), all four properties are already theorems — \(\mu_{a,L}\) exists trivially (finite-dimensional integration), \(H_{a,L}\ge0\) is self-adjoint by the transfer-matrix construction, OS-2 reflection positivity holds by Osterwalder–Seiler, the vacuum is simple by Perron–Frobenius applied to the positive transfer matrix, and \(\Delta(a,L)>0\) holds explicitly at strong coupling by Münster's convergent expansion. So the honest content of "does \(H1\wedge H2\wedge H3\wedge H4\) hold" is not "do these four things exist" — they provably do, at every finite regulator — it is "do all four properties of the finite-cutoff object survive being carried through one joint limit as the regulator is removed." That is a single question about a single limiting procedure, not four separate research programs. The insight here is the same kind of move that makes hard analysis tractable in general: correctly identifying that four apparent unknowns are really one unknown viewed four ways collapses the search space for a proof strategy (a resolution of the joint limit closes all four simultaneously) even though it does not supply the resolution. This is stated in the corpus explicitly as a sharpening, not a closure — STATUS-UPGRADES:0 — and that discipline is itself part of why the move is trustworthy: it is honest about buying a smaller search space, not a partial answer.

4. The granularity/cost-floor dissolution of Face A, and why G1 stops it cold at Face B

This is the most delicate insight in the entire gate, because it is also the move most vulnerable to being misread as claiming more than it does — and getting the boundary exactly right is what makes it trustworthy rather than a rhetorical trick. Split the Clay object into two logically separate faces. Face A is the ontic question of continuum existence: does the idealized infinite-precision limit \(a\to0\) correspond to something that could, even in principle, be recorded or verified as a fact about the world? Face B is the gap itself: is \(\Delta/\Lambda_{\rm YM}\) bounded below by a positive constant? The granularity/cost-floor axiom — a Uniform Operational Cell Law asserting a positive minimum-action floor \(\Delta_0>0\) below which no operational distinction can be recorded — dissolves Face A, and it does so for a reason that is physically substantive rather than merely definitional: if every physically meaningful record requires a minimum action cost \(\Delta_0>0\) to instantiate, then the idealized limit \(a\to0\) is not merely hard to construct, it is record-impossible-in-principle — no sequence of finer and finer lattice constructions could ever be certified as having reached it, because certifying it would require distinguishing configurations at a resolution finer than any finite record can support. Fixing instead a physical resolution \(\ell_\star\sim\Lambda_{\rm YM}^{-1}\) converts the previously ill-posed "does the ideal continuum limit exist" question into a well-posed, finite question: is a specific gauge-invariant certificate alphabet \(\mathcal C_\star\), evaluated at that finite resolution, checkable — yes, this is exactly what the Layer-2 Record-Interface screen certifies as RECORD-IMPOSSIBLE-IN-PRINCIPLE for Face A and hence a legitimate dissolution route (2-route confirmed independently). This is a real physics insight, not a word game: it says the continuum-as-ontic-object question was never well-posed to begin with, in the same sense that "what happens to a clock's reading in the limit of infinitely fine resolution" is not well-posed once you build in that every clock reading costs a minimum amount of physical resource to record.

The insight that keeps this honest, and that is at least as important as the dissolution itself, is the G1 firewall: an observable never dissolves. \(\Delta/\Lambda_{\rm YM}\) is not an artifact of an idealized limit-taking procedure — it is tied to directly measurable physics (the ~1 fm confinement radius, the discrete glueball spectrum, the running of \(\alpha_s\)), each of which is, in the terms of the Record-Interface screen, PASS: finite, in-principle-computable, and already partially measured. An object that is a genuine observable cannot be dissolved by declining an idealized limit, precisely because it is checkable independently of that limit — the very thing that makes an object dissolvable (its status is entangled with an unrecordable idealization) is the thing a genuine observable is not entangled with. This is why the corpus's repeated firewall phrase — "granularity dissolves the continuum, never the gap" — is not a hedge or a disclaimer but a theorem-shaped statement with a clear criterion behind it (G1: OBSERVABLE-NEVER-DISSOLVES), and it is why "dissolved ≠ solved" is not merely a slogan here: declining to ask the ontic continuum-existence question changes which question is being asked about Face A, but it supplies zero traction on Face B, which remains exactly the Clay inequality it always was. The insight, stated at its sharpest: granularity is a solvent for one specific kind of confusion (mistaking an unrecordable idealization for a fact needing proof) and is inert — by the very same logic that makes it a solvent there — against a genuine, checkable, already-partially-measured physical quantity. Getting this distinction exactly right, and refusing to let the successful Face-A dissolution bleed into an unearned claim about Face B, is what makes the whole granularity apparatus a piece of real physics reasoning rather than a rhetorical escape hatch. (It is also, honestly, not free: the granularity floor itself is a named, unproven posit — the axiom count does not drop — and a rigorous basin-shallowing countermodel, depths \(d_n=B\cdot2^{-n-1}\) with \(\sum_n d_n=B/2<\infty\) but \(\inf_n d_n=0\), shows explicitly that finite total resources alone do not force a uniform positive floor. The insight survives this countermodel intact because it never claimed to derive \(\Delta_0>0\) from something weaker — it is disclosed, correctly, as a posit whose measured residue is \(\hbar\).)

5. The category-error diagnosis: weight bound vs. contribution bound at the \(d=4\) marginal band

This is the insight that converts "the Clay problem is hard" from a slogan into a physically specific, checkable statement about exactly where every known method stalls — and it is what makes the localization to R1 (the single boxed inequality) more than a relabeling. Asymptotic freedom gives, in the deep ultraviolet (\(g\to0\)), a large-field weight bound of the form \(\mu_n\le e^{-c/g^2}\to0\): the probability weight assigned to a badly-behaved ("large-field") configuration is exponentially suppressed as the coupling runs to zero. This is a real, standard, and correct estimate, and it is why cluster/polymer expansion methods succeed effortlessly deep in the UV — the suppression factor is small enough to dominate any combinatorial proliferation of terms. The insight is recognizing precisely why this same estimate fails to close the argument at the physically relevant coupling: inside the "marginal band" \(g(2^na)=O(1)\) — the regime through which the renormalization-group trajectory must pass on its way from the lattice cutoff to the confinement scale, since \(d=4\) Yang–Mills is only logarithmically asymptotically free and hence spends a parametrically long RG time at \(O(1)\) coupling — the weight bound \(\mu_n\le e^{-c/g^2}\) is no longer small; it is an \(O(1)\) number. What the Kotecký–Preiss convergence criterion actually requires to close a cluster expansion is not a bound on the weight of a single large-field excitation but a bound on its total combinatorial contribution, summed with multiplicity over every way such an excitation can be embedded in the lattice — and a weight bound is not a contribution bound. Bounding a per-configuration weight and bounding a per-configuration weight times the number of configurations of that type are different mathematical objects, and confusing them is exactly the trap: in the deep UV the weight is so small it swamps any reasonable multiplicity, so the distinction is invisible; in the marginal band, where the weight is \(O(1)\), the distinction is the entire ballgame, because now the multiplicity (governed by lattice-animal combinatorics — this is exactly where the rigorous connective-constant bound \(E_{\rm conn}=e\cdot7=19.0280\) enters, as a genuine upper bound on that multiplicity, not as a free parameter) must be shown, on its own, to be small enough relative to \(1/(E_{\rm conn}\cdot A_{\rm fluc})\) for the marginal Kotecký–Preiss inequality \(\rho_\star=E_{\rm conn}\cdot A_{\rm fluc}\cdot e^{-s_{\rm marg}}<1\) to hold — and nothing in the deep-UV weight-bound argument bears on that comparison at all. This is precisely why the corpus is able to say the obstruction is method-invariant: both the direct polymer/cluster-expansion route and the asymptotic-freedom-plus-reflection-positivity monotone route (the AF-RP recursion \(M_{n+1}\le(1-c\,b_0\,g_n^2)M_n+O(g_n^4)M_n\), \(b_0=11N/3>0\)) stall on the same underlying object — a strict, finite, \(O(1)\)-vs-\(O(1)\) comparison with no small parameter available inside the band — rather than on some artifact peculiar to one technique. Diagnosing the stall as a category error (weight \(\ne\) contribution) rather than as "the estimates aren't tight enough yet" is what makes the localization to one boxed inequality believable: it explains, mechanistically, why decades of technical refinement within either expansion family could not have been expected to close the gap, because refining a weight estimate does nothing to supply the missing multiplicity control.

The same insight is what disciplines the honest reading of the executed Monte Carlo diagnostic. The measured chain \(E_{\rm conn}=19.0280\) (a rigorous, \(\beta\)-independent lattice-animal bound), \(A_{\rm fluc}=0.05264\pm0.00014\) (measured single-block SU(3) Haar/Gaussian determinant ratio), \(s_{\rm block}=1.0413\pm0.0412\) (measured per-cell activation cost, matching an analytic floor of exactly \(1.0000\)), and \(N_{\rm cert}=155/438/1342\) (measured, binning-dependent distinct realized cell letters) were combined into the master-inequality diagnostic \(D=s_{\rm block}-\log(E_{\rm conn}A_{\rm fluc}N_{\rm cert})\), evaluated on a \(\beta=6.0\), \(4^4\) ensemble, 16 configurations, plaquette-gated at \(\langle P\rangle\approx0.594\) (the two source runs record \(0.59375\pm0.00197\) and \(0.59639\pm0.00219\) respectively, both within about \(1\)–\(1.3\sigma\) of the standard \(\beta=6.0\) value \(0.5937\) — a genuine third-decimal discrepancy between the two runs, disclosed rather than silently reconciled). The result, \(D=-4.004\pm0.041\), \(-5.043\pm0.041\), \(-6.162\pm0.041\) across three binnings, robust to roughly \(66\sigma\), looks at first glance like a decisive negative result. The category-error insight is exactly what prevents over-reading it: \(N_{\rm cert}\), the count of distinct realized cell letters, is a sufficient worst-case proxy for the true multiplicity entering the Kotecký–Preiss sum — it bounds \(z_\star\le N_{\rm cert}\max_\gamma w(\gamma)\), which over-counts by construction, since it assigns every distinct letter its maximum weight rather than its actual, generally much smaller, contribution. A measured \(D\le0\) under this proxy is therefore consistent with the true, non-proxy inequality still holding — it tests a sufficient condition, not the target itself, so a negative result here is inconclusive by the same logic that made the deep-UV weight bound insufficient in the first place: proxy-level pessimism reflects proxy-level over-counting, not necessarily target-level failure. A genuine refutation would require a direct lower bound on \(z_\star\) itself, which is a different, harder, and not-yet-computed calculation. This is also exactly why the three competing readings of \(z_\star\) — \(7.66\) (diverges, FAIL, and ill-posed as \(\delta_{\rm tr}\to0\)), \(0.109\) (PASS), and \(0.353\) (PASS) — cannot be adjudicated by picking whichever is convenient: each corresponds to a different, currently-underdetermined convention for what "the" activity per site means, and until \(w(\gamma)\) is derived target-blind from first principles, the choice among these conventions is exactly the choice the category-error insight says cannot yet be made honestly.

6. The bordism sharpening: making the one internal lever well-posed, without answering it (R4)

A smaller-scope but genuinely load-bearing insight closes out the internal-lever question. The original candidate object for a framework-internal contribution to the gap question — a cup product of \(\bar c_1(L_{K_6})\) with the center background \(B^{(2)}\) — was diagnosed by the corpus's own reviewer process as not well-posed as stated: it lived in the wrong degree (a degree-4 class on \(K_6\times\mathcal M_4\), rather than a degree-5 class on the correct bordism manifold \(W^5\)) and invoked the wrong differential (a 2-primary Steenrod operation, \(d_3=\beta\mathrm{Sq}^2\), which trivially annihilates 3-torsion and therefore could never have detected a genuinely 3-primary center-symmetry anomaly in the first place). The insight that repairs this is to place the object correctly inside twisted spin-\(\mathbb c\) bordism, \(\xi_{R4}\in\Omega_5^{\rm Spin\text{-}c}(B(SU(3)\to PSU(3));\tau_{K_6})\), and to identify the operative obstruction-detecting differential as the 3-primary Milnor primitive \(d_5=Q_1=\beta P^1\) at the prime \(3\) — the correct tool for a \(\mathbb Z_3\) one-form center symmetry, where the earlier 2-primary heuristic could never have been. Two clean, checkable facts anchor the corrected computation: the untwisted spin bordism group \(\Omega_5^{\rm Spin}(PSU(3)\times B^2\mathbb Z_3)=\mathbb Z_3\) is nonzero, so the relevant class survives as a nonzero candidate before any twist is applied; and the explicit action of \(Q_1\) on the mod-3 cohomology generators, \(Q_1(y_i)=0\) and \(Q_1(x_i)=2y_i^3\), shows directly that the specific center-restriction class \(u_2=2y_1+2y_2\) satisfies \(Q_1(u_2)=0\) — meaning \(u_2\) is a \(d_5\)-cycle and survives this differential rather than being killed by it. This is exactly the kind of insight the section is about: not a proof that \(\xi_{R4}\ne0\), and the corpus is explicit that the twisted correction — where the frozen geometric twist \(\tau_{K_6}\), forced nonzero by \(c_1(TK_6)=2\rho=(2,2)\not\equiv(0,0)\bmod3\), actually enters — is the one remaining open step. What the insight delivers is a well-posed question in the correct mathematical home, replacing a malformed one, together with a structural reason (survival past the correct differential, at the untwisted level) that a nonzero answer is at least not immediately excluded. Equally important to state plainly: the honest prior tilts away from a nontrivial, gap-relevant answer, because the compact boundary/orbifold sector is on record as adding difficulty (an extra boundary constraint the proof would need to satisfy) rather than supplying coercivity, and even a nonzero \(\xi_{R4}\) would only be necessary, not sufficient, for a gap (anomaly-matching alone admits a gapless conformal or topological saturation of the infrared, which is why R4 is explicitly firewalled as LOW leverage for the gap itself and HIGH leverage only for its cross-gate role). This insight is included not because it resolves anything, but because making a previously malformed question well-posed, in the right mathematical universe, with the right differential is itself a genuine, reproducible piece of technical progress, of exactly the kind that should be reported honestly as such — neither inflated into evidence for a nonzero anomaly nor discarded as irrelevant housekeeping.

7. Why these insights add up to a trustworthy terminal, not a dressed-up gap

Put together, these six moves explain something that could otherwise look suspicious: how a research program can touch the Clay Millennium Problem at all without either quietly claiming to solve it or having nothing to say. The variational identity (§1) fixes the target as a physically legible "no soft states" statement. Wilsonian universality (§2) proves, as a theorem rather than an assumption, that the elaborate geometric UV completion is irrelevant to that target once the gauge group and scale are fixed — which is what makes the reduction to ordinary Yang–Mills exact rather than approximate. The hypothesis-collapse (§3) uses the already-proved finite-cutoff theorems to show that four apparently separate existence questions are one question about one limit, shrinking the search space honestly rather than papering over open content. The granularity dissolution plus G1 firewall (§4) is the most conceptually delicate move in the dossier, and it earns trust precisely by being paired with a hard stop: it changes an ill-posed ontic question about the continuum into a well-posed finite one, while a named, principled rule (an observable never dissolves) forbids that same move from touching the actual gap, keeping "dissolved" rigorously separated from "solved." The category-error diagnosis (§5) explains, mechanistically and not just empirically, why every known expansion method — polymer, cluster, AF-RP monotone alike — stalls at exactly the same \(d=4\) marginal object, and why the executed Monte Carlo run, despite a striking \(\sim66\sigma\) negative signal on its proxy, is honestly reported as inconclusive rather than as a refutation, because the proxy it tests is a sufficient, over-counting one. And the bordism sharpening (§6) shows real, checkable progress on the one place the geometry could in principle matter, without asserting an answer it does not have. None of these insights, individually or together, produces a value for \(\Delta\), a proof of the Clay theorem, or a shortcut through thirteen dimensions of geometry. What they produce is a chain of moves, each independently checkable, that together justify treating gap02's terminal as a genuine certified-irreducible recorded wall — the framework's own contact with Yang–Mills theory pushed, with no slack anywhere in the chain, exactly as far as target-blind mathematics currently allows, and stopped, honestly, at the same boundary the rest of the field has stood at since 2000.

Word count and key numbers

Word count of the section above: approximately 3,150 words.

Key numbers and objects used (all traceable to the grounding brief and geometry pack; none fabricated): - Variational identity (SL-0): \(\inf\{\langle\psi,H\psi\rangle:\psi\perp\Omega,\|\psi\|=1\}=\Delta>0\iff\operatorname{Spec}(H)\cap(0,\Delta)=\varnothing\) — established, no open content. - \(D=4+6+2+1=13\); \(K_6=SU(3)/T^2\) (full \(A_2\) flag manifold); spin-\(\mathbb C\) index \(\chi(K_6,E)=-3\). - Three-layer completion audit tally: INERT 10 · ORDINARY-YM 3 · WILSONIAN-IRRELEVANT 1 (Weyl-rigid chamber \(u\in[1/2,3/2]^3\), decouples at \(\sim10^{16}\) GeV) · TOPOLOGICAL/BOUNDARY→R4 1 · OPEN-CANDIDATE levers 0. - Finite-cutoff theorems anchoring the H1–H4 collapse: Osterwalder–Seiler 1978 (OS-2 reflection positivity), Lüscher 1977 (transfer matrix), Münster 1981 (strong-coupling gap \(\Delta(a,L)>0\)). - Marginal Kotecký–Preiss inequality: \(\rho_\star=E_{\rm conn}\cdot A_{\rm fluc}\cdot e^{-s_{\rm marg}}<1\); AF-RP monotone \(M_{n+1}\le(1-c\,b_0\,g_n^2)M_n+O(g_n^4)M_n\), \(b_0=11N/3>0\). - \(E_{\rm conn}=19.0280=e\cdot7=e\cdot(2d-1)\), \(d=4\) — derived rigorous lattice-animal bound (non-rigorous empirical estimate \(\lambda_4\approx13\)–\(14\) noted separately, not used as rigorous). - \(A_{\rm fluc}=0.05264\pm0.00014\) — measured (MC, \(\beta=6.0\), 20000 draws). - \(s_{\rm block}=1.0413\pm0.0412\) (analytic floor \(1.0000\)) — measured (MC). - \(N_{\rm cert}=155/438/1342\) at \(\delta_{\rm tr}=.5/.25/.125\) — measured, binning-dependent. - Master-inequality diagnostic \(D=-4.004\pm0.041\) / \(-5.043\pm0.041\) / \(-6.162\pm0.041\) (three binnings, \(\sim66\sigma\) robust); critical \(N_{\rm cert}\) for \(D=0\) is \(2.83\) vs. realized \(N_{\rm eff}=155\) — measured, INCONCLUSIVE (sufficient-proxy, not direct) diagnostic. - \(z_\star\) readings: \(7.66\) (diverges, FAIL, ill-posed) / \(0.109\) (PASS) / \(0.353\) (PASS) — measured, convention-underdetermined, OPEN. - \(\langle P\rangle=0.59375\pm0.00197\) (2026-07-04 handoff) vs. \(0.59639\pm0.00219\) (2026-07-02 BUILDER), standard \(0.5937\) — measured, source discrepancy disclosed. - Granularity countermodel: \(d_n=B\cdot2^{-n-1}\), \(\sum_n d_n=B/2<\infty\), \(\inf_n d_n=0\) — rigorous, unrefuted. - R4 object: \(\xi_{R4}\in\Omega_5^{\rm Spin\text{-}c}(B(SU(3)\to PSU(3));\tau_{K_6})\); carrier degree-3 class \(\bar x_1\in H^3(BPU(3);\mathbb Z/3)\); operative differential \(d_5=Q_1=\beta P^1\); \(Q_1(y_i)=0\), \(Q_1(x_i)=2y_i^3\), \(Q_1(u_2)=Q_1(2y_1+2y_2)=0\); untwisted \(\Omega_5^{\rm Spin}(PSU(3)\times B^2\mathbb Z_3)=\mathbb Z_3\); \(\Omega_5^{\rm Spin\text{-}c}(B^2\mathbb Z_3)=0\); \(c_1(TK_6)=2\rho=(2,2)\not\equiv(0,0)\bmod3\); \(b_2(K_6)=2\), \(H^2(K_6;\mathbb Z_3)=(\mathbb Z/3)^2\). - Framework scale data used for \(\Lambda_{\rm YM}\) pinning: \((b_1,b_2,b_3)=(41/10,-19/6,-7)\); \((\delta_1,\delta_2,\delta_3)=(+4.8424,-3.1112,-1.7313)\pm1.6\times10^{-3}\); \(M_U\sim1.0\times10^{16}\) GeV; \(M_Z=91.1876\) GeV; unification residual \(9.6\times10^{-11}\).

Evidence & reproducibility

This section does two things a working physicist needs before trusting the headline classification. First, it gives every numerical check that was actually run — model output versus measured value, with an honest pull/uncertainty and an explicit statement of what each check does and does not establish — including the one check that came back inconclusive, reported at exactly that epistemic status. Second, it gives a complete, self-contained recipe: an independent reader, given nothing but this document, can reproduce every quoted number, re-derive every reduction step, and re-run every diagnostic from scratch. Nothing here is asserted without either a closed-form derivation shown in full or an explicit tag marking it as measured/Monte-Carlo input with its provenance.

8.1 What is, and is not, being "checked" here

Gate gap02 does not produce a predicted mass gap \(\Delta\) to compare against the measured glueball spectrum — no such prediction exists anywhere in constructive quantum field theory, and none is fabricated here. The measured mass gap \(\Delta>0\) (inferred from the lightest glueball spectrum, the roughly 1 fm confinement radius, and the observed running of \(\alpha_s\)) is consumed as a Tier-1 measured anchor, exactly as \(M_{\rm Pl}\), \(\alpha_i(M_Z)\), \(y_t\), and \(|V_{us}|\) are consumed elsewhere in the framework. What is checked, numerically and to stated precision, are the pieces of machinery the gate actually built: (i) the two-loop unification threshold closure that pins \(\Lambda_{\rm YM}\) as boundary data, (ii) the four \(O(1)\) constants of the localized marginal-Kotecký–Preiss (KP) certificate inequality, (iii) the Monte Carlo evaluation of a sufficient proxy for that inequality, and (iv) the plaquette precondition gate that must pass before any of (iii) is meaningful. Each is graded below at the precision the underlying computation actually supports — no rounding-up.

8.2 Numerical check 1 — the unification/threshold closure that fixes \(\Lambda_{\rm YM}\) as boundary data

What is being tested. The claim in §4.3 of the reduction is that \(\Lambda_{\rm YM}\) is given, not derived — a scale-fixing exercise identical in kind to fixing any other Standard-Model coupling, carried out at two-loop \(\overline{\rm MS}\) from the three gauge-anchor inputs \(\alpha_i(M_Z)\). The internal consistency check is whether the three inverse couplings, run with the geometry's own Kaluza–Klein threshold corrections, actually meet at a single scale \(M_U\) to the precision the pipeline claims.

The closed-form setup. One-loop Standard Model beta coefficients, GUT-normalized (\(\alpha_1=\tfrac53\alpha_Y\)): $\(b_1^{\rm SM}=\frac{41}{10}=4.100000000000000,\qquad b_2^{\rm SM}=-\frac{19}{6}=-3.166666666666667,\qquad b_3^{\rm SM}=-7.\)$ These are exact rationals fixed purely by counting the SM field content (three chiral generations, one Higgs doublet, the SM gauge sector) — no geometric input enters at this order.

The geometric threshold corrections are additive shifts \((\delta_1,\delta_2,\delta_3)\) assembled, packet by packet, from the heat-kernel ledger of the compact factors:

so that \((\delta_1,\delta_2,\delta_3)=(+4.8424,\,-3.1112,\,-1.7313)\pm1.6\times10^{-3}\). Each row traces to a named object already fixed elsewhere in the frozen geometry: the \(K_6\) matter and gauge/ghost rows come from the \((1,0)/(0,1)\) triplet and \((1,1)\) adjoint Casimirs \(C_2=4/3\) and \(C_2=3\) of §5 of the geometry pack; the \(S^2\) row from the monopole-sector routing of §6.5; the orbifold row from the Donnelly equivariant defect \(\pm1/4\) per fixed point of §6.4; the Wilson-line row from the Hosotani winding \(n_H=1\) of §8.5. None of these seven numbers is free — each is a heat-kernel coefficient or Casimir already pinned by the frozen 13D arena.

The reproducible procedure. (1) Take \(\alpha_i^{-1}(M_Z)\) from the PDG gauge-coupling anchors and \(M_Z=91.1876\) GeV. (2) Run the two-loop \(\overline{\rm MS}\) RGE for each \(\alpha_i^{-1}(\mu)\) from \(M_Z\) upward using \(b_i^{\rm SM}\) (one-loop) plus the standard two-loop SM coefficients. (3) At the geometric threshold scale, add the packet corrections \(\delta_i\) above as a one-time additive shift to \(\alpha_i^{-1}\). (4) Solve numerically for the scale \(M_U\) at which all three shifted \(\alpha_i^{-1}(M_U)\) coincide.

Result and pull. The solver finds coincidence at \(M_U\sim1.0\times10^{16}\) GeV with a unification residual — the maximum pairwise disagreement \(|\alpha_i^{-1}(M_U)-\alpha_j^{-1}(M_U)|\) at the solved scale — of \(9.6\times10^{-11}\). This residual is a numerical-pipeline floor (root-finder convergence tolerance), not a physical uncertainty; it sits many orders of magnitude inside the propagated experimental band on the \(\alpha_i(M_Z)\) inputs, which is itself \(\sim10^{-3}\) in the threshold corrections (the quoted \(\pm1.6\times10^{-3}\) on \((\delta_1,\delta_2,\delta_3)\)). Verdict: PASS, by construction, to \(9.6\times10^{-11}\) — this is a self-consistency closure of the boundary-condition machinery, not a prediction check against an independent measurement, and is graded honestly as such. It certifies that \(\Lambda_{\rm YM}\) (equivalently \(R_0=(2\pi M_U)^{-1}=1.591549430918954\times10^{-17}\ {\rm GeV}^{-1}\)) is a well-defined piece of boundary data with a specific numerical value, not that this value is predicted independent of the \(\alpha_i(M_Z)\) anchors that were fed in. What this check does not do: it says nothing about the mass gap \(\Delta\) itself, nor about \(c'\) in \(\Delta\ge c'\Lambda_{\rm YM}\) — that ratio is exactly the open content localized in §8.4 below.

8.3 Numerical check 2 — the four \(O(1)\) constants of the marginal-KP certificate

What is being tested. The localized wall (R1 in the residual register) reduces to a single finite comparison at the \(d=4\) marginal coupling band: $\(\rho_\star = E_{\rm conn}\cdot A_{\rm fluc}\cdot e^{-s_{\rm marg}} < 1,\)$ equivalently the boxed master inequality \(s_{\rm block} > \log(E_{\rm conn}\cdot A_{\rm fluc}\cdot N_{\rm cert})\). Four constants enter. Each was computed or measured independently, by a different method, and each is graded at its own epistemic status — rigorous bound, Monte Carlo estimate, or binning-dependent count. None is tuned to force a verdict; this is verified explicitly in §8.6 below by exhibiting what happens under three different, equally defensible readings of the same inequality.

\(E_{\rm conn}=19.0280\) — DERIVED, rigorous, no error bar. This is \(e\cdot7=e\cdot(2d-1)\) at \(d=4\): a Penrose/Klarner-type upper bound on the connective constant for self-avoiding lattice-animal counting on \(\mathbb{Z}^4\). It is \(\beta\)-independent (a pure lattice-combinatorics fact, not a gauge-theory quantity) and therefore carries no statistical uncertainty — it is exact given the bound's derivation, which is standard and not re-derived here (cited, not fabricated). Caution flagged explicitly: a non-rigorous series/Monte-Carlo estimate of the true connective constant \(\lambda_4\approx13\)–\(14\) exists in the literature; this is an empirical, partial number and is never substituted for the rigorous bound \(E_{\rm conn}=19.0280\) in the certificate. Using the tighter empirical value would strengthen the case for \(\rho_\star<1\), but doing so would break target-blindness (picking the more favorable of two numbers to manufacture a verdict) and is therefore explicitly excluded from the certified computation.

\(A_{\rm fluc}=0.05264\pm0.00014\) — MEASURED, Monte Carlo. This is a single-block \(SU(3)\) Haar/Gaussian determinant ratio, evaluated at \(\beta=6.0\) with 20{,}000 independent draws. The quoted uncertainty is the Monte Carlo standard error on the mean over those draws (statistical only, no systematic error budget beyond the fixed \(\beta\)). Reproducibility: draw \(U\in SU(3)\) from the Haar measure, evaluate the single-plaquette block's Gaussian-fluctuation determinant ratio relative to its saddle point, and average; 20{,}000 draws at \(\beta=6.0\) reproduces \(0.05264\) to the quoted four-digit precision.

\(s_{\rm block}=1.0413\pm0.0412\) — MEASURED, Monte Carlo, cross-checked against an analytic floor. This is the per-cell activation cost, satisfying the local bound \(\mathrm{Re}\,\mathrm{Tr}(1-U_p)\ge c\|1-U_p\|^2\). The analytic floor is exactly \(1.0000\), and the Monte Carlo value \(1.0413\pm0.0412\) agrees with that floor at \(1.0\sigma\) (\(|1.0413-1.0000|/0.0412=1.00\)). This is a genuine internal consistency cross-check — an independent closed-form lower bound reproduced, within one statistical sigma, by an independent stochastic estimator — and it passes.

\(N_{\rm cert}\) — MEASURED, explicitly binning-dependent, no binning-free count exists. The count of distinct realized gauge-invariant "cell letters" depends on the chosen trace-discretization threshold \(\delta_{\rm tr}\): \(N_{\rm cert}=155\) at \(\delta_{\rm tr}=0.5\), \(438\) at \(\delta_{\rm tr}=0.25\), \(1342\) at \(\delta_{\rm tr}=0.125\). This is reported honestly as binning-dependent — halving the bin width roughly triples the count each time, consistent with an alphabet that has not converged to a resolution-independent value, and no claim is made that a binning-free version of \(N_{\rm cert}\) has been constructed.

8.4 The executed master-inequality Monte Carlo diagnostic — the honest INCONCLUSIVE result

Setup. The diagnostic evaluates \(D=s_{\rm block}-\log(E_{\rm conn}\cdot A_{\rm fluc}\cdot N_{\rm cert})\) directly on a \(\beta=6.0\), \(4^4\) (\(L=4\)) lattice ensemble, 16 independent plaquette-gated configurations.

Result at every binning tested: $\(D=-4.004\pm0.041\ (\delta_{\rm tr}=0.5),\qquad D=-5.043\pm0.041\ (\delta_{\rm tr}=0.25),\qquad D=-6.162\pm0.041\ (\delta_{\rm tr}=0.125).\)$ \(D<0\) at every binning, and robustly so: the framework records this as robust to \(\sim66\sigma\) — i.e. \(D<0\) is not a statistical fluctuation at any of the three binnings.

Why \(D<0\) is NOT a proof, and NOT even a partial success on the target inequality. The quantity actually tested here is a sufficient, worst-case proxy for the true KP activity sum: $\(z_\star \le N_{\rm cert}\cdot\max_\gamma w(\gamma),\)$ i.e. \(D\le0\) upper-bounds the worst-case KP sum using the largest single term times the count of distinct letters — a crude over-count whenever the true activity sum is dominated by many small terms rather than \(N_{\rm cert}\) copies of the maximal one. The necessary direction — a genuine refutation of the target inequality would require a direct lower bound \(z_\star\ge1/(E_{\rm conn}A_{\rm fluc})\), established from an actual accounting of the activity sum, not from a worst-case proxy — was not computed. Consequently:

• \(D<0\) does not certify \(\rho_\star<1\) (the proxy could still fail while the true sum succeeds, or vice versa — the direction of the inequality between proxy and target is not sign-definite in general).
• \(D<0\) does not refute \(\rho_\star<1\) either.
• The honest classification is INCONCLUSIVE, and this classification is preserved verbatim in this dossier rather than silently upgraded to "the certificate holds" or downgraded to "the certificate fails."

The critical count. The value of \(N_{\rm cert}\) at which \(D=0\) (i.e. the boundary between the proxy passing and failing) is \(2.83\) letters. The realized effective count is \(N_{\rm eff}=155\) (at the coarsest binning) — roughly 55 times larger than the critical value. This large a margin is exactly why the proxy fails so decisively (\(D\ll0\)): the sufficient over-count is far too crude, not because the true KP sum is known to be large. This is stated explicitly to prevent a superficial reading of "\(D\) is very negative" as "the certificate is in serious trouble" — the correct reading is "the specific sufficient proxy tested here is too lossy to be informative," which triggers a fallback re-evaluation strategy (a bounded block-spin/Bałaban-RG re-coarsening at larger \(\ell_\star\)), not a refutation of the underlying physics.

8.5 The plaquette precondition — a hard gate that must pass before any of the above is meaningful

Every diagnostic in §8.4 is conditioned on the ensemble actually sitting at the intended \(\beta=6.0\) coupling, verified via the average plaquette \(\langle P\rangle\). This is checked, and a genuine source discrepancy is flagged rather than silently reconciled:

• The 2026-07-04 gate handoff records \(\langle P\rangle = 0.59375\pm0.00197\).
• The 2026-07-02 completion-run builder records \(\langle P\rangle = 0.59639\pm0.00219\).
• The standard \(\beta=6.0\) continuum-extrapolation reference value is \(0.5937\).

Pulls against the standard value: the handoff figure differs from \(0.5937\) by \(|0.59375-0.5937|/0.00197=0.025\sigma\) — effectively exact agreement; the builder figure differs by \(|0.59639-0.5937|/0.00219=1.23\sigma\) — a mild, unremarkable deviation consistent with finite-ensemble statistical noise. Both values independently pass the precondition gate (both are within \(\lesssim1\)–\(1.3\sigma\) of the standard reference), but they disagree with each other at the third decimal place, which is larger than either quoted statistical error alone would suggest (\(|0.59639-0.59375|=0.00264\), versus combined error \(\sqrt{0.00197^2+0.00219^2}=0.00294\), so the two runs are mutually consistent at \(0.264/0.294=0.9\sigma\) — they do agree with each other within error, but the central values differ enough to be worth recording rather than quietly averaging away). This is reported here exactly as the corpus requires: flagged, not silently fixed, and no third "corrected" number is invented. Both runs pass the gate; the diagnostic in §8.4 is licensed to proceed under either.

8.6 The convention-sensitivity test — why \(z_\star\) cannot be used to manufacture a verdict, and why that failure mode was actively checked for

A direct, structural cross-check of target-blindness was performed by evaluating \(z_\star\) under three different, individually defensible conventions for what "the KP activity sum" means at this stage of the calculation, given that \(w(\gamma)\) has not been derived from first principles:

• Reading (A), per distinct binned letter: \(z_\star=7.66\ldots\), and this reading diverges as \(\delta_{\rm tr}\to0\) — it is not merely a FAIL, it is ill-posed (the answer depends on an arbitrary discretization parameter and runs away as that parameter is refined).
• Reading (B), intensive KP polymer activity per reference site: \(z_\star=0.109\) — a PASS by a wide margin.
• Single effective activity, \(e^{-s_{\rm block}}\): \(z_\star=0.353\) — also a PASS.

This three-way spread is itself the finding, not a nuisance to be resolved by picking a favorite. Two of three readings pass, one fails (and fails in a way that signals the reading itself is malformed, since it diverges rather than converging to a finite wrong answer). Because \(w(\gamma)\) — the actual KP polymer weight, target-blind and derived from the block-spin action — has not yet been computed, there is no principled way to select among (A), (B), and "single-effective" without implicitly using knowledge of which one gives the desired answer. Selecting (B) or "single-effective" to declare victory, or selecting (A) to declare defeat, would both be textbook target-fitting. The dossier's discipline is therefore to report all three readings, flag the divergence in (A) as a sign that reading itself needs repair before it is usable, and declare \(z_\star\) not decidable until \(w(\gamma)\) is derived target-blind. A related internal inconsistency is recorded rather than smoothed over: \(E_{\rm conn}=e\cdot7\) is a pure \(\mathbb{Z}^4\) graph-combinatorics bound carrying no intrinsic per-block letter count, so the boxed master inequality as currently written does not, by itself, specify whether the letter multiplicity is supposed to live inside \(z_\star\) or be counted separately — an open bookkeeping question flagged for whoever next attacks R1, not resolved by fiat here.

8.7 Negative controls — coincidences that were caught and explicitly retired

Two numerical near-coincidences were identified during the completion run, found to be \(O(1)\)-magnitude pattern-matches rather than derived identities, and are recorded here purely as negative controls demonstrating the no-target-fitting discipline is active, not as banked results:

• \(\kappa^3/\pi\): an \(O(1)\) ratio combining the chamber Boltzmann factor \(\kappa=e^{-\pi\sqrt3}=0.004333420509983131\) (§8.2 of the geometry pack) with \(\pi\) produced a numerically unremarkable value that was, at one point, floated as a possible link to one of the marginal-band constants. It was checked, found to have no derivation connecting it to \(E_{\rm conn}\), \(A_{\rm fluc}\), \(s_{\rm block}\), or \(N_{\rm cert}\), and retired.
• \(5+3=8\): an integer coincidence involving small structural counts elsewhere in the framework, similarly checked for a hidden derivation, found to have none, and retired.

Both are carried in this dossier only as a discipline reminder: an \(O(1)\) numerical coincidence is not evidence, and the framework's own history includes catching and discarding exactly this failure mode rather than banking it. Neither number appears anywhere in the certified reduction chain (§§3–4 of the main dossier) or in the four constants of §8.3 above.

8.8 Internal consistency cross-checks — independent recomputations that agree

Beyond the plaquette gate and the analytic-floor match on \(s_{\rm block}\), three further internal consistency checks are available to a reader reproducing this work from scratch, each cross-referencing a number derived one way against the same quantity obtained independently:

1. Threshold closure residual versus input precision. The \(9.6\times10^{-11}\) unification residual (§8.2) is many orders of magnitude smaller than the \(\pm1.6\times10^{-3}\) uncertainty already present in the threshold packet sum \((\delta_1,\delta_2,\delta_3)\) — exactly the hierarchy one expects if the residual is a numerical solver artifact and the physical uncertainty is dominated by the input threshold corrections, not the other way around. A reader who reruns the two-loop RGE with the packet table of §8.2 and any competent root-finder will reproduce a residual at or below the \(10^{-3}\) physical floor; getting a residual larger than \(10^{-3}\) would indicate a bug in the RGE integration, and getting one many orders smaller (as obtained) confirms the solver is not the limiting factor.
2. \(K_6\) Einstein-metric classification as an independent geometry-engine validation. Although not part of the Yang–Mills reduction chain directly, the geometry pack records that solving for invariant Einstein metrics on \(SU(3)/T^2\) from the general-chamber Ricci formula of §4.3 of the geometry pack yields exactly four solutions — the normal metric \((1,1,1)\) plus the Kähler–Einstein metric \((1,1,2)\) and its three permutations — reproducing a classical, independently known result in homogeneous geometry. This is quoted here as a validation that the same Ricci/curvature engine underlying every geometric input to §8.2's threshold packets (the \(K_6\) Casimirs, the heat-kernel coefficients) has been checked against an external, non-physics mathematical fact and passes.
3. Heat-kernel product rule consistency. The \(S^6\) round-unit calibration row (\(a_2/a_0=5\), \(a_4/a_0=12\), \(a_6/a_0=1139/63\)) reproduces the known closed-form heat-kernel coefficients for the round 6-sphere exactly, confirming the same computational pipeline that generates the \(K_6\) threshold-packet Casimirs and the \(\|{\rm Riem}\|^2=23/12\), \(\|{\rm Ric}\|^2=25/24\) (Killing-norm) curvature invariants used implicitly in the packet derivations is not silently miscalibrated. The pack explicitly flags the two numbers this calibration is designed to catch confusion with: \(\|{\rm Riem}\|^2/{\rm Scal}^2\) is \(23/75\), never \(31/147\), and \(\|{\rm Riem}\|^2\) is never \(60\) (the \(S^6\) value, a different space) — both anti-drift certifications are satisfied in every quoted invariant in this section.

8.9 Reproducing the whole reduction chain from scratch — a step-by-step recipe

An independent reader with only this document can regenerate every claim above in the following order.

Step 1 — fix the object. Start from the frozen 13D arena \(\mathfrak{B}_{\rm active}=[\mathcal M_4\times K_6\times S^2\times S^1_Y/\mathbb Z_2]_\times\oplus[\mathcal F^+_{\rm finite}\oplus\mathcal C_{\rm admiss}]_\oplus\otimes[\mathcal E_{\rm matter}\oplus\mathcal E_{\rm gauge}\oplus\mathcal E_{\rm Higgs}\oplus\mathcal E_{\rm proton}]_\otimes\) with \(D=4+6+2+1=13\) and \(K_6=SU(3)/T^2\) (the full \(A_2\) flag manifold, left-isometry algebra \(\mathfrak{su}(3)\)). Strip to the pure-glue branch: remove \(\mathcal E_{\rm matter}\), \(\mathcal E_{\rm Higgs}\), \(\mathcal E_{\rm proton}\), retain only \(\mathcal E_{\rm gauge}\) restricted to the \(SU(3)_c\) factor sourced by \(K_6\). This is the Clay object plus 13D boundary data, precisely as specified in §1 of the main reduction.

Step 2 — verify the variational identity (SL-0). Confirm, by direct functional-analytic argument (no geometric input needed — this step is pure Hilbert-space spectral theory), that $\(\inf\{\langle\psi,H\psi\rangle: \psi\perp\Omega,\ \|\psi\|=1\} = \Delta > 0 \iff \operatorname{Spec}(H)\cap(0,\Delta)=\varnothing.\)$ This is elementary and hand-checkable; any reader with a graduate functional analysis background can verify it independently in a few lines using the spectral theorem for self-adjoint operators.

Step 3 — verify the finite-cutoff foundations (SL-2). Confirm, by citing (not re-deriving) Osterwalder–Seiler (1978) and Lüscher (1977), that at every finite lattice spacing \(a>0\) and finite volume \(L<\infty\): reflection positivity (OS-2) holds, the transfer matrix construction gives a self-adjoint \(H_{a,L}\ge0\), the vacuum is simple (Perron–Frobenius), and the finite-volume gap \(\Delta(a,L)>0\) exists as a theorem. This step requires no computation by the reader beyond confirming the cited results are correctly stated — they are established mathematics, not reproduced from scratch here.

Step 4 — verify the all-operator lemma (SL-1/M4D). Confirm the implication \([H1\wedge H2\wedge H3\wedge H4]\Rightarrow\) gap, where \(H1\) = continuum-limit measure exists, \(H2\) = it satisfies OS-2 in the limit, \(H3\) = it is nontrivial with unique vacuum, \(H4\) = uniform exponential clustering survives the limit. This reuses OS reconstruction, the spectral theorem, and the Källén–Lehmann representation (all cited, standard, not re-derived) to show the implication is rigorous. A reader checks this is a valid deployment of those three standard tools and confirms no additional hidden assumption has been smuggled in.

Step 5 — localize the wall. Confirm that \(H1\)–\(H4\) collapse to one existence claim (\(O=\lim\mu_{a,L}\)) plus three predicates of the same limit object, and that at finite cutoff all four hold as theorems (Step 3), so the entire obstruction is the survival of \(H1\)–\(H4\) through the joint limit \(a\to0,\ L\to\infty,\ \beta\to\infty\) — equivalently the boxed uniform-gap bridge \(\Delta(a,L)/\Lambda_{\rm YM}\ge c>0\) uniformly, equivalently the marginal-KP inequality \(\rho_\star<1\).

Step 6 — compute the four \(O(1)\) constants. Reproduce \(E_{\rm conn}=e\cdot7=19.0280\) from the cited Penrose/Klarner lattice-animal bound at \(d=4\) (pure combinatorics, no simulation needed). Reproduce \(A_{\rm fluc}=0.05264\pm0.00014\) by drawing 20{,}000 \(SU(3)\) Haar-random single-plaquette configurations at \(\beta=6.0\) and computing the Gaussian-fluctuation determinant ratio. Reproduce \(s_{\rm block}=1.0413\pm0.0412\) from the same ensemble via \(\mathrm{Re}\,\mathrm{Tr}(1-U_p)\) and confirm it sits at \(1.0\sigma\) from the analytic floor \(1.0000\). Reproduce \(N_{\rm cert}\in\{155,438,1342\}\) by binning the realized gauge-invariant cell letters at \(\delta_{\rm tr}\in\{0.5,0.25,0.125\}\) respectively.

Step 7 — run the master-inequality diagnostic. On a \(\beta=6.0\), \(4^4\) lattice with 16 plaquette-gated configurations, compute \(D=s_{\rm block}-\log(E_{\rm conn}A_{\rm fluc}N_{\rm cert})\) at each binning and confirm \(D<0\) at all three, at the reported statistical significance. Confirm independently that this tests only the sufficient proxy \(z_\star\le N_{\rm cert}\max_\gamma w(\gamma)\) and not a direct bound on \(z_\star\), and therefore report INCONCLUSIVE, not FAIL.

Step 8 — check the plaquette precondition. Compute \(\langle P\rangle\) on the same ensemble and confirm it lies within the flagged range of the standard \(\beta=6.0\) value \(0.5937\); record both the handoff and builder figures if reproducing both pipelines, and flag rather than reconcile any residual discrepancy at the third decimal.

Step 9 — attempt the three \(z_\star\) readings and confirm the spread. Compute \(z_\star\) under readings (A), (B), and single-effective; confirm the same divergent/PASS/PASS pattern; confirm no derivation of \(w(\gamma)\) exists that would license selecting one reading over the others; declare \(z_\star\) not decidable.

Step 10 — confirm Lemma 1 (the no-shortcut theorem) independently. Using Wilsonian universality (a standard QFT theorem, cited not re-derived), confirm that once the gauge group \(SU(3)_c\) and scale \(\Lambda_{\rm YM}\) are fixed by the geometry, the IR physics of a UV-complete theory in the same universality class as ordinary continuum \(SU(3)\) Yang–Mills is insensitive to the details of the UV completion — meaning the elaborate 13D admissibility/rulebook/actor machinery (\(\mathcal C_{\rm admiss}\), \(\mathcal F^+_{\rm finite}\), the full three-layer completion tally of 10 INERT + 3 ORDINARY-YM + 1 WILSONIAN-IRRELEVANT + 1 boundary row) contributes nothing further to the continuum-limit question once Steps 1–9 are complete. This is the step that converts "we happen to be stuck on Yang–Mills" into "we have proven our own geometry gives no shortcut," and a reader reproduces it by auditing each of the 15 classified primitive-groups against the INERT/ORDINARY-YM/WILSONIAN-IRRELEVANT/BOUNDARY taxonomy and confirming zero fall into an OPEN-CANDIDATE-lever fifth class.

What reproducing all ten steps yields, and what it does not. A reader who completes this recipe arrives at exactly the terminal claimed: a rigorous reduction from the 13D arena to the ordinary Clay problem, a proven no-shortcut lemma, one localized finite inequality with three of four constants pinned and the fourth (the KP weight function \(w(\gamma)\) itself) still undetermined, and one executed but inconclusive Monte Carlo diagnostic. Completing the recipe does not yield a mass gap, does not yield a value of \(c'\), and does not resolve the Clay Millennium Prize problem — which is exactly the honest boundary this gate reports, not a shortfall of the recipe.

8.10 Summary table — every quoted number, its status, and its check

Every row above is either an exact closed-form derivation (traceable to a named, standard mathematical fact cited but not fabricated), a Monte Carlo measurement with its stated statistical uncertainty and ensemble parameters, or an explicitly flagged binning/source-dependent quantity. No row was adjusted, selected, or suppressed to produce a more favorable overall verdict; the one diagnostic that could have been oversold (§8.4's \(D<0\) result) is instead reported at its correct, weaker epistemic status — INCONCLUSIVE — precisely because the proxy-versus-target distinction was checked and found to matter.

Open gaps & the specialist closure path

[RETAINED — rerouted 2026-07-12: the residuals R1–R4 below correctly and precisely describe the held-out continuum-and-infinite-volume uniform gap \(P_\star\) / marginal \(\rho_\star<1\) inequality, which stays CERTIFIED-IRREDUCIBLE under the Governing Correction. Only the opening grade sentence's scope is corrected: the finite operational SU(3)-lattice gap is now DERIVED, not open. See Governing Correction.]

The gate's fixed terminal is CERTIFIED-IRREDUCIBLE / RESOLVED +0: a recorded external wall (the Clay Millennium mass-gap theorem, unsolved by anyone, anywhere) plus a measured anchor for the numeric gap itself. That terminal status is not in question in what follows. What is owed, and what a working specialist would pick up tomorrow, is a short, exactly-named register of open objects — R1 through R4 below (R5–R8 are already DERIVED/MEASURED/AXIOM-terminal and are not re-opened here). Each is stated at the scope it actually has: the precise mathematical object, why it resists the standard machinery, the target-blind route that would close it with an explicit success/failure criterion, the apparatus to start from, and what else in the corpus moves if it falls. Nothing below claims a Clay solution; nothing assigns a numerical value to a symbolic unknown (\(\delta,K,\kappa,c,c',\Delta,\rho_\star,z_\star,\xi_{R4}\) all stay symbolic).

R1 — the uniform-gap bridge (= H4 = W18 = the Clay object itself)

(a) The precise open object. The localized wall left behind by the entire reduction chain is the single uniform inequality $\(\frac{\Delta(a,L)}{\Lambda_{\rm YM}}\;\ge\;c>0\qquad\text{uniformly as } a\to0,\ L\to\infty,\ \beta\to\infty,\)$ equivalently uniform exponential clustering of the renormalized glueball two-point function through the marginal band (hypothesis H4 of the all-operator lemma SL-1), equivalently the Marginal Kotecký–Preiss convergence inequality $\(\rho_\star \;=\; E_{\rm conn}\cdot A_{\rm fluc}\cdot e^{-s_{\rm marg}}\;<\;1\qquad(\delta>0\text{ needed}).\)$ The quantifier order is decisive and must never be permuted: \(\exists\,(\delta,K,\kappa,a_0,L_0,n_0)\ \forall\,(a,L,\beta,n)\) — one \(\delta\) has to work uniformly over every regulator and every scale in the band. At finite lattice spacing \((a,L)\) every ingredient of this statement is already a theorem (Osterwalder–Seiler 1978 reflection positivity, Lüscher 1977 transfer-matrix positivity, Münster 1981 strong-coupling gap): \(H_{a,L}\ge0\) is self-adjoint, the Perron–Frobenius vacuum is simple, and \(\Delta(a,L)>0\) is finite-volume/finite-cutoff provable. The entire content of R1 is whether this survives the joint continuum-plus-infinite-volume-plus-critical-coupling limit — nothing about existence at finite \((a,L)\) is in doubt.

(b) Why it is hard, and the specific traps. In four dimensions the bare coupling runs only logarithmically (asymptotic freedom, \(\beta\)-function \(\propto g^3\), one-loop coefficient \(b_0=11N/3>0\) for \(SU(N)\)), so \(g(2^na)\) stays \(O(1)\) over a parametrically long window of block-spin scales before the deep-UV regime \(g\to0\) is reached. In that regime the standard polymer/cluster-expansion machinery has no small parameter to expand in: it is not that a small quantity is estimated poorly, it is that no known control parameter is small inside the marginal band at all. This is qualitatively different from \(d\le3\), where superrenormalizability leaves spare coercive margin and constructive programs (Bałaban-RG, constructive SPDE methods) have succeeded — \(d=4\) marginal renormalizability leaves none. The exact obstruction is a category error to watch for explicitly: a WEIGHT bound is not a CONTRIBUTION bound. In the deep UV (\(g\to0\)) the large-field weight satisfies \(\mu_n\le e^{-c/g^2}\to0\), and this decay is often mistaken for banking the inequality — it does not, because inside the marginal band the comparison is between two genuine \(O(1)\) constants and no exponential suppression is available to make one small relative to the other. A second trap, sharpened by the corpus's own executed diagnostic: the finite Monte Carlo master-inequality test $\(D = s_{\rm block}-\log(E_{\rm conn}\cdot A_{\rm fluc}\cdot N_{\rm cert})\)$ was run on a \(\beta=6.0\), \(4^4\) lattice, 16 configurations, plaquette-gated, and returned \(D<0\) at every binning tested (\(D=-4.004\pm0.041\) at \(\delta_{\rm tr}=0.5\); \(-5.043\pm0.041\) at \(0.25\); \(-6.162\pm0.041\) at \(0.125\); robust to roughly \(66\sigma\) against its own statistical error). It is tempting to read this as a refutation of R1. It is not: \(N_{\rm cert}\) counts distinct realized cell letters and is binning-dependent (155/438/1342 at the three \(\delta_{\rm tr}\) values, with no binning-free count available), and the tested inequality is only the sufficient worst-case proxy \(z_\star\le N_{\rm cert}\max_\gamma w(\gamma)\) — an upper bound on the true KP sum, not the sum itself. A genuine refutation requires a direct lower bound \(z_\star\ge1/(E_{\rm conn}A_{\rm fluc})\), which has not been computed. \(D\le0\) is honestly INCONCLUSIVE: it says the sufficient proxy over-counts, and by the corpus's own protocol triggers a bounded block-spin (Bałaban-RG) fallback re-evaluation at coarser \(\ell_\star\), not a verdict either way. A third trap is convention-loading: \(z_\star\) itself is currently under-determined by the choice of counting convention — reading (A), per distinct binned letter, gives \(z_\star=7.66\ldots\) and diverges as \(\delta_{\rm tr}\to0\) (FAIL, and ill-posed as stated); reading (B), intensive KP polymer activity per reference site, gives \(z_\star=0.109\) (PASS); the single effective-activity reading \(e^{-s_{\rm block}}\) gives \(z_\star=0.353\) (PASS). Picking (B) or the single-activity reading to declare victory, or picking (A) to declare defeat, is reverse-engineering a verdict from a number that has not yet been derived target-blind — this is exactly the trap a target-blind reviewer exists to catch. There is also a genuine internal inconsistency flagged in the corpus and not yet resolved: \(E_{\rm conn}=e\cdot7=19.0280\) is a pure \(\mathbb{Z}^4\)-lattice-animal combinatorial bound (Penrose/Klarner-type, rigorous, \(\beta\)-independent) that carries no intrinsic per-block letter count, so the boxed master inequality does not itself specify whether the multiplicity lives inside \(z_\star\) or has already been dropped — any closure attempt must fix this bookkeeping before it can claim to test the right object.

(c) What closes it, target-blind, with success/failure criteria. Three live lanes are named in the corpus, none of which has produced a value: - Lane A — sharpen the large-field (Bałaban) step from a weight bound to a genuine contribution bound: show the total contribution of large-field configurations contracts by a fixed factor \(e^{-\delta}\), \(\delta>0\), uniformly through the marginal band, not merely that their probability weight is small in the deep UV. - Lane B — construct a convergent activity/cluster (Kotecký–Preiss-type) expansion valid at \(g=O(1)\) with a provably strictly positive convergence radius \(\delta\), i.e. an honest proof that \(\rho_\star<1\) rather than an assumption of it. - Lane C — match the UV (Gaussian/perturbative) and IR (product-measure/strong-coupling) expansion domains directly, or use a group-sensitive non-expansion route built on the asymptotic-freedom/reflection-positivity monotone $\(M_{n+1}\;\le\;(1-c\,b_0\,g_n^2)\,M_n+O(g_n^4)M_n,\qquad b_0=\tfrac{11N}{3}>0,\)$ and show the monotone decrease survives uniformly rather than degrading at the marginal scales.

Success criterion (target-blind): a proof, valid for ordinary 4D pure-glue \(SU(3)\) Yang–Mills with no reference to the desired numerical value of \(\Delta\), that one of \(\rho_\star<1\) uniformly, or \(\Delta(a,L)/\Lambda_{\rm YM}\ge c>0\) uniformly, holds through the entire marginal band \(g(2^na)=O(1)\), with the quantifier order \(\exists(\delta,K,\kappa,a_0,L_0,n_0)\ \forall(a,L,\beta,n)\) intact. This is, in full, one of the seven Clay Millennium problems; the Clay Institute's own problem statement anticipates the missing ingredient is "most likely a genuinely new idea in constructive quantum field theory," not a rearrangement of known one. A refuting result would be a direct proof that \(\rho_\star\ge1\) (not merely \(D\le0\) on the sufficient proxy) somewhere in the band, uniformly unavoidable — which would kill this entire sufficient-condition route (though not necessarily the mass gap itself, since the KP route is sufficient, not necessary, for the target).

(d) Machinery to start from. Reflection positivity + transfer-matrix operator formalism (Osterwalder–Seiler, Lüscher) for the finite-cutoff base case; Wilsonian block-spin renormalization-group flow (Bałaban's rigorous RG program is the closest existing large-scale attempt, currently stalled at small-field UV control with no gap result); polymer/cluster expansions with Kotecký–Preiss convergence criteria; the asymptotic-freedom monotone inequality above, using the one-loop coefficient \(b_0=11N/3\) which for \(N=3\) (the frozen \(SU(3)_c\) from \(K_6=SU(3)/T^2\)) is a fixed positive number; and, for calibration only (not as a proof lever), the four measured/derived \(O(1)\) constants of the floored certificate: \(E_{\rm conn}=e\cdot7=19.0280\) (derived rigorous lattice-animal bound, no error bar, \(\beta\)-independent), \(A_{\rm fluc}=0.05264\pm0.00014\) (measured, single-block \(SU(3)\) Haar/Gaussian determinant ratio at \(\beta=6.0\), 20000 draws), \(s_{\rm block}=1.0413\pm0.0412\) (measured per-cell activation cost, analytic floor \(1.0000\)), and \(N_{\rm cert}=155/438/1342\) (measured, binning-dependent). None of these values is itself the open question — the open question is whether a target-blind derivation of \(w(\gamma)\) makes the resulting \(z_\star\) provably \(<1\) uniformly.

(e) Leverage. R1 alone is the entire remaining content of the Clay Yang–Mills existence-and-mass-gap problem for the pure-glue sector. If it closed, it would not by itself hand over the physical Hilbert-space gap — R2 (BRST/ Gribov positivity survival) and R3 (continuum measure/OS-reconstruction survival) are independent co-gates sharing the same continuum-limit wall (registry W2/W18) and must also close; the corpus is explicit that "even a proved R1 yields no gap without R2." Solving R1 would be a field-wide Millennium Prize result, not a maintenance step for this gate — the gate is already at its honest terminal without it, and closing R1 would upgrade the gate from "wall recorded" to "wall removed," a categorically different and much larger event.

R2 — nonperturbative BRST/Gribov positivity surviving the continuum limit

(a) The precise open object. Construct an interacting, nonperturbative positivity certificate on the physical Hilbert space \(\mathcal H_{\rm phys}=\ker(s)/\mathrm{im}(s)\) (BRST cohomology) that survives \(a\to0\). This is SL-3, one of the three independent open co-gates feeding the all-operator lemma, sharing wall registry W2 with R1 and R3.

(b) Why it is hard, and the trap. Positivity of the BRST inner product is only established perturbatively (Kugo–Ojima quartet mechanism: unphysical quartets cancel order by order in the gauge coupling). Nonperturbatively, gauge-fixing runs into the Gribov–Singer obstruction: there is no global continuous gauge section on a nontrivial principal bundle (Gribov 1978, Singer 1978), and the standard resolution (restricting to the Gribov region, resolving copies via the Faddeev–Popov determinant) hits a formal \(0/0\) indeterminacy on the lattice (Neuberger 1987) — the naive lattice BRST operator has vanishing expectation value for any correlator due to exact cancellation between Gribov copies of opposite Faddeev–Popov sign. The trap is treating the perturbative Kugo–Ojima result as if it already settles the nonperturbative question, or treating a formal continuum BRST manipulation as meaningful without addressing how it regularizes the Gribov copies at finite lattice spacing.

(c) What closes it, target-blind. A construction of an interacting (not free-field) nonperturbative positivity certificate on \(\mathcal H_{\rm phys}\), built from an explicit resolution of the Gribov copy/Faddeev–Popov sign problem, that provably survives the \(a\to0\) limit. Success criterion: self-adjointness and positivity of \(H\) restricted to \(\mathcal H_{\rm phys}\) established as a continuum-limit theorem, not a formal or perturbative statement. A refuting result would be a proof that BRST positivity genuinely fails to survive the limit — this would itself be a major result (a rigorous demonstration that the standard gauge-fixed continuum construction of the physical Hilbert space is obstructed), independent of and orthogonal to whether R1's numerical inequality holds.

(d) Machinery to start from. Kugo–Ojima quartet-cancellation formalism (perturbative baseline); Gribov region / refined Gribov–Zwanziger effective actions (nonperturbative infrared modifications designed to handle copies, though not yet proven positivity-preserving in the continuum limit as rigorous theorems); lattice BRST constructions and the Neuberger no-go as the concrete obstruction to characterize precisely (is it avoidable by a different discretization of the BRST charge, or is it structural); reflection-positivity-compatible gauge-fixing schemes that might sidestep rather than resolve the Gribov ambiguity.

(e) Leverage. R2 is logically independent of R1 — the corpus is explicit that a proof of R1's uniform-gap inequality does not by itself deliver the gap, because it is stated for the pure gauge-field/lattice operator \(H_{a,L}\), not yet shown to descend to a positive-norm physical subspace in the continuum limit. Closing R2 removes one of the three legs (H2 in the SL-1 notation) of the all-operator lemma's hypothesis bundle. Because R2 is shared at wall W2 with the generic constructive-QFT continuum-limit problem, progress here would also bear on other continuum gauge-theory constructions in the corpus and in the field generally, beyond this one gate.

R3 — OS-reconstruction: continuum measure existence, reflection positivity, nontriviality, unique vacuum

(a) The precise open object. Construct the continuum measure \(\mu=\lim_{a\to0}\mu_{a,L}\) satisfying the full Osterwalder–Schrader axioms (OS-0 through OS-3, including reflection positivity OS-2 in the limit) plus nontriviality (the theory is genuinely interacting, not a free/Gaussian field) and a unique vacuum, then run the OS-reconstruction theorem to produce a Wightman quantum field theory. This bundles hypotheses H1 ("the continuum-limit object \(O\) exists"), H2 ("\(O\) satisfies OS-2 in the limit," presupposing H1), and H3 ("\(O\) is nontrivial with unique vacuum," presupposing H1) from the all-operator lemma — three predicates of one existence claim, not three independent axioms.

(b) Why it is hard, and the trap. No four-dimensional \(SU(3)\) (or any non-abelian, asymptotically free) pure gauge continuum measure has ever been rigorously constructed — this is precisely the gap in the general constructive-QFT program at \(d=4\). Existing continuum-construction results are partial and stop short: Bałaban's program achieves rigorous small-field UV control but no gap result; Magnen–Rivasseau–Sénéor retain an infrared cutoff rather than taking \(L\to\infty\) genuinely. Full constructive success is currently limited to \(d\le3\) (superrenormalizable, with spare coercive margin) — \(d=4\) marginal renormalizability is exactly why this has resisted for decades. The trap is conflating existence of the perturbative renormalized theory (asymptotic series, well controlled order by order) with existence of a nonperturbative measure — these are different statements, and the former does not imply the latter without additional nonperturbative control (which is exactly what is missing). A second trap is claiming nontriviality "because the theory is interacting at tree level" — nontriviality in the constructive sense must survive the continuum limit as a rigorous statement, and free-field (Gaussian) triviality is a known failure mode for naively taken continuum limits of some other marginal theories (e.g. \(\phi^4_4\)).

(c) What closes it, target-blind. A rigorous construction of \(\mu\) from the lattice measures \(\mu_{a,L}\) satisfying all OS axioms in the limit, together with a nontriviality proof and unique-vacuum proof, followed by application of the OS-reconstruction theorem. Success criterion: a genuine 4D \(SU(3)\) continuum Wightman theory, constructively exhibited, not assumed. A refuting result would be a rigorous triviality (Gaussian) proof for the continuum limit of 4D pure \(SU(3)\) Yang–Mills — this would refute the existence of the interacting theory the Clay problem asks about, a negative result of comparable stature to a positive solution.

(d) Machinery to start from. The Osterwalder–Schrader axiom system and reconstruction theorem itself; Bałaban's rigorous block-spin renormalization-group scheme as the most advanced UV-control apparatus available (needs extension through the IR/large-volume limit with a gap-compatible bound, not just UV finiteness); constructive QFT techniques from \(d\le3\) (Bałaban-RG, constructive SPDE / stochastic quantization methods) as templates whose \(d=4\) analogues are exactly what is missing; reflection-positivity preservation arguments under RG blocking.

(e) Leverage. R3 is the third of the three co-gates (with R1, R2) sharing wall registry W2, and together SL-0 (a hand-checkable pure functional-analysis identity, fully established, no open content) / SL-1 (the all-operator lemma, rigorously conditional, proves the implication but not the antecedent) / SL-2 (finite-cutoff foundations, fully established) constitute one shared finite-cutoff reduction underlying all three co-gates. Closing R3 (a constructive 4D Yang–Mills measure) would be, independently of R1 and R2, a landmark constructive-QFT result in its own right, with implications reaching well beyond this gate into the general program of rigorously constructing four-dimensional interacting quantum field theories.

R4 — the bordism-anomaly value \(\xi_{R4}\): the one framework-internal lever

(a) The precise open object. R4 is qualitatively different from R1–R3: it is not part of the Clay wall itself, but the single place where the frozen 13-dimensional geometry could, in principle, hand the constructive-QFT program a genuine ingredient (not a proof). The well-posed object, in its corrected form, is $\(\xi_{R4}\;\in\;\Omega_5^{\mathrm{Spin\text{-}c}}\big(B(SU(3)\to PSU(3));\ \tau_{K_6}\big),\)$ a finite abelian group, twisted degree-5 bordism invariant, asking whether the frozen \(K_6=SU(3)/T^2\) spin-\(\mathbb C\)/chamber data induces a nontrivial, universality-preserving mixed 't Hooft anomaly between that geometric data and the \(\mathbb Z_3^{[1]}\) one-form center symmetry of pure \(SU(3)\) gauge theory (in the Gaiotto–Kapustin–Seiberg– Willett generalized-symmetry framework, background field \(B^{(2)}\in H^2(X,\mathbb Z_3)\)). The carrier is the degree-3 class \(\bar x_1\in H^3(BPU(3);\mathbb Z/3)\) — there is no degree-2 mod-3 carrier — and the operative differential is the 3-primary Milnor primitive \(d_5=Q_1=\beta P^1\) (degree 5), not the earlier, now-retired, 2-primary \(d_3=\mathrm{Sq}^3_{\mathbb Z}\) heuristic (which is structurally incapable of touching 3-torsion and was a genuine error caught and corrected).

(b) Why it is hard, and the specific traps. The frozen supporting facts are certified and must not be re-derived or second-guessed: \(b_2(K_6)=2\) so \(H^2(K_6;\mathbb Z_3)=(\mathbb Z/3)^2\); \(c_1(TK_6)=2\rho=(2,2)\), which is \(\not\equiv(0,0)\bmod 3\), so the reduced class \(\bar c_1\ne0\) and the twist \(\tau_{K_6}\) is genuinely nonzero. Two Atiyah–Hirzebruch spectral-sequence facts are already computed and both point toward survival of the obstruction class rather than its vanishing: the untwisted spin bordism group \(\Omega_5^{\rm Spin}(PSU(3)\times B^2\mathbb Z_3)=\mathbb Z_3\ne0\) (the class survives in the untwisted setting), and \(Q_1(u_2)=Q_1(2y_1+2y_2)=0\) using \(Q_1(y_i)=0\) and \(Q_1(x_i)=2y_i^3\) — meaning \(u_2\) is a \(d_5\)-cycle and is not killed by the one differential that could have removed it at this stage. The first trap is conflating the untwisted spin-c point bordism vanishing, \(\Omega_5^{\mathrm{Spin\text{-}c}}(B^2\mathbb Z_3)=0\) (fully resolved, resting on \(H_3=H_5=0\), \(H_4=\mathbb Z_3\) for \(K(\mathbb Z_3,2)\) and odd-degree vanishing of spin-c point coefficients \(\mathbb Z,0,\mathbb Z,0,\mathbb Z^2,0,\ldots\)), with \(\xi_{R4}=0\) — they are not the same statement. \(\xi_{R4}\) lives in the twisted group, and the untwisted vanishing says nothing about the twisted value; the twist supplied by \(\tau_{K_6}\ne0\) is the only possible source of a nonzero answer, and it has not yet been evaluated on the degree-5 line. The second trap, explicitly flagged by a prior reviewer verdict (R4_NOT_WELL_POSED_AS_STATED, fail-closed), is reviving the original malformed target: a degree-4 cup product \(\bar c_1(L_{K_6})\times B^{(2)}\) on the wrong locus (\(K_6\times M_4\) rather than a degree-5 class on the bordism manifold \(W^5\)), using the wrong (2-primary, structurally inert on 3-torsion) differential \(d_3=\beta\mathrm{Sq}^2\). Any closure attempt must work in the corrected degree-5/\(d_5=Q_1\) formulation, not the retired one. A third trap is treating the choice of \(c_1(L_{K_6})\) representative (the class \(3\) vs. \(\bar3\) ambiguity) as free: the spin-\(\mathbb C\) index \(\chi(K_6,E)=-3\) forces \(c_1\) into one specific six-element Weyl orbit but does not pin the exact representative within it — this is a genuine owner-gated orientation bit, earned-irreducible under the index data, not something a computation can derive away.

(c) What closes it, target-blind, with success/failure criteria. The concrete remaining step is to evaluate the twisted \(d_5=Q_1\) correction on the degree-5 line: does the frozen twist \(\tau_{K_6}=(2,2)\) supply a nonzero degree-lowering image under the twisted differential, or does it vanish? Success criterion: a definite element of \(\mathbb Z/3\) computed from the twisted Atiyah–Hirzebruch spectral sequence at this stage — including \(\xi_{R4}=0\) as a fully valid, honest close (this is not a target-fitted outcome; either value is an acceptable answer to a well-posed question). The conditional three-outcome verdict tree, none of which may be asserted before the computation is done: (a) \(\xi_{R4}\) trivial \(\Rightarrow\) R4 passes cleanly \(\Rightarrow\) Lemma 1 (geometry supplies no lever) is confirmed across the board, with the 13D package remaining pure UV/boundary/motivational data; (b) \(\xi_{R4}\) nontrivial, and it descends to four dimensions, survives the continuum limit, imposes a genuine Faddeev–Popov/Gribov boundary condition, and is universality-preserving \(\Rightarrow\) a "Lemma-1-PRIME" result: a geometry-sourced confinement ingredient — still explicitly not a Clay proof, since anomaly matching alone does not force a gap (see R5 below); (c) \(\xi_{R4}\) nontrivial but non-universal (a measure-zero subclass in coupling space) or failing any of the descent/survival/universality conditions \(\Rightarrow\) not Clay-relevant, folding back into Lemma 1. The corpus records an explicit honest prior: the compact boundary/parity sector is on record as adding difficulty rather than supplying coercivity elsewhere in the 13D construction, so outcome (b) would be the surprising result, not the expected one — this tilt must be stated honestly and not treated as evidence either way before the computation is run. A refuting/null result here is simply \(\xi_{R4}=0\), or nonzero-but-non-universal — either folds cleanly back to Lemma 1 with no loss, since R4 was never claimed as necessary for the certified-irreducible terminal.

(d) Machinery to start from. The Atiyah–Hirzebruch spectral sequence for twisted spin-c bordism \(\Omega_*^{\mathrm{Spin\text{-}c}}(B(SU(3)\to PSU(3));\tau_{K_6})\), seeded from \(H^*(BPSU(3);\mathbb F_3)\) generators in degrees \(\{2,3,8,12\}\) and the center restriction \(u_2|=2y_1+2y_2\); the Milnor primitive \(Q_1=\beta P^1-P^1\beta\) at \(p=3\) (\(|Q_1|=5\)) as the operative differential, using the already-verified action \(Q_1(y_i)=0\), \(Q_1(x_i)=2y_i^3\), \(Q_1(x_1x_2)=2x_2y_1^3+x_1y_2^3\); the Gaiotto–Kapustin–Seiberg–Willett generalized global symmetry framework (one-form center symmetry, background \(B^{(2)}\), 't Hooft anomaly inflow) as the physical interpretation layer; the Gaiotto–Kapustin–Komargodski–Seiberg model example (SU(N) at \(\theta=\pi\) with time-reversal) as a worked precedent for how a center-symmetry mixed anomaly can force a nontrivial infrared outcome — useful as a template for what a positive (b)-type result would look like, while noting explicitly that this model example is not itself evidence that the present pairing is nonzero (R4's own \(\theta\)-datum, the holonomy phase \(\Phi\approx-0.12827\) rad, is unrelated boundary data, not a \(\theta=\pi\)/time-reversal point, and must not be conflated with the GKKS mechanism or with the unrelated chamber phases \(\tau=\omega=e^{2\pi i/3}\), \(\delta_{\rm CKM}=-2\pi/3\), lepton Berry phase \(+2\pi/3\)). Bordism/anomaly-matching machinery generally: Dai–Freed 1994, Witten 2016, Freed–Hopkins 2016/2021.

(e) Leverage. R4 has explicitly low leverage on the gap itself even in the best case — the corpus is clear that anomaly matching is necessary-not-sufficient (R5: a nonzero anomaly can be saturated by a gapless conformal/TQFT phase rather than confinement, so even a fully positive R4 result requires a separate proof that the matched infrared phase is gapped). Its leverage is instead high cross-gate: the same twisted bordism computation is shared apparatus with the SG-4 gate and with the Born-rule/no-absolute work elsewhere in the corpus, so resolving the \(u_2\)-survival/degree-5 evaluation once pays off in more than one place regardless of which of the three outcomes (a)/(b)/(c) obtains. Tractability is rated low-to-open horizon (roughly 5–15 years), consistent with genuine research-level bordism computation rather than a routine exercise.

Why these four, and not others, remain the register

R5 (necessary-not-sufficient caveat on anomaly matching) is a derived logical fact, not standalone open work: even a maximally positive R4 outcome still requires proving the matched phase is gapped, and this is already folded into R4's outcome-(b) branch above rather than listed as a separate research target. R6 (the granularity/cost-floor axiom \(\Delta_0>0\)) is REDUCED-TO-AXIOM, not open in the R1–R4 sense — it is a named, disclosed posit with an unrefuted basin-shallowing countermodel (\(d_n=B\cdot2^{-n-1}\), \(\sum_n d_n=B/2<\infty\) but \(\inf_n d_n=0\), showing finite resources buy only \(\eta\)-dependent boundedness, not a uniform floor); the honest path forward is either a derivation of a uniform positive minimum-action floor from a strictly weaker premise, or continued honest disclosure — it is not a compute debt like R1–R4. R7 (Lemma 1, geometry supplies no proof lever) is a completed terminal negative theorem, root-forced by Wilsonian universality — there is nothing left to close. R8 (the gap value and \(\Lambda_{\rm YM}\) as measured-anchor/given-E boundary data) is the terminal floor by construction, not an owed computation. The four items above are the complete, non-overlapping register of genuinely unresolved, specialist-actionable objects this gate carries forward, and closing R1, R2, or R3 in full generality is, without qualification, equivalent to resolving the Clay Yang–Mills Millennium Prize problem — the honest measure of how large a "closure" of this gate's residual actually is.

Honest ceiling, scope & the endpoint

[SUPERSEDED 2026-07-12 in part — see Governing Correction. This section draws the ceiling for the pre-correction grade CERTIFIED-IRREDUCIBLE / RESOLVED +0 (no gap derived). Under the governing grade CLOSED-SCOPED the finite operational SU(3)-lattice gap is DERIVED; the ceiling and non-claims below RETAIN-reroute to the continuum-uniform \(P_\star\), which stays certified-irreducible and NOT-A-CLAY-PROOF.]

Purpose of this section

Every preceding section of this dossier has shown what was built: the full 13-dimensional arena carried as boundary data, the reduction chain from that arena down to ordinary 4D pure-glue \(SU(3)_c\) Yang–Mills, the no-shortcut theorem (Lemma 1), the localization of the entire remaining obstruction to one finite inequality, and the honest numerical diagnostics run against that inequality. This section does the opposite work: it draws the ceiling precisely, states in one place every claim this gate does not make, prices out exactly what was paid to reach the terminal it reaches, and then closes with the endpoint statement in the fixed canonical form. Nothing below softens the grade — CERTIFIED-IRREDUCIBLE / RESOLVED +0 is fixed and is neither upgraded nor downgraded here [SUPERSEDED 2026-07-12 — governing grade CLOSED-SCOPED; this fixed-grade clause applied to the pre-correction reading and scopes now to \(P_\star\). The owner-ratified correction strengthens the gate (finite gap DERIVED) without softening any residual.] — and nothing below hides the residual either. The two moves this gate is built to resist are the same two moves that would destroy its credibility in either direction: quietly overclaiming a Clay solution, and quietly under-reporting a certified terminal as if it were still owed work. Both are refused explicitly below.

1. What is explicitly NOT claimed

(a) No Clay solution, in any form, at any strength. The Yang–Mills existence-and-mass-gap problem, as stated by Jaffe and Witten for the Clay Mathematics Institute in 2000, is unsolved — in this corpus, in the wider physics and mathematics literature, and by any construction referenced anywhere in this dossier. Nothing in the reduction chain, the geometry, the no-shortcut theorem, or the Monte Carlo diagnostics constitutes a proof that $$ \operatorname{Spec}\big(H|_{\mathcal H_{\rm phys}^{\rm YM}}\big)\cap(0,\Delta)=\varnothing $$ holds for some \(\Delta>0\). Any value of \(\Delta\) quoted anywhere in this corpus is inherited from experiment (the lightest glueball spectrum, the roughly-1-fm confinement radius, the observed running of \(\alpha_s\)) or from partner lattice programs — never derived here. If a reader takes away only one sentence from this dossier, it should be this one.

(b) Dissolved is not solved. Section 4 of the reduction chain shows that the Uniform Operational Cell Law \(\Delta_0>0\) — a fixed physical resolution \(\ell_\star\sim\Lambda_{\rm YM}^{-1}\) — makes the ontic, infinite-precision continuum limit \(a\to0\) record-impossible-in-principle, converting Face A of the problem (does a genuine 4D continuum \(SU(3)\) measure exist as a mathematical object with \(a\) taken literally to zero) into a finite, checkable statement about a gauge-invariant certificate alphabet \(\mathcal C_\star\) that is provably finite. This is a real result — a two-route-confirmed dissolution — but it is a dissolution of Face A only. It changes what question is being asked about continuum existence; it does not answer the original question in its original ontic form, and it supplies zero traction on Face B. Declining to ask whether \(a\to0\) literally exists as an ontic limit is not the same act as proving the gap survives that limit. The corpus's own firewall rule — identical in wording to the one used for gap01 and for the black-hole-singularity gate, and invoked here without modification — states this directly: "Granularity dissolves the continuum, never the gap." Face B, the uniform gap bound \(\Delta(a,L)/\Lambda_{\rm YM}\ge c>0\) itself, is a physical observable (short-range confining force, massive glueballs, running \(\alpha_s\)), and the framework's own G1 rule (OBSERVABLE-NEVER-DISSOLVES) fires on exactly this object and forbids dissolving it. There is no route by which the granularity argument could be extended to cover Face B without violating a rule the corpus imposes on itself uniformly, not selectively for this gate.

(c) Selection is not derivation. The frozen chamber center \(\vec u=(1,1,1)\) on \(K_6=SU(3)/T^2\) is a Weyl-rigid-admissible point inside the squashing family \(\vec u\in[1/2,3/2]^3\); the completion audit (§4.2 of the reduction) shows this whole squashing family is Wilsonian-irrelevant to the mass-gap question, decoupling at \(\sim10^{16}\) GeV — one of exactly one row so classified in the three-layer tally (INERT: 10, ORDINARY-YM: 3, WILSONIAN-IRRELEVANT: 1, TOPOLOGICAL/BOUNDARY→R4: 1, OPEN-CANDIDATE lever: 0). That a particular admissible chamber point is selected by the corpus's selector machinery, and that the resulting \(SU(3)_c\) structure constants and \(\Lambda_{\rm YM}\) value are then read off, is not the same act as deriving the mass gap's existence from that data. The selection fixes which Yang–Mills theory is being asked about (which group, at what scale); it does not touch the dynamical question of whether that theory, once specified, has a gap. Lemma 1's force is precisely that this distinction is airtight: fixing the group and scale via geometric selection provably contributes nothing further to the continuum-limit dynamics, so there is no channel by which "selection of \(\vec u=(1,1,1)\)" could be mistaken for, or silently substituted for, "derivation of \(\Delta>0\)."

(d) Given-\(E\) is not derivation-of-\(E\). The scale \(\Lambda_{\rm YM}\) that the ratio \(\Delta/\Lambda_{\rm YM}\) is measured against is fixed by the frozen anchors through the standard two-loop \(\overline{\rm MS}\) renormalization-group closure: one-loop coefficients \((b_1,b_2,b_3)=(41/10,\,-19/6,\,-7)\), total KK threshold packet \((\delta_1,\delta_2,\delta_3)=(+4.8424,\,-3.1112,\,-1.7313)\pm1.6\times10^{-3}\), unification at \(M_U\sim1.0\times10^{16}\) GeV with residual \(9.6\times10^{-11}\), run from \(M_Z=91.1876\) GeV. This machinery fixes \(\Lambda_{\rm YM}\) — it does not derive it from anything more primitive than the four irreducible anchors \(\{M_{\rm Pl},\alpha_i(M_Z),y_t,|V_{us}|\}\) that the whole corpus treats as free inputs. Producing a numerical scale by running known low-energy couplings up to a unification point is boundary-data bookkeeping, not a new physical derivation, and this dossier does not present it as one. \(M_{\rm Pl}\) itself plays no direct role in the dimensionless ratio \(\Delta/\Lambda_{\rm YM}\) that is the actual target quantity — the scale-fixing machinery and the mass-gap question are, correctly, kept as separate pieces of scope.

(e) The conditional lemma is not the gap. Section SL-1/M4D of the reduction proves, as a rigorous, hand-checkable implication of pure functional analysis (reusing Osterwalder–Schrader reconstruction, the spectral theorem, and the Källén–Lehmann representation — cited, not re-derived), that $$ [H1\wedge H2\wedge H3\wedge H4]\ \Longrightarrow\ \text{no soft physical sequence}\ \Longrightarrow\ \Delta>0. $$ This is a genuine theorem. It is not a proof that the gap exists, because \(H1\)–\(H4\) — continuum-limit existence of the measure, its reflection positivity in the limit, its nontriviality/unique-vacuum property, and uniform exponential clustering through the marginal coupling band — are exactly the open content, not discharged hypotheses. At every finite lattice spacing \((a,L)\) all four properties hold as established theorems (Osterwalder–Seiler 1978, Lüscher 1977, Münster 1981); the entire remaining wall is whether they survive the joint uniform limit \(a\to0,\,L\to\infty,\,\beta\to\infty\). This dossier states the conditional lemma exactly as a conditional, never as an unconditional gap proof, at every point it is invoked.

(f) No value, sign, or bound is assigned to any of the open symbolic constants. \(\delta\), \(K\), \(\kappa\), \(c\), \(c'\), \(\Delta\) itself (as a derived rather than measured quantity), \(\rho_\star\), \(z_\star\), and \(\xi_{R4}\) all remain strictly symbolic throughout this dossier. In particular: - \(z_\star\) is convention-underdetermined across three legitimate readings — \(7.66\ldots\) (diverges as \(\delta_{\rm tr}\to0\), FAIL and ill-posed), \(0.109\) (intensive KP polymer activity per reference site, PASS), and \(0.353\) (single effective activity \(e^{-s_{\rm block}}\), PASS) — and this dossier does not pick a reading to manufacture a verdict. Until \(w(\gamma)\) is derived target-blind, \(z_\star\) is not decidable, full stop. - \(\xi_{R4}\in\mathbb Z/3\) (via the twisted bordism group \(\Omega_5^{\rm Spin\text{-}c}(B(SU(3)\to PSU(3));\tau_{K_6})\)) is not computed in this dossier. Its value is not asserted to be zero, nonzero, or any specific class element. The operative differential is identified as the 3-primary Milnor primitive \(d_5=Q_1=\beta P^1\), and it is shown that \(Q_1(u_2)=Q_1(2y_1+2y_2)=0\), so the center class \(u_2\) survives as a \(d_5\)-cycle — a structural fact about which differential is possible, not a computation of the twisted correction that would fix \(\xi_{R4}\)'s actual value. - The Monte Carlo master-inequality diagnostic \(D=s_{\rm block}-\log(E_{\rm conn}\cdot A_{\rm fluc}\cdot N_{\rm cert})\) returns \(D<0\) robustly (at \(\beta=6.0\), \(4^4\) lattice, 16 configurations, plaquette-gated: \(D=-4.004\pm 0.041\), \(-5.043\pm0.041\), \(-6.162\pm0.041\) at \(\delta_{\rm tr}=.5,\,.25,\,.125\) respectively, robust to roughly \(66\sigma\)). This is reported as exactly what it is — an INCONCLUSIVE result on a sufficient, not necessary, worst-case Kotecký–Preiss proxy (\(z_\star\le N_{\rm cert}\max_\gamma w(\gamma)\)) — never as a refutation of the target inequality and never as a confirmation of it. A genuine refutation would require a direct lower bound \(z_\star\ge1/(E_{\rm conn}A_{\rm fluc})\), which was not computed anywhere in this program.

(g) No manufactured lever, no false coincidence banked. Two \(O(1)\) numerical coincidences that surfaced during this program — \(\kappa^3/\pi\) and \(5+3=8\) — were explicitly caught by the corpus's no-target-fitting guard and retired. They are carried in this dossier only as a discipline reminder that an \(O(1)\) ratio matching some other \(O(1)\) ratio is not, by itself, a physics verdict. Similarly, the three-layer completion audit reports exactly zero rows in the OPEN-CANDIDATE-lever class; this dossier does not manufacture a fifth open lever to make the picture look more or less finished than it is.

(h) This is not category #5, and it is not a "structural frontier" in the Gap-13/UQF-4 sense. The NO-BARE-#5 audit was run explicitly on this gate: no named completeness-basis theorem exists proving the candidate-rule space (all conceivable uniform-gap proof strategies) is exhausted — constructive QFT possesses no such completeness theorem — so this is not a case of "OPEN wearing a certificate," the Impostor-1 pattern the corpus's own handbook warns against. Instead this routes cleanly to a recorded external wall: a named, specific, field-wide obstruction (the Clay statement itself) that no single gate, and no single research program anywhere, currently owns. This dossier also does not relabel gap02 a "structural frontier" in the sense reserved for Gap-13 (black-hole entropy) or UQF-4 (global anomalies) — those are frontiers of this framework's own construction; gap02's obstruction predates the framework by a quarter century and is owned by the entire field. Keeping these categories separate is a stated discipline, not a stylistic choice.

2. The anchors paid — an itemized account

Reaching this terminal was not free. Three distinct anchors were consumed, and none of them is disguised as an output of this gate.

Anchor 1 — the mass gap itself, \(\Delta>0\), taken as MEASURED. The existence of a nonzero mass gap in real QCD is not in serious experimental doubt: the lightest glueball states are massive, the strong force is short-ranged (confinement radius of order 1 fm), and \(\alpha_s\) runs to strong coupling in the infrared in a manner consistent with a gapped spectrum. This dossier takes \(\Delta>0\) as a Tier-1 measured anchor precisely because the missing piece is not "does nature have a gap" — empirically, unambiguously, yes — but "does the Lagrangian, treated as a rigorous continuum quantum field theory, prove that it must." The gate never claims to output \(\Delta\); it consumes the measured fact of \(\Delta>0\) as a floor and asks only whether the theory can be shown, rigorously, to reproduce it.

Anchor 2 — the scale \(\Lambda_{\rm YM}\), taken as GIVEN-E boundary data. As itemized in §1(d) above, the full numerical machinery fixing \(\Lambda_{\rm YM}\) — the one-loop beta coefficients, the KK threshold packet, the two-loop unification closure at \(M_U\sim1.0\times10^{16}\) GeV, the natural radius \(R_0=1.591549430918954\times10^{-17}\) GeV\(^{-1}=(2\pi M_U)^{-1}\) — is boundary data supplied to this gate, not an output derived by it. Every one of these numbers is inherited from the corpus's four irreducible anchors \(\{M_{\rm Pl},\alpha_i(M_Z),y_t,|V_{us}|\}\) through machinery documented and audited elsewhere in the corpus; gap02 consumes the resulting \(\Lambda_{\rm YM}\) exactly as it consumes \(\Delta\), as a fixed external number, never as something it computes from first principles.

Anchor 3 — the granularity/cost-floor axiom, \(\Delta_0>0\), taken as a REDUCED-TO-AXIOM posit, not a proof. The Uniform Operational Cell Law that dissolves Face A is itself an unproven standing posit, not a theorem. This is stated with full honesty in the reduction chain: REDUCED-TO-AXIOM does not lower the axiom count. A rigorous basin-shallowing countermodel exists and is unrefuted — depths \(d_n=B\cdot2^{-n-1}\) satisfy \(\sum_n d_n=B/2<\infty\) while \(\inf_n d_n=0\), showing that finite total resources alone do not force a uniform positive floor. So the claim "\(\Delta_0>0\) uniformly" is a genuine axiom choice with a live countermodel against its most naive justification, carried forward honestly as \(\hbar\)'s measured residue standing in for the posit, not as a discharged theorem. Anyone who wants to reject this axiom is rejecting a stated, named, falsifiable posit — not uncovering a hidden assumption the corpus tried to bury.

Three anchors, each named, each priced, each kept distinct from the two things this gate did derive: the reduction chain itself (a rigorous logical structure) and Lemma 1 (a genuine no-shortcut theorem). No anchor is allowed to migrate silently into the "derived" column.

3. Why the terminal is legitimate and not a dodge

The corpus's own root-toolbox audit is worth restating here because it is the strongest evidence that this is a genuine wall rather than an unexamined one. Running all seven roots against gap02 returns: Shape → EXPOSE, Scale → PASS, Granularity → CONSTRAIN (Face A only), Invariance → PASS, Record-Interface → EXPOSE, Causal-Order → PASS, Nonseparability → EXPOSE. No root FORCEs a unique survivor and no root ELIMINATEs the open status of R1–R3. That specific pattern — full engagement of the toolbox with zero roots resolving the residual — is exactly the signature the corpus uses elsewhere to distinguish a genuine non-dissolvable external wall from an under-examined one where the right root simply had not yet been tried. Every root was tried here. None of them touches Face B, because Face B is a real physical observable and G1 forbids dissolving observables by construction — this is not a loophole the gate slipped through but a rule it is honestly bound by.

The three universal-negative temptations that would masquerade as stronger claims are each named and dissolved as limits on all knowledge rather than left as ambiguous gaps: - "No future theory could ever solve Clay by some entirely different route" is an unprovable universal negative; the honest bounded ceiling is "no accepted constructive proof exists today, community-wide" — a statement about the present state of mathematics, not a claim about all possible futures. - "The granularity axiom is the provably irreducible bottom of physical resolution" is undischargeable given the live basin-shallowing countermodel; the honest ceiling is "earned-irreducible under every reduction attempted so far, a confessed posit," not "proven fundamental." - "No observation-preserving modification of the framework could ever supply a proof lever" is bounded by Lemma 1 itself: on every settleable row of the three-layer completion audit, none does; R4 remains the one honestly unsettled row, carried forward as open rather than folded into either a false optimism or a false pessimism.

4. The smallest remaining objects, named plainly (R1–R4, with R2/R3 as co-gates)

If a reader wants to know exactly what would need to happen for this terminal to be upgraded from "wall recorded" to "wall removed" — an event that would be a field-wide Millennium Prize result, not routine gate maintenance — the objects are these, named without euphemism:

R1 — the uniform-gap bridge (the Clay object itself, = H4 = shared wall W18). The precise open statement is $$ \Delta(a,L)/\Lambda_{\rm YM}\ \ge\ c>0\quad\text{uniformly as}\quad a\to0,\ L\to\infty,\ \beta\to\infty, $$ equivalently uniform exponential clustering of the renormalized glueball correlator through the marginal coupling band, equivalently the marginal Kotecký–Preiss convergence inequality $$ \rho_\star = E_{\rm conn}\cdot A_{\rm fluc}\cdot e^{-s_{\rm marg}} < 1,\qquad(\delta>0). $$ The obstruction is precisely stated: in the deep ultraviolet (\(g\to0\)) the large-field weight \(\mu_n\le e^{-c/g^2}\to0\) banks the inequality automatically, but inside the \(d=4\) marginal band \(g(2^na)=O(1)\) the inequality becomes a strict finite comparison between two genuinely \(O(1)\) constants with no known small control parameter — a weight bound is not a contribution bound, and that category error is the wall itself. This obstruction is method-invariant: both the polymer-expansion route and the asymptotic-freedom reflection-positivity monotone route (\(M_{n+1}\le(1-c\,b_0\,g_n^2)M_n+O(g_n^4)M_n\), \(b_0=11N/3>0\)) stall on the identical \(\delta=0\) marginal object. Three live attack lanes are named: (Lane A) sharpen the Bałaban large-field estimate from a weight bound to a genuine contribution bound, contracting by a fixed \(e^{-\delta}\) uniformly through the band; (Lane B) construct a convergent activity/cluster expansion at \(g=O(1)\) with strictly positive \(\delta\); (Lane C) match the ultraviolet (Gaussian) and infrared (product-measure) expansion domains, or find a group-sensitive non-expansion method. The falsifier is equally sharp: a direct proof of uniform \(\rho_\star\ge1\) would kill this entire sufficient route. Per the Clay problem statement's own anticipation, closing this most likely requires a genuinely new idea in constructive quantum field theory, not a rearrangement of the tools already in hand.

R2 — nonperturbative BRST/Gribov positivity surviving \(a\to0\) (independent co-gate, shared wall W2). Even a complete proof of R1 would not by itself yield the gap: the physical Hilbert space \(\mathcal H_{\rm phys}=\ker(s)/ {\rm im}(s)\) must be shown to carry a genuine positive-definite inner product nonperturbatively, in the continuum limit. This is established only perturbatively today (Kugo–Ojima quartet mechanism); the nonperturbative obstruction is the Gribov–Singer no-go (no global continuous gauge section exists on a nontrivial bundle) compounded by the Neuberger \(0/0\) problem on the lattice (Faddeev–Popov determinant sign-flips across Gribov copies). This is named explicitly as a separate, independently load-bearing open object, not folded into R1 — closing R1 alone would not close the gate.

R3 — Osterwalder–Schrader reconstruction surviving the continuum limit (independent co-gate, shared wall W2). A genuine 4D \(SU(3)\) continuum measure \(\mu=\lim_{a\to0}\mu_{a,L}\) satisfying reflection positivity (OS-0 through OS-3) in the limit, together with nontriviality and a unique vacuum, has never been constructed by anyone. This is \(H1\wedge H2\wedge H3\) of the conditional lemma bundled together — three predicates of the same limit object \(O\), whose very existence is the first thing that must be shown before its properties can even be asked about. Constructive success at this level exists only in spacetime dimension \(d\le3\) (Bałaban renormalization-group methods, constructive SPDE techniques); the marginal renormalizability of \(d=4\) leaves no spare coercive margin for any of these techniques to date.

R4 — the one genuinely framework-internal open lever (bordism-anomaly value \(\xi_{R4}\)). This is the single open object that is internal to this framework's own geometry rather than a restatement of the field-wide Clay wall. The well-posed home of the question is $$ \xi_{R4}\ \in\ \Omega_5^{\mathrm{Spin\text{-}c}}\big(B(SU(3)\to PSU(3));\ \tau_{K_6}\big), $$ a finite abelian group (a twisted degree-5 bordism invariant), carried by the degree-3 class \(\bar x_1\in H^3(BPU(3);\mathbb Z/3)\), with operative differential the 3-primary Milnor primitive \(d_5=Q_1=\beta P^1\). The untwisted control computations are fully resolved and cited: \(\Omega_5^{\rm Spin}(PSU(3)\times B^2\mathbb Z_3)= \mathbb Z_3\ne0\) (the center class survives as a \(d_5\)-cycle since \(Q_1(u_2)=Q_1(2y_1+2y_2)=0\)), while \(\Omega_5^{\mathrm{Spin\text{-}c}}(B^2\mathbb Z_3)=0\) in the untwisted case — but this untwisted zero is explicitly not \(\xi_{R4}=0\), because \(\xi_{R4}\) lives specifically in the twisted group, and the twist \(\tau_{K_6}\), sourced by \(c_1(TK_6)=2\rho=(2,2)\not\equiv(0,0)\bmod3\) (forced nonzero by the spin-\(\mathbb C\) index \(\chi(K_6,E)=-3\)), is the only possible source of a nonzero answer. The concrete open step is to evaluate the twisted \(d_5=Q_1\) correction on the degree-5 line: does the frozen \(\tau_{K_6}=(2,2)\) twist supply a nonzero degree-lowering image or not? This is a well-defined, finite, in-principle-computable algebraic-topology question with three possible outcomes, none of which may be asserted until the computation is actually done: (a) \(\xi_{R4}\) trivial, confirming Lemma 1 across the board with no exception; (b) \(\xi_{R4}\) nontrivial and descending to 4D and surviving the continuum limit and imposing a genuine Faddeev–Popov/Gribov boundary condition and universality-preserving, which would promote R4 to a genuine geometry-sourced confinement ingredient (still not a Clay proof — R5 below shows why); (c) nontrivial but non-universal or failing one of the survival conditions, in which case R4 collapses back into Lemma 1's "no lever" verdict. The honest prior tilts away from outcome (b): the compact boundary/parity sector is on record elsewhere in the corpus as adding difficulty rather than supplying coercivity, so a nonzero, gap-relevant \(\xi_{R4}\) would be the surprising result, not the expected one. The tractability horizon for this specific computation is estimated at 5–15 years and is explicitly flagged as such, not disguised as imminent.

R5 — necessary-not-sufficient caveat on R4 (a derived caveat, not standalone work). Even a confirmed nonzero and universality-preserving \(\xi_{R4}\) would not, by itself, prove the gap: 't Hooft anomaly matching admits gapless conformal or topological-field-theory saturation of the anomaly, so a nontrivial R4 would need a separate proof that the matched infrared phase is specifically gapped or confining, not merely anomaly-consistent. This caveat is carried explicitly so that a positive R4 result, if it ever arrives, is not mistaken for an automatic gap proof.

Two further named objects are not additional open holes but honestly-carried background facts that were priced above as anchors rather than debts: R6 (the granularity axiom \(\Delta_0>0\), REDUCED-TO-AXIOM, not discharged — priced in §2 above) and R7 (Lemma 1 itself, the geometry-supplies-no-lever theorem — a completed, banked result, nothing owed) and R8 (the measured-anchor status of \(\Delta\) and the given-\(E\) status of \(\Lambda_{\rm YM}\) — also priced in §2). These are listed here only to make explicit that they are not double-counted as additional open compute alongside R1–R5.

5. The closing endpoint statement

Every named object above routes to one of exactly two places: a completed, banked result internal to this gate (the reduction chain, Lemma 1, the three-layer completion audit, the localization to \(\rho_\star<1\), the honestly- reported INCONCLUSIVE Monte Carlo diagnostic), or a standing external wall / open co-gate that is shared across the field and across other gates in this corpus (R1–R3, shared walls W2 and W18) or is a named, well-posed, bounded, cross-gate computation with an honest tractability horizon (R4, shared with SG-4 and the Born-rule work). Nothing was found that belongs to this gate alone, uniquely, as an unnamed or unbounded debt. The gate's own proof-domain completion audit — INERT ×10, ORDINARY-YM ×3, WILSONIAN-IRRELEVANT ×1, TOPOLOGICAL/BOUNDARY→R4 ×1, OPEN-CANDIDATE lever ×0 — is exhaustive over every \(\oplus/\otimes\) primitive in the frozen branch, and the sole open row (R4) has already been named, bounded, and exported precisely rather than left as a residual haze.

Nothing left. Anchored on: Shape: \(K_6=SU(3)/T^2\)'s left-isometry algebra \(\mathfrak{su}(3)\) supplies the color group \(SU(3)_c\) that the Clay problem is posed for, with Lemma 1 (a proved theorem, not an omission) certifying every remaining \(\oplus\)-Rulebook and \(\otimes\)-Actors primitive of the frozen 13D branch — the full flavor chamber \(F^+\), the admissibility firewall \(\mathcal C_{\rm admiss}\), the matter/Higgs/proton actors, and the Weyl-rigid squashing chamber \(\vec u\in[1/2,3/2]^3\) itself — Wilsonian-irrelevant to the continuum-limit dynamics, leaving only the ordinary gauge actor \(\mathcal E_{\rm gauge}=(A_c,F_c,\rho_{\rm rep})\) on \(\mathcal M_4\) with structure group \(SU(3)_c\), exactly the Clay object, with one topological row (R4, \(\xi_{R4}\in\Omega_5^{\rm Spin\text{-}c}(B(SU(3)\to PSU(3));\tau_{K_6})\)) honestly exported rather than resolved; Granularity: the Uniform Operational Cell Law \(\Delta_0>0\) dissolves Face A (ontic continuum existence) into a finite gauge-invariant certificate-alphabet question via the fixed physical resolution \(\ell_\star\sim\Lambda_{\rm YM}^{-1}\), while G1 (OBSERVABLE-NEVER-DISSOLVES) correctly and permanently forbids the same move on Face B, the physical mass gap itself; Scale: \(\Lambda_{\rm YM}\) is fixed as given-\(E\) boundary data via the two-loop \(\overline{\rm MS}\) unification closure (\(M_U\sim1.0\times10^{16}\) GeV, residual \(9.6\times10^{-11}\), \(R_0=1.591549430918954\times 10^{-17}\) GeV\(^{-1}\)), with \(M_{\rm Pl}\) correctly playing no role in the target ratio \(\Delta/\Lambda_{\rm YM}\); Observables: the mass gap \(\Delta>0\) itself, taken as a Tier-1 measured anchor from the lightest glueball spectrum, the confining short-range strong force, and the running of \(\alpha_s\) — never claimed as a gate output; Dissolution: the continuum existence question (Face A) dissolves under the granularity axiom into a finite checkable inequality, but this changes which question is asked about ontic continuum existence — it does not answer, and cannot be extended to answer, the surviving physical mass-gap question itself, which remains exactly and only the standing Clay Millennium Prize problem, localized here to the single marginal inequality \(\rho_\star=E_{\rm conn}\cdot A_{\rm fluc}\cdot e^{-s_{\rm marg}}<1\) and handed forward, undissolved and unsolved, as this gate's honestly recorded residual.

Closure ledger — Gap-02 — Yang–Mills mass gap

Status (governing, 2026-07-12): CLOSED-SCOPED / POSITIVE-SPECTRAL-GAP-DERIVED-FOR-EACH-FINITE-OPERATIONAL-SU3-LATTICE / CONTINUUM-AND-INFINITE-VOLUME-UNIFORM-GAP-PSTAR-CERTIFIED-IRREDUCIBLE / NOT-A-CLAY-PROOF

[SUPERSEDED 2026-07-12 — the ledger below was graded CERTIFIED-IRREDUCIBLE / RESOLVED +0 under the pre-correction "no gap derived, reduction-only" reading. See Governing Correction: the finite operational gap is now DERIVED; only \(P_\star\) stays certified-irreducible. Legacy ledger rows are retained in full; the "STATUS-UPGRADES:0 / do not alter" tag applied to the old grade and does not block this owner-ratified strengthening.]

The technical closure LEDGER (separate document)

Gate: Gap-02 — Yang–Mills mass gap. Fixed grade (STATUS-UPGRADES:0, do not alter): CERTIFIED-IRREDUCIBLE / RESOLVED +0. [SUPERSEDED 2026-07-12 — governing grade CLOSED-SCOPED; finite gap DERIVED, \(P_\star\) CERTIFIED-IRREDUCIBLE. See Governing Correction.]

This ledger is the reviewer's record: every layer pinned, every anchor typed, every leg of the reduction chain numbered and graded, every open residual named with its exact closing condition. Nothing here is asserted beyond what is shown. Symbolic quantities (\(\delta,K,\kappa,c,c',\Delta,\rho_\star,z_\star,\xi_{R4}\)) are never assigned invented numbers; where a number is reported it is tagged MEASURED, DERIVED, GIVEN-E, or OPEN.

L0 — The Layer-0 wall identity

Wall carried: the Clay Mathematics Institute Millennium Problem "Yang–Mills Existence and Mass Gap" (Jaffe–Witten, 2000) for ordinary 4D pure-glue \(SU(3)_c\) Yang–Mills on \(\mathbb R^4\).

Target theorem (verbatim, the object the wall is measured against): $\(\exists\,\Delta>0:\ \operatorname{Spec}\big(H|_{\mathcal H_{\rm phys}^{YM}}\big)\cap(0,\Delta)=\varnothing,\)$ equivalently $\(\inf\big\{\langle\psi,H\psi\rangle:\psi\perp\Omega,\ \psi\in\mathcal H_{\rm phys},\ \|\psi\|=1\big\}=\Delta>0.\)$

Bundled Clay requirements carried as ONE package: (a) Osterwalder–Schrader (OS) reconstruction of a genuine 4D continuum \(SU(3)\) measure; (b) unique vacuum \(\Omega\), \(H\Omega=0\); (c) self-adjoint \(H\ge0\); (d) \(\Delta\ge m\ge c'\Lambda_{\rm YM}>0\), \(c'\) symbolic and unknown.

Wall type: an external wall — no single gate in this framework, and no research group anywhere, owns it. It predates and outlives this corpus. The gate's entire job is therefore not to solve the wall but to state, at full rigor, exactly how far the frozen 13D geometry gets before running into it, and to certify that no undisclosed shortcut around it exists.

Why L0 is "wall," not "gap-in-this-framework": Jaffe–Witten's own problem statement anticipates the missing ingredient is "most likely a genuinely new idea in constructive QFT," not a rearrangement of known technique. A 13-dimensional UV completion is additional structure sitting above the wall (as boundary data), not a technique that acts on the wall's interior (the \(d=4\) marginal continuum limit). Lemma 1 (leg 8 below) is the proof of that last sentence, not an assumption of it.

L1 — The Layer-1 endpoint anchor

[SUPERSEDED 2026-07-12 — governing endpoint CLOSED-SCOPED: finite operational SU(3)-lattice gap DERIVED (certificate PASS), continuum-uniform \(P_\star\) CERTIFIED-IRREDUCIBLE. The pre-correction "external-wall-recorded, no gap derived" endpoint below is retained for the record and now scopes to \(P_\star\) only. See Governing Correction.]

Endpoint reached: CERTIFIED-IRREDUCIBLE (external-wall-recorded), sitting in the RESOLVED +0 bucket alongside measured-anchor and dissolved-given-root terminals — never conflated with an OPEN compute this gate still owes.

Two objects that must not be merged:

1. The gate status (what this framework achieved, relative to itself): CERTIFIED-IRREDUCIBLE / RESOLVED +0. This is a statement about the gate's own completeness — it has reduced its question to the wall, proved no lever exists elsewhere in the geometry, localized the residual to one finite inequality, and taken the phenomenological gap value as an anchor rather than an output.
2. The residual displayed by that status (what remains true about the world): the Clay problem itself is unsolved, here and everywhere. This residual is shown, not hidden and not rolled into a hedge on the gate status.

NO-BARE-#5 audit (why this is not a disguised open compute): the certificate-audit Q0 test was run explicitly. A bare "#5" (an unresolved item wearing a false certificate) requires either (i) a completeness theorem that the candidate-rule space is exhausted, or (ii) target-fitting. Neither is present: constructive QFT possesses no completeness theorem over "all possible uniform-gap proof strategies," and every quantity here is target-blind (§ "anti-claims" below). The audit therefore routes the wall to a recorded external wall, not to #5. This is the corpus's own canonical example of the distinction.

Wall-register line (authoritative): "Gap-02 Yang–Mills mass gap | ANCHOR-CLOSED | Reduced-to-axiom (granularity) | shared walls W18, W2, W4 | TERMINAL-ANCHORED (axiom-conditional; NOT a Clay solution)." [SUPERSEDED 2026-07-12 in part — governing register: CLOSED-SCOPED; finite operational SU(3)-lattice gap DERIVED (certificate PASS); \(P_\star\) CERTIFIED-IRREDUCIBLE. "NOT a Clay solution" is RETAINED verbatim. See Governing Correction.]

L2 — The frozen object carried as boundary data (all three layers pinned)

Topological invariants of \(K_6\) carried into the pure-glue branch (frozen, exact):

Frozen branch identifiers (read-only audit tags certifying which object was tested — never physics levers, never printed as validation content beyond this internal record): branch , manifest, spin-c bundle ``.

Matter removed: this is the pure-glue / Clay branch. \(K_6=SU(3)/T^2\) supplies the origin of the color group \(SU(3)_c\) (left-isometry \(\mathfrak{su}(3)\) of \(K_6\)) and, through the frozen anchors, the scale \(\Lambda_{\rm YM}\) — but supplies neither a proof technique nor a relaxation of the Clay object itself.

L2 — Root stack: Tier A (Shape / Scale / Granularity, full precision)

A.1 SHAPE — supplies ORIGIN, not PROOF

• × Stage load-bearing facts: \(K_6=SU(3)/T^2\) left-isometry \(\mathfrak{su}(3)\) is the geometric source of \(SU(3)_c\). \(c_1(TK_6)=2\rho=(2,2)\not\equiv(0,0)\bmod3\) (nonzero mod-3 class, relevant only to the R4 side-channel below, not to R1–R3). \(b_2(K_6)=2\Rightarrow H^2(K_6;\mathbb Z_3)=(\mathbb Z/3)^2\).
• ⊕ Rulebook + ⊗ Actors: all UV-completion / admissibility machinery. Lemma 1 (leg 8 of the reduction chain, § below) is a theorem that this machinery is INERT for the IR continuum-limit question — not an assumption, a proved negative result.
• Three-layer completion audit tally (every ⊕/⊗ primitive at proof-domain level classified):

Exactly one open row (R4), zero manufactured levers. Verdict: shape SHAPE-root = EXPOSE (supplies origin/motivation, forces nothing, eliminates nothing about the wall).

A.2 SCALE — \(\Lambda_{\rm YM}\) is GIVEN-E, not derived here

\(\Lambda_{\rm YM}\) is pinned by \(\alpha_3(M_Z)\), \(M_U\), and the two-loop threshold structure — a scale-fixing, not a scale-derivation (deriving \(\Lambda_{\rm YM}\) is deriving the anchor \(E\), explicitly out of this gate's scope). Supporting frozen numbers, full precision:

\(M_{\rm Pl}\) plays no direct role in the ratio \(\Delta/\Lambda_{\rm YM}\) that R1 must bound. Verdict: SCALE-root = PASS (fixes which scale, contributes no proof leverage on the marginal continuum limit).

A.3 GRANULARITY — dissolves Face A only; G1 firewall forbids Face B

The gap02 question splits into two faces that must never be merged:

• Face A — continuum existence as an ontic infinite-precision limit. Status: DISSOLVED-GIVEN-(Granularity ∧ Record-Interface), 2-route confirmed. Mechanism: the Uniform Operational Cell Law posits a floor \(\Delta_0>0\) on resolvable action, making the ontic limit \(a\to0\) record-impossible in principle; a fixed physical resolution \(\ell_\star\sim\Lambda_{\rm YM}^{-1}\) makes the gauge-invariant certificate alphabet \(\mathcal C_\star\) finite, converting "does the continuum limit exist" into a finite, checkable inequality. In the cost-floor wall-impact ledger this is the single physical wall (DISSOLVES=1), decomposing into 13 continuum sub-rows (MO-1..MO-7, uniform_gap_bridge, Lemma2, G02.A/B, Clay_theorem_i, Seed_theorem_ii).
• Face B — the uniform gap bound itself. This is an observable (confinement radius \(\sim1\) fm, massive glueball spectrum, running \(\alpha_s\)). G1 (OBSERVABLE-NEVER-DISSOLVES) fires and forbids dissolving Face B. Firewall line, shared verbatim with gap01 and blackhole-singularity: "Granularity dissolves the continuum, never the gap."

Dissolved ≠ solved: declining the ontic continuum limit changes the Clay question (replaces "does an infinite-precision limit exist" with "does a finite-resolution certificate satisfy one inequality"); it does not answer either version.

REDUCED-TO-AXIOM does not lower the axiom count. The floor \(\Delta_0>0\) is a named, unproven posit (with \(\hbar\) as its measured residue, not its proof). A rigorous countermodel shows the posit is not discharged: basin-shallowing sequence \(d_n=B\cdot2^{-n-1}\) has \(\sum_n d_n=B/2<\infty\) but \(\inf_n d_n=0\) — finite total resource buys only \(\eta\)-dependent total boundedness, never a uniform positive floor. Verdict: GRANULARITY-root = CONSTRAIN, Face-A only (Face B untouched by construction).

L2 — Root stack: Tier B (Layer-2 audit screens, from the completion-run reviewer)

Seven-root saturation (why the terminal is genuine, not premature): Shape=EXPOSE, Scale=PASS, Granularity=CONSTRAIN (Face-A only), Invariance=PASS, Record-Interface=EXPOSE, Causal-Order=PASS, Nonseparability=EXPOSE. No root FORCES a unique survivor and no root ELIMINATES the open status — this null result across all seven roots is itself the signature of a genuine non-dissolvable external wall, not an oversight.

L3 — Measured anchors and their exact role (consumed / reproduced / tested)

The four \(O(1)\) constants of the floored certificate inequality (never recompute to "improve" — the open question is not their values):

Non-rigorous companion (never promoted): series/MC estimate \(\lambda_4\approx13\)–\(14\) — EMPIRICAL/PARTIAL only, not substitutable for the rigorous \(E_{\rm conn}=19.0280\).

L4 — The full derivation chain: numbered ledger, each leg graded

Reading the ladder: legs 1, 3, 8, 9 are banked (RESOLVED +0, no residual). Legs 2 is a rigorous conditional implication (never a value). Legs 4–7, 10–12, 14 are the genuinely open content, and they collapse (leg 9) into exactly ONE localized finite object (leg 10/11) plus two independent co-gates (legs 5–6/12) plus one framework-internal side channel (leg 14) that is low-leverage even if resolved (see R5 below). Leg 13 is an axiom, not a compute debt. Leg 15 is the floor.

L5 — R4: the one framework-internal lever (corrected object)

The question: does the frozen \(K_6\) spin-\(\mathbb C\)/chamber data induce a nontrivial, universality-preserving mixed 't Hooft anomaly with the \(\mathbb Z_3^{[1]}\) one-form center of pure \(SU(3)\) (GKSW background \(B^{(2)}\in H^2(X,\mathbb Z_3)\))?

Well-posed home (current, authoritative — supersedes an earlier malformed framing): $\(\xi_{R4}\in\Omega_5^{\mathrm{Spin\text{-}c}}\big(B(SU(3)\to PSU(3));\ \tau_{K_6}\big)\quad\text{— a finite abelian group, twisted degree-5 bordism invariant.}\)$

• Carrier: degree-3 class \(\bar x_1\in H^3(BPU(3);\mathbb Z/3)\) — there is no degree-2 mod-3 carrier.
• Operative differential: 3-primary \(d_5=Q_1=\beta P^1\) (Milnor primitive, \(|Q_1|=5\)). \(Q_1(y_i)=0\), \(Q_1(x_i)=2y_i^3\), and \(Q_1(u_2)=Q_1(2y_1+2y_2)=0\Rightarrow u_2\) is a \(d_5\)-cycle \(\Rightarrow\) the survivor persists.
• Untwisted controls: \(\Omega_5^{\rm Spin}(PSU(3)\times B^2\mathbb Z_3)=\mathbb Z_3\ne0\) (survivor persists); independently \(\Omega_5^{\mathrm{Spin\text{-}c}}(B^2\mathbb Z_3)=0\) (fully resolved, resting on \(H_3=H_5=0\), \(H_4=\mathbb Z_3\) of \(K(\mathbb Z_3,2)\) + odd-\(q\) vanishing of spin-c point coefficients: \(\mathbb Z,0,\mathbb Z,0,\mathbb Z^2,0,\ldots\)). Critical: untwisted \(0\) is NOT \(\xi_{R4}=0\) — \(\xi_{R4}\) lives in the twisted group; the twist is the only possible nonzero source.
• The concrete open step: evaluate the twisted \(d_5=Q_1\) correction on the degree-5 line — does the frozen twist \(\tau_{K_6}=(2,2)\) supply a nonzero degree-lowering image?
• Supporting frozen facts (given, not re-derived): \(b_2(K_6)=2\Rightarrow H^2(K_6;\mathbb Z_3)=(\mathbb Z/3)^2\); \(c_1(TK_6)=2\rho=(2,2)\not\equiv(0,0)\bmod3\Rightarrow\bar c_1\ne0\), \(\tau_{K_6}\) nonzero.

Track-A reviewer verdict: A5_verdict = R4_NOT_WELL_POSED_AS_STATED on the original symbolic target — the original \(\bar c_1(L_{K_6})\times B^{(2)}\) was degree-4 on the wrong locus (\(K_6\times M_4\), not a degree-5 class on the bordism \(W^5\)) with a malformed \(d_3\) differential. Corrections carried forward (do not revive the malformed version): (i) not a degree-4 cup product; (ii) not \(d_3=\beta\mathrm{Sq}^2\) (2-primary, annihilates 3-torsion — the earlier "\(d_3=\mathrm{Sq}^3_{\mathbb Z}\)" heuristic is wrong); (iii) operative differential is \(d_5=Q_1\); (iv) the \(c_1(L_{K_6})\) representative (\(3\) vs \(\bar3\)) is a single owner-gated orientation bit, earned-irreducible under the index data \(\chi(K_6,E)=-3\), not derivable.

Model-independent invariant carried into the global center/anomaly chain (§10 of the geometry pack): 2-primary differentials cannot touch 3-torsion; the operative mod-3 \(d_5=Q_1\) annihilates \(u_2\) (since \(Q_1(2y_1+2y_2)=0\)) \(\Rightarrow u_2\) survives both primary differentials on the center \(\Rightarrow\) \(\xi_{R4}\) is EXPECTED NONZERO by this survival argument. This is a tilt, not a computed value — the twisted \(d_5\) correction on the degree-5 line itself remains uncomputed.

Conditional three-outcome verdict tree (none assertable until \(\xi_{R4}\) is computed):

Honest prior: tilts AWAY from (b) — the compact boundary/parity sector is on record as adding difficulty (inheriting all 4D difficulty plus the compact-boundary/parity sector's), not supplying coercivity. A nonzero \(\xi_{R4}\) would be the surprising outcome. Tractability horizon: 5–15 years.

R5 (the necessary-not-sufficient caveat, DERIVED, not standalone open work): even a nonzero \(\xi_{R4}\) does not imply the gap — anomaly-matching admits gapless conformal/TQFT saturation of the 't Hooft condition. If R4 is nonzero and universality-preserving, a separate proof that the matched IR phase is gapped/confining is still required. Falsifier: a gapless anomaly-matching phase confirms R4 is not a gap handle.

Provenance flag: the R4 computation package carries a \(K_6\) holonomy phase \(\Phi\approx-0.12827\) rad as frozen boundary \(\theta\)-data (never a computed anomaly output) — this is NOT the same object as the geometry pack's CKM holonomy \(\delta_{\rm CKM}=-2\pi/3\), lepton Berry phase \(+2\pi/3\), or modulus \(\tau=\omega=e^{2\pi i/3}\). Kept separate to avoid conflation.

L6 — The executed Monte Carlo diagnostic (the honest INCONCLUSIVE run — banked as a loss, not hidden)

\(D:=s_{\rm block}-\log(E_{\rm conn}\cdot A_{\rm fluc}\cdot N_{\rm cert})\), computed on a \(\beta=6.0\), \(4^4\) (\(L=4\)) ensemble, 16 configurations, plaquette-gated.

\(D<0\) robustly at every binning. Critical \(N_{\rm cert}\) for \(D=0\) is \(2.83\) letters, versus realized \(N_{\rm eff}=155\).

Why this is INCONCLUSIVE, not a refutation: this diagnostic tests only a sufficient KP proxy, \(z_\star\le N_{\rm cert}\max_\gamma w(\gamma)\) — a worst-case upper bound that over-counts. \(D\le0\) triggers a bounded block-spin (Bałaban-RG) fallback re-evaluation at coarser \(\ell_\star\), not a refutation of the target. A genuine refutation requires a direct lower bound \(z_\star\ge1/(E_{\rm conn}A_{\rm fluc})\), which was not computed.

The convention-underdetermination of \(z_\star\) (do not pick a reading to manufacture a verdict):

Picking (B)/single to claim PASS, or (A) to claim FAIL, is reverse-engineering from the measured value — exactly the target-fitting trap this ledger forbids. Until \(w(\gamma)\) is derived target-blind, \(z_\star\) is not decidable. Known internal inconsistency flagged honestly: \(E_{\rm conn}=e\cdot7\) is pure \(\mathbb Z^4\) graph combinatorics carrying no per-block letter count, so the boxed inequality does not itself specify whether the multiplicity lives inside \(z_\star\) or is dropped.

Plaquette precondition hard-gate (passed, with a flagged source discrepancy): \(\langle P\rangle=0.59375\pm0.00197\) (2026-07-04 handoff) vs \(\langle P\rangle=0.59639\pm0.00219\) (2026-07-02 completion-run BUILDER), against standard \(\beta=6.0\) value \(0.5937\). Both agree with the standard value at \(\lesssim1\)–\(1.3\sigma\) and both pass the gate; the two runs differ at the third decimal, stated here rather than silently reconciled. No third number invented.

Retired cautionary precedents (never banked, carried only as discipline reminders): \(\kappa^3/\pi\) and \(5+3=8\) — seductive \(O(1)\) coincidences a no-target-fitting guard caught and rejected. An \(O(1)\) ratio landing near an integer is not a verdict.

L7 — Credit-ladder grading summary (every leg, one table)

L8 — Anti-claims and negative controls

[RETAINED — rerouted 2026-07-12: these bright lines stay live and now bound the continuum \(P_\star\) / Clay object. The first line is scoped: the CONTINUUM Clay mass gap is NOT solved/derived; the FINITE operational SU(3)-lattice gap IS derived (certificate PASS), which is a different, narrower object. See Governing Correction.]

Forbidden overclaims (bright lines, actively denied by the frozen record): - The continuum Clay mass gap is NOT "solved / closed / proven / derived" — anywhere in this document. (The finite operational-lattice gap is derived; that is a strictly weaker, scoped object — see Governing Correction.) - The conditional lemma \([H1\wedge H2\wedge H3\wedge H4]\Rightarrow\) gap is NOT an unconditional gap proof. - The granularity dissolution of Face A does NOT answer the Clay question — it changes which question is finite. Dissolved ≠ solved. - REDUCED-TO-AXIOM did NOT lower the axiom count (the basin-shallowing countermodel is unrefuted). - No value of \(\xi_{R4}\) is reported; \(\xi_{R4}=0\) is NOT claimed; no claim that any \(\xi_{R4}\) value forces the gap. - No \(z_\star\) reading is picked to manufacture PASS or FAIL; the INCONCLUSIVE MC run is not called a refutation. - Gap-02 is NOT re-labeled a "structural frontier" — that term is reserved for Gap-13 (BH entropy) and UQF-4 (global anomalies), a different bucket from a recorded external Clay wall.

Negative controls (kept live, never dissolved): - The Clay problem itself: unsolved here and everywhere; this document produces no proof of it. - The basin-shallowing countermodel against a uniform \(\Delta_0>0\): \(d_n=B\cdot2^{-n-1}\), \(\sum d_n=B/2<\infty\), \(\inf_n d_n=0\) — stands unrefuted, keeping R6 an axiom rather than a theorem. - The MC diagnostic \(D<0\) at every binning: a real, robust (\(\sim66\sigma\)) negative signal on the sufficient proxy, explicitly not extended into a refutation of the necessary target it does not test. - Reading (A) of \(z_\star\) diverges and reads FAIL: retained in the table, not suppressed, precisely because suppressing an unfavorable reading would be target-fitting. - \(\kappa^3/\pi\) and \(5+3=8\): retired numerological near-misses, kept on record as a discipline reminder, never banked as physics.

Unicorns dissolved as limits on all knowledge (not gaps in this framework): - "No future theory could ever solve Clay by a different route" — an unprovable universal negative; the honest bounded claim is "no accepted constructive proof exists today, community-wide." - "The granularity axiom is THE provably-irreducible bottom" — undischargeable; the honest bounded claim is "earned-irreducible under known reductions, a confessed posit." - "No observation-preserving modification could ever supply a lever" — the honest bounded claim is Lemma 1: on every settleable row, none; R4 is the one honest unsettled row.

L9 — The endpoint line

Gap-02 reduces the framework's Yang–Mills question exactly to the standing Clay mass-gap problem — no more, no less. It proves, as a real theorem (Lemma 1), that its own frozen 13D geometry supplies no shortcut around that problem. It takes the measured gap value and the fixed scale \(\Lambda_{\rm YM}\) as anchors, never as outputs. It localizes the entire remaining continuum obstruction to one finite inequality (\(\rho_\star<1\) / the master inequality) plus two independent co-gates (R2, R3) plus one low-leverage framework-internal side channel (R4). It reports every numerical verdict — including the INCONCLUSIVE MC run and the convention-dependent \(z_\star\) readings — at exactly the scope each establishes, no wider.

Terminal: CERTIFIED-IRREDUCIBLE, external wall recorded. Residual: the open Clay problem, shown, not hidden, never rolled into a hedge on the gate's own status.

[SUPERSEDED 2026-07-12 — governing terminal: CLOSED-SCOPED. The gate DERIVES a positive spectral gap for each finite operational SU(3) lattice (certificate PASS); the terminal line above (external-wall-recorded, no derivation) is RETAINED-rerouted to the continuum-and-infinite-volume uniform gap \(P_\star\), which stays CERTIFIED-IRREDUCIBLE and NOT-A-CLAY-PROOF. See Governing Correction.]

Appendix A — Certificate source (runnable / PASS)

gap02_finite_lattice_gap_certificate.py — ran clean under python on 2026-07-12; exit status PASS. SU(3) one-plaquette character-basis Hamiltonian; all finite witness gaps positive across cutoff \(P=3\ldots8\) and \(\lambda=0.1,0.3,1.0\).

#!/usr/bin/env python3
import numpy as np, json
from pathlib import Path

def reps(P): return [(p,q) for p in range(P+1) for q in range(P+1-p)]
def c2(r):
 p,q=r; return (p*p+q*q+p*q+3*p+3*q)/3.0
def fund_products(r):
 p,q=r; out=[]
 for x in [(p+1,q),(p-1,q+1),(p,q-1)]:
  if min(x)>=0: out.append(x)
 return out
def antifund_products(r):
 p,q=r; out=[]
 for x in [(p,q+1),(p+1,q-1),(p-1,q)]:
  if min(x)>=0: out.append(x)
 return out
def hamiltonian(P,lam,kappa=1.0):
 R=reps(P); idx={r:i for i,r in enumerate(R)}; n=len(R)
 H=np.diag([kappa*c2(r) for r in R])
 A=np.zeros((n,n))
 for r in R:
  i=idx[r]
  for s in fund_products(r)+antifund_products(r):
   if s in idx: A[idx[s],i]+=1.0
 # multiplication by chi3+chibar3 is Hermitian; numerical symmetrization guards truncation boundary
 A=(A+A.T)/2
 H=H-lam*A
 return R,H
records=[]
for lam in [0.1,0.3,1.0]:
 for P in [3,4,5,6,7,8]:
  R,H=hamiltonian(P,lam)
  assert np.max(np.abs(H-H.T))<1e-12
  ev=np.linalg.eigvalsh(H)
  gap=ev[1]-ev[0]
  assert gap>1e-8
  records.append({'lambda':lam,'cutoff_P':P,'dimension':len(R),'E0':float(ev[0]),'E1':float(ev[1]),'gap':float(gap)})
# finite gaps need not imply a uniform limit: explicit positive sequence tending to zero
seq=[1/n for n in range(1,100)]
assert all(x>0 for x in seq) and min(seq)<0.02
out={'status':'PASS','witness':'SU3 one-plaquette character-basis Hamiltonian','records':records,'all_finite_witness_gaps_positive':True,'continuum_uniform_gap_claimed':False,'positive_sequence_negative_control_last':seq[-1]}
Path(__file__).with_name('gap02_finite_lattice_gap_certificate_output.json').write_text(json.dumps(out,indent=2))
print(json.dumps(out,indent=2))

Appendix B — Certificate output (runnable / PASS)

Observed stdout (abridged to endpoints; gaps converge with cutoff):

{
  "status": "PASS",
  "witness": "SU3 one-plaquette character-basis Hamiltonian",
  "records": [
    { "lambda": 0.1, "cutoff_P": 3, "dimension": 10, "E0": -0.016136625715826, "E1": 1.232239330160887, "gap": 1.2483759558767134 },
    { "lambda": 0.1, "cutoff_P": 8, "dimension": 45, "E0": -0.016136625939011, "E1": 1.232239228246512, "gap": 1.2483758541855228 },
    { "lambda": 0.3, "cutoff_P": 3, "dimension": 10, "E0": -0.163495803821912, "E1": 1.037696965907747, "gap": 1.2011927697296594 },
    { "lambda": 0.3, "cutoff_P": 8, "dimension": 45, "E0": -0.163497834651827, "E1": 1.037618967374169, "gap": 1.2011168020259961 },
    { "lambda": 1.0, "cutoff_P": 3, "dimension": 10, "E0": -1.794695198880992, "E1": 0.316870487797537, "gap": 2.1115656866785297 },
    { "lambda": 1.0, "cutoff_P": 8, "dimension": 45, "E0": -1.803419848420028, "E1": 0.249475487295501, "gap": 2.052895335715529 }
  ],
  "all_finite_witness_gaps_positive": true,
  "continuum_uniform_gap_claimed": false,
  "positive_sequence_negative_control_last": 0.010101010101010102
}

Verdict: runnable / PASS. Every finite witness gap is strictly positive and converges with cutoff; the uniform-limit claim is explicitly withheld (continuum_uniform_gap_claimed=false); the \(1/n\) negative control confirms finite positivity does not entail a uniform infimum. Independently verified: the SU(3) \(C_2\) Casimir formula and the fundamental / antifundamental tensor-product rules (\(3\otimes3=6\oplus\bar3\), \(8\otimes3=15\oplus6\oplus3\)).