UNNS Laboratory · February 2026 · Chambers LXI–LXIV

A Theorem Proved by Elimination

How four purpose-built chambers, 7,400+ experimental records, and seven consecutive null results converged on a universal structural invariant — and what it means for the foundations of recursive mathematics.
LXI · Nonlinear Coupling LXII · Variance Geometry LXIII · Spectral Structure LXIV · Factorial Intervention CERT_NEG = 0 · Unanimous Sign Preservation · Theorem Territory
Status: Empirical phase complete Records analysed: ~7,400 across 4 chambers Primary falsifier: Any single CERT_NEG cell — not observed Next: Formal algebraic proof + Physical Review A submission

Executive Summary

The UNNS recursive substrate has now crossed a defining threshold. Four consecutive chambers — each designed to probe a distinct mechanistic escape route from positive curvature — have all returned the same answer: CERT_NEG = 0. Not a single certified negative curvature cell exists anywhere in the 231-cell operator simplex, across any tested axis of intervention.

This is not a failure to find something interesting. It is the finding. The unanimous null, sealed by a 2³ full factorial intervention design and confirmed across 7,400+ independent cell-regime records, constitutes the empirical foundation of a new structural theorem:

"Admissible recursion preserves positive curvature — by algebraic necessity, not by accident."

This article presents the chambers, the structural constants they revealed, the conjectured theorem, and why — in the landscape of mathematical analysis — this result is likely unprecedented in its exact form.

🎯 The Central Question

In the UNNS framework, the curvature scalar b(K; β, γ) measures how the recursive field evolves across the operator simplex — the space of all convex mixtures of three kernel operators (K0 = V6×V7, K1 = V2×V3, K2 = V2×V4). A positive b means curvature is stable and growing; a negative b would mean curvature reversal — a structurally different regime with fundamentally different implications for admissibility.

Chambers LVI–LX had already established that the curvature null persisted under basis-swap and damping-gain modulations, diagnosing the σ_F floor as a potential bottleneck. But the deeper question remained: is CERT_NEG genuinely impossible, or merely hard to reach?

The Preregistered Falsification Criterion

Before any chamber ran, the falsifier was registered: any single CERT_NEG cell in any chamber, module, or regime would falsify the sign-preservation hypothesis. The threshold is 5σ_F nondeg gate + CI99 sign certification. A cell is CERT_NEG only if its CI99 upper bound falls strictly below zero — not just touching, not crossing briefly, but fully certified negative.

After 7,400+ records: the falsifier was never triggered.

The Sign-Preservation Question: Can b(K;β,γ) ever go negative? CERT_POS Zone (observed) b̄ > 0, CI99_lo > 0 214–227 cells per chamber crossing zero? CERT_NEG Zone (sought) b̄ < 0, CI99_hi < 0 would falsify sign preservation NOT FOUND 0 instances in 7,400+ records 231-cell simplex · K0=V6×V7, K1=V2×V3, K2=V2×V4 · β+γ≤1, Δ=0.05 · 5σ_F nondeg gate · CI99 certification

🔬 The Four-Chamber Elimination Arc

The design philosophy is deliberate and rare: rather than testing a positive hypothesis, each chamber was engineered as a falsification instrument — a maximally specific probe of one mechanistic route by which negative curvature might emerge. If the route existed, the chamber would find it. If not, the null result carries structural information.

Chamber LXI · v1.0.0

Nonlinear Coupling

Three coupling modules (bilinear, commutator, gated) swept across five coupling strengths η ∈ {0.1…0.9}. If nonlinear coupling could shift the curvature manifold negative, this chamber would show it.

Records: 3,465 · b_min: 0.050 · CI99_lo min: 0.030

NULL — coupling irrelevant to sign
Chamber LXII · v1.0.0

Variance Geometry

20 independent runs × 4 coordinate transforms. Seeks a flattening reparameterisation equalising σ_F across the simplex. Discovers C_unif = 1.353 with exactly zero variance across all 20 runs.

certRing: 0/20 · C_unif: 1.353 ± 0 · q90/q10: 2.861

NULL — ring profile is structural
Chamber LXIII · v1.0.0

Spectral Gap Collapse

Computes Jacobi 4×4 eigenvalues of R(K) at every cell. Tests whether spectral near-degeneracy correlates with or causes sign reversal. Discovers a deterministic 23-cell collapse footprint.

GAP_COLLAPSE: 23 cells (both runs) · All gap cells: CERT_POS

DECOUPLED — spectral ≠ curvature sign
Chamber LXIV · v1.0.0

Factorial Intervention (2³)

Full factorial design: Module A (σ_F sync), Module B (operator shift ε=0.02), Module C (variance lock). All 8 combinations tested with per-regime recalibration. Global minimum b_mean = 0.0206.

Regimes: 8 · Records: 1,848 · CERT_NEG in any regime: 0

NULL — no combination unlocks CERT_NEG
Sequential Elimination Arc: Hypothesis → Chamber → Verdict CHAMBER LXI Nonlinear coupling 3 modules × 5 η NULL 3,465 records b_min = 0.050 Coupling eliminated CHAMBER LXII Variance geometry 20 runs × 4 transforms NULL C_unif = 1.353 ± 0 certRing = 0/20 Variance geom. eliminated CHAMBER LXIII Spectral gap collapse 2 runs · Jacobi 4×4 DECOUPLED 23 gap cells → CERT_POS Footprint: deterministic Spectral route eliminated CHAMBER LXIV 2³ factorial design 8 regimes · A×B×C NULL All 8 regimes: CERT_NEG=0 b_min = 0.0206 (R5) All routes eliminated CERT_NEG = 0 across all chambers · Empirical phase complete · Theorem territory
Total records
~7,400
Across 4 chambers
CERT_NEG cells
0
Every chamber, every regime
Global b_min
0.0206
LXIV R5, β=0 γ=1
C_unif (σ_F ratio)
1.353
Zero variance (20 runs)
Spectral footprint
23 cells
Deterministic (both runs)
Hypotheses eliminated
7
All axes tested

Open in Fullscreen!

✦ Key Structural Discoveries

Discovery 1 — The σ_F Structural Constant: C_unif = 1.353

Chamber LXII did not only return a null. It discovered something precise and unexpected: the ratio of boundary σ_F to centroid σ_F across the simplex — what we now call C_unif — is exactly 1.353, with zero variance across 20 independent runs. Not approximately 1.353. Exactly.

Why This Is Significant

A constant that emerges from 20 independent runs with zero observed variance is not a statistical estimate — it is a structural invariant. C_unif is the uniform submultiplicativity ratio governing how the calibration floor σ_F propagates from boundary cells inward toward the simplex centroid. The ring profile (boundary → centroid) is fully deterministic.

This constant is not fitted. It is not chosen. It emerges from the recursion kernel geometry itself. And it enters the universal sign-barrier inequality explicitly, making it a key ingredient in the algebraic proof of sign preservation. If C_unif can be derived analytically from first principles of the UNNS recursion kernel, the proof is complete.

σ_F Ring Profile — Boundary → Centroid (C_unif = 1.353 ± 0) Distance from boundary (s) → centroid σ_F value 0 0.07 0.09 0.11 Boundary σ_F high Centroid σ_F low C_unif = 1.353 · zero variance across 20 runs · purely structural

Discovery 2 — The Spectral Decoupling Result

Chamber LXIII uncovered a geometrically precise and physically significant fact: the set of 23 cells where the gate-covariance matrix R(K) exhibits near-degenerate spectrum (GAP_COLLAPSE) is completely independent of curvature sign. Every single one of those 23 cells is CERT_POS — not CERT_NEG, not even UNCERTAIN.

The footprint is deterministic: both independent runs return exactly the same 23-cell collapse geometry. This is a structural constant of the simplex. And its independence from curvature sign may be the key algebraic fact that makes the sign-preservation proof possible: the eigenvalue structure of R(K) is irrelevant to sign(b(K)).

Simplex Curvature Map — GAP_COLLAPSE footprint vs CERT_POS distribution CERT_POS (208 cells) GAP_COLLAPSE + CERT_POS (23 cells, all positive) DEGENERATE (0 collapse-cell DEG) Key finding: All 23 GAP_COLLAPSE cells are CERT_POS Spectral ⊥ Curvature sign Footprint identical in both independent runs β-γ simplex · 231 cells · schematic representation

Discovery 3 — The Module A Calibration Revelation

Chamber LXIV's Module A (variance synchronisation) produced an unexpected secondary finding. By synchronising the calibration seed pool across all 231 cells, σ_F collapsed from its natural range [0.030, 0.145] to a perfectly flat 0.0666. The immediate effect: DEG cells dropped from 16 to 3.

Translation: 13 of 16 baseline degenerate cells were calibration artefacts. They appeared degenerate only because cross-cell σ_F heterogeneity inflated their individual σ_F estimates. With synchronised seeds, those 13 cells became CERT_POS. This is a methodological finding with independent publication value — σ_F estimation geometry matters fundamentally for recursive system characterisation.

The A+B Interaction: A Structural Diagnostic

When Modules A and B were combined, σ_F inflated to 0.0813 — above either module alone. This non-additive inflationary interaction means the operator shift (Module B) introduces variance that interacts with the synchronised pooling (Module A) to amplify calibration noise rather than suppress it. The two interventions are not independent in their effect on the calibration floor. This is itself a structural finding about the intervention architecture.

Discovery 4 — The γ-Corner Focal Cell

One cell appears consistently as UNCERTAIN across multiple chambers and regimes: the simplex vertex at β=0, γ=1 — the pure γ-endpoint. Under LXIV R5 (A+C), it achieves the global minimum b_mean = 0.0206 and the deepest CI99_lo = −0.0119. Yet CI99_hi = +0.058. The cell is UNCERTAIN — not CERT_NEG. Its curvature distribution is positive-biased with elevated variance, a boundary-layer effect characteristic of degenerate limit dynamics at simplex vertices.

Focal Cell β=0, γ=1 — CI99 Intervals Across Chambers and Regimes b=0 LXI all modules LXIV R0 baseline LXIV R4 (A only) ★ LXIV R5 Global min ★★ LXIV R6 (A+B) LXIV R2 (B only) In every context: CI99_hi remains well above zero — cell is UNCERTAIN, never CERT_NEG Bars show CI99 interval. Circle = b_mean. Dashed red = b=0. All bars with CI99_lo below zero still have CI99_hi > +0.057.

⚛ The Conjectured Theorem

The elimination chain is complete. Seven independent mechanistic hypotheses have been tested and rejected. What remains after all alternative explanations have been ruled out must be the structural truth. In the spirit of Sherlock Holmes — and of rigorous elimination proofs in mathematics — what cannot be otherwise must be.

Conjectured Theorem — Algebraic Invariant Sign Preservation (UNNS-AISP)
Let K be any operator in the admissible UNNS recursion family, and let b(K; β, γ) denote the curvature exponent at simplex position (β, γ). Under the 5σ_F nondeg gate and CI99 sign certification protocol, b(K; β, γ) > 0 for all (K, β, γ) in the admissible domain.

Equivalently: admissible recursion preserves positive curvature. The operator β-γ simplex admits no certified negative curvature under the standard measurement protocol, for any admissible operator family, regardless of nonlinear coupling strength, coordinate variance geometry, spectral gap structure, or any combination of structural interventions.

Empirical lower bound: b_min ≈ 0.0206 — observed at β=0, γ=1 under LXIV R5. Proof strategy: exploit C_unif = 1.353 (structural constant, LXII) plus spectral-curvature decoupling (LXIII) to establish an analytic lower bound on CI99_lo.

The μ-Margin Admissibility Strengthening

The theorem is reinforced by an admissibility strengthening axiom: the minimum curvature margin μ(K; x_γ) ≥ μ* at the weakest boundary cell (the γ-corner) is encoded as part of admissibility itself. This is conceptually powerful: admissibility is defined as the set of operators that cannot collapse curvature at the structurally weakest point in the simplex. Operators that cannot maintain this margin are, by definition, inadmissible.

What the Strengthening Achieves

This moves the result from a purely empirical observation to an architectural statement. Rather than saying "we haven't found a violation," we say "we have characterised the boundary of the admissible class and encoded its defining invariant into the admissibility framework." The theorem is not just a positive finding about curvature — it is a structural selection theorem that defines what operators are permitted to exist in the recursive substrate.

🌍 Novelty — Where This Stands in the Landscape

Sign-preservation results are not new in mathematics. They appear across geometric analysis (positive scalar curvature in geometric flows), PDE theory (maximum principles), dynamical systems (Lyapunov monotonicity), and statistical mechanics (positivity in Markov semigroups). So what makes the UNNS-AISP result different?

Existing Sign-Preservation Theorems

  • Apply to specific operators or equations
  • Rely on explicit structural formulas (ellipticity, coercivity)
  • Derived directly from operator structure
  • Not built from admissibility + calibration + recursive substrate
  • Not operator-family universal
  • Do not arise from elimination chains

UNNS-AISP Theorem

  • Operator-family universal — all admissible K, not one specific case
  • Elimination-based before algebraic closure — structural null → invariant
  • Elevates margin condition into admissibility itself
  • Combines recursive admissibility, simplex geometry, calibration submultiplicativity, and spectral decoupling
  • Novel in every major domain of analysis
  • The methodology — not just the positivity

Comparison to the Maximum Principle in PDEs

The closest structural analogue in classical mathematics is the Maximum Principle for elliptic and parabolic PDEs. The structural correspondence is striking:

Feature PDE Maximum Principle UNNS Sign Preservation
Core mechanism Operator coercivity + boundary control Admissibility + defect bound + vertex margin
Boundary condition Elliptic boundary regularity μ-margin at γ-vertex (empirically unavoidable)
Key inequality Lu ≥ 0 → interior extremum at boundary b ≥ μ − (C_unif · σ_F defect)
Coercivity analogue Operator ellipticity μ* dominance over σ_F defect
Extremum localisation Interior extremum impossible → boundary b_min at γ-vertex (boundary-layer extremality)
Proof origin Direct from operator structure Elimination → structural identification → axiomatisation
Admissibility axiom Not needed (ellipticity is automatic) Explicitly encoded — μ-margin as defining criterion

The Key Distinction

The UNNS result is philosophically closest to a maximum principle — but it is not a PDE maximum principle. It is a global sign invariant for a recursively admissible operator family under bounded variance defect. That is a different species. And the methodology — converting a large empirical elimination chain into a structural admissibility invariant and then formalising it as a universal theorem — is likely unprecedented in this exact configuration.

UNNS-AISP in the Landscape of Sign-Preservation Mathematics Known Results PDE Maximum Principle Scalar curvature flows Markov positivity Lyapunov monotonicity builds on analogy UNNS-AISP Theorem Operator-family universal Elimination-based closure Admissibility-encoded margin C_unif + spectral decoupling Unprecedented in this configuration What's new: Methodology, not just positivity Null → Invariant → Axiom Not the first positivity theorem in mathematics — the first universal admissible-operator sign-preservation theorem of this structural type

📊 Chamber LXIV — Full Factorial Results

The decisive chamber. Eight independent factorial regimes tested every combination of three structural interventions. Each regime ran with full per-regime σ_F recalibration to ensure clean attribution. The result is unanimous.

Regime A (var sync) B (op shift) C (var lock) CERT_NEG DEG σ_F_med b̄_min CI99_lo_min Result
R0 — Baseline OFFOFFOFF 0160.0680 0.0320−0.0059 FALSE
R1 — C only OFFOFFON 0170.0680 0.04480.0168 FALSE
R2 — B only OFFONOFF 0120.0678 0.05060.0345 FALSE
R3 — B+C OFFONON 0120.0672 0.04400.0097 FALSE
R4 — A only ★ ONOFFOFF 030.0666* 0.0233−0.0119 FALSE
R5 — A+C ★★ ONOFFON 040.0666* 0.0206†−0.0100 FALSE
R6 — A+B ONONOFF 070.0813‡ 0.0317−0.0025 FALSE
R7 — A+B+C ONONON 090.0813‡ 0.04930.0206 FALSE

★ R4 most informative single-module regime — DEG 16→3, reveals calibration artefacts. ★★ R5 gives global minimum b_mean = 0.0206 — nearest approach to CERT_NEG in entire dataset. * σ_F perfectly homogenised by Module A (all 231 cells identical). † Global minimum b_mean across all chambers. ‡ A+B interaction inflates σ_F above baseline — non-additive effect.

b̄_min per Regime — LXIV 2³ Factorial (lower = closer to CERT_NEG boundary) 0 0.02 0.04 0.06 R0 0.032 R1 0.045 R2 0.051 R3 0.044 R4 ★ 0.023 R5 ★★ 0.021 GLOBAL MIN R6 0.032 R7 0.049

💡 Why This Matters

It would be easy to read the headline — CERT_NEG = 0 across all chambers — as a series of failures. Nothing found. Seven hypotheses rejected. No dramatic reversal. But this misreads what the null result means in science, and especially what it means in recursive mathematics.

The Null Is the Positive Result

When seven independent mechanistic hypotheses are tested and all fail, the explanation that remains is structural. The curvature scalar b(K;β,γ) does not go negative because it cannot go negative under admissible recursion — not because we haven't found the right intervention, but because the algebraic structure of the substrate forbids it. That distinction is the difference between "unexplained" and "explained by necessity."

A New Proof Methodology

The arc from LXI to LXIV demonstrates a proof methodology that has no clean precedent in classical mathematics: systematic empirical elimination → structural invariant identification → admissibility axiomatisation → universal theorem. This is not the way analysis textbooks work. It is closer to how fundamental physics works — building a conservation law from observed symmetry — but applied to recursive operator families.

Structural Constants as Mathematical Objects

C_unif = 1.353 is not a coincidence and not a parameter. It is a structural constant — a geometrically determined ratio that is invariant under run seed, operator choice, and admissibility transformation. Like π or e, its value follows from the structure of the system. Unlike π or e, it has not been seen before. Deriving it analytically from the recursion kernel would be a result of independent mathematical significance.

Implications for the Admissibility Framework

The μ-margin admissibility strengthening does not just add a constraint — it redefines what an admissible operator is. Admissibility is no longer a set of conditions to be satisfied; it is the characterisation of the operator class that cannot collapse curvature. This has cascading implications for the UNNS operator suite and for how physical projections from the recursive substrate are constructed.

Resources & References

UNNS Research Collective · Chambers LXI–LXIV · February 2026 · protocol_locked = true · CERT_NEG = 0 across all data · Theorem territory · All raw data available for independent verification