UNNS Research Collective · February 2026 · Theoretical Backbone

Structure Above Constants
The Backbone Theorem

How a single architectural insight — separating structural necessity from numeric certification — transformed the LXI–LXIV elimination programme into a constant-agnostic universal theorem, and elevated the UNNS dispersion functional from a measurement tool into a quasi-subadditive potential.
Chamber LXII · Empirical Origin LXI–LXIV Arc · Theorem Layer Log-Potential ψ = log σ_F Quasi-Subadditivity · Fekete-Type Asymptotic Depth Exponent · Proven Companion to Duality Theorem
Document type: Foundational admissibility paper Empirical source: Chamber LXII (C₄₂₂ discovery) Theorem scope: All admissible (K, x) — constant-agnostic Open questions: Analytic derivation of C_*, μ-margin from first principles
PDF
Backbone Paper — Full Text
Admissibility, Depth Submultiplicativity, and Universal Sign Preservation
UNNS Research Collective · February 2026 · Theoretical backbone of the LXI–LXIV arc
↓ Download PDF unns.tech/media/chamber_lxii/

Executive Summary

This paper is not a chamber report. It sits above the chambers. Born from the structural constant discovered in Chamber LXII and generalising across the LXI–LXIV elimination arc, it extracts an invariant that was always implicit in the empirical work — but had never been stated independently of any specific numeric value.

The central result: under depth submultiplicativity and a positive margin condition, no admissible operator can certify persistent negative curvature — by algebraic necessity, not by luck. The theorem is constant-agnostic: it requires only the existence of some C* > 1, not its exact value. The UNNS empirical chambers then provide certification that such a constant exists, rather than serving as the proof itself.

"The chambers discovered the structure. This paper names it."

🏛 Where This Paper Sits

The UNNS theoretical hierarchy now has three clearly separated layers. Understanding which layer each document occupies is essential to reading them correctly.

UNNS Theoretical Hierarchy LAYER I — FOUNDATIONAL BACKBONE Admissibility, Depth Submultiplicativity & Universal Sign Preservation (this paper) + The UNNS Observability–Admissibility Duality Theorem Pure theory Constant-agnostic LAYER II — ELIMINATION ARC (LXI–LXIV) LXI Nonlinear Coupling → LXII Variance Geometry → LXIII Spectral Decoupling → LXIV Factorial Discovers C_* > 1 empirically · Certifies sign preservation · 7,400+ cell-regime records Empirical + partial theory scaffold LAYER III — CHAMBER SUBSTRATE (XII–LXIV) Raw experimental data · Operator simulations · Grid explorations · Certification source Pure empirical LXII → Backbone abstraction

Primary Structural Alignment: Chamber LXII

The backbone paper formalises exactly what Chamber LXII discovered empirically: the C₄₂₂ ratio, C_med and C_q95 measurements, and the submultiplicativity envelope. If someone asks "where did this theory originate?" the correct answer is Chamber LXII. The backbone paper abstracts LXII into pure structure.

Paper Empirical? Theoretical? Chamber Link Role
Chamber LXII Yes — variance geometry Partial structure Source of C_* Discovers the constant
This Backbone Paper No Pure structural Formalises LXII Extracts the invariant
LXI–LXIV Arc Yes — elimination Mixed scaffold Uses C_* Eliminates mechanisms
Duality Theorem No Pure theoretical XL, XXXI, XXXII Observability principle

✦ Three Architectural Upgrades

The paper delivers three distinct structural advances, each addressing a fragility in how sign preservation was previously understood. Together they move the result from chamber-dependent empirics into the language of pure operator theory.

Upgrade A — Structural Independence from Numeric Constants

Before

  • Sign preservation depended on chamber constants (1.353, 2.136)
  • C* entangled with structural claims
  • Theorem fragile if a constant shifted
  • "It works because the constant equals 2.1363"

After

  • Universal Sign Preservation Theorem is constant-agnostic
  • Only existence of C* > 1 required
  • Empirical constants are certification, not foundations
  • "It works because depth submultiplicativity + margin imply it"

Upgrade B — Precise Definitions of σ_F and L(k)

Previously, the dispersion functional σ_F was described in chamber prose — understood empirically, but never formally specified. The backbone paper gives it a precise mathematical definition that any external reader can evaluate without access to chamber implementation details.

The Operational Chain (now fully explicit)

Step 1. For each seed ω, compute the depth-n sensitivity envelope L_n(K; x; ω) — the maximum absolute gate-to-gate slope from the log-response profile.

Step 2. Fit the OLS slope of log L_n across depth n → this gives curvature estimator b̂_n(K; x).

Step 3. σ_F(n; K, x) := Std_ω(b̂_n) — the seed-to-seed standard deviation over the admissible seed pool.

Upgrade C — Log-Potential and Asymptotic Exponent

The deepest structural move: defining ψ_n = log σ_F and proving that Depth Submultiplicativity implies quasi-subadditivity of ψ with a fixed additive defect. This connects the paper to classical Fekete-type analysis and yields the existence of a well-defined asymptotic depth exponent.

📐 The Five Admissibility Axioms

The paper identifies the minimal structural conditions sufficient to prove universal sign preservation. Five axioms characterise the admissible operator class — no fewer are needed, and none is redundant.

Axiom 1 — Deterministic Kernel Law
For each admissible K and control parameter τ, there exists a deterministic update map T_τ : M → M which is Lipschitz in τ. This is the regularity floor — no chaotic kernel behaviour.
Axiom 2 — Convex Closure
If K₁, K₂ ∈ A and θ ∈ [0, 1], then θK₁ + (1−θ)K₂ ∈ A. The admissible class is convex — mixtures of valid operators remain valid. This is what makes the asymptotic exponent convex in the operator (Lemma 5).
Axiom 3 — Margin Propagation (depth 2 → 4)
There exists κ > 0 such that b₄(K; x) ≥ κ · b₂(K; x) for all admissible (K, x). Positive curvature at depth 2 propagates with factor κ to depth 4.
Axiom 4 — Margin Dominance
There exists μ > 0 such that E[b₂(K; x)] ≥ μ for all admissible (K, x). The depth-2 curvature has a strictly positive expected value — the fundamental positivity condition.
Axiom 5 — Depth Submultiplicativity
There exists C* > 1 such that σ_F(4; K, x) ≤ C* · σ_F(2; K, x)² for all admissible (K, x). Dispersion at depth 4 is controlled quadratically by depth-2 dispersion. This is the core variance-control hypothesis — empirically certified by Chamber LXII (C_q95 = 2.1363 > 1).
Axiom Flow: How the Five Conditions Combine to Produce Sign Preservation Ax 1 Kernel Law Regularity Ax 2 Convex Closure Exponent convex Ax 4 Margin Dominance E[b₂] ≥ μ > 0 Ax 3 Margin Propagation b₄ ≥ κ · b₂ Ax 5 Depth Submultipl. σ_F(4) ≤ C*·σ_F(2)² Universal Sign Preservation No admissible (K,x) can certify persistent negative curvature

📊 Separating the Three Certification Constants

One of the paper's surgical fixes is ending a conflation that could have exposed the result to referee attack: the difference between a high-quantile envelope and a uniform bound. Three structurally distinct constants now have precise, separate identities.

C_med — Typical Behaviour
1.3529
Median of C₄₂₂ over the 231-cell simplex grid. Describes what a typical cell looks like. Not used to certify the theorem — used to characterise the distribution centre.
C_q95 — 95% Worst-Case Envelope
2.1363
95th percentile of C₄₂₂ over the grid. A high-quantile envelope: 95% of cells have σ_F(4) ≤ 2.1363 · σ_F(2)². Confirms C* > 1 with strong coverage. Not a uniform bound.
C_max — True Uniform Constant
max(C₄₂₂)
The true uniform grid constant: the maximum of C₄₂₂ over all cells. This is what Axiom 5 (Depth Submultiplicativity) structurally requires. C_q95 provides a lower bound on C_max sufficient to certify C* > 1.
C₄₂₂ Distribution Across 231-Cell Simplex Grid 1.0 1.2 1.4 1.6 1.8 2.0+ C₄₂₂ value C_med = 1.353 C_q95 = 2.136 (95th pct) 5% tail C_max →

Why This Distinction Matters

The theorem requires only existence of some C* > 1. The empirical data shows C_q95 = 2.1363 > 1, which implies C_max ≥ 2.1363 > 1. The structural result follows. No claim of global uniformity from a quantile is made.

〜 The Log-Potential Discovery

The deepest conceptual contribution of the paper is a reframing. Before, σ_F was treated as a nuisance statistic — something to control so the margin dominates. After the log-potential definition, σ_F becomes a structural potential governing admissibility.

Definition — Log-Potential
ψ_n(K, x) := log σ_F(n; K, x)

Under Depth Submultiplicativity, taking logarithms of σ_F(4) ≤ C* · σ_F(2)² yields:
ψ₄(K, x) ≤ 2 · ψ₂(K, x) + δ

where δ = log C* is a universal additive defect independent of (K, x).

From Multiplicative Inequality to Additive Potential Theory

This is the key move. The multiplicative submultiplicativity inequality becomes an additive quasi-subadditivity inequality on ψ. This connects the UNNS framework to classical potential theory and subadditive ergodic theory — specifically Fekete's lemma.

Log-Potential Quasi-Subadditivity: ψ_{2k} ≤ k · ψ₂ + (k−1)δ n=2 n=4 n=6 n=8 n=10 recursion depth 2n ψ_{2n} = log σ_F(2n) upper bound k·ψ₂+(k-1)δ ψ_{2n} (actual) slope = λ(K,x) defect δ Gap between bound and actual shrinks per unit depth → asymptotic exponent λ = lim ψ_{2n}/2n exists (Fekete-type)

Quasi-Subadditivity Lemma

With δ := log C*, we have ψ₄ ≤ 2ψ₂ + δ — a bounded additive defect from subadditivity. The iterated bound extends this to arbitrary even depth: ψ_{2k} ≤ k·ψ₂ + (k−1)δ, equivalently σ_F(2k) ≤ C*^{k−1} · σ_F(2)^k. Dispersion growth is multiplicative only up to a fixed exponential distortion factor.

Asymptotic Depth Exponent — Existence Proof

The quasi-subadditivity of ψ, combined with a Fekete-type argument, implies the existence of a well-defined asymptotic depth exponent:

λ(K, x) = limn→∞ ψ2n(K, x) / (2n) = infn≥1 [an(K,x) + δ] / n

The proof proceeds by defining b_n = a_n + δ (where a_n = ψ_{2n}), showing (b_n) is subadditive, applying Fekete's lemma to obtain the limit β = lim b_n/n, and noting that δ/n → 0 so the limit of a_n/n equals β. The asymptotic exponent λ = β/2 is the growth rate of log σ_F per unit recursion depth — a well-defined, finite quantity for every admissible (K, x).

Convexity of the Exponent Under Convex Closure

If the log-potential is convex under operator mixing (i.e. ψ_{2n}(θK₁ + (1−θ)K₂, x) ≤ θ·ψ_{2n}(K₁, x) + (1−θ)·ψ_{2n}(K₂, x) for all n), then λ(·, x) is convex in the operator. The proof divides by 2n and takes n → ∞: convexity is preserved under pointwise limits. This means no hidden corner of the convex admissible class can blow up depth growth.

✦ Key Findings and Discoveries

Structural upgrades
3
Each addresses a distinct fragility
Axioms required
5
Minimal sufficient set
C* requirement
> 1
Existence only — no exact value needed
C_q95 certified
2.136
95th pct · confirms C* > 1
Exponent λ(K,x)
Exists
Proven via Fekete-type argument
Additive defect δ
log C*
Universal — independent of (K,x)

Finding 1 — Depth Submultiplicativity ⟹ Structural Variance Control

This is the core finding. Axiom 5 implies three things simultaneously: no explosive variance cascade across recursion depths; controlled exponential distortion bounded by C*^{k−1}; and a universal additive defect δ = log C* governing the growth potential. The analogy to bounded curvature in geometric flows is legitimate — the recursion kernel imposes the same kind of large-scale regularity that Ricci flow curvature bounds impose on geometry.

Finding 2 — Margin + Submultiplicativity ⟹ CI Locking

The combination of Axioms 3–5 implies CI99 locking: under the condition μ₄ > 2.576 · C* · σ̄₂², the 99% confidence interval for b̂₄(K; x) lies strictly above zero for every admissible (K, x). This is not statistical luck. It is algebraically constrained — sign preservation is a necessary consequence of the admissibility structure, not an observation about a particular run.

Finding 3 — Convex Closure ⟹ Convex Asymptotic Exponent

Combining Axiom 2 (Convex Closure) with the log-potential convexity assumption yields a strong stability result: the asymptotic depth exponent λ is convex in the operator. No corner of the convex admissible class can have an anomalously large depth growth rate. The dispersion geometry is globally well-behaved.

Three Key Implication Chains Depth Submultiplicativity σ_F(4) ≤ C* · σ_F(2)² [Axiom 5] Controlled Variance No explosive cascade δ = log C* universal Margin + Submultipl. Axioms 3, 4, 5 combined μ₄ > 2.576·C*·σ̄₂² CI99 Locking CI99 lo > 0 guaranteed for all admissible (K,x) Convex Closure Axiom 2 + log-convexity of log-potential Convex Exponent λ(K,x) convex in K no corner blowup UNIVERSAL SIGN PRESERVATION THEOREM No admissible (K,x) can certify persistent negative curvature

🔭 Broader Theoretical Relations

The backbone paper's results connect to several established areas of mathematics — not by importing those frameworks, but by deriving parallel structures from first principles. Each analogy is legitimate; none is claimed as identity.

Relation to Fekete's Lemma / Subadditive Ergodic Theory

The existence proof for λ(K, x) mirrors Fekete's lemma exactly: a subadditive sequence b_n satisfies lim b_n/n = inf b_n/n. Here b_n = a_n + δ is made subadditive by the defect shift, and the limit exists by the classical argument. The novelty is that this structure arises from a dispersion functional on a recursion substrate, not from a linear operator or ergodic system.

Relation to Lyapunov Exponents / Spectral Radius Theory

The asymptotic depth exponent λ(K, x) behaves analogously to a log spectral radius or Lyapunov exponent — it captures the exponential growth rate of a quantity under iterated application of an operator. But here the operator is the recursion kernel and the quantity is the dispersion functional σ_F, not a linear map acting on a vector space. The result is genuinely new: a Lyapunov-type exponent for dispersion.

Relation to Maximum Principles (PDE analogy)

The Universal Sign Preservation Theorem structurally parallels maximum principles for elliptic PDEs: positivity + bounded distortion ⟹ sign preserved. In PDE theory, the positivity condition is the boundary data; in UNNS, it is the margin axiom. The bounded distortion is curvature control in PDE; in UNNS, it is depth submultiplicativity. The analogy is structurally precise, though the setting is entirely different.

Relation to Statistical Physics / Thermodynamic Limits

Quasi-subadditivity + convexity + exponent existence closely resembles free energy density arguments in statistical mechanics: the free energy per site has a well-defined thermodynamic limit precisely because it is (quasi-)subadditive in system size. Here, recursion depth plays the role of system size, and the log-potential ψ plays the role of free energy density. The structural parallel is conceptually illuminating, if not formally identical.

💡 What This Paper Actually Changes

The most important question a reader should ask is not "what does this prove?" but "what changes in UNNS because this paper exists?"

Before the Backbone Paper

  • Sign preservation depended on chamber elimination + μ-margin axiom
  • Constants (1.353, 2.136) entangled with structural claims
  • σ_F described in chamber prose, not formally specified
  • No asymptotic depth theory — only finite-depth empirics
  • Theorem fragile if any constant shifted or chamber protocol changed
  • Chambers served as proof foundations

After the Backbone Paper

  • Universal Sign Preservation Theorem is constant-agnostic
  • Only existence of C* > 1 required
  • σ_F formally defined via L_n, OLS slope, seed-pool std
  • Asymptotic depth exponent λ(K,x) proven to exist
  • Theorem stable: chambers certify, they do not prove
  • Chambers serve as certification of axiom satisfaction

The Architectural Shift

If you removed the backbone paper, UNNS sign preservation would revert to being an empirical claim about 7,400 cell-regime records that happened to show no negative curvature. Strong evidence — but evidence, not proof. With the backbone paper, sign preservation is a structural theorem: admissible recursion cannot certify persistent negative curvature, and the chambers are the verification that the abstract axioms are satisfied in the UNNS substrate. That is the difference between science and mathematics.

⚖ Positioning: Backbone vs Duality Theorem

UNNS now has two flagship theoretical papers with distinct and complementary roles. Understanding the difference is important for reading either correctly.

Dimension Duality Theorem Backbone Paper (this)
Conceptual novelty Very high — new principle Moderate–high — formalisation
Mathematical depth Moderate High — Fekete, convexity, CI locking
Cross-field relevance High (physics, QKD, philosophy) Mostly internal + math analysis
Internal structural importance High Very high — eliminates fragility
External paradigm effect Yes — observability is operator-relative No — stabilises existing theorem
Risk level Higher (interpretive claims) Low (formal proofs)

They Are Complementary, Not Competing

The Duality Theorem asks: why does operator structure matter for what is observable? It has philosophical and physics implications. The Backbone Paper asks: given that operator structure matters, what can we prove about the admissible class? It has mathematical rigour implications. UNNS needs both: the Duality Theorem gives UNNS its philosophical core; the Backbone Paper gives its LXI–LXIV arc mathematical permanence.

UNNS Flagship Paper Ecosystem OBSERVABILITY–ADMISSIBILITY DUALITY Flagship · Paradigm-level Observability is operator-relative, not ontological Affects Bell tests · DI-QKD · null experiments Cross-links Chambers XL, XXXI, XXXII External Impact: ★★★★★ Complementary ADMISSIBILITY BACKBONE PAPER Structural Engine · Formal Proves sign preservation under axiomatic conditions Asymptotic exponent · Quasi-subadditivity Companion theory to Chamber LXII Internal Rigor: ★★★★★

🔍 What This Paper Does NOT Do

Intellectual honesty is a load-bearing element of UNNS methodology. The backbone paper's precision makes its open questions equally precise — which is itself progress.

Three Remaining Open Questions

  • Analytic derivation of C*. The paper certifies that C* > 1 exists. It does not derive the value of C* from the recursion kernel structure analytically. C_med = 1.353 and C_q95 = 2.136 are empirically measured, not proven from first principles. Deriving C* analytically would be a significant independent result.
  • μ-margin from first principles. Axiom 4 (Margin Dominance) asserts E[b₂(K;x)] ≥ μ > 0 for all admissible (K,x). This is empirically verified across the 231-cell simplex, but not proven from the kernel structure. The proof strategy (C_unif · spectral curvature decoupling → analytic lower bound on CI99_lo) remains open.
  • Global uniform σ_F vertex bound. The paper does not establish a global uniform upper bound on σ_F across all admissible (K,x). The grid provides certification over a finite simplex; the analytic extension remains open.

Why Isolating Them Is Progress

Before the backbone paper, these open questions were entangled with the structural result itself — it was unclear which parts of sign preservation were proven and which were assumed. Now the separation is clean: the theorem follows from five axioms, and what remains open is precisely the analytic justification of those axioms from the recursion kernel. That is intellectually tractable. The proof strategy is visible.

Resources & References

UNNS Research Collective · February 2026 · Foundational Admissibility Paper · Theoretical Backbone of the LXI–LXIV Elimination Programme (Variance-Geometry Formalisation) · Companion theory paper to Chamber LXII · Not a chamber · Not an experiment · A structural extraction.