UNNS Substrate Research Program · Phase Mapping · April 2026

The Geometry
That Refuses to Break: Mapping the Regimes of Structure

The first systematic deformation study of realizability structure: 22,817 evaluations across 93 datasets and 11 physical domains reveal a locally rigid, boundary-governed structural observable — and expose representation, not deformation, as the dominant structural variable.
hero image
STRUC-PERC-I v2.4.1 93 datasets · 22,817 evaluations Zero inter-class transitions Zero non-trivial commutators 11 physical domains Principle 1 — corpus-supported Field Generator v1.0
Instrument: STRUC-PERC-I v2.4.0/2.4.1 Pipeline: Field Generator (Python + Node.js + Puppeteer) Grid: 17×17 joint (α, μ) · Ω = [0.80, 1.20]² Date: April 2026 · unns.tech
Executive Summary

We investigated how the realizability classification of ordered physical sequences responds when their defining parameters are continuously deformed. Using a fully automated field generator pipeline and the STRUC-PERC-I percolative chamber, we subjected 93 ladder datasets to a joint (α, μ) deformation grid spanning ±20% around physical values.

The outcome is unambiguous. Across all 22,817 evaluated grid points, zero inter-class transitions were observed. Every tested ladder stayed in a single realizability class — Full, Tail, or Hard — across the entire parameter space. The structural commutator was identically zero at every point. Phase maps were monochromatic throughout.

"Realizability in the UNNS Substrate is a locally rigid, boundary-governed structural observable — governed not by continuous response but by discrete boundary conditions in gap space."

Simultaneously, the same corpus reveals that representation — not deformation — is the primary structural variable. The same physical system encoded differently can yield a categorically different realizability class, while bounded deformation of any single encoding never does.

Total datasets
93
B1(34) · B2(3) · B3(56)
Grid evaluations
22,817
17×17 per ladder
Full (at physical α)
72
77.4% of corpus
Hard (at physical α)
16
Structural fragmentation
Tail (at physical α)
5
Outlier-dominated
Inter-class transitions
0
Zero across all 22,817 points
Non-trivial commutators
0
Δ𝒞 = ΔGR = Δκ = 0 everywhere
Physical domains
11
Atomic to cosmic web

Principle 1 — Bounded Structural Rigidity of Realizability

The central result of this work is not simply a null observation — it is the identification of a structural principle. The realizability coordinate ℛ(L) of an admissible ladder is not continuously sensitive to parameter variation. It exhibits piecewise-invariant behaviour: deformation space is partitioned into regions within which ℛ(L) is constant, separated by boundaries where discrete class changes may occur.

Principle 1 · Bounded Structural Rigidity of Realizability
Let L ∈ ℳadm be an admissible ladder. There exists a finite deformation domain ΩL ⊂ ℝ² containing the physical point (α, μ) = (1, 1) such that:

(i) The realizability class 𝒞(L) is invariant over ΩL.
(ii) The connectivity structure Gκ(L) — in particular GR and κconn — is invariant over ΩL.

The tested domain Ω = [0.80, 1.20]² lies within ΩL for all 93 corpus datasets, with zero inter-class transitions across 22,817 evaluations. This is a bounded, local property of the realizability coordinate; it is not a universal structural invariant.

Three Non-Negotiable Constraints

Locality

The stability domain is finite: ΩL ⊊ ℝ². No claim is made that 𝒞(L) is globally invariant under all deformations. Breakdown scans at α ∈ [1.0, 4.0] (a separate protocol) confirm that rigidity does not extend to all scales.

Ladder-dependence

ΩL depends on L. Different physical systems and different encodings of the same system have different stability radii. No universal Ω is asserted — each ladder carries its own stability region.

Coordinate restriction

Principle 1 applies only to the realizability coordinate ℛ(L). It makes no claim about the admissibility coordinate ρ̄(L) or the full two-coordinate state 𝒮(L) = (ρ̄(L), ℛ(L)). Both coordinates are needed for a complete structural description.

The Geometry: Stratified, Not Smooth

The (α, μ)-plane does not support a smooth response manifold for realizability. Instead it is partitioned into rigidity domains and transition loci. The phase maps observed in this corpus correspond to a single rigidity domain containing the tested region Ω.

Stratified Deformation Space — Ω₍L₎ and Transition Loci ℝ² Stability region Ω₍L₎ 𝒞(L) invariant throughout Tested Ω [0.80, 1.20]² 289 pts · 22,817 evaluations Zero transitions (1,1) physical ∂Ω₍L₎ Transition locus ∂Ω₍L₎ α (scale deformation) 0.80 1.00 1.20 Tested region Ω ⊆ Ω₍L₎ Stability region Ω₍L₎ (extent unknown) Transition boundary ∂Ω₍L₎ μ (intensity deformation) on vertical axis · Ω₍L₎ is ladder-specific · boundary location not determined by this corpus

Derived Corollaries

Three corollaries follow directly from Principle 1 and are attested across the full corpus:

Corollary 1 — Degenerate Local Phase Space

Within ΩL, the phase map ΦL(α, μ) = 𝒞(L) is constant. No phase structure — no transitions, no boundaries — is accessible within the tested region. Every phase map in this corpus is monochromatic.

Corollary 2 — Emergent Commutativity

Within ΩL, C(α, μ; L) = 0. Operator commutativity is a consequence of realizability invariance, not a fundamental property of α and μ. It is emergent, not primitive. Outside ΩL, the same operators may produce non-trivial asymmetric structural responses.

Corollary 3 — Local Operator Irrelevance

Within ΩL, both operator order (α∘μ vs. μ∘α) and operator magnitude are structurally irrelevant for the realizability coordinate. The pipeline may apply α and μ in any order without affecting the structural outcome.

Monochromatic Phase Maps — Zero Transitions

For every one of the 93 tested datasets, the phase map ΦL : Ω → {Full, Tail, Hard} is a constant function — a single uniform colour across all 289 (α, μ) grid points. No dataset exhibits a verdict transition anywhere in the tested parameter space.

Figure 1 — Phase Maps
Figure 1 - Three monochromatic 17×17 phase maps:
He (QM-I) → Full (green, all 289 pts), Na Zeeman n=1999 → Tail (amber, all 289 pts), Na (QM-I) → Hard (red, all 289 pts).
Caption: "Monochromatic phase maps — zero inter-class transitions across all 22,817 evaluations.
17×17 Grid — Each cell is one (α, μ) point — All cells identical within each system He (QM-I) — Full κ_conn ≈ 10⁶ · GR = 1.000 · all 289 pts FULL × 289 Na Zeeman n=1999 — Tail GR ∈ [0.955, 0.984] · outlier-dominated TAIL × 289 Na (QM-I) — Hard Persistent fragmentation · no κ_conn HARD × 289

What this means

A monochromatic phase map is not a trivial outcome. A priori, deforming scale and intensity by ±20% could shift gap ratios across critical thresholds, reconfigure the vulnerability graph, and move the ladder into a different regime. The corpus shows this never happens for any of the 93 tested systems. The realibility coordinate behaves as a categorical structural observable defined on stratified regions of deformation space — not as a smooth response variable.

Representation Dominance — The Primary Structural Variable

While bounded deformation within Ω never changes the realizability class of a fixed ladder, the choice of how to encode the physical system as a ladder does — sometimes categorically. This is the most striking finding in the corpus and the strongest empirical confirmation of Theorem 7.5 of the Dual Observability manuscript.

Figure 2 — Representation Dominance
Figure 2 — He: QM-I encoding yields Full (GR=1.000, dense connectivity graph, green) vs. Zeeman encoding yields Tail (GR < 1, sparse, amber). Same physical system · Different ladder construction · Categorical regime shift.
Representation Splits — Same Physical System, Different Encodings He (QM-I) Standard level ladder → FULL · GR = 1.000 same atom He (Zeeman) Field-split encoding → TAIL · GR < 1 Na (QM-I) Standard level ladder → HARD · no backbone same atom Na (Zeeman) Field-split encoding → TAIL · GR < 1 HD — Parity Splits Same molecular system, different sub-ladders Lower sub-ladder → FULL · GR = 1.000 Combined / Odd parity → HARD · GR = 0.800 ΔGR = 0.200

The Canonical Ladder Problem

These findings crystallise the highest-priority open question in the UNNS framework: Is there a canonical encoding for each physical system that makes the realizability class an invariant of the physical system rather than the measurement procedure? The Dual Observability manuscript (Proposition 7.3) identifies this as open; the present corpus provides the strongest empirical attestation that the problem is real and non-trivial.

Intra-deformation vs. inter-representation variation

Intra-grid GR variation (within a single encoding, across all 289 (α, μ) grid points): 0.000 for every dataset.
Inter-representation GR variation (HD combined vs. HD lower): 0.200.

These differ by a factor of infinity. They measure fundamentally different things. Deformation does not explore the structural landscape; re-encoding does.

Documented Cross-Representation Splits

SystemEncoding 1ClassEncoding 2Class
HeQM-I (standard)FullZeeman (field-split)Tail
NaQM-I (standard)HardZeeman (field-split)Tail
LiQM-IFullZeemanFull
HDCombined (full vibrational)HardLower sub-ladderFull
HDEven parityFullOdd parityHard
GeoidPhysical α = 1.0FullAmplified α = 1.1 (non-physical)Hard
Crystalsper_atom normalisationvariescell_volume normalisationvaries

Sub-Regime Metric Variation — Hidden Continuous Structure

Realizability class is categorical — it assigns every ladder to one of four discrete regimes. But within a regime, the connectivity threshold κconn is a continuous structural coordinate that varies by over six orders of magnitude across the Full class alone. This variation is invisible to the verdict, yet constitutes genuine structural information.

Figure 3 — Sub-Regime Metric Variation
Figure 3 — Bar chart: κ_conn across Full-class systems on log scale. CH4/ERA5/CO2 (≈0.3, near-immediate) · CMB R3.01 (1.0–7.9) · GOES-XRS Solar (16.6) · Cosmic Web DESI (5,908–15,658) · Na QM-I (161,260) · Li QM-I (376,546) · He QM-I (≈10⁶, maximum adaptive extension). Caption: "κ_conn spans over six orders of magnitude — invariant across the entire (α, μ) 17×17 grid.
κ_conn Across Full-Class Systems — Log Scale Each value is constant across the full 17×17 (α, μ) grid — sub-regime metric structure is a property of the ladder, not the deformation 10⁻¹ 10⁰ 10¹ 10² 10³ 10⁴ 10⁵ 10⁶ κ_conn (log scale) ≈0.3 CH4/ERA5 /CO2 ≈0.3 1.0–7.9 CMB R3.01 1–7.9 16.6 GOES-XRS Solar 16.6 5,908– 15,658 Cosmic Web DESI ~10⁴ 161,260 Na QM-I ~1.6×10⁵ 376,546 Li QM-I ~3.8×10⁵ ≈10⁶ He QM-I max ext. All FULL same class, different depth

Categorical class vs. continuous coordinate

Realizability class 𝒞(L) is categorical: it partitions ladders into discrete, mutually exclusive regimes. The connectivity threshold κconn is a continuous structural coordinate: it measures the depth of connectivity within a class. These are complementary, not redundant — same class does not imply same structure at the metric level. κconn is itself constant across the (α, μ) grid for every fixed dataset, confirming that sub-regime metric structure is a property of the ladder encoding, not of the deformation parameters.

Open in Fullscreen!

The Connectivity-Margin Mechanism — A Structural Explanation

The corpus results call for a structural explanation of why realizability is locally rigid. We propose a candidate mechanism grounded in the geometry of the vulnerability graph.

Mechanism Candidate 1 — Connectivity-Margin Mechanism

A ladder L possesses a non-zero local rigidity region ΩL when its gap vector lies at a positive distance from the nearest realizability-class boundary induced by the vulnerability-graph predicate. In that case, sufficiently small (α, μ) deformations preserve the decisive connectivity relations underlying class membership, and the realizability coordinate ℛ(L) remains unchanged.

Define the connectivity margin m(L) as the minimum normalised distance of any decisive gap-ratio pair from the critical connectivity threshold κ*. When m(L) > 0: the class-defining connectivity relations of Gκ(L) are preserved under bounded deformation, and the structural commutator C(α, μ; L) = 0.

Status: candidate mechanism — consistent with the entire corpus but not yet a formally proved theorem. It is the most parsimonious structural account of the observed monochromatic phase maps and zero commutator.

Connectivity Margin — Gap Vector Distance from Realizability Boundary Realizability boundary FULL region gap vector Δ(L) m(L) > 0 deformed versions — all FULL HARD region HARD No backbone forms No κ_conn within range

The Field Generator — Automated Phase-Mapping Pipeline

All 22,817 corpus evaluations were produced by an open, deterministic pipeline called the Field Generator: a multi-layer system coupling a Python orchestration layer to the STRUC-PERC-I engine via a Puppeteer-bridged Node.js process.

Field Generator — Evaluation Pipeline Ladder Files .txt / .csv 93 datasets Python Layer runner.py α×μ grid (17×17) Apply operators Dispatch + aggregate Node.js Bridge engine_server.js Puppeteer headless JSON RPC Line-oriented protocol STRUC-PERC-I v2.4.0 / v2.4.1 Vulnerability graph κ sweep (17 pts) Verdict / GR / κ_conn Results JSON · full grid Forward + Reverse Commutator analysis Regime classifications UNNS_MODE=phase · break · mu Deterministic · byte-identical outputs

Operating Modes

ModeUNNS_MODEGridUse
Phase mappingphase (default)α ∈ [0.80, 1.20] × μ ∈ [0.80, 1.20], 17×17 ptsPrimary corpus — all B1 results
Breakdown scanbreakα ∈ [1.0, 4.0], step 0.25 × μ = 1.0 fixedEstablishing outer boundary of ΩL
μ-sweepmuα = 1.0 fixed × μ ∈ [0.95, 1.05], 21 ptsGeoid and cosmic web B3 datasets

Downloads

Regime Taxonomy — The PRP Four-Tier Classification

STRUC-PERC-I assigns each ladder to one of four realizability classes defined by the Percolative Realizability Principle. The phase-mapping corpus realises three of the four; Giant is present in the formal taxonomy but not observed in the current datasets.

Realizability Class Taxonomy — PRP Four-Tier Exhaustive Partition FULL Satisfies all 4 PRP conditions GR → 1 · Dense global conn. κ_conn finite and exists Corpus: 72 datasets Simple polyatomics to cosmic web κ_conn <0.3 to ~10⁶ GIANT Satisfies conditions 1–4 Dominant backbone · GR≥thresh Not all nodes participate Corpus: 0 datasets Present in PRP taxonomy; not yet observed in corpus TAIL Satisfies 1–3; fails condition 4 Persistent outlier nodes GR below full-percolation thresh Corpus: 5 datasets Zeeman n=1999 constructions GR ∈ [0.955, 0.984] HARD Fails conditions 1–3 No backbone forms; fragmented Not a USL violation Corpus: 16 datasets Na QM-I · CO · HD combined 9 crystallographic datasets

Hard ≠ USL violation

Hard fragmentation is a structural condition of the gap-ratio distribution, not a violation of the Universal Structural Law. All 16 Hard datasets at physical values are admissible under the USL; they simply do not form a connected backbone under the vulnerability-graph criterion. Selection (the USL constraint) does not uniquely fix organisation (the percolative class).

Implications and Open Problems

What this work establishes

Before this work

  • Realizability was a verdict attached to a fixed ladder
  • Deformation behaviour of verdicts was unexplored
  • Phase space structure was unknown
  • Operator order was assumed to be structurally irrelevant (informal)

After this work

  • Realizability is a locally rigid piecewise-invariant observable
  • The deformation space is stratified, not smooth
  • Operator commutativity is emergent, not fundamental
  • Representation is the dominant structural variable

Open Questions

  1. Canonical ladder problem. Is there a physically motivated canonical encoding for each class of system that makes the realizability class an invariant of the system rather than the procedure? (Proposition 7.3 of Dual Observability — open)
  2. Universality of local rigidity (Conjecture 1). Does every admissible ladder possess a non-zero stability region ΩL? A proof would elevate Principle 1 to a theorem. A counterexample would fundamentally revise the theory.
  3. Non-commutative systems. Is there any ladder for which C(α, μ; L) ≠ 0? Such a ladder must sit near a structural boundary with asymmetric gap-distribution response.
  4. Margin functional m(L). Can the connectivity margin be operationalised within STRUC-PERC-I, computed across the corpus, and used to predict ΩL extent?
  5. Higher-dimensional deformation. Does rigidity extend to deformations involving fundamental constants αs or αG?

Falsifiability

Principle 1 is falsified by the existence of any admissible ladder L for which, for every ε > 0, there exists (a, m) with ‖(a, m) − (1, 1)‖ < ε and 𝒞(αam(L))) ≠ 𝒞(L). No such ladder has been found in the corpus. The construction of a falsifying example is an explicit open problem. Practical search strategies focus on near-boundary ladders, highly irregular gap structures, and fine-grained deformation studies near ∂ΩL.

Corpus Coverage — 11 Physical Domains

The 93-dataset corpus spans the following physical domains, all returning zero inter-class transitions at physical parameter values:

DomainDatasetsPredominant classκ_conn range
Atomic spectroscopy (QM-I)He, Li, Na, K + Zeeman variantsFull / Hard (repr.-dependent)<1 – 10⁶
Molecular vibrational (simple)CH₄, CO₂, H₂O, NH₃, O₃Full< 0.3 — near-immediate
Molecular (combined)CO, HD combinedHard
Nuclear (U-238)5 sub-constructionsmixed
Atmosphere / ERA58 constructionsFull0.7 – 2.0
CMB power spectra (Planck R3.01)TT, TE, EEFull1.0 – 7.9
Cosmic web (DESI, SDSS, 2MRS)7 constructionsFull27.9 – 15,658
Geoid (Earth, Mars, Moon)EIGEN-6C4, JGM85, AIUB-GRL350AFull at physical α0.7 – 6.3
Solar / GOES-XRS1 (first solar domain)Full16.6
Crystallography (Materials Project)15 valid; 3 excludedFull / Hard (norm.-dependent)varies
UNNS Substrate Research Program · unns.tech · April 2026
Instruments: STRUC-PERC-I v2.4.0/v2.4.1 · Field Generator v1.0 · 93 datasets · 22,817 evaluations · 11 physical domains
All corpus evaluations produced by the field generator pipeline (Python + Node.js + STRUC-PERC-I). Engine source at struc_perc_i_v2_4_0.html. Full corpus results in JSON format available for download above.