WP-82 · The Missing Floor — Principia Orthogona Vol VI

Noncommutative geometry appears in fifteen files of this corpus. Spectral triples appear in seven. K-theory appears in none. A spectral triple is a device for computing an index; Connes' geometry is built on K-homology and the Chern character. The series has been standing on its top rung without having built the one beneath it.

DATA measured MODEL derived in-series VALUE a stated number ASSUME a choice, not a finding OPEN not established

Renumbered 2026-08-29 · WP-81 → WP-82

This paper was published on 2026-08-28 as WP-81. That number was already held by WP-81 · The Conviction and the Kernel, published 2026-08-27 and carrying five inbound citations. The number was taken without checking whether it was free — the same class of unmeasured absence this paper is about. Moved to WP-82; the collision and its cause are recorded here rather than quietly corrected.

§1

The measurement

A circulating graphic titled 33 Levels of Mathematical Mastery orders the mathematical fields by ascending difficulty and terminates at rung 33, Noncommutative Geometry. The list is useful here not as a curriculum but as a ruler: this series can be held against it and asked, rung by rung, whether the vocabulary is present in the corpus at all.

The count below is the number of tracked *.html and *.md files containing at least one case-insensitive match, taken at commit 654fb06 on 2026-08-29. DATA

Re-measured 2026-09-16, same method, at commit d97154e. The 2026-08-29 column is kept: it reproduces exactly, twelve of twelve, when the method is re-run at the commit it names, and it is what the argument below was built on. book6/wp82-verify.py recomputes both columns at the commits they name and fails if either drifts. DATA

Correction, 2026-09-17 — the ruler is inside the set it measures. The two columns below did not range over the same corpus. At 654fb06 this paper is not in the tree at all — book6/wp81 is the last working paper in that commit — so the first column counts a corpus without the ruler. At d97154e the paper is in the tree and matches all twelve of its own patterns, because it prints them; so the second column counted itself, twelve times, and the drift between the columns carried a spurious +1 in every row. Principia Mathematica's rule (Whitehead–Russell, Vol I, Introduction ch. II) is that no object may be defined in terms of a totality containing itself, and its remedy is to stratify the range of the variable. Both columns now exclude the ruler. At 654fb06 that changes nothing, which is how the exclusion is known to be the right one rather than a convenient one. Rung 28's total is 2 → 17 and rung 33's 31 → 56; the inversion narrows to 3.3 : 1 rather than 3.0 : 1. The finding survives and the numbers move. book6/wp82-verify.py block [2] recomputes both the stratified and the unstratified column and asserts both, and tools/self_reference.py is the instrument that found it. DATA

Correction, 2026-09-16 (second measurement of the day). The second column was first published against HEAD rather than a named commit, and six of its twelve rows moved within the same day as the corpus was written to — k-theory 7 → 9, index theorem 4 → 6, Atiyah 3 → 5, von Neumann 11 → 12, sheaf 6 → 7, spectral triple 14 → 15. A date is not a commit. Both columns now name one, and both are recomputed rather than remembered. DATA

Rung	Field	Pattern	2026-08-29	2026-09-16 @ d97154e stratified
28	K-Theory & Index Theory	k-theory	0	8
28	↳ index theorem	index theorem	1	5
28	↳ Atiyah	atiyah	1	4
29	Operator Algebras	operator algebra	76	99
29	↳ von Neumann	von neumann	7	11
30	Higher Category Theory	∞-categor / infinity-categor	0	2
31	Derived Algebraic Geometry	sheaf / sheaves	1	6
32	Motivic / Langlands	motivic / langlands	3	6
33	Noncommutative Geometry	noncommutative	9	16
33	↳ Connes	connes	15	26
33	↳ spectral triple	spectral triple	7	14
—	Monstrous Moonshine	moonshine	25	27

The finding

The distribution is inverted. The corpus reaches rung 33 in thirty-one file-mentions across three vocabularies, and rung 28 in two. K-theory, the subject on which the index theorem and therefore the spectral triple depend, occurs zero times.

This is the same shape as three defects recorded elsewhere in this corpus and its tooling: a vacuity scan whose grep anchor could never match, so its green meant nothing; a Lean header claiming five honest admits where the kernel reported six; and a theorem whose hypothesis had silently become sorry under Mathlib drift, so that #print axioms returned the same four axioms before and after the statement stopped meaning anything. In each case the surface reads correct and rests on something never checked to exist. DATA

§2

Why the gap is load-bearing

A spectral triple (𝒜, ℋ, D) is not a description; it is an instrument. The point of packaging an algebra, a Hilbert space and a Dirac operator together is that the pairing of D with a K-homology class produces an integer — an index — and that integer is the geometric invariant. Strip out K-theory and the index theorem and a spectral triple is notation with nothing to compute.

So the fifteen files invoking Connes are not wrong so much as ungrounded. They use a vocabulary whose semantics live one rung below, on a rung this corpus has never written. That is a statement about the series' architecture, not about the mathematics, which is settled and independent of us. OPEN

The hinge from Part I

Part I already contains the argument that forces Part II, and states it in Volume I: Liouville's theorem forbids attractors on compact symplectic manifolds. The series needed post-fold stability, could not have it symplectically, and escaped into contact geometry, where the result does hold.

Part II asks the next question that move implies. If the obstruction was the manifold, the escape into contact geometry was a small step — a change of structure on the same kind of object. The larger step is to stop requiring a manifold at all and let the algebra of functions be the space. That is rung 33, and it is where Part I has already been pointing without saying so.

Part I asked what generative transitions are: folds, limit cycles, bifurcations classified against Whitney A₁–A₃, an operator chain G = U∘F∘K∘C acting on manifolds. Part II asks what kind of space they live on when it stops being a manifold. Six volumes, one per rung, ascending from the floor to the ceiling the corpus has already been quoting.

§3

Part II · Volumes XI–XVI

Vol	Rung	Subject	Kind	Seed already in the corpus
XI	28	K-Theory & Index Theory	FLOOR	Not new ground — the floor under material already published. Whitney A₁–A₃ fold classification; the transverse Floquet multiplier λ⊥ = e^(μ_maxT*) = e^−4π as an analytic index; the Morse condition from Vol I.
XII	29	*Operator Algebras — C- and von Neumann**	CONSOLIDATION	"The operator algebra of C→K→F→U" is used across 76 files and has never had a volume. Gather it, state it once, verify it. Forced position: NCG is what a C*-algebra becomes when taken seriously as a space, so this must precede XVI.
XIII	30	Higher Category Theory & ∞-Categories	NEW	CatGT and GOMC are the entry point. Zero files use the vocabulary. Where the operator chain stops being a diagram and becomes a structure. 2026-09-16: the vocabulary is now in three files, and Volume XIII on disk (book13/, nine chapters, stubbed 2026-08-28 in `9d78ff8` — after the commit this table was taken at) is where it is. That volume was written before XI and XII for a reason recorded in `book13/ch-mathlib-verify.py`: Mathlib has no K-theory, so Volume XI cannot have a machine-checked core, while `CategoryTheory/` is large enough that XIII's work is instantiation rather than construction.
XIV	31	Derived Algebraic Geometry	NEW	E8 at the octonion rung as the integral-octonion lattice; homological algebra already present from the category-theoretic work. Sheaves appear once.
XV	32	Motivic Homotopy / Langlands	BRIDGE	Monstrous Moonshine — 25 files, and Vol VIII entire — is the road from the Monster to modular functions, which is half the distance already walked. Correction, 2026-09-11: this seed is wrong in a way worth stating. Moonshine supplies modular functions — the automorphic side. Langlands is a correspondence, and the other side is Galois representations, which Moonshine does not produce; half of a correspondence is not half the distance. Two replacements, both already held. Arithmetic: `book4/ladder-polynomials.html` carries concrete Galois groups for the n-bonacci family, generically the full symmetric group and so non-solvable from n = 5; the honest entry is the question of which Artin L-function results, not a claim that it is automorphic. Geometric: W-algebras. Feigin–Frenkel gives the centre of the affine vertex algebra at critical level as the classical W-algebra of the Langlands dual, z(ĝ) ≅ W(ᴱg), with a quantum duality relating W^k(g) and W^k′(ᴱg) through the levels — Langlands duality inside vertex algebra theory, needing no Galois side at all, and reachable from the VOA material Vol VIII already holds. The corpus has 7 files using “vertex operator”, 5 using “Virasoro”, and 0 using “W-algebra”.
XVI	33	Noncommutative Geometry	CEILING	Where the 31 existing file-mentions of Connes, spectral triples and noncommutativity finally acquire a floor and a home.

Admissibility — the bar Part II must clear

This corpus's standing rule is that an entry is written only when there is a recorded artifact behind it. A six-volume plan is a plan, not a claim, and a rung is not occupied because a volume is named after it. Each of XI–XVI requires a machine-checked core — a Lean file that elaborates clean against a pinned Mathlib and reports its axioms — or it is a reading list with a DOI on it. Volume XI's core is the first test of that, and the easiest to state: an index, computed, verified, and reported by the kernel.

Correction, 2026-09-17 — the obstacle was measured once and then inherited. The sentence “Mathlib has no K-theory, so Volume XI cannot have a machine-checked core” has been repeated in this corpus without being re-measured. Re-measured today against the pinned checkout (mathlib 81a5d257c8, toolchain v4.32.0) it is true of the name and overstates the obstacle.

There is still no Mathlib/KTheory/ and still zero files containing K_0. But Mathlib now carries Mathlib/GroupTheory/MonoidLocalization/GrothendieckGroup.lean (Best & Dillies, 2025), which defines GrothendieckGroup M := Localization (⊤ : Submonoid M) with instCommGroup and lift : (M →* G) ≃ (GrothendieckGroup M →* G). That equivalence is the universal property, and group completion is the whole of the K₀ construction. Alongside it: RingTheory/ClassGroup.lean, and NumberTheory/NumberField/ClassNumber.lean with classNumber, classNumber_pos and classNumber_eq_one_iff — positivity proved from Fintype.card, so finiteness of the class group is established rather than assumed.

Since K₀(𝒪_K) ≃ ℤ ⊕ Cl(K) (Weibel, K-book ch. I–II), the right-hand side of Volume XI's easiest theorem is already machine-checked. What is missing is the monoid of finitely generated projective modules under ⊕, and the structure theorem joining the two — a quotient-type construction and a bridge lemma, not a field. The admissibility verdict for XI is restated accordingly: not “no core is possible” but “the core needs two named objects built on arithmetic Mathlib already holds.”

The distance is now written down rather than guessed, and as of 2026-09-18 it is closed. book6/lean/VolXI_K0_Floor.lean states the instrument checks, the first theorem, the class-number facts Mathlib supplies, and the obligation itself — and it now compiles with no sorry against the pinned checkout. Its axiom report reads [propext, Classical.choice, Quot.sound], none of which is sorryAx. That is exactly this paper’s bar, met. DATA

What the remaining theorem turned out to be. The group completion of ℕ is ℤ — and the proof is not a construction. Mathlib’s Localization.mulEquivOfQuotient already carries a localization map to an isomorphism, and GrothendieckAddGroup M is an abbrev for AddLocalization (⊤), reducibly the same type. So the whole task was certifying that the cast ℕ → ℤ is a localization map at ⊤: every integer is an additive unit, every integer is a difference of two naturals, and the cast collapses nothing. Three arithmetic conditions, all of which omega discharges. Two earlier attempts built the map by hand and spent themselves proving it bijective; the file keeps them, because the rule they illustrate is general — when a universal property is available, constructing the map by hand is work you have chosen, not work the theorem requires.

What is still absent upstream is unchanged: the monoid of finitely generated projective modules under ⊕. So §2 of that file proves the second half of “K₀ of a field is ℤ” and states the first half as the thing Mathlib does not yet have. That is a gap in Mathlib, not in the file. book6/wp82-k0-floor-verify.py measures the Mathlib claims above and recomputes the class numbers from scratch by counting reduced binary quadratic forms.

§3b

The one index candidate the corpus already holds — measured

The table above gives Volume XI a seed: “the transverse Floquet multiplier λ_⊥ = e^(μ_maxT*) = e^−4π as an analytic index.” That is the only concrete candidate for an index anywhere in this corpus, and it is the thing Volume XI would have to ground. It has now been computed rather than quoted, against the exact equations in Volume II §4.3, and it does not behave the way the sentence implies. DATA

ṙ = r(1 − r²) + 2(r − 1)e^−z, θ̇ = 1, ż = r² − 2(r − 1)²e^−z

on (ℝ²_>0 × ℝ, α = dz − r²dθ), with Γ = {r = 1}

What holds exactly. Γ is invariant: the radial field vanishes identically on r = 1, for every z. The transverse eigenvalue is the radial derivative there, and it is exactly the function Volume II states, λ(z) = ∂_rṙ|_r=1 = 1 − 3 + 2e^−z = −2(1 − e^−z) — confirmed to twelve digits. And the asymptotic rate is right: λ(z) → −2 as z → ∞, so with T* = 2π the product μ_maxT* is −4π. DATA

What does not hold. On Γ the third equation reads ż = 1. The height is not a parameter of the orbit; it is a coordinate the orbit climbs, at unit rate, forever. Γ is a helix, not a closed orbit — it closes in the (r, θ) projection and in no other. So the monodromy of one turn is not e^λT* for any single λ. It is the integral of a moving one, and it has a closed form:

∫₀^2π λ(z₀ + t) dt = −4π + 2e^−z₀(1 − e^−2π)

RK4 integration of the variational equation agrees with that expression to 1×10⁻¹³ at every height tested. The consequence is the finding: DATA

The multiplier is a function of where you start

z₀	multiplier over one turn	÷ e^−4π
0	2.567 × 10⁻⁵	7.36
1	7.268 × 10⁻⁶	2.08
2	4.569 × 10⁻⁶	1.31
5	3.535 × 10⁻⁶	1.014
10	3.488 × 10⁻⁶	1.0001
→ ∞	3.487 × 10⁻⁶	1

e^−4π is the limit, not the value. At the neutral height z₀ = 0 the true multiplier is 7.36 times larger. The two agree to 1% only above z₀ ≈ 5.

Why this matters for Volume XI specifically, and not as a correction. An index is not merely a number attached to an operator. Its defining property is that it does not move: it is invariant under continuous deformation, which is what lets it be computed topologically on one side of Atiyah–Singer and analytically on the other. The quantity above moves. It is a smooth, strictly monotone function of the base point z₀, it takes a continuum of values, and it is not an integer at any of them. Whatever else it is, it is not yet an index, and no amount of K-theory will make a continuously varying real number into one. OPEN

What the calculation does supply is a better-posed target. Deformation invariance is exactly what is missing, so the question Volume XI inherits is sharp: is there a K-theory class whose pairing is constant along this helix? The candidates are the ones the drift itself suggests — the z → ∞ limit as a boundary or asymptotic index, or a relative class on the pair (M, {z ≤ c}), or a Conley-type invariant of the isolated invariant set, none of which is e^−4π as written. That is more than this paper had before, and it is still a question rather than a result. OPEN

Computation: book7/ch-grothendieck-verify.py, blocks [3] and [4], standard library only. The same script checks the K-theory construction this rung is named after. The history — why the group is called K, what Grothendieck built it for in 1957, and why his own route to the Weil conjectures is still open — is in Book 7 · Alexander Grothendieck. The index-theorem half of rung 28 belongs to the Weil chapter’s neighbours and is not yet written. The ceiling this floor is meant to hold up is Book 7 · Alain Connes.

§4

Known limits of this paper

One spelling per concept. The counts are single-pattern greps. $K$-theory, K theory, or KK-theory would not have matched, so 0 means "zero in that form", not "absent from the corpus". The finding is strong enough to survive a second pass but has not had one. OPEN

The instrument fails in three ways, and this paper originally recorded one. Added 2026-09-16, each checked on every run of a named script. DATA

Wrong spelling — the paragraph above. A 0 means zero in that spelling.
Substring inflation. An unanchored pattern matches inside longer words, so a non-zero count can be pure noise. Measured at d97154e: /gns/ matches 68 files and /\bGNS\b/ matches 0 — the hits are designs and assignments; /bott/ matches 789 and bott periodicity matches 1 — the rest are bottom. Checked by book7/ch-conley-verify.py block [7].
Entity blindness — the one that makes false zeros on visible text. The corpus is HTML and writes accented names as character entities, which no plain pattern matches. Entity-aware at d97154e: Liénard 5, not 2; Poincaré 68, not 58; Gödel 27, not 23; Poincaré–Bendixson 14, not 12. Sixty-eight tracked HTML files carry at least one accented entity, so every published count in this corpus for a name with a diacritic is low until it is re-taken. The instrument is tools/corpus_count.py; the case that found it is ch-van-der-pol, whose block [6] reported 0 chapters for a word printed twice on the page doing the counting.

A fourth, smaller, belongs with them because it is the same class: git ls-tree -r --name-only HEAD -- '*.html' returns nothing and exits 0 where ls-files and grep honour the same pathspec — a silent empty answer. DATA

The ruler is damaged. The source graphic breaks its own numbering — rung 9 carries no numeral — and jumps from 28 straight to 33, so rungs 29 through 32 are absent from it. Of the four, only 29 is forced by the mathematics; the placement of higher category theory, derived algebraic geometry and motivic homotopy in XIII–XV is an ordering chosen here, not one recovered from the source. ASSUME

Part I is not fully surveyed. Volume X carries a deposit DOI in the AXLE header but no title appears anywhere in the repository, so the arc from IX to XI has a segment this paper cannot see. OPEN

This paper contains no formal results. Its contribution is a dated measurement of the corpus against an external ruler, and the observation that the measurement is inverted: the series is heaviest where it is least founded. The remedy proposed — build the floor first — is an architectural recommendation and is not established by the measurement, only motivated by it.

References

Principia Orthogona, Volume I — The Mathematics of Generative Transitions. Concept DOI 10.5281/zenodo.19117399. Source of the Liouville obstruction and the contact reformulation.
Principia Orthogona, Volume II — Contact Realization of Generative Transitions.
Principia Orthogona, Volume VIII — The Monster. The moonshine material on which Volume XV would build.
A. Connes, Noncommutative Geometry, Academic Press, 1994. The spectral triple and its index pairing.
M. F. Atiyah and I. M. Singer, The Index of Elliptic Operators, Ann. Math., 1968–1971. Rung 28.
33 Levels of Mathematical Mastery, circulating graphic, source unverified. Used here as a ruler, not as an authority; its rungs 29–32 are absent and its rung 9 is unnumbered.
Defect record: AXLE/main/AXLE_v8_1.lean, header corrected from five honest admits to six on 2026-08-28 after the kernel probe reported sorryAx on all six declarations; and vol1-proofs/tools/run.sh, vacuity fixture failure of the same date.