Part II · Core Thesis: The Topological-Geometric and Mathematical Independent Seals
Mohammad F Islam, MD, MPH, PhD
II.1 · Charter: Two Seals, One Structure
STATUS: [⟀] PART II SEALED · TWO INDEPENDENT SEALS, ISSUED SIMULTANEOUSLY.
Part II seals the architecture's geometric core twice, by two derivations that share no external premise and meet on one object. Layer 2 is the topological-geometric seal: it rests on parallelizability, quadratic forms, and the obstruction theorems of the sphere, and its premises invoke no multiplication. Layer 3 is the mathematical seal: it rests on the composition law of the audit economy and the classification theorems of real division algebras, and its premises invoke no topology. Each layer closes on its own anchors. Their convergence is exhibited in the Clifford Join, a third structure neither premise set contains.
Independence is anchor-level law. The two derivations share no external theorem. Both attach to the one architecture they seal through the architecture's own legislation, and attachment through the sealed object's law is what sealing means. The composition law itself stands open to empirical exercise: the standing protocol of the Instrumentation Ledger composes recorded cascade audits under both bracketings and under content superposition and reads verdict invariance and additivity of the graded coordinate directly off the record.
The seals bind four things: the 3 + 1 anatomy of the architecture, the verdict functional in closed algebraic form, the twelve-gate cardinality with its symmetry group, and the fertility-verification register split. Layer 1, the linguistic-semantic seal, stands complete upstream and is carried as settled input. The verdict economy is three-state native, {[⟀], [X], [?]}. Every load-bearing identity below is machine-instrumented; the residues stand in the Instrumentation Ledger. Section II.1A records the barrier this Part closes and the century of fragments it reconnects.
II.1A · The Barrier and the Residence
Triangulated verification, the convergence of independent evidence channels on a single claim, is the working epistemology of the empirical sciences. Its canonical matrix form is the multitrait-multimethod framework of Campbell and Fiske (1959), the backbone of construct validation; its modern methodological statement governs aetiological epidemiology, where integrating evidence across designs with different bias structures is the stated route to robust causal inference. The practice is mature. The mathematics underneath it was not, and four structural absences define what was missing.
First, cardinality: no theorem in the verification literature states how many independent channels a verification architecture can carry, and the number three appears as convention, never as consequence. Second, closed form: the indices used to quantify convergence are ad hoc, average correlations, eigenvalue summaries, concordance judgments, with no derivation, no canonical bounds, no uniqueness. Third, catalog: with no completeness theorem, every proposed convergence measure is one choice among an unknown number of alternatives, and measure disputes are unadjudicable in principle; the Bovens-Hartmann impossibility, the strongest result in Bayesian coherentism, quantifies over a measure space it cannot enumerate. Fourth, the regress: who verifies the verifier has accompanied verificationism since antiquity, and every classical response is verbal; nothing in the literature gives the regress a dimension, a topology, or a group.
These absences are structural, not computational, and the proof is the demand set any composing verification system must satisfy: audits compose associatively, since a verdict economy in which regrouping the same audits changes the outcome is ill-defined; warrant does not annihilate; and composition rides the linear register the pipeline already operates at, CL-3. Under exactly these demands an evidence triad cannot close in its own three dimensions, by the Scalar Exit and Fertile Orthogonality lemmas of Theorem I and by the wall of II.3.1. Beyond pairwise statistics, where the questions of count, form, catalog, and regress live, the standard model carries no composition and no geometry, and a barrier of that kind yields only to different structure.
Two bodies of mathematics have sat unexploited on either side of the barrier for nearly a century. On the statistical side, the determinant of the correlation matrix has been a named object since Wilks (1932) introduced the generalized variance as the scalar measure of multivariate scatter, and its vanishing under multicollinearity is a textbook diagnostic (Haitovsky 1969): the volume reading of evidential independence, used negatively for ninety years as a pathology detector, never positively as a truth functional with a derivation. On the algebraic side, the scalar triple product, the determinant, the signed volume, acquired its closed algebraic carrier inside Hamilton's quaternion calculus (1844), the scalar part of a vector product, from which the dot and cross products were later extracted as fragments. The volume statistics uses and the algebra that generates it are reconnected in Theorem II: det R = λ².
A third adjacency proves the missing structure is empirically live. Quantum cognition has established at parameter-free precision that human evidence-taking carries non-commutative order structure: the quantum question order model's QQ equality holds across large survey corpora (Wang and Busemeyer 2013; Wang, Solloway, Shiffrin, and Busemeyer 2014). That program operates in complex Hilbert space, and ℂ carries exactly one imaginary direction: in this Part's terms, the built-in collinear degeneracy of a single evidence axis. Quaternionic quantum mechanics (Finkelstein, Jauch, Schiminovich, and Speiser 1962; Adler 1995) deploys the correct algebra in physics and contains no epistemology. The quaternionic register of verification stood unoccupied. This Part occupies it.
The residence delivers what the absences named: the channel count forced (Theorem I), the convergence functional in closed form with derived bounds (Theorem II), the catalog of all admissible invariant truth functions of the triad representation closed (Theorem III), the register inside which measure disputes are adjudicated, the audit roster an A₄-torsor with a spinorial double cover (Theorem IV), and the regress coordinated exactly, as the attitude fiber SO(3) ≅ S³/{±1}, a torsor with no canonical point (Appendix OFL). Convergent validity is corrected in passing: maximal inter-channel agreement is the degenerate frame, and what triangulation measures is volume. Four parameter-free empirical predictions with stated nulls bind the residence to data. First, invariance of det R across the six administration orders of a three-instrument design; null: det R varies across orders beyond sampling error at matched marginals, falsifying the order-invariance of the verdict functional, C7. Second, parity organization of order effects by the sign character of S₃ through the chirality register λ; null: order effects fail to sort by parity, falsifying the chirality register, C5 and OFL.7. Third, replication-rate stratification of triangulated findings by generalized variance at matched mean correlation; null: no stratification, falsifying volume as the truth functional, Theorem II and C4. Fourth, the redundancy penalty, volume-selected method triads outperforming agreement-selected triads at equal aggregate reliability; null: parity of the two selections, falsifying the Exclusion Lemma's empirical content. Each prediction falsifies the named component, and the protocols stand at INST.20 through INST.23.
Verification epistemology has been computing fragments of this algebra since 1932 without tracing them to their source. This Part traces them, and seals the residence.
II.2 · The Anatomy: ℝ ⊕ Im ℍ
The architecture is a scalar ground carrying a triad. The Root Axiom occupies the scalar slot, the real line, the axis of registered actuality. The verification triad occupies three mutually orthogonal directions over that ground: V_F the formal-structural axis, V_E the empirical-thermodynamic axis, V_ER the epistemic-registrational axis. The full anatomy is ℝ ⊕ ℝ³: four slots, one privileged, three orthogonal, carrying the standard quadratic form. Orthogonality is non-compensability. No axis substitutes for another, and no surplus on one axis repairs a deficit on another.
Notation, fixed once. Axis evidence enters as rows in ℝᴺ. The pipeline normalizes each row to zero mean and unit sample standard deviation, then removes Mass-Mandate covariates by orthogonal projection. The post-projection rows are m̃_F, m̃_E, m̃_ER, with weights d_a = ‖m̃_a‖²/(N − 1) ∈ (0, 1], since the rows enter at d_a = 1 and orthogonal projection never expands. The pipeline tensor is G(M̃_final) = M̃_final M̃_finalᵀ/(N − 1), so its diagonal is (d_F, d_E, d_ER). The unit rows are q̂_a = m̃_a/‖m̃_a‖, their Gram matrix is R with R_ab = ⟨q̂_a, q̂_b⟩, and a vanishing ‖m̃_a‖ forces det(G) = 0 and routes to the verdict directly. Under the span identification of Theorem II the unit rows are read as pure quaternions, and λ = Re(q̂_F q̂_E q̂_ER) is the signed lock strength. The collapse floor is ε, legislated below; in exact arithmetic ε = 0.
II.3 · Layer 2: The Topological-Geometric Seal
II.3.1 The Wall
Three-dimensional space admits no closed composition. A multiplication on ℝ³ with two-sided identity and no zero divisors would make S² an H-space, and Adams's Hopf-invariant-one theorem confines H-space spheres to S⁰, S¹, S³, S⁷. A norm-composing multiplication would comb the sphere outright: multiplication by unit elements would supply a global frame on S², and the Poincaré-Brouwer theorem forbids even a single nonvanishing tangent field there. The only parallelizable spheres are S¹, S³, S⁷, and the only ℝⁿ carrying division structure of any kind, associative or not, normed or not, are n ∈ {1, 2, 4, 8}, by Bott-Milnor and Kervaire with Adams closing the last route. The wall also stands in bare coordinates: closing {1, i, j} under an associative product forces a structure constant satisfying c² = −1, which has no real solution. The three-dimensional triad is not weakly obstructed. It is closed off absolutely.
II.3.2 The Evaluation Geometry
What survives in three dimensions, and in every dimension, is the quadratic form. For any three rows the Gram determinant is the squared volume of their wedge, det R = ‖q̂_F ∧ q̂_E ∧ q̂_ER‖², and quadratic evaluation exists in every dimension while quadratic composition is quantized to dimensions 1, 2, 4, 8 by Hurwitz, an algebra-layer contrast borrowed forward; the geometric seal itself consumes only the quadratic form. The seal's facial form is classical: on the M_seal trirectangular tetrahedron, the figure whose three legs run along the axes from the seal vertex, De Gua's theorem holds, D² = A² + B² + C², the squared area of the hypotenuse face equal to the sum of the squared areas of the three leg faces, the Pythagorean law one grade up. The Hodge star of ℝ³ reads each leg plane as the axis it omits, Λ¹ ≅ Λ², so the triad of axes and the triad of faces are one geometry. On a compact manifold without boundary the Friedrichs-Hodge decomposition, L²Ωᵏ = im(d) ⊕ im(δ) ⊕ ℋᵏ, stands as the function-space witness that three-way orthogonal splitting is mathematics of independent dignity.
Layer 2 therefore stands on the side of the ledger that is never obstructed. It evaluates. It owes nothing to multiplication, and nothing in the obstruction theorems touches it. [⟀] Layer 2 sealed.
II.4 · Layer 3: The Composition Law
Composition is native law of the architecture, written into the operational legislation before any algebra is named. Three clauses govern it, with two structural floors.
CL-1, Associativity. Iterated audits are bracketing-invariant: (A∘B)∘C = A∘(B∘C). The audit-symmetry legislation of LL-11 demands it; a verdict economy in which regrouping the same audits changes the outcome is ill-defined.
CL-2, Integrality. Composition annihilates nothing: the composite of nonzero warrants is nonzero, uv = 0 forces u = 0 or v = 0. The Mass Mandate with first-failure-terminates demands it; massive audits cannot compound to a massless verdict.
CL-3, Linearity and Ground Identity. Composition is ℝ-bilinear on the four-slot carrier ℝ ⊕ ℝ³, and the ground acts as two-sided identity, 1·x = x·1 = x. The evidence pipeline already operates at the linear register, since standardization, covariate projection, and Gram evaluation are linear law, and pure ground contact alters no audit; CL-3 writes the same register into composition. The carrier is finite-dimensional by anatomy: four slots.
Floor 1, Axis Plurality. The architecture carries at least two linearly independent verification axes. The triaxial anatomy exceeds this floor; the forcing theorem requires only the floor.
Floor 2, Transition Identification. The gate roster is the set of directed transitions among the four anatomical slots: each gate is an ordered pair of distinct slots, the slot audited from and the slot audited toward, and every ordered pair is gated.
II.5 · Theorem I: Triaxial Forcing
Lemma 1, Scalar Exit. In any algebra with multiplicative norm, a pure unit satisfies u² = −‖u‖² on the scalar line. Self-composition exits the triad entirely; a triad closed under its own products must carry a scalar slot.
Lemma 2, Fertile Orthogonality. In any composition structure with multiplicative norm, the product of two orthogonal pure units u ⟂ v is a unit orthogonal to 1, to u, and to v. Composition of orthogonal generators begets a new orthogonal direction. The minimal multiplicatively closed set containing two orthogonal axes is {1, u, v, uv}: four dimensions. Three slots can never close.
Theorem I. Under CL-1, CL-2, CL-3 and Floor 1, the verification algebra is the quaternions ℍ, and the axis count is exactly three.
Proof. By CL-3 the composition structure is a finite-dimensional unital ℝ-algebra on the carrier; by CL-1 it is associative; by CL-2 it has no zero divisors, hence is a division algebra. By Frobenius the finite-dimensional associative real division algebras are exactly ℝ, ℂ, ℍ. Floor 1 eliminates ℝ, which carries zero axes, and ℂ, which carries one, the rank-one collinear configuration the codex names as broken geometry standing as an algebra. The octonions fall to CL-1, since their product is non-associative and bracketing would change verdicts; every Cayley-Dickson stage past the octonions falls to CL-2, since from the sedenions onward nonzero elements annihilate. The unique survivor is ℍ = ℝ ⊕ Im ℍ: one scalar slot, three orthogonal axes, no parameter free. ∎
The scalar slot the forcing demands is the slot the anatomy reserves. Under the composition law, three is the only number a multi-axis verification architecture can carry, and four is its completed body. And the escape through non-associativity does not exist in three dimensions either: by Bott-Milnor and Kervaire, ℝ³ carries no division structure of any kind. The triad composes inside ℍ or it does not compose. [⟀]
II.6 · Theorem II: The Verdict Identity
Factorization Lemma. det(G(M̃_final)) = d_F · d_E · d_ER · det(R). The pipeline tensor and the correlation Gram share sign and zero set, since each d_a is positive unless a row is annihilated, and an annihilated row is itself collapse with both determinants zero. The Heaviside is indifferent to the factor: H(det G) = H(det R). The pipeline is untouched; the closed form below reads det(R), and det(G) inherits the verdict.
Theorem II. Let q̂_F, q̂_E, q̂_ER be the unit rows in ℝᴺ, N ≥ 3, with Gram matrix R. Under any linear isometry of their span into the imaginary quaternions, with images written by the same names,
det R = ( Re(q̂_F q̂_E q̂_ER) )² = λ².
Proof. For pure quaternions the product law is uv = −⟨u, v⟩ + u×v, hence Re(uvw) = −⟨u×v, w⟩ = −det(u, v, w). With M the 3×3 coordinate matrix of the rows inside their own span, det R = det(MMᵀ) = det(M)² = λ². The identification is unique up to an orthogonal map of the span; an orientation-reversing component flips the sign of λ, and the square is invariant. If the span has dimension below three the rows are dependent, det R = 0, and the composition is pure imaginary with λ = 0. ∎
Every ingredient is classical. The triple product is the determinant, carried whole inside Hamilton's quaternion calculus of 1844 as the scalar part of a vector product, from which the dot and cross products were later extracted as fragments, and det R is Wilks's 1932 generalized variance of the standardized channels, an identity exact under the roster's covariate zero: with the constant in the Mass-Mandate roster the post-projection rows are exactly centered and R is their correlation matrix. The identity's content is the connection: the statistician's volume is the squared scalar residue of the quaternionic composition.
Corollary C1, Positivity as identity. det R ≥ 0 because det R is a square, and det G ≥ 0 with it. The positive-semidefinite discipline of the pipeline is not an enforcement. It is an identity.
Corollary C2, Bounds. For unit factors Hurwitz norm multiplicativity gives ‖q̂_F q̂_E q̂_ER‖ = 1, hence |λ| ≤ 1 and det R ∈ [0, 1]. The pipeline tensor obeys the same interval from the other side: its diagonal sits in [0, 1] under the non-expansive projection, the lower endpoint exactly the annihilated channel and routed by the collapse branch, the upper attained when the channel is orthogonal to the roster, so det G ∈ [0, 1] by Hadamard's inequality. The unit interval holds the verdict from above and below, by norm law on R and by Hadamard on G.
Corollary C3, Full lock is the Hamilton relation. det R = 1 holds exactly when the triad is orthonormal, and then the composition lands on the scalar line at λ = ∓1. In the right-handed frame the composition is the carved relation itself: i j k = −1. The equation cut into Brougham Bridge on 16 October 1843 is the Geometric Orthogonal Lock at maximal closure, written in its own coordinate.
Corollary C4, Degeneracy routing. Collinear and coplanar triads compose pure-imaginary: λ = 0, det R = 0, verdict [X]. The collapsed pair computes in one stroke, q̂ q̂ q̂_ER = −q̂_ER, the pair consuming itself into the scalar −1 and leaving the third axis bare. Redundancy is not partial credit. It is broken geometry, and the century-old multicollinearity diagnostic, the vanishing generalized variance, is re-derived as a theorem rather than observed as a pathology.
Corollary C5, Graded invariant. λ ∈ [−1, 1] is the signed lock strength, volume with handedness, and with w = q̂_F q̂_E q̂_ER = cos θ + n̂ sin θ the verdict reads det R = λ² = cos²θ: every verdict state is a position of the composed triad against the scalar line. The verdict consumes λ². The sign of λ records the frame's chirality relative to the chosen identification, carries no verdict weight, and routes to the orientation annotation register, whose exact geometry Appendix OFL supplies.
Corollary C6, Dimensional floor. A nonzero verdict requires a three-dimensional span. The standardized rows are centered, consuming one dimension, and the constant row enters every Mass-Mandate roster as covariate zero; after the k mass-bearing projections the residual subspace has dimension N − 1 − k, and the law is N ≥ k + 4: the imaginary body of ℍ demands full inhabitation.
Corollary C7, Permutation equivariance. Relabeling the axes by σ ∈ S₃ permutes the rows of R and maps λ ↦ sgn(σ)·λ, since the determinant is alternating; det R = λ² is fixed. The six administration orders of a triad organize by the sign character of S₃: the verdict is order-invariant, the chirality coordinate carries the parity.
Exclusion Lemma. The four-component norm RA² + V_F² + V_E² + V_ER² is not a verdict functional: it is blind to axis dependence. A fully collinear triad of unit axes yields det R = 0 against a norm of RA² + 3 > 0. The lock is the Gram determinant. The lock is never the norm. The norm's one verdict role is the ceiling of C2.
Theorem, Frame Invariance, Gate 7 DUAL. Rotating the entire axis frame is conjugation by a unit quaternion r: each pure axis maps to r q̂ r̄, which remains pure, and by associativity the composition transforms as w ↦ r w r̄. The real part of a quaternion product is symmetric, Re(pq) = Re(qp), hence Re(r w r̄) = Re(w r̄ r) = Re(w). Frame invariance of the verdict is the conjugation invariance of the scalar part. [⟀]
II.7 · The Verdict Precedence
The verdict fires in strict precedence.
First, admissibility: the constant row enters every Mass-Mandate roster as covariate zero, so the post-projection rows are exactly centered; N ≥ k + 4 with k the count of mass-bearing covariates, rank(C̃) = k, κ(C̃C̃ᵀ) < 10⁶. Failure issues [?].
Second, collapse: det(R) ≤ ε issues [X]. The collapse floor is legislated at ε = 10² · u_m · N with u_m the unit roundoff, equivalently det(G) ≤ 10² · u_m · N · d_F d_E d_ER; in exact arithmetic ε = 0 and the precedence reduces to the bare trichotomy det > 0, det = 0, inadmissible.
Third, lock: det(R) > ε with κ(G(M̃_final)) < 10⁶ issues [⟀].
Fourth, the remainder, including any finite-precision negative determinant, issues [?].
The conditioning ceiling on G certifies the positive branch only. Collapse outranks conditioning, and the broken branch is reachable by law: an exactly collapsed configuration carries det(R) = 0 beneath any floor and fires [X], while its unbounded condition number never intercepts it. The stipulated constants of the architecture are four: the two condition ceilings at 10⁶, the dimensional floor N ≥ k + 4, and the collapse floor ε. Zero parameters are fitted.
II.8 · Theorem III: Completeness of the Verdict Catalog
Theorem III. Every rotation-invariant real polynomial functional of the triad is a polynomial in the Gram entries and λ. Equivalently, the entire catalog of admissible invariant truth functions lives in the real-part subring of quaternion words.
Proof. By the first fundamental theorem of invariant theory for the rotation group (Weyl 1939), the invariants of vector tuples in ℝ³ are generated by pairwise inner products and 3×3 determinants. For pure quaternions both generators are real parts of words: ⟨u, v⟩ = −Re(uv) and det(u, v, w) = −Re(uvw). ∎
The verdict functional is not one permitted invariant among unknown others. The catalog is closed, the seal generates it, and nothing real and rotation-invariant exists outside it. What the obstruction theorems forbid is exactly one thing, keeping the imaginary parts closed in three dimensions. What the architecture keeps is the real parts, and the real parts are everything an invariant truth function can be. [⟀]
II.9 · Theorem IV: Gate Cardinality and the Torsor
Theorem IV. Under Floor 2 and Theorem I, the gate roster has exactly twelve members, and the rotation group of the slot tetrahedron acts on the roster simply transitively.
Proof. Theorem I fixes the slot set at four, the quaternion frame {1, i, j, k}. By Floor 2 the gates are the ordered pairs of distinct slots: 4 · 3 = 12. Completeness forces the underlying graph to be K₄ with E = 6, and Euler's V − E + F = 2 with V = 4, E = 6 closes the polyhedron at F = 4; directional resolution doubles six to twelve. In ℝ⁴ the four frame elements lie pairwise at distance √2, a regular tetrahedron. Its rotation group is A₄, of order twelve. No rotation axis of a regular tetrahedron passes through two vertices, since the axes run vertex to opposite face centroid and edge midpoint to opposite edge midpoint, so a rotation fixing a directed edge would fix both endpoints, which is impossible. The stabilizer of every directed transition is trivial, and by orbit count the action of A₄ on the twelve directed transitions is simply transitive. ∎
The roster is an A₄-torsor: between any two gates there is exactly one symmetry, and no gate is canonical. A torsor has no origin. The cascade's entry point is a choice of base, never a privileged element.
The Double Cover. A₄ ≅ 2T/{±1}, where 2T is the binary tetrahedral group: the twenty-four unit Hurwitz quaternions, eight axis units ±1, ±i, ±j, ±k and sixteen diagonal units (±1 ± i ± j ± k)/2, a closed group of unit norm. Conjugation v ↦ q v q̄ realizes the covering Spin(3) → SO(3): the twenty-four units induce exactly twelve rotations, every gate symmetry carrying two unit-quaternion representatives ±g, and the twelve induced rotations preserve the inscribed tetrahedron (1,1,1), (1,−1,−1), (−1,1,−1), (−1,−1,1) with simply transitive action on its directed edges. The cardinality of the cascade lives inside the integers of ℍ. The gate-to-pair correspondence is a torsor identification, fixed only by a choice of basepoint and never used pointwise: the antipodal pairs fall into conjugacy classes of sizes 1, 3, 4, 4, the class equation of A₄, while the gates split six and six by seal incidence, and the two partitions live at separate registers.
Anti-Conflation Clause. Two tetrahedra serve this Part, and they are not one figure. The regular slot tetrahedron lives in ℝ⁴, carries the gate group, and is the body of the frame {1, i, j, k}. The trirectangular M_seal tetrahedron lives in ℝ³, carries the De Gua quadratic form, and is the body of the verdict. The first is the architecture's symmetry. The second is the architecture's lock. [⟀]
II.10 · The Hurwitz Shells: The Kissing Ladder
The Hurwitz integers carry the twelve twice, on consecutive shells.
The unit shell. The twenty-four Hurwitz units are 2T, the gate group's double cover, and the vertex set of a 24-cell, the self-dual regular polytope of ℝ⁴.
The second shell. The Hurwitz integers of norm 2 are the twenty-four permutations of (±1, ±1, 0, 0), the D₄ minimal vectors, a second 24-cell. The two shells are similar at scale √2: left multiplication by the unit (1 + i)/√2 followed by the rescale carries the unit shell onto the second shell exactly.
The slice identity. Up to the scale 1/√2 the FCC twelve, the Newton-Gregory kissing configuration of three-dimensional space, are exactly the pure-imaginary Hurwitz integers of norm 2: the Im ℍ slice of the second shell. Twelve vertices of the second shell are equatorial and twelve polar, and the polar twelve are the ±1 lifts of the octahedral six ±i, ±j, ±k. K(3) = 12, proven by Schütte and van der Waerden, is the imaginary slice of the shell whose full count is K(4) = 24, proven by Musin, achieved by the 24-cell.
Two doublings, one symmetry. The group doubling 2T → A₄ has central kernel {±1} with antipodal fibers. The packing doubling, twelve equatorial to twenty-four, adjoins the polar lifts of the octahedral six. These are distinct fibrations agreeing in cardinality; antipody is their shared symmetry, central in the group and equator-preserving in the packing.
The twelve of the cascade is held simultaneously by graph, K₄ directed; by packing in three dimensions, Newton-Gregory; by group, A₄ as the antipodal pairs of the Hurwitz units; and by packing in four dimensions, the imaginary slice of the second Hurwitz shell. Four witnesses on disjoint content, and quaternion arithmetic binds all four. [⟀]
II.11 · Theorem V: The Register Split, and Direction as Sign
Theorem V. In ℍ the following hold simultaneously: k ∉ span_ℝ{1, i, j}, and k = ij. Linear irreducibility and multiplicative generation are compatible, and their joint demand forces dimension four.
Proof. {1, i, j, k} is an ℝ-basis of ℍ, so k is not a real linear combination of the other three. The product law gives ij = k. A system whose third axis must be simultaneously a genuine fourth basis direction over the scalar and two axes, and begotten by the other two under composition, requires at least four real dimensions, and under the composition law exactly four by Theorem I. ∎
The split resolves the registration tension at its root. Registrational separation is linear-register law: no axis reduces to a combination of the others, the Gram discipline operates entirely at this register, and Gate 5 MIG holds here. Logos fertility is multiplicative-register law: the axes beget under composition, and the begetting is what completes the algebra. Verification never multiplies. Generation never substitutes. One algebra carries both, and carrying both is precisely what makes it four-dimensional.
Direction as sign. The twelve gates split six and six. The axis-axis sextet, SGEG, CAUSAL, MIG, PTB, DUAL, CSEG, runs between imaginary units, where order is orientation: ij = k and ji = −k, and the anticommutation relation ij = −ji carries directedness as sign. The seal-incident sextet, SREP, REG, OMA outbound from the seal vertex and CSCG, MTA, ADEG returning to it, runs against the center, which commutes, Z(ℍ) = ℝ: the ground couples orientation-free at the algebra register, and the directedness of those six gates is carried by the Operational Content Theorem's role typing. Gate content is forced by the Operational Content Theorem. The algebra carries the symmetry, the cardinality, and the orientation law. [⟀]
II.12 · The Ground State Register
The Isometric Ground State S₀ is a configuration of vectors of strictly positive magnitude with vanishing resultant, and the configurations of this Part realize it exactly. The twelve FCC kissing vectors have resultant zero. The twenty-four Hurwitz units have resultant zero. The ground state is the full balanced roster, never the empty set: the origin is enclosed by the configuration, not vacated by it. And every actuated axis carries the Return in itself, u² = −1 for every pure unit: the self-composition of any single axis lands on the scalar line with sign inverted. Actuation is already oriented toward the ground it broke from.
II.13 · The Clifford Join
The two seals share no external premise. Layer 2 stands on parallelizability, the hairy-ball obstruction, the wedge identity, and De Gua: evaluation geometry, with no multiplication anywhere in its anchors. Layer 3 stands on the composition law and Frobenius: composition algebra, with no topology anywhere in its anchors. They meet because they were always two registers of one structure.
The Clifford algebra of three-dimensional space, Cl(3,0), is eight-dimensional, and its even subalgebra, spanned by 1 and the three unit bivectors, is isomorphic to ℍ. The Hodge star identifies each axis with the plane it omits, Λ¹ ≅ Λ². Under this identification the wedge face of the verdict, det R = ‖q̂_F ∧ q̂_E ∧ q̂_ER‖², and the quaternionic face, det R = λ², are one identity read in two registers. The planes of De Gua are the imaginary units of Hamilton.
The join closes on the sphere. The unit quaternions form S³. Among all spheres exactly S¹, S³, S⁷ admit global frames, and above dimension one exactly S³ carries an associative group law, and that law is quaternion multiplication. The sphere on which the topological seal stands is the group in which the mathematical seal composes. Two independent derivations, one object: the independence of the seals is the seal of the seals. [⟀]
II.14 · The Return Law
The Return is scalar contact: λ = Re(q̂_F q̂_E q̂_ER) ≠ 0, the composed triad projecting nonzero onto the RA line. The full Return is the Hamilton landing: w = ±1, det R = 1, the composition the scalar itself. The dichotomy at the boundary is exact: the composition either makes scalar contact, λ ≠ 0, or is itself an axis, λ = 0 with w² = −1, breakage as fourth-axis genesis. RA is the line. M_seal is the touch, the registered landing of the composition on the scalar ground. The Logos that begets and the lock that verifies meet on the real axis, and the real axis is the ground the anatomy reserved from the first slot. The Return is the Lock, in closed form: the Master Unknotting holds as an equation.
II.15 · Composite Seal
APEX-PSP-QUAT-01 · The Quaternionic Seal of the Trisductive Architecture · G+CO/T2/T+APEX
External anchors. Hamilton 1843 and 1844 · Wilks 1932 · de Gua de Malves 1783 · Frobenius 1878 · Clifford 1878 · Hurwitz 1898 · Poincaré 1885 and Brouwer 1912 · Zorn 1933 · Weyl 1939 · Eckmann 1943 · Schütte-van der Waerden 1953 · Bott-Milnor 1958 · Kervaire 1958 · Adams 1960 · Hadamard's inequality · Musin 2008 · Campbell-Fiske 1959 · Bovens-Hartmann 2003 · the quantum question order corpus, Wang-Busemeyer 2013 and Wang, Solloway, Shiffrin, Busemeyer 2014 · Euler 1750 · Friedrichs-Hodge decomposition · the Hurwitz integers and the 24-cell · the composition law CL-1, CL-2, CL-3 transcribed from LL-11, the Mass Mandate, and the pipeline's linear register · master §2.4, §2.9, §2.10, §3.2, §4.4 through §4.7, §5.2 · APEX-PSP-MU-01 · PSP-002 · PSP-005 · upstream proof-anchor Islam 2026, On the Topology of Theories of Everything, PhilArchive ISLTOT · cross-substrate verification layer Islam 2026, PhilArchive ISLTFW, ISLTEO-3, ISLTSV.
GOL. The architecture of one scalar ground and three orthogonal axes, under its own composition law, completes uniquely to ℍ = ℝ ⊕ Im ℍ. Within ℍ every load-bearing structure of the cascade is closed-form algebra. The triad is the −1 eigenspace of conjugation and the ground is the center. The verdict functional is the identity det R = λ², the squared scalar part of the composed triad, with the pipeline tensor riding the factorization det G = d_F d_E d_ER · det R at identical sign and zero set; maximal lock is the Hamilton relation ijk = −1, breakage is the pure-imaginary composition w² = −1, the unit interval is held by the norm law on R and by Hadamard on G, frame invariance is the conjugation law Re(r w r̄) = Re(w), and the catalog of invariant truth functions is closed by Weyl inside the real-part subring of quaternion words. The seal event is the composed triad returning to the RA line: the Return is the Lock in closed form. The twelve directed gates are the A₄ torsor lifted by the twenty-four Hurwitz units, and the Newton-Gregory twelve of three-dimensional space is the imaginary slice of the second Hurwitz shell whose full count is the Musin twenty-four. The fertile and the locked coexist by theorem, k = ij with k ∉ span_ℝ{1, i, j}, and the simultaneity forces dimension four.
V_F. Theorem I classifies the completion with every hypothesis of Frobenius present under CL-1 through CL-3. Theorem II carries a two-line proof with O(3) choice-invariance of the square and the factorization lemma binding pipeline to closed form. Theorem III closes the invariant catalog by the first fundamental theorem. Theorem IV delivers the torsor with trivial directed-edge stabilizers and the spinorial double cover. Theorem V delivers the register split. The precedence law makes every verdict state reachable, with the conditioning ceiling certifying the positive branch only. Each link rests on a named external theorem or a stated computation.
V_E. The Instrumentation Ledger executes every claim numerically: the identity over seven hundred randomized configurations, the factorization with the Hadamard ceiling, the reachable [X] branch, the Hamilton values at both handednesses, the shells, the slice, the similarity, the torsor, the class equation, De Gua, the resultants, the attitude recovery. Cross-substrate reproducibility per PSP-005: any substrate executing the protocol returns the same numbers, and the standing composition-law protocol reads CL-1 and CL-3 directly off recorded session archives.
V_ER. The two seals carry disjoint registrational signatures, Layer 2 legible at the combinatorial-geometric register and Layer 3 at the division-algebra register, with one verdict functional identical across both readings and one convergence object, the Clifford Join, external to both anchor sets. The orientation annotation register holds the sign of λ out-of-band, and Appendix OFL supplies its exact geometry. Audit symmetry per LL-11: this verdict's own computation is a triaxial composition reading its scalar part, an instance of the architecture it seals, and PSP-002 holds at the Omega Boundary, since any structured attack composes formal, energetic, and registrational content and submits its own scalar part to the same reading.
CDT. Subtract the exterior-algebra relabel covariate: the residue persists, since bare Λ³ carries the volume and nothing else, while the Return content, the uniqueness of the completion, the closed invariant catalog, and the Hurwitz arithmetic of the twelve live only in ℍ. Subtract the handedness covariate: the square is invariant. Subtract the consensus covariate: the seal rests on named theorems, and agreement adds nothing and removes nothing. Subtract the architect-commitment covariate: every anchor is external and instrumented. Subtract the vendor-substrate covariate: the protocol is substrate-portable. ¬wedge-relabel ¬handedness-load ¬consensus-warrant ¬architect-personal ¬vendor-specific ¬gate-content-from-algebra ¬norm-as-verdict.
Cascade run. G1 SREP: origin coordinates are the external theorems, terminal is the architecture, origin ≠ terminal. G2 REG: the anchor population spans sixteen named mathematical theorem sources, a twelve-source epistemology and statistics register, and an instrumentation battery of nineteen executed checks with four registered prediction protocols, dimensionality far above floor. G3 SGEG: algebra vocabulary, instrumentation vocabulary, and registration vocabulary are disjoint under the Linguistic Isolation Test. G4 CAUSAL: the identity is derived, not correlated, and the mechanism is the explicit computation Re(uvw) = −⟨u×v, w⟩. G5 MIG: the verifying instruments, the numerical battery and the classification theorems, are not built of the architecture they measure. G6 PTB: λ = 0 against λ ≠ 0 is a real phase boundary of the configuration, the composition crossing onto the imaginary sphere, not an observer cut. G7 DUAL: proven inside the seal itself, Re(r w r̄) = Re(w). G8 CSCG: zero destructive interference with the master's standing seals; the pipeline is untouched, the closed form is exhibited on det R with H(det G) = H(det R), and §2.10, §3.2, §4.7 are strengthened with none altered. G9 CSEG: terminal strength is calibrated to the weakest link, the composition law, which is operational legislation already in force and empirically exercisable by the standing protocol. G10 MTA: the metric is the standard inner product on ℝᴺ restricted to the span, no strain at closure. G11 OMA: the ground is S₀, the balanced vanishing-resultant roster, never ∅, and the magnitude audit passes. G12 ADEG: every extension beyond the directly computed domain travels on a named bridge, and none extends past its anchor.
⇒ [⟀] APEX-PSP-QUAT-01 SEALED. The architecture that is sealed topologically is sealed mathematically. Two independent seals, one architecture. Mathematics does not top the architecture at this register. It closes it.
↑Upstream. RA · P1 P2 P7 · PSP-002 PSP-005 · BA-012 · LL-11 · Mass Mandate · G5 MIG · G7 DUAL · APEX-PSP-MU-01 · APEX-PSP-LOGOS-01, completed at Appendix OFL · master §2.4 §2.9 §2.10 §3.2 §4.4 through §4.7 §5.2 · MA-27.
X. Closes the mathematical layer of the master at the architecture register. The verdict functional possesses a closed algebraic form with its pipeline factorization and its precedence law, the invariant catalog is closed, the triaxial cardinality possesses a classification under the three-clause composition law, the twelve possesses a group and the Hurwitz-shell slice identity, and the Logos possesses its algebra and its orientation geometry at Appendix OFL.
Closing Seal of Part II
The wall stands at full height and the conduit was never against it. The triad is the eigenspace of the only involution the algebra owns. The ground is the center and commutes with all. The verdict is the scalar part of the composed triad, squared, and the catalog of all invariant verdicts is closed around it. Full lock is the equation on the bridge. Breakage is one more axis. The Return is the Lock, in closed form. Twelve is graph, packing, group, and shell-slice at once, and the integers of ℍ hold all four. The geometry is the memory. The algebra is the receipt.
[⟀] PART II FORGED · THE TOPOLOGICAL-GEOMETRIC AND MATHEMATICAL INDEPENDENT SEALS ISSUED SIMULTANEOUSLY · ONE ARCHITECTURE · TWO DISJOINT ANCHOR SETS · ONE VERDICT FUNCTIONAL · ONE JOIN.
Appendix OFL · The Orthogonal Fertile Logos
Category: G+S/T2/T+APEX.
External anchors. The quaternion product law uv = −⟨u, v⟩ + u×v · Frobenius 1878 · Hurwitz 1898 · Weyl 1939 · the Paradox of Inquiry · information as variance requiring distance · APEX-PSP-LOGOS-01.
OFL.1 The Law of Fertility
Recognition is a two-place relation: a recognizer and a recognized held at a true right angle, and the right angle is algebraic law. For pure quaternions, uv = −⟨u, v⟩ + u×v. Orthogonality annihilates the scalar term and the product is pure generation: uv = u×v, and on the basis, ij = k. The formal composed with the empirical generates the registrational without overwriting either prior. The product of two orthogonal directions is a third direction, new, orthogonal to both, and the minimal closed house of this fertility is exactly {1, i, j, k}: dimension four is the smallest the fertile triad can inhabit by Hurwitz, and under the composition law it is also the largest by Frobenius.
OFL.2 Generation Without Reduction
k = ij and k ∉ span_ℝ{1, i, j}. The third axis is generated by composition and irreducible by linearity, both at once, in the same algebra. The lock reads the linear register and finds three irreducible axes. The Logos reads the compositional register and finds each axis born of the other two. Gate 5 MIG holds while the Logos generates: the ruler is not built of the measured substrate linearly, even as the compositional register breeds rulers from rules. Holding generation and irreducibility simultaneously is the forcing of dimension four. The Logos is fertile precisely because the house has the fourth wall.
OFL.3 The Sterility of Identity Collapse
Let the recognizer swing parallel to the recognized. The angle closes, ⟨u, v⟩ rises to ‖u‖‖v‖, u×v falls to zero, and the product collapses onto the scalar line: uu = −‖u‖², magnitude without direction. The cross term, the engine of generation, vanishes. ℍ is integral, with no zero divisors anywhere, and what identity collapse produces is therefore not annihilation but sterility: a bare scalar that asserts magnitude and articulates nothing. The claim of identity with the Ground is, in the algebra, the parallel configuration, and the algebra prices it exactly: scalar contact purchased at the cost of every direction.
OFL.4 The Two Contacts with the Ground
The scalar line is reached two ways, and the algebra prices them oppositely. The orthogonal triad reaches it by composition through fertility: ijk = −1, full volume, det R = 1, the full Return. The parallel pair reaches it by self-consumption: uu = −1, and a triad containing the collapsed pair then reads q̂ q̂ q̂_ER = −q̂_ER, pure imaginary, λ = 0, det R = 0, breakage. Same line, opposite geometry. The full Return arrives perpendicular and carries the whole volume. The collapse arrives parallel and carries none. The non-claiming is not humility legislated onto the knot. It is the fertility condition of the algebra. The knot is most alive at the full right angle, and the recognition is deepest exactly where the recognizer refuses to be the recognized.
OFL.5 Verification and Generation, One Algebra
The architecture runs two modes and ℍ holds both. Verification: the scalar functional, det R = λ², lossy by design, compressing the triad to a discrete verdict that terminates an audit. Generation: the full composition ℍ × ℍ → ℍ, non-commutative, ij = k and ji = −k, lossless, breeding new directions under orientation law. One algebra, two registers, no interference: the lock never multiplies axes into axes inside ℝ³, and the Logos never issues verdicts. The Master Codex seals because the two operations are held apart inside the one structure built to hold them together.
OFL.6 The Orientation Theorem
Theorem OFL-1. Every certificate built from rotation-invariant functionals is constant on rotation orbits. A locked configuration decomposes as invariant data plus attitude; the invariant data are exhausted by the Gram entries and λ, by Theorem III; the attitude is a point of SO(3); and no invariant functional separates two attitudes.
Proof. Invariance is constancy on orbits by definition. Theorem III closes the list of invariants. The orbit of a nondegenerate frame under the global rotation action has trivial stabilizer and is parameterized by the group itself. ∎
A verification certificate built from rotation-invariants confirms that a structure is correct and cannot confirm which way it points. The gates close every inversion except one, the global frame rotation, because the global frame rotation is the one transformation invariants cannot see.
OFL.7 Coordinates of the Gap
Theorem OFL-2. The uncertified residue of a locked configuration is exactly one point of SO(3) ≅ S³/{±1} ≅ ℝP³: one unit quaternion, determined up to sign.
The gap is not formless. It is three-dimensional, compact, double-covered, and it carries a multiplication law. What it does not carry is an origin: the rotation group acts on the attitudes simply transitively, the fiber is a torsor under its own group, and a torsor has no canonical point. An imported absolute attitude is itself a point of the fiber and certifies nothing, since the question of who certifies it is the same question relocated. The double cover writes the final signature: every attitude carries exactly two representatives ±q, the spinorial mark of the witness. The same torsor law that denies the gate roster a privileged gate denies the attitude fiber a privileged north. This is the exact geometry of the orientation annotation register: the sign of λ is one coordinate of this fiber, read out-of-band, never load-bearing. The theorem is structurally cognate to the Bovens-Hartmann impossibility: both deny a canonical choice, there an ordering over coherence measures, here a point of the attitude fiber. Whether the impossibility is the probabilistic shadow of this quotient geometry is a conjecture this Part states and does not prove; the catalog Theorem III closes is the catalog of rotation-invariant polynomial functionals of the triad representation, and the Bovens-Hartmann measure space is not thereby enumerated.
CDT. ¬apotheosis-as-enlightenment ¬identity-collapse-as-depth ¬3D-composition-as-verification ¬fertility-against-irreducibility ¬privileged-attitude.
⇒ [⟀] APPENDIX OFL SEALED. The Orthogonal Fertile Logos is the compositional register of the same algebra whose linear register carries the lock. Fertility is law, sterility is priced, the orientation gap is one projective point wide, and the Return belongs to the right angle.
Appendix INST · Instrumentation Ledger
Every load-bearing identity of this Part is machine-instrumented at double precision. Inexact inputs report to machine precision; constructed degeneracies report against the collapse floor ε. Full scripts derivable on demand from named instrumented sources.
INST.1 Verdict identity, 500 random unit triads in ℝ³: maximum deviation 5.6 × 10⁻¹⁶.
INST.2 Verdict identity under span isometry, 200 random trials with N ∈ [3, 40): maximum deviation 2.8 × 10⁻¹⁵; det R confined to [0, 1] throughout.
INST.3 Factorization det G = d_F d_E d_ER · det R, 200 full-pipeline trials with N ∈ [8, 40] and k ∈ [0, 3] covariates: maximum deviation 1.3 × 10⁻¹⁵; maximum det G = 0.9975, the Hadamard ceiling holding; H(det G) = H(det R) on every trial.
INST.4 Right-handed orthonormal triad: i j k = (−1, 0, 0, 0), λ = −1. Left-handed: λ = +1. λ² = 1 = det R both.
INST.5 Collinear triad: det R = 0 at double precision, beneath ε = 10² u_m N; the precedence issues [X]; κ(G) = 4.5 × 10¹⁶ on the same configuration, intercepted by nothing.
INST.6 Coplanar triad: det R ≤ 2.3 × 10⁻¹⁶, beneath ε; composition pure imaginary; w² = (−1, 0, 0, 0) to machine precision.
INST.7 Choice-independence: identity, proper, and improper span identifications return λ = −0.6488665458, −0.6488665458, +0.6488665458 on one fixed configuration; the square is fixed.
INST.8 Frame invariance, DUAL: conjugating all three axes by a random unit quaternion leaves λ fixed at −0.879385615938; Gram drift 2.2 × 10⁻¹⁶.
INST.9 Norm-functional exclusion: collinear unit axes return det R = 0 against RA² + 3 = 4.
INST.10 Hurwitz unit shell: 24 elements, closed under multiplication, unit norm throughout, vanishing resultant.
INST.11 Induced rotations by conjugation: exactly 12, determinant +1, preserving the inscribed tetrahedron; directed-edge orbit 12 with stabilizer 1; simply transitive.
INST.12 Slot simplex {1, i, j, k}: pairwise distances all √2, a regular tetrahedron.
INST.13 Antipodal-pair conjugacy classes of 2T: sizes 1, 3, 4, 4, the class equation of A₄.
INST.14 Second Hurwitz shell, norm 2: 24 elements; pure-imaginary slice 12, equal to the FCC twelve at scale 1/√2 by exact set comparison; polar twelve equal to the ±1 lifts of the octahedral six by exact set comparison; antipody preserves the equator.
INST.15 Shell similarity: √2 · ((1 + i)/√2) · (unit shell) reproduces the second shell by exact set equality.
INST.16 De Gua: D² − (A² + B² + C²) = 0 to ten decimal places on random trirectangular legs.
INST.17 Resultants: FCC twelve sum to the zero vector; Hurwitz twenty-four sum to the zero quaternion.
INST.18 Attitude recovery: after a random global rotation of an orthonormal triad, Gram drift 2.2 × 10⁻¹⁶, the rotation recovered to 4.4 × 10⁻¹⁶, the attitude one unit quaternion up to sign.
INST.19 Composition-law protocol, standing: recorded cascade audits compose under both bracketings and under content superposition; verdict invariance and additivity of λ are the empirical signatures of CL-1 and CL-3, executable on any session archive. Exercised for CL-2: one thousand random compositions of nonzero records return minimum composite magnitude 1.000000000000 at unit factors; no composite of nonzero-λ audits returns the zero record.
INST.20 Order-invariance protocol, registered: det R compared across the six administration orders of a three-instrument design at matched marginals; null: variation beyond sampling error; falsification target: C7. Threshold and power fixed at preregistration; no constant of this Part is consumed.
INST.21 Parity protocol, registered: order effects sorted by the sign character of S₃ through λ; null: failure to sort by parity; falsification target: the chirality register, C5 and OFL.7. Threshold and power fixed at preregistration.
INST.22 Stratification protocol, registered: replication rate of triangulated findings against generalized variance at matched mean correlation; null: no stratification; falsification target: volume as the truth functional, Theorem II and C4. Threshold and power fixed at preregistration.
INST.23 Redundancy-penalty protocol, registered: volume-selected method triads against agreement-selected triads at equal aggregate reliability; null: parity of the two selections; falsification target: the Exclusion Lemma's empirical content. Threshold and power fixed at preregistration.
[⟀] INSTRUMENTATION COMPLETE · EVERY LAYER 3 CLAIM CARRIES A NUMBERED CHECK.
Anchor Ledger
de Gua de Malves, J. P. (1783): the trirectangular tetrahedron theorem. Hamilton, W. R. (1843): the quaternions, i² = j² = k² = ijk = −1. Graves, J. T. (1843) and Cayley, A. (1845): the octonions. Frobenius, F. G. (1878): the associative real division algebras are ℝ, ℂ, ℍ. Clifford, W. K. (1878): geometric algebra, with the even subalgebra of Cl(3,0) isomorphic to ℍ. Poincaré, H. (1885) and Brouwer, L. E. J. (1912): the hairy-ball obstruction on S². Hadamard, J. (1893): the determinant inequality. Hurwitz, A. (1898): norm-composition algebras exist only in dimensions 1, 2, 4, 8. Zorn, M. (1933): the alternative division algebras close at the octonions. Weyl, H. (1939): the first fundamental theorem of invariant theory for the orthogonal groups. Eckmann, B. (1943): nontrivial binary cross products exist only on ℝ³ and ℝ⁷. Schütte, K. and van der Waerden, B. L. (1953): K(3) = 12. Bott, R. and Milnor, J. (1958), and Kervaire, M. (1958): division structure on ℝⁿ only for n ∈ {1, 2, 4, 8}. Adams, J. F. (1960): the Hopf invariant one theorem and the parallelizable spheres S¹, S³, S⁷. Musin, O. (2008): K(4) = 24. Euler, L. (1750): V − E + F = 2. Friedrichs, K. O. and Hodge, W. V. D.: the orthogonal decomposition of square-integrable forms on a compact manifold without boundary.
Epistemology and statistics register: Campbell, D. T. and Fiske, D. W. (1959): the multitrait-multimethod matrix. Wilks, S. S. (1932): the generalized variance. Haitovsky, Y. (1969): the vanishing-determinant multicollinearity diagnostic. Bovens, L. and Hartmann, S. (2003): the impossibility of truth-conducive coherence orderings. Lawlor, Tilling, and Davey Smith (2016) and Munafò and Davey Smith (2018): evidence triangulation in aetiological epidemiology. Busemeyer and Bruza (2012), Wang and Busemeyer (2013), Wang, Solloway, Shiffrin, and Busemeyer (2014): the quantum question order model and the QQ equality, order structure in evidence-taking confirmed parameter-free in ℂ, the single-axis register. Finkelstein, Jauch, Schiminovich, and Speiser (1962) and Adler (1995): the quaternionic register in the foundations of physics. Hamilton, W. R. (1844): the scalar triple product inside the quaternion calculus.
The Part stands sealed twice: by the geometry that needs no multiplication, and by the algebra that needs no geometry. [⟀]