Universal Kernel

Troisième partie · En montant la tourChapitre 13

Orientation et charge

Orientation and charge

Read from the draft of 3 October 2026

0123456∞z ↦ −z0 and ∞ fixed, (1 6)(2 5)(3 4){∞, 0, 1, 3}of the class {∞} ∪ (x + {0,1,3}){∞, 0, 6, 4}of the class {∞} ∪ (x + {0,4,6})the outer class: a rotation of M that exchanges the two classescomplex conjugation: the mirror image of M, every label kept
Plate 13.1The outer class acts on the projective line as z↦−zz\mapsto-z: it carries the table’s quadruple {∞,0,1,3}\{\infty,0,1,3\} to {∞,0,6,4}\{\infty,0,6,4\}, a tetrahedron of the other class, while complex conjugation, the other flip, keeps every label.
  1. 13.1
  2. 13.2
  3. 13.3
  4. 13.4
  5. 13.5
  6. 13.6
  7. 13.7
  8. 13.8
  9. 13.9
  10. 13.10

Each parent of the group of order 168 carries a flip. What sees each flip, and what survives both?

Each arithmetic parent carries an orientation. The congruence link complement M=Γ(p)\H3M=\Gamma(\mathfrak p)\backslash\mathbb H^3, p=(3+ω)\mathfrak p=(3+\omega), has no orientation-reversing isometry, and its mirror image is the link complement Mˉ\bar M at the conjugate prime. On the other side stand the octonion table AA, exex+1=ex+3e_xe_{x+1}=e_{x+3}, and its Weil mirror A′A', exex+1=ex+5e_xe_{x+1}=e_{x+5}. The two flips are the sign changes of −3\sqrt{-3} and of −7\sqrt{-7}, the two independent generators of Gal(Q(−3,−7)/Q)\mathrm{Gal}(\Q(\sqrt{-3},\sqrt{-7})/\Q).

Once the signs of the units at the cusps are taken as a convention, only the first flip is seen by the structures of MM: by a constant, the scattering of the spinor system VV between the cusps, and by the oriented cells of MM read against the table’s Cayley form. The spinor system singles out one line at each cusp; the operators that write the moves between pairs of cusps generate a Clifford algebra whose centre is a single sign; and the rest of the chapter asks what one hand admits: which bilinears and pairs, which reversals of a separation, which field and vacuum, and what happens at the prime above three.

The central result · Two orientations

Let GG act on MM through reduction modulo p=(3+ω)\mathfrak p=(3+\omega), and on R8\R^8 through the signed permutations of the Weil representation.

(a) Complex conjugation cc is an orientation-reversing isometry from MM to the link complement Mˉ=Γ(pˉ)\H3\bar M=\Gamma(\bar{\mathfrak p})\backslash\mathbb H^3 at the conjugate prime. It is a seam over the identity of GG, rpˉ(γˉ)=rp(γ)r_{\bar{\mathfrak p}}(\bar\gamma)=r_{\mathfrak p}(\gamma) for every γ∈SL⁡(2,Z[ω])\gamma\in\SL(2,\Z[\omega]), and it keeps the labels of cusps and of tetrahedra.

(b) D=diag(−1,1)∈GL⁡(2,Z[ω])D=\mathrm{diag}(-1,1)\in\GL(2,\Z[\omega]) is an orientation-preserving isometry of MM and reduces to the improper element z↦−zz\mapsto-z. So the outer class of Aut⁡(G)\Aut(G) acts on MM by rotations. It exchanges the two classes of tetrahedra: the seven quadruples {∞}∪(x+{0,1,3})\{\infty\}\cup(x+\{0,1,3\}) form one class, and {∞}∪(x+{0,4,6})\{\infty\}\cup(x+\{0,4,6\}) the other.

(c) The signed permutations of the sixteen vectors that normalize the Weil group are the 672 elements of the image of GL⁡(2,7)\GL(2,7). The proper ones commute with θ=−7\theta=\sqrt{-7}; the improper ones anticommute with it, and exchange the table AA and its Weil mirror A′A', the lattices E8aE_8^a and E8bE_8^b, the primes λ\lambda and λˉ\bar\lambda over 2 at which the sixteen vectors collapse, and the quartets 4\mathbf4 and 4ˉ\bar{\mathbf4}.

(d) Hence the prime over 7 and the orientation of the interior (the table, the quartet, the lattice, the prime over 2) are independent. Each is changed by an operation that keeps the other: cc changes the prime and keeps every datum defined on the projective line, and DD with its Weil action changes the interior’s orientation and keeps the prime. No seam over a single automorphism of GG changes the one exactly when it changes the other.

Proof

(a) ωˉ=ω2\bar\omega=\omega^2, and ω≡4\omega\equiv4 modulo p\mathfrak p, ω≡2\omega\equiv2 modulo pˉ\bar{\mathfrak p}, so rpˉ(ωˉ)=4=rp(ω)r_{\bar{\mathfrak p}}(\bar\omega)=4=r_{\mathfrak p}(\omega). The identity holds on the elementary generators, whose reductions generate SL⁡(2,7)\SL(2,7), and both sides are homomorphisms; conjugation (z,t)↦(zˉ,t)(z,t)\mapsto(\bar z,t) reverses the orientation of H3\mathbb H^3 and carries Γ(p)\Gamma(\mathfrak p) to Γ(pˉ)\Gamma(\bar{\mathfrak p}). (b) DD normalizes SL⁡(2,Z[ω])\SL(2,\Z[\omega]) and fixes p\mathfrak p, so it normalizes the kernel; its reduction has determinant −1-1, a non-square modulo 7, and it acts on the sphere at infinity as z↦−zz\mapsto-z, preserving orientation. (c) A signed permutation normalizing the Weil group is determined by the image of v∞v_\infty and by where it sends two generators; propagating from every candidate image finds exactly 672, the image of GL⁡(2,7)\GL(2,7), by machine. (d) follows from (a)–(c).

Status

The chapter’s results are proved by hand or computed exactly: the independence of the two flips, the scattering constant, the gauge content of the Cayley data, the cusp lines, the commit algebra, the bilinears and pair structures, the obstruction of the ramified prime, the vacuum and the empty level. Some are computations to a stated range: the level ratios at the separations the prime above three marks, by a sieve to cutoffs of a few million, and the reversal search in a box. A few were found afterwards and are marked so: the reason the table’s mirror differs from the classical one, the counter standing in the clock’s place, the mechanism of the uneven levels, the exact equality of counts at finite three-adic depth, and the bounded levels. Whether four vectors of determinant −11-11 are reversed by larger elements, and whether the record field is local at the place above three, are open.

The last part of the chapter rests on additions it names: a filled band on the twenty-eight observers, Fermi statistics, a filling and spin as a spectator; there the thresholds are built conditionally, the contact condensing the pair term is refuted, and so is a coupling that runs on the refinement at the prime above three. The colour field’s propagator is exact and its exchange is built in the static limit; its time direction is a dial that the record does not fix, and the one-step range of the static field on the whole tower is computed, its mechanism not yet derived. The physical readings of all of this, of the flips as hands, of VV as a spinor of matter and of the fields and vacua, belong to the volume, in Chapters XV, XVI, XVIII and XXIII.

Deux orientationsTwo orientations

0123456∞z ↦ −z0 and ∞ fixed, (1 6)(2 5)(3 4){∞, 0, 1, 3}of the class {∞} ∪ (x + {0,1,3}){∞, 0, 6, 4}of the class {∞} ∪ (x + {0,4,6})the outer class: a rotation of M that exchanges the two classescomplex conjugation: the mirror image of M, every label kept
Plate 13.1The outer class acts on the projective line as z↦−zz\mapsto-z: it carries the table’s quadruple {∞,0,1,3}\{\infty,0,1,3\} to {∞,0,6,4}\{\infty,0,6,4\}, a tetrahedron of the other class, while complex conjugation, the other flip, keeps every label.

On a fixed basis 1,e0,…,e61,e_0,\dots,e_6 there are 480 octonion multiplication tables with eaeb=±ece_ae_b=\pm e_c: thirty Fano planes, each with sixteen orientations. A table’s 3-form orients R7\R^7, and the 480 fall into two classes of 240; a table and its opposite lie in different classes. The table AA and its Weil mirror A′A' lie in the same class, while the mirror obtained by relabelling the units by y↦−yy\mapsto-y, and the opposite of AA, lie in the other: the Weil mirror is not the classical mirror of the 480 tables. The group 2⋅A72{\cdot}A_7 of the Hermitian E8E_8 acts on Q(−7)4\Q(\sqrt{-7})^4 through a character with values (1±−7)/2(1\pm\sqrt{-7})/2 on elements of order 7; its invariant lattice is E8E_8 over the integers of Q(−7)\Q(\sqrt{-7}), and the group of order 168 is the stabilizer of one cross.

In MM the table’s ordered quadruple (∞,x,x+1,x+3)(\infty,x,x+1,x+3) lifts to the ideal tetrahedron (∞,a,a+1,a+1+ωˉ)(\infty,a,a+1,a+1+\bar\omega), negatively oriented, and the Weil mirror’s (∞,x,x+1,x+5)(\infty,x,x+1,x+5) to (∞,a,a+1,a+1+ω)(\infty,a,a+1,a+1+\omega), positively oriented, since 1+ωˉ≡31+\bar\omega\equiv3 and 1+ω≡51+\omega\equiv5 modulo p\mathfrak p while Im⁡(1+ωˉ)<0<Im⁡(1+ω)\operatorname{Im}(1+\bar\omega)<0<\operatorname{Im}(1+\omega). In Mˉ\bar M both signs reverse. The sign is changed by cc and, in these conventions, by the outer class; the second change will turn out to be a change of convention.

Proposition(The two orientations as Galois conjugations)

In Q(ζ21)⊃Q(−3,−7)\Q(\zeta_{21})\supset\Q(\sqrt{-3},\sqrt{-7}) the Frobenius at 2, σ2\sigma_2, negates −3\sqrt{-3} and fixes −7\sqrt{-7}, so it carries p=(3+ω)\mathfrak p=(3+\omega) to pˉ\bar{\mathfrak p} and fixes λ\lambda and λˉ\bar\lambda; σ13\sigma_{13} fixes −3\sqrt{-3} and negates −7\sqrt{-7}; complex conjugation negates both. The link complement is defined over Q(−3)\Q(\sqrt{-3}) and the Weil data over Q(ζ7)⊃Q(−7)\Q(\zeta_7)\supset\Q(\sqrt{-7}). On these fields the change cc acts as σ2\sigma_2 and the improper Weil action as σ13\sigma_{13}: the two orientations are the two independent generators of Gal(Q(−3,−7)/Q)≅C2×C2\mathrm{Gal}(\Q(\sqrt{-3},\sqrt{-7})/\Q)\cong C_2\times C_2.

Proof

σa(−3)=(a3)−3\sigma_a(\sqrt{-3})=\bigl(\tfrac a3\bigr)\sqrt{-3} and, −7\sqrt{-7} being a Gauss sum, σa(−7)=(a7)−7\sigma_a(\sqrt{-7})=\bigl(\tfrac a7\bigr)\sqrt{-7}. Here 2≡−12\equiv-1 modulo 3 and 2 is a square modulo 7, while 13≡113\equiv1 modulo 3 and 13≡613\equiv6 is not a square modulo 7. cc acts on Q(ω)\Q(\omega) as conjugation and trivially on the data defined on the projective line; DD has rational entries, and its Weil action is antilinear in −7\sqrt{-7}.

La diffusion du système spinorielThe scattering of the spinor system

∞∞00112233445566ba+1: b − a a square, {1, 2, 4}, and ∞’s row−1: b − a a non-square, {3, 5, 6}, and ∞’s columnantisymmetric; P² = −7I, so its eigenvalues are ±i√7
Plate 13.2The Paley matrix Θ\Theta on the eight cusps ∞,0,1,…,6\infty,0,1,\dots,6: zero diagonal, +1+1 along the row of ∞\infty, −1-1 down its column, and the Legendre symbol of b−ab-a elsewhere. The boundary scattering of VV is y Θy\,\Theta.

The orientation of MM appears in a constant. Let VV be the local system on MM given by the defining representation of SL⁡(2,Z[ω])\SL(2,\Z[\omega]), restricted to Γ(p)\Gamma(\mathfrak p). At a cusp with translations tbt_b a cocycle of VV restricts to z(b)=(sb2/2+Ab+Bbˉ, sb)z(b)=(sb^2/2+Ab+B\bar b,\ sb), with ss holomorphic in the parameter of the cusp and BB antiholomorphic, and the boundary image of H1(M;V)H^1(M;V) is the graph s=LVBs=L_VB of an antisymmetric matrix LVL_V with zero diagonal, indexed by the eight cusps. The lattices stable under SL⁡(2,Z[ω])\SL(2,\Z[\omega]) are all homothetic to Z[ω]2\Z[\omega]^2, and it fixes the scattering up to one sign: in the basis (ue1,ve2)(ue_1,ve_2) the ratio s/Bs/B is multiplied by u2vˉ/(v2uˉ)u^2\bar v/(v^2\bar u), which for units is (u/v)3(u/v)^3, equal to 1 exactly when diag(u,v)\mathrm{diag}(u,v) reduces to an inner automorphism of SL⁡(2,7)\SL(2,7).

The outer class D=diag(−1,1)D=\mathrm{diag}(-1,1) acts by a cochain map, by a signed permutation RDR_D over c↦−cc\mapsto-c on the coordinates BB and by −RD-R_D on ss, with RDΘ=−ΘRDR_D\Theta=-\Theta R_D. It exchanges the quartets and leaves LVL_V and yy unchanged: the sign that tells the quartets apart belongs to Θ\Theta. The mirror cc carries LVL_V to its complex conjugate, y↦yˉy\mapsto\bar y, and since vp(y)−vpˉ(y)=−2v_{\mathfrak p}(y)-v_{\bar{\mathfrak p}}(y)=-2 while a common rescaling of the cusps changes this difference by multiples of 3, in no rational normalization is yy even or odd under cc. The scattering matrices and cup products of C\C, VV, Vˉ\bar V and Sym2V\mathrm{Sym}^2V are carried to themselves by DD, and every structure defined on P1(F7)\Proj^1(\F_7) through the reduction is invariant under cc: each orientation is blind to the other’s flip.

Galois type is not parity. The eigenvalue v4=−(24−7+221)/49v_{\mathbf4}=-(24\sqrt{-7}+2\sqrt{21})/49 has a component along 21\sqrt{21}, which σ2\sigma_2 and σ13\sigma_{13} each negate, but it is not odd under both flips, since the outer class negates Θ\Theta and exchanges the quartets. Around a cusp the parabolic (1101)\left(\begin{smallmatrix}1&1\\0&1\end{smallmatrix}\right), of order 7, acts on 4\mathbf4 with eigenvalues ζk\zeta^k, k∈{0,1,2,4}k\in\{0,1,2,4\}, and on 4ˉ\bar{\mathbf4} with k∈{0,3,5,6}k\in\{0,3,5,6\}: one fixed line together with the quadratic residues, or with the non-residues.

Theorem(The scattering constant) computed

Let Θ\Theta be the Paley matrix on the cusps ∞,0,…,6\infty,0,\dots,6, with border Θ∞a=1\Theta_{\infty a}=1, Θa∞=−1\Theta_{a\infty}=-1 and inner entries the Legendre symbol (b−a7)\bigl(\frac{b-a}7\bigr). In the integral framing, with the deck group labelled by reduction modulo p\mathfrak p in that basis:

(a) LV=y ΘL_V=y\,\Theta with no further signs, where

y=−2ω−3(3+ω)2=22−4ω49,∣y∣2=1249;y=\frac{-2\omega\sqrt{-3}}{(3+\omega)^2}=\frac{22-4\omega}{49},\qquad |y|^2=\frac{12}{49};

(b) the eigenspace of Θ\Theta for −−7-\sqrt{-7} is the quartet 4\mathbf4, on which (1101)\left(\begin{smallmatrix}1&1\\0&1\end{smallmatrix}\right) has trace (1+−7)/2(1+\sqrt{-7})/2, so LVL_V acts on 4\mathbf4 by v4=−−7 yv_{\mathbf4}=-\sqrt{-7}\,y and on 4ˉ\bar{\mathbf4} by −v4-v_{\mathbf4};

(c) in the cusp-adapted basis (pe1,e2)(pe_1,e_2), p=3+ωp=3+\omega, the constant is −2ω−3/pˉ-2\omega\sqrt{-3}/\bar p, and the entries Lc∞L_{c\infty}, c≠∞c\neq\infty, are 2−32\sqrt{-3} times the corresponding entries ω/pˉ\omega/\bar p for the trivial local system.

Proof

The cochain complex of the triangulation of MM by its 28 tetrahedra is exact over Q(ω)\Q(\omega). All 64 cocycles were restricted to the eight cusps exactly, and their image is the graph of an antisymmetric, zero-diagonal LVL_V with entries in Q(ω)\Q(\omega); the constant y=L0iL0jLij−1Θij(Θ0iΘ0j)−1y=L_{0i}L_{0j}L_{ij}^{-1}\Theta_{ij}(\Theta_{0i}\Theta_{0j})^{-1} is the same for all pairs. The deck group acts on the coordinates by signed permutations; the trace of (1101)\left(\begin{smallmatrix}1&1\\0&1\end{smallmatrix}\right) is 1 on 4⊕4ˉ\mathbf4\oplus\bar{\mathbf4}, and tr⁡(uLV)=7y\operatorname{tr}(uL_V)=7y. Part (c) uses the framing law and the constant of the trivial system, recomputed exactly; it was found after the first run.

La forme de Cayley sur les cellulesThe Cayley form on the cells

∞0123456∞ 0 1 3−+∞ 1 2 4−−∞ 2 3 5−−∞ 3 4 6−−∞ 0 4 5−+∞ 1 5 6−−∞ 0 2 6−+2 4 5 6−−0 3 5 6−+0 1 4 6−+0 1 2 5−+1 2 3 6−−0 2 3 4−+1 3 4 5−−εΦs0 = −1the sign gaugea sign at each cusp reaches 16 patternsnot +1 on every cell, the mirror’s patternS = Σ εΦ takes the values −14, 0, 228 triples T1, T2, T1 △ T2 cover each cusp evenly:their products, −1 in M and +1 in M̄, survive
Plate 13.3The fourteen Cayley cells of the table against the eight cusps, each with εΦA=−1\varepsilon\Phi_A=-1 in MM: a change of sign at one cusp reverses the seven cells through it, the changes reach sixteen patterns, and +1+1 on every cell, the mirror’s pattern, is not among them.

Attach to the cusp ∞\infty the unit 1 and to the cusp xx the unit exe_x. The Cayley 4-form Φ(w,x,y,z)=⟨w, x×y×z⟩\Phi(w,x,y,z)=\langle w,\,x\times y\times z\rangle, with x×y×z=12(x(yˉz)−z(yˉx))x\times y\times z=\tfrac12(x(\bar yz)-z(\bar yx)), is invariant under Spin(7)\mathrm{Spin}(7), and on a basis of units it is ±1\pm1 exactly on the fourteen quadruples that span Cayley planes, the blocks of a Steiner system S(3,4,8)S(3,4,8), and 0 on the other fifty-six. For the table AA these are the 14 tetrahedra of MM of the class {∞}∪(x+{0,1,3})\{\infty\}\cup(x+\{0,1,3\}) together with the complements of the lines x+{0,1,3}x+\{0,1,3\}, and for A′A' the 14 of the other class. Equivalently, the product of the four units of a quadruple of cusps, in any order and association, is ±1\pm1 exactly on the Cayley quadruples, and ±\pm an imaginary unit on the other 56.

Paired with the oriented cells of MM, Φ\Phi gives a sum S=∑Tε(T)Φ(T)S=\sum_T\varepsilon(T)\Phi(T) of the shape of Dijkgraaf and Witten’s actions, the pairing of a 3-cochain with the fundamental cycle of the end compactification of MM; no classical instance of it is known, and no novelty is claimed. In the cyclic gauges, where AA reads exex+1=+ex+3e_xe_{x+1}=+e_{x+3} and A′A' reads exex+1=+ex+5e_xe_{x+1}=+e_{x+5}, every Cayley cell of AA has εΦA=−1\varepsilon\Phi_A=-1 in MM, so S(A)=−14S(A)=-14, while S(A′)=0S(A')=0; in Mˉ\bar M every orientation reverses. But SS is not an invariant, and the outer class carries the table to its mirror in the opposite gauge, where the sign agrees with the table’s: no gauge-invariant function of the oriented Cayley data is odd under each flip separately.

Theorem(The sign gauge, and what survives it) computed

Changing the signs scs_c of the units at the cusps multiplies ε(T)Φ(T)\varepsilon(T)\Phi(T) by ∏c∈Tsc\prod_{c\in T}s_c. (a) On the 14 Cayley cells of AA these changes realize exactly 16 sign patterns, and the pattern −1-1 on every cell, which is what the mirror cc does, is not one of them. (b) Pulled back by DD, the pattern of A′A' in its cyclic gauge equals the pattern of AA times the pattern of the signs +1+1 at ∞\infty and −1-1 elsewhere. (c) On the gauge orbit of AA‘s pattern, SS takes the values −14-14, 0 and 2. (d) For each of the 28 triples {T1,T2,T1△T2}\{T_1,T_2,T_1\triangle T_2\} of Cayley cells, which cover every cusp an even number of times, ∏εΦA=−1\prod\varepsilon\Phi_A=-1 in MM, and likewise for A′A'; in Mˉ\bar M every such product is +1+1.

So the gauge-invariant content of the oriented Cayley data is odd under cc and invariant under the outer class.

Proof

Over F2\F_2 the map from cusp signs to cell signs has as kernel the extended Hamming code spanned by the cells, self-dual of dimension 4, so its image has dimension 4. The all-ones vector is not in the image, since no set ss of cusps meets all fourteen blocks oddly: if ∣s∣|s| is odd, a block and its complement meet ss in sizes summing to ∣s∣|s|; if ∣s∣∈{2,4}|s|\in\{2,4\}, of the three blocks through two points a,b∈sa,b\in s at most one contains each further point of ss, so one meets ss in exactly {a,b}\{a,b\}; if ∣s∣=6|s|=6, a block through the two points outside ss meets ss in two points; and ∣s∣∈{0,8}|s|\in\{0,8\} meets every block evenly. Part (b) is the oriented sum read as a change of gauge. All parts were also enumerated by machine.

Le système spinoriel aux pointesThe spinor system at the cusps

∞(1, 0)0(0, 1)1(1, 1)1 + ω(1 + ω, 1)pairs sharing a cusp: parabolicu∞(1): z ↦ z + 1holomorphic: keeps the class of V∞(1, 0)0(0, 1)1(1, 1)1 + ω(1 + ω, 1)opposite pairs: order 4 in 2T∞ ↔ 1 + ω, 0 ↔ 1holomorphic: keeps the class of V
Plate 13.4The tetrahedron T0T_0 with cusps ∞\infty, 0, 1, 1+ω1+\omega: four cusp lines and six pairs of cusps. A move between two pairs that share a cusp is parabolic, a move to the opposite pair has order 4 in 2T2T, and every move keeps the class of VV.

The cusps of one tetrahedron single out a line of VV at each cusp. Let T0T_0 have cusps ∞,0,1,1+ω\infty,0,1,1+\omega, with primitive vectors ξ∞=(1,0)\xi_\infty=(1,0), ξ0=(0,1)\xi_0=(0,1), ξ1=(1,1)\xi_1=(1,1) and ξ1+ω=(1+ω,1)\xi_{1+\omega}=(1+\omega,1), the cusp of (x,y)(x,y) being x/yx/y. They are pairwise unimodular, the null vectors Nc=ξcξc†N_c=\xi_c\xi_c^\dagger have ⟨Na,Nb⟩=12\langle N_a,N_b\rangle=\tfrac12 and form a Z\Z-basis of Herm2(Z[ω])\mathrm{Herm}_2(\Z[\omega]), and at each cusp uc(t)=I+t ξcξcTJu_c(t)=I+t\,\xi_c\xi_c^{\mathsf T}J, J=(01−10)J=\left(\begin{smallmatrix}0&1\\-1&0\end{smallmatrix}\right), is parabolic for t≠0t\ne0 and fixes exactly the line of ξc\xi_c. Of the 24 permutations of the cusps exactly the 12 even ones are induced by Möbius maps, all in PSL⁡(2,Z[ω])\PSL(2,\Z[\omega]), with lifts forming the binary tetrahedral group 2T2T; modulo p\mathfrak p the stabilizer of the quadruple and its complement has 48 elements with the element orders of 2O2O, whose elements of order 8, of traces 3 and 4, act on MM without fixed points.

The six pairs of cusps give six unit timelike vectors Tab=Na+NbT_{ab}=N_a+N_b, with ⟨Tab,Tac⟩=32\langle T_{ab},T_{ac}\rangle=\tfrac32 and ⟨Tab,Tcd⟩=2\langle T_{ab},T_{cd}\rangle=2. Each of the 24 moves {c,a}→{c,b}\{c,a\}\to\{c,b\} is realized by a parabolic element fixing ξc\xi_c, and each of the 6 moves to the opposite pair by an element of order 4 of 2T2T; all are holomorphic, so each keeps the class of VV, while the reflections of T0T_0 are anti-Möbius. Restricted to 2O2O the quartets have no doublet summand: SL⁡(2,7)\SL(2,7) has no two-dimensional irreducible representation, 4\mathbf4 and 4ˉ\bar{\mathbf4} both restrict to 4s=E+⊗2=E−⊗2\mathbf4_s=E_+\otimes\mathbf2=E_-\otimes\mathbf2, and the faithful doublets E±=ρ±E_\pm=\rho_\pm occur only in 6s\mathbf6_s, 6s′\mathbf6_s' and 8s\mathbf8_s.

Theorem(The spinor system at the cusps) computed

Let ρ\rho be either of the two faithful two-dimensional representations ρ±\rho_\pm of 2O2O on C2\C^2. (a) The line of ξc\xi_c is the unique line of VV fixed by the parabolic stabilizer of cc. (b) Reduction modulo p\mathfrak p maps the stabilizer of T0T_0 in SL⁡(2,Z[ω])\SL(2,\Z[\omega]) isomorphically onto the even part 2T2T of 2O2O, and there ρ\rho is equivalent to VV; the stabilizer permutes the four cusp lines as it permutes the cusps. (c) The odd part of 2O2O acts on MM without fixed points, so its values under ρ\rho are values of VV at no point; on VV it acts only by transport between fibers, and VV does not choose between ρ+\rho_+ and ρ−\rho_-. (d) The cusp lines transform as VV and not as Vˉ\bar V, and in the mirror link complement Mˉ\bar M they transform as Vˉ\bar V.

Proof

(a) ξcξcTJ\xi_c\xi_c^{\mathsf T}J has rank one and trace ξcTJξc=0\xi_c^{\mathsf T}J\xi_c=0, so uc(t)u_c(t) has determinant 1 and trace 2, and its fixed line is the kernel, the line of ξc\xi_c. (b) The groups, traces and element orders were enumerated exactly, and the map from lines to cusps is equivariant for ψ↦gψ\psi\mapsto g\psi. (c) An elliptic element of PSL⁡(2,Z[ω])\PSL(2,\Z[\omega]) of order nn has a lift of trace ±2cos⁡(πk/n)\pm2\cos(\pi k/n), real and in Z[ω]\Z[\omega], hence in Z\Z, so n≤3n\le3; an element of order 4 with a fixed point on MM would lift to one of order divisible by 4. The two representations agree on 2T2T and differ by 2↦−2\sqrt2\mapsto-\sqrt2 on the elements of order 8. (d) For g=(1ω01)g=\left(\begin{smallmatrix}1&\omega\\0&1\end{smallmatrix}\right) the line gξ0g\xi_0 is the cusp ω=g⋅0\omega=g\cdot0, while gˉξ0\bar g\xi_0 is the cusp ωˉ\bar\omega; and V≇VˉV\not\cong\bar V, since (1ω11+ω)\left(\begin{smallmatrix}1&\omega\\1&1+\omega\end{smallmatrix}\right) has trace 2+ω≠2+ω‾2+\omega\neq\overline{2+\omega}.

L’algèbre de consignationThe commit algebra

1234567a1 = (2, 3)a2 = (4, 5)a3 = (6, 7)the counterZ = +1Z = −1I ⊗ Z, in the place of 1six Leₐ ⊗ X at the other points(−1)N = ±DCl7 on seven generators, centre spanned by I and Dage = D ⊗ Z
Plate 13.5The commit algebra at the point 1: the six other points are the six generators Lea⊗XL_{e_a}\otimes X, and the three lines through 1 pair them into the three modes whose occupation NN gives (−1)N=±D(-1)^N=\pm D.

In C⊗O\C\otimes\Oct fix an imaginary unit epe_p and put D=−iLepD=-iL_{e_p}. For each of the six units eae_a orthogonal to epe_p, Lea2=−IL_{e_a}^2=-I and LeaL_{e_a} anticommutes with LepL_{e_p}, hence with DD, and the product of the six is ±Lep\pm L_{e_p}, whatever the signs of the units: by alternativity LxLy+LyLx=−2⟨x,y⟩IL_xL_y+L_yL_x=-2\langle x,y\rangle I for imaginary x,yx,y. So along a word of moves of T0T_0, each paired with the left multiplication by a unit orthogonal to epe_p, the spinor factor keeps the class of VV while DD changes sign at every move, and after nn moves the preserved operator is (−1)nD(-1)^nD. The commit algebra asks which operators commute with everything that writes the moves.

For the three lines {p,q,r}\{p,q,r\} through pp, ordered so that eqer=epe_qe_r=e_p, put aj=(Leq+iLer)/2a_j=(L_{e_q}+iL_{e_r})/2 and N=∑jaj†ajN=\sum_ja_j^\dagger a_j, and on a second factor C2\C^2 write XX and ZZ for the Pauli matrices. On O\Oct alone the seven LexL_{e_x} make a single module of Cl7\mathrm{Cl}_7, whose volume element is a scalar, and there is one sector. The counter’s I⊗ZI\otimes Z stands in the place of LepL_{e_p}, the volume element becomes ±Lep⊗Z\pm L_{e_p}\otimes Z, which is not a scalar, and its two eigenspaces are the sectors. With the counter replaced by ℓ2(N)\ell^2(\mathbb N) the commutant is still spanned by II and D⊗(−1)nD\otimes(-1)^n: the parity carries all of the counter’s contribution.

What the centre leaves out: a function of NN, tensored with II or with ZZ, commutes with the algebra only if it lies in span{I,(−1)N⊗Z}\mathrm{span}\{I,(-1)^N\otimes Z\}; the products LeaLebL_{e_a}L_{e_b} generate M4(C)M_4(\C) on each eigenspace of DD; and J0=Lu/7J_0=L_u/\sqrt7, u=e0+⋯+e6u=e_0+\dots+e_6, is not in the algebra, since LeaLuLea−1=2Lea−LuL_{e_a}L_uL_{e_a}^{-1}=2L_{e_a}-L_u is never ±Lu\pm L_u. The +i+i-eigenspaces of J0J_0 and of LepL_{e_p} are isoclinic, at the angle with cos⁡2=12(1±1/7)\cos^2=\tfrac12(1\pm1/\sqrt7).

Theorem(The commit algebra is Cl7\mathrm{Cl}_7 with two sectors)

The commit algebra at epe_p is Ap=alg{Lea⊗X (a≠p), I⊗Z}⊂End((C⊗O)⊗C2)\mathcal A_p=\mathrm{alg}\{L_{e_a}\otimes X\ (a\neq p),\ I\otimes Z\}\subset\mathrm{End}((\C\otimes\Oct)\otimes\C^2), with Dage=D⊗ZD_{\mathrm{age}}=D\otimes Z. For each of the seven clocks: (1) the six LeaL_{e_a}, a≠pa\neq p, generate M8(C)M_8(\C), with commutant C\C; (2) the seven generators of Ap\mathcal A_p anticommute in pairs, six square to −I-I and one to +I+I, and their product is ±Lep⊗Z\pm L_{e_p}\otimes Z, so Ap\mathcal A_p is the complex Clifford algebra Cl7\mathrm{Cl}_7, of dimension 128; (3) the commutant of Ap\mathcal A_p is its centre, span{I,Dage}\mathrm{span}\{I,D_{\mathrm{age}}\}, and Ap≅M8(C)⊕M8(C)\mathcal A_p\cong M_8(\C)\oplus M_8(\C), with central projections P±=12(I±Dage)P_\pm=\tfrac12(I\pm D_{\mathrm{age}}) of rank 8; (4) (−1)N=±D(-1)^N=\pm D, so Dage=±(−1)N⊗ZD_{\mathrm{age}}=\pm(-1)^N\otimes Z; (5) complex conjugation of C⊗O\C\otimes\Oct fixes every generator and exchanges P+P_+ with P−P_-.

Proof

(1) The six LeaL_{e_a} anticommute and square to −I-I, so they generate a quotient of Cl6≅M8(C)\mathrm{Cl}_6\cong M_8(\C), which is simple; on C8\C^8 the image is all of M8(C)M_8(\C). (2) (Lea⊗X)(Leb⊗X)=LeaLeb⊗I(L_{e_a}\otimes X)(L_{e_b}\otimes X)=L_{e_a}L_{e_b}\otimes I anticommutes for a≠ba\neq b, each Lea⊗XL_{e_a}\otimes X anticommutes with I⊗ZI\otimes Z, and the six LeaL_{e_a} multiply to ±Lep\pm L_{e_p}. (3) On an odd number of generators the centre of the complex Clifford algebra is spanned by II and the volume element, here ±iDage\pm iD_{\mathrm{age}}; since tr⁡Dage=0\operatorname{tr}D_{\mathrm{age}}=0, both simple summands act. (4) 1−2aj†aj=iLeqLer1-2a_j^\dagger a_j=iL_{e_q}L_{e_r}, so (−1)N=i3∏jLeqjLerj=∓iLep(-1)^N=i^3\prod_jL_{e_{q_j}}L_{e_{r_j}}=\mp iL_{e_p}. (5) The generators are real matrices, and Dage=−iLep⊗ZD_{\mathrm{age}}=-iL_{e_p}\otimes Z is imaginary. All five were also checked for all seven clocks.

Formes bilinéaires d’une seule mainBilinears of one hand

11e0e0e1e1e2e2e3e3e4e4e5e5e6e6τ0, the octonion normsymmetricpair for fermions 4√2, for bosons 011e0e0e1e1e2e2e3e3e4e4e5e5e6e6τD, the matrix of D6antisymmetricpair for fermions 0, for bosons 4√2neutral for all seven clocks: only the multiples of τ0
Plate 13.6The two pair tensors at one clock as 8×88\times8 matrices: τ0\tau_0, the octonion norm, symmetric, makes a pair only for fermions; τD\tau_D, the matrix of DpD_p, antisymmetric, only for bosons; each pair has norm 424\sqrt2.

VV and its mirror Vˉ\bar V are inequivalent, and that fixes the bilinear invariants: for G=SL⁡(2,Z[ω])G=\SL(2,\Z[\omega]), dim⁡(V⊗V)G=1\dim(V\otimes V)^G=1, spanned by ε\varepsilon, dim⁡(V⊗Vˉ)G=0\dim(V\otimes\bar V)^G=0 and dim⁡(Vˉ⊗Vˉ)G=1\dim(\bar V\otimes\bar V)^G=1, so HomG(V,Vˉ)=0\mathrm{Hom}_G(V,\bar V)=0. At a clock epe_p, with the quartet 4\mathbf4 the −1-1-eigenspace of DD, the fifteen products LeaLebL_{e_a}L_{e_b} span su(4)\mathfrak{su}(4), and an element of (V⊗V)⊗(4⊗4ˉ)(V\otimes V)\otimes(\mathbf4\otimes\bar{\mathbf4}) invariant under GG and su(4)\mathfrak{su}(4) is unique up to scale, ε⊗∑iei⊗eˉi\varepsilon\otimes\sum_ie_i\otimes\bar e_i.

A tensor annihilated by Dp⊗I+I⊗DpD_p\otimes I+I\otimes D_p for all seven pp is a multiple of τ0=1⊗1+∑kek⊗ek\tau_0=1\otimes1+\sum_ke_k\otimes e_k, the octonion norm’s, which is symmetric; at one clock the su(4)\mathfrak{su}(4)-invariant ones are spanned by τ0\tau_0 and the antisymmetric tensor τD\tau_D of DpD_p, with ∑iei⊗eˉi=12(τ0−τD)\sum_ie_i\otimes\bar e_i=\tfrac12(\tau_0-\tau_D). So the invariant pair has a half that makes a pair only for fermions and a half only for bosons. The symmetric pair tensors invariant under the SU(3)\mathrm{SU}(3) fixing epe_p and annihilated by Dp⊗I+I⊗DpD_p\otimes I+I\otimes D_p are spanned by τ0\tau_0 and τ0∘(Pℓ−13Pq)\tau_0\circ(P_\ell-\tfrac13P_q), with PℓP_\ell and PqP_q the projections onto ⟨1,ep⟩\langle1,e_p\rangle and onto the six other units, and Pℓ−13Pq=−σ(B−L)DpP_\ell-\tfrac13P_q=-\sigma(B-L)D_p for B−L=23N−1B-L=\tfrac23N-1 and Dp=σ(−1)ND_p=\sigma(-1)^N: a second structure, with the factor −13-\tfrac13 between its parts.

Holes in a fermionic Fock space over a finite-dimensional space transform by (g−1)T=εgε−1(g^{-1})^{\mathsf T}=\varepsilon g\varepsilon^{-1}, so the holes of VV carry VV again, whatever positive inner product is used, and none is invariant, since (1101)\left(\begin{smallmatrix}1&1\\0&1\end{smallmatrix}\right) is a nontrivial Jordan block. The vectors the moves pass between do carry invariant forms: for positive TT of determinant one, ⟨ξ,η⟩T=ξ†T−1η\langle\xi,\eta\rangle_T=\xi^\dagger T^{-1}\eta makes every g∈SL⁡(2,C)g\in\SL(2,\C) an isometry onto ⟨ ,⟩gTg†\langle\,,\rangle_{gTg^\dagger}, and on ℓ2(G⋅T0)⊗V\ell^2(G\cdot T_0)\otimes V the action is unitary, the discrete form of the representation of SL⁡(2,C)\SL(2,\C) induced from SU(2)\mathrm{SU}(2). On the cusp lines a hole carries the conjugate character, χˉ\bar\chi where the cusp’s vector carries χ\chi.

Theorem(Pairs fix the exchange sign relative to the interior) computed

Let τ∈(C⊗O)⊗2\tau\in(\C\otimes\Oct)^{\otimes2} have symmetry sτ=±1s_\tau=\pm1 under the swap of its factors, and put κ=ε⊗τ\kappa=\varepsilon\otimes\tau, a tensor on the sixteen modes of V⊗(C⊗O)V\otimes(\C\otimes\Oct). In the Fock space of these modes with exchange sign η\eta, −1-1 for fermions and +1+1 for bosons, the pair ∑A,BκAB cA†cB† Ω\sum_{A,B}\kappa_{AB}\,c_A^\dagger c_B^\dagger\,\Omega is nonzero if and only if η=−sτ\eta=-s_\tau. As a two-particle tensor, κ\kappa has eigenvalue −sτ-s_\tau under the swap of the particles and −1-1 under a rotation by 2π2\pi of one of them. The two agree exactly when τ\tau is symmetric.

Proof

Since cA†cB†=η cB†cA†c_A^\dagger c_B^\dagger=\eta\,c_B^\dagger c_A^\dagger, the vector depends only on the part of κ\kappa of symmetry η\eta. κ\kappa has symmetry −sτ-s_\tau because ε\varepsilon is antisymmetric, and the rotation by 2π2\pi is −I-I on VV. A pairing realized by operators as [ψ(f),ψ(g)]±=⟨f,g⟩[\psi(f),\psi(g)]_\pm=\langle f,g\rangle has the symmetry of its bracket, which is why the pairing of the spinor alone, antisymmetric and of Frobenius–Schur indicator −1-1 on the dicyclic group of order 12, selects no statistics. Checked with Jordan–Wigner fermions and truncated bosons on the sixteen modes: the pair through ε⊗τ0\varepsilon\otimes\tau_0 has norms 424\sqrt2 and 0, and the pair through ε⊗τD\varepsilon\otimes\tau_D norms 0 and 424\sqrt2.

Renverser les séparationsReversing separations

474 primitive vectors with det X < 0reversed in SL(2, ℤ[ω]): 276reversed only with a unit determinant: 46blocked by the prime above 3: 148not reached in the box: 425507580−152−298−336−464−532−648−78−816−98−1012−118−124−138−14det X
Plate 13.7The 474 primitive vectors of the search box with det⁡X<0\det X<0, by determinant: reversed in SL⁡(2,Z[ω])\SL(2,\Z[\omega]) (276), reversed only with a unit determinant (46 more), blocked by the prime above 3 (148, all with 3∣det⁡X3\mid\det X), and four of determinant −11-11 that the box does not reach.

GG acts on Λ=Herm2(Z[ω])\Lambda=\mathrm{Herm}_2(\Z[\omega]) by X↦gXg†X\mapsto gXg^\dagger, preserving det⁡X\det X and the cone of positive vectors. In the continuum every XX with det⁡X<0\det X<0 is carried to −X-X by some element of SL⁡(2,C)\SL(2,\C). Over Z[ω]\Z[\omega] the ramified prime π=1−ω\pi=1-\omega blocks this at every XX with 3∣det⁡X3\mid\det X that does not vanish modulo π\pi.

A search finds the obstruction and almost nothing else in its range. In the box ∣x1∣,∣x2∣≤2|x_1|,|x_2|\le2, z=u+vωz=u+v\omega with ∣u∣,∣v∣≤2|u|,|v|\le2, there are 474 primitive vectors with det⁡X<0\det X<0. The elements of SL⁡(2,Z[ω])\SL(2,\Z[\omega]) with entry coordinates in [−2,2][-2,2] reverse 276 of them, and those of GL⁡(2,Z[ω])\GL(2,\Z[\omega]) with unit determinant reverse 322. The unreversed are the 148 the theorem covers and 4 of determinant −11-11. At a prime that is not ramified there is no invariant of this kind, since at an inert prime every scalar of the residue field is a norm and at a split prime the two factors scale independently, so the four are presumably reversed by larger elements, which were not searched. No element reverses a vector with det⁡X>0\det X>0.

Theorem(The ramified prime blocks the reflection) proved

Let X∈ΛX\in\Lambda have det⁡X<0\det X<0 and 3∣det⁡X3\mid\det X, and suppose X≢0X\not\equiv0 modulo π=1−ω\pi=1-\omega. Then XX modulo π\pi is a symmetric form of rank one over F3\F_3, λ ℓℓT\lambda\,\ell\ell^{\mathsf T}, and the Legendre symbol (λ3)\bigl(\tfrac{\lambda}{3}\bigr) is invariant under X↦gXg†X\mapsto gXg^\dagger for every g∈GL⁡(2,Z[ω])g\in\GL(2,\Z[\omega]), and under X↦XTX\mapsto X^{\mathsf T}. Since (−13)=−1\bigl(\tfrac{-1}{3}\bigr)=-1, no such map sends XX to −X-X.

Proof

Since ω≡ωˉ≡1\omega\equiv\bar\omega\equiv1 modulo π\pi, reduction turns gXg†gXg^\dagger into gˉXgˉT\bar gX\bar g^{\mathsf T} over F3\F_3, and λ ℓℓT↦λ (gˉℓ)(gˉℓ)T\lambda\,\ell\ell^{\mathsf T}\mapsto\lambda\,(\bar g\ell)(\bar g\ell)^{\mathsf T}. The coefficient λ\lambda is defined up to squares, and the reduction of XTX^{\mathsf T} is the same form.

Le champ d’une seule main et son videThe field of one hand and its vacuum

records at the levels of diag(2, −3)−k+kN(k)/N(−k)k = 1/258,59010,0515.83k = 155,50655,2901.004k = 246,45246,6620.995k = 5/2none48,744empty at −kk = 7/234,50034,4101.003⟨T, R⟩ ≤ 20,000; the level −5/2 is empty for every cutoff
Plate 13.8The records at the levels ±j/2\pm j/2 of X=diag(2,−3)X=\mathrm{diag}(2,-3), counted up to ⟨T,R⟩≤20000\langle T,R\rangle\le20000 for the unit frame RR orthogonal to XX: the level −52-\tfrac52 is empty, −12-\tfrac12 is thinned about six to one, and the other levels agree to within a percent.

Write OO for the records, the positive T∈ΛT\in\Lambda with det⁡T=1\det T=1. They are one orbit: by Hermite reduction, with the covering radius 1/31/\sqrt3 of Z[ω]\Z[\omega], every record is gg†gg^\dagger for some g∈Γg\in\Gamma, which is class number one. The field of one hand lives on ℓ2(O)⊗V⊗F\ell^2(O)\otimes V\otimes F, F=C⊗OF=\C\otimes\Oct, each record’s spinor measured by its form ⟨ ,⟩T\langle\,,\rangle_T. Choosing gTg_T with gTgT†=Tg_Tg_T^\dagger=T, its coefficients x(T,s)=gTesx(T,s)=g_Te_s and y(T,s)=−gTεesy(T,s)=-g_T\varepsilon e_s satisfy ∑sxx†=T\sum_sxx^\dagger=T, ∑syy†=T\sum_syy^\dagger=T, ∑sxyT=ε\sum_sxy^{\mathsf T}=\varepsilon and T−1x=εyˉT^{-1}x=\varepsilon\bar y exactly over Z[ω]\Z[\omega], and ggTg†−1ggTg_{gTg^\dagger}^{-1}gg_T lies in the stabilizer of II, the dicyclic group of order 12. For masses with m/πm/\pi irrational, a quasi-free state invariant under the translations of Λ\Lambda has no anomalous part in the record modes, since two records never sum to zero; a ground state for one frame is their Fock vacuum; and that vacuum is invariant under Γ\Gamma, with no mesh entering.

Locality becomes a statement about levels. For spacelike X∈ΛX\in\Lambda let OX(k)O_X(k) be the records with ⟨T,X⟩=k\langle T,X\rangle=k, and n~X(k)\tilde n_X(k) the number of orbits of the stabilizer of XX on it, each weighted by the inverse order of its stabilizer. The anticommutators of the field at the separation XX are sums over k>0k>0 of n~X(k)−n~X(−k)\tilde n_X(k)-\tilde n_X(-k) against sin⁡mk\sin mk and kcos⁡mkk\cos mk, so the Fermi field is local at XX, for all masses, exactly when n~X\tilde n_X is even, and the Bose field is local at no spacelike XX. A reversal of XX in the lattice’s group makes n~X\tilde n_X even. The antiunitary JT=gTεKgT−1J_T=g_T\varepsilon Kg_T^{-1} is native and fixes the vacuum, but for a two-point function it yields only W(−X)=W(X)‾W(-X)=\overline{W(X)}, which holds anyway; the cancellation needs the linear reversal W(X)=W(−X)W(X)=W(-X).

At the separations the prime above three marks, the levels are uneven. Counted by a sieve to cutoffs of a few million, the ratios of the levels ±k\pm k at diag(1,−3)\mathrm{diag}(1,-3), diag(2,−3)\mathrm{diag}(2,-3) and diag(1,−6)\mathrm{diag}(1,-6) agree with 1 to 10−310^{-3}, except at the levels where the level quadric factors, where they are near 3/73/7, 6 and 1/61/6, or a level is empty. Since OnX(nk)=OX(k)O_{nX}(nk)=O_X(k), the field is not local on any lattice of positions commensurable with Λ\Lambda.

Theorem(An empty level) proved

Let X=diag(2,−3)X=\mathrm{diag}(2,-3), a separation the prime above three obstructs. No record TT has 2⟨T,X⟩=−52\langle T,X\rangle=-5, while T+=(11+2ω1+2ωˉ4)T_+=\left(\begin{smallmatrix}1&1+2\omega\\1+2\bar\omega&4\end{smallmatrix}\right) is a record with 2⟨T+,X⟩=52\langle T_+,X\rangle=5. So n~X(52)>0=n~X(−52)\tilde n_X(\tfrac52)>0=\tilde n_X(-\tfrac52), and the Fermi field is not local at XX: its anomalous anticommutator contains sin⁡(5m/2)\sin(5m/2) with a positive weight. Likewise at X=diag(1,−6)X=\mathrm{diag}(1,-6) the level +52+\tfrac52 is empty and −52-\tfrac52 contains II.

Proof

For X=diag(a,−b)X=\mathrm{diag}(a,-b), 2⟨T,X⟩=at2−bt12\langle T,X\rangle=at_2-bt_1. At 2t2−3t1=−52t_2-3t_1=-5, t1=2u+1t_1=2u+1 is odd and t2=3u−1≥1t_2=3u-1\ge1, so N(w)=t1t2−1=(3u+2)(2u−1)N(w)=t_1t_2-1=(3u+2)(2u-1) with u≥1u\ge1. A positive integer is a norm from Z[ω]\Z[\omega] exactly when every prime q≡2q\equiv2 modulo 3 divides it to an even power. Since 3u+2≡23u+2\equiv2 modulo 3, some such qq divides 3u+23u+2 to an odd power, and q∤2u−1q\nmid2u-1 because gcd⁡(3u+2,2u−1)\gcd(3u+2,2u-1) divides 7, which is 1 modulo 3. So N(w)N(w) is not a norm. For diag(1,−6)\mathrm{diag}(1,-6) at level +52+\tfrac52, N(w)=(6t−1)(t+1)N(w)=(6t-1)(t+1) with the gcd dividing 7. The examples are direct checks.

La place au-dessus de troisThe place above three

v0the tree at the prime above 313 even vertices: records of depth ≤ 1at distance two: 4 copies of K4 at v0each carries a copy of the recordscounts × 13 at depth 1, × 121 at depth 2
Plate 13.9The four-valent tree at π=1−ω\pi=1-\omega round v0v_0: the SS-records of depth at most one sit at the thirteen even vertices of the ball of radius two, each carrying a copy of the records, and joined at distance two those vertices are four copies of K4K_4 sharing v0v_0.

Inverting 3 supplies the missing reversal. An SS-record is a positive T∈Herm2(Z[ω,13])T\in\mathrm{Herm}_2(\Z[\omega,\tfrac13]) with det⁡T=1\det T=1, and its depth is the least rr with 3rT∈Λ3^rT\in\Lambda. The SS-arithmetic group acts on H3\mathbb H^3 times the four-valent tree T\mathcal T at π\pi, and the SS-records of depth at most rr sit at the even vertices of the ball B(v0,2r)B(v_0,2r), 13 for r=1r=1 and 121 for r=2r=2, each carrying an isometric copy of OO: the counts grow by those factors, 12.81, 13.12, 13.04 and 119.5, 121.6, 121.3 at three cutoffs. Over Z[ω,13]\Z[\omega,\tfrac13] the four separations diag(1,−3)\mathrm{diag}(1,-3), diag(2,−3)\mathrm{diag}(2,-3), diag(1,−6)\mathrm{diag}(1,-6) and diag(4,−3)\mathrm{diag}(4,-3) are reversed, by elements of determinant −1-1 such as (0π−1π0)\left(\begin{smallmatrix}0&\pi^{-1}\\\pi&0\end{smallmatrix}\right), and each reversal fixes the vertex toward the report that obstructed it.

At finite depth the levels even out. At depth at most one the factorable levels of diag(2,−3)\mathrm{diag}(2,-3) and diag(1,−6)\mathrm{diag}(1,-6) agree to 0.003, and at diag(1,−3)\mathrm{diag}(1,-3), level 1, exactly; in all 51 computed cases the counts agreed exactly if and only if the reflection through X⊥X^\perp was integral at that depth. An asymmetric level reappears only at depths up to the 3-denominator of kk: the obstruction moves to the finest depth. When X⊥X^\perp is anisotropic at 3 each level lies on finitely many vertices and its counts stop changing; when it is isotropic they keep growing. Whether the record field is local at this place is open.

The refinement at π\pi does not touch the antiflags. For every vertex vv at distance two from v0v_0 and every nn there is hh with hv0=vhv_0=v and h≡Ih\equiv I modulo pn\mathfrak p^n, so the records at vv naming any pair, at any scale up to nn, are the images of those at v0v_0 naming the same one. On the thirteen frames joined at tree distance two, four copies of K4K_4 sharing v0v_0, every cycle of length at most six projects to zero, no heptagon is a sum of shorter refined cycles, and the short cycles span a subspace of codimension exactly 15: the coarse coefficient is 13 times the fine one, in parallel. The building of SL⁡2\SL_2 at a finite place is a tree, with no 2-cells to subdivide a loop.

Proposition(Every SS-record has one 3-adic rest frame) computed

(1) Every SS-record is gg†gg^\dagger with gg in the SS-arithmetic group ΓS\Gamma_S, and gg is unique up to right multiplication by the dicyclic group of order 12. (2) Hence v(T)=g v0v(T)=g\,v_0 is a well-defined vertex of T\mathcal T, and T↦(T,v(T))T\mapsto(T,v(T)) embeds the SS-records as one discrete ΓS\Gamma_S-orbit in H3×T\mathbb H^3\times\mathcal T. (3) d(v(T),v0)=2 depth(T)d(v(T),v_0)=2\,\mathrm{depth}(T). In particular SS-records sit only at even vertices, and every even vertex v=hv0v=hv_0 carries the full orbit hOh†hOh^\dagger.

Proof

(1) Over the completion at π\pi the binary Hermitian form TT has determinant 1, a norm, so it has a unimodular lattice; glued to Z[ω]2\Z[\omega]^2 at the other places it gives a unimodular positive Hermitian Z[ω]\Z[\omega]-lattice, which is standard by class number one. For uniqueness, the unitary matrices over Z[ω,13]\Z[\omega,\tfrac13] are the twelve of the dicyclic group: N(a)+N(b)=3kN(a)+N(b)=3^k forces π∣a,b\pi\mid a,b for k≥1k\ge1, because the norms of π\pi-units are 1 modulo 3. (3) d(gv0,v0)=−2min⁡vπ(gij)d(gv_0,v_0)=-2\min v_\pi(g_{ij}) for det⁡g=1\det g=1, and the diagonal entries of gg†gg^\dagger, sums of norms of a row, do not cancel at the minimal valuation. Checked on 300 random elements and on all 1627 positive S∈ΛS\in\Lambda with det⁡S=9\det S=9 and s1+s2≤30s_1+s_2\le30.

Le contact, la bande et les liensThe contact, the band and the links

the Coxeter graph’s adjacencygap06−√2 − 17−16√2 − 1821313 negative modes, summing to −(13 + 6√2)multiplicities fixed by tr Ak, k ≤ 4, and (A − 3)(A − 2)(A2 + 2A − 1)(A + 1) = 0
Plate 13.10The spectrum of the Coxeter graph: 3 once, 2 eight times, 2−1\sqrt2-1 six times, −1-1 seven times and −2−1-\sqrt2-1 six times. The thirteen negative modes are the filled part, and the gap runs from −1-1 to 2−1\sqrt2-1.

The contact acts through the antisymmetric part ω(x,y)\omega(x,y) of the product of two letters. It has no matrix element on two quanta at one record, it has rank one with ⟨ω∣ω⟩=6\langle\omega|\omega\rangle=6, and V2=6gVV^2=6gV, so its repetitions sum in closed form within one channel, 1+(e−6ig−1)Πω\mathbf1+(e^{-6ig}-1)\Pi_\omega for one occurrence of strength gg, with no sum over momenta. In second quantization it commutes with the number of quanta and annihilates the vacuum and every one-quantum state, so the free vacuum stays an exact eigenstate at every order and the one-quantum energies are not shifted.

A filled state needs four additions: a band on the twenty-eight antiflags, Fermi statistics, the filling at zero and spin as a spectator. With them the band is the Coxeter graph’s adjacency tensored with I8I_8, the filled part is its thirteen negative modes for each internal state, and the gap runs from −1-1 to 2−1\sqrt2-1. Every weight of the pair susceptibility is at most 1/(2(2−1))1/(2(\sqrt2-1)) and its normalized form is positive definite, so for g>0g>0 nothing pairs at any strength. For g<0g<0 the band at fixed filling stays stable down to a g∗g^* between −0.43-0.43 and −0.54-0.54 in band units, depending on the form of the contact and its family of four-sets, with closed forms in the signed meeting case, where ∑QsQ=A2+A−3I\sum_Qs_Q=A^2+A-3I; and every internal pair state of the allowed exchange symmetry has the same threshold, under a symmetry U(16)U(16), so the contact selects no internal channel.

The comparisons on the meetings carry a quadratic form of their own. Expanded about the flat class the plaquette weight is β12∑a∥CAa∥2\tfrac{\beta}{12}\sum_a\|CA^a\|^2, where K=CTCK=C^{\mathsf T}C has spectrum {027,87,148}\{0^{27},8^7,14^8\}; its kernel is exactly the gradients, because the 24 heptagons span the cycle space, and its pseudo-inverse is K+=(186K−11K2)/6272K^+=(186K-11K^2)/6272. The static exchange it induces never acts on the pairs through ⟨1,ep⟩\langle1,e_p\rangle, where T1⋅T2=0T_1\cdot T_2=0, and under the four additions it pairs the colour singlet first. A time direction for this field is a dial the record does not fix. If one is required, Gauss’s law on the closed network leaves only pairs of zero total colour, a singlet’s energy is a fixed multiple of the effective resistance between its sites, 0, 914\tfrac9{14}, 1314\tfrac{13}{14}, 2928\tfrac{29}{28} or 1514\tfrac{15}{14} at Coxeter distance 0 to 4, and the pairs through ⟨1,ep⟩\langle1,e_p\rangle then pair first. On the whole tower of scales, roughly isometric to the tree at 7 and so non-amenable and transient, every such field is gapped at Gaussian order, and its static response reaches one scale step and no further.

Theorem(The interacting vacuum is the free vacuum) computed

In second quantization, Hc=g2∑ω(x′,y′) ω(x,y) ax′i†ay′j†ayjaxiH_{\mathrm c}=\tfrac g2\sum\omega(x',y')\,\omega(x,y)\,a^\dagger_{x'i}a^\dagger_{y'j}a_{yj}a_{xi} commutes with the number of quanta and annihilates the vacuum and every one-quantum state. With the free record field, whose vacuum is the empty Fock state of the record modes, the vacuum is therefore an exact eigenstate of the interacting dynamics at every order, and the one-quantum energies are not renormalized: the contact generates no condensate.

Proof

Each term has two annihilators to the right. Checked on the cube’s six letters with two spin states, 12 modes in a space of dimension 4096: HcH_{\mathrm c} commutes with the number operator, annihilates the vacuum, and vanishes on all 12 one-quantum states.

Each parent’s flip is seen by its own structures and not by the other’s. The link complement’s orientation is seen by the scattering constant and by the oriented cells; the table’s is an exact symmetry of everything defined on MM, once the signs of the units are a convention. The spinor system carries the first at the cusps, the commit algebra keeps one sign, and one hand admits exactly the bilinears, pairs and reversals found here, with the prime above three marking where reversal and locality fail.

The next chapter leaves the primes for the continua: real and complex geometries in which the finite objects reappear, as configurations or as classes modulo a congruence subgroup.