by igor.kh · Nov 13, 2026 · Mathlib 0df444a (Lean v4.33.1)
monoidal-categoriesquantum-groupsspin-networks
Let C be a K-linear preadditive monoidal category, K=C(q), with additive and K-linear tensor product. The chosen objects Vn for natural labels satisfy V0≅1 and have exact self-pairings. Assume SuQ2Axioms: unique scalar endomorphisms of each Vn, zero morphisms between different labels, the admissible SL(2) fusion decompositions, and the loop values Δn. Choose trivalent vertices W, and assume VertexNormalized W: theta normalization and cyclicity. Scalars act through ι:Q(q)↪K, fixing q. Composition f;g means first f, then g.
Also assume the separate initial data TetSeed W: for every natural a,b,c,d,j, T(a,b,c,d;j,2)=ι(Tetsum(a,b,c,d;j,2))id1. These are seed values, not an all-channel network-to-sum identity.
Prove QuantumSixJ.NetworkRecurrence. For every a,b,c,d∈N and every signed channels j,l∈Z, the graphical Tet of equation (6), extended by zero when a channel is negative, satisfies
The graphical network is T(a,b,c,d;j,l)=X(a,d,c,b;j);HR(a,d,c,b;l) on nonnegative channels. Scalar coefficients act on endomorphisms of the tensor unit. No eigenvector equation, matrix-entry evaluation, general tetrahedral symmetry, or all-channel network-to-sum equality is assumed beyond the stated category, normalization and seed data. Total field division includes the boundary and inadmissible cases. The pure scalar recurrence is not an assumption of this statement.
Independent mathematical read-back
Graphical tetrahedral recurrence
Statement and universe scope
Let u,v be arbitrary universe levels. The assertion quantifies over every category C whose objects have size u and whose morphisms have size v, equipped with a preadditive structure, a K-linear structure, a monoidal structure, and compatible monoidal-preadditive and monoidal-linear structures, where K=C(q). Thus morphism sets are additive groups and K-modules; composition and tensoring satisfy the indicated additivity and linearity requirements. No ordering between u and v is imposed by the assertion.
For every choice of labelled self-duality data S and trivalent vertices W described below, assume all the category axioms, the two vertex-normalization conditions, and the separate tetrahedral seed condition below. Then for everya,b,c,d∈N and j,l∈Z,
Products of a field coefficient and a morphism here mean the K-scalar action, and zero is the zero morphism. Set hr=h(a,b,c,d,r), Lr,s=L(a,b,c,d,r,s), and Tr,l=W.tet(a,b,c,d,r,l). Natural labels in scalar expressions are coerced to integers. The map ι:Q(q)→C(q) is the coefficient embedding induced by Q↪C, fixing the indeterminate q.
The three denominators are respectively the norms at j−2,j,j+2, not a common denominator hj. There are no admissibility or nonzero-denominator hypotheses on the displayed indices. This assertion does not assume the scalar recurrence or an all-channel equality between the graphical tetrahedron and the explicit sum.
Scalar definitions used in the assertion and seed
In F=Q(q), define for integers n
[n]=q−q−1qn−q−n,[n]!={0∏k=1n[k]n<0,n≥0.
The empty product is one. All divisions in F and K are total field division, so z/0=0, including 0/0=0. All divisions of integer indices by two mean Euclidean integer division, i.e. floor division for divisor two. Integer powers are used in expressions such as (−1)S.
Write A(a,b,c) for
a,b,c≥0,a≤b+c,b≤c+a,c≤a+b,a+b+c≡0(mod2).
Write Hr for A(a,d,r)∧A(b,c,r), and Vl for A(a,b,l)∧A(c,d,l). These predicates do not denote objects of C. Define
All other entries vanish. Thus Lr,s=Ls,r, with only the diagonal and step-two neighboring diagonals possibly nonzero. In particular L0,0=0 by an explicit branch. The lower neighbor uses
hj[(a+d−j+2)div2][(b+c−j+2)div2] when Hj−2∧Hj; the upper neighbor uses
hj+2[(a+d−j)div2][(b+c−j)div2] when Hj∧Hj+2. Otherwise those entries are zero.
The seed uses an independently defined explicit sum T(a,b,c,d,j,l). Define
otherwise T=0. This is an inclusive sum over integer indices, empty when the upper bound is negative. No separate support guard is present. Negative factorial arguments yield zero, and a zero factorial denominator makes the corresponding quotient zero by total division. Admissibility makes the relevant half-sums integral without rounding.
Labelled self-duality data
For every a∈N, choose an object Va, with an isomorphism V0≅1, and an exact self-pairing. Write its coevaluation and evaluation as
ua:1⟶Va⊗Va,ea:Va⊗Va⟶1.
Both snake identities hold: the composites
(ua⊗1a);(1a⊗ea) and
(1a⊗ua);(ea⊗1a), with their canonical unitors and associators, are each 1a. Here and below f;g means first f, then g. These are chosen pairings, not Hermitian inner products.
All category axioms
Scalar endomorphisms, uniquely: for every a∈N and f:Va→Va, there exists exactly one s∈K with f=s1a.
Disjoint labels: for all a,b∈N with a=b, every morphism Va→Vb is zero.
Fusion: for every a,b∈N, let
R(a,b)={c∈N:0≤c≤a+b,A(a,b,c)}.
There exist families of morphisms, defined for all natural c,
ic:Vc→Va⊗Vb and pc:Va⊗Vb→Vc, such that
No condition on these families outside R(a,b) is included, nor is a chosen identification with the vertex maps assumed.
4. Loops: for every natural a,
ua;ea=ι(Δ(a))11.
No braiding, symmetry, dagger, positivity, specified concrete representation category, or general pivotal-compatibility law is an additional hypothesis.
Vertices, their bends, and the precise tetrahedron
Choose for every natural triple a morphism
wabc:Va⊗Vb⊗Vc→1.
To specify all bends without suppressing their order, write 1a for 1Va, 1 for an identity on the object determined by the surrounding expression, and f⋆g for successive composition with only canonical monoidal associators and unitors inserted between the displayed morphisms. Multiple tensor factors retain their displayed order; no permutation or crossing is inserted.
For f:Va⊗Vb⊗Vc→1, its cyclic bend is
rot(f)=1⋆(ua⊗1Vb⊗Vc⊗Va)⋆(1a⊗(f⊗1a))⋆ea.
It is a morphism Vb⊗Vc⊗Va→1.
The all-up vertex, splitting map, and fusing map are defined, not independently chosen, by
Their respective types are 1→Va⊗Vb⊗Vc, Vl→Va⊗Vb, and Va⊗Vb→Vl. These formulas specify nested cups and the reversed vertex order in wabc and slab; they are linear categorical constructions, not complex conjugations.
Their types are respectively 1→Va⊗Vb⊗Vc⊗Vd and the reverse. For signed channels the graphical tetrahedron is exactly
T(a,b,c,d,j,l)={X(a,d,c,b,j);Hcov(a,d,c,b,l),0,j≥0 and l≥0,otherwise.
Only in the first branch are the nonnegative integers converted to natural channels. The boundary order is a,d,c,b, the channel in X is j, and the inner rung in the covector is l. This is not a definition using Hj;Xlcov, nor a definition of a tetrahedron by the recurrence. The definition has no separate admissibility guard for nonnegative channels.
Exactly two vertex-normalization conditions
For all natural a,b,c, assume
wabc;wabc=ι(θ(a,b,c))11,rot(wabc)=wbca.
These are theta normalization and cyclicity, including inadmissible triples. There is no tetrahedral evaluation in this normalization package.
Separate tetrahedral seed data
Independently assume, for every natural a,b,c,d,r,
T(a,b,c,d,r,2)=ι(T(a,b,c,d,r,2))11.
The scalar arguments are a,b,c,d,r,2, while the network's boundary presentation is a,d,c,b. This is an assumed last-slot-two evaluation, including zero values of the explicit sum on inadmissible labels. It is seed data used as an input to the assertion, not a conclusion or a consequence declared as part of vertex normalization. It does not separately assume penultimate-slot-two values or equality to the explicit sum at every inner channel.
Degenerate cases and meaning
All signed channels are included, with no boundary-index restrictions. A negative channel makes its graphical tetrahedron zero by definition. An inadmissible horizontal channel has zero scalar norm; every coefficient with that norm as denominator is zero by total field division. Adjacent entries require both horizontal channels to be admissible, and the diagonal at zero is separately set to zero. Thus no cancellation or division by a nonzero norm is tacitly required to interpret the equation.
The conclusion is a morphism-valued recurrence with independently specified scalar coefficients. No scalar-to-network bridge beyond the stated last-slot-two seed, no eigenvector equation, and no numerical matrix-entry evaluation is an extra assumption. Neither the scalar formulas nor the category assertion specialize q to a complex number; total division here is not a prescription for values at meromorphic poles. This testimony specifies the mathematical claim and its hypotheses, not a certification of a proof.
Preamble
import Definitions.Def_QuantumSixJRecurrences
universe v u
Formal statement
namespace QuantumSixJ
theorem network_recurrence : NetworkRecurrence.{v, u} := by sorry
end QuantumSixJ
Source
I. Khavkine, Recurrence relation for the 6j-symbol of su_q(2) as a symmetric eigenvalue problem, International Journal of Geometric Methods in Modern Physics (2015), https://doi.org/10.1142/S0219887815501170; https://arxiv.org/abs/1009.2261. Eqs. (23), (24), graphical-network recurrence.
Confirmed by the moderator at approval.