The Appendix-sum and categorical recurrence propositions
DefinitionQuantumSixJRecurrencesThis module defines two separately named propositions, QuantumSixJ.SumRecurrence and QuantumSixJ.NetworkRecurrence. Both describe equation (24); the graphical Tet is defined in equation (6). These are definitions of propositions, not assumptions or proofs of either recurrence.
For fixed external labels, denotes the symmetric tridiagonal matrix, with and unless . Its entries are indexed on a fixed parity lattice; . In Lean, QuantumSixJ.L a b c d j k is the two-index matrix entry . The off-diagonal formulas give , including zero extensions; symmetry is not an extra hypothesis.
SumRecurrence: Appendix sum. For every , the explicit finite Appendix sum, normalized by , satisfies
Here in , with the published zero extensions and no additional hypotheses. Equivalently, , displaying the symmetric generalized eigenvalue problem.
NetworkRecurrence: categorical network. In every -linear preadditive monoidal category with additive, linear tensor product, labeled self-dual objects and trivalent vertices satisfyingSuQ2Axioms,VertexNormalized, and the separate initial-data assumptionTetSeed, for every and ,
The coefficients act on endomorphisms of the tensor unit. The graphical network of equation (6) is when both channels are nonnegative and zero otherwise, with composition read left to right. The last channel is the inner rung.
The category assumptions are scalar endomorphisms of each label object, vanishing morphisms between different labels, the fusion decompositions and loop values. Vertex normalization requires only theta evaluation and cyclicity. The separate assumption TetSeed supplies the last-slot-2 Tet values as initial data for the recurrence solution and as input to its derivation. No second numerical Tet evaluation, general rotation, or all-channel equality between the network and Appendix sum is assumed. In both clauses field division by zero is zero; complex specialization is not part of the statement. The scalar clause is independent of all categorical hypotheses.
The scalar Tet is defined by QuantumSixJ.tetSum, and the graphical Tet by QuantumSixJ.Network.Vertex.tet. Both recurrence assertions are part of the conclusion, not hypotheses.
A theorem with statement SumRecurrence, or with statement NetworkRecurrence at the required universe levels, can be introduced independently. The joint theorem recurrences asserts their conjunction.
import Definitions.Def_QuantumSixJScalar
import Definitions.Def_QuantumSixJNetwork
open CategoryTheory MonoidalCategory
open QuantumSixJ.Network
namespace QuantumSixJ
/-- Equation (24) for the explicit Appendix tetrahedral sum. -/
def SumRecurrence : Prop :=
∀ a b c d j l : ℤ,
L a b c d j (j - 2) * normalizedTetSum a b c d (j - 2) l +
(L a b c d j j - Λ a b l * h a b c d j) * normalizedTetSum a b c d j l +
L a b c d j (j + 2) * normalizedTetSum a b c d (j + 2) l = 0
/-- Equation (24) for the graphical Tet of equation (6), under the
specified category axioms, vertex normalization, and separate Tet seed data.
-/
def NetworkRecurrence.{v, u} : Prop :=
∀ {C : Type u} [Category.{v} C] [Preadditive C] [Linear K C]
[MonoidalCategory C] [MonoidalPreadditive C] [MonoidalLinear K C],
∀ (S : SuQ2Data C) (W : Vertex S), SuQ2Axioms S → VertexNormalized W → TetSeed W →
∀ (a b c d : ℕ) (j l : ℤ),
(ι (L a b c d j (j - 2)) / ι (h a b c d (j - 2))) • W.tet a b c d (j - 2) l +
(ι (L a b c d j j - Λ a b l * h a b c d j) / ι (h a b c d j)) • W.tet a b c d j l +
(ι (L a b c d j (j + 2)) / ι (h a b c d (j + 2))) • W.tet a b c d (j + 2) l = 0
end QuantumSixJ
Read-back
What the Lean code literally says, in plain math · openai/gpt-6.1-sol
Two tetrahedral recurrence propositions
Scalar quantities and conventions
Let , where is an indeterminate. All scalar operations below are operations in this rational-function field, not evaluation at a complex number. In particular, inversion is total, with , and division by zero has value zero. For each integer , put
The empty product is one, so . Every integer quotient by below means Euclidean integer division (rounding down), including on negative inputs; it is not rational division. Powers with integer exponents are field powers.
A triple of integers is admissible precisely when all three entries are nonnegative, each is at most the sum of the other two, and is even. Write for admissibility of both and , and for admissibility of both and . Define
Here and below always uses the displayed external labels ; and mean the same three-argument function with exactly those substitutions.
For every pair of integer indices , the coefficient is defined as follows:
In particular even if channel zero is admissible. The off-diagonal formula uses the norm at the larger index and the two brackets at the smaller index; thus . All other diagonals are zero. In the recurrence, the lower-neighbor coefficient is , the central coefficient is , and the upper-neighbor coefficient is . In particular, when their admissibility tests hold, , whereas .
For six integer labels define the four-entry and three-entry lists
with the integer-division convention above. Put
The defining sum is finite and is zero when its integer interval is empty. The summand has no extra support restriction: a negative factorial argument makes that factorial zero, and a resulting zero denominator makes the entire quotient zero. All other quotients, including when , use the same total-field convention.
Sum recurrence
SumRecurrence. For every six integers , the quantities just defined satisfy the following equality in :
This proposition has no hypotheses: external labels and both channels range over all integers, including negative labels, zero, inadmissible triples, both parities, and indices with missing admissible neighbors. The neighbors are the signed integers and , not truncated natural-number differences; their values are governed solely by the definitions above. No categorical data, seed values, specialization of , nonvanishing assumptions, or separate admissibility assumptions are quantified or required.
Categorical data and the three separate assumptions
Fix universes and . Let be any category whose objects form a type in universe and whose morphisms have universe , and put . Require that is preadditive (each morphism set is an abelian group and composition is additive in each argument), -linear (each morphism set is a -module and composition is -linear in each argument), monoidal, monoidally preadditive, and monoidally -linear (tensoring by any fixed object preserves addition and scalar multiplication on morphisms). Write for the tensor unit and for the coefficient inclusion induced by , fixing the indeterminate . Scalar multiplication of morphisms is over . The morphism zero below is the additive zero, and denotes the identity of an object .
The label data consist of objects for every , a chosen isomorphism , and a chosen exact self-pairing for every . Concretely, each pairing supplies a cup and cap satisfying both snake identities
Here means first , then , and throughout these formulas the canonical monoidal associators and unitors are inserted to match source and target; no permutation or crossing of tensor factors is implicit. The vertex data consist of an arbitrary chosen morphism for every triple of natural labels, not merely admissible triples.
For clarity, the following derived morphisms specify exactly which cups, caps, and vertices occur. Write in tensor expressions for the object , and omit only the canonical associators and unitors. Define
Their types are respectively , , , , and . The superscript is just a name for this covector, not complex conjugation or a Hermitian adjoint. For , define its rotation by
The closed tetrahedral morphism, for natural external labels and signed integer channels, is
In its nonzero branch the channels are converted to their equal natural-number values. The intermediate boundary order is exactly ; is the channel of and is the channel of , with no additional admissibility guard in this definition.
The assumption SuQ2Axioms on has four fields: (i) for every natural and every endomorphism of , there exists a unique such that ; (ii) for distinct natural labels , every morphism is zero; (iii) for every natural , there exist families and indexed by all natural , such that, writing , one has for every and ; the families have no stated conditions outside this finite set, and no further pairwise orthogonality condition is a field; (iv) for every natural , . The existence of the fusion families is asserted separately for each ; these families are not required to equal morphisms derived from .
The separate assumption VertexNormalized on has exactly two fields: for every natural triple , ; and for every such triple, . These equalities apply also to inadmissible triples, where the scalar theta is zero. The further separate assumption TetSeed on has exactly one field: for every five natural labels ,
The scalar on the right uses external arguments in that order; only the closed-network construction uses boundary order . This seed condition is restricted to last channel and natural first channel, but includes zero and all inadmissible choices; it is not an assumed evaluation at every last channel and is not a field of either of the other two assumptions.
Network recurrence
NetworkRecurrence at universes . For every category with object universe and morphism universe , for every choice of each of the six structures specified above (category, preadditive, -linear, monoidal, monoidally preadditive, and monoidally -linear), for every label datum and vertex datum on , if SuQ2Axioms holds for , VertexNormalized holds for , and, separately, TetSeed holds for , then for every and every one has the following equality in :
Each product here is scalar multiplication of a morphism, not composition; each fraction lies in and uses total division, so it is zero whenever its displayed denominator is zero. Every scalar uses the natural external labels included in the integers, and use their own respective signed channel indices. There is no assumption that those denominators are nonzero, that either channel or any neighbor is admissible, or that the external labels are positive. All zero external labels, negative channels, both parities, , and absent neighbors are included; negative-channel tetrahedral morphisms are zero by definition. For each fixed pair of universe levels this is a proposition universally quantifying , its structures, , and , followed by the three implications and then the six label quantifiers; universe levels themselves are parameters of the family of propositions, not additional mathematical objects quantified inside it. It asserts no existence of data satisfying the three assumptions and makes no assertion about a choice failing any one of them; for a choice where those assumptions cannot all hold, the implication imposes no conclusion. It neither assumes the scalar recurrence nor asserts that closed tetrahedral morphisms agree with the scalar finite sum at arbitrary last channel.