5.6. Function-field branches and the semismallness budget
The paper states Theorem 3.9 for a configuration over a finitely generated
extension K/k and reads it, equivalently, as a codimension bound on
subvarieties of X^\circ_{\mathcal{T}}. The formalization keeps only the
first reading. Its object is a function-field branch: a placement
\mathrm{pos} : V \to K^3 of the live vertices, together with proofs that
the coordinates generate K over k, that the placement is injective,
and that every terminal triple is collinear. The branch dimension is the
transcendence degree \trdeg_k K, and no codimension, stress bound, or
height conclusion is a field of the structure.
/-- The function field is generated by the labelled live coordinates. -/
def GeneratedByLiveCoordinates (a : V → Fin 3 → K) : Prop :=
IntermediateField.adjoin k (Set.range (liveCoordinateFamily a)) = ⊤/-- All labelled points in `X` lie on one affine line. The direction is
not required to be nonzero; pairwise distinctness supplies nondegeneracy on
the three-terminal flags where it is needed. -/
def AffinelyCollinearOn (a : V → Fin 3 → K) (X : Finset V) : Prop :=
∃ origin direction : Fin 3 → K,
∀ v ∈ X, ∃ t : K, a v = origin + t • direction/-- Every active provenance triple is collinear. -/
def FlagCollinearities (S : State V Flag) (a : V → Fin 3 → K) : Prop :=
∀ t : Flag, AffinelyCollinearOn a (S.terminals t)/-- The direction-stress dimension at the generic point of the branch. -/
def FunctionFieldBranch.stressDim
(S : State V Flag) (Y : FunctionFieldBranch (k := k) (K := K) S) : ℕ :=
DirectionStress.directionStressDim S.edges Y.position/--
The flag-semismall numerical statement in subtraction-free form:
`stress dimension + branch dimension + 2 * number of active flags`
is at most the full `3 * number of live vertices` ambient dimension.
All dimensions in this definition are read from actual Lean objects:
`ker`, `trdeg`, and the cardinalities of the exact active types.
-/
def FunctionFieldBranch.SemismallBudget
(S : State V Flag) (Y : FunctionFieldBranch (k := k) (K := K) S) : Prop :=
(Y.stressDim S : Cardinal) + Algebra.trdeg k K +
2 * (Fintype.card Flag : Cardinal) ≤
3 * (Fintype.card V : Cardinal)
The last definition is the inequality the whole induction proves. As noted
in the deletion chapter, the paper's defect
\Delta(G, \Gamma, a) = s + \trdeg_k K + 2|\Gamma| - 3|V| has no Lean
counterpart as a quantity; the statement \Delta \le 0 appears instead as
this predicate, in the subtraction-free form
s + \trdeg_k K + 2|\Gamma| \le 3|V|,
where s is
stressDim, the
self-stress dimension of the live graph at the branch's placement. The
paper's bookkeeping identities (3.16) and (3.18) below likewise have no
counterparts as equations: each induction step proves the parent's budget
from the child's budget and a local payment, and the arithmetic of the
defect is carried out inside those proofs.
Let (G = (V, E), \Gamma) be a sparse collinearity-flag system, let K/k
be a finitely generated field extension, and let a : V \to K^3 be an
injective configuration that realizes all the flags and whose coordinates
generate K; then
\dim_K \ker D_G(a)^T + \trdeg_k K + 2|\Gamma| \le 3|V|.
(Zheng, 2026, Theorem 3.9)
Lean code for Theorem5.6.1●2 declarations
Associated Lean declarations
-
theoremdefined in RB31EndToEnd/NullCellule/ProvenanceFlagSemismallnessFinal.leancomplete
theorem RB31E2E.ProvenanceFlag.provenanceFlag_semismallness.{u, v, w, x} {k : Type u} {K : Type v} {V : Type w} {Flag : Type x} [Field k] [Field K] [Algebra k K] [Fintype V] [DecidableEq V] [Fintype Flag] [DecidableEq Flag] (S : RB31E2E.ProvenanceFlag.State V Flag) (hSparse : S.CompletionSparse) (Y : RB31E2E.ProvenanceFlag.FunctionFieldBranch S) : RB31E2E.ProvenanceFlag.FunctionFieldBranch.SemismallBudget S Y
theorem RB31E2E.ProvenanceFlag.provenanceFlag_semismallness.{u, v, w, x} {k : Type u} {K : Type v} {V : Type w} {Flag : Type x} [Field k] [Field K] [Algebra k K] [Fintype V] [DecidableEq V] [Fintype Flag] [DecidableEq Flag] (S : RB31E2E.ProvenanceFlag.State V Flag) (hSparse : S.CompletionSparse) (Y : RB31E2E.ProvenanceFlag.FunctionFieldBranch S) : RB31E2E.ProvenanceFlag.FunctionFieldBranch.SemismallBudget S Y
**Provenance-flag semismallness.** For every literal flagged state whose private-ghost completion is `(2,2)`-sparse, every coordinate-generated injective function-field branch satisfies the stress/codimension budget. The structures `State` and `FunctionFieldBranch` contain no semismallness or height conclusion.
-
defdefined in RB31EndToEnd/NullCellule/ProvenanceFlagBranch.leancomplete
def RB31E2E.ProvenanceFlag.FunctionFieldBranch.SemismallBudget.{u_1, u_2, u_3, u_4} {k : Type u_1} {K : Type u_2} {V : Type u_3} {Flag : Type u_4} [Field k] [Field K] [Algebra k K] [Fintype V] [DecidableEq V] [Fintype Flag] [DecidableEq Flag] (S : RB31E2E.ProvenanceFlag.State V Flag) (Y : RB31E2E.ProvenanceFlag.FunctionFieldBranch S) : Prop
def RB31E2E.ProvenanceFlag.FunctionFieldBranch.SemismallBudget.{u_1, u_2, u_3, u_4} {k : Type u_1} {K : Type u_2} {V : Type u_3} {Flag : Type u_4} [Field k] [Field K] [Algebra k K] [Fintype V] [DecidableEq V] [Fintype Flag] [DecidableEq Flag] (S : RB31E2E.ProvenanceFlag.State V Flag) (Y : RB31E2E.ProvenanceFlag.FunctionFieldBranch S) : Prop
The flag-semismall numerical statement in subtraction-free form: `stress dimension + branch dimension + 2 * number of active flags` is at most the full `3 * number of live vertices` ambient dimension. All dimensions in this definition are read from actual Lean objects: `ker`, `trdeg`, and the cardinalities of the exact active types.
Write n = |V|, j = |\Gamma|, s = \dim_K \ker D_G(a)^T and
\Delta(G, \Gamma, a) = s + \trdeg_k K + 2j - 3n; the assertion is
\Delta \le 0, proved by strong induction on n. If n = 0 then
E = \Gamma = \emptyset and K = k, so \Delta = 0. Otherwise the
selection lemma gives a vertex v of one of its two types.
An outside vertex of degree at most three. Delete v and keep every
flag; the child completion embeds in \widehat{G} - v, so the child system
is sparse, and the ledger gives
\Delta(G, \Gamma, a) = \Delta(H, \Gamma, a_H) + (u + \delta_v - 3). If
u + \delta_v \le 3 the induction hypothesis applies to H = G - v.
Otherwise the low-degree classification gives \delta_v = 3, u = 1 and
three distinct collinear neighbours N whose pairwise rigidity rows lie in
\operatorname{row}_L D_H(a_H). If \widehat{G} - v is not complete on
N, the outside augmentation lemma gives an addable nonedge f; it is a
certified response edge, so adding it raises the self-stress dimension of
H from t to t + 1 = s, since the ledger gives s = t + u with
u = 1. Every retained flag remains valid, and
\Delta(H + f, \Gamma, a_H) = \Delta(G, \Gamma, a) with one live vertex
fewer. If instead the neighbour triangle is complete, its three edges are
live and no two of its vertices share a flag; delete one triangle edge
d_N and register (T_\star, d_\star) = (N, d_N) with a new ghost as a
new flag. Mapping the new ghost to v embeds the new completion into
\widehat{G}, so the enlarged system is sparse. Removing one collinear
triangle row preserves the rank, so the stress dimension becomes s - 2,
while j increases by one and n decreases by one; hence \Delta is
unchanged and the induction hypothesis applies again.
A private support vertex of degree at most two. Pivot so that
d_\gamma = vq with vp, pq live; the pivot changes neither the
completion nor the stress dimension. Delete v and the flag \gamma
together; the child completion embeds in \widehat{G}, and the ledger now
reads
\Delta(G, \Gamma, a) = \Delta(H, \Gamma \setminus \{\gamma\}, a_H) +
(u + \delta_v - 1). If u + \delta_v \le 1 the induction hypothesis
applies. Otherwise the private classification forces \delta_v = u = 1,
degree exactly two, and the two candidate rows r_{pz}, r_{qz} in the
retained row space; by the private augmentation lemma one of the two
candidates is an addable edge f of the child completion, its insertion is
a certified response edge, the stress dimension is restored to s, and
\Delta(H + f, \Gamma \setminus \{\gamma\}, a_H) = \Delta(G, \Gamma, a)
with one live vertex fewer.
The formal theorem is
provenanceFlag_semismallness:
for every state whose completion is sparse and every function-field branch
on it, the semismallness budget holds. The strong induction is on the size
Fintype.card of the live vertex type, which strictly
decreases at every recursive call because each move constructs its child on
the subtype of remaining vertices. The base case is
the empty live type, where the state's own fields force \Gamma to be
empty and coordinate generation forces \trdeg_k K = 0. Each of the four
cases of the induction is proved in its own module: the two nonexceptional
deletions rest on the local payments proved in the deletion ledger; in the
outside exceptional case an edge is inserted or the neighbour triple is
registered, with the stress and flag terms changing by exactly two each in
the registered alternative; and in the private exceptional case the branch
is transported across the pivot before the deletion. We describe those
transitions one by one in the closing section.
One hypothesis differs in form. The paper redefines the retained data after
each move and reuses the ambient field K; the formalization measures each
child by the intermediate field generated by the child's own coordinates
inside K, proves that direction-stress rank is unchanged under that
restriction of scalars, and descends collinearity of the retained triples to
the child field. The equivalence between the two measurements is the
intrinsic-placement interface of the deletion ledger, and it is why the
induction can pass to a subtype without rebuilding the function field at
every step.
Let F be a finite simple (2,2)-sparse graph on the vertex set V,
let K/k be a finitely generated field extension, and let
a : V \to K^3 be injective with coordinates generating K; then
\dim_K \ker D_F(a)^T + \trdeg_k K \le 3|V|.
(Zheng, 2026, Theorem 1.2)
Lean code for Theorem5.6.2●2 theorems
Associated Lean declarations
-
theoremdefined in RB31EndToEnd/NullCellule/ProvenanceFlagGroundedPF.leancomplete
theorem RB31E2E.ProvenanceFlagGroundedPF.groundedPF_of_provenanceFlag_semismallness (hFlagPF : ∀ {k K V Flag : Type} [inst : Field k] [inst_1 : Field K] [inst_2 : Algebra k K] [inst_3 : Fintype V] [inst_4 : DecidableEq V] [inst_5 : Fintype Flag] [inst_6 : DecidableEq Flag] (S : RB31E2E.ProvenanceFlag.State V Flag), S.CompletionSparse → ∀ (Y : RB31E2E.ProvenanceFlag.FunctionFieldBranch S), RB31E2E.ProvenanceFlag.FunctionFieldBranch.SemismallBudget S Y) (V : Type) [Fintype V] [DecidableEq V] (F : RB31E2E.SimpleEdgeSet V) : RB31E2E.Sparse22 F → ∀ (root : V) {K : Type} [inst : Field K] [inst✝ : Algebra ℚ K] (a : V → Fin 3 → K), a root = 0 → Function.Injective a → IntermediateField.adjoin ℚ (Set.range fun x ↦ a x.1 x.2) = ⊤ → RB31E2E.DirectionStress.directionStressDim F a + Cardinal.toNat (Algebra.trdeg ℚ K) ≤ Nat.card (RB31E2E.GroundedTwistSplit.SpatialVariable root)
theorem RB31E2E.ProvenanceFlagGroundedPF.groundedPF_of_provenanceFlag_semismallness (hFlagPF : ∀ {k K V Flag : Type} [inst : Field k] [inst_1 : Field K] [inst_2 : Algebra k K] [inst_3 : Fintype V] [inst_4 : DecidableEq V] [inst_5 : Fintype Flag] [inst_6 : DecidableEq Flag] (S : RB31E2E.ProvenanceFlag.State V Flag), S.CompletionSparse → ∀ (Y : RB31E2E.ProvenanceFlag.FunctionFieldBranch S), RB31E2E.ProvenanceFlag.FunctionFieldBranch.SemismallBudget S Y) (V : Type) [Fintype V] [DecidableEq V] (F : RB31E2E.SimpleEdgeSet V) : RB31E2E.Sparse22 F → ∀ (root : V) {K : Type} [inst : Field K] [inst✝ : Algebra ℚ K] (a : V → Fin 3 → K), a root = 0 → Function.Injective a → IntermediateField.adjoin ℚ (Set.range fun x ↦ a x.1 x.2) = ⊤ → RB31E2E.DirectionStress.directionStressDim F a + Cardinal.toNat (Algebra.trdeg ℚ K) ≤ Nat.card (RB31E2E.GroundedTwistSplit.SpatialVariable root)
Empty-flag semismallness implies the rooted direction-stress inequality. Grounding and all cardinal conversions are proved in this declaration.
-
theoremdefined in RB31EndToEnd/NullCellule/GroundedPFEndToEnd.leancomplete
theorem RB31E2E.endToEndBodyPinStatement_of_groundedPF (hPF : ∀ (V : Type) [Fintype V] [inst : DecidableEq V] (F : RB31E2E.SimpleEdgeSet V), RB31E2E.Sparse22 F → ∀ (root : V) {K : Type} [inst_1 : Field K] [inst_2 : Algebra ℚ K] (a : V → Fin 3 → K), a root = 0 → Function.Injective a → IntermediateField.adjoin ℚ (Set.range fun x ↦ a x.1 x.2) = ⊤ → RB31E2E.DirectionStress.directionStressDim F a + Cardinal.toNat (Algebra.trdeg ℚ K) ≤ Nat.card (RB31E2E.GroundedTwistSplit.SpatialVariable root)) : RB31E2E.EndToEndBodyPinStatement
theorem RB31E2E.endToEndBodyPinStatement_of_groundedPF (hPF : ∀ (V : Type) [Fintype V] [inst : DecidableEq V] (F : RB31E2E.SimpleEdgeSet V), RB31E2E.Sparse22 F → ∀ (root : V) {K : Type} [inst_1 : Field K] [inst_2 : Algebra ℚ K] (a : V → Fin 3 → K), a root = 0 → Function.Injective a → IntermediateField.adjoin ℚ (Set.range fun x ↦ a x.1 x.2) = ⊤ → RB31E2E.DirectionStress.directionStressDim F a + Cardinal.toNat (Algebra.trdeg ℚ K) ≤ Nat.card (RB31E2E.GroundedTwistSplit.SpatialVariable root)) : RB31E2E.EndToEndBodyPinStatement
Grounded semismallness for every sparse simple edge system implies the body--pin theorem.
The paper proves this by applying Theorem 3.9 with G = F and
\Gamma = \emptyset, so that \widehat{G} = F and (3.14) is (1.5). The
formalization makes the same specialization:
emptyFlagState puts
the empty flag type on F, and its completion is F itself. It never
states (1.5) as a standalone theorem, however. The assembly of
the main theorem uses a grounded form instead:
the placement sends a root vertex to zero, the base field is \mathbb{Q},
and the right-hand side is 3(|V| - 1), the number of spatial variables
left after grounding.
groundedPF_of_provenanceFlag_semismallness
derives that form from flag semismallness. It adjoins three independent
translation variables to K and translates the placement by them;
translatedPlacement
leaves every direction row, hence the stress space, unchanged, while the
transcendence degree grows by exactly three. Applying the empty-flag budget
over the enlarged field and cancelling those three dimensions against the
grounding gives the grounded inequality, which then enters
endToEndBodyPinStatement_of_groundedPF
as its sole hypothesis, and
endToEndBodyPinStatement_of_provenanceFlag_semismallness
is the composite, from the flag theorem to the body–pin statement. The
grounded model itself is the subject of
the strata chapter.