Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma 38 — effect of letters

Proved
BuchiRabinIndex.LowerBound.lemma_38

by mikedeng1 · Nov 10, 2026 · Mathlib 0df444a (Lean v4.33.1)

automata-theoryfinite-combinatoricsp2o-batch-stoc27ap2o-gran-per-chapterp2o-plan-paperp2o-v1rabin-index

Let H=⟨T,l⟩H=\langle T,l\rangleH=⟨T,l⟩ be any simple history of FnF_nFn​, n≥1n\ge1n≥1, with root state set PPP. The four source cases have the following exact effects.

  1. If qn∉Pq_n\notin Pqn​∈/P and qp∈Pq_p\in Pqp​∈P, generation by idP∪{(qp,qn)}\mathrm{id}_P\cup\{(q_p,q_n)\}idP​∪{(qp​,qn​)} inserts the new rightmost child of the node owning qpq_pqp​, with label {qn}\{q_n\}{qn​}. Every old own set is unchanged; both annotation sets are empty.
  2. Every injection π:P→Q∖{qn}\pi:P\to Q\setminus\{q_n\}π:P→Q∖{qn​} gives the image-labelled tree ⟨T,π∘l⟩\langle T,\pi\circ l\rangle⟨T,π∘l⟩, with both annotation sets empty.
  3. If a leaf xxx has label {qs}\{q_s\}{qs​} and an injection π:P→Q\pi:P\to Qπ:P→Q sends qsq_sqs​ to qnq_nqn​, its image-labelled tree has accepting set exactly {x}\{x\}{x} and unstable set empty.
  4. For a youngest-child leaf zzz labelled {qs}\{q_s\}{qs​}, deletion has accepting set empty and unstable set exactly {z}\{z\}{z}. If qn∉P∖{qs}q_n\notin P\setminus\{q_s\}qn​∈/P∖{qs​}, all other own sets stay fixed. Otherwise, when qnq_nqn​ labels another leaf xxx, the primed deletion makes xxx own qsq_sqs​ and leaves every other surviving own set fixed.

Every successor is required to exist and remain simple. These are exact successor and annotation equations, not merely existence of some tree having the indicated shape.

In particular, the exact annotation pairs in cases G, P, A and D are

(acc⁡,unst⁡)=(∅,∅), (∅,∅), ({x},∅), (∅,{z}),\bigl(\operatorname{acc},\operatorname{unst}\bigr)=(\varnothing,\varnothing),\ (\varnothing,\varnothing),\ (\{x\},\varnothing),\ (\varnothing,\{z\}),(acc,unst)=(∅,∅), (∅,∅), ({x},∅), (∅,{z}),

respectively; the primed deletion has the same annotation pair as D.

Differences from the source:

  • encoding: nodes are finite words of positive integers → List Nat with positivity in node and tree predicates; reason: children retain the source’s indices starting at 1
  • encoding: states q_1,...,q_n and alphabet 2^(Q×Q) → Fin n with q_i represented by i−1 and Finset transition relations; reason: this is a bijective indexing of all states and letters
  • encoding: labels are functions defined on the node set → labels extend by the empty set outside the node set, with a validity subtype; reason: the extension is unique and adds no histories or transitions
  • encoding: deterministic transitions may fail to have a successor → Option history trees, with a dying trace rejected by every condition; reason: root deletion has no history-tree successor; no sink is added
  • reading: step 4 calls nodes whose descendants are deleted accepting → only the topmost resetting nodes survive to receive accepting annotations; reason: a nested reset below one of these nodes is deleted and any old such node is unstable
  • encoding: injective maps with domain P → total maps on Fin n whose restrictions to P are injective; reason: the finite nonempty target admits an extension, and values off P are never used
Preamble
import Definitions.Def_BuchiRabinIndex_LowerBound_Witness

set_option autoImplicit false
Formal statement
namespace BuchiRabinIndex.LowerBound

/-- STOC 2027 submission #1731, lemma 38; statement only. -/
theorem lemma_38 (n : ℕ) (hn : 1 ≤ n) (H : History (Fin n)) (hH : Simple H) :
    (∀ p ∈ stateSet H.val, lastState n hn ∉ stateSet H.val →
      Effect hn H (genLetter hn (stateSet H.val) p) (generatedData hn H.val p) ∅ ∅ ∧
      (∀ x ∈ H.val.nodes, own (generatedData hn H.val p) x = own H.val x) ∧
      (generatedData hn H.val p).label
        (owner H.val p ++ [degree H.val.nodes (owner H.val p) + 1]) = {lastState n hn}) ∧
    (∀ π : Fin n → Fin n,
      Set.InjOn π {q | q ∈ stateSet H.val} →
      (∀ q ∈ stateSet H.val, π q ≠ lastState n hn) →
      Effect hn H (permLetter (stateSet H.val) π) (mapData H.val π) ∅ ∅) ∧
    (∀ x s (π : Fin n → Fin n), Leaf H.val.nodes x → H.val.label x = {s} →
      Set.InjOn π {q | q ∈ stateSet H.val} → π s = lastState n hn →
      Effect hn H (permLetter (stateSet H.val) π) (mapData H.val π) {x} ∅) ∧
    (∀ z s, Leaf H.val.nodes z → Youngest H.val.nodes z → H.val.label z = {s} →
      (lastState n hn ∉ (stateSet H.val).erase s →
        Effect hn H (delLetter (stateSet H.val) s) (deletedData H.val z s) ∅ {z} ∧
        ∀ x ∈ H.val.nodes, x ≠ z → own (deletedData H.val z s) x = own H.val x) ∧
      (∀ x, x ≠ z → Leaf H.val.nodes x → H.val.label x = {lastState n hn} →
        lastState n hn ∈ (stateSet H.val).erase s →
        Effect hn H (delPrimeLetter hn (stateSet H.val) s)
          (deletedPrimeData hn H.val z s) ∅ {z} ∧
        own (deletedPrimeData hn H.val z s) x = {s} ∧
        ∀ u ∈ H.val.nodes, u ≠ z → u ≠ x →
          own (deletedPrimeData hn H.val z s) u = own H.val u)) := by sorry

end BuchiRabinIndex.LowerBound
Source
Cong Tian, Runzhe Ma, Wensheng Wang, Zhenhua Duan, Büchi Determinization with Optimal States and Square Root Index Improvement, STOC 2027 submission #1731 (October 2026), Lemma 38, p. 27; https://github.com/sigact-stoc/stoc2027/releases/download/papers-2/1731.pdf
Human review
  • Endorsed by Shuze Chen · Nov 10, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Nov 10, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me