Theorem 2.4 — semantic ΣΠΣΠ-PC simulates semantic Res(PC) over any field
ProvedSemAC02.Supplement.theorem_2_4_res_pc_simulationalgebraic-circuitsp2o-gran-per-chapterp2o-plan-paperp2o-v1pigeonhole-principlepolynomial-calculusproof-complexitysemantic-proof-systems
Let be any field and a CNF formula with polynomial encoding over . If has an arity-2 semantic Res(PC) refutation in which every polynomial is written as a sum of monomials, then has an arity-2 semantic ΣΠΣΠ-PC refutation with
- total circuit size , the constant absolute;
- if has lines;
- a ΣΠΣΠ[]-PC refutation whenever is a semantic Res(PC) refutation.
The simulation preserves the inference arity.
Lower bounds for semantic ΣΠΣΠ[w]-PC therefore apply to semantic Res(PC).
Formalization Note Total circuit size is csize (shared subcircuits) and is resPCSize. Both systems have arity 2, the arity the paper uses throughout.
Differences from the source:
- encoding: proof lines are displayed circuits representing classes in B_n; initial lines are the axioms → ACirc terms; a line is an axiom line when its value on the cube equals an axiom; initial lines equal to 0 are allowed; reason: p.15: a line represents its residue class; p.54: initial tautologies 0 = 0; a zero line follows from any line, so no refutation or measure changes
- reading: total circuit size of π′ (the theorem does not say whether subcircuits shared across lines are counted once) → csize: distinct subterms across all lines (the C-size of p.15, sharing across lines); reason: the proof reuses shared polynomial subcircuits across the derivation (p.19); the per-line sum |π|₄ of (32) can exceed size(π) by a factor L when one polynomial occurs in many lines, so only the sharing reading makes the statement true
- reading: size(π) counts distinct rooted subcircuits across the derivation → resPCSize counts distinct lines, polynomials, monomials and variables; reason: each line is the gate joining its polynomials; the proof adds O(1) gates per distinct line, so the lines are counted
- reading: axiom lines are the clauses of ϕ → literal x_i written as 1 − x_i = 0, ¬x_i as x_i = 0 (litPolyF), in any order; reason: the paper's truth-value polynomials (p.14); the translated axiom is then ⟦C⟧
- reading: Res(PC_w): every polynomial has degree at most w → totalDegree of each displayed polynomial ≤ w; reason: the degree of the polynomial as written; this is what gives bottom fan-in ≤ w
- encoding: the simulation preserves the inference arity → π and π' are both arity-2 semantic refutations; reason: the paper uses arity 2 throughout (p.16)
- encoding: total circuit size O(size(π)) → ∃ K : ℝ chosen before the field, csize π' ≤ K·resPCSize π; reason: the O-constant is absolute
Preamble
import Mathlib import Definitions.Def_LiftSAParity_CNF import Definitions.Def_SemAC02_Supplement_ACirc import Definitions.Def_SemAC02_Supplement_IsSemPCRef import Definitions.Def_SemAC02_Supplement_clauseAx import Definitions.Def_SemAC02_Supplement_ResPC
Formal statement
namespace SemAC02.Supplement
open ACirc
/-- **Theorem 2.4 (Simulation of Res(PC))** (Alekseev–Gryaznov–Itsykson–Pitassi–Robere–Tzameret,
*Hard Instances for Strong Semantic Proof Systems*, manuscript (STOC 2027 submission), Theorem 2.4,
p.19): let `F` be any field and `ϕ` a CNF formula. Suppose that `ϕ` has a semantic Res(PC)
refutation `π`, every polynomial represented explicitly as a sum of monomials. Then the polynomial
encoding `F_ϕ` has a semantic ΣΠΣΠ-PC refutation `π'` of total circuit size `O(size(π))`. The
simulation preserves the inference arity. If `π` has `L` lines, then `S(π') ≤ L`. If `π` is a
semantic Res(PC_w) refutation, then `π'` is a semantic ΣΠΣΠ[w]-PC refutation.
**Formalization Note** `O(size(π))` is an absolute constant `K` chosen before the field. The total
circuit size of `π'` is its `C`-size `csize` (sharing across lines) and `size(π)` is `resPCSize`.
Both systems are arity-2 semantic systems (the paper uses arity 2 throughout, p.16), which is the
preserved arity. Res(PC_w): every polynomial of `π` has degree at most `w`. -/
theorem theorem_2_4_res_pc_simulation :
∃ K : ℝ, ∀ (F : Type) [Field F] (n : ℕ) (φ : LiftSAParity.CNF (Fin n))
(π : List (List (MvPolynomial (Fin n) F))), IsSemResPCRef φ π →
∃ π' : List (ACirc F (Fin n)),
IsSemPCRef (IsSPSP ⊤) Set.univ (clauseAx (φ : Set (LiftSAParity.Clause (Fin n)))) π' ∧
(csize π' : ℝ) ≤ K * resPCSize π ∧
topS Set.univ π' ≤ π.length ∧
∀ w : ℕ, (∀ D ∈ π, ∀ f ∈ D, f.totalDegree ≤ w) → ∀ c ∈ π', IsSPSP (w : ℕ∞) c := by sorry
end SemAC02.Supplement
Source
Alekseev, Gryaznov, Itsykson, Pitassi, Robere, Tzameret, Hard Instances for Strong Semantic Proof Systems: CNFs for Depth-Four PC and Deep Instances for AC0[2]-Frege, manuscript (STOC 2027 submission, sha256 c228d68b89b45b5c6c547aa994f7d1799441bb22b2a85e36c6cd400bea8cdb6a), Theorem 2.4, p.19 (PDF p.21)