Theorem 5 — total and level Fourier mass from the accepting-set size
OpenIQPForrelation832.SingleQuery.Supplement.theorem_5computational-complexityforrelationiqp-circuitsp2o-batch-stoc27sp2o-gran-per-chapterp2o-plan-paperp2o-strengthenedp2o-v1quantum-information
Let be the actual acceptance probability of any -query IQP circuit with any accepting subset of the -bit computational basis. For every restriction , the restricted function has its multilinear Fourier expansion of degree at most , and
In particular, for every natural level ,
The sum includes the constant coefficient. The constant is exactly one, the acceptance event is the unconditioned Born event, and can be empty. Every restriction, query overlap, repeated address, ancilla width and zero-query case is retained. This companion theorem bounds total Fourier mass by the actual accepting-set size, independently of the query-dependent bound of Theorem 6.
Differences from the source:
- generalization: polynomial-size IQP middle blocks → arbitrary finite diagonal-unitary middle gate lists; reason: the Section 5 arguments do not use the number of non-oracle gates; every source circuit is included
- encoding: sign assignments on the free set Gamma and its subsets → Boolean signs on the exact free-coordinate subtype; reason: true represents +1 and false represents -1; all assignments and restrictions correspond bijectively, including the empty free set
Preamble
import Definitions.Def_IQPForrelation832_SingleQuery_Supplement_OracleFourier set_option autoImplicit false
Formal statement
namespace IQPForrelation832.SingleQuery.Supplement
open QAlgorithms QAlgorithms.Complexity
theorem theorem_5 {n m d : ℕ} (c : PhaseMiddle n m)
(hc : PhaseIQP c) (hd : OracleCircuit.queryCount c 0 = d)
(F : Finset (Qubits m)) (ρ : Restriction n) :
MultilinearDegreeLE (phaseBorn c F) ρ (2*d) ∧
totalMass (phaseBorn c F) ρ ≤ Real.sqrt (F.card : ℝ) ∧
(∀ ell : ℕ, ell ≤ 2*d → levelMass (phaseBorn c F) ρ ell ≤ Real.sqrt (F.card : ℝ)) ∧
(∀ ell : ℕ, 2*d < ell → levelMass (phaseBorn c F) ρ ell = 0) := by sorry
end IQPForrelation832.SingleQuery.Supplement
Source
Quentin Buzet, André Chailloux, IQP circuits for 2-Forrelation, STOC 2027 submission #832 (October 2026), p. 26, Theorem 5; proof pp. 26-27. https://github.com/sigact-stoc/stoc2027/releases/download/papers-1/832.pdf