Alphabeta Math
LemmaStatement: Literature-sourcedProof: AI-adaptedPipeline-generatedprecheck passjudge pass (gpt-6-luna)
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

Symmetric BAOAB one-step weak expansion

Statement

Assume the underdamped model in Underdamped Langevin equation and canonical Gibbs density, with V∈C8, ∇V of linear growth, and DjV globally bounded for 2≤j≤8. Let ϕ∈CP6(R2d) as defined in that item, and let Kh be the one-step BAOAB transition operator and Ph the exact SDE transition operator. For 0<h≤1 there are C,r<∞ such that ∣Khϕ(z)−Phϕ(z)∣≤Ch3(1+∣z∣r)(z∈R2d). The exponent r may depend on the polynomial-growth bounds for ϕ and its derivatives. If ϕ∈Cb6, one may take r=3; the constant is uniform over test families with a common bound on all derivatives of orders zero through six. This is a finite-time local weak estimate; it is not an invariant-measure bias statement.

Facts & Assumptions

Given: AC, the smoothness and growth conditions in the statement, 0<h≤1, and the BAOAB and exact SDE transition operators.

[F1]

Underdamped Langevin equation and canonical Gibbs density and Multidimensional Ito formula for Brownian-driven processes: writing F=−∇V and S=2γβ−1M1/2, the SDE has drift b(q,p)=(M−1p,F(q)−γp) and diffusion matrix G=(0,S)T. It^o's formula therefore gives its generator Lψ=b⋅∇ψ+12(GGT):D2ψ=(M−1p)⋅∇qψ+(F(q)−γp)⋅∇pψ+γβ−1tr⁡(MDpp2ψ). Define Aψ=(M−1p)⋅∇qψ, Bψ=F(q)⋅∇pψ, and Oψ=−γp⋅∇pψ+γβ−1tr⁡(MDpp2ψ). The deterministic A and B subflows have these first-order generators by the chain rule; the O subflow has generator O by It^o's formula applied to the OU equation. Thus L=A+B+O under the displayed model convention.

[F2]

BAOAB underdamped Langevin splitting and Exact Ornstein–Uhlenbeck momentum transition and covariance: for s≥0 and a test function ψ, define the exact subflow transition operators by PsAψ(q,p)=ψ(q+sM−1p,p),PsBψ(q,p)=ψ(q,p+sF(q)), PsOψ(q,p)=E ψ ⁣(q,e−γsp+β−1(1−e−2γs) M1/2ξ),ξ∼Nd(0,Id). The A and B formulas are their deterministic subflow maps, and the O formula is the exact OU transition. Also write Psψ(z)=Ezψ(Qs,Ps) for the exact SDE transition. Operator products are in state-update order: the leftmost transition is applied first. Hence the BAOAB one-step operator is the palindromic composition Kh=Ph/2BPh/2APhOPh/2APh/2B.

[F3]

Multidimensional Ito formula for Brownian-driven processes: It^o's formula holds for C2 spatial test functions of the Brownian It^o process, with localized stochastic integrals; expectation and repeated integral identities require the moment bounds established below.

[F4]

Global strong well-posedness for Lipschitz Itô SDEs: the globally Lipschitz, linearly growing drift gives a global strong solution with a finite second path moment; higher polynomial moments from deterministic starting states are proved below.

[F5]

Fatou's lemma: the integral of a nonnegative pointwise limit is at most the limit inferior of the integrals.

[A1]

The Axiom of Choice: AC is inherited through [F3]–[F5]; no further choice enters the operator expansion.

Proof

Given: The assumptions above.

1.1A1F1F2F3givenalgebra

Choose an integer p≥0 and K<∞ such that ∣Djϕ(z)∣≤K(1+∣z∣p) for 0≤j≤6, and put Wr(z)=(1+∣z∣2)r/2. For a bounded Cb6 family choose p=0 and a common K. The explicit A and B maps are (q,p0)↦(q+sM−1p0,p0) and (q,p0)↦(q,p0+sF(q)), where F=−∇V. They have linear growth and uniformly bounded spatial derivatives of orders one through six for 0≤s≤1, because DjF is bounded for 1≤j≤6. The O map is affine with centered Gaussian noise of uniformly bounded moments of every fixed order by [F2]. Differentiating these explicit maps and taking Gaussian expectations therefore shows that each PsC, C∈{A,B,O}, preserves bounds ∣Djψ(z)∣≤K0Wr(z) through any stated order k≤6, uniformly for s≤1. Each generator maps such bounds through order k to bounds through order k−2 with weight Wr+1: its drift coefficients grow at most linearly, their positive-order derivatives are bounded, and its second-order coefficients are constant. Thus a composition with at most three generator applications, even with subflow semigroups between them, uses at most six test derivatives and is bounded by CWp+3(z).

2.1A1F1F3F4F5givenstep 1.1algebra

For any r≥0, direct differentiation of Wr and the linear-growth drift and constant diffusion give ∣LWr∣≤CrWr. Itô's formula [F3] stopped on a ball, the integral Gronwall estimate, and Fatou [F5] as the stopping radius increases give PuWr(z)≤CrWr(z) for 0≤u≤1. These are estimates from a fixed deterministic starting point; higher weights can be used to justify the localization limits. In particular, ∣L3ϕ(z)∣≤CWp+3(z) by the same coefficient and derivative count as in step 1.1, and ∣PuL3ϕ(z)∣≤CWp+3(z). For the bounded-test branch this is the cubic weight W3, using the same Lyapunov function with exponent 3/2, not a higher initial-moment assumption.

2.2F1F2F3givenstep 1.1algebra

Put (C1,…,C5)=(B,A,O,A,B), (c1,…,c5)=(1/2,1/2,1,1/2,1/2), and Ψ(u,z)=(Pc1uC1⋯Pc5uC5ϕ)(z). The subflow integral identity PsCψ−ψ=∫0sPvCCψ dv follows from the explicit A/B maps and Itô's formula for O. It and the weighted derivative bounds in step 1.1 justify differentiating this five-factor product three times, with right derivatives at zero. Explicitly, its third derivative is the finite ordered sum. [F1, F2, F3, given, step 1.1, algebra] ∂u3Ψ(u)=∑ν1+⋯+ν5=33!∏i=15ciνi∏i=15νi!(Pc1uC1C1ν1⋯Pc5uC5C5ν5ϕ). Every summand has total generator length three, so step 1.1 bounds it by CWp+3(z); no term with more than three generators is formed. The first derivative at zero is Lϕ. Grouping the second derivative at zero gives the pure squares A2+B2+O2 and each ordered mixed pair AB,BA,AO,OA,BO,OB once, hence ∂u2Ψ(0)=L2ϕ. Taylor's formula with its third-derivative integral remainder now gives Khϕ=ϕ+hLϕ+12h2L2ϕ+12∫0h(h−u)2∂u3Ψ(u) du, whose remainder is bounded by Ch3Wp+3(z). This is a second-order Taylor expansion of the product itself, rather than multiplication of five truncated polynomials with higher differential cross-products.

3.1A1F3F4F5givenstep 1.1step 2.1algebra

Applying the localized Itô integral identity successively to ϕ, Lϕ and L2ϕ in which the three test functions have respectively six, four and two continuous spatial derivatives, gives the following identity. [A1, F3, F4, F5, given, step 1.1, step 2.1, algebra] Phϕ=ϕ+hLϕ+12h2L2ϕ+12∫0h(h−u)2PuL3ϕ du. The polynomial bounds from steps 1.1 and 2.1 justify taking expectations and removing the stopping times in these three identities. Step 2.1 bounds the remainder by Ch3Wp+3(z).

4.1step 1.1step 2.1step 2.2step 3.1algebra∎

Subtracting steps 2.2 and 3.1 cancels all terms through order two and gives ∣Khϕ(z)−Phϕ(z)∣≤Ch3Wp+3(z)≤C′h3(1+∣z∣p+3). This proves the original polynomial-growth claim with some finite exponent. When the test family has a common Cb6 bound, p=0 in every preceding estimate, so the same argument proves the uniform cubic-growth bound Ch3(1+∣z∣3). Neither conclusion is an invariant-measure statement.

Depends on

Used by

Dependency tree · two levels

62 results within two dependency steps of this one, each drawn at its shortest distance from it. An arrow runs from a result to what uses it, so the chart reads left to right and ends at this result, which carries a heavier outline. Every node is a link to that result. Click elsewhere on the chart to enlarge it.

Sources