A Birkhoff Normal Form Theorem for Partial Differential Equations on torus

Duohui Xiang; Jianjun Liu

REVIEW 1 major objections 4 minor 26 references

A Birkhoff Normal Form Theorem for Partial Differential Equations on torus

T0 review · 1 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A complete Birkhoff normal form, up to arbitrary finite order, is established for Hamiltonian partial differential equations on the torus.

desk verdict The stress-test's central objection to Lemma 3.6 does not hold up once you read the norm definition; the paper's core theorem and applications look sound. read the letter →

arxiv 2411.13312 v1 pith:WMAXIYDN submitted 2024-11-20 math.AP math.DS

Jianjun Liu , Duohui Xiang This is my paper · ORCID

classification math.APmath.DS MSC 37K5535B3535L0535Q55

keywords BirkhoffnormalformHamiltonianPDEslongtimestabilitynonlinearwaveequationSchrödingertorusSobolevspacesnon-resonantconditions

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper establishes an abstract Birkhoff normal form theorem for Hamiltonian partial differential equations on the torus: under a non-resonant condition on the frequencies, every Hamiltonian H0 + P can be transformed by a canonical change of variables into H0 + Z + R, where Z is a resonant polynomial of order at most r + 2 and R has order at least r + 3, with explicit constants in all estimates. This makes the normal form complete up to arbitrary finite order, without the truncation in Fourier space that earlier torus results required. The theorem is applied to the one-dimensional nonlinear wave equation and to high-dimensional nonlinear Schrödinger equations, giving long-time stability in Sobolev spaces with considerably smaller index s than previously known, and, when the nonlinearity is analytic, sub-exponentially long stability times.

What carries the argument

The key object is the s,N-norm (2.7) for homogeneous polynomials, defined by summing over momentum b with weight ⟨b⟩^{N-s} the supremum of |(j*_1...j*_l)^s ̃f_j|. Its main property is that the Poisson bracket is bounded by a constant factor times the product of the two norms with the same s and N (Lemma 3.3, inequality (3.8)), so the norm is preserved through the Lie transformations that implement the normal form. The second component is the non-resonant condition (2.18), which bounds the small divisors by γ times the rational weight (j*_1/(⟨M(j)⟩ j*_2 ... j*_l))^τ; this weight, together with the norm, lets every non-resonant monomial of order at most r + 2 be eliminated without introducing a truncation parameter N, while keeping all constants explicit in r.

What would settle it

A concrete observation that would settle the claim: compute, for a fixed frequency family such as ω_a = √(a² + m) with m ∈ [1,2], the quantity |δ1ω_a1 + ⋯ + δlω_al| for some l ≤ r and a tuple of modes j outside J_l; if for a set of m of positive measure this quantity falls below γ(j*_1/(⟨M(j)⟩ j*_2 ⋯ j*_l))^τ with the paper's choice of γ and τ, the homological equation estimate in Lemma 3.6 loses its bound and Theorem 2.1 fails for those parameters. More directly, one can numerically integrate the measure estimate in Lemma 4.2 for l = 3 over m ∈ [1,2] and compare the observed length of the resonant set with the claimed $e^{{12·3}}$$γ^{{1/(7·9)}}$.

Watch

Extended reading notes

Core claim

The paper's central claim is Theorem 2.1: if the Hamiltonian perturbation P satisfies the growth estimate (2.11) in the s,N-norm introduced in (2.7), and the frequencies satisfy the non-resonant condition (2.18) up to order r + 2, then there is a canonical transformation φ with the estimate (2.15) such that H ◦ φ = H0 + Z + R. Here Z is a (γ,τ)-normal form polynomial of order at most r + 2, R has order at least r + 3, and both satisfy the quantitative bounds (2.16) and (2.17). The regularity requirement is s ≥ s0 + rτ + (d+1)/2, which improves on the earlier O($r^{2}$ α) bound from the truncation approach by removing the dependence on a large Fourier truncation parameter. As corollaries, the normal form implies that the Sobolev norm of a small solution stays bounded for times of order $ε^{{-(r+1)}}$, and that the super actions Ja change by at most $ε^{{3-ν}}$ over that time.

Load-bearing premise

The load-bearing premise is the non-resonant condition (2.18): for every Fourier mode and every order up to r + 2, the frequency combination either vanishes exactly (the resonant set J_l) or is bounded below by γ times the stated rational weight; in the applications this is only proved to hold outside a small exceptional set of parameters m or V.

Editorial extensions

If this is right

For the 1D nonlinear wave equation utt − uxx + mu + f(x,u) = 0 with m in a set of full measure, solutions with initial Sobolev size ε stay bounded by 2ε for |t| ≤ c1 ε^{-r} in H^s with s = O(r^4), improving the earlier s = O(r^8).
For the d-dimensional nonlinear Schrödinger equation i∂tu = −Δu + V*u + g(x,u,̅) with d ≥ 2 and generic potentials V, the same stability holds with s = O(r) instead of O(r^3).
With analytic nonlinearity, choosing r depending on ε gives sub-exponential stability times e^{(1/2)(ln 1/ε)^{1+λ}} for the wave equation and ρ^{(logρ 1/ε)^2/(24τ logρ logρ 1/ε)} for NLS, with the latter improving to ρ^{(1/46)(logρ 1/ε)^2} when the nonlinearity is independent of x.
The super actions (mode energies) are almost conserved: ⟨a⟩^{2s}|Ja(t) − Ja(0)| ≤ c2 ε^{3−ν} for times of order ε^{-(r+ν)}.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The same abstract theorem should apply to other Hamiltonian PDEs on the torus whose nonlinearity is smooth in x and analytic in u, provided a non-resonant condition of the form (2.18) can be verified with the rational weight; the weight is tailored to the convolution structure of the torus and may be more restrictive than necessary for equations with a different geometry.
The explicit dependence of all constants on r opens the possibility of optimizing the iterative step count to push the stability time beyond sub-exponential, toward exponential, for nonlinearities with stronger decay in Fourier space.
The norm (2.7) compresses the information of high Sobolev indices into a single weight, so the framework could be tested numerically on finite truncations: the predicted estimates (2.15)–(2.17) are quantitative enough to be checked by computing the normal form coefficients for small r.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The paper is a solid contribution to the Birkhoff normal form literature. The abstract theorem — a complete normal form up to arbitrary finite order on the torus, with explicit constants — is genuinely new, and the applications lower the Sobolev thresholds for 1D NLW and high-dimensional NLS. The new s,N norm, which is preserved under Poisson brackets, is the key technical idea and it works.

The stress-test's claim that Lemma 3.6 is algebraically false does not survive reading the norm. The norm in (2.7) includes ⟨M(j)⟩ in the denominator: the weight is (j*_1/(⟨M(j)⟩j*_2...j*_l))^s, not (j*_1...j*_l)^s. Under that definition, the small-divisor denominator cancels exactly when shifting from s to s+τ. The concrete example j=(a,-a,1) gives a constant factor 4^τ and then ⟨1⟩^{-τ} brings it down, so the estimate holds with no additional a-dependence. The stress-test's calculation omits the ⟨M(j)⟩ term in the norm.

The reader's concern about Lemma 4.2 is also a sign error. For γ<1 and l≥3, the exponents satisfy 1/l > 1/(7l) > 1/(7l^2), so γ^{1/l} < γ^{1/(7l)} < γ^{1/(7l^2)}. Bounding the first two terms by γ^{1/(7l^2)} is valid; the inequality is in the correct direction.

The measure-theoretic parts are largely imported from Bambusi with added detail, and the high-dimensional NLS measure argument on the infinite product space needs a careful check of the finite-dimensional approximation. But these are standard concerns in this area, not fatal flaws. The analyticity assumptions for sub-exponential times are clearly stated.

Overall: the central theorem is proved carefully, the constants are tracked, and the applications deliver the promised improvements. This deserves serious peer review. I would cite it if I worked on long-time stability of Hamiltonian PDEs.

Referee Report

1 major / 4 minor

Summary. The paper proves an abstract Birkhoff normal form theorem (Theorem 2.1) for Hamiltonian PDEs on the torus: under the non-resonance condition (2.18), the Hamiltonian H = H0 + P is conjugate to H0 + Z + R, where Z is a (γ,τ)-normal form of order at most r+2 and R has order at least r+3, with explicit estimates (2.15)-(2.17). Corollary 2.1 derives polynomial-in-time stability of Sobolev norms and of super-actions. The abstract result is then applied to the one-dimensional nonlinear wave equation and to the high-dimensional nonlinear Schrödinger equation, yielding improved Sobolev-index requirements and, by taking r depending on the initial datum size, sub-exponentially long stability times.

Significance. If the results stand, the paper gives a genuinely self-contained abstract normal-form theorem with concrete constants, and it improves the known regularity requirements for the two applications: s = O(r^4) for 1D NW instead of O(r^8), and s = O(r) for high-dimensional NLS instead of O(r^3). The use of the s,N-norm (2.7) is well adapted to the Poisson bracket, and the constants in the iterative construction are tracked explicitly. I specifically checked the homological-equation estimate in Lemma 3.6 that could appear problematic: it is correct, because the factor ⟨M(j)⟩^{N−s} in the norm makes the shift s → s+τ exactly cancel the small-divisor denominator in (2.18). Thus the main abstract theorem is not undermined by that concern.

major comments (1)

[§4.2, Lemma 4.3] The proof of Lemma 4.3 contains an invalid estimate in the smooth (non-analytic) case: it bounds |Pn|_{0,N} = ∑_{b∈Z^d} ⟨b⟩^N |\hat G_n(−b)| by using the decay |\hat G_n(b)| ≤ C̃_N (2R′)^{-n} ⟨b⟩^{-N−2} and then treats ∑_{b∈Z^d} ⟨b⟩^{-N−2} as bounded by 3. This is false for d ≥ 2 when N is small; for example when N=0 the series diverges for d=2. The lemma as stated, 'for any N ≥ 0', is therefore not proved and is false as written. The fix is local: either require N > d−2 (which is automatically satisfied in Theorem 4.2, where N = s+1 with s ≥ rτ+(d+1)/2), or replace the decay exponent by ⟨b⟩^{-N−d−1} so that the sum converges for all N ≥ 0. This correction does not affect the abstract Birkhoff theorem or the nonlinear wave application, but it is a load-bearing step in the proof of the NLS application and must be repaired.

minor comments (4)

[§1 and §2, formula (2.12)] The constant ρ appears to be 2^{d+4}3^d, as used in Lemma 3.1 and in the proof of Theorem 4.1 where ρ=96 for d=1; the displayed '2 d+43d' in the text is a typographical corruption and should be corrected.
[§4.2, Lemma 4.3] In the analytic case of Lemma 4.3, the claimed estimate C_N = C′((N+2)/(eμ))^{N+2} is plausible, but the proof should state which norm of the x-analytic function is used to justify the exponential decay of the Fourier coefficients uniformly in the Taylor degree n.
[Throughout] There are several typographical errors that should be corrected in a revision: 'miantained' and 'ineqaility' in Section 1, 'Moverover' in Lemma 3.6, 'Schr¨odinger' versus 'Schr"odinger', and inconsistent uses of 'tours' for 'torus'.
[§4.2, Lemma 4.4] The measure estimates in Lemma 4.4 are written for an infinite product space of parameters {v_a}_{a∈Z^d}; the proof implicitly uses finite-dimensional marginal measure. The authors should state this convention explicitly, since it is essential for the meaning of |V′|.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the abstract theorem is proved from explicit non-resonance hypotheses, and the self-citations are contextual rather than load-bearing.

full rationale

The derivation chain is conditional and self-contained: Theorem 2.1 assumes the quantitative non-resonance condition (2.18), defines the (gamma,tau)-normal form by (2.9), solves the homological equation in Lemma 3.6, and iterates via Lemma 3.7. No parameter is fitted to the stability conclusion; the constants in (2.15)-(2.17) are computed from the assumptions in (2.11) and (2.13). The applications verify (2.18) by measure estimates (Lemmas 4.2 and 4.4), so the exceptional sets are inputs rather than outputs. Self-citations appear: [Liu22] for the coefficient structure and [CLSY18, CLY16] as contextual references, but the norm is defined and its key Poisson-bracket property is proved in Lemma 3.3, and the non-resonance verification is carried out in the text, so these citations are not load-bearing. The skeptical concern about Lemma 3.6 - that the s -> s+tau weight shift may not absorb the small-divisor denominator - is a mathematical correctness objection, not a circularity objection: the estimate is asserted and proved (or not) independently of the theorem's conclusion. Likewise the reliance on Bambusi's lemmas is imported external support, not a self-citation chain. Hence no reduction of the claimed results to their own inputs is present; the appropriate score is at the bottom of the scale.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The abstract theorem introduces no fitted parameters; its constants are explicit. The applications introduce hand-chosen parameters τ, γ, r to satisfy the measure estimates, and rely on standard Fourier analysis, imported measure lemmas, and the non-resonance hypothesis.

free parameters (3)

τ (non-resonance exponent) = τ ≥ 63(r+2)^3 for NLW; τ ≥ 15(m+2d+1) for NLS
Hand-chosen lower bound in the measure estimates Lemmas 4.2 and 4.4; the stability indices and times depend on it.
γ (small divisor constant) = sufficiently small, e.g., γ=e^{-91(ln ε^{-1})^{3λ}} in Corollary 4.1
Smallness parameter in the non-resonance condition; chosen to make the exceptional parameter set small.
r (normal form order) = r ≈ (ln ε^{-1})^λ in Corollaries 4.1-4.3
Integer order chosen as a function of the initial size to obtain sub-exponential stability.

assumptions (4)

standard math Fourier analysis and Sobolev spaces on the torus
Background used throughout.
standard math Lemma 5 (determinant estimate) and Lemma 6 (measure estimate) imported from [Bam03]/[Bam99b] without proof
Used in Lemma 7 and Lemma 8 for the measure of resonant parameter sets; the paper quotes them verbatim.
domain assumption Assumption (2.11): the perturbation P has finite s0,N norm; the nonlinearities are smooth/analytic in x and analytic in u
Needed to apply Theorem 2.1; verified in Lemma 4.1 and 4.3 for the wave and Schrödinger examples.
domain assumption Non-resonant condition (2.18) holds for frequencies up to order r+2
This is the hypothesis of the abstract theorem; in applications it is shown to hold outside a small parameter set.

how reviews work

0 comments

Cite this review

Pith. "Pith review of A Birkhoff Normal Form Theorem for Partial Differential Equations on torus." pith.science (2026). https://pith.science/paper/WMAXIYDN

@misc{pith2026241113312,
  author       = {Pith},
  title        = {Pith review of: A Birkhoff Normal Form Theorem for Partial Differential Equations on torus},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WMAXIYDN}},
  note         = {Machine review of arXiv:2411.13312}
}

abstract

We prove an abstract Birkhoff normal form theorem for Hamiltonian partial differential equations on torus. The normal form is complete up to arbitrary finite order. The proof is based on a valid non-resonant condition and a suitable norm of Hamiltonian function. Then as two examples, we apply this theorem to nonlinear wave equation in one dimension and nonlinear Schr\"{o}dinger equation in high dimension. Consequently, the polynomially long time stability is proved in Sobolev spaces $H^s$ with the index $s$ being much smaller than before. Further, by taking the iterative steps depending on the size of initial datum, we prove sub-exponentially long time stability for these two equations.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 23 canonical work pages

[1]

D. Bambusi. Nekhoroshev theorem for small amplitude solutions in nonlinear S chr\" o dinger equations. Math. Z. , 230(2):345--387, 1999

work page 1999
[2]

D. Bambusi. On long time stability in H amiltonian perturbations of non-resonant linear PDE s. Nonlinearity , 12(4):823--850, 1999

work page 1999
[3]

D. Bambusi. Birkhoff normal form for some nonlinear PDE s. Comm. Math. Phys. , 234(2):253--285, 2003

2003
[4]

D. Bambusi. A B irkhoff normal form theorem for some semilinear PDE s. In Hamiltonian dynamical systems and applications , NATO Sci. Peace Secur. Ser. B Phys. Biophys., pages 213--247. Springer, Dordrecht, 2008

work page 2008
[5]

Berti and J.-M

M. Berti and J.-M. Delort. Almost global solutions of capillary-gravity water waves equations on the circle , volume 24 of Lecture Notes of the Unione Matematica Italiana . Springer, Cham; Unione Matematica Italiana, [Bologna], 2018

work page 2018
[6]

Bambusi, J.-M

D. Bambusi, J.-M. Delort, B. Gr\' e bert, and J. Szeftel. Almost global existence for H amiltonian semilinear K lein- G ordon equations with small C auchy data on Z oll manifolds. Comm. Pure Appl. Math. , 60(11):1665--1690, 2007

work page 2007
[7]

Bernier, E

J. Bernier, E. Faou, and B. Gr\' e bert. Long time behavior of the solutions of NLW on the d -dimensional torus. Forum Math. Sigma , 8:12--26, 2020

work page 2020
[8]

Bernier, E

J. Bernier, E. Faou, and B. Gr\' e bert. Rational N ormal F orms and S tability of S mall S olutions to N onlinear S chr\" o dinger E quations. Ann. PDE , 6(2):14, 2020

work page 2020

Show all 26 references

[9]

Bambusi and B

D. Bambusi and B. Gr\' e bert. Birkhoff normal form for partial differential equations with tame modulus. Duke Math. J. , 135(3):507--567, 2006

2006
[10]

Bernier and B

J. Bernier and B. Gr\' e bert. Long time dynamics for generalized K orteweg-de V ries and B enjamin- O no equations. Arch. Ration. Mech. Anal. , 241(3):1139--1241, 2021

2021
[11]

Biasco, J.-E

L. Biasco, J.-E. Massetti, and M. Procesi. An abstract B irkhoff normal form theorem and exponential type stability of the 1d NLS . Comm. Math. Phys. , 375(3):2089--2153, 2020

2020
[12]

Bourgain

J. Bourgain. Construction of approximative and almost periodic solutions of perturbed linear S chr\" o dinger and wave equations. Geom. Funct. Anal. , 6(2):201--230, 1996

1996
[13]

Bourgain

J. Bourgain. On diffusion in high-dimensional H amiltonian systems and PDE . J. Anal. Math. , 80:1--35, 2000

2000
[14]

H. Cong, J. Liu, Y. Shi, and X. Yuan. The stability of full dimensional KAM tori for nonlinear S chr\" o dinger equation. J. Differential Equations , 264(7):4504--4563, 2018

2018
[15]

H. Cong, C. Liu, and P. Wang. A N ekhoroshev type theorem for the nonlinear wave equation. J. Differential Equations , 269(4):3853--3889, 2020

2020
[16]

H. Cong, J. Liu, and X. Yuan. Stability of KAM tori for nonlinear S chr\" o dinger equation. Mem. Amer. Math. Soc. , 239(1134):vii+85, 2016

2016
[17]

H. Cong, L. Mi, and P. Wang. A N ekhoroshev type theorem for the derivative nonlinear S chr\" o dinger equation. J. Differential Equations , 268(9):5207--5256, 2020

2020
[18]

H. Cong, L. Mi, X. Wu, and Q. Zhang. Exponential stability estimate for the derivative nonlinear S chr\" o dinger equation. Nonlinearity , 35(5):2385--2423, 2022

2022
[19]

J.-M. Delort. A quasi-linear B irkhoff normal forms method. A pplication to the quasi-linear K lein- G ordon equation on S^1 . Ast\' e risque , (341):vi+113, 2012

2012
[20]

J.-M. Delort. Quasi-linear perturbations of H amiltonian K lein- G ordon equations on spheres. Mem. Amer. Math. Soc. , 234(1103):vi+80, 2015

2015
[21]

Faou and B

E. Faou and B. Gr\' e bert. A N ekhoroshev-type theorem for the nonlinear S chr\" o dinger equation on the torus. Anal. PDE , 6(6):1243--1262, 2013

2013
[22]

Gr\' e bert, R

B. Gr\' e bert, R. Imekraz, and \' E . Paturel. Normal forms for semilinear quantum harmonic oscillators. Comm. Math. Phys. , 291(3):763--798, 2009

2009
[23]

J. Liu. Birkhoff N ormal F orm for the D erivative N onlinear S chr\" o dinger E quation. Acta Math. Sin. (Engl. Ser.) , 38(1):249--262, 2022

2022
[24]

Yuan and J

X. Yuan and J. Zhang. Long time stability of H amiltonian partial differential equations. SIAM J. Math. Anal. , 46(5):3176--3222, 2014

2014
[25]

Q. Zhang. Long-time existence for semi-linear K lein- G ordon equations with quadratic potential. Comm. Partial Differential Equations , 35(4):630--668, 2010

2010
[26]

J. Zhang. Almost global solutions to H amiltonian derivative nonlinear S chr\" o dinger equations on the circle. J. Dynam. Differential Equations , 32(3):1401--1455, 2020

2020

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

D. Bambusi. Nekhoroshev theorem for small amplitude solutions in nonlinear S chr\" o dinger equations. Math. Z. , 230(2):345--387, 1999

work page 1999

[2] [2]

D. Bambusi. On long time stability in H amiltonian perturbations of non-resonant linear PDE s. Nonlinearity , 12(4):823--850, 1999

work page 1999

[3] [3]

D. Bambusi. Birkhoff normal form for some nonlinear PDE s. Comm. Math. Phys. , 234(2):253--285, 2003

2003

[4] [4]

D. Bambusi. A B irkhoff normal form theorem for some semilinear PDE s. In Hamiltonian dynamical systems and applications , NATO Sci. Peace Secur. Ser. B Phys. Biophys., pages 213--247. Springer, Dordrecht, 2008

work page 2008

[5] [5]

Berti and J.-M

M. Berti and J.-M. Delort. Almost global solutions of capillary-gravity water waves equations on the circle , volume 24 of Lecture Notes of the Unione Matematica Italiana . Springer, Cham; Unione Matematica Italiana, [Bologna], 2018

work page 2018

[6] [6]

Bambusi, J.-M

D. Bambusi, J.-M. Delort, B. Gr\' e bert, and J. Szeftel. Almost global existence for H amiltonian semilinear K lein- G ordon equations with small C auchy data on Z oll manifolds. Comm. Pure Appl. Math. , 60(11):1665--1690, 2007

work page 2007

[7] [7]

Bernier, E

J. Bernier, E. Faou, and B. Gr\' e bert. Long time behavior of the solutions of NLW on the d -dimensional torus. Forum Math. Sigma , 8:12--26, 2020

work page 2020

[8] [8]

Bernier, E

J. Bernier, E. Faou, and B. Gr\' e bert. Rational N ormal F orms and S tability of S mall S olutions to N onlinear S chr\" o dinger E quations. Ann. PDE , 6(2):14, 2020

work page 2020