REVIEW 3 major objections 7 minor 31 references
A KAM theorem for the Hamiltonian with finite zero normal frequencies and its applications
T0 review · 3 major / 7 minor · reviewed 2026-08-14 · deepseek-v4-flash
Pith's one-line read A single constant quantity decides whether a Hamiltonian with finitely many zero normal frequencies has KAM tori.
desk verdict A serious attempt at a genuinely open KAM problem, but Theorem 1.1 as printed is vacuous because Assumption (A) fails on the zero modes; the intended fix is clear and the paper deserves a careful referee. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central object is the constant quantity $\delta_0=\sqrt{|\breve N^{z_0}(\xi)|_2^2+|\breve N^{\bar z_0}(\xi)|_2^2}$, computed from the coefficients of the $z_0$ and $\bar z_0$ linear terms in the final normal form; it is invariant under the iteration by construction and decides between torus existence and torus absence. The argument that carries the proof is a non-standard KAM step in which the five zero-mode coefficient classes $\hat R^{z_0}(0,\xi)$, $\hat R^{\bar z_0}(0,\xi)$, $\hat R^{z_0z_0}(0,\xi)$, $\hat R^{z_0\bar z_0}(0,\xi)$, $\hat R^{\bar z_0\bar z_0}(0,\xi)$ are promoted into the next normal form instead of being eliminated. The resulting homological equation splits into four types, and the zero-mode blocks are inverted by Kronecker products and column straightening, with new small-divisor conditions requiring non-vanishing determinants of finite matrices such as $i\langle k,\omega_m\rangle I_{3b^2}-B_{1m}(\xi)$. The normal form preservation is what gives $\delta_0$ a meaning at the limit of the iteration.
What would settle it
Work out the first Newton step for the minimal case of one zero normal frequency ($b=1$) and a perturbation whose only low-order term is a nonzero constant coefficient $\hat R^{z_0}(0,\xi)$. If the iteration produces $\delta_0>0$ and the flow estimate for $|z_0(1)|$ stays below $\varepsilon_m^{7/6}$ for initial data with $\|z^*(0)\|_{a,p}+\|\bar z^*(0)\|_{a,p}\le \varepsilon_m^{7/6}$, the no-torus conclusion fails; if the computed drift exceeds $\varepsilon_m^{7/6}$, the dichotomy is confirmed in the cleanest possible case.
Extended reading notes
Core claim
On its own terms the paper's discovery is Theorem 1.1: for a parameter-dependent Hamiltonian $N+R$ whose normal form is $\langle\omega(\xi),y\rangle+\langle\Omega_0(\xi)z_0,\bar z_0\rangle+\langle\Omega(\xi)z,\bar z\rangle$ with $\Omega_0\equiv 0$ on a finite block, and under assumptions of nondegeneracy, spectral asymptotics, regularity and smallness, a Newton-type KAM iteration conjugates $H$ to a normal form that keeps the zero-mode terms. Whether $\delta_0$ is zero or positive then decides the geometry: when $\delta_0=0$, the zero block drops out of the linear flow and the embedded torus survives with shifted frequency; when $\delta_0>0$, the zero-mode linear term acts as a persistent drift that forces every trajectory starting near the torus out of the shrinking domain $\Phi_{m-1}(\Xi_m\times\{\xi\})$, so no invariant torus exists there. The paper also establishes that in the NLS application the quantities $\breve N^{z_0}$ and $\breve N^{\bar z_0}$ vanish at every iteration, so the application lands on the existence side of the dichotomy.
Load-bearing premise
The theorem's load-bearing premise is the non-resonance assumption (A): if the index $l$ is allowed to be supported entirely on the zero-mode coordinates with $k=0$, then $\langle l,\Omega(\xi)\rangle\equiv 0$, which contradicts the required $\langle l,\Omega(\xi)\rangle\neq 0$; hence the theorem is only non-vacuous when (A) is read as restricted to the nonzero normal frequencies.
Editorial extensions
If this is right
- Zero normal frequencies need no longer be excluded from KAM theorems: any system satisfying the stated assumptions with a finite zero block has its torus problem settled by the constant $\delta_0$.
- In the $\delta_0=0$ case the persistence is quantitative: the torus embedding is $\varepsilon$-close to the identity, the frequency shift is $O(\varepsilon)$, and the good parameter set has measure $\mathrm{Meas}\,\Pi\,(1-O(\gamma))$.
- In the $\delta_0>0$ case the theorem gives a certified torus-free region $\Phi_{m-1}(\Xi_m\times\{\xi\})$ rather than merely failing to construct a torus, because the linear zero-mode term produces a flow that escapes the shrinking domain.
- The periodic nonlinear Schr\"odinger equation $iu_t-u_{xx}+|u|^2u=0$ with periodic boundary conditions and even symmetry has many quasi-periodic solutions, despite the $q_0$ zero mode, because the relevant Fourier coefficients vanish at every iteration.
- The normal form reached by the iteration contains only quadratic or higher terms besides the zero-mode block, so the dynamics near the survived torus is governed by the constants $\breve N^{z_0}$, $\breve N^{\bar z_0}$ and the quadratic zero-block matrix.
Reading between the lines
- Editorial inference: $\delta_0$ should be computable at first order as the projection of the perturbation onto the zero-mode linear terms; if that is true, one can decide torus persistence for concrete PDEs by a one-step calculation before running the full Newton scheme.
- Editorial inference: the same normal-form strategy may transfer to finite blocks with small nonzero normal frequencies or with eigenvalue limit points, since only finite-dimensionality of the exceptional block is used; the finite-limit-point shallow-water results could be interpreted as a special case of this mechanism.
- Editorial inference: in generic perturbations with a zero mode but no symmetry, $\delta_0$ should be nonzero, so torus persistence is the exceptional, symmetry-forced outcome; testing a family of perturbations that breaks the parity symmetry would locate the transition at $\delta_0=0$.
- Editorial inference: one could try to read off $\delta_0$ from the original perturbation as $\hat R^{z_0}(0,\xi)^2+\hat R^{\bar z_0}(0,\xi)^2$ plus higher-order corrections; the paper's verifications suggest the leading term is often the exact vanishing condition.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. This paper proposes a KAM theorem for infinite-dimensional Hamiltonian systems with finitely many zero normal frequencies. The main result (Theorem 1.1) states that, under nondegeneracy, spectral asymptotics, and regularity assumptions, the Hamiltonian can be conjugated to a normal form containing zero-mode terms, and the outcome is governed by the constant quantity δ0 = sqrt(|N^{z0}(ξ)|^2 + |N^{bar z0}(ξ)|^2): if δ0 = 0, a rotational torus persists for most parameters; if δ0 > 0, no torus exists in a certain shrinking domain. As an application, the authors show that the periodic NLS equation iu_t - u_xx + |u|^2u = 0 with an even symmetry condition possesses many quasi-periodic solutions despite having a zero normal frequency.
Significance. If the hypothesis issue identified below is repaired, the paper would address a recognized open problem in KAM theory: Kuksin's degenerate case of zero normal frequencies ([21] mentions that no preservation theorem is known in this case). The paper's new technical content includes the treatment of zero-mode terms via finite matrix non-resonance conditions, the derivation of the δ0 criterion, and a detailed verification of the structural assumptions for the NLS application. Strengths are the explicit decomposition of the homological equations into four types with solution estimates (Section 2), the iterative measure estimates (Section 4), and the careful structural verification in Section 7, where the zero-mode coefficients are shown to vanish at every step. There is no circularity: δ0 is read off from the limit normal form and is not used to construct the tori. However, as printed, Theorem 1.1 applies to no Hamiltonian with zero normal frequencies, so the advertised applications do not currently follow.
major comments (3)
- [Theorem 1.1, Assumption (A), eqs. (1.7)-(1.8)] Assumption (A) is not satisfiable when zero normal frequencies are present. Since Ω_{j_m}=0 for j_m∈J, taking k=0 and l=e_{j_m} (with |l|=1) gives ⟨l,Ω(ξ)⟩ ≡ 0 on Π, so (1.8) fails identically and the set in (1.7) is all of Π. Because Π is assumed to have positive measure, the assumption can never hold for any Hamiltonian with a zero normal frequency. Consequently Theorem 1.1 as stated is vacuous, and the NLS application in Section 7, which has Ω^0_0=0, cannot invoke it. The intended hypothesis is evidently to restrict (1.7)-(1.8) to l supported on the nonzero normal frequencies N+\J, while zero-mode resonances are controlled by the finite matrix conditions (2) and (4) in Section 2.2, as the solvability analysis in (2.23)-(2.31) and the small-divisor sets in (2.12) and Section 4 already do. This is a load-bearing logical gap rather than a typo, though the repair is local and the rest of the proof appears designed for it.
- [Lemma 2.1, proof] The proof of the first iterative step is entirely delegated to [23] with the sentence "These results can be seen clearly in [23]." This is the base step of the KAM scheme and is central to the theorem. Since [23] does not treat zero normal frequencies, and the present paper's novelty is precisely the handling of zero modes, the lemma should at least state which results from [23] are used and how they are adapted, especially the measure estimate (2.17) and the solution estimates (2.14)-(2.16) in the presence of the zero-mode terms. As written, the proof of the key lemma is not self-contained in a way that supports the intended extension.
- [Section 6, nonexistence branch] The proof that δ0 > 0 implies nonexistence of tori shows that any solution of H_m starting in Ξ_m leaves Ξ_m by time 1, using the estimates (6.4)-(6.5). This does establish that no invariant torus is contained in Φ_{m-1}(Ξ_m × {ξ}). However, the domain Ξ_m depends on m and shrinks as ε_m → 0, so the conclusion is only local in the shrinking neighborhoods. The statement in Theorem 1.1 is formally correct as written ("there is no torus in the domain Φ_{m-1}(Ξ_m × {ξ})"), but the authors should clarify in Section 6 whether a fixed-size neighborhood is intended, and if so, provide the additional argument needed to extend the nonexistence to that fixed neighborhood.
minor comments (7)
- [Affiliations] "Furan University" should read "Fudan University."
- [Section 7, Lemma 7.1] "If a curse I → l^{a,p}" should read "If a curve I → l^{a,p}."
- [Notation] The symbol ⋖ is used throughout without being defined; it should be introduced as "a ≤ c b for a constant c depending on n and τ" or replaced by explicit inequalities.
- [Theorem 1.1, conclusion] In the final part of Theorem 1.1, "a district Ξ_m" should be "a domain Ξ_m."
- [Section 7, eq. (7.9)] The normal form in (7.9) omits the zero-mode term Ω^0_0 z0 \bar z0 (with Ω^0_0=0) that appears in (7.12); this is likely intentional but should be stated explicitly to avoid confusion.
- [References] Reference [1] is incomplete: it lists only a title and an arXiv number, with no authors; the full citation should be supplied.
- [Section 2.2, eq. (2.50)] The exponent of K_1 in (2.50) appears as (10b^2+2)τ+10b^2 in the display but as (10b^2+2)τ+10b^2-1 in one place in the text; the exponent should be consistent.
Circularity Check
No significant circularity: the δ0 dichotomy is read off from the limit normal form rather than fitted, and the NLS application's δ0=0 verification is an independent computation.
full rationale
The paper's central dichotomy is not circular. Theorem 1.1 first constructs, via Newton iteration, a conjugated Hamiltonian whose normal form contains the zero-mode terms ⟨N^{z0}, z0⟩, ⟨N^{bar z0}, bar z0⟩ etc.; only then is δ0 = sqrt(|N^{z0}|^2 + |N^{bar z0}|^2) introduced as a case criterion. The existence branch is the standard statement that T^n×{0} is invariant once the linear zero-mode coefficients vanish (Section 6, equations (6.1)-(6.3)); the non-existence branch is an ODE estimate showing the zero-mode coordinates leave the shrinking domain when δ0>0 (Section 6, (6.4)-(6.15)). Neither branch uses δ0 to construct the torus, and δ0 is not fitted to data. The NLS conclusion ˘N^{z0}=˘N^{bar z0}=0 is proved by the parity/momentum cancellations in Lemmas 7.3-7.6, an independent calculation, not by invoking the theorem's conclusion. The paper's reliance on [23] and [28] is ordinary use of prior KAM machinery and measure estimates, not a self-citation chain, and [30] (same author) is cited only as background. One non-circular correctness problem must be flagged: as printed, Assumption (A) (Theorem 1.1, equations (1.7)-(1.8)) quantifies over all (k,l) with 1≤|l|≤2 even though the normal frequencies include Ω_{j_m}=0 by definition. Taking k=0 and l=e_{j_m} makes ⟨l,Ω(ξ)⟩≡0, so (1.8) fails identically and the resonance set in (1.7) is all of Π; Theorem 1.1 as stated therefore has no instances. The intended restriction to l supported on N_+\J is suggested by the paper's own remark 1.2 ('Conditions (1.7)-(1.9) are the same as those in [28] when we only consider non-zero normal frequencies') and by the actual small-divisor conditions (2.12) and Section 4, which use nonzero normal frequencies for scalar Melnikov conditions and finite matrix conditions for zero modes. This is a repairable logical gap, not circularity, so the circularity score remains 0.
Assumptions & free parameters
assumptions (5)
- domain assumption The tangential frequency map ω(ξ) is a Lipeomorphism onto its image, and the standard non-resonance conditions hold for the nonzero normal frequencies as intended by (1.7)-(1.8).
- domain assumption Normal frequencies have spectral asymptotics Ω_j = j^d + O(j^δ) with δ < d-1, plus a separation condition for d=1.
- domain assumption The perturbation vector field is real analytic and maps P_{a,p} into P_{a,bar p} with p - bar p ≤ δ < d-1 and small weighted norm ε ≤ cγ.
- standard math The standard KAM estimates of Kuksin-Pöschel [23] and Pöschel [28] remain valid for the diagonal part and for the modified normal form with preserved zero-mode terms.
- standard math The cubic NLS Hamiltonian admits a real-analytic Birkhoff normal form leaving a quartic resonant part and a remainder of order six, as stated in Proposition 7.2.
Cite this review
Pith. "Pith review of A KAM theorem for the Hamiltonian with finite zero normal frequencies and its applications." pith.science (2026). https://pith.science/paper/YIIKSHJ6
@misc{pith2026190811072,
author = {Pith},
title = {Pith review of: A KAM theorem for the Hamiltonian with finite zero normal frequencies and its applications},
year = {2026},
howpublished = {\url{https://pith.science/paper/YIIKSHJ6}},
note = {Machine review of arXiv:1908.11072}
}
read the original abstract
In this paper, we investigate the existence of KAM tori for an infinite dimensional Hamiltonian system with finite number of zero normal frequencies. By constructing a constant quantity we show that, for "most" frequencies in the sense of Lebesgue measure, either if the quantity is zero, there is a KAM tori or if the quantity is not zero, there is no KAM tori in some domain. As application, we show that the nonlinear Schr\"{o}dinger equation with a zero frequency possesses many quasi-periodic solutions.
Reference graph
Works this paper leans on
-
[23]
S. B. Kuksin and J. P¨ oschel. Invariant cantor manifold s of quasi-periodic oscillations for a nonlinear Schr¨ odin ger equation. Annals of Mathematics , 143(1):149–179, 1996
work page 1996
- [28]
-
[21]
S. B. Kuksin. Nearly integrable infinite-dimensional Hamiltonian syste ms. Springer-Verlag, 1993
work page 1993
-
[1]
Large KAM tori for quasi-linear perturbations of KdV
Large KAM tori for quasi-linear perturbations of KdV. arXiv:1908.08768, 2019
work page Pith review arXiv 1908
-
[2]
V. I. Arnold. Proof of a theorem by A.N. Kolmogorov on the i nvariance of quasi-periodic motions under small perturbations of the Hamiltonian. Russian Mathematical Surveys , 18(5):9–36, 1963
work page 1963
- [3]
- [4]
- [5]
Show all 31 references
-
[6]
Berti and L
M. Berti and L. Biasco. Branching of Cantor manifolds of e lliptic tori and applications to PDEs. Communications in mathematical physics , 305(3):741–796, 2011
2011
-
[7]
Berti, L
M. Berti, L. Biasco, and M. Procesi. KAM theory for the Ham iltonian derivative wave equation. Annales scientifiques de l’Ecole Normale Superieure. Societe Mathe matique de France , 46(2):301–373, 2013
2013
-
[8]
Bourgain
J. Bourgain. Construction of quasi-periodic solutions for Hamiltonian perturbations of linear equations and applications to nonlinear PDE. Internat. Math. Res. Notices , (11):475–497, 1994
1994
-
[9]
Bourgain
J. Bourgain. Construction of periodic solutions of nonl inear wave equations in higher dimension. Geometric Functional Analysis GAF A, 5(4):629–639, 1995
1995
-
[10]
Bourgain
J. Bourgain. On Melnikovs persistency problem. Mathematical Research Letters , 4(4):445–458, 1997
1997
-
[11]
Bourgain
J. Bourgain. Quasi-periodic solutions of Hamiltonian perturbations of 2D linear Schr¨ odinger equations. Annals of Mathematics , 148(2):363–439, 1998
1998
-
[12]
Bourgain
J. Bourgain. Green’s function estimates for lattice Sc hr¨ odinger operators and applications.(am-158). Annal of Mathmatics Studies , 158, 2004
2004
-
[13]
Craig and C
W. Craig and C. E. Wayne. Newton’s method and periodic so lutions of nonlinear wave equations. Communica- tions on Pure and Applied Mathematics , 46(11):1409–1498, 1993
1993
-
[14]
L. H. Eliasson. Perturbations of stable invariant tori for Hamiltonian systems. Annali Della Scuola Normale Superiore Di Pisa Classe Di Scienze , 15(1):115–147, 1988
1988
-
[15]
L. H. Eliasson, B. Gr´ ebert, and S. B. Kuksin. KAM for the nonlinear beam equation. Geom. Funct. Anal. , 26(6):1588–1715, 2016
2016
-
[16]
L. H. Eliasson and S. B. Kuksin. KAM for the nonlinear Sch r¨ odinger equation. Annals of mathematics , 172(1):371–435, 2010
2010
-
[17]
Feola and M
R. Feola and M. Procesi. Quasi-periodic solutions for f ully nonlinear forced reversible Schr¨ odinger equations. Journal of Differential Equations , 259(7):3389–3447, 2015
2015
-
[18]
Geng and J
J. Geng and J. You. KAM tori of Hamiltonian perturbation s of 1D linear beam equations. J. Math. Anal. Appl. , 277(1):104–121, 2003
2003
-
[19]
Kappeler and J
T. Kappeler and J. P¨ oschel. KdV & KAM. A Series of Modern Surveys in Mathmatics , 45, 2003
2003
-
[20]
S. B. Kuksin. Perturbation of conditionally periodic s olutions of infinite-dimensional Hamiltonian systems. Engl. Transl. in Math. USSR Izv. , 32(1):41–63, 1989
1989
-
[22]
S. B. Kuksin. Analysis of Hamiltonian PDEs . Oxford University Press, 2000
2000
-
[24]
Liu and X
J. Liu and X. Yuan. A KAM theorem for Hamiltonian partial differential equations with unbounded perturba- tions. Communications in mathematical physics , 307(3):629–673, 2011
2011
-
[25]
V. K. Melnikov. On some cases of conservation of almost p eriodic motions under a small change of Hamiltonian function. Soviet Math. Dokl. , 6(3):1592–1596, 1965
1965
-
[26]
V. K. Melnikov. A certain family of conditionaly period ic solutions of a Hamiltonian systems. Soviet Math. Dokl., 9(3):882–886, 1968
1968
-
[27]
P¨ oschel
J. P¨ oschel. On elliptic lower dimensional tori in Hami ltonian systems. Mathematische Zeitschrift , 202(4):559– 608, 1989
1989
-
[29]
C. E. Wayne. Periodic and quasi-periodic solutions of n onlinear wave equations via KAM theory. 127(3):479–528, 1990
1990
-
[30]
X. Yuan. KAM theorem with normal frequencies of finite li mit-points for some shallow water equations. arXiv preprint arXiv:1809.05671 , 2018
2018 arXiv
-
[31]
Zhang, M
J. Zhang, M. Gao, and X. Yuan. KAM tori for reversible par tial differential equations. Nonlinearity, 24(4):1189– 1228, 2011. School of Mathematical Sciences, Furan University, Shangh ai 200433, P. R. China E-mail address : 14110840003@fudan.edu.cn School of Mathematical Scienc...
2011
Reviewed August 14, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.