{"id":"0f0e57ca-9250-4514-9777-acc06ed1fd1d","arxiv_id":"2502.00898","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"For almost every direction on a translation surface, the invariant surface of the flat geodesic flow persists within a finite-codimension family of smooth Hamiltonian perturbations.","lead":"A new proof addresses the persistence of invariant surfaces of flat geodesic flows on translation surfaces under smooth perturbations, for almost every direction. It establishes finite-codimension stability, extending a form of KAM behavior to higher-genus pseudo-integrable systems.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The finite-codimension conclusion rests on the unproved surjectivity of the obstruction map dP at H0; the stated linearization uses false values M[u0]=Id and S[u0]=0, contradicting Lemma 5.1 (M[u0]=diag(I2,-I2), S[u0]=-I2), so the submersion claim is not established.","rationale":"I agree with the reader's identification: the weakest point is the submersion step. The reader's statement that Lemma 5.1 and Section 6 contradict each other is correct and locates the gap precisely. My attack sharpens the point: the wrong values M[u0]=Id and S[u0]=0 are not cosmetic. With the correct values, the linearized para-cohomological equation has the form v2 - Xξ v1 + ... = ... and -Xξ v2 + ... = ..., so the obstruction coefficients in the first block depend on a solution of the second block. Surjectivity then requires controlling a triangular map involving a Green operator; it is plausible and likely repairable, for example by independent perturbations h=a(x)ξ1 and h=b(x)ξ2, but it is not demonstrated. Since the theorem's conclusion of finite codimension is exactly the submersion claim, this is the most load-bearing concern. I do not see a reason to change the reader's CONDITIONAL verdict: the concern is a potentially repairable gap, not a demonstrated falsehood. The suggested test would settle it.","tokens_in":15726,"tokens_out":17940,"duration_ms":184512,"concrete_test":"Redo the linearization of (3) at (H0,u0) using M[u0]=diag(I2,-I2) and S[u0]=-I2, and compute the differential of P∘u explicitly. For perturbations h(x,ξ)=a(x)ξ1+b(x)ξ2, compute f=DHF(H0,u0)(h), solve the second block for v2 with obstruction coefficients di,2=-Di(f2), then compute di,1=-Di(f1-v2) up to the signs fixed by the calculation. Check whether the triangular map (a,b)↦(di,1,di,2) is onto R^m, using only the linear independence of {Di} on smooth functions on M and the fact that a and b can be chosen independently. If the image has positive codimension, the submersion claim fails; if it is onto, write the argument in the paper, since the current text omits the v2 term and therefore does not establish the needed surjectivity.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Theorem 1.1 the finite codimension of Hs(ξ) is obtained by setting P[u(H)]=0 and invoking the implicit function theorem. The load-bearing step is therefore the claim that the differential of H↦P[u(H)] at H0 is surjective onto the finite-dimensional obstruction space. This is the only place where 'finite codimension' enters. The proof of that claim in Section 6 is not a proof: it states the linearization as Xξ v + Σ di[χi] = -DHF(H0,u0)(h), 'since M[u0]=Id, S[u0]=0'. But Lemma 5.1 and the computation preceding it give M[u0]=diag(I2,-I2) and S[u0]=-I2. With those values the correct linearization is a coupled system whose first block contains the unknown v2 (the vertical variation of the surface) in addition to Xξ v1. Consequently the coefficients di,1 are determined only after solving the second block, via a Green operator for Xξ, and the assertion that one can realize any prescribed vector of coefficients 'as long as the obstructions are linearly independent' does not follow from linear independence alone. Surjectivity of the resulting map h↦(di,1,di,2) is exactly what must be checked, and it is not checked anywhere in the manuscript. Without it, P[u(H)]=0 need not be a finite-codimension submanifold and the main theorem is not established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims to prove (Theorem 1.1) that for almost every direction ξ on a translation surface (M,ω), the invariant two-dimensional surface Mξ of the flat geodesic flow in a fixed energy level is stable with finite codimension under smooth Hamiltonian perturbations: there is a finite-codimensional local subvariety Hs(ξ) of Hamiltonians close to H0 such that each H in Hs(ξ) admits an invariant surface M_H^ξ that is an H^t graph over Mξ and the restricted flow is H^t-conjugated to the translation flow. The codimension is claimed to grow linearly in the Sobolev regularity s, the genus of M, and the number of conical singularities. A corollary is stated for rational polygonal billiards. The proof combines the author's cohomological equation results [F97, F21] with the para-differential approach of Alazard-Shao [AlSh] and invokes a fixed-point argument for a para-cohomological equation with counter-terms.","tokens_in":16063,"tokens_out":10272,"duration_ms":97868,"significance":"If the proof were complete, this would be a significant advance: it would establish a persistence result for higher-genus invariant surfaces in pseudo-integrable systems where standard KAM iteration is obstructed by distributional obstructions of growing Sobolev order, and it would transfer the finite-codimension phenomenon for the linearized cohomological equation to a nonlinear finite-codimension stability statement. The manuscript is honest about relying on published theorems rather than introducing fitted parameters, and the overall strategy of para-linearizing the invariant-surface equation and solving a cohomological equation with counter-terms is coherent. However, the central submersion step connecting the linearized equation to the finite codimension of the subvariety is not demonstrated, and there is an internal contradiction in the linearization; the main theorem should not be accepted in the present form.","major_comments":[{"comment":"Section 6 claims that the linearization of the para-cohomological equation at H=H0 is Xξ v + Σ di[χi] = -D_H Fξ(H0,u0)(h), 'since M[u0]=Id, S[u0]=0'. This contradicts Lemma 5.1, where M[u0]=diag(I2,-I2) and S[u0]=-I2, and it also contradicts the statement in Section 3 that M[u] is close to diag(I2,-I2). With the values computed in the paper, the linearized system is coupled: in the variable w=(w1,w2) it contains a term S[u0]w2 in the first line together with Xξ w1, so the coefficients di,1 can be determined only after solving the second block for w2. The sentence 'It is clearly possible to find a variation h such that the values of the coefficients (di) is any given vector of coefficients' is therefore not justified by linear independence of the obstructions. Since the finite codimension of Hs(ξ) is obtained solely from the claim that the map h ↦ P[u(H)] has surjective differential at H0, this gap is load-bearing and the main theorem is not established.","section":"Section 6 (proof of Theorem 1.1)"},{"comment":"Even under the simplified linearization assumed in Section 6, no proof is given that the map h ↦ (di) is surjective onto the finite-dimensional obstruction space. The expression D_H Fξ(H0,u0)(h) is never computed, and no argument links an arbitrary linear combination of the invariant distributions χi to a Hamiltonian perturbation h. Linear independence of the χi gives injectivity of the coefficient-to-functional map, not surjectivity of the perturbation-to-coefficient map. A submersion theorem requires the latter, and the implicit function theorem cannot be invoked until this surjectivity is established.","section":"Section 6 (surjectivity of dP)"},{"comment":"The acknowledgments state that N. Tedesco pointed out a calculation mistake in Lemma 4.1, but the manuscript does not include the corrected statement. Lemma 4.1 is the starting point of the para-linearized equation used in Section 4, in Lemma 5.1, and again in the final identity for Fξ(H,u) in Section 6. Without a corrected Lemma 4.1, the derivation of the operator S[u] and the subsequent solution of the para-cohomological equation cannot be checked. This is not a presentation issue; it affects the core of the proof.","section":"Acknowledgments and Lemma 4.1"}],"minor_comments":[{"comment":"The sentence 'L[u]=0 if and only if the differential of the pull-back under u of the standard symplectic form on M×R^2 is a closed 2-form' is inaccurate: the pull-back of a symplectic form is always closed, and the Lagrangian condition is u^*Ω=0.","section":"Section 3 (before Lemma 3.3)"},{"comment":"In the solution of the first equation, the inequality 'Σ |ci,2| ≤ ... ∥(T^{-1}_{M[u]} f)_1 - TS[u] v̂2∥' should refer to ci,1, not ci,2; as written it concerns the wrong set of constants.","section":"Lemma 5.1 proof"},{"comment":"The notation switches between P[u(H)], P_i[u], and the operator P[u] introduced in Lemma 5.1; the text should define explicitly which finite-dimensional vector is being set to zero in the condition P[u(H)]=0.","section":"Sections 5 and 6"},{"comment":"There are numerous typos, including 'he genus' in Theorem 1.1, 'the the' in Theorem 1.1 and Corollary 1.2, and 'Inv ariant' in the title; these should be corrected in a revision.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The main theorem is not established in the current version: the submersion argument is a one-sentence assertion and contradicts the paper's own computation of S[u0] in Lemma 5.1. The acknowledgment of a mistake in Lemma 4.1 makes it impossible for a reader to verify the para-differential core. I would be willing to look at a revised version that corrects Lemma 4.1 and gives a complete proof of the surjectivity of the differential of P[u(H)]."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take. The paper claims a genuinely new result: finite-codimension stability of invariant higher-genus surfaces for flat geodesic flows on translation surfaces, extending KAM persistence beyond Liouville tori. The strategy—transfer the Alazard–Shao para-differential fixed-point scheme and use your cohomological equation theorem as the linear input—is credible. The paper is honest about the literature and credits prior work properly.\n\nThe problem is that the main theorem is not established by the written proof. The finite-codimension conclusion depends on one step: the differential of H -> P[u(H)] at H0 must be surjective onto the obstruction space. Section 6 attempts to prove this, but it states the linearization as Xξ v + Σ d_i [χ_i] = -D_H F(H0,u0)(h) 'since M[u0]=Id, S[u0]=0'. Earlier, Lemma 5.1 and the computation before it give M[u0]=diag(I2,-I2) and S[u0]=-I2. These are incompatible. With the correct values, the first block of the linearized system contains an extra S[u0] v2 term, coupling the two components. The subsequent claim that you can realize any prescribed vector of coefficients 'as long as the obstructions are linearly independent' does not follow from linear independence alone; you need a proof of surjectivity of the actual map, and that is absent. The acknowledgment that Lemma 4.1 contained a calculation mistake adds to the uncertainty about which formulas are correct.\n\nThis is a load-bearing gap, not a cosmetic typo. If the surjectivity can be proved, the theorem would be significant and likely correct. But the current manuscript is not ready. I would send it to a serious referee—the result and the program deserve that—but the referee should be told to focus on the linearization and the submersion step. For my own work, I would not cite it yet. The paper is coherent in its overall plan, but the internal contradiction makes it not coherent at the proof level.","headline":"The finite-codimension theorem is a real new claim, but the proof's linearization is internally inconsistent (Lemma 5.1 vs Section 6) and the required surjectivity of the obstruction map is unproved, so the main result is not established.","tokens_in":16544,"tokens_out":3091,"would_cite":false,"duration_ms":29004,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["37C75","37C83","35S50"],"pacs":[],"model":"deepseek-v4-flash","headline":"Almost every invariant surface survives smooth perturbations","keywords":["translation surfaces","rational polygonal billiards","invariant surfaces","finite codimension stability","para-differential calculus","cohomological equation","KAM theory","geodesic flow"],"falsifier":"Evaluate the matrix $S[u_0]$ directly from its defining formula in Lemma 4.1 for the flat Hamiltonian $H_0$ and the identity section $u_0$; substituting $A[u_0]$ and $Du_0$ settles whether the value is $-I_2$ or $0$. Then, for a translation surface of genus two with a single cone point, compute the linearized obstruction map at $(H_0,u_0)$ by varying $H$ through a basis of smooth Hamiltonian perturbations $h$; the claimed finite codimension requires that the resulting coefficients $d_i$ range over the full finite-dimensional obstruction space. Finding one invariant distribution not attained by any such variation disproves the surjectivity the proof needs.","tokens_in":15526,"feed_emoji":"📐","tokens_out":14542,"duration_ms":116175,"temperature":0.7,"pith_summary":"The paper establishes that, for almost every direction on a translation surface, the two-dimensional invariant surface of the flat geodesic flow is stable with finite codimension under smooth perturbations of the Hamiltonian. In concrete terms, there is a finite-codimension local subvariety of nearby Hamiltonians for which the perturbed flow still has an invariant surface, and that surface is a slightly regular graph over the original one with the restricted flow smoothly conjugate to the translation flow. This is a higher-genus analogue of KAM persistence, applied to pseudo-integrable systems such as rational polygonal billiards, whose phase spaces are foliated by invariant surfaces rather than tori. As a corollary, for rational polygonal billiards the invariant surfaces in typical directions survive perturbations within a finite-codimension family of metrics, and the codimension grows linearly with the differentiability class and with the genus and number of conical singularities.","feed_headline":"Almost every invariant surface survives smooth perturbations","feed_subtitle":"A finite-codimension family of Hamiltonians has a nearby invariant surface whose flow is a translation.","key_machinery":"Para-differential operators, a frequency-localized form of pseudodifferential calculus used to linearize nonlinear expressions, are applied to the invariant-surface equation $F_\\xi(H,u)=0$. The central object is the para-differential cohomological equation with counter-terms,\n$$T_{M[u]}\\begin{pmatrix}0&S[u]\\\\0&0\\end{pmatrix}T_{M[u]}^{-1}v - T_{M[u]}X_\\xi(T_{M[u]}^{-1}v) + \\sum_i c_i[\\chi_i] = f,$$\nwhere $M[u]$ is a matrix built from the differential of $u$, $S[u]$ is a $2\\times2$ matrix constructed from the Lagrangian measurement $L[u]=(Du)^t J Du$, and $\\chi_i$ are a basis of invariant distributions for the translation vector field $X_\\xi$. The invariant distributions are the obstructions: the linearized equation is solvable only when the data annihilate them, and the counter-terms $c_i$ are chosen to cancel those obstructions. The proof shows that for Hamiltonians for which the counter-terms vanish, the nonlinear equation is solved by a fixed-point argument, yielding the finite-codimension submanifold.","core_discovery":"Main theorem: let $(M,\\omega)$ be any translation surface. For almost all directions $\\xi \\in P^1(\\mathbb{R}^2)$, the invariant two-dimensional surface $M_\\xi$ in a fixed energy level of the flat geodesic flow is stable with finite codimension in the following sense. There exists $s_0>0$ such that for every $s>s_0$ there is a finite-codimension local subvariety $H_s(\\xi)$ of the space of Hamiltonians near the flat Hamiltonian $H_0$ in the Sobolev space $H^s(M)$; every $H \\in H_s(\\xi)$ admits an invariant surface $M_H^\\xi$ that is an $H^t$ graph over $M_\\xi$ for $t<s-s_0$, and the Hamiltonian flow of $H$ on $M_H^\\xi$ is $H^t$-conjugate to the translation flow on $M_\\xi$. The codimension $h_s$ grows linearly in $s$, in the genus of $M$, and in the number of conical singularities of the translation structure. The proof para-linearizes the invariant-surface equation and reduces it to a cohomological equation with counter-terms; the obstructions are the invariant distributions of the translation flow, and requiring all obstructions to vanish carves out the finite-codimension submanifold of admissible Hamiltonians. The same statement holds for billiards in rational polygons after unfolding to a translation surface.","pith_inferences":["The author suggests in the introduction that the para-differential counter-term method may lead to a new proof of the linearization theorem for interval exchange transformations under weaker Diophantine conditions than the Roth-type condition; carrying out that program would be a natural next step.","One could test the conclusion numerically on a low-genus translation surface by checking that small Hamiltonian perturbations inside the finite-codimension family leave a nearby invariant graph, while generic perturbations destroy it.","The finite-codimension formulation raises a quantitative question the paper does not address: whether the codimension equals the dimension of the invariant-distribution obstruction space, so that the stability is optimal."],"forward_implications":["For rational polygonal billiards, almost every direction has an invariant surface that survives perturbations within a finite-codimension family of metrics, so there are finite-codimension families of non-ergodic perturbations of these pseudo-integrable systems.","The codimension of the stable family grows linearly with the Sobolev regularity $s$, which means that in the $C^\\infty$ limit the allowed perturbations are infinite codimension, consistent with the known infinite codimension of the range of the cohomological equation in the smooth category.","The result gives a positive answer to the conjecture that typical translation flows are stable with finite codimension under smooth perturbations, at least for almost all directions.","The para-differential fixed-point approach succeeds where a Nash-Moser iteration would fail because the smoothing operators make the distributional obstructions of growing Sobolev order blow up."],"supporting_citations":[{"why":"Supplies the para-differential fixed-point method for solving the nonlinear stability problem by standard fixed point theorems.","marker":"[AlSh]"},{"why":"Proves the finite-codimension range of the cohomological equation for translation flows, providing the invariant-distribution obstructions used throughout.","marker":"[F97]"},{"why":"Establishes the weighted Sobolev space scale and loss-of-derivative estimates for solutions of the cohomological equation, used in the a priori bounds.","marker":"[F21]"},{"why":"KAM theory without action-angle variables; the source of the Lagrangian measurement identity in Lemma 3.3 and of the linearization formulas in Lemma 4.1.","marker":"[LGJV05]"},{"why":"Gives improved regularity estimates for the cohomological equation on Roth-type interval exchange transformations, cited as the low-loss alternative to the harmonic analysis method.","marker":"[MMY05]"},{"why":"Shows that the Roth-type condition holds for almost every direction on every translation surface, justifying the 'almost all' statement in the theorem.","marker":"[CE15]"}],"fun_headline_variants":["Most invariant surfaces survive Hamiltonian perturbations","Finite exceptions block surface instability in translation flows","Stability of invariant surfaces up to finite codimension","Almost all invariant surfaces persist: finite obstructions only","Translation surfaces have finite-codimension stable surfaces"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof depends on the linearized obstruction map from Hamiltonian variations to the coefficients $d_i$ of the invariant-distribution obstructions being surjective at the unperturbed Hamiltonian, so that the zero set of the obstructions has the claimed finite codimension; the paper asserts this is 'clearly possible' without supplying the derivation, and the linearization is internally inconsistent about the value of the matrix $S[u_0]$ (Lemma 5.1 computes it as $-I_2$, while Section 6 states it is $0$).","fun_headline_variants_meta":{"raw":{"variants":["Most invariant surfaces survive Hamiltonian perturbations","Finite exceptions block surface instability in translation flows","Stability of invariant surfaces up to finite codimension","Almost all invariant surfaces persist: finite obstructions only","Translation surfaces have finite-codimension stable surfaces"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000228,"raw_usage":{"total_tokens":1446,"prompt_tokens":889,"completion_tokens":557,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":505,"completion_tokens_details":{"reasoning_tokens":487}},"tokens_in":505,"tokens_out":557,"duration_ms":6107,"temperature":1.0,"reasoning_tokens":487,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T17:20:06.794245+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the matrix $S[u_0]$ directly from its defining formula in Lemma 4.1 for the flat Hamiltonian $H_0$ and the identity section $u_0$; substituting $A[u_0]$ and $Du_0$ settles whether the value is $-I_2$ or $0$. Then, for a translation surface of genus two with a single cone point, compute the linearized obstruction map at $(H_0,u_0)$ by varying $H$ through a basis of smooth Hamiltonian perturbations $h$; the claimed finite codimension requires that the resulting coefficients $d_i$ range over the full finite-dimensional obstruction space. Finding one invariant distribution not attained by any such variation disproves the surjectivity the proof needs.","supporting_citations":[],"review_version":1}