Notes on tangent bicharacteristics and ill-posedness of the Cauchy problem

Enrico Bernardi; Tatsuo Nishitani

REVIEW 3 major objections 5 minor 7 references

Notes on tangent bicharacteristics and ill-posedness of the Cauchy problem

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read For the hyperbolic operator $P=-D_0^2+2x_1D_0D_n+D_1^2+x_1^2(x_1+\varepsilon x_r^\ell)D_n^2$, the Cauchy problem is not locally solvable in any Gevrey class of order $s>\max\{5,1+N/\ell\}$, hence not in $C^\infty$.

desk verdict Real advance on spectral transition, but the Rouché exponent and phase-convergence steps need patching before the proof is airtight. read the letter →

arxiv 2506.08481 v2 pith:VRGE7BWR submitted 2025-06-10 math.AP math.DS

Enrico Bernardi , Tatsuo Nishitani This is my paper

classification math.APmath.DS

keywords hyperbolicoperatorsdoublecharacteristicsHamiltonmapspectraltransitionGevreyclassesCauchyproblemill-posednesstangentbicharacteristics

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

The paper's target is a family of second-order hyperbolic operators with double characteristics whose Hamilton map changes spectral type along the double manifold, a situation not previously analyzed. For the model $P=-D_0^2+2x_1D_0D_n+D_1^2+x_1^2(x_1+\varepsilon x_r^\ell)D_n^2$ with $2\le r\le n-1$, $\ell\in\mathbb{N}$, $\varepsilon=\pm1$, the authors prove that the Cauchy problem is not locally solvable at the origin in the Gevrey class of order $s>\max\{5,1+N/\ell\}$ for some $N>1$ independent of $\ell$, so it is also ill-posed in $C^\infty$. Gevrey classes of order $s$ are smooth classes whose $k$-th derivatives grow like $(k!)^s$, so the statement gives a precise threshold: no local solution is guaranteed above that order. Because the model has analytic coefficients and zero subprincipal symbol, the obstruction lies in the principal symbol's geometry rather than in lower-order terms, and the proof builds a WKB-type approximate solution that contradicts any Gevrey a priori estimate.

What carries the argument

The workhorse is the one-dimensional ordinary differential equation $-Y''(x)+(x^3+\zeta x+\epsilon)Y(x)=0$ and its globally defined solutions that decay in prescribed sectors at infinity. The paper uses the connection coefficients $C_k(\zeta,\epsilon)$ (Stokes coefficients) between neighboring sector solutions; the key fact is a zero $\zeta_0$ of $C_0(\zeta,0)$ with $\mathrm{Im}\,\zeta_0<0$, which lets the authors select a complex frequency $\zeta(\rho^{-1},x_{n-1})$ near a fractional-power branch $\eta(\rho^{-1})$ by a standard perturbation argument. The approximate solution $U=\exp(-i\rho^5 x_n+\tfrac{i}{2}\zeta\rho(T-x_0))w(\rho^2x_1)$ then has the exponential decay needed for the final contradiction, while the nonzero moment $\int Y(x;\zeta_0,0)x^j\,dx$ supplies the lower bound that the pairing cannot be small.

What would settle it

Compute the connection coefficients for $-Y''+(x^3+\zeta x+\epsilon)Y=0$ from the asymptotics of its sector solutions and check whether $C_0(\zeta,0)$ really has a zero with negative imaginary part, whether the nearby branches separate with an exponent $\mu$ bounded by an integer $N$ independent of $\ell$, and whether some moment $\int Y(x;\zeta_0,0)x^j\,dx$ with $j\le2$ is nonzero. A direct check that one of these fails would invalidate the constructed contradiction.

Watch

Extended reading notes

Core claim

The central claim is that spectral transition of the Hamilton map, combined with a null bicharacteristic tangent to the doubly characteristic set, forces failure of local solvability at the origin. On the model $P$, for every $\ell$ there is a tangent bicharacteristic (written explicitly for $\ell\ne1$ and obtained by integrating the $\xi_r$ equation for $\ell=1$), and the authors show that no Gevrey-class solution can exist for orders $s>\max\{5,1+N/\ell\}$. The mechanism is a lower bound on a pairing between the approximate solution and any purported Gevrey solution that cannot be absorbed by the Gevrey estimates; the exponential rates are controlled by a zero of the connection coefficient $C_0(\zeta,0)$ with negative imaginary part. The result is the first ill-posedness theorem for double-characteristic hyperbolic operators whose Hamilton map undergoes a spectral transition.

Load-bearing premise

The proof leans on three imported facts about the special solutions of one ordinary differential equation—that a certain connection coefficient has a zero in the lower half-plane, that the nearby zeros split with an exponent leaving room for an integer $N$ independent of the tangency parameter $\ell$, and that a certain integral of the special solution is nonzero—and if any of those facts were false the contradiction would not go through.

Editorial extensions

If this is right

For the displayed operator $P$, the Cauchy problem is not locally solvable at the origin in any Gevrey class of order $s>\max\{5,1+N/\ell\}$, with $N>1$ independent of $\ell$; in particular it has no $C^\infty$ local solution.
The obstruction is stable under adding arbitrary lower-order terms $\sum_{j=0}^n b_jD_j$ to $P$, so the failure is a property of the principal part.
Every member of the family carries a null bicharacteristic tangent to the doubly characteristic set, and the Gevrey threshold depends on the tangency order through the $N/\ell$ term.
The result is the first ill-posedness statement for double-characteristic operators whose Hamilton map undergoes a spectral transition, extending earlier constant-spectral-type results.

Reading between the lines

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

Not a paper claim, but if $N$ were made explicit in terms of the fractional-power exponent $\mu$, the threshold $\max\{5,1+N/\ell\}$ would become a computable function of the tangency order $\ell$; for large $\ell$ it would approach $5$.
Not a paper claim, but the paper's comparison with the $k\ge2$ family suggests that the tangent bicharacteristic, rather than the spectral transition alone, is the decisive feature; that family is $C^\infty$ well-posed despite the same transition.
Not a paper claim, but the mechanism should transfer to nearby principal symbols with the same transition and tangency pattern, with the Gevrey threshold governed by the tangency order; this could be tested by perturbing $P$ by a full lower-order term beyond the $\sum b_jD_j$ form treated in the paper.
Located in the introduction, the paper corrects a predecessor reference: part of the Gevrey-5 argument in [1, Section 5.2] should be read as [6, Section 6.2]; the present proof relies on [6] directly, so the correction is bibliographic rather than a gap in the current argument.

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

Two things up front. The result is genuinely new: it is the first ill-posedness theorem for double-characteristic hyperbolic operators where the Hamilton map changes spectral type along the double manifold, and the Gevrey threshold depends on the degree l of the vanishing term. The second thing is that the proof as written has a repairable exponent error in the Rouché step, so the key estimate (2.7) is not actually established as written.

What the paper does well: it picks a concrete family P = -D_0^2 + 2x_1 D_0 D_n + D_1^2 + x_1^2(x_1 + ε x_r^l) D_n^2 and shows how Sibuya's global asymptotics plus a scaling argument can force a contradiction in Gevrey classes above the threshold. The reduction to the cubic oscillator (2.3) is clean, and the organization—deferring the symplectic geometry to [2]—is honest. The claimed threshold max{5, 1+N/l} is plausible and the dependence on l is the right kind of new information.

Now the soft spots, in order of importance.

First, the stress-test note is correct. In §2.1 the parameter s has been set to ρ^{-2} (via the -ζ^2ρ^{-2}/4 term). The lower bound on C0 on the contour |ζ-ζ1(s)| = c1|s|^ν is c2|s|^μ, so after s=ρ^{-2} it is c2ρ^{-2μ}. The perturbation is bounded by C'|α| ≤ Cρ^{-N}. Rouché therefore requires N>2μ, not N>μ. As written, (2.7) is not derived. This is a minor fix—choose N>max{2,2μ} and the theorem's existential N still works.

Second, after (2.7) the proof uses ν>1 to get ρ^{1-ν}→0 and pass to the limit in the phase factor. The text never proves ν>1, and the Puiseux exponent could be less than 1. This is also patchable: for a simple zero the actual root movement is O(|α|+ρ^{-2}) = O(ρ^{-N}+ρ^{-2}) by the implicit function theorem, and with N>2 that gives the needed o(ρ^{-1}). But the proof doesn't say that; as written the convergence claim is not justified.

Third, the reliance on the companion paper [2] for the geometric part is a real limitation of the arXiv version, and the moment condition from [6, Lemma 6.6] is imported. These are not red flags—[6] is published LNM and the moment condition is an external lemma—but a referee will need to check them.

Bottom line: the central claim is credible and not disproved. The proof is not watertight as written, but the gaps are local and mechanical. This paper deserves a serious referee; send it out with a request to fix the Rouché exponent and the phase-convergence argument.

Referee Report

3 major / 5 minor

Summary. The paper studies the second-order hyperbolic operator P = -D_0^2 + 2x_1 D_0 D_n + D_1^2 + x_1^2(x_1 + \epsilon x_r^l)D_n^2 and proves (Theorem 1.1) that, for every l \in \mathbb{N} and \epsilon = \pm 1, there exists N > 1 independent of l such that the Cauchy problem for P is not locally solvable at the origin in the Gevrey class of order s > \max\{5, 1+N/l\}; the authors conclude in particular that C^\infty local solvability also fails. The proof introduces a large-parameter rescaling P_\rho, constructs exact solutions U of P_\rho U = 0 using Sibuya/WKB solutions of the ODE -Y'' + (x^3 + \zeta x + \epsilon_0)Y = 0, and obtains a contradiction by comparing an upper bound of order e^{C\rho^\kappa} for the solution v_\rho to the Cauchy problem with a lower bound for the scalar product (U(0), D_0 v_\rho(0)) that decays faster than any Gevrey-5 function.

Significance. If the proof can be repaired, the result is significant: it appears to provide the first ill-posedness example for double-characteristic hyperbolic operators whose Hamilton map undergoes a spectral transition along the double manifold, and it links the Gevrey threshold to the geometry of a tangent bicharacteristic. The paper is concise and the overall strategy is coherent; the main computation is explicit, and the proof does not assume the target theorem. The authors are candid about relying on earlier technical results, especially the Sibuya estimates from [7] and the normal-form/Stokes results from [6] and the companion preprint [2]. However, as written the central Rouch\'e step contains an exponent error, and a later convergence claim requires a Puiseux exponent bound that is not proved. These issues are local and apparently repairable without changing the statement of Theorem 1.1, but they are load-bearing and must be resolved before the theorem is established.

major comments (3)

[§2.1, Eq. (2.7)] The Rouché step is not justified as written. The lower bound on C_0(\zeta, -\zeta^2 s/4) is stated as |C_0(\zeta, -\zeta^2 s/4)| \ge c_2 |s|^\mu on the circle |\zeta - \zeta_1(s)| = c_1 |s|^\nu. In the application one has s = \rho^{-2}, because the second argument of C_0 in (2.4) contains -\zeta^2 \rho^{-2}/4. Hence the contour is |\zeta - \eta(\rho^{-1})| = c_1 \rho^{-2\nu} and the lower bound is c_2 \rho^{-2\mu}, while (2.6) gives a perturbation of size O(\rho^{-N}). Rouché's theorem therefore requires N > 2\mu, not N > \mu as stated, and (2.7) should read |\zeta - \eta| \le c_1 \rho^{-2\nu}. Until this is corrected, the uniform decay (2.8) and the subsequent lower bound for |(U(0), D_0 v_\rho(0))| are not established. This is a local, repairable gap: choosing N > \max\{2, 2\mu\} preserves the theorem's requirement that N be independent of l, since \mu comes from C_0 alone.
[§2.2, after Eq. (2.7)] The passage 'From (2.7) we have |\zeta(\rho^{-1}, x_{n-1}) - \eta(\rho^{-1})|\rho \le c_1 \rho^{1-\nu} with \nu > 1' is used to pass to the limit in I_\rho via dominated convergence. However, the preceding Puiseux argument establishes only the existence of some \nu > 0, not that \nu > 1; the intermediate assertion 'necessarily \mu > \nu' is itself not justified and can fail when the distinguished zero is simple. Without a lower bound on \nu, the phase factor e^{i(\zeta-\eta)\rho T/2} need not converge to 1, so the claimed limit \rho^{2j} e^{-i\eta\rho T/2} I_\rho \to a_j \neq 0 is not established. Since the nonzero constant a_j is what forces the contradiction (2.19), this is load-bearing. If (2.7) is corrected as in the previous comment, the needed condition would become 2\nu > 1; the manuscript still needs to prove such a bound on the Puiseux separation exponent.
[§2.1–2.2, imported technical facts] The proof uses several very specific external results without restating their exact hypotheses: Lemma 2.1 (existence of \zeta_0 with Im \zeta_0 < 0, quoted from [6, Proposition 6.1]), the moment condition on \int Y(x_1;\zeta_0,0)x_1^j dx_1 ([6, Lemma 6.6]), and the Sibuya asymptotics from [7]. The moment condition is used to select j \in \{0,1,2\} and to assert a_j \neq 0; the selected j enters the lower bound (2.19) through \rho^{-2j}. The authors should either state these facts as explicit hypotheses or give precise published references with all conditions verified. This is not a circularity, but it is a completeness issue, especially because the geometric normal form is taken from the unpublished preprint [2].

minor comments (5)

[§2] The heading 'Proof of Therem' should read 'Proof of Theorem'.
[§2.1, Eq. (2.6)] In Eq. (2.6), the first argument of C_0 is written as \zeta - \alpha^3/3; from the definition w(y) = Y(y + \alpha/3; \zeta - \alpha^2/3, \ldots) it should be \zeta - \alpha^2/3.
[§2.1, Eq. (2.8)] In the sentence following (2.8), the condition 'Im(\zeta(\rho^{-1},x_{n-1}) - \alpha^3/3) \le -c_1' should read 'Im(\zeta - \alpha^2/3) \le -c_1'.
[References] Reference [6] lists the year as '2202'; it should be '2017'.
[§1] The phrase 'it is even more so for C^\infty' in Theorem 1.1 is awkward; the intended meaning is that local solvability also fails in C^\infty, and this should be stated directly.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the ill-posedness proof is self-contained modulo cited technical lemmas; self-citations to [2] and [6] supply auxiliary facts, not the target theorem.

full rationale

The derivation chain in Theorem 1.1 does not reduce to its own inputs. The Gevrey non-solvability is obtained by constructing explicit solutions U of P_rho U=0 through Sibuya asymptotics, locating a Stokes zero via Rouche's theorem, and then combining an energy identity with a Holmgren uniqueness argument to force a contradiction on the Fourier transform of the Cauchy datum theta. No parameter is fitted to a subset of data and then renamed a prediction; the Gevrey threshold arises from the exponent N in the construction and the scaling rho^{N/l}. The prior results invoked are technical prerequisites: [7] supplies Sibuya's Stokes theory, [6] supplies the zero zeta_0 and the nonzero moment condition, and [2] supplies the geometric normal form and spectral-transition classification. These are auxiliary facts with stated hypotheses that do not include the theorem being proved, and none of them is defined in terms of Gevrey solvability at the origin. Although [2] and [6] are written by the present authors, they are independent support in the sense of the review rules: they are parameter-free technical/geometric lemmas, not the ill-posedness conclusion. The apparent exponent concern in the Rouche step is a possible correctness gap, not a circularity: even if (2.7) as written requires a larger N, the proof does not assume the target ill-posedness. Therefore the paper has no significant circularity.

Assumptions & free parameters 1 free parameters · 7 assumptions · 0 invented entities

No new particles, forces, or dimensions are introduced. The main ledger items are the existential constant N and the imported Sibuya/Stokes lemmas. These are mathematical background rather than fitted empirical parameters.

free parameters (1)

N
The theorem's ill-posedness threshold s>max{5,1+N/l} depends on an integer N whose existence is proved (N>μ) but whose value is never computed. It is chosen by hand in the proof to satisfy a Rouche-type condition.

assumptions (7)

standard math Sibuya's global asymptotic theory: -Y''+(x^3+ζx+ε)Y=0 has entire solutions Y_k with Stokes relations and asymptotic expansion (see [7]).
Invoked in §2.1 to construct w via (2.4) and to justify (2.5) and (2.8).
standard math C0(ζ,0) has a zero ζ0 with Im ζ0 < 0 ([6, Proposition 6.1]).
Used in Lemma 2.1 to start the Puiseux/Rouche construction of ζ(ρ^{-1},x_{n-1}). This is a prior result from the same author's book.
standard math Weierstrass preparation and Puiseux expansion for C0(ζ,-ζ²s/4).
Used in §2.1 to obtain η(ρ^{-1}) and the separation bound |C0| ≥ c2|s|^μ with μ>ν.
standard math Rouche's theorem for perturbing C0(ζ,-ζ²s/4) by |α|≤Cρ^{-N}.
Used in §2.1 to solve (2.7) when N>μ.
standard math Holmgren uniqueness and local solvability lemma (Mizohata [4]) giving the a priori estimate (2.12).
Used in §2.2 to pass from assumed local solvability to uniform C^1 bounds.
standard math Paley-Wiener/Gevrey Fourier decay: |θhat(ξ)| ≤ C e^{-c|ξ|^{1/s}} characterizes γ^{(s)}.
Used at the end of §2.2 to convert the bound (2.19) into θ∈γ^{(5)}.
domain assumption The operator P is hyperbolic with double characteristics and analytic coefficients; Gevrey spaces γ^{(s)} are defined as in (2.11).
This sets the framework of Theorem 1.1; all constructions are local near |x_i|≪1.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Notes on tangent bicharacteristics and ill-posedness of the Cauchy problem." pith.science (2026). https://pith.science/paper/VRGE7BWR

@misc{pith2026250608481,
  author       = {Pith},
  title        = {Pith review of: Notes on tangent bicharacteristics and ill-posedness of the Cauchy problem},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VRGE7BWR}},
  note         = {Machine review of arXiv:2506.08481}
}

read the original abstract

We exhibit a family of second-order hyperbolic differential operators presenting spectral transition of the Hamilton map. As a consequence we prove that the Cauchy problem is not locally solvable at the origin in Gevrey classes of order greater than some fixed value. The main feature of these operators is that they may all have bicharacteristics tangent to the double manifold.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

7 extracted references · 6 canonical work pages

[2]

E.Bernardi, T.Nishitani, http://arxiv.org/abs/2505.21078: Geometric results for hyperbolic operators with spectral transition of the Hamilton map

arXiv
[6]

Springer, Cham, 2017

T.Nishitani,Cauchy problem for differential operators with double char- acteristics,2202, Lecture Notes in Mathematics. Springer, Cham, 2017

work page 2017
[7]

Y.Sibuya,Global Theory of a Second Order Linear Ordinary Differential Equation with a Polynomial Coefficient, North-Holland Mathematical Studies,18, Elsevier, 1975. 12

work page 1975
[1]

Analyse Math.,105 (2008) 197–240

E.Bernardi, T.Nishitani, On the Cauchy problem for noneffectively hy- perbolic operators, the Gevrey 5 well posedness, J. Analyse Math.,105 (2008) 197–240

work page 2008
[3]

Ja.,Correctness of the Cauchy problem for nonstrictly hyper- bolic operators

Ivri ˘i, V. Ja.,Correctness of the Cauchy problem for nonstrictly hyper- bolic operators. III. The energy integral,Trudy Moskov. Mat. Obˇ sˇ c.,34 (1977) 151-170

work page 1977
[4]

S.Mizohata,The theory of partial differential equations, Cambridge Uni- versity Press, 1973

work page 1973
[5]

T.Nishitani, Note on some noneffectively hyperbolic operators,Sci. Rep. College Gen. Ed. Osaka Univ.,32(1983) 9–17

work page 1983

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [2]

E.Bernardi, T.Nishitani, http://arxiv.org/abs/2505.21078: Geometric results for hyperbolic operators with spectral transition of the Hamilton map

arXiv

[2] [6]

Springer, Cham, 2017

T.Nishitani,Cauchy problem for differential operators with double char- acteristics,2202, Lecture Notes in Mathematics. Springer, Cham, 2017

work page 2017

[3] [7]

Y.Sibuya,Global Theory of a Second Order Linear Ordinary Differential Equation with a Polynomial Coefficient, North-Holland Mathematical Studies,18, Elsevier, 1975. 12

work page 1975

[4] [1]

Analyse Math.,105 (2008) 197–240

E.Bernardi, T.Nishitani, On the Cauchy problem for noneffectively hy- perbolic operators, the Gevrey 5 well posedness, J. Analyse Math.,105 (2008) 197–240

work page 2008

[5] [3]

Ja.,Correctness of the Cauchy problem for nonstrictly hyper- bolic operators

Ivri ˘i, V. Ja.,Correctness of the Cauchy problem for nonstrictly hyper- bolic operators. III. The energy integral,Trudy Moskov. Mat. Obˇ sˇ c.,34 (1977) 151-170

work page 1977

[6] [4]

S.Mizohata,The theory of partial differential equations, Cambridge Uni- versity Press, 1973

work page 1973

[7] [5]

T.Nishitani, Note on some noneffectively hyperbolic operators,Sci. Rep. College Gen. Ed. Osaka Univ.,32(1983) 9–17

work page 1983