Skip to main content
Log in

Perturbation at Blow-Up Time of Self-Similar Solutions for the Modified Korteweg–de Vries Equation

  • Published:
Archive for Rational Mechanics and Analysis Aims and scope Submit manuscript

Abstract

We prove a first stability result of self-similar blow-up for the modified Korteweg–de Vries equation on the line. More precisely, given a self-similar solution and a sufficiently small regular profile, there is a unique global solution which behaves at \(t=0\) as the sum of the self-similar solution and the smooth perturbation.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Similar content being viewed by others

Data Availability

There is no data associated to this manuscript.

References

  1. Banica, V., Vega, L.: Scattering for 1D cubic NLS and singular vortex dynamics. J. Eur. Math. Soc. 14(1), 209–253, 2012

    MathSciNet  Google Scholar 

  2. Banica, V., Vega, L.: Stability of the self-similar dynamics of a vortex filament. Arch. Ration. Mech. Anal. 210(3), 673–712, 2013

    Article  MathSciNet  Google Scholar 

  3. Banica, V., Vega, L.: Evolution of polygonal lines by the binormal flow. Ann. PDE 6(1), Paper No. 6, 53, 2020

    Article  MathSciNet  Google Scholar 

  4. Bourgain, J., Wang, W.: Construction of blowup solutions for the nonlinear Schrödinger equation with critical nonlinearity. Ann. Scuola Norm. Sup. Pisa Cl. Sci. (4) 25(1–2), 197–215, 1998 1997. Dedicated to Ennio De Giorgi

    Google Scholar 

  5. Christ, M., Colliander, J., Tao, T.: Asymptotics, frequency modulation, and low regularity ill-posedness for canonical defocusing equations. Am. J. Math. 125(6), 1235–1293, 2003

    Article  MathSciNet  Google Scholar 

  6. Correia, S., Côte, R., Vega, L.: Asymptotics in Fourier space of self-similar solutions to the modified Korteweg–de Vries equation. J. Math. Pures Appl. 9(137), 101–142, 2020

    Article  MathSciNet  Google Scholar 

  7. Correia, S., Côte, R., Vega, L.: Self-similar dynamics for the modified Korteweg–de Vries equation. Int. Math. Res. Not. 13, 9958–10013, 2021

    Article  MathSciNet  Google Scholar 

  8. Deift, P., Zhou, X.: A steepest descent method for oscillatory Riemann–Hilbert problems, asymptotics for the mkdv equation. Ann. Math. 137(2), 295–368, 1993

    Article  MathSciNet  Google Scholar 

  9. Deift, P.A., Zhou, X.: Asymptotics for the Painlevé II equation. Commun. Pure Appl. Math. 48(3), 277–337, 1995

    Article  Google Scholar 

  10. Dunst, K., Kokocki, P.: On global solutions of defocusing mKdV equation with specific initial data of critical regularity. Phys. D 417, 132810, 2021

    Article  MathSciNet  Google Scholar 

  11. Germain, P., Pusateri, F., Rousset, F.: Asymptotic stability of solitons for mKdV. Adv. Math. 299, 272–330, 2016

    Article  MathSciNet  Google Scholar 

  12. Goldstein, R.E., Petrich, D.M.: Soliton’s, Euler’s equations, and vortex patch dynamics. Phys. Rev. Lett. 69(4), 555–558, 1992

    Article  ADS  MathSciNet  Google Scholar 

  13. Grünrock, A., Vega, L.: Local well-posedness for the modified KdV equation in almost critical \(\widehat{H^s_r}\)-spaces. Trans. Am. Math. Soc. 361, 5681–5694, 2009

    Article  Google Scholar 

  14. Gutierrez, S., Vega, L.: On the stability of self-similar solutions of 1D cubic Schrödinger equations. Math. Ann. 356(1), 259–300, 2013

    Article  MathSciNet  Google Scholar 

  15. Harrop-Griffiths, B.: Long time behavior of solutions to the mKdV. Commun. Partial Differ. Equ. 41(2), 282–317, 2016

    Article  MathSciNet  Google Scholar 

  16. Harrop-Griffiths, B., Killip, R., Visan, M.: Sharp well-posedness for the cubic NLS and mKdV in \(H^s(\mathbb{R})\). Preprint, arXiv:2003.05011

  17. Hastings, S.P., McLeod, J.B.: A boundary value problem associated with the second Painlevé transcendent and the Korteweg–de Vries equation. Arch. Rational Mech. Anal. 73(1), 31–51, 1980

    Article  ADS  MathSciNet  Google Scholar 

  18. Hayashi, N., Naumkin, P.I.: Large time behavior of solutions for the modified Korteweg–de Vries equation. Int. Math. Res. Not. 8, 395–418, 1999

    Article  MathSciNet  Google Scholar 

  19. Hayashi, N., Naumkin, P.I.: On the modified Korteweg–de Vries equation. Math. Phys. Anal. Geom. 4(3), 197–227, 2001

    Article  MathSciNet  Google Scholar 

  20. Kenig, C.E., Ponce, G., Vega, L.: Well-posedness and scattering result for the generalized Korteweg–De Vries equation via contraction principle. Commun. Pure Appl. Math. 46, 527–620, 1993

    Article  MathSciNet  Google Scholar 

  21. Kenig, C.E., Ponce, G., Vega, L.: On the ill-posedness of some canonical dispersive equations. Duke Math. J. 106(3), 617–633, 2001

    Article  MathSciNet  Google Scholar 

  22. Perelman, G., Vega, L.: Self-similar planar curves related to modified Korteweg–de Vries equation. J. Differ. Equ. 235(1), 56–73, 2007

    Article  ADS  MathSciNet  Google Scholar 

Download references

Acknowledgements

We would like to thank Luis Vega for his encouragement and insightful remarks, and the anonymous referee for his careful reading and comments which improved the manuscript.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Raphël Côte.

Ethics declarations

Conflict of interest

The authors are not aware of any Conflict of interest regarding this manuscript.

Additional information

Communicated by N. Masmoudi.

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

S.C. was partially supported by Fundação para a Ciência e Tecnologia, through CAMGSD, IST-ID (Projects UIDB/04459/2020 and UIDP/04459/2020) and through the project NoDES (PTDC/MAT-PUR/1788/2020). The research of R.C. has benefitted from support provided by the University of Strasbourg Institute for Advanced Study (USIAS) for a Fellowship within the French national programme “Investment for the future” (IdEx-Unistra). This work of the Interdisciplinary Thematic Institute IRMIA++, as part of the ITI 2021-2028 program of the University of Strasbourg, CNRS and Inserm, was supported by IdEx Unistra (ANR-10-IDEX-0002), and by SFRI-STRAT’US project (ANR-20-SFRI-0012) under the framework of the French Investments for the Future Program.

Appendix A: Proof of Estimate (2.1)

Appendix A: Proof of Estimate (2.1)

Proof of Lemma 9

As we are to prove a trilinear estimate, we can rescale and we will assume, without loss of generality, that

$$\begin{aligned} \Vert f\Vert _{\mathscr {E}(t)} = \Vert g\Vert _{\mathscr {E}(t)} = \Vert h \Vert _{\mathscr {E}(t)} =1, \end{aligned}$$

so that

$$\begin{aligned} \Vert \tilde{f} \Vert _{L^\infty } \leqq 1 \quad \text {and} \quad \Vert \partial _{\xi }\tilde{f} \Vert _{L^2} \leqq t^{1/6}, \end{aligned}$$
(A.1)

and the same for \({\tilde{g}}\) and \({\tilde{h}}\).

As in Lemma 11, given a domain \(\mathscr {D} \subset \mathbb {R}^2\), we write

$$\begin{aligned} J(\mathscr {D}) = \iint _{\mathscr {D}} e^{it\Phi }\tilde{f}(\xi _1)\tilde{g}(\xi _2)\xi _3\tilde{h}(\xi _3)\textrm{d}\xi _1 \textrm{d}\xi _2. \end{aligned}$$

We will only do the proof when the integral is restricted to the domain

$$\begin{aligned} \mathscr {A}:= \{ (\xi _1,\xi _2) \in \mathbb {R}^2: |\xi _3|\geqq |\xi _2|\geqq |\xi _1| \}, \end{aligned}$$

where \(\xi _3\) is the largest frequency: this is actually the worst-case scenario, and it is clear from the computations below how to adapt the estimates to the others domains.

  1. 1)

    For small frequencies, we have a crude bound: let \(\mathscr {A}_0 = \{ (\xi _1,\xi _2) \in \mathscr {A}: |\xi _3| \leqq t^{-1/3} \}\). Then

    $$\begin{aligned} |J(\mathscr {A}_0)| \leqq \iint _{\mathscr {A}_0} |\xi _3| \textrm{d}\xi _1 \textrm{d}\xi _2 \Vert {\tilde{f}}\Vert _{L^\infty } \Vert {\tilde{g}} \Vert _{L^\infty } \Vert {\tilde{h}} \Vert _{L^\infty } \leqq \frac{1}{t}. \end{aligned}$$
  2. 2)

    We now consider the case when \(|\xi _2|\) is significantly smaller than \(|\xi _3|\) (and so \(|\xi _1|\) as well):

    $$\begin{aligned} \mathscr {A}_1 = \left\{ (\xi _1,\xi _2) \in \mathscr {A} {\setminus } \mathscr {A}_0: |\xi _3| - |\xi _2| \geqq \frac{|\xi _3|}{10} \right\} . \end{aligned}$$

    Notice that, on \(\mathscr {A}_1\),

    $$\begin{aligned} |\partial _{\xi _1} \Phi | = 3(\xi _3^2 - \xi _1^2) \gtrsim |\xi _3|^2 \text{ and } |\partial _{\xi _1}^2 \Phi | \lesssim |\xi _3,| \end{aligned}$$

    which allows us to perform an IBP in \(\xi _1\) using that \(e^{i t \Phi } = \frac{1}{it \partial _{1} \Phi } \partial _{\xi _1} (e^{it \Phi })\):

    $$\begin{aligned} J(\mathscr {A}_1)&= \iint _{\mathscr {A}_1} e^{it\Phi } \partial _{\xi _1} \left( \frac{\xi _3}{it \partial _{1} \Phi } \tilde{f}(\xi _1)\tilde{g}(\xi _2) \tilde{h}(\xi _3) \right) \textrm{d}\xi _1 \textrm{d}\xi _2 \nonumber \\ {}&\qquad + \int _{\partial \mathscr {A}_1} \frac{\xi _3 e^{it\Phi }}{it \partial _{1} \Phi } \tilde{f}(\xi _1)\tilde{g}(\xi _2) \tilde{h}(\xi _3) \textrm{d}\sigma (\xi _1,\xi _2). \end{aligned}$$
    (A.2)

Unfortunately, one can check that a direct bound on these integrals (after distributing the derivative \(\partial _{\xi _1}\) on all factors of the first integral) lead to at least a logarithmic divergence.

For the terms in (A.2) where the derivative in \(\xi _1\) does not fall on \(\tilde{h}(\xi _3)\), we perform a second IBP, this time in \(\xi _2\), writing this time \(e^{i t \Phi } = \frac{1}{it \partial _{2} \Phi } \partial _{\xi _2} (e^{it \Phi })\). Notice that on \(\mathscr {A}_1\), \(|\partial _{2} \Phi | \gtrsim |\xi _3|^2\) and \(|\nabla ^2 \Phi | \lesssim |\xi _3|\), \(|\nabla ^3 \Phi | \lesssim 1\). For example, when the derivative in (A.2) falls on the phase \(\partial _{1} \Phi \):

$$\begin{aligned} J_1&: =\left| \iint _{\mathscr {A}_1} e^{it\Phi } \left( \frac{\xi _3 \partial _{1 1}^2 \Phi }{it (\partial _{1} \Phi )^2} \tilde{f}(\xi _1)\tilde{g}(\xi _2) \tilde{h}(\xi _3) \right) \textrm{d}\xi _1 \textrm{d}\xi _2 \right| \nonumber \\ {}&\leqq \left| \iint _{\mathscr {A}_1} e^{it\Phi } \partial _{\xi _2} \left( \frac{\xi _3 \partial _{1 1}^2 \Phi }{t^2 (\partial _{1} \Phi )^2 \partial _{2} \Phi } \tilde{f}(\xi _1)\tilde{g}(\xi _2) \tilde{h}(\xi _3) \right) \textrm{d}\xi _1 \textrm{d}\xi _2 \right| \\ {}&\qquad + \left| \int _{\partial \mathscr {A}_1} \frac{e^{it\Phi } \xi _3 \partial _{1 1}^2 \Phi }{t^2 (\partial _{1} \Phi )^2 \partial _{2} \Phi } \tilde{f}(\xi _1)\tilde{g}(\xi _2) \tilde{h}(\xi _3) \textrm{d}\sigma (\xi _1,\xi _2) \right| \\ {}&\lesssim \frac{1}{t^2} \iint _{\mathscr {A}_1} \frac{1}{|\xi _3|^5} \textrm{d}\xi _1 \textrm{d}\xi _2 \Vert {\tilde{f}} \Vert _{L^\infty } \Vert {\tilde{g}} \Vert _{L^\infty } \Vert {\tilde{h}} \Vert _{L^\infty } \\ {}&\qquad + \frac{1}{t^2}\iint _{\mathscr {A}_1} \Vert {\tilde{f}} \Vert _{L^\infty } (|\partial _\xi {\tilde{g}}(\xi _2)| \Vert {\tilde{h}} \Vert _{L^\infty } +\Vert {\tilde{g}} \Vert _{L^\infty } |\partial _\xi {\tilde{h}}(\xi _3)| ) \frac{\textrm{d}\xi _1 \textrm{d}\xi _2}{|\xi _3|^4} \\ {}&\qquad + \frac{1}{t^2} \int _{\partial \mathscr {A}_1} \frac{\textrm{d}\sigma (\xi _1,\xi _2)}{|\xi _3|^4} \Vert {\tilde{f}} \Vert _{L^\infty } \Vert {\tilde{g}} \Vert _{L^\infty } \Vert {\tilde{h}} \Vert _{L^\infty }. \end{aligned}$$

Now with (A.1) in mind, and with the change of variable in \(\xi _2\) defined by \(\xi _2' = \xi _3 =\xi -\xi _1-\xi _2\) and \(\mathscr {B}_1 = \{(\xi _1,\xi _2'): (\xi _1, \xi -\xi _1-\xi _2') \in \mathscr {A}_1 \}\) we estimate

$$\begin{aligned} \iint _{\mathscr {A}_1} \frac{\textrm{d}\xi _1 \textrm{d}\xi _2}{|\xi _3|^5}&= \iint _{\mathscr {B}_1} \frac{\textrm{d}\xi _1 \textrm{d}\xi _2'}{|\xi _2'|^5} \leqq \int _{|\xi _2'| \geqq t^{-1/3}} \left( \int _{|\xi _1| \leqq |\xi _2'|} \frac{\textrm{d}\xi _1 }{|\xi _2'|^5} \right) \textrm{d}\xi _2' \\&= 2\int _{|\xi _2'| \geqq t^{-1/3}} \frac{\textrm{d}\xi _2'}{|\xi _2'|^4} \lesssim t. \end{aligned}$$

Similarly, using the Cauchy-Schwarz inequality,

$$\begin{aligned} \iint _{\mathscr {A}_1} |\partial _\xi {\tilde{g}}(\xi _2)| \frac{\textrm{d}\xi _1 \textrm{d}\xi _2}{|\xi |^4}&\leqq \int _{|\xi _1| \geqq t^{-1/3}} \Vert \partial _\xi {\tilde{g}} \Vert _{L^2}\left( \int _{|\xi _2| \geqq |\xi _1|} \frac{\textrm{d}\xi _2}{|\xi _2|^8} \right) ^{1/2} \textrm{d}\xi _1 \\ {}&\leqq \int _{|\xi _1| \geqq t^{-1/3}} \frac{\textrm{d}\xi _1}{|\xi _1|^{7/2}} \Vert \partial _\xi {\tilde{g}} \Vert _{L^2} \lesssim t^{5/2 \cdot 1/3 + 1/6} \lesssim t, \end{aligned}$$

which allows to take care of the two terms the second last line. As for the boundary term,

$$\begin{aligned} \int _{\partial \mathscr {A}_1} \frac{\textrm{d}\sigma (\xi _1,\xi _2)}{|\xi _3|^4} = \int _{\partial \mathscr {B}_1} \frac{\textrm{d}\sigma (\xi _1,\xi _2')}{|\xi _2'|^4} \lesssim t. \end{aligned}$$

We hence obtained

$$\begin{aligned} J_1 \lesssim \frac{1}{t}. \end{aligned}$$

As mentioned, similar computations can be done for all terms, except for the term with a derivative on \({\tilde{h}}\):

$$\begin{aligned} J_1'&:= \iint _{\mathscr {A}_1} \left( \frac{e^{it\Phi } \xi _3}{it \partial _{1} \Phi } \tilde{f}(\xi _1)\tilde{g}(\xi _2) \partial _\xi \tilde{h}(\xi _3) \right) \textrm{d}\xi _1 \textrm{d}\xi _2 \end{aligned}$$

Here, we first perform the change of variable in \(\xi _1\) defined by \(\xi _1':= \xi - \xi _1 - \xi _2 = \xi _3\) so that \(\xi _3':= \xi - \xi _1' - \xi _2 = \xi _1\), with \(\mathscr {B}_1' = \{(\xi _1',\xi _2): (\xi - \xi _1'-\xi _2, \xi _2) \in \mathscr {A}_1 \}\):

$$\begin{aligned} J_1'&= \iint _{\mathscr {B}_1'} \left( \frac{e^{it\Phi } \xi _2'}{it \partial _{1} \Phi (\xi _3', \xi _2)} \tilde{f}(\xi _3')\tilde{g}(\xi _2) \partial _\xi \tilde{h}(\xi _1') \right) \textrm{d}\xi _1' \textrm{d}\xi _2. \end{aligned}$$

Then we are in a position to perform once again an IBP in \(\xi _2\) (because there will be no derivative falling on \(\partial _\xi {\tilde{h}}\)). On \(\mathscr {B}_1'\),

$$\begin{aligned} \partial _{1} \Phi (\xi _3', \xi _2) = 3|\xi _1'^2 - \xi _3'| \gtrsim |\xi _1'|^2, \quad |\partial _{2} \Phi (\xi _3',\xi _2)| = 3|\xi _1'^2 - \xi _2^2| \gtrsim |\xi _1'|^2, \end{aligned}$$

\(|\xi _1'| \geqq t^{-1/3}\) and \(|\xi _1'| \geqq |\xi _2| \geqq |\xi _3'|\). These estimates allow us to argue as for \(J_1\) and derive

$$\begin{aligned} |J(\mathscr {A}_1)| \lesssim \frac{1}{t}. \end{aligned}$$
  1. 3)

    We then consider the case when only \(\xi _1\) is significantly smaller:

    $$\begin{aligned} \mathscr {A}_2 = \left\{ (\xi _1,\xi _2) \in \mathscr {A} {\setminus } \mathscr {A}_0: | |\xi _3| - |\xi _2| | \leqq \frac{|\xi _3|}{10} \text { and } |\xi _3| - |\xi _1| \geqq \frac{|\xi _3|}{10} \right\} . \end{aligned}$$

    We start as in 2), first performing an IBP in \(\xi _1\):

    $$\begin{aligned} J(\mathscr {A}_2)&= \iint _{\mathscr {A}_2} e^{it\Phi } \partial _{\xi _1} \left( \frac{\xi _3}{it \partial _{1} \Phi } \tilde{f}(\xi _1)\tilde{g}(\xi _2) \tilde{h}(\xi _3) \right) \textrm{d}\xi _1 \textrm{d}\xi _2 \nonumber \\ {}&\qquad + \int _{\partial \mathscr {A}_2} \frac{\xi _3 e^{it\Phi }}{it \partial _{1} \Phi } \tilde{f}(\xi _1)\tilde{g}(\xi _2) \tilde{h}(\xi _3) \textrm{d}\sigma (\xi _1,\xi _2). \end{aligned}$$
    (A.3)

    Except on the term where the derivative \(\partial _{\xi _1}\) falls on \({\tilde{h}}\) (see step 3.3 below), we perform on (A.3) an IBP in \(\xi _2\), using the slighlty different identity

    $$\begin{aligned} e^{i t \Phi } = \frac{1}{1 + it \xi _2 \partial _2 \Phi } \partial _{\xi _2} (\xi _2 e^{i t \Phi }). \end{aligned}$$

    For example, if in (A.3) the \(\partial _{\xi _1}\) derivative fell on \({\tilde{f}}\), we are to bound

    $$\begin{aligned} J_2&: = \left| \iint _{\mathscr {A}_2} e^{it\Phi } \left( \frac{\xi _3}{it \partial _{1} \Phi } \partial _\xi \tilde{f}(\xi _1) \tilde{g}(\xi _2) \tilde{h}(\xi _3) \right) \textrm{d}\xi _1 \textrm{d}\xi _2 \right| \\ {}&\leqq \left| \iint _{\mathscr {A}_2} e^{it\Phi } \partial _{\xi _2} \left( \frac{\xi _2 \xi _3}{(1 + it \xi _2 \partial _2 \Phi )(it \partial _{1} \Phi )} \partial _\xi \tilde{f}(\xi _1) \tilde{g}(\xi _2) \tilde{h}(\xi _3) \right) \textrm{d}\xi _1 \textrm{d}\xi _2 \right| \\ {}&\qquad + \left| \iint _{\partial \mathscr {A}_2} e^{it\Phi } \frac{\xi _2 \xi _3}{(1 + it \xi _2 \partial _2 \Phi )(it \partial _{1} \Phi )} \partial _\xi \tilde{f}(\xi _1) \tilde{g}(\xi _2) \tilde{h}(\xi _3) \textrm{d}\sigma (\xi _1,\xi _2) \right| \end{aligned}$$

We now split into two cases.

3.1) Let \(\mathscr {B}_2 = \{ (\xi _1, \xi _2) \in \mathscr {A}_2: |\xi _3 - \xi _2| \leqq \frac{|\xi _3|}{100} \}\).

On \(\mathscr {B}_2\), \(|\partial _1 \Phi | =3|\xi _3^2 - \xi _1^2| \gtrsim |\xi _3|^2\) and

$$\begin{aligned} |\partial _2 \Phi | =3|\xi _3^2 - \xi _2^2| \gtrsim |\xi _3| |\xi _3-\xi _2| = |\xi _3| |\xi -\xi _1 - 2 \xi _2|. \end{aligned}$$

Also using that \(|\xi _2| \gtrsim |\xi _3| \geqq |\xi _2|\) and \(|\nabla ^2 \Phi | \lesssim |\xi _3|\), we infer

$$\begin{aligned} J_2&\lesssim \iint _{\mathscr {A}_1} \frac{|\xi _2| | \partial _\xi {\tilde{f}}(\xi _1)|}{(1+t|\xi _2|^2 |\xi _3-\xi _2|) t |\xi _2|^2} \textrm{d}\xi _1 \textrm{d}\xi _2 \Vert {\tilde{g}} \Vert _{L^\infty } \Vert {\tilde{h}}\Vert _{L^\infty } \end{aligned}$$
(A.4)
$$\begin{aligned}&\qquad + \iint _{\mathscr {A}_1} \frac{|\xi _2|^3 | \partial _\xi {\tilde{f}}(\xi _1)|}{(1+t|\xi _2|^2 |\xi _3-\xi _2|)^2 t |\xi _2|^2} \textrm{d}\xi _1 \textrm{d}\xi _2 \Vert {\tilde{g}} \Vert _{L^\infty } \Vert {\tilde{h}}\Vert _{L^\infty } \nonumber \\ {}&\qquad + \iint _{\mathscr {A}_1} \frac{|\xi _2|^2 | \partial _\xi {\tilde{f}}(\xi _1)| ( | \partial _\xi {\tilde{g}}(\xi _2)| \Vert {\tilde{h}}\Vert _{L^\infty } + \Vert {\tilde{g}}\Vert _{L^\infty } | \partial _\xi {\tilde{h}}(\xi _3)| )}{(1+t|\xi _2|^2 ||\xi _3-\xi _2||) t |\xi _2|^2} \textrm{d}\xi _1 \textrm{d}\xi _2 \nonumber \\ {}&\qquad + \iint _{\partial \mathscr {A}_1} \frac{|\xi _2|^2 | \partial _\xi {\tilde{f}}(\xi _1)|}{(1+t|\xi _2|^2 ||\xi _3-\xi _2||) t |\xi _2|^2} \textrm{d}\sigma (\xi _1,\xi _2) \Vert {\tilde{g}} \Vert _{L^\infty } \Vert {\tilde{h}}\Vert _{L^\infty } \end{aligned}$$
(A.5)

Then for fixed \(\xi _2\), we can bound

$$\begin{aligned} \int _{\xi _1} \frac{ | \partial _\xi {\tilde{f}}(\xi _1)|}{(1+t|\xi _2|^2 |\xi _3-\xi _2|)} \textrm{d}\xi _1&\leqq \left( \int \frac{\textrm{d}\xi _1}{(1+t|\xi _2|^2 |\xi -\xi _1-2\xi _2|)^2} \right) ^{1/2} \Vert \partial _\xi {\tilde{f}} \Vert _{L^2} \\&\lesssim \frac{1}{t^{1/2} |\xi _2|}\Vert \partial _\xi {\tilde{f}} \Vert _{L^2} \end{aligned}$$

so that for the term in (A.4)

$$\begin{aligned} \iint _{\mathscr {A}_1} \frac{|\xi _2| | \partial _\xi {\tilde{f}}(\xi _1)|}{(1+t|\xi _2|^2 |\xi _3-\xi _2|) t |\xi _2|^2} \textrm{d}\xi _1 \textrm{d}\xi _2&\lesssim \frac{1}{t^{3/2}} \int _{|\xi _2| \geqq t^{-1/3}} \frac{\textrm{d}\xi _2}{|\xi _2|^{2}} \Vert \partial _\xi {\tilde{f}} \Vert _{L^2} \\&\lesssim t^{-3/2+ 1/3+1/6} \leqq \frac{1}{t}. \end{aligned}$$

Similarly, for (A.4)

$$\begin{aligned}&\iint _{\mathscr {A}_1} \frac{|\xi _2|^2 | \partial _\xi {\tilde{f}}(\xi _1)| | \partial _\xi {\tilde{g}}(\xi _2)|}{(1+t|\xi _2|^2 |\xi _3-\xi _2|) t |\xi _2|^2} \textrm{d}\xi _1 \textrm{d}\xi _2\\ \lesssim \ {}&\frac{1}{t} \int _{|\xi _2| \geqq t^{-1/3}} | \partial _\xi {\tilde{g}}(\xi _2)| \left( \int _{\xi _1} \frac{ | \partial _\xi {\tilde{f}}(\xi _1)| \textrm{d}\xi _1}{(1+t|\xi _2|^2 |\xi -\xi _1-2\xi _2|) } \right) \textrm{d}\xi _2 \\ \lesssim \ {}&\frac{1}{t^{3/2}} \int _{|\xi _2| \geqq t^{-1/3}}|\partial _\xi {\tilde{g}}(\xi _2)| \frac{\textrm{d}\xi _2}{|\xi _2|} \Vert \partial _\xi {\tilde{f}} \Vert _{L^2} \lesssim t^{-3/2+1/3 \cdot 1/2} \Vert \partial _\xi {\tilde{g}} \Vert _{L^2} \Vert \partial _\xi {\tilde{f}} \Vert _{L^2} \lesssim \frac{1}{t}. \end{aligned}$$

One can bound the other terms in \(J_2\) in the same fashion.

3.2) Let \(\mathscr {C}_2 = \{ (\xi _1, \xi _2) \in \mathscr {A}_2: |\xi _3 + \xi _2| \leqq \frac{|\xi _3|}{100} \}\). Then on \(\mathscr {C}_2\), \(|\partial _1 \Phi | =3|\xi _3^2 - \xi _1^2| \gtrsim |\xi _3|^2\) and

$$\begin{aligned} |\partial _2 \Phi | =3|\xi _3^2 - \xi _2^2| \gtrsim |\xi _3| |\xi _3+\xi _2| = |\xi _3| |\xi -\xi _1|. \end{aligned}$$

We still have \(|\xi _2| \gtrsim |\xi _3| \geqq |\xi _2| \gtrsim |\xi |, |\xi _1|\) and \(|\nabla ^2 \Phi | \lesssim |\xi _3|\).

We can then proceed as in 3.1), integrating first in \(\xi _1\) for \(\xi _2\) fixed.

Summing up, as \(\mathscr {A}_2 = \mathscr {B}_2 \cup \mathscr {C}_2\), we get that

$$\begin{aligned} J_2 \lesssim \frac{1}{t}. \end{aligned}$$

3.3) To complete the bound on \(J(\mathscr {A}_2)\), it remains to consider the term in (A.3) when the derivative \(\partial _{\xi _1}\) of the (first) IBP falls on \({\tilde{h}}\):

$$\begin{aligned} J_2': = \left| \iint _{\mathscr {A}_2} e^{it\Phi } \left( \frac{\xi _3}{it \partial _{1} \Phi } \tilde{f}(\xi _1) \tilde{g}(\xi _2) \partial _\xi \tilde{h}(\xi _3) \right) \textrm{d}\xi _1 \textrm{d}\xi _2 \right| \\ \end{aligned}$$

As in case 2), we perform the change of variable in \(\xi _1\) defined by \(\xi _1':= \xi - \xi _1 - \xi _2 = \xi _3\) so that \(\xi _3':= \xi - \xi _1' - \xi _2 = \xi _1\), with \(\mathscr {B}_2' = \{(\xi _1',\xi _2): (\xi - \xi _1'-\xi _2, \xi _2) \in \mathscr {A}_2 \}\):

$$\begin{aligned} J_2'&= \iint _{\mathscr {B}_2'} \left( \frac{e^{it\Phi } \xi _2'}{it \partial _{1} \Phi (\xi _3', \xi _2)} \tilde{f}(\xi _3')\tilde{g}(\xi _2) \partial _\xi \tilde{h}(\xi _1') \right) \textrm{d}\xi _1' \textrm{d}\xi _2, \end{aligned}$$

and then we are in a position for an IBP in \(\xi _2\). We conclude by doing the same computations as in 3.1) and 3.2), and we obtain, as desired,

$$\begin{aligned} J(\mathscr {A}_2) \lesssim \frac{1}{t}. \end{aligned}$$

4) We are now in the case when \(|\xi _1|, |\xi _2|, |\xi _3|\) are similar, which corresponds to the stationary points. There are two cases: either they all have same signs (corresponding to \((\xi /3,\xi /3,\xi /3)\)), or they don’t (corresponding to \((\xi ,\xi ,-\xi )\) and its two symmetric). We thus consider

$$\begin{aligned} \mathscr {A}_3 := \left\{ (\xi _1,\xi _2) \in \mathscr {A} : |\xi _3-\xi _2|, |\xi _3-\xi _1| \leqq \frac{|\xi _3|}{10} \right\} \\ \mathscr {A}_4 : = \left\{ (\xi _1,\xi _2) \in \mathscr {A}: |\xi _3 + \xi _2|, |\xi _3+\xi _1| \leqq \frac{|\xi _3|}{10} \right\} . \end{aligned}$$

The other two stationary points can be treated as for \(\mathscr {A}_4\).

In these stationary regions, we must exploit as much as possible the phase oscillations. To that end, we adapt our reference frame to the directions of maximal oscillation (see (A.6)). This analysis was crucial to derive the refined estimates in [18] and [7]. The computations below follow closely the arguments therein.

As \(\xi _1\), \(\xi _2\), \(\xi _3\) are all of the order of \(\xi \), it is convenient to rescale and denote \(q_i = \xi _i/\xi \) so that \(q_1+q_2+q_3 =1\), and the phase

$$\begin{aligned} \Phi (\xi _1,\xi _2) = \xi ^3 Q(q_1,q_2) = 3 \xi ^3 (1-q_1)(1-q_2)(1-q_3). \end{aligned}$$

By symmetry, we can also assume without loss of generality that \(\xi >0\), and we denote the rescaling factor

$$\begin{aligned} \tau := t \xi ^3 \geqq 1 \quad \text {on } \mathscr {A}_3 \cup \mathscr {A}_4. \end{aligned}$$

(\(|\tau | \leqq 1\) corresponds to \(\mathscr {A}_0\)). Denote the nonlinear function without the \(\xi _3\) factor

$$\begin{aligned} F:= {\tilde{f}}(\xi q_1) {\tilde{g}}(\xi q_2) {\tilde{h}}(\xi q_3). \end{aligned}$$

We now introduce the change of variable

$$\begin{aligned} 1-q_1=\lambda - \mu ,\quad 1-q_2= \lambda + \mu ,\quad 1-q_3=2(1-\lambda ), \end{aligned}$$

so that

$$\begin{aligned} Q(q_1,q_2) = 6(1-\lambda ) (\lambda -\mu )(\lambda +\mu ). \end{aligned}$$

The stationary point correspond to \((\lambda ,\mu ) = (2/3,0)\) and (0, 0), respectively. We integrate by parts using the relation

$$\begin{aligned} e^{i\tau Q}=\frac{1}{1+12i\tau \mu ^2(1-\lambda )}\partial _ \mu ( \mu e^{i\tau Q}) \end{aligned}$$

so that, with \(\mathscr {D}' = \{ (\lambda , \mu ): (1-(\lambda -\mu ), 1-(\lambda +\mu ) \in \mathscr {D} \}\), where \(\mathscr {D}\) is an integration domain and

$$\begin{aligned} A = 1+12i\tau \mu ^2(1-\lambda ),\quad \partial _\mu A = 24 i \tau \mu (1-\lambda ), \end{aligned}$$

and depending whether the derivative falls on the phase or not,

$$\begin{aligned} | J(\mathscr {D})|&= 2\xi ^3 \iint _{\mathscr {D}} e^{i \tau Q} F (2\lambda -1) \textrm{d}\lambda \textrm{d}\mu = - 2\xi ^3 \iint _{\mathscr {D}} \frac{1}{A} \partial _\mu ( \mu e^{i \tau Q}) F (2\lambda -1) \textrm{d}\lambda \textrm{d}\mu \\ {}&= 2\xi ^3 \iint _{\mathscr {D}'} e^{i\tau Q} F \frac{\mu \partial _\mu A (2\lambda -1)}{A^2} \textrm{d}\lambda \textrm{d}\mu + 2\xi ^3 \iint _{\mathscr {D}'} e^{i\tau Q}\partial _{1} F \frac{ \mu (2\lambda -1)}{A}\textrm{d}\lambda \textrm{d}\mu \\ {}&\qquad - 2\xi ^3 \iint _{\mathscr {D}'} e^{i\tau Q}\partial _{2} F \frac{ \mu (2\lambda -1)}{A}\textrm{d}\lambda \textrm{d}\mu \\ {}&=: 2\xi ^3 (J_\sharp (\mathscr {D}') + J_\flat (\mathscr {D}') + J_\natural (\mathscr {D}')). \end{aligned}$$

(Observe that there is no derivative on \({\tilde{h}}\) in this computation.) By symmetry between \(\partial _1 F\) and \(\partial _2 F\), it suffices to bound \(J_\sharp (\mathscr {D}')\) and \(J_\flat (\mathscr {D}')\).

5) Here we bound \(J_\sharp \). We do yet another change of variable of \(\mu \) defined by \(\nu = \mu \sqrt{1-\lambda }\) so that Q has separate variables:

$$\begin{aligned} Q = 6 (1-\lambda ) \lambda ^2 - 6\nu ^2, \quad \partial _\lambda Q = 6 \lambda (2-3\lambda ). \end{aligned}$$
(A.6)

We perform an integration by parts using, for fixed \(\lambda _0 \in \{ 0, 2/3 \}\)

$$\begin{aligned} e^{i\tau Q}=\frac{1}{1 + 6i\tau (\lambda -\lambda _0) \lambda (2-3\lambda )}\partial _\lambda ((\lambda -\lambda _0)e^{i\tau Q}). \end{aligned}$$

Denoting

$$\begin{aligned} B = 1 + 6 i \tau (\lambda -\lambda _0) \lambda (2-3\lambda ), \quad \partial _\lambda B = 6i\tau (\lambda (2-3\lambda ) + 2 (\lambda -\lambda _0)(1-3\lambda )), \end{aligned}$$

we have

$$\begin{aligned} J_\sharp (\mathscr {D}')&= \iint _{\mathscr {D}'} e^{i\tau Q}F \frac{24i\tau (1-\lambda ) \mu ^2}{A^2}\textrm{d}\lambda \textrm{d}\mu \\ {}&= \iint _{\mathscr {D}'} \partial _{\lambda } (\lambda - \lambda _0) e^{i\tau Q}) F \frac{24i\tau \nu ^2}{A^2B} \frac{2\lambda -1 }{\sqrt{1-\lambda }} \textrm{d}\lambda \textrm{d}\nu \\ {}&= - \iint _{\mathscr {D}'} e^{i\tau Q}\left[ - \partial _1 F - \partial _2 F + 2 \partial _3 F \right] \times \frac{24i\tau (\lambda -\lambda _0) \nu ^2 }{A^2 B} \frac{2\lambda -1}{\sqrt{1-\lambda }} d \lambda \textrm{d}\nu \\ {}&\quad + \iint _{\mathscr {D}'} e^{i\tau Q} F \frac{24i\tau (\lambda - \lambda _0) \nu ^2}{A^2 B} \left( 2+ \frac{1}{2(1-\lambda ) } + \frac{\partial _\lambda B}{B}\right) \frac{2\lambda -1}{\sqrt{1-\lambda }} \textrm{d}\lambda \textrm{d}\nu \\ {}&\quad + \int _{\partial \mathscr {D}'} (\lambda - \lambda _0) e^{i\tau Q} F \frac{24i\tau \nu ^2}{A^2B}\frac{2\lambda -1}{\sqrt{1-\lambda }} \textrm{d}\sigma (\tilde{\lambda },\nu ). \end{aligned}$$

5.1) On \(\mathscr {A}_3'\), we choose \(\lambda _0 = 2/3\). Then we have \(|\lambda - 2/3| \leqq 1/10\) and \(|\nu | \leqq 1/10\) so that

$$\begin{aligned} \lambda , 1-\lambda , 1-3\lambda , 2\lambda -1 \end{aligned}$$

can be bounded (above and below) by uniform constants. In particular, letting \(\tilde{\lambda }:= \lambda - 2/3\),

$$\begin{aligned} |A| \gtrsim 1+ \tau \nu ^2, \quad |B| \gtrsim 1+ \tau \tilde{\lambda }^2, \end{aligned}$$

and we can bound

$$\begin{aligned} |J_\sharp (\mathscr {A}_3')|&\lesssim \iint _{\mathscr {A}_3^\sharp } \frac{\tau \nu ^2 |\tilde{\lambda }|}{(1+\tau \nu ^2)^2(1+ \tau \tilde{\lambda }^2)}(|\partial _1 F| + |\partial _2 F| + |\partial _3 F|)\textrm{d}\tilde{\lambda }\textrm{d}\nu \nonumber \\&\quad + \iint _{\mathscr {A}_3^\sharp } \frac{\tau \nu ^2 |\tilde{\lambda }|}{(1+\tau \tilde{\lambda }^2) (1+\tau \nu ^2)^2} \left( 1 + \frac{\tau | \tilde{\lambda }|}{1 + \tau \tilde{\lambda }^2} \right) |F| \textrm{d}\tilde{\lambda }\textrm{d}\nu \nonumber \\&\quad + \int _{\partial \mathscr {A}_3^\sharp } \frac{\tau \nu ^2 \tilde{|}\lambda |}{(1+\tau \tilde{\lambda }^2) (1+\tau \nu ^2)^2} |F| \textrm{d}\sigma (\tilde{\lambda },\nu ). \end{aligned}$$
(A.7)

First consider the first term. Recall that \(\Vert F \Vert _{L^\infty } \lesssim \Vert {\tilde{f}} \Vert _{L^\infty } \Vert {\tilde{g}} \Vert _{L^\infty } \Vert {\tilde{h}} \Vert _{L^\infty } \lesssim 1\) and

$$\begin{aligned} \left( \int |\partial _1 F|^2 \textrm{d}\tilde{\lambda }\right) ^{1/2} \lesssim |\xi |^{1/2} \Vert \partial _\xi {\tilde{f}} \Vert _{L^2} \Vert {\tilde{g}} \Vert _{L^\infty } \Vert {\tilde{h}} \Vert _{L^\infty } \lesssim \tau ^{1/6}, \end{aligned}$$

and similarly for \(\partial _2 F\) and \(\partial _3 F\). Keeping in mind that \(\tilde{\lambda }\in [-1,1]\) uniformly on \(\mathscr {A}_3^\sharp = \{ (\tilde{\lambda },\nu ): (\lambda ,\mu ) \in \mathscr {A}_3' \}\), we bound

$$\begin{aligned}&\int _{-1}^1 \frac{|\tilde{\lambda }|}{1+ \tau \tilde{\lambda }^2} |\partial _j F| \textrm{d}\tilde{\lambda }\lesssim \Vert \partial _j F \Vert _{L^2} \left( \int _{-1}^1 \frac{\tilde{\lambda }^2 \textrm{d}\tilde{\lambda }}{(1+\tau \tilde{\lambda }^2)^2} \right) ^{1/2} \lesssim \tau ^{1/6-3/4},\quad j=1,2,3. \end{aligned}$$

Hence the first term in (A.7) is bounded by

$$\begin{aligned}&\tau ^{1/6-3/4} \int \frac{\tau \nu ^2}{(1+\tau \nu ^2)^2} \textrm{d}\nu \lesssim \tau ^{1/6-3/4- 1/2} \lesssim \tau ^{-13/12} \end{aligned}$$

(we used that \(\tau \geqq 1\)).

Similarly, for the second term, observe that

$$\begin{aligned} \int _{-1}^1 \frac{|\tilde{\lambda }|}{1+ \tau \tilde{\lambda }^2} \left( 1 + \frac{\tau |\tilde{\lambda }|}{1 + \tau \tilde{\lambda }^2} \right) \textrm{d}\tilde{\lambda }\lesssim \tau ^{-1} \ln (1+\tau ) + \tau ^{-1/2} \lesssim \tau ^{-1/2}, \end{aligned}$$

so that the second term is bounded by

$$\begin{aligned} \tau ^{-1/2} \int \frac{\tau \nu ^2}{(1+\tau \nu ^2)^2} \textrm{d}\nu \lesssim \tau ^{-1}. \end{aligned}$$

The boundary term can be bounded accordingly by

$$\begin{aligned} \int _{-1}^1 \frac{\tau |\rho ^3| \textrm{d}\rho }{(1+ \tau \rho ^2)^3} \lesssim \tau ^{-1}. \end{aligned}$$

Hence, as \(\tau \geqq 1\), we obtain

$$\begin{aligned} |J_\sharp (\mathscr {A}_3')| \lesssim \tau ^{-1}. \end{aligned}$$

5.2) On \(\mathscr {A}_4'\), we choose \(\lambda _0 = 0\). Then we have \(|\lambda | \leqq 1/10\) and \(|\nu | \leqq 1/10\) so that

$$\begin{aligned} 1-\lambda , 1-3\lambda , 2 - 3\lambda \end{aligned}$$

can be bounded (above and below) by uniform constants and

$$\begin{aligned} |A| \gtrsim 1+ \tau \nu ^2, \quad |B| \gtrsim 1+ \tau \lambda ^2. \end{aligned}$$

Hence, we can bound (with \(\mathscr {A}_4^\sharp = \{ (\lambda , \nu ): (\lambda ,\mu ) \in \mathscr {A}_4' \}\))

$$\begin{aligned} |J_\sharp (\mathscr {A}_4')|&\lesssim \iint _{\mathscr {A}_4^\sharp } \frac{\tau \nu ^2 |\lambda |}{(1+(\tau \nu ^2)^2)(1+ \tau \lambda ^2)}(|\partial _1 F| + |\partial _2 F| + |\partial _3 F|)d \lambda \textrm{d}\nu \\ {}&\quad + \iint _{\mathscr {A}_4^\sharp } \frac{\tau \nu ^2 |\lambda |}{(1+\tau \lambda ^2) (1+\tau \nu ^2)^2} \left( 1 + \frac{| \lambda |}{1 + \tau \lambda ^2} \right) |F| d \lambda \textrm{d}\nu \\ {}&\quad + \int _{\partial \mathscr {A}_4^\sharp } \frac{\tau \nu ^2 |\lambda |}{(1+\tau \lambda ^2) (1+\tau \nu ^2)^2} \textrm{d}\sigma ( \lambda ,\nu ). \end{aligned}$$

The exact same computations as in 5.1) give the bound

$$\begin{aligned} |J_\sharp (\mathscr {A}_4')| \lesssim \tau ^{-1}. \end{aligned}$$

6) Here we bound \(J_\flat \). We consider the change of variables

$$\begin{aligned} \mu =\frac{3\zeta +\chi -2}{2},\quad \lambda =\frac{2-\zeta +\chi }{2}. \end{aligned}$$

One may obtain this transformation by going back to the \(\xi \) variables, switching \(q_1\) with \(q_3\) and then redoing the \(\lambda ,\mu \) transformation. In this way, \(q_1\) depends on a single variable \(\zeta \), and more precisely

$$\begin{aligned} q_1 = 2 \zeta , \quad 1- q_2 = \zeta +\chi , \quad 1-q_3 = \zeta - \chi . \end{aligned}$$

This permits the integration by parts in \(\chi \) without the introduction of second-order derivatives in f. In this coordinate system, the stationary points are

$$\begin{aligned} (\chi ,\zeta )=(0,2/3) \text { and } (-1,1). \end{aligned}$$

Also

$$\begin{aligned} Q = 6 ( \zeta +\chi ) (\zeta - \chi ) (1-\zeta ), \quad \partial _\chi Q = 12 \chi (\zeta -1). \end{aligned}$$

6.1) On \(\mathscr {A}_3'\), we use the relation

$$\begin{aligned} e^{i\tau Q}=\frac{1}{1+12i\tau \chi ^2(1-\zeta )}\partial _\chi (\chi e^{i\tau Q}). \end{aligned}$$

Define

$$\begin{aligned} A{} & {} =1+4i\tau \mu ^2(1-\lambda )=1+i\tau (3\zeta +\chi -2)^2(\zeta -\chi )/2, \quad \\ C{} & {} = 1+12i\tau \chi ^2(1-\zeta ). \end{aligned}$$

We now integrate by parts:

$$\begin{aligned} J_\flat (\mathscr {A}_3')&= \iint _{\mathscr {A}_3'} e^{i\tau Q}\frac{ \mu \partial _1 F}{A}\textrm{d}\lambda \textrm{d}\mu = \iint _{\mathscr {A}_3^\flat } \frac{1}{C} \partial _{\chi } (\chi e^{i\tau Q}) \frac{\mu }{2A} \partial _1 F \textrm{d}\zeta \textrm{d}\chi \nonumber \\ {}&= - \iint _{\mathscr {A}_3^\flat } \chi e^{i\tau Q} \left[ - \partial _{12} F+ \partial _{13} F \right] \frac{\mu }{2AC} \textrm{d}\chi \textrm{d}\zeta \nonumber \\ {}&\qquad - \iint _{\mathscr {A}_3^\flat } \chi e^{i\tau Q} \frac{\partial _{1} F}{2AC} \left( 1 - \mu \left( \frac{\partial _\chi A }{A} + \frac{\partial _\chi C}{C} \right) \right) \textrm{d}\chi \textrm{d}\zeta \nonumber \\ {}&\qquad - \iint _{\partial \mathscr {A}_3^\flat } e^{i\tau Q} \frac{\chi \mu }{2AC} \partial _1 F \textrm{d}\sigma (\zeta ,\chi ). \end{aligned}$$

On \(\mathscr {A}_3^\flat \), \(1-\zeta \), \(\zeta -\chi \) are bounded above and below and

$$\begin{aligned} A \gtrsim 1 + \tau \mu ^2, \quad C \gtrsim 1 + \tau \chi ^2, \quad |\partial _\chi A| \lesssim \tau |\mu |, \quad |\partial _{\chi } C| \lesssim \tau |\chi |. \end{aligned}$$

Therefore, after changing again the variable \(\zeta \) back for \(\mu \) (and \(\mathscr {A}_3^\flat = \{ (\chi , \mu ): (\lambda ,\mu ) \in \mathscr {A}_3' \}\)),

$$\begin{aligned}&| J_\flat (\mathscr {A}_3')| \lesssim \iint _{\mathscr {A}_3^\flat } \frac{|\chi \mu |}{(1+\tau \mu ^2)(1+\tau \chi ^2)} | \partial _{12} F| + | \partial _{13} F|) \textrm{d}\chi \textrm{d}\mu \end{aligned}$$
(A.8)
$$\begin{aligned}&\quad + \iint _{\mathscr {A}_3^\flat } \frac{|\chi |}{(1+\tau \mu ^2)(1+\tau \chi ^2)} \left( 1 +\frac{|\tau | \mu ^2}{1+\tau \mu ^2} + \frac{\tau |\mu \chi |}{1+\tau \chi ^2} \right) |\partial _1 F| \textrm{d}\chi \textrm{d}\mu \end{aligned}$$
(A.9)
$$\begin{aligned}&\quad + \iint _{\partial \mathscr {A}_3^\flat } \frac{|\chi \mu |}{(1+\tau \mu ^2)(1+\tau \chi ^2)} | \partial _{1} F| \textrm{d}\sigma (\chi ,\mu ). \end{aligned}$$
(A.10)

By the Cauchy-Schwarz inequality,

$$\begin{aligned} \left( \iint | \partial _{12} F|^2 \textrm{d}\chi \textrm{d}\mu \right) ^{1/2} \lesssim |\xi | \Vert \partial _\xi {\tilde{f}} \Vert _{L^2} \Vert \partial _\xi {\tilde{g}} \Vert _{L^2} \Vert {\tilde{h}} \Vert _{L^\infty }&\lesssim \tau ^{1/3}, \\ \left( \int | \partial _{1} F|^2 \textrm{d}\chi \right) ^{1/2} \lesssim |\xi |^{1/2} \Vert \partial _\xi {\tilde{f}} \Vert _{L^2} \Vert {\tilde{g}} \Vert _{L^\infty } \Vert {\tilde{h}} \Vert _{L^\infty }&\lesssim \tau ^{1/6}, \quad \text {so that} \\ \Vert \partial _1 F \Vert _{L^\infty _\mu L^2_\chi } , \Vert \partial _1 F \Vert _{L^\infty _\chi L^2_\mu }, \Vert \partial _1 F \Vert _{L^2(\partial \mathscr {A}_3^\flat )}&\lesssim \tau ^{1/6}. \end{aligned}$$

and the first bound also holds with \(\partial _{13} F\). Hence the integral in (A.8) is bounded by

$$\begin{aligned}&\left( \iint \frac{|\mu |^2}{(1+\tau \mu ^2)^2} \frac{|\chi |^2}{(1+\tau \chi ^2)^2} \textrm{d}\mu \textrm{d}\chi \right) ^{1/2} \left( \iint (|\partial _{12} F| + |\partial _{13} F|)^2 \textrm{d}\chi \textrm{d}\mu \right) ^{1/2} \\&\lesssim \tau ^{-3/2+1/3} \lesssim \tau ^{-7/6}. \end{aligned}$$

The first two terms of the integral in (A.9) can be controlled as

$$\begin{aligned} \int \frac{\textrm{d}\mu }{1+\tau \mu ^2} \left( \int \frac{\chi ^2 \textrm{d}\chi }{1+(\tau \chi ^2)^2} \right) ^{1/2} \Vert \partial _1 F \Vert _{L^\infty _\mu L^2_\chi } \lesssim \tau ^{-1/2- 3/4+ 1/6} \lesssim \tau ^{-13/12}, \end{aligned}$$

while the third term in (A.9) is bounded by

$$\begin{aligned} \int \frac{\tau \chi ^2 \textrm{d}\chi }{(1+\tau \chi ^2)^2} \left( \int \frac{|\mu |^2 \textrm{d}\mu }{(1+\tau \mu ^2)^2} \right) ^{1/2} \Vert \partial _1 F \Vert _{L^\infty _\chi L^2_\mu } \lesssim \tau ^{-1/2-3/4+ 1/6} \lesssim \tau ^{-13/12}. \end{aligned}$$

Finally, the boundary term (A.10) can be bounded by

$$\begin{aligned} \Vert \partial _1 F \Vert _{L^2(\partial \mathscr {A}_3^\flat )} \left( \int \left( \frac{\rho ^2}{(1+\tau \rho ^2)^2} \right) ^2 \textrm{d}\rho \right) ^{1/2} \lesssim \tau ^{1/6-5/4} \lesssim \tau ^{-13/12} \end{aligned}$$

and we obtain

$$\begin{aligned} J_\flat (\mathscr {A}_3') \lesssim \tau ^{-13/12}. \end{aligned}$$

6.2) On \(\mathscr {A}_4'\), we use the relation

$$\begin{aligned} e^{i\tau Q}=\frac{1}{1+12i\tau \chi (\chi +1)(1-\zeta )}\partial _\chi ((\chi +1) e^{i\tau Q}). \end{aligned}$$

Let

$$\begin{aligned} A{} & {} =1+4i\tau \mu ^2(1-\lambda )=1+i\tau (3\zeta +\chi -2)^2(\zeta -\chi )/2, \quad \\ D{} & {} = 1+12i\tau \chi (\chi +1)(1-\zeta ). \end{aligned}$$

The integration by parts yields

$$\begin{aligned} J_\flat (\mathscr {A}_4')&= \iint _{\mathscr {A}_4'} e^{i\tau Q}\frac{ \mu \partial _1 F}{A}\textrm{d}\lambda \textrm{d}\mu = \iint _{\mathscr {A}_4^\flat } \frac{1}{D} \partial _{\chi } ((\chi +1) e^{i\tau Q}) \frac{\mu }{2A} \partial _1 F \textrm{d}\zeta \textrm{d}\chi \\ {}&= - \iint _{\mathscr {A}_4^\flat } (\chi +1) e^{i\tau Q} \left[ - \partial _{12} F+ \partial _{13} F \right] \frac{\mu }{2AD} \textrm{d}\chi \textrm{d}\zeta \\ {}&\qquad - \iint _{\mathscr {A}_4^\flat } (\chi +1) e^{i\tau Q} \frac{\partial _{1} F}{2AD} \left( 1 - \mu \left( \frac{\partial _\chi A}{A} + \frac{\partial _\chi D}{D} \right) \right) \textrm{d}\chi \textrm{d}\zeta \\ {}&\qquad - \iint _{\partial \mathscr {A}_4^\flat } (\chi +1) e^{i\tau Q} \frac{\chi \mu }{2AD} \partial _1 F \textrm{d}\sigma (\zeta ,\chi ). \end{aligned}$$

On \(\mathscr {A}_4^\flat \), \(1-\zeta \), \(\zeta -\chi \) are bounded above and below. Moreover, denoting \(\tilde{\chi }= \chi +1\), \(\tilde{\zeta }= 1-\zeta \), there holds

$$\begin{aligned} A \gtrsim 1 + \tau \mu ^2, \quad D \gtrsim 1 + \tau |\tilde{\chi }\tilde{\zeta }|, \quad |\partial _\chi A| \lesssim \tau |\mu |, \quad |\partial _{\chi } D| \lesssim \tau |\tilde{\zeta }|. \end{aligned}$$

After we perform once again the change of variable \((\mu ,\chi ) \rightarrow (\tilde{\zeta }, \tilde{\chi })\) with \(\tilde{\zeta }= 1-\zeta = \frac{2\mu -\tilde{\chi }}{3}\)), we find

$$\begin{aligned}&\ |J_\flat (\mathscr {A}_4')| \lesssim \iint _{\mathscr {A}_4^\flat } \frac{|\tilde{\chi }\mu |}{(1+\tau \mu ^2)(1+ \tau |\tilde{\chi }\tilde{\zeta }|)} (|\partial _{12} F|+ |\partial _{13} F) \textrm{d}\tilde{\chi }\textrm{d}\mu \end{aligned}$$
(A.11)
$$\begin{aligned}&\quad + \iint _{\mathscr {A}_4^\flat } \frac{|\tilde{\chi }|}{(1+\tau \mu ^2)(1+ \tau |\tilde{\chi }\tilde{\zeta }|)} \left( 1 + \frac{\tau \mu ^2}{1+\tau \mu ^2} + \frac{\tau |\mu \tilde{\zeta }|}{1+\tau |\tilde{\chi }\tilde{\zeta }|} \right) |\partial _{1} F| \textrm{d}\tilde{\chi }\textrm{d}\mu \end{aligned}$$
(A.12)
$$\begin{aligned}&\quad + \int _{\partial \mathscr {A}_4^\flat } \frac{|\tilde{\chi }\mu |}{(1+\tau \mu ^2)(1+ \tau |\tilde{\chi }\tilde{\zeta }|)} |\partial _1 F| \textrm{d}\sigma (\tilde{\chi },\mu ). \end{aligned}$$
(A.13)

As in 6.1),

$$\begin{aligned} \left( \iint (|\partial _{12} F|+ |\partial _{13} F)^2 \textrm{d}\tilde{\chi }\textrm{d}\mu \right) ^{1/2}&\lesssim \tau ^{1/3}, \\ \Vert \partial _1 F \Vert _{L^\infty _{\tilde{\chi }} L^2_\mu }, \Vert \partial _1 F \Vert _{L^\infty _\mu L^2_{\tilde{\chi }}}, \Vert \partial _1 F \Vert _{L^2(\mathscr {A}_4^\flat )}&\lesssim \tau ^{1/6}. \end{aligned}$$

Hence, rescaling \(\check{\chi }= \sqrt{\tau } \tilde{\chi }\), \(\check{\mu }= \sqrt{\tau } \mu \) and using

$$\begin{aligned} \int \frac{\check{\chi }^2 \textrm{d}\check{\chi }}{(1+|\check{\chi }(2\check{\mu }- \check{\chi })|)^2}{} & {} \lesssim \int _{|2 \check{\mu }- \check{\chi }| \leqq 1/|\check{\mu }|} \check{\chi }^2 \textrm{d}\check{\chi }+ \int _{ |2 \check{\mu }- \check{\chi }| \geqq 1/|\check{\mu }|} \frac{\textrm{d}\check{\chi }}{(2\check{\mu }- \check{\chi })^2}\nonumber \\{} & {} \lesssim |\check{\mu }|, \end{aligned}$$
(A.14)

we can bound the integral in (A.11) by

$$\begin{aligned}&( \Vert \partial _{12} F \Vert _{L^2} + \Vert \partial _{13} F \Vert _{L^2} ) |\tau |^{-3/2} \left( \iint _{| \check{\mu }|, |\check{\chi }| \leqq 10 \sqrt{\tau }} \left( \frac{|\check{\chi }\check{\mu }|}{(1+ \check{\mu }^2)(1+|\check{\chi }(2 \check{\mu }- \check{\chi })|)} \right) ^2 d \check{\chi }\textrm{d}\check{\mu }\right) ^{1/2}\\ {}&\lesssim \tau ^{1/3-3/2}\left( \int _{|\check{\mu }|\leqq 10\sqrt{\tau }} \frac{|\check{\mu }|^3}{(1+\check{\mu }^2)^2}\textrm{d}\check{\mu }\right) ^{1/2} \lesssim \tau ^{-7/6} \ln (1+\tau ) \lesssim \tau ^{-1}. \end{aligned}$$

We now consider the integral in (A.12). For the first two terms, we do Cauchy-Schwarz in \(\tilde{\chi }\) and rescale as before \(\check{\chi }= \sqrt{\tau } \tilde{\chi }\) and \(\check{\mu }= \sqrt{\tau } \mu \), so as to get the bound

$$\begin{aligned}&\Vert \partial _1 F \Vert _{L^\infty _{\mu } L^2_{\tilde{\chi }}} \int _{|\mu | \leqq 1} \left( \int _{|\tilde{\chi }| \leqq 1} \frac{\tilde{\chi }^2d \tilde{\chi }}{(1+|\tau \tilde{\chi }\tilde{\zeta }|)^2} \right) ^{1/2} \frac{\textrm{d}\mu }{(1+\tau \mu ^2)} \\ {}&\lesssim \tau ^{1/6-1/2-3/4} \int _{|\check{\mu }| \leqq 10 \sqrt{\tau }} \left( \int _{|\check{\chi }| \leqq 10 \sqrt{\tau }} \frac{\check{\chi }^2 d \check{\chi }}{(1+| \check{\chi }\check{\zeta }|)^2} \right) ^{1/2} \frac{\textrm{d}\check{\mu }}{(1+\check{\mu }^2)} \lesssim \tau ^{-13/12}, \end{aligned}$$

where we used once again (A.14).

For the third term of the integral in (A.12), we do Cauchy-Schwarz in \(\tilde{\chi }\) and rescale again \(\check{\chi }= \sqrt{\tau } \tilde{\chi }\), \(\check{\mu }= \sqrt{\tau } \mu \), and get the bound

$$\begin{aligned}&\Vert \partial _1 F \Vert _{L^\infty _{\mu } L^2_{\tilde{\chi }}} \int _{|\mu | \leqq 1} \left( \int _{|\tilde{\chi }| \leqq 1} \frac{d \tilde{\chi }}{(1+|\tau \tilde{\chi }\tilde{\zeta }|)^2} \right) ^{1/2} \frac{\mu \textrm{d}\mu }{(1+\tau \mu ^2)} \\ {}&\lesssim \tau ^{1/6- 1-1/4} \int _{|\check{\mu }| \leqq 10 \sqrt{\tau }} \left( \int _{|\check{\chi }| \leqq 10 \sqrt{\tau }} \frac{d \check{\chi }}{(1+| \check{\chi }\check{\zeta }|)^2} \right) ^{1/2} \frac{|\check{\mu }|\textrm{d}\check{\mu }}{(1+\check{\mu }^2)} \lesssim \tau ^{-13/12}. \end{aligned}$$

Indeed, we show that the integral \(\lesssim 1\) as follows. We can assume \(| \check{\mu }| \geqq 10\). Then, for fixed \(\check{\mu }\),

$$\begin{aligned} \int _{10 \leqq |\check{\chi }|} \frac{d \check{\chi }}{(1+| \check{\chi }(2\check{\mu }- \check{\chi })|)^2}&\lesssim \int _{|2 \check{\mu }- \check{\chi }| \leqq 1/|\check{\mu }|} \textrm{d}\check{\chi }\\&\qquad + \int _{1/|\check{\mu }| \leqq |2 \check{\mu }- \check{\chi }| \leqq |\check{\mu }|/100 } \frac{\textrm{d}\check{\chi }}{ \check{\mu }^2 (2\check{\mu }- \check{\chi })^2} \\ {}&\qquad + \int _{|2 \check{\mu }- \check{\chi }| \geqq |\check{\mu }|/100 } \frac{\textrm{d}\check{\chi }}{(2\check{\mu }- \check{\chi })^2} \lesssim \frac{1}{|\check{\mu }|}. \end{aligned}$$

Hence

$$\begin{aligned} \int _{|\check{\mu }| \leqq 10 \sqrt{\tau }} \left( \int _{|\check{\chi }| \leqq 10 \sqrt{\tau }} \frac{d \check{\chi }}{(1+| \check{\chi }\check{\zeta }|)^2} \right) ^{1/2} \frac{\check{\mu }\textrm{d}\check{\mu }}{(1+\check{\mu }^2)} \lesssim \int \frac{|\check{\mu }|^{1/2} \textrm{d}\check{\mu }}{(1+\check{\mu }^2)} \lesssim 1. \end{aligned}$$

Finally, we consider the boundary term (A.13): there, \(\tilde{\chi }\), \(\mu \) and \(\tilde{\zeta }\) are proportional; by Cauchy-Schwarz inequality, this term is bounded up to a constant by

$$\begin{aligned} \Vert \partial _1 F \Vert _{L^2(\mathscr {A}_4^\flat )} \left( \int _{[-1,1]} \frac{\rho ^4 \textrm{d}\rho }{(1+\tau \rho ^2)^4} \right) ^{1/2} \lesssim \tau ^{1/6-5/4} \lesssim \tau ^{-13/12}. \end{aligned}$$

7) In summary, (recalling \(\tau \geqq 1\)), we proved in 5) and 6) that

$$\begin{aligned} |J_\sharp (\mathscr {A}_3)|+ |J_\flat (\mathscr {A}_3)| + |J_\sharp (\mathscr {A}_4)| + |J_\flat (\mathscr {A}_4)| \lesssim \tau ^{-1}. \end{aligned}$$

Hence \(|J(\mathscr {A}_3)| + |J(\mathscr {A}_4)| \lesssim \tau ^{-1} \xi ^3 \lesssim 1/t\). Summing up with the bounds in 1)-3), we infer that \(|J(\mathscr {A})| \lesssim 1/t\), and the same estimate holds for \(J(\mathbb {R}^2)\), as claimed. \(\square \)

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Correia, S., Côte, R. Perturbation at Blow-Up Time of Self-Similar Solutions for the Modified Korteweg–de Vries Equation. Arch Rational Mech Anal 248, 25 (2024). https://doi.org/10.1007/s00205-024-01969-x

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1007/s00205-024-01969-x

Navigation