Hostname: page-component-7b7567c8c-fvjsw Total loading time: 0 Render date: 2026-11-16T09:54:50.168Z Has data issue: false hasContentIssue false

High-frequency instabilities of Stokes waves

Published online by Cambridge University Press:  28 February 2022

Ryan P. Creedon*
Affiliation:
Department of Applied Mathematics, University of Washington, Seattle, WA 98195, USA
Bernard Deconinck
Affiliation:
Department of Applied Mathematics, University of Washington, Seattle, WA 98195, USA
Olga Trichtchenko
Affiliation:
Department of Physics and Astronomy, The University of Western Ontario, London, ON N6A 3K7, Canada
*
†Email address for correspondence: creedon@uw.edu

Abstract

Euler's equations govern the behaviour of gravity waves on the surface of an incompressible, inviscid and irrotational fluid of arbitrary depth. We investigate the spectral stability of sufficiently small-amplitude, one-dimensional Stokes waves, i.e. periodic gravity waves of permanent form and constant velocity, in both finite and infinite depth. We develop a perturbation method to describe the first few high-frequency instabilities away from the origin, present in the spectrum of the linearization about the small-amplitude Stokes waves. Asymptotic and numerical computations of these instabilities are compared for the first time, with excellent agreement.

Information

Type
JFM Papers
Copyright
© The Author(s), 2022. Published by Cambridge University Press
Figure 0

Figure 1.A schematic of 1-D gravity waves in finite depth $h$. In this work, the surface displacement $\eta$ and velocity field ${\boldsymbol {u}} = (\phi _x,\phi _z)$ are $2{\rm \pi} /\kappa$-periodic in the $x$-direction.

Figure 1

Figure 2.The stability spectrum of a $2{\rm \pi}$-periodic Stokes wave with amplitude $\varepsilon = 0.01$ and (a) $h = \infty$, (b) $h=1.5$, (c) $h=1.4$ and (d) $h=1$. The Benjamin–Feir figure eight is coloured blue. The high-frequency isolas are coloured red. Purely imaginary eigenvalues are coloured black. A zoom-in of the Benjamin–Feir and high-frequency instabilities are inlaid in the top, left plot.

Figure 2

Figure 3.(a) The high-frequency isola closest to the origin for a $2{\rm \pi}$-periodic Stokes wave in depth $h = 1.5$ with amplitude $\varepsilon = 2 \times 10^{-3}$ (orange), $\varepsilon = 4 \times 10^{-3}$ (red), $\varepsilon = 6 \times 10^{-3}$ (magenta), $\varepsilon = 8 \times 10^{-3}$ (purple) and $\varepsilon = 10^{-2}$ (blue). The imaginary axis is recentred to show the drift of the isola from the collided eigenvalues at $\lambda _0$. The isolas are computed using the perturbation method developed in this paper. (b) The interval of Floquet exponents that parameterizes the isola closest to the origin in depth $h=1.5$ as a function of the amplitude. The solid black lines indicate the boundaries of this interval, while the dashed black line gives the Floquet exponent corresponding to the most unstable eigenvalue on the isola. The coloured lines give the Floquet exponents corresponding to the similarly coloured isolas in the left figure. The Floquet axis is recentred to show the drift of the parameterizing interval from the Floquet exponent $\mu _0$ that corresponds to the collided eigenvalues. The parameterizing interval is also computed using the perturbation method in this paper.

Figure 3

Figure 4.(a) The amplitude vs velocity bifurcation diagram of $2{\rm \pi}$-periodic Stokes waves when $\alpha = 1$ (dashed line), $\alpha = 1.5$ (dotted line), $\alpha = 2$ (dot-dashed line) and $\alpha = \infty$ (solid line), according to our ${O}(\varepsilon ^{4})$ asymptotic calculations. The zeroth-order contribution $c_0$ is removed for better visibility. The numerical results are given by the coloured dots. Red dots correspond to $\alpha = 1$, magenta dots correspond to $\alpha = 1.5$, purple dots correspond to $\alpha = 2$ and blue dots correspond to $\alpha = \infty$. (b) Expansions of $\eta _S/\varepsilon$ to ${O}(\varepsilon ^{4})$ with $\varepsilon = 0.1$ for $\alpha = 1, 1.5, 2$ and $\infty$ (arranged from top to bottom using the same line styles as in panel a). A sampling of the numerical results is given by the coloured dots using the same colour scheme as in panel (a).

Figure 4

Figure 5.(a) A plot of $\mathcal {S}_2$ vs $\alpha$ (solid red). The zero of $\mathcal {S}_2$ for $\alpha > 0$ is $\alpha _1 = 1.8494040837\dots$ (gold star). (b) The real part $\lambda _{r,*}$ of the most unstable eigenvalue on the $p=2$ isola as a function of $\alpha$ according to our asymptotic calculations (solid red). The real part of the eigenvalue is normalized by $\varepsilon ^{2}$ for better visibility. We zoom-in around $\alpha = \alpha _1$ (gold star) in the inlay. The real part of the most unstable eigenvalue on the isola vanishes as $\alpha \rightarrow \alpha _1$ according to our asymptotic calculations, which agrees with our numerical results using the FFH method with $\varepsilon = 0.01$ (blue dots).

Figure 5

Figure 6.(a) The $p=2$ isola with $\alpha = 1.5$ and $\varepsilon = 0.01$. The most unstable eigenvalue $\lambda _{*}$ is removed from the imaginary axis for better visibility. The solid red curve is the ellipse obtained by our asymptotic calculations. The blue dots are a subset of eigenvalues from the numerically computed isola using the FFH method. (b) The Floquet parameterization of the real (blue) and imaginary (red) parts of the isola on the left. The most unstable eigenvalue $\lambda _*$ and its corresponding Floquet exponent $\mu _*$ are removed from the imaginary and Floquet axes, respectively, for better visibility. The solid curves are our asymptotic results. The coloured dots are our numerical results using the FFH method. (c,d) Same with $\alpha = 1$.

Figure 6

Figure 7.(a) The interval of Floquet exponents parameterizing the $p=2$ isola as a function of $\varepsilon$ for $\alpha = 1.5$. The zeroth-order correction of the Floquet exponent is removed from the Floquet axis for better visibility. The solid blue curves are the boundaries of this interval according to our asymptotic calculations. The blue dots are the boundaries computed numerically by the FFH method. The solid red curve gives the Floquet exponent of the most unstable eigenvalue on the isola according to our asymptotic calculations. The red dots are the Floquet exponent of the most unstable eigenvalue as computed by the FFH method. (b) The real (blue) and imaginary (red) parts of the most unstable eigenvalue of the $p=2$ isola with $\alpha = 1.5$ as a function of $\varepsilon$. The zeroth-order correction of the eigenvalue is removed from the imaginary axis for better visibility. The solid curves are our asymptotic calculations. The coloured dots are our numerical results using the FFH method. (c,d) Same with $\alpha = 1$.

Figure 7

Figure 8.A plot of $M_{2,+}M_{2,-}$ vs $\alpha$ (solid red). We find $M_{2,+}M_{2,-} < 0$ only when $\alpha \in (0.8643029367\dots, 1.0080416077\dots )$ (solid black). If $M_{2,+}M_{2,-} < 0$, the boundaries of the Floquet exponents parameterizing the $p=2$ isola have opposite concavities at $\varepsilon =0$. Only then does $\mu _0$ remain in the interval of Floquet exponents parameterizing the isola for positive $\varepsilon$.

Figure 8

Figure 9.(a) The $p=2$ isola with $\alpha =\infty$ and $\varepsilon = 0.01$. The most unstable eigenvalue $\lambda _{*}$ is removed from the imaginary axis for better visibility. The solid red curve is the ellipse obtained by our asymptotic calculations. The blue dots are a subset of eigenvalues from the numerically computed isola using the FFH method. (b) The Floquet parameterization of the real (blue) and imaginary (red) parts of the isola. The most unstable eigenvalue $\lambda _*$ and its corresponding Floquet exponent $\mu _*$ are removed from the imaginary and Floquet axes, respectively, for better visibility. The solid curves are our asymptotic results. The coloured dots are our numerical results using the FFH method.

Figure 9

Figure 10.(a) The interval of Floquet exponents parameterizing the $p=2$ isola as a function of $\varepsilon$ for $\alpha = \infty$. The most unstable Floquet exponent $\mu _*$ is removed from the Floquet axis for better visibility. The solid blue curves are the boundaries of this interval according to our asymptotic calculations. The blue dots are the boundaries computed numerically by the FFH method. The solid red curve gives the Floquet exponent of the most unstable eigenvalue on the isola according to our asymptotic calculations. The red dots are the Floquet exponent of the most unstable eigenvalue as computed by the FFH method. (b) The real (blue) and imaginary (red) parts of the most unstable eigenvalue of the $p=2$ isola with $\alpha = \infty$ as a function of $\varepsilon$. The zeroth-order correction of the eigenvalue is removed from the imaginary axis for better visibility. The solid curves are our asymptotic calculations. The coloured dots are our numerical results using the FFH method.

Figure 10

Figure 11.(a) A plot of $\mathcal {S}_3$ vs $\alpha$ (solid red). The zero of $\mathcal {S}_3$ for $\alpha > 0$ is $\alpha _2 = 0.8206431673\dots$ (gold star). (b) The real part $\lambda _{r,*}$ of the most unstable eigenvalue on the $p=3$ isola as a function of $\alpha$ according to our asymptotic calculations (solid red). The real part of the eigenvalue is normalized by $\varepsilon ^{3}$ for better visibility. We zoom-in around $\alpha = \alpha _2$ (gold star) in the inlay. The real part of the most unstable eigenvalue on the isola vanishes as $\alpha \rightarrow \alpha _2$ according to our asymptotic calculations, which agrees with our numerical results using the FFH method with $\varepsilon = 0.01$ (blue dots).

Figure 11

Figure 12.(a) The $p=3$ isola with $\alpha = 1.5$ and $\varepsilon = 0.01$. The most unstable eigenvalue $\lambda _{*}$ is removed from the imaginary axis for better visibility. The solid and dashed red curves are the ellipses obtained by our ${O}(\varepsilon ^{4})$ and ${O}(\varepsilon ^{3})$ asymptotic calculations, respectively. The blue dots are a subset of eigenvalues from the numerically computed isola using the FFH method. (b) The Floquet parameterization of the real (blue) and imaginary (red) parts of the isola on the left. The most unstable eigenvalue $\lambda _*$ and its corresponding Floquet exponent $\mu _*$ are removed from the imaginary and Floquet axes, respectively, for better visibility. The solid teal and orange curves are our asymptotic results for the real and imaginary parts of the Floquet parameterization, respectively, to ${O}(\varepsilon ^{4})$. The dashed blue and red curves are the same results to ${O}(\varepsilon ^{3})$. The blue and red dots are the numerically computed real and imaginary parts of the Floquet parameterization, respectively, using the FFH method. (c,d) Same with $\alpha = 1$.

Figure 12

Figure 13.(a) The interval of Floquet exponents parameterizing the $p=3$ isola as a function of $\varepsilon$ for $\alpha = 1.5$. The most unstable Floquet exponent $\mu _*$ is removed from the Floquet axis for better visibility. The solid and dashed blue curves are the boundaries of this interval according to our ${O}(\varepsilon ^{4})$ and ${O}(\varepsilon ^{3})$ asymptotic calculations, respectively. The blue dots are the boundaries computed numerically by the FFH method. The solid and dashed red curves give the Floquet exponent of the most unstable eigenvalue on the isola according to our ${O}(\varepsilon ^{4})$ and ${O}(\varepsilon ^{3})$ asymptotic calculations, respectively. The red dots are the Floquet exponent of the most unstable eigenvalue as computed by the FFH method. (b) The real (blue) and imaginary (red) parts of the most unstable eigenvalue of the $p=3$ isola with $\alpha = 1.5$ as a function of $\varepsilon$. The zeroth-order correction of the eigenvalue is removed from the imaginary axis for better visibility. The solid teal and orange curves are our asymptotic calculations for the real and imaginary parts of the most unstable eigenvalue to ${O}(\varepsilon ^{4})$, respectively. The dashed blue and red curves are the same results to ${O}(\varepsilon ^{3})$. The blue and red dots are the numerically computed real and imaginary parts of the most unstable eigenvalue using the FFH method. (c,d) Same with $\alpha = 1$.