betaFeedback

WorksMathematics

A Method for Finding Exact Solutions to the 2D and 3D Euler-Boussinesq Equations in Lagrangian Coordinates

Abstract

We study the Boussinesq approximation for the incompressible Euler equations using Lagrangian description. The conditions for the Lagrangian fluid map are derived in this setting, and a general method is presented to find exact fluid flows in both the two-dimensional and the three-dimensional case. There is a vast amount of solutions obtainable with this method and we can only showcase a handful of interesting examples here, including a Gerstner type solution to the two-dimensional Euler-Boussinesq equations. In two earlier papers we used the same method to find exact Lagrangian solutions to the homogeneous Euler equations, and this paper serves as an example of how these same ideas can be extended to provide solutions also to related, more involved models.

The first author was supported by Finnish Cultural Foundation.

Introduction

We continue our investigations of finding explicit solutions to the incompressible Euler equations in the Lagrangian framework. In [12, 13] we considered the homogeneous case while here we treat the Boussinesq approximation of the heterogeneous case. The idea behind the Boussinesq approximation is that the density of the fluid does not fluctuate very much around the mean value, so that one supposes that the density is constant, except in the term that represents the gravity. There are many physical situations where this simplification is justified. The general introduction to this type of models can be found in [11], while in regard to the Lagrangian description in general we refer to [3].

One of the first explicit solutions to the homogeneous Euler equations in Lagrangian formulation is due to Gerstner [9], and it turns out that solutions of the Gerstner type show up in many related models. For example the Gerstner solution plays a prominent role also in the heterogeneous Euler equations, even without using the Boussinesq approximation. In [8] it was shown that the Gerstner wave can also be a barotropic flow, i.e. a flow where pressure and density are functions of each other, see also a modern exposition by Stuhlmeier [14]. Gerstner type solutions are also relevant in many physical models that are somehow modifications of the Euler equations. For example Constantin [4] found a barotropic Gerstner type exact solution to the equatorial water wave equations with the beta-plane approximation of the Coriolis effect. For more papers on the applicability of the Gerstner waves see for example [10, 7, 6, 5] and especially the survey articles [2, 15]. In [16] there is also a Gerstner type solution to the first and second terms of the asymptotic expansion to the heterogeneous Euler equations. One remarkable property of Gerstner type solutions that makes them popular, is that they can in many cases be interpreted as free surface solutions, or in other words one can model the interface of two different fluids. Apparently there are no other known explicit solutions with this property.

Many explicit solutions to stratified fluid flows are shear flows, which are flows with two-dimensional horizontal motion that only depends on height. Most of the studies use the Eulerian framework. As for the Lagrangian description, Yakubovich and Shrira [17] found solutions with columnar motion using the Boussinesq approximation, but it seems that elsewhere the Eulerian framework is used.

As stated above there are numerous models for geophysical fluid flows, depending on the particular context. We have chosen to analyze the Euler–Boussinesq model in the present article. For a thorough discussion of this model and its applicability we refer to [11]. As in [12, 13], we use a separation of variables method to compute our solutions, that is, we search for fluid particle maps that can be expressed as the product of a time-only dependent matrix and a spatial-only dependent vector. This approach leads for example to a two-dimensional Gerstner type flow (23), as well as a plethora of other solutions both in the 2D and the 3D setting. Solution (23) does not satisfy the free boundary condition like most Gerstner type flows, but it still gives valid internal flows, which are an important application of the Euler–Boussinesq equations. Note in particular that our approach is not restricted to the Euler equations and Euler–Boussinesq equations, and we are confident that it could be used profitably to analyze other related models. For example Abrashkin [1] derived the governing Lagrangian equations for equatorial beta-plane flows, allowing anyone to readily attempt the separation of variables strategy that we use below. Also the Gerstner type solutions in all those different models discussed above could have been found by our method.

This paper is organized as follows. In Section 2 we give basic definitions and specify the model which will be analyzed. In Section 3 we discuss columnar flows and generalize one family of solutions which was given in [17]. In Section 4 we find exact solutions to the two-dimensional Euler–Boussinesq equations while utilizing the similarities to the homogeneous situation [12]. In Section 5 we consider the three-dimensional case and use the framework outlined in [13] to find solutions to the Euler–Boussinesq equations.

In some cases we are able to find explicit periodic and nonperiodic solutions. In other cases one can show that a solution exists for all times for convenient parameter values. Finally there are cases where all one can say that a local solution is well-defined. We allow for both stably and unstably stratified fluids as both types of situations can often be covered by a single formula, though when we give specific examples we concentrate on the stably stratified case. Some solutions that we find are neither stably nor unstably stratified. We also note that it has not been possible to explore all different cases in the present article.

Euler–Boussinesq equations

Notation

Let A:R→Rn×mA:\mathbb{R}\to\mathbb{R}^{n\times m} and v:Rn→Rmv:\mathbb{R}^{n}\to\mathbb{R}^{m}. We denote the columns of AA by AiA_i and its entries by aija_{ij}. Now the minors of AA are denoted by

pij=det⁡(Ai,Aj),if n=2,pijk=det⁡(Ai,Aj,Ak),if n=3.(1)\begin{aligned} p_{ij} &= \det(A_i,A_j), \quad\text{if } n=2,\\ p_{ijk} &= \det(A_i,A_j,A_k), \quad\text{if } n=3. \tag*{(1)} \end{aligned}

Similarly the minors of dvdv are

qij=det⁡(∇vi,∇vj),if n=2,qijk=det⁡(∇vi,∇vj,∇vk),if n=3.(2)\begin{aligned} q_{ij} &= \det(\nabla v^i,\nabla v^j), \quad\text{if } n=2,\\ q_{ijk} &= \det(\nabla v^i,\nabla v^j,\nabla v^k), \quad\text{if } n=3. \tag*{(2)} \end{aligned}

It will also be useful to define

Qij=⟨Ai′,Aj⟩−⟨Aj′,Ai⟩,Gij=∇vi×∇vj,(3)\begin{aligned} Q_{ij} &= \langle A'_i,A_j\rangle-\langle A'_j,A_i\rangle,\\ G_{ij} &= \nabla v^i\times\nabla v^j, \tag*{(3)} \end{aligned}

where Ai′A'_i is the (time) derivative of AiA_i and GijG_{ij} is useful only when n=3n=3. For derivatives of vv we use multiindices so that for example

v2011=∂3v1∂z12∂z3.v^{1}_{201}=\frac{\partial^{3}v^{1}}{\partial z_{1}^{2}\partial z_{3}}.

At times we will also say that functions v1v^1 and v2v^2 are an anti-Cauchy–Riemann pair, or an anti-CR pair, if

{v101+v012=0v011−v102=0\begin{cases} v^1_{10}+v^2_{01}=0\\ v^1_{01}-v^2_{10}=0 \end{cases}

Model

The nn-dimensional heterogeneous incompressible Euler equations are given by the system

∇⋅u=0ρ~(ut+u∇u+gen)+∇p=0ρ~t+⟨u,∇ρ~⟩=0.(4)\begin{aligned} \nabla\cdot u&=0\\ \widetilde{\rho}(u_t+u\nabla u+g e_n)+\nabla p&=0\\ \widetilde{\rho}_t+\langle u,\nabla\widetilde{\rho}\rangle&=0. \tag*{(4)} \end{aligned}

Here uu is the velocity field, ρ~\widetilde{\rho} is the density, gg is the acceleration due to gravity, and ene_n is the vertical unit vector parallel to gravity. Let us write these equations in the Lagrangian framework. Let D⊂RnD\subset\mathbb{R}^n be a domain and let us consider a family of diffeomorphisms φt:D→Ωt=φt(D)\varphi^t:D\to\Omega_t=\varphi^t(D). The coordinates in DD are denoted by zz and in Ωt\Omega_t by xx. We can also define

φ:D×R→Rn,φ(z,t)=φt(z).\varphi:D\times\mathbb{R}\to\mathbb{R}^n,\qquad\varphi(z,t)=\varphi^t(z).

Now given such φ\varphi we can define the associated vector field uu by the formula

∂∂tφ(z,t)=u(φ(z,t),t).(5)\frac{\partial}{\partial t}\varphi(z,t)=u(\varphi(z,t),t). \tag*{(5)}

Then (u,ρ^,p^)(u,\widehat{\rho},\widehat{p}) solves (4) if det⁡(dφ)≠0\det(d\varphi)\ne0 and

ddtdet⁡(dφ)=0(6)\frac{d}{dt}\det(d\varphi)=0 \tag*{(6)}
ρ^(dφT(φ′′+gen))+∇p^=0(7)\widehat{\rho}\left(d\varphi^T(\varphi''+g e_n)\right)+\nabla\widehat{p}=0 \tag*{(7)}
ddtρ^=0.(8)\frac{d}{dt}\widehat{\rho}=0. \tag*{(8)}

where p^=p∘φ\widehat{p}=p\circ\varphi and ρ^=ρ~∘φ\widehat{\rho}=\widetilde{\rho}\circ\varphi. Hence ρ^\widehat{\rho} is a function of spatial variables only, which is one great advantage of using the Lagrangian description.1 Typically one cannot explicitly recover uu from φ\varphi, since it requires computing the inverse of φt\varphi^t. One exception to this is obtained when φ=A(t)z\varphi=A(t)z for some square matrix AA, yielding u=A′A−1xu=A'A^{-1}x.

The standard way to apply the Boussinesq approximation is to assume that ρ^\widehat{\rho} is constant in every term except when it is multiplied by gg. Thus the Boussinesq approximation takes into account how density variations affect buoyancy. This would mean that Newton’s second law (7) would be replaced by the equation

dφT(ρ‾φ′′+ρ^gen)+∇p^=0,(9)d\varphi^T(\overline{\rho}\varphi''+\widehat{\rho}g e_n)+\nabla\widehat{p}=0, \tag*{(9)}

where ρ‾\overline{\rho} is the average density. However, as was shown in [17], we can make the model slightly more accurate without making the equations any more difficult to study. Supposing that n=3n=3 and taking the curl of (7), we obtain

∇ρ^×(dφTφ′′)+ρ^(∑i=13∇φi′′×∇φi)+g∇ρ^×∇φ3=0.(10)\nabla\widehat{\rho}\times(d\varphi^T\varphi'')+\widehat{\rho}\left(\sum_{i=1}^{3}\nabla\varphi_i''\times\nabla\varphi_i\right)+g\nabla\widehat{\rho}\times\nabla\varphi_3=0. \tag*{(10)}

Now assuming only ∇ρ^×(dφTφ′′)=0\nabla\widehat{\rho}\times(d\varphi^T\varphi'')=0 in (10), dividing by ρ^\widehat{\rho} and defining ρ=gln⁡ρ^\rho=g\ln\widehat{\rho} we obtain

∑i=13∇φi′′×∇φi+∇ρ×∇φ3=0.(11)\sum_{i=1}^{3}\nabla\varphi_i''\times\nabla\varphi_i+\nabla\rho\times\nabla\varphi_3=0. \tag*{(11)}

If we had used (9), we would have obtained the same equation but with ρ\rho defined as ρ=gρ^/ρ‾\rho=g\widehat{\rho}/\overline{\rho} instead of ρ=gln⁡ρ^\rho=g\ln\widehat{\rho}. Hence it should be kept in mind that ρ\rho is not really density, but the density ρ^\widehat{\rho} can be recovered from ρ\rho using either ρ=gρ^/ρ‾\rho=g\widehat{\rho}/\overline{\rho} or ρ=gln⁡ρ^\rho=g\ln\widehat{\rho}. Integrating the left hand side of (11) with respect to tt we obtain the equivalents of what are the Cauchy invariants for the Euler equations:

h=(h1,h2,h3)=∑i=13∇φi′×∇φi+∇ρ×∇∫φ3 dt.(12)h = (h^1,h^2,h^3) = \sum_{i=1}^{3} \nabla\varphi'_i \times\nabla\varphi_i + \nabla\rho\times\nabla\int\varphi_3\,dt . \tag*{(12)}

This integrated form is often useful since it removes the second-order dependence of φ1\varphi_1 and φ2\varphi_2. But calling the components of hh the ”Cauchy invariants of the Euler–Boussinesq equations” is perhaps a bit of a stretch since the time integral of φ3\varphi_3 produces an arbitrary function of zz and so there is no canonical way to choose hh.

In the two-dimensional case hh is just a scalar and it is convenient to write it in the following form:

h=∑i=12det⁡(∇φi′,∇φi)+det⁡(∇ρ,∇∫φ2 dt).(13)h = \sum_{i=1}^{2} \det\left(\nabla\varphi'_i,\nabla\varphi_i\right) + \det\left(\nabla\rho,\nabla\int\varphi_2\,dt\right). \tag*{(13)}

We have now established the conditions for a solution to the 2D and 3D Euler–Boussinesq equations:

References

  1. [1]A. A. Abrashkin. Generalization of cauchy invariants for equatorial β-plane flows. Deep-Sea Res. II, 160:3–6, 2019.DOI
  2. [2]A. A. Abrashkin and E. N. Pelinovsky. Gerstner waves and their generalizations in hydrodynamics and geophysics. Phys. Usp., 65:453–467, 2021.
  3. [3]A. Bennett. Lagrangian fluid dynamics. Cambridge Monographs on Mechanics. Cambridge University Press, Cambridge, 2006.DOI
  4. [4]A. Constantin. An exact solution for equatorially trapped waves. J. Geophys. Res., Oceans, 117(C05029), 2012.DOI
  5. [5]A. Constantin. Comments on: Nonlinear wind-drift ocean currents in arctic regions. Geophys. Astrophys. Fluid Dyn, 116(2):116–121, 2022.DOI
  6. [6]A. Constantin. Nonlinear wind-drift ocean currents in arctic regions. Geophys. Astrophys. Fluid Dyn, 116(2):101–115, 2022.DOI
  7. [7]A. Constantin and S. G. Monismith. Gerstner waves in the presence of mean currents and rotation. J. Fluid Mech., 820:511–528, 2017.DOI
  8. [8]M.-L. Dubreil-Jacotin. Sur les ondes de type permanent dans les liquides heterogenes. Atti Accad. Naz. Lincei, 15:814–819, 1932.
  9. [9]F. Gerstner. Theorie der Wellen samt einer daraus abgeleiteten Theorie der Deichprofile. Ann. Phys., 2:412–445, 1809.
  10. [10]D. Henry. An exact solution for equatorial geophysical water waves with an underlying current. Eur. J. Mech. B Fluids, 38:18–21, 2013.DOI
  11. [11]A. Majda. Introduction to PDEs and Waves for the Atmosphere and Ocean, volume 9 of Courant Lecture Notes. American Mathematical Society, New York, 2003.
  12. [12]T. Saleva and J. Tuomela. On the explicit solutions of separation of variables type for the incompressible 2D Euler equations. J. Math. Fluid Mech., 23(39), 2021.DOI
  13. [13]T. Saleva and J. Tuomela. Explicit solutions to the 3D incompressible Euler equations in Lagrangian formulation. J. Math. Fluid Mech., 24(98), 2022.DOI
  14. [14]R. Stuhlmeier. On edge waves in stratified water along a sloping beach. J. Nonlinear Math. Phys., 18:127–137, 2011.DOI
  15. [15]R. Stuhlmeier. Gerstner’s water wave and mass transport. J. Math. Fluid Mech., 17:761–767, 2015.DOI
  16. [16]J. E. H. Weber. A lagrangian study of internal gerstner- and stokes-type gravity waves. Wave Motion, 88:257–264, 2019.DOI
  17. [17]E. I. Yakubovich and V. I. Shrira. Non-steady columnar motions in rotating stratified boussinesq fluids: exact lagrangian and eulerian description. J. Fluid Mech., 691:417–439, 2012.DOI

Paper details

vlnc:W_h2ygzn2f
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents