No label defined (Q4700638)

This monograph presents Hamiltonian formalisms to study the stability of solutions to partial-differential equations describing the evolution of quantities related to two-dimensional fluid flows (the stream function of the Euler equation, the Charney-Hasegawa-Mima equation, the Korteweg-de Vries equation). In the introduction, the author presents ``chaos'' and solitons as two paradigms for understanding turbulent flows: orderered structures arise at the large scale level from a ``chaotic'' small scale level. Next, historical considerations on Arnol'd's work who adapted Hamiltonian stability theory to the infinite-dimensional situation of fluid flow are provided. Finally, an outline of the book is provided. In chapter two, the Hamiltonian dynamics formalism is presented as well as elements for the study of the stability of Hamiltonian systems in finite dimensions. The first-differential of the Hamiltonian has to vanish at the equilibrium and its second-differential has to be definite for stability to hold. In chapter three, the author provides a Hamiltonian formulation of the vorticity equation of the two-dimensional incompressible Euler equation for a fluid in \(\Omega\subset \mathbb{R}^2\) which is open with smooth boundary having \(n\) connected components. It can be written \[ q_t= [q,H]= J \frac{\delta H}{\delta q} \] with Poisson bracket given for \(F\), \(G\) functionals from a space of smooth functions, by \[ [F,G]= \iint_\Omega q\partial \Biggl( \frac{\delta F}{\delta q}, \frac{\delta G}{\delta q} \Biggr) dx dy= \iint_\Omega \frac{\delta F^t}{\delta q} J\frac{\delta G}{\delta q} dx dy \] where \(q= \Delta\Psi\) is the vorticity, \(\Psi\) is the stream function and \[ \partial(A,B)= A_xB_y- A_yB_x, \quad J(.)= -\partial(q,.), \] \[ H= \tfrac 12 \iint_\Omega \nabla\Psi. \nabla\Psi dx dy- \sum_{i=1}^{i=n} \lambda_1 \oint_{\partial\Omega_i} n_i.\nabla \Psi ds. \] The first term is the kinetic energy. \(n_i\) a normal unit vector to the boundary \(\partial\Omega_i\), \(\lambda_i= \Psi_{\partial\Omega_i}\) is seen to be constant and \(\frac{\delta}{\delta q}\) is the Fréchet derivative. It is shown that this Poisson bracket is the transformed version of the canonical Poisson bracket in the Lagrange description of fluid motion. Casimir functionals which commute with all the functionals are not necessarily trivial (i.e. constant) because their Fréchet derivative belongs to \(\ker J\) so that they are functions of \(q\) only. Finally, Noether's theorem is exploited to find integrals in the case of spatial or temporal translational invariance and rotational invariance. In chapter four, the author uses the formalism from the previous chapter and develops the methods from chapter two to study the stability of steady Euler flows. First the stationary solutions are obtained by considering a constrained Hamiltonian which is obtained by adding a Casimir (obtained from the stationarity condition) to \(H\). So one has \(\frac{\delta \widetilde{H}(q_s)} {\delta q}=0\), where \(\widetilde{H}\) is the modified Hamiltonian and \(q_s\) is the stationary vorticity. Linear stability is investigated next by computing the second variation of \(\widetilde{H}\). Sufficient conditions for it to be positive or negative definite allow to show stability. For the proofs, one needs that the second variation of \(\widetilde{H}\) at \(q_s\) does not change with time and to choose suitable norms. The assumptions are based on the boundedness (away from zero) of the derivative at \(q_s\) of a function \(\Phi\) whose existence follows from the stationarity condition. (In this case, \(\Psi_s= \Phi(q_s)\)). In the second case, a Rayleigh-Ritz variational principle is used. This method of Arnol'd (1965) allows to recover the necessary conditions for instability of Rayleigh (1880) and Fjorthoft (1950) which impose constraints on the velocity profile of a parallel shear flow; they arise from perturbing the stationary stream function by a term of the form \(\varphi_0(y) \exp(ik(x-ct))\), substituting this expression in the derived stability equations and after some manipulations, one looks at the real and imaginary parts of the obtained expression, respectively. Nonlinear versions are presented too but now, the nonlinear terms in infinite dimensions may be an obstacle for the definiteness of \(\delta^2 \widetilde{H}(q_s)\). This is repaired by exploiting the convexity of \(\Phi\). Andrew's theorems (1984) show the existence of situations where the framework of Arnold's does not cover more cases than the parallel shear flow configuration of Rayleigh and Fjorthoft. Finally, the cases of parallel shear flows and circular flows are examined from the same point of view but through other variational principles which exploit the specific symmetries of the problem at hand. One considers other invariant functionals which replace the Hamiltonian. Chapter five follows the same lines as the previous one, Now the Charney-Hasegawa-Mima equation is considered \[ (\Delta-1) \Psi_t+ \Psi_x+ \partial (\Psi, \Delta\Psi)=0. \] The term \(\Psi_x- \Psi_t\) has been added. It models large scale dynamics in oceans and the atmosphere (a preoccupation of the author) and \(\Psi\) is the deviation from the mean of the free surface of a rotating fluid. A corresponding Hamiltonian structure due to Weinstein (1983) allows to define the stationary solution and to investigate linear and nonlinear stability as before. The parallel shear flow situation is described too as well as a pendant to Andrew's theorem. A section is devoted to steadily-travelling solutions (Rossby waves, modons). The first variation of several Hamiltonians vanishes at the stationary solution but a stability theory following Arnold cannot be established because Andrew type theorems exclude the existence of flows satisfying the convexity assumptions on the pendant of \(\Phi\). So stability remains an open problem in this case. The last chapter presents a similar study for the Korteweg-de Vries equation. First, a derivation is presented based on a model of a two-dimensional stratified (in the \(z\) direction) fluid. The stream function is supposed to take the form \[ \Psi= \varphi(x,t) \Phi(z) \] After scaling, \(\varphi\) satisfies \[ \varphi_t+ \varphi\varphi_x+ \varphi_{xxx}=0. \] Several Hamiltonian formulations are indicated but the one from Gardner (1971) is studied in detail. Periodic and soliton solutions are given directly and analytically. The Hamiltonian has to be modified by adding Casimirs so that its first variation vanishes at these solutions. The proof of Jeffrey and Kakutani in the early seventies for linear stability of the KdV soliton by a direct analytical method is given but one has to take care of the fact that the perturbed solution in the stability analysis shifts away from the nominal solution. Finally, nonlinear stability holds and the proof uses a sliding pseudo-metric for handling the above mentioned shift, a functional more general than the second variation of the Hamiltonian, i.e. containing higher-order terms and the spectral properties of the density of the second variation of the Hamiltonian (at the soliton). The definiteness of the second variation of the Hamiltonian was not straightforward. The book is clearly written and accessible to an advanced undergraduate student in mathematics. The computations can be followed directly and they are sometimes superfluous (page 58 the fact that \(H\) is an invariant of motion follows from \([H,H]= 0\) which follows from the antisymmetry of the Poisson bracket so the last property only has to be checked). Also formula (3.18) has been used in the derivation of (3.31) so that later reestablishing it using (3.31) is tautological. Some typos have been detected, \(d^2\) page 30 vanishes always (so there is no need to be at the equilibrium to get rid of the term where it appears), page 82 ``it'' is missing in the middle, page 122 \(\delta\varphi\) has to be replaced by \(\varphi\), page 210 (6.24) has to be equated to zero and the derivation of (6.27) is not natural in the sense that the term above is correct but is easier to obtain from the result than from the preceding line. The book attempts to present a systematic and unifying approach for the study of the stability of certain quantities arising in two-dimensional fluid dynamics via a Hamiltonian formalism, but as the author points out, there is no systematic way of devising an adequate Hamiltonian structure (which may not be unique), and once it is available, the definiteness of the second variation of the Hamiltonian is first sufficient (only) for stability and second it may not be obtained in a systematic way. Finally let us quote a note of the author before the introduction for meditation. Science is also about persons ``engaged in conversation'' which ``is not about who is better, but is simply the process of discovering the truth'' and which has to be distinguished from ``the language of the conversation''.