Abstract
The two-dimensional (2-D) Euler equations of a perfect fluid possess a beautiful geometric description: they are reduced geodesic equations on the infinite-dimensional Lie group of symplectomorphims with respect to a right-invariant Riemannian metric. This structure enables insights to Eulerian and Lagrangian stability via sectional curvature and Jacobi equations. The Zeitlin model is a finite-dimensional analogue of the 2-D Euler equations; the only known discretization that preserves the rich geometric structure. Theoretical and numerical studies indicate that Zeitlin’s model provides consistent long-time behaviour on large scales, but to which extent it truly reflects the Euler equations is mainly open. Towards progress, we give here two results. First, convergence of the sectional curvature in the Euler–Zeitlin equations on the Lie algebra \(\mathfrak {su}(N)\) to that of the Euler equations on the sphere. Second, \(L^2\)-convergence of the corresponding Jacobi equations for Lagrangian and Eulerian stability. The results allow geometric conclusions about Zeitlin’s model to be transferred to Euler’s equations and vice versa, which could expedite the ultimate aim: to characterize the generic long-time behaviour of perfect 2-D fluids.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
The motion of an incompressible fluid can be described in two equivalent ways: either by following the trajectory of a fixed fluid particle—the Lagrangian point of view—or by considering the velocity of fluid particles passing by a fixed point in space—the Eulerian point of view. The two viewpoints give rise to separate notions of stability with respect to perturbation of the initial conditions.
In his pioneering work, Arnold [3] showed that ideal fluid motions describe geodesics on the group of volume-preserving diffeomorphisms endowed with a right-invariant \(L^2\) Riemannian metric. This beautiful description enables tools from Riemannian geometry in the study of ideal fluids. Now, Riemannian geodesic motion is stable in the Lagrangian sense if a perturbation of the initial conditions yields a geodesic that stays close to the unperturbed one. Infinitesimal perturbations of geodesics are given by the Jacobi fields, whose evolution is governed by the sectional curvature. Roughly speaking, negative curvature suggests instability, whereas positive curvature suggests stability. This leads to an interest in sectional curvature of volume-preserving diffeomorphism groups on various domains. Using Fourier series, Arnold first derived a formula for the sectional curvature when the fluid domain is the two-torus. Lukatsky [17], Arakelyan-Savvidy [2], Dowker-Wei [10] and Yoshida [32] then derived corresponding formulae for the sphere, Nakamura [23] for the three-torus case, whereas Lukatsky [18] and Preston [28] exposed general computations for two-dimensional (2-D) compact surfaces. Regarding the link with stability, Misiołek [19] adapted Rauch comparison theorem to prove that negative sectional curvature of a plane spanned by the velocity and a Jacobi field implies that the \(L^2\) norm of the Jacobi field grows at least linearly in time.
Misiołek’s result established “slow” (less than exponential) Lagrangian instability in time. Arnold, however, early advocated that mostly negative sectional curvature should imply “fast” (exponential) instability and based thereon he concluded that long-term weather forecasts are intrinsically unreliable (see [4, Preface and Ch. 4B]). The question was further clarified by Preston [26], who demonstrated that, although the sectional curvature is non-positive and mostly strictly negative, the Jacobi fields do not necessarily grow “fast”. More precisely, he studied a splitting of the Jacobi equations and proved that (i) the sign of sectional curvature alone could not provide information about exponential Lagrangian instability, and (ii) exponential Eulerian instability always imply Lagrangian exponential instability. In summary, the notion of Lagrangian and Eulerian stability and their connection to sectional curvature is an important tool in the analysis of ideal hydrodynamics.
Another important tool for understanding 2-D ideal hydrodynamics is Zeitlin’s model [33, 34], which in turn is based on quantization results of Hoppe [12, 13]. Zeitlin’s model is the only known (spatial) finite-dimensional approximation that fully adopts Arnold’s geometric description in that it also describes geodesics on a Lie group equipped with a right-invariant Riemannian metric. This structure gives rise to conservation of Casimirs, such as enstrophy, critical for the understanding of 2-D-specific turbulence phenomenology as described by Kraichnan [16]. In this context, Zeitlin’s model provide a coherent approach to simulating the qualitative long-time behaviour of 2-D Euler equations [9, 20]. Furthermore, since Zeitlin’s model establish a link between hydrodynamics and matrix theory, results from the latter enable new techniques [22]. An example is canonical decomposition of the vorticity field along the stabilizer of the stream function, which captures the dynamics of large and small vortex scales formations [21]. Numerical evidences show that Zeitlin’s model, contrary to traditional discretization, retain the correct qualitative behaviour, for example the spectral power laws in the inverse energy cascade [8]. Local convergence of solutions to Zeitlin’s model to solutions of the Euler equations was established by Gallagher [11]. But a rigorous understanding of the observed superior long-time behaviour remains largely open.
Our objective here is to answer the following question: how well does Zeitlin’s model capture the Lagrangian and Eulerian stability properties of the 2-D Euler equations? In particular, does the sectional curvature in Zeitlin’s discretization converge to the sectional curvature of the Euler equations? We answer this and related questions for the case when the fluid domain is the sphere; the most relevant domain for applications in geophysical fluid dynamics (cf. [25, 35]).
In Sect. 2 we recall the geometric formulation of Euler equations on the sphere. In Sect. 3 we then present Zeitlin’s finite-dimensional analogue of the continuous Euler equations. In Sect. 4 we recall the notions of Eulerian and Lagrangian stability and the link to sectional curvature. As a first main result, we show in Sect. 6 that the sectional curvature of the Euler–Zeitlin equations converges to the sectional curvature of the Euler equations when the degrees of freedom in the model tend to infinity. In particular, this result implies that for Zeitlin’s model with enough degrees of freedom, the sign of the finite- and infinite-dimensional sectional curvatures are the same. Thus, Zeitlin’s model preserve the Lagrangian stability behaviour implied by the sectional curvature. The second main result, proved in Sect. 7, concerns stationary solutions of both the Euler–Zeitlin and the Euler equations. We show that Lagrangian perturbations (i.e. Jacobi fields) and Eulerian perturbations of the Zeitlin system converge in a certain sense toward corresponding perturbations of the continuous Euler equations. Thus, Zeitlin’s model also preserves the stable or unstable nature of stationary solutions to the Euler equations. We are not aware of any other discretization of the 2-D Euler equations with this property.
2 Euler’s Equations on the Sphere
In his seminal work, Arnold [3] described ideal incompressible fluid flows on a Riemannian manifold (M, g) with volume form \(\mu \), as an equation for geodesics on the infinite-dimensional group of volume preserving diffeomorphisms \({\text {Diff}}_{\mu }(M)\), with respect to the standard \(L^2\) metric. More precisely, Arnold showed that the geodesic equation on \({\text {Diff}}_{\mu }(M)\), right translated to the Lie algebra \(\mathfrak {X}_\mu (M)\) of divergence-free vector fields, yields the incompressible Euler equations
where \(\dot{u} {:=}\partial _t u\) denotes differentiation with respect to time, and \(\nabla _u u\) denotes the co-variant derivative of u along u. The equivalent Hamiltonian formulation on the cotangent bundle is described by the kinetic energy Hamiltonian, which is also right-invariant. Hamilton’s equations can then be reduced by translation to the corresponding Lie-Poisson system on the smooth dual of the Lie algebra \(\mathfrak {X}_\mu (M)^* \simeq \Omega ^1(M)/\Omega ^0(M)\) (cf. [4]).
When M is the unit sphere \(\mathbb {S}^2\), there is a Lie–Poisson isomorphism between \(\mathfrak {X}_\mu (M)^*\) and the space \(C^\infty _0(\mathbb {S}^2)\) of smooth function with vanishing mean. The isomorphism is given by \(\star d\), where \(\star \) is the Hodge star and d is the exterior derivative. When identifying a one-form to a divergence-free velocity field via the metric, the aforementioned isomorphism consists in taking the \({\text {curl}}\) of the velocity field which gives the vorticity function \(\omega \in C^\infty _0(\mathbb {S}^2)\). The significance of the vorticity function is that it is transported by the vector field u. In turn, the divergence free vector field u is the Hamiltonian vector field \(X_\psi \) for some Hamiltonian \(\psi \in C_0^\infty (\mathbb {S}^2)\) called the stream function. The Euler Eq. (1) can then be written entirely in terms of the vorticity and stream function
The Poisson bracket on \(\mathbb {S}^2\) is given by
where the gradients are taken in \(\mathbb {R}^3\), extending constantly on rays functions on the sphere. The vorticity Eq. (2) is an infinite-dimensional Lie–Poisson system on \(C^\infty _0(\mathbb {S}^2)\) for the Lie-Poisson bracket given by
where \(\mathcal {F},\mathcal {G}\) are functionals \(:C^\infty _0 (\mathbb {S}^2) \rightarrow \mathbb {R}\), and for the specific Hamiltonian
In addition to total energy, and contrary to the 3-D Euler equations, the 2-D system possesses infinitely many constants of motion, called Casimirs: for any function \(f\in C^\infty (\mathbb {R},\mathbb {R})\),
is conserved. A distinguished class of Casimirs are those with \(f(x) = x^k \; (k\in \mathbb {N})\). In particular, the quadratic Casimir is called enstrophy, and, as mentioned, is critical in 2-D turbulence phenomenology [16]. Higher-order Casimirs also play a role in the formation of large-scale coherent vortex structures [1].
Notice that the Hamiltonian \(\mathcal {H}\) in Eq. (3) is equal to the kinetic energy (\(L^2\) norm) of the velocity field which in turn is the \(H^1\) norm of the stream function
where \(u = X_\psi \). The connection between solutions to the vorticity Eq. (2) and geodesics on \({\text {Diff}}_\mu (\mathbb {S}^2)\) is established as follows: if \(\omega (t)\) is a solution and \(\psi (t)\) the corresponding path of stream functions, then a geodesic curve \(\gamma (t) \in {\text {Diff}}_\mu (\mathbb {S}^2)\) is obtained by integrating the corresponding non-autonomous ordinary differential equation
3 Zeitlin’s Model on the Sphere
Zeitlin’s insight was to use quantization theory to spatially discretize the vorticity Eq. (2) by replacing the Poisson algebra of smooth functions with the matrix Lie algebra \(\mathfrak {u}(N)\) of skew-Hermitian \(N\times N\) complex matrices [33, 34]. To achieve this, Zeitlin used an explicit quantization scheme developed by Hoppe [12, 13] initially within the context of relativistic membranes. Hoppe’s quantization is an example of Toeplitz quantization [5, 6].
3.1 \(L_\alpha \) approximation
Bordemann, Hoppe, Schaller, and Schlichenmaier [5] proposed a set of axioms to characterize a family of matrix algebras \((\mathfrak {g}_N,\left[ \cdot ,\cdot \right] _N)\) as an approximation of an arbitrary Lie algebra \((\mathfrak {g},[\cdot ,\cdot ])\) (typically an infinite-dimensional Poisson algebra). They called it \(L_\alpha \) approximation. Given, for each Lie algebra \(\mathfrak {g}_N\), a distance \(d_N\) and a projection \(p_N:\mathfrak {g}\rightarrow \mathfrak {g}_N\), the family \((\mathfrak {g}_N,[\cdot ,\cdot ]_N,d_N,p_N)\) is an \(L_\alpha \) approximation of \(\mathfrak {g}\) if for each pair \(x,y\in \mathfrak {g}\)
-
(1)
\(d_N(p_N(x),p_N(y)) \rightarrow 0\) as \(N\rightarrow \infty \) implies \(x=y\), and
-
(2)
\(d_N([p_N(x),p_N(y)]_N - p_N([x,y])) \rightarrow 0\) as \(N\rightarrow \infty \).
3.2 Quantization of the sphere
Recall the \(L^2\) orthonormal basis for \(C^\infty (\mathbb {S}^2,\mathbb {C})\) provided by the spherical harmonics. Expressed in inclination-azimuthal coordinates \((\theta ,\phi )\in [0,\pi ]\times [0, 2\pi )\) they are
where \(P_{lm}\) are the associated Legendre polynomials for \(l\ge 0\) and \(m\in \{-l,\cdots ,l\}\). The spherical harmonics are eigenfunctions of the Laplace–Beltrami operator on \(\mathbb {S}^2\)
Using the spherical harmonics basis, Hoppe [12] gave in his thesis an explicit quantization of the Poisson algebra \(\mathfrak {g} = (C^\infty (\mathbb {S}^2,\mathbb {C}),\{\cdot ,\cdot \} )\) of complex valued smooth functions (see [5, Example 3] for an exposition in terms of \(L_\alpha \) approximations). The approximating Lie algebras are \(\mathfrak {g}_N = \mathfrak {gl}(N,\mathbb {C})\) with \([\cdot ,\cdot ]_N = \frac{1}{\hbar _N}[\cdot ,\cdot ]\), where
and \([\cdot ,\cdot ]\) is the matrix commutator.
Let us introduce the following rescaled inner products on \(\mathfrak {gl}(N,\mathbb {C})\)
The distances \(d_N\) are given by the induced norms of the inner products, and the projections are defined via \(p_N\mathcal {Y}_{lm}:=iT^N_{lm}\in \mathfrak {gl}(N,\mathbb {C})\) for
where ( : : : ) is the Wigner 3j-symbol. Note that the quantized harmonics \((T^N_{lm})\) form an orthonormal basis of \(\mathfrak {gl}(N,\mathbb {C})\) with respect to the rescaled inner product. Thus, a function expanded in spherical harmonics \(f = \sum _{l=0}^{\infty } \sum _m f^{lm}\mathcal {Y}_{lm}\) is projected onto
Recall from Sect. 2 that the vorticity and stream functions for the Euler equations on the sphere are real-valued. That is, we consider the Poisson subalgebra \(C^\infty (\mathbb {S}^2)\) of real valued functions. The corresponding Lie subalgebras of \(\mathfrak {gl}(N)\) are given by skew-Hermitian complex matrices \(\mathfrak {u}(N)\). Furthermore, the smaller, trace free subalgebras \(\mathfrak {su}(N)\subset \mathfrak {u}(N)\) correspond to \(C^\infty _0(\mathbb {S}^2)\).
The last ingredient we need in order to approximate the vorticity formulation (2) of the 2-D Euler equations is an approximation \(\Delta _N :\mathfrak {su}(N)\rightarrow \mathfrak {su}(N)\) of the Laplace–Beltrami operator. Given the construction above, it is natural to define it directly in terms of the basis \(T_{lm}^N\) as
so that, up to truncation \(l < N\), it corresponds to the Laplacian on \(C^\infty (\mathbb {S}^2)\). It turns out that the quantized Laplacian \(\Delta _N\) defined this way admits a beautiful, canonical description in the theory of unitary representation theory of \(\mathfrak {so}(3)\); see the work of Hoppe and Yau [14] for details. In short, if \(x^1,x^2,x^3\in C^\infty (\mathbb {S}^2)\) denotes the Euclidean coordinate functions (for \(\mathbb {S}^2\) embedded as the unit sphere in \(\mathbb {R}^3\)), and if \(X^i_N = p_N x^i\), then \(X^1_N,X^2_N,X^3_N\) are (scaled) generators of a representation of \(\mathfrak {so}(3)\) in \(\mathfrak {u}(N)\) and the quantized Laplacian for \(F\in \mathfrak {u}(N)\) is given by
We now have all the components we need to define Zeitlin’s model on the sphere. For more details on the projection operator \(p_N\) and the Hoppe–Yau Laplacian \(\Delta _N\), including efficient computer implementation, we refer to the work of Modin and Viviani [20] and Cifani, Modin, and Viviani [9].
3.3 The Euler–Zeitlin equations
Using the aforementioned matrix algebra approximation of \(\big ( C^\infty (\mathbb {S}^2),\{\cdot ,\cdot \}\big )\), the Euler–Zeitlin equations are given by the following matrix flow for \(W = W(t) \in \mathfrak {su}(N)\)
where \(W=W(t)\in \mathfrak {su}(N)\) is the vorticity matrix and \(P = P(t)\in \mathfrak {su}(N)\) is the stream matrix. The remarkable feature of these equations is that they completely capture Arnold’s description, but in a finite-dimensional setting: the Euler–Zeitlin equations constitute a Lie–Poisson system on the dual \(\mathfrak {su}(N)^*\), which via the Frobenius inner product is identified with \(\mathfrak {su}(N)\), for the Hamiltonian given by
The associated right invariant Riemannian metric on the matrix group is determined from the inner product on the Lie algebra \(\mathfrak {su}(N)\) given by
The reconstruction equation for geodesics on the matrix group \({\text {SU}}(N)\) is
This equation is the direct analogue to Eq. (4) for geodesics on the infinite-dimensional group \({\text {Diff}}_\mu (\mathbb {S}^2)\).
The Euler–Zeitlin system (6) is isospectral, which reflects the transport nature of the vorticity Eq. (2). Indeed, in addition to the Hamiltonian, the system has the following constants of motion (Casimirs)
As \(N\rightarrow \infty \) these converge to the corresponding Casimirs \(\mathcal {C}_k(\omega )\) of the continuous system (see [29, Corollary 8.1.2]).
4 Stability Results
4.1 Lagrangian stability and sectional curvature
For a fluid domain given by a compact Riemannian manifold M, let us first review the notion of Lagrangian stability on \({\text {Diff}}_\mu (M)\) and how it is related to sectional curvature. Consider a geodesic \(\gamma = \gamma (t)\) on \({\text {Diff}}_{\mu }(M)\), and a smooth family \(\gamma _s\) of geodesics such that \(\gamma _0 = \gamma \). The corresponding Jacobi field is the vector field along \(\gamma \) defined by
It satisfies the Jacobi equation, which in abstract notation can be written
where \({\bar{\nabla }}_{{\dot{\gamma }}}\) is the “big” co-variant derivative on \({\text {Diff}}_\mu (\mathbb {S}^2)\) (as an infinite-dimensional Riemannian manifold) and R is the corresponding “big” Riemann curvature tensor (see [19] for details on the functional analytic setting). The fluid motion \(\gamma \) is Lagrangian stable if every Jacobi field along \(\gamma \) remains bounded in the Riemannian metric along the trajectory \(t\mapsto \gamma (t)\).
From the curvature tensor one can extract sectional curvature for the plane spanned by two tangent vectors \(U,V\in T_\varphi {\text {Diff}}_\mu (M)\) as
where \(\left\langle \cdot ,\cdot \right\rangle _\varphi \) denotes the Riemannian metric and \(\Vert \cdot \Vert _{\varphi }\) the corresponding norm. Due to right invariance, it follows that \(C_\varphi (U,V) = C_{{\text {id}}}(u,v) {=:}C(u,v)\) for the divergence free vector fields on M given by \(u=U\circ \varphi ^{-1}\) and \(v = V\circ \varphi ^{-1}\). Furthermore, Arnold [3] gave the explicit formula
where \(B:\mathfrak {X}_\mu (M)\times \mathfrak {X}_\mu (M)\rightarrow \mathfrak {X}_\mu (M)\) is defined by
For the case \(M=\mathbb {S}^2\), with \(u=X_f\) and \(v=X_g\) for \(f,g\in C^\infty _0(\mathbb {S}^2)\), the sectional curvature is
where \(\langle \cdot ,\cdot \rangle _{H^{1}} = \langle -\Delta (\cdot ),\cdot \rangle _{L^2}\) and correspondingly for \(\Vert \cdot \Vert _{H^{1}}\). A direct analogue of this formula in the finite-dimensional case yields the sectional curvature \(C_N:\mathfrak {su}(N)\times \mathfrak {su}(N)\rightarrow \mathbb {R}\) of \({\text {SU}}(N)\) in Zeitlin’s model (see Sect. 6 below).
Sectional curvature provide important information about the Lagrangian stability of the fluid. For example, Misiołek [19, Lemma 4.2] proved that if the sectional curvature of the plane spanned by \(\dot{\gamma }\) and a Jacobi field J remains non-positive along \(\gamma \), then J grows at least linearly in time, and, consequently, \(\gamma \) is at least weakly Lagrangian unstable. Similar results are valid also in finite dimensions, as pointed out by Preston [26].
We are now ready to state the first main theorem of the paper.
Theorem 1
The sectional curvature of \({\text {SU}}(N)\) (with Zeitlin’s metric) converges to the sectional curvature of \({\text {Diff}}_\mu (\mathbb {S}^2)\) (with Arnold’s metric) as follows: for any \(f,g \in H^7(\mathbb {S}^2)\)
where the constant \(\alpha >0\) is independent of f, g, N.
4.2 Splitting of Jacobi equation and Eulerian stability
Just as the Euler equations are expressed in the Eulerian (right reduced) variable \(u={\dot{\varphi }}\circ \varphi ^{-1}\), it is natural to express the corresponding Jacobi equation (8) also in Eulerian variables. As explored by Preston [26, 27], such a rewriting splits the second order Jacobi equation into two first order equations that exposes the links between Eulerian and Lagrangian stability. For the family of geodesics \(\gamma _s\) as before, let \(u_s = \dot{\gamma _s}\circ \gamma ^{-1}_s\) be the associated Eulerian velocity fields, each one a solution to the Euler equations. The right reduced version of the Jacobi field J in Eq. (7) is \(j = J \circ \gamma ^{-1}\). Since the Jacobi equation is of second order, we also need a variable corresponding to the derivative of J, namely the variation of the velocity field
Under this change of variables \((J,\dot{J})\leftrightarrow (j, z)\), the Jacobi equation (8) turns into the first order system
As expected, Eq. (10b) is the linearized Euler equation about the solution u, whereas Eq. (10a) is a reconstruction equation for the reduced Jacobi field j.
On the sphere we obtain a “vorticity formulation” expressed in the functions \(\upsilon ,\zeta \in C^\infty _0(\mathbb {S}^2)\) defined by \(j = X_\upsilon \) and \(\zeta = {\text {curl}} z\). Geometrically, \(\upsilon \) is an element of the Lie algebra, whereas \(\zeta \) is an element of the dual. The Eq. (10) expressed in these variables become
where \(\omega \) fulfil the vorticity Eq. (2). Now, the solution \(\omega =\omega (t)\) is called Eulerian stable (with respect to some norm) if every solution \(\zeta \) of the system (11) is bounded uniformly in time.
Remark 4.1
On an arbitrary Lie group G endowed with a right-invariant metric, the corresponding splitting of the Jacobi equations become the following system of equations on the direct product of \(\mathfrak {g}=T_e G\) with itself:
where \(W=W(t)\in \mathfrak {g}\) is a solution to the Euler–Arnold equation for geodesic curves on G, \({\text {ad}}^*_X\) is the adjoint of the linear operator \({\text {ad}}_X = \left[ X, \cdot \right] \), and \(\mathcal {A}:\mathfrak {g}\rightarrow \mathfrak {g}^*\) is the inertia operator that defines the inner product on \(\mathfrak {g}\).
For details, see for example [15, sec. 4].
For the Euler–Zeitlin equations (6) on \(\mathfrak {su}(N)\) we get the analogue of (11)
The interpretation of these equations is two-fold: they describe at the same time a discretization of Preston’s reduced Jacobi equations (11) and, independently of the connection to the Euler equations, the reduced form of the Jacobi equations for the Zeitlin model describing geodesics on \({\text {SU}}(N)\).
To study the correspondence of Eulerian and Lagrangian stability between the Euler and Euler–Zeitlin systems we now restrict to stationary solutions of both systems. It means that
It is an open problem to characterize in which situations stationary Euler (13) implies stationary Euler–Zeitlin (14) (although we always have that if \(\omega \) is stationary then Eq. (14) is fulfilled in the limit \(N\rightarrow \infty \)). However, the following situations are straightforward to check (cf. Viviani [31]):
-
If, for some fixed l, the vorticity is \(\omega = \sum _{m=-l}^l \omega ^{lm}\mathcal {Y}_{lm}\) then \(\omega \) and \(p_N\omega \) are stationary solutions.
-
If \(\omega \) is zonal (there exists a choice of north pole on the sphere for which \(\omega \) is constant on fixed latitudes) then \(\omega \) and \(p_N\omega \) are stationary solutions.
Define the embedding \(\iota _N:\mathfrak {u}(N)\rightarrow C^\infty (\mathbb {S}^2)\) by
Our second main theorem states that corresponding stationary solutions to the Euler and the Euler–Zeitlin equations share the same Eulerian and Lagrangian \(L^2\)-stability as \(N\rightarrow \infty \).
Theorem 2
Let \(\omega \) be a stationary solution of the vorticity Eq. (2) such that \(W = p_N \omega \) is a stationary solution of the Euler–Zeitlin equation (6). Let \(\upsilon (t)\) and \(\zeta (t)\) be solutions of the reduced Jacobi equations (11). Furthermore, let Y(t) and Z(t) be corresponding solutions of the finite-dimensional reduced Jacobi equations (12) with \(Y(0) = p_N\upsilon (0)\) and \(Z(0) = p_N \zeta (0)\). Then, for any fixed t,
Moreover, the convergence is uniform on bounded intervals of t.
One interpretation is that the Zeitlin discretization (12) of the reduced Jacobi equations (11) is convergent in the sense of numerical analysis. Indeed, the proof is based on concepts of stability and consistency and is given in Sect. 7. The theorem implies that Euler–Zeitlin model preserves the stable or unstable nature of stationary solutions of Euler equations, so that results on either the continuous or the discretized system can be transferred to the other. For example, the result by Taylor [30, Thm. 4.1.1], that zonal stationary solutions that are strictly monotonous in the meridional direction are Eulerian stable.
5 Bracket Convergence and Preliminary Estimates
Before proving the two main theorems, we expose a central result used in the proofs, and then give some needed estimates.
The projection \(p_N\) can be understood as Toeplitz quantization on the sphere. The central result of interest to us is that the Poisson bracket is approximated by the Lie algebra bracket in a stronger sense than the one shown by Hoppe. To state it we first introduce the matrix operator norm, given for \(A\in \mathfrak {u}(N)\) by
where \(\left\Vert \cdot \right\Vert \) denotes the Euclidean norm on \(\mathbb {C}^N\). The notation of the norm is motivated by the following consistency result.
Theorem 3
(Bordemann, Meinrenken, Schlichenmaier [6]) For every \(f\in C^\infty (\mathbb {S}^2)\) there exists \(c>0\) such that
In the same paper the authors also prove convergence of the brackets as \(\hbar _N\rightarrow 0\). Indeed, they prove the following result:
Theorem 4
(Bordemann, Meinrenken, Schlichenmaier [6]) For every \(f,g \in C^\infty (\mathbb {S}^2)\)
Charles and Polterovich [7, Prop. 3.6, 3.9] exposed a sharper estimate of this quantum-classical correspondence, namely
where \(c_0>0\) is a constant independent of f and g, and \(\left\Vert f \right\Vert _{C^k} = \underset{i \le k}{{\text {max}}} \; {\text {sup}} \left|\nabla ^i f \right|\). Via Sobolev embeddings we then obtain the following bound
where \(\left\Vert \cdot \right\Vert _{H^{5}}\) is a Sobolev \(H^5\) norm (with a suitable, fixed scaling).
In addition to the spectral norm \(\left\Vert \cdot \right\Vert _{L^\infty _N}\), we shall use the following inner products and norms on \(\mathfrak {su}(N)\)
where \(\lambda _1,\ldots ,\lambda _N\) are the eigenvalues of \(A\in \mathfrak {su}(N)\). The first result is that the inner products converge spectrally to their continuous analogues.
Lemma 5
Let \(q\in \{ -1,0,1\}\) and \(s\ge q\). For every \(f,g \in H^s(\mathbb {S}^2)\)
Proof
The embedding \(\iota _N\), defined in Eq. (15), is isometric for any pair \(H^q_N,H^q\) of inner products on \(\mathfrak {u}(N)\) and \(C^\infty (\mathbb {S}^2)\).
Let \(\Pi _N\) denote the \(L^2\)–projection on spherical harmonics with \(l\ge N\). Then
These (non-sharp) estimates conclude the proof. \(\square \)
The second lemma concerns the comparison of the different inner products and norms.
Lemma 6
For every \(A,B\in \mathfrak {su}(N)\)
Proof of Lemma 6
(i) The quantized harmonics \(iT_{lm} = p_N \mathcal {Y}_{lm}\) form an \(\left\langle \cdot , \cdot \right\rangle _{L^2_N}\)-orthonormal basis for \(\mathfrak {gl}(N,\mathbb {C})\). If we expand \(A\in \mathfrak {su}(N)\subset \mathfrak {gl}(N,\mathbb {C})\) in that basis
we have
The estimates (ii)–(iii) follow since \(2 \le l(l+1) \le N(N+1)\).
Let \(i\lambda _1,\ldots ,i\lambda _N\) denote the eigenvalues of A. The estimate (iv) then follows since
The last estimate is direct from the definition of \(\left\Vert A \right\Vert _{L^1_N}\). \(\square \)
6 Proof of the Convergence of Sectional Curvature
In this section we prove Theorem 1, which states that sectional curvature of the quantized matrix algebras \((\mathfrak {su}(N)\), \( \left[ \cdot ,\cdot \right] _N\), \(\left\langle \cdot ,\cdot \right\rangle _{H^{1}_N})\) converges as \(N\rightarrow \infty \) to the sectional curvature of infinite-dimensional system
Arnold [3] gave the general formula for sectional curvature on a Lie group G equipped with a right invariant Riemannian metric determined by an arbitrary inner product \(\left\langle \cdot ,\cdot \right\rangle \) and its corresponding norm \(\Vert \cdot \Vert \) on the Lie algebra \(\mathfrak {g}\):
where \(B:\mathfrak {g}\times \mathfrak {g}\rightarrow \mathfrak {g}\) is defined by
In our specific case, when \(\mathfrak {g}\) is the Poisson algebra of smooth functions on the sphere and the metric is Sobolev \(H^{1}\), we have
Substitution into Arnold’s formula (17) yields Eq. (9) above for the sectional curvature \(C(X_f,X_g)\). It is convenient to express it in the \(L^2\) inner product (because \(L^2\) is bi-invariant with respect to the Poisson algebra)
The same calculation for \(\mathfrak {su}(N)\) equipped with the right invariant metric determined by the inner product \(\left\langle \cdot ,\cdot \right\rangle _{H^1_N}\) yields the analogous formula
Now, the aim is to prove that
To this end, we carry out estimates for each corresponding term in the formulae (18) and (19). The “problematic” term is the third, since the \(H^1_N\) norm is not bounded by \(L^\infty _N\).
6.1 The first, second, and fourth terms
By construction of the quantized Laplacian we have that \(p_N \circ \Delta = \Delta _N \circ p_N\). Thus,
If \(\Pi _N:C^\infty (\mathbb {S}^2)\rightarrow C^\infty (\mathbb {S}^2)\) denotes the projection onto spherical harmonics with \(l\ge N\), then for \(a,b,c,d\in C^\infty _0(\mathbb {S}^2)\) and \(q\in \{ 0,1\}\)
The last inequality follows from the well-known result:
Lemma 7
Let \(f,g \in C^1(\mathbb {S}^2)\). Then
Proof
Let \(J:T\mathbb {S}^2\rightarrow T\mathbb {S}^2\) denote the complex structure on \(\mathbb {S}^2\). By Cauchy–Schwartz
The estimate now follows since \(\mathbb {S}^2\) is Kähler, so J is an isometry. \(\square \)
By choosing a, b, c, d from the set \(\{f,g,\Delta f, \Delta g \}\), to match the corresponding terms, using Sobolev embeddings and estimating \(\left|r_N \right|\) with Lemma 5, we obtain
for a constant \(\alpha >0\) independent of f, g, N.
6.2 The third term
The term
cannot be treated directly by the estimate above, since the \(H^1_N\) norm is not controlled by the \(L^\infty _N\) norm as \(N\rightarrow \infty \). To overcome this problem we shall need the following results for the quantized Laplacian \(\Delta _N\).
Lemma 8
For the coordinate functions \(x^i\in C^\infty (\mathbb {S}^2)\) and the corresponding matrices \(X^i_N = p_N x^i\), let \(\nabla ^{\bot ,i} = \{x^i,\cdot \}\) and \(\nabla _N^{\bot ,i} = [X^i_N,\cdot ]/\hbar _N\). Then \(\nabla ^{\bot ,i}\) and \(\nabla _N^{\bot ,i}\) are \(p_N\)-related:
Proof
For any \(l\ge 1\), the Lie group \({\text {SO}}(3)\) acts on \(\{ f\in C^\infty (\mathbb {S}^2)\mid \Delta f = l(l+1)f \}\) via its action on \(\mathbb {S}^2\). The corresponding infinitesimal representation of \(\mathfrak {so}(3)\) is Hamiltonian with respect to the Poisson bracket with generators \(x^1,x^2,x^3 \in C^\infty (\mathbb {S}^2)\), i.e., given exactly by \(\nabla ^{\bot ,i}\). From the work of Hoppe and Yau [14] it follows that \(X_N^1,X_N^2,X_N^3\) are corresponding generators for the infinitesimal representation of \(\mathfrak {so}(3)\) on \(\{F\in \mathfrak {su}(N)\mid \Delta _N F = l(l+1)F \}\) and that \(p_N\) is an algebra morphism. This proves the result. \(\square \)
Lemma 9
Let \(F,G\in \mathfrak {su}(N)\) and \(f,g\in C^3(\mathbb {S}^2)\). Then
and
Proof
The first result follows from the Hoppe–Yau formula (5) and application of the Jacobi identity twice.
The second result is a direct continuous analogue. \(\square \)
From these results we obtain
Each pair of corresponding quantized and continuous brackets can now be treated by the same estimate as the second term above, which gives
for constants \(\alpha >0\), \(\beta >0\), and \(\gamma >0\) independent of f, g, N. From the Sobolev norm relations \(\left\Vert f \right\Vert _{H^q} \le \left\Vert f \right\Vert _{H^p}\) for \(q\le p\) and from Lemma 5 we then obtain the result in Theorem 1. This concludes the proof.
7 Proof of the Convergence of the Reduced Jacobi Equation
In this section we prove the second main result stated in Theorem 2 above.
Let us first rewrite the continuous and quantized reduced Jacobi equations (11) and (12) for corresponding stationary solutions \(\omega _0 \in C_0^\infty (\mathbb {S}^2)\) and \(W_0 = p_N \omega _0 \in \mathfrak {su}(N)\) as linear evolution equations on \(C^\infty (\mathbb {S}^2 \times \mathbb {S}^2)\) and \((\mathfrak {su}(N))^2\) respectively. The continuous reduced Jacobi equation is then
where
The corresponding quantized reduced Jacobi equation is
where
Recall that the embedding \(\iota _N: \mathfrak {su}(N)\rightarrow C^\infty (\mathbb {S}^2)\) maps quantized harmonics to continuous ones. Let \(\pi _N: C^\infty (\mathbb {S}^2)\rightarrow C^\infty (\mathbb {S}^2)\) be the truncation of the (continuous) spherical harmonics expansion up to \(l \le N-1\), i.e., \(\pi _N = \iota _N p_N\). In order to directly compare the quantized and continuous Jacobi equations, we introduce the operator \(\mathcal {L}_N {:=}\iota _N\circ \Lambda _N\circ p_N\) such that if \({\varvec{X}}(t)\) is a solution of the matrix dynamical system (20), then \({\varvec{\xi }}_N (t)= \iota _N {\varvec{X}}(t) + \Pi _N {\varvec{\xi }}_0 \) is the solution of the continuous system
In order to prove the convergence \({\varvec{\xi }}_N (t)\) to \({\varvec{\xi }}(t)\), we will use the following result of Trotter and Kato (see, e.g., Pazy [24, Thm. 3.4.2]).
Theorem 10
(Trotter, Kato) For a Banach space \((X,\left\Vert \cdot \right\Vert )\), let \((L_N)_{N \ge 0} \) and L be linear operators \(D(L) \subset X \rightarrow X\). Let \(\left\Vert \cdot \right\Vert \) denote also the operator norm and make the following assumptions.
- Well-posedness:
-
there exist scalars \(M, \omega \) such that L is the infinitesimal generator of a \(C^0\) semigroup T(t) satisfying \( \left\Vert T(t) \right\Vert \le M e^{\omega t}\).
- Stability:
-
for any \(N \ge 0\), \(L_N\) is the infinitesimal generator of a \(C^0\) semigroup \(T_N(t)\) satisfying \( \left\Vert T_N(t) \right\Vert \le M e^{\omega t}\) (for the same M and \(\omega \)).
- Consistency:
-
for every \(x\in X\) and \(\lambda \in \mathbb {C}\) with \({\text {Re}} \lambda > \omega \)
$$\begin{aligned} \left\Vert (I - \lambda L_N)^{-1}x - (I - \lambda L)^{-1}x \right\Vert \rightarrow 0 \quad \text {as}\quad N\rightarrow \infty . \end{aligned}$$
Then, for every \(x \in X\) and \(t \ge 0\),
Moreover, the convergence is uniform on bounded intervals of t.
Remark 7.1
The classical formulation of the Trotter & Kato result does not include convergence rates. In the special case of Zeitlin’s model, the convergence rate is \(O(\hbar _N)\). In Appendix A we have included a proof of Theorem 10 with this convergence rate.
7.1 Well-posedness of the continuous system
Here we prove that \(\mathcal {L}\) is the infinitesimal generator of a \(C^0\) semigroup with respect to the \(L^2\) norm.
Proposition 11
The operator \(\mathcal {L}:C^\infty (\mathbb {S}^2 \times \mathbb {S}^2)\rightarrow L^2(\mathbb {S}^2 \times \mathbb {S}^2)\) is the generator of a \(C^0\) semigroup \(\mathcal {T}(t)\) with
Proof
We split \(\mathcal {L} = \mathcal {L}_1 + \mathcal {L}_2\), where
Now, \(\mathcal {L}_1\) is the generator for the semigroup \(\mathcal T_1(t)\) given explicitly by
where \(\eta _t\) is the diffeomorphism on \(\mathbb {S}^2\) generated by the finite-dimensional Hamiltonian vector field \(X_{\Delta ^{-1}\omega _0}\) (the solution exists for all times since \(X_{\Delta ^{-1}\omega _0}\) is a smooth vector field on a compact domain). Notice that \(\mathcal {T}_1(t)\) is bounded with operator norm 1, since
For the second part, \(\mathcal {L}_2\) is a bounded operator, since
Thus, \(\mathcal {L}_2\) generates a \(C^0\) semigroup \(\mathcal {T}_2(t)\). By the Lie–Trotter formula we then obtain the \(C^0\) semigroup \(\mathcal {T}(t)\) via
which is generated by \(\mathcal {L}\). The operator norm estimate also follows from the Lie–Trotter formula. \(\square \)
7.2 Stability of the semi-discrete method
We now prove that the quantized system (21) is stable, which is the semi-discrete correspondence to Proposition 11.
Proposition 12
For any \(N\ge 0\), the operator \(\mathcal {L}_N:C^\infty (\mathbb {S}^2 \times \mathbb {S}^2)\rightarrow L^2(\mathbb {S}^2 \times \mathbb {S}^2)\) is the generator of a \(C^0\) semigroup \(\mathcal {T}_N(t)\) with
Proof
The proof is a direct analogue of the proof of Proposition 11. For the operator splitting \(\mathcal {L}_N = \mathcal {L}_{N,1} + \mathcal {L}_{N,2}\), the semigroup for the generator \(\mathcal {L}_{N,1}\) is
where \(E_t = \exp (t \Delta _N^{-1}W_0/\hbar _N)\), \(Y = p_N \upsilon \), and \(Z = p_N \zeta \). We then have
The estimate for \(\mathcal {L}_{N,2}\) follows as in the proof of Proposition 11, using the estimates in Lemma 6. \(\square \)
7.3 Consistency of the semi-discrete method
First, let us prove the following lemma.
Lemma 13
There exist \(\alpha >0\) independent of \(\omega _0, \hbar _N\) and \({\varvec{\xi }}= \begin{bmatrix} f \\ g \end{bmatrix} \), such that
Proof
Using estimates from Sect. 5, we have
where \(\alpha >0\) is a constant. \(\square \)
Then we can prove consistency, namely the following result.
Proposition 14
For every \({\varvec{\xi }}\in C^\infty (\mathbb {S}^2 \times \mathbb {S}^2)\) and \(\lambda > c {:=}\sqrt{\frac{1}{4} + \frac{1}{2} \left\Vert \omega _0 \right\Vert _{H^{1}}^2}\),
where \(\tilde{{\varvec{\xi }}} = (I - \lambda \mathcal {L})^{-1} {\varvec{\xi }}\) and \(\alpha >0\) a constant as in Lemma 13.
Proof
Recall from Proposition 12 that \(\mathcal {L}_N\) is the infinitesimal generator of a \(C_0\) semigroup \(\mathcal {T}_N(t)\) satisfying \(\left\Vert \mathcal {T}_N(t) \right\Vert _{L^2} \le e^{ct}\). The Hille–Yosida theorem (see, e.g., Pazy [24, Thm. 1.5.3]) then implies that
Thus, we can proceed to the following estimates:
Let \({\tilde{{\varvec{\xi }}}} = (I - \lambda \mathcal {L})^{-1} {\varvec{\xi }}\). We then have from Lemma 13 that
Recall that \(\mathcal {L}_N = \iota _N \Lambda _N p_N\), so \(\mathcal {L}_N {\tilde{{\varvec{\xi }}}}\) is a finite sum of spherical harmonics with \(l \le N-1\). Thus, \(\pi _N \mathcal {L}_N = \mathcal {L}_N\), and
As \({\varvec{\xi }}\in C^\infty (\mathbb {S}^2 \times \mathbb {S}^2)\), we have \(\mathcal {L} \tilde{{\varvec{\xi }}} = \mathcal {L}(I - \lambda \mathcal {L})^{-1} {\varvec{\xi }}\in C^\infty (\mathbb {S}^2 \times \mathbb {S}^2) \subset H^2(\mathbb {S}^2 \times \mathbb {S}^2)\). Thus, by Lemma 5,
\(\square \)
We now have everything to apply Theorem 10 above: well-posedness from Proposition 11; stability from Proposition 12; consistency from Proposition 14. This concludes the proof of our second main result, stated in Theorem 2 above.
References
Abramov, R.V., Majda, A.J.: Statistically relevant conserved quantities for truncated quasigeostrophic flow. Proc. Nat. Acad. Sci. 100(7), 3841–3846 (2003)
Arakelyan, T.A., Savvidy, G.K.: Geometry of a group of area-preserving diffeomorphisms. Phys. Lett. B 223(1), 41–46 (1989)
Arnold, V.I.: Sur la géométrie différentielle des groupes de Lie de dimension infinie et ses applications à l’hydrodynamique des fluides parfaits. Annal. Inst. Four. 16(1), 319–361 (1966)
Arnold, V.I., Khesin, B.A.: Topological Methods in Hydrodynamics, vol. 125. Springer (1999)
Bordemann, M., Hoppe, J., Schaller, P., Schlichenmaier, M.: gl(\(\infty \)) and geometric quantization. Comm. Math. Phys. 138(2), 209–244 (1991)
Bordemann, M., Meinrenken, E., Schlichenmaier, M.: Toeplitz quantization of Kähler manifolds. Comm. Math. Phys. 165(2), 281–296 (1994)
Charles, L., Polterovich, L.: Sharp correspondence principle and quantum measurements. St. Petersburg Math. J. 29(1), 177–207 (2018)
Cifani, P., Viviani, M., Luesink, E., Modin, K., Geurts, B.: Casimir preserving spectrum of two-dimensional turbulence. Phys. Rev. Fluids 7, L082601 (2022)
Cifani, P., Viviani, M., Modin, K.: An efficient geometric method for incompressible hydrodynamics on the sphere. J. Comput. Phys. 473, 111772 (2023)
Dowker, J.S., Mo-zheng, W.: Area-preserving diffeomorphisms and the stability of the atmosphere. Class. Quant. Grav. 7(12), 2361 (1990)
Gallagher, I.: Mathematical analysis of a structure-preserving approximation of the bidimensional vorticity equation. Num. Math. 91(2), 223–236 (2002)
Hoppe, J.: Quantum Theory of a Relativistic Surface and a Two-Dimensional Bound State Problem, PhD Thesis, MIT (1982)
Hoppe, J.: Diffeomorphism groups, quantization, and \(\rm SU (\infty )\). Int. J. Mod. Phys. A 04(19), 5235–5248 (1989)
Hoppe, J., Yau, S.-T.: Some properties of matrix harmonics on \(S^2\). Comm. Math. Phys. 195(1), 67–77 (1998)
Khesin, B.A., Lenells, J., Misiołek, G., Preston, S.C.: Curvatures of Sobolev metrics on diffeomorphism groups. Pure Appl. Math. Quart. 9(2), 291–332 (2013)
Kraichnan, R.H.: Inertial ranges in two-dimensional turbulence. Phys. Fluids 10(7), 1417–1423 (1967)
Lukatskii, A.M.: Curvature of groups of diffeomorphisms preserving the measure of the 2-sphere. Func. Anal. Appl. 13(3), 174–177 (1979)
Lukatskii, A.M.: Structure of the curvature tensor of the group of measure-preserving diffeomorphisms of a compact two-dimensional manifold. Sib. Math. J. 29(6), 947–951 (1988)
Misiołek, G.: Stability of flows of ideal fluids and the geometry of the group of diffeomorphisms. Indiana Math. J. 42(1), 215–235 (1993)
Modin, K., Viviani, M.: A Casimir preserving scheme for long-time simulation of spherical ideal hydrodynamics. J. Fluid Mech. 884, A22 (2020)
Modin, K., Viviani, M.: Canonical scale separation in 2d incompressible hydrodynamics. J. Fluid Mech. 943, A36 (2022)
Modin, K., Viviani, M.: Two-Dimensional Fluids Via Matrix Hydrodynamics (2024)
Nakamura, F., Hattori, Y., Kambe, T.: Geodesics and curvature of a group of diffeomorphisms and motion of an ideal fluid. J. Phys. A 25(2), L45–L50 (1992)
Pazy, A.: Semigroups of Linear Operators and Applications to Partial Differential Equations. Springer, New York (1983)
Pedlosky, J.: Geophysical Fluid Dynamics. Springer (2013)
Preston, S. C.: Eulerian and Lagrangian Stability of Fluid Motions, PhD Thesis, State University of New York at Stony Brook (2002)
Preston, S.C.: For ideal fluids, Eulerian and Lagrangian instabilities are equivalent. Geom. Func. Anal. 14(5), 1044–1062 (2004)
Preston, S.C.: Nonpositive curvature on the area-preserving diffeomorphism group. J. Geom. Phys. 53(2), 226–248 (2005)
Rios, P.M., Straume, E.: Symbol Correspondences for Spin Systems. Birkhäuser, Cham (2014)
Taylor, M.: Euler equation on a rotating surface. J. Func. Anal. 270(10), 3884–3945 (2016)
Viviani, M.: Symplectic Methods for Isospectral Flows and 2D Ideal Hydrodynamics, PhD Thesis, Chalmers University of Technology (2020)
Yoshida, K.: Riemannian curvature on the group of area-preserving diffeomorphisms (motions of fluid) of 2-sphere. Phys. D 100(3–4), 377–389 (1997)
Zeitlin, V.: Finite-mode analogs of 2D ideal hydrodynamics: coadjoint orbits and local canonical structure. Phys. D 49, 353–362 (1991)
Zeitlin, V.: Self-consistent finite-mode approximations for the hydrodynamics of an incompressible fluid on nonrotating and rotating spheres. Phys. Rev. Lett. 93(26), 264501 (2004)
Zeitlin, V.: Geophysical Fluid Dynamics: Understanding (Almost) Everything with Rotating Shallow Water Models. Oxford University Press (2018)
Acknowledgements
The first author was supported by the Swedish Research Council (grant number 2022-03453) and the Knut and Alice Wallenberg Foundation (grant number WAF2019.0201). The second author was supported by a PhD fellowship from Ecole Normale Supérieure Paris. The authors would also like to thank Petra Flurin for her help on quantization estimates.
Funding
Open access funding provided by Chalmers University of Technology.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors have no relevant financial or non-financial interests to disclose. Data sharing is not applicable to this article as no datasets were generated or analysed during the current study.
Additional information
Communicated by P. Chrusciel.
To the memory of Vladimir Zeitlin, whose beautiful model guides us still.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Trotter and Kato theorem with Convergence Rate
Trotter and Kato theorem with Convergence Rate
In this section, we adapt the proof of the Trotter–Kato Theorem 10 from Pazy [24]. Let us denote \(\mathcal {R}_{\lambda }:= (\lambda I - \mathcal {L})^{-1}\) and \(\mathcal {R}_{\lambda ,N}:= (\lambda I - \mathcal {L}_N)^{-1}\). Fix \({\varvec{\xi }}\in D(\mathcal {L})\) and an interval [0, T]. For any \(t \in [0,T]\):
Using the stability of \(\mathcal {L}_N\) (Proposition 12) and the consistency estimate (Proposition 14) we have
and
where we have used the fact that \(\mathcal {R}_{\lambda } \mathcal {T}(t) = \mathcal {T}(t) \mathcal {R}_{\lambda }\). For the term \(N_2\), we first use the identity (e.g. Pazy, [24, Lemma 3.4.1])
Thus, we have
Then it follows from the above estimates that for any \({\varvec{\xi }}\in D(\mathcal {L})\):
uniformly on [0, T]. Since the Hille–Yosida theorem implies that \(D(\mathcal {L})\) is dense in \(C^\infty (\mathbb {S}^2 \times \mathbb {S}^2)\), it follows that the previous statement holds for every \({\varvec{\xi }}\in C^\infty (\mathbb {S}^2 \times \mathbb {S}^2)\). Moreover, it is clear that regarding only the N dependency, the convergence is \(O(\hbar _N)\).
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Modin, K., Perrot, M. Eulerian and Lagrangian Stability in Zeitlin’s Model of Hydrodynamics. Commun. Math. Phys. 405, 177 (2024). https://doi.org/10.1007/s00220-024-05047-x
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00220-024-05047-x