Abstract
We describe the geometry of the incompressible porous medium (IPM) equation: we prove that it is a gradient dynamical system on the group of area-preserving diffeomorphisms and has a special double-bracket form. Furthermore, we show its similarities and differences with the dispersionless Toda system. The Toda flow describes an integrable interaction of several particles on a line with an exponential potential between neighbours, while its continuous version is an integrable PDE, whose physical meaning was obscure. Here we show that this continuous Toda flow can be naturally regarded as a special IPM equation, while the key double-bracket property of Toda is shared by all equations of the IPM type, thus manifesting their gradient and non-autonomous Hamiltonian origin. Finally, we comment on Toda and IPM modifications of the QR diagonalization algorithm, as well as describe double-bracket flows in an invariant setting of general Lie groups with arbitrary inertia operators.
Similar content being viewed by others
Notes
If Q is an infinite-dimensional Fréchet manifold, the cotangent bundle \(T^*Q\) is given in terms of the regular dual (cf. [16]), defined so that \(T_q^*Q\simeq T_qQ\).
References
Balehowsky, T., Karlsson, C.-J., Modin, K.: Shape analysis via gradient flows on diffeomorphism groups. Nonlinearity 3(6), 862 (2022)
Bloch, A.M.: Steepest descent, linear programming, and Hamiltonian flows. Contemp. Math. 114, 77–88 (1990)
Bloch, A., Flaschka, H., Ratiu, T.: A Schur-Horn-Kostant convexity theorem for the diffeomorphism group of the annulus. Invent. Math. 113, 511–529 (1993)
Bloch, A.M., Brockett, R.W., Ratiu, T.S.: Completely integrable gradient flows. Commun. Math. Phys. 147, 57–74 (1992)
Brenier, Y.: Various formulations and approximations of incompressible fluid motions in porous media. Ann. Math. Qué. 4(6), 195–206 (2022)
Brockett, R.W., Bloch, A.M.: Sorting with the dispersionless limit of the Toda lattice. In: Proceedings of the CRM Workshop on Hamiltonian Systems, Transformation Groups and Spectral Transform Methods, Publications CRM, Montreal, 1–11 (1990)
Brockett, R.W.: Dynamical systems that sort lists, diagonalize matrices, and solve linear programming problems. Linear Algebra Appl. 146, 79–91 (1991)
Cifani, P., Viviani, M., Modin, K.: An efficient geometric method for incompressible hydrodynamics on the sphere. J. Comput. Phys. 473, 111772 (2023)
Córdoba, D., Gancedo, F., Orive, R.: Analytical behavior of two-dimensional incompressible flow in porous media. J. Math. Phys. 48, 065206 (2007)
Deift, P., Nanda, T., Tomei, C.: Ordinary differential equations and the symmetric eigenvalue problem. SIAM J. Numer. Anal. 20, 1–22 (1983)
Deift, P., Trogdon, T.: Universality for the Toda algorithm to compute the eigenvalues of a random matrix. Commun. Pure Appl. Math. 73, 505–536 (2018)
Ebin, D.G., Marsden, J.E.: Groups of diffeomorphisms and the notion of an incompressible fluid. Ann. Math. 92, 102–163 (1970)
Flaschka, H.: The Toda lattice. Phys. Rev. B 9, 1924–1925 (1974)
Hoppe, J.: Diffeomorphism groups, quantization, and \({{\rm SU}} (\infty )\). Int. J. Mod. Phys. A 04, 5235–5248 (1989)
Hoppe, J., Yau, S.-T.: Some properties of matrix harmonics on \(S^2\). Commun. Math. Phys. 195, 67–77 (1998)
Khesin, B., Wendt, R.: The Geometry of Infinite-Dimensional Groups. Springer, Berlin (2009)
Marsden, J.E., Ratiu, T.S.: Introduction to Mechanics and Symmetry, 2nd edn. Springer, New York (1999)
Modin, K.: Generalized Hunter–Saxton equations, optimal information transport, and factorization of diffeomorphisms. J. Geom. Anal. 25, 1306–1334 (2015)
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.: Lie–Poisson methods for isospectral flows. Found. Comput. Math. 20, 889–921 (2020)
Moser, J.: Finitely many mass points on the line under the influence of an exponential potential–an integrable system. In: Dynamical Systems, Theory and Applications. Springer, pp. 467–497 (1975)
Otto, F.: Evolution of microstructure in unstable porous media flow: a relaxational approach. Commun. Pure Appl. Math. 52(7), 873–915 (1999)
Palais, R.S.: Foundations of Global Non-linear Analysis. Benjamin, New York (1968)
Zeitlin, V.: Finite-mode analogs of \(2\)D ideal hydrodynamics: coadjoint orbits and local canonical structure. Physica D 49, 353–362 (1991)
Acknowledgements
We thank the anonymous referee for helpful suggestions. Research of B.K. was partially supported by an NSERC Discovery Grant. K.M. was supported by the Swedish Research Council (grant number 2022-03453) and the Knut and Alice Wallenberg Foundation (grant number WAF2019.0201). The computations were enabled by resources provided by the Swedish National Infrastructure for Computing (SNIC) at C3SE partially funded by the Swedish Research Council through grant agreement no. 2018-05973.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors have no competing interests to declare that are relevant to the content of this article. Data sharing is not applicable to this article as no datasets were generated or analysed during the current study.
Additional information
Communicated by C. Liverani.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix A: Universal Double Bracket Flows
1.1 The gradient flow on a Lie group
Here we consider a framework for Riemannian gradient flows confined to group orbits. This leads to the proper setting for the double bracket flows in the case of an arbitrary inertia operator. A similar framework, with focus on shape analysis, is also given in [1].
Let Q be a configuration manifold, possibly infinite-dimensional. Furthermore, let G be a Lie group (or a Fréchet–Lie group in the infinite-dimensional case) acting on Q from the left by a smooth action map \(\Phi :G\times Q\rightarrow Q\); the action of \(g\in G\) on \(q\in Q\) is denoted \(g\cdot q\). This action is typically neither free, nor transitive. The orbit of \(q\in Q\) is
We think of Q as the space of ‘shapes’ and the orbit \(\textrm{Orb}(q)\) represents all possible deformations of q. For a given template shape \(q_0\), our objective is to study gradient flows on \(\textrm{Orb}(q_0)\). Notice, however, that we do not assume that Q is Riemannian; the Riemannian structure on \(\textrm{Orb}(q_0)\) is instead inherited from G. Before we state the main result of this section, we need a few concepts from geometric mechanics (cf. [17]).
By differentiating the action map at the identity we obtain the infinitesimal action map \({\mathfrak {g}}\times Q\rightarrow T Q\), where \({\mathfrak {g}}= T_e G\) is the Lie algebra of G. The infinitesimal action of \(v\in {\mathfrak {g}}\) on \(q\in Q\) is denoted \(v\cdot q\). This linear map in v is the (cotangent bundle) momentum map:
Definition 5.1
The momentum map \(J:T^*Q\rightarrow {\mathfrak {g}}^*\) is defined by
where \(T^*Q\) denotes the cotangent bundle of Q.Footnote 1
Next, we introduce a Riemannian structure on G.
Definition 5.2
A Riemannian metric \(\langle \!\langle \cdot ,\cdot \rangle \!\rangle :TG\times TG\rightarrow {{\mathbb {R}}}\) on G is called right-invariant if
where \(u\cdot g\) denotes the tangent lifted right action of g on u.
A right-invariant metric is completely determined by the inner product \(\langle \!\langle \cdot ,\cdot \rangle \!\rangle _e\).
Definition 5.3
Let \(\langle \!\langle \cdot ,\cdot \rangle \!\rangle \) be a right-invariant metric on G. Then the inertia operator \(A:{\mathfrak {g}}\rightarrow {\mathfrak {g}}^*\) is defined by
Since G acts on Q from the left, the action map induces a Riemannian structure on \(\textrm{Orb}(q)\). To see this, we first need the notion of horizontal vectors on G.
Definition 5.4
The vertical distribution associated with the action of G on \(q\in Q\) is the subbundle of TG given by
If \(\langle \!\langle \cdot ,\cdot \rangle \!\rangle \) is a right-invariant metric on G, then the corresponding horizontal distribution is given by
where the complement is taken with respect to \(\langle \!\langle \cdot ,\cdot \rangle \!\rangle \).
Lemma 5.5
Assume that \(\textrm{Orb}(q)\) is a submanifold of Q. Then any right-invariant Riemannian metric \(\langle \!\langle \cdot ,\cdot \rangle \!\rangle \) induces a Riemannian metric \({\textsf{g}}\) on \(\textrm{Orb}(q)\) fulfilling
Proof
Since \(\textrm{Orb}(q)\) is a manifold, the mapping \(\pi :G\rightarrow Q\) defined by \(\pi (g) = g\cdot q\) is a submersion. Thus, \(T_g \pi :\textrm{Hor}_g \rightarrow T_{\pi (g)}\textrm{Orb}(q)\) is a linear isomorphism. Now, for \((x,\dot{x}) \in T\textrm{Orb}(q)\), take any g such that \(x = \pi (g)\) (which exist by definition of the group orbit). Define the metric at x by
By right-invariance, this metric satisfies (5.1) and is independent of the choice of g. \(\square \)
Let G be a Lie group acting from the left on a manifold Q of shapes. Let \(q_0\in Q\) and let \(F:Q\rightarrow {{\mathbb {R}}}\) be a function on Q. We are interested in finding the minimum of F on the G-orbit of \(q_0\), that is, we want to minimize the function \(E:G\rightarrow {{\mathbb {R}}}\) defined by
If G is equipped with a right-invariant Riemannian metric \({\textsf{g}}\), defined by an inertia operator \(A:{\mathfrak {g}}\rightarrow {\mathfrak {g}}^*\), one can define the gradient vector field \(\nabla E\) on G by
Our aim is to solve the minimization problem by considering the gradient flow
Proposition 5.6
The gradient \(\nabla E\) is given by
where \(\xi \in {\mathfrak {g}}\) is given by
Proof
By definition of the gradient we have
where \(\xi = \dot{g}\cdot g^{-1}\). Now from the definition of the momentum map it follows that
The result follows since the metric is right invariant. \(\square \)
Proposition 5.7
The gradient flow (5.2) induces a gradient flow on the G-orbit of \(q_0\), given by
where
Proof
Follows from Prop. 5.6. \(\square \)
Definition 5.8
The double-bracket flow on the dual Lie algebra \({\mathfrak {g}}^*\) with the inertia operator A and a potential function F on \({\mathfrak {g}}^*\) is as follows:
For the quadratic potential \(F(m) = \langle m, A^{-1}m\rangle \) the corresponding flow is
Proposition 5.9
The double-bracket flow on \({\mathfrak {g}}^*\) is the gradient for the restriction of F on each coadjoint orbit.
Proof
Indeed, consider the special case where \(Q = {\mathfrak {g}}^*\). The action is given by \(g\cdot m = {\text {Ad}}_g^* m\). The momentum map is thereby given by
Then from (5.3) we obtain the equation(s) in Definition 5.8, and the statement follows from Proposition 5.7. The corresponding flow tries to minimize the energy on a specific co-adjoint orbit. \(\square \)
Note also that this flow is always orthogonal (w.r.t. \(A^{-1}\)) to the (Hamiltonian) Euler-Arnold flow: one of them is tangent to levels of the Hamiltonian, while the other is orthogonal to the levels of the same function regarded as a potential. One can also see this directly, since
1.2 Contraction property of the flow
Consider the gradient flow (5.2).
Proposition 5.10
Let \(d:G\times G\rightarrow {{\mathbb {R}}}\) denote the Riemannian distance function induced by the right-invariant metric associated with the gradient flow (5.2). Let \(\gamma :[0,T)\rightarrow G\) be a solution curve. Then, for all \(t\in [0,T)\)
Proof
First notice that
for the vector field v(t) generating the curve \(\gamma (t)\). Denote by \(\Vert v\Vert ^2_A:= \langle \!\langle v,v \rangle \!\rangle _e = \left\langle Av,v \right\rangle \). Since any curve between \(\gamma (0)\) and \(\gamma (t)\) cannot exceed the length of the geodesic between the points we have
where the last inequality is Cauchy–Schwartz on \(L^2([0,t])\). \(\square \)
Remark 5.11
Theorem 5.10 in relation to the IPM flow then gives the following result: the “Arnold fluid-distance” between \(\varphi (0)\) and \(\varphi (t)\) is bounded by
Appendix B: A Direct Continuous Toda Limit
Start from the Hamiltonian for the finite-dimensional Toda lattice in \((q_j,p_j)\) variables. Take the limit \(n\rightarrow \infty \) to obtain a continuous system on \(T^*\textrm{Dens}({\mathbb {R}})\). Under this limit we have
where \(\varphi \) is the diffeomorphism describing how each point on the line has moved. Since we also have \(\varphi '(z) > 0\) and therefore \(\rho = \textrm{Jac}(\varphi ^{-1}) = 1/\varphi '\circ \varphi \), we get the potential
where the potential V is
The gradient of U is computed as
Newton’s equations on the space of diffeomorphisms is thereby
The fluid formulation of this system is
Let us now make the following change of variables
Since the density is uniform, the variable b is interpreted physically as the Lagrangian momentum. Direct calculations now yield
and, likewise,
This is the same equation as in the first continuous limit, which can be rewritten as
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.
About this article
Cite this article
Khesin, B., Modin, K. The Toda Flow as a Porous Medium Equation. Commun. Math. Phys. 401, 1879–1898 (2023). https://doi.org/10.1007/s00220-023-04680-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00220-023-04680-2