05.14.06 · symplectic / topological-hydrodynamics

KdV and Camassa-Holm as geodesics on the Bott-Virasoro group

shipped3 tiersLean: none

Anchor (Master): Ovsienko-Khesin 1987 *Korteweg-de Vries superequation as an Euler equation* (Funct. Anal. Appl. 21, 329-331, originator of the KdV-as-Bott-Virasoro-geodesic theorem); Misiołek 1998 *A shallow water equation as a geodesic flow on the Bott-Virasoro group* (J. Geom. Phys. 24, 203-208); Camassa-Holm 1993 *An integrable shallow water equation with peaked solitons* (Phys. Rev. Lett. 71, 1661-1664); Khesin-Misiołek 2003 *Euler equations on homogeneous spaces and Virasoro orbits* (Adv. Math. 176, 116-144); Segal 1981 *Unitary representations of some infinite dimensional groups* (Comm. Math. Phys. 80, 301-342); Arnold-Khesin *Topological Methods in Hydrodynamics* (Springer Applied Math. Sci. 125, 2nd ed. 2021) Ch. I §10-§11

Intuition Beginner

A soliton is a solitary water wave that keeps its shape as it travels, and the Korteweg-de Vries equation is the famous equation that governs it. Written out, it says $u_{t} + 3 u u_{x} + u_{xxx} = 0$ : the wave height $u$ changes in time through a steepening term and a dispersive term. For more than a century this equation was studied as a piece of analysis. The surprise, found by Ovsienko and Khesin in 1987, is that it is secretly a statement in geometry.

The geometric picture is this. Imagine all the ways to smoothly rearrange the points of a circle. These rearrangements form a single huge curved space, an infinite-dimensional group of circle motions. On a curved space one can ask for the straightest possible path between two configurations: a geodesic, the analogue of a great circle on a globe. Ovsienko and Khesin showed that the KdV equation is exactly the equation for such a straightest path, once the space is given the right notion of length.

There is one extra ingredient, and it is the heart of the matter. The plain group of circle motions does not produce the dispersive $u_{xxx}$ term on its own. One must enlarge the group by a single extra dimension, a central extension, which records a subtle twisting of the rearrangements. That one extra dimension is precisely what manufactures the dispersion of the soliton.

The same group carries a second natural way of measuring length. With that second metric the straightest paths obey a different equation, the Camassa-Holm equation, a shallow-water model whose waves can develop sharp peaks.

Visual Beginner

Picture a curved surface shaped like a vast dome, standing in for the group of circle rearrangements. A marble released on a dome rolls along a geodesic, the path it would take with no sideways push. Each point of the dome is one way of rearranging the circle, and the rolling marble traces a family of rearrangements unfolding in time. The KdV soliton is the shadow this rolling marble casts back onto the circle.

The extra vertical axis in the picture is the central direction, the single added dimension of the extension. Tilt the metric on the dome from one shape to another and the same marble now rolls along a different straightest path: that second path is the Camassa-Holm wave, which can sharpen into a peak.

Worked example Beginner

Take the simplest soliton solution of the KdV equation $u_{t} + 3 u u_{x} + u_{xxx} = 0$ and watch it move. The single-soliton profile is a travelling bump $$ u(x, t) = \frac{c}{2}, \operatorname{sech}^2!\left( \frac{\sqrt{c}}{2},(x - c t) \right), $$ a hump of height $c /2$ that glides to the right at speed $c$ without changing shape.

Step 1. Read off the dependence on $x - c t$ . The whole profile depends on space and time only through the combination $x - c t$ , so the bump is rigid: it simply slides. Taller bumps (larger $c$ ) are also narrower and faster. This is the signature of a balance between the steepening term $3 u u_{x}$ , which tries to pile the wave up, and the dispersive term $u_{xxx}$ , which tries to spread it out.

Step 2. See where the geometry enters. In the geodesic picture, this travelling bump is the shadow of a rolling motion on the group of circle rearrangements. The rigid travel of the bump corresponds to the motion settling onto a steady rolling pattern, the geometric counterpart of a wave that neither grows nor decays.

Step 3. Note the role of the extra dimension. If the central direction were removed, the dispersive term $u_{xxx}$ would vanish and the bump could not hold together; it would steepen and break. The single added dimension of the central extension is the geometric source of the balance that lets the soliton survive.

What this tells us: a famous closed-form wave from analysis is the same object as a steady straightest path on a curved infinite-dimensional group.

Check your understanding Beginner

Exercise (easy, multiple choice).

In the geodesic interpretation of the KdV equation, the dispersive term $u_{xxx}$ comes from

A. the steepening of the wave
B. the central extension that enlarges the group of circle motions by one dimension
C. the choice of starting configuration
D. friction in the fluid

Hint

The plain group of circle rearrangements does not produce the dispersion on its own; one extra ingredient is needed.

Answer

B. The dispersive term is manufactured by the central extension — the one extra dimension added to the group of circle motions. Feedback-correct: correct; this central term is the geometric fingerprint of the Bott-Virasoro extension. Feedback-wrong: steepening is the other, nonlinear term $3 u u_{x}$ ; the starting configuration only sets which geodesic one travels along; and the equation describes an ideal dispersive balance with no friction at all.

Exercise (easy, true-false).

True or false: the Camassa-Holm equation and the KdV equation are geodesic equations on the very same group, differing only in the choice of metric used to measure length.

Hint

Both are geodesic equations on the enlarged group of circle motions; what changes between them is how length is measured.

Answer

True. Both equations are straightest-path equations on the same enlarged group of circle rearrangements. KdV is the geodesic equation for one way of measuring length (the $L^{2}$ metric); Camassa-Holm is the geodesic equation for a second way (the $H^{1}$ metric, which also penalises how fast the wave varies in space). Feedback: changing the metric changes which paths count as straightest, so a single curved group hosts a whole family of wave equations, one per metric.

Formal definition Intermediate+

Notation. Let $Diff (S^{1})$ denote the group of smooth orientation-preserving diffeomorphisms of the circle $S^{1} = R /2 π Z$ , a Fréchet Lie group. Its Lie algebra is the space $Vect (S^{1})$ of smooth vector fields $f (x) \partial_{x}$ on the circle, written for brevity as functions $f \in C^{\infty} (S^{1})$ , with the Lie bracket $$ [f, g] = f g' - f' g $$ (the negative of the commutator of the corresponding flows, in the convention that makes the Euler-Arnold sign come out for a right-invariant metric). Throughout, a prime denotes $\partial_{x}$ , and $⟨ f, g ⟩_{L^{2}} = \int_{S^{1}} f g d x$ .

Definition (Bott-Virasoro cocycle and the Virasoro algebra). The Gelfand-Fuchs / Bott-Virasoro cocycle is the bilinear map $c : Vect (S^{1}) \times Vect (S^{1}) \to R$ , $$ c(f, g) = \frac{1}{2}\int_{S^1} f'(x), g''(x), dx . $$ It is antisymmetric and satisfies the cocycle identity, so it defines a one-dimensional central extension. The Virasoro algebra is $$ \widehat{\operatorname{Vect}}(S^1) = \operatorname{Vect}(S^1) \oplus \mathbb{R}, \qquad \big[(f, a), (g, b)\big] = \big([f, g],; c(f, g)\big) , $$ with the $R$ summand central 03.11.03; the cocycle is the infinitesimal trace of the group-level Bott cocycle on the Virasoro group $Diff (S^{1})$ , the central extension of $Diff (S^{1})$ by $R$ 03.11.01.

Definition (Euler-Arnold equation for a right-invariant metric). Let $G$ be a Lie group with Lie algebra $g$ , and let $⟨ \cdot, \cdot ⟩_{A}$ be an inner product on $g$ given by a symmetric positive inertia operator $A : g \to g^{*}$ , so $⟨ u, w ⟩_{A} = ⟨ A u, w ⟩$ . Extend it to a right-invariant (weak Riemannian) metric on $G$ . A geodesic of this metric, expressed through its right-logarithmic velocity $u (t) \in g$ and the momentum $m = A u \in g^{*}$ , satisfies the Euler-Arnold equation 05.09.05 $$ \dot{m} = -\operatorname{ad}^_u m , \qquad m = A u , $$ where $\operatorname{ad}^ $i s t h eco a d j o in t a c t i o n d u a l t o$ \operatorname{ad}_u w = [u, w]$.

Definition (the two metrics on Virasoro). On the Virasoro algebra $Vect (S^{1})$ take inner products of the form $⟨(f, a), (g, b)⟩ = \int_{S^{1}} (A f) g d x + ab$ , where $A$ is a differential inertia operator acting on the $Vect (S^{1})$ part:

the $L^{2}$ metric, with $A = Id$ ;
the $H^{1}$ metric, with $A = Id - \partial_{x}^{2}$ .

The Euler-Arnold geodesic equation of the right-invariant extension of the $L^{2}$ metric is the KdV equation; that of the $H^{1}$ metric is the Camassa-Holm equation. The momentum variable in the $H^{1}$ case, $m = (Id - \partial_{x}^{2}) u = u - u_{xx}$ , is the CH momentum, whose peakon (sharp-peaked) configurations correspond to sums of Dirac masses.

Counterexamples to common slips

The cocycle is not optional decoration. Dropping $c$ collapses $Vect (S^{1})$ to $Vect (S^{1})$ , and the $L^{2}$ Euler-Arnold equation becomes the inviscid Burgers equation $u_{t} + 3 u u_{x} = 0$ , which steepens and breaks. The term $u_{xxx}$ is the cocycle's contribution alone.
Left versus right invariance flips a sign. The fluid-mechanical Euler-Arnold convention uses a right-invariant metric on a diffeomorphism group; choosing left invariance reverses the sign of $ad^{*}$ and changes the equation. The KdV identification is for the right-invariant extension.
The central charge is a coordinate, not a parameter to tune. In the geodesic reduction the central component of the momentum is conserved and enters as a fixed coefficient $c_{0}$ multiplying $u_{xxx}$ ; rescaling $u$ and $x$ normalises it, so KdV with any nonzero dispersion coefficient is the same geodesic flow up to rescaling.
Camassa-Holm is the same group, not a new one. The shift from KdV to CH is entirely in the inertia operator $A$ , hence in the metric; the underlying group and its coadjoint action are unchanged.

Key derivation Intermediate+

Theorem (Ovsienko-Khesin 1987; Misiołek 1998). Let $Diff (S^{1})$ be the Virasoro group with right-invariant metric induced by the inner product $$ \big\langle (f, a), (g, b)\big\rangle = \int_{S^1} (A f), g, dx + a b $$ on $Vect (S^{1})$ . Writing the geodesic velocity as $(u, c_{0})$ with $c_{0}$ the conserved central component, the Euler-Arnold geodesic equation is $$ m_t = -\big( 2 u_x m + u, m_x + c_0, u_{xxx} \big), \qquad m = A u . $$ For $A = Id$ (the $L^{2}$ metric) this is the Korteweg-de Vries equation $u_{t} + 3 u u_{x} + c_{0} u_{xxx} = 0$ . For $A = Id - \partial_{x}^{2}$ (the $H^{1}$ metric) it is the Camassa-Holm equation $m_{t} + 2 u_{x} m + u m_{x} = 0$ with $m = u - u_{xx}$ .

Proof. Compute the coadjoint action on $Vect (S^{1})^{*}$ . Pair a momentum $(m, c_{0}) \in Vect (S^{1})^{*} \oplus R$ with a velocity $(g, b)$ by $⟨(m, c_{0}), (g, b)⟩ = \int_{S^{1}} m g d x + c_{0} b$ , identifying $Vect (S^{1})^{*}$ with quadratic differentials $m (x) (d x)^{2}$ . By definition of the coadjoint action, $$ \big\langle \operatorname{ad}^{(f,a)}(m, c_0),, (g, b)\big\rangle = \big\langle (m, c_0),, \operatorname{ad}{(f,a)}(g,b)\big\rangle = \big\langle (m, c_0),, ([f,g],, c(f,g))\big\rangle . $$ The right side equals $\int_{S^{1}} m (f g^{'} - f^{'} g) d x + c_{0} \cdot \frac{1}{2} \int_{S^{1}} f^{'} g^{''} d x$ . Integrate by parts to move all derivatives off $g$ . For the bracket term, $$ \int m (f g' - f' g), dx = \int \big( -(m f)' - m f' \big) g, dx = \int \big( - 2 m f' - m' f \big) g, dx . $$ For the cocycle term, two integrations by parts give $\frac{1}{2} \int f^{'} g^{''} d x = \frac{1}{2} \int f^{'''} g d x$ . Collecting the coefficient of $g$ , $$ \operatorname{ad}^{(f, a)}(m, c_0) = \Big( 2 m f' + m' f + \tfrac{1}{2} c_0, f''' ,; 0\Big) , $$ the central component being annihilated because the extension is central (so $ad^{*}$ leaves $c_{0}$ fixed and $c_{0}$ is a conserved Casimir of the motion). Substituting $f = u = A^{- 1} m$ into the Euler-Arnold equation $\dot{m} = -\operatorname{ad}^*{(u, c_0)}(m, c_0)$ yields $$ m_t = -\big( 2 m u_x + u, m_x + \tfrac{1}{2} c_0, u_{xxx}\big), $$ which after absorbing the factor $\frac{1}{2}$ into the normalisation of $c_{0}$ is the stated equation.

For $A = Id$ the momentum equals the velocity, $m = u$ , so $2 m u_{x} + u m_{x} = 2 u u_{x} + u u_{x} = 3 u u_{x}$ , and the equation becomes $u_{t} + 3 u u_{x} + c_{0} u_{xxx} = 0$ , the KdV equation. The dispersive third-derivative term is exactly the cocycle term $c_{0} u_{xxx}$ ; with $c_{0} = 0$ one recovers inviscid Burgers.

For $A = Id - \partial_{x}^{2}$ the momentum is $m = u - u_{xx}$ . The central term is absent (the $H^{1}$ Camassa-Holm reduction is taken with $c_{0} = 0$ , on the non-extended part of the metric), and $m_{t} + 2 u_{x} m + u m_{x} = 0$ with $m = u - u_{xx}$ is the Camassa-Holm equation. $□$

Bridge. The computation above is the foundational reason the two equations share a home: both are the Euler-Arnold equation $\overset{m}{˙} = - ad_{u}^{*} m$ on the Virasoro group, and the only difference is the inertia operator $A$ that converts velocity to momentum. This is exactly the same coadjoint-action machinery that the Euler-Arnold framework 05.09.05 applies to the volume-preserving diffeomorphism group to produce the equations of an ideal fluid; the present unit builds toward the realisation that integrable PDEs are not a separate species but the geodesic flows of particular metrics on particular groups. The central term that distinguishes KdV from Burgers generalises the abstract central extension 03.11.01 to a concrete analytic effect, the soliton dispersion. The dual statement — that the resulting Hamiltonian structure is the Lie-Poisson bracket on the Virasoro dual — appears again in the Advanced results, where the bi-Hamiltonian pencil of KdV 05.09.08 is recovered as a frozen-plus-linear pair of brackets, putting these together into a single coadjoint-orbit picture; the bridge is that the dispersion, the integrability, and the geometry are three faces of one central extension.

Exercises Intermediate+

Exercise 4 (medium, short-answer).

Explain why removing the central extension turns the $L^{2}$ Euler-Arnold equation into the inviscid Burgers equation, and what is lost.

Hint

Set $c_{0} = 0$ in the reduced equation and ask which term disappears.

Answer

Without the central extension the cocycle is gone, so $c_{0} = 0$ and the dispersive term $c_{0} u_{xxx}$ vanishes. The $L^{2}$ Euler-Arnold equation on the bare group $Diff (S^{1})$ is then $u_{t} + 3 u u_{x} = 0$ , the inviscid Burgers equation. What is lost is dispersion: Burgers solutions steepen until the gradient becomes infinite and a shock forms in finite time, whereas the KdV dispersion balances the steepening and supports smooth travelling solitons. The central extension is thus the precise geometric origin of the smoothing dispersion. Rubric: full credit for identifying the lost term, naming Burgers, and explaining shock formation versus dispersive balance.

Exercise 5 (medium, short-answer).

Identify the Camassa-Holm momentum $m = u - u_{xx}$ and explain in what sense a peakon $u (x) = e^{- ∣ x ∣}$ (on the line) is a delta-function of momentum.

Hint

Apply the operator $Id - \partial_{x}^{2}$ to $e^{- ∣ x ∣}$ in the distributional sense; recall $- \partial_{x}^{2} e^{- ∣ x ∣}$ produces a delta.

Answer

For $u (x) = e^{- ∣ x ∣}$ , away from the origin $u_{xx} = u$ , so $u - u_{xx} = 0$ there. At the origin the kink in $u$ contributes a distributional second derivative: $\partial_{x}^{2} e^{- ∣ x ∣} = e^{- ∣ x ∣} - 2 δ (x)$ , hence $m = u - u_{xx} = e^{- ∣ x ∣} - (e^{- ∣ x ∣} - 2 δ) = 2 δ (x)$ . So the peakon is a single Dirac mass of CH momentum. This is why multi-peakon solutions reduce to a finite-dimensional Hamiltonian system for the peak positions and strengths: the momentum is a sum of point masses, and the CH flow moves them. Rubric: full credit for the distributional computation $m = 2 δ$ and the point-mass interpretation.

Exercise 6 (hard, short-answer).

Sketch why peakon solutions show that the $H^{1}$ geodesic flow on the Virasoro group is geometrically incomplete (geodesics can leave the smooth-diffeomorphism domain in finite time).

Hint

A geodesic of a right-invariant metric on $Diff (S^{1})$ stays in the group only as long as its velocity profile remains a smooth diffeomorphism-generating field. Wave breaking destroys smoothness.

Answer

A geodesic of the $H^{1}$ right-invariant metric is a curve $t \mapsto φ (t) \in Diff (S^{1})$ whose velocity is the Camassa-Holm field $u (t)$ . Constantin-Escher 1998 proved that for a large class of smooth initial data the slope $u_{x}$ blows up to $- \infty$ in finite time while $u$ itself stays bounded — wave breaking. At that instant the velocity field is no longer smooth, so the geodesic has run off the smooth part of $Diff (S^{1})$ : the metric is geodesically incomplete. The peakon $e^{- ∣ x ∣}$ is the prototype of the resulting non-smooth profile, a continuous field with a corner. Geometrically, the curvature of the $H^{1}$ metric (computed by Misiołek 1998) is sign-indefinite, and the breaking is the analytic shadow of geodesics being focused off the smooth stratum. Rubric: full credit for connecting wave breaking to loss of smoothness of the geodesic, citing Constantin-Escher, and naming the peakon as the limiting profile.

Exercise 7 (hard, short-answer).

Explain how the second Hamiltonian structure of KdV is the linear Lie-Poisson bracket on the dual Virasoro algebra, and what plays the role of the first structure.

Hint

The Lie-Poisson bracket on $g^{*}$ is ${F, G} (m) = ⟨ m, [d F, d G]⟩$ ; for Virasoro the bracket includes the cocycle term.

Answer

On the dual Virasoro algebra the Lie-Poisson bracket is ${F, G} (m, c_{0}) = \int m [d F, d G] d x + c_{0} c (d F, d G)$ , where $d F, d G$ are the variational derivatives viewed as elements of $Vect (S^{1})$ . Written as an operator on densities this is the Gardner-Zakharov-Faddeev / Magri second structure $B_{2} = c_{0} \partial_{x}^{3} + (m \partial_{x} + \partial_{x} m)$ — exactly the cocycle term $\partial_{x}^{3}$ plus the coadjoint term. The first structure of the KdV bi-Hamiltonian pair 05.09.08 is the constant (frozen) bracket $B_{1} = \partial_{x}$ , obtained by freezing the Lie-Poisson bracket at a constant momentum. The two brackets are compatible (any linear combination is Poisson), and Magri's theorem then generates the KdV hierarchy. Thus the bi-Hamiltonian structure, discovered analytically, is the coadjoint-orbit Lie-Poisson geometry of the Virasoro algebra. Rubric: full credit for identifying $B_{2}$ with the Virasoro Lie-Poisson bracket including the cocycle, naming the frozen bracket as $B_{1}$ , and noting compatibility.

Lean formalization Intermediate+

Mathlib contains neither the Fréchet Lie group $Diff (S^{1})$ , the Virasoro central extension, nor the Euler-Arnold reduction on a group with a right-invariant weak metric, so the central theorem cannot be stated. A schematic aspirational signature, in the absence of that infrastructure, would read:

-- Aspirational, not currently realisable in Mathlib.
theorem kdv_is_bott_virasoro_geodesic
    (u : ℝ → C∞(circle, ℝ)) (c0 : ℝ)
    (hgeo : IsEulerArnoldGeodesic VirasoroGroup L2Metric (u, c0)) :
  ∀ t, deriv (u t) ... = - (3 • (u t) * deriv (u t) + c0 • iteratedDeriv 3 (u t)) :=
sorry

Even this statement presupposes VirasoroGroup (the central extension of Diff(circle)), L2Metric as a right-invariant weak Riemannian metric, IsEulerArnoldGeodesic as the geodesic predicate after reduction, and a coadjoint-action API on the dual algebra — none of which exist. A faithful Camassa-Holm statement would additionally require the inertia operator $Id - \partial_{x}^{2}$ as an invertible map on a Sobolev scale and the Constantin-Escher wave-breaking theorem to express geodesic incompleteness. The unit's correctness gates are therefore the human-review surfaces documented in the unit metadata: the cocycle normalisation, the inertia operators $A = Id$ and $A = Id - \partial_{x}^{2}$ , the coadjoint computation producing $u_{xxx}$ from the central term, and the attributions to Ovsienko-Khesin 1987 and Misiołek 1998.

Advanced results Master

The Ovsienko-Khesin theorem in context. Ovsienko and Khesin 1987 ^{[Ovsienko-Khesin 1987]} proved that the KdV equation is the Euler-Arnold equation of the right-invariant $L^{2}$ metric on the Bott-Virasoro group, completing the programme Arnold had opened in 1966 by identifying the ideal-fluid Euler equations with geodesics on the volume-preserving diffeomorphism group 05.09.05. The conceptual content is that the central extension is not a formal nicety: its sole physical effect, after reduction, is to insert the dispersive operator $\partial_{x}^{3}$ into the equation of motion. Without the cocycle the reduction yields inviscid Burgers; with it, KdV. The dispersion of the soliton is therefore a cohomological invariant of $Diff (S^{1})$ , the generator of $H^{2} (Vect (S^{1}); R)$ , made dynamical.

Misiołek's Camassa-Holm theorem. Misiołek 1998 ^{[Misiołek 1998]} showed that the Camassa-Holm equation, introduced by Camassa and Holm in 1993 ^{[Camassa-Holm 1993]} as an integrable shallow-water model, is the Euler-Arnold geodesic equation of the right-invariant $H^{1}$ metric on the same Virasoro group. The shift of metric from $L^{2}$ to $H^{1}$ changes the inertia operator from the identity to $Id - \partial_{x}^{2}$ and replaces dispersive smoothing by a transport of the momentum $m = u - u_{xx}$ . Misiołek computed the sectional curvature of the $H^{1}$ metric and found it sign-indefinite, in contrast with the predominantly negative curvature of the $L^{2}$ fluid metric; the indefiniteness is the geometric companion of the coexistence of global smooth solutions with finite-time wave breaking.

Wave breaking and geodesic incompleteness. Constantin and Escher 1998 ^{[Constantin-Escher 1998]} proved that Camassa-Holm solutions can break: $u$ remains bounded while $u_{x} \to - \infty$ in finite time. In the geodesic dictionary this is the statement that the $H^{1}$ right-invariant metric on $Diff (S^{1})$ is geodesically incomplete — a geodesic can reach the boundary of the smooth-diffeomorphism stratum at finite arclength. The peakon $u = c e^{- ∣ x - c t ∣}$ , a travelling wave with a corner, is the canonical broken profile, and multipeakon configurations reduce the partial differential equation to a finite-dimensional integrable Hamiltonian system for the peak positions and momenta.

Lie-Poisson recovery of the bi-Hamiltonian structure. The momentum equation lives on the dual Virasoro algebra, where the natural Poisson structure is the linear Lie-Poisson bracket ${F, G} (m, c_{0}) = \int m [d F, d G] d x + c_{0} c (d F, d G)$ . As an operator pencil this is the second Gardner-Zakharov-Faddeev / Magri structure $B_{2} = c_{0} \partial_{x}^{3} + m \partial_{x} + \partial_{x} m$ of KdV; the first structure $B_{1} = \partial_{x}$ is the same bracket frozen at a constant momentum. The compatibility of $B_{1}$ and $B_{2}$ — the defining property of a bi-Hamiltonian system 05.09.08 — is, in this language, the affine-linear dependence of the Lie-Poisson bracket on the momentum, which is automatic for any Lie algebra. Magri's recursion $B_{2} = R B_{1}$ then produces the KdV hierarchy, and the conserved densities are the Casimirs and Hamiltonians on Virasoro coadjoint orbits.

Coadjoint orbits and the Kirillov picture. The symplectic leaves of the Lie-Poisson structure are the coadjoint orbits of the Virasoro group, classified by Segal 1981 ^{[Segal 1981]} and Kirillov via the projective structures and Hill's-equation monodromy attached to a quadratic differential $m (d x)^{2} + c_{0}$ . The KdV flow is a Hamiltonian flow on these orbits, and the integrals of motion are orbit invariants. This embeds the analytic integrability of KdV in the orbit method: the action variables are the monodromy data of the associated Hill operator $c_{0} \partial_{x}^{2} + m$ , the Schrödinger operator whose isospectral deformation is the Lax form of KdV.

The Hunter-Saxton and a family of metrics. Khesin and Misiołek 2003 ^{[Khesin-Misiołek 2003]} showed that the homogeneous $\dot{H}^{1}$ metric ( $A = - \partial_{x}^{2}$ ) on the Virasoro group yields the Hunter-Saxton equation $(u_{t} + u u_{x})_{x} = \frac{1}{2} u_{x}^{2}$ , an integrable model of liquid-crystal director fields, and that this metric has constant positive curvature, making the corresponding geodesic flow especially transparent. The pattern is that an entire family of integrable shallow-water and director-field equations — KdV, Camassa-Holm, Hunter-Saxton — are the geodesic equations of a one-parameter family of right-invariant metrics on a single group, distinguished only by their inertia operators.

Synthesis. Putting these together, the foundational reason KdV, Camassa-Holm, and Hunter-Saxton stand together is that each is the Euler-Arnold geodesic equation on the one group $Diff (S^{1})$ , and the choice that separates them is the inertia operator $A$ alone. This is exactly the geometric content the analytic and spectral theories had not exposed: the soliton dispersion of KdV is the cocycle term $\partial_{x}^{3}$ , and the bi-Hamiltonian pencil of 05.09.08 generalises into the frozen-plus-linear pair of Lie-Poisson brackets on the Virasoro dual, so the second Magri bracket is dual to the coadjoint action while the first is its constant-coefficient freezing. The central insight is that the abstract central extension 03.11.01 and the integrable hierarchy are the same datum read twice: the bridge is the Euler-Arnold reduction, which appears again in the ideal-fluid story of 05.09.05 and builds toward a uniform account of integrable PDEs as geodesic flows. Camassa-Holm's wave breaking, finally, is dual to the incompleteness of the $H^{1}$ metric, binding the analysis of blow-up to the geometry of geodesics leaving the smooth stratum.

Full proof set Master

Proposition (cocycle identity). The map $c (f, g) = \frac{1}{2} \int_{S^{1}} f^{'} g^{''} d x$ is a Lie-algebra two-cocycle on $Vect (S^{1})$ : it is antisymmetric and satisfies $c ([f, g], h) + c ([g, h], f) + c ([h, f], g) = 0$ for all $f, g, h \in C^{\infty} (S^{1})$ .

Proof. Antisymmetry was shown by one integration by parts: $\int f^{'} g^{''} d x = - \int f^{''} g^{'} d x = - \int g^{'} f^{''} d x$ . For the cocycle (Jacobi) identity, write $[f, g] = f g^{'} - f^{'} g$ . A computation of $c ([f, g], h)$ expands $([f, g])^{'} = (f g^{'} - f^{'} g)^{'} = f g^{''} - f^{''} g$ and pairs it against $\frac{1}{2} h^{''}$ . Summing the three cyclic terms, every monomial in $f, g, h$ and their derivatives that appears does so with a cancelling partner after integration by parts on the circle, where all boundary contributions vanish by periodicity. The total integrand is a total derivative, so the cyclic sum integrates to zero. Hence $c$ is a cocycle and defines the central extension $Vect (S^{1})$ 03.11.01. $□$

Proposition (coadjoint action on the Virasoro dual). Identifying $\operatorname{Vect}(S^1)^ $w i t h q u a d r a t i c d i f f er e n t ia l s v ia t h e$ L^2 $p ai r in g, t h eco a d j o in t a c t i o n o f$ (f, a) \in \widehat{\operatorname{Vect}}(S^1) $o n$ (m, c_0) \in \widehat{\operatorname{Vect}}(S^1)^$ is $$ \operatorname{ad}^*_{(f, a)}(m, c_0) = \big( 2 f' m + f m' + \tfrac{1}{2} c_0, f''',; 0 \big). $$

Proof. By definition $⟨ ad_{(f, a)}^{*} (m, c_{0}), (g, b)⟩ = ⟨(m, c_{0}), [(f, a), (g, b)]⟩ = \int m (f g^{'} - f^{'} g) d x + c_{0} \cdot \frac{1}{2} \int f^{'} g^{''} d x$ . Integrate by parts to expose the coefficient of $g$ . From $\int m f g^{'} d x = - \int (m f)^{'} g d x = - \int (m^{'} f + m f^{'}) g d x$ and $- \int m f^{'} g d x = - \int m f^{'} g d x$ , the bracket term contributes $- \int (m^{'} f + 2 m f^{'}) g d x$ ; including the overall sign in the pairing convention $\overset{m}{˙} = - ad^{*}$ , the surviving coefficient is $2 f^{'} m + f m^{'}$ . The cocycle term gives $\frac{c _{0}}{2} \int f^{'} g^{''} d x = \frac{c _{0}}{2} \int f^{'''} g d x$ , contributing $\frac{c _{0}}{2} f^{'''}$ . The central component is zero because $b$ does not appear on the right, so $c_{0}$ is conserved. $□$

Proposition (KdV reduction). With the $L^{2}$ inertia operator $A = Id$ , so that $m = u$ , the Euler-Arnold equation $\dot m = -\operatorname{ad}^{(u, c_0)}(m, c_0) $i s t h eK d V e q u a t i o n$ u_t + 3 u u_x + \tfrac{1}{2} c_0, u{xxx} = 0$.*

Proof. Substitute $m = u$ and $f = u$ in the coadjoint formula: $ad_{(u, c_{0})}^{*} (u, c_{0}) = (2 u^{'} u + u u^{'} + \frac{1}{2} c_{0} u^{'''}, 0) = (3 u u_{x} + \frac{1}{2} c_{0} u_{xxx}, 0)$ . The Euler-Arnold equation $u_{t} = - (3 u u_{x} + \frac{1}{2} c_{0} u_{xxx})$ rearranges to KdV. Rescaling $x$ and $u$ normalises $\frac{1}{2} c_{0}$ to $1$ . $□$

Proposition (Camassa-Holm reduction). With the $H^{1}$ inertia operator $A = Id - \partial_{x}^{2}$ and central component switched off, the Euler-Arnold equation is $m_{t} + 2 u_{x} m + u m_{x} = 0$ with $m = u - u_{xx}$ , the Camassa-Holm equation.

Proof. The coadjoint formula with $c_{0} = 0$ reads $ad_{u}^{*} m = 2 u_{x} m + u m_{x}$ . Euler-Arnold gives $m_{t} = - (2 u_{x} m + u m_{x})$ , i.e. $m_{t} + 2 u_{x} m + u m_{x} = 0$ . Substituting $m = u - u_{xx}$ and expanding reproduces the standard form $u_{t} - u_{t xx} + 3 u u_{x} = 2 u_{x} u_{xx} + u u_{xxx}$ . The momentum $m = (Id - \partial_{x}^{2}) u$ is recovered from $u$ by convolution with the Green's function of $Id - \partial_{x}^{2}$ , the peakon kernel $\frac{1}{2} e^{- ∣ x ∣}$ on the line. $□$

Proposition (peakon as point momentum). On the line, the peakon profile $u (x) = e^{- ∣ x ∣}$ has Camassa-Holm momentum $m = u - u_{xx} = 2 δ_{0}$ , a single Dirac mass at the origin.

Proof. For $x \neq = 0$ , $u_{xx} = e^{- ∣ x ∣} = u$ , so $u - u_{xx} = 0$ . Across $x = 0$ the first derivative jumps from $+ 1$ to $- 1$ , a jump of $- 2$ , so $u_{xx}$ contains $- 2 δ_{0}$ in the distributional sense: $u_{xx} = e^{- ∣ x ∣} - 2 δ_{0}$ . Therefore $m = u - u_{xx} = e^{- ∣ x ∣} - (e^{- ∣ x ∣} - 2 δ_{0}) = 2 δ_{0}$ . A sum of translated peakons has momentum a sum of point masses, reducing the CH dynamics to a finite-dimensional Hamiltonian system. $□$

Connections Master

Upstream — the Euler-Arnold engine. This unit is a direct application of the Euler-Arnold reduction 05.09.05: the same theorem that turns geodesics on the volume-preserving diffeomorphism group into the ideal-fluid Euler equations turns geodesics on the Virasoro group into KdV and Camassa-Holm. The shift from a metric on $SDiff (M)$ to a metric on the central extension of $Diff (S^{1})$ is the only structural change, which is why the topological-hydrodynamics chapter keeps the Arnold-Khesin geometric framing of both stories together.

Upstream — the central extension. The Bott-Virasoro cocycle is the concrete two-cocycle that builds the central extension 03.11.01 of $Diff (S^{1})$ , and the Virasoro algebra 03.11.03 is its Lie-algebra incarnation. The present unit gives that abstract cohomology class its sharpest dynamical meaning: the generator of $H^{2} (Vect (S^{1}); R)$ is, after Euler-Arnold reduction, the dispersive operator $\partial_{x}^{3}$ that creates the KdV soliton.

Lateral — the bi-Hamiltonian framing. The KdV bi-Hamiltonian structure of 05.09.08 is the same equation seen through a complementary lens: there KdV is a Hamiltonian evolution equation with a Magri pencil of compatible Poisson brackets; here that pencil is recognised as the frozen and the linear Lie-Poisson brackets on the dual Virasoro algebra. The two units describe one object — the second Magri bracket is the Virasoro Lie-Poisson bracket, cocycle term included — so reading them together yields the unification of integrability with the coadjoint-orbit (Kirillov) method.

Lateral — coadjoint orbits and integrable systems. The symplectic leaves of the Virasoro Lie-Poisson structure are coadjoint orbits whose invariants are the monodromy data of an associated Hill operator; this places KdV inside the broader programme of integrable systems realised as Hamiltonian flows on coadjoint orbits, alongside the Toda and other Euler-Arnold hierarchies. The Hunter-Saxton equation (Khesin-Misiołek 2003) sits in the same family, a different right-invariant metric on the same group.

Downstream — geometry of diffeomorphism groups. The sign-indefinite curvature of the $H^{1}$ metric and the geodesic incompleteness signalled by Camassa-Holm wave breaking feed into the geometric analysis of diffeomorphism groups and the study of when right-invariant metrics are complete, a theme that recurs in optimal transport and shape analysis and connects forward to the Wasserstein and Korteweg-Madelung units of the optimal-transport chapter.

Historical & philosophical context Master

The Korteweg-de Vries equation entered mathematics in 1895 as a model of long waves in a shallow canal, and for most of a century its remarkable properties — solitons, infinitely many conservation laws, the inverse-scattering solution — were uncovered by analytic and algebraic means: the Gardner-Greene-Kruskal-Miura inverse-scattering transform of 1967, the Lax pair of 1968, and the Gardner-Zakharov-Faddeev and Magri Hamiltonian structures of the 1970s. None of these revealed why the equation should be so special. Arnold's 1966 reinterpretation of ideal hydrodynamics as geodesic flow on a diffeomorphism group suggested that integrable PDEs might be geodesic equations in disguise, and Ovsienko and Khesin 1987 ^{[Ovsienko-Khesin 1987]} confirmed this for KdV by identifying it as the Euler-Arnold equation on the Bott-Virasoro group. The philosophical point is striking: a single equation that had been studied as analysis, as algebra, and as spectral theory turned out to be, at bottom, the law of inertia — a free particle gliding along a geodesic — on a curved infinite-dimensional group.

The Camassa-Holm equation arrived later, derived by Camassa and Holm in 1993 ^{[Camassa-Holm 1993]} as a shallow-water model retaining stronger nonlinearity than KdV, and notable for its peaked solitons. Misiołek 1998 ^{[Misiołek 1998]} placed it in the same geometric frame by changing the metric from $L^{2}$ to $H^{1}$ , and Constantin and Escher 1998 ^{[Constantin-Escher 1998]} showed that its solutions can break in finite time, the analytic face of a geodesic running off the smooth diffeomorphisms. Segal's 1981 study of the Virasoro group ^{[Segal 1981]} and Khesin and Misiołek's 2003 survey ^{[Khesin-Misiołek 2003]} consolidated the picture, in which an entire family of shallow-water equations is organised by the single geometric idea of a right-invariant metric on the central extension of the circle-diffeomorphism group, the canonical modern reference being the Arnold-Khesin monograph ^{[Arnold-Khesin 2021]}. The Bott cocycle ^{[Bott 1977]} that names the extension was first written down in the study of characteristic classes of diffeomorphism groups, an origin entirely outside fluid mechanics, which makes its reappearance as the source of soliton dispersion all the more telling.

Bibliography Master

@article{OvsienkoKhesin1987,
  author  = {Ovsienko, V. Yu. and Khesin, B. A.},
  title   = {Korteweg-de Vries superequation as an Euler equation},
  journal = {Funct. Anal. Appl.},
  volume  = {21},
  pages   = {329--331},
  year    = {1987}
}

@article{Misiolek1998,
  author  = {Misiołek, G.},
  title   = {A shallow water equation as a geodesic flow on the Bott-Virasoro group},
  journal = {J. Geom. Phys.},
  volume  = {24},
  pages   = {203--208},
  year    = {1998}
}

@article{CamassaHolm1993,
  author  = {Camassa, R. and Holm, D. D.},
  title   = {An integrable shallow water equation with peaked solitons},
  journal = {Phys. Rev. Lett.},
  volume  = {71},
  pages   = {1661--1664},
  year    = {1993}
}

@article{KhesinMisiolek2003,
  author  = {Khesin, B. and Misiołek, G.},
  title   = {Euler equations on homogeneous spaces and Virasoro orbits},
  journal = {Adv. Math.},
  volume  = {176},
  pages   = {116--144},
  year    = {2003}
}

@article{Segal1981,
  author  = {Segal, G.},
  title   = {Unitary representations of some infinite dimensional groups},
  journal = {Comm. Math. Phys.},
  volume  = {80},
  pages   = {301--342},
  year    = {1981}
}

@article{ConstantinEscher1998,
  author  = {Constantin, A. and Escher, J.},
  title   = {Wave breaking for nonlinear nonlocal shallow water equations},
  journal = {Acta Math.},
  volume  = {181},
  pages   = {229--243},
  year    = {1998}
}

@article{Bott1977,
  author  = {Bott, R.},
  title   = {On the characteristic classes of groups of diffeomorphisms},
  journal = {Enseign. Math.},
  volume  = {23},
  pages   = {209--220},
  year    = {1977}
}

@book{ArnoldKhesin2021,
  author    = {Arnold, V. I. and Khesin, B. A.},
  title     = {Topological Methods in Hydrodynamics},
  series    = {Applied Mathematical Sciences},
  volume    = {125},
  edition   = {2nd},
  publisher = {Springer},
  year      = {2021}
}

@book{KhesinWendt2009,
  author    = {Khesin, B. and Wendt, R.},
  title     = {The Geometry of Infinite-Dimensional Groups},
  series    = {Ergebnisse der Mathematik},
  volume    = {51},
  publisher = {Springer},
  year      = {2009}
}

Prerequisites

05.09.05
05.09.08
03.11.03
03.11.01

Tier anchors

beginner: Arnold-Khesin *Topological Methods in Hydrodynamics* (Springer Applied Math. Sci. 125, 2nd ed. 2021) Ch. I §1 informal; Drazin-Johnson *Solitons: an Introduction* (Cambridge 1989) Ch. 1 informal exposition of the KdV soliton
intermediate: Arnold-Khesin *Topological Methods in Hydrodynamics* Ch. I §10-§11; Khesin-Wendt *The Geometry of Infinite-Dimensional Groups* (Springer Ergebnisse 51, 2009) Ch. II §2; Misiołek 1998 *A shallow water equation as a geodesic flow on the Bott-Virasoro group* (J. Geom. Phys. 24) §2-§3
master: Ovsienko-Khesin 1987 *Korteweg-de Vries superequation as an Euler equation* (Funct. Anal. Appl. 21, 329-331, originator of the KdV-as-Bott-Virasoro-geodesic theorem); Misiołek 1998 *A shallow water equation as a geodesic flow on the Bott-Virasoro group* (J. Geom. Phys. 24, 203-208); Camassa-Holm 1993 *An integrable shallow water equation with peaked solitons* (Phys. Rev. Lett. 71, 1661-1664); Khesin-Misiołek 2003 *Euler equations on homogeneous spaces and Virasoro orbits* (Adv. Math. 176, 116-144); Segal 1981 *Unitary representations of some infinite dimensional groups* (Comm. Math. Phys. 80, 301-342); Arnold-Khesin *Topological Methods in Hydrodynamics* (Springer Applied Math. Sci. 125, 2nd ed. 2021) Ch. I §10-§11

References

Ovsienko, Khesin 1987 — Korteweg-de Vries superequation as an Euler equation · Funct. Anal. Appl. 21, 329-331; the theorem that KdV is the Euler-Arnold geodesic equation on the Bott-Virasoro group for the L2 metric
Misiołek 1998 — A shallow water equation as a geodesic flow on the Bott-Virasoro group · J. Geom. Phys. 24, 203-208; Camassa-Holm as the Euler-Arnold geodesic for the H1 right-invariant metric on the Virasoro group, and sectional curvature
Camassa, Holm 1993 — An integrable shallow water equation with peaked solitons · Phys. Rev. Lett. 71, 1661-1664; derivation of the Camassa-Holm equation, peakon solutions, bi-Hamiltonian structure
Khesin, Misiołek 2003 — Euler equations on homogeneous spaces and Virasoro orbits · Adv. Math. 176, 116-144; Hunter-Saxton equation, homogeneous-metric Euler-Arnold equations, coadjoint orbits
Segal 1981 — Unitary representations of some infinite dimensional groups · Comm. Math. Phys. 80, 301-342; geometry of the Virasoro group and the Bott cocycle
Arnold, Khesin — Topological Methods in Hydrodynamics · Springer Applied Math. Sci. 125, 2nd ed. 2021, Ch. I §10-§11 (Bott-Virasoro group, KdV and Camassa-Holm as geodesics)
Constantin, Escher 1998 — Wave breaking for nonlinear nonlocal shallow water equations · Acta Math. 181, 229-243; finite-time blow-up of Camassa-Holm solutions and the geometric loss of smoothness of geodesics
Bott 1977 — On the characteristic classes of groups of diffeomorphisms · Enseign. Math. 23, 209-220; the Bott cocycle and the group-level central extension of Diff(S1)

Estimated time

beginner: 18m
intermediate: 45m
master: 85m