Groups and topology in the Euler

hydrodynamics and KdV

Boris Khesin

Abstract We survey applications of group theory and topology in fluid mechanics

and integrable systems. The main reference for most facts in this paper is [1], see
also details in [4].

1 Euler equations and geodesics.

1.1 The Euler hydrodynamics equation.

Consider an incompressible fluid occupying a domain M in Rn . The fluid motion

is described by a velocity field v(t, x) and a pressure field p(t, x) which satisfy the
classical Euler equation:
∂t v + (v · ∇)v = −∇p, (1)
where div v = 0 and the field v is tangent to the boundary of M. The function p
is defined uniquely modulo an additive constant by the condition that v has zero
divergence at any moment t.
The flow (t, x) 7→ g(t, x) describing the motion of fluid particles is defined by its
velocity field v(t, x):

∂t g(t, x) = v(t, g(t, x)), g(0, x) = x.

Acceleration of particles is given by ∂t2 g(t, x) = (∂t v + (v · ∇)v)(t, g(t, x)), according
to the chain rule, and hence the Euler equation (1) is equivalent to

∂t2 g(t, x) = −(∇p)(t, g(t, x)).

Boris Khesin, Department of Mathematics, University of Toronto, ON M5S 2E4, Canada, e-mail:
[email protected]

1
2 Boris Khesin

The latter form of the Euler equation (for a smooth flow g(t, x)) says that the ac-
celeration of the flow is given by a gradient and hence it is L2 -orthogonal to the set of
volume-preserving diffeomorphisms, which satisfy the incompressibility condition
det(∂x g(t, x)) = 1. More precisely, it is L2 -orthogonal to the tangent to this set, the
space of divergence-free fields. In other words, the fluid motion g(t, x) is a geodesic
line on the set of such diffeomorphisms of the domain M with respect to the in-
duced L2 -metric. Note that this metric is invariant with respect to reparametrizing
the fluid particles, i.e. it is right-invariant on the set of volume-preserving diffeo-
morphisms (a reparametrization of the independent variable is the right action of a
diffeomorphism).
More generally, the Euler equation describes an ideal incompressible fluid fill-
ing an arbitrary Riemannian manifold M, see [1, 5]. It defines the geodesic flow on
the group of volume-preserving diffeomorphisms of M. It turns out that the group-
geodesic point of view, developed in [1] is quite fruitful for topological and qualita-
tive understanding of the fluid motion, as well as for obtaining various quantitative
results related to stability and first integrals of the Euler equation.

1.2 Geodesics on Lie groups.

In [1] V. Arnold suggested a general framework for the Euler equations on an arbi-
trary group, which we recall below. In this framework the Euler equation describes
a geodesic flow with respect to a suitable one-sided invariant Riemannian metric on
the given group.
More precisely, consider a (possibly infinite-dimensional) Lie group G, which
can be thought of as the configuration space of some physical system. (Examples
from [1]: the group SO(3) for a rigid body and the group SDiff(M) of volume-
preserving diffeomorphisms for an ideal fluid filling a domain M.) The tangent space
at the identity of the Lie group G is the corresponding Lie algebra g. Fix some
(positive definite) quadratic form, the energy, on g. We consider right translations of
this quadratic form to the tangent space at any point of the group (the “translational
symmetry” of the energy). This way the energy defines a right-invariant Riemannian
metric on the group G. The geodesic flow on G with respect to this energy metric
represents the extremals of the least action principle, i.e., the actual motions of our
physical system. (For a rigid body one has to consider left translations.)
To describe a geodesic on the Lie group with an initial velocity v(0) = ξ , we
transport its velocity vector at any moment t to the identity of the group (by using
the right translation). This way we obtain the evolution law for v(t), given by a
(non-linear) dynamical system dv/dt = F(v) on the Lie algebra g (Fig.1).

Theorem 1. The system on the Lie algebra g, describing the evolution of the velocity
vector along a geodesic in a right-invariant metric on the Lie group G, is called the
Euler equation corresponding to this metric on G.
Groups and topology in the Euler hydrodynamics and KdV 3

e
G
g(t)

Fig. 1 The vector ξ in the Lie algebra g is the velocity at the identity e of a geodesic g(t) on the
Lie group G.

1.3 Geodesic description for various equations.

A similar Arnold-type description via the geodesic flow on a Lie group can be given
to a variety of conservative dynamical systems in mathematical physics. Below we
list several examples of such systems to emphasize the range of applications of this
approach. The choice of a group G (column 1) and an energy metric E (column 2)
defines the corresponding Euler equations (column 3).

Group Metric Equation

SO(3) < ω , Aω > Euler top

SO(3)+̇R3 quadratic forms Kirchhoff equations for a body in a fluid
SO(n) Manakov0s metrics n−dimensional top
Diff(S1 ) L2 Hopf (or, inviscid Burgers) equation
Virasoro L2 KdV equation
Virasoro H1 Camassa − Holm equation
Virasoro Ḣ 1 Hunter − Saxton (or Dym) equation
SDiff(M) L2 Euler ideal fluid
SDiff(M)+̇SVect(M) L 2 + L2 Magnetohydrodynamics
1
Maps(S , SO(3)) H −1 Landau − Lifschits equation

In some cases these systems turn out to be not only Hamiltonian, but also bi-
hamiltonian. More detailed descriptions and references can be found in the book
[4].

2 Topology of steady flows.

2.1 Arnold’s classification of steady fluid flows.

The stationary Euler equation in the domain M has the form

(v · ∇)v = −∇p
4 Boris Khesin

on a divergence-free vector field v. In 3D this equation can be rewritten as follows:

v × curl v = −∇α ,

i.e. the cross-product of the fields v and curl v is a potential vector field. Here α =
p + |v|2/2 is called the Bernoulli function. (Another way to express this is to say that
the field v commutes with its vorticity curl v. The latter commutativity condition is
valid in any dimension.)

Theorem 2. [2, 3] Let M be a three-dimensional manifold without boundary. Then

all non-critical level sets of α are 2-tori. Furthermore, both fields v and curl v are
tangent to these levels and define there the R2 -action.
On a manifold M with boundary, the α -level sets are either 2-tori or annuli. On
tori the flow lines are either all closed or all dense, and on annuli all flow lines are
closed.

The proof of the theorem is based on the observation that v is always tangent to
the level sets of α , i.e. the function α is a first integral of the equation. On non-
critical sets one has ∇α 6= 0, which implies that v 6= 0. Thus the α -level sets are
two-dimensional orientable surfaces which admit a non-vanishing tangent vector
field. Thus these surfaces must be tori, since their Euler characteristic is 0. For M
with boundary, the α -level sets could intersect boundary, in which case they are
annuli, see Fig.2.

(a) (b)

Fig. 2 The flow lines of steady flows typically lie on tori or annuli: see the cases of M without
boundary (a) and with boundary (b).

Remark 1. [2, 3] i) Analyticity assumptions on M and v imply that there is a finite

number of cells between the critical levels of α , which are foliated by tori or annuli.

ii) The R2 -action on tori is given by two commuting vector fields v and curl v. In
particular, locally around a non-critical level of α there are coordinates {φ 1 , φ2 , z}
such that the α -levels are given by {z = const} and

v = v1 (z)∂φ1 + v2 (z)∂φ2 ,
Groups and topology in the Euler hydrodynamics and KdV 5

curl v = w1 (z)∂φ1 + w2 (z)∂φ2 .

This way a steady 3D flow looks like a completely integrable Hamiltonian system
with two degrees of freedom.
iii) It could happen that ∇α = 0 everywhere, i.e. α = const. Then

v × curl v = 0,

and hence v is collinear with curl v at every point. Such fields are called force-free.
If v 6= 0 everywhere, we can express curl v as curl v = κ (x)v for a smooth function
κ (x) on M. Then κ is a first integral of our dynamical system given by the field v.
Indeed,
0 = div(curl v) = div(κ v) = ∇κ · v .
Again, the vector field v is tangent to the level sets of κ . On these sets there is only
the R-action.
iv) Another interesting case is when κ (x) = const. Then

curl v = λ v,

i.e. v is an eigen field for the curl operator: curl ξ = λ ξ . Such fields are called
Beltrami fields (or flows). One famious example is given by the so called ABC
flows on a 3D-torus, which exhibit chaotic behavior and draw special attention in
fast dynamo constructions:

vx = A sin z +C cosy

vy = B sin x + A cosz
vz = C sin y + B cosx
v) There is a two-dimensional version of the above Arnold’s theorem. Any area-
preserving field in dimension 2 is a Hamiltonian field (with possibly a multival-
ued Hamiltonian function): v = sgrad ψ . This Hamiltonian function ψ is called the
stream function for the field v. The condition that v is a steady flow, i.e. that it com-
mutes with its vorticity curl v, amounts in 2D to the fact that the stream function ψ
and its Laplacian ∆ ψ have the same level curves. In other words, locally there is a
function F : R → R such that ∆ ψ = F(ψ ).

2.2 Variational principles for steady flows.

The stationary solutions of the Euler equation come by as extremals from two dif-
ferent variational principles [3, 8].
i) The magneto-hydrodynamic (“MHD”) variational principle: consider the en-
ergy functional
6 Boris Khesin
Z
E(v) = |v|2 d 3 x
M
on divergence-free vector fields v on a 3D manifold M. Then extremals of the energy
functional among the fields diffeomorphic to a given one are singled out by the same
condition as the steady Euler flows: such fields must commute with their vorticities.
(This problem on conditional extremum corresponds to the restriction of the energy
E to the adjoint orbits of the diffeomorphism group.)
ii) The ideal hydrodynamic (“IHD”) principle: steady fields are extremal fields
for the energy functional among the fields with diffeomorphic vorticities, i.e. among
isovorticed fields. (The latter corresponds to the energy restriction to the coadjoint
orbits of the same group.) In this sense these principles are dual to each other, but
give the same sets of extremal fields.

3 Euler equations and integrable systems.

3.1 Hamiltonian reformulation of the Euler equations.

The differential-geometric description of the Euler equation as a geodesic flow on a

Lie group has a Hamiltonian reformulation. Fix the notation E(v) = 12 hv, Avi for the
energy quadratic form on g which we used to define the Riemannian metric. Identify
the Lie algebra and its dual with the help of this quadratic form. This identification
A : g → g∗ (called the inertia operator) allows one to rewrite the Euler equation on
the dual space g∗ .
It turns out that the Euler equation on g∗ is Hamiltonian with respect to the natural
Lie–Poisson structure on the dual space [1]. Moreover, the corresponding Hamilto-
nian function is minus the energy quadratic form lifted from the Lie algebra to its
dual space by the same identification: −E(m) = − 12 hA−1 m, mi, where m = Av. Here
we are going to take it as the definition of the Euler equation on the dual space g ∗ .
(The minus is related to the consideration of a right-invariant metric on the group.
It changes to plus for left-invariant metrics.)

Definition 1. (see, e.g., [4]) The Euler equation on g∗ , corresponding to the right-
invariant metric E(m) = 21 hAv, vi on the group, is given by the following explicit
formula:
dm
= −ad∗A−1 m m, (2)
dt
as an evolution of a point m ∈ g∗ . Here ad∗ is the coadjoint operator, dual to the
operator defining the structure of the Lie algebra g.

Below we explain the meaning of this operator in the case of the Virasoro algebra,
“responsible” for several equations of mathematical physics.
Groups and topology in the Euler hydrodynamics and KdV 7

3.2 The Virasoro algebra and the KdV equation.

Definition 2. The Virasoro algebra vir = Vect(S1 ) ⊕ R is the vector space of pairs
which consist of a smooth vector field on the circle and a number. This space is
equipped with the following commutation operation:

∂ ∂ ∂
 Z 
[( f (x) , a), (g(x) , b)] = ( f 0 (x)g(x) − f (x)g0 (x)) , f 0 (x)g00 (x) dx ,
∂x ∂x ∂ x S1

for any two elements ( f (x)∂ /∂ x, a) and (g(x)∂ /∂ x, b)Rin vir.

The bilinear skew-symmetric expression c( f , g) := S1 f 0 (x)g00 (x)dx is called
the Gelfand–Fuchs 2-cocycle.

There exists a Virasoro group, an extension of the group of smooth diffeomor-

phisms of the circle, whose Lie algebra is the Virasoro algebra vir. Fix the L 2 -energy
quadratic form in the Virasoro Lie algebra:

∂ 1
Z 
E( f (x) , a) = f 2 (x) dx + a2 .
∂x 2 S1

Applying the construction of Section 1 to the Virasoro group, one can equip this
group with a (right-invariant) Riemannian metric and consider the corresponding
Euler equation, i.e., the equation of the geodesic flow generated by this metric on
the Virasoro group.

Theorem 3. [9] The Euler equation corresponding to the geodesic flow (for the
above right-invariant metric) on the Virasoro group is a one-parameter family of
the Korteweg–de Vries (KdV) equations:

∂t u + u∂xu + c∂x3u = 0; ∂t c = 0

on a time-dependent function u on S 1 . Here c is a (constant) parameter, the “depth”

of the fluid.

Proof. The space vir ∗ can be identified with the set of pairs

(u(x)(dx)2 , c)| u(x) is a smooth function on S1 , c ∈ R .



Indeed, it is natural to contract the quadratic differentials u(x)(dx) 2 with vector

fields on the circle, while the constants are to be paired between themselves:

∂
Z
h(v(x) , a), (u(x)(dx)2 , c)i = v(x) · u(x) dx + a · c.
∂x S1

The coadjoint action of a Lie algebra element ( f ∂ /∂ x, a) ∈ vir on an element

(u(x)(dx)2 , c) of the dual space vir ∗ is

ad∗( f ∂ /∂ x,a) (u(dx)2 , c) = (2(∂x f )u + f ∂x u + c∂x3 f , 0).

8 Boris Khesin

It is obtained from the identity

∂ ∂ ∂
h[( f , a), (g , b)], (u(dx)2 , c)i = h(g , b), ad∗( f ∂ , a) (u(dx)2 , c)i,
∂x ∂x ∂x ∂x

which holds for every pair (g ∂∂x , b) ∈ vir.

The quadratic energy functional E on the Virasoro algebra vir determines the
“tautological” inertia operator A : vir → vir ∗ , which sends a pair (u(x)∂ /∂ x, c) ∈ vir
to (u(x)(dx)2 , c) ∈ vir∗ .
In particular, it defines the quadratic Hamiltonian on the dual space vir ∗ ,
1
Z
E(u(dx)2 , c) = ( u2 dx + c2)
2
1 ∂ 1 ∂ ∂
= h(u , c), (u(dx)2 , c)i = h(u , c), A(u , c)i.
2 ∂x 2 ∂x ∂x
The corresponding Euler equation for the right-invariant metric defined by E on the
group (according to the general formula (2) above) is given by

∂
(u(dx)2 , c) = −ad∗A−1 (u(dx)2 ,c) (u(dx)2 , c).
∂t
Making use of the explicit formula for the Virasoro coadjoint action ad ∗ for

( f ∂ /∂ x, a) = A−1 (u(dx)2 , c) = (u ∂ /∂ x, c),

we obtain the required Euler equation:

∂t u = −2(∂x u)u − u∂xu − c∂x3u = −3u∂x u − c∂x3 u, ∂t c = 0.

The coefficient c is preserved in time, and the function u satisfies the KdV equation.
QED.

3.3 Equations–relatives and conservation laws.

For different metrics on the Virasoro group, other interesting equations can appear
from the same scheme. The Euler equation on the Virasoro group with respect to
the right-invariant H 1 -metric gives the Camassa-Holm equation:

∂t u − ∂xxt u = −3u∂x u + 2(∂x u)∂xx u + u∂xxx u + c∂x3u,

see [7]. Similarly, the homogeneous Ḣ 1 -metric gives the Hunter-Saxton equation
(an equation in the Dym hierarchy):

∂xxt u = −2(∂x u)∂xx u − u∂xxx u,

Groups and topology in the Euler hydrodynamics and KdV 9

see [6].

Remark 2. It turns out that all these three equations (KdV, CH, and HS) are bihamil-
tonian systems, and hence admit an infinite family of conservation laws. The cor-
responding Hamiltonian (or Poisson) structures are naturally related to the Virasoro
algebra.
For instance, for the KdV equation these conserved quantities can be expressed
in the following way. Consider the KdV equation on (u(x)(dx) 2 , c) as an evolution
d2
of Hill’s operator c dx 2 + u(x). The monodromy M(u) of this operator is a 2 × 2-
matrix with the unit determinant. Look at the following function of the monodromy
for a family of Hill’s operators:

hλ (u) := log(trace M(u − λ 2 )),

2
where M(u − λ 2) is the monodromy of the Hill operator dx
d 2
2 + u(x) − λ .
Now, the expansion of the function hλ in λ produces the first integrals of the
KdV equation:
∞
hλ (u) ≈ 2πλ − ∑ cn h2n−1(u)λ 1−2n ,
n=1

where
1
Z Z Z  
2
h1 = u(x) dx, h3 = u (x) dx, h5 = 3
u (x) − (ux (x))2 dx, . . .
S1 S1 S1 2

and c1 = 1/2, cn = (2n − 3)!!/(2nn!) for n > 1. One can recognize here the familiar
form of higher KdV integrals. Their appearance in this expansion is due to the fact
that the trace of the monodromy M(u) is a Casimir function for the Virasoro alge-
bra, while the coefficients in a Casimir expansion provide a hierarchy of conserved
charges for any bihamiltonian systems, see more details on this and other equations
in [6].

Acknowledgements This paper presents the lecture notes for author’s minicourse on Mathemati-
cal Fluid Dynamics, given at the Fields Institute in September 2003. I am obliged to Walter Craig
for the invitation to give this course, encouragement and careful notes. I am also grateful to the
University of Geneva and Anton Alekseev for hospitality during the preparation of this survey.
The present work was partially sponsored by an NSERC research grant.
10 Boris Khesin

References

1. Arnold, V.I. (1966) Sur la géométrie différentielle des groupes de Lie de dimension infinie et
ses applications à l’hydrodynamique des fluides parfaits. Ann. Inst. Fourier 16, 316–361.
2. Arnold, V.I. (1973) The asymptotic Hopf invariant and its applications. Proc. Summer School
in Diff. Equations at Dilizhan, Erevan (in Russian); English transl.: Sel. Math. Sov. 5 (1986),
327–345.
3. Arnold, V.I. (1974) Mathematical methods in classical mechanics. English transl.: Graduate
Texts in Mathematics, v.60. Springer-Verlag, New York, 1989, 516pp.
4. Arnold, V.I. & Khesin, B.A. (1998) Topological methods in hydrodynamics. Applied Mathe-
matical Sciences, vol. 125, Springer-Verlag, New York, pp. xv+374.
5. Ebin, D. & Marsden, J. (1970) Groups of diffeomorphisms and the notion of an incompressible
fluid. Ann. of Math. (2) 92, 102–163.
6. Khesin, B. & Misiołek, G. (2003) Euler equations on homogeneous spaces and Virasoro orbits.
Advances in Math. 176, 116–144.
7. Misiołek, G. (1998) A shallow water equation as a geodesic flow on the Bott-Virasoro group.
J. Geom. Phys. 24:3, 203–208; Classical solutions of the periodic Camassa-Holm equation.
Geom. Funct. Anal. 12:5 (2002), 1080–1104.
8. Moffatt, H.K. & Tsinober, A. (1992) Helicity in laminar and turbulent flow. Annual Review of
Fluid Mechanics 24, 281–312.
9. Ovsienko, V.Yu. & Khesin, B.A. (1987) Korteweg-de Vries super-equation as an Euler equa-
tion. Funct. Anal. Appl. 21:4, 329–331.