Distributions, Frobenious Theorem & Transformations

Harry G. Kwatny
Department of Mechanical Engineering & Mechanics Drexel University

Outline

Distributions & integral surfaces Involutivity Frobenious Theorem Codistributions & computing Invariant distributions Transformation of vector fields Involutive closure

Distributions
p = span {v( p)1 , , vr ( p)} is a subspace of TM p
Definition: A smooth distribution on M is a map which assigns to each point pM, a subspace of the tangent space to M at p, pTMp such that p is the span of a set of smooth vector fields v1,..,vr evaluated at p. We write =span{v1,..,vr}. Definition: An integral submanifold of a set of vector fields v1,..,vr is a submanifold NM whose tangent space TNp is spanned by {v1(p),..,vr(p)} for each pN. The set of vector fields is (completely) integrable if through every point pM there passes an integral submanifold.

v1 , , vr is a set of vector fields on M

Involutive Distributions
Definition: A system of smooth vector fields {v1,..,vr} on M is in involution if there exist smooth real valued functions ckij(p), pM and i,j,k = 1,..,r such that for each i,j
ij = v , v c i j k vk k =1 r

Proposition: (Froebenius) Let {v1,..,vr} be an involutive system of vector fields with dim [span{v1,..,vr}]=k on M. Then the system is integrable with all integral manifolds of dimension k. Proposition: (Hermann) Let {v1,..,vr} be a system of smooth vector fields on M. Then the system is integrable if and only if it is in involution.

Example
M = R3

2 zx y 2 yz = span{v, w} v = x w = 2 2 2 z 1 x y + 0

[v , w] 0 so the distribution is completely integrable. The distribution is singular because dim = 2 everywhere except on the z-axis x = 0, y = 0 2 2 and on the circle x + y = 1, z = 0 where dim = 1 The z-axis and the
circle are one-dimensional integral manifolds. All others are the tori:

Tc = ( x, y , z ) R 3 ( x 2 + y 2 ) 1 / 2 ( x 2 + y 2 + z 2 + 1) = c > 2

Example

5 2.5 0 -2.5 -5 -10 0 10 -5 0 5

10

Maximal Integral Submanifolds

Definition: A maximal integral submanifold N of a distribution on a manifold M is a connected integral manifold of which is not a proper subset of any other connected integral manifold of . A distribution on M is said to have the maximal integral manifolds property if through every point pM passes a maximal integral submanifold of . Proposition: Let be a d-dimensional, involutive, smooth distribution on M. Let pM. Then through p there passes a unique maximal connected integral manifold of .

Covector Fields
A covector field w on M assigns to each point pM an element w(p)T*Mp. A codistribution on M is a mapping which assigns a subspace (p) of T*Mp to each point pM. As with distributions we write = span{w1,..,wr}. Distributions are sometimes associated with special codistributions. As an example, for each pM, the annihilator of the distribution (p) is the set of all covectors which annihilate vectors in (p) (p) := {wT*Mp : w(v)=0 for all v(p)} Conversely, given a codistribution , we define its annihilator, a distribution (p):= {vTMp : v(w)=0 for all w(p)}.

Representing Integral Surfaces

= span{v1 ,, vk } a k -dim, involutive distribution on m-dim M . Consider x = ( z, x0 ), where

( z1 ,, zk , x0 ) = 1z1 2z2
x3

t2 notation : 1t1 2 ( x ) = ( t1 , ( t2 , x ) )

z m ( x0 )
k

z2

x = ( z1 , z2 , x0 )

x2
z1 x1

integral surface passing through x0

Local Coordinate Transformation

= span{v1 ,, vk } a k -dim distribution on m-dim M define vk +1 ,, vm such that span{v1 ,, vm } = R m consider x = ( z, x0 ), where

( z1 ,, zm , x0 ) = 1z1 2z2
x3 z2
x = ( z1 , z2 , c, x0 )

z m ( x0 )
m

First k coordinates map onto integral surface.

x2

z3

z1 x1

c identifies the surface. Mapping is defined on a neighborhood of x0 .

Invariant Distributions
Definition: A distribution = {v1 ,, vr } on M is invariant with respect to a vector field f on M if the Lie bracket [f,vi], for each i = 1, , r is a vector field of .

Notation:

[ f , ] = span{[ f , vi ], i = 1, , r}

that is invariant with respect to f may be stated [f,] . In general +[f,] = +span{[f,vi], i=1,..,r } = span{v1,..,vr,[f,v1],..,[f,vr]}

Example: Invariant Linear Subspace

V = span{v1 ,, v r }
that is invariant under the linear mapping A, i.e., AVV. Define a distribution on Rn

vi R n

V ( x) = span{v1 , , v r }

and a vector field

f A ( x) = Ax

We prove that V is invariant under the vector field fA. We need to show that [ f A , vi ] V Compute

vi f A [ f A , vi ] = fA vi = Avi x x By assumption

Avi V

Transformation of Vector Fields: Lemma

Lemma: Let be a be an involutive distribution of constant dimension d on an open subset U of R n and suppose that is invariant under a vector field f. Then at each point x0 U there exists a neighborhood U 0 of x 0 in U and a coordinate transformation z = ( x), defined on it, in which the vector field f is of the form
f1 ( z1 , z d , z d +1 ,, z n ) f d ( z1 , z d , z d +1 ,, z n ) f ( z) = f ( z , , z ) d +1 d +1 n f n ( z d +1 ,, z n )

Transformation of Vector Fields sketch of proof ~ (a)

has dimension d integral surfaces through every x0 U 0
z = ( x), x U 0 with span {d d +1 , d n } =

1i i 0 k i i define the basis for TM z : ( z ) = , k = 1 k = i i n f i f f , i = z = z i

Transformation of Vector Fields sketch of proof ~ (b)

note : = span 1 , , d

v1 , v v = vd in z coords 0

invariant w.r.t. f f , f k = 0, zi d + 1 k n, 1 i d

Transformation of Vector Fields

z2

z = ( x)

x2

1
2
x0

z1
integral manifolds

x1

z1 = f1 ( z1 , z2 ) x = f ( x) z2 = f 2 ( z2 )

Slice
z1 = f1 ( z1 , z2 ) z2 = f 2 ( z2 )

Sef c = x U 0 z2 ( x) = z2 (e f c)

Sc = { x U 0 z2 ( x) = z2 (c)}

Example

1 0 1 0 = span {v1 , v2 } =span , 0 0 x 2 x1

is involutive and invariant w.r.t. f.

x2 x 3 f = x3 x4 x1 x2 x3 2 + + sin x x x x 3 2 1 3

1 0 0 0 0 0 0 0 0 1 0 0 v3 = v4 = span {v1 , v2 , v3 , v4 } = span , , , 1 0 0 0 1 0 0 1 0 0 0 1

( z ) = [ z1 z2 z3 z1 z2 + z4 ] , 1 ( x) = [ x1 x2 x3 x1 x2 + x4 ]

z2 z 3 f = z3 z 4 sin z3

Involutive Closure ~ 1

Problem 1: find the smallest distribution with the following properties

It is nonsingular It contains a given distribution It is involutive It is invariant w.r.t. a given set of vector fields, 1,,q

1 , q

Involutive Closure ~ 2a

Problem 2: find the largest distribution with the following properties

It is nonsingular It is contained in a given distribution It is involutive It is invariant w.r.t. a given set of vector fields, 1,,q

Involutive Closure ~ 2b

Problem 2 is equivalent to: find the smallest codistribution with the following properties

It is nonsingular It contains a given codistribution It is spanned by a set of exact covector fields It is invariant w.r.t. a given set of vector fields, 1,,q

1 , q

Algorithms
Algorithm for Problem 1: 0 =
k = k 1 + i =1 [ i , k 1 ]
q

stop when k = k 1

Algorithm for Problem 2:

0 = = k = k 1 + i =1 L i k 1
q

stop when k * = k * 1 k*