LECTURE 1: DIFFERENTIAL FORMS

1. 1-forms on Rn In calculus, you may have seen the dierential or exterior derivative df of a function f (x, y, z ) dened to be f f f df = dx + dy + dz. x y z The expression df is called a 1-form. But what does this really mean? Denition: A smooth 1-form on Rn is a real-valued function on the set of all tangent vectors to Rn , i.e., : T Rn R with the properties that 1. is linear on the tangent space Tx Rn for each x Rn . 2. For any smooth vector eld v = v (x), the function (v ) : Rn R is smooth. Given a 1-form , for each x Rn the map x : Tx Rn R is an element of the dual space (Tx Rn ) . When we extend this notion to all of Rn , we see that the space of 1-forms on Rn is dual to the space of vector elds on Rn . In particular, the 1-forms dx1 , . . . , dxn are dened by the property that for any vector v = (v 1 , . . . , v n ) Tx Rn , dxi (v ) = v i . The dxi s form a basis for the 1-forms on Rn , so any other 1-form may be expressed in the form
n

=
i=1

fi (x) dxi .

If a vector eld v on Rn has the form v (x) = (v 1 (x), . . . , v n (x)),

JEANNE NIELSEN CLELLAND

then at any point x Rn ,

n

x (v ) =
i=1

fi (x) v i (x).

2. p-forms on Rn The 1-forms on Rn are part of an algebra, called the algebra of dierential forms on Rn . The multiplication in this algebra is called wedge product, and it is skew-symmetric: dxi dxj = dxj dxi . One consequence of this is that dxi dxi = 0. If each summand of a dierential form contains p dxi s, the form is called a p-form. Functions are considered to be 0-forms, and any form on Rn of degree p > n must be zero due to the skew-symmetry. A basis for the p-forms on Rn is given by the set {dxi1 dxip : 1 i1 < i2 < < ip n}. Any p-form may be expressed in the form =
|I |= p

fI dxi1 dxip

where I ranges over all multi-indices I = (i1 , . . . , ip ) of length p. Just as 1-forms act on vector elds to give real-valued functions, so p-forms act on p-tuples of vector elds to give real-valued functions. For instance, if , are 1-forms and v, w are vector elds, then ( )(v, w) = (v ) (w) (w) (v ). In general, if 1 , . . . , p are 1-forms and v1 . . . , vp are vector elds, then (1 p )(v1 , . . . , vp ) =
Sp

sgn( ) 1 (v(1) ) 2 (v(2) ) n (v(n) ).

3. The exterior derivative The exterior derivative is an operation that takes p-forms to (p + 1)-forms. We will rst dene it for functions and then extend this denition to higher degree forms. Denition: If f : Rn R is dierentiable, then the exterior derivative of f is the 1-form df with the property that for any x Rn , v Tx Rn , dfx (v ) = v (f ), i.e., dfx (v ) is the directional derivative of f at x in the direction of v .

LIE GROUPS AND THE METHOD OF THE MOVING FRAME

It is not dicult to show that

n

df =
i=1

f i dx . xi

The exterior derivative also obeys the Leibniz rule d(f g ) = g df + f dg and the chain rule d(h(f )) = h (f ) df. We extend this denition to p-forms as follows: Denition: Given a p-form = ative d is the (p + 1)-form d =
|I |= p

fI dxi1 dxip , the exterior deriv|I |= p

dfI dxi1 dxip .

If is a p-form and is a q -form, then the Leibniz rule takes the form d( ) = d + (1)p d. Very Important Theorem: d2 = 0. i.e., for any dierential form , d(d) = 0. Proof: First suppose that f is a function, i.e., a 0-form. Then
n

d(df ) = d(
i=1

f i dx ) xi

=
i,j

2f dxj dxi xi xj ( 2f 2f )dxi dxj xj xi xi xj

=
i<j

=0 because mixed partials commute. Next, note that dxi really does mean d(xi ), where xi is the ith coordinate function. So by the argument above, d(dxi ) = 0. Now suppose that =
|I |= p

fI dxi1 dxip .

JEANNE NIELSEN CLELLAND

Then by the Leibniz rule, d(d) = d(

|I |= p

dfI dxi1 dxip ) [d(dfI ) dxi1 dxip dfI d(dxi1 ) dxip + . . . ]

= = 0.

|I |= p

Denition: A p-form is closed if d = 0. is exact if there exists a (p 1)-form such that = d . By the Very Important Theorem, every exact form is closed. The converse is only partially true: every closed form is locally exact. This means that given a closed p-form on an open set U Rn , any point x U has a neighborhood on which there exists a (p 1)-form with d = . 4. Differential forms on manifolds Given a smooth manifold M , a smooth 1-form on M is a real-valued function on the set of all tangent vectors to M such that 1. is linear on the tangent space Tx M for each x M . 2. For any smooth vector eld v on M , the function (v ) : M R is smooth. So for each x M , the map x : Tx M R is an element of the dual space (Tx M ) . Wedge products and exterior derivatives are dened similarly as for Rn . If f : M R is a dierentiable function, then we dene the exterior derivative of f to be the 1-form df with the property that for any x M, v Tx M , dfx (v ) = v (f ). A local basis for the space of 1-forms on M can be described as before in terms of any local coordinate chart (x1 , . . . , xn ) on M , and it is possible to show that the coordinate-based notions of wedge product and exterior derivative are in fact independent of the choice of local coordinates and so are well-dened. More generally, suppose that M1 , M2 are smooth manifolds and that F : M1 M2 is a dierentiable map. For any x M1 , the dierential dF (also denoted F ) : Tx M1 TF (x) M2 may be thought of as a vector-valued 1-form, because it is a linear map from Tx M1 to the vector space TF (x) M2 . There is an analogous map in the opposite direction for dierential forms, called the pullback and denoted F . It is dened as follows.

LIE GROUPS AND THE METHOD OF THE MOVING FRAME

Denition: If F : M1 M2 is a dierentiable map, then 1. If f : M2 R is a dierentiable function, then F f : M1 R is the function (F f )(x) = (f F )(x). 2. If is a p-form on M2 , then F is the p-form on M1 dened as follows: if v1 , . . . , vp Tx M1 , then (F )(v1 , . . . , vp ) = (F (v1 ), . . . , F (vp )). In terms of local coordinates (x1 , . . . , xn ) on M1 and (y 1 , . . . , y m ) on M2 , suppose that the map F is described by y i = y i (x1 , . . . , xn ), 1 i m. Then the dierential dF at each point x M1 may be represented in this coordinate system by the matrix y i . xj The dxj s are forms on M1 , the dy i s are forms on M2 , and the pullback map F acts on the dy i s by
n

F (dy i ) =
j =1

y i j dx . xj

The pullback map behaves as nicely as one could hope with respect to the various operations on dierential forms, as described in the following theorem. Theorem: Let F : M1 M2 be a dierentiable map, and let , be dierential forms on M2 . Then 1. F ( + ) = F + F . 2. F ( ) = F F . 3. F (d) = d(F ). 5. The Lie derivative The nal operation that we will dene on dierential forms is the Lie derivative. This is a generalization of the notion of directional derivative of a function.

JEANNE NIELSEN CLELLAND

Suppose that v (x) is a vector eld on a manifold M , and let : M (, ) M be the ow of v . This is the unique map that satises the conditions (x, t) = v ((x, t)) t (x, 0) = x. In other words, t (x) = (x, t) is the point reached at time t by owing along the vector eld v (x) starting from the point x at time 0. Recall that if f : M R is a smooth function, then the directional derivative of f at x in the direction of v is f (t (x)) f (x) v (f ) = lim t0 t ( ( f ) f )(x) = lim t . t0 t Similarly, given a dierential form we dene the Lie derivative of along the vector eld v (x) to be Lv = lim t . t0 t Fortunately there is a practical way to compute the Lie derivative. First we need the notion of the left-hook of a dierential form with a vector eld. Given a p-form and a vector eld v , the left-hook v of with v (also called the interior product of with v ) is the (p 1)-form dened by the property that for any w1 , . . . , wp1 Tx Rn , (v )(w1 , . . . , wp1 ) = (v, w1 , . . . , wp1 ). For instance, (dx dy + dz dx) = dy dz. x Now according to Cartans formula, the Lie derivative of along the vector eld v is Lv = v d + d(v ).

Exercises
1. Classical vector analysis avoids the use of dierential forms on R3 by converting 1-forms and 2-forms into vector elds by means of the following one-to-one correspondences. (1 , 2 , 3 will denote the standard basis 1 = [1, 0, 0], 2 = [0, 1, 0], 3 = [0, 0, 1].)

LIE GROUPS AND THE METHOD OF THE MOVING FRAME

f1 dx1 + f2 dx2 + f3 dx3 f1 1 + f2 2 + f3 3 f1 dx2 dx3 + f2 dx3 dx1 + f3 dx1 dx2 f1 1 + f2 2 + f3 3 Vector analysis uses three basic operations based on partial dierentiation: 1. Gradient of a function f :
3

grad(f ) =
i=1 3

f i xi

2. Curl of a vector eld v =

i=1

v i (x) i : v 1 v 3 x3 x1
3

curl(v ) =

v 3 x2

v 2 x3

1 +

2 +

v 2 v 1 x1 x2

3. Divergence of a vector eld v =

i=1 3

v i (x) i : v i xi

div(v ) =
i=1

Prove that all three operations may be expressed in terms of exterior derivatives as follows: 1. df grad(f ) 2. If is a 1-form and v , then d curl(v ). 3. If is a 2-form and v , then d div(v ) dx1 dx2 dx3 . Show that the identities curl(grad(f )) = 0 div(curl(v )) = 0 follow from the fact that d2 = 0. 2. Let f and g be real-valued functions on R2 . Prove that df dg =
f x g x f y g y

dx dy.

(You may recognize this from the change-of-variables formula for double integrals.) 3. Suppose that , are 1-forms on Rn . Prove the Leibniz rule d( ) = d d.

JEANNE NIELSEN CLELLAND

4. Prove the statement above that if F : M1 M2 is described in terms of local coordinates by y i = y i (x1 , . . . , xn ), 1 i m then
n

F (dy i ) =
j =1

y i j dx . xj

5. Let (r, ) be coordinates on R2 and (x, y, z ) coordinates on R3 . Let F : R2 R3 be dened by F (r, ) = (cos , sin , r). Describe the dierential dF in terms of these coordinates and compute the pullbacks F (dx), F (dy ), F (dz ).