Advanced Calculus of Several Variables (1973)
Part V. Line and Surface Integrals; Differential Forms and Stokes' Theorem
Chapter 8. CLOSED AND EXACT FORMS
Let ω be a differential k-form defined on the open set U in n. Then ω is called closed (on U) if dω = 0, and exact (on U) if there exists a (k − 1)-form α on U such that dα = ω.
Since d(dω) = 0 by Proposition 5.1 , we see immediately that every exact form is closed. Our object in this section is to discuss the extent to which the converse is true.
According to Theorem 2.5 , every closed 1-form, which is defined on all of 2, is exact. However, we have seen in Section 2 that the 1-form
is closed on 2 − 0, but is not exact; that is, there is no function f : 2 − 0 → such that df = ω. These facts suggest that the question as to whether the closed 1-form ω on U is exact, depends upon the geometry of the set U.
The “Poincar lemma” (Theorem 8.1 below) asserts that every closed form defined on a star-shaped region is exact. The open set is called star-shaped with respect to the point a provided that, given , the line segment from a to x is contained in U. The open set U is star-shaped if there exists such that U is star-shaped with respect to a. For example, n itself is star-shaped (with respect to any point), as is every open ball or open n-dimensional interval in n.
Theorem 8.1 Every closed differential k-form defined on a star-shaped open subset U of n is exact.
PROOF We may assume (after a translation if necessary) that U is star-shaped with respect to the origin 0. We want to define, for each positive integer k, a certain function I, from k-forms on U to (k − 1)-forms, such that I(0) = 0.
First we need the following notation. Given a k-tuple i = (i1, i2, . . . , ik), write
Note that
and, if ji = (j, i1, . . . , ik), then
(Exercise 8.4). Now, given a k-form
on U, we define the (k − 1)-form Iω on U by
Note that this makes sense because U is star-shaped with respect to 0, so if and . Clearly I(0) = 0, and
If ω is closed on U, then
Upon addition of (Eqs. (4) and (5), we obtain
Thus we have found a (k − 1)-form whose differential is ω, as desired.
Example 1 Consider the closed 2-form
on 2. In the notation of the proof of Theorem 8.1,
and
Equation (3) therefore gives
a 1-form whose differential is ω.
Theorem 8.1 has as special cases two important facts of vector analysis. Recall that, if F = (P, Q, R) is a vector field on an open set U in 3, then its divergence and curl are defined by
It follows easily that div F = 0 if F = curl G for some vector field G on U, while curl F = 0 if F = grad f = f for some scalar function f. The following theorem asserts that the converses are true if U is star-shaped.
Theorem 8.2 Let F be a vector field on the star-shaped open set . Then
(a)curl F = 0 if and only if there exists f : U → such that F = grad f,
(b)div F = 0 if and only if there exists G : U → 3 such that F = curl G.
PROOF Given a vector field F = (P, Q, R), we define a 1-form αF, a 2-form βF, and a 3-form γF by
Then routine computations yield the formulas
To prove that curl(grad f) = 0, note that
by (6), (7), and Proposition 5.1 . To prove that div(curl G) = 0, note that
by (7), (8), and Proposition 5.1 .
If curl F = 0 on U, then dαF = βcurl F = 0 by (7), so Theorem 8.1 gives a function f : U → such that df = αF, that is
so f = F. This proves (a).
If div F = 0 on U, then dβF = γdiv F = 0 by (8), so Theorem 8.1 gives a 1-form
such that dω = βF. But if G = (G1, G2, G3), then
by (7), so it follows that curl G = F. This proves (b).
Example 2 As a typical physical application of the Poincar lemma, we describe the reduction of Maxwell‘s electromagnetic field equations to the inhomogeneous wave equation. Maxwell’s equations relate the electric field vector E(x, y, z, t), the magnetic field vector H(x, y, z, t), the charge density ρ(x, y, z), and the current density J(x, y, z); we assume these are all defined on 3 (for each t). With the standard notation div F = · F and curl F = × F, these equations are
Equation (10) asserts that there are no magnetic sources, while (9), (11), and (12) are, respectively, Gauss’ law, Faraday’s law, and Ampere's law.
Let us introduce the following differential forms on 4:
where E = (E1, E2, E3), H = (H1, H2, H3), and J = (J1, J2, J3). Then Eqs. (10) and (11) are equivalent to the equation
while Eqs. (9) and (12) are equivalent to
Thus (13) and (14) are Maxwell's equations in the notation of differential forms.
Because of (13), the Poincar lemma implies the existence of a 1-form
such that
Of course the 1-form α which satisfies (15) is not unique; so does α + df for any differentiable function f : 4 → . In particular, if f is a solution of the in-homogeneous wave equation
then the new 1-form
satisfies both the equation
and the condition
Computing dβ, we find that (17) implies that
Substituting these expressions for the components of E and H, and making use of (18), a straightforward computation gives
Comparing this result with Eq. (14), we conclude that the vector field G = (G1, G2, G3) on 3 and the function g satisfy the inhomogeneous wave equations
Thus the solution of Maxwell's equations reduces to the problem of solving the inhomogeneous wave equation. In particular, if G and g satisfy Eq. (18) and 20), then the vector fields
defined by Eqs. (19), satisfy Maxwell's equations (9–(12).
Exercises
8.1For each of the following forms ω, find α such that dα = ω.
(a)ω = (3x2y2 + 8xy3) dx + (2x3y + 12x2y2 + 4y) dy on 2.
(b)ω = (y2 + 2xz2) dx + (2xy + 3y2z3) dy + (2x2z + 3y3z2) dz on 3.
(c)ω = (2y − 4) dy dz + (y2 − 2x) dz dx + (3 − x − 2yz) dx dy.
(d)ω = (xy2 + yz2 + zx2) dx dy dz.
8.2Let ω be a closed 1-form on 2 − 0 such that . Prove that ω is exact on 2 − 0. Hint: Given , define f(x, y) to be the integral of ω along the path γ(x,y) which follows an arc of the unit circle from (1, 0) to (x/ x, yy), and then follows the radial straight line segment to (x, y). Then show that df = ω.
8.3If ω is closed 1-form on 3 − 0, prove that ω is exact on 3 − 0. Hint: Given , define f(p) to be the integral of ω along the path γp which follows a great-circle arc on S2 from (1, 0, 0) to , and then follows the radial straight line segment to p. Apply Stokes' theorem to show that f(p) is independent of the chosen great circle. Then show that df = ω.
8.4Verify formulas (1) and (2).
8.5Verify formulas (6), (7), and (8).
8.6Let ω = dx + z dy on 3. Given f(x, y, z), compute d(fω). Conclude that ω does not have an integrating factor (see Exercise 5.4).