Week 8
Week 8
Week 8 Notes
May 17 and 19, 2016
Green’s Theorem
Let’s quickly recall how we defined the line integral of a vector field F along a curve C. We
imagined a thin wire submersed in some fluid, as in the following figure:
To perform a naı̈ve analysis, we’ll restrict our attention even further. Let’s consider a
rectangle
D = {(x, y) : a ≤ x ≤ b, c ≤ y ≤ d}
which lies in the domain of a vector field F on R2 , and we take our curve C to be the
boundary of this rectangle, written C = ∂D. A picture of our setup can be seen in Figure
1 As before, we would like to compute the work done by F along the curve ∂D, oriented
counterclockwise, but this time we’ll take a different approach. Consider the work done along
the bottom line of our rectangle — the work done from (a, c) to (b, c). At each point along
this line, our positively-oriented unit tangent vector T is the vector h1, 0i, so the quantity
F · T is given by
F · T = hF1 , F2 i · h1, 0i = F1 .
1
Figure 1: The curve C = ∂D.
We find the work done along the bottom line by integrating with respect to x (this is the
variable that changes as we move along this line):
Z b
work done along bottom line = F1 (x, c)dx.
a
Next, consider the work done along the top line of the rectangle — the line from (a, d) to
(b, d). This time our positively-oriented tangent vectors are pointing to the left, so we have
T = h−1, 0i, meaning that F · T = −F1 . So
Z b
work done along top line = −F1 (x, d)dx.
a
This means that the total work done along the top and bottom lines is given by
Z b
work done along top and bottom lines = (F1 (x, c) − F1 (x, d))dx.
a
Now because we’re looking at a rectangle which lies in the domain of F, we can use the
Fundamental Theorem of Calculus to rewrite the quantity inside of this integral, for each
a ≤ x ≤ b: Z c Z d
∂F1 ∂F1
F1 (x, c) − F1 (x, d) = (x, y)dy = − (x, y)dy.
d ∂y c ∂y
Substituting this in, we see that
Z bZ d
∂F1
work done along top and bottom lines = − (x, y)dydx. (1)
a c ∂y
2
A similar analysis works for the left and right lines. On the right, T = h0, 1i, so F · T = F2 ,
while on the left T = h0, −1i and F · T = −F2 . Then
Z d Z d Z b
∂F2
work along left and right lines = (F2 (b, y) − F2 (a, y))dy = (x, y)dxdy. (2)
c c a ∂x
Of course we find the total work done along ∂D by summing the values in (1) and (2). Using
Fubini’s theorem to interchange the order of integration, we see that the total work is given
by
Z bZ d Z dZ b Z bZ d
∂F1 ∂F2 ∂F2 ∂F1
− (x, y)dydx + (x, y)dxdy = − dydx.
a c ∂y c a ∂x a c ∂x ∂y
That is,
I ZZ
∂F2 ∂F1
F · dr = − dA. (3)
∂D D ∂x ∂y
We call the result in (3) Green’s Theorem, and it can be used to simplify a lot of cal-
culations. Notice that the integrand here is precisely the z-coordinate of curl(F), where we
consider F to be a vector field in three space. So in words Green’s theorem is saying that
the work done along ∂D is given by integrating the curl of F over D. We have a
helpful reality check here: if F is a conservative vector field, then the line integral of F about
the closed loop ∂D must be zero. But we also know that conservative vector fields have zero
curl, so Green’s theorem gives zero, as desired.
Note. It’s important here that D lies in the domain of F — otherwise, we might not be able
to use the FTC. This subtlety is related to the requirement that a vector field’s domain be
simply connected in order for the curl test for conservative vector fields to work. If D were
not simply connected, we could not always apply the FTC, and the situation would become
much hairier.
Note. Not all hope is lost, though. Despite the fact that we’ve only given an explanation
for Green’s theorem in the case that D is a rectangle, the equation continues to hold as long
as D is a region in the domain of F whose boundary consists of a finite number of simple,
closed curves, and we orient these curves so that D is always to the left.
H
Example. (Section 18.1, Exercise 8) Compute C (ln x+y)dx−x2 dy, where C is the rectangle
with vertices (1, 1), (3, 1), (1, 4), and (3, 4), oriented clockwise.
H
(Solution) In our symbolic notation, we’re being asked to compute C F · dr, where F =
hln x + y, −x2 i. Since we don’t like integrating terms such as ln x, this is a very difficult line
integral to compute a priori. Green’s theorem simplifies it quite a bit though, since
∂F2 ∂F1
= −2x and = 1.
∂y ∂y
3
Since C is the boundary of the rectangle D = {(x, y) : 1 ≤ x ≤ 3, 1 ≤ y ≤ 4}, we have
I ZZ Z 3Z 4
2
(ln x + y)dx − x dy = (−2x − 1)dA = (−2x − 1)dydx
C D 1 1
Z 3
3
(2x + 1)dx = −3 x2 + x 1
= −3
1
= −3(9 + 3 − 1 − 1) = −30.
So we see that Green’s theorem is a nice tool for turning difficult line integrals into double
integrals (or vice versa). In particular, it’s occasionally helpful
RR to use Green’s theorem to
compute the area of a region by rewriting the double integral D 1dA as a line integral along
∂D. The trick is to find a vector field F so that ∂F∂x
2
− ∂F
∂y
1
= 1. Three such vector fields are
given by
1 1
F(x, y) = h0, xi, F(x, y) = h−y, 0i, F(x, y) = h− y, xi,
2 2
meaning that area can be written as
ZZ I I I
1
Area = 1dA = xdy = −ydx = xdy − ydx.
D ∂D ∂D 2 ∂D
2 2
Example. Use one of these vector fields to compute the area of the ellipse xa2 + yb2 = 1,
where a, b > 0.
H
(Solution) We’ll use the expression ∂D xdy to compute the area, but first we’ll need a
parametrization of ∂D. The standard parametrization of the boundary of an ellipse is given
by
r(t) = (a cos t, b sin t),
where t varies from 0 to 2π. Then the area is
Z Z 2π
A= xdy = h0, a cos ti · h−a sin t, b cos tidt
∂D 0
Z 2π
ab 2π
Z
2
= ab cos tdt = (1 + cos 2t)dt
0 2 0
2π
ab 1 ab
= t + sin t = 2π = πab.
2 2 0 2
The first non-trivial step of the above computation can be thought of in this way: the
expression xdy in a line integral means that we have F(x, y) = h0, xi, so F(r(t)) = h0, a cos ti.
Notice that if we tried to compute the area as a double integral we would have something
like
Z Z √ 2 2
a b 1−x /a
A= √ 1dydx
−a −b 1−x2 /a2
in rectangular coordinates, and the polar coordinates setup isn’t nice either. In either case,
our line integral is much simpler.
4
Flux
Recall that the divergence of a vector field is a measure of the extent to which the vector
field is expanding or contracting at a given point. In week 4 we derived a formula for the
divergence of a vector field F = hF1 , F2 i by drawing a small rectangle around a given point
and measuring the degree to which matter was leaving or entering this rectangle. This sounds
eerily similar to a flux computation, and we can play a game similar to what we did above
for work to explicitly characterize this relationship. Consider the rectangle D is Figure 1,
with all the same assumptions as before. If we’d like to compute the flux through ∂D, we can
compute this line-by-line, as we did with work. Along the bottom line our outward-pointing
unit normal vector is given by h0, −1i, so
Z b
flux through bottom line = −F2 (x, c)dx
a
Just as with Green’s theorem, this equation also a nice interpretation: the total flux out
of D is given by integrating the divergence of F across D. This is an unsurprising
relationship, especially if we go back and look at some of the figures from week 4.
5
Example. (Section 18.1, Exercise 38) Compute the flux of a fluid flow with velocity vector
field F(x, y) = hcos y, sin yi across the boundary of the rectangle 0 ≤ x ≤ 2, 0 ≤ y ≤ π2 .
(Solution) Since
∂ ∂
div(F) = (cos y) + (sin y) = cos y,
∂x ∂y
the flux is given by
I Z π/2 Z 2
F · n ds = cos ydxdy = 2 [sin y]π/2
0 = 2.
∂D 0 0
Note that computing the flux as a line integral would require us to parametrize the rectangle
0 ≤ x ≤ 2, 0 ≤ y ≤ π2 . This parametrization would almost certainly be defined in four
pieces, so we’d have to compute four different normal vectors and then evaluate four different
integrals. This approach is much cleaner.