Line integrals

What to know:
1. Be able to parametrize curves
2. Be able to set up and compute line integrals with respect to arc length (ds) and of vector
fields (·d~r).
3. Be able to compute line integrals dx and dy (interpreted as special cases of integrals of vector
fields).
4. Know which line integrals depend on the parametrization of the curve and which ones don’t.
In this section, we’ll see how to integrate over curves. The curves we’ll talk about are called
piecewise smooth curves, which means that they are finite unions of smooth curves, parametrized
as c(t) = (x(t), y(t)), t ∈ [a, b], like in figures 1 and 21 .

Figure 2: Another piecewise

smooth curve

Figure 1: A piecewise
smooth curve

Line integrals of scalar functions

Suppose that f (x, y) is a continuous non-negative function. Our goal is to be able to evaluate the
area of a fence that lies above the curve c and under the graph of f .
To do this, we approximate the area in the following way: We first approximate the curve
c by line segments [aj , aj+1 ] and form rectangles (living in R3 ) with base [aj , aj+1 ] and height
f (a∗j ), where a∗j is the midpoint of each line segment. Then, we evaluate the area of each of
those rectangles and sum the areas. As we make the approximation of the curve more and more
accurate, this procedure gives rise to a special type of integral, called line integral with respect
to to arc length.
(you can find an animated graph of this procedure here).
1
The precise definition of a piecewise smooth curve is that you want the derivative ċ(t) = (ẋ(t), ẏ(t)) to exist
and be continuous for all except maybe finitely many t ∈ [a, b], and the side limits limt→t± ẋ(t) to exist for all
0
t0 ∈ [a, b]. This is to exclude some pathological but quite exciting cases, like the one in figures 3 and 4, or even
space filling curves!

1
 Figure 3: The graph of Figure 4: A curve con-
f (x) = x sin(1/x), not sisting of infinitely many
piecewise smooth line segments, not piecewise
smooth

Definition 1. The line integral with respect to arc length of a continuous function f (x, y) along
a piecewise smooth curve c(t) = (x(t), y(t)), a ≤ t ≤ b, is defined to be
s 
Z Z b 2  2
dx dy
f (x, y)ds = f (x(t), y(t)) + dt.
c a dt dt

Remarks:
• Note that if we replace f by the function 1 we obtain the length of c!

• Another physical interpretation of a line integral with respect to arc length is that when you
have the density function ρ of a wire lying on a plane curve c, its mass is given by
Z
m = ρ(x, y)ds.
c

R
Example 1. Find c
xds, where c is the right half of a circle with radius 2.
Solution. We first parametrize c: a way to do it is by choosing c(t) = (2 cos(t), 2 sin(t)), −π/2 ≤
t ≤ π/2. Then, compute x0 (t) = −2 sin(t), y 0 (t) = 2 cos(t) and write
Z Z π/2 q
xds = (2 cos(t)) 4 sin2 (t) + 4 cos2 (t)dt = 8
c −π/2

Wait! How do we know that this is the correct way to parametrize this curve? We know
that there are many different ways this can be done. Does it matter how fast we go or in which
direction?
Exercise 1. Now parametrize the curve of the last example using c(t) = (2 cos(−2t), 2 sin(−2t)),
for −π/4 ≤ t ≤ π/4 (this follows the same path, but twice as fast and in opposite direction). What
do you find?
In fact, line integrals with respect to arc length do not depend on the parametrization,
as long as the curve is only transversed once: (for example, a case where you would find different
answers is if you try to parametrize the unit circle by c1 (t) = (cos(t), sin(t)), 0 ≤ t ≤ 2π and

2
c2 (t) = (cos(t), sin(t)), 0 ≤ t ≤ 3π). This makes the question asked in Example 1 a valid one: no
matter which way we choose to parametrize the curve, we’ll find the same answer.
Now that you know that the parametrization doesn’t matter as long as you have one, you can
find the parametrizations for curves that appear frequently in a handout, here.
The definition of the line integral with respect to arc length for a function f (x, y, z) is com-
pletely analogous:
Definition 2. The line integral with respect to arc length of a continuous function f (x, y, z) along
a piecewise smooth curve c(t) = (x(t), y(t), z(t)), a ≤ t ≤ b, is defined to be
s 
Z Z b 2  2  2
dx dy dz
f (x, y, z)ds = f (x(t), y(t), z(t)) + + dt.
c a dt dt dt

Line Integrals of Vector Fields

So far, we’ve seen how we can integrate scalar functions. Now we want to make sense of integration
of vector fields along curves, and the motivation comes from physics: In physics, we are interested
in knowing the work produced by a force field when an object moves along a curve c given by the
vector function ~r(t), t ∈ [a, b] inside the force field F~ (x, y).
We know from physics that the work produced by a constant force G ~ while an object is displaced
by ~x is given by ~x · G. In our case, we have a variable force (a force field) F~ and a variable direction
~
of movement. So, when trying to make sense of the total work produced by F~ on an object moving
along c, the natural thing to do is to split the curve into very short curves, where the displacement
is happening in an almost constant direction (~r 0 (t)!) and the force is also almost constant and
equal to F~ (~r(t)). Then the infinitesimal work produced by F~ for time t would be

F~ (~r(t)) · ~r 0 (t).

We’d finally sum these infinitesimal quantities, that is, integrate!

We have:
Definition 3. Let F~ (x, y) be a continuous vector field and c(t) = ~r(t), t ∈ [a, b] be a curve in R2 .
We define the line integral of F~ along c to be
Z Z b
F~ · d~r = F~ (~r(t)) · ~r 0 (t)dt.
c a

Example 2. Compute c F~ · d~r, where F~ (x, y) = hx2 , −xyi and c is the line segment from (1,0)
R

to (0,1) parametrized by c(t) = ~r(t) = (1 − t, t), t ∈ [0, 1].

Solution. We have ~r 0 (t) = h−1, 1i, so
Z Z 1
~
F · d~r = −(1 − t)2 − t(1 − t)dt = −1/2.
c 0

Exercise 2. Now try to parametrize the segment in the above example differently: Use c(t) =
~r(t) = (2t, 1−2t), t ∈ [0, 1/2] (this travels along the segment in double speed and opposite direction).

3
 In the last exercise, you should find the answer to be 1/2. You may have noticed that the sign
is different from the one we found in the solution before. However, nothing else seems to have
changed. This is not an accident. The line integrals of vector fields do not depend on the speed of
the parametrization, but they do depend on the direction that the curve is transversed.
This means that if a problem asks to compute such an integral, it must provide a parametrization,
or at least direction in which the curve is transversed.
If c is a curve, let us denote by −c a curve that consists of the same set of points as c, but is
transversed in the opposite direction (regardless of the speed). Then
Z Z
F~ · d~r = − F~ · d~r. (1)
−c c

So, if we are trying to solve a problem that that asks to compute a line integral of a vector
field along a curve and only gives us the direction in which a curve is transversed but not an exact
parametrization, here’s what we can do: We find a parametrization and check if our parametriza-
tion agrees with the direction given in the problem: if it does, we’re fine; if not, we integrate with
the parametrization we found and then use (1) to say that the final answer has the opposite sign
of the one we found.
Now, by unwinding the definition, if c(t) = ~r(t) = (x(t), y(t)), t ∈ [a, b] and F~ (x, y) =
hP (x, y), Q(x, y)i,
Z Z b
~
F · d~r = P (x(t), y(t))x0 (t) + Q(x(t), y(t))y 0 (t)dt. (2)
c a

Notation: we write Z b Z
0
P (x(t), y(t))x (t)dt =: P (x, y)dx (3)
a c
and similarly Z b Z
0
Q(x(t), y(t))y (t)dt =: Q(x, y)dy. (4)
a c
So, (2) can be written as
Z Z Z
F~ · d~r = P (x, y)dx + Q(x, y)dy. (5)
c c c

You will often see this last notation abbreviated as

Z Z Z
P (x, y)dx + Q(x, y)dy =: P (x, y)dx + Q(x, y)dy.
c c c

Remark: Many textbooks (including Stewart’s) define the right hand sides of (3) and (4) as two
more types of line integrals of scalar functions, besides the line integral R with respect to arc length
and deduce (5) as a consequence. In our setting, we can interpret c f (x, y)dx as the line integral
of the vector field F~ (x, y) = hf (x, y), 0i along c, and similarly for c f (x, y)dy. To compute them,
R

(3) and (4) give

Z Z b
f (x, y)dx = f (x(t), y(t))x0 (t)dt
c a
R
and similarly for c f (x, y)dy.
The definitions and properties related to line integrals of vector fields carry over unchanged
when we have vector fields in R3 and curves in R3 : Just include a variable z! We have

4
Definition 4. Let F~ (x, y, z) be a continuous vector field and c(t) = ~r(t) = hx(t), y(t), z(t)i,
t ∈ [a, b] be a piecewise smooth curve in R3 . We define the line integral of F~ along c to be
Z Z b
F~ · d~r = F~ (~r(t)) · ~r 0 (t)dt.
c a