In Chapter 7, the magnitude ‖⃗𝑟′(𝑡)‖ measured the speed of a moving
point. We now use that same quantity to measure distance along the curve. Once
we know how to add up little pieces of length, we can also weight those pieces
by a density or some other scalar quantity.
8.1Arc Length
One way to estimate the length of a curve is to choose points along it, join
successive points with line segments, and add the lengths of those segments.
The estimate improves as we use more and shorter segments.
Figure 8.1 Join a handful of points on the curve and measure the polygon instead. With six
chords the answer is visibly short—the polygon cuts every corner and misses the
loop almost entirely; with fifty it is short by less than you can see. Each
side of that polygon is one Δ⃗𝑟, and shrinking it is what turns
‖Δ⃗𝑟‖ into
𝑑𝑠=‖⃗𝑟′(𝑡)‖𝑑𝑡. This cubic has no elementary formula for its
length, so the number the sum climbs toward is one you could not have written
down another way.
Definition 8.1 (Infinitesimal Arc Length). If ⃗𝑟(𝑡) is a parametric curve, its infinitesimal arc length is measured by
𝑑𝑠=‖⃗𝑟′(𝑡)‖𝑑𝑡
This makes sense: after all the derivative ⃗𝑟′(𝑡) is the velocity, ‖⃗𝑟′(𝑡)‖ is the speed, and 𝑑𝑡 is an infinitesimal length of time. Thus, the product ‖⃗𝑟′(𝑡)‖𝑑𝑡 is an infinitesimal bit of distance - a small length along the curve. To take this infinitesimal information and get something useful out - we need to integrate along the curve.
Definition 8.2 (Arc Length). If ⃗𝑟(𝑡) is a parametric curve, its length between 𝑡=𝑎 and 𝑡=𝑏 is given by
𝐿=∫𝑏𝑎𝑑𝑠=∫𝑏𝑎‖⃗𝑟′(𝑡)‖𝑑𝑡
Figure 8.2 The arrow is the velocity; the graph is its length. The two are drawn to the
same scale on purpose—stand the arrow upright and it is the height of the
graph, which is the whole content of the word "speed." As the point moves it
draws the graph out behind it, and the area it has swept out is the distance it
has travelled. Send it around twice and the picture on the right never changes
while the area on the left keeps climbing, which is why length belongs to the
traversal and not only to the image.
Example 8.3 (Arc Length of a Helix). Find the arc length of ⃗𝑟(𝑡)=(cos(𝑡),sin(𝑡),𝑡) from 𝑡=0 to 𝑡=2𝜋. First, we need to find the velocity ⃗𝑟′:
⃗𝑟′(𝑡)=⟨−sin(𝑡),cos(𝑡),1⟩
Next, we need to take this velocity and find the speed:
‖⃗𝑟′(𝑡)‖=√(−sin𝑡)2+(cos𝑡)2+1=√2
Finding arc length is just integrating this over the domain:
∫2𝜋0‖⃗𝑟′(𝑡)‖𝑑𝑡=∫2𝜋0√2𝑑𝑡=2𝜋√2
Arc-length integrals can be challenging because of the square root, but with some algebra and integration techniques, much can be learned.
From this idea, we can define the arc-length function, which measures the length of a curve ⃗𝑟(𝑡) from a starting point 𝑡=𝑎:
Definition 8.4 (The Arc-Length Function). If ⃗𝑟(𝑡) is a parametric curve, for any given starting point 𝑡=𝑎 we may define the arc-length function which measures the length of curve between 𝑎 and 𝑡:
𝑠(𝑡)=∫𝑡𝑎‖⃗𝑟′(𝑢)‖𝑑𝑢
(Note we have changed the variable of integration so that 𝑡 is not used in two different contexts)
What is the arc-length function for the helix in our earlier example, starting from 𝑡=0? Since ‖⃗𝑟′(𝑡)‖=√2, we see that
𝑠(𝑡)=∫𝑡0√2𝑑𝑢=√2𝑡
This tells us that after 𝑡 seconds, we have traced out √2𝑡 units of arc length. How could we reparameterize this curve so that its arc-length function is just 𝑠(𝑡)=𝑡 (tracing out 𝑡 units of arc in 𝑡 units of time)?
Definition 8.5 (Unit Speed Curve). A curve ⃗𝑐(𝑡) is unit speed if ‖⃗𝑐′(𝑡)‖=1 for all 𝑡. If the motion begins at 𝑡=0, this means that after 𝑡 seconds the curve has traversed 𝑡 units of length. For this reason, we also call unit-speed curves arc-length-parameterized curves.
In our example, to make the helix unit speed we need to slow it down by a factor of √2: that is, we need ⃗𝑟(𝑡/√2):
⃗𝑟(𝑡√2)=(cos𝑡√2,sin𝑡√2,𝑡√2)
To trace the same portion of the helix, the new parameter runs from 0 to
2𝜋√2. Differentiating with the chain rule confirms that the new speed
is 1.
More generally, suppose ⃗𝑐(𝑢)=⃗𝑟(𝜙(𝑢)) is a reparameterization.
The chain rule gives
⃗𝑐′(𝑢)=⃗𝑟′(𝜙(𝑢))𝜙′(𝑢),
so
‖⃗𝑐′(𝑢)‖=‖⃗𝑟′(𝜙(𝑢))‖|𝜙′(𝑢)|.
If 𝜙 traces the original parameter interval exactly once, a substitution
shows that the two parameterizations have the same length. The absolute value
makes this true even when the direction is reversed. But if the
reparameterization retraces part of the curve or goes around it several times,
that distance is counted each time.
8.2Scalar Line Integrals
Suppose a thin wire follows a curve 𝐶 and has linear density 𝜌. A small
piece of wire with length Δ𝑠 has mass approximately
𝜌Δ𝑠. Adding these small masses and taking a limit leads to the
scalar line integral
∫𝐶𝜌𝑑𝑠.
Figure 8.3 Cut the wire into pieces of equal length. Each piece has its own density, so it
has its own mass 𝜌Δ𝑠—and on the right, its own rectangle of width
Δ𝑠 and height 𝜌. The bead and the rectangle are the same number,
so the mass of the wire is the area under its density graph, and
∫𝐶𝜌𝑑𝑠 is an ordinary single-variable integral once the wire is
straightened out against 𝑠. Choose 𝜌=1 and the graph flattens to height
one: its area is 𝐿, and the line integral has become the arc length. Traverse
the wire backwards and the graph flips while its area does not move.
We can denote the domain 𝐶 as a subscript just like we do for double and triple integrals. When 𝐶 is a closed curve one may optionally modify the integral sign to denote this, writing ∮𝐶𝜌𝑑𝑠. Our first goal is to try and figure out how to compute this in terms of integrals we know how to do. First: recall that we can represent a curve 𝐶 by parameterizing it, writing it as the image of a function ⃗𝑐(𝑡)=(𝑥(𝑡),𝑦(𝑡)) in the plane, or ⃗𝑐(𝑡)=(𝑥(𝑡),𝑦(𝑡),𝑧(𝑡)) in 3 dimensions. We saw above how to express a small bit of arc length along a parametric curve:
𝑑𝑠=‖⃗𝑐′(𝑡)‖𝑑𝑡
To evaluate a density 𝜌(𝑥,𝑦,𝑧) along the curve ⃗𝑐(𝑡), we simply plug the curve into the density. This gives a concrete quantity to integrate.
Definition 8.6 (Scalar Line Integral). If ⃗𝑐(𝑡), 𝑎≤𝑡≤𝑏, parameterizes a curve 𝐶, then
∫𝐶𝜌𝑑𝑠=∫𝑏𝑎𝜌(⃗𝑐(𝑡))‖⃗𝑐′(𝑡)‖𝑑𝑡.
Note the simplest line integrals are just when 𝜌=1. This is the integral just adding up the infinitesimal arc lengths ∫𝐶𝑑𝑠=∫‖⃗𝑐′‖𝑑𝑡. We already met this integral long ago - this gives the arc length of the curve!
Figure 8.4 The same wire, lying flat, with its density standing up over it. The strip above
one piece is Δ𝑠 wide and 𝜌 tall, so its area is the term the sum
was built from—and the area of the whole sheet is ∫𝐶𝜌𝑑𝑠. It is the
same number as the figure above, arrived at by hanging the graph over the curve
instead of straightening the curve out under the graph.
If 𝐿(𝐶) is the length of the wire, its average density is its total mass
divided by its total length:
𝜌avg=1𝐿(𝐶)∫𝐶𝜌𝑑𝑠.
Density is the main interpretation, but the same definition works for any
scalar quantity assigned to the points of a plane or space curve.
Example 8.7 (A Straight-Line Integral). Find ∫𝐶𝑥𝑦𝑑𝑠 for 𝐶 the diagonal of the unit square going from (0,0) to (1,1).
We parameterize the diagonal by
⃗𝑐(𝑡)=(𝑡,𝑡),0≤𝑡≤1.
Along the curve, 𝑥𝑦=𝑡2, while
‖⃗𝑐′(𝑡)‖=‖(1,1)‖=√2.
Therefore
∫𝐶𝑥𝑦𝑑𝑠=∫10𝑡2√2𝑑𝑡=√23.
Example 8.8 (A Circular Integral). Evaluate ∮𝐶𝑥2𝑑𝑠 for 𝐶 the circle of radius 2, traversed clockwise.
A clockwise parameterization is
⃗𝑐(𝑡)=(2cos𝑡,−2sin𝑡),0≤𝑡≤2𝜋.
Here 𝑥2=4cos2𝑡 and ‖⃗𝑐′(𝑡)‖=2, so
∮𝐶𝑥2𝑑𝑠=∫2𝜋08cos2𝑡𝑑𝑡=8𝜋.
Using a counterclockwise parameterization gives exactly the same integral.
This illustrates an important difference between scalar line integrals and the
vector line integrals we will meet later: scalar line integrals are
orientation-independent. Reversing a parameterization changes the sign of
its parameter derivative, but not its magnitude ‖⃗𝑐′(𝑡)‖.
Sometimes a curve 𝐶 might be piecewise, and has a corner where the edges join up. In this case, we evaluate the line integral by doing each segment of the curve separately, and adding the results:
∫𝐶1∪𝐶2𝜌𝑑𝑠=∫𝐶1𝜌𝑑𝑠+∫𝐶2𝜌𝑑𝑠
The pieces must cover the desired curve between them, but they do not need to
be parameterized in matching directions. Reversing any one piece leaves its
scalar line integral unchanged.
Figure 8.5 A path in three pieces, each carrying its own sheet, and the three areas simply
add. The middle side is deliberately traced from its far end backwards, and it
makes no difference at all: a sheet has an area whichever way you sweep it out.
The pieces have to cover the path once between them; they do not have to agree
about which way they are going.
Example 8.9 (A Piecewise Curve). Find the line integral of 𝑥 along the curve 𝐶 that traces the triangle with vertices (0,0), (1,0), and (0,1), starting at the origin and going counterclockwise.
We break this into three curves:
⃗𝑐1(𝑡)=(𝑡,0)⃗𝑐2(𝑡)=(1−𝑡,𝑡)⃗𝑐3(𝑡)=(0,1−𝑡)
Each segment is traced from 𝑡=0 to 𝑡=1. We then just compute the line integral along each and add up the results.
On 𝐶1, the integrand is 𝑥=𝑡 and the speed is 1. On 𝐶2, the
integrand is 𝑥=1−𝑡 and the speed is √2. On 𝐶3, the integrand is
𝑥=0. Thus,
∫𝐶𝑥𝑑𝑠=∫10𝑡𝑑𝑡+∫10(1−𝑡)√2𝑑𝑡+∫100𝑑𝑡=12+√22.
It is straightforward to generalize this to integrals along space curves instead of curves in the plane:
Example 8.10 (A Space-Curve Integral). Compute the line integral ∫𝐶𝑓𝑑𝑠 for 𝑓(𝑥,𝑦,𝑧)=𝑦sin𝑧 and 𝐶 the helix 𝑥=cos𝑡, 𝑦=sin𝑡 and 𝑧=𝑡 for 𝑡∈[0,2𝜋].
Along the helix,
𝑓(⃗𝑐(𝑡))=sin𝑡sin𝑡=sin2𝑡,
and from our earlier arc-length calculation,
‖⃗𝑐′(𝑡)‖=√2.
Therefore
∫𝐶𝑓𝑑𝑠=√2∫2𝜋0sin2𝑡𝑑𝑡=𝜋√2.
Scalar line integrals measure quantities accumulated against length, so their
values do not depend on which direction the curve is traversed. Later, vector
line integrals will measure work and circulation; for those integrals,
orientation will matter.