Parametric Curves

(Com S 477/577 Notes)

Yan-Bin Jia
Oct 17, 2013

Introduction

A curve in R2 (or R3 ) is a differentiable function : [a, b] R2 (or R3 ). The initial point is [a]
and the final point is [b]. The domain of the curve is the interval [a, b]. A portion of defined on
an interval [c, d] [a, b] is called a curve segment.
Example 1. Straight Line
linear. Explicitly, the curve

The line is the simplest curve in the plane as its coordinate functions are

(t) = p + tv = (x0 + tu, y0 + tv),

where v 6= 0,

(1)

is a straight line through the reference point p = (0) = (x0 , y0 ) in the direction v = (u, v). Here, t is the
signed distance from a point (t) on the line to p as scaled by kvk.
As shown on the left, the vector from p to a point
(x,
y)
on the line must be either in the direction of (u, v)
( x, y)
or in its opposite direction. Hence, the cross product of
( x0 , y0)
the two vectors must be zero, that is,
( u, v)

(x x0 , y y0 ) (u, v) = 0.
Expansion of the above cross product yields an implicit
equation of the line that relates the x and y coordinates
of every incident point:
vx uy vx0 + uy0 = 0.

Example 2. Helix1
The curve t (a cos t, a sin t, 0) travels around a circle of radius a > 0 in
the x-y plane. If we allow this curve to rise (or fall) at a constant rate, we obtain
a helix
= (a cos t, a sin t, bt),
where a > 0 and b 6= 0.

Example 3. The curve : R R3 such that

(t) = (et , et , 2t)

The figure is from [1, p. 16].

(2)

shares with the helix in Example 2 the property of rising constantly. However, it lies over the hyperbola
xy = 1 in the x-y plane instead of a circle.

A curve (t) = (x(t), y(t)) is said to be smooth at t = t0 if its kth derivative



(k)
(k)
(k)
(t) = x (t), y (t)
exists for any integer k > 0. A piecewise smooth curve has a domain which is
the union of a finite number of subintervals over each of which is smooth.
Example 4. A line (t) = p + tq is a smooth curve. Here (t) = q and (k) = 0 for k > 1. A polygon,
on the other hand, is a piecewise smooth curve, where each edge determines a subdomain.
Example 5. Cuspidal cubic The curve (t) = (t2 , t3 ) is smooth.2
We have
(t)
(t)
(t)
(k) (t)

= (2t, 3t2 ),
= (2, 6t),
= (0, 6),
= 0,
k 4.

Consider a plane curve : [a, b] R2 . It is called a closed parametric curve if (a) = (b). A
point of self-crossing is a point (t1 ) for which there exist finitely many distinct values t1 , . . . , tn
[a, b], n 2, which satisfy (t1 ) = (t2 ) = = (tn ), and in the case n = 2, [t1 , t2 ] 6= [a, b].
Example 6. A circle is closed. The other three curves all have self-crossings.

This figure is from [2, p. 60].

Velocity, Speed, and Arc Length

Let (t) be a curve. The velocity vector of at t is (t). The speed at t is the length k (t)k. The
meaning is clear if we see (t) as the location of a moving point at time t. The parametrization
(t) is unit-speed if k (t)k = 1 for all values of t. A point where (t) = 0 is called a cusp on the
curve.
The curve (t) is regular if all velocity vectors are different from zero, that is, (t) 6= 0 for
all t.
Example 7. The origin on the cuspidal cubic in Example 5 is a cusp.
Example 8. Consider the curve () = (a cos , a sin ). It has velocity
() = a(cos sin , sin + cos ),
and speed
p
p
k ()k = |a| (cos sin )2 + (sin + cos )2 = |a| 1 + 2 6= 0.

Therefore the parametrization is regular.

The velocity and speed depend on the parametrization. Generally, they vary with the parametrization of the same geometric curve. Non-regularity at a point may be just a property of the
parametrization, and need not correspond to any special feature of the geometric curve. For a
different parametrization the curve may have a non-zero velocity at the corresponding point.
To formulate the length of , we note that the portion
over [t, t + t] is nearly a straight line when t is very small.
(t )
So the length over [t, t + t] can be approximated by
(t +t )

k(t + t) (t)k,
which again is approximated by
k (t)kt.

We divide up into segments, each of which corresponds to a small increment t. As t tends to

zero, we will obtain the exact length. The arc length of from t = a to t = b is thus defined as
Z

b
a

k (t)k dt.

Example 9. Logarithmic spiral3

The curve
(t) = (et cos t, et sin t),
has a spiral motion. We obtain that
(t) =
k (t)k
3

et (cos t sin t), et (sin t + cos t) ,
t
2e .

The figure originally appears in [3, p. 8].

Hence the arc length of starting at (0) = (1, 0), for instance, is
Z t

s=
2eu du = 2(et 1).
0

Reparametrization

Let I and J be intervals. Let : I R3 be a curve and let h be a differentiable function. Then
the composite function = h is a curve called the reparametrization of by h.

I
s

Example 10.

( s ) = (h (s ))

Suppose (t) = ( t, t t, 1 t) on (0, 4). If h(s) = s2 on (0, 2), then
(s) = (h(s)) = (s2 ) = (s, s3 , 1 s2 ).

The curve has been reparametrized by h to yield the curve .

At each time s in the interval J, the curve is at the point (s) = (h(s)) reached by the
curve at time h(s) in the interval. Thus does follow the route of , but it reaches a given point
on the route at a different time than does.
Sometimes one is interested only in the route followed by a curve and not in the particular
speed at which it traverses its route. One way to ignore the speed of a curve is to reparametrize
to a curve
which has unit speed k
k = 1.
Theorem 1 If is a regular curve, then there exists a reparametrization
that has unit speed.

Proof

Consider the arc length function

s(t) =

k (u)k du,

where c is a number in the domain of . It then follows that

s (t) = k (t)k;
namely, the derivative of s is the speed function k (t)k. Since is regular, 6= 0 everywhere;
hence ds
dt > 0 always holds. By a standard theorem of calculus, the function s has an inverse
function t(s), and
1
1
dt
= ds =
.

ds
k (t)k
dt
Now we let (s)

= (t(s)) be the reparametrization of . Then

(s) = (t(s))
Hence, the speed of
is
k
(s)k = k (t(s))k

dt
.
ds
1

k (t(s))k

= 1.

The unit-speed curve

is said to have arc-length parameterization, since the arc length of

from s = a to s = b, a < b, is just b a.

Example 11.

Let us consider the helix = (a cos t, a sin t, bt) in Example 2 again. It has velocity
(t) = (a sin t, a cos t, b).

Hence
k (t)k2 = (t) (t) = a2 sin2 t + a2 cos2 t + b2 = a2 + b2 .
Thus has constant speed:
p
a2 + b 2 .

c = k k =
The arc length from t = 0 is then
s(t) =

c du = ct.

Hence, t(s) = sc . Substituting this into the formula for , we get the unit-speed reparametrization

s 
s
s bs
(s)

=
= a cos , a sin ,
.
c
c
c c

Although every regular curve has a unit-speed reparametrization, this may be very complicated,
or even impossible to write down explicitly, as the following examples show.
Example 12.

The logarithmic spiral

(t) = (et cos t, et sin t),

has speed

t
2e > 0.

So it is regular. The arc length starting at (1, 0) was found in Example 9 to be s = 2(et 1). Hence,
t = ln( s2 + 1), so a unit-speed reparametrization of is given by the rather unwieldy formula

 
 

 


s
s
s
s
+ 1 cos ln + 1
, + 1 sin ln + 1
.
(s)

=
2
2
2
2
Example 13. Twisted cubic4
This is the space curve given by
(t) = (t, t2 , t3 ),

< t < .

We have
(t) =
k (t)k

(1, 2t, 3t2 ),

p
1 + 4t2 + 9t4 .

Since the speed k (t)k is not zero everywhere, is regular. And the
arc-length starting at (0) = 0 is
Z tp
1 + 4u2 + 9u4 du.
s=
0

The above integral has a horrendous closed form not in terms of familiar functions.

Tangent and Normal

The standard method of studying the geometry of a curve at a point is to attach orthonormal
vectors to the point and see how the directions of these vectors change as the point moves on the
curve for an infinitesimal distance. We choose tangent and normal vectors at a regular point.
Let (t) = (x(t), y(t)) be a curve. At a regunormal line
lar point (t) we have a (non-zero) tangent vectangent line
tor (t) = (x (t), y (t)). So the tangent vector
t increasing
represents the velocity of the curve at the point.
The normal vector at (t) is given by rotating

( x( t) ,y ( t) )
the tangent vector counterclockwise through an
(y ( t) ,x ( t) )
angle 2 . It is given by (y (t), x (t)). Note
that (x (t), y (t)) (y (t), x (t)) = (x (t))2 +
(y (t))2 > 0.
If (t) is a unit-speed curve, then both the tangent vector and the normal vector are unit vectors.
By convention they are denoted as T and N , respectively, with the cross product T N = 1.
For a parametric curve we have a tangent line and a normal line at each regular point (t).
The tangent line to the curve at (t) passes through (t) and is parallel to (t) 6= 0. So it has
the parametric equation


x(s), y(s) = (t) + s (t),
s (, ),
4

The figure originally appears in [3, p. 14].

or equivalently, the algebraic equation


 

(x, y) (t) y (t), x (t) = 0.
The normal line at (t) passes through the point and is parallel to (y (t), x (t)). So its equations
are of the form




x(s), y(s) = (t) + s y (t), x (t) ,
s (, ),
or equivalently,



x(s), y(s) (t) (t) = 0.
Example 14. Crunodal cubic5 is described as


(t) = t2 1, t(t2 1) .
Find its tangent and normal lines of the curve at the points t = 1, 0.
We obtain that
(t) =
(1) =
(1) =

(0) =
(1) =

(2t, 3t2 1),

(2, 2),
(2, 2),
(0, 1),
(0, 0).

Here = (0, 0) is referred to as a double point since it is attained at both

t = 1 and t = 1. The tangent lines at this double point are respectively
(x, y) = s(1, 1),

or equivalently,

y = x,

and
(x, y) = s(1, 1),

or equivalently,

y = x.

The normal lines at the double point are respectively

(x, y) = s(1, 1),

or equivalently,

y = x,

and
(x, y) = s(1, 1),

or equivalently,

y = x.

At t = 0, we have (0) = (0, 1), and the tangent line at (0) is

(x, y) = (1, 0) + s(0, 1),

or equivalently,

x = 1.

or equivalently,

y = 0.

The normal line at (0) is

(x, y) = (1, 0) + s(1, 0),

This figure is taken from [2, p. 60].

References
[1] B. ONeill. Elementary Differential Geometry. Academic Press, Inc., 1966.
[2] J. W. Rutter. Geometry of Curves. Chapman & Hall/CRC, 2000.
[3] A. Pressley. Elementary Differential Geometry. Springer-Verlag London, 2001.