3.

1 Vector-Valued Functions
Vectors differ from regular numbers because they have both a magnitude
(length) and a direction. In our three dimensional world, vectors have one
component for each direction, and are denoted by

r(t) = (x, y, z) . (1)

Another common notation uses the unit vectors i, j and k for the x, y and z
direction respectively. The vector is then written as

r(t) = xi + yj + zk . (2)

These numbers x, y and z give a position relative to some reference point,

1.0
y
0.5

0.0
1.0

z 0.5

0.0
0.0

0.5

x
1.0

Figure 1: r(t) = i + j + k

which we call the origin and has coordinates (0, 0, 0). An analogy is how you
might give someone directions (before the advent of Google Maps!) Imagine
that you lived in an apartment in a city with perpendicular streets (like
American cities.) You could direct someone by saying that you live in the
third street west (x position) and seventh street from the north (y position)
from their apartment, and that also you live on the sixth floor (z position.)
Only with all three pieces of information could your friend find you. In other
words, vectors give complete information about position. Recall that a we can

1
 y

-4 -2 2 4 x

-2

-4

Figure 2: r(t) = 5 cos ti + 5 sin tj

add two vectors r1 = (1, 3, 1) and r2 = (0, 3, −2) to get r1 + r2 = (1, 6, −1),
and so on. The length of a vector is given by its magnitude, the formula
for which is p
||r|| = x2 + y 2 + z 2 . (3)
Similarly, when we think of functions, we are used to parametric functions
such as
x(t) = 3t , x(t) = t2 , (4)
etc. However, it is reasonable to think that this can be generalised to three (or
more!) dimensions. In this instance they describe shapes in three-dimensions.
We define vector-valued functions to be functions of a real variable with
several component functions depending on a parametric variable t as

r(t) = f (t)i + g(t)j + h(t)k . (5)

An important word to take note of is “real”. In this course we will only

be concerned with functions of real variables t ∈ R. The domain of r(t),
denoted D(r), is the set of values of t for which r(t) is defined. Let us consider
an example:
Example: What is the domain of
√
r(t) = ti + tj + ln(t − 3)k . (6)

2
 5- 5
x
y 0
0 5

-5

15

z 10

Figure 3: r(t) = 5 cos ti + 5 sin tj + tk

Solution: Where is r(t) defined? Let’s look at it component by compo-

nent. Firstly, f (t) = t is √
defined for any real number, and so has the range
(−∞, ∞). Next, g(t) = t is defined for any non-negative number, so the
range is [0, ∞). Finally, the component h(t) = ln(t − 3) is well-defined when
the argument of ln is positive; in other words, the range is (3, ∞). So what
is the range of r(t)? It is only defined where all its components are defined.
Therefore its range is (3, ∞).

In this course we will only consider vector-valued functions in two or three

dimensions, with values in t ∈ R2 and t ∈ R3 respectively . Let us look at
simple examples. In two dimensions, the parametric equations of a circle of
radius 5 are
f (t) = 5 cos t , g(t) = 5 sin t , (7)
which shown in Figure 2. we now consider a three-dimensional example with

3
equations
f (t) = 5 cos t , g(t) = 5 sin t , h(t) = t , (8)
which is instead a spiral (circular helix) and is drawn in Figure 3.

3.1.1 Vector Form of a Line Segment

How do we write a line segment in the form of a parametric equation? Con-
sider Figure 4. If we consider the line that connects the positions r0 and r1 ,

tHr1- r0 L
r0 r r1

Figure 4: Line segment

we see that to move along this line, we could start at position r0 and move
to other point r via
r = r0 + vt , (9)
where v is the rate at which the position changes (the “velocity”) and the
parameter t describes how much “time” has elapsed in which the change has
occurred. Let r0 be at t = 0 and r1 be at t = 1. Then
r1 = r0 + v , (10)
and so
v = r1 − r0 . (11)
This means we arrive at
r = r0 + (r1 − r0 )t
(12)
= (1 − t)r0 + tr1 .
This is the two-point form of a line. Notice that if 0 ≤ t ≤ 1 then it describes
the line segment from r0 to r1 .

4
3.1.2 Limits and continuity
We define the limit of a vector-valued function to be

lim r(t) = L , (13)

t→a

if for r(t) = (x(t), y(t), z(t))

lim x(t) = Lx , lim y(t) = Ly , lim x(t) = Lz , (14)

t→a t→a t→a

with L = (Lx , Ly , Lz ). Using this definition of a limit, we say that a vector-

valued function is continuous at the point t = a if

lim r(t) = r(a) . (15)

t→a

A vector-valued function is continuous in an interval if it is continuous at

ΘHtL
2.0

1.5

1.0

0.5

-2 -1 1 2 t
-0.5

-1.0

Figure 5: Heaviside step function

all points in the interval. We can see what this means with an example.
Clearly, the vector-valued function r(t) = (t, t2 , t + t3 ) is continuous because
the limit at t → a is r(a) = (a, a2 , a + a3 ), and is continuous for all real
values of a, i.e. on the interval (−∞, ∞). However, let us instead consider
r(t) = (θ(t), t2 , t + t3 ), with the new function given by
(
1 if t ≥ 0
θ(t) = (16)
0 if t < 0 .

which is the Heaviside step function. It is shown in Figure 5. Notice that

it is continuous on the intervals [0, ∞) and (−∞, 0), but if we consider the

5
interval [−2, 0], it is discontinuous at the point t = 0. To see this, we notice
that the limit at t = 0 in this interval must be 0 since we must take the limit
from below. However, the value at t = 0 is 1. Therefore, because one of the
components of r(t) = (θ(t), t2 , t + t3 ) is discontinuous r(t) is discontinuous.

3.1.3 Derivatives
We now define derivatives of vector-valued functions using limits. In order
to be differentiable, the vector-valued function must be continuous, but the
converse does not hold. The derivative is defined as
r(t + h) − r(t)
r0 (t) = lim , (17)
t→h h
provided the limit exists. Clearly, it exists only when the function is contin-
uous. This is shown in Figure 6. Notice that the derivative r0 (t) is tangent
to the curve traced out by r(t), and points in the direction of increasing t.
In mechanics, r0 (t) = v(t). Alternative notations include r0 , dtd r(t), and dr
dt
.
Let’s look at an example:
3 √
r(t) = et i + 1 + t2 j − sin t k ,
3 t (18)
⇒ r0 (t) = 3t2 et i + √ j − cos t k .
1 + t2
Let us recall some properties of derivatives that apply to vector-valued func-
tions :

Rules for Differentiation

d
1. dt
c = 0,
d
2. dt
(kr) = k dtd r ,
3. d
(r
dt 1
+ r2 ) = r01 + r02 ,
4. d
(r
dt 1
− r2 ) = r01 − r02 ,
5. d
dt
(ar1 + br2 ) = ar01 + br02 ,
6. d
dt
(f (t)r) = f 0 (t)r + f (t)r0 ,
where c is a constant vector and a, b, k are constants, and f is any function
of t. The last of these is the chain rule.

6
 1.0 1.0

0.8 rHt+hL-rHtL 0.8 r'HtL

0.6 rHt+hL 0.6

0.4 rHtL 0.4 rHtL

0.2 0.2

0.2 0.4 0.6 0.8 1.0 0.2 0.4 0.6 0.8 1.0

Figure 6: Derivatives of vectors

3.1.4 Tangent Lines and Tangent Vectors

We mentioned in the last section that the derivative of a vector-valued func-
tion is tangent to the vector-valued function at that point. From this, we
define the tangent line to r(t) at t0 to be the line parallel to the derivative
r0 (t0 ). The equation for the tangent line is

R(t) = r0 + v0 t , (19)

with r0 = r(t0 ) and v0 = r0 (t0 ). This is clearly the vector form of a line
segment.
Example: Find the tangent line of a circular helix with the equation

r(t) = ρ cos t i + ρ sin t j + ct k , (20)

Solution: The derivative is

r0 (t) = −ρ sin t i + ρ cos t j + c k . (21)

Let us calculate the tangent line at t = π. We then have

r0 = r(π) = −ρ i + πc k , (22)

and
v0 = r0 (π) = −ρ j + c k . (23)

7
 1.0

0.8 r'HtL

0.6 RHtL

0.4 rHtL

0.2

0.2 0.4 0.6 0.8 1.0

Figure 7: Tangent Line

Therefore the tangent line has the equation

R(t) = (−ρ i + πc k) + (−ρ j + c k)t
(24)
= −ρ i − ρt j + c(t + π) k .

3.1.5 Derivatives involving vectors: dot and cross prod-

ucts
Recall that for vectors r1 = (x1 , y1 , z1 ) and r2 = (x2 , y2 , z2 ) the dot product
is given by
r1 · r2 = x1 x2 + y1 y2 + z1 z2 , (25)
√
and the magnitude is given by ||r|| = r · r. The cross product is given by
r1 × r2 = (y1 z2 − z1 y2 )i + (z1 x2 − x1 z2 )j + (x1 y2 − y1 x2 )k , (26)
and its magnitude is
||r1 × r2 || = ||r1 || · ||r2 || sin θ , (27)
where θ is the angle between the vectors. Some useful identities are
r1 · r2 = r2 · r1 , r1 × r2 = −r2 × r1 . (28)

8
Then, applying the derivatives to these products gives
d
(r1 · r2 ) = r01 · r2 + r1 · r02 ,
dt (29)
d
(r1 × r2 ) = r01 × r2 + r1 × r02 .
dt
Finally in this section, we use these rules to prove a theorem. You probably
remember that the tangents you met previously are perpendicular (normal)
to the curve. Is that true here? The theorem below says yes in certain
conditions.
Theorem 1.1 If r(t) is a real vector-valued function with constant magnitude
||r(t)||, then r · r0 = 0, which means that r0 is perpendicular to r, i.e. r ⊥ r0 .
Proof
||r||2 = r · r
d (30)
⇒ 0 = (r · r) = r0 · r + r · r0 = 2r0 · r ,
dt
as required.

3.1.6 Definite Integrals of Vector-Valued Functions

Let r(t) be a continuous (not necessarily differentiable) function on an inter-
val a ≤ t ≤ b. The definite integral is defined as
Z b N
X
r(t) dt = lim r(tk )∆tk
a max∆tk →0
k=1 (31)
Z b Z b Z b
= x(t) dt i + y(t) dt j + z(t) dt k .
a a a

If desired, this can be understood by considering Figure 8. However, the de-

tails are not necessary, and the formula (31) suffices for our purposes. This
gives rise to the following useful properties:

Rules of Integration
Rb Rb
1. a (kr(t)) dt = k a r(t) dt ,

9
 y
60

50

40

30

20

10

3.6 3.8 4.0 4.2 4.4 4.6 x

Figure 8: Riemann Integral

Rb Rb Rb
2. a
(c r1 (t) + d r2 (t)) dt = c a
r1 (t) dt + d a
r2 (t) dt .

Here k, c, and d are constants. Also recall that if the graph f lies above the
graph g, then the area between the graphs is given by
Z b


Area = f (x) − g(x) dx . (32)
a

Also recall that the volume of a solid that has cross-sectional area A(x) has
volume Z b
Volume = A(x) dx . (33)
a
Let us consider an example. √
Example: Find the integral of r(t) = t3 i + tj − sin πt 2
k with limits 0 and
2.
Solution:
Z 2 √

3 πt
t i + tj − sin k dt
0 2
  2
1 4 2 3/2 2 πt
= t i + t j + cos k
4 3 π 2
" √ # 
0
 (34)
4 2 2 2
= 4i + j − k − 0i + 0j + k
3 π π
√
4 2 4
= 4i + j − k.
3 π

10
We made use of the fact that cos 0 = 1 and cos π = −1.

Next we define the antiderivative for a vector-valued function r(t), which

is itself a vector-valued function R(t) which is given by

R0 (t) = r(t) . (35)

Hence, Z
r(t) dt = R(t) + C . (36)

Again, let’s look at an example,

Z  
1
i + cos 2t j dt
t−1 (37)
1
= log |t − 1| i + sin 2t j + C .
2
There are the following properties
1. dtd r(t) dt = r(t) ,
R

2. r0 (t) dt = r(t) + C ,
R

The last thing we consider in this section is The Fundamental Theorem

of Calculus, which states
Z b
b
r(t) dt = R(t) a = R(b) − R(a) . (38)
a

Let us return to the previous example, which gives

Z 3 
1
i + cos 2t j dt
2 t−1
  3
1
= log |t − 1| i + sin 2t j
2 2 (39)
1
= (log 2 − log 1) i + (sin 6 − sin 4)j
2
1
= log 2 i + (sin 6 − sin 4)j ,
2
where of course log 1 = 0.

11
3.1.7 Arc Length and Changing Parameters
We say that a curve is smoothly parameterised by r(t), or that r(t) is a
smooth function of t if r0 (t) exists, is continuous and that r0 (t) 6= 0 for all
t. If a function is smooth, we can calculate its arc length, which is given by
s
Z b Z b  2  2  2
dr dx dy dz
L= dt = + + dt . (40)
a dt a dt dt dt
As an example, lets find the arc length of r(t) = 5 cos ti + 5 sin tj + tk from
0 to π2 . Then r0 (t) = −5 sin ti + 5 cos tj + k, and
p √
||r0 (t)|| = (−5 sin t)2 + (5 cos t)2 + 1 = 26 , (41)
which gives arc length
Z π/2 √ √ π
L= 26 dt = 26 . (42)
0 2

3.1.7.1 Parameterising a curve using arc length

In order to use the arc length to parameterise a curve, we select a point to
be our reference point, P , and choose an orientation so that one direction is
positive and the other negative. Then, the “signed” arc length parameterises
the curve via
x = x(s) , y = y(s) , z = z(s) . (43)
In other words, we view the position on the curve as a function of its arc
length (“distance”) from the reference point, P . This is known as arc length
parameterisation. To understand how it works, let us consider an example
of a circle of radius 5. It has the parametric equation
r(t) = 5 cos t i + 5 sin t j , 0 ≤ t ≤ 2π , (44)
in which t plays the role of a radian angle measured from the point O on the
x-axis to an arbitrary point P (x, y), as shown in Figure 9. The arc length is
Z tp
s≡L= (−5 sin t)2 + (5 cos t)2 dt = 5t , (45)
0
and therefore
s = 5t , or t = s/5 . (46)
As a result, the circle is now parameterised by
x = 5 cos(s/5) , y = 5 sin(s/5) , 0 ≤ s ≤ 10π . (47)

12
 y

4 PHx,yL

t
-4 -2 2 4 x
H5,0L
-2

-4

Figure 9: Arc length parameterisation of a circle

3.1.7.2 Change of parameter

More generally, we can change the parameter of a curve using the chain rule.
If r(t) is a vector-valued function that is differentiable with respect to t, we
can change parameter to τ using t = g(τ ), where g is differentiable with
respect to τ . Then r(g(τ ) is differentiable with respect to τ and we have

dr dr dt
= . (48)
dτ dt dτ
If g(τ ) is smooth, then the change is called a smooth change of parameter.
dt
Also, if dτ > 0 for all τ , then it is a positive change of parameter.
dt
Alternatively, if dτ < 0 for all τ , then it is a negative change of parameter.
Using these notations, we state that if C is the graph of a smooth vector-
valued function r(t), with some reference point r(t0 ), then the arc length
parameter is given by a positive change of parameter and is found from the
formula
s
Z t Z t  2  2  2
dr dx dy dz
s= du = + + du . (49)
t0 du t0 du du du

For an example, let us consider the vector-valued function r(u) = 2 cos u i +

2 sin u j + u k, where u is some parameter replacing t in order to avoid poor

13
notation in the integral. We will use t0 = 0 as our reference point. Then
dr √
= 5, (50)
du
and so Z t√ √
s= 5du = 5t . (51)
0
Therefore, the reparameterisation give us
   
s s s
r(s) = 2 cos √ i + 2 sin √ j + √ k. (52)
5 5 5

Important note: Although, this “new” variable u might seem confusing,

it really is the same thing as t. It is a dummy variable representing t for
the purpose of integration. Otherwise we would end up R t with equally con-
fusing and mathematical incorrect expressions such as 0 ||dr/dt||dt, which
obviously we want to avoid.
Although we can define many different reparameterisations, the arc length
parameterisation has some special properties. These are
(a) If r(t) is a vector-valued function with parameter t, with arc length
parameter s,
dr ds
= . (53)
dt dt

(b) If r(t) is a vector-valued function and arc length parameter s, the tan-
gent vector for any s has length given by
dr
= 1. (54)
ds

(a) If r(t) is a vector-valued function with ||dr/dt|| = 1 for any value of

parameter t, then s = t − t0 is an arc length parameter with reference
point t0 .
The proof of these is simple. For (a), apply the Fundamental Theorem of
Calculus to (49). Then the antiderivative is s and therefore, ds/dt is the
argument of the integral, which is ||dr/dt||, and we get the result. (b) comes

14
directly from (a) if we take t = s. Finally, (c) comes from setting ||dr/dt|| = 1
in (49), which gives
Z t Z t
dr
s= du = du = u|tt0 = t − t0 , (55)
t0 du t0

as required.

3.1.8 Tangent, Normal and Binormal Vectors

We previously defined the tangent vector for the curve of a vector-valued
function to be the vector r0 . We often instead want the unit tangent vector
which is the unit vector along the tangent line, i.e.

v(t) r0 (t)
T= = 0 , (56)
||v(t)|| ||r (t)||

where of course ||v(t)|| is the magnitude of the vector. This is important to

allow us to define the normal vector.

3.1.8.1 Normal vectors

Recall that we said that if a vector-valued function has a constant norm ||r||,
then the tangent (and hence unit tangent) vector to the curve C is orthogonal
to r. If we apply the same reasoning to T, which has constant norm 1, we
see that T0 must be orthogonal to T. We say that T0 is normal to C, and,
provided T0 6= 0, we define the (principal) unit normal vector to C at t
as
T0 (t)
N(t) = . (57)
||T0 (t)||
Note that the condition T0 6= 0 means that we cannot define the normal
vector for straight lines. An indication of these vectors is given in Figure 10.

Example: Find the unit tangent and unit normal vectors to the curve

r(t) = 3 cos ti + 3 sin tj + 4tk .

15
 y
10

N(t)
8
T(t)
6

2
0 1 2 3 4 x

Figure 10: Unit tangent and unit normal vectors.

Solution: We first find the tangent vector,

r0 (t) = −3 sin ti + 3 cos tj + 4k ,

which has magnitude

p
||r0 (t)|| = (−3 sin t)2 + (3 cos t)2 + 42 = 5 ,

which gives us the unit tangent vector

3 3 4
T(t) = − sin ti + cos tj + k .
5 5 5
Then
3 3
T0 (t) = − cos ti − sin tj ,
5 5
which has magnitude
s 2  2
0 3 3 3
||T (t)|| = − cos t + − sin t = ,
5 5 5

which gives us the unit normal vector

N(t) = − cos ti − sin tj .

Note that this vector is the parametric form of a circle in a plane with a
minus sign. Hence the normal vector is parallel to the xy-plane and points
from the curve (spiral) towards the z-axis, i.e. in the opposite direction to

16
the vector that would describe a circle.
When the curve is parameterised by its arc length parameter, the calculation
of these vectors is simpler. We have the formula

T(s) = r0 (s) ,
r00 (s) (58)
N(s) = 00 .
||r (s)||

Example: Calculate the unit tangent and unit normal vectors of

r(t) = 3 cos ti + 3 sin tj , (0 ≤ t ≤ 2π) .

Solution: The arc length parameter is

Z t
s= 3du = 3t ,
0

and therefore
r(s) = 3 cos s/3i + 3 sin s/3j .
We then find the tangent vector,

T(s) = r0 (s) = − sin s/3i + cos s/3j .

Then
1 1
r00 (s) = − cos s/3i − sin s/3j ,
3 3
which has magnitude
s 2  2
00 1 1 1
||r (s)|| = − cos s/3 + − sin s/3 = ,
3 3 3

which gives us the unit normal vector

N(s) = − cos s/3i − sin s/3j .

17
3.1.8.2 Binormal vectors
The binormal vector is defined by

B(t) = T(t) × N(t) . (59)

By definition it is orthogonal to both T and N, and since both these vectors

are unit vectors, B is also a unit vector, since

||T(t) × N(t)|| = ||T||||N|| sin π/2 = 1 . (60)

These three vectors define three mutually perpendicular planes at any point
0.0
0.5
1.0
1.0

B rHtL

0.5

N
T
0.0
0.0

0.5

1.0

Figure 11: Normal, binormal and tangent vectors, and the planes containing
them.

on a curve C of a vector-valued function r(t). These are the TB-plane or

rectifying plane, the TN-plane or osculating plane, and the NB-plane
or normal plane. This is shown in Figure 11. This defines a coordinate
system known as the TBN-frame or Frenet frame, which is a frame that
has its origin move along the curve C, with its axes rotating as we move
along the curve (since T, N and B rotate). We can express B directly in
terms of the original vector-valued function by

r0 (t) × r00 (t)

B(t) = 0 , (61)
||r (t) × r00 (t)||

18
which in the situation where we use the arc length parameter, becomes
r0 (s) × r00 (s)
B(s) = . (62)
||r00 (s)||
This is a good moment to pause to think again about why we might want
to use arc length parameterisation. We have already seen that it simplifies
the simplifies the expressions for the unit tangent, unit normal and binormal
vectors. This is itself useful, but it is in the definition of curvature that it
becomes vital.

3.1.9 Curvature
For a curve C of a vector-valued function r, the unit tangent vector T is a
measure of how quickly we move along C. If we then consider the derivative
dT/ds, this is a measure of how quickly the motion changes direction or
“curves” at that point. Therefore, we define the curvature of C to be
dT
κ(s) = = ||r00 (s)|| , (63)
ds
which is a number which tells us how much a curve bends. Of course, this
is a bit simplistic: a curve in three dimensions can bend in three different
directions, so how can one number give us all this information? The answer
is that it doesn’t, and in fact, we should also consider dN/ds and dB/ds
to gives us a full picture. The concept of T changing along the curve is
represented in Figure 12.
y y
3.0 2.0

2.5 T
1.5 T
2.0

1.5 T 1.0

T
1.0
0.5
0.5
T
0.0
0 1 2 3 4 x 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 x

Figure 12: T along a line and a generic curve.

Example: Find the curvature of

19
 (a) The circle r(s) = 3 cos s/3i + 3 sin s/3j , (0 ≤ s ≤ 6π) .

(b) The line r(s) = r0 + sv .

Solution: We apply the formula for the curvature:

(a) The second derivative is

1 1
r00 (s) = − cos s/3i − sin s/3j ,
3 3
and so the curvature is
s 2  2
1 1 1
κ(s) = − cos s/3 + − sin s/3 = .
3 3 3

(b) The first derivative is

r0 (s) = 0 + v = v ,

since r0 and v are constants, and the second derivative will vanish as
a result:
r00 (s) = 0 .
This implies the curvature is zero:

κ(s) = 0 .

More generally, we find that for a circle of radius a, the curvature is 1/a
and so the the larger a circle, the smaller the curvature, which makes sense
intuitively. Also, any line has zero curvature, which again makes sense.
If we want to use more a more general parameter t, we can write the curvature
as
||T0 (t)||
κ(t) = ,
||r0 (t)||
(64)
||r0 (t) × r00 (t)||
or κ(t) = ,
||r0 (t)||3

but the expression is much simpler for the arc length parameter. We will not
prove these, but proofs can be found in section 12.5 of the textbook. Often,

20
the second of these formulae is more practical to use. These are necessary
when calculating the arc length parameter is complicated. For example if the
argument in the square root in equation (49) cannot be reduced to a square
it might be very difficult to integrate the result. In these cases the formulae
for the general parameter are more appropriate.
Example: Consider the vector-valued function r = 2 cos ti + 3 sin tj for
0 ≤ t ≤ 2π. Find the curvature at the points t = π/2 and t = π.
Π
2
3

-2 -1 1 2 0

-1

-2

-3

Figure 13: The graph of r = 2 cos ti + 3 sin tj.

Solution: The graph of this vector-valued function is an ellipse, shown in

Figure 13. You should recognise it as the polar coordinates of an ellipse.
Therefore we are looking for the curvature at the endpoints of the major and
minor axes. Since the curvature equations are three-dimensional, we view
the ellipse to sit in the xy-plane of a three-dimensional coordinate system
and write
r = 2 cos ti + 3 sin tj + 0k .
Then
r0 = −2 sin ti + 3 cos tj ,
r00 = −2 cos ti − 3 sin tj ,

21
and we find the cross-product of these
i j k
r0 × r00 = = 6 sin2 t + 6 cos2 t k = 6k .
 
−2 sin t 3 cos t 0
−2 cos t −3 sin t 0
In other words, this vector, which is parallel to the binomial vector points
along the z axis. We now need its norm, which is obviously 6. We also need
the norm of r0 , which is
p p
||r0 || = (−2 sin t)2 + (3 cos t)2 = 4 sin2 t + 9 cos2 t .

This is the reason we would be unwise to use the arc length parameter here
since this square root appears in the definition of the arc length parameter.
Obviously this makes life difficult, so instead we find the curvature via (64)
to find
||r0 (t) × r00 (t)|| 6
κ(t) = 0 3
=
3/2 .
||r (t)|| 4 sin2 t + 9 cos2 t
We now can find the particular values we are interested in
6 3
κ(π/2) = 3/2
= ,
(4) 4
6 2
κ(π) = 3/2
= .
(9) 9
Notice that the curvature is greater at the end of the major axis than at the
end of the minor axis. Between this points the curvature is between these
values as we get a mixture of the sin and cos terms.

3.1.9.1 Radius of Curvature

Our previous example showed us that the curvature of an ellipse varies as we
go round the ellipse. This is not surprising, but it highlights that the circle is
special with a constant curvature given by 1/a, where a is the radius of the
circle. We can make use of this fact to define osculating circle or circle of
curvature at a point P on a curve to be the circle with the same curvature
as the curve at P that also shares a tangent line, and lies on the concave
side of the curve. Hence, the circle has radius ρ = κ1 . ρ is called the radius
of curvature and the centre of the osculating circle is called the centre of
curvature. This is shown in Figure 14.

22
 Centre of
curvature
Ρ=1/Κ

Figure 14: The centre of curvature and radius of curvature for a curve.

3.1.9.2 Curvature in two dimensions

In two dimensions, we can use an angle, φ defined with respect to the x-axis
to parameterise curvature. You might notice that this is essentially the same
idea as a polar angle for polar coordinates. This is shown in Figure 15. We
y
2.0
T
1.5 Φ
T
1.0
Φ
0.5
Φ

0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 x

Figure 15: Curvature in two dimensions.

then express the tangent vector along the i and j directions as

T = cos φi + sin φj , (65)

23
which of course gives us
dT
= − sin φi + cos φj . (66)
dφ
The key point is that we can relate this to the derivative with respect to the
arc length parameter using the chain rule. We find
dT dT dφ
= . (67)
ds dφ ds
Therefore the curvature becomes
dT dφ dT p dφ dφ
κ(s) = = = (− sin φ)2 + (cos φ)2 = , (68)
ds ds dφ ds ds
which we present compactly as
dφ
κ(s) = , (69)
ds
This is the rate of change of φ with respect to s. In other words, in two
dimensions, curvature can be interpreted as how quickly the polar coordinate
changes.

3.1.10 Mechanics
You might have noticed some striking similarities between what we have been
doing and your experience in Newtonian mechanics. This is not a coincidence.
Although many mechanics problems are of linear motion or circular motion,
and vectors tend to be treated separately when considering work, moments
and so on, in fact we should always think three-dimensionally. Think again
about arc length. For a straight line is it the same as distance. For circular
motion, it is a portion of the orbit. Either way, the instantaneous speed
along the curve is ds/dt, which is the rate of change of the arc length. More
generally, we can define the velocity as
ds
v(t) = T(t) , (70)
dt
since T is a unit vector which gives the direction of the rate of change of
position. Of course, the acceleration is just the derivative of the velocity,

24
 dv
a= dt
.Therefore, we have the following equations
dr
velocity = v(t) = ,
dt
dv d2 r
acceleration = a(t) = = 2, (71)
dt dt
ds
speed = ||v(t)|| = .
dt
What about displacement? We can simply integrate the velocity to give this,
assuming the velocity is known via
Z t2 Z t2
dr(t)
∆r = v(t) dt = dt = r(t2 ) − r(t1 ) . (72)
t1 t1 dt
Of course if we already know the position vector (i.e. the vector-valued
function for whose graph gives the motion of the particle), we can directly
substitute or recognise r as the antiderivative of v using the Fundamental
Theorem of Calculus. This is something you probably understand very well,
but it is helpful to look at it in this mathematical way to help us to under-
stand what all the previous work on vector-valued functions was about. The
distance travelled comes from integrating the norm of the velocity (speed)
over the time interval, which is
Z t2 Z t2
dr
s= dt = ||v|| dt . (73)
t1 dt t1

Although we use s here, this is not actually an arc length parameter unless
we say t1 is a reference point. The use of s is an unfortunate coincidence due
to the fact that in mechanics s is the standard way to represent displacement.
This is actually an arc length over an interval.
Returning to the acceleration, you know from circular motion that we should
expect both a tangential acceleration and radial acceleration in general. This
radial acceleration is in fact along the direction of the normal vector, and
using v = ds/dtT, we have the following decomposition of the acceleration
d2 s
 
ds
a = 2T+κ N. (74)
dt dt
We can define two separate components of acceleration
a = aT T + aN N , (75)

25
with
d2 s
 
ds
aT = 2 , aN = κ . (76)
dt dt
We call aT the tangential scalar component of acceleration and aN
the normal scalar component of acceleration and aT T the tangen-
tial vector component of acceleration and aN N the normal vector
component of acceleration. Finally, in terms of velocity and acceleration,
these are given by
v·a ||v × a|| ||v × a||
aT = , aN = , κ= . (77)
||v|| ||v|| ||v||3

Example: Find the velocity, acceleration and tangential and normal accel-
erations of the motion

r(t) = 2 cos ti + 3 sin t cos θj + 3 sin t sin θk ,

assuming θ is a constant.
Solution: Firstly, this is in fact an ellipse rotated out of the xy-plane by
and angle θ about the x-axis. Hence the y and z components have changed.
This is shown in Figure 17. It describes the orbit of a planet around a star
if the orbit is at an angle to the xy-plane. The velocity is given by

v(t) = −2 sin ti + 3 cos t cos θj + 3 cos t sin θk ,

with magnitude given by

p p
||v|| = (−2 sin t)2 + (3 cos t cos θ)2 + (3 cos t sin θ)2 = 4 sin2 t + 9 cos2 t .

The acceleration is then

a = −2 cos ti − 3 sin t cos θj − 3 sin t sin θk .

To find the components of acceleration, we use equation (77). The dot prod-
uct is
v · a = 2 − 9 sin2 (θ) sin(2t) ,

and the cross product is

v × a = −6 sin θj + 6 sin θk ,

26
 -2
x
-1
0
1
2

Θ
0 z

-1

2
0
-2

Figure 16: An ellipse in the xy-plane (red) and rotated away from the plane
(blue).

with norm √
||v × a|| = 6 2 sin θ
Note that this is 0 if θ = 0. In other words if the motion is in the xy-plane
there is no normal acceleration. The components of acceleration are
2 − 9 sin2 (θ) sin(2t)


v·a
aT = = √ ,
||v|| 4 sin2 t + 9 cos2t
√ (78)
||v × a|| 6 2 sin θ
aN = =
3/2 .
||v||3 4 sin2 t + 9 cos2 t

3.1.11 Planetary Orbits - Kepler’s Laws

Recall from section 1.6.2 Kepler’s Laws of Planetary Motion:
Kepler’s Laws of Planetary Motion:
1. Law of Orbits: The motion of each planet traces an ellipse with the
Sun at one of the foci.
2. Law of Areas: The line joining the Sun to the centre of the planet
sketches out equal areas in equal times.

27
 3. Law of Periods: The square of the orbital period of a planet is pro-
portional to the cube of the semi-major axis of its orbit.

How did Kepler derive these? We now have the tools to understand this. If a
particle moves under the influence of a single force acting from a fixed point,
we say the particle is moving in a central force field, with the force called
a central force. If we look at the problem in a coordinate system with this
fixed point as the origin, the acceleration acts along the radius vector (since
the force does) but in the opposite direction. This means

r × a = 0. (79)

As a result, we have another conserved quantity, l = r × v since

dl d
= (r × v) = (v × v) + (r × a) = 0 × 0 = 0 . (80)
dt dt
If you are familiar with the concept of angular momentum, this is essen-
tially the same thing: for angular momentum replace v with p, the ordinary
momentum.

3.1.11.1 Newton’s Universal Law of Gravitation

The relevance of central forces is due to Newton’s Universal Law of Gravita-
tion, which you should recognise as
GM m
||F|| = . (81)
r2
If you are not yet familiar with the vector version of this force, note that it
acts between two bodies. If M >> m, say in the case of the Sun and the
Earth, then we treat the centre of M as fixed and the force acts along the
radius vector, and points to the centre of M . Hence,
 
r  r
F = ||F|| − = ||F|| − , (82)
||r|| r
r
where − ||r|| is a unit vector anti-parallel to r and

GM m
F=− r. (83)
r3

28
Therefore the acceleration is
GM
a=− r. (84)
r3
Take the initial position and velocity to be

r0 = r0 i , v0 = v0 j , (85)

which implies
l = r0 i × v0 j = r0 v0 k , (86)
which is constant. If we now define a unit vector

u = cos θi + sin θj , (87)

which follows the planet, we can write the radius vector in polar coordinates
using this convenient unit vector:

r = r cos θi + r sin θj = ru , (88)

and hence
GM
a=− u. (89)
r2
Now,
d d du dr
v= r = (ru) = r + u, (90)
dt dt dt dt
and since l is constant,
 
du dr du dr du
l = r × v = (ru) × r + u = r2 u × + r u × u = r2 u × , (91)
dt dt dt dt dt

since u × u = 0. Since u depends on θ we must use the chain rule to find its
derivative
du du dθ dθ
= = (− sin θi + cos θj) , (92)
dt dθ dt dt
and so
du dθ
u× = k, (93)
dt dt
and therefore
dθ
l = r2 k . (94)
dt

29
If we now take the cross product between a and l, we get
 
GM 2 dθ
a × l = − 2 (cos θi + sin θj) × r k
r dt
(95)
dθ du
= GM (− sin θi + cos θj) = GM .
dt dt
Since l is constant, we find
d dv du
(v × l) = × l = a × l = GM , (96)
dt dt dt
which we integrate to find

v × l = GM u + C , (97)

where C is a constant vector. Since this is a constant, we can find it’s value
by substituting the values of v and u at t = 0, i.e. v = v0 j and u = i and
we get
C = (r0 v02 − GM )i . (98)
We will now make use of this to determine the position as a function of θ.
Consider r · (v × l) and recall the identity u · (v × w) = (u × v) · w. Then

r · (v × l) = (r × v) · l = l · l = r02 v02 . (99)

Alternatively,
 r
+ ru · (r0 v02 − GM )i


r · (v × l) = r · (GM u + C) = r · GM
r (100)
= GM r + r(r0 v02 − GM ) cos θ .

Comparing these two equations, we see

r02 v02 = GM r + r(r0 v02 − GM ) cos θ , (101)

or
r02 v02
GM k
r=   = , (102)
r0 v02 1 + e cos θ
1+ GM
− 1 cos θ
where
r02 v02 r0 v02
k= , e= − 1. (103)
GM GM
30
 e>1
4

e=1
2

0<e<1

-8 -6 -4 -2

-2

-4

Figure 17: Planetary orbits as conic sections.

In other words, this is the equation of a conic section with k = de and k and
e given above. This can result in an ellipse (0 < e < 1), parabola (e = 1) or
hyperbola (e > 1). The conic that the orbit sketches is the result of the mass
of the body causing the gravitational force, as well as the initial position
and velocity. If e ≥ 1, the orbit is not closed, i.e. the bodies escapes the
gravitational pull. The condition e = 1 give the escape velocity,
r
2GM
vesc = . (104)
r0

3.1.11.2 Deriving Kepler’s Laws

For the first law, since we know that the planets don’t escape, the orbit
must be an ellipse by the previous discussion. The fact that the sun is at a
focus comes from equation (102). Since r is the distance to a focus and the
distance to the centre of the gravitational force, in the case our the planets
in our solar system, the Sun must be at the ellipse. To find the second law,
we equate the two expressions for l (86) and (94). This gives us

dθ
r2 = r0 v0 . (105)
dt

31
Since the curve should be described by a function of the angle, r = f (θ), and
hence form the formula for area in polar coordinates,
Z θ2
1
A= [f (θ)]2 dθ , (106)
θ1 2

we can find the rate of change of the area

dA dA dθ 1 dθ 1 dθ 1
= = [f (θ)]2 = r2 = r0 v0 , (107)
dt dθ dt 2 dt 2 dt 2
which is a constant. Since the area changes at a constant rate, the areas
traced out in equal times are equal.

For the third law, we begin with the area of an ellipse

πab , (108)

where a and b are the semi-major and semi-minor axes. Although we haven’t
actually covered this is is quite easy to prove (try!). In one period, the area
the radial line sweeps out is
Z T Z T
dA 1 1
dt = r0 v0 dt = r0 v0 T , (109)
0 dt 0 2 2
and therefore
1
r0 v0 T = πab . (110)
2
We can square this to obtain

4π 2 a2 b2
T2 = . (111)
r02 v02

We know, however, that c2 = a2 − b2 for an ellipse and also

√
c a2 − b 2
e= = , (112)
a a
which implies
b2 = a2 (1 − e2 ) . (113)

32
Moreover, we can see from equation (103) that

r02 v02
= k = a(1 − e2 ) , (114)
GM
and so
4π 2 a3 4π 2 a3 r02 v02 4π 2 3
T2 = k = = a , (115)
r02 v02 r02 v02 GM GM
and taking the square root gives the result,
2π 3/2
T =√ a . (116)
GM

33