3 Vector functions of a single scalar variable.

Applications to
basic differential geometry of curves. (Reading: Anton §13)

In §2 we discussed the parametric form of a line in 3-D. In this section we extend the discussion
to more general parametric curves.

3.1 Examples of parametric curves in 3-D (Anton §13.1)

0 r(t)h P

r(t1 )

Let P be a point whose position vector r varies with “time” t, so that,

r = r(t) = x(t)i + y(t)j + z(t)k.

The components x(t), y(t), z(t) are real-valued (scalar) functions of the real variable t. So, a
vector-valued function r(t) may be interpreted as a triple of real-valued functions. As t varies,
the point P traces out a curve C.
Example 1: If a and l are constant vectors then r(t) = a + tl represents the equation of a
straight line, L say.
As t varies from −∞ to +∞, P traces out the straight line in direction of l, through A.
Example 2: (Circular helix)

z
C

a y

33
 Let

r(t) = (a cos t)i + (a sin t)j + c t k (a, c positive constants )

⇔ x(t) = a cos t, y(t) = a sin t, z(t) = ct.

Note that then x2 + y 2 = a2 , and therefore the curve C lies on this cylinder x2 + y 2 = a2 but
increases in z with t. The resulting curve C is called a circular helix. 2

There is a nice introduction in Anton §13.1 (p. 863–869).

3.2 Calculus of vector-valued functions (Anton §13.2 p. 869)

We are going to define differentiation of vector functions in the obvious way, but first we need
to know what a limit is.

Definition 3.1 (Limit) r(t) → r0 as t → t0 iff

kr(t) − r0 k → 0 as t → t0 .

(Recall that the latter means that for all positive ε there exists δ > 0 such that kr(t) − r0 k < ε
as long as |t − t0 | < δ.)
If we write r(t) = (x(t), y(t), z(t)) and r0 = (x0 , y 0 , z 0 ) then obviously r(t) → r0 as t → t0
iff
x(t) → x0 , y(t) → y 0 , z(t) → z 0 .

Here we use the fact that a vector valued function is a triple of real valued scalar functions, and
we know how to take the limits of real functions.

Definition 3.2 (Continuity) We say that r(t) is continuous if for all t0 , r(t) → r(t0 ) as
t → t0 . Obviously, r(t) is continuous if and only if all its components x(t), y(t) and z(t) are
continuous.

Definition 3.3 (Derivative) Given r(t) the derivative r′ (t) is

 
dr ′ r(t + h) − r(t)
= r (t) := lim , (3.1)
dt h→0 h

provided this limit exists.

34
It follows that, provided the scalar derivatives x′ (t), y ′ (t) and z ′ (t) all exist,

r′ (t) = x′ (t)i + y ′ (t)j + z ′ (t)k, (3.2)

so we differentiate the components. From now on, we assume that all r(t) we consider are “good
enough” for the derivatives to exist. We then say r(t) is differentiable.
dr d  
Notation: r′ (t), r′ , , r(t) are equivalent.
dt dt
Geometrical interpretation of derivative:
(See a diagram in the lectures:) As t varies r traces out the curve C. Now if P and Q are
the end points of the position vectors r(t) and r(t + h) respectively,
−−→
r(t + h) − r(t) PQ
=
h h
r(t + h) − r(t)
and so, if h → 0, Q → P , and tends to a vector in the direction of the tangent
h
to C at P . Hence,

Definition 3.4 Assuming r′ (t) 6= 0, r′ (t) is called a tangent vector to the curve C at point
P . Also the tangent points in the direction of increasing t.

Example: For helix r(t) = a cos t i + a sin t j + c t k the tangent = r′ (t), is

r′ (t) = − a sin t i + a cos t j + c k. 2

Rules of differentiation
Let r(t), r1 (t), r2 (t) be differentiable vector-valued functions, λ(t) a differentiable real-valued
function, and c a constant vector.
Then:

d
1. (c) = 0
dt
d  d  d 
2. r1 (t) + r2 (t) = r1 (t) + r2 (t)
dt dt dt
d   d   d 
3. λ(t)r(t) = λ(t) r(t) + λ(t) r(t)
dt dt dt
d  dr1 dr2
4. For the dot product: r1 . r2 = . r2 + r1 .
dt dt dt
d  dr1 dr2
5. For the vector product, r1 ∧ r2 = ∧ r2 + r1 ∧ (N.B. Keep correct ordering)
dt dt dt

35
Thus the usual rules of differentiation apply. Proofs of (1) - (5) directly follow from definition
(3.1), or component form (3.2).
(See Examples 1 and 2 in the lectures.)
For a position vector r(t), kr(t)k is a scalar function of t.
Example 1: Show that for r 6= 0
d  1
krk = r . r′
dt krk
Deduce that r.r′ = 0 if krk = constant.
Solution: (i) Recall, krk2 = r.r. Differentiate both sides with respect to t
d  2 d
krk = (r . r)
dt dt
d  dr dr dr
∴ 2krk krk = .r + r = 2 r.
dt dt dt dt
d   1
∴ krk = r.r′ . (3.3)
dt krk
We use this result often.
d 
(ii) If krk = constant ∴ krk = 0
dt
d 
∴ r.r′ = krk krk = 0 2
|dt {z }
=0

Integration By analogy with differentiation, we have

 b   b   b 
Zb Z Z Z
r(t)dt :=  x(t)dt i +  y(t)dt j +  z(t)dt k.
a a a a

Also, the usual integration rules apply.

Anti-derivatives
Z Z 
′ d
r (t)dt = r(t) + c, r(t)dt = r(t).
dt
(See Examples 1-3 in the lectures.)
Example 4: If r : R → R3 is a twice differentiable function such that r′′ (t) ≡ 0, show that r(t)
represents a straight line.
Solution: In this case,
d ′

r′′ (t) ≡ 0 =⇒ r (t) = 0
dt
=⇒ r′ (t) = c,

36
where c is a constant vector (analogous to the constant of integration). Therefore,
Z
r′ (t)dt = ct + d,

where d is a constant vector, and thus

r(t) = tc + d, ∀t ∈ R,

which is the equation of a straight line through arbitrary point d in an arbitrary direction c, as
required. 2

3.3 Change of parameter, arc length (Reading: Anton §13.3)

We shall assume that the tangent r′ (t) is continuous and r′ 6= 0.

The arc length of a section of a curve is defined as a limit of lengths of piecewise straight
lines (see Anton §6.4 for 2-D case). Selecting a point P0 on a curve C corresponding to t = t0 ,
we introduce the arc length, s say, from P as a signed length: s is positive for t > t0 and negative
for t < t0 .
It is often convenient to use the arc length s as a parameter to describe C (instead of t).
Change of variable: (Chain Rule)
If r = r(t) and t = t(s) then
dr dr dt
= (3.4)
ds dt ds
(Prove using components).
Example: For a circle of radius a in the xy-plane, r = (x, y, 0), where x(t) = a cos t, y(t) =
a sin t, and for the arc length s(t) = at. Hence
 s s
dr dr dt dr ds 1
= = = r′ = − i sin + j cos 2
ds dt ds dt dt a a a

To compute Arc length

Let r(t) be a position vector of P on C, a smooth curve in 3-D, (so can be described by r(t)
such that r′ exists and r′ 6= 0).
Let P0 ∈ C be a given reference point and let s(t) be the arc length along C from P0 to point
P having position vector r(t). So s(t0 ) = 0; s(t) is length from P0 to P and s(t + h) is length
from P0 to Q. Then, for small h,
−−→
s(t + h) − s(t) kP Qk kr(t + h) − r(t)k r(t + h) − r(t)
≃ = = .
h h h h

37
 C

Q
t+h

t1 P

t0 P0

Now taking the limit as h → 0,

ds dr
= = kr′ k (3.5)
dt dt

Integrating (3.5) from t0 to t

s
Zt Zt 2  2  2
′ dx dy dz
s(t) = r (τ ) dτ = + + dτ
dτ dτ dτ
t0 t0

(Here τ is a dummy variable of integration which, strictly speaking, could be anything but t
which is the upper limit; we nevertheless often use in similar context t also as the integration
variable, for notational simplicity.)
See an example in lectures on the arc length for a circular helix; see also the example in the
end of the next subsection.
Warning! In general
dr d 
6= kr(t)k .
dt dt
(The length of a derivative is generally different from the derivative of a length.)

3.4 Unit tangent and principal unit normal vectors

We have seen that r′ (t) is a vector tangent to C at r(t), assuming r′ (t) 6= 0. We apply the usual
recipe to construct a vector of unit length pointing in the tangent direction:

Definition 3.5 With r′ (t) 6= 0,

r′ (t)
T̂(t) := (3.6)
kr′ (t)k

is called the unit vector tangent to C at t (or unit tangent vector).

38
If we choose the arc length s as the parameter for the curve C, then using the Chain Rule and
(3.6)
dr dr ds dr ′
r′ = = = r ∴ from (3.6)
dt ds dt ds

dr
T̂ = (3.7)
ds
 dr 
Note that = 1.
ds

3.4.1 Unit normal vector and curvature

We have
kT̂k = 1 ∴ kT̂k2 = T̂ · T̂ ≡ 1
dT̂
and differentiating the latter, T̂ · =0
dt
dT̂
Thus is orthogonal to the tangent vector T̂ and so is normal to the curve C (see the
dt
diagram in lectures).

dT̂
Definition 3.6 Assume T̂′ := 6= 0, then
dt

T̂′
N̂ := (3.8)
kT̂′ k

is called the principal unit normal.

(The word “principal” means that this is “the” unit normal to C pointing in the direction in
which the curve actually curves, or in other words lying in the plane “best approximating”
the curve C near the considered point P . More precise statements would require use of more
advanced Analysis.)
ds
In the s-parametrisation, replacing in (3.8) t by s (and recalling that dt > 0)

dT̂
N̂ = ds ,
dT̂
ds

or, re-arranging,
dT̂ dT̂
= κ N̂, κ := . (3.9)
ds ds

39
 dT̂
Definition 3.7 The length of vector is called the curvature of curve C at the given point
ds
t.

By the chain rule and using also (3.5),


dT̂ dT̂ ds kT̂′ (t)k
κ = = = (3.10)
ds dt dt kr′ (t)k

Example: [cf.Qn 5 Sheet 6] Consider again the circular helix. Then

r(t) = (a cos t)i + (a sin t)j + (ct)k (a, c positive constants )

r′ (t) = −a sin t i + a cos t j + ck

 1 p
2
kr′ k = a2 (sin2 t + cos2 t) + c2 = a2 + c2

So
ds p
= kr′ k = a2 + c2
dt
Take our start point as t = 0 i.e. r(0) = (a, 0, 0).
Rt √
(1) The arc length from (a, 0, 0) to r(t) is s(t) = kr′ kdt = a2 + c2 t.
0

(2) The unit tangent vector:

r′ −(a sin t)i + (a cos t)j + ck
T̂ = = √ .
kr′ k a2 + c2
(3) To find the principal unit normal, now
dT̂ −(a cos t)i − (a sin t)j
T̂′ = = √
dt a2 + c2
a T̂′
∴ T̂′ = √ ∴ N̂ = = − cos t i − sin t j.
a + c2
2
T̂′

(4) Hence, from (3.9)-(3.10), for the curvature,

d
T̂
dt a 1 a
κ = = √ √ = 2 ,
kr′ k a2 + c2 a2 + c2 a + c2
with
 √ 1 (−a cos ti − a sin tj)
dT̂ dT ds a2 +c2 −(a cos t)i − (a sin t)j
= = √ = = κ N̂.
ds dt dt 2
a +c 2 a2 + c2
1
Note that, in particular, for c = 0 (so that C is a circle) κ = , the inverse radius. 2
a
More examples are found in Sheet 6 (Qns 4 and 5).

40