MATH3968 Lecture 2: DR Emma Carberry 29 July 2009
MATH3968 Lecture 2: DR Emma Carberry 29 July 2009
MATH3968 Lecture 2: DR Emma Carberry 29 July 2009
Dr Emma Carberry
29 July 2009
Parameterised Curves
α = β ◦ φ.
Example 3.
α : R → R3 , α(t) = (2 cos(t), 3 sin(t), t)
α : R → R3 , α(t) = (2 cos(2t), 3 sin(2t), 2t)
α : R → R2 , α(t) = (t2 , t3 )
Definition 4. α0 (t) is called the velocity vector of α at t. If α0 (t) 6= 0, then there is a unique line in Rn
that contains the point α(t) and is parallel to α0 (t); we call this the tangent line of α at t.
On Rn assume given the standard inner product, for which there are two standard notations:
v · w = hv, wi = v1 w1 + · · · vn wn
1
Definition 5. The arc-length of a parameterised smooth curve α : (a, b) → Rn after time t ∈ (a, b) is
Z t
s(t) = |α0 (t)|dt;
a
Active Learning
Question 6. Can every parameterised smooth curve be reparameterised by arc-length?
i.e. Given smooth α : (a, b) → Rn , is t 7→ s(t) smooth with a smooth inverse?
Answer 7. The map t 7→ s(t) is always smooth, but it does not always have a smooth inverse. Some
counterexamples:
1.
α : (0, 1) → R3 , α(t) = (1, 2, 3),
has s(t) ≡ 0.
2.
α : R → R2 , α(t) = (t2 , t3 )
ds dt
If t 7→ s(t) has a smooth inverse, then 1 = , so s0 (t) 6= 0 for all t. But s0 (0) = 0.
dt ds
3.
α : (0, π) → R2 α(t) = (0, sin t)
has s0 π
2 = 0.
This last curve has a trace which could be smoothly parameterised by arc length, whereas the first two
do not.
Definition 8. A parameterised smooth curve α : I → Rn is regular if α0 (t) 6= 0 for all t ∈ I.
We have just seen that being regular is necessary in order that α may be reparameterised by arc-length.
Is it sufficient?
Theorem 9 (Inverse Function Theorem). • Let W ⊂ Rn be an open set, and
W → Rn
x = (x1 , . . . , xn ) 7→ (f 1 (x), . . . , f n (x))
∂f i
is invertible, where fji = ∂xj .
2
Example 10.
α: R → R2
t 7→ (a cos t, b sin t), a, b > 0
Finding the arc length of an ellipse can be done using elliptic functions, which you may learn about in a
Riemann surfaces class.
• Write
α00 (s) =: k(s)n(s),
where
α00 (s)
k(s) = |α00 (s)|, n(s) =
|α00 (s)|
• n(s) is defined only if k(s) 6= 0; we shall restrict ourselves to this (generic) situation
• Then k(s) is the magnitude of the rate of change of the unit tangent vector t at s, and we call it
the curvature of α at s.
• The function k is called the curvature of α.
Since α0 (s) has constant length, it and its derivative are orthogonal.
α00 (s)
α0 (s)
t · t = constant
t · t + t · t0 = 0
0
t0 · t = 0
1
t(u) ≡ p (a1 , . . . , an ), k ≡ 0, n not defined
a21 + · · · + a2n
3
Example 13.
α(u) = (r cos(u), r sin(u))
α
hence
dt dt du
=
ds du ds
1
= (− cos(u), − sin(u)),
r
so
1
n(u) = −(cos(u), sin(u)), k(u) = .
r
Helix
Circle
Question 14.
How should the curvatures of a circle and a helix with the same radius compare?
Answer 15. The curvature of the helix should be less. The fact that it is stretched vertically means
that there is less turning per unit length.
A parameterisation by arc-length is
s s as
α(s) = r cos √ , r sin √ ,√
r2 + a2 r2 + a2 r2 + a2
and
1 s s
t(s) = √ −r sin √ , r cos √ ,a ,
r2 + a2 r2 + a2 r2 + a2
s s
n(s) = − cos √ , sin √ ,0 ,
2
r +a 2 r + a2
2
r
k(s) = 2
r + a2
4
n
Answer 15 (continued).
(t, n0 ) = A(e1 , e2 ),
n0 n0
e2 1
curvature = radius
e1