Math 442

Winter 2019
Solutions to Homework 2

1-5.1. a. dα/ds = (−(a/c) sin(s/c), (a/c) cos(s/c), b/c) and thus

r
0 a2 2 s a2 2
s b2
|α (s)| = sin + cos +
c2 c c2 c c2
r
a2 + b2
= = 1,
c2
since c2 = a2 + b2 . Therefore α is parametrized by arc length.
b. α00 (s) = (−(a/c2 ) cos(s/c), −(a/c2 ) sin(s/c), 0) so k(s) = |α00 (s)| = a/c2 . To
compute the torsion, start with

c2  a
 
a s a s b s a s 
b(s) = t(s) ∧ n(s) = − sin , cos , ∧ − 2 cos , − 2 sin , 0
c c c c c a c c c c
  
a s a s b s s 
= − sin , cos , ∧ − cos , − sin , 0
c c c c c c c
 
b s b s a
= sin , − cos , .
c c c c c

Then b0 (s) = ((b/c2 ) cos(s/c), b/c2 sin(s/c), 0) = −(b/c2 )n(s), so that τ (s) = −b/c2 .
c. The plane passing through α(s) = (a cos(s/c), a sin(s/c), bs/c) with normal
b(s) = ((b/c) sin(s/c), −(b/c) cos(s/c), a/c) is the osculating plane; it is given by
the set of (x, y, z) with
 
b s  s b s  s a bs
sin · x − a cos − cos · y − a sin + z− = 0.
c c c c c c c c

Simplify this expression to get

   
b s b s a bas
sin x− cos y+ z= 2 .
c c c c c c

d. Since n(s) = (− cos(s/c), − sin(s/c), 0), it follows that the line through α(s)
in the direction of n(s) is parallel to the xy-plane. It also intersects the z-axis at
the point (0, 0, bs/c) and must therefore do so at an angle of π/2.
e. e3 · (α0 (s)/|α0 (s)| = cos θ, where θ is the angle between α0 (s) and the z-axis.
Since
α0 (s) b
e3 · 0 = ,
|α (s)| c
it follows that this angle is constant.
Typeset by AMS-TEX
1
1-5.5. a. Without loss of generality, we may assume that α is parametrized by
arc length; we may also assume, by translating if necessary, that the fixed point
through which all the tangent lines pass is the origin. This implies that the line
through the origin and α(s) is the tangent line of α at s; hence

(1) α(s) = c(s)α0 (s)

for some scalar c(s). In fact, c(s) = α(s) · α0 (s), so that c is a differentiable function
of s. Now differentiate both sides of (1) and rearrange to get

c(s)α00 (s) = (1 − c0 (s))α0 (s).

But α00 (s) is perpendicular to α0 (s) (since α is parametrized by arc length); the
above formula then implies that c(s)α00 (s) = 0 for all s. I claim that in fact
α00 (s) = 0 for all s and hence, by a previous homework exercise (1-2.3), the trace
of α is a straight line (segment). Indeed, let s be in the domain of α. If c(s) 6= 0,
then α00 (s) = 0. If c(s) = 0, then α(s) = 0. But α0 (s) 6= 0, so by the definition of
the derivative, it follows that there exists an  > 0 such that α(t) 6= 0 whenever
|t − s| <  and t 6= s. This implies that c(t) 6= 0 and hence α00 (t) = 0 for all such t.
But α00 is continuous; therefore α00 (s) = 0 as well. This completes the proof.
b. If α is not regular, the conclusion of part a) need not hold. For example,
consider α : R → R2 given by

− 1

 e t2 (t, −t)

 t<0
α(t) = (0, 0) t=0

 − t12

e (t, t) t > 0.

1
(The factor e− t2 is introduced so that α is (infinitely) differentiable. Recall that
the function (
0 t=0
f (t) = − t12
e t 6= 0

is an infinitely differentiable function on R all of whose derivatives vanish at 0.)

However, the trace of α is the graph of the function y = |x| for |x| < 1, yet all of
the tangent lines of α pass through the origin.
1-5.6. a. Let ρ : R3 → R3 be an orthogonal transformation, and let u, v be elements
of R3 . We have
|u|2 = u · u = ρ(u) · ρ(u) = |ρ(u)|2 ,
so the norm of a vector is invariant under an orthogonal transformation. Similarly,
let θ0 be the angle between u and v, and let θ1 be the angle between ρ(u) and ρ(v),
0 ≤ θi ≤ π. Then

|u||v| cos θ0 = u · v = ρ(u) · ρ(v) = |ρ(u)||ρ(v)| cos θ1 .

Since |u| = |ρ(u)| and |v| = |ρ(v)|, we have that cos θ0 = cos θ1 (since we must have
u and v nonzero to talk about the angle between them) and hence θ0 = θ1 .
2
 b. Let ρ : R3 → R3 be an orthogonal transformation with positive determinant
and let u,v ∈ R3 . If u and v are collinear, then u ∧ v = 0 and hence ρ(u ∧ v) = 0.
On the other hand, ρ(u) and ρ(v) are also collinear, so ρ(u) ∧ ρ(v) = 0 = ρ(u ∧ v).
Now suppose that u and v are not collinear. Then ρ(u) and ρ(v) are not collinear
(why?), and ρ(u) ∧ ρ(v) is the unique vector of length |ρ(u)||ρ(v)| sin θ1 | perpen-
dicular to both ρ(u) and ρ(v) and such that (ρ(u), ρ(v), ρ(u) ∧ ρ(v)) is positively
oriented. Once again, θ is the angle between ρ(u) and ρ(v). We will show that
ρ(u ∧ v) has these same properties. Indeed, u ∧ v is perpendicular to both u and
v; hence ρ(u ∧ v) is perpendicular to ρ(u) and ρ(v). Since (u, v, u ∧ v) is positively
oriented and ρ has positive determinant, so is (ρ(u), ρ(v), ρ(u ∧ v)). Moreover, by
part a,
|ρ(u ∧ v)| = |u ∧ v| = |u||v|| sin θ0 | = |ρ(u)||ρ(v)|| sin θ1 |;
therefore ρ(u ∧ v) = ρ(u) ∧ ρ(v).
This proof also shows that if ρ is orthogonal with negative determinant, then
ρ(u) ∧ ρ(v) = −ρ(u ∧ v).
c. It is trivial to check that the arc length, curvature, and torsion are invariant
under a translation.
Now let ρ : R3 → R3 be an orthogonal transformation with positive determinant,
and let α : I → R3 be a parametrized curve. Consider also the curve ρ ◦ α : I → R3 .
Since ρ is a linear transformation, it follows that

(1) (ρ ◦ α)0 (t) = ρ(α0 (t)),

so

(2) |(ρ ◦ α0 (t)| = |ρ(α0 (t))| = |α0 (t)|

by the orthogonality of ρ. It is clear from this that the arc length of ρ ◦ α from t0
to t is the same as the arc length of α from t0 to t. Moreover, if α is parametrized
by arc length, so is ρ ◦ α. In such a case, k(s), the curvature of α, is given by

k(s) = |α00 (s)|,

and kρ (s), the curvature of ρ ◦ α, is given by

kρ (s) = |(ρ ◦ α)00 (s)|.

But (ρ ◦ α)0 = ρ ◦ α0 , so the same arguments as in (1) and (2) show that

(ρ ◦ α)00 (s) = ρ(α00 (s))

and
|(ρ ◦ α)00 (s)| = |ρ(α00 (s))| = |α00 (s)|.
Thus k(s) = kρ (s).
Finally, let t(s), n(s), b(s), and τ (s) be defined for α as in do Carmo; and let
tρ (s), nρ (s), bρ (s), and τρ (s) be the corresponding functions for ρ ◦ α. We have
already shown that

tρ (s) = ρ(t(s))
nρ (s) = ρ(n(s));
3
hence, by part a),

bρ (s) = tρ (s) ∧ nρ (s) = ρ(t(s)) ∧ ρ(n(s)) = ρ(t(s) ∧ n(s)) = ρ(b(s)).

It then follows as in (1) that

b0ρ (s) = ρ(b0 (s)),
so that
b0ρ (s) = ρ(τ (s)n(s)) = τ (s)ρ(n(s)) = τ (s)nρ (s)
and therefore τρ (s) = τ (s).
1-5.9. Let α(s) be as stated in the problem. Then α0 (s) = (cos θ(s), sin θ(s)), so α
is parametrized by arc length. Moreover,

α00 (s) = θ0 (s)(− sin θ(s), cos θ(s)) = k(s)(− sin θ(s), cos θ(s)),

and (cos θ(s), sin θ(s)), (− sin θ(s), cos θ(s)) is positively oriented. Thus the curva-
ture of α at s is k(s).
Now suppose β : I → R2 is a curve parametrized by arc length whose curvature
is the given function k(s). Since |β 0 (s)| = 1, we may follow the hint to write

β 0 (s) = (cos ψ(s), sin ψ(s))

for some differentiable function ψ : I → R. Then

β 00 (s) = ψ 0 (s)(− sin ψ(s), cos ψ(s)),

and since (cos ψ(s), sin ψ(s)), (− sin ψ(s), cos ψ(s)) is positively oriented, it follows
that ψ 0 (s) = k(s). Thus Z
ψ(s) = k(s) ds + ω

for some constant ω. Moreover, β is an antiderivative of β 0 , so

Z Z 
β(s) = cos ψ(s) ds + u, sin ψ(s) ds + v

for some constants u and v.

If u and v are changed, the curve β is translated, and if ω is changed, β 0 is
rotated. This implies that its antiderivative is rotated (and perhaps translated).
Rt
1-5.12. Define φ(t) = t0 |α0 (u)| du and β = α ◦ φ−1 .
a. This problem asks you to compute d(φ−1 )/ds and d2 (φ−1 )/ds2 . First of all,

dφ−1 1 1
= 0 −1 = 0 −1 .
ds φ (φ (s)) |α (φ (s))|
1
Now, since |α0 (t)| = (α0 (t) · α0 (t)) 2 , we have

d α0 (t) · α00 (t)

(|α0 (t)|) = ,
dt |α0 (t)|
4
so
α0 (t) · α00 (t)
 
d 1
0
=− .
dt |α (t)| |α0 (t)|3
Therefore,

d2 φ−1 α0 (φ−1 (s)) · α00 (φ−1 (s)) dφ−1 α0 (φ−1 (s)) · α00 (φ−1 (s))
= − · = − .
ds2 |α0 (φ−1 (s))|3 ds |α0 (φ−1 (s))|4

b. We need to compute the curvature of β at s, where t = φ−1 (s). Now

dφ−1 α0 (t)
β 0 (s) = α0 (φ−1 (s)) · = 0 .
ds |α (t)|

Hence
 0 
00 d α (t) dt
β (s) =
dt |α0 (t)| ds
 00
(α0 (t) · α00 (t))α0 (t)

α (t) 1
= − · 0
|α0 (t)| |α0 (t)|3 |α (t)|
(α0 (t) · α0 (t))α00 (t) − (α0 (t) · α00 (t))α0 (t)
= .
|α0 (t)|4

In this expression, the dot product of the numerator with itself is given by

|α00 (t)|2 |α0 (t)|4 − 2(α0 (t) · α00 (t))2 |α0 (t)|2 + |α0 (t)|2 (α0 (t) · α00 (t))2 ,

which simplifies to

|α0 (t)|2 [|α00 (t)|2 |α0 (t)|2 − (α0 (t) · α00 (t))2 ].

But
|α00 (t)|2 |α0 (t)|2 − (α0 (t) · α00 (t))2 = |α0 (t) ∧ α00 (t)|2 ,
where we have used the formula
 
u · u u · v 
(u ∧ v) · (u ∧ v) =  .
u·v v·v

Thus
|α0 (t)|2 |α0 (t) ∧ α00 (t)|2
|β 00 (s)|2 = ,
|α0 (t)|8
so
|α0 (t) ∧ α00 (t)|
|β 00 (s)| = .
|α0 (t)|3
This is the curvature of β at s.
c. Let bβ (s) be the binormal vector of β at s, where t = φ−1 (s). Then

β 00 (s)
bβ (s) = β 0 (s) ∧
|β 00 (s)|
α0 (t) ∧ α00 (t)
= 0 .
|α (t) ∧ α00 (t)|
5
Now write  
d 1 g(t)
=
dt |α (t) ∧ α00 (t)|
0 |α0 (t) ∧ α00 (t)|3
with
g(t) = −(α0 (t) ∧ α00 (t)) · (α0 (t) ∧ α(3) (t)).
Then
α0 (t) ∧ α(3) (t) g(t)(α0 (t) ∧ α00 (t)) dt
 
b0β (s) = +
|α0 (t) ∧ α00 (t)| |α0 (t) ∧ α00 (t)|3 ds
 0 (3) 0 00

α (t) ∧ α (t) g(t)(α (t) ∧ α (t)) 1
= 0 00
+ 0 00 3 0
.
|α (t) ∧ α (t)| |α (t) ∧ α (t)| |α (t)|

Let τβ (s) be the torsion of β at s. Then

β 00 (s)
τβ (s) = b0β (s) ·
|β 00 (s)|
 0
(α (t) ∧ α(3) (t)) · α00 (t)

1
= 00 ,
|β (s)||α0 (t)| |α0 (t)|2 |α0 (t) ∧ α00 (t)|

since α0 (t) ∧ α(3) (t) is orthogonal to α0 (t), and α0 (t) ∧ α00 (t) is orthogonal to both
α0 (t) and α00 (t). Now

(α0 (t) ∧ α(3) (t)) · α00 (t) = −(α0 (t) ∧ α00 (t)) · α(3) (t)

(why?); we therefore obtain

(α0 (t) ∧ α00 (t)) · α(3) (t)

τβ (s) = − .
|α0 (t) ∧ α00 (t)|2

d. In terms of x(t) and y(t), we have that

(x0 (t), y 0 (t))

β 0 (s) = 1 ,
(x0 (t)2 + y 0 (t)2 ) 2

so that the normal vector is given by

1
n(s) = (x0 (t)2 + y 0 (t)2 )− 2 (−y 0 (t), x0 (t)).

But, from part b),

(x00 (x0 )2 + x00 (y 0 )2 − (x0 )2 x00 − x0 y 0 y 00 , y 00 (x0 )2 + y 00 (y 0 )2 − y 0 x0 x00 − (y 0 )2 y 00 )

β 00 (s) =
[(x0 )2 + (y 0 )2 ]2
00 0 2 0 0 00 00 0 2 0 0 00
(x (y ) − x y y , y (x ) − y x x )
=
[(x0 )2 + (y 0 )2 ]2
3
= [(x0 )2 + (y 0 )2 ]− 2 (x0 y 00 − x00 y 0 )n(s).

This proves that the signed curvature of β at s is

3
(x0 (t)2 + y 0 (t)2 )− 2 (x0 (t)y 00 (t) − x00 (t)y 0 (t)).
6
1-5.14. We may assume that α is parametrized by arc length. At t0 , the function
f (t) = α(t) · α(t) attains a maximum. Therefore, f 0 (t0 ) = 0 and f 00 (t0 ) ≤ 0. Now

f 0 (t) = 2α0 (t) · α(t)

and

f 00 (t) = 2α00 (t) · α(t) + 2α0 (t) · α0 (t)

= 2[(α00 (t) · α(t)) + 1],

since α is parametrized by arc length. We then have

α00 (t0 ) · α(t0 ) + 1 ≤ 0,

so
|α00 (t0 ) · α(t0 )| ≥ 1.
But, by the Cauchy-Schwarz inequality,

|α00 (t0 ) · α(t0 )| ≤ |α00 (t0 )||α(t0 )| = |k(t0 )||α(t0 )|,

so
1
|k(t0 )| ≥ .
|α(t0 )|