The University of Sydney

School of Mathematics and Statistics

Solutions to Tutorial Week 2

MATH3968: Differential Geometry Semester 2, 2009
Lecturer: Emma Carberry

”Lecture notes” refers to Lecture Notes for Differential Geometry, MATH3968 by Nigel
O’Brien.
”do Carmo” refers to Differential Geometry of Curves and Surfaces, by Manfredo do
Carmo.
Solutions to exercises in the class notes are posted separately; below are solutions to the
remaining exercises.

Required Problems
1. do Carmo §1.5 Q12 and Lecture Notes Exercise Set 2, Q5
Solution:

a. By the chain rule, ds dt

dt ds
= 1. Since s = s(t) is arc length along α, ds
dt
= |α0 |.
Thus
dt 1
= 0 .
ds |α |
We need to differentiate this again with respect to s, which will involve differ-
entiating |α0 |. Recall that |α0 | = hα0 , α0 i1/2 . In addition, for any u, v : R → R3 ,

d d
hu, vi = (u1 v1 + u2 v2 + u3 v3 )
dt dt
= u01 v1 + u1 v10 + u02 v2 + u2 v20 + u03 v3 + u3 v30
= hu0 , vi + hu, v 0 i.

Thus we have
d 0 d
|α | = hα0 , α0 i1/2
dt dt
hα00 , α0 i + hα0 , α00 i
=
2hα0 , α0 i1/2
hα , α00 i
0
= .
|α0 |

1
 What we actually want is:
d2 t d 1
=
ds2 ds |α0 |
dt d 1
=
ds dt |α0 |
 
dt 1 d 0
= − 02 |α |
ds |α | dt
  0 00
1 1 hα , α i
= 0 − 02
|α | |α | |α0 |
hα0 , α00 i
=− .
|α0 |4

b. Recall the definition of curvature:

 2 
 d α(s) 
k(s) = 
 .
ds2 
One solution is given in example 2.5 of the class notes, but the following
solution enables one to see how the answer could have been arrived at without
being given the formula first. We have
d2 α
 
d dt dα
=
ds2 ds ds dt
d2 t dα dt d2 α dt
= 2 +
ds dt ds dt2 ds
0 00
hα , α i 0 α00
=− α + 0 2,
|α0 |4 |α |

Therefore,

 00 hα0 , α00 i 0 
 2   
d α 1
k =  2  = 0 2 α − α  .
ds |α |  |α0 |2
If the angle between the vectors α0 and α00 is θ, then recall that the projection
of α00 onto α0 is
1 0 hα0 , α00 i 0
projα0 α00 = (|α00 | cos θ) α = α.
|α0 | |α0 |2

α00

θ
α0
projα0 α00
hα0 ,α00 i 0
Therefore, the vector α00 − |α0 |2
α has length |α00 | sin θ = |α0 × α00 |/|α0 |. It
follows that
|α0 × α00 |
k= .
|α0 |3

2
 c. Recall the vector triple product, a scalar quantity depending on three vectors
(u, v and w) which gives the (oriented) volume of the parallelepiped spanned
by those vectors. The base has area |u × v| and the perpendicular height is
|w| cos θ (where θ is the angle between w and u×v), so the volume is hu×v, wi,
which is called the vector triple product of u, v and w. This value is cyclically
symmetric in u, v and w:
hu × v, wi = hv × w, ui = hw × u, vi.
See the solutions to question 5 of exercise sheet 2 in the class notes.
d. The absolute value of the signed curvature was found in part b:
|α0 × α00 | |(x0 y 00 − x00 y 0 )k| |x0 y 00 − x00 y 0 |
|k(t)| = = 3 = .
|α0 |3 ((x0 )2 + (y 0 )2 )3/2
p
(x0 )2 + (y 0 )2
Let θ be the angle traversed anticlockwise going from the vector α0 to α00 . We
know that α0 × α00 = (x0 y 00 − x00 y 0 )k points in the positive z-direction if and
only if sin θ > 0, i.e. θ is in the range (0, π)—in other words, if and only if
(α0 , α00 ) is a positively oriented basis of R2 . Thus x0 y 00 − x00 y 0 > 0 if and only
if (α0 , α00 ) is a positively oriented basis of R2 , which is true exactly when the
signed curvature k(t) is positive. Therefore
x0 y 00 − x00 y 0
k(t) = .
((x0 )2 + (y 0 )2 )3/2

2. Lecture Notes, Exercise Set 2, Q1bc, Q2, Q3

3. do Carmo §5.7 p404 Q1bc

Solution:
a. −3.
b. −4.
c. 1.
d. 0.

4. Let A be an invertible n × n matrix with real entries.

(a) Show that the following are equivalent:
(i) AAt = I, At A = I;
(ii) for all v, w ∈ Rn , hAv, Awi = hv, wi;
(iii) the columns of A form an orthonormal basis for Rn ;
(iv) the rows of A form an orthonormal basis for Rn .
Such matrices are called orthogonal, and we denote the set of all such matrices by
O(n).
(b) Show that O(n) is a subgroup of the group of invertible n × n matrices under
matrix multiplication, i.e. it contains the identity, the product of any two
orthogonal matrices is orthogonal, and the inverse of an orthogonal matrix
is orthogonal.

3
 (c) Show that an orthogonal matrix has determinant either 1 or −1.
An orthogonal matrix with determinant 1 is called special orthogonal; the set of
such matrices is denoted SO(n) and is also a group.
Solution:
(a) i =⇒ ii. For any v, w ∈ Rn , considering them as column vectors, we have
hAv, Awi = (Av)t Aw = v t At Aw = v t Iw = v t w = hv, wi.
ii =⇒ iii. Denote by ei the ith standard basis vector. As the standard basis
is orthonormal, hei , ej i = δij for all i, j. The ith column of A is Aei , and
by hypothesis hAei , Aej i = hei , ej i = δij for all i, j, so the columns of A
are orthonormal. Since A has n columns, they form an orthonormal basis
of Rn .
iii =⇒ i. The (i, j)th entry of At A is the inner product of the ith and jth
columns of A. Since the columns of A are orthonormal, this value is δij .
Hence At A = I and consequently AAt = I also.
i ⇐⇒ iv. We have the following chain of equivalences: AAt = I ⇐⇒
At (At )t = I ⇐⇒ the columns of At are an orthonormal basis of Rn ⇐⇒
the rows of A are an orthonormal basis of Rn . In the second equivalence
we used the fact that i ⇐⇒ iii, established above.
(b) Since II t = I, I ∈ O(n), so O(n) contains the identity. If A, B ∈ O(n) then
AAt = I and BB t = I, so AB(AB)t = ABB t At = AIAt = AAt = I, so
AB ∈ O(n). In other words, the product of any two orthogonal matrices
is orthogonal. For inverses, note that (At )−1 = (A−1 )t (since At (A−1 )t =
(A−1 A)t = I t = I). Thus if AAt = I then also A−1 (A−1 )t = A−1 (At )−1 =
(At A)−1 = I −1 = I, so A−1 ∈ O(n). Thus the inverse of an orthogonal matrix
is orthogonal. This establishes that O(n) is a subgroup of the group of n × n
invertible matrices.
(c) Suppose A is orthogonal. Then AAt = I. Taking determinants of both sides,
we get det(A) det(At ) = det(I) = 1. But det(At ) = det(A) by definition of
the determinant. Thus det(A)2 = 1, so det(A) = 1 or −1.

Recommended Problems
5. Lecture Notes, Exercise Set 1, Q6

6. do Carmo §1.5 p22 Q1cde

Solution: It was computed in lectures that s equals arc length, and the tangent,
normal, binormal and curvature are given by
1 s s 
t(s) = −a sin , a cos , b ,
c s
c c
s 
n(s) = − cos , − sin , 0 ,
c c
b s b s a
b(s) = t(s) × n(s) = sin i − cos j + k,
c c c c c
a
k = 2.
c

4
 c. The osculating plane is the plane through α(s) perpendicular to b(s) as given
above.
d. Note that these lines are in the direction n(s) and furthermore that hn(s), ki =
0 by the above. Therefore the lines are always perpendicular to the x-axis.
e. The tangent lines are in the direction t(s). Since t(s) is a unit vector, the
quantity t(s) · k is equal to the sine of the angle which the tangent line to α
at α(s) makes with the x-axis. But t(s) · k = b as above, which is a constant,
so the tangent lines to α make a constant angle with the x-axis.

7. do Carmo §5.7 p405 Q5

Solution:
a. Suppose that p0 q were tangent to C at q. By definition of a convex curve, all
points of C lie in a closed half-plane bounded by p0 q. The interior of C must
therefore lie in the interior of that half-plane, which does not contain p0 (as
p0 lies on the boundary of the half-plane). This contradicts the fact that p0
is in the interior of C.
b. Let α : [0, l] → R2 be a regular parameterisation of C. Let φ : [0, l] → R
and θ : [0, l] → R be continuous maps such that the vectors q(s) − p0 and
α0 (s) point in the directions at angles θ(s) and φ(s) respectively (measured
anticlockwise from i) for all s ∈ [0, l]. Note that q(s) − p0 and α0 (s) are
collinear for no s ∈ [0, l], by part a, so we know that θ(s) − φ(s) is never a
multiple of π.
By adding or subtracting multiples of 2π if necessary, we may insist that
θ(0) − φ(0) lie in the range (−π, π). Since θ(s) − φ(s) is never a multiple
of π, it follows that θ(s) − φ(s) ∈ (−π, π) for all s ∈ [0, l]. In particular,
θ(l) − φ(l) ∈ (−π, π).
However, θ(0) and θ(l) represent the same physical direction, so they must
differ by a multiple of 2π, and the same is true of φ(0) and φ(l). Therefore
θ(l) − φ(l) differs from θ(0) − φ(0) by a multiple of 2π. We conclude that
θ(l) − φ(l) = θ(0) − φ(0). Hence θ(l) − θ(0) = φ(l) − φ(0). Dividing by 2π,
we see that the winding number of C relative to p0 equals the rotation index
of C.
c. This question is false as stated—one must assume additionally that C is sim-
ple. (Without this assumption, we could simply trace out the same closed
curve repeatedly.) Assuming that C is simple, let α, p0 and θ be as in part
b. If ever θ(s) = θ(s0 ) + 2kπ, for some s 6= s0 ∈ [0, l) and k ∈ Z, then we
must have α(s) 6= α(s0 ) (as C is simple); hence one of α(s) and α(s0 ) will lie
outside the half-plane bounded by the tangent at the other, contradicting that
C is convex. Therefore θ([0, l)) lies within an interval of length at most 2π.
It follows that the winding number of C about p0 , which is (θ(l) − θ(0))/2π,
must be 0, 1 or −1.
Furthermore θ is monotonic, because it is injective on the interval [0, l) and
continuous. Therefore the winding number cannot be 0; it must be 1 or −1.
As noted in part b, the rotation index of C is equal to this winding number.