A First Course in Differential Geometry Surfaces in Euclidean Space Solutions Manual - Woodward Problems

1
A FIRST COURSE IN DIFFERENTIAL GEOMETRY

Woodward and Bolton
Solutions to exercises
Chapter 1
1.1. A sketch of the astroid is given in Figure 1(a). It is clear that all
points in the image of α satisfy the equation of the astroid. Conversely, if
x2/3 + y 2/3 = 1, then there exists u ∈ R such that (x1/3 , y 1/3 ) = (cos u, sin u).
Thus every point of the astroid is in the image of α.
Trigonometric identities may be used to show that α0 = (3/2) sin 2u(− cos u, sin u),
which is zero only when u is an integer multiple of π/2. The corresponding points
of the astroid are the cusps in Figure 1(a).
The required length is
3 π/2
Z
3
sin 2u du = .
2 0 2
(a) Astroid (b) An epicycloid
Figure 1
1.2. A sketch of the trace of an epicycloid is given in Figure 1(b). Trigono-

metric identities may be used to show that
α0 = 4r sin(u/2) sin(3u/2), cos(3u/2) .

R 2π
So, for 0 ≤ u ≤ 2π, |α0 | = 4r sin(u/2), and required length is 4r 0
sin(u/2)du =
16r.
1.3. When r = 1, a calculation shows that α0 = tanh u sech u(sinh u, −1),

so that, for u ≥ 0, t = sech u(sinh u, −1). It follows that α + t = (u, 0). A
sketch of the trace of a tractrix is given in Figure 2(a).
2 2
1.4. Here, |α0 | = (1 + g 0 )1/2 and t = (1 + g 0 )−1/2 (1, g 0 ). Hence n =
2 2
(1 + g 0 )−1/2 (−g 0 , 1). A calculation shows that t0 = g 00 (1 + g 0 )−3/2 (−g 0 , 1), so
that
dt 1 g 00
= 0 t0 = n.
ds |α | (1 + g 0 2 )3/2
2 SOLUTIONS TO EXERCISES
Figure 2: (a) shows a tractrix and (b) shows three catenaries
2
Hence κ = g 00 (1 + g 0 )−3/2 .
Taking x(u) = u, y(u) = g(u) in the formula given in Exercise 1.8 gives the
same formula for κ.
1.5. Use the method of Example 2 of §1.3. For u ≥ 0, |α0 | = tanh u and
t = (tanh u, −sech u). It follows that dt/ds = (|α0 |)−1 t0 = n/ sinh u. Hence
κ = cosech u.
1.6. EITHER: use Exercise 1.4 to show that the curvature of the catenary
α(u) = (u, cosh u) is given by κ = sech2 u,
OR: use the method of Example 2 of §1.3, and proceed as follows:-
α0 = (1, sinh u), so that |α0 | = cosh u and t = (sech u, tanh u). Hence n =
(−tanh u, sech u), and
dt 1 1 1
= 0 t0 = 2 (−tanh u, sech u) = n.
ds |α | cosh u cosh2 u
Hence κ = sech2 u. A sketch of the traces of three catenaries is given in Figure

2(b).
1.7. Differentiating with respect to u, we see that, using Serret-Frenet,
α` 0 = α0 + `n0 = |α0 |(t − κ`t) = |α0 |(1 − κ`)t .
It follows that |α` 0 | = |α0 | |1 − κ`| and t` = t, where = (1 − κ`)/|1 − κ`|.
Hence n` = n, so, if s` denotes arc length along α` , we have
dt` 1
= 0 t0 = 0 t0 .
ds` |α | |1 − κ`| ` |α | |1 − κ`|
Using Serret-Frenet, t0 = |α0 |κn = |α0 |κn` , from which the result follows.
2 2
1.8. Since α0 = (x0 , y 0 ), we have that |α0 | = (x0 + y 0 )1/2 . Hence t =
2 2 2 2
(x , y 0 )/(x0 + y 0 )1/2 and n = (−y 0 , x0 )/(x0 + y 0 )1/2 . Hence
0
α00 α0 (x0 x00 + y 0 y 00 )

t0 = 2 1/2 − ,
(x0 2 0
+y ) (x0 2 + y 0 2 )3/2
A FIRST COURSE IN DIFFERENTIAL GEOMETRY 3
and a short calculation shows that

dt 1
0 00 0 0 00 0 0 00 00 0

= 02 2 y (x y − x y ), x (x y − x y )
ds (x + y 0 )2
x0 y 00 − x00 y 0
= 02 n,
(x + y 0 2 )3/2
and the result follows.
1.9. (i) Let sα be arc length along α measured from u = 0. Since α0 =

(1, sinh u) we see that dsα /du = |α0 | = cosh u. Hence sα (u) = sinh u and
tα = (sech u, tanh u). The result follows from formula (1.9) for the involute.
(ii) The evolute of α is given by
1
β =α+ nα .
κα
Here, we have (from Exercise 1.6) that κα = sech2 u and nα = (−tanh u, sech u).
A direct substitution gives the result.
A short calculation shows that β 0 = 0 if and only if u = 0, so this gives the
only singular point of β (where the curve β has a cusp). A sketch of the traces
of α and β is given in Figure 3.
Figure 3: A catenary and its evolute
1.10. Let sα denote arc length along α starting at u = u0 . Then, using

(1.9) and the notation used there, we see that
β 0 = α0 − sα 0 tα − sα tα 0 = −sα tα 0 .
It follows that β 0 = −sα |α0 |κα nα , so the only singular point of β is when
sα = 0, that is at u = u0 .
1.11. For ease, assume that κα > 0, and restrict attention to u0 < u1 < u.
Then, from (1.12), we have that κ0 = 1/s0 and κ1 = 1/s1 .
Let ` be the length of α measured from α(u0 ) to α(u1 ). Then ` = s0 −s1 > 0,
so the definition of involute gives that β 1 = β 0 + (s0 − s1 )tα = β 0 + `n0 . Hence
β 1 is a parallel curve to β 0 , and
κ0 1/s0 1 1
= = = = κ1 .
|1 − κ0 `| |1 − `/s0 | |s0 − `| s1
1.12. A sketch of the trace of α is given in Figure 4.

Figure 4: The curve in Exercise 1.12
√
A short calculation shows that |α0 | = 3eu and
1
t = √ (cos u − sin u, sin u + cos u, 1) .
3
When z = λ0 we have that u = log λ0 , and when z = λ1 we have that
u = log λ1 . So, required length is
√ Z log λ1 u √
3 e du = 3(λ1 − λ0 ) .
log λ0
Also, using the method of Example 2 of §1.5, we find that

dt 1
= u (− sin u − cos u, cos u − sin u, 0) ,
ds 3e
√
so that κ = 2/(3eu ) and
1
n = √ (− sin u − cos u, cos u − sin u, 0) .
2
It then follows that

1
b = t × n = √ (sin u − cos u, − cos u − sin u, 2) ,
6
so that
db 1
= √ (sin u + cos u, − cos u + sin u, 0) .
ds 3 2eu
Hence τ = −1/(3eu ).
0
√ 1.13. Calculations similar to√those of Example2 2 of §1.5 show that |α | =
2 cosh u, t = (tanh u, 1, sech u)/
√ 2, κ = (1/2)sech u, n = (sech u, 0, −tanh u),
and b = (−tanh u, 1, −sech u)/ 2. Differentiating one more time, we find that
db/ds = (1/2)sech2 u(−sech u, 0, tanh u), so that τ = −(1/2)sech2 u.
1.14. Assume that α(s) is a smooth curve in R3 parametrised by arc length.

If α has zero curvature then dt/ds = 0, so that t is a constant unit vector t0 ,
say. Since dα/ds = t0 , it follows that α(s) = st0 + v 0 , for some constant vector
v 0 . Hence α is a line. Conversely, if α is the line through v 0 in direction of unit
vector t0 , say, then α may be parametrised as α(s) = st0 + v 0 , and it quickly
follows that κ = 0.
1.15. Using Serret-Frenet, we have
α0 = |α0 |t , α00 = κ|α0 |2 n + |α0 |0 t ,
so that α0 × α00 = κ|α0 |3 b. Equating the lengths of both sides, we obtain the
required formula for κ. Differentiating one more time, we find that
α000 = −κτ |α0 |3 b + terms involving t and n ,
so that (α0 × α00 ).α000 = −κ2 τ |α0 |6 . Using the expression we have just found for
κ now gives the required expression for τ .
1.16. (i) Assume first that α(u) = (a cos u, a sin u, bu). The tangent√ vector
to α(u) makes angle φ with the generating lines where cos φ = b/ a2 + b2 .
Hence φ is a constant different from 0 and π/2. We saw in Example 2 of §1.5
that α(u) has non-zero torsion, so it follows that α(u) is a helix with value
(a, 0, 0) when u = 0.
Conversely, assume that α(v) = (a cos θ(v), a sin θ(v), v + c) describes a helix
with value (a, 0, 0) when v = 0. Then c = 0 and, denoting differentiation with
respect to v by 0 , we have
α0 = (−aθ0 sin θ, aθ0 cos θ, 1) ,
so the angle φ of the tangent vector to the z-axis is given by cos φ = (1 +

2
a2 θ0 )−1/2 .
Hence, our assumption that α(v) is a helix implies that θ0 is a non-zero
constant, so that θ = c1 v + c2 , where c1 6= 0 and c2 are constants. The initial
condition α(0) = (a, 0, 0) shows that c2 is an integer multiple of 2π, so if we let
u = c1 v then α(u) = (a cos u, a sin u, u/c1 ), which is of the required form.
(ii) Assume that α(s) has constant κ and τ . It follows immediately from Serret-
Frenet that the derivative of τ t − κb is zero, so that τ t − κb is constant, X 0 ,
say.
Let a = κ/(κ2 + τ 2 ). Then, using Serret-Frenet,
d κ
(α + an) = t + 2 (−κt − τ b)
ds κ + τ2
τ
= 2 (τ t − κb) .
κ + τ2
Hence α + an has constant rate of change τ X 0 /(κ2 + τ 2 ), so that

τ
α + an = Y 0 + s X0 ,
κ2 + τ2
for some constant vector Y 0 . Since n is perpendicular to t and to X 0 , it follows
that α lies on the cylinder S of radius a whose axis of rotation is the line through
Y 0 in direction X 0 . Finally, we note that t.X 0 = τ , which is constant, so that
α is a helix on S.
1.17. Assume there is a unit vector X 0 such that t.X 0 = c, a constant.

Then n.X 0 = 0, so that X 0 = ct + c1 b for some constant c1 . Then 0 = X 0 0 =
|α0 |(cκ + c1 τ )n, so that κ/τ = −c1 /c which is constant.
Conversely, if κ/τ = k, a constant, the Serret-Frenet formulae show that
(t − kb)0 = |α0 | κn − (κ/τ )τ n = 0 ,

so that t − kb is constant. The result follows since t.(t − kb) = 1.
1.18. The assumption on α implies the existence of a smooth function r(u)

such that α + rn = p. If we differentiate this expression, use Serret-Frenet, and
then equate the coefficients of t, n and b to zero, we find that r0 = 0 (so that
r is a non-zero constant), κ = 1/r, and τ = 0. The result now follows from
Lemma 1 of §1.5 and Example 8 of §1.3.
1.19. The given information implies that nα = ±nβ .

(i) Differentiating tα .tβ (with respect to u), and using Serret-Frenet, we find
that
(tα .tβ )0 = |α0 |κα nα .tβ + |β 0 |κβ tα .nβ = 0 ,
so that tα .tβ is constant.
(ii) The given information implies the existence of a smooth function r(u) such
that β = α + rnα . Differentiating this, using Serret-Frenet, and taking inner
product with nα gives that r0 = 0. Hence result.
√ √
1.20. Here,
√ |α0 | = a2 + b2 and tα = (−a sin u, a cos u, b)/ a2 + b2 . Hence
sα (u) = u a2 + b2 , from which it follows that
β(u) = (a cos u + au sin u, a sin u − au cos u, 0) ,
so, in particular, the third component of β is zero. The circle of intersection

of the plane z = 0 with the cylinder x2 + y 2 = a2 may be parametrised as
γ(u) = (a cos u, a sin u, 0), and the exercise may now be completed using the
formula for the involute by noting that γ 0 = (−a sin u, a cos u, 0), so that tγ =
(− sin u, cos u, 0) and sγ (u) = au.
1.21. First assume that α(s) lies on a sphere with centre p and radius r,
or, equivalently, that (α − p).(α − p) = r2 . We shall differentiate repeatedly to
find an expression for α − p in terms of t, n and b.
So, differentiate once to find that (α − p).t = 0. Differentiating again and
using Serret-Frenet, we obtain (α − p).n = −1/κ. Differentiating this and using
Serret-Frenet gives (α − p).b = −κ0 /(τ κ2 ). It now follows that
1 κ0
α − p = − n − 2b .
κ τκ
The derivative of the left hand side, and hence of the right hand side, of the
above equation is t. In particular, the coefficient of b of the derivative of the
right hand side is zero, which gives the desired relation between κ and τ .
Conversely, assume that κ and τ for a regular curve α(s) are related as in
the given formula, and let
1 κ0
p(s) = α + n + 2b .
κ τκ
The given relation between κ and τ may be used to show that p0 = 0 so that p
is constant. It now follows (again by differentiating) that (α − p).(α − p) is also
constant. Since α is not constant we have that |α − p| = r for some positive
constant r, so that α lies on the sphere with centre p and radius r.
Chapter 2
2.1. The line through (u, v, 0) and (0, 0, 1) may be parametrised by α(t) =
2
t(u, v, 0) + (1 − t)(0, 0, 1). This line intersects S 2 (1) when tu, tv, (1 − t) = 1,
and a short calculation gives that t = 0 or t = 2/(u2 + v 2 + 1). Since t = 0
corresponds to (0, 0, 1), we quickly see that x(u, v) is as claimed.
The formula for F follows from consideration of similar triangles, OR we
may use the following method which is similar to the one used in the earlier
part of the solution to this exercise. The line through (x, y, z) and (0, 0, 1) may
be parametrised by β(t) = t(x, y, z) + (1 − t)(0, 0, 1). For z 6= 1, this line cuts
the xy-plane when t = (1 − z)−1 , which gives the point (1 − z)−1 (x, y, 0) on the
line β. The formula for F now follows.
That F x(u, v) = (u, v) is a routine calculation (and also follows from the
geometrical construction). That x is a local parametrisation as claimed is now
immediate from conditions (S1) and (S2), taking U = R2 and W = R3 \ P .
2.2. Let V be an open subset of Rn such that X = V ∩ S is non-empty. If

p ∈ X, let x : U → Rn be a local parametrisation of S whose image contains
p. Then x−1 (V ) = {(u, v) ∈ U : x(u, v) ∈ V } is an open subset of R2 and the
restriction of x to x−1 (V ) is a local parametrisation of X whose image contains
p. To see this, if F : W → R2 is a map satisfying condition (S2) for x then
consider the restriction of F to V ∩ W .
2.3(a). (i) There are many ways. The one which perhaps is closest to that
given in Example 4 of §2.1 is to cover the cylinder by four local parametrisations.
2 + 3
Firstly, let U = {(u, v)
√ ∈ R : −1 < u < 1, v ∈ R} and let x 3 : U → R be
+ 2
given by x (u, v) = ( 1 − u , u, v). If we let W = {(x, y, z) ∈ R : x > 0} and
let F : W → R2 be given by F (x, y, z) = (y, z), then conditions (S1) and (S2)
are satisfied. Hence x+ is a local parametrisation of the given cylinder S, and
the corresponding coordinate neighbourhood is shown in the left hand picture
of Figure 5. The whole of S may be covered by x+ and an additional three
local parametrisations with domain U given by
p
x− (u, v) = (− 1 − u2 , u, v) ,
p p
y + (u, v) = (u, 1 − u2 , v) , y − (u, v) = (u, − 1 − u2 , v) .
In fact, the whole of S may be covered by just one parametrisation. For

instance, let U = R2 \ {(0, 0)} and define x̃ : U → R3 by

u v 1 2 2
x̃(u, v) = , , log(u + v ) .
(u2 + v 2 )1/2 (u2 + v 2 )1/2 2
Then, if we take W to be R3 , and let F : W → R2 be given by F (x, y, z) =

(xez , yez ), it is routine to show that (S1) and (S2) both hold.
(ii) If x2 + y 2 = 1 then (x cos u − y sin u)2 + (x sin u + y cos u)2 = 1 so
Figure 5: Coordinate neighbourhoods on a cylinder
that the cylinder is setwise invariant under rotations about the z-axis. It is
obtained by rotating the curve α(v) = (1, 0, v) about the z-axis. The coordinate
neighbourhoods corresponding to the two local parametrisations x and y as
discussed in §2.2 are shown in Figure 5. The coordinate neighbourhood for x
is the cylinder with one line deleted (shown in broken bold face) in the centre
picture of Figure 5, and for y is the cylinder with one line deleted (shown in
bold face) in the picture on the right hand side of Figure 5.
(iii) If f (x, y, z) = x2 + y 2 − 1 then grad f = (2x, 2y, 0), which is zero on and
only on the z-axis. Since f maps the z-axis to −1, and since, for example,
f (1, 0, 0) = 0, the equation f (x, y, z) = 0 defines a surface by Theorem 1 of
§2.4.
√
2.3(b). (i) Use the local parametrisation x(u, v) = (u, v, u2 + v 2 + 1) (as
given in Example 2 of §2.1) to cover the upper sheet, taking W = {(x, y, z) ∈
R3 : z > 0}, and F (x, y, z) = (x, y) for all (x, y, z) ∈√W .
Cover the lower sheet by using y(u, v) = (u, v, − u2 + v 2 + 1), taking W =
{(x, y, z) ∈ R3 : z < 0} and F (x, y, z) = (x, y) for all (x, y, z) ∈ W .
(ii) If x2 +y 2 = z 2 −1 then (x cos u−y sin u)2 +(x sin u+y cos u)2 = z 2 −1, so that
this hyperboloid of two sheets is setwise invariant under rotations about the z-
axis. It is obtained by rotating the union of the curves α(v) = (sinh v, 0, cosh v),
v ≥ 0, and β(v) = (sinh v, 0, − cosh v), v ≥ 0, about the z-axis.
(iii) If f (x, y, z) = x2 + y 2 − z 2 + 1, then the given hyperboloid of two sheets S
has equation f (x, y, z) = 0. This is non-empty (eg, f (0, 0, 1) = 0) and grad f is
zero at (0, 0, 0) only, which is not on S.
2.4. Note that f isn’t smooth on the z-axis, so in order to apply Theorem
1 of §2.4 we must restrict the domain of definition of f to W = R3 \ z-axis. On
W,
p p !
x2 + y 2 − a x2 + y 2 − a
grad f = 2 x p ,y p ,z .
x2 + y 2 x2 + y 2
Hence, grad f = 0 on and only on the circle x2 + y 2 = a2 in the plane z = 0,

and all these points are mapped by f to 0.
If 0 < b then, by Theorem 1 of §2.4, the set S of points of W satisfying the
equation f (x, y, z) = b2 is a surface. If 0 < b < a then every point satisfying the
equation f (x, y, z) = b2 is actually in W , so that S is equal to the set of points
in R3 satisfying f (x, y, z) = b2 , so this latter set is indeed a surface. It is a torus
of revolution since it is setwise invariant under rotations about the z-axis, and
intersects the half-plane y = 0, x > 0 in the circle centre (a, 0, 0) radius b (and,
in particular, x > 0 on this circle).
If 0 < a < b, there are points on the z-axis satisfying f (x, y, z) = b2 , so
we cannot conclude that the set of points of R3 satisfying this equation form a
surface. In fact, this set of points is obtained by rotating the circles of Figure
6 round the z-axis. You will see that “non-surface like” things happen on the
z-axis.
Figure 6: f (x, y, z) = b2 when 0 < a < b
2.5(a). Since grad f = 2(x + y + z − 1)(1, 1, 1), it is clear that grad f = 0

exactly when x + y + z = 1. These are the points of R3 which are mapped
to zero under f . We can conclude that the equation f (x, y, z) = k defines a
surface for any k > 0 (namely, the union of two parallel planes), while, if k < 0
then f (x, y, z) = k is the empty set (and so not a surface). Note that, although
grad f vanishes at EVERY point satisfying f (x, y, z) = 0, this set is still a
surface, namely the plane x + y + z = 1.
This example shows that Theorem 1 of §2.4 is not an “if and only if” theorem.
2.5(b). If f (x, y, z) = xyz 2 , then grad f = 0 if and only if either z = 0 or

x = y = 0. It follows that if k 6= 0 then f (x, y, z) = k defines a surface. The set
xyz 2 = 0 is the union of the three coordinate planes, which is not a surface.
2.6. We have
α0 = |α0 |t and α00 = |α0 |t0 + |α0 |0 t .
Hence,
xu = |α0 | + v|α0 |0 t + v|α0 |t0 xv = |α0 |t .

and
If κ(u) = 0 then t0 is zero, so that xu and xv are linearly dependent. If κ(u) 6= 0
then t0 = |α0 |κn, so that xu and xv are linearly dependent if and only if v = 0.
The next statement is immediate from Theorem 1 of §2.5.
2.7. The ellipsoid S is defined by f (x, y, z) = 1, where f (x, y, z) = (x2 /a2 )+

(y /b2 ) + (z 2 /c2 ). A routine check shows that grad f is never zero on S, so that
2
S is a surface.
(i) One suitable choice of F : W → R2 is obtained by taking
W = {(x, y, z) ∈ R3 : |z| < c, and if y = 0 then x > 0} ,
and x y z
F (x, y, z) = Arg +i, arcsin .
a b c
(ii) We check conditions (1), (2) and (3) of that theorem. Firstly, f x(u, v) =
cos2 v + sin2 v = 1, so (1) holds. Secondly, if x(u1 , v1 ) = x(u2 , v2 ) then com-
paring the third component gives that v1 = v2 . The first two components then
show that u1 = u2 . Finally, it is clear that xu and xv are linearly independent
unless cos v = 0, which never happens for the given range of values of v.
2.8. (i) Use Theorem 1 of §2.4 to show that S is a surface by checking

that if f (x, y, z) = x2 − y 2 − z then grad f doesn’t vanish on the non-empty
subset of R3 defined by equation f (x, y, z) = 0. We use Theorem 4 of §2.5 to
show that x is a local parametrisation. We first note that f x(u, v) = 0 for all
(u, v) ∈ R2 , so that the first condition holds. If x(u1 , v1 ) = x(u2 , v2 ) then the
first two components show that cosh u1 − sinh u1 = cosh u2 − sinh u2 , so the
third component shows that v1 = v2 from which it quickly follows that u1 = u2 .
Finally, since sinh u 6= cosh u, a comparison of the first two components of xu
and xv shows that xu and xv are linearly independent at all points.
(ii) This may be done using similar methods. We first note that the components
of x(u, v) satisfy the given equation for all 0 < u < 2π, v ∈ R, so that the first
condition holds. If x(u1 , v1 ) = x(u2 , v2 ) then the first and third components
show that v1 = v2 , and then, using all three components, we may prove that
sin u1 = sin u2 and cos u1 = cos u2 so that u1 = u2 . Finally, it may be checked
that xu × xv is never zero so that xu and xv are linearly independent at all
points.
2.9. Show that S is a surface by applying Theorem 1 of §2.4 to the function

f (x, y, z) = x sin z − y cos z. It is clear that f x(u, v) = 0 for all (u, v) ∈ R2 , so
that the image of x is a subset in S. To show that the image of x is the whole
of S, we note that if (x, y, z) ∈ S then x(z, x cos z + y sin z) = (x, y, z).
Part (i) is straightforward; for instance, xu ×xv = (− sin u, cos u, −v), which
is never zero. For (ii), we note that our checking of the surjectivity of x shows
that we can take F (x, y, z) = (z, x cos z + y sin z).
2.10. The statements in the given hint are straightforward to check, and
it follows from these that f u and f v are linearly independent unless f u = 0.
However, if f u = 0 then the third component of f u shows that either v = 0 or
u = (2n + 1)π. If v = 0 then f u = 2α0 6= 0, while if u = (2n + 1)π then the
second component of f u is equal to ±v/2. It follows that f u is never zero, so
that f u and f v are linearly independent for all values of u and v.
We now define two maps x and y by restricting the domain of definition of
f . Specifically, we let
x(u, v) = f (u, v) , −π < u < π , −1 < v < 1 ,
y(u, v) = f (u, v) , 0 < u < 2π , −1 < v < 1 .
It is clear that S is the union of the images of x and y. We now show that x
and y are both injective. There are several ways - here is a geometrical one. The
line segment on S through 2α(u) projects down (under orthogonal projection)
to the line segment 2 + v cos(u/2) α(u) in the xy-plane and it is clear that,
since −1 < v < 1, no two distinct segments intersect. Injectivity of x and y
now follows.
The exercise may be completed by applying Theorem 4 of §2.5 to show that
x(u, v) and y(u, v) are both local parametrisations of S.
2.11. First check that all points in the image of x satisfy the given equations.
Then check that x is injective (on its domain of definition). Finally check that
xu and xv are linearly independent at all points. The image of x is the flat
torus with two circles omitted. The whole of the flat torus may be covered by
considering three further parametrisations which have the same formula as x
but have domains (0, 2π) × (0, 2π), (0, 2π) × (−π, π), and (−π, π) × (0, 2π).
Chapter 3
3.1. Here, xu = (1, 0, 2u) and xv = (0, 1, 3v 2 ). Since these vectors span the
tangent plane, we would need both (1, 0, 2u).(−1, 1, 0) and (0, 1, 3v 2 ).(−1, 1, 0)
to be zero. Clearly, neither expression can be zero!
3.2. Clearly x(2, −1) = (2, −1, 3) ∈ S. At this point, xu = (1, 0, 4) and
xv = (0, 1, 3). Also (1, −1, 1) = (1, 0, 4) − (0, 1, 3) and so is in T(2,−1,3) S. For
the last part, we need scalars λ and µ such that λ(1, 0, 4)+µ(0, 1, 3) is orthogonal
to (1, −1, 1). This gives 5λ + 2µ = 0, so we could take λ = 2 and µ = −5. Hence
2(1, 0, 4) − 5(0, 1, 3) = (2, −5, −7) is a vector in T(2,−1,3) S which is orthogonal
to (1, −1, 1).
3.3. Let f (x, y, z) = 2x2 − xy + 4y 2 . Then grad f = (4x − y, −x + 8y, 0).

In particular, (grad f )(0, 1/2, 2) = (−1/2, 4, 0), so the unit normal there is
√
(−1, 8, 0)/ 65. A basis for the tangent plane of S at (0, 1/2, 2) is provided by
any pair of linearly independent vectors orthogonal to the normal, for instance
(0, 0, 1) and (8, 1, 0).
3.4. Let f (x, y, z) = (x2 /a2 ) + (y 2 /b2 ) + (z 2 /c2 ). Then the ellipsoid S has
equation f (x, y, z) = 1. Also, grad f = 2(x/a2 , y/b2 , z/c2 ), which is never zero
on S so that (grad f )(p) is a non-zero vector orthogonal to S at p. This is parallel
to (1, 1, 1) if and only if (x/a2 ) = (y/b2 ) = (z/c2 ). Using this to eliminate x
and y in the equation of the ellipsoid, we find that z 2 = c4 /(a2 + b2 + c2 ). So
the tangent plane√ at a point (x, y, z) of the ellipsoid is orthogonal to (1, 1, 1) at
±(a2 , b2 , c2 )/ a2 + b2 + c2 .
3.5. Let f (x, y, z) = x2 + y 2 − z 2 . Then S has equation f (x, y, z) = 1. Also

grad f = (2x, 2y, −2z), which is never zero on S so if p = (x, y, z) ∈ S then
(grad f )(p) is a non-zero vector orthogonal to S at p. So, if p = (x, y, z) ∈ S,
then T(x,y,z) S is orthogonal to (1, 0, −1) if and only if x2 + y 2 − z 2 = 1, x = z,
y = 0, which is not possible. However T(x,y,z) S is orthogonal to (1, 1, 1) if and
only if x2 + y 2 − z 2 = 1 and x = y = −z. The two possibilities are ±(1, 1, −1).
3.6. . Let f (x, y, z) = x2 + y 2 + z 2 − ax and g(x, y, z) = x2 + y 2 + z 2 − by.

Then grad f = (2x − a, 2y, 2z) and grad g = (2x, 2y − b, 2z). Grad f = 0 only
at (a/2, 0, 0), and since f (a/2, 0, 0) = −a2 /4 6= 0 it follows that the equation
f (x, y, z) = 0 defines a surface S1 in R3 . Similarly, g(x, y, z) = 0 defines a
surface S2 . If (x0 , y0 , z0 ) ∈ S1 ∩ S2 then x0 2 + y0 2 + z0 2 = ax0 = by0 , from
which it follows that, at (x0 , y0 , z0 ), grad f.grad g = 0. It follows that S1 and
S2 intersect orthogonally at all points of intersection. The intersections of S1
and S2 with x2 + y 2 + z 2 = cz are dealt with in a similar manner.
3.7. We use the notation from the solution of Exercise 3.4. Since (grad f )(p)
√
is orthogonal to√S at p, the equation of the tangent plane based at (a/2, b/2, c/ 2)
is x/a + y/b + 2z/c = 2.
3.8. Here, xu = (cos v, sin v, 1) and xv = (−u sin v, u cos v, − tan v). Hence
E = xu .xu = 2, and, similarly, F = − tan v, and G = u2 + tan2 v.
3.9. Let f1 (x1 , x2 , x3 , x4 ) = x1 2 + x2 2 and let f2 (x1 , x2 , x3 , x4 ) = x3 2 + x4 2 .

Then grad f1 = (2x1 , 2x2 , 0, 0) and grad f2 = (0, 0, 2x3 , 2x4 ), which are linearly
independent unless either x1 = x2 = 0 or x3 = x4 = 0. Since f (r1 , 0, r2 , 0) =
(r1 2 , r2 2 ), it follows that (r1 2 , r2 2 ) is a regular value of f for each pair of positive
real numbers r1 , r2 .
We now check the conditions for the application of Theorem 4 of §2.5. Firstly,
it is clear that f x(u, v) = (r1 2 , r2 2 ) so that (1) holds. That (2) holds is imme-
diate. Finally, xu = (−r1 sin u, r1 cos u, 0, 0) and xv = (0, 0, −r2 sin v, r2 cos v),
so that xu and xv are linearly independent for all (u, v).
The coefficients of the first fundamental form are E = r1 2 , F = 0, G = r2 2 .
3.10. Routine calculations show that the surface S has E = G = cosh2 v,

F = 0. That xu = x̃v and xv = −x̃u is straightforward.

Also,
Eθ = (cos θ xu + sin θ x̃u ).(cos θ xu + sin θ x̃u )
= (cos θ xu − sin θ xv ).(cos θ xu − sin θ xv )
= E cos2 θ − 2F cos θ sin θ + G sin2 θ
= cosh2 v .
That Fθ = 0 and Gθ = cosh2 v are proved in a similar manner.
Finally, we note that the tangent plane to Sθ is spanned by cos θ xu −sin θ xv
and cos θ xv + sin θ xu , which is parallel to the plane spanned by xu and xv .
Ru p
3.11. The length of αv from t = u0 to t = u1 is u01 E(t, v) dt, which is
independent of v since E is independent of v.
3.12. (i) Here α0 = xu + xv so that |α0 |2 = E + 2F + G, where the right

hand side of the equation is evaluated at (t, t). This gives that |α0 | = 2 cosh t,
R1
so the required length is 2 0 cosh t dt = 2 sinh 1.
(ii) The two given curves intersect when t = r = 0, which gives u = v = 0. The
result follows since α0 (0).β 0 (0) = (xu + xv ).(xu − xv )(0, 0) = (E − G)(0, 0) = 0.
3.13. √The curves intersect when t = π/3 and √ r = 1/2, which gives u = 1/2
and v = 3/2.√ At this point, E = 2, F = 1/ 3, G = 4/3. √ We also have that
α0 (π/3) = −( 3/2)xu + (1/2)xv , while β 0 (1/2) = xu + 3xv . A calculation
using |α0 (π/3)||β 0 (1/2)| cos θ = α0 (π/3).β 0 (1/2) yields the given answer.
R π/2 p
3.14.Here, L(u) = −π/2 G(u, t) dt. Hence, differentiating under the in-
R π/2 √
tegral sign, 2L0 (u) = −π/2 Gu / G dt. A short calculation shows that Gu =
(b2 − a2 ) sin2 v sin 2u, which is identically zero when u = −π/2, 0, π/2, and π.
Each curve αu forms one half of the intersection of S with the plane in R3
containing the z-axis and having equationp xb sin u = ya cos u. This is an ellipse
with principal semi-axes of lengths a2 cos2 u + b2 sin2 u and c. The critical
lengths |a| and |b| of the first of these semi-axes correspond to u = 0, π and
u = ±π/2 respectively.
3.15. Here
2
xu = (−u2 + v 2 + 1, −2uv, 2u) ,
(u2 + v 2 + 1)2
so that
4 4
(−u2 + v 2 + 1)2 + 4u2 v 2 + 4u2 = 2

E= .
(u2 2
+ v + 1) 4 (u + v 2 + 1)2
That F = 0 and G = 4(u2 + v 2 + 1)−2 may be shown in a similar manner.
3.16. Let α(u)R = x(u, 0), u ∈ R. Then |α0 |2 = E(u, 0), so the required
∞
length is given by −∞ 2/(u2 + 1) du = 2[arctan u]∞
−∞ = 2π.
3.17. Let f (u + iv) = x(u, v) + iy(u, v), where x(u, v) and y(u, v) are the
real and imaginary parts of f . Then x(u, v) = u, v, x(u, v), y(u, v) so that
xu = (1, 0, xu , yu ) and xv = (0, 1, xv , yv ). That E = G, and F = 0, follows from
the Cauchy-Riemann equations, xu = yv and xv = −yu .
3.18. Here, E = 1 + v 2 , F = 1 and G = 2. Also, F is given by φ(u, v) =

constant, where φ(u, v) = u+v. It follows that the tangent vectors to the family
F are scalar multiples of xu − xv .
π π π
A short calculation shows that if x(u, v) = √ , , then u = π/6
2 3 6 2
and v = π/3, at which point E = 1 + π 2 /9, F = 1 and G = 2. Then
|xu ||xu − xv | cos θ = xu .(xu − xv ) ,
so that E(E − 2F + G) cos2 θ = (E − F )2 . Substituting the values of E, F , G

at (π/6, π/3), we find that (1 + π 2 /9)2 cos2 θ = (π 2 /9)2 . It follows that cos θ is
as claimed.
We also see that β(r) = x u(r), v(r) is an orthogonal trajectory of F if
and only if (xu du/dr + xv dv/dr).(xu − xv ) = 0 which holds if and only if
Edu/dr + F (dv/dr − du/dr) − Gdv/dr = 0 which quickly gives that
du 1 dv
= 2 .
dr v dr
Integrating, we see that the orthogonal trajectories of F are given by u +
1/v = constant. Since x(0, 1) = (1, 0, 1), we see that the orthogonal trajectory of
F through (1, 0, 1) has equation u+1/v = 1. Since x(π/2, −π/2) = (0, −π/2, 0),
we see that the orthogonal trajectory of F through (1, 0, 1) does not pass through
(0, −π/2, 0).
3.19. Here, E = G = 1 and F = 0. The family F is given by φ(u, v) =

constant, where φ(u, v) = v sin u/(1 − cos u), and it follows that if x u(t), v(t)
is a member of F then
(v 0 sin u + vu0 cos u)(1 − cos u) = u0 v sin2 u .
This simplifies to give u0 v = v 0 sin u, so that the tangent vectors to F are

scalar multiples of sin u xu + vxv . It then follows that β(r) = x u(r), v(r) is
an orthogonal trajectory of F if and only if
du dv
sin u = −v .
dr dr
Integrating, we see that the orthogonal trajectories of F are given by ψ(u, v) =
constant where ψ(u, v) = 2 cos u − v 2 . Since |x(u, v)|2 = 2(1 − cos u) + v 2 , we
see that |x(u, v)| is constant on each of these trajectories.
3.20. A non-zero vector X bisects the angle between the coordinate curves if
and only if X.xu /|xu | = ±X.xv /|xv |. This quickly leads to the given condition.
Let x(u, v) = (u, v, u2 − v 2 ). We note that x(1, 1) = (1, 1, 0), and, since
xu (1, 1) = (1, 0, 2) and xv (1, 1) = (0, 1, −2), we see that E(1, 1) = G(1, 1) = 5
and F (1, 1) = −4. This leads to α = ±β, so that X is any non-zero scalar
multiple of either (1, 1, 0) or (1, −1, 4).
3.21. Here, E = v 2 + 1, F = 0, G = 1. So, using the result of Exercise

3.20, it follows that a curve β(r) = x u(r), v(r) bisects the angles between the
coordinate curves if and only if
du dv 1
=± √ .
dr dr v 2 + 1
Integrating, we see that u ± arcsinh v = constant.
3.22. If α(t) = x u(t), v(t) is a member of F, then α0 (t) = u0 xu + v 0 xv

where u0 φu + v 0 φv = 0. Hence u0 /v 0 = −φv /φu , and it follows that tangent

vectors to the family F at x(u,
v) are the scalar multiples of φv xu − φu xv .
If x(u, v) = u, v, φ(u, v) then E = 1 + φu 2 , F = φu φv , G = 1 + φv 2 . The
family F of level curves of φ (or, in other words, the contours on S) are given
by φ(u, v) = constant, and the paths of steepest
descent are the orthogonal
trajectories of F. So, if β(r) = x u(r), v(r) is a path of steepest descent then

du dv
xu + xv . (φv xu − φu xv ) = 0 .
dr dr
Using the expressions for E, F and G given above, this equation simplifies to
give
du dv
φv − φu = 0 .
dr dr
If φ(u, v) = u3 + v 3 , then the orthogonal trajectories have v 2 du 2 dv
dr − u dr = 0,
so that
1 du 1 dv
2
− 2 =0.
u dr v dr
Integrating, we see that the paths of steepest descent are given by ψ(u, v) =
1 1
constant, where ψ(u, v) = − .
u v
3.23. The image of x is a cone having vertex at the origin with axis of
rotation being the z-axis. That xu and xv are linearly dependent if and only
if v = 1 is quick. The curve α is the intersection of the cone with the plane
z = −1, and x(u, 1) = (0, 0, 0) for all u ∈ R.
3.24. The given line lies on S if and only if, for all λ ∈ R,
(p1 + λv1 )(p2 + λv2 ) = (p3 + λv3 ) .
Taking into account that (p1 , p2 , p3 ) ∈ S, we see that the above holds if and
only if v1 p2 + p1 v2 = v3 and v1 v2 = 0. This implies that (v1 , v2 , v3 ) is a scalar
multiple of either (0, 1, p1 ) or (1, 0, p2 ). This shows that S is a doubly ruled
surface, and the rulings intersect orthogonally if and only if p1 p2 = 0. Hence

result.
3.25. (i) The left hand side of the equation is equal to (xu × xv )(u0 , v0 ).
(ii) Let γ(u) = α(u) + λ(u)β(u) for some smooth function λ(u). A short
calculation shows that γ 0 .β 0 = 0 if and only if λ = −α0 .β 0 /|β 0 |2 .
(iii) Since β 0 is orthogonal to both γ 0 and β, it follows that β 0 is parallel to
γ 0 × β. If we put γ 0 × β = µβ 0 , then a short calculation shows that y u × y v = 0
if and only if µ(u) = v = 0 (and v = 0 corresponds to the striction curve).
3.26. Let S be a ruled surface of revolution, and assume first that the rul-
ings of S are complete lines (rather than just line segments). Let ` be a line of
the ruling, let a ≥ 0 be the perpendicular distance of ` from the axis of rotation,
and let θ be the angle between ` and the axis of rotation.
By applying a rigid motion of R3 , we may assume that the axis of rota-
tion is the z-axis, and that ` is the line parametrised by α(t) = (a, 0, 0) +
t(0, sin θ, cos θ).
If a = 0 we have a cone or a plane, so we now assume a > 0. When the point
α(t) is rotated about
p the z-axis to be in the half-plane y = 0, x > 0, we obtain
the point β(t) = ( a2 + t2 sin2 θ, 0, t cos θ), so that β is a parametrisation of
the curve in the xz-plane with equation x2 − z 2 tan2 θ = a2 . If θ = 0, then S
is the cylinder x2 + y 2 = a2 ; if θ = π/2, then S is a plane (with, arguably, the
possibility that a disc centred on the origin may be removed). Otherwise, S is
the hyperboloid of revolution x2 /a2 + y 2 /a2 − z 2 /c2 = 1, where c2 = a2 cot2 θ.
So, the surfaces of revolution ruled by complete lines are planes, cones, cylin-
ders, and those hyperboloids of one sheet discussed in Example 2 of §3.6 for
which a = b. Similar arguments show that surfaces of revolution ruled by line
segments are rotationally invariant open subsets of the surfaces just described.
3.27. If we parametrise S 2 (1) by rotating (cos v, 0, sin v), −π/2 < v < π/2,
about the z-axis, then E = cos2 v, F = 0 and G = 1. Hence the area of the
southern hemisphere is (as expected!)
Z 0 Z π
cos v du dv = 2π .
−π/2 −π
3.28. Here, EG − F 2 = (1 − u2 )−1 (1 − v 2 )−1 . Required area is π 2 /4.
3.29. We parametrise S 2 (1) as in the solution to Exercise 3.27, and let E,

F and G be the coefficients of the first fundamental form.
If we define a local parametrisation of the cylinder by letting x̃(u, v) =
f x(u, v) = (cos u, sin u, sin v), then routine calculations show that Ẽ G̃ − F̃ 2 =
EG − F 2 , and the result follows from (3.20).
R 2π R 2π
3.30. Here E = r1 2 , F = 0 and G = r2 2 . Required area is 0 0
r1 r2 dudv =
4π 2 r1 r2 as claimed.
3.31. (i) If the coordinate curves Rof xpform a Tchebycheff net, then, for
u
each fixed u0 < u1 , the length L(v) = u01 E(t, v) dt of the coordinate curve
u 7→ x(u, v), u0 < u < u1 ,√is independent of v. By differentiating under the
integral sign, we see that ( E)v = 0, and hence Ev = 0. Similar reasoning
shows that Gu =√ 0. This argument may be reversed √ to give the converse.
(ii) Here, ũu = E, ũv = 0, ṽu = 0 and ṽv = G. In particular, if u (resp.
v) is constant then so is ũ (resp. ṽ). The statement concerning the families of
coordinate curves now follows. It also follows that
√ √
xu = x̃ũ E , xv = x̃ṽ G ,
p
so that E = ẼE and G = G̃G. Hence Ẽ = G̃ = 1, and cos θ = F̃ / Ẽ G̃ = F̃ .
Chapter 4
4.1. Here, xu = (− sinh v sin u, sinh v cos u, 1) and xv = (cosh v cos u, cosh v sin u, 0).
It follows that xu × xv = cosh v(− sin u, cos u, − sinh v). Hence
N = sech v(− sin u, cos u, − sinh v) .
This is not injective since it is periodic in u.
4.2. If we parametrise
the surface of revolution as usual using a generating
curve f (v), 0, g(v) , f (v) > 0 for all v, then the formula in Example 2 of §4.1
2
shows p that the parallel ofpS through v = v0 maps to the parallel of S (1) through
2 2 2 2
(g 0 / f 0 + g 0 , 0, −f 0 / f 0 + g 0 )(v0 ), while the meridian u = u0 maps to the
meridian of S 2 (1) through (cos u0 , sin u0 , 0).
p
2 2
√ f = (2x, 2y, −1), so that N = (2x, 2y, −1)/ 4x + 4y + 1 =
4.3.(i) Grad
(2x, 2y, −1)/ 4z + 1. Hence the image of N is contained in the lower hemi-
sphere of S 2 (1). However, if X 2 + Y 2 + Z 2 = 1 with −1 ≤ Z < 0, then

X Y 1 2
N − ,− , (1 − Z ) = (X, Y, Z) ,
2Z 2Z 4Z 2
so it follows that the image of N is the lower hemisphere (excluding the equator).
ALTERNATIVELY: Since S is a surface of revolution, the image of N will
be obtained by rotating the image under N of the generating curve (v, 0, v 2 ),
v ≥ 0 of S. Since N (v, 0, v 2 ) is the unit 2
p vector in direction (grad f )(v, 0, v )
2 2
we find that N (v, 0, v ) = (2v, 0, −1)/ (1 + 4v ). If (X, Z) = (cos θ, sin θ) for
−π/2 ≤ θ < 0, then, when v = (−1/2) cot θ ≥ 0, N (v, 0, v 2 ) = (X, 0, Z). Hence
the image of N is as obtained above.
(ii) This surface is obtained by rotating the curve (cosh v, 0, sinh v), v ∈ R,
about the z-axis. Also,
N (cosh v, 0, sinh v) = (cosh v, 0, − sinh v)(1 + 2 sinh2 v)−1/2 .

√
The third component takes all values between ±1/ 2. Hence result as given.
(iii) This surface is obtained by rotating the union of the two curves (sinh v, 0, ± cosh v),
v ≥ 0, about the z-axis. Also,
N (sinh v, 0, ± cosh v) = (sinh v, 0, ∓ cosh v)(2 cosh2 v − 1)−1/2 .

√
It follows that the image of the Gauss map is {(x, y, z) ∈ S 2 (1) : |z| > 1/ 2}.
4.4. We may use the method of the alternative solution to Exercise 4.3(i),
since the catenoid is obtained by rotating (cosh v, 0, v), v ∈ R, about the z-axis.
Here N (cosh v, 0, v) = (1, 0, − sinh v)/ cosh v, and this equals (cos θ, 0, sin θ),
−π < θ ≤ π if and only if −π/2 < θ < π/2 and sinh v = − tan θ. The result
follows.
4.5. Here, h is the restriction to S of the map g : R3 → R given by the same

formula. Since g is linear, it is smooth and is equal to its own derivative at each
point. It follows that h : S → R is smooth and if X ∈ Tp S then dhp (X) = X.v.
This map is identically zero if and only if v is orthogonal to S at p.
4.6. This is similar to the above, but with g(p) = (q − p).(q − p). This is
smooth on R3 and the derivative is dgp (X) = 2X.(p − q). Now argue as in the
solution to Exercise 4.5.
4.7. The hypotheses imply the existence of a point q0 ∈ R3 and a function

f : S → R, such that, if p ∈ S, then p + f (p)N (p) = q0 . The result of the
previous exercise shows that f is smooth (except at q0 if this latter point is in
S). If we differentiate the equation, we find that if X is a tangent vector to S
then X + df (X)N + f dN (X) = 0. Taking the component of this orthogonal
to S, we deduce that df (X) = 0, so that f is constant r, say. Since r is clearly
non-zero, we see that q0 ∈ / S and S is an open subset of the sphere centre q0
and radius r.
4.8. Let (X, Y, Z) ∈ T(x,y,z) S 2 (1). Then xX+yY +zZ = 0, and df(x,y,z) (X, Y, Z) =
(aX, bY, cZ). The vector (x/a, y/b, z/c) is orthogonal to S̃ at (ax, by, cz), and
the inner product of (x/a, y/b, z/c) with (aX, bY, cZ) is zero.

4.9. One example is f (x, y, 0) = a cos(x/a), a sin(x/a), y . We check this
is a local isometry using the conditions given in Proposition 2 of §4.3. First
parametrise the plane using x(u, v) = (u, v, 0), so that f (u, v) = (a cos(u/a), a sin(u/a), v).
Then fu .fu = 1 = E, fu .fv = 0 = F , and fv .fv = 1 = G. This shows that f is
a local isometry, while f is surjective because if x2 + y 2 = a2 then (x/a, y/a) is
on the unit circle and so is equal to (cos θ, sin θ) for some −π < θ ≤ π. Hence
f (aθ, z, 0) = (x, y, z).
4.10. We first note that f is well-defined since, if (r cos θ, r sin θ) = (r̃ cos φ, r̃ sin φ)
then r = r̃ and φ and θ differ by an integer multiple of 2π, so that (r cos nθ, r sin nθ, br) =
(r̃ cos nφ, r̃ sin nφ, br̃). It is clear that the image of f is on S̃, while f is sur-
jective since every point of S̃ may be written in the form (r cos µ, r sin µ, br),
r > 0. We parametrise the plane S (minus the non-positive real axis) by us-
ing polar coordinates; x(r, θ) = (r cos θ, r sin θ, 0), r > 0, −π < θ < π. Then
f (r, θ) = (1/n)(r cos nθ, r sin nθ, br), so that fr = (1/n)(cos
√ nθ, sin nθ, b) and
fθ = (−r sin nθ, r cos nθ, 0). It follows that if b = n2 − 1 then fr .fr =
(1 + b2 )/n2 = 1 = E, fr .fθ = 0 = F , and fθ .fθ = r2 = G. The result fol-
lows from Proposition 2 of §4.3.
To see the geometry of f , note that f (r̃eiθ̃ ) = f (reiθ ) if and only if r̃ = r
and θ̃ = θ + (2k/n)π for some integer k. So, to model the map, cut a line from
the edge of the paper to the centre, then make a cone with vertex at the centre
of the paper by sliding the edges of the cut past each other until they line up
again after n − 1 circuits.
4.11. It is clear that the image of the restriction of f to S lies on S̃. Surjec-
tivity follows from the fact that f (r1 cos θ, r1 sin θ, r2 φ) = (r1 cos θ, r1 sin θ, r2 cos φ, r2 sin φ).
If we parametrise S by x(u, v) = (r1 cos u, r1 sin u, v), then E = r1 2 , F = 0 and
G = 1. Also, fu = (−r1 sin u, r1 cos u, 0, 0) and fv = 0, 0, − sin(v/r2 ), cos(v/r2 ) ,
so that fu .fu = r1 2 , fu .fv = 0 and fv .fv = 1. The result follows from Proposi-
tion 2 of §4.3.
4.12. Let p and q be points of S̃, and let f be the map described in Example
4 of §4.3. By first applying a translation of the plane, we may assume that
f (πr1 , πr2 , 0) = p. Then there exists 0 < u ≤ 2πr1 , 0 < v ≤ 2πr2 such that
f (u, v, 0)
√ = q. The line segment joining (πr1 , πr2 , 0) to (u, v, 0) has length at
most π r1 2 + r2 2 , and the image of this line segment under f is a curve on S̃
joining p to q.
4.13. If we parametrise the helicoid as in Example 2 of §4.5, then E =

G = cosh2 v, F = 0. A calculation gives that N u .N u = N v .N v = cosh−2 v
and N u .N v = 0, so Proposition 3 of §4.4 shows that N is conformal with
conformal factor λ = cosh−2 v.
4.14. The formula for f may be proved as in the solution to Exercise 2.1.
We note that a = 1/r1 > r2 /r1 = b, so that Ta,b is indeed a torus of
revolution. A calculation shows that p if x1 2 + x2 2 = r1 2 and x3 2 + x4 2 = r2 2 , and
if f (x1 , x2 , x3 , x4 ) = (x, y, z), then ( x2 + y 2 − a)2 + z 2 = b2 . Since r2 2 < 1,
S ⊂ X, so that f maps S to Ta,b . If we parametrise S by taking

u u v v
x(u, v) = r1 cos , r1 sin , r2 cos , r2 sin ,
r1 r1 r2 r2
then E = G = 1 and F = 0. −2

We now show conformality. Calculations show that fu .fu = 1−r2 sin(v/r2 )
and fu .fv = 0. A little more work using r1 2 + r2 2 = 1 shows that fv .fv =
−2
1 − r2 sin(v/r2 ) , so the conformality of f follows from Proposition 3 of §4.4.
It also follows that fu and fv are linearly independent at each point, so the
discussion at the beginning of §4.3 implies that if we show that f is bijective
then it will follow that f is a diffeomorphism. To show that f is bijective, it is
enough to show that the map

v v
r1 , 0, r2 cos rv2
r1 , 0, r2 cos , r2 sin 7→
r2 r2 1 − r2 sin rv2
defines a bijective map from the circle x1 = r1 , x2 = 0, x3 2 + x4 2 = r2 2 in R4

onto the circle (x − a)2 + z 2 = b2 , y = 0, in R3 . We do this by showing (with
some work) that the equation

r1 , 0, r2 cos rv2 r2

1

= (cos φ, 0, sin φ) + , 0, 0
1 − r2 sin rv2 r1 r1
has the unique solution

v v (r1 sin φ, 0, r2 + cos φ)
cos , 0, sin = .
r2 r2 1 + r2 cos φ
4.15. That the correspondence is an isometry is immediate from Proposi-

tion 1 of §4.5. Also N θ is a positive scalar multiple of (cos θxu + sin θx̃u ) ×
(cos θxv + sin θx̃v ) = (cos θxu − sin θxv ) × (cos θxv + sin θxu ) = xu × xv , which
is independent of θ.
4.16. We use the notation of Example 1 of §4.6. The coordinate plane x = 0

in R3 intersects S 2 (1) in the meridian through (0, 1, 0), and the image of this
under stereographic projection is the line u = 0 in the complex plane. This line
is mapped by f to the line u = 1, and x maps this line to the intersection of
S 2 (1) with the plane x + z = 1. The image of the equator of S 2 (1) is found
in a similar manner. (Brief details: The image of the equator under f may be
parametrised by (1+cos θ, sin θ) and the image of this under x consists of points
of the form
(2 + 2 cos θ, 2 sin θ, 1 + 2 cos θ)
.
3 + 2 cos θ
These points form the circle of intersection of S 2 (1) with the plane 2x − z = 1.)
Figure 7 shows the curves of intersection of S 2 (1) with (i) x = 0, (ii) z = 0,
Figure 7: For the solution to Exercise 4.16

and also shows their images under f˜. The circle x2 + z 2 = 1, y = 0, is left
setwise fixed.
4.17. (i) If p ∈ S then B(p, p) = −1, so, differentiating, we find that

B(X, p) = 0.
(ii) The line through (0, 0, −1) and (u, v, 0) may be parametrised as (0, 0, −1) +
λ(u, v, 1). If u2 + v 2 < 1, this line intersects S at the unique point at which
λ = 2/(1 − u2 − v 2 ). At this point (0, 0, −1) + λ(u, v, 1) is the given formula for
f (u, v).
(iii) The map f is also surjective since, if (x1 , x2 , x3 ) ∈ R3 , then the line
through (0, 0, −1) and (x1 , x2 , x3 ) intersects the plane x3 = 0 at (x̃1 , x̃2 , 0) =
(1 + x3 )−1 (x1 , x2 , 0), and a quick check shows that if (x1 , x2 , x3 ) ∈ S then
(x̃1 )2 + (x̃2 )2 < 1. Finally, a calculation shows that
2
fu = (1 + u2 − v 2 , 2uv, 2u) ,
(1 − u2 − v 2 )2
and another calculation shows that hfu , fu i = B(fu , fu ) = 4/(1 − u2 − v 2 )2 = Ẽ.

Similar calculations show that hfu , fv i = 0 = F̃ , while hfv , fv i = G̃. The result
now follows from Proposition 2 of §4.3.
√
4.18. First show that α0 (v) = (−v −2 , 0, v −2 v 2 − 1).
It is clear that the image of f is equal to S, and calculations similar to those
used to compute the coefficients of the first fundamental form of a surface of
revolution show that fu .fu = fv .fv = 1/v 2 , and fu .fv = 0. The result now
follows from Proposition 3 of §4.4.
4.19. (i) Here, f˜ = xf x−1 so the conformal factor of f˜ at x(w) is the

product of the conformal factors of x−1 at x(w), f at w and x at f (w). The
result follows.
(ii) This follows from (i) using the fact that f 0 (w) = (ad − bc)/(cw + d)2 .
(iii) This follows since the limit as w → ∞ of the conformal factor of f˜ at x(w)
is |ad − bc|/(|a|2 + |c|2 ), which is non-zero.
(iv) Assuming that ad − bc = 1, the map f˜ is an isometry if and only if
1 + ww̄ = (cw + d)(c̄w̄ + d) ¯ + (aw + b)(āw̄ + b̄) for all w. Comparing constants,
¯ 0 = cd¯+ ab̄, and
coefficients of w, and coefficients of ww̄, we obtain 1 = bb̄ + dd,
1 = aā + cc̄. The result follows.
4.20. We compute the conformal factor of the map f given in Example 4 of

Appendix 2. We have that
(ad − bc)2
g(fu , fu ) = .
(cz + d)2 (cz̄
+ d)2 (im f (z))2
A short calculation shows that if z = u + iv then the imaginary part of f (z)

is equal to (ad − bc)v(cz + d)−1 (cz̄ + d)−1 . It then follows that g(fu , fu ) = 1/v 2 ,
so that the conformal factor of f is equal to 1.
Chapter 5
√
5.1. In this case, N = (−gu , −gv , 1)/ D, where D = 1 + gu 2 + gv 2 . Also,
xuu = (0, 0, guu√), xuv = (0, 0, guv ), xvv = √(0, 0, gvv ). It follows
√ that L =
xuu .N = guu / D, and similarly M = √ guv / D, and N = gvv / D. √ For the
given function g, we have that L = 2/ D, M = 0, and N = 2/ D, where
D = 1 + 4u2 + 4v 2 . Hence (LN − M 2 ) = 4/D > 0, so that K = (LN −
M 2 )/(EG − F 2 ) > 0.
5.2. Finding E, F and G is straightforward. Further calculations show that

N = (−2u, 2v, −D + 2)/D, where D = 1 + u2 + v 2 . Since xuu = (−2u, 2v, 2),
xuv = (2v, 2u, 0), and xvv = (2u, −2v, −2), more calculations show that L =
xuu .N = 2, and similarly, M = 0 and N = −2. That H = 0 follows immediately
from Lemma 2 of §5.4.
5.3. The coefficients of the first fundamental form are quickly found to be
E = G = cosh2 v, F = 0. Also, N = (− sin u, cos u, − sinh v)/ cosh v, while
xuu = (− sinh v cos u, − sinh v sin u, 0) ,
xuv = (− cosh v sin u, − cosh v cos u, 0) ,

xvv = (sinh v cos u, sinh v sin u, 0) .
Hence L = xuu .N = 0, and similarly M = 1 and N = 0. Hence K = (LN −
M 2 )/(EG − F 2 ) = − cosh−4 v, and, using Lemma 2 of §5.4, we find that H =
0. It then follows from Lemma 2 of §5.5 that the principal curvatures are
± cosh−2 v.
5.4. We found in Example 1 of §3.2 that
E = 1 + gu 2 , F = gu gv , G = 1 + gv 2 ,
while L, M and N were found in Exercise 5.1. It follows that K = (LN −

M 2 )/(EG − F 2 ) = (guu gvv − guv 2 )/D2 , where D = 1 + gu 2 + gv 2 . Also, from
(5.19),
gvv (1 + gu 2 ) − 2gu gv guv + guu (1 + gv 2 )
H= .
2D3/2
5.5. (a) Here, xu = α0 + vα00 , xv = α0 , so that N is the unit vector in

direction α00 × α0 . This is independent of v.
(b) Here, N is the unit vector in direction α0 × α.
(c) Here, N is the unit vector in direction α0 × e.
5.6. Let x(u, v) = α(u) + vβ(u) sweep out a ruled surface as described
in §3.6. Then xvv = 0, so that N v .xv = −N .xvv = −N = 0. Hence K =
−M 2 /(EG − F 2 ) so that K = 0 if and only if M = 0. Also, N v .xu = −M , so
that N v = 0 if and only if M = 0.
5.7. Differentiating, we find that if X is a tangent vector to S, then

df λ (X) = X + λdN (X).
(i) Since dN maps vectors tangential to S to vectors tangential to S, we see
that N λ f λ = N .
(ii) It follows that, if X is tangential to S, then dN λ (df λ (X)) = dN (X). In
particular, if dN (X) = −κX then df λ (X) = (1 − λκ)X, so that
−κX = dN (X) = dN λ df λ (X) = (1 − λκ)dN λ (X) ,


(iii) Let κ1 and κ2 be the principal curvatures of S. Then, since K λ is the
product of the principal curvatures of S λ , we have that
κ1 κ2 κ1 κ2
Kλ = = ,
(1 − λκ1 )(1 − λκ2 ) 1 − λ(κ1 + κ2 ) + λ2 κ1 κ2
which leads to the desired expression for K λ . The expression for H λ is obtained
in a similar manner.
(iv) This is a direct substitution.
5.8. Corollary 5 of §5.5 says that K = −f 00 /f .

(i) Here, f 00 + f = 0, so, replacing v by v + c for a suitable constant c, we can
take f (v) = A cos v for a positive constant A and −π/2 < v < π/2. However,
since |α0 | = 1 we need |f 0 (v)| < 1 (since K = 0 if g 0 (v) = 0). Hence the domain
of the generating curve is as stated in the exercise.
Having determined f (v), then g(v) is given by the indefinite integral
Z q
2
g(v) = 1 − f 0 (v) dv .
Figure 8 gives sketches of the requested generating curves. Note that the
Figure 8: For the solution to Exercise 5.8(i)
case A = 1 gives the unit sphere.

(ii) Here f 00 = 0, so that f (v) = Av + B for suitable constants A and B, and we
need −1 ≤ A ≤ 1. If A = 0 then B > 0, g(v) = ±v+C, and we√obtain a cylinder.
If A 6= 0, we may assume that B = 0. In this case, g(v) = ±v 1 − A2 + C, and
hence S is a plane if A = 1 and a cone otherwise.
(iii) Here, f 00 − f = 0, so that f (v) = λev + µe−v .

If λ and µ are both non-zero and have the same sign (which must be positive),
then, replacing v by v +c for a suitable constant c, we may write f (v) = A cosh v
for some positive constant A. Then the domain of the generating curve is
(−v0 , v0 ) where A sinh v0 = 1.
If one of λ and µ is zero, we may assume that f (v) = e−v , v > 0. The
corresponding surface is the pseudosphere.
If λ and µ are both non-zero and have opposite signs, then, replacing v by
±v + c for a suitable constant c, we may write f (v) = B sinh v, v > 0, for some
positive constant B. This time the domain of the generating curve is (0, v0 )
where B cosh v0 = 1. In particular, 0 < B < 1.
Figure 9 gives sketches of the requested generating curves.
Figure 9: For the solution to Exercise 5.8(iii)
5.9. We may find κ1 and κ2 by putting f (v) = v in (5.23), and the expres-
sion for K = κ1 κ2 is then immediate. In particular, we note that the sign of K
is the same as the sign of g 0 g 00 . The given surface
is the surface of revolution ob-
tained by rotating the curve v, 0, 1/(1 + v 2 ) , v > 0, about the z-axis (together
with a pole at (0, 0, 1)) and, from working out the derivatives, it follows that √
the sign of K is equal to√the sign of 1 − 3v 2 . So, K > 0 when 0 < v < 1/ 3,
and K < 0 when v > 1/ 3. At the pole, the curvature is obtained by taking
the limit as v → 0, and this gives curvature 4 at the pole.
Figure 10 gives a sketch of the generating curve and its reflection in the
z-axis, and indicates those parts of the curve which give points of positive Gaus-
sian curvature and points of negative Gaussian curvature on the corresponding

surface of revolution.
5.10. This is a short calculation using the formula for H given in Proposition
4 of §5.6.
5.11. It follows from the formulae in Example 6 of §5.6 that the torus of
revolution has points where H = 0 if and only if we can find a value of v such
that 2b cos v = −a. Such a v exists if and only if a ≤ 2b.
5.12. Parametrise S by taking

x(u, v) = (af (v) cos u, bf (v) sin u, v) .
Calculations show that N = (b cos u, a sin u, −abf 0 )/D, where D is the modulus
of the numerator. Then L = xuu .N = −abf /D, M = 0, N = −abf 00 /D. Hence
K is a positive scalar multiple of −a2 b2 f f 00 , and the result follows.
5.13. (i) If we parametrise the (upper half of the) cone as usual by taking
x(u, v) = (v cos u, v sin u, v) , v>0,
√
then calculations give that E = v 2 , F = 0, G = 2, while L = −v/ 2, M = N =
t 2t
0. At α(t) we have u = t and √ v = e , so that, at α(t), we have E = e , F = 0
and G = √ 2, while L = −et / 2, M = N = 0. Formula (5.33) now shows that
κn = −(3 2et )−1 .
(ii) Methods described in Chapter 1 show that if t is the unit tangent vector
to α then
dt 1
= t (− sin t − cos t, cos t − sin t, 0) .
ds 3e
√
Usual methods show that the unit normal to the cone at α(t) is (cos t, sin t, −1)/ 2,
and a calculation using the formula for κn given in the exercise yields the same
answer as that found in (i).
5.14. Let x(u, v) be a local parametrisation whose image is V . Then

(x.N )u = xu .N = 0, and, similarly, (x.N )v = 0. Hence x.N is constant,
and V is contained in a plane.
5.15. Here E = 1 + 4u2 , F = 4uv, 2

√ and G = 1 + 4v . In the √ solution
to Exercise 5.1 we found that L = 2/ D, M = 0, and N = 2/ D, where
D = 1 + 4u2 + 4v 2 . Hence, using (5.41), we may show that the given curve is a
line of curvature if and only if the given equation holds.
So, either uu0 + vv 0 = 0 or u0 v − uv 0 = 0. The first integrates up to give
u + v 2 = constant and the second gives u/v = constant.
2
In fact, S is a surface of revolution, and the families of curves we have found

are the parallels and the meridians. We noted near the start of §5.9 that these
are indeed the lines of curvature on a surface of revolution.
5.16. Let x(u, v) be a local parametrisation of S whose image contains no

umbilics and whose coordinate curves are lines of curvature. Lemma 1 of §5.6
shows that F = M = 0. It follows that K = LN/EG, so we may assume that

L = 0 and N 6= 0. It then follows that N u = 0 and N v = −(N/G)xv . For each
fixed v0 , consider the coordinate curve α(u) = x(u, v0 ). Then N is constant
along α, and we now show that xuu is a scalar multiple of xu which will show
that α is a straight line. To do this, we note that xuu .N = −L = 0, while
xuu .xv is zero if and only if xuu .N v = 0. However, this latter follows because
xuu .N v = (xu .N v )u − xu .N uv = 0.
5.17. We found E, F , G; L, M and N for Enneper’s surface in Exercise 5.2.

(i) Since F = M = 0, the principal curvatures are L/E and N/G.
(ii) This is immediate from Lemma 1 of §5.6.
(iii) It follows from (5.43) that the asymptotic curves are given by u02 − v 02 = 0,
that is, u ± v = constant.
Fix k1 and k2 and consider the curves u + v = k1 and u − v = k2 . These
intersect at the point corresponding to u = (k1 + k2 )/2, v = (k1 − k2 )/2.
5.18. In the solution to Exercise 5.3, we found that L = N = 0 and

M = 1. Hence the asymptotic curves satisfy u0 v 0 = 0, and so are the families of
coordinate curves. These are the z-axis, the lines of the ruling of the helicoid,
and the helices obtained by intersecting the helicoid with the cylinders whose
axis is the z-axis.
5.19. Routine calculations show that E = sech2 v, F = 0 and G = tanh2 v.

We may use (5.22) to show that L = −N = −sech v tanh v, and M = 0. It
now follows from (5.43) that the asymptotic curves are u ± v = constant. The
tangent vectors to the asymptotic curves are xu ∓ xv , and the angle θ between
these is given by
(xu + xv ).(xu − xv ) sech2 v − tanh2 v

cos θ = = = 2 sech2 v − 1 .
|xu + xv | |xu − xv | sech2 v + tanh2 v
5.20. A line has zero curvature, so the normal curvature of a line on a

surface is also zero. It follows from Proposition 4 of §5.7 that the principal
curvatures at a point of a ruled surface cannot both have the same sign. Hence
K ≤ 0.
5.21. Let α be an asymptotic curve on a surface S, and assume that α is

parametrised by arc length. Then α00 .N = 0, so that the principal normal n of
α is also orthogonal to N . Hence, if b is the binormal of α then (by choosing
the correct sign for N ), we have that b = N . Hence b0 = N 0 = τ n, so, if θ is the
angle between α0 and the principal direction with principal curvature κ1 , then,
by (5.35), τ 2 = |N 0 |2 = κ1 2 cos2 θ + κ2 2 sin2 θ. Since α is an asymptotic curve,
0 = II(α0 ) = κ1 cos2 θ + κ2 sin2 θ, so that τ 2 = −κ1 κ2 (sin2 θ + cos2 θ) = −K.
5.22. Let e1 and e2 be unit vectors in the two principal directions, and let
κ1 and κ2 be the corresponding principal curvatures. Then cos θe1 + sin θe2 is
in an asymptotic direction if and only if κ1 cos2 θ + κ2 sin2 θ = 0, which happens
if and only if tan2 θ = −κ1 /κ2 . This implies that cos θe1 + sin θe2 is in an
asymptotic direction if and only if cos(−θ)e1 + sin(−θ)e2 is in an asymptotic
direction, so the result follows.
Since lines of curvature intersect orthogonally, the equivalence of the three
conditions follows because θ = π/4 if and only if κ1 = −κ2 .
5.23. The vector α00 is a linear combination of α0 and n. Hence n.N = 0

if and only if α00 .N = 0. This holds if and only if II(α0 ) = 0.
5.24. Let α(t) be a parametrisation of the regular curve C. Then
N 1 .N 2 = const ⇐⇒ N 1 0 .N 2 + N 1 .N 2 0 = 0 ⇐⇒ κα0 .N 2 + N 1 .N 2 0 = 0 .
However, α0 is orthogonal to N 2 , and N 1 .N 2 0 = 0 if and only if N 2 0 is

orthogonal to both N 1 and N 2 . This happens if and only if N 2 0 is a scalar
multiple of α0 .
5.25.√ Using the notation

√ of the solution
√ to Exercise 5.1, we have that
L = 6u/ D, M = −6v/ D, and N = −6u/ D. Hence LN − M 2 < 0 except
at u = v = 0, in which case LN − M 2 = 0.

5.26. Let α(t) = x(t), y(t), z(t) be a regular curve on S. It follows from
(5.34) that the sign of the normal curvature κn of α is the same as that of
α00 .grad f .
Differentiating x4 + y 4 + z 4 = 1 with respect to t we find that
x3 x0 + y 3 y 0 + z 3 z 0 = 0 ,
and, differentiating again,

2 2 2
3(x2 x0 + y 2 y 0 + z 2 z 0 ) + x3 x00 + y 3 y 00 + z 3 z 00 = 0 .
Hence,
2 2 2
α00 .grad f = 4(x3 x00 + y 3 y 00 + z 3 z 00 ) = −12(x2 x0 + y 2 y 0 + z 2 z 0 ) .
Thus κn is always negative unless:

(i) any two of x, y, z are zero, in which case all normal curvatures are zero;
(ii) exactly one of x, y, z is zero. If x = 0, then the rate of change of N in
direction (1, 0, 0) is zero, but, for instance, the rate of change of N in direction
(0, −z 3 , y 3 ) is non-zero.
The result now follows.
5.27. For a general local parametrisation x(u, v) we have
(u2 xuu + 2uvxuv + v 2 xvv ).N = u2 L + 2uvM + v 2 N

= II(uxu + vxv ) .
The result follows by applying this to the given local parametrisation at (0, 0).
5.28. Parametrise S by setting
x(u, v) = (v cos u, v sin u, v 2 ) , v > 0 , −π < u < π .
Then E = v 2 , F = 0, G = 1 + 4v 2 . Also, using Exercise 5.9, K = 4(1 + 4v 2 )−2 .

Hence, ZZ Z π Z ∞
|K|dA = 4v(1 + 4v 2 )−3/2 du dv = 2π.
S −π 0
This is equal to the area of the lower hemisphere of S 2 (1), which, as found in
the solution to Exercise 4.3, is the image of the Gauss map of the paraboloid of
revolution.
5.29. We use the argument and notation of the proof of Theorem 4 of

§5.12. If α(t) is a curve on S with α(0) = p0 we have that α0 (0).q0 = 0 and
α00 (0).q0 ≤ 0. It follows that N (p0 ) = q0 , and all the normal curvatures of S at
p0 are non-positive. Hence p0 isn’t a hyperbolic point.
Chapter 6
6.1. No; for instance, there is a local isometry from the plane to the cylinder.
6.2. The expressions for the Christoffel symbols follow immediately from
equations (6.4). The Gauss formula (6.9) now shows that GK = −Guu /2 +
Gu 2 /4G, from which the result follows. The formula for the Gaussian curvature
may also be obtained using equation (6.10) for K in orthogonal coordinates.
6.3. Equations (6.4) may be used to show that Γ122 = Γ222 = 0, and that
Γ112 = −Γ212 = v/(1 + v 2 ). The Gauss formula (6.9) may then be used to obtain
the desired expression for K.
Alternatively, since the Gauss formula is difficult to remember(!), we may
mimic the method of proof of that formula to find the expression for K as
follows.
LN = LN .N N = (xuu − Γ111 xu − Γ211 xv ) .(xvv − Γ122 xu − Γ222 xv )

= (xuu − Γ111 xu − Γ211 xv ) .xvv = xuu .xvv ,
while M 2 = M N .M N , which, after a little work gives
v2
M 2 = xuv .xuv − .
1 + v2
Hence
v2
LN − M 2 = xuu .xvv − xuv .xuv + .
1 + v2
Then, using (6.8), we find that LN −M 2 = −1/(1+v 2 ), so obtaining the desired
expression for K.
6.4. This is very quick using the formula for the Gaussian curvature of a
Tchebycheff parametrisation given in Example 2 of §6.2.
6.5. Let F = cos θ. Then Fu = −θu sin θ, Fv = −θv sin θ, and Fuv =
−θuv sin θ − θu θv cos θ. The required formula now follows from the formula for
the Gaussian curvature of a Tchebycheff parametrisation given in Example 2 of
§6.2.
6.6. The given parametrisation

√ x of the helicoid has E = 1 + v 2 , F = 0,
G = 1, L = N = 0, M = 1/ 1 + v . Hence K = −(1 + v 2 )−2 . The formula
2
given in Exercise 5.9 may be used to show that if K̃ is the Gaussian curvature
of S̃ then K̃ = −(1 + v 2 )−2 also. However, the given map isn’t an isometry
since Ẽ = v 2 + v −2 6= E.
6.7. If L = M = N = 0, then N is constant and S is an open subset of a

plane.
(i) Equations (6.4) show that all the Christoffel symbols are zero, so that
xuu = LN , xuv = M N , and xvv = N N . So, from (6.7) and (6.8), we see that
LN − M 2 = 0. Hence K = 0.
(ii) If L = N = 0 then M = 0 also. Hence, assuming that S is not a plane, we
may assume that L 6= 0.
From now on, we assume that L 6= 0.
(iii) Then au = LM N −LM N = 0 and similarly av = 0. Since S is connected,
it follows that a is a non-zero constant vector.
(iv) That a.a = L(EN − 2F M + GL) is quick using LN − M 2 = 0.
(v) We first show that bu .xu = 0. In fact
xu .a
bu .xu = xu .xu − a.xu + cN u .xu
a.a
(EM − F L)2
=E− − cL ,
L(EN − 2F M + GL)
which, on substituting for c and simplifying, is seen to be zero.
Similar calculations show that bu .xv = bv .xu = bv .xv = bu .N = bv .N = 0,
so that bu = bv = 0. It now follows that b is constant.
(vi) Just show that a.b = 0.
(vii) It now follows from part (v) that
(b1 , b2 , 0) = (x1 , x2 , x3 ) − (0, 0, x3 ) + cN ,
so that |(b1 − x1 , b2 − x2 , 0)|2 = c2 , and the result follows.
6.8. It follows from Lemma 1 of §5.6 that the coordinate curves are lines of
curvature if and only if F = M = 0. In this case κ1 = L/E and κ2 = N/G,
and the first Codazzi-Mainardi equation (6.16) becomes 2Lv = Ev (κ1 + κ2 ).
However, Lv = (Eκ1 )v = Ev κ1 + E(κ1 )v , so we see that the first Codazzi-
Mainardi equation holds if and only if the first of the given equations holds.
The equivalence of the other pair of equations is proved similarly (or simply
interchange u and v).
6.9. Using the equations from Exercise 6.8, we see that the first Codazzi-
Mainardi equation gives
(L − N )v = {E(κ1 − κ2 )}v = Ev (κ1 − κ2 ) + E(κ1 − κ2 )v
= −2E(κ1 )v + E(κ1 − κ2 )v = −E(κ1 + κ2 )v .
Hence, if κ1 + κ2 is constant then (L − N )v = 0. Interchanging u and v shows
that (L − N )u = 0 also, so that L − N is constant.
6.10. We saw in Example 2 of §6.2 that if a surface S has a Tchebycheff

parametrisation with E = G = 1 and F = cos θ, with θ satisfying the sine-
Gordon equation, then S has constant Gaussian curvature K = −1. If we also
want the coordinate curves of x to be asymptotic curves then, by (5.43), we
need to have L = N = 0. In order to satisfy (EG − F 2 )K = LN − M 2 with
K = −1 we also need to have M = ± sin θ, and, reflecting if necessary, we
assume M = sin θ (using the unit normal in direction xu × xv ).
We now show we can apply Bonnet’s theorem for the above values of E, F ,
G, L, M , N . Example 2 of §6.2 showed that the right hand side of (6.9) is
equal to −θuv sin θ, so (6.9) reduces to 1 − cos2 θ = θuv sin θ, which holds since
θuv = sin θ.
We see from the first pair of equations of (6.5) that Γ111 sin θ = θu cos θ so
the right hand side of (6.14) reduces to −θu cos θ, as does the left hand side. A
similar argument (or just reverse the roles of u and v) shows that (6.15) also
holds. Hence we can apply Bonnet’s theorem to prove the required result.
6.11. Let λ and µ be the principal curvatures of a compact connected surface

S in R3 with constant mean curvature H and everywhere positive Gaussian
curvature K. Since S is compact, the globally defined continuous function
(λ − µ)2 = 4(H 2 − K) attains its maximum at some point p0 ∈ S. If p0 is an
umbilic point, then every point of S is an umbilic point and S is a sphere. So,
assume that, at p0 , λ > µ > 0. Then λ − µ has a maximum at p0 and, since
λ + µ is constant, λ has a maximum at p0 and µ has a minimum. Hence, at p0 ,
λu = λv = 0, λuu ≤ 0, λvv ≤ 0, and µu = µv = 0, µuu ≥ 0, µvv ≥ 0. If we
differentiate (6.19) at p0 , we find that, at p0 ,
Evv (λ − µ) = −2Eλvv ≥ 0 , and Guu (λ − µ) = 2Gµuu ≥ 0 .
Continuing as in the proof of Liebmann’s theorem (Theorem 1 of §6.4), we find
that K(p0 ) ≤ 0, which contradicts one of our initial assumptions.
6.12. If K > 0 then Theorem 5 of §6.4 shows that S is a sphere. Hilbert’s

theorem shows that K can’t be a negative constant, so we now assume that
K = 0. In this case, by Massey’s theorem, S is a generalised cylinder, and so
has a parametrisation of the form
y(u, v) = α(u) + ve ,
where α0 is never a scalar multiple of the constant non-zero vector e. As-
suming without loss of generality that e = (0, 0, 1), we see that if α(u) =

x(u), y(u), z(u) then the curve β(u) = (x(u), y(u), 0) is also regular, and, by
replacing v with v + z(u), we may assume that
y(u, v) = β(u) + v(0, 0, 1) .
Now assume that β(u) is parametrised by arc length, and let n(u) = (−y 0 (u), x0 (u), 0)
be its principal normal. If κ is the curvature of β and N is the unit normal of
S, then y u = β 0 , y v = (0, 0, 1), N = −n, y uu = β 00 = κn, y uv = y vv = 0.
Hence E = 1, F = 0, G=1, L = −κ, M = 0, N = 0. Since H is constant, so
is κ. Hence the trace of β is a line or a circle, so that S is a plane or a round
cylinder.
Chapter 7
7.1. We note that dα/ds = α0 /|α0 |, so that
d2 α α00

1 d 1
= 02+ 0 α0 .
ds2 |α | |α | dt |α0 |
It follows from (7.1) that κg = (d2 α/ds2 ).(N × dα/ds), and the desired formula
follows if we substitute the expressions obtained above for dα/ds and d2 α/ds2 .
7.2. The Gauss map of S 2 (1) is the identity map, so if α is a regular curve
on S 2 (1) then the first formula of §7.1 shows that α00 = κg α × α0 + κn α. Hence,
if κg = c then
v 0 = α × α00 + cα0 = −cα0 + cα0 = 0 .
Also, α.v = c, so that α lies on the plane√perpendicular to v whose perpen-
dicular distance from the origin is c/|v| = c/ 1 + c2 .
7.3. This may be done using a very similar method to that employed in
Example 2 of §7.1. Specifically, α0 = r(− cos v0 sin t, cos v0 cos t, 0), so that
|α0 | = r cos v0 and α00 = −r(cos v0 cos t, cos v0 sin t, 0). Also, N = (1/r)α so a
short calculation using (7.2) shows that κg = (1/r) tan v0 as required.
7.4. Parametrise Cv as usual by setting

α(u) = f (v) cos u, f (v) sin u, g(v) .
Then find α0 , α00 and N (using, for instance, Example 2 of §4.1). Then use
(7.2) to find κg .
7.5. (a) Note that (g 0 , 0, −f 0 ) is orthogonal to S at (f, 0, g), and the line
through (f, 0, g) in direction (g 0 , 0, −f 0 ) is given by (f, 0, g) + λ(g 0 , 0, −f 0 ). This
meets the z-axis when λ = −f /g 0 , at which point the z-coordinate is given by
h = g + f f 0 /g 0 .
(b) A short calculation shows that |(0, 0, h) − (f cos u, f sin u, g)|2 = r2 , so
that Cv lies on S̃. Also, using (a), the outward unit normal Ñ to S̃ at
(f cos u, f sin u, g) is given by
1 f
Ñ = (f cos u, f sin u, g − h) = 0 (g 0 cos u, g 0 sin u, −f 0 ) .
r gr
The unit normal N of S is given by
(g 0 cos u, g 0 sin u, −f 0 )
N= p ,
f 0 2 + g0 2
so that
Ñ = N , where = g 0 /|g 0 | .
which proves (b).
(c) It follows from Exercise 7.3 that if −π/2 < µ < π/2 is such that tan µ =
(g − h)/f then
κ̃g = (1/r) tan µ .
However,
2 2
r2 = f 2 + (g − h)2 = f 2 + (f f 0 /g 0 )2 = (f 0 + g 0 )(f /g 0 )2 ,
while
tan µ = (g − h)/f = −f 0 /g 0 ,
from which the given formula for κ̃g follows.
(d) It follows from Exercise 7.4 that κ̃g = κg , which is in accordance with
Proposition 3 of §7.1 since Ñ = N .
7.6. That α lies on the given cone is easy to check. Also,

√
α0 = et (cos t − sin t, sin t + cos t, 1) , so that |α0 | = 3et ,
while
α00 = et (−2 sin t, 2 cos t, 1) .
The unit normal N to the cone at (x, y, z) is in direction grad (x2 + y 2 − z 2 ),

so that √
N α(t) = (cos t, sin t, −1)/ 2 .
A calculation
√ using formula (7.2) for geodesic curvature now shows that
κg = 1/( 6et ).
7.7. Let α(u) = (u, v0 , u2 − v0 2 , 2uv0 ). A quick check shows that x(u, v)
is an orthogonal (in fact, isothermal) parametrisation, so that X = xv /|xv | is
a unit vector tangential to S and orthogonal to α0 . Formula (7.3) for geodesic
curvature quickly gives that
κg = −4v0 (1 + 4u2 + 4v0 2 )−3/2 .

7.8. (a) Let α(s)

√ = x(u(s), v0 ) be a parametrisation of v = v0 by arc length.
Then du/ds = 1/ E, and, since xu and xv are orthogonal, N × dα/ds =
xv /|xv |. Hence, using (7.1) (or (7.3) for a surface in Rn ),
d2 α xv

1 d 1 xv 1 Ev
κg = . = √ √ x u .√ = √ xuu .xv = − √ ,
ds2 |xv | E du E G E G 2E G
which leads to the given answer.
(b) Substitute the expressions for E, F and G in terms of f and g.
(c) Using (a), we
√ see that the geodesic curvature of the coordinate curve u =
constant is (1/2 E)(log G)u , and the result follows.
(d) The standard parametrisation of a surface of revolution has F = 0 and
Gu = 0.
7.9. Here, xv = b, and xs = α0 + vb0 . Hence α00 .xv = κn.b = 0, while,

when v = 0, α00 .xs = α00 .α0 = 0. Hence α00 is orthogonal to S and the result
follows.
7.10. (a) Let Ñ be the (constant) unit normal to P . Then Ñ and α0

are non-zero orthogonal vectors in Tα(t) S, and hence span Tα(t) S. However,
0
α00 .Ñ = α0 .Ñ = 0, while, since α is parametrised proportional to arc length,
α00 .α0 = 0. It follows that α00 is orthogonal to S, and so α is a geodesic on S.
(b) Let f (x, y, z) = x4 + y 6 + z 8 . The unit normal to S at (x, y, z) is in
direction grad f = (4x3 , 6y 5 , 8z 7 ). If P is the yz-plane, then Ñ = (1, 0, 0) which
is orthogonal to N (0, y, z). The result for P now follows from (a), and the other
coordinate planes are handled similarly.
7.11. (a) Using the standard notation, the assumptions imply that N 0 =
λα and α00 = µN for some scalar functions λ and µ. It follows that (N ×α0 )0 =
0
0 so that N × α0 is constant. It now follows that α.(N × α0 ) is also constant,

so that α lies on a plane with normal N × α0 and the result follows.
(b) This time the assumptions imply that, if Ñ is the (constant) unit normal to
P , then α0 .Ñ = 0 and α00 = µN with µ never zero. Then 0 = α00 .Ñ = µN .Ñ ,
which implies that N .Ñ = 0, so that N 0 .Ñ = 0. Since Ñ is tangential to S
and orthogonal to α0 , it follows that N 0 is a scalar multiple of α0 and the result
follows.
7.12. First show that xu is orthogonal to xv .

(i) and (iii) Let γ(t) = x(π, t) = π, 1 + cosh t, 4 sinh(t/2) /2. This is the
parabola x = π/2, y − 1 = z 2 /4.
Since xu and xv are orthogonal, to show that (when parametrised propor-
tional to arc length) the coordinate curve u = π is a geodesic, it is enough to
prove that γ 00 (t) is orthogonal to xu γ(t) . This is a routine calculation.
(ii) and (iii) This time we take γ(t) = (t − sin t, 1 − cos t, 0)/2. This is a cycloid
in the xy-plane, and it is quick to show that γ 00 (t) is orthogonal to xv γ(t) .

7.13. Every planar point is an umbilic. If p ∈ S is not a planar point, then,

by Exercise 7.11, there are an infinite number of lines of curvature through p.
This implies that p is also an umbilic, so the result follows from Theorem 1 of
§5.8.
7.14. Assume that α(t) is a closed geodesic on a helicoid, and let it attain
its maximum height at α(t0 ), say. Then α is tangential to the straight line
on the helicoid through α(t0 ), so, by uniqueness of geodesics, the trace of α(t)
must be this line. Hence α(t) is not a closed curve.
7.15. (a) The segment of the unit circle which lies in the sector has length
φ, and when we bend the sector to form the cone, this segment maps to the
Figure 11: For the solution to Exercise 7.15(a)
2 2 2
parallel on the cone
p which is a circle of radius
p r where r + β r = 1 (Figure
2 2
11). Hence r = 1/ 1 + β so that φ = 2π/ 1 + β , and the result follows.
(b),(c) and (d) Cut the plane along a half-line to the origin, and then form
a cone by sliding the edges of the cut past each other until the line segments
forming an angle φ at the origin line up to give a generator of the cone. A line `
in the plane not intersecting the cut gives a maximal geodesic on the cone, and
two points p1 and p2 on ` give the same point on the geodesic if and only if they
are equidistant from the origin 0 and the angle ∠p1 0p2 is an integer multiple of
φ.
If φ ≥ π, then no geodesic has self-intersections. If π > φ ≥ π/2 then
every geodesic which is not a meridian has exactly one self-intersection. If
π/n > φ ≥ π/(n + 1) for some integer n then every geodesic which is not a
meridian has exactly n self-intersections. Figure 12 illustrates this for n = 2.
In this figure, the rays from the origin drawn in a solid line map to the same
meridian of the cone, and those drawn in a dashed line map to the opposite
meridian. The points p1 and p2 (resp. q1 and q2 ) give a point where the
geodesic has a self-intersection.
The result now follows from (a).
Figure 12: The horizontal line gives a geodesic on the cone (Exercise 7.15)
7.16. (a) A short calculation using equation (7.11) shows that (u0 /v 2 )0 = 0
if and only if (7.11) holds. We also note that, in this case, u00 = 2u0 v 0 /v.

(b) The curve u(t), v(t) in H is parametrised proportional to arc length if and
only if
0 2 0 2
u v
+ = const . (1)
v v
However, a calculation shows that if u00 = 2u0 v 0 /v then
0 2 0 2 ! 2 2
!
d u v 0 u0 00 v0
+ =v +v − .
dt v v v v
Hence (1) holds if and only if either (7.12) holds or v is constant. However, if v
is constant, then (7.12) shows that u is also constant.
(c) The curves (u0 , v0 ekt ) are geodesics. If u0 = cv 2 with c 6= 0, and if u02 +v 02 =
kv 2 with k 6= 0, then √
v0 k − c2 v 2
= ,
u0 |c|v
so that Z
v dv
u − u0 = |c| √ ,
k − c2 v 2
which integrates up to give the desired answer.

7.17. Let α(s) = u(s), v(s) be a curve on H parametrised by arc length
s, and let 0 denote differentiation with respect to s. Then
2 2
u0 + v 0 = v 2 , (2)
while (7.14) gives that

κg = (2u0 /v) + φ0 . (3)
However, cos φ = u0 /v so a short calculation gives that
u0 v 0 − u00 v
φ0 = .
vv 0
So, if κg is equal to a constant, k, say, then (3) quickly gives that
u0 v 0
u00 − 2 + kv 0 = 0 . (4)
v
We now solve (2) and (4). We begin by noting that (4) integrates to give
u0 = v 2 c + (k/v) for some constant c, so, from (2), we deduce that
2
2 v 0 (cv + k)2
u0 = .
1 − (cv + k)2
Taking square roots and integrating, we find that the curves in H of constant
geodesic curvature are the intersection with the upper half-plane of (Euclidean)
circles in R2 (together with the lines u = constant).
7.18. The map f (u, v) = (cos u, sin u, cos v, sin v) is a local isometry of the
plane onto the flat torus, so the geodesics on the latter are the images under f
of straight lines in the plane. The geodesics through (1, 0, 1, 0) come from lines
through the origin, and so take the form t 7→ (cos kv, sin kv, cos v, sin v). This
returns to (1, 0, 1, 0) (in which case the geodesic closes up) if and only if k is a
rational number.
7.19. Use either Exercise 7.8 or (7.14) to see that

Ev Gu
νg = − √ , µg = √ .
2E G 2G E

Hence, from (7.14), if α(s) = x u(s), v(s) , then
√ du √ dv dφ
κg = νg E + µg G + .
ds ds ds
However,
xu du E √ du
cos φ = α0 . = √ = E ,
|xu | ds E ds
√
and, similarly, sin φ = Gdv/ds. The result now follows.
7.20. Here we are rotating (v, 0, βv) about the z-axis, so Clairaut’s relation
gives v cos θ = (1/β) cos θ0 . In particular, z(t) = βv ≥ cos θ0 . Note that this
result may also be shown using the ideas of Exercise 7.15 - the parallels of
the cone come from circles with centre at the origin of the plane; now use the
fact that if ` is a line in the plane cutting the unit circle at angle θ0 then the
perpendicular distance of ` from the origin is cos θ0 .
7.21. Here we are rotating (cosh v, 0, v) about the z-axis. Let x(u,
v) be the
standard parametrisation of the surface, and let α(t) = x u(t), v(t) be a closed
geodesic. Assume that cosh v(t) attains its maximum at t = t0 . Then |v(t)| also
attains its maximum at t = t0 . Hence v 0 (t0 ) = 0 so that α is tangential to
the parallel at α(t0 ). Hence from Corollary 3 of §7.6, cosh v(t) also attains its
minimum at t0 , so that cosh v(t), and hence v(t), is constant and α is a parallel.
However, Lemma 1 of §7.6 shows that the parallel v = v0 is a geodesic if and
only if sinh v0 = 0, so that the only closed geodesic is the parallel v = 0 (when
parametrised proportional to arc length).
7.22. We may obtain this surface by rotating (cosh v, 0, sinh v) about the
z-axis. The solution is now word-for-word the same as for Exercise 7.21.
7.23. Here we are rotating (v, 0, 1/v 2 ) about the z-axis. Arguing
as in
the solution to Exercise 7.21, we see that if α(t) = x u(t), v(t) is a closed
geodesic and if v(t) attains its maximum at t = t0 , then v 0 (t0 ) = 0 so that α
is tangential to the parallel at α(t0 ). It follows from Corollary 3 of §7.6 that α
must be a parallel, but, for this surface, Lemma 1 of §7.6 shows that no parallels
are geodesics.

7.24. As usual, we assume that S is obtained by rotating f (v), 0,
g(v) ,
f (v) > 0 ∀v, about the z-axis. If α(t) = x u(t), v(t) , then f v(t0 ) is the
maximum value of f v(t) so that fv v(t0 ) v 0 (t0 ) = 0. If fv v(t0 ) = 0 then, by
Lemma 1 of §7.6, the parallel through α(t0 ) is a geodesic (when parametrised
proportional to arc length). Now assume that v 0 (t0 ) = 0. Then α is tangential
to the parallel at α(t0 ), so Corollary 3 of §7.6 shows that α is a parallel (which
is thus a geodesic).
The result does not hold if “maximum” is replaced by “minimum”. Take,
for example, any non-equatorial great sphere on S 2 (1) or any geodesic on the
cone which is not a meridian.

7.25. Clairaut’s relation says that, along the geodesic α(t) = x u(t), v(t) ,
we have
(a + b cos v) cos θ = (a + b) cos θ0 .
If α also intersects the parallel v = π at an angle θ1 , say, then
(a − b) cos θ1 = (a + b) cos θ0 .
For this to have a solution θ1 we need that cos θ0 ≤ (a − b)/(a + b). However,
if we have equality, then cos θ1 = 1 so that the geodesic α is tangential to the
parallel v = π. Since this parallel is itself a geodesic, the uniqueness theorem
for geodesics would imply that α is also the parallel v = π, which gives a
contradiction.
Conversely, assume that for some > 0 we have that
cos θ0 = (a − b − )/(a + b) .
2
We also assume that α is parametrised
√ by arc length, which implies that u0 E +
02 0
v G = 1 and that cos θ = u E. Since G = b2 , these equations imply that
2
v 0 b2 = 1 − cos2 θ . (5)
Clairaut’s relation gives that
(a + b cos v) cos θ = (a + b) cos θ0 = a − b − ,
so that
a−b−
cos θ = ≤1− ,
a + b cos v a−b
so, from (5),
2
2
v 0 b2 ≥ 1 − 1 − ≥1− 1− = .
a−b a−b a−b
It follows that v 0 is bounded away from zero, so that |v| tends to infinity as
|t| tends to infinity. In particular, α must cross the parallel v = π.

7.26. The generating curve f (r), 0, g(r) is the geodesic from p with ini-
tial vector (f 0 (0), 0, g 0 (0)). Since the pole isn’t a singular point it follows that
g 0 (0) = 0, so if we parametrise the generating curve by arc length, then the
generating curve is α(1,0,0) (r), the geodesic with initial vector (1, 0, 0). For each
θ, let αθ be the meridian αθ (r) = f (r) cos θ, f (r) sin θ, g(r) . Then α0θ (0) =
(cos θ, sin θ, 0) so αθ (r) = α(cos θ,sin θ,0) (r). Hence x(r, θ) = x̃(r cos θ, r sin θ) =
α(cos θ,sin θ,0) (r) = f (r) cos θ, f (r) sin θ, g(r) , as required.
7.27. (a) We have that xu = βv 0 . Since xv (0, 0) = α0 (0) we see that

xu (0, 0) and xv (0, 0) are orthogonal unit vectors. It follows from Theorem
3 of §2.5 that x is a local parametrisation of S on a sufficiently small open
neighbourhood U of (0, 0). We also have that E = 1, while
Fu = xuu .xv + Ev /2 = β v 00 .xv + 0 = 0 .
However, F (0, v) = β v 0 (0).α0 (v) = 0, so it follows that F = 0.

Finally, G(0, v) = α0 .α0 = 1, and, since Gu = 2xuv .xv = −2xu .xvv , we see
that
Gu (0, v) = −2xu (0, v).xvv (0, v) = −2β v 0 (0).α00 (v) = 0 .
(b) In this case we have that α(v) = (cos v, sin v, 0). Also, β 0 (u) = (cos u, 0, sin u)
while β v (u) is obtained from β 0 (u) by rotation about the z-axis through angle
v. Hence
x(u, v) = (cos u cos v, cos u sin v, sin u) ,
which is the standard parametrisation of the unit sphere as surface of revolution,
but with u and v interchanged. In particular, E = 1, F = 0 and G = cos2 u.
(c) We prove Minding’s theorem by showing that E, F and G are uniquely
determined when K is a given constant. Arguing as in the proof of Minding’s √
theorem given in §7.7, it is enough to show that if K is constant and if g = G
then the solutions of guu + Kg = 0 are uniquely determined by the conditions
on G determined in (a).
CASE 1: K = 0. The general solution of guu = 0 is g(u, v) = uh1 (v) + h2 (v).
The conditions on G then imply that g(u, v) = 1.
CASE 2: K = k 2 > 0. The general solution of guu + k 2 g = 0 is g(u, v) =
h1 (v)
√ sin ku + h2 (v) cos ku. The conditions on G then imply that g(u, v) =
cos Ku.
CASE 3: K = −k 2 < 0. The general solution of guu − k 2 g = 0 is g(u, v) =
h1 (v)√sinh ku + h2 (v) cosh ku. The conditions on G then imply that g(u, v) =
cosh −Ku.
7.28. This is clear since a local isometry maps a geodesic on S1 parametrised

by arc length to a geodesic on S2 parametrised by arc length.
7.29. We use geodesic polar coordinates

R π px(r, θ).
(i) Equation (7.33) says that `(r) = −π G(r, θ) dθ. If K is constant, then
the proof of Minding’s theorem shows that:
if K = 0 then G = r2 so that `(r) = 2πr;
if K = k 2 > 0 then G = (1/k)2 sin2 kr so that `(r) = (2π/k) sin kr;

if K = −k 2 < 0 then G = (1/k)2 sinh2 kr so that `(r) = (2π/k) sinh kr.
(ii) Using the given formula for κg we see that if K = 0 then κg = ±1/r; if
K = k 2 > 0 then κg = ±k cot kr; while if K = −k 2 < 0 then κg = ±k coth kr.
(iii) Integrating the results obtained in (i), we find that if K = 0 then A(r) =
πr2 ; if K = k 2 > 0 then A(r) = 2π(1 − cos kr)/(k 2 ); while K = −k 2 < 0 then
A(r) = 2π(cosh kr − 1)/(k 2 ).
The final statement is a routine check.
7.30. Let p = (0, 0, 1) ∈ S 2 (1). The circumference of the geodesic circle in

2
S (1) radius r is 2π sin r. Hence,
2πr − L(r) 2πr − 2π sin r r − sin r 2π π

lim = lim = 2π lim = = .
r→0 r3 r→0 r3 r→0 r3 3! 3
7.31. For rolling without slipping, we have

0
Ñ .(Ñ × t̃) = (A0 N + AN 0 ).A(N × t) = −AN .A0 (N × t) + N 0 .(N × t) . (6)
If there is no twisting, then by Proposition 1 of §7.9, we have that κg = κ̃g .

Also, the equation displayed in the Exercise follows immediately from condition
(7.36) for no twisting, and equation (6) above.
Assume now that κg = κ̃g and the equation displayed in the Exercise holds.
Then, using the proof of Proposition 1 of §7.9, we find that
A0 (N × t).At = 0 .
Also, using (6) and the equation displayed in the Exercise, we have
A0 (N × t).AN = 0 ,
while A0 (N × t).A(N × t) is also zero. It now follows that A0 (N × t) = 0, so

the rolling takes place without twisting.
The surface S rolls without twisting (or slipping) on a straight line in a plane
if and only if κg = 0 and N 0 .(N × t) = 0, which happens if and only if α is a
geodesic and N 0 is a scalar multiple of t (so that α is a line of curvature). The
final statement is now immediate from Exercise 7.11.
Chapter 8
8.1. The first set has non-compact interior. The boundary of the second set
is not a disjoint union of piecewise regular simple closed curves. The third one
is not the boundary of an open set.
8.2. The interior angles at v1 , v2 , v3 and v4 are π/2, 3π/2, π/2 and 3π/2
respectively. The exterior angles are π/2, −π/2, π/2, −π/2 respectively. Figure
13 shows the exterior angles at v1 , v2 and v4 for the indicated orientation. Note
that the exterior angles at v2 and v4 are negative, which reflects the fact that
the angle of rotation from the incoming to outgoing vectors at these vertices is
in the opposite direction to that given by the orientation.
8.3. A generalisation of the Theorem of Turning Tangents is obtained by

replacing “Let T be a triangle contained in the image of a local parametrisation
x : D → S of a surface S in Rn ” by “Let R be a regular region of a surface
S in Rn which is the image of an n-gon in D under a local parametrisation
x : D → S”, and then replacing the final equation by
Z
dθ X
ds = αi − (n − 2)π ,
∂R ds
where the sum is taken over the interior angles at the vertices of ∂R.
The proof is by induction on n. It may be obtained R by dividing the n-gon
into an (n − 1)-gon and a triangle, and noting that dθ ds ds cancels over the
common edge of the (n − 1)-gon and the triangle.
8.4. (i) Two edges of T are meridians, and hence have zero geodesic cur-
vature. The third edge is a segment of a parallel, and may be parametrised
by
α(u) = f (v1 ) cos u, f (v1 ) sin u, g(v1 ) , u0 ≤ u ≤ u1 .
Then, along α,
nin = − f 0 (v1 ) cos u, f 0 (v1 ) sin u, g 0 (v1 ) ,

since these are the unit vectors pointing back down the meridians. A routine
calculation shows that
d2 α 1
2
=− (cos u, sin u, 0) ,
ds f (v1 )
so that, using (8.2), κg = f 0 (v1 )/f (v1 ). This is constant as expected, since
rotations of S about the z-axis are isometries. The length of α is equal to
(u1 − u0 )f (v1 ) so that
Z
κg ds = (u1 − u0 )f 0 (v1 ) .
∂T
(ii) The Gaussian curvature K was found in Corollary 5 of §5.5 to be given by

K = −f 00 /f , so
ZZ Z v1 Z u1 00 p
f
KdA = − f 2 du dv
T 0 u0 f
Z v1
f 00 dv = (u1 − u0 ) f 0 (0) − f 0 (v1 ) .

= −(u1 − u0 )
0
Since the pole of S is not a singular point, S intersects its axis of rotation
orthogonally (Example 6 of §4.2), so that g 0 (0) = 0. Hence f 0 (0) = 1, so that
ZZ
KdA = (u1 − u0 ) 1 − f 0 (v1 ) .

T
(iii) The interior angles of T are π/2, π/2 and u1 −u0 , so we see that (8.4) holds
for T , which confirms the Gauss-Bonnet theorem for a triangle in this case.
8.5. Since S is compact, Theorem 4 of §5.10 says that S has an elliptic

point p0 , say. Hence, if S had no hyperbolic points the Gauss-Bonnet theorem
would imply that the Euler characteristic of S was positive. In this case, the
Classification theorems in §8.4 would imply that S is diffeomorphic to a sphere.
So, let p1 be a hyperbolic point, and join p0 to p1 by a curve on S. Applying
the Intermediate Value theorem to K along this curve shows there is a point on
the curve with K = 0.
8.6. Parametrise the surface S as a surface of revolution,

x(u, v) = (v cos u, v sin u, v 2 ) , −π < u < π , v > 0 .
Routine calculations show that K = (LN − M 2 )/(EG − F 2 ) = 4(1 + 4v 2 )−2
(or use (5.25)), so that
ZZ Z aZ π p
1
KdA = 2
K EG − F du dv = 2π 1 − √ .
R 0 −π 1 + 4a2
The boundary ∂R is the parallel v = a, and nin = −xv /|xv |. Hence, using
(8.3),
1 xv Ev 1
κg = − xuu . = √ = √ .
|xu |2 |xv | 2E G a 1 + 4a2
R √
The length of the parallel is 2πa so that ∂R κg ds = 2π/ 1 + 4a2 . Since
χ(R) = 1, we now see that the Gauss-Bonnet theorem holds for R.
8.7. We assume that the generating curve is parametrised by arc length.

Along the parallel v = v0 we have that nin = xv , so, using (8.3) (or Exercise
7.4), we see that along v = v0 we have κg = xuu .xv /(|xu |2 ) = −f 0 (v0 )/f (v0 ).
Similarly, along v = v1 we have κg = f 0 (v1 )/f (v1 ).
Hence, Z
κg ds = 2π f 0 (v1 ) − f 0 (v0 ) .

∂R
Also, from Corollary 5 of §5.5, we have that K = −f 00 /f . Hence

ZZ Z v1
f 00 dv = 2π f 0 (v0 ) − f 0 (v1 ) .

KdA = −2π
R v0
This verifies the Gauss-Bonnet Theorem in this case, since R is an annulus and
hence has Euler characteristic zero.
8.8. This surface is diffeomorphic to a cylinder. Moreover, using the result of

Exercise 5.12 (or just do the calculations) we see that the sign of K is opposite
to the sign of cosh00 z. Hence K < 0 everywhere on S. It now follows from
Theorem 6 of §8.6 that S has at most one simple closed geodesic. The surface
intersects the plane z = 0 orthogonally (with the curve of intersection being an
ellipse) so it follows from Exercise 7.10 that this ellipse is a geodesic on S.
8.9. It follows from Exercise 5.26 that K ≥ 0, and K = 0 only on the

intersection of S with the coordinate planes. Since S is compact, Theorem 1
of §8.6 shows that S is diffeomorphic to S 2 (1). If geodesics Γ1 and Γ2 don’t
intersect then they bound a region R homeomorphic RR to an annulus. On this
annulus, K ≥ 0, and K > 0 somewhere. Hence R KdA > 0, so the Gauss-
Bonnet theorem would imply that χ(annulus) > 0 which gives a contradiction.
Chapter 9
9.1. Let S be a CMC surface in R3 with constant mean curvature c 6= 0,

and let f : R3 → R3 denote scalar multiplication by c. If S̃ is the image of
S under f , it is quick to prove that S̃ is a surface in R3 (if x(u, v) is a local
parametrisation of S then x̃(u, v) = cx(u, v) is a local parametrisation of S̃).
Since f is linear, its derivative at any point is equal to f itself, from which
it follows that Tf (p) S̃ = Tp S for each point p ∈ S. It follows from this that
Ñ f = N , so that c dÑ f (p) = dN p . Hence the mean curvature of S̃, being
minus half the trace of dÑ , is equal to c/c = 1.
ALTERNATIVELY: in the standard notation, it is quick to show that Ẽ =
(1/c2 )E, F̃ = (1/c2 )F , G̃ = (1/c2 )G, L̃ = (1/c)L, M̃ = (1/c)M , Ñ = (1/c)N ,
so it follows from (9.3) that S̃ has constant mean curvature 1.
9.2. This follows by taking h = H in equation (9.4) to see that A0 (0) < 0
for the given variation of any such open neighbourhood.
9.3. (i) This is immediate.

2
(ii) Integrating the given equation, we find that log(f 0 + 1) = 2(log f + log k)
for some positive constant k, so that
2
f 0 + 1 = (kf )2 .
(iii) It follows that
f0
p = ±1 ,
(kf )2 − 1
so, integrating, we find that arccosh (kf ) = ±k(v + c) for some constant c, so
that f = (1/k) cosh k(v + c) . Hence the graph of f is a catenary.
9.4. (i) In Example 1 of §3.2 we found that, for the standard parametrisation
of S as a graph,
E = 1 + gu 2 , F = gu gv , G = 1 + gv 2 ,
while the solution of Exercise 5.1 gives that if D = 1 + fu 2 + fv 2 then

√ √ √
L = fuu / D , M = fuv / D , N = fvv / D .
The result now follows quickly from equation (9.3).

(ii) If f (x, y) is a function of x only, then the minimal graph equation becomes
f 00 = 0, which integrates to give f (x) = ax + b for some √ constants a and b.
Hence S has equation √ z − ax = b so that N = (−a, 0, 1)/ 1 + a2 . Conversely,
if N = (−a, 0, 1)/ 1 + a2 for some constant a then S has equation z − ax = b
for some constant b, and S is the graph of f (x) = ax + b.
(iii) This time, denoting differentiation with respect to the appropriate variable
by 0 , we obtain
2 2
(1 + h0 )g 00 + (1 + g 0 )h00 = 0 .
Hence g 00 = 0 if and only if h00 = 0, and in this case the graph of f is a plane.
Otherwise, using the usual separation of variables argument,
2 2
(1 + h0 )/h00 = −(1 + g 0 )/g 00 = c
for some non-zero constant c.

The first equation integrates to give

h(y) = −c log cos (y − k̃)/c + `˜ ,
where k̃ and `˜ are constants, while the second one gives

g(x) = c log cos (x − k)/c + ` ,
where k and ` are constants.

Hence f takes the form given in the exercise, and the result follows.
9.5. The conjugate minimal surface is given by the imaginary part of the
C3 -valued function ψ given in Example 7 of §9.8, namely
v3 u3

1
y(u, v) = v − u2 v + ,u + − uv 2 , 2uv .
2 3 3

If x̃(u, v) = x̃1 (u, v), x̃2 (u, v), x̃3 (u, v) , then
u3 v3

1
x̃1 = √ u− + uv 2 + v − + u2 v ,
2 2 3 3
u3 v3

1
x̃2 = √ u− + uv 2 − v + − u2 v ,
2 2 3 3
1
x̃3 = (u2 − v 2 ) .
2
A calculation now shows that if ũ and ṽ are as given, then x̃(ũ, ṽ) = y(u, v)
as required.
9.6. If y is the harmonic conjugate of x then xθ = cos θ x + sin θ y. Hence
(xθ )u = cos θ xu + sin θ y u = cos θ xu − sin θ xv ,
with a similar expression for (xθ )v . The unit normal N to Sθ is independent of

θ so that
Lθ = −(xθ )u .N u = L cos θ − M sin θ ,
and the expressions for Mθ and Nθ may be found in a similar manner.
9.7. (i) Complex bilinearity shows that
φ.φ = (xu .xu − xv .xv ) + 2ixu .xv ,
which equals zero if and only if x is isothermal. Similarly,
φ.φ̄ = (xu .xu + xv .xv ) ,

(ii) We note that φ is holomorphic if and only if both xuu = −xvv and xuv =
xvu . The second equation always holds, and the first holds if and only if x is
harmonic. The final statement follows immediately from equation (9.12).
9.8. With f and g as given, short calculations show that

1
φ(z) = 1 − cos z, sin z, −2i sin(z/2) ,
2
so that
Z
1
ψ(z) = φ(z)dz = z − sin z, − cos z, 4i cos(z/2) + (c1 , c2 , c3 ) ,
2
where c1 , c2 , c3 are (complex) constants of integration. Hence,
1
x(u, v) = u − sin u cosh v, − cos u cosh v, 4 sin(u/2) sinh(v/2) + (k1 , k2 , k3 ) ,
2
and we take k1 = k3 = 0 and k2 = 1/2 to obtain the given formula for x.
9.9. Here,
1 1 1
φ(z) = z2 − , i z 2
+ , 2 ,
2 z2 z2
so that
z3 1 z3

1 1
ψ(z) = + ,i − , 2z ,
2 3 z 3 z
which gives that
u3 v3

1 2 u 2 v
x(u, v) = − uv + 2 , −u v + − 2 , 2u .
2 3 u + v2 3 u + v2
9.10. Arguing as in Example 7 of §9.8, we see after some calculation that

x̃ is the real part of the function ψ̃ defined in the exercise. The final equation
is now clear.
9.11. Routine calculations show that, for the parametrisation of Enneper’s

surface given in Exercise 9.5,
N = (2u, 2v, u2 + v 2 − 1)/(u2 + v 2 + 1) .
It follows that πN = u + iv, so that Theorem 2 of §9.9 is checked in this

case.
9.12. The given parametrisation of Enneper’s surface arises from taking

f (z) = 1, g(z) = z. It follows from Proposition 4 of §9.9 that L = −N = −1
and M = 0, so the required solution ω of the Liouville equation is given by
ω = log E. A routine calculation
shows that 4E = (1 + u2 + v 2 )2 , so that
ω = 2 log (1 + u2 + v 2 )/2 .
9.13. The given parametrisation of the helicoid has L − N − 2iM = −2i. It

follows from Lemma 1 of §9.13 that, in order to obtain an isothermal parametri-
sation x̃(ũ, ṽ) with L̃ = −Ñ = −1 and M̃ = 0, we should take ũ + iṽ = h(u, v)
where (h0 )2 = i. So, taking h(z) = eiπ/4 z, we have
√ √
ũ = (u − v)/ 2 , ṽ = (u + v)/ 2 .
Hence, our required isothermal parametrisation of the helicoid is given by
x̃(ũ, ṽ) = (sinh v sin u, − sinh v cos u, u) ,
where √ √
u = (ũ + ṽ)/ 2 , v = (−ũ + ṽ)/ 2 .
The corresponding Ẽ, F̃ and G̃ may be found either by direct calculation,

or by using (3.32). Either way, we find that
√
Ẽ = G̃ = cosh2 (−ũ + ṽ)/ 2 ,
and the required solution ω̃ of the Liouville equation is given by ω̃ = log Ẽ.
9.14. It follows from Exercise 9.6 that
Lθ − iMθ = e−iθ (L − iM ) = −e−iθ .
Hence, from Lemma 1 of §9.13, we need a change of variable z̃ = h(z) such that
2
h0 = e−iθ , so we take
z̃ = h(z) = ±e−iθ/2 z .
The final equation in the proof of Lemma 1 of §9.13 says that if x̃θ (z̃) = xθ (z)
then
(xθ )u − i(xθ )v = ± (x̃θ )ũ − i(x̃θ )ṽ e−iθ/2 ,

so, in particular, since Eθ = E we see that Ẽθ (z̃) = E(z).

Hence, the required solution of the Liouville equation is ω̃(ũ, ṽ) = ω(u, v)
where u + iv = eiθ/2 (ũ + iṽ).

A First Course in Differential Geometry Surfaces in Euclidean Space Solutions Manual - Woodward Problems

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A First Course in Differential Geometry Surfaces in Euclidean Space Solutions Manual - Woodward Problems

Uploaded by

Copyright:

Available Formats

1

A FIRST COURSE IN DIFFERENTIAL GEOMETRY

(a) Astroid (b) An epicycloid

1.2. A sketch of the trace of an epicycloid is given in Figure 1(b). Trigono-

α0 = 4r sin(u/2) sin(3u/2), cos(3u/2) .

1.3. When r = 1, a calculation shows that α0 = tanh u sech u(sinh u, −1),

Figure 2: (a) shows a tractrix and (b) shows three catenaries

Hence κ = sech2 u. A sketch of the traces of three catenaries is given in Figure

1.7. Differentiating with respect to u, we see that, using Serret-Frenet,

α` 0 = α0 + `n0 = |α0 |(t − κ`t) = |α0 |(1 − κ`)t .

α00 α0 (x0 x00 + y 0 y 00 )

and a short calculation shows that

1.9. (i) Let sα be arc length along α measured from u = 0. Since α0 =

Figure 3: A catenary and its evolute

1.10. Let sα denote arc length along α starting at u = u0 . Then, using

1.12. A sketch of the trace of α is given in Figure 4.

Figure 4: The curve in Exercise 1.12

Also, using the method of Example 2 of §1.5, we find that

It then follows that

1.14. Assume that α(s) is a smooth curve in R3 parametrised by arc length.

1.15. Using Serret-Frenet, we have

α0 = |α0 |t , α00 = κ|α0 |2 n + |α0 |0 t ,

α000 = −κτ |α0 |3 b + terms involving t and n ,

α0 = (−aθ0 sin θ, aθ0 cos θ, 1) ,

so the angle φ of the tangent vector to the z-axis is given by cos φ = (1 +

Hence α + an has constant rate of change τ X 0 /(κ2 + τ 2 ), so that

1.17. Assume there is a unit vector X 0 such that t.X 0 = c, a constant.

(t − kb)0 = |α0 | κn − (κ/τ )τ n = 0 ,

so that t − kb is constant. The result follows since t.(t − kb) = 1.

1.18. The assumption on α implies the existence of a smooth function r(u)

1.19. The given information implies that nα = ±nβ .

β(u) = (a cos u + au sin u, a sin u − au cos u, 0) ,

so, in particular, the third component of β is zero. The circle of intersection

2.2. Let V be an open subset of Rn such that X = V ∩ S is non-empty. If

In fact, the whole of S may be covered by just one parametrisation. For

Then, if we take W to be R3 , and let F : W → R2 be given by F (x, y, z) =

Figure 5: Coordinate neighbourhoods on a cylinder

Hence, grad f = 0 on and only on the circle x2 + y 2 = a2 in the plane z = 0,

Figure 6: f (x, y, z) = b2 when 0 < a < b

2.5(a). Since grad f = 2(x + y + z − 1)(1, 1, 1), it is clear that grad f = 0

2.5(b). If f (x, y, z) = xyz 2 , then grad f = 0 if and only if either z = 0 or

α0 = |α0 |t and α00 = |α0 |t0 + |α0 |0 t .

2.7. The ellipsoid S is defined by f (x, y, z) = 1, where f (x, y, z) = (x2 /a2 )+

W = {(x, y, z) ∈ R3 : |z| < c, and if y = 0 then x > 0} ,

2.8. (i) Use Theorem 1 of §2.4 to show that S is a surface by checking

2.9. Show that S is a surface by applying Theorem 1 of §2.4 to the function

3.3. Let f (x, y, z) = 2x2 − xy + 4y 2 . Then grad f = (4x − y, −x + 8y, 0).

3.5. Let f (x, y, z) = x2 + y 2 − z 2 . Then S has equation f (x, y, z) = 1. Also

3.6. . Let f (x, y, z) = x2 + y 2 + z 2 − ax and g(x, y, z) = x2 + y 2 + z 2 − by.

3.9. Let f1 (x1 , x2 , x3 , x4 ) = x1 2 + x2 2 and let f2 (x1 , x2 , x3 , x4 ) = x3 2 + x4 2 .

3.10. Routine calculations show that the surface S has E = G = cosh2 v,

F = 0. That xu = x̃v and xv = −x̃u is straightforward.

3.12. (i) Here α0 = xu + xv so that |α0 |2 = E + 2F + G, where the right

3.18. Here, E = 1 + v 2 , F = 1 and G = 2. Also, F is given by φ(u, v) =

|xu ||xu − xv | cos θ = xu .(xu − xv ) ,

so that E(E − 2F + G) cos2 θ = (E − F )2 . Substituting the values of E, F , G

3.19. Here, E = G = 1 and F = 0. The family F is given by φ(u, v) =

(v 0 sin u + vu0 cos u)(1 − cos u) = u0 v sin2 u .

This simplifies to give u0 v = v 0 sin u, so that the tangent vectors to F are

3.21. Here, E = v 2 + 1, F = 0, G = 1. So, using the result of Exercise

3.22. If α(t) = x u(t), v(t) is a member of F, then α0 (t) = u0 xu + v 0 xv

where u0 φu + v 0 φv = 0. Hence u0 /v 0 = −φv /φu , and it follows that tangent