Solutions to the exercises from sections 4.1, 4.

2,
4.3
Solution to Exercise 4.3. That (1) implies (2! ) is immediate since (1) implies (2), and (2) implies (2! ).
Now suppose that (2! ) holds. Let (xn )n∈N be a sequence contained in I that converges to c ∈ I. Then
the new sequence x1 , c, x2 , c, x3 , c, . . . also converges to c. By the hypothesis (2! ), it follows that the image
of this sequence under f , namely the sequence f (x1 ), f (c), f (x2 ), f (c), f (x3 ), f (c), . . . must be convergent,
say with limit L. But then each of its subsequences must also be convergent with the same limit L. By
looking at the even indexed terms of this sequence, we obtain that the subsequence f (c), f (c), f (c), . . .
converges to L, and so L = f (c). On the other hand, by looking at the odd indexed terms, we conclude
that the sequence f (x1 ), f (x2 ), f (x3 ), . . . must be also convergent with limit L (=f (c)). Thus (2) holds.
As (2) implies (1), it follows that (1) holds. Hence we have shown that (2! ) implies (1).

Solution to Exercise 4.6. Let x "= 0. We consider two possible cases.

1◦ Let x be rational. Then we can find a sequence (yn )n∈N of irrational numbers that converges
to x, by the result in Exercise 1.40. If f was continuous at x, then (f (yn ))n∈N = (−yn )n∈N
would have to be convergent with limit f (x) = x. But then we obtain x = −x and so x = 0, a
contradiction.
2◦ Let x be irrational. Then we can find a sequence (rn )n∈N of rational numbers that converges to
x, by the result in Exercise 1.39. If f was continuous at x, then (f (rn ))n∈N = (rn )n∈N would
have to be convergent with limit f (x) = −x. But then we obtain x = −x and so x = 0, a
contradiction.
So f is not continuous at x whenever x "= 0. We now show that f is continuous at 0. Let ! > 0. Let
δ := ! > 0. Since f (x) = x if x is rational, and −x if x is irrational, it follows that in either case,
|f (x)| = |x|. Thus, whenever x ∈ R and |x − 0| = |x| < δ = !, we have
|f (x) − f (0)| = |f (x) − 0| = |f (x)| = |x| < δ = !.
So f is continuous at 0.

Solution to Exercise 4.7. Let x be a rational number. Then f (x) > 0. We can find a sequence (yn )n∈N
of irrational numbers that converges to x, by the result in Exercise 1.40. If f was continuous at x, then
(f (yn ))n∈N = (0)n∈N would have to be convergent with limit f (x) > 0, which is clearly not true. Thus f
is not continuous at x.

Now suppose that x is irrational. Then x "= 0, and so there exists an integer N such that x ∈ (N, N +1).
Let ! > 0. Suppose that r is a rational number in (N, N + 1) for which f (r) ≥ !. Let r = n/d, where
n, d are integers without any common divisors and d > 0. Then we have f (r) = 1/d ≥ ! and so d ≤ 1/!.
So there are just finitely many possibilities for d. Moreover, since N < n/d < N + 1, it follows that
N d < n < (N + 1)d, and so there are only finitely many possibilities for d as well. So there are just finitely
many rational numbers r in (N, N + 1) for which f (r) ≥ !. Now among these finitely many rational
numbers, let r∗ be one which is the closest to x, and set δ := min{|x − r∗ |/2, |x − N |, |x − (N + 1)|} > 0.
Then whenever z ∈ R and |z − x| < δ, we are guaranteed that z ∈ (N, N + 1), and moreover if z is rational,
then it can’t be one of the rational r’s in (N, N + 1) for which f (r) ≥ !, and so it must be the case that
f (z) < !. If z is irrational, then f (z) = 0 by definition of f , and so then too we have f (z) = 0 < !. So
we have that for all z ∈ R such that |z − x| < δ, |f (z) − f (x)| = |f (z) − 0| = |f (z)| = f (z) < !. This
completes the proof that f is continuous at every irrational number.

Solution to Exercise 4.8. Imagine a directed line # such that the pancake lies entirely to one side of
the line # (say if we look along the direction of the line, the pancake appears to our right). Now translate
the line to the right parallel to itself till the whole pancake appears to the left of the line. Suppose
that the total distance by which the line is translated is d. At each intermediate distance x ∈ [0, d],
let A(x) denote the area of the part of the pancake that lies to the right of our line. Thus if S is the
total area of the pancake, then A(0) = A and A(d) = 0. As the map A : [0, d] → R is continuous and
f (0) = S ≥ S/2 ≥ 0 = f (d), it follows by the Intermediate Value Theorem, that there is an y ∈ [0, d] such
that A(y) = S/2, or in other words, at some position of our line, the pancake is divided into two parts
with equal areas.
As the direction of the line in the above process was inconsequential, given any direction, we can
choose a straight line cut having that direction which divides the pancake into two equal parts.
Solution to Exercise 4.10. For nonzero x ∈ R2 , we have
! !
! !
! x1 x2 ! |x1 ||x2 | 'x'2 'x'2
!" 2 != ≤ = 'x'2 .
! x1 + x22 ! 'x'2 'x'2

Let ! > 0. Set δ := ! > 0. Then if 0 "= x ∈ R2 and 'x − 0'2 = 'x'2 < δ = !, we have
! !
! !
! x1 x2 !
|f (x) − f (0)| = |f (x) − 0| = |f (x)| = ! " 2 ! ≤ 'x'2 < δ = !.
! x1 + x22 !

On the other hand, if x = 0, then |f (x) − f (0)| = 0 < !, trivially.

Thus f is continuous at 0.

Solution to Exercise 4.12. If x, y ∈ X we have by the triangle inequality that

d(x, p) ≤ d(x, y) + d(y, p),

and so d(x, p) − d(y, p) ≤ d(x, y). Similarly,

d(y, p) ≤ d(y, x) + d(x, p) = d(x, y) + d(x, p),

and so d(y, p) − d(x, p) ≤ d(x, y). Consequently, |d(x, p) − d(y, p)| ≤ d(x, y). Let ! > 0, and set δ := ! > 0.
Then whenever y ∈ X and d(x, y) < δ = !, we have

|d(x, p) − d(y, p)| ≤ d(x, y) < δ = !.

This shows that the map f is continuous at x ∈ X. As the choice of x ∈ X was arbitrary, it follows that
f is continuous on X.

Solution to Exercise 4.13. Let (x0 , y0 ) ∈ R2 . We have for all (x, y) ∈ R2 that

|x + y − (x0 + y0 )| = |(x − x0 ) + (y − y0 )| ≤ |x − x0 | + |y − y0 |
≤ '(x, y) − (x0 , y0 )'2 + '(x, y) − (x0 , y0 )'2 = 2'(x, y) − (x0 , y0 )'2 .

Thus, given ! > 0, we can set δ := !/2 > 0, and then for all (x, y) ∈ R2 satisfying the inequality
'(x, y) − (x0 , y0 )'2 < δ, we have

|x + y − (x0 + y0 )| ≤ 2'(x, y) − (x0 , y0 )'2 ≤ 2δ = !.

As the choice of (x0 , y0 ) ∈ R2 was arbitrary, it follows that addition is continuous.

Let (x0 , y0 ) ∈ R2 . We have for all (x, y) ∈ R2

|xy − x0 y0 | = |xy − x0 y + x0 y − x0 y0 )| ≤ |y||x − x0 | + |x0 ||y − y0 |

≤ |y|'(x, y) − (x0 , y0 )'2 + |x0 |'(x, y) − (x0 , y0 )'2
= (|y| + |x0 |)'(x, y) − (x0 , y0 )'2 .

But if (x0 , y0 ) is close to (x, y), then |y| is close to |y0 |. So the continuity will follow from the above
inequality. We give the precise details below. Let ! > 0. Set
# $
!
δ = min 1, > 0.
1 + |y0 | + |x0 |
Then whenever '(x, y) − (x0 , y0 )'2 < δ, we have first of all that

|y| = |y0 + (y − y0 )| ≤ |y0 | + |y − y0 | ≤ |y0 | + '(x, y) − (x0 , y0 )'2 < |y0 | + δ ≤ |y0 | + 1.

So if '(x, y) − (x0 , y0 )'2 < δ, then

!
|xy − x0 y0 | ≤ (|y| + |x0 |)'(x, y) − (x0 , y0 )'2 < (|y0 | + 1 + |x0 |)δ ≤ (|y0 | + 1 + |x0 |) = !.
1 + |y0 | + |x0 |

As the choice of (x0 , y0 ) ∈ R2 was arbitrary, it follows that multiplication is continuous.

Solution to Exercise 4.14. Let f ∈ C[0, 1]. If g ∈ C[0, 1], then we have for x ∈ [0, 1] that
|(Sg)(x) − (Sf )(x)| = |(g(x))2 − (f (x))2 | = |g(x) + f (x)||g(x) − f (x)|
≤ (|g(x)| + |f (x)|)'g − f '∞
≤ ('g'∞ + 'f '∞ )'g − f '∞ .
Thus 'Sg − Sf '∞ ≤ ('g'∞ + 'f '∞ )'g − f '∞ . If 'g − f '∞ < δ, then we have
'g'∞ ≤ 'g − f '∞ + 'f '∞ < δ + ∞,
and so the above yields that 'Sg − Sf '∞ < (δ + 2'f '∞ )δ. Let ! > 0. Define
# $
!
δ = min ,1 .
1 + 2'f '∞
Then whenever g ∈ C[0, 1] and 'g − f '∞ < δ, we have
'Sg − Sf '∞ < (δ + 2'f '∞ )δ ≤ (1 + 2'f '∞ )δ ≤ !.
Hence S is continuous at f . As the choice of f ∈ C[0, 1] was arbitrary, it follows that S is continuous on
C[0, 1].

Solution to Exercise 4.16.

(1) Let x, y ∈ C. Define γ : [0, 1] → C by γ(t) = (1 − t)x + ty, t ∈ [0, 1]. Then γ is well-
defined since γ(0) = x ∈ C, γ(1) = y ∈ C and for all t ∈ (0, 1), C being convex, we have
γ(t) = (1 − t)x + ty ∈ C as well.
Also, if s, t ∈ [0, 1], we have γ(t) = x + t(y − x) and γ(s) = x + s(y − x), and so
'γ(t) − γ(s)'2 = 'x + t(y − x) − (x + s(y − x))'2 = '(t − s)(y − x)'2
= |t − s|'y − x'2 .
From here it follows that γ is continuous. Hence γ is the desired path joining x to y. Conse-
quently, C is path connected.
(2) We verify below that R is reflexive, symmetric and transitive.
(R1) (Reflexivity) Let x ∈ S. Define γ : [0, 1] → S by γ(t) = x (t ∈ [0, 1]). Clearly γ is
continuous and γ(0) = γ(1) = x. So xRx.
(R2) (Symmetry) Let x, y ∈ S and suppose that xRy. Then there exists a continuous path
γ : [0, 1] → S such that γ(0) = x and γ(1) = y. Define the new map µ : [0, 1] → S by
µ(t) := γ(1 − t) (t ∈ [0, 1]). Then µ(0) = γ(1) = y, µ(1) = γ(0) = x, and since µ is the
composition of two continuous maps, namely the continuous map γ with the continuous
map t )→ 1 − t : [0, 1] → [0, 1], it follows that µ is continuous too. Hence yRx as well.
(R3) (Transitivity) Let x, y, z ∈ S be such that xRy and yRz. We would like to show that xRz.
As xRy and yRz, there exist continuous maps γ1 , γ2 : [0, 1] → S such that γ1 (0) = x,
γ1 (1) = y, γ2 (0) = y and γ2 (1) = z. Define the map γ : [0, 1] → S as follows:
#
γ1 (2t) if t ∈ [0, 21 ),
γ(t) =
γ2 (2t − 1) if t ∈ [ 12 , 1].
Then γ(0) = γ1 (0) = x and γ(1) = γ2 (1) = z. Also, it is not hard to check that γ is
continuous. Hence xRz.
Thus R is an equivalence relation.
(3) The circle S1 := {(x, y) ∈ R2 : x2 + y 2 = 1} is path connected. Indeed, if a, b ∈ S1 , then
there are real numbers θa , θb ∈ R such that a = (cos θa , sin θb ) and b = (cos θb , sin θb ). Define
the map θ : [0, 1] → R by θ(t) = (1 − t)θa + tθb , t ∈ [0, 1]. Now define γ : [0, 1] → R by
γ(t) = (cos(θ(t)), sin(θ(t))) (t ∈ [0, 1]). Then γ is continuous, γ(0) = a and γ(1) = b. Thus S1 is
path connected. (Alternately, using the fact that R is an equivalence relation, it is easy to see
that it is enough to check the path connectedness of the two arcs
A+ := {(x, y) ∈ R2 : x2 + y 2 = 1, y ≥ 0},
A− := {(x, y) ∈ R2 : x2 + y 2 = 1, y ≤ 0}
lying in the upper half-plane and lower half plane, respectively. If a = (x1 , ya ) and b = (xb , yb )
belong to A+ , then define x : [0, 1] → R by x(t) = (1−t)xa +txb (t ∈ [0, 1]). Define γ : [0, 1] → R2
as follows: % " &
γ(t) = x(t), 1 − (x(t))2 (t ∈ [0, 1]).
Then γ is continuous, γ(0) = aand γ(1) = b. This shows that A+ is path connected. For
showing that A− is path connected, we proceed similarly, except that we use
% " &
γ(t) = x(t), − 1 − (x(t))2 (t ∈ [0, 1])

as our path.)
The set {(x, y) ∈ R2 : xy = 0} is just the union of the two straight lines, namely the
horizontal “x-axis” and the vertical “y-axis” when we identify R2 with the plane in the usual
manner. It is now pictorially evident that this set is path connected. Indeed, if x, y belong
to this set then we have three cases. If they both lie on the x-axis we just join them by the
horizontal straight line segment which is contained entirely in the x-axis. If they both lie on
the y-axis, we join them by the vertical line segment which is contained entirely in the y axis.
Finally, if one lies on the x axis and the other on the y-axis, then we join the point on the x-axis
to the origin by a horizontal straight line segment, and then join the origin to the point on the
y-axis with a vertical line segment. Figure 6 illustrates this.

Figure 6. Paths in the set {(x, y) ∈ R2 : xy = 0}.

The set H := {(x, y) ∈ R2 : xy = 1} is not path connected, and we prove this by con-
tradiction. Clearly the points (1, 1) and (−1, −1) both belong to H. Suppose that there is a
continuous path γ : [0, 1] → R2 such that γ(0) = (−1, −1) and γ(1) = (1, 1). Composing γ with
the linear transformation which is projection onto the x-axis, namely the map

x(t) := the first component of γ(t) (t ∈ [0, 1]),

we have that x : [0, 1] → R is continuous and x(0) = −1 and x(1) = 1. As we have that
x(−1) = −1 < 0 < 1 = x(1), it follows from the Intermediate value Theorem that there is a
t0 ∈ [0, 1] such that x(t0 ) = 0. But this means that the first component of x(t0 ) is 0 and so the
product of the two components of γ(t0 ) is 0. Thus γ(t0 ) "∈ H, a contradiction. Consequently,
H is not path connected.
In fact, H has two path components, namely,

H+ := {(x, y) ∈ R2 : xy = 1 and x > 0},

H− := {(x, y) ∈ R2 : xy = 1 and x < 0}.

It is easy to see that H+ and H− are path connected. We just show it for H+ . Suppose
that a = (xa , ya ) and b = (xb , yb ) are points in H+ . Let the map x : [0, 1] → R be defined by
x(t) = (1 − t)xa + txb (t ∈ [0, 1]). Then x(t) > 0 for all t ∈ [0, 1]. Define γ : [0, 1] → R2 as
follows:
' (
1
γ(t) = x(t), (t ∈ [0, 1]).
x(t)

Then γ is continuous, γ(0) = a and γ(1) = b.

 H+

H−

Figure 7. Path components of the set {(x, y) ∈ R2 : xy = 1}.

Solution to Exercise 4.17. We have

f −1 ({−1, 1}) = {nπ : n ∈ Z},

f −1 ({1}) = {2nπ : n ∈ Z},

f −1 ([−1, 1]) = f −1 (R) = R,

'' (( )% )' (
1 1 π π π π& π 2π
f −1 − , = (2n + 1) − , (2n + 1) + = nπ + , nπ + .
2 2 n∈Z
2 6 2 6 n∈Z
3 3

Solution to Exercise 4.18. Since the cosine function cos is periodic with period 2π (that is, there holds
f (x) = f (x + 2π) for all x ∈ R), it follows that f (R) = f ([0, 2π]) = f ([δ, δ + 2π]) = [−1, 1].

Solution to Exercise 4.21. (“If” part) Suppose that for every closed F in Y , f −1 (F ) is closed in X.
Now let V be open in Y . Then Y \ V is closed in Y . Thus
f −1 (Y \ V ) = f −1 (Y ) \ f −1 (V ) = X \ f −1 (V )
is closed in X. Hence f −1 (V ) = X \ (X \ f −1 (V )) is open in X. So for every open V in Y , f −1 (V ) is
open in X. Consequently, by Theorem 4.19, f is continuous on X.

(“Only if” part) Suppose that f is continuous. Let F be closed in Y . Then Y \ F is open in Y . Hence
f −1 (Y \F ) = f 1 (Y )\f −1 (F ) = X \f −1 (F ) is open in X. Consequently, we have that f −1 (F ) = X \f −1 (F )
is closed in X. So for every closed F in Y , f −1 (F ) is closed in X.

Solution to Exercise 4.23. Let x ∈ (f ◦ g)−1 (W ). Then (f ◦ g)(x) ∈ W , that is, f (g(x)) ∈ W . So
g(x) ∈ f 1 (W ). Hence x ∈ g −1 (f −1 (W )). So we have the inclusion (f ◦ g)−1 (W ) ⊆ g −1 (f −1 (W )).
Now let x ∈ g −1 (f −1 (W )). This means that g(x) ∈ f −1 (W ). Thus (f ◦ g)(x) = f (g(x)) ∈ W .
Hence x ∈ (f ◦ g)1 (W ). So we also have the inclusion g −1 (f −1 (W )) ⊆ (f ◦ g)−1 (W ). Consequently,
(f ◦ g)−1 (W ) = g −1 (f −1 (W )).

Solution to Exercise 4.26. We show by induction that f (n) = nf (1) for all natural numbers n. For
n = 1, we have f (n) = f (1) = 1 · f (1) = nf (1), and so the claim is true when n = 1. If for some n ∈ N
there holds f (n) = nf (1), then we have f (n + 1) = f (n) + f (1) = nf (1) + f (1) = (n + 1)f (1). This
completes the induction step, and so the result holds for all natural numbers.
 Also f (0) = f (0 + 0) = f (0) + f (0) shows that f (0) = 0 = 0 · f (1). So f (n) = nf (1) for all nonnegative
integers. Let m be a negative integer. Then

0 = f (0) = f (m + (−m)) = f (m) + f (−m) = f (m) + (−m)f (1),

and so f (m) = −(−m)f (1) = mf (1). So we have that f (n) = nf (1) for all integers n.
Now every rational number r can be expressed as r = n/d for some integers n, d with d > 0. Then

nf (1) = f (n) = f (d · (n/d)) = f (n/d) + · · · + f (n/d) = d · f (n/d).

* +, -
d times

Consequently, f (n/d) = (nf (1))/d, that is, f (r) = rf (1).

Let x ∈ R. Then we can find a sequence (rn )n∈N of rational numbers which converges to x. Using
the continuity of f we obtain
% &
f (x) = f lim rn = lim f (rn ) = lim rn f (1) = xf (1).
n→∞ n→∞ n→∞

Consequently, f (x) = xf (1) for all x ∈ R.

Solution to Exercise 4.27. First note that if we substitute x = 0 into the given identity, we obtain
f (0) + f (0) = 0, hence f (0) = 0.
Clearly f (2x) = −f (x) for all x ∈ R, and so

f (x) = −f (x/2) = +f (x/4) = −f (x/8) = +f (x/16) = · · · = +f (x/4n )

for all x ∈ R and n ∈ N. That is, for any x ∈ R the sequence

f (x), f (x/4), f (x/16), . . . , f (x/4n ), . . . ,

is a constant sequence with limit f (x). Since the sequence (x/4n )n∈N converges to 0, and f is continuous
at 0, we also obtain that the above sequence converges to f (0) = 0. Since limits are unique, we obtain
that f (x) = 0 for all x ∈ R.
So if f is continuous and it satisfies the given identity then it must be the constant function 0.
Conversely, note that the constant function 0 is indeed continuous and also satisfies the identity
f (2x) + f (x) = 0 for all x ∈ R.

Solution to Exercise 4.29. Clearly the sequence given by (1, (−1)n /n), n ∈ N, is convergent with
limit (1, 0). So if f −1 were continuous, then (f −1 ((1, (−1)n /n)))n∈N should converge to f −1 (1, 0) = 0.
However, if we look at the odd indexed terms of (f −1 ((1, (−1)n /n)))n∈N , then we obtain the subsequence
(f −1 ((1, −1/n)))n∈N , which clearly converges to 1 "= 0 = f −1 (1, 0). Hence f −1 is not continuous at the
point (1, 0).

Solution to Exercise 4.30. See Figure 8. Look at a spherical cap Ox around the point x on the sphere.
Imagine a light source at the center O of the (transparent) sphere, which casts a shadow U of the (coloured)
spherical cap on the plane R2 (screen) perpendicular to the line Ox. It is clear then that the sets Ox and
U are homeomorphic. This shows that S2 is a manifold of dimension 2.

Solution to Exercise 4.31. Consider the two sequences (vn )n∈N and (wn )n∈N defined by
. 1 / . /
0
vn := n and wn := 1 (n ∈ N),
0 n

both of which are convergent to 0. But (f (vn ))n∈N = (1)n∈N , while (f (wn ))n∈N = (−1)n∈N , and clearly
these cannot have the same limit f (0) = c, no matter what c is. Hence it follows from Theorem 4.25 that
f is not continuous at 0.
 R2

screen

x
Ox

O
light source

Figure 8. S2 is a 2-dimensional manifold.

Solution to Exercise 4.32. The determinant of M is given by the sum of expressions of the type
±m1p(1) m2p(2) m3p(3) . . . mnp(n) ,
where p : {1, 2, 3, . . . , n} → {1, 2, 3, . . . , n} is a permutation. Since each of the maps
M )→ m1p(1) m2p(2) m3p(3) . . . mnp(n)
is easily seen to be continuous using the characterization of continuous functions provided by Theorem 4.25,
it follows that their linear combination is also continuous.
{0} is closed in R, and so its inverse image under the continuous map det : Rn×n → R is also closed.
In other words, det−1 {0} = {M ∈ Rn×n : det M = 0} is closed. Thus its complement, namely the set
{M ∈ Rn×n : det M "= 0} is open. But this is precisely the set of invertible matrices, since M ∈ Rn×n is
invertible if and only if det M "= 0.

Solution to Exercise 4.33. Let X = (0, 1], Y = R with the usual Euclidean metrics. Define f : (0, 1] →
R by
1
f (x) = (x ∈ (0, 1]).
x 0 1
Then f is continuous. Take (xn )n∈N to be the Cauchy sequence n1 n∈N . Then f (xn ) = n, and we have
|f (xn ) − (xm )| = |n − m| ≥ 1 for all n "= m, showing that the sequence (f (xn ))n∈N is not Cauchy.

Solution to Exercise 4.34. If U is open in X = R with the Euclidean metric, then f (U ) = U is also
open in Y = R with the discrete metric d, since all subsets of (R, d) are open. So the map f is open.
Let x0 ∈ R and let ! = 1/2 > 0. If f is continuous at x0 , then there exists a δ > 0 such that whenever
|x − x0 | < δ, we have d(f (x), f (x0 )) < ! = 1/2. Now take x = x0 + δ/2. Then |x − x0 | = δ/2 < δ, but
f (x) = x = x0 + δ/2 "= x0 = f (x0 ) and so d(f (x), f (x0 )) = 1 "< 1/2 = !. Thus we arrive at a contradiction.
Consequently, f is not continuous at x0 .

Solution to Exercise 4.35.

2
(1) Since (x ± y 2 )2 ≥ 0, we obtain by rearranging that ± x2xy +y 4
≤ 1
2
and so
! !
! xy 2 !
|f (x, y)| = !! 2 !≤ 1
x + y4 ! 2
whenever (x, y) "= (0, 0).
(2) Consider the ball B(0, r) with center (0, 0) and radius r > 0. The idea is to take points (xn , yn )
in B(0, r) of the type
1 1
xn := m , yn := !
n n
with large enough natural numbers n and suitable exponents #, m so that so that
1
nm+2! nm+4!
g(xn , yn ) = = 6!
1 1 n + n2m
+ 6!
n2m n
grows unboundedly with increasing n. Clearly this can be arranged if m + 4l > 6# and moreover
m + 4# > 2m, that is if 4# > m > 2#. One choice guaranteeing this is # = 1 and m = 3. Then
1
n5 n
g(xn , yn ) = = .
1 1 2
+ 6
n6 n
2 2
1 1 1 1 n→∞
Clearly for all large enough n’s, '(xn , yn )'2 = + < r since + 2 −→ 0. So the
n6 n2 n6 n
n
points (xn , yn ) belong to B(0, r) for all n’s large enough. But as g(xn , yn ) = , we see that g
2
is not bounded on B(0, r).
1 1
(3) We have that the sequence ((an , bn ))n∈N given by an = and bn = (n ∈ N), converges to
n2 n
(0, 0). However,
1
n4 1
f (an , bn ) = = (n ∈ N),
1 1 2
4
+ 4
n n
and so (f (an , bn ))n∈N = (1/2)n∈N does not converge to f (0, 0) = 0. Consequently, f is not
continuous at (0, 0).
(4) Suppose that g is continuous at (0, 0). Then with ! := 1 > 0, there exists a r > 0 such that
whenever (x, y) ∈ B(0, r), we have that
|g(x, y) − g(0, 0)| = |g(x, y) − 0| = |g(x, y)| < ! = 1.
In particular, g is bounded on the ball B(0, r). However, we have already seen in part (2) that
this is false. Hence g is is not continuous at (0, 0).
(5) f (0, y) = 0 for all y ∈ R, and so the restriction of f to the line x = 0 is continuous. Similarly,
since f (x, 0) = 0 for all x ∈ R, the restriction of f to the line y = 0 is continuous. For a point
p = (x, mx) on the line y = mx with m "= 0, and x "= 0, we have that
x(mx)2 m2 x
f (p) = = .
x2 + (mx)4 1 + m4 x 2
Now if pn = (xn , mxn ) (n ∈ N) is a sequence of points on the line y = mx that converges to
p = (x, mx), then it is clear that (xn )n∈N converges to x, and so
m2 x n
f (pn ) =
1 + m4 x2n
converges to
m2 x n
f (p) =
1 + m4 x2n
by the algebra of limits. (If x = 0, then xn converges to x = 0, and so f (pn ) clearly converges
to 0 = f (0, 0) = f (p).) Thus the restriction of f to the line y = mx is continuous.
g(0, y) = 0 for all y ∈ R, and so the restriction of f to the line x = 0 is continuous. Similarly,
since g(x, 0) = 0 for all x ∈ R, the restriction of f to the line y = 0 is continuous. For a point
p = (x, mx) on the line y = mx with m "= 0, and x "= 0, we have that
x(mx)2 m2 x
g(p) = = .
x2 + (mx)6 1 + m6 x 4
 Now if pn = (xn , mxn ) (n ∈ N) is a sequence of points on the line y = mx that converges to
p = (x, mx), then it is clear that (xn )n∈N converges to x, and so
m2 x n
g(pn ) =
1 + m6 x4n
converges to
m2 x n
g(p) =
1 + m6 x4n
by the algebra of limits. (If x = 0, then xn converges to x = 0, and so g(pn ) clearly converges
to 0 = g(0, 0) = g(p).) Thus the restriction of g to the line y = mx is continuous.

Solution to Exercise 4.40. We had seen in Exercise 1.37 that a singleton set in any metric space is
closed. In particular, the set {0} is closed in Rn . As the linear transformation TA : Rm → Rn is continuous,
it follows that its inverse image under f , namely the set
f −1 (0) = {x ∈ Rm : Ax = 0} = ker A
is closed in Rm .

Solution to Exercise 4.41. Let V be a subspace of Rn , and let {v1 , . . . , vk } be a basis for V . Extend
this to a basis {v1 , . . . , vk , vk+1 , . . . , vn } for Rn . By using the Gram-Schmidt orthogonalization procedure,
we can find an orthonormal set of vectors {u1 , . . . , un } such that for each i ∈ {1, . . . , n}, the span of the
vectors v1 , . . . , vi coincides with the span of u1 , . . . , ui . Now define A ∈ R(n−k)×n as follows:
 ) 
uk+1
A =  ...  .
 

u)
n

It is clear from the orthonormality of the ui ’s that Au1 = Au2 = Au3 = · · · = Auk = 0, and so it follows
that also any linear combination of u1 , . . . , uk lies in the kernel of A. In other words, V ⊆ ker A.
On the other hand, if x = α1 u1 + · · · + αn un , where α1 , . . . , αn are scalars and if Ax = 0, then it
follows that  )   
uk+1 (α1 u1 + · · · + αn un ) αk+1
0 = Ax = 
 ..   .. 
. = . ,
)
un (α1 u1 + · · · + αn un ) α n
and so αk+1 = · · · = αn = 0. Thus x = α1 u1 + · · · + αk uk ∈ V . Hence also ker A ⊆ V . Consequently
V = ker A, and by the result of the previous exercise, it now follows that V is closed.