18.

100B Problem Set 7 Solutions

Sawyer Tabony
1) We have ai > 0 and ai+1 ≤ ai for all i = 0, 1, 2, ..., and lim ai = 0, and we want to show the
i→∞
convergence of
∞
�
(−1)i ai = a0 − a1 + a2 − ...
i=0

So we define sn to be the partial sums of the first n + 1 terms of the sum:

n
�
sn = (−1)i ai
i=0

So for any k ∈ N, we have s2k − s2k−2 = (−1)2k−1 a2k−1 + (−1)2k a2k = a2k − a2k−1 ≤ 0 by the
monotonicity of an . Therefore s2k ≤ s2k−2 , so s2k is decreasing. Similarly, s2k−1 is increasing.
Also, s2k − s2k−1 = (−1)2k a2k = a2k > 0, so s2k > s2k−1 . These combine to give s2k−1 < s2k� for
any k, k � = 0, 1, 2, ..., since choosing N > max{k, k � }, we have

s2k−1 ≤ s2N −1 < s2N ≤ s2k� .

So s2k and s2k−1 are both monotonic and bounded, so they each converge. However, s2k −s2k−1 =
a2k −→ 0, so they must converge to the same limit. Therefore sn converges, so the sum is
convergent.

2) Our function f is defined on (0, 1) by

⎨ 1 if, in lowest terms, x = p ∈ Q,

⎧

f= q q
⎩0 if x �∈ Q.

p 1
First we show f is discontinuous at every rational. For r = ∈ Q (in lowest terms), let ε = .
q 2q
1 1 1
Then for any δ > 0, ∃x ∈ � Q with r < x < r + δ, so |f (r) − f (x)| = | − 0| = > = ε. This
q q 2q
proves discontinuity at r.
Now we want to show continuity at irrationals, so let x �∈ Q, 0 < x < 1. Given ε > 0, we
need to find a δ > 0 such that for every y with |x − y| < δ, |f (x) − f (y)| < ε. Since f (x) = 0
1
and f (y) ≥ 0, we need f (y) < ε. But ε > 0 means we can find N ∈ N with < ε. So
N
|f (y)| > ε means y, when written in lowest terms, has denominator smaller than N . But there
are only finitely many fractions with denominator less than N between 0 and 1, so for some
M , {yi }M
i=1 = {y ∈ (0, 1)|f (y) > ε}. So we let δ = min |x − yi |, which exists and is greater
1≤i≤M
than 0 since there are only finitely many yi . Then if y ∈ (0, 1) is such that |x − y| < δ, y =
� yi
∀i ∈ {1, 2, ..., M }, so f (y) < ε. Therefore f is continuous at x.

3) We have f, g : M −→ N , and Q ⊆ M is dense.

1
 a) We need to show f (Q) is dense in f (M). So let K ⊆ N be closed, with f (Q) ⊆ K. Then by
continuity of f , f −1 (K) is closed, and f −1 (K) contains f −1 (f (Q)) ⊇ Q. Since Q is dense in
M, f −1 (K) = M. Hence f (M) ⊆ K, so f (Q) is dense in f (M).
b) Now we have f = g on Q. Now consider the function φ : M −→ R, with
φ(x) = dN (f (x), g(x))
for dN the distance function on N . Since dN , f , and g are all continuous, φ is also continuous.
Therefore φ−1 (0) is a closed set in M. But since f = g on Q,
∀x ∈ Q, f (x) = g(x) =⇒ φ(x) = dN (f (x), g(x)) = 0.
Thus Q ⊆ φ−1 (0), which is closed, so by density,
φ−1 (0) = M =⇒ ∀x ∈ M, 0 = φ(x) = dN (f (x), g(x)) =⇒ f (x) = g(x).

4) a) So we must find a continuous f : E −→ R with E ⊆ R bounded and f (E) unbounded. Let

1
E = (0, 1), f (x) =
x
So E is clearly bounded, f (E) = (1, ∞) is unbounded, and f is continuous: at x ∈ (0, 1),
given ε > 0, let δ = min{ x2 , 13 x2 ε} > 0. Then
1 2
1 1 δ x ε 2
f (x − δ) − f (x) = − = ≤ 3 x = ε < ε.
x−δ x x(x − δ) x( 2 ) 3
And similarly, f (x) − f (x + δ) < ε. Since f is monotonically decreasing, this shows f is
continuous.
b) Now we have that f is uniformly continuous, and E is bounded. So for ε = 1, ∃δ > 0 such that
∀x, y ∈ E, if |x − y | < δ, |f (x) − f (y)| < 1. E is bounded, so let B ∈ N such that E ⊆ [−B, B].
4B δ
Then we can divide [−B, B] into � � closed intervals of length , say Ii for i = 1, 2, ..., M .
δ 2
Then choose xi ∈ Ii ∩ E, when Ii ∩ E = � ∅. Let C = max{|f (xi )|} + 1. Then for any x ∈ E,
i
δ
x ∈ Ii for some i, so |x − xi | ≤ < δ. Thus, |f (x) − f (xi )| < 1, so |f (x)| < 1 + |f (xi )| ≤ C.
2
So C bounds f (E).
c) This is the easiest one: let E = R, and f (x) = x. Then f is uniformly continuous by choosing
δ = ε, since |f (x) − f (y)| = |x − y|, but E = f (E) = R are both unbounded.

5) f : M −→ N is a uniformly continuous map between metric spaces.

a) We need to show that f preserves Cauchy sequences. So we are given that (xn ) is Cauchy.
To show (f (xn )) is Cauchy, let ε > 0. Then by uniform continuity, ∃δ > 0 such that if
x, y ∈ M with dM (x, y) < δ, then dN (f (x), f (y)) < ε. Since (xn ) is Cauchy, ∃N ∈ N such
that ∀n, m > N , dM (xn , xm ) < δ. Therefore, dN (f (xn ), f (xm )) < ε, for all n, m > N . So
(f (xn )) is Cauchy.
b) We have g(x) = x2 on R. g is continuous, so given a Cauchy sequence (xn ) in R, it converges
by the completeness of R, so (g(xn )) is convergent by the continuity of g, so it is Cauchy. But
 1
g is not uniformly continuous: for any δ > 0, letting x = we have
δ
δ δ2
g(x + )2 − g(x) = (x2 + 1 + ) − x2 > 1.
2 4
So for ε = 1, no δ exists that satisfies uniform continuity on all of R.

6) We have defined
dE (x) = inf d(x, z),
z∈E
and we want to show that
|dE (x) − dE (y)| ≤ d(x, y).
So fix ε > 0. Then ∃z0 ∈ E with d(x, z) < dE (x) + ε. Then d(y, z) ≤ d(x, z) + d(x, y) <
dE (x) + ε + d(x, y) by the triangle inequality. Therefore,
dE (y) = inf d(y, z) ≤ dE (x) + ε + d(x, y) =⇒ dE (y) − dE (x) ≤ d(x, y) + ε
z∈E
for any ε > 0. So dE (y) − dE (x) ≤ d(x, y). By symmetry, dE (x) − dE (y) ≤ d(x, y), so
|dE (x) − dE (y)| ≤ d(x, y).
Uniform continuity follows immediately by letting δ = ε.

7) So now we have K, F ⊆ M, with K ∩ F = ∅, F closed and K compact. By the previous exercise,

dF is uniformly continuous, and thus continuous, positive function on K. Therefore dF attains its
minimum on K, by compactness. So ∃x ∈ K with dF (x) = inf y∈K dF (y). Suppose this infimum
is 0. Thus dF (x) = 0, so x is a limit point of F . But F is closed, so x ∈ F =⇒ x ∈ K ∩ F = ∅, a
contradiction. Therefore inf y∈K dF (y) = δ > 0. Therefore ∀p ∈ K, q ∈ F , d(p, q) > 2δ > 0.
This doesn’t hold for arbitrary K, F ⊆ M closed. For a counterexample, take M = R2 and
look at
1
K = {(x, y) ∈ R2 |y = 0}, F = {(x, y) ∈ R2 |x ≥ 0, y ≥ }
x
1 1
Both are closed, and for any δ > 0, for N > , (N, 0) ∈ K and (N, ) ∈ F , and the distance
δ N
1
betweent these is < δ.
N
MIT OpenCourseWare
http://ocw.mit.edu

18.100B Analysis I
Fall 2010

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.