Homework Solutions #4

CAS MA 511

Problem (5.2.2). (a) is possible. Consider f (x) to be the Dirichlet function and

0 if x ∈ Q
g(x) =
1 if x ∈
6 Q

(b) is possible. Consider g(x) = x and


1 if x 6= 0
x
f (x) =
1 if x = 0

(c) is impossible. If g and f + g are differentiable, then by the algebraic differentiability theorem,
the difference (f + g) − g = f must be differentiable.

(d) is possible. Let d(x) be the Dirichlet function and consider f (x) = x2 d(x). If f (x) were
differentiable at any point other than 0, then it would be continuous at that point and hence
f (x)
x2
= d(x) would be continuous which is a contradiction. So, now we just need to show that f (x)
is indeed differentiable at x = 0. We have:
x2 d(x) − 02 d(0) x2 d(x)
lim = lim = lim xd(x) = 0
x→0 x−0 x→0 x x→0

Thus, f (x) is differentiable at x = 0 (and you can see why we needed to multiply by x2 .)
Problem (5.2.3). For (a), we have:
1 c−x
x
− 1c 1 1
lim = lim xc = − lim =− 2
x→c x−c x→c x − c x→c xc c
Thus, h0 (x) = − x12 .

For (b), from part (a) and the chain rule, we have:
!0
1 g 0 (x)
=−
g(x) [g(x)]2
Thus, by theorem 5.2.4 part (iii), we have:
!0 !0
f 01 1 g(x) g 0 (x) f 0 (x)g(x) − f (x)g 0 (x)
(x) = f (x) + f (x) = f 0 (x) − f (x) =
g g(x) g(x) [g(x)]2 [g(x)]2 [g(x)]2

For (c), we have:

f (x) f (c)
g(x)
− g(c) 1 f (x)g(c) − f (c)g(x) 1 f (x)g(c) − f (c)g(c) + f (c)g(c) − f (c)g(x)
= =
x−c x−c g(x)g(c) x−c g(x)g(c)

1
 !
1 f (x)g(c) − f (c)g(c) + f (c)g(c) − f (c)g(x) 1 f (x) − f (c) g(x) − g(c)
= g(c) − f (c)
x−c g(x)g(c) g(x)g(c) x−c x−c
Now, when we take a limit as x → c, we have:

f 0 (x)g(x) − f (x)g 0 (x)

!
1 f (x) − f (c) g(x) − g(c)
lim g(c) − f (c) =
x→c g(x)g(c) x−c x−c [g(x)]2

Problem (5.2.12). Let f : [a, b] → R be continuous and differentiable on its domain, with f 0 (x) 6=
0. Then there exists a function f −1 defined on its range by f −1 (y) = x where f (x) = y. Thus, by
the chain rule:

f −1 (f (x)) = x
(f −1 (f (x)))0 = (x)0
(f −1 )0 (f (x))f 0 (x) = 1
1
(f −1 )0 (f (x)) = 0
f (x)
1
(f −1 )0 (y) = 0
f (x)

where y = f (x). The second to last equality is where we use that f 0 (x) 6= 0.

Problem (5.3.6). For (a), let g(x) : [0, a] → R be differentiable, suppose g(0) = 0 and |g 0 (x)| ≤ M
on [0, a]. Then, for x ∈ [0, a] by the mean value theorem, there is some 0 ≤ c ≤ x ≤ a such that:

g(x) − g(0) g(x)

g 0 (c) = =
x−0 x
and hence |g(x)| = |g 0 (c)x| ≤ M x.

For (b), suppose that h(x) : [0, a] → R is twice differentiable with h0 (0) = h(0) = 0 and |h00 (x)| ≤
M on [0, a]. If we try to apply the technique of (a), twice, we will get that h(c) ≤ M x2 , so we will
2
try something different. Consider f1 (x) = h(x) − M x2 . Then, f 00 (x) = h00 (x) − M ≤ 0 on [0, a]
and hence by applying the mean value theorem twice, we see that f1 (x) ≤ 0 on [0, a] and hence
2 2
h(x) ≤ M x2 on [0, a]. Similarly, if we let f2 (x) = h(x) + M x2 , then we will find that f2 (x) ≥ 0 so
2 2
that h(x) ≥ −M x2 on [0, a] and hence |h(x)| ≤ M x2 on [0, a]

For (c), we can show that if l(x) : [0, a] → R is differentiable n times with l(n) (0) = · · · = l0 (0) = 0
n
and |l(n) (x)| ≤ M on [0, a], then |l(x)| ≤ M xn! . Then, by applying the mean value theorem n times
n n
to f1 (x) = l(x) − M xn! and f2 (x) = l(x) + M n xn! we can show that f1 (x) ≤ 0 and f2 (x) ≥ 0 on
(n) (n)
[0, a], since we have f1 (x) = l(n) (x) − M ≤ 0 and f2 (x) = l(n) (x) + M ≥ 0. Thus, we have
xn xn xn
l(x) ≤ M n! and l(x) ≥ −M n! and hence |l(x)| ≤ M n! . (Note that we need the n!1 to cancel all
of the factors introduced by taking the n derivatives of xn .)

Problem (5.3.10). First, of all we have:

1
lim g(x) = lim e− x2 = 0
x→0 x→0

since 1
x2
→ ∞ as x → 0 and e−x → 0 as x → ∞. Next, we have:

1
 
1
lim |f (x)| = lim |x| sin e− x2 ≤ lim |xg(x)| = 0
x→0 x→0 x4 x→0

2
since g(x) → 0 as x → 0. Since we have |f (x)| → 0 as x → 0, we must have that f (x) → 0 as
x → 0. Thus both f (x) and g(x) tend to 0 as x tends to 0. Now, consider the quotient:
f (x) 1
 
lim = lim x sin 4 =0
x→0 g(x) x→0 x
 
since −x ≤ x sin x14 ≤ x, which allows us to use the squeeze theorem. Finally, consider the
quotient of derivatives:
  1   1   1
0
f (x) sin 1
x4
e− x2 − 4x−4 cos 1
x4
e− x2 + 2x−2 sin 1
x4
e− x2 +
lim = lim 1
x→0 g 0 (x) x→0 −3 − x2
   2x e  
sin x14 − 4x−4 cos x14 + 2x−2 sin x14
= lim
x→0 2x−3
1 1 1 1
     
= lim x3 sin 4 − 4x−1 cos 4 + 2x sin 4
2 x→0 x  x x
1
= −2 lim x−1 cos 4
x→0 x
where the second to last equality follows by two applications of the squeeze theorem. Now this
final limit diverges, since we can always find some small

x > 0 such

that x14 = 2kπ or (2k + 1)π.
This allows us to find sequences (xn ) → 0 such that x−1 1
n cos x4n → ∞ or −∞. This does not
contradict L’Hospital’s rule, since the limit of the quotient of derivatives does not converge, so we
cannot apply it.
1 1
Problem (5.4.2). First of all, we have 0 ≤ h(2n x) ≤ 1, and thus we have 0 ≤ 2n
h(2n x) ≤ 2n
.
Furthermore, we have:
∞
X 1 1
= =2
n=0 2
n 1 − 12
P∞ 1 n
Thus, by the comparison test, n=0 2n h(2 x) also converges.
Problem (6.2.12). For (a), we have the graphs of f1 , f2 , and f3 below (spaced out so we can see
all 3 without overlap):

Indeed, we observe that each is continuous, increasing and constant on [0, 1] \ Cn .

For (b), we can continue this process to obtain a sequence

h i (fn ). Now, we want to use the Cauchy
1 2
criterion for uniform convergence. First of all, if x ∈ 3 , 3 :

|fn (x) − fm (x)| = 0

h i
Now, if x ∈ 0, 13 , then:
1 1
|fn (x) − fm (x)| = |fn−1 (3x) − fm−1 (3x)| ≤ |fn−1 (x) − fm−1 (x)|
2 2
3
where the last inequality follows since each fn is increasing. Thus, we have:
1 1 1 1
|fn (x)−fm (x)| ≤ |fn−1 (x)−fm−1 (x)| ≤ 2 |fn−2 (x)−fm−2 (x)| ≤ · · · ≤ m |fn−m (x)−f0 (x)| ≤ m
2 2 2 2
1
where this final inequality holds since f0 (x) = x ≤ 1 and each 0 ≤ fn ≤ 1. Since, 2m
→ 0, we have
by the Cuachy criterion for uniform convergence that (fn ) converges uniformly.

For (c), let f = lim fn . Since each fn is continuous, by theorem 6.2.6, f is also continuous. Now,
consider x, y ∈ [0, 1] with x ≤ y. Then, since fn is increasing, we have fn (x) ≤ fn (y). Taking
a limit as n → ∞, we obtain that f (x) ≤ f (y), so that f is also increasing. Since fn (0) and
f1 (1) = 1 for all n, we have f (0) = 0 and f (1) = 1.

Now, consider x ∈ [0, 1] \ C. Then, since Cn ⊂ Cn+1 , there must be some N such that x 6∈ Cn for
all n ≥ N . Then, x is in some open interval of length 31N on which fn is constant for n > N . In
particular, f is also constant on this interval and hence has f 0 (x) = 0. (Note that to say something
about the derivative at x we need to know what is going on in a small neighborhood around x.)
Since x was arbitrary, f 0 (x) = 0 on [0, 1] \ C.
Problem (6.2.15). Let (fn ) be a bounded, equicontinuous sequence on [0, 1]. For (a), using exercise
6.2.13, since Q ∩ [0, 1] is countable and (fn ) is bounded, we can produce a subsequence fnk that
converges on Q ∩ [0, 1]. Denote this subsequence by gk .

For (b), since (fn ) is equicontinuous, so is (gk ). Thus, there exists a δ > 0 such that |x − y| < δ
implies that |gk (x) − gk (y)| < 3ε .

Now, since [0, 1] is compact it can be covered in a finite number of δ-neighborhoods centered at
rational points r1 , . . . , rm . Thus, since (gk ) converges to g, we can find a single N (by taking a
maximum over the finite number δ-neighborhoods) such that for s, t ≥ N we have |gs (ri )−gt (ri )| <
ε
3
for all i. If the number of points weren’t finite the maximum might not exist.

For (c), for each x ∈ [0, 1], we must have that x is in some δ-neighborhood of one of the ri and
hence for s, t ≥ N , we have:

|gs (x) − gt (x)| = |gs (x) − gs (ri ) + gs (ri ) − gt (ri ) + gt (ri ) − gt (x)|
≤ |gs (x) − gt (ri )| + |gt (ri ) − s(ri )| + |gt (ri ) − gt (x)|
ε ε ε
< + + =ε
3 3 3
Thus, (gk ) converges uniformly on [0, 1], by the Cauchy criterion.
Problem (6.3.4). For any x, we have:
1
|hn (x)| ≤ √ → 0
n

and hence hn (x) → 0 regardless of x, so that the convergence is uniform. On the other hand,
consider the derivatives:
cos(nx) √
h0n (x) = n √ = n cos(nx)
n

√ cos(nx) = 0 for all n ≥ N , since otherwise the sequence

Suppose to reach a contradiction that
will be unbounded because of the n. In particular, cos(N x) = 0. However, this implies that
cos(2N x) 6= 0, which is a contradiction.
√ Thus, for any x, we can find arbitrarily large n for which
0
cos(nx) 6= 0 and hence (h (x)) = (( n cos(nx)) is unbounded.

4
Problem (6.3.7). Let (fn ) be a sequence of differentiable functions defined on the closed interval
[a, b], and assume (fn0 ) converges uniformly on [a, b]. Suppose there exists x0 ∈ [a, b] where fn (x0 )
is convergent. Now, since fn (x) − fm (x) is differentiable, we can apply the mean value theorem to
find some c between x0 and x such that:
fn (x) − fm (x) − (fn (x0 ) − fm (x0 ))
fn0 (c) − fm
0
(c) =
x − x0
Now, since the derivatives converge uniformly, we can choose N1 such that for n, m ≥ N1 we have
0
that |fn0 (c) − fm (c)| < min{ 2ε , 2(b−a)
ε
} and hence:
ε
0
|fn (x) − fm (x) − (fn (x0 ) − fm (x0 ))| = |x − x0 ||fn0 (c) − fm 0
(c)| ≤ |b − a||fn0 (c) − fm (c)| <
2
Now, since (fn ) converges at x0 , can choose N2 such that for n, m ≥ N2 |fn (x0 ) − fm (x0 )| < 2ε .
Thus, for any x ∈ [a, b] and n, m ≥ max N1 , N2 , we have:
ε ε
|fn (x) − fm (x)| ≤ |fn (x) − fm (x) − (fn (x0 ) − fm (x0 ))| + |fn (x0 ) − fm (x0 )| < + =ε
2 2
Thus, (fn ) converges uniformly on [a, b].
Problem (6.5.5). For (a), we have:
(n + 1)sn n+1
lim n−1
= n→∞
lim s =s<1
n→∞ ns n
Thus, by the ratio test, ∞ n−1
converges. Since this series converges, we must have nsn−1 → 0
P
n=0 ns
and hence this sequence must be bounded.

For (b), suppose that f (x) = ∞ n

n=0 an x converges on (−R, R) and pick t such that |x| < t < R.
P

Then, we will show that the differentiated series converges absolutely at t and hence converges on
[−t, t] for any such t, which implies that it converges on (−R, R). Since |x| t
< 1, (a) says that
|x| n−1
 
n t
< M for some M :
 !n−1 
∞ ∞ ∞ ∞
|x| M X
|nan xn−1 | = (tn−1 |an |) n tn−1 |an | = tn |an |
X X X
 ≤M
n=1 n=1 t n=1 t n=1

Now, since f (x) = ∞ n

P
n=0 an x converges on (−R, R) it converges absolutely on [−t, t] by the-
P∞ n
orem 6.5.1 and theorem 6.5.2. Thus, n=1 t |an | converges so that by the comparison test,
P∞ n−1
n=1 |nan x | converges, so we are done.
Problem (6.5.8b). Suppose that f (x) = ∞ n 0
P
n=0 an x converges on (−R, R) and f (x) = f (x) for
all x ∈ (−R, R) and f (0) = 1. Then, since f 0 (x) = f (x), differentiating term by term, we have:
∞ ∞ ∞
an xn = f (x) = f 0 (x) = nan xn−1 = (n + 1)an+1 xn
X X X

n=0 n=1 n=0

So, by the uniqueness of power series representations (i.e. 6.5.8a), we have:

an
an+1 =
n+1
Now, since f (0) = 1, we have a0 = 1, so that:
an−1 an−2 a0 1
an = = = ... = =
n n(n − 1) n! n!
What function has this special property that f 0 (x) = f (x)?

5
Problem (6.6.5). For (a), since f (n) (x) = ex , we have that the Taylor coefficients are:

f (n) (0) e0 1
= =
n! n! n!
P∞ Rn
Now, let R > 0 and consider the series n=0 n! . Then, we have:
Rn+1
(n+1)! R
lim Rn = n→∞
lim =0<1
n→∞
n!
n+1

Thus, by the ratio test, this series converges. Moreover, the convergence is absolute (since the
xn
terms are positive) so that by theorem 6.5.2, the Taylor series ∞
P
n=0 n! converges uniformly on the
closed interval [−R, R].

For (b), we differentiate term by to term to compute f 0 (x), as follows:

∞ ∞ ∞
xn−1 xn−1 xn
f 0 (x) =
X X X
n = = = f (x)
n=1 n! n=1 (n − 1)! n=0 n!

For (c), we have:

∞ ∞
(−x)n (−1)n xn
e−x = f (−x) =
X X
=
n=0 n! n=0 n!
Then, we have:
∞ ∞
xn (−1)n xn
! !
x −x
X X
e e =
n=0 n! n=0 n!
! !
x2 (−x)2 x3 x2 (−x)2 (−x)3
= 1 + (x − x) + + x(−x) + + + (−x) + x + + ···
2 2 6 2 2 6
!
3 3 3 3
x x x x
= 1 + (x − x) + (x2 − x2 ) + − + − + ···
6 2 2 6
=1