+ All Categories
Home > Documents > STANFORD MATH 113 Elementary-Analysis-Solutions

STANFORD MATH 113 Elementary-Analysis-Solutions

Date post: 21-Jan-2016
Category:
Upload: spitzersglare
View: 570 times
Download: 0 times
Share this document with a friend
Description:
X
68
Elementary Analysis Kenneth A. Ross Selected Solutions Angelo Christopher Limnios
Transcript
Page 1: STANFORD MATH 113 Elementary-Analysis-Solutions

Elementary Analysis

Kenneth A. Ross

Selected Solutions

Angelo Christopher Limnios

Page 2: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 1.2

Claim:P (n) = 3 + 11 + · · ·+ (8n− 5) = 4n2 − n ∀n ∈ N

Proof :

By induction. Let n = 1. Then 3 = 4(1)2 − (1) = 3, which will serve as theinduction basis. Now for the induction step, we will assume P (n) holds trueand we need to show that P (n+ 1) holds true. P (n+ 1) = 3 + 11 + · · ·+ (8n−5) + (8(n+ 1)− 5) = 4(n+ 1)2 − (n+ 1).

Now, 3 + 11 + · · · + (8n − 5) + (8(n+ 1)− 5) = 4n2 − n + 8(n + 1) − 5. So inorder to show that P (n+ 1) is true, we need to show that 4(n+ 1)2− (n+ 1) =4n2 − n+ 8(n+ 1)− 5, which, when evaluated is true, as desired. ♠

EXERCISE 1.3

Claim:P (n) = 13 + 23 + · · ·+ n3 = (1 + 2 + · · ·+ n)2 ∀n ∈ N

Proof :

By induction. Let n = 1. Then, P (1) = 13 = 1 = 12, which will serve as theinduction basis. Now for the induction step, we will assume P (n) holds trueand we need to show that P (n+ 1) holds true.

P (n+ 1) = 13 + 23 + · · ·+ n3 + (n+ 1)3 = (1 + 2 + · · ·+ n)2 + (n+ 1)3

= (1 + 2 + · · ·+ n)2 + [(n+ 1)3 − (n+ 1)2] + (n+ 1)2

= (1 + 2 + · · ·+ n)2 + (n+ 1)2[n+ 1− 1] + (n+ 1)2

= (1 + 2 + · · ·+ n)2 + 2(n+ 1)n(n+ 1)/2 + (n+ 1)2

= (1 + 2 + · · ·+ n)2 + 2(n+ 1)(1 + 2 + · · ·+ n) + (n+ 1)2

= 1(1 + 2 + · · ·+ (n+ 1))2,

as desired. ♠

1(a+ b)2 = a2 + 2ab+ b2, where a = 1 + 2 + · · ·+ n and b = n+ 1.

1

Page 3: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 1.4

(a) For

n = 1, the expression is 1.

n = 2, the expression is 4.

n = 3, the expression is 9.

n = 4, the expression is 16.

We note that 1 + 3 + · · · (2n− 1) = n2, which is the proposed formula.

(b) Claim:P (n) = 1 + 3 + · · · (2n− 1) = n2 ∀n ∈ N

Proof :

By induction. We already showed the case when n = 1. Now for the inductionstep, we will assume P (n) holds true and we need to show that P (n+ 1) holdstrue.

P (n+ 1) = 1 + 3 + · · · (2(n+ 1)− 1) = 1 + 3 + · · ·+ (2n− 1) + (2n+ 1)

= n2 + 2n+ 1 = (n+ 1)2,

as desired. ♠

EXERCISE 1.6

Claim:P (n) = (11)n − 4n is divisible by 7 when n ∈ N

Proof :

By induction. P (1) = 11− 4 = 7 is divisible by 7. Now for the induction step,we will assume P (n) holds true and we need to show that P (n+ 1) holds true.

P (n+ 1) = (11)n+1 − 4n+1 = 11n+1 − 4 · 11n + 4 · 11n − 4n+1

= 11n(11− 4) + 4(11n − 4n) = 7 · 11n + 4(11n − 4n).

Since we assumed that 11n − 4n was divisible by 7 because we assumed P (n)was true, we can see that

11n+1 − 4n+1

is divisible by 7, as desired. ♠

2

Page 4: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 1.9

(a) The inequality does not hold for n = 2, 3, 4. It holds true for all other n ∈ N.

(b) It is true by by inspection for n = 1 and 24 = 42 also holds for n = 4.Implement the induction step. For n ≥ 4, if 2n ≥ n2, then 2(n+1) > (n + 1)2.But 2(n+1) = 2 · 2n ≥ 2n2 > (n + 1)2 iff (n + 1) <

√2n, for example when

n > 1√2−1

=√

2 + 1, which includes n ∈ N : n ≥ 4.

EXERCISE 1.10

Claim:

(2n+ 1) + (2n+ 3) + (2n+ 5) + · · ·+ (4n− 1) = 3n2 ∀n ∈ N.

Proof :

Using the results from 1.4, we can avoid induction. Observe that

(2n+1)+(2n+3)+ · · ·+(4n−1) = (1+3+ · · ·+(4n−1))−(1+3+ · · ·+(2n−1))

= (1 + 3 + · · ·+ 2(2n)− 1)− (1 + 3 + · · ·+ (2n− 1))

= (2n)2 − n2 = 4n2 − n2 = 3n2,

as desired. ♠

EXERCISE 1.12 (b) and (c)

(b) Observe the following:

1k

+1

n− k + 1=

(n− k + 1) + k

k(n− k + 1)=

n+ 1k(n− k + 1)

.

Hence,

n!k!(n− k)!

+n!

(k − 1)!(n− k + 1)!=

n!(k − 1)!(n− k)!

[1k

+1

n− k + 1

]=

n!(k − 1)!(n− k)!

[n+ 1

k(n− k + 1)

]=

(n+ 1)!k!(n− k + 1)!

.

(c) For n = 1 : (a+ b)n = a+ b =(11

)a+

(11

)b. Now, for n ≥ 1, we get

(a+ b)n =n∑k=0

(n

k

)akbn−k.

3

Page 5: STANFORD MATH 113 Elementary-Analysis-Solutions

Hence,

(a+ b)(n+1) = (a+ b)n∑l=0

(n

l

)albn−l =

n+1∑k=0

[(n

k − 1

)+(n

k

)]akbn−k+1

=n+1∑k=0

(n+ 1k

)akbn−1+k.

EXERCISE 2.1

Show that√

3 is not rational (The exact same technique can be used to showthe other numbers are not rational).

Claim: √3 /∈ R

Proof :

By Contradiction. If√

3 ∈ R, then ∃ an r ∈ Q : pq =

√3, where r = p

q =⇒(pq

)2

= 3 =⇒ p2 = 3q2 =⇒ p2 is divisible by 3 =⇒ p is divisible by 3. Let p =

3k. Then 9k2 = 3q2 =⇒ 3k2 = q2 =⇒ q2 is divisible by 3 =⇒ q is divisible by 3,which is a contradiction. ♠

EXERCISE 2.5

Let’s assume [3+√

2]23 does represent a rational number. Further, let’s call this

rational number q. This implies q3 = [3 +√

2]2 = 9 + 6√

2 + 2 = 11 + 6√

2,which in turn implies

√2 = (q3−11)

6 , which is a rational number. But we know√2 /∈ Q - a contradiction.

4

Page 6: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 3.3

This problem will be done in two parts.

Part I: Show that (−a)(−b) = ab ∀ (a, b) ∈ R.

Proof:

(−a)(−b) = ab (1)⇐⇒ (−a)(−b) + (−ab) = ab+ (−ab) (2)⇐⇒ (−a)(−b) + (−a)b = ab+ (−ab) (3)⇐⇒ (−a) [(−b) + b] = ab+ (−ab) (4)⇐⇒ (−a) [b+ (−b)] = ab+ (−ab) (5)

⇐⇒ (−a)(0) = ab+ (−ab) (6)0 = ab+ (−ab), (7)

which is true by A4. Hence, (−a)(−b) = ab2 ♠

Part II: ac = bc and c 6= 0 imply a = b ∀ (a, b, c) ∈ R.

Proof:

ac(c−1) = bc(c−1) (8)

a(cc−1) = b(cc−1) (9)a(1) = b(1) (10)

Hence, a = b3 ♠

EXERCISE 3.6

a. Claim: |a+ b+ c| ≤ |a|+ |b|+ |c|

Proof: One iteration of the triangle inequality can be used to construct:

|a+ b+ c| = |(a+ b) + c| ≤ |a+ b|+ |c| .

A second iteration of the triangle inequality yields:

|a+ b|+ |c| ≤ |a|+ |b|+ |c| ,2(1) used (i), (2) used (iii), (3) used DL, (4) used A2, (5) used A4, (7) reintroduced

(1) and the conclusion reused (i).3(8) was built from multiplying both sides of the equality by the same element to preserve

equality, (9) used M1 and (10) used M4.

5

Page 7: STANFORD MATH 113 Elementary-Analysis-Solutions

as desired. ♠b. We want to show |a1 + a2 + · · ·+ an| ≤ |a1|+ |a2|+ · · · |an|

We will use the principal of mathematical induction to show the above is true.

Proof:

P1 is certainly true, since |a1| = |a1|, and thus will serve as our basis for induc-tion. Our induction hypothesis is:

Assume ∃ an n ∈ R : Pn holds, i.e. Pn : |a1 + a2 + · · ·+ an| ≤ |a1|+|a2|+· · · |an| .

Implementing the induction step and the triangle inequality yield:

|a1 + a2 + · · ·+ an + an+1| ≤ |a1 + a2 + · · ·+ an|+ |an+1| .

The induction step and O4 yield:

|a1 + a2 + · · ·+ an + an+1| ≤ |a1|+ |a2|+ · · ·+ |an|+ |an+1| ,

as desired. ♠

EXERCISE 4.2, (a) through (n)

This exercise is asking us to list three lower bounds for the set; if the set is notbounded below, it will be labeled ”NOT BOUNDED BELOW” or ”NBB”. Therule outlined in the text we will follow is Definition 4.2 (b):

If a real number m satisfies m ≤ s ∀ s ∈ S, then m is called a lower bound of Sand the set S is said to be bounded below.

a. S = [0, 1]; −1, −2 and −3 satisfy Definition 4.2 (b).

b. S = [0, 1); −1, −2 and −3 satisfy Definition 4.2 (b).

c. S = {2, 7}; −1, −2 and −3 satisfy Definition 4.2 (b).

d. S = {π, e}; −1, −2 and −3 satisfy Definition 4.2 (b).

e. S ={

1n : n ∈ N

}; −1, −2 and −3 satisfy Definition 4.2 (b).

f. S = {0}; −1, −2 and −3 satisfy Definition 4.2 (b).

6

Page 8: STANFORD MATH 113 Elementary-Analysis-Solutions

g. S = [0, 1] ∪ [2, 3]; −1, −2 and −3 satisfy Definition 4.2 (b).

h. S =⋃∞n=1 [2n, 2n+ 1]; −1, −2 and −3 satisfy Definition 4.2 (b).

i. S =⋂∞n=1

[− 1n , 1 + 1

n

]; −1, −2 and −3 satisfy Definition 4.2 (b).

j. S ={

1− 13n : n ∈ N

}; −1, −2 and −3 satisfy Definition 4.2 (b).

k. S ={n+ (−1)n

n : n ∈ N}

; −1, −2 and −3 satisfy Definition 4.2 (b).

l. S = {r ∈ Q : r < 2}; “NBB”. The set Q has no maximum or minimum4.S has an upper bound, but since there is no lower bound outlined, this set isunbounded from below.

m. S ={r ∈ Q : r2 < 4

}; −2

1 , −31 and −4

1 satisfy Definition 4.2 (b).

n. S ={r ∈ Q : r2 < 2

}; −2

1 , −31 and −4

1 satisfy Definition 4.2 (b).

EXERCISE 4.4, (a) through (n), ADDITIONALLY, DETERMINE IF THEMINIMUM OF THE SET EXISTS.

This exercise is asking us to give the infima of each set S. The rule outlined inthe text we will follow is Definition 4.3 (b):

If S is bounded below and S has a greatest lower bound, then we will call it theinfimum of S and denote it by infS.

Additionally, we are asked to determine if the minimum of the set (denotedminS) exists. Note that, unlike the minimum of the set S, infS need not belongto the set S5

a. S = [0, 1]; infS = minS = 0.

b. S = (0, 1); infS = 0, minS D.N.E.

c. S = {2, 7}; infS = minS = 2.

d. S = {π, e}; infS = minS = e.

4Page 20, Example 1 (c).5Page 21, first paragraph below Definition 4.3, first sentence.

7

Page 9: STANFORD MATH 113 Elementary-Analysis-Solutions

e. S ={

1n : n ∈ N

}; infS = 0, however, minS does not exist.

f. S = {0}; infS = minS = 0.

g. S = [0, 1] ∪ [2, 3]; infS = minS = 0.

h. S =⋃∞n=1 [2n, 2n+ 1]; infS = minS = 2.

i. S =⋂∞n=1

[− 1n , 1 + 1

n

]; infS = 0, minS D.N.E.

j. S ={

1− 13n : n ∈ N

}; infS = minS = 2

3 .

k. S ={n+ (−1)n

n : n ∈ N}

; infS = minS = 0.

l. S = {r ∈ Q : r < 2}; both infS and minS do not exist since S is “NBB”.

m. S ={r ∈ Q : r2 < 4

}; infS = −2, however, minS does not exist.

n. S ={r ∈ Q : r2 < 2

}; infS = −

√2, however, minS does not exist.

EXERCISE 4.8

Let S and T be non-empty subsets of R, with s ≤ t ∀s ∈ S and t ∈ T .

(a) Since it is given that s ≤ t ∀s ∈ S and t ∈ T , we know that any elementof T will bound S from above, so ∃ supS. Conversely, since it is given thats ≤ t ∀s ∈ S and t ∈ T , we know that any element of S will bound T frombelow, so ∃ infT .

(b) Claim:supS ≤ infT

This will be done in two parts.

Proof :

Given s ≤ t ∀t ∈ T , it follows that s is a lower bound for T . By definition, infTis the greatest lower bound for T . Hence, s ≤ infT ∀s ∈ S.

Since s ≤ infT ∀s ∈ S, infT is an upper bound for S. It follows that since supSis the least upper bound for S,

supS ≤ infT,

8

Page 10: STANFORD MATH 113 Elementary-Analysis-Solutions

as desired. ♠(c) Let S = (0, 1] and T = [1, 2). It is readily observable that supS = infT . Itis also readily observable that

S ∩ T 6= ∅.

(d) Let S = (0, 1) and T = (1, 2). It is readily observable that supS = infT . Itis also readily observable that

S ∩ T = ∅.

EXERCISE 4.12

Claim:

Given a < b, ∃ x ∈ R \Q : a < x < b.

Proof :

Following the hint, we know r +√

2 ∈ R \ Q when r ∈ Q. By contradiction,if r ∈ Q and if the number x = r +

√2 ∈ Q, then

√2 = x − r would have

been ∈ Q, a contradiction. Now, due to the denseness of Q in R, we find anr ∈ Q : a−

√2 < r < b−

√2. Then we have x = r +

√2 ∈ R \Q and we have

a < x < b, as desired. ♠

EXERCISE 4.14 (a)

Let A and B be nonempty bounded subsets of R, and let S be the set of allsums a+ b where a ∈ A and b ∈ B.

Claim: supS = supA + supB.

Proof :

The information we are given is that A and B are both subsets of R whichare non-empty and bounded from above (since they are bounded, this impliesthey are bounded from above and below). Since both subsets A and B are bothnon-empty and bounded from above, they both possess a least upper bound,i.e., both supA and supB exist6. If S is to be defined as the set whose elementsare all of the sums a + b, where a ∈ A and b ∈ B and we are given both Aand B bounded, we know S is bounded; more specifically bounded from belowand above. Hence, by the Completeness Axiom, S is bounded above and as aconsequence supS exists. If supS exists, then ∃ a number O ∈ R : r ≤ 0 ∀ r ∈ Sand whenever O1 < O, ∃ r1 ∈ O : r1 < O1..

6Otherwise, the “Completeness Axiom” wouldn’t hold.

9

Page 11: STANFORD MATH 113 Elementary-Analysis-Solutions

If both supA and supB exist, then there exists a number M ∈ R : s ≤M ∀ s ∈ Aand there exists a number N ∈ R : t ≤ N ∀ t ∈ B; whenever M1 < M ∃ s1 ∈A : s1 < M1 and whenever N1 < N ∃ t1 ∈ B : t1 < N1.

As a consequence of the above identities, it is obvious that since S ≡ set of allsums a + b where a ∈ A and b ∈ B, O ≡ M + N , and hence supS = supA +supB, as desired. ♠

EXERCISE 4.15

Claim:

Let a, b ∈ R. If a ≤ b+ 1n ∀n ∈ N, then a ≤ b.

Proof :

By contradiction. Assume a > b. This implies a− b > 0. By the “ArchimedianProperty” of R ∃ n ∈ N : a− b > 1

n . Using this specific n, we have a > b+ 1n , a

contradiction, as desired. ♠

EXERCISE 4.16

Claim: sup{r ∈ Q : r < a} = a for each a ∈ R.

Proof :

Let S ≡ {r ∈ Q : r < a}. From this, we can see that S is certainly boundedfrom above, since we can chose any number a + n with n ∈ N and Definition4.2 (b) will still be satisfied. Now by the denseness of Q in R, we know that7

if (a, b) ∈ R with a < b, then ∃ an r ∈ Q : a < r < b. So although we knowS is bounded and we can certainly povide many upper bounds for this set, ifwe search for a least upper bound for this set, the search will go on indefinitely,since whatever rational upper bound is discovered, it is always possible to findone smaller. Thus, although a ∈ R, but a /∈ S, this does not disqualify it frombeing the supS. However, since the search for a least upper bound for S goeson indefinitely due to the denseness of Q, by the definition of the supremum ofa set, sup{r ∈ Q : r < a} = a for each a ∈ R, as desired. ♠

EXERCISE 7.3 (b), (d), (f), (h), (j), (l), (n), (p), (r), (t)

(b) bn = n2+3n2−3 converges to 1.

(d) tn = 1 + 2n converges to 1.

(f) sn = (2)1n converges to 1.

(h) dn = (−1)nn diverges.

7Proof ommitted, as it was done in class

10

Page 12: STANFORD MATH 113 Elementary-Analysis-Solutions

(j) 7n3+8n2n3−31 converges to 7

2 .

(l) sin(nπ2

)diverges.

(n) sin(

2nπ3

)diverges.

(p) 2n+1+52n−7 = 2n(2)+5

2n−7 converges to 2.

(r)(1 + 1

n

)2 converges to 1.

(t) 6n+49n2+7 converges to 0.

EXERCISE 8.4

If (sn) is a sequence which converges to 0 and (tn) is a bounded sequence, thenthe sequence (sntn) converges to 0.

We will argue the fact that we are given a sequence (sn) which converges to 0so we know that

limn→∞

(sn) = 0,

and we are given a sequence (tn) which is a bounded sequence, so we know that

∃ a constant M : |tn| ≤M ∀n ∈ N,

which means, in a geometric sense, that we can find an interval [−M,M ] thatcontains every term in the sequence tn.

With the above given, we can almost assuredly argue with what we know aboutproducts that if one sequence which arbitrarily closes in on the value zero ismultiplied by another sequence which is bounded by a constant we can con-clude that their product will eventually close in on the value zero since “zero”× “a constant” = zero (eventually).

Claim: Given (sn), a sequence which converges to 0 and (tn), a bounded se-quence,

limn→∞

(sntn) −→ 0.

Proof :

Let ε > 0 be given. Since we know limn→∞ sn → 0, we can always find ann > N : |sn − 0| < ε < ε

M8, for some constant M > 1. Further, ∃ a constant

M : |tn| ≤M ∀n ∈ N. Then, for n > N , it holds that

|sntn − 0| = |sn| |tn| <( ε

M

)(M) = ε. ♠

8Proof ommitted since convergence to 0 is taken as given.

11

Page 13: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 8.7

Claim:

sn = cos(nπ3

)does not converge.

Proof :

For n = 1, 2, · · · , 6, the terms of the sequence are 12 ,−

12 ,−

11 ,−

12 ,

12 ,

11 ,. . . .

Hence, for any s and any N , we can come up with n > N : sn = 1 =⇒sn+3 = −1 =⇒ |sn − s| ≥ 1 or |sn+3 − s| ≥ 1, by the triangle inequality, whichproves this sequence does not converge, as desired. ♠

EXERCISE 8.8 (c)

Claim:

limn→∞[√

4n2 + n− 2n] = 14 .

Proof : √4n2 + n− 2n =

n√4n2 + n+ 2n

=1

2√

1 + 14n + 2

If 1 < a, then 1 < a < a2 which then implies

1 ≤ limn→∞

√1 +

12n≤ limn→∞

[1 +

12n

]= 1 + 0 = 1.

Now, applying the limit theorems, we see that

limn→∞

1

2√

1 + 14n + 2

=1

2 · 1 + 2=

14,

as desired. ♠

12

Page 14: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 8.10

We are given s > a. Let limn→∞ sn = s.

Claim:

∃ a number N : n > N =⇒ sn > a.

Proof :

s > a =⇒ there is an ε > 0 : s − ε > a9 For such chosen ε ∃ Nε : n > Nε =⇒sn ∈ (s− ε, s+ ε) =⇒ n > Nε =⇒ sn > s− ε > a, as desired. ♠

EXERCISE 9.4

Let s1 = 1 and for n ≥ 1 let sn+1 =√sn + 1.

(a) The first four elements are 1,√

2,√√

2 + 1,√√√

2 + 1 + 1.

(b) Assume sn converges.

Claim: limn→∞ sn = 12 (1 +

√5).

Proof :

We are given the assumption that sn converges. We will name some σ asthe element sn converges to. If sn converges, sn+1 inherently converges; morespecifically, sn+1 converges to σ, as well. From the problem, we have

(sn+1)2 = sn + 1,

which when utilized with the assumption laid out will converge to

(σ)2 10 = σ + 1⇐⇒ σ2 − σ − 1 = 0.

Implementing the quadratic formula11 to find a pair of solutions to the equationyields

σ1 =1 +√

52

and σ2 =1−√

52

.

Since the greatest lower bound of the sequence sn is 1, the solution we wantto pick will reflect that the limit will trivially be positive, eliminating σ2 as asolution.

Hence, limn→∞ sn = σ1 = 12 (1 +

√5). ♠

9For example ε := s−a2

.10For this, we will use the fact that the limit of a product of sequences is the product of the

limits, i.e. if lim sn → s and lim tn → t, then lim(sntn)→ st.

11The quadratic formula, learned from principals, is x = −b±√b2−4ac2a

.

13

Page 15: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 9.5

Assume lim tn exists and is defined to be t. Then lim tn+1 = t as well. For alln, we get 2tntn+1 = t2n + 2. Implementing the limit theorems, we see that

2t2 = t2 + 2 =⇒ t = ±√

2.

Since we are given t1 = 1 and the fact that the sequence stays positive, we caneliminate −

√2 and conclude that the limit equals

√2.

EXERCISE 9.11

(a) Show that if lim sn = +∞ and inf{tn : n ∈ N} > −∞, then lim (sn + tn) =+∞.

Let M > 0 and let m = inf {tn : n ∈ N}. We want sn+ tn > M for n > N . Thiswill be sufficed by sn + m > M or sn > M −m for n > N . So we will choosean N : sn > M −m for n > N . Let µ = M −m. From the constraints above,we see that 0 ≤ µ <∞. Since we can always find an N : sn > µ for n > N , wecan conclude that lim (sn + tn) = +∞.

(b) Show that if lim sn = +∞ and lim tn > −∞, then lim (sn + tn) = +∞.

If lim tn > −∞, we can conclude that inf{tn : n ∈ N} exists. We can then usethe same argument as above by letting M > 0 and m = inf {tn : n ∈ N}. Wewant sn+tn > M for n > N . This will be sufficed by sn+m > M or sn > M−mfor n > N . So we will choose an N : sn > M −m for n > N . Let µ = M −m.From the constraints above, we see that 0 ≤ µ < ∞. Since we can always findan N : sn > µ for n > N , we can conclude that lim (sn + tn) = +∞.

(c) Show that if lim sn = +∞ and if (tn) is a bounded sequence, then lim (sn + tn) =+∞.

Let M > 0. If (tn) is a bounded sequence, then ∃ a constant ω : |tn| ≤ ω ∀n ∈ N,which implies tn is bounded from above, but more importantly for this proof,bounded from below by −ω. We want sn + tn > M for n > N . This will besufficed by sn − ω > M or sn > M + ω for n > N . So we will choose anN : sn > M + ω for n > N . Let µ = M + ω. From the constraints above, wesee that 0 ≤ µ <∞. Since we can always find an N : sn > µ for n > N , we canconclude that lim (sn + tn) = +∞.

EXERCISE 9.12 (a)

Choose a : L < a < 1. If ε = a−L > 0, there exists an N : ∀n ≥ N | sn+1sn−L| ≤

ε∣∣∣ sn+1sn

∣∣∣ < L+ε = a < 1. More specifically, |sN+1| < a |sN | , |sN+2| < a∣∣sN+1

∣∣ <a2 |sN | . . .. Thus, by induction, |sN+n| < an |sN | ∀ n ∈ N . To summarize,

limn→∞

|sn| = limn→∞

|sN+n| ≤ limn→∞

an |sN | = |sN | limn→∞

an = 0

14

Page 16: STANFORD MATH 113 Elementary-Analysis-Solutions

given |a| < 1.

EXERCISE 9.15

Let sn = an

n! =⇒ sn+1sn

= a(n+1) → 0 as n→∞ =⇒ lim sn = 0

EXERCISE 10.1 (b), (d) and (f)

(b) (−1)n

n2 . The first term, −1, is less than the second term, 14 . Hence, the

sequence is not nonincreasing. The second term is greater than the third term,− 1

9 . Hence, the sequence is not nondecreasing. Since∣∣∣ (−1)n

n2

∣∣∣ =∣∣n−2

∣∣ ≤ 1 ∀n ∈N, the sequence is bounded.

(d) sin(nπ7

). The first term is positive, however, the 7th term is 0 =⇒ the

sequence is not nondecreasing. The second term is greater than the first termand hence the sequence is not nonincreasing. Since |sinx| ≤ 1 ∀x, the sequenceis bounded.

(f) n3n . The sequence is nonincreasing since an+1

an< 1. As all terms are positive,

the sequence is bounded from below and since the sequence is nonincreasing, itis bounded from above =⇒ the sequence is bounded.

EXERCISE 10.2

Claim:

All bounded monotone sequences converge.

Proof :

Let (sn) be a bounded nonincreasing sequence. Let S be defined as the set{sn : n ∈ N} and let v = inf S. S bounded =⇒ v exists and is real. We willnow show that lim sn = u. Let ε > 0. Since v + ε is not an upper bound for S,∃N : sN < v + ε. Since sn is nonincreasing, =⇒ sN ≥ sn ∀n > N . sn ≥ v ∀nand hence, n > N =⇒ v + ε > sn ≥ v =⇒ |sn − v| < ε =⇒ lim sn = v, asdesired. ♠

EXERCISE 10.5

Claim: If (sn) is an unbounded non-increasing sequence, then lim sn = −∞.

We want to show that

lim sn = −∞⇐⇒ for each M < 0 ∃ a number N : n > N =⇒ sn < M.

15

Page 17: STANFORD MATH 113 Elementary-Analysis-Solutions

Proof :

Let (sn) be an unbounded non-decreasing sequence. Let M < 0. Since the set{sn : n ∈ N} is unbounded and it is bounded above by s1, it must be unboundedbelow, since for this sequence to be non-increasing, the condition sn ≥ sn+1 mustbe fulfiilled. Hence, for some N ∈ N we have sN < M . Clearly, n > N =⇒M > sN ≥ sn, so

lim sn = −∞,as desired. ♠

EXERCISE 10.6

(a) Let (sn) be a sequence such that

|sn+1 − sn| < 2−n ∀n ∈ N

Claim: (sn) is a Cauchy sequence and hence a convergent sequence.

Proof :

Given an arbitrary ε > 0, for n > N , choose N : 2−N < ε2 . Then ∀m > n > N ,

we can see that

|sm − sn| ≡

∣∣∣∣∣m−1∑k=n

sk+1 − sk

∣∣∣∣∣ .Now, ∣∣∣∣∣

m−1∑k=n

sk+1 − sk

∣∣∣∣∣ ≤m−1∑k=n

|sk+1 − sk| ≤m−1∑k=n

2−k (11)

*Note: The second inequality from line (1) is justified by ”Let (sn) be a sequencesuch that |sn+1 − sn| < 2−n ∀n ∈ N” in the first line of the problem.

m−1∑k=n

2−k ≤∞∑k=N

2−k = 2−N+1 < ε, (12)

as desired. ♠(b) Is the result in (a) true if we only assume that |sn+1 − sn| < 1

n ∀n ∈ N?

If we only assume that |sn+1 − sn| < 1n ∀n ∈ N, our immediate reaction is to

say that the new constraint makes the result in (a) false, since the harmonicseries,

∑∞n=1

1n , is taken to be divergent.

As a counter-example, let us choose sn =∑∞n=2

1n−1 , which will still fulfill

|sn+1 − sn| ≤ 1n , but will inevitably diverge. Thus, as we travel down the se-

quence (N →∞), arbitrary sms and sns are not encroaching towards eachother,violating the Cauchy criterion.

16

Page 18: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 10.7

Choose n ∈ N. Construct sn ∈ S : supS − sn < 1n and sn > sn−1 for n > 1 =⇒

sn ≡ an increasing sequence converging to supS.

Pick s1 ∈ S : supS − 1 < supS is not an upper bound of S. Implement induc-tion. Assume s1 < · · · < sn−1 exists. Since supS /∈ S =⇒ sn−1 < supS =⇒∃ sn ∈ S : supS ≥ sn > sn−1 and supS − sn < 1

n . This is possible given thatneither sn−1 < supS nor supS − 1

n < supS is an upper bound for S.

EXERCISE 10.8

Let (sn) be a non-decreasing sequence of positive numbers and define

σn =1n

(s1 + s2 + · · ·+ sn)

Claim: (σn) is a non-decreasing sequence.

Proof :

From the assumption of sn being a non-decreasing sequence, we will proclaim

sn ≤ sn+1

Hence,

nsn ≤ nsn+1 (13)=⇒ (s1 + s2 + · · · sn) ≤ nsn ≤ n(sn+1) (14)

=⇒ n(s1 + s2 + · · · sn) + (s1 + s2 + · · · sn) ≤ n(s1 + s2 + · · · sn) + n(sn+1)(15)

=⇒ (n+ 1)(s1 + s2 + · · · sn) ≤ n(s1 + s2 + · · · sn + sn+1) (16)

=⇒ 1n

(s1 + s2 + · · · sn) ≤ 1n+ 1

(s1 + s2 + · · · sn + sn+1),

(17)

Thus,σn ≤ σn+1,

as desired. ♠

EXERCISE 11.8

(a) Use Definition 10.6 and Exercise 5.4 to prove that lim inf sn = −lim sup(−sn).

It is readily observable that

lim sup(−sn) = limN→∞

sup {(−sn) : n > N} = limN→∞

−inf {(sn) : n > N} = −lim inf(sn).

17

Page 19: STANFORD MATH 113 Elementary-Analysis-Solutions

Hence,

lim sup(−sn) = −lim inf(sn)⇐⇒ lim inf(sn) = −lim sup(−sn),

as desired.

(b) Let (tk) be a monotonic subsequence of (−sn) converging to lim sup(−sn).Show that (−tk) is a monotonic subsequence of (sn) converging to lim inf sn.Observe that this completes the proof of Corollary 11.4.

We are given a monotonic subsequence of (−sn), denoted (tk). Then

tk = −snk ∀k, (tk monotonic)⇐⇒ −tk = snk ∀k, (−tk monotonic).

When taken with the results from part (a), the desired result is apparent.

EXERCISE 11.9

(a) Let (xn) be a convergent sequence such that a ≤ xn ≤ b ∀n ∈ N =⇒a ≤ limn→∞ xn ≤ b =⇒ [a, b] is a closed set.

(b) No. We know that the set of any subsequential limits of any set mus beclosed. The interval (0, 1) is not closed.

EXERCISE 11.10

Let (sn) be the sequence of numbers in Figure 11.2 listed in the indicated order.

(a) Find the set S of subsequential limits of (sn).

Laying out the numbers listed in the indicated order gives the sequence

1,12, 1, 1,

12,

13,

14,

13,

12, 1, 1, ...

Rearranging,

1/1,1/2, 1/1,1/1, 1/2, 1/3,1/4, 1/3, 1/2, 1/1,1/1, 1/2, 1/3, 1/4, 1/5,1/6, 1/5, 1/4, 1/3, 1/2, 1/1,1/1, 1/2, 1/3, 1/4, 1/5, 1/6, 1/7,1/8, 1/7, 1/6, 1/5, 1/4, 1/3, 1/2, 1/1,

It seems that ∃ 2 sequences, both monotonic, one increasing the other decreas-ing. The even rows of the pyramid represent a monotonically increasing se-quence and the odd rows correspond to a monotonically decreasing sequence.Upon further inspection, we see that with respect to the decreasing sequences,

18

Page 20: STANFORD MATH 113 Elementary-Analysis-Solutions

the denominator of the last entry corresponds to the row number of the pyramidfor which the decreasing sequence is located, i.e., the last entry of row 3 has13 as its last entry. Conversely, with respect to the increasing sequences, thedenominator of the entry for which the increasing sequence commences corre-sponds to the row number of the pyramid for which the increasing sequence islocated, i.e., the first entry of row 6 is 1

6 . Since the odd rows all commence at11 and progressively converge to 1

n , we can say that one subsequential limit is0. Conversely, since the even rows all seem to converge to 1 from a fractionprogressively closer to 1

n , we could say the other subsquential limit is 1.

Hence the set of subsequential limits is

S ={

1n

: n ∈ N}⋃{

0}

EXERCISE 12.1

Let (sn) and (tn) be sequences and suppose that there exists N0 : sn ≤ tn ∀n >N0.

Claim:

lim inf sn ≤ lim inf tn and lim sup sn ≤ lim sup tn.

Proof :

Refer to Exercise 4.8 (b). From there, we see that

sup {sn : n > N} ≤ inf {tn : n > N} ∀N ≥ N0 =⇒limN→∞

sup {sn : n > N} ≤ limN→∞

inf {tn : n > N} ⇐⇒

lim sup(sn) ≤ lim inf(tn) (∗)

Since

lim inf(sn) ≤ lim sup(sn) andlim inf(tn) ≤ lim sup(tn),

the desired inequalities are obtained implementing (*), as desired. ♠

19

Page 21: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 12.2

Prove that lim sup |sn| = 0 if and only if lim sn = 0.

We will use Theorem 11.7 throughout this proof.

Claim: lim sup |sn| = 0⇐⇒ limsn = 0.

Proof :

“⇐= ” :limsn = 0⇒ lim|sn| = 0⇒ lim supsn = 0.12

“ =⇒ ” :lim sup |sn| = 0⇒ lim inf |sn| = 0, since |sn| ≥ 0.

As N → ∞, lim sup |sn| → lim sn, since Theorem 11.7 tells us that lim sup |sn|is exactly the largest subsequential limit of sn. Hence,

lim sup |sn| = 0⇐⇒ lim sn = 0,

as desired. ♠

EXERCISE 12.3

(a) lim inf sn + lim inf tn = 0 + 0 = 0

(b) lim inf(sn + tn) = 1

(c) lim inf sm + lim sup tn = 0 + 2 = 2

(d) lim sup(sn + tn) = 3

(e) lim sup sn + lim sup tn = 2 + 2 = 4

(f) lim inf sntn = 0

(g) lim sup sntn = 2

EXERCISE 12.4

Claim:

lim sup(sn + tn) ≤ lim sup sn + lim sup tn for bounded (sn) and (tn).

Proof :

For any n > N ,

sn + tn ≤ sup {sn : n > N}+ sup {tn : n > N} =⇒sup {(sn + tn) : n > N} ≤ sup {sn : n > N}+ sup {tn : n > N} .

12By Theorem 11.7.

20

Page 22: STANFORD MATH 113 Elementary-Analysis-Solutions

Hence,

limN→∞

sup {(sn + tn) : n > N} ≤ limN→∞

(sup {sn : n > N}+ sup {tn : n > N})

= limN→∞

sup {sn : n > N}+ limN→∞

sup {tn : n > N} ,

where the last equality holds since(sup {sn : n > N}

)∞N=N0

and(

sup {tn : n > N})∞N=N0

are both convergent and monotonic, as desired. ♠

EXERCISE 12.8

Let (sn) and (tn) be bounded sequences of nonnegative numbers. Prove that

lim sup sntn ≤ (lim sup sn)(lim sup tn).

Give an example where the strict inequality holds.

Claim:lim sup sntn ≤ (lim sup sn)(lim sup tn).

Proof :

Since sn and tn are both bounded sequences of nonnegative numbers,

sn ≤ sup {sn : n > N} ∀n ∈ N

andtn ≤ sup {tn : n > N} ∀n ∈ N,

by definition. This implies

(sn)(tn) ≤ (sup {sn : n > N})(sup {tn : n > N}) ∀n ∈ N,

by multiplicativity and since, once again, sn and tn are both bounded andnonnegative. Now since the right hand side of the inequality serves as an upperbound for the left hand side,

sup {sntn : n > N} ≤ (sup {sn : n > N})(sup {tn : n > N}) ∀n ∈ N

If sn and tn are both sequences which converge to s and t with sn ≤ tn ∀n ∈ N,we can take limits and conclude

lim sup sntn ≤ (lim sup sn)(lim sup tn),

as desired. ♠

21

Page 23: STANFORD MATH 113 Elementary-Analysis-Solutions

An example where the strict inequality would hold would be the case when snand tn both converge to s and t, respectively, but sn < tn ∀n ∈ N.

EXERCISE 12.10

Claim:

(sn) is bounded ⇐⇒ lim sup |sn| < +∞Proof :

lim sup = +∞ =⇒ ∃ a subsequence (snk) : lim |snk | = +∞ =⇒ this subsequenceis unbounded. Conversely, (sn) unbounded =⇒ for any k ∈ N ∃ snk : |snk | > k.Assume that n1 < n2 < · · · < nk < · · · and thus get a subsequence (snk) :lim |snk | = +∞, i.e., lim sup |sn| = +∞. ♠

EXERCISE 12.11

Claim:

lim inf∣∣∣∣sn+1

sn

∣∣∣∣ ≤ lim inf |sn|1n

Proof :

Let α = lim inf |sn|1n . Let L = lim inf

∣∣∣ sn+1sn

∣∣∣. We need to show that α ≥ L forany L1 < L.

L = lim inf∣∣∣∣sn+1

sn

∣∣∣∣ = limn→∞

(inf{∣∣∣∣sn+1

sn

∣∣∣∣ : n > N

)}> L1

=⇒ ∃ N ∈ N : inf{∣∣∣∣sn+1

sn

∣∣∣∣ : n ≥ N}> L1

=⇒ ∀n ≥ N we have∣∣∣∣sn+1

sn

∣∣∣∣ > L1

It then follows that

|sn| > Ln−N1 |sn| , ∀n > N

|sn| > Ln1(L−N1 |sn|

)︸ ︷︷ ︸=α

, ∀n > N

|sn| > Ln1α, ∀n > N =⇒ |sn|1n > L1α

1n , ∀n > N

Since limn→∞ α1n = 1, the result is that lim sup |sn|

1n ≥ L1. ♠

22

Page 24: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 13.3 (a)

Let B be the set of all bounded sequences x = (x1, x2, ...) and define d(x,y) =sup {|xj − yj | : j = 1, 2, ...}.Show that d is a metric for B.

To show that d is indeed a metric for B, we need to show that it satisfies thethree conditions of Definition 13.1. As we will see the first 2 conditions aretrivial.

D1. d(x, x) = 0, since sup {|xj − xj | : j = 1, 2, ...} = sup {0} = 0, andd(x, y) > 0, since for bounded sequences x = (x1, x2, ...) and y = (y1, y2, ...),sup {|xj − yj | : j = 1, 2, ...} > 0.

D2. d(x,y) = sup {|xj − yj | : j = 1, 2, ...} = sup {|yj − xj | : j = 1, 2, ...} =d(y,x).

D3. If x, y, z ∈ B, then for each j = 1, 2, ..., k,

d(x, z) = sup {|xj − zj | : j = 1, 2, ...} (18)= sup {|xj − yj + yj − zj | : j = 1, 2, ...} (19)= sup {|(xj − yj) + (yj − zj)| : j = 1, 2, ...} (20)≤ sup {|xj − yj | : j = 1, 2, ...}+ sup {|yj − zj | : j = 1, 2, ...} (21)≤ d(x, y) + d(y, z), (22)

where the inequality follows from the triangle inequality and the nature of thesupremum. Hence d is a metric.

EXERCISE 13.6

We will prove each part of Proposition 13.9 in turn.

(a) Claim:

The set E is closed if and only if E = E−.

Proof :

”=⇒”

Assume E is closed =⇒ E is a closed set containing E =⇒ the intersection ofall closed sets containing E ≡ E =⇒ E− = E.

”⇐=”

Assume E− = E. By definition, E is a closed set since any intersection of closedsets is closed (by Definition 13.8). E ≡ closed set =⇒ E is closed. ♠

23

Page 25: STANFORD MATH 113 Elementary-Analysis-Solutions

(b) Claim:

The set E is closed if and only if it contains the limit of every convergentsequence of points in E.

Proof :

Assume E is closed and for purposes of contradiction, assume the limit of everyconvergent sequence of points /∈ E =⇒ ∃ a convergent sequence of points ∈E =⇒ for some arbitrary limit point in E ∃ some ε-ball around the limit pointsuch that all points are inside the ball. But, it then becomes a problem (acontradiction) to have a sequence from E converge to its limit point if for someε there does not exist elements of E contained in such a ball. Hence, all limitpoints of E must be contained in E− ≡ E, as desired. ♠(c) Claim:

An element is in E− if and only if it is the limit of some sequence of points inE.

Proof :

”=⇒”

Let x ∈ E− be arbitrary. If this x is not the limit point =⇒ ∃ some ε-ballaround this x: x is the only element of E− in this ball =⇒ Eo is not the interiorof E because any (s 6= x ∈ S) ∈ the aforementioned ε-ball is not contained inEo, which is a contradiction. Now if x ∈ E− \ Eo =⇒ we can create a similarε-ball around x: some of the interior of E is missing, another contradiction =⇒any such point in E− is a limit point.

”⇐=”

see part (b) above, as desired, lol. ♠(d) Claim:

A point is in the boundary of E if and only if it belongs to the closure of bothE and its complement.

Proof :

Let E ≡ Eo =⇒ S \ E closed =⇒ S \ E contains its boundary =⇒ any s ∈boundary of S \E must also be in the boundary of E, otherwise either ∃ somex: x is neither ∈ E or S \ E, or x ∈ both E and S \ E, which both contradictS \ E ≡ Ec. Let E ≡ E− =⇒ S \ E is open =⇒ S \ E does not contain itsboundary =⇒ any s ∈ boundary of S \ E must also be ∈ the boundary of Eand ∈ E =⇒ S \ E ≡ Ec, as desired. ♠

24

Page 26: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 13.8 (b)

We will verify each assertion step by step.

1. In Rk, open balls {x : d(x,x0) < r} are open sets. We can verify this byshowing that every point in the set is interior to the set. Choose an arbitrary x1

in the set. Then |x0|+r represents the boundary. We can then take (|x0|+r)−|x1|2 ,

which represents the distance between the boundary and x1, but divided by 2.Hence all of the points in the set are interior to the set.

2. In Rk, closed balls {x : d(x,x0) ≤ r} are closed sets. We want to show thatthe complement of closed balls are by definition open sets =⇒ closed balls areclosed sets. WLOG define our closed ball to be

Bε(0) := {x : d(x, 0) ≤ ε} ,

WLOG, define the complement of the closed ball by

Bε(0) := {x : d(x, 0) > ε} .

To show the complement is open, let α be arbitrary. Define δ := the magnitudeof α - ε. Define a new ε′-ball, with radius:= δ

2 . All points in this new ball arecontained in the complement of our closed ball.

3. The boundary of each of these sets is

{x : d(x,x0) = r}

The boundary of each of these sets is the boundary of the ball with radius r.By Proposition 13.9 (d),

A point is in the boundary of E if and only if it belongs to the closureof both E and its complement.

Consider the ”neighborhood” of this set

{x : d(x,x0) < r} = (x− r,x + r).

whose closure, the intersection of all closed sets containing (x − r,x + r), is[x− r,x + r]. The complement of {x : d(x,x0) < r} is defined to be {x : d(x,x0) ≥ r}.Hence,

[x− r,x + r] ∩ {x : d(x,x0) ≥ r} = {x : d(x,x0) = r} .

Assertion 1{(x1, x2) : x1 > 0} is open.

Discussion: Noting that any element in this set is, by definition, interior to it,we see this is an open set.

25

Page 27: STANFORD MATH 113 Elementary-Analysis-Solutions

Assertion 2{(x1, x2) : x1 > 0 and x2 > 0} is open.

Discussion: Once, again, noting the strict inequalties, this defines an open set.Conversely, if ∃ non-strict inequalities, i.e., >→≥, we can note that the set willinclude its limit points, hence closing the set.

Assertion 3

{(x1, x2) : x1 > 0 and x2 ≥ 0} is neither open nor closed.

Discussion: This set fails to be open or closed. As counterexamples, consider(1, 0) which is not interior to this set and (0, 0), which is a limit point, but isnot ∈ the set.

EXERCISE 13.10

(a) We want to show that the interior of the set{

1n : n ∈ N

}= ∅.

Claim:

int

{1n

: n ∈ N}

= ∅

Proof :

Define a ”neighborhood” in this set as:

1n+ 1

≤ 1n≤ 1n− 1

.

Now define an ε-ball with radius

r =1n −

1n+1

2.

This makes ε-balls around an arbitrary element in the set such that only thearbitrary element is contained in the ball, as desired. ♠(b) Using the fact that Q contains “gaps”,13 we want to show that

int(Q) = ∅

Proof :

Let q ∈ Q and r > 0 be arbitrary. Consider the ”neighborhood”

{s ∈ R : |s− q| < r} = (s− r, s+ r)

13This is resolved by the Completeness Axiom.

26

Page 28: STANFORD MATH 113 Elementary-Analysis-Solutions

By the ”denseness” of the irrationals14, we know that ∃ an irrational numberz ∈ (s − r, s + r) =⇒ the ”neighborhood” (s − r, s + r) is not contained in Q.But s was given as arbitrary =⇒ q cannot be defined as an interior point. Butq was given as arbitrary =⇒ interior of Q = ∅, as desired. ♠(c) Claim:

The interior of the Cantor set is empty.

Proof:

By the definition of the Cantor Set outlined in the text, the Cantor Set ≡ theintersection of closed sets =⇒ the Cantor Set is closed. Now define the different”articulations” (labeled Tn) of the Cantor Set as follows:

T0 = [0, 1], T1 = [0, 1] \(

13,

23

), ..., T∞

15 = ∩ of all Tn

Recalling the above, T∞ is closed =⇒ T∞ ≡ T∞, the closure of the Cantor Set.

Now,int(T∞) =

(T c∞)c

and note that we are still in the metric space [0, 1]. Thus, T c∞ = [0, 1] \ T∞ =⇒T c∞ = ([0, 1] \ T∞) ∪ {x ∈ [0, 1] : x is a limit point of [0, 1] \ T∞}. Since T∞ ⊆{x ∈ [0, 1] : x is a limit point of [0, 1] \ T∞} ⊆ [0, 1], we know

T c∞ = ([0, 1] \ T∞) ∪ T∞ ≡ [0, 1].

In other words, the closure of the Cantor Set is T0, the interval from 0 to 1.Hence,

int(T∞) =(T c∞)c ≡ [0, 1]c ≡ ∅,

as desired. ♠

EXERCISE 13.12

(a) Let (S, d) be any metric space.

Claim:

If E is a closed subset of a compact set F , then E is also compact.

Proof :

Let E be a closed subset of a compact set F . Let a collection of open sets U bean open cover for E. E closed =⇒ Ec open. Now, Ec + U ≡ an open cover ofF . Given F compact, ∃ a finite subcover that covers F . Now given E ⊆ F andF covered by some finite subcover, E ≡ compact. ♠

14For every real numbers x and y, with x < y, ∃ a rational α and an irrational β such thatx < α < y and x < β < y.

15T∞ ≡ THE Cantor Set

27

Page 29: STANFORD MATH 113 Elementary-Analysis-Solutions

(b) Let (S, d) be any metric space.

Claim:

The finite union of compact sets ∈ S is compact.

Proof :

Let C1, C2, ..., Cn : n <∞ be compact sets, each with a finite subcover Si : i =1, 2, ..., n. Since each Si is the union of open sets and contains finite open sets, itis open. Now consider the union of all of these finite subcovers (S1∪S2∪· · ·∪Sn),which trivially covers the union of C1, C2, ..., Cn. The union of all of these finitesubcovers (S1 ∪ S2 ∪ · · · ∪ Sn) is, by definition, open, since it is the union ofopen sets and further, it is a finite collection of open covers, since it is a unionof finitely many finite subcovers. Hence, since ∃ a finite subcover of the unionof C1, C2, ..., Cn =⇒ the union of C1, C2, ..., Cn ≡ compact. ♠

EXERCISE 13.13

We will show that inf E belongs to E and the case for the sup E is similar.

Claim: If E is a nonempty subset of R, then inf E ∈ E.

Proof :

Assume, by contradiction, that inf E /∈ E. Since E is nonempty and compact,we know, by the Heine-Borel Theorem, that a subset E of Rk is compact if andonly if it is closed and bounded. Since E is closed and bounded, ∃ a sequence(sn) in E where inf E = lim sn = sup E. Furthermore, if the set E is closed,this implies that it contains the limit of every covergent sequence of points inE, including inf E, a contradiction. ♠

Claim: If E is a nonempty subset of R, then sup E ∈ E.

Proof :

Assume, by contradiction, that sup E /∈ E. Since E is nonempty and compact,we know, by the Heine-Borel Theorem, that a subset E of Rk is compact if andonly if it is closed and bounded. Since E is closed and bounded, ∃ a sequence(sn) in E where inf E = lim sn = sup E. Furthermore, if the set E is closed,this implies that it contains the limit of every covergent sequence of points inE, including sup E, a contradiction. ♠Alternatively, both of these proofs can be combined into one proof.

Assume E is a compact subset of R. We know, by the Heine-Borel theorem E isclosed and bounded. Call Lu and Gl the least upper and greatest lower bound,respectively. Then we know Lu ≡ supE and Gl ≡ infE. Now consider the se-quences Lu− 1

n and Gl+ 1n which are clearly ∈ E. Furthermore, limLu− 1

n = Luand limGl + 1

n = Gl. E closed =⇒ Lu, Gl ∈ E. But then by the definition of

28

Page 30: STANFORD MATH 113 Elementary-Analysis-Solutions

Lu and Gl, (supE, infE) ∈ E. ♠

EXERCISE 14.2

(a)∑

n−1n2 = 16

∑nn2 −

∑1n2 =

∑1n −

∑1n2 . Since

∑1n diverges, the entire

series diverges.

(b) The series∑

(−1)n fails to converge because it doesn’t satisfy the Cauchycriterion17. In other words, the terms of the sequence an do not arbitrarily growcloser to eachother as n→∞.

(c)∑

3nn3 =

∑3n2 = 3

∑1n2 , which we know is 3 times a convergent series, thus

the series converges.

(d) If an = n3

3n , then an+1/an = (n+1)33n

3n+1n3 , so lim|an+1/an| = 13 . Hence the

series converges by the Ratio Test.

(e) If an = n2

n! , then an+1/an = (n+1)2n!(n+1)!n2 , so lim|an+1/an| = 0. Hence the series

converges by the Ratio Test.

(f) If an = 1nn =

(1n

)n, then lim sup|an|1n = 0. Hence the series converges by

the Root Test.

(g) If an = n2n , then an+1/an = (n+1)2n

2n+1n , so lim|an+1/an| = 12 . Hence the series

converges by the Ratio Test.

EXERCISE 14.4

(a) We will use induction and the Comparison Test to show that the series

∞∑n=2

1[n+ (−1)n]2

converges.

To accomplish this task we need to show

n+ (−1)n ≥ 12n,

so that∞∑n=2

1[n+ (−1)n]2

≤∞∑n=2

4n2

= 4∞∑n=2

1n2,

since, by the Comparison Test, this would show the series is convergent.16This single series has been split into 2 seperate series. The rule I am follow-

ing is that the sum of two series will converge is both of the sums converge. Hencethe series will diverge, if we can show that one of the sums diverges. Reference:http://www.sosmath.com/calculus/series/examples/examples.html

17Proof ommitted, as it wasn’t required.

29

Page 31: STANFORD MATH 113 Elementary-Analysis-Solutions

The summand index commences at n = 2, so this will serve as the inductionbasis.

2 + (−1)2 ≥ 12

(2) =⇒ 3 ≥ 1,

which is trivial.

Let us now implement the induction step, n + 1, and show the inequality stillholds.

(n+ 1) + (−1)n+1 ≥ 12

(n+ 1) (23)

⇐⇒ n+ 1 + (−1)n(−1) ≥ 12n+

12

(24)

⇐⇒ n+ 1− (−1)n ≥ 12n+

12

(25)

⇐⇒ 12n+

12− (−1)n ≥ 0 (26)

12n+

12≥ (−1)n, (27)

which holds for n ≥ 2, as desired. Hence,

n+ (−1)n ≥ 12n (28)

⇐⇒ (n+ (−1)n)2 ≥(

12n

)2

(29)

⇐⇒ 1(n+ (−1)n)2

≤ 1(12n)2 =

4n2

(30)

⇐⇒∞∑n=2

1[n+ (−1)n]2

≤∞∑n=2

4n2

= 4∞∑n=2

1n2, (31)

as desired.

(b) Since∑[√n+ 1−

√n]

=∑ 1√

n+ 1 +√n≥∑ 1

2√n+ 1

≥∑ 1

2√

2n=

12√

2

∑ 1√n,

which is a divergent ”p-series”18, the series diverges.

(c) We will show the series ∑ n!nn

18A ”p-series” is a series of the form∑ 1

np. Such a series converges if p > 1 and diverges if

p ≤ 1.

30

Page 32: STANFORD MATH 113 Elementary-Analysis-Solutions

converges via the Ratio Test. We want to show

lim∣∣∣∣an+1

an

∣∣∣∣ < 1.

lim∣∣∣∣an+1

an

∣∣∣∣ = lim∣∣∣∣ (n+ 1)!(n+ 1)n+1

· nn

n!

∣∣∣∣ = lim∣∣∣∣ n!(n+ 1)nn

(n+ 1)n(n+ 1)n!

∣∣∣∣ = lim∣∣∣∣ nn

(n+ 1)n

∣∣∣∣ = lim∣∣∣∣( n

n+ 1

)n∣∣∣∣ .Since

lim∣∣∣∣(n+ 1

n

)n∣∣∣∣ = lim∣∣∣∣(1 +

1n

)n∣∣∣∣ =⇒ e,

this suffices to show that

lim∣∣∣∣( n

n+ 1

)n∣∣∣∣ =⇒ 1e< 1.

Hence, the series converges by the Ratio Test.

EXERCISE 14.7; assume p ∈ Z : p > 1

We want to show that if we have a known convergent series∑an and raise it

to a power p > 1, it will simply converge quicker. An example would be theconvergent p-series. We know that ∑ 1

np

converges for values of p > 1. Now if the series is further raised by a power ofp > 1 the original p will be even greater, and thus wll still be convergent.

Claim:∑an := convergent and an ≥ 0 ∀n ∈ N =⇒

∑(an)p converges.

Proof :

If we know that∑an is a convergent series, then

limn→∞

an = 0 =⇒ ∃N ∈ N : an ∈ (0, 1) ∀n ≥ N =⇒ (an)p < an ∀p > 1 and ∀n ≥ N

(an)p < an ∀p > 1 and ∀n ≥ N =⇒∑

(an)p <∑

an ∀p > 1 and ∀n ≥ N.

Hence,∑

(an)p converges by the Comparison Test, as desired. ♠

31

Page 33: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 14.10

Consider this series:∞∑n=0

2(−1)n+n

EXERCISE 15.4

Determine which of the following series converge.

(a)∑∞n=2

1√n logn

. Since log n <√n for large n =⇒ 1√

n< 1

logn =⇒ 1n <

1√n logn

=⇒∑∞n=2

1n <

∑∞n=2

1√n logn

, and hence is divergent by comparisonwith the harmonic series.

(b)∑∞n=2

lognn . This problem trivially diverges when compared19 to the har-

monic series for values of n > 1.

(c)∑∞n=4

1n(logn)(log logn) . Implement the integral Test.

∫∞n=4

1n(logn)(log logn)dn

can be evaluated with a substitution. Let u = log log n. The integral nowbecomes∫ log log∞

log log 4

1udu =

[log u

]log log∞

log log 4= log log log∞− log log log 4 =∞,

hence the series diverges by the integral test.

(d)∑∞n=2

lognn2 . The integral of logn

n2 is −(logn+1)n , so implementing the integral

test yields:

limn→∞

−(log x+ 1)x

∣∣∣∣∣n

2

=− log nn

− 1n

+log 2

2+

12,

which converges to log 22 + 1

2 using L’Hospital’s Rule. Hence, since the integralconverges to an existent finite number, the series converges.

EXERCISE 15.6

(a) A divergent series∑an for which

∑a2n converges is the harmonic series,

∞∑n=1

1n.

(b) If∑an is a convergent series of nonnegative terms, then

∑a2n also con-

verges. This can be shown with a similar method as the proof done in ProblemNo. 3.

19Note: the index on the series begins at n = 2 =⇒ logn > 1 for n ≥ 2.

32

Page 34: STANFORD MATH 113 Elementary-Analysis-Solutions

Claim:∑an := convergent and an ≥ 0 ∀n ∈ N =⇒

∑(an)2 converges.

Proof :

If we know that∑an is a convergent series, then

limn→∞

an = 0 =⇒ ∃N ∈ N : an ∈ (0, 1) ∀n ≥ N =⇒ (an)2 < an ∀n ≥ N

(an)2 < an ∀n ≥ N =⇒∑

(an)2 <∑

an ∀n ≥ N.

Hence,∑

(an)2 converges by the Comparison Test, as desired. ♠(c) Consider this series:

∞∑n=1

(−1)n√n

.

EXERCISE 17.5

(a) We will use induction to prove f(x) = xm is continuous. This will be doneby first showing that when m = 1 we are dealing with f(x) = x, which willbe taken as continuous (proof is provided in the Appendix). Then by assumingthat f(x) = xm is continuous for some m ∈ N, we will show our inductionstep, f(x) = xm+1 is continuous, by noting that f(x) = xm+1 = xmx is hencecontinuous given Theorem 17.4 (ii), which will then imply that f(x) = xm iscontinuous for m ∈ N.

Claim:

If m ∈ N, then the function f(x) = xm is continuous on R.

Proof :

This proof will use mathematical induction. For m = 1, f(x) = x, whichwill be taken as continuous (proof provided in the Appendix). f(x) = x willthen serve as our induction basis. We can now assume our induction hypothesis,f(x) = xm is continuous. But we need to show continuity holds for the inductionstep, m + 1. But we know f(x) = xm+1 ≡ xm · x, which is the product of twocontinuous functions, our inductions hypothesis, and our induction basis, andis hence continuous by Theorem 17.4 (ii). ♠(b) Claim:

Every polynomial function p(x) = a0 + a1x+ · · ·+ anxn is continuous on R.

Proof :

Given the solution from part (a), p(x) is simply the sum and product of con-tinuous functions and is hence continuous, by Theorem 17.4 (i) and Theorem17.3, as desired. ♠

33

Page 35: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 17.6

Claim:

Every rational function is continuous.

Proof :

A rational function is composed of constants, f(x) = c and the continuous func-tion f(x) = x by multiplication, addition and division. Since f(x) = x andf(x) = c are trivially continuous =⇒ rational functions are continuous by thecontinuity theorems of sums, products and quotients of continuous functions, asdesired. ♠

EXERCISE 17.7 (b)

Claim:

|x| is a continuous function on R.

Proof :

|x| is continuous at any x0 since it coincides with x for x > 0 and −x for x < 0.At x = 0, the function f(x) = |x| is continuous because for any δ > 0 we have:|x− 0| < ε =⇒ |f(x)− f(0)| = |x| < ε. ♠

EXERCISE 17.8

(a) Claim:

min(f, g) =12

(f + g)− 12|f − g|

Proof :

Case 1 : Let f(x) ≤ g(x). Then

min(f, g) = f(x) =12

(f + g)− 12

(g − f) (32)

=12

(f + g)− 12|f − g| (33)

Case 2 : Let f(x) ≥ g(x). Then

min(f, g) = g(x) =12

(f + g)− 12

(f − g) (34)

=12

(f + g)− 12|f − g| . (35)

Hence,

min(f, g) =12

(f + g)− 12|f − g| ,

as desired. ♠

34

Page 36: STANFORD MATH 113 Elementary-Analysis-Solutions

(b) Claim:min (f, g) = −max (−f,−g)

Proof :

Case 1 : Let f(x) ≤ g(x) =⇒ −f(x) ≥ −g(x)=⇒ min (f, g) = f(x) = − (−f(x)) = −max(−f,−g)

Case 2 : Let f(x) ≥ g(x) =⇒ −f(x) ≤ −g(x)=⇒ min (f, g) = g(x) = − (−g(x)) = −max(−f,−g).

Hence,min (f, g) = −max (−f,−g) ,

as desired. ♠(c) Claim:

f and g continuous at x0 ∈ R =⇒ min(f, g) is continuous at x0

Proof :

Recall Theorems 17.3 and 17.4 (i):

Theorem 17.3: Let f be a real-valued function with dom(f) ⊆ R.If f is continuous at x0 in dom(f), then |f | and kf , k ∈ R, arecontinuous at x0.

Theorem 17.4 (i): Let f and g be real-valued functions that arecontinuous at x0 ∈ R. Then f + g is continuous at x0.

In combination with the results from part (a), i.e.,

min(f, g) =12

(f + g)− 12|f − g| ,

we see that min(f, g) is simply the sum, difference and composition of functionswhich are continuous at x0, and hence is itself continuous at x0, as desired. ♠.

EXERCISE 17.9

(a) Claim:f(x) = x2 is continuous at x0 = 2

Proof :

Let ε > 0 be given. We want to show that |x2 − 4| < ε provided |x − 2| issufficiently small, i.e., less than some δ. We observe that |x2− 4| = |(x+ 2)(x−2)| ≤ |x + 2| · |x − 2|. We need to get a bound for |x − 2| that doesn’t dependon x. We notice that if |x − 2| < 1, say, then |x + 2| < 5, so it suffices to get|x−2| ·5 < ε. So by setting δ = min

{1, ε5}

, we see that f(x) = x2 is continuousat x0 = 2, as desired. ♠

35

Page 37: STANFORD MATH 113 Elementary-Analysis-Solutions

(b) Claim:f(x) =

√x is continuous at x0 = 0

Proof :

Let ε > 0 be given. We want to show that |√x − 0| < ε provided |x − 0|

is sufficiently small, i.e., less than some δ. We observe that |√x − 0| =

√x.

Since we want this to be less than ε, we set δ = ε2. Then |x − 0| < δ implies√x <√δ = ε, so

|x− 0| < δ =⇒ |f(x)− f(0)| < ε,

as desired. ♠(c) Claim:

f(x) = xsin(

1x

)for x 6= 0 and f(0) = 0 is continuous at x0 = 0

Proof :

Let ε > 0 be given. We want to show that |xsin(

1x

)− 0| < ε provided |x − 0|

is sufficiently small, i.e., less than some δ. We see that |xsin(

1x

)− 0| ≤ x ∀x.

Since we want this to be less than ε, we set δ = ε. Then |x − 0| < δ impliesx < δ = ε, so

|x− 0| < δ =⇒∣∣∣∣xsin

(1x

)− 0∣∣∣∣ < ε,

as desired. ♠(d) Claim:

g(x) = x3 is continuous at x0 arbitrary

Proof :

By the hint given,

x3 − x30 = (x− x0)(x2 + x0x+ x2

0) = (x− x0)[(x2 − x0)2 + 3xx0)

]and we know that

|x| = |x− x0 + x0| ≤ |x− x0|+ |x0|,

by the triangle inequality. Hence,

|g(x)− g(x0)| = |x3 − x30| ≤ |x− x0|(|x− x0|2 + 3|x− x0||x0|+ 3x2

0).

Now let ε > 0 be given. We show that ∃ a δ = δ(x0, ε) > 0 such that

|x− x0| < δ =⇒ |g(x)− g(x0)| < ε.

36

Page 38: STANFORD MATH 113 Elementary-Analysis-Solutions

Let δ = min(

1, ε3 ,ε

g|x0|+1 ,ε

gx20+1

). Then |x− x0| < δ implies

|g(x)− g(x0)| ≤ δ(δ2 + 3δ|x0|+ 3x20) ≤ 20δ(1 + 3|x0|+ 3x2

0)

= δ + 3δ|x0|+ 3δx20 <

ε

2+ 3|x0|

ε

g|x0|+ 1+ 3x2

0

ε

g|x0|2 + 1,

≤ ε

3+ε

3+ε

3= ε,

as desired. ♠.

Alternatively,

Assume g(x) = x2, x0 arbitrary. Observe that (x3−x30) = (x−x0)(x2+x0x+x2

0).Now we will make the assumption that |x − x0| < 1 =⇒ |x| < |x0| + 1, whichenables us to see

|x2 +x0x+x20| ≤ |x2|+ |x0x|+ |x2

0| < |x0|2 +2|x0|+1+ |x0| · ||x0|+1|+ |x20| ≡ k,

where the material to the right of the last inequality is all greater than 0.

We can then set δ = min{

1, εk}

. Hence,

|x−x0| < δ =⇒ |g(x)−g(x0)| = |(x−x0)(x2+x0x+x20)| < |x−x0||k| <

ε

k·k = ε,

as desired. ♠

EXERCISE 17.12

(a) Let f be a continuous real-valued function with domain (a, b).

Claim:

If f(r) = 0 for each rational number r ∈ (a.b), then f(x) = 0 ∀x ∈ (a.b).

Proof :

We are given f(x) = 0 ∀x ∈ Q. If x ∈ R \ Q, then ∃ a sequence of rationalnumbers, (rn), which converges to x. Hence, by continuity, rn → x =⇒ f(rn)→f(x). But f(rn) = 0 ∀n, given the conditions in the claim, so 0 → f(x) =⇒f(x) = 0 ∀x ∈ (a, b), as desired. ♠(b) Let f and g be continuous real-valued functions on (a, b) : f(r) = g(r) foreach rational number r ∈ (a, b).

Claim:

f(x) = g(x) ∀x ∈ (a, b).

Proof :

Using the limit concept of sequences, for any x ∈ (a, b) ∃ a sequence of rationalnumbers, (rn) : limfn = x =⇒ f(x) = limf(rn) = limg(rn) = g(x).

20δ ≤ 1

37

Page 39: STANFORD MATH 113 Elementary-Analysis-Solutions

Hence, f(x) = g(x), as desired. ♠

EXERCISE 17.13 (b)

Let h(x) = x ∀x ∈ Q and h(x) = 0 ∀x ∈ R \Q.

Claim:

h is continuous at x = 0 and no other point.

Proof :

For any ε > 0, if |x− 0| < ε, then |h(x)− h(0)| is either 0 (if x is ∈ R\Q) or |x|(if x ∈ Q, and thus in both cases < ε). Thus h is continuous at x = 0. ∀ otherx, consider two sequences with limit x, one (rn) ∈ Q, and another, (xn) ∈ R\Q.Then limh(xn) = 0 and limh(rn) = x 6= 0. Hence, ∃ disconituity for h at x 6= 0,as desired. ♠

EXERCISE 17.14

Claim:

f is continuous at each point of R \Q and discontinuous at each point of Q.

Proof :

For x = pq ∈ Q, define a sequence xn ∈ R\Q : limxn = x =⇒ lim f(xn) = 0, but

f(x) = 1q 6= 0 =⇒ f is discontinuous at x. For an irrational x and any ε > 0, let

δ > 0 be defined as the distance from x to the closest irreducible fraction pq with

denominator q ≤ 1ε . Then for any x′ : |x′ − x| < δ =⇒ |f(x′)− f(x)| < 0 =⇒ f

continuous at x ∈ R \Q, as desired. ♠

EXERCISE 18.2

The limit x0 (or y0) of the subsequence xnk ∈ (a, b) (or ynk ∈ (a, b)) may bean endpoint a or b of the interval and thus lie outside of the domain of thefunction.)

38

Page 40: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 18.4

Let S ⊆ R and suppose there exists a sequence (xn) in S that converges to anumber x0 /∈ S.

Claim:

∃ an unbounded continuous finction on S.

Proof :

Let x0 /∈ S. We are given a sequence (xn) ∈ S which converges to x0. Then|xn − x0| := the distance to x0 is continuous and strictly positive on S. Definef := 1

|xn−x0| =⇒ f is well-defined and continuous on S =⇒ limn→∞ f = ∞, asdesired. ♠

EXERCISE 18.6

Claim:

x = cosx for some x ∈(0, π2

).

Proof :

We know f(x) = x−cosx is continuous ∈[0, π2

], < 0 at x = 0 and > 0 at x = π

2 .Implement the Intermediate Value Theorem. Hence, ∃ an x ∈

(0, π2

): f(x) = 0,

as desired. ♠

EXERCISE 18.8

Suppose that f is a real-valued function continuous on R and that f(a)f(b) < 0for some a, b ∈ R.

Before commencing, we will state the properties that

0 · a = 0 ∀a ∈ Z

anda · b < 0⇐⇒ a < 0, b > 0 or a > 0, b < 0 ∀a, b ∈ Z

Claim:

∃ an x between a and b : f(x) = 0.

Proof :

Given f(a)f(b) < 0, either

Case 1 : f(a) < 0 =⇒ f(b) > 0 =⇒ f(a) < 0 < f(b), or

Case 2 : f(a) > 0 =⇒ f(b) < 0 =⇒ f(b) < 0 < f(a).

In either case, the Intermediate Value Theorem tells us that ∃ x ∈ (a, b) : f(x) =0, as desired. ♠

39

Page 41: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 18.10

Suppose that f is continuous on [0, 2] and f(0) = 2.

Claim:

x, y ∈ [0, 2] : |x− y| = 1 and f(x) = f(y).

Proof :

Let g(x) = f(x + 1) − f(x) =⇒ g is continuous on [0, 1] and g(0) = f(1) −f(0) = f(1) − f(2) = −g(1). Implement the Intermediate Value Theorem.∃x ∈ [0, 1] : g(x) = 0, as desired. ♠

EXERCISE 19.2

(a) Claim:f(x) = 3x+ 1 is uniformly continuous on R.

Proof :

Let ε > 0 be given. We want |f(x) − f(y)| = |(3x + 1) − (3y + 1)| < ε for|x− y| < δ with x, y ∈ R. We know |(3x+ 1)− (3y+ 1)| = |3x− 3y| = 3|x− y|.Take δ := ε

3 . Then

|x−y| < δ =⇒ |x−y| < ε

3=⇒ 3|x−y| < ε =⇒ |3x−3y| < ε =⇒ |3x−3y+1−1| < ε =⇒

|3x+ 1− 3y − 1| < ε =⇒ |(3x+ 1)− (3y + 1)| < ε =⇒ |f(x)− f(y)| < ε,

as desired. ♠

(b) Claim:f(x) = x2 is uniformly continuous on [0, 3] .

Proof :

Let ε > 0 be given. We want |f(x)− f(y)| = |x2 − y2| < ε for |x− y| < δ withx, y ∈ [0, 3]. We know |x2 − y2| = |(x − y)(x + y)| = |x − y| · |x + y|. Withx, y ∈ [0, 3], let |x + y| ≤ |3 + 3| = 6. Define δ := ε

6 . Then for any x, y ∈ [0, 3]with |x− y| < δ,

|f(x)− f(y)| = |x2 − y2| = |(x− y)(x+ y)| = |x− y| · |x+ y| < |x− y| · 6 = ε,

as desired. ♠(c) Claim:

f(x) =1x

is uniformly continuous on[

12,∞).

40

Page 42: STANFORD MATH 113 Elementary-Analysis-Solutions

Proof :

Let ε > 0 be given. We want |f(x) − f(y)| = | 1x −1y | < ε for |x − y| < δ, with

x, y ∈[12 ,∞

). We know | 1x −

1y | = |

y−xxy | = |(

1xy )(y − x)| = |x − y| · | 1

xy |. Withx, y ∈

[12 ,∞

), let 1

xy ≤1

12 ·

12

= 114

= 4. Define δ := ε4 . Then for any x, y ∈

[12 ,∞

)with |x− y| < δ,

|f(x)− f(y)| =∣∣∣∣ 1x − 1

y

∣∣∣∣ =∣∣∣∣y − xxy

∣∣∣∣ = |x− y| ·∣∣∣∣ 1xy

∣∣∣∣ < |x− y| · 4 = ε,

as desired. ♠.

EXERCISE 19.4(a) Claim:If f is uniformly continuous on a bounded set S, then f is a bounded functionon S.Proof :

By contradiction. Assume that f is uniformly continuous on a bounded set Sand an unbounded function on S. Then ∃ a sequence, (xn) in the domain off such that |f(xn)| ≥ n. By the Bolzano-Weierstrass Theorem, the sequencecontains a convergent subsequence (xnk), since the domain is bounded. A con-vergent subsequence is Cauchy (by definition) and hence the sequence of valuesf(xnk) is Cauchy by the property of uniformly continuous functions. But thesequence |f(xnk)| ≥ nk is unbounded, contradicting our assumption in the out-set of this proof. Hence, if f is uniformly continuous on a bounded set S, thenf is a bounded function on S, as desired. ♠

(b) Claim:1x2

is not uniformly continuous on (0, 1).

Proof :

We want to show that if 1x2 is not a bounded function on (0, 1), then 1

x2 is notuniformly continuous on (0, 1)21. So it will suffice to show that

f(x) =1x2

is not a bounded function on (0, 1).

Let M > 1 be arbitrary =⇒ 0 < 1M2 < 1. We want to show that 1

x2 > M forsome x ∈ (0, 1) =⇒ x < 1√

M. Let x = 1√

M+1, since 0 < 1√

M+1< 1√

M< 1. But

then for f(x) = 1x2 , we see that

1x2

=1(1√M+1

)2 = M + 1 > M,

21This is the contrapositive of what was proved in part (a).

41

Page 43: STANFORD MATH 113 Elementary-Analysis-Solutions

which shows that f(x) = 1x2 is not a bounded function on (0, 1), as desired. ♠

EXERCISE 19.6

(a) Claim:

f(x) =√x is uniformly continuous on (0, 1], although f ′ is unbounded.

Proof :

f ′(x) = 12√x→ +∞ as x → 0 =⇒ f ′(x) is unbounded. f continuous on

the closed interval [0, 1] =⇒ f uniformly continuous on [0, 1] =⇒ f uniformlycontinuous on the subset (0, 1], as desired. ♠Let f(x) =

√x for x ≥ 1.

(b) Claim:

f is uniformly continuous on [1,∞).

Proof :

Let ε > 0 be given. We want |f(x)− f(y)| = |√x−√y| < ε for |x− y| < δ with

x, y ∈ R. We know |√x−√y| = |x−y|

|√x+√y| . Take δ := 2ε. Then

|x− y| < δ =⇒ |x− y| < 2ε =⇒ |x− y|2

≤ |x− y||√x+√y|

= |√x−√y| < ε,

as desired. ♠

EXERCISE 19.10

The limit

limx→0

g(x)x

= limx→0

x sinx1x

= 0,

in other words, g is differentiable at x = 0 (g′(0) = 0). At x 6= 0,

g′(x) = 2x sin1x

+ x2 cos1x·(− 1x2

)= 2x sin

1x− cos

1x,

which is bounded. Since |cos y| 6= 1and |sin y| ≤ |y| ∀y =⇒ for y = 1x :∣∣∣∣2x sin

1x− cos

1x

∣∣∣∣ =∣∣∣∣2y sin y − cos y

∣∣∣∣ ≤ 2 + 1 = 3.

Thus, g′ is both bounded and defined on R =⇒ g is uniformly continuous.

42

Page 44: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 20.11 (b)

Claim:

limx→b

√x−√b

x− b=

12√b

Proof :

√x−√b

x− b=

(√x−√b) ·

√x+√b√

x+√b

x− b=

x−b√x+√b

x− b=

1√x+√b

Hence,

limx→b

1√x+√b−→ 1√

b+√b

=1

2√b

EXERCISE 20.14

Claim:limx→0+

1x

= +∞ and limx→0−

1x

= −∞

We will split this up into 2 cases/claims.

Claim 1 :limx→0+

1x

= +∞

Proof :

Let M > 0 and δ := 12M . Then 0 < x < 0 + δ =⇒ 0 < x < δ =⇒

f(x) = 1x >

1δ = 1

12M

= 2M > M . Hence, limx→0+1x = +∞, as desired. ♠

Claim 2 :limx→0−

1x

= −∞

Proof :

Let M < 0 and δ := −12M . Then 0 − δ < x < 0 =⇒ −δ < x < 0 =⇒

f(x) = 1x <

−1δ = 1

12M

= 2M < M . Hence, limx→0−1x = −∞, as desired. ♠

EXERCISE 20.16

Suppose that the limits L1 = limx→a+ f1(x) and L2 = limx→a+ f1(x) exist.

(a) Claim:

If f1(x) ≤ f2(x) ∀x in some interal (a, b), then L1 ≤ L2.

43

Page 45: STANFORD MATH 113 Elementary-Analysis-Solutions

Proof :

By contradiction. Let f1(x) ≤ f2(x) and assume L1 > L2. Corollary 20.8 says

limx→a+

f(x) = L⇐⇒ ∀ε > 0 ∃ δ > 0 : a < x < a+ δ =⇒ |f(x)− L| < ε

Define ε := 12 (L1 − L2) =⇒ for ε > 0 :

∃ δ1 : x ∈ (a, a+ δ1), f1(x) > L1 − ε = L1 −[

12

(L1 − L2)]

=12

(L1 + L2)

∃ δ2 : x ∈ (a, a+ δ2), f2(x) < L2 + ε = L2 +[

12

(L1 − L2)]

=12

(L1 + L2).

Let δ := min {δ1, δ2} =⇒

f1(x) >12

(L1 + L2) > f2(x),

which is a contradiction. Hence, if f1(x) ≤ f2(x) ∀x in some interal (a, b), thenL1 ≤ L2, as desired. ♠(b) Suppose f1(x) < f2(x) ∀x in some interval (a, b). Can you conclude thatL1 < L2.

Consider f1(x) = x and f2(x) = 2x, x ∈ (a, b) =⇒ f1(x) < f2(x) ∀x ∈ (a, b).Consider a = 0 and b = 1 =⇒ (a, b) = (0, 1). We can plainly see that

limx→0

f1(x) = limx→0

f2(x) = 022,

a contradiction =⇒ we cannot conclude that L1 < L2 if f1(x) < f2(x) ∀x insome interval (a, b).

EXERCISE 20.18

Claim:

For f(x) =√

1 + 3x2 − 1x2

, limx→0

f(x) =32.

Proof :

Non-formal. Rearranging f(x),√

1 + 3x2 − 1x2

=√

1 + 3x2 − 1x2

·√

1 + 3x2 + 1√1 + 3x2 + 1

=(1 + 3x2)− 1

(√

1 + 3x2 + 1)x2=

3√1 + 3x2 + 1

,

which we see is a composition of continuous functions which happen to behavewell near x = 0. Hence, limx→0 f(x)→ f(0) = 3

2 , as desired. ♠

22Since both f1 and f2 extend continuously to 0 where they take on the value 0.

44

Page 46: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 21.6

Let (S1, d1), (S2, d2), and (S3, d3) be metric spaces.

Claim:

f : S1 → S2 and g : S2 → S3 continuous =⇒ g ◦ f continuous from S1 into S3

Proof :

From both the claim and Theorem 21.3, f−1(U) is an open subset of S1 for everyopen subset U of S2, i.e., f−1(U) = {s ∈ S1 : f(s) ∈ U} and g−1(V ) is an opensubset of S3 for every open subset V of S3, i.e., g−1(V ) = {s ∈ S3 : f(s) ∈ V }.Define a new mapping, h(S1) = g ◦f : S1 → S3. Then h−1(V ) is an open subsetof S1 for every open subset V of S3, i.e. h−1(V ) = {s ∈ S1 : h(S) ∈ V }, by thecontinuity of composite functions outlined in Theorem 17.5 and the fact thatg and f are both coninuous throughout S. Thus, by Theorems 17.5 and 21.3,h(S1) := (g ◦ f)(S1) : S1 → S3 is continuous, as desired. ♠

EXERCISE 21.10 (b) and 21.11 (b)

For 21.10 (b), in order to show ∃ continuous functions mapping (0, 1) → R,consider

f(x) =log(2x)1− x

As x → 0, f(x) → −∞ and as x → 1, f(x) → +∞. We see f(x) is the com-position of continuous functions and the denominator 6= 0 through f ’s domainand is hence continuous, as desired.

For 21.11 (b), we can use informal contradiction:

f : [0, 1] → R =⇒ R is compact, since [0, 1] is compact, by Theorem 21.4 (i).But R is unbounded and thus cannot be compact, which is a contradiction.Hence, there do not exist continuous functions mapping [0, 1] onto R.

EXERCISE 23.1 (b), (d), (f) and (h)

(b)∑(

xn

)n =∑

xn

nn =∑(

1n

)nxn. If an =

(1n

)n, then lim sup |an|1n = 0.

Therefore, β = 0, R = +∞ and this series has a radius of convergence +∞ andhence an interval of convergence of (−∞,+∞).

(d)∑(

n33n

)xn. If an = n3

3n , then an+1an

= (n+1)3

3n+1 · 3n

n3 , so lim∣∣∣an+1an

∣∣∣ = 13 . There-

fore, β = 13 and R = 3. This series diverges for both x = 3 and x = −3, hence

the radius of convergence is 3 and the interval of convergence is (−3, 3).

(f)∑(

1(n+1)22n

)xn. If an = 1

(n+1)22n , then an+1an

= (n+1)22n

(n+2)22n+1 , so lim∣∣∣an+1an

∣∣∣ =12 . Therefore, β = 1

2 and R = 2. This series converges at both x = 2 and

45

Page 47: STANFORD MATH 113 Elementary-Analysis-Solutions

x = −2, hence the radius of convergence is 2 and the interval of convergence is[−2, 2].

(h)∑(

(−1)n

n2·4n

)xn. If an = (−1)n

n2·4n , then an+1an

= (−1)n+1

(n+1)2·4n+1 · n2·4n

(−1)n , so lim∣∣∣an+1an

∣∣∣ =14 . Therefore, β = 1

4 and R = 4. This series converges at both x = 4 and x = −4,hence the radius of convergence is 4 and the interval of convergence is [−4, 4].

EXERCISE 23.2

(a)∑√

nxn. If an =√n, then an+1

an=√n+1√n

, so lim∣∣∣√n+1√

n

∣∣∣ = 1. Therefore,β = 1 and R = 1. This series diverges at x = 1 and x = −1, hence the radiusof convergence is 1 and the interval of convergence is (−1, 1).

(b)∑

1n√nx

n. If an = 1n√n , then |an|

1n =

∣∣∣∣ 1

n1√n

∣∣∣∣ and lim sup∣∣∣∣ 1

n1√n

∣∣∣∣ = 1.

At x = 1, the series converges with comparison to∑

1np with p > 1 and for

x = −1, the series diverges by the alternating series theorem since the limit of|an| approaches 1 and not 0. Hence, the radius on convergence is 1 and theinterval of convergence is (−1, 1].

(c)∑xn!. When |x| ≥ 1, the series diverges since limxn! does not tend to

0 as n → ∞. Now for |x| < 1, the series converges absolutely by comparisonwith

∑|x|m23. Therefore, β = 1 and R = 1. This series diverges at x = 1 and

x = −1, hence the radius of convergence is 1 and the interval of convergence is(−1, 1).

(d)∑

3n√nx2n+1. For this series, the radius of convergence is:

R =1

lim(

3n√n

) 12n+1

=1√3

lim(3n)1

4n+2 =1√3.

When x = ± 1√3, the series explodes in both directions24. Hence, the radius of

convergence is 1√3

and the interval of convergence is(−1√

3, 1√

3

).

EXERCISE 23.6 (b)

An example of such a series is ∑n>0

(−x)n

n.

23Note that∑xn! =

∑amxm with am = 1 when m = n! and am = 0 when m 6= n!.

24The series turns into∑± 1√

3n.

46

Page 48: STANFORD MATH 113 Elementary-Analysis-Solutions

This series converges to − ln(1 + x) when |x| < 1, diverges at x = −1 and con-verges at x = +1.

EXERCISE 23.8

fn(x)→ 0 since∣∣n−1 sinnx

∣∣ 6= 1n → 0 as n→∞. But f ′n(x) = cosnx→ (−1)n

when x = π, which has no limit.

EXERCISE 24.2

For x ∈ [0,∞), let fn(x) = xn .

(a) For fn(x) = xn , limn→∞ fn(x) = limn→∞

xn = 0. Hence, f(x) = 0.

(b) YES.

Claim: fn → f uniformly on [0, 1].

Proof :

Let ε > 0 be given. Let N := 1ε . Then, for n > N , |fn(x) − 0| =

∣∣ xn

∣∣ ≤ 1ε

n =1nε < ε, as desired. ♠(c) NO.

Claim: fn does not converge uniformly to f on [0,∞).

Proof :

By contradiction. Using the above, let ε := 1 =⇒ ∃ an N :∣∣ xn

∣∣ < 1 ∀n >N =⇒ |x| < n ∀n > N =⇒ |x| < n + ξ for some ξ > 0. But since x ∈ [0,∞),x is unbounded, contradicting |x| < n + ξ for some ξ > 0. Hence, fn does notconverge uniformly to f on [0,∞), as desired. ♠

EXERCISE 24.6

Let fn(x) =(x− 1

n

)2 for x ∈ [0, 1].

(a) YES.

Claim:

fn(x) =(x− 1

n

)2 for x ∈ [0, 1] converges pointwise on the set [0, 1].

Proof :

fn(x) =(x− 1

n

)2 = x2 − 2xn + 1

n2 With x ∈ [0, 1], let n grow arbitrarily large=⇒ fn(x) −→ x2, since with n large, 2x

n −→ 0 and 1n2 −→ 0 =⇒ fn(x) −→

f(x) := x2. Hence, given x arbitrary, fn(x) converges pointwise for x ∈ [0, 1],as desired. ♠The limit function will thus be f(x) = x2.

(b) YES.

47

Page 49: STANFORD MATH 113 Elementary-Analysis-Solutions

Claim:

fn(x) =(x− 1

n

)2 for x ∈ [0, 1] converges uniformly on the set [0, 1].

Proof :

Let ε > 0 be given and N := 1ε2 . Then for x ∈ [0, 1], we have:∣∣∣∣∣

(x− 1

n

)2

− x2

∣∣∣∣∣ =∣∣∣∣1− 2xn

n2

∣∣∣∣ ≤ 1√n<

1√N

= ε,

as desired. ♠.

EXERCISE 24.14

Let fn(x) = nx1+n2x2

(a) Claim:

fn −→ 0 pointwise on R.

Proof :

Let x ∈ R. Let ε > 0 be arbitrary. Let N := max{x2,√x√ε

}. Then for n > N ,

|fn − 0| =∣∣∣∣ nx

1 + n2x2

∣∣∣∣ =∣∣∣∣ x

1n + nx2

∣∣∣∣ ≤ ∣∣∣∣ x1n + n2

∣∣∣∣ ≤ ∣∣∣ xn2

∣∣∣ ≤ x

N2< ε.

Hence, fn(x) converges uniformly to zero, as desired. ♠(b) Claim:

fn(x) does not converge to 0 uniformly on [0, 1].

Proof :

Implement Remark 24.4. Hence,

f ′n(x) =n

1 + n2x2− nxn22x

(1 + n2x2)2=n+ n3x2 − 2n3x2

(1 + n2x2)2=

n− n3x2

(1 + n2x2)2

Assume, for contradiction, n−n3x2

(1+n2x2)2 → 0, which implies x2 = 1n2 which implies

x = ± 1n . But we see that fn

(1n

)= 1

2 6= 0, a contradiction, since x ∈ [0, 1]satisfies x = ± 1

n . Hence, fn(x) does not converge to 0 uniformly on [0, 1] asdesired. ♠(c) Claim:

fn(x) converges to 0 uniformly on [1,∞).

48

Page 50: STANFORD MATH 113 Elementary-Analysis-Solutions

Proof :

Implement Remark 24.4. With x ∈ [1,∞), x is unable to always satisfy ± 1n .

Using methods from calculus we can see that fn(x) = nx1+n2x2 assumes its max-

imum at x = 1, which is ∈ [1,∞). Since fn(1) = n1+n2 → 0 as n → ∞, fn(x)

converges to 0 uniformly on [1,∞) as desired. ♠

EXERCISE 25.2

Let fn(x) = xn

n .

Claim: (fn) is uniformly convergent on [−1, 1].

Proof :

Let ε > 0 be given. Let N := 1ε . Then for n > N and ∀x ∈ [−1, 1],

|fn(x)− f(x)| =∣∣∣∣xnn − 0

∣∣∣∣ ≤ 1n

25 ≤ 1N

= ε,

as desired. ♠The limit function is fn(x)→ f(x) = 0 for large n.

EXERCISE 25.6

(a) Show that if∑|ak| < ∞, then

∑akx

k converges uniformly on [0, 1] to acontinuous function.

Implement Theorem 25.5 and the Weierstrass M-test. Since∑|ak| <

∞ and that∑akx

k ≤∑ak because x ∈ [0, 1], we know

∑akx

k convergesuniformly on S. Now since the series converges uniformly on S and akx

k iscontinuous, then

∑akx

k represents a continuous function on S.

(b) Does∑∞n=1

1n2x

n represent a continuous function on [−1, 1]?

This is a series which converges at both x = −1 (by the alternating series test)and at x = 1 (convergent p-series). Now consider the interval −1 ≤ a ≤ 1and note that

∑∞n=1

1n2 a

n converges. Since |n−2xn| ≥ |n−2an| =(an

n2

)for

x ∈ [−a, a], the Weierstrass M-test shows that the series∑∞n=1

1n2x

n convergesuniformly to a function on [−a, a]. Since |a| can be any number ≤ 1, we con-clude that f represents a continuous function on [−1, 1]26.

25This function takes its max value at x = 1.26Note:

∑∞n=1

∣∣∣ 1n2 x

n∣∣∣ ≤ π2

6for x ∈ [−1, 1].

49

Page 51: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 25.14

Claim:

If∑gk converges uniformly on a set S and if h is a bounded function on S,

then∑hgk converges uniformly on S.

Proof :

If the series∑gk converges uniformly on a set S, then

∑gk is uniformly Cauchy

on S. If h is a bounded function on S, then ∃ an M : |h| ≤ M . Let ε > 0 begiven and let N := ε

M . Then

n ≥ m > N =⇒

∣∣∣∣∣n∑

k=m

hgk

∣∣∣∣∣ ≤∣∣∣∣∣n∑

k=m

Mgk

∣∣∣∣∣ = M

∣∣∣∣∣n∑

k=m

gk

∣∣∣∣∣ ≤M · εM = ε,

as desired. ♠

EXERCISE 27.2

Show that if f is continuous on R, then there exists a sequence (pn) of polyno-mials such that pn → f uniformly on each bounded subset of R. Hint : Arrangefor |f(x)− pn(x)| < 1

n for |x| ≤ n.

Claim:

If f is continuous on R, then there exists a sequence (pn) of polynomials suchthat pn → f uniformly on each bounded subset of R.

Proof :

Given |x| ≤ n, suppose I = [−n, n], a closed and bounded interval; hence x ∈ Iand f : I −→ R is a continuous function. Let g : [0, 1] −→ [−n, n] be a bijectivemap defined by g(x) = −n+ x(2n), and hence continuous, i.e., g(0) = −n andg(1) = n. Since f is continuous, the composite function, f ◦ g : [0, 1] −→ Ris continuous. Hence, for any ε > 0,∃N > 0 such that for any n ≥ N , theBernstein Polynomial Bn(f ◦ g) satisfies

|(f ◦ g)(x)−Bn(f ◦ g)(x)| < ε ∀x ∈ [0, 1]

Now g is a continuous injective map and so g has a continuous inverse functiondefined by

g−1(x) =x+ n

2nx ∈ [−n, n]

Thus, for all x ∈ [−n, n], |f(x)−BN (f ◦ g)(g−1(x))| < ε.

Hence, ∣∣∣∣f(x)−BN (f ◦ g)(x+ n

2n

)∣∣∣∣ < ε ∀x ∈ [−n, n]

50

Page 52: STANFORD MATH 113 Elementary-Analysis-Solutions

SinceBN (f◦g) is a polynomial function, pε(x) = BN (f◦g)(x+n2n

)is a polynomial

function in x and |f(x)− pε(x)| < ε ∀x ∈ I. If we let qn(x) = BN (f ◦ g)(x+n2n

),

then

qn(x) = BN (f ◦ g)(x+ n

2n

)=

n∑k=0

f ◦ g(k

n

)(n

k

)(x+ n

2n

)k (1−

(x+ n

2n

))n−k(36)

=n∑k=0

(f ◦ g)(k

n

)(n

k

)(x+ n

2n

)k (n− x

2n

)n−k(37)

=n∑k=0

f

(a+

k

n(2n)

)(n

k

)(x+ n

2n

)k (n− x

2n

)n−k(38)

It then follows from |(f ◦ g)(x) − Bn(f ◦ g)(x)| < ε∀x ∈ [0, 1] that qn −→ funiformly on [−n, n], as desired. ♠

EXERCISE 27.6

Claim:

If Bnf → f uniformly on [0, 1], then f is continuous on [0, 1].

Proof :

If Bn(f)→ f uniformly on [0, 1], then

∀ε > 0 ∃N ∀x ∈ [0, 1] ∀n > N : |Bnf(x)− f(x)| < ε

2

Let ε > 0 be given. Let N := δ. Then x ∈ [0, 1] and |x− x0| < δ =⇒

|f(x)−f(x0)| = |Bnf(x)−f(x0)+f(x)−Bnf(x)| ≤ |Bnf(x)−f(x0)|+|Bnf(x)−f(x)| ≤ ε

2+ε

2= ε,

as desired. ♠

EXERCISE 28.2

Use the definition of the derivative to calculate the derivatives of the followingfunctions at the indicated points.

(a) f(x) = x3 at x = 2.

f ′(2) = limx→2

x3 − 8x− 8

= limx→2

(x− 2)(x2 + 2x+ 4)x− 2

= limx→2

x2+2x+4 = (2)2+2(2)+4 = 12

51

Page 53: STANFORD MATH 113 Elementary-Analysis-Solutions

(b) g(x) = x+ 2 at x = a.

f ′(a) = limx→a

(x+ 2)− (a+ 2)x− a

= limx→a

x+ 2− a− 2x− a

= limx→a

x− ax− a

= 1

(c) f(x) = x2 cosx at x = 0.

f ′(0) = limx→0

x2 cosx− (0)2 cos(0)x− 0

= limx→0

x2 cosx− 0x− 0

= limx→0

x cosx = 0 · 1 = 0

(d) r(x) = 3x+42x−1 at x = 1

r′(1) = limx→1

3x+42x−1 −

71

x− 1= limx→1

−11x+112x−1

x− 1= limx→1

−11x+ 112x− 1

· 1x− 1

= limx→1

−112x− 1

= −11

EXERCISE 28.4

Let f(x) = x2 sin 1x for x 6= 0 and f(0) = 0.

(a) Use Theorems 28.3 and 28.4 to show that f is differentiable at each a 6= 0and calculate f ′(a). Use, without proof, the fact that sinx is differentiable andthat cosx is its derivative.

Using the definition, we see that

f ′(a) = limx→a

x2 sin(

1x

)− a2 sin

(1a

)x− a

= limx→a

x2 sin(

1x

)− sin

(1a

)x− a

+ sin(

1a

)x2 − a2

x− a,

where thesin( 1

x )−sin( 1a )

x−a limit represents sin′(

1x

). We are given that a 6= 0,

which gives the function sin(

1a

)some meaning in terms of differentiability and

hence, using Theorems 28.3 and 28.4 and the given fact that sin′ x = cosx,

f ′(a) = a2 sin′(

1a

)+ sin

(1a

)(a2)′ = a2 cos

(1a

)(−1a2

)+ 2a sin

(1a

)= 2a sin

(1a

)− cos

(1a

).

(b) Use the definition to show that f is differentiable at x = 0 and that f ′(0) =0.

limx→0

f(x)− f(0)x− 0

= limx→0

f(x)x

= limx→0

x2 sin 1x

x= limx→0

x sin1x

= 0 · ξ,

where |ξ| ≤ 1, and hence the whole expression equals zero, as desired.

(c) Show that f ′ is not continuous at x = 0.

52

Page 54: STANFORD MATH 113 Elementary-Analysis-Solutions

Implement Theorem 28.4, which implies f is differentiable everywhere. Further,f ′(x) = x sin

(1x

)− cos

(1x

)at x 6= 0. We know x sin

(1x

)→ 0 as x → 0 and

cos(

1x

)doesn’t have a limit at x = 0. Hence, the sum f ′(x) has no limit at

x 6= 0 and thus is discontinuous at x = 0.

EXERCISE 28.8

Let f(x) = x2 for x rational and f(x) = 0 for x irrational.

(a) Claim:

f is continuous x = 0.

Proof :

Let ε > 0 be given. Let |x− 0| <√ε. Then f(x) is either equal to x2 ∈ [0, ε) or

0. In either of these 2 cases, |f(x)− f(0)| < ε =⇒ continuity at 0, as desired. ♠

(b) Claim:

f is discontinuous at all x 6= 0.

Proof :

This will be done in 2 cases.

Case I: x 6= 0, x ∈ QLet ε > 0 be given. Let δ > 0 be given. Further, let ε := x2. Due to thedenseness of the rationals, ∃ an irrational number q in (a− δ, a+ δ). But while|x− q| < δ and |f(x)− f(q)| = ε, δ can be made arbitrarily small (once again,due to the denseness property) =⇒ f is not continuous at x, as desired.♠Case II: x 6= 0, x ∈ R \Q

Let ε > 0 be given. Let 0 < δ < |x|2 be given. Further, let ε := x2

10 . Due tothe denseness of the irrationals, ∃ a rational number q in (a− δ, a+ δ). By thetriangle inequality, |q| > |x|

2 =⇒ f(q) = x2

4 =⇒ |f(x)− f(q)| = x2

4 > ε. Since δcan be made arbitrarily small =⇒ f is not continuous at x, as desired.♠

(c) Claim:

f is differentiable at x = 0.

Proof :

Let x = 0 and a 6= 0. Then limx→af(x)−f(a)

x−a = f(a)a , which will will equal a if

a ∈ Q and 0 otherwise. Both cases show that the limit→ 0 as a→ 0, and hencef is differentiable at x = 0 with derivative equal to 0, as desired. ♠

53

Page 55: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 28.15

Proof of Leibniz’ Rule

Claim:

(fg)(n)(a) =n∑k=0

(n

k

)f (k)(a)g(n−k)(a)

Proof :

By Induction on n27. Let Leibniz’ Rule hold for n = m. Then

(fg)(m+1) = (f ′g + fg′)(m) =m∑k=0

(m

k

)f (k+1)g(m−k) +

m∑k=0

(m

k

)f (k)g(m+1−k)

=m+1∑k=0

[(m

k − 1

)+(m

k

)]f (k)g(m+1−k)

=m+1∑k=0

(m+ 1k

)f (k)g(m+1−k),

where, using Pascal’s triangle28 (from the Binomial Theorem handout), the lastequality holds. Hence, Leibniz’ Rule holds true, as desired. ♠

EXERCISE 29.4

Let f and g be differentiable functions on an open interval I. Suppose that a, bin I satisfy a < b and f(a) = f(b) = 0.

Claim:f ′(x) + f(x)g′(x) = 0 for some x ∈ (a, b).

Proof :

Consider the function h(x) = f(x)eg(x). Since f(a) = f(b) = 0 =⇒ h(a) =h(b) = 0, implementing Rolle’s Theorem, we know ∃ some x ∈ (a, b) : h′(x) = 0.Differentiating h(x) yields:

h′(x) = f(x)g′(x)eg(x) + f ′(x)eg(x)

= eg(x) (f(x)g′(x) + f ′(x)) .

Given h′(x) = 0 for some x ∈ (a, b), either eg(x) has to equal 0 for some x ∈ (a, b)or f(x)g′(x) +f ′(x) has to equal 0 for some x ∈ (a, b). Since we know eg(x) can-not equal 0 for any value g(x), we can safely conclude that f(x)g′(x)+f ′(x) = 0for some x ∈ (a, b), as desired. ♠

27For n = 1, Leibniz’ Rule turns into the product rule, i.e., (fg)′ = f ′g + fg′.28(m+1k

)=( mk−1

)+(mk

).

54

Page 56: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 29.8

Claim:

f is strictly decreasing is f ′(x) < 0 ∀x ∈ (a, b).

Proof :

Consider x1, x2 where a < x1 < x2 < b. By the Mean Value Theorem, for somex ∈ (x1, x2) we have

f(x2)− f(x1)x2 − x1

= f ′(x) < 0.

Since x2 − x1 > 0 and f(x2)− f(x1) < 0 =⇒ f(x2) < f(x1), as desired. ♠

Claim:

f is increasing if f ′(x) ≥ 0 ∀x ∈ (a, b).

Proof :

Consider x1, x2 where a < x1 ≤ x2 < b. By the Mean Value Theorem, for somex ∈ (x1, x2) we have

f(x2)− f(x1)x2 − x1

= f ′(x) ≥ 0.

Since x2 − x1 > 0 and f(x2)− f(x1) ≥ 0 =⇒ f(x2) ≥ f(x1), as desired. ♠

Claim:

f is decreasing if f ′(x) ≤ 0 ∀x ∈ (a, b).

Proof :

Consider x1, x2 where a < x1 ≤ x2 < b. By the Mean Value Theorem, for somex ∈ (x1, x2) we have

f(x2)− f(x1)x2 − x1

= f ′(x) ≤ 0.

Since x2 − x1 > 0 and f(x2)− f(x1) ≤ 0 =⇒ f(x2) ≤ f(x1), as desired. ♠

EXERCISE 29.10

Let f(x) = x2 + x2 sin 1

x at x 6= 0, and f(0) = 0.

Claim:

f ′(0) > 0, but f is not increasing on any interval containing 0. Compare thisresult with Theorem 29.7 (i).

55

Page 57: STANFORD MATH 113 Elementary-Analysis-Solutions

Proof :

We know

f ′(0) = limx→0

f(x)x

= limx→0

[12

+ x sin1x

]=

12

+ 0 > 0

Conversely, at x 6= 0, we get

f ′(x) = 2x sin1x

+12− cos

1x.

The first term → 0 as x → 0. The last term oscillates between +1 and −1infinitely many times in any neighborhood of x = 0. 1

2 − 1 < 0 =⇒ in anyneighborhod of x = 0, f ′ stay < 0 on some intervals. Implement Theorem 29.7.The function f is therefore decreasing on these intervals =⇒ f is not increasingin any neighborhod of x = 0. If f ′ were continuous at x = 0, then it wouldremain positive in some neighborhood of x = 0. Hence, ∃ discontinuity of f ′ atx = 0, as desired. ♠

EXERCISE 29.15

We know that (xm)′ = mxm−1 for m ≥ 0,(

1x

)′ = − 1x2 and

(x

1n

)′= x

1n−1. We

can thus calculate(xmn

)′ using the chain rule as follows:

d

dx

(xmn

)=

d

dx(xm)

1n =

y1n−1

n

∣∣∣y=xm

·mxm−1 =m

nxm( 1

n−1)xm−1 =m

nxmn −1.

EXERCISE 29.18

Let f be be differentiable on R with a = sup {|f ′(x)| : x ∈ R} < 1. Select s0 ∈ Rand define sn = f(sn−1) for n ≥ 1. Thus s1 = f(s0), s2 = f(s1),...etcetra.

Claim:

(sn) is a convergent sequence.

Proof :

Implement the Mean Value Theorem. Then, for each n > 0,

|sn+1 − sn| = |f(sn)− f(sn−1)| = |f ′(y)(sn − sn−1)| ≤ a|sn − sn−1|,

which implies|sn+1 − sn| ≤ an|s1 − s0|, ∀n > 0,

by induction. Further,

|sm+1 − sn| ≤m∑k=n

|sk+1 − sk| ≤ (s1 − s0)m∑k=n

ak, ∀m ≥ n > 0.

56

Page 58: STANFORD MATH 113 Elementary-Analysis-Solutions

We know the geometric series∑ak converges if a < 1 and further, its partial

sums,∑nk=0 form a Cauchy Sequence. Therefore, by the previous estimate, (sn)

is a Cauchy sequence, and hence converges, as desired. ♠

EXERCISE 30.2

Find the following limits if they exist.

(a) limx→0x3

sin x−x = 00 , which is an indeterminant form, so we apply L’Hospital’s

rule and get limx→03x2

cos x−1 = 00 , which is an indeterminant form, so we apply

L’Hospital’s rule and get limx→06x− sin x = −6 · limx→0

xsin x = −6 · 1 = −6.

(b) limx→0−xx3 = 0

0 , which is an indeterminant form, so we apply L’Hospital’srule and get limx→0

sec2 x−13x2 = 0

0 , which is an indeterminant form, so we applyL’Hospital’s rule and get limx→0

2 sec2 x tan x6x = 0

0 , which is an indeterminantform, so we apply L’Hospital’s rule and get limx→0

sec4 x+2 sec2 x tan2 x3 = 1

3

(c) limx→0

[1

sin x −1x

]=∞−∞, which we can rearrange to suffice for an appli-

cation of L’Hospitals rule. limx→0

[1

sin x −1x

]= limx→0

x−sin xx sin x = 0

0 , which is anindeterminant form, so we apply L’Hospital’s rule and get limx→0

1−cos xx cos x+sin x =

00 , which is an indeterminant form, so we apply L’Hospital’s rule and getlimx→0

sin x2 cos x−x sin x = 0

2 = 0.

(d) limx→0(cosx)1x2 = 1∞, which when manipulated, can yield an indetermi-

nant form sufficient for the application of L’Hospital’s rule.

Let y = limx→0(cosx)1x2 . Then ln y = ln limx→0(cosx)

1x2 = limx→0 ln(cosx)

1x2 =

limx→0ln(cos x)x2 = 0

0 , which is an indeterminant form, so we apply L’Hospital’srule and get limx→0

− tan x2x = 0

0 , which is an indeterminant form, so we applyL’Hospital’s rule and get limx→0

− sec2 x2 = −1

2 . Thus ln y = −12 =⇒ eln y =

e−12 =⇒ y = 1√

e. Therefore, the limit as x→ 0 = 1√

e.

EXERCISE 30.4

Let f be a function defined on some interval (0, a), and define g(y) = f(

1y

)for

y ∈ (a−1,∞); here we set a−1 = 0 if a =∞.

Claim:

limx→0+ f(x) exists if and only if limy→∞ g(y) exists, in which case they areequal.

Proof :

In two parts.

“limx→0+ f(x) exists =⇒ limy→∞ g(y) exists”.

57

Page 59: STANFORD MATH 113 Elementary-Analysis-Solutions

Assume limx→0+ f(x) exists and is equal to L. Then for each ε > 0 ∃ δ > 0 : 0 <x < δ =⇒ |f(x)− L| < ε. Now, define y := 1

x . Then 0 < x < δ =⇒ 1δ < y <∞.

Then for f defined on an interval (c,∞), for each ε > 0 ∃ α < ∞ : α < 1y =⇒∣∣∣f ( 1

y

)− L

∣∣∣ < ε =⇒ limy→∞ g(y) exists and is equal to L, as desired.

“ limy→∞ g(y) =⇒ limx→0+ f(x) exists”.

limy→∞ g(y) = limy→∞ f(

1y

), which when implemented with the same trans-

formation, y := 1x =⇒ x = 1

y , we see that limy→∞ f(

1y

)≡ limx→0+ f(x) and

hence limx→0+ f(x) exists and is finite.

EXERCISE 30.7

For x ∈ R, let

f(x) = x+ cosx sinx and g(x) = esin x(x+ cosx sinx).

(a) Since | cosx| ≤ 1 and | sinx| ≤ 1 =⇒ | cosx sinx| ≤ 1, we can conclude

limx→∞

f(x) = x+ cosx sinx ≥ limx→∞

x− 1 = +∞,

and hence, limx→∞ f(x) = x+cosx sinx = +∞. Since we know 1e ≤ limx→∞ esin x ≤

e, this implies

limx→∞

1e

(x+ cosx sinx) ≤ limx→∞

esin x(x+ cosx sinx) ≤ limx→∞

e(x+ cosx sinx)

which implies1e· ∞ ≤ lim

x→∞esin x(x+ cosx sinx) ≤ e · ∞

which implies limx→∞ esin x(x+ cosx sinx) = +∞, by the Squeeze Theorem.

(b) Implement Theorem 28.3 and the trigonometric identity sin2 x+cos2 x = 1.

f(x) = x+ cosx sinxf ′(x) = 1 + cosx(cosx) + sinx(− sinx)

= 1 + cos2 x− sin2 x

= cos2 x+ (1− sin2 x)

= cos2 x+ cos2 x

= 2(cosx)2

g(x) = esin x(x+ cosx sinx)

g′(x) = esin x(2 cos2 x) + (x+ cosx sinx) cosxesin x

= esin x(2 cos2 x) + esin x cosx [f(x)]

= esin x cosx [2 cosx+ f(x)]

58

Page 60: STANFORD MATH 113 Elementary-Analysis-Solutions

f ′(x)g′(x)

=2(cosx)2

esin x cosx [2 cosx+ f(x)]

=2e− sin x cos2 x

cosx [2 cosx+ f(x)]

=2e− sin x cosx2 cosx+ f(x)

(d) Using the fact that | sinx| ≤ 1 and | cosx| ≤ 1, we see that

limx→∞

2 cosxesin x(2 cosx+ x+ cosx sinx)

≤ limx→∞

22e+ xe+ e

≤ limx→∞

2e(3 + x)

→ 0 as x→∞,

so we can conclude that f ′(x)g′(x) tends to 0 for large x. However,

limx→∞

x+ cosx sinxesin x(x+ cosx sinx)

= limx→∞

1esin x

,

which will oscillate between e and 1e as x→∞, which shows that limx→∞

f(x)g(x)

doesn’t exist.

EXERCISE 31.2

sinhx =∑n≥1

x2n−1

(2n− 1)!and coshx =

∑n≥0

x2n

(2n)!.

Both of these results follow from the series expansion for ex ≡∑

xn

n! and thefact that (sinhx)′ = coshx. Convergence for both series can be shown byimplementing the Ratio Test.

lim sup∣∣∣∣ x2n+2

(2n+ 2)!· (2n)!x2n

∣∣∣∣ = lim sup∣∣∣∣ x2

2n(2n+ 1)

∣∣∣∣→ 0 < 1,

lim sup∣∣∣∣ x2n+1

(2n+ 1)!· (2n− 1)!x2n−1

∣∣∣∣ = lim sup∣∣∣∣ x2

(2n+ 1)(2n+ 2)

∣∣∣∣→ 0 < 1.

EXERCISE 32.2

Let f(x) = x for rational x and f(x) = 0 for irrational x.

(a) To calculate the upper Darboux integral for f on the interval [0, b], we needto come up with an upper bound. Consider the partition

P = {0 = t0 < t1 < · · · < tn = b} where tk = bk

nfor each k.

59

Page 61: STANFORD MATH 113 Elementary-Analysis-Solutions

This implies

U(f, P ) =n∑k=1

kb

n· bn

29 =(b

n

)2 n∑k=1

k =(b

n

)2n2 + n

2=

12b2(1 + 2n−1).

12b

2(1 + 2n−1)→ 12b

2 as n→∞ =⇒ U(f) ≤ 12b

2.

We now need to come up with a lower bound. Consider the partition

P = {0 = t0 < t1 < · · · < tn < b}

which implies

U(f, P ) =n∑k=1

tk(tk−tk−1)30 ≥n∑k=1

(tk + tk−1)2

(tk−tk−1)31 =12

n∑k=1

(t2k−t2k−1)32 =12b2

which implies U(f) = 12b

2.

To calculate the lower Darboux integral for f on the interval [0, b], for eachpartition

P = {0 = t0 < t1 < · · · < tn < b} ,

∃ an irrational number in [ti−1, ti] for each i, which implies the minimal valueof f on [ti−1, ti] is 0 =⇒ L(f, P ) = 0 for each P =⇒ L(f) = 0.

(b) Assume b > 0 =⇒ U(f) 6= L(f) =⇒ f is not integrable on [0, b], since theTheorems of this chapter won’t hold for a degenerate interval.

EXERCISE 32.6

Let f be a bounded function on [a, b]. Suppose there exist sequences (Un) and(Ln) of upper and lower Darboux sums for f such that lim(Un−Ln) = 0. Showf is integrable and

∫ baf = limUn = limLn.

We want to show that limLn = limUn. Assume that limLn < limUn. Thenlim(Un − Ln) > 0 which is a contradiction. Observe that Ln ≤ L(f) ≤ U(f) ≤Un. Taking limits of both sides and implementing the Squeeze Theorem yieldslimLn ≤ L(f) ≤ U(f) ≤ limUn = limLn. Thus L(f) = U(f) by the SqueezeTheorem which implies f is integrable by Theorem 32.9.

29Each interval has length bn

; on the interval [ti−1, ti], the maximum value of f is attained

at ti = in

.30Each interval has length b

n; the supremum of f on [ti−1, ti] is ti since you can pick rational

numbers in the interval arbitrarily close to ti.31Since tk > tk−1, tk >

tk+tk−12

.32Note that this is a telescoping series, lol.

60

Page 62: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 33.3 (a)

A function f on [a, b] is called a step-function if ∃ a partition P = {a = u0 < u1 < · · · < um = b}of [a, b] such that f is constant on each interval (uj−1, uj), say f(x) = cj for xin (uj−1, uj).

Claim:

f is integrable.

Proof :

Consider a sub-partition of P , called P ′, where

P ′ ={u0 < u1 < u1 < u1 < · · · < un−1 < un−1 < un−1 < un

},

where ui − ui < ε ∀i.We want to show |U(f, P )− S| < ε and |S − L(f, P )| < ε. From P ′,

U(f, P ′)− L(f, P ′) =∑

(ui − ui) · |ci − ci−1| =∑|ci − ci−1| ≤ ε ·max {ci} ,

which will tend to 0 as ε tends to 0. Hence,

L(f, P ′) ≤n∑j=1

cj(uj − uj−1) ≤ U(f, P ′),

and since L(f, P ′) = U(f, P ′) from above,∫ baf =

∑nj=1 cj(uj − uj−1), as de-

sired. ♠

EXERCISE 33.4

Give an example of a function f on [0, 1] that is not integrable for which |f | isintegrable.

Consider the intervale [0, 1] and let f(x) = 1 for rational x ∈ [0, 1], and letf(x) = −1 for irrational x ∈ [0,−1]. For any partition

P = {0 = t0 < t1 < · · · < tn = 1} ,

we have

U(f, P ) =n∑k=1

M(f, [tk−1, tk]) · (tk − tk−1) =n∑k=1

1 · (tk − tk−1) = 1

and

L(f, P ) =n∑k=1

(−1) · (tk − tk−1) = −1.

61

Page 63: STANFORD MATH 113 Elementary-Analysis-Solutions

It follows that U(f) 6= L(f) which implies that f is not integrable. However, ifthe absolute value of f is taken, we see that f(x) = 1 ∀x ∈ R, and hence willbe integrable since U(f) = L(f).

EXERCISE 33.7

Let f be a bounded function on [a, b], so that ∃ B > 0 : |f(x)| ≤ B ∀x ∈ [a, b].

(a)Claim:

U(f2, P )− L(f2, P ) ≤ 2B[U(|f |, P )− L(|f |, P )

]∀ partitions P of [a, b]

Proof :n∑k=1

[M(f2, [tk−1, tk])−m(f2, [tk−1, tk)

]· (tk − tk−1)

=n∑k=1

([M(|f |, [tk−1, tk]) +m(|f |, [tk−1, tk)] [M(|f |, [tk−1, tk])−m(|f |, [tk−1, tk)]

)· (tk − tk−1)

≤ 2Bn∑k=1

[M(|f |, [tk−1, tk])−m(|f |, [tk−1, tk])

]· (tk − tk−1),

as desired33. ♠.

(b)Claim:

If f is integrable on [a, b], then f2 is integrable on [a, b].

Proof :

From part (a), we know

U(f2, P )− L(f2, P ) ≤ 2B[U(|f |, P )− L(|f |, P )

]∀ partitions P of [a, b],

and by Theorem 32.5, if f is integrable on [a, b], then for ε > 0, U(f, P ) −L(f, P ) < ε.

But,

U(f2, P )−L(f2, P ) ≤ 2B[U(|f |, P )−L(|f |, P )

]< ε =⇒ U(f2, P )− L(f2, P )

2B< ε,

which impliesU(f2, P )− L(f2, P ) < 2Bε.

So define a new ε, ε′ := 2Bε > 0 and we thus have

U(f2, P )− L(f2, P ) < ε′,

33The first equality in this proof is from factoring the perfect squares.

62

Page 64: STANFORD MATH 113 Elementary-Analysis-Solutions

which implies that if f is integrable on [a, b], then f2 is integrable on [a, b], asdesired. ♠

EXERCISE 33.8

Let f and g be integrable functions on [a, b].

(a) Claim:

fg is integrable on [a, b].

Proof :

Implement Theorem 33.3. Let f and g be integrable on [a, b]. Then, by Theo-rem 33.3, f + g and f − g are integrable on [a, b]. Further, if f + g and f − gare integrable on [a, b], then (f + g)2 and (f − g)2 are both integrable on [a, b],as per the results of exercise 33.7 (b). Since 1

4

[(f + g)2 − (f − g)2

]= fg, and

Theorem 33.3 tells us a constant times an integrable function is integrable, thisshows that if f and g are integrable on [a, b], then fg is integrable on [a, b], asdesired. ♠

(b) Claim:

max(f, g) and min(f, g) are integrable on [a, b].

Proof :

Implement Theorem(s) 33.3 and 33.5. We know

max(f, g) =12

(f + g) +12|f − g|

min(f, g) =12

(f + g)− 12|f − g|,

which are compositions of functions and constants integrable on [a, b] and henceare integrable by Theorems 33.3 and 33.5, as desired. ♠.

EXERCISE 33.10

Let f(x) = sin(

1x

)for x 6= 0 and f(0) = 0. Show that f is integrable on [−1, 1].

Let ε > 0 be given. Since f is piece-wise continuous, by Definition 33.7, on [ ε4 , 1],∃ a partition P1 of [ ε4 , 1] : U(f1, P )−L(f1, P ) < ε

2 . Similarly, ∃ a partition P2 of[−1,− ε

4 ] : U(f2, P )− L(f2, P ) < ε2 . Define P = P[1 d P2, a partition of [−1, 1].

Since {M(f,

[− ε

4,ε

4

])−m(f,

[− ε

4,ε

4

])}·{ ε

4−(− ε

4

)}< ε,

which, when combined with Theorem 32.5, shows f is integrable on [−1, 1].

63

Page 65: STANFORD MATH 113 Elementary-Analysis-Solutions

EXERCISE 33.14

Suppose f and g are continuous functions on [a, b] and that g(x) ≥ 0 ∀x ∈ [a, b].

(a) Claim:

∃ x ∈ [a, b] : ∫ b

a

f(t)g(t)dt = f(x)∫ b

a

g(t)dt.

Proof :

Given that ∫ b

a

f(t)g(t)dt =∞∑i=1

f(ti)g(ti) · (ti − ti−1),

where ti ∈ [ui−1, ui] and f(ti) = ai and g(ti) = bi. Further,

aibi ≥ min {ai} · bi =⇒ ai ≥ min {ai}aibi ≤ max {ai} · bi =⇒ ai ≤ max {ai} .

Then, ∫ b

a

f(t)g(t)dt ≥ fmin

∫ b

a

g(t)dt∫ b

a

f(t)g(t)dt ≤ fmax

∫ b

a

g(t)dt.

Note: if bi = 0 then the equality holds, trivially. Hence, by the IntermediateValue Theorem, ∃ x ∈ [a, b] :

∫ baf(t)g(t)dt = f(x)

∫ bag(t)dt., as desired. ♠.

(b) Let g(x) := 1b−a . Then

1b− a

∫ b

a

f(t)dt =∫ b

a

f(t)g(t)dt = f(x) ·∫ b

a

g(t)dt = f(x) · 1 = f(x).

EXERCISE 34.2

(a)

limx→0

1x

∫ x

0

et2dt

To calculate this integral, let us use the formulation

F (x)− F (x0)x− x0

=1

x− x0

∫ x

x0

f(t)dt for x 6= x0

Hence,

f(x0) =1

x− x0

∫ x

x0

f(x0)dt.

64

Page 66: STANFORD MATH 113 Elementary-Analysis-Solutions

Thus, by the Fundamental Theorem of Calculus II,

limx→0

1x

∫ x

0

et2dt = e0

2= 1.

(b)

limh→0

1h

∫ 3+h

3

et2dt

Using the same argument as above,

limh→0

1h

∫ 3+h

3

et2dt = lim

h→0

1h

∫ 3

0

e(t+3)2dt = e32

= e9.

EXERCISE 34.6

Let f be a continuous function on R and define

G(x) =∫ sin x

0

f(t)dt for x ∈ R.

If f is continuous at x ∈ R, then f(sinx) is differentiable on R as a com-position of differentiable functions. Hence, f is differentiable at sinx andG′ = f(sinx) cosx, per the chain rule.

Let ε > 0 and B > 0 : |f(sinx)| ≤ B ∀x ∈ R. Now ∀x ∈ R : | sinx − sin 0| ≤| sinx| < ε, we have |F (sinx)− F (0)| ≤ |

∫ sin x

0f | ≤ | sinx| < ε, which shows G

is continuous.

EXERCISE 36.1

Show that if f is integrable on [a, b] as in Definition 32.1, then

limd→b−

∫ d

a

f(x)dx =∫ b

a

f(x)dx.

It suffices to show that if |f | is bounded by some number B, then∣∣∣∣∣∫ d

a

f(x)dx−∫ b

a

f(x)dx

∣∣∣∣∣ ≤ B(b− d).

Hence, we want to show that for |b− d| < δ =⇒ |B(b− d)| < ε. Choose δ := εB .

Then|B(b− d)| = B|(b− d)| < Bδ = ε.

65

Page 67: STANFORD MATH 113 Elementary-Analysis-Solutions

Hence as d approaches b from the left, the difference between the integrals con-verges to zero, showing they are equivalent.

EXERCISE 36.6

Let f and g be continuous functions on (a, b) : 0 ≤ f(x) ≤ g(x) ∀x ∈ (a, b); acan be −∞ and b can be +∞.

(a) Claim: ∫ b

a

g(x)dx <∞ =⇒∫ b

a

f(x)dx <∞

Proof :

We know ∫ b

a

f(x)dx =∫ 0

a

f(x)dx+∫ b

0

f(x)dx,

by Theorem 33.6. Likewise,∫ b

a

g(x)dx =∫ 0

a

g(x)dx+∫ b

0

g(x)dx.

From the fact that 0 ≤ f(x) ≤ g(x) ∀x ∈ (a, b), we can assume(∫ 0

a

f(x)dx+∫ b

0

f(x)dx

)≤

(∫ 0

a

g(x)dx+∫ b

0

g(x)dx

).

Implementing Definition 36.1 and Theorem 19.6 (Extensions from bounded in-tervals to unbounded intervals being uniformly continuous and hence integrable):(

lima→−∞

∫ 0

a

f(x)dx+ limb→+∞

∫ b

0

f(x)dx

)≤

(lim

a→−∞

∫ 0

a

g(x)dx+ limb→+∞

∫ b

0

g(x)dx

).

Define a continuous function h(x) := g(x) − f(x) =⇒ h(x) ≥ 0, since g(x) ≥f(x). Hence,∫

h(x)dx :=∫ (

g(x)− f(x))dx =⇒

∫g(x)dx =

∫ (h(x) + f(x)

)dx.

We can now see that∫g(x)dx <∞ =⇒

∫ (h(x)+f(x)

)dx <∞ =⇒

∫f(x)dx <∞, since

∫h(x)dx ≥ 0,

as desired. ♠(b) Claim: ∫ b

a

f(x)dx =∞ =⇒∫ b

a

g(x)dx =∞

66

Page 68: STANFORD MATH 113 Elementary-Analysis-Solutions

Proof :

Using the result from part (a),∫f(x)dx =

∫g(x)dx−

∫h(x)dx,

and therefore,∫f(x)dx =∞ =⇒

∫g(x)dx−

∫h(x)dx =∞ =⇒

∫g(x)dx =∞,

as desired. ♠.

67


Recommended