Calculus A: Problems with Solutions (Lecture 1)

Problem 1.1 Show that if a and b are any numbers and a is not equal to 0, then there is one and only
one number x such that a · x = b, and that this number is given by x = b · a−1 .

We are required to show two things: the existence (”there is a number x”) and the uniqueness (”one
and only one”) of a solution.
• Existence: It is enough to check that the number x = b · a−1 satisfies the equation a · x = b.

a · x = a · [b · a−1 ]
= a · [a−1 · b] by axiom (M1), commutativity
−1
= [a · a ]·b by axiom (M2), associativity
= 1·b by axiom (M4), existence of inverse
= b by axiom (M3), existence of 1

Therefore, a · x = b and so x = b · a−1 is a solution.

• Uniqueness: Let y be any solution of the equation a · x = b. Then,

a·y = b
a · (a · y) = a−1 · b
−1
by axiom (M4), existence of inverse
[ −1 ]
a · a · y = a−1 · b by axiom (M2), associativity
−1
1·y = a ·b by axiom (M4), existence of inverse
−1
y = a ·b by axiom (M3), existence of 1
−1
y = b·a by axiom (M1), commutativity

We have shown that if y is a solution then y = x, which means that x is the unique solution.
Actually, the second part (Uniqueness) of the proof is enough to show the statement of the problem
(why?).

Notice that application of axiom (M4) requires a ̸= 0.

Problem 1.2 Using just the ordered field axioms, prove that if ab < 0, then a and b have opposite signs.

There are several methods for proving this statement. We will give here only two of them.
(In this proof we assume that ”a · 0 = 0 ∀a ∈ R” is already proved.)

Method 1: checking all possible cases

According to axiom (O1), there can be only three cases for a: a > 0 or a < 0 or a = 0. In each case
we show that the statement of the problem holds.
 • a > 0: If we can show that this implies a−1 > 0, then we have

ab < 0
−1
a (ab) < a−1 0 by axiom (O4)
−1
(a a)b < 0 by axiom (M2)
b < 0 by axioms (M4) and (M3),

hence a and b have opposite signs.

The fact that a−1 > 0 can be proved by contradiction: assume that a−1 < 0 or a−1 = 0, then we
use (O4) and multiply the inequality a−1 < 0 by a > 0 two times, getting 1 < 0 and a < 0, a
contradiction with a > 0. Ruling out a−1 = 0 is similar.

• a < 0: In this case −a > 0 (follows by adding (−a) two times to each side of the inequality and
using (A3), (A4)). Since (−a)b = −ab > 0 (this follows from (−a)b+ab = [(−a)+a]b = 0b = 0
and the axioms), we have similarly as in the previous case that (−a)−1 > 0 and thus

(−a)b > 0
(−a) [(−a)b] > (−a)−1 0
−1
by axiom (O4)
[ −1
]
(−a) b > 0 by axiom (M2)
b > 0 by axioms (M4) and (M3),

In this case a and b have different signs, too.

• a = 0: This case cannot happen because then it would hold that ab = 0b = 0, which is in
contradiction with our assumption ab < 0.

In both possible cases a > 0 and a < 0, the assumption ab < 0 implied that a and b have different signs,
so our proof is complete.

Method 2: by contradiction

We assume that ”a and b have different signs” does not hold and derive a contradiction. It means
we assume that a and b have both the same sign or that one of them is zero. In each case we derive
contradiction:
• a = 0: Then ab = 0b = 0, which is in contradiction with ab < 0 (according to axiom (O1)).

• b = 0: This is analogous to the previous case.

• a > 0 and b > 0: Then by axiom (O3), we can multiply b to both sides of a > 0, obtaining
ab > 0, which is contradiction with ab < 0.

• a < 0 and b < 0: Then −a > 0 and −b > 0, so it is sufficient to show that (−a)(−b) = ab, since
then it follows as in the previous case that (−a)(−b) = ab > 0, a contradiction. Actually, we
have to show only a · (−b) = −(ab) because applying this two times yields (−a)(−b) = ab. We
write a(−b) = a(−b) + ab + [−(ab)] = a[(−b) + b] + [−(ab)] = a0 + [−(ab)] = −(ab), where
we have used several axioms (try to list all of them).

April 11, 2017 2 Karel Švadlenka

 Problem 1.3 Find sup E, inf E for the following sets and decide whether these sets have a maximum
and minimum. (1) { n+1n
: n ∈ N} (2) {p ∈ Q : |p| ≤ π} .

(1) If we write the specific form of the elements of this set, we see that

E = { 12 , 32 , 34 , . . . }.

It is clear that the numbers are increasing. We can show this by proving the inequality

(n + 1) n
> ,
(n + 1) + 1 n+1

which says that a member in the sequence is always greater than its predecessor. (The proof of the
inequality is easy - just multiply both sides by (n + 1)(n + 2) and cancel identical terms on both
sides.)
Hence the smallest element of the set is the first one ( 12 ) and this element is both the infimum and
the minimum of E since it belongs to E.
What about the supremum and maximum? One can expect that the supremum will be 1 because
n
limn→∞ n+1 = 1. To show this precisely, we use the following equivalent definition of the
supremum:
”M is the supremum of E if and only if the following two conditions hold:
• e ≤ M for all e ∈ E.
• For any ε > 0 there exists f ∈ E such that f > M − ε.”

For M = 1, the first condition is clear and the second one follows from the archimedean property.
Indeed, by the equivalent statement of the archimedean property, for any ε > 0 there is a natural
1
number n such that n+1 < ε. Subtracting 1 from each side of this inequality and multiplying the
resulting inequality by −1, we get 1 − n+11
> 1 − ε, which is exactly what we wanted to prove.
Or, we can just say that we set f to be n+1 , where n is some natural number greater than 1−ε
n
ε –
check that then f > 1 − ε.
Since sup E = 1 and 1 is not a member of the set E, the maximum of E does not exist.

(2) Let us think about the supremum and maximum first. The number π is obviously an upper bound
of E. We will show that it is the supremum of E. For that we will again use the following
equivalent definition of the supremum:
”s is the supremum of E if and only if the following two conditions hold:
• e ≤ s for all e ∈ E.
• For any ε > 0 there exists f ∈ E such that f > s − ε.”

The first condition is satisfied and the second one means that we have to find a rational number in
the interval (π − ε, π). But this follows from the density of rational numbers in R. Therefore, π
is the supremum.
Since the supremum π is not a rational number and thus does not belong to the set E, the maximum
for E does not exist in this case.
For infimum and minimum, the proof is analogous leading to the conclusion that −π is the infi-
mum of E and there is no minimum. Here we use the equivalent definition of infimum:
”m is the infimum of E if and only if the following two conditions hold:
• e ≥ m for all e ∈ E.

April 11, 2017 3 Karel Švadlenka

 • For any ε > 0 there exists f ∈ E such that f < m + ε.”

Problem 1.4 Using the completeness axiom, show that every nonempty set E of real numbers that is
bounded below has a greatest lower bound.

The idea of the proof is to reduce the situation to the case of least upper bound for which we have
the completeness axiom (this problem is actually just the completeness axiom for the ”other end” of the
set E). This is done by considering a set F which is the set E but with ”opposite sign”. If we have this
idea then the rest are just technical details which are however important and follow below.

Since E is bounded below, it has a lower bound. Let us take one such lower bound m, so that e ≥ m
for all e ∈ E. Since we are required to use the completeness axiom, we have to create a situation, where
we have a set that is bounded from above. One of the possibilities is to use the following trick: define
the set F by
F = {−x : x ∈ E}.
Now, F is bounded above by −m because for any f ∈ F there is an element e of E so that f = −e and
then
f = −e ≤ −m,
by the fact that m is a lower bound of E.

Next, we use the completeness axiom to deduce that the set F has the supremum (least upper bound)
M . We would like to show that −M is the infimum (greatest lower bound) of E. First, −M is definitely
a lower bound of E because for any e ∈ E there is an f ∈ F so that −f = e and then

e = −f ≥ −M,

since M is the supremum of F (and therefore f must be ≤ M ).

It remains to show that −M is the greatest lower bound. We can do it by contradiction: if it is not the
greatest lower bound it means that there is a lower bound of E that is still greater than −M , we denote
it by −N (i.e., −M < −N ). Then, similarly as above, we can show that N is an upper bound for F
which is smaller than M = sup F , a contradiction!

Problem 1.5 Show by induction that the number 5n − 4n − 1 is divisible by 16 for all natural numbers
n.

The proof by induction proceeds in two steps:

1. Show that the statement holds for n = 1: In this case 51 − 4 · 1 − 1 = 0, which is divisible by 16,
so the statement is true.

2. Show that if the statement holds for n then it holds also for n + 1: Assume that 5n − 4n − 1 is
divisible by 16 for some fixed n and show that 5n+1 − 4(n + 1) − 1 is also divisible by 16. Since
we want to use the fact that 5n − 4n − 1 is divisible by 16, we artificially create this expression as

April 11, 2017 4 Karel Švadlenka

 follows:

5n+1 − 4(n + 1) − 1 = 5n+1 − 4n − 5

= 5 · 5n − 5 · 4n − 5 + 5 · 4n − 4n
= 5 (5n − 4n − 1) + 20n − 4n
= 5(5n − 4n − 1) + 16n.

We see that 5n+1 − 4(n + 1) − 1 is a sum of two terms both of which are divisible by 16 (the first
one is divisible by 16 due to the induction assumption) and therefore itself is divisible by 16.

√
Problem 1.6 Show that the numbers of the form m 2/10n for m ∈ Z and n ∈ N are dense in R.

According to the definition, we √interval (a, b),√we are able

√ have to show that for any given nonempty
to find a number of the form m 2/10n in the interval. Denoting x = a/ 2 and y = b/ 2, it is the
same as finding a number of the form m/10n in the interval (x, y). By the archimedean theorem, we
can find a natural number k such that
1
< y − x.
k
Moreover, we can find a natural number n so that 10n > k (for example, n = k will always work).
1
It means that the number x + 10n is between x and y, and, multiplying this relation by 10n , we find
that
10n x < 10n x + 1 < 10n y.
Since the numbers 10n x and 10n x + 1 are 1 apart, there is an integer m ∈ Z such that

10n x < m ≤ 10n x + 1 < 10n y.

Dividing this
√ inequality by 10n we find
√ that
n
√ the number m/10 is in the interval (x, y). Hence, the
n
number m 2/10 is in the interval ( 2x, 2y) = (a, b).
√
For any interval (a, b) we have found a number of the form m 2/10n in this interval, which means
that numbers of this form are dense in R.

April 11, 2017 5 Karel Švadlenka

 Calculus A: Problems with Solutions (Lecture 2)

Problem 2.1 Show that |x| − |y| ≤ |x − y| holds for any real numbers x, y and find a condition for the
equality to hold.

By the definition of the absolute value, a ≤ |a| and −b ≤ |b| hold for any a, b ∈ R. We consider four
cases:
(1) x ≥ 0 and y ≥ 0: Then |x| − |y| = x − y ≤ |x − y| (we put a = x − y above). Equality holds if
and only if x − y ≥ 0, i.e., x ≥ y.
(2) x ≥ 0 and y < 0: Then |x| − |y| = x + y < x − y ≤ |x − y| (we put a = x − y above). Equality
never holds in this case.
(3) x < 0 and y ≥ 0: Then |x| − |y| = −x − y ≤ −x + y ≤ |x − y| (we put b = x − y above).
Equality holds if and only if y = 0.
(4) x < 0 and y < 0: Then |x| − |y| = −x + y ≤ |x − y| (we put b = x − y above). Equality holds
if and only if −x + y ≥ 0, i.e., x ≤ y.

We have proved that the inequality |x| − |y| ≤ |x − y| holds for all x, y ∈ R. The equality holds if and
only if x and y satisfy one of the following
• x, y ≥ 0 and x ≥ y
• x, y < 0 and x ≤ y
• y=0

Report Problem 2.2 Assuming the triangle inequality |a + b| ≤ |a| + |b|, show that another form of this
inequality ||x| − |y|| ≤ |x − y| holds.

There are many ways how to prove this inequality. One of the ways is to consider all the possible
combinations of signs for x and y, such as x ≥ 0 & y ≥ 0, etc., and in each case check that the
inequality holds. This method is simple but a little tedious, so we show another approach.

Using the triangle inequality |a + b| ≤ |a| + |b| with a = x − y and b = y, we get

|x| = |(x − y) + y| ≤ |x − y| + |y| ⇒ |x| − |y| ≤ |x − y|.
On the other hand, the triangle inequality |a + b| ≤ |a| + |b| with a = y − x and b = x implies
|y| = |(y − x) + x| ≤ |y − x| + |x| = |x − y| + |x| ⇒ |y| − |x| ≤ |x − y|.
(We have used the fact that |x − y| = |y − x| which follows from |a| = | − a| for any a.)

Since ||x| − |y|| is either equal to |x| − |y| or |y| − |x|, we see from the above two inequalities that
in either case ||x| − |y|| ≤ |x − y| holds.
Problem 2.3 Prove that if sn → ∞ then (sn )2 → ∞ also.

We want to show that for any M there is a number N ∈ N so that

(sn )2 ≥ M for all n ≥ N.

From the fact that sn → ∞ we conclude that there is N1 so that sn ≥ 1 for n ≥ N1 (we have chosen
M = 1 in the definition of divergence to infinity).

Moreover, we can find N2 so that

sn ≥ M for all n ≥ N2 .

Now set N = max{N1 , N2 }. Then (since sn ≥ 1 for n ≥ N and so (sn )2 ≥ sn for n ≥ N ),

(sn )2 ≥ sn ≥ M for all n ≥ N,

and the proof is finished.

Problem 2.4 Suppose that {sn } is a sequence of positive numbers converging to a positive limit. Show
that there is a positive number c so that sn > c for all n.

The main idea of the proof is that if a sequence converges to some positive number L then if we are
far enough in the sequence it must be close to L in the sense that the members are larger than L/2. Since
L/2 is a positive number, we are done with the tail of the sequence. The remaining members of the
sequence are only finite in number, so they can be dealt with just by taking their minimum. The formal
details follow.

Let us denote the limit of {sn } by L. We know that L > 0, so we can choose ε = L/2 in the
definition of the limit of {sn } and find a number N so that

|sn − L| < L
2 for all n ≥ N.

Then for n ≥ N we have

sn = L + sn − L ≥ L − |sn − L| > L − L
2 = L
2.

Hence sn > L2 for n ≥ N and the remaining members of the sequence {s1 , . . . , sN −1 } are all greater
than the half of their minimum (which is a positive number), so the required c can be defined by

c = min{ 21 s1 , 12 s2 , . . . , 12 sN −1 , L2 }.

This c is positive and satisfies sn > c for all n.

April 19, 2017 2 Karel Švadlenka

 √ √
Problem 2.5 Consider the sequence defined recursively by x1 = 2, xn = 2 + xn−1 .
Show that xn < 2 for all n ∈ N, and that xn < xn+1 for all n ∈ N.

Both statements can be shown by induction.

To prove xn < 2,
√
• first check that it holds for n = 1, i.e., that x1 < 2. But this is obvious since x1 = 2.
• Next show that if xn < 2, then the statement holds for n + 1, that is xn+1 < 2. It is enough to
note that √ √
xn+1 = 2 + xn < 2 + 2 = 2,
and we are done.

To prove xn < xn+1 ,

√ √ √
• first check that it holds for n = 1, i.e., that x1 < x2 . This is easy, since x1 = 2 < 2+ 2 =
x2 .
• Next show that if xn < xn+1 , then the statement holds also for n + 1, that is xn+1 < xn+2 .
Starting from the induction assumption, we deduce
xn < xn+1
xn + 2 < xn+1 + 2
√ √
xn + 2 < xn+1 + 2
xn+1 < xn+2 ,
and the proof by induction is complete.

√ xn < xn+1 can also be proved directly using the fact that xn < 2. To see this, notice
The inequality
that xn+1 = 2 + xn , so that
√ 2 + xn − x2n (2 − xn )(1 + xn )
xn+1 − xn = 2 + xn − xn = √ = √ .
2 + xn + xn 2 + xn + xn
Since xn < 2 and obviously xn > 0, we see that the last expression is always positive, which means
that xn+1 > xn .

Problem 2.6 Decide whether the following statements are true and give a reason for your answer.
(1) If {sn } and {tn } are both divergent then so is {sn tn }.
(2) If {sn } and {tn } are both convergent then so is {sn tn }.

(1) This statement is false. To prove it, it is sufficient to give an counterexample. Set for example
sn = (−1)n and tn = sn . Then sn tn = 1 for all n so this sequence is convergent but both {sn }
and {tn } are divergent.
(2) This statement is true. To prove it, we have to give a proof according to the definition. See the
proof of Theorem 2.16 in the textbook. (Or, since we have already proved Theorem 2.16 in the
lecture, it is enough to say that the statement follows from Theorem 2.16.)

April 19, 2017 3 Karel Švadlenka

Problem 2.7 Suppose that {sn } and {tn } are sequences of positive numbers and that
sn
lim =α∈R and sn → ∞.
n→∞ tn

What can you conclude for the sequence {tn }?

Since sn → ∞ and after dividing by tn the sequence converges to a finite number, we may expect
that tn → ∞, too. Let us prove it by contradiction.

We assume that tn does not diverge to ∞, which means that we can find a number M1 so that

for any N there is N1 ≥ N so that tN1 < M1 .

(Check that this is really the negation of the definition of ”tn → ∞”.)

Furthermore, by the assumptions of the problem, we know that there are numbers N2 , N3 so that
sn
<α+1 for all n ≥ N2 , (1)
tn
sn > M1 (α + 1) for all n ≥ N3 .
(In the first statement we have chosen ε = 1 in the definition and in the second one, M = M1 (α + 1).)

Let us set N = max{N2 , N3 }. Then both the above statements are true for n ≥ N , while from the
assumption that tn does not diverge to ∞ we find a number N1 ≥ N so that tN1 < M1 , which implies

sN1 M1 (α + 1)
> = α + 1.
tN1 M1

But this is a contradiction with the convergence of {sn /tn } (inequality (1))!

April 19, 2017 4 Karel Švadlenka

 Calculus A: Problems with Solutions (Lecture 3)

Problem 3.1 Consider the sequence s1 = 1, sn = s22 . We argue that if sn → L then L = L22 and so
√ √ n−1
L = 3 2. Our conclusion is that limn→∞ sn = 3 2. Do you have any criticism of this argument?

Writing several initial terms of the sequence:

1 1
1, 2, , 8, , 2048, . . . ,
2 32
we see that this sequence does not converge, so the conclusion above must be wrong.

The reason is that the relation L = L22 holds only under the condition that the sequence is convergent.
This condition is not fulfilled in this case and hence it does not make sense to say that the limit satisfies
some relation.

Problem 3.2 Establish which of the following statements are true for an arbitrary sequence {sn }.
(1) If all monotone subsequences of a sequence {sn } are convergent, then {sn } is bounded.
(2) If all monotone subsequences of a sequence {sn } are convergent, then {sn } is convergent.
(3) If all convergent subsequences of a sequence {sn } converge to 0, then {sn } converges to 0.
(4) If all convergent subsequences of a sequence {sn } converge to 0 and {sn } is bounded, then {sn }
converges to 0.

(1) This is true, which can be proved by contradiction.

Assume that {sn } is not bounded above (the case for {sn } unbouded below is similar). Then we
can find N1 such that sN1 ≥ 1.
Next, we can find N2 > N1 such that sN2 ≥ 2 (if there is not such N2 it would mean that all
terms sn are less than max{s1 , . . . , sN1 , 2}, which is in contradiction with {sn } being unbounded
above).
Similarly, we can find N3 > N2 such that sN3 ≥ 3, and so on. This gives a subsequence {sNk }
of {sn } whose Nk -th term is at least k. This means that this subsequence diverges to ∞, which is
a contradiction.
(2) This is false as the sequence 1, −1, 1, −1, 1, −1, . . . shows.
(3) This is also false as the sequence 0, 1, 0, 2, 0, 3, 0, 4, 0, 5, 0, 6, . . . shows.
(4) This is true (notice that the counterexample in (3) is excluded by the new assumption that {sn }
is bounded). We can prove it again by contradiction. Assume that {sn } does not converge to 0,
which means that there exists ε > 0 so that

for all N there is n > N such that |sn | ≥ ε.

 We construct a subsequence that does not converge to 0.
To do so first find some N1 ≥ 1 such that |sN1 | ≥ ε.
Next find N2 > N1 such that |sN2 | ≥ ε. Such N2 exists by the above assumption that {sn }
does not converge to 0. In this way we construct a subsequence {sNk } such that all its members
are greater than or equal to ε in absolute value. This subsequence is bounded because {sn } itself
is bounded, so it must contain a monotonic convergent subsequence by the Bolzano-Weierstrass
theorem. However, this monotonic subsequence cannot converge to 0 because all its terms are out
of the interval (−ε, ε). A contradiction!

Problem 3.3 Show that every bounded monotonic sequence is Cauchy.

We know that a sequence {sn } is bounded:

there is M such that |sn | ≤ M for all n,

and monotonic:
s1 ≤ s2 ≤ · · · ≤ sn ≤ sn+1 ≤ · · ·
(we consider the case of nondecreasing sequence; the case of increasing sequences is a subset of this
case, and for nonincreasing and decreasing sequences the proof is analogous),

and we want to show that {sn } is a Cauchy sequence:

for any ε > 0 there is N such that |sn − sm | < ε for all n, m ≥ N.

Since the sequence is nondecreasing, for n > m we can write this as sn − sm < ε or sn < sm + ε.

We will try to use the proof by contradiction. We assume that the sequence is not Cauchy, which
means that we can find ε > 0 such that

for any N there are m, n ≥ N with n > m such that sn − sm > ε.

The idea is that if we add this ε to s1 sufficiently many times, we will surpass the upper bound M of the
sequence.

Precisely, set K to be the natural number which is closest to but greater than M −s
ε . We find m1 , n1
1

so that sn1 − s1 ≥ sn1 − sm1 > ε, so that sn1 > s1 + ε. Next, we find m2 , n2 > n1 so that
sn2 − sn1 ≥ sn2 − sm2 > ε, so that sn2 > sn1 + ε > s1 + 2ε. We proceed in this way K times to obtain
snK satisfying snK > s1 + Kε. But from the definition of K, we get

M − s1
snK > s1 + Kε > s1 + ε = M.
ε
We got a contradicition with the boundedness of {sn } and the proof is complete.

April 26, 2017 2 Karel Švadlenka

 Calculus A: Problems with Solutions (Lecture 4)

Problem 4.1 Determine the set of interior points, accumulation points, isolated points, and boundary
points for the set
E = (0, 1) ∪ (1, 2) ∪ (2, 3) ∪ · · · ∪ (n, n + 1) ∪ · · · .

(1) The set of interior points are all positive numbers except natural numbers. Natural numbers do
not belong to E, so they cannot be interior points. All other positive real numbers belong to some
of the intervals (n, n + 1) and since every point of an open interval is its interior point, they also
have to be an interior point of E.

(2) The set of accumulation points is {x ∈ R : x ≥ 0} (all nonnegative numbers). Any point of an
open interval is its accumulation point, so we have to check only the natural numbers and zero.
For any n ∈ N ∪ {0} and c > 0 the interval (n − c, n + c) contains infinitely many points of E
(actually, except for n = 0, all points of this interval except n belong to E), so natural numbers
are also accumulation points.

(3) There are no isolated points. Any point in an open interval is not isolated, so only the points
corresponding to natural numbers could be isolated. But as we said in (2), for any n ∈ N and
c > 0 the interval (n − c, n + c) contains infinitely many points of E, so natural numbers are not
isolated points of E.

(4) The set of boundary points is the set of natural numbers plus 0. Indeed, as said above, for any
n ∈ N and c > 0 the interval (n − c, n + c) contains infinitely many points of E, but at the same
time the point n does not belong to the interval. So each such interval contains both points from
E and points which are not in E. Moreover, any point from an open interval is not its boundary
point.

Problem 4.2 Show that every interior point of a set must also be an accumulation point of that set, but
not conversely.

By definition, if x is an interior point of E, then there exists c > 0 so that the interval (x − c, x + c) is
contained in E. Take any d > 0 and set m = min{c, d} > 0. Then the interval (x − d, x + d) contains
the interval (x − m, x + m) and this interval (x − m, x + m) belongs to E and contains infinitely many
points. This shows that (x − d, x + d) contains infinitely many points of E for any d > 0 and thus, by
definition, x is an accumulation point.

The converse ”Any accumulation point is an interior point” is not true since, for example, a is an
accumulation point of the open interval E = (a, b) but it does not even belong to this interval, so it
cannot be its interior point.

Problem 4.3 Show that a set E is closed if and only if E = E.

 Let us denote by E ′ the set of accumulation points of E.
• Let the set E be closed. Then every accumulation point belongs to E, so that E ′ ⊂ E, which
implies E = E ∪ E ′ = E.

• Let E = E. This means that E ∪ E ′ = E so that E ′ ⊂ E must hold. This says that the set of
accumulation points E is a subset of E, that is, every accumulation point of E belongs to E. The
set E is closed.

Problem 4.4 Show that the closure operation has the property E1 ∪ E2 = E1 ∪ E2 .

We will first show that (E1 ∪ E2 )′ = E1′ ∪ E2′ . The result then immediately follows since

E1 ∪ E2 = (E1 ∪ E2 ) ∪ (E1 ∪ E2 )′ = E1 ∪ E2 ∪ E1′ ∪ E2′ = (E1 ∪ E1′ ) ∪ (E2 ∪ E2′ ) = E1 ∪ E2 .

We show the above equality by showing the following two inclusions:

• (E1 ∪ E2 )′ ⊂ E1′ ∪ E2′
Let x be any point in (E1 ∪E2 )′ . Then for any c > 0 there are infinitely many points from E1 ∪E2
in the interval (x − c, x + c). There are three possibilities:
– For any c > 0 there are infinitely many points from E1 in the interval (x − c, x + c). Then
x ∈ E1′ .
– For any c > 0 there are infinitely many points from E2 in the interval (x − c, x + c). Then
x ∈ E2′ .
– For any c > 0 there are infinitely many points from both E1 and E2 in the interval (x −
c, x + c). Then x ∈ E1′ ∩ E2′ .

It is shown by contradiction that no other case is possible. Indeed, the remaining case is that there
is a number c > 0 so that the interval (x − c, x + c) contains only finitely many points of E1 and
only finitely many points of E2 . However, this is a contradiction with the fact that (x − c, x + c)
contains infinitely many points of E1 ∪ E2 .
In either case, we have shown that x ∈ E1′ ∪ E2′ , which implies the desired statement.

• (E1 ∪ E2 )′ ⊃ E1′ ∪ E2′

Let x be any point in E1′ ∪ E2′ . Let c > 0. Then the interval (x − c, x + c) contains infinitely
many points from E1 (if x ∈ E1′ ) or infinitely many points from E2 (if x ∈ E2′ ). In either case
this interval contains infinitely many points from E1 ∪ E2 , which shows that x ∈ (E1 ∪ E2 )′ and
the second inclusion is proved.

Problem 4.5 Show that the intersection of an arbitrary collection of closed sets is closed.

Let us denote the collection of sets by {Eα }α∈A , where A is some index set, and their intersection
by E.

Let x be an accumulation point of the intersection E. To show that E is closed, it is sufficient to

show that x belongs to the intersection E. Since x is an accumulation point of E, for any c > 0 there
are infinitely many points in the set
(x − c, x + c) ∩ E.

May 10, 2017 2 Karel Švadlenka

These points have to be also members of each set Eα since they belong to the intersection of all these
sets. Hence the intersection (x − c, x + c) ∩ Eα contains infinitely many points for every α ∈ A and
c > 0, which means that x is an accumulation point of every set Eα , α ∈ A. But since these sets Eα are
closed, x must belong to each of these sets, and consequently to the intersection E of all these sets. The
proof is complete.

Problem 4.6 Show directly that the interval [0, ∞) does not have the Bolzano-Weierstrass property.

The Bolzano-Weierstrass property for a set E is: ”Every sequence of points chosen from the set has a
subsequence that converges to a point that belongs to E.” We take the sequence 1, 2, 3, 4, . . . of natural
numbers. This sequence is contained in [0, ∞). However, every subsequence of this sequence diverges
to infinity, hence this set does not satisfy the Bolzano-Weierstrass propety.

Problem 4.7 Show directly that the union of two sets with the Bolzano-Weierstrass property must have
the Bolzano-Weierstrass property.

Let the sets A and B have the Bolzano-Weierstrass property. We want to show that A ∪ B also has
the same property. To this end, choose any sequence from A ∪ B. The task is to find a subsequence
which converges to a point in A ∪ B.

By contradiction we can easily show that the sequence has to contain infinitely many points from at
least one of the sets A or B (if not, it contains only finitely many points from A and finitely many points
from B, so it must be finite and thus it is not a sequence). Without loss of generality, assume that the
sequence contains infinitely many points of A. Since A has the Bolzano-Weierstrass property, we can
find a subsequence converging to a point x that belongs to A. But then the subsequence is contained in
A ∪ B and the limit point x also belongs to A ∪ B. Hence we have found a subsequence converging to
a point in A ∪ B.

May 10, 2017 3 Karel Švadlenka

 Calculus A: Problems with Solutions (Lecture 5)

Problem 5.1 Using both the ε − δ definition and the sequential definition of the limit, give two different
proofs of the following statement:

If lim f (x) = L then lim |f (x)| = |L|.

x→x0 x→x0

1. ε − δ definition:
We know that limx→x0 f (x) = L by the ε − δ version of the definition of limit means that

for every ε > 0 there is δ > 0 so that |f (x) − L| < ε whenever 0 < |x − x0 | < δ.

Then, by triangle inequality we deduce

|f (x)| − |L| ≤ |f (x) − L| < ε whenever 0 < |x − x0 | < δ,

which shows that limx→x0 |f (x)| = |L|.

2. sequential definition:
We know that limx→x0 f (x) = L by the sequential version of the definition of limit means that
for every sequence {en } with en ̸= x0 , and en → x0 as n → ∞,

lim f (en ) = L.
n→∞

Therefore, we have (see Theorem 2.22)

lim |f (en ) − L| = | lim (f (en ) − L)| = 0.

n→∞ n→∞

Then, using the following triangle inequality:

−|f (en ) − L| ≤ |f (en )| − |L| ≤ |f (en ) − L|,

we get, by the Squeeze Theorem (see Theorem 2.20), that

lim (|f (en )| − |L|) = lim |f (en ) − L| = 0,

n→∞ n→∞

which is what we wanted to show.

Problem 5.2 Formulate a definition for the statements

lim f (x) = L, and lim f (x) = ∞,

x→∞ x→∞

and find the limits

lim xp
x→∞
for various real numbers p.

First, we give the definitions:

 (1) We write
lim f (x) = L
x→∞

if for every ε > 0 there is a K > 0 so that

|f (x) − L| < ε whenever x > K.

(2) We write
lim f (x) = ∞
x→∞

if for every M > 0 there is a K > 0 so that

f (x) > M whenever x > K.

Next, we calculate lim xp for various real numbers p. We consider three cases:
x→∞

(1) p < 0: We guess that the limit is 0. To show it, for any given ε > 0, we choose K = ε1/p > 0.
Then, since p < 0, we get

|xp − 0| = |xp | = xp < K p = ε whenever x > K > 0.

(Note that we can assume that x is positive since we are interested only in large values of x.)
We conclude
lim xp = 0 for p < 0.
x→∞

(2) p = 0: xp = x0 = 1 for every x > 0, so we can immediately write

lim x0 = 1.
x→∞

(3) p > 0: In this case, the limit will be +∞. Indeed, for every M > 0, we choose K = M 1/p > 0.
Then, since p > 0, we get

xp > K p = M whenever x > K.

Therefore, we conclude
lim xp = ∞ for p > 0.
x→∞

Problem 5.3 Prove the following theorem using an ε − δ proof and also using the sequential definition
of limit.

Theorem 5.15 Suppose that the limit limx→x0 f (x) exists and that C is a real number. Then
( )
lim Cf (x) = C lim f (x) .
x→x0 x→x0

May 17, 2017 2 Karel Švadlenka

 1. ε − δ definition:
Let L = limx→x0 f (x). We need to prove that for every ε > 0 there is δ > 0, such that

|Cf (x) − CL| < ε whenever 0 < |x − x0 | < δ.

Hence take an arbitrary ε > 0. If C = 0, there is nothing to prove, so we may assume C ̸= 0. By

the properties of absolute value

|Cf (x) − CL| = |C||f (x) − L|,

so if we choose δ > 0 so that |f (x) − L| < ε/|C| whenever 0 < |x − x0 | < δ (which we always
can because limx→x0 f (x) = L), we have

|Cf (x) − CL| = |C||f (x) − L| < |C|(ε/|C|) = ε.

This is precisely the statement that limx→x0 Cf (x) = CL.

2. Sequential definition:
We need to prove that for every sequence {en } with en ̸= x0 , and en → x0 as n → ∞,

lim Cf (en ) = CL.

n→∞

Here, by the sequential definition of limit, limx→x0 f (x) = L means that

lim f (en ) = L for any sequence {en } as above.

n→∞

Therefore, using Theorem 2.14, we deduce that

lim Cf (en ) = C( lim f (en )) = CL,

n→∞ n→∞

which is the relation that we wanted to prove.

May 17, 2017 3 Karel Švadlenka

 Calculus A: Problems with Solutions (Lecture 6)

Problem 6.1 Show that the following definitions of continuity for a function f defined in a neighbor-
hood of x0 are equivalent.
(1) Function f is continuous at x0 provided lim f (x) = f (x0 ).
x→x0

(2) Function f is continuous at x0 if for each ε > 0 there exists δ such that |f (x) − f (x0 )| < ε
whenever |x − x0 | < δ.

(3) Function f is continuous at x0 if lim f (en ) = f (x0 ) for every sequence {en } → x0 .
n→∞

The equivalence will be shown if we prove the following four implications:

1. (1) ⇒ (2)
By the ε − δ definition of limit, for each ε > 0 there exists δ so that

|f (x) − f (x0 )| < ε whenever |x − x0 | < δ and x ̸= x0 .

Hence we only need to check the case x = x0 , but this is easy because

|f (x) − f (x0 )| = |f (x0 ) − f (x0 )| = 0 < ε.

2. (2) ⇒ (1)
From (2) we immediately get that for each ε > 0 there exists δ so that

|f (x) − f (x0 )| < ε whenever 0 < |x − x0 | < δ.

This is the ε − δ definition of limx→x0 f (x) = f (x0 ).

3. (2) ⇒ (3)
The proof follows the arguments on p.270 in the textbook. By (2), for any ε > 0 there exists a
positive number δ such that

|f (x) − f (x0 )| < ε if |x − x0 | < δ.

Now, consider any sequence {en } → x0 . This means that there is a number N such that

|en − x0 | < δ for all n ≥ N,

where δ is the number obtained above. Putting these together we find that

|f (en ) − f (x0 )| < ε if n ≥ N,

which is the definition (3).

 4. not (2) ⇒ not (3)
Assume that (2) does not hold. We want to show that then (3) also does not hold. That is, we have
to find a sequence {en } converging to x0 such that f (en ) does not converge to f (x0 ).
If (2) is not true, then there must be some ε0 > 0 so that for any δ > 0 there will be points x in
the domain of f with

|x − x0 | < δ and |f (x) − f (x0 )| ≥ ε0 .

Applying this to δ = 1, 21 , 31 , . . . (i.e., δ = n1 ), we obtain a sequence of points en such that

1
|en − x0 | < and |f (en ) − f (x0 )| ≥ ε0 .
n
Then {en } is the desired sequence since en converges to x0 but f (en ) does not converge to f (x0 ).

Problem 6.2 Prove that the function f (x) = |x| is continuous at every point of R using the δ − ε
definition of continuity.

Let x0 be any point of R and ε any positive number. Since any such x0 is an interior point of R, our
task is to show that there is δ (it can depend on x0 ) such that

|x| − |x0 | < ε whenever |x − x0 | < δ.

Since by the triangle inequality

|x| − |x0 | ≤ |x − x0 |,
we see that it is enough to take δ = ε.

To summarize, given any x0 ∈ R and ε > 0, we set δ = ε. Then for any x satisfying |x − x0 | < δ, it
holds
|x| − |x0 | ≤ |x − x0 | < ε.
This by definition says that |x| is continuous at x0 and since x0 was arbitrary, |x| is continuous on R.

Problem 6.3 Suppose f is uniformly continuous on each of the compact sets X1 , X2 , . . . , Xn . Prove
that f is uniformly continuous on the set
∪
n
X= Xi .
i=1

Show that this need not be the case if the sets Xk are not closed and need not be the case if the sets Xk
are not bounded.

We know that ”f is uniformly continuous on Xi ” means that for any ε > 0 there exists δi > 0 such
that
|f (x) − f (y)| < ε whenever x, y ∈ Xi and |x − y| < δi .

To begin with, we consider n = 2, that is X = X1 ∪ X2 .

We first prove that f is continuous on X. For any x ∈ X, x is either in X1 or in X2 , so we can assume
without loss of generality that x ∈ X1 .

May 24, 2017 2 Karel Švadlenka

 Let d(x, X2 ) = inf{|x − y| : y ∈ X2 } be the ”distance” of x from the set X2 . We consider two
cases:

(1) If d(x, X2 ) > 0, we have

y ∈ X and |x − y| < d(x, X2 ) ⇒ y ∈ X1 .

Therefore, we get that

|f (x) − f (y)| < ε whenever y ∈ X and |x − y| < min{δ1 , d(x, X2 )}.

Thus f is continuous at x in this case.

(2) If d(x, X2 ) = 0, we have x ∈ X2 because X2 is compact set (provide the details!). Hence
x ∈ X1 ∩ X2 and we conclude that

|f (x) − f (y)| < ε whenever y ∈ X and |x − y| < min{δ1 , δ2 }.

Thus f is also continuous at x in this case.

Next we need to prove that f is uniformly continuous on X, but this follows from Theorem 5.48 since
f is continuous on X and X is compact.

We have proved that f is uniformly continuous

∪k on∪X if n = 2. For general n, we use induction. Sup-
pose that f is uniformly continuous on i=1 Xi . As ki=1 Xi is compact, if we use the above argument
(∪k ) ∪k+1
to i=1 Xi ∪ Xk+1 , then we find that f is uniformly continuous on i=1 Xi , which completes the
proof.

Finally, we give counterexamples.

(1) not closed case : Let {

0 for x < 1
f (x) =
1 for x ≥ 1
and X1 = (0, 1), X2 = [1, 2). Then we have
{
0 for x ∈ X1
f (x) =
1 for x ∈ X2 ,

so f is uniformly continuous on both X1 and X2 . However f is not continuous on X1 ∪ X2 =

(0, 2).

(2) not bounded case : Let f (x) = x2 and

X1 = {1, 2, 3, . . . , n, n + 1, . . . }
1 1 1 1
X2 = {2 + , 3 + , . . . , n + , n + 1 + , . . . }.
2 3 n n+1
Since all points in both of the above sets are isolated, f is uniformly continuous on both sets.
Indeed, we can see that by taking δ = 1/2. Then

x, y ∈ X1 and |x − y| < 1/2 ⇒ x = y,

and
x, y ∈ X2 and |x − y| < 1/2 ⇒ x = y,

May 24, 2017 3 Karel Švadlenka

 by definition of X1 , X2 . So for any ε > 0, we have

|f (x) − f (y)| = 0 < ε whenever x, y ∈ Xi and |x − y| < δ,

for i = 1, 2, which shows that f is uniformly continuous on both X1 and X2 .

Now, define xn , yn ∈ X1 ∪ X2 as xn = n and yn = n + 1/n. Then |yn − xn | = 1/n → 0 as
n → ∞, but
( 1 )2 1
|f (yn ) − f (xn )| = | n + − n2 | = 2 + 2 > 2.
n n
Hence, f is not uniformly continuous on X1 ∪ X2 .

Problem 6.4* Show the following theorem (Theorem 5.50 in the textbook) using a Bolzano-Weierstrass
argument.

Let f be continuous on [a, b]. Then f possesses both an absolute maximum and an absolute minimum
on [a, b].

Let M = sup{f (x) : a ≤ x ≤ b}. By Theorem 5.48, f is uniformly continuous on [a, b]. Thus,
by Theorem 5.49, M < ∞. We need to prove, using Bolzano-Weierstrass argument, that there exists
x0 ∈ [a, b] so that f (x0 ) = M .

For every number n ∈ N let Un = {x ∈ [a, b] : f (x) > M − 1/n}.

First, we prove that Un ’s are not empty for every n. If Un is empty for some n, then we have

f (x) ≤ M − 1/n for any x ∈ [a, b].

This contradicts the fact that M is the least upper bound of f (x) in [a, b], so Un cannot be empty.

Next, for each n, let xn be an element of Un . Then, the sequence {xn } is an infinite sequence, and by
the Bolzano-Weierstrass Theorem, this sequence has a convergent subsequence {xnk } such that {xnk }
converges to a point x0 ∈ [a, b]. Then, by the definition of Un , we have

f (xnk ) > M − 1/nk for every k ∈ N.

Because f is continuous on [a, b], we get

f (x0 ) = lim f (xnk ) ≥ lim (M − 1/nk ) = M.

k→∞ k→∞

Hence we conclude that f (x0 ) = M due to the definition of M .

A similar proof would show that f has an absolute minimum on [a, b].

Problem 6.5* Show that a nondecreasing function with the Darboux property (or Intermediate Value
property, see Definition 5.27 in the textbook) must be continuous.

We prove it by contradiction. Accordingly, suppose that there is a point x0 ∈ R where f fails to be

continuous. Then, by the nondecreasing property (see Textbook, p. 335), f has a jump at x0 and

lim f (x) < lim f (x).

x→x0− x→x0+

Let L = limx→x0− f (x). Then there is an increasing sequence {xn } such that

xn → x0 and f (xn ) → L.

May 24, 2017 4 Karel Švadlenka

 We now assume that L < f (x0 ). Let M ∈ R be such that L < M < f (x0 ).
Then for xn we have
f (xn ) ≤ L < M < f (x0 ),
so the Intermediate Value property of f on the interval [xn , x0 ] implies the existence of yn ∈ (xn , x0 )
such that f (yn ) = M .

Since xn → x0 , we have yn → x0 by the Squeeze Theorem. Thus, given n, there is m such that
yn < xm , but
f (yn ) = M > L ≥ f (xm ),
which contradicts the fact that f is nondecreasing.

A similar argument is valid if f (x0 ) < limx→x0+ f (x).

Problem 6.6* Let f be a continuous function on an open interval (a, b). Suppose that f has no local
maximum or local minimum at any point. Show that f must be monotonic.

We prove this by contradiction. To this end, we first consider the following claim.

Claim. If f is not monotonic on (a, b), there exist c, d ∈ (a, b) such that

a<c<d<b and f (c) = f (d).

Assuming that this claim is true, we can easily show that f has a local maximum or minimum at a
point in (c, d). Indeed, by Theorem 5.50, f possesses an absolute maximum and an absolute minimum
on [c, d]. If either of these points is in (c, d), then f attains a local minimum or maximum at this point.
Otherwise, if both points correspond to one of the endpoints c or d, then we have

f (x) = f (c) = f (d) whenever x ∈ [c, d].

Hence, each point of (c, d) is a point of local minimum and maximum. Since we have shown that f has
a local minimum or maximum in (a, b), this is a contradiction, and the proof is done.

Finally we prove the above claim.

Proof of Claim. Since f is not monotonic, there exist x, y, z ∈ (a, b) with x < y < z such that

either f (x) < f (y) and f (y) > f (z), or f (x) > f (y) and f (y) < f (z).

Suppose f (x) < f (y) and f (y) > f (z). We consider the following three cases:
(1) f (x) = f (z): In this case, if we define c = x, d = z, the claim is valid.

(2) f (x) < f (z): In this case, we have f (x) < f (z) < f (y).
As f is continuous on [x, y], the Intermediate Value Property (see Theorem 5.53) can be applied.
Then, there exists c ∈ (x, y) so that f (c) = f (z). Since c < z, if we define d = z, the claim is
valid.

(3) f (x) > f (z): In this case, we have f (z) < f (x) < f (y).
Again, as f is continuous on [y, z], the Intermediate Value Property can be applied. So there exists
d ∈ (y, z) so that f (d) = f (x). Since x < d, if we define c = x, the claim is valid.

May 24, 2017 5 Karel Švadlenka

A similar argument can be used if f (x) > f (y) and f (y) < f (z), so the claim is proved.

Problem 6.7* Suppose f is increasing on an interval I. Let x0 be an interior point of I. Prove that

lim f (x) ≤ f (x0 ) ≤ lim f (x).

x→x0 − x→x0 +

First, since the function is increasing (and therefore monotonic), by Theorem 5.60 we find that both
the limits
lim f (x) and lim f (x)
x→x0 − x→x0 +

exist (see the proof of the theorem in the textbook).

Since f is increasing on I, we have

f (x) < f (x0 ) for all x ∈ I such that x < x0 .

The point x0 is an accumulation point of the set {x ∈ I : x < x0 } because it is an interior point of the
interval I. The left-hand limit
lim f (x)
x→x0 −

takes into account only points x, which are ”to the left of x0 ” (that is, which satisfy x < x0 ), and thus
by the order properties of limit (Theorem 5.19) we have

lim f (x) ≤ f (x0 ).

x→x0 −

We can treat the right-hand limit in completely the same way. Since f is increasing on I, we have

f (x0 ) < f (x) for all x ∈ I such that x > x0 .

The right-hand limit

lim f (x)
x→x0 +

takes into account only points x, which are ”to the right of x0 ” (that is, which satisfy x > x0 ), and thus
by the order properties of limit we have

f (x0 ) ≤ lim f (x).

x→x0 +

May 24, 2017 6 Karel Švadlenka

 Calculus A: Problems with Solutions (Lecture 7)

Problem 7.1 Decide whether the following functions are uniformly continuous on the given interval.
(1) f1 (x) = 2x + 3 on I1 = R.

(2) f2 (x) = x4 on I2 = [−10, 6].

√
(3) f3 (x) = x on I3 = (0, 1).

(4) f4 (x) = x3 on I4 = (0, ∞).

(1) f1 (x) = 2x + 3 on I1 = R
The function is uniformly continuous on R, which follows by the definition. To confirm that, take
any ε and set δ = 2ε . Then for any pair of numbers x, y satisfying |x − y| < δ we obtain

|f1 (x) − f1 (y)| = |(2x + 3) − (2y + 3)| = 2|x − y| < 2δ = ε,

which is the definition of uniform continuity.

(2) f2 (x) = x4 on I2 = [−10, 6]

f2 is uniformly continuous on I2 , which immediately follows from Theorem 5.48 (this theorem
says that any continuous function on a closed and bounded interval is uniformly continuous on
the interval).
√
(3) f3 (x) = x on I3 = (0, 1)
We can show the uniform continuity of f3 on I3 in several ways. For example, since f3 is con-
tinuous on the closed and bounded interval [0, 1], it is uniformly continuous on [0, 1] by Theorem
5.48. Hence it is also uniformly continuous on any subset of [0, 1], in particular, on the interval
(0, 1).
Another proof is the direct one. Take any ε > 0 and any two numbers x, y in the interval (0, 1).
Then we need to find δ > 0 such that we can estimate
√ √
|f3 (x) − f3 (y)| = | x − y|
ε2
by ε when |x − y| < δ. We show that setting δ = 4 will work.
ε2 2
• If x < 16 , then taking δ ≤ ε4 we have
√ √ √ √ √ √
| x − y| ≤ x + y = x + (y − x) + x
√ √ √ √ √
≤ x + |y − x| + x < 2 x + δ
ε ε
≤ 2 + =ε
4 2
ε2
• If x ≥ 16 then we can estimate
√ √ |x − y| |x − y| 4 4δ
| x − y| = √ √ ≤ √ ≤ |x − y| < ,
x+ y x ε ε
 ε2 √ √
so if we take δ ≤ 4 then | x − y| < ε.

(4) f4 (x) = x3 on I4 = (0, ∞)

This function is not uniformly continuous on I4 . We prove it by showing that the difference
|f4 (x) − f4 (y)| = |x3 − y 3 | can be made arbitrarily large however small the difference |x − y|
may be. ”Can be made arbitrarily large” precisely means that for any given M (assumed large)
and δ (assumed small) we can find x, y ∈ (0, ∞) such that |x − y| < δ but |x3 − y 3 | > M .
As
|x3 − y 3 | = |x − y|(x2 + xy + y 2 ) ∀x, y ∈ (0, ∞),
√
it is sufficient to take x = 2M δ and y = x + 2 . Then of course |x − y| = 2 < δ and
δ δ

δ δ δ 2M
|x3 − y 3 | = |x − y|(x2 + xy + y 2 ) = (x2 + xy + y 2 ) > x2 = = M.
2 2 2 δ

There are no other solved problems this time. See solved problems number 6 for examples on continuous
functions and number 8 for examples on derivatives.

May 25, 2016 2 Karel Švadlenka

 Calculus A: Problems with Solutions (Lecture 8)

Problem 8.1 Let {

x2 , if x ≥ 0
f (x) =
ax, if x < 0.
(1) For which values of a is f differentiable at x = 0?

(2) For which values of a is f continuous at x = 0?

(3) When f is differentiable at x = 0, does f ′′ (0) exist?

4
−0.7*x
x2
3.5

2.5

1.5

0.5

0
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2

(1) For which values of a is f differentiable at x = 0?

If f is differentiable at x = 0 then the left-hand derivative and the right-hand derivative at x = 0
have to exist and be equal. In this case it is easy to compute
f (x) − f (0) ax − 0
f−′ (0) = lim = lim =a
x→0− x−0 x→0− x
f (x) − f (0) x2 − 0
f+′ (0) = lim = lim = 0.
x→0+ x−0 x→0+ x
Since these two derivatives have to be equal, we get the condition a = 0.

(2) For which values of a is f continuous at x = 0?

The function is continuous at x = 0 if both the left-hand limit and the right-hand limit at x = 0
exist and are equal. Here we calculate

lim f (x) = lim ax = 0

x→0− x→0−
lim f (x) = lim x2 = 0,
x→0+ x→0+

so the function is continuous at x = 0 for any value of a ∈ R.

 (3) When f is differentiable at x = 0, does f ′′ (0) exist?
When f is differentiable at x = 0 (that is, form (1), when a = 0), then we have the following
function: { 2
x , if x ≥ 0
f (x) =
0, if x < 0.
This function is differentiable on R and the derivative is
{
′ 2x, if x ≥ 0
f (x) =
0, if x < 0

(we have computed the derivative at x = 0 in (1)).

Our task is to decide whether this function f ′ (x) is differentiable at x = 0. The function f ′ (x) is
continuous at x = 0, so we cannot say anything about differentiability using continuity. (If f ′ (x)
were discontinuous at x = 0, we could immediately say that it is not differentiable but this is not
the case here.) We have to check differentiability again by comparing the left-hand and right-hand
derivatives of f ′ (x) at x = 0:

f ′ (x) − f ′ (0) 0−0

f−′′ (0) = lim = lim =0
x→0− x−0 x→0− x
f ′ (x) − f ′ (0) 2x − 0
f+′′ (0) = lim = lim = 2.
x→0+ x−0 x→0+ x
The one-sided derivatives are different, so f ′′ (0) does not exist.

Problem 8.2 Suppose that a function has an infinite derivative at a point. What, if anything, can you
conclude about the continuity of that function at that point?

We show by giving two examples, that the function can be both continuous and discontinuous.
• Take { √
x
√ if x ≥ 0
f (x) = .
− −x if x < 0
This function is obviously continuous at x = 0 since

lim f (x) = 0 = f (0).

x→0

However, the derivative at x = 0 is infinite, as we can see from

f (x) − f (0) 1
lim = lim √ = ∞.
x→0 x−0 x→0 |x|

1.5

0.5

−0.5

−1

−1.5
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2

June 7, 2017 2 Karel Švadlenka

 • On the other hand, define {
1/x if x ̸= 0
f (x) = .
0 if x = 0
This function is not continuous at x = 0 because the limits limx→0+ f (x) and limx→0− f (x) are
not equal to the value f (0). At the same time, the derivative is infinity as we can see from

f (x) − f (0) 1
lim = lim 2 = ∞.
x→0 x−0 x→0 x

10

−2

−4

−6

−8

−10
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2

Problem 8.3 Obtain the rule

d 1 f ′ (x)
=−
dx f (x) f (x)2
from Theorem 7.7 in the textbook and also directly from the definition of the derivative.

Theorem 7.7 says that if g, h are functions differentiable at x0 and h(x0 ) ̸= 0, then

d (g) g ′ (x0 )h(x0 ) − g(x0 )h′ (x0 )

(x0 ) = .
dx h g(x0 )2

Here we take g(x) = 1 (constant function) and h(x) = f (x), so that if f is differentiable at x0 and
f (x0 ) ̸= 0, then we have

d 1 1′ · f (x0 ) − 1 · f ′ (x0 ) f ′ (x0 )

(x0 ) = = − ,
dx f f (x0 )2 f (x0 )2
which is what we wanted to prove.

We prove the same result directly from the definition of the derivative, assuming that f is differen-
tiable at x0 and f (x0 ) ̸= 0.

d 1
1
f (x) − 1
f (x0 ) f (x0 ) − f (x) f (x) − f (x0 ) 1
(x0 ) = lim = lim = lim − .
dx f x→x0 x − x0 x→x0 f (x0 )f (x)(x − x0 ) x→x0 x − x0 f (x0 )f (x)

Now, since f is differentiable at x0 , it is continuous at x0 and thus the limit limx→x0 f (x) exists and is
equal to f (x0 ). Moreover, f (x0 ) ̸= 0, so we get by Theorem 5.18 (limit of quotients)

d 1 1 f (x) − f (x0 ) 1 1 1 f ′ (x0 )

(x0 ) = − lim lim =− f ′ (x0 ) =− .
dx f f (x0 ) x→x0 x − x0 x→x0 f (x) f (x0 ) f (x0 ) f (x0 )2
In the calculation above we were allowed to split the limit of product of two functions into product of
limits, because the limits exist (see Theorem 5.17). Since f (x1 0 ) is a constant, we also used Theorem

June 7, 2017 3 Karel Švadlenka

5.15 to bring it in front of the limit sign.

Problem 8.4 Find a formula for the derivative of the function arctan x assuming that the usual formula
for
d
tan x = sec2 x
dx
has been found.

sin x
The function tan x = cos x is not defined at points x where cos x = 0 and it is not one-to-one on the
whole R but if we restrict its domain to the interval (− π2 , π2 ) then the function is defined and one-to-one,
so its inverse function arctan x is well-defined. Moreover, tan x maps the interval (− π2 , π2 ) to the whole
interval (−∞, ∞) and thus the inverse function arctan x is defined on the whole of R. (Draw the graph
of tan x and arctan x.)

The derivative of tan x is

( )
′ sin x ′ sin′ x cos x − sin x cos′ x cos2 x + sin2 x 1
tan x = = 2
= 2
= = sec2 x,
cos x cos x cos x cos2 x
hence we see that it never becomes 0 in the interval (− π2 , π2 ).

To calculate the derivative of arctan x, we use Theorem 7.32, which gives the formula
1
(f −1 )′ (f (x)) = .
f ′ (x)
This means that the derivative of the inverse function of f at the point f (x) is equal to one over the value
of the derivative of f at the point x.

Inserting tan into f , we have

1
arctan′ (tan x) = = cos2 x.
tan′ x
Since we want to have a formula for the derivative of arctan at a given point y, we perform the change
of variables y = tan x. Then we have to express cos2 x in terms of y = tan x. This is done by
cos2 x 1 1
cos2 x = 2 = 2x = .
2
cos x + sin x sin
1 + cos2 x 1 + y2

Therefore,
1
(arctan y)′ = ,
1 + y2
which is the desired formula.

Problem 8.5 Find the maximum and minimum values of the function
x+1
f (x) =
x2 + 1
on the interval [−1, 21 ].

The function is differentiable at all points of the interval [−1, 12 ]. Thus it is sufficient to check the
values of the function at points x where f ′ (x) = 0 and at the endpoints −1 and 12 .

June 7, 2017 4 Karel Švadlenka

 • We find all points x satisfying f ′ (x) = 0.
(x + 1)′ (x2 + 1) − (x + 1)(x2 + 1)′ x2 + 1 − (x + 1) · 2x −x2 − 2x + 1
f ′ (x) = = =
(x2 + 1)2 (x2 + 1)2 (x2 + 1)2
and so the derivative equal to zero at points
√ √
x1 = −1 + 2, x2 = −1 − 2.

The point x1 belongs to the considered interval but the point x2 does not, which leads to taking
into account only √ √ √
−1 + 2 + 1 2 1+ 2
f (x1 ) = √ = √ = .
(−1 + 2)2 + 1 4−2 2 2

• The values at the endpoints are

1 6
f (−1) = 0, f ( ) = = 1.2,
2 5

√ √
and since 2 = 1.414 . . . , we conclude that the maximum on the interval is achieved at x = −1 + 2
and the minimum at x = −1. This is also visible from the graph of the function below.

1.4
(x+1)/(x2+1)

1.2

0.8

0.6

0.4

0.2

0
−1 −0.5 0 0.5

Problem 8.6 Use Rolle’s theorem to explain why the cubic equation

x3 + ax + b = 0

cannot have more than one solution whenever a > 0.

Let us assume that there exist two different solutions x1 and x2 . Without loss of generality, assume
that x1 < x2 . Setting
f (x) = x3 + ax + b,
we see that f (x1 ) = f (x2 ) = 0 and hence, by Rolle’s theorem, there exists a number c ∈ (x1 , x2 ) such
that f ′ (c) = 0.

We compute the derivative of f as follows:

f ′ (x) = 3x2 + a.

June 7, 2017 5 Karel Švadlenka

Since a > 0 and 3x2 is always nonnegative, we conclude that the derivative of f is always positive. This
is a contradiction with the existence of c such that f ′ (c) = 0 and the statement is proved.

Problem 8.7 Suppose that f is continuous on [a, b] and differentiable on (a, b). If
lim f ′ (x) = C,
x→a+

what can you conclude about the right-hand derivative of f at a?

By the Mean value theorem, for any x ∈ (a, b) there exists a number c(x) (depending on x) between
a and x, such that
f (x) − f (a)
= f ′ (c)
x−a
(check that the assumptions of the Mean value theorem are satisfied).

Since the right-hand derivative of f at a (if it exists) is defined as the limit

f (x) − f (a)
f+′ (a) = lim ,
x→a+ x−a
we find that
f+′ (a) = lim f ′ (c(x)).
x→a+
Now, because c(x) ∈ (a, x), we see that y = c(x) → a+ as x → a+. Therefore,
f+′ (a) = lim f ′ (y) = C.
y→a+

Problem 8.8 Prove that if a function F is differentiable on a neighborhood of x0 with F ′ (x0 ) > 0 and
F ′ is continuous at x0 , then F is increasing on some neighborhood of x0 .

The statement that F is differentiable on a neighborhood of x0 means that there exists a number a > 0
such that F is defined on (x0 −a, x0 +a) and the derivative of F exists at every point x ∈ (x0 −a, x0 +a).
Moreover we know that F ′ is continuous at x0 , which means that
lim F ′ (x) = F ′ (x0 ) > 0
x→x0

by assumption.

Then, similarly as in Theorems 5.11 or 5.12, we can find a number c > 0 (smaller than a), such that
F ′ (x)> 0 for x ∈ (x0 − c, x0 + c). Let us prove this. Denoting F ′ (x0 ) > 0 by L, from the definition
of limit, taking ε = L2 , we can find δ > 0 such that
|F ′ (x) − F ′ (x0 )| < L
2 for every x ∈ (x0 − δ, x0 + δ).
This implies
F ′ (x) > F ′ (x0 ) − L
2 =L− L
2 = L
2 > 0,
and hence if we take c = δ, we have
F ′ (x) > 0 on (x0 − c, x0 + c).

By Theorem 7.24(ii), the function F is increasing on (x0 − c, x0 + c), which is a neighborhood of

x0 . (If the above proof is confusing due to the appearance of the derivative F ′ , one can replace F ′ by
some other symbol, such as f , and work with f .)

June 7, 2017 6 Karel Švadlenka

 Calculus A: Problems with Solutions (Lecture 9)

Problem 9.1 Let f be convex on an open interval (a, b). Must f be bounded above? Must f be bounded
below?

The function
1
f (x) =
, x ∈ (0, 1)
x
is convex on the interval (0, 1) (because the second derivative f ′′ (x) = 2/x3 is positive in this interval)
but it is not bounded above, which gives a counterexample to the first question.

To answer the second question, notice that f must be continuous (Theorem 7.34) and that there are
only three possibilities for a convex function on (a, b):
1. f is nonincreasing on (a, b),
2. f is nondecreasing on (a, b),
3. there is a number c such that f is nonincreasing on (a, c] and nondecreasing on [c, b).

(The proof is obtained by the fact that in all other cases there would exist points x1 < x2 < x3 with
f (x1 ) < f (x2 ) and f (x2 ) > f (x3 ), which is not possible for a convex function.)

In the third case, the value at c is the minimum value of f on (a, b), so the function is bounded below.

In the second case, let us take any two points x1 , x2 ∈ (a, b) such that x1 < x2 . Then the values
of the function f in the interval (a, x1 ) cannot lie below the chord L connecting the points (x1 , f (x1 ))
and (x2 , f (x2 )). (If there is a point (x3 , f (x3 )) lying below the chord, then the point (x2 , f (x2 )) would
lie above the chord connecting (x3 , f (x3 )) and (x1 , f (x1 )), which would yield a contradiction with the
convexity of f .) Since the function is nondecreasing on (a, b), its values on the interval [x1 , b) must be
greater than or equal to any value of the function in the interval (a, x1 ). But these values are above the
chord L, and thus the function is bounded below by the value of the y-coordinate of the intersection of
the chord L and the line x = a.

y b

f (x)
L
b

b
b

a b
b x1 x2 x

The first case is dealt with in the same way as the second case. Hence, we have shown that a convex
function on an interval (a, b) is always bounded below.
Problem 9.2 Consider the function
3x − 2x
h(x) = , x ̸= 0.
x
(1) What value should be assigned to h(0) in order that h be everywhere continuous?
(2) Does h′ (0) exist if this value is assigned to h(0)?
(3) Would it be correct to calculate h′ (0) by computing instead h′ (x) by the usual rules of the calculus
and finding limx→0 h′ (x)?

(1) What value should be assigned to h(0) in order that h be everywhere continuous?
If the limit limx→0 h(x) = L exists, then by defining h(0) = L the function will be continuous
everywhere (because it is continuous at all other points except 0). Setting
f (x) = 3x − 2x , g(x) = x,
the limit is of the 00 form, so we consider applying L’Hospital’s rule. Obviously, the assumptions
of Theorem 7.38 are fulfilled, so if the limit on the right-hand side below exists, we have
f (x) f ′ (x)
lim h(x) = lim = lim ′ .
x→0 x→0 g(x) x→0 g (x)

We calculate the last limit as follows:

f ′ (x) 3x ln 3 − 2x ln 2 3
lim ′
= lim = ln 3 − ln 2 = ln .
x→0 g (x) x→0 1 2
Therefore, if we set h(0) = ln 23 , the function h will be continuous on R.
(2) Does h′ (0) exist if this value is assigned to h(0)?
We attempt to find the derivative from definition (assuming that it exists):
3x −2x
′ h(x) − h(0) x − ln 23 3x − 2x − x ln 32
h (0) = lim = lim = lim .
x→0 x−0 x→0 x x→0 x2
This limit is again of the type 00 , so checking the assumptions of L’Hospital’s rule, we apply it
twice to get
3x − 2x − x ln 23 3x ln 3 − 2x ln 2 − ln 32 3x ln2 3 − 2x ln2 2 ln2 3 − ln2 2
lim = lim = lim = .
x→0 x2 x→0 2x x→0 2 2
(3) Would it be correct to calculate h′ (0) by computing instead h′ (x) by the usual rules of the calculus
and finding limx→0 h′ (x)?
Calculating the derivative h′ (x) for x ̸= 0, we obtain
3x (x ln 3 − 1) − 2x (x ln 2 − 1)
h′ (x) = ,
x2
and using L’Hospital’s rule we find that
ln2 3 − ln2 2
lim h′ (x) = .
x→0 2
Hence, the result is identical to the result of (2).
This calculation is justified by solved Problem 8.7 provided the limit limx→0 h′ (x) exists.

June 14, 2017 2 Karel Švadlenka

Problem 9.3 Calculate the limit
ex − 1 − x + x2
lim .
x→0 x2

Setting f (x) = ex − 1 − x + x2 and g(x) = x2 , we see that f (0) = 0 and g(0) = 0, so the limit
cannot be computed just by substituting x = 0 into the function. This leads to applying the L’Hospital’s
rule of the form 00 . We check the assumptions of Theorem 7.38:
• f (0) = g(0) = 0 is true as we already saw above.

• For every neighborhood of x = 0 (without the point x = 0) the function g(x) = x2 is positive, so
it never equals to zero.
f ′ (x)
• To see whether lim exists, we compute
x→0 g ′ (x)

f ′ (x) ex − 1 + 2x
=
g ′ (x) 2x

but this is again an expression of the type 00 as x → 0. Therefore, we have to apply L’Hospital’s
rule again to this function. To this end, set f1 (x) = ex − 1 + 2x and g1 (x) = 2x and confirm that
the assumptions of Theorem 7.38 are satisfied.
– f1 (0) = g1 (0) = 0 is true.
– For every neighborhood of x = 0 (without the point x = 0) the function g1 (x) = 2x is not
equal to zero.
f ′ (x)
– The limit lim 1′ exists:
x→0 g1 (x)

f1′ (x) ex + 2 3
lim ′ = lim = .
x→0 g1 (x) x→0 2 2

Therefore, the assumptions of L’Hospital’s theorem 7.38 for the limit limx→0 fg(x)
(x)
are fulfilled and we
conclude
ex − 1 − x + x2 f ′ (x) f1′ (x) 3
lim 2
= lim ′
= lim ′ = .
x→0 x x→0 g (x) x→0 g1 (x) 2

Problem 9.4 Prove a second-order version of the mean value theorem:

Let f be continuous on [a, b] and twice differentiable on (a, b). Then there exists c ∈ (a, b) such that
1
f (b) = f (a) + (b − a)f ′ (a) + (b − a)2 f ′′ (c).
2
(Hint: Imitate the proof of the mean value theorem in the textbook but instead of subtracting a linear
function, subtract a quadratic function of the form

h(x) = f (x) − f (a) − f ′ (a)(x − a) − β(x − a)2

for an appropriate number β.)

Define the quadratic function Q(x) by

Q(x) = f (a) + f ′ (a)(x − a) + β(x − a)2

June 14, 2017 3 Karel Švadlenka

and the function h(x) by
h(x) = f (x) − Q(x).
We want to apply Rolle’s theorem to h(x) on (a, b), so we choose β, so that h(a) = h(b), that is we take
f (b) − f (a) − f ′ (a)(b − a)
β= .
(b − a)2
Then we have h(a) = h(b) = 0 for h which is continuous on [a, b] and differentiable on (a, b), so by
Rolle’s theorem, there is a point d ∈ (a, b), such that h′ (d) = 0. Moreover, since
h′ (x) = f ′ (x) − Q′ (x) = f ′ (x) − f ′ (a) − 2β(x − a),
we see that h′ (a) = h′ (d) = 0. Additionally, h′ (x) is continuous on [a, d] (since f ′ (a) exists) and
differentiable on (a, d), so we can apply Rolle’s theorem once again to h′ (x) on (a, d) to get the existence
of a point c ∈ (a, d) such that h′′ (c) = 0. As h′′ (c) = f ′′ (c) − 2β = 0, we conclude that
f (b) − f (a) − f ′ (a)(b − a) 1
β= = f ′′ (c),
(b − a)2 2
which is the identity to be shown.

There is another easier method of proof, which, however, requires a clever idea. We define β as above
and set [ ]
g(x) = f (b) − f (x) + f ′ (x)(b − x) + β(b − x)2 .
Then by the definition of β, we see that g(a) = g(b) = 0, while g is continuous on [a, b] and differen-
tiable on (a, b), and thus by Rolle’s theorem there is c ∈ (a, b) such that g ′ (c) = 0. By differentiating g
we compute that g ′ (c) = (b − c)(2β − f ′′ (c)) = 0, which gives 2β = f ′′ (c) since c ̸= b.

Problem 9.5* Suppose f is convex on an open interval I. Prove that if f is differentiable on I, then f ′
is continuous on I.

Since f is convex and differentiable on I, we immediately get from Theorem 7.35 that f ′ is nonde-
creasing on I. Hence, by Theorem 5.60 and the considerations following it, the function f ′ can have
only a countable number of jump discontinuities (because both right-hand and left-hand limits exist at
each point, which is also contained in Theorem 7.34).

Assume that the derivative f ′ is discontinuous at least at one point x0 ∈ I. Then it means that
lim f ′ (x) < lim f ′ (x) and f ′ (x0 ) ∈ [ lim f ′ (x), lim f ′ (x)].
x→x0 − x→x0 + x→x0 − x→x0 +

Thus, f ′ (x0 ) must be different from at least one of the one-sided limits
L− = lim f ′ (x), L+ = lim f ′ (x).
x→x0 − x→x0 +

Without loss of generality, assume that f ′ (x0 ) < L+ . Then setting

f ′ (x0 ) + L+
γ= ,
2
the function f ′ (x) does not take the value γ at any point of I (because it is nondecreasing, f ′ (x0 ) < γ
and the right-hand limit at x0 is greater than γ). But this is a contradiction with the fact that derivative
has the Darboux property (Theorem 7.31).

June 14, 2017 4 Karel Švadlenka

 Calculus A: Problems with Solutions (Lecture 10)

Problem 10.1 Calculate

∫ 1
x2 dx
0
by partitioning [0, 1] into subintervals of equal length.

The function x2 is continuous on [0, 1], so the integral is defined. We partition the interval [0, 1] into
n subintervals [xk−1 , xk ], k = 1, 2, . . . , n of equal length, where
1 2 n−1
x0 = 0, x1 = , x2 = , . . . , xn−1 = , xn = 1.
n n n
Then since xk − xk−1 = 1
n for all k = 1, 2, . . . , n, we have for the Riemann sums

∑
n
1∑ 2
n
Sn (f ) = f (ξk )(xk − xk−1 ) = ξk .
n
k=1 k=1

Here ξk is any point in the interval [xk−1 , xk ].

Since we know from theorem 8.1 that the Riemann sums converge to the given definite integral for
any associated points ξk , we select ξk so that the sum is easy to compute. For example, we can take
ξk = xk = nk , that is, the associated point is the right endpoint of the interval [xk−1 , xk ]. Then we have
( )2
1∑ 1 ∑ 2
n n
k 1 n(n + 1)(2n − 1)
Sn (f ) = = 3 k = 3 .
n n n n 6
k=1 k=1

The definite integral is the limit of these Riemann sums as the length of the subintervals goes to zero:
∫ 1
1 n(n + 1)(2n − 1) 1
x2 dx = lim Sn (f ) = lim 3 = .
0 n→∞ n→∞ n 6 3

Problem 10.2 Let b > a > 0 and p ∈ R. Calculate

∫ b
xp dx
a

by partitioning [a, b] into subintervals [a, aq], [aq, aq 2 ], [aq 2 , aq 3 ], . . . , [aq n−1 , b], where aq n = b.

Since the function f (x) = xp is continuous on the interval [a, b] (if b > a > 0), the integral is defined
and is equal to the limit of arbitrary Riemann sum. Setting the partition as xk = aq k , k = 0, 1, . . . , n
and the associated points as ξk = xk−1 = aq k−1 , we have the Riemann sum
∑
n ∑
n ∑
n
Sn (f ) = f (ξk )(xk − xk−1 ) = (aq ) (aq − aq ) = a (q − 1)
k−1 p k k−1 p+1
q (p+1)(k−1) .
k=1 k=1 k=1
By the formula for the partial sum of geometric progression

∑
n−1
dn − 1
dk =
d−1
k=0

applied to d = q p+1 , we have

∑
n−1
q n(p+1) − 1 ( ab )p+1 − 1
Sn (f ) = ap+1 (q − 1) (q p+1 )k = ap+1 (q − 1) = a p+1
(q − 1) ,
q p+1 − 1 q p+1 − 1
k=0

where in the last equality we used the fact that aq n = b.

Now we need to take the limit as n → ∞ in the above. When n → ∞ then q = (b/a)1/n → 1+
since b/a > 1. Hence,

( ab )p+1 − 1 ( ) q−1
lim Sn (f ) = lim ap+1 (q − 1) = a p+1
( b p+1
) − 1 lim .
n→∞ q→1+ q p+1 − 1 a q→1+ q p+1 − 1

The last limit is of the type 00 , so we use l’Hospital’s rule (check the assumptions!) to find

q−1 1 1
lim = lim = .
q→1+ q p+1 − 1 q→1+ (p + 1)q p p+1

Putting all the results together, we conclude

∫ ( ) 1
b
bp+1 − ap+1
xp dx = ap+1 ( ab )p+1 − 1 = ,
a p+1 p+1
a well-known formula.

Problem 10.3 If f is continuous and f (x) ≥ 0 for all x in [a, b], show that
∫ b
f (x) dx ≥ 0.
a

Since the function f is continuous, the integral is well-defined and equals to the limit of Riemann
sums
∑n
f (ξk )(xk − xk−1 ),
k=1

when the largest length of the subintervals [xk−1 , xk ] of the partition

a = x0 < x1 < · · · xn−1 < xn = b

goes to zero.

Here, xk−1 < xk and f (ξk ) ≥ 0 for every ξk ∈ [a, b], so f (ξk )(xk − xk−1 ) ≥ 0 and we see that any
possible Riemann sum must be greater than or equal to zero. Since the definite integral above is given
as the limit of the Riemann sums, it follows by the order property of limits that it is nonnegative.

June 20, 2017 2 Karel Švadlenka

Problem 10.4 If f is continuous and m ≤ f (x) ≤ M for all x in [a, b], show that
∫ b
m(b − a) ≤ f (x) dx ≤ M (b − a).
a

From the monotone property of the integral we know that if f (x) ≤ M for all x ∈ [a, b] then
∫b ∫b
a f (x) dx ≤ a M dx (here g(x) in 8.6 of the textbook is taken as the constant function g(x) = M ).
Since the last integral is equal to M (b − a), we have proved the second inequality.

We can prove the statement also directly from the definition of the integral. Since the function f is
continuous, the integral is well-defined and equals to the limit of Riemann sums

∑
n
f (ξk )(xk − xk−1 )
k=1

as n → ∞ (or, more precisely, as the length of the longest interval in the partition a = x0 < x1 <
· · · xn−1 < xn = b goes to zero). Now, f (ξk ) ≤ M for every possible ξk and therefore

∑
n ∑
n
f (ξk )(xk − xk−1 ) ≤ M (xk − xk−1 ) = M (b − a).
k=1 k=1

By the order property of limit, we conclude that the second inequality in the problem holds.

The proof of the first inequality is analogous.

Problem 10.5 Express

( )
1∑
n
k
lim f
n→∞ n n
k=1

as a definite integral, where f is continuous on [0, 1].

Let us set xk = nk , k = 0, 1, . . . , n. Then these points form a partition of the interval [0, 1] into
subintervals [xk−1 , xk ], k = 1, 2, . . . , n of length n1 , whence the above sum can be written as
( ) ∑ ( )
1∑
n n
k k
f = f (xk − xk−1 ).
n n n
k=1 k=1

This is exactly a Riemann sum for ∫ 1

f (x) dx
0

with the partition x0 , x1 , . . . , xn and associated points nk = xk ∈ [xk−1 , xk ]. Taking n → ∞ makes the
length of subintervals go to zero, and, since f is continuous on [0, 1], Theorem 8.1 guarantees that the
limit is equal to the above definite integral.

June 20, 2017 3 Karel Švadlenka

Problem 10.6 If f is continuous and nonnegative on an interval [a, b] and
∫ b
f (x) dx = 0,
a

show that f is identically equal to zero there.

Let us prove the statement by contradiction. Assume f is nonnegative but not identically equal to
zero on [a, b]. This means that there exists a point c ∈ [a, b], such that d = f (c) > 0. Hence, the
function is greater than d2 on some neighborhood of the point c, that is, there is δ > 0 such that

d
f (x) > for every x ∈ (c − δ, c + δ).
2
This can be proved, for example, as in Theorem 5.12.

We have by the additive property 8.4 of the integral,

∫ b ∫ c−δ ∫ c+δ ∫ b
f (x) dx = f (x) dx + f (x) dx + f (x) dx.
a a c−δ c+δ

Here the first and third integral on the right-hand side are nonnegative by the monotone property 8.6 of
the integral (because f (x) ≥ 0). Therefore, dropping these two integrals the right-hand side will not
increase, and we obtain
∫ b ∫ c+δ
d
f (x) dx ≥ f (x) dx > ((c + δ) − (c − δ)) = dδ > 0.
a c−δ 2

The second inequality follows from the monotone property, since f (x) > d2 on the interval (c − δ, c + δ).
We got a contradiction with the assumption that the integral of f is zero, which concludes the proof.

Problem 10.7 If f is continuous and m ≤ f (x) ≤ M for all x in [a, b], show that
∫ b ∫ b ∫ b
m g(x) dx ≤ f (x)g(x) dx ≤ M g(x) dx
a a a

for any continuous, nonnegative function g.

Since the function g is nonnegative on [a, b], we can multiply the number g(x) ≥ 0 for a fixed
x ∈ [a, b] to the inequality m ≤ f (x) ≤ M without changing its validity, thus obtaining

mg(x) ≤ f (x)g(x) ≤ M g(x) for all x ∈ [a, b].

Then we just apply the monotone property of the integral

∫ b ∫ b
h1 , h2 continuous, h1 (x) ≤ h2 (x) for all x ∈ [a, b] ⇒ h1 (x) dx ≤ h2 (x) dx
a a

to the pairs of functions h1 (x) = mg(x), h2 (x) = f (x)g(x) and h1 (x) = f (x)g(x), h2 (x) = M g(x),
respectively, obtaining the statement of the problem.

June 20, 2017 4 Karel Švadlenka

Problem 10.8 Calculate the following integrals:
∫ 1/2
(1) ex sin πx dx
0
∫ 2
sin(ln x)
(2) dx
1 x

(1) Since we have a product of two functions where ex when differentiated does not change and sin
when differentiated twice becomes a multiple of sin, we expect that integration by parts will lead
to the result.
In this case, setting g(x) = g ′ (x) = ex , f (x) = sin πx, we see that f, g, f ′ , g ′ are all continuous
on [0, 21 ] and thus we can use the integration by parts formula
∫ b ∫ b
′
f (x)g (x) dx = [f (b)g(b) − f (a)g(a)] − f ′ (x)g(x) dx
a a

to obtain
∫ 1/2 [ 1 ] ∫ 1/2 ∫ 1/2
π √
e sin πx dx = e sin − e sin 0 − π
x 2
0
e cos πx dx = e − π
x
ex cos πx dx.
0 2 0 0

Applying the integration by parts this time to g(x) = g ′ (x) = ex and f (x) = cos πx, we have
∫ 1/2 [ 1 ] ∫ 1/2 ∫ 1/2
π
ex cos πx dx = e 2 cos − e0 cos 0 + π ex sin πx dx = −1 + π ex sin πx dx.
0 2 0 0

Plugging this into the previous identity, we get

∫ ( ∫ )
1/2 √ 1/2
e sin πx dx = e − π −1 + π
x
ex sin πx dx ,
0 0

hence ∫
2
1/2 √
(1 + π ) ex sin πx dx = e + π,
0
which yields
∫ 1/2 √
x e+π
e sin πx dx = .
0 1 + π2

(2) Since we have here the composed function sin(ln x) and the derivative of ln x is x1 , which also ap-
pears in the integral, we can probably use integration by substitution. g(x) = ln x is differentiable
on [1, 2] and f (x) = sin x is continuous on R, so the assumptions for the formula
∫ b ∫ g(b)
′
f (g(x))g (x) dx = f (s) ds
a g(a)

to be valid, are fulfilled. Hence (noting that g ′ (x) = x1 ),

∫ 2 ∫ ln 2
sin(ln x)
dx = sin s ds = cos 0 − cos(ln 2) = 1 − cos(ln 2).
1 x ln 1

June 20, 2017 5 Karel Švadlenka

 Calculus A: Problems with Solutions (Lecture 11)

Problem 11.1 Let f and g be continuous functions on (a, b] and such that |f (x)| ≤ |g(x)| for all a <
∫b ∫b
x ≤ b. If the integral a g(x) dx is absolutely convergent, show that so also is the integral a f (x) dx.

The function f may be discontinuous (infinite) at a. In that case the integral becomes improper and
(if it exists) is defined as ∫ b ∫ b
f (x) dx = lim f (x) dx.
a X→a+ X

By the assumption |f (x)| ≤ |g(x)| and the monotone property of the integral, we have
∫ b ∫ b
|f (x)| dx ≤ |g(x)| dx for any X > a.
X X

Note that since X > a, both the above integrals are defined since the integrands are continuous on [X, b].

Now by the order properties of limit

∫ b ∫ b
lim |f (x)| dx ≤ lim |g(x)| dx.
X→a+ X X→a+ X

But the last expression is the definition of the (improper) integral

∫ b
|g(x)| dx,
a

which, according to our assumption, is finite.

∫b
Therefore, we have shown that also a |f (x)| dx is finite, in other words, the integral of f is abso-
lutely convergent.

∫ 1
Problem 11.2 For what values of p is the integral xp dx convergent?
0

If p ≥ 0 then the integral is a usual integral of a continuous function, so there is no question of

convergence. However, the integral becomes improper when p < 0, so in this case we write by definition
∫ 1 ∫ 1
p
x dx = lim xp dx.
0 X→0+ X

There are two cases (when we compute the primitive function of xp ):

• p = −1: in this case we have
∫ 1
lim xp dx = lim [ln x]1X = lim (− ln X) = +∞,
X→0+ X X→0+ X→0+

so in this case the integral is not convergent.

 • p ̸= −1: in this case the primitive function is again a power of x, in particular,
∫ 1 [ ]1
xp+1 1 1 1
lim p
x dx = lim = lim (1 − X p+1 ) = − lim X p+1 .
X→0+ X X→0+ p+1 X p + 1 X→0+ p + 1 p + 1 X→0+

The last limit is finite (actually, zero) if and only if p > −1 (we are assuming p ̸= −1 here).

To conclude, the integral is convergent if and only if p > −1.

∫∞
Problem 11.3 Let f be a continuous function on [1, ∞) such that the integral 1 f (x) dx converges.
Can you conclude that limx→∞ f (x) = 0?

The function does not necessarily have to converge to zero at infinity. To see this, consider the
following function (draw a graph of it):
 2
 n x + 1 − n3 x ∈ [n − 1/n2 , n], n ∈ N
f (x) = −n2 x + 1 + n3 x ∈ [n, n + 1/n2 ], n ∈ N .

0 otherwise

This function has peaks of height 1 around each natural number and the area of the n-th peak is 1/n2 .
∑ 1
Since the series n2
converges, the integral of f is finite:
∫ ∞ ∫ X ∑
n ∞
∑
1 1
f (x) dx = lim f (x) dx = lim 2
= < ∞.
0 X→∞ 0 n→∞ k k2
k=1 k=1

However, the function values do not converge to 0 for x → ∞ since the function is equal to 1 at each
natural number.

∫∞
Problem 11.4 Let f be a continuous,∑∞ decreasing function on [1, ∞). Show that the integral 1 f (x) dx
converges if and only if the series n=1 f (n) converges. Give examples showing that both implications
in the statement are not true if the assumption of f being decreasing is omitted.

The improper integral is given by

∫ N
lim f (x) dx.
N →∞ 1

We consider a partition x1 , x2 , . . . , xN of the interval (1, N ) into subintervals [1, 2], [2, 3], [3, 4], . . . , [N −
1, N ] , i.e., xk = k and the corresponding Riemann sum for the above integral

∑
N −1 ∑
N
f (ξk )(xk+1 − xk ) = f (ξk ).
k=1 k=1

Since the function is decreasing, taking ξk = xk = k (the left end-point of the subinterval) as the
associated point, the Riemann sum will not be smaller than the value of the integral, and similarly taking
ξk = xk+1 = k + 1 (the right end-point of the subinterval), the Riemann sum will not be greater than
the value of the integral:
N∑−1 ∫ N ∑
N −1
f (k + 1) ≤ f (x) dx ≤ f (k).
k=1 1 k=1

June 28, 2017 2 Karel Švadlenka

Taking N to infinity, we deduce
∞
∑ ∫ ∞ ∞
∑
f (k) ≤ f (x) dx ≤ f (k),
k=2 1 k=1

which immediately implies that the convergence of the series and the integral are equivalent.

One counterexample for a function that is not decreasing is the function from∑the above problem
11.3. Indeed, the value of the function at natural numbers is always 1, so the series f (n) diverges but
the integral of f converges. Moreover, if we construct a function that is zero at every natural number
but has a peak of area 1 between every pair of neighboring natural numbers, then it will serve as a
counterexample for the opposite implication.

Problem 11.5 Find the value of the integral

∫ 0 1
ex
dx.
−∞ x2

The integral looks ”doubly” improper: it is over an infinite interval and the integrated function might
blow up to infinity at x = 0. However, the integrated function is continuous (bounded) in the left
neighborhood of zero:
1
ex
lim = lim y 2 ey = 0,
x→0− x2 y→−∞

since the exponential function is stronger at infinity than any polynomial (use l’Hospital’s rule to confirm
it).

Hence, we can define the integral as

∫ 0 1 ∫ 0 1
ex ex
dx = lim dx,
−∞ x2 X→ −∞ X x2

and a change of variables f (y) = e−y , g(x) = − x1 , g ′ (x) = 1

x2
(or y = − x1 , dy = 1
x2
dx) leads to
∫ 0 1 ∫ 0 ∫ g(0)
ex ′
lim dx = lim f (g(x))g (x) dx = lim f (y) dy.
X→ −∞ X x2 X→ −∞ X X→ −∞ g(X)

In fact, the value g(0) in the last interval is not defined, so we correct the computation as follows:
∫ 0 ∫ Z ∫ g(Z)
′ ′
lim f (g(x))g (x) dx = lim lim f (g(x))g (x) dx = lim lim f (y) dy.
X→ −∞ X X→ −∞ Z→0− X X→ −∞ Z→0− g(X)

This can already be calculated by standard methods:

∫ −1/Z ( )
lim lim e−y dy = lim lim e1/X − e1/Z = lim e1/X = 1.
X→ −∞ Z→0− −1/X X→ −∞ Z→0− X→ −∞

June 28, 2017 3 Karel Švadlenka

 Calculus A: Problems with Solutions (Lecture 12)

Problem 12.1 Find the following primitive functions.

∫
1
(1) 2
dx
x +x
∫
x
(2) √ dx
1 − x4
∫
x3 + x + 2
(3) dx
x4 + 2x2 + 1
∫
(4) sec x dx

Each primitive function is determined up to a constant. We omit these constants in the calculations
below.
(1) Since 1
x2 +x
= 1
x(x+1) = 1
x − 1
1+x , we have
∫ ∫ ( )
1 1 1 x
2
dx = − dx = ln |x| − ln |1 + x| = ln .
x +x x 1+x 1+x

We use the absolute value to account for the possibility of negative argument, i.e., the primitive
function of x1 is ln x + C for x > 0 and ln(−x) for x < 0, which is written compactly by
employing the absolute value.

(2) We use the integration by substitution formula

∫ ∫
′
f (g(x))g (x) dx = f (y)dy

with g(x) = x2 and f (y) = √ 1 . Then g ′ (x) = 2x and

2 1−y 2
∫ ∫ ∫ ∫
x ′ 1
√ dx = f (g(x))g (x) dx = f (y) dy = √ dy
1 − x4 2 1 − y2
1 1
= arcsin y = arcsin(x2 ).
2 2
It is usually easier to use the following formalism instead of the functions f, g above: Set the new
dy
variable y = x2 , then dx = 2x and so formally xdx = 12 dy, hence replacing x dx by 21 dy in the
integral and changing x2 to y, we get the integral
∫
1
√ dy,
2 1 − y2

which is the same as above.

 x3 +x+2 x(x2 +1)+2 x 2
(3) As x4 +2x2 +1
= (x2 +1)2
= x2 +1
+ (x2 +1)2
, we can split the integral into two parts:
∫ ∫ ∫
x3 + x + 2 x 2
dx = dx + dx.
x4 + 2x2 + 1 2
x +1 (x2 + 1)2

Now, the first integral is simple - we use the substitution y = x2 + 1 with dy = 2x dx to get
∫ ∫
x 1 1 1
2
dx = dy = ln y = ln(x2 + 1),
x +1 2y 2 2
so it remains to calculate the second integral.
( )′
x 1 − x2
We notice that = , and thus we can write
x2 + 1 (x2 + 1)2
∫ ∫ ( )
2 1 − x2 1 x
dx = + dx = 2 + arctan x.
(x2 + 1)2 (x2 + 1)2 x2 + 1 x +1

Therefore, the final answer is

∫
x3 + x + 2 1 x
dx = ln(x2 + 1) + 2 + arctan x.
x4 + 2x2 + 1 2 x +1

(4) To get a form where substitution can be used, we first rearrange as

∫ ∫ ∫ ∫
cos x cos x cos x
sec x dx = dx = dx = dx.
2
cos x 1 − sin2 x
Now we can substitute y = sin x with dy = cos x dx to get
∫ ∫ ∫
cos x 1
sec x dx = dx = dy.
1 − sin2 x 1 − y2
This integral can already be calculated by decomposing into simpler rational functions:
∫ ∫ ( )
1 1 1 1 1 1 1+y
dy = + dy = (ln |1 + y| − ln |1 − y|) = ln .
1−y 2 2 1+y 1−y 2 2 1−y

Hence our integral is ∫

1 1 + sin x
sec x dx = ln ,
2 1 − sin x
1+sin x
which can be also written as (after multiplying by 1+sin x )
∫
1 + sin x
sec x dx = ln = ln |sec x + tan x| .
cos x

July 5, 2017 2 Karel Švadlenka

 Calculus A: Problems with Solutions (Lecture 13)

∑∞
Problem 13.1 If the series k=1 ak converges, what can you say about the series
∞
∑
(a2k + a2k−1 ) ?
k=1

∑∞
This series converges and has the same sum as k=1 ak .

To see this, first write out the new series:

∞
∑
(a2k + a2k−1 ) = (a1 + a2 ) + (a3 + a4 ) + (a5 + a6 ) + · · · + (a2K−1 + a2K ) + · · · .
k=1

Hence we see that the sequence of partial sums of this series is equal to

∑
2K
s2K = ak , K = 1, 2, 3, . . . ,
k=1

which is a subsequence of the partial sums of the original series

∑
K
sK = ak .
k=1
∑∞
Since the meaning of ”series k=1 ak converges and its sum is S” is that ”the sequence of partial sums
{sK } converges and its limit is S”, we conclude
∑∞ that any subsequence of {sK }, so also {s2K }, must
converge to S, which means that the series k=1 (a2k + a2k−1 ) converges and its sum is S.

The key to a correct answer here is to consider first only the partial sums, which are finite and
therefore, we can change order of summation. This will be even more pronounced in problem 13.4.

∞
∑ ∞
∑
Problem 13.2 If the series (a2k + a2k−1 ) converges, what can you say about the series ak ?
k=1 k=1

, i.e., the sequence −1, 1, −1, 1, −1, . . . shows.

We cannot say anything as the sequence ak = (−1)k∑
Here a2k + a2k−1 = 0 for any k which implies that ∞ k=1 (a2k + a2k−1 ) converges (it is a sum of
∑K
zeros). However, the partial
∑∞ sums s K = a
k=1 k are −1, 0, −1, 0, −1, 0, . . . , which is not a convergent
sequence, so the series k=1 ak does not converge.
 ∞
∑ ( )
k+1
Problem 13.3 Does the series ln converge?
k
k=1

Using the properties of logarithm, we can write the partial sums of this series as
∑ ( ) ∑
K
k+1
K
( )
sK = ln = ln(k + 1) − ln k
k
k=1 k=1

and thus it is a telescoping series with partial sums sK = ln(K + 1). When K → ∞, sK diverges to
infinity, hence the given series is divergent.

∑
Problem 13.4 Show that a series ∞ k=1 ak is absolutely convergent if and only if at least two of the
series
∑∞ ∑∞ ∑∞
ak , +
[ak ] , and [ak ]− (1)
k=1 k=1 k=1
converge.

Here, [ak ]+ means the positive part of ak , that is, [ak ]+ = max{0, ak } (we are summing only the
positive terms in the series). Similarly, [ak ]− = − min{0, ak } is the negative part.

Since it is an equivalence statement (”if and only if”) we have to prove two implications.
∑∞
• First,
∑∞ let us prove that if the series k=1 ak is absolutely convergent (that is the number S =
k=1 |ak | is finite), then the three series (1) converge.
∑
For the first series ∞ k=1 ak this is exactly Theorem 3.14 in the textbook. Moreover,

∑
K ∑
K ∞
∑
[ak ] ≤ +
|ak | ≤ |ak | = S < ∞,
k=1 k=1 k=1

so the sequence of partial sums for the second series in (1) is nondecreasing (because all [ak ]+
are nonnegative) and bounded above by S, which means that it must converge (by the monotone
convergence theorem). For the third series in (1) the argument is completely analogous.
∑
• Next, let us prove that if two of the series (1) converge then ∞
k=1 ak is absolutely convergent.
Consider, for example, the case when the first two series converge. We denote their sums as
follows:
∞
∑ ∑∞
a k = S1 , [ak ]+ = S2 .
k=1 k=1
We notice that
∑
K K (
∑ ) ∑K ( ) ∑
K ∑
K
|ak | = [ak ]+ − [ak ]− = [ak ]+ − (ak − [ak ]+ ) = 2 [ak ]+ − ak .
k=1 k=1 k=1 k=1 k=1

∑
The sequence of partial sums for the series ∞ k=1 |ak | is nondecreasing because all terms |ak | are non-
negative, and moreover it is bounded - to show this we use the above identity in the following way:
∑
K ∑
K ∑
K ∑
K ∑
K
|ak | = 2 [ak ] −
+
ak ≤ 2 +
[ak ] + ak ≤ 2S2 + |S1 | < ∞.
k=1 k=1 k=1 k=1 k=1

July 11, 2017 2 Karel Švadlenka

 ∑∞
An application of the monotone convergence theorem yields the convergence of k=1 |ak |.

Other combinations of converging sums in (1) are proved using a similar idea (do it in detail).

Problem 13.5 For which values of x does the following series converge?
∞
∑ xn
log n
n=2

We will consider several cases for x and check the convergence in each case by different methods
(just to show how to use them). If there is no log n in each term, the boundary value for convergence
would be x = ±1. Since log n does not have much strength comparing to the exponential growth of xn ,
we try to divide the interval for x in the same way.
1. x > 1
We use for example the root test (3.30 in the textbook):
√ n
x x x
lim n = lim √ n
≥ lim √n
= x > 1,
n→∞ log n n→∞ log n n→∞ n
√
so the series diverges. Here we used the fact that log n < n for n ≥ 2 and that limn→∞ n
n = 1.

2. x = 1
In this case we have the series
∞
∑ 1
.
log n
n=2
We can use the comparison test II (3.20 in the textbook) with harmonic series:
∞
∑
1 1 1
< and diverges,
n log n n
n=2

so the above series also diverges.

Note that the root test would not give us any information here because the limit would be 1.

3. x ∈ [0, 1)
Let us use the ratio test (3.28 in the textbook):

xn+1 log n log n

lim = lim x
n→∞ log(n + 1) xn n→∞ log(n + 1)
n
log(n + 1) + log n+1
= lim x
n→∞ log(n + 1)
( n )
log n+1
= lim x 1 +
n→∞ log(n + 1)
= x < 1,

so the series converges.

4. x ∈ (−1, 0)
∑∞
n=2 |x
In this case, the series of absolute values n / log n| is the same as in the previous case, so
the series converges absolutely.

July 11, 2017 3 Karel Švadlenka

 5. x = −1
Here the series becomes
∞
∑ (−1)n
,
log n
n=2
so we can use the alternating series test (3.43 in the textbook)
∑ to say that the series converges.
However, it does not converge absolutely since the series ∞ 1
n=2 log n diverges.

6. x < −1
We can use the trivial test:
xn
lim
n→∞ log n

x2n
does not exist (because the terms change signs and the subsequence log(2n) diverges to infinity),
so the series cannot converge.

In summary, the series converges absolutely for x ∈ (−1, 1), converges nonabsolutely for x = −1
and diverges otherwise.

Problem 13.6 Determine the values of s, for which the series

∞
∑ 1
n(log n)s
n=2

converges.

If we try to use the root test or ratio test, we end up with the indecisive limit 1, so these tests cannot
be used.

For s ≤ 0, we can use the comparison with harmonic series:

∞
∑
1 1 1
s
≥ and diverges,
n(log n) n n
n=1

so our series diverges for s ≤ 0.

For general s, the integral test (3.35 in the textbook) works: if s ̸= 1 then
∫ X ∫ log X [ ]
1 1 1 1 log X
lim dx = lim dy = lim −
X→∞ 2 x(log x)s X→∞ log 2 y s X→∞ s − 1 y s−1 log 2
[ ] ( )
1 1 X 1 1 1
= lim − = − lim
X→∞ s − 1 y s−1 log 2 s − 1 (log 2)s−1 X→∞ X s−1

We have used the change of variables y = log x. We see that the limit of the integral is finite if and only
if s > 1.

In the special case when s = 1, we have

∫ X ∫ log X
1 1
lim dx = lim dy = lim [log y]log X
lim log(log X) − log(log 2)
log 2 = X→∞
X→∞ 2 x log x X→∞ log 2 y X→∞

In this case, the limit becomes infinite, so the series does not converge.

July 11, 2017 4 Karel Švadlenka

 REFERENCES

In summary, our series converges if and only if s > 1.

(See also http://math.feld.cvut.cz/mt/txte/2/txe3ec2.htm for more examples of testing convergence of

series.)

Additional Problems
Problem 1 ([2], p.489) Decide whether each of the following series is convergent or divergent:
∞
∑ sin nθ 1 1 1 1 2 1 2 1 2 1
(1) , θ∈R (2) 1−+ − +··· (3) 1 − + − + − + − + · · ·
n2 3 5 7 2 3 3 4 4 5 5
n=1
∑∞ ∑∞ ∑∞ ∑∞ ∑∞
log n 1 1 n2 log n
(4) (−1)n (5) √ (6) √ (7) (8)
n=1
n
n=2
3
n −1
2
n=1
3 2
n +1 n=1
n!
n=1
n
∞
∑ 1 ∑∞ ∑∞ ∑∞
1 1 1
(9) (10) , k∈N (11) n
(12) (−1)n
log n (log n) k (log n) (log n)n
n=2 n=2 n=2 n=2
∑∞ ∑∞ ∑∞ ∑ ∞
n2 1 1 1
(13) 3
(14) sin (15) (16)
n +1 n n log n n(log n)2
n=1 n=1 n=2 n=2
∑∞ ∞
∑ ∞
∑ ∑ ∞
1 n! 2n n! 3n n!
(17) (18) (19) (20) .
n2 (log n) nn nn nn
n=2 n=1 n=1 n=1

Problem 2 ([1], p.413) Test the following series for convergence and absolute convergence:
∞
∑ ∑∞ ∑∞ ∑∞
n n! −n 1
(1) (−1) n
(2) (3) ne (4) √
3n − 2 2n
1 + n3
n=1 n=1 n=1 n=1
∑∞ ∑∞ ∑ ∞
n 1 nq
(5) (−1)n−1 2 (6) (−1)n−1 (7) , q∈R
n +1 log(log n) 1 + nq
n=1 n=3 n=1
∑∞ ∑∞ ∑∞
rn 1 cos(nθ)
(8) , r ∈ R (9) (−1) n−1
sin , p ∈ R (10) √ .
1 + r2n np n
n=1 n=1 n=1

References
[1] S.R. Ghorpade, B.V. Limaye: A Course in Calculus and Real Analysis, Springer, 2006.

[2] M. Spivak: Calculus, Publish or Perish, 2008.

July 11, 2017 5 Karel Švadlenka

 Calculus B: Problems with Solutions (Lecture 1)

Problem 1.1 Let

sin nx
fn (x) = √ .
n
Show that limn→∞ fn (x) = 0 but limn→∞ fn′ (0) = ∞.

The pointwise limit of the sequence of functions can be easily obtained from the squeeze theorem.
Notice that because sin x is bounded above by 1 and below by −1, we have that
1 sin nx 1
−√ ≤ √ ≤√ .
n n n
Since both expressions − √1n and √1n tend to 0 as n → ∞, the quantity fn (x) must also tend to 0 for
every x ∈ R, by the squeeze theorem.

We compute
n cos nx √ √
fn′ (0) = √ = n cos nx = n.
n x=0 x=0

Therefore, obviously fn′ (0) → ∞ as n → ∞.

The reason why we cannot interchange limit and differentiation in this case is that the assumption of
Theorem 9.34 that ”the sequence {fn′ (x)} converges uniformly on [a, b]” is not satisfied for any interval
√
around 0. Indeed, fn′ (x) = n cos nx does not even converge pointwise for any neighborhood of 0.

Problem 1.2 Examine the uniform limiting behavior of the sequence of functions
fn (x) = x2 e−nx .
On what sets can you determine uniform convergence?

First, we check the pointwise convergence to find the limit function, and then investigate if the con-
vergence is uniform.

We see that for x < 0 the exponent −nx is positive and increasing to infinity for n → ∞, so that
fn (x) → ∞. If x = 0 then fn (0) = 0 for all n so that the limit is 0. If x > 0 then the exponent −nx
is negative and decreases to −∞ as n → ∞, which implies fn (x) → 0. (To be precise we should show
it from definition, that is for any given ε find N so that x2 e−nx < ε when n ≥ N . But that is easy, just
take N > (2 ln x − ln ε)/x. The same for the case x < 0.) We conclude
{
∞ if x < 0
lim fn (x) = .
n→∞ 0 if x ≥ 0

Since ∞ is not a number and therefore the limit function is not defined for x < 0, we investigate
uniform convergence on the interval [0, ∞) only. The graph of the function fn (x) looks like in the
following figure (here n = 3).
 0.07

0.06

0.05

0.04

0.03

0.02

0.01

0
0 0.5 1 1.5 2 2.5 3 3.5 4

It is easy to check that it is increasing on x ∈ (0, n2 ) and decreasing on ( n2 , ∞), where the maximum
is attained at x = n2 and is equal to n42 e−2 .

In order for the function to converge uniformly to the zero function on [0, ∞), it is necessary that the
values of fn are smaller than any given ε > 0 if n is large enough (that is, if n ≥ N for some N ). In this
case this is obviously true because the maximum of fn on [0, ∞) is equal to n42 e−2 , which decreases to
√
0 when n → ∞. Hence, given ε > 0, we set N > 2/(e ε) and then |fn (x) − 0| < ε for any n ≥ N .
This shows that {fn (x)} converges uniformly to f (x) = 0 on the interval [0, ∞).

We could show it also by the equivalent definition using sup as follows:

4 −2
lim sup |fn (x) − f (x)| = lim sup |fn (x)| = lim e = 0.
n→∞ x∈[0,∞) n→∞ x∈[0,∞) n→∞ n2

Problem 1.3 Prove that fn → f converges uniformly on D if and only if

lim sup |fn (x) − f (x)| = 0. (1)

n→∞ x∈D

First assume that (1) holds. We will show that {fn } converges to f according to Definition 9.9. Let
ε > 0. Then according to the definition of the limit in (1), we can find N , so that

sup |fn (x) − f (x)| < ε when n ≥ N.

x∈D

But the above inequality means (by the definition of supremum) that

|fn (x) − f (x)| < ε for all x ∈ D if n ≥ N.

This is exactly the statement in Definition 9.9.

On the contrary, assume that {fn } converges uniformly to f on D according to Definition 9.9. Let
ε > 0. Then there is N such that

|fn (x) − f (x)| < ε/2 for all x ∈ D and all n ≥ N.

This means that for every n ≥ N we have

sup |fn (x) − f (x)| ≤ ε/2 < ε.

x∈D

October 4, 2017 2 Karel Švadlenka

But this is the definition of the limit in identity (1) (i.e., for every ε > 0 we could find N so that etc. -
this is the definition of limit).

∑∞ k,
Problem 1.4 Verify that the geometric series k=0 x which converges pointwise on (−1, 1), does
not converge uniformly there.

The geometric series has the sum 1−x1

for |x| < 1, which shows that it converges pointwise on
(−1, 1). In Example 9.15 it is shown that the series converges uniformly on any interval [0, η], where
0 < η < 1. In the same way we could show that it converges uniformly on any interval [−η, η], where
0 < η < 1.

Let us show that it does not converge uniformly on (−1, 1). This means that the sequence of partial
sums
∑n
1 − xn+1
Sn (x) = xk =
1−x
k=0
1
does not converge uniformly to the function S(x) = 1−x . To prove it, we will use the equivalent
definition of uniform convergence from Problem 1.3. That is the series converges uniformly if and only
if
lim sup |Sn (x) − S(x)| = 0. (2)
n→∞ x∈(−1,1)

We compute the difference in absolute values:

1 − xn+1 1 |x|n+1
|Sn (x) − S(x)| = − = .
1−x 1−x 1−x

Checking the supremum on (−1, 1), we find

|x|n+1
sup |Sn (x) − S(x)| = sup = ∞,
x∈(−1,1) x∈(−1,1) 1 − x

since the function |x|n+1 /(1 − x) diverges to ∞ when x → 1−. Hence the limit in (2) cannot be 0 and
the series does not converge uniformly on (−1, 1).

The problem here is with the point x = 1 because there |Sn (x) − S(x)| diverges to ∞. One can see
from the above calculation that if we restrict ourselves to an interval [−η, η] with 0 < η < 1, then we
get uniform convergence because

|η|n+1
|Sn (x) − S(x)| ≤ →0 as n → ∞.
1−η

Problem 1.5 Let {fn } be a sequence of functions each of which is uniformly continuous on an open
interval (a, b). If fn → f uniformly on (a, b) can you conclude that f is also uniformly continuous on
(a, b)?

Yes. We can show that for any given ε > 0, we can find δ > 0, so that

|f (x1 ) − f (x2 )| < ε whenever |x1 − x2 | < δ and x1 , x2 ∈ (a, b)

in the following way.

October 4, 2017 3 Karel Švadlenka

 Let ε > 0 and from the definition of uniform convergence of {fn } find N such that
ε
|fn (x) − f (x)| < for every n ≥ N and all x ∈ (a, b).
3
Moreover, from the definition of uniform continuity of fN find δ such that
ε
|fN (x1 ) − fN (x2 )| < whenever |x1 − x2 | < δ and x1 , x2 ∈ (a, b).
3
Then, for any x1 , x2 ∈ (a, b) such that |x1 − x2 | < δ we have
ε ε ε
|f (x1 ) − f (x2 )| ≤ |f (x1 ) − fN (x1 )| + |fN (x1 ) − fN (x2 )| + |fN (x2 ) − f (x2 )| ≤ + + = ε.
3 3 3
This completes the proof of uniform continuity of f .

Problem 1.6 Prove that ∫ π

sin nx
lim dx = 0.
n→∞ π nx
2

We can use Theorem 9.26 on the sequence of functions

sin nx
fn (x) = , x ∈ [ π2 , π].
nx
To this end, we just need to check the assumptions of the theorem:
• lim fn (x) = f (x) for all x ∈ [ π2 , π].
n→∞
If we fix any x ∈ [ π2 , π], we compute

sin nx
f (x) = lim fn (x) = lim = 0.
n→∞ n→∞ nx

(This limit can be computed by the squeeze theorem.)

• Each function fn is continuous on [ π2 , π].

This is obvious, since sin is continuous and the denominator never becomes zero on the interval
[ π2 , π].

• The convergence of fn is uniform on [ π2 , π].

We use the equivalent definition from solved problem 1.3. First we estimate

sin nx 1 1
0 ≤ sup ≤ sup = π.
x∈[ π2 ,π] nx x∈[ π ,π] nx n2
2

Since we have sandwiched the supremum between 0 and nπ2

, which both tend to 0 as n → ∞, we
conclude that
sin nx
lim sup = 0,
n→∞ x∈[ π ,π] nx
2

which means that the convergence is uniform on [ π2 , π].

October 4, 2017 4 Karel Švadlenka

Since all the assumptions of the theorem are fulfilled, we conclude
∫ π ∫ π ∫ π
lim fn (x) dx = f (x) dx = 0 dx = 0,
n→∞ π π π
2 2 2

which implies ∫ π
sin nx
lim dx = 0.
n→∞ π nx
2

Note that it would be very difficult to calculate the integral without first interchanging the order of limit
and integration.

Problem 1.7 Show that if fn → f uniformly on [a, b] and each fn is continuous then the sequence of
functions ∫ x
Fn (x) = fn (t) dt
a
also converges uniformly on [a, b].

Let ε > 0. From the uniform convergence of fn to f , we can find N such that
ε
|fn (t) − f (t)| < for every t ∈ [a, b]. (3)
b−a

Define ∫ x
F (x) = f (t) dt.
a
This integral is well-defined because f is continuous on [a, b] (it is a uniform limit of a sequence of
continuous functions, so it must be continuous).

Then, if x is any point in [a, b] and n ≥ N , we have

∫ x ∫ x ∫ x ∫ x
|Fn (x) − F (x)| = fn (t) dt − f (t) dt = (fn (t) − f (t)) dt ≤ |fn (t) − f (t)| dt.
a a a a

In the last estimate we used the absolute property 8.7 of integral. Thus by (3),
∫ x ∫ x ∫ b
ε ε ε
|Fn (x) − F (x)| ≤ |fn (t) − f (t)| dt ≤ dt ≤ dt = (b − a) = ε.
a a b − a a b − a b − a
This shows, by definition, the uniform convergence of Fn to F on [a, b].

sin nx
Problem 1.8 Can the sequence of functions fn (x) = be differentiated term by term? How about
∑ n3
the series ∞ sin kx
k=1 k3 ?

For the sequence, we use Theorem 9.34, which requires us to check two main conditions:
• fn′ converges to uniformly to some function g
Notice that Theorem 9.34 requires a bounded interval [a, b] so we select any R > 0 and check the
uniform convergence on [−R, R]. We easily find that g(x) = 0 and so we calculate
cos nx 1 1
lim sup |fn′ (x) − g(x)| = lim sup ≤ lim sup = lim 2 = 0.
n→∞ x∈[−R,R] n→∞ x∈[−R,R] n2 n→∞ x∈[−R,R] n2 n→∞ n

October 4, 2017 5 Karel Švadlenka

 Thus fn′ converges uniformly to the zero constant function (we use the squeeze theorem to show
that the above limit is 0).

• fn converge pointwise to some function f

This is easy: pick any x ∈ [−R, R] and set f (x) = 0, then

| sin nx| 1
lim |fn (x) − f (x)| = lim 3
≤ lim 3 = 0.
n→∞ n→∞ n n→∞ n
Therefore, fn converge pointwise to the zero function f (x) = 0 on [−R, R].

Theorem 9.34 now implies that we can switch the order of limit and differentiation for the sequence
{fn (x)} on any bounded interval, which is the same as saying that the sequence can be differentiated
term by term.

∑
Now, let us look at the∑series ∞k=1 k3 . Note that even if the sequence {fn } converges uniformly,
sin kx

that does not imply that fk (x) converges∑ uniformly, since we are ”adding up” the functions fk (e.g.,
fn (x) = 1/n converges uniformly to 0 but 1/n = ∞).

To get the answer, we use Corollary 9.35, which again requires us to check two basic assumptions
(we check them again on any interval [−R, R]):
∑ ′
• fk converges uniformly to some function g
We use the M -test (Theorem 9.16). Setting Mk = 1/k 2 we find

| cos kx|
|fk′ (x)| = ≤ Mk .
k2
Moreover,
∞
∑ ∞
∑ 1
Mk = <∞
k2
k=1 k=1
∑
converges (p-harmonic series with p = 2). Hence, the M -test implies that fk′ converges uni-
formly on [−R, R].
∑
• fk converge pointwise to some function f
This is again easier than the first assumption: pick any x ∈ [−R, R], then
∞
∑ ∞
∑ ∞
∑
| sin kx| 1
fk (x) ≤ ≤ < ∞.
k3 k3
k=1 k=1 k=1
∑
Therefore, fk converges pointwise on [−R, R].

By Corollary 9.35, we can switch the order of summation and differentiation.

October 4, 2017 6 Karel Švadlenka

Problem 1.9 Compute the limit
∫ 1
e−nt
lim √ dt.
n→∞ 0 t

Since we want to apply Theorem 9.33, we first use a change of variables to transform the improper
integral of a discontinuous function
√ on a bounded interval to an integral of a continuous function on an
unbounded interval. Set x = 1/ t, then t = 1/x2 and dt = −2/x3 dx, and we get
∫ ∫ ∫ ∞
e−nt e−n/x −2 e−n/x
1 1 2 2

√ dt = 1 dx = 2 dx.
0 t ∞ x
x3 1 x2

We see that we are in the situation of Theorem 9.33 with a = 1 and

1 −n/x2
fn (x) = e , x ∈ [1, ∞).
x2

Now, to apply the theorem and change the order of limit and integration, we have to check the
following assumptions:
• fn converge uniformly to a function f on any [1, b], b > 1
We expect f (x) = 0, so let us compute
1 −n/x2
lim sup |fn (x) − f (x)| = lim sup e .
n→∞ x∈[1,b] n→∞ x∈[1,b] x2

The surest way to find the above supremum is to investigate the extrema of the functions fn (x).
We find that
−2 1 2 2n 2
fn′ (x) = 3 e−n/x + 2 e−n/x 3 = 3 e−n/x (−1 + xn2 ).
2 2

x x x x
√ √
The derivative is positive on [1, n) and negative on ( n, ∞), so the maximum is attained at
√ √ 1 √
x = n, where fn ( n) = en (if b is less than n then the maximum is even smaller than this
value). Hence,
1
lim sup |fn (x) − f (x)| = lim = 0,
n→∞ x∈[1,b] n→∞ en

and we conclude that the sequence converges uniformly on [1, b] to f (x) = 0 for any b > 1.

• there is an integrable dominating function g(x) for fn (x)

We have to find g so that |fn (x) ≤ g(x) for all x ∈ [1, ∞) and such that g is integrable on (1, ∞).
Note that g must be independent of n (that is, one function g has to dominate all the functions in
the sequence {fn }). Since e−n/x ≤ 1 for x ∈ [1, ∞), we can try
2

1
g(x) = .
x2
∫∞ 1
By direct computation we check that this function is integrable ( 1 x2
dx = 1).

All assumptions of Theorem 9.33 are satisfied and we conclude

∫ ∫ ∞ ∫ ∞ ∫ ∞
e−nt e−n/x e−n/x
1 2 2

lim √ dt = lim 2 dx = 2 lim dx = 2 0 dx = 0.

n→∞ 0 t n→∞ 1 x2 1 n→∞ x2 1

October 4, 2017 7 Karel Švadlenka

 Calculus B: Problems with Solutions (Lecture 2)

There are no solved problems for this lecture since they are included in the material for previous lecture
- see ”Problems with Solutions (Lecture 1)”.
 Calculus B: Problems with Solutions (Lecture 3)

Problem 3.1 Find the radius of convergence for the series

∞
∑
k!xk .
k=0

√
k
Since it is not so easy to calculate the limit of k!, we try to use the ratio test instead of the root test.
Setting ak = k!xk we compute

ak+1 (k + 1)!xk+1
lim = lim = lim (k + 1)|x| = ∞ for any x ̸= 0.
k→∞ ak k→∞ k!xk k→∞

Therefore, by the ratio test, the series converges absolutely only for x = 0. This gives us information
only on absolute convergence since the ratio test applies only to series with positive terms (we have
applied it to the series of absolute values). However, since we know that the interval of convergence for
the power series is an interval centered at x = 0, and we have just shown that the series diverges for all
positive x, we conclude that the radius of convergence is R = 0.

Another proof uses the trivial test noticing that

lim k!xk
k→∞

is either ∞ (when x > 0) or does not exist (when x < 0) (why?). In either case the trivial test says that
the series cannot converge and we immediately get R = 0.

Problem 3.2 If the limit

ak
lim
k→∞ ak+1
∑∞
exists or equals ∞, then show that it
∑∞ gives the radius of convergence of the power series k
k=0 ak x .
k
Give an example of a power series k=0 ak x for which the radius of convergence R satisfies

1
R= √
limk→∞ k
|ak |

but
ak
lim
k→∞ ak+1
does not exist.

We use the inequalities (proved later)

ak+1 √ √ ak+1
lim inf ≤ lim inf k |ak | ≤ lim sup k |ak | ≤ lim sup .
k→∞ ak k→∞ k→∞ k→∞ ak
If the limit
ak
lim
k→∞ ak+1
exists and is not equal to 0 (and also is not ∞), then the limit
ak+1 1
lim =
k→∞ ak limk→∞ ak
ak+1

also exists and is neither 0 nor ∞, which means that lim inf and lim sup of this sequence are equal (and
equal to the limit), so all the four terms
√in the above chain of inequalities must be equal. Hence, the
lim inf and lim sup of the sequence { k |ak |} are equal, which means that the limit of this sequence
exists and equals the lim inf and lim sup. We conclude
ak 1
lim = √ .
k→∞ ak+1 limk→∞ k |ak |
The case of zero or infinite limit are similar. For example, if the limit of {|ak |/|ak+1 |} is 0, then the
limit of {|ak+1 |/|ak |} is infinity and the lim inf is also ∞ and all the terms in the chain of inequalities
are ∞ (because they are defined).

To find the required counterexample, set

{
1 when k is odd
ak = ,
k when k is even

that is, {ak } = 1, 2, 1, 4, 1, 6, 1, 8, 1, 10, 1, . . . . Since

√ √ √
1 ≤ k ak ≤ k,
k k

we see by the squeeze theorem that

√
lim k
ak = 1.
k→∞
However, { }
ak 1 1 1
= , 2, , 4, , 6, . . . ,
ak+1 2 4 6
so the limit of this sequence does not exist.

It remains to prove the chain of inequalities. We prove only the last inequality
√ ak+1
lim sup k
|ak | ≤ lim sup , (1)
k→∞ k→∞ ak
the first inequality being analogous and the second one following from Theorem 2.46. Let β be any
a
number larger than lim supk→∞ k+1 ak . Then, by the definition of lim sup, there exists N such that

|ak+1 |
<β for every k ≥ N.
|ak |
This implies
√
√ |aN |
|aN +1 | < β|aN | ⇒ |aN +k | < β |aN |
k
⇒ N +k
|aN +k | < β N +k
.
βN

Keeping N fixed and taking k → ∞ in the last inequality, we get

√ √
lim sup k |ak | = lim sup N +k |aN +k | ≤ β.
k→∞ k→∞

October 18, 2017 2 Karel Švadlenka

 √
This means that lim supk→∞ k
|ak | is less than or equal to any number which is greater than
a
lim supk→∞ k+1
ak , hence the inequality (1) is proved.

∑∞
Problem 3.3 If the coefficients {ak } of a power series k=0 ak x
k form a bounded sequence, show that
the radius of convergence is at least 1.

If the radius of convergence R of the series is ∞, then the statement is true and there is nothing to
show, so let us assume that R is finite. Since {ak } is bounded, there exists M > 0 such that |ak | ≤ M
for all k = 0, 1, 2, . . . . This implies
√ √ √ √ 1 1
|ak | ≤ M ⇒ lim sup k |ak | ≤ lim sup M ⇒ √ ≥ √
k k k
.
k→∞ k→∞ lim supk→∞ k
|ak | lim supk→∞ k
M
√
But we know (see Example 2.33) that limk→∞ k M = 1 for any positive M , so the last inequality above
says that R ≥ 1, which is what we wanted to prove.

∑
Problem 3.4 Show that if ∞ k
k=0 ak x converges uniformly on an interval (−r, r), then it must in fact
converge uniformly on [−r, r].

Let ε > 0 be given. If the power series converges uniformly on (−r, r), then according to the Cauchy
criterion (Theorem 9.14), we see that there is N such that

∑
n
ε
ak xk < for any m, n ≥ N and any x ∈ (−r, r).
2
k=m

Fix any m, n ≥ N . Then the left-hand side of the above inequality is a continuous function in x. Thus,
taking the limit as x → r− and x → −r+, respectively, we deduce

∑
n
ε ∑
n
ε
ak r k ≤ < ε, ak (−r)k ≤ < ε.
2 2
k=m k=m

This means that

∑
n
ak xk < ε for any m, n ≥ N and any x ∈ [−r, r],
k=m

and this, in turn, means by the Cauchy criterion that the series converges uniformly on [−r, r].

October 18, 2017 3 Karel Švadlenka

 Calculus B: Problems with Solutions (Lecture 4)

Problem 4.1 Obtain power series expansion about x = 0 for

x
1 + x2
and find the largest interval where this expansion is valid.

Since
∑ ∞
1
= 1 − x2
+ x4
− x6
+ · · · = (−1)k x2k for x ∈ (−1, 1)
1 + x2
k=0
by the sum of geometric series, we formally have
∑ ∞
x
= x − x3
+ x5
− x7
+ · · · = (−1)k x2k+1
1 + x2
k=0

but we have to check for what values of x this formula is valid.

Since
∑ the coefficients series are either 0 or 1 or −1 (when we write the series in the
ak of this power√
form ∞ k=0 a k xk ), we have lim sup
k→∞
k
|ak | = 1 and get from the formula for radius of convergence
that
1
R= √ = 1,
lim supk→∞ k |ak |
so the formula is valid for x ∈ (−1, 1). It is easy to check that it is not valid for the endpoints x = 1 and
x = −1.

We can also use the arguments in Section 10.6, which say that the radius of convergence of the series
2 is at least the smaller value of the radius of convergence for x (R1 = ∞) and of the radius of
x
for 1+x
1
convergence of the series for 1+x 2 (R2 = 1).

Problem 4.2 Let

{
0, if x = 0
f (x) = −1/x2
e , if x ̸= 0
Prove that f is infinitely differentiable on the real line. Show that f (k) (0) = 0 for all k ∈ N. Explain
why the Taylor series for f about x = 0 does not represent f in any neighborhood of zero.

The function f is obviously infinitely differentiable on the interval (−∞, 0) and on the interval
(0, ∞) because the derivatives of f in these intervals
( )
′ 2 −1/x2 6 4
f (x) = − 4 + 6 e−1/x ,
′′ 2
f (x) = 3 e , etc.
x x x
are functions of the type ( )
∑
n
ak
e−1/x ,
2
f (m) (x) =
xk
k=1
which are continuous on any interval not including x = 0.

It remains to check the differentiability at x = 0. We show by induction that f (m) (0) = 0 for all
m ∈ N ∪ {0}. For m = 0 this holds by the definition of the function f . If f (m) (0) = 0, let us compute
f (m+1) (0): ( n )
f (m) (x) − f (m) (0) ∑ ak
e−1/x .
2
f (m+1) (0) = lim =
x→0 x−0 xk+1
k=1

If we show that limx→0 x1k e−1/x = 0 for any k = 1, 2, . . . , then the above limit will be zero and we are
2

done. But this limit can be calculated by squeeze theorem and L’Hospital’s rule from

1 −1/x2 yk
lim e = lim y2 = 0.
x→0 |x|k y→∞ e

Since all derivatives of f at x = 0 vanish, the Taylor series about the point x = 0 would be the zero
series. However, since the function is nonzero (positive) in any neighborhood of the origin x = 0, the
Taylor series cannot represent f on any neighborhood of x = 0.

Problem 4.3 Verify that the function cos x2 is analytic at x = 0, and write its Taylor series about x = 0.

We can use Theorem 10.33. in particular cos x2 is a composition of two functions f (x) = cos x and
g(x) = x2 , where g(x) is a polynomial, so it can be considered as a power series centered at x0 = 0 and
with its first term equal to C = 0:

g(x) = 0 + 0x + x2 + 0x3 + · · · .

If we expand f (x) about the point C = 0 as follows

∞
∑ (−1)k x2k
x2 x4 x6
cos x = 1 − + − + ··· =
2! 4! 6! (2k)!
k=0

(we know from the textbook that cos x is analytic on R and that the above expansion holds for every
x ∈ R), Theorem 10.33 asserts that the power series for the composition f (g(x)), obtained by inserting
the expansion of g into the expansion of f , has a positive radius of convergence R (actually, both f
and g have infinite radius of convergence, and so does f (g(x))). This means that the function cos x2 is
analytic in a neighborhood of the origin (since it can be represented by a power series in (−R, R)) and
in this neighborhood

∑ (−1)k x4k∞
(x2 )2 (x2 )4 (x2 )6 x4 x8 x12
cos x2 = 1 − + − + ··· = 1 − + − + ··· = .
2! 4! 6! 2! 4! 6! (2k)!
k=0

October 25, 2017 2 Karel Švadlenka

 Problem 4.4 Find first four terms of the series expansion of tan x about x = 0 using the following
three methods:
• long division (quotient of series),
• product of series,
• usual formula for Taylor series.

The power series for sin x and cos x are

∑ (−1)k x2k+1 ∞
x3 x5 x7
sin x = x − + − + ··· =
3! 5! 7! (2k + 1)!
k=0
∞
∑
x2 x4 x6 (−1)k x2k
cos x = 1 − + − + ··· = ,
2! 4! 6! (2k)!
k=0

where these expansions are valid for all x ∈ R. We briefly describe the three methods to obtain the
Taylor series for
tan x = c0 + c1 x + c2 x2 + c3 x3 + · · · .
Since we are required to find the first four terms, it is enough to calculate c0 , c1 , c2 , c3 .
• quotient of series (see Section 10.6.1):
Theorem 10.32 says that the quotient series for sin x/ cos x is valid in some neighborhood of zero.
We can find the coefficients by division or from the equation
( )
x2 x4 x3 x5
(c0 + c1 x + c2 x + c3 x + · · · ) 1 −
2 3
+ − ··· = x − + − ··· .
2! 4! 3! 5!
Multiplying out the left-hand side, we get
( c ) ( c ) x3
0 1
c0 + c1 x + − + c2 x2 + − + c3 x3 + · · · = x − + ··· .
2 2 3!
Comparing the coefficients we immediately see that c0 = c2 = 0, c1 = 1, c3 = 31 . Hence
1
tan x = x + x3 + · · · .
3
• product of series (see Section 10.6):
We take the function tan x as the product of functions sin x and cos1 x , write the power series for
these two functions and multiply them. Since both the functions are analytical in the neighborhood
of 0, the power series obtained by multiplication will also be valid in some neighborhood of zero.
The derivatives of cos1 x are
( )′ ( )′′ ( )′′′
1 sin x 1 1 + sin2 x 1 sin x(5 + sin2 x)
= , = , = .
cos x cos2 x cos x cos3 x cos x cos4 x
( )′ ( )′′ ( )′′′
Therefore, cos1 x |x=0 = 0, cos1 x |x=0 = 1, cos1 x |x=0 = 0, and in a neighborhood of
x = 0,
1 1
= 1 + x2 + 0x3 + · · · .
cos x 2
Multiplying this expansion with the expansion for sin x we have
( )( )
1 2 x3 1
tan x = 1 + x + · · · x− + · · · = x + x3 + · · · .
2 3! 3

October 25, 2017 3 Karel Švadlenka

 • usual formula for Taylor series:
The first three derivatives of tan x are
1 2 sin x 2(1 + 2 sin2 x)
(tan x)′ = , (tan x)′′ = , (tan x)′′′ = ,
cos2 x cos3 x cos4 x
so we obtain tan 0 = 0, tan′ (0) = 1, tan′′ (0) = 0, tan′′′ (0) = 2. Plugging these values into
the formula for Taylor series, we have

tan′′ (0) 2 tan′′′ (0) 3 1

tan x = tan 0 + tan′ (0)x + x + x + · · · = x + x3 + · · · .
2! 3! 3

October 25, 2017 4 Karel Švadlenka

 Calculus B: Problems with Solutions (Lecture 5)

Problem 5.1 Show that

∥x∥∞ = max(|x1 |, |x2 |, . . . , |xn |)
is a norm (i.e., that it has the properties listed in Theorem 11.6 in the textbook).

We check the properties (i)–(iii) one by one.

(i) ∥x∥∞ ≥ 0 and ∥x∥∞ = 0 if and only if x = 0
Since ∥x∥∞ is the maximum of a finite number of nonnegative numbers, it must be itself
nonnegative. Hence, ∥x∥∞ ≥ 0.
Now, if x = 0 then obviously ∥x∥∞ = max(0, . . . , 0) = 0. On the other hand, if ∥x∥∞ = 0
then the maximum of n nonnegative numbers |x1 |, . . . , |xn | being zero means that all the
numbers have to be zero. Therefore, x = 0.
(ii) ∥αx∥∞ = |α| ∥x∥∞ , α ∈ R
∥αx∥∞ = max(|αx1 |, |αx2 |, . . . , |αxn |) = max(|α||x1 |, |α||x2 |, . . . , |α||xn |)
= |α| max(|x1 |, |x2 |, . . . , |xn |) = |α| ∥x∥∞ .

(iii) ∥x + y∥∞ ≤ ∥x∥∞ + ∥y∥∞

∥x + y∥∞ = max(|x1 + y1 |, |x2 + y2 |, . . . , |xn + yn |)
≤ max(|x1 | + |y1 |, |x2 | + |y2 |, . . . , |xn | + |yn |)
≤ max(|x1 |, |x2 |, . . . , |xn |) + max(|y1 |, |y2 |, . . . , |yn |)
= ∥x∥∞ + ∥y∥∞ ,
where we used the triangle inequality for real numbers in the first inequality and the fact
that
max(a1 + b1 , a2 + b2 ) ≤ max(a1 , a2 ) + max(b1 , b2 )
(which can be proved by checking all the possible cases) in the second inequality.

Problem 5.2 Prove that a set E ⊂ Rn is closed if and only if it contains all its accumulation
points.

We have defined a closed set E as a set for which Rn \ E is an open set.

• Assume that E is closed and that there is an accumulation point x0 of E that does not
belong to E. Then x0 ∈ Rn \ E and also there exists a sequence of points {xk } ⊂ E such
that xk → x0 . Therefore, for any ε > 0 there is a point xK such that ∥xK − x0 ∥ < ε.
This means that in any ball B(x0 , ε) centered at x0 there is some point belonging to E.
However, this is a contradiction with Rn \ E being open (for x0 ∈ Rn \ E there must exist
a radius r > 0 so that B(x0 , r) ⊂ Rn \ E but we have proved above that each such ball
contains a point from E, that is, a point which is not in Rn \ E), which proves the first
implication.
 • To prove the converse implication assume that E contains all its accumulation points but
is not closed. This implies that F = Rn \ E is not open, so there is a point x0 ∈ F
such that for any n ∈ N there exists a point xn ∈ B(x0 , n1 ), which does not belong to F
(and thus belongs to E). Then the sequence {xn } is contained in E and converges to x0
(because ∥xn − x0 ∥ ≤ n1 ). Hence, x0 is an accumulation point of E but it does not belong
to E, which is a contradiction, and the second implication is proved.

Problem 5.3 Let {xk } be a sequence in Rn and let α ∈ R. If limk→∞ xk = x, show that then
lim (αxk ) = αx.
k→∞

Let ε > 0 and assume that α ̸= 0 (since otherwise there is nothing to prove). We use the
second property of the norm to find
∥(αxk ) − (αx)∥ = ∥α(xk − x)∥ = |α| ∥xk − x∥.
Since xk → x, for ε̃ = ε/|α| there exists K such that
∥xk − x∥ < ε̃ for any k ≥ K.
From the above estimate we then obtain
∥(αxk ) − (αx)∥ = |α| ∥xk − x∥ ≤ |α|ε̃ = ε for every k ≥ K,
which is what we had to prove.

Problem 5.4 Define the norm ∥x∥1 = |x1 | + · · · + |xn |. Show that for a sequence {xk } in Rn
the following holds:
lim xk = x if and only if ∥xk − x∥1 → 0 as k → ∞.
k→∞

This means that convergence does not depend on which of the norms ∥ · ∥ (euclidean norm) or
∥ · ∥1 we use.

We have
√
∥xk − x∥ = |xk1 − x1 |2 + · · · + |xkn − xn |2 ≤ |xk1 − x1 | + · · · + |xkn − xn |,
which can be rewritten as
0 ≤ ∥xk − x∥ ≤ ∥xk − x∥1 .
Hence, if ∥xk − x∥1 → 0 then by the squeeze theorem we conclude that ∥xk − x∥ → 0, which
is the definition of xk → x as k → ∞.

To prove the converse implication, we use Theorem 11.15, which says that if xk → x then
every coordinate converges:
xkj → xj as k → ∞, i.e., |xkj − xj | → 0 as k → ∞.
Therefore,
( )
lim ∥xk − x∥1 = lim |xk1 − x1 | + · · · + |xkn − xn | = lim |xk1 − x1 | + · · · + lim |xkn − xn | = 0,
k→∞ k→∞ k→∞ k→∞

which is the fact we wanted to show.

November 1, 2017 2 Karel Švadlenka

 Calculus B: Problems with Solutions (Lecture 6)

Problem 6.1 Suppose that E ⊂ Rn , f : E → Rm , g : E → Rm , and x0 is an accumulation

point of E. Show that if limx→x0 f (x) = Y and limx→x0 g(x) = Z then

lim (f (x) + g(x)) = Y + Z.

x→x0

Let ε > 0. According to the definition of limit, we have to show that there exists δ > 0 such
that
∥f (x) + g(x) − (Y + Z)∥ ≤ ε whenever x ∈ E and 0 < ∥x − x0 ∥ < δ.
From the assumption that limx→x0 f (x) = Y and limx→x0 g(x) = Z, we can find δ1 > 0 and
δ2 > 0 such that

∥f (x) − Y ∥ ≤ 21 ε whenever x ∈ E and 0 < ∥x − x0 ∥ < δ1

∥g(x) − Z∥ ≤ 1
2ε whenever x ∈ E and 0 < ∥x − x0 ∥ < δ2 .

Hence, if we take any δ > 0 smaller than or equal to min{δ1 , δ2 }, both the above relations hold
for x ∈ E such that 0 < ∥x − x0 ∥ < δ and for such x we have by triangle inequality

∥f (x) + g(x) − (Y + Z)∥ ≤ ∥f (x) − Y ∥ + ∥g(x) − Z∥ ≤ 12 ε + 12 ε = ε,

which is what we wanted to show.

Problem 6.2 Let f (r, θ) = (r cos θ, r sin θ). Is f continuous at the origin (0, 0)? Is f continuous
on all of R2 ?

To show that f is continuous at the origin, for any ε > 0 we have to find δ > 0 such that

∥f (r, θ) − f (0, 0)∥ ≤ ε whenever (r, θ) ∈ R2 and ∥(r, θ) − (0, 0)∥ < δ.

Since f (0, 0) = (0, 0), we calculate the above norms as

√
∥f (r, θ) − f (0, 0)∥ = ∥(r cos θ, r sin θ) − (0, 0)∥ = (r cos θ)2 + (r sin θ)2 = |r|

and √
∥(r, θ) − (0, 0)∥ = r2 + θ2 .

√
Now it is easy to see that |r| ≤ r2 + θ2 for any θ ∈ R, so taking δ = ε we have
√ √
∥f (r, θ) − f (0, 0)∥ = |r| ≤ r2 + θ2 < δ = ε whenever r2 + θ2 < δ.

This is the statement in the beginning of this proof.

 Next we check continuity at any other point (r0 , θ0 ) ∈ R2 . Since

∥f (r, θ) − f (r0 , θ0 )∥ = ∥(r cos θ, r sin θ) − (r0 cos θ0 , r0 sin θ0 )∥

√
= (r cos θ − r0 cos θ0 )2 + (r sin θ − r0 sin θ0 )2
√
= r2 + r02 − 2rr0 (cos θ cos θ0 + sin θ sin θ0 )
√
= (r − r0 )2 + 2rr0 (1 − cos(θ − θ0 ))
√
θ − θ0
= (r − r0 )2 + 4rr0 sin2 ,
2
and √
∥(r, θ) − (r0 , θ0 )∥ = (r2 − r0 )2 + (θ − θ0 )2 ,
√ √
we need to bound the expression (r − r0 )2 + 4rr0 sin2 θ−θ 2
0
by (r2 − r0 )2 + (θ − θ0 )2 , when
the latter expression is small enough.
√
If we consider only δ’s that are less than or equal to 1, then from (r2 − r0 )2 + (θ − θ0 )2 <
δ ≤ 1 we can deduce √
|r − r0 | ≤ (r2 − r0 )2 + (θ − θ0 )2 < δ ≤ 1,
so that |r| < |r0 | + 1.

Moreover, for δ < 1 we show in the same way that |θ − θ0 | < δ ≤ 1 and thus

θ − θ0 1
sin ≤ |θ − θ0 |.
2 2

Putting all together, we have

√
θ − θ0
∥f (r, θ) − f (r0 , θ0 )∥ = (r − r0 )2 + 4rr0 sin2
2
√
√ θ − θ0
≤ (r − r0 )2 + 4|rr0 | sin2
2
√
≤ |r − r0 | + 4(|r0 | + 1)|r0 |( 12 (θ − θ0 ))2
√
= |r − r0 | + (|r0 | + 1)|r0 ||θ − θ0 |
√
≤ δ + (|r0 | + 1)|r0 |δ.

Therefore, taking { }
ε
δ = min 1, √ ,
1+ (|r0 | + 1)|r0 |
we have shown the continuity at (r0 , θ0 ).

Problem 6.3 Prove that if f : E → Rm is uniformly continuous on E ⊂ Rn , and E is star-

shaped and bounded, then f is bounded on E.

Since E is bounded, there exists M > 0 so that E ⊂ B(0, M ) (open ball of radius M centered
at the origin 0). Since f is uniformly continuous on E, there is δ > 0 such that

∥f (x) − f (y)∥ < 1 for any x, y ∈ E such that ∥x − y∥ < δ.

November 8, 2017 2 Karel Švadlenka

 The idea is that, since E is star-shaped, we can reach from the origin to any point in E along
a straight line in at most 2M/δ steps of length δ/2. Precisely said, the equation of a straight
line joining the origin and a point y ∈ E is {ty, t ∈ R}. Setting the points x0 , . . . , xN (where
N < 2M/δ + 1) as
x0 = 0, x1 = 2δ y, . . . , xn = n 2δ y, . . . , xN = y,
we see that
∥xn − xn+1 ∥ ≤ δ
2 < δ,
so that by triangle inequality and the above assumptions
∥f (y)∥ ≤ ∥f (0)∥ + ∥f (x1 ) − f (0)∥ + · · · + ∥f (y) − f (xN −1 )∥ ≤ ∥f (0)∥ + N
(notice that each difference term ∥f (xn )−f (xn−1 )∥ above is less than 1 since ∥xn −xn+1 ∥ < δ.)

Hence, we have shown that the value of f at an arbitrary point of E is less than
2M
∥f (0)∥ + N < ∥f (0)∥ + + 1,
δ
which is a fixed finite number.

(Compare this proof to the proof of Theorem 5.47.)

Problem 6.4 Decide if the limit

sin2 x sin2 y
lim
(x,y)→(0,0) x4 + 3y 4

exists or not and calculate its value if it exists.

Since sin x behaves like x when x is close to zero, we expect that the function will be close to
x2 y 2
x4 +3y 4
,
when x and y are close to 0. For this type of function we already know that the limits
when approaching by different paths to the origin may be different. In fact, approaching (0, 0)
along the x-axis or y-axis, the limit is zero because the function is identically equal to zero on
both axes (remember that we do not care about the value at (0, 0) itself). On the other hand,
approaching along the line y = x, we get
x2 y 2 t4 1
lim 4 4
= lim 4
= .
(x=t,y=t)→(0,0) x + 3y t→0 4t 4
Hence, we suspect that the limit does not exist.

To prove it precisely, we just show that approaching (0, 0) in two different ways yields two
different limits:
0 sin2 y
lim f (0, y) = lim = lim 0 = 0
(0,y)→(0,0) y→0 3y 4 y→0
( )
4
sin t 1 sin t 4 1
lim f (t, t) = lim = lim =
(t,t)→(0,0) t→0 4t4 t→0 4 t 4

Therefore, approaching along the y-axis gives the limit 0, while approaching along the straight
line y = x yields the limit 0.25. This means that the function takes values that are close to both
0 and 0.25 in any (however small) neighborhood of the origin, implying that the limit cannot
exist.

November 8, 2017 3 Karel Švadlenka

 Calculus B: Problems with Solutions (Lecture 7)

Problem 7.1 Find the equation for the tangent plane to the paraboloid z = f (x, y) = x2 + 14 y 2
at the point (x, y) = (1, 2). Moreover, find the rate of change of the function f at the same
point in the direction (3, 4).

We know that if f is differentiable at (a1 , a2 ) then the tangent plane is defined by

{(x, y, z) ∈ R3 | z = f (a1 , a2 ) + fx (a1 , a2 )(x − a1 ) + fy (a1 , a2 )(y − a2 )}.
Note that we do not have to remember the above formula because it is a plane passing through
the point (a1 , a2 , f (a1 , a2 )), so its equation must have the form z = f (a1 , a2 )+b(x−a1 )+c(y−a2 )
(if (x, y) = (a1 , a2 ), we get z = f (a1 , a2 )). It only remains to find the coefficients b, c – these
are however the rates of change of the plane at the point (a1 , a2 ) in the x and y directions,
respectively, which should match the rates of changes of the function f at the point (a1 , a2 ) in
those directions, which are the partial derivatives of f at the point (a1 , a2 ) with respect to x
and y, respectively.

Here we have (a1 , a2 ) = (1, 2) and

(1 )
fx (1, 2) = (2x) (x,y)=(1,2)
= 2, fy (1, 2) = 2y (x,y)=(1,2)
= 1,

thus the equation for the tangent plane to f at (1, 2) is

z = 2 + 2(x − 1) + 1(y − 2),
or
2x + y − z − 2 = 0.

Next, the unit vector u with same direction as the vector (3, 4) is u = ( 35 , 45 ). Therefore, the
rate of change of f at (1, 2) in the direction (3, 4) is given by
f (1 + 53 h, 2 + 45 h) − f (1, 2) 13 2
2h + 25 h
Du f (1, 2) = lim = lim = 2.
h→0 h h→0 h

Problem 7.2 Compute the mixed partial derivatives fxy and fyx of the function
{ 2
x arctan xy − y 2 arctan xy ̸ 0 and y ̸= 0
if x =
f (x, y) = .
0 otherwise

1. Derivative at points (x, y) such that x ̸= 0 and y ̸= 0:

By definition of f and chain rule, we have
x3 x xy 2
fy = − 2y arctan +
x2 + y 2 y x2 + y 2
2
∂ f x −y
2 2
fyx = = 2 .
∂x∂y x + y2
 Similarly we have
y3 y x2 y
fx = − + 2x arctan −
x2 + y 2 x x2 + y 2
∂2f x2 − y 2
fxy = = 2 .
∂y∂x x + y2
2. Derivative at points (x, 0) such that x ̸= 0:

f (x, h) − f (x, 0)
fy (x, 0) = lim
h→0 h
1 2 h x
= lim (x arctan − h2 arctan )
h→0 h x h
x2 h x
= lim arctan − lim h arctan .
h→0 h x h→0 h
Here the first limit is of the 0/0 type, so we use L’Hospital’s rule to get
x
x2 arctan hx x2 h2 +x 2
lim = lim = x.
h→0 h h→0 1
The second limit becomes zero because arctan is bounded by π/2 and h → 0. Hence, we
obtain
fy (x, 0) = x.
Also,
f (x + h, 0) − f (x, 0) 0−0
fx (x, 0) = lim = lim = 0.
h→0 h h→0 h
So we get
fy (x + h, 0) − fy (x, 0)
fyx (x, 0) = lim
h→0 h
(x + h) − x
= lim = 1,
h→0 h
fx (x, h) − fx (x, 0)
fxy (x, 0) = lim
h→0 h
3
x2 h
− x2y+h2 + 2x arctan hx − x2 +h2
= lim
h→0 h
h2 2x h x2
= lim (− 2 + arctan − )
h→0 x + h2 h x x 2 + h2
= 2 − 1 = 1.

3. Derivative at points (0, y) such that y ̸= 0:

By similar calculation as in the previous case, we have
fx (0, y) = −y,
thus
fxy (0, y) = −1 = fyx (0, y).

4. Derivative at the point (0, 0):

From the above calculations we deduce,
fy (h, 0) − fy (0, 0) h
fyx (0, 0) = lim = = 1,
h→0 h h
fx (0, h) − fx (0, 0) −h
fxy (0, 0) = lim = = −1.
h→0 h h

November 15, 2017 2 Karel Švadlenka

Problem 7.3 Show that the function
{
x2 y
x2 +y 2
if (x, y) ̸= 0
f (x, y) =
0 if (x, y) = (0, 0)

is continuous and has first-order partial derivatives on R2 but is not differentiable at (0, 0).

First we check the continuity of f . It is clear that f is continuous at (x, y) ̸= (0, 0), so we
show the continuity of f at (0, 0). To see this, we use the polar √ coordinate system (x, y) =
√ that (x, y) → (0, 0) means ∥(x, y)∥ = x + y → 0 and in the polar
(r cos θ, r sin θ). Notice 2 2

system this becomes r2 cos2 θ + r2 sin θ = r → 0. Then we have

2

|f (x, y) − f (0, 0)| = |r(cos θ)2 sin θ − 0| ≤ r → 0 as r → 0.

This implies that f is continuous at (0, 0). (We can show it also directly by estimating

x2 |y| (x2 + y 2 )|y| √

|f (x, y) − f (0, 0)| = ≤ = |y| ≤ x2 + y 2 → 0 as (x, y) → (0, 0).)
x2 + y 2 x2 + y 2

Next we show that f has first-order partial derivatives on R2 . First,

2xy 3 x2 (x2 − y 2 )
fx (x, y) = fy (x, y) =
(x2 + y 2 )2 (x2 + y 2 )2

for (x, y) ̸= (0, 0), and

f (h, 0) − f (0, 0)
fx (0, 0) = lim =0
h→0 h
f (0, h) − f (0, 0)
fy (0, 0) = lim = 0.
h→0 h
Therefore f has first-order partial derivatives on R2 .

Finally we show that f is not differentiable at (0, 0). We prove this by contradiction. If f
was differentiable at (0, 0), we would get using Theorem 12.18 that

Du f (0, 0) = fx (0, 0)u1 + fy (0, 0)u2 = 0

for any unit vector u = (u1 , u2 ). However, when u = √1 (1, 1), we have
2

f (0 + hu1 , 0 + hu2 ) − f (0, 0) u2 u2 1

Du f (0, 0) = lim = 2 1 2 = √ ̸= 0,
h→0 h u1 + u2 2 2
thus we obtain a contradiction.

November 15, 2017 3 Karel Švadlenka

 Calculus B: Problems with Solutions (Lecture 8)

Problem 8.1 Find the direction in which the function f (x, y, z) = x2 + xz + y 3 decreases the
fastest at the point (1, 3, 2).

Let u = (u1 , u2 , u3 ) be any unit vector. Since f is differentiable at (1, 3, 2), by Theorem
12.18 we have
Du f (1, 3, 2) = (∇f (1, 3, 2)) · u
Here

∇f (1, 3, 2) = (fx , fy , fz )(1, 3, 2) = (2x + z, 3y 2 , x)|(x,y,z)=(1,3,2) = (4, 27, 1) ̸= (0, 0, 0).

By the law of cosines we can write

Du f (1, 3, 2) = (∇f (1, 3, 2)) · u = ∥∇f (1, 3, 2)∥ ∥u∥ cos θ = ∥∇f (1, 3, 2)∥ cos θ,

where θ is the angle between the vectors ∇f (1, 3, 2) and u. Thus the minimum rate of change
of f is the direction corresponding to θ = π (where cos θ achieves its minimum). This occurs
when u and −∇f (1, 3, 2) have the same direction. Therefore the direction in which f decreases
the fastest at (1, 3, 2) is (−4, −27, −1).
To be precise, we should√ give a unit vector
√ for the direction, which is the vector (−4, −27, −1)
divided by its length 42 + 272 + 12 = 746, i.e., the vector √746 1
(−4, −27, −1).

Problem 8.2 Verify the following formula for the differential of a ratio of two functions f and
g: ( )
f g df − f dg
d = (g ̸= 0).
g g2

By definition of the differential, we have

( ) ( ) ( )
f ∂ f ∂ f
d = dx1 + · · · + dxn
g ∂x1 g ∂xn g
fx g − f gx1 fx g − f gxn
= 1 2 dx1 + · · · + n 2 dxn
g g
g (fx1 dx1 + · · · + fxn dxn ) − f (gx1 dx1 + · · · + gxn dxn )
=
g2
g df − f dg
= .
g2

Problem 8.3 Give the chain rule for ∂G/∂s, ∂G/∂t and ∂G/∂u, where G : R3 → R is defined
by
G(s, t, u) = F (s, t, f (s, t), g(s, u), h(s, t, u)),
with F : R5 → R, f, g : R2 → R and h : R3 → R differentiable functions.

Let us write
G(s, t, u) = F (x1 , x2 , x3 , x4 , x5 ) = F (x),
where each xi is taken as a function of s, t, u according to

x1 (s, t, u) = s, x2 (s, t, u) = t, x3 (s, t, u) = f (s, t), x4 (s, t, u) = g(s, u), x5 (s, t, u) = h(s, t, u).

Then we can mechanically apply the chain rule as follows:

∂G ∂F ∂x1 ∂F ∂x2 ∂F ∂x3
(s, t, u) = (x) (s, t, u) + (x) (s, t, u) + (x) (s, t, u)
∂s ∂x1 ∂s ∂x2 ∂s ∂x3 ∂s
∂F ∂x4 ∂F ∂x5
+ (x) (s, t, u) + (x) (s, t, u)
∂x4 ∂s ∂x5 ∂s
∂F ∂F ∂F ∂f
= (x) · 1 + (x) · 0 + (x) (s, t)
∂x1 ∂x2 ∂x3 ∂s
∂F ∂g ∂F ∂h
+ (x) (s, u) + (x) (s, t, u)
∂x4 ∂s ∂x5 ∂s
∂F ∂F ∂f ∂F ∂g ∂F ∂h
= (x) + (x) (s, t) + (x) (s, u) + (x) (s, t, u).
∂x1 ∂x3 ∂s ∂x4 ∂s ∂x5 ∂s
Hence we obtain
∂G ∂F ∂F ∂f
(s, t, u) = (s, t, f (s, t), g(s, u), h(s, t, u)) + (s, t, f (s, t), g(s, u), h(s, t, u)) (s, t)
∂s ∂x1 ∂x3 ∂s
∂F ∂g
+ (s, t, f (s, t), g(s, u), h(s, t, u)) (s, u)
∂x4 ∂s
∂F ∂h
+ (s, t, f (s, t), g(s, u), h(s, t, u)) (s, t, u),
∂x5 ∂s
which is often written in a short form as
∂G ∂F ∂F ∂f ∂F ∂g ∂F ∂h
= + + + .
∂s ∂s ∂x3 ∂s ∂x4 ∂s ∂x5 ∂s

November 22, 2017 2 Karel Švadlenka

Similarly for the other partial derivatives
∂G ∂F ∂x1 ∂F ∂x2 ∂F ∂x3
(s, t, u) = (x) (s, t, u) + (x) (s, t, u) + (x) (s, t, u)
∂t ∂x1 ∂t ∂x2 ∂t ∂x3 ∂t
∂F ∂x4 ∂F ∂x5
+ (x) (s, t, u) + (x) (s, t, u)
∂x4 ∂t ∂x5 ∂t
∂F ∂F ∂F ∂f
= (x) · 0 + (x) · 1 + (x) (s, t)
∂x1 ∂x2 ∂x3 ∂t
∂F ∂F ∂h
+ (x) · 0 + (x) (s, t, u)
∂x4 ∂x5 ∂t
∂F ∂F ∂f ∂F ∂h
= (x) + (x) (s, t) + (x) (s, t, u),
∂t ∂x3 ∂t ∂x5 ∂t
∂G ∂F ∂x1 ∂F ∂x2 ∂F ∂x3
(s, t, u) = (x) (s, t, u) + (x) (s, t, u) + (x) (s, t, u)
∂u ∂x1 ∂u ∂x2 ∂u ∂x3 ∂u
∂F ∂x4 ∂F ∂x5
+ (x) (s, t, u) + (x) (s, t, u)
∂x4 ∂u ∂x5 ∂u
∂F ∂F ∂F
= (x) · 0 + (x) · 0 + (x) · 0
∂x1 ∂x2 ∂x3
∂F ∂g ∂F ∂h
+ (x) (s, u) + (x) (s, t, u)
∂x4 ∂u ∂x5 ∂u
∂F ∂g ∂F ∂h
= (x) (s, u) + (x) (s, t, u).
∂x4 ∂u ∂x5 ∂u

Problem 8.4 Express the wave equation

∂2f 2
2∂ f
= c
∂x2 ∂y 2

for c > 0 and f (x, y) : R2 → R in the new coordinates ξ = y + cx, η = y − cx.

We define the function g of new coordinate variables that corresponds to f in the following
way: ( )
ξ−η ξ+η
g(ξ, η) = f (x, y) = f , ,
2c 2
or, from another viewpoint, by

f (x, y) = g(ξ, η) = g(y + cx, y − cx). (1)

ξ−η ξ+η
(Here, the expressions 2c , 2 were obtained by solving the linear system of two equations

ξ = y + cx (2)
η = y − cx

for x and y.)

To transform the wave equation into the new coordinate system means to rewrite the equation
fxx = c2 fyy in terms of g and the variables ξ, η. To do so we use the relation (1) and the relation
between the old and new variables (2) in order to compute the partial derivatives of f in terms

November 22, 2017 3 Karel Švadlenka

of partial derivatives of g. Applying the chain rule we have
∂f ∂g ∂ξ ∂g ∂η
(x, y) = (ξ, η) (x, y) + (ξ, η) (x, y)
∂x ∂ξ ∂x ∂η ∂x
∂g ∂g
= (ξ, η) · c + (ξ, η) · (−c)
∂ξ ∂η
[ ]
∂g ∂g
= c (ξ, η) − (ξ, η) .
∂ξ ∂η

Hence, we can compute the second order partial with respect to x as

( )
∂2f ∂ ∂f
(x, y) = (x, y)
∂x2 ∂x ∂x
( [ ])
∂ ∂g ∂g
= c (ξ, η) − (ξ, η)
∂x ∂ξ ∂η
[ ]
∂ ∂g ∂ ∂g
= c (ξ, η) − (ξ, η)
∂x ∂ξ ∂x ∂η
[( 2 )
∂ g ∂ξ ∂2g ∂η
= c (ξ, η) (x, y) + (ξ, η) (x, y)
∂ξ 2 ∂x ∂η∂ξ ∂x
( 2 )]
∂ g ∂ξ ∂2g ∂η
− (ξ, η) (x, y) + 2 (ξ, η) (x, y)
∂ξ∂η ∂x ∂η ∂x
[( 2 2
) ( 2 )]
∂ g ∂ g ∂ g ∂2g
= c (ξ, η) · c + (ξ, η) · (−c) − (ξ, η) · c + 2 (ξ, η) · (−c)
∂ξ 2 ∂η∂ξ ∂ξ∂η ∂η
[ 2 2 2
]
∂ g ∂ g ∂ g
= c2 2
(ξ, η) − 2 (ξ, η) + 2 (ξ, η) ,
∂ξ ∂η∂ξ ∂η

where we have assumed that g is twice continuously differentiable, so that the cross partials
∂ 2 g/∂η∂ξ and ∂ 2 g/∂ξ∂η are equal.

For the partial derivative with respect to y we calculate in an analogous way

∂f ∂g ∂g
(x, y) = (ξ, η) + (ξ, η),
∂y ∂ξ ∂η
∂2f ∂2g ∂2g ∂2g
(x, y) = (ξ, η) + 2 (ξ, η) + (ξ, η).
∂y 2 ∂ξ 2 ∂η∂ξ ∂η 2

Substituting these results for partial derivatives into the wave equation

∂2f ∂2f
2
(x, y) = c2 2 (x, y),
∂x ∂y
we obtain
[ 2 ] [ 2 ]
∂ g ∂2g ∂2g 2 ∂ g ∂2g ∂2g
c2 (ξ, η) − 2 (ξ, η) + (ξ, η) = c (ξ, η) + 2 (ξ, η) + (ξ, η) .
∂ξ 2 ∂η∂ξ ∂η 2 ∂ξ 2 ∂η∂ξ ∂η 2

Since most of the terms cancel, we finally arrive (after dividing by −4c2 ) at the relation

∂2g
(ξ, η) = 0,
∂η∂ξ
which is the wave equation in new coordinates.

November 22, 2017 4 Karel Švadlenka

 Calculus B: Problems with Solutions (Lecture 9)

Problem 9.1 Show that the equation

x2 y 2 + 2exy − 4 − 2e2 = 0

can be solved for y in terms of x in a neighborhood of the point x = 1 with y(1) = 2. Calculate
dy
dx at x = 1 in two ways: by applying Theorem 12.40 and via implicit differentiation of the
equation.

If we write
F (x, y) = x2 y 2 + 2exy − 4 − 2e2 ,
then in view of Fy (x, y) = 2x2 y + 2xexy we have

F (1, 2) = 0, Fy (1, 2) = 2 · 2 + 2 · e2 = 4 + 2e2 ̸= 0.

Then by Theorem 12.40, F (x, y) = 0 can be solved for y in terms of x in a neighborhood of the
point x = 1 with y(1) = 2.
dy
Next we calculate dx at x = 1.
1. applying Theorem 12.40
Using the formula of Theorem 12.40, we obtain

dy Fx (1, y(1)) 8 + 4e2

(1) = − =− = −2.
dx Fy (1, y(1)) 4 + 2e2

2. implicit differentiation of the equation

On the other hand, in the neighborhood of the point x = 1,
d dy
0 = F (x, y(x)) = Fx (x, y(x)) + Fy (x, y(x)) (x)
dx dx
[ ] [ ] dy
2 xy(x)
0 = 2xy (x) + 2y(x)e + 2x y(x) + 2xexy(x)
2
(x)
dx
holds (because F (x, y(x)) = 0 for all x in the neighborhood), so that inserting x = 1 we
have
dy
0 = (8 + 4e2 ) + (4 + 2e2 ) (1),
dx
dy 8 + 4e2
(1) = − = −2.
dx 4 + 2e2
 Problem 9.2 Suppose that F (x, y, z) = 0 defines x as a function of y and z, and also y as a
function of x and z and moreover z as a function of x and y. Show that under appropriate
conditions on the function F ,
∂y ∂z ∂x
= −1.
∂x ∂y ∂z

If F has continuous first-order partial derivatives and

Fx (x0 , y0 , z0 ), Fy (x0 , y0 , z0 ), Fz (x0 , y0 , z0 ) ̸= 0

holds at some point (x0 , y0 , z0 ), then by Theorem 12.44 we can express x as a function of y and
z, and also y as a function of x and z and moreover z as a function of x and y in a neighborhood
of the point (x0 , y0 , z0 ). Moreover, the theorem states that

∂y Fx (x0 , y(x0 , z0 ), z0 )
(x0 , z0 ) = − ,
∂x Fy (x0 , y(x0 , z0 ), z0 )
∂z Fy (x0 , y0 , z(x0 , y0 ))
(x0 , y0 ) = − ,
∂y Fz (x0 , y0 , z(x0 , y0 ))
∂x Fz (x(y0 , z0 ), y0 , z0 )
(y0 , z0 ) = − .
∂z Fx (x(y0 , z0 ), y0 , z0 )

Since x(y0 , z0 ) = x0 , y(x0 , z0 ) = y0 and z(x0 , y0 ) = z0 , these identities imply

( )( )( )
∂y ∂z ∂x Fx Fy Fz
(x0 , y0 , z0 ) = − − − = −1.
∂x ∂y ∂z Fy Fz Fx

(In the equation above, Fx , Fy , Fz are all evaluated at the point (x0 , y0 , z0 ).)

Notice that we cannot treat the symbols ∂x, ∂y, ∂z as some quantities because upon canceling
them on the left-hand side of the above equation, we would get +1 instead of the correct result
−1.

Problem 9.3 Decide whether the following system

x2 − 2y 2 − 2uy − v 3 − 12 = 0
xy + 2y 2 − 3u2 + x2 v + 7 = 0

can be solved for u, v as functions of x, y in the neighborhood of the point p0 = (x0 , y0 , u0 , v0 ) =

(2, 1, −1, −2). Moreover find the values of the derivatives ∂u/∂x, ∂v/∂x at the point (x0 , y0 ).

According to Theorem 12.48, for the system to be solvable for u, v in a neighborhood of the
point p0 , it is sufficient that the condition

∂(F 1 , F 2 )
(x0 , y0 , u0 , v0 ) ̸= 0
∂(u, v)

is satisfied, where

F 1 (x, y, u, v) = x2 − 2y 2 − 2uy − v 3 − 12
F 2 (x, y, u, v) = xy + 2y 2 − 3u2 + x2 v + 7.

November 29, 2017 2 Karel Švadlenka

We compute

∂(F 1 , F 2 ) Fu1 Fv1

(x0 , y0 , u0 , v0 ) =
∂(u, v) Fu2 Fv2 (x,y,u,v)=(x0 ,y0 ,u0 ,v0 )
−2y0 −3v02
=
−6u20 x20
= −2x20 y0 − 18u20 v02 = −80,

so that the condition is fulfilled.

To find the partial derivatives, we differentiate the given system with respect to x, remem-
bering that u, v are functions of x and y:

∂F 1 ∂u ∂v
(x0 , y0 , u0 , v0 ) = 2x0 − 2y0 (x0 , y0 ) − 3v 2 (x0 , y0 ) (x0 , y0 ) = 0
∂x ∂x ∂x
∂F 2 ∂u ∂v
(x0 , y0 , u0 , v0 ) = y0 − 6u(x0 , v0 ) (x0 , y0 ) + 2x0 v(x0 , y0 ) + x20 (x0 , y0 ) = 0.
∂x ∂x ∂x
This yields the following system of linear equations for ∂u/∂x, ∂v/∂x at (x0 , y0 ):

∂u ∂v
2y0 (x0 , y0 ) + 3v02 (x0 , y0 ) = 2x0
∂x ∂x
∂u ∂v
6u0 (x0 , y0 ) − x20 (x0 , y0 ) = y0 + 2x0 v0 .
∂x ∂x
Solving by usual methods we find

∂u 2x30 + 3y0 v02 + 6x0 v03 17

(x0 , y0 ) = 2 2 =
∂x 2x0 y0 + 18u0 v0 16
∂v 6x0 u0 − y0 − 2x0 y0 v0
2 21
(x0 , y0 ) = 2 2 = .
∂x x0 y0 + 9u0 v0 32

Problem 9.4 The formulas relating spherical to rectangular coordinates read

x = r cos θ sin ϕ
y = r sin θ sin ϕ
z = r cos ϕ.

Can these equations be solved for r, θ, ϕ in terms of x, y, z?

The Jacobian determinant is

cos θ sin ϕ −r sin θ sin ϕ r cos θ cos ϕ
∂(x, y, z)
= sin θ sin ϕ r cos θ sin ϕ r sin θ cos ϕ = −r2 sin ϕ.
∂(r, θ, ϕ)
cos ϕ 0 −r sin ϕ

For any point (r0 , θ0 , ϕ0 ) ∈ R3 such that r0 ̸= 0 and ϕ0 ̸= nπ for any n ∈ Z, the Jacobian
determinant will not be equal to zero and thus there will be a neighborhood of (r0 , θ0 , ϕ0 ) on
which the equations can be solved for r, θ, ϕ in terms of x, y, z. Notice that this neighborhood
cannot be taken too large since the above mapping may not be one-to-one on a larger domain
due to the periodicity of the trigonometric functions.

November 29, 2017 3 Karel Švadlenka

 On the other hand, when r = 0 (that is, x = y = z = 0), we cannot determine the angles
(θ, ϕ) from the equations. Also, when sin ϕ = 0 (that is, x = y = 0), we cannot determine
the angle θ from the equations. Note, however, that the fact that the Jacobian is zero does
not imply that the equations cannot be solved in a neighborhood, since the implicit function
theorem gives only a sufficient condition.

November 29, 2017 4 Karel Švadlenka

 Calculus B: Problems with Solutions (Lecture 10)

Problem 10.1 Write the second order Taylor’s formula for the function

1
f (x, y) =
xy 2

at the point (x0 , y0 ) = (2, 1).

We see that the function has continuous partial derivatives of any order in any neighborhood
of the point (2, 1) which does not contain the lines x = 0 or y = 0. For example, we can
take as the neighborhood the open ball B((2, 1), 1) of radius 1 centered at (2, 1). Hence, the
differentiability assumptions of Taylor’s theorem are fulfilled and we can write

f (x0 + h, y0 + k) = f (x0 , y0 ) + fx (x0 , y0 )h + fy (x0 , y0 )k (1)

1
[ 2 2
]
+ 2 fxx (x0 , y0 )h + 2fxy (x0 , y0 )hk + fyy (x0 , y0 )k + R,

where R is the remainder.

We calculate the partial derivatives as follows:

−1
fx (x, y) =
x2 y 2
−2
fy (x, y) =
xy 3
2
fxx (x, y) =
x3 y 2
2
fxy (x, y) =
x2 y 3
6
fyy (x, y) = .
xy 4
Therefore,
1 −1 1 1
f (2, 1) = , fx (2, 1) = , fy (2, 1) = −1, fxx (2, 1) = , fxy (2, 1) = , fyy (2, 1) = 3,
2 4 4 2
which yields upon substituting into (1),
1 1 1 1 3
f (2 + h, 1 + k) = − h − k + h2 + hk + k 2 + R(h, k).
2 4 8 2 2
 1.5

1.4

1.3

1.2

1.1

0.9

0.8

0.7

0.6

0.5
1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3 2.4 2.5

Contour lines of the original function f (color) and its second order Taylor approximation (black).

Problem 10.2 Find all critical points of the function

f (x, y) = x sin(x + y)

and determine whether they are local minimum points, local maximum points or saddle points.

The function is infinitely differentiable on the whole R2 , so there are only critical points for
which ∇f = 0. Consequently, it is sufficient to compute the partial derivatives fx , fy , and check
when both of them are zero:

fx (x, y) = sin(x + y) + x cos(x + y), fy (x, y) = x cos(x + y).

If both partials have to be zero, we get the conditions

x cos(x + y) = 0 sin(x + y) = 0.

The second condition implies that x + y = kπ, where k ∈ N. But for such x, y it holds that
cos(x + y) ̸= 0, and we see that x must be equal to 0. We conclude that all critical points of f
are given by
(x, y) = (0, kπ), k ∈ N.

To find whether these points yield local extrema or are saddle points, we compute the Hessian
( ) ( )
fxx fxy 2 cos(x + y) − x sin(x + y) cos(x + y) − x sin(x + y)
= ,
fyx fyy cos(x + y) − x sin(x + y) −x sin(x + y)

and its determinant

( )( ) ( )2
D(x, y) = 2 cos(x + y) − x sin(x + y) − x sin(x + y) − cos(x + y) − x sin(x + y)
= − cos2 (x + y).

Since
D(0, kπ) = − cos2 (kπ) = −1 < 0,
all the critical points are saddle points.

December 13, 2017 2 Karel Švadlenka

 8

z
−2

−4

−6

−8
6

−2

−4
y 6 8
−6 2 4
−2 0
−6 −4
−8
x

Graph of the function f (x, y).

Problem 10.3 If f : R2 → R is an even differentiable function, show that (0, 0) is a critical

point of f .

A function f : R2 → R is called even if f (−x, −y) = f (x, y) for all x, y ∈ R. We only have
to show that fx (0, 0) = fy (0, 0) = 0.

By definition

f (h, 0) − f (0, 0) f (−h, 0) − f (0, 0)

fx (0, 0) = lim , fx (0, 0) = lim ,
h→0 h h→0 −h
so (since by differentiability of f the above two limits exist) we get by adding these identities,

f (h, 0) − f (−h, 0)
fx (0, 0) = lim .
h→0 2h
However, by the evenness of the function f , the numerator is equal to zero for any h, which
immediately yields fx (0, 0) = 0.

The partial derivative with respect to y is treated analogously:

f (0, k) − f (0, −k) 0

fy (0, 0) = lim = lim = 0.
k→0 2k k→0 2k

Problem 10.4 Find the distance of the point (2, 3, 0) and the plane x − 4y + z = 3.

The distance of a point and a plane is defined as the distance of the point to the nearest point
on the plane. Any point on the plane can be expressed as (x, y, 3 − x + 4y) and the distance of
this point and the point (2, 3, 0) is
√
d(x, y) = (x − 2)2 + (y − 3)2 + (3 − x + 4y − 0)2 .

December 13, 2017 3 Karel Švadlenka

 Our task is to find x, y for which the above value d(x, y) is the smallest. Since the square
root is an increasing function on (0, ∞), notice that it is equivalent to finding x, y that minimize

D(x, y) = (x − 2)2 + (y − 3)2 + (3 − x + 4y)2 .

Note also that this is a quadratic function that increases to infinity when either x → ∞ or
y → ∞, so we do not have to care about what happens for large x and y and need only to
check all critical point. (Otherwise, since we deal with minimization on the whole R2 , which
is not a compact set, we would not be able to prove the existence of a minimum. What we
said above means that for (x, y) far enough from the origin – let us say, further than M – the
function values are greater than 22, which is the value D(0, 0). Therefore, we can restrict the
minimization to the closed ball B((0, 0), M ), which is a compact set.)

Computing the partial derivatives we have

Dx (x, y) = 2(x − 2) − 2(3 − x + 4y), Dy (x, y) = 2(y − 3) + 8(3 − x + 4y).

Hence the condition on vanishing partials gives the system of linear equations

2x − 4y = 5
4x − 17y = 9.
49 1
Solving this system we obtain x = 18 , y = 9 and the corresponding distance

13
d( 49 1
18 , 9 ) =
√ .
18
Since this is the only critical point, it must yield the global minimum because the function is
quadratic and coercive (i.e., increases without bounds as ∥(x, y)∥ → ∞). We could also check
the sign of the Hessian to see that it is a local minimum and from the coercivity conclude that
it is a global minimum.

December 13, 2017 4 Karel Švadlenka

 Calculus B: Problems with Solutions (Lecture 11)

Problem 11.1 Sketch the level curves and gradient vector field for the functions f (x, y) =
√
x2 + y 2 and g(x, y) = x − y1 .

We want to plot the points which satisfy f (x, y) = C, where C’s are various constants. This
means
x2 + y 2 = C 2 ,
thus the level curves of f are circles centered at the origin. The radius of the circle, where f is
equal to C is exactly C.

The gradient plot can be sketched by drawing vectors that are perpendicular to the level
curves and point in the direction of increase of f (that is away from the origin), or we can
directly compute
1
∇f (x, y) = √ (x, y),
x + y2
2

which means that the gradient vector at a point (x, y) is a vector connecting the origin and this
point, and whose size is always 1.

We proceed similarly for the function g. In particular, g(x, y) = C implies

1
y= ,
x−C
hence the level curves look like in the figure below. The gradient field is

∇g(x, y) = (1, y12 ).

The gradient vector is thus almost parallel to the y-axis when y is small and almost parallel
to the x-axis when y is large, changing smoothly in between. The gradient vector field is
normalized to 1 in the figure, so that it is clearly visible (so we can see only its direction but
not the actual size here).

2 2

1.5 1.5

1 1

0.5 0.5

0 0

−0.5 −0.5

−1 −1

−1.5 −1.5

−2 −2
−2.5 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2 2.5 −2.5 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2 2.5
 Problem 11.2 Find the minimum of the function f (x, y, z) = x2 + y 2 + z 2 subject to the
constraints 2y + z = 6 and x − 2y = 4.

We present two methods of solution.

• The first method is based on the fact that we can reduce the problem to minimization of
a function of one variable without any constraints. Indeed, from the first constraint we
see that z = 6 − 2y and from the second constraint that x = 4 + 2y, so substituting these
expressions in the function, we obtain
f (x, y, z) = (4 + 2y)2 + y 2 + (6 − 2y)2 ,
which is a function of one variable, let us denote it by g(y).
Now it is enough to find the minimum of g. To this end, we compute its derivatives:
g ′ (y) = 4(4 + 2y) + 2y − 4(6 − 2y) = 18y − 8
g ′′ (y) = 18.
From the first derivative we see that the only candidate for minimum is y = 4/9 and from
the second derivative we see that it is really a minimum (these facts are obvious if we
notice that g is just a quadratic function).
From the constraints we then get the values of x and z: x = 4 + 2y = 35/9, z = 6 − 2y =
46/9. Finally, the minimum of f is (35/9)2 + (4/9)2 + (46/9)2 = 373/9.
• The other method uses the general theory of Lagrange multipliers. Denote g1 (x, y, z) =
2y + z − 6 and g2 (x, y, z) = x − 2y − 4. We check the following points:
1. Points (x, y, z) satisfying
∇f (x, y, z) = λ1 ∇g1 (x, y, z) + λ2 ∇g2 (x, y, z), g1 (x, y, z) = 0, g2 (x, y, z) = 0,
for some real numbers λ1 , λ2 . When we rewrite these equations, we get
2x = λ1 · 0 + λ2 · 1
2y = λ1 · 2 + λ2 · (−2)
2z = λ1 · 1 + λ2 · 0
2y + z − 6 = 0
x − 2y − 4 = 0.
This system can be easily solved by expressing λ1 from the third equation (λ1 = 2z),
then λ2 from the first equation (λ2 = 2x) and plugging these values into the second
equation. In this way we obtain (from the second, fourth and fifth equations) a system
of linear equations for x, y, z. Solving this system we obtain x = 35/9, y = 4/9 and
z = 46/9.
2. Points (x, y, z) satisfying the constraint, where ∇g1 (x, y, z) and ∇g2 (x, y, z) are par-
allel vectors. Since ∇g1 (x, y, z) = (0, 2, 1) and ∇g2 (x, y, z) = (1, −2, 0) these vectors
can never be parallel and there are no points in this category.
3. The endpoints of the constraint curve. In this case the constraints are two planes
and their intersection (i.e., set of points that satisfies both constraints) is a straight
line, which extends to ”infinity”, so we can say that the endpoints are at infinity.
However, if we look at the function f , we see that it increases without bounds when
either of the variables x, y or z goes to infinity. Since we are concerned with minimum
of f here, we do not have to care about far-away points (f has large value there).

December 20, 2017 2 Karel Švadlenka

 Above we obtained only one candidate for minimum of f . From the form of the function
we conclude that this candidate in paragraph 1 is indeed a point of global minimum.

Problem 11.3 Find the extreme values of the function f (x, y) = x − y 2 under the constraints
x3 − y 2 = 0 and −1 ≤ y ≤ 1.

There are two ways how to solve this problem.

• The first method is based on the fact that we can reduce the problem to minimization of
a function of one variable. Indeed, from the first constraint we find y 2 = x3 and hence
the function restricted to the constraint curve reads

f (x, y) = x − x3 ,

which is a function of one variable – let us denote it by h(x).

Now it is enough to find the extrema of g. To this end, we compute its derivatives:

h′ (x) = 1 − 3x2
h′′ (x) = −6x.
√ √
The first derivative is equal to zero when either x = 1/ 3 or x = −1/ 3. However, for
the second negative value, there is no y which would satisfy the constraint y 2 = x3 . Hence,
we are left with the first value, which is a local maximum since the second derivative is
negative. It is not a global maximum of x − x3 because this function tends to infinity
when x → −∞. The corresponding value of y is ±3−3/4 , which is obtained by solving
y 2 = x3 .
It remains to take into account the constraint −1 ≤ y ≤ 1. Both points (3−1/2 , 3−3/4 ) and
(3−1/2 , −3−3/4 ) obtained above satisfy this constraint. As we found above, these points
cannot give extrema since f tends to ±∞ when x → ∓∞. However, due to the second
constraint −1 ≤ y ≤ 1, the set on which we look for extrema is compact and theory
guarantees the existence of both a global minimum and a global maximum.
The curve is depicted in the figure below. Notice that it does not have a tangent vector at
the point (0, 0). (This is the same as saying that the function y 2/3 does not have derivative
at y = 0.) Therefore, we should also check the point (0, 0) when looking for extrema.
x3 −y2 = 0
2

1.5

0.5

0
y

−0.5

−1

−1.5

−2
−1 −0.5 0 0.5 1 1.5 2 2.5 3
x

The constraint curve – the blue part shows the points satisfying both constraints, the green part satisfies only the
first constraint.

December 20, 2017 3 Karel Švadlenka

 The extrema are obtained by comparing the values of f at the above two points and at
the endpoints of the constraint curve, that is, the points (1, −1) and (1, 1):
√
f (3−1/2 , 3−3/4 ) = f (3−1/2 , −3−3/4 ) = 2 3/9, f (0, 0) = f (1, −1) = f (1, 1) = 0.

Hence, the maximum is attained at the points (3−1/2 , ±3−3/4 ) and minimum at the points
(1, ±1) and (0, 0).

• The second method is based on the theory of Lagrange multipliers. Denote g(x, y) =
x3 − y 2 . We check the following points:
1. Points (x, y) satisfying

∇f (x, y) = λ∇g(x, y), g(x, y) = 0,

for some real number λ. When we rewrite these equations, we get

1 = λ · (3x2 )
−2y = λ · (−2y)
x3 − y 2 = 0.

From the second √ equation we immediately have λ = 1, whence the first equation
yields x = ±1/ 3. Considering also the third equation, we get the same points as
in the first method: (3−1/2 , 3−3/4 ) and (3−1/2 , −3−3/4 ).
2. Points (x, y) satisfying the constraint, where ∇g(x, y) = 0. Here ∇g(x, y) =
(3x2 , −2y), so there is only one such point: (0, 0).
3. The endpoints of the constraint curve. These are (1, −1) and (1, 1).

We then compare the values of the function f at the above obtained five points. Since
these points are the same as in the first method, we arrive at the same conclusion.

December 20, 2017 4 Karel Švadlenka

 Calculus B: Problems with Solutions (Lecture 12)

Problem 12.1 Set R = [−1, 1] × [−2, 3]. Find

∫∫ √
1 − x2 dA.
R

The integral
√ represents the volume of the region above the rectangle R and below the graph
of f (x, y) = 1 − x2 . Since f does √not depend on y, the graph of f is obtained by translating
the graph of the 1-variable function 1 − x2 (which is a half-circle) along the y-axis. The region
in question is then the upper half of a cylinder with radius 1 and axis running along the y-axis
between the planes y = −2 and y = 3. (Draw a picture.)

The volume of the full cylinder is π × 12 × (3 − (−2)) = 5π and thus the integral equals to
∫∫ √
5
1 − x2 dA = π.
R 2

Problem 12.2 Derive the formula for the volume of a cone of base radius r and height h.

We apply Cavalieri’s principle. Set up the cone, for example, in such a way that the center
of the base is at the origin and the axis of the cone runs along the x-axis. Then the base is in
the yz-plane.

The cone is the region obtained by rotating the graph of the function
r
f (x) = r − x, 0≤x≤h
h
about the x-axis.
z
r y
z = r − hr x
h
x

Hence the volume is computed by the well-known formula

∫ h ∫ h( [ ]
r )2 h ( r )3 h 1 2
V =π 2
(f (x)) dx = π r − x dx = π − r− x = πr h.
0 0 h 3r h 0 3
Problem 12.3 Evaluate (∫ )
∫ 2 y/2 √
x x2 + y 2 dx dy.
1 0

∫ y/2 √
First we compute 0 x x2 + y 2 dx, where y is considered as a fixed parameter (constant).
√ √
Setting u = x2 + y 2 , we have du = x/ x2 + y 2 dx = x/u dx and thus
∫ ∫ √
y/2 √ 5y/2
1( √ ) 1( √ )
2 2
x x + y dx = u2 du = ( 5y/2)3 − y 3 = ( 5/2)3 − 1 y 3 .
0 y 3 3

Here the interval of integration

√ with respect to u is obtained by substituting x = 0 and x = y/2
in turn in the formula u = x2 + y 2 .

Next, we integrate the result with respect to y:

∫ 2
1( √ ) 1( √ ) [ 1 ]2 5(5√5 − 8)
( 5/2) − 1 y dy =
3 3
( 5/2) − 1
3
y4 = .
1 3 3 4 1 32

This is the desired value of the given iterated integral.

Problem 12.4 Evaluate ∫∫

(2xy − y) dA,
D
where
D = {(x, y); 0 ≤ y ≤ 1, −y ≤ x ≤ 1 + y} .

The region D is depicted in the figure below.

y x=1+y
−y = x 1

0 1 x

We can integrate in both directions (that is, first in the x-direction and then in the y-
direction, or the other way round) but in this case it is easier to integrate first with respect to
x and then with respect to y (if we integrate with respect to y first, then we have to split the
integration into 3 parts).

As can be seen from the figure, the integral becomes

∫∫ ∫ 1 ∫ 1+y
(2xy − y) dA = (2xy − y) dx dy.
D 0 −y

December 27, 2017 2 Karel Švadlenka

Now it remains only to calculate the above iterated integral:
∫ 1 ∫ 1+y ∫ 1( ∫
[ 2 ]1+y ) 1
(2xy − y) dx dy = x y − xy x=−y dy = 0 dy = 0.
0 −y 0 0

Problem 12.5 Find the area of the ellipse with semiaxes a and b.

The above ellipse is defined as all points (x, y) satisfying

x2 y 2
+ 2 ≤ 1.
a2 b
We use Cavalieri’s principle to compute the area. Therefore, we first compute the length of the
cross section of the ellipse at a position x along the x-axis and then integrate this value along
the x-axis (from −a to a because for other values of x the above length is zero).

y
r
b y =b 1− x2
a2

x a x
r
x2
y = −b 1 − a2

√
The length of cross section at position x is obviously 2b 1 − x2 /a2 and thus the area
∫ a √ ∫ 1√ [ √ ]1
x2 1 1
A= 2b 1 − 2 dx = 2ba 1 − u2 du = 2ab u 1 − u2 + arcsin u = πab,
−a a −1 2 2 −1

where we have made the change of variables

√ u = x/a and the integral is evaluated by parts
′
(using the functions f (u) = 1 and g(u) = 1 − u2 ).

Problem 12.6 Find the volume of the region in the first octant between the planes x + y + z = 1
and x + y + 2z = 1.

The first plane crosses the x-axis at position x = 1, the y-axis at position y = 1 and the
z-axis at position z = 1. Similarly, the second plane crosses the respective axes at x = 1, y = 1
and z = 12 . Hence, the situation looks like in the following figure:

December 27, 2017 3 Karel Švadlenka

 z

1 D
y

1 C
2
1
E

B x
A 1

We can calculate the volume without using any integrals, just by subtracting the volume of
the tetrahedron ABCE from the volume of the tetrahedron ABCD:
1 1 1
V = − = .
6 12 12

Another method is to integrate the difference of the two planes z = 1 − x − y and 12 (1 − x − y)

in the z-direction, that is 12 (1 − x − y), over the rectangle in the xy-plane with vertices (0, 0, 0),
(1, 0, 0) and (0, 1, 0). This triangle is given by the inequalities 0 ≤ x ≤ 1 and 0 ≤ y ≤ 1 − x, so
we obtain
∫ 1 ∫ 1−x ∫ [ ]
1 1 1 1 2 1−x
V = (1 − x − y) dy dx = (1 − x)y − y dx
0 0 2 2 0 2 y=0
∫ [ ]1
1 11 1 1 1
= (1 − x) dx =
2
− (1 − x) 3
= .
2 0 2 4 3 0 12

December 27, 2017 4 Karel Švadlenka

 Calculus B: Problems with Solutions (Lecture 13)

Problem 13.1 Evaluate the integral

∫ 3 (∫ 9 )
x cos(2y 2 ) dy dx.
0 x2

If we try to integrate first with respect to y, we will fail, as it is not possible to integrate
cos(2y 2 ) in a closed form. Therefore, we try to change the order of integration. The iterated
integral can be written as a double integral
∫∫
x cos(2y 2 ) dA,
D

where D = {(x, y); 0 ≤ x ≤ 3, x2 ≤ y ≤ 9} is depicted in the figure below.

y
y = x2
9

0 3 x

The region D can be also expressed as

√
D = {(x, y); 0 ≤ y ≤ 9, 0 ≤ x ≤ y}.

Hence, we can write this double integral as the iterated integral

∫ (∫ √ 9 y
)
x cos(2y 2 ) dx dy.
0 0

This integral is easily evaluated:

∫ (∫ √
9 y
) ∫ 9[ ]x=√y
2 1 2 2
x cos(2y ) dx dy = x cos(2y ) dy
0 0 0 2 x=0
∫ 9
1
= y cos(2y 2 ) dy
2 0
[ ]9
1 1 2
= sin(2y )
2 4 y=0
1
= (sin 162 − 1).
8
 Problem 13.2 Find the volume of the region under the surface z = xy and above the triangle
in the xy-plane with vertices (0, 1), (1, 1) and (1, 2).

The triangular region in the xy-plane looks like

2
D
1

0 1 x

We denote it by D.

Since the given region is the set of points below the graph of f (x, y) = xy and above D, the
volume of this region is equal to the double integral
∫∫
f (x, y) dA.
D

We can write D = {(x, y); 0 ≤ x ≤ 1, 1 ≤ y ≤ x + 1}, and thus the above integral is rewritten
and evaluated as the iterated integral
∫ 1 (∫ x+1 ) ∫ 1[ ]y=x+1 ∫ 1
1 2 1 11
xy dy dx = xy dx = (x3 + 2x2 ) dx = .
0 1 0 2 y=1 2 0 24

The required volume is 11/24.

Problem 13.3 Let θ be a real number between 0 and 2π and a, b be any real numbers. Define
( )
cos θ − sin θ
A= .
sin θ cos θ

Describe the image of the square [0, 1] × [0, 1] by the mapping

( )
a
T (u) = Au + .
b

Let us split the mapping T into two steps: the first step multiplies by the matrix A, which
maps coordinates (u, v) to (x′ , y ′ ), and the second step adds the vector (a, b), which maps
coordinates (x′ , y ′ ) to (x, y).

The usual method to find the image is to check how horizontal and vertical lines are mapped
by T . However, since some kind of angle come into play, we take a different approach and
consider the following two properties of T :

January 10, 2018 2 Karel Švadlenka

 • How does the distance of a point u = (u, v) from the origin change by multiplying the
point by the matrix A?
To see this, we just have to compute
∥Au∥ = ∥(u cos θ − v sin θ, u sin θ + v cos θ)∥
√ √
= (u cos θ − v sin θ)2 + (u sin θ + v cos θ)2 = u2 + v 2 = ∥u∥.
This equation says that the distance of a point from the origin does not change if we
multiply it by the matrix A.
• How are rays passing through the origin mapped by A?
To see this, take any ray, i.e., line v = cu passing through the origin and having angle ϕ
with the u-axis (then obviously c = tan ϕ is the slope). Then any point (u, cu) on this
line is mapped to
( )
′ ′ u
(x , y ) = A = (u(cos θ − c sin θ), u(sin θ + c cos θ)).
cu
Therefore, the ray is mapped to another ray with a different slope d from the original ray.
Let us compute this slope:
sin θ + c cos θ sin θ + tan ϕ cos θ sin θ cos ϕ + sin ϕ cos θ sin(θ + ϕ)
d= = = = = tan(θ+ϕ).
cos θ − c sin θ cos θ − tan ϕ sin θ cos θ cos ϕ − sin ϕ sin θ cos(θ + ϕ)
This means that the new ray forms an angle θ + ϕ with the x′ -axis, and thus is obtained
by rotating the original ray by the angle θ.

Putting together the above two pieces of information, we conclude that multiplying by the
matrix A to a point rotates the point by the angle θ around the origin. Adding the vector (a, b)
just translates the resulting point by this vector. Hence, the mapping T rotates the square by
the angle θ and translates it by the vector (a, b).

(x′ , y ′)T = A(u, v)T (x, y)T = (x′ , y ′)T + (a, b)T
(a, b)T
v y′ y
1 b
1 b 1 b

b
θ b b

0 1 u 0 1 x′ 0 1 x

Problem 13.4 Evaluate ∫∫

x+y
dA, sin
D x−y
where D is the region bounded by the lines x − y = 1, x − y = 5 and the coordinates axes.

It might be difficult to compute the integral by usual methods. Therefore, we examine the
integrated function f (x, y) and the region D to see if we can exploit some of their properties.
The region D is depicted in the figure below:

January 10, 2018 3 Karel Švadlenka

 y x−y =1 x−y =5
1 5
0 x
(−b, −a) b
D
−1
b

(a, b)

x+y =0

−5

We see that the region is symmetric with respect to the line x + y = 0. Now, let us calculate
the value of f at two points which are symmetric with respect to this line.

As in the figure, we take any (a, b) ∈ D and first compute its reflection with respect to the
line x + y = 0. Since the lines x − y = C, where C is any constant, are perpendicular to
x + y = 0, the reflected point must lie on the line x − y = a − b. Moreover, since (a, b) lies on
the line x = a, the reflected point must lie on the line y = −a. (Similarly, the point (a, b) lies
on the line y = b, so the reflected point lies on x = −b – draw a picture.) We conclude that the
reflected point is (−b, −a).

The values of the integrated function at these two points are

a+b
f (a, b) = sin
a−b ( )
−b − a a+b a+b
f (−b, −a) = sin = sin − = − sin = −f (a, b).
−b + a a−b a−b

Therefore, for each point in the region below the line x + y = 0 the reflected point in D above
this line yields the negative value of the function at the original point. Due to this symmetry,
we can say that the integral equals to zero, without even computing it.

To give a rigorous proof, we would argue that the function is continuous on D, and therefore
integrable on D, which implies that the integral can be obtained as a limit of Riemann sums.
Then we would consider partitions of D that are symmetric with respect to x + y = 0 and show
that if we select the associated points in a suitable way, the Riemann sum corresponding to the
part of D above x + y = 0 can be made exactly minus the sum for the part below x + y = 0.
We will not give the details here.

Problem 13.5 Compute (∫ √ )

∫ 1 1−x2
y
arctan dy dx.
−1 0 x

√
The region over which we integrate is D = {(x, y); −1 ≤ x ≤ 1, 0 ≤ y ≤ 1 − x2 }. This is
a half-disk as shown below.

January 10, 2018 4 Karel Švadlenka

 y
1 √
1 − x2

D1 D2

−1 0 1 x

Since we are dealing with a disk, it is natural to use polar coordinates x = r cos θ, y = r sin θ,
where 0 ≤ r ≤ 1 and 0 ≤ θ ≤ π. If we plug these into our function, we get
y r sin θ
arctan = arctan = arctan(tan θ).
x r cos θ
However, this is not always equal to θ for the range of θ ∈ [0, π] that we consider. Indeed, for
example, arctan(tan 34 π) = arctan(−1) = − 14 π ̸= 43 π. The identity arctan(tan θ) = θ holds for
θ ∈ [0, 12 π), so we consider splitting the integration region into two quarter-disks: one region
D1 to the left of y-axis and another D2 to the right of y-axis. Since arctan xy is an odd function
when viewed as function of x only, and therefore | arctan xy | is even in x, we have
∫∫ ∫∫
y y
arctan dA = 2 arctan dA,
D x D2 x
and it is enough to compute the integral over the quarter-disk D2 (hence 0 ≤ θ < 12 π).

This is obtained by first calculating the Jacobian of the transformation:

( )
∂(x, y) cos θ r(− sin θ)
= = r,
∂(r, θ) sin θ r cos θ
and performing the change of variables in the integral:
∫∫ ∫ 1 ∫ π/2
y r sin θ ∂(x, y)
arctan dA = arctan dθ dr
D2 x 0 0 r cos θ ∂(r, θ)
∫ 1 ∫ π/2
= θr dθ dr
) (∫ )
0 0
(∫ 1 π/2
= r dr θ dθ
0 0

π2
= .
8
Therefore, the integral from the problem is equal to π 2 /4.

Problem 13.6 Compute the value of the triple integral

∫∫∫
|y| dV,
D

where D is the three-dimensional region bounded by the paraboloids z = 12 − x2 − y 2 and

z = 2x2 + 2y 2 .

The region D is sketched in the following figure:

January 10, 2018 5 Karel Švadlenka

 z
12
z = 12 − x2 − y 2 z = 2x2 + 2y 2
b

b
8
z = 12 − x2 − y 2

D y
2
z = 2x
b
+ 2y 2

−2
b

(x, y) 0 2 x
−2
D̃

In this case it is natural to first integrate with respect to z along the blue segment, that is,
∫ 12−x2 −y2
g(x, y) = |y| dz.
2x2 +2y 2

Then we integrate the resulting function g of two variables over the two-dimensional region
determined by the set of (x, y) such that the paraboloid z = 12 − x2 − y 2 is above the paraboloid
z = 2x2 + 2y 2 . Solving
12 − x2 − y 2 ≥ 2x2 + 2y 2 ,
we find that it corresponds to a disk of radius 2 centered at the origin: D̃ = {(x, y); x2 +y 2 ≤ 4}.
Hence,
∫∫∫ ∫ ∫ ∫ 12−x2 −y2
|y| dV = |y| dz dA.
D D̃ 2x2 +2y 2

However, the function |y| is defined differently in {y > 0} and in {y < 0}, so we split the
integral into two parts: over the region (half-disk) D̃+ = {(x, y); x2 + y 2 ≤ 4, y ≥ 0} and over
the region D̃− = {(x, y); x2 + y 2 ≤ 4, y ≤ 0}. From the symmetry of the function |y| and the
integration domain, we see that
∫∫∫ ∫ ∫ ∫ 12−x2 −y2 ∫ ∫ ∫ 12−x2 −y2
|y| dV = y dz dA + (−y) dz dA
D D̃+ 2x2 +2y 2 D̃− 2x2 +2y 2
∫∫ ∫ 12−x2 −y 2
= 2 y dz dA.
D̃+ 2x2 +2y 2

The last integral is evaluated by standard methods for double integrals. In particular we
employ polar coordinates to find
∫ ∫ ∫ 12−x2 −y2 ∫∫ ∫ 2∫ π
128
y dz dA = 3 y(4 − x2 − y 2 ) dA = 3 r sin θ(4 − r2 )r dθ dr = .
2
D̃+ 2x +2y 2 D̃+ 0 0 5
The given integral is equal to 256/5 = 51.2.

Problem 13.7 Let b > a > 0. Find the volume of the 3D region above the plane z = a and
inside the sphere x2 + y 2 + z 2 = b2 .

We have to find the volume of the spherical cap D depicted in the figure:

January 10, 2018 6 Karel Švadlenka

 z
b
y
√
b2 − a2 b
a D
b
−b 0 b x

√ intersection of the ball x + y + z ≤ b with the plane z = a is a disk of ra-

The 2 2 2 2

dius b2 √ − a2 centered at (0, 0, a). Hence, we obtain the volume of D by integrating the
function b2 − x2 − y 2 − a, representing the graph of the upper hemisphere, over the disk
E = {(x, y); x2 + y 2 ≤ b2 − a2 }. We subtracted a from the function because without this sub-
traction, we would obtain the volume of the spherical cap D (which is to be calculated in this
problem) plus the
√ volume of the cylinder with basis E and height a. (Another method would
be to integrate b2 − x2 − y 2 over E and then subtract the volume of the cylinder, which is
πa(b2 − a2 ).)

Therefore, using polar coordinates x = r cos θ, y = r sin θ with Jacobian r,

∫ ∫ (√ ∫ √ ∫
) b2 −a2 2π (√ )
V = b − x − y − a dA =
2 2 2 b2 − r2 − a r dθ dr
E 0 0
∫ √
b2 −a2 (√ )
= 2π b2 − r2 − a r dr
0
[ ]√ 2 2
1 2 1 2 b −a
= 2π − (b − r ) − ar
2 3/2
3 2 r=0
π( 3 )
= 2b + a − 3ab .
3 2
3

Problem 13.8 Evaluate the integral

∫∫ √
2x2 + 2y 2 + 3 dA,
D

where D is the disk x2 + y 2 ≤ 1.

Using polar coordinates x = r cos θ, y = r sin θ, the disk D is expressed by

D = {(r, θ) ∈ R2 ; 0 ≤ r ≤ 1, 0 ≤ θ ≤ 2π},

January 10, 2018 7 Karel Švadlenka

so we have
∫∫ √ ∫ 2π ∫ 1√
2x2 + 2y 2 + 3 dA = 2r2 + 3 · r drdθ
D 0 0
(∫ 2π ) (∫ 1 √ )
= dθ · 2
r 2r + 3 dr
0 0
[ ]1
1 3
= 2π (2r2 + 3) 2
6 0
π √ √
= (5 5 − 3 3).
3
Do not forget to multiply by the Jacobian r on the right-hand side of the first equality.

Problem 13.9
(1) Using a double integral, compute the volume of a sphere of radius a > 0.

(2) Let 0 < b ≤ a. Find the volume of the region inside the sphere x2 + y 2 + z 2 = a2 and
inside the cylinder x2 + y 2 = b2 .

0 | ≤ a), i.e., the intersection of the sphere

(1) The cross section of the sphere at height z0 (|z√
with the plane z = z0 , is the disk of radius a2 − z02 . Since the area of this disk is
π(a2 − z02 ), we calculate
∫ a (∫ ) ∫ a
1 dA dz = π(a2 − z 2 ) dz
−a x +y ≤a −z
2 2 2 2 −a
∫ a
= 2π (a2 − z 2 ) dz
[0 ]a
z3
= 2π a z −
2
3 0
4 3
= πa .
3

There is the another way to compute the volume of the sphere. Let D(a) be the closed
disk of radius a (that is, the intersection of the sphere with the xy-plane):
{ }
D(a) = (x, y) ∈ R2 ; x2 + y 2 ≤ a2 .

Then the sphere is written as

{ √ √ }
(x, y, z) ∈ R3 ; (x, y) ∈ D(a), − a2 − x2 − y 2 ≤ z ≤ a2 − x2 − y 2 .

Therefore, we can express the volume of the given sphere as follows:

∫∫ (∫ √ 2 2 2 ) ∫∫
a −x −y √
√ 1 dz dA = 2 a2 − x2 − y 2 dA
D(a) − a2 −x2 −y 2 D(a)

January 10, 2018 8 Karel Švadlenka

 Using polar coordinates, we calculate the above integral as follows:
∫∫ √ ∫ 2π (∫ a √ )
2 a2 − x2 − y 2 dA = 2 a2 − r2 · r dr dθ
D(a) 0 0
∫ a √
= 4π r a2 − r2 dr
[0 ]
−1 ( 2 )3 a
= 4π a −r 2 2
3 0
4 3
= πa .
3

(2) The given region is expressed by

{ √ √ }
(x, y, z) ∈ R3 ; (x, y) ∈ D(b), − a2 − x2 − y 2 ≤ z ≤ a2 − x2 − y 2 .

We can use a method similar to the second method of the first part (1) of this problem
above. Hence, the volume is
∫∫ (∫ √ 2 2 2 ) ∫∫
a −x −y √
√ 1 dz dA = 2 a2 − x2 − y 2 dA
D(b) − a2 −x2 −y 2 D(b)
∫ 2π (∫ b √ )
=2 a − r · r dr dθ
2 2
0 0
∫ b √
= 4π r a2 − r2 dr
0
[ ]
−1 ( 2 )3 b
= 4π a −r 2 2
3 0
4π { 3 ( 2 )3 }
= a − a −b 2 2
.
3

In particular, if a = b, we get the same answer as in (1). This is natural because the two
regions given in (1) and (2) are exactly the same when a = b.

Problem 13.10 Compute the value of the triple integral

∫∫∫
(|x| + y) dV,
D

where D is the three-dimensional region below the paraboloid z = 1−x2 −y 2 , inside the cylinder
x2 + y 2 = 1 and above the xy-plane.

Since the intersection of the paraboloid and cylinder is the unit circle in the xy-plane, we
can write D as
{ }
D = (x, y, z) ∈ R3 ; x2 + y 2 ≤ 1 , 0 ≤ z ≤ 1 − x2 − y 2 .

Integrating with respect to z, we have

∫∫∫ ∫∫ ∫ 1−x2 −y 2 ∫∫
(|x| + y)dV = (|x| + y) dA dz = (|x| + y)(1 − x2 − y 2 ) dA, (1)
D Ω 0 Ω

January 10, 2018 9 Karel Švadlenka

 { }
where Ω is the two dimensional unit disk: Ω = (x, y) ∈ R2 ; x2 + y 2 ≤ 1 . Hence, we can
calculate the above integral as the iterated integral:
∫ (∫ √ 2
1 1−x
)
(1) = √ (|x| + y)(1 − x2 − y 2 ) dy dx
−1 − 1−x2
∫ ( ∫ √1−x2 )
1
= 2 |x|(1 − x − y ) dy
2 2
dx
−1 0
∫ 1 [ 3
]√1−x2
y
= 2|x| (1 − x2 )y − dx
−1 3 0
∫
4 1 3
= |x|(1 − x2 ) 2 dx
3 −1
∫
8 1 3
= x(1 − x2 ) 2 dx
3 0
[ ]
8 −1 ( )5 1
= 1−x 2 2
3 5 0
8
= .
15
In the above calculation, we have used the following important property of odd (and even)
functions: If f (u) is an odd function (i.e., f (−u) = −f (u)), we have
∫ a
f (u) du = 0,
−a

while if f (u) is an even function (i.e., f (−u) = f (u)), we have

∫ a ∫ a
f (u) du = 2 f (u) du
−a 0

for any a > 0. For example, f (u) = u2n is an even function, and f (u) = u2n+1 is an odd
function (n ∈ N).

(We can also calculate the integral in (1) using polar coordinates. Try by yourself.)

Problem 13.11 Sketch the curve r = cos(2θ) and find the area of the region enclosed by one of
its loops. (The equation of the curve above is given in usual polar coordinates.)

The curve r = cos 2θ is depicted below. Note that all the four loops are closed and meet at
the origin (this is not visible in the figure due to software limitation).

January 10, 2018 10 Karel Švadlenka

 1

0.8

0.6

0.4

0.2

−0.2

−0.4

−0.6

−0.8

−1

−1 −0.5 0 0.5 1

Let D be the region enclosed by one of its loops contained in {(x, y) ∈ R2 ; x ≥ 0, −x ≤ y ≤

x}. We will denote the corresponding region in polar coordinates by D̃:

D̃ = {(r, θ) ∈ R2 ; −π/4 ≤ θ ≤ π/4, 0 ≤ r ≤ cos(2θ)}.

Then, transforming coordinates, we obtain the area of D as follows:

∫∫ ∫∫
1 dA = r drdθ
D D̃
∫ π/4 (∫ cos 2θ )
= r dr dθ
−π/4 0
∫ π/4 [ ]cos 2θ
1 2
= r dθ
−π/4 2 0
∫ π/4
1
= cos2 (2θ) dθ
2 −π/4
∫ π/4
1
= (1 + cos(4θ)) dθ
4 −π/4
[ ]π/4
1 1
= θ + sin(4θ)
4 4 −π/4
π
= .
8

∫∫
Problem 14.12 Compute the integral D xy 3 dA, where D is the region in the first quadrant
bounded by the lines x = 1 and x = 2 and the hyperbolas xy = 1 and xy = 3.

D is expressed by { }
1 3
(x, y) ∈ R2 ; 1 ≤ x ≤ 2 , ≤y≤ .
x x

January 10, 2018 11 Karel Švadlenka

Thus, we can calculate the given integral as follows:
∫∫ ∫ (∫ 3
)
2 x
xy 3 dA = x y 3 dy dx
1
D 0 x
∫ 2 ]3 [
y4 x
= x dx
1 4 1
∫ 2 ( 4 )
x

x 3 1
= − dx
1 4 x4 x4
∫ 1
dx
= 20 3
0 x
[ −2 ]2
x
= 20
−2 1
2−2 − 1
= 20 ·
−2
3
= 20 ·
8
15
= .
2

January 10, 2018 12 Karel Švadlenka