I. INTRODUCTION.

ELEMENTS OF MATHEMATICAL
LOGIC AND SET THEORY. REAL NUMBERS
Propositional logic.
The language of propositional logic consists of
• a set of primitive symbols (syntactical variables, e.g. A, B, ϕ, ψ. . .)
• logical operators: & (in other texts, the mark ∧ is used), ∨, ¬ (∼ is used as
well), ⇒, ⇔
• auxiliary symbols: (,)
Definition 1.1. A formula of propositional logic:
(1) Any syntactical variable is a formula (so called atomic formula).
(2) If ϕ and ψ are formulae, then so is ¬ϕ (negation of ϕ), ϕ & ψ (conjunction of
ϕ and ψ), ϕ ∨ ψ (disjunction or alternative of ϕ and ψ), ϕ ⇒ ψ (implication: ϕ
implies ψ) and ϕ ⇔ ψ (equivalence of ϕ and ψ).
(3) Every formula is built inductively from atomic formulae using the previous step.

The truth tables.

1 stands for the truth value true, 0 for false. The following table shows how to com-
pute truth values of conjunction, disjunction, implication and equivalence of formulas
A and B knowing the truth values of A, B.

A B ¬A A & B A∨B A⇒B A⇔B

1 1 0 1 1 1 1
1 0 0 0 1 0 0
0 1 1 0 1 1 0
0 0 1 0 0 1 1

Proposition 1.1. Any implication is equivalent to its contrapositive:

(A ⇒ B) ⇔ (¬B ⇒ ¬A).

Proposition 1.1 gives an important method of proving implications (see below).

Proposition 1.2 (De Morgan Laws).

¬(A & B) ⇔ (¬A ∨ ¬B)

¬(A ∨ B) ⇔ (¬A & ¬B)

A formula which is true regardless of the truth values of its atomic subformulas is
called tautology.
First order logic (predicate calculus).
Mathematical theories are expressed using first order logic. It differs from proposi-
tional logic by its use of quantified variables: (∀x ∈ A), (∀y ∈ R), (∀f1 ∈ F ) etc.; we
often write (∀x), (∀n), when it is clear to which set x (n, respectively) belongs. Further,
language of first order logic involves function symbols (e.g., +, . are binary function
1
2

√
symbols, is unary) and predicate symbols (e.g., <, ≤, ∈, =, ⊆ and 6= are binary
predicate symbols). Function symbols represent operations with mathematical objects,
e.g. with numbers and functions, predicate symbols represent relations between such
objects. Use of logical operators and auxiliary symbols remains the same as in proposi-
tional logic. Formulae are built inductively using logical operators and quantifying over
individual variables.
Sets.
A set is a collection of well defined and distinct objects. E.g., R is the set of all real
numbers, N is the set of all natural numbers. Another example is a set of points in the
plane with given property or the set of all functions with nonnegative values.
Notation and terminology.
• x ∈ A . . . x is an element of A, x belongs to A
• x 6∈ A . . . x is not an element of A
• A ⊆ B or A ⊂ B . . . the set A is a subset of B (inclusion)
• A = B . . . the sets A and B have the same elements, equivalently A ⊂ B &
B⊂A
• ∅ . . . the empty set, ∅ = {x; x 6= x}
• A ∪ B . . . union of the sets A and B; A ∪ B = {x; x ∈ A ∨ x ∈ B}
• A ∩ B . . . intersection of the sets A and B; A ∩ B = {x; x ∈ A & x ∈ B}
• A ∩ B = ∅ . . . A and B are disjoint, i.e. they have no common element
• A \ B . . . relative complement, set theoretic difference of the sets A and B;
A \ B = {x ∈ A; x 6∈ B};
 e.g. R \ Q = I – the set of irrationals
• A1 × A2 × · · · × An = [a1 , . . . , an ]; a1 ∈ A1 & · · · & an ∈ An . . . Cartesian
product of the the sets A1 , . . . , An
Let I be a nonempty set of indices, and let Aα be a set for each α ∈ I.
S
• α∈I Aα = {x; (∃α ∈ I) x ∈ Aα } . . . union of all Aα s, the set of all elements
belonging to at least one Aα
T
• α∈I Aα = {x; (∀α ∈ I) x ∈ Aα } . . . intersection of all Aα s, the set of all
elements belonging to Aα for every α

Formulae with quantifiers.

With given set M and formula ϕ (assertion about x),
• (∀x ∈ M ) ϕ(x) means ’for all x ∈ M ϕ(x) holds’,
• (∃x ∈ M ) ϕ(x) means ’there is at least one x ∈ M such that ϕ(x) holds’,
• (∃!x ∈ M ) ϕ(x) is an abbreviation for ’there exists a unique (i.e. exactly one)
x ∈ M such that ϕ(x) holds’.
Notice that
• (∀x ∈ M ) ϕ(x) means (∀x) (x ∈ M ⇒ ϕ(x)),
• (∃x ∈ M ) ϕ(x) means (∃x) (x ∈ M & ϕ(x)).
Negation of formulae with quantifiers.
• ¬(∀x ∈ M ) ϕ(x) is the same as (∃x ∈ M ) ¬ϕ(x),
• ¬(∃x ∈ M ) ϕ(x) is the same as (∀x ∈ M ) ¬ϕ(x).
 3

Proof methods.
• Direct proof. The aim is to prove A ⇒ B. We do it via proving sequence of
assertions: A ⇒ C1 ⇒ C2 ⇒ · · · ⇒ Cn ⇒ B.
• Indirect proof. Proving a contrapositve ¬B ⇒ ¬A instead of A ⇒ B.
• Proof by contradiction. We want to prove an assertion A. To this aim, we assume
¬A holds and deduce (in several steps) a contradiction from it. I.e. a formula of the
form B & ¬B (typically A & ¬A). Since we have come to a nonsense, the assumption
¬A was wrong and A holds true.
• Mathematical induction. We want to prove an assertion ϕ(n) for every natural
number n (i.e. n = 1, 2, 3, . . . ). First, we prove ϕ(1). Then, assuming ϕ(n) holds (such
an assumption is called inductive assumption), we prove ϕ(n + 1). Since n has been
chosen arbitrarily, we can conclude ϕ(n) holds for every natural n.
Theorem 1.3 (De Morgan Laws for sets). Suppose I 6= ∅, S, Aα (for every α ∈ I)
are sets. Then
[ \
S\ Aα = (S \ Aα ),
α∈I α∈I
\ [
S\ Aα = (S \ Aα ).
α∈I α∈I

Sketch of the proof. The equalities can be proven directly (regarding the notion of
equality): suppose x is an element of the left hand side and show it is also an element
of the right hand side. And vice versa. 
√ √
Example. Irrationality of 2. 2 6∈ Q, i.e. if x ∈ R and x2 = 2, then x is
irrational.
√
Proof. Assume for contradiction 2 = x = pq so that p, q ∈ N and the fraction is
2
irreducible (i.e. p, q are coprime). Hence pq2 = x2 = 2, it means p2 = 2q 2 and,
consequently, p is even. Let p = 2r, r ∈ N. Then p2 = 4r2 = 2q 2 , thus 2r2 = q 2 and q
is even as well. This contradicts assumption on p, q being coprime. So, the assumption
that x can be expressed as pq was false, and x is irrational. 
n(n+1)
Example. (∀n ∈ N) 1 + 2 + 3 + · · · + n = 2 .
Proof. Easily by induction on n. 
Example. Binomial theorem. For every a, b ∈ R and every n ∈ N,
n  
n
X n
(a + b) = ak bn−k .
k
k=0

Recall that for natural numbers n and k, n ≥ k, the binomial coefficient

 
n n! n · (n − 1) . . . (n − k + 1)
= =
k k! · (n − k)! k!
4

n n



is equal to the

number of k-element subsets of an n-element set. Further 0 = n = 1,
n
of course, 1 = n. In the proof, the equality
     
n+1 n n
= +
j j j−1
Pn
uses. Finally, k=0 ak = a0 + a1 + · · · + an .
Proof of binomial theorem. By induction
on n. For

n = 1, the left hand side, L, equals
1 0 1 1 1 0
a + b, while the right hand side R = 0 a b + 1 a b = b + a (k ∈ {0; 1} here). Hence
L = R.
Suppose binomial theorem holds for given n and let us prove it for n + 1. L =
(a + b)n+1 = (a + b) · (a + b)n . According to the induction assumption, the latter equals
to
n   n   n  
X n k n−k X n k+1 n−k X n k n−k+1
(a + b) · a b = a b + a b .
k k k
k=0 k=0 k=0

Now, let us rename the variable k in both sums: we put j = k + 1 in the first sum and
j = k in the second sum, respectively. Thus
n+1
X  n  
n j n−j+1
X n j n−j+1
L= a b + a b
j=1
j−1 j=0
j
n        
X n n j n−j+1 n 0 n+1 n n+1 0
= + a b + a b + a b .
j=1
j − 1 j 0 n

According to the equalities mentioned above, this equals the right hand side for n+1. 
Sets of numbers.
• The set of all natural numbers N = {1, 2, 3, . . . }.
• The set of integers Z = N ∪ {0} ∪ {−n;
n n ∈ N} = {. . . , −2,
o −1, 0, 1, 2, . . . }.
p
• The set of all rational numbers Q = q ; p ∈ Z & q ∈ N , where

p1 p2
= ⇔ p1 q2 = p2 q1 .
q1 q2

Definition 1.2. The set of all real numbers R is a set with operations + (addition) and
· (multiplication) and with relation ≤ (ordering), in which the following three groups of
axioms are satisfied:
I. properties of addition and multiplication and their relationships,
II. properties of ordering and its relationships to addition and multiplication,
III. Infimum Axiom.
Once we specify the groups of properties I, II, III, Definition 1.2 will be complete.
I. Properties of addition and multiplication.
• + and · are commutative: (∀x, y ∈ R) (x + y = y + x & x · y = y · x)
• + and · are
associative: (∀x, y, z ∈ R) x + (y + z) = (x + y) + z & x · (y · z) =
(x · y) · z
 5

• existence of additive identity element, zero: (∃0 ∈ R) (∀x ∈ R) x + 0 = x

• existence of multiplicative

identity element, one: (∃1 ∈ R) 0 6= 1 & (∀x ∈
R) (x 6= 0 ⇒ x · 1 = x)
• distributivity of · over +: (∀x, y, z ∈ R) x · (y + z) = x · y + x · z
• existence of additive inverses: (∀x ∈ R) (∃ − x ∈ R) x + (−x) = 0
• existence of multiplicative inverses: (∀x ∈ R) x 6= 0 ⇒ (∃x−1 ∈ R) x · x−1 = 1

II. Properties of ordering.

• ≤ is transitive: (∀x, y, z ∈ R) (x ≤ y & y ≤ z) ⇒ x ≤ z

• ≤ is weakly antisymmetric: (∀x, y ∈ R) (x
≤ y & y ≤ x) ⇒ x = y
• dichotomy of ≤: (∀x, y ∈ R) x ≤ y∨y
≤ x (any two elements can be compared)
• (∀x, y, z ∈ R) x ≤ y ⇒ x + z ≤ y + z

• (∀x, y ∈ R) (0 ≤ x & 0 ≤ y) ⇒ 0 ≤ x · y
Definition 1.3. We say that a set M ⊂ R is bounded from below, if there exists
a number a ∈ R such that (∀x ∈ M ) a ≤ x. Such a is called lower bound of M .
Analogously, we define sets bounded from above and upper bound. We say that a set
M ⊂ R is bounded if it is bounded from below and from above.
III. Infimum Axiom. Let M be a nonempty set bounded from below. Then there
exists a unique number i ∈ R with the following properties:
(i) (∀x ∈ M ) x ≥ i (i.e. i is a lower bound of M ),
′ ′ ′


(ii) (∀i ∈ R) i > i ⇒ (∃x ∈ M ) x < i (i.e. i is the biggest lower bound).
Definition 1.4. The i ∈ R from Infimum Axiom is called infimum of M and denoted
i = inf M . Analogously, for a nonempty set N bounded from above, supremum of N ,
sup N defines as the least upper bound of N .
Remarks.
(1) The set R exists and is defined uniquely by I–III (a deep theorem of set theory).
(2) (∀x ∈ R) 0 · x = x · 0 = 0 (0 makes every element vanish).
(3) (∀x ∈ R) − x = (−1) · x
(4) (∀x ∈ R \ {0}) (∀n ∈ N) x−n can be defined as (xn )−1 = (x−1 )n (xn defines
inductively: x1 = x, xn+1 = xn · x for every x ∈
R).
(5) (∀x ∈ R) (∀y ∈ R) (x > 0 & y > 0) ⇒ x · y > 0
Intervals.
Let a, b ∈ R, a < b.
• (a, b) = {x ∈ R; a < x < b} . . . open interval
• ha, bi = {x ∈ R; a ≤ x ≤ b} . . . closed interval
• (a, bi = {x ∈ R; a < x ≤ b} . . . semiopen interval
• ha, b) = {x ∈ R; a ≤ x < b} . . . semiclosed interval
• unbounded intervals:
(a, +∞) = {x ∈ R; a < x}
(−∞, a) = {x ∈ R; a > x}
ha, +∞) = {x ∈ R; a ≤ x}
(−∞, ai = {x ∈ R; a ≥ x}
6

Theorem 1.4 (Supremum Theorem). If M ⊂ R is nonempty bounded from above

then there exists the unique s = sup M .
Sketch of the proof. sup M = − inf(−M ), where −M = {x ∈ R; −x ∈ M }. 
Definition 1.5. Let M ⊂ R. We call a ∈ R maximum of M and write a = max M if
a ∈ M and (∀x ∈ M ) x ≤ a. Minimum of M , min M , defines analogously.
Theorem 1.5 (Existence of integer part).

(∀x ∈ R) (∃!k ∈ Z) k ≤ x < k + 1.

Such k is called integer part of x and is denoted by [x] or ⌊x⌋.

Proof. Let x ∈ R and put M = {n ∈ Z; n ≤ x}. Clearly, the set is bounded from above.
We shall prove M is nonempty.
Assume for contradiction M = ∅. From dichotomy of ≤, it follows that (∀z ∈ Z) z >
x. Under the assumption, the set Z is nonempty and bounded from below, hence by
Infimum Axiom there exists y = inf Z ∈ R. It follows that (∀z ∈ Z) z − 1 ≥ y, i.e.,
(∀z ∈ Z) z ≥ y + 1 – a contradiction with definition of infimum. We can conclude that
M 6= ∅.
Now we are ready to use Supremum Theorem: let G = sup M . Since G is the lowest
upper bound of M , G−1 is not an upper bound, thus there is k ∈ M such that G−1 < k.
I.e. G < k + 1 which is therefore not in M . Hence k ≤ x < k + 1 for such k. 
Theorem 1.6 (Archimede Property).

(∀x ∈ R) (∃n ∈ N) x ≤ n.

Proof. Put n = max{1, ⌊x⌋ + 1}. 

Theorem 1.7 (Existence of n-th root).

(∀x ∈ h0, ∞)) (∀n ∈ N) (∃!y ∈ h0, ∞)) y n = x.

Theorem 1.8 (Density of Q and R \ Q in R). Let a, b ∈ R, a < b. Then

(∃q ∈ Q) a < q < b,

(∃j ∈ R \ Q) a < j < b.
 7

II. SEQUENCES. LIMITS

Definition 2.1. An assignment n 7→ an , where an ∈ R for each n ∈ N, is called
sequence, ve denote it {an }∞
n=1 or simply {an }; an is called n-th member of the sequence
{an }n=1 . Two sequences are equal, {an }∞
∞ ∞
n=1 = {bn }n=1 iff (∀n ∈ N) an = bn .

Definition 2.2. We say that a sequence {an }∞ n=1 is bounded from below (bounded from
above, bounded, respectively), if the corresponding set {an ; n ∈ N} is bounded from
below (bounded from above, bounded, resp.).
Definition 2.3 (Monotonicity). We say that {an }∞
n=1 is
• increasing, if (∀n ∈ N) an < an+1
• decreasing, if (∀n ∈ N) an+1 < an
• nonincreasing, if (∀n ∈ N) an+1 ≤ an
• nondecreasing, if (∀n ∈ N) an ≤ an+1
• monotone, if it is nonincreasing or nondecreasing
• strictly monotone, if it is increasing or decreasing

Examples.
 ∞
(1) n1 n=1 = 1, 12 , 31 , 14 , . . . is decreasing.

(2) {an }∞
n=1 , where an = n for each n ∈ N, is increasing.
(3) Fibonacci sequence (0, 1, 1, 2, 3, 5, 8, 13, . . . ), i.e. a1 = 0, a2 = 1, an+2 = an +
an+1 , is nondecreasing. It is a subsequence (see Definition 2.5 below) of the
previous sequence.
(4) {(−1)n } = (−1, 1, −1, 1, −1, . . . ) is not monotone.

Definition 2.4. We say that a sequence {an }∞n=1 has limit A ∈ R (or converges to A),
if
(∀ε ∈ R, ε > 0) (∃n0 ∈ N) (∀n ∈ N, n ≥ n0 ) |an − A| < ε.
We shall denote this fact by

lim an = A, or briefly lim an = A, an −−−→ A, or an → A.

n→∞ n→∞

A sequence {an } is called convergent if it has a limit A ∈ R.

Example. A useful limit is
1
lim = 0.
n→∞ n
This proves from the definition of limit using Archimede Property (Theorem 1.6).
Theorem 2.1. Each sequence has at most one limit.
Theorem 2.2. Each convergent sequence is bounded.
Definition 2.5. Let {an }∞ ∞
n=1 be a sequence of real numbers. We say that {bk }k=1 is a
subsequence of {an }∞ ∞
n=1 , if there is an increasing sequence {nk }k=1 of natural numbers
such that (∀k ∈ N) bk = ank .
8

Theorem 2.3 (Subsequences preserve limits). Let {an }∞ ∞

n=1 be a sequence, {bk }k=1
its subsequence, limn→∞ an = A. Then limk→∞ bk = A.
Example. {(−1)n } has no limits. Indeed, its subsequence {(−1)2k }∞k=1 = (1, 1, 1, . . . )
2k+1 ∞
has limit 1, while another subsequence {(−1) }k=1 = (−1, −1, −1, . . . ) converges to
−1.
Definition 2.6. For sequences {an }, {bn } and a constant λ ∈ R, we define
• {an } + {bn } = {an + bn }
• {an } · {bn } = {an · bn }
• λ · {an } = {λ · an } n o
{an }
• if (∀n ∈ N) bn 6= 0, then {bn } = abnn

Theorem 2.4 (Arithmetics of limits). Let lim an = A ∈ R, lim bn = B ∈ R. Then

(1) lim(an + bn ) = A + B
(2) lim(an · bn ) = A · B
(3) if, moreover, bn 6= 0 for each n and B 6= 0 then lim abnn = B
A

Definition 2.7. For a ∈ R the absolute value of a defines

a if a ≥ 0,

|a| =
−a if a < 0.

Proposition 2.5 (The Triangle Inequality). For every x, y ∈ R,

|x + y| ≤ |x| + |y|.

Example.
5 11
3n2 − 5n + 11 3− n + n2 3−0+0 3
lim 2
= lim 4 7 = = ,
n→∞ 2n + 4n − 7 n→∞ 2 + − 2+0−0 2
n n2

because lim n1 = 0 and arithmetics of limits can be applied.

Theorem 2.6. Let an → 0, let {bn } be a bounded sequence. Then an · bn → 0.
Theorem 2.7 (Limits preserve ordering). Let lim an = A ∈ R and lim bn = B ∈ R.
(i) Suppose there exists n0 ∈ N such that an ≤ bn for each n ≥ n0 . Then A ≤ B.
(ii) Let A < B. Then there exists n0 ∈ N such that an < bn for each n ≥ n0 .
Example.
 1 ≤1 cannot be replaced by < in the previous theorem: consider the sequences
− n and n – they have common limit 0.
Theorem 2.8 (Two Policemen or Sandwich Theorem). Let {an }, {bn } be con-
vergent sequences and let {cn } be a sequence such that
(i) (∃n0 ∈ N) (∀n ≥ n0 ) an ≤ cn ≤ bn and
(ii) lim an = lim bn = A.
Then {cn } is convergent and lim cn = A.
 9

Definition 2.8. We say that a sequence {an } has limit +∞ if

(∀L ∈ R) (∃n0 ∈ N) (∀n ∈ N) an > L.
We say that a sequence {an } has limit −∞ if
(∀K ∈ R) (∃n0 ∈ N) (∀n ∈ N) an < K.
The structure of the real line now extends by adding two limit elements, +∞ and
−∞:
R∗ = R ∪ {−∞, +∞}.
We have to define ordering < and operations + and · on this extended structure.
• (∀a ∈ R) − ∞ < a < +∞
• (∀a ∈ R) ± ∞ + a = a + ±∞ = ±∞ 
 +∞ if a > 0
• (∀a ∈ R∗ \ {0}) + ∞ · a = a · +∞ =
−∞ if a < 0
−∞ if a > 0 

& −∞ · a = a · −∞ =
+∞ if a < 0
• ±∞ + ±∞ = ±∞
a
• (∀a ∈ R) ±∞ =0
Important remark. The following expressions are not defined:
a ∞
”∞ − ∞”, ”0 · ±∞”, ” ”, ” ”.
0 ∞
Practically, it means that if arithmetics of limits produces such an expressesion, other
methods have to be used to compute the limit, including, of course, rearrangement of
the original expression.
√ √
Example. limn→∞ n + 1 − n leads to ”∞ − ∞” at the first sight (because n + 1
and n as well as their square roots have limit +∞). Let us rearrange the problem:
√ √
√ √ √ √
n+1+ n n+1−n 1
n+1− n= n+1− n · √ √ =√ √ =√ √ → 0,
n+1+ n n+1+ n n+1+ n
√ √
because n + 1 + n → +∞.
Theorem 2.9 (Arithmetics of limits). Let lim an = A ∈ R∗ and lim bn = B ∈ R∗ .
Then
(i) lim(an + bn ) = A + B, if the right side is defined,
(ii) lim(an · bn ) = A · B, if the right side is defined,
(iii) lim abnn = B
A
, if the right side is defined.
Remark. Each sequence of real numbers has at most one limit in R∗ . Limits preserve
≤ in R∗ and an obvious modification of Sandwich Theorem holds.
Theorem 2.10. Each monotone sequence has a limit.

n
Example. lim 1 + n1 defines e ≈ 2.71. To see that the sequence is convergent, it
suffices to show it is increasing and bounded from above; both facts are non-obvious.
Detailed comment later.
Theorem 2.11 (Bolzano–Weierstrass). Each bounded sequence contains a conver-
gent subsequence.
10

III. FUNCTIONS
Definition 3.1. Let A and B be nonempty sets. A mapping is an assignment

f : A → B,
x 7→ f (x),

such that (∀x ∈ A) (∃!y ∈ B) y = f (x).

Definition 3.2. Let f : A → B be a mapping. Its domain defines as Df = A, its range
Rf = {f (x); x ∈ A}. For X ⊂ A, image of X is f [X] = {f (x); x ∈ A}, for Y ⊂ B,
preimage of Y equals f −1 [Y ] = {x ∈ A; (∃y ∈ Y )f (x) = y}.
The graph of f is defined as Gf = {[x, y] ∈ A × B; y = f (x)}.
Definition 3.3. A mapping f : A → B is onto if Rf = B. It is injective or one-to-one
if

(∀x1 , x2 ∈ A) f (x1 ) = f (x2 ) ⇒ x1 = x2 .
A mapping is called bijective if it is injective and onto.
Let f : A → B, g : B → C be mappings. The symbol g ◦ f stands for their
composition, i.e. a mapping from A to C defined by

(g ◦ f )(x) = g(f (x)), x ∈ A.

Let f : A → B be injective and onto. Inversion mapping f −1 : B → A is defined by

f −1 (y) = x, where x satisfies f (x) = y.
Definition 3.4. A mapping f is a function of one real variable (a function for short)
if f : M → R, where M ⊂ R.
Definition 3.5. A function f : J → R is increasing on an interval J, if for each pair
x1 , x2 ∈ J, x1 < x2 , the inequality f (x1 ) < f (x2 ) holds. The notions of decreasing,
nondecreasing, nonincreasing functions are defined in an analogous way.
By monotone function (strictly monotone function, respectively) on the interval J
we mean a function, which is nondecreasing or nonincreasing (increasing or decreasing
respectively) on J.
Definition 3.6. We say that a function f : Df → R is
• odd, if for each x ∈ Df , −x ∈ Df and f (−x) = −f (x),
• even, if for each x ∈ Df , −x ∈ Df and f (−x) = f (x).

Definition 3.7. A function f : Df → R is called periodic with period a ∈ R, a > 0, if

for each x ∈ Df , x + a ∈ Df , x − a ∈ Df and f (x + a) = f (x − a) = f (x).
Examples.
(1) Every f : x 7→ x2n with n ∈ N is even on R. Each f : x 7→ x2n+1 is odd.
(2) Functions sin and cos are periodic on R,their period is 2π, but also 4π, 6π etc.
(3) tg is periodic on its domain, i.e. on R \ π2 + kπ; k ∈ Z .
(4) Constant function f : x 7→ c with c ∈ R is periodic with any period a ∈ R.
 11

Remark. If for a periodic function the smallest period a ∈ R exists (which is the case
of sin, cos with 2π and tg with π), then such an a is sometimes called primitive period.
Definition 3.8. Let f be a function, M ⊂ Df . We say that f is
• bounded from above on M , if
(∃K ∈ R) (∀x ∈ M ) f (x) ≤ K,
• bounded from below on M, if
(∃K ∈ R) (∀x ∈ M ) f (x) ≥ K,
• bounded on M, if
(∃K > 0) (∀x ∈ M ) |f (x)| ≤ K,
• constant on M, if f (x) = f (y) for each x, y ∈ M .
Definition 3.9. Let c ∈ R and let ε > 0. We define
• Bε (c) = (c − ε, c + ε) (open) neighborhood of c,
• Pε (c) = Bε (c) \ {c} punctured neighborhood of c,
• Pε (+∞) = Bε (+∞) = (1/ε, +∞) neighborhood and punctured neighborhood of
+∞,
• Pε (−∞) = Bε (−∞) = (−∞, −1/ε) neighborhood and punctured neighborhood
of −∞.
Definition 3.10. We say that A ∈ R∗ is a limit of function f at the point c ∈ R∗ if
(∀ε > 0) (∃δ > 0) (∀x ∈ Pδ (c)) f (x) ∈ Bε (A)
and denote this fact by limx→c f (x) = A.
Remark. Notice that for δ in the previous definition, Pδ (c) ⊂ Df , i.e. f is defined on
some punctured neighbourhood of c.
Definition 3.11. Let c ∈ R, ε > 0. We define
• Bε+ (c) = hc, c + ε) right neighbourhood of c,
• Bε− (c) = (c − ε, ci left neighbourhood of c,
• Pε+ (c) = (c, c + ε) right punctured neighbourhood of c,
• Pε− (c) = (c − ε, c) left punctured neighbourhood of c.
Further, we put
• Bε− (+∞) = Pε− (+∞) = Bε (+∞),
• Bε+ (−∞) = Pε+ (−∞) = Bε (−∞).
Definition 3.12. Let A ∈ R∗ , c ∈ R ∪ {−∞}. We say that A is limit from the right of
a function f at c if
(∀ε > 0) (∃δ > 0) (∀x ∈ Pδ+ (c)) f (x) ∈ Bε (A)
and denote it limx→c+ f (x) = A. Similarly, limit from the left, limx→c− f (x), defines for
c ∈ R ∪ {+∞}.
12

Example.
1 1 1
lim = +∞, lim = −∞, lim does not exist.
x→0+ x x→0− x x→0 x

Definition 3.13. We say that a function f is continuous at c ∈ R if limx→c f (x) = f (c).

Definition 3.14. We say that a function f is continuous at c ∈ R from the right (from
the left, respectively) if limx→c+ f (x) = f (c) (limx→c− f (x) = f (c), resp.).
√
Example. f (x) = x : h0, +∞) → R is continuous at each x ∈ R+ (why?) and
continuous from the right at x = 0.
Theorem 3.1. Let c ∈ R∗ . Each function has at most one limit at c.
Proof. By contradiction: assume two different values a, b satisfy the definition of limit
of f at c. If both a, b ∈ R, put ε = |b−a| 3 . Notice that Bε (a) ∩ Bε (b) = ∅ then. If some
of a, b is infinite, it is still easy to find ε > 0 such that Bε (a) ∩ Bε (b) = ∅. Since a is a
limit of f at c, there exists δ1 > 0 such that

(∀x ∈ Pδ1 (c)) f (x) ∈ Bε (a).

And since b is a limit of f at c, there exists δ2 > 0 such that

(∀x ∈ Pδ2 (c)) f (x) ∈ Bε (b).

Let δ = min{δ1 , δ2 } and take arbitrary x ∈ Pδ (c). For such x, f (x) ∈ Bε (a) ∩ Bε (b) – a
contradiction. 
Theorem 3.2. Suppose that a function f has a proper limit at c ∈ R∗ (i.e., limx→c f (x) ∈
R). Then there exists δ > 0 such that f is bounded on Pδ (c).
Proof. Denote limx→c f (x) = A ∈ R and put ε = 1. According to the definition of limit,
there is δ > 0 such that

(∀x ∈ Pδ (c)) f (x) ∈ Bε (A) = (A − 1, A + 1).

Hence A − 1 ∈ R is a lower bound and A + 1 ∈ R is an upper bound of f on Pδ (c). 

Definition 3.15. Let J ⊂ R be an interval. We say that f : J → R is continuous on
J if
(1) it is continuous at every interior point of J,
(2) it is continuous from the left at the right endpoint of J, if it belongs to J,
(3) it is continuous from the right at the left endpoint of J, if it belongs to J.

Theorem 3.3 (Arithmetics of Limits). Let c ∈ R∗ , let limx→c f (x) = A ∈ R∗ ,

limx→c g(x) = B ∈ R∗ . Then
(1) limx→c (f + g)(x) = A + B, if A + B is defined,
 ·g)(x) = A · B, if A · B is defined,
(2) limx→c (f
(3) limx→c fg (x) = B A A
, if B is defined (in particular, if B 6= 0).
 13

Remark. An analogous theorem holds for limx→c+ and limx→c− .

sin x
Example. We shall see later that limx→0 x = 1. Let us apply this fact to compute
limx→0 1−cos
x2
x
.
We rearrange the expression:
1 − cos x 1 − cos x 1 + cos x 1 − cos2 x sin2 x 1
2
= 2
· = 2 = 2
· .
x x 1 + cos x x · (1 + cos x) x 1 + cos x
The first function has limit 12 = 1 at 0 (we use arithmetics of limit), while the second
one is continuous at 0, so it suffices to put x = 0 there. Hence limx→0 1−cos
x2
x
= 12 .
Proposition 3.4. Let c ∈ R∗ . Let g > 0 on some Pδ (c) and limx→c g(x) = 0. Further,
(x)
let limx→c f (x) = A > 0, A ∈ R∗ . Then limx→c fg(x) = +∞.
Theorem 3.5 (Limits and Inequalities). Let c ∈ R, let f , g, h be functions.
(1) Let limx→c f (x) > limx→c g(x). Then there exists a punctured neighbourhood
Pδ (c) such that (∀x ∈ Pδ (c)) f (x) > g(x).
(2) Let f (x) ≤ g(x) on Pδ (c), let limx→c f (x) and limx→c g(x) exist. Then limx→c f (x) ≤
limx→c g(x).
(3) (Sandwich Theorem) Let (∀x ∈ Pδ (c)) f (x) ≤ h(x) ≤ g(x). Suppose that
limx→c f (x) = limx→c g(x). Then limx→c h(x) exists and is equal to limx→c f (x).
Remark. The same theorems hold for limx→c+ and limx→c− .
Theorem 3.6 (Limit of Composition). Let c, D, A ∈ R∗ , limx→c g(x) = D,
limy→D f (y) = A and at least one of the following conditions is satisfied
(1) (∃η > 0) (∀x ∈ Pη (c)) g(x) 6= D or
(2) f is continuous at D.
Then limx→c f (g(x)) = A.
Remark. The first conditions says that the inner function does not meet its limit on
some punctured neighbourhood of c. The second condition expresses continuity of the
outer function at the respective point.
Example. The signum function returns the sign of a given real number:

 −1 for x < 0,

sgn x = 0 for x = 0,

1 for x > 0.


It is bounded and monotone, but neither connected at 0 from the left nor from the right.
Example. If none of the conditions of Theorem 3.6 is satisfied, then the limit of the
composed function need not be as expected. We consider c = D = 0, A = 1. Let
0 for x 6= 0,

g(x) =
1 for x = 0,
then limx→0 g(x) = 0, but (1) fails – to the contrary, g(x) ≡ 0 on every Pη (0). Let f (x) =
|sqn x| – a function discontinuous at 0. Now, limy→0 f (y) = 1, while limx→0 f (g(x)) =
limx→0 f (0) = 0.
14

Theorem 3.7 (Heine). Let A, c ∈ R∗ , let f be defined on Pδ (c). Then the following
are equivalent:
(1) limx→c f (x) = A,

(2) (∀xn ∈ Df , xn 6= c) xn → c ⇒ f (xn ) → A .

Example. The function sin x has no limit at +∞: consider sequences {an } and {bn }
with an = nπ and bn = π/2 + 2nπ (n ∈ N). Then an → +∞ and bn → +∞, sin an → 0
and sin bn → 1. If there was a limit limx→+∞ sin x, then the two sequences {sin an } and
{sin bn } would have the same limit equal to limx→+∞ sin x.
We can derive that there is no limit of sin x1 in 0 either.
Example on limit of composition. Let f be continuous at 0. Then

lim f (1/x) = f (0).

x→+∞

How Theorem 3.6 applies here? We find c = +∞, D = 0 = limx→+∞ x1 , A = f (0). The
outer function f is continuous at D as well as the inner function x1 does not reach its
limit 0 on any Pδ (+∞).
Remark. On computation of limits of composed functions, conditions of Theorem 3.6
must always be verified.
Example. We shall compute
sin(x − π6 )
limπ
x→ 6 x − π6
using the ’tabular’ limit
sin x
lim = 1.
x→0 x

Here, the outer function, f (y) = siny y is not continuous at 0! Further, c = π6 , g(x) =
x − π6 , D = limx→ π6 g(x) = 0, A = limx→0 sinx x = 1. We have to verify the second
condition. Indeed, g(x) 6= 0 outside π6 , i.e. on any Pδ ( π6 ), because g is one-to-one.
Theorem 3.8 (Limit of monotone function). Let f be monotone on (a, b), a,
b ∈ R∗ . Then there exists limx→a+ f (x) and limx→b− f (x).
In the sequel, we shall deal with functions continuous on intervals.
Theorem 3.9 (Bolzano, Darboux). Let f be a continuous function on ha, bi, a,
b ∈ R, f (a) < f (b). Then for every c ∈ (f (a), f (b)) there exists ξ ∈ (a, b) such that
f (ξ) = c.
Theorem 3.10. Let J be an interval, let f : J → R be continuous on J. Then f [J] is
an interval or a one-point set.
The key is to prove the following lemma.
Lemma 3.11. Let ∅ 6= C ⊂ R be convex, i.e. a, b ∈ C, a < c < b imply c ∈ C. Then
C is an interval or a one-point set.
 15

Theorem 3.12. Let f be a continuous function on an interval ha, bi. Then f is bounded
on ha, bi.

The proof (by contradiction) uses Bolzano-Weierstrass Theorem.

Definition 3.16. Let M ⊂ R, x ∈ M and let M ⊂ Df for a function f . We say that

f attains at x its
• maximum on M (and denote f (x) = maxM f ) if (∀y ∈ M ) f (y) ≤ f (x),
• minimum on M (and denote f (x) = minM f ) if (∀y ∈ M ) f (y) ≥ f (x).
The point x is called point of maximum of f (point of minimum of f , respectively).

Definition 3.17 = 3.16’. Let M ⊂ R, x ∈ M and let M ⊂ Df for a function f . We

say that f attains at x
• local maximum with respect to M if there exists δ > 0 such that (∀y ∈ Pδ (x) ∩
M ) f (y) ≤ f (x),
• local minimum with respect to M if there exists δ > 0 such that (∀y ∈ Pδ (x) ∩
M ) f (y) ≥ f (x).

Examples.
(1) sin x attains (local) maximum at every π2 + 2kπ (k ∈ Z), (local) minimum at
− π2 + 2kπ (k ∈ Z).
(2) A typical local maximum which is not maximum is attained at 0 by the function
|x| − 1 (draw the graph!).

Theorem 3.13. Let f be continuous on ha, bi. Then f attains its maximum and min-
imum on ha, bi.

Remark. Notice that the requirement on the interval being closed is essential: there
are functions that map a bounded open interval onto the whole R.

Theorem 3.14. Let f be an increasing continuous function on an interval J. Then

f −1 is continuous and increasing on f [J].

Elementary functions.

Theorem 3.15 + Definition. There exists a unique function logarithm (log) with the
following properties:
(L1) Dlog = R+ = (0, +∞)

and log is increasing on (0, +∞),
(L2) ∀x, y ∈ (0, +∞) log(x · y) = log x + log y,
(L3) limx→1 log x
x−1 = 1.

Definition 3.17. Exponential function (x 7→ exp(x) or ex ) is defined as the the inverse

function to log.

Definition 3.18. Let a, b ∈ R, a > 0. The number ab is defined as ab = exp(b · log a).
16

Remarks.
(1) Is the last definition correct? I.e., is an = exp(n · log a), a−n = exp(−n · log a),
1
a n = exp n1 · log a for every n ∈ N?

(2) What are Dexp and further properties of the function? To answer (1) and (2),
we have to prove more about log.
Proposition 3.16 (Further properties of log).
(1) log 1 = 0,
(2) (∀x ∈ (0, +∞)) log x1 = − log x,

(3) (∀x ∈ (0, +∞)) (∀n ∈ Z) log (xn ) = n · log x,

(4) limx→+∞ log x = +∞, limx→0+ log x = −∞,
(5) log is continuous on (0, +∞),
(6) Rlog = R.
Proposition 3.17 (Properties of exp).
(1) Dexp = R, Rexp = (0, +∞),
(2) exp is increasing on R,
(3) exp is continuous on R, limx→+∞ exp x = +∞, limx→−∞ exp x = 0,
(4) exp 0 = 1,
(5) (∀x, y ∈ R) exp(x + y) = exp x · exp y,
(6) limx→0 expxx−1 = 1.
Definition 3.19. e is the unique number such that log e = 1.
1 n


Theorem 3.18. The number e is irrational, e=2.71828,
˙ e = limn→∞ 1 + n .
log x
Definition 3.20. Let a > 0, a 6= 1. Logarithm of x to the base a defines as loga x = log a
for every x ∈ (0, +∞).
Theorem 3.19 + Definition. There exists a unique π > 0, π ∈ R and a unique
function sine (sin) such that
(S1) Dsin = R,
(S2) sin is increasing on h− π2 , π2 i,
(S3) sin 0 = 0,
(S4) (∀x, y ∈ R) sin(x + y) = sin x · sin π2 − y + sin π2 − x · sin y,

(S5) limx→0 sinx x = 1.

Proposition 3.20 (Further properties of sin).
(1) sin π2 = 1,
(2) (∀x ∈ R) | sin x| ≤ 1,
(3) sin is continuous on R,
(4) sin is odd and 2π-periodic.
Definition 3.21 (Further trigonometric functions). The following functions are
defined:
• cosine: cos x = sin π2 − x , x ∈ R,


sin x π
• tangent: tg x = cos x , x ∈ R \ {(2k + 1) 2 ; k ∈ Z},
• cotangent: cotg x = cos x
sin x , x ∈ R \ {kπ; k ∈ Z}.
 17

Definition 3.22 (Cyclometric functions). The symbol ↾ stands for restriction of

a particular function to a given domain.
−1
• arcsine: arcsin = (sin ↾ h−π/2, π/2i) ,
−1
• arccosine: arccos = (cos ↾ h0, πi) ,
−1
• arctangent: arctg = (tg ↾ (−π/2, π/2)) ,
−1
• arccotangent: arccotg = (cotg ↾ (0, π)) .

Remark. It is easy to see that Darcsin = Darccos = h−1, 1i and Darctg = Darccotg = R.
Proposition 3.21. All the functions defined in 3.21–3.22 are continuous on their do-
mains.

Derivative.
Definition 3.23. Let f be a real function, a ∈ R. Derivative of f at a is defined by
the formula
f (a + h) − f (a)
f ′ (a) = lim ,
h→0 h
if the limit on the right exists.
Remark. Existence of derivative in a involves the fact that f is defined not only in a
but on some neighbourhood Bδ (a).
Definition 3.24. Let f be a real function, a ∈ R. Derivative of f at a from the right
(from the left, respectively) is defined by the formula
 
′ f (a + h) − f (a) ′ f (a + h) − f (a)
f+ (a) = lim f− (a) = lim , resp. ,
h→0+ h h→0− h

if the limits on the right exist.

Examples.
(1) Let us compute f ′ (x) for f (x) = xn with n ∈ N at given (but arbitrary) x ∈ R.

f (x + h) − f (x) (x + h)n − xn
f ′ (x) = lim = lim
h→0 h h→0 h
n−1
+ (x + h)n−2 · x + · · · + xn−1



(x + h) − x · (x + h)
= lim
h→0 h
n−1 n−2
· x + · · · + xn−1


= lim (x + h) + (x + h)
h→0
= n · xn−1 .

(2) For f (x) = sgn x (cf. Example after Theorem 3.6), f ′ (0) can be computed
directly as +∞. Notice that, then, sgn is a function discontinuous in 0, but
has a derivative there. The following theorem precises the relation between
continuity and existence of derivative.
18

Theorem 3.22. If a function f has a proper derivative in a point a (i.e. f ′ (a) ∈ R),
then f is continuous in a.
Remarks.
f (x)−f (a)
(1) Alternatively, derivative of f at a can be defined as f ′ (a) = limx→a x−a .
(2) f ′ (a) either exists and

is proper, i.e. f ′ (a) ∈ R,



is improper, i.e. f ′ (a) ∈ {+∞, −∞},

or does not exist.

(3) Geometric meaning of derivative f ′ (a) is the slope of the tangent line of the
graph of f in the point a.

Theorem 3.23 (Arithmetics of Derivatives). Let f , g have proper derivatives at

a ∈ R. Then
(i) (f + g)′ (a) = f ′ (a) + g ′ (a), (αf )′ (a) = α · f ′ (a) for every α ∈ R,
(ii) (f · g)′ (a) = f ′ (a) · g(a) + f (a) · g ′ (a),
′
(a)·g ′ (a)
(iii) if g(a) 6= 0 then (f /g)′ (a) = f (a)·g(a)−f g 2 (a) .

Derivatives of some elementary functions.

• log′ x = x1 , x ∈ (0, +∞)
• (ex )′ = ex , x ∈ R
• sin′ x = cos x, x ∈ R
• cos′ x = sin x, x ∈S R
tg′ x = cos12 x , x ∈ k∈Z kπ − π2 , kπ + π2


•
cotg′ x = − sin12 x , x ∈ k∈Z (kπ, (k + 1)π)
S
•
Theorem 3.24 (Derivative of Composed Function). Let x0 , y0 ∈ R, g(x0 ) = y0 ,
g ′ (x0 ) ∈ R, f ′ (y0 ) ∈ R. Then (f ◦ g)′ (x0 ) = f ′ (y0 ) · g ′ (x0 ).

• We have proved that (xn )′ = n · xn−1 , x ∈ R, for every n ∈ N (in particular,

const′ = 0). Let us extend this formula to arbitrary exponent α ∈ R. Notice
that we apply Theorem 3.24.

′
(xα )′ = exp(α · log x) = exp′ (α · log x) · α · log′ x
1 1
= exp(α · log x) · α · = xα · α · = α · xα−1
x x

for every x ∈ (0, +∞).

Theorem 3.25 (Derivative of Inverse Function). Let f be continuous and increas-
ing (decreasing, respectively) on an interval (a, b). Let f have a proper nonzero f ′ (x0 )
at x0 ∈ (a, b). Then f −1 has (f −1 )′ at y0 = f (x0 ) and

1
(f −1 )′ (y0 ) =
.
f′ f −1 (y 0)
 19

Theorem 3.25 applies in computing derivatives of cyclometric functions.

• arcsin′ x = √1−x
1
2
, x ∈ (−1, 1)
′ 1
• arccos x = − √1−x2 , x ∈ (−1, 1)
1
• arctg′ x = 1+x 2, x ∈ R
′ 1
• arccotg x = − 1+x 2, x ∈ R

Derivative and its relation to local extrema.

Theorem 3.26 (Necessary Condition for Local Extremum). Let x0 be a point
of local extremum of f . Then either f ′ (x0 ) does not exist or f ′ (x0 ) = 0.
Remark. Let f : ha, bi → R. The function can attain its maxima/minima on ha, bi at
(1) points a, b,
(2) points x0 ∈ (a, b) such that f ′ (x0 ) does not exist,
(3) points x0 ∈ (a, b) such that f ′ (x0 ) = 0 (Theorem 3.26).
Notice that contnuity of a function defined on a closed bounded interval ha, bi guar-
antees existence of points of maxima and minima on ha, bi (Theorem 3.13).

Deeper theorems on derivatives.

Theorem 3.27 (Rolle). Let a, b ∈ R, a < b. Let a function f satisfy
(i) f is continuous on ha, bi,
(ii) f has (proper or improper) derivative at every point of (a, b),
(iii) f (a) = f (b).
Then there is ξ ∈ (a, b) such that f ′ (ξ) = 0.
Theorem 3.28 (Lagrange). Let a, b ∈ R, a < b, let f be continuous on ha, bi and
have (proper or improper) derivative on (a, b). Then there is ξ ∈ (a, b) such that

f (b) − f (a)
f ′ (ξ) = .
b−a

Remarks.
(1) There can be more than one point ξ in Theorems 3.27, 3.28.
(2) Theorems 3.27, 3.28 are also referred to as Mean Value Theorems.

Definition 3.25. For interval J with endpoints a, b ∈ R∗ , a < b, we denote int J its
interior, i.e. int J = (a, b).
Theorem 3.29 (Monotonicity and the Sign of the Derivative). Let J ⊂ R be
an interval, f continuous on J and let f ′ exist at each point of int J. Then
(1) if f ′ (x) > 0 for each x ∈ int J then f is increasing on J,
(2) if f ′ (x) < 0 for each x ∈ int J then f is decreasing on J,
(3) if f ′ (x) ≥ 0 for each x ∈ int J then f is nondecreasing on J,
(4) if f ′ (x) ≤ 0 for each x ∈ int J then f is nonincreasing on J.
20

Theorem 3.30 (l’Hospital Rule). Let f , g have proper derivatives f ′ , g ′ on some

′
Pδ (a), a ∈ R∗ , and suppose limx→a fg′ exists. If, moreover,
(i) limx→a f (x) = limx→a g(x) = 0 or
(ii) limx→a |g(x)| = +∞,
f f′
then limx→a g exists and is equal to limx→a g′ .

Remarks.
(1) Conditions (i), (ii) of Theorem 3.30 must be verified before use of l’Hospital
Rule. Always write that the limit is, e.g., ’of the type 00 ’ or ’of the type ∞
∞ ’.
Generally, the rule does not hold for other values of limx→a f , limx→a g – try to
find a counterexample.
(2) The Rule can be used repeatedly, e.g.

ex ex ex ex
lim = lim = lim = lim = +∞.
x→+∞ x3 x→+∞ 3x2 x→+∞ 6x x→+∞ 6

∞
We apply l’Hospital Rule for ’limit of the type ∞ ’ three times here.
′ ′
(3) Computing with derivatives f , g instead of f , g does not always ease the situa-
tion, e.g., if derivative of product or composed functions occurs in the numerator
and/or denominator of the expression.

Theorem 3.31. Let f be continuous from the right at a ∈ R and let limx→a+ f ′ (x)
′
exist. Then f+ (a) exists and
′
f+ (a) = lim f ′ (x).
x→a+

Similarly from the left.

Example. The function arcsin is continuous on h−1, 1i, in particular, continuous from
the right at −1 and from the left at 1.

1
lim arcsin′ x = lim √ = +∞ = lim arcsin′ x,
x→−1+ x→−1+ 1 − x2 x→1−

hence arcsin′+ (−1) = arcsin′− (1) = +∞.

Definition 3.26. Let n ∈ N, a ∈ R, let f have proper n-th derivative f (n) on a

neighbourhood of a. Then (n + 1)-st derivative of f at a is defined by

′ f (n) (a + h) − f (n) (a)

f (n+1) (a) = f (n) (a) = lim .
h→0 h

Formally, we put f (0) = f . Small order derivatives have duplicit notation: f ′ = f (1) ,
f ′′ = f (2) , f ′′′ = f (3) .

Convex and Concave Functions.

 21

Definition 3.27. Let f have a proper first derivative at a ∈ R. We call the set

Ta = [x, y] ∈ R; y = f (a) + f ′ (a) · (x − a)



tangent line to the graph of f at [a, f (a)].

We say that [x, f (x)] is
• below the tangent line Ta if f (x) < f (a) + f (a) + f ′ (a) · (x − a),
• above the tangent line Ta if f (x) > f (a) + f (a) + f ′ (a) · (x − a).

Definition 3.28. Let f ′ (a) ∈ R. We say that a is an inflection point of f if there exists
a ∆ > 0 such that


(i) ∀x ∈ (a − ∆, a)
[x, f (x)] is below Ta and
(ii) ∀x ∈ (a, a + ∆) [x, f (x)] is above Ta
or


(i) ∀x ∈ (a − ∆, a)
[x, f (x)] is above Ta and
(ii) ∀x ∈ (a, a + ∆) [x, f (x)] is below Ta .

Theorem 3.32. Let a ∈ R be an inflection point of f . Then f ′′ (a) does not exist or is
equal to 0.
Remarks.
(1) (Analogy to search for extrema.) Let f have proper derivative everywhere on
(a, b). Then inflection points of f on (a, b) are points c at which either f ′′ (c)
does not exist or f ′′ (c) = 0.
(2) f ′′ (c) = 0 does not imply c is an inflection point of f – consider, e.g., f (x) = x4 ,
c = 0. Here f ′′ (c) = 0 but all the graph is above the tangent line which is the
x-axis in this case.

Theorem 3.33. Let f have a continuous derivative on (a, b) and x0 ∈ (a, b). Suppose
that (∀x ∈ (a, x0 )) f ′′ (x) > 0 and (∀x ∈ (x0 , b)) f ′′ (x) < 0. Then x0 is an inflection
point of f .
Definition 3.29. Let I be an interval. We say that f is


• convex on I if (∀x1 , x2 ∈ I) (∀λ ∈ h0, 1i) f λx1 + (1 − λ)x2 ≤ λf (x1 ) + (1 −
λ)f (x2 )

• concave on I if (∀x1 , x2 ∈ I) (∀λ ∈ h0, 1i) f λx1 + (1 − λ)x2 ≥ λf (x1 ) + (1 −
λ)f (x2 )


• strictly convex on I if (∀x1 , x2 ∈ I, x1 6= x2 ) ∀λ ∈ (0, 1) f λx1 + (1 − λ)x2 <
λf (x1 ) + (1 − λ)f (x2 )


• strictly concave on I if (∀x1 , x2 ∈ I, x1 6= x2 ) ∀λ ∈ (0, 1) f λx1 + (1 − λ)x2 >
λf (x1 ) + (1 − λ)f (x2 )

Remark. λx1 + (1 − λ)x2 with λ ∈ h0, 1i expresses a typical element of the segment
connecting points x1 and x2 .
22

Examples. Natural logarithm log is strictly concave on (0, +∞), exp is strictly convex
on R. This is obvious from the shapes of the graphs but computation and estimates
using definition might be difficult. The following theorem gives an easy criterion of
convexity/concaveness for a class of functions.
Theorem 3.34 (Second Derivative and Convexity). Let f have a proper second
derivative f ′′ on (a, b), a < b.
(1) If f ′′ (x) > 0 for every x ∈ (a, b) then f is strictly convex on (a, b).
(2) If f ′′ (x) < 0 for every x ∈ (a, b) then f is strictly concave on (a, b).
(3) If f ′′ (x) ≥ 0 for every x ∈ (a, b) then f is convex on (a, b).
(4) If f ′′ (x) ≤ 0 for every x ∈ (a, b) then f is concave on (a, b).

Example. For f (x) = log x is f ′ (x) = x1 , f ′′ (x) = − x12 which is negative on all the
domain of log. It follows that log is indeed strictly concave on (0, +∞).
Definition 3.30. We say that a function x 7→ ax + b, a, b ∈ R, is asymptote of f at
+∞ (at −∞, resp.) if




lim f (x) − ax − b = 0 lim f (x) − ax − b = 0, resp. .
x→+∞ x→−∞

Theorem 3.35. A function f has asymptote x 7→ ax + b in +∞ if and only if

f (x)

lim = a ∈ R and lim f (x) − ax = b ∈ R.
x→+∞ x x→+∞

Remarks.
(1) Analogous theorem holds for x → −∞.
(2) Theorem 3.35 describes the way to compute parameters a, b of an asymptote
(or to show that a function has no asymptote).

Investigation of a function f .
(1) Determine the domain Df and the set of all points of continuity of f .
(2) Find out if the function is odd, even or periodic.
(3) Compute limits at all endpoints of Df (if Df is a union of intervals there may
be more than two limits to investigate).
(4) Compute the first derivative f ′ in all points in which it exists, including deriva-
tives from the right/left in x ∈ Df in which f ′ (x) does not exist. Use f ′ to
find intervals of monotonicity of f , its local and global maxima/minima and the
range Rf .
(5) Compute the second derivative f ′′ in all points in which it exists. Use it to find
intervals of convexity/concaveness of f and inflection points.
(6) Find asymptotes at ±∞ if they exist.
(7) Draw the graph of f . It may involve further computation, e.g. f (x) at important
points, f ′ (x) at inflection points etc.