Functional Equations PDF

Functional Equation
Pang-Cheng, Wu
September 9, 2018
Contents
Preface 5
0 Preliminaries 7
0.1 Notations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1 Basic Concepts 9
1.1 What are functional equations ? . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2 Assertion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3 Injectivity and Surjectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4 Cauchy’s Functional Equation . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.5 Establishing Inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.6 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.6.1 2014 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.6.2 2015 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.6.3 2016 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.6.4 2017 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2 Cauchy FE 27
2.1 Taiwanese Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2 Variation and Extension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.4 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3 Double Counting 39
3.1 Switching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2 Three variables method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.3 Double Counting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.4 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3
4 Calculus (Supplement) 51
4.1 Do You Know How to Count? . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.2 Construction of Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.3 Limit and Continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.4 Differentiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.5 Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.6 Ordinary Differential Equation . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.7 Lebesgue Monotone Differentiable Theorem . . . . . . . . . . . . . . . . . . 63
4.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5 Functional Equation in N/Z 67

5.1 Mathematical Induction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.2 Number Theory Facts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.2.1 About Divisibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.2.2 Euler Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.2.3 Perhaps Useful . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5.3 Recursive Chain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.4 v2 method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6 Polynomial 85
6.1 Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.2 Factorization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
6.3 Irreducibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.4 Polynomial Functional Equation . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.4.1 Naive Method: Compare the Coefficients . . . . . . . . . . . . . . . . 91
6.4.2 Bounding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.4.3 Taiwanese Transformation . . . . . . . . . . . . . . . . . . . . . . . . 92
6.5 n − m | P (n) − P (m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
7 Appendix 95
7.1 A. Removed Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
7.2 B. Removed Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Preface
In the past, I learned a lot from AoPS. Many users generously shared their solutions and
even the ideas. They also kindly helped me to resolve my doubts. I hope to be such a user,
so I began this project: LaTeXing a handout about functional equation.
I referred some handouts (not only about FE) and I found out a common phenomenon: full
of solutions, no motivations. As a past contestant of IMO, I think it’s much important to
learn the way of thinking. Therefore, I tried my best to illustrate my ideas.
Of course, the goal of this handout is to investigate common and useful techniques for
solving functional equations. In chapter 1, we briefly introduce the basic skills: substitution,
injective, surjective, inequality; In the next 3 chapters, we’ll study special methods for R+
functional equations. After, the readers may have the ability to bash R+ FEs by Three
Variables Method or using calculus. Then, we back to functional equations with discrete
domains: Z, N and polynomial functional equations. I deeply grateful to and would like to
thank my friends, especially
• clavin9403: Correc typos and give some advises.
• USJL: Suggest suitable examples and exercises.
• YaWNeeT: Complete chapter 4 (I’m not familiar with this method, in fact).
Pang-Cheng, Wu
September 9, 2018
5
Chapter 0
Preliminaries
0.1 Notations
• N : the set of all positive integers
• Z: the set of all integers
• Q: the set of all rational numbers
• R: the set of all real numbers
• N0 : defined by N0 = N ∪ {0}
• R+ : the set of all positive reals
7
Chapter 1
Basic Concepts
Solving functional equations is a main part in Olympiad Algebra. We will first introduce
the basic ideas. For example, apply injectivity to reduce the equation. Or establish some
inequalities to find the value of given point. Then, we’re going to solve problems in past
contests. Finally, we’ll learn supplement materials, which are used to bash.
1.1 What are functional equations ?

We begin with a system of equations
2x − 5y = 8
3x + 9y = −12
This is a special case of functional equations. To be precise, we translate the question to
Example 1. Find all functions f : {1, 2} → R such that
2f (1) − 5f (2) = 8
3f (1) + 9f (2) = −12
In fact, when the domain of f is a finite subset A of N, it’s just a system of equations
with variables {f (a)}a∈A . So, it’s nature to consider finding all functions with particular
properties as a generalization. Here is an example,
Example 2. Find all functions f : R → R such that
2f (x) − 5f (y) = 8, ∀x, y ∈ R
9
As dealing with systems of equations, we want to choose several suitable equations and
solve. For instance, for f (0), the simplest identity is
2f (0) − 5f (0) = 8
which gives f (0) = − 83 .
f (x) + xf (1 − x) = x
for all x ∈ R
Solution. For any a, the identities in which f (a) appears are
f (a) + af (1 − a) = a
f (1 − a) + (1 − a) f (a) = 1 − a
a2
Direct to get f (a) = 1−a+a2
. It’s over? No, we should verify our answer. As following:
x2 x (1 − x)2
f (x) + xf (1 − x) = +
1 − x + x2 1 − (1 − x) + (1 − x)2
x2 x − 2x2 + x3
= +
1 − x + x2 1 − x + x2
=x
Remark 1. The question, now, is why we need to check our answer. That’s because, in
most problems, we won’t find all identities in which a given variable appears. The value may
contradicts to other equations.
Remark 2. In fact, our central idea is:
• If f (a) is unknown for some a, choose an equation in which f (a) appears
• If the variables ( f (something) ) in the chosen equation are known, go to the next step.
Otherwise, back to Step 1.
• Solve the system of equations

1.2 Assertion
f (x − y) = f (x) + f (y) − 2xy
holds for all x, y ∈ R
Solution. There are infinite equations being the concern of a given variable. We must choose
some of them and solve it. Well, it’s clear that for all a ∈ R, we have
f (0) = 2f (a) − 2a2
Once we know f (0), we get f (a). In fact, the above equation also tells us f (0) = 0 (why?)
So the only possibility of the function is f (a) = a2 for all a. Indeed,
f (x − y) = (x − y)2 = x2 + y 2 − 2xy = f (x) + f (y) − 2xy
Note that what we have done is to choose some equations and solve systems of equations.
Equivalently, we assert values to variables. In the previous example, taking
f (0) = 2f (a) − 2a2
and asserting x = a, y = a are the same. As the result, I’ll say ”set x =?, y =?” instead
of ”choose the following equation”. Assertion is the easiest and most important step in
solving functional equations. Consequently, let’s do more examples. We’ll see that for most
problems, f (0) is important.
Example 5. (Komal) Determine all functions f : R → R such that
f (x + y) + f (x) f (y) = x2 y 2 + 2xy (1.1)
for all x, y ∈ R
Solution. Set y = 0 into (1.1):
(1 + f (0)) f (x) = 0
If f (0) 6= −1, then f (x) = 0 for all x ∈ R, which is absurd. So f (0) must be −1.
Now, we set (x, y) = (a, a) , (a, −a) , (2a, −a) to get
f (2a) + f (a)2 = a4 + 2a2
f (a) f (−a) − 1 = a4 − 2a2
f (a) + f (2a) f (−a) = 4a4 − 4a2
Those implies f (a) = a2 − 1, which is indeed a answer to (1.1)
In this problem, we have solved the system of equations. In many ”easy” functional
equations, solving the system of equations is the key-point.
Example 6. Find all functions f : N → N so that

p
3
f (x + y) = f (x) + f (y) + 3 (x + y) f (x) f (y)
holds for all x, y ∈ N
Solution. Notice that q

f (2x) = 2f (x) + 6x f (x)2
3
p
Therefore, we can find a function g : N → N defined by g (x) = 3
f (x) (why) Then g
satisfies the following equation
g (x + y)3 = g (x)3 + g (y)3 + 3 (x + y) g (x) g (y)
for all x, y ∈ N. If we know the values of g (1) and g (2), it’s easy to show g is the identity
map by mathematical induction. As the previous example suggested, when one encounters
the relation between h (x) , h (y) , h (xy) , h (x + y) for some function h, it’s quite useful to get
h (1) , h (2) by taking the following assertions:
(x, y) = (1, 1) , (1, 2) , (1, 3) , (2, 2)
The remains are left as an exercise.
Example 7. (2015 Baltic Way) Find all functions f : R → R satisfying for all reals x, y,
|x|f (y) + yf (x) = f (xy) + f x2 + f (f (y))

Solution. First, consider the assertion (x, y) = (1, 1), we get
2f (1) = 2f (1) + f (f (1)) → f (f (1)) = 0
Then, take x = 0, y = 1:
f (0) = 2f (0) + f (f (1))
√
So f (0) = 0. It follows that for a ≥ 0, f (a) = 0 by setting (x, y) = ( a, 0)
Finally, since f (f (b)) = 0 for all b ∈ R (why) , f (c) = −f (−1) c for all negatives c (why)
1.3 Injectivity and Surjectivity
Sometimes, it’s not clear to show, after assertion, what f should be. Perhaps, there are
too many f , we aren’t sure that whether f (a) appears in equations. Therefore, we hope to
reduce the numbers of f . There are two ways: injectivity and surjectivity.
Definition 1. A function f : A → B is called injective if f (a1 ) = f (a2 ) implies a1 = a2
Obviously, whenever we prove that a function is injective, the equation
f (something 1) = f (something 2)
becomes useful.
Example 8. Find all functions f : Z → Z that satisfy f (0) = 1 and
f (f (n)) = f (f (n + 2) + 2) = n
for all integers n
Solution. Notice that f (f (n)) = n implies f is injective (why) Therefore,
f (f (n)) = f (f (n + 2) + 2) → f (n) = f (n + 2) + 2
Then, by mathematical induction, f (2m) = 1 − 2m for all m ∈ Z Now, given an odd integer
k, write k = 2l + 1. Our goal is to find the value of f (2l + 1). However, note that
f (f (n)) = n and f (2m) = 1 − 2m
It follows that
f (2l + 1) = f (1 − 2 (−l)) = f (f (2 (−l))) = −2l = 1 − (2l + 1)
So we conclude that f (n) = 1 − n for all n ∈ Z
Remark 3. For convenience, let P (x1 , x2 , · · · , xn ) be the assertion of a functional equation

with n variables. Assume there are two distinct elements a1 , a2 in the domain of the function
f with f (a1 ) = f (a2 ). We should derive a contradiction by assertions. For many cases, just
compare two equations. For instance, P (a1 , x2 , · · · , xn ) and P (a2 , x2 , · · · , xn ).
In the previous remark, we see a standard way for proving injectivity: contradiction
method. Although the following example can be solved by exchanging x, y, we still give a
proof with injectivity.
Example 9. Find all functions f : Z → Z such that
f (x + f (y)) = f (x) + y
for all x, y ∈ Z
Solution. If f (a) = f (b) for some a, b ∈ Z. Compare y = a, b in the equation:
f (x) + a = f (x + f (a)) = f (x + f (b)) = f (x) + b → a = b
In other words, f is injective. As mentioned before, to use the injectivity, it suffices to choose
an equation of the form
f (something 1) = f (something 2)
Indeed, one may assert y = 0 and get
f (x + f (0)) = f (x)
It follows that f (0) = 0. We then see f (f (y)) = y by taking x = 0 in the equation. Plugging
back to the equation,
f (x + y) = f (x) + f (y)
In particular, f (x + 1) = f (x) + f (1), which implies
f (x) = xf (1)
Most of them do not satisfy the original equation, the only possibilities are
f (x) = x, ∀x ∈ Z or f (x) = −x, ∀x ∈ Z
Remark 4. Here is a useful result: If g ◦ f is injective, then so is f
Proof. Use the contradiction method, again. Suppose f is not injective, then there are two
elements a, b such that f (a) = f (b). So we have
g (f (a)) = g (f (b)) → a = b
since g ◦ f is assumed to be injective. Thus, we conclude that f must be injective.

For more information, https://en.wikipedia.org/wiki/Injective_function
Definition 2. A function f : A → B is called surjective if for any b ∈ B, there exists a ∈ A
satisfying f (a) = b. Equivalently, the pre-image of any element of B is not empty.
Example 10. Find all functions f, g : R → R such that g is injective and
f (g (x) + y) = g (x + f (y))
for all x, y in R
Solution. Since g is injective, we tend to fix LHS. Thus, set y = −g (x) to get
f (0) = g (x + f (−g (x)))
There must be a constant c such that
x + f (−g (x)) = c, ∀x ∈ R
This equation, in particular, shows that f is surjective. Now, the original condition tells us
g is also surjective. Let a be a real number with g (a) = 0. Then
f (y) = g (a + f (y)) → y − a = g (y) , ∀y ∈ R
by using the surjectivity of f . Substitute into the original condition,
f (x + y − a) = x + f (y) − a
which is easy to get f is a linear function (left as an exercise)
Indeed, for every surjective function, we can replace f (something) to an arbitrary number
whenever something can be any number of the codomain.
Definition 3. A function f : A → B is called bijective if f is both injective and surjective.
Sometimes, it’s not obvious that a function is injective because the condition is compli-
cate. However, if we know that the function is surjective, then we can reduce the condition
to a simpler form. Then, maybe it’s possible to show the injectivity.
Example 11. (2002 A1) Find all functions f from reals to reals such that
f (f (x) + y) = 2x + f (f (y) − x)
for all x, y in R
Solution. Since there is already a term ”2x”, which is a surjective function. So if we want
to show f is surjective, we may hope x is ”free”. Therefore, one should choose a suitable y
such that one of the terms f (f (x) + y) and f (f (y) − x) is a constant.
P (x, −f (x)) → f (f (−f (x)) − x) = f (0) − 2x
So f is surjective. Now, suppose that a, b are reals such that f (a) = f (b). Then,
P (a, y) , P (b, y) → 2a + f (f (y) − a) = 2b + f (f (y) − b)
We hope a = b, which is equivalent to f (f (y) − a) = f (f (y) − b) for some y. Note that, so

far, we have only one known pair (x, y) such that f (x) = f (y), namely (a, b)
Choose y to be the real with f (y) = a + b, the conclusion follows. Finally, one would like to
set x = 0 (why)
f (f (0) + y) = f (f (y)) → f (y) = y + f (0)
by using injectivity. Those functions are solutions as
LHS = f (x + f (0) + y) = x + y + 2f (0)

= 2x + f (y + f (0) − x) = 2x + f (f (y) − x) = RHS
Example 12. (2015 Swiss) Find all functions f : R → R such that for arbitrary x, y ∈ R,
(y + 1) f (x) + f (xf (y) + f (x + y)) = y
Solution. First of all, rewrite the condition,
f (xf (y) + f (x + y)) = (1 − f (x)) y − f (x)
So f is surjective (why) Now, assume a, b ∈ R s.t. f (a) = f (b). Consider
P (0, a) → (a + 1) f (0) + f (f (a)) = a
P (0, b) → (b + 1) f (0) + f (f (b)) = b

If f (0) 6= 1, then f is injective. However, f (0) 6= 1 since
P (0, c) → (c + 2) f (0) = c
where c is a real with f (c) = 0. We conclude that f is bijective. Next,
P (c, 0) → f (cf (0)) = 0 = f (c)
It follows that f (0) = 0. From here, one can easily find the desired function f is
f (x) = −x, ∀x ∈ R
1.4 Cauchy’s Functional Equation
So far, one may observe that the domains of functional equations are usually ”R”. Thus, I
would like to introduce a result of real functions.
Definition 4. A function f : A → B is called additive if
f (a1 + a2 ) = f (a1 ) + f (a2 )
holds for all a1 , a2 , a1 + a2 ∈ A
We consider the situation that A = B = R. By using induction, it’s easy to check
f (n) = nf (1)
for all n ∈ Z. (ref. example 9) For rationals, we have

p 1 p
f = f (p) = f (1)
q q q
The first equality follows from f (nx) = nf (x), which can be proved by induction, also. Now,
we may expect that f (x) = xf (1) for all x ∈ R. However, it’s false! (why?) Fortunately,
we have a criterion.
Theorem 1. Anyone of following conditions implies an additive f : R → R must be linear.
• f is continuous on some interval
• f is bounded on some interval
• f is monotonic on some interval
Proof. For the first one, notice that if f is continuous at one point, then f is continuous
everywhere. Next, for any real r, choose a sequence of rationals {an }n≥1 converging to that
number. Since f is continuous,
f (rx) = lim f (an x) = lim an f (x) = rf (x)
n→∞ n→∞
For the second one, suppose there is a real r s.t. f (r) 6= rf (1). WLOG f is increasing and
f (r) > rf (1). Choose a rationals q so that ff (1)
(r)
> q > r then
f (r) < f (q) = qf (1) < f (r)

a contradiction. For the third, let an to be any rationals such that nx − an ∈ [a, b]. Then
|f (nx − an ) | = |n (f (x) − xf (1)) − (an − nx) f (1) | ≥ n|f (x) − xf (1) | − | (nx − an ) f (1) |
which means f (x) must be xf (1), done!
Note that the method that we used to prove the second (or third) part of theorem 1 is
useful. That’s because it’s usually easier to determine the function on rationals and one can
use rationals to approach any real number.
Example 13. Find all functions f : R → R such that for all reals x, y, we have
f (x)2 + f (y)2 + f (x + y)2 = 2x2 + 2y 2 + f (xf (y) + yf (x))

Solution. The original equation is complicate, especially the term f (xf (y) + yf (x)). To
kick it out, we may hope to have f (0) = 0 (why) This is indeed the case,
P (0, 0) → 2f (0)2 + f (0) = f (0)
implies f (0) = 0. Next, subtract P (x, 0) and P (y, 0) from P (x, y), we have the following
Q (x, y) : f (x + y)2 = f x2 + f y 2 + f (xf (y) + yf (x))

Then, compare
P (x, 0) : f (x)2 + f x2 = 2x2

P (−x, 0) : f (−x)2 + f x2 = 2x2

we see that f (−x) = ±f (x) holds for all x. Assume f (a) = f (−a) = 0 for some a
P (a, −a) → 0 = 4a2 ⇔ a = 0
Now, if f (b) = f (−b) for some b
Q (b, −b) → 0 = 2f b2

so b2 = 0. In other words, b = 0. From here, one can conclude that f must be odd. Finally,
add two equations Q (x, y) and Q (x, −y),
f (x + y)2 + f (x − y)2 = 2f x2 + 2f y 2

If we set g (x) = f (x2 ), it becomes
g (x + y) + g (x − y) = 2g (x) + 2g (y)
The remains are left to readers:
Exercise. 1. Show that g (nx) = n2 g (x) for all n ∈ N and x ∈ R
2. Notice that g (x) ≤ 2x2 . Try to prove that g (x) = g (1) x2 holds for all x ∈ R
We return to an application of the theorem.
Example 14. (2015 Korea) Find all functions f : R → R such that
f x2015 + f (y)2015 = f (x)2015 + y 2015

holds for all reals x, y
Solution. Start with
f f f x2015 + f (y)2015 = f f (x)2015 + y 2015 = x2015 + f (y)2015

By using the surjectivity of x2015 , we get f (f (x)) = x and thus, f is bijective. Now,
P (x, f (0)) → f x2015 = f (x)2015 + f (0)2015

and then
P (x, f (y)) → f x2015 + y 2015 = f (x)2015 + f (y)2015 = f x2015 + f y 2015 − 2f (0)2015

Again, use the surjectivity of x2015 , the above becomes
f (x + y) = f (x) + f (y) − 2f (0)2015
One should be able to determine the value of a (hope so) Thus, we have
f (x + y) = f (x) + f (y) and f x2015 = f (x)2015

We claim that f is linear: Given any real number x. We have, for any rational number q,
f (x + q)2015 = (f (x) + f (q))2015 = f (x)2015 qf (1) f (x)2014 + · · · + q 2015 f (1)2015
On the other hand,
f (x + q)2015 = f (x + q)2015 = f (x)2015 + 2015qf x2014 + · · · + q 2015 f (1)2015

Since q can be arbitrary, f (x2014 ) = f (1) f (x)2014 (why) In particular, the second condition
in the theorem is fulfilled. The conclusion follows.
Example 15. (2017 CSP) Find all functions f : R+ → R such that

x
f (x) − f (x + y) = f f (x + y)
y
holds for all x, y ∈ R+
Solution. Because we expect the only non-trivial solution is

1
f (x) = , ∀x ∈ R+
x
we want to show that f (x) 6= 0 for all x ∈ R+
Assume a ∈ R+ so that f (a) = 0. Given b < a,
P (b, a − b) → f (b) = 0
Also, notice that

P (a, large) → f (large) = 0
Combine those results, we conclude that f ≡ 0
Now, suppose f (x) 6= 0 for all x ∈ R+ . It’s easy to see f (1) = 1 (why) So
P (x, x) → f (x) = 2f (2x)
We have
x
f (x) f y
+1
P (x, y) , P (y, x) → = y

f (y) f x
+1
and thus,

x 1 x
f y
+1 f (x) f (2x) 2
f y
+1
x y
y
= = = y
⇔f f =1
2f x
+2 2f (y) f (y) 2f x
+1 y x
It follows that f is multiplicative and
f (x) f (y) = (f (x) + f (y)) f (x + y)
By using theorem 1, we know that (why)
1
f (x) = , ∀x ∈ R+
x
1.5 Establishing Inequalities
Sometimes, it’s not straightforward to find a proof of f (a) = b for some given a, b. Instead,
we try to prove the following inequalities:
f (a) ≥ b and f (a) ≤ b
which maybe much easier to handle.
Example 16. Find all functions f : N → N with f (1) = 1 and
f (x + 5) ≥ f (x) + 5 and f (x + 1) ≤ f (x) + 1
holds for all x ∈ N
Solution. By the second inequality,
f (x + 5) ≤ f (x) + 5
Combine it with the first inequality, we get
f (x) + 5 ≤ f (x + 5) ≤ f (x) + 5
So the equality of f (x + 5) ≤ f (x) + 5 must hold, which implies f (x + 1) = f (x) + 1. A

simple induction will give f (n) = n for all n ∈ N
Because the condition is already an inequality, it’s very straightforward to come up with
such a solution.
Exercise. Find all functions f : R → R such that
f (x) + f (y) ≥ 2f (x + y)
Let’s , now, do a ”true” functional equation.
Example 17. (2017 OAO Shortlist) Find all functions f : N → N such that
n
X 1
=1
i=1
fi (n)
holds for all natural number n.

Solution. We want to find two relative equations. Here is a simple way, compare P (n) , P (f (n))
n f (n)
X 1 X 1
i
= 1 = i
i=1
f (n) i=1
f (f (n))
From above, we get that f (n) > n − 1 (why). In other words, f (n) ≥ n for all n ∈ N.
Therefore,
n n
X 1 X 1
1= i
≤ =1
i=1
f (n) i=1
n
It follows that f (n) = n must hold for all natural n, which is clearly a solution of the
functional equation.
Exercise. (Socrates) Find all functions f : N → N such that for each positive integer n,
there exists a positive integer k with
k
X
f i (n) = kn
i=1
To establish an inequality, we should take large numbers into consideration. For instance,
Example 18. (2014 Taiwan) Find all increasing functions f : N → N satisfying f (2) = 7,
f (mn) = f (m) + f (n) + f (m) f (n)
for all positive integers m and n
Solution. Recall the well-known identity:
(x + 1) (y + 1) = xy + x + y + 1
One tends to rewrite the condition:
f (mn) + 1 = (f (m) + 1) (f (n) + 1)
Let g (n) := f (n) + 1, we then have g (mn) = g (m) g (n) holds for all m, n ∈ N. Now,
y 1 1 3y
nx > 2y ⇔ log2 n > ⇔ g (n) = g (nx ) x > g (2y ) x = 2 x
x
This gives the lower-bound of g (n), namely n3 . A similar argument gives g (n) ≤ n3
Note that whenever the codomain of the function is R+ or N, we have a trivial inequality.
Example 19. (2016 Poland) Find all functions f : R+ → R+ such that
2f (f (f (x))) + 5f (f (x)) = f (x) + 6x
holds for all x ∈ R+
Solution. If one does not familiar to recurrence, take a look: https: // en. wikipedia.
org/ wiki/ Recurrence_ relation . Given any x ∈ R. Let ai = f i (x). Then conditions
P (x) , P (f (x)) , · · · becomes
2ai+3 + 5ai+2 = ai+1 + 6ai
for all i ∈ N0 . To get the general formula of ai , we should solve
2x3 + 5x2 − x − 6 = 0
i
The roots are 1, −2, − 32 , ai = b + c (−2)i + d − 32 If b or c is not zero, then ai < 0 for some
integer i (why) This means the sequence {ai }i≥0 is constant. In other words, f (x) = x. Note
that the choice of x can be arbitrary, we conclude the answer to this problem is
f (x) = x, ∀x ∈ R+
In most problems, one needs to establish the recurrence.
Example 20. (Socrates) Determine all functions f : R+ → R+ such that
f (f (x) + y) + f (x + y) = 2x + 2f (y)
Solution. We have the following equations:
P (x, f (1)) → f (f (x) + f (1)) + f (x + f (1)) = 2x + 2f (f (1))
P (x, 1) → f (f (x) + 1) + f (x + 1) = 2x + 2f (1)
P (1, x) → f (x + f (1)) + f (x + 1) = 2 + 2f (x)
P (1, f (x)) → f (f (x) + f (1)) + f (1 + f (x)) = 2 + 2f ((x))
So,
f (f (x)) = 2x − f (x) + (f (f (1)) + f (1) − 2)
Denote f (f (1)) + f (1) − 2 by c, then
f (f (x)) = 2x − f (x) + c
A simple induction gives
1 c 1 c
f n (x) = f (x) + 2x + nc − + (−2)n x − f (x) +
3 3 3 3
The remains are easy.
1.6 Exercise
To get a better feeling, I strongly suggest that the readers do lots of functional equations
because we usually come up with a great idea due to our experience.
1.6.1 2014
Problem 1. (Baltic Way) Find all functions f defined on all reals and taking real values
such that
f (f (y)) + f (x − y) = f (xf (y) − x)
for all real numbers x, y
Problem 2. (ELMO) Find all triples (f, g, h) of injective functions from the set of real
numbers to itself satisfying
f (x + f (y)) = g (x) + h (y)

g (x + g (y)) = h (x) + f (y)
h (x + h (y)) = f (x) + g (y)
for all real numbers x and y
Problem 3. (IZHO) Does there exist a surjective function f : R → R satisfying
f (f (x)) = (x − 1) f (x) + 2
for all real x
Problem 4. (Kazakhstan) Find all functions f : Q × Q → Q such that for all x, y, z ∈ Q,
f (x, y) + f (y, z) + f (z, x) = f (0, x + y + z)
Problem 5. (USA) Find all functions f : Z → Z such that
f (x)2
xf (2f (y) − x) + y 2 f (2x − f (y)) = + f (yf (y))
x
holds for all x, y ∈ Z with x 6= 0

1.6.2 2015
Problem 6. (Belarus) Find all functions f (x) determined on interval [0, 1] satisfying
f (f (x)) = f (x)
{f (x)} sin2 x + {x} cos f (x) cos x = f (x)

Problem 7. (Canada) Find all functions f : N → N such that
(n − 1)2 < f (n) f (f (n)) < n2 + n
for every positive integer n

Problem 8. (Greece) Find all functions f : R → R satisfying
yf (x) + f (y) ≥ f (xy)
for all x, y ∈ R
Problem 9. (Turkey) Find all the functions f : R → R such that
f x2 + 4xy 2 f (y) = f (x − y) + y 2 (f (x + y) + f (y))

holds for every x, y ∈ R
1.6.3 2016
Problem 10. (Balkan) Find all injective functions f : R → R such that for every real
number x and every positive integer n,

X n
i (f (x + i + 1) − f (f (x + i))) < 2016

i=1
Problem 11. (EMMO) Find all functions f : N → R such that

n−1
X f (k)
x+ = bxf (n)c
k=0
n
holds for all x ∈ R and n ∈ N

Problem 12. (HMMT) Let f : N → N be a function such that for w, x, y, z ∈ N
f (f (f (z))) f (wxf (yf (z))) = z 2 f (xf (y)) f (w)
Show that f (n!) ≥ n! for every positive integer n

Problem 13. (Netherland) Find all functions f : R → R such that
f (xy − 1) + f (x) f (y) = 2xy − 1

Problem 14. (Norway) Find all functions f : R → R such that

xy + 1
f (x) f (y) = |x − y| · f
x−y
holds for all x 6= y ∈ R
Problem 15. (Romania) Determine all functions f : R → R such that
f a2 − f b2 ≤ (f (a) + b) (a − f (b))

1.6.4 2017
Problem 16. (Belarus) Find all functions f : R+ → R+ satisfying the following equation
f (x + f (xy)) = xf (1 + f (y))
Problem 17. (Greece) Find all functions f, g : R → R such that g (1) = −8 and
f (x − 3f (y)) = xf (y) − yf (x) + g (x)

Problem 18. (IZHO)Find all functions f : R → R such that
x + y 2 f (yf (x)) = xyf y 2 + f (x)


Problem 19. (Pakistan) Find all functions f : R+ → R+ such that for all distinct x, y, z,
f x2 − f (y) f (z) = f (xy ) f (y) f (z) (f (y z ) − f (z x ))

Problem 20. (Turkey) Given a real number a, try to find all functions f : R → R so that
f (xy + f (y)) = yf (x) + a
for every x, y ∈ R
Chapter 2
Cauchy FE
In chapter one, we have introduced some terminologies. Recall the statement of theorem 1:
For an additive f : R → R, any of the following conditions guarantees that f is linear.
• f is continuous on some interval
• f is bounded on some interval
• f is monotonic on some interval
We’re not contented because additivity and others are strong conditions.
2.1 Taiwanese Transformation

If one makes a suitable transformation, then he/she will get another useful result.
Example 21. All monotonic functions f : R → R+ such that
f (x + y) = f (x) f (y)
are f (x) = ax , ∀x ∈ R for some constant a.

Solution. Define g : R → R>a by g (x) = log f (x). It’s also monotonic since it’s the
composition of two monotonic functions. Now, the original equation becomes
g (x + y) = g (x) + g (y)
We deduce g (x) = cx, ∀x ∈ R for some constant c, which means
f (x) = ax , ∀x ∈ R
for some constant a (why)
27
Exercise. Find all continuous functions f : R → R so that
f (xy) = f (x) f (y)
for all x, y ∈ R. PS: Such a function is called multiplicative.
Remark 5. A multiplicative function satisfying Cauchy condition must be zero or identity.
Proof. Notice that if a function is multiplicative, then
f x2 = f (x)2 ≥ 0

So f is bounded on some interval. It follows that f is linear.
Exercise. Find all continuous functions f : R → R so that
f (xy) = f (x) + f (y)
for all x, y ∈ R
Example 22. Given a constant a. Find all functions f : R+ → R+ so that
f (x + y) = f (x) + f (y) + a
Solution. Why we want to transform the function? Because we hope to show that f is
linear by using Theorem 1! Thus, after guessing the answer to the FE, we should apply the
corresponded transformation. In this example, one may guess
f (x) = cx − a, ∀x ∈ R
for some constant c ∈ R are the only functions satisfying the conditions. So, we tend to
define g : R+ → R such that
g (x) = f (x) + a
Then the identity becomes
g (x + y) = g (x) + g (y)
which is trivail?
Another example,
Example 23. Given a constant d. Find all continuous functions f : R → R so that
f (x + y + d) = f (x) + f (y)
holds for x, y ∈ R
Solution. This is similar to the previous one. First of all, one need to guess the answer.
Assume that f is linear:
f (x) = ax + b
for some constant a, b ∈ R. Then we must have
ax + ay + ad + b = LHS = RHS = ax + ay + 2b ⇔ b = ad
In other words,
f (x) = cx + cd = c (x + d) , ∀x ∈ R
for some constant c ∈ R. Define g : R → R by the following rule
g (x) = f (x + d) → g (x + y) = g (x) + g (y)
Exercise. Given two real numbers a, b. Find all continuous functions f : R → R such that
f (x + y + a) = f (x) + f (y) + b
holds for all pairs (x, y) ∈ R × R
Before the next subsection, I would like to summary our previous work. (under a suitable
situation i.e., bounded)
Condition Transformation Solution

f (x + y) = f (x) + f (y) cx
f (x + y) = f (x) f (y) g (x) = log f (x) ax
f (xy) = f (x) f (y) g (x) = log f (ax ) xloga b
f (xy) = f (x) + f (y) g (x) = f (ax ) loga x
f (x + y) = f (x) + f (y) + a g (x) = f (x) + a cx − a
f (x + y + a) = f (x) + f (y) g (x) = f (x + a) c (x + a)
f (x + y + a) = f (x) + f (y) + b g (x) = f (x + a) + b c (x + a) − b
2.2 Variation and Extension
Return to additive functions, we have proved that under suitable conditions, they must be
linear. In this section, we’ll investigate more ”suitable conditions”.
Example 24. Find all additive functions f : R → R such that

1
f f (x) = 1, ∀x ∈ R∗
x
Solution. Notice that if f is a solution, then−f is also satisfies the conditions. So, let’s
assume that f (1) = 1. We’ll calculate f x21+x in two different ways.

1 1 1 1 1 1
f 2
=f −f = − = 2
x +x x x+1 f (x) f (x) + 1 f (x) + f (x)
On the other hand,

1 1 1
f = =
x2 + x f (x2 + x) f (x2 ) + f (x)
It follows that
f x2 = f (x)2 , ∀x ∈ R

And therefore, by the corollary, f must be the identity.
This calculation is tricky, but it helps in many variations. More examples,
Example 25. (2008 Korea) Find all additive functions f : R → R such that

1 f (x)
f = 2 , ∀x ∈ R∗
x x
Solution. This one looks very similar to the previous. We still assume f (1) = 1 and
1
calculate the same expression f x2 +x .
(2x + 1) f (x) − x2

1 1 1 f (x) f (x) + 1
f =f −f = 2 − =
2
x +x x x+1 x (x + 1)2 (x2 + x)2
It equals to
f (x2 + x) f (x2 ) + f (x)
=
(x2 + x)2 (x2 + x)2
which is equivalent to
f x2 + x2 = 2xf (x)

Now, it’s not obvious to show f is the identity. So we tend to calculate another expression
2 !
2f x1

1 2 1 2 1
f x+ = f x + 2 + 2 = 2xf (x) − x + 2 + − 2
x x x x
From early result, it must equal to

1 1 1
2 x+ f x+ − x2 − 2 − 2
x x x
After reduction, we find that f (x) = x, ∀x ∈ R
Sometimes, we have a partial Cauchy’s condition.
Example 26. Find all continuous functions f : R → R such that
f (x + y) = f (x) + f (y) , ∀x ∈ R and y ∈ R+
Solution. For any a, b ∈ R, choose a large c > 0 with a + b + 2c, b + 2c > 0, we have
f (a + b + 2c) = f (a + b) + f (2c)
On the other hand,
f (a + b + 2c) = f (a) + f (b + 2c) = f (a) + f (b) + f (2c)
Thus, f is additive and must be linear.
To derive the Cauchy’s condition from a partial Cauchy’s condition, we usually take
another sufficiently nice number and count a specify expression in two different ways.
Example 27. Find all continuous functions f : R → R so that
f (x + y) = f (x) + f (y)
for all x, y ∈ R satisfying |x − y| ≤ 1.
Solution. I’ll show that f (x + y) = f (x) + f (y) for all |x − y| ≤ 2. Indeed,

x+z y+z
2f = f (x) + f (z) , 2f = f (y) + f (z)
2 2
x+y
where z = 2
.Now, notice that

x+z y+z x + y + 2z
2f + 2f = 2f = 2f (2z) = 4f (z)
2 2 2
Therefore,
f (x) + f (y) = 4f (z) − 2f (z) = 2f (z) = f (x + y)
One can use induction to establish the Cauchy’s condition.
As the previous example suggests, extending several times may be much easier
Example 28. Find all functions f : R+ → R+ satisfying the equation
f (z + x) + f (z + y) = f (2z + x + y) , ∀x, y, z ∈ R+ with x, y < z 2
Solution. I’ll use induction to prove that f is ”almost” additive. Given a small , we claim
f (z + ) + f (w) = f (z + + w) for all w > z ≥ 133
By the condition, it’s true for w ∈ (z, z + z 2 ). Suppose it’s true for w ∈ z, z + n2 z 2

Consider w ∈ z + n2 z 2 − , z + n+1 2

2
z . Note that
r r
n+1 2 n+1 2 n+1 2
z+ z ++w < z+ z ++z+ z < 2z + nz 2
2 2 2
Thus, r !
n+1 2
f z+ z + + f (w) = 2f (a)
2
where a ∈ z, z + n2 z 2 . According to the inductive hypothesis,

r !
n+1 2 z++w
f (z + ) + 2f (a) = f z+ z + + 2f
2 2
z++w

As the result, f (z + ) + f (w) = 2f 2
= f (z + + w), as desired.
We have already proved that f (x) + f (y) = f (x + y) for large x, y ∈ R+ . Thus, f (x) = cx
for all large x ∈ R+ . Finally, because f (2x) = 2f (x) , ∀x ∈ R+ , we get f (x) = cx, ∀x ∈ R+
In fact, in these examples, we just extend the ”size” of a particular set. Formally, let
S0 = {(x, y) |f (x + y) = f (x) + f (y) is proved}
Our goal is to find sets S1 , S2 , . . . , Sk such that
S0 ⊂ S1 ⊂ S2 ⊂ · · · ⊂ Sk = R × R
and it’s easy to extend Si−1 to Si . We can modify our set S0 . For instance, consider
S0 = {x|f (x) = x3 + x is proved}
This method is used whenever a nice property holds for sufficient large numbers.
Example 29. (2014 Iran) Find all functions f : R+ → R+ st for all positive reals x, y,

y x+1
f +f = f (y)
f (x + 1) xf (y)
Solution. Notice that the domain of this function is R+ , so we can’t have
x+1
= y =⇒ yf (y) ≤ 1, ∀y ∈ R+
xf (y)
As I said in chapter one, establishing inequalities is helpful for R+ functional equation,

y x+1 f (x + 1) xf (y)
f (y) = f +f ≤ +
f (x + 1) xf (y) y x+1
which is equivalent to
(x + 1) f (x + 1) ≥ yf (y)
Therefore, we have
1
f (x) = , ∀x > 1
x
According to above, we hope to have
y x+1
> 1 and >1
f (x + 1) xf (y)
1 x
We know the first one implies y > f (x + 1) = x+1
. Under this condition, if y ≥ x+1
x+1 y (x + 1)
> ≥1
xf (y) x
From here, we can get

1 1
f (x) = , ∀x >
x 2
Repeat this argument (formally, use mathematical induction),
1
f (x) = , ∀x ∈ R+
x
Remark 6. There are some related problems. See the following link:
https: // artofproblemsolving. com/ community/ u141363h1115981p5108071

2.3 Applications
Example 30. Find all functions f : R → R satisyfing
• f (x) f (y) ≤ f (xy)
• f (x) + f (y) ≤ f (x + y)
Solution. For non-trivial case (f ≡ 0), We hope to reverse ≤. A common way is to multiply
−1 on both sides. So we want to show that f is an odd function, i.e
f (x) + f (−x) = 0, ∀x ∈ R
First of all, we have
f (0)2 ≤ f (0) , 2f (0) ≤ f (0)
which implies f (0) = 0. From here, we get
f (x) + f (−x) ≤ f (0) = 0
To show the other one, it seems that we need f (−1) = −1. Notice that there exists a number
α such that f (α) 6= 0, therefore,
0 = f (0) ≥ f (α) + f (−α) ≥ (f (1) + f (−1)) f (α)
Similarly,
0 = f (0) ≥ f (α) + f (−α) ≥ (f (−1) + f (1)) f (−α)
As the result, f (−1) = −f (1) = −1 (try to find out f (1))
Example 31. (2017 Thailand) Determine all function f : R → R such that
f (f (x) + yz) = x + f (y) f (z)
holds for every x, y, z ∈ R
Solution. Clearly, f is a bijective function. Suppose a is the root of f . Then
f (az) = f (f (a) + az) = a + f (a) f (z) = a
It follows that a must be zero and so (by setting y = 0, x = 0, respectively)
f (f (x)) = x, f (yz) = f (y) f (z)
To make use of the first identity, we may substitute x by f (x) and get
f (x + yz) = f (f (f (x)) + yz) = f (x) + f (y) f (z) = f (x) + f (yz)
which means f is additive. We conclude that f is identity.
Example 32. (2018 Korea) Find all functions f : R → R satisfying the following conditions:
• f (x + y) − f (x) − f (y) ∈ {0, 1} for all x, y ∈ R
• bf (x)c = bxc for all real x
Solution. At first glance, we wander when we have
f (x + y) = f (x) + f (y)
and when
f (x + y) = f (x) + f (y) + 1
happens. Notice the second condition is equivalent to
bxc ≤ f (x) < bxc + 1
So it seems that there is only one way: establishing inequalities.

Also, it’s natural to consider the interval [0, 1) because of the floor function condition. For
any two numbers x, y ∈ [0, 1), we have
0 ≤ f (x) + f (y) < 2
Thus, if f (x + y) lies in [0, 1), then f (x + y) = f (x) + f (y).

According those discussions, we know at least f is additive on the interval [0, 21 ) and bounded.
1
f (x) = cx, ∀x ∈ [0, )
2
for some constant c. Now, one should be able to extend this result to the following
f (x) = cx, ∀x ∈ [0, 1)
and conclude that (

f (x) = x, ∀x ∈ R
f (x) = bxc, ∀x ∈ R
Remark 7. CEH mentioned that even if the first condition is replaced by

1 1
f (x + y) − f (x) − f (y) ∈ {0, 1, , , . . . }
2 3
the problem is still solvable. I didn’t try this generalization, so I’m not sure ...
Example 33. Find all f : R → R satisfying
f (x + f (xy)) = f (x) f (y) + f (−y)
Solution. In the condition, take x = 0 and get
f (f (0)) = f (0) f (y) + f (−y)
If f (0) = 1, then
f (x + 1) = f (x) + 1
by setting y = 0. In particular, f (1) = 2 and the condition can be rewritten as follows.
f (x + f (xy)) = (f (x) − 1) f (y) + 2

1

Because we could take 1 out of the brackets., we tend to consider P x, y + x
,

1
f (x) f y + − f (y) = 1
x
Consequently,
1 1
f y+ − f (y) = f − f (0)
x x
which means f − 1 is additive (replace x1 by 0, the identity still holds) and thus, f is multi-
plicative (why?). According to the corollary, f is linear and so
f (x) = x + 1, ∀x ∈ R
In the above partial solution, it’s natural to divide the problem into two cases. That’s because
if f (0) 6= 1, then we can get a useful identity, immediately. This case is left as an exercise
for the readers.
2.4 Exercise
Problem 21. Find all f : R+ → R+ such that
f (f (x + y)) − f (y) − f (x − y) = y
Problem 22. (2017 Iran) Find all functions f : R+ → R+ such that

x + f (y) 1 1
=f +f
xf (y) y x
Problem 23. (1979 ISL) Given a function f : R → R such that
f (xy + x + y) = f (x) + f (y) + f (xy) , ∀x, y ∈ R
Show that
f (x + y) = f (x) + f (y) , ∀x, y ∈ R
Problem 24. (Serbia) Suppose a non-constant function f : R → R satisfies
f (x) f (x − y) + f (y) f (x + y) = f (x)2 + f (y)2
Show that f is additive.
Problem 25. (2018 SRMC) Determine all functions f : R → R such that
f (x + 1) = f (x) + 1, f x4 − x2 = f (x)4 − f (x)2

Problem 26. (2016 APMO) Find all functions f : R+ → R+ such that
(z + 1) f (x + y) = f (xf (z) + y) + f (yf (z) + x)
Problem 27. (2014 Iran) Find all continuous functions f : R≥0 → R≥0 such that
f (xf (y)) + f (f (y)) = f (x) f (y) + 2, ∀x, y ∈ R≥0
Problem 28. (2002 A4) Find all functions f : R → R such that
(f (x) + f (z)) (f (y) + f (t)) = f (xy − zt) + f (xt + yz)
for all real x, y, z, t

Problem 29. (2003 A5) Find all functions f : R+ → R+ such that
√ √ √
f (xyz) + f (x) + f (y) + f (z) = f ( xy) f ( yz) f zx
and f (x) < f (y) when 1 ≤ x < y
f (xf (x + y)) = f (yf (x)) + x2 , ∀x, y ∈ R
f (1 + xy) − f (x + y) = f (x) f (y) , ∀x, y, ∈ R
and f (−1) 6= 0.
Problem 32. (2013 Brazil) Find all the injective functions f : R∗ → R∗ such as
f (x + y) · (f (x) + f (y)) = f (xy)
for all x, y ∈ R∗ with x + y 6= 0
Problem 33. (2016 EMC) Determine all functions f : R → R such that equality
f (x + y + yf (x)) = f (x) + f (y) + xf (y)
holds for all real numbers x, y
Problem 34. (2017 Azerbaijan) Find all functions f : R → R such that
f x + yf x2 = f (x) + xf (xy)


Chapter 3
Double Counting
3.1 Switching
Basically, LHS and RHS, in the equation, must hold the same properties. For instance,
if LHS is an increasing function of x, then RHS is also an increasing function of x. In
particular, if LHS is symmetric w.r.t the variables, then so is RHS. If this is the case, we
simply exchange the variables and observe what would happened.
Example 34. (Socrates) Find all functions f : R → R such that
f (f (x) + f (y)) = f x2 + 2x2 f (y) + f (y)2

Solution. Notice that the left hand side is already symmetric. Switch x, y, we see that
f x2 + 2x2 f (y) + f (y)2 = f y 2 + 2y 2 f (x) + f (x)2

We now want to choose y that minimizes the number of uncertain values. After solving
y 2 − y = 0 ⇔ y = 0, 1
we tend to ”take” two equations:
P (x, 1) → f x2 + 2x2 f (1) + f (1)2 = f (1) + 2f (x) + f (x)2

P (x, 0) → f x2 + 2x2 f (0) + f (0)2 = f (0) + f (x)2

From here, one can easily get the answers
f (x) = 0, ∀x ∈ R or f (x) = x2 , ∀x ∈ R
39
Indeed, switching the variables is equivalent to comparing P (x, y) , P (y, x). We usually
get a simple identity, which is much easier to solve the function. The only bad news is: In
most of the problems, LHS and RHS aren’t symmetric.
To deal with this situation, we can assert a special form to one of the variables.
Example 35. Find all functions f : N → N so that
f (f (x) + y) = x + f (y) , ∀x, y ∈ N
Solution. As the previous comment, to make LHS become symmetric, substitute f (y) for
y, then we get
f (f (x) + f (y)) = x + f (f (y))
Now, the LHS is symmetric and we can switch x and y (Hooray!)
x + f (f (y)) = f (f (x) + f (y)) = f (f (y) + f (x)) = y + f (f (x))
This means
f (f (z)) = z + c, ∀z ∈ N
for some constant c (why) Recall, in chapter one, I have mentioned: Always use the little
result to get another identity. In this example,
P (x, f (y)) → f (x + y + c) = f (x) + f (y)
So f is eventually linear. Like what we do in chapter three: for any given x, take a sufficient
large y, we can see that x, f (x) satisfy the same linear equation and we conclude that f is
linear. The rest is easy.
Remark 8. In chapter one, we solve this FE by using the injectivity.
Example 36. Find all functions f : R → R satisfying the following equation
f (xf (y)) − x = f (xy) , ∀x, y ∈ R
Solution. It seems that we could do our magic trick
f (xf (y)) − x = f (xy) = f (yx) = f (yf (x)) − y
This identity isn’t very useful. Thus, we need to find another one. Rewrite the condition
f (xf (y)) = f (xy) + x
Replace x by f (x),
f (f (x) f (y)) = f (f (x) y) + x = f (xy) + y + f (x)

Since now LHS is symmetric,
f (xy) + y + f (x) = f (yx) + x + f (y)
must hold. And thus, f must be linear (why) Assume f (x) = ax + b, we have
LHS = axf (y) + b = ax (ay + b) + b = a2 xy + abx + b

RHS = axy + b + x
Compare the coefficient, a = 1, b = 1. In other words,
f (x) = x + 1, ∀x ∈ R
Example 37. (2016 Iran) Find all functions f : N → N such that for all a, b ∈ N, we have
(f (a) + b) f (a + f (b)) = (a + f (b))2
Solution. Similar to example 33, we have
(f (f (a)) + b) f (f (a) + f (b)) = (f (a) + f (b))2
As the result,
f (f (n)) = n + c
for some constant c. Note that f has fixed point:
P (a, a) → f (a + f (a)) = a + f (a)
It follows that f (f (n)) = n for all n ∈ N Now,
(f (a) + f (b)) f (a + b) = (a + b)2
The rest is easy. Example 6
Although this trick is useful, there are still many problems can’t be solved in this way.
For example,
Example 38. (Based on: 2014 Romania) Find all functions f : R+ → R+ such that
f (x + 3f (y)) = f (x) + f (y) + 2y, ∀x, y ∈ R+
Fortunately, if we modify our trick a little bit, the previous FE becomes trivial.
3.2 Three variables method
As the previous discussion, we hope to find a more useful trick. Our idea is simple! Usually,
we got stuck after switching variables because there are many restrictions. Consequently, we
introduce a ”free” variable in order to ignore the restrictions.
Example 39. (Socrates) Find all solutions f : R+ → R+ satisfying the equation
f (x + y)2 = f (x)2 + 2f (xy) + f (y)2 , ∀x, y ∈ R+
Solution. Despite the consequences, replace y by y + z:
f (x + y + z)2 = f (x)2 + 2f (x (y + z)) + f (y + z)2
Notice that the term f (y + z)2 can be further simplified:
f (y + z)2 = f (y)2 + 2f (yz) + f (z)2
As the result,
f (x + y + z)2 = f (x)2 + f (y)2 + f (z)2 + 2f (x (y + z)) + 2f (yz)
Cyclic the variables, we get
f (yz) + f (xy + zx) = f (zx) + f (yz + xy) = f (xy) + f (zx + yz)
Since there is always a solution to the system of equations: yz = a, zx = b, xy = c, the

identity can be rewrote as
f (a) + f (b + c) = f (b) + f (c + a) = f (c) + f (a + b)
Fix b, c, we see that

f (a + c) − f (a) = f (b + c) − f (b)
In other words,
f (x + c) = f (x) + (f (b + c) − f (b))
Define g : R+ → R by g (c) = f (b + c) − f (b), then
f (x + y) = x + g (y) , ∀x, y ∈ R+
which is an easy exercise.
Remark 9. Note that the function g is well-defined

Example 40. (Based on: 2014 Romania) Find all functions f : R+ → R+ such that
f (x + 3f (y)) = f (x) + f (y) + 2y, ∀x, y ∈ R+
Solution. Observe the equation
f (x + 3f (y + z)) = f (x) + f (y + z) + 2 (y + z)
We hope that there is another way to calculate LHS. According to the condition, we tend to
replace z by 3f (z):
LHS = f (x + 3f (y) + 3f (z) + 6z)

= f (x + 3f (y) + 6z) + f (z) + 2z
= f (x + 6z) + f (y) + 2y + f (z) + 2z
On the other hand,
RHS = f (x) + f (y + 3f (z)) + 2y + 6f (z)

= f (x) + f (y) + f (z) + 2z + 2y + 6f (z)
We conclude that
f (x + 6z) = f (x) + 6f (z)
holds for all x, y ∈ R+ . Now,
f (6x + 6z) = f (6x) + 6f (z) → 6f (z) = f (6z) + c, ∀z ∈ R+
for some constant c. Therefore,
f (x + z) = f (x) + f (z) + c
which implies f is linear (why) and thus,
f (x) = x, ∀x ∈ R+
Remark 10. Because we want to calculate an expression in two different ways, we usually
”iterate” our condition. Besides, since we exchange two variables, we usually need to deal
with something like Cauchy FE and get the additivity.
Example 41. (Socrates) Find all functions f : R+ → R+ such that
f (2x + 2f (y)) = x + f (x) + 2y
Solution. Substitute 2y + 2f (z) for y,
LHS = f (2x + 2f (2y + 2f (z)))

= f (2x + 2y + 2f (y) + 4z)
= x + y + 2z + f (x + y + 2z) + 2y
On the other hand,
RHS = f (2x + 2f (2y + 2f (z)))

= x + f (x) + 2 (2y + 2f (z))
After simplification,
f (x + y + 2z) + 2z = f (x) + y + 2f (z)
By switching the variables, we see that f is linear and thus,
f (x) = x, ∀x ∈ R+
Remark 11. In fact, we can solve this problem by switching x, y directly. We have
f (2f (x) + 2f (y)) = f (x) + f (f (x)) + 2y
Thus,
f (x) + f (f (x)) = 2x + c
for some constant c. According to Example 20, we conclude that
f (x) = x, ∀x ∈ R+
However, switching doesn’t help for the following problem:

Determine all functions f : R+ → R so that
f (2x + 2f (y)) = x + f (x) + 2y
holds for all positive reals x, y
In fact, I’m inspired by

Example 42. (2007 A4) Find all functions f : R+ → R+ so that
f (x + f (y)) = f (x + y) + f (y) , ∀x, y ∈ R+
Solution. The solution is due to Andreas Dwi Maryanto Gunawan, from Indonesia.
For any positive reals z, we have that
f (x + f (y)) + z = f (x + y) + f (y) + z
Then, take f on both side
f (f (x + f (y)) + z) = f (f (x + y) + f (y) + z)
Use the condition to reduce expressions:
f (x + f (y) + z) + f (x + f (y)) = f (x + y + f (y) + z) + f (x + y)
Again, we have
f (x + y + z) + f (y) + f (x + y) + f (y) = f (x + 2y + z) + f (y) + f (x + y)
⇒ f (x + y + z) + f (y) = f (x + 2y + z)
The rest is a routine work.
Remark 12. In general, if our condition is
f (x + g (y)) = g (y) + h (x, y)
for some functions g, h. Then we can try to compare P (x, y + g (z)) and P (x + h (y, z) , y)
Exercise. (Socrates) Find all functions f : R+ → R+ so that for all x, y ∈ R+
f (x + f (y)) = 3f (x) − 2x + f (y)
Sometimes, we need to use this method twice.
Exercise. (2016 Taiwan) Find all functions f : R+ → R+ such that
f (x + y + f (y)) = 4030x − f (x) + f (2016y) , ∀x, y ∈ R+

3.3 Double Counting
Indeed, switching and three variables method are kinds of double counting. The central
concept is: for a particular expression, we calculate the value in two different ways.
Example 43. (2015 APMO) Let S = {2, 3, 4, · · · } denote the set of integers that are greater
than or equal to 2. Does there exists a function f : S → S such that
f (a) f (b) = f a2 b2 for all a, b ∈ S with a 6= b

Solution. Notice that we have
f (21 )f (22 )f (23 )f (24 )f (25 ) = f (23 )f (24 )f (25 )f (26 )

= f (25 )f (26 )f (214 )
= f (214 )f (222 )
= f (272 )
On the other hand,
f (21 )f (22 )f (23 )f (24 )f (25 ) = f (21 )f (22 )f (24 )f (216 )

= f (22 )f (24 )f (234 )
= f (24 )f (272 )
So f (24 ) = 1, which is absurd! Thus, we conclude that there is not such a function.
Remark 13. If we apply the following transformation: Let g : log S → log S defined by
g (x) = log f (ex )
Then we find that the condition becomes
g (a) + g (b) = g (2a + 2b)
Consequently,
g (a) + g (b) + g (2c) = g (2a + 2b) + g (2c) = g (4a + 4b + 4c)
Switching the variables, we get
g (2d) = g (d) + c, ∀d ∈ log S
And so,
2g (d) = g (4d) = g (2d) + c = g (d) + 2c → g (d) = 2c
It follows that g ≡ 0, which is impossible!
Of course, we need to have some identities to ”open” the terms. Notice that we, initially,
have only the condition. Therefore, we usually use the given condition to do the trick.
Example 44. (2014 Poland) Find all functions f : Q+ → Q+ satisfying
f (f (f (. . . f (f (q)) . . . ))) = f (nq)

| {z }
n
for all integers n ≥ 1 and rational numbers q > 0
Solution. We want to calculate the following expression
f (f (f (. . . f (f (q)) . . . )))
| {z }
n
in another way. By the definition of composition,
LHS = f (f (f (. . . f (f (f (f (f (. . . f (f (q) . . . )))) . . . )))

| {z }| {z }
a n−a
It follows that
f (a (n + 1 − a) q) = f (nq)
holds for all a ∈ {1, 2, · · · , n}. In particular,
f a a2 − a + 1 − a q = f a2 − a q → f (q) = f (mq) , ∀m ∈ N

0
by simple substitution. Now, for any x, y ∈ Q+ , write x, y as qr , qr0 , respectively. Then
f (x) = f (q 0 r · x) = f (qq 0 ) = f (qr0 · y) = f (y)
We conclude that f is constant, as desired.
Remark 14. Since we expect the only function is constant, we may want to show that for
any x, y ∈ Q+, there are some a < n ∈ N and q ∈ Q+ such that
(
x = a (n − a) q
y = nq
This is not so obvious. The readers can try to give it a proof or a counter example.
In most situations, it suffices to find out ”short” (because we hope to get a simple result)
identities in the beginning.
Example 45. (Socrates) Find all functions f : R+ → R+ so that
f x3 − f y 3 = (x − y) f x2 + f (xy) + f y 2 , ∀x, y ∈ R+

Solution. WLOG, assume f (1) = 1. Notice that
f x3 − f z 3 = f x3 − f y 3 + f y 3 − f z 3

Therefore, we have
x2 − 1 f x4 + f x2 + 1 = x2 − x f x4 + f x3 + f x2

+ (x − 1) f x2 + f (x) + 1

After simplification,
f x4 = x (x − 1) f x2 + x2 − x + 1 f (x) + x (x − 1)

For convenience, let g, h be two polynomials such that
f x3 = g x, f (x) , f x2 and f x4 = h x, f (x) , f x2

3 4
Our idea is to calculate f (x12 ) in two ways. Since (x4 ) = x12 = (x3 ) , it follows
f x12 = h x3 , f x3 , f x6

= h x3 , g x, f (x) , f x2 , g x2 , f x2 , f x4

= h x3 , g x, f (x) , f x2 , g x2 , f x2 , h x, f (x) , f x2

and
f x12 = g x4 , f x4 , f x8

= g x4 , h x, f (x) , f x2 , h x2 , f x2 , f x4

= g x3 , h x, f (x) , f x2 , h x2 , f x2 , h x, f (x) , f x2

From here, one can show that
f x2 + f (x) = x2 + x, ∀x ∈ R+

Finally,
f x3 = (x − 1) f x2 + f (x) + 1 + 1 = (x − 1) x2 + x + 1 + 1 = x3

Usually, we use two operations which are commute to realize our idea.
3.4 Exercise
We begin with standard exercises.
Problem 35. (2006 MOP) Find all functions f : R → R satisfying
f (x + f (y)) = x + f (f (y))
Problem 36. Determine all functions f : R+ → R+ such that
f (x + y + f (y)) + f (x + z + f (z)) = f (2f (z)) + f (2y) + f (2x)
holds for all x, y, z ∈ R+
Problem 37. (Socrates) Find all functions f : R+ → R+ such that
f (x + f (x) + y) = f (y) + 2x, ∀x, y ∈ R+
f (x + 2f (y)) = f (x + y) + y, ∀x, y ∈ R+
Problem 39. (Mohammed Jafari) Determine all functions f : R+ → R+ such that
f (x + f (x) + 2y) = f (2x) + y + f (y)
Problem 40. (2007 Romania) Let f : Q → R be a function such that
|f (x) − f (y) | ≤ (x − y)2
for all x, y ∈ Q. Prove that f is constant.
Problem 41. (Own) Find all functions f : N → N satisfying that for all a, b ∈ N,
f (a + f (b)) = f (a)f (b)
Problem 42. (MDS) Find all functions f : R+ → R+ such that
f (x + f (y)) + x = f (x + f (x)) + y
holds for all positive reals x, y

In fact, for most problems, double counting is used to find some useful identities. Try to
bash the following problems! :)
Problem 43. (Crux) Find all real functions f such that
f (xy) = f (f (x) + f (y))
holds for all real x, y
Problem 44. (2017 Iran) Find all functions f : R+ × R+ → R+ that satisfy the following
conditions: for all positive reals x, y, z,
f (f (x, y) , z) = x2 y 2 f (x, z)
and
f (x, 1 + f (x, y)) ≥ x2 + xyf (x, x)
Problem 45. (2010 A5) Determine all functions f : Q+ → Q+ which satisfy the following
equation for all x, y ∈ Q+ :
f f (x)2 y = x3 f (xy)

f x2 + yf (z) + f (x) = xf (x) + zf (y) + x, ∀x, y, z ∈ R+

Problem 47. (2005 A2) Find all functions f : R+ → R+ which have the property:
f (x) f (y) = 2f (x + yf (x))
for all positive real numbers x and y
Problem 48. Find all functions f : R → R such that
f (x) f (yf (x) − 1) = x2 f (y) − f (x) , ∀x, y ∈ R
Problem 49. (2013 USA) Find all functions f : N → N that satisfy the equation
f abc−a (abc) + f abc−b (abc) + f abc−c (abc) = a + b + c
for all a, b, c ≥ 2
Chapter 4
Calculus (Supplement)
In this chapter, we’ll fulfill some weird imaginations. I believe that for many people, at
least for me, when they first know what a functional equation is, they will claim that only
continuous functions or polynomials can be the solution. Yet, this imagination will be broken
as soon as they learned some weird solutions, e.g. solutions of Cauchy functional equation.
However, most of the functional equations have ”regular” solutions, and the word ”regular”
may refer to continuous or differentiable. Therefore, I’ll state some definitions, theorems
and properties in calculus and demonstrate how to obtain these properties in a functional
equation. Also, the proves of those theorems and properties can be easily found in any
calculus textbook, which means I’ll omit the proves here. In addition, Since this technique
is about calculus, we always need to derive some inequalities from the equation, and hence
it only works on some R+ functional equations and few R functional equations.
4.1 Do You Know How to Count?

Question. Given two sets, when would we say they have the same size? Does {dog, cat,
rabbit, duck} have the same size as {1, 2, 4, 5} ? How about the set of all positive integers
and integers? the set of all real numbers?
Answer. First we notice that if each set is finite, they have the same size iff there is a
bijection between two sets; if one is larger then the other, there is a injection from the
smaller one to the larger one. So we want to generalize this to the infinite case.
Definition 5. Given two sets A, B, we say |A| = |B| if there is a bijection f : A → B
51
Property. For any sets A, B, C, we have
1. |A| = |A|
2. If |A| = |B|, then |B| = |A|
3. If |A| = |B| and |B| = |C|, then |A| = |C|

Definition 6. Given two sets A, B, we say |B| ≥ |A| if there is an injection f : A → B
Remark 15. If there is an injection f : A → B, then there is a surjection g : B → A. The
converse is true if we assume axiom of choice.
Property. For any sets A, B, C, we have
1. If |A| ≥ |B|, |B| ≥ |C|, then |A| ≥ |C|
2. If |A| ≥ |B|, |B| ≥ |A|, then |A| = |B|

Definition 7. If |A| ≤ |N|, we say A is countable.
Property. If A is an infinite set , and A is countable, then |A| = |N|
Exercise. Prove
• |N| = |N ∪ {dog}|
• |N| = |Z|
• |N| = |N × N|
• |N| = |Q|,
Theorem 2. |R| |N|
Proof. Although we don’t actually know what real numbers are, and the construction will
be derived in the next section, we’ll first assume real numbers to be a integer plus a decimal
part, but we say that 999... after some digit is same plus 1 in the previous digit. For example,
2.999 . . . = 3
In fact, we can prove that the size of real numbers in (0, 1] is (strictly) larger than N. Let’s
assume the statement is false, which means that we can label the real numbers in (0, 1] by
integers. Assume the i-th number is ai , and write it in decimal system with infinite decimal
(not end in 000...). Next, we can choose a real number x ∈ (0, 1] such that the i-th digit is
different from the i-th digit of ai , and x is not a finite decimal. Therefore, x was not counted,
and we get a contradiction!
4.2 Construction of Numbers
Before learning calculus, we should first know what the numbers we’re familiar with are, so
in this section, I’ll first introduce the construction of integers, rational numbers, and real
numbers, and deal with some basic properties of them.
Definition 8. (Peano axioms) The set N are constructed as the following:
1. 1 is a positive integer.
2. For every positive integer a, there is a unique successor a0 , which is also a positive
integer.
3. Two positive integers are equivalent iff their successors are equivalent.
4. 1 is not a successor of any positive integer.
5. For any statement about positive integers, if one can prove that it’s true for 1, and for
n0 under the assumption that it’s true for n, then the statement is true for N
Definition 9. A relation ∼ on S is a collection of (a, b) ∈ S × S ( denote as a ∼ b ), if a

relation ∼ satisfies the following:
1. x ∼ x
2. x ∼ y ⇒ y ∼ x
3. x ∼ y, y ∼ z ⇒ x ∼ z
then we say ∼ is an equivalent relation.
Definition 10. Let S := {(a, b)|a, b ∈ Z, b 6= 0}, and we define an equivalent relation ∼ by
(a, b) ∼ (c, d) ⇔ ad − bc = 0
then the set of rational numbers is defined to be Q = S/ ∼
Definition 11. Define a sequence a1 , a2 , ... converges to a as the following:
∀ε > 0, ∃N ∈ N, s.t. n > N ⇒ |an − a| < ε
denoted by limn→∞ an = a, and a is called the limit of the sequence.

Definition 12. Let {an } stands for the sequence a1 , a2 , ..., and let
S := {({an }, {bn })| ∀n ∈ N, an , bn ∈ Q, bn > an ,
{an } non-decreasing, {bn } non-increasing, lim (bn − an ) = 0}

n→∞
and define an equivalent relation ∼ by
({an }, {bn }) ∼ ({cn }, {dn }) ⇔ lim (cn − an ) = 0

n→∞
then the set of all real numbers is defined to be R = S/ ∼
Definition 13. If {an } satisfies that
∀ε > 0, ∃N ∈ N, s.t. m, n > N ⇒ |an − am | < ε
then we call it a Cauchy sequence.
Property. Every convergent sequence of real numbers is a Cauchy sequence.
Theorem 3. (Completeness of real numbers) All real-valued Cauchy sequence {an } converges
in R , i.e. there is a real number a such that limn→∞ an = a
4.3 Limit and Continuity

Definition 14. We say that a real-valued function f (x) converges to a when x tends to
infinity if
∀ε > 0, ∃N ∈ R, s.t. x > N ⇒ |f (x) − a| < ε
denoted by limx→∞ f (x) = a, and a is called the limit of f (x) as x tends to infinity.
Definition 15. We say that a real-valued function f (x) converges to a when x tends to b if
∀ε > 0, ∃δ > 0, s.t. 0 < |x − b| < δ ⇒ |f (x) − a| < ε
denoted by limx→b f (x) = a, and a is called the limit of f (x) as x tends to b.
Remark 16. From 0 < |x − b| < δ in the previous definition, we should notice that f (b)
does not necessarily equal a.
Property. The limit does not necessarily exist, but once it exists, it must be unique.
Property. If both limn→∞ an and limn→∞ bn exist, then
1. limn→∞ an ± limn→∞ bn = limn→∞ an ± bn
2. limn→∞ an × limn→∞ bn = limn→∞ an × bn
3. Moreover, if limn→∞ bn 6= 0, we have limn→∞ an ÷ limn→∞ bn = limn→∞ an ÷ bn
Theorem 4. (Squeeze theorem) If an ≤ bn ≤ cn , and
lim an , lim bn , lim cn

n→∞ n→∞ n→∞
all exist, then limn→∞ an ≤ limn→∞ bn ≤ limn→∞ cn .

Moreover, if an ≤ bn ≤ cn , and both
lim an , lim cn
n→∞ n→∞
exist and are equal, then limn→∞ bn exists and equals the others.
Similarly, we also have the functional version of the previous properties, and their proofs
are almost the same.
Property. Let b ∈ [−∞, ∞], if both limx→b f (x), limx→b g(x) exist, then
1. limx→b f (x) ± limx→b g(x) = limx→b f (x) ± g(x)
2. limx→b f (x) × limx→b g(x) = limx→b f (x) × g(x)
3. Moreover, if limx→b g(x) 6= 0, then limx→b f (x) ÷ limx→b g(x) = limx→b f (x) ÷ g(x)
Theorem 5. (Squeeze Theorem) Let b ∈ [−∞, ∞], if f (x) ≤ g(x) ≤ h(x), and
lim f (x), lim g(x), lim h(x)

x→b x→b x→b
all exist, then limx→b f (x) ≤ limx→b g(x) ≤ limx→b h(x)

Moreover, if f (x) ≤ g(x) ≤ h(x), and both
lim f (x), lim h(x)

x→b x→b
exist and are equal, then limx→b g(x) exists and equals the others.
After understanding the definition of limit, we can define least upper bound and greatest
lower bound, and take limits of them in order to understand more about a function or
sequence.
Definition 16. Given a set S ⊆ R, we say that S has an upper bound if there exists M ∈ R,
such that for all x ∈ S, x ≤ M , and M is called an upper bound of S; similarly we can
define a lower bound of S, and if S has both upper and lower bound, we say S is bounded.
Property. A non-decreasing sequence with an upper bound must converge; similarly a non-
increasing sequence with lower bound must converge.
Definition 17. Given a set S ⊆ R, if S has an upper bound, then among all upper bounds
of S, there is a smallest one, and this value is called the least upper bound ( supremum )
of S, denoted by sup S; similarly we can define the greatest lower bound ( infimum ) of S,
denoted by inf S.
Definition 18. Define the limit superior ( lim sup ) as the following:
lim sup an = lim (sup am )

n→∞ n→∞ m>n
lim sup f (x) = lim (sup f (y))

x→∞ x→∞ y>x
lim sup f (x) = lim+ ( sup f (x))

x→b ε→0 x∈[b−ε,b+ε],x6=b
Similarly we can define the limit inferior ( lim inf ).

Property. If lim sup = lim inf, then the limit exists and equals the others.
Definition 19. Define a function f (x) to be continuous at x0 as the following:
lim f (x) = f (x0 )

x→x0
or in other words, the limit of f (x) when x tends to x0 is f (x0 ). We say f (x) is continuous
on a set if f (x) is continuous at all points in the set.
Property. Polynomials, rational functions, exponential function, logarithm, and trigono-
metric functions are continuous at all points where they can be defined.
Property. If both f (x), g(x) are continuous functions, then f (g(x)) is continuous at all
points wherever it’s well-defined.
Theorem 6. (Intermediate Value Theorem) If f is continuous on [a, b], then for any given
y between f (a) and f (b), there must exist a c ∈ [a, b] such that f (c) = y.
Property. A continuous function f on [a, b] must reach its supremum.
Theorem 7. (Bolzano Weierstrass Theorem)Given an infinite sequence a1 , a2 , ..., where
ai ∈ [a, b] holds for all i, then we can find a subsequence ai1 , ai2 , ... which converges.
Theorem 8. A non-decreasing (or non-increasing) function on (a, b) ⊆ R is continuous at
all but countable points.
Definition 20. A real-valued function f is uniform continuous if
∀ε > 0, ∃δ > 0, s.t. 0 < |x − y| < δ ⇒ |f (x) − f (y)| < ε
Property. A function which is continuous on a closed interval must be uniform continuous.
Example 46. Find all functions f : R+ → R+ satisfying that for all x, y ∈ R+ ,
xf (x + y) (yf (x) + 1) = 1
Proof. First, for all a < b, we can always find 0 < c < a. By choosing (x, y) = (c, a −
c), (c, b − c), we can obtain f (a) > f (b). Therefore, f is a decreasing function, so we can
apply theorem 8. In other words, f is continuous on all but countable points. In particular,
there is at least one point t such that f is continuous at t. Given any r > t, by taking
y = r − t, we can obtain
1 1
lim f (x + r − t) = lim = = f (r)
x→t x→t x(yf (x) + 1) t(yf (t) + 1)
Hence, f is continuous at r. Similarly, we can prove that f (r) is continuous for all r < t.
Therefore, we have
1 1
f (x) = lim+ f (x + y) = lim+ =
y→0 y→0 x(yf (x) + 1) x
Also, we can check it is a solution
Example 47. Find all functions f : R+ → R+ satisfying that for all a, b, c ∈ R+ ,
a, b, c forms a triangle ⇔ f (a), f (b), f (c) forms a triangle
Proof. If lim supx→0+ f (x) is not zero or doesn’t exist, then we can find t, k > 0 and a
sequence a1 , a2 , ... such that
1. limn→∞ an = 0
2. For all n ∈ Nf (an ) > k
3. f (t) < 2k
By choosing (a, b, c) = (t, an , an ), we can get a contradiction as long as n large enough. On
the other hand, lim inf x→0+ f (x) ≥ 0, hence by combining this with the previous result, we
proved that limx→0+ f (x) = 0 Given f (a) > ε > 0, we can set (a, b, c) = (a, δ, x), where
δ is small enough such that f (δ) < ε. As a result, for all x ∈ (a − δ, a + δ), we have
f (x) ∈ (f (a) − ε, f (a) + ε), which means f is continuous.
If there exists (a, b, x0 ) with x0 ≥ a + b and f (x0 ) ≤ |f (a) − f (b)|, by intermediate value
theorem we can obtain that for all x ≥ a + b, we have f (x) ≤ |f (a) − f (b)|. By taking
y ≥ 2a + 2b, and let w ∈ {a, b} be the one which its f -value is larger, we can plug in (w, w, y)
and get a contradiction from the fact that w, w, y can’t form a triangle but f (w), f (w), f (y)
can. Hence f (a + b) ≥ f (a) + f (b)
By plugging in ((a, b, a + b − ε)), we can get that f (a + b − ε) < f (a) + f (b). Let ε → 0, since
f is a continuous function, we obtain that f (a + b) ≤ f (a) + f (b). Therefore, f is additive,
and by combining this with f ≥ 0 we can get that f (x) = cx for some constant c, which is
indeed a solution.
Example 48. (2016 APMO) Find all functions f : R+ → R+ such that for all x, y, z ∈ R+ ,
(z + 1)f (x + y) = f (xf (z) + y) + f (yf (z) + x)
Proof. If there is a positive sequence a1 , a2 , ... such that
1. limn→∞ an = 0
2. limn→∞ f (an ) = ∞
We can WLOG assume that all an < 1 by forgetting first few terms. By plugging y =
1 − x, z = an in the original equation we can get
f (xf (an ) + (1 − x)) < (an + 1)f (1) < 2f (1)
It tells us that f ((1, f (an ))) bounded by 2f (1). By taking n goes to infinite, we can get that
f ((1, ∞)) is bounded by 2f (1). On the other hand, if we take x > 1, y > 1 and z very large
in the original equation, RHS is smaller than 4f (1), but LHS can be arbitrary large, hence
we get a contradiction!
As a result, such sequence an don’t exist. In other words, lim supx→0+ f (x) exists. Assume
lim supx→0+ f (x) = k, and we can find a sequence xn > 0 such that
1. limn→∞ xn = 0
2. limn→∞ f (xn ) = k
xn
By plugging in x = y = 2
and by taking n → ∞, we can obtain that
(z + 1)k = lim (z + 1)f (xn ) = lim 2f (xn f (z) + xn ) ≤ 2k

n→∞ n→∞
Therefore, k = 0, and hence limz→0+ f (z) = 0.

By forgetting one term in the RHS, we have that
f (xf (z) + y)
f (x + y) ≥
1+z
If there exists a > b with f (a) < f (b), we can first solve x + y = a, xf (z) + y = b,
a−b b − af (z)
x= ,y =
1 − f (z) 1 − f (z)
Also, because of limz→0 f (z) = 0 we can choose z very small and f (z) very small, and obtain
a contradiction! Therefore, f is an increasing function.
We can take z → 0 and let g(x) = limh→0+ f (x+h), hence we have that f (x+y) = g(x)+g(y).
Moreover, take y as y + ε and let ε → 0+ , we have that g(x + y) = g(x) + g(y). Join with
the fact that g ≥ 0, we obtain that g(x) = cx and hence f (x) = cx. Finally, we can get that
c = 1 and check it’s indeed a solution.
4.4 Differentiation
Differentiation is a method to understand what happens to a function near some point, and
its definition is just the limit of the slope of its secant lines.
Definition 21. A real-valued function f (x) is said to be differentiable at x0 if
f (x) − f (a)
lim
x→a x−a
df
exists and we denote the value by dx
(x0 ) or f 0 (x0 ), which is the derivative of f (x) at x0
Property. If f is differentiable at x0 , then f is continuous at x0
Property. If both f, g are differentiable, then
1. (f ± g)0 = f 0 ± g 0
2. (f g)0 = f 0 g + g 0 f
f 0 g−g 0 f
3. Moreover, if g(x0 ) 6= 0, then( fg )0 = g2
Property. If f is differentiable at x0 , and g is differentiable at f (x0 ), then
(g · f )0 (x0 ) = g 0 (f (x0 ))f 0 (x0 )
Property. If f and g are the inverse function of each other, f is differentiable at x0 and
f 0 (x0 ) 6= 0, then g is differentiable at f (x0 ) and
1
g 0 (f (x0 )) =
f 0 (x 0)
Property. The following differentiations are all taken w.r.t. x:

1. (xa )0 = axa−1
2. (ex )0 = ex (where e := limn→∞ (1 + n1 )n )
3. (log x)0 = x1 (where log = loge )
4. (sin x)0 = cos x
5. (cos x)0 = − sin x
Theorem 9. (Mean Value Theorem of Differentiation) If f is differentiable on [a, b], then
there exists c ∈ [a, b] such that
f (b) − f (a)
f 0 (c) =
b−a
4.5 Integration
Here, integral means Riemann integral , but actually we follow the definition of Darboux.
Definition 22. Define S to be a partition of [a, b] if S = {x0 , ..., xn |a = x0 < x1 < ... <
xn = b}, and |S| means maxni=1 (xi − xi−1 ), which is the length of the greatest interval in S
Definition 23. Define U (f, S) := Σni=1 fi (xi − xi−1 ), where fi is the least upper bound of f in
[xi−1 , xi ], and similarly we can define L(f, S) by changing the least upper bound into greatest
lower bound. Furthermore, we can define U (f ) := inf S U (f, S), and L(f ) := supS L(f, S)
Definition 24. A real-valued function f on [a, b] is said to be (Riemann) integrable if
U (f ) = L(f )
and we can define the integral of f on [a, b] by

Z b
f (x)dx := U (f )
a
Property. If f is continuous on [a, b], then f is integrable on [a, b]
Property. If f is integrable on [a, b], then f is integrable on [c, d] ⊆ [a, b]
Definition 25. Define the indefinite integral of f , F by

Z x
F (x) = f (u)du
a
Property. If f is integrable on both [a, b], [b, c], then f is integrable on [a, c] and
Z b Z c Z c
f (x)dx + f (x)dx = f (x)dx
a b a
Ra Rb
Definition 26. If b > a, then b
f (x)dx := − a
f (x)dx
Theorem 10. (Mean Value Theorem of Integration) If f is a continuous function on [a, b],
then there exists c ∈ [a, b] such that
Z b
1
f (c) = f (x)dx
b−a a
Theorem 11. (Fundamental Theorem of Calculus) If f is continuous, then the derivative

of the integral of f is still f
Theorem 12. (Fundamental Theorem of Calculus) If f is differentiable and its derivative

is continuous, there the difference between f and the indefinite integral of f 0 can only be a
constant.
Property. (Change of Variables) If f is continuous on [a, b]; g is a monotone and differen-

tiable [a, b]-valued function, then
Z b Z g −1 (b)
f (x)dx = f (g(u))g 0 (u)du
a g −1 (a)
where g −1 is the inverse function of g
Property. (Integration by Parts) If f, g are both integrable on [a, b], then

Z b Z b
0
f (x)g (x)dx = f (b)g(b) − f (a)g(a) − f 0 (x)g(x)dx
a a
4.6 Ordinary Differential Equation
ODE refers to a functional equation consisting of functions and their derivatives with only
one variable. However in this section, I’ll only go through some common form with only one
function in the equation to be solved, and the function is required to be differentiable in the
region; furthermore, how many times we could differentiate are usually consistent with the
equation.
Property. If an O.D.E. is of the form an u(n) + ... + a1 u0 + a0 u = 0, where u(k) means the
k-th derivative of u, and let x1 , ..., xr be all solutions of an xn + ... + a1 x + a0 = 0, where xi
has multiplicity mi , then all solutions of the previous O.D.E. are of the form
u(x) = Σri=1 Pi (x)exi x
where Pi is a polynomial with its degree not greater then mi
Property. If an O.D.E. is of the form an u(n) + ... + a1 u0 + a0 u = f , and g is a solution of

it , then all solutions are of the form
u(x) = g(x) + Σri=1 Pi (x)exi x
where Pi is a polynomial with its degree not greater then mi
Property. If a differential equation is of the form du

dx
= P (u)Q(x), then we can first rewrite
the equation as
du
= Q(x)dx
P (u)
and then take integral on both side,
Z Z
du
= Q(x)dx
P (u)
then we can find out u, while we should notice that there is a constant difference after the
indefinite integral.
4.7 Lebesgue Monotone Differentiable Theorem
In this section, I’m going to talk about Lebesgue monotone differentiable theorem. However,
it’s a strong theorem and its proof is too complex. Therefore, I won’t prove it here and will
instead demonstrate the usage of it.
Definition 27. We say that a set S ⊆ R has measure zero if

∞
[
∀ε > 0, ∃[ai , bi ], i = 1, 2, ..., s.t. S ⊆ [ai , bi ], Σ∞
i=1 bi − ai < ε
i=1
Property. The union of countably many arbitrary measure zero sets is still measure zero,
which tells us an arbitrary countable set is measure zero.
Definition 28. We say that a property holds almost everywhere (a.e.) if it fails to hold only
on a measure zero set.
Theorem 13. (Lebesgue Monotone Differentiable Theorem) If f is non-decreasing (or non-

increasing), then f is differentiable a.e.
Definition 29. We say that a subset S ⊆ X is a dense subset if every point x ∈ X is a

limit point of S, i.e. x = limn→∞ an for some {an }∞
n=1 ⊆ S
Property. Let S = (a, b) − (a measure zeor set) ⊆ R, where a, b might be ±∞. Then S is
dense in (a, b)
f (x)f (y) = f (x + yf (x))
Proof. If there exists f (a) < 1, then we can choose (x, y) = (a, 1−fa(a) ) and get that f (a) = 1,
which gives a contradiction!
Hence, for all x, we have that f (x) ≥ 1. Therefore, f is an increasing function. Moreover,
we can rewrite the equation as (1):
f (x + yf (x)) − f (x) f (y) − 1

=
yf (x) y
By applying Lebesgue monotone differentiable theorem, we can find a point x0 where f is

differentiable at. By setting x = x0 , y → 0+ , we can obtain that
f (y) − 1
lim+
y→0 y
exists. Let’s assume the value is c, hence we have that
f (x + yf (x)) − f (x)
lim+ =c
y→0 yf (x)
b−a
holds for all x. Also, if we plug x = a, y = f (a)
in (1), where b > a, we can get that
f (b) − f (a) f ( fb−a

(a)
)−1
= b−a
b−a f (a)
By taking a → b− , we have that b−a

f (a)
→ 0+ . Therefore,
f (b) − f (a)
lim− =c
a→b b−a
Hence, f is differentiable on R+ , and
f 0 (x) = c
Therefore, f (x) = cx + d, and we can check which pairs of c, d satisfy the original equation.
Example 50. Given a positive integer n ≥ 2, find all f : R+ → R+ satisfying that for all
x, y ∈ R+ ,
n−1
X
n n
f (x ) − f (y ) = (x − y) f (xk y n−1−k )
k=0
Proof. First, we should notice that f is an increasing function. P

Next, we are going to prove f
n−1
is continuous everywhere. Let y → x in the equation, and since k=0 f (xk y n−1−k ) is bounded
when y is close to x, RHS goes to 0. Therefore, LHS suggests us that f is continuous at xn ,
and since the image of xn is R+ , f is continuous everywhere.
Then we divide by xn − y n on both side and get that
Pn−1 k n−1−k
f (xn ) − f (y n ) k=0 f (x y )
n n
= n−1 k n−1−k
x −y
P
k=0 x y
By taking y goes to x, we can obtain that f is differentiable everywhere and

f (xn−1 )
f 0 (xn ) =
xn−1
Yet, we don’t know how to solve such kind of ODE which is involved in different points. It
is natural to think about changing the RHS, and to achieve this we can take y → 0+ and
assume limx→0+ f (x) = k, which exists since it is decreasing and have a lower bound, we can
obtain that
f (xn ) − k = x(f (xn−1 ) + (n − 1)k)
and the fact k = 0 can be derived from the previous equation by plugging in x = 1. Therefore,
f (xn−1 ) f (xn )
=
xn−1 xn
The rest thing is just solving the ODE on R+
f (x)
f 0 (x) =
x
which tells that f (x) = cx for some constant x. Also, we can check they are solutions
f (x) = f (x + y) + f (x + f (y) − f (x + y))
Proof. First, it can be easily observed from the original equation that f is a decreasing
function. If f is bounded above, i.e.
lim f (x) < ∞,

x→0+
then by taking y = x → 0+ , we can get that x, x + y, x + f (y) − f (x + y) all go to 0, which

means limx→0+ f (x) = 0, and it contradict the fact that f is a decreasing function with
positive value.
If f is not continuous at some point, say x0 , then we can choose two sequences a1 , a2 , ... and
b1 , b2 , ... such that an → x− +
0 and bn → x0 . Now, by plugging in y = an , x = bn − an , we have
f (bn − an ) = f (bn ) + f (bn − an + f (an ) − f (bn )) ≤ f (x0 ) + f (k),
where k = limx→x−0 f (x)−limx→x+0 f (x) > 0, and we obtain a contradiction by taking n → ∞

since bn − an → 0+ .
By taking x → ∞, we can make x, x + y, x + f (y) − f (x + y) all go to infinity, and hence we
can conclude that limx→∞ f (x) = 0. By rewriting the original equation, we can get
f (x + f (y) − f (x + y)) − f (x) −f (x + y)
=
f (y) − f (x + y) f (y) − f (x + y)
Also, by Lebesgue monotone differentiable theorem, we know f 0 exist almost everywhere.
Therefore, for all x such that f is differentiable, we have
−f (x + y)
f 0 (x) = lim
y→∞ f (y) − f (x + y)
because f (y) − f (x + y) → 0 as y → ∞. Moreover, we can rewrite the equation above,
f 0 (x) − 1 f (y)
0
= lim .
f (x) y→∞ f (x + y)
Note that for arbitrary x, if R.H.S. exists and does not equal 1, then f 0 (x) also exists because
−1
we can obtain f 0 (x) by composing L.H.S. with t−1 . Note that for the case R.H.S. goes to
0
+∞, f (x) = 0 still exists. However, we have
f (y) f (y) f (y) f (a + y) f (y)
lim lim = lim lim = lim
y→∞ f (a + y) y→∞ f (b + y) y→∞ f (a + y) y→∞ f (a + b + y) y→∞ f (a + b + y)
f (y) f (y)
if both limy→∞ f (a+y)
and limy→∞ f (b+y)
exist. Also, the last term must be greater then 1
f 0 (a)−1 f 0 (b)−1
since f 0 (a), f 0 (b) ≤ 0 and f 0 (a)
to prove f 0 (x) exists it’s
, f 0 (b) > 1. Hence, for arbitrary x,
0 0
enough to find a, b such that a + b = x and both f (a) and f (b) exist.
Suppose f 0 exists on a set S with R \ S measure zero, then for any x, (0, x) ∩ S ∩ (x − S) is
non-empty since (R \ S) ∪ (R \ (x − S)) is still measure zero. Take any a ∈ (0, x) ∩ S ∩ (x − S),
and set b = x − a we can conclude that f 0 (x) exists. Next, let
f 0 (x) − 1
g(x) = log ,
f 0 (x)
and we can get that g : R+ → (0, ∞] satisfies the Cauchy equation. Note that we don’t need
and we can’t subtract ∞ from ∞, but we can still say that g satisfies the Cauchy equation.
However, it’s easy to check g can’t take both finite and infinite value. Moreover, we can
conclude that g(x) = rx for some r might be ∞. Therefore,
f 0 (x) − 1
log = rx.
f 0 (x)
By solving the ODE above, we can conclude that
1
f (x) = x − log(erx − 1)
r
if r ∈ (0, ∞) and f (x) = constant if r = ∞, and the second case is invalid. Also, we can
check that the functions in the first case are solutions!
4.8 Problems
Problem 50. Find all functions f : R+ → R+ satisfying that for all x, y ∈ R+ ,
f (y)f (xf (y)) = f (x + y)
Problem 51. Find all functions f : R → R satisfying that for all x, y ∈ R,
f (x2 + f (x)y) = f (x)2 + xf (y)
Chapter 5
Functional Equation in N/Z
Recently, there are more and more functional equations in integers. For those problems,
our previous ideas may not be enough to derive a complete solution. For instance, a linear
function f (x) = ax + b is surjective if f : R → R, but it’s not necessarily surjective when
f : Z → Z. Therefore, we are going to study such problems in details.
5.1 Mathematical Induction

There are many great properties for integers. In my opinion, the well-ordering principle is
one of the most important things. One simple way to solve FEs in positive integers is:
1. Bash out f (1) , f (2) , . . . , f (k) where k is a ”small” positive integer
2. Use induction to find solutions
Example 52. (2015 Canada) Find all functions f : N → N such that for all n ∈ N,
(n − 1)2 < f (n) f (f (n)) < n2 + n
Solution. It’s easy to see that f (1) = 1. Suppose we have proved f (k) = k for all k < n,
we must have f (m) ≥ n for all m ≥ n. That’s because if f (m) = k for some k < n,
(m − 1)2 < f (m) f (f (m)) = k 2 = (n − 1)2
We get a contradiction. Now,
nf (n) ≤ f (n) f (f (n)) < n2 + n
which means f (n) < n + 1. So f (n) = n, as desired.
67
Example 53. (Base on: 2015 Mexico) Find all functions f : N → N such that
f (a + b + ab) = f (ab)
holds for all positive integers a, b
Solution. It’s natural to guess the answer is the constant function. Note that we have
f (2) = f (5) = f (11) = f (6) = f (13) = f (27) = f (9) = f (3) = f (1)
Assume that f (1) = f (2) = · · · = f (k − 1) for some k ≥ 3, then


k

 k ≡ 0 mod 3, f (k) = f 3 = f (1)

2k+1

k ≡ 1 mod 3, f (k) = f (2k + 1) = f 3
= f (1)

k ≡ 2 mod 3, f (k) = f n−2

= f (1)

3
Notice our inductive step doesn’t work for k = 2, that’s why we need to bash f (2). To find
the number k that we need to bash f (1) , . . . , f (k) for induction, one can do the inductive
step first and use the constraints to find out k.
We demonstrate this idea in the next example.
Example 54. (Own) Find all functions f : N0 → N0 such that
f (f (x) + f (y)) + 2f (xy) = xf (y + 1) + yf (x + 1) , ∀x, y ∈ N0
Solution. Stick to one of our principles, substitute zeros for variables whenever it’s possible.
Set x = y = 0, we have
f (2f (0)) + 2f (0) = 0
Since f is non-negative, f (0) must be 0. Now, in the original condition, take y = 0:
f (f (x)) = xf (1)
It’s easy to check that f ≡ 0 when f (1) = 0. For the other case, f is injective and thus,
f (1) should be 1 (why). So far, the only useful identity (for induction) is
f (f (x) + 1) + 2f (x) = 2x + f (x + 1)
IDEA: If f (x) = x, the above equation gives nothing. However, if f (x + 1) = x + 1,
f (f (x) + 1) + 2f (x) = 3x + 1
Assume f (1) = 1, . . . , f (x − 1) = x − 1, then the only possibility for f (x) is x (why)

This suggests us to proceed the mathematical induction
P (1) , . . . , P (2k) → P (2k + 2) → P (2k + 1)
where P (i) means that f (i) = i holds. So it seems that we need P (2) and the corre-
sponding equation (y = 2 in the original condition)
f (f (x) + 2) + 2f (2x) = xf (3) + 2f (x + 1)
From above, we require P (3) and 2k ≥ k + 3. That’s because we hope to get P (2k + 2)
by setting x = k + 1. In summary, it suffices to bash P (3) , . . . , P (6)
To get the value of f (3), one should notice that 4f (3) = 3f (4), combines this fact with
f (f (3) + 1) + 2f (3) = 6 + f (4)
we must have f (3) = 3. Since if f (3) > 3,
2
f (f (3) + 1) + f (3) ≥ 3 + 4 = 7 > 6
3
which is impossible. Finally, choose (x, y) = (3, 2),
f (5) + 2f (6) = 17
According to the injectivity of f , f (5) , f (6) ≥ 5 and therefore, f (5) = 5, f (6) = 6
Let’s left this exercise as the end of the section. Try induction!
Exercise. (2017 Taiwan) Determine all surjective functions f : Z → Z satisfying
f (xyz + xf (y) + yf (z) + zf (x)) = f (x) f (y) f (z)
holds for all x, y, z ∈ Z

5.2 Number Theory Facts
5.2.1 About Divisibility
We briefly review two basic knowledges of divisibility:
• If a | b for some a, b ∈ Z then | a |≤| b | or b = 0
• Every prime pk has only k + 1 factors, namely 1, p, · · · , pk
The first one extremely important because we can usually use
a | b −→ a | b − ka
to rewrite the divisibility condition to the form
g (m, n, f (m) , f (n)) | h (m, f (m))
for some functions g, h. Then, we may have h (m, f (m)) = 0 whenever g is unbounded for
a fixed m. See the following problem.
Example 55. (2017 Balkan) Find all functions f : N → N such that
n + f (m) | f (n) + nf (m)
holds for all m, n ∈ N
Solution. We have the equivalence
n + f (m) | f (n) + nf (m)

⇐⇒ n + f (m) | f (n) − n2
If there exists a s.t f (a) 6= a2 , then f is bdd. Obviously, f (1) is 1 (why). So, in particular,
n + f (1) | f (n) − n2
⇐⇒ n + f (1) | f (n) − 1
It follows f ≡ 1. In conclusion,
(
f (x) = x2 , ∀x ∈ N
f (x) = 1, ∀x ∈ N
Example 56. (2013 N1) Find all functions f : Z>0 → Z>0 such that
m2 + f (n) | mf (m) + n
holds for all positive integers m and n
Solution. In the relation, take m = n:
m2 + f (m) | mf (m) + m
−→ m2 + f (m) ≤ mf (m) + m
So when m ≥ 2, f (m) ≥ m. Then, take m = f (n),
f (n)2 + f (n) | f (n) f (f (n)) + n
which implies f (n) | n and thus, f (n) ≤ n

Clearly, we must have f (n) = n holds for all n ∈ N
The second small fact is also useful. The reason is, once we have
f (p) | pk
We have only few possibilities for f (p). Then we plug p into one variable of the condition
to determine the other values.
Solution. This time, set m = f (p) , n = p where p is a prime,
f (p)2 + f (p) | f (p) f ((p)) + p
By our assumption, f (p) = 1 or p. Notice that for sufficient large p, f (p) = 1. Otherwise,
p2 + f (n) | p + n
which is absurd. Plug n = p in the original relation (p large):
m2 + p | mf (m) + p
=⇒ m2 + p | mf (m) − m2
Consequently, f (m) = m, ∀m ∈ N
Exercise. (2016 MEMO) Find all functions f : N → N so that
f (a) + f (b) | 2 (a + b − 1) , ∀a, b ∈ N

In fact, a common strategy for divisibility problems is
• Bash f (1)
• Try to find f (p) for large primes p
• Use those values to find f (usually by establishing inequalities)

Example 57. (forget) Find all functions f : N → N such that
a + f (b) | a2 f (a) + b3
holds for all a, b ∈ N

Solution. As the second solution to 2013 N1, we tends to assert a = f (p) , b = p,
2f (p) | f (p)2 f (f (p)) + p3
So f (p) = 1, p, p2 or p3 . Because f (1) = 1 (figure out!),
p + 1 | p2 f (p) + 1
It remains two cases f (p) = p or p3 However, if f (q) = q 3 for some prime q, then
a + q 3 | a2 f (a) + q 3
This means q is ”small” (think a minute)

In other words, we know that f (p) = p for ”large” p. The remain part is left as an exercise.
Exercise. (2016 N6) Find all functions f : N → N such that
f (m) + f (n) − mn | mf (m) + nf (n)
holds for all positive integers m and n

The following problem looks somehow crazy. But it’s actually a standard one for training
induction and divisibility.
Exercise. (2008 N5) For every n ∈ N, let d(n) denote the number of (positive) divisors of
n. Find all functions f : N → N with the following properties:
d (f (x)) = x, ∀x ∈ N
and
f (xy) | (x − 1) y xy−1 f (x) , ∀x, y ∈ N
5.2.2 Euler Theorem
Solve FEs in integers as FEs in reals, we sometimes get
(
f (cn) = cf (n)
f (n + d) = f (n) + e
for some constants c, d, e. Then,

!
ck − 1 n ck − 1 n
ck f (n) = f ck n = f

n+ · d = f (n) + ·e
d d
if d | ck − 1. Notice that if c, d are co-prime, we can apply Euler theorem to find such a k
and thus, it’s easy to conclude
en
f (n) = , ∀n ∈ N
d
Theorem 14. (Euler) If n and a are co-prime positive integers, then
aφ(n) ≡ 1 mod n
Example 58. (2014 A4) Determine all functions f : Z → Z satisfying
f (f (m) + n) + f (m) = f (n) + f (3m) + 2014
for all integers m and n
Solution. For convenience, let f (0) = c. Then
f (n + c) = f (n) + 2014
IDEA: It’s natural to proceed induction and get
f (kc) = f ((k − 1) c) + 2014 = · · · = c + 2014k
We want to compare P (n, m) and P (n, m + kc) for some k. Notice that m is ”wrapped”
in two f , we need that k is a multiple of c.
The simplest choice is k = c and
f n + f c2 + f c2 = f (n) + f 3c2 + 2014 −→ c = 1007

IDEA: If we hope to use the method mentioned in the beginning of this section, then
we should perform taiwanese transformation f → g so that g (0) = 0 and show
(
g (cn) = cf (n)
g (n + d) = g (n) + e
Define g = f − 2014, then the condition can be rewritten as
g (g (m) + n) + g (m) = g (n) + g (3m)
By induction, it’s easy to show
g (n) + 2014g (m) = g (n + 1007g (m))

= g (n) + 1007g (3m) − 1007g (m)
which implies g (3m) = 3g (m), as desired.
5.2.3 Perhaps Useful

IMO, there are many number theory facts that seems to be helpful for solving FEs. However,
I can’t find any suitable examples. So I decide to list a part of them in this short section.
Theorem 15. (Chinese Remainder Theorem) Given m1 , . . . , mr ∈ Z such that (mi , mj ) = 1

for all i 6= j. Then for any c1 , . . . , cr ∈ Z There exists C satisfying
C ≡ ci (mod mi ), ∀i
Theorem 16. (Wilson) For a integer p > 1,
p | (p − 1)! + 1 ⇔ p is a prime
Theorem 17. (Burton) An integer n has a primitive root if it is of the form
2, 4, pa or 2pa
where p is an odd prime and a ≥ 1

5.3 Recursive Chain
For problems that concern the relation between f, f 2 , . . . , f k for some k. It’s natural to
consider the recursive chains:
{x, f (x) , . . . }, x = 1, 2, . . .
Definition 30. We say that s is an ancestor of t if there is a positive integer of u such

that
f (f (. . . f (s) . . .)) = t
| {z }
u
In this case, we call that t is a posterity of s
Remark 17. If a is an ancestor of b 6= a and is a posterity of b, then f contains a cycle.
Example 59. (2017 Taiwan) Find all injective functions f : N → N such that
f f (a) (b) f f (b) (a) = f (a + b)2
holds for all a, b ∈ N
Solution. Obviously, the pre-image of 1 is an empty set (why) and one shall notice that
f f (n)−1 (n) = 2n
for all n ∈ N. Indeed, just set a = b = n into the condition and use the injectivity of f . In
particular, there is an integer c such that f (c) = 2. Then
2c = f f (c)−1 (c) = f (c) = 2
So c = 1. From here, by induction, it’s easy to see that 1 is an ancestor of any power of 2.
Now, for any positive integer n, take a natural integer m so that f (n + m) is a power of 2
(why we can make such a choice?) Then, by our condition,
f f (m) (n) | f (n + m)2
Thus, f f (m) (n) is a power of 2. It follows that 1 is an ancestor of n
Exercise. Finish the solution of the above example

Example 60. (2017 IMOC) Given a positive integer k. Find all functions f : N → N s.t
f n (n) = n + k, ∀n ∈ N
IDEA: As we solve combinatorics, we shall work for small cases: For k = 1, It’s quite
clear that m is an ancestor of n whenever m < n. So if f (n) 6= n for some n ∈ N, then
we’ll have that the recursive chain of n is a non-trivial cycle i.e
| {n, f (n) , . . . } |< ∞
However, this is impossible because every recursive chain is unbounded. Now, for k = 2:
If n > m such that n ≡ m (mod 2), then n is a posterity of m. Similar to the previous
case, we should have
f (n) 6≡ n (mod 2), ∀n ∈ N
Consider a partition of N:
A = {1, 3, 5, . . . } = {2m − 1 | m ∈ N}
B = {2, 4, 6, . . . } = {2m | m ∈ N}
Then f maps A to B and maps B to A. It seems fine at the first sight. But it implies
n ∈ A −→ f 2 (n) ∈ A
Combine with the fact that f n (n) = n + 2, we conclude that f contains a cycle (why),
which is again absurd.
Solution. Consider a partition of N:

A1 = {km + 1 | m ∈ N0 }
A2 = {km + 2 | m ∈ N0 }
..
.
Ak−1 = {km + (k − 1) | m ∈ N0 }
We hope to generalize our idea, that is, find a positive integer n ∈ A1 and
m < n satisfying f m (n) ∈ A1
Indeed, just choose a large n such that
min{n, f (n) , . . . } > k
then n has the desired property (why)
Example 61. (2017 Korea) Find all functions f : N → N satisfying the following conditions:
• For every n ∈ N, f n (n) = n
• For every m, n ∈ N, | f (mn) − f (m)f (n) |< 2017
Exercise. Prove that if unbounded function f : N → N satisfies
| f (mn) − f (m) f (n) |< M, ∀m, n ∈ N
for some M , then f is multiplicative.
Solution. Since f is multiplicative, it suffices to determine f (p) for all primes p. As usual,
consider the recursive chain of p,
{p, f (p) , . . . , f p−1 (p)}
If f (p) = p, then we have nothing to do. If f (p) 6= p, then all the elements in the chain must
be distinct (why). Moreover, f p−1 (p) should be a multiple of p. That’s because the recursive
chains of p and f p−1 (p) are the same i.e.
p−1 (p)+p−2
{f p−1 (p) , f p (p) , . . . , f f (p)} = {p, f (p) , . . . , f p−1 (p)}
Assume that f p−1 (p) = cp for some c > 1,
p = f p (p) = f (cp) = f (c) f (p) > f (p)
However, we have f (p) is a multiple of p, again. Thus, f (p) = p holds for all p.
Remark 18. For any finite recursive chain,
{x, f (x) , . . . }
One could consider the minimal s ∈ N s.t f s (x) = x to get some useful results.
Exercise. (2018 Japan) Let S = {1, 2, . . . , 999} . Consider a function f : S → S such that
for any n ∈ S,
f n+f (n)+1 (n) = f nf (n) (n) = n
Prove that there exists a ∈ S, such that f (a) = a

5.4 v2 method
The main idea is
1. Establish a v2 relation
2. Use the inequality to determine some special function values
Example 62. (2017 Taiwan) Find all surjective functions f : Z → Z satisfying
f (y + xf (y)) = f (x) f (y)
holds for all x, y, z ∈ Z
Solution. First of all, it’s easy to see that f (0) = 1. Suppose that f (a) = 2, then
f (2x + a) = 2f (x)
We now claim that

v2 (f (x)) ≥ v2 (x + a) , ∀x ∈ Z
Indeed, this inequality holds when x + a is odd. Assume that it holds for all v2 (x + a) ≤ k.
For v2 (x + a) = k + 1,

x−a x−a x+a
LHS = v2 f 2 · +a = v2 f ≥ v2 = k + 1 = RHS
2 2 2
as desired. Next, we try to show that f (2ka) = 2k + 1. Suppose not, let l to be an integer
such that f (l) = 2l + 1, then take (x, y) = ((2n − 1) a, l) in the condition:
f (2n (2k + 1) a − l − 2ka − a) = (2k + 1) 2n
By using the v2 inequality,
v2 (LHS) ≥ v2 (2n (2k + 1) a − l − 2ka)
We hope that v2 (2n (2k + 1) a − l − 2ka) > n so we choose n = v2 (l − 2ka). So far, we get
f (2ka) = 2k + 1, ∀k ∈ Z
Hence, f (ka) = k + 1, ∀k ∈ Z and a = ±1. In conclusion,

(
x + 1, ∀x ∈ Z
f (x) =
−x + 1, ∀x ∈ Z
Example 63. (2018 Taiwan) Find all functions f : Z → Z such that
f (x + f (y)) f (y + f (x)) = (2x + f (y − x)) (2y + f (x − y))
Exercise. Suppose that F is a solution of this functional equation.
• Figure out f (0)
• Show that f (x + f (x)) = 2x for all x ∈ Z
• Prove that f is odd and injective.
Solution. Assume the results. We have
2f (x) = f (x) + f (f (x)) = f (2x)
We have two cases:
If f (Z) contains an odd integer, then f (x) = x, ∀x ∈ Z:

Suppose d is an odd (why) integer so that f (d) is odd. Substitute x + d for y in the
original FE, since
RHS = (2x + f (d)) (2x + 2d − f (d)) ≡ 1 (mod 2)
by our assumption. So f (x + d + f (x)) must also be odd and it follows (why)
v2 (f (x)) = v2 (x)
x−f (x)
Now, substitute 2
for y. After simplification, one shall have

x − f (x)
f x+f = 2x − f (x)
2

Notice that 2x − f (x) = f (f (x)). By injectivity, x + f x−f2(x) = f (x). Take v2 on
both side,
x − f (x) x − f (x)
v2 (f (x) − x) = v2 f = v2
2 2
From here, it’s easy to get f (x) = x for all x ∈ Z
If f (Z) contains only even integers, then f (x) = −2x, ∀x ∈ Z
Apply Taiwanese transformation, define g : Z → Z by f (x) = −2g (x). Then we can
rewrite our previous results
1. 2g (g (x)) = g (x) + x holds for all integer x
2. g (2g (x) − x) holds for all x ∈ Z
3. g is odd, injective and satisfies g (2x) = 2g (x) , ∀x ∈ Z
Furthermore, the original FE becomes
g (2g (x) − y) g (2g (y) − x) = (x + g (x − y)) (y − g (y − x))
In this equation, replace y by g (x):

2
2 x + g (x)
LHS = g (g (x)) g (2g (g (x)) − x) = g (g (x)) =
2
and 2
x + g (x)
RHS = (x + g (x − g (x))) (g (x) + g (g (x) − x)) ≤
2
So we must have
x + g (x − g (x)) = g (x) + g (g (x) − x)
which means
g (x) − x
g (x − g (x)) = , ∀x ∈ Z
2
Again, by g (2x) = 2g (x),
v2 (g (x − g (x))) ≥ v2 (g (x − g (x)))
by induction. So g (x) = x is the only possibility, which is indeed a solution.
Remark 19. This problem can be solved without using v2 , sad. Anyway, if one has
f (px + q) = pf (x)
for some prime p and q ∈ Z, then one could consider to establish a vp relation.
5.5 Problems
Problem 52. Find all functions f : Z → Z such that for all a, b, c ∈ Z,
f a3 + b3 + c3 = f (a)3 + f (b)3 + f (c)3

Problem 53. (2003 USA) Find all functions f : N → N so that
f (m + n) f (m − n) = f m2

holds for all positive integers m, n
Problem 54. (2012 IMO) Find all functions f : Z → Z such that for all a, b, c ∈ Z with
a + b + c = 0, we have
f (a)2 + f (b)2 + f (c)2 = 2f (a) f (b) + 2f (b) f (c) + 2f (c) f (a)
Problem 55. (2014 A6) Find all functions f : Z → Z such that or all n ∈ Z,
n2 + 4f (n) = f (f (n))2
Problem 56. (2004 N3) Find all functions f : N → N satisfying

2
f (m)2 + f (n) | m2 + n
for m, n ∈ N
Problem 57. (2011 N3) Let n ≥ 1 be an odd integer. Determine all functions f : Z → Z
so that
f (x) − f (y) | xn − y n
holds for all x, y ∈ Z
Problem 58. (2016 Turkey) Find all functions f : N → N such that for all m, n ∈ N,
f (mn) = f (m) f (n)
and
m + n | f (m) + f (n)
Problem 59. (2012 A5) Let f : N → N be a function. Suppose that for every n ∈ N there
exists a k ∈ N such that f 2k (n) = n + k, and let kn be the smallest such k. Prove that the
sequence k1 , k2 , . . . is unbounded.
Problem 60. (2007 N5) Find all surjective functions f : N → N such that for m, n ∈ N
and every prime p,
p | f (m + n) ⇔ p | f (m) + f (n)
Problem 61. (2015 A5) Find all functions f : Z → 2Z + 1 satisfying
f (x + f (x) + y) + f (x − f (x) − y) = f (x + y) + f (x − y)
holds for every x, y ∈ Z
Problem 62. (2018 China) Functions f, g : Z → Z satisfy
f (g (x) + y) = g (f (y) + x)
for any integers x, y. Show that if f is bounded, prove that g is periodic.
Problem 63. (2015 N6) Consider a function f : N → N with the following two properties
f n (m)−m
1. If m, n ∈ N, then n
∈N
2. The set N \ {f (n) | n ∈ N} is finite
Prove that the sequence f (1) − 1, f (2) − 2, f (3) − 3, . . . is periodic.
Problem 64. (2018 China) Let M, a, b, r be non-negative integers with a, r ≥ 2, and suppose
there exists a function f : Z → Z satisfying the following conditions:
• For all n ∈ Z, f r (n) = an + b
• For all n ≥ M , f (n) ≥ 0
• For all n > m > M , n − m | f (n) − f (m)
Show that a is a perfect r-th power
Problem 65. (2013 A5) Find all the functions f : Z≥0 → Z≥0 satisfying the relation
f (f (f (n))) = f (n + 1) + 1
for all n ∈ Z≥0
Problem 66. (2018 Kazakhstan) Does there exist a function f : N → N such that
f (mf (n)) = f (m) f (m + n) + n
holds for all m, n ∈ N

Problem 67. (2018 Taiwan) Find all functions f : N → N such that for x, y ∈ N,
f (x + yf (x)) = x + f (x) f (y)
Problem 68. (2011 Iran) Suppose that f : N → N is a function for which the expression
af (a) + bf (b) + 2ab
is always a perfect square for all a, b ∈ N. Prove that f (a) = a for all a ∈ N
Problem 69. (2013 N6) Determine all functions f : Q → Z satisfying

f (x) + a x+a
f =f
b b
for all x ∈ Q, a ∈ Z and b ∈ N
Problem 70. (2016 Taiwan) Find all functions f : Z → Z such that
f (f (x) + f (y)) + f (x) f (y) = f (x + y) f (x − y)
for all integers x, y ∈ Z

Chapter 6
Polynomial
In the chapter, we are going to discuss some problems about polynomials. The first type is
polynomial functional equation, there are two general ideas to solve such problems:
1. Consider the root with absolute value α. Show that if α 6= c for some constant c, then
we could generate infinitely many distinct roots
2. Guess the solution and apply Taiwanese transformation. After that, rewrite the equa-
tion and compare the coefficients.
The second type is about Z[x], we usually use n − m | P (n) − P (m), which follows by
x − y | xn − y n = (x − y) xn−1 + · · · + y n−1

But, as usual, we should review some basic knowledges about polynomials (one variable).
6.1 Terminology
Roughly speaking, a polynomial is a special function of the form
an xn + an−1 xn−1 + · · · + a0
for some positive integer n (called degree) and ”coefficients” a0 ∼ an . For example,
x2 + 3x + 3, x3 + 3x2 + 3x + 3, x4 + 3x3 + 3x2 + 3x + 3
are polynomials of the variable x. If the coefficients of a polynomial are all integers, then
we’ll say it an integral polynomial. For convenience, we use the notation Z[x] to represent
the set of all integral polynomials. Similarly, we have rational polynomials Q[x], complex
polynomials C[x], very complex polynomials QAQ[x], etc.
85
6.2 Factorization
Just like what we did to integers, we want to ”factorize” polynomials.
Definition 31. A polynomial P (x) is factorizable in K if there are two non-constant poly-
nomials Q (x) , R (x) ∈ K[x] such that
P (x) = Q (x) R (x)
holds for all x. (Here K can be Z, Q, R, C or anything you want) Moreover, we say that P
is divided by Q and R
For complex polynomials, this issue is not so complex. That’s because we have
Theorem 18. (Fundamental theorem of algebra) Every polynomial P (x) with complex co-
efficients has a complex root.
Combine this with the following theorem,
Theorem 19. If α is a root of P (x) i.e P (α) = 0, then
(x − α) | P (x)
Proof. Like numbers, we can perform long division algorithm.
it’s easy to see that every polynomial P (x) with complex coefficients has exactly deg (P )
complex roots. So every complex polynomial can be split into linear factors:
(x − α1 ) (x − α2 ) . . . (x − αn )
And for real polynomials, we need a small fact: If β is a complex root of P ∈ R[x], then β
is also a root of P . It turns out that every real polynomial can be split into
(x − α1 ) . . . (x − αn ) x2 + β1 x + γ1 . . . x2 + βm x + γm

Back to Q[x], Z[x]. In fact, we have no general answer, but one could provide partial results.
q
Theorem 20. If p
is a root of P ∈ Z[x] and
P (x) = an xn + an−1 xn−1 + · · · + a0
then p | an and q | a0 must hold.

Proof. Apply the previous result, we have
(px − q) | P (x)
Furthermore, we know that P (x) = (px − q) Q (x) for some Q (x) ∈ Z[x]. Finally, compare
the coefficients, it’s easily to see p | an and q | a0
Surprisingly, they are essentially the same.
Theorem 21. (Gauss) If a polynomial P ∈ Z[x] is irreducible over Q, then it’s also irre-
ducible over Z. That is, if
P (x) = A (x) B (x) =⇒ P (x) = Ã (x) B̃ (x)
where A, B ∈ Q[x] and Ã, B̃ ∈ Z[x]
Definition 32. A integral polynomial an xn + an−1 xn−1 + · · · + a0 is called primitive if
(an , an−1 , . . . , a0 ) = 1
Proof. WLOG, assume that P is a primitive polynomial. Then, choose four positive integers
pA , qA , pB , qB such that A = pqAA Ã, B = pqBB B̃. In addition, Ã, B̃ are primitive (How?). So
P (x) = A (x) B (x) −→ pA pB P (x) = qA qB A (x) B (x)
Clearly, we must have qA qB = 1 because our assumption: P is primitive. It remains to show

pA pB = 1. Suppose
Ã (x) = xm + cm−1 xm−1 + · · · + c0

B̃ (x) = xn + dn−1 xn−1 + · · · + d0
If pA pB 6= 1, one can find a prime p and i, j so that ci and dj aren’t divided by p while
ci−1 , . . . , c0 , dj−1 , . . . , d0 are divided by p. This is also based on the fact Ã, B̃ are primitive.
However, it follows that
ci+j d0 + ci+j−1 d1 + · · · + ci dj + ci−1 dj−1 + · · · + c0 di+j
isn’t divided by p, which is absurd.

According to this lemma, we only need to focus on factorization of Z[x].
6.3 Irreducibility
There are lots of criterions about irreducibility of polynomials. In this section, we list out
some of them.
Theorem 22. (Eisenstein) Given a prime p and a polynomial P ∈ Z[x] of the form
an xn + an−1 xn−1 + · · · + a0
If it satisfies two conditions
1. an 6≡ an−1 , . . . , a0 ≡ 0 (mod p)
2. a0 6≡ 0 (mod p2 )
Then P is irreducible over Z
Proof. Suppose on the contrary, P is the product of two polynomials A, B. WLOG
A (x) = xp + cp−1 xp−1 + · · · + c0
B (x) = xq + dq−1 xq−1 + · · · + d0
Assume that ci and dj aren’t divided by p while ci−1 , . . . , c0 , dj−1 , . . . , d0 are divided by p.
Then, compare the coefficients, we have
ai+j = ci dj + p × (something)
This is possible only when i + j = n. However, this contradicts to the second condition.
Example 64. (1993 IMO) Prove that for all n ∈ N≥2 ,
xn + 5xn−1 + 3
is irreducible over Z
Solution. Although there are not sufficient conditions to apply Eisenstein, we could still use
the spirit. Now, we don’t know whether xn + 5xn−1 + 3 is irreducible or not. But we could
easily factorize xn + 5xn−1 into
xn−1 (x + 5)
So we firstly work on Z3 then back to Z. In Z3 [x],
xn + 5xn−1 + 3 ≡ xn + 2xn−1 ≡ xn−1 (x + 2)
If xn + 5xn−1 + 3 is a product of two polynomials A, B, then they have the following form,
A (x) = xk , B (x) = xn−k−1 (x + 2)
in Z3 [x] Moreover, it’s clearly that k 6= 0 and k 6= n − 1. Therefore, the constant term of
AB must be divided by 9, which is absurd.
Remark 20. Whenever the polynomial becomes simpler in Zp [x] for some prime p, one can
obtain informations about the polynomial. In the previous example. after analyzing, we know
that many coefficients are divided by p = 3.
Another common way to determine the irreducibility is bounding the roots. We demon-
strate this method by proving an interesting theorem (though, useless for olympiad)
Theorem 23. (Cohn) Given a number b and a polynomial P (x) = an xn + an−1 xn−1 + · · · +
a0 ∈ Z[x] such that
1. 0 ≤ an , an−1 , . . . , a0 < b
2. (an an−1 . . . a0 )b is a prime number.
Then it’s irreducible over Z. In other words, if you have a prime in the base b, then you can
generate a irreducible polynomial.
Proof. Notice that by definition,
P (b) = an bn + an−1 bn−1 + · · · + a0 = (an an−1 . . . a0 )b
So if P is a product of two polynomials A, B, then one of |A (b) |, |B (b) | must be 1. To

derive a contradiction, we show |A (b) |, |B (b) | > 1: Indeed, factorize P into
(x − α1 ) (x − α2 ) . . . (x − αn )
Then it suffices to show (b − αi ) > 1 for every i. When |x| > b − 1 with Re (x) > 0,
|P (x) | = |an xn + an−1 xn−1 + · · · + a0 |
≥ |an xn + an−1 xn−1 | − |an−2 xn−2 | − · · · − |a0 |
≥ Re an xn + an−1 xn−1 − (b − 1) |xn−2 | + · · · + 1

b−1
≥ 1− 2 xn > 0
x −x
The following theorem leaves for the readers.

Theorem 24. (Perron) For an integral polynomial P (x) = xn + an1 xn−1 + · · · + a0 with
a0 6= 0, if
|an−1 | > 1 + |an−2 | + · · · + |a0 |
then it’s irreducible over Z
Example 65. (2007 Iran) Is there an infinite sequence of pairwise co-prime positive integers,
a0 , a1 , a2 , . . . , satisfying
Xn
ai x i
i=0
is irreducible for all n ∈ N?
Solution. Take ai to be a large prime (compare with a0 , . . . , ai−1 ). We claim that this
sequence works. If not, there exists two polynomials A, B such that
Xm
ai xi = A (x) B (x)
i=0
m
ai xi have absolute values
P
for some m. However, because am is big enough, the roots of
i=0
less than 1. In particular, it follows that a0 can’t be any positive integer, a contradiction.
We close this section by learning a special trick.
Example 66. Given a prime p, find all pairs of integral polynomials (f, g) so that
xp − 1
f (g (x)) =
x−1
Solution. The magic step is to differentiate both sides w.r.t x:
0
0 0 p−1 p p−2 p
f (g (x + 1)) g (x + 1) = x + x + ··· +
1 p−1

p−2 p p−3 p
= (p − 1) x + (p − 2) x + ··· +
1 p−2
According to Eisenstein criterion, the last polynomial is irreducible. Thus, there are two
possibilities, f 0 (g (x)) is constant or g 0 (x) is a constant. Assume that f is linear, then
xp−1 + xp−2 + · · · + x + (1 − b)
g (x) =
a
for some integers a, b. Since g has only integral coefficients, we must have a = ±1. So in
this case, our answer is
p
x −1
f (x) = ± (x + b) , g (x) = ± −b
x−1
For the other case, it’s obvious (consider the leading coefficient) that g (x) = ± (x + b) and
± (x − b)p − 1
f (x) =
± (x − b) − 1
for some integer b.
6.4 Polynomial Functional Equation
6.4.1 Naive Method: Compare the Coefficients
It’s obvious that solving polynomial function equations is equivalent to solving a system of
equations. For instance, if we want to find all polynomials satisfying
P (x)2 = P x2

We can write P (x) = an xn + an−1 xn + · · · + a0 and try to solve

2
an xn + an−1 xn−1 + · · · + a0 = a2n x2n + a2n−1 x2n−2 + · · · + a20
which means (for convenience, ai = 0 for all i < 0)

Xn a2i if i ≡ 0 (mod 2)
2
aj ai−j =
0 otherwise

j=0
This is not easy to solve. Anyway, there are still some advantages. Perhaps we could
determine the leading coefficient and the constant term by the naive method.
6.4.2 Bounding
Example 67. Find all polynomials P ∈ R[x] such that
P (x) P (x + 2) = P x2

Solution. If P is non-constant, consider any root α of P . Simply plug α into the equation,
P α2 = P (α) P (α + 2) = 0

we know that α2 is also a root of P , which implies |α| = 1. Otherwise, α, α2 , α4 , . . . are

distinct roots of P , which is absurd.
√
Here is another way to generate roots. We substitute x by α,
√ √
P α P α + 2 = P (α) = 0
There are two possibilities
√
• α is a root of P
√
• α + 2 is a root of P
Notice
√ √ that the second possibility can hold only when α = 1, so if P has a root β 6= 1, then
β, β, 4 β, . . . are distinct roots of P . which is impossible. We conclude
P (x) = 0, P (x) = 1 and P (x) = (x − 1)n for some n ∈ N
are solutions.
6.4.3 Taiwanese Transformation
Example 68. (Revisited) Find all polynomials P ∈ R[x] such that
P (x) P (x + 2) = P x2

Solution. Again, we ignore the constant answers. Notice that the leading coefficient an of
P satisfies (compare the coefficient of x2 deg(p) )
a2deg(P ) = adeg(P )
So it must be 1 (it can’t be zero, of course). Suppose that deg (P ) = n, we hope to show
P (x) = (x − 1)n under this assumption. Define Q (x) = P (x) − (x − 1)n , then it satisfies
n
(Q (x) + (x − 1)n ) (Q (x + 2) + (x + 1)n ) = Q x2 + x2 − 1

After reduction,
Q (x) Q (x + 2) + (x − 1)n Q (x + 2) + (x + 1)n Q (x) = Q x2

This equation will never happen. Because the degree (as a polynomial) of LHS is
max (2 deg (Q) , n + deg (Q)) = n + deg (Q)
and that of RHS is 2 deg (Q) < n + deg (Q)
Remark 21. The steps of this method are
1. Guess the solution
2. Find some properties by simply compare the coefficients
3. Apply Taiwanese transformation and rewrite the equation
4. Compare the degrees
Moreover, this method seems that it works when a family of answers exists.
Exercise. (2017 Taiwan) Find all polynomials P with real coefficients which satisfy
P (x) P (x + 1) = P x2 − x + 3

∀x ∈ R
6.5 n − m | P (n) − P (m)
This lemma is quite trivial and quite important. Let’s see how it works.
Example 69. Find all polynomials P (x) ∈ Z[x] satisfying for all positive integers n,
P (n) | 2n − 1
Solution. If P is not constant, choose any n ∈ N so that |P (n) | > 1. Find an arbitrary
prime p so that p | P (n). Since p | P (n + p)−P (n), we get that p | P (n + p). In particular,
p | 2n − 1, p | 2n+p − 1
which is impossible by Fermat’s little theorem.
Example 70. (2016 ELMO) Find all polynomials P (x) ∈ Z[x] such that
n | P (2n )
holds for nature numbers.
Solution. We hope to get something like
n | P (a)
and conclude that P (a) = 0. According to the condition, we have
n | P (2mn )
To use the lemma, it suffices to find m so that
2mn ≡ a (mod n)
A natural way is to choose n to be a prime because of Fermat’s little theorem. If so,
n | P (2m )
Just take n to be large enough, it follows P (2m ) = 0 for all m and so P ≡ 0
Remark 22. For many integral polynomial problems with divisibility conditions, it’s common
to take a prime that divides LHS. Then, use our crucial lemma n − m | P (n) − P (m) and
the well-known Fermat’s little theorem
ap ≡ a (mod p)
Exercise. Find all polynomials P (x) ∈ Z[x] such that
P (p) | (p − 1)! + 1
holds for all prime numbers p
Exercise. (2004 Iran) Find all integral polynomial P (x) satisfying
(m, n) = 1 → (P (m) , P (n)) = 1
Exercise. (2018 TSTST) Determine all functions θ : Z[x] → Z so that for any p, q ∈ Z[x],
• θ (p + 1) = θ (p) + 1
• If θ (p) 6= 0, then θ (p) | θ (p · q)
Example 71. (IMO 2006) Given n ∈ N and a non-constant, integral polynomial P (x).
Prove that the polynomial
Q (x) := P (P (. . . P (x) . . .))
| {z }
n
has at most deg P fixed points.
Solution. Suppose a1 > · · · > ak are fixed points of Q. Then,
ai − aj | P (ai ) − P (aj ) | · · · | Q (ai ) − Q (aj ) = ai − aj
This means P (ai ) − P (aj ) = ± (ai − aj ). By triangle inequality,

k−1
X
a1 − ak = |P (a1 ) − P (ak ) | = | (P (ai ) − P (ai+1 ))|
i=1
k−1
X
≤ |P (ai ) − P (ai+1 ) | = a1 − ak
i=1
It follows P (ai ) − P (ai+1 ) have same signs. Therefore, P (ai ) = ai + c for some constant c.
So a1 , . . . , ak are distinct roots of P − c, the conclusion follows.
Remark 23. Like FEs with divisibility conditions, don’t be hesitate to replace | by ≤
Chapter 7
Appendix
7.1 A. Removed Exercises

These problems are originally in the section 1.6. I removed them from 1.6 because of difficulty.
They are not friendly for beginners. Also, some of them are messy, one should handle lots
of cases. This is not a necessary skill for that moment (just completely understand chapter
1). But we can learn other ”unpopular” techniques by solving these functional equations.
For instance, consider the special set (fixed point or root) and analysis (not only extend to
the whole domain, what we did in chapter 2).
Problem 71. (2014 EGMO) Determine all functions f : R → R satisfying the condition
f y 2 + 2xf (y) + f (x)2 = (y + f (x)) (x + f (y))


Problem 72. (2014 Turkey) Find all functions from real numbers to itself such that for all
reals x, y,
f f (y) + x2 + 1 + 2x = y + f (x + 1)2

holds.
Problem 73. (2015 India) Find all functions f : N0 → N0 such that
f m2 + mf (n) = mf (m + n)

holds for all m, n ∈ N0

Problem 74. (2015 IZHO) Find all functions f : R → R such that
f x3 + y 3 + xy = x2 f (x) + y 2 f (y) + f (xy)

95
Problem 75. (2015 MEMO) Find all surjective functions f : N → N such that for all
positve integers a and b, exactly one of the following equations is true
f (a) = f (b)
f (a + b) = min{f (a) , f (b)}

Problem 76. (2015 Moldova) Find all functions f : N → N that satisfies
f (mf (n)) = n + f (2015m)
for all m, n ∈ N
Problem 77. (2016 Korea) Determine all functions f : R → R satisfying the following
f (xf (y) + yf (z) + zf (x)) = yf (x) + zf (y) + xf (z)
for all x, y, z ∈ R
Problem 78. (2016 MEMO) Find all functions f : R → R such that
f (x) f (y) = xf (f (y − x)) + xf (2x) + f x2


Problem 79. (2016 Iran) Find all functions f : R+ → R+ such that
f (y) f (x + f (y)) = f (x) f (xy)

Problem 80. (2016 Japan) Find all functions f : R → R such that
f (yf (x) − x) = f (x) f (y) + 2x

Problem 81. (2017 ELMO) Find all functions f : R → R such that for all real numbers
a, b and c:
1. If a + b + c ≥ 0, then
f a3 + f b3 + f c3 ≥ 3f (abc)

2. If a + b + c ≤ 0, then
f a3 + f b3 + f c3 ≤ 3f (abc)

7.2 B. Removed Examples
In the beginning of this project, I planned to use shortlist problems to demonstrate some
ideas. But I finally realized that it’s better to leave them as simulation tests for the readers.
Example 72. (2014 A2) Determine all functions f : Z → Z satisfying
f (f (m) + n) + f (m) = f (n) + f (3n) + 2014
for all integer m and n

Example 73. (2014 A6) Find all functions f : Z → Z such that
n2 + 4f (n) = f (f (n))2
for all n ∈ Z
Example 74. (2015 A2) Determine all functions f : Z → Z with the property that
f (x − f (y)) = f (f (x)) − f (y) − 1
holds for all x, y ∈ Z

Example 75. (2015 A4) Determine all functions f : R → R that satisfy the equation
f (x + f (x + y)) + f (xy) = x + f (x + y) + yf (x)
for all real numbers x, y

Example 76. (2016 A4) Find all functions f : R+ → R+ such that
xf x2 f (f (y)) + f (yf (x)) = f (xy) f f x2 + f f y 2


Solution. As usual, consider some simple assertions: P (1, 1) , P (x, 1) , P (1, y), we have
f (f (y)) = f (y) f f y 2 and xf x2 = f (x)

One should find out the following identity

f (f (x))
f f x2 = f f (x)2

=
f (x)
So we tend to show that f is injective (why) Suppose α, β ∈ R+ such that f (α) = f (β) Then
f (f (α)) f (f (β))
f f α2 = f f β2

= =
f (α) f (β)
Compare P (α, y) and P (β, y), we know f (αy) = f (βy) holds for all y ∈ R+ . Now,
αyf y 2 = αyf α2 y 2 = f (αy) = f (y) = f (βy) = βyf β 2 y 2 = βyf y 2

Thus, α = β, as desired. The injectivity implies that f (x2 ) = f (x)2 , it follows that
1
f (x) = , ∀x ∈ R+
x
which is indeed a solution!
Example 77. (2016 A7) Find all functions f : R → R such that f (0) 6= 0 and
f (x + y)2 = 2f (x) f (y) + max{f x2 + f y 2 , f x2 + y 2 }

Solution. Because there is a strange condition f (0) 6= 0, I tend to find out f (0),
P (0, 0) → 0 = f (0)2 + max{2f (0) , f (0)}
Since f (0)2 > 0, we get max{2f (0) , f (0)} < 0. So f (0) < 0 and thus, f (0) = −1. Now,
because we know the value of f (0), we tend to assert some zeros to variables
P (x, 0) → f (x)2 = −2f (x) + f x2 ⇔ f x2 = f (x)2 + 2f (x)

Replace x by −x,
f x2 = f (−x)2 + 2f (−x) → f (x) = f (−x) or f (x) + f (−x) = −2

Always remember to determine the special values. The above implies f (1) = −1 or 0
We first consider the case that f (1) = 0: In this situation, f (−1) = 0 or f (−1) = −2. We
claim that f (−1) = 0 is impossible. Suppose on the contrary, f (−1) = 0
P (x, 1) , P (x, −1) → f (x + 1)2 = f (x − 1)2
In particular, f (2) = ±1 and f (−2) = ±1. Notice that
P (1, 1) → f (2)2 = max{0, f (2)} → f (2) = 0 or 1
We conclude that f (2) = 1 and
P (x, 2) → f (x)2 = f (x + 2)2 = 2f (x) + max{f x2 + 3, f x2 + 4 } ≥ 2f (x) + 2

which is clearly impossible (why) Therefore, f (−1) = −2 and
P (x, 1) , P (x, −1) → f (x + 1)2 − f (x − 1)2 = 4f (x)
It follows that f (−2) = −3 and f (2) = 1. Again,
P (x, 2) , P (x, −2) → f (x + 2)2 − f (x − 2)2 = 8f (x)
Now, a simple trick
8f (x) = f (x + 2)2 − f (x − 2)2 = f (x + 2)2 − f (x)2

+ f (x)2 − f (x − 2)2 = 4f (x + 1) + 4f (x − 1)

tells us 2f (x) = f (x + 1) + f (x − 1). And then
4f (x) = f (x + 1)2 − f (x − 1)2 = 2f (x) (f (x + 1) − f (x − 1))
We claim that there isn’t any real b 6= 1 such that f (b) = 0

If not, there are two cases: f (−b) = 0 and f (−b) = −2. Assume f (−b) = 0,
P (b, y) , P (b, −y) → f (b + y)2 = f (b − y)2
Recall that f (x + 1)2 − f (x − 1)2 = 4f (x),
4f (b + y + 1) = f (b + y + 2)2 − f (b + y)2 = f (b − y − 2)2 − f (b − y)2 = −4f (b − y − 1)
So f (b + y + 1) + f (b − y + 1) = 0. Similarly, we have
f (−b + y + 1) + f (−b − y + 1) = 0
and one can see that f is periodic. However, f can’t be eventually periodic (why) On the
other hand, consider f is not periodic and f (−b) = −2, then
P (b, 1) , P (−b, −1) → f (b + 1)2 = f (−b − 1)2 − 8 ⇔ 2f (−b − 1) − 2f (b + 1) = 8
So f (b + 1) = 1 and according to f (b + y + 1) + f (b − y − 1) = 0, we have
f (b − 1) = −1, f (2b − 2) = 2
Then,
P (x, b − 1) , P (x, 1 − b) → f (x + b − 1)2 = f (x − 1 + b)2
P (x, 2b − 2) → f x2 + (2b − 2)2 = f x2

which is again a contradiction. From this result, we have
f (x + 1) − f (x − 1) = 2
for all x ∈ R. Now if f (a) = f (−a) for some non-zero a,
P (a, a) , P (a, −a) → f (2a) = ±1
So f (2x ± 1) = 0 and thus, 2a ± 1 = 1 → a = 0 or 1, absurd. Consequently,
f (x) + f (−x) = −2, ∀x ∈ R
and this also implies f (x) f (−x) = −f (x2 ) by Vieta’s theorem.
P (x, −x) → 1 = −2f x2 + f 2x2

In other words, f (2x) = 2f (x) + 1 whenever x ≥ 0. But we have a relation between f (x)
and f (−x), namely f (x) + f (−x) = −2. We get f (2x) = 2f (x) + 1. Finally
P (x, y) , P (x, −y) → (f (x + y) + f (x − y)) (f (x + y) − f (x − y)) = 2f (x) (f (y) − f (−y))
Replace x, y by 2x, 2y, respectively, and use f (2x) = 2f (x) + 1,
f (x + y) + f (x − y) = 2f (x) ⇔ (f (x + y) + 1) + (f (x − y) + 1) = (f (2x) + 1)
As the result, the function g = f + 1 satisfies the condition of theorem 1 (why) and it follows
f (x) = x − 1 holds for all x ∈ R
Next, let’s consider the case that f (1) = −1: We still want to know f (2) , f (−2),
P (1, −1) → f (2) = −1, f (−2) = −1
Then, as usual, we assert y = 2,
P (x, 2) → f x2 = f x2 + 4

From
P (x, y) , P (x, −y) → f (x + y)2 − f (x − y)2 = 2f (x) (f (y) − f (−y))
Given any y > 0, choose a sufficient large x with f (x) = −1 (how) then we get
f (−y − 4) = f (−y)
One conclude that f (x) = f (x + 4) holds for all x ∈ R

Because we expect the only solution in this case is the constant function −1, we thus want
to show f (x) = f (−x). Suppose that there is a real a such that k = f (a) − f (−a) 6= 0,
P (x, a) , P (x, −a) → f (x + a)2 − f (x − a)2 = 2kf (x)
If f (2a) 6= f (−2a), then we can apply the previous simple trick
P (a, a) , P (a, −a) → f (2a)2 − f (−2a)2 = 2 f (a)2 − f (−a)2

which is non-zero (why) So
f (x + 2a)2 − f (x − 2a)2 = 2 (f (2a) − f (−2a)) = 4kf (x)
Thus, it’s easily to see 2f (x) = f (x + a) + f (x − a) and f must satisfy f (x + a) = f (x)

since it’s already periodic. But this is absurd since f (a) 6= f (−a). Therefore, f is even
P (x, y) , P (x, −y) → f (x + y)2 = f (x − y)2
and thus (why)

f (x) = −1 for all x ∈ R
Example 78. (2017 A6) Find all functions f : R → R such that
f (f (x) f (y)) + f (x + y) = f (xy)
Solution. First of all,
P (x, 0) → f (f (x) f (0)) + f (x) = f (0)
Clearly, if f (0) = 0, then f ≡ 0. So we consider the case that f (0) 6= 0. To get a simple
expression, we solve the equation
x
x + y = xy ⇔ x 6= 1 and y =
x−1
Then
x x
P x, →f f (x) f = 0, ∀x 6= 0
x−1 x−1
In particular, f (x) 6= 0 whenever x 6= 1
It follows that f (1) = 0 and f (0) = ±1. Assume f (0) = −1
P (x, 1) → f (x + 1) = f (x) + 1
Apply ”Taiwanese Transformation”: g = f + 1, the condition becomes
g (g (x) g (y) − g (x) − g (y)) + g (x + y) = g (xy)
Since g (x + 1) = g (x) + 1,
Q (x, y) : g (g (x) g (y)) + g (x + y) = g (xy + x + y)
To avoid complicate calculation, substitute 1, −1 for y:
g (g (x)) + g (x + 1) = g (2x + 1) → g (g (x)) + g (x) = g (2x)
g (−g (x)) + g (x − 1) = g (−1) → g (−g (x)) = −g (x)
Combine those results, we have g (−2g (x)) = −2g (x) and thus,
Q (x, −2) → g (−2g (x)) + g (x − 2) = g (−x − 2) → g is odd
Consequently, g (g (x)) = g (x) and g (nx) = ng (x) , ∀n ∈ N by easy induction. Note that
Q (nx, ny) → ng (g (x) g (y)) + g (x + y) = g (nxy + x + y)
Now, consider any two reals a < b, we hope to find m and x, y such that

x + y = a

x + y + mxy = a+b2

x + y + 2mxy = b

That is,
b−a
x + y = a, xy =
2m
We just need show the discriminant is non-negative. The discriminant is
4 (b − a)
a2 − >0
2m
Take m be a sufficient large integer, we get the desired result and so g is Cauchy! From
Q (x, y), we have
g (g (x) g (y)) = g (xy)
Finally, assume g (α) = g (β), then g (α − β) = 0 → α − β = 0 (notice that g is injective at
the point 0) and g (x) = x. In other words,
f (x) = x − 1, ∀x ∈ R
The case that f (0) = 1 can be solved in a similar way and get another solution:
f (x) = 1 − x, ∀x ∈ R

Functional Equations PDF

Uploaded by

Copyright:

Available Formats

Functional Equations PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Functional Equations PDF

Uploaded by

Copyright:

Available Formats

Functional Equation

5 Functional Equation in N/Z 67

• clavin9403: Correc typos and give some advises.

• USJL: Suggest suitable examples and exercises.

• Z: the set of all integers

• Q: the set of all rational numbers

• R: the set of all real numbers

• R+ : the set of all positive reals

1.1 What are functional equations ?

This is a special case of functional equations. To be precise, we translate the question to

Example 1. Find all functions f : {1, 2} → R such that

Example 2. Find all functions f : R → R such that

2f (x) − 5f (y) = 8, ∀x, y ∈ R

which gives f (0) = − 83 .

Example 3. Find all functions f : R → R such that

Solution. For any a, the identities in which f (a) appears are

Remark 2. In fact, our central idea is:

• If f (a) is unknown for some a, choose an equation in which f (a) appears

• Solve the system of equations

Example 6. Find all functions f : N → N so that

holds for all x, y ∈ N

Solution. Notice that q

g (x + y)3 = g (x)3 + g (y)3 + 3 (x + y) g (x) g (y)

(x, y) = (1, 1) , (1, 2) , (1, 3) , (2, 2)

The remains are left as an exercise.

|x|f (y) + yf (x) = f (xy) + f x2 + f (f (y))

Solution. First, consider the assertion (x, y) = (1, 1), we get

2f (1) = 2f (1) + f (f (1)) → f (f (1)) = 0

Definition 1. A function f : A → B is called injective if f (a1 ) = f (a2 ) implies a1 = a2

Obviously, whenever we prove that a function is injective, the equation

Example 8. Find all functions f : Z → Z that satisfy f (0) = 1 and

for all integers n

Solution. Notice that f (f (n)) = n implies f is injective (why) Therefore,

f (f (n)) = n and f (2m) = 1 − 2m

f (2l + 1) = f (1 − 2 (−l)) = f (f (2 (−l))) = −2l = 1 − (2l + 1)

So we conclude that f (n) = 1 − n for all n ∈ Z

Remark 3. For convenience, let P (x1 , x2 , · · · , xn ) be the assertion of a functional equation

Solution. If f (a) = f (b) for some a, b ∈ Z. Compare y = a, b in the equation:

f (x) + a = f (x + f (a)) = f (x + f (b)) = f (x) + b → a = b

f (x) = x, ∀x ∈ Z or f (x) = −x, ∀x ∈ Z

Remark 4. Here is a useful result: If g ◦ f is injective, then so is f

since g ◦ f is assumed to be injective. Thus, we conclude that f must be injective.

Example 10. Find all functions f, g : R → R such that g is injective and

f (0) = g (x + f (−g (x)))

There must be a constant c such that

f (y) = g (a + f (y)) → y − a = g (y) , ∀y ∈ R

by using the surjectivity of f . Substitute into the original condition,

which is easy to get f is a linear function (left as an exercise)

Definition 3. A function f : A → B is called bijective if f is both injective and surjective.

P (x, −f (x)) → f (f (−f (x)) − x) = f (0) − 2x

P (a, y) , P (b, y) → 2a + f (f (y) − a) = 2b + f (f (y) − b)

We hope a = b, which is equivalent to f (f (y) − a) = f (f (y) − b) for some y. Note that, so

LHS = f (x + f (0) + y) = x + y + 2f (0)

(y + 1) f (x) + f (xf (y) + f (x + y)) = y

Solution. First of all, rewrite the condition,

f (xf (y) + f (x + y)) = (1 − f (x)) y − f (x)

So f is surjective (why) Now, assume a, b ∈ R s.t. f (a) = f (b). Consider

P (0, a) → (a + 1) f (0) + f (f (a)) = a

P (0, b) → (b + 1) f (0) + f (f (b)) = b