New Zealand Mathematical Olympiad Committee

Functional equations
Arkadii Slinko

1 Composition of functions. Groups of functions.

Let f : R → R and g : R → R be two functions. Then the composition of f and g, written f ◦ g, is the function
defined by (f ◦ g)(x) = f (g(x)).
Note that the domain of f ◦ g consists of all x from the domain of g such that g(x) belongs to the domain of f .
Exercise 1. Find the domains of f ◦ g and g ◦ f , where
√
f (x) = 1 − x2 , g(x) = x.

Exercise 2. For the functions

x−2 2x − 3
f (x) = , g(x) =
3x + 4 5x − 1
find f ◦ g and g ◦ f and their domains.
Exercise 3. Find functions f, g such that f ◦ g 6= g ◦ f . Prove that for all functions f, g, h it is true that
(f ◦ g) ◦ h = f ◦ (g ◦ h).
Exercise 4. Let
x
f (x) = √ .
1 − x2
Find the n-fold composition f (n) (x) = f ◦ f ◦ f ◦ . . . ◦ f (x).

We say that two functions f and g commute if f ◦ g = g ◦ f .

Exercise 5. Prove that for every function f the functions f (n) andf (m) commute for all m, n.

The function e(x) = x has the property that e ◦ f = f ◦ e = f for any other function f ; e is sometimes called the
identity function. A function g is called the inverse of f if f ◦ g = g ◦ f = e or equivalently f (g(x)) = g(f (x)) = x
for all x ∈ R. In this case we denote the inverse of f by f −1 .
A set G of functions is called a group if

1. e ∈ G;
2. If f, g ∈ G, then f ◦ g ∈ G; and,
3. If f ∈ G, then f −1 ∈ G.
Exercise 6. Prove that the set G of linear functions f (x) = ax + b, a 6= 0, b arbitrary, is a group.
Exercise 7. Prove that the functions
g1 = x, g2 = 1 − x
form a group. Write down the “multiplication table” for this group.
Exercise 8. Prove that the functions
1 x−1
g1 = x, g2 = , g3 =
1−x x
form a group. Write down the “multiplication table” for this group.

1
Exercise 9. Prove that the functions
x−1 1 x+1
g1 = x, g2 = , g3 = − , g4 =
x+1 x 1−x
form a group. Write down the “multiplication table” for this group.

We will use Exercise 7 to solve the functional equation

2f (1 − x) + 1 = xf (x).

We substitite 1 − x for x. As these two functions form a group we will again have an equation relating f (x) and
f (1 − x). Namely,
2f (x) + 1 = (1 − x)f (1 − x).
Now, making f (1 − x) the subject of the first equation and substituting it into the second, we get
1
2f (x) + 1 = (1 − x)f (1 − x) = (1 − x) · (xf (x) − 1)
2
and
x−3
f (x) = .
x2 − x + 4
This is a consequence of the original equation. Therefore we have to check that this function is indeed a solution.
This can be easily done. I am not doing it here but it is a necessary part of the solution!
 
1
Exercise 10. Solve the functional equation 2xf (x) + f 1−x = 2x.
 
x−1
Exercise 11. Solve the functional equation xf (x) + 2f x+1 = 1.

2 Cauchy’s equation.
The equation
F (x + y) = F (x) + F (y)
is called the Cauchy equation. It is clear that any linear homogeneous function F (x) = ax satisfies this equation.

Exercise 12. Let F (x) be any function satisfying the Cauchy equation. Prove that for a = F (1) we have
F (x) = ax for all rational numbers x.
Exercise 13. Prove that a continuous function satisfying the Cauchy equation must be a function F (x) = ax
for all real numbers x and a = F (1).

But apart from linear functions there are many discontinuous solutions to the Cauchy equation.
In 1905 the German mathematician G. Hamel published in Mathematische Annalen the following result: he
constructed a subset G ⊆ R with the property that every real number x ∈ R can be uniquely represented as a
finite linear combination
x = n1 g1 + n2 g2 + · · · + nk gk
of some g1 , g2 , . . . gk ∈ G with nonzero integer coefficients n1 , n2 , . . . nk . The construction of this set G, which
is known now as Hamel’s basis, is quite tricky and cannot be given here.
Using this basis we can construct infinitely many solutions to the Cauchy equation, as follows. Let us put an
arbitrary real number a(g) in correspondence to an element g ∈ G and define

F (x) = n1 a(g1 ) + n2 a(g2 ) + · · · + nk a(gk ).

2
To prove that F (x) satisfies the Cauchy equation let us take also

y = m1 g1 + m2 g2 + · · · + mk gk .

Without loss of generality we may assume that y is a linear combination of the same elements of G because we
can always add some additional terms with zero coefficients. As

x + y = (n1 + m1 )g1 + (n2 + m2 )g2 + · · · + (nk + mk )gk

we have
F (x + y) = (n1 + m1 )a(g1 ) + (n2 + m2 )a(g2 ) + · · · + (nk + mk )a(gk ) = F (x) + F (y),
so F satisfies the Cauchy equation, as claimed.
Discontinuous solutions are strangely behaved, and very hard to imagine. In fact, the graph of any discontinuous
solution to the Cauchy equation is everywhere dense in the plane! As an example, we will prove one result
about discontinous solutions.
Theorem 1. Let F (x) be a function satisfying the Cauchy equation. Suppose that F (x) is bounded from above
at least on one interval (p, p + s), where s is a positive number. Then F (x) = ax for all real numbers x, where
a = F (1).

Proof. If the function F (x) is bounded on the interval (p, p + s), then it is bounded from above on the interval
(0, s). Indeed, for an arbitrary x ∈ (0, s) there exists an upper bound M such that F (p + x) < M . Thus
F (p + x) = F (p) + F (x) < M and F (x) < M − F (p) for an arbitrary x ∈ (0, s). We may assume that s is
rational. Let us consider now another function
F (s)
G(x) = F (x) − x.
s
Clearly this function also satisfies the Cauchy equation and it is also bounded on the interval (0, s). In addition
we have G(s) = 0 and hence
G(x + s) = G(x) + g(s) = G(x)
which means that s is a period of G(x)! It follows now that the function G(x) is bounded on the whole real line
as every value of G(x) is equal to some value of this function on the interval (0, s). The function G(x) must now
be identically zero. Indeed, if G(x0 ) 6= 0, then G(nx0 ) = nG(x0 ), so G gets arbitrary large for large n which
contradicts the fact that G(x) is bounded. Thus F (x) = F (s)s x = F (1)x. The theorem is proved.

Exercise 14. Analyse the Cauchy equation

F (x + y) = F (x)F (y)

by reducing it to the Cauchy equation by means of logarithmic transformation.

3 Problems
1. (Russia 1977)
(a) For an arbitrary α ∈ R find all polynomials of degree at most three, which commute with the
polynomial P (x) = x2 − α.
(b) Prove that there exists no more than one polynomial of degree n which commutes with the quadratic
polynomial P (x) = x2 + px + q.
(c) Given a quadratic polynomial P (x) = x2 + px + q, find all polynomials of degrees four and eight
which commute with P (x).
(d) Prove that if two polynomials Q(x) and R(x) both commute with a fixed polynomial P (x) = x2 +px+q
of degree 2, then they commute, i.e. Q ◦ R = R ◦ Q.

3
 (e) Prove that there exists a sequence of polynomials P1 , P2 , . . . Pk , . . . , such that the degree of Pk is
equal to k, P2 (x) = x2 − 2, and every two polynomials in this sequence commute. Hint: Use the
Chebycheff polynomials Tk (x) for which Tk (cos t) = cos kt.
2. (IMO 1973) Let P1 (x) = x2 − 2 and Pj (x) = P1 (Pj−1 (x)) for j = 2, 3, . . . . Prove that for any n ∈ N all
roots of the equation Pn (x) = x are real and distinct.
3. (IMO 1973) A nonempty set G of linear functions f (x) = ax + b, where a and b are real numbers, does
not contain constant functions and has the following properties:

(i) For arbitrary f, g ∈ G, we have f ◦ g ∈ G, i.e. the set G is closed under composition of functions;
(ii) If f (x) = ax + b ∈ G, then f −1 (x) = a1 (x − b) ∈ G; and,
(iii) For every f ∈ G there exists a number af ∈ R such that f (af ) = af .
Prove that there exists a number a ∈ R such that f (a) = a for all f ∈ G.

4. (APMO 1989) Determine all functions from the reals to the reals for which
(i) f (x) is strictly increasing; and,
(ii) f (x) + g(x) = 2x for all real x, where g(x) is the composition inverse function to f (x).
(Note: f and g are said to be composition inverses, if f (g(x)) = x and g(f (x)) = x for all real x.)

5. (IMO Shortlist 1997)

(a) Do there exist functions f : R → R and g : R → R such that

f (g(x)) = x2 and g(f (x)) = x3 for all x ∈ R?

(b) Do there exist functions f : R → R and g : R → R such that

f (g(x)) = x2 and g(f (x)) = x4 for all x ∈ R?

(c) Find all functions f with the following properties: for all t,
 
1 1
f (−t) = f (t), f (t + 1) = f (t) + 1, f = 2 f (t).
t t

January 12, 2004

http://www.mathsolympiad.org.nz