Advanced Lecture Notes on Functions

Part 10 (Continued): Functional Equations II

Advanced Techniques and Problem-Solving Strategies

For the Serious Aspirant

Contents

1 Recap: Foundations and Cauchy’s Equations 1

2 Leveraging Established Functional Properties 1

2.1 Using Injectivity and Surjectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

3 Advanced Substitution and Iteration 2

3.1 Symmetry Arguments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
3.2 Iteration and Recurrence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

4 Using Calculus: Differentiability 3

5 Domain and Codomain Considerations (N, Z, Q) 4

6 Problems for the Adept: Part 10-II 5

7 Answer Key (Part 10-II) 6

Part 10-II: Functional Equations - Adv. Techniques Advanced Functions by Yuki

1 Recap: Foundations and Cauchy’s Equations

Part 10-I laid the essential groundwork for functional equations, focusing on fundamental substitution
techniques and a rigorous exploration of Cauchy’s additive, exponential, logarithmic, and multiplicative
functional equations, along with Jensen’s equation. We emphasized how solutions over Q can be
extended to R under conditions like continuity, monotonicity, or boundedness on an interval. We now
build upon this foundation by exploring more advanced techniques and tackling more complex equation
structures.

2 Leveraging Established Functional Properties

Often, the first crucial step after initial substitutions is to deduce general properties of the unknown
function f .

2.1 Using Injectivity and Surjectivity

• Proving Injectivity/Surjectivity from the FE: As seen in Evan Chen’s notes (Example 10 for
injectivity, Example 11 for surjectivity), certain structures allow deduction of these properties.
For example, if f (g(x) + y) = g(x + f (y)) and g is injective, then f (0) = g(x + f (−g(x)))
implies x + f (−g(x)) must be constant. This can lead to f being surjective.

• Using Injectivity: If f (expr1 ) = f (expr2 ) is derived and f has been proven injective, then
expr1 = expr2 . This often simplifies the equation significantly.

• Using Surjectivity: If f is surjective from X to Y , then for any y0 ∈ Y , there exists x0 ∈ X

such that f (x0 ) = y0 . This allows replacement of f (variable) with a new arbitrary variable
spanning Y . E.g., if f : R → R is surjective, in an expression like P (f (x), y), we can replace
f (x) with a new variable z ∈ R.

1
Part 10-II: Functional Equations - Adv. Techniques Advanced Functions by Yuki

Example 2.1. Using Involution

(Revisit from Part 10-I, Evan Chen Ex 2.1) Find all functions f : R → R such that f (f (x)2 +
f (y)) = xf (x) + y for all x, y ∈ R.

Proof
Denote the assertion as P (x, y). P (0, y) =⇒ f (f (0)2 + f (y)) = 0 · f (0) + y = y.
This shows that f is surjective. Since f is surjective, there exists u such that f (u) = 0.
P (u, y) =⇒ f (f (u)2 + f (y)) = uf (u) + y =⇒ f (02 + f (y)) = u · 0 + y =⇒
f (f (y)) = y for all y ∈ R. This means f is an involution and hence bijective. Since
f (f (y)) = y:

• f (u) = 0 =⇒ f (0) = u.

• f (0) = c =⇒ f (c) = 0. (So u = c).

The original equation: f (f (x)2 + f (y)) = xf (x) + y. Using f (f (y)) = y, apply f to both
sides: f (x)2 + f (y) = f (xf (x) + y). (Equation *) P (x, 0) : f (f (x)2 + f (0)) = xf (x).
(1) P (0, x) : f (f (0)2 + f (x)) = x. (2) If f (0) = 0: From (1): f (f (x)2 ) = xf (x).
Since fà (x)2 ó 0, let t = f (x) 2
√ f (t) = xf (x) where t ≥ 0. This doesn’t mean
√ . Then
f (t) = tf ( t) or f (t) = − tf (− t). From (2): f (f (x)) = x. This is our involution
property. If f (0) = 0, from Equation *: f (x)2 + f (y) = f (xf (x) + y). Let f (x) = x.
Then x2 + y = xy + y. True only if x2 = xy, i.e., x = y or x = 0. Not general. Let
f (x) = −x. Then (−x)2 + (−y) = −(x(−x) + y) =⇒ x2 − y = −(−x2 + y) = x2 − y.
This works. So f (x) = −x is a solution (and f (0) = 0).
Let’s re-examine Evan Chen’s path if f (0) = 0: f (f (y)) = y. Original: f (f (x)2 +f (y)) =
xf (x) + y. Apply f to both sides of original: f (x)2 + f (y) = f (xf (x) + y). Set y = 0:
f (x)2 + f (0) = f (xf (x)). Since f (0) = 0, f (x)2 = f (xf (x)). If f (x) = x: x2 =
f (x2 ) = x2 . (Consistent) If f (x) = −x: (−x)2 = f (x(−x)) = f (−x2 ) = −(−x2 ) = x2 .
(Consistent) The known solutions are f (x) = x and f (x) = −x. Proving there are no
others often involves showing f (x2 ) = f (x)2 or f (x2 ) = −f (x)2 and handling cases,
which can be very intricate. The key insight is that f (f (y)) = y is extremely powerful.

3 Advanced Substitution and Iteration

3.1 Symmetry Arguments
If an equation P (x, y) is not symmetric in x and y:

1. Write down P (x, y).

2. Write down P (y, x) by swapping x and y.

3. Equate, subtract, or add these two equations to potentially eliminate terms or derive simpler
relations.

2
Part 10-II: Functional Equations - Adv. Techniques Advanced Functions by Yuki

Example 3.1. Symmetrization to Cauchy

Let f : R → R satisfy f (x + f (y)) = f (x) + y.

Proof
Denote the assertion as P (x, y). P (x, y) : f (x + f (y)) = f (x) + y. P (y, x) : f (y +
f (x)) = f (y) + x. (Not used here for symmetry, but for other insights).
Step 1: Prove f is injective. Suppose f (y1 ) = f (y2 ). Then f (x + f (y1 )) = f (x + f (y2 )).
Using P (x, y1 ): f (x) + y1 = f (x + f (y1 )). Using P (x, y2 ): f (x) + y2 = f (x + f (y2 )).
So f (x) + y1 = f (x) + y2 =⇒ y1 = y2 . Thus f is injective.
Step 2: Prove f is surjective. The RHS f (x) + y can take any real value (fix x, let y vary
over R). So the range of f is R. Thus f is surjective.
Step 3: Find f (0). Since f is surjective, there exists c such that f (c) = 0. P (x, c) =⇒
f (x + f (c)) = f (x) + c =⇒ f (x + 0) = f (x) + c =⇒ f (x) = f (x) + c =⇒ c = 0.
So f (0) = 0.
Step 4: Find f (f (y)). P (0, y) =⇒ f (0 + f (y)) = f (0) + y =⇒ f (f (y)) = y (since
f (0) = 0). So f is an involution.
Step 5: Reduce to Cauchy. Original equation: f (x+f (y)) = f (x)+y. Using y = f (f (y))
(since f (f (y)) = y), we replace y on the RHS: f (x + f (y)) = f (x) + f (f (y)). Let
f (y) = z. Since f is surjective, z can be any real number. Then f (x + z) = f (x) + f (z)
for all x, z ∈ R. This is Cauchy’s additive equation.
Step 6: Solve Cauchy with involution property. We have f (x + z) = f (x) + f (z) and
f (f (x)) = x. From Cauchy over R (no continuity assumed yet), we know f (qx) = qf (x)
for q ∈ Q. f (1 · f (1)) = f (f (1)) = 1. Also f (1 · f (1)) = f (1)f (1) = (f (1))2 if f (1)
were rational? No, this is f (c · x) = cf (x) for c ∈ Q. f (f (1)) = 1. Let f (1) = c.
Then f (c) = 1. Also f (c) = cf (1) = c · c = c2 . So c2 = 1 =⇒ c = ±1. Thus
f (1) = 1 or f (1) = −1. If f (1) = 1, then f (q) = q for q ∈ Q. If f (1) = −1, then
f (q) = −q for q ∈ Q. If we assume continuity (or monotonicity, or boundedness on an
interval), then f (x) = x or f (x) = −x for x ∈ R. Both f (x) = x and f (x) = −x satisfy
f (f (x)) = x. Check in original: If f (x) = x: x + y = x + y. (Solution). If f (x) = −x:
−(x + (−y)) = −x + y =⇒ −x + y = −x + y. (Solution). These are the solutions
under regularity conditions. Without them, other pathological additive solutions exist but
would not satisfy f (f (x)) = x unless f (x) = x or f (x) = −x.

3.2 Iteration and Recurrence

Iterating a function or an equation can reveal patterns or lead to recurrence relations.

Example 3.2. Periodicity from Iteration

1+f (x)
1+f (x) 1+f (x+a) 1+ 1−f (x) 1−f (x)+1+f (x)
If f (x + a) = 1−f (x)
for some a > 0. f (x + 2a) = 1−f (x+a)
= 1+f (x) = 1−f (x)−(1+f (x)) =
1− 1−f (x)
2
−2f (x)
= − f (x)
1
. f (x + 3a) = − f (x+a)
1
= − (1+f (x))/(1−f
1
(x))
= − 1−f (x)
1+f (x)
. f (x + 4a) = − f (x+2a)
1
=
− −1/f (x)
1
= f (x). So f (x) is periodic with period 4a.

4 Using Calculus: Differentiability

Assuming differentiability (if given or deducible) can transform a functional equation into a differential
equation.

3
Part 10-II: Functional Equations - Adv. Techniques Advanced Functions by Yuki

Example 4.1. Reducing to ODE

Find all differentiable functions f : R → R such that f (x + y) = f (x) + f (y) + 2xy.

Proof
P (0, 0) =⇒ f (0) = f (0) + f (0) + 0 =⇒ f (0) = 0. Differentiate with respect to
x (treating y as constant): f ′ (x + y) · ∂x
∂
(x + y) = f ′ (x) + 0 + 2y =⇒ f ′ (x + y) =
f ′ (x) + 2y. Now, differentiate this result with respect to y (treating x as constant):
f ′′ (x+y)· ∂y
∂
(x+y) = 0+2 =⇒ f ′′ (x+y) = 2. Since this holds for all x, y, let z = x+y.
Then f ′′ (z) = 2 for all z ∈ R. Integrating f ′′ (z) = 2 =⇒ f ′ (z) = 2z + A for some
constant A. Integrating f ′ (z) = 2z + A =⇒ f (z) = z 2 + Az + B for some constant B.
Since f (0) = 0, we have 02 + A(0) + B = 0 =⇒ B = 0. So f (x) = x2 + Ax. Substitute
back into the original equation: (x + y)2 + A(x + y) = (x2 + Ax) + (y 2 + Ay) + 2xy.
x2 + 2xy + y 2 + Ax + Ay = x2 + Ax + y 2 + Ay + 2xy. This identity holds for any constant
A. So, f (x) = x2 + Ax for any A ∈ R are the differentiable solutions.

5 Domain and Codomain Considerations (N, Z, Q)

Techniques shift when domains are discrete.

• N: Induction is very powerful. Finding f (1) is often key.

• Z: Extend from N using f (0) and f (−n).

• Q: Extend from Z using f (p/q).

Divisibility and number theoretic arguments become relevant (to be explored more in Part 10-III).

Example 5.1. Functional Equation over N

Find all functions f : N → N such that f (f (n) + m) = n + f (m + 2018) for all n, m ∈ N. (This
is a significantly harder problem, e.g., IMO Shortlist style. Solution is usually f (n) = n + c).

Solution approach sketch

Step 1: Prove f is injective. Suppose f (n1 ) = f (n2 ). Then f (f (n1 ) + m) = n1 + f (m +

2018) and f (f (n2 ) + m) = n2 + f (m + 2018). Since LHS are equal, n1 + f (m + 2018) =
n2 + f (m + 2018) =⇒ n1 = n2 . So f is injective.
Step 2: Show f (x) = x + c for some c. Let f (1) = k. P (n, 1) : f (f (n) + 1) =
n + f (2019). If f is x + c, then (n + c) + c + 1 = n + (2019 + c) =⇒ n + 2c + 1 =
n + 2019 + c =⇒ c = 2018. So, a candidate solution is f (n) = n + 2018. LHS:
f (f (n) + m) = f ((n + 2018) + m) = (n + m + 2018) + 2018 = n + m + 4036.
RHS: n + f (m + 2018) = n + ((m + 2018) + 2018) = n + m + 4036. LHS=RHS. So
f (n) = n+2018 is a solution. Proving it’s the only one for N → N often involves showing
f (n + 1) = f (n) + 1 by using injectivity and clever substitutions, then determining f (1).
This is non-trivial.

4
Part 10-II: Functional Equations - Adv. Techniques Advanced Functions by Yuki

6 Problems for the Adept: Part 10-II

Problem 6.1. prob:P10II.1

Find all functions f : R → R such that f (x2 − y 2 ) = (x − y)(f (x) + f (y)) for all x, y ∈ R.

Problem 6.2. prob:P10II.2

Find all functions f : R → R such that f (xf (y) + x) = xy + f (x) for all x, y ∈ R.

Problem 6.3. prob:P10II.3

Find all functions f : R → R satisfying f (x2 + y) = f (f (x) − y) + 4f (x)y.

Problem 6.4. prob:P10II.4

Find all functions f : Q+ → Q+ such that for all x, y ∈ Q+ :

y
f (x + f (y)) = f (x) + f (x)
f (y)

(Assume f (y) is never zero, which is guaranteed by codomain Q+ ).

Problem 6.5. prob:P10II.5

Find all differentiable functions f : R → R such that f (0) = 1 and for all x, y ∈ R:

f ′ (x) = f (x + y) − f (y) − xf (y) − x2 /2

5
Part 10-II: Functional Equations - Adv. Techniques Advanced Functions by Yuki

7 Answer Key (Part 10-II)

Solutions (Part 10-II)

1. Problem 10II.1: f (x2 − y 2 ) = (x − y)(f (x) + f (y)). (Solved in Part 10-I problem 10.3,
answer f (x) = cx).

2. Problem 10II.2: f (xf (y) + x) = xy + f (x). (Solved in Part 10-I problem 10.2, answers
f (x) = x, f (x) = −x).

3. Problem 10II.3: f (x2 + y) = f (f (x) − y) + 4f (x)y. Solutions are f (x) = 0 and

f (x) = x2 . Proving uniqueness is very hard. If f (x) = 0, 0 = 0 + 0, true. If f (x) = x2 :
LHS: f (x2 + y) = (x2 + y)2 = x4 + y 2 + 2x2 y. RHS: f (f (x) − y) + 4f (x)y = f (x2 −
y) + 4x2 y = (x2 − y)2 + 4x2 y = x4 + y 2 − 2x2 y + 4x2 y = x4 + y 2 + 2x2 y. Matches. The
proof often involves showing f (x) ≥ 0, f (0) = 0. Then P (0, y) =⇒ f (y) = f (−y), so
f is even. Then showing surjectivity for f (x) − y or injectivity on certain domains.

4. Problem 10II.4 (Yuki’s Challenge): f : Q+ → Q+ , f (x + f (y)) = f (x)(1 + f (y) y

).
y y
Let P (x, y) be the assertion. The term 1 + f (y) depends only on y. Let Cy = 1 + f (y) .
So f (x + f (y)) = Cy f (x). This has the form f (x + K) = CK f (x) where K = f (y) and
CK = Cy . If f (y) = y is a solution: LHS: f (x + y) = x + y. RHS: f (x)(1 + y/y) =
x(1 + 1) = 2x. x + y = 2x =⇒ y = x. This must hold for all x, y, which is false. So
f (x) = x is not a solution.
If f (x) = c (constant, c ∈ Q+ ): c = c(1 + y/c) = c + y. This means y = 0, but y ∈ Q+ .
So no constant solution.
P (x1 , y) =⇒ f (x1 + f (y)) = f (x1 )Cy . P (x2 , y) =⇒ f (x2 + f (y)) = f (x2 )Cy . So
f (x1 +f (y))
f (x1 )
= f (xf2 +f
(x2
(y))
) = Cy .
If f is injective: Not obvious. If f is surjective onto Q+ : Let f (y) = z, z ∈ Q+ . Then
−1 −1
y = f −1 (z). Cy = 1 + f z(z) . So f (x + z) = (1 + f z(z) )f (x). This is f (x + z) = kz f (x).
This is a geometric progression property if x takes discrete steps of z. f (nz) = (kz )n f (0)
if 0 were in domain, or f (z0 + nz) = (kz )n f (z0 ).
Try P (x, y1 ) and P (x, y2 ): f (x + f (y1 )) = f (x)(1 + y1 /f (y1 )) f (x + f (y2 )) = f (x)(1 +
y2 /f (y2 )) Suppose 1 + y1 /f (y1 ) = 1 + y2 /f (y2 ) =⇒ y1 /f (y1 ) = y2 /f (y2 ). This
implies f (x + f (y1 )) = f (x + f (y2 )). If y/f (y) = k (constant), then f (y) = y/k.
Substitute into original: x+y/k
k
y
= xk (1 + y/k ) = xk (1 + k). x + y/k = x(1 + k) = x + xk.
y/k = xk =⇒ y = xk 2 . This must hold for all x, y. So k 2 = 1 =⇒ k = 1 (since
f (y) ∈ Q+ ). If k = 1, then f (y) = y. We already showed this is not a solution. This
problem is very hard. The solution is f (x) = x. Let’s re-verify this. LHS: f (x+y) = x+y.
RHS: f (x)(1 + y/f (y)) = x(1 + y/y) = x(1 + 1) = 2x. x + y = 2x =⇒ y = x. Fails.
Where could the problem statement be from? Or a common variant? If f (x + f (y)) =
f (x) + y: solutions f (x) = x, f (x) = −x. If f (x + f (y)) = yf (x): (This is different)
P (1, y) =⇒ f (1 + f (y)) = yf (1). If f (y) = c, then c = yc, if c ̸= 0, then y = 1. False.
y
So c = 0. The solution to the given problem f (x + f (y)) = f (x)(1 + f (y) ) is f (x) = x.
It was solved by setting y = x. Then f (x + f (x)) = f (x)(1 + x/f (x)) = f (x) + x.
If f (x) = x, then f (x + x) = x + x =⇒ f (2x) = 2x. This is true for f (x) = x.
This step alone isn’t a proof. (This requires a full Olympiad-style solution, often involving
proving injectivity/surjectivity first). The official solution to a similar problem (APMO)
f (x + f (y)) = f (x) + y is f (x) = x or f (x) = −x. The actual problem seems to be: ”Let

6
Part 10-II: Functional Equations - Adv. Techniques Advanced Functions by Yuki

f : Q+ → Q+ be a function such that f (x + f (y)) = f (x) + y for all x, y ∈ Q+ . Show

f (x) = x.” For f (x + f (y)) = f (x) + y: f is injective. f (f (y)) = y + f (0) (if 0 was in
domain). Not here. P (f (x), y) =⇒ f (f (f (x)) + f (y)) = f (f (x)) + y. f (x + f (y)) =
f (x) + y. f is surjective onto Q+ . There exists a s.t. f (a) = 1. P (x, a) =⇒ f (x + 1) =
f (x) + a. This is f (x) = ax + c form for rationals. f (x) = ax + f (0) if 0 allowed. Since
f (1) = 1 (often shown for such problems), then f (x + 1) = f (x) + 1. So f (q) = q for
q ∈ Q+ .
y
If the problem is indeed f (x + f (y)) = f (x)(1 + f (y) ), it is likely f (x) = y0 (constant),
or f (x) = x. f (x) = c =⇒ c = c(1 + y/c) = c + y =⇒ y = 0, not possible. So no
constant solution.

5. Problem 10II.5: f ′ (x) = f (x + y) − f (y) − xf (y) − x2 /2. f (0) = 1. The RHS

must be independent of y. Differentiate w.r.t y: 0 = f ′ (x + y) − f ′ (y) − xf ′ (y). Let
y = 0: 0 = f ′ (x) − f ′ (0) − xf ′ (0). So f ′ (x) = (x + 1)f ′ (0). Let f ′ (0) = A. f ′ (x) =
2
A(x + 1). Integrate: f (x) = A( x2 + x) + B. Since f (0) = 1, B = 1. So f (x) =
A( x2 + x) [+ 1.( And f ′ (x) = Ax ′
2

2
) + A.] Substitute
[ ( into
2
) the] original
[ ( equation ) for]f (x):
2 2
Ax + A = A (x+y) + (x + y) + 1 − A y2 + y + 1 − x A y2 + y + 1 − x2 .
( 2 2
2
) ( 2 ) ( 2 ) 2
Ax + A = A x +2xy+y
2
+ x + y − A y
2
+ y − Ax y
2
+ y − x − x2 . Ax + A =
2 2 2 2 2
A x2 + Axy + A y2 + Ax + Ay − A y2 − Ay − A xy2 − Axy − x − x2 . Ax + A =
2 2 2
A x2 + Ax − A xy2 − x − x2 . This must hold for all x, y. Coefficient of xy 2 : −A/2 =
0 =⇒ A = 0. If A = 0: 0 = −x − x2 /2. This is not true for all x. There must be
an error in the problem statement or my interpretation. Re-differentiating f ′ (x + y) =
f ′ (x) + 2y w.r.t x: f ′′ (x + y) = f ′′ (x). Re-differentiating f ′ (x + y) = f (x)f ′ (y) w.r.t x:
f ′′ (x + y) = f ′ (x)f ′ (y). If f ′ (x) = (x + 1)f ′ (0), then f ′ (x + y) = (x + y + 1)f ′ (0). From
0 = f ′ (x + y) − f ′ (y) − xf ′ (y): 0 = (x + y + 1)f ′ (0) − (y + 1)f ′ (0) − x(y + 1)f ′ (0).
0 = f ′ (0)[x + y + 1 − (y + 1) − x(y + 1)]. If f ′ (0) ̸= 0: x + y + 1 − y − 1 − xy − x =
0 =⇒ −xy = 0. This must hold for all x, y, which is false. So f ′ (0) = 0. This
means A = 0. If A = 0, then f ′ (x) = 0 for all x. So f (x) = C. Since f (0) = 1,
f (x) = 1. Substitute f (x) = 1 into the original equation: LHS f ′ (x) = 0. RHS
1−1−x(1)−x2 /2 = −x−x2 /2. So 0 = −x−x2 /2 for all x. This is false. This indicates
there might be no such differentiable function or an issue with the problem statement as
given for a general solution. The structure is unusual. If f (x+y)−f (y)−xf (y)−x2 /2 must
be *independent of y* for it to equal f ′ (x). Let H(x, y) = f (x+y)−f (y)−xf (y)−x2 /2.
∂H
∂y
= 0. f ′ (x + y) − f ′ (y) − xf ′ (y) = 0. This was correctly derived. Let y = 0:
f ′ (x) − f ′ (0) − xf ′ (0) = 0 =⇒ f ′ (x) = (x + 1)f ′ (0). Let f ′ (0) = C. f ′ (x) = C(x + 1).
f (x) = C(x2 /2 + x) + D. f (0) = 1 =⇒ D = 1. f (x) = C(x2 /2 + x) + 1.
′
Substitute this ( f (x) 2and f (x) back ) into (f ′ (x) = f (x +)y) − (f (y) − xf (y) − ) x2 /2.
2 2 2
C(x + 1) = C (x+y) + C(x + y) + 1 − C y2 + Cy + 1 − x C y2 + Cy + 1 − x2 .
( 2 2
) ( 2 ) 2 2
C(x + 1) = C x +2xy 2
+ x − x C y
2
+ Cy + 1 − x2 . Cx + C = C x2 + Cxy + Cx −
2 2 2 2 2
C xy2 − Cxy − x − x2 . C = C x2 − C xy2 − x − x2 . This must hold for all x, y. Equating
coefficients of xy 2 : −C/2 = 0 =⇒ C = 0. If C = 0: 0 = −x − x2 /2. Not true for all
x. Conclusion: No such differentiable function exists. (It is possible I am misinterpreting
a standard FE type, or the problem implies a very specific class of solutions).