Let P(x,y) be the assertion f(x2)+f(2y2)=(f(x+y)+f(y))(f(x−y)+f(y)).
P(0,x) gives us
f(0)+f(2x2)=2f(x)(f(x)+f(−x))(1)
and P(0,−x) gives us
f(0)+f(2x2)=2f(−x)(f(x)+f(−x)).(2)
By combining (1) and (2) we get
f(x)2=f(−x)2.(3)
P(0,0) gives us 2f(0)=4f(0)2, thus we have two cases:
Case 1. f(0)=21.
P(x,0) gives us
f(x2)=(f(x)+21)2−21,(4)
while P(−x,0), gives us
f(x2)=(f(−x)+21)2−21(5)
Combining (4) and (5) and using (3) we get
f(x)=f(−x)(6)
The assertion P(x2,x2) can be written as
f(x4)+f(2x4)=f(2x2)+f(x2)(21+f(x2))(7)
For an arbitrary x∈R, let us denote a=f(x). Using (4) we get:
f(x2)f(x4)=(a+21)2−21,=(f(x2)+21)2−21=(a+21)4−21.
Using (1) and (6) we get
f(2x2)f(2x4)=4f(x2)−21=4a2−21,=4f(x2)2−21=4((a+21)2−21)2−21.
Plugging the last 4 equations in (7) we get:
(a+21)2+4(a+21)2−21−1=(4a2−1+a+21)(a+21)2
which is equivalent to
(a+21)2(4a−2)=0.
Therefore a=±21 and f(x)=±21. Now if we use (6) in (1) we get
f(0)+f(2x2)=4(f(x))2=1
so f(2x2)=21 for every x, now using (6) we conclude f(x)=21 for all x which is easily checked to be a solution.
Case 2. f(0)=0.
We immediately see using P(x,0) that
f(x2)=f(x)2.(8)
By comparing P(x,y) and P(x,−y) and using (3) we get:
(f(y)−f(−y))(f(x+y)+f(x−y))=0
If there exists c∈R such that f(c)=f(−c) we have for all x
f(x+c)=−f(x−c)
Plugging in x+c in x here gives us:
f(x+2c)=−f(x).(9)
Specially, f(2c)=0. Now, P(2c−y,y):
f((2c−y)2)+f(2y2)(−f(−y))2+f(2y2)f(2y2)=(f(2c)+f(y))(f(2c−y)+f(y)),=f(y)f(2c−2y)+f(y)2=f(y)f(2c−2y)=−f(y)f(−2y)(10)
Let S(x) denote the statement (x=0)∧(f(x)=f(−x)=0). If there is no d∈R such that S(d) then f(x)=−f(−x) for all x∈R. P(0,x) gives us
f(2x2)2f(x)(f(x)+f(−x))=0,
which gives us another solution f(x)=0. Now, let us assume that there exists d∈R such that S(d) holds. Obviously, S(−d) holds, as well. P(0,d) gives us
f(2d)=4f(d)2
and (10) gives us
f(2d2)f(−2d)f(2d)=−f(d)f(−2d)=−4f(d)=−4f(−d)=−4f(d)=f(−2d).
Therefore, S(2d) also holds. Inductively, we deduce that S(2nd) holds for every n∈N. Also, f(2nd)=(−4)nf(d), which means that f is unbounded.
P(x,c), using the fact f(x2)=f(x)2:
f(x)2+f(2c2)=f(x+c)f(x−c)+f(c)(f(x+c)+f(x−c))+f(c)2,
and since f(x+c)=−f(x−c) and f(2c2)=0 (this follows from P(0,c)) we have
f(x)2+f(x+c)2=f(c)2
which implies that f is bounded and that is contradiction. Therefore, there is no c∈R such that f(c)=−f(c) and therefore
f(x)=f(−x), for all x∈R.(11)
P(0,x):
f(2x2)=4f(x)2=4f(x2).
Therefore, using (11):
f(2x)=4f(x), for all x∈R.(12)
P(x,y) can now be written as follows:
f(x)2+3f(y)2=f(y)(f(x+y)+f(x−y))+f(x+y)f(x−y)
and similarly, P(y,x) can be written as:
f(y)2+3f(x)2=f(x)(f(x+y)+f(x−y))+f(x+y)f(x−y).
Subtracting the previous two equalities:
(f(x)−f(y))(2f(x)+2f(y)−f(x+y)−f(x−y))=0.(13)
Assume that for some x,y∈R, f(x)=f(y)=a. Let f(x+y)=b and f(x−y)=c.
Now we have:
4a2=bc+ab+ac(14)
P(x+y,x−y):
f(x+y)2+4f(x−y)2=(f(2x)+f(x−y))(f(2y)+f(x−y))
i.e.:
b2+4c2=(4a+c)2.(15)
If we plug in x→x+y, y→x−y in (13) we get:
(f(x+y)−f(x−y))(2f(x+y)+2f(x−y)−f(2x)−f(2y))=0
i.e.:
(b−c)(2b+2c−8a)=0.
If b=c (15) gives us:
5b2=(4a+b)2
b2=4a2+2ab
while (14) gives us:
4a2=b2+2ab.
Thus, ab=0 and a=b=c=0 which implies 2a+2a−b−c=0. On the other hand, if b=c we also have 2a+2a−b−c=0.