Olympiad Maths Prep

Track / Stage 9 / 59 of 80 #1939 of 2000

Problem 1939

IMO P2/P5; hard shortlist
Algebra Difficulty 9.2 Prove it IMO HK TST · Hong Kong

Let SS be a proper subset of R\mathbb{R} (i.e. SRS \neq \mathbb{R}) having at least two elements. Suppose there exists a function f:RRf: \mathbb{R} \to \mathbb{R} satisfying the following conditions:
(i) f(a+x+y)+f(f(a))+f(x)+f(y)=x+yf(a + x + y) + f(f(a)) + f(x) + f(y) = x + y; and
(ii) f(axy)+f(a)+f(x)f(y)=xyf(axy) + f(a) + f(x)f(y) = xy
for any real numbers aSa \notin S and x,ySx, y \in S. Find all such function(s) ff.

This one wants a proof. Work it on paper, read the official solution, then mark yourself honestly — the ladder only means something if the record is true.

Official solution

The only solution is f(a)=0f(a) = 0 for aSa \notin S and f(x)=xf(x) = x for xSx \in S.
Label the equations as follows.
f(a+x+y)+f(f(a))+f(x)+f(y)=x+y(1) f(a + x + y) + f(f(a)) + f(x) + f(y) = x + y \quad (1)
f(axy)+f(a)+f(x)f(y)=xy(2) f(axy) + f(a) + f(x)f(y) = xy \tag{2}

Case 1. 0S0 \notin S
Putting a=0a = 0 in (2), we obtain
f(x)f(y)=xy2f(0)(3) f(x)f(y) = xy - 2f(0) \tag{3}
for all x,ySx, y \in S. In particular, by putting x=yx = y, we get
f(x)2=x22f(0)(4) f(x)^2 = x^2 - 2f(0) \tag{4}
for all xSx \in S. Then we have
(x22f(0))(y22f(0))=f(x)2f(y)2=(xy2f(0))2, (x^2 - 2f(0))(y^2 - 2f(0)) = f(x)^2 f(y)^2 = (xy - 2f(0))^2,
which implies 2f(0)(x2+y2)=4f(0)xy2f(0)(x^2 + y^2) = 4f(0)xy. This means
2f(0)(xy)2=0. 2f(0)(x - y)^2 = 0.
Since SS has at least two elements, we can choose distinct x,ySx, y \in S to conclude
that f(0)=0f(0) = 0. Therefore, by (4),
f(x)2=x2f(x)^2 = x^2
for all xSx \in S. Since f(x)f(y)=xyf(x)f(y) = xy by (3), we see that the choice of the sign for f(x)f(x) is independent of xx. This means f(x)=xf(x) = x for all xSx \in S or f(x)=xf(x) = -x for all xSx \in S.
Now, (2) is reduced to
f(axy)+f(a)=0.(5) f(axy) + f(a) = 0. \tag{5}

For fixed nonzero aSa \notin S, if 1aS\frac{1}{a} \in S, we may put y=1ay = \frac{1}{a} in (5) to get f(x)+f(a)=0f(x) + f(a) = 0. But this cannot be true as we can choose two different values for xx (and hence f(x)f(x)). Therefore, we must have 1aS\frac{1}{a} \notin S. From this, we see that xSx \in S implies 1xS\frac{1}{x} \in S, since otherwise x=(x1)1Sx = (x^{-1})^{-1} \notin S.
Now, we can put y=1xy = \frac{1}{x} in (5) to obtain 2f(a)=02f(a) = 0, i.e. f(a)=0f(a) = 0 for any aSa \notin S. Equation (1) becomes
f(a+x+y)±(x+y)=x+y. f(a + x + y) \pm (x + y) = x + y.
By putting x=yx = y, we get
f(a+2x)±2x=2x. f(a + 2x) \pm 2x = 2x.
If the negative sign is chosen, then we have f(a+2x)=4xf(a + 2x) = 4x. Since f(a+2x)f(a + 2x) can only be (a+2x)-(a+2x) or 00 (depending on whether a+2xSa+2x \in S or not), we need a=6xa = -6x for any aSa \notin S and xSx \in S (note we cannot have 0=4x0 = 4x). This is impossible as we can find two distinct elements in SS. Therefore, f(a)=0f(a) = 0 for aSa \notin S and f(x)=xf(x) = x for xSx \in S.

Case 2. 0S0 \in S
Putting y=0y = 0 in (2), we obtain
f(0)+f(a)+f(x)f(0)=0(6) f(0) + f(a) + f(x)f(0) = 0 \qquad (6)
for all aSa \notin S, xSx \in S. In particular, by putting x=0x = 0, we get
f(a)=f(0)f(0)2.(7) f(a) = -f(0) - f(0)^2. \qquad (7)
Then equation (6) becomes
f(x)f(0)=f(0)2. f(x)f(0) = f(0)^2.
If f(0)0f(0) \neq 0, we need f(x)=f(0)f(x) = f(0) for all xSx \in S. Consider equation (1). It now becomes
f(a+x+y)+f(f(0)f(0)2)+2f(0)=x+y. f(a + x + y) + f(-f(0) - f(0)^2) + 2f(0) = x + y.
Note that the left-hand side can take at most two values (as f(a+x+y)f(a+x+y) can be f(0)f(0)2-f(0) - f(0)^2 or f(0)f(0)). However, as there are at least two elements in SS, say 0 and z0z \neq 0, the right-hand side can take at least three different values, namely, 0, zz, 2z2z. This is a contradiction, and hence f(0)=0f(0) = 0. By (7), we obtain f(a)=0f(a) = 0 for aSa \notin S.
Next, we prove that aSa \notin S implies aS-a \notin S. Indeed, suppose aS-a \in S. We put y=ay = -a in (1) to get
2f(x)+f(a)=xa.(8) 2f(x) + f(-a) = x - a. \qquad (8)
f(a)=a,(9) f(-a) = -a, \tag{9}
and hence 2f(x)=x2f(x) = x for any xSx \in S by (8). Now, by putting x=ax = -a, we obtain 2f(a)=a2f(-a) = -a. But then we have f(a)=af(-a) = -a by (9). This forces a=0Sa = 0 \in S, which is a contradiction. So the claim is true. From this, we see that xSx \in S implies xS-x \in S. Thus, we can put y=xy = -x in (1). This gives
f(x)+f(x)=0(10) f(x) + f(-x) = 0 \tag{10}
for any xSx \in S.
Similarly, we shall prove that z,wSz, w \in S implies z+wSz+w \in S. Suppose on the contrary that z+wSz+w \notin S. Note that wS-w \in S from above. So we can put a=z+wa = z+w and x=0x = 0, y=wy = -w in (1) to obtain
f(z)+f(w)=w. f(z) + f(-w) = -w.
Using (10), we get
f(z)f(w)=w. f(z) - f(w) = -w.
By symmetry, we also have
f(w)f(z)=z. f(w) - f(z) = -z.
Adding these, we obtain z+w=0Sz+w=0 \in S, which is a contradiction.
It is now clear that a+x+ySa+x+y \notin S for any aSa \notin S and x,ySx, y \in S. Otherwise, if a+x+ySa+x+y \in S, since xy=(x)+(y)S-x-y = (-x)+(-y) \in S, we have
a=(a+x+y)+(xy)S. a = (a + x + y) + (-x - y) \in S.
It follows that (1) is reduced to f(x)+f(y)=x+yf(x)+f(y)=x+y. Putting x=yx=y, we get f(x)=xf(x)=x for any xSx \in S.
In any case, f(a)=0f(a) = 0 for aSa \notin S and f(x)=xf(x) = x for xSx \in S is the only possible function. One can check that this is indeed a solution if and only if a+x+y,axyS{0}a+x+y, axy \notin S - \{0\} for any aSa \notin S and x,ySx, y \in S.

Source: MathNet, licensed CC-BY-4.0. Statement and solution reproduced as published; topic, difficulty and ordering added by this site.