1. Let P(x,y) denote the given assertion:
P(x,y):f(x+y)≤yf(x)+f(f(x))
for all real numbers x and y.
2. First, consider P(x,0):
P(x,0):f(x+0)≤0⋅f(x)+f(f(x))⟹f(x)≤f(f(x))
This implies:
f(f(x))≥f(x)
Call this result (1).
3. Next, consider P(x,f(x)−x):
P(x,f(x)−x):f(x+(f(x)−x))≤(f(x)−x)f(x)+f(f(x))⟹f(f(x))≤(f(x)−x)f(x)+f(f(x))
Simplifying, we get:
0≤(f(x)−x)f(x)
This implies:
f(x)(f(x)−x)≥0
Call this result (2).
4. Claim: f(x)≤0 for all x∈R.
Proof:
- For an arbitrary real z, consider P(x,f(z)−x):
P(x,f(z)−x):f(x+(f(z)−x))≤(f(z)−x)f(x)+f(f(x))⟹f(f(z))≤(f(z)−x)f(x)+f(f(x))
Interchanging variables x and z and adding the inequalities, we get:
f(f(z))+f(f(x))≤(f(z)−x)f(x)+f(f(x))+(f(x)−z)f(z)+f(f(z))
Simplifying, we get:
2f(f(z))≤f(z)f(x)−xf(x)+f(x)f(z)−zf(z)
2f(f(z))≤2f(z)f(x)−xf(x)−zf(z)
2f(z)f(x)≥xf(x)+zf(z)
Call this result (3).
- Suppose there exists a real number t0≥0 for which f(t0)>0. Consider P(t0,y):
P(t0,y):f(y+t0)≤yf(t0)+f(f(t0))
Letting y→−∞, we get:
y→−∞limf(y+t0)=−∞
This implies f(x)→−∞ as x→−∞.
- Putting z=t0 and letting x→−∞ in (3), we get:
0>2f(x)f(t0)≥xf(x)+t0f(t0)≥0
This is a contradiction.
- Now assume t1,t2<0 are reals for which f(ti)>0 for i∈{1,2}. Without loss of generality, let t1<t2. Consider P(t2,t1−t2):
P(t2,t1−t2):f(t1)≤(t1−t2)f(t2)+f(f(t2))
Since t1−t2<0 and f(t2)>0, we get:
f(t1)≤0+0=0
This is a contradiction.
- Finally, if f(x)>0 holds for only one point x=t3, then f(t3)>0>t3. From (1) and the previous paragraph, we get:
0≥f(f(t3))≥f(t3)>0
This is a contradiction.
Hence, the claim must hold: f(x)≤0 for all x∈R.
5. Using (2), we get x→f(t) implies:
f(f(t))(f(f(t))−f(t))≥0
Since f(f(t))≥f(t) by (1), we see that f(f(t))=0⟹f(f(t))=f(t).
6. Define:
S:={x∈R:f(f(x))=0}
For any x,z∈S, replacing (x,z)→(f(x),f(z)) in (3), we get:
2f(x)f(t)=2f(f(x))f(f(t))≥f(x)f(f(x))+f(z)f(f(z))=f(x)2+f(z)2
This implies:
(f(x)−f(z))2≤0
So:
f(x)=f(z)
7. We are left with showing rather straightforward details to complete our proof. Note that f≡0 since otherwise, there is nothing to see.
Case 1: R=S.
- Note that there exists z0 such that f(f(z0))=0. Put z=f(z0) in (3) to get:
0=2f(f(z0))f(x)≥f(z0)f(f(z0))+xf(x)=xf(x)
So for all x<0, we must have f(x)=0. To see f(0)=0, simply observe that:
0=f(f(0))≤f(0)≤0
Case 2: R=S.
- Note that f(f(x)) is the constant function now. Therefore, set y=f(t)−x for any t∈R to get:
0≤f(x)(f(t)−x)
Thus, for all x<f(t)≤0, we have f(x)=0. Now, take y>0 and x<f(t)−y to get:
0=f(x+y)≤yf(x)+f(f(x))=f(0)≤0
Which gives f(0)=0. This contradicts the hypothesis of this case, since f(f(0))=0.
Hence, the conclusion must hold. ■