Answer: f(n)=n+1, for all positive integers n.
Let f be such a solution.
First step. The pre-image of 1 is empty.
Obviously, if there is an integer x so that f(x)=1. Set a=b=x into (1), then
f(2x)2=ff(x)(x)ff(x)(x)=f(x)2=1
This implies f(2x)=1. But, it's impossible since f is injective, as desired.
Second step. ff(n)−1(n)=2n for all n∈N. Moreover, f(1)=2
It's also clear. Set a=b=n into (1) and use the injectivity of f:
ff(n)(n)2=f(2n)2→ff(n)(n)=f(2n)→ff(n)−1(n)=2n(1)
In particular, there is an integer c such that f(c)=2. Then
2c=ff(c)−1(c)=f(c)=2.
So c=1, as desired.
Third step. The pre-image of 5 is non-empty.
According to the previous work, we get that the pre-images of even integers are non-empty. There must be an integer d such that f(d)=4. By using (1),
2d=ff(d)−1(d)=f3(d)=f2(4)(2)
Substitute a=1,b=4 into the problem statement:
f2(4)ff(4)(1)=f(5)2(3)
Combining the problem statement with (1), it follows f(5) is an even integer, say 2e. By (1) again,
f(5)=2e=ff(e)−1(e)
Clearly, e=1 since f is injective, so f(e)>2 and 5 is in the image of f.
Because the pre-image of 1 is empty, f2(g−1) and ff(g−1)(1) both are
3. Notice that
f2(g−1)=3=f(g)→f(g−1)=g
Furthermore, take a=2,b=g−2 in the problem statement
f5(g−2)ff(g−2)(2)=f(g)2=9(5)
By the result of the first step, we then get
fg(1)=ff(g−1)(1)=3=ff(g−2)(2)=ff(g−2)+1(1)
Therefore, f(g−2) equals to g−1. Also, (3) tells us
3=f5(g−2)=f2(3).
Since
ff(3)−1(3)=6
The only possibility is f(3)=6. This means e=29, which is absurd!
fh(1)=ff(h−1)(1)=5=ff(h−2)(2)=ff(h−2)(2)=ff(h−2)+1(1)
which implies f(h−2)=h−1. And we'll have
5=ff(2)(h−2)=ff(2)−3(5)
If f(2)=3, then the sequence 5,f(5),f2(5),… contains only finite distinct values. On the other hand, (1) tells us that
ff(5)−1(5)=10,ff(10)−1(10)=20,…,ff(5⋅2m)−1(5⋅2m)=5⋅2m+1,…
In particular, the sequence 5,f(5),f2(5),… contains 5⋅2m for all m∈N∪{0}. Thus, our assumption is false. That is, f(2)=3 and 4=ff(2)−1(2)=f(3)
Sixth step. f(n)=n+1 for all n∈N
We have shown that the statement f(n)=n+1 is true for n=1,2,3. Assume that it's true for all n<k. If k is odd, write k=2l+1 take a=l,b=l+2:
f(2l+2)2=ff(l)(l+2)ff(l+2)(l)=fl+1(l+2)fl+3(l)=fl+1(l+2)2⇒2l+2=fl+1(l+2)=f(2l+1)
For the other case: if k=2l, take a=l,b=l+1:
f(2l+1)2=ff(l)(l+1)ff(l+1)(l)=fl+1(l+1)2⇒2l+1=fl(l+1)=f(2l)
Thus, use the principle of induction, the conclusion follows.