Let P∈R[x] satisfy the given condition.
First, we notice that if b=0 then xP(x−a)=xP(x)⇒P(x−a)=P(x) (for all x=0), hence P(x)≡ a constant (since a=0), and we can recheck that any constant polynomial satisfies the given condition.
Now, let b=0, and put ab=k. We need only consider 2 cases.
1. k∈N : Substituting x=0 into the given condition, we have
−bP(0)=0⇒P(0)=0.
Substituting x=a,2a,3a,…,ka (in succession) into the given condition, we also have
aP(0)=(a−b)P(a)⇒P(a)=02aP(a)=(2a−b)P(2a)⇒P(2a)=03aP(2a)=(3a−b)P(3a)⇒P(3a)=0…(k−1)aP((k−2)a)=((k−1)a−b)P((k−1)a)⇒P((k−1)a)=0.
Thus, 0,a,2a,…,(k−1)a are roots of P(x)=0; therefore, by Bezout's theorem, P(x) can be written in the form
P(x)=x(x−a)(x−2a)⋯(x−(k−1)a)Q(x)
for some Q∈R[x].
The given condition is then equivalent to:
x(x−a)⋯(x−ka)Q(x−a)=x(x−a)⋯(x−ka)Q(x)⇔Q(x−a)=Q(x)∀x∈/{0,a,2a,…,ka}⇔Q(x)≡c⇔P(x)=cx(x−a)(x−2a)⋯(x−(k−1)a)∀x∈R
for any real constant c.
2. k∈/N : In the same way, it is easy to check that all numbers ma, with m∈N, are roots of P(x)=0; which implies that P(x)≡0; and we can re-check that the zero polynomial satisfies the given condition.
In summary:
- If b=0, then P(x)≡c.
- If ab=k∈N, then P(x)≡cx(x−a)(x−2a)⋯(x−(k−1)a).
- If ab=k∈/N, then P(x)≡0.