Denote P(x)=(x−a1)(x−a2)⋯(x−a100) with a1,a2,…,a100 are real roots of polynomial P(x). So
P(2x2−4x)=0⇔2x2−4x−ai=0 for 1≤i≤100.
This equation cannot have 1 root since Δ=16+8ai=0. So each equation can have 0 or 2 roots. Note that P(2x2−4x) has 130 roots so there are 2130=65 equations have 2 roots, which mean there are 65 numbers ai>−2 and 35 numbers ai<−2.
By the same way, we have
P(4x−2x2)=0⇔2x2−4x+ai=0
and Δ=16−8ai for any 1≤i≤100. And there are 65 numbers ai<2 and 35 numbers ai>2.
By applying the principle of inclusion and exclusion, there are
65+65−100=30 numbers ai∈(−2;2).
Suppose that a1<a2<…<a35<−2<a36<…<a64<2<a65<…<a100; and denote M,N,K as the subsets of these numbers with the indices 1→35,36→64,65→100.
Take A1(x)=∏m∈M(x−m), B(x)=∏n∈N(x−n), A2(x)=∏k∈K(x−k) then we will prove that
∣A1(x0)⋅A2(x0)∣>∣B(x0)∣
for any number x0∈(−1;1).
Note that ∀x0∈(−1;1) then ∣x0−m∣>1,∀m<−2 and ∣x0−k∣>1,∀k>2. We can suppose that x0>0 and for any n∈N, we have two cases:
1. If n>0 then n∈(0;2) and ∣x0−n∣<∣x0−2∣<∣x0−k∣ with any k>2.
2. If n<0 then n∈(−2;0) and ∣x0−n∣<∣x0−(−2)∣<∣x0−m∣ with any m<−2.
So in all cases of n respect to the factor x0−n in B(x0), we can choose factor from A1(x0) or A2(x0) with absolute value greater than it, and since ∣M∣,∣K∣>∣N∣, we always can do that. Hence ∣A1(x0)⋅A2(x0)∣>∣B(x0)∣,∀x0∈(−1;1) which mean this equation has no solution.
Therefore, we can choose A(x)=A1(x)A2(x) and B(x) to satisfy the given condition. □