Note that
r(a,b):=a+a1+b+b1=ab(a+b)(ab+1).
We put a=ut and b=wv, where t,u,v and w are positive integers such that both t and u and also v and w are coprime. Then we get r(a,b)=tuvw(tv+uw)(tw+uv), whence the Diophantine equation
tu(v2+w2)+vw(t2+u2)=kptuvw(6)
has to be investigated. Now gcd(tu,t2+u2)=1. Therefore, (6) implies tu∣vw. As we get similarly vw∣tu, too, we infer
tu=vw(7)
and (6) becomes
v2(v2+t2)(v2+u2)=t2+u2+v2+w2=kptu.
Therefore, p has to divide either v2+t2 or v2+u2. In the case p≡−1(mod4), i.e. when −1 is a quadratic non-residue mod p, this means that p divides v (and t or u). But since the same argument is valid for w instead of v, we have p∣v,w contradicting the coprimality of v and w. Thus the infinitely many primes with p≡−1(mod4) have no fantastic multiple and part (a) is solved.
For part (b) we choose v=1 and substitute w=tu. Thus we are looking for integers t and u such that
1+t2+u2+t2u2=kptu.
Here we choose t=F2l+1, u=F2l−1 and use the identity 1+F2l+12=F2l+3F2l−1 to obtain
(1+t2)(1+u2)=(1+F2l+12)(1+F2l−12)=F2l+3F2l−1F2l+1F2l−3=kpF2l+1F2l−1,
i.e. F2l+3F2l−3=kp. Therefore every prime factor of the Fibonacci number F2l+3 has a fantastic multiple.
In view of the well-known formula gcd(Fa,Fb)=Fgcd(a,b) it is clear that Fa and Fb are relatively prime, if a and b are different prime numbers. Hence we know that infinitely many prime numbers have a fantastic multiple, which solves part (b).