Let's define the radical of the positive integer $n$ as $$\operatorname{rad}(n)=\prod_{\substack{p\mid n\\ p\text{ prime}}}p$$ and consider the following Fibonacci-like sequence $$a_{n+1}=\operatorname{rad}(a_{n})+\operatorname{rad}(a_{n-1})$$ If $a_1=1,\,a_2=1$ the sequence coincides with OEIS A121369 $$1, 1, 2, 3, 5, 8, 7, 9, 10, 13, 23, 36, 29, 35, 64, 37, 39, 76, 77, ...$$ If $a_1=2,\,a_2=2$ the sequence becomes $$2, 2, 4, 4, 4, ...$$ If $a_1=3,\,a_2=3$ the sequence becomes $$3, 3, 6, 9, 9, 6, 9, ...$$ If $a_1=5,\,a_2=5$ the sequence becomes $$5, 5, 10, 15, 25, 20, 15, 25, ...$$ If $a_1=7,\,a_2=7$ the sequence becomes $$7, 7, 14, 21, 35, 56, 49, 21, 28, 35, 49, 42, 49, 49, 14, 21, ...$$ The above sequences, except for the first, are all periodic. Continuing with the successive prime numbers, we obtain:
for $\,p=11,\,$ a sequence with a period length of $\,9$,
for $\,p=13,\,$ a sequence with a period length of $\,81$,
but for $\,p=17\,$ and $\,p=19\,$ two apparently divergent sequences.
Other primes that generate periodic sequences are (the respective period lengths in brackets):
$$23 (9), 29 (12), 31 (207), 37 (27), 41 (36), 47 (39), 73 (198), 79 (60)$$
Some questions arise from the previous experimental observations:
is the period length always a multiple of $3$ (not considering the case $p=2$)?
also in the doubtful cases mentioned above, does the sequence become periodic at some point?
given the starting prime number, is it possible to predict the length of the period of the generated sequence or, at least, to identify some pattern?
I have posted a more general question of the same nature here.
Edit
For the calculation of $\,\operatorname{rad}(n)\,$ I used the sympy.primefactors() method inside Python:
from sympy import primefactors
def rad(num):
primes = primefactors(num)
value = 1
for p in primes:
value *= p
return value
(a0, a1) = (17, 17)
for n in range(2, 10001):
a2 = rad(a1) + rad(a0)
print(n, a2)
a0 = a1
a1 = a2