Z[] AND THE FIBONACCI SEQUENCE MODULO n

SAMIN RIASAT

Abstract. It has long been known that the Fibonacci sequence modulo n is periodic for any integer n > 1. In this paper we present an elementary approach of proving properties of this period by working on Z[] and also deduce some new results. In the last section a method of proving identities is shown using examples.

1. Periodicity Modulo n We will use the following notations: n is a positive integer. Fi is the i-th Fibonacci number, with F0 = 0, F1 = 1 and Fi+1 = Fi + Fi1 for all i 1. Li is the i-th Lucas number, with L0 = 2, L1 = 1 and Li+1 = Li + Li1 for all i 1. = 1+2 5 , the golden ratio. And all congruences are taken modulo n unless otherwise stated. Denition 1.1. For n > 1, k(n) is the smallest positive index such that n | Fk(n) . We will often denote k(n) simply by k for brevity. Example: k(2) = 3, k(10) = 15 etc. Denition 1.2. For n > 1, (n) is the length of the period of the Fibonacci sequence modulo n. Example: (2) = 3, (10) = 60 etc. The following theorem was stated by Wall in [5]. Theorem 1.3. The Fibonacci sequence mod n is periodic. Proof. The terms of the Fibonacci sequence mod n can take only n possible values, namely 0, 1, . . . , n 1. Note that if a sub-sequence Fk , Fk+1 repeats at some point, the whole sequence will repeat from that point, since Fk+2 = Fk+1 + Fk and so on. There are at most n2 possible choices for the sub-sequence Fk , Fk+1 . So it must re-appear at some point and hence the sequence will become periodic. Corollary 1.4. Every positive integer divides innitely many Fibonacci numbers. Proof. Since the Fibonacci sequence mod n is periodic for every positive integer n, there exists innitely many positive integers k such that Fk+1 Fk+2 1. Then Fk 1 1 = 0 for all such k and the conclusion follows immediately.
1

SAMIN RIASAT

Proposition 1.5. (n) = k ordn (Fk+1 ). Proof. Since Fk 0, we have Fk+1 Fk1 (0 n 1) and from the Fibonacci recurrence it follows that Fk+i Fi for all i. Let g = ordn (). Then by the recursion we obtain Fgk+1 Fgk+2 g 1. Also, since g is the order, there does not exist g < g such that Fg k+1 Fg k+2 g 1. Hence we conclude that (n) = gk = k ordn (). Corollary 1.6. n | Fm k(n) | m. Proposition 1.7. (n) {k, 2k, 4k} n > 1. Proof. We will work on Z[]. Since Fk 0, we have k (1 )k k+1 (1 )k k+1 (1 )k+1 (1 )k (2 1) k+1 (1 )k+1 5k Fk+1 k 1
k

(mod n).

On the other hand, using 1 = 1/ we get k 2k (1)k 4k 1. (1.1)

If k 1 then Fk+1 1 and (n) = k. Otherwise, (i) if k is even then F2k+1 2k 1 which implies (n) = 2k. (ii) if k is odd then F4k+1 4k 1 implying (n) = 4k. Hence the conclusion. Remark: It is not dicult to see that k = Fk + Fk1 holds for all k. Thus Fk+1 Fk1 k follows from here as well. From Fk+1 k (mod n) we can propose a new denition for (n): Denition 1.8. (n) = ordn () n > 1.

Now we present a very short proof of another theorem of Wall in [5]. Theorem 1.9. (n) is even for n > 2. Proof. Assume that (n) {2k, 4k}. Then (n) = k implies k 1. Hence 2k 1 and by (1.1), (1)k 1. Therefore k is even (since n > 2) and the conclusion follows. Proposition 1.10. If n > 2 and k(n) is odd, then (n) = 4k(n). Proof. From the last theorem it follows that (n) = k. Suppose that (n) = 2k. Then (1.1) implies, 1 2k (1)k 1, a contradiction. Therefore (n) = 4k. Now we shall prove the central theorems of this section. From here onwards p will represent a prime. Theorem 1.11. If p > 3, n > 1 and n | Fp , then k(n) = p and (n) = 4p.

Z[] AND THE FIBONACCI SEQUENCE MODULO n

Proof. It is well known that gcd(Fi , Fj ) = Fgcd(i,j) . Hence i 0 (mod p) we have gcd(Fp , Fi ) = 1, which implies gcd(n, Fi ) = 1. Thus we conclude that p is the smallest positive integer such that n | Fp , i.e. k(n) = p. Since 3 p, Fp is odd, implying n is odd, and so n > 2. Since k(n) is odd, by Proposition 1.10 we conclude that (n) = 4p. Theorem 1.12. If n is prime and p > 3, then (n) = 4p n | Fp . Proof. Using Theorem 1.11, we need only prove that (n) = 4p n | Fp . From Proposition 1.7 we have 4p {k(n), 2k(n), 4k(n)}. Hence k(n) {p, 2p, 4p}. If k(n) = p we are done. So assume that k(n) = 2p. Since n | F2p = Fp Lp and n Fp we must have n | Lp = p + (1/)p . Therefore 2p 1, which implies (n) = 2p, contradiction. Now suppose that k(n) = 4p. This implies n | F4p = F2p L2p and since n F2p , n | L2p = 2p + (1/)2p . But then 4p 1, contradiction. Therefore k(n) = p. Theorem 1.13. If q is a prime and p > 3, then (q n ) = 4p q n | Fp . Proof. q = 2; otherwise 3 = (2) | (2n ) = 4p, which contradicts p > 3. Thus q is odd. Now Li = Fi+1 + Fi1 implies gcd(Fi , Li ) {1, 2} i. The rest of the proof is similar to that of Theorem 1.12. Theorem 1.14. If p > 3 and (n) = 4p, then n has a prime factor q with multiplicity r 1 such that q r | Fp . Proof. Let n = pa1 pj j be the prime factorization of n. Then 1 (n) = lcm( (pa1 ), . . . , (pj j )) = 4p. 1
a a

(1.2)

Therefore (pai ) {2, 4, p, 2p, 4p} i. But there is no x such that (x) {2, 4, p}. i Hence (pai ) {2p, 4p} i. If (pai ) = 4p for some i, we are done by Theorem i i 1.13. Otherwise, (pai ) = 2p i, which implies from (1.2) (n) = 2p, a contradiction. i Hence the result. As a consequence of these results we arrive at the following conclusion. Proposition 1.15. If q is an odd prime and r 2, then the following statements are equivalent, and they imply that q is a Wall-Sun-Sun prime. (i) q r | Fp . (ii) k(q r ) = k(q r1 ) = = k(q 2 ) = k(q) = p. (iii) (q r ) = (q r1 ) = = (q 2 ) = (q) = 4p. Proof. The above theorems imply that (i), (ii) and (iii) are equivalent. On the other hand, it is well known that q | Fq( q ) for all odd primes q, where a is the Legendre b
q symbol. Since p is the smallest index such that q | Fp , we must have p | q 5 i.e. Fp | Fq( q ) . Therefore q 2 | q r | Fp | Fq( q ) , implying q must be a Wall-Sun-Sun 5 5 prime, as desired.
5

It should, however, be noted that no prime p has yet been found such that q 2 | Fp , and the equivalence of statements (i), (ii) and (iii) may suggest a possible approach for investigating the existence of such primes.

SAMIN RIASAT

2. The Range of In 1913, R. D. Carmichael proved the following theorem: Theorem 2.1. Every Fibonacci number except F1 , F2 , F6 and F12 has a prime divisor which does not divide any smaller Fibonacci number. Such prime divisors are called characteristic divisors. Based on his result, we will attempt to nd X, the range of . Proposition 2.2. (2) = 3 is the only odd element of X. Proof. This follows directly from Theorem 1.9. Proposition 2.3. 8n + 4 X n. Proof. For n = 1 we have (8) = 12. Otherwise, let p be a characteristic divisor of F2n+1 . Then k(p) = 2n + 1 and from Proposition 1.10, (p) = 4(2n + 1) = 8n + 4, as desired. Proposition 2.4. 4n + 2 X n. Proof. For n = 1 we have (4) = 6. Otherwise, let p be a characteristic divisor of F4n+2 = F2n+1 L2n+1 . Then p | L2n+1 = 2n+1 + (1/)2n+1 which implies 4n+2 1 (mod p). Thus (p) = 4n + 2. Proposition 2.5. 8n X n. Proof. For n = 3 we have (6) = 24. Otherwise, let p be a characteristic divisor of F4n = F2n L2n . Then p | L2n = 2n + (1/)2n which implies 4n 1 (mod p). Thus 8n 1 (mod p) and we conclude that (p) = 8n. The above results may be summarized into the following theorem: Theorem 2.6. The elements of X are precisely 3 and all even numbers > 4.

3. Proving Identities In this section we will extensively use the following facts, which are very easy to prove, to prove some identities. If a, b, c, d are integers then a + b = c + d a = b, c = d. (a + b) + (c + d) = e + f for integers e, f such that e = a + c, f = b + d. (a+b)(c+d) = k+l for integers k, l such that k = ac+bd, l = ad+bc+bd. n = Fn + Fn1 . First Identity.
n

Fn = Fn+2 1.
i=1

(3.1)

Z[] AND THE FIBONACCI SEQUENCE MODULO n

Proof. Let Sn =

n i=1 Fn .

We have
n

n+1 1 1= 1 =

k
k=1 n

(Fk + Fk1 )
k=1

= Sn + Sn1 . On the other hand, n+1 1 = Fn+1 + Fn 1. Hence Fn+1 + Fn 1 = Sn + Sn1 + 1 1 Fn+1 + Fn = Sn (2 ) + (Sn1 + 1) Sn1 Fn+1 + Fn = (Sn1 + 1) + Sn Sn1 .

Therefore we conclude that Sn1 + 1 = Fn+1 , as desired. Second Identity. Fm+n1 = Fm Fn + Fm1 Fn1 , Proof. Since m+n = m n , we get Fm+n + Fm+n1 = (Fm + Fm1 )(Fn + Fn1 ) = (Fm Fn + Fm1 Fn + Fm Fn1 ) + Fm Fn + Fm1 Fn1 = (Fm Fn+1 + Fm1 Fn ) + Fm Fn + Fm1 Fn1 . Hence (3.2) follows. Third Identity.
n

or, Fm+n = Fm Fn+1 + Fm1 Fn .

(3.2)

Fkn+c =
i=0

n i ni Fk Fk1 Fc+i . i

(3.3)

Proof. From kn+c = (k )n c , we can write Fkn+c + Fkn+c1 = (Fk + Fk1 )n c

n

=
i=0 n

n ni i Fk i Fk1 i n i ni Fk Fk1 c+i i

=
i=0 n

=
i=0 n

n i ni Fk Fk1 (Fc+i + Fc+i1 ) i n i ni Fk Fk1 Fc+i + i

n i=0

=
i=0

n i ni Fk Fk1 Fc+i1 i

SAMIN RIASAT

Thus Fkn+c =

n i=0

n i ni Fk Fk1 Fc+i . i

It is clear that many other identities, if not all, can be proven in similar ways, and new identities may as well be deduced. Finally, the methods discussed here can easily be generalized to other Fibonacci-like sequences.

References
[1] [2] [3] [4] [5] Wikipedia, Fibonacci number, http://en.wikipedia.org/wiki/Fibonacci number Wikipedia, Pisano period, http://en.wikipedia.org/wiki/Pisano period Mathworld, Pisano period, http://mathworld.wolfram.com/PisanoPeriod.html Wikipedia, Carmichaels theorem, http://en.wikipedia.org/wiki/Carmichaels theorem D. D. Wall, Fibonacci Series Modulo m, American Mathematical Monthly 67 (1960), 525532.

Samin Riasat, Dhaka, Bangladesh E-mail address: nayel71@gmail.com