← Back to maths posts

Self-Fibonacci Numbers

number theory
research
open
fibonacci
exploration
For which n does n divide the nth Fibonacci number?
Author

Apurva Nakade

Published

August 11, 2025

Here’s a seemingly simple question that lacks a clean solution:

For what positive integers \(n\) does \(n\) divide \(F_n\), where \(F_n\) denotes the \(n^{\text{th}}\) Fibonacci number?

According to the On-Line Encyclopedia of Integer Sequences (OEIS), such numbers are called “Self-Fibonacci numbers”: https://oeis.org/A023172. The sequence begins:

1, 5, 12, 24, 25, 36, 48, 60, 72, 96, 108, 120, 125, 144, 168, 180, 192, 216, 240, 288, 300, 324, 336, 360, 384, 432, 480, 504, 540, 552, 576, 600, 612, 625, 648, 660, 672, 684, 720, 768, 840, 864, 900, 960, 972, 1008, 1080, 1104, 1152, 1176, 1200, 1224, 1296, 1320

One approach to this question is as follows:

Let \(\alpha(n)\) be the smallest positive integer \(k\) such that \(n \mid F_k\). Note that if \(n \mid F_k\), then \(n \mid F_{mk}\) for all positive integers \(m\). Therefore, the self-Fibonacci numbers are exactly those for which \(\alpha(n) \mid n\).

This sequence is fascinating, with apparently no simple formula! I explored the following related question:

For what positive integers \(n\) does \(\alpha(n) = n\)?

Here’s the list of such numbers up to 10,000.

This is a much simpler question, and the sequence appears to consist of all numbers that are powers of 5 or 12 times a power of 5. The OEIS entry for this sequence is https://oeis.org/A289586.

ImportantConjecture

The only integers \(n\) such that \(\alpha(n) = n\) are of the form \(5^k\) or \(12 \cdot 5^k\) for some non-negative integer \(k\).

I’m not certain this is true, but it holds for all integers up to 10,000. I’m also interested in understanding the distribution of \(n / \alpha(n)\). When plotting \(n / \alpha(n)\), most values fall above 1, though some are less than or equal to 1.

Below is a table of \(n / \alpha(n)\) for all integers \(n\) up to 10,000. Surprisingly, the most common value is not 1 but \(4/3\).

I wonder what patterns exist in the ratios \(n / \alpha(n)\)…

NoteQuestions
  • What is the distribution of the ratios \(n / \alpha(n)\)?

  • For every integer \(k\), is there some integer \(n\) such that \(n / \alpha(n) = k\)?

  • Can we exploit the fact that

    \[ F_k = \dfrac{1}{\sqrt{5}} \left( \phi^k - \psi^k \right) \]

    where \(\phi = \frac{1 + \sqrt{5}}{2}\) and \(\psi = \frac{1 - \sqrt{5}}{2}\), to find a closed form for \(\alpha(n)\)? We know that \(\mathbb{Z}[\phi]\) is a Euclidean domain so we can have access to number theoretic tools such as divisibility and the Euclidean algorithm.