5

For all prime $p \ge 7$, the decimal representation of $1/p$ repeats. Let $l_p$ be the length of the period or the non-repeating part of the decimal representation of $1/p$. It can be shown that $l_p \le p-1$. Thus $\frac{l_p}{p-1}$ is a measure of the length of $l_p$ relative to the magnitude of $p$.

Definition: We define the relative magnitude of the length of the period of a prime $p$ as $\frac{l_p}{p-1}$.

Question 1: What is the limiting value of

$$ c_n = \frac{1}{n}\sum_{k = 1}^n\frac{l_{p_n}}{p_n-1} $$

Question 2: Is it true that

$$ \lim_{n \to \infty}\frac{1}{n}\sum_{k = 1}^n\frac{l_{p_n}}{p_n} = \lim_{n \to \infty}\frac{\sum_{k = 1}^n l_{p_n}}{\sum_{k = 1}^n p_n} = \prod_{p} \Big (1 - \frac{p}{p^3-1}\Big ) \approx 0.57596 $$

The plot of $c_n$ vs. $n$ for $n \le 2460000$ and for this value of, $c_n \approx 0.5763$.

enter image description here

Source code

p = 7
s1 = s2 = s3 = k = 0
step = target = 10^4

while True: k = k + 1 d = divisors(p-1) l = len(d) i = 1 while i < l: e = d[i] if (10^e)%p == 1: s1 = s1 + 1/e.n() s2 = s2 + (e/(p-1)).n() s3 = s3 + e break i = i + 1

if k &gt;= target:
    print(k, s1, s2/k, s3)
    target = target + step
p = next_prime(p)

  • 2
    I do not think that this helps, but we even have $l_p\mid p-1$. One open problem is whether $l_p=p-1$ occurs infinite many times. – Peter Mar 29 '22 at 08:17
  • 1
    @NilotpalSinha What did you attempted in order to prove or disprove it? –  Mar 29 '22 at 16:24
  • @Peter I am running a separate program to check the proportion of primes for which $l_p = p-1$. The proportion looks to be stable at about $0.375$. Will post the results in a separate post after a bigger search range. – Nilotpal Sinha Mar 30 '22 at 04:32
  • 2
    Nilotpal Sinha. Period length $p-1$ happens if and only if $10$ is a primitive root modulo $p$. As $10$ is square-free, Artin's conjecture on primitive roots would imply this to happen with probability $0.373955\ldots$. However, the conjecture is still open. – Jyrki Lahtonen Mar 30 '22 at 04:48
  • @JyrkiLahtonen Thanks. This explains why the experimental proportion is close to the value predicted by Artin's conjecture – Nilotpal Sinha Mar 30 '22 at 04:53
  • 3
    https://arxiv.org/abs/1509.01752 is related. https://math.dartmouth.edu/~carlp/PDF/ordertalk.pdf is a bit closer. It has a $0.5723$ in it. – Gerry Myerson Mar 30 '22 at 06:09
  • Should the sum not start with $k=4$ ? – Peter Mar 30 '22 at 08:08
  • @GerryMyerson Thanks. Just an FYI, $0.5723$ agrees with fist four digits after the decimal point of the mean value of $\frac{a_2}{p-1}$ where $a_2$ is the smallest exponent such that $2^{a_2}-1$ is divisible by $p$. Not sure if the mean of $\frac{l_p}{p-1}$ will converge to the same value though. – Nilotpal Sinha Mar 30 '22 at 14:12
  • 1
    The length of the period of the decimal for $1/p$ is the smallest exponent $a_{10}$ such that $10^{a_{10}}-1$ is divisible by $p$. – Gerry Myerson Mar 31 '22 at 10:48
  • @GerryMyerson Yes, in fact I have used this property to compute $c_n$ for $n \le 2460000$. Added the source code – Nilotpal Sinha Mar 31 '22 at 11:37
  • The only question is whether the limit for base ten is the same as the limit for base two, and maybe if you follow the proof in the link you can work out whether you can modify it to work for base ten. In another matter, your definition of $c_n$ needs attention. You have a sum on $k$ of a summand that doesn't have $k$ in it, and you have $c_n$ defined as a limit as $n\to\infty$. – Gerry Myerson Mar 31 '22 at 12:00
  • Care to engage with my comment, Nilotpal? – Gerry Myerson Apr 02 '22 at 12:43
  • @GerryMyerson Hello. In your first link, the constant $0.5723$ appears in page 14 in the average order of $l(p)$. But my comment that the mean value of $\frac{a_2}{p-1}$ was obtained from experimental data. It could be a coincidence the first four digits agree or there could be a deeper connection. The proofs and other historical results mentioned in the second link are for a general base $a$ and the RHS of these results are independent of the base $a$. So I am not sure if these proofs can even be modified for base $10$. – Nilotpal Sinha Apr 02 '22 at 16:12
  • @GerryMyerson I don't think its a coincidence and also limit for base $2$ is not the same as the limit of base $10$. As defined in the link, let $$c = \Pi_p (1 - \frac{p}{p^3-1}) \approx 0.576$$ which agrees with the experimental constant for base $10$. If we compute $\frac{159c}{160}$ (a fraction which also appears in page 14 of the first link) we get approximately $0.5723$. This shows that when we drop from base $10$ to base $2$ the constant drops by a known fraction of $\frac{159}{160}$. It will be useful to see the derivation of this fraction in the original paper of Balazard and Kurlberg – Nilotpal Sinha Apr 02 '22 at 16:26

0 Answers0