1

It is an open question as to whether there are infinite Mersenne primes. However, by the prime number theorem the probability that a large random number $z$ is prime is approximately 1/ln($z$). We can apply this to the Mersenne numbers ($z$ = 2^n-1), to figure out the expected number of Mersenne primes up to (2^n -1) (let's call this function E(n)).

E(n) = 1/ln(2^2-1) + 1/ln(2^3-1) + 1/ln(2^4-1) + ... + 1/ln(2^n-1)

E(n) > 1/ln(2^2) + 1/ln(2^3) + 1/ln(2^4) + ... + 1/ln(2^n)

E(n) > 1/(2ln2) + 1/(3ln2) + 1/(4ln2) + ... 1/(nln2)

E(n) > 1/ln(2) * (1/2 + 1/3 + 1/4 + ... + 1/n)

E(n) > ln(n)/ln(2)

Thus, as n approaches infinity, E(n) approaches infinity. But does this mean that the actual number of Mersenne primes is infinite?

The question: can this sort of probabilistic reasoning be used to prove the infinitude of Mersenne primes?

The only way there could not be infinite Mersenne primes is if the Mersenne numbers were systematically less likely to be prime than other numbers in their vicinity (to use a trivial example, the even numbers are less likely to be prime).

We can test how well this model matches reality using the list of 50 known Mersenne primes. This is the result.

enter image description here

The x-axis is on a logarithmic scale, such that the y-axis is the number of Mersenne primes up until 2^(2^x)-1. The orange line gives the approximation E(n) described above. As E(n) = approx ln(n)/ln(2); E(2^x) = x.

As you can see, the blue line (actual number of Mersenne primes) is consistently and increasingly above the orange line (expected number). I realised that this may be because Mersenne numbers cannot be even, and hence are twice as likely to be prime as a random number of that size. In other words, the formula for expected number of Mersenne primes should perhaps be E(2^x) = 2*ln(2^x)/ln(2) = 2x. This is represented by the yellow line, which is consistently above the actual number of Mersenne primes.

However, there is no particular reason to stop there. Compared to a random integer, Mersenne numbers are more likely to be multiples of 3. 2^(2n)-1 is a multiple of 3, meaning that a random Mersenne number has a 1/2 chance of being a multiple of 3 (and thus is less likely to be prime). Taking this into account, our formula becomes E(2^x)=1.5x. This seems to be a better approximation (green line).

There is no limit to how many prime factors we could take into account. For any prime except 2, a Mersenne number is more likely (compared to a random integer) to be a multiple of it, as 2^(p-1)-1 = pk. The more prime factors we take into account, the more we reduce our estimate of how many Mersenne primes there are.

Question: It seems that the actual number of Mersennes exceeds what this probabilistic model suggests. Why?

Thomas Delaney
  • 1,683
  • I'm not an expert on what you can and cannot prove about prime distributions with probability arguments (which is why this is only a comment), but (i) your proof strategy would require being Mersenne isn't one of those ruinous properties the way being even is, so you'd need a systematic theory of which properties "count", and (ii) the only reason we say $n$ has probability $1/(\ln n)$ of being prime is because $\pi(n)\sim\frac{n}{\ln n}$, a fact that's consistent with all sorts of fluctuations in $\frac{\pi(n)\ln n}{n}-1$ that probably make such a systematic theory very difficult to formulate. – J.G. Apr 29 '18 at 08:36
  • No, this heuristic is no proof. Especially because it does not consider the special structure of the Mersenne-numbers. A number of the form $2^p-1$ can only be prime if $p$ is prime and then, every factor of $2^p-1$ has the form $kp+1$. But even, if a heuristic did consider this as well, there would remain the possibility of "hidden forced factors" for sufficiently large primes $p$ which cannot be ruled out. – Peter Apr 29 '18 at 08:44
  • @Peter Ah, that's a good point; the requirement that $p$ be prime means the "right" sum to consider is $\sum_{n\ge 1}\frac{1}{\ln (2^{p_n}-1)}\sim\sum_n\frac{1}{n\ln n}$ (which is still divergent). – J.G. Apr 29 '18 at 08:46
  • Incredibly, it is even unknown whether $2^p-1$ is COMPOSITE for infinite many prime numbers $p$. Intuitive, this should be easy to prove, but apparently, it isn't. – Peter Apr 29 '18 at 08:47
  • @J.G. In fact, the probability for a number $2^p-1$ with prime exponent $p$ to be prime is much higher than the probability that a random number with this magnitude is prime. In fact, a reasonable guess is that infinite many Mersenne primes exist, but a proof seems to be out of reach. – Peter Apr 29 '18 at 08:50
  • Just to do feedback about your question, is that I think that using analysis (or probability) is impossible to deduce facts about the sequence of Mersenne primes. My intuition (I can not explain my ideas more) tells me that the only thing that can help us to understand the distribution of Mersenne primes is the light (I believe, but my ideas don't arise from a good scientific reasoning, that the only mechanism that is able to analyze this problem is a relativistic and theoretical tool, the so-called Einstein's clock or light clock). Insn't required a response of this comment, and good week. –  Apr 29 '18 at 09:11
  • @user243301 How on earth do you come to relativity theory in a question about Mersenne-primes ? Relativity theory to solve one of the most difficult questions in mathematics ? This would be stranger than Volker Dittmar's try to prove that no god exists with the relativity theory, which I consider to be absurd although I am an atheist. – Peter Apr 29 '18 at 13:00
  • Is a long story @Peter , the thoughts that I have created about this topic and others (as a model to understand the nature of certain problems) don't arise from the scientific method and now I cann't explain it (because it's a very crazy theory and I do not have any predictions). –  Apr 29 '18 at 19:16
  • 1
    Hi everyone, please check out the edits & expansion I've made. – Thomas Delaney May 04 '18 at 01:26

0 Answers0