7

Let $q$ and $r$ be fixed coprime positive integers, $$ 1 \le r < q, \qquad \gcd(q,r)=1. $$ Suppose that two prime numbers $p$ and $p'$, with $p<p'$, satisfy $$ p \equiv p' \equiv r \ ({\rm mod}\ q), \tag{1} $$ and no other primes between $p$ and $p'$ satisfy $(1)$. Then we have the following

Naive generalization of Cramer's conjecture to primes in residue class $r$ mod $q$: $$ p'-p ~<~ \varphi(q)\,(\ln p')^2. \tag{2} $$

(PrimePuzzles Conjecture 77, A. Kourbatov, 2016). See arXiv:1610.03340, "On the distribution of maximal gaps between primes in residue classes" for further details, including the motivation for the $\varphi(q)$ constant. Here, as usual, $\varphi(q)$ denotes Euler's totient function.

Note: In the inequality $(2)$ we take the logarithm of the prime $p'$ at the larger end of the "gap". Very few counterexamples to $(2)$ are known; see Appendix 7.4 in arXiv:1610.03340. Definitely no counterexamples for $q=2, \ p<2^{64}$; also none for $1\le r < q \le 1000$, $ \ p<10^{10}$.

This conjecture (mostly in a less-naive "almost always" form) is mentioned in the following OEIS sequences listing maximal (record) gaps between primes of the form $p=qk+r$, $ \ \gcd(q,r)=1$: A084162, A268799, A268925, A268928, A268984, A269234, A269238, A269261, A269420, A269424, A269513, A269519.

Question 1: Find a counterexample to conjecture $(2)$.

Question 2: Find a counterexample to $(2)$, with prime $q$ and prime $r$.

Question 3: Find a counterexample to $(2)$, with $$ {p'-p \over \varphi(q)(\ln p')^2} > 1.1 \tag{3} $$ (A.Granville predicts that such counterexamples exist even for $q=2$, with the above ratio greater than $1.12$ -- more precisely, Granville expects that the ratio should exceed or come close to $2e^{-\gamma}$).

Question 4: Find a counterexample to $(2)$, with the additional condition $p'-p>q^2$.

Hint: Counterexamples are very rare. To find one, you will likely need to write a program and run it long enough. Good luck!

Alex
  • 4,873
  • I do expect a counterexample. I only think that such counterexamples are very rare. Thank you! – Alex May 08 '17 at 20:06
  • No you don't except someone will find a counter-example if you already made some programs. That's why you need to make it clear... – reuns May 08 '17 at 20:13
  • Is that $\log \log p' $ or $(\log p')^2 $? – daniel May 15 '17 at 15:47
  • That's $(\log p')^2$. – Alex May 15 '17 at 16:42
  • Did you try discussing the random models underlying those things ? (in particular what is Maier's heuristic under which $\log^2(p)$ should be replaced by $\log^{2+\epsilon}(p)$) – reuns Nov 14 '17 at 15:13
  • I do not think Maier ever suggested replacing $\log^2p$ by $\log^{2+\epsilon}p$. It was Adleman and McCurley who suggested $\log^{2+\epsilon}p$; whereas Maier and Pomerance (1990) just cautiously said that "we are a long way from proving these statements" - i.e. statements like $G(x)\sim\log^2x$ or $\limsup(G(x)/\log^2x)=1$. – Alex Nov 14 '17 at 19:10
  • Maier's theorem just tells us that, basically, average prime gaps near $x$, over relatively short intervals $[x,x+\log^\lambda x]$ infinitely often deviate up and down from the expected $\sim\log x$ and the deviation is so large that the limit of $$\frac{\pi(x+(\log x)^\lambda)-\pi(x)}{(\log x)^{\lambda-1}}$$ does not exist as $x\to\infty$. Based on this, Granville heuristically argues that maximal prime gaps near $x$ will also deviate from the expected $\sim\log^2 x$. – Alex Nov 14 '17 at 19:16

1 Answers1

2

Here are two counterexamples.

(A) Take $q=1605$, $r=341$, and consider the primes $p=3415781$ and $p'=3624431$.

It is not difficult to check that $$ p \equiv p' \equiv r \ ({\rm mod} \ q), \tag{1} $$ and between $p$ and $p'$ there are no other primes satisfying $(1)$. We have $\varphi(1605)=848$, and the exceptionally large gap is $$ 3624431 - 3415781 = 208650 > \varphi(q) \cdot (\log3624431)^2 = 193434.64\ldots $$ (This only answers question 1.)

(B) Take $q=18692$, $r=11567$, and consider the primes $p=190071823$ and $p'=193978451. \ $ We check that $(1)$ holds for $p$ and $p'$ -- and for no other primes between $p$ and $p'$. We have $\varphi(18692)=9344$; our exceptionally large gap is $$ 3906628 = 193978451 - 190071823 > \varphi(q) \cdot (\log193978451)^2 = 3402811.2255\ldots $$ (This answers questions 1 and 3.)

As of December 2019, questions 2 and 4 are still open.

Alex
  • 4,873