Are there an infinite number of primes of the form $\lfloor \pi n \rfloor$?

This is sort of a clickbait title. I would really like to show that, for any real irrational $r > 1$, there are an infinite number of primes of the form $\lfloor r n \rfloor$ for positive integer $n$.

I can show that there are an infinite number of primes of the form $\lfloor r n \rfloor$ for $r < 1+1/g$ for a fixed $g > 1$, but the best known value of $g$ is $246$ unconditionally and $6$ assuming an unproved conjecture.

To prove this, I use the idea of Beatty sequences (https://en.wikipedia.org/wiki/Beatty_sequence). For a real $r > 1$, let $B(r) =\{\lfloor nr \rfloor \mid n \in \mathbb{N}^+\}$. ($\mathbb{N}^+$ is the set of positive integers.) Then Beatty's theorem states that $B(r)$ and $B(r/(r-1))$ make up a disjoint partition of $\mathbb{N}^+$.

If there are only a finite number of primes in $B(r)$, then all the primes above a certain value are in $B(r/(r-1))$.

We have $\lfloor (n+1)r \rfloor =\lfloor nr+r \rfloor \ge \lfloor nr \rfloor +\lfloor r \rfloor $. Therefore, if $r > 3$, $B(r)$ can not contain any twin primes. Therefore, if there are an infinite number of twin primes, $B(r/(r-1))$ must contain an infinite number of primes for all $r > 3$, or $r/(r-1) \lt 3/2 $.

The use of an unproved conjecture can be removed at the cost of a weaker conclusion by using the recent results on prime gaps (https://en.wikipedia.org/wiki/Prime_gap). It has been shown that there is a constant $g$ such that there are an infinite number of consecutive primes that differ by at most $g$.

It has been shown that $g \le 246$ unconditionally and that, assuming the Elliott–Halberstam conjecture (https://en.wikipedia.org/wiki/Elliott%E2%80%93Halberstam_conjecture), $g \le 12$ ($g \le 6$ assuming a generalized form of the conjecture.).

Arguing as before, if $r > g+1$, then $B(r)$ can not contain all the primes above any finite value. For, if it did, there are an infinite numer of consecutive primes $p$ and $q$ such that $q-p \le g$, and $p$ and $q$ can not both be in $B(r)$.

Therefore $B(r/(r-1))$ must contain an infinite number of primes for $r > g+1$. Restating, $B(r)$ must contain an infinite number of primes for $r < (g+1)/g =1+1/g$.

I don't know how to go beyond this.


There is a conjecture on the least prime in an arithmetic progression that would imply the answer to your question is yes. For coprime positive integers $a$, $d$, write $p(a,d)$ for the smallest prime congruent to $a$ modulo $d$.

Conjecture 1: For every $\epsilon>0$, the bound $p(a,d) =O_\epsilon(d^{1+\epsilon})$ holds for all $a$, $d$.

The statement is known to hold if $1+\epsilon$ is replaced by $5$, and if we assume GRH it is know to hold with $1+\epsilon$ replaced by $2+\epsilon$. The conjecture and related results are discussed here.

Proof that Conjecture $1$ implies an affirmative answer to your question:. There are infinitely many pairs of coprime positive integers $p$, $q$ such that $$ 0<\frac{p}{q}-r< \frac{1}{q^2}. $$ For such $p$, $q$, the integers $$ \lfloor qr\rfloor,\lfloor 2qr\rfloor,\ldots,\lfloor q^2r\rfloor $$ form an arithmetic progression with initial term $p-1$ and common difference $p$. The largest term is about $p^2/r$, so once $p$ is sufficiently large, one of these terms must be prime.

Unconditionally, this argument shows that there are infinitely many primes of the form $\lfloor rn\rfloor$ if the irrationality measure $\mu(r)$ satisfies $\mu(r)> 5$, and on GRH it is sufficient that $\mu(r)>2$.

| cite | improve this answer | |

Current mathematics can prove your result unconditionally: If a prime $p$ satisfies $\{\frac{1}{\pi} p \} > 1-\frac{1}{\pi}$, then it is of the form $\lfloor n \pi \rfloor$ (in fact on can check that this condition on the fractional part is equivalent to $p$ being of the form $\lfloor \pi n \rfloor$). Indeed, if $\{\frac{1}{\pi} p \} > 1-\frac{1}{\pi}$ set $n = \lceil \frac{p}{\pi} \rceil$.

Then $\pi n \ge \pi \frac{p}{\pi} = p$

While $\pi n < \pi (\frac{p}{\pi}+\frac{1}{\pi}) = p+1$

Hence $\lfloor \pi n \rfloor = p$, so $p$ is of this form.

However it is known that for any irrational $\alpha, \{ \alpha p \}$ is dense in $[0,1]$ (in fact equidistributed) as $p$ varies over the primes. See for instance this MO question.

So not only we can prove that there are infinitely many primes of the form $\lfloor \pi n \rfloor$, but (as expected!) a proportion $\frac{1}{\pi}$ of primes are of this form, since a proportion $\frac{1}{\pi}$ of the fractional parts of $\{\pi p\}$ will lie in the interval $(1-\frac{1}{\pi},1)$

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.