Is there an $n$ such that $p|n^2+1$ with $2n<p<2n+\sqrt n$?

72 Views Asked by At

Is there an integer $n$ such that $n^2+1$ is divisible by a prime $p$ with $2n<p<2n+\sqrt n$?

It's complicated to describe my interest, but these are near-missed for arc-cotangent reducible numbers (A002312) which would provide a counterexample that certain types of numbers are equivalent to that sequence.

The chance that a random number is divisible by some prime in that range is roughly $$ \frac{1}{2\log n\sqrt n} $$ the sum of which diverges far too rapidly to explain the absence of examples for $n<10^7.$ (I recognize that these heuristics are too rough for the situation in at least two ways, but it helps show that the situation is interesting.)

1

There are 1 best solutions below

0
On BEST ANSWER

Your intuition is right to suspect an algebraic obstacle to the random behavior the heuristic predicts. Suppose that $k\ge1$ is such that $2n+k$ divides $n^2+1$. Writing $n^2+1=(2n+k)d$, we see that $d$ must be greater than $\frac n2-\frac k4$, since $(2n+k)(\frac n2-\frac k4) = n^2-\frac{k^2}4$ is too small. But then the smallest that $d$ can be, given that it is an integer, is $\frac n2-\frac k4+\frac14$. This implies that $$ n^2 + 1 = (2n+k)d \ge (2n+k)\big( \tfrac n2-\tfrac k4+\tfrac14 \big) = n^2 + \tfrac n2 + \tfrac k4 - \tfrac{k^2}4, $$ and so $(\frac k2-\frac14)^2 = \frac{k^2}4 - \frac k4 + \frac1{16} \ge \frac n2 - \frac{15}{16}$. We conclude that $$ (2n+k)\mid (n^2+1) \implies k\ge 2\sqrt{\tfrac n2 - \tfrac{15}{16}} + \tfrac14 \ge \sqrt{2n} $$ (the latter inequality holding for $n\ge30$). Note that this holds for all divisors, not just prime divisors.