# Asymptotic Distribution of Primes

Given an integer $n$ and let $1\leq m\leq n$ be such that $n$ and $m$ are coprimes define $$\mathcal{N_{n,m}}:=\text{the set of primes p such that p\equiv{m}\hspace{0.1cm}\mathrm{mod}(n)}.$$ Let $\mathcal{P}$ be the set of all primes. I seem to recall that the following result is true: $$\varphi(n)^{-1}=\lim_{k\to\infty}{\frac{|\mathcal{N_{n,m}}\cap\{1,2,\ldots,k\}|}{|\mathcal{P}\cap\{1,2,\ldots,k\}|}}.$$ where $\varphi$ is the Euler's function.

My question is two fold:

• Does anyone have a reference for the previous fact? I was unsuccesful finding it.
• Are there finer results along these lines? Second order results?

Thanks!

-
This is due to Dirichlet. See en.wikipedia.org/wiki/… – wood Apr 6 '11 at 12:59
@wood: Clearly not, it implies the PNT. – Charles Apr 6 '11 at 14:29
@Charles, in principle one could prove that the primes in one residue class are $1/\phi(n)$ of all the primes without knowing the asymptotics of either counting function, thus, without PNT, no? I agree that Dirichlet did not do this. – Gerry Myerson Apr 7 '11 at 0:07

## 2 Answers

A good way to find the result you mentioned is to search for Dirichlet's (prime number) theorem; while Dirichlet only proved the infinitude of the set in question, nowadays one will frequently find the more precise assertion you mentioned when this result is discussed.

A more common way to state it is that the number of primes congruent to $m$ modulo $n$ smaller than $x$ is asymptotically equal to $\varphi(n)^{-1} x/log (x)$ (assuming coprimeness as you did), which in combination with the prime number theorem implies what you are looking for.

There are a variety of results related to finer aspects of this problem; key words e.g. Bombieri-Vinogradov Theorem or Siegel-Walfiz Theorem.

See for example the wikipedia article on Dirichlet's theorem here which also links to the keywords I mentioned for a quick overview.

Other than that as Gerry Myerson said any typical book on Analytic Number Theory will contain something on this subject (how much depends of course on the book).

-
The $x/\log x$ result can't be Dirichlet's, as it would imply the prime number theorem, which came about 60 years after Dirichlet's work. – Gerry Myerson Apr 6 '11 at 13:19
+1 for the pointer to Siegel-Walfisz. – Gerry Myerson Apr 6 '11 at 13:21
I don't have the reputation to comment but what about the second question? Are there more precise results? – Val Apr 6 '11 at 13:27
@Gerry Myerson, of course you are right I was imprecise (and will change this). And, sorry for the wrong spelling of your name. – user9072 Apr 6 '11 at 13:31
Yes, see 2nd paragraph of unknown (google)'s answer for keywords. – Gerry Myerson Apr 6 '11 at 13:31

It's just the prime number theorem for primes in arithmetic progression, no? Should be in any analytic number theory text that does the prime number theorem.

-