Limit to Infinity of Summation of Euler Phi Function over Square

From ProofWiki
Jump to navigation Jump to search

Theorem

$\displaystyle \lim_{n \mathop \to \infty} \dfrac {\map \Phi n} {n^2} = \dfrac 3 {\pi^2}$

where:

$\map \Phi n = \displaystyle \sum_{k \mathop = 1}^n \map \phi k$
$\map \phi k$ is the Euler $\phi$ function of $k$.


Numerically, this evaluates to:

$\dfrac 3 {\pi^2} \approx 0 \cdotp 30396 35509 \ldots$

This sequence is A104141 in the On-Line Encyclopedia of Integer Sequences (N. J. A. Sloane (Ed.), 2008).


Proof

We find the probability that the two random numbers not less than $n$ is coprime.

We pick randomly $x, y \in \set {1, \dots, n}$.

There are $n^2$ ways to do this.

There are three cases:


If $x > y$, for each $x$, there are $\map \phi x$ values of $y$ so that $x \perp y$.

Since $1 \le x \le n$, there are a total of $\map \phi 1 + \dots + \map \phi n = \map \Phi n$ valid values of $y$.

If $x < y$, the result is similar: there are $\map \Phi n$ valid values of $x$.

If $x = y$, $x \perp y$ only if $x = y = 1$.


Therefore the probability that the two random numbers not less than $n$ is coprime is:

$\dfrac {\map \Phi n + \map \Phi n + 1} {n^2}$


Taking limit to infinity and using Probability of Two Random Integers having no Common Divisor:

\(\ds \frac 6 {\pi^2}\) \(=\) \(\ds \lim_{n \mathop \to \infty} \frac {\map \Phi n + \map \Phi n + 1} {n^2}\)
\(\ds \) \(=\) \(\ds \lim_{n \mathop \to \infty} \frac {2 \map \Phi n} {n^2} + \lim_{n \mathop \to \infty} \frac 1 {n^2}\)
\(\ds \) \(=\) \(\ds \lim_{n \mathop \to \infty} \frac {2 \map \Phi n} {n^2}\) Basic Null Sequences
\(\ds \leadsto \ \ \) \(\ds \lim_{n \mathop \to \infty} \frac {\map \Phi n} {n^2}\) \(=\) \(\ds \frac 3 {\pi^2}\)

$\blacksquare$



Historical Note

This result was cited by Leonard Eugene Dickson in his History of the Theory of Numbers, Volume I of $1919$.

He attributed it to J. Perrot about whom nothing can be found.


Sources