# Seeking a proof of $\sum_{d|n}\phi(\frac{n}{d})a^d\equiv 0 \mod{n}$, where $\phi$ is the Euler Totient Function.

I need to prove the proposition.

Let $a$ be an arbitrary integer. Then for every positive integer $n$, we have $$\sum_{d \mid n}\phi\left(\frac{n}{d}\right)a^d\equiv0\pmod{n}.$$

-
 What have you tried? – Qiaochu Yuan Oct 20 '12 at 1:49 I used the Burnside's Lemma (Cauchy-Frobenius), this is Group and rings Theory applied, but I want to know if I can use just number theory, anyone tolds me that I can find a solution to this problem using "Pólya counting" But I don't understand it – Elmo goya Oct 20 '12 at 1:57 You shouldn't think of mathematical subjects as being divided like that. Burnside's lemma is Burnside's lemma. You could call it group theory. It would be equally valid to call it combinatorics. This particular application could safely be called number theory. The labels don't matter. Polya enumeration is a special case of Burnside's lemma. – Qiaochu Yuan Oct 20 '12 at 2:00 I posted just now, but didn't see the two comments that appeared very recently. I hope my post is useful in spite of the overlap with these comments. – Marko Riedel Oct 20 '12 at 2:18 Check it directly when $n$ is a prime power. For other $n$, pick a prime factor $p$ of $n$ and write $n$ as a product of a $p$-power and another number relatively prime to $p$. Divisors of $n$ break up in a similar way to this, and $\varphi$ is multiplicative, so you should be able to derive the result by induction on the number of different prime factors of $n$. – KCd Jan 25 at 6:58

The cycle index of the cyclic group on $n$ elements is $$Z(C_n) = \frac{1}{n} \sum_{d|n} \phi(n/d) x_{n/d}^d$$ where the $x_d$ are the variables.
Hence, by the Polya Enumeration Theorem, the quantity $$Z(C_n)_{x_1 = a, x_2 = a, x_3 = a, \ldots}$$ counts the number of distinct necklaces on $n$ elements under rotation where the slots on the necklace hold one of $a$ colors.
Hence, combinatorially, the following quantity must be an integer $$\frac{1}{n} \sum_{d|n} \phi(n/d) a^d$$ because it counts the number of necklaces.
 The OP has indicated that he already knows the solution using Burnside's lemma, and PET is a special case of this. – Qiaochu Yuan Oct 20 '12 at 2:19 Ty, but... The answer I need is without the $\frac{1}{n}$, I think that the answer comes near – Elmo goya Oct 20 '12 at 2:24