I'm asked to find all integers $n$ such that $\phi(n)+\sigma(n)=2n.$ I know that when $n$ is a prime, $\phi(n)+\sigma(n)=(n-1)+(n+1)=2n.$ My guess is that $n$ can be primes only, and I want to derive a contradiction when $n$ is a composite number. But I can only show that when $n=pq$ ($p$, $q$ distinct primes), $\phi(n)+\sigma(n)=2pq+2=2n+2.$ How do I generalize from here?
Asked
Active
Viewed 2,357 times
7
-
For what it's worth, $\phi(n) = 2n - \sigma(n) > 0$, so that $n$ must be deficient. – Jose Arnaldo Bebita Dris Apr 01 '15 at 03:34
-
Note further that $n = 1$ is a solution, as $\phi(1) + \sigma(1) = 2\cdot1$. – Jose Arnaldo Bebita Dris Apr 01 '15 at 03:43
1 Answers
10
Start with the definitions of $\sigma$ and $\phi$:
$ \sigma(n) = \sum_{d \vert n}{d}$
$ \phi(n) = \prod_i{(p_i - 1) \cdot p_i^{e_i-1}} $
Expand the product and rewrite as:
$ \phi(n) = \sum_{d \vert n}{a_d \cdot d}$
with coefficients $a_d \in \{-1,0,1\}$. So now we have
$ \sigma(n) + \phi(n) = \sum_{d \vert n}{(a_d + 1) \cdot d} $
and since $a_n = 1$:
$ \sigma(n) + \phi(n) = 2 \cdot n + \sum_{d \vert n, d < n}{(a_d + 1) \cdot d}$
Now $\sigma(n) + \phi(n) = 2 \cdot n$ if and only if $a_d = -1$ for all proper divisors $d$. This is only true for prime $n$, and $n=1$. Otherwise some positive terms remain.
Dan Brumleve
- 18,041