3
$\begingroup$

Let $n=\prod_{p|n}p^{\alpha_p}$. While considering arithmetic functions of the projection $$\textrm{rad}_k(n)=\prod_{p|n}p^{\textrm{min}(\alpha_p,k)},$$ a basic question has arisen that is more difficult to answer than I had supposed.

Let $a,f:\mathbb{N}\to\mathbb{C}$ and define $$Af(n)=\sum_{d|n}a(d)f(d)$$

so that $$A^2f(n)=\sum_{d|n}a(d)\sum_{e|d}a(e)f(e)$$ and so on. The question is simply this:

If $A^{k+1}=A$ for some $k\in\mathbb{N}$ and $a(1)=1$, does it follow that $a(n)$ is multiplicative?

Multiplicative solutions exist for every $k$; these are the functions $$a(n)=\delta_k(n)z^{\Omega(n)}$$ where $z$ is a primitive $k$th root of unity and $\delta_k(n)=0,1$ depending on whether $n$ is divisible by a $k$th power or not. There are $\phi(k)$ primitive roots, and an affirmative answer to the question would mean that accounts for all solutions.

Interchanging the order of summation, the statement says that $$a(m)=a(m)\sum_{d_1|n}a(d_1m)\cdots\sum_{d_k|d_{k-1}}a(d_km)$$ for all $m,n$ from which one can infer that $$a^{k+1}(n)=a(n)$$ for all $n$ and $$a(m)a(n)=a(mn)$$ whenever $a(m)$, $a(n)$ and $a(mn)$ are all non-zero.

From the stated property of $A$ we have that $A^k$ is a projection and when $a(n)$ is multiplicative that $$A^kf(n)=f(\textrm{rad}_{k-1}(n)).$$

Is every solution as such?

$\endgroup$

1 Answer 1

2
$\begingroup$

It's not the case when $k=2$ even: other solutions are, for any integer $j\ge2$, $$ a(n) = \begin{cases} 1, &\text{if } n=1, \\ -1, &\text{if } n=j, \\ 0, &\text{otherwise.} \end{cases} $$ In particular, this function is not multiplicative when $j=6$. Indeed, there seem to be lots of solutions that involve setting $a(p)=0$ for all primes $p$ and $a(j)\in\{-1,0\}$ for integers $j$ with $\Omega(j)=2$.

Similar remarks seem to hold for higher $k$, with $-1$ replaced by a $k$th root of unity.

$\endgroup$
3
  • $\begingroup$ Thank you, Greg. I have realised that my expecting solutions to be multiplicative came from starting with the primes and assuming $a(p)$ to be non-zero. $\endgroup$ Commented Oct 11, 2016 at 8:43
  • 1
    $\begingroup$ I'm fairly sure that $$ a(n) = \begin{cases} 1, &\text{if } n=1, \\ -1, &\text{if $n$ is prime}, \\ 0, &\text{otherwise.} \end{cases} $$ is also a (non-multiplicative) solution for $k=2$. There are probably hybrid solutions between this and $a(n)=\mu(n)$ as well. $\endgroup$ Commented Oct 11, 2016 at 16:53
  • $\begingroup$ I think it is too. I noticed that this afternoon but I can't quite see how to prove it, nor how to characterise the general case. $\endgroup$ Commented Oct 11, 2016 at 20:03

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .