4
$\begingroup$

Let $\omega(n)$ denote the number of distinct prime divisors of $n>1$, with $\omega(1)=0$.

(a) Show that $2^{\omega(n)}$ is a multiplicative function.

(b) Prove that $$\sigma(n^2)=\sum_{d\mid n} 2^{\omega(d)}.$$

I have done the part (a) and I am stuck by (b).

First, I set $d=p_1^{e_1}\cdots p_k^{e_{k}}$ be a factor of $n$. Then I don't know what's the next step.

$\endgroup$
  • $\begingroup$ it is enough to show that $\sigma(p^{2k}) = \sum_{d | p^k} 2^{w(d)} $ and that $f(n) = \sigma(n^2)$ is multiplicative $\endgroup$ – reuns Mar 28 '16 at 17:00
  • $\begingroup$ (because if $g(n)$ is multiplicative, then $h(n) = \sum_{d | n} g(d)$ is multiplicative, and to prove that two multiplicative functions $f(n) ,g(n)$ are equal, it is enough to show that $f(p^k) = g(p^k)$ for every prime power $p^k$) $\endgroup$ – reuns Mar 28 '16 at 17:04
  • $\begingroup$ Is this even true? With $n$ a prime number $p$ we get $$\sigma(n^2) = 1+p+p^2$$ and $$\sum_{d|n} 2^{\omega(d)} = 1 + 2 + 2 = 5,$$ a constant. $\endgroup$ – Marko Riedel Mar 28 '16 at 22:24
4
$\begingroup$

Suppose we seek to show that $$\tau(n^2) = \sum_{d|n} 2^{\omega(d)}.$$

This can be done using Dirichlet series and Euler products. We have for the RHS and $$\sum_{n\ge 1} \frac{1}{n^s} 2^{\omega(n)}$$ the Euler product $$\prod_p \left(1 + \frac{2}{p^s} + \frac{2}{p^{2s}} + \frac{2}{p^{3s}} +\cdots\right).$$ which is $$\prod_p \left(-1 + 2\frac{1}{1-1/p^s}\right) = \prod_p \frac{-1+1/p^s+2}{1-1/p^s} \\ = \prod_p \frac{1+1/p^s}{1-1/p^s} = \prod_p \frac{1-1/p^{2s}}{(1-1/p^s)^2} = \frac{\zeta(s)^2}{\zeta(2s)}.$$

Therefore $$\sum_{n\ge 1} \frac{1}{n^s} \sum_{d|n} 2^{\omega(d)} = \frac{\zeta(s)^3}{\zeta(2s)}.$$

On the other hand we have $$\sum_{n\ge 1} \frac{1}{n^s} \tau(n^2) \\= \prod_p \left(1 + (2+1) \frac{1}{p^s} + (4+1) \frac{1}{p^{2s}} + (6+1) \frac{1}{p^{3s}} + (8+1) \frac{1}{p^{4s}} + \cdots\right).$$

This is $$\prod_p \left(1+\frac{1/p^s}{1-1/p^s} + \sum_{k\ge 1} \frac{2k}{p^{ks}} \right) \\ = \prod_p \left(1+\frac{1/p^s}{1-1/p^s} + 2 \frac{1/p^s}{(1-1/p^s)^2} \right).$$

To aid in simplification we put $z=1/p^s$ to get for the inner term

$$1 + \frac{z}{1-z} + \frac{2z}{(1-z)^2}$$

This simplifies to $$\frac{1+z}{(1-z)^2}.$$

On the other hand $$\frac{\zeta(s)^3}{\zeta(2s)} = \prod_p \frac{1-z^2}{(1-z)^3} = \prod_p \frac{1+z}{(1-z)^2}.$$

We have equality, QED.

$\endgroup$
4
$\begingroup$

A Combinatorial Proof (Sketch)

I assume that, by $\sigma$, you mean $\sigma_0=\tau$, the divisor-counting function. Let $$S(n):=\Big\{(t,d)\in\mathbb{N}\times\mathbb{N}\,\Big|\,t\mid n^2,\,d\mid n,\text{ and }d\text{ is smallest such that }n^2\mid td^2\Big\}\,.$$ First, verify that, for each $t\in\mathbb{N}$ such that $t\mid n^2$, there exists a unique $d\in\mathbb{N}$ such that $(t,d)\in S(n)$. This proves that $\big|S(n)\big|=\sigma_0\left(n^2\right)$.

Next, justify that, for each $d\in\mathbb{N}$ such that $d\mid n$, there are precisely $2^{\omega(d)}$ values of $t\in\mathbb{N}$ such that $(t,d)\in S(n)$. This shows that $\big|S(n)\big|=\sum\limits_{d\mid n}\,2^{\omega(d)}$. Therefore, $\sigma_0\left(n^2\right)=\sum\limits_{d\mid n}\,2^{\omega(d)}$.

In fact, one can show also that $$\sigma_0\left(n^k\right)=\sum\limits_{d\mid n}\,k^{\omega(d)}$$ for all $k\in\mathbb{N}_0$ (where $0^0:=1$). One needs to only count the number of the elements of the set $$S(n,k):=\Big\{(t,d)\in\mathbb{N}\times\mathbb{N}\,\Big|\,t\mid n^k,\,d\mid n,\text{ and }d\text{ is smallest such that }n^k\mid td^k\Big\}\,.$$

The hidden portion below is a remark, almost completely unrelated to the question.

Maybe, we can use this result to extend $\sigma_0$ onto $\mathbb{Q}_{>0}$, by defining $\sigma_0\left(\frac{1}{n}\right):=\sum\limits_{d\mid n}\,(-1)^{\omega(d)}$ and $\sigma_0\left(\frac{u}{v}\right):=\sigma_0(u)\,\sigma_0\left(\frac{1}{v}\right)$ if $\gcd(u,v)=1$. For example, if $p$ is a prime natural number and $j\in\mathbb{N}_0$, then $\sigma_0\left(\frac{1}{p^j}\right)=1-j$. This result is consistent with a more natural definition $\sigma_0\left(\prod\limits_{i=1}^\ell\,p_i^{\alpha_i}\right):=\prod\limits_{i=1}^\ell\,\left(1+\alpha_i\right)$ if $p_1,\ldots,p_\ell$ are pairwise distinct prime natural numbers and $\alpha_1,\ldots,\alpha_\ell\in\mathbb{Z}$.
We also have the identity $$\sigma_0\left(n^k\right)=\sum\limits_{d\mid n}\,k^{\omega(d)}$$ for all $n\in\mathbb{N}$ and $k\in\mathbb{Z}$. This identity can be yet again extended in the form $$\sigma_0\left(r^k\right)=\sum\limits_{q\mid r}\,s(q)\,k^{\omega(q)}$$ for all $r\in\mathbb{Q}_{>0}$ and $k\in\mathbb{Z}$ as follows. For $q,r\in\mathbb{Q}_{>0}$, we say that $q\mid r$ if $q=\frac{u}{v}$ and $r=\frac{x}{y}$, with $u,v,x,y\in\mathbb{N}$, $\gcd(u,v)=1$, and $\gcd(x,y)=1$, satisfy $u\mid x$ and $v\mid y$. For $q\in\mathbb{Q}_{>0}$, $\omega(q):=\omega(u)+\omega(v)$ and $s(q):=(-1)^{\omega(v)}$ if $q=\frac{u}{v}$ with $u,v\in\mathbb{N}$ and $\gcd(u,v)=1$.
Indeed, if $\sigma_z(n):=\sum\limits_{d\mid n}\,n^z=\prod\limits_{i=1}^\ell\,\left(\frac{p_i^{z\left(1+\alpha_i\right)}-1}{p_i^z-1}\right)$, where $z\in\mathbb{C}\setminus\{0\}$ and $n=\prod\limits_{i=1}^\ell\,p_i^{\alpha_i}$ with distinct primes $p_1,\ldots,p_\ell$ and positive integers $\alpha_1,\ldots,\alpha_\ell$, then we obtain a similar identity $$\sigma_z\left(n^k\right)=\sum\limits_{d\mid n}\,\left(\frac{n}{d}\right)^{kz}\,\varpi_z^k(d)\,,$$ where $\varpi_z^k(n):=\prod\limits_{\substack{{p\mid n}\\p\text{ is prime}}}\,\left(\frac{p^{kz}-1}{p^z-1}\right)$ for $z\neq 0$ and $\varpi_0^k(n):=k^{\omega(n)}$. We extend $\sigma_z$ onto $\mathbb{Q}_{>0}$ similarly. Furthermore, if $q=\frac{u}{v}$ where $u,v\in\mathbb{N}$ with $\gcd(u,v)=1$, then we define $\varpi^k_z(q):=\varpi^k_z(u)\,\varpi^k_z(v)$ and $s_z(q):=\prod\limits_{\substack{{p\mid v}\\{p\text{ is prime}}}}\,\left(\frac{-1}{p^z}\right)=\frac{s(q)}{\big(\text{rad}(v)\big)^z}$. Finally, we get the identity $$\sigma_z\left(r^k\right)=\sum\limits_{q\mid r}\,s_z(q)\,\left(\frac{r}{q}\right)^{kz}\,\varpi^k_z(q)$$ for all rational numbers $r>0$, for every integer $k$, and for every complex number $z\neq 0$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.