Multiplicative function and Euler product

365 Views Asked by At

Theorem 1

Let $G \subset \mathbb{C}$ be an area and $\sum_{n=1}^{\infty}a_n e^{-\lambda _ns} $ and $\sum_{n=1}^{\infty}b_n e^{-\lambda _ns} $ two Dirichlet-series that converge on $G$ and represents the same funciton. Then $a_n = b_n$ for all $n$.

Theorem 2

Let $f: \mathbb{N} \to \mathbb{C}$ be a function and $F(s) = \sum_{n=1}^{\infty} f(n)n^{-s}$. Then $f$ is multiplicative if and only if $ \displaystyle F(s) = \prod_{p \in \mathbb{P}} \left( \sum_{l =0}^{\infty} f(p^l) p^{-ls} \right)$, where the product is absolut convergent.


We proofed Theorem 2 in the lecture and there is one step at the end (have a look at that first) that I don't understand. Here is an excerpt of the proof (proof from the right to the left).

For $N \in \mathbb{N}$ let $\{p_1, \ldots, p_r \} = \{ p \in \mathbb{P}; p \leq N \}$. It is

$\displaystyle F(s) = \sum_{n=1}^{N} f(n)n^{-s} = \sum_{p_1^{v_1} \cdots p_r^{v_r}\leq N} f(p_1^{v_1} \cdots p_r^{v_r}) \cdot (p_1^{v_1} \cdots p_r^{v_r})^{-s}$

and

$\displaystyle \prod_{p \leq N} \left( \sum_{l =0}^{\infty} f(p^l) p^{-ls} \right) = \sum_{p_1^{v_1} \cdots p_r^{v_r}\leq N} f(p_1^{v_1}) \cdots f(p_r^{v_r}) \cdot (p_1^{v_1} \cdots p_r^{v_r})^{-s} + \sum_{p_1^{v_1} \cdots p_r^{v_r} > N} f(p_1^{v_1}) \cdots f(p_r^{v_r}) \cdot (p_1^{v_1} \cdots p_r^{v_r})^{-s} $.

With the assumption we get

\begin{align}\displaystyle 0 &= \lim_{N \to \infty} \left(\sum_{n=1}^{N} f(n)n^{-s} - \prod_{p \leq N} \left( \sum_{l =0}^{\infty} f(p^l) p^{-ls} \right)\right) \\ &= \lim_{N \to \infty} \left ( \sum_{p_1^{v_1} \cdots p_r^{v_r}\leq N} \left[ f(p_1^{v_1} \cdots p_r^{v_r}) - f(p_1^{v_1}) \cdots f(p_r^{v_r}) \right]\cdot (p_1^{v_1} \cdots p_r^{v_r})^{-s} - \sum_{p_1^{v_1} \cdots p_r^{v_r} > N} f(p_1^{v_1}) \cdots f(p_r^{v_r}) \cdot (p_1^{v_1} \cdots p_r^{v_r})^{-s} \right) \end{align}

Since

$\displaystyle \sum_{p_1^{v_1} \cdots p_r^{v_r} > N} f(p_1^{v_1}) \cdots f(p_r^{v_r}) \cdot (p_1^{v_1} \cdots p_r^{v_r})^{-s}$

vanishes for $N \to \infty$, we have

\begin{equation}\displaystyle \sum_{r=1}^{\infty} \sum_{v_1,\ldots, v_r \in \mathbb{N}_0} \left[ f(p_1^{v_1} \cdots p_r^{v_r}) - f(p_1^{v_1}) \cdots f(p_r^{v_r}) \right]\cdot (p_1^{v_1} \cdots p_r^{v_r})^{-s} = 0 \end{equation}

Now here comes the step that isn't clear for me.

Using Theorem 1 we have, that all coefficients are 0.

My question is: Why can we apply Theorem 1? The left side of the equation resembles a Dirichlet-series but I don't see exactly why it is one and I need it to be one if I want to use Theorem 1...

Any help is much appreciated.

1

There are 1 best solutions below

0
On BEST ANSWER

If the summation is absolutely convergent (something that is not immediately obvious to me), then you should be able to rearrange the summation into the form $$\begin{equation}\displaystyle \sum_{n=0}^{\infty} \left[ f(p_{1n}^{v_{1n}} \cdots p_{r_nn}^{v_{r_nn}}) - f(p_{1n}^{v_{1n}}) \cdots f(p_{r_nn}^{v_{r_nn}}) \right]\cdot (p_{1n}^{v_{1n}} \cdots p_{r_nn}^{v_{r_nn}})^{-s} = 0 \end{equation}$$

for some clever sequencing of the indices of the multiple summation. From there, just let $$a_n = f(p_{1n}^{v_{1n}} \cdots p_{r_nn}^{v_{r_nn}})$$ $$b_n = f(p_{1n}^{v_{1n}}) \cdots f(p_{r_nn}^{v_{r_nn}})$$ $$\lambda_n = \log (p_{1n}^{v_{1n}} \cdots p_{r_nn}^{v_{r_nn}}).$$