Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I want someone to clarify me the next step in page 11. They write for $\delta=\prod p^l$ ,$k=\prod p^{\lambda}$ the sum: $\sum_{\delta |k} \delta^{1-s}\mu(\frac{k}{\delta})\prod_{p|\delta}(l+1-lp^{-s})$ is $$ \begin{align*} k^{1-s}&\prod_{p|k}(\lambda+1-\lambda p^{-s})-\sum_{p|k} (\frac{k}{p})^{1-s}(\lambda-(\lambda-1)p^{-s})\prod_{p'|k, p'\neq p} (\lambda+1-\lambda (p')^{-s})\\ &+\sum_{pp'|k}(\frac{k}{pp'})^{1-s} (\lambda-(\lambda-1)p^{-s})(\lambda-(\lambda-1)(p')^{-s})\prod_{p'|k, p''\neq p,p'} (\lambda+1-\lambda(p'')^{-s})-...\\ &= k^{1-s}\prod_{p|k}((\lambda+1-\lambda p^{-s})-\frac{1}{p^{1-s}}(\lambda-(\lambda-1)p^{-s})) \end{align*} $$

If I understand correctly, the LHS of the above is from exclusion-inclusive principle, but I don't understand the RHS, I don't see how did they get the second term in the brackets of the product, i.e $\frac{1}{p^{1-s}}(\lambda-(\lambda-1)p^{-s})$.

Any hints?


share|cite|improve this question
up vote 2 down vote accepted

The notation used in the book is somewhat confusing as, although there is nothing to show it, $\lambda$ is a function of $p$, and $l$ is a function of $p$ and $\delta$. To indicate this, I will write $l_\delta(p)$ and $\lambda(p)$ instead of $l$ and $\lambda$.

In the sum we start with, $$\Xi:=\sum_{\delta \mid k} \delta^{1-s}\mu(\frac{k}{\delta})\prod_{p\mid \delta}(l_\delta(p)+1-l_\delta(p)p^{-s}),$$ only values of $\delta$ for which $\mu(k/\delta)$ is nonzero contribute to the sum. Now, recall that $\mu(i)$ is nonzero only if $i$ is squarefree. In this case, if $i$ has $j$ prime factors, $\mu(i)$ equals $(-1)^j$. If $i$ is not squarefree, $\mu(i)$ is $0$. Let the long sum starting $k^{1-s} \prod_{p\mid k}\ldots$ be called $\Sigma$, and call its first piece $\Sigma_0$, its second piece $\Sigma_1$, and so on, so that $$\Sigma=\Sigma_0+\Sigma_1+\Sigma_2+\cdots.\qquad (*)$$ The expansion $(*)$ comes from setting $i:=k/\delta$ in $\Xi$ and looking successively at the cases where $j=0$, $1$, $2$, and so on. For example, if $j$ is $0$, we must have $i=1$, so $\delta=k$ and $$ \Sigma_0=\delta^{1-s}\mu(\frac{k}{\delta})\prod_{p\mid \delta}(l_\delta(p)+1-l_\delta(p)p^{-s})= k^{1-s}\prod_{p\mid k}(\lambda(p)+1-\lambda(p)p^{-s}). $$ This gives the first term of $\Sigma$. The second term of $\Sigma$ comes from the case when $j=1$. In this case, $i$ is prime, so $\delta=k/p$, for some $p$ dividing $k$. Therefore, the contribution to $\Xi$ is $$ \Sigma_1=\sum_{p\mid k, \delta=k/p} \delta^{1-s}\mu(\frac{k}{\delta})\prod_{p\mid \delta}(l_\delta(p)+1-l_\delta(p)p^{-s})$$ $$ =-\sum_{p\mid k} (\frac{k}{p})^{1-s} (\lambda(p)-(\lambda(p)-1)p^{-s}) \prod_{p'\mid k,\ p'\ne p}(\lambda(p')+1-\lambda(p')p'^{-s}). $$ The third term of $\Sigma$ comes from the case where $j=2$. In this case, $i=p p'$ for distinct primes $p$ and $p'$ dividing $k$, so $\delta=k/(p p')$. The contribution to $\Xi$ is then $$ \Sigma_2=\sum_{p, p'\mid k,\ p\ne p'} (\frac{k}{p p' })^{1-s} (\lambda(p)-(\lambda(p)-1)p^{-s}) (\lambda(p')-(\lambda(p')-1)p'^{-s})$$ $$ \prod_{p''\mid k,\ p''\ne p,\ p''\ne p'}(\lambda(p'')+1-\lambda(p'')p''^{-s}). $$ Continuing in this way, we can write down all of the expansion $(*)$. This proves that $\Xi=\Sigma$.

To see that the product that we finally end up with, $$\Pi= k^{1-s}\prod_{p\mid k}((\lambda(p)+1-\lambda(p) p^{-s})-\frac{1}{p^{1-s}}(\lambda(p)-(\lambda(p)-1)p^{-s})),$$ is equal to $\Sigma$, we can write $$\Pi=k^{1-s} \prod_{p\mid k} (A_p + B_p),\qquad (**)$$ where $$A_p=\lambda(p)+1-\lambda(p) p^{-s},\ \ \ B_p=-\frac{1}{p^{1-s}}(\lambda(p)-(\lambda(p)-1)p^{-s}).$$ Then, if we expand $(**)$ and sum together all terms which contain exactly $j$ factors of the form $B_p$, we recover $\Sigma_j$. For example, if we take $j=0$, we get $k^{1-s} \prod_{p\mid k} A_p$, which equals $\Sigma_0$. If we take $j=1$, we get $$k^{1-s} \sum_{p\mid k} B_p \prod_{p'\mid k, \ p'\ne p} A_{p'},$$ which equals $\Sigma_1$, and so on. This shows that $\Sigma=\Pi$.

Another way to derive $\Pi$ is to go back to the sum we started with, $$ \Xi:=\sum_{\delta \mid k} \delta^{1-s}\mu(\frac{k}{\delta})\prod_{p\mid \delta}(l_\delta(p)+1-l_\delta(p)p^{-s}). $$ If we rewrite this in terms of $\delta':=k/\delta$, and write $\delta'=\prod_{p\mid k} p^{l_{\delta'}(p)}$, we get $$ \Xi=\sum_{\delta'\mid k} (\frac{k}{\delta'})^{1-s} \mu(\delta') \prod_{p\mid k} (\lambda(p)-l_{\delta'}(p)+1-(\lambda(p)-l_{\delta'}(p))p^{-s}). $$ We can rewrite this as $$ \Xi=k^{1-s} \prod_{p\mid k} (\lambda(p)+1-\lambda(p) p^{-s}) \sum_{\delta'\mid k} G(\delta'),\qquad (***)$$ where $$G(\delta')= \frac{1}{\delta'^{1-s}} \mu(\delta') \prod_{p\mid \delta'}\frac{\lambda(p)-l_{\delta'}(p)+1-(\lambda(p)-l_{\delta'}(p))p^{-s}}{\lambda(p)+1-\lambda(p) p^{-s}}. $$ But now $G(\delta')$ is multiplicative, $G(1)=1$, and $G(\delta')$ vanishes unless $\delta'$ is squarefree. Therefore $$ \sum_{\delta'\mid k} G(\delta')=\prod_{p\mid k} (1 + G(p)) =\prod_{p\mid k} (1 - \frac{1}{p^{1-s}} \frac{\lambda(p)-(\lambda(p)-1)p^{-s}}{\lambda(p)+1-\lambda(p) p^{-s}}). $$ Plugging this into $(***)$ gives $\Xi=\Pi$.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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