3
$\begingroup$

Show that if $p$ is and odd prime and $h$ is an integer, $1\le h \le p$, then

$$\displaystyle\sum_{n=1}^{p}\left(\sum_{m=1}^{h}\left(\frac{m+n}{p}\right)\right)^2=h(p-h)$$ where $\left(\frac{m+n}{p}\right)$ denotes the Jacobi symbol.

My solution:

For $h=1$, we have $\left(\frac{1+n}{p}\right)^2$ is always 1 or 0. It is 0 only when $n=p-1$. So the sum comes out to be $p-1$, which is accordance with $h(p-h)$. Similarly, for $h=p$, and the sum is zero.

But I am having trouble when $ h \neq 1 or p$, how will I proceed in that case?

This question has been taken from the book : An introduction to theory of numbers by Niven, Zuckerman, Montgomery. Section 3.3., question 19. Thanks in advance.

$\endgroup$
3
$\begingroup$

The square of the inner sum is $$\sum_{m_1,m_2=1}^h\left(\frac{(m_1+n)(m_2+n)}{p}\right).$$ The whole sum is $$\sum_{m_1,m_2=1}^h\sum_{n=1}^p\left(\frac{(m_1+n)(m_2+n)}{p}\right) =\sum_{m_1,m_2=1}^hS(m_1,m_2)$$ say. When $m_1=m_2$ then $S(m_1,m_2)=p-1$. When $m_1\ne m_2$ then $$S(m_1,m_2)=\sum_{n=1}^p\left(\frac{n(n+2m')}p\right)$$ where $2m'\equiv m_1-m_2\not\equiv0\pmod p$. We can drop the $n=p$ term, and then let $n'$ be the mod $p$ inverse of $n$ gives $$S(m_1,m_2)=\sum_{n=1}^{p-1}\left(\frac{n(n+2m')}p\right) =\sum_{n'=1}^{p-1}\left(\frac{1+2m'n'}p\right)=-1.$$ Therefore the original sum is $$h(p-1)-(h^2-h)=h(p-1-(h-1))=h(p-h).$$

$\endgroup$
  • $\begingroup$ I am having difficult in understanding the following:$$S(m_1,m_2)=\sum_{n=1}^p\left(\frac{n(n+2m')}p\right)$$ where $2m'\equiv m_1-m_2\not\equiv0\pmod p$. We can drop the $n=p$ term, and then let $n'$ be the mod $p$ inverse of $n$ gives $$S(m_1,m_2)=\sum_{n=1}^{p-1}\left(\frac{n(n+2m')}p\right) =\sum_{n'=1}^{p-1}\left(\frac{1+2m'n'}p\right)=-1.$$ can you please explain this part? Thanks for the solution. $\endgroup$ – Epsilon Delta Oct 17 at 6:47
  • $\begingroup$ @EpsilonDelta (i) is just a change of variable, (ii) is extracting a factor $(n^2/p)$. $\endgroup$ – Lord Shark the Unknown Oct 17 at 17:38
  • $\begingroup$ Thanks for the help $\endgroup$ – Epsilon Delta Oct 21 at 17:08

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.