According to the theorem, if

$$1+\frac{1}{2}+\frac{1}{3}+\frac{1}{4}+\frac{1}{5}+\cdots+\frac{1}{p-1} =\frac{r}{q}$$

then we have to prove that $r\equiv0 \pmod{p^2}$.

(Given $p>3$, otherwise $1+\dfrac{1}{2}=\dfrac{3}{2}$, $3 \not\equiv 0 \pmod 9$.)

I guess there's a $(\bmod p)$ solution for this, but I don't really get how to start it.


The group theoretic proof on Wikipedia is good.

Alternatively, another nice way is, the second line is working in $\mathbb{Z}/p\mathbb{Z}$,

$$ \frac2p\sum_{n=1}^{p-1} \frac{1}{n} = \frac2p \sum_{n=1}^{(p-1)/2} \frac{1}{n} + \frac{1}{p-n} = 2 \sum_{n=1}^{(p-1)/2} \frac{1}{n(p - n)}$$ $$ \equiv -\sum_{n=1}^{p-1} \frac{1}{n^2} \equiv - \sum_{n=1}^{p-1} n^2 \equiv -\frac{(p-1)p(2p - 1)}{6}$$

which appears in this article by Christian Aebi and Grant Cairns.

A more hands-on approach can be found in these notes by Timothy Choi.

Finally, an excellent survey and additional results here by Romeo Mestrovic.

  • 1
    $\begingroup$ I love this argument, and of course +1 upvote! $\endgroup$ Mar 9 '13 at 15:25
  • $\begingroup$ How can we show a congruence modulo $p^2$ by working in $\mathbb{Z}/p\mathbb{Z}$ ? It seems to me your computation only shows that your sum is $0$ modulo $p$. You can get that property by $\sum_{k \in \mathbb{Z}/p\mathbb{Z}^*} k^{-1} = \sum_{k \in \mathbb{Z}/p\mathbb{Z}^*} k$ and notice that $\sum_{k \in \mathbb{Z}/p\mathbb{Z}^*} k = 0$ when $p > 2$ because it is translation invariant. $\endgroup$
    – Joel Cohen
    Jun 9 '15 at 14:33
  • $\begingroup$ It is the same answer as Mercio, the wording was inadequate $\endgroup$
    – reuns
    May 22 '20 at 23:50

$ (1 + \frac 1 {p-1}) + (\frac 1 2 + \frac 1 {p-2}) + \ldots + (\frac 1 {(p-1)/2} + \frac 1 {(p+1)/2}) = \frac p {p-1} + \frac p {2(p-2)} + \ldots + \frac p {((p-1)(p+1)/4)} \\ = p \sum_{k=1}^{(p-1)/2} \frac 1 {k(p-k)} = \frac p 2 \sum_{k=1}^{p-1} \frac 1 {k(p-k)} $.

Now, working on this second sum in the field $\Bbb Z / p \Bbb Z$, and assuming $p>3$ you get :

$\sum_{k=1}^{p-1} \frac 1 {k(p-k)} = - \sum_{k=1}^{p-1} \frac 1 {k^2} = - \sum_{k=1}^{p-1} k^2 = -(p-1)p(2p-1)/6 = 0$.

If you don't know the sum of the first consecutive squares formula, you can still get it with a bit more work :

Pick a number $a \neq 1$ such that $a$ is a square modulo $p$ (again assuming $p>3$). Then, $\sum k^2 (1-a) = \sum k^2 - \sum (ak^2) = \sum k^2 - \sum k^2 = 0$. Since $1-a \neq 0$, $\sum k^2 = 0$.

(when $p = 3$, $1+ \frac 1 2 = \frac 3 2$ which is not a multiple of $3^2$)


Note that $\displaystyle \sum_{1 \le i \le p-1}\dfrac{1}{i}=\displaystyle \sum_{1 \le i \le \frac{p-1}{2}}\left ( \dfrac{1}{i}+\dfrac{1}{p-i} \right )=p \displaystyle \sum_{1 \le i \le \frac{p-1}{2}}\dfrac{1}{i(p-i)}=p \cdot \dfrac{a}{b}$. Thus, enough to show that $a \equiv 0 \mod p$ or, equivalently, that integer

$$S=\displaystyle \sum_{1 \le i \le \frac{p-1}{2}}\dfrac{(p-1)!}{i(p-i)}$$

is a multiple of $p$. Let $r_i \in \mathbb{Z}_p$ st $ir_i \equiv 1 \mod p$. Note que $r_{p-i} \equiv -r_i \mod p$. So,

$$S \equiv \displaystyle \sum_{1 \le i \le \frac{p-1}{2}}\dfrac{(p-1)!}{i(p-i)}\cdot ir_i \cdot (p-i)r_{p-i} \mod p$$ $$S \equiv \displaystyle \sum_{1 \le i \le \frac{p-1}{2}}r_i^2 \mod p$$ by Wilson's Theorem. Therefore,

$$S \equiv \displaystyle \sum_{1 \le i \le \frac{p-1}{2}}i^2 = \dfrac{p(p^2-1)}{24} \mod p$$

So, $S$ is a multiple of $p$.


I second Joel's comments. The proof given by Christian Aebi and Grant Cairns, in a sense, is misleading & harmful. Directly jumping into $\mathbb{Z}/p\mathbb{Z}$ without carefully observing suitability and justification is misleading and wrong. For example, when computing in $\mathbb{Z}/p\mathbb{Z}$, let $p=5$, then $\frac{1}{2} + \frac{1}{3}\equiv 3 + 2 \equiv0,$ but the numerator $mod\ 25$ is 5.

The original problem asks to prove the numerator $mod\ p^2$ is zero. One needs to be very careful when using $\mathbb{Z}/p\mathbb{Z}.$ For example, $\frac{3}{7} + \frac{1}{13}$'s numerator $mod\ 17$ is 12. But if calculating in $\mathbb{Z}/p\mathbb{Z}$ for $p=17$, one gets $\frac{1}{7}\equiv5, \frac{1}{13}\equiv4$ and $\frac{3}{7} + \frac{1}{13}\equiv2$. One can use $\mathbb{Z}/p\mathbb{Z}$ when handling $\sum_{k=1}^{(p-1)/2}\frac{1}{r(p-r)}$ is due to the guarantee of Wilson's theorem in the background, and the fact that we only care about if the modulus result is zero.


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.