When I read an introductory algebraic number theory textbook, there is a problem like this:

If $p$ is an odd prime, prove $\displaystyle\sum_{i=1}^{p-1}\sqrt[p]{\frac{i}{p}} $ is an algebriac integer.

Note that it is in the first chapter, so I think that it does not require much knowledge, but I have no idea how to solve it.

  • $\begingroup$ @DietrichBurde Sorry, I am new to this. I saw that an algebraic integer is a root of some monic polynomial with $\mathbb{Z}$ coefficients, but $i/p$ is not an integer, why is it an algebraic integer? I also don't get what you mean by "$\overline{Q}$". Can you explain it more? Thank you. $\endgroup$ – Eric Sun May 27 at 9:24
  • $\begingroup$ Which textbook, please? $\endgroup$ – Gerry Myerson May 27 at 9:39
  • $\begingroup$ Eric, I thought you meant algebraic numbers $\overline{\Bbb {Q}}$. For algebraic integers see for example this question. $\endgroup$ – Dietrich Burde May 27 at 9:41
  • $\begingroup$ @GerryMyerson A Conversational Introduction to Algebraic Number Theory written by Paul Pollack $\endgroup$ – Eric Sun May 27 at 9:48
  • $\begingroup$ @DietrichBurde The method in that link does work I think. It is not just a finite sum. $p$ can be arbitrarily large. $\endgroup$ – Eric Sun May 27 at 9:50

It is enough to show that $$ \alpha = \sqrt[p]{\frac{i}{p}} + \sqrt[p]{\frac{p-i}{p}} $$ is an algebraic integer for $1\leq i\leq (p-1)/2$, since the original sum is sum of such numbers. We have $$ \alpha^{p} = \frac{(\sqrt[p]{i} + \sqrt[p]{p-i})^{p}}{p} = \frac{1}{p} \left(i + (p-i) + \sum_{k=1}^{p-1} \binom{p}{k}i^{k/p}(p-i)^{(p-k)/p}\right) \\= 1 + \sum_{k=1}^{p} \frac{1}{p}\binom{p}{k}i^{k/p}(p-i)^{(p-k)/p}. $$ Since $\binom{p}{k}$ is a multiple of $p$ for $1\leq k\leq p-1$ and $i^{k/p}, (p-i)^{(p-k)/p}$ are obviously algebraic integers, so is $\alpha^{p}$. So $f(\alpha^{p}) = 0$ for some monic $f(x) \in \mathbb{Z}[x]$, which implies that $g(\alpha) = 0$ for $g(x) = f(x^{p})$, which is clearly monic. Thus $\alpha$ is an algebraic integer.


To show this sum (call it $S$) is an algebraic integer, you want to find a monic polynomial that makes this sum vanish. You can use a few facts about algebraic integers:

  1. If for some integer $k$, a number $\alpha^k$ is an algebraic integer, then $\alpha$ is an algebraic integer: take $P$ monic with integer coefficient such that $P(\alpha^k) = 0$, then $Q = P \circ X^k$ is also monic, with integer coefficients, such that $Q(\alpha) = 0$.
  2. Any quantity of the form $\sqrt[a]{b}$ for $a$ and $b$ integers is an algebraic integer: it is the root of $X^a - b$.
  3. The sum of two algebraic integers is an algebraic integer. You will probably find a proof of this fact in your textbook, the main ingredient is that a number $\alpha$ is an algebraic integer if and only if $\mathbb{Z}[\alpha]$ is a finitely generated $\mathbb{Z}$-module, so that $\mathbb{Z}[\alpha + \beta]$ is a submodule of $\mathbb{Z}[\alpha][\beta]$, which is finitely generated, so that $\mathbb{Z}[\alpha + \beta]$ is finitely generated, hence $\alpha + \beta$ is an algebraic integer.

Now, to show $S$ is an algebraic integer, it is enough to show that S so a suitable power is a sum of known algebraic integers. This suitable power is obviously $p$: \begin{align} S^p &= \left(\sum\limits_{k=1}^{p-1}\sqrt[p]{\frac{k}{p}}\right)^p\\ &= \sum\limits_{i_1 + \cdots + i_{p-1} = p}\dbinom{p}{i_1,\cdots,i_{p-1}}\prod\limits_{j=1}^{p-1}\left(\sqrt[p]{\frac{j}{p}}\right)^{i_j}\\ &= \sum\limits_{i_1 + \cdots + i_{p-1} = p}\dbinom{p}{i_1,\cdots,i_{p-1}}\left(\sqrt[p]{\prod\limits_{j=1}^{p-1}\frac{j^{i_j}}{p^{i_j}}}\right)\\ &= \sum\limits_{i_1 + \cdots + i_{p-1} = p}\dbinom{p}{i_1,\cdots,i_{p-1}}\left(\sqrt[p]{\frac{\prod\limits_{j=1}^{p-1}j^{i_j}}{p^{\sum\limits_{j=1}^{p-1}i_j}}}\right)\\ &= \sum\limits_{i_1 + \cdots + i_{p-1} = p}\dbinom{p}{i_1,\cdots,i_{p-1}}\sqrt[p]{\frac{\prod\limits_{j=1}^{p-1}j^{i_j}}{p^{p}}}\\ &= \sum\limits_{i_1 + \cdots + i_{p-1} = p}\frac{1}{p}\dbinom{p}{i_1,\cdots,i_{p-1}}\sqrt[p]{\prod\limits_{j=1}^{p-1}j^{i_j}}\\ \end{align}

But $p$ divides $\dbinom{p}{i_1,\cdots,i_{p-1}}$ if there is no $i_j$ equal to $p$, indeed $\dbinom{p}{i_1,\cdots,i_{p-1}} = \frac{p!}{i_1!i_2!\cdots i_{p-1}!}$, looking at $p$-valuations, if no $i_j$ is $p$, then $p$ divides this coefficient.

For the remaining coefficients, one has the sum \begin{align} \sum\limits_{j=1}^{p-1}\frac{j}{p} &= \frac{p(p-1)}{2p}\\ &= \frac{p-1}{2} \end{align} which is an integer because $p$ is odd.

Thus the $p$-th power of S is the sum of agebraic integers. Hence an algebraic integer, by the first fact listed above, $S$ is an algebraic integer.


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.