Let $p, q \in \mathbb Q$, $n \in \mathbb Z^+$ and label $a = \sqrt[n]p, b=\sqrt[n]q$.

Conjecture: If $a + b$ is a non-zero rational, then both $a$ and $b$ are rational.

(Preliminary question: is this a known result I'm not aware of?)

I believe I have found a partial proof of the above. Namely,

  • I first proved that $ab$ is rational for $n = 1, 2, 3$.
    • For $n = 2$, $(a + b)^2 = a^2 + 2ab + b^2$, so $$ab = \frac{(a+b)^2-a^2-b^2}{2} \in \mathbb Q$$
    • For $n = 3$, $(a + b)^3 = a^3 + 3ab(a + b) + b^3$, so $$ab = \frac{(a+b)^3-a^3-b^3}{3(a+b)} \in \mathbb Q$$
  • As it turns out, an additional assumption that $ab \in \mathbb Q$ allows proving the conjecture.
  • The question: how can I prove $ab \in \mathbb Q$ for $n > 3$? (or what numbers are a counterexample?)
  • The details of the proof assuming $ab \in \mathbb Q$ follow.
    • Consider the polynomial $(x - a)(x - b) = x^2 - (a + b)x + ab$. Note that its coefficients are rational.
    • This means that $\Pi \in \mathbb Q[x]$, the minimal polynomial of $a$, is of degree at most 2.
    • Hence, $\deg \Pi \in \{1, 2\}$. If $\deg \Pi = 1$, then $a$ is rational, which was to be proven, so let's assume that $\deg \Pi = 2$ and hope for a contradiction.
    • As the minimal polynomial is unique, $\Pi = (x - a)(x - b)$. Moreover, $\Pi$ divides $x^n - p$, since the latter has a root at $a$.
    • Hence, the roots of $\Pi$ are a subset of the roots of $x^n - p$.
    • For odd $n$, we have $\{a, b\} \subseteq \{a\}$, so $a = b$.
    • For even $n$, we have $\{a, b\} \subseteq \{a, -a\}$. $b$ can't be equal to $-a$, since roots of even degree are nonnegative (or imaginary, but in that case $a + b$ wouldn't be rational). Hence, $a = b$.
    • In both cases, from the assumption $a + b \in \mathbb{Q} \setminus \{0\}$, we get $2a \in \mathbb{Q} \implies a \in \mathbb Q$ and therefore $\deg \Pi = 1$, which is a contradiction.
  • 1
    $\begingroup$ Are you familiar/comfortable with field theory? You are essentially asking if the nth roots are linearly independent over Q. $\endgroup$ – Calvin Lin Nov 29 '19 at 23:46
  • $\begingroup$ @Calvin Lin No, I've started studying abstract algebra a few days ago after someone pointed me to the concept of minimal polynomials. (In a hope to eventually learn about algebraic number theory in general). $\endgroup$ – NieDzejkob Nov 30 '19 at 11:18

It may not be true if $n$ is odd and $q=-p$. But otherwise your statement is correct. Let's prove it in the case $p$, $q \ge 0$. The following is a standard trick that avoids the Galois theory. It is enough to show that $\sqrt[n]{p}$ is a rational fraction in $\sqrt[n]{p} + \sqrt[n]{q}$ with rational coefficients. Consider the two polynomials $X^n - p$ and $(\sqrt[n]{p} + \sqrt[n]{q}-X)^n - q$. The first polynomial has the roots $\omega \cdot \sqrt[n]{p}$, while the second polynomial has the roots $(\sqrt[n]{p} + \sqrt[n]{q}) -\omega' \cdot \sqrt[n]{q}$. where $\omega$, $\omega'$ are $n$-th roots of $1$. These polynomials have exactly one common root $\sqrt[n]{p}$. Therefore, their gcd is $(X-\sqrt[n]{p})$. Now, the coefficients of the gcd of two polynomials can be expressed rationally in terms of the coefficients of the given polynomials. We conclude that there exists a rational function $R(t) \in \mathbb{Q}(t)$ such that $$\sqrt[n]{p} = R(\sqrt[n]{p} + \sqrt[n]{q})$$

  • $\begingroup$ Note that $q = -p$ violates the assumption that $a + b$ is a non-zero rational. I'll try to find some information about the gcd of polynomials and understand your proof. Thank you. $\endgroup$ – NieDzejkob Nov 30 '19 at 13:26
  • $\begingroup$ @NieDzejkob: You are welcome! The most general result would be that roots of positive rationals are linearly independent over $\mathbb{Q}$ if they are pairwise incomensurable. you can check for instance math.stackexchange.com/questions/158722/… $\endgroup$ – orangeskid Dec 2 '19 at 0:45
  • $\begingroup$ I thought quite a bit about various properties of the gcd of polynomials and I think I understand now. Though, I can't see where the assumption that $p, q \geq 0$ is used. Does this proof break for negative $p$ or $q$? I haven't noticed any such moment. $\endgroup$ – NieDzejkob Dec 2 '19 at 14:45
  • $\begingroup$ @NieDzejkob: the proof uses the fact that $\sqrt[n]{p}+ \sqrt[n]{q}= \omega\sqrt[n]{p}+ \omega' \sqrt[n]{q}$ implies $\omega = \omega' = 1$. Now this is straightforward if $ p$,$q$ are positive , but not that simple if say $q$ is negative. $\endgroup$ – orangeskid Dec 2 '19 at 18:25

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.