13

Assume there exist some rationals $a, b$ such that $\sqrt[3]{a}, \sqrt[3]{b}$ are irrationals, but:

$$\sqrt[3]{a} + \sqrt[3]{b} = \frac{m}{n}$$

for some integers $m, n$

$$\implies \left(\sqrt[3]{a} + \sqrt[3]{b}\right)^3 = \frac{m^3}{n^3}$$

$$\implies a + b + 3 \cdot \sqrt[3]{ab}\left(\sqrt[3]{a} + \sqrt[3]{b}\right) = \frac{m^3}{n^3}$$

Since $a +b$, $\sqrt[3]{a} + \sqrt[3]{b}$ are rational, $\sqrt[3]{ab}$ must be rational as well.

For convenience let us say $\sqrt[3]{a} = p, \sqrt[3]{b} = q \implies pq$ is rational. This means, for all $i$, $p^iq^i$ is rational.

$$\implies (p + q)^2 = \frac{m^2}{n^2}$$

$$ = p^2 + q^2 + 2pq = \frac{m^2}{n^2}$$

Since $pq$ is rational, $2pq$ is rational and so is $p^2 + q^2$.

Assume, for some $i$ that $p^i + q^i$ and $p^{i-1} + q^{i-1}$ is rational.

$$\implies (p^i + q^i)(p + q) - pq(p^{i - 1} + q^{i - 1}) = p^{i+1} + q^{i+1}$$

is rational as well.

So for all $i$,

$$a^{\frac{i}{3}} + b^{\frac{i}{3}}$$

and

$$a^{\frac{i}{3}}b^{\frac{i}{3}}$$

are rational.

I know I'm really close to the answer, but it somehow just keeps slipping through my fingers.

ShBh
  • 6,054
Gerard
  • 4,264
  • 3
    $a = 2,\ b = -2 \Rightarrow \sqrt[3]{a} + \sqrt[3]{b} = \sqrt[3]{2} + \sqrt[3]{-2} = 0$ is rational, but $\sqrt[3]{2}$ is not. You should probably exclude the case $a = -b$ from your statement. – Vedran Šego Oct 18 '13 at 10:05
  • 1
    Yes, at first sight I was thinking that "if $\sqrt[3]{a}$ and $\sqrt[3]{b}$ are irrational, then $\sqrt[3]{a} + \sqrt[3]{b}$ is irrational" is false. You are only right if you exclude the case $a = -b$ – wilsonw Oct 18 '13 at 10:07
  • See the end of the accepted answer to https://math.stackexchange.com/questions/951609 – Watson Dec 25 '16 at 14:22
  • 1
    Related: https://math.stackexchange.com/questions/890821 – Watson Nov 25 '18 at 17:44
  • Related: https://math.stackexchange.com/questions/479092 – Watson Nov 28 '18 at 13:36

2 Answers2

10

As you have proved, $p^2+q^2\in\Bbb Q$ and $pq\in \Bbb Q$, so $$p-q=\frac{a-b}{p^2+pq+q^2}\in \Bbb Q.$$ Combining this with $p+q\in \Bbb Q$, we are done.

23rd
  • 16,234
  • 44
  • 70
  • 2
    As reminded by Vedran Šego and wilsonw in their comments, this answer is valid under the assumption that $p+q\ne 0$. – 23rd Oct 18 '13 at 22:06
3

Statement for the general case (for any $n>1$ in place of $3$)

Let $a,b$ be two positive rational numbers such that $\sqrt[n]{a}+\sqrt[n]{b}$ is a rational number for some natural number $n>1$. Then both $\sqrt[n]{a}$, $\sqrt[n]{b}$ are rational numbers.

Proof

Let $s=\sqrt[n]{a}+\sqrt[n]{b}\in\mathbb{Q}$. Then $\sqrt[n]{a}$ is a root of the polynomial $p(X)=X^n-a$ (If $\sqrt[n]{a}$ is not rational then $p$ is in fact the minimal polynomial). Again $\sqrt[n]{a}$ is also a root of $q(X)=(s-X)^n-b$. Now we show that $\sqrt[n]{a}$ is the unique common root of $p$ and $q$. Let $z$ be a common root of $p$ and $q$. Then $z=\sqrt[n]{a}\omega_1$ and $s-z=\sqrt[n]{b}\omega_2$ for some $n^{th}$ roots of unity $\omega_1,\omega_2$. Then $$\sqrt[n]{a}+\sqrt[n]{b}=\sqrt[n]{a}\omega_1+\sqrt[n]{b}\omega_2\tag{1}$$ Since $|\omega_1|=|\omega_2|=1$ therefore $|\mathrm{Re}(\omega_i)|\leq1$ with equality if and only if $\omega_i=\pm1$ for $i\in\{1,2\}$. Now comparing real parts in the equation $(1)$ we can easily conclude that $\omega_1=\omega_2=1$ implying that $\sqrt[n]{a}$ is the unique common root.

Then $$\gcd(p(X),q(X))=X-\sqrt[n]{a}$$ But $p(X),q(X)\in\mathbb{Q}[X]$ implies $X-\sqrt[n]{a}$ is in $\mathbb{Q}[X]$ then $\sqrt[n]{a}\in\mathbb{Q}$. Similarly one can prove that $\sqrt[n]{b}$ is also rational.

ShBh
  • 6,054
  • If not positive then there is a chance that $\sqrt[n]{a}$ or $\sqrt[n]{b}$ to be complex numbers. I just avoided that case. – ShBh Jul 31 '20 at 15:21
  • @BarryCipra Okay, I'm posting that. – ShBh Jul 31 '20 at 19:27
  • @BarryCipra Look, I have expanded that part. – ShBh Jul 31 '20 at 19:43
  • Thanks! @BarryCipra – ShBh Jul 31 '20 at 19:52
  • @BarryCipra In the original question, the positivity assumption is not needed because we are referring to the real cube roots of $a,b$. But in the general case when $n$ is even then we must need $a,b$ to be positive. Otherwise $\sqrt[n]{a},\sqrt[n]{b}$ would be purely imaginary and hence neither rational nor irrational. That's why I added the positivity assumption. – ShBh Jul 31 '20 at 19:58
  • Actually you do need positivity for the cube root case: For example, neither $\sqrt[3]7$ nor $\sqrt[3]{-7}$ is rational, but their sum is rational (namely $0$). – Barry Cipra Jul 31 '20 at 20:56
  • @BarryCipra Wow! It's right! I can see in previous comments that it was pointed out long time ago but the question was not updated. I think the OP missed this somehow. I think you should edit the question. – ShBh Aug 01 '20 at 00:34