I was proving $\sqrt 6 \notin \Bbb Q$, by assuming its negation and stating that: $\exists (p,q) \in \Bbb Z \times \Bbb Z^*/ \gcd(p,q) = 1$, and $\sqrt 6 = (p/q)$.

$\implies p^2 = 2 \times 3q^2 \implies \exists k \in \Bbb Z; p = 2k \implies 2k^2 = 3q^2$ and found two possible cases, either $q$ is even or odd, if even we get contradiction that $\gcd(p, q) \neq 1$, if odd we get contradiction that $2k^2 = 3q^2$.

Is it a right path for reasoning it?

  • $\begingroup$ Is the $k$ a typo? What is $p, k$ and $q$? Why is $q$ being even a condradiction with $\gcd(p,q) = 1$? And why is $2k^2 = 3q^2$ a contradiction? You just said $2k^2 = 3q^2$ so that isn't a contradiction? You are on the right track but you haven't explained any of your arguments. $\endgroup$ – fleablood Nov 15 '18 at 2:13
  • $\begingroup$ Why is $q$ even a contradiction? What if $k$ or $p$ (whats the difference) is odd? $\endgroup$ – fleablood Nov 15 '18 at 2:20
  • $\begingroup$ I rearranged it! $\endgroup$ – Papa Nov 15 '18 at 3:51

How about a proof by descent?

First show that $2^2<6<3^2$. Then if $\sqrt{6}$ is to be rational it must have a form $p/q$ where $p,q\in \mathbb{Z}, q>0, 2q<p<3q$. By simple algebra the square root is also equal to $6q/p$, thus

$p/q=6q/p\text{.....Eq. 1}$.

Now if $a/b=c/d$ then also


for any coefficients $m,n$ where the denominator is nonzero. In particular, Eq. 1 implies

$p/q=(3p-6q)/(3q-p)\text{.....Eq. 2}$

where we already have $2q<p<3q$ and thus $0<3q-p<q$. So the proposed rational fraction $p/q$ must be equal to an alternative rational fraction with a smaller positive denominator. This causes an infinite descent contradiction forcing the assumption of a rational value to be false.

We can form a similar proof for the square root of any natural number that is not a squared integer. "Not a squared integer" is needed because the square root must be strictly between two adjacent integers to obtain a descent of positive denominators.


Assume $\sqrt{6} = {a \over b}$ with $a, b \in \mathbb{Z}$ . This implies

$$2 \cdot 3 b^2 = a^2$$

Use the fundamental theorem of algebra to decompose both sides into a unique prime factorization. There are an even number of factors of $2$ on the rhs and an odd number of factors of $2$ on the lhs... a contradition. Same for factors of $3$.

Thus $\sqrt{6} \neq {a \over b}$, i.e., is irrational.

  • $\begingroup$ Yeah, I tried to reach it without prime factorisation. A second look? $\endgroup$ – Papa Nov 15 '18 at 2:42
  • $\begingroup$ You are considering the value of the rhs and lhs of the equation. I am instead looking at the number of prime factors. Frankly, I think that is so much more elegant... but I suppose this is a matter of taste. Mine applies to $\sqrt{5/3}$ whereas yours doesn't... at least not as directly. $\endgroup$ – David G. Stork Nov 15 '18 at 2:47
  • $\begingroup$ Our professor used the value way, so I am following his steps for this, though, thanks! $\endgroup$ – Papa Nov 15 '18 at 2:50
  • $\begingroup$ Show him/her the better way! And stand out as a student! $\endgroup$ – David G. Stork Nov 15 '18 at 2:51
  • $\begingroup$ Well, that doesn t apply in Morocco. Pray for me I win visa lottery so I can join a better faculty in the US :) $\endgroup$ – Papa Nov 15 '18 at 3:49

I'll play more generally and see what happens.

Suppose $\sqrt{m} =\dfrac{u}{v} $ where $m = \prod_{p \in P} p^{m_i}, u = \prod_{p \in P} p^{u_i}, v = \prod_{p \in P} p^{v_i} $.

Then $\prod_{p \in P} p^{m_i} =\dfrac{u^2}{v^2} =\dfrac{ \prod_{p \in P} p^{2u_i}}{\prod_{p \in P} p^{2v_i}} $ so $\prod_{p \in P} p^{m_i}\prod_{p \in P} p^{2v_i} =\prod_{p \in P} p^{2u_i} $ or $\prod_{p \in P} p^{m_i+2v_i} =\prod_{p \in P} p^{2u_i} $.

By unique factorization, $m_i+2v_i =2u_i $, so $m_i =2u_i-2v_i =2(u_i-v_i) $.

Therefore $m =\prod_{p \in P} p^{m_i} =\prod_{p \in P} p^{2(u_i-v_i)} =\left(\prod_{p \in P} p^{u_i-v_i}\right)^2 $ so $m$ is a perfect square.

Therefore the square root of an integer is rational only if it is a square.

I'll now try to generalize this to $k$-th roots, with as much cut-and-paste as possible.

Suppose $\sqrt[k]{m} =\dfrac{u}{v} $ where $m = \prod_{p \in P} p^{m_i}, u = \prod_{p \in P} p^{u_i}, v = \prod_{p \in P} p^{v_i} $.

Then $\prod_{p \in P} p^{m_i} =\dfrac{u^k}{v^k} =\dfrac{ \prod_{p \in P} p^{ku_i}}{\prod_{p \in P} p^{kv_i}} $ so $\prod_{p \in P} p^{m_i}\prod_{p \in P} p^{kv_i} =\prod_{p \in P} p^{ku_i} $ or $\prod_{p \in P} p^{m_i+kv_i} =\prod_{p \in P} p^{ku_i} $.

By unique factorization, $m_i+kv_i =ku_i $, so $m_i =ku_i-kv_i =k(u_i-v_i) $.

Therefore $m =\prod_{p \in P} p^{m_i} =\prod_{p \in P} p^{k(u_i-v_i)} =\left(\prod_{p \in P} p^{u_i-v_i}\right)^k $ so $m$ is a perfect $k$-th power.

Therefore the $k$-th root of an integer is rational only if it is a $k$-th power.


When you get $2p^2 = 3q^2$ that means that $2|q^2$ so $2|q$ and that $3|p^2$ so $3|p$.

Then if you replace $p = 3p'$ for some integer $p'$ and $q=2q'$ for some integer $q'$ you get

$2(3p')^2 = 3(2q')^2$

$18p'^2 = 12q'^2$

$3p'^3 = 2q'^2$.

Can you finish from there?

This assumes you know Euclid's lemma that 1) prime numbers exist and 2) if $p$ is prime and $p|a*b$ then either $p|a$ or $p|b$ (or both).

  • $\begingroup$ I tried to reach it without prime factorisation, can you have a second look? $\endgroup$ – Papa Nov 15 '18 at 2:42

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.