# Is it possible to express $\sqrt{2}$ as rational combination of $1, \sqrt{2}, \sqrt{2}$?

Is it possible to express $\sqrt{2}$ as rational combination of $1, \sqrt{2}, \sqrt{2}$?

So can we find $a,b,c \in \mathbb{Q}$ such that $\sqrt{2}=a+ b\sqrt{2}+c \sqrt{2}$?

My guess is no and one should try to prove by contradiction. And the contradiction should show that either at least one number from ${a, b, c}$ is irrational or at least one number from ${\sqrt{2}, \sqrt{2}}$ is rational.

However, my attempts failed because they lead to some unmanageable (for me at least) long computations.

Can you please prove some smart way to handle this?

Thanks a lot!

• This is question about field extensions of the rational numbers. Are you familiar with this subject? – Hans Engler Oct 29 '17 at 16:35
• Yes, but I need elementary proof for that or at least scheme of the proof. – Hedgehog Oct 29 '17 at 16:36
• If you could, then you could double te cube with an unmarked sraightedge and compasses. Uhhh... no. – Oscar Lanzi Oct 29 '17 at 16:40
• @OscarLanzi I hope this problem is simpler since we are much more limited compared with sraightedge and compasses constructions. – Hedgehog Oct 29 '17 at 16:46
• Raise everything to the sixth power; then the right hand side is a ${\mathbb Q}$-linear combination of the powers of $\sqrt{2}$. – franz lemmermeyer Oct 29 '17 at 16:58

By Eisenstein's Criterion, you know that $x^6-2$ is irreducible over $\mathbb{Q}$. Observe that $\sqrt{2}$ is a root of this polynomial.
Suppose (for contradiction) that $\sqrt{2}=a+b\sqrt{2}+c\sqrt{2}$ for some rational $a$, $b$, and $c$. Let $y=a+b\sqrt{2}+c\sqrt{2}$. If you look at $y^0,\cdots,y^4$ as vectors in the span of $1$, $\sqrt{2}$, $\sqrt{2}$, and $\sqrt{8}$, then this is a collection of $5$ vectors in a vector space of dimension at most $4$, so there is some nontrivial combination of these which equals zero. This tells you that there is a nonzero polynomial $p$ of degree at most $4$ with rational coefficients with $y$ as a root (since $y^0,\dots,y^4$ is a linear dependent set).
Therefore, $\sqrt{2}$ is a root of both $x^6-2$ and $p$, so it is a root of their gcd (which cannot be $1$ because the gcd needs a root). This, however, is impossible because such a gcd would need to be of degree at most $4$ and at least $1$ and divide $x^6-2$, but $x^6-2$ is irreducible. This is a contradiction.
Hint: By Eisenstein's Criterion, $x^{12}-2$ is irreducible over $\mathbb{Q}$. Therefore, there are no $a,b,c,d\in\mathbb{Z}$ so that $ax^6+bx^3+cx^2+d=0$ where $x=2^{1/12}$.