The proof I'm familiar with that the algebraic numbers $\mathbb A$ form a field uses the fact that the resultant of two polynomials $p,q\in\mathbb Q[x]$ satisfies the following properties:

  • It is $0$ iff $p$ and $q$ have a common factor.
  • It is a polynomial in the coefficients of $p$ and $q$.

We then introduce a new variable and cleverly manipulate $p$ and $q$ to get polynomials which vanish at the sums and products of their roots. This is in some ways a nice proof, e.g. it is constructive and so can be converted into an algorithm to find such polynomials (which I in fact just finished doing in C). But I don't find it very enlightening; it seems like the fact that $\mathbb A$ is a field is simply an accident. Is there a more enlightening proof of this fact?


If $a,b$ are algebraic then $[F(a,b):F]$ is finite so $a+b, ab \in F(a,b)$ are algebraic too. I find this proof to feel the most enlightening.

If you specifically wanted to see why the constructive proof is not merely an accident, then see this link- after proving that the sum of algebraic numbers is algebraic in the usual way, they add:

Now let us analyse the above argument a little. We see that what made it work was that the vector space of all rational combinations of powers of x+y was finite-dimensional. We then observe that we deduced this from the fact that the corresponding vector spaces for powers of x and powers of y were also finite-dimensional - and also that we could build a spanning set for the (x+y)-space out of the spanning sets for the x-space and the y-space.

  • $\begingroup$ Ah, I see. I was missing the fact that $[F(a,b):F(a)]$ is finite hence $[F(a,b):F]$ is. I'm not sure why that didn't occur to me. $\endgroup$ – Alex Becker Mar 15 '13 at 8:06
  • 1
    $\begingroup$ "we could build a spanning set for the (x+y)-space out of the spanning sets for the x-space and the y-space." How? $\endgroup$ – Martin Brandenburg Mar 16 '13 at 20:14
  • 5
    $\begingroup$ In my opinion this standard proof is not constructive at all. It does not help us to find explicitly an algebraic equation for $a+b$ when we have found one for $a$ and $b$. It only shows the existence. $\endgroup$ – Martin Brandenburg Mar 16 '13 at 20:17

My favourite proof of this goes through matrices. A complex number is an eigenvalue of a square matrix of rational numbers if and only if it is algebraic (e.g. any monic polynomial has a companion matrix of which it is the characteristic polynomial).
If $A$ is an invertible matrix with eigenvalue $\alpha$, then $A^{-1}$ has eigenvalue $1/\alpha$. If $A$ and $B$ are square matrices with rational entries and eigenvalues $\alpha$ and $\beta$ for eigenvectors $u$ and $v$ respectively, then $A \otimes B$ and $A \otimes I + I \otimes B$ have eigenvalues $\alpha \beta$ and $\alpha + \beta$ for eigenvector $u \otimes v$.

This also shows, BTW, that the algebraic integers are closed under addition and multiplication: they are the eigenvalues of matrices with integer entries.

  • 13
    $\begingroup$ +1. This proof is constructive and also works for ring extensions, it shows that the set of integral elements is a subring. $\endgroup$ – Martin Brandenburg Mar 16 '13 at 20:15
  • 6
    $\begingroup$ Beautiful proof. $\endgroup$ – Prism Aug 31 '14 at 5:49
  • 2
    $\begingroup$ Robert Israel, can you explain in more detail on how to calculate companion matrix of Kronecker symbol like $A\otimes B, A\otimes I+I\otimes B$, and post it on question math.stackexchange.com/questions/1277753/…? Thanks. $\endgroup$ – hermes May 15 '15 at 5:22

I find another charming proof in Lecture on the theory of algebraic numbers by Hecke. It is constructive in appearance, but is impractical in fact.
Let $p(x)$ and $q(x)$ be two polynomials with rational coefficients, and with roots $\alpha$ and $\beta$ respectively. Let $\{\alpha_i|i=0,\ldots,m\}$ and $\{\beta_j|j=0,\ldots,n\}$ be their conjugates, i.e. roots of $p$ and $q$ respectively. Then set $$r(x)=\prod_{i=0}^m\prod_{j=0}^n(x-\alpha_i-\beta_j).$$ We kow that $\alpha+\beta$ is a root of $r(x)$, and that coefficients of $r(x)$ are symmetric functions of roots of $p$ and $q$, and hence could be written as polynomials of coefficients of the two polynomials, thus is rational. Therefore $\alpha+\beta$ is also algebraic. A similar proof goes for products. Finally, write $$p(x)=\sum_0^ma_ix^i.$$Then $$\sum_0^ma_i\alpha^i=0,$$ so that $$\sum_0^ma_{m-i}\alpha^{-i}=0,$$ and hence $\alpha^{-1}$ also is algebraic.
Inform me of any ambiguity or error, thanks.

  • 1
    $\begingroup$ If we replace $\mathbb{Q}$ by an arbitrary field, this proof only works for separable $\alpha,\beta$, right? Or can it be modified? $\endgroup$ – Martin Brandenburg Mar 16 '13 at 20:13
  • $\begingroup$ @MartinBrandenburg I cannot see the reason it only works for separable numbers: it does not use Galois theory at all; rather, it makes use of the fundamental theorem of symmetric functions. Per chance this works only for separable extensions? In any case, thanks for your attention. $\endgroup$ – awllower Mar 17 '13 at 5:35
  • $\begingroup$ This is the "other" approach to finding a polynomial for $\alpha+\beta$, that of using resultants. Your $r(x)$ is essentially Res$_y(p(y),$Res$_z(q(z),x-y-z))$. $\endgroup$ – Jose Brox Nov 4 '20 at 16:28

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.