there already exists a proof for this theorem: http://www.proofwiki.org/wiki/Common_Divisor_Divides_GCD
This one, however, uses Bêzout's Identity. I'm not allowed to use this for the proof.
So, I had two ideas:
(1) Use prime factorization (and here, I believe, I am not sure to use this, since it'll appear in later lesson - however, let's try this one as well!)
So let $a, b \in \mathbb{Z}$ non-negative. Let $a=\displaystyle\prod_{i}{p_i}^{\alpha_i}$ and $b=\displaystyle\prod_i{p_i}^{\beta_i}$, then $\gcd(a,b)$ will contain the exponents $\min(\alpha_i,\beta_i)$.
And what next?
(2) Let now $a,b \in \mathbb{Z}$ non-negative with $d = gcd(a,b)$. If now c is a common divisor of both a and b, I'll get:
$a = c^n \cdot p, b = c^n \cdot q$ for $p,q \in \mathbb{Z}, n \in \mathbb{N}$.
If $c$ or $c^n$ is the GCD, then it divides itself, which seems clear to me. If neither $c$ nor $c^n$ is the GCD, then.. ? Yes, what then?
I'll appreciate any comments ;-)
(1) Next, consider any common divisor $c$, and make a note about how the exponents in its prime factorization compare to min$(\alpha_i, \beta_i)$.
(2) From here, consider what would happen if the greatest common divisor could not be written in the form $c^nr$. You can arrive at a contradiction by using the definition of gcd.