0
$\begingroup$

Let $h(t,x)$ and $f ( t, x)$ be nonzero elements of $\mathbb C[t,x]$. Suppose that $h$ is not divisible by any polynomial of the form $t-\alpha$. If $h$ divides $f$ in $\mathcal F[x]$, then $h$ divides $f$ in $\mathbb C[t,x]$. This proposition can be found in the book "Algebra" by Artin.

The proof seemed to be easy but when I tried to embed the intuition into logic, I got stuck.

Proof:

we have, $f = h q$ , where $q \in \mathcal F[x]$, it is a polynomial in $x$ whose coefficients are rational functions in $t$. We multiply both sides of the equation $f = h q$ by a monic polynomial in $t$ to clear denominators in these coefficients. This gives us an equation of the form $u ( t) f ( t, x) = h (t, x) q_1(t, x )$.

Let $\alpha $ be a root of $u(t)$, then $t-\alpha$ divedes RHS as well, $t — \alpha$ does not divide $h$, so $h(\alpha,x)$ is not zero. Since the polynomial ring $\mathbb C[x]$ is an integral domain, $q_1 (\alpha, x) = 0$, and the lemma shows that $t-\alpha$ divides $q_1(t, x)$. We cancel $t-\alpha$ from $u$ and $q_1$.

Problem: (All the implications below are mine, so can thus be wrong)

We have $q \in \mathcal F[x]$ where $\mathcal F = \mathbb C (t)=$ a rational field of $\mathbb C [t]$.

So, $q=\frac{a_n(t)}{b_n(t)}x^n+\frac{a_{n-1}(t)}{b_{n-1}(t)}x^{n-1}+\frac{a_{1}(t)}{b_{1}(t)}x+\frac{a_{0}(t)}{b_{0}(t)}$

where $ q_1 \in \mathbb{C}[t,x]$. Let $q_1=u(t)*q=c_{n}(t)x^n+c_{n-1}(t)x^{n-1}+\cdots+c_{1}(t)x+c_{0}(t)$, where, $u(t)=lcm(b_{n}(t),b_{n-1}(t),\cdots, b_{0}(t))$ and $c_j(t)=\frac{u(t)}{b_j(t)}a_{j}(t)$.

$(t-\alpha)|q_1 \Rightarrow (t-\alpha)| c_{j}(t) \; \forall \; j={1,\cdots,n} $. let $(t-\alpha)$ has multiplicity $k$ in $u(t)$ which is due to some $b_{p}(t)$. Now, $c_{p}(t)$ will not have any power of $(t-\alpha)$ which would mean $(t-\alpha)$ does not divide $c_j(t) \; \forall j$, which is a contradiction to what I have assumed initially. What am I doing wrong here?

$\endgroup$
7
  • $\begingroup$ What you write is correct, but what is the question/problem? $\endgroup$ Feb 15, 2017 at 12:14
  • $\begingroup$ $q_1(\alpha , x)=0 \Rightarrow (t-\alpha)$ is a factor of $q_1$ but what I wrote in Problem section, suggests that there will be term in $q_1$ which doesn't get divided by $(t-\alpha)$ and hence, $q_1(\alpha , x) \neq 0$. $\endgroup$
    – MUH
    Feb 15, 2017 at 12:21
  • 1
    $\begingroup$ Yes, and so what's the problem? We get a contradiction and celebrate. $\endgroup$ Feb 15, 2017 at 12:26
  • $\begingroup$ In the proof section, it is proven that $q_1(\alpha ,x)=0$ but in the problem section, I am concluding that $q_1(\alpha,x) \neq 0$. So, there must be some fallacy or gaps in my argument. I just want to know, where the fallacy is? $\endgroup$
    – MUH
    Feb 15, 2017 at 12:48
  • 1
    $\begingroup$ You really should clear up the meaning of your notation: your $\mathcal F$ is what the rest of the world would write as $\mathbf C(t)$. Anyway, this is Gauss' lemma. Find a proof of it first in $\mathbf Z[x]$ before trying to understand it in $\mathbf C[t,x] = \mathbf C[t][x]$. Ultimately it is about polynomials in $x$ with coefficients in a PID, but learn the most basic case like $\mathbf Z$ first. $\endgroup$
    – KConrad
    Feb 15, 2017 at 12:59

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.