5
$\begingroup$

Let $R$ be a ring (not commutative in general) with identity and let $\Psi(x) = x^m-\sum_{j=0}^{m-1}\psi_jx^j$ be a monic polynomial over $R$. I want to construct a ring extension $K$ of $R$, which contains a root of $\Psi(x)$. One construction follows from non-commutative Hamilton-Caley Theorem.

But I also think that there is another construction, which seems to be very interesting.

Let $M = R[x]/R[x]\Psi(x)$ be a left module over $R[x]$ and $A = \mathrm{Ann}_{R[x]}M\subset R[x]$ be an annihilator of $M$. I want to prove that $\Psi(x)$ has a root in the ring $$ R[x]/A. $$

The proof is very simple in the case, where $R$ is a commutative ring. So the main interest is when $R$ is not commutative.

$\endgroup$
5
  • $\begingroup$ So you want to show that $\Psi(x + A) = \Psi(x) + A$. Isn't $x$ central in $R[x]$? $\endgroup$
    – Luc Guyot
    Aug 20, 2017 at 11:42
  • $\begingroup$ Yes, x is central in R[x]. $\endgroup$ Aug 20, 2017 at 11:45
  • $\begingroup$ We certainly have $\Psi(x + A) + A = \Psi(x) + A$, but $\Psi(x) \in A$ holds if and only if $\Psi(x) R[x] \subset R[x] \Psi(x)$. So I would expect a negative answer. $\endgroup$
    – Luc Guyot
    Aug 20, 2017 at 12:58
  • 1
    $\begingroup$ If $a$ and $b$ are two non-commuting elements of $R$ and $\Psi(x) = x - a$ then $\Psi(x)(x - b) \in \Psi(x) R[x] \setminus R[x] \Psi(x)$. So $\Psi(x) \notin A$. Therefore $x + A$ is not a root of $\Psi$ in $R[x]/A$. But you are asking about any root in $R[x]/A$? $\endgroup$
    – Luc Guyot
    Aug 20, 2017 at 13:21
  • $\begingroup$ @LucGuyot, yes, x+A is not a root of $\Psi(x)$ in general case (but in commutative is). Moreover, my experiments show that the root can be choosen in the form $x+H(x)\Psi(x)+A$ for some polynomyal $H(x)$. $\endgroup$ Aug 20, 2017 at 17:17

1 Answer 1

3
$\begingroup$

No, $R[x]/A$ does not always have a root.

Let $R=\mathbb{Z}\langle a,b\rangle$ be the free ring on two noncommuting elements $a$ and $b$. Let $\Psi=x^2+a\in R[x]$, and let $M$ and $A$ be the objects you defined.

In this case, $A=0$. Suppose $f=r_nx^n+\cdots+r_0\in A$. In particular, there is some $g=s_n x^n+\cdots+s_0\in R[x]$ such that $f\cdot b=g\cdot (x^2+a)$. Matching up coefficients, we find that $r_0b=s_0a$, $r_1b=s_1a$, $r_2b=s_2a+s_0$,$\dots$, $r_ib=s_ia+s_{i-2}$. In the free ring $R$ the first two equations force $r_0=s_0=0$ and $r_1=s_1=0$, and by induction the rest of the equations force $f=g=0$. Hence $A=0$.

Finally, the ring $R[x]/A\simeq R[x]$ doesn't have any roots of $\Psi$.

$\endgroup$
0

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.