Let $A$ be an integral domain, $a$ and $b \in A-\{0\}$, and let $B = A[X]/(aX+b)$. Show that, if $Aa \cap Ab=Aab$, then $B$ is an integral domain.

My attempt at proof (following a hint). Denote by $K$ the field of fractions of $A$. Let $\phi: A[X] \to K$, with $\phi(X)=-b/a$ and $\phi(y)=y, y \in A$.

Then $\phi(aX+b)=-b+b=0$. If $p(X) \in A[X]$ and $p(X) \not \in (aX+b)$, then $p(X)=q(X)(aX+b)+r$, $r \in A$. Therefore, $\phi(p(X))=\phi(r)=r$, so $p(X) \in \ker(\phi) \iff r=0 \iff p(X) \in (aX+b)$.

Hence $\ker(\phi)=(aX+b)$; by the first isomorphism theorem, $A[X]/(aX+b)$ must be isomorphic to Im$(\phi)$. In conclusion $B$ is isomorphic to a subfield of $K$, so it's an integral domain.

The problem is, I don't know where I used the condition $Aa \cap Ab=Aab$. Am I doing something wrong?

  • 2
    $\begingroup$ Be careful where you consider your principal ideals. Note that $X^2 + X$ is not in the ideal generated by $2X + 2$ in the ring of integral polynomials. $\endgroup$ – quid Feb 8 '15 at 20:32
  • $\begingroup$ Why should $\operatorname{img} ϕ ⊂ K$ be a subfield? (Of course, the conclusion that it’s integral would still hold.) $\endgroup$ – k.stm Feb 8 '15 at 20:33
  • $\begingroup$ I thought that resulted from the isomorphism theorem, but looking back at it, I'm actually not sure. $\endgroup$ – odnerpmocon Feb 8 '15 at 20:35
  • $\begingroup$ @quid: True, but can you explain how it applies here? $\endgroup$ – odnerpmocon Feb 8 '15 at 20:37
  • $\begingroup$ @odnerpmocon What quid was getting at was that if the ideal is in $K[x]$ instead of $A[x]$ you will come to difference conclusions, since different rings have different workings. Also, the division algorithm only works if you have a Euclidean domain, in the case $A$ is not a field, then no such $q,r$ are guaranteed to exist, which undermines that entire approach! $\endgroup$ – Adam Hughes Feb 8 '15 at 21:41

It suffices to prove $\,f = ax+b\,$ is prime in $\,A[x],\,$ which follows as below.

Theorem $ $ Suppose that $\,D\,$ is a domain, and $\,0\ne a,b \in D\,$ satisfy $\,a,b\mid c\, \color{#c00}\Rightarrow\, ab\mid c\,$ for all $\,c \in D.\,$ Then $\, f = ax+b\,$ is prime in $\,D[x].$

Proof $\ $ Assume $\,f\mid gg'\,$ for $\,g,g'\! \in D[x].\,$ $\,f = a(x + \frac{b}a)= a\bar f$ in $\,\bar D = D[a^{-1}]\,$ where $\,\bar f\,$ is prime, hence $\,\bar f\mid g\,$ or $\,\bar f\mid g'\,;\,$ wlog $\,\bar f\mid g,\,$ so $\, g = \bar f \bar h,\ \bar h \in \bar D[x].\, $ Scaling by $a^n$ for big enough $\,n\,$ yields $\, a^n g = f h,\ h\in D[x],\,$ so $\,f\mid a^n g\,\Rightarrow\,f\mid g,\,$ by iterating below Lemma. Hence $\,f\,$ is prime $\ \ $ QED

Lemma $\,\ f\mid ag\,\Rightarrow\,f\mid(f\!-\!ax)g=bg\,\Rightarrow\,f\mid ag,bg\,\Rightarrow\,a,b\mid abg/f\,\color{#c00}\Rightarrow\,ab\mid abg/f\,\Rightarrow\,f\mid g$

Remark $ $ If localizations like $\,D[a^{-1}]\,\cong\, D[\,t]/(at\!-\!1)\,$ are unfamiliar then we can eliminate it by instead using the nonmonic form of the polynomial division algorithm, namely $\, a^n g = q\, f + r\,$ where $\,a\,$ is the lead coef of $\,f,\,$ which has a simple inductive proof. I presented it here in local form so to show how this Theorem generalizes to localizations. The results there generalize as below. Above we used $\, S^{-1}D = D[a^{-1}],\,$ whereas there we used the $\,S^{-1}D = K = $ full fraction field.

Theorem $\,\ f\,$ is prime in $\,D[x] \!\iff\! f\,$ is prime in $\,S^{-1} D[x]\,$ and $\,f\,$ is $S$-superprimitive.

where $\,f\,$ is $\,S$-superprimitive if it satisfies one of the following equivalent properties

$(1)\quad c\mid gf\,\Rightarrow\, c\mid g\quad$ for all $\ c\in S,\ g\in D[x]$

$(2)\quad c\mid df\,\Rightarrow\, c\mid d\quad$ for all $\ c\in S,\ d\in D$

$(3)\quad f\mid g\,$ in $\,S^{-1} D[x]\,\Rightarrow\, f\mid g\,$ in $\,D[x]\quad $ for all $\, g\in D[x]$

$(4)\quad f\mid cg\,\Rightarrow\,f\mid g\quad$ for all $\ c\in S,\ g\in D[x]$

$(5)\quad gf\in D[x]\,\Rightarrow\, g \in D[x]\quad$ for all $\,g\in S^{-1}D[x]$


You write:

If $p(X) \in A[X]$ and $p(X) \not \in (aX+b)$, then $p(X)=q(X)(aX+b)+r, r \in A$.

Yet how do you actually justify it? One cannot divide by a polynomial in an integral domain in general; one needs the leading coefficient is invertible (or some other assumption).

So, let us work over the quotient field instead. We can define the $\phi$ just as well over $K[x]$ and then we get that the kernel for the map defined on $K[X]$ is the ideal generated by $aX + b$ in $K[X]$.

Now, restrict the map to $A[X]$. Then the kernel is the intersection of $A[X]$ with the ideal generated by $aX+b$ in $K[X]$. In general this is not the same as the ideal generated by $aX+ b$ in $A[X]$ but here you can then use your assumption.

  • 1
    $\begingroup$ Could you explain how I could use the assumption? $\endgroup$ – odnerpmocon Feb 9 '15 at 1:59
  • $\begingroup$ The linked to proposed duplicate question contains various arguments along these lines. $\endgroup$ – quid Feb 9 '15 at 11:25

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.