UFD and field of fractions

Let $A$ be a UFD, $K$ its field of fractions, and $f$ an element of $A[T]$ a monic polynomial.

I'm trying to prove that if $f$ has a root $\alpha \in K$, then in fact $\alpha \in A$.

I'm trying to exploit the fact of something about irreducibility, will it help? I havent done anything with splitting fields, but this is something i can look for.

-
If you are familiar with The Rational Root Theorem, I should think you could recast it in your more general setting. –  Gerry Myerson May 27 '12 at 23:55

The proof follows exactly like the proof of the Rational Root Theorem.

Let $\alpha\in K$ be a root. We can express $\alpha$ as $\frac{a}{b}$ with $a,b\in A$, and using unique factorization we may assume that no irreducible of $A$ divides both $a$ and $b$.

If $f(x) = x^n + c_{n-1}x^{n-1}+\cdots+c_0$, then plugging in $\alpha$ and multiplying through by $b^n$ we obtain $$a^n + c_{n-1}ba^{n-1}+\cdots + c_0b^n = 0.$$ Now, $c_{n-1}ba^{n-1}+\cdots+c_0b^n$ is divisible by $b$, hence $a^n$ is divisible by $b$. Since no irreducible of $A$ divides both $a$ and $b$, it follows that $b$ must be a unit by unique factorization. Hence $\alpha\in A$.

-

You can find a proof of your theorem, here

EDIT To people who down voted, would you please kindly tell me the reason so that I could improve my answer?

-
It is usually a good idea to at least summarise the contents of the link you are pointing to—or to post this type of answers in the form of comments. –  Mariano Suárez-Alvarez May 28 '12 at 3:16
It's just a proof of the theorem the OP is asking. I have no idea what else he needs. By the way, the proof is exactly the same as Arturo's which was posted a little later than mine. –  Makoto Kato May 28 '12 at 8:56
@Makoto, you summarized the job well and did everything I could ask for –  James R. May 28 '12 at 20:43
@Eric Many questions posed here are easily answered by simple searches, which quickly locate answers on Wikipedia, etc. Such questioners probably did not perform any web searches. So, generally, one can make no assumptions about whether or not providing such links is redundant. I am surprised that anyone might find such links impolite. I am also surprised by the 3 downvotes - which are rather harsh in my opinion. –  Bill Dubuque May 30 '12 at 0:28
I agree with Bill. Makoto took the time to answer the question correctly. One is entitled to think that he should have done this or that to give a better answer, but there is absolutely no reason to downvote him. –  Georges Elencwajg May 30 '12 at 6:44

The well-known proof of the Rational Root Test immediately generalizes to any UFD or GCD domain. The sought result is simply a monic special case. One can present the proof in a form that works for both gcds and cancellable ideals by using only universal laws common to gcds and ideals, e.g. commutative, associative, distributive laws. Below I give such a universal proof for degree $3$ (to avoid notational obfuscation). It is clear from this how it generalizes to any degree.

Let $\rm\:D\:$ be a domain and suppose monic $\rm\:f(x)\in D[x]\:$ hase root $\rm\:a/b,\ a,b\in D.\:$ The notation $\rm\:(d,e,\ldots)\:$ denotes $\rm\:gcd(d,e,\ldots)\:$ (or the ideal $\rm\:dD + eD +\:\!\ldots\:$ in the ideal case).

$\rm\qquad f(x) = c_0 + c_1 x + c_2 x^2 + x^3,\:$ and $\rm\:b^3\:\! f(a/b) = 0\:$ yields

$\rm\qquad c_0 b^3 + c_1 a b^2 + c_2 a^2 b = -a^3\$

$\rm\qquad\qquad\ \ \Rightarrow\ (b^3, a b^2, a^2 b)\ |\ a^3,\$ since the gcd divides the LHS of above so also the RHS

$\rm\qquad\ (b,a)^3 = \, (b^3, a b^2, a^2 b,\, a^3),\$ hence, by the above divisibility

$\rm\qquad\qquad\quad\:\! =\, (b^3, a b^2, a^2 b)$

$\rm\qquad\qquad\quad =\, b\, (b,a)^2,\$ so cancelling $\rm\:(b,a)^2\:$ yields

$\rm\qquad\ \, (b,a) =\, b\:\Rightarrow\: b\:|\:a,\$ i.e. $\rm\: a/b \in D.\ \$ QED

The degree $\rm\:n> 1\:$ case has the same form: one cancels $\rm\:(b,a)\:$ from $\rm\:(b,a)^n = b\,(b,a)^{n-1}.\:$

The ideal analog is the same, except replace "divides" by "contains", and assume that $\rm\:(a,b)\:$ is invertible (so cancellable), e.g. in any Dedekind domain. Thus the above yields a uniform proof that PIDs, UFDs, GCD and Dedekind domains satisfy said monic case of the Rational Root Test, i.e. that they are integrally closed.

The proof is more concise if one knows about fractional gcds and ideals. Now, with $\rm\:r = a/b,\:$ one simply cancels $\rm\:(r,1)\:$ from $\rm\:(r,1)^n = (r,1)^{n-1}\:$ so $\rm\:(r,1) = (1),\:$ i.e. $\rm\:r \in D.\:$ For more, see my posts in this 2009/5/22 sci.math thread, which includes discussion of how most elementary irrationality proofs are simply unwindings of the elegant one-line proof employing Dedekind's notion of conductor ideal (universal denominator ideal).

-