6
$\begingroup$

For a prime power $q$, consider polynomials $f_1,f_2 \in \mathbb{F}_q[x]$. Then, how do we check whether there exists an algebra isomorphism between:

$$\frac{\mathbb{F}_q[x]}{\langle f_1 \rangle} \text{and} \frac{\mathbb{F}_q[x]}{\langle f_2 \rangle}?$$

This probably has something to do with factorising $f_1$ and $f_2$, and using Chinese Remainder Theorem, but I don't see how exactly.

$\endgroup$
4
  • 2
    $\begingroup$ Do you know that if $f_1$ and $f_2$ are irreducible, the quotients are isomorphic if and only if $f_1$ and $f_2$ have the same degree? $\endgroup$ Apr 26, 2017 at 6:31
  • $\begingroup$ If the polynomials are squarefree, the problem is very easy. If they are not, the question is whether $\mathbb F_q[x]/(f^n)$ and $\mathbb F_q[x]/(g^n)$ are isomorphic if $f,g$ are irreducible of the same degree. $\endgroup$
    – MooS
    Apr 26, 2017 at 6:32
  • $\begingroup$ Yes, @MooS I realised that. $\endgroup$
    – MathManiac
    Apr 26, 2017 at 7:02
  • $\begingroup$ Possible duplicate of math.stackexchange.com/questions/2235563/…. $\endgroup$
    – lhf
    Apr 26, 2017 at 11:40

1 Answer 1

4
$\begingroup$

Here is a proof for the following proposition:

If $q$ is any prime power and $f,g \in \mathbb F_q[x]$ irreducible of the same degree $d$, then there is an isomorphism of $\mathbb F_q$-algebras $\mathbb F_q[x]/(f^n) \cong \mathbb F_q[x]/(g^n)$.

This proposition answers the question, using the Chinese remainder theorem.

First let me proof the following easy lemma:

If $l$ is a prime power and $A \supset \mathbb F_l$ an algebra with $l^2$ elements and an element $0 \neq a \in A$ with $a^2=0$, then $A$ is isomorphic to $\mathbb F_l[\varepsilon]/(\varepsilon^2)$.

Proof. We clearly get a morphism $\mathbb F_l[\varepsilon]/(\varepsilon^2) \to A$ by sending $\varepsilon \mapsto a$, which is injective because $\varepsilon$ is not contained in the kernel. Comparing the number of elements shows that it is an isomorphism.

Now we can proof the proposition above:

We know the statement is true for $n=1$, i.e. we have an isomorphism $$\mathbb F_q[x]/(f) \to \mathbb F_q[x]/(g).$$ Choosing such an isomorphism is nothing else but choosing a polynomial $h \in \mathbb F_q[x]$ with $f(h(x)) = 0 \pmod{g(x)}$.

If $f(h(x)) \neq 0 \pmod{g(x)^2}$ we are done, because in this case we have that $f(h(x))^n \neq 0 \pmod{g(x)^n}$ and $f(h(x))^{n-1} \neq 0 \pmod{g(x)^n}$, thus we get a well-defined injective morphism $$\mathbb F_q[x]/(f^n) \to \mathbb F_q[x]/(g^n)$$ by sending $x \mapsto h(x)$, which is again an isomorphism by comparing number of elements.

If $f(h(x)) = 0 \pmod{g(x)^2}$, we get a well-defined morphism $\mathbb F_q[x]/(f) \to \mathbb F_q[x]/(g^2)$, i.e. $\mathbb F_q[x]/(g^2)$ is an algebra over $\mathbb F_{q^d}$ which clearly satisfies the assumptions of the lemma above, i.e. it is isomorphic to $\mathbb F_{q^d}[\varepsilon]/(\varepsilon^2)$.

So - for $n=2$ - we have shown the following:

If $f(h(x)) \neq 0 \pmod{g(x)^2}$, then we have the desired isomorphism. If $f(h(x)) = 0 \pmod{g(x)^2}$, then $\mathbb F_q[x]/(g^2)$ is isomorphic to $\mathbb F_{q^d}[\varepsilon]/(\varepsilon^2)$

Of course we can switch roles of $f$ and $g$ and get an $h'$ with:

If $g(h'(x)) \neq 0 \pmod{f(x)^2}$, then we have the desired isomorphism. If $g(h'(x)) = 0 \pmod{f(x)^2}$, then $\mathbb F_q[x]/(f^2)$ is isomorphic to $\mathbb F_{q^d}[\varepsilon]/(\varepsilon^2)$

Summarizing, we have shown the result for $n=2$, i.e. we have an isomorphism $$\mathbb F_q[x]/(f^2) \to \mathbb F_q[x]/(g^2).$$ Now we repeat our argument: Choosing such an isomorphism is nothing else but choosing a polynomial $h \in \mathbb F_q[x]$ with

(i) $f(h(x))^2 = 0 \pmod{g(x)^2}$, but

(ii)$f(h(x)) \neq 0 \pmod{g(x)^2}$.

The first property yields (iii) $f(h(x)) = 0 \pmod{g(x)}$. Like already explained (ii),(iii) yield an isomorphism $$\mathbb F_q[x]/(f^n) \to \mathbb F_q[x]/(g^n).$$


Edit:

For the original question, I claimed a condition in the comments, which is clearly sufficient. Here is an easy argument that it is necessary:

W.l.o.g let $n_1$ be the largest among all multiplicities $n_i,m_j$. Then $\mathbb F_q[x]/(f)$ has an nilpotent element of nilpotence order $n_1$, namely $f_1$. So the ring has $\mathbb F_q[x]/(g)$ has also such an element, say $\phi(f_1)$. This shows some $m_j$ must be at least $n_1$. By the maximality of $n_1$, we get that some $m_j$ is equal to $n_1$, w.l.o.g it is $m_1$. After dividing out $f_1$ in $\mathbb F_q[x]/(f)$ and its image in $\mathbb F_q[x]/(g)$, we get an isomorphism $\mathbb F_q[x]/(f_1) \cong \mathbb F_q[x]/(\phi(f_1))$, thus we deduce that $\deg f_1 = \deg \phi(f_1)$ and $\phi(f_1)$ is irreducible. This yields $\varphi(f_1)=g_1$. So we have shown that the multiplicity and the degree of one of the irreducible factors coincides. The rest is an induction argument.

$\endgroup$
8
  • $\begingroup$ A pretty elegant argument. But, I think we still have some work to do to solve the original problem. For example, suppose $f=a^2$ and $g=bc$, where $a,b,c$ are irreducible polynomials of equal degree. Then, can we claim something about the quotient rings? If isomorphic, why? If not, why not? $\endgroup$
    – MathManiac
    Apr 26, 2017 at 9:51
  • $\begingroup$ Of course not isomorphic. Think about nilpotents. $\endgroup$
    – MooS
    Apr 26, 2017 at 10:02
  • $\begingroup$ Yes. So, coming back to the original question, what exactly should be the conditions for isomorphism? $\endgroup$
    – MathManiac
    Apr 26, 2017 at 10:04
  • 1
    $\begingroup$ If $f=f_1^{n_1} \dotsb f_s^{n_s}$ and $g=g_1^{m_1}\dotsb f_t^{m_t}$ where the irreducible factors are ordered by degree (and if the degree is the same, we order them by the multiplicity), we must have $\deg f_i=\deg g_i$ and $n_i=m_i$ for any $i$. In particular $s=t$. $\endgroup$
    – MooS
    Apr 26, 2017 at 10:07
  • 1
    $\begingroup$ See my edit, I gave an argument for the converse. $\endgroup$
    – MooS
    Apr 26, 2017 at 10:56

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.