This is the continuation of another question I did some days ago. Here.

I have been working on it and I would like to know if my try to prove it is right or not. I would appreciate a lot any feedback I can get.

I want to prove that $\texttt{Nil}_n$ is represented by $(\mathbb{Z}[x]/(x^n), \tau_R)$. Meaning that there exists a family of functions for every ring $R$, $\tau_R: h^{\mathbb{Z}[x]/(x^n)}(R) \longrightarrow \texttt{Nil}_n(R): f \mapsto f(\overline{x})$ that is an isomorphism (for each $R$)(bijective application, since it is a function between sets). And $\tau_R$ is natural, i.e. the following map commutes for every map $g:A\longrightarrow B$ of rings:

$$\require{AMScd}\begin{CD}h^{\mathbb{Z}[x]/(x^n)}(A) @>g \circ - >> h^{\mathbb{Z}[x]/(x^n)}(B) \\ @V\tau_AVV @V\tau_BVV\\\texttt{Nil}_n(A) @>>\texttt{Nil}_n(g)> \texttt{Nil}_n(B) \end{CD}$$



This means that we have to check $\tau_S \circ h^{\mathbb{Z}[x]/(x^n)}(f)= \texttt{Nil}_n (f) \circ \tau_R$.

Let $h\in h^{\mathbb{Z}[x]/(x^n)}$ be an arbitrary element.

On the one hand, we have $(\tau_S \circ h^{\mathbb{Z}[x]/(x^n)}(f))(h)= \tau_S(h^{\mathbb{Z}[x]/(x^n)}(f)(h)) = \tau_S (f\circ h) = (f\circ h)(\overline{x})\in \texttt{Nil}_n(S)$.

On the other hand, we have $(\texttt{Nil}_n (f) \circ \tau_R)(h)=\texttt{Nil}_n (f)(\tau_R(h))= \texttt{Nil}_n (f)(h(\overline{x}))= f(h(\overline{x}))= (f\circ h)(\overline{x}) \in \texttt{Nil}_n(S)$.

Hence, $\tau_S \circ h^{\mathbb{Z}[x]/(x^n)}(f)= \texttt{Nil}_n (f) \circ \tau_R$.


Let $a,a'\in h^{\mathbb{Z}[x]/(x^n)}$ be two different elements.

$a: h^{\mathbb{Z}[x]/(x^n)} \longrightarrow R: \overline{x} \mapsto a(\overline{x})$

$a': h^{\mathbb{Z}[x]/(x^n)} \longrightarrow R: \overline{x} \mapsto a'(\overline{x})$

Since $\texttt{Nil}_n(R) \subseteq R$, $\tau_R(a)$ or $\tau_R(a')$ is nothing else than the restriction of $a,a'$ to the subset $\texttt{Nil}_n(R)$. Hence, $\tau_R(a)\neq \tau_R(a')$ when $a\neq a'$.


Let $b\in \texttt{Nil}_n(R)$. Since it is a nilpotent element, we know that $b^n=0$. We know that there exists an homomorphism in $h^{\mathbb{Z}[x]/(x^n)}$, namely $h$, such that $h(0)=0$. That means $b^n=0=h(0)$. But since $\overline{0}=\overline{x^n}$, $b^n=0=h(0)=h(\overline{x^n})=h(\overline{x})^n$.

Hence $b=h(\overline{x})$.

Are Naturality, Injectivity and Surjectivity well-proven?

Thank you.


Naturality looks good to me.

Injectivity and surjectivity both look problematic to me though.

Again, let $A=\Bbb{Z}[x]/(x^n)$ for notational brevity.


You say $\tau_R(a)$ is the restriction of $a$ to $\newcommand\Nil{\operatorname{Nil}}\Nil_n(R)$, which isn't right. $\tau_R(a)=a(\bar{x})$. Thus you need to prove that if $a\ne a'$, then $a(\bar{x})\ne a'(\bar{x})$. However, this is immediate, since $a$ and $a'$ are maps from $A$ to $R$, which is generated over $\Bbb{Z}$ by $\bar{x}$, so if $a(\bar{x})=a'(\bar{x})$, then $a(p(\bar{x}))=a'(p(\bar{x}))$ for any polynomial $p\in \Bbb{Z}[x]$, and any element of $A$ can be expressed as $p(\bar{x})$ for some polynomial, so we must have $a=a'$.

Note I am being much more explicit above than I usually would be, since I'm trying to explain it. It should suffice to write:

If $\tau_R(a)=\tau_R(a')$, then $a(\bar{x})=a'(\bar{x})$, so $a=a'$, since $A$ is generated over $\Bbb{Z}$ by $\bar{x}$.


Your argument for surjectivity doesn't make sense. You're saying there exists $h\in h^A$ such that $h(0)=0$, and then concluding that therefore $h(\bar{x})=b$. That doesn't make any sense, since every element $h\in h^A$ satisfies $h(0)=0$, but we just showed in the injectivity section that if $h(\bar{x})=h'(\bar{x})$, then $h=h'$, so they can't possibly all have $h(\bar{x})=b$.

Instead, do the following.

Define $\phi : \Bbb{Z}[x]\to R$ by $x\mapsto b$. Since $b^n=0$, $x^n$ is in the kernel of $\phi$, so $\phi$ induces a map $\tilde{\phi}:A\to R$ with $\tilde{\phi}(\bar{x})=b$. $\tilde{\phi}$ is thus the desired map.

  • $\begingroup$ Here, in surjectivity, why do you know $b^n=0$ if we only know that $b\in R$? Should $R$ be $Nil_n(R)$? @jgon $\endgroup$ – idriskameni Jan 2 at 15:01
  • 1
    $\begingroup$ @idriskameni, yes $b$ should be in $\Nil_n(R)$, sorry for not being more explicit there. $\endgroup$ – jgon Jan 2 at 15:32

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.