Let $R$ be a commutative ring, $a_1, ..., a_n$ its elements and

$\phi: R[x_1, ..., x_n] \to R$

defined by

$ \phi(f(x_1, ..., x_n)) = f(a_1, ... ,a_n)$

a ring homomorphism.

Prove: $\ker \phi = (x_1-a_1, ..., x_n-a_n)$

It is obvious that $(x_1 - a_1, ..., x_n -a_n) \subseteq \ker \phi$. I'm not sure how to prove the converse.

At this point I don't know any division algorithms for multivariable polynomials, only for the ones in $R[x]$(and the book from where I taken the exercise doesn't assume the reader to know something beyond basics of rings and ideals, and the division algorithm for $R[x]$). Though I know this could be solved for $i = 1$ by dividing by $x-a$:

Let $f(x) \in \ker \phi$, divide by $x-a: f(x) = q(x)(x-a) + r, f(a) = r = 0$, so $f(x) = q(x)(x-a) \in (x-a)$.

  • $\begingroup$ Hint: $f(x_1,\cdots, x_n)+(x_1-a_1, \cdots, x_n-a_n)=f(a_1,\cdots, a_n)+(x_1-a_1, \cdots, x_n-a_n)$. Also since $(x_1-a_1, \cdots, x_n-a_n)\subset \ker\phi$ we have an induced map $R[x_1, \cdots, x_n]/(x_1-a_1, \cdots, x_n-a_n)\to R$, what is this map? Is it surjective? Was $\phi$ surjective? What can you say about $\ker\phi$ now? $\endgroup$ – Hamed Jul 10 '16 at 1:16
  • $\begingroup$ You say you can do the $n=1$ case, but really every polynomial ring is a polynomial ring in one variable since $R[x_1, \ldots, x_{n-1}, x_n] = (R[x_1, \ldots, x_{n-1}])[x_n]$. I wonder if you can factor your evaluation map into one-variable evaluation maps and extend your proof... $\endgroup$ – André 3000 Jul 10 '16 at 5:16
  • 1
    $\begingroup$ I found math.stackexchange.com/questions/1352606 $\endgroup$ – Alphonse Feb 7 '17 at 13:42

In problems like this, instead of explicitly showing that every element of the kernel is in $I=(x_1-a_1,\dots,x_n-a_n)$, it is easier to think about the quotient ring $S=R[x_1,\dots,x_n]/I$. Since $I\subseteq\ker(\phi)$, $\phi$ induces a surjective homomorphism $\tilde{\phi}:S\to R$ such that $\phi$ is equal to the composition of $\tilde{\phi}$ with the quotient map $\psi:R[x_1,\dots,x_n]\to S$. Let us now show that $\tilde{\phi}$ is injective, so it is actually an isomorphism and $\ker(\phi)=\ker(\psi)=I$.

To prove this, note that in $S$, $\psi(x_i)=\psi(a_i)$ for each $i$. Since $\psi$ is a homomorphism, it follows that given any polynomial $f(x_1,\dots,x_n)$, $\psi(f(x_1,\dots,x_n))=\psi(f(a_1,\dots,a_n))$. This means that every element of $S$ is actually $\psi(r)$ for some $r\in R$. But for $r\in R$, $\tilde{\phi}(\psi(r))=\phi(r)=r$. In particular, $\tilde{\phi}(\psi(r))=0$ iff $r=0$. It follows that the kernel of $\tilde{\phi}$ consists of only $\psi(0)=0$, so $\tilde{\phi}$ is injective.


This proof is pretty notation-heavy for which I apologize, but the idea is not complicated: a polynomial $f$ is usually written as a Taylor series centered at $(0,\ldots, 0)$. We use the Binomial Theorem to re-center it at $(\alpha_1, \ldots, \alpha_n)$ and show that the constant term of this series is $f(\alpha_1, \ldots, \alpha_n)$.

Lemma: Let $R$ be a unital commutative ring and $R[x_1, \ldots, x_n]$ be the $n$-variable polynomial ring over $R$. For $\alpha_1, \ldots, \alpha_n \in R$, let \begin{align*} \varphi = \text{eval}_{\alpha_1, \ldots, \alpha_n}: R[x_1, \ldots, x_n] &\to R\\ x_1, \ldots, x_n &\mapsto \alpha_1, \ldots, \alpha_n \end{align*} be the evaluation homomorphism. Then $\ker(\varphi) = (x_1-\alpha_1, \ldots, x_n-\alpha_n)$ and the induced map $\overline{\varphi}: R[x_1, \ldots, x_n]/(x_1-\alpha_1, \ldots, x_n-\alpha_n) \to R$ is an isomorphism.

Proof: Certainly $(x_1-\alpha_1, \ldots, x_n-\alpha_n) \subseteq \ker(\varphi)$ so it remains to show the reverse inclusion. Given $f \in \ker(\varphi)$, we may write $f(x_1, \ldots, x_n) = \sum_{\gamma} a_\gamma {x_1}^{\gamma_1} \cdots {x_n}^{\gamma_n}$ where $a_\gamma = 0$ for all but finitely many multi-indices $\gamma$. Writing $x_i = (x_i - \alpha_i) + \alpha_i$ and expanding by the binomial theorem, we have \begin{align*} f(x_1, \ldots, x_n) &= \sum_{\gamma} a_\gamma {x_1}^{\gamma_1} \cdots {x_n}^{\gamma_n} = \sum_{\gamma} a_\gamma ((x_1 -\alpha_1) +\alpha_1)^{\gamma_1} \cdots ((x_n -\alpha_n) +\alpha_n)^{\gamma_n}\\ &= \sum_{\gamma} a_\gamma \left(\sum_{\kappa_1=0}^{\gamma_1} \binom{\gamma_1}{\kappa_1} (x_1 -\alpha_1)^{\kappa_1}\alpha_1^{\gamma_1 - \kappa_1}\right) \cdots \left(\sum_{\kappa_n=0}^{\gamma_n} \binom{\gamma_n}{\kappa_n} (x_n -\alpha_n)^{\kappa_n}\alpha_n^{\gamma_n - \kappa_n}\right)\\ &= \sum_{\gamma} a_\gamma \sum_{\kappa_1=0}^{\gamma_1} \cdots \sum_{\kappa_n=0}^{\gamma_n} \binom{\gamma_1}{\kappa_1} \cdots \binom{\gamma_n}{\kappa_n}\alpha_1^{\gamma_1 - \kappa_1} \cdots\alpha_n^{\gamma_n - \kappa_n} (x_1 -\alpha_1)^{\kappa_1} \cdots (x_n -\alpha_n)^{\kappa_n}\\ &= \sum_{\gamma} a_\gamma \sum_{\kappa} \binom{\gamma_1}{\kappa_1} \cdots \binom{\gamma_n}{\kappa_n}\alpha_1^{\gamma_1 - \kappa_1} \cdots\alpha_n^{\gamma_n - \kappa_n} (x_1 -\alpha_1)^{\kappa_1} \cdots (x_n -\alpha_n)^{\kappa_n}\\ &=\sum_{\kappa} \underbrace{\left(\sum_{\gamma}a_\gamma\binom{\gamma_1}{\kappa_1} \cdots \binom{\gamma_n}{\kappa_n}\alpha_1^{\gamma_1 - \kappa_1} \cdots\alpha_n^{\gamma_n - \kappa_n}\right)}_{b_\kappa} (x_1 -\alpha_1)^{\kappa_1} \cdots (x_n -\alpha_n)^{\kappa_n}\\ &=\sum_{\kappa} b_\kappa (x_1 -\alpha_1)^{\kappa_1} \cdots (x_n -\alpha_n)^{\kappa_n}. \end{align*} Note that the constant term $b_{(0,\ldots, 0)}$ is \begin{align*} b_{(0,\ldots, 0)} = \sum_{\gamma} a_\gamma\binom{\gamma_1}{0} \cdots \binom{\gamma_n}{0}\alpha_1^{\gamma_1} \cdots\alpha_n^{\gamma_n} = f(\alpha_1, \ldots,\alpha_n) = \varphi(f) = 0 \end{align*} since $f \in \ker(\varphi)$. Now all other terms in the sum $\sum_{\kappa} b_\kappa (x_1 -\alpha_1)^{\kappa_1} \cdots (x_n -\alpha_n)^{\kappa_n}$ have a factor of $(x_i - \alpha_i)^{\kappa_i}$ for some $i$ with $\kappa_i \geq 1$, hence belong to $(x_1 - \alpha_1, \ldots, x_n - \alpha_n)$. Thus $f(x_1, \ldots, x_n) = \sum_{\kappa} b_\kappa (x_1 -\alpha_1)^{\kappa_1} \cdots (x_n -\alpha_n)^{\kappa_n} \in (x_1 - \alpha_1, \ldots, x_n - \alpha_n)$.


Observe $(x_1-a_1,\dots, x_n-a_n)$ is a maximal ideal of $k[x_1,\dots,x_n]$, since $$k[x_1,\dots,x_n]/(x_1-a_1,\dots, x_n-a_n)\simeq k,$$ and it is contained in the ideal $\ker\phi$.

As constants are mapped to themselves, $\phi \ne 0$, so $\ker\phi\ne k[x_1,\dots,x_n]$, and by definition of a maximal ideal, $$(x_1-a_1,\dots, x_n-a_n)=\ker\phi$$

  • $\begingroup$ OP is asking about a polynomial ring over a ring $R$ not necessarily a field. $\endgroup$ – Hamed Jul 10 '16 at 1:29
  • $\begingroup$ Oh! I misread the question. Thanks. $\endgroup$ – Bernard Jul 10 '16 at 1:32
  • 6
    $\begingroup$ In any case, this seems essentially circular--showing that $k[x_1,\dots,x_n]/(x_1-a_1,\dots, x_n-a_n)\simeq k$ is basically what this problem is asking you to do. $\endgroup$ – Eric Wofsey Jul 10 '16 at 1:35
  • $\begingroup$ No, this is trivial, and is true over any ring. $\endgroup$ – Bernard Jul 10 '16 at 1:40
  • $\begingroup$ @Bernard But this is essentially a rephrasing of the question to some extent. Even if it does seem trivial to you, perhaps you could expand on it to help the OP. $\endgroup$ – John Don Feb 13 at 3:03

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.