Let $K$ be any field. Is it true that:

  • If $f_1,\dots, f_k \in K[x_1,\dots,x_n]$ are polynomials of degree $1$, then $I=(f_1,\dots, f_k)$ is all $K[x_1,\dots,x_n]$, or, if it is a proper ideal, then is it prime?

  • In particular, is it true that $K[x_1,\dots,x_n]/I$ is isomorphic to a polynomial ring (in the case that $I$ is proper and hence prime by the statement above)?

If they are true, how can i prove these statements in an elementary but rigorous way?

P.S.: Birth of my question: I found on my book that "affine varieties given by linear equations are irreducibile", so I translated (I don't know if I have made it right) this geometric statement into the analogous algebraic version.

  • $\begingroup$ Maybe you know that an ideal is prime if and only if the quotient by this ideal is a domain $\endgroup$ – Dario Oct 30 '17 at 16:25
  • $\begingroup$ Yes, in fact my strategy was to prove that $K[x_1,...,x_n]/I$ is isomorphic to some polynomial ring over $K$, so thaht this last one is a domain, hence $I$ is prime. The question was: How can i take a polynomial ring to be isomorphic to $K[x_1,...,x_n]/I$? $\endgroup$ – Minato Oct 30 '17 at 16:35
  • 1
    $\begingroup$ If you know Groebner basis theory - then a Groebner basis of $I$ ends up having all linear terms, with at most one generator having leading monomial $x_i$ for each $i$. From there, it's easy to see the quotient is isomorphic to $K[x_j]$ where $j$ ranges over values such that $x_j$ is not a leading monomial of any generator. (The Groebner basis calculation ends up just being equivalent to a Gaussian elimination on the coefficient matrix.) $\endgroup$ – Daniel Schepler Oct 30 '17 at 21:45

Yes, this is true. Here's probably the easiest way to prove it. Let $V$ be the vector space of linear polynomials in $K[x_1,\dots,x_n]$. Note that $\{1,x_1,\dots,x_n\}$ is a basis for $V$. Let $T:V\to V$ be any invertible linear map such that $T(1)=1$. Note that $T$ then induces an automorphism $T^*$ of the ring $K[x_1,\dots,x_n]$ by $T^*f(x_1,\dots,x_n)=f(Tx_1,\dots,Tx_n)$. This is an automorphism because it has inverse $f(x_1,\dots,x_n)\mapsto f(T^{-1}x_1,\dots,T^{-1}x_n)$. (To prove these operations really are inverse, note that when restricted to $V$, they are just $T$ and $T^{-1}$, since $T(1)=1$ and $T^{-1}(1)=1$. So their compositions when restricted to $V$ are the identity, which means their compositions on all of $K[x_1,\dots,x_n]$ are the identity since it is generated by $V$ as a ring. Geometrically, you can think of $T$ as an affine linear automorphism $K^n\to K^n$ and then $T^*$ is the induced automorphism of $K[x_1,\dots,x_n]$ when you think of polynomials as functions on $K^n$.)

Now let $I\subseteq K[x_1\dots,x_n]$ be an ideal generated by linear polynomials. Let $W=I\cap V$, which by assumption generates $I$. If $1\in W$, then $I$ is the entire ring, so we may assume $1\not\in W$. Let $y_1,\dots,y_r$ be a basis for $W$. Then $1,y_1,\dots,y_r$ are linearly independent, so there exists a linear automorphism $T$ of $V$ such that $T(1)=1$ and $T(y_i)=x_i$ for $i=1,\dots,r$.

Now consider the induced automorphism $T^*$ of $K[x_1,\dots,x_n]$. Since $T^*(y_i)=x_i$, $T^*$ maps $I$ to the ideal $(x_1,\dots,x_r)$. So, since $T^*$ is an isomorphism of rings, it suffices to answer your question in the special case $I=(x_1,\dots,x_r)$.

But in that case, your question is very easy, since the quotient $K[x_1,\dots,x_n]/(x_1,\dots,x_r)$ is just isomorphic to the polynomial ring $K[x_{r+1},\dots,x_n]$. So the quotient is indeed isomorphic to a polynomial ring, and in particular the ideal is prime since polynomial rings are domains.

  • $\begingroup$ A stupid question: to obtain a linear automorphism $T$ of $V$ such that $T(1)=1$ and $T(y_i)=x_i$ for $i=1,\dots,r$ i have to complete the basis (for $W$ ) $y_1,\dots,y_r$ to a basis for $V$ picking elements from the basis $1,x_1,\dots,x_n$ and than define $T$ such that is verified the above condition and such that T is bijective, right? Or there is something else? $\endgroup$ – Minato Nov 1 '17 at 18:31
  • $\begingroup$ More or less. You extend $1,y_1,\dots,y_r$ to a basis for $V$ however you want, and then extend $1\mapsto 1$, $y_i\mapsto x_i$ to a bijection between that basis and $1,x_1,\dots,x_n$. $\endgroup$ – Eric Wofsey Nov 1 '17 at 18:43

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.