Krull dimension of a ring $R$ is the supremum of the number of strict inclusions in a chain of prime ideals.

Question 1. Considering $R = \mathbb{C}[x_1, x_2, x_3, x_4]/\left< x_1x_3-x_2^2,x_2 x_4-x_3^2,x_1x_4-x_2 x_3\right>$, how does one calculate the Krull dimension of $R$? This variety is well-known as the twisted cubic in $\mathbb{P}^3$.

Question 2. In general for any ring $R$, how are the Krull dimension of $R$ and the dimension of Spec$(R)$ related?

Thank you.

  • 5
    $\begingroup$ Nitpick: the variety is not the twisted cubic but the affine cone over that cubic, which explains why the Krull dimension of $R$ is $2$. $\endgroup$ – Georges Elencwajg Jul 6 '12 at 6:52
  • $\begingroup$ Thank you Georges! Then what are the equations for a twisted cubic? I thought these were the ones (I thought I saw this from several books)... $\endgroup$ – math-visitor Jul 6 '12 at 6:54
  • 2
    $\begingroup$ Dear math-visitor: the confusing point is that the equations are exactly the same! However in the case of the twisted cubic they are to be interpreted as defining a curve in $\mathbb P^3$, rather than a cone un $\mathbb A^4$. So to keep things straight, we write $Proj(R)$ for cubic the and $Spec (R)$ for the cone. Let me make an analogy: the single equation $x=0$ defines a line in $\mathbb A^2$; however the very same equation defines a plane in $\mathbb A^3$. $\endgroup$ – Georges Elencwajg Jul 6 '12 at 7:09
  • $\begingroup$ Many thanks again Georges! I had to read this twice and in the future, I hope not to confuse myself and others! =) $\endgroup$ – math-visitor Jul 6 '12 at 7:14
  • 2
    $\begingroup$ By the way, you can find an elementary, completely explicit calculation that the ideal defining the cubic is prime here. $\endgroup$ – Georges Elencwajg Jul 6 '12 at 7:15

Question 1: It is a theorem that $\mathrm{dim}\ A[x]=\mathrm{dim}\ A+1$ for any Noetherian ring $A$, where $\mathrm{dim}$ denotes Krull dimension. Thus $\mathrm{dim}\ \mathbb C[x_1,x_2,x_3,x_4]=4$, as $\mathrm{dim}\ \mathbb C=0$ trivially. The easiest way to compute the dimension of $R$ is to verify that $$P=\langle x_1x_3-x_2^2,x_2 x_4-x_3^2,x_1x_4-x_2 x_3\rangle$$ is a prime ideal, i.e. by the methods I employ here, and computing its height. As a caution, I have not proved that $P$ is prime myself but I believe that it is. To determine its height, notice that $\mathbb C[x_1,x_2,x_3,x_4]$ is a UFD so all prime ideals of height $1$ are principal, while $P$ is easily shown not to be principal. By the Generalized Principal Ideal Theorem, $P$ has height at most it's number of generators, i.e. $3$. Hence we need only determine whether $P$ has height $2$ or $3$. In fact, it has height $3$, which can be shown by verifying that $$0\subset \langle x_1x_3-x_2^2\rangle\subset \langle x_1x_3-x_2^2,x_2 x_4-x_3^2\rangle\subset P$$ is a strict chain of prime ideals, e.g. by the methods in the post I linked to.

Question 2: Yes, the dimensions are equal. See for example Wikipedia.

Edit: Not doing calculations can get you in trouble. As Mariano's answer shows, $P$ has height $2$ rather than $3$ (so apparently the second ideal in the chain I wrote is not prime), hence the dimension of $R$ is $4-2=2$.

  • $\begingroup$ Thank you Alex. I will read what you wrote above carefully. And thanks for the additional link reference! $\endgroup$ – math-visitor Jul 6 '12 at 3:32
  • $\begingroup$ @math-visitor If you have Mathematica I recommend using it, but Sage is also a good option. Also, the calculations are tedious so I haven't done them myself, so there's a chance that my intuition is failing me and one of the ideal I claim is prime, isn't. $\endgroup$ – Alex Becker Jul 6 '12 at 3:34
  • $\begingroup$ Thanks Alex! I thought about doing the above calculations by hand, but I will definitely consider using a computer program. =) $\endgroup$ – math-visitor Jul 6 '12 at 3:36
  • 3
    $\begingroup$ @math-visitor Computing Groebner bases by hand? That would be horrid. $\endgroup$ – Alex Becker Jul 6 '12 at 3:40
  • $\begingroup$ For completeness' sake, the second ideal in the chain you wrote contains the elements $$x_4(x_1x_3-x_2^2)+x_2(x_2x_4-x_3^2)=-x_3(x_1x_4+x_2x_3),$$ $$x_3(x_1x_3-x_2^2)+x_1(x_2x_4-x_3^2)=-x_2(x_1x_4+x_2x_3),$$ which can be used to show that the ideal is not prime. $\endgroup$ – Servaes Jun 12 at 23:59

Your ideal is generated by binomials, so one can be smart about it.

There is an algebra map $\mathbb C[x_1,x_2,x_3,x_4]\to \mathbb C[s, t]$ such that $x_1\mapsto s^3$, $x_2\mapsto s^2t$, $x_3\mapsto st^2$ and $x_4\mapsto t^3$, and this map maps your ideal $\mathfrak p$ to zero, so it induces $\phi:\mathbb C[x_1,x_2,x_3,x_4]/\mathfrak p\to \mathbb C[s, t]$. This is, in fact, an injection —try to prove this by hand— so that $\mathfrak p$ is prime because the codomain of $\phi$ is a domain.

It follows, in particular, that $A=\mathbb C[x_1,x_2,x_3,x_4]/\mathfrak p$ is a domain, so its Krull dimension is equal to the trascendence degree of its fraction field $F=\operatorname{Frac}(A)$. In $F$ we have $x_1=x_2^2/x_3$ and $x_4=x_3^2/x_2$, so the inclusion $\mathbb C(x_2,x_3)\to F$ is actually surjective. This makes it obvious that $F$ is of trascendence degree $2$.

  • $\begingroup$ Wow, cool! Thank you Mariano! I wish all quadratic polynomial problems are cool and simple like this. $\endgroup$ – math-visitor Jul 6 '12 at 4:21
  • 1
    $\begingroup$ @math-visitor, every variety can be cut by cuadrics (in some sense: mathoverflow.net/questions/11488/varieties-cut-by-quadrics) so you are out of luck :) $\endgroup$ – Mariano Suárez-Álvarez Jul 6 '12 at 4:38
  • $\begingroup$ Thank you for the link-- a lot of info there! Well, I was hopeful for the time-being... $\endgroup$ – math-visitor Jul 6 '12 at 4:40

I'll prove the following result:

$$K[x_1, x_2, x_3, x_4]/\left< x_1x_3-x_2^2,x_2 x_4-x_3^2,x_1x_4-x_2 x_3\right>\simeq K[s^3,s^2t,st^2,t^3],$$ where $K$ is a field.

Let $\varphi: K[x_1,x_2,x_3,x_4]\to K[s, t]$ be the ring homomorphism that maps $x_1\mapsto s^3$, $x_2\mapsto s^2t$, $x_3\mapsto st^2$ and $x_4\mapsto t^3$. Obviously $\operatorname{Im}\varphi=K[s^3,s^2t,st^2,t^3]$; this is a subring of $K[s,t]$ and the extension $K[s^3,s^2t,st^2,t^3]\subset K[s,t]$ is integral, hence $\dim K[s^3,s^2t,st^2,t^3]= \dim K[s,t]=2.$

It remains to prove that $\ker\varphi=\left< x_1x_3-x_2^2, x_2 x_4-x_3^2, x_1x_4-x_2 x_3\right>$. By definition $\varphi(f(x_1,x_2,x_3,x_4))=f(s^3,s^2t,st^2,t^3)$. In particular, this shows that the polynomials $g_1=x_1x_3-x_2^2$, $g_2=x_2 x_4-x_3^2$ and $g_3=x_1x_4-x_2 x_3$ belong to $\ker\varphi$.

Conversely, let $f\in\ker\varphi$, i.e. $f(s^3,s^2t,st^2,t^3)=0$. We want to show that $$f\in\left< x_1x_3-x_2^2, x_2 x_4-x_3^2, x_1x_4-x_2 x_3\right>.$$ The initial monomials of $g_1$, $g_2$, resp. $g_3$ with respect to the lexicographical order are $x_1x_3$, $x_2x_4$, resp. $x_1x_4$. The remainder on division of $f$ to $G=\{g_1,g_2,g_3\}$, denoted by $r$, is a $K$-linear combination of monomials none of which is divisible by $x_1x_3$, $x_2x_4$, resp. $x_1x_4$. This shows that the monomials of $r$ can have one the following forms: $x_1^ix_2^j$ with $i\ge 1$ and $j\ge 0$, $x_2^kx_3^l$ with $k\ge 1$ and $l\ge 0$, respectively $x_3^ux_4^v$ with $u,v\ge 0$. In order to prove that $f\in\left< x_1x_3-x_2^2, x_2 x_4-x_3^2, x_1x_4-x_2 x_3\right>$ it is enough to show that $r=0$. But we know that $f(s^3,s^2t,st^2,t^3)=0$ and therefore $r(s^3,s^2t,st^2,t^3)=0$. The monomials (in $s$ and $t$) of $r(s^3,s^2t,st^2,t^3)$ are of the following types: $s^{3i+2j}t^j$, $s^{2k+l}t^{k+2l}$, respectively $s^ut^{2u+3v}$. Now check that there is no possible cancelation between these monomials (because they can't be equal), so $r=0$.

Now it follows that $\dim R=2$.

  • 4
    $\begingroup$ This proof can be easily generalized in order to show that, in general, the coordinate ring of a rational normal scroll is isomorphic to $K[s^d,s^{d-1}t,\dots,st^{d-1}t^d]$. $\endgroup$ – user26857 Jan 9 '13 at 20:10

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.