3

Let $\mathbb{C}[z_1,z_2,...,z_n]$ be the ring of multivariate polynomials in the variables $z_1,z_2,...,z_n$ with complex coefficients. This ring is spanned by the countably infinite basis of monomials

$$e_{i_1,i_2,...,i_n}=z_1^{i_1}z_2^{i_2}\cdots z_n^{i_n}$$

for $i_j=0,1,2,...$ where $j\in\{1,2,...,n\}$.

Next, consider taking the quotient ring of $\mathbb{C}[z_1,z_2,...,z_n]$ by an ideal of $n$ known multivariate polynomials $\langle p_1,p_2,...,p_n\rangle$ in variables $z_1,z_2,...,z_n$ with complex coefficients:

$$Q=\frac{\mathbb{C}[z_1,z_2,...,z_n]}{\langle p_1,p_2,...,p_n\rangle}.$$

If $Q$ turns out to have finite dimension, in the sense that it is spanned by a finite subset of monomials $e_{i_1,i_2,...,i_n}$ (not known explicitly), how does one then compute the dimension of $Q$ in general? In other words, how does one compute the overall number of linearly independent $e_{i_1,i_2,...,i_n}$ that are a basis of $Q$?

user26857
  • 52,094
Kagaratsch
  • 2,239
  • So basically, what you are saying is that one should first reduce the ideal to a Groebner basis, eliminate all but one variable, and then treat the remaining univariate quotient ring as shown in this link? http://www.maths.manchester.ac.uk/~pas/code/notes/part13.pdf – Kagaratsch Jul 16 '16 at 20:00
  • Which would mean that the number of linearly independent monomials would be equal to the degree of the single variable (highest degree) polynomial in the Groebner basis? – Kagaratsch Jul 16 '16 at 20:01
  • 2
    In general of course this is difficult. But, if in addition you had the $p_i$s to be homogeneous, then Bezout's theorem will tell you that the dimension is just the product of the degrees of the $p_i$. – Mohan Jul 16 '16 at 21:55
  • Well, this is an interesting question, firstly, We need to note that The Provided ideals are of $\mathbb{C}[z_1,...,z_n]$ Next you can proceed by induction. In case $ n=1$ it's deg $p_i$ in case $n=2$ the given set $Q_2$ is you will and element is of the form $f(z_1,z_2)+a(z_1,z_2)p_1(z_1,z_2)+b(z_1,z_2)p_2(z_1,z_2)$ you can guess that it's $A_2=\max (\deg{p_i})_{i=1}^{2}$ Now induct. In case I am not forgetting the definition of ideal generated by $<c_1,...,c_2>$ the answer should be $A_n$ where $A$ indexed defines corollary as before as $A_2$ – Safal Das Biswas May 23 '23 at 21:19
  • I am answering this question wait a bit, Let me think it more properly. – Safal Das Biswas May 23 '23 at 21:33

1 Answers1

0

The ideal $\langle p_1,p_2,...,p_n\rangle$ can be equivalently expressed in terms of the corresponding Groebner basis $G$. In the following we consider the special case where the polynomials $\langle g_1,g_2,...,g_n\rangle$ of the Groebner basis have the following structure:

$$g_1=z_1-P_1(z_n)\\g_2=z_2-P_2(z_n)\\\vdots\\g_{n-1}=z_{n-1}-P_{n-1}(z_n)\\g_n=P_n(z_n)$$

where $P_i(z_n)$ for $i\in\{1,2,...,n\}$ are certain polynomials in the single variable $z_n$, while we assume that $z_n$ is the highest weight variable in the employed monomial ordering.

On the support of the ideal $\langle g_1,g_2,...,g_n\rangle$ we have the simultaneous condition $g_i=0$ for all $i\in\{1,2,...,n\}$. Therefore, $g_1$ through $g_{n-1}$ can be used to eliminate the variables $z_1$ through $z_{n-1}$ in the quotient ring $Q$, producing a new, equivalent, but effectively univariate quotient ring

$$Q\rightarrow Q'=\frac{\mathbb{C}'[z_n]}{\langle g_{n}\rangle}$$

In the univariate case, it is well known that for a quotient ring with an ideal containing a single polynomial of degree $m$, a minimal basis is spanned by the set of univariate monomials $e_i=z_n^{i-1}$ for $i\in\{1,2,...,m\}$ (see i.e. proposition 28 here). Therefore, $Q'$ has dimension $\dim_{\mathbb{C}}(Q')=m$, and by equivalence the same is true for $Q$:

$$\dim_{\mathbb{C}}(Q)=\dim_{\mathbb{C}}(Q')=\deg(P_n(z_n))=m$$

Due to the above, it seems that in order to obtain the dimension of a multivariate quotient polynomial ring over a zero dimensional ideal, one has to determine the degree of the univariate equation in the Groebner basis for the polynomial ideal (if the Groebner basis satisfies the structure above). Which can be rather difficult to compute. But conceptually, this should be a possible way to go about. Any objections anyone?

user26857
  • 52,094
Kagaratsch
  • 2,239
  • Why would a Groebner basis have that form? – Mariano Suárez-Álvarez Jul 17 '16 at 03:22
  • @MarianoSuárez-Alvarez because per definition a Groebner basis always exists that linearizes the relation between all but one variable to the highest weight variable. That is the whole purpose of computing a Groebner basis in the first place, as far as I know. – Kagaratsch Jul 17 '16 at 14:59
  • @MarianoSuárez-Alvarez admittedly, this point is not always made completely clear. Take the example on page 2 in http://people.math.umass.edu/~norman/462_11/notes/m462notes13.pdf for instance. What they claim to be the minimal Groebner basis can actually still be reduced to just $g_1=y^3$ and $g_2=x-2y^2$, which displays the structure claimed above (in a different variable weight order). I know its just an example, but I don't seem to be able to track down a general reference right now. – Kagaratsch Jul 17 '16 at 15:10
  • 1
    A groebner basis of the form you describe can only exist if the function mapping a point of the zero set of the ideal to its last coordinate is injective, and it is easy to find examples of ideas for which this does not work for any ordering of the components. – Mariano Suárez-Álvarez Jul 17 '16 at 16:50
  • Oh, I did not know this! Sorry about that. I will edit the answer to emphasize this point. – Kagaratsch Jul 18 '16 at 19:15