MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $k[x_0,\cdots,x_N]$ be the polynomial ring. Suppose that we have the monomial orders on $k[x_0,...,x_m]$ and $k[x_m,...,x_N]$ then is there a monomial order on $k[x_0,\cdots,x_N]$ such that this order induces given monomial orders on $k[x_0,\cdots,x_m]$ and $k[x_m,\cdots,x_N]$? I think it may be possible, but i dont know how to construct the monomial order formally.

share|cite|improve this question
You need at least a compatibility condition, namely that the orders induced on $k[x_m]$ coincide. – Alberto García-Raboso Feb 9 '12 at 15:40

It is a result of Lorenzo Robbiano's (Term orderings on the polynomial ring; see also section 1.2 of Greuel and Pfister's A Singular Introduction to Commutative Algebra) that every monomial ordering on a polynomial ring in $n$ variables can be obtained from a matrix $A \in \mathrm{GL}(n, \mathbb{R})$ in the following way: $$ x^\alpha >_A x^\beta \qquad\Leftrightarrow\qquad A\alpha > A\beta $$ where you arrange the exponents as a column vector and the ordering on the right-hand side is the lexicographical ordering on $\mathbb{R}^n$.

Notice that different matrices can give rise to the same monomial ordering. E.g., there are only two different monomial orderings on $k[x]$: the lexicographical ($x > 1$) and the negative lexicographical ($1 < x$). Any 1x1 matrix with positive entry will give the first one; the second one can be obtained from any 1x1 matrix with negative entry.

Choose matrices $$ A = (a_{ij})_{i,j = 0, \ldots, m} \in \mathrm{GL}(m+1, \mathbb{R}) $$ and $$ B = (b_{ij})_{i,j = m, \ldots, N} \in \mathrm{GL}(N-m+1, \mathbb{R}) $$ that give rise to your original monomial orderings $>_A$ on $k[x_0, \ldots, x_m]$, and $>_B$ on $k[x_m, \ldots, x_N]$, respectively. The signs of $a_{mm}$ and $b_{mm}$ determine the restrictions of $>_A$ and $>_B$ to $k[x_m]$. If you want to "glue" this orderings then it is necessary that these coincide, i.e., that the signs of $a_{mm}$ and $b_{mm}$ be the same. I claim that this condition is also sufficient.

We thus want to construct a matrix $$ C = (c_{ij})_{i,j = 0, \ldots, N} \in \mathrm{GL}(N+1, \mathbb{R}) $$ Start by scaling $B$ in such a way that $a_{mm} = b_{mm}$ (you can always multiply a matrix by a positive number without changing the associated ordering). Define then

  • $c_{ij} = a_{ij}$ for $0 \leq i, j \leq m$,
  • $c_{ij} = b_{ij}$ for $m \leq i, j \leq N$,
  • $c_{ij} = 0$ otherwise.

If this matrix is invertible, then it defines an ordering $>_C$ on $k[x_0, \ldots, x_N]$ satisfying your requirements.

Let me denote by $A_i$ the columns of $A$. Since $A$ is invertible, these are $m+1$ linearly independent vectors in $\mathbb{R}^{m+1}$. The same can be said about the columns $B_i$ of $B$ as vectors in $\mathbb{R}^{N-m+1}$. Let now $C_i$ be the columns of $C$, and suppose they satisfy a linear relation $$ \lambda_0 C_0 + \ldots + \lambda_N C_N = 0 $$ in $\mathbb{R}^{N+1}$. Projecting onto the first $m+1$ coordinates yields $$ \lambda_0 A_0 + \ldots + \lambda_m A_m = 0 $$ which implies $\lambda_0 = \cdots = \lambda_m = 0$. Similarly, projection onto the last $N-m+1$ coordinates results in the equation $$ \lambda_m B_m + \ldots + \lambda_N B_N = 0 $$ forcing the vanishing of the remaining $\lambda$'s and proving that the columns of $C$ are linearly independent.

share|cite|improve this answer
@willy: if my post answers your question, you should accept it. – Alberto García-Raboso Feb 9 '12 at 21:56
The matrix C which you make is not invertible in general. and moreover the C does not induce the order on $k[x_m,\cdots,x_N]$ what i given. finally, $a_{mm}$ and $b_{mm}$ may be zeros – willy Feb 20 '12 at 10:00
@willy: $a_{mm}$ cannot be zero, for then $>_A$ would induce the zero order on $k[x_m]$, which is not an order. Equivalently, $a_{mm} = 0$ would say that be saying the vector $x_m$ is an eigenvalue of $A$ with eigenvalue zero, contradicting the invertibility of $A$. – Alberto García-Raboso Feb 20 '12 at 13:24
Consider the permutation matrix $A$, this matrix can have the $a_mm=0$, but this matrix defines the order on $k[x_0,\cdots,x_m]$, moreover $a_mm=0$ does not imply the matrix is not invertiable, for example, the matrix (0,1:1,0) is invertible and this matrix defines the lexicographic order on $k[x_0,x_1]$ such that $x_1>x_0$. – willy Feb 21 '12 at 17:50

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.