In practice, the lex order Gröbner bases for many problems have the following shape. (I don't give the technical conditions that ensure this because (a) I don't know them, and (b) it doesn't matter anyway, just compute the Gröbner basis and see if it has the shape. The theorem says that it will usually happen).

That is, once is determined, all the other variables are determined by evaluation of polynomials at . This simple shape is quite convenient to work with, on occasion.

** Example.** Eigenvalue problems.
Eigenvalue problems with simple eigenvalues
(i.e. no repeated roots) lead to Gröbner bases which have the specified shape.
That is,
we start with the polynomial system
in the variables .
Then our ordinary analysis gives the equivalent system
from the characteristic polynomial, and once is
identified we solve the linear system for the **x**'s.
This
gives (on first blush) a solution rational
function of (e.g. by
Cramer's rule on a nonsingular submatrix),
but remember that if
we can transform any rational function of
into a polynomial in
by simplification (e.g.
).
(I'm not certain simplicity of the eigenvalues is required).

This is clearly a Gröbner basis, and has the required shape.

Tue Mar 12 21:09:19 EST 1996