Laurent exponents in connected families #
Suppose a function on a product has, near every point of a distinguished hyperplane, a Laurent-power expression in the transverse variable with an analytic nonvanishing unit. For a fixed global function the integer exponent is locally constant along the hyperplane: two such expressions at nearby points overlap, and uniqueness of the order of an analytic germ identifies their exponents. Consequently the exponent is constant on a preconnected set of parameters.
The pointwise uniqueness statement laurentExponent_eq is also exported, so that exponents
obtained from independent constructions can be compared.
The globalization result is stated simultaneously for an arbitrary family of functions. It is used for global root labels in parameterized Puiseux factorizations: the labels having finite limits are precisely those with nonnegative exponent, so their set cannot change over a connected base.
References #
- S. McCallum, A. ParusiΕski, L. Paunescu, Validity proof of Lazard's method for CAD construction, Journal of Symbolic Computation 92 (2019), 52β69, Β§5.2.
Uniqueness of the Laurent exponent at a point of the hyperplane. If, near (x, 0),
the function r is both b.2 ^ m * u b and b.2 ^ n * v b off the hyperplane b.2 = 0,
with u and v analytic and nonvanishing at (x, 0), then m = n. No connectedness is
needed: the two expressions are compared on the transverse line through (x, 0).
Laurent exponents are constant in preconnected families. Suppose that, near every
(x, 0) with x β S, every function r i is a power of the transverse coordinate times
an analytic unit. Then one exponent can be chosen for each global label i, independently
of x. The analytic unit may still depend on the point.
No finiteness assumption on the label type is needed. Any pointwise property P i e of the
exponents, such as a common lower bound, is retained. Negative exponents and an empty label type
are allowed.