Extension of Constants,
Rigidity
and
Two Conjectures Transparent to the B(ranch)
C(ycle) L(emma)
There is an exposition on the BCL
in deflist-cov/Branch-Cycle-Lem.html.
The 1st point of this paper was to take the most practical examples of
problems that had been around for years, and to show how immediately
the BCL can produce counterexamples to them.
The group theory is elementary, requiring just the differentiation of
some conjugacy classes (from their cycle type) in An,
the alternating group of degree n.
The 2nd point was to show how polynomial
covers f: x f(x)=z,
provide an elmentary yet, distinctly nontrivial setting to practice
using families of covers. The author had seen these conjectures
announced for several years, explained their falseness to various
lecturers and authors, and had yet found there was no response. Thus,
this paper. A fuller – and deeper – gamut of elementary BCL
uses appear in paplist-mt/fried-kop97.html,
paplist-cov/GQpresentation.html
and paplist-cov/ser_gal.html
their attached pdf files.
We assume understood the
monodromy group, Ĝf,K
of f over a field K: The Galois group of the
splitting field of f(x)=z
over K(z), with z a transcendental. Of course, we
assume f has coefficients in K,
the case K=Q being the territory of the stated
conjectures. When K is the
algebraic closure of Q, then
we drop the hat, and use
just the notation Gf.
This is the geometric
monodromy of f. Recall that
the cyclic polynomial of degree n
is xn, and the nth Chebychev polynomial is – for
an algebraist – the famed polynomial map from calculus cos(ϑ) →
cos(nϑ).
The Three Problems: To
remove trivial counterexamples known
to the proposers of
these problems we have two assumptions:
- f does not decompose as
two positive degree
polynomials of lower degree (it is indecomposable);
and
- f is neither, by affine
change of x or z, a cyclic nor a Chebychev polynomial.
Note: As was proved in an elementary, albeit essential, piece of the
resolution of the Schur Conjecture (say, [§4.b, paplist-ff/sch-carlitz.pdf]),
indecomposability of f (just for polynomials, not rational functions)
is independent of the choice of its definition field. Here is the
statement of the three conjectures, under assumptions #1 and #2.
- Cohen: There is no extension
of
constants (that is, Ĝ f,K=Gf).
- Chowla-Zassenhaus: For a prime p
large, f(x)-b
does not split into linear
factors for any b ∈Z/p.
The Upshot of the Paper:
Problems #3 and #4 are equivalent, and for
each odd non-square integer n,
there is a polynomial counterexample to each conjecture. The
counterexamples are polynomials we call (An,Sn)-realizations:
Geometric monodromy group is An,
but over Q the monodromy
(arithmetic) is Sn.
Further, using Muller's classification of the monodromy groups of
polynomials, we show n must
be odd.
Later, Muller completed the result by showing that our conditions were
exact, nonsquare is necessary. The paper also needed to eliminate
polynomials in the Davenport families of polynomials that arose
in such considerations as the genus 0 problem.
Finding all actual Nielsen
classes that can possibly produce (An,Sn)-realizations is the
toughest problem left by the paper. Our examples tie nicely between
these elementary results and considerably harder problems that require
some version of Hurwitz spaces to complete the analysis.