In the first part of the paper we prove an existence theorem for gauge invariant
–normal
neighborhoods of the reduction loci in the space
of oriented connections on a fixed Hermitian 2–bundle
.
We use this to obtain results on the topology of the moduli space
of
(non-necessarily irreducible) oriented connections, and to study the Donaldson
–classes
globally around the reduction loci. In this part of the article we use essentially the
concept of harmonic section in a sphere bundle with respect to an Euclidean
connection.
Second, we concentrate on moduli spaces of instantons on definite 4–manifolds
with arbitrary first Betti number. We prove strong generic regularity results
which imply (for bundles with “odd" first Chern class) the existence of a
connected, dense open set of “good" metrics for which all the reductions in the
Uhlenbeck compactification of the moduli space are simultaneously regular. These
results can be used to define new Donaldson type invariants for definite
4–manifolds. The idea behind this construction is to notice that, for a good metric
, the geometry
of the instanton moduli spaces around the reduction loci is always the same, independently of
the choice of .
The connectedness of the space of good metrics is important, in order to prove that
no wall-crossing phenomena (jumps of invariants) occur. Moreover, we notice that,
for low instanton numbers, the corresponding moduli spaces are a priori compact and
contain no reductions at all so, in these cases, the existence of well-defined
Donaldson type invariants is obvious. Note that, on the other hand, there
seems to be very difficult to introduce well defined numerical Seiberg–Witten
invariants for definite 4–manifolds. For instance, the construction proposed by
Okonek and the author in [Seiberg–Witten invariants for 4–manifolds with
,
from: ”Complex analysis and algebraic geometry”, (T Peternell,
F O Schreyer, editors), de Gruyter, Berlin (2000) 347–357] gives a
–valued
function defined on a countable set of chambers.
The natural question is to decide whether these new Donaldson type invariants
yield essentially new differential topological information on the base manifold, or
have a purely topological nature.