RENDICONTI DEL SEMINARIO MATEMATICO

Explosions in dimensions one through three 3forward iteration.DEFINITION 2 (Chain explosions). A chain explosion point (x,λ 0 ) is a pointsuch that x is chain recurrent for f λ0 , but there is a neighborhood N of x such that onone side of λ 0 (i.e. either for all λ < λ 0 or for all λ > λ 0 ), no point in N is chainrecurrent for f λ .Note that in the above definition, at f λ0 , x is not necessarily an isolated pointof the chain recurrent set. A well studied example of this is the explosion that occursat a saddle node bifurcation on an invariant circle. The chain recurrent set consists oftwo fixed points prior to bifurcation and the whole circle at and in many cases afterbifurcation. In subsequent usage, if the distinction is not important, we will refer torecurrent points rather than always saying chain recurrent.2.1. One dimensionThis section describes a classification of explosions via homoclinic tangencies in onedimension which appears in [2]. Although one dimension would seem to be the easiestcase, there are some key differences between one- and two-dimensional explosionswhich are not simplifications in one dimension. For example, for a diffeomorphism,homoclinic and heteroclinic orbits require the existence of saddle points with stableand unstable manifolds of dimension at least one. However, since a one-dimensionalmap is in general noninvertible, it is possible to have fixed or periodic points with onedimensionalunstable manifolds and a non-trivial zero-dimensional stable manifolds.Marotto terms such points snap-back repellers [22]. It is not possible to reverse thedimensions of the stable and unstable manifolds; the existence of a homoclinic orbitto an attracting fixed point requires a multivalued map [29]. In addition, the chainrecurrent set is not invariant under backwards iteration of a noninvertible map, so thediscussion of explosions in one dimension includes cases in which a point is not anexplosion point, but the preimages are explosion points. As this is a broad survey, thestatements and proofs of the results below are only sketches. The full details appearin [2].Let f be a one variable function with a repelling fixed point x 0 . Let y andk be such that y is a k th preimage of x 0 , and assume that a sequence of preimagesof y converge to x 0 . Then y is contained in an orbit which limits both forwards andbackwards to x 0 . That is, y is a homoclinic point for x 0 . Homoclinic points for periodicorbits are defined by replacing f with some appropriate iterate f m . Notice that fordiffeomorphisms, all orbits through homoclinic points are homoclinic orbits. For onedimensionalmaps, there may be many non-homoclinic orbits through a homoclinicpoint.Since the stable manifold of a homoclinic point is zero-dimensional, a homoclinictangency is a tangency of the graph of the map at a homoclinic point. That is, ahomoclinic tangency occurs if the graph of f has a horizontal tangent at a homoclinicpoint.