We discuss the quasianalytic properties of various spaces of functions suitable for one-dimensional small divisor problems. These spaces are formed of functions C 1 -holomorphic on certain compact sets K j of the Riemann sphere (in the Whitney sense), as is the solution of a linear or non-linear small divisor problem when viewed as a function of the multiplier (the intersection of K j with the unit circle is defined by a Diophantine-type condition, so as to avoid the divergence caused by roots of unity). It turns out that a kind of generalized analytic continuation through the unit circle is possible under suitable conditions on the K j 's.
IntroductionFollowing V. Arnold and M. Herman, and in the same line of research as in [MS03], we consider "monogenic functions" in the sense ofÉmile Borel with a view to small divisor problems. In these problems of dynamical origin, there is a complex parameter q, called multiplier, which must be kept off the roots of unity in order to solve a functional equation; typically, q is the eigenvalue at a fixed point of a onedimensional complex map that one wants to linearize and one studies the equation (corresponding to the so-called Siegel problem)(where G(z) ∈ zC{z} is given, with G ′ (0) = 1, and h(q, . ) is sought in a Banach space of functions holomorphic in the variable z), or the linearized equation h(q, qz)− qh(q, z) = qg(z) (with g(z) ∈ z 2 C{z} given), or the more complicated non-linear equation corresponding to the conjugacy between a circle map and a rigid rotation with rotation number 1 2πi log q (see equation (2)). We are interested in the dependence of the solution on the multiplier q. Roots of unity act as resonances, because the coefficients of the solution of the problem are inductively defined by expressions which involve division by q k − 1, k ≥ 1. On the other hand the case where |q| = 1 is particularly interesting from the dynamical point of view. One is thus led to define compact sets K j of the Riemann sphere C by removing smaller and smaller neighbourhoods of the roots of unity. It is shown in [He85] and [CM08] for the above-mentioned non-linear problems and in [MS03] for the linear one, that the solution is Whitney smooth on the K j 's, which gives rise to an example of "monogenic" function (the definition is recalled in Section 2). In all the cases we consider, the union of the K j 's on which our monogenic functions are defined contains C \ S, where S denotes the unit circle, and also a subset of S defined