Abstract.In this paper we show that for SI , a starlike Lipschitz domain, the dual of the space of harmonic functions in LP{S1) need not be the harmonic functions in Lq(Sl), where \/p + \/q = 1 . We show that, as a consequence, the harmonic Bergman projection for Í2 need not extend to a bounded operator on Lp(Sl) for all 1 < p < oo . The duality result is a partial answer to a question of Nakai and Sario [9] posed initially in the Proceedings of the London Mathematical Society in 1978. We treat the duality question as a biharmonic problem, and our result follows from the failure of uniqueness for the biharmonic Dirichlet problem in domains with sharp intruding corners.
IntroductionIn this paper we give a partial answer to a question posed by Nakai and Sario concerning the duals of harmonic Bergman spaces. Their question, initially posed in the Proceedings of the London Mathematical Society [9, p. 345] was whether, for 1 < p < oo, D. c R" , the dual of the space of harmonic functions in LP(Q.) is the space of harmonic functions in Lq(Q), where l/p + l/q = 1, 1 < p, q < oo. When the answer is affirmative, we will say that the duality property holds (for the given value of p). We note that every domain has the property for p = 2, and domains which are bounded and have a sufficiently smooth boundary have the property for every p .In 1981 Nakai and Sario [10] showed that the duality property can fail for Í2 = R \{z,, z2, ... , zn}, and one of the authors subsequently discovered that duality also fails for the punctured disk.In this paper we show that there is a p between 1 and 2 and a Lipschitz domain with an intruding corner for which the duality property fails. We show that, as a consequence, there are Lipschitz domains on which the harmonic Bergman projection does not extend to a bounded operator on a full range of p's between 1 and oo .Our approach is to reduce the duality question to a biharmonic problem. We first show that the duality property is equivalent to the direct sum decomposition