Some obstacles in characterising the boundedness of bi-parameter singular integrals

The famous T1 theorem for classical Calderón–Zygmund operators is a characterisation for their boundedness in L². In the bi-parameter case, on the other hand, the current T1 theorem is merely a collection of sufficient conditions. This difference in mind, we study a particular dyadic bi-parameter singular integral operator, namely the full mixed bi-parameter paraproductP, which is precisely the operator responsible for the outstanding problems in the bi-parameter theory. We make several remarks about P, the common theme of which is to demonstrate the delicacy of the problem of finding a completely satisfactory product T1 theorem. For example, P need not be unconditionally bounded if it is conditionally bounded—a major difference compared to the corresponding one-parameter model operators. Moreover, currently the theory even lacks a characterisation for the potentially easier unconditional boundedness. The product BMO condition is sufficient, but far from necessary: we show by example that unconditional boundedness does not even imply the weaker rectangular BMO condition.