We consider the convolution inequality a * u≥v for given functions a and v, and we then investigate conditions on a and v that force the unknown function u to be positive or monotone or convex. We demonstrate that these results for abstract convolution equations can be specialized to yield new insights into the qualitative properties of fractional difference and differential operators. Finally, we apply our results to finite difference methods for fractional differential equations, and we show that our results yield insights into the qualitative behavior of these types of numerical approximations.1. Introduction. The division problem is the following: Let a and v be functions defined on a time scale T ⊂ R. Can we find a function u which satisfiesThis problem was posed and solved in complete generality by B. Malgrange and L. Ehrenpreis in the fifties and sixties. However, more precise questions, e.g., the regularity of the function u, were left open. This was called the Problem B by Ehrenpreis [14]. In this paper, we address following form of the Problem B : Under which conditions on a and, eventually, on initial values of u, can we find a positive, monotone or convex function u satisfying (1) -more precisely, we replace (1) with the somewhat more general convolution inequality a * u ≥ v? As we will explain momentarily, it turns out that this question has many applications in the theory of fractional differential and difference operators as well as finite difference methods for fractional differential equations.Convolution is a widely used technique in mathematical analysis, statistics and approximation theory, and has many applications, e.g., image and signal processing.