The Mokobodzki condition is a mathematical criterion that determines whether a "sandwich" problem has a solution. Specifically, given two random processes that act as lower and upper boundaries over time, the condition asks whether there exists a well-behaved process called a semimartingale that fits between them. Semimartingales are a broad and important class of processes in probability theory, underpinning much of modern mathematical finance and stochastic analysis. Knowing when such a process exists matters for problems like optimal stopping, game theory, and the pricing of financial derivatives with constraints.
The paper introduces a new numerical quantity, called the mean co-variation functional, that precisely measures how "compatible" two boundary processes are with each other. When the two boundaries coincide, this quantity reduces to a classical measure of process variation studied by the mathematician Rao. The authors prove that this functional is essentially equivalent to the smallest possible size, measured in a precise probabilistic norm, of any semimartingale that fits between the two boundaries. In other words, instead of just asking yes or no whether a solution exists, the paper provides a quantitative gauge of how tightly or loosely the boundaries can be sandwiched by a well-behaved process.
The result is proven in a general setting, without restrictive assumptions on the underlying probability space, and covers a full range of cases indexed by a parameter. The most technically delicate case, corresponding to the boundary of the parameter range, requires a different mathematical scale tied to a classical notion in probability called Doob's class D, which captures processes that are not too wild at stopping times. Overall, the work refines and extends classical results in stochastic process theory by turning a qualitative existence question into a precise, measurable quantity.