Backward stochastic differential equations (BSDEs) are a class of equations used widely in financial mathematics, stochastic control, and probability theory. Unlike ordinary differential equations that evolve forward in time from an initial condition, BSDEs work backward from a specified endpoint, called the terminal condition. Solving them becomes much harder when the equations involve random, time-varying coefficients that can grow large in an uncontrolled way, and when the time horizon itself is random rather than fixed. This paper tackles exactly that challenging setting, asking when well-behaved solutions still exist under such general and unpredictable conditions.
The authors introduce a carefully chosen weighted function space to handle the difficulty. By multiplying the usual measure of solution size by an exponential weighting factor that depends on a process related to the equation's growth coefficients, they are able to tame the unbounded randomness and work in a tractable mathematical environment. A key technical contribution is a relaxed condition on a controlling process called rho, which captures how fast the equation's coefficients can grow. Compared to a closely related recent paper, the authors require a slightly weaker condition on rho when the parameter p is greater than 1, which broadens the range of equations their results cover.
Within this weighted framework, the paper delivers several concrete results: conditions under which a solution exists, conditions under which a solution is unique, identification of smallest and largest solutions when multiple solutions might exist, and two comparison theorems that tell you how solutions change when the inputs are ordered relative to each other. Taken together, these results unify and extend a scattered collection of earlier findings in the literature, and the techniques developed to handle the interplay between unbounded stochastic coefficients and the weighted space are new contributions in their own right.