Algebraic conditions for second-moment stability boundaries of linear, time-invariant stochastic delay-differential equations

For linear, time-invariant stochastic delay-differential equations with a single constant delay and both multiplicative and additive noise, this paper derives optimal semi-analytic algebraic equality conditions that can be used to identify second-moment stability boundaries without the use of problem discretization.

Successful validation against Monte Carlo simulations and published results for several low-dimensional models clarifies limitations of stability conditions proposed in the literature and demonstrates considerable savings in computational effort relative to discretization-based approaches.

In particular, using the theory derived in this paper, second-moment stability boundaries are shown to be computable using parameter continuation techniques applied to discretization-free equality conditions that scale only with the square of the problem dimension.

For the case of one-dimensional stochastic delay-differential equations, in particular, the analysis is entirely closed form with a stability condition expressed entirely in terms of elementary functions.

These results are enabled by the derivation of an advection-type boundary-value problem with non-local boundary conditions for a three-variable correlation function followed by a reduction to a delay-differential boundary-value problem for a two-variable correlation function.

For the former problem, observations regarding the spectral abscissa of the discretization of the corresponding infinitesimal generator, particularly that second-moment stability is lost when a real eigenvalue passes through the origin, motivate identification of second-moment stability boundaries with a loss of uniqueness of stationary solutions to the latter problem.