Stability boundaries for nonreducible distributed delays (Q1091292)

This work is an important contribution in the domain of biological models involving delays, i.e. with dynamics depending on the earlier history of a variable quantity. Generally in such models one uses either discrete delays or distributed delays. In the first case the current rate of change of one or more variables depends on the value of a variable at a particular earlier time; in the other case it depends on an integral over all earlier values of the variable. The most widely used distributed delays are the gamma distributions which have the property to give a finite number of roots in the stability analysis. A characteristic of gamma distributed delays is that they have a slowly decaying memory of the distant past. The work presents a new type of delays with memory confined to a finite range of the past. Other properties of these delays are that the stability analysis is as simple as for gamma distributions, they go over as a natural limit to discrete delays, and give stability boundaries intermediate between those for discrete delays and those for gamma distributed delays. Two distributions are described in the paper, one is made up of two linear sections, the other of three smoothly joined quadratic sections, both are zero beyond a value equal to twice the mean delay. The particular features of these new distributed delays are shown by their use in two cases: a logistic population model and a general linear characteristic equation.