On the Integrate-and-Fire model with superlinear drift

Abstract

The Nonlinear Integrate-and-Fire equation is a fundamental model for neural assemblies, describing the dynamics of the density of neurons with a given potential. Its standard formulation uses a given firing potential, and is known to undergo finite-time blow-up for excitatory connections. To fit experimental observations, it has been advocated to use superlinear drifts. We show how, in this situation, it makes sense to set the equation on the whole line. The main question we address is whether in these circumstances the solution still exhibits finite-time blow-up. We study the construction of solutions by a fixed point and, using the framework of \cite{carrilloQualitativePropertiesSolutions2015}, we develop novel a priori estimates which use specifically the superlinearity of the drift at $+\infty$. In particular, we obtain Ces{à}ro-mean bounds on the activity, as well as exponential convergence towards the unique steady state in the weakly connected regime. Several critical steps rely on controls depending solely on the $L^1$ norm of the solution or coefficients.