Modeling the bursting effect in neuron systems (Q358160)

Bursting in systems of neurons can be characterized by alternating periods of dramatic changes (several spikes) and periods of almost no change of the membrane potential. This behavior can be modeled by means of singularly perturbed systems of ordinary differential equations. The authors present another approach based on a scalar singularly perturbed delay differential equation of the form \[ e{du\over dt}= f(u(t-h))- g(u(t-1))\tag{\(*\)} \] with \(0<\varepsilon\ll 1\), \(h> 0\), and where \(f\) and \(g\) are continuously differentiable functions with the properties (i) \(f(0)= 1\), \(g(0)= 0\), (ii) \(f(u)=- a_0+ O({1\over u})\), \(uf'(u)= O({1\over u})\) as \(u\to+\infty\), (iii) \(g(u)= b_0+ O({1\over u})\), \(ug'(u)= O({1\over u})\) as \(u\to\infty\), \(a_0> 0\), \(b_0> 0\). The authors prove to any fixed natural number \(n\) that there are parameters \(h\), \(a_0\), \(b_0\) such that, for sufficiently small \(\varepsilon\), equation \((*)\) has an exponentially orbitally stable cycle \(u= u_*(t,\varepsilon)\) with a period \(T_*(\varepsilon)\) satisfying \[ \lim_{\varepsilon\to 0} T_*(\varepsilon)= T_*> 0 \] with the property that on a time interval of length \(T_*(\varepsilon)\) the solution \(u_*(t,\varepsilon)\) has exactly \(n\) consecutive spikes whose height is of order \(\exp({h\over\varepsilon})\) and duration of order \((1+{1\over a_0})/\varepsilon\), on the remaining part of the interval it is asymptotically small.