Ordinary differential equations: Bifurcations
Bistability and hysteresis
We consider the differential equation If then we find the equilibria by detecting the roots of the cubic polynomial . These are and . By local linearization one can verify (do it yourself!) that is a repeller, and that and are attractors. Because of the latter property we speak of a bistable differential equation.
The parameter shifts the cubic polynomial upwards or downwards. In such a translation, three roots of the cubic polynomial remain as long as .
If then and there are two equilibria: a semi-stable equilibrium and a stable equilibirum (check the nature of the equilibria yourself!).
If then and there are two equilibria: a semi-stable equilibrium and a stable equilibrium (check the nature of the equilibria yourself!).
This implies that the two bifurcations and are saddle-node bifurcations.
With a larger translation there exists only one root of the polynomial and this leads to a repelling equilibrium.
We explore what happens when we start with in the equilibrium and then gradually increase the parameter value with small steps and wait after each change until a new equilibrium has settled. With increasing , the point first remains in the vicinity on the lower branch of the stable equilibria; the solution goes in the course of time to a new equilibirum in the neighbourhood of where it came form along the branch in bifurcation diagram where the point is present. This continues until we reach at a saddle-node bifurcation; as soon as is slightly greater than , than the solution moves away to the stable equilibrium in the upper branch of the stable equilibria. When we now decrease the parameter value step by step to , then the point remains on the upper branch of the stable equilibria and we arrive for at the equilibrium . The parameter value returned to its original value, but the equilibrium not! This phenomenon of lack of reversibility is called hysteresis. It is visualized in the figure below.