Lecture 17 - Thu 2020 10 29 - Virtual % ============================================================================== % Re: last lecture. Please read the notes on asymptotic versus convergent on the web page. NEW TOPIC: BIFURCATIONS IN THE PLANE. First: bifurcations involving an isolated CP. % ---------------------------------------------------------------------------- % 1. Bifurcations of critical points involving loss of stability by a single eigenvalue: Saddle node, transcritical, pitchfork. Typically this means that all directions but one remain "stable", so that all the dynamics occurs along a 1-D curve [leading order: line through unstable eigenvector]. Thus: just take the 1-d cases and add a stable direction! Examples: Show that all involve interactions of saddles and nodes. In particular: Explain why name: "saddle-node" Soft pitchfork: two nodes and a saddle. Hard pitchfork: two saddles and a node. Alternative way to understand them (in the plane): look at the geometry of what the null-clines do. For example: Nullclines crossing/uncrossing yields a saddle-node. Adding restrictions [nullcline crossing cannot be destroyed, or symmetry] yields the other cases. % ---------------------------------------------------------------------------- % 2. Bifurcation of a critical point involving a 2-eigenvalue stability loss. NEW case: Hopf bifurcation. Complex pair of eigenvalues crosses imaginary axis. Bifurcation arises because of interplay between linearity and nonlinearity. Unstable linear terms and stabilizing nonlinearity produces stable limit cycle [soft Hopf]. Reverse produces unstable limit cycle [hard Hopf]. Situation is "similar" to what happens for a pitchfork; except that the effect of the nonlinearity is given by an average as the orbit goes around. Simple example/canonical equation: \dot{r} = a*r-r^3 and \dot{theta} = 1. [A] Explain why it has to be r^3; cannot be r^2: Analysis explanation: with r^2 equations are not smooth at r = 0 [write them in terms of x and y. Geometrical explanation: quadratic terms cancel at leading order in the radial direction. Their contribution to the vector field is the same at (x, y) and (-x, -y). Perturbative explanation: the quadratic terms produce no resonances, hence no secular terms in the solution [will see this more clearly later]. From [A], note: (i) Super and sub-critical cases. (ii) Bifurcation diagram. Amplitude of the limit cycle goes like sqrt{|a|}, where a is the departure from the bifurcation point at which linearized system has a center. This is generic: balance of linear and nonlinear terms happens at cubic level. Linear term epsilon*amplitude; where epsilon=imaginary part eigenvalues. Cubic nonlinearity amplitude^3 Balance when amplitude ~ sqrt{epsilon} This can be justified with asymptotics for Hopf bifurcations. % % ---------------------------------------------------------------------------- % % % ============================================================================== EOF