Topics for APMA E4101 - Fall, 2018, v1.

0
Lightly modified from the 2017 Study Guide

Marc Spiegelman∗
December 3, 2018

Sources: Strogatz, Hirsch-Smale-Devaney, lectures

Some General Comments

• The final exam is cumulative but will emphasize later material

• The main text for the course was Strogatz (mainly Chapters 2, 3, 5, 6, 7, 8, 9 and a bit of 10),
but there was some additional material covered in class that is not in Strogatz.

• An efficient approach to studying is to work on problems which “reach back” to earlier

topics. So, for example, a problem requiring one to sketch the phase portrait for a nonlin-
ear systems of equations with equilibria, requires one to use one’s understanding of linear
systems in analyzing the local behavior near equilibria .

• Look at the Student Solution guide to Strogatz

Topics:
1. Explicit solutions of scalar linear and nonlinear ODEs Solution of first order initial value
problems x0 = f (x, t) with initial condition x(0) = x0 . In particular, the following equations
(see flowchart for more details):

(a) x0 = b(t), where b(t) is a given function

(b) x0 = ax, (a, a constant)
(c) separable problems x0 = f (x)
(d) linear problems x0 = a(t)x + b(t) - solution via an integrating factor
(e) Bernoulli equations x0 + a(t)x = b(t)xn
(f) Exact Equations
(g) Non-linear Homogeneous equations
∗
Dept. of Applied Physics & Applied Mathematics, and Dept. of Mathematics, Columbia University, New York,
NY 10027

1
2. Existence and Uniqueness of Ẋ = F (X), X(0) = X0 for the vector field F : Rn → Rn ,
e.g. n = 2    
x f (x, y)
7→
y g(x, y)

(a) Assume F is a smooth (for example, C 1 ) vector field. Existence, uniqueness of solu-
tions on initial data for the initial value problem (IVP)

X 0 = F (X), X(t0 ) = x0 (1)

the “flow” φ(t; X0 ) is the unique solution of the initial value problem (1).
(b) Examples illustrating:
i. Local-in-time (but not global) existence (ẋ = x2 and finite time blow-up) vs.
global-in-time existence (ẋ = x)
1
ii. uniqueness vs. non-uniqueness (ẋ = x 3 , x(0) = 0)
(c) Continuous dependence of solutions on initial data - see HSD

3. Numerical Methods for Integration of ODE IVP’s: Understand algorithms and errors for
the following Explicit methods

(a) Euler’s method

(b) Improved Euler’s method (RK2)
(c) 4th order runge-kutta

4. 1D systems: flows on the line

(a) Flows on the line: 1d phase line for an autonomous equation: x0 = f (x)
(b) Identify fixed points (f (x) = 0) and apply local linear stability analysis (as well as
graphical analysis)
(c) Example: the Logistic equation: x0 = x(1 − x);

(d) Direction / Slope field (in the (x, t)-plane) for the scalar non-autonomous eqn x0 =
f (x, t). Note: A non-autonomous system may be viewed as an autonomous sys-
tem by introducing t as a dependent variable, t = t(s), E.g.

x0 = f (x, t), x(0) = x0

is equivalent to
       
d x f (x, t) x(s)  x0
= , =
ds t 1 t(s) s=0
 0

which is really a 2-D system (see below)

(e) Bifurcations: for ẋ = f (x; r) for some parameter r

2
 i. Classification/ Normal forms of Saddle Node, Pitchfork and Transcritical bifurca-
tions for scalar ODEs
ii. identify critical points where f (x; r) = 0 and fx (x; r) = 0
iii. Draw bifurcation diagrams. example Logistic equation with constant-rate harvest-
ing, x0 = x(1 − x) − h
5. Higher Dimensional phase portraits
(a) F : Rn → Rn , n− dimensional phase space of an autonomous system of ODEs
(b) n = 2, ODE flows in the plane

ẋ = f (x, y), ẏ = g(x, y) (2)

(c) Understand why two trajectories of a smooth flow cannot cross in finite time.
(d) Understand that they can intersect at a fixed point at infinite time (t → ∞). Why?
6. Linear Dynamical systems: X 0 = AX, where A is a constant 2 × 2 matrix:
(a) Transforming the general 2 × 2 systems X 0 = AX, where A is real to a real canonical
form: Y 0 = CY , where T is real, invertible and T −1 AT = C; See HSD.
(b) Find general solutions and phase portraits in the transformed canonical frame.
(c) Plot phase portraits in the original frame (i.e. X = T Y ) also see HSD
(d) Draw phase portraits for two dimensional systems: X 0 = AX, where A is a real 2 × 2
matrix
(e) Classification of planar linear systems using the Tr(A) and |A| diagram.
(f) Exploring how linear systems evolve on the Tr(A) and |A| diagram as a function of
changing parameters A(p).
7. Non-linear Dynamical Systems
(a) Review concepts of equilibrium (fixed point), stability, instability, asymptotic stability
(attracting) of solutions of X 0 = F (X).
(b) Writing a single scalar equation of higher order as a first order system of equations (e.g
a non-linear oscillator ẍ + µ(x)ẋ + x = 0)
(c) Equilibrium solutions: F (X ? ) = 0 at fixed points X ?
(d) Linearized stability analysis of equilibria: linearized perturbation equation
(e) Nonlinear systems with multiple equilibria. e.g. “rabbits and sheep”. Use of lineariza-
tion to study the local phase portrait near each equilibrium.
(f) More global view of phase portrait: Stable and unstable manifolds of a hyperbolic
equilibrium (fixed) point
(g) Examples where linearization fails to predict correct local phase portrait
(h) Notion of a hyperbolic fixed point (Section 6.3 of Strogatz, Chapter 8 of Hirsch-Smale-
Devaney)

3
 (i) Local vs. Global behavior – linearization gives, at best, local information
(j) Nullclines, Conserved energy and their uses in understanding global characteristics of
phase portrait
(k) Conservative systems
(l) Hamiltonian system (a special class of conservative system) - volume preserving char-
acter of the phase-flow and its consequences;
(m) The phase portait of one-dimensional Newtonian conservative systems: mẍ(t) = −V 0 (x(t)),
where V (x) is a potential. Let y = ẋ. Then, E(x(t), y(t)), kinetic plus potential energy
is conserved on any solution. It follows that the phase space is made up of the level sets
of the energy function E(x, y).
(n) Nonlinear Centers (Theorem 6.5.1 of Strogatz and ideas behind its proof)
(o) Application of a conserved energy to conclude that an equilibrium which is “linear
center” (linearized system having a pair complex conjugate eigenvalues) is in fact a
“nonlinear center”.
(p) Nonlinear pendulum: θ̈(t) + sin θ(t) = 0
(q) Damped nonlinear pendulum: θ̈(t) + sin θ(t) = −µθ̇, µ > 0.
Use of energy-dissipation: dtd E(θ, θ̇) = −µ(θ̇)2 ≤ 0
(r) Liapunov function
(s) Gradient dynamical systems, Ẋ = −∇X V (X).
(t) Index Theory Given a vector field, (x, y) 7→ (f (x, y), g(x, y)). IC , the index of a
closed curve C ⊂ R2 , which does not pass through a fixed point of the vector field
1
= 2π [∆φ]C
(u) Computation of indices of elementary fixed points of X 0 = AX, where A is a real 2 × 2
matrix.
(v) Limit Cycles: Isolated closed orbit in the two-dimensional phase plane ( path traced
out by non-constant time-periodic solutions). Isolated means that in some neighbor-
hood about the orbit there are no other periodic orbits.
i. Poincaré-Bendixson Theorem and how to apply it
ii. Ruling out limit cycles in a region of phase space.
iii. Supercritical and subcritical Hopf bifurcations
(w) Bifurcations in higher-dimensional systems

8. The Lorenz Equations

(a) Understand the physical meaning (if not the derivation) of the Lorenz Equations.
(b) Demonstrate contraction of volume in dissipative systems
(c) Find and classify the fixed points and bifurcations in Lorenz
(d) Define (roughly) Chaos and Attractor.
(e) Sensitive dependence on initial conditions and Liapunov Exponent

4
 (f) The Lorenz Map: argue for lack of periodic orbits

9. Discrete-time dynamical systems iteration of a function:

x0 , x1 = F (x0 ), . . . , xn = F (xn−1 ), . . .

For example, the tent-map or logistic map (or Lorenz Map)

(a) Find fixed points of discrete-time dynamical systems and determine their stability.
(b) Find p-cycles of discrete-time dynamical systems and find stability