Sem#4: Problem 7-02 (from 2020)

[Document copied from 2020 - hopefully I fixed the links!]

 

Just a clarification of ambition when it comes to phase diagrams. When I teach - and this applies now as well, since I'm responsible for the exam - I promise to never ask you to "draw" a phase diagram, only "sketch" it. 

The difference is that it does not need to be pretty as long as it is qualitatively correct: your null-isoclines should look about the way they actually do; paths across an x-isocline should be vertical or reasonably close; you should not reverse arrows or directions. 

I mean, look at the the front page of the Norwegian book. Links to an external site. Yes that is a phase diagram; the circle and the parabola are null-isoclines, and that curve that goes all over the picture from (-3,-1) to (4,3) is the convergent saddle path. Who wants to spend an hour of an exam making that thing pretty?
But I expect you to be able to draw a parabola and a circle, and please take note how the paths cross the nullclines; vertically or horizontally. 

 

System: 

LaTeX: \begin{align}\dot x&= ax+2y+\alpha\\
\dot y&= 2x+ay+\beta\end{align}˙x=ax+2y+α˙y=2x+ay+β

Because the question says "for all", we must expect that for some parameters, the matrix-based approach will get us into trouble (repeated roots, ...), so let us do it like 7-01: 

  • Deduce a diff.eq. for x
  • Solve that
  • Find y. From the first equation, LaTeX: y=\tfrac12\big(\dot x-ax-\alpha\big)y=12(˙xaxα), so when x is found, we differentiate it and insert and that is it.

 

A differential equation for x

... will be of the form LaTeX: \ddot x-\dot x\textsf{ tr }\mathbf A+x\textsf{ det }\mathbf A=f(t)¨x˙x tr A+x det A=f(t). (Here trace is 2a and determinant is a2-4.) You can look up in the book a formula for f(t), or you can just do the job: 

  • Differentiate the first eq.: LaTeX: \ddot x=a\dot x+2\dot y¨x=a˙x+2˙y
  • Insert for LaTeX: \dot y˙y: LaTeX: \ddot x=a\dot x+2(2x+ay+\beta)¨x=a˙x+2(2x+ay+β)
  • Eliminate the 2y using the first equation that says LaTeX: 2y=\dot x-ax-\alpha2y=˙xaxα, to get LaTeX: \ddot x=a\dot x+4x+2\beta+a(\dot x-ax-\alpha)¨x=a˙x+4x+2β+a(˙xaxα)
  • Reorganize into LaTeX: \ddot x-2a\dot x+(a^2-4)x=2\beta-a\alpha¨x2a˙x+(a24)x=2βaα

 

Solve the homogeneous equation for x:

Like 7-01, except with a parameter: 

  • Characteristic roots: LaTeX: r=a\pm\sqrt{a^2-(a^2-4)}=a\pm 2r=a±a2(a24)=a±2, two distinct and real.
  • Homogeneous equation has general solution LaTeX: C_1e^{(a-2)t}+C_2e^{(a+2)t}C1e(a2)t+C2e(a+2)t.
  • For the inhomogeneous equation, the problem arises if LaTeX: a=\pm 2a=±2, for then the RHS solves the homogeneous equation. We therefore split up into those two cases (see next)

 

The inhomogeneous equation and y: case LaTeX: |a|\neq 2.|a|2.

In this case, we have LaTeX: x(t)=C_1e^{(a-2)t}+C_2e^{(a+2)t}+dx(t)=C1e(a2)t+C2e(a+2)t+d, with the constant d fit to the right-hand side: inserting into the equation, that gives LaTeX: 0-2a0+(a^2-4)d=2\beta-a\alpha02a0+(a24)d=2βaα so LaTeX: x^*\equiv d=\frac{2\beta-a\alpha}{a^2-4}xd=2βaαa24and LaTeX: \displaystyle x(t)=C_1e^{(a-2)t}+C_2e^{(a+2)t}+\frac{2\beta-a\alpha}{a^2-4}x(t)=C1e(a2)t+C2e(a+2)t+2βaαa24.

To get y, we have LaTeX: y=\tfrac12\big(\dot x-ax-\alpha\big)y=12(˙xaxα)
[Here you just do the rest.]

 

The inhomogeneous equation and y: case |a|=2

Here we must try some first-order ct for x*. (Why not ct+d? No problem, but we already have an arbitrary constant, so d is redundant.)
Inserting x* = ct into the equation, we get LaTeX: -2ac = 2\beta-a\alpha2ac=2βaα and LaTeX: c = \frac{a\alpha-2\beta}{2a}=\frac\alpha2-\frac\beta ac=aα2β2a=α2βa.
Then we have LaTeX: \displaystyle x(t)=C_1e^{(a-2)t}+C_2e^{(a+2)t}+\big(\frac\alpha2-\frac\beta a\big)tx(t)=C1e(a2)t+C2e(a+2)t+(α2βa)t.

  • [Here you can split into the two cases a=-2 or a=2, and insert into LaTeX: y=\tfrac12\big(\dot x-ax-\alpha\big)y=12(˙xaxα)to get the answer on page 36.]
  • [Alternatively, if you want to be a bit fancy: Either a=2 so a-2=0 and a+2=2a, or a=-2 so a-2=2a and a+2=0; in either case, one root is zero and the other is 2a, so you can write x as LaTeX: \displaystyle x(t)=D+Ce^{2at}+\Big(\frac\alpha2-\frac\beta a\Big)tx(t)=D+Ce2at+(α2βa)t and take it from there.]

 

Part (c) - the phase diagram for the special case LaTeX: \begin{align}\dot x&= -x+2y-4\\
\dot y&= 2x-y-1\end{align}˙x=x+2y4˙y=2xy1; 

Download More sketches for the phase diagram here.

A bit of theory first: the associated matrix has determinant (-1)*(-1)-2*2=-3<0. Negative determinant => saddle point! 
So what is the stationary state [a.k.a. "equilibrium point"]? It satisfies 
-x+2y=4
2x-y=1
Scale the second by 2 and add up to get 3x=6 so x=2; then y=2x-1=4-1=3. So the system is equivalent to 
LaTeX: \frac d{dt}\begin{pmatrix}x-2\\y-3\end{pmatrix}=\begin{pmatrix}-1&2\\2&-1\end{pmatrix}\begin{pmatrix}x-2\\y-3\end{pmatrix}ddt(x2y3)=(1221)(x2y3)
which apart from moving the origin two steps horizontally and three steps vertically, behaves the same as LaTeX: \begin{pmatrix}\dot x\\\dot y\end{pmatrix}=\begin{pmatrix}-1&2\\2&-1\end{pmatrix}\begin{pmatrix}x\\y\end{pmatrix}(˙x˙y)=(1221)(xy)

 

Null-isoclines:

  • To get the null-isocline for x: LaTeX: \dot x=0 ˙x=0 when 2y=x+4 i.e. y=2+x/2. Across that line, there is no left--right motion; the paths cross that line vertically.
  • To get the null-isocline for y: LaTeX: \dot y=0˙y=0 when y=2x-1. Across that line, there is no up--down motion; the paths cross that line horizontally.

Directions: Null-isoclines solve an equality. You are interested in inequalities:

  • LaTeX: \dot x>0˙x>0 (i.e. rightward motion - that does not mean "straight to the right", it can be NE or SE, but not W):
    When 2y-x-4>0. That is, when y>2+x/2, that is, above that null-isocline. So above that line, there is motion in the rightward direction. 
    Likewise, below that line, LaTeX: \dot x=2y-x-4˙x=2yx4 is < 0 and we have motion to the left.
  • LaTeX: \dot y>0˙y>0 (i.e. upwards motion - that means NE or NW or possibly straight N): 
    When 2x-y-1>0. That is, when y<2x-1, that is, below that null-isocline. So below that line, there is upward motion. 
    Likewise, above that line, LaTeX: \dot y˙y is < 0 and we have motion downwards. 

Saddle path: 

This is probably not needed for this question - but you could have been asked for the convergent non-constant path, so I refresh that theory. Above, we saw that the associated matrix has eigenvalues a plus/minus 2, and here, a=-1. A negative and a positive eigenvalue (that gives the saddle property), and the negative eigenvalue is -1. An associated eigenvector solves
(-1-(-3))x+2y=0 (exercise: why? And why only one equation?), that is, x+y=0 and the eigenvector has a slope of -1 (-45 degrees).