6.2 Pointcaré Map and Periodic Stability

Nonlinear Stability of Periodic Orbits

If we have a T-periodic orbit γ(t)={ϕt(x0),0tT} where T is the minimal-period.
We have the distance to the orbit:

dist(x,γ)=min0sT{|xγ(s)|0}

Orbitally L-Stable

ε>0,δ>0

dist(y0,γ)<δt0,dist(ϕt(y0),γ)<ε

If not LStable then it is unstable

Orbitally ωAttracting

δ>0 s.t

dist(y0,γ)<δlimtdist(ϕt(y0),γ)=0

Orbitally Asymptotically Stability

if Lstable and ωattracting

Poincaré Map

Also called the first return map.
Assume we have the system x˙=f(x) for xRn. Assume we have xΣ where:

For xΣ τ(x)=min{t:ϕt(x)Σ}. We call the cross-section Σ a Poincaré section for the flow and the map P(x)=ϕτ(x)(x) the Poincaré map.

Example: Poincaré Map for Harmonic Oscillator

First consider the most simple example

x˙=y;y˙=x.

Consider the point (x,y)=(1,0), and the hyperplane

Σ={(x,y),y=0}.

Then the Poincaré map for a point near (1,0) can be calculated in the following way: In polar coordinates, the system reads

r˙=0;θ˙=1.

The time to return to Σ for points near (1,0) is 2π. Moreover θ(2π)=2π, one has that

x(2π)=r(2π)cosθ(2π)=r,y(2π)=r(2π)sinθ(2π)=0.

Then the Poincaré map for a point near (1,0) is P(x,0)=(r(2π),0)=(x,0), which is the identity map.

Example: Poincaré Map for Anharmonic Oscillator

First, consider the system:

x˙=y+x(1x2y2)y˙=x+y(1x2y2)

Consider the point (x,y)=(1,0) and the cross-section (the line where we measure the return):

Σ={(x,y),y=0}

To find the Poincaré map for a point near (1,0), we switch to polar coordinates (x=rcosθ,y=rsinθ). The system simplifies to:

r˙=r(1r2)θ˙=1

Let (x,0)=(r(0),0)Σ be a starting point near (1,0).

From θ˙=1, we know that θ(t)=t. Therefore, the time T required to make one full rotation and return to the cross-section Σ is 2π (since the angle goes from 0 to 2π).

So, we need to find the value of the radius r(t) at time t=2π.

x(2π)=r(2π)cos(θ(2π))=r(2π)y(2π)=r(2π)sin(θ(2π))=0

so at 2π we are back on the map.

Step-by-Step Derivation of the Integral

We need to solve the differential equation for r:

drdt=r(1r2)
  1. Separation of Variables

Move all r terms to the left and dt to the right:

1r(1r2)dr=dt
  1. Setting up the Definite Integral

We integrate from the start time t=0 to the return time t=2π.

r(0)r(2π)1r(1r2)dr=02πdt

The right side is simple: 02πdt=2π.

The left side requires Partial Fraction Decomposition.

  1. The Decomposition

We want to break 1r(1r2) into simpler fractions. Note that 1r2=(1r)(1+r).

The decomposition is:

1r(1r2)=Ar+B1r+C1+r

Solving for the constants (A=1,B=1/2,C=1/2), we get:

1r(1r2)=1r+12(1r)+12(1+r)
  1. Performing the Integration

Now we integrate term by term:

(1r+12(1r)+12(1+r))dr=ln|r|12ln|1r|+12ln|1+r|=ln|r|+12ln|1+r1r|

The text simplifies the integral result into a combined log form:

12ln(r21r2)

(Checking the derivative of this expression confirms it matches the original integrand).

  1. Evaluating the Bounds

We evaluate [12ln(r21r2)] from r(0) to r(2π):

12ln(r2(2π)1r2(2π))12ln(r2(0)1r2(0))=2π

Multiply the entire equation by 2 to clear the fraction:

ln(r2(2π)1r2(2π))ln(r2(0)1r2(0))=4π

Solving for the Poincaré Map

We want to find the mapping P(x)=r(2π) based on the input x=r(0).

Exponentiate both sides to remove the natural logs:

r2(2π)1r2(2π)r2(0)1r2(0)=e4π

Let u=r2(2π) (the unknown output squared) and v=r2(0) (the known input squared).

u1u=e4π(v1v)

Now, solve for u:

u=e4π(v1v)(1u)u(1+e4πv1v)=e4πv1v

After simplifying the algebra to isolate u (which corresponds to r2(2π)), and taking the square root, we arrive at the result shown in the image.

As a result:

P((x,0))=r(2π)=((1+e4π(x21))1/2,0),x>0

Characteristic Multipliers

Let x be a point on γ(t) and Σ be a Poincaré section containing x with the Pincaré map denoted by P(x). The eigenvalues of the Jacobian DP(x) is called the characteristic multipliers.

if every |λi|<1 then γ(t) is orbitally asymptotically stable
if there exists |λi|>1 then γ(t) is not obritally Lyapunov stable.