invariant manifolds
Stable Manifold Theorem
You should know: fixed points stability, hamiltonian systems
Overview
The stable manifold theorem (Hadamard-Perron theorem) describes the geometry of trajectories near a hyperbolic fixed point. At such a point, the phase space locally decomposes into three invariant manifolds: the stable manifold (trajectories approaching x* exponentially as t -> +inf), the unstable manifold (trajectories approaching x* exponentially as t -> -inf), and the center manifold (corresponding to eigenvalues with zero real part, where stability requires further nonlinear analysis). The Hartman-Grobman theorem says the local flow is topologically equivalent to the linearization; the stable manifold theorem adds that the invariant manifolds exist as smooth submanifolds tangent to the eigenspaces of the Jacobian. Stable and unstable manifolds organize all trajectories in phase space, and their intersections (homoclinic/heteroclinic orbits) are the seeds of chaos.
Intuition
Near a saddle point, some directions attract and others repel. The stable manifold is the set of all points that eventually flow into the saddle -- it is a surface (in 3D) or curve (in 2D) threading through the saddle. The unstable manifold is the set that flows out of the saddle. These are like the two rails of a train track intersecting at the saddle: one rail leads to the station (stable manifold), the other leaves it (unstable manifold). When a stable manifold of one saddle intersects the unstable manifold of another (or itself), a heteroclinic (or homoclinic) orbit forms, and Poincare showed this creates infinitely many periodic orbits nearby and sensitive dependence -- the geometry of chaos.
Formal Definition
Let x* be a hyperbolic fixed point of x' = F(x), with Jacobian J = DF(x*) having eigenvalues with real parts negative (E^s), positive (E^u), or zero (E^c).
Notation
| Notation | Meaning |
|---|---|
| Stable manifold of x*: trajectories approaching x* forward in time | |
| Unstable manifold of x*: trajectories approaching x* backward in time | |
| Center manifold tangent to eigenspace of eigenvalues with zero real part | |
| Stable, unstable, and center eigenspaces of DF(x*) |
Theorems
Worked Examples
- 1
The fixed point is (0,0). Jacobian J = [[1,0],[0,-1]] with eigenvalues lambda_1=1 (unstable, eigenvector (1,0)) and lambda_2=-1 (stable, eigenvector (0,1)).
- 2
Stable eigenspace E^s: span{(0,1)} = the y-axis. Unstable eigenspace E^u: span{(1,0)} = the x-axis.
- 3
Since the system is linear, stable and unstable manifolds equal the eigenspaces exactly: W^s = {x=0} (y-axis), W^u = {y=0} (x-axis).
- 4
Verification: on W^s: x(t)=0*e^t=0, y(t)=y_0*e^{-t}->0 as t->+inf. On W^u: x(t)->inf as t->+inf, but x(t)->0 as t->-inf (backward time).
✓ Answer
W^s(0,0) is the y-axis {x=0} and W^u(0,0) is the x-axis {y=0}. Trajectories starting off these axes form hyperbolas xy = const.
Practice Problems
Explain what a homoclinic orbit is and why its transverse intersection implies chaos via the Smale-Birkhoff theorem.
What is the purpose of the center manifold theorem, and give an example of how it reduces stability analysis to a lower-dimensional problem.
Common Mistakes
Thinking stable and unstable manifolds are flat (equal to the eigenspaces globally).
The eigenspaces E^s and E^u are only the tangent spaces to W^s and W^u AT the fixed point. The actual manifolds are curved nonlinear surfaces that agree with the eigenspaces only to first order.
Applying the Hartman-Grobman theorem when eigenvalues have zero real part.
Hartman-Grobman requires hyperbolicity (no eigenvalue with zero real part). When center eigenvalues exist, the nonlinear flow is NOT topologically equivalent to the linear flow; center manifold theory is required.
Confusing heteroclinic and homoclinic orbits.
A homoclinic orbit connects a saddle to itself (W^s(x*) intersects W^u(x*)). A heteroclinic orbit connects two different fixed points (W^u(x1) intersects W^s(x2)).
Assuming the center manifold is unique.
The center manifold is generally NOT unique (unlike stable and unstable manifolds, which are unique). Different center manifolds correspond to different Taylor expansions but agree to finite order.
Quiz
Historical Background
Poincare first recognized the importance of stable and unstable manifolds (separatrices) in the late 19th century; he showed that their transverse intersection (homoclinic points) leads to complex, chaotic behavior in the three-body problem. Hadamard (1901) and Perron (1928) independently proved the existence of stable and unstable manifolds for hyperbolic fixed points, giving the result its two names. The center manifold theorem was proved by Pliss (1964) and Kelley (1967), completing the local decomposition. Smale's horseshoe construction (1960s) explicitly showed how homoclinic orbits give rise to chaotic invariant sets, connecting the geometry of stable/unstable manifolds to Devaney chaos.
- 1892-99
Poincare discovers homoclinic orbits and their connection to chaos in the three-body problem
Henri Poincare
- 1901
Hadamard proves the existence of stable/unstable manifolds for geodesic flows
Jacques Hadamard
- 1928
Perron provides a general proof for hyperbolic fixed points of ODEs
Oskar Perron
- 1964-1967
Pliss and Kelley prove the center manifold theorem
V. A. Pliss, A. Kelley
- 1965
Smale uses horseshoe maps to show transverse homoclinic orbits create chaos
Stephen Smale
Summary
- The stable manifold theorem: near a hyperbolic fixed point, W^s and W^u are smooth submanifolds tangent to the stable/unstable eigenspaces of the Jacobian.
- The center manifold theorem: when zero-real-part eigenvalues exist, stability analysis reduces to the dynamics on the center manifold W^c.
- The Smale-Birkhoff theorem: transverse intersections of W^s and W^u (homoclinic points) imply chaotic dynamics containing a topological horseshoe.
- Stable and unstable manifolds organize global phase space structure; their geometry determines whether chaos, heteroclinic connections, or regular behavior dominates.
References
- BookGuckenheimer, J. and Holmes, P. Nonlinear Oscillations, Dynamical Systems, and Bifurcations of Vector Fields. Springer, 1983. Chapters 1-2.
Mathematics