se
6 Index theory (Strogatz 6.8)
A general vector field f (x) is characterized by a magnitude and a
direction for most x. However, in singular points x∗ where the mag-
nitude of the vector field is zero (fixed points in dynamical systems),
the direction is undefined. The index of a two-dimensional vector field
(in the plane) is an integer that describes global information about
the phase portrait around isolated zeroes (the value of the index is
determined by how the vector field orients around the zeroes).
6.1 Illustration
The index IC of any closed, non-intersecting curve C not passing
through any fixed point is defined as the winding of a vector field
as we traverse one counter-clockwise lap on C:
Here the vector field is f = (ẋ, ẏ). At each point along the curve it
forms an angle ϕ = atan(ẏ/ẋ) (chosen in the correct quadrant) to the
x-axis.
Let ∆ϕ = ϕstart − ϕstop be the angular change as one lap is traversed.
For the Fig. above, the angular change can be visualized by labelling
of all vectors in counter-clockwise order and then parallel transport
them to a common origin (right panel). Traversing the vectors through
the labels in order we see that the angle ϕ has increased by 2π when
we come back to the starting vector.
1
Dynamical systems 2019 [email protected]
For a general closed curve, since the start and end positions are the
same, ∆ϕ = IC2π, where the index of curve C, IC, is an integer.
For the example above IC = ∆ϕ/(2π) = +1.
Remark 1 We only use information of the vector field on the curve
that encloses a black box we know nothing about.
Using the index of the curve, IC, it is possible to deduce information
on the possible content of the black box.
Remark 2 ϕ is undefined at fixed points (where ẋ = ẏ = 0). This
is why C is not allowed to pass fixed points.
6.1.1 Mathematical formulation
When the dynamical system is given in terms of equations
ẋ = f (x, y)
ẏ = g(x, y)
rather than a graphical vector field one can either
• Plot the flow as a vector field
• Use an analytical evaluation:
ẏ g(x, y)
ϕ = atan = atan
ẋ f (x, y)
∂ϕ ∂ϕ g f
dϕ = df + dg = − 2 df + dg
∂f ∂g f + g2 f 2 + g2
∆ϕ 1 1 f dg − gdf
I I
⇒ IC ≡ = dϕ =
2π 2π C 2π C f 2 + g 2
2
Dynamical systems 2019 [email protected]
6.2 Properties of indices
1. A continuous transformation of curve C to new curve C 0 without
passing a fixed point ⇒ IC = IC 0 .
Proof: IC must vary continuously, but since IC can only take
discrete values it must remain constant. (Crossing a fixed point
could change the index because the orientation is undefined at
the fixed point.)
2. A curve C that does not enclose any fixed points has IC = 0.
Proof: Shrink C to a very small curve C 0 in which vectors are
close to parallel due to smoothness of the flow ⇒ IC = IC 0 = 0.
3. Reversing all arrows in vector field (t → −t) leaves index un-
changed.
Proof: All angles change from ϕ to ϕ + π ⇒ Index unchanged.
4. The index of a closed orbit (i.e. the curve C is the closed orbit),
has IC = +1.
Proof: Flow is tangential to the curve at all times ⇒ vector
field rotates once as curve is traversed.
5. Let x∗ be an isolated fixed point. The index of a fixed point I is
equal to IC for any curve that encloses x∗ but does not enclose
any other fixed points.
6. A curve C surrounding n isolated fixed points has index IC =
I1 + I2 + · · · + In.
3
Dynamical systems 2019 [email protected]
6.2.1 Indices of fixed points
The index of a saddle point can be obtained using the method above:
The change in ϕ is −2π, so the index of a saddle point is I = −1. In
this way it is possible to show that all other isolated fixed points of
the types found in linear systems (nodes, spirals, centers, degenerate
nodes, stars) have I = +1. Fixed points with −1 and +1 each cover
one half-plane in the classification diagram from Lecture 3:
Along the line ∆ = 0 the linear system has a line of fixed points
and the index cannot be defined (allowing the discontinuity in I as ∆
passes zero).
Fixed points corresponding to higher-order roots of f = 0 may give
other integer values on I.
4
Dynamical systems 2019 [email protected]
6.2.2 Consequences (in the plane)
• Closed orbits must encircle fixed points whose indices sum to
+1 ⇒ at least one fixed point (not a single saddle) inside any
closed orbit in the phase plane.
Example: Double-well potential revisited:
Closed orbits may surround the centers (+1 each), or surround
both the centers and the saddle (IC = 2 · (+1) − 1 = +1). Tra-
jectories surrounding one center and one saddle can not happen.
The separating homoclinic orbits do not have well defined indices
(the curves cross the fixed point).
• Sometimes index theory can be used to rule out closed orbits in
parts of the phase plane.
Example: Since any closed orbit must encircle fixed points
whose indices sum to +1, it is sometimes possible to rule out
closed orbits, c.f. Strogatz Example 6.8.5 (Rabbits vs. sheep):
5
Dynamical systems 2019 [email protected]
Curves C1 and C2 can be ruled out because the indices of the
encircled fixed points does not sum to +1.
The trajectories connecting the origin to (1, 0) and the origin to
(0, 1) are heteroclinic trajectories, preventing any trajectories to
encircle the fixed points on the axes, ruling out curve C3 (the
flow move vertically along the y-axis in contradiction with the
blue trajectory that cut the y-axis at an angle).
• Index theory provides information about trajectories around
higher-order fixed points (not achievable with linearisation).
• Index theory is important for bifurcation theory to answer which
types of bifurcations are allowed (next lecture).
