Dynamical Systems: Part 2

Bifurcation Theory

In practical applications that involve dierential equations it very often happens that the dierential equation contains parameters and the value of these parameters are often only known approximately. In particular they are generally determined by measurements which are not exact. For that reason it is important to study the behavior of solutions and examine their dependence on the parameters. This study leads to the area referred to as bifurcation

42

theory. It can happen that a slight variation in a parameter can have signicant impact on the solution. Bifurcation theory is a very deep and complicated area involving lots of current

1.4 Phase Line and Bifurcation Diag

Technical publications may use special diagrams to display qu research. A complete examination of of the eld would be impossible. information about the equilibrium points of the dierential e
A xed point (or equilibrium point) of a dierential equation y = f (y) is a root of the equation f (y) = 0. As we have already seenfor autonomous problems xed points can be

(1)

y = f (y).

This equation of independent of x, hence there are no externa very useful in determining the long time behavior is solutions.
Qualitative information about the equilibrium points of the dierential equation y = f (y) can be obtained from special diagrams called phase diagrams.

terms that depend on x. Due to the lack of external controls, tion is said to be self-governing or autonomous.

A phase line diagram for the autonomous equation y sink, source segment with labels each root of segment with labels = f (y) is a line or node, one for for so-called sinks, sources or nodes, one forequilibrium;f see = 0, i.e. each equilibrium. i.e. each each root of (y) Figure 11.

A phase line diagram for the autonomous equation y = f (y

source y0

sink y1

Figure 11. A phase line diagram autonomous equation y = f (y).

The labels are borrowed are dened as follows: The names are borrowed from the theory of uids and they from the theory of uids, and they following special denitions:6 Sink y = y0
1. Sink An equilibrium y0 which attracts nearby solutions at t = , i.e., there exists M > 0 so that if |y(0) y0 | < M , then |y(x) y0 | 0

The equilibrium y = y0 attracts nearby solu t x = : for some H > 0, |y(0) y0 | < H |y(x) y | decreases to 0 as x . 2. Source An equilibrium y1 which repels nearby solutions at 0t = , i.e., here exists Source y = y1 The equilibrium y = y1 repels nearby solut M > 0 so that if |y(0) y1 | < M , then |y(x) y1 | increases as t . x = : for some H > 0, |y(0) y1 | < H that |y(x) y1 | increases as x . 3. Node An equilibrium y2 which is neither a sink or a source. In uids, sink means uid Node y = y2 The equilibrium y = y2 is neither a sink nor a
is lost and source means uid is created.

In uids, sink means uid is lost and source means uid is cr 1 memory device for these concepts is the kitchen sink, wherein t is the source and the drain is the sink. The stability test

Stability Test: The term stable means that solutions that start near the equilibrium will stay nearby as t . The term unstable means not stable. Therefore, a sink is stable and a source is unstable. Precisely, an equilibrium y0 is stable provided for given > 0 there

exists some > 0 such that |y(0) y0 | < implies y(t) exists for t 0 and |y(t)?y0 | < . Theorem 2.1 (Stability Conditions). Let f and f be continuous. The equation y = f (y) has a sink at y = y0 provided f (y0 ) = 0 and f (y0) < 0. An equilibrium y = y1 is a source provided f (y1 ) = 0 and f (y1 ) > 0. There is no test when f is zero at an equilibrium. Our objective in this section (for rst order equations) is to briey examine the three simplest types of bifurcations: 1) Saddle Node; 2) Transcritical; 3) Pitchfork .

2.1

Saddle Bode Bifurcation

We begin with the Saddle Node bifurcation (also called the blue sky bifurcation) corresponding to the creation and destruction of xed points. The normal form for this type of bifurcation is given by the example

x = r + x2

(1)

The three cases of r < 0, r = 0 and r > 0 give very dierent structure for the solutions.

r<0

r=0

r>0

We observe that there is a bifurcation at r = 0. For r < 0 there are two xed points given by x = r. The equilibrium x = r is stable, i.e., solutions beginning near this equilibrium converge to it as time increases. Further, initial conditions near r diverge from it. 2

At r = 0 there is a single xed point at x = 0 and initial conditions less than zero give solutions that converge to zero while positive initial conditions give solutions that increase without bound. Finally if r > 0 there are no xed points at all. For any initial condition solutions increase without bound. There are several ways we depict this type of bifurcation one of which is the so called bifurcation diagram.

Note that if instead we consider x = r x2 the the so-called phase line can be drawn as

r<0

r=0

r>0

Exercise: Analyze the bifurcation properties of the following following problems. 1. x = 1 + rx + x2 2. x = r cosh(x) 3. x = r + x ln(1 + x)

2.2

Transcritical Bifurcation

Next we consider the transcritical bifurcation corresponding to the exchange of stability of xed points. The normal form for this type of bifurcation is given by the example

x = rx x2

(2)

In this case there is either one (r = 0) or two (r = 0) xed points. When r = 0 the only xed point is x = 0 which is semi-stable (i.e., stable from the right and unstable from the left). For r = 0 there are two xed points given by x = 0 and x = r. So we note in this case x = 0 is a xed point for all r. For r < 0 the nonzero xed point is unstable but for r > 0 the nonzero xed point becomes stable. Thus we say that the stability of this xed point has switched from unstable to stable.

r<0

r=0

r>0

Bifurcation diagram for a transcritical bifurcation.

Exercise: Analyze the bifurcation properties of the following following problems. 1. x = rx + x2 2. x = rx ln(1 + x) 4

3. x = x rx(1 x)

2.3

Pitchfork Bifurcation

Finally we consider the pitchfork bifurcation. The normal form for this type of bifurcation is given by the example

x = rx x3

(3)

The cases of r 0 and r > 0, once again, give very dierent structure for the solutions.

r<0
x

r=0

r>0

Super Critical Pitchfork Bifurcation Diagram Now consider the example x = rx + x3 . (4)

For this example we obtain the so-called sub-critical pitchfork bifurcation. Notice that solutions blow-up in nite time, i.e., satisfy x(t) as t a < .

Sub Critical Pitchfork Bifurcation Diagram Exercise: Analyze the bifurcation properties of the following following problems. 1. x = x + tanh(x) 2. x = rx 4x3 3. x = rx sin(x) 4. x = rx + 4x3 5. x = rx sinh(x) 6. x = rx 4x3 7. x = x + rx 1 + x2

2.4

Hysteresis: a more complicated bifurcation

In this subsection we consider an even more complicated example which contains pitchforkand saddle node bifurcations. Consider the example x = rx + x3 x5 . (5)

1. For small initial conditions the bifurcation diagram looks just like the sub-critical bifurcation diagram. The origin is locally stable for r < 0 and the two branches are unstable. The two backward unstable branches bifurcated from r = 0. The term x5 6

has now created a new phenomenon: at a value of r < 0, denoted by r , the unstable branches turn around around and become stable. These new branches exist for all r > r
Sub Critical Pitchfork Bifurcation Diagram Exercise: Analyze the bifurcation properties of the following following problems.

2. Note that for r < r < 0 there are three stable solutions. The initial condition determines which of these three xed points the solution converges to as time increases.
2. x = rx 4x3 1. x = x + tanh(x)

3. This example demonstrates an important physically observed phenomenon known as Hysteresis. If we start the system with an initial condition close to x = 0 r
3.9.4 Hysteresis: a more complicated bifurcation In this subsection we consider an even more complicated example which contains pitchforkand x saddle node bifurcations. Consider the example x = rx + x3 x5 .
r

3. x = rx sin(x)

(14)

33 Bifurcation Diagram showing Hysteresis