Transcript 投影片 1
II. Towards a Theory of Nonlinear Dynamics & Chaos
3.
4.
5.
6.
7.
8.
Dynamics in State Space: 1- & 2- D 3-D State Space & Chaos Iterated Maps Quasi-Periodicity & Chaos Intermittency & Crises Hamiltonian Systems
3.
Dynamics in State Space: 1- & 2- D
Concepts to be introduced: State space / Phase Space H. Poincare J.W. Gibbs Fixed points ( equilibium / stationary / critical / singular ) points Limit Cycles Stability (attractor) / Instability (repellor) Bifurcations : Change of stability / Birth of f.p. or l.c.
State Space
Degrees of freedom : 1. Classical mechanics ( phase space ) : number of (q,p) pairs.
2. Dynamical systems ( state space ) : number of independent variables.
Spring obeying Hooke ’s law :
x
t
x
sin
t m x
k x
Cycle : Closed periodic trajectory
k m
Systems of 1
st
Order ODEs
u
u i
f i
Autonomous
i
1, ,
n
DoF = n
u
u i
f i
Non Autonomous Dimension of state space = number of 1st order autonomous ODEs.
N-DoF non autonomous → (N+1)-DoF autonomous
u i
f i i
1, ,
n
1 Autonomous
u n
1
u n
1
t
f n
1 1 Non-crossing theorem is applicable only to autonomous systems
One n th order ODE ~ n 1 st order ODEs Mass spring: 2 2
u
f u
u u
1 , 2
f
f f
1 , 2
y
, 2
x
u
2 , 2
u
1 2 nd order ODE Two 1 st order ODEs
u
1
u
2 2 2 1
Given
u
u* is a fixed point if
0
Caution : Autonomous version of a non-autonomous system requires special treatment [ u n+1 = 1 0 ].
All dynamical systems can be converted to a set of 1st order ODEs.
For some systems this requires DoF = ∞, e.g., • PDEs • integral – differential eqs • memory eqs If the system is dissipative, only a few DoFs will remain active eventually.
No-Intersection Theorem
• A state space trajectory cannot cross itelf.
• 2 distinct state space trajectories cannot intersect in a finite amount of time.
Physical implication : Determinism Mathematical origin : Uniqueness solutions of ODE that satisfy the Lipschitz condition (f bounded).
Apparent violations: • Asymptotic intersects.
• Projections
Dissipative Systems & Attractors
• Transients not important in dissipative systems ( long time final states independent of IC ) • Attractor : Region of state space to which some trajectories converge.
• Basin of an attractor : Region of state space through which all trajectories converge to that attractor.
• Separatrix : Boundary between the basins of two different attractors.
•Miscellaneous: –Fractal basin boundaries.
–Riddled basins of attraction.
–Dimension of the state space.
Evolution eq. : Fixed point :
X
1-D State Space
X
Types of fixed points in 1-D state spaces: • Nodes / sinks / stable fixed points • Repellors / sources / unstable fixed points • Saddle points
Type Determination
Let X 0 be a fixed point:
X
0
d f d X
= characteristic value ( eigenvalue ) of X 0 0 For λ > 0 For λ < 0
X
0
X
0
X
0
for X
X
0
X
0
X
0
X
0
for X
X
0
X
0
X
0
X
0
X
0
λ = 0 2
d f d X
2 0 0 0
for for X X
X
0
X
0
d X
2 0 0 0
for for X X
X
0
X
0
λ = 0 2
d f d X
2 0 0
X
0
X
0
X
0
X
0
for X
X
0
X
0 Repells Attracts
Saddle points
λ = 0
d X
2 0 0
X
0
for X
X
0
X
0
X
0
X
0
X
0 Attracts Repells Convex Concave
Structural Instability
Saddle point is structurally unstable
Taylor series Expansion
X 0 = fixed point 0
X
X
0
d f d X
0 1 2
X
X
0 2 2
d f d X
2 0
X
X
X
0
X
0
X
x e
t
d f d X
0 0
X
0
e
t
X
0 Lyapunov exponent > 0 X 0 repellor < 0 X 0 node = 0 X 0 s.p. / node / rep.
Trajectories in 1-D State Space
local behavior Global behavior determined by matching fixed point basins.
→ joining arrows pointing toward (away) from nodes (repellors).
f continuous neighboring fixed points cannot be • both nodes or both repellors • saddle points of different types Exercises 3.8- 3,4
Bounded systems
: Outermost fixed points must be • nodes or • • type I saddle point on the left type II saddle point on the right.
A node must be on the repelling side of a saddle point.
When Is A System Dissipative ?
Defining characteristics of a dissipative system : Motion reduced asymptotically to a few active DoFs. Cluster of ICs ( those that lead to fixed points excluded ).
C.f., statistical ensembles.
d L
d t d d t
X B
X A
d L Ld t d f d X A X B
X A
B
A d f d X A
X B
X A
System is dissipative near X A if df/dX < 0 .
e.g., near a node.
Divergence theorem
X
1
X
2
f
1
f
2
X X
1 , 2
X X
1 , 2
2-D State Space
0 0
f
1
f
2
X
10 ,
X
10 ,
X
20
X
20
X
1
f
1
X
1 0
X
1
X
10
f
1
X
2 0
X
2
X
20
X
2
f
2
X
1 0
X
1
X
10
f
2
X
2 0
X
2
X
20
x
1
x
2
f
1 , 1
f
2 , 1
x
1
x
1
f
1 , 2
f
2 , 2
x
2
x
2
x
f x
x j
X j
X j
0
f i
,
j
f i
x j
0 Fixed point
λ 1 < 0 λ 2 < 0 λ 1 > 0 λ 2 > 0
Special case
f
1 , 1
f
2 , 1 1 0
x
1
x
2 1 2
x
1
x
2
f
1 , 2
f
2 , 2 0 2 λ 1 > 0 λ 2 < 0 λ 1 < 0 λ 2 > 0
y
Saddle point at (x,y) = (-π/2 , 0 ) 1 0.5
0 -0.5
-1
Hyperbolic point : λ 0 • • • Invariant manifold: all trajectories along the principal axes of a hyperbolic saddle point.
Stable (invariant) manifold ( in-set ): Trajectories heading towards the hyperbolic saddle point.
Unstable (invariant) manifold ( out-set ): Trajectories heading away from the hyperbolic saddle point.
In-sets & outsets serves as separatrices.
1-D saddle point are non-hyperbolic since λ = 0
→ Fixed points:
Brusselator
X
B
1
X
2
X Y Y
BX
2
X Y X A X A
X
0
BX
2
X Y
0
X
0 ,
Y
0
A
,
B A
A,B > 0
x
1
x
2
f x
11 1
f x
21 1
f x
12 2
f x
22 2 General 2-D Case
x
1
f x
11 1
f x
12 2
f x
11 1
f
12
f x
21 1
f x
22 2
f x
11 1
f
11
f
12
f x
21 1
f
22
f
22
x
1
f f
12 21
x
1
f
12
f x
11 1
f f
11 22
x
1
x
1
Ce
t
→ 2
f
11
f
22
f f
11 22
f f
12 21 0 1 2
f
11
f
22
f
11
f
22 2 4
f f
11 22
f f
12 21 1 2
f
11
f
22
f
11
f
22 2 4
f f
12 21
x
1
t
t
State variables can be any pair from
x
2
B e
t
x x x x
1 , 1 , 2 , 2
t
Ex 3.11
Complex Characteristic Values
1 2
f
11
f
22
f
11
f
22 2 4
i
21
R
1 2
f
11
f
22 Re
f
11
f
22 2 4
f f
12 21 Im
f
11
f
22
2 4 21 Note :
f
11
f
22
2 4 21 is either real or purely imaginary
x
1
e R t C e i
t
x
2
e R t B e i
t
t t
Spirals, inward if R < 0 ( focus ) outward if R > 0 Limit cycle if R = 0
C
C
A
2
B
B
A
2
i x
1
Ae R t
cos
t x
2
Ae R t
sin
t
R < 0 R > 0
Dissipation & Divergence Theorem 2-D state space Area :
A
X
1
C
X
1
B
X
2
C
X
2
B
d A
d t f
1
X
1
C
,
X
2
B
f
1
X
1
B
,
X
2
B
X
2
C
X
2
B f
1
f
2
X
1
C
,
X
1
B
,
X
2
B X
2
C
f
1
f
2
X
1
B
,
X
1
B
,
X
2
B X
2
B X
1
C
X
1
B
f
2
X
1
B
,
X
2
C
X
1
C X
2
C
X
1
B X
2
B
f X
1 1
f
2
X
2
X
1
B
,
X
2
B
X
1
B
,
X
2
B
f
2
X
1
B
,
X
2
B
d A
d t f
11
X
1
B
,
X
2
B
X
1
C
X
1
B
X
2
C
X
2
B
1
d A
A d t f
11
f
22
f
f
< 0 dissipative
X
1
C
X
1
B
X
2
C
X
2
B
f
22
X
1
B
,
X
2
B
1
dV V d t
j N
1
f j j
f
t
0
X i
f i
Jacobian Matrix at Fixed Point
X i
X i
0
x i
→
x i
j N
1
f x i j j x i
Ce
t
→
x j
j N
1
f x i j j
→
J
x
x
J
x f i j
f
11
f N
1
f
1
N
f NN
Jacobian matrix
f
11
f N
1
f
1
N f NN
0
Tr f
i N
1
f i i
i N
1
i
det
f
i N
1
i
2-D State Space 1 2
f
11
f
22
f
11
f
22 2 4
f f
11 22
f f
12 21 1 2
Tr J
Tr J
2 4
det J
1 2
Tr J
Δ < 0 Δ > 0 det J > 0 Δ > 0 det J det J < 0 Tr J < 0 Both λ complex, Real part negative Spiral node Both λ real & negative node Both λ real & of of opposite signs, Saddle point Tr J > 0 Both λ complex, Real part positive Spiral repellor Both λ real & positive repellor Both λ real & of of opposite signs, Saddle point
Example: The Brusselator
X Y
BX
B
1
X
X
0 ,
Y
0
A
,
B A J
B
B
1
A
2
A
2
Tr J B
det
J
A
2
A
2 1 2
B A
2
B
Set A = 1 & let B be control parameter :
A
2
2 4
A
2 1 2
B
• B < 2, spiral node • 2 < B < 4, spiral repellor (converge to another limit cycle) • B > 4, do exercise 3.14-2.
B
2 2 4
Limit Cycles
Limit cycle : closed loop in state space to (from) which nearby trajectories are attracted (repelled). vortex Invariant set : region in state space where a trajectory starting in it will remain there forever.
Poincare-Bendixson theorem : Let R be a finite invariant set in a 2-D state space, then any trajectory in it must, as t → ∞, approach a 1.
2.
fixed point , or limit cycle.
Implications : • no chaos in 2-D systems.
• limit cycle in Brusselator.
Delayed DE
u
: DoF = : IC for t [-T,0] needed Topology : Poincare index theorem
Poincare Sections
Poincare section in n-D state space: An (n-1)-D hyper-surface that cuts through the trajectory of a n-D continuous flow and reduces it to a (n-1)-D discrete map.
Example : Limit cycle in 2-D state space
Poincare Map
P n
1 Exercise 3.16-1 Let Fixed point of F : Near P*:
d n
P n
P
*
P
*
P
2
P
* →
d
2
d F d P P
*
d
1
Md
1
d F d P P
*
P
1
P
*
M
d F d P P
* = (characteristic / Floquet / Lyapunov) multiplier
d n
Md n
1 →
d n
M n
1
d
1 Characteristic exponent ln
M
M < 1 M > 1 M = 1 Attracting Repelling Saddle (rare in 2-D )
Bifurcation Theory
Study of changes in the character of fixed points.
( limit cycles are fixed points in Poincare sections ) Appendix B 2 types of bifurcation diagrams: • control parameter vs location of fixed point.
• control parameter vs characteristic value.
x
1
x
a
Bifurcations in 1-D
y
y
Normal form δ> 0 repellor δ< 0 node Bifurcation at δ= 0
x x
2
d f d x
2
x
No Fixed points if μ< 0.
2 fixed points for μ> 0
x
* node
x
* repellor
d x
2 2 -4 -2 -2 -4 -6 -8 -10 4 2 • For μ= 0, x* = 0 is a saddle point. • For μ> 0, x* = ± μ form repellor-node pair.
• μ= 0 is repellor-node bifurcation point.
• Other names: saddle-node / tangent / fold bifurcation Note that for μ= 0, x* = 0 is structurally unstable.
2 4
Lifted (Suspended) State Space
Flow along extra dimension X 2 always towards original axis X 1 .
x
1
x
1 2
x
2
x
2 Repellor ↓ Saddle point Node ↓ Node No fixed point
Bifurcation in 2-D The Brusselator 1 2
B
B
2 2 4 B > 4, λ ± real (Transcritical) bifurcation at B = 2.
Node → Repellor + limit cycle 4 B 0, λ ± complex
Normal Form Equations
Fixed point at
x
=
0
. Bifurcation at μ = 0.
Saddle – node bifurcation :
x x
2 1
x
2
x
1 2 μ> 0 : node at saddle at μ< 0 : no fixed point
x x
1 , 2
x x
1 , 2 ,0 ,0 μ= 0 : bifurcation
Transcritical bifurcation :
x
x
y
y
x
2 fixed points switching types of stability Pitchfork bifurcation :
x y
x
y
x
2 1 → 3 fixed points
Limit Cycle Bifurcations
Spiral-in-spiral-out bifurcation at Re( λ) = 0 Hopf bifurcation : birth of stable limit cycle Poincare section of limit cycle in 2 D → 1-D dynamics Normal form:
x
1 2
x
1
x
2
x
2
x
1 2
x
1 2
x
2 2
x
2 2 Polar coordinates:
r
r
r
2 1
r
x
1 2
x
2 2 tan 1
x
2
x
1 0
t
r
r
2
f
3
r
2 Fixed points: for μ< 0 r* = 0 spiral node for μ> 0 r* = 0 spiral repellor
r
* limit cycle, period = 2π Hopf bifurcation at μ= 0 Limit cycle: asymptotic time-dependent behavior of dissipative system.