An Introduction to theQualitative Theory
of Nonautonomous Dynamical Systems
Martin Rasmussen
Imperial College London
9th Meeting of the
European Study Group on Cardiovascular Oscillations
Lancaster, UK, 10–14 April 2016
11 April 2016
G.D. Birkhoff (1884–1944) Dynamical Systems (1927)
Dynamical systemsas abstractions and generalisations of
ordinary differential equations.
G.D. Birkhoff (1884–1944) Dynamical Systems (1927)
Dynamical systemsas abstractions and generalisations of
ordinary differential equations.
Birkhoff was strongly influenced by the two fathers of theQualitative Theory of Dynamical Systems
A.M. Lyapunov (1857–1918) H. Poincare (1854–1912)
Theory of ordinary differential equations before the era ofLyapunov/Poincare/Birkhoff:
Concentration on finding analytical expressions forsolutions of specific differential equations
Explicit expressions of solutions are unavailablefor the most interesting differential equations!
Numerical integration
Approximate solutions vianumerical schemes such as Euler
and Runge–Kutta method.
Qualitative theory
Study of qualitative propertieswithout having the knowledge of
solutions
Theory of ordinary differential equations before the era ofLyapunov/Poincare/Birkhoff:
Concentration on finding analytical expressions forsolutions of specific differential equations
Explicit expressions of solutions are unavailablefor the most interesting differential equations!
Numerical integration
Approximate solutions vianumerical schemes such as Euler
and Runge–Kutta method.
Qualitative theory
Study of qualitative propertieswithout having the knowledge of
solutions
Theory of ordinary differential equations before the era ofLyapunov/Poincare/Birkhoff:
Concentration on finding analytical expressions forsolutions of specific differential equations
Explicit expressions of solutions are unavailablefor the most interesting differential equations!
Numerical integration
Approximate solutions vianumerical schemes such as Euler
and Runge–Kutta method.
Qualitative theory
Study of qualitative propertieswithout having the knowledge of
solutions
Theory of ordinary differential equations before the era ofLyapunov/Poincare/Birkhoff:
Concentration on finding analytical expressions forsolutions of specific differential equations
Explicit expressions of solutions are unavailablefor the most interesting differential equations!
Numerical integration
Approximate solutions vianumerical schemes such as Euler
and Runge–Kutta method.
Qualitative theory
Study of qualitative propertieswithout having the knowledge of
solutions
Differential Equations Random Systems Control Systems
Nonautonomous Dynamical Systems
Natural Sciences Medicine Financial Markets
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Contents
1 Basic theory
2 Attractor theory
3 Linear theory
4 Nonlinear theory
5 Bifurcation theory
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Process formulationTranslation invarianceInvarianceSkew product flow formulation
Nonautonomous differential equations
An initial value problem is the combination of
• a nonautonomous ordinary differential equation (ODE)
x :=dx
dt= f (t, x) ,
where f : R× Rd → Rd is assumed to be continuous, and
• an initial condition x(t0) = x0 for some fixed t0 ∈ R and x0 ∈ Rd .
Process formulation
Assuming global existence and uniqueness of solutions, the solutions forma mapping
(t, t0, x0) 7→ ϕ(t, t0, x0) ∈ Rd for all (t, t0, x0) ∈ R× R× Rd .
This mapping is called the process induced by the above ODE.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Process formulationTranslation invarianceInvarianceSkew product flow formulation
The initial value and cocycle property
The initial value and cocycle property
A process is a continuous mapping (t, t0, x0) 7→ ϕ(t, t0, x0) ∈ Rd whichsatisfies
initial value condition. ϕ(t0, t0, x0) = x0,
cocycle property. ϕ(t2, t0, x0) = ϕ(t2, t1, ϕ(t1, t0, x0)
).
t0t1 t2
x0
ϕ(t1,t0,x0)
ϕ(t2,t0,x0)
=ϕ(t2,t1,ϕ(t1,t0,x0))
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Process formulationTranslation invarianceInvarianceSkew product flow formulation
The autonomous case
The situation is much simpler in the autonomous case, where x = f (x).
Translation invariance
The process ϕ(t, t0, x0) induced by an autonomous differential equationdoes not depend separately on
initial time t0 and actual time t,
but only on the elapsed time t − t0.
This induces a flow φ(t − t0, x0) := ϕ(t, t0, x0), fulfilling
• initial value condition: φ(0, x0) = x0,
• group property: φ(t + s, x0) = φ(t, φ(s, x0)).
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Process formulationTranslation invarianceInvarianceSkew product flow formulation
The autonomous case
The situation is much simpler in the autonomous case, where x = f (x).
Translation invariance
The process ϕ(t, t0, x0) induced by an autonomous differential equationdoes not depend separately on
initial time t0 and actual time t,
but only on the elapsed time t − t0.
This induces a flow φ(t − t0, x0) := ϕ(t, t0, x0), fulfilling
• initial value condition: φ(0, x0) = x0,
• group property: φ(t + s, x0) = φ(t, φ(s, x0)).
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Process formulationTranslation invarianceInvarianceSkew product flow formulation
Flow versus process
Autonomous ODEx = f (x)
Nonautonomous ODEx = f (t, x)
Flow φ(t, x0)
depends on one parameter(initial state x0)
Process ϕ(t, t0, x0)
depends on two parameters(initial time t0 and initial state x0)
All interesting objects aresubsets of thephase space(x0-space)
All interesting objects aresubsets of the
extended phase space((t0, x0)-space)
We call an arbitrary subset of theextended phase space a
nonautonomous set
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Process formulationTranslation invarianceInvarianceSkew product flow formulation
Invariant sets
The long-term behaviour of a flow is characterised by its invariant sets.
Definition (Invariance)
A subset M ⊂ Rd is called invariant w.r.t. the flow φ : R× Rd → Rd if
φ(t,M) = M for all t ∈ R .
Examples are given by
equilibria periodic orbits or fractals.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Process formulationTranslation invarianceInvarianceSkew product flow formulation
Invariant nonautonomous sets
For a nonautonomous set M ⊂ R× Rd and some time t ∈ R, the t-fiberof M is given by
M(t) :={x ∈ Rd : (t, x) ∈ M
}.
Definition (Invariant nonautonomous set)
A nonautonomous set M ⊂ R× Rd for a process ϕ : R× R× Rd → Rd
is called invariant if
M(t) = ϕ(t, t0,M(t0)
)for all t, t0 ∈ R .
R
Rd
t
M(t)
M
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Process formulationTranslation invarianceInvarianceSkew product flow formulation
A simple example
Consider the nonautonomous differential equation
x = 2tx
which induces the process
φ(t, t0, x0) = x0et2−t20 .
Then the set
M :={
(t, x) : −et2
≤ x ≤ et2}
is an invariant nonautonomous set. t
x
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Process formulationTranslation invarianceInvarianceSkew product flow formulation
A simple example
Consider the nonautonomous differential equation
x = 2tx
which induces the process
φ(t, t0, x0) = x0et2−t20 .
Then the set
M :={
(t, x) : −et2
≤ x ≤ et2}
is an invariant nonautonomous set.
M
t
x
It is exactly given by all solution curves with initial values x0 ∈ [−1, 1]starting at time t0 = 0.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Process formulationTranslation invarianceInvarianceSkew product flow formulation
An interesting observation
If ϕ : R× R× Rd → Rd is the process generated by the nonautonomousdifferential equation
x = f (t, x) ,
then the mapping(t, t0, x0) 7→ (t, ϕ(t, t0, x0))
is a flow!
A process is also a flow
This flow is induced by the (d + 1)-dimensional autonomous differentialequation
t = 1 ,
x = f (t, x) .
Why do we need a nonautonomous theory?
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Process formulationTranslation invarianceInvarianceSkew product flow formulation
An interesting observation
If ϕ : R× R× Rd → Rd is the process generated by the nonautonomousdifferential equation
x = f (t, x) ,
then the mapping(t, t0, x0) 7→ (t, ϕ(t, t0, x0))
is a flow!
A process is also a flow
This flow is induced by the (d + 1)-dimensional autonomous differentialequation
t = 1 ,
x = f (t, x) .
Why do we need a nonautonomous theory?
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Process formulationTranslation invarianceInvarianceSkew product flow formulation
Skew product flows
But we can learn something from the example
t = 1 , x = f (t, x) .
nonautonomous influence
t-system on Rx-system on Rd
The drawback: All solutions of the t-system are unbounded!
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Process formulationTranslation invarianceInvarianceSkew product flow formulation
Skew product flows
But we can learn something from the example
t = 1 , x = f (t, x) .
nonautonomous influence
t-system on Rx-system on Rd
The drawback: All solutions of the t-system are unbounded!
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Process formulationTranslation invarianceInvarianceSkew product flow formulation
Skew product flows
nonautonomous influence
flow θ on a compact set P
x-system on Rd
Definition (Skew product flow)
A skew product flow (θ, ϕ) consists of a flow θ : R× P → P on a baseset P and a cocycle ϕ : R× P × Rd → Rd over θ, i.e., one has
initial value condition: ϕ(0, p, x) = x ,
cocycle property: ϕ(t + s, p, x) = ϕ(t, θ(s, p), ϕ(s, p, x)
).
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Process formulationTranslation invarianceInvarianceSkew product flow formulation
A short summary
Nonautonomous differential equations give rise to processes, whichdepend on two parameters (initial time t0 and initial value x0) incontrast to flows (which only depend on the initial value x0).
The interesting sets in the nonautonomous case are nonautonomoussets, which are subsets of the extended phase space.
An invariant nonautonomous set can be seen as union of solutioncurves.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous attractorsNonautonomous attractorsForward attractorsPullback attractorsExistence of attractors
Attractor theory
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous attractorsNonautonomous attractorsForward attractorsPullback attractorsExistence of attractors
The autonomous case
An attractor describes attraction in forward time.
Definition (Attractor)
Let φ : R× Rd → Rd be a flow. A compactinvariant set A is called global attractor if
dist(φ(t,B),A)→ 0 as t →∞
for all bounded sets B ⊂ Rd .
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous attractorsNonautonomous attractorsForward attractorsPullback attractorsExistence of attractors
The nonautonomous case
Let ϕ : R× R× Rd → Rd be a process, and let A ⊂ R× Rd be aninvariant nonautonomous set with compact fibers. There are essentiallytwo possibilities of attraction in the nonautonomous case.
Forward and pullback attraction
For all bounded sets B ⊂ Rd , one has
dist(ϕ(t, t0,B),A(t)
)→ 0
as t →∞ with t0 fixed
(forward attraction)as t0 → −∞ with t fixed
(pullback attraction).
Definition (Forward and pullback attractor)
In case of forward attraction, the nonautonomous set A is called forwardattractor; and it is called pullback attractor in case of pullback attraction.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous attractorsNonautonomous attractorsForward attractorsPullback attractorsExistence of attractors
The nonautonomous case
Let ϕ : R× R× Rd → Rd be a process, and let A ⊂ R× Rd be aninvariant nonautonomous set with compact fibers. There are essentiallytwo possibilities of attraction in the nonautonomous case.
Forward and pullback attraction
For all bounded sets B ⊂ Rd , one has
dist(ϕ(t, t0,B),A(t)
)→ 0
as t →∞ with t0 fixed
(forward attraction)as t0 → −∞ with t fixed
(pullback attraction).
Definition (Forward and pullback attractor)
In case of forward attraction, the nonautonomous set A is called forwardattractor; and it is called pullback attractor in case of pullback attraction.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous attractorsNonautonomous attractorsForward attractorsPullback attractorsExistence of attractors
Forward attractor
Forward attractor
A forward attractor A fulfills for all x0 ∈ Rd that
dist(ϕ(t, t0, x0),A(t)
)→ 0 as t →∞ with t0 fixed.
t0
x0
t
x ϕ(·, t0, x0)
A
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous attractorsNonautonomous attractorsForward attractorsPullback attractorsExistence of attractors
Pullback attractor
Pullback attractor
A pullback attractor A fulfills for all x0 ∈ Rd that
dist(ϕ(t, t0, x0),A(t)
)→ 0 as t0 → −∞ with t fixed.
t0t0t0 t
x0
ϕ(·, t0, x0)
A(t)
A
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous attractorsNonautonomous attractorsForward attractorsPullback attractorsExistence of attractors
An example
Consider the nonautonomous differential equation
x = 2tx
which induces the process
φ(t, t0, x0) = x0et20−t
2
.
Then
the nonautonomous setR× {0} is a pullbackattractor,
there exists no forwardattractor.
t
x
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous attractorsNonautonomous attractorsForward attractorsPullback attractorsExistence of attractors
An example
Consider the nonautonomous differential equation
x = −2tx
which induces the process
φ(t, t0, x0) = x0et2−t20 .
Then
there exists no pullbackattractor,
every invariantnonautonomous set withcompact fibers is a forwardattractor.
t
x
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous attractorsNonautonomous attractorsForward attractorsPullback attractorsExistence of attractors
Pullback versus forward attraction
The following general statements about pullback and forward attractorshold:
The concepts of pullback and forward attraction are independent ofeach other.
Pullback attraction describes attraction for the past of the system,whereas forward attraction concerns the future of a system.
If a pullback attractor exists, it is unique.
Forward attractors are intrinsically nonunique.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous attractorsNonautonomous attractorsForward attractorsPullback attractorsExistence of attractors
Existence of attractors
The existence of autonomous attractors follows from absorbing sets.
Definition (Absorbing set)
A nonempty compact set D ⊂ Rd is called an absorbing set for a flow φif for every bounded set B ⊂ Rd , there exists a T > 0 such thatφ(t,B) ⊂ D for all t ≥ T .
Theorem (Existence of global attractors)
Suppose that a flow φ has an absorbing set D. Then φ has a uniqueglobal attractor A, given by
A =⋂s≥0
⋃t≥s
φ(t,D) .
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous attractorsNonautonomous attractorsForward attractorsPullback attractorsExistence of attractors
Existence of pullback attractors
Definition (Pullback absorbing set)
Let ϕ be a process. A nonempty compact set D ⊂ Rd is called pullbackabsorbing if for each t ∈ R and every bounded set B ⊂ Rd , there exists aT > 0 such that
ϕ(t, t0,B) ⊂ D for all t0 ≤ t − T .
Theorem (Existence of pullback attractors)
Let ϕ be a process with a pullback absorbing set D. Then there exists apullback attractor A uniquely determined by
A(t) =⋂τ≥0
⋃t0≤−τ
ϕ(t, t0,D) for all t ∈ R .
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Linearisations in the autonomous caseLinearisations in the nonautonomous caseExponential dichotomySacker–Sell spectrum
Linear theory
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Linearisations in the autonomous caseLinearisations in the nonautonomous caseExponential dichotomySacker–Sell spectrum
Linearisations in the autonomous case
Consider the autonomous differential equation
x = f (x) ,
with a smooth function f : Rd → Rd , having an equilibrium x∗ ∈ Rd ,i.e., f (x∗) = 0.
Spectrum
The linearisation in this equilibrium is then given by the linear system
x = Df (x∗) x ,
the spectrum of which is defined by
σ :={
Reλ : λ is an eigenvalue of Df (x∗)}
= {a1, . . . , an} .
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Linearisations in the autonomous caseLinearisations in the nonautonomous caseExponential dichotomySacker–Sell spectrum
Spectral decomposition
The spectrum describes the exponential growth behaviour of the system.
Spectral decomposition
Suppose that σ = {a1, . . . , an} is the spectrum of x = Df (x∗) x . Thenfor all elements ai ∈ σ, there is a invariant linear subspace
Wi :=⊕
λ∈C,Reλ=ai
ker(A− λ1)d
such thatsolutions in Wi grow as fast as eai t .
The subspaces Wi form a linear decomposition of Rd , i.e.,
W1 ⊕ · · · ⊕Wn = Rd .
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Linearisations in the autonomous caseLinearisations in the nonautonomous caseExponential dichotomySacker–Sell spectrum
Hierarchy of invariant subspaces
Assume that σ = {a1, a2, a3} with a1 < a2 = 0 < a3 andW1 ⊕W2 ⊕W3 = Rd . This gives a hierarchy of invariant subspaces.
Hierarchy of invariant subspaces
W1 ⊂ W1 ⊕W2 ⊂ Rd
∪ ∪W2 ⊂ W2 ⊕W3
∪W3
Stable subspace W1
Contains all trajectories which converge exponentially to 0 in forwardtime.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Linearisations in the autonomous caseLinearisations in the nonautonomous caseExponential dichotomySacker–Sell spectrum
Hierarchy of invariant subspaces
Assume that σ = {a1, a2, a3} with a1 < a2 = 0 < a3 andW1 ⊕W2 ⊕W3 = Rd . This gives a hierarchy of invariant subspaces.
Hierarchy of invariant subspaces
W1 ⊂ W1 ⊕W2 ⊂ Rd
∪ ∪W2 ⊂ W2 ⊕W3
∪W3
Center-stable subspace W1 ⊕W2
Contains all trajectories which grow not too fast in forward time.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Linearisations in the autonomous caseLinearisations in the nonautonomous caseExponential dichotomySacker–Sell spectrum
Hierarchy of invariant subspaces
Assume that σ = {a1, a2, a3} with a1 < a2 = 0 < a3 andW1 ⊕W2 ⊕W3 = Rd . This gives a hierarchy of invariant subspaces.
Hierarchy of invariant subspaces
W1 ⊂ W1 ⊕W2 ⊂ Rd
∪ ∪W2 ⊂ W2 ⊕W3
∪W3
Unstable subspace W3
Contains all trajectories which converge exponentially to 0 in backwardtime.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Linearisations in the autonomous caseLinearisations in the nonautonomous caseExponential dichotomySacker–Sell spectrum
Hierarchy of invariant subspaces
Assume that σ = {a1, a2, a3} with a1 < a2 = 0 < a3 andW1 ⊕W2 ⊕W3 = Rd . This gives a hierarchy of invariant subspaces.
Hierarchy of invariant subspaces
W1 ⊂ W1 ⊕W2 ⊂ Rd
∪ ∪W2 ⊂ W2 ⊕W3
∪W3
Center-unstable subspace W2 ⊕W3
Contains all trajectories which grow not too fast in backward time.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Linearisations in the autonomous caseLinearisations in the nonautonomous caseExponential dichotomySacker–Sell spectrum
Hierarchy of invariant subspaces
Assume that σ = {a1, a2, a3} with a1 < a2 = 0 < a3 andW1 ⊕W2 ⊕W3 = Rd . This gives a hierarchy of invariant subspaces.
Hierarchy of invariant subspaces
W1 ⊂ W1 ⊕W2 ⊂ Rd
∪ ∪W2 ⊂ W2 ⊕W3
∪W3
Center subspace W2
Contains trajectories which are both in W1 ⊕W2 and W2 ⊕W3 (inparticular, bounded trajectories).
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Linearisations in the autonomous caseLinearisations in the nonautonomous caseExponential dichotomySacker–Sell spectrum
Linearisations in the nonautonomous case
We have linearised an autonomous differential equation
x = f (x)
in an equilibrium x∗.
In the nonautonomous context,it is possible to linearise along an arbitrary solution.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Linearisations in the autonomous caseLinearisations in the nonautonomous caseExponential dichotomySacker–Sell spectrum
Linearisations in the nonautonomous case
We have linearised an autonomous differential equation
x = f (x)
in an equilibrium x∗.
In the nonautonomous context,it is possible to linearise along an arbitrary solution.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Linearisations in the autonomous caseLinearisations in the nonautonomous caseExponential dichotomySacker–Sell spectrum
Linearisations in the nonautonomous case
Consider the nonautonomous differential equation
x = f (t, x) ,
with a smooth function f : R× Rd → Rd . Let µ : R→ Rd be a solutionof this differential equation.
1. Transformation to the trivial equilibrium. Introduce a new variabley = x − µ(t). This yields the differential equation
y = f (t, y + µ(t))− f (t, µ(t)) ,
the so-called equation of perturbed motion.
2. Identifying linear and nonlinear part. The y -equation reads as
y = D2f (t, µ(t)) y︸ ︷︷ ︸linear part
+ f (t, y + µ(t)) + f (t, µ(t))− D2f (t, µ(t)) y︸ ︷︷ ︸nonlinear part
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Linearisations in the autonomous caseLinearisations in the nonautonomous caseExponential dichotomySacker–Sell spectrum
Linearisations in the nonautonomous case
This means that the solution µ corresponds to the zero solution of adifferential equation of the form
y = A(t)y + r(t, y) .
We now study the linear part of this equation, which is the so-calledvariational equation
y = D2f (t, µ(t))y .
Remark
In the autonomous context, such a transformation is only possible forequilibria (without leaving the class of autonomous systems).
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Linearisations in the autonomous caseLinearisations in the nonautonomous caseExponential dichotomySacker–Sell spectrum
Linearisations in the nonautonomous case
Consider the variational equation
x = D2f (t, µ(t)))︸ ︷︷ ︸=:A(t)
x .
with an induced process ϕ.
Warning
The eigenvalues of A(t) do not give information about the stability of thesystem. This can be see by the example
A(t) :=
(−2 cos2 t −1− 2 cos t sin t
1− 2 cos t sin t −2 sin2 t
).
For each t ∈ R, the matrix A(t) has double eigenvalue −1, but thereexist solutions which converge to ∞.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Linearisations in the autonomous caseLinearisations in the nonautonomous caseExponential dichotomySacker–Sell spectrum
Exponential dichotomy
Definition (Exponential dichotomy)
Let α < β. ϕ admits an exponential dichotomy with gap (α, β) if thereare K ≥ 1 and invariant nonautonomous sets S+ and S− with
S+(t)⊕ S−(t) =Rd for all t ∈ R ,
‖ϕ(t, t0, x0)‖ ≤Keα(t−t0)‖x0‖ for all t ≥ t0 and x0 ∈ S+(t0) ,
‖ϕ(t, t0, x0)‖ ≥ 1
Keβ(t−t0)‖x0‖ for all t ≥ t0 and x0 ∈ S−(t0) .
R
RdS+
S−
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Linearisations in the autonomous caseLinearisations in the nonautonomous caseExponential dichotomySacker–Sell spectrum
Sacker–Sell spectrum
Definition (Sacker–Sell spectrum)
The Sacker–Sell spectrum of the above linear system is defined by
Σ :=
R \⋃
ED with gap (α,β)
(α, β)
∪{−∞} ∪ {∞}︸ ︷︷ ︸possibly
.
Remark
All Lyapunov exponents are included in the Sacker–Sell spectrum.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Linearisations in the autonomous caseLinearisations in the nonautonomous caseExponential dichotomySacker–Sell spectrum
Spectral Theorem
Theorem (Spectral Theorem)
The Sacker–Sell spectrum is given by
Σ = [a1, b1] ∪ · · · ∪ [an, bn] ,
where −∞ ≤ a1 ≤ b1 < a2 ≤ · · · < an ≤ bn ≤ ∞ and 1 ≤ n ≤ d , and foreach spectral interval [ai , bi ], there exists a linear and invariantnonautonomous set Si , a so-called spectral manifold with
S1(t)⊕ · · · ⊕ Sn(t) = Rd for all t ∈ R .
Remark
Solutions lying in the spectral manifolds admit an exponential growthbehaviour, which is determined by the corresponding spectral interval.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Invariant manifoldsAn exampleThe importance of invariant manifolds
Nonlinear theory
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Invariant manifoldsAn exampleThe importance of invariant manifolds
Invariant manifolds
Consider the nonlinear autonomous differential equation
x = Ax + r(x) .
We assume the following:
Hypothesis on linear part.σ(A) = {a1, a2, a3}.
Hypothesis on nonlinearity.Dr(0) = 0.
Existence of invariant manifolds
Then, in a neighbourhood of 0, there exist invariants manifolds, whichinherit the exponential growth behaviour from the linear subspaces.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Invariant manifoldsAn exampleThe importance of invariant manifolds
Invariant manifolds
Consider the nonlinear autonomous differential equation
x = Ax + r(x) .
We assume the following:
Hypothesis on linear part.σ(A) = {a1, a2, a3}.Hypothesis on nonlinearity.Dr(0) = 0.
Existence of invariant manifolds
Then, in a neighbourhood of 0, there exist invariants manifolds, whichinherit the exponential growth behaviour from the linear subspaces.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Invariant manifoldsAn exampleThe importance of invariant manifolds
Nonlinear systems
We have seen that an arbitrary solution µ of a nonautonomousdifferential equation
x = f (t, x)
can be transformed to the trivial solution of a differential equation of theform
x = A(t)x + r(t, x) ,
where r(t, 0) = 0 for all t ∈ R. We assume that r is a C 1-function.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Invariant manifoldsAn exampleThe importance of invariant manifolds
Invariant manifolds
Theorem (Existence of local invariant manifolds)
We assume:
x = A(t)x admits an exponential dichotomy with gap (α, β).
The nonlinearity fulfills limx→0 supt∈R∥∥ ∂r∂x (t, x)
∥∥ = 0.
Then, locally, there exist invariant nonautonomous sets T+ and T−,so-called invariants manifolds, which inherit the growth behaviour fromlinear manifolds S+ and S−.
R
RNS+
S−
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Invariant manifoldsAn exampleThe importance of invariant manifolds
Invariant manifolds
Theorem (Existence of local invariant manifolds)
We assume:
x = A(t)x admits an exponential dichotomy with gap (α, β).
The nonlinearity fulfills limx→0 supt∈R∥∥ ∂r∂x (t, x)
∥∥ = 0.
Then, locally, there exist invariant nonautonomous sets T+ and T−,so-called invariants manifolds, which inherit the growth behaviour fromlinear manifolds S+ and S−.
R
RNS+
S−
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Invariant manifoldsAn exampleThe importance of invariant manifolds
Invariant manifolds
Theorem (Existence of local invariant manifolds)
We assume:
x = A(t)x admits an exponential dichotomy with gap (α, β).
The nonlinearity fulfills limx→0 supt∈R∥∥ ∂r∂x (t, x)
∥∥ = 0.
Then, locally, there exist invariant nonautonomous sets T+ and T−,so-called invariants manifolds, which inherit the growth behaviour fromlinear manifolds S+ and S−.
R
RNT+
T−
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Invariant manifoldsAn exampleThe importance of invariant manifolds
Example
We consider the system
x =− x + 0.4 cos(t)(sin(x) + sin(y)) ,
y = y + 0.4 sin(t) sin(x) .
The system fulfills the conditions of the invariant manifold theorem, andwe obtain a stable (green) and an unstable (yellow) manifold.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Invariant manifoldsAn exampleThe importance of invariant manifolds
The importance of invariant manifolds
Invariant manifolds are important for the study of bothlocal and global
dynamical behavior.
Locally
• Stable and unstable manifolds describe the saddle point structurearound hyperbolic equilibria.
• Center manifolds capture the essential dynamics, which makes thema main object in bifurcation and stability theory.
Globally
• Invariant manifolds serve as separatrix between different domains ofthe space.
• Attractors consist of unstable manifolds.
• Inertial manifolds allow a reduction to finite-dimensional dynamics.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous pitchfork bifurcationNonautonomous pitchfork bifurcation
Bifurcation theory
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous pitchfork bifurcationNonautonomous pitchfork bifurcation
Autonomous pitchfork bifurcation
αx We consider for α ∈ R the differential equa-
tionx = αx − x3 ,
which is a prototype for an autonomouspitchfork bifurcation.
• For α < 0, we have only oneequilibrium (x = 0), which isattractive.
• For α > 0, x = 0 becomes repulsive,and two other equilibria are created(x = ±
√α), which are attractive.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous pitchfork bifurcationNonautonomous pitchfork bifurcation
Autonomous pitchfork bifurcation
αx
A(α)
• For α < 0, we have only oneequilibrium (x = 0), which isattractive.
• For α > 0, x = 0 becomes repulsive,and two other equilibria are created(x = ±
√α), which are attractive.
Alternative interpretation
A(α) := [−√α,√α] is an attractor, which
shrinks down in the limit α↘ 0.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous pitchfork bifurcationNonautonomous pitchfork bifurcation
Autonomous pitchfork bifurcation
αx
A(α)
• For α < 0, we have only oneequilibrium (x = 0), which isattractive.
• For α > 0, x = 0 becomes repulsive,and two other equilibria are created(x = ±
√α), which are attractive.
Alternative interpretation
A(α) := [−√α,√α] is an attractor, which
shrinks down in the limit α↘ 0.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous pitchfork bifurcationNonautonomous pitchfork bifurcation
Autonomous pitchfork bifurcation
αx
A(α)
• For α < 0, we have only oneequilibrium (x = 0), which isattractive.
• For α > 0, x = 0 becomes repulsive,and two other equilibria are created(x = ±
√α), which are attractive.
Alternative interpretation
A(α) := [−√α,√α] is an attractor, which
shrinks down in the limit α↘ 0.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous pitchfork bifurcationNonautonomous pitchfork bifurcation
Autonomous pitchfork bifurcation
Recall the prototype of an autonomous pitchfork bifurcation:
x = αx − x3 .
Autonomous pitchfork bifurcation
We obtain a pitchfork bifurcation of an autonomous differential equation
x = a(α)x + b(α)x3 + r(α, x)
under the following conditions:
• Transversal exchange of stability: a(0)=0 , dadα (0) > 0,
• Non-vanishing third Taylor coefficient: b(0) < 0,
• Conditions on the remainder: r(α, x) = O(x4, x2α, xα2).
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous pitchfork bifurcationNonautonomous pitchfork bifurcation
Autonomous pitchfork bifurcation
Recall the prototype of an autonomous pitchfork bifurcation:
x = αx − x3 .
Autonomous pitchfork bifurcation
We obtain a pitchfork bifurcation of an autonomous differential equation
x = a(α)x + b(α)x3 + r(α, x)
under the following conditions:
• Transversal exchange of stability: a(0)=0 , dadα (0) > 0,
• Non-vanishing third Taylor coefficient: b(0) < 0,
• Conditions on the remainder: r(α, x) = O(x4, x2α, xα2).
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous pitchfork bifurcationNonautonomous pitchfork bifurcation
Nonautonomous pitchfork bifurcation
Consider the nonautonomous differential equation
x = a(t, α)x + b(t, α)x3 + r(t, α, x) ,
for which we want to formulate nonautonomous conditions for
• transversal exchange of stability,
• non-vanishing third Taylor coefficient,
• conditions on the remainder.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous pitchfork bifurcationNonautonomous pitchfork bifurcation
Nonautonomous pitchfork bifurcation
Consider the nonautonomous differential equation
x = a(t, α)x + b(t, α)x3 + r(t, α, x) ,
Transversal exchange of stability
The autonomous case:spec
α
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous pitchfork bifurcationNonautonomous pitchfork bifurcation
Nonautonomous pitchfork bifurcation
Consider the nonautonomous differential equation
x = a(t, α)x + b(t, α)x3 + r(t, α, x) ,
Transversal exchange of stability
The nonautonomous case:spec
α
β1
β2
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous pitchfork bifurcationNonautonomous pitchfork bifurcation
Nonautonomous pitchfork bifurcation
Consider the nonautonomous differential equation
x = a(t, α)x + b(t, α)x3 + r(t, α, x) ,
Transversal exchange of stability
There exist functions β1, β2 and K ≥ 1 such that for τ ≤ t ∈ I , we have∫ t
τ
a(s, α)ds ≤ β1(α)(t − τ) logK ,
∫ τ
t
a(s, α)ds ≤ β2(α)(t − τ) logK
and
lim supα→0
lim supx→0
supt∈R
2K |r(t, x , α)||x |max
{− β1(α), β2(α)
} < 1 .
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous pitchfork bifurcationNonautonomous pitchfork bifurcation
Nonautonomous pitchfork bifurcation
Consider the nonautonomous differential equation
x = a(t, α)x + b(t, α)x3 + r(t, α, x) ,
Non-vanishing third Taylor coefficient
The autonomous condition b(0) < 0 is generalised by
−∞ < lim infα→0
inft∈R
b(t, α) ≤ lim supα→0
supt∈R
b(t, α) < 0 .
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous pitchfork bifurcationNonautonomous pitchfork bifurcation
Nonautonomous pitchfork bifurcation
Consider the nonautonomous differential equation
x = a(t, α)x + b(t, α)x3 + r(t, α, x) ,
Conditions on the remainder
The autonomous condition r(α, x) = O(x4, x2α, xα2) is generalised by
limx→0
supα∈(−x2,x2)
supt∈R
|r(t, x , α)||x |3
= 0 .
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous pitchfork bifurcationNonautonomous pitchfork bifurcation
Nonautonomous pitchfork bifurcation
Nonautonomous pitchfork bifurcation (supercritical case)
Under the above hypotheses, the nonautonomous differential equation
x = a(t, α)x + b(t, α)x3 + r(t, α, x)
admits a bifurcation of pullback attractors; more precisely, for negative αnear 0, the trivial solution is a local pullback attractor, which bifurcatesto a nontrivial pullback attractor Aα for positive α near 0. One has thelimit relation
limα↘0
dist(Aα(t), {0}
)= 0 for all t ∈ R .
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous pitchfork bifurcationNonautonomous pitchfork bifurcation
Summary
We have seen that there are two different types of attraction in thenonautonomous case: forward and pullback attraction.
The existence of pullback attractors follows via absorbing sets.
Hyperbolicity for nonautonomous systems is described by theconcept of an exponential dichotomy.
The Sacker–Sell spectrum gives information about the exponentialgrowth behaviour of linear systems.
The growth rates are attained in the spectral manifolds.
The spectral manifolds persist under nonlinear perturbation and givethe invariant manifolds.
Nonautonomous bifurcations can be described as attractorbifurcations; we have seen how the autonomous pitchfork bifurcationpattern translates to the nonautonomous case.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous pitchfork bifurcationNonautonomous pitchfork bifurcation
Questions?
If you have questions...
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous pitchfork bifurcationNonautonomous pitchfork bifurcation
Questions?
If you have questions...
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems
Basic theoryAttractor theory
Linear theoryNonlinear theory
Bifurcation theory
Autonomous pitchfork bifurcationNonautonomous pitchfork bifurcation
Thanks!
Thanks for your attention!And thanks to:
L. Arnold, A. Berger, B. Aulbach, T. Caraballo, A. Carvalho, D. Cheban,C. Chicone, I. Chueshov, F. Colonius, W. Coppel, H. Crauel, T.S. Doan,
R. Fabbri, L. Grune, T. Huls, P. Imkeller, R. Johnson, J. Langa,C. Nunez, W. Kliemann, P.E. Kloeden, Y. Latushkin, S. Novo, R. Obaya,
R. Ortega, K. Palmer, C. Potzsche, J. Robinson, R. Sacker,B. Schmalfuß, G. Sell, S. Siegmund, T. Wanner, Y. Yi.
Martin Rasmussen Qualitative Theory of Nonautonomous Dynamical Systems