NONLINEAR SYSTEMS

We consider a system of differential equations * =

f(t, x)

where
xeEcR", f Gf.fn) and f
= =
f(t,x, 2 ,n)
If f depends explicitly on t, the system is said to be time dependent or non-autonomous
or forced equation.

If f does not depend explicitly on t, the system is called autonomous.

The system is called nonlinear autonomous if f is nonlinear as well as autonomous.

A point xo is called an equilibrium point (or critical point or rest point or fixed point or

steady state) of the nonlinear autonomous system x =

f(x), if f(x0 ) = 0.
Jheorem: Let E bean open subset of R" containing xpand assume that fe C (E). Then

there exists a real number a>0 such that the nonlinear autonomous initial value

problemn

X= f(x) x(0) , =*o

has a unique solution x(t) on the interval [-a, a).
Phasespace:
We consider a system of differential equations * = f(t,x), x() = *o (1)
where x eE CR", f =ff2, r, fn) and fi =
fi(t,X1, X2,Xn).

The space (E) in which we describe the behaviour of the variables (X1, X2,.,Kn)
parameterized by t, is called phase-space of dimension n.
A point in phase-space with coordinates (x1 (t), X2 (t), .., xn (t)) for certain t, is called a
phase-point. In general, for increasing t, a phase-point shall move through phase-
space.
The solution of the differential equations (1) is called an integral curve or orbit or

trajectory through xo.

The phase portrait of a system of differential equations (1) with xeR" is the set of all
solution curves of (1) in the phase space R".
Theflow defined bya differential equation
We consider a system of differential equations *= f(x), x{0) = Xo (1)

where xeE CRT.

For xo ¬E, let o(t,x0) be

Definition: Let E be an open subset of R and let feci (E).
the solution of the initial value problem (1) defined on its maximal interval of existence

I(xo). Then for t e (xo), the set of mappings o defined by

:(xo) = ot, xo)

is called the flow of the differential equation (1) or the flow defined by the differential
equation (1); pt, is also referred to as the flow of the vector field f(X) or evolution

operator. The solution d; gives the past (t<0) or future (t>0) positions or evolution of
the phase-point through xo. So, existence and uniqueness of solutions are required for

t o be well defined.
The flow de also satisfies the following group properties
() ia where i is the identity map on E.

(ii) p =
p' o p', i.e. p**x g' (g'x) for
=
all x e E and t,s e T.
Linearization of dynamical systems:
We consider nonlinear system *
a =
f(x), xeR" (1)
with an equilibrium point xo, we assume that f is differential each of the n variables of

vectors x. A local linear approximation of the nonlinear system near the equilibrium

point o is given by the expansion in a Taylor series of the function f truncated after the

1st order terms.

Let X-zo then = fG +x) = fxo) + Df(z,)E +D'fx,) + 2

Considering the 1st order term on the r.h.s. of ) we get the local linear approximation

of (1) as =
Df(x%)E, i.e.
the system reduced to a liner system * = Ax (2)
around the equilibrium point Xo . The matrix Df(xo) is called Jacobian matrix of f at Xo

and Ax Df(xo)x is the linear

=
part of f at xo.

It is important to note that how much of the dynamic characterization of the nonlinear
system (1) is preserved by the linear approximation equation (2). There exists a general
result in the theorem of differential equation known as Hartman-Grobman theorem.
Hartman-Grobman theorem states that the qualitative (topologically equivalent)
behavior of the solution set of a nonlinear system of ordinary differential equations near
a hyperbolic equilibrium point is typically the same as the qualitative behavior of the
solution set of the corresponding linearized system near the equilibrium point.

Definition: An equilibrium point xo ER" is called a hyperbolic equilibrium point of (1) if

none of the eigenvalues of the matrix Df (x%) have zero real part.

Definition: An equilibrium point xo of (1) is called a sink if all of the eigenvalues of the

matrix Df(x0) have negative real part; it is called source if all of the eigenvalues of the

matrix Df(xo) have positive real part; and it is called a saddle if it is a hyperbolic

equilibrium point and if Df(%o) has at least one eigenvalue with a positive real part and
at least one eigenvalue with a negative real part.
 open subset of R", let feC (E), and let o:E> E be the flow
Definition: Let E be an

R. Then a set S cE is called invariant with

of the nonlinear system (1) defined for all te

respect to the flow o, if p(S) c S for all te R and S is called positively (or negatively)
invariant with respect to the flow , if o(S) c S for all t> 0 (or t < 0).

Theorem:(Stable Manifold Theorem)

and let pe be the flow

Let E be an open subset of R" containing the origin, let fec1 (E),
of the nonlinear system (1). Suppose that f(O) =0 and that Df(0) has k eigenvalues with
Then there exist a k-
negative real part and n-k eigenvalues with positive real part.
linear
dimensional differentiable manifold S tangent to the stable subspace Es of the

system (2) at 0 such that for all t2 0, p: (S) c S and for all x E S,limt-o p¢(xo) =
0;

and there exist an n-k dimensional differentiable manifold U tangent to the unstable
and for all
subspace F" of the linear system (2) at 0 such that for all t 0, d: (U) cU
0.
Xo EU, lim- Pe(xo) =

Hartman Theorem:
Let E be an open subset of R" containing the point xo ¬R", let feC (E), and let ¢e be
the flow of the nonlinear system (1). Suppose that f(x0)=0 and that all of the

eigenvalues ,Az, .. , A,of the matrix A=Df(X0) have negative (or positive) real part.

Then there exists a C-diffeomorphism H [i.e. H is homeomorphism of R with

H,H-1e C1(R")] of a neighborhood U of Xo onto an open set V containing the origin such

thatfor each x ¬l there is an open interval l(X) CR containing zero such that for all

x eU and teI(x)
H op,(x) =
etH(x)
i.e., H maps trajectories of (1) near the point Xo onto trajectories of (2) near the origin
and preserves the sense of the orbits or parameterization (orientation) by time.
Some formal definition of stability of equilibrium point of a nonlinear system:

Ne consider a nonlinear autonomous system x = f(x), x¬R" ( 1 )

with an isolated equilibrium point N", such that f(N°) = 0. Let N(t) be the solution of

Euclidean
the system at time t with initial condition N(to)=No and |I. | indicates the
distance in R".

Definition
A critical point x = N* of the equation x = f(x), x¬R" is called a positive attractor

if there exists a neighbourhood y cR" of x = N* such that x(t%) E AN implies

lim- X(to) = N°. If a critical point x= N* has this property for t ^ - o , then X =N* is

called a negative attractor.

Definition: An equilibrium point N" is Lyapunov stable if, for any e> 0, there exist a

8(e)>0 such that, for all initial condition N(t%)=N, satisfying INo- N*'|l <8, we have

lINCE)-N"|| <e V t> to OR

Definition: An equilibrium point N* is asymptotically stable if it is stable and if there
exist a p>0 such that lim-N() -

N"|| = 0v N, satisfying |N -

N"| <p. OR

An equilibrium point N is asymptotically stable if it is stable and there is a

neighbourhood cA cR"
R of N" such that every trajectory which passes through Qy
approaches N' as t tends to infinity.

Definition: An equilibrium point N* is stable but not asymptotically stable is said to be

neutrally stable.
Definition: An equilibrium point N' of the system (1) is called unstable if it is not stable.
Unstable means there exist a e> 0 if an orbit starts from a point No arbitrary closed to
the equilibrium point N* such that the distance between the orbit from the
equilibrium
point becomes at least as large as e at a finite time, i.e. ||N(t) -

N°|| 2 e for some t> to

Definition: Let N be an asymptotically stable equilibrium point of (1), then the set
B(N°) = {xER°: lim, lIN() - N"|| = 0) is the domain or basin of attraction of N".

I B(N") = R", then N* is said to be globally asymptotically stable. If stability only holds

in a neighbourhood of N*, then is to be locally asymptotically stable

Remark:
If det A#0, the system i Ax has
=
a unique equilibrium point at the origin. Forthose
systems asymptotically stability is always global. Nonlinear system, on the other hand,
may have any finite or a countably infinite equilibrium points, local stability does not
always imply global stability

Liapunov's second method or Direct method of Lyapunov:

The direct method for studying the stability of nonlinear system is very different
from the method of linearization.
This method makes use of certain functions
(a distance function like form) and attempts
to determine the stability (not necessarily local) for both
simple and non simple fixed
point by evaluating the sign of their time derivatives along the orbits of the system.

We consider a nonlinear system x = f(x), xeRT

(1)
with an isolated equilibrium point x0 which we assume to be at origin.
Definition: Let E be an open subset of R" containing the point xo = 0 ER". A
continuously differentiable real valued function V:E R
»
(Vec'(E)) is said to be
positive (negative) definite in E if V(x) > 0 (V(x) < 0) for E\{0} and if V(0) = 0.
 the point Xo = 0 ER". A
Definition: Let E be an open subset of R" containing
(Vec'(E)) is said to be
continuously differentiable real valued function V:E» R

E\{0} and if V(0) = 0.

positive (negative) semi-definite in E if V(x) 2 0 ( V(x) < 0) for

point Xo = 0 ¬R". If feC"(E),

Definition: Let E be an open subset of R" containing the

VEC"(E) and is the flow of the nonlinear system (1), then for x e E the derivative of
o
thefunction V(x) along the solution «(x)is

V=DV(fE) 1f:x).
dxi
=

Kiapunov stability theorem and Liapunov functions:

Theorem: Let E be an open subset of R" containing the point 0 ¬R". Suppose that

fec (E) and that f(0)=0. Suppose further that there exists a real valued function

Vec(E) satisfying V(0) = 0 and V(x) > 0 E\{O}. Then

(1) if V(x) s 0 vxeE, then Xo =0 is stable;

(i) if V(x) <0 vxe E\{0}, then Xo = 0 is asymptotically stable;

(i) if V(x) >0 v* eE\{0}, then xo = 0 is unstable.

A function V: R" R satisfying the hypotheses of the above theorem is called a

Liapunov function.