Chapter 1

Introduction to Steady State

Bifurcation Theory

In this chapter, we present some classical theorems on steady state bifur-

cations, including the Lyapunov-Schmidt procedure, bifurcation theorems
from eigenvalues of odd multiplicity, and bifurcation theorems for potential
operators. The presentation we adopt here is to use the Lyapunov-Schmidt
procedure to link all these theorems in a natural way.
The version of the Lyapunov-Schmidt procedure presented here differs
slightly from the one given in classical textbooks. The later is done by
decomposing the space into the direct sum of the eigenspace and its com-
plement. While the Lyapunov-Schmidt procedure presented here is based
on the decomposition of the space into the direct sum of the generalized
eigenspace and its complement. This Lyapunov-Schmidt procedure is more
nature, and much more convenient to study steady state bifurcations. In
fact, it is this difference, together with other ingredients, including in par-
ticular the spectral theorem in Chapter 3, that made many problems more
accessible.
Another important ingredient for the Lyapunov-Schmidt procedure pre-
sented here is the introduction of the normalization of the bifurcation equa-
tion (1.19), which is crucial for the new steady state bifurcation theory from
higher order terms regardless of the multiplicity of the linearized eigenprob-
lems given in Chapter 4.

1.1 Implicit Function Theorem

Let X , Y, Z be Banach spaces, U c X x Y an open set, and F : U -, Z a

mapping. Let (xo,yo) E U satisfy
2 Bifurcation Theory and Applications

We consider the solvability near (xo,yo) of the equation

The following well known implicit function theorem is of fundamental

importance in the bifurcation theory.
Theorem 1.1 (Implicit Function Theorem). Let F E ck(U, Z ) be k -
t h order differentiable ( k 2 l), and satisfy (1.1). Assume that the derivative
operator D,F(xo, yo) : X -+ Z has a bounded inverse. Then, the following
assertions hold true.

(1) There exist a neigborhood V c X of xo and a neighborhood W C Y of

yo such that as y E W , (1.2) has a unique solution x = @(y) E V .
(2) The mapping @ : W -+ V is k-th diigeerentiable, especially i f F is ana-
lytic, then @ is also analytic.
(3) If DyF(x0,yo) = 0 , then we have

Remark 1.1 Actually, if the derivations of F with respect to y vanish

up to m-th order at (xo,yo):

then the implicit function satisfies that

1.2 Basics of Topological Degree Theory

1.2.1 Brouwer degree

a
Let fi c Rn be a bounded open set, F : -+ Rn a continuous mapping.
p E Rn with p @ F(8R). We begin with the definition of Brouwer degree,
which assigns each triple (F, R,p) to an integer, denoted by deg(F, R,p),
which is called the Brouwer degree of F in R with respect to p.
A point xo E R is called a regular point of F if the Jacobian matrix
DF(xo) is invertible; otherwise xo is called a critical point of F. A point
p E Rn is called a regular value of F if either F W 1 ( p )= 8 or each point
 Introduction t o Steady State Bifurcation Theory 3

x E F-l(p) is a regular point. A point p E Rn is called a critical value

of F if it is not a regular value. By the Sard theorem [Milnor, 19651, for
F E C1(R, Rn) the set of critical values of F has measure zero in Rn.
Definition 1.1 We define for each triple (F,R,p) with p $! F(BR) the
Brouwer degree as follows,
(1) if F E C1(n, Rn) and p E Rn is a regular value of F, then

where det DF(y) is the determinant of DF(y);

(2) if F E C1(fi,Rn) and p E Rn is a critical value of F, then we take a
regular value pl of F with Ipl - pl < dist(p, F(dS2)) and define that

(3) if F E C(O,Rn) is a continuous function, we take a C1 function Fl E

C1( n , Rn) such that

and define the Brouwer degree by

It is routine, but lengthy, to justify that the Brouwer degree is well-

defined, and enjoys the following basic properties; see among others [Niren-
berg, 2001; Chen, 19811:

where id : R 4 Rn is the identity.

2. DOMAINADDITIVITY:If R1, R2 c R with nflz = 0, p $! F ( n \ (QI U
Q2)), then

3. HOMOTOPY INVARIANCE: Let H : fi x [0, 11 -, Rn be continuous, and

for each t E [ O , l ] , p 4 H(dQ t ) . Then
4 Bijurcation Theory and Applications

where Ho(x) = H(x, O), and HI (x) = H(x, 1).

4. KRONECKER'S EXISTENCETHEOREM:If deg(F,R,p) # 0, then the
equation F(x) = p has at least one solution in R.
5. EXCISION PROPERTY: If D c 0 is a closed set, and p @ F ( D ) , then

6. CONNECTED DOMAIN PROPERTY:Let p : IR1 -, fi E CO. If p(t) $!

F(d0) for any t E R1, then
deg(F, a , p(t)) = const.
PROPERTY:If F, G E C(QRn) with F J a n = GIan, then
7. BOUNDARY

W F , a , P) = deg(G, 0 , P).
8. POINCARE-BOHL Let F,G
THEOREM: E C(ft,Rn). If

tF(x) + (1- t)G(x) # p, Vt E [0, l],Vx E dR,

then

1.2.2 Basic theorems of Brouwer degree

We introduce here four basic theorems for the Brouwer degree, which will
be useful in the later discussion. They are the Reduction Theorem, Multi-
plication Formula, Borsuk Theorem, and Even Mapping Theorem.
Theorem 1.2 (Reduction Theorem). Let Rm c Rn (n > m), and
R c Rn a bounded open set, f E C(Q,Rm). I f p E Rm and p $ F(BR),
where F(x) = x - f (x), then

where G = FInnw-.
Example 1.1 Consider the following two mappings
 Introduction to Steady State Bifurcation Theory 5

Theorem 1.3 (Multiplication Formula). Let R1, R2,c Rn be

bounded open sets, F E C ( n l , R n ) , G E C(Q2,Itn), F ( n l ) C R2. Let
R, be the connected domains of Rz \ F(BR1). If p +! G o F(BR1) u G(BRz),
then

where pa E R,. B y the connected domain property, the degree deg(G o

F,Rl, p) is independent of pa E R,.
Theorem 1.4 (Borsuk Theorem). Let R c Rn be a bounded open set,
O E R , and F EC(fi,Rn). I f R is symmetric o n x = O , i.e. X E R * -x E
R, and

then deg(F, R, 0 ) = odd.

Theorem 1.5 (Even Mapping Theorem). Let R be as in Theo-

rem 1.4, and F : R --+ R" be an even function, i.e. F ( z ) = F(-x), Q
x E 8 0 . Then we have

deg(F, R, 0) = even.

Remark 1.2 The Even Mapping Theorem here will be applied to de-
rive bifurcations from higher-order nondegenerate singularities in Chapter 4
later, which is easily derived from the Sard theorem and Definition 1.1.

1.2.3 Leray-Schauder degree

The Leray-Schauder degree is a generalization of the Brouwer degree to
mappings defined on an infinite dimensional Banach space. Let X be a
Banach space, and R c X a bounded open set. A mapping F : R -+
X is called compact mapping or completely continuous mapping if F is
continuous and maps any bounded set of R to a precompact set of X. The
following mapping is called completely continuous field

It is known that for a completely continuous mapping F , there exists

finite dimensional spaces Xn C X and continuous mappings Fn : R -+ X,,
6 Bzfurcation Theory and Applications

such that

sup llF(x) - Fn(x)II +0, as n + oo.

XEQ

Definition 1.2 Let Xn cX be finite dimensional spaces, p E Xn, and

Fn : fi 4 Xn satisfy

Then, we define the Leray-Schauder degree for the completely continuous

field f = id-F on R at p as follows

where fn = id-Fn.

It is known that the Leray-Schauder degree is well-defined, and has the

folIowing three basic properties.

Theorem 1.6 For a completely continuous field f = id -F, the Leray-

Schauder degree enjoys the properties

2. DOMAIN ADDITIVITY:if R1, C12 c R with R1 n R2 = 0, p $ f (fi \ 521 U

Rz), then we have

3. HOMOTOPY INVARIANCE: let H :?i x [O,1] --+ X be a compact mapping

h(x, t ) = x - H(x, t ) , and p 4 h(8R, t ) for any t E [O, 11. Then

where ho = h(.,O), hl = h(., 1).

The basic properties 5-8 in subsection 1.2 hold true for the Leray-
Schauder degree as well, which can derived using Theorem 1.6.
 Introduction to Steady State Bzfurcation Theory

1.2.4 Indices of isolated singularities

We consider the equation

f ( x )= x - F ( x ) = 0 (1.4)
where F : R 4 X is a compact operator. Solutions of (1.4) are called
singularities of f .
Let xo E fl be an isolated solution of (1.4). Then we define the index of
f = id - F at the singular point xo by

where U is a small neighborhood of xo.

We know that if F : R 4 X is a differentiable compact mapping, then
the derivative operator of F at xo E R

is a linear compact operator. By the Riesz-Schauder theory, the eigenvalues

of the following equation are isolated

and the algebraic multiplicity of each eigenvalue X of (1.5) is finite, which

is the dimension of space
00

Ex = U{ y E X I (id - XDF(z0))"y = 0 ) .
n= 1
The following index formula is important for the bifurcation theory.
Theorem 1.7 Let xo E R be a singular point of a completely continuous
field f = id- F , and F is differentiable at xo. If X = 1 is not an eigenvalue
of D F ( x o ) , then xo is an isolated singular point o f f , and

where p is the sum of the multiplicities of all eigenvalues of D F ( x 0 ) in

(01 1).
We remark here that the Leray-Schauder degree theory is applicable for
the class of nonlinear operators given in the following, which cover many
problems in differential equations.
Let H and H I be two Hilbert spaces, HI 4 H a dense and compact
inclusion, A : H I + H a linear homeomorphism, and G : H I -, H be
8 Bafu~cationTheory and Applications

a compact operator. Assume that there exist a real eigenvalue sequence

{ p k ) C R1 and an eigenvector sequence { e k ) c H I of A such that

and { e k ) constitutes a basis of H.

We can define the Leray-Schauder degree for the operator A G on a +
bounded open set R C H1 at a point p E H, which satisfies the domain
additivity property, the homotopy invariance and the normalization. In
particular, we have

In this case, we also call the operator A + G : HI + H a completely

continuous field.

1.3 Lyapunov-Schmidt Method

1.3.l Preliminaries
Let X, Y be two Banach spaces, and L : X x R1 -+ Y be a linear bounded
operator and G : X x R1 -t Y a continuous mapping satisfying

Consider an operator equation

It is clear that (u, A) = (0, A) is a trivial solution of (1.7). The bifurca-

tion problem of (1.7) is to seek a nontrivial solution (ux,A) # (0, A) of (1.7)
from some point (0, Ao) such that

as shown in Figure 1.1.

More precisely, we give the following definition of bifurcation.
Definition 1.3 We say that the equation (1.7) bifurcates from (0, Ao) a
solution (ux,A) E X x R if there exists a sequence of solutions (u,, A,) of
 Introduction to Steady State Bifurcation Theory

Fig. 1.1

(1.7) which satisfy

lim A, = Ao, and lim (JunJJ

= 0.
n-w n+co

By the implicit function theorem (Theorem 1.1), one immediately de-

rives the necessary condition of bifurcation for (1.7).
Theorem 1.8 Under the condition (1.6), i f (1.7) has a bifurcation solu-
tion from (0, Ao), then the linear operator LA, : X -+ Y is not invertible.
+
Especially, i f LA is a linear completely continuous field, i.e. LA = Ax B x ,
Ax : X -+ Y a linear homeomorphism, and Bx : X 4 Y a linear compact
operator, then there exist u E X with u # 0 such that Lx,u = 0.

1.3.2 Lyapunov-Schmidt procedure

We begin with an example to introduce the Lyapunov-Schmidt procedure.
Consider a system of two algebraic equations given by

where aij = aij (A) are continuous functions of A, and f l , f 2 are C w func-
tions satisfying that
10 Bifurcation Theory and Applications

Obviously, (x, A) = (0, A) is a trivial solution of (1.8). We shall use the

Lyapunov-Schmidt procedure to investigate the bifurcation of (1.8). For
simplicity, we assume that the matrix

has two eigenvalues in a neighborhood of A. as follows

Under a coordinate transformation, the equations (1.8) are rewritten as

where Fi(x) = o(lxI), i = 1,2.

By the implicit function theorem-Theorem 1.1, (1.10) has a solution
near xl = 0:

Inserting (1.11) into (1.9) we arrive at

Thus, the bifurcation of (1.8) is equivalent to that of (1.12). By the index

theorem-Theorem 1.7, we find

which implies that (1.12) has a bifurcation from (0, Ao). Hence, (0, Ao) is a
bifurcation point of (1.8).
We now introduce the Lyapunov-Schmidt method in general setting.
Consider

where u E X , X a Banach space, A : X -+ X a linear compact operator,

and G : X x R -+ X a continuous mapping satisfying (1.6).
 Introduction to Steady State Bifurcation Theory 11

Assume that A;' is a real eigenvalue of A with algebraic multiplicity

m 2 1. Thus, the space X can be decomposed into the direct sum of two
invariant subspaces of A as follows

Eo = U{x E X 1 (id - XoA)"x = O),

nEN
dimEo = m.

The linear operator A : X -+ X can be decomposed into

Let Po : X 4 Eo and PI = id - Po : X -+ El be the canonical projections.

Then (1.13) is decomposed into

It is clear that id - XoAl : El -+ El is invertible. Based on the implicit

function theorem, (1.15) can be solved near x = 0 and y = 0:

Putting (1.16) into (1.14), we have

which is an m-dimensional equation, called the bifurcation equation of

(1.13). The following theorem verifies the validity of the Lyapunov-Schmidt
procedure.
Theorem 1.9 Let 'A: be a real eigenvalue of A with algebraic multiplicity
m 2 1. Then the bifurcation problem of (1.13) near X = Xo is equivalent to
that of (1.17).
We remark here that the Lyapunov-Schmidt procedure presented here
differs slightly from the one usually presented in classical textbooks. The
later is done by decomposing X into the direct sum of the kernel Xo of
id - XoA and its complement XI:
12 Bifurcation Theory and Applications

While the Lyapunov-Schmidt procedure presented here is based on the

decomposition X = Eo @ El with Eo being generalized eigenspace. As
we shall, it is easy to see that this Lyapunov-Schmidt procedure is more
nature, and much more convenient to study bifurcation of (1.13). In fact,
it is this difference, together with other ingredients presented in this book
that made many problems more accessible.

1.3.3 Normalization
Let the nonlinear term G : X x JR1 -+ X be Cm. Then there is a Ic 2 2
such that

where Gn(u,A) is an n-multilinear mapping, and

Gn(au,A) = anGn(u,A), b' a E R1.

By (1.16), y(x, A) is a higher order function of x. Hence, (1.17) has the

following form

If x = 0 is an isolated singular point of PoGk,then the bifurcation problem

of (1.18) is reduced to that of the following equation:

There are many bifurcation problems in science and engineering, where

the bifurcation equation (1.19) can be explicitly calculated, providing an
important step toward to their bifurcation analysis.
For example, we consider a special case where A : X -+ X is a symmetric
linear compact operator, X is a Hilbert space, A;' is an eigenvalue of A
with multiplicity two, and el, ez are the eigenvectors corresponding to Ao,
which are orthogonal

Let k = 2 and Gz be the second order multilinear operator of G. The

equation (1.19) can be expressed as the following system of %dimensional
 Introduction to Steady State Bifurcation Theory

algebraic equations

where

Remark 1.3 The method that reduces the bifurcation equation (1.17)
to its normalized form (1.19), together with the spectrum theorem and
the attractor bifurcation theorems developed in [Ma and Wang, 2004e;
Ma and Wang, 2004d], is a very useful tool in the bifurcation analysis
in many problems. The bifurcation theory with applications developed by
the authors is summarized in this book and as well as in a forthcoming
book on hydrodynamic stability and bifurcation.

1.4 Krasnosel'ski Bifurcation Theorems

1.4.1 Bifurcation from eigenvalues with odd multiplicity

When the eigenvalue ' A; of A has odd multiplicity, the following well-
known Krasnosel'ski bifurcation theorem says that the equation (1.13) has
a bifurcation from (0, Xo).
Theorem 1.10 (Krasnosel'ski Theorem). Under the condition (1.6),
if A : X -, X is a linear compact operator, and X i 1 E R1 is a real eigen-
value of A with odd algebraic multiplicity, then ( u ,A) = (0, XO) is a bifur-
cation point of (1.13).
Proof. This theorem can immediately be derived by the index theorem
(Theorem 1.7) and the Lyapunov-Schmidt theorem (Theorem 1.9). Indeed,
without loss of generality, we assume that Xo > 0. Then the bifurcation
equation (1.17) reads

By Theorem 1.7, we have

if X < Xo,
ind(id - XAo + g, 0) = ind(id - AAo, 0) =
-1 ifX>Xo.
 14 Bifu~catzonTheory and Applications

It implies that (x,A) = (O,Ao) is a bifurcation point of (1.20). Thus, by

Theorem 1.9 we obtain the Krasnosel'ski theorem. 0
Consider a special case where the algebraic multiplicity m = 1 of A;',
i.e. A;' is a simple eigenvalue of A. The following theorem is due to Cran-
dall and Rabinowitz.
Theorem 1.11 Assume that the conditions of Theorem 1.10 hold true.
If A;' is a simple eigenvalue of A, then the equation (1.13) bifurcates from
(0, Xo) to exactly two branches rl and r z , which are of the following form

r : (uxlA), u~ = te + tv(t), A = Ao + p(t),

where v(t) E X , ~ ( t E) R1 are continuous on t with v(0) = 0, p(0) = 0,
and e is the eigenvector of A corresponding to A;'.

This theorem can be derived from (1.20). In fact, if g

t > 0,
- 0, then for any

rl : UA = te + y(te, Ao),
r2: UA = -te + y(-te, Ao),
where y (x, A) is given by (1.16).
If g $ 0 in (1.20), and g(re, A) = p(r, A)e, p(r, A) has the Taylor expan-
sion near T = 0 as follows

then the equation (1.20) can be written as

which has two branches of solutions as shown in Figure 1.2.

1.4.2 Krasnosel'ski theorem for potential operators

Let X be a Hilbert space. For the equation (1.13) we assume that

(A1) A : X + X is a self-adjoint linear compact operator, and

(A2) there is a functional F E CT(X x IR1,R1), with

D,F(u, A) = G(u, A), (u, A) E X x IW',

and G(u, A) satisfies (1.6).

 Introduction to Steady State Bifurcation Theory

(4
Fig. 1.2 (a) Case: Ic =odd, cu(Xo) > 0; (b) case: k =odd, a ( X o ) < 0; ( c ) case: k =even.

The following theorem is due to Krasnosel'ski.

Theorem 1.12 Under the hypotheses (Al) and (Az), if X i 1 E R1 is

an eigenualue of A, then the equation (1.13) has at least two bifurcated
branches from (u, A) = (0, Ao).

To see this theorem, we assume that the eigenvalue Xo has multiplicity

m 2 1 with orthogonal eigenvectors {el,. . . , e m ) , F = F,, Ic 2 2,xr==k+l
and F, is the n-multilinear functional of F. Then, the bifurcation equation
16 Bifurcation Theory and Applications

of (1.13) is of the form

where x = ( X I , . . , x m ) E Rm.
If x = 0 is an isolated singular point of VFk+1(CPlxj e j , X o ) , it suffices
t o consider the bifurcation of the following system of equations

It is clear that the critical points ( x ( X ) A)

, of the functional

are the bifurcated solutions of (1.22).

One can obtain the critical points of (1.23) by the following method.
Let ( a l , .. . , a m )E Rm with laI2= 1 satisfy

Putting (XI,. - . ,x,) = ( a l t , ... , amt)in (1.23) we get

Without loss of generality, let F k + l ( C E l aiei, X O ) > 0. Then, for X > Xo

we take

Thus, this point x ( X ) = ( a l t x ,. . . , crmtx)is a critical point of (1.23).

In the same fashion, we can obtain another critical point of (1.23) by
taking a E Rm as follows
 Introduction to Steady State Bifurcation Theory

When the operator G = D F is odd, i.e.

we have the following theorem, due to D.C. Clark.

Theorem 1.13 Under the hypotheses ( A 1 ) and (Az), if G(u,X) is a n
odd operator, and A;' is a n eigenvalue of A with multiplicity m, then there
are at least m distinct pairs of bifurcated branches of (1.13) from (u,A) =
(0, Xo) .
The basic idea of finding critical points x(X) of (1.23) is as follows. We
take ( a l , . . . , a m ) E lRm such that

Because Fk+'is even, the values (al, , a m ) can be derived for each
1 5 r 5 m, and the m distinct pairs of points

.. . , a r m ) satisfies (1.25)
are the critical points of (1.23), where a, = (aTl,
and t~ satisfies (1.24).

1.5 Rabinowitz Global Bifurcation Theorem

In this section, we introduce the Rabinowitz global bifurcation theorem.

Let
I u # 0, (u,X) satisfy (1.13)).
B = {(u,X) E X x IW'
Theorem 1.14 Let G : X x JR1 --+ X be a compact operator satisfying
(1.6), and A : X + X a linear compact operator. If X i ' is a real eigenvalue
of A with odd algebraic multiplicity, then the connected component C c B
which contains (0, Xo) satisfies one of the following assertions:
(1) C is unbounded i n X x R, or
(2) C contains odd number of points (0, Xi) # (0, Xo) such that X i ' are the
eigenvalues of A with odd algebraic multiplicities.
This theorem is, in essence, a corollary of the homotopy invariance of the
Leray-Schauder degree. The assertions (1) and (2) of Theorem 1.14 amount
to saying that the connected component C c B containing (0, Xo) must be
18 Bifurcation Theory and Applications

one of the two types as shown in Figure 1.3(a) and (b). If otherwise, then C
must be bounded in X x R1as shown in Figure 1.4. By the basic properties
of the Leray-Schauder degree, we immedaitely deduce a contradiction.
More precisely, by the excision property (see Figure 1.4), we have

deg(id - X2A + G, BR,0) = deg(id - X2A + G, Br, 0)

= ind(id - X2A,0) = (-1)'.

Then the homotopy invariance of the degree shows that

deg(id - X2A + G, BR,0) = deg(id - plA + G, BR, 0),

deg(id - X2A + G, B,, 0) = deg(id - X1A + G, Br, 0).

It follows then from (1.26)-(1.28) that

deg(id - plA + G, BR,0) = deg(id - XIA + G, BR,0) = (-1)'.

Furthermore, by the index theorem (Theorem 1.7),

deg(id - plA + G, BR,0) = ind(id - pl A, 0) = (-I)'-~,

where m = odd is the algebraic multiplicity of Xo, and BR = {x E X I
llxll < R). Thus, we arrive at a contradiction from (1.29) and (1.30).

(a) (b)
Fig. 1.3 (a) C is unbounded; (b) C is bounded.
 Introduction t o Steady State Bifurcation Theory

Fig. 1.4 The larger rectangular box is BR x [ p l , X2], and the smaller rectangular box
is Br x [ X i , Xz].

1.6 Notes

1.1 The proof of the classical implicit function theorem can be found in
many standard textbooks; see, among others, [Chen, 1981; Chow and
Hale, 1982; Nirenberg, 2001; Zhong et al., 19981.
1.2 Both the Brouwer degree theory and the Leray-Schauder degree theory
can be found in classical books, including, among many others, [Chen,
1981; Krasnosel'ski'i, 1956; Nirenberg, 2001; Zeidler, 19861.
1.3 The standard Lyapunov-Schmidt procedures are discussed in [Niren-
berg, 2001; Chow and Hale, 1982; Golubitsky and Schaeffer, 1985;
Golubitsky et al., 19881.
1.4 Theorems 1.10 and 1.12, are due to the pioneering work by
[Krasnosel'ski'i, 19561. Theorem 1.11 is due to [Crandall and Rabi-
nowitz, 19711. Theorem 1.13 is proved by [Clark, 19751, and is gener-
alized by [Rabinowitz, 19771.
1.5 The Rabinowitz global bifurcation theorem, Theorem 1.14, is due to
[Rabinowitz, 19711.