IEj(z.
k" .�)-Ej
(z:
k" .�)I �
Kz(E; )Ilz - z*lIlX.}
1F;(
t.
z. �k 'Cr) - F;(t.
z: �k ' Cr)1 � Kz (F;)lIz - z*lIcx . (4.3.14) For the purposes of our existence proof. we firstly look at the monotone system'�n . Bn
where we assume our coupling functionsJi andFj
are monotone nondecreasing in Cj andCj.
respectively for all }.4.3.1 Upper and Lower Solutions and Monotone Iteration
We shall now introduce the concepts of upper and lower solutions relative to the monotone system
Sn. Bn .
Definition 4.3.1 .
Assume that
(i) For components
i
E I. where Di > O.fj
andCj
are continuous functions in.Q
x A with continuousfirst order Xj derivatives in
.Q
x A and continuous second order Xj derivatives in.Q
x A ;(ii) For components
i
E I. whereDj
= II i = O. fi and C, arc continuous functions in £2 x A ;(iii) For components
i
E 1. where�j
> O. C andCj
are continuous functions in if. with continuous first order Zj derivatives in if and continuous second order Zj derivatives in A;(iv) For components
i
E 1. where�i
= O.u ·
'\lCi $0. �i and Ci are continuous functions in A . with continuous first order Zj derivatives in A;4.3 EXISTENCE OF SOLUTIONS TO THE STEADY STATE PROBLEM 1 05
The ordered pair of functions
(fi' �i)
and(ci' Ci)
withfi
�Ci
on nx A
and�i
�Ci
on If arc said to belower
andupper solutions
ofSn' Bn
respectively, if Ihey satisfy:and -D
iV;f
i � h(x, c;.;) in,axA,
dC·
-=:!.. � 0 ond,alXA,
dn
d
c·
D -=:!.. < H (C· - c· ) ond,a2XA
Ian
- ' -' -I '-!llV2 c.+u· VC.+ H.
I _ I _ _I IJ
aDz(
_IC
.- c-I -)
� F( I z, (;:J') _ inA,
-D;V; Ci
� h (x, Cj) in.(lxA,
de·
� 0 ond,alxA,
dc·
- _ D·I -dn
' - I >I-f.(C.
I-c·
I·)
ond,a2XA
,VC+ l-f.J
(E -c »
F(zC )
inA
I I I aD2 I I - I ' J '-
dE
VI C·+
I9).
I __ , >VIC.
I ondAI
dnl
- I. 'dC·
__ , � o ondAa,
a = 2, 3,dna
respectively.We nOLe from the counterexample at !he end of section 3 . 1 , !hat comparison !heorems analogous to Theorems 3 .2. 1 1 and 3.2. 1 2 do not hold in general in the case of the corresponding steady state or time independelll problem
Sn' Bn .
Hence, if there exist(fi' �i)
and(ci' C;)
which arelower
andupper
solutions of the steady stale problemSn' Bn
and(Ci, Ci)
is a solution ofSn' Bn ,
then in contrast to the unsteady state problemSn, Bn,
we cannot assert thatfi
�ci
�Ci
and�i
�Ci
�Ci .
However, the method of monotone iteration is still applicable and shows !he existence of at least one solution
(Ci, Ci)
ofSn' Bn
lying between(fi' �i)
and(ci' Ci) ·
Lower and upper solutions may not always exist for elliptic equations. Therefore, as a result of this, certain unstable solutions cannot be obtained by monotone iteration (PARTER [230] , KELLER and COHEN [ 1 39], AMANN [9]). However, it must be noted that as with parabolic equations (PAO [222]), there are geometric conditions which the nonlinear reaction functions f; and
Fi
may satisfy which guarantee the existence of either lower or upper solutions for elliptic equations (AMANN [9]). These lower and upper solutions may not necessarily exist simultaneously.As for the system
Sn,
Bn, it is important to note that the lower and upper solutions provide lower and upper bounds for solutions of,�'" Bn
which can be improved by monotone iteralive procedures.4.3 EXISTENCE OF SOLUTIONS TO THE STEADY STATE PROBLEM 106
In order to establish an existence theorem for
Sn' Bn
in terms of upper and lower solutions, we define a transformation !!i, byand consider the sequences
{(c�k), Cfk» }
wherec�k)
is obtained from the linear systemand
Cj(k)
is obtained from the linear systemC(k)
- C
a
VI,
' + VI ' I on'"
ac(k)
-a '
= 0 onaAa,
a=2, 3, na 'th <(k-I)
< - dC
<C(k-I)
<C-
f' k - 1 WI £j _ Cj _j _ j _ or - , .... (4.6.15) (4.3.16) (4.3.17) (4.3.18) (4.3.19) (4.3.20) (4.3.21)For each k, the system (4.3.16) consists of
n(J)
linear, completely uncoupled boundary value problems with boundary conditions given by (4.3. 1 7)-(4.3. 1 8) and this system is uncoupled from the system (4.3. 1 9) which also consists ofn(J)
linear, completely uncoupled boundary valuc problems with boundary conditions given by (4.3.20)-(4.3.21).Since c
fk)
(x, z) is not differentiated with respect to z in (4.3 . 1 6),A
may be considered to be a parameter space in (4.3 . 1 6)-(4.3. 18). For functionsc[k)(x,
z) whereDj
=Hj =
0, n may also be considered to be a paramctcr spacc and for functionsC[k)(z),
where9)j =
0, U ·vcfk-1)
== 0, wc may similarly u'catA
as a parameter space. The cxistence and uniqucness of sequences{(c?), Cj(k» }
may therefore fol low from solving standard scalar systems of linear elliptic equations (LADYSHENKA YA [ 154] or GILBARG and TRUDINGER [ 1 06]) which may or may not depend on parameters, systems of first order partial differential equations which may depend on parameters (LAKSIIMI KANTHAM etal. I I GO]) and systcms of algebraic
equations in many variables.The nonlinear algebraic equations h (x, Cj )
=
0 obtained whenDj
=Hj
=°
may be expressed as Cj
= fi(x,
C j)+ Cj
in order to perfonn functional iterations to findCj
in terms ofCj.
We shall develop a general theory for a broad class of monotone iterations involving such algebraic equations. This class of iterations includes Newton's method as wcll as a family of mCI,hods, which arc Ncwton-Gauss-Seidel processes(ORTEGA and R I IElNBOLT 1 206, 207J, LADDE et
al.
1 1 53, p. 36 1). It is the Lipschitz propcrty that may be used instead of differentiability in many other iterative methods. Note that fi (x, Cj )+ Cj
satisfy the Lipschitz and HOlder continuity properties that are assumed on our original fi (x,Cj)
as well as the monotonicity in Cj (see Lemma 4. 1 . 1). We may therefore rewrite (4.3 . 1 6) asCfk)=
h (X,C;k-l» +cfk-1)
when D, =Hj
=O .
The existence and uniqueness of sequences( c?) }
follows from the uniquely dcfined solutions to algebraic equations.4.3 EXISTENCE OF SOLUTIONS TO THE STEADY STATE PROBLEM 1 07
The rest of these theorems will require Holder continuity properties on the functions !;(x,
C)k-I»
and F (z I,
J IC�k-I)
I which are satisfied if eitherC(k-I)
J ECl+a,a [{),
xA Rn(l)] C(k-I)
' ' J ECa,a
[{),
xA Rn(l)]
, , J ECI+a[A Rn(J)]
' orC
J (k-
I)
ECa[A Rn(J)]
, .These theorems will also require Holder continuity properLies in the boundary conditions and so we make the following assumptions on
Ci,l'
We will assume thatdQ
anddA
belong to classC2+a.
(H' ) 3
C.
1,1 ECI+a[A Rn(J)]
' •The velocity distribution vector function u(z) is also required to satisfy the following Holder continuity property
where
UI(Z)
is chosen without loss of generality to be the first component that is nonzero. For components i E J, where9Ji
= 0,u·VCi
¥'O, we shall also need the following additional assumptions(Hs) Assume that
A
==A I
XA
n-I , whereA I
E R.(i) For each
(ZIO,
zo) EA
I XA n-l ,
there exists a unique solutionZ(ZI, ZIO,
zo) ofdz �z) -
- == -- ,
Z(ZIO)
= zo, onA I,
dZIul
(z)where
ZI
correspondes to the nonzero component UI(Z);(ii)
Z(ZIo ZIO, zo)
is continuously differentiable with respect to(ZIO, zo);
(iii) The relationshipholds.
(H6) Assume that
A I
is the interval [a, b](i) For each
Zo
EAn-I
andYiO
ERn(J),
there exists a unique solutionYi(Z\,
a,YiO;
zo
) ofF( (
) C(k-I» H
sYY. HJ
(k-I)(
( »}
dJi
== IZ), Z z), zlO, zo '
J- I I + I
aaz ci z
lo
X,Z Zio zlO, Zo ,
dZ)
UI(z)
Yi(
a) =YIO,
on
AI,
whereZ(ZI, ZIO,
zo
) is the unique solution of (4.3.22);(ii)
Yi(ZI,
a,YiO;
zo
) is cont.inuously differentiable with respect to(YiO,
zo).(4.3.22)
(4.3.23)
4.3 EXISTENCE OF SOLUTIONS TO THE STEADY STATE PROBLEM 108
Note that assumptions (H3)-(H6) will hold in either our original system