ItIXTHEMhTICAL
263
BIOSCIENCES
Existence Theorems for a Nonlinear Partial Equation of Viscous, Incompressible Flow A.
R.
Differential
ELCRAT
Mathematics
Department,
Communicated
by
Indiana
Cniuersity,
Richard
Bellman
of a solution
is shown
Bloomington,
Indiana+
ABSTRACT Existence partial
differential
incompressible in large adheres
fluid
vessels
[2].
a Banach
The problem
Solutions
a generalization sequences problems
1.
of the tube,
stating
that
which
is intended
boundary
there
may
any two
condition
be many
regular
by
Altman
elements
are
applying ;9].
solutions
through
is guaranteed
by
the
These
imposed
operator
Frkhet
Kantorovich-Newton are defined
Friedman
a theorem
asymptotic
continuously
flow
the fluid
in the equation
boundary-value of
blood
is that
however,
in L” maps
differen-
theorem
or
by convergent
problems
second order are uniformly
theorems
of a viscous,
to represent
are time
solutions
of
for a nonlinear
flow
solutions;
solutions
space Y and is twice
are obtained
and their derivatives
existence
tube
problem
in time-dependent
is set so that the differential
of it by
whose
velocity
the only
space X into a Banach
tiable.
Their
Since
for a boundary-value
for the
in an elastic
[l].
to the wall
has been proven norm
equation
for
linear
HZjlder continuous.
[4].
INTRODUCTION
The goal of this paper is to prove nonlinear
partial
differential
equation
work in [l] and [2] on mathematical Only “classical” derivatives values
solutions
*
for a certain
comes
out of the
that is, ones which have continuous
in the equation
and take on boundary
continuously.
In the past it has been found relatively classical
theorems which
models for blood flow in large vessels.
are sought,
of all orders appearing
existence problem
solutions
Present
address:
of nonlinear Wichita
partial
State
0
1965 by
equations,
but
of
often
University. Mathematical
Copyright
easy to prove uniqueness
differential
American
Biosciences
Elsevier
2, 263-292
Publishing
Company,
(1965) Inc.
very difficult
to show existence.
have weakened theorems.
As a consequence,
the sense of existence
Unfortunately,
uniqueness
theorems
have not always been possible to obtain, about the relationships there
reasons
about
problems
the physical to weaken
and, perhaps,
of a smooth
thus make existence
asymptotic
uniqueness”
incompressible section
11 i has
squared
conditions
tion that the fluid adheres (linear equation)
explicit
series
integro-differential even though that
this approach
operator. solution
square
Thus, the complete
Existence ential
with
and applying
theorems
initial-boundary
These
solutions
are twice
to this one.
which lies behind
an
are proven here.
with in 11:
However,
it must
described
above.
is that
value problem,
It is
the work in
is dealt
that is, the one which
unique (in the usual sense),
up the problems continuously
equations
so that the differ-
Frechet
of L. V. Kantorovich
are constructive,
are constructed,
considered
he obtains
to an unbounded
sense, since he knows
uniqueness”
to make a solution
of operator
* The suggestion problem
In the case of a rigid
for nonlinear
differentiable,
or one of 31. Altman operators
space into another by means of a Newton-type
theorems
condi-
here.
involved
on the solution
dictated
gradient
iteration
the sense of uniqueness
either a theorem
one real Ranach
he avoids imposing
asyvmptotic”
problems
will be shown by setting
operators
pressure
square
asymptotic
has enough such conditions
over a fixed tube
goes to zero as time goes
form of uniqueness
Picard
square
tube flow of a viscous,
in the conventional
to fluid dynamic “Mean
in mind that
is not dealt
of “mean
He is then able to use this special solution,
is “mean
and 121, and only existence be kept
easier to prove.
idea
to the wall of the tube.
it is not unique
any other
this paper.
solution
the
other than the physically
by adapting
in such
to be dropped,
of pulsatile
with an analytic
solution
For these
conditions
of two solutions
By using this weakened
any initial-boundary tube
of these solutions below.
fluid, this being simply that the integral
of the difference
to infinity.
introduced
in his study
In addition,
the sense of uniqueness
and initial
M. Lieberstein
solutions.
significance
a way as to allow some boundary H.
to go with these results
such as the one considered
it seems worthwhile
in the field
and there has been some confusion
between various generalized
is some question
in fluid dynamic
workers
in order to be able to prove existence
in that
the elements
of trying subsequently
sequences
converging
of these sequences
some Newton-type was made in [l,
procedure footnote
mapping
procedure.* to the
being solutions
on the fluid dynamic 1.7, p. 751.
EXISTENCE
of linear equations. of nonlinear
This represents
equations
of its variants an element
where
is employed,
for finding a solution
the
and no method here.
a marked Schauder
of approximate
theorem
with this matter,
existence but they
partial
the solution
are not pursued
which will be expected
THE
PHYSICAL
Suppose tube
made
straight blood
FIG.
will show that
is
for dealing In addition,
the physical
problem
in view of the minimal
PRORLEM
the flow of a viscous,
of compliant
through
this problem
here
equation,
imposed.
we observe
section
is provided
differential
in this paper.
may have many solutions,
”
On the other hand
possibilities
theorem
conditions
or one
a method
of this linear equation
will suggest
the proof using Altman’s boundary
theorem
solutions
for a linear
proofs
over treatments
does not provide
once it is known that one exists.
for approximating
The
advantage fixed-point
since that theory
of the sequence
only by an existence given
265
THEOREMS
material,
restricting
a large blood vessel, is referred
fluid in a
our attention
This is intended
of this tube.
the reader
incompressible to represent
to a fairly the flow of
but for the physiological to
setting
of
[I-31.
1.
The tube section in E3.
and z axis are taken perpendicular denoted
at time t is thought
The axis of the tube section as rectangular
to the x axis.
Cartesian
The lateral
by S(t), and the cross section
the x axis at x by A(x, t).
Various
have to be made subsequently,
of as a domain
is taken
R(t) contained
as the x axis and the y axis coordinate
surface
axes in a plane
of the tube
section
cut by the plane perpendicular
assumptions
but it is assumed
on the regions immediately,
is to
R(t) will for clarity
266
A.
of exposition,
that
R(t)
which the divergence surface
is a simply
theorem
connected
holds),
that
S(t) from the x axis is given by a function
differentiable
on its domain,
dimensional
Green’s
imaginable
region bounded
a smooth
on S(t), rubber
As in [lj driving
[2],
the pressure
Hence,
and U, v, and w denote equation
gradient
if fi denotes
the differential
the function
u’w, +
z’wuy +
(constant)
of viscosity.
density
provides
components
of momentum,
in with
(1)
v>,y+ v,,) = - Py
f-qw,, + w>,,,+
of the fluid and ,U is the
[8] f or a derivation equations.)
the boundary
of these
(constant) equations,
To go with these equations condition
the fluid and the wall.
surface of the tube section has negligible
(2)
WI,)= - p,,
that the fluid be at
to the wall at the wall at all times,
be “no slip” between
values
of motion,
ww*)-
(See
of as the force
w, = 0,
8” +
called the Navier-Stokes
viscous flow theory rest with respect
of
of mass),
d(v, + uvx + vvy + WV,)- p(vxx +
usually
assumption
if he thinks
giving the pressure
the velocity
conservation
(conservation
equations
is thought
giving
respectively,
26, +
d is the
two-
for all x and t.
to be given by the functional
the functions
the x, y, and z directions,
coefficient
on the
tube.
of a given function.
where
curve,
nearly any smoothness
and the reader will not be led astray
and
d(w, +
(one for
which is continuously
by a Jordan
the fluid, and it is assumed
the continuity
region of points
and that A(x, t) is a simply connected
Of course, the physical problem justifies
yields
Green’s
the distance
R. ELCRAT
that is, that
In particular,
there
if the lateral
motion, this gives us the boundary
conditions u=v=w=C) We might However,
use Eqs.
(l),
on S(t)
(2), and
a mathematically
(3) as a model
simpler system
a model for blood flow that is subject assuming
the pressure
to be nearly
that p, and p, are negligible, of velocity,
for all t.
is needed in [l] to formulate
and assuming
2. 263-
292
(1968)
verification.
in cross sections that
v and W, are such that terms involving
~Wathematical Biosciemxx
for our flow problem.
to experimental constant
(3)
the radial
Thus,
A(%, t), so components
them and their deriv-
EXISTENCE
atives
267
THEOREMS
can be neglected,
conservation
of momentum
implies
the single
equation +,
+ SbU,) - /-+,,
for the axial component The
“no slip”
+ GYY+ u,,) = -
of velocity
condition
P,,
(4)
where p, is a function
of x and t only.
now becomes on S(t) for all t.
‘U = 0 (For further
discussion
of the physical
components
of velocity
and pressure gradient
and [2].)
Proving
existence
main
of this
paper.
task
Although
in Eq.
of a solution
velocity
term
W,
appears
axis
is thought
behind neglecting
radial
the reader is referred
to [l]
of Eqs.
(4) the cross components
just as they are in Poisseuille the axial
reasoning
varies
(4) and (5) will be the
of velocity
flow and in pulsatile
with the axial
in the equation. of as arising
(W
distance
so that
This variation
from
the
are neglected
flow in a rigid tube of velocity
pressure
relief
[l],
the nonlinear along the
due to the wall
compliance. Comparing problem
Eqs.
(4) and
for a parabolic
hope for a solution boundary
values
the solution
(5) with
equation,
of Eqs.
the
initial-boundary-value that
(4) and (5) to be unique.
at the ends of the tube
function
usual
we see immediately sections
there
In fact,
and initial
would have to be prescribed
to obtain
is no
additional values
for
uniqueness.
This is just the sort of thing which will not be done.
There are two reasons
for this.
values other than the
First,
physically
imposition
dictated
is not desirable
of initial
condition
and boundary
that the fluid adhere to the wall of the tube
in the model
of physiological
phenomena
the work in this paper because such data are virtually experimentally. boundary-value
Second, problem
in the Introduction; to treat Thus, Eqs.
existence
(4) and (5).
existence
for
impossible the
that is, for the complete to attack
in the
to obtain initialdiscussed
problem it may be necessary
sense or by a nonconstructive the incomplete
problem
The price paid is that the weakened
was discussed
inspired
complete
for Eq. (4) may lead to the difficulties
in a generalized
it seems prudent
which
proving
that
Introduction
must
method.
consisting
of
sense of uniqueness
be adopted.
3. PREREQUISITES We
now introduce
the subsequent
notations
and give results
that
are prerequisite
to
work. Mathewzatical
Hiosciences
2, 263-292
(1968)
A function
p defined
Holder continuous one such
on a region W in (x, y, z, t) space is uniformly
with exponent
SCif u is a real number between zero and
that
b(P) -
1< const. [W, Q)l", P,QE~,
v(Q)
where
d(P,Q) = [(x - x’)~ + (v - Y’)~ + (2 - x’)~ + jt - t’ll’P if P = (x, y, For
such functions
2,
q,
It can be shown
(see [4, Chapter
denoted
is a Banach
C,(W)*,
We denote
Q = (x’, y’, z’, t’).
and
t)
31) that
space
the set of all such functions,
under
the norm
th e set of all functions
by CzTn(@
q defined
on L@ such
that
It can be shown (see [4, Chapter
31) that C, t,(9?)
is a Banach
space under
the norm
The following proposition in later + with the
sections:
Tills is slightly two different
notation
a somewhat
on extension
A function different types
C,(a) different
from
of Banach
f uniformly the notation
will be used free15
Hiilder continuous of Friedman
spaces of HGlder
for the space described space.
of functions
above
on a set in
in [4], where
continuous
functions.
and the notation
he deals He uses
used above
for
EXISTENCE
269
THEOREMS
(x, y, z, t) space can be extended to the whole @ace in such a *way that it is still uniformly A proof
Hiilder continuous
can be found
in
and its norm
[7, Chapter
The regions 3 which are considered and are contained simply
connected,
and t = t,, respectively, for t, < f < t,.
will be bounded regions
2 depicts
2. Composition
such
n {(x,
2,
a region
of the
where h is a function
.Y
We see that
tz).
g.
space-time
region
H.
E if there is a four-dimensional
?’ = h(.r, 2, t),
all of whose derivatives
and are Holder continuous
(exponent
a).
by one of
x = h(y, z, t), through
second order exist
9 has property
E if every point
has it.
The nornzal boundary of a region 3’ such as the one described is the
of
on t = t,
manifold
V of the point such that V fl 9’ can be represented z = 12(x, ‘J’, t),
of 9’
R(t,)
regions as well.
z) E aR(t), t, < t <
t)I(x,y,
A point of .Y is said to have property
-
by the closures and
y, z, I) 11= t>
three-dimensional
9 = {(x, y,
neighborhood
K(t,)
t = constant,
and by an open three-dimensional
will be simply connected,
FIG.
two hyperplanes
The sets
R(t) = .&
Figure
They
three-dimensional
not been increased.
in this paper are simply connected
in the region between
say t = t, and t = t, (tl < tz).
If 1, has
71.
set
R(t,) U .‘9’ (the base
and side of the
“cylinder”
above
depicted
in
270
A. Ii.
Fig. 2). As a consequence [4, Chapter
should be prescribed For
of the maximum princi$le for parabolic
21 it is this manifold
example,
the maximum
principle
y=o
zero in 9?‘, so that
$ defined
C, +a if there
a solution
functions,
on the normal
is a function FE
such that
function unique.
if
boundary
on the normal
and g are continuous
A function class
that,
equation
of 9,
of
in 99,
P=g
f
implies
on the normal
VAT-vp,=f
where
of a parabolic
in :‘R,
vdrp-q+=O
then p is identically
equations
on which values of a solution
to make the solution
ELCKAT
IJ = !P on the normal
boundary
of 3,
is unique. boundary
of 9
is said to be of
!P, C,+.(@, boundary
of W.
For such functions
VW
define
I*1
2+a
inf{lYl,+,
--
j YEC2+.(B)
and Z+!I = Y-’ on the normal The following theorem For
two theorems,
for the heat equation,
proofs, THEOREM
see
[4, Chapter
1. Suppose
boundary
an a priori inequality are fundamental
of W}. and an existence
to the work in this paper.
41.
that 9! has property
is of class C, f 1 on the normal boundary of W.
E, that
f E C,(S),
a?zd $
Then, if y is a soktion
of
in M,
v&--t=f
on the normal boundary
P=#
of &‘,
and y E C, +.(9?), there is a constant B depending only on v, u, and 3? such that
bl Mathewmtical
Hzoscienc~s
2+a
2, 263-292
G BW,. (196X)
+
IfI,).
EXISTENCE
THEOREMS
THEOREM
2.
271 W has property
Suppose
E, f E C,(W),
and f is zero on
Then there is a (uniqzhe) solution of the problem
ai?(
in 2,
yAP,-vi=f
on the normal boundary
cp=o
of 99,
und this solution is an element of C, ,.(92). In later sections we will investigate from one Banach
space to another
them and linear operators precisely,
suppose
continuously Then,
of nonlinear
operators
connections
between
which are local approximations
P maps
an open
into the Banach
subset
space
if there is a linear operator
linear operators
the properties by establishing
A of the
Y, and that
L in B(X;
to them. Banach
More
space
x is an element
X
of A.
Y), the space of continuous
from X into Y, such that P(x + 1%)= P(x)
+ L(h) + N(h),
(6)
where
we say that
P is Fre’chet differentiable
x, and we call L the F-differential F-differential denoted
is unique if it exists
as P’(x).
Then
the Eq.
that
satisfies
Eq.
In a normed the set of points
an operator
F-differentiable
at
It can be shown that the
(see [S, Chapter
81).
In this case L is
(6) can be written
J’(x + h) = P(x) where o(h) denotes
or, more simply,
of P at x.
+ P’(x)(h)
defined
+ o(h),
in a neighborhood
of zero in X
(7). space the “segment” a + E(b -
I\ proof of the following
a) where
important
connecting 8 varies
proposition
two points
a and b is
over the interval
[0, 11.
can be found in [6, Chapter
Xl. THEOREM
3.
Suppose
that P maps an open neighborhood
A of the
segment S connecting the points a and b of the Banach space X’ into the Banach space Y, and that P is F-differentiable
at each point of ,4.
Then, for each
3’ in A IP(b) -
p(a) -
P’(y)(b
-
a)\\< [lb ,~Iathematical
all 2~
IlP’(x)
Biosciemxs
-
P’(y)Il.
2, 263-292
(1968)
Suppose
now that
the mapping
1’ is F-differentiable
x + P’(x)
is F-differentiable
of A into B(X;
at a point
While this mapping
x we denote
is an element
with an element
of 23(X, X;
from XxX
into
Y, since H(X;
isomorphic
(see 16, Chapter
Fdifferentiable into
B(X,
X;
at each Y)
differentiable,
at each point of A and that
Y) is continuous.
of B(X;
13(X;
point
is continuous,
H(X;
Y))
as P”(x).
Y)), it can be identified
Y), the space of continuous 51).
If this mapping
its F-differential
and H(X, X;
bilinear
operators
Y) are naturally
Thus, it can bc shown that if I-’ is twice of A and the mapping that
is, if 1’ is twice
x -
P”(x)
of A
continuously
F-
then
P(x + h) = P(x) \vhere the last neighborhood
term
+ I”(x)(h)
on the right
of zero in X that
+ JP”(x)(h, represents
vanishes
h) + O(ljhJz),
an operator
more rapidly
than
defined
in a
I~Jz]~~as lIh!l
goes to zero. \Ve are now ready to state the following found
in [5, Chapter
‘THEOREM 4 (Kantorovich). the s@ere
A = (x 1 Ijx -
Sf&bose
The proof is to be
P is a continuous
ma$fiizg
v/
xOjj < I?> in the Banuck @ace X &to the Banach
s$ace Y, and that P is twice continwusly A, = (x ! 11%-
theorem.
17j.
x0:] < 71, where Y < R.
(1)
[P’(xa)J-i
exists,
(4)
l[P”(X)lj
.2:
in
diJferentiable 012the closed sfihere
F?lrtllrr,
slcfiposc that
a-1.
‘I‘lzela, if h = h’B2q < 4 and Y,, = (1 I’(x)
111 -
2/a)Bq/h < Y, t/w equation
= 0
has a solzltion x* z~t the sphere {x ) l/x0 ~ x;j < Yo} to which the seqccence
I -‘cn; 1 = %, converges.
fn
addition,
the sepence
P’(~“)l-lwc,)
EXISTESCE
is
273
THEOKEMS
well defined
an.d converges
The
to x*.
following
ewor
hold:
bounds
II%* - x,,jl < (Zh)2”?gz2”, and
the solution
x*
is unique
in
A,.
The process of forming the sequence the Newton Newtolz
space that
be a one-to-one Y.
above
{x,}
of forming theorem
mapping
in the above theorem the sequence
requires
that
of the Banach
We shall give a theorem
P’(x,,)
{xl]
is called
the modified
the linear
operator
space X onto the Banach
below that
removes
the restriction
be one-to-one.
Supposing space
The
method.
P’(xJ
and that
method,
X, to be a closed linear set in a normed space X, the vector
of cosets
X/X,={x/x=x+X,,xEX} is a normed
space
with
the norm
defined
and it can be shown that X/X,, is complete suppose
that L is a bounded
X onto (not necessarily
b>
if X is [5, Chapter
linear operator
in a one-to-one
mapping
manner)
121.
Thus,
the Banach
space
the Banach
space Y, and
let
x, = Then, let 9
be the linear operator
L-l(O).
from X/X,
onto Y induced by L, that
is,
q-q 9
is a bounded
in a one-to-one
= L(x)
linear operator manner,
where
mapping
so the bounded
XEX
one Banach inverse
;IlatWematical
space onto another
theorem
Nioscie~zces
implies 1, 263-
that
the
292 (1968)
_\. Ii. ELCRAT
274 inverse
mapping,
is a continuous
Z-l,
P to be a continuous space
X into
(possibly
the Banach
linear
nonlinear)
space
Y,
operator.
operator
that
P
Now, suppose
mapping
the Banach
is F-differentiable
in the
sphere S(X,, 7) = {x j 11~~- ~11< Y>, and that L = P’(x,J
maps X onto
mapping
Y induced
Let
X/X,
onto
E be an arbitrary
positive x1 =
There
is an element
-
there
in this
number.
be the operator as follows:
Set
Set
/ 1x1:i < j IX1i 1 + F.
x-lP(x,
+ Xl).
ljxz -
x1/( < jlX_salready
Zr(j(l + E).
If
been defined
let Xn=
The method
2
we proceed
X1, . . . , X,,_ 1 and xi, . . . , x,, _ 1 have
manner,
and choose
letting
9-lP(x,).
is an x2 in X2 such that
the elements
Then,
x1 in Xi such that X2 = x1 -
Then
Y.
by L as above,
Xn_l
-
Z-lP(x, + x,,_ 1)’
x, in %+,such that
used in forming
the sequence
{?c,, + xx} which is defined
in
this way will be called the underdetermined modified Newton method. Theorem 5, which gives conditions from
for the convergence
of this sequence,
[9].
THEOREM
5 (A&man).
Szt$$ose that I’ is as above, that
and that
IIP’(x) -
P'(x,)//
for
XES(X,,
Then, if BC
Bioscimces
and
Lzrj< r(1 -
2, 263- 292 (196X)
K),
Y).
is taken
EXISTENCE
275
THEOREMS
the sequence {x,, + x,} to a solution
defined above is contained in S(x,, Y) and converges
x* of the equation P(x) = 0.
This solution
x* is an element of the closed sphere S(x,, Y).
\Ve shall prove
the following
our later needs using essentially Suppose
6.
THEOREM
simpler
theorem
Altman’s
proof
P is a continuously
which is tailored from
to
[9].
F-differentiable
ma$$Gag
of a Banach space X into a Banach space Y, that P’(0) maps X onto Y, and that IlP’(x) -
P’(y)11 < IIx -
Let dp be the mapping
of X/X,
y1i
for all x and y i~z X.
onto Y induced by L = P’(O), and assume II-Lp-ljI < B.
Then, if r is chosen strictly less that l/B
and if
IlP(O)//< “17< ~(1 the sequence {xn}
defined previously
Br)IB,
(x0 = 0 here) is contained in S(0, Y)
and converges to an element x* of S(0, V) which is a solution of the equation P(x) = 0. Proof:
Since 1
Br< there is a positive
number
Hr(l + E) < 1 We use this
E
and
to define
Brj < r(l -
Br),
E such that and
By + e < ~(1 -
the underdetermined
Br(1 + F)).
Newton
sequence
{x~>.
We have
jlXllid Ij%ll+
E<
Ij~-lll /lp(o)/I+
E<
+
+
E<
r(l - BQ + E)) < y
(8) so that x1 E S(0, Y).
Recalling
the definition Mathematical
of z,,, we have Biosciences
2, 263-
292
(1968)
270
a\. Ii.
x2 -
$1 = 21 -
and,
applying
9-l~P(x1) Theorem
ljxg- x11/d 11x2 - q(l
- P(0): == - z-‘[P(x,)
-
P(0) -
BLCR.1-I’
P’(O)(x,)],
3,
+ &I < (1 + 4B
IIWx,) - ~‘(O)I~ * +&
SUP lJ:.:bI
-wl
<
Let
cc denote
Kr(l
jjxzii < 11x2-
+ 8).
Then,
+ ~)llXll,.
using Eqs.
xl/j +- (1x1/1< (1 + 4 = (1 -
so that x, -
so that
xz t S(0, r). x,* __1 = -
For
n larger
22-1 [I+,‘
J -
than I’(.x,,
(8) and
(0) (!I),
lxllj < (1 + a)(1 -- 4~
r-f)y < Y, two the definitions .1) -
I”(O)(X,~~ 1 -
imply
N,,_:!)l,
EXISTENCE
so that limit,
THEOIIEBIS
{xPL} has a limit
0 =
and
x* in S(0,
we
-
lim Z-~P(.X~ 71+ rzI
Remark.
By
see that
11P(O);l).
maximizing
the
choice
Y =
th is case
In
l/(SS) the
Eq.
(12)
gives
approximate
4. SETTING
theorem
us the
_FIP(x*),
earlier,
precisely,
K(t)
solutions
our main
P(X*).
YE (0, l/W, restrictive
on q (and,
hence,
on
l/(4.32). bound
x,Il <
jb’Y(1 + F)]‘lY
{ x~}. PROBLEM goal in this paper
of the boundary-value
problem
is to prove
consisting
the existence
of Eqs.
(4) and (5).
if g
where
-
requires
error
THE MATHEMATICAL
As stated of a solution More
{A?~} also has a
function
is least
i/x* the
=
LZY)/B,
7j <
for
_ I) =
9(O)
f(Y) = Y(l -
Blso,
the sequence
have 0 =
we
Then
Y).
so
= {(x,
y>z, t) 1(% y>2) E n(t), t, < t e tz},
is the tube section
described
in Section
Mathematical
2, we wish to prove
Biosciences
4,
263 - 292
the
(1968)
278 existence
of a solution*
(4
ZLt +
(b)
U= 0
ZLU,
of Y(U,, +
~
on S(t)
IL,.).
(13)
for all t,
where / = $,/d and S(t) is the lateral surface of R(t). this by proving space-time
existence
region,
\+‘e suppose
and
that
of a function which
in .9,
UJ + i = 0
+
\Ve shall accomplish
which has as its domain
solves
Eqs.
(13) when
a larger
restricted
to 2.
the manifold
.Y’, = {(X, y, 2, t) ~(x, y, z) E S(t), t, < t < tZ} has Property
E at each of its points,
We make two successive in a “spacelike” extension,
manner,
which
is assumed
ends” to the three-dimensional
region,
to be such
which we shall denote that
9, = ((x, y, has Property
2,
E.
A&, is accomplished
between
region 9, the first manner.
The first
by adding,
region R(t),
as R,(t).
t) I (x,
3’,
at
to form
This extension
z) E aR,(t), t, < t < &} that is, so that the space-time
The second extension, by adding
a “finite
from W, to a space-time history”
region region
to .%?I. More precisely,
92 = {(x, J’> 2, t) 1(x, y, z) E R,(t), t, < t < tz}, where
of C,(W).
the manifold
E at each of its points,
B1 has Property
of the space-time in a “timelike”
as -tiI, is accomplished
WC shall denote
each time t, “smooth an “ellipsoidal”
extensions the second
and that f is an element
to -c t,,
R,(t) is a simply connected three-dimensional region for each t t, and t,, K,(t) = K,(t) for t, < t < t,, 2, has Property E, and
the set
* See the
first
paragraph
of the
Introduction
EXISTESCE
THEOREMS
279
where
is a cylinder with respect to the time coordinate. of W are depicted
in Fig.
The successive
extensions
4.
FIG. 4.
The above extensions theorems linear
of &? are made in order to be able to apply the
given in Section
equations
3 on solutions
in our later
ends” to the tube sections that
have Property
preventing
work.
of boundary-value
In particular,
problems
addition
for
of “smooth
is done in order to work in space-time
regions
E, the sharp edges at the ends of the tube sections
this condition
from being fulfilled
by 92.
The addition
of the
part of giz between t, and t, is made so that we may make a (mathematically) necessary
assumption
about
f
is an element
of C,(&)
and that theorem,
It is clear now that
a function
IfI,
(a)
L’t+ vu, -
(b)
v=o
a function
f
f to
imposing
this restric-
92s in such a way that
is zero on aR,(t,,) without
Section
u which is a solution
Y(v,,X + zj,, t
increasing
3).
a,,) +
f= 0
of the problem in W,, (14)
on 9,
is such that its restriction of such
without
In fact, we extend
(extension
its norm,
f
the function
tion on the physical problem.
to W is a solution
v is what
we shall
of Eqs.
(13).
subsequently
Mathematical
Biosciences
The existence
establish. 2, 263-
292
(1968)
280 5.
.A. R. ELCRAT
PRELIRIINARY
REStlLTS
For ZI an element
of C, , .(A$) we define
P(u) = PR0~0sIT10ii
1.
7ft
+
P(26)
IUf,
iS
-
Y(?f,, + uv,. + u,,) + f.
UTZ &W6&
Of
c&).
PYOO!: It suffices to show that gh E C,(.Z,) if g and b do. We note that
so that
where
and
the result PROPOSITION
Proof:
is proven. 2.
P maps C,.,_,(9,,)
into C,(W,)
continuously.
IA “6°ECl!+1(‘%3,
and let t: be a positive
number.
121- ZZO;Z+n< 6 Since
me have
We seek 6 = a(.~, u”) such that
implies
Iw
-
P(u”) lx < E.
EXISTENCE
281.
THEORERIS
TO deaI with the second term on the right we note that (fqx so that,
(ZP),
= (U2 -
zql
= (U,+-
z&26 + 240) -t (24 -
ZP)(ZL,+ U(1).
using the fact that Ikl, G IflA&
we have 1(2b2)xFinally,
(ldo2)xi2 <
(‘t41z +
using the fact JR -
it follows
4,
1U”/z)114,-
24:1, +
(jU&
+
~Z~~~J~~t”-
%I,.
that
< c
implies
that
:gj, < IhI, + c>
that 1P(UO) -
P(U) ja < 62 + S(1 -+ 3V + IZLOlq + lzqa)
when ever 1~4~ - 7ii2 +_r< 6. The
proof
is completed.
PROPOSITION
Proof:
3.
P is F-differentiable
at each point of C,l .(a,),
with
By definition P(u + 9) = W
+ L(q) + h(p),
where
We must first show that L is an element linearity
and that L(q) E C,(W,)
to show that
L is bounded.
of B(C,+z(W,);
if goE C,_,x(ZJ
are immediate,
Let r+~be an element Mathenmtical
C,(&?a)).
it suffices
of C,+,(9$\,).
Bioscietices
Since Then
2, 263 - 292 (196X)
A. R. ELCRAT
282
I-WI, g IPtln+ l4l%l~ + I%l,lP;,+ ~$P,zI,+ byvIa+ Id) < max (1, v, [u(,,[~~,(a)(p[2~_n, so that
L is indeed
bounded,
and
IWl d max (1, v, I& Now, noting
that
h(v) is an element
Iu,J,).
of C,(W,),
we have
1~1~ or /qlija is zero, h(q) is zero, so, after
If either
PROPOSITION
4.
The
noting
that
ma$@g 21 + P’(u)
o/ C, tu(Si?P) into R(C,
La(J?2) ; C,(&))
sa tis~f ies
the
uniform
Liflschitz
tion
Proof:
IYe set M = P’(d)
so that
M is the operator M(p) = (Zllpl), -
in R(CzI (&p),
-
,(a,);
= (26; -
Thus,
Mathematical
Riosciences
2,
263-
292
P’(d),
(1968)
C,(%,)) u;,p
given
+ (u1 -
bl
212)p*
condi-
EXISTENCE
283
THEOREMS
PROPOSITION
The ma$#Gg
6.
of C, +a(92) into B(C,+,(W,); F-differential
C,(W,))
at u corresponds
F(h, k) = (hk) ,, Referring
Proof:
is F-differentiable
at each u, and its
to the bilinear operator
to Proposition
h, k E C,(%). 3, we have
J”(tt + h)(f) = P’(u)(k) + (hk),, so that P’(u + h) = P’(u) + T(h), where
T maps
Jz into
the linear
operator
T,
defined
by
k E C, + .(9q.
T,(k) = (hk),, Since
The
mapping
T is linear,
and
liTI = so
that
that
T is an element
sup
of B(C,+,(XJ;
T is the F-differential
llTbll G I B(C,,
m(W2); C,(S?,))).
It follows
of the mapping u -+ P’(U)
at each
zc.
Using
the identification P”(h, k) = T,(k),
which
was mentioned
in Section
3. the result Mathematical
follows.
Biosciences
2, 263-292
(1968)
lie~~zark 1. is twice
It now follows
continuously
continuous),
immediately
differentiable
(a constant
Kemnark 2.
mapping
P
is certainly
< 1.
If X and Y are closed linear manifolds
respectively,
such that
hold with
Y
into Y for each ZL,all the above results
C, Lr*(L%Z)and C,(.@,)
Actually,
in C, i_,(@,,) and
P maps X into Y, P’(U) maps X into
for each 21, and P”(U) maps XxX
section
5 that
and that
p”(z,)II
C,(%,),
from Proposition
replaced
by X
and Y, respectivel!-.
this is the form in which we shall use the material
of this
in \chat follows.
fi. .4 SOI.LU’IOS
U’e shall Section
LISINC; THE KAXTOROVICH
now give
a proof
THEOREM
of the existence
theorem
promised
in
4.
Let X be the subset of C,+,(d,) identically consisting
consisting
zero on the normal boundary of those functions
identically
of those functions
zero on the surface
the sum of zeros is zero and the limit of a sequence
aA’,(t,).
and Y are closed linear manifolds
in C,,,(W,)
5 maps X into Y because
to be zero on aA’,(t,),
and, moreover,
and all of its derivatives 8R,(t,).
Similar
and that
reasoning
P”(U)
The problem a solution
that
maps
into
respectively.
f
is defined
in X is such that
appear in the definition
shows that
XxX
and C,(W,),
any function
Since
of zeros is zer-o, X
The operator
P defined in Section
which are
of B2 and Y the subset of C,(&)
it
of P are zero on
P’(U) maps X into
Y for each 21,
Y for each 1~.
which we shall now solve is that of proving
existence
of
ZI in X of the equation P(M) = 0,
where P is thought
of as an operator mapping X into Y. Since the manifold
yJJILis a subset of the normal boundary of Eqs.
(l-4)) and its restriction
\Ve ha1.c seen that that
of Si?,, such a function
to .oRis therefore
P is twice continuously
u is a solution
a solution F-differentiable
of Eqs.
(13).
on X and
EXISTENCE
285
THEOREMS
for u in X in Section
5. Theorems
1 and 2 enable us to prove the following
lemma. 1.
L~~li.4
Tlze
bounded
linear
operator
L, = P’(0)
E R(X;
Y)
is
aj+earing
in the a jwiori inequalit_y given by Theorenz 1.
inxtertible, and
7elhere B is the co&ant Proo,/:
1Ve have
seen that
L,(q) = cp*- vhp. If wt. show that
has a unique
the equation
solution
in X for each h in Y, or, equivalently,
initial-boundary-value
YAP; = h
yr -
theorem
solution
will imply
boundary
of BZ
f or each h in Y, the bounded
in C, +,(W,)
X).
Since W, has Property
E and an It in Y is zero on aR,(t,)
Theorem
the existence
Theorem
2 guarantees principle
1, with
inverse
that LO’ E R(Y;
The maximum
the
in ‘g2, on the normal
q=o has a unique
that
problem
of a solution
implies that this solution
by definition,
of the above equation. is unique.
In addition,
4 = 0, yields
so that
For clarity immediate
of exposition
corollary
we state
of Theorem
the following
lemma
which
is an
4. Mnthematicnl
Rioscienc~s
2.
263-292
(1968)
286
A. R.
LEMMA 2. mapping
Suppose
that P is a twice continuously
ETLKAT
IT-differentiable
of a Banach space X7 into a Balzach space Y such that P’(0)
invertible,
is
and that
(1) II[P’(W1/I < Br (2)
lip”(u) 11< 1
for each u i% X.
Then, if
where ljl <
the Newton
1/(2fq,
sequence 21, +, =
u, -
[p'(%)l-lp(a%J, uug= 0,
is well defined and converges to a solution u* of the equation P(u) = 0, where
u*
is an element of the closed sphere S(u, Y,, =
(1 -
‘Jr-
Y,,),
B2$/B.
The erYoY bougtd j (u* holds for the sequence {un}.
u,,/( < (2B2~)2”/2”B
Further, the solution u* is unique in the sfihere
S(u, YJ where y1 = (1 + vl The
requirements
of Lemma
B2r)/B.
2 are satisfied
P(0) = f satisfies ;f InG 171 where
Mathematical
Biosciences
2, 263 - 292
(1968)
by our operator
P if
EXISTENCE
287
THEOREMS
and B is the constant this condition
referred
to in Lemma
1.
We are forced to fulfill
by fiat. The
i%SUMPTION.
pressure-gradient
function
p, satisfies
IPxlx< WB2) in the region B (and, hence, its extension
to W, satisfies the same inequality
there). Summarizing
we have
If the $ressure-gradient
THEOREM.
the problem
our results,
(13) has a solution
Eqs.
proven
the following
function satisfies the above condition
u. Further,
u is aw element
of C, +,(.9Q,
and 2 +a < l/B. IUUI 1.
Remark
Our application
not only yields existence solution can be generated. {u,},
CT1+ (WP) -
y&
zero in .cZ~, and u,,+r is the solution
on the normal
The solution that
is the unique F1 -
is, generating solution
is contained the sequence
(Theorem
Thus,
where 2,; = uO, and I&, , problem
in W,,
be included
by applying
in the results
Altman’s
each
only by means 2).
as such for
using the modified Newton
{u:}
on the normal
well we will “obtain” solutions
and uniqueness
of the initial-boundary-value
Ydp, = P(&
will, in fact,
are obtained
of 9,.
in the proof of the Kantorovich
could also be generated
$V=o This
existence
boundary
and the proof of the fact that it has a unique solution
at each stage in the iteration method,
of the
in B?s,
= JY%)
we have not investigated
this linear problem, theorem.
the above theorem
problem
$9=0 However,
2 to obtain
but points out a way in which this
In fact, the solution u is the limit of a sequence
where u,, is identically
initial-boundary-value
of Lemma
of a solution
element
in the
of 9s
of the next
theorem.
of an existence
we have
boundary
However, sequence
theorem
section
which
in this case as of aphroximate
for a linear
problem
the error bound Ttiatlzematical
Biosciences
2, 263-292
(1968)
288
guaranteed modified
by Lemma 2, and we could give other, less desirable ones for the Newton
do not really leaves
method
have
open the question
of the above
(one is given in the next
the elements
linear
of how one might
problems
with such approximations. IZew~~r/z2. solution space X.
and combine
This matter
1Ve have included
of the operator Actually,
equation
Ilt + 2111 .\-
is “globally
VA I6 + f = 0
this
the results
in Lemma
the solution
on the normal
problem
Eqs.
(15).
with arbitrary
ever, this approach
does not appear
problem
in 9
A
Thus,
in,
SOLUTIOK
We shall
USING
again
to the problem A function
problem
being
for the “complete”
existence
technique
of a solution
of
values of class C, 1x. How-
to be applicable estimate
to such an initial-
edges at the ends of
such as that in Theorem
the sense of uniqueness
the problem
theorem
of Eqs.
for the problem
will be somewhat
has helped in
different
(14) in order
of Eqs. here,
on the normal
class C, + 1 there will be called an mdisturbed if
(15)
of S2,
THEOREM
# which is defined
to be ,mdisturbed,
of the
solution.
ALTMAN’S
attack
the desired existence
this
theorem
owing to the sharp
weakening
for a smooth
boundary
In fact, the general
prescribed
the tube sections which prevent an a$riori
5.
obtained
to be unique.
the similar problem
the quest
the
for that one, and a solution
which we have used could be used for proving
1 from holding.
that
it is the problem of Eqs. (13)
we are interested
\C’e have proven an existence
initial-boundary-value
boundary-value
above
in d2,
[4, Chapter 2 1. However,
one which
(13) is not expected
Kenzurk 3.
obtained
2 the statement
merely a device with which to prove existence of Eqs.
This
the solutions
problem
unique”
than
approximate
is unique in some sphere in the Banach
U= 0
rather
but we
sequences.
is not dealt with in this paper.
it could be shown that
initial-boundary-value
paragraph),
of the approximating
(13).
to obtain
The approach
however. boundary
boundary
of W, and of
functiolz,
or said
EXISTENCE
THEOREMS
2x9
(1)
$J = 0 on ,Y’,_, and
(2)
+, = #, = #,, = $J,,,, = lClfZ = 0 on aW,).
We define
X’ = (111zl
E
ZI has undisturbed
C,+&%,),
and let Y be the Banach
space so designated
1’ are closed linear manifolds the operator and P”(u)
map X’ and X’xX’, the existence
in Section
in C,_+ .(.%?,) and C,@,),
P defined in Section
shall prove
boundary
respectively,
6.
Then X’ and
respectively,
5 maps X’ into Y.
of a solution
values}
and
In addition
P’(U)
into Y for each 21 in X’.
zc in X’
\Vr
of the equation
P(u) = 0, where P is thought
of as an operator
this will yield the desired will be accomplished the results X’
existence
by applying
of Section
5 are that
mapping
X’ into Y, and, as before,
theorem
for Eqs.
Theorem
6.
In fact,
P is continuously
(13).
The proof
included
and that
its F-differentials
l,E.\IMA.
The lineur operator L, = P’(0) E B(X ; Y) ma/w X’
If X;, = L;l(O)
SF1 is invertible,
where B is the constant a$pearin, Proof:
Since a function boundary
u in
Lemma
in CztU (9,)
where X is the space introduced Lemma
maps X’ onto 1 of Section
1 of the previous
which is identically initial-boundary
of g2
section
zero on the values,
X’
in the previous
paragraph.
Y since it maps the subset
Thus,
X of X’ onto
L,
1’ 1)~
6.
\I’e shall denote by @ the class of all functions boundary
onto Y.
and
of A?, has undisturbed
xc certainly
on
satisfy
and 5ifl is the one-to-o+ae operator nzajq%zg X1/X:, onto
1’ that L, induces,
normal
among
F-differentiable
which
defined on the normal
al-e undisturbed. Mnthematical
Biosciences
2, 263-292
(19&S)
Suppose k is an element initial-boundary-value -k(P)
of Y, and let ~,4be an element
solution
on the normal
of a solution
of d,
of X’.
The existence
4 in X of the initial-boundary-
problem, in &,
h
L,@) = f/6=0 where
k = F, -
on the normal
L,(Y),
Y
being
with #on the normal boundary The
boundary
cp, and this p is an element
follows from the existence value
Then the
in 9&.,
= h
cp=# has a unique
of 4!.
problem
existence
implies that maximum
of Q follows L,(Y)
an element
of 9s.
Hence,
of d,,
of C,, %(&?a) which
agrees
In fact, we need only set v = Q!?+ If/.
from Theorem
is zero on 3R,(t,).
principle.
only one element
boundary
2 since
Uniqueness
to each $ in %! there
$ being
undisturbed
of cp follows from the corresponds
one, and
of the set L, ‘(h) c X’.
Since
we have
where
B is the constant
Theorem
1.
Therefore,
appearing
jj9”;‘jj and the lemma
Mathematical
inequality
given
by
< B,
is proven.
Now, all the requirements will be fulfilled
in the a priori
we have
if we assume
Biosciences
2, 263-
of Theorem that
292
P(0)
(1968)
6, with Y chosen to be 1/(2B), =
f
satisfies
EXISTESCE
291
THEOKEMS
where 7 < 1/(4P). We make
this assumption.
sequence
{zt*},
ah in X’
Then,
as defined
an underdetermined
prior to Theorem
modified
6, converges
Newton
to a solution
of the equation P(U) = 0.
In addition
the error
bounds
hold where E is the positive We could
Remark.
of the previous
number
equally
paragraph.
used in defining
well apply
Then
Theorem
the operator
one to one, and the underdetermined
the sequence
P’(0)
6 to the in that
{.u,$}.
problem
theorem
is
Newton process reduces to the mod-
ified Newton method. The number E used in defining the sequence of approximate
solutions
can then
be taken
to be zero, and the error
holds (if we choose Y to be 1/(2B)).
This error bound is markedly
than the one provided by Kantorovich’s method
as applied to our problem.
of Altman’s
theorem
SUMMARY
OF
partial
in flow of a viscous,
This solution
that the velocity
bound
satisfies
differential
on the C,-norm
of
stating
VLlZd+ p, = 0 incompressible
the physically
of a solution
in a larger,
set in such a way that
that there exists
equation
fluid in a compliant
dictated
be zero at the wall of the tube.
existence
for this equation being
as we have seen, application
a smaller
result of this paper is a theorem
of the nonlinear
for the velocity
by proving
better Newton
function.
d(Zl, + UU,) -
tube.
for the modified
RESULTS
The central a solution
theorem
However,
here requires
the pressure-gradient
bound
boundary
of an initial-boundary-value
“nonphysical” a solution
region,
this second
of it restricted
Mathematical
condition
The result is obtained
Biosciences
problem problem
to the original 2, 263-
292 (196s)
region
is a solution
problem equation showing
for an operator that,
problem.
The assumptions
Existence
of finding
one Ranach
assumptions,
of Kantorovich’s
space
pressure-gradient
function
a certain
among these assumptions.
proportional
an a jwiori
inequality
and
r-t, Chapter
another
involved
or a generalization
and
satisfies
of it due to
be uniformly
bound.
of the bound on the C, norm of the pr(xure-gradient is inversely
into
the operator
theorem
with its C, norm satisfying
the most restrictive
for the second
a zero of an operator
involved are that a certain manifold be “smooth”
and that the (prescribed) continuous
mapping
under certain
the hypotheses Altman.
of the original
is shown by posing a problem
to the square derived
function
is by far
In fact, the required
of a constant
by Friedman
Hijlder
The assumption
which
(see Theorem
4_!), and we gi1.e no upper bound
bound
appears
in
1, Section
3,
for the value of this
constant. It should be pointed problem adhere obtained
only
the
out that we ha\.e imposed
above-mentioned
to the wall of the tube. to be unique,
of in light uniqueness” HEFEl
of those has been
boundary
Thus,
and tile results
in
[l]
and
demonstrated
[a]
on the fluid dJ.namic
condition
that
the
fluid
there is no hope for the solution of this paper where
“mean
for this problem.
are to be thought squarcl
asymptotic