Beruflich Dokumente
Kultur Dokumente
A Dissertation
Presented to
the Graduate School of
Clemson University
In Partial Fulfillment
of the Requirements for the Degree
Doctor of Philosophy
Mathematical Sciences
by
John C. Chrispell
August 2008
Accepted by:
Dr. Vincent J. Ervin, Committee Chair
Dr. Eleanor W. Jenkins, Co-Chair
Dr. Christopher E. Kees
Dr. Hyesuk K. Lee
Dr. Daniel D. Warner
ABSTRACT
The accurate numerical approximation of viscoelastic fluid flow poses two difficul-
ties: the large number of unknowns in the approximating algebraic system (corresponding
to velocity, pressure, and stress), and the different mathematical types of the modeling
splitting technique that may be used to decouple mathematical equations of different types
as well as separate the updates of distinct modeling equation variables when modeling mixed
In this work a fractional step -method is described and analyzed for the numerical
computation of both the time dependent convection-diffusion equation and the time depen-
dent equations of viscoelastic fluid flow using the Johnson-Segalman constitutive model. For
convection-diffusion the -method presented allows for a decoupling within time steps of the
parabolic diffusion operator from the hyperbolic convection operator. The hyperbolic con-
The analysis given for the convection-diffusion equation serves as a template for the analysis
The -method implementation analyzed for the viscoelastic modeling equations al-
lows the velocity and pressure approximations within time steps to be decoupled from the
stress, reducing the number of unknowns resolved at each step of the method. Additionally
the -method decoupling results in the approximation of the nonlinear viscoelastic modeling
system using only the solution of linear systems of equations. Similar to the scheme im-
a SUPG-method. For both the convection-diffusion and the viscoelastic modeling equa-
tions a priori error estimates are established for their -method approximations. Numerical
ii
computations supporting the theoretical results and demonstrating the -methods are also
included.
iii
To My Parents
ACKNOWLEDGMENTS
This work would not have been possible without the support of many others. I
would like to start by acknowledging my advisors: Dr. Vince Ervin, and Dr. Lea Jenkins.
I do not have the words to express the debt of gratitude I feel for the knowledge they have
shared, guidance they have given, and patience and kindness they have shown me. They
both possess a love of mathematics that comes through in and outside the class room. It
has been a privilege to be their student, and I hope to work with both Dr. Jenkins and Dr.
I would like to thank Dr. Chris Kees for taking me under his wing during my
summer internships with the Army Corps of Engineers in Vicksburg, as well as for serving
on my committee. It was an honor to work under his guidance, and I look forward to future
I would also like to thank the other members of my committee Dr. Dan Warner
and Dr. Hyesuk Lee. Their experience, suggestions, and expertise have been a tremendous
benefit.
Throughout the completion of this work I was supported by grants from the National
Science Foundation and the U.S. Army Research Office. These grants allowed me to focus
my efforts over the last four years on research. I will be forever grateful for having had that
opportunity.
I would like to thank my parents, my brother Jared (you will always be my Cap-
tain), and Grandma for all their support and love. Their encouragement and reassurance
throughout my life has always been a blessing. They believed in me before I believed in
Over the last few years I had the privilege to get to know Sandy Berry and her
family during my summer stays in Vicksburg Mississippi. She and her family made a boy
v
During my time at Clemson I was lucky enough to share an office with Jason Howell.
Jason is a tremendous colleague and friend. Knowing him has enhanced my life both
My friends have been a immense benefit while completing this work. I couldnt have
done it without their support. Specifically, I would like to thank Denise Guinnane for her
vi
TABLE OF CONTENTS
Page
TITLE PAGE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i
ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
DEDICATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
ACKNOWLEDGMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x
CHAPTER
1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 The Fractional Step -method . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 The Johnson-Segalman Model for Viscoelastic Fluid Flow . . . . . . . . 5
1.4 Mathematical Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2. CONVECTION-DIFFUSION . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Convection-Diffusion Model and Notation . . . . . . . . . . . . . . . . . 11
2.3 -method for Convection-Diffusion . . . . . . . . . . . . . . . . . . . . . 12
2.4 Variational Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.5 Unique Solvability of the Scheme . . . . . . . . . . . . . . . . . . . . . 14
2.6 A Priori Error Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.7 Optimal Value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.8 Numerical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 -method for Viscoelastic Fluid Flow . . . . . . . . . . . . . . . . . . . 27
3.3 Mathematical Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.4 Unique Solvability of the Scheme . . . . . . . . . . . . . . . . . . . . . 32
3.5 A Priori Error Estimates for the Conservation Equations and Constitu-
tive Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.5.1 Analysis of the conservation equations . . . . . . . . . . . . . . 38
3.5.2 Analysis of the constitutive equation . . . . . . . . . . . . . . . 59
3.6 Establishing the Full -method Error Estimate . . . . . . . . . . . . . . 115
vii
Table of Contents (Continued)
Page
3.6.1 The first time step (0, t] . . . . . . . . . . . . . . . . . . . . . 118
3.6.2 The general time step (nt, (n + 1)t] . . . . . . . . . . . . . . 153
3.7 Induction Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
APPENDICES
BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
viii
LIST OF TABLES
Table Page
4.1 Approximation errors and convergence rates for |ku uh k| and |kp ph k| at
T = 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
4.3 Approximation errors and convergence rates for |ku uh k|,0 , |k h k|,0 ,
and |kp ph k|,0 at T = 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
4.4 Approximation errors and convergence rates for |ku uh k|0,1 , |ku uh k|0,0 ,
|k h k|0,0 , and |kp ph k|0,0 at T = 2. . . . . . . . . . . . . . . . . . . . 171
ix
LIST OF FIGURES
Figure Page
2.3 Plots of the true and approximate solution for convection-diffusion Example 2. 24
2.4 Plots of the true and approximate solution for convection-diffusion Example 3. 26
4.4 Streamlines and magnitude of velocity contours for u at t = 0.125 and t = 0.375. 173
4.5 Streamlines and magnitude of velocity contours for u at t = 0.5 and t = 0.75. 174
x
CHAPTER 1
INTRODUCTION
1.1 Motivation
Viscoelastic fluids are encountered everyday. Examples of these fluids include: food
products like egg whites, mayonnaise, and molten chocolate, biological fluids such as mu-
cus and blood, as well as industrial materials like paint, adhesives, multi-grade oils, and
polymers [48]. From this list of examples it can be seen that viscoelastic fluids share the
properties of both viscous fluids and elastic solids. In order to describe the motion of a
viscoelastic fluid the constitutive law modeling the relationship between the fluids stress
and deformation must incorporate a history of the fluids deformation. The complicated na-
ture of the equations governing viscoelastic flow make finding analytical solutions difficult,
if not impossible. Hence, the study of good numerical methods to obtain approximations
to the modeling equations of viscoelastic flow is a topic of interest for both academic and
industrial researchers, as it can give useful insights into applications such as biological fluid
modeling, as well as fiber and film processing. To this end, in this dissertation a fractional
Computational methods for modeling viscoelastic fluid flow are difficult for a variety
of reasons. The modeling equations (assuming slow flow) represent a Stokes system for
the conservation of mass and momentum equations, coupled with a non-linear hyperbolic
equation describing the constitutive relationship between the fluids velocity and the (ex-
tra) stress. The numerical approximation requires the determination of the fluids velocity,
pressure and stress (a symmetric tensor). For an accurate approximation a direct approxi-
mation technique requires the solution of a very large nonlinear system of equations at each
time step.
The fractional step -method [55, 56, 58] is an appealing numerical approximation
technique for several reasons. The -method decouples the approximation of velocity and
1
pressure from the approximation of stress, thereby reducing the size of the algebraic systems
which have to be solved at each sub-step. This decoupling also allows for appropriate
approximation techniques to be applied when resolving the resulting parabolic model for
velocity and pressure, and hyperbolic constitutive relation for stress. An additional benefit
of the -method [58] is that by decoupling velocity and pressure from the stress, the algebraic
Volterra, in the late eighteen hundreds, but it was the work of Oldroyd in 1950 that produced
a constitutive model that worked well when modeling fluids with large deformations [46, 62].
Since Oldroyds original work, many constitutive equations have been formulated to describe
the motion of viscoelastic fluids. These include the models of Giesekus [25], Oldroyd [47],
and Phan-Thien and Tanner [50], as well as the Johnson and Segalman [36] constitutive
Error analysis of finite element approximations to steady state viscoelastic flow was
first done by Baranger and Sandri in [4] using a discontinuous Galerkin (DG) formulation
of the constitutive equation. In [54] Sandri presented analysis of the steady state problem
dependent problem was first analyzed by Baranger and Wardi in [5], using an implicit Euler
Ervin and Miles analyzed the problem using an implicit Euler time discretization and a
SUPG discretization for the stress in [21]. Analysis of a modified Euler-SUPG approxima-
tion to the transient viscoelastic flow problem was presented by Bensaada and Esselaoui
in [6]. The temporal accuracy of the approximation schemes studied in [5, 6, 21] are all
O(t). The work of Machmoum and Esselaoui in [40] examined time dependent viscoelas-
tic flow using a characteristics method that has accuracy O((h2 / t) + t). Ervin and
Heuer proposed a Crank-Nicolson time discretization method [19] which they showed was
O t2 . Their method uses a three level scheme to approximate the nonlinear terms in
the equations. Consequently their approximation algorithm only requires linear systems of
2
equations be solved. In [9] Bonito, Clement, and Picasso use an implicit function theorem
to analyze a simplified time dependent viscoelastic flow model where the convective terms
were neglected.
The fractional step -method was introduced, and its temporal approximation accu-
racy studied, for a symmetric, positive definite spatial operator, by Glowinski and Periaux
in [27]. The method is widely used for the accurate approximation of the Navier-Stokes
equations (NSE) [33, 59, 60]. In [39], Kloucek and Rys showed, assuming a unique solution
existed, that the -method approximation converged to the solution of the NSE as the spa-
tial and mesh parameters went to zero (h, t 0+ ). The temporal discretization error for
the -method for the NSE was studied by Muller-Urbaniak in [44] and shown to be second
order.
The implementation of the fractional step -method in [58] for viscoelasticity dif-
fers significantly from that for the NSE. For the NSE at each sub-step the discretization
contains the stabilizing operator u. For the viscoelasticity problem the middle substep
when resolving the stress is a pure convection (transport) problem that requires stabiliza-
tion in order to control the creation of spurious oscillations in the numerical approximation.
Marchal and Crochet [42] were the first to use streamline upwinding to stabilize the hyper-
Prior to presenting analysis of the fractional step -method for viscoelastic fluid flow
system and therefore share some of the characteristics observed in the viscoelastic model.
The analysis conducted for the convection-diffusion problem serves as a template for the
work done regarding viscoelasticity. Chapter 5 contains a brief summary and a description
of future work. The remainder of the current Chapter gives an introduction to the -method,
3
establishes the mathematical model considered for viscoelastic fluid flow, and presents rel-
mathematical setting.
In order to introduce the fractional step -method consider the following time de-
u
+ F (u, x, t) = 0 in (0, T ]
t
subject to u(x, t) = 0 x (0, T )
u(x, 0) = u0 (x) x .
Discretizing the temporal derivative and choosing a value of (0, 1/2), the -method
advances across a generic nth time step of size t according to the following three step
procedure:
(n+)
Step 1. Compute an approximation to uh : 6
n+1
t
(n+) (n)
uh uh
+ 1 F (n+) = 2 F (n) . t
t
(n+1)
uh uh
(n+) (1 2)t
+ 2 F (n+1) = 1 F (n+) .
(1 2)t tn+ x
(n+1)
Step 3. Compute an approximation to uh : t
uh
(n+1) (n+1)
uh 1 (n+1) 2 (n+)
tn
+ F = F .
t
4
It will be seen in the analysis and corresponding computations for both convection-
diffusion and viscoelastic fluid flow, that for the optimal choice of = 1 2/2 the fractional
u = 0 on , (1.4)
Here (1.1) is the constitutive equation relating the fluids velocity u to the stress , and (1.2)
and (1.3) are the conservation of momentum and conservation of mass equations. The fluid
the product of a characteristic strain rate and the relaxation time of the fluid [8]. Note that
if the value of is set to zero in (1.1) the well known modeling equations for a Newtonian
fluid, introduced by Navier (1827) and Stokes (1845), are obtained [45, 53]. Re denotes the
The body forces acting on the fluid are given by f , and (0, 1) denotes the fraction of
5
The ga term and deformation tensor d(u) are defined as:
1a 1+a
ga (, u) := u + (u)T (u) + (u)T
2 2
and
1
d(u) = u + (u)T .
2
The gradient of u is defined such that (u)i,j = ui /xj . For the remainder of this
model is obtained. Proofs of the existence and uniqueness of solutions to (1.1)-(1.6) can be
for 1 r <
Z
|f |r dA < .
For the case of L () the space of all essentially bounded functions on the norm
is used, where the essential supremum of |f | on is the greatest constant lower bound K
for which |f (x)| K almost everywhere on [1, 18]. The L2 () inner product and norm
6
and for tensors and
Z Z 1
2
(, ) := : dA, and kk := : dA .
The products are defined as follows for vectors u and v, and tensors and
d
X d X
X d
uv = ui vi , and : = ij ij .
i=1 i=1 j=1
L20 () = f L2 () : f = 0 on .
The standard Sobolev space [1] of order k is denoted Wpk (), and its norm is given
represent the Sobolev space W2k , and kkk denotes the norm in H k .
To describe the finite element framework used in the analysis, let Th be a triangu-
lation of the discretized domain Rd . Then
= K, K Th .
c1 h hK c2 K ,
where hK is the diameter of triangle K, K is the diameter of the greatest ball (sphere)
Several different continuous and discrete norms are used in the analysis. When
7
For N ZZ+ let T /N = t, and define the discrete norms
N
! 12
X
2
|||v|||,k := max
v (n)
, |||v|||0,k := t
v (n)
.
1nN k k
n=1
v (n) v (n)
d v (n) := ,
t
v (n) v (n1)
dt v (n) := .
t
element space norms. Thus, for P , a vector space of functions defined on a bounded domain
K in Rd ,
Lemma 1.4.1 (Inverse Estimate [11]) Let {Th }, 0 < h 1, denote a quasi-uniform
family of subdivisions of a polyhedral domain Rd . Let (K, P, N ) be a reference fi-
nite element such that P Wpl (K) Wqm (K), where 1 p , 1 q , and
for all v Vh .
The discrete Gronwalls inequality [31] is used when establishing a priori error estimates for
the fractional step -method applied to both convection-diffusion and the viscoelastic fluid
modeling equations.
8
Lemma 1.4.2 (Discrete Gronwalls Inequality) Let t, H, and an , bn , cn , n (for in-
9
CHAPTER 2
CONVECTION-DIFFUSION
2.1 Introduction
The fractional step -method for the linear convection-diffusion equation is analyzed
can be divided into two approaches: (i) additive decomposition methods and (ii) product
a sum of several operators. At each sub-step in the approximation algorithm the spatial
operator is replaced by its additive decomposition, with some of the operators evaluated at
the current time (i.e. treated implicitly) and the others at past times (i.e. treated explicitly).
Examples of this approach are the Alternating Direction Implicit (ADI) methods [17, 20,
43, 49] and the IMplicit EXplicit (IMEX) schemes [2, 32]. With product decomposition
methods, to advance the approximation from time tn1 to tn , a pure convective operator is
applied to obtain an initial estimate at tn . This estimate is then taken as the initial data at
tn1 and a pure diffusion operator is used to determine the approximation at tn . Examples
of this approach include the work of Dawson and Wheeler [15, 16], Khan and Liu [38], and
Evje and Karlsen [23]. A survey of these methods can be found in [22].
decomposition method, with the features of a product decomposition method. In the first
and third sub-steps of the three sub-step algorithm a pure diffusion problem is approxi-
the convection-diffusion problem will provide insight for the viscoelastic fluid flow problem
because the approximation scheme studied for convection-diffusion is similar to that in [57]
for viscoelastic fluid flow. The middle sub-step in both applications is a pure convection
(transport) problem, and the first and third sub-steps are parabolic problems. In this work
10
the -method is applied to the linear convection-diffusion equations is thoroughly outlined.
The remainder of this chapter is organized as follows. In the next section the
convection-diffusion problem is specified, and the mathematical notation used that is specific
to its analysis is given. In Section 2.3 the fractional step -method for the convection-
diffusion equation is described, computability of the algorithm is shown, and a priori error
estimates for the method are established. A discussion on the optimal choice of the
parameter is given in Section 2.7. Numerical examples demonstrating the method are
u
4u + b u + c u = f in (0, T ] (2.1)
t
u(x, t) = 0, x (0, T ] (2.2)
where b = [b1 (x, t), b2 (x, t)]T is an incompressible velocity field (i.e b = 0), c(x, t) cmin
is used in the variational formulations in the analysis. The corresponding finite element
Xh := v X C : v K Pk (K) K Th .
Let U be the L2 projection of u onto Xh , and use u(n) := u (, nt). Used in the error
(n)
(n) := u(n) U (n) , E (n) := U (n) uh .
11
Note that is the error between the true solution and the best solution in the discrete space,
and E is the difference between the achieved approximation and the best approximation
available in the discrete space. Having defined the following interpolation properties of
N
X N
X
n 2
t k k tCI2 h2k+2 ku(, tn )k2k+1 Ch2k+2 |kuk|20,k+1 (2.6)
n=0 n=0
(n) (n1)
2
N
X 2
t
Ch2k+2 kut k0,k+1 (2.7)
t
n=1
where k denotes the order of the elements used in the interpolation space.
Using the outline of the -method given in Section 1.2, an abstract representation
of (2.1)-(2.3) is:
u
+ F (u, x, t) = 0 in (0, T ] (2.8)
t
subject to u(x, t) = 0, x (0, T ) (2.9)
where
F = u + b u + cu f.
where
1 c
F (u, x, t) = u + u f (2.12)
2
2 c
F (u, x, t) = b u + u. (2.13)
2
12
(n) (n+1)
For 0 12 , the fractional step method for advancing the approximation of uh to uh
(n+) (n)
uh uh
+ 1 F (n+) = 2 F (n)
t
(n+) (n+) c (n+) (n) c (n)
d uh uh + uh = f (n+) b uh uh . (2.14)
2 2
(n+1)
Step 2. Compute the approximation uh :
(n+1) (n+)
uh uh
+2 F (n+1) = 1 F (n+)
(1 2)t
(n+1) (n+1) c (n+1) (n+) c (n+)
d(12) uh + b uh + uh = f (n+) + uh uh . (2.15)
2 2
(n+1)
Step 3. Compute the approximation to uh :
(n+1) (n+1)
uh uh
+1 F (n+1) = 2 F (n+1)
t
(n+1) (n+1) c (n+1) (n+1) c (n+1)
d uh uh + uh = f (n+1) b uh uh . (2.16)
2 2
Analysis requires variational formulations of each substep.
(n+) c (n+) (n+)
d uh + uh , v + uh , v
2
(n) c (n)
= f (n+) b uh uh , v , v Xh , (2.17)
2
(n+1)
(n+1) c
d(12) uh , v + b uh + u(n+1) , vb
2
c (n+)
(n+)
(n+)
= f (n+) uh , vb uh , v + 4uh , b v , v Xh , (2.18)
2
13
(n+1) c (n+1) (n+1)
d uh + uh , v + uh , v
2
(n+1) c (n+1)
= f (n+1) b uh uh , v , v Xh . (2.19)
2
The first step in the analysis is to show that the scheme (2.17)(2.19) is computable.
That is, the associated coefficient matrices on the left hand sides of (2.17) - (2.19) are
invertible.
(n+)
Lemma 2.5.1 There exists a unique solution uh Xh satisfying (2.17).
where
1 c
A(w, z) = (w, z) + (w, z) + ( w, z) .
t 2
Note that (2.21) represents a square linear system of equations Ac = f . The fact that
1 c
A(w, w) = (w, w) + (w, w) + ( w, w) > 0
t 2
guarantees that ker(A) = {0}. It follows that (2.17) has a unique solution.
The unique solvability of (2.19) follows exactly as for (2.17). For (2.18) the same approach,
14
together with the divergence free assumption for b (i.e. b = 0), establishes the unique
solvability. The unique solvability of the -method algorithm and the subsequent a priori
error estimates for the convection-diffusion equation may be shown without the divergence
free assumption for b provided the time step t is assumed sufficiently small.
error estimate. This estimates given in Theorem 2.6.1, and a discussion of the proof is
Theorem 2.6.1 For a sufficiently smooth solution u, with t Ch2 , the fractional step
where
G(t, h, ) = C(t)2 kuttt k0,0 + kutt k0,1 + kutt k0,0 + kftt k0,0
+ Ct kut k0,2 + kut k0,1 + kut k0,0 + kft k0,0
A key step in the analysis of the fractional step -method is the construction of
unit stride expressions for the error at successive time steps. These are expressions where a
2
2
bound is known on the difference of error terms at times t apart i.e.
E (n+1)
E (n)
.
2
Assuming the initial error
E (0)
= 0, the summation of a sequence of this form telescopes
2
to
E (l)
. Choosing an optimal value for , it will be seen that the first order terms in
15
the interpolation error are zero, yielding a second order temporal discretization. Further
Step 1. Form the following linear combinations of equations (2.17), (2.18), and (2.19) to
(n+1) (n) (n+1) (n) (n+) (n1+)
obtain equations involving uh uh , uh uh , and uh uh , respec-
tively.
t (2.17)
+ (1 2) t (2.18)
+ t (2.19) (2.24)
with nn with nn with nn
t (2.17) + (1 2) t (2.18) + t (2.19) (2.25)
with nn with nn with nn1
t (2.17) + (1 2) t (2.18) + t (2.19) (2.26)
with nn with nn1 with nn1
Step 2. Subtract equations (2.24), (2.25), and (2.26) from (2.20) evaluated at the midpoint
(n+1) (n) (n+1)
(u(n+1) uh ) (u(n) uh ) , vh + t Gpos ((u(n+1) uh ), vh )
(n+1) (n+) (n)
= t Grem (t, f, u, uh , uh , uh , vh ), (2.27)
(n+1) (n) (n+1)
(u(n+1) uh ) (u(n) uh ) , vh + t Hpos ((u(n+1) uh ), vh )
(n+) (n) (n)
= tHrem (t, f, u, uh , uh , uh , vh ), (2.28)
(n+) (n1+) (n+)
(u(n+) uh ) (u(n1+) uh ) , vh + tKpos ((u(n+) uh ), vh )
(n) (n) (n1+)
= t Krem (t, f, u, uh , uh , uh , vh ) , (2.29)
where Gpos , Hpos , and Kpos denote the positive part of the operators.
16
Step 3. Use u uh = + E, choose vh = E (n+1) , vh = E (n+1) , and vh = E (n+) in
(n+1)
2
(n)
2
E
E
+ tGpos (E (n+1) , E (n+1) )
tR1 t, f, u, (n) , (n+) , (n+1) , (n+1) , E (n) , E (n+) , E (n+1) . (2.30)
(n+1)
2
(n)
2
E
E
+ tHpos (E (n+1) , E (n+1) )
tR2 t, f, u, (n) , (n) , (n+) , (n+1) , E (n) , E (n) , E (n+) . (2.31)
(n+)
2
(n1+)
2
E
E
+ tKpos (E (n+) , E (n+) )
tR3 t, f, u, (n) , (n) , (n+1) , (n+) , E (n) , E (n) , E (n+1) . (2.32)
Use E (0) = 0 and sum (2.30), (2.31), and (2.32) from (n = 0 to (N 1)), (n = 1 to (N 1)),
(N )
2
(N )
2
(N 1+)
2
E
+
E
+
E
N
X 1 N
X 1
+ t Gpos (E (n+1) , E (n+1) ) + t Hpos (E (n+1) , E (n+1) )
n=0 n=1
N
X 1
+ t Kpos (E (n+) , E (n+) )
n=1
N
X 1
t R1 t, f, u, (n) , (n+) , (n+1) , (n+1) , E (n) , E (n+) , E (n+1)
n=0
N
X 1
+ t R2 t, f, u, (n) , (n) , (n+) , (n+1) , E (n) , E (n) , E (n+)
n=1
N
X 1
+ t R3 t, f, u, (n) , (n) , (n+1) , (n+) , E (n) , E (n) , E (n+1)
n=1
2
2
+
E (1)
+
E ()
. (2.33)
17
2
Step 5. Apply discrete Gronwalls inequality (see Lemma 1.4.2) with al =
E (N )
+
(N )
2
(N 1+)
2
E
+
E
. The lemma requires that
C2 2 C 2 2 C4
t C1 + 2 + 4 3 + 2 1, (2.34)
h h h
and h and the upwinding parameter . If is chosen such that ( Ch), (2.34) becomes
Ct h2 1. (2.35)
This is computationally very restrictive. This constraint is not enforced for the numerical
results in Section 2.8. It is an open question if this condition is necessary for (2.22) and
(2.23).
(N ) (N )
Step 6. Use the triangle inequality to obtain the error estimate for
u uh
+
(N ) (N )
(N 1+)
u uh
+
u(N 1+) uh
.
Numerical results are presented in Section 2.8 for a continuous, piecewise linear
approximation to u (i.e. k = 1). For this case, (2.22) and (2.23) give the following estimate.
Ch, and u sufficiently smooth, the approximation uh satisfies the error estimate:
In [27] Glowinski and Periaux studied the convergence and stability of the -method
for
du
+ Au = 0 ,
dt
18
where A was assumed to be a constant pp symmetric, positive definite matrix and u IRp .
The decomposition they considered (see 2.11) was, for (0, 1),
Au = Au + (1 )Au,
i.e. F1 (u, t) = Au and F2 (u, t) = (1 )Au. Using an eigenvalue analysis, the authors
were able to establish that for = 1 2/2 the fractional step -method was second order
An eigenvalue analysis approach is not possible for the approximation method de-
In Grem :
1
u(n+1) + (1 ) u(n+) u(n+ 2 ) , (2.38)
n+ 12
(1 ) u(n+) + u(n) u( ). (2.39)
In Hrem :
1
u(n) + (1 ) u(n+) u(n+ 2 ) , (2.40)
1
(1 2) u(n+) + u(n) + u(n) u(n+ 2 ) . (2.41)
In Krem :
1
u(n) + (1 ) u(n) u(n+ 2 ) , (2.42)
1
(1 2) u(n+1) + u(n) + u(n+) u(n+ 2 ) . (2.43)
Suitable estimates for these expressions are obtained using Taylor series expansions about
(n + 1/2)t, (n + 1/2 )t, and (n + 1/2)t for the Grem , Hrem , and Krem expressions,
respectively. The first order terms in these expansions, i.e. the coefficients of t, all reduce
to a constant multiple of
22 4 + 1. (2.44)
The roots of (2.44) are = 1 2/2. Thus, in order to have a second order temporal
discretization error the only possible choice for satisfying 0 < < 1/2 is = 1 2/2.
Verification that (2.38) yields a second order accurate temporal discretization is shown in
19
the proof of Lemma 2.7.1. Second order accuracy for the other five terms (2.39) - (2.43) is
shown in a manner similar to the proof of Lemma 2.7.1 and is given in [12].
Lemma 2.7.1 For the optimal value = 1 2/2
N 1
1
2
t
u(n+1) + (1 ) u(n+) u(n+ 2 )
C(t)4 kutt k20,0 .
X
n=0
1
Proof: Expanding u(n+1) and u(n+) about u(n+ 2 ) gives
1
Z t(n+1)
u (n+1)
= u(n+ 2 ) + ut (, t) dt
t
( n+ 1
2 )
Z t(n+1)
) + 1 t u(n+ 2 ) +
1
n+ 21
= u(
t utt (, t) t(n+1) t dt, (2.45)
2 t
( n+ 1
2 )
and
Z t
1 (n+ 12 )
u (n+)
= u(n+ 2 ) ut (, t) dt
t(n+)
Z t 1
1 1 (n+ 1 ) (n+ 2 )
= u(n+ 2 )
t ut 2 + utt (, t) t t(n+) dt. (2.46)
2 t(n+)
20
Z ( Z t(n+1)
= utt (, t) t(n+1) t dt
t
(n+ 12 )
)2
Z t
(n+ 12 )
+ (1 ) utt (, t) t t(n+) dt dA
t(n+)
2
Z ( Z t(n+1)
2
2 utt (, t) t(n+1) t dt
t
( n+ 1
2 )
!2 )
Z t
( n+ 1 )
+ (1 )2 2
utt (, t) t t(n+) dt dA
t(n+)
Z ( Z t(n+1) Z t(n+1) 2
2 2
(utt (, t))2 dt t(n+1) t dt
t t
( n+ 1
2 ) (n+ 1
2 )
!)
Z t Z t
2 (n+ 12 ) 2 (n+ 12 ) 2
+ (1 ) (utt (, t)) dt t t(n+) dt dA
t(n+) t(n+)
Z ( 3 Z t(n+1)
1 t
= 2 2 (utt (, t))2 dt
3 2 t
(n+ 12 )
3 Z t !)
2 1 1 (n+ 12 ) 2
+ (1 ) t (utt (, t)) dt dA
3 2 t(n+)
3 Z ( Z t(n+1)
2 t
= 2 (utt (, t))2 dt
3 2 t
(n+ 12 )
Z t 1 !)
(n+ )
+ (1 )2 (1 2)3 2
(utt (, t))2 dt dA
t(n+)
3 (Z )
oZ t(n+1)
2 t n
2
2
max , (1 ) , (1 2)3
(utt (, t))2 dt dA
3 2 t(n+)
3 Z t(n+1)
(t)
= (1 )2 kutt (, t)k2 dt.
12 t(n+)
Thus,
N 1
1
2
t
u(n+1) + (1 ) u(n+) u(n+ 2 )
X
n=0
N 1
(t)4 t(n+1)
X Z
(1 )2 kutt (, t)k2 dt
12 t(n+)
n0
C(t) 4
kutt k20,0 .
21
The optimal value was investigated numerically by calculating experimental con-
vergence rates for the convection-diffusion problem given in (2.1)(2.3) for b = [1, 1]T ,
c = 1.0, = (0, 1) (0, 1), Xh the space of continuous piecewise linear functions, and f
4.5
u(x, y, t) = 10xy(1 x)(1 y)ex (1 t4 ). (2.47)
The meshes used for these calculations were obtained by dividing the spatial (h)
and temporal (t) discretization parameters on each successive mesh by 2. As the spatial
discretization scheme is second order, it is expected that the experimental convergence rate
is determined by the temporal discretization. Figure 2.1 indicates that when = 1 2/2
the method has second order convergence with respect to both h and t.
Figure 2.2 displays the error |ku uh k|0,0 at T = 1 on a mesh with t = 1/128 and
h = 2/320 for different values of . The minimum error corresponds to = 1 2/2.
4
x 10
3
2.1 2.04
2.02
Convergence Rate at time T = 1
2
2
1.98
T =1
1.96
2.5
1.9 1.94
||| u uh |||0,0 at
1.5
50 100 150 200 250 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
1/h Value
22
2.8 Numerical Results
Three examples are used to verify the theoretical convergence rates and demonstrate
the effectiveness of the fractional step -method (2.17)(2.19) for convection diffusion prob-
orders of magnitude less than the magnitude of the velocity field. The solution in Ex-
ample 3 represents a steep moving front propagating through the domain. The optimal
value of = 1 2/2 was used in all computations. In each example a sequence of con-
tinuous, piecewise linear approximations uh were computed, by dividing the time step t
and the spatial mesh parameter h by two. The experimental convergence rates are com-
puted for various choices of . From Theorem 2.6.1, the predicted convergence rates are
Example 1. This test problem was described in Section 2.7 when the optimal value of
was examined, and is a modified version of a steady state problem given in [37].The true
solution is given in (2.47). The solution is a slightly skewed bubble function which decays
to zero as t 1. The numerical results for this example are presented in Table 2.1.
Example 2. In this example, taken from [61], the approximation of u(x, y, t) satisfies
u
k4u + b u = f in (0, T ], (2.48)
t
for k = 0.0001, b = [4y, 4x]T , and = (0.5, 0.5) (0.5, 0.5). The true solution is given
by
2 2 (x + 0.25)2 + y 2
u(x, y, t) = 2 exp , (2.49)
2 + 4kt 2 2 + 4kt
where x = x cos(4t) + y sin(4t), y = x sin(4t) + y cos(4t) and = 0.0477. The initial and
boundary conditions are given by u0 (x, y) = u(x, y, 0), and u(x, y, t)| = u(x, y, t) 0.
The solution represents a Gaussian pulse being convected by a rotating velocity field. Table
2.2 lists the errors in the numerical approximation and the experimental convergence rates.
The approximation is illustrated in Figures 2.3(a)2.3(b), for h = 2/64 and = h2 .
23
Table 2.1 Approximation errors and convergence rates for Example 1.
=1 2/2 Time T = 1.0
(t, h) 1
( 10 , 82 ) 1
( 20 , 162 ) 1
( 40 , 322 ) 1
( 80 , 642 ) 1
( 160 , 1282 )
0 |ku uh k|0,1 4.092e-1 2.184e-1 1.117e-1 5.628e-2 2.823e-2
Cvge. Rate - 0.9 1.0 1.0 1.0
|ku uh k|,0 2.039e-2 5.358e-3 1.359e-3 3.411e-4 8.537e-5
Cvge. Rate - 1.9 2.0 2.0 2.0
h
2
|ku uh k|0,1 5.407e-1 3.061e-1 1.576e-1 7.757e-2 3.772e-2
Cvge. Rate - 0.8 1.0 1.0 1.0
|ku uh k|,0 7.055e-2 3.275e-2 1.524e-2 7.178e-3 3.443e-3
Cvge. Rate - 1.1 1.1 1.1 1.1
3/2
h
2
|ku uh k|0,1 4.337e-1 2.259e-1 1.135e-1 5.668e-2 2.831e-2
Cvge. Rate - 0.9 1.0 1.0 1.0
|ku uh k|,0 3.782e-2 1.176e-2 3.632e-3 1.138e-3 3.649e-4
Cvge. Rate - 1.7 1.7 1.7 1.6
2
h
2
|ku uh k|0,1 4.124e-1 2.188e-1 1.117e-1 5.629e-2 2.823e-2
Cvge. Rate - 0.9 1.0 1.0 1.0
|ku uh k|,0 2.613e-2 6.793e-3 1.709e-3 4.256e-4 1.058e-4
Cvge. Rate - 1.9 2.0 2.0 2.0
0.9
1 0.8
0.8 0.7
0.6 0.6
0.4 0.5
0.2 0.4
0.3
1 1 0.2
0.9 0.1
0.8 0.8 0
0.7
0.6 0.6
0.5
0.4
0.4 0.3
0.2
0.2 0.1
0
0 -0.1
0.6 0.6
0.4 0.4
0.2 0.2
-0.6 0 -0.6 0
-0.4 -0.4
-0.2 -0.2 -0.2 -0.2
0 0
0.2 -0.4 0.2 -0.4
0.4 0.4
0.6 -0.6 0.6 -0.6
Figure 2.3 Plots of the true and approximate solution for convection-diffusion Example 2.
24
Table 2.2 Approximation errors and convergence rates for Example 2.
=1 2/2 Time T = 0.3
(t, h) 1
( 10 , 82 ) 1
( 20 , 162 ) 1
( 40 , 322 ) 1
( 80 , 642 ) 1
( 160 , 1282 )
0 |ku uh k|0,1 1.242e-0 7.999e-1 3.394e-1 1.543e-1 7.604e-2
Cvge. Rate - 0.6 1.2 1.1 1.0
|ku uh k|,0 8.114e-2 4.071e-2 1.127e-2 2.572e-3 6.338e-4
Cvge. Rate - 1.0 1.9 2.1 2.0
h
2
|ku uh k|0,1 9.445e-1 7.978e-1 5.932e-1 4.079e-1 2.637e-1
Cvge. Rate - 0.2 0.4 0.5 0.6
|ku uh k|,0 6.609e-2 5.751e-2 4.353e-2 3.104e-2 2.057e-2
Cvge. Rate - 0.2 0.4 0.5 0.6
3/2
h
2
|ku uh k|0,1 9.899e-1 7.375e-1 3.912e-1 1.797e-1 8.356e-2
Cvge. Rate - 0.4 0.9 1.1 1.1
|ku uh k|,0 6.619e-2 4.548e-2 2.135e-2 8.268e-3 3.003e-3
Cvge. Rate - 0.5 1.1 1.4 1.5
2
h
2
|ku uh k|0,1 1.076e-0 7.519e-1 3.428e-1 1.553e-1 7.616e-2
Cvge. Rate - 0.5 1.1 1.1 1.0
|ku uh k|,0 7.102e-2 4.040e-2 1.281e-2 3.111e-3 7.702e-4
Cvge. Rate - 0.8 1.7 2.0 2.0
Example 3. This example problem of approximates the solution to a steep front moving
where
0.1A + 0.5B + C
w(, t) = ,
A+B+C
and b = [w(x, t), w(y, t)]T . The boundary and initial conditions are determined by the true
solution.
25
This example does not satisfy the assumptions for Theorem 2.6.1, as b 6= 0.
Nonetheless, the numerical results presented in Table 2.3 are consistent with those predicted
in (2.22) and (2.23). The numerical approximation using h = 2/32 and = h3/2 is
0.9 1
0.8 0.9
0.7 0.8
0.6 0.7
0.5 0.6
0.4 0.5
0.3 0.4
1 0.2 1 0.3
0.9 0.1 0.9 0.2
0.8 0.8 0.1
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
1 1
0.8 0.8
0 0.6 0 0.6
0.2 0.4 0.2 0.4
0.4 0.4
0.6 0.2 0.6 0.2
0.8 0.8
1 0 1 0
Figure 2.4 Plots of the true and approximate solution for convection-diffusion Example 3.
26
CHAPTER 3
3.1 Introduction
formulated. At each time increment the -method decouples the velocity-pressure and stress
updates. Additionally the systems for the velocity-pressure update and the stress update
are linear. Motivated by this decoupling, the analysis of the method is approached similarly.
Analysis of a Stokes system for the velocity and pressure is examined first, assuming the
stress is known. The hyperbolic constitutive model is analyzed next, assuming the velocity
and pressure are known. The estimates obtained when analyzing each of the decoupled
parts are then used to establish a priori error estimates for the full implementation of the
-method.
The remainder of this chapter is organized as follows. In Section 3.2 the -method
for the viscoelastic modeling equations is formulated. Section 3.3 contains the mathematical
notation and variational formulations used in the analysis. Computability of the algorithm
is shown in Section 3.4. A priori error estimates for the -method are stated and proved in
Section 3.5.
and is ignored) viscoelastic fluid flow using a Johnson-Segalman constitutive equation are:
27
+ + u + ga (, u) 2d(u) = 0 in , (3.1)
t
u
Re + p 2(1 ) d(u) = f in , (3.2)
t
u = 0 in , (3.3)
u = 0 on , (3.4)
and (3.2). Use the splitting parameters and (0, 1), and define:
Constitutive equation:
1 G := , (3.7)
Conservation of momentum:
Let t denote the temporal increment between times tn and tn+1 , and for c {, , , a, }
(n+) (n)
+ 1 G (n+) = 2 G (n) .
t
28
Step 1b: (Solve for velocity and pressure.)
u(n+) u(n)
Re + p(n+) + 1 Fu (n+) = 2 Fu (n) ,
t
u(n+) = 0.
u(n+) u(n+)
Re + p(n+) + 2 Fu (n+) = 1 Fu (n+) ,
(1 2) t
u(n+) = 0.
(n+) (n+)
+ 2 G (n+) = 1 G (n+) .
(1 2) t
Step 3a and Step 3b: The temporal advancement to time tn+1 is com-
pleted
by repeating, Step 1a, and Step 1b with (n) and (n + ) replaced by
n + and (n + 1), respectively.
the approximation of the non-linear system of equations only requiring the solution of linear
systems of equations.
The following function spaces are defined for use in the analysis:
X := H01 () := u H 1 () : u = 0 on ,
n o
S := = (ij ) : ij = ji ; ij L2 (); 1 i, j d
= (ij ) : u L2 (), u X ,
Z
2 2
Q := L0 () = q L () : q dx = 0 ,
Z
Z := vX: q( v) dx = 0, q Q .
(q, v)
inf sup > 0. (3.11)
qQ vX kqk kvk1
29
A variational formulation of (3.1)-(3.3), found by multiplication of the modeling
equations by test functions and integrating over , is: Given u0 X and 0 S find
As the velocity and pressure spaces X and Q satisfy the inf-sup condition (3.11), an equiv-
= K, K Th .
As Pk (A) denotes the space of polynomials on A of degree no greater than k and C()d the
space of vector valued functions with d components which are continuous on . Then the
30
associated finite element spaces are defined by:
n o
v X C()d : vK Pk (K) K Th ,
Xh :=
n o
S C()dd : K Pm (K) K Th ,
Sh :=
Qh := q Q C() : q K Pq (K) K Th ,
Zh := {v Xh : (q, v) = 0 q Qh } .
Analogous to the continuous spaces assume that Xh and Qh satisfy the discrete inf-sup
condition:
(q, v)
inf sup > 0. (3.21)
qQh vXh kqk kvk1
The analysis for the -method is accomplished in three parts: the analysis of Stokes-
like problem using a known true stress; the analysis of the constitutive model assuming a
known true velocity and pressure; and finally a coupling of these estimates establishes the
full a priori error estimates for the -method. For the analysis it is helpful to define
uh := uh uh ,
h := h h .
method for viscoelastic flow described by Steps 1a - 3b above. Letting U and S denote
(n)
(n) = u(n) U (n) , E (n) = U (n) uh ,
(n)
(n) = (n) S (n) , F (n) = S (n) h ,
(n) (n) (n) (n)
eu = u(n) uh , e = (n) h .
The following properties of Sobolev and finite element spaces are needed in the
analysis. From [26] for (u, p) H k+1 ()d H q+1 () there exists (U, P) Zh Qh such
31
that
A skew symmetric formulation is used when bounding some of the convective terms [29].
Define
c(u, , ) := (u , ) , (3.25)
1
c(u, , ) := (c(u, , ) c(u, , )) , (3.26)
2
c(u, , ) = 0.
Computability implies that the coefficient matrix associated with the variational formu-
lation of each step of the -method algorithm is invertible. To stabilize the hyperbolic
to avoid spurious oscillations in the approximation. This is implemented by testing all terms
in the constitutive equation (except the discretized temporal derivative) against modified
()
() := + uh , (3.27)
h
and is a small positive constant. Note that = 0 gives the standard Galerkin method.
The variational formulations for the steps in the -method approximation are as follows.
32
(n+)
Step 1a: Find h Sh such that
(n+)
(n+)
h , + h , (n)
t h
(n)
(n)
(n) (n)
= h , h , (n) uh h , (n)
t h h
(n) (n) (n)
ga ( h , uh ), (n) + 2 d(uh ), (n) , Sh . (3.28)
h h
(n+)
Step 1b: Find uh Zh such that
Re (n+)
(n+)
Re (n)
uh , v + 2(1 ) d(uh ), d(v) = uh , v
t t
(n) (n+)
2(1 ) d(uh ), d(v) + f (n+) , v h , d(v) , v Zh . (3.29)
(n+)
Step 2a: Find uh Zh such that
Re (n+) (n+) Re
(n+)
uh , v + 2(1 ) d(uh ), d(v) = uh ,v
(1 2)t (1 2)t
(n+) (n+)
2(1 ) d(uh ), d(v) + f (n+) , v h , d(v) , v Zh . (3.30)
(n+)
Step 2b: Find h Sh such that
(n+) (n+) (n+) (n+)
h , + h , (n+) + uh h , (n+)
(1 2)t h h
(n+) (n+) (n+)
+ ga ( h , uh ), (n+) 2 d(uh ), (n+)
h
h
(n+)
(n+)
= h , h , (n+) , Sh . (3.31)
(1 2)t h
(n+1)
Step 3a: Find h Sh such that
(n+1) (n+1) (n+) (n+)
h , + h , (n+) = h , h , (n+)
t h t h
(n+) (n+) (n+) (n+)
uh h , (n+) ga ( h , uh ), (n+)
h h
(n+)
+ 2 d(uh ), (n+) , Sh . (3.32)
h
33
(n+1)
Step 3b: Find uh Zh such that
Re (n+1)
(n+1)
Re (n+)
uh , v + 2(1 ) d(uh ), d(v) = uh ,v
t t
(n+)
(n+1)
2(1 ) d(uh ), d(v) + f (n+1) , v h , d(v) , v Zh . (3.33)
The following induction hypothesis is used when proving the lemmas that establish
Induction Hypothesis 1 Under the assumptions of Theorem 3.6.1 there exists a constant
Lemma 3.4.1 (Step 1a) Assume Induction Hypothesis 1 is true. For Ch and t
(n+)
sufficiently small there exists a unique solution h Sh satisfying (3.28).
(n+)
(n) (n)
A1 h , = h , (1 ) h , (n)
t h
(n) (n) (n) (n)
uh h + ga ( nh , uh ) , (n) + 2 d(uh ), (n) , Sh , (3.35)
h h
where
(n+)
(n+)
(n+)
A1 h , := h , + h , (n) .
t h
Here (3.35) represents a square linear system of equations Ax = b. With the choice =
(n+)
h , the individual terms in A1 are
34
and
(n)
(n+)
2
1
(n+) (n) (n+)
h , uh h d2 Ch1
uh
h
(n+)
2
1
K2 d2 Ch1
h
.
1 (n+) (n+)
Provided Ch and t /K2 d2 C, then A1 ( h , h ) > 0. Thus, ker (A1 ) =
(n+)
Lemma 3.4.2 (Step 1b) There exists a unique solution uh Zh satisfying (3.29).
where
(n+)
Re (n+)
(n+)
A 2 uh , v := uh , v + 2(1 ) d(uh ), d(v) .
t
(n+)
Note that choosing v = uh
(n+) (n+)
Re (n+) (n+)
(n+) (n+)
A 2 uh , uh = uh , uh + 2(1 ) d(uh ), d(uh ) > 0.
t
Thus, ker(A2 ) = {0}, and existence and uniqueness of a solution to (3.29) has been shown.
(n+)
Lemma 3.4.3 (Step 2a) There exists a unique solution uh Zh satisfying (3.30).
where
(n+) Re (n+) (n+)
A 3 uh ,v := uh , v + 2(1 ) d(uh ), d(v) .
(1 2)t
35
(n+) (n+) (n+)
For v = u , A 3 uh , uh > 0. Thus, ker(A3 ) = {0}, and it then follows that
Lemma 3.4.4 (Step 2b) Assume Induction Hypothesis 1 is true. For Ch and t
(n+)
sufficiently small there exists a unique solution h Sh satisfying (3.31).
with
(n+) (n+) (n+)
A4 h , := h , + (1 ) h , (n+)
(1 2)t h
(n+) (n+) (n+) (n+)
+ uh h , (n+) + ga ( h , uh ), (n+) .
h h
(n+) (n+)
Bounding the terms in A4 h , h yields
(n+)
2
(n+) (n+)
h , h =
,
(1 2)t (1 2)t
h
(n+)
2
(n+) (n+)
(1 ) h , h = (1 )
h
,
(n+)
2
(n+) (n+) (n+) 1
1
(1 ) h , uh h (1 )d 2 Ch K2
h
,
(n+)
2
(n+) (n+) (n+) 1
1
uh h , h d 2 Ch K2
h
,
2
(n+) (n+) (n+) (n+)
(n+) (n+)
uh h , uh h =
uh
h
,
36
(n+) (n+) (n+)
(n+) ( n+)
(n+)
ga ( h , uh ), h 4
h uh
h
2
1
(n+)
(n+)
4d2
uh
h
(n+)
(n+)
2
1
1
4d2 Ch
uh
h
(n+)
2
1
1
4d Ch K2
h
2
,
(n+) (n+) (n+) (n+)
ga ( h , uh ), uh h
(n+) ( n+)
(n+) (n+)
4
h uh
u
h h
1
(n+)
(n+)
(n+) (n+)
4d 2
uh
h
uh
h
(n+)
2
2
2 h2 K2 2
4dC
+ 1
u(n+) (n+)
.
h
h h
1
Thus,
(n+) (n+)
A4 ( h , h )
!
2 h2 K2 2
(n+)
2
1
1 4dC
+ (1 ) ((1 ) + 2) d Ch K2
2
h
(1 2)t 1
2
(n+) (n+)
+ (1 1 )
uh
h
.
1 (n+) (n+)
Choosing 1 = 2, Ch, and t Ch establishes A4 ( h , h ) > 0. Hence
The unique solvability of (3.32) and (3.33), representing the third step in the algo-
3.5 A Priori Error Estimates for the Conservation Equations and Constitutive Equation
A priori error estimates are formulated first for a -method for the Stokes-like prob-
lem given by (3.29), (3.30), and (3.33), assuming the stress is known. Then a priori error
estimates for a -method for the constitutive equation given by (3.28), (3.31), and (3.32)
37
are found assuming the velocity and pressure are known. These estimates are given in
t Ch2 , the fractional step -method approximation, uh given by Step 1b, Step 2a, and
Step 3b converges to u on the interval (0, T ] as t, h 0, and satisfies the error estimates:
where
Proof: For notational simplicity in this proof the tilde is dropped from uh and ph , the
discrete approximations of velocity and pressure using the true stress. Following the analysis
of the -method for the convection-diffusion equation, linear combinations are formed to
n+
( )
(n+1) (n) (n) (n+) (n+1)
obtain expressions for uh uh , uh uh , and uh uh .
38
This combination gives
(n+1) (n)
!
uh uh
Re ,v + (n+) , d(v) + (n+1) , d(v)
t
(n) (n+)
+2(1 ) d(uh ), d(v) + 2(1 ) d(uh ), d(v)
(n+)
(n+1)
+2(1 ) d(uh ), d(v) + 2(1 ) d(uh ), d(v)
= f (n+) , v + f (n+1) , v , v Zh . (3.41)
Thus,
B1n (uh , v) = f (n+) , v + f (n+1) , v , (3.42)
where B1n (uh , v) denotes the left hand side of (3.41). The true solution of the momentum
where
(n+ 21 )
1
I1n (v) := Re dt u (n+1)
ut , v + (n+) + (n+1) (n+ 2 ) , d(v)
+ 2 d(u(n) ) + d(u(n+) ) + d(u(n+1) )
(n+) (n+ 12 )
+ d(u )d u , d(v) , v Zh . (3.45)
39
Linear combination 2: The combination
(3.29) + (1 2) (3.30) + (3.33) , (3.47)
with nn with nn with nn1
yields
(n+) (n)
uh uh
(n+)
Re , v + (n+) , d(v) + (n) , d(v) + 2 d(uh ), d(v)
t
(n+)
(n)
+ 2 d(uh ), d(v) + (1 2) 2 d(uh ), d(v)
(n)
+ 2 d(uh ), d(v) = f (n+) , v + f (n) , v , v Zh . (3.48)
Giving,
B2n (uh , v) = f (n+) , v + f (n) , v , (3.49)
where B2n (uh , v) denotes the left hand side of (3.48). The true solution of the momentum
where
(n+ 12 ) 1
I2n (v) := Re dt u (n+)
ut , v + (n+) + (n) (n+ 2 ) , d(v)
+ 2 d(u(n+) ) + d(u(n) ) + (1 2) d(u(n+) )
(n) (n+ 12 )
+ d(u )d u , d(v) , v Zh . (3.52)
40
The error expression is obtained by subtracting (3.48) from (3.51), giving
1
1
B2n (eu , v) = f (n+ 2 ) f (n+) f (n) , v + p(n+ 2 ) , v + I2n (v), v Zh .
(3.53)
yields
(n+) (n+1)
!
uh uh
(n)
(n+)
Re ,v + 2 d(uh ), d(v) + 2 d(uh ), d(v)
t
(n) (n+1)
+ 2 d(uh ), d(v) + 2 (1 2) d(uh ), d(v)
+ (n+) , d(v) + (1 2) (n+1) , d(v) + (n) , d(v)
= f (n+) , v + (1 2) f (n+1) , v + f (n) , v , v Zh . (3.55)
Thus,
B3n (uh , v) = f (n+) , v + (1 2) f (n+1) , v + f (n) , v , (3.56)
where B3n (uh , v) denotes the left hand side of (3.55). The true solution of the momentum
41
with
(n+ 21 )
I3n (v) := Re dt u (n+)
ut ,v
1
+ (n+) + (1 2) (n+1) + (n) (n+ 2 ) , d(v)
+ 2 d(u(n) ) + d(u(n+) ) + d(u(n) )
(n+1) (n+ 12 )
+ (1 2) d(u ) d(u ), d(v) , v Zh . (3.59)
Subtracting (3.55) from (3.58) gives the final error expression for the conservation equations:
1
B3n (eu , v) = f (n+ 2 ) f (n+) (1 2) f (n+1) f (n) , v
1
+ p(n+ 2 ) , v + I3n (v), v Zh . (3.60)
The analysis is completed by bounding the terms in (3.46),(3.53), and (3.60), with the test
Bounding B1n (eu , E (n+1) ), B2n (eu , E (n+) ), and B3n (eu , E (n+) ):
The following estimates are used to bound the terms in B1n (eu , E (n+1) ), B2n (eu , E (n+) ),
and
(n+1) Re
(n+1)
(n+1)
2
(n)
2
Re dt E ,E
E
E
, (3.61)
2t
Re
2
2
Re dt (n+1) , E (n+1)
dt (n+1)
+ Re
E (n+1)
. (3.62)
4
Similarly,
(n+) (n+)
Re dt eu ,E = Re dt E (n+) , E (n+) + Re dt (n+) , E (n+) ,
42
(n+) (n+)
Re
(n+)
2
(n)
2
Re dt E ,E
E
E
, (3.63)
2t
and
Re
2
2
Re dt (n+) , E (n+)
dt (n+)
+ Re
E (n+)
. (3.64)
4
Also
(n+)
Re dt eu , E (n+) = Re dt E (n+) , E (n+) + Re dt (n+) , E (n+) ,
(n+) (n+)
Re
(n+)
2
(n+1)
2
Re dt E ,E
E
E
, (3.65)
2t
and
Re
2
(n+)
2
Re dt (n+) , E (n+) d (n+)
+ Re E
. (3.66)
t
4
2 2 2
2
2
2 d((n+1) ), d(E (n+1) ) (n+1)
+ d(E (n+1)
)
. (3.68)
1
1
The term
(n+)
2 d(eu ), d(E (n+1) ) = 2 d(E (n+) ), d(E (n+1) )
+ 2 d((n+) ), d(E (n+1) ) ,
is bounded using
2 2 2
2
2
2 d(E (n+) ), d(E (n+1) ) E (n+)
+ d(E (n+1)
)
, (3.69)
2
2
and
2 2 2
2
2
2 d((n+) ), d(E (n+1) )
(n+)
+ 3
d(E (n+1) )
. (3.70)
3
43
Next
(n)
2 d(eu ), d(E (n+1) ) = 2 d(E (n) ), d(E (n+1) )
+2 d((n) ), d(E (n+1) ) ,
and
2 2 2
2
2
2 d((n) ), d(E (n+1) )
(n)
+ 5
d(E (n+1) )
. (3.72)
5
Write
(n+)
2 d(eu , d(E (n+1)
) = 2 d(E (n+) ), d(E (n+1) )
+ 2 d( ( n+)
), d(E (n+1)
) ,
and
2 2 2
2
2
2 d((n+) ), d(E (n+1) ) (n+)
+ d(E (n+1)
)
. (3.74)
7
7
+ 2 d((n+) ), d(E (n+) ) ,
with
2 2 2
2
2
2 d(E (n+)
), d(E (n+)
)
E
(n+)
+ 8
d(E
(n+)
)
, (3.75)
8
44
and
2 2 2
2
2
2 d((n+) ), d(E (n+) )
(n+)
+ 9
d(E (n+) )
. (3.76)
9
Next
2 d(eu ), d(E (n+) ) = 2 d(E (n) ), d(E (n+) )
(n)
+2 d((n) ), d(E (n+) ) ,
with
2 2
2
2
2 d(E (n)
), d(E (n+)
) (n)
E
+ 10
d(E
(n+)
)
(3.77)
10
and
2 2
2
2
2 d((n) ), d(E (n+) )
(n)
+ 11
d(E (n+) )
. (3.78)
11
The term
(n+)
2 (1 2) d(eu ), d(E (n+)
) = 2 (1 2) d(E (n+)
), d(E (n+)
)
+ 2 (1 2) d((n+) ), d(E (n+) ) ,
and
2 (1 2) d((n+) ), d(E (n+) ) (3.80)
(1 2)2 2 2
2
2
(n+)
+ 12
d(E (n+) )
. (3.81)
12
The term
2 d(eu ), d(E (n+) ) = 2 d(E (n) ), d(E (n+) )
(n)
+2 d((n) ), d(E (n+) ) ,
45
gives
2 2 2
2
2
2 d(E (n) ), d(E (n+) )
E (n)
+ 13
d(E (n+) )
, (3.82)
13
and
2 2 2
2
2
2 d((n) ), d(E (n+) )
(n)
+ 14
d(E (n+) )
. (3.83)
14
Set
(n+)
2 d(eu ), d(E (n+) ) = 2 d(E (n+) ), d(E (n+) )
+ 2 d((n+) ), d(E (n+) ) ,
and
2 2 2
2
2
2 d((n+) ), d(E (n+) )
(n+)
+ 15
d(E (n+) )
. (3.85)
15
Also
(n)
2 d(eu ), d(E (n+) ) = 2 d(E (n) ), d(E (n+) )
+2 d((n) ), d(E (n+) ) ,
where
2 2
2
2
(n) (n+) (n)
(n+)
2 d(E ), d(E )
E
+ 16
d(E )
, (3.86)
16
and
2 2
2
2
2 d((n) ), d(E (n+) )
(n)
+ 17
d(E (n+) )
. (3.87)
17
46
Next,
(n+1)
2 (1 2) d(eu ), d(E (n+) ) = 2 (1 2) d(E (n+1) ), d(E (n+) )
+ 2 (1 2) d((n+1) ), d(E (n+) ) ,
with
2 (1 2)2 2
2
2 (1 2) d(E (n+1) ), d(E (n+) )
E (n+1)
18
2
+ 18
d(E (n+) )
(3.88)
and
2 (1 2)2 2
2 (1 2) d((n+1) ), d(E (n+) )
(n+1)
19
2
+ 19
d(E (n+) )
. (3.89)
Lastly,
(n)
2 d(eu ), d(E (n+) ) = 2 d(E (n) ), d(E (n+) )
+ 2 d((n) ), d(E (n+) ) ,
where
2 2 2
2
2
2 d(E (n) ), d(E (n+) )
E (n)
+ 20
d(E (n+) )
(3.90)
20
and
2 2 2
2
2
2 d((n) ), d(E (n+) )
(n)
+ 21
d(E (n+) )
. (3.91)
21
47
Bounds for I1n (E (n+1) ), I2n (E (n+) ), and I3n (E (n+) ) (see (3.45) , (3.52), and (3.59)):
1
+ (n+) + (n+1) (n+ 2 ) , d(E (n+1) )
1
+ 2 d(u(n+1) ) + d(u(n+) ) d u(n+ 2 ) , d(E (n+1) )
1
+ 2 d(u(n) ) + d(u(n+) ) d u(n+ 2 ) , d(E (n+1) ) . (3.92)
1
(n+) + (n+1) (n+ 2 ) , d(E (n+1) )
1
1
2
2
+ (n+1) (n+ 2 )
+ 22
d(E (n+1) )
, (3.94)
(n+)
422
1
2 d(u(n+1) ) + d(u(n+) ) d u(n+ 2 ) , d(E (n+1) )
2 2
(n+1) (n+)
(
2
n+ 12 )
2
(n+1)
d(u ) + d(u )d u
+ 23
d(E )
, (3.95)
23
and
1
2 d(u(n) ) + d(u(n+) ) d u(n+ 2 ) , d(E (n+1) )
2 2
1
2
2
d(u(n) ) + d(u(n+) ) d(u(n+ 2 ) )
+ 24
d(E (n+1) )
. (3.96)
24
48
Now examine the bounds for I2n term in (3.52) with v = E (n+) .
1
(n+ ) (n+)
I2n (E (n+) ) := Re dt u(n+) ut 2 ,E
1
+ (n+) + (n) (n+ 2 ) , d(E (n+) )
1
+2 d(u(n+) ) + d(u(n) ) d u(n+ 2 ) , d(E (n+) )
+2 (1 2) d(u(n+) ) + d(u(n) )
1
+d(u(n) ) d u(n+ 2 ) , d(E (n+) ) . (3.97)
Then
(n+) (n+ 21 )
Re dt uh ut , E (n+)
2
Re
(n+) (n+ 12 )
(n+)
2
d t u u t
+ Re E
, (3.98)
4
1
(n+) + (n) (n+ 2 ) , d(E (n+) )
1
1
2
2
+ (n) (n+ 2 )
+ 25
d(E (n+) )
, (3.99)
(n+)
425
1
2 d(u(n+) ) + d(u(n) ) d u(n+ 2 ) , d(E (n+) )
2 2
1
2
(1 )d(u(n+) ) + d(u(n) ) d u(n+ 2 )
26
2
+ 26
d(E (n+) )
, (3.100)
and
1
2 (1 2) d(u(n+) ) + d(u(n) ) + d(u(n) ) d u(n+ 2 ) , d(E (n+) )
2 2
1
2
(1 2) d(u(n+) ) + d(u(n) ) + d(u(n) ) d u(n+ 2 )
27
2
+ 27
d(E (n+) )
. (3.101)
49
Next consider the bounds for I3n term in (3.59) with v = E (n+) .
(n+ 12 )
I3n (E (n+) ) := Re dt u (n+)
ut ,E (n+)
1
+ (n+) + (1 2) (n+1) + (n) (n+ 2 ) , d(E (n+) )
1
+ 2 d(u(n+) ) + d(u(n) ) + (1 2) d(u(n+1) ) d(u(n+ 2 ) ), d(E (n+) )
1
+ 2 d(u(n) ) + d(u(n) ) d(u(n+ 2 ) ), d(E (n+) ) . (3.102)
1
(n+) + (1 2) (n+1) + (n) (n+ 2 ) , d(E (n+) )
1
(n+) (n+1) (n) (
2
n+ 12 )
+ (1 2) +
428
2
+ 28
d(E (n+) )
, (3.104)
1
2 d(u(n+) ) + d(u(n) ) + (1 2) d(u(n+1) ) d(u(n+ 2 ) ), d(E (n+) )
2 2
(n+) (n) (n+1) (
2
n+ 21 )
d(u ) + d(u ) + (1 2) d(u ) d(u )
29
2
+ 29
d(E (n+) )
(3.105)
and
1
2 d(u(n) ) + d(u(n) ) d(u(n+ 2 ) ), d(E (n+) )
2 2
1
2
2
d(u(n) ) + d(u(n) ) d(u(n+ 2 ) )
+ 30
d(E (n+) )
. (3.106)
30
50
Bounding the f and p terms in (3.46),(3.53), and (3.60):
Bounds for the terms containing the forcing function f are given by
1
1
(n+ 12 )
2
f (n+ 2 ) f (n+) f (n+1) , E (n+1)
f f (n+) f (n+1)
4
2
+
E (n+1)
, (3.107)
1
1
(n+ 12 )
2
f (n+ 2 ) f (n+) f (n) , E (n+)
f f (n+) f (n)
4
2
+
E (n+)
, (3.108)
and
1
f (n+ 2 ) f (n+) (1 2) f (n+1) f (n) , E (n+)
1
1
2
2
f (n+ 2 ) f (n+) (1 2) f (n+1) f (n)
+
E (n+)
. (3.109)
4
The pressure terms are bounded making note that E (n+1) , E (n+) , and E (n+) Zh . Thus,
1
1 1
p(n+ 2 ) , E (n+1) = p(n+ 2 ) P (n+ 2 ) , E (n+1)
1
1
1
2
p(n+ 2 ) P (n+ 2 )
d2
E (n+1)
1
d2
(n+ 21 ) (
2
n+ 12 )
2
(n+1)
p P
+
E
, (3.110)
4
d21
2
2
(n+ 21 ) (n+)
(n+ 21 ) 1
p , E
p P (n+ 2 )
+
E (n+)
, (3.111)
4
and
1
d21
2
2
(n+ 21 ) n+ 12 )
p(n+ 2 ) , E (n+) p P ( + E (n+)
. (3.112)
4
51
Using the estimates given in (3.61) - (3.112), three error expressions are written as
Re
(n+1)
2
(n)
2
E
E
2t
2
+
d(E (n+1) )
(2 1 2 3 4 5
Re
(n+)
2
(n)
2
E
E
2t
2
+
d(E (n+) )
(2 (1 2) 8 9 10 11 12
and
Re
(n+)
2
(n+1)
2
E
E
2t
2
+
d(E (n+) )
(2 15 16 17 18 19
52
where,
1
(n+ 21 )
2
Hf1 :=
f f (n+) f (n+1)
,
4
1
(n+ 21 )
2
Hf2 :=
f f (n+) f (n)
,
4
1
(n+ 12 )
2
Hf3 :=
f f (n+) (1 2) f (n+1) f (n)
,
4
1
d2
(n+ 12 ) 1
2
Hp1 :=
p P (n+ 2 )
,
4
1
d2
(n+ 12 )
2
(n+ 21 )
Hp2 := p P
,
4
1
d2
(n+ 21 ) 1
2
Hp3 :=
p P (n+ 2 )
,
4
Re
2 2 2 2
2 2 2 2
2
H1 :=
dt (n+1)
+
(n+1)
+
(n+)
4 1 3
2 2 2
2 2 2
2
2
+
(n)
+
(n+)
,
5 7
Re
2 2 2 2
2 2 2
2
H2 :=
dt (n+)
+
(n+)
+
(n)
4 9 11
2 2 2
2 2 2
(1 2)
2
2
+
(n+)
+
(n)
,
12 14
Re
2 2 2 2
2 2 2
2
H3 :=
dt (n+)
+
(n+)
+
(n)
4 15 17
2 2 2 2 2 2
(n)
2 (1 2)
(n+1)
+ +
, (3.116)
21 19
2
Re
(n+1)
+ 1
(n+ 12 )
(n+) (n+1)
2
(n+ 21 )
HI n 1 :=
d t u u t +
4
422
2 2
1
2
+
d(u(n+1) ) + d(u(n+) ) d u(n+ 2 )
23
2 2
1
2
+
d(u(n) ) + d(u(n+) ) d(u(n+ 2 ) )
, (3.117)
24
53
2
Re
(n+)
+ 1
(n+ 21 )
(n+) (n)
2
(n+ 12 )
HI n 2 :=
d t u u t +
4
425
2 2
(n+) (n)
(
2
n+ 12 )
+ (1 )d(u ) + d(u ) d u
26
2 2
1
2
+
(1 2) d(u(n+) ) + d(u(n) ) + d(u(n) ) d u(n+ 2 )
,(3.118)
27
and
2
Re
(n+) (n+ 21 )
HI n 3 :=
dt u ut
4
1
(n+)
(n+1) (n) (
2
n+ 12 )
+ + (1 2) +
428
2 2
1
2
+
d(u(n+) ) + d(u(n) ) + (1 2) d(u(n+1) ) d(u(n+ 2 ) )
29
2 2
1
2
+
d(u(n) ) + d(u(n) ) d(u(n+ 2 ) )
. (3.119)
30
54
2t
Multiplying (3.113), (3.114), and (3.115) by Re and summing each from (n = 0 to N 1),
(N )
2
(N 1+)
2
(N 1+)
2
E
+
E
+
E
+ (2 1 2 3 4 5
l
X 2t
2
6 7 22 23 24 )
d(E (n+1) )
Re
n=0
+ (2 (1 2) 8 9 10 11 12
l
2t
2
d(E (n+) )
X
13 14 25 26 27 )
Re
n=1
+ (2 15 16 17 18 19
l
X 2t
2
20 21 28 29 30 )
d(E (n+) )
Re
n=1
N 1 N 1
2t
(n+1)
2 X 2t
2
(2Re + 1)
E (n+)
X
(2Re + 1)
E
+
Re Re
n=0 n=1
N 1
2 NX 1
2t 2 2 2
X 2t
2
+ (2Re + 1)
E (n+)
+ E (n)
Re Re 4
n=1 n=0
N 1
2 NX 1
2t 2 2 2 2
X 2t
2
+ +
E (n)
+
E (n+1)
Re 10 16 Re
n=1 n=0
N 1
2 NX 1
!
X 2t
(n+)
2t 2 2 2
( n+)
2
+ E + E
Re Re 6
n=1 n=0
N 1
2 NX 1
! !
X 2t 2 2 2
2t 2 2 2
2
+ 1+
E (n+)
+
E (n+)
Re 8 Re 2
n=1 n=0
N 1
!
X 2t 2 2 2 2 2 2
2
(n)
+ +
E
Re 13 20
n=1
N 1
!
X 2t 2 (1 2)2 2
2
+
E (n+1)
Re 18
n=1
2
2
2
+H +
E (0)
+
E ()
+
E ()
, (3.120)
55
where
N 1 N 1 N 1
2t 2t 2t
H
X X X
= H1 + H2 + H
Re Re Re 3
n=0 n=1 n=1
N 1 N 1 N 1
X 2t X 2t X 2t
+ HI n 1 + HI n 2 + HI n 3
Re Re Re
n=0 n=1 n=1
N 1 N 1 N 1
X 2t X 2t X 2t
+ Hf + Hf + Hf
Re 1 Re 2 Re 3
n=0 n=1 n=1
N 1 N 1 N 1
X 2t X 2t X 2t
+ Hp1 + Hp2 + Hp3 . (3.121)
Re Re Re
n=0 n=1 n=1
For sufficiently small values of i , i {1. . . . , 30}, the terms on the left hand side of expres-
2 2 2 2 2 2 2
2t 2
1 2Re + 1 + Ch 1+ + + , (3.122)
Re 4 10 16
!!
2t 2 2 2 2 2 2 2 2 2
1 2Re + 1 + Ch2 1+ + + , (3.123)
Re 6 13 20
!!
2t 2 2 2 2 2 2 2 2 (1 2)2 2
1 2Re + 1 + Ch 1+ + + . (3.124)
Re 2 8 18
t C + Ch2 1.
t Ch2 . (3.125)
For i = 1, 2, 3, with j = 0 if i = 1, and j = 1 if i > 1, bounds are found for the expressions
PN 1 2t PN 1 2t PN 1 2t PN 1 2t
n=j Re Hi , n=j Re HI n i , n=j Re Hfi , and n=j Re Hpi in (3.120) using interpo-
lation properties, and approximation properties of the div-free velocity space Zh and the
56
PN 1 2t
pressure space Qh . Typical n=j Re Hi terms are bounded as
N 1 N 1
Z tn+1
2
X 1
2 Xt
1
t
dt (n+1)
=
dt
2 2 t tn
t
n=0 n=0
N 1
!
2
Z Z tn+1 Z tn+1
X 1
1 dt dt dA
2t tn tn t
n=0
N 1 Z Z tn+1
2
X 1
dt dA
2 tn t
n=0
Ch2k+2 kut k20,k+1 (3.126)
and
N
X 1
2 N
X 1
2
(n+1)
2k
tC1
Ch t
u(n+1)
k+1
n=0 n=0
Ch2k |kuk|20,k+1 . (3.127)
Thus,
N 1
X 2t
H Ch2k+2 kut k20,k+1 + Ch2k |kuk|20,k+1 . (3.128)
Re i
n=j
PN 1 2t
Taylor series expansions are used to bound the n=j Re HI i
n terms. The results follow
and
N 1
X 2t
Hp Ch2q+2 |kpk|20,q+1 . (3.131)
Re
n=0
57
Next, apply the discrete Gronwalls inequality to (3.120) to obtain
(N )
2
(N 1+)
2
(N 1+)
2
E
+
E
+
E
2
+ (2 1 2 3 4 5
Re
NX1
2
6 7 22 23 24 ) t
d(E (n+1) )
n=0
2
+ (2 (1 2) 8 9 10 11 12
Re
N 1
2
t
d(E (n+) )
X
13 14 25 26 27 )
n=0
2
+ (2 15 16 17 18 19
Re
N
X 1
2
20 21 28 29 30 ) t
d(E (n+) )
n=0
Gu (t, h), (3.132)
where
+ C(t)4 CT . (3.133)
2
2
Note in order obtain suitable error estimates for
E ()
and
E ()
the -method is
initialized using a Crank-Nicolson method for Steps 1b and 2a of the first time step. A
2
2
description of the specific scheme implemented and bounds for
E ()
and
E ()
are
58
Thus,
N
(n)
2
X
|kEk|20,0 = t
E
n=1
N
X
Gu (t, h)t 1
n=1
= T Gu (t, h). (3.135)
Hence
and
Theorem 3.5.2 (Assuming u and p are known) For sufficiently smooth solutions ,
u, p such that
59
2
with = 1 2 and t Ch2 , the fractional step -method approximation, h given by
Step 1a, Step 2b, and Step 3a converges to on the interval (0, T ] as t, h 0 and satisfies
where
2
F (t, h, ) := C(t) k ttt k0,0 + k tt k0,1 + k t k0,1 + kk0,1
+ k tt k0,0 + k t k0,0 + kk0,0 + CT
+ C(t) kk0,1 + k t k0,1 + kk0,0 + k t k0,0 + CT
Proof: For notational simplicity the tilde is dropped from h , the discrete approximation
using the true u and p, until the proof of Theorem 3.5.2 is complete. The upwinded test
To start the proof three linear combinations are established to obtain expressions for
n+
( )
(n+1) (n) (n) (n+) (n+1)
h h , h h , and h h .
60
yields
(n+1) (n+)
B4n ( h , ) := dt h , + h , (n)
(n+) (n+1)
+ (1 2) h , (n+) + h , (n+)
(n+) (n+)
+ h
(n)
, (n+) + h , (n) + u ( n+)
h , (n+)
(n+)
(n) (n)
+ u h , (n) + ga ( h , u (n+)
), (n+)
+ ga ( nh , u(n) ), (n) (1 )2 d(u(n+) ), (n+)
2 d(u(n) ), (n) = 0, Sh . (3.145)
The true solution for the constitutive equation evaluated at tn+ 1 satisfies:
2
1
!
(n+ 2 ) 1
1
0 = , (n+ 2 ) , + 2 d(u(n+ 2 ) ),
t
1 1 1 1
u(n+ 2 ) (n+ 2 ) + ga ( (n+ 2 ) , u(n+ 2 ) ), , S. (3.146)
61
where
(n+ 21 ) (n+ 12 )
I4n ( ) := dt (n+1)
t , + t (n)
, u
(n+ 12 )
1
+ t , u (n+)
+ (n+) (n+ 2 ) , (n)
(n+1) (n+ 21 ) (n+) (n+ 12 )
+ , (n+) + (1 2) , (n+)
1
1
+ (n) (n+ 2 ) , (n) + (n+) (n+ 2 ) , (n+)
1 1
+ u(n) (n) u(n+ 2 ) (n+ 2 ) , (n)
1 1
+ u(n+) (n+) u(n+ 2 ) (n+ 2 ) , (n+)
1 1
+ ga ( (n) , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) ), (n)
1 1
+ ga ( (n+) , u(n+) ) ga ( (n+ 2 ) , u(n+ 2 ) ), (n+)
1
+ 2 d(u(n+ 2 ) ) d(u(n) ), (n)
1
+ 2 d(u(n+ 2 ) ) d(u(n+) ), (n+) . (3.149)
62
gives
(n+)
(n+) (n+)
B5n ( h , ) := dt h ,
+ h , (n) + (1 2) h , (n+)
(n+)
(n) (n)
+ h , (n) + (1 2) h , (n+) + h , (n)
(n+)
(n)
+ h , (n) + (1 2) u (n+)
h , (n+)
(n) (n)
+ u(n) h , (n) + u(n) h , (n)
(n+)
+ (1 2) ga ( h , u ( n+) (n)
), (n+) + ga ( h , u(n) ), (n)
+ ga ( h
(n)
, u (n)
), (n) (1 2) 2 d(u (n+)
), (n+)
2 d(u(n) ), (n) 2 d(u(n) ), (n) = 0, Sh . (3.152)
The true solution to the constitutive equation evaluated at the midpoint between tn and
giving
63
where
(n+) (n+ 21 ) (n+ 12 )
I5n ( ) := dt t , + t (n)
, u
(n+ 21 ) (n+ 12 ) (n+ 12 ) (n)
+ (1 2) t , u + t , u
(n+) (n+ 12 ) (n+) (n+ 12 )
+ , (n) + (1 2) , (n+)
1
1
+ (n) (n+ 2 ) , (n) + (1 2) (n+) (n+ 2 ) , (n+)
1
1
+ (n) (n+ 2 ) , (n) + (n) (n+ 2 ) , (n)
1 1
+ (1 2) u(n+) (n+) u(n+ 2 ) (n+ 2 ) , (n+)
1 1
+ u(n) (n) u(n+ 2 ) (n+ 2 ) , (n)
1 1
+ u(n) (n) u(n+ 2 ) (n+ 2 ) , (n)
1 1
+ (1 2) ga ( (n+) , u(n+) ) ga ( (n+ 2 ) , u(n+ 2 ) ), (n+)
1 1
+ ga ( (n) , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) ), (n)
1 1
+ ga ( (n) , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) ), (n)
1
+2 d(u(n+ 2 ) ) d(u(n) ), (n)
1
+ (1 2) 2 d(u(n+ 2 ) ) d(u(n+) ), (n+)
1
+2 d(u(n+ 2 ) ) d(u(n) ), (n) . (3.156)
64
gives
(n+) (n+) (n)
B6n ( h , ) := dt h , + h , (n) + h , (n)
(n+1) (n) (n)
+ (1 2) h , (n) + h , (n) + h , (n)
(n) (n)
+ u(n) h , (n) + u(n) h , (n)
(n)
+ ga ( h , u(n) ), (n) + ga ( nh , u(n) ), (n)
2 d(u(n) ), (n) 2 d(u(n) ), (n) = 0, Sh . (3.159)
Evaluate the true solution at the midpoint between tn+1 and tn+ to get
1
!
(n+ 2 ) 1
1
0 = , (n+ 2 ) , + 2 d(u(n+ 2 ) ),
t
1 1 1 1
u(n+ 2 ) (n+ 2 ) + ga ( (n+ 2 ) , u(n+ 2 ) ), , S. (3.160)
65
where
(n+ 21 ) (n+ 21 )
I6n ( ) := dt (n+)
t , + t (n)
, u
(n+ 21 )
1
+ t , u (n)
+ (n+) (n+ 2 ) , (n)
(n) ( n+ 21 )
+ , (n)
1
+ (1 2) (n+1) (n+ 2 ) , (n)
1
1
+ (n) (n+ 2 ) , (n) + (n) (n+ 2 ) , (n)
1 1
+ u(n) (n) u(n+ 2 ) (n+ 2 ) , (n)
1 1
+ u(n) (n) u(n+ 2 ) (n+ 2 ) , (n)
1 1
+ ga ( (n) , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) ), (n)
1 1
+ ga ( n , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) ), (n)
1
+2 d(u(n+ 2 ) ) d(u(n) ), (n)
1
+2 d(u(n+ 2 ) ) d(u(n) ), (n) . (3.163)
This is the final error expression needed for the analysis. Bounds are now found for
Define
()
F () := F () + u() F () . (3.165)
66
Then
(n+1) (n+) (n+1)
B4n (, F (n+1) ) B4n ( h , F (n+1) ) = dt (e ), F (n+1) + e , F (n)
(n+1) (n+1) (n+) (n+1)
+ e , F (n+) + (1 2) e , F (n+)
( ) (n+1)
n+
(n) (n+1)
+ e , F (n+) + e , F (n)
(n+) (n+1)
+ u (n+)
e
(n)
, F (n+) + u(n) e , F (n)
(n+1)
(n+)
+ ga (e , u (n+) (n+1) (n)
), F (n+) + ga (e , u(n) ), F (n)
(n+1)
. (3.166)
where
(n+1) (n+1) (n) (n+1)
dt (F (n+1) ), F (n+1) = F ,F F ,F
t
(n+1)
2
(n+1)
(n)
F
F
F
t
(n+1)
2
(n)
2
F
F
(3.168)
2t
and
2
2
dt ((n+1) ), F (n+1)
dt (n+1)
+
F (n+1)
. (3.169)
4
Each of the next eight terms in (3.166) is rewritten as the sum of four terms.
(n+) (n+1)
e , F (n) = F (n+) , F (n+1) + F (n+) , u(n) F (n+1)
+ (n+) , F (n+1) + (n+) , u(n) F (n+1) , (3.170)
where
F (n+) , F (n+1)
F (n+)
F (n+1)
(n+)
2
(n+1)
2
F
+
F
, (3.171)
2 2
67
F (n+) , u(n) F (n+1)
F (n+)
u(n) F (n+1)
1
F (n+)
u(n)
d2
F (n+1)
1
K1 d 2
F (n+)
F (n+1)
K1 2 d
(n+)
2
2
F
+ 2
F (n+1)
, (3.172)
4
(n+)
2
(n+1)
2
(n+) , F (n+1) + F
, (3.173)
2 2
and
K1 2 d
(n+)
2
2
(n+) , u(n) F (n+1)
+ 2
F (n+1)
. (3.174)
4
Next,
e , F (n+) = F (n+1) , F (n+1) + F (n+1) , u(n+) F (n+1)
(n+1) (n+1)
+ (n+1) , F (n+1) + (n+1) , u(n+) F (n+1) , (3.175)
with bounds
2
F (n+1) , F (n+1) =
F (n+1)
, (3.176)
K 2 d
(n+1)
2
2
F (n+1) , u(n+) F (n+1)
1
F
+ 2
F (n+1)
, (3.177)
4
(n+1)
2
(n+1)
2
(n+1) , F (n+1)
+
F
, (3.178)
2 2
and
K 2 d
(n+1)
2
2
(n+1) , u(n+) F (n+1)
1
+ 2
F (n+1)
. (3.179)
4
Next
(n+) (n+1)
(1 2) e , F (n+) = (1 2) F (n+) , F (n+1)
+ (1 2) F (n+) , u(n+) F (n+1) + (1 2) (n+) , F (n+1)
+ (1 2) (n+) , u(n+) F (n+1) , (3.180)
68
has bounds
(1 2)
(n+)
2 (1 2)
(n+1)
2
(1 2) F (n+) , F (n+1)
F
+
F
, (3.181)
2 2
(1 2) K1 2 d
(n+)
2
(1 2) F (n+) , u(n+) F (n+1)
F
4
2
+ (1 2) 2
F (n+1)
, (3.182)
(1 2)
(n+)
2 (1 2)
(n+1)
2
(1 2) (n+) , F (n+1)
+ F
, (3.183)
2 2
and
(1 2) K1 2 d
(n+)
2
(1 2) (n+) , u(n+) F (n+1)
4
2
+ 2 (1 2)
F (n+1)
. (3.184)
Also
(n+) (n+1)
e , F (n+) = F (n+) , F (n+1) + F (n+) , u(n+) F (n+1)
+ (n+) , F (n+1) + (n+) , u(n+) F (n+1) ,
has bounds
(n+)
2
(n+1)
2
F (n+)
,F (n+1)
F
+
F
, (3.185)
2 2
K 2 d
(n+)
2
2
F (n+) , u(n+) F (n+1)
1
F
+ 2
F (n+1)
, (3.186)
4
(n+)
2
(n+1)
2
(n+) , F (n+1)
+
F
, (3.187)
2 2
and
K 2 d
(n+)
2
2
(n+) , u(n+) F (n+1)
1
+ 2
F (n+1)
. (3.188)
4
69
Write
(n) (n+1)
e , F (n) = F (n) , F (n+1) + F (n) , u(n) F (n+1)
+ (n) , F (n+1) + (n) , u(n) F (n+1) , (3.189)
where
(n)
2
(n+1)
2
F (n) , F (n+1)
F
+
F
, (3.190)
2 2
K 2 d
(n)
2
2
1
F (n) , u(n) F (n+1) F + 2
F (n+1)
, (3.191)
4
(n)
2
(n+1)
2
(n) , F (n+1)
+
F
, (3.192)
2 2
and
K 2 d
(n)
2
2
1
(n) , u(n) F (n+1) + 2
F (n+1)
. (3.193)
4
with
12
dK
(n+1)
2
2
u(n+) F (n+) , F (n+1)
F
+
F (n+)
, (3.195)
4
dK 12
2
u(n+) F (n+) , u(n+) F (n+1)
F (n+1)
2
12
dK
2
+
F (n+)
, (3.196)
2
1 2 h2
dK
(n+1)
2
2
u(n+) (n+) , F (n+1)
F
+ h2
(n+)
, (3.197)
4
70
and
2 dK 12
2
u(n+) (n+) , u(n+) F (n+1)
(n+)
2
12
dK
2
(n+1)
+
F
. (3.198)
2
Similarly
(n) (n+1)
u(n) e , F (n) = u(n) F (n) , F (n+1)
+ u(n) F (n) , u(n) F (n+1) + u(n) (n) , F (n+1)
+ u(n) (n) , u(n) F (n+1) , (3.199)
with
12
dK
(n+1)
2
2
u(n) F (n) , F (n+1)
F
+
F (n)
, (3.200)
4
u(n) F (n) , u(n) F (n+1)
12
dK 12
2 dK
2
F (n+1)
+
F (n)
, (3.201)
2 2
1 2 h2
dK
(n+1)
2
2
u(n) (n) , F (n+1)
F
+ h2
(n)
, (3.202)
4
and
12
dK 12
2 2 dK
2
(n) (n) (n) (n+1) (n+1)
(n)
u , u F
F
+
. (3.203)
2 2
71
and bound each terms as
ga (F (n+) , u(n+) ), F (n+1)
ga (F (n+) , u(n+) )
F (n+1)
4d
u(n+)
F (n+)
F (n+1)
4dK1
F (n+)
F (n+1)
2
2
4dK 1
F (n+)
+ dK
1
F (n+1)
, (3.205)
ga (F (n+) , u(n+) ), u(n+) F (n+1) (3.206)
ga (F (n+) , u(n+) )
u(n+) F (n+1)
2
4d2
u(n+)
F (n+)
F (n+1)
2 2
(n+)
(n+1)
4d K1
F
F
2
2
4d4 K1 4
F (n+)
+ 2
F (n+1)
, (3.207)
2
2
ga ((n+) , u(n+) ), F (n+1) 4dK
(n+)
+ dK
1
1
F (n+1)
, (3.208)
and
ga ((n+) , u(n+) ), u(n+) F (n+1)
2
2
4d4 K1 4
(n+)
+ 2
F (n+1)
. (3.209)
Similarly,
(n) (n+1)
ga (e , u(n) ), F (n) = ga (F (n) , u(n) ), F (n+1)
+ ga (F (n) , u(n) ), u(n) F (n+1) + ga ((n) , u(n) ), F (n+1)
+ ga ((n) , u(n) ), u(n) F (n+1) , (3.210)
(n)
2
(n+1)
2
ga (F (n)
, u (n)
), F (n+1)
4dK1
F
+ dK1
F
, (3.211)
2
2
ga (F (n) , u(n) ), u(n) F (n+1) 4d4 K1 4
F (n)
+ 2
F (n+1)
, (3.212)
72
2
2
1
ga ((n) , u(n) ), F (n+1) 4dK
1
(n)
+ dK
F (n+1)
,
(3.213)
and
2
2
ga ((n) , u(n) ), u(n) F (n+1) 4d4 K1 4
(n)
+ 2
F (n+1)
. (3.214)
The bounds for each term on the right hand side of (3.215) follow. Write
(n+)
dt e ,F (n+)
= dt F (n+) , F (n+)
+ dt (n+) , F (n+) , (3.216)
with
(n+)
(n+)
(n+)
2
(n)
2
dt F ,F
F
F
, (3.217)
2t
and
2
2
dt (n+) , F (n+)
F (n+)
+
dt (n+)
. (3.218)
4
73
Write
(n+)
= F (n+) , F (n+) + F (n+) , u(n) F (n+)
(n+)
e , F (n)
+ (n+) , F (n+) + (n+) , u(n) F (n+) , (3.219)
Next
(n+)
= (1 2) F (n+) , F (n+)
(n+)
(1 2) e ,F
(n+)
+ (1 2) F (n+) , u(n+) F (n+) + (1 2) (n+) , F (n+)
+ (1 2) (n+) , u(n+) F (n+) , (3.224)
(1 2) F (n+) , u(n+) F (n+)
(1 2) 2
(n+)
2
2
F
+ 2 (1 2) 1
u(n+) F (n+)
, (3.226)
41
(1 2)
(n+)
2 (1 2)
(n+)
2
(1 2) (n+) , F (n+)
+
F
, (3.227)
2 2
74
and
(1 2) (n+) , u(n+) F (n+)
(1 2) 2
(n+)
2
2
+ 2 (1 2) 1
u(n+) F (n+)
. (3.228)
41
Write
(n+)
= F (n) , F (n+) + F (n) , u(n) F (n+)
(n)
e , F (n)
+ (n) , F (n+) + (n) , u(n) F (n+) , (3.229)
Next
(n+) (n+)
(1 2) e , F (n+) = (1 2) F (n+) , F (n+)
+ (1 2) F (n+) , u(n+) F (n+) + (1 2) (n+) , F (n+)
+ (1 2) (n+) , u(n+) F (n+) , (3.234)
75
(1 2) F (n+) , u(n+) F (n+)
(1 2) 2
(n+)
2
2
F
+ 2 (1 2) 1
u(n+) F (n+)
, (3.236)
41
(1 2)
(n+)
2 (1 2)
(n+)
2
(1 2) (n+) , F (n+)
+
F
, (3.237)
2 2
and
(1 2) (n+) , u(n+) F (n+)
(1 2) 2
(n+)
2
2
+ 2 (1 2) 1
u(n+) F (n+)
. (3.238)
41
The term
(n+)
= F (n) , F (n+) + F (n) , u(n) F (n+)
(n)
e , F (n)
+ (n) , F (n+) + (n) , u(n) F (n+) , (3.239)
has bounds
(n)
2
(n+)
2
F (n) , F (n+) F + F
, (3.240)
2 2
K 2 d
(n)
2
2
F (n) , u(n) F (n+)
F
+ 2
F (n+)
,
1
(3.241)
4
(n)
2
(n+)
2
(n) , F (n+)
+
F
, (3.242)
2 2
and
K 2 d
(n)
2
2
(n) , u(n) F (n+) (n+)
1 2
+ F
. (3.243)
4
The term
(n+)
= F (n) , F (n+) + F (n) , u(n) F (n+)
(n)
e , F (n)
+ (n) , F (n+) + (n) , u(n) F (n+) , (3.244)
has bounds
(n)
2
(n+)
2
F (n) , F (n+)
F
+
F
, (3.245)
2 2
76
K 2 d
(n)
2
2
F (n) , u(n) F (n+)
+ 2
F (n+)
,
1
F (3.246)
4
(n)
2
(n+)
2
(n) , F (n+)
+
F
, (3.247)
2 2
and
K 2 d
(n)
2
2
(n)
, u (n)
F (n+)
1
2
+
F
(n+)
. (3.248)
4
Z
( u)q dA = 0, q L20 ().
Thus,
Z
u(n+) F (n+) , F (n+) = u(n+)F (n+) : F (n+) dA
Z Z
= u (n+)
(v v0 ) dA + u(n+) v0 dA
Z Z
= v0 u (n+)
n ds v0 u (n+)
1 dA
= 0 (3.250)
77
Using (3.250) the first term on the right hand side of (3.249) yields
(1 2) u(n+) F (n+) , F (n+) = (1 2) u(n+) F (n+) , F (n+)
(1 2) u(n+) F (n+) , F (n+)
= (1 2) u(n+) F (n+) , F (n+) .
(1 2) u(n+) F (n+) , u(n+) F (n+)
2
= (1 2)
u(n+) F (n+)
, (3.252)
(1 2) u(n+) (n+) , F (n+)
2
(1 2) K1 2 dh
(n+)
2
2
F
+ (1 2) h2
(n+)
, (3.253)
4
and
(1 2) u(n+) (n+) , u(n+) F (n+)
12
(1 2) dK
2 2 (1 2) K 2 d
2
F
(n+)
+
1
(n+)
. (3.254)
2 2
78
K1 2 d
2
u(n) F (n) , u(n) F (n+)
F (n+)
2
K1 2 d
2
+
F (n)
, (3.257)
2
2
K 2 dh
(n+)
2
2
u(n) (n) , F (n+)
1 2
(n)
F + h
, (3.258)
4
and
u(n) (n) , u(n) F (n+)
K1 2 d
2 2 K 2 d
2
F (n+)
+
1
(n)
. (3.259)
2 2
Next
(n+)
= u(n) F (n) , F (n+)
(n) (n)
u e , F (n)
+ u(n) F (n) , u(n) F (n+) + u(n) (n) , F (n+)
+ u(n) (n) , u(n) F (n+) , (3.260)
with bounds
K 2 d
(n+)
2
2
u(n) F (n) , F (n+)
1
F
+
F (n)
, (3.261)
4
K1 2 d
2
u(n) F (n) , u(n) F (n+) F (n+)
2
K1 2 d
2
+
F (n)
, (3.262)
2
2
K 2 dh
(n+)
2
2
u(n) (n) , F (n+)
1
F
+ h2
(n)
, (3.263)
4
and
K1 2 d
2
u(n) (n) , u(n) F (n+) F (n+)
2
2 K1 2 d
2
+
(n)
. (3.264)
2
79
For the ga (, ) terms write
(n+) (n+)
(1 2) ga (e , u (n+)
), F (n+) = (1 2) ga (F (n+) , u(n+) ), F (n+)
+ (1 2) ga (F (n+) , u(n+) ), u(n+) F (n+)
+ (1 2) ga ((n+) , u(n+) ), F (n+)
+ (1 2) ga ((n+) , u(n+) ), u(n+) F (n+) , (3.265)
(1 2) ga (F (n+) , u(n+) ), u(n+) F (n+)
2
2
4 (1 2) d4 K1 4
F (n+)
+ 2 (1 2)
F (n+)
, (3.267)
(1 2) ga ((n+) , u(n+) ), F (n+)
2
2
4 (1 2) K1 2 d2
(n+)
+ (1 2)
F (n+)
, (3.268)
and
(1 2) ga ((n+) , u(n+) ), u(n+) F (n+)
2
2
4 (1 2) d4 K1 4
(n+)
+ 2 (1 2)
F (n+)
. (3.269)
Next,
(n+)
= ga (F (n) , u(n) ), F (n+)
(n)
ga (e , u(n) ), F (n)
+ ga (F (n) , u(n) ), u(n) F (n+) + ga ((n) , u(n) ), F (n+)
+ ga ((n) , u(n) ), u(n) F (n+) , (3.270)
80
which is bound using
2
2
ga (F (n) , u(n) ), F (n+) 4K1 2 d2
F (n+)
+
F (n)
, (3.271)
2
2
ga (F (n) , u(n) ), u(n) F (n+) 4K1 4 d4
F (n)
+ 2
F (n+)
, (3.272)
2
(n)
2
(n)
ga ( , u ), F (n) (n+) 2 2
(n+)
4K1 d
F
+
, (3.273)
and
2
2
ga ((n) , u(n) ), u(n) F (n+) 4K1 4 d4
(n)
+ 2
F (n+)
. (3.274)
Lastly write
(n+)
= ga (F (n) , u(n) ), F (n+)
(n) (n)
ga (e , u ), F (n)
+ ga (F (n)
, u (n)
), u (n)
F (n+)
+ ga ( (n)
, u (n)
), F (n+)
+ ga ((n) , u(n) ), u(n) F (n+) , (3.275)
where
2
(n)
2
ga (F (n)
, u (n)
), F (n+) 2 2
(n+)
4K1 d
F
+
F
, (3.276)
2
ga (F (n)
, u (n)
), u (n)
F (n+) 4 4
4K1 d
F (n)
2
+ 2
F (n+)
, (3.277)
2
2
ga ((n) , u(n) ), F (n+) 4K1 2 d2
F (n+)
+
(n)
, (3.278)
and
2
ga ((n) , u(n) ), u(n) F (n+) 4K1 4 d4
(n)
2
2
+
F
(n+)
. (3.279)
81
Next examine B6n (, ) B6n ( h , ) in (3.164) with = F (n+) . Thus,
(n+) (n+) (n+)
B6n (, F (n+) ) B6n ( h , F (n+) ) := dt e , F (n+) + e , F (n)
(n) (n+) (n+1) (n+)
+ e , F (n) + (1 2) e , F (n)
(n) (n+) (n) (n+)
+ e , F (n) + e , F (n)
(n) (n+) (n) (n+)
+ u(n) e , F (n) + u(n) e , F (n)
(n) (n+) (n) (n+)
+ ga (e , u(n) ), F (n) + ga (e , u(n) ), F (n) . (3.280)
and
2
2
dt (n+) , F (n+)
dt (n+)
+
F (n+)
. (3.283)
4
Next,
(n+) (n+)
e , F (n) = F (n+) , F (n+) + F (n+) , u(n) F (n+)
+ (n+) , F (n+) + (n+) , u(n) F (n+) . (3.284)
K 2 d
(n+)
2
2
1
F (n+) , u(n) F (n+)
F
+ 2
F (n+)
, (3.286)
4
(n+)
2
(n+)
2
(n+) (n+)
,F
+
F
, (3.287)
2 2
82
and
K 2 d
(n+)
2
2
1
(n+) , u(n) F (n+)
+ 2
F (n+)
. (3.288)
4
The term
(n) (n+)
e , F (n) = F (n) , F (n+) + F (n) , u(n) F (n+)
+ (n) , F (n+) + (n) , u(n) F (n+) , (3.289)
has bounds
(n)
2
(n+)
2
F (n) , F (n+)
F
+
F
, (3.290)
2 2
K 2 d
(n)
2
2
1
F (n) , u(n) F (n+)
F
+ 2
F (n+)
, (3.291)
4
(n)
2
(n+)
2
(n) , F (n+) + F
, (3.292)
2 2
and
K 2 d
(n)
2
2
1
(n) , u(n) F (n+)
+ 2
F (n+)
. (3.293)
4
Next
(n+1) (n+)
(1 2) e , F (n) = (1 2) F (n+1) , F (n+)
+ (1 2) F (n+1) , u(n) F (n+) + (1 2) (n+1) , F (n+)
+ (1 2) (n+1) , u(n) F (n+) , (3.294)
with bounds
(n+1) (n+)
(n+1)
2 (1 2)
(n+)
2
(1 2)
(1 2) F ,F
F
+
F
, (3.295)
2 2
(1 2) F (n+1) , u(n) F (n+)
(1 2) K1 2 d
(n+1)
2
2
F
+ 2 (1 2)
F (n+)
, (3.296)
4
83
(1 2)
(n+1)
2 (1 2)
(n+)
2
(1 2) (n+1) , F (n+)
+
F
, (3.297)
2 2
and
(1 2) (n+1) , u(n) F (n+)
(1 2) K1 2 d
(n+1)
2
2
2
(n+)
+ (1 2) F
. (3.298)
4
Also
(n) (n+)
e , F (n) = F (n) , F (n+) + F (n) , u(n) F (n+)
+ (n) , F (n+) + (n) , u(n) F (n+) , (3.299)
with
(n)
2
(n+)
2
F (n) , F (n+)
F
+
F
, (3.300)
2 2
K 2 d
(n)
2
2
1
F (n) , u(n) F (n+)
F
+ 2
F (n+)
, (3.301)
4
(n)
2
(n+)
2
(n) , F (n+)
+
F
, (3.302)
2 2
and
K 2 d
(n)
2
2
1
(n) , u(n) F (n+)
+ 2
F (n+)
. (3.303)
4
Write
(n) (n+)
e , F (n) = F (n) , F (n+) + F (n) , u(n) F (n+)
+ (n) , F (n+) + (n) , u(n) F (n+) . (3.304)
84
and
K 2 d
(n)
2
2
1
(n) , u(n) F (n+)
+ 2
F (n+)
. (3.308)
4
(n) (n+)
u(n) e , F (n) = u(n) F (n) , F (n+)
+ u(n) F (n) , u(n) F (n+) + u(n) (n) , F (n+)
+ u(n) (n) , u(n) F (n+) , (3.309)
where
12
dK
(n+)
2
2
u(n) F (n) , F (n+)
F
+
F (n)
, (3.310)
4
dK 12
2
u(n) F (n) , u(n) F (n+)
F (n)
2
12
dK
2
(n+)
+ F
, (3.311)
2
1 2 h2
dK
(n+)
2
2
(n) (n) (n+) 2
(n)
u ,F
F
+ h
, (3.312)
4
and
12
dK
2
u(n) (n) , u(n) F (n+)
F (n+)
2
12
2 dK
2
+
(n)
. (3.313)
2
(n) (n+)
u(n) e , F (n) = u(n) F (n) , F (n+)
+ u(n) F (n) , u(n) F (n+) + u(n) (n) , F (n+)
+ u(n) (n) , u(n) F (n+) , (3.314)
85
where
12
dK
(n+)
2
2
u(n) F (n) , F (n+)
F
+
F (n)
, (3.315)
4
dK
12
12
2 dK
2
(n) (n) (n) (n+) (n)
(n+)
u F , u F
F
+
F
,
2 2
(3.316)
1 2 h2
dK
2
2
u(n) (n) , F (n+)
(n+)
F
+ h2
(n)
, (3.317)
4
and
12
dK 12
2 2 dK
2
u(n) (n) , u(n) F (n+)
F (n+)
+
(n)
. (3.318)
2 2
(n) (n+)
ga (e , u(n) ), F (n) = ga (F (n) , u(n) ), F (n+)
+ ga (F (n) , u(n) ), u(n) F (n+) + ga ((n) , u(n) ), F (n+)
+ ga ((n) , u(n) ), u(n) F (n+) , (3.319)
where
2
2
ga (F (n) , u(n) ), F (n+) 4K1 2 d2
F (n)
+
F (n+)
, (3.320)
2
ga (F (n) , u(n) ), u(n) F (n+) 4K1 4 d4
F (n)
2
+ 2
F (n+)
, (3.321)
2
2
ga ((n) , u(n) ), F (n+) 4K1 2 d2
(n)
+
F (n+)
, (3.322)
and
2
ga ((n) , u(n) ), u(n) F (n+) 4K1 4 d4
(n)
2
+ 2
F (n+)
. (3.323)
86
Also
(n) (n+)
ga (e , u(n) ), F (n) = ga (F (n) , u(n) ), F (n+)
+ ga (F (n) , u(n) ), u(n) F (n+)
+ ga ((n) , u(n) ), F (n+) + ga ((n) , u(n) ), u(n) F (n+) , (3.324)
and is bounded as
2
2
ga (F (n) , u(n) ), F (n+) 4K1 2 d2
F (n)
+
F (n+)
, (3.325)
2
2
ga (F (n) , u(n) ), u(n) F (n+) 4K1 4 d4
F (n+)
+ 2
F (n+)
,
(3.326)
2
2
ga ((n) , u(n) ), F (n+) 4K1 2 d2
F (n+)
+
(n)
, (3.327)
and
2
2
ga ((n) , u(n) ), u(n) F (n+) 4K1 4 d4
(n)
+ 2
F (n+)
. (3.328)
The terms in I4n (F (n+1) ) from (3.149) are bounded by separating the standard
with
(n+1) (n+1) (n+ 12 ) (n+1)
Int 1 G (F ) = dt t ,F
1
+ (n+) + (n+1) (n+ 2 ) , F (n+1)
1
+ (n) + (n+) (n+ 2 ) , F (n+1)
1 1
+ u(n) (n) + u(n+) (n+) u(n+ 2 ) (n+ 2 ) , F (n+1)
1 1
+ ga ( (n) , u(n) ) + ga ( (n+) , u(n+) ) ga ( (n+ 2 ) , u(n+ 2 ) ), F (n+1)
1
+ 2 d(u(n+ 2 ) ) d(u(n) ) d(u(n+) ), F (n+1) ,
87
and
(n+1) (n+ 21 ) (n) (n+1)
Int 1 up (F ) = t , u F
(n+ 1 )
+ t 2 , u(n+) F (n+1)
(n+) (n+ 12 ) (n) (n+1)
+ , u F
1
+ (n+1) (n+ 2 ) , u(n+) F (n+1)
1
+ (1 2) (n+) (n+ 2 ) , u(n+) F (n+1)
1
+ (n) (n+ 2 ) , u(n) F (n+1)
1
+ (n+) (n+ 2 ) , u(n+) F (n+1)
1 1
+ u(n) (n) u(n+ 2 ) (n+ 2 ) , u(n) F (n+1)
1 1
+ u(n+) (n+) u(n+ 2 ) (n+ 2 ) , u(n+) F (n+1)
1 1
+ ga ( (n) , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) ), u(n) F (n+1)
1 1
+ ga ( (n+) , u(n+) ) ga ( (n+ 2 ) , u(n+ 2 ) ), u(n+) F (n+1)
1
+2 d(u(n+ 2 ) ) d(u(n) ), u(n) F (n+1)
1
+2 d(u(n+ 2 ) ) d(u(n+) ), u(n+) F (n+1) .
1
1
2
(n+) + (n+1) (n+ 2 ) , F (n+1) + (n+1) (n+ 2 )
(n+)
4
2
+
F (n+1)
, (3.330)
1
(n) + (n+) (n+ 2 ) , F (n+1)
1
2
2
(n) + (n+) (n+ 2 )
+
F (n+1)
, (3.331)
4
88
1 1
u(n) (n) + u(n+) (n+) u(n+ 2 ) (n+ 2 ) , F (n+1)
1 1
2
2
u(n) (n) + u(n+) (n+) u(n+ 2 ) (n+ 2 )
+
F (n+1)
, (3.332)
4
1 1
ga ( (n) , u(n) ) + ga ( (n+) , u(n+) ) ga ( (n+ 2 ) , u(n+ 2 ) ), F (n+1)
(n) (n) (n+) (n+) (n+ 21 ) (
2
n+ 12 )
ga ( , u ) + ga ( , u ) ga ( , u )
4
2
+
F (n+1)
, (3.333)
and
1
2 d(u(n+ 2 ) ) d(u(n) ) d(u(n+) ), F (n+1)
1
2
2
d(u(n+ 2 ) ) d(u(n) ) d(u(n+) )
+ 2
F (n+1)
. (3.334)
2
2
2 K1 2 d
2
(n+ 21 ) (n+ 12 )
t , u(n+) F (n+1)
t
+
F (n+1)
, (3.336)
4
1
2 K1 2 d
2
n+ 21 )
(n+)
(n+ 2 ) , u(n) F (n+1) (
(n+)
4
2
+
F (n+1)
, (3.337)
1
2 K1 2 d
1
2
(n+1) (n+ 2 ) , u(n+) F (n+1) (n+ 2 )
(n+1)
4
2
(n+1)
+
F
, (3.338)
1
(1 2) (n+) (n+ 2 ) , u(n+) F (n+1)
(1 2) 2 K1 2 d
1
2
2
(n+ 2 )
+ (1 2)
F (n+1)
, (3.339)
(n+)
4
89
1
2 K1 2 d
1
2
(n) (n+ 2 ) , u(n) F (n+1)
(n+ 2 )
(n)
4
2
+
F (n+1)
, (3.340)
1
2 K1 2 d
1
2
(n+) (n+ 2 ) , u(n+) F (n+1)
(n+)
(n+ 2 )
4
2
+
F (n+1)
, (3.341)
1 1
u(n) (n) u(n+ 2 ) (n+ 2 ) , u(n) F (n+1)
2 K1 2 d
1 1
2
2
u (n) u(n+ 2 ) (n+ 2 )
+
F (n+1)
, (3.342)
(n)
4
1 1
u(n+) (n+) u(n+ 2 ) (n+ 2 ) , u(n+) F (n+1)
2 K1 2 d
(n+) (n+) (n+ 12 ) (
2
n+ 21 )
2
(n+1)
u u + F
, (3.343)
4
1 1
ga ( (n) , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) ), u(n) F (n+1)
2 K1 2 d
1 1
2
2
ga ( (n) , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) )
+
F (n+1)
, (3.344)
4
1 1
ga ( (n+) , u(n+) ) ga ( (n+ 2 ) , u(n+ 2 ) ), u(n+) F (n+1)
2 K1 2 d
1 1
2
2
ga ( (n+) , u(n+) ) ga ( (n+ 2 ) , u(n+ 2 ) )
+
F (n+1)
, (3.345)
4
1
2 d(u(n+ 2 ) ) d(u(n) ), u(n) F (n+1)
2 K1 2 d
1
2
2
d(u(n+ 2 ) ) d(u(n) )
+ 2
F (n+1)
, (3.346)
2
1
2 d(u(n+ 2 ) ) d(u(n+) ), u(n+) F (n+1)
2 K1 2 d
1
2
2
d(u(n+ 2 ) ) d(u(n+) )
+ 2
F (n+1)
. (3.347)
2
90
Next decompose I5n (F (n+) ) from (3.156) into its standard Galerkin and upwinded compo-
nents as
with
1
(n+ ) (n+)
I5n G (F (n+) ) = dt (n+) t 2 ,F
1
+ (n+) + (1 2) (n+) + (n) (n+ 2 ) , F (n+)
1
+ (1 2) (n+) + (n) + (n) (n+ 2 ) , F (n+)
+ (1 2) u(n+) (n+) + u(n) (n) + u(n) (n)
(n+ 12 ) ( n+ 12 ) (n+)
u ,F
+ (1 2) ga ( (n+) , u(n+) ) + ga ( (n) , u(n) )
(n) (n) (n+ 12 ) (n+ 12 ) (n+)
+ ga ( , u ) ga ( , u ), F
1
+ 2 d(u(n+ 2 ) ) d(u(n) ) (1 2) d(u(n+) )
d(u (n)
), F (n+)
,
91
and
(n+ 21 )
I5n up (F (n+) )
= t , u (n)
F (n+)
(n+ 21 ) ( n+ 12 ) (n+)
+ (1 2) t , u F
(n+ 1 )
+ t 2 , u(n) F (n+)
1
+ (n+) (n+ 2 ) , u(n) F (n+)
1
+ (1 2) (n+) (n+ 2 ) , u(n+) F (n+)
(n) ( n+ 12 ) (n) ( n+)
+ , u F
1
+ (1 2) (n+) (n+ 2 ) , u(n+) F (n+)
1
+ (n) (n+ 2 ) , u(n) F (n+)
1
+ (n) (n+ 2 ) , u(n) F (n+)
(n+) (n+) (n+ 12 ) (n+ 12 ) (n+) (n+)
+ (1 2) u u , u F
1 1
+ u(n) (n) u(n+ 2 ) (n+ 2 ) , u(n) F (n+)
1 1
+ u(n) (n) u(n+ 2 ) (n+ 2 ) , u(n) F (n+)
+ (1 2) ga ( (n+) , u(n+) )
( n+ 12 ) (n+ 12 ) (n+) (n+)
ga ( , u ), u F
1 1
+ ga ( (n) , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) ), u(n) F (n+)
1 1
+ ga ( (n) , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) ), u(n) F (n+)
1
+ 2 d(u(n+ 2 ) ) d(u(n) ), u(n) F (n+)
1
+ (1 2) 2 d(u(n+ 2 ) ) d(u(n+) ), u(n+) F (n+)
1
+ 2 d(u(n+ 2 ) ) d(u(n) ), u(n) F (n+) .
92
1
(n+) + (1 2) (n+) + (n) (n+ 2 ) , F (n+)
1
2
2
(n+) + (1 2) (n+) + (n) (n+ 2 )
+
F (n+)
, (3.350)
4
1
(1 2) (n+) + (n) + (n) (n+ 2 ) , F (n+)
(n+) (n) (n) (
2
n+ 12 )
(n+)
2
(1 2) + +
+
F
, (3.351)
4
(1 2) u(n+) (n+) + u(n) (n)
1 1
+ u(n) (n) u(n+ 2 ) (n+ 2 ) , F (n+)
(1 2) u(n+) (n+) + u(n) (n)
4
2
(n) (n) (n+ 12 ) (n+ 12 )
+ u u
2
+
F (n+)
, (3.352)
(1 2) ga ( (n+) , u(n+) ) + ga ( (n) , u(n) )
1 1
+ ga ( (n)
, u (n)
) ga ( (n+ 2 ) , u(n+ 2 ) ), F (n+)
(1 2) ga ( (n+) , u(n+) ) + ga ( (n) , u(n) )
4
2
(n) (n) (n+ 12 ) (n+ 12 )
+ ga ( , u ) ga ( , u )
2
+
F (n+)
, (3.353)
and
1
2 d(u(n+ 2 ) ) d(u(n) ) (1 2) d(u(n+) ) d(u(n) ), F (n+)
1
2
d(u(n+ 2 ) ) d(u(n) ) (1 2) d(u(n+) ) d(u(n) )
2
2
+ 2
F (n+)
. (3.354)
93
Bounding the terms in I5n up (F (n+) ) gives
2
2 K1 2 d
2
(n+ 21 ) (n+ 21 )
t , u (n)
F (n+)
t
+
F (n+)
, (3.355)
4
(n+ 21 ) 1
(1 2) t , u(n+ 2 ) F (n+)
2
(1 2) 2 K1 2 d
2
(n+ 12 )
+ (1 2)
F (n+)
, (3.356)
t
4
2
2 K1 2 d
(n+ 21 ) 1
(n+ 2 )
(n)
F (n+)
t , u t
4
2
+
F (n+)
, (3.357)
1
(n+) (n+ 2 ) , u(n) F (n+)
2 K1 2 d
(n+) (
2
n+ 12 )
(
2
n+)
+ F
, (3.358)
4
1
(1 2) (n+) (n+ 2 ) , u(n+) F (n+)
2 (1 2)2 2 dK
12
1
2
2
(n+ 2 )
+
F (n+)
, (3.359)
(n+)
4
1
(n) (n+ 2 ) , u(n) F (n+)
2 K1 2 d
1
2
2
(n+ 2 )
+
F (n+)
, (3.360)
(n)
4
1
(1 2) (n+) (n+ 2 ) , u(n+) F (n+)
2 (1 2)2 2 dK
12
(n+) (
2
n+ 12 )
(
2
n+)
+
F
, (3.361)
4
1
(n) (n+ 2 ) , u(n) F (n+)
2 K1 2 d
1
2
2
(n+ 2 )
+
F (n+)
, (3.362)
(n)
4
94
1
(n) (n+ 2 ) , u(n) F (n+)
2 K1 2 d
1
2
2
(n+ 2 )
+
F (n+)
, (3.363)
(n)
4
1 1
(1 2) u(n+) (n+) u(n+ 2 ) (n+ 2 ) , u(n+) F (n+)
2 (1 2)2 2 dK
12
(n+) 1 1
2
u (n+) u(n+ 2 ) (n+ 2 )
4
2
+
F (n+)
, (3.364)
1 1
u(n) (n) u(n+ 2 ) (n+ 2 ) , u(n) F (n+)
2 K1 2 d
1 1
2
2
u (n) u(n+ 2 ) (n+ 2 )
+
F (n+)
, (3.365)
(n)
4
1 1
u(n) (n) u(n+ 2 ) (n+ 2 ) , u(n) F (n+)
2 K1 2 d
1 1
2
2
(n) u(n+ 2 ) (n+ 2 )
+
F (n+)
, (3.366)
(n)
u
4
1 1
(1 2) ga ( (n+) , u(n+) ) ga ( (n+ 2 ) , u(n+ 2 ) ), u(n+) F (n+)
2 (1 2)2 2 dK
12
(n+) (n+) (n+ 12 ) (
2
n+ 12 )
g ( , u ) g ( , u )
a a
4
2
+
F (n+)
, (3.367)
1 1
ga ( (n) , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) ), u(n) F (n+)
2 K1 2 d
(n) (n) (n+ 12 ) (
2
n+ 12 )
(
2
n+)
g ( , u ) g ( , u ) + F
, (3.368)
a a
4
1 1
ga ( (n) , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) ), u(n) F (n+)
2 K1 2 d
1 1
2
ga ( (n) , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) )
4
2
+
F
(n+)
, (3.369)
95
1
2 d(u(n+ 2 ) ) d(u(n) ), u(n) F (n+)
2 K1 2 d
1
2
2
d(u(n+ 2 ) ) d(u(n) )
+ 2
F (n+)
, (3.370)
2
1
(1 2) 2 d(u(n+ 2 ) ) d(u(n+) ), u(n+) F (n+)
1
2
2
2 (1 2)2 2 dK
d(u(n+ 2 ) ) d(u(n+) )
+
F (n+)
, (3.371)
12
and
1
2 d(u(n+ 2 ) ) d(u(n) ), u(n) F (n+)
2 2 K1 2 d
(n+ 12 )
2
(n)
(
2
n+)
d(u ) d(u )
+ 2
F
. (3.372)
4
where
(n+ 21 ) (n+)
I6n G (F (n+) ) = dt (n+) t ,F
1
+ (n) + (n) (n+ 2 ) , F (n+)
1
+ (n+) + (n) + (1 2) (n+1) (n+ 2 ) , F (n+)
1 1
+ u(n) (n) + u(n) (n) u(n+ 2 ) (n+ 2 ) , F (n+)
+ ga ( (n) , u(n) ) + ga ( n , u(n) )
1 1
ga ( (n+ 2 ) , u(n+ 2 ) ), F (n+)
1
+ 2 d(u(n+ 2 ) ) d(u(n) ) d(u(n) ), F (n+)
96
and
(n+ 21 )
I6n up (F (n+) )= t , u F (n) (n+)
(n+ 21 ) (n) (n+)
+ t , u F
(n+) ( n+ 12 ) (n) (n+)
+ , u F
1
+ (n) (n+ 2 ) , u(n) F (n+)
1
+ (1 2) (n+1) (n+ 2 ) , u(n) F (n+)
1
+ (n) (n+ 2 ) , u(n) F (n+)
1
+ (n) (n+ 2 ) , u(n) F (n+)
1 1
+ u(n) (n) u(n+ 2 ) (n+ 2 ) , u(n) F (n+)
1 1
+ u(n) (n) u(n+ 2 ) (n+ 2 ) , u(n) F (n+)
1 1
+ ga ( (n) , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) ), u(n) F (n+)
1 1
+ ga ( n , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) ), u(n) F (n+)
1
+ 2 d(u(n+ 2 ) ) d(u(n) ), u(n) F (n+)
1
+ 2 d(u(n+ 2 ) ) d(u(n) ), u(n) F (n+) .
1
(n) + (n) (n+ 2 ) , F (n+)
1
2
2
(n) + (n) (n+ 2 )
+
F (n+)
, (3.375)
4
1
(n+) + (n) + (1 2) (n+1) (n+ 2 ) , F (n+)
(n+) (n) (n+1) (
2
n+ 12 )
(n+)
2
+ + (1 2)
+
F
, (3.376)
4
97
1 1
u(n) (n) + u(n) (n) u(n+ 2 ) (n+ 2 ) , F (n+)
1 1
2
u(n) (n) + u(n) (n) u(n+ 2 ) (n+ 2 )
4
2
+
F (n+)
, (3.377)
1 1
ga ( (n)
, u (n)
) + ga ( , u n (n)
) ga ( (n+ 2 ) , u(n+ 2 ) ), F (n+)
1 1
2
ga ( (n) , u(n) ) + ga ( n , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) )
4
2
+
F (n+)
, (3.378)
1
2 d(u(n+ 2 ) ) d(u(n) ) d(u(n) ), F (n+)
(n+ 21 ) (n)
2
(n)
(n+)
2
d(u ) d(u ) d(u )
+ 2
F
. (3.379)
2
1
2
2 K1 2 d
2
(n+ 21 )
(n+ 2 )
+
t , u(n) F (n+) t F (n+)
,
4
(3.381)
1
(n+) (n+ 2 ) , u(n) F (n+)
2 K1 2 d
1
2
2
(n+ 2 )
+
F (n+)
, (3.382)
(n+)
4
1
(n) (n+ 2 ) , u(n) F (n+)
2 K1 2 d
(n) (
2
n+ 12 )
2
(n+)
+
F
, (3.383)
4
1
(1 2) (n+1) (n+ 2 ) , u(n) F (n+)
(1 2) 2 K1 2 d
1
2
2
(n+ 2 )
+ (1 2)
F (n+)
, (3.384)
(n+1)
4
98
1
(n) (n+ 2 ) , u(n) F (n+)
2 K1 2 d
1
2
2
(n+ 2 )
+
F (n+)
, (3.385)
(n)
4
1
(n) (n+ 2 ) , u(n) F (n+)
2 K1 2 d
1
2
2
(n+ 2 )
+
F (n+)
, (3.386)
(n)
4
1 1
u(n) (n) u(n+ 2 ) (n+ 2 ) , u(n) F (n+)
2 K1 2 d
1 1
2
2
(n) u(n+ 2 ) (n+ 2 )
+
F (n+)
, (3.387)
(n)
u
4
1 1
u(n) (n) u(n+ 2 ) (n+ 2 ) , u(n) F (n+)
2 K1 2 d
(n) (n) (n+ 21 ) (
2
n+ 21 )
2
(n+)
u u
+
F
, (3.388)
4
1 1
ga ( (n) , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) ), u(n) F (n+)
2 K1 2 d
1 1
2
ga ( (n) , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) )
4
2
+
F (n+)
, (3.389)
1 1
ga ( n , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) ), u(n) F (n+)
2 K1 2 d
1 1
2
2
ga ( n , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) )
+
F (n+)
, (3.390)
4
1
2 d(u(n+ 2 ) ) d(u(n) ), u(n) F (n+)
2 K1 2 d
(n+ 21 )
2
(n)
2
(n+)
d(u ) d(u ) + 2 F
, (3.391)
2
99
1
2 d(u(n+ 2 ) ) d(u(n) ), u(n) F (n+)
2 K1 2 d
1
2
2
d(u(n+ 2 ) ) d(u(n) )
+ 2
F (n+)
. (3.392)
2
Bringing all the bounds on (3.150), (3.157), and (3.164) together to form the following three
expressions:
(n+1)
2
(n)
2
2
F
F
+
F (n+1)
2t
!
2
(n)
2
K 1 d
F
+ 4K1 d + 4K1 4 d4 +
2 4
2
+
F (n+1)
1 + 4 + 2 + + + (1 2) +
2
2 2 2 2
K1 d dK1 dK1 h
+ + + + 2dK1
4 4 4
!
(n+)
2 K1 d (1 2) (1 2) K1 d
2 2
+
F + + +
2 4 2 4
!
(n+)
2 K1 d
2
+
F + 1 + 4K1 d
+ 4dK 4 4
2 4
!!
2 12
dK
2 12
dK
(n)
+
F
1 +
+
F
(n+)
+
2 2
!
2 dK 12
1 2 dK
+
F (n+1)
1 + 3 + 2 + + + 2 2 + 2 2 + 2 2
2 2
+ H1 + HG1 + 2 H upA1 + 2 H upB1 , (3.393)
100
(n+)
2
(n)
2
F
F
2t
2
2
+ (1 2)
F (n+)
+ (1 2) ( 41 )
u(n+) F (n+)
!
2
2 K 1 d
F (n)
+ + 4K1 4 d4 +
2 4
!
(n+)
2 K1 d (1 2) (1 2) 2
2
+
F + + +
2 4 2 41
!
2 K1 2 d
+
F (n)
+ + + 4K1 4 d4
2 4
(n+)
2
+
F
1 + 4 + + + (1 2) + (1 2) + 2
(1 2) K1 2 d K1 2 dh 2 (1 2) K1 2 d
+ + + +
2 2 2 4
(1 2) 2
4 4
+4 (1 2) K1 d + 4 (1 2) K1 d + 16K1 d +2 2
41
!
2
2 K1 d
+
F (n)
+
2
!!
2 12
dK
(n)
+
F
1 +
2
5K1 2 d
2
+
F (n+)
5 + + 2 + 5 + 4 +
4
1
(1 2) dK 2
2 2
+ + K1 d + 2 ( + 2)
2
+ H2 + HG2 + 2 H upA2 + 2 H upB2 , (3.394)
101
and
(n+)
2
(n+1)
2
2
F
F
+
F (n+)
2t
!
(n)
2 2 2 K1 2 d K1 2 d
F
+ 4K1 d + +
2 4 4
!!
2 (1 2) K1 2 d
+
F (n+1)
1+
2 2
2 K1 2 d
+
F (n+)
1 + 5 + + 2 + (1 2) + + + +
2 4
!
1 2 dK
dK 1 2 h2
+ + + + 4K1 4 d4 + 4K1 2 d2
2 4 4
!
2 K1 2 d 2 2 4 4
+
F (n)
+ + + 4K1 d + 4K1 d
2 4
! !!
2
(n)
K1 2 d
2
(n)
12
dK
+
F
+ +
F
1 +
2 2
2 12
dK
+
F (n+)
1 + 3 + 2 + (1 + )
2
2
+ + 2 + 2 + (1 2) + 2
where terms Hi , HGi , H upAi , and HupBi for i {1, 2, 3} in (3.393)(3.395) are defined by
2
(n+1) (n+ 21 )
HG1 :=
dt t
4
1
2
+
(n+) + (n+1) (n+ 2 )
4
1
2
+
(n) + (n+) (n+ 2 )
4
1 1
2
+
u(n) (n) + u(n+) (n+) u(n+ 2 ) (n+ 2 )
4
1 1
2
+
ga ( (n) , u(n) ) + ga ( (n+) , u(n+) ) ga ( (n+ 2 ) , u(n+ 2 ) )
4
1
2
+
d(u(n+ 2 ) ) d(u(n) ) d(u(n+) )
, (3.396)
2
102
2
(n+) (n+ 12 )
HG2 :=
dt t
4
1
2
+
(n+) + (1 2) (n+) + (n) (n+ 2 )
4
1
2
+
(1 2) (n+) + (n) + (n) (n+ 2 )
4
+
(1 2) u(n+) (n+) + u(n) (n)
4
2
(n) (n) (n+ 12 ) (n+ 12 )
+u u
+
(1 2) ga ( (n+) , u(n+) ) + ga ( (n) , u(n) )
4
2
(n) (n) (n+ 21 ) (n+ 12 )
+ ga ( , u ) ga ( , u )
(n+ 21 ) (n) (n+)
2
(n)
+
d(u ) d(u ) (1 2) d(u ) d(u )
, (3.397)
2
2
(n+) (n+ 12 )
HG3 :=
dt t
4
1
2
+
(n+) + (n) + (1 2) (n+1) (n+ 2 )
4
1
2
+
(n) + (n) (n+ 2 )
4
1 1
2
+
u(n) (n) + u(n) (n) u(n+ 2 ) (n+ 2 )
4
(n) (n) n (n) (n+ 12 ) (
2
n+ 21 )
+
ga ( , u ) + ga ( , u ) ga ( , u )
4
1
2
+
d(u(n+ 2 ) ) d(u(n) ) d(u(n) )
, (3.398)
2
2
K1 2 d
2
+ K1 d
2
(n+ 21 )
(n+) (n+ 12 )
H upA1 =
t
4
4
K1 2 d
1
2
(1 2) K1 2 d
1
2
(n+ 2 )
+ (n+ 2 )
(n+1)
(n+)
+
4 4
2
2
K1 d
(n) (
n+ 12 )
2 K1 d
(n+)
(
2
n+ 12 )
+
+
4 4
K1 2 d
1 1
2
u (n) u(n+ 2 ) (n+ 2 )
(n)
+
4
K1 2 d
(n+) 1 1
2
+
u (n+) u(n+ 2 ) (n+ 2 )
, (3.399)
4
103
2
K1 2 d
2
+ K1 d
2
(n+ 21 )
(n+) (n+ 21 )
H upA2 =
t
4
4
(1 2)2 2 dK
12
1
2
(n+ 2 )
(n+)
+
4
(1 2)2 2 dK
12
(n+) 1
2
+
(n+ 2 )
4
K1 2 d
(n) (
2 K 2 d
n+ 21 )
1
(n) (
2
n+ 12 )
+ +
4 4
K1 2 d
1
2
(n+ 2 )
(n)
+
4
(1 2)2 2 dK 12
(n+) 1 1
2
+
u (n+) u(n+ 2 ) (n+ 2 )
4
K1 2 d
(n) (n) (n+ 12 ) (
2
n+ 21 )
+ u u
4
K1 2 d
1 1
2
(n) u(n+ 2 ) (n+ 2 )
,
(n)
+
u (3.400)
4
2
K1 2 d
2
+ K1 d
2
(n+ 21 )
(n+) (n+ 12 )
H upA3 =
t
4
4
K1 2 d
1
2
(1 2) K1 2 d
1
2
(n+ 2 )
+ (n+ 2 )
(n)
(n+1)
+
4 4
K1 2 d
1
2 K 2 d
1
2
(n+ 2 )
+ (n+ 2 )
(n) 1
(n)
+
4 4
2
K1 d
(n) (n) (n+ 12 ) (
2
n+ 12 )
+
u u
4
K1 2 d
1 1
2
u (n) u(n+ 2 ) (n+ 2 )
,
(n)
+ (3.401)
4
K1 2 d
1 1
2
H upB1 =
ga ( (n) , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) )
4
K1 2 d
1 1
2
+
ga ( (n+) , u(n+) ) ga ( (n+ 2 ) , u(n+ 2 ) )
4
K1 2 d
(n+ 12 )
2
(n)
+
d(u ) d(u )
2
K1 2 d
1
2
+
d(u(n+ 2 ) ) d(u(n+) )
, (3.402)
2
104
(1 2)2 2 dK
12
1 1
2
H upB2 =
ga ( (n+) , u(n+) ) ga ( (n+ 2 ) , u(n+ 2 ) )
4
K1 2 d
1 1
2
+
ga ( (n) , u(n) ) ga ( (n+ 2 ) , u(n+ 2 ) )
4
K1 2 d
(n) (n) (n+ 12 ) (
2
n+ 12 )
+ g ( , u ) g ( , u )
a a
4
K1 2 d
1
2
+
d(u(n+ 2 ) ) d(u(n) )
2
1
2
+ (1 2)2 2 dK 12
d(u(n+ 2 ) ) d(u(n+) )
2K1 2 d
(n+ 21 )
2
(n)
+ d(u ) d(u )
, (3.403)
4
2
H1 :=
dt (n+1)
4 !
(n+)
2 K1 d (1 2) (1 2) K1 d
2 2
+
+ + +
2 4 2 4
!
2 K 2 d
1
+
(n+1)
+
2 4
2 K 2 d
(n+)
1 4 4
+
+ + 4K1 d + 4K1 d
2 4
!
2
(n)
2 K d
1
+
+ 4K1 d + + 4K1 4 d4
2 4
! !
2 2 dK 2
2 2 K 2 d
+
(n+)
h2 +
1 1
+
(n)
h2 + ,
2 2
105
!
2
2 K1 2 d
H2 :=
dt (n+)
+
(n)
+ + + 4K1 4 d4
4 2 4
!
(n+)
2 K1 d (1 2) 2
2
+
+
4 41
(n+)
2 (1 2) (1 2) 2
2 2 4 4
+
+ + 4 (1 2) K1 d + K1 d
2 41
!
2 2
2 K 1 d K 1 d
+ 4K1 4 d4
(n)
+
+ + +
2 4 4
! !
(n)
2 2 K1 2 d
2
(n)
12
2 dK
+
+ +
+
2 2
!
(
2
n+)
(1 2) K1 2 d
+
(1 2) h2 + ,
2
and
!
2
2 K 2
1
H3 :=
dt (n+)
+
(n)
+ + 4K1 2 d2 + 4K1 4 d4
4 2 4
!
2 K 2 d
1
+
(n+)
+
2 4
!
(n+1)
2 (1 2) (1 2) K1 d
2
+
+
2 4
!
(n)
2
K1 2 d K1 2 d 4 4
+
+ + + + 4K1 d
2 4 4
! !
2 2 K 2 d
2 12
2 dK
1
+
(n)
h2 + +
(n)
h2 + .
2 2
Apply an inverse inequality to the kF k terms on the right hand side of (3.393), (3.394),
and (3.395) then multiply each expression by 2t/ and sum from (n = 0 to N 1), (n = 1
2
2
to N 1), and (n = 1 to N 1), respectivly. Bounds for
F ()
and
F (1)
are given
106
Thus,
(N ) 2 (N ) 2 (N 1+) 2
F + F + F
N 1 N 1 N 1
2 2 2 2 2 2
(1 2) F (n+) + t
X X X
+t F (n+1) + t F (n+)
n=0
n=1
n=1
N 1
2 2
(1 2) ( 41 ) u(n+) F (n+)
X
+t
n=1
N 1
!!
X (n) 2 2
K1 2 d 12
dK
t F + 4K1 d + 4K1 4 d4 + + Ch2 1+
n=0
2 4 2
N 1
X (n+1) 2 2
K1 2 d dK
12
1
+ t F 1 + 4 + 2 + + + (1 2) + + + + 2dK
n=0
2 4 4
!!
1 2 h2
dK 12
dK 12
dK
+ + Ch2 1 + 3 + 2 + + + 2 2 + 2 2 + 2 2
4 2 2
N 1
!
X 2 2
(n+) K1 d 2 (1 2) (1 2) K1 d2
+ t F + + +
n=0
2 4 2 4
N 1
!
K1 2 d 12
X (n+) 2 2 dK
4 4 2
+ t F + + 4dK1 + 4K1 d + Ch +
n=0
2 4 2
N 1
!!
X (n) 2 2
K1 2 d 12
dK
+ t F + + 4K1 4 d4 + + 4K1 2 d2 + 2Ch2 1 +
n=1
2 2
N 1
!!
X
2 2 (1 2)
K1 2 d
+ t F (n+1) 1+
n=1
2 2
N 1
X (n+) 2 2
(1 2)
+ t F 1 + 5 + + 2 + (1 ) + + +
n=1
2 2
(1 2) 2 K1 2 d 12
dK
+ + + + + 4K1 4 d4 + 4K1 2 d2
41 2 2 4
!
1 2 h2
dK
12
dK ff
+ + Ch2 1 + 3 + 2 + (1 + ) + 2 + 2 + 2 + (1 2) + 2
4 2
N 1
X (n) 2 2
(1 + ) (1 + )K1 2 d
+ t F 2 + +
n=1
2 4
!!
2 2 4 4 2 K1 2 d
+4K1 d + 4(1 + )K1 d + Ch +
2
N 1
X (n+) 2 2
(1 2) K1 2 d
+ t F 1 + 4 + + + (1 2) + (1 2) + 2 + +
n=1
2 2
(1 2) 2 2
K1 2 dh
+4 (1 2) K1 d + 4 (1 2) K1 4 d4 + 16K1 2 d2 + +
41 2
2
(1 2) K1 2 dh
5K1 2 d
+ + Ch2 5 + + 2 + 5 + 4 +
4 4
2 ff!
(1 2) dK1
+ + K1 2 d + 2 2 ( + 2)
2
H
2 2 2
+ F (0) + F (1) + F () + , (3.405)
107
where
N 1 N 1 N 1 N 1
2 X 2 X 2 X 2
H = t
X
H1 + t HG1 + t 2 H upA1 + t 2 H upB1
n=0 n=0 n=0 n=0
N 1 N 1 N 1 N 1
X 2 X 2 X 2 2 X 2 2
+ t H + t HG2 + t H upA2 + t H upB2
2
n=0 n=0 n=0 n=0
N 1 N 1 N 1 N 1
X 2 X 2 X 2 2 X 2 2
+ t H + t HG3 + t H upA3 + t H upB3 . (3.406)
3
n=0 n=0 n=0 n=0
(
2 (1 ) K1 2 d
1 t 1 + 5 + + 2 + + + + (1 2) +
2 2 4
(1 )K1 2 d (1 2) 2 dK 12 1 2 h2
dK
+ + + + + 4K1 4 d4 + 4K1 2 d2 +
2 41 2 4 4
2 )
dK 1
+ Ch2 1 + 3 + 2 + (1 + ) + 2 + 2 + 2 + (1 2) + 2 , (3.408)
2
and
(
2 (1 2) K1 2 d
1 t 2 + 5 + + (1 2) + 2 + +
2 2
2
(1 2) 2 1 + 2K1 d
+ 4 (1 2) K1 d + 4 (1 2) K1 4 d4 + + 4dK
41 4
2 2 2 2
K1 dh (1 2) K1 dh
+ 4K1 2 d2 + 16K1 2 d2 + 8K1 4 d4 + +
2 4
2 2
5K1 d (1 2) dK1
+ Ch2 5 + 3 + 2 + 4 + 4 + + + K1 2 d
4 2
)
2 K1 2 d dK
12
+ 2 ( + 2) + + , (3.409)
2 2
108
(n)
2
(n+1)
2
(n+)
2
(n+1)
2
coming from the coefficients on
F
,
F
,
F
,
F
, and
(n+)
2
((n))
2
F
,
F
terms respectively. Summarizing expressions (3.407), (3.408),
t Ch2 . (3.411)
For i = 1, 2, 3, with j = 0 if i = 1, and j = 1 if i > 1, bounds are now found for the
P 1 2 PN 1 2 PN 1 2 2 PN 1 2 2
expressions t N n=j Hi , t n=j HGi , t n=j H upAi , and t n=j H upBi
in (3.405). The H terms are bounded using interpolation properties and approximation
P 1 2
properties of the stress space Sh . Typical t N
n=j Hi terms are bounded as follows.
Analogous to (3.126)
N 1
X t
2
2
dt (n+1)
Ch2m+2 k t k0,m+1 , (3.412)
2
n=0
N 1 N 1
!
X
2 X
2
t (C1 )
(n+)
2m+2
t
(n+)
Ch
m+1
n=0 n=0
Ch2m+2 |kk|20,m+1 , (3.413)
and
N 1 N 1
!
X
2 X
2
t C2 h2 + 2 C3
(n)
Ch2m+2 + 2 Ch 2m
t
(n)
m+1
n=0 n=0
Ch2m+2 |kk|20,m+1 + Ch 2 2m
|kk|20,m+1 .(3.414)
Thus,
N 1
X 2
C h2m+2 + 2 h2m |kk|20,m+1
t H
i
n=j
109
PN 1 2
The t n=j HGi terms are bounded using Taylor series expansions similar to those given
by (A.3) and (A.6) in Appendix A. Typical terms are
N 1
2 N 1
2
X t
dt (n+1) t(n+ 2 )
=
1
X t
1
(n+1) 1 (n) t(n+ 2 )
1
n=0
2
n=0
2
t t
N 1 Z 2
X t 1 (n+1) 1 (n) (n+ 12 )
= t dA
n=0
2 t t
N 1
t (n+ 12 ) (t)2 (n+ 12 ) 1 tn+1
Z Z
t 1
(n+ 2 ) +
X 1
= t + tt + ttt (, t)(tn+1 t)2 dt
n=0
2 t 2 8 2 t 1
n+
2
2
1 t ( n+ 1
) (t) (n+ 1
)
(n+ 2 )
1
2
+ tt 2
t 2 t 8
2
1 tn+ 21
Z
(n+ 1 )
ttt (, t)(t(n+1) t)2 dt t 2 dA
2 tn
N 1 Z 2 Z tn+1 Z t 1 2
X t 1 n+
2
= ttt (, t)(tn+1 t)2 dt + ttt (, t)(tn+1 t)2 dt dA
n=0
2 2t tn+ 1 tn
2 2
N 1 Z 2 Z t
X t 1 n+1
2 ttt (, t)(t(n+1) t)2 dt
2 2t
n=0 t 1 n+
2
Z tn+ 1 2 !
2
+ ttt (, t)(tn+1 t)2 dt dA
tn
N 1 Z 2 Z tn+1 Z tn+1
X t 1 2
2 ttt (, t) dt (tn+1 t)4 dt
n=0
2 2t tn+ 1 tn+ 1
2 2
Z tn+ 1 Z tn+ 1
2 2
+ ttt (, t)2 dt (tn+1 t)4 dt dA
tn tn
N 1
t3 tn+1
Z Z Z t 1
X t 2
n+
2
= 6
ttt (, t) dt + ttt (, t)2 dt dA
n=0
2 5 (2 ) t 1 tn
n+
2
N 1 4 Z tn+1
X t 2
= k ttt (, t)k dt
n=0
5 (27 ) tn
4 2
C(t) k ttt k0,0 , (3.416)
N 1 N 1 Z tn+1
t
1
2
+ (n+1) (n+ 2 )
X
(n+) X 2
C(t)4 k tt (, t)k dt
2
n=0 n=0 tn+
2
C(t)4 k tt k0,0 , (3.417)
110
N 1
t
1
2
u (n) + u(n+) (n+) u(n+ 2 ) (n+ 2 )
X
(n) 1
n=0
2
N 1
(t)4 tn+1
X Z
2
k(u )tt k dt
n=0
48 tn
N 1
X (t)4 tn+1
Z
2
= ku tt + 2ut t + utt k dt
n=0
48 tn
N 1
X (t) 4 Z tn+1
2 2 2
4 ku tt k + 4 k2ut t k + 4 kutt k dt
n=0
48 tn
N 1
(t)4 tn+1
X Z
2 2 2 2
4 kuk dk tt k + 16 kut k dk t k
n=0
48 tn
2 2
+4 kutt k dkk dt
N 1
(t)4 tn+1
X Z
2 2 2
4K1 2 dk tt k + 16K1 2 dk t k + 4K1 2 dkk dt
n=0
48 tn
2 2 2
(t)4 C k tt k0,1 + k t k0,1 + kk0,1 . (3.418)
Next
N 1
t
1 1
2
ga ( (n) , u(n) ) + ga ( (n+) , u(n+) ) ga ( (n+ 2 ) , u(n+ 2 ) )
X
2
n=0
N 1
(t)4 tn+1
X Z
k(ga (, u))tt k2 dt
48 tn+
n=0
N
X 1 Z tn+1
4
(t) C1 k(u)tt k2 dt
n=0 tn+
N
X 1 Z tn+1
= (t)4 C1 k(u)tt + 2 t (u)t + tt uk2 dt
n=0 tn+
N
X 1 Z tn+1
(t)4 C2 k(u)tt k2 + k t (u)t k2 + k tt uk2 dt
n=0 tn+
N
X 1 Z tn+1
(t)4 C2 k(u)tt k2 d2 kk2 + k(u)t k2 d2 k t k2
n=0 tn+
+ kuk2 d2 k tt k2 dt
N
X 1 Z tn+1
(t)4 C2 K1 2 d2 kk2 + k t k2 + k tt k2 dt
n=0 tn+
(t)4 C k tt k20,0 + k t k20,0 + kk20,0 . (3.419)
111
PN 12
The last typical t n=j HGi term to consider is
N 1
2
X t
(n+ 1
) (n) (n+ )
d(u 2 ) d(u ) d(u )
n=0
N 1
(t)4 tn+1
X Z
kd(u)tt k2 dt
24 tn
n=0
N 1
X (t)4 tn+1Z
k(u)tt k2 dt
24 tn
n=0
N 1
X (t)4 tn+1Z
k(u)tt k2 d2 k1k2 dt
24 tn
n=0
(t)4 CCT , (3.420)
PN 1 PN 1
For t n=j H upAi and t n=j H upBi the typical terms are bounded as
N 1
2 K1 2 d X
(n+ 12 )
2
t
t
C 2 |k t k|2 , (3.422)
0,0
2
n=0
N 1 N 1
2 K1 2 d X
2 K1 2 d(t)2 tn+ 1
Z
1
2
t
(n+) (n+ 2 )
X 2
k t k2 dt
2 2 tn+
n=0 n=0
C(t) 2 2
k t k20,0 , (3.423)
N 1
2 K1 2 d X
1 1
2
t
u(n+) (n+) u(n+ 2 ) (n+ 2 )
2
n=0
N 1 2 2 Z tn+1
X K1 2 d(t)
kut + u t k2 dt
2 t 1
n=0 n+ 2
N 1 tn+1
2 K1 2 d(t)2
X Z
kut k2 + ku t k2 dt
t
n=0 n+ 1
2
C (t) kk0,1 + k t k20,1 ,
2 2 2
(3.424)
112
N 1
2 K1 2 d X
1 1
2
t
ga ( (n+) , u(n+) ) ga ( (n+ 2 ) , u(n+ 2 ) )
2
n=0
N 1 2 2 Z tn+1
X K1 2 d(t)
k(ga (, u))t k2 dt
2 t 1
n=0 n+ 2
N
X 1 Z tn+1
C1 2 (t)2 k(u)t k2 dt
n=0 tn+ 1
2
N
X 1 Z tn+1
= C1 2 (t)2 k t u + ut k2 dt
n=0 tn+ 1
2
N
X 1 Z tn+1
C2 2 (t)2 k t uk2 + kut k2 dt
n=0 tn+ 1
2
2
C (t) 2
k t k20,0 + kk20,0 . (3.425)
Lastly
N 1
2 K1 2 d X
1
2
t
d(u(n+ 2 ) ) d(u(n) )
n=0
N 1 2 2 Z tn+ 1
X K1 2 d(t)
2
kd(u)t k2 dt
tn
n=0
N 1 2 tn+ 1
K1 2 d(t)2
X Z
2
k(u)t k2 dt
tn
n=0
(t)2 CCT .
2
(3.426)
This gives
N 1 N 1
X 2 2 X 2
tH upAi + 2 tH upBi
n=j n=j
C 2 (t)2 kk20,1 + k t k20,1 + kk20,0 + k t k20,0
113
Apply the discrete Gronwall lemma to (3.405) to obtain
N
2
2
2
F N
+
F N 1+
F
+
N 1
2
2
(1 2) ( 51 )
u(n+) F (n+)
G (t, h, ), (3.428)
X
+ t
n=0
where
G (t, h, ) := C(t) 4
k ttt k20,0 + k tt k20,1 + k t k20,1 + kk20,1
2 2 2
+ k tt k0,0 + k t k0,0 + kk0,0 + CT
2 2 2 2 2 2
+ C(t) kk0,1 + k t k0,1 + kk0,0 + k t k0,0 + CT
By (3.428)
Then
and
114
3.6 Establishing the Full -method Error Estimate
The error estimates obtained in Theorems 3.5.1 and 3.5.2 are now used to obtain
an a priori error estimate for the full -method described by Steps 1a - 3b in Section 3.2.
2
with = 1 2 and t Ch2 , Ch, the fractional step -method approximations
where
1 1
F (t, h, ) := C 1 + 2 + Fu (t, h) + C 1 + 2 F (t, h, ), (3.435)
and Fu (t, h) and F (t, h, ) are defined by (3.39) and (3.142) in Theorems 3.5.1 and
3.5.2, respectively.
Theorem 3.6.1 is established using Theorems 3.5.1 and 3.5.2. Recall that
uh := uh uh , and h := h h .
Assume
(0) (0) (0)
uh = u(0) Xh , h = (0) Sh , and ph = p(0) Qh . (3.436)
115
The error estimates in Theorems 3.5.1 and 3.5.2 establish the following bounds that will be
()
2 ()
2
2
2
()
u uh
, t
d(u() uh )
,
E ()
, and
()
Fu
2
(t, h), (3.437)
and
()
2
()
2
() 2
h
, and
F
F (t, h, ), (3.438)
and
2
F (t, h, ) = C(t) 4
k ttt k20,0 + k tt k20,1 + k t k20,1 + kk20,1
2 2 2
+ k tt k0,0 + k t k0,0 + kk0,0 + CT
+ C(t)2 2 kk20,1 + k t k20,1 + kk20,0 + k t k20,0 + CT
The proof of Theorem 3.6.1 requires Induction Hypothesis 1 stated in (3.34) and Induction
Hypothesis 2.
Induction Hypothesis 2 Under the assumptions of Theorem 3.6.1 there exists a constant
K3 such that
(n)
(n+)
(n+1)
h
,
h
, and
h
< K3 . (3.441)
116
Verification of Inductive Hypothesis 2: Assume expression (3.441) holds for n =
d d d
C t2 h 2 + th 2 + hm 2
d d d
+hm 2 + hm+1 2 + h 2 + K1 . (3.442)
(N )
d
As the resulting bound for
h
is independent of N , for m 2 1 and h, t, and
such that
d d
hm+1 2 1, and t2 , h 2 . (3.443)
5. The triangle inequality and the results of stated in Theorems 3.5.1 and 3.5.2 are
then used to establish Theorem 3.6.1.
117
3.6.1 The first time step (0, t]
2 (t, h, ) defined in (3.440) and the assumption in (3.436), () ()
Step 1a: With F h = h ,
and
()
2 ()
2
()
() 2
h
=
h
F (t, h, ). (3.444)
()
2
2
uh
tCF (t, h, ), (3.445)
Proof: In the initial time step the variational formulation using a known stress for the
Re ()
()
Re (0)
uh , v + 2(1 ) d(uh ), d(v) = u , v
t t h
(0)
1 (0) 1 () 1 (0) 1 ()
2(1 ) d(uh ), d(v) + f + f ,v + , d(v) . (3.446)
2 2 2 2
The full -method approximation for Step 1b (using an approximated stress) gives the
variational formulation
Re ()
()
Re (0)
uh , v + 2(1 ) d(uh ), d(v) = u ,v
t t h
(0)
1 (0) 1 () 1 (0) 1 ()
2(1 ) d(uh ), d(v) + f + f ,v + h , d(v) . (3.447)
2 2 2 h 2
(0) (0) (0) (0)
Subtract (3.447) from (3.446), making note that uh = uh and h = h , as u(0) Xh
Re ()
()
1
()
uh , v + (1 ) d(uh ), d(v) + () h , d(v) = 0. (3.448)
t 2
()
Choose v = uh Xh and examine (3.448) term by term giving
Re () () Re
()
2
uh , uh =
uh
, (3.449)
t t
()
2
() ()
(1 ) d(uh ), d(uh ) = (1 )
d(uh )
, (3.450)
118
and
1 () () ()
1
() ()
2
1
()
2
h , d(uh )
h
+ (1 )
d(uh )
. (3.451)
2 8(1 ) 2
Re
()
2 (1 )
()
2 1 ()
2
()
uh
+
d(uh )
h
. (3.452)
t 2 8(1 )
()
2
2
uh
tC1b1 F (t, h, ) (3.453)
and
()
2
2
t
d(uh )
tC1b2 F (t, h, ), (3.454)
where
1
C1b1 := , and C1b2 := . (3.455)
8Re(1 ) 4(1 )2
()
2
2
uh
tCF (t, h, ) + (t)5 C. (3.456)
()
Proof: Implementing a Crank-Nicolson method for Step 2a would require h , which
()
is not obtained until Step 2b. Instead, a second order extrapolated value, exh , obtained
()
using (0) and h , is used in Step 2a (see Lemma A.4 in the Appendix). The variational
formulation for Step 2a in the initial step of the -method algorithm using a known solution
for is
Re
()
()
Re
()
uh , v + 2(1 ) d(uh ), d(v) = uh , v
(1 2)t (1 2)t
1 1 ()
() () ()
2(1 ) d(uh ), d(v) + f , v + , d(v) , (3.457)
2 2
119
and the variational formulation for Step 2a using the extrapolated stress is
Re
()
()
Re
()
uh , v + 2(1 ) d(uh ), d(v) = uh , v
(1 2)t (1 2)t
1 1
1 () 1 ()
()
2(1 ) d(uh ), d(v) + f () + f () , v exh + h , d(v) . (3.458)
2 2 2 2
()
Subtract (3.458) from (3.457) with v = uh to get
Re () () ()
() ()
uh uh , uh + t2(1 ) d(uh ), d(uh )
(1 2)
t
() () ()
+ t2(1 ) d(uh ), d(uh ) + () ex
()
, d(u h )
2 h
t
() ()
+ () h , d(uh ) = 0. (3.459)
2
Re () Re
()
2 Re
()
2
() ()
uh uh , uh
uh
uh
, (3.460)
(1 2) 2 (1 2) 2 (1 2)
2
() () ()
t2(1 ) d(uh ), d(uh ) = t2(1 )
d(uh )
, (3.461)
() ()3t 2 (1 )
()
2
t2(1 ) d(uh ), d(uh )
d(uh )
t(1 )
2
()
+
d(uh )
, (3.462)
3
t () 3t t(1 )
2
()
2
() ()
() ()
h , d(uh )
h
+
d(uh )
, (3.463)
2 16(1 ) 3
and
t () 3t
2 t(1 )
2
() ()
() ()
h , d(uh )
(ex)h
+
d(uh )
. (3.464)
2 16(1 ) 3
120
Bring (3.460) - (3.464) together to get
()
2
Re Re
2
()
2
()
uh
uh
+ t(1 )
d(uh )
2 (1 2) 2 (1 2)
3t 2 (1 )
()
2
3t
() ()
2
d(uh )
+
h
16(1 )
3t
()
2
+
(ex)h
. (3.465)
16(1 )
()
For the second order temporal extrapolation of exh (see Lemma A.4)
!
2
()
2
() ( )
(ex)h
=
() h (0)
!
2
() ()
( ) (0) () ()
=
+ h
2 2
!
() ()
( ) (0)
() ()
2
2
+
h
2
()
2
= (t)4 Cext +
()
h
, (3.466)
()
where Cext denotes a constant obtained from the extrapolation of h . Thus, (3.465) yields
()
2
()
2
6 (1 2) 2 (1 )
()
2
uh
uh
t
d(uh )
Re
!
3 (1 2) 2
() ()
2
3 (1 2)
+ t 1+
h
+ (t)5 Cext . (3.467)
8Re(1 ) 8Re(1 )
121
Thus, from (3.465)
2
()
2 ()
2
()
t
d(uh )
C2a4
uh
+ C2a5 t
d(uh )
()
2
+tC2a6
() h
+ (t)5 C2a7 Cext
2
tC 2a F (t, h, ) + (t)5 C2a7 Cext , (3.468)
con2
and
()
2
2
h
tC 2a F
u (t, h, ) + (t)5 C2a3 Cext . (3.469)
con1
2 (t, h), and F 2 (t, h, ) are defined in (3.439), and (3.440), respectively.
where Fu
() () () () ()
uh h u() h = (u() uh ) h
() () () ()
+ (u() uh ) ( () h ) (u() uh ) () u() h , (3.471)
and
() () () () ()
ga ( h , uh ) ga ( h , u() ) = ga h , (u() uh )
() () ()
+ ga ( () h ), (u() uh ) ga () , (u() uh )
()
ga h , u() . (3.472)
122
The difference of general upwinded terms A and B is written
() () ()
A, uh h B, u() h
() () ()
= A B, u() h A, u() uh h . (3.473)
Hence, the variational formulation for Step 2b of the -method algorithm using a known
solution for u is
()
()
()
()
h , + h , () + u() h , () + ga ( h , u() ), ()
(1 2)t
()
()
2 d(u() ), () h , + h , () = 0. (3.474)
(1 2)t
Using the approximated velocity uh in Step 2b gives the fully discrete problem
()
() () () () ()
h , + h , () + uh h , () + ga ( h , uh ), ()
(1 2)t h h h
()
()
()
2 d(uh ), () h , + h , () = 0. (3.475)
h (1 2)t h
()
Subtract (3.474) from (3.475) and let = h to obtain
() ()
F2b ( h ) + F2bup ( h ) = 0, (3.476)
()
() () ()
() ()
F2b ( h ) := h h , h + h , h
(1 2)t
() () () () () () () ()
+ uh h , h u() h , h + ga ( h , uh ), h
() () () () () ()
ga ( h , u() ), h + 2 d(u() uh ), h + h , h , (3.477)
123
and
() () () () () ()
F2bup ( h ) := h , uh h h , u() h
() () () () () ()
+ uh h , uh h u() h , u() h
() () () () () ()
+ ga ( h , uh ), uh h ga ( h , u() ), u() h
() () () ()
2 d(uh ), uh h + 2 d(u() ), u() h
() () () () ()
+ h , uh h h , u() h . (3.478)
()
The first two terms in F2b ( h ) are bounded as
()
2
()
2
() () ()
h h , h
h
h
(3.479)
(1 2)t (1 2)t
and
()
2
() ()
h , h =
h
. (3.480)
()
Using (3.471) the convective terms in F2b ( h ) may be written as
() () () () ()
uh h , h u() h , h
() () () () () ()
= (u() uh ) h , h + (u() uh ) ( () h ), h
() () () ()
(u() uh ) () , h u() h , h . (3.481)
Three of the terms on the right hand side of (3.481) are bounded via
1
2
() () () ()
()
u() uh h , h d2 Ch1
u() uh
h
, (3.482)
() ()
u() h , h = 0, (3.483)
() () ()
()
(u() uh ) () , h d
()
u() uh
h
2 d2 K1 2
()
2
()
()
2
u uh
+ 0
h
. (3.484)
40
124
The last unbounded term in (3.481) is written as
() () () () ()
(u() uh ) ( () h ), h = (u() uh ) () , h
() () () () () ()
(u() uh ) h , h uh h , h . (3.485)
2 d2 CK3 2 2
()
2
Fu (t, h) + 2 h2
h
, (3.487)
42
and
() () ()
2 d2 K3 2 Ch2
()
2
()
2
uh h , h
uh
+ 3
h
. (3.488)
43
()
Use (3.472) to rewrite the ga terms in F2b ( h ) as
() () () () () () () ()
ga ( h , uh ), h ga ( h , u() ), h = ga h , (u() uh ) , h
() () () () ()
ga h , (u() uh ) , h ga h , u() , h . (3.489)
Split the second term on the right hand side of (3.489) such that
() () () () ()
ga h , (u() uh ) , h = ga h , E () , h
() () () () ()
+ ga h , () , h + ga h , uh , h . (3.491)
125
Use inverse inequalities and interpolation properties to obtain
() ()
()
()
ga h , E () , h 4d
h
E ()
h
2 2
4 d K3 C
2
()
2
2
2
()
E
+ 4 h
h
4
42 d2 K3 2 C 2
()
2
Fu (t, h) + 4 h2
h
, (3.492)
4
()
()
42 d2 K3 2 2
()
2
ga h , () , h Fu (t, h) + 5
h
, (3.493)
5
and
() ()
()
42 d2 K3 2 Ch2
()
2
()
2
ga h , uh , h
uh
+ 6
h
. (3.494)
6
()
2
2
()
2
2 d(() ), h
()
+ 8
h
8
2 2
()
2
Fu (t, h) + 8
h
, (3.497)
8
and
() ()
2 Ch2
()
2
()
2
2 d(uh ), h
uh
+ 9
h
. (3.498)
9
126
The last term in (3.477) is bounded using
() ()
2
()
2
()
2
h , h
h
+ 10
h
. (3.499)
410
()
The first two terms in F2bup ( h ) can be rewritten as
() () () () ()
h , uh h h , u() h
() () () () ()
= h , u() h h , u() uh h
() () () () ()
= h , u() h + h , u() uh h
() () ()
+ () h , u() uh h
() ()
() , u() uh h . (3.500)
127
The two ga (, ) terms are rewritten
() () () () () ()
ga ( h , uh ), uh h ga ( h , u() ), u() h
() () ()
= ga ( h , (u() uh )), u() h
() () ()
ga h , (u() uh ) , u() h
() ()
ga ( h , u() ), u() h
() () () ()
+ ga ( h , u() uh ), u() uh h
() () () ()
ga ( h , u() uh ), u() uh h
() () ()
ga ( h , u() ), u() uh h
() () ()
+ ga ( h , u() ), u() uh h . (3.502)
1
2
() () () () 1
() ()
()
h , u uh h d Ch
u uh
h
,
2 (3.506)
128
and
() () ()
() h , u() uh h
2
()
2
()
()
2
()
h
+ 11
u() uh h
. (3.507)
411
Rewrite
() () () ()
() , u() uh h = () , u() uh h
() ()
+ () , uh h ,
and
2 2 d2 K 2 Ch2
()
2
()
2
() () 1
() , uh h u
h
+ 13
h
.
(3.509)
413
() () ()
2 (u() uh ) h , u() h
() ()
2 K1 2
2 2 dCh
()
2
14
(u() uh ) h
+
h
, (3.510)
14
() () () ()
u() uh h , uh h
2 d3 C 2 K2 2 K3 2 2
()
2
Fu (t, h) + 2 15 h4
h
, (3.511)
415
()
() () () 2 d3 C 2 K2 2 K3 2
()
2
()
2
uh h , uh h
uh
+ 2 16 h4
h
, (3.512)
416
2
() () ()
u() h , u() h =
u() h
, (3.513)
129
and
2
() () () () () ()
u() uh h , u() uh h =
u() uh h
. (3.514)
3
()
()
()
d 2
u
h
u() uh
h
( )
2
d C K1 K3 2
3 2 2
2
()
2
() 2 4
()
u u h
+ 17 h
h
.
(3.515)
417
Next consider bounding the upwinded ga (, ) terms in (3.502). The first term is
() () ()
ga ( h , (u() uh )), u() h
3
2
()
()
4d2 K1 Ch1
(u() uh )
h
. (3.516)
() () () () ()
ga h , (u() uh ) , u() h = ga h , E () , u() h
() () () () ()
+ ga h , () , u() h + ga h , uh , u() h ,
() ()
ga h , E () , u() h
42 d3 C 2 K1 2 K3 2 2
()
2
Fu (t, h) + 2 18 h4
h
, (3.517)
18
42 d2 K 2
2
() () 3 ()
ga h , () , u() h 2
Fu (t, h)+19
u() h
, (3.518)
19
130
and
() () ()
ga h , uh , u() h
42 d2 K3 2 Ch2
()
2
2
()
uh
+ 20
u() h
. (3.519)
20
Next
()
2
3
() ()
ga ( h , u() ), u() h 4d2 K1 2 Ch1
h
. (3.520)
Similarly
() () () ()
ga ( h , u() uh ), u() uh h
() () () () ()
= ga ( h , E () ), u() h ga ( h , E () ), uh h
() () ()
+ ga ( h , () ), u() uh h
() () () ()
+ ga ( h , uh ), u() uh h .
() ()
ga ( h , E () ), u() h
42 d3 C 2 K1 2 K3 2 2
()
2
Fu (t, h) + 2 21 h4
h
, (3.521)
21
() () ()
ga ( h , E () ), uh h
42 d3 C 2 K2 2 K3 2 2
()
2
Fu (t, h) + 2 22 h4
h
, (3.522)
22
() () ()
ga ( h , () ), u() uh h
42 d2 K3 2 2
()
2
()
Fu (t, h) + 23
u() uh h
, (3.523)
23
131
and
() () () ()
ga ( h , uh ), u() uh h
42 d2 K3 2 Ch2
()
2
()
2
()
uh
+ 24
u() uh h
. (3.524)
24
Also
() () () ()
ga ( h , u() uh ), u() uh h
42 d2
2
2
()
()
()
2
()
u() uh
h
+ 25
u() uh h
. (3.525)
25
For
() () ()
ga ( h , u() ), u() uh h
() () () () () ()
= ga ( h , u() ), u() uh h + ga ( h , u() ), uh h ,
() () ()
ga ( h , u() ), u() uh h
4 2 2 d4 CK1 2 K3 2 2
()
2
Fu (t, h) + 26 h2
h
(3.526)
26
and
2 42 d4 K 2 K 2 Ch2
()
2
()
2
() () () 1 2
ga ( h , u() ), uh h
uh
+ 27
h
. (3.527)
27
() () ()
ga ( h , u() ), u() uh h
42 d2 K1 2
()
2
()
2
()
h
+ 28
u() uh h
. (3.528)
28
132
and bounded as
() ()
2 d(E () ), u() h 2
d(E () )
u() h
1
()
2d2
u()
E ()
h
2 2
dC K1 2 2
()
2
Fu (t, h) + 2 29 h4
h
, (3.529)
29
2
2
() ()
2 d(() ), u() h 2
Fu (t, h) + 30
u() h
, (3.530)
30
and
2 Ch2
()
2
2
() () ()
2 d(uh ), u() h
()
u
h
+ 31
u h
. (3.531)
31
Similarly
() () ()
2 d(u() uh ), u() uh h
() () ()
= 2 d(E () ), u() h 2 d(E () ), uh h
() ()
+2 d(() ), u() uh h
() () () ()
+2 d(uh ), u uh h ,
and
()
2 K1 2
2 dC
()
2
2 d(E () ), u() h 2
Fu (t, h) + 2 32 h4
h
, (3.532)
32
() ()
2 K2 2
2 dC
()
2
2 d(E () ), uh h 2
Fu (t, h) + 2 33 h4
h
, (3.533)
33
() ()
2 d(() ), u() uh h
2 2
()
2
()
Fu (t, h) + 34
u() uh h
, (3.534)
34
133
() () ()
2 d(uh ), u() uh h
2 Ch2
()
2
()
2
()
uh
+ 35
u() uh h
. (3.535)
35
with
2 2 CK 2
()
2
() () 1
2 d(u() ), u() uh h 2
Fu (t, h) + 36 h2
h
, (3.536)
36
and
2 2 K 2 Ch2
()
2
()
2
() () 1
2 d(u() ), uh h
uh
+ 37
h
. (3.537)
37
Next bound the terms in (3.504):
() ()
2
()
2
2
()
h , u() h
h
+ 38
u() h
, (3.538)
438
() () ()
() h , u() uh h
2
() ()
2
() ()
2
()
h
+ 39
u uh h
, (3.539)
439
and
()
()
()
2
()
2
()
2
()
h , u() uh h
h
+ 40
u() uh h
. (3.540)
440
Write
() () () () () () () ()
, u uh h = , u
uh h
() ()
+ () , uh h ,
134
with bounds
2 d2 2 CK 2
()
2
() () 1
() , u() uh h 2
Fu (t, h) + 41 h2
h
, (3.541)
441
and
2 d2 2 K 2 Ch2
()
2
()
2
() () 1
() , uh h
uh
+ 42
h
. (3.542)
442
135
Thus,
(
()
2
(1 2)
h
1 + t 0 1 3 5 6 8 9 10 13
27 37 42 h2 (2 + 4 + 7 + 12 + 26 + 36 + 41 )
()
2
(1 2)
2
()
h
+ t ( 19 20 30 31 38 )
u() h
(1 2)
+t 11 14 23 24 25
2
() ()
28 34 35 39 40
u() uh h
!
(1 2)t 2
2
2
()
2
+ +
h
410 440 438
( !)
(1 2)t 2 d3 C 2 K2 2 K3 2 2 2 C 2 d2 K3 2 C 42 d2 K3 2 C
+ +h + +
416 9 43 6
( )
42 d2 K3 2 C 42 d2 K3 2 C 2 C 2 C
+h2 + + +
20 24 31 35
( )!
2 2
2 2 2 d C 42 d4 K2 2 C 2 C d2 2 C
()
2
+ h K1 + + +
uh
413 27 37 442
(1 2)t 2 1 1
()
()
2
+ +
h
411 439
!
(1 2)t 2 2 2 1 1 32
dK
()
2
()
+ d K1 + +
u uh
40 41 417
(
d2 C 4d2 C 4d2 4d3 C 2 K1 2 4d3 C 2 K1 2 4d3 C 2 K2 2
+t (1 2)K3 2 + + + + +
42 4 5 18 21 22
) " #
d3 C 2 K2 2
(1 2) 2 C 1 K12 2 1 1 1
+ + + + dC + +
415 7 8 29 32 33
( )
(1 2) 42 d2 K3 2 42 d2 K3 2
1 1
+ + + 2 +
19 23 30 34
( )!
2 (1 2)K1
2
2 C 42 d4 CK3 2 2 d2 C d2 2 C 2
+ + + + Fu (t, h). (3.543)
36 26 412 441
136
Define
(
(1 2)
J2b (h, ) = 0 1 3 5 6 8 9 10 13
27 37 42 h2 (2 + 4 + 7 + 12 + 26 + 36 + 41 )
and constants
(1 2)
C2b1 := 19 20 30 31 38 ,
(1 2)
C2b2 := 11 14 23 24 25 28 34 35 39 40 ,
(1 2) 2
C2b3 := ,
410
(1 2) 2 2
C2b4 := + ,
440 438
(1 2)d3 C 2 K2 2 K3 2
C2b5 := ,
416
!
2 C d2 K3 2 C 4d2 K3 2 C
C2b6 := (1 2) + + ,
9 43 6
( )
(1 2) 42 d2 K3 2 C 42 d2 K3 2 C 2 C 2 C
C2b7 := + + + ,
20 24 31 35
( )
(1 2)K1 2 2 d2 C 42 d4 K2 2 C 2 C d2 2 C
C2b8 := + + + ,
413 27 37 442
(1 2) 2 1 1
C2b9 := + ,
411 439
!
(1 2) 2 2 2 1 1 32
dK
C2b10 := 2 d K1 + + ,
40 41 417
(3.545)
137
(
d2 C 4d2 C 4d2 4d3 C 2 K1 2
C2b11 := (1 2)K3 2 + + +
42 4 5 18
)
4d3 C 2 K1 2 4d3 C 2 K2 2 d3 C 2 K2 2
+ + +
21 22 415
" #
(1 2) 2 C 1 K1
2 2 1 1 1
+ + + dC + + ,
7 8 29 32 33
( )
(1 2) 42 d2 K3 2 42 d2 K3 2
2 1 1
C2b12 := + + + ,
19 23 30 34
( )
(1 2)K1 2 2 C 42 d4 CK3 2 2 d2 C d2 2 C
C2b13 := + + + ,
36 26 412 441
C 2b := C2b5 + C2b10 ,
con1
C 2b := C2b10 + C2b11 .
con2
Thus,
!
()
2
()
2
1 + tJ2b (h, )
h
h
2
2
()
() ()
+tC2b1
u() h
+ tC2b2
u() uh h
()
2
(tC2b3 + tC2b4 )
h
!
()
2
+ t(C2b5 + C2b10 ) + th2 C2b6 + th2 C2b7 + 2 th2 C2b8
uh
2
()
2
()
+tC2b9
() h
+ tC2b10
u() uh
!
+t C2b11 + C2b12 + 2 C2b13 Fu
2
(t, h)
()
2
(tC2b3 + tC2b4 )
h
!
()
2
+ tC 2b + th2 C2b6 + th2 C2b7 + 2 th2 C2b8
uh
con1
!
2
+tC2b9 F (t, h, ) + 2
tC 2b + tC2b12 + 2 tC2b13 Fu (t, h). (3.546)
con2
There are two general cases to consider when setting up the coefficients on the
telescoping sum.
138
Case 1 (J2b (h, ) 0): In this case (3.546) gives:
()
2
()
2
2
2
()
() ()
h
h
+ tC2b1
u() h
+ tC2b2
u() uh h
()
2
(tC2b3 + tC2b4 )
h
!
()
2
+ tC 2b + th2 C2b6 + th2 C2b7 + 2 th2 C2b8
uh
con1
!
2 2 2
+tC2b9 F (t, h, ) + tC 2b + tC2b12 + tC2b13 Fu (t, h).
con2
Case 2 (J2b (h, ) < 0): Set t sufficiently small such that (1+tJ2b (h, )) remains positive.
()
2
Assume that t|J2b (h, )| < 12 and subtract tJ2b (h, )
h
from both sides of (3.546).
Thus,
()
2
()
2
(1 + tJ2b (h, ))
h
h
2
2
()
() ()
+tC2b1
u() h
+ tC2b2
u() uh h
()
2
()
2
t|J2b (h, )|
h
+ (tC2b3 + tC2b4 )
h
!
()
2
+ tC 2b + th2 C2b6 + th2 C2b7 + 2 th2 C2b8
uh
con1
!
2
+tC2b9 F (t, h, ) + tC 2b + tC2b12 + 2 tC2b13 Fu
2
(t, h). (3.547)
con2
139
Divide (3.547) by the positive factor (1 + tJ2b (h, )) to obtain
()
2
()
2
h
h
2
2
t
() ()
() ()
()
+ C2b1
u h
+ C2b2
u uh h
(1 + tJ2b (h, ))
()
2
()
2
2t|J2b (h, )|
h
+ 2 (tC2b3 + tC2b4 )
h
!
()
2
+2 tC 2b + th2 C2b6 + th2 C2b7 + 2 th2 C2b8
uh
con1
!
2
+t2C2b9 F (t, h, ) + 2 tC 2b + tC2b12 + 2 tC2b13 Fu
2
(t, h). (3.548)
con2
Case 1 and Case 2 described above lead to bounds of the same convergence order
once the discrete Gronwalls inequality (Lemma grolma) has been applied. However, Case 2
has a more restrictive condition that must be satisfied for the discrete Gronwalls inequality
to be appled. Thus, without loss of generality for the remainder of the proof only Case 2
()
2
will be considered. (Note that
h
in the first time step (0, t] is zero. Including these
()
2
h
2 t2 C 2b C 2a + t2 h2 C2b6 C 2a + tC2b9
con1 con1 con1
!
+t2 h2 C2b7 C 2a + 2 t2 h2 C2b8 C 2a 2
F (t, h, )
con1 con1
!
+2 tC 2b + tC2b12 + 2 tC2b13 Fu
2
(t, h)
con2
+2 tC 2b + th2 C2b6
con1
!
2 2 2
+th C2b7 + th C2b8 (t)5 C2a3 Cext . (3.549)
140
Step 3a: The variational formulation for Step 3a of the -method algorithm using a known
(1)
(1)
()
()
h , + h , () h , + h , ()
t t
() ()
+ u() h , () + ga ( h , u() ), () 2 d(u() ), () = 0. (3.550)
The variational formulation for Step 3a of the -method algorithm using an approximated
velocity is:
(1) (1) () ()
h , + h , () h , + h , ()
t h t h
() () () () ()
+ uh h , () + ga ( h , uh ), () 2 d(uh ), () = 0. (3.551)
h h h
(1)
Subtract (3.551) from (3.550)and let = h to get
(1) (1)
F3a ( h ) + F3aup ( h ) = 0, (3.552)
and
(1) (1) (1) (1) () (1)
F3aup ( h ) := h , u() h h , uh h
() (1) () () (1)
+ h , u() h h , uh h
() (1) () () () (1)
+ u() h , u() h uh h , uh h
() (1) () () () (1)
+ ga ( h , u() ), u() h ga ( h , uh ), uh h
(1) () () (1)
2 d(u() ), u() h + 2 d(uh ), uh h . (3.554)
141
The first three terms in (3.553) are bounded as
(1)
(1)
2
()
2
() (1)
h h , h
h
h
, (3.555)
t 2t
(1)
2
(1) (1)
h , h =
h
, (3.556)
and
() (1)
2
()
2
(1)
2
h , h
h
+ 1
h
. (3.557)
41
Rewrite the convective terms
() (1) () () (1)
u() h , h uh h , h
() () (1) () () (1)
= uh h , h + (u() uh ) h , h . (3.558)
Inductive Hypothesis 1 is used to to bound the first term on the right hand side of (3.558)
as
() () (1)
2 2 Ch2
2 dK
()
2
(1)
2
uh h , h
h
+ 2
h
. (3.559)
42
where
() () () (1)
2 d2 K3 2 C 2
2
2
(1)
(u uh ) h , h Fu (t, h) + 3 h
h
(3.561)
43
and
() () (1)
2 d2 K3 2 Ch2
()
2
(1)
2
(uh ) h , h
uh
+ 4
h
. (3.562)
44
142
Next write
() (1) () () (1)
ga ( h , u() ), h ga ( h , uh ), h
() () (1) () () (1)
= ga h , uh , h + ga h , (u() uh ) , h . (3.563)
Rewrite
() () (1) () (1)
ga h , (u() uh ) , h = ga h , E () , h
() (1)
+ ga h , () , h
() () (1)
+ ga h , uh , h ,
()
(1)
42 d2 K3 2 2
(1)
2
ga h , () , h Fu (t, h) + 7
h
, (3.566)
7
and
() ()
(1)
42 d2 K3 2 Ch2
()
2
(1)
2
ga h , uh , h
uh
+ 8
h
. (3.567)
8
143
(1)
2 2
(1)
2
2 d(() ), h Fu (t, h) + 10
h
, (3.569)
10
and
2 Ch2
()
2
(1)
2
() (1)
2 d(uh ), h
uh
+ 11
h
. (3.570)
11
(1)
Each of the upwinded terms in F1aup ( h ) is written as
(1) (1) (1) () (1)
h , u() h h , uh h
(1) () (1) (1) (1)
= h , u() uh h + h , u() h
(1) () (1)
+ h , (u() uh ) h , (3.571)
() (1) () () (1)
h , u() h h , uh h
() () (1) () () (1)
= h , uh h + h , (u() uh ) h , (3.572)
() (1) () () () (1)
u() h , u() h uh h , uh h
() () () (1) () () () (1)
= (u() uh ) h , uh h + uh h , uh h
() () (1)
+ u() h , (u() uh ) h , (3.573)
() (1) () () () (1)
ga ( h , u() ), u() h ga ( h , uh ), uh h
() () () (1) () () () (1)
= ga ( h , (u() uh )), uh h + ga ( h , uh ), uh h
() () (1)
+ ga ( h , u() ), (u() uh ) h , (3.574)
and
(1) () () (1)
2 d(u() ), u() h 2 d(uh ), uh h
() () (1) () (1)
= 2 d(uh ), u() uh h + 2 d(u() uh ), u() h . (3.575)
144
Bound the terms in (3.571) as
()
(1)
2
1
(1) () (1)
h , u() uh h d2 Ch1
u() uh
h
, (3.576)
(1) (1)
h , u() h = 0, (3.577)
and
(1) () (1)
(1)
() ()
(1)
h , (u() uh ) h d
h
u u h
h
2 d2 2 K3 2 C
2
()
(1)
2
u uh
+ 12 h2
h
. (3.578)
()
412
and
() () (1)
2 d2 2 K3 2 C
2
()
2
2
(1)
h , (u()
()
uh ) h
u uh
+ 14 h
h
. (3.580)
414
() () () (1)
2 2 d2 K2 4 Ch4
()
2
(1)
2
uh h , uh h
h
+ 16
h
, (3.582)
416
and
() () (1)
2 d3 K1 2 K3 2 C
2
()
u() h , (u() uh ) h
()
u u h
417
2
(1)
+17 2 h4
h
. (3.583)
145
The first term on the right hand side of (3.574) is written as
() () () (1)
ga ( h , (u() uh )), uh h
() () (1)
= ga ( h , E () ), uh h
() () (1)
+ ga ( h , () ), uh h
() () () (1)
+ ga ( h , uh )), uh h . (3.584)
Then,
() () (1) ()
() (1)
ga ( h , E () ), uh h
ga ( h , E () )
uh h
3
()
()
()
(1)
4d
h
E
uh
h
2
2 3 2 2
4 d K2 K3 C
()
2
(1)
2
E
+ 18 2 h4
h
18
4 d K2 2 K3 2 C
2 3
(1)
2
2
Fu (t, h) + 18 2 h4
h
, (3.585)
18
() () (1)
ga ( h , () ), uh h
4 2 2 d3 K2 2 K3 2 C 2
2
2
(1)
Fu (t, h) + 19 h
h
, (3.586)
19
and
4 2 2 d3 K 2 K 2 Ch4
()
2
(1)
2
() () () (1) 2 3
ga ( h , uh ), uh h
uh
+ 20
h
. (3.587)
20
146
For the last term on the right hand side of (3.574) write
() () (1)
ga ( h , u() ), (u() uh ) h
() (1)
= ga ( h , u() ), E () h
() (1)
+ ga ( h , u() ), () h
() () (1)
+ ga ( h , u() ), uh h , (3.589)
which is bound as
() (1)
ga ( h , u() ), E () h
4 2 2 d4 K1 2 K3 2 C 2
(1)
2
Fu (t, h) + 22 h2
h
, (3.590)
22
() (1)
ga ( h , u() ), () h
4 2 2 d4 K1 2 K3 2 C 2
(1)
2
Fu (t, h) + 23 h2
h
, (3.591)
23
and
() () (1)
4 2 2 d4 K1 2 K3 2 Ch2
()
2
(1)
2
ga ( h , u() ), uh h
uh
+ 24
h
. (3.592)
24
where
2 d2 K 2 C
(1)
2
() (1) 2
2 d(uh ), E () h 2
Fu (t, h) + 25 2 h4
h
, (3.593)
25
2 d2 K 2 C
(1)
2
() (1) 2
2 d(uh ), () h 2
Fu (t, h) + 26 2 h4
h
, (3.594)
26
147
2 2 d2 K 2 Ch4
()
2
(1)
2
() () (1) 2
2 d(uh ), uh h
uh
+ 27
h
. (3.595)
27
Next
() (1) (1)
2 d(u() uh ), u() h = 2 d(E () ), u() h
(1)
+2 d(() ), u() h
() () (1)
+2 d(uh ), u h ,
with bounds
(1)
12C
2 dK
(1)
2
2 d(E () ), u() h 2
Fu (t, h) + 28 2 h4
h
, (3.596)
28
(1)
12C
2 2 dK
(1)
2
2 d(() ), u() h 2
Fu (t, h) + 29 h2
h
, (3.597)
29
and
() (1)
1 2 Ch4
2 2 dK
()
2
(1)
2
2 d(uh ), u() h
uh
+ 30
h
. (3.598)
30
148
Bring all the bounds together for Step 3a. Thus,
(
(1)
2
2
h
1 + t 1 + 2 + 5 + 7 + 8 + 10 + 11 + 13
+16 + 20 + 21 + 24 + 27 + 30
2
h 3 + 6 + 9 + 12 + 14 + 19 + 22 + 23 + 29
2 4
h 15 + 17 + 18 + 25 + 26 + 28
)!
()
2
1
()
d2 Ch1
u() uh
h
()
2
2 2 2 2 Ch2 42 d2 K2 2 Ch2
2 dK
h
t + +
41 42 5
2 2 4 4 !
2 d K2 Ch 42 d3 K2 4 Ch4 2 dK
2 2 Ch2
+ + +
416 21 413
2 42 d2 K3 2 Ch2 2 Ch2
2 4 2 2 2
()
2 2 4 d K1 K3 Ch
+
uh
t + +
8 11 24
!
42 d3 K2 2 K3 2 Ch4 2 d2 K2 2 Ch4 2 dK
1 2 Ch4
+ + +
20 27 30
2
()
2 2 d2 K2 2 K3 2 C 2 d3 K1 2 K3 2 C
+
u() uh
t +
415 417
2 2 2 !
2 d K3 C d2 2 K3 2 C
+ +
412 414
2 d2 K3 2 C 42 d2 K3 2 C 42 d2 K3 2 2 C 2
2
+t + + + +
43 6 7 9 10
42 d3 K2 2 K3 2 C 2 d2 K2 2 C 2 d2 K2 2 C 12C
2 dK
+ + + +
18 25 26 28
2 3 2 2
4 d K2 K3 C 4 d K1 K3 C
2 4 2 2
+ 2 +
19 22
!
42 d4 K1 2 K3 2 C 12C
2 dK 2
+ + Fu (t, h) (3.599)
23 29
149
Define
(
2
J3a (h, ) = 1 + 2 + 5 + 7 + 8 + 10 + 11 + 13 + 16 + 20 + 21
2
+24 + 27 + 30 h 3 + 6 + 9 + 12 + 14 + 19 + 22 + 23 + 29
)
1
()
2 h4 15 + 17 + 18 + 25 + 26 + 28 d2 Ch1
u() u
,(3.600)
h
and constants
2 2
C3a1 := ,
41
!
22C
dK 4d2 K2 2 C
C3a2 := 2 + ,
42 5
22C
2 2 dK
C3a3 := ,
413
!
2 4 1 4d
C3a4 := 2d K2 C + ,
416 21
!
2 42 d2 K3 2 C 2 C
C3a5 := + ,
8 11
8d4 K1 2 K3 2 C
C3a6 := ,
24
!
2 42 d3 K2 2 K3 2 C 2 d2 K2 2 C 12C
2 dK
C3a7 := + + ,
20 27 30
!
d2 K2 2 d3 K1 2
C3a8 := 2K3 2 C + ,
415 417
!
2 d2 2 K3 2 C d2 2 K3 2 C
C3a9 := + ,
412 414
2 2 d2 K3 2 C 42 d2 K3 2 C 42 d2 K3 2 2 C 2
C3a10 := + + + +
43 6 7 9 10
2 3
4 d K2 K3 C2 2 2 2
d K2 C 2 2
d K2 2 C
2 12C
2 dK
+ + + + ,
18 25 26 28
2 42 d3 K2 2 K3 2 C 42 d4 K1 2 K3 2 C 42 d4 K1 2 K3 2 C 12C
2 dK
C3a11 := + + + ,
19 22 23 29
150
C 3a := 2C3a8 + C3a10 , and C 3a := 2C3a9 + C3a11 .
con1 con2
(1)
2
()
2
(1 + tJ3a (h, ))
h
h
()
2
tC3a1 + th2 C3a2 + 2 th2 C3a3 + 2 th4 C3a4
h
()
2
+ th2 C3a5 + 2 th2 C3a6 + 2 th4 C3a7
uh
2
2
() ()
+ tC3a8 + tC3a9
u uh
+ tC3a10 + 2 tC3a11 Fu
2
(t, h)
()
2
tC3a1 + th2 C3a2 + 2 th2 C3a3 + 2 th4 C3a4
h
()
2
2 2 2 2 2 4
+ t2C3a8 + th C3a5 + t2C3a9 + th C3a6 + th C3a7
uh
2
+ tC 3a + 2 tC 3a Fu (t, h). (3.601)
con1 con2
As in (3.546), assume that J3a (h, ) < 0 and t is small enough so that t|J3a (h, )| < 21 .
()
2
Subtract tJ3a (h, )
h
from both sides of (3.601) and divide the resulting expression
(1)
2
()
2
h
h
()
2
2 tC3a1 + th2 C3a2 + 2 th2 C3a3 + 2 th4 C3a4
h
()
2
+2 t2C3a8 + th2 C3a5 + 2 t2C3a9 + 2 th2 C3a6 + 2 th4 C3a7
uh
()
2
+2t|J3a (h, )|
h
+ 2 tC 3a + 2 tC 3a Fu2
(t, h). (3.602)
con1 con2
Step 3b: The variational formulation for Step 3b using a known stress for the -method
algorithm is
Re (1)
(1)
uh , v + 2(1 ) d(uh ), d(v)
t
Re ()
()
= uh , v 2(1 ) d(uh ), d(v) + f (1) , v (1) , d(v) . (3.603)
t
151
The variational formulation for the full -method approximation for Step 3b (using an
approximated stress) is
Re (1)
(1)
uh , v + 2(1 ) d(uh ), d(v)
t
Re ()
()
(1)
= uh , v 2(1 ) d(uh ), d(v) + f (1) , v h , d(v) . (3.604)
t
(1)
Subtract (3.604) from (3.603) and choose v = uh Xh :
Re (1) () (1)
(1) (1)
uh uh , uh + 2(1 ) d(uh ), d(uh )
t
() (1) (1) (1)
+ 2(1 ) d(uh ), d(uh ) + (1) h , d(uh ) = 0. (3.605)
(1)
2
()
2
(1) () (1)
2 uh uh , uh
uh
uh
, (3.606)
2t 4t(1 )
(1)
2
(1) (1)
2(1 ) d(uh ), d(uh ) =
d(uh )
, (3.607)
Re Re
2t
() (1)
4t 2 (1 )
2
()
2(1 ) d(uh ), d(uh )
d(uh )
Re Re
t(1 )
(1)
2
+
d(uh )
, (3.608)
Re
and
(1)
2
()
2 (1)
2
1
uh
uh
+ (C3b1 ) t
d(uh )
2
(1)
2
()
(1)
C3b2 t
d(uh )
+ C3b3 t
h
, (3.610)
152
where
Re 4 2 (1 )
C3b1 := , C3b2 := , C3b3 := .
2(1 ) Re Re(1 )
A bound for a general time step of the -method algorithm is needed to complete the
analysis. The bound across a general time step is found by extending the bounds obtained
above for the initial time step to the general time step (nt, (n + 1)t].
Step 1a: Step 1a and Step 3a of the -method for viscoelastic fluid flow are identical in
(n+)
2
(n)
2
h
h
(n)
2
2 tC3a1 + th2 C3a2 + 2 th2 C3a3 + 2 th4 C3a4
h
(n)
2
+2 t2C3a8 + th2 C3a5 + 2 t2C3a9 + 2 th2 C3a6 + 2 th4 C3a7
uh
(n)
2
+2t|J3a (h, )|
h
+ 2 tC 3a + 2 tC 3a Fu 2
(t, h). (3.611)
con1 con2
(n)
2
(n+)
2
2
C3b2 t
d(uh )
+ 2C3b3 t
h
+ 2C3b3 tF (t, h, ). (3.612)
Step 2a: To obtain the general bound for Step 2a, define
Re 4 (1 2) 2 (1 2)
C2a8 := , C2a9 := , and C2a10 := .
2 (1 2) Re Re
(n+)
2
(n+)
2
2
1 (n+)
uh
uh
+ (C2a8 ) t
d(uh )
2
(n+)
2
(n+)
2
tC2a9
d(uh )
+ 2tC2a10
h
+ 2tC2a10 F (t, h, ). (3.613)
153
Step 2b: The bound for Step 2b is found by examining (3.548).
(n+)
2
(n+)
2
h
h
2
t
(n+)
+ C2b1
u(n+) h
(1 + tJ2b (h, ))
2
(n+) (n+) (n+)
+C2b2
u uh h
(n+)
2
(n+)
2
2t|J2b (h, )|
h
+ 2 (tC2b3 + tC2b4 )
h
!
(n+)
2
+2 tC 2b + th2 C2b6 + th2 C2b7 + 2 th2 C2b8
uh
con1
!
2 22
+t2C2b9 F (t, h, ) + 2 tC 2b + tC2b12 + tC2b13 Fu (t, h). (3.614)
con2
Step 3a: From (3.602) the following bound for Step 3a is obtained:
(n+1)
2
(n+)
2
h
h
(n+)
2
2 tC3a1 + th2 C3a2 + 2 th2 C3a3 + 2 th4 C3a4
h
(n+)
2
+2 t2C3a8 + th2 C3a5 + 2 t2C3a9 + 2 th2 C3a6 + 2 th4 C3a7
uh
2
(n+)
+2t|J3a (h, )|
h
+ 2 tC 3a + 2 tC 3a Fu2
(t, h). (3.615)
con1 con2
Step 3b: The bound for Step 3b at the general nth step follows from (3.610).
154
For the first unit stride sum (3.611) - (3.616):
(n+1)
2
+ (C3b1 )1 t
d(uh )
2
t
(n+)
+ C2b1
u(n+) h
(1 + tJ2b (h, ))
2
(n+) (n+) (n+)
+C2b2
u uh h
(n)
2
2 2 2 2 4
2 tC3a1 + th C3a2 + th C3a3 + th C3a4
h
(n+)
2
+2 (tC2b3 + tC2b4 )
h
(n+)
2
+2 tC3a1 + th2 C3a2 + 2 th2 C3a3 + 2 th4 C3a4
h
2
2
2
(n+)
(n+)
(n+1)
+2tC3b3
h
+ 2tC2a10
h
+ 2tC3b3
h
155
For the second unit stride sum (3.611) - (3.616) with (n (n + 1)) in expressions (3.611)
(n+1)
2
1
+ (C3b1 ) t
d(uh )
2
t
(n+)
+ C2b1
u(n+) h
(1 + tJ2b (h, ))
2
(n+) (n+) (n+)
+C2b2
u uh h
(n+1)
2
2 tC3a1 + th2 C3a2 + 2 th2 C3a3 + 2 th4 C3a4
h
2
(n+)
+2 (tC2b3 + tC2b4 )
h
(n+)
2
+2 tC3a1 + th2 C3a2 + 2 th2 C3a3 + 2 th4 C3a4
h
2
2
2
(n+1+)
(n+)
(n+1)
+2C3b3 t
h
+ 2tC2a10
h
+ 2C3b3 t
h
+2 t(2C3a8 ) + th2 C3a5 + 2 t(2C3a9 )
(n+1)
2
2 2 2 4
+ th C3a6 + th C3a7
uh
!
(n+)
2
+2 tC 2b + th2 C2b6 + th2 C2b7 + 2 th2 C2b8
uh
con1
+2 t(2C3a8 ) + th2 C3a5 + 2 t(2C3a9 )
(n+)
2
2 2 2 4
+ th C3a6 + th C3a7
uh
156
For the third unit stride, sum (3.611) - (3.616) with (n (n + 1)) in (3.611), (3.612),
157
Sum (3.617), (3.618), and (3.619), applying inverse estimates to the right hand side defor-
mation terms to obtain:
(n+1) 2 (n) 2 (n+1+) 2 (n+) 2 (n+1+) 2 (n+) 2
h h + h h + h h
2 2 2 2 2
(n+) 2
(n+1) (n) (n+1+) (n+) (n+1+)
+ uh uh + uh uh + uh uh
2 2 2
+ (C3b1 )1 t d(uh ) + 2 (C2a8 )1 t d(uh ) + 3 (C3b1 )1 t d(uh
(n+) (n+) (n+1)
)
2 2
+2 (C3b1 )1 t d(uh ) + (C2a8 )1 t d(uh
(n+1+) (n+1+)
)
2 ff
t
2
(n+) (n+) (n+)
(2C2b1 ) u(n+) h
(n+)
+ + (2C2b2 ) u uh h
(1 + tJ2b (h, ))
2 ff
t
2
(n+1+) (n+1+) (n+) (n+) (n+1+)
+ C2b1 u h + C2b2 u uh h
(1 + tJ2b (h, ))
(n) 2
+ t(2C3a1 ) + th2 (2C3a2 ) + 2 th2 (2C3a3 ) + 2 th4 (2C3a4 ) h
`
(n+) 2
+ (t(4C2b3 + 4C2a10 + 2C3b3 ) + t(4C2b4 )) h
(n+1) 2
+ t(4C3a1 + 6C3b3 ) + th2 (4C3a2 ) + 2 th2 (4C3a3 ) + 2 th4 (4C3a4 ) h
`
2
+ t(6C3a1 ) + th2 (6C3a2 ) + 2 th2 (6C3a3 ) + 2 th4 (6C3a4 ) h
` (n+)
2
(n+1+)
+ (t(2C2b3 + 4C3b3 + 2C2a10 ) + t(2C2b4 )) h
(n+) 2
+th2 (2CC2a9 ) uh
(n+1+) 2
+th2 CC2a9 uh
!
(n+1+) 2
2 2 2 2
+ t(2C 2b ) + th (2C2b6 ) + th (2C2b7 ) + th (2C2b8 ) uh
con1
158
where
2 2
HF 2 = t(12C 3a + 6C 2b ) + t(6C2b12 ) + t(6C2b13 + 12C 3a ) Fu (t, h)
con1 con2 con2
2
+ t(12C3b3 + 6C2a10 ) + t(6C2b9 ) F (t, h, ). (3.621)
(0) (0) ()
Sum expression (3.620) from n = 0 to N 1, and noting uh , h , and h are zero, yields
a bound at time (N )t = T . Thus,
(N ) 2 (N +) 2 (N +) 2 (N ) 2 (N +) 2 (N +) 2
h + h + h + uh + uh + uh
1 1 1
1 NX
(n+) 2
N 2
(n+) 1 NX
(n+1) 2
) + 2 (C2a8 )1
` X `
+ C3b1 t d(uh t d(uh ) + 3 C3b1 t d(uh )
n=0 n=0 n=0
N 1 N 1 2
(n+1+) 2
1
(n+1+)
) + (C2a8 )1
` X X
+2 C3b1 t d(uh t d(uh )
n=0 n=0
N 1 2 N 1 2
(2C2b1 ) X
(n+) C2b1 X
(n+1+)
+ t u(n+) h + t u(n+1+) h
(1 + tJ2b (h, )) n=0 (1 + tJ2b (h, )) n=0
N 1 2
(2C2b2 ) X
(n+)
(n+)
+ t u(n+) uh h
(1 + tJ2b (h, )) n=0
N 1 2
C2b2 X
(n+)
(n+1+)
+ t u(n+) uh h
(1 + tJ2b (h, )) n=0
N 1
s + u + u () 2 () 2 () 2
X
h + h + uh + HF 2 , (3.622)
n=0
where
N 1
(n) 2 `
s X
2 2 2 2 4
= h t(2C3a1 ) + th (2C3a2 ) + th (2C3a3 ) + th (2C3a4 )
n=0
N 1
(n+) 2
X
+ h (t(4C2b3 + 4C2a10 + 2C3b3 ) + t(4C2b4 ))
n=0
N 1
(n+1) 2 `
X
2 2 2 2 4
+ h t(4C3a1 + 6C3b3 ) + th (4C3a2 ) + th (4C3a3 ) + th (4C3a4 )
n=0
N 1
(n+) 2 `
X
2 2 2 2 4
+ h t(6C3a1 ) + th (6C3a2 ) + th (6C3a3 ) + th (6C3a4 )
n=0
N 1
(n+1+) 2
X
+ h (t(2C2b3 + 4C3b3 + 2C2a10 ) + t(2C2b4 ))
n=0
N 1 N 1 N 1
(n) 2 (n+) 2 (n+) 2
X X X
+2t|J3a (h, )| h + 4t|J2b (h, )| h + 6t|J3a (h, )| h
n=0 n=0 n=0
N 1 N 1
(n+1) 2 (n+1+) 2
X X
+4t|J3a (h, )| h + 2t|J2b (h, )| h , (3.623)
n=0 n=0
159
and
N 1
(n) 2
u X
= uh t(4C3a8 ) + th2 (CC3b2 + 2C3a5 )
n=0
!
2 2 2 2 4
+ t(4C3a9 ) + th (2C3a6 ) + th (2C3a7 )
N 1 N 1
(n+) 2 (n+1+) 2
X X
2 2
+ uh th (2CC2a9 ) + uh th CC2a9
n=0 n=0
N 1
(n+) 2
X
+ uh t(12C3a8 + 4C 2b ) + th2 (3CC3b2 + 6C3a5 + 4C2b6 ) + 2 t(12C3a9 )
con1
n=0
!
2 2 2 2 4
+th (4C2b7 ) + th (4C2b8 + 6C3a6 ) + th 6C3a7
N 1
(n+1) 2
X
+ uh t(8C3a8 ) + th2 (2CC3b2 + 4C3a5 )
n=0
!
+ 2 t(8C3a9 ) + 2 th2 (4C3a6 ) + 2 th4 (4C3a7 )
1
N
!
(n+1+) 2
X
2 2 2 2
+ uh t(2C 2b ) + th (2C2b6 ) + th (2C2b7 ) + th (2C2b8 ) .(3.624)
con1
n=0
The discrete Gronwalls inequality (Lemma 1.4.2) requires inequalities (3.625) - (3.630) be
(n)
2
(n+1)
2
satisfied. The coefficients on the
h
and
h
terms in s give the inequality
(n+)
2
(n+1+)
2
The
h
and
h
terms yield the inequality
(n+)
2
and the
h
terms give
160
2
(n+1)
2
From u the
u(n)
h
and
uh
terms yield
(n+)
2
(n+1+)
2
the
uh
and
uh
terms give
(n+)
2
(n+1+)
2
and the coefficients on the
uh
and
uh
terms yield the inequality
and
(n+) (n+)
u
uh
K1 + K2 .
Thus,
161
Apply the discrete Gronwalls inequality (Lemma 1.4.2) to (3.622). Hence,
(N )
2
(N )
2
h
+
uh
N 1 N 1
2
(n+)
2
(n+)
+ (C3b1 )1 )
+ 2 (C2a8 )1
X X
t
d(uh t
d(uh )
n=0 n=0
N 1 N 1
2
1
X
(n+1)
2
(2C2b1 ) X
(n+) (n+)
+3 (C3b1 ) t
d(uh )
+ t
u h
(1 + tJ2b (h, ))
n=0 n=0
N 1
2
(2C2b2 ) X
(n+)
(n+)
+ t
u(n+) uh h
(1 + tJ2b (h, ))
n=0
G(t, h, ), (3.632)
where
G(t, h, ) = CT 1 + + 2 Fu
2 2
(t, h) + CT 1 + F (t, h, ). (3.633)
Thus,
N
(n)
2
X
|kuh k|20,0 = t
uh
n=0
N
X
tG(t, h, ) 1
n=0
= T G(t, h, ),
and similarly
|k h k|20,0 T G(t, h, ).
162
Apply Korns inequality (see [10]) to (3.634) to obtain
C3b1
|kuh k|20,0 G(t, h, ), (3.635)
3
and
|k h k|20,0 T G(t, h, ) + F
2
(t, h, ).
Finally,
holds true for n = 1, 2, . . . , N 1. Interpolation properties, and inverse estimates (see [11])
163
give
(N )
(N )
uh
uh u(N )
+
u(N )
(N )
(N )
(N )
E
+
+
uh
+ K1
d
d
d
(N )
Ch 2
E (N )
+ Ch 2
(N )
+ Ch 2
uh
+ K1 . (3.637)
Using (3.132)
d
d d d
Ch 2
E (N )
C hk 2 + hq+1 2 + (t)2 h 2 . (3.638)
From (3.632)
d d d
d2
(N )
Ch
uh
C hk 2 + hm+1 2 + hq+1 2
d d d d
+(t)2 h 2 + (t)h 2 + hm 2 + h 2 , (3.639)
d
d
d
Ch 2
(N )
Chk+1 2
u(N )
Chk+1 2 . (3.640)
k+1
d d
an expression independent of N . Hence, provided k, m 2 1, q 2 1, and values of
d d d d
hk 2 , hm+1 2 , hq+1 2 1, and t2 , h 2 ,
Then
(N )
uh
< K1 + 7C.
(N +)
(N +1)
Similarly it follows that
uh
, and
uh
< K1 + 7C.
164
CHAPTER 4
In this chapter numerical results for the -method applied to viscoelastic fluid flow
are presented using two test problems. The first example uses a known analytical solution
to verify numerical convergence rates for the -method. The second example simulates vis-
coelastic flow through a 4:1 planar contraction, a prototypical problem for viscoelastic fluid
flow. Finite element computations were done using the FreeFem++ integrated development
environment [30]. Continuous piecewise quadratic elements were used for modeling the ve-
locity, and continuous piecewise linear elements were used for the pressure and stress. The
errors are O((t)2 ). The influence of the value for on the numerical approximations
with = 0. The remaining computations used the optimal value. The constitutive and
conservation equation splitting parameters and were set to 1/2 for the computations
presented in Tables 4.1, 4.2, 4.3, and 4.4. The analysis presented in Chapter 3 used a Crank-
Nicolson method in Steps 1b and 2a of the initial time step. The computations presented in
this chapter were all implemented using the algorithm as stated in Section 3.2 of Chapter 3.
4.1 Example 1
The theoretical convergence rates were verified by considering fluid flow across a
165
5
x 10
7 2.5
6.5
5.5 2
4.5
= 0.2
4 1.5
= 0.27
3.5 = Optimal
= 0.315
3 = 0.385
2.5 1
0.2 0.22 0.24 0.26 0.28 0.3 0.32 0.34 0.36 0.38 0 5 10 15 20 25
Value 1/h
= 2d(u). (4.3)
(4.1)-(4.3).
known stress, , are presented in Table 4.1. These results correspond to the analysis of
166
Step 1b, Step 2a, and Step 3b as stated in Theorem 3.5.1. Note the following corollary to
Theorem 3.5.1.
Corollary 4.1.1 For Xh the space of continuous, piecewise quadratic functions, and Qh the
there exists a constant C such that the approximation uh satisfies the error estimate:
The numerical convergence rates observed in Table 4.1 are consistent with those predicted
in Corollary 4.1.1. Note the proof of Theorem 3.5.1 required the restriction t Ch2 .
Table 4.1 Approximation errors and convergence rates for |ku uh k| and |kp ph k| at
T = 2.
1 2 1 2 1 2 1 2 1 2 1 2
(t, h) ,
2 4 ,
4 8 ,
8 16 ,
16 32 ,
32 64 ,
64 128
|ku uh k|0,1 1.4552e-1 5.0959e-2 1.2837e-2 3.2943e-3 8.5448e-4 2.2329e-4
Cvge. Rate - 1.5 2.0 2.0 2.0 1.9
|ku uh k|0,0 1.5039e-2 2.8838e-3 5.2841e-4 1.0373e-4 2.1197e-5 4.4002e-6
Cvge. Rate - 2.4 2.5 2.3 2.3 2.3
|ku uh k|,0 1.4446e-2 3.1673e-3 6.3994e-4 1.3203e-4 2.7402e-5 5.7571e-6
Cvge. Rate - 2.2 2.3 2.3 2.3 2.3
|kp ph k|0,0 6.6917e-0 1.1223e-1 1.4973e-2 6.9976e-3 3.4928e-3 1.7535e-3
Cvge. Rate - 5.9 2.9 1.1 1.0 1.0
|kp ph k|,0 4.7931e-0 8.5149e-2 1.6671e-2 8.6430e-3 4.4883e-3 2.2925e-3
Cvge. Rate - 5.8 2.4 0.9 0.9 1.0
167
4.1.2 Approximating h with u and p known
Next the approximation of the stress h assuming the velocity and pressure functions
are known is investigated. This is done by implementing Step 1a, Step 2b, and Step 3a of
the -method algorithm. The following corollary is obtained from Theorem 3.5.2.
Corollary 4.1.2 For Sh the space of continuous, piecewise linear functions, t Ch2 ,
and , u, p sufficiently smooth, there exists a constant C such that the approximation h
The numerical convergence rates presented in Table 4.2 are consistent with those predicted
in Corollary 4.1.2. In Table 4.2 the effect of the upwinding parameter on |k h k|0,0
The proof of Theorem 3.5.2 also requires the restriction t Ch2 . Here the nu-
Tables 4.3 and 4.4 contain the results for the approximation of u, p and using
the -method described by Step 1a - Step 3b. The following corollary is obtained from
Theorem 3.6.1.
Corollary 4.1.3 For Xh the space of continuous, piecewise quadratic functions, Sh and
smooth, there exists a constant C such that the approximations uh and h satisfy the error
estimates
168
Table 4.2 Approximation errors and convergence rates for |k h k| at T = 2.
2 1 2 1 2 1 2 1 2
(t, h) 1, 2 2, 4 4, 8 8 , 16 16 , 32
0 |k h k|0,0 2.1235e-1 6.6773e-2 1.9191e-2 5.0437e-3 1.2830e-3
Cvge. Rate - 1.7 1.8 1.9 2.0
|k h k|,0 1.7214e-1 5.7062e-2 1.6462e-2 4.4028e-3 1.1387e-3
Cvge. Rate - 1.6 1.8 1.9 2.0
h |k h k|0,0 2.0070e-1 8.4563e-2 3.7449e-2 1.6980e-2 8.0428e-3
2
Cvge. Rate - 1.2 1.2 1.1 1.1
|k h k|,0 1.6409e-1 6.1898e-2 3.1010e-2 1.5523e-2 7.62823e-3
Cvge. Rate - 1.4 1.0 1.0 1.0
3
2
h |k h k|0,0 2.0174e-1 7.3678e-2 2.3575e-2 6.9629e-3 2.0645e-3
2
Cvge. Rate - 1.5 1.6 1.8 1.8
|k h k|,0 1.6474e-1 5.8665e-2 1.7012e-2 5.4369e-3 1.7269e-3
Cvge. Rate - 1.5 1.8 1.6 1.7
2
h
2
|k h k|20,0 2.0346e-1 6.9501e-2 2.0245e-2 5.3281e-3 1.3546e-3
Cvge. Rate - 1.5 1.8 1.9 2.0
|k h k|,0 1.65915e-1 5.7620e-2 1.6546e-2 4.4105e-3 1.1399e-3
Cvge. Rate - 1.5 1.8 1.9 2.0
169
The numerical convergence rates in Tables 4.3 and 4.4 are consistent with Corol-
lary 4.1.3. Pressure is treated implicitly in all steps of the algorithm, and a first order
convergence rate is observed for both |kp ph k|0,0 and |kp ph k|,0 .
Table 4.3 Approximation errors and convergence rates for |ku uh k|,0 , |k h k|,0 ,
and |kp ph k|,0 at T = 2.
1 2 1 2 1 2 1 2 1 2
(t, h) ,
2 4 ,
4 8 ,
8 16 ,
16 32 ,
32 64
0 |ku uh k|,0 2.6018e-3 5.0502e-4 1.2559e-4 3.1530e-5 7.9659e-6
Cvge. Rate - 2.4 2.0 2.0 2.0
|k h k|,0 5.6338e-2 1.6455e-2 4.3977e-3 1.1373e-3 2.8942e-4
Cvge. Rate - 1.8 1.9 2.0 2.0
|kp ph k|,0 8.1077e0 1.2817e-1 8.5460e-3 4.4574e-3 2.2822e-3
Cvge. Rate - 6.0 3.9 0.9 1.0
h |ku uh k|,0 2.8998e-3 1.1712e-3 5.4285e-4 2.5489e-4 1.2285e-4
2
Cvge. Rate - 1.3 1.1 1.1 1.1
|k h k|,0 6.1238e-2 2.9514e-2 1.3939e-2 6.6777e-3 3.2636e-3
Cvge. Rate - 1.1 1.1 1.1 1.0
|kp ph k|,0 8.1079e0 1.3145e-1 1.5620e-2 7.6836e-3 3.8197e-3
Cvge. Rate - 6.0 3.1 1.0 1.0
3
2
h |ku uh k|,0 2.7188e-3 7.2552e-4 2.2485e-4 6.9023e-5 2.2812e-5
2
Cvge. Rate - 1.9 1.7 1.7 1.7
|k h k|,0 5.8004e-2 1.7633e-2 5.5220e-3 1.7079e-3 5.3942e-4
Cvge. Rate - 1.7 1.7 1.7 1.7
|kp ph k|,0 8.1078e0 1.28850e-1 8.6091e-3 4.4585e-3 2.2812e-3
Cvge. Rate - 6.0 3.9 0.9 1.0
2
h |ku uh k|,0 2.6484e-3 5.7890e-4 1.4843e-4 3.7558e-5 9.4995e-6
2
Cvge. Rate - 2.2 2.0 2.0 2.0
|k h k|,0 5.6933e-2 1.6543e-2 4.4059e-3 1.1381e-3 2.8949e-4
Cvge. Rate - 1.8 1.9 2.0 2.0
|kp ph k|,0 8.1078e0 1.2834e-1 8.5574e-3 4.4574e-3 2.2821e-3
Cvge. Rate - 6.0 3.9 0.9 1.0
As was the case in Section 4.1.2 the effect of the upwinding parameter can be seen
170
Table 4.4 Approximation errors and convergence rates for |ku uh k|0,1 , |ku uh k|0,0 ,
|k h k|0,0 , and |kp ph k|0,0 at T = 2.
1 2 1 2 1 2 1 2 1 2
(t, h) ,
2 4 ,
4 8 ,
8 16 ,
16 32 ,
32 64
0 |ku uh k|0,1 5.3126e-2 1.3580e-2 3.5106e-3 9.2452e-4 2.4611e-4
Cvge. Rate - 2.0 2.0 1.9 1.9
|ku uh k|0,0 2.8471e-3 6.6322e-4 1.6269e-4 4.1253e-5 1.0498e-5
Cvge. Rate - 2.1 2.0 2.0 2.0
|k h k|0,0 6.5800e-2 1.9868e-2 5.3430e-3 1.38326e-3 3.5382e-4
Cvge. Rate - 1.7 1.9 1.9 2.0
|kp ph k|0,0 1.0359e1 1.5906e-1 7.7094e-3 3.5582e-3 1.7673e-3
Cvge. Rate - 6.0 4.4 1.1 1.0
h |ku uh k|0,1 5.6155e-2 1.8351e-2 7.1198e-3 3.1101e-3 1.4553e-3
2
Cvge. Rate - 1.6 1.4 1.2 1.1
|ku uh k|0,0 3.4867e-3 1.4315e-3 6.1829e-4 2.8092e-4 1.3303e-4
Cvge. Rate - 1.3 1.2 1.2 1.1
|k h k|0,0 8.2889e-2 3.6696e-2 1.6088e-2 7.4667e-3 3.6056e-3
Cvge. Rate - 1.2 1.2 1.1 1.1
|kp ph k|0,0 1.0360e1 1.6301e-1 1.9587e-2 9.5799e-3 4.7499e-3
Cvge. Rate - 6.0 3.1 1.0 1.0
3
2
h |ku uh k|0,1 5.4420e-2 1.4811e-2 4.1097e-3 1.1722e-3 3.4391e-4
2
Cvge. Rate - 1.9 1.8 1.8 1.8
|ku uh k|0,0 3.1541e-3 9.2163e-4 2.6985e-4 8.0835e-5 2.4851e-5
Cvge. Rate - 1.8 1.8 1.7 1.7
|k h k|0,0 7.2634e-2 2.4166e-2 7.1489e-3 2.1020e-3 6.3693e-4
Cvge. Rate - 1.6 1.8 1.8 1.7
|kp ph k|0,0 1.0360e1 1.5990e-1 9.8979e-3 4.2639e-3 1.9876e-3
Cvge. Rate - 6.0 4.0 1.2 1.1
2
h |ku uh k|0,1 5.3686e-2 1.3925e-2 3.6220e-3 9.55202e-4 2.5414e-4
2
Cvge. Rate - 1.9 1.9 1.9 1.9
|ku uh k|0,0 2.9926e-3 7.4891e-4 1.8748e-4 4.7623e-5 1.2108e-5
Cvge. Rate - 2.0 2.0 2.0 2.0
|k h k|0,0 6.8587e-2 2.1004e-2 5.6463e-3 1.4598e-3 37310e-4
Cvge. Rate - 1.7 1.9 2.0 2.0
|kp ph k|0,0 1.0359e1 1.5928e-1 8.1413e-3 3.6559e-3 1.7892e-3
Cvge. Rate - 6.0 4.3 1.2 1.0
171
4.2 Example 2
channel is presented for a second example. This has been a long standing benchmark
problem for viscoelastic flow [41, 48, 51]. A diagram of the flow geometry is given in
Figure 4.2. It is assumed that the channel lengths are sufficiently long for fully developed
Poiseuille flow at both the inflow and outflow boundaries. In the computations the value of
16L
4L
16L
L
The flow at t = 0 is assumed to be stationary and then slowly increased for t > 0
using A(t) = 1 et . The boundary conditions at the inflow of the channel are defined by
1 2
32 1 y
u = A(t) , (4.6)
0
and
where
172
No slip boundary conditions are imposed for the velocity on the solid walls of the contraction,
and a symmetry condition is imposed along the bottom of the computational domain. The
computations were performed on a uniformly refined version of the mesh shown in Figure
0
0 4 8
The computations were done using the full -method approximation given by (3.28)
- (3.33) for an Oldroyd B fluid (a = 1), with = 2 and Re = 1. The value of was set
to 8/9, which is commonly used in the literature [48]. The time step size and upwinding
parameter were set to t = 1/32 and = (2ymin )2 , respectively. Figures 4.4, 4.5, and 4.6
show streamlines for the fluid at times t = 1/8, 3/8, 1/2, 3/4, 1, and 2 superimposed on
a contour plot showing the magnitude of velocity. Note that, consistent with expectations,
as the velocity is increased, a vortex appears in the upper corner of the domain and grows
1 1
0.8 0.8
0.6 0.6
y
0.4 0.4
0.2 0.2
0 0
3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8 5 3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8 5
x x
0 0.005 0.01 0.015 0.02 0.025 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035
Figure 4.4 Streamlines and magnitude of velocity contours for u at t = 0.125 and
t = 0.375.
173
1 1
0.8 0.8
0.6 0.6
y
y
0.4 0.4
0.2 0.2
0 0
3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8 5 3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8 5
x x
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0 0.01 0.02 0.03 0.04 0.05 0.06
Figure 4.5 Streamlines and magnitude of velocity contours for u at t = 0.5 and t = 0.75.
1 1
0.8 0.8
0.6 0.6
y
0.4 0.4
0.2 0.2
0 0
3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8 5 3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8 5
x x
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
(a) t = 1 (b) t = 2
174
CHAPTER 5
tion was presented in Chapter 2. The -method analyzed for convection-diffusion was shown
to be second order accurate in time. The discretization of the modeling equations using the
-method allowed for a decoupling of the hyperbolic convective update from the parabolic
diffusion operator update within each time step. The streamline upwind Petrov-Galerkin
In Chapter 3 a fractional step -method was analyzed for slow or creeping viscoelas-
tic fluid flow modeled using a Johnson-Segalman constitutive equation. The -method
presented had the following benefits. The temporal discretization technique was proved to
be second order accurate, and allowed for the decoupling of the velocity and pressure up-
date from the stress update. By decoupling the modeling systems variables, the -method
was that only linear systems of equations are solved at each step in the approximation.
The decoupling used in this analysis is not unique. Future research may show other
splittings of the viscoelastic modeling system also have second order temporal accuracy. The
method analyzed used the splitting parameters and picked in [0,1]. Further study of
these parameters may provide insight into optimal values for these parameters which would
reduce the computation time for the approximation. SUPG was chosen as the stabilization
different stabilization techniques for the presented constitutive model or even the use of
175
other constitutive models may reduce the computational time or improve the accuracy of
the approximation.
drops the inertial term (uu) from the conservation of momentum equation. An extension
of the work presented herein would be to analyze the -method including the inertial term.
The modeling equations of two-phase flow in porous media are a coupled system of
media is challenging because the modeling equations are non-linear, and the character
(parabolic or hyperbolic) of the modeling equations can vary in the problem domain. The
the modeling equations, and a method of handling different model phenomena using spe-
176
APPENDICES
Appendix A
Common Bounds
In Appendix B and C bounds are established for the initial steps in the -method algorithm
applied to viscoelastic flow. For the proof of Theorem 3.5.1 where an estimate for the Stokes
2
Problem is obtained assuming a known true stress function bounds are needed for
E ()
and
2
()
E
. Lemmas B.1 and B.2 establish these estimates. For the constitutive equation the bounds
2
2
given by Lemmas C.1 and C.2 in Appendix C for
F ()
and
F ()
which are used to establish
the error estimates given in Theorem 3.5.2. In this Appendix some lemmas are proved that establish
bounds on common terms occurring in both the previously presented -method analysis, as well as
in Appendix B and Appendix C.
Taylor Expansions: Using repeated integration by parts
Z t
= u( 2 ) +
u() ut (, t) dt (A.1)
t
2
Z t
1 ()
= u( ) + t ut 2 +
2 utt (, t) (t t) dt (A.2)
2 t
2
t
( ) 1 (t)2 ( 2 ) 1
Z
1
= u( 2 ) + t ut 2 +
2
utt + uttt (, t) (t t) dt, (A.3)
2 2 4 2 t
2
and
Z t
= u( 2 )
(0) 2
u ut (, t) dt (A.4)
t0
Z t
1 ()
= u( 2 ) t ut 2 +
2
utt (, t) (t t0 ) dt (A.5)
2 t0
t
( ) 1 (t)2 ( 2 ) 1
Z
1
= u( 2 ) t ut 2 +
2 2
utt uttt (, t) (t t0 ) dt. (A.6)
2 2 4 2 t0
178
2
Z Z t Z t
1 2 1 2 2
= uttt (, t) (t t) dt + uttt (, t) (t t0 ) dt dA
2t t 2t t0
2
2
Z 2 Z t
1 2
2 uttt (, t) (t t) dt
2t t
2
Z t 2 )
2 2
+ uttt (, t) (t t0 ) dt dA
t0
Z 2 Z t Z t
1 2 4
2 (uttt (, t)) dt (t t) dt
2t t t
2 2
Z t Z t
2 2 2 4
+ (uttt (, t)) dt (t t0 ) dt dA
t0 t0
Z 7 Z t
12 3 2
(t) (uttt (, t)) dt dA
25 t0
Z t
2
Ct3 kuttt (, t)k dt.
t0
Proof: Using (A.2) and (A.5) the proof follows analogous to the work shown for the proof of
Lemma A.1.
179
Proof: Note that
Z t
u() u(0) = ut dt
t0
Z t 12 Z t 21
2
1 dt (ut ) dt
t0 t0
Z t 21
1 1
2
= t 2 2 (ut ) dt .
t0
Thus,
2 Z Z t 21 !2
() 1 1
u u(0)
t (ut )2 dt dA
2 2
t0
Z Z t
2
t (ut ) dt dA
t0
Z t
2
t kut k dt
t0
2 2
CK1 t .
()
Lemma A.4 For f C 3 (0, 1) with (0, 1) and given f (0) and f () then f (0) + f () is
an O(t2 ) approximation for f () .
Multiplying (A.10) by t and (A.11) by t and add the resulting equations. Hence
() (0) () (0)
t f 0(0) = f f f + f + O(t3 ),
yielding
!
0(0) 1
f = f () f () + f (0) + O(t2 ). (A.12)
t( ) t( ) t( )
180
Substitute (A.12) into (A.13) giving
! !! !
2
f () = 1 + f (0) f () + O(t2 ),
( ) ( ) ( )
181
Appendix B
in Chapter 3 is used. Note: In order to achieve second order temporal convergence of the -method
() ()
the approximations for uh and uh are found using modifications of the general algorithm steps 1b
and 2a given in (3.29) and (3.30) respectivly. Thus, for the first time step of the -method algorithm
Steps 1b and 2a are described as follows:
Step 1b:
() (0)
uh u h () ()
Re + ph 2(1 ) d(uh )
t
1
() (0)
1
(0)
= h + h + f () + f (0) + 2(1 ) d(uh ), (B.1)
2 2
()
uh = 0. (B.2)
Step 2a:
(1) ()
uh uh (1) (1)
Re + ph 2(1 ) d(uh )
(1 2)t
1
(1) ()
1
()
= h + h + f (1) + f () + 2(1 ) d(uh ), (B.3)
2 2
(1)
uh = 0. (B.4)
() ()
Existance and uniqueness of solutions uh and uh obtained using the modified algorithm steps in
(B.1) and (B.3) follow exactly the proof of Lemmas 3.4.1 and 3.4.3 respectivly.
1
Lemma B.1 Under the assumptions of Theorem 3.5.1 with = = 2 and using (B.1), then
()
2
2
E
and t
d(E ()
2
2
2
Ch2k+2
u()
+ Ch2k+2
u(0)
+ tCh2k
u()
k+1 k+1 k+1
2 Z t Z t
2 2
+tCh2k
u(0)
+ Ct4 k(utt )k dt + Ct5 kuttt k dt
k+1 t0 t0
Z t Z t
2
k tt k dt + tCh2q+2
p( 2 )
2 2
+Ct5 kftt k dt + Ct4 .
t0 t0 q+1
182
Proof: Subtract the variational formulation of (B.1) from (3.13) evaluated at time t to obtain
2
( ) Re () 1
uh uh , v + 2(1 ) d(u( 2 ) ), d(v) 2(1 )
(0) ()
Re ut 2 , v d(uh ), d(v)
t 2
1 1 () 1 (0)
f + f f ( 2 ), v
(0)
= 2(1 ) d(uh ), d(v)
2 2 2
1 () 1 (0)
+ ( 2 ) , d(v) + p( 2 ) , v , v Zh . (B.5)
+
2 2
183
Consider bounding each term on the right hand side of (B.7) as
(0)
2
1
2
E (0) , E () =
E
+ 0
E ()
, (B.8)
40
(1 )t (1 )2 2 t
2
2
d(E (0) ), d(E () ) = (0)
) + t ()
)
, (B.9)
1
Re 4Re2 1
d(E
d(E
()
2
1
2
() , E ()
+ 2
E ()
, (B.10)
42
(1 )t (1 )2 2 t
2
2
d(() ), d(E () )
()
+ 3 t
d(E () )
, (B.11)
Re 4Re 32
(0)
2
1
2
(0) , E ()
+ 4
E ()
, (B.12)
44
(1 )t (1 )2 2 t
2
2
d((0) ), d(E () ) (0)
+ t ()
, (B.13)
5
Re 4Re2 5
d(E
2(1 )t 1 1
d(u() ) + d(u(0) ) d(u( 2 ) ), d(E () )
Re 2 2
2
(1 )2 2 t
2
1 () 1 (0) (
)
()
u + u u 2
+ 6 t
)
, (B.14)
Re2 6 2 2
d(E
()
u u(0)
()
t ut 2 , E ()
t
2
2 t2
()
u u(0) ( 2 )
2
()
u t
+ 7
,
47
t
E
t 1 () 1 (0)
f + f f ( 2 ) , E ()
Re 2 2
2
2 t2
2
1 f () + 1 f (0) f ( 2 )
+ 8
()
, (B.15)
4Re2 8
2 2
E
t 1 () 1 (0)
( 2 ) , d(E () )
Re 2 2
2
2 t
2
1 () 1 (0) ( )
()
+ 2
+ 9 t
)
, (B.16)
4Re2 9
2 2
d(E
t ( 2 ) C 2 2 t
2
2
(2) ( 2 )
p , E () P + t ()
)
. (B.17)
10
Re Re2 10
p
d(E
184
Using (B.8)-(B.17) with (B.7) gives
2 (1 )
2
(1 0 2 4 7 8 )
E ()
+ 1 3 5 6 9 10 t
d(E () )
Re
(0)
2 (1 )2 2 t
2
1
1
()
2 1
(0)
2
E
+
d(E (0) )
+
+
40 4Re 12 42 44
2 2 2 2
(1 ) t
2 (1 ) t
2
+
()
+
(0)
4Re2 3 4Re2 5
2
2
(1 )2 2 t
2 t2
()
u u(0)
1 () 1 (0) (
)
( 2 )
+
2 u + 2 u u
+ 47
ut
2
Re2 6 t
2 2
2 t2
2
1 f () + 1 f (0) f ( 2 )
+ t
1 () + 1 (0) ( 2 )
+ 2 2
4Re 8 2
2
4Re 9 2
2
C 2 2 t
( 2 )
2
+ 2
P ( 2 )
. (B.18)
Re 10
p
2
2
Choose 0 , 2 , 4 , 7 , 9 = 10 1
, and 1 , 3 , 5 , 6 , 9 , 10 = 12Re
. Note that
E (0)
=
d(E (0) )
= 0.
Use lemmas from Appendix A and interpolation properties with (B.18) to obtain
()
2 (1 )
1
2
E
+ t
d(E () )
2 2Re
2
2
Ch2k+2
u()
+ Ch2k+2
u(0)
k+1 k+1
2
2
+tCh2k
u()
+ tCh2k
u(0)
k+1 k+1
Z t Z t
2 2
+Ct4 k(utt )k dt + Ct5 kuttt k dt
t0 t0
Z t Z t
2 2
+Ct5 kftt k dt + Ct4 k tt k dt
t0 t0
2
+tCh2q+2
p( 2 )
. (B.19)
q+1
1
Lemma B.2 Under the assumptions of Theorem 3.5.1 with = = 2 and using (B.3), then
()
2
()
2
()
2
(0)
2
2k+2
E
Ch
u
+
u
+
u
k+1 k+1 k+1
2
2
(0)
2
2k
()
()
+tCh
u
+
u
+
u
k+1 k+1 k+1
Z t Z t
2 2
+Ct4 k(utt )k dt + Ct5 kuttt k dt
t0 t0
Zt Z t
2 2
+Ct5 kftt k dt + Ct4 k tt k dt
t0 t0
2
1
2
+tCh2q+2
p( 2 )
+
p( 2 )
.
q+1 q+1
185
Proof: The proof of Lemma B.2 follows analogously the proof of Lemma B.1, making note that
()
2
2
E
, and t
d(E () )
are bound using (B.19).
186
Appendix C
()
2
2
C h2m+2 + t2 h2m+2 + 2 t2 h2m
()
F
m+1
2
+C h2m+2 + t2 h2m+2 + t2 h2m + 2 t2 h2m + 2 t2 h2m2
(0)
m+1
Z t
2
2 (0)
+Ct4 + Ct4 2 h2 + Ct5 k ttt (, t)k dt + C 2 t2 h2
t
. (C.1)
t0
()
()
(0)
(0)
(0)
e , + e , (0) = e , (1 ) e , (0) u(0) e , (0)
t h t h h
() (0)
(0)
(0)
ga (e , u(0) ), (0) + t ,
h t
(0)
(0) () , (0) + t , u(0) . (C.2)
h
()
Recall that e = () + F () . Let = F () and (C.2) yields
()
2
()
2 t
t
F
+ () , F () +
F
+ () , F ()
t () ()
t
+ F , u F(0)
+ () , u(0) F ()
(0) () (0) ()
= F ,F + ,F
t t
F (0) , F () F (0) , u(0) F ()
t (0) () t (0)
,F , u(0) F ()
t u(0) F (0) , F () t u(0) F (0) , u(0) F ()
t u(0) (0) , F () t u(0) (0) , u(0) F ()
t ga (F (0) , u(0) ), F () t ga (F (0) , u(0) ), u(0) F ()
t ga ((0) , u(0) ), F () t ga ((0) , u(0) ), u(0) F ()
t () (0) , F () t () (0) , u(0) F ()
()
(0)
(0) (0)
+t t , F () + t t , u(0) F () . (C.3)
t
187
Appropriate bounds are now placed on each of the terms in (C.3). Thus,
()
2
1
2
() , F ()
+ 1
F ()
, (C.4)
41
t () () 2 2 (t)2
()
2
()
2
,F + 2
, (C.5)
42 2
F
t ()
F , u(0) F () = 0, (C.6)
t () 1
2 2 2 (t)2
2 K 2 dCh
()
2
2
, u(0) F ()
+ 3
F ()
, (C.7)
43 2
F (0) , F () = 0, (C.8)
(0)
2
1
2
(0) , F ()
+ 4
F ()
, (C.9)
44
t (0) ()
F ,F = 0, (C.10)
t
F (0) , u(0) F () = 0, (C.11)
t (0) () 2 2 (t)2
(0)
2
2
,F
+ 5
F ()
(C.12)
2
4 5
t (0) 1
2 2 (t)2 Ch2
2 K 2 d
(0)
2
()
2
, u(0) F () + 6
, (C.13)
42 6
F
t u(0) F (0) , F () = 0, (C.14)
t u(0) F (0) , u(0) F () = 0, (C.15)
K 2 d
1
2 (t)2
2
2
t u(0) (0) , F ()
(0)
+ 7
F ()
, (C.16)
47
2 d2 K 4 2 (t)2 Ch2
2
2
1
t u(0) (0) , u(0) F ()
(0)
+ 8
F ()
,(C.17)
48
t ga (F (0) , u(0) ), F () = 0, (C.18)
t ga (F (0) , u(0) ), u(0) F () = 0, (C.19)
4K 2 d2 2 (t)2
(0)
2
2
1
t ga ((0) , u(0) ), F ()
+ 9
F ()
, (C.20)
9
t ga ((0) , u(0) ), u(0) F ()
188
2 (t)2 2
2
2
t () (0) , F ()
(0)
+ 13
F ()
()
(C.24)
413
t () (0) , u(0) F ()
1 2 Ch2
2 2 (t)2 2 dK
2
2
(0)
+ 14
F ()
()
(C.25)
414
()
2
2
C h2m+2 + t2 h2m+2 + 2 t2 h2m
()
F
m+1
2
2m+2 2 2m+2 2 2m 2 2 2m
+C h + t h + t h + t h + 2 t2 h2m2
(0)
m+1
Z t
(0)
2
2
+Ct4 + Ct4 2 h2 + Ct5 k ttt (, t)k dt + C 2 t2 h2
t
. (C.29)
t0
189
Lemma C.2 Under the assumptions of Theorem 3.5.1,
()
2
F
S(h, t, ), (C.30)
where
2
S(h, t, ) = C h2m+2 + t2 h2m+2 + 2 t2 h2m
()
m+1
2
+C h2m+2 + t2 h2m+2 + t2 h2m + 2 t2 h2m + 2 t2 h2m2
(0)
m+1
Z t
2
+Ct4 + Ct4 2 h2 + Ct5 k ttt (, t)k dt
t0
2
(0)
+C 2 t2 h2
t
(1 + t2 + t)
2
+C h2m+2 + t2 h2m+2 + th2m+2
()
m+1
2
+C h2m+2 + t2 h2m+2 + th2m+2 + t2 h2m + th2m
()
m+1
Z t
2
+C(t4 + 2 t4 h2 + Ct5 k ttt (, t)k dt + C 2 t2 .
t
()
()
()
()
e , + e , () + u e , () + ga (e , u() ), ()
(1 2)t h h h
!
() ()
()
()
()
= e , e , () + t ,
(1 2)t h (1 2) t
()
+ t , u() () () , () . (C.31)
h
190
Choose = F () . Then from (C.31)
()
2
F
+ () , F ()
(1 2) t
()
2 (1 2) t
+
F
+ F () , u() F ()
(1 2) t () () (1 2) t ()
+ ,F + , u() F ()
2
+ (1 2) t u F () , F () + (1 2) t
u F ()
+ (1 2) t u () , F () + (1 2) t u () , u() F ()
+ (1 2) t ga (F () , u() ), F () + (1 2) t ga (F () , u() ), u() F ()
+ (1 2) t ga (() , u() ), F () + (1 2) t ga (() , u() ), u() F ()
= F () , F () + () , F ()
(1 2) t () () (1 2) t ()
F ,F F , u() F ()
(1 2) t () () (1 2) t ()
,F , u() F ()
!
() ()
()
()
+ (1 2) t t , F () + (1 2) t t , u() F ()
(1 2) t
(1 2) t ()
() , F () + u() F () . (C.32)
()
2
1
2
() , F ()
+ 1
F ()
, (C.33)
41
(1 2) t ()
F , u() F () = 0, (C.34)
2
(1 2) t () () 2 (1 2) t2
()
2
2
,F
+ 2
F ()
, (C.35)
2
42
(1 2) t () 2 (1 2) t
()
2
, u() F ()
2 43
2
+ 3 (1 2) t
u() F ()
, (C.36)
(1 2) t u F () , F () = 0, (C.37)
(1 2)2 t2 K 2 d
2
2
1
(1 2) t u () , F ()
()
+ 4
F ()
, (C.38)
44
(1 2) tK 2 d
2
1
(1 2) t u () , u() F ()
()
45
2
+5 (1 2) t
u() F ()
, (C.39)
191
2
(1 2) t ga (F () , u() ), F () 4 (1 2) tK1 d
F ()
, (C.40)
(1 2) t ga (F () , u() ), u() F ()
2 4 2 (1 2)2 t2 K 2 d2
2
1
6
F ()
+
u F ()
,
()
(C.41)
6
4 (1 2)2 t2 K 2 d2
()
2
2
1
(1 2) t ga (() , u() ), F ()
+ 7
F ()
, (C.42)
7
4 (1 2) tK 2 d2
()
2
1
(1 2) t ga (() , u() ), u() F ()
8
2
+ (1 2) t8
u() F ()
, (C.43)
()
2
1
2
F () , F ()
F
+ 9
F ()
, (C.44)
49
1
()
2
2
() , F ()
+ 10
F ()
, (C.45)
410
2
(1 2) t () () 2 (1 2) t2
()
2
()
2
F ,F + 11
, (C.46)
42 11
F
F
(1 2) t () 2 (1 2) t
()
2
F , u() F ()
2 412
F
2
+ (1 2) t12
u() F ()
, (C.47)
2
(1 2) t () () 2 (1 2) t2
()
2
2
,F
+ 13
F ()
, (C.48)
2
413
(1 2) t () 2 (1 2) t
()
2
, u() F ()
2 414
2
+14 (1 2) t
u() F ()
, (C.49)
!
2
2
() () () () (1 2) t2
() ()
()
(1 2) t t , F t
(1 2) t 415
(1 2) t
2
+15
F ()
. (C.50)
Using u = 0, and u = 0,
() ()
(1 2) t t , u() F () = (1 2) t u() , t : F ()
()
(1 2) t u() t , F ()
()
= (1 2) t u() t , F ()
2
(1 2) 2 t2
2
()
2
u t
+ 16
F ()
. (C.51)
()
416
192
Lastly
(1 2) t () 2 (1 2)2 t2
2
()
2
() , F ()
() ()
+ 17
, (C.52)
42 17
F
and
2 1 2 Ch2
(1 2) t () 2 2 (1 2) t2 dK
2
() , u() F ()
() ()
42 18
2
+ 18
F ()
. (C.53)
!
4 (1 2) tK1 2 d2
6
!
2
2 1 2 (1 2) t2 2 (1 2) t
F ()
+ +
49 42 11 2 412
!
2
()
2
1 2 (1 2) t2 2 (1 2) t
+
+ +
410 2 413 2 414
2
2 1 2 (1 2) t2 2 (1 2) t
+
()
+ +
41 2 42 2 43
!
2
4 (1 2) t2 K1 2 d2 4 (1 2) tK1 2 d2
+ +
7 8
!
2 (1 2)2 t2 K 2 d (1 2) tK 2 d
1 1
+
()
+
44 45
!
2 2 (1 2)2 t2 2
2 2 (1 2) t2 dK 1 2 Ch2
() ()
+
+
42 17 42 18
2
2 2
(1 2) t2
() ()
()
(1 2) 2 t2
()
2
()
+ t
+
u t
. (C.54)
415
(1 2) t 416
Choosing i (i {1, . . . , 18}) and t sufficiently small the terms on the left hand side of (C.54)
remain positive. Lemma A.1 yields
2
2 Z t
() ()
(1 2) t2
()
5 2
t
Ct k ttt (, t)k dt
415
(1 2) t
t
193
and making note that < K1 , and using Lemma A.3 gives
!
2 2 1 2 Ch2
2 (1 2) t2 2 2 (1 2) t2 dK
2
C(t4 + 2 t4 h2 ).
() ()
+
42 17 42 18
Thus, using interpolation properties, approximation properties of the stress space Sh establishes
()
2
F
S(h, t, ).
194
BIBLIOGRAPHY
1. R.A. Adams. Sobolev spaces. Academic Press, New York-London, 1975. Pure and
Applied Mathematics, Vol. 65.
2. U.M. Ascher, S.J. Ruuth, and B.T.R. Wetton. Implicit-explicit methods for time-
dependent partial differential equations. SIAM J. Numer. Anal., 32(3):797823,
1995.
6. M. Bensaada and D. Esselaoui. Error estimates for a stabilized finite element method
for the Oldroyd B model. J. Math. Anal. Appl., 325(2):10421059, 2007.
8. R.B. Bird, R.C. Armstrong, and O. Hassager. Dynamics of Polymeric Liquids. Wiley-
Interscience, 1987.
10. D. Braess. Finite Elements. Cambridge University Press, Cambridge, second edition,
2001.
11. S.C. Brenner and L.R. Scott. The Mathematical Theory of Finite Element Methods.
Springer-Verlag, New York, second edition, 2002.
12. J.C. Chrispell, V.J. Ervin, and E.W. Jenkins. A fractional step -method for convection-
diffusion using a SUPG approximation. Technical Report TR2006_11_CEJ,
Clemson University, 2006.
13. J.C. Chrispell, V.J. Ervin, and E.W. Jenkins. A fractional step -method for convection-
diffusion problems. J. Math. Anal. Appl., 333(1):204218, 2007.
195
14. J.C. Chrispell, V.J. Ervin, and E.W. Jenkins. A fractional step -method for viscoelastic
fluid flow using a supg approximation. International Journal of Computational
Science, Accepted 2008.
15. C. Dawson. Godunov-mixed methods for advection-diffusion equations in
multidimensions. SIAM J. Numer. Anal., 30(5):13151332, 1993.
16. C.N. Dawson and M.F. Wheeler. Time-splitting methods for advection-diffusion-
reaction equations arising in contaminant transport. In ICIAM 91 (Washington,
DC, 1991), pages 7182. SIAM, Philadelphia, PA, 1992.
17. J. Douglas, Jr. Alternating direction methods for three space variables. Numer. Math.,
4:4163, 1962.
18. A. Ern and J. Guermond. Theory and Practice of Finite Elements. Springer-Verlag,
New York, 2004.
19. V.J. Ervin and N. Heuer. Approximation of time-dependent, viscoelastic fluid
flow: Crank-Nicolson, finite element approximation. Numer. Methods Partial
Differential Equations, 20(2):248283, 2004.
20. V.J. Ervin and W.J. Layton. A robust and parallel relaxation method based on algebraic
splittings. Numer. Methods Partial Differential Equations, 15(1):91110, 1999.
21. V.J. Ervin and W.W. Miles. Approximation of time-dependent viscoelastic fluid flow:
SUPG approximation. SIAM J. Numer. Anal., 41(2):457486 (electronic), 2003.
22. M.S. Espedal and K.H. Karlsen. Numerical solution of reservoir flow models based on
large time step operator splitting algorithms. In Filtration in Porous Media and
Industrial Application (Cetraro, 1998), volume 1734 of Lecture Notes in Math.,
pages 977. Springer, Berlin, 2000.
23. S. Evje and K.H. Karlsen. Viscous splitting approximation of mixed hyperbolic-
parabolic convection-diffusion equations. Numer. Math., 83(1):107137, 1999.
24. E. Fernandez-Cara, F. Guillen, and R.R. Ortega. Mathematical modeling and analysis
of viscoelastic fluids of the Oldroyd kind. In Handbook of Numerical Analysis,
Vol. VIII, pages 543661. North-Holland, Amsterdam, 2002.
25. H. Giesekus. A unified approach to a variety of constitutive models for polymer fluids
based on the concept of configuration dependent molecular mobility. Rheol.
Acta, 21, 1982.
26. V. Girault and P.A. Raviart. Finite Element Methods for Navier-Stokes Equations.
Springer-Verlag, Berlin, 1986.
27. R. Glowinski and J.F. Periaux. Numerical methods for nonlinear problem in fluid
dynamics. In Supercomputing, pages 381479. North-Holland, Amsterdam, 1987.
28. C. Guillope and J.-C. Saut. Existence results for the flow of viscoelastic fluids with a
differential constitutive law. Nonlinear Anal., 15(9):849869, 1990.
196
29. M.D. Gunzburger. Finite Element Methods for Viscous Incompressible Flows. Academic
Press Inc., Boston, MA, 1989.
33. V. John. Large Eddy Simulation of Turbulent Incompressible Flows, volume 34.
Springer-Verlag, Berlin, 2004. Analytical and numerical results for a class of
LES models.
34. C. Johnson. Numerical Solution of Partial Differential Equations by the Finite Element
Method. Cambridge University Press, Cambridge, 1987.
35. C. Johnson, U. Navert, and J. Pitkaranta. Finite element methods for linear hyperbolic
problems. Comput. Methods Appl. Mech. Engrg., 45(1-3):285312, 1984.
36. M.W. Johnson and D. Segalman. A model for viscoelastic fluid behavior which allows
non-affine deformation. J. Non-Newtonian Fluid Mech., 2:255270, 1977.
37. C.T. Kelley. Iterative Methods for Linear and Nonlinear Equations. Society for
Industrial and Applied Mathematics (SIAM), Philadelphia, PA, 1995.
38. L.A. Khan and P.L.-F. Liu. Numerical analyses of operator-splitting algorithms for the
two-dimensional advection-diffusion equation. Comput. Methods Appl. Mech.
Engrg., 152(3-4):337359, 1998.
39. P. Kloucek and F.S. Rys. Stability of the fractional step -scheme for the nonstationary
Navier-Stokes equations. SIAM J. Numer. Anal., 31(5):13121335, 1994.
41. J.M. Marchal and M.J. Crochet. Hermitian finite elements for calculating viscoelastic
flow. J. Non-Newtonian Fluid Mech., 20:187207, 1986.
42. J.M. Marchal and M.J. Crochet. A new mixed finite element for calculation of
viscoelastic flow. J. Non-Newtonian Fluid Mech., 26:77114, 1987.
43. G.I. Marchuk. Splitting and alternating direction methods. In Handbook of Numerical
Analysis, Vol. I, pages 197462. North-Holland, Amsterdam, 1990.
197
45. J. Neustupa and P. Penel, editors. Mathematical Fluid Mechanics: Recent Results and
Open Questions. Birkhauser, 2001.
46. J.G. Oldroyd. On the formulation of rheological equations of state. Proc. Roy. Soc.
London, A200:523541, 1950.
47. J.G. Oldroyd. Non-Newtonian effects in steady motion of some idealized elastico-viscous
liquids. Proc. Roy. Soc. London, A245:278297, 1958.
48. R.G. Owens and T.N. Phillips. Computational Rheology. Imperial College Press,
London, 2002.
49. D.W. Peaceman and H.H. Rachford, Jr. The numerical solution of parabolic and elliptic
differential equations. J. Soc. Indust. Appl. Math., 3:2841, 1955.
50. N. Phan-Thien and R.I. Tanner. A new constitutive equation derived from network
theory. J. Non-Newtonian Fluid Mech., 2:353365, 1977.
51. T.N. Phillips and A.J. Williams. Viscoelastic flow through a planar contraction using a
semi-Lagrangian finite volume method. J. Non-Newtonian Fluid Mech., 87:215
246, 1999.
52. M. Renardy. Mathematical Analysis of Viscoelastic Flows, volume 73. Society for
Industrial and Applied Mathematics (SIAM), Philadelphia, PA, 2000.
53. R.H. Sabersky and A.J. Acosta. Fluid Flow: A First Course in Fluid Mechanics. The
Macmillan Company, 1964.
54. D. Sandri. Finite element approximation of viscoelastic fluid flow: Existence of
approximate solutions and error bounds. Continuous approximation of the
stress. SIAM J. Numer. Anal., 31(2):362377, 1994.
55. P. Saramito. A new -scheme algorithm and incompressible FEM for viscoelastic fluid
flows. RAIRO Model. Math. Anal. Numer., 28(1):135, 1994.
56. P. Saramito. Efficient simulation of nonlinear viscoelastic fluid flows. J. Non-Newt.
Fluid Mech., 60:199223, 1995.
57. J. Sun, M.D. Smith, R.C. Armstrong, and R.A. Brown. Finite element method for the
viscoelastic flows based on the discrete adaptive viscoelastic stress splitting and
the discontinuous Galerkin method: DAVSS-G/DG. J. Non-Newt. Fluid Mech.,
86:281307, 1999.
58. R. Sureshkumar, M.D. Smith, R.C. Armstrong, and R.A. Brown. Linear stability and
dynamics of viscoelastic flows using time-dependent numerical simulations. J.
Non-Newt. Fluid Mech., 82:57104, 1999.
59. S. Turek. A comparative study of time-stepping techniques for the incompressible
Navier-Stokes equations: From fully implicit non-linear schemes to semi-implicit
projection methods. Internat. J. Numer. Methods Fluids, 22(10):9871011,
1996.
198
60. S. Turek. Efficient Solvers for Incompressible Flow Problems, volume 6. Springer-Verlag,
Berlin, 1999.
61. H. Wang, H.K. Dahle, R.E. Ewing, M.S. Espedal, R.C. Sharpley, and S. Man. An
ELLAM scheme for advection-diffusion equations in two dimensions. SIAM J.
Sci. Comput., 20(6):21602194 (electronic), 1999.
62. J.L. White and A.B. Metzner. Development of constitutive equations for polymeric
melts and solutions. Journal of Applied Polymer Science, 7:18671889, 1963.
199