This action might not be possible to undo. Are you sure you want to continue?
Harvard Lomax and Thomas H. Pulliam NASA Ames Research Center David W. Zingg University of Toronto Institute for Aerospace Studies August 26, 1999
Contents
1 INTRODUCTION
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.1 Problem Speci cation and Geometry Preparation . . . . . . 1.2.2 Selection of Governing Equations and Boundary Conditions 1.2.3 Selection of Gridding Strategy and Numerical Method . . . 1.2.4 Assessment and Interpretation of Results . . . . . . . . . . . 1.3 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4 Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Conservation Laws . . . . . . . . . . . . 2.2 The NavierStokes and Euler Equations . 2.3 The Linear Convection Equation . . . . 2.3.1 Di erential Form . . . . . . . . . 2.3.2 Solution in Wave Space . . . . . . 2.4 The Di usion Equation . . . . . . . . . . 2.4.1 Di erential Form . . . . . . . . . 2.4.2 Solution in Wave Space . . . . . . 2.5 Linear Hyperbolic Systems . . . . . . . . 2.6 Problems . . . . . . . . . . . . . . . . . . 3.1 Meshes and FiniteDi erence Notation 3.2 Space Derivative Approximations . . . 3.3 FiniteDi erence Operators . . . . . . 3.3.1 Point Di erence Operators . . . 3.3.2 Matrix Di erence Operators . . 3.3.3 Periodic Matrices . . . . . . . . 3.3.4 Circulant Matrices . . . . . . . iii . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1
1 2 2 3 3 4 4 5
2 CONSERVATION LAWS AND THE MODEL EQUATIONS
7 8 12 12 13 14 14 15 16 17
7
3 FINITEDIFFERENCE APPROXIMATIONS
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
21
21 24 25 25 25 29 30
3.4 Constructing Di erencing Schemes of Any Order . . . . . 3.4.1 Taylor Tables . . . . . . . . . . . . . . . . . . . . 3.4.2 Generalization of Di erence Formulas . . . . . . . 3.4.3 Lagrange and Hermite Interpolation Polynomials 3.4.4 Practical Application of Pade Formulas . . . . . . 3.4.5 Other HigherOrder Schemes . . . . . . . . . . . . 3.5 Fourier Error Analysis . . . . . . . . . . . . . . . . . . . 3.5.1 Application to a Spatial Operator . . . . . . . . . 3.6 Di erence Operators at Boundaries . . . . . . . . . . . . 3.6.1 The Linear Convection Equation . . . . . . . . . 3.6.2 The Di usion Equation . . . . . . . . . . . . . . . 3.7 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
31 31 34 35 37 38 39 39 43 44 46 47
4 THE SEMIDISCRETE APPROACH
4.1 Reduction of PDE's to ODE's . . . . . . . . . . . . . . . . . . . . . . 4.1.1 The Model ODE's . . . . . . . . . . . . . . . . . . . . . . . . 4.1.2 The Generic Matrix Form . . . . . . . . . . . . . . . . . . . . 4.2 Exact Solutions of Linear ODE's . . . . . . . . . . . . . . . . . . . . 4.2.1 Eigensystems of SemiDiscrete Linear Forms . . . . . . . . . . 4.2.2 Single ODE's of First and SecondOrder . . . . . . . . . . . . 4.2.3 Coupled FirstOrder ODE's . . . . . . . . . . . . . . . . . . . 4.2.4 General Solution of Coupled ODE's with Complete Eigensystems 4.3 Real Space and Eigenspace . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 De nition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2 Eigenvalue Spectrums for Model ODE's . . . . . . . . . . . . . 4.3.3 Eigenvectors of the Model Equations . . . . . . . . . . . . . . 4.3.4 Solutions of the Model ODE's . . . . . . . . . . . . . . . . . . 4.4 The Representative Equation . . . . . . . . . . . . . . . . . . . . . . 4.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 Basic Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Model Equations in Integral Form . . . . . . . . . . . . . . . . . . . 5.2.1 The Linear Convection Equation . . . . . . . . . . . . . . . 5.2.2 The Di usion Equation . . . . . . . . . . . . . . . . . . . . . 5.3 OneDimensional Examples . . . . . . . . . . . . . . . . . . . . . . 5.3.1 A SecondOrder Approximation to the Convection Equation 5.3.2 A FourthOrder Approximation to the Convection Equation 5.3.3 A SecondOrder Approximation to the Di usion Equation . 5.4 A TwoDimensional Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
51
52 52 53 54 54 55 57 59 61 61 62 63 65 67 68 72 73 73 74 74 75 77 78 80
5 FINITEVOLUME METHODS
71
5.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
83 86 87 88 89 90 91 91 92 93 93 95 95 96 97 97 98 99 100 101 102 102 103 105 106 107 110 110 111 112 115 117
6 TIMEMARCHING METHODS FOR ODE'S
6.1 Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Converting TimeMarching Methods to O E's . . . . . . . . . . . . 6.3 Solution of Linear O E's With Constant Coe cients . . . . . . . . 6.3.1 First and SecondOrder Di erence Equations . . . . . . . . 6.3.2 Special Cases of Coupled FirstOrder Equations . . . . . . . 6.4 Solution of the Representative O E's . . . . . . . . . . . . . . . . 6.4.1 The Operational Form and its Solution . . . . . . . . . . . . 6.4.2 Examples of Solutions to TimeMarching O E's . . . . . . . 6.5 The ; Relation . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.5.1 Establishing the Relation . . . . . . . . . . . . . . . . . . . . 6.5.2 The Principal Root . . . . . . . . . . . . . . . . . . . . . . 6.5.3 Spurious Roots . . . . . . . . . . . . . . . . . . . . . . . . 6.5.4 OneRoot TimeMarching Methods . . . . . . . . . . . . . . 6.6 Accuracy Measures of TimeMarching Methods . . . . . . . . . . . 6.6.1 Local and Global Error Measures . . . . . . . . . . . . . . . 6.6.2 Local Accuracy of the Transient Solution (er j j er! ) . . . 6.6.3 Local Accuracy of the Particular Solution (er ) . . . . . . . 6.6.4 Time Accuracy For Nonlinear Applications . . . . . . . . . . 6.6.5 Global Accuracy . . . . . . . . . . . . . . . . . . . . . . . . 6.7 Linear Multistep Methods . . . . . . . . . . . . . . . . . . . . . . . 6.7.1 The General Formulation . . . . . . . . . . . . . . . . . . . . 6.7.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.7.3 TwoStep Linear Multistep Methods . . . . . . . . . . . . . 6.8 PredictorCorrector Methods . . . . . . . . . . . . . . . . . . . . . . 6.9 RungeKutta Methods . . . . . . . . . . . . . . . . . . . . . . . . . 6.10 Implementation of Implicit Methods . . . . . . . . . . . . . . . . . . 6.10.1 Application to Systems of Equations . . . . . . . . . . . . . 6.10.2 Application to Nonlinear Equations . . . . . . . . . . . . . . 6.10.3 Local Linearization for Scalar Equations . . . . . . . . . . . 6.10.4 Local Linearization for Coupled Sets of Nonlinear Equations 6.11 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1 Dependence on the Eigensystem 7.2 Inherent Stability of ODE's . . 7.2.1 The Criterion . . . . . . 7.2.2 Complete Eigensystems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
85
7 STABILITY OF LINEAR SYSTEMS
121
122 123 123 123
7.3
7.4 7.5 7.6
7.7
7.8 7.9
7.2.3 Defective Eigensystems . . . . . . . . . . . . . . Numerical Stability of O E 's . . . . . . . . . . . . . . 7.3.1 The Criterion . . . . . . . . . . . . . . . . . . . 7.3.2 Complete Eigensystems . . . . . . . . . . . . . . 7.3.3 Defective Eigensystems . . . . . . . . . . . . . . TimeSpace Stability and Convergence of O E's . . . . Numerical Stability Concepts in the Complex Plane . 7.5.1 Root Traces Relative to the Unit Circle . . . 7.5.2 Stability for Small t . . . . . . . . . . . . . . . Numerical Stability Concepts in the Complex h Plane 7.6.1 Stability for Large h. . . . . . . . . . . . . . . . 7.6.2 Unconditional Stability, AStable Methods . . . 7.6.3 Stability Contours in the Complex h Plane. . . Fourier Stability Analysis . . . . . . . . . . . . . . . . 7.7.1 The Basic Procedure . . . . . . . . . . . . . . . 7.7.2 Some Examples . . . . . . . . . . . . . . . . . . 7.7.3 Relation to Circulant Matrices . . . . . . . . . . Consistency . . . . . . . . . . . . . . . . . . . . . . . . Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
123 124 124 125 125 125 128 128 132 135 135 136 137 141 141 142 143 143 146
8 CHOICE OF TIMEMARCHING METHODS
8.1 Sti ness De nition for ODE's . . . . . . . . . . . . 8.1.1 Relation to Eigenvalues . . . . . . . . . . 8.1.2 Driving and Parasitic Eigenvalues . . . . . . 8.1.3 Sti ness Classi cations . . . . . . . . . . . . 8.2 Relation of Sti ness to Space Mesh Size . . . . . . 8.3 Practical Considerations for Comparing Methods . 8.4 Comparing the E ciency of Explicit Methods . . . 8.4.1 Imposed Constraints . . . . . . . . . . . . . 8.4.2 An Example Involving Di usion . . . . . . . 8.4.3 An Example Involving Periodic Convection . 8.5 Coping With Sti ness . . . . . . . . . . . . . . . . 8.5.1 Explicit Methods . . . . . . . . . . . . . . . 8.5.2 Implicit Methods . . . . . . . . . . . . . . . 8.5.3 A Perspective . . . . . . . . . . . . . . . . . 8.6 Steady Problems . . . . . . . . . . . . . . . . . . . 8.7 Problems . . . . . . . . . . . . . . . . . . . . . . . .
149
149 149 151 151 152 153 154 154 154 155 158 158 159 160 160 161
9 RELAXATION METHODS
9.1 Formulation of the Model Problem . . . . . . . . . . . . . . 9.1.1 Preconditioning the Basic Matrix . . . . . . . . . . . 9.1.2 The Model Equations . . . . . . . . . . . . . . . . . . 9.2 Classical Relaxation . . . . . . . . . . . . . . . . . . . . . . 9.2.1 The Delta Form of an Iterative Scheme . . . . . . . . 9.2.2 The Converged Solution, the Residual, and the Error 9.2.3 The Classical Methods . . . . . . . . . . . . . . . . . 9.3 The ODE Approach to Classical Relaxation . . . . . . . . . 9.3.1 The Ordinary Di erential Equation Formulation . . . 9.3.2 ODE Form of the Classical Methods . . . . . . . . . 9.4 Eigensystems of the Classical Methods . . . . . . . . . . . . 9.4.1 The PointJacobi System . . . . . . . . . . . . . . . . 9.4.2 The GaussSeidel System . . . . . . . . . . . . . . . . 9.4.3 The SOR System . . . . . . . . . . . . . . . . . . . . 9.5 Nonstationary Processes . . . . . . . . . . . . . . . . . . . . 9.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
163
164 164 166 168 168 168 169 170 170 172 173 174 176 180 182 187
10 MULTIGRID
10.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191 10.1.1 Eigenvector and Eigenvalue Identi cation with Space Frequencies191 10.1.2 Properties of the Iterative Method . . . . . . . . . . . . . . . 192 10.2 The Basic Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192 10.3 A TwoGrid Process . . . . . . . . . . . . . . . . . . . . . . . . . . . 200 10.4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202 11.1 11.2 11.3 11.4 OneSided FirstDerivative Space Di erencing The Modi ed Partial Di erential Equation . . The LaxWendro Method . . . . . . . . . . . Upwind Schemes . . . . . . . . . . . . . . . . 11.4.1 FluxVector Splitting . . . . . . . . . . 11.4.2 FluxDi erence Splitting . . . . . . . . 11.5 Arti cial Dissipation . . . . . . . . . . . . . . 11.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
191
11 NUMERICAL DISSIPATION
203
204 205 207 209 210 212 213 214
12 SPLIT AND FACTORED FORMS
12.1 The Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217 12.2 Factoring Physical Representations  Time Splitting . . . . . . . . . 218 12.3 Factoring Space Matrix Operators in 2{D . . . . . . . . . . . . . . . . 220
217
12.4 12.5 12.6 12.7
12.3.1 Mesh Indexing Convention . . . . . . . . . . . 12.3.2 Data Bases and Space Vectors . . . . . . . . . 12.3.3 Data Base Permutations . . . . . . . . . . . . 12.3.4 Space Splitting and Factoring . . . . . . . . . SecondOrder Factored Implicit Methods . . . . . . . Importance of Factored Forms in 2 and 3 Dimensions The Delta Form . . . . . . . . . . . . . . . . . . . . . Problems . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . .
220 221 221 223 226 226 228 229
13 LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS
13.1 The Representative Equation for Circulant Operators . . . . . . . 13.2 Example Analysis of Circulant Systems . . . . . . . . . . . . . . . 13.2.1 Stability Comparisons of TimeSplit Methods . . . . . . . 13.2.2 Analysis of a SecondOrder TimeSplit Method . . . . . . 13.3 The Representative Equation for SpaceSplit Operators . . . . . . 13.4 Example Analysis of 2D Model Equations . . . . . . . . . . . . . 13.4.1 The Unfactored Implicit Euler Method . . . . . . . . . . . 13.4.2 The Factored Nondelta Form of the Implicit Euler Method 13.4.3 The Factored Delta Form of the Implicit Euler Method . . 13.4.4 The Factored Delta Form of the Trapezoidal Method . . . 13.5 Example Analysis of the 3D Model Equation . . . . . . . . . . . 13.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.1 A.2 A.3 A.4 A.5 B.1 B.2 B.3 B.4 Notation . . . . . . . . . . De nitions . . . . . . . . . Algebra . . . . . . . . . . Eigensystems . . . . . . . Vector and Matrix Norms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
233
233 234 234 237 238 242 242 243 243 244 245 247 249 250 251 251 254 257 258 259 260 260 261 262 263
A USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA 249
B SOME PROPERTIES OF TRIDIAGONAL MATRICES
Standard Eigensystem for Simple Tridiagonals . . Generalized Eigensystem for Simple Tridiagonals . The Inverse of a Simple Tridiagonal . . . . . . . . Eigensystems of Circulant Matrices . . . . . . . . B.4.1 Standard Tridiagonals . . . . . . . . . . . B.4.2 General Circulant Systems . . . . . . . . . B.5 Special Cases Found From Symmetries . . . . . . B.6 Special Cases Involving Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
257
C THE HOMOGENEOUS PROPERTY OF THE EULER EQUATIONS265
Chapter 1 INTRODUCTION
1.1 Motivation
The material in this book originated from attempts to understand and systemize numerical solution techniques for the partial di erential equations governing the physics of uid ow. As time went on and these attempts began to crystallize, underlying constraints on the nature of the material began to form. The principal such constraint was the demand for uni cation. Was there one mathematical structure which could be used to describe the behavior and results of most numerical methods in common use in the eld of uid dynamics? Perhaps the answer is arguable, but the authors believe the answer is a rmative and present this book as justi cation for that belief. The mathematical structure is the theory of linear algebra and the attendant eigenanalysis of linear systems. The ultimate goal of the eld of computational uid dynamics (CFD) is to understand the physical events that occur in the ow of uids around and within designated objects. These events are related to the action and interaction of phenomena such as dissipation, di usion, convection, shock waves, slip surfaces, boundary layers, and turbulence. In the eld of aerodynamics, all of these phenomena are governed by the compressible NavierStokes equations. Many of the most important aspects of these relations are nonlinear and, as a consequence, often have no analytic solution. This, of course, motivates the numerical solution of the associated partial di erential equations. At the same time it would seem to invalidate the use of linear algebra for the classi cation of the numerical methods. Experience has shown that such is not the case. As we shall see in a later chapter, the use of numerical methods to solve partial di erential equations introduces an approximation that, in e ect, can change the form of the basic partial di erential equations themselves. The new equations, which 1
2
CHAPTER 1. INTRODUCTION
are the ones actually being solved by the numerical process, are often referred to as the modi ed partial di erential equations. Since they are not precisely the same as the original equations, they can, and probably will, simulate the physical phenomena listed above in ways that are not exactly the same as an exact solution to the basic partial di erential equation. Mathematically, these di erences are usually referred to as truncation errors. However, the theory associated with the numerical analysis of uid mechanics was developed predominantly by scientists deeply interested in the physics of uid ow and, as a consequence, these errors are often identi ed with a particular physical phenomenon on which they have a strong e ect. Thus methods are said to have a lot of \arti cial viscosity" or said to be highly dispersive. This means that the errors caused by the numerical approximation result in a modi ed partial di erential equation having additional terms that can be identi ed with the physics of dissipation in the rst case and dispersion in the second. There is nothing wrong, of course, with identifying an error with a physical process, nor with deliberately directing an error to a speci c physical process, as long as the error remains in some engineering sense \small". It is safe to say, for example, that most numerical methods in practical use for solving the nondissipative Euler equations create a modi ed partial di erential equation that produces some form of dissipation. However, if used and interpreted properly, these methods give very useful information. Regardless of what the numerical errors are called, if their e ects are not thoroughly understood and controlled, they can lead to serious di culties, producing answers that represent little, if any, physical reality. This motivates studying the concepts of stability, convergence, and consistency. On the other hand, even if the errors are kept small enough that they can be neglected (for engineering purposes), the resulting simulation can still be of little practical use if ine cient or inappropriate algorithms are used. This motivates studying the concepts of sti ness, factorization, and algorithm development in general. All of these concepts we hope to clarify in this book.
1.2 Background
The eld of computational uid dynamics has a broad range of applicability. Independent of the speci c application under study, the following sequence of steps generally must be followed in order to obtain a satisfactory solution.
1.2.1 Problem Speci cation and Geometry Preparation
The rst step involves the speci cation of the problem, including the geometry, ow conditions, and the requirements of the simulation. The geometry may result from
1.2. BACKGROUND
3
measurements of an existing con guration or may be associated with a design study. Alternatively, in a design context, no geometry need be supplied. Instead, a set of objectives and constraints must be speci ed. Flow conditions might include, for example, the Reynolds number and Mach number for the ow over an airfoil. The requirements of the simulation include issues such as the level of accuracy needed, the turnaround time required, and the solution parameters of interest. The rst two of these requirements are often in con ict and compromise is necessary. As an example of solution parameters of interest in computing the ow eld about an airfoil, one may be interested in i) the lift and pitching moment only, ii) the drag as well as the lift and pitching moment, or iii) the details of the ow at some speci c location.
Once the problem has been speci ed, an appropriate set of governing equations and boundary conditions must be selected. It is generally accepted that the phenomena of importance to the eld of continuum uid dynamics are governed by the conservation of mass, momentum, and energy. The partial di erential equations resulting from these conservation laws are referred to as the NavierStokes equations. However, in the interest of e ciency, it is always prudent to consider solving simpli ed forms of the NavierStokes equations when the simpli cations retain the physics which are essential to the goals of the simulation. Possible simpli ed governing equations include the potential ow equations, the Euler equations, and the thinlayer NavierStokes equations. These may be steady or unsteady and compressible or incompressible. Boundary types which may be encountered include solid walls, in ow and out ow boundaries, periodic boundaries, symmetry boundaries, etc. The boundary conditions which must be speci ed depend upon the governing equations. For example, at a solid wall, the Euler equations require ow tangency to be enforced, while the NavierStokes equations require the noslip condition. If necessary, physical models must be chosen for processes which cannot be simulated within the speci ed constraints. Turbulence is an example of a physical process which is rarely simulated in a practical context (at the time of writing) and thus is often modelled. The success of a simulation depends greatly on the engineering insight involved in selecting the governing equations and physical models based on the problem speci cation.
1.2.2 Selection of Governing Equations and Boundary Conditions
1.2.3 Selection of Gridding Strategy and Numerical Method
Next a numerical method and a strategy for dividing the ow domain into cells, or elements, must be selected. We concern ourselves here only with numerical methods requiring such a tessellation of the domain, which is known as a grid, or mesh.
4
CHAPTER 1. INTRODUCTION
Many di erent gridding strategies exist, including structured, unstructured, hybrid, composite, and overlapping grids. Furthermore, the grid can be altered based on the solution in an approach known as solutionadaptive gridding. The numerical methods generally used in CFD can be classi ed as nitedi erence, nitevolume, niteelement, or spectral methods. The choices of a numerical method and a gridding strategy are strongly interdependent. For example, the use of nitedi erence methods is typically restricted to structured grids. Here again, the success of a simulation can depend on appropriate choices for the problem or class of problems of interest.
1.2.4 Assessment and Interpretation of Results
Finally, the results of the simulation must be assessed and interpreted. This step can require postprocessing of the data, for example calculation of forces and moments, and can be aided by sophisticated ow visualization tools and error estimation techniques. It is critical that the magnitude of both numerical and physicalmodel errors be well understood.
1.3 Overview
It should be clear that successful simulation of uid ows can involve a wide range of issues from grid generation to turbulence modelling to the applicability of various simpli ed forms of the NavierStokes equations. Many of these issues are not addressed in this book. Some of them are presented in the books by Anderson, Tannehill, and Pletcher 1] and Hirsch 2]. Instead we focus on numerical methods, with emphasis on nitedi erence and nitevolume methods for the Euler and NavierStokes equations. Rather than presenting the details of the most advanced methods, which are still evolving, we present a foundation for developing, analyzing, and understanding such methods. Fortunately, to develop, analyze, and understand most numerical methods used to nd solutions for the complete compressible NavierStokes equations, we can make use of much simpler expressions, the socalled \model" equations. These model equations isolate certain aspects of the physics contained in the complete set of equations. Hence their numerical solution can illustrate the properties of a given numerical method when applied to a more complicated system of equations which governs similar physical phenomena. Although the model equations are extremely simple and easy to solve, they have been carefully selected to be representative, when used intelligently, of di culties and complexities that arise in realistic two and threedimensional uid ow simulations. We believe that a thorough understanding of what happens when
1.4. NOTATION
5
numerical approximations are applied to the model equations is a major rst step in making con dent and competent use of numerical approximations to the Euler and NavierStokes equations. As a word of caution, however, it should be noted that, although we can learn a great deal by studying numerical methods as applied to the model equations and can use that information in the design and application of numerical methods to practical problems, there are many aspects of practical problems which can only be understood in the context of the complete physical systems.
1.4 Notation
The notation is generally explained as it is introduced. Bold type is reserved for real physical vectors, such as velocity. The vector symbol ~ is used for the vectors (or column matrices) which contain the values of the dependent variable at the nodes of a grid. Otherwise, the use of a vector consisting of a collection of scalars should be apparent from the context and is not identi ed by any special notation. For example, the variable u can denote a scalar Cartesian velocity component in the Euler and NavierStokes equations, a scalar quantity in the linear convection and di usion equations, and a vector consisting of a collection of scalars in our presentation of hyperbolic systems. Some of the abbreviations used throughout the text are listed and de ned below. PDE ODE O E RHS P.S. S.S. kD O( )
~ bc
Partial di erential equation Ordinary di erential equation Ordinary di erence equation Righthand side Particular solution of an ODE or system of ODE's Fixed (timeinvariant) steadystate solution kdimensional space Boundary conditions, usually a vector A term of order (i.e., proportional to)
6
CHAPTER 1. INTRODUCTION
Chapter 2 CONSERVATION LAWS AND THE MODEL EQUATIONS
We start out by casting our equations in the most general form, the integral conservationlaw form, which is useful in understanding the concepts involved in nitevolume schemes. The equations are then recast into divergence form, which is natural for nitedi erence schemes. The Euler and NavierStokes equations are brie y discussed in this Chapter. The main focus, though, will be on representative model equations, in particular, the convection and di usion equations. These equations contain many of the salient mathematical and physical features of the full NavierStokes equations. The concepts of convection and di usion are prevalent in our development of numerical methods for computational uid dynamics, and the recurring use of these model equations allows us to develop a consistent framework of analysis for consistency, accuracy, stability, and convergence. The model equations we study have two properties in common. They are linear partial di erential equations (PDE's) with coe cients that are constant in both space and time, and they represent phenomena of importance to the analysis of certain aspects of uid dynamic problems.
2.1 Conservation Laws
Conservation laws, such as the Euler and NavierStokes equations and our model equations, can be written in the following integral form:
Z
V (t2 )
QdV ;
Z
V (t1 )
QdV +
Z t2 I
t1
S (t)
n:FdSdt =
Z t2 Z
t1
V (t)
PdV dt
(2.1)
In this equation, Q is a vector containing the set of variables which are conserved, e.g., mass, momentum, and energy, per unit volume. The equation is a statement of 7
8
CHAPTER 2. CONSERVATION LAWS AND THE MODEL EQUATIONS
the conservation of these quantities in a nite region of space with volume V (t) and surface area S (t) over a nite interval of time t2 ; t1 . In two dimensions, the region of space, or cell, is an area A(t) bounded by a closed contour C (t). The vector n is a unit vector normal to the surface pointing outward, F is a set of vectors, or tensor, containing the ux of Q per unit area per unit time, and P is the rate of production of Q per unit volume per unit time. If all variables are continuous in time, then Eq. 2.1 can be rewritten as d Z QdV + I n:FdS = Z PdV (2.2) dt V (t) S (t) V (t) Those methods which make various numerical approximations of the integrals in Eqs. 2.1 and 2.2 and nd a solution for Q on that basis are referred to as nitevolume methods. Many of the advanced codes written for CFD applications are based on the nitevolume concept. On the other hand, a partial derivative form of a conservation law can also be derived. The divergence form of Eq. 2.2 is obtained by applying Gauss's theorem to the ux integral, leading to @Q + r:F = P (2.3) @t where r: is the wellknown divergence operator given, in Cartesian coordinates, by ! @ +j @ +k @ : (2.4) r: i @x @y @z and i j, and k are unit vectors in the x y, and z coordinate directions, respectively. Those methods which make various approximations of the derivatives in Eq. 2.3 and nd a solution for Q on that basis are referred to as nitedi erence methods.
2.2 The NavierStokes and Euler Equations
The NavierStokes equations form a coupled system of nonlinear PDE's describing the conservation of mass, momentum and energy for a uid. For a Newtonian uid in one dimension, they can be written as @Q + @E = 0 (2.5) @t @x with 2 3 2 3 u 3 2 0
6 6 Q=6 6 6 4
e
7 7 u7 7 7 5
E
6 6 =6 6 6 4
u(e + p)
7 6 7 6 2+p 7;6 u 7 6 7 6 5 4
4 @u 3 @x 4 3
u @u + @x
@T @x
7 7 7 7 7 5
(2.6)
2.2. THE NAVIERSTOKES AND EULER EQUATIONS
9
where is the uid density, u is the velocity, e is the total energy per unit volume, p is the pressure, T is the temperature, is the coe cient of viscosity, and is the thermal conductivity. The total energy e includes internal energy per unit volume (where is the internal energy per unit mass) and kinetic energy per unit volume u2=2. These equations must be supplemented by relations between and and the uid state as well as an equation of state, such as the ideal gas law. Details can be found in Anderson, Tannehill, and Pletcher 1] and Hirsch 2]. Note that the convective uxes lead to rst derivatives in space, while the viscous and heat conduction terms involve second derivatives. This form of the equations is called conservationlaw or conservative form. Nonconservative forms can be obtained by expanding derivatives of products using the product rule or by introducing di erent dependent variables, such as u and p. Although nonconservative forms of the equations are analytically the same as the above form, they can lead to quite di erent numerical solutions in terms of shock strength and shock speed, for example. Thus the conservative form is appropriate for solving ows with features such as shock waves. Many ows of engineering interest are steady (timeinvariant), or at least may be treated as such. For such ows, we are often interested in the steadystate solution of the NavierStokes equations, with no interest in the transient portion of the solution. The steady solution to the onedimensional NavierStokes equations must satisfy @E = 0 (2.7) @x If we neglect viscosity and heat conduction, the Euler equations are obtained. In twodimensional Cartesian coordinates, these can be written as @Q + @E + @F = 0 (2.8) @t @x @y with 2 3 2 3 2 2 q1 u 3 v 3 6 7 6 7 6 2+ 7 6 7 Q = 6 q2 7 = 6 u 7 E = 6 u uv p 7 F = 6 v2uv p 7 6 7 6 7 6 7 6 (2.9) 4 q3 5 4 v 5 4 5 4 5 + 7 q4 e u(e + p) v ( e + p) where u and v are the Cartesian velocity components. Later on we will make use of the following form of the Euler equations as well: @Q + A @Q + B @Q = 0 (2.10) @t @x @y
@E @F The matrices A = @Q and B = @Q are known as the ux Jacobians. The ux vectors given above are written in terms of the primitive variables, , u, v, and p. In order
10
CHAPTER 2. CONSERVATION LAWS AND THE MODEL EQUATIONS
to derive the ux Jacobian matrices, we must rst write the ux vectors E and F in terms of the conservative variables, q1 , q2, q3 , and q4 , as follows:
2 6 6 6 6 6 =6 6 6 6 6 6 4
E1 E2 E3 E4
E
2 3 6 7 6 7 6 7 6 7 6 7 6 7=6 7 6 7 6 7 6 7 6 7 6 5 6 6 4
(
3 7 7 7 2 2 7 3; q2 ; ;1 q3 7 ; 1)q4 + 2 q1 2 q1 7 7 7 7 q3 q2 7 7 q1 7 7 7 5 3 2q q4 q2 ; ;1 q2 + q3 2 2 2 q1 2 q1 q1
q2
(2.11)
F
2 6 6 6 6 6 =6 6 6 6 6 6 4
3 2 F1 7 6 7 6 7 6 6 F2 7 6 7 6 7=6 7 6 7 6 F3 7 6 7 6 7 6 5 6 4 F4
q3
q3 q2 q1
( ; 1)q4 + 3; 2
q4 q 3 q1
2 q3 q1
;
+
;
2q ;1 q2 2 3
2
q1
3 7 7 7 7 7 7 7 7 2 7 ;1 q2 7 2 q1 7 7 7 5 3 q3 2 q1
(2.12)
We have assumed that the pressure satis es p = ( ; 1) e ; (u2 + v2)=2] from the ideal gas law, where is the ratio of speci c heats, cp=cv . From this it follows that the ux Jacobian of E can be written in terms of the conservative variables as
2 6 6 6 6 6 @Ei = 6 6 6 @qj 6 6 6 6 6 4
0
1 (3 ; ) q2 q1 q q3 1
0 (1 ; ) q3 q1 q2 q1
0
A=
a21 q ; q2 q1 q1 3 a41
!
;1
0
a42
!2
a43
q2 q1
3 7 7 7 7 7 7 7 7 7 7 7 7 7 5
(2.13)
where
a21 =
; 1 q3 2 ; 3 ; 2 q1 2
q2 q1
2.2. THE NAVIERSTOKES AND EULER EQUATIONS
a41 = (
2 ; 1)4 !
11
q2 q1
!3
+ q3 q1
2
!2 !2
q2 q1
!3 5; !2 3 5
q4 q1
!
q2 q1
!
q4 ; ; 1 43 q2 a42 = q1 2 q1 ! ! q2 q3 a43 = ;( ; 1) q q 1 1 and in terms of the primitive variables as
2 6 6 6 6 6 A=6 6 6 6 6 6 4
+ q3 q1
(2.14)
3 7 7 7 7 7 7 7 7 7 7 7 5
0
1
0
0 0
a21 (3 ; )u (1 ; )v ( ; 1)
;uv
a41 a21 =
v a42
u a43
(2.15)
u
where
a41 = ( ; 1)u(u2 + v2) ; ue a42 = e ; ; 1 (3u2 + v2 ) 2
; 1 v2 ; 3 ; u2 2 2
a43 = (1 ; )uv (2.16) Derivation of the two forms of B = @F=@Q is similar. The eigenvalues of the ux Jacobian matrices are purely real. This is the de ning feature of hyperbolic systems of PDE's, which are further discussed in Section 2.5. The homogeneous property of the Euler equations is discussed in Appendix C. The NavierStokes equations include both convective and di usive uxes. This motivates the choice of our two scalar model equations associated with the physics of convection and di usion. Furthermore, aspects of convective phenomena associated with coupled systems of equations such as the Euler equations are important in developing numerical methods and boundary conditions. Thus we also study linear hyperbolic systems of PDE's.
12
2.3 The Linear Convection Equation
2.3.1 Di erential Form
CHAPTER 2. CONSERVATION LAWS AND THE MODEL EQUATIONS
The simplest linear model for convection and wave propagation is the linear convection equation given by the following PDE: @u + a @u = 0 (2.17) @t @x Here u(x t) is a scalar quantity propagating with speed a, a real constant which may be positive or negative. The manner in which the boundary conditions are speci ed separates the following two phenomena for which this equation is a model: (1) In one type, the scalar quantity u is given on one boundary, corresponding to a wave entering the domain through this \in ow" boundary. No boundary condition is speci ed at the opposite side, the \out ow" boundary. This is consistent in terms of the wellposedness of a 1storder PDE. Hence the wave leaves the domain through the out ow boundary without distortion or re ection. This type of phenomenon is referred to, simply, as the convection problem. It represents most of the \usual" situations encountered in convecting systems. Note that the lefthand boundary is the in ow boundary when a is positive, while the righthand boundary is the in ow boundary when a is negative. (2) In the other type, the ow being simulated is periodic. At any given time, what enters on one side of the domain must be the same as that which is leaving on the other. This is referred to as the biconvection problem. It is the simplest to study and serves to illustrate many of the basic properties of numerical methods applied to problems involving convection, without special consideration of boundaries. Hence, we pay a great deal of attention to it in the initial chapters. Now let us consider a situation in which the initial condition is given by u(x 0) = u0(x), and the domain is in nite. It is easy to show by substitution that the exact solution to the linear convection equation is then u(x t) = u0(x ; at) (2.18) The initial waveform propagates unaltered with speed jaj to the right if a is positive and to the left if a is negative. With periodic boundary conditions, the waveform travels through one boundary and reappears at the other boundary, eventually returning to its initial position. In this case, the process continues forever without any
2.3. THE LINEAR CONVECTION EQUATION
13
change in the shape of the solution. Preserving the shape of the initial condition u0(x) can be a di cult challenge for a numerical method. We now examine the biconvection problem in more detail. Let the domain be given by 0 x 2 . We restrict our attention to initial conditions in the form u(x 0) = f (0)ei x (2.19) where f (0) is a complex constant, and is the wavenumber. In order to satisfy the periodic boundary conditions, must be an integer. It is a measure of the number of wavelengths within the domain. With such an initial condition, the solution can be written as u(x t) = f (t)ei x (2.20) where the time dependence is contained in the complex function f (t). Substituting this solution into the linear convection equation, Eq. 2.17, we nd that f (t) satis es the following ordinary di erential equation (ODE) df = ;ia f (2.21) dt which has the solution f (t) = f (0)e;ia t (2.22) Substituting f (t) into Eq. 2.20 gives the following solution u(x t) = f (0)ei (x;at) = f (0)ei( x;!t) (2.23) where the frequency, !, the wavenumber, , and the phase speed, a, are related by != a (2.24) The relation between the frequency and the wavenumber is known as the dispersion relation. The linear relation given by Eq. 2.24 is characteristic of wave propagation in a nondispersive medium. This means that the phase speed is the same for all wavenumbers. As we shall see later, most numerical methods introduce some dispersion that is, in a simulation, waves with di erent wavenumbers travel at di erent speeds. An arbitrary initial waveform can be produced by summing initial conditions of the form of Eq. 2.19. For M modes, one obtains
2.3.2 Solution in Wave Space
u(x 0) =
M X m=1
fm (0)ei mx
(2.25)
14
CHAPTER 2. CONSERVATION LAWS AND THE MODEL EQUATIONS
where the wavenumbers are often ordered such that 1 2 M . Since the wave equation is linear, the solution is obtained by summing solutions of the form of Eq. 2.23, giving
u(x t) =
M X
Dispersion and dissipation resulting from a numerical approximation will cause the shape of the solution to change from that of the original waveform.
m=1
fm (0)ei m(x;at)
(2.26)
2.4 The Di usion Equation
2.4.1 Di erential Form
Di usive uxes are associated with molecular motion in a continuum uid. A simple linear model equation for a di usive process is
Therefore, u must vary linearly with x at steady state such that the boundary conditions are satis ed. Other steadystate solutions are obtained if a source term g(x) is added to Eq. 2.27, as follows:
@u = @ 2 u (2.27) @t @x2 where is a positive real constant. For example, with u representing the temperature, this parabolic PDE governs the di usion of heat in one dimension. Boundary conditions can be periodic, Dirichlet (speci ed u), Neumann (speci ed @u=@x), or mixed Dirichlet/Neumann. In contrast to the linear convection equation, the di usion equation has a nontrivial steadystate solution, which is one that satis es the governing PDE with the partial derivative in time equal to zero. In the case of Eq. 2.27, the steadystate solution must satisfy @2u = 0 (2.28) @x2
@u = @t
" 2 @u @x2
; g(x)
#
(2.29)
giving a steady statesolution which satis es
@ 2 u ; g(x) = 0 @x2
(2.30)
2.4. THE DIFFUSION EQUATION
15
In two dimensions, the di usion equation becomes " # @u = @ 2 u + @ 2 u ; g(x y) (2.31) @t @x2 @y2 where g(x y) is again a source term. The corresponding steady equation is @ 2 u + @ 2 u ; g(x y) = 0 (2.32) @x2 @y2 While Eq. 2.31 is parabolic, Eq. 2.32 is elliptic. The latter is known as the Poisson equation for nonzero g, and as Laplace's equation for zero g.
2.4.2 Solution in Wave Space
We now consider a series solution to Eq. 2.27. Let the domain be given by 0 x with boundary conditions u(0) = ua, u( ) = ub. It is clear that the steadystate solution is given by a linear function which satis es the boundary conditions, i.e., h(x) = ua + (ub ; ua)x= . Let the initial condition be
u(x 0) =
M X m=1
fm(0) sin m x + h(x)
(2.33)
where must be an integer in order to satisfy the boundary conditions. A solution of the form M X u(x t) = fm(t) sin mx + h(x) (2.34) satis es the initial and boundary conditions. Substituting this form into Eq. 2.27 gives the following ODE for fm : dfm = ; 2 f (2.35) m m dt and we nd m fm(t) = fm(0)e; 2 t (2.36) Substituting fm(t) into equation 2.34, we obtain
m=1
u(x t) =
M X m=1
m fm (0)e; 2 t sin m x + h(x)
(2.37)
The steadystate solution (t ! 1) is simply h(x). Eq. 2.37 shows that high wavenumber components (large m) of the solution decay more rapidly than low wavenumber components, consistent with the physics of di usion.
16
2.5 Linear Hyperbolic Systems
CHAPTER 2. CONSERVATION LAWS AND THE MODEL EQUATIONS
The Euler equations, Eq. 2.8, form a hyperbolic system of partial di erential equations. Other systems of equations governing convection and wave propagation phenomena, such as the Maxwell equations describing the propagation of electromagnetic waves, are also of hyperbolic type. Many aspects of numerical methods for such systems can be understood by studying a onedimensional constantcoe cient linear system of the form @u + A @u = 0 (2.38) @t @x where u = u(x t) is a vector of length m and A is a real m m matrix. For conservation laws, this equation can also be written in the form @u + @f = 0 (2.39) @t @x @f where f is the ux vector and A = @u is the ux Jacobian matrix. The entries in the ux Jacobian are @f aij = @ui (2.40) j The ux Jacobian for the Euler equations is derived in Section 2.2. Such a system is hyperbolic if A is diagonalizable with real eigenvalues.1 Thus = X ;1AX (2.41) where is a diagonal matrix containing the eigenvalues of A, and X is the matrix of right eigenvectors. Premultiplying Eq. 2.38 by X ;1 , postmultiplying A by the product XX ;1, and noting that X and X ;1 are constants, we obtain =0 (2.42) @t + @x With w = X ;1u, this can be rewritten as @w + @w = 0 (2.43) @t @x When written in this manner, the equations have been decoupled into m scalar equations of the form @wi + @wi = 0 (2.44) i @t @x
1
@X ;1u
z } { @ X ;1AX X ;1u
See Appendix A for a brief review of some basic relations and de nitions from linear algebra.
2.6. PROBLEMS
17
The elements of w are known as characteristic variables. Each characteristic variable satis es the linear convection equation with the speed given by the corresponding eigenvalue of A. Based on the above, we see that a hyperbolic system in the form of Eq. 2.38 has a solution given by the superposition of waves which can travel in either the positive or negative directions and at varying speeds. While the scalar linear convection equation is clearly an excellent model equation for hyperbolic systems, we must ensure that our numerical methods are appropriate for wave speeds of arbitrary sign and possibly widely varying magnitudes. The onedimensional Euler equations can also be diagonalized, leading to three equations in the form of the linear convection equation, although they remain nonlinear, of course. The eigenvalues of the ux Jacobian matrix, or wave speeds, are q u u + c, and u ; c, where u is the local uid velocity, and c = p= is the local speed of sound. The speed u is associated with convection of the uid, while u + c and u ; c are associated with sound waves. Therefore, in a supersonic ow, where juj > c, all of the wave speeds have the same sign. In a subsonic ow, where juj < c, wave speeds of both positive and negative sign are present, corresponding to the fact that sound waves can travel upstream in a subsonic ow. The signs of the eigenvalues of the matrix A are also important in determining suitable boundary conditions. The characteristic variables each satisfy the linear convection equation with the wave speed given by the corresponding eigenvalue. Therefore, the boundary conditions can be speci ed accordingly. That is, characteristic variables associated with positive eigenvalues can be speci ed at the left boundary, which corresponds to in ow for these variables. Characteristic variables associated with negative eigenvalues can be speci ed at the right boundary, which is the inow boundary for these variables. While other boundary condition treatments are possible, they must be consistent with this approach.
2.6 Problems
1. Show that the 1D Euler equations can be written in terms of the primitive variables R = u p]T as follows: @R + M @R = 0 @t @x where 2 3 u 0 M = 6 0 u ;1 7 4 5 0 p u
18
CHAPTER 2. CONSERVATION LAWS AND THE MODEL EQUATIONS
Assume an ideal gas, p = ( ; 1)(e ; u2=2). 2. Find the eigenvalues and eigenvectors of the matrix M derived in question 1. 3. Derive the ux Jacobian matrix A = @E=@Q for the 1D Euler equations resulting from the conservative variable formulation (Eq. 2.5). Find its eigenvalues and compare with those obtained in question 2. 4. Show that the two matrices M and A derived in questions 1 and 3, respectively, are related by a similarity transform. (Hint: make use of the matrix S = @Q=@R.) 5. Write the 2D di usion equation, Eq. 2.31, in the form of Eq. 2.2. 6. Given the initial condition u(x 0) = sin x de ned on 0 x 2 , write it in the form of Eq. 2.25, that is, nd the necessary values of fm (0). (Hint: use M = 2 with 1 = 1 and 2 = ;1.) Next consider the same initial condition de ned only at x = 2 j=4, j = 0 1 2 3. Find the values of fm(0) required to reproduce the initial condition at these discrete points using M = 4 with m = m ; 1. 7. Plot the rst three basis functions used in constructing the exact solution to the di usion equation in Section 2.4.2. Next consider a solution with boundary conditions ua = ub = 0, and initial conditions from Eq. 2.33 with fm(0) = 1 for 1 m 3, fm (0) = 0 for m > 3. Plot the initial condition on the domain 0 x . Plot the solution at t = 1 with = 1. 8. Write the classical wave equation @ 2 u=@t2 = c2 @ 2 u=@x2 as a rstorder system, i.e., in the form @U + A @U = 0 @t @x where U = @u=@x @u=@t]T . Find the eigenvalues and eigenvectors of A. 9. The CauchyRiemann equations are formed from the coupling of the steady compressible continuity (conservation of mass) equation @ u+@ v =0 @x @y and the vorticity de nition @v ! = ; @x + @u = 0 @y
2.6. PROBLEMS
19
where ! = 0 for irrotational ow. For isentropic and homenthalpic ow, the system is closed by the relation = 1 ; ; 1 u2 + v2 ; 1 2
;1
1
Note that the variables have been nondimensionalized. Combining the two PDE's, we have @f (q) + @g(q) = 0 @x @y where v q= u f = ;v u g = ;u v ; One approach to solving these equations is to add a timedependent term and nd the steady solution of the following equation: @q + @f + @g = 0 @t @x @y (a) Find the ux Jacobians of f and g with respect to q. (b) Determine the eigenvalues of the ux Jacobians. (c) Determine the conditions (in terms of and u) under which the system is hyperbolic, i.e., has real eigenvalues. (d) Are the above uxes homogeneous? (See Appendix C.)
20
CHAPTER 2. CONSERVATION LAWS AND THE MODEL EQUATIONS
¡ ¢
¦ H B d 4 Y H h 5 D h H h H H 2 2 5B D T D 5 2 h 2 T Y 7 d H D H F D H 2 n H h 5 D Y R Y DB 7 d H H d R ` n YB F D q 5 2 5B XQW6IkGÆQI6QIÅ8EGVG¢I¬I¢9W6GeIWII6IkGIwGE9W6sI8U EIG¢Å8EGD r 4 H Y 2B 2 i ¥ H d R `B ¤ 2B 2 @ 5 F Y YB H 4 T Y h 2 T H 7B D F D 5 ` 2B by 5 b 2B F Y H 7 D Y Hy 7B Y H F QWIEej pWI8EvXCcIWvXuQVUWeIV9XGW¢ IXx88oXQ96XI9XIut ¶ Ã ¹ » ´ » Ã 9QavÄÂ ± Á ¶ ± À ± ¿ ¹ ½ ¼ ± » ¹ ¶ ¹ ej)6¾hfv6egº¸ · ¶ ´ ² ± ³ ² ewµ± ° ¯ ® c
2 5B D T 4By T y T 4B D 4 T d 8EGVQXEII¢VgXW6¢WI ` 2B Y 5y d 5 H bB D 4B d D Y H d iy R h 2 R ` 2B H D R 5 F DB IX8X988EW6EWGfQ¬EIIpIrIEQ9p8IWEC@ QGfx«VW6Gf¢ªQ6¢UXISQ©¢Iv@ 7 H DY i Y 2 TB Y H Dd T h H 4 T YB R P H 2 T H Y R H YB Y iy T 2 T i 4 T d R 4 4 T ` 2B @ 5yy 5q H F D q 5 F 4 R 7 2B t 2 5 Y T H d YB F D d 5 ¤ Y 2 5B D T R P H h H 7 d 5q Y 2 T d XWx¢U¢6VGI6Q¢IECcXX¢9IW¢I AX§8WVgWwXpG8UeQp8XWVUSQAQ9W8fI¢WGD H F D q 5 Y D 2 HB 4 ¨ H 5 4 Hy TB d T b 5 D 2B h H d 5 Y T YB 2 5B D TB d T b 4B d D H 7 5 H ` H F D yy T 7 d 5qB 2 R YB } pGpVvGo6XQ96XIVWV¢CGoEwQW¢WI¢§E8XWV¢VXWW698Q¢vIW¢XV¢W8IIXhxp3 ¢l¢£xlIu§8EWp8XvuEwCcI¢sxp3V¢l¢£xl#QEX¢Q3VuuXE¢VQWoXB yx x ¢ { H F t ¦ ¥ H d R `B ¤ 2B 2 @ 5 F Y Y T t } yx x ¢ { h Hyy T 4r 5 Y T 2B Yy T b d H D 2 h H 4 T YB R P H F DB @ F Y H 7 2 TB Y H D d T 7 d 5qB 2 R T 5 D h H 7 d 5q Y 2 T d D H 2 T 4 } y 2 Q6¢UXISQGCW6#¢QW¢¡8EpIGAQ9W8fI¢GW9rVgxp3v¢WcxUeXB Y H D T 2B h d 5 5 4 d T H 2ByB b d R 4 y T d H 2 H ` 2B h H Y Y H d H Y H F Y H 7 q 5 Y H Y Y Ty 4 y T d H 2 H ` H DB R P D T F D D 4 T 6GVIXI86¢QIXEXxI6¢G6IQ¢EeQWQIQIQ9vVf6WWVQVGQpQ8uGISCVUW¢q H F D H B Y T F 7 H H 7 H D Y i Y 2 TB Y H D d T T 2B h H D 2 H Y H d YB @ 5y H y TB d H D T 7 H F D q 5 yy pG QEW¢II#6 wQWxe¢XW6GVeXAQWS6W6WIEsEQ¢QGldIWj¢vEi d H D T F 2B h H Y Y R 4 YB h H yyB @ Y 2 5B D T 7B 5 d T H 4 2 H d H gB h r H DB 2 n H Y H F If6GpVUX6WIQXIuhXECI8EGV#XgWpI¢uQpQW6hI3GIUuWQpGD ` 2B iy T d 5q Y HB ` H D T d D d H D T F 2B h H D T H d D H d T Y H bB D T bB d H h H 7B D Y H bB D T bB d H h y TB D T pXxEIp¢u86X86GVWGxCQG¢UXe6GVQWWW¢u68GVVEQIf#EGp6Q¢EGl¢W6Ih¢XGVIWY 2 5 YB d H D T F YB F D 2B YB Y T F 7 H d R Y H bB D T bB d H h H 7B D d 5 Y H bB D T bB d H h y TB D T Y 5 D d H F DB 8XQG¢UEIGXfXWVUI9QpQ68GVVEQIw9XGC8fQ8GVVWQpm¢XGVIWdWf6IGQH h HBy T H 2 T 4 H Y H F t h H D 2 H Y H d YB Y H bB D T bB d H h y TB D d T 5 D Y 2 5B D T 7B 5 d T H 4 2 H d H gB 6XEII¢#sr¢gQIuQWoQW6WIuEwQ¢EWVVEW6Ij¢XGVUWI8XWV#XgWIIVQpQW6hIh r H DB 2 n q 5 D H 4 2 5 4 H F D t d H D T F 4 YB F D 2 1 H d H F h H B Y T F 7 H F 4 T 5 d T H F D YB YB F t h H by 5 GIUm¢p6QI¢QwIGUQ6GpVUfEIG3VWQpQ6X¢UI9Q¢oWpI¢aIGuXCEIu68E¢WY H 2 T 4 F 4B F @ Y 2 5B D T R P H H 4 2 H d H gB h q 5 D H Y T Y H 4 R h 5 d F 4B F @ x } { y { ~ } {yx rVg)EICwp8XWVUSQ9QI6W6hI¢w6wQQpIIrEIC9QUdxxG¢93dzkT 5 D ` 2B h T Hy t D H 2 h H D T 7B 5 d T H d T Y H bB D T bB d H h H 7B D H F t Y 2 5B D T R P H y TB D 2 H d H gB h i d T 2B WwIEI¢QXpQxI#QGldEpgII¢W¢vQ¢EGl¢W6If9XWfIuj6I8EGVUS6VGo6WQsEp¢UEIh r d 5 q 5 7 H D Y i Y T ` 2B 4 R h 5 d i H d H F D t D Y d n Y H bB D T bB d H h y TB D T Y H F D H D T 7B 5 d T 2 T 4 H 2 f8¢pQGfxdepXQpIIo6WQpGhQWIQ¢EGl¢W6ImVGlUW pGuGV#XgWIp¢k¢gwI85 t8Q8GVIWQWEfjQI8EGVIx6QpQW6hIwpXGIW6WfIG¢E8eQIWdI¢9WI¢QI¢GEIwXWWQpGD iy H bB D T 2 d H Dy i Y 2 5B D T R P H H 4 2 H d H gB h ` 2B Dy R Y H d H F D H by 5 Y 2 H F D h 2 T iy Y R 5 H 2 T Dy R 7B Y H Y H F H D T 7B 5 d T 2 T 4 H 2 H 7B D h 2 T H 4 T Y F D 5 5 D D 4 H Y H d F DB @ Y H bB D T bB d H h y TB D d T 2B T D 2 5 GV#XgWIp¢¢gIV#EG I¢C6¢UG¢GQWQ GCvQ¢EWVVEW6Ix¢W¢UwX¢GS¢Q4 Y 2 5B D T R P H y H h 5 7 d R 5 t @ 5 q hB R q i h T H D Y 2 R ` 2B 2 d H b 5 ` Y 2 5B D T R P H H F D F DB @ 2 5 7 7 5 4 2 p8XWVUSQQIx9wI8vusIrXIIpVgGfIIeIEIW68c8aI8XWVUSQ#IGGECA89#86)31
0 £
' ) $
%
( " #!
'
$ &
&
%
© ¨
¦ §¥
¤
£
D 2 H 2 5 H 2 QoQI8¢T ¢WXI¢IXWXp9Wf6WQpGoQ¢U 7 5 dq DB F YB R ` 2B D YB h 5 D Y H Y H F D 2 H d T F DB @ h H Y 5y 4 2 H iyy T R Y R YB h 2 T D B d 4 Y d H R Y T Y i T @y T YB H 7B D t d H F D H ` 5 D h H Y R 2 H F @ D R WECÅQ8X6IQEE¢UWp X¬I¢9XW6WQIWxg¢E#EGjQ6IG88W¬6WIÆQpCpI t D B d 4 Y R Y T Y T d T H T Y H 4B h 2B H 4 T Y h 2 T H 7B D H F D F D 5 t H 2 5y T h H Y R 2 H F @ D T F D H 4B D 5 QpXQIIWC¢¢Qsp¢uQ6XIpXwQ¢UI¢9XWupGeG¢pVI8E¢6WIQICulUG6XGVfò ä ç ñ õ ø Þà ß î þ ø ßæ ¡p«¬mÝ ä ç ß î þ ø ßæ ¾s ¬vÝ ä ç ñ õ ø Þæ ¡p«mÝ ç ¥ 8¥ pæ
ü ¢í ¡£Ý ¤ý ü ¥Icï ü í Ý ä ç ß îû þ ø ù æ ¡xÿsXÆê mÝ
ý ü ï Ý ä ç ñ û õ ø ìù æ IgUðëx©9úmÝ
Û Y @ 5yy 5q Y T YB ` 2B H h 2B D B d 4 Y d H R Y h 2 T D B d 4 Y R Y d 5q 2 5B D 2 H b 2 5 4 H F QCcXE8u¢CXIEIIEpXQWQIWI¢fpEW6WIIu8¢XGo68o8QwIut Y R 5B b 5 ` 2B 2 T H 7 H F D H ÷ T 7 hy R 5 F 6I8Eo8pXI¢Q#uIW¢¢#XI8pWY DxWo8Qu¢QECp¢WW6QQp96#EG#8#EºXpCe6W8WpI#6IGV#8©QWp¢s8£vì H D 2 5 4 y T 4 5y t ç i d T Y Y H 4 H 2 Y H 7B D H 7 5 Y YB F 4B F @æ Y H Y 5 d R d H F D 5 d 5q h H Y R H d T ö à õ à êà 2 H F i T @ d TyB 7B Y T 2B â h 2 T á d 5q h H Y R H d T ö h 2 T õ d H D T ô D R 5 F ` R 5 d F D Þ î ä ñ D T F Qpª8gv¢XX9XWX#I¢¢QIuW¢jI¢6GVvgp¢I8I8IGóeVIGD HD 5 ò W¢f pwWp8XvXeCcI¢sXdpXQVUWaIGu p¢AEdIEQ¢UwIWuXk W6IC@ ¥ H d R `B ¤ 2B 2 @ 5 F Y Y T t Þ 2B ` 2B 4 T Y H F D Þ î h 2 T ß 2B ` 2B 4 T Y H F D YB ß î H d H F ç pæ ¥ ç ¦ ¥ cEpæ ñ ì ä Þ î ì ä ï ª ða3Þ ß î ê ä í S«uxß ä ä ¡Þ ß
hB d ` h H 4 T YB R P H 2 T 2 5 2 H F t ì è YB X86Q¢IWXpx6¢¢Qpujvs¢é¢wEÞ d 5q D T F D h 2 T t ê è YB ß d 5q H h 2B F Y H 7 H F t h H 7 R Y Y T YB d H F D 5 H F D 2 5 H 4 2 H h 2 H H ¢uVUGI¢ëUc¢é¢wXC¢eIpXQ9apu6#IW¢CEf6IGVwpG8 QpQII6s6Ih t h H D 5 2 H h YB Hy TB d T b H 2 5 iy 2 5 2 H F ç Þ à ßæ Ý ä Ý H Y T 4 ã r ¦ H F D iyB d T 7B d d H hB Y 2 5 4 H hQGVIQIXwXI¢XW¢Vwp8I8kQpªuWmåeW¢QÏÔ6eIWdXWVdEWI6IXI8Qv@ Y d H D T F 4 y T d H b H Y D Y d n H F D d 5 ¤ â à á à ß h 2 T t Þ Y Hy TB d T b D 2 H h 2 H H h 2B H F D q 5 Y 2 5B D 4 2 Rq H d W6GpVUh¢WQ¢6WUfIWj8UjpGdIVpjQEI¢XW¢VoQpIQQIpXpG¢vI8EG6pI¢T t Hy 7 T H d 5q t Ý t Y Hy TB d T b D 2 H h 2 H H h H F D ty T d H 2 H ` 2 1 Y 2 5B D T 7B 5 d T H 4 2 H d H gB h H DB 2 ¢Xp#¢p8QQEI¢XW¢VCoQII6s6IIWVGQpQ836I8XWV#XgWIIVuQI6WQsEpGIUn Y H 4 T Y y T 2 5B D T D R 7 5 4 h 2 T y T 4B Y i F Q6Q¢IWhVU8XWVI#¢QI¢¢QXWxojÜ Û ¦ ¥ H d R `B 8puWI¢Xv¤
Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u§fj«wÔzmQÔv«SÏmÇ
y
x
8
7 d 5qB 2 R YB ` 2B 4 T Y H F D t ß î d 5 Þ î 2 5 D B d 4 Y R Y 5 2 YB H d H F D 2 H F 8EpIuXIEQ¢UuIGU C8 8fpXQIIWIuEaQIW6Iª ç 9 ¥ ¢%pæ 4D ¢W6H à ï Ý 6 4 ü ï Ý ä ï Ý UÅ85cUªaëî à í ß 6 4 üí ß ä í ß sxr'5xªCx î à ïÞ 6 4 ü ï I3r75c3Þ ä ïÞ a3 î
D T F D F 4 R Y h H 2 n H h YB F 4B F @ t î y 5 7 i VUWIQpU6IuEEICIe8 xY H F D YB 2 5B D H 4 H H 2 5 H F t ` 2B 2 T H 7 H YB 4 H d Y DB H 2B 7 d H D H h 5 D h H Y R H d T Y i T @ d H F D pGXk8EGp6Qp8IuuIIEI¢g9WEQ6WIWE pX9WQW6IWk6WI¢wxcQpG S( y 5 7 i Y H F D 2B h HB d d T 4 D 5 2 YB 2 5B D T 7B 5 d T H F D q 5 ç d H h d 5 h 2 Tæ H d R D T 2 H YB 4 H d H F D D R 8swfuIGE6XW¢QC¢IuXd8XWVdEcII¢upG§¢6IW8I¢9WIWVUuWEQ6WIuIWupI ç ¥ 8 pæ
0 ¥21¥ (
à 321 ( 0
t Hy 7 T H d 5q t D T F D F 4 R Y H bB D T bB d H h T 5 D 2 5B D T 7B 5 d T H 4 2 H d H gB h ¢Xp#¢pu8pglUGIWu8GVVWQp9Wd8XWV#XgWIIVw6IQQsEIT D 2 H Y H d H d 5 D h H Y R YB ( y 5 7 i Y H F D D T F D 2 T H 7 H @ YB F D i & H R PB 2 R D 5 2 YB DB ç 2 5B D H 4 H H 2 oQQWpQWGaQWp§Eu)¢s xfvIWmlUGu¢g9vXIWv'VISXIpv¢p§E§Eë8XWpQ6I85 F DB @æ D R i F 5 Y 5yB F H 7 T Y H F D Y @ 5yy 5q Y 2 5B D T 7B 5 d T H 4 2 H d H gB h d 5q 2 5B D T D 5 2 H F WECpI Sp8W8EXIpf9¢GIWCsEX8p8XWVdEcII¢QI6W6hI88XWV¢pIut ß 5 D D 4 H Y H d F DB @ Ý q 5 H bB D T bB d H h D Y d n H F D D 2 H Y H d H d 5 D h H Y R H D 5 2 yyB 9WQW6WGECvV8EWVVEQIUIGfoQW6WI6WGQIs¢phXEC@ ISut Ý Y R F Y H bB D T bB d H h Y Y H d H 5 D h H Y R d H b H 2 H d T Y Hy TB d T b D 2 H h 2 H H h 2 5 Y D B d 4 Y R Y D H D YB F D 2 QQ¢EWVVEW6IuW6WIdGQIQ¢6IaW¢fQEI¢XW¢VoQIpQQI¢GpEWQIIxGCXpG31 ç $ ¥ ¢%pæ Þ Ý
"
" ä #!Ý
H Y R H @ ã q 5 i h R D Y H F D 2B H bB D T bB d H h H 7B D i d T 2B h d 5 H F D d 5 IuuY u¢ppGuIWeX8GVVWQp#EG#¢UEIW¢wpG8U¤ ç ¥ o pæ 4D ¢W6H à
ß Ý
ä eÝ ¥ à
Þ
Ý
ä eÝ à
ß Ý
ä eÝ
iy T H ` 2 T F 4 d H D 2B H Y R H @ H 7B D d 5 H 4 T Y 2B Y H bB D T bB d H I¢Q8I¢UQWSEaIw9XW8wQ¢UeXCQ¢EWVVEW6Ih y TB D d T d 5q Y R F t Y 2 5B D 2 H b 2 5 4 y T R Y R H F D 5 D ` 2B h d 5 4 4 ¢W¢I8#ISu©QI8EGoQ¢o8QC¢UIIGeWAIEIW¢QQVT QWQIWV#68GVVEQã h H Y Y H d H H d T Y H bB D T bB d H D 2 H 7 H ` 2 T d d T hB d ` H 7B Dr H 4 T goQ9Q¢I¢G¢X89XG6¢U Û pI8Ev¤ ¥ H d R `B
©
©
¥
Ó × Ò Ë É Ë × u§IawÓ
Ì Ç Ó Ì Í Ì Ñ Ñ Ò ÖÕ Ì Ë Ò Ó Ò Ñ Ö Ó É Ú Ì È Ú Ì Ù Ð ¦Ð fvÔmQfÔvvº8§xÎ
n1
n+1
n
t
j2
∆t
j1
j
∆x
j+1
j+2
Grid or Node Points
x
©
¨
ç pzæ ¦ ¥
ç zdø ß îæ
4 í Ý 6 4 üí AxÅ5Ý
ß î
ä
í W ß V Ý
7 d 5q H F D 2B h H D 2 H Y H d Dy R Y H d i d T 7 7 R Y H F 8IGeXQGo6WQICEpWQf¢d9IwpGD Y H H Y 2 H Dq 5 H 2 5 h 2 T t 2 5B D T 7B 5 d T H 4 2 H d H gB h h H d H D 2 H 4 D 2B 5 r H H d F D H F D Y T 5 D h H d d Hq H d Y Q6W9QW¢fp8#p¢o8EGV#XgWpI¢fQI6WQsEpeQW6GoQ6oX83QWpGCIG¢aG#6WQ6XB d H D D Ty H F t H bB D T bB d H h D Y d n T 5 D 2 5B D T 7B 5 d T d H h d 5 r h 2 5 4 H Y T YB ç ß QGVpu¢¢EGl¢W6IuUwW#8EGV#XgWIp¢CQI83xI8Q6WwXvs î IÔAxk£5xæ æ rç4 í Ý 6 4 üí Ý u D T F D Y @ 5 F Y ¦ ¦ ¥ P t iy d TyB 7B lUGjCcIWu¢EpSëj8W¢XX9X y í xWw8XWV#XgWIIVQI83fWIX#V6935xæ w u t 5 D 2 5 B D T 7 B 5 d T d H h d 5 r D Y d n T Y B ß î r ç í Ý 6 4 ü í Ý 2 5B Y Y H d H H F D D T F D H H Y H @ 8EWW6WIuIGmVUG6WuvVt ¡ Ep8j8ëvxIG#pGë¢v6IW¢uIGv688jQvcsjfQvcXy ¦ ¥ P 7 5 d ¤ h 5 F D H 7 H F D q 5 d H h d 5 H F D Y H bB ` d H @ 5 D Y H @ 5 H F D 5 D ` 2B 4 T Y hB d ` H F D ` 2B 2B T D 2 5 4 7 d H D d 5 d d H H F t d H h d 5 2B T D d H 4 T 5 D ` 2B 4 T Y hB d pGWIX6¢UrX8dpGIXIE¢o86¾QWw8W6Iu6QI8rE¢fQ6WIX6¢UrX8` H F D 2 5 Y h 2 H H h d 5 d d H H F D D T F D d T Hy 4 5 Yy T YB D1 2 `B Y y T R P H H F D q 5 H hB Y D F `B d H F D 2 pG¬8#pIQQI8W6pGVUW ¢gEQkWE¢X9£Å¢XWC¢Ix6IG¢IXo8XIG¬85 Y d T H T F 4B F @ d 5 d d H q 5 D 2 R 5 7 T 2B T D d H 4 T F DB @ H bB D T bB d H h D Y d n T D 2 H Y H d H d 2 5B D T R P H H F ¢gI¢dXICj8W6¢oI89¢#X¢G6QudWECu8GVVEQIfWUuoQQIQ#8EGVIx6fpGD q 5 H hB Y Dq Hy H F D 2 5 Y 7 d H D H D H d 4 YB h H F D D T F D d T Hy 4 YB DB t d H 2 2 T 7 YB F D 2B h H Y Y H d H 2 H F ¢uIEWC6EapG8C#QGW6W6WEIapGCVUW¢gEQCXCUQQpI¢#EIGeEeQWQI6Iª ç ¦ ¦ ¥ s8pzæ
xx8í W ß ¡ V ¦ V
ß
î Ý ¦
ø
í W ß V î ß Ý ¦ ä
í W ß ß î V 6 Ý 4 í Ý 6 4 üí ¥x)5WxÝ
2 5B D T R P H ` 2B @ 5yy 5 8XWVUSQwIXCcXE8q H F pGD 8ÆWeEIQApGªI8W¢QWðI¢«9p8¢8GVWQ9ISkIGD X9W6GeIGðI¢Uv 7 d 5q 5 D Dy R Y H d H F D R 5 d ` H d h 2 T ê D R 5 T d 5 D T d H 7 R 2 H F 2B Y 7 d H D H F D h 2 T s î pf¥xÅ75xA8EGV#XgWIp¢w6IQQsEIwQVUWuIW6IXI8QfxXVduXe î ç ß æ r ç 4 í Ý 6 4 ü í Ýæ 2 5B D T 7B 5 d T H 4 2 H d H gB h H 4 T Y H F D d H hB Y 2 5 4 D H ò yy T 7 Y YB ß u y w u t d 5 q 2 5 B D T 7 B 5 d T H y T 2 5 Y T H d T Y B ß î r ç í Ý 6 4 ü í Ý æ 2 5 B Y Y H d H H F D Y R F Y T ` 2 5y Y VI8Ev¢T í xv18w¢XGldEpgII¢vXIVU8WVgWfvE# ¢sQxaq5Wx¢XWQIIWvISut ç ¡ pzæ ¦ ¥
Y Y Y aa`ø
í W ß V ç ß î ¡s æ Ý ¦ ø
í Wpß ß î V iä í Ý 6 4 üí x)85Ý Ý @ 5 cfò
2B T D 5 5 D ß î i H hB bB h h 2 T í Ý D 4 T d D R X¢Gp89GeSpXxEII¢¢GWpIWY ç h ¥ 8dpæ
d T H 2B ¢gpXXy
í W í í W b ï ß V s æ 8ì saagø W ß V sæ ø ß ¡s æ xªj5xÝ Y Y Y aa`ø ø Y Y Y V ç ß î ø í Ý ä 4 üí ç ß î ç ß î ¦ ¦ Ý Ý ï Ý ï Û 4 ü í Ý d 5q 2 5B Y 2 T H Y HB d H Y d 5y i T t H F D d H hB Y 2 5 4 t Hy 7 T H d 5 5xu8¢XWp¢UQEWQu8XxgmwpGuQIEWI¢QIVXI9¢pu8U¤ Y Y Y aaSà ¥8f#8fIWxI¢xm¢I8EGVIXI ¢Q4 e à ¦ e ä õ d 5q ý ü í Ý h 2 T í Ý q 5 Y 2 5B D T 2B 7 5 7 5 dq h H 7 d 5q H 2 T 4 H bB D T bB d H h y TB D d T 2 H bB ` T 5 D Y 2 5B D T 7B 5 d T H 4 2 H d H gB h y T 4 5 8WuQ9W8jaVg8GVVEQI¢¢W¢UwQ¢E8CGI¢XGldEpgII¢QpQW6hIo¢Qô
í W í í W b ï ß V s pæ 7ì caa`ø W ß V s pzæ ø Xß ¡s pzæ vðCIWxÝ ç8dpmaa`ø c ¥ æ Y Y Y ç ß î õ ø Y Y Y ç ß î õ V ç ß î õ ø í Ý ä ý üí Ý ï ¦ Ý ¦ Ý ï Y H bB ` ß D R 5 T 7 d H D d H D D Ty H F D ` 2B h 2 T ý ü í Ý ä ç ß î Q¢E8up8¢µW6G6GVuIW9IEII¢IpvUsIWxÿs põ ø ß î êæ # xmÝ sð#mAp¢) Sê p¥ pG pQSjTXkQ¢E8A8EGoQ¢S¢Qe8EGVG¢I2 ä ç ß î õ ø ßæ Ý h 2 T ß î ä ß t ¥ 5 D ¦ ¥ Y P 2B 2 H bB ` 2 5B D 2 H b 2 5 4 2 5B D T D 5 H F D ` 2B @ 5yy 5q t 2 H F t h H n Þ F DB @ ç Þ à ßæ Ý d H hB Y 2 5 4 t Hy 7 T H d 5 pGkIXCcXE8uQpuóQp©GECWm#QIEWI86u¢EI#VI#¢ë¤ 8EWI¢IpQEW6WY 2 5B Y 2 T H Y HB d H d 5y i T t Hy 7B Y T q 5 Y 2 T H 7 i h H D T Ry T b H d 5 h H D T d H 2 H ` H 2 T 4 2 5B D T 7B 5 d T H 4 2 H d H gB h ¢XxgvjEI#EWë¢vp¢g9wSaQWVUXV¢m¢ QWVG6IQ8 ¢gu8EGV#XgWpI¢6IQQsEIÆi ² ¶ Ã ¹ » e#6ja´
R ¹ P Ã À C C SQjeIHG
± E ¹ » ´ E ¹ À 9FjwwF± ½
± Á ´ C wDB
@® A
Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u§fj«wÔzmQÔv«SÏmÇ
° ¯ § ¢ « ¨ © ¢ « § ¢ ® « ¥ § ¥ ¢ ¬ « ª © ¨ § ¥ ¢ ¦d¦q£¦A¤sF'¦d¥¥dsa1¦£sq7¦d¤£a¡ddvo
H F Y H 7 H F D 2B D 2B 5 i d H b H D T $ ¦ ¥ P i 2 H bB ` 2 5B D T 7B 5 d T H 4 2 H d H gB h h H d H D 2 H Q9apGXoX8fQ¢6VqEsëjoQ¢E8e¢XGldEpgII¢aQI6W6hIeQQGo6Q4 H F D HY R h 2 T pGeWI¬IVuÝ 9msÝ ç ¡ mjQp8XWEII¢QÅf¢UII¢sa6EIXX8#EB cfò ä ç æ Ý à ä æ Ý t Y 2 5B DB h 2 5 4 i d T h 2 R 5 D Hy F 4B dB ã H Y 5 7 @ 5 ç ¦ sªø
æ r ä ß î F Y H 7 D 2B 5 d R 5 I
vhWQ9uoX¢suI¢ë¤
6
Y
¥
6
Y
6
¦
¦
6
¡
ä ê ä eß
Y HB d T h 2 R 5 H F 6QEW¢UpI8wpGD ` 2B h Ry 4 H Y D 2B 5 y }x y q 5 d H 7 R 2 H F D 2 T H 7 H @ t F Y H 7 T 2B Y D 2B 5 q 5 d H 7 R 2 H F IEIIEQCGoE8sl3UVf6s pxwIW¢Q#wvëpW6#ECGoX8V6swISaIWD q¢¢gWveQICfVUWwQEG¢ucXQkCcIWuTI¢VGoX8¢UpI8GCWQ97 5 ÷ T H Y H @ 2 H F @ D T F D H 4B D 5 ò @ 5y H 2 @ 5 F Y h 2 T D T Y D 2B 5 i d T h 2 R 5 F DB @ F Y H DSE8p8wpGCQIEWI8h8XWVdEcII¢a9¢GuIWC¢8EGVGS6W6WI6WC8GlGQ8w¢9{ 2B 5 d R 5q H F D d H hB Y 2 5 2 5B D T 7B 5 d T H 7 T Y H F D d 5q 2 5B D T D 2 H Y H d H d d 5 D T d H 5 y x T H bB d H h Y R D Hy @ 5 ò H bB D T bB d H h h 2 5 4 H Y T 5 D 2 5B D T 7B 5 d T H 4 2 H d H gB h D 2B 5 T YB F 4B F ¢EQIuI6Eÿcj¢¢EGl¢W6II8Q6W W#8EGV#XgWpI¢6IQQsEISE8wuXeEIC@ ç $ ¦ ¥ 8qpzæ ç4 í Ý ø í fAxÆvÝ 85Wxæ 6 4 üí Ý
ß î ¦
ä íç ( zsÝ ¥ æ 2 5B D Ty H d H F D d H hB Y 2 5 8XWV6WuIGu6IXI8
} r e j  r p y w p t i p r p ng l g r j U¡7mxvu{f o!¡7
o
2 5B D Ty R 7 d 5q h H D T 4B D YB F 5 h8XWVIwW8eQWVgEGfXII¢WY H d 5 7 T Y H dB R P H d 2 5B D T 7 d 5q 2B y T 2 5B DB h h T F 4 R h H 4 d 5q 2 H H d T Y 2 5B DB h 2 5 4 i d T h 2 R 5 W89#QWEISQe8XWVd8pE¢U8EGEpI¢eIxhQQ8p6a¢p8XWEII¢QVUII8 cIF @ 5 d 5 h H 4 2 H d H gB h H d T F Y H 7 H F D 2B Y D 2B 5 d H F D ¢rQ6IQQsEI9WV)W6#9IWAEwWoX8a6IGV5 cIapjXQWaI¢XWQIdQIW¢wQpGE6I2 @ 5 F Y R yy H D Y 2 5B Y Y H d H H Y H F D q 5 d H F DB H t6Q8cQGoE8saIXpII8WI¢#QG9XQ9a#XvoX8W6W6WEII¢Vf8GVVWQpaT d H b H @ 5 Y D 2B 5 ` 2B h 2 R 5 d d R Y q 5 Y 7 d H D 2B F Y H 7 T 2B D 2B 5 H D H d 4 YB h H 2 5 D T H bB D T bB d H h 5 D 2 5B D T 7B 5 d T 2 T H bB ` i H F D H 4 2B Y x } } } } y x y d 5q YB Y T H F D H d T H Y H F W¢XGldEpgII¢d¢¢E896pGf6IXGlÔ¢UhG ¢UlUj8XWVUIWf¢W6Iut ç ¦ ¥ Sgpzæ ç ¥ ¦ ¥ ocpzæ
ç z~f¥r75xæ ß îæ ø ç4 í Ý 6 4 üí Ý
í W ß ß î ç ß îzæ~øçf4¥íÝ¬øíxÝ 6845üíæ Ý ä V ¦ Ý í W ß V Ý ß î ¦ ä
Û Y H bB D T bB d H h h 2 5 4 H Y h 2 T D Y d n H F D d 5q Y 2 5B D T 7B 5 d T H 4 2 H d H gB hr h H d H D 2 H 4 QQ¢EWVVEW6Ip8Q6WI¢fUIG¢I8EGV#XgWIp¢ 6IQQsEI£QQGo6Q4 D 2B 5 r H H d F D H F D H d T Y Ty R 7 d 5q H 4 2 H d H gB hr H DB 2 n q 5 Y Hy 7 T H 2 5 7 7 5 4 D Y 5 7 H F D Y T F d H SE836WIWaIWw¢¢p W86IQQsEI£GIUVf6XI9¢p89#8689apGIVUW6mÜ
} r e j  r p y w p t i p r p ng l j ig e h¡7s{zxvusqo2mkshfd
² À Ã » ´ À ± C 9wr ± Á ¶ ± À ± ¿ CejÅ6¹ ½ ¼ ± » ¹ ¶ ¹ hfegº¸
s
®
$
Ú Í × Ë É Í Ì Ê fuu×
Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Ð vÔmQfÔm«xÎ xÎ
H d T F Y H 7 D 2B 5 r d R 5q T 2 5 Y 2 5B DB h 2 5 W¢W6#uoE8s£fI¢8up8XWEII¢Q4 i d T h 2 R 5 D Hy F 4B dB ã F DB @ H bB D T bB d H h h 2 5 4 H Y h 2 T D Y d n T d 5q Y 2 5B D T 7B 5 d T H 4 2 H d H gB ¢III8u6EIXXGEC¢EGl¢W6I)I86QrI¢wUw8p8XWVdEcII¢#QpQW6hIh ry T d D 2 H 4 D 2B 5 r H H d F D H F D ` 2B D 2 H Y H d H d Y d 5 D T d H 5 B d D T 7 H F t H 7 T Y H F D YB Y H 2By H F D q ¢WGoQ6oX83QWpG IWIEGoQQIQW8WVG6s¢ WWVdIuw¢9¢WIGEQpXXpG¢5 H 2 5 2 t h 5 F D H 7 y T d D 4 H Y T ` 2B D 2 H Y H d H d d 5 D T d H 5 H 4 2 H d H gB h B d D T 7 H F D q 5 H Y T 4 H 7 H d D I¢IUxIG#§¢GW66s#eIEGoQQIQ¢GVWQ8w6IQQsEIWWV#wIWv¢aW¢Qw9QWWxH H F D 2 1 h H nB h 5 7 H B d D T 7 H F D q 5 5 D d pG3QIEIx##GV#uIG¢8W85 ¢GWVsuIW¢gIu¢uVu6IXEwpGVIGGl6EIh 7 5 D D 5 H F D d T H 2 d 5 D T Y H 2By H F D D T F D H D T D 4B i T 7 Y 2 5B DB h 2 5 4 i d T h 2 R 5 t Hy 7 T H d 5 ¤ H 7 T Y H F D iyB d T Y Y H 4 H 2 D 5 2 H d T H 2By 5 D H 2B 7 5 d gdjI¢XGII86VUII8o¢EI9¢pj8UjV#¢WCIG X¢GQQ6I¢Iu¢CIXEuGIEXy ¢Wq h H Y R Y d 5 D T d H 5 D 2B 5 H F D D R t d 5 D T d H 5 H 4 2 H d H gB h D 2B 5 T 2 5 h H Y T YB d 5 D T d H 5 H 4 2 H d H QIW8WVG6s8voX8pGpIxQ8WVG6s8fQI6W6hISE8ae8dQWVUvE8GVWQ86IQQq rqB h B d D T 7 T q 5 H 2By F 4 T d 5 D T d H 5 H 4 2 H d H gB h B d D T 7 T Y H D T d D Y RyyB Hy 7 T H YB F EI WGldv¢aIXE¢½Q¢GVWQ8QI6W6hIWWV#6GVWGpXXE Xp#¢pXIut ¡ ¥ ç ¢ pzæ
y A ¼D» e² Ý ¥ ( º² t ø ²Ý ä
Y T 9 ¦ ¥ P H DB d @ H d 2 T 4 H ¢
qpëjWEC6W¢Qav@ 6 ¦ ¦ ¦
ç h ¦ ¥ oapzæ
¹ · ·· ¶ !Ý ¹ ·
¡ ¡
6 ¦ ¦
6 ¦
³´´
6
´µ ß
î ¦
ä
»
h 2 I¢T
ç c ¦ ¥ oapzæ
·· ¸¶ !Ý
³´´
´µ ß î ¦
ä y ¦ t º² à
·
·· 4 ¸¶ £Ý
Ý ¹ ´µ Ý ´ Ý ³´
ä ² Ý
` 2B 4 R h 5 d D 2 1 7 d 5q y TB 4 H Y i d H b T Y T F B d D T 7 ` 2B Dy R Y H IEQIpGo3W8¢XQ6sQ¢ fVU WWVdIEGIW6Wd H F D D T F D t d H F D d Rq h 2 T t 7 d 5q B d D T 7 r d 5 D 4 H b T 2 pGwVUG§Q6IGfIAp¢§8eGV#3Ô8GQ¢XB QpGwQWppr¢QdvuVUG¢QX6wEwEB 7 H F D Y Y H d H 2 T 4 H @ D T F D d T Hy 4 YB D
Ý !ø
ç 9 ¦ ¥ 8qpzæ
Ý 6 Ý ø
ß î r ¢sç ß î r ¢sç ß î r ¢sç ß î r ¢sç
Ý Ý 6 ø Ý
Ý
æ æ
ä æ ä ä ä
Ý ±£Ý 6 4 ø Ý 4 £Ý 6
ä ä ä ä
Ý !µæ
ç hÝ ¥ ç hÝ ¥ ç hÝ ¥ 4 hÝ ¥ ç
(æ (æ (æ ( æ
7 d 5q H bB D Y H ` ` R Y H d 5 7 H F D 2B Y 2 5B D T R P H H Y H F D ` 2B DB d W8u8GfQ88pWW¢#uIWECI8XWVUSQuQIW IXWEW ç ¦ ¥ ocpzæ ç Ý 3!Æø
Ý Ý 6 æ
ç ( sÝ ¥ æ ç ( sÝ ¥ æ
ç Æø Ý ç
Ý ç ÝÆø Ý 6 æ Ý 8£æ 6 4 Ý ¬vÝ Ý ø 4 6 Ý k!æ
ç ( sÝ ¥ æ
ß î ¦ ß î ¦ ß î ¦ ß î ¦
4 sÝ ¥ æ ç (
Û Y 2 5B D T R P H d R 5q H F D D T H bB d d 6I8XWVUSQup8IGCVu¢EW¢T
Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u§fj«wÔzmQÔv«SÏmÇ
y T 2 5 ` TB h y T d D 2 H z¢U8¢¢IVGGo6Q4 H F D 5 D Y d Hq H d è H 2 5 y T d D 2 H 4 H F D h 2 T d H 7 R 2 2B h h 5 Y i T @y T H d T B d D T 7 h H h 2 T pGdWW66Wc¢é¢ep8v¢WGoQ6 IWI¢Q ISXIfxcEVW¢WWV#eQIIVUdT d¢GoQ9I8¢uIujQ² º c² ë² ¢6WQWppseVgCII¢IIGaI8XVfWo¢fI8Qp8IWEC@ 5q Y D 2 H 7 R ` d T H F t ç à à æ Á Y T h H Y Y H d H H 2 T 4 Y h 2 T H F D ` 2 5y T Y D 2 T D Y 2 5 4 D R 5 F DB B d D T 7 y T 2 5 ` TB hB d D i i DByB B Y Y 5 T YB Y h 2 T q 5 d H 7 R 2 i 2 T F ` R 5 F Dy T B d D T 7 y y GV#¢VU88¢XIXGÅ83XEIXW8f§X§IIVUI¢§Q ISuo¢a8I8pGEVuGV#¢lxc¢x Hy 7B Y T d 5q 2 H bB ` YB 2 5B D T d D Y RyyB H F D h 2 T y T 2 5B D 5 YB D 2 H 7 R ` d T H F D 2 XI9Xw¢Q88E8XWVGWIEXXfIGp¢¢U8EGp8vXS6#p8W¢IG#XB 8¢IW6IC@ q 5 HY R HdH F ç ¥ 8$ pzæ
¹ · ··
¦
º
··
·· ¸¶
º Ä º
´´µ ´´ ´´ ³´
ä ç º à à jQÃsÃëÂÛ
æ ¤eÁ
2 5B D T D 5 2 H F D H 4 R h 5 d D 2B h 2 T y x { } À T B d D T 7 T F 4 R Y yy T 4 H y T 2 5 ` TB 8EGVG¢IuIG6IIxWGoEp¢w¢99Wx¢e WGldpWE¢gumVU88¢XIh yVGGo6QpG«Ip8W¢ª6W6GpXQW¢ªEICdIp¢UAI8XVeW8pG#¢pQ68Q6E¢¢T T d D 2 H 4 H F D h 2 R 5 d T h H d H D Y Ry 4 H d T F 4B F @ Y h 2 T ` 2 5y T H Y 5 F D d 5q D H 4 H Y 5 d H yy T H d D T F D Y D 2 H 7 Hy H F DB @ B d D T 7 H d T R P Y T YB Y d 5 D T d H 5 H 4 2 H d H gB h B d D T 7 H Y H F D q 5 F 4 T lUGGo6#6XQkWEC WWVd9WVUSWeuXuW¢GVWQ8QI6WQsEpeGV#9WQpG¢r¢
ç ¥ S pzæ
¹
¦
¥ ¢r 6 ¦
¥¢r 6 ¦ ¦
·
·· ¸¶
6 ¦
³´´
6
´µ ßî ¦
ä ¥ (
i 2 H bB oQ88` YB D F `B d H F D 2 5 2 5B DB h 2 5 4 2 2 T 7 R H ò T d 5q yx {y } ~ } h 2 T Y D 2B 5 Eo8XpG¬8¬8XWEII¢QÆI¢#IQA98el¢98FGoxlW¢lckóp¢9GoX8 d 5B d H D 2B D T H 7 H F 4 Y ` 2B 4 2 H d H gB hry T d D 2 H 4 D 2B 5 r H H d F D T d 5q d 5 D T d H 5 B d D T 7 H F D 2 H F 8XQGoEdV#6IWIX6IQQsEI£¢WGoQ6doX¢s£QIGAd88WVG6s8Æ WWVdIGº6Iut ç ¥ o¥ pzæ
Wß V Ý
ä
¿¡¾¦ Whß V Ý
YB 2 5B DB h 2 5 4 i d T h 2 R 5 H F D H d H F Xe¢XGII86eVUII8uIGuW6IC@
¥ ç 8 pzæ ç
ß ß r6 æ î ¥ 6 W ß V î ¥ Ý Ý ¦ Ý ¦ ä
ç ( sÝ ¥ æ
D 2B 5 ` 2B @ 5yy 5q H F SE8ÅIXCcXE8kIGD
Ûç p p8XW6Qxw6A¢XGII86eVUII8I¢#IQ9C¢8GVWQ85 ¥ 2 5B D 4 H H H æ 2 5B DB h 2 5 4 i d T h 2 R 5 2 2 T 7 R H ò T d 5q d 5 D T d H i ¦ ¥ P 2B H 2By F D d R 5q H F D H 4 Ty H d t Hy 7 T H d H F D d Rq T Y o¬cEfëj`XeIEXðGfI8kIGkQ¢XIQVXI9¢pQIWIÅi ¦ ¦ 6 ¦ ¦
ç ¥ s¦ pzæ
¹
¦
6
·
·· ¶
6 ¦
6
´µ ß î ³´´ ¦
ä ¥ ¥à (
¹
¦
¡ ¦
¦w6 ¡ ¦
·
·· ¶
¦w6 ¡
´µ
¦w6 ¡
³´´
ß î ¦
ä (
9
Ú Í × Ë É Í Ì Ê fuu×
Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Ð vÔmQfÔm«xÎ xÎ
° ¬ « § ª « ÂFÕ¤¥sxd § ® © ª Ö ª ® © ¬ « « § © ¬ « ® © ¢ ¢ « ª § ¬ Ô ¨ ª ª ¬ © ¯ ¬ « Ô § Ñ « ¬ « § ª ¬ Ò Ñ © ÎÐ § ¨ § § Í ¦d¤¥Xv¸£sv1q¦Q¤'qFs¤¤¦ÉS£É£d£ÏFsX£dqÕ¦1kdv¦Ó¦dquk% Ï¥¤p¦« sÌ
iDByB T D Y q 5 Y 2 5B D T d H hB Y 2 5 4 2B d H D Ty y Rq H Y R H yyB @ F 4B F @ Y 2 5B D 4 R d D Y 2 5 ¢3EEXI¢Gm¢CI8EGVWQIEWI86eXu6GVX§p6WpsEXCXICuI¢XGIWWI¢Q4 y¢G6IQ8wGEpxVWI
pGCGE9W6sQX¢UIp8swIG9IXIpXQpXEQUwpGEGoE8sfI8EW¢Vb T d H 2 H ` H DB R P q 5 H Y R H F D Y DB 7 d H Y HB d T h 2 R 5 H F D ` 2B h Ry 4 2B hy H n H F D 2B Y D 2B 5 Y R 5B d T H F D D T 2 5B D T 7B 5 d T H F D q 5 H d R D T 2 D 4 T H H F D d 5 D T d H 5 H bB D T bB d H h B d D T 7 H F D 2B iqB 4 H pGjVw8XWVdEcII¢CpGU¢uIGVI6¢pupGv8WVG6s¢8GVVEQIa WGlduIWwXwE6QWY 5 D i DByB T H F t Y HB d T h 2 R 5 H F D D T D H 4 H hy H n H F D 2B D 2B 5 i d H b H D T h HBy T 2 H H Y T W#3XXp¢wpuv6QX¢UIp8swIWlp6QX6UIGXCoX¢s68fl6XEII¢QQ¢UF d 5 D T d H 5 D 2B 5 H 7 T Y H F D D T F D d T Hy 4 YB DB h 2 T h H nB 4 H Y iy H R PB 2 R 2 H H H b T F Y 2 5B DB h 2 5 8GVWQ8oE8sw9¢G
pGlUGC¢gEQfXfEIVkQUQQWQISXpIeQ6s
¢gUfp8XWEII¢Q4 id T h 2 R 5 H F D t ¥ h 2 T ¢III8apGpph pI¢9 pQSj3xqEsëjDoQ88u8WVG6s8CoE8suIWVUGe8EGD ¥ Y P 2 1 $ ¦ ¥ P i 2 H bB ` d 5 D T d H 5 D 2B 5 H F D 2 T F D 2 5B r T 7 d 5q 2B H d 5 7 i d d T 4 l#W8XW89¢gwh I¢w9 pp6SËo68E¢v8GVWQ8 WWVdfIGVUW6XW¢fò ¥ h 2 T ¥ Y P i 2 H bB ` Y d 5 D T d H 5 B d D T 7 H F D D T F D H 4B D 5
Æ
Ê W ß V Ý
È ß É î ¥ à Y Y Y aaxà ¡ à ¡ sÝ à ß
î ¦
ä ä ä
y ¦ t º² ² ²
Hd H F W6IC@
Æ ûV¢r 6xà6YaYaxà xà xà ù ¥ Y 6 6 6 Æ V¢r aax8Q8¦ ù û ¥ à Y Y Y à ¦ à
ç ¥ oh pzæ
y A Çw² Ý sQl² U² eÁ º² t ø ç ¦ à à æ
ß
î ¦
ä
² Ý ¥ (
h 2 n H @ IUuvpt 8 p¥ h 2 T IV psQSjfE¢¢pXWfo8XapG¢p8XWEII¢Q¢III8IVdI6fwsEWQQIwv£1 ¥ Y P 2B Y T t H hB Y D F `B d H F D 2 5 Y 2 5B DB h 2 5 4 i d T h 2 R 5 2 2 T 7 R H ò H B d 4 Y H d H @ q ç ¥ oc pzæ
Æû Ý à Y Y Y à §!Gà ¡ 6aaSà ¡ s!úÝù
ß î ¦
ä y A t º²
Û Y H hB Y D F `B d h 2 T Dq H 6QIEWCo8EWI¢Xy H F D 2 5 Y 2 5B DB h 2 5 4 i d T h 2 R 5 D Hy F 4B dB ã H F D ` 2B hy 5 F d 5 D 4 H b H F D d 5q Y h 2 T D pG)8wI8EGEpI8QVUII8a6XpXWEeIWIEIX¢I8GQ¢dIGw8wIp¢fY y A ~W6IC@ º² t H d H F ç ¥ 89 pzæ
y A ¼w² Ý sgà x8ÔeÁ º² t ø ç ¦ 6 à ¦æ
ß î ¦
ä e² Ý ¥ (
i o² Ý ¢¢EGl¢W6II¢QQuIGWVdEcII¢¢Qav@ q 5 H bB D T bB d H h h 2 5 4 H Y H F D H D T 7B 5 d T 2 T 4 H
ç ¥ o pzæ
·· ··
Å ¹ Ý ´µ ´ ´ · Ý ´´ Ý ³´´ 4 ·· ¶ £Ý
ä ² Ý
Y T ` 2B 2 n H ã Y Hy 7 T H Y R 5B b H d d R 5 H By T d H 2 H @ 5 2 2 T 4 H V² Ý IXpU66QXp#¢pp8Xx6ICI86XXVGQpQ8` sp¢Qur Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u§fj«wÔzmQÔv«SÏmÇ
c
Ý ¥ ¹ Ý Aº ·· ·· ·· ·· Ý ¸¶ 3
º
ç ¥ pzæ ¥
º
º
4 xº ´µ ´´ ´´ ´´ Þ¡ ³´ 4 º S 4
ä ç à à Û æ Ü jQ² º s² ² zokÁ
i F Y H 7 D 2B 5 r T d 5q 2 H bB ` YB Y HB d D 2 oQ9uoX83s#C8e68E¢uXQXGoQH W8IIS8IWEC@ 7 d 5qB 2 R 2 5 2 F DB d 5 D T d H 5 B d D T 7 y T 2 5 ` TB hB d D 4B h 5B d H y T 4B i D i B d D T 7 y y } T Y T 5 D h H d d Hq H d YB F Y H 8GVWQ8 WWVdV¢U8¢¢IEWWvXIxXQS¢gEIx3¬sGV#h¢QU§¢fGa6WQ6XwWQ97 4B h 5B d H T 2 5 Y 2 5B D T R P H d Ty T 4 Y d 5q Y H 7 H F 4 Y ` 2B 4 2 H d H gB h Y D 2 H Y H d H d D T F D B d D T 7 H F XIxXQw#8I8EGVIx6¢¢QWj8Q9QIuIX6IQQsEIWoQW6WI6WVIG WWVdIut
r xs eQ9XpEWQ¢QIXE8uWoX8Co8Xj ä ß î F Y H 7 4B h 5B d H d T H 2By T 2 5 Y D 2B 5 D F `B 6
Y Y aaY
¦
Y c 6
Y 9 6
Y
6
Y $ 6
Y
6
Y
¥
6
¦
6
¦
¡ ¡
c
6 9
6
Y Y aaY
ä ê ä eß
8EGVQEuIXW¢GjGapXI6QQIvIÚoX¢sCIWlwsIIQH E¢wI8EVW¢GuIXQ IS2 2 5B D T 4 5y ` 2B D d T D Y Y DB ` 2B h H 4 H d D Y R Û D 2B 5 H F D D T Y h 2 DB Y T ` 2 5y Y T Y D d T D Y ` 2B d H 7 R H F D H d H F @ d H D D T 7 iyy T H d 2 Y H 5 h DB t Hy 4 dB 4 T q 5 d H D H 7B d H H F D 2 5 D R 5 hB Ty YB F Y H 7 H F pGW6ICf6GWldEE¢guD ÚWQxIEVXQX6VQW69XQIG8p8E¢XeW6#pGD 2 H F ¥ H d R `B ¤ 2B Y T F Y H 7 d Ty R 4 dB 4 T 2 5 iy H bB D Y H ` ` R Y H d 5 7 d 5 t Y HB d D 2 H h H D T H H d F DB Qpªmp¥ puI8Xm#Ej¢#WQ9¢XIQX6u98 EQ¢EWQ¢8IWW89j8SQ6XWWoQ9QGlgQWWEC@ F Y H 7 d T H 2By T 2 5 h H D 2 H Y H d H d H F DB H 2 T 4 YB F t @ 5y H 2 @ 5 F Y F Y H 7 4B h 5B d H D 2B 5 r D F `B Q9v¢QIXE89QGo6WQIQpGE6¢gvXIumusEQCspWW6#XIxXQvoE8s£o8EQH H F D d H hB Y 2 5 Y H ` 2 T F 4 d 5 D T d H 5 B d D T 7 H F D q pGuQIEWI8eQQ¢I¢Uu8WVG6s8GV#IGv¢5 W¢9IW¢QU ¢wI8XWEIp8Qef¢T 7 d 5q H F D t y y } H d T Y 2 5B DB h 2 5 4 i d r h 2 R 5 H F D q1 h H n H d T Y 2 5B DB h 2 5 4 i d T h 2 R 5 H F D F 4B F @ 2B Y H Y T 4 h H D T d D Y RyyB H b 5 T H F xII8CIGs£jQpuWVI8XWEIp8Qf¢UIp8sIW9EICXQWVgQWVGWIEXEf¢s¢fIut
} p tg r j v!¡7
tg Ø eg r p ÙQX!sfd
í Ý q 5 D 2 HB 4 ¨ H 5 sxm¢oQX696xQ4 H F D YB D T F D d T Hy 4 DB Y H ÷ T 7 F 4B F @ D 2 H 7 Hy H F Dq n H F D 2B 5 d H T q 5 2 5B DB h h T H F D H D 5 pGwE9ÓlUGw¢gEQ QV¢#ÆEICwo6#6XQÅWIIWrXeQQVr8EGEpI¢eIWdGVfò ç ¦ ¥ sp¥ pzæ à 4 º à à 4 æ Á ä ²Ý ² Ý ç ¡ a¥c¥¡¥à ef5 ( Y XB ¡ ¥ pj7¢CoQE¢VEISQdGV#uIGCISujQ¢GVWQ8 WWVduIWeXCEpWY ¥ P q 5 D 2 Hy T bB R P H B d D T 7 H F D Y R F t d 5 D T d H 5 B d D T 7 H F D 2B DqB F y¢U8¢¢IWwII8W6W8Qw8GlGQ8aoX8dIWAEwIW£¢ê e6I8XWVgxXwI8EW¢V9IGwVT T 2 5 ` TB h T 5 D Y h 2 5 Y H d d 5 4 d 5 D T d H 5 D 2B 5 H F D 2B DqB F Yr i Y 2 5B D T 4 5y Y R 5B d T b H F D D 2 5B D 4 2 Rq H F D q 5 Y H Ry T b H F D 5 D Y D F `B H @ H F D ` 2B bB ` Y D 2 HB 4 ¨ H 5 4 H F D Y D 2 H 7 R ` d T Y DB d 5q Y T 8EGIIIWm¢u6IXV¢pGGuGo8EQvIG pXxE8CWS6X6#6IGuWS6#p8W¢GEu¢¢UF d 5 D T d H 5 B d D T 7 ` 2B h 2 5 Y H d d 5 4 H F t H bB D T bB d H h D Y d n T d 5q d 5 D T d H 5 D 2B 5 H F D H D F `B 8GVWQ8 WWVdwIEII8W6WW¢Qpuv¢¢EWVVEW6IvWIu88WVG6s¢voE8sIWCo8X97 ç ¡ ¥ pzæ ¥
4 üí Ý4 º ø í Ý ø 4 í Ý4 5Wxfxjxh¥f¡~ø AxÝ ×uhÝ æ í ä íç (
Hy 7 T EI9¢pH d8Uj8EG6pIIW¢6IXV¢uIEW8S8X6I¢#QG#X vSE8QI6WQWfIGvV¢EWVVEW6Ih 5 ¤ 2 5B D 4 2 Rq H F D q 5 Y H Ry T b ` 2B d 5 F `B H 2 q 5 Y 7 d H D 2B ê D 2B 5 H 4 2 H d Hq H d H F D D T H bB D T bB d H H 7 5 Y d 5q 2 H D DB d @ iyy T d H 2 H ` YB d 5 D T d H 5 D 2B 5 98u86GEWC#EE¢G6IQ8CEu8GVWQ8oX8 vQIWQ63se¢XGlQ9XXp¢fQ9GD i 7 H F D 2 H H @D H 2 5B D Ty H d T F YBy T DY H 5 D 2 T D d 5 7B YB DB t Y d 5 D T d H 5 D 2B 5 h 2 T B d D T 7 F D 5 q 5 H Y R Hy T d H hB Y 2 5 4 H ÷ T 7 H @ H 4 2B o¢f8s9X§X§¢Q8GlGQ8§oX8 IVwGV#wW¢sp¢IvEI¢WQIEWI86vV¢#vvQpX
h
Ú Í × Ë É Í Ì Ê fuu×
Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Ð vÔmQfÔm«xÎ xÎ
h 2 T D Y d n T d 5q Y 2 5B D T 7B 5 d T ` 2B 4 2 H d H gB h ry T d D 2 H 4 D 2B 5 r H H d F D h d T h 2 T D Y H F D 2 H F IVuU8p8XWVdEcII¢aIEQI6WQsEp3VGGo6QoX83QWpG#¢UIVfIG#QIª ç ¥ S¢¥ pzæ
¹ · ··
¦
º
º
º º
··
·· ¸¶
´´µ ´´ ´´ ³´
ä ç º à à 6¥c¥UÂÛ
æ Ü xkÁ
i 2 H bB oQ88` YB B d D T 7 D 2 Ty R 4 dB 4 y T 2 5 ` TB hB d D T q 5 H Y T 4 y TB 4 H Y H F t DB H b 5 T H 2 5 H F D q 5 D F `B d H F EkGV# o¢XIQX6¢U8¢¢IEWW¢¢g¢6QWeIug8c¢eI¢dIWu¢wo8EWpGD 5 D D 2 H 7 Hy H H 2 5 ç ¥ H d R `B ¤ H H Yæ h H DqB F Y YB B d D T 7 D 2 Ty R 4 dB 4 T q WfoQ9QEQwI8uo¥ pwWI8Ev 6WÅQWEIfEe WGldo¢pQWEQ9¢5 cW¢gfVUWw6XGVfò @ 5 d F 4 T H D T F D H 4B D 5
ç ¥ ¥ o8¥ pzæ
ä A æS 4 ¹ 4 ä x±¥ ´´µ 4 ä ³´ · åS±¥ ·· ¸¶ åSÄ¥ 4 ä
YB d H h d 5 q 5 B d D T 7 D 2 Ty R 4 dB 4 y T d H 2 H ` D Y 5 7 H F t d H D T E6IW¢v¢dGV#Co¢pQWEQ¢WQI68C¢#uIuj6QGly 7 H F D D R 5 T i T Y 5 D H d 5 7 F 4 R 7 H b T F yyB @ H YB Y iy T 2 T d R 5 2B Hy 5 d y T DB b T i Ty Y H 4B d D T QIW8sVdgWwGu8#dI ¢gUUXECvQEWxEVU¢vI89XCX8U¢xw9gI6QXGV#7 DSVI6WXgo¢GI86W¢aIIVUI8EW¢VIGeI¢X¢wGo6#6XQIW6ICeQ9W8sGV#7 2 Ty R 4 dB D 2 T D Y 2 5 4 H d T Y h 2 T Y R 5B d T b H F D ` 2 5y T Y D 2 H 7 Hy H H F D 2 H F @ h H 7 d 5q t B d D T DSVI6WX6vpGXGV#jXIxXQp¢v6WpIWo¢XQ6sS6Q8c'go¢GI¢Q§¢IWVuGV#7 2 Ty R 4 dB 4 H F D YB B d D T 7 4B h 5B d H T q 5 D H Y R Y y TB 4 H Y T t d H b H @ 5 D 2 T D Y 2 5 4 D 5 2 H d T B d D T 4B h 5B d H T q 5 Yy T 2 5 ` TB h H F D ` 2 5y T Y D 2 H 7 Hy H H F D t Hy 7 T H H F D 2B 2 @ 5 F Y Y T ty T d H 2 H ` 2 XIxXQv¢fE¢U8¢¢IpG#p8X¢uWoQ9QX6 IWh¢EI#VI IWXACcIWV¢WQI68e31
} p tg r j sv£!¡7
F Y H 7 d Ty R 4 dB 4 T 2 5 Y D 2B 5 D F `B W6#¢I6WEQ¢uGoX8Co8Xj
j i ã â t rg mÉsáà
ß
Û ¥ H d R `B p¥ puI8Xm¤
Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u§fj«wÔzmQÔv«SÏmÇ
7
6 8 1
¡ ¥
5 4 2 3
h H D T Ry T b QWVUE¢V6H H 5 D YB H bB D T bB d H h H F D F 4B F @ D T D 2B 5 H F D D R 5 «WeXA8EWVVEQIAIGðXICVoX8AIGp8¢T 6WQWoQQÿ8XW6IpAIWw¢QpX¢Vb h H d H D 2 H 4 2 5B D 4 2 Rq H F D q 5 Y H Ry T H H d F D ` 2B Y R H bB D T bB d H 6WIWIEWI8EWVVEQIh I¢QQru¢Æ8EGV#XgWIp¢T ¢#8XI¢G#8ExcmrWS6W6WI6Wd h 2 5 4 H Y T q 5 2 5B D T 7B 5 d 2 T d 5q Hy T D d 5y i T t T Y D 2 H Y H d H F 4B F @ ¦ ¥ Hy T t d H hB Y 2 5 4 t Hy 7 T H 2 T Y i c ¥ P i 2 H bB ` 2 5B Y 2 T H H F D q 5 H Y XpCpEI¢md6IXI8Q¢Xp#¢pº¢fªpdëjzoº688º8EWI¢IpIW¢IR H bB Y 2 H D H Y H ÷ T 7 F 4 B F @ t H y T D d 5 y i T 8WI6GxQV¢#XICu¸¢Xp¢8Xxgmt 6IGfI8QwGXp¢vEIGfW¢QaW9g#¢f8q T D 4 R d D Y 2 5 4 5 D YB D R 5 YB F D i d d T 4 5 D i T @ h d T @ d 5 D F `B T d D Y h 2 T Hy 7B S¢X¢WGªIVXp#EWY QQpW85 o¢a¢#W8WVG6s¢3oE8sQI6WQsEp3WEII¬pXGIWWI¢Q4 i d H hd i 2 T q 5 Y d 5 D T d H 5r D 2B 5 H 4 2 H d H gB hr H DB 2 n ` 2B D 4 R d D Y 2 5 d 5q Y 2 T H 7 T Y H hB b 5 d D 2B 5 h H n T D R 5 T Y 2 5B D 4 2 Rq q 5 2 5B Y 2 T H Y HB d H Y d 5y i T t H F ¢wI¢Q#6IXxcWI SE8Æ6pp8¢I¢XGIIC¢¬8EWI¢IpQEWQ 8EgmeIut
} pã ø  ö r eã ÷  sËgU£õvËö
À ± · ¼ vÀ
ß sõ
ó ¶ G ñ Ã ² ô¾òÓð±
R ± ³ Á B
ï ¶ ¹ Á ¶ ± À ± ¿¹ ½ ï ¶ ¹ » Á î À » ² ¶ ðgejÅ6¾¡ðgjuejme9Ã
í
ì® £
Y H by H Y 7 H F D Y H 4B d D T 7 H F D 5 D d 5B d H D 2B yy T YB 2 5B D T 7 d 5q 2B YB F D H 4 2B Y Y d 5 D 4 H b 2 5B DB h 2 5 QQ¢E6W9QIWQQEWWVduIWG8EWQWoXhE¢CXd8XWVd8pEuXIWu6IXf8GQ¢8XWEII¢Q4 i d T h 2 R 5 5 2 H d T H d H F D D T F D H 4B D 5 ò iD d H 5 d y TB 4 H Y YB F D q 5 H ` T D 2 T b h T H ÷ T D H @ 2 ¢III8kIWVWQpG VIGd6XGVf836s¢WIv¢6QW EIGC¢¢¢o¢Vp¢V¢ev)85 d H D T ô Y d 5 D T d H 5 D 2 Ty R 4 dB 4 5 Yy T H d T Y d 5 D T d H 5 4B h 5B d H q 5 Y H Y T 4 y TB 4 H Y H Y H F D t iy d T Hy 6GVaQ8GVWQ8o¢XIQXQeX¢ W¢u8GVWQ8wEIxXW6s¢u6W¢Qv¢XQQWQIWh8W¢QX 6 ¦ ¦ ¦ ¦ ¦ ¦ 6 ¦
ç $ ¥ 83¥ pzæ
ß
î çsQà 6Sà8¦fæxÜkÁ ¦ ¦
ä
¹ · ·· ¸¶ ¦
6
³´´
6
´µ ß î ¦
ä Ü ç ¥ æ (
h 2 I¢T ¡ ¦ ¦
ç ¦ à ¦ 6æ Ü sQà ¡ 8w£xkÁ
ß î ¦ ä
¹ ·
¦w6 ¡ ¦
¦
·· ¦ ¸¶ a6
¦w6 ¡
´µ
¦w6 ¡
³´´
ß î ¦
ä Ü ç æ (
7d 5q H F W¢upGD H ÷ T D i H F D t Y 2 5B DB h 2 5 4 i d T h 2 R 5 4B h 5B d H F DB @ h H Y R H d T t ¥ P H H Y t H bB D T bB d H h h 2 5 4 H V¢a6IW¢QI¢XGII86 f¢UII¢svXIxXQGEC6WIu¢o8¦ poëjXQ8¢¢EGl¢W6I#p8Q6WY ¦ p¥ Í Ì Ö Í fu×
ê Ó É Ñ × Ú Ì Ù Ì È Ç ë¬uvvfÚ qÔvwfÔÖ {Ô§Ïmuuóx%xÎ é ÓÒ Ç Ó Ì Í Ì Ñ ÑÒ é Ó Ò Ë Ç è Í Ë Ú Ó × Ç Ð çÐ
YB Y T 6X¢U QGeo QW8CQEW6WY 7d H D i 7 d H D T 2 5 Y HB d H d 5y i T t q 5 Y 2 5B D T 2B 7 5 4 d T H 2By ` 2B 7 d 5q q 5 i T @ D 2 HB 2 H b 2 5 4 T iy 7B Y YB Hy T D d 5y i T ¢Xxgv¢fI8XWVUEI 86¢QIXEdIE#8¢eguoQEIQ¢o8QEp#EWE XIV8Excmt i
í W ß ço8¥ p¥zæ«YaYaY`ø ß V Y î Ý í W ß V Y î ß Ý ¦
í W ß Y çs¦fæoYÉºø ß ç ¦æ Y º V Y î Y sfxkø ¦ Ý ¦ í Wpß ¦ V Y ß î hsÔxkvxº Y ç ¦æ Y º ø í Ý Ý ¦
Y cfxk~ø ç ¦æ Y º
ä
4 üí Ý 5º
Û5Wx£V9¢XWp¢Uj6XW6Wj8ExcmuIWGII8W6WW¢QEI¢pG X©cWjf¢upG8¢EI9¢pH 4 ü í Ý º q 5 2 5B Y 2 T H Y HB d H Y d 5y i T t H F D 5 D Y h 2 5 Y H d d 5 4 Hy T D H F D 2B @ 5 d D Y Ty H F D t Hy 7 T d8UauceIXIp8sQW86 IGjV6E IWGD WQWwIWVe8XI¢UQXQ8Xxgm IGXB 5 ¤ @ 5 d ` 2B h 2 5 Y H d d 5 4 H F D q 5 Dq Hy H F D 5 7 d H D H F D q 5 2 5B Y 2 T H Y HB d H Y d 5y i T t H F D 2 2 7 Ry 5 4 ` 2B h 2 5 Y H d d 5 4 H F D q 5 5 D H F D D 9IE8QIEII8W6W8QuIWë¢8GpGVT W6GIWVS6X6#6uIGmXfGoQ¢'VXIVD 7 d H D H F D q 5 D 2 HB 4 ¨ H 5 4 H F D YB id D 2 H F 4 T Hy T d 5y i T t H F D R H ÷ T 7 t Y ` 2B h T H F H F D d H h 2 R t H 2 5 D Y 5 7 Dq Hy H F D q 5 D F `B d H F D 5 D Y 2 7 Ry 5 4 H F ¢XxgvuIW9pV¢#oQ8IEI¢QICIGjQpIIS¢I¢j89GXCIWs¢S¢XWCIWGI9IX¢QCIut
Y Y Y à ¦ aaxà 8gà ¡ #õ ä
í W ý ß V Y î ß ý ý Ý
t YB D T F D t ê D 2B 5 H F D D R 5 6XCVUWpUoX8uIGup¢s¢T WQW¢QvV8EWIVUH 7 d H D F 4 T H q 5 2 5B Y 2 T H F D 2B Y d R 4 4 5 D T F D d 5 D 4 Tq 2 5 7 7 5 4 H F D Y d T H T H d H F D 2 7 Ry 5 4 F 4 T H q 5 h T H F H F D D T D T F pGXCIQ6ulUG8W6¢#8#986IG¢gI¢uQIW#IE8Q¢QmVVgIIWuVCVIGD H 4B D 5 2 2 H F t Hy T D H F D 5 D 2B D R H 5 D Y 7 d H D H F D Y HqBy 7B Y F 4B F 6XW¢Iº6Iu«ç VXI¢GpGkGoX#pIsAG9WQWpG QXI9X«EIC@ W6G«¢QWoXB 7d H D F 4 T H 5D 2 h HBy B Dy R 7 2 H H Y T QEXIEGEp Q6sf¢UF h¢WVgXIWXUh8I8IWE¢EI¢GaIW¢6EwIWfV9IX86#XB ß î t H Y T 4 YB F D 2B t F ` R 5 F Dy Tæ Hy T D H F D q 5 Dq Hy H F D D T 2 7 Ry 5 4 T 2 dVgI¢#8XWVUSQfIGX9WQWfIWh¢UE¢vVUW6XGVfe!8EGV#XgWIp¢XIWh¢fWIIGo T H T 2 5B D T R P H H F D 2B Y 7 d H D H F D q 5 yy T D T F D H 4B D 5 ò ù 2 5B D T 7B 5 d T YB F D q 5 H Y R H F D i h H Y R T 4 d 5 d d H y T 4 5y H F D YB D T F @ t iy H 7 T 2 d H @ Y 2 T 2 T 4 Hy T D YB F D q 5 i h R D Y T D T F D Y 2 5B D Y H R QI¢Qj8WQpVgxXuIWXVUCV8E6#VUp6QvI¢9¢QXp¢EIGë¢ppGlUGjI8EGfQISP HpG¢I8wWoQW6WI6WEIuëW¢#¢XGfQISeAGCA8EWQWppV#6W9aXI¢GIG¢5 F D q 5 H 2 5 Y D 2 H Y H d H d YB F t ÷ d T 7 2 5B D Y H R P T F DB @ 2 5B Y Y H d H 2 T H H Y H @ Hy T D H F D q 5 D H F D D i h H nBy 7B Y YB T d H `y T H F D q 5 H 7 5 Y D T F D 5 Y h H D 4 R d D Y 2 5 4 YB Hy T D H F 8WeIG)6UEEI9XWEGI68X¢IWf¢e98#lUGkWº6GIWWI86XeEI¢Iut H bB D T bB d H ¢8GVVWQph h 2 5 4 H Y T 5 D 2 5B D T 7B 5 d T 2 TB ` 2 T d ` T ô D 2B 5 r ¥ h H d H D 2 H 4 d 5q Hy T D d 5y i T t ¦ ¥ Hy T I8Q6WeGr8EGV#XgWpI¢¬¢X8I¢W8¢oX83sºQQWS6Q8EI¢G 8Xxgv¬8pEI¢mt
4 4
Y sfSÉº ç ¦æ Y 4 4 Y sWx ç ¦ræ Y
Ý Y sfokº ç ¦æ Y 4 4
Y sfxkº ç ¦æ Y 4 4 Y sx ç ¦ ræ Y
¦
4 hsfxhº Y ç ¦æ Y º 4 hsGS{ Y ç ¦ ræ Y
Ý Y sWx½ ç ¦ræ Y
4 üí Ý Y 5xckº í Ý Y xck 4 í Ý Y ¥Q{ í y ú ¤¦uu t Y î w ú ß
í y ü Aut wüu Y î ß
í y û ¤Auvt w û u Y î ß
í y ú ¤¦uvt w ú u Y î ß
í y xuvt w u Y ß î
í xÝ
ù ä ç4 üí Ýº ø í Ý ø 4 í Ý ×f5Wxjxh¥xSæ
ß
î ¦
ø
í W ß V Ý
Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u§fj«wÔzmQÔv«SÏmÇ
¥
D Y d n T 5 D 2 5B D T 7B 5 d T 2 TB ` 2 T d ` T ô D 2B 5 r UÅ©W«8EGV#XgWIp¢ª¢¢I¢G¢¢oX83l¥
h d T @ ÷ 4 T d 5q Hy T D d 5y i T W¢o¢Ue8EI¢G8Xxgvt
¢8GVVWQph H bB D T bB d H pkEI¢mt ¥ Hy T
4 4
Y sWx7¡ ç ¦ræ Y 4 Y ç xY ræ 4
4Ý
Ý Y sGx'¡ ç ¦ ræ Y 4 Y ç xY ræ 4
í y û u w û Au t Y
î ß
4
Y cx8¡ ç ¦ ræ Y 4 Y ç xY ræ 4
4 UsS'¡ Y ç ¦ ræ Y 4 44 Uç SY Y ræ
¦
4
í Ý Y xck
4 í Ý Y 4 ¥xc7¡ ¥xcY í Ý í y xuu t f î w Y ß
í y ü u wü Au t Y î ß
í y ú u w ú ¤¦u t Y î ß
í y xuu t w Y ß î
í xÝ
í W ß ß î ù ä çí Ý ø 4 í Ý4 f6~hAxf¡ø AxÝ æ í ø V Ý ¦ uXI¢mXCspWuX¢EWVVEW6Ih ¥ Hy T t 2B 2 @ 5 F Y YB H bB D T bB d H D Y d n T ` 2B D 2 H Y H d H d d 5 D T d H 5 ` 2B 4 2 H d H gB h r h d T @ ÷ 4 T D 2B 5 r ¥ T d 5q Hy T D d 5y i T t H F fWUa9IEGoQQIQ8GlGQ8IEQI6W6hI3¢SVUoX¢s£l u8XIV8ExcmIut í W ß î ß V ço¥ pzæ c ¥ ç zd×f5¬xÝ ¥æ ß îæ ä ç4 üí Ý ø í 6 4 í Ý 6 Ý ! ¦ H bB D T bB d H ¢EWVVEW6Ih h 2 5 4 H Y T d 5q d 5 D T d H 5 D 2B 5 ` 2B 4 2 H d H gB h y T d D 2 H 4 D 2B 5 r ¥ d TByB 7 Tq H F D h H bB d H h D Y R Û H b T I86Qw88WVG6s8oX8wIEQI6W6hIh¢WGoQ6oE8s£lj¢XXE#¢IG68WQpuI8gUF H D 2 H D YB Y 2 5 4 7 d H D d 5 d d H H F D H ÷ T 7 5 D F ` R 5 d F D h H hB bB h 2 H H Y T ugoQWEWI86µW6Gu8WQaIG¢Vd#G¢I8WpGQIEI6Q¢IF IWCVUGuGVfò ß î H FD D T F D H D 5 ç 9 ¥ 83¥ pzæ í W ß ¦ V ß Ý î ä í W ß V ß 5 î Ý
¢ £
º
ø
¡ ß î ¦
ä xþ ÿ
Y h 2 n H 2 5 h 2 T ç i d D H 7 7 i Y i F YB 2 T b yyB @ Y 2 7 Ry 5 IIIup8I¢upG#9xoWEI¢VsEXCI9IX¢Q4 2 H b H yy T Hy 7 T H YB F D d 5qæ Hy T D H F D 2B 2 7 Ry 5 4 F Dq n H F D D T Y d R 4 4 Q¢6uXVEI9¢pXIW9¢XIVIG¬XÆ#pX86«GpIW9lWpQQx5 xðWVg9XIW©31 ÿ þ H Y T 4 YB F D 2 5D d Hq Hd h 2 T ÿ þ H 5 W66WI¢t x½dWD d 5 d d H Y HB d H Y d 5y i T t H F D Y T D Q8W66XQWC8ExcmuIGVEB 7 R Y ` 2B F YB 2 T b 2 5 2 D Y d n H F D H D T 2 `B Y H h H IW9IXpWXp¢VS¢IfWUuIWuWVU8EW6Ia
d 5 d d H H F D H D R DB D Y 2 5 4 5 d H 5 D 7 R Y D 5 2 5 h D T F D Y 2 7 Ry 5 4 H F Q8WQuIWGpWEWI86#Q6#W§IWCVI#pCVUGp#IE8QwIut
¹
6 ¡
³´µ
·¶ ¡
û ¦ QQà S8¦ ws¥c¥UýS68E¢uXe8EGpE8WwIut 6 à ù 6 ä û º à à ù i 2 H b B ` Y B 2 5 B D R y 5 Y H F ¡ ¦ ¹ º ³´µ ¹ ¦ ¡ w6 ³´µ ä ¦ ¦ ·¶ ·¶ ¦ ¦ ¦
2 5B D T R P H H F D iq YB D T Y H @ qB 5 d H H d T Y 2 7 Ry 5 8EGVIx6uIGfXWVGuvm#QQ¢CI9IX¢Q4 H H d F D D Y d n H F D q 5 Y 7 R Y H F D D T F D @ 5 F Y iyB Y T H 2 T 4 H 2 5 d H H 5 D Y 7 R Y H Y H F D t 4 h 2 6WIWwfWUIW¢u9IWIWVUWÏcIWEX¢gA¢Q#I#IQQGu#I#QIW¢AI¢T tëpj¢wQE8IQ8WpaIWSëVQW¢I¢fo8EW9GC6Eµ8WQQ6IwvUxIG#wIGm¢5 t T q 5 H 4B 5 F 4 d H 5 d H F D i t H 4 d 5q h 2 T D F `B d 5 D Dq Hy 7 5 dq h H H 4 5 d H @ t h 5 F D H 7 H F D q i 4 T d R 4 4 T q 5 d H h d 5 H F D H B 7B T 7 5 t Y 2 7 Ry 5 4 H Y H F D q 5 F 4 T H q 6¢WIQ6¢u¢6IW¢#IWa6X9Xl#§QI9IE8Q9W6IGv¢VgV5 IWIWw6IXI8 7 R Y H F D d H hB Y 2 5 ¥ 8¥ Í Ì Ö Í fu×
ê Ó É Ñ × Ú Ì Ù Ì È Ç ë¬uvvfÚ qÔvwfÔÖ {Ô§Ïmuuóx%xÎ é ÓÒ Ç Ó Ì Í Ì Ñ ÑÒ é Ó Ò Ë Ç è Í Ë Ú Ó × Ç Ð çÐ
ç pzæ ¥
ý Ýç ßæ ý Ôs8S
ä ¾ý
(
ä ç ßæ jsmÝ
i 2 H bB ` YB y TB 7 5 2 iy 5 2 5B D Ty 5 d H D 2B 2 TB ` 2 T d ` T ô H F o68E¢uX¢98IxE¢s¢XGl8W6GoXV8IVG8¢aIut
}ã g e i ÷ã e d i eg j ã e y r p j i ' p j £#¨uõhfæh£7£h{!s¡ÉÂ¡og
r p % Ø i  p " i  r "  £sÙ&u¼U$h#Ó!
ß õ
YB Y iy T 2 QEWxEVU¢T y T 4B d H 7 R 2 2 5 Y D H D i 2 T 7 2B h H Y Y R 4 YB h H d T Y Ty R 7 d 5q H Y H F t Y Ty R 7 d 5q 2 5B D Ty 5 d H D 2 ¢gEWQ9ISº8dGxWÆo¢#«EªQWIQXIW¢9¢XI 8W6Iuÿ6¢p W8Å8EGVX8WQWoXB q 5 @ HB b q 5 D 2B 5 H F D 7 5 dq YB D T F D t i T @ D 2 H d H gB h iy D F `By Y T 2B D 4 H Û R Y H F D F 4 T 5 d ¢§6Xx¢oX8IG¾8WXlUG¢goQQsEIEWS¢XXAÆEdQ¡IpG¬¢oWpI¢T Y R D Hy t YB F D 5 h 5 D d H h d 5 2 1 h H By T d H 2 H ` d H F D d Rq H 2 T 4 DB D T F D D 4 Tq H F D YB t d H b H @ 5 I 6Xj6XIWeIeG 6IW8¬3AQ6XE¢G6IQ86IGfIÆVg VUW 6¢pG XvQ686vcIF tgo¢G89X8 QI¢XWpQ9XIQp8XIfGQpIQWwIVw6GV#8G¢v6Gpp#86wEI8Qm£1 D 2 T D d 5 7B H d 5 Y 2 5B Y 2 H 7B h d H F `B F 5 D h H h 2 H D H h 2 T h H D T 7 5 D R T d H D R 7 5 4 H hy R 5 4 D ¢8GVVWQpo¢8jQI8o¢h¢W8WVG6s¢joX8ë¢GWS6Qj8pQv6VWxW¢o¢US¢f8q H bB D T bB d H h i 2 T d 5q d H h d 5 i 2 T q 5 Y d 5 D T d H 5 D 2B 5 y T d D 2 H 4 d 5 t h H @ H ÷ Y t h d T @ ÷ 4 T t h d T @ d 5 h 2 n 5 D h H Y R H 2 T 4 Y Y H 4 5 d YB F D iy d T Hy d H h d 5 2 H bB ` T q 5 Y 2 5B D T 7B 5 d T H 4 R h 5 d IUGÅQIdÅ¢Q W6QxWIaXIWAW¢QXQQpW8Å68E¢¢uI8EGV#XgWIp¢eQpII 5 D Y Hy T t d 5y i T t q 5 H Y R H F D F ` R 5 d F D h H 2B 7 d H D H h H 5 D Y D 2 HB 4 ¨ H 5 4 H d WaQEI¢mw8Excmu¢dWIpGr8p8WIWrQpX9WQW6I#sGwWS6X6#6d¢T IWdW6IC@ H FD Hd H F ç ¦ ¥ sI pzæ
ÿ þ xfä
üí xÝ
Ü ¾
ø
í W ß V Ý
Y T Y D 2B 5 ` 2B d 5 F `B H 2 ¦ ø ø q 5 Y 7 d H D 2B D Y T ¢CGoX¢sIEW8S8X6Iuð C¢u9W6GeXVg4 H 2 T 4 ê D 2B 5 hB d ` D T H bB D T bB d H h F D þ H F D 5 D 2 5B D T 7B 5 d T H 4 2 H d H gB h T ty T d H 2 H ` 2 A¢guSE8AEW8ul8EWVVEQIAGIWW8EGV#XgWIp¢d6IQQsEIez¢G6IQ8A31
} ã â r e © p t i p r p ng l § e i eg j  ¥gã  r p i p Ç£Uuvu{fD¨æk7v¦£su8×¤
ç ¡ pzæ ¥ ç
fjx¢©h¥x8Å6 ß îæ ä ç í Ý ¥ ø 4 í Ý ß Axæ
î 6 í W ß V í Ý ¦ Ý
ß oõ
Û H bB D T bB d H ¢¢EWVVEW6Ih D Y d n T q 5 2 5B D T 7B 5 d T H 4 2 H d H gB h h d T @ ÷ 4 T d H h d 5r h 2 5 4 H Y T h H bB d H h H b T F H @ Y R F fWU¢)8EGV#XgWIp¢QI6WQsEp«WVo¢U6IW8£I86QA©Q¢EW6I8gUISut ç h ¥ o¥ pzæ í W ß ¥ V Ý î ß ä í W ß ¢ ß V î Ý 4 ¡
6 6 c 4
ß î ¦
ä xþ ÿ
Û 7 d H D d 5 d d H 2 5 B D T 4 2 R d D ` 2 B h T H W6G8W6¢XG lgIIG IEI¢QXy H F D Y H hB b 5 d 2 7 Ry 5 4 F D d R 5q H F D H Y T 4 YB F D 21 û ¥ à 6 à pG#6IXxcWp©9IX¢QÆGp8IW¢g9XpG©3«VGIfS8¦ ù 6 uÃÃà ¸Q88«EIC@ ä û à 4 ù Y H bB ` F 4B F
¹
·¶ ¡
¡ ´ ¦ ³µ
¦ ¦ ¦ ¦
¡ ¹ ¹ ¡ ä ¡ ³µ 4 ´ ·¶ ·¶ ¦ qB 5 d H 5 E9WQ6GD
³´µ
7 R Y Y 2 7 Ry 5 4 H H d F D D Y d n H F D H 7B D YB F IuI#pX86w6WIWCWIupG9XWCEIut ¢¥
Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u§fj«wÔzmQÔv«SÏmÇ
DSE8up8¢u6XW6C8ExcmIEWIQpI¢UufI X¢XpC8«CIVuÆuGoX8 2B 5 D R 5 T Y HB d H d 5y i T t ` 2B Y R h H h 2 T H H D Y R 7 5 Yy T F 4B F @ t ¦ ø ê h 2 T ¦ 6 ê Y D 2B 5 DlCQ8GVVWQpw¢QIIEQIEWE¢fpIhhQGo6WQIQ#oX¢sVu¢EGl¢W6IuIGEEp8¢I2 T Y H bB D T bB d H h H d T h H h Ry 4 2B 5 Yy T D R t h H D 2 H Y H d H d ê D 2B 5 D T H bB D T bB d H h H F D YB iy 2 5 D 5 H d H ¥ Hy T t q 5 5 D H F D D T h H D T d D Y RyyB YB Ty R 7 d 5q Hy 7 T H 2 i ¥ P 5 D Y R 5 ` 5y T 2 Qf7p¥ puXp¢vëV8GIWVwQGlGGfIXEXmXXI W¢vXp#¢p9fmp8 oëQWmI¢88XVU¢T ç $ ¥ 8Â pzæ
ÿ þ x`ä
üí WxÝ
Ü ¾
ø
ü í W ß 8 9 ¾ V Ý
7
t B t Y D 2B 5 ` 2B d 5 F `B H 2 D T Y H bB D T bB d H h H h Ry 4 2 ¢H pQGoE8spXW¢sS8EQICV68EWVVEQIIpXQpXB 2 T 4 ¦ ¥ P q 5 2 5B D T By T d H 2 H ` T t ¥ P ¢QI mëjk¢8EGVQXXVGQpQ8eI8 p§ëj ¢6W¢qpIwW88XWV8W6GoXWE9W6f 2 5 h H Y T & Ty R 7 d 5q 2 5B D Ty 5 d H D 2B H DB 7 d H T q 5 H Y R H F D 7 5 dq ` 2B YB d T H bB D T bB d H h D Y d n T d 5q H 7 H F 4 Y ` 2B 4 2 H d H gB h H 4 T Y xy y {y 2 ¢9WIIWp8WpXWEW¢¢EWVVEW6IwUu¢99QIIX6IQQsEI9Q¢IWéÆ¢T d 5q Hy T D d 5y i T t H F D D H 2 d H hB Y 2 5 Y Ty R 7 d 5q 2 5B D Ty 5 d H D 2B 2 TB ` 2 T d ` T ô 7 5 dq Y d 5 D T d ¢wXIV8EgmIGuxIu6IXI86¢IwW8¢XGl8W6GoXV8IVG8¢p8WW¢GVWQH r 5 H 4 2 H d H gB h D 2B 5 xy y } h H D 4 R d D Y 2 5 4 Hy T D d 5y i T t T q 5 Y Hy 7 T H Y R 5B b H d H F 8w6IQQsEIoX¢sé6s©6G6pWGfI86aEI¢u8Excm#¢QXp#¢pI8Ex6WpwIut D H 4 2 5 4 H F D iy 2 5 h H H 2 H @ H d H F D R t Y h 5 F D H gp6QI86uIG#EI8Q6IuuW6ICpIU6IpG697 y¢gEWQ9IS8u6QI6WQWo¢dXeII¢wseVgCX¢X#¢IxE8W6IGj¢e¢XWIQXIwG6EI T 4B d H 7 R 2 2 5 Y H 4 2 H d Hq H d i 2 T 7 2B h 2 R 5q H 2 T 4 Yy TB 7 5 2 iy 5 H Y H F D q 5 2 5B Y Y R 4 YB h H D Hy r 7 5 i H D T D 4B h Y 7 Hy 5 d H F D Y T h H h Ry 4 2B H hy R 5 4 Y H bB D T bB d H h d H h d 5r d H F `B F iy Y R 5B b 864 vVGVG6Xp#6XI¢WIuIW¢QpIX6IXuXI86Q¢EWVVEW6I6IW8£f6I8Xp EI8Eo ç ¥ S8 pzæ ý W ß ç ßæ ý V so Ý
ø ý Ýç ßæ ý ëÔsoo
ä ç ßæ jsvÝ
2 5B Y Y H d H H F 8EWW6WIwpGD ` 2B 4 R h 5 d t h H Y R H d T H bB D T bB d H h D Y d n Y DB h 2 T 2 5B D 4 2 Rq H F D q 5 Y H Ry T b H D H d 4 YB h F 4B F @ 2 pXQpII6WI9WVd¢EGl¢W6IwfWUuGrIV)8EG6pI9pG¢wQpX¢VG6QXI)EICkXB H Y T 4 H F D d 5q YB 2 5B D T d D Y RyyB d R H 4 T Y 2B Y D 2B 5 2 H bB ` D T 53 } 1yx 1y } xy 2 5B D 4 2 R ¢gfIG8X8EGVWGpXXEvpvVQ¢U#EWSE8dQ¢E8vVu6z4¢2z¢l2xw ¢e8EGIIq H F D q 5 Y H Ry T b H Y R Y Ty R 7 d 5q H DB 7 d H t ç ßæ Ý d 5q y TB 7 5 2 iy 5 T D 4 R d D Y 2 5 4 5 t 2 5B D Ty 5 d H D 2 pGhV6IXV¢CWpj¢IwW8WE9WQ¡sm8S¢98Ix8su6IGp8Qam8EGVX8WQWoXB 2 TB DB 7 d H ` 2B Y R i D R 5 T D F ` R 5 d YB F 4 T 5 d T 2 TB ` 2 T d ` T ô H F D q 5 2 5B D T By T d H 2 H ` ¢WE9WQ{pXWpS8sVo8p8WIE#¢8WIIV#V8IVG8¢IWh¢8EGVQXE¢G6IQ8ªi Y 2 5B D T 7B 5 d QI8EGV#XgWpI¢T 2 TB ` 2 T d ` T ô Y T 5 D h H dd Hq H d H d ¢8p¢G8V§a¢GQWQWW¢T p§ëjå8WAQ¢EW6I9sA¢gwVIGa6#6IW9QpQW6hIh ¥ P 7 5 dq h H bB d H h H 2 T 4 D T F D Y H 7 H F 4 Y H 4 2 H d H gB H DB 2B ¤ h H D 2 H Y H d D Y R Û Y 2 5B D T 7B 5 d T H 4 2 H d H gB h H DB 2 n H F D H bB d H h H d H @ t D 2B 5 H D H d GIXv)QWS6W6WIfIÉI8XWV#XgWIIVQI6WQsEpGIUdIW8WQpQWvvgoE8seW6Q4 r YB h H D TB d 5 d T H F D D T Y H bB D T bB d H h H Y H F D H D T Ry T b H h 2 T t F Y H 7 h H 4 T YB R P H 2 T H Y 5 7 fXpdGlWp8WIp¢IGwVQ¢EWVVEW6IeW6IG#GVUE¢V6ÅIVvhWQ9¬Q6¢UWEISQÅ¢dW89XB tsm¢a8GVVEQIep8Q6W8UwIWw¢¢G j£Q6 IGXaIX¢VwW6QWEI6IG¢o¢T ç ßæ Ý q 5 H bB D T bB d H h h 2 5 4 H Y d 5 D Y d n H F D H ÷ T D H @ q1 D H Y H F D 2B H Ry T b H D H d 4 YB h d H F D 5 i 2 Y H ÷ T D ß 2 H F @ 5d H h 2 T t ý 6¢¢G©¬QICeQQÅp¢¢ëß ä Æ6ICI8 EdëÆ¢g¢oQEQ96xQepG VIGd6XGVfò ß 2 H F @ H 2 5 YB ý Ý F 4 T H q 5 D 2 HB 4 ¨ H 5 4 H F D D T F D H 4B D 5 ç ¥ ¥ oo pzæ ç )'£Qç )U!æ ß 6 4 ßæ ß 6 ä ß Ý ø ç ß 6 4 ßæ ç ß 6 ä ß s)'£Qs)U!æ ßæ ç4 ß 6 ä ß ßæ ç ä ß 6 4 ß çW£)6 Q£)U!æ 4 ß ç ä ß !r6 Q!)7£æ 4 Ý ä Ý ä ç ßæ ß æ ç ß 6 ä ß £¬ø sr6 Qsr7£æ !ªsmÝ ç ß s)6 Qs)U!æ ç ß ßæ ç ß 6 4 ß
Y D 2B 5 h H 4 T YB R P Hr 2 5 2 F DB @ ç 2 5B GoX¢sQ6¢UWEISQ£8IGCë8EGD rl8¢WGxu8A8XWV¢sQGoEwEGVWI¢USC8wXI W¢V8IVG8¢uEI#EWwIGD 8WQ¢VD Ty 5 T d D H d 5æ 2 5B D Ty 5 d H D 2B 4B D T d h T R P d 5q Ty R 7 d 5q 2 TB ` 2 T d ` T ô Hy 7B Y H F 7 5dq 2 H ÷ T H 2 T 4 ç ßæ ý H F D q 5 2 5B D 4 R d D Y 2 5 4 H F t ) H H d ` H h q 5 ß 2B Yy TB 7 5 2 iy 5 H d T ç ßæ ý H d H F ¢gs8opGvV8EGIWWI86uIuv)06W86IvVeeXX¢X#8p8w¢us8SW6IC@
$ 3¥
Í Ì Ö Í fu×
ê Ó É Ñ × Ú Ì Ù Ì È Ç ë¬uvvfÚ qÔvwfÔÖ {Ô§Ïmuuóx%xÎ é ÓÒ Ç Ó Ì Í Ì Ñ ÑÒ é Ó Ò Ë Ç è Í Ë Ú Ó × Ç Ð çÐ
ç 9 ¥ 8Â pzæ
í W ß ¡ V ¦ V Ý î ß
ä xþ ÿ
F D B Y H F D Y 2 5B DB h 2 5 4 H Y H F D d H h 2 H û ¦ à ¦ à ¥ 6 à W W#IGuI8EGEpI8Q#W6IGw6III6g8gpfxà ¡ p¥ ù
4
¦ ¦ ¦ ¦ ¦
5 D Y 7 R Y 2 7 Ry 5 Gu9IWe9IX¢Q4 " ä û þ à º à à ù 2 5B D Ry 5 Y H F D ` 2 B b T Q¥à s¥c¥U¸«8EGpE8W#IGIXxgUF ¡ ¡ ¡ ¦ ¡
·
¹ ¡ ´µ ´
¦ ·· w6 ¡ ´´
¡
·· ¶ ¡
³´
· ·· ¦
¹ ¥
´µ ´´ ³´´
þ ¹ " ´µ ´´ ä · º ´ ³´ ·· ·· ¶
·· ¶ ¡
6 ¥ 6 ¦ ¡
¦ ¦w6 ¦ ¦w6 ¦
2 5B D Ty H d H F D iq YB D T Y DY R 7 H @ t i 4 T d R 4 4 T q 5 d H h d 5 H F D H B 7B T 7 5 ¢XGlQuIG#XWVGfI wI¢6¢WIQ6¢V6IW¢wpGf6X9Xl##§t H bB D T bB d H h D Y d n T 5 D 2 5B D T 7B 5 d T 2 TB DB 7 d H D 2B 5 r ¥ y T d D 2 H 4 d 5q Hy T D d 5y i T t ¥ Hy T ¢¢EGl¢W6IWI9W8XWVdEcII¢¢WE9WQpoX83s¢GWS6Qu8EI¢G8Xxgvp¥ puEI¢mt
4 4 4
Y sfSÉº ç ¦æ Y
Ý ç ¦æ Y 4 Y sÔxkº Ý Y sWx½ ç ¦ræ Y 4 Y sÔxXþ ç ¦æ Y 4 4
4
Y sfxkº ç ¦æ Y Y sx ç ¦ ræ Y 4 hcfxpþ Y ç ¦æ Y 4 4
4 hsfxhº 4 Y ç ¦æ Y º 4 hsGS{ Y ç ¦ ræ Y þ
¦
ä 4 Y sfxhº ç ¦æ Y 4 ä 4 Y sGS{ ç ¦ ræ Y 4 Y sfSXþ ç ¦æ Y 4
Y sWx ç ¦ræ Y Ý Y sfxpþ ç ¦æ Y
"
í xuu t y w Y ß
î ø í xÝ
" ç ¦ ræ 4 Y cGxY í y F u w F ¤Gu t Y î ß
Ý Y cxY " ç ¦ ræ 4 í y ü u wü Au t Y î ß
" ç ¦ræ 4 Y sWxY í y û u w û ¤Au t Y
î ß
ß
î ¦ ø
4 ç ¦ræ " 4 YhsWxY í y ú u w ú ¤¦u t Y î ß
4 üí Ý Y 5Wxckº í Ý Y xch 4 í Ý Y Axc 4 ü y w u t þ Y ß 5í xuv!q î í !xut î y w u Y ß 4 w Aí y xuu t " î Y ß
ù ä ç4 üí Ýº ø í Ý ø 4 í Ý 2jf5zv~¥xSæ
45üWí Wß í ß W þ V ø V Ý Ý
4Aí Wß " V Ý
@ 5y H h H 2By D R 5 Y T F 4 T 5 d usEQeQIEXWp8C¢¢oWpI¢T Hy T t d 5y i T t H F D ` 2B Y R h H 2B 7 d H D H h H D Y R 7 F 4B F @ ç T d H `y T H F D iqBy 7B Y 5 D H 2 5 Y EI¢m8XxgvIWApXWp©6IX9W6G6p#I ÆXpCaoGI68XVIWkEXp#EWAGI8¢T 2 H ÷ T D YB D 2B 5 ê H F D 2 5 D 2 HB 4 ¨ H 5 4 H F Dæ Y D 2 HB 4 ¨ H 5 4 H b T F @ 5 2 Y 7 d H D H bB D T bB d H h H F QV¢ EoX89IWÆ89oQEQ96xQIWGoQEQ96xQ8gU§cI#QW8EWVVEQIIut ç ¥ oo pzæ
í Ê DEW b ä ¾B ï ß V ï ß sæ 7ì CÈ ç ß î õ ï ¦ ï Ý
@ Aä
ý ü Ií W ß V Ý
Û c ¥ P dpëëj 2B 2 H bB ` 2 5B Y 2 T H Y HB d H Y d 5y i T t H F D q 5 2 5B D T By T d H 2 H ` ` 2B @ 5yy 5q H F D Y H dB R P H d YB F t EkQ88k8XI¢UQXQ8XxgmIG¢¢XGlgXE¢G6IQ¢dIECsEX8 pGQXIS6WXpuaUê Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u§fj«wÔzmQÔv«SÏmÇ 8¥
° © ¢ ¬ Ò « e ¨ § « e § q i ssd¦qqF$¦sA rh pÐ ¬ ¦ Ð ¤Avg¦sqÕ¦¥UadÕu%q¦ÚvFÕ¤¸q¦£1¤Aqq¦vfq¤ÉddqÕ¦¤S¦£¦¤o¦q¤¦£8d¬ ¦a ¨ § « e Ñ « © Ò ¥ Ò ª ¬ Ò « © ¬ « § ¬ « ª ¯ ¢ ¬ Ò « e ¨ § c ¬ ¬ § ¬ « ª ¢ « § ª ¥ ¨ § © ª © ¨ § b
h
Y d H by 5 Y h H h 2 T D 2 HB 4 ¨ H q 5 H Y R H F D i h H D 2 H 7 Hy 7B H 2 T 4 i H F D 2 H F @ y T 4B D 4 T d H QQ88WQIp¢UoQX696m¢aIuIGdSe6GoQ9QEI9Xus¢Qe6IW6ICh¢QXG¢Gpw 2 T 4 Y 2 5 B D T 7 B 5 d T `H h T Ü d 5 2 T B D B 7 d H t y T d H 2 H ` 2 1 h H Y R i y H h B @ h 2 T t 2 R d 5 D D 2 H B 4 ¨ ¢QuI8EGV#XgWIp¢w¨¢v8¢XGE9W6f!z¢G6IQ836WIeQIECIVshIW9GoQX696H t H h 5 4 5 D Hy 7B Y YB F 4B F @ t d H by 5 ¢IxQkGXp#EW EÆEICjI688WY ¢I88¢XIXGf¢I¢Q#roÅII8wEB y T 2 5 ` TB hB d D T q 5 Y 2 T H 7 i h 2 R 5q Y ûç ¦ à à ¦æ Á 3sgIW8feýù 4 q 5 2 5B D T Ry T b H H F t y T 2 5 ` TB hB d D T H Y T 4 YB F D 2B t F D hB @ h 2 T yy T 7 Y T ` 2B b T F Y H 4B d D T ¢¬8EGVUE¢V6puº¢I88¢XIXGAW¢Q#EIGÆXuWpXCpp¢UuXVdr)IExcI#6QXGV#7 h H h 2 T q 5 7 d 5q H F D 2B d T H T i H F t Y 2 5B D T 4By T H 4 2 H d H gB h H DB 2 n 2B 2 5 7 7 5 4 i d H b H d QpI¢Uh¢ÏW¢CpGX¢gI¢#Ium6I8EGVQXXpI¢fQI6WQsEpWEIIE8#986#fQ¢C¢T Y d 5 D T d H 5 H Y H F t Y 2 5B D T R P H d T H 2By q 5 D H Y h Hy R 5 4 T q 5 2 5B D Ry 5 Y H F D Y HBy 7B d 5 D T d H 8GlGQ8QIuQp8XWVUSQu¢QIXEu¢Wr6XIp8Qu¢¢XGX8dIWQXEI9Xw8GVWQ85 B d D T 7 H Y d H b 2B 2 i 2 5B D T D H d d H D 2B Hy 7B Y T 2 H bB ` H 2 T 4 yyB D Y DB D R t d 5 D T d H 5 H Y d H b 2 GV#wWW68oX8EGVG6WpWQWoXXp#EWQ¢E8u¢QXEGfCEfpIUQ¢GVWQ8wWQ¢oXB 2 T q 5 2 5B D T Ry T b H H F D ` 2B h 2 T 7 H h YB DB H 4 2B Y t Hy D R Y H d 5 7 YB @ 5 d h 2 5 4 H Y H F D q 5 ` 2B 2 T H ¢m¢d8XWVUXV¢IG pXIIVd6IXEQpXWUVXWpIWW89XÿcWI86QWpG§¢ IEI¢g97 H F t Iu§²W Ý gG¢ WVpQ9WQWoX IWEEIQ IGaW8WIVQp8XWEII¢Q¢III8 IWI¢T i T d d T H D TB h H 7 d H D 2B H F D 2B Dy R Y H d H F D H d 5 D Y h 2 T t Y 2 5B DB h 2 5 4 i d T h 2 R 5 H F D h h tgQI6W6hIC² Ý gWW¢8W668pG¢V¬YxgG¬I9XWd£1 6¢gEQºXp8IWk6IQIS6WY DB H 4 2 H d H gB h t i T d d T d 5 D 4 H b H F D H ÷ T D X Y i T Y iy 7B Y D d T Hy 4 H hy R 5 F Y H 4 2 H R P H YB F D 2B d 5 D 4B h H d H F D q 5 ` 2B 2 T H 7 H F D t 2 5B D T D 2 H 7 Hy 7B y T 4B D 4 T d 5 D D 4 H Y H d F DB EIGX 8W6EIQIeIGC¢IXIVg9IG§8EGVGS6#6XI9Xv¢gEG¢GIG66sQWÅWEª ç ¦ $ ¥ sdpzæ ä ä
²Ý ( ²W Ý
ûç ¦ à à ¦æ ù W² Ý ÔcgIG8ÔeýÁf 4 ß y A Çw² Ý cgà ¡ 8w£Á î º² t ø ç ¦ à ¦ 6 æ ¦
H 4 2 H R P H Y d 5 D 4 H d d 5 4rd 5 D 4B h H d 6IQIS6W8W66WW¢Q3Ô8GXI6WI pGeoQW6WI6Wu¢QEIC@ H F D i h H Y Y H d H H d H 2 T 4 F 4B F ç ¡ dpzæ $ ¥
¢
ß y A Çw² Ý sQà ¡ 8w£eÁ î º² t ø ç ¦ à ¦ 6 æ û ç ¦ à à ¦ æ Áù ä ÔsgpG8f¸Uð² Ý ( 4 ¦
5 D D 2 Hy T bB R P WuS6X¢VISQH YB YB F D iyy T 4B D T 7 H F D T EfEIGdEXVgXWV#QIWV Q8#EeQQ ¢cIfI¢XGII86¢UpI8fXIEWEkXICXB h H Y 5 7B 2 H H H b T F Y 2 5B DB h 2 5 4 i d T h 2 R 5 D Hy F 4B dB ã F 4B F @ 2 ç h ¥ o3 pzæ
y A t º²
ø ç ¦ à ¦ 6æ w² Ý cgà ¡ 8w£Á
ßî ¦
ä e² Ý scgIG8ÔeÁ (ç ¦ à à ¦æ ¦
YB c ¥ P d 5q 2 5B D T D 5 2 B d D T 7 h H h 2 T X3 pëjp88XWVVI WWVdQpI¢U i Q8WVG6s¢9WWV# §¢5 W8pGdEpG699¢WIGuQWpphEeI6IQI#¢Q WdVgH d 5 D T d H 5 B d D T 7 T q 7 d 5q H F D 2B h 5 F D H 7 H 7 T Y H F D Y Y H d H H @ qB h 2 H F H d 7 5 4 5 D i Y T H DB R P YB 2 5B D T R DB Y H F D t d H b H @ 5 h HBy T H 2 T 4 c ¥ P @ 5 F D Y R Û D Y d n D T D 2 H hB b H D 5 WEISE8EGVUWEIGëQ686vcfhQXEII¢u¢g pIjfcICI'fWUlS6IXxC¢I2 iy T 5 d YB DB d 5 D T d H 5 D 2B 5 T q 5 7 d 5q H F D 21 H h 5 4 d H D R 7 5 4 T 2 7 H F D D 2 H 7 Hy 7B 5 EIVU8WpjXEj¢GVWQ8oX8wh¢Ï8uIG3VI6QGI#¢Qu#EB QIWvoQ9QEI9XwGD d H F D 5 2 T H DB R P h 2 T D H 4 2 5 4 2 TB DB 7 d H H F D ` 2B Y R Y h 5 F D H 7 D 4 R d D Y 2 5 4 5 D ` 2B F D H 2 5 YB D 6IGVI¢aGEpxI¢pQ6I8QeVG#Q{IW#IEWIIxIG#f6pWGfI86dGIEIGuI8CXf£1
}  ã â r e © Up Ø  d § e i e g j  t g ã y y T ã  t g j t  Ç£UuVfD¨ih7vs`QvÚr d
ß ß õ
T y R 7 d 5 q R} P T Y T 5 D h H d d H q H d 5 Y y T Y B Y B F I 8SouQ¢9GQWQWX¢CEEIut ç c ¥ o3 pzæ ç ß îæ ä ç4 üí Ý ø 4 í Ý 6 zdff5Wx¬¥xfæ ßî ¥
6
45üí W ß V Ý
ø
í W ß 4 Aí W ß V ¬ø V Ý Ý
Y T h HYY Hd H H 2 T 4 h 5 FD H 7 H F D h 2 ¢k6WQWppus¢QxIW69wpGI¢T
9 3¥
Í Ì Ö Í fu×
ê Ó É Ñ × Ú Ì Ù Ì È Ç ë¬uvvfÚ qÔvwfÔÖ {Ô§Ïmuuóx%xÎ é ÓÒ Ç Ó Ì Í Ì Ñ ÑÒ é Ó Ò Ë Ç è Í Ë Ú Ó × Ç Ð çÐ
YB Y iy T 2 T d 5 d d H d HB d R 5 ¤ ` 2B Y 6XWx¢U¢C8W6uQXI8U9IXIR h H 2B T D 5 H 2 T 4 H 7 H F 4 Y H 4 2 H d H gB hr H DB 2 n T q 5 d 5B b T F H d 5 d d H H F D D R 5 T 2 5B D T 7 d 5q 2 6IX¢Gp8¬¢ge9QpW6IQQsEI£GIUf¢ ¢XxgUQ 8WQepG p¢s¢8EGV#W8XB d6IGfIUj8EGpEW6WQp96G#EXWXVauQpXxcWIvEx¢IW¢Q#sI6IwXXIWEXIªvVXIVD H F D d R ¤ 2 5B D B d 4 Y H h h H DB 7By iy dB Tq T Y H hB b 5 d DB t H d R Y T H 7 y Rq H Y R T YB YB F D HyB F Hy T d¢Xxgvp8WeQpX9WQW6IpWQW8WQepXI¢QXIWvVQI8IG¢9W6GXrQWQIH 5y i T t T 7 5 dq h H 2B 7 d H D H h 7 d H D d 5 d d H ` 2B h T Hy H F D q 5 d H h d 5 H F D q 5 Y 7 d H D 2B h H Y Y H d 2 H Dq 5 YB d 5 D T d H 5 2 T q 5 i 4 T d R 4 4 T H F t Y H 7 H F 4 Y H D T hB h 2 T 4 H F D q 5 i 4 T d R 4 4 T H F D Y Y H Y Y T 5 6G¢Xj¢GVWQ8#V¢¢GI6Q¢fIuQ6#6IWCWVUXpI¢gIWs¢9¢GI6Q¢IWvQW¢aGD Hy T H D Y R 7 H 2 5 2 5B D T 4By T 2 H bB ` T d 5q H 7 H F 4 Y H 4 2 H d H gB hr H DB 2 n T D 4 Hy H Y 5 D d H h d 5 2 XI¢wsfp ap88XWVgEXIIVQ¢E89¢a9QIa6IQQsEI£GIU66X6W9Gf6IW¢31 ² ¹ ² ó ´ g¦¶ ç $ $ ¥ 8dpzæ
G À Ã À À À ±¹ À î Ã i9ó6eëe¸ 6
ß
®
ß y
Wð`² Ý saSpcgà ¡ xcg8w£xkÁ î ¦
ß ît ä ç ¦ 6 à ¦ ¥ 6 à ¦ à ¦ 6æ Ü ¦
ß y
Wð×d² Ý swxp¥à ¡ p1S8fxkÁ
î ¦
ß ît ä ç ¦ 6 à c à c 6 à ¦æ Ü ¦
ß ² Ý
ç $ ¥ SVdpzæ
6
²Ý
Û ç Y H 7 H F 4 Y H Y H F D ` 2B iy T 2B Y H R Y YB y T B 4 2B d H F D q 5 H 2 5 YB YB F D F ` R 5 F Dy T t Y 2 5B DB h 2 5 6#6IWuQIWpXxEIp¢XCQIWE¢UEQIEWIuIWm¢uI¢uXuXpG8I¢IG¢UQp8XWEII¢Q4 id T h 2 R 5 H FD i h HD T Hd ¢III8IWdoeQWVgQ4 QEI8WpIGW8p8XvW6I W¢6¢WIQ6¢§¢6IW¢9¢GpGD 7 Hy 5 d H F D H d 5 2 `B H @ H d H Fæ H d T i 4 T d R 4 4 T q 5 d H h d 5 H 7 T Y H F F DB @ Y H 7 H F 4 Y 2 TB ` 2 T d ` T ô 5 @D t d Ty R 4B D d T 21 Y H 4B h 2B H 4 T Y q 5 h T H d Y d H hB @ T H b T F yyB WECv6#6IWV8IVG8¢w386¢I6XW¢U#3v6QQEIIEQVUWh¢d¢QWIQIECu8gUIEXC@ i H FD D R 6IWvppot dpIVw pQSjËX9Q88IxIG#IG¢9¢GI6Q¢¢QI8#¢WCIG8gUF $ ¥ h 2 T c ¥ Y P 2B 2 H bB ` Y h 5 F D H 7 H F D Y T i 4 T d R 4 4 T q 5 d H h d 5 H 7 T Y H F D H b T D T F D h H bB d H h H 2 T 4 Y 2 5B D T 7B 5 d T 2 TB ` 2 T d ` T ô ` 2B Y R Y H 7 H F 4 Y D 2B 5 H b n D T F D t H Y d R 5 lUG68EQIfsd¢gI8EGV#XgWpI¢#¢8p¢G8V§wIXIuQ9QpWvSE88IVUWp¢WI¢Q4 q 5 t Y R 5B b 5 YB D1 Y D 2 HB h T d ` H H D Y q 5 i DB 2B 4B b H F D 2B h 2 T Y HB d T h 2 R 5 D T Y R 5 H ` T D 2 T b h T H ¢SQp8Xxo8Xv£QWoQXp¢G8dQQWh¢ 3IX6XxpGXdp¢QX¢UIp8sVvI868¢GSV¢¢ 2 T 4 YB F D t 2 5B D T 4By T 2 1 h H D T Ry T b H ` 2B H H d T Y H bB D T bB d H h H F D H d H F @ D 2B 5 H F D 5 D y T 4 5 ¢QCXIWx¢XGlgXEII¢3QWVUE¢V6IX6su¢Q8GVVWQpwpGuWQpCuSE8wpGGhVgxXy yyB D Y YB D T F D 2 5B D T 7 d 5q 2B ` 2B Y R i 4 T d R 4 4 T 2B H Y T H d 4 2B YB F D D H ` i H F D D T F D YB D 2B 5 Hy D R XEGfXVUGe8EGV#W¢pXpXWp¢GI6Q¢X ¢gQIEXpG86IWVUWfEoX¢s EGppWY Hd 5 7 T t d H b H @ 5 W896Q8c 8¢GpQQV©6W¢QW6IXaXQpGC¢WIVgQQ9QIr¢X8I¢W8¢oE8s£l¥ i 4 T d R 4 4 T h H Y T H d 4 2B dB H F D q 5 H Y R T 4 H Y H 7 H F 4 Y 2 TB ` 2 T d ` T ô D 2B 5 r d H b 5 H ` T D 2 T b h T 2 T H b T F i H F D iy d T Hy Y 2 5B D T 7B 5 d T 2 TB DB 7 d H 2 5 h d 5 @ y T 2 n 68cu8¢o¢VpV¢a8gUIGW¢QXvQp8XWVdEcII¢VG#Qf8W8v¢UIi ¦ $ ¥ 5 D c 8dp9G 3 p¥ 6S XwI8XW6WI9IW¢ko¢Gj¢gEGoQpXuX8GVVWQpwfWUu¢aIIXEIW9IWaVIGD Y P 2B Y 2 5B Y Y H d H H F D q 5 i 2 T 5 D y T 4B D 2 H hB YB H bB D T bB d H h D Y d n T q 5 D n H 2By Y H F D D T F H D 5 2 H 2 @ 5 2 ÷ D 5 2 YB D n H 2By Y T 5 D 2B DyB R iDB R 2B D 2 5 4 y T 5y ` h H Y T H d 4 2B H F D i D H Y g W¢IhCspx¢IEuIIEXI GoXvEEIId3EISXWo8Q¢U8E8Q¢gQIEIG#SCWs¢5 YB i 4 T d R 4 4 T 2B 2 5B D 4 R h H d YB F D F 4 R EA6¢WIQ6¢)E)8XW6IpQWaXIWrI 7 cfç z I8wX6¢WIQ6¢C¢aQI8aWEwpI @ 5 ß îæ iy 2 5 YB i 4 T d R 4 4 T q 5 d H h d 5 Y DB D R ß ¢ y ¦ t w² Ý sgà x¢feÁ î ç ¥ $ ¥ o¢dpzæ º² ø ç ¦ 6 à ¦æ û ç ¦ à à ¦ æ Áù ä 3sgIG¢feýÿe² Ý ¥ ( 4 ¦ i o 2 H b B ` Y B D 1 d 5 D T d H 5 H 4 2 H d H g B h B d D T 7 `H h T Ü T q 68E¢vE£mQ8WVG6s8fQI6W6hIGV#¨¢vw¢5 8p8X8XWV#XgWIIVIEXIfpGD 7 d 5q H 2 5 YB 2 5B D T 7B 5 d T H 2By Y H F D T F D h H 2 5B D 2 H 7 H hy R 5 F Y D1 Y 2 5B D T d H 5 y T 2 5 ` TB hB d D iy 2 5 q 5 H Y R Y H ÷ T 7 h 2 T lUG9QI8EGoQ9s#XI8pWv£vQp8XWVG6s8p¢U88VIEWGI8h¢I6¢¢#dIVç hXB ß îæ Y ß ¢ y A t ² Ý sgà x8ÔeÁ î ç dpzæ $ ¥ º² ø ç ¦ 6 à ¦æ ûç ¦ ¦ à ¦æ Ôcgà ¡ g¢feÁ ù å² Ý ¥ ( ¦ ä 4 ¦ Hy 7 T H d 5 ¤ Hy T D d 5y i T Xp#¢pu8U¢Xp¢8Excmt T q 5 Y 2 T H 7 i h H bB d H h iyB Y T H H 5 Yy T 2 T 4 H bB D T bB d H h h 2 5 4 H Y H F D q 5 Y 7 d 5q 2 TB DB 7 d H ¢wI¢g9)oÅ68EQIAXWVgWE¢¬¢Q8GVVEQI¬I¢QQIWu¢w9W8A¢WE9W6f
} p p u t r p Ø r w y r p u " xCuXsÂsu¡xg
Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u§fj«wÔzmQÔv«SÏmÇ
% r p u wsuvj
w
t ß 6õ
c ¥
H F t H 7 H F 4 Y `H h T Ü d H h d 5 r F D d R 5 q H F D h 2 T H 7 H F 4 Y H 4 2 H d H g B h h H d H D 2 H 4 d H h d 5 r F D d R 5 q h d T IuV#6IW¨¢vvQI83xGp8fIWdI¢#6IW6IQQsEI6WQWoQQQpW8£GfI8d¢Uh r 2 T D Y H F D d 5q Y 2 5B D Ty H d d TyB 7B Y F DB @ ` 2 5y T t ¥ H d R `B x¢fApGe8I8EGVXQ#¢XX9XÿWECÅI8E¢I pkI8Xm¤ Eð6G¢XIE¬%pÿ8EGVUSj 2B h H D D 5y YB c $ ¥ 2 5B D T R P
xS
ø
ß î
6 Ókä
ß î ß î 2B UfXY
H 4 2B Y t h H D 4 H H H 5 D YB Y T t i 4 T d R 4 4 T d H h d 5r h 2 5 4 H Y 5 D Y H D T 7B 5 d QIEWQW66s9GXC¢ë¢6¢WIQ6¢fQI83p8Q6WGf6GV#XgWpI¢T ç c $ ¥ ovdpzæ ß î ß î 2B ëfEWY ä h vv
h jlUGGVfò DT FD HD5
i 2 H bB oQ88` YE6swISQ8gºQUIx#kIWd8GlGQ8kQpQW6hIh 6W6GoQ66IW8£I86QeIG8U¤ B d H 7 R 2 H b T @ h H nB h 5 7 H F D d 5 D T d H 5 H 4 2 H d H gB h Hd H D 2 H 4 d H hd 5r h 2 5 4 HY H F D d 5 2 5B D T 7B 5 d T H F D q 5 i 4 T d R 4 4 T H F D q h8XWVdEcII¢wIWmV6¢WIQ6¢wIG§¢5 HIWVg9X6s px68g¢UW6¢IG§QGldEpgII¢Q IS68gwQUIx#pGXICfGD d R Y T H 7 T YB d H 7 R 2 H b T @ y T R D 4 T H F D Y H D T 7B 5 d T d H 7 R 2 H b T @ h H nB h 5 7 H F D F 4B F @ 5 H6W86IvpGmpxuv8XW6WIv6¢pvIWwEm¢gI¢ViQ6s px68gIWvQICW¢Qsp¢mEB H d ` H h H F D Y R F t 2 5B Y Y H d H D 4 T H H F D 2B Y d T H T t t d H 7 R 2 H b T @ H F D H d H F @ Y d T H T D H Y R T 4 H h H 7 T 2 5 Y YB d H 7 R 2 H b T @ h H nB h 5 7 H F t d H 7 R 2 H b T @ h H nB h 5 7 H F D Y I¢g6saQ9¢UmEjQ ISQ¢guQUIx#vIujQ6s px68gw6UEIx9IW§EB h SW6IC@ HdH F ç 9 $ ¥ 8dpzæ
þh ¥v6
ä
ä ä ä
þ Ôs
UfXog6
U8Qª's
ëfEWoV9
ë¢Qù ûç ß î 2B Y ß î Y 5 4æ 6 ç ß î 2B Y ø ß î Y 5 4æ
ß î
ß î þ ß UfXY î 2B ß î ¦
þ sç
e
ß î
þ 6 ¤45Wí ü ß î ¢
e G e
þ æ þ
ä
¢ ¤4 ¥í
e
þ 6
4 í Ý 6 4 üí AxÅ5WxÝ
t
ä
íç ( Ý æ
d ¦
þ ä í Ý 5 D d 5 D T d H 5 H 4 2 H d H gB h h H d H D 2 H 4 d H h d 5r h 2 5 4 H Y T t Hy 7 T H d 5q t iy T H @ q #xWu¢GVWQ8d6IQQsEI)6W6GoQ6wQpW8£I86QWm¢Xp#¢pw¢§¢EIIVdv£1 þ 6ä
ß þ
D H ` H @ t ß î ê ä í ß Hd H F 68wI S«uxuW6IC@
ç $ ¥ o¢dpzæ
Y XB þ q 5 H b B D T b B d H h D Y d n D 4 T H H F ¢¢EWVVEW6IUC6¢pwIut Y H bB D T bB d H h d H F `B F 5 D Hy T 4By 6Q8GVVWQpuQI8EIGEI¢QXXpI¢T iyy T R P H YB YB Y iy T 2 T H F D F ` R 5 F Dy T t Y 2 5B D T 7B 5 d T H bB D T bB d H h D Y d n 2 5 t H d H F t H D T d D 2 H 4 2 5 EE¢USQXXxE¢UVIW98p8IG¢SQp8XWVdEcII¢f8GVVWQpU#¢p¢QI¢WVGWS6QI¢Q4 yXECwr U¢Q¢EWVVEW6I6GV#XgWpI¢¢GVWQ8 QpQW6hI3GIUQ¢E8vX6@ cIF yB @ H þ q 5 Y H bB D T bB d H h Y H D T 7B 5 d T d 5 D T d H 5 H 4 2 H d H gB h r H DB 2 n 2 H bB ` T yy H @ 5 HpX9¢puG6WQWoXVf86QIWXv£6Q ISQ¢gCIGEd$QICot þ W8uIW9E¢T 2B 7 T H 5 D D Y H d H D 2B q 5 H d 5q H d H F D YB D1 d H 7 R 2 H b T @ H F D YB H d H F @ 7 d 5q H F D 2B H d F 4B F @ t Y D 2 H 2 5 7 5 4 d HB d R 5 ¤ Y DB 5 D 2B h H Y 5 7 5 4 H h H 2 T 4 2 5B D 4 2 Rq 4B h 5B d H i d T d DB d T 2 XpC8QGo6I8#86v6XWp8UGGoX6W898Q6IusVg8EG6pIEIxXW6saVGGIWVi
r e j  r p y w ã g j  y U7s{zD7uõ h ¥
e j i eg j  tgã y y ¡ k7v£×T
t sÉ6
Ú Ò Ú ê É Ó É Í × Í Í Ì Í Ì Ò Í è × Ñ Ð Ð fxfóCfóÔCmÓ«SxÎ
{
i s u § { z y s w u Ô « Ñ § e « § ¢ ® ¢ § Ô Ô ¯ § ¬ ¨ § t s ¢ ° « § ¢ y ® ~} rq § Ô 5« § ¢Ñ ¦ § ¢ Ò ¤® d¢ª ¢ Ñ ÉÔ Ô Ô ¯ ¢ © ¤§ q § b ¬ Gi gCExvs¦¤¢qd¬A¤qGÚd©£Adª¤£kÉxd©¤F¦qxfn¦¤¥z £v¦pQ¤vs¢ ¥FS£dqA£'%ªk¤)¦5Éx¤£X¤¬¨ pp A « Ô ¬ ª « ©
ç h $ ¥ ovdpzæ
ø ä
ç ß î ¥ 2B ø ß î s ëofEWY U EWY UjX 2B ø ß î 2B Y 4 æ " ß " ën86x4 " æ vä " ù î ç ß î ¥ Y 5 s Upë8Q4 ø ß î U 8Q4 Y 5 ø ß î Y 5 4 ø ¦
h v6
ä íç ( zsÝ 7 æ h 2 I¢T
YB d H 7 R 2 H b T @ h H nB h 5 7 ` 2B h 2 5 Y H d d 5 4 H F XfQ ISQ¢cQIEIx#IEII8W6W8QwIut
" û í Ý Ôç ¥xÆø Wxæ üí Ý
ø ç ¥xÆø í Ý
ß Wxæ " f¥¬h5Wx4 " jxGä " ù î üí Ý ø ç4 í Ý ø 4 üí Ýæ ø í Ý ¦
û í Ý 3ç Ax)6
ß æ ç ¥x)6 Wxæ ~fAxÅ5Q¡¸ù î üí Ý ø í Ý üí Ý ø ç4 í Ý 6 4 üí Ýæ 4 ¦
ä íç ( usÝ æ
2 H F D t ¥ ø ê 5 D ¥ 6 ê 7 5 dq ` 2B h 2 H D H Y H 7 H F 4 Y 5 D D Y H d H D 2B d R 5 D 4B d D Y H 6IGI©wW¬½Ï89pXII6GxfQ9QIGC6WQWoXuI8XWW6Wd H @ q vv£1 Q¢GVWQ8 XG#9xuXwzç 7 ©I¢8GVWQ8wEWW69#xEGo¢A¢Eaç W6IC@ d 5 D T d H 5 4B d D H 7 7 i Y T YB í (æ h 2 T d 5 D T d H 5 4B d D H 7 7 i YB D 2 T 2 T YB í (æ H d H F
í (æ ø í ( zç 7 ÿzç æ
ä í ( Cç æ
7 d 5q H F D 2B 2 H D DB d @ H 2 T 4 Y d 5 D T d H 5 H 4 2 H d H gB hr H DB 2 n ty T d H 2 H ` 2 W¢upGeXQWWWCs¢QuW8WVG6s8u6IQQsEI£GIUpz¢G6IQ¢31 l t H B Y 2 H bB ` T q 5 F Y H 7 T 2B Hy TyB T b j l
ë m¡ ¢6X68E¢v¢W6#XfXI¢XX¢VgT ß î H ` 2 T d d H 7 R 2 H b T @ H dB D 2 H H F D d 5q h H by 5 Y H d YB 2 5B D Ry 5 Y H F D q 5 D 2 H 2 5 7 5 4 d H 7 R 2 H b T ¢I¢Gu6s
px68gwWEGoQ
pGf¢k688W6WE8EGpE8W
pGj¢o6I8#86uQ
ISQ¢c@ 2 H bB ` T F 4B F @ F DB @ i 4 T d R 4 4 T H F D Y H hB b 5 d d H 7 R 2 H b T @ h H nB h 5 7 H F D d 5q 2 5B Y Y H d 68E¢¬XIC)WECr¢GpQQVeIG
6IXxcWp96swISQ8grQIEIx#dIGw¢¬¢XWQIH Y H 7 H F 4 Y Y R 5B d T b d 5q d H 7 R 2 H b T @ h H nB h 5 6Q9QIuI8X¢V8f6swISQ8geQUI
0 0 0.5 1
Û ¥ H d R `B I uWI8Ev¤
Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u§fj«wÔzmQÔv«SÏmÇ
κ*∆ x
0.5
1.5
2.5
1
2
3
2 Central
nd
κ∆ x
1.5
th
4 Central
4 Pade
th
2
2.5
3
¡
¨ § c ¬ § Ò ¥ ¬ uddAÕ¦¥£q¦
° ¬ Ñ ¨ Sds§ ¸q© 642VQÐ c §§ ¯ e ¬ ¨ ¢ u © ¨ª Ò© © Ô¢ § § ¨§ § Í ¦0dd¬d½xsda¦¤A xz Ð ¦S¦ÏU¤£¤¥¤¦« s v " ß î h if ¢ ëfEWY fä Þ " 6 ä 6 ß î 2B
Û ç Þæ d 5q h H 2B T D 5 YB ã f¢6IXVp8uE© ` 2B @ 5yy 5q H F D t 7 d H D y TB D T Y H F D 5 D h HBy T H d T Y H 4 2 H d H gB h h H d H D 2 H 4 d H h d 5 r h 2 5 4 H Y q IXCcEX8uIGUW6Gh¢XGVIWwIG9GeQEXIIVw¢6QI6W6hIeQQGo6QQI83p8Q6Wj£1
¤ ¡
¢
þ æ
ä çÞà ßæ sç ¡ jWvÝ
Þ
Y T 2 5B D Ry 5 Y D 4 T H H F D Y H bB V8XWpX8¢puIWC688` ¡ ëjëGoE9pXGGEWIpWI¢uvf8 IXx8 ¥ P 5 D 2B ` 2B D R DB D Y R Y h 2 T ç Þæ d 5q ` 2B by 5
6 £if¼ä
ç ¡ pzæ ¥
"
" ã H F D Y H n YB D T Y ç Þæ H d H F aIWC6UEGVWW6IC@
þçÞæ ä çÞà ßæ sjWvÝ
YB d H 7 R 2 H b T @ F DB @ 2 5B D 4 2 Rq 4B 2 5 7 d T F T i h H D TB DB 2 E6swISQ8geGECe8EGIIEI89W¢Uo6GVXGEpXB 2 5B D Ry 5 Y T D T F ¢XGX8VUWD p¥ 8XW6QxÄ¢WX¢QQC GD $6 8eIXIpQGxrXVd8p¬85 2 5B D 4 H 7 5 dq yy T 4 H 5 7 5 dq ` 2B h 2 H D H 2B T 7 5 h T 2 ß Þ ¡ ä ø Ý Ý 7d 5 8q H F D 2B 2 5B D T R P H 2 5B D 4 H b 2 5 4 d T H 2By H F D 2B T ` T H 4 2 5 d H hB Y 2 5 2 5B D T D H d d H D 2B y T 4B Y i F pGE8EGVUS6e8XW6Q¢o8Q¢gpXXaIGXVo¢6I86IXI88EGVG6WpWQWoX§¢gEWxoIdT 2 H bB ` H 2 T 4 Y d 5 d d H H F D t 2 5B D T R P H 2 5B D 4 H b 2 5 4 d T H 2By H F D q 5 D H D 2 5 4 H F D 21 Y 2 5B D T 7 68E¢#r¢Qa8WQ#IG8EGVUS6)8XW668o8QVgIEX9IGV xWS¢Q9IGA3#6I8XWV#XB r 5 d T H 4 2 H d H gB h ` 2B Y Y H Y Y T d 5q y 5 5 D y Rq H Y R T Y H hB b 5 d d H 7 R 2 H b T @ h H nB h 5 7 H F SgWpI¢QpQW6hIeIEWW6W¢88xGvp6WpduQIExsIuQ ISQ¢gQUI9Iut ß îçs U8Q¬ø æ ß î Y 5 4 ß î 2B Y UfEWor¥ ç ä h £ Y H bB ` F 4B F 688EIC@
e
þ á6
e
þ æ
ß î ¥ ä
e þh ø h
£mv6o¬ø G e
þh ¥£
Y H n YB D T Y H 7 H F 4 Y YB F D d 5q d H 7 R 2 H b T @ h H nB h 5 7 H F 6UEGVWa9QIuXIW8f6swISQ8gQUIx#wIut ç4 í Ý 6 4 üí Ý f¥Å5xæ ß î ¥ ä 4 üíç (æ ø íç (æ ø 4 íç ( 5WhÝ ÿjzÝ UÆAÝ æ
i 2 H b B ` Y B H 7 H F 4 Y `H h T Ü d H h d 5 r F D d R 5 oQ88CXu9QIa¡VvCQpW8£GfI8q H F t d 5 d d H iy H dB D 2 H ` 2B H D 2 H 2 5 7 5 4 i d T 2B ` T 7B H F D F DB @ t Hy 7 5 4 YB d H 7 R 2 H b T Iu¬68WQÅQWEGoQkIEQ9S6I8#¢QA¢IX8¢#XIWÆWECjsEI#¢QX#Q ISQ¢c@ h H nB h 5 7 H F D t D 2 H 2 5 7 5 4 4B d D H 7 7 i Y T Y H h Ry 4 2B d 5 D T d H 5 H F D 2 H F y T H d iy H d R YB d H 6UEp9IWxQoQI898QXG699faQIIEQIEv¢GVWQ8IWQIª§¢QW9E6WIpvEQ r 7 R 2 H b T @ h H nB h 5 7 H F D t ç h H d H D 2 H 4æ 4B d D H 7 7 i YB D 2 T YB d 5 D T d H 5 H 4 2 H d H gB hr H DB 2 n H F D 2 H F ISQ¢c6UEIx9jIGlë6WQWoQQuEWG#9xEGo¢E8GlGQ8j6IQQsEI£GIUIWw6Iª ¦ I Ú Ò Ú ê É Ó É Í × Í Í Ì Í Ì Ò Í è × Ñ Ð Ð fxfóCfóÔCmÓ«SxÎ
h H dB R P H d H d T Y d 5 D T d H 5 i d T h 2 R 5 hQWEISQw¢CW¢GVWQ8eVUII8 y TB 4 H Y 5 2 t Y 2 5B DB h 2 5 4 i d T h 2 R 5 4B h 5B d H q 5 H Y T 4 H F D 2 1 2 5B Y R gB h h 2 T 2 5B D 4 H b 2 5 4 d 5 ¢6QWdpUQI8EGII86¢UpI8XIxXQ¢aW¢QaIW38EWIsEII¢e¢XGQ¢S¢QC8q Y 2 5B D T R P H y H h 5 7 d R 5 d 5q h H h H H 2 Y d 5 D T d H 5 i d T h 2 R 5 H F D d H hB Y 2 5 4 H @ t 2 5B D 4 H Y YB F D 2 p8XWVUSQ§QIx9I88eQpQQpW¢GVWQ8f¢UIp8saIGfQIEWI86 vs8EGQWfXIW31 QQEW¢III8dIWal VIGrIVE¢#8IdpG¢u8EW6GoX#IGAXk8XWV#XgWIIVQpQW6hIh Y HB d T h 2 R 5 H F D D T D T F D h 2 T 2B T 7 5 h H F D q 5 d 5B d H D 2B H F D 2B 2 5B D T 7B 5 d T H 4 2 H d H gB H F D F D 5 Y H D T d 5 d 5 4 2B d 5 D T d H 5 H 4 2 H d H gB h B d D T 7 T pGªG¢eQWVG8W¢QIE8GlGQ8kQpQW6hI«WWV#¬t p¥ ÿ¢XGQÿEB QWI6WEIi ¥ 2 5B D 4 H 2 h H Y Y R 4 YB h Y ² ± ¹ À ´ · ¶ î 6weeëÃ
¤ » w´
² À Ã » ´ À ± C 9aj
± Á ¶ ± À ± ¿ CejÅ6¹
½
¢® £
88A8XW66X6WIQXICQpGpXCXIum¢IpGEEI9¢XB ¦ ¦ 2 5B D 4 H 2B h H Y Y R 4 YB h d H F D d Rq YB YB F t H h R DBy 7 T 2 d 5 d d H H F D 2 5B D d 5 4B d D H 7 7 i Y H F D h 2 T h H H Y 2B d 5 d d H H F D Y H 2B 7 d H D H h d 5 D T d H 5 H 4 2 H d H gB ¢WW6CIG 8EGf8sXWW69#xfpGI¢dQQWEj8WQCIG6IX9W6GIj8GlGQ8CQpQW6hIh y TB D T Y H F D q 5 2 5B D d 5 4B d D H 7 7 i YB D 2 T H F D Y R F t 2 5B D Ry 5 Y H F D q 5 H h R DBy 7 T H F D 2B d 5 d d ¢WVUIGVd8EG¢sXG699fXWSVIWISuv8EGpE8WIG¢CIpGEEI9¢IWdXv8WQH 2 T 5 D Y h T Hy D d T i d T 2B ` T 7B H F D HyB F @ t h H H Y 2 5B D T ` T 5 d H F D 2B d 5 d d 2 T 5 D Y h T H ¢)G#p¢gEe¢IÅ¢IX8¢#XIGXEIChQQW©8XWVo¢I8WIIG«Ed8W6H ¢ÅG9I¢QXy d H 7 R 2 H b T @ h H nB h 5 7 H F D q 5 D d T y T H d H F D D T F D H H Y H @ t ¦ ¥ P q 5 7 d 5q H F Q IS68g6UEp9IG§¢Cf¢U¢QWIWuVUWu6Wuv8p pëëj8¢§8IGD 8WU¤ 7 5d y H b Hy d 5 d d H H 7 T Q¢6E¢WW6#VGY H F D H b HB F 4 T 5 D iy H bB D 4 H Y H pG86EI¢kW)E68EW66sQWd ¡ ªIVqeWEISQW9QI¨¢v6IW¢3WI8pGD ¦ h 2 T $ ¦ H d B R P H d H 7 H F 4 Y `H h T Ü d H h d 5 r F D d R 5 q H F h 2 T H 7 H F 4 Y h H d H D 2 H 4 d H h d 5r F D d R 5q D 2B 5 r $ H F t D 2 H 4 d H ¦ IV#6IWk6WQWoQQfQI83xGp8CSE83sÓpugoQ6W6s ¡ VUGWQE¢eQ6s WVUI 2 T F D Y Y Hy q 5 h H H Y H Y T F 2B d 5 d d 2 T H 4 R h 5 d 5 Ed8W6H ¢eQpIIrWD 0 ¡ ôQWEISQe#6IW6IQQsEIª6WQWoQQdQI83p8Q6WY c Y H dB R P H d H 7 H F 4 Y H 4 2 H d H gB h h H d H D 2 H 4 d H h d 5r h 2 5 4 H H F D t Hy 7 T H d 5 ¤ y H b Hy h H nB 4 H Y T @ 5y H Y d 5 d d H H 4 R h 5 d 5 D h H dB R P H pG¢Xp#¢p8UÅ686E©6UE6QWeµcEQW8WQ6IIxWIAGÅQXIS6Wd H FD q 0¡IG¢5 Y 7 d H D 2B h H Y Y H d H H 2 T 4 H 7 H F 4 Y T q 5 i 4 2 HB 4 ¨ H ` 2B by 5 Y H d H F t ß î r 9W6G)E«QW6WIds¬¢Q9QpW¢ÅIQEQ96epXxE8QeIu Uo oÅQ88` i 2 H bB YB H by 5 Y H d YB H b T @ 2 H bB ` T F 4B F @ i E)hQ¢E8QaEe8gQ88¬EICkoç ¢xX¢r#UuUlU¢w6swISIut æ x } } 1 è } x y q 5 d H 7 R 2 H F iy Y 5B b H d h H d H hB Y 2 5 4 Y H 7 H F 4 Y H F D d 5q h H H Y H Y T F y T 4B d H 7 R 2 H F D Y @ 5 F Y $ ¥ H d R `B 8EIR8EWI6WQpXWp8QQ9QIuIGm8QQWC¢UI¢gEWQ9ISuIWCcI%uWI¢Xv¤ 7 d 5q y T 2B `B d W8ë¢UE8X85 Y DB Y Y 5y iy T R D H b H Y 2 H 2 5 7 5 4 d H 7 R 2 H b T @ D 2 H d H gB h i 2 T 7 q 5 ` 2B D YB Y 2 5 4 7 d 5q H b T WEuQH8XAXy¢UWo2Q8aWDoQI898Q 6swISQ8gao6WQsEpeo¢d¢pXGfXI8Q¡86¢c@ T t Dy R H d T uQEIYQrYfÿV8EW6HsXII¢5I#EÏC¢UX¢XW¢IWÆ8p8IG¢u¢XWQWEId6QIIxWWoXB i H bB Y d YB h 2 2 YB ã Ü y T 2B `B d 5 H F D F ` R 5 F Dy T t 2 5B Y d H YB h Y H 4 R h 5 d D 2 2 5B D T 7 5 d T y T B d H 7 2 H F t d 7 R 2 H b T @ H F D q 5 2 5B D 4 2 Rq T Y 8XWVdEBcII¢¢g4EWQ9IRSIWDmQ6HswISQ8gIG¢e8EG6pIuXB h ë6IX8ECcEWexGD H 4 2B iy @ 5y Y 5 5 Y6GV8¢U85I«85EGpE85W¢QXWd6#px2eHpGD¢EI9¢pdEIG98¦ l r h fQIE ¢EE¢gEWQ9IS2 H D T ` T d 2 B D Ry Y y T 4B H 7 R F t Hy 7 T H YB F D d 5q H 4 2B iyy T 4B d H 7 R h D T ` T 5 YB 5B D 4 Rq 4B 2 7 d T 6HGVoVU8WdpX28EG6p2IvXp85#¢UFuTaFXBICVaQ6sIGEa6QWW¢Up¢¢QXW6#pxIut 4 F @ D T h H H Y H F D YB h H H Y H Y T F y T 4B d H 7 R 2 H F ß î U ä h ß î 2B UEWY t Hy 7 T H H b 5 T H F D d 5 ¢EI#VIu8csVapG8U¤
i d H 7 R 2 H b T ofQ ISQ¢c@ h H nB h 5 7 H F D 5 D h H D Ty H d YB F 4B F @ t h H H Y H Y T F ç h H nB h 5 7 d 5æ y T 4B d H 7 R 2 H F D Y 6UEp9IWwG#QGlQvEeXICQ6sW¢IIUQUI98v¢gEW6#ISfIGXB h W6IC@ HdH F ç ¦ ¥ sp pzæ r 5 d T y TB D T Y H F SgWpI¢ë¢XGVUCIGD
h
ä
h
¤ ¡
¢
þ æ ä ç Þ à ß y T 4B d H 7 R £ç ¡ kæ ¢QXW6#px2 Ý
2B T D 5 H @ E¢8upt ¡ sëjëGoXIEGpWEWIII¢pEI¢8XWV#XB ¥ P 5 D 2B ` 2B D R DB D Y R Y h 2 T ç iy 2 5 2 5B D T 7 7 5 dq d 5 d d H H F D ` 2B d H hB Y 2 5 4 H d T H @ H 4 2B Yæ iy D 4 T H ã YB F D ` 2B by 5 ¢Wj8WQIGIEW6IXI8QC¢vuQIEWG¢pvXpGIXx8
Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u§fj«wÔzmQÔv«SÏmÇ
¹
ç ¥ S¢ pzæ
¡ ¡
´´µ ³´´
¹
à
··
ß î ¦ w6
ä y A t º²
ä Ý ·· ¶ !6
ç ¥ ¥ o8 pzæ
··
¦
·· ¶ y ¦ t º²
ø ²Ý D»
¦
¡ ¦
¦w6 ¡
´´µ
¦ w6 ¡
³´´
ß î ¦ w6 ä
»
F DB WEC@
ä ²Ý £ (
2B T D 5 H @ Y H 4 2 H d H gB h h H d H D 2 H 4 d H h d 5r h 2 5 4 H Y F DB @ t Hy 7 T H d 5 ¤ i d T h 2 R 5 YB F D d 5 X¢Gp8wCQ6IQQsEIQW6GoQ6f6IW¢3I¢QQGEC¢Xp#¢pu8Uv¢¢III8uXIW8q h H dB R P H d YB D 2 H 7 D T H d D y TB 4 H Y 5 2 2 H F D t 4 í Ý h 2 5 i H h 2 H D H D 5 2 Y H 5 h 2 5B D T 7B 5 d QXIS6WwX o6#WVgG¢6QWIrQIWv¥¬I¢86ÅI6Gxw¢Iw6p8EGV#XgWpI¢T H 4 2 H d H gB h d 5B d H D 2B H F D Y T ` 2 5y Y T D T F D d T Hy 4 YB D1 i d T h 2 R 5 @ 5 q 2B H F D D Y d n d H hB Y 2 5 6IQQsEI§8XQWSEIWv¢I¢X¢VUWv¢gEQmXj£m8VUII8ÆcUpXIGfWU6IXI8 h H nB 4 H Y YB ä Ý h 2 6UE6QWuXC!I¢T ç pzæ ¥
Æ aCà û Å Ý
xS
à
C£Ý Ïe² Ý Ý à 4 ù ä
YB Y 2 @ 5 2 ÷ 2 R q 5 d 5 D 4 H b H F D Y R F t i d T h 2 R 5 EfICcISIpv¢8W6Q¢IGCISu8VUII8 @ 5 q D R 5 H F D YB i d T h 2 R 5 h 2 T Fr D F `B d H F D h 2 T t i d T h 2 R 5 @ 5 q 2B H F D YB i d T h 2 R 5 h 2 T cUWp¢IGEVUII8eI¢I3o8EWIWkp¢8f¢UII¢såcUpXIWX¢UpI8Ap¢UF r6EIGpxW¢¢EGW8X QQW98g IGVUG 9IW¢9v WQq¢¢UpI8 cUp85 Dq Hy H F D Y R F D H bB DB Y 5 YB h H H Y H b T @ H F D D T F D H 7 RYY T H @ Hd H id T h 2 R 5 @ 5 qD R H F D D T h H nB 4 H Y YB 2 5B DB h 2 5 4 5 2 HyB F @ t i d T h 2 R 5 @ 5 q 2B H F D D T 2 H bB ` YB 2 5B DB h 2 5 4 i d pGVAQIEQ6sX8XWEII¢QeIaXXpC8VUII8cUpXpGVA68E¢X8XWEII¢Qf¢T rxII8j6EIXXapVW¢Q6GVIG93j¢x3«cUW83cUpE¢8uEIEW6sQpGE6fdVg4 h 2 R 5 D Hy F 4B dB ã T t H Y T 4 d H D D Ty H F D 21 H iD @ 5 q D R 5r @ 5 q 2B q 5 d 5 4B h 5B d H d H F DB H H 2 T 2 5B D T R P H 2 5B D 4 H b 2 5 4 d T H 2By H F D d 5q Y 2 5B DB h 2 5 4 i d T h 2 R 5 H F D t 8EGVIx68EG668o86VgIEXwpGC8p8XWEII¢QVUII8wpGëp¥ 8EG66eW#IEWQ6C 2 5B D 4 H 5 D ` 2B d d Hq H
i eg j  â ¬ « i eg j t p © i h£7uX&kÚhªke 2susëugö à r  p ig ! p u
Y H 7 H F 4 Y Y R 5B d T b d 5q h H H Y H Y T F y T 4B d H 7 R 6Q9QIuI8EW¢V8QQWwWVUIhVgXQ9Ifò
0 0 0.5 1
¨ s
Û $ ¥ H d R `B x%puI8Ev¤
¥ o
Ú Ì Ò Í É Ö Ó è × § Ë É Ú Í × Ë É Í Ì Ê × Ì Ç Ó Ì Í Ì Ñ Ñ Ò Ö Ð ¥Ð ÔfmÓVefCCCó fvÔ«o¦xÎ
a* _ a
0.2
0.4
0.6
0.8
1.2
1
2 Central
nd
4 Central
th
κ ∆x
4 Pade
1.5
th
2
2.5
3
H dB R P H XIS6Wd
q 5 Y @ 5 d 5 @D D Y Ty H F D d 5 D T d H 5 d 5B d H D 2B h H d H D 2 H 4 d H h d 5r F D d R 5q T q 5 H Y T 4 H F D 2 » VCcWv3vf¢fIG8GVWQ88XQWSEQW6GoQ66IW8£GfI8a¢WVgIWd31
2 5B D T 4 nB h 5 ¢XGlgUI97
¹
ç c ¥ o pzæ
· ·· ··
¡ ¡
¡
´µ ´´ ´´ ³´´
¹
ß
î ¦ a6 ä y ¦ t º² à ··
· ·· ·· ·· ¶
¦
¡
6
´µ ´´
¦ ¡
¦ w6 ¦
´´ ´´
¦ w6 ¡
ä Ý ·· ¶ 6
¦a6 ¡ ¦
³´´
ß î ¦ w6 ä
»
F DB @ d 5 D T d H 5 H 4 2 H d H gB h T Y H 4 R h 5 d YB F GC8GVWQ8uQI6W6hI96QIIxWpEIut ç 9 ¥ 83 pzæ ç4 Å Ý 6 Å Ý )Ùæ ß î ¦ ä
Å sÝ æ ç (
Û h H Y R H 2 T 4 Ty R 7 d 5q h d T @ ÷ 4 T d H h d 5r D Y d n ` 2B @ 5yy 5 hQWpw¢g XI 8¢o¢I6IW¢3fWUaIXCcXE8q HpGjQ85WVG6Hs¢ 8XdQWSEQI83p8Q6WeÆGEª)hQWpp p W8e6IQQsEIÆW¢@ F D t d D T d 5 d 5B H D 2B d H h d 5r h 2 5 4 H Y T F DB h H Y R H D Y R 7 Ty R 7 d 5q H 4 2 H d H gB h h d T r ÷ T 84¢U i2 lT6WIRu¢Ip¢gIG5WxuQXIS6WXICxQ8WVG6s¢a6IQQsEI£QQGo6Q4 ä ê D h H Y H D 5 2 2 T 4 t 4 ü í Ý Y H dB R P H d F 4B F @ t d 5 D T d H 5 H 4 2 H d H gB hr h H d H D 2 H d H h 5r h 5 4 H Y F D Y R t 4 Å Ý D R 5 T 2 5B D T 7 d 5q 2B 5 2 F DB 6IWd8£I286QeHIG ISFuk5ü 8sVr8EGV#W¢pXeIrGEC@ DpI VT ß 5Ý WV#XB H h 5 2 D r HD T 7 r 5d T Y R 7 SgWpI¢wDI Hd6UnE6QHWYuXBk8EGEpI8Q¢UpI8Im8f¢UII¢s sIWp89IWai h H B 4 Y 2 5B DB h 2 5 4 i d T h 2 R 5 5 2 t i d T h 2 R 5 @ 5 qD R 5 H FD D H D T d R 4 4 T d H h d 5 r F D d R 5q iyy T 5y ` YB 2 5B D T 7B 5 d T YB F ¢WVGI6Q¢C6IW¢3WI8X¢U¢X8uE8XWV#XgWIIVfEIut
¹
à
ç ¥ o8 pzæ
ä y ¦ t º²
¹ ´´µ î ¦ ß ¡ ³´´ w6 ¦ ä Ý ·· ·· ä Ý ¸¶ !86
··
¦
¦ a6
c
¡ 6
·· ¸¶
c
¡ ¦
c õ6
¦
´´µ c õ6
f6
³´´
ß î ¦ ¦ w6 ä
»
F DB @ ¥ ¥ P q 5 7 d 5 WEC 8 pj7¢§8q H F D Y T F d 5 D T d H 5 H 4 2 H d H gB h ` 2B Dy R Y H d H F t Hy T D d 5y i T t T ` 2B Y R h H bB d H h iyB Y T H YB F 4B F pG¢U8GlGQ8QI6W6hIuIXWEIQWIuVXIVj8XxgmwaIXId68WQp EEW¢QXdEIC@ ç $ ¥ 83 pzæ ç ß î Å6 ¢Æh£¢¬Ý æ 8 Ý Ý ø 4 Ý ¥ 6 ä 6 ¦ ä 4 sÝ æ ç (
Û ¦ ä ê D T d 5 D Td H 8lC8GVWQ85 d H h d 5 r h dB F D ` 2B @ 5yy 5q H F D H Y R 2 T 4 H @ t Y H 4 2 H d H gB h h H d H D 2 H 4 d H h d 5r F D d R 5q F DB @ t Hy 7 T QI83XIWaIECcXX¢IGWpVguI6Q6IQQsEIQW6GoQ6QpW8£GfI8GCp¢EI9¢pH d8U ¢9QpW¢XW6GoXpGj¢lUGvVUSQ§XXC6¢WIQ6¢vVU8X¢ IGeIV¢9QI8EWQWoXB 5 ¤ H 7 H F 4 Y d 5B d H D 2B H F D q 5 D T F D y T R P H yyB @ i 4 T d R 4 4 T y T 5y ` H F D h 2 T t H 7 H F 4 Y d 5B d H D 2 H F D q 5 D T F D 2 T F D } è } } YB F 4B F @ i 4 T d R 4 4 T q 5 d H h d 5 2 T H b T F 2 T 4 t } { } pG¢9VIG©VUGwlXexlehlX©XIC¬¢GI6Q¢¢QI8«V8gU©¢gdUc ¢oxl5¢WdUd¢CcIS68GlGQ8#¢Iv8¢GpQQV¢6IW¢fWVI8WpI¢pGD ¢ À y } { ¢ T Y T 2 @ 5 2 ÷ t d 5 D T d H 5 2 T F 4 R i 4 T d R 4 4 T q 5 d H h d 5 H D TBd 5 d T H F ` 2B b T F HyB F @ t ¦ 6 ê 5 D iy 2 5 Y h 2 H D H F 4B F @ IExcIXXpC§8ÿë#GEI8aII6GxXIC¦ l)QXIS6WuXu8WVG6s8woQQsEIT ä ê D T h H dB R P H d YB d 5 D T d H 5 D 2 H d H gB h í Ý q 5 H Ry T b T Y H dB R P H d ¦ ä ê D T 2 5B D T 7B 5 d H 4 2 H 2B T 7 5 h H F D H hB Y D R 5 YB F 4B F @ 6IQXVd8pwpGIEWG8uXXpCIt Axm¢IE¢V f6WEISQWuåwVe8EGV#XgWpI¢T H F D 2 H F D t $ ¥ P 2 2 H bB ` d 5 D T d H 5 d 5B d H D 2B d H h d 5 r F D d R 5q H F D H Y R H @ qB t d H b H @ 5 pGªQIWCpV%uëgEB Q¢E8¢GVWQ8d8EW6GoX6IW¢3WI8pGWpAvEQQ¢6vcf Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u§fj«wÔzmQÔv«SÏmÇ 8
ç 9 ¥ SVdpzæ
W ß 4 ü 5Å W ß V iä V Ý Ý
YB D T F D t ¦ XVUWp8ðø ä ê D T h H nB 4 H Y YB AV¬6UE6QWwX©ß ÚÝ VUWd#pWWVe8EGII86rf¢UII¢s)p¢dpQfw8U¤ r D T F D H 7 R Y Y T H @ t 2 5B DB h 2 5 4 i d T h 2 R 5 2 2 T 7 R H ò T d 5 ¥ P 2 9 pëëjXB 2 H bB ` d 5 D T d H 5 H 4 2 H d H gB h H F D 5 D ` 2B h T Hy t Y HB d T h 2 R 5 d T H 2 h H dB R P H d H d T Y 2 5B D T 4 nB h 5 7 5 68E¢§¢GVWQ86IQQsEIIGGuIEI¢gEV6QEW¢UpI8m¢QI 6WXpx6WW¢I8XWVgIEIx#fI2 ç ¥ 9 ¥ o¢dpzæ ç4 í Ý ø í fAxÆvÝ 85Wxæ 6 4 üí Ý
ß î ¦
ä í sÝ ¥ æ ç (
d 5 D T d H 5 d 5B d H D 2B h H d H D 2 H 4 d H h d 5r h 2 5 4 H 8GlGQ8¢XW6GoXQQGo6QQI83p8Q6WY HpGGª¢8c¢AQWI6WXpe8XWVUSQk8EG668o86 IG8¢UpI8åcUX IW¢QIEXy F D F DB H b 5 T h H Y Y R 4 YB h 2 5B D T R P H 2 5B D 4 H b 2 5 4 H F D d 5q i d T h 2 R 5 @ 5 q 2B H F D Y T Y H 2B H 7 T Y H F D ` 2 5y T Y h H H 4 5 d 2 5B DB h 2 5 4 i d T h 2 R 5 D Hy F 4B dB ã T q 5 D 2 H 7 D T H d D H F t Y 2 5B DB h 2 5 9¢WuIGI¢X¢I6QQxWp 8XWEIp8Qf¢UIp8sm6EIXXuh¢joQ9GVQWWCpuvQp8XWEII¢Q4 i d T h 2 R 5 2 2 T 7 R H ò h 2 T D Hy F 4B dB ã d H hB Y 2 5 4 D Y R 7 H @ 2 5B D T R P H 2 5B Y R gB h H F D ` 2B by 5 Y 2 ¢III8I¢#IQI¢6XpXWEuQpXWp8QCfI v8XWVUSQ¢XWphIIW IXxE¢We31
i eg j  â ¬ « i eg } â ng l p u h£7uX&h£¡uÏo2Ùugö
¨ o
9 ¥ P 5 D y T 4B D 2 H hB YB F 4B F pëjëGh¢QXWS6IXuEEIC@ ç dpzæ 9 ¥ ä ä ç4 Å Ý 6 4 ü Å Ý ¡)855æ ßî ¦ ä
ç4 Å Ý 6 4 Å Ý 6 Å ÔÅ7)ÙÝ æ
ß î çf4ÅÝÅ6ÙÅ5Ýæ ¦ ßî ¦
Å sÝ æ ç (
Y H bB Q¢E8` H h 5 2 D T h HBy T d 5 D T d H 5 H 4 2 H d H gB h h H d H D 2 H 4 d H h d 5 r h 2 5 4 H Y H F D 5 D 2B YB F D ` 2B D R DB D Y R pIVeQEXIp¢8WVG6s¢QI6WQsEp6WQWoQQ6IW¢3I¢QQIG#GoXEIG#IXWpGGpI ç ¦ 9 ¥ sdpzæ
4 Å Ý 6 Å Å2Ý j5Ý ä 4 ü Å
Û 4 ü Å Ý 5 D 2 5B D T 7B 5 d T d H h d 5 r D Y d n ` 2B @ 5yy 5q H F D Y H bB ` F 4B F 59Wd8EGV#XgWIp¢QI83fWIwIXCcEX8IWC688EIC@
ç ¡ dpzæ 9 ¥ ¡
ä 4 Å Ý ø Å ÆðÝ 85ÿ55Ôç 6 4 ü Å Ý ä 4 ü Å Ý
® ø
4
®
pfæ 6 ¦
2B T D 5 H X¢Gp8w@ d~GEC¢Xp#¢p8U8ºë°X8EGD ä F DB @ t Hy 7 T H d 5 ¤ ¦ 6 YB 2 5B r T 7B 5 d T H F D q 5 d H h d 5 H F D h 2 T 4 ü í Ý ä í Ý ® i h H 2 n H h d 5 D T d H 5 DqB F Y H F D Y ® H d H F l#XgWIIVIGVQpW8pGdIV5Wxðw¯oQpU6I8WVG6s8EIIWEB W6IC@ ç h ¥ o pzæ ¡ ä 4 ü Å 5Ý ç
Ü 4
® 6 ¦ Vpfæ
Û hB d ` h H 4 T YB R P H 2 T 2 5 d H h d 5 i d T d DB d T 5 D T D T hXW¢Q6¢UXISQ¢8CQpW8d¢WGEpW¢9G GVUh d8XQGoEvIWó8WuQWV¢s¢WGxvsGh5§CcEX¢Cp W8pXCcXE8vIu¢XGl8¢GWxH 5B d H D 2B H F D 7 5 dq h H D Ty 5 T d D H H 5 D 4 ü Å Ý Y @ 5yy T Ty R 7 d 5q ` 2B @ 5yy 5q H F t 2 5B D Ty 5 T d D H 4 T Y q 5 Y 7 d H D 2B YB Y H 7 H F 4 Y i d T h 2 R 5 q 5 D 2 H 7 5y H b H h H F D 5 D F 4 T 5 d T d H F D 5 2 6¢UWv¢9W6GXXwQ9QIVUII8v¢wS6#p8X686pIGeGe¢8WIIVa6IGVIi
$ Â
Ú Ì Ò Í É Ö Ó è × § Ë É Ú Í × Ë É Í Ì Ê × Ì Ç Ó Ì Í Ì Ñ Ñ Ò Ö Ð ¥Ð ÔfmÓVefCCCó fvÔ«o¦xÎ
ç c 9 ¥ ovdpzæ
ç zdø ß îæ
Ê¤45üÅ W ß V ß ð î Ý
ø 4 Å Ý 6 Å Ý )Ù8
È ¥
¦
ä 4 ü Å j5Ý Y H bB ` 4 ü Å Ý d 5q ` 2B by 5 Q88v5u8 IXx8
ç 9 9 ¥ 8dpzæ
ç z~5¢©ª8Å8æ ß îæ ø ç4 ü Å Ý ¥ ø Å Ý 6 4 Å Ý
ß î ¦ ä
45üÅ W ß V Ý
Û ¦ 8ªø IxICV6XEII¢8EGV#XgWpI¢aQI6WQsEp3¢o¢If6IW8£I86Q6IXI8 H h 5 2 D T h HBy T 2 5B D T 7B 5 d T H 4 2 H d H gB hr h d T @ ÷ 4 T d H h d 5r h 2 5 4 H Y T d H hB Y 2 5 pgIE8EGVX8sVGGx#6¢UWpGIXI¢dp§jTE8GVWQ8IGeE¢p¢eWXVr¢Qr T H hB 2 5B D Ty 5 T d D H H 4 T Y H F D ` 2B Y R 9 ¥ P 2B d 5 D T d H 5 H F D 2B T D 5 5 Yy T 2 T 4 H H 7 H F 4 Y d 5B d H D 2B H F D 2 T F D d H @ 5y d H h d 5 H 2 5 H 2 T 4 F 4B F ¢9QI8EWQWoXuIG¢UWuQvsEuQI8uI8u¢QEIC@ iy Y R 5B b H d h H B d 4 Y H h Y H 7 H F 4 Y i d T h 2 R 5 y T 4B d H 7 R 2 H F D F DB @ Y D Y T d D 2 5 4 YB F t 2 5B D Ty R EI8XxWIQXW6W6If6#6IW¢UpI8¢gEW6#ISuIWWECCGf¢GWS¢QCXIum8EGVXI 7 rf8uGVWIQ6¢fQI83p8Q6W9EE¢U8E89E¢p¢#Wf¢GQQ6ICXuEIGph8XWEII¢QVUII8 d 5q H D T d R 4 4 T d H h d 5r h 2 5 4 H Y iyy T 5y ` T 2B T D 5 5 D i d T Y Y H 4 H 2 YB YB F D t 2 5B DB h 2 5 4 i d T h 2 R 5 2 2 T 7 R H ò T 5 D 2 5B D T 7B 5 d T y T 4B d H 7 R 2 T q 5 H Y T 4 H F D 21 H D T d R 4 4 T d H h d 5r h 2 5 4 H Y Y IVdI6fGk8XWVdEcII¢v¢QXW6#px¢WVgIW)3¢WVGpQQVQI83xI8Q6WaXB d 5 D T d H 5 YB F D D T F D H 4B D 5 ò ¥ P 2B 2 H bB ` d 5 D T d H 5 H 4 2 H d H gB h H F D Y H 4 R h 5 d F 4B F 8GVWQ8XpGVIGw6XGVfaph pëjTXkQ¢E8¢GVWQ8QI6W6hIIG6QIIxWpAEIC@ ç 9 ¥ o¢dpzæ
45üÅ W ß V Ý
ß¢¥ î
ø ç Å Ý Ý æ 6 4 Å
ß
8¥ î ¦
ä
Å sÝ ¥ æ ç (
Û d 5 D T d H 5 ` 2B @ 5yy 5q H F D 2B T D 5 H @ t º h 2 T t t d 5q ` 2B by 5 Q8WVG6s¢ IXCcEX8IWX¢Gp8wpsI¢ëc!Up8 IXx8 2 5B DB h 2 5 4 i d T h 2 R 5 2 2 T 7 R H ò d 5q Hy T D d 5y i T t ¥ Hy T ¢XGII86eVUII8I¢#IQC8XIV8EgmjI puXIVvt
4 4
Y sfokº ç ¦æ Y
º 4 Y ç ¦ ræ Y 4 hsGS{
í y xuu t w Y ß
î
u
Ý Y sÔxkº ç ¦æ Y 4
ç ¦ræ Y 4 Y sWx
í y ü w ü uu t Y
î ß
Ý Y sWx½ ç ¦ræ Y
4 hsÔxkº Y ç ¦æ Y 4 ç ¦ ræ Y 4 Y sx
¦ í y ú w ú uu t Y
î ß ß î
45Wí y ¥u t ÉaY î ü w Y º ß í Ý Y xck 4 í Ý Y ¥Q{ í y ú w ú ut Y î ß u
í y û w û uu t Y î ß
í xÝ
ù ×ä
45üí W ß V Ý
º
ø ç í Ý ø 4 í Ý ~¥xSæ
ß
î ¦
ø
í W ß V Ý
¥ Hy T t 2B 2 @ 5 F Y Y I aEI¢meXCspWC¢T t Hy T D d 5y i T t T ` 2B Y R h H 2B 7 d H D H h H iyB Y T H 2 T 4 F 4B F @ Y D 2 T D Y 2 5 4 H d T º h 2 T t t H d H F ¢EI¢¢Xxgv9IXIQIE#QGIus#EEW¢QVgXICuGo¢GI¢QuW¢fIVësëW6IC@ ç $ 9 ¥ 8dpzæ ß î
45üÅ W ß V Ý
º
ø ç4 Å Ý ø Å Ý f5~ÿoæ
ß
î ¦
ä
Å hÝ ¥ æ ç (
Û 7 d 5q ` 2B @ 5yy 5q H F D 2B YB F 4B F W8wpXCcXE8IGeXuEEIC@
H h 5 2 D T d 5 D T d H 5 2 T 2 `B Y H h H @ Y R F {pIClC8GVWQ8¢8XQIwCISut
o
Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u§fj«wÔzmQÔv«SÏmÇ
7 d H D d 5 d d H ` 2B h T Hy H F D h 2B QW8WQIEI¢gEuIGeIEv¤ ç xgh5WxÝ üí Ý þ ø 4 üí
"
ø í Ý º ø 4 í Ý vx~¥x~ø
ß Axoæ î í Ý ¦
ä íç ( uÝ ¥¥ æ
7 d 5q H F D 2B H bB D T bB d H h h dB F D T 5 D 2 5B D T 7B 5 d T H 4 2 H d H gB h r H DB 2 n T H bB d H ã W8IWeX8GVVWQpXIW #W8XWV#XgWIIVuQI6WQsEp3WEIIu¢EW6$ ¡ ä
"
F DB GC@ 8EG6ISVgQCfI 2 5B D Y H R P D T H H
$ ä ê D T 2 H bB ` YB Ý h 2 2l68E¢uXIVT ¡ VQ88WVkß 5Ý I¢Ý ä ê D T 2 H bB ` H d T r h 2 T FD 5 2 H F W¢sk6IC@ 8EG6ISXk68EQI9QpW IWdIEWISQ9W88W668¢XGII864 2 5B D Y H R P 2B h H bB d H h H 7 H F 4 Y H F D ` 2B Y R i h H 7 d 5q d 5 D 4 H b 2 5B DB h 2 5 r i d T h 2 R 5 H F D h 2 T Y H 4B d D T 7 ¸ H F D D R 5 H DB d @ t F Y H 7 D 2B 5 ç d 5B d H D 2Bæ T ` 2B Y H of¢UII¢sIWI¢QQEWWVdsIWp8GECpQ9oX8G8XQWSE pXWfp¥ 7 d H D d 5 d d H ` 2B h T Hy H F D h 2B QW8WQIEI¢gEuIGeIEv¤ ç4 üí f5Ý
"
ø í Ý º ø 4 í Ý x~Axæ
ß î ¦
ä íç (æ ø 4 íç (æ uhÝ ÿhAhÝ
7 d 5q H F D 2B H bB D T bB d H h D Y d n T 5 D 2 5B D T 7B 5 d T H 4 2 H d H gB hr H DB 2 n T H bB d H ã 8IGeX¢EWVVEW6IU #W8XWV#XgWIIVuQI6WQsEp3WEIIu¢EW6 7 d H D d 5 d d H ` 2B h T Hy H F D h 2B QW8WQIEI¢gEuIGeIEv¤ ç4 üí f5WxÝ
"
ø í Ý º ø 4 í Ý vxzh¥x~ø
ß Axoæ î í Ý ¦
ä íç ( uhÝ æ
7d 5 8q H F D 2B H bB D T bB d H h D Y d n T 5 D 2 5B D T 7B 5 d T H 4 2 H d H gB hr H DB 2 n d H h d 5 r h dB F D T H bB d H ã pGXa8GVVEQIufWUdG¢XGldEpgII¢6IQQsEI£WEIUQI83xWXpGd¢EW68¦ ²
R ± ¶ Ã Q·YjÀ
µ
´® G
9 ¥ P 5 D y T 4B D 2 H hB YB F 4B F pV%pëjëGh¢QXWS6IXuEEIC@ ç ¡ Õpzæ c ¥
45üÅ WUß V Ý
ß 8¥ î
ø ç Å 5Ý 85Ý æ 6 4 Å
ß
8¥ î ¦
ä ä ä
Ê¤45üÅ W ß V ß ð
î Ý
ç h 9 ¥ ovdpzæ
ÈD
8¥ ß î ø4ÅÝÅ62ÅÝ8ÆøðÅÝ¢Å684Å¢¥ Ý ¦ ß
î ¦
ç4 Å Ý ø Å ÆªÝ
6 4 ü Å Ý 85æ
Å sÝ ¥ æ ç (
Y H bB Q¢E8` ½IpCVQXEII¢u¢EWVVEW6Ih H h 5 2 D T h HBy T H bB D T bB d H h 2 5 4 H Y T d 5q d 5 D T d H 5 H 4 2 H d H gB h h H d H D 2 H 4 d H h d 5 r h 2 5 4 H Y H F D 5 D 2B YB F D ` 2B D R DB D Y R I86Qºe8¢GVWQ8ÅQI6W6hIå6W6GoQ6A6IW¢3I¢QQ)IWºGoXEIG©IXWpGGpI
9 Â
Ú Ù Ì § × Í Ê Ð ±Ð ª³«c²xÎ
H 7 H F 4 Y F 4 T H d 5q d 5 d d H 2 5B D T 4 2 R d D H F D 2B Y 7 d H V#6IW¢gu¢C¢WW68XWVgIpWGIWeXu9WQWD l d 5q ß î Y b ß q 5 Y D d T id T 2B ` T 7 `IXIVgXev3dIW#¢EW6j l U ¹¡ ¢Æ U#6xÆ î h f¢wGVUf¢UX¢¢dEB 2B h T Hy 5 @D H F D H bB d H ã ß î h 2 T y T H d H F D D 5y Ü Y H 7 H F 4 Y H Y H F D q 5 F 4 T H d 5q d H 7 R 2 H b T @ h H nB h 5 7 H F D h 2B IVu¢geIGw¢Xj¬Q6#6IWW6IGC¢Æ¢Q 8Q ISQ¢c)QIEIx#epG)IEv¤
í Ý h ø 4 í Ý c ¦ 6 í Ý ¦ ¦ ç ß î æ r s¢zIç AxÝ 6 Ax¢~hAx¢a«hxg8Ôæ í
ä íç ( uhÝ æ
í Ý ø 4 í Ý 6 í Ý ¥æ ä íç ( ç ß sî Iç AxÆh¥¢)hx¢ÏuzÝ æ æ r
ß î r ç4 í Ý 6 í Ýæ ä íç (
¢f¥ÅUxÏuÝ æ Û iy H bB D 4 H Y H d t d H h d 5 r h dB F D h 2 Q8GQWQp6QI83XIWp¢T t r h 2 5 4 H Y t r D Y d n H d T F 4B F @ t Y H 7 H F 4 Y ` 2B 4 2 H d H gB h h H hB Yr H 2 5 ` 2B @ 5yy 5q H F D d H hB Y 2 5 GI86QGUf¢eEICpQ6#6IWaIX6IQQsEI#QIEW£I8uIECcXX¢fIWvQpXWp8f ¡ ¦ H bB D T bB d H h D Y d n T 5 D Y 2 5B D T 7B 5 d T h H d H ¢¢EGl¢W6IfWU
9GvI8XWV#XgWIIV6WQWD r 2 H 4 D 4 T 7 5 4 h 2 T D 4 T 7 5 4 2 5 2 d H hd 5r FD BY d 5q D 2 H 4d H ¦ xQQC¢U98QIVC6¢I#86I8Iu6IW¢3W W¢o6QW6su ¡ ¢UWW6Xu8WQQ6sY 2 T F D Y Y Hy d 5 d d H h H H H Y T F T H b HB F 4 T 5 D D 2 H 7 H dB R P H d ç F D ` 2 Hy H b T @r d H r Y D 2B 5 r hB d ` H F D h 2B ¤ ¢UIr86EI¢ÅWoQ9QXISQ
0æ W¢I6XQ¢gfQ3ÔGoX¢s£X8IW©IEvph
l d 5q ß î Y j l U m¡ 8 UQxb ß î ß î h f¢Xjp¢XGfQISeE68WQp8WVG6s8QI83WI¢6¢I#86IGp¢I8dp¥ D 5y Ü 2 5B D Y H R P 2B h H bB d H h d 5 D T d H 5 d H h d 5 r F D d R 5 q D 4 T 7 5 4 H F D h 2 T t ç $ $ ëj8WVG6s¢96IW¢3WI896VU#¢QI8pIWvV8EWVVEQIºp8Q6Wk#¢¢GVWQ85 P æ d 5 D T d H 5 d H h d 5r F D d R 5q D 4 T 7 5 4 2 5 2 H F D t H bB D T bB d H h h 2 5 4 H Y T d 5q d 5 D T d H H 4 2 H d H gB h h H d H D 2 H 4 d H h d 5r h 2 5 4 H Y H F D d 5q d H 7 R 2 H b T @ h H nB h 5 7 H F D h 2B ¤ 2 5B 6IQQsEI¬QQWS6Q 6IW¢3I¢QQdIWw¢6swISQ8gA6UEp9IWIEv8XWD r T 7B 5 d T H bB D T bB d H h h 2 5 4 H Y T d 5 l#XgWIIV8EWVVEQIAI86QWu¢q h Q ISQ¢g6UEIx9IGeIEIU6pwpxWD d H 7 R 2 H b T @ h H nB h 5 7 H F D ` 2B 2 n H h Y R F
6 ä í þ h uç e í
þ ¥ æ (
Y H bB ` H bB D T bB d H h h 2 5 4 H Y H F D d 5q d 5 D T d H 5 H 4 2 H d H gB h T q 5 2 5B D T 4By 68E¢8EWVVEQII8Q6WuIW88GlGQ8uQpQW6hI¢d8XWVgEXIIfi
Y H bB Q¢E8`
ß þ $6 ä þ þ 2 5B D 4 2 Rq H F D 5 D d 5 D T d H 5 H bB D T bB d H hr h 2 5 4 H Y H F D q 5 2 5B D T 4By i 8XW6IpupGG8WVG6s¢w¢EWVVEW6I3xI8Q6WuIWm¢d8XWVgEXIIfCpc
ç $ ¥ P æ d 5 D T d H 5 h H d H D 2 H 4 d H h d 5 r F D d R 5 q H F SVdpëëw8GlGQ86W6GoQ6uQI83WI¢upGD ¢WeQIE¢8VUGeWECuVU 7 5 dq h H 2B T D 5 D T F D F DB @ D d T l d 5q ß î Y b ß q 5 Y D d T i d T 2B ` T 7B h 2 T y T H y T Hd H F D Hd T 7 5 ¢gIG¢U98j l U m¡ 8 U£6 î h #¢GVU9¢UE8¢#XeIV¢gd ß î H F D D 5y Ü ¦ 2 5B D Y H R P 2B h H bB d H h d 5 D T d H 5 H F D d 5q d H 7 R 2 H b T @ h H nB h 5 7 H F D h 2B ¤ pGVXj8e8EG6ISdE68WQp8WVG6s8IG8uQ ISQ¢gd6UEp9IW#IEv9 7 d H D d 5 d d H ` 2B h T Hy H F D h 2B QW8WQIEI¢gEuIGeIEv¤ ç4 üí Ý º ø í Ý ø 4 í Ý 5Wxv~¥xSæ
ß î ¦
ä 4 üíç (æ þ ø íç (æ ø 4 íç ( j5WzÝ ¥ S~jzÝ ¥ ÿAzÝ ¥ æ
"
7 d 5q H F D 2B H bB 8IGEu8EWD r T b B d H h h 2 5 4 H Y T 5 D 2 5 B D T 7 B 5 d T H 4 2 H d H g B h r H D B 2 n ç `H h T Ü d 5 æ D 4 T 7 5 4 T H b B d H ã lVEW6Ip8Q6W9Gd8EGV#XgWpI¢uQpQW6hI3GIUc¨¢mC8a¢U98Qu¢EW6p Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u§fj«wÔzmQÔv«SÏmÇ
c 3
Chapter 4 THE SEMIDISCRETE APPROACH
One strategy for obtaining nitedi erence approximations to a PDE is to start by di erencing the space derivatives only, without approximating the time derivative. In the following chapters, we proceed with an analysis making considerable use of this concept, which we refer to as the semidiscrete approach. Di erencing the space derivatives converts the basic PDE into a set of coupled ODE's. In the most general notation, these ODE's would be expressed in the form d~ = F (~ t) u ~u (4.1) dt which includes all manner of nonlinear and timedependent possibilities. On occasion, we use this form, but the rest of this chapter is devoted to a more specialized matrix notation described below. Another strategy for constructing a nitedi erence approximation to a PDE is to approximate all the partial derivatives at once. This generally leads to a point di erence operator (see Section 3.3.1) which, in turn, can be used for the time advance of the solution at any given point in the mesh. As an example let us consider the model equation for di usion @u = @ u @t @x Using threepoint centraldi erencing schemes for both the time and space derivatives, we nd 2 n 3 ujn ; ujn; uj ; 2ujn + ujn 5 ; = 4 2h x or h i ujn = ujn; + 2h ujn ; 2ujn + ujn (4.2) ; x 51
2 2 ( +1) ( 1) ( ) +1 ( ) 2 ( ) 1 ( +1) ( 1) 2 ( ) +1 ( ) ( ) 1
52
CHAPTER 4. THE SEMIDISCRETE APPROACH
Clearly Eq. 4.2 is a di erence equation which can be used at the space point j to advance the value of u from the previous time levels n and n ; 1 to the level n + 1. It is a full discretization of the PDE. Note, however, that the spatial and temporal discretizations are separable. Thus, this method has an intermediate semidiscrete form and can be analyzed by the methods discussed in the next few chapters. Another possibility is to replace the value of ujn in the right hand side of Eq. 4.2 with the time average of u at that point, namely (ujn + ujn; )=2. This results in the formula
( ) ( +1) ( 1)
ujn
( +1)
un = ujn; + 2h 4ujn ; 2 @ j x
( 1) 2 ( ) +1 ( +1)
2
0
( +1)
+ ujn; A n 5 + uj ; 2
( 1) ( ) 1
1
3
(4.3)
which can be solved for u n and time advanced at the point j . In this case, the spatial and temporal discretizations are not separable, and no semidiscrete form exists. Equation 4.2 is sometimes called Richardson's method of overlapping steps and Eq. 4.3 is referred to as the DuFortFrankel method. As we shall see later on, there are subtle points to be made about using these methods to nd a numerical solution to the di usion equation. There are a number of issues concerning the accuracy, stability, and convergence of Eqs. 4.2 and 4.3 which we cannot comment on until we develop a framework for such investigations. We introduce these methods here only to distinguish between methods in which the temporal and spatial terms are discretized separately and those for which no such separation is possible. For the time being, we shall separate the space di erence approximations from the time di erencing. In this approach, we reduce the governing PDE's to ODE's by discretizing the spatial terms and use the welldeveloped theory of ODE solutions to aid us in the development of an analysis of accuracy and stability.
4.1 Reduction of PDE's to ODE's
4.1.1 The Model ODE's
First let us consider the model PDE's for di usion and biconvection described in Chapter 2. In these simple cases, we can approximate the space derivatives with di erence operators and express the resulting ODE's with a matrix formulation. This is a simple and natural formulation when the ODE's are linear.
4.1. REDUCTION OF PDE'S TO ODE'S
53
Model ODE for Di usion
For example, using the 3point centraldi erencing scheme to represent the second derivative in the scalar PDE governing di usion leads to the following ODE di usion model d~ = u B (1 ;2 1)~ + (bc) u ~ (4.4) dt x ~ with Dirichlet boundary conditions folded into the (bc) vector.
2
Model ODE for Biconvection
The term biconvection was introduced in Section 2.3. It is used for the scalar convection model when the boundary conditions are periodic. In this case, the 3point centraldi erencing approximation produces the ODE model given by
d~ = ; a B (;1 0 1)~ u u (4.5) dt 2 x p where the boundary condition vector is absent because the ow is periodic. Eqs. 4.4 and 4.5 are the model ODE's for di usion and biconvection of a scalar in one dimension. They are linear with coe cient matrices which are independent of x and t.
4.1.2 The Generic Matrix Form
The generic matrix form of a semidiscrete approximation is expressed by the equation
d~ = A~ ; ~ (t) u u f dt
(4.6)
Note that the elements in the matrix A depend upon both the PDE and the type of di erencing scheme chosen for the space terms. The vector ~ (t) is usually determined f by the boundary conditions and possibly source terms. In general, even the Euler and NavierStokes equations can be expressed in the form of Eq. 4.6. In such cases the equations are nonlinear, that is, the elements of A depend on the solution ~ and u are usually derived by nding the Jacobian of a ux vector. Although the equations are nonlinear, the linear analysis presented in this book leads to diagnostics that are surprisingly accurate when used to evaluate many aspects of numerical methods as they apply to the Euler and NavierStokes equations.
54
4.2 Exact Solutions of Linear ODE's
CHAPTER 4. THE SEMIDISCRETE APPROACH
In order to advance Eq. 4.1 in time, the system of ODE's must be integrated using a timemarching method. In order to analyze timemarching methods, we will make use of exact solutions of coupled systems of ODE's, which exist under certain conditions. The ODE's represented by Eq. 4.1 are said to be linear if F is linearly dependent on u (i.e., if @F=@u = A where A is independent of u). As we have already pointed out, when the ODE's are linear they can be expressed in a matrix notation as Eq. 4.6 in which the coe cient matrix, A, is independent of u. If A does depend explicitly on t, the general solution cannot be written whereas, if A does not depend explicitly on t, the general solution to Eq. 4.6 can be written. This holds regardless of whether or ~ not the forcing function, f , depends explicitly on t. As we shall soon see, the exact solution of Eq. 4.6 can be written in terms of the eigenvalues and eigenvectors of A. This will lead us to a representative scalar equation for use in analyzing timemarching methods. These ideas are developed in the following sections.
4.2.1 Eigensystems of SemiDiscrete Linear Forms
Complete Systems
An M M matrix is represented by a complete eigensystem if it has a complete set of linearly independent eigenvectors (see Appendix A) . An eigenvector, ~ m , and its x corresponding eigenvalue, m, have the property that A~ m = m~ m or x x A ; mI]~ m = 0 x (4.7) The eigenvalues are the roots of the equation det A ; I] = 0 We form the righthand eigenvector matrix of a complete system by lling its columns with the eigenvectors ~ m: x h i X = ~ ~ : : : ~M x x x The inverse is the lefthand eigenvector matrix, and together they have the property that
1 2
X ; AX =
1
(4.8)
where is a diagonal matrix whose elements are the eigenvalues of A.
4.2. EXACT SOLUTIONS OF LINEAR ODE'S
55
Defective Systems
If an M M matrix does not have a complete set of linearly independent eigenvectors, it cannot be transformed to a diagonal matrix of scalars, and it is said to be defective. It can, however, be transformed to a diagonal set of blocks, some of which may be scalars (see Appendix A). In general, there exists some S which transforms any matrix A such that S ; AS = J where 2J 3 6 J 7 6 7 6 7 ... 6 7 J =6 7 6 7 Jm 4 5 ... and 2 3 m 1 6 7 1 ... 6 7 ... m n =6 Jm 6 7 . . . 1 7 ... 4 5 m n The matrix J is said to be in Jordan canonical form, and an eigenvalue with multiplicity n within a Jordan block is said to be a defective eigenvalue. Defective systems play a role in numerical stability analysis.
1 1 2 ( )
4.2.2 Single ODE's of First and SecondOrder
FirstOrder Equations
The simplest nonhomogeneous ODE of interest is given by the single, rstorder equation du = u + ae t (4.9) dt where , a, and are scalars, all of which can be complex numbers. The equation is linear because does not depend on u, and has a general solution because does not depend on t. It has a steadystate solution if the righthand side is independent of t, i.e., if = 0, and is homogeneous if the forcing function is zero, i.e., if a = 0. Although it is quite simple, the numerical analysis of Eq. 4.9 displays many of the fundamental properties and issues involved in the construction and study of most popular timemarching methods. This theme will be developed as we proceed.
56
CHAPTER 4. THE SEMIDISCRETE APPROACH
The exact solution of Eq. 4.9 is, for 6= ,
u(t) = c e
1 1
t+
ae t
;
where c is a constant determined by the initial conditions. In terms of the initial value of u, it can be written
t; t u(t) = u(0)e t + a e ; e
The interesting question can arise: What happens to the solution of Eq. 4.9 when = ? This is easily found by setting = + , solving, and then taking the limit as ! 0. Using this limiting device, we nd that the solution to
du = u + ae t dt
is given by
(4.10)
u(t) = u(0) + at]e t As we shall soon see, this solution is required for the analysis of defective systems.
SecondOrder Equations
The homogeneous form of a secondorder equation is given by
d u + a du + a u = 0 (4.11) dt dt where a and a are complex constants. Now we introduce the di erential operator D such that d D dt
2 2 1 0 1 0
and factor u(t) out of Eq. 4.11, giving
2
(D + a D + a ) u(t) = 0
1 0
The polynomial in D is referred to as a characteristic polynomial and designated P (D). Characteristic polynomials are fundamental to the analysis of both ODE's and O E's, since the roots of these polynomials determine the solutions of the equations. For ODE's, we often label these roots in order of increasing modulus as , , , m ,
1 2
4.2. EXACT SOLUTIONS OF LINEAR ODE'S
1 2
57
, M . They are found by solving the equation P ( ) = 0. In our simple example, there would be two roots, and , determined from P( ) = + a + a = 0 (4.12) and the solution to Eq. 4.11 is given by u(t) = c e 1 t + c e 2 t (4.13) where c and c are constants determined from initial conditions. The proof of this is simple and is found by substituting Eq. 4.13 into Eq. 4.11. One nds the result c e 1t( + a + a ) + c e 2t( + a + a ) which is identically zero for all c , c , and t if and only if the 's satisfy Eq. 4.12.
2 1 0 1 2 1 2 1 2 1 1 1 0 2 2 2 1 2 0 1 2
4.2.3 Coupled FirstOrder ODE's
A Complete System
A set of coupled, rstorder, homogeneous equations is given by u0 = a u + a u u0 = a u + a u (4.14) which can be written " # 0 a a ~ = A~ ~ = u u ]T u u u A = (aij ) = a a Consider the possibility that a solution is represented by u = c x e 1t + c x e 2t u = c x e 1t + c x e 2t (4.15) By substitution, these are indeed solutions to Eq. 4.14 if and only if " #" # " # " #" # " # a a x = x a a x = x (4.16) a a x x a a x x Notice that a higherorder equation can be reduced to a coupled set of rstorder equations by introducing a new set of dependent variables. Thus, by setting u = u0 u =u we nd Eq. 4.11 can be written u0 = ;a u ; a u u0 = u (4.17) which is a subset of Eq. 4.14.
1 2 11 21 1 1 12 22 2 2 1 2 11 21 12 22 1 2 1 1 11 21 2 2 12 22 11 12 11 21 21 22 1 11 21 11 12 22 12 22 21 2 12 22 1 2 1 1 1 1 0 2 2
58
CHAPTER 4. THE SEMIDISCRETE APPROACH
1 2
A Derogatory System
Eq. 4.15 is still a solution to Eq. 4.14 if = = , provided two linearly independent vectors exist to satisfy Eq. 4.16 with A = . In this case 0 0 1 = 0 1 0 and 0 0 0 = 1 0 1
provide such a solution. This is the case where A has a complete set of eigenvectors and is not defective.
A Defective System
If A is defective, then it can be represented by the Jordan canonical form
"
# " #" # u0 = 0 u u0 1 u
1 2 1 2
(4.18)
whose solution is not obvious. However, in this case, one can solve the top equation rst, giving u (t) = u (0)e t . Then, substituting this result into the second equation, one nds du = u + u (0)e t dt which is identical in form to Eq. 4.10 and has the solution
1 1 2 2 1
u (t) = u (0) + u (0)t]e t
2 2 1
From this the reader should be able to verify that
u (t) = a + bt + ct e t
3 2
is a solution to
2 03 2 6 u0 7 = 6 1 4u 5 4 u0 1
1 2 3
32 3 76u 7 54u 5 u
1 2 3
if
a = u (0)
3
b = u (0)
2
1 c = 2 u (0)
1
(4.19)
The general solution to such defective systems is left as an exercise.
4.2. EXACT SOLUTIONS OF LINEAR ODE'S
59
4.2.4 General Solution of Coupled ODE's with Complete Eigensystems
Let us consider a set of coupled, nonhomogeneous, linear, rstorder ODE's with constant coe cients which might have been derived by space di erencing a set of PDE's. Represent them by the equation
d~ = A~ ; ~ (t) u u f (4.20) dt Our assumption is that the M M matrix A has a complete eigensystem and can be transformed by the left and right eigenvector matrices, X ; and X , to a diagonal matrix having diagonal elements which are the eigenvalues of A, see Section 4.2.1. Now let us multiply Eq. 4.20 from the left by X ; and insert the identity combination XX ; = I between A and ~ . There results u ~ X ; du = X ; AX X ; ~ ; X ; ~ (t) u f (4.21) dt Since A is independent of both ~ and t, the elements in X ; and X are also indepenu dent of both ~ and t, and Eq. 4.21 can be modi ed to u
1 1 1 1 1 1 1 1 1
d X ; ~ = X ; ~ ; X ; ~ (t) u f dt u ~ Finally, by introducing the new variables w and ~ such that g
1 1 1
~ w = X; ~ u
1
~ (t) = X ; ~ (t) g f
1
(4.22)
we reduce Eq. 4.20 to a new algebraic form
~ dw = w ; ~ (t) ~ g (4.23) dt It is important at this point to review the results of the previous paragraph. Notice that Eqs. 4.20 and 4.23 are expressing exactly the same equality. The only di erence between them was brought about by algebraic manipulations which regrouped the variables. However, this regrouping is crucial for the solution process because Eqs.
In the following, we exclude defective systems, not because they cannot be analyzed (the example at the conclusion of the previous section proves otherwise), but because they are only of limited interest in the general development of our theory.
1
60
CHAPTER 4. THE SEMIDISCRETE APPROACH
4.23 are no longer coupled. They can be written line by line as a set of independent, single, rstorder equations, thus
w0 = ... 0 wm = ... 0 wM =
1
1
w ; g (t)
1 1
m wm ; gm (t) M wM
; gM (t)
(4.24)
For any given set of gm (t) each of these equations can be solved separately and then recoupled, using the inverse of the relations given in Eqs. 4.22: ~ (t) = X w(t) ~ u =
M X m=1
wm(t)~ m x
(4.25)
where ~ m is the m'th column of X , i.e., the eigenvector corresponding to m . x We next focus on the very important subset of Eq. 4.20 when neither A nor ~ has f any explicit dependence on t. In such a case, the gm in Eqs. 4.23 and 4.24 are also time invariant and the solution to any line in Eq. 4.24 is w (t) = c e m t + 1 g
m m m m
where the cm are constants that depend on the initial conditions. Transforming back to the usystem gives ~ (t) = X w(t) ~ u = = = =
M X m=1 M X m=1 M X m=1 M X m=1
wm(t)~ m x cm e m t ~ m + x
m=1 m M X 1 ~ gm xm
cm e m t ~ m + X x
;1 X ;1~ f
1
cm e m t ~ m + A; ~ x f

Transient
{z
} 
Steadystate
{z
}
(4.26)
4.3. REAL SPACE AND EIGENSPACE
1
61
Note that the steadystate solution is A; ~ , as might be expected. f The rst group of terms on the right side of this equation is referred to classically as the complementary solution or the solution of the homogeneous equations. The second group is referred to classically as the particular solution or the particular integral. In our application to uid dynamics, it is more instructive to refer to these groups as the transient and steadystate solutions, respectively. An alternative, but entirely equivalent, form of the solution is ~ (t) = c e 1 t ~ + + cme m t ~ m + + cM e M t ~ M + A; ~ u x x x f (4.27)
1 1 1
4.3 Real Space and Eigenspace
Following the semidiscrete approach discussed in Section 4.1, we reduce the partial di erential equations to a set of ordinary di erential equations represented by the generic form d~ = A~ ; ~ u u f (4.28) dt The dependent variable ~ represents some physical quantity or quantities which relate u to the problem of interest. For the model problems on which we are focusing most of our attention, the elements of A are independent of both u and t. This permits us to say a great deal about such problems and serves as the basis for this section. In particular, we can develop some very important and fundamental concepts that underly the global properties of the numerical solutions to the model problems. How these relate to the numerical solutions of more practical problems depends upon the problem and, to a much greater extent, on the cleverness of the relator. We begin by developing the concept of \spaces". That is, we identify di erent mathematical reference frames (spaces) and view our solutions from within each. In this way, we get di erent perspectives of the same solution, and this can add signi cantly to our understanding. The most natural reference frame is the physical one. We say If a solution is expressed in terms of ~ , it is said u to be in real space. There is, however, another very useful frame. We saw in Sections 4.2.1 and 4.2 that pre and postmultiplication of A by the appropriate similarity matrices transforms A into a diagonal matrix, composed, in the most general case, of Jordan blocks or, in the
4.3.1 De nition
62
CHAPTER 4. THE SEMIDISCRETE APPROACH
simplest nondefective case, of scalars. Following Section 4.2 and, for simplicity, using only complete systems for our examples, we found that Eq. 4.28 had the alternative form ~ dw = w ; ~ ~ g dt which is an uncoupled set of rstorder ODE's that can be solved independently for ~ the dependent variable vector w. We say
~ If a solution is expressed in terms of w, it is said to be in eigenspace (often referred to as wave space). The relations that transfer from one space to the other are: ~ w = X; ~ u ~ = X; ~ g f
1 1
~ = Xw ~ u ~ = X~ f g
The elements of ~ relate directly to the local physics of the problem. However, the u ~ are linear combinations of all of the elements of ~ , and individually elements of w u they have no direct local physical interpretation. When the forcing function ~ is independent of t, the solutions in the two spaces f are represented by ~ (t) = X cm e m t ~ m + X ; X ; ~ u x f and wm(t) = cm e m t + 1 gm m=1 2 M
1 1
for real space and eigenspace, respectively. At this point we make two observations: 1. the transient portion of the solution in real space consists of a linear combination of contributions from each eigenvector, and 2. the transient portion of the solution in eigenspace provides the weighting of each eigenvector component of the transient solution in real space.
m
4.3.2 Eigenvalue Spectrums for Model ODE's
It is instructive to consider the eigenvalue spectrums of the ODE's formulated by central di erencing the model equations for di usion and biconvection. These model ODE's are presented in Section 4.1. Equations for the eigenvalues of the simple
4.3. REAL SPACE AND EIGENSPACE
63
tridiagonals, B (M : a b c) and Bp(M : a b c), are given in Appendix B. From Section B.1, we nd for the model di usion equation with Dirichlet boundary conditions ;2 + 2 cos Mm+ 1 m = x ! m ;4 sin m=1 2 M (4.29) = 2(M + 1) x and, from Section B.4, for the model biconvection equation ;ia sin 2m m=0 1 M ;1 m = x M
2 2 2
where
= ;i m a
m
m=0 1
M ;1
(4.30)
= sin m x m=0 1 M ;1 (4.31) x is the modi ed wavenumber from Section 3.5, m = m, and x = 2 =M . Notice that the di usion eigenvalues are real and negative while those representing periodic convection are all pure imaginary. The interpretation of this result plays a very important role later in our stability analysis.
4.3.3 Eigenvectors of the Model Equations
The Di usion Model
Next we consider the eigenvectors of the two model equations. These follow as special cases from the results given in Appendix B. Consider Eq. 4.4, the model ODE's for di usion. First, to help visualize the matrix structure, we present results for a simple 4point mesh and then we give the general case. The righthand eigenvector matrix X is given by 2 sin (x ) sin (2x ) sin (3x ) sin (4x ) 3 6 sin (x ) sin (2x ) sin (3x ) sin (4x ) 7 6 6 sin (x ) sin (2x ) sin (3x ) sin (4x ) 7 7 4 5 sin (x ) sin (2x ) sin (3x ) sin (4x ) The columns of the matrix are proportional to the eigenvectors. Recall that xj = j x = j =(M + 1), so in general the relation ~ = X w can be written as u ~
1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4
uj =
M X
m=1
wm sin mxj
j=1 2
M
(4.32)
64
CHAPTER 4. THE SEMIDISCRETE APPROACH
For the inverse, or lefthand eigenvector matrix X ; , we nd
1
2 sin (x ) 6 sin (2x ) 6 6 sin (3x ) 4 sin (4x )
1 1 1 1
sin (x ) sin (2x ) sin (3x ) sin (4x )
2 2 2 2
sin (x ) sin (2x ) sin (3x ) sin (4x )
3 3 3 3
sin (x ) 3 sin (2x ) 7 7 5 sin (3x ) 7 sin (4x )
4 4 4 4 1
The rows of the matrix are proportional to the eigenvectors. In general w = X ; ~ ~ u gives
wm =
M X j =1
uj sin mxj
m=1 2
M
(4.33)
In the eld of harmonic analysis, Eq. 4.33 represents a sine transform of the function u(x) for an M point sample between the boundaries x = 0 and x = with the condition u(0) = u( ) = 0. Similarly, Eq. 4.32 represents the sine synthesis that companions the sine transform given by Eq. 4.33. In summary, For the model di usion equation:
~ w = X ; ~ is a sine transform from real space to (sine) wave u space. ~ = Xw ~ u is a sine synthesis from wave space back to real space.
1
The Biconvection Model
Next consider the model ODE's for periodic convection, Eq. 4.5. The coe cient matrices for these ODE's are always circulant. For our model ODE, the righthand eigenvectors are given by
~ m = ei j x
With xj = j
M ;1 X m=0
(2
m=M )
j =0 1 m =0 1
M ;1 M ;1 M ;1
x = j 2 =M , we can write ~ = X w as u ~ uj = wmeimxj j=0 1
(4.34)
4.3. REAL SPACE AND EIGENSPACE 2w 6w 6 6w 4 w 3 21 1 1 1 7 1 6 1 e; i = e; i = e; i = 7= 6 ;i = ;i = ; i = 7 4 61 e 5 4 e e ; i = e; i = e ; i = 1 e
2 4 6 4 4 4 4 8 4 4 6 12 18 12 4 1
65
For a 4point periodic mesh, we nd the following lefthand eigenvector matrix from Appendix B.4:
1 2 3 4 4 4 4
32u 76u 76 76u 54 u
1 2 3 4
3 7 7 = X; ~ 7 u 5
1
X 1 M ; u e;imxj wm = M m=0 1 M ;1 j j This equation is identical to a discrete Fourier transform of the periodic dependent variable ~ using an M point sample between and including x = 0 and x = 2 ; x. u For circulant matrices, it is straightforward to establish the fact that the relation ~ = X w represents the Fourier synthesis of the variable w back to ~ . In summary, ~ ~ u u
=0
In general
For any circulant system:
~ w = X ; ~ is a complex Fourier transform from real space to u wave space. ~ = Xw ~ u is a complex Fourier synthesis from wave space back to real space.
1
4.3.4 Solutions of the Model ODE's
The Di usion Equation
For the di usion equation, Eq. 4.27 becomes
We can now combine the results of the previous sections to write the solutions of our model ODE's.
uj (t) =
where
M X m=1
cme m t sin mxj + (A; f )j
1
j=1 2
M
(4.35)
m
m = ;4 sin 2(M + 1) x
2 2
!
(4.36)
66
CHAPTER 4. THE SEMIDISCRETE APPROACH
(4.37)
With the modi ed wavenumber de ned as 2 m x m = x sin 2 and using m = m, we can write the ODE solution as
uj (t) =
M X m=1
cme;
m 2 t sin m xj
+ (A; f )j
1
j=1 2
M
(4.38)
This can be compared with the exact solution to the PDE, Eq. 2.37, evaluated at the nodes of the grid:
uj (t) =
M X m=1
cm e;
m 2 t sin
m xj + h(xj )
j=1 2
M
(4.39)
We see that the solutions are identical except for the steady solution and the modi ed wavenumber in the transient term. The modi ed wavenumber is an approximation to the actual wavenumber. The di erence between the modi ed wavenumber and the actual wavenumber depends on the di erencing scheme and the grid resolution. This di erence causes the various modes (or eigenvector components) to decay at rates which di er from the exact solution. With conventional di erencing schemes, low wavenumber modes are accurately represented, while high wavenumber modes (if they have signi cant amplitudes) can have large errors.
The Convection Equation
For the biconvection equation, we obtain
uj (t) =
where
M ;1 X m=0
cme m t ei mxj
j=0 1
M ;1
(4.40)
m
= ;i m a
(4.41)
with the modi ed wavenumber de ned in Eq. 4.31. We can write this ODE solution as
uj (t) =
M ;1 X m=0
cme;i m at ei mxj
j=0 1
M ;1
(4.42)
4.4. THE REPRESENTATIVE EQUATION
1
67
and compare it to the exact solution of the PDE, Eq. 2.26, evaluated at the nodes of the grid: M; X uj (t) = fm (0)e;i mat ei m xj j=0 1 M ;1 (4.43) Once again the di erence appears through the modi ed wavenumber contained in . As discussed in Section 3.5, this leads to an error in the speed with which various m modes are convected, since is real. Since the error in the phase speed depends on the wavenumber, while the actual phase speed is independent of the wavenumber, the result is erroneous numerical dispersion. In the case of noncentered di erencing, discussed in Chapter 11, the modi ed wavenumber is complex. The form of Eq. 4.42 shows that the imaginary portion of the modi ed wavenumber produces nonphysical decay or growth in the numerical solution.
m=0
4.4 The Representative Equation
In Section 4.3, we pointed out that Eqs. 4.20 and 4.23 express identical results but in terms of di erent groupings of the dependent variables, which are related by algebraic manipulation. This leads to the following important concept: The numerical solution to a set of linear ODE's (in which A is not a function of t) is entirely equivalent to the solution obtained if the equations are transformed to eigenspace, solved there in their uncoupled form, and then returned as a coupled set to real space. The importance of this concept resides in its message that we can analyze timemarching methods by applying them to a single, uncoupled equation and our conclusions will apply in general. This is helpful both in analyzing the accuracy of timemarching methods and in studying their stability, topics which are covered in Chapters 6 and 7. Our next objective is to nd a \typical" single ODE to analyze. We found the uncoupled solution to a set of ODE's in Section 4.2. A typical member of the family is dwm = w ; g (t) (4.44) m m m dt The goal in our analysis is to study typical behavior of general situations, not particular problems. For such a purpose Eq. 4.44 is not quite satisfactory. The role of m is clear it stands for some representative eigenvalue in the original A matrix. However,
68
CHAPTER 4. THE SEMIDISCRETE APPROACH
the question is: What should we use for gm(t) when the time dependence cannot be ignored? To answer this question, we note that, in principle, one can express any one of the forcing terms gm(t) as a nite Fourier series. For example X ;g(t) = ak eikt
X ak eikt k ik ; From this we can extract the k'th term and replace ik with . This leads to
w(t) = ce t +
The Representative ODE dw = w + ae t dt (4.45)
for which Eq. 4.44 has the exact solution:
k
which can be used to evaluate all manner of timemarching methods. In such evaluations the parameters and must be allowed to take the worst possible combination of values that might occur in the ODE eigensystem. The exact solution of the representative ODE is (for 6= ): t (4.46) w(t) = ce t + ae
;
4.5 Problems
1. Consider the nitedi erence operator derived in question 1 of Chapter 3. Using this operator to approximate the spatial derivative in the linear convection equation, write the semidiscrete form obtained with periodic boundary conditions on a 5point grid (M = 5). 2. Consider the application of the operator given in Eq. 3.52 to the 1D di usion equation with Dirichlet boundary conditions. Write the resulting semidiscrete ODE form. Find the entries in the boundarycondition vector. 3. Write the semidiscrete form resulting from the application of secondorder centered di erences to the following equation on the domain 0 x 1 with boundary conditions u(0) = 0, u(1) = 1: @u = @ u ; 6x @t @x
2 2
4.5. PROBLEMS
69
4. Consider a grid with 10 interior points spanning the domain 0 x . For initial conditions u(x 0) = sin(mx) and boundary conditions u(0 t) = u( t) = 0, plot the exact solution of the di usion equation with = 1 at t = 1 with m = 1 and m = 3. (Plot the solution at the grid nodes only.) Calculate the corresponding modi ed wavenumbers for the secondorder centered operator from Eq. 4.37. Calculate and plot the corresponding ODE solutions. 5. Consider the matrix
A = ;Bp(10 ;1 0 1)=(2 x)
corresponding to the ODE form of the biconvection equation resulting from the application of secondorder central di erencing on a 10point grid. Note that the domain is 0 x 2 and x = 2 =10. The grid nodes are given by xj = j x j = 0 1 : : : 9. The eigenvalues of the above matrix A, as well as the matrices X and X ; , can be found from Appendix B.4. Using these, compute and plot the ODE solution at t = 2 for the initial condition u(x 0) = sin x. Compare with the exact solution of the PDE. Calculate the numerical phase speed from the modi ed wavenumber corresponding to this initial condition and show that it is consistent with the ODE solution. Repeat for the initial condition u(x 0) = sin 2x.
1
70
CHAPTER 4. THE SEMIDISCRETE APPROACH
Chapter 5 FINITEVOLUME METHODS
In Chapter 3, we saw how to derive nitedi erence approximations to arbitrary derivatives. In Chapter 4, we saw that the application of a nitedi erence approximation to the spatial derivatives in our model PDE's produces a coupled set of ODE's. In this Chapter, we will show how similar semidiscrete forms can be derived using nitevolume approximations in space. Finitevolume methods have become popular in CFD as a result, primarily, of two advantages. First, they ensure that the discretization is conservative, i.e., mass, momentum, and energy are conserved in a discrete sense. While this property can usually be obtained using a nitedi erence formulation, it is obtained naturally from a nitevolume formulation. Second, nitevolume methods do not require a coordinate transformation in order to be applied on irregular meshes. As a result, they can be applied on unstructured meshes consisting of arbitrary polyhedra in three dimensions or arbitrary polygons in two dimensions. This increased exibility can be used to great advantage in generating grids about arbitrary geometries. Finitevolume methods are applied to the integral form of the governing equations, either in the form of Eq. 2.1 or Eq. 2.2. Consistent with our emphasis on semidiscrete methods, we will study the latter form, which is
d Z QdV + I n:FdS = Z PdV dt V (t) S (t) V (t)
(5.1)
We will begin by presenting the basic concepts which apply to nitevolume strategies. Next we will give our model equations in the form of Eq. 5.1. This will be followed by several examples which hopefully make these concepts clear. 71
72
5.1 Basic Concepts
CHAPTER 5. FINITEVOLUME METHODS
The basic idea of a nitevolume method is to satisfy the integral form of the conservation law to some degree of approximation for each of many contiguous control volumes which cover the domain of interest. Thus the volume V in Eq. 5.1 is that of a control volume whose shape is dependent on the nature of the grid. In our examples, we will consider only control volumes which do not vary with time. Examining Eq. 5.1, we see that several approximations must be made. The ux is required at the boundary of the control volume, which is a closed surface in three dimensions and a closed contour in two dimensions. This ux must then be integrated to nd the net ux through the boundary. Similarly, the source term P must be integrated over the control volume. Next a timemarching method1 can be applied to nd the value of
Z
V
QdV
(5.2)
at the next time step. Let us consider these approximations in more detail. First, we note that the average value of Q in a cell with volume V is
Q
and Eq. 5.1 can be written as
1 Z QdV V
V
(5.3)
I Z d V dt Q + S n:FdS = V PdV
(5.4)
for a control volume which does not vary with time. Thus after applying a timemarching method, we have updated values of the cellaveraged quantities Q. In order to evaluate the uxes, which are a function of Q, at the controlvolume boundary, Q can be represented within the cell by some piecewise approximation which produces the correct value of Q. This is a form of interpolation often referred to as reconstruction. As we shall see in our examples, each cell will have a di erent piecewise approximation to Q. When these are used to calculate F(Q), they will generally produce di erent approximations to the ux at the boundary between two control volumes, that is, the ux will be discontinuous. A nondissipative scheme analogous to centered di erencing is obtained by taking the average of these two uxes. Another approach known as uxdi erence splitting is described in Chapter 11. The basic elements of a nitevolume method are thus the following:
1
Timemarching methods will be discussed in the next chapter.
5.2. MODEL EQUATIONS IN INTEGRAL FORM
73
1. Given the value of Q for each control volume, construct an approximation to Q(x y z) in each control volume. Using this approximation, nd Q at the controlvolume boundary. Evaluate F(Q) at the boundary. Since there is a distinct approximation to Q(x y z) in each control volume, two distinct values of the ux will generally be obtained at any point on the boundary between two control volumes. 2. Apply some strategy for resolving the discontinuity in the ux at the controlvolume boundary to produce a single value of F(Q) at any point on the boundary. This issue is discussed in Section 11.4.2. 3. Integrate the ux to nd the net ux through the controlvolume boundary using some sort of quadrature. 4. Advance the solution in time to obtain new values of Q. The order of accuracy of the method is dependent on each of the approximations. These ideas should be clari ed by the examples in the remainder of this chapter. In order to include di usive uxes, the following relation between rQ and Q is sometimes used: Z I rQdV = nQdS (5.5) or, in two dimensions,
V S
Z
where the unit vector n points outward from the surface or contour.
A
rQdA = C nQdl
I
(5.6)
5.2 Model Equations in Integral Form
5.2.1 The Linear Convection Equation
A twodimensional form of the linear convection equation can be written as
@u + a cos @u + a sin @u = 0 @t @x @y
(5.7)
This PDE governs a simple plane wave convecting the scalar quantity, u(x y t) with speed a along a straight line making an angle with respect to the xaxis. The onedimensional form is recovered with = 0.
74
CHAPTER 5. FINITEVOLUME METHODS
For unit speed a, the twodimensional linear convection equation is obtained from the general divergence form, Eq. 2.3, with
Q = u F = iu cos + ju sin P = 0 d Z udA + I n:(iu cos + ju sin )ds = 0 dt A C
(5.8) (5.9) (5.10)
Since Q is a scalar, F is simply a vector. Substituting these expressions into a twodimensional form of Eq. 2.2 gives the following integral form (5.11) where A is the area of the cell which is bounded by the closed contour C . The integral form of the twodimensional di usion equation with no source term and unit di usion coe cient is obtained from the general divergence form, Eq. 2.3, with
5.2.2 The Di usion Equation
Q = u F = ;ru ! @u + j @u = ; i @x @y P = 0
Using these, we nd
(5.12) (5.13) (5.14) (5.15)
!
to be the integral form of the twodimensional di usion equation.
d Z udA = I n: i @u + j @u ds dt A @x @y C
(5.16)
5.3 OneDimensional Examples
We restrict our attention to a scalar dependent variable u and a scalar ux f , as in the model equations. We consider an equispaced grid with spacing x. The nodes of the grid are located at xj = j x as usual. Control volume j extends from xj ; x=2 to xj + x=2, as shown in Fig. 5.1. We will use the following notation:
xj;1=2 = xj ; x=2
xj+1=2 = xj + x=2
(5.17)
5.3. ONEDIMENSIONAL EXAMPLES
j1/2 ∆ x j+1/2
75
j2
j1
L R
j
L R
j+1
j+2
Figure 5.1: Control volume in one dimension. With these de nitions, the cellaverage value becomes 1 Z xj+1=2 u(x t)dx u (t)
j
uj 1=2 = u(xj 1=2) x
fj 1=2 = f (uj 1=2)
(5.18) (5.19)
Zx d ( xu ) + f ; fj;1=2 = x j+1=2 Pdx (5.20) j j +1=2 dt j ;1=2 Now with = x ; xj , we can expand u(x) in Eq. 5.19 in a Taylor series about xj (with t xed) to get 2 3 ! ! ! 1 Z x=2 4u + @u + 2 @ 2 u + 3 @ 3 u + : : :5 d uj x ; x=2 j @x j 2 @x2 j 6 @x3 j ! ! x2 @ 2 u + x4 @ 4 u + O( x6 ) = uj + 24 @x2 (5.21) 4 j 1920 @x j
and the integral form becomes
xj;1=2
or
(5.22) where uj is the value at the center of the cell. Hence the cellaverage value and the value at the center of the cell di er by a term of second order.
uj = uj + O( x2)
5.3.1 A SecondOrder Approximation to the Convection Equation
In one dimension, the integral form of the linear convection equation, Eq. 5.11, becomes u (5.23) x ddtj + fj+1=2 ; fj;1=2 = 0
76
CHAPTER 5. FINITEVOLUME METHODS
with f = u. We choose a piecewise constant approximation to u(x) in each cell such that u(x) = uj xj;1=2 x xj+1=2 (5.24) Evaluating this at j + 1=2 gives fjL+1=2 = f (uL+1=2) = uL+1=2 = uj (5.25) j j where the L indicates that this approximation to fj+1=2 is obtained from the approximation to u(x) in the cell to the left of xj+1=2 , as shown in Fig. 5.1. The cell to the right of xj +1=2, which is cell j + 1, gives fjR =2 = uj+1 (5.26) +1 Similarly, cell j is the cell to the right of xj;1=2 , giving fjR 1=2 = uj (5.27) ; and cell j ; 1 is the cell to the left of xj;1=2, giving fjL;1=2 = uj;1 (5.28) We have now accomplished the rst step from the list in Section 5.1 we have de ned the uxes at the cell boundaries in terms of the cellaverage data. In this example, the discontinuity in the ux at the cell boundary is resolved by taking the average of the uxes on either side of the boundary. Thus 1 f^j+1=2 = 2 (fjL+1=2 + fjR =2) = 1 (uj + uj+1) (5.29) +1 2 and 1 f^j;1=2 = 2 (fjL;1=2 + fjR 1=2) = 1 (uj;1 + uj ) (5.30) ; 2 where f^ denotes a numerical ux which is an approximation to the exact ux. Substituting Eqs. 5.29 and 5.30 into the integral form, Eq. 5.23, we obtain u 1 u x ddtj + 2 (uj + uj+1) ; 1 (uj;1 + uj ) = x ddtj + 1 (uj+1 ; uj;1) = 0 (5.31) 2 2 With periodic boundary conditions, this point operator produces the following semidiscrete form: d~ = ; 1 B (;1 0 1)~ u u (5.32) dt 2 x p
5.3. ONEDIMENSIONAL EXAMPLES
77
This is identical to the expression obtained using secondorder centered di erences, except it is written in terms of the cell average ~ , rather than the nodal values, ~ . u u Hence our analysis and understanding of the eigensystem of the matrix Bp(;1 0 1) is relevant to nitevolume methods as well as nitedi erence methods. Since the eigenvalues of Bp(;1 0 1) are pure imaginary, we can conclude that the use of the average of the uxes on either side of the cell boundary, as in Eqs. 5.29 and 5.30, can lead to a nondissipative nitevolume method.
5.3.2 A FourthOrder Approximation to the Convection Equation
Let us replace the piecewise constant approximation in Section 5.3.1 with a piecewise quadratic approximation as follows u( ) = a 2 + b + c (5.33) where is again equal to x ; xj . The three parameters a, b, and c are chosen to satisfy the following constraints: 1 Z ; x=2 u( )d = u
x ;3 x=2
1
j ;1
x
Z x=2
; x=2
x=2
u( )d = uj
(5.34)
1 These constraints lead to
x
Z 3 x=2
u( )d = uj+1
j a = uj+1 ;22ux2+ uj;1
u b = uj+12 ;x j;1
(5.35)
the following values at the cell boundaries: uL+1=2 = 1 (2uj+1 + 5uj ; uj;1) j 6
c = ;uj;1 + 26uj ; uj+1 24 With these values of a, b, and c, the piecewise quadratic approximation produces
(5.36)
78
CHAPTER 5. FINITEVOLUME METHODS uR;1=2 = 1 (;uj+1 + 5uj + 2uj;1) j 6 uR+1=2 = 1 (;uj+2 + 5uj+1 + 2uj ) j 6
1 uL;1=2 = 6 (2uj + 5uj;1 ; uj;2) j (5.37) (5.38) (5.39)
using the notation de ned in Section 5.3.1. Recalling that f = u, we again use the average of the uxes on either side of the boundary to obtain f^j+1=2 = 1 f (uL+1=2) + f (uR+1=2)] j 2 j 1 (5.40) = 12 (;uj+2 + 7uj+1 + 7uj ; uj;1) and 1 f^j;1=2 = 2 f (uL;1=2) + f (uR;1=2)] j j = 1 (;uj+1 + 7uj + 7uj;1 ; uj;2) 12
(5.41)
Substituting these expressions into the integral form, Eq. 5.23, gives u 1 x ddtj + 12 (;uj+2 + 8uj+1 ; 8uj;1 + uj;2) = 0 (5.42) This is a fourthorder approximation to the integral form of the equation, as can be veri ed using Taylor series expansions (see question 1 at the end of this chapter). With periodic boundary conditions, the following semidiscrete form is obtained: d~ = ; 1 B (1 ;8 0 8 ;1)~ u u (5.43) dt 12 x p This is a system of ODE's governing the evolution of the cellaverage data.
In this section, we describe two approaches to deriving a nitevolume approximation to the di usion equation. The rst approach is simpler to extend to multidimensions, while the second approach is more suited to extension to higher order accuracy.
5.3.3 A SecondOrder Approximation to the Di usion Equation
5.3. ONEDIMENSIONAL EXAMPLES u x ddtj + fj+1=2 ; fj;1=2 = 0
Zb
79 (5.44) (5.45)
In one dimension, the integral form of the di usion equation, Eq. 5.16, becomes with f = ;ru = ;@u=@x. Also, Eq. 5.6 becomes
a
We can thus write the following expression for the average value of the gradient of u over the interval xj x xj+1: 1 Z xj+1 @u dx = 1 (u ; u ) (5.46)
@u dx = u(b) ; u(a) @x
x
xj
@x
x j+1
j
From Eq. 5.22, we know that the value of a continuous function at the center of a given interval is equal to the average value of the function over the interval to secondorder accuracy. Hence, to secondorder, we can write ! @u f^ = ; = ; 1 (u ; u ) (5.47)
j +1=2
@x j+1=2
x j+1
j
Similarly, Substituting these into the integral form, Eq. 5.44, we obtain x duj = 1 (u ; 2u + u ) or, with Dirichlet boundary conditions, d~ = 1 B (1 ;2 1)~ + bc u u ~ 2
f^j;1=2 = ; 1x (uj ; uj;1) dt x
j ;1 j
(5.48) (5.49) (5.50)
j +1
This provides a semidiscrete nitevolume approximation to the di usion equation, and we see that the properties of the matrix B (1 ;2 1) are relevant to the study of nitevolume methods as well as nitedi erence methods. For our second approach, we use a piecewise quadratic approximation as in Section 5.3.2. From Eq. 5.33 we have
dt
x
@u = @u = 2a + b @x @
(5.51)
80
CHAPTER 5. FINITEVOLUME METHODS
j +1=2 j +1=2
with a and b given in Eq. 5.35. With f = ;@u=@x, this gives f R = f L = ; 1 (u ; u )
x j+1
j
(5.52) (5.53)
Notice that there is no discontinuity in the ux at the cell boundary. This produces duj = 1 (u ; 2u + u ) (5.54) which is identical to Eq. 5.49. The resulting semidiscrete form with periodic boundary conditions is d~ = 1 B (1 ;2 1)~ u u (5.55) which is written entirely in terms of cellaverage data.
fjR 1=2 = fjL;1=2 = ; 1x (uj ; uj;1) ; dt x2
j ;1 j j +1
dt
x2
p
5.4 A TwoDimensional Example
The above onedimensional examples of nitevolume approximations obscure some of the practical aspects of such methods. Thus our nal example is a nitevolume approximation to the twodimensional linear convection equation on a grid consisting of regular triangles, as shown in Figure 5.2. As in Section 5.3.1, we use a piecewise constant approximation in each control volume and the ux at the control volume boundary is the average of the uxes obtained on either side of the boundary. The nodal data are stored at the vertices of the triangles formed by the grid. The control volumes are regular hexagons with area A, is the length of the sides of the triangles, and ` is the length of the sides of the hexagons. The following relations hold between `, , and A. 1 ` = p 3 p A = 3 2 3 `2 ` = 2 (5.56) A 3 The twodimensional form of the conservation law is
d Z QdA + I n:Fdl = 0 dt A C
(5.57)
5.4. A TWODIMENSIONAL EXAMPLE
c b
2 3
81
p d
4
1
l
∆
0 5
a
e
f
Figure 5.2: Triangular grid. where we have ignored the source term. The contour in the line integral is composed of the sides of the hexagon. Since these sides are all straight, the unit normals can be taken outside the integral and the ux balance is given by
5 d Z Q dA + X n dt A =0
Z
Fdl = 0
where indexes a side of the hexagon, as shown in Figure 5.2. A list of the normals for the mesh orientation shown is given in Table 5.1. Side 0 1 2 3 4 5 Outward Normal n (i ; 3 j)=2 pi (i + p j)=2 3 (;i + 3 j)=2 ;i p (;i ; 3 j)=2
p
Table 5.1. Outward normals, see Fig. 5.2. i and j are unit normals along x and y, respectively.
82
CHAPTER 5. FINITEVOLUME METHODS
For Eq. 5.11, the twodimensional linear convection equation, we have for side
n
Z
Fdl = n (i cos + j sin )
Z `=2
;`=2
u ( )d
(5.58)
where is a length measured from the middle of a side . Making the change of variable z = =`, one has the expression
u( )d = ` ;1=2 u(z)dz (5.59) Then, in terms of u and the hexagon area A, we have " # 5 d Z u dA + X n (i cos + j sin ) ` Z 1=2 u(z)dz = 0 (5.60) dt A ;1=2 =0 The values of n (i cos + j sin ) are given by the expressions in Table 5.2. There
;`=2
Z `=2
Z 1=2
are no numerical approximations in Eq. 5.60. That is, if the integrals in the equation are evaluated exactly, the integrated time rate of change of the integral of u over the area of the hexagon is known exactly. Side 0 1 2 3 4 5
n (i cos + j sin )
(cos ; 3 sin )=2 cos p (cos + p sin )=2 3 (; cos + 3 sin )=2 ; cos p (; cos ; 3 sin )=2
p
Table 5.2. Weights of ux integrals, see Eq. 5.60. Introducing the cell average,
Z
and the piecewiseconstant approximation u = up over the entire hexagon, the approximation to the ux integral becomes trivial. Taking the average of the ux on either side of each edge of the hexagon gives for edge 1: Z up + ua Z 1=2 dz = up + ua u(z)dz = 2 ;1=2 (5.62) 2 1
A
u dA = Aup
(5.61)
5.5. PROBLEMS
Similarly, we have for the other ve edges:
Z
83
u(z)dz = up + ub 2 2 u(z)dz = up + uc 2 3
(5.63) (5.64) (5.65) (5.66) (5.67)
Z Z
u(z)dz = up + ud 2 4 u(z)dz = up + ue 2 5
Z Z
Substituting these into Eq. 5.60, along with the expressions in Table 5.2, we obtain
u(z)dz = up + uf 2 0
p u ` A ddtp + 2 (2 cos )(ua ; ud) + (cos + 3 sin )(ub ; ue) p +(; cos + 3 sin )(uc ; uf )] = 0
(5.68)
or
dup + 1 (2 cos )(u ; u ) + (cos + p3 sin )(u ; u ) a d b e dt 3 p +(; cos + 3 sin )(uc ; uf )] = 0
(5.69)
The reader can verify, using Taylor series expansions, that this is a secondorder approximation to the integral form of the twodimensional linear convection equation.
5.5 Problems
1. Use Taylor series to verify that Eq. 5.42 is a fourthorder approximation to Eq. 5.23. 2. Find the semidiscrete ODE form governing the cellaverage data resulting from the use of a linear approximation in developing a nitevolume method for the linear convection equation. Use the following linear approximation:
u( ) = a + b
84 where b = uj and
CHAPTER 5. FINITEVOLUME METHODS u a = uj+12 ;x j;1
and use the average ux at the cell interface. 3. Using the rst approach given in Section 5.3.3, derive a nitevolume approximation to the spatial terms in the twodimensional di usion equation on a square grid. 4. Repeat question 3 for a grid consisting of equilateral triangles.
Chapter 6 TIMEMARCHING METHODS FOR ODE'S
After discretizing the spatial derivatives in the governing PDE's (such as the NavierStokes equations), we obtain a coupled system of nonlinear ODE's in the form These can be integrated in time using a timemarching method to obtain a timeaccurate solution to an unsteady ow problem. For a steady ow problem, spatial discretization leads to a coupled system of nonlinear algebraic equations in the form ~u F (~ ) = 0 (6.2) As a result of the nonlinearity of these equations, some sort of iterative method is required to obtain a solution. For example, one can consider the use of Newton's method, which is widely used for nonlinear algebraic equations (See Section 6.10.3.). This produces an iterative method in which a coupled system of linear algebraic equations must be solved at each iteration. These can be solved iteratively using relaxation methods, which will be discussed in Chapter 9, or directly using Gaussian elimination or some variation thereof. Alternatively, one can consider a timedependent path to the steady state and use a timemarching method to integrate the unsteady form of the equations until the solution is su ciently close to the steady solution. The subject of the present chapter, timemarching methods for ODE's, is thus relevant to both steady and unsteady ow problems. When using a timemarching method to compute steady ows, the goal is simply to remove the transient portion of the solution as quickly as possible timeaccuracy is not required. This motivates the study of stability and sti ness, topics which are covered in the next two chapters. 85
d~ = F (~ t) u ~u dt
(6.1)
86
CHAPTER 6. TIMEMARCHING METHODS FOR ODE'S
Application of a spatial discretization to a PDE produces a coupled system of ODE's. Application of a timemarching method to an ODE produces an ordinary di erence equation (O E ). In earlier chapters, we developed exact solutions to our model PDE's and ODE's. In this chapter we will present some basic theory of linear O E's which closely parallels that for linear ODE's, and, using this theory, we will develop exact solutions for the model O E's arising from the application of timemarching methods to the model ODE's.
6.1 Notation
Using the semidiscrete approach, we reduce our PDE to a set of coupled ODE's represented in general by Eq. 4.1. However, for the purpose of this chapter, we need only consider the scalar case Although we use u to represent the dependent variable, rather than w, the reader should recall the arguments made in Chapter 4 to justify the study of a scalar ODE. Our rst task is to nd numerical approximations that can be used to carry out the time integration of Eq. 6.3 to some given accuracy, where accuracy can be measured either in a local or a global sense. We then face a further task concerning the numerical stability of the resulting methods, but we postpone such considerations to the next chapter. In Chapter 2, we introduced the convention that the n subscript, or the (n) superscript, always points to a discrete time value, and h represents the time interval t. Combining this notation with Eq. 6.3 gives u0n = Fn = F (un tn) tn = nh Often we need a more sophisticated notation for intermediate time steps involving temporary calculations denoted by u, u, etc. For these we use the notation ~ ~ u0n+ = Fn+ = F (~n+ tn + h) ~ u The choice of u0 or F to express the derivative in a scheme is arbitrary. They are both commonly used in the literature on ODE's. The methods we study are to be applied to linear or nonlinear ODE's, but the methods themselves are formed by linear combinations of the dependent variable and its derivative at various time intervals. They are represented conceptually by (6.4) un+1 = f 1 u0n+1 0u0n ;1u0n;1 0 un ;1 un;1
du = u0 = F (u t) dt
(6.3)
6.2. CONVERTING TIMEMARCHING METHODS TO O E'S
87
With an appropriate choice of the 0s and 0s, these methods can be constructed to give a local Taylor series accuracy of any order. The methods are said to be explicit if 1 = 0 and implicit otherwise. An explicit method is one in which the new predicted solution is only a function of known data, for example, u0n, u0n;1, un and un;1 for a method using two previous time levels, and therefore the time advance is simple. For an implicit method, the new predicted solution is also a function of the time derivative at the new time level, that is, u0n+1. As we shall see, for systems of ODE's and nonlinear problems, implicit methods require more complicated strategies to solve for un+1 than explicit methods.
6.2 Converting TimeMarching Methods to O E's
Examples of some very common forms of methods used for timemarching general ODE's are:
un+1 = un + hu0n un+1 = un + hu0n+1
and
(6.5) (6.6)
According to the conditions presented under Eq. 6.4, the rst and third of these are examples of explicit methods. We refer to them as the explicit Euler method and the MacCormack predictorcorrector method,1 respectively. The second is implicit and referred to as the implicit (or backward) Euler method. These methods are simple recipes for the time advance of a function in terms of its value and the value of its derivative, at given time intervals. The material presented in Chapter 4 develops a basis for evaluating such methods by introducing the concept of the representative equation
un+1 = un + hu0n ~ 1 ~ ~ un+1 = 2 un + un+1 + hu0n+1]
(6.7)
du = u0 = u + ae t (6.8) dt written here in terms of the dependent variable, u. The value of this equation arises
from the fact that, by applying a timemarching method, we can analytically convert
MacCormack's method, which is a fullydiscrete method, will be presented in Section 11.3
1 Here we give only MacCormack's timemarching method. The method commonly referred to as
88
CHAPTER 6. TIMEMARCHING METHODS FOR ODE'S
such a linear ODE into a linear O E . The latter are subject to a whole body of analysis that is similar in many respects to, and just as powerful as, the theory of ODE's. We next consider examples of this conversion process and then go into the general theory on solving O E's. Apply the simple explicit Euler scheme, Eq. 6.5, to Eq. 6.8. There results un+1 = un + h( un + ae hn) or un+1 ; (1 + h)un = hae hn (6.9) Eq. 6.9 is a linear O E , with constant coe cients, expressed in terms of the dependent variable un and the independent variable n. As another example, applying the implicit Euler method, Eq. 6.6, to Eq. 6.8, we nd un+1 = un + h un+1 + ae h(n+1) or (1 ; h)un+1 ; un = he h ae hn (6.10) As a nal example, the predictorcorrector sequence, Eq. 6.7, gives un+1 ; (1 + h)un = ahe hn ~ 1 u (6.11) ; 2 (1 + h)~n+1 + un+1 ; 1 un = 1 ahe h(n+1) 2 2 which is a coupled set of linear O E's with constant coe cients. Note that the rst line in Eq. 6.11 is identical to Eq. 6.9, since the predictor step in Eq. 6.7 is simply the explicit Euler method. The second line in Eq. 6.11 is obtained by noting that u0n+1 = F (~n+1 tn + h) ~ u = un+1 + ae h(n+1) ~ (6.12) Now we need to develop techniques for analyzing these di erence equations so that we can compare the merits of the timemarching methods that generated them.
6.3 Solution of Linear O E's With Constant Coe cients
The techniques for solving linear di erence equations with constant coe cients is as well developed as that for ODE's and the theory follows a remarkably parallel path. This is demonstrated by repeating some of the developments in Section 4.2, but for di erence rather than di erential equations.
6.3. SOLUTION OF LINEAR O E'S WITH CONSTANT COEFFICIENTS
89
6.3.1 First and SecondOrder Di erence Equations
FirstOrder Equations
The simplest nonhomogeneous O E of interest is given by the single rstorder equation
un+1 = un + abn
(6.13)
where , a, and b are, in general, complex parameters. The independent variable is now n rather than t, and since the equations are linear and have constant coe cients, is not a function of either n or u. The exact solution of Eq. 6.13 is where c1 is a constant determined by the initial conditions. In terms of the initial value of u it can be written Just as in the development of Eq. 4.10, one can readily show that the solution of the defective case, (b = ), un+1 = un + a n is h i un = u0 + an ;1 n This can all be easily veri ed by substitution.
n; un = u0 n + a b b ; n
un = c1 n + bab ;
n
SecondOrder Equations
The homogeneous form of a secondorder di erence equation is given by
un+2 + a1 un+1 + a0un = 0
(6.14)
d Instead of the di erential operator D dt used for ODE's, we use for O E's the di erence operator E (commonly referred to as the displacement or shift operator) and de ned formally by the relations
un+1 = Eun
un+k = E k un
Further notice that the displacement operator also applies to exponents, thus
b bn = bn+ = E bn
90
CHAPTER 6. TIMEMARCHING METHODS FOR ODE'S
where can be any fraction or irrational number. The roles of D and E are the same insofar as once they have been introduced to the basic equations the value of u(t) or un can be factored out. Thus Eq. 6.14 can now be reexpressed in an operational notion as (E 2 + a1E + a0)un = 0 (6.15) which must be zero for all un. Eq. 6.15 is known as the operational form of Eq. 6.14. The operational form contains a characteristic polynomial P (E ) which plays the same role for di erence equations that P (D) played for di erential equations that is, its roots determine the solution to the O E. In the analysis of O E's, we label these roots 1 , 2 , , etc, and refer to them as the roots. They are found by solving the equation P ( ) = 0. In the simple example given above, there are just two roots and in terms of them the solution can be written
un = c1( 1 )n + c2( 2 )n
(6.16)
where c1 and c2 depend upon the initial conditions. The fact that Eq. 6.16 is a solution to Eq. 6.14 for all c1 c2 and n should be veri ed by substitution.
6.3.2 Special Cases of Coupled FirstOrder Equations
A Complete System
Coupled, rstorder, linear homogeneous di erence equations have the form
u(1n+1) = c11u(1n) + c12 u(2n) u(2n+1) = c21u(1n) + c22 u(2n) ~ n = hu(1n) u(2n)iT u c12
(6.17)
which can also be written
~ n+1 = C~ n u u
(c11 ; E )
C = c11 c12 c21 c22
The operational form of Eq. 6.17 can be written
u1 (n) = C ; E I ]~ = 0 un c21 (c22 ; E ) u2 which must be zero for all u1 and u2. Again we are led to a characteristic polynomial, this time having the form P (E ) = det C ; E I ]. The roots are found from P ( ) = det (c11c ; ) (c c12 ) = 0 21 22 ;
6.4. SOLUTION OF THE REPRESENTATIVE O E'S
91
Obviously the k are the eigenvalues of C and, following the logic of Section 4.2, ~ are its eigenvectors, the solution of Eq. 6.17 is if x 2 ~ n = X ck ( k )n~ k u x
k=1
where ck are constants determined by the initial conditions. The solution of O E's with defective eigensystems follows closely the logic in Section 4.2.2 for defective ODE's. For example, one can show that the solution to
A Defective System
2 3 2 un+1 7 6 6 un+1 5 = 4 1 4^ un+1 1
32 3 7 6 un 7 ^ 5 4 un 5 un
is
un = u0 n h i un = u0 + u0n ;1 n ^ ^ " # ;1 + u n(n ; 1) ;2 un = u0 + u0n ^ 0 2
n
(6.18)
6.4 Solution of the Representative O E's
6.4.1 The Operational Form and its Solution
Examples of the nonhomogeneous, linear, rstorder ordinary di erence equations, produced by applying a timemarching method to the representative equation, are given by Eqs. 6.9 to 6.11. Using the displacement operator, E , these equations can be written
E ; (1 + h)]un = h ae hn
(1 ; h)E ; 1]un = h E ae
hn
(6.19) (6.20) (6.21)
"
;(1 + h) # " u # = h " 1 # ae hn ~ 1E 1 (1 + h)E E ; 1 u n ;2 2 2
E
92
CHAPTER 6. TIMEMARCHING METHODS FOR ODE'S
All three of these equations are subsets of the operational form of the representative O E P (E )un = Q(E ) ae hn (6.22) which is produced by applying timemarching methods to the representative ODE, Eq. 4.45. We can express in terms of Eq. 6.22 all manner of standard timemarching methods having multiple time steps and various types of intermediate predictorcorrector families. The terms P (E ) and Q(E ) are polynomials in E referred to as the characteristic polynomial and the particular polynomial, respectively. The general solution of Eq. 6.22 can be expressed as K h X un = ck ( k )n + ae hn Q(e h) (6.23) P (e ) k=1 where k are the K roots of the characteristic polynomial, P ( ) = 0. When determinants are involved in the construction of P (E ) and Q(E ), as would be the case for Eq. 6.21, the ratio Q(E )=P (E ) can be found by Kramer's rule. Keep in mind that for methods such as in Eq. 6.21 there are multiple (two in this case) solutions, one for un and un and we are usually only interested in the nal solution un. Notice also, ~ the important subset of this solution which occurs when = 0, representing a time invariant particular solution, or a steady state. In such a case K X Q(1) un = ck ( k )n + a P (1) k=1
6.4.2 Examples of Solutions to TimeMarching O E's
As examples of the use of Eqs. 6.22 and 6.23, we derive the solutions of Eqs. 6.19 to 6.21. For the explicit Euler method, Eq. 6.19, we have P (E ) = E ; 1 ; h Q(E ) = h (6.24) and the solution of its representative O E follows immediately from Eq. 6.23:
un = c1(1 + h)n + ae hn
For the implicit Euler method, Eq. 6.20, we have P (E ) = (1 ; h)E ; 1 Q(E ) = hE
e
h;1;
h
h
(6.25)
6.5. THE ; RELATION
so
93
In the case of the coupled predictorcorrector equations, Eq. 6.21, one solves for the nal family un (one can also nd a solution for the intermediate family u), and there ~ results # " E ;(1 + h) = E E ; 1 ; h ; 1 2h2 P (E ) = det ; 1 (1 + h)E E ; 1 2 2 2 h Q(E ) = det ; 1 (1 E h)E 1 hE = 1 hE (E + 1 + h) 2 2 + 2 The root is found from 1 P( ) = ; 1 ; h ; 2 2 h2 = 0 which has only one nontrivial root ( = 0 is simply a shift in the reference index). The complete solution can therefore be written 1 h 1 2h2 n + ae hn 2 h e + 1 + h un = c1 1 + h + 2 (6.26) e h ; 1 ; h ; 1 2h2 2
he h 1 n un = c1 1 ; h + ae hn (1 ; h)e h ; 1
6.5.1 Establishing the Relation
6.5 The ; Relation
We have now been introduced to two basic kinds of roots, the roots and the roots. The former are the eigenvalues of the A matrix in the ODE's found by space di erencing the original PDE, and the latter are the roots of the characteristic polynomial in a representative O E found by applying a timemarching method to the representative ODE. There is a fundamental relation between the two which can be used to identify many of the essential properties of a timemarch method. This relation is rst demonstrated by developing it for the explicit Euler method. First we make use of the semidiscrete approach to nd a system of ODE's and then express its solution in the form of Eq. 4.27. Remembering that t = nh, one can write
n u(t) = c1 e 1 h ~ 1 + x
+ cm e
mh n
~m + x
+ cM e
Mh n
~ M + P:S: (6.27) x
94
CHAPTER 6. TIMEMARCHING METHODS FOR ODE'S
where for the present we are not interested in the form of the particular solution (P:S:). Now the explicit Euler method produces for each root, one root, which is given by = 1 + h. So if we use the Euler method for the time advance of the ODE's, the solution2 of the resulting O E is
un = c1( 1 )n ~ 1 + x
+ cm ( m )n ~ m + x
+ cM ( M )n ~ M + P:S: x
(6.28)
where the cm and the ~ m in the two equations are identical and m = (1 + m h). x Comparing Eq. 6.27 and Eq. 6.28, we see a correspondence between m and e m h. Since the value of e h can be expressed in terms of the series 1 e h = 1 + h + 2 2h2 + 1 3h3 + 6 1 + n! nhn +
the truncated expansion = 1 + h is a reasonable3 approximation for small enough h. Suppose, instead of the Euler method, we use the leapfrog method for the time advance, which is de ned by
un+1 = un;1 + 2hu0n
(6.29)
Applying Eq. 6.8 to Eq. 6.29, we have the characteristic polynomial P (E ) = E 2 ; 2 hE ; 1, so that for every the must satisfy the relation
m ; 2 mh m ; 1 = 0
2
(6.30)
Now we notice that each produces two roots. For one of these we nd
m
q h+ 1+ m 1 = 1 + mh + 2
=
2 h2 ; 1 4 h4 +
m
2 h2 m
(6.31) (6.32)
mh
8
m
This is an approximation to e m h with an error O( 3h3). The other root, q 1 + 2 h2 , will be discussed in Section 6.5.3. m
2 Based on Section 4.4. 3 The error is O( 2 h2 ).
;
6.5. THE ; RELATION
95
6.5.2 The Principal Root
Based on the above we make the following observation: Application of the same timemarching method to all of the equations in a coupled system linear ODE's in the form of Eq. 4.6, always produces one root for every root that satis es the relation (6.33) 1 2h2 + + 1 k hk + O hk+1 =1+ h+ 2 k! where k is the order of the timemarching method.
We refer to the root that has the above property as the principal root, and designate it ( m )1 . The above property can be stated regardless of the details of the timemarching method, knowing only that its leading error is O hk+1 . Thus the principal root is an approximation to e h up to O hk . Note that a secondorder approximation to a derivative written in the form ( t u)n = 1 (un+1 ; un;1) (6.34) 2h has a leading truncation error which is O(h2), while the secondorder timemarching method which results from this approximation, which is the leapfrog method:
un+1 = un;1 + 2hu0n
(6.35)
has a leading truncation error O(h3). This arises simply because of our notation for the timemarching method in which we have multiplied through by h to get an approximation for the function un+1 rather than the derivative as in Eq. 6.34. The following example makes this clear. Consider a solution obtained at a given time T using a secondorder timemarching method with a time step h. Now consider the solution obtained using the same method with a time step h=2. Since the error per time step is O(h3), this is reduced by a factor of eight (considering the leading term only). However, twice as many time steps are required to reach the time T . Therefore the error at the end of the simulation is reduced by a factor of four, consistent with a secondorder approximation.
6.5.3 Spurious Roots
We saw from Eq. 6.30 that the ; relation for the leapfrog method produces two roots for each . One of these we identi ed as the principal root which always
96
CHAPTER 6. TIMEMARCHING METHODS FOR ODE'S
has the property given in 6.33. The other is referred to as a spurious root and designated ( m )2 . In general, the ; relation produced by a timemarching scheme can result in multiple roots all of which, except for the principal one, are spurious. All spurious roots are designated ( m )k where k = 2 3 . No matter whether a root is principal or spurious, it is always some algebraic function of the product h. To express this fact we use the notation = ( h). If a timemarching method produces spurious roots, the solution for the O E in the form shown in Eq. 6.28 must be modi ed. Following again the message of Section 4.4, we have un = c11 ( 1 )n ~ 1 + + cm1( m )n ~ m + + cM 1 ( M )n ~ M + P:S: 1x 1x 1x n~ + n~ + +c12 ( 1)2 x1 + cm2 ( m )2 xm + cM 2 ( M )n ~ M 2x n~ + n~ + +c13 ( 1)3 x1 + cm3 ( m )3 xm + cM 3 ( M )n ~ M 3x +etc., if there are more spurious roots (6.36) Spurious roots arise if a method uses data from time level n ; 1 or earlier to advance the solution from time level n to n + 1. Such roots originate entirely from the numerical approximation of the timemarching method and have nothing to do with the ODE being solved. However, generation of spurious roots does not, in itself, make a method inferior. In fact, many very accurate methods in practical use for integrating some forms of ODE's have spurious roots. It should be mentioned that methods with spurious roots are not self starting. For example, if there is one spurious root to a method, all of the coe cients (cm )2 in Eq. 6.36 must be initialized by some starting procedure. The initial vector ~ 0 u does not provide enough data to initialize all of the coe cients. This results because methods which produce spurious roots require data from time level n ; 1 or earlier. For example, the leapfrog method requires ~ n;1 and thus cannot be started using u only ~ n. u Presumably (i.e., if one starts the method properly) the spurious coe cients are all initialized with very small magnitudes, and presumably the magnitudes of the spurious roots themselves are all less than one (see Chapter 7). Then the presence of spurious roots does not contaminate the answer. That is, after some nite time the amplitude of the error associated with the spurious roots is even smaller then when it was initialized. Thus while spurious roots must be considered in stability analysis, they play virtually no role in accuracy analysis.
6.5.4 OneRoot TimeMarching Methods
There are a number of timemarching methods that produce only one root for each root. We refer to them as oneroot methods. They are also called onestep methods.
6.6. ACCURACY MEASURES OF TIMEMARCHING METHODS
97
1 The method represents the explicit Euler ( = 0), the trapezoidal ( = 2 ), and the implicit Euler methods ( = 1), respectively. Its ; relation is ) = 1 +1(1 ; h h ; It is instructive to compare the exact solution to a set of ODE's (with a complete eigensystem) having timeinvariant forcing terms with the exact solution to the O E's for oneroot methods. These are ~ (t) = c1 e 1 h n ~ 1 + + cm e m h n ~ m + + cM e M h n ~ M + A;1~ u x x x f ~ n = c1( 1 )n ~ 1 + + cm( m )n ~ m + + cM ( M )n ~ M + A;1~ u x x x f (6.37) respectively. Notice that when t and n = 0, these equations are identical, so that all the constants, vectors, and matrices are identical except the ~ and the terms inside u the parentheses on the right hand sides. The only error made by introducing the time marching is the error that makes in approximating e h.
They have the signi cant advantage of being selfstarting which carries with it the very useful property that the timestep interval can be changed at will throughout the marching process. Three oneroot methods were analyzed in Section 6.4.2. A popular method having this property, the socalled method, is given by the formula h i un+1 = un + h (1 ; )u0n + u0n+1
6.6 Accuracy Measures of TimeMarching Methods
6.6.1 Local and Global Error Measures
There are two broad categories of errors that can be used to derive and evaluate timemarching methods. One is the error made in each time step. This is a local error such as that found from a Taylor table analysis, see Section 3.4. It is usually used as the basis for establishing the order of a method. The other is the error determined at the end of a given event which has covered a speci c interval of time composed of many time steps. This is a global error. It is useful for comparing methods, as we shall see in Chapter 8. It is quite common to judge a timemarching method on the basis of results found from a Taylor table. However, a Taylor series analysis is a very limited tool for nding the more subtle properties of a numerical timemarching method. For example, it is of no use in:
98
CHAPTER 6. TIMEMARCHING METHODS FOR ODE'S
nding spurious roots. evaluating numerical stability and separating the errors in phase and amplitude. analyzing the particular solution of predictorcorrector combinations. nding the global error.
The latter three of these are of concern to us here, and to study them we make use of the material developed in the previous sections of this chapter. Our error measures are based on the di erence between the exact solution to the representative ODE, given by
t u(t) = ce t + ae ;
(6.38)
and the solution to the representative O E's, including only the contribution from the principal root, which can be written as
h un = c1 ( 1)n + ae hn Q(e h ) P (e )
(6.39)
j j
6.6.2 Local Accuracy of the Transient Solution (er
Transient error
er! )
The particular choice of an error measure, either local or global, is to some extent arbitrary. However, a necessary condition for the choice should be that the measure can be used consistently for all methods. In the discussion of the  relation we saw that all timemarching methods produce a principal root for every root that exists in a set of linear ODE's. Therefore, a very natural local error measure for the transient solution is the value of the di erence between solutions based on these two roots. We designate this by er and make the following de nition
er
e h; 1
The leading error term can be found by expanding in a Taylor series and choosing the rst nonvanishing term. This is similar to the error found from a Taylor table. The order of the method is the last power of h matched exactly.
6.6. ACCURACY MEASURES OF TIMEMARCHING METHODS
99
Amplitude and Phase Error
Suppose a eigenvalue is imaginary. Such can indeed be the case when we study the equations governing periodic convection which produces harmonic motion. For such cases it is more meaningful to express the error in terms of amplitude and phase. Let = i! where ! is a real number representing a frequency. Then the numerical method must produce a principal root that is complex and expressible in the form (6.40) 1 = r + i i ei!h From this it follows that the local error in amplitude is measured by the deviation of j 1 j from unity, that is
q era = 1 ; j 1 j = 1 ; ( 1)2 + ( 1 )2 i r er! !h ; tan;1 ( 1 )i=( 1 )r )]
and the local error in phase can be de ned as (6.41) Amplitude and phase errors are important measures of the suitability of timemarching methods for convection and wave propagation phenomena. The approach to error analysis described in Section 3.5 can be extended to the combination of a spatial discretization and a timemarching method applied to the linear convection equation. The principal root, 1( h), is found using = ;ia , where is the modi ed wavenumber of the spatial discretization. Introducing the Courant number, Cn = ah= x, we have h = ;iCn x. Thus one can obtain values of the principal root over the range 0 x for a given value of the Courant number. The above expression for er! can be normalized to give the error in the phase speed, as follows ;1 (6.42) er = er! = 1 + tan ( 1)i =( 1)r )]
p
where ! = ;a . A positive value of erp corresponds to phase lag (the numerical phase speed is too small), while a negative value corresponds to phase lead (the numerical phase speed is too large).
!h
Cn x
6.6.3 Local Accuracy of the Particular Solution (er )
The numerical error in the particular solution is found by comparing the particular solution of the ODE with that for the O E. We have found these to be given by 1 P:S:(ODE) = ae t ( ; )
100 and
CHAPTER 6. TIMEMARCHING METHODS FOR ODE'S
h P:S:(O E) = ae t Q(e h) P (e )
respectively. For a measure of the local error in the particular solution we introduce the de nition
er
P:S: h P:S:(O E) ; 1 (ODE )
(
)
(6.43)
The multiplication by h converts the error from a global measure to a local one, so that the order of er and er are consistent. In order to determine the leading error term, Eq. 6.43 can be written in terms of the characteristic and particular polynomials as
er = co ;
where
n
( ; )Q e
h
;P e h
o
(6.44)
co = h!0 h( ;h ) lim P e
The value of co is a methoddependent constant that is often equal to one. If the forcing function is independent of time, is equal to zero, and for this case, many numerical methods generate an er that is also zero. The algebra involved in nding the order of er can be quite tedious. However, this order is quite important in determining the true order of a timemarching method by the process that has been outlined. An illustration of this is given in the section on RungeKutta methods.
6.6.4 Time Accuracy For Nonlinear Applications
In practice, timemarching methods are usually applied to nonlinear ODE's, and it is necessary that the advertised order of accuracy be valid for the nonlinear cases as well as for the linear ones. A necessary condition for this to occur is that the local accuracies of both the transient and the particular solutions be of the same order. More precisely, a timemarching method is said to be of order k if
er = c1 ( h)k1+1 er = c2 ( h)k2+1 where k = smallest of(k1 k2)
(6.45) (6.46) (6.47)
6.6. ACCURACY MEASURES OF TIMEMARCHING METHODS
101
The reader should be aware that this is not su cient. For example, to derive all of the necessary conditions for the fourthorder RungeKutta method presented later in this chapter the derivation must be performed for a nonlinear ODE. However, the analysis based on a linear nonhomogeneous ODE produces the appropriate conditions for the majority of timemarching methods used in CFD.
6.6.5 Global Accuracy
In contrast to the local error measures which have just been discussed, we can also de ne global error measures. These are useful when we come to the evaluation of timemarching methods for speci c purposes. This subject is covered in Chapter 8 after our introduction to stability in Chapter 7. Suppose we wish to compute some timeaccurate phenomenon over a xed interval of time using a constant time step. We refer to such a computation as an \event". Let T be the xed time of the event and h be the chosen step size. Then the required number of time steps, is N, given by the relation
T = Nh
Global error in the transient
A natural extension of er to cover the error in an entire event is given by
Er
e T ; ( 1 ( h))N
(6.48)
Global error in amplitude and phase
If the event is periodic, we are more concerned with the global error in amplitude and phase. These are given by
Era = 1 ;
and
q "
( 1 )2 + ( 1)2 i r
N
(6.49)
( ) Er! N !h ; tan;1 ( 1) i 1r ;1 ( ) =( ) ] = !T ; N tan 1i 1r
!#
(6.50)
102
CHAPTER 6. TIMEMARCHING METHODS FOR ODE'S
Global error in the particular solution
Finally, the global error in the particular solution follows naturally by comparing the solutions to the ODE and the O E. It can be measured by
Er
( ; )
Qeh ;1 P eh
6.7 Linear Multistep Methods
In the previous sections, we have developed the framework of error analysis for time advance methods and have randomly introduced a few methods without addressing motivational, developmental or design issues. In the subsequent sections, we introduce classes of methods along with their associated error analysis. We shall not spend much time on development or design of these methods, since most of them have historic origins from a wide variety of disciplines and applications. The Linear Multistep Methods (LMM's) are probably the most natural extension to time marching of the space di erencing schemes introduced in Chapter 3 and can be analyzed for accuracy or designed using the Taylor table approach of Section 3.4.
6.7.1 The General Formulation
When applied to the nonlinear ODE
all linear multistep methods can be expressed in the general form
1 X
k=1;K k un+k = h
du = u0 = F(u t) dt
1 X
k=1;K
k Fn+k
(6.51)
where the notation for F is de ned in Section 6.1. The methods are said to be linear because the 's and 's are independent of u and n, and they are said to be K step because K timelevels of data are required to marching the solution one timestep, h. They are explicit if 1 = 0 and implicit otherwise. When Eq. 6.51 is applied to the representative equation, Eq. 6.8, and the result is expressed in operational form, one nds
0 1 @ X
k=1;K
1 0 1 X k k E Aun = h@
k=1;K
1 k hn k E A( un + ae )
(6.52)
6.7. LINEAR MULTISTEP METHODS
103
We recall from Section 6.5.2 that a timemarching method when applied to the representative equation must provide a root, labeled 1, that approximates e h through the order of the method. The condition referred to as consistency simply means that ! 1 as h ! 0, and it is certainly a necessary condition for the accuracy of any time marching method. We can also agree that, to be of any value in time accuracy, a method should at least be rstorder accurate, that is ! (1 + h) as h ! 0. One can show that these conditions are met by any method represented by Eq. 6.51 if X X X k = 0 and k = (K + k ; 1) k Since both sides of Eq. 6.51 can be multiplied by an arbitrary constant, these methods are often \normalized" by requiring X k=1 Under this condition co = 1 in Eq. 6.44.
k k k k
6.7.2 Examples
There are many special explicit and implicit forms of linear multistep methods. Two wellknown families of them, referred to as AdamsBashforth (explicit) and AdamsMoulton (implicit), can be designed using the Taylor table approach of Section 3.4. The AdamsMoulton family is obtained from Eq. 6.51 with k = ;1 ;2 (6.53) 1=1 0 = ;1 k=0 The AdamsBashforth family has the same 's with the additional constraint that 1 = 0. The threestep AdamsMoulton method can be written in the following form un+1 = un + h( 1u0n+1 + 0 u0n + ;1 u0n;1 + ;2u0n;2) (6.54) A Taylor table for Eq. 6.54 can be generated as 1 1 1 un+1 1 1 2 6 24 ;un ;1 1 ;h 1 u0n+1 ;1 ;1 ; 12 ; 11 6 ;h 0 u0n ;0 1 1 ;h ;1 u0n;1 ; ;1 ; ;1 2 ;1 6 ;1 1 ;h ;2 u0n;2 ;(;2)0 ;2 ;(;2)1 ;2 ;(;2)2 ;2 2 ;(;2)3 ;2 1 6 Table 6.1. Taylor table for the AdamsMoulton threestep linear multistep method.
un
h u0n
h2 u00 n
h3 u000 n
h4 u0000 n
104
CHAPTER 6. TIMEMARCHING METHODS FOR ODE'S
This leads to the linear system 2 1 1 1 1 32 3 213 1 6 2 0 ;2 ;4 7 6 0 7 6 1 7 6 7 6 7 6 3 0 3 12 7 6 76 (6.55) 4 5 4 ;1 7 = 6 1 7 5 4 5 4 0 ;4 ;32 1 ;2 to solve for the 's, resulting in (6.56) 1 = 9=24 0 = 19=24 ;1 = ;5=24 ;2 = 1=24 which produces a method which is fourthorder accurate.4 With 1 = 0 one obtains 2 32 3 2 3 1 1 1 0 7 617 6 0 ;2 ;4 7 6 ;1 5 = 4 1 5 (6.57) 4 54 0 3 12 1 ;2 giving (6.58) 0 = 23=12 ;1 = ;16=12 ;2 = 5=12 This is the thirdorder AdamsBashforth method. A list of simple methods, some of which are very common in CFD applications, is given below together with identifying names that are sometimes associated with them. In the following material AB(n) and AM(n) are used as abbreviations for the (n)th order AdamsBashforth and (n)th order AdamsMoulton methods. One can verify that the Adams type schemes given below satisfy Eqs. 6.55 and 6.57 up to the order of the method.
Explicit Methods
un+1 un+1 un+1 un+1
= un + hu0n Euler 0 = un;1 + 2hhun Leapfrog 1 h 3u0n ; u0n;1i = un + 2 h 0 i AB2 h = un + 12 23un ; 16u0n;1 + 5u0n;2 AB3
Implicit Methods
un+1 un+1 un+1 un+1
4 Recall from Section 6.5.2 that a kthorder timemarching method has a leading truncation error term which is O(hk+1 ).
= un + hu0n+1 Implicit Euler 1 hhu0n + u0n+1i = un + 2 Trapezoidal (AM2) 1 h4un ; un;1 + 2hu0n+1i =3 2ndorder Backward h h5u0n+1 + 8u0n ; u0n;1i = un + 12 AM3
6.7. LINEAR MULTISTEP METHODS
105
6.7.3 TwoStep Linear Multistep Methods
High resolution CFD problems usually require very large data sets to store the spatial information from which the time derivative is calculated. This limits the interest in multistep methods to about two time levels. The most general twostep linear multistep method (i.e., K=2 in Eq. 6.51), that is at least rstorder accurate, can be written as h i (1 + )un+1 = (1 + 2 )un ; un;1] + h u0n+1 + (1 ; + ')u0n ; 'u0n;1 (6.59) Clearly the methods are explicit if = 0 and implicit otherwise. A list of methods contained in Eq. 6.59 is given in Table 6.2. Notice that the Adams methods have = 0, which corresponds to ;1 = 0 in Eq. 6.51. Methods with = ;1=2, which corresponds to 0 = 0 in Eq. 6.51, are known as Milne methods.
'
1=2 1 3=4 1=3 1=2 5=9 0 0 0 1=3 5=12 1=6 0 1 0 0 0 1=2 0 ;1=2 ;1=2 ;1=6 ;1=2 0 ;5=6 ;1=6 0 ;1=2 0 0 0 0 ;1=4 ;1=3 ;1=2 ;2=9 0 1=2 ;1=3 0 1=12 ;1=6
Method Euler Implicit Euler Trapezoidal or AM2 2nd Order Backward Adams type Lees Type Two{step trapezoidal A{contractive Leapfrog AB2 Most accurate explicit Third{order implicit AM3 Milne
Order 1 1 2 2 2 2 2 2 2 2 3 3 3 4
Table 6.2. Some linear one and twostep methods, see Eq. 6.59. One can show after a little algebra that both er and er are reduced to 0(h3) (i.e., the methods are 2ndorder accurate) if 1 '= ; +2 The class of all 3rdorder methods is determined by imposing the additional constraint =2 ;5 6 1 Finally a unique fourthorder method is found by setting = ;' = ; =3 = 6 .
106
6.8 PredictorCorrector Methods
CHAPTER 6. TIMEMARCHING METHODS FOR ODE'S
There are a wide variety of predictorcorrector schemes created and used for a variety of purposes. Their use in solving ODE's is relatively easy to illustrate and understand. Their use in solving PDE's can be much more subtle and demands concepts5 which have no counterpart in the analysis of ODE's. Predictorcorrector methods constructed to timemarch linear or nonlinear ODE's are composed of sequences of linear multistep methods, each of which is referred to as a family in the solution process. There may be many families in the sequence, and usually the nal family has a higher Taylorseries order of accuracy than the intermediate ones. Their use is motivated by ease of application and increased e ciency, where measures of e ciency are discussed in the next two chapters. A simple onepredictor, onecorrector example is given by
un+ = un + h 0n ~ hu i un+1 = un + h u0n+ + u0n ~
(6.60)
where the parameters and are arbitrary parameters to be determined. One can analyze this sequence by applying it to the representative equation and using the operational techniques outlined in Section 6.4. It is easy to show, following the example leading to Eq. 6.26, that
h P (E ) = E E ; 1 ; ( + ) h ; Q(E ) = E h E + + h]
2 h2 i
(6.61) (6.62)
Considering only local accuracy, one is led, by following the discussion in Section 6.6, to the following observations. For the method to be secondorder accurate both er and er must be O(h3). For this to hold for er , it is obvious from Eq. 6.61 that + =1 =1 2 which provides two equations for three unknowns. The situation for er requires some algebra, but it is not di cult to show using Eq. 6.44 that the same conditions also make it O(h3). One concludes, therefore, that the predictorcorrector sequence
is a secondorder accurate method for any .
un+ = un + hu0n ~ 1 un+1 = un + 2 h 1 u0n+ + 2 ; 1 u0n ~
(6.63)
5 Such as alternating direction, fractionalstep, and hybrid methods.
6.9. RUNGEKUTTA METHODS
107
A classical predictorcorrector sequence is formed by following an AdamsBashforth predictor of any order with an AdamsMoulton corrector having an order one higher. The order of the combination is then equal to the order of the corrector. If the order of the corrector is (k), we refer to these as ABM(k) methods. The AdamsBashforthMoulton sequence for k = 3 is i 1 h un+1 = un + 2 h 3u0n ; u0n;1 ~ i hh u un+1 = un + 12 5~0n+1 + 8u0n ; u0n;1 (6.64) Some simple, speci c, secondorder accurate methods are given below. The Gazdag method, which we discuss in Chapter 8, is 1 hu ~ i un+1 = un + 2 h 3~0n ; u0n;1 ~ h i un+1 = un + 1 h u0n + u0n+1 ~ ~ (6.65) 2 The Burstein method, obtained from Eq. 6.63 with = 1=2 is 1 un+1=2 = un + 2 hu0n ~ un+1 = un + hu0n+1=2 ~ and, nally, MacCormack's method, presented earlier in this chapter, is un+1 = un + hu0n ~ 1 un+1 = 2 un + un+1 + hu0n+1] ~ ~ Note that MacCormack's method can also be written as un+1 = un + hu0n ~ un+1 = un + 1 h u0n + u0n+1] ~ 2 from which it is clear that it is obtained from Eq. 6.63 with = 1.
(6.66)
(6.67)
(6.68)
6.9 RungeKutta Methods
There is a special subset of predictorcorrector methods, referred to as RungeKutta methods,6 that produce just one root for each root such that ( h) corresponds
explicit RungeKutta methods here.
6 Although implicit and multistep RungeKutta methods exist, we will consider only singlestep,
108
CHAPTER 6. TIMEMARCHING METHODS FOR ODE'S
to the Taylor series expansion of e h out through the order of the method and then truncates. Thus for a RungeKutta method of order k (up to 4th order), the principal (and only) root is given by (6.69) = 1 + h + 1 2h2 + + 1 k hk 2 k! It is not particularly di cult to build this property into a method, but, as we pointed out in Section 6.6.4, it is not su cient to guarantee k'th order accuracy for the solution of u0 = F (u t) or for the representative equation. To ensure k'th order accuracy, the method must further satisfy the constraint that er = O(hk+1) (6.70) and this is much more di cult. The most widely publicized RungeKutta process is the one that leads to the fourthorder method. We present it below in some detail. It is usually introduced in the form k1 = hF (un tn) k2 = hF (un + k1 tn + h) k3 = hF (un + 1k1 + 1k2 tn + 1 h) k4 = hF (un + 2k1 + 2k2 + 2 k3 tn + 2h) followed by u(tn + h) ; u(tn) = 1 k1 + 2k2 + 3k3 + 4k4 (6.71) However, we prefer to present it using predictorcorrector notation. Thus, a scheme entirely equivalent to 6.71 is b un+ = un + hu0n b un+ 1 = un + 1 hu0n + 1hu0n+ ~ b un+ 2 = un + 2 hu0n + 2hu0n+ + 2hu0n+ 1 ~ 0 + hu0 + hu0 (6.72) un+1 = un + 1hun 2 bn+ 3 ~n+ 1 + 4 hu0n+ 2 Appearing in Eqs. 6.71 and 6.72 are a total of 13 parameters which are to be determined such that the method is fourthorder according to the requirements in Eqs. 6.69 and 6.70. First of all, the choices for the time samplings, , 1, and 2 , are not arbitrary. They must satisfy the relations = 1 = 1+ 1 (6.73) 2 = 2+ 2+ 2
6.9. RUNGEKUTTA METHODS
109
The algebra involved in nding algebraic equations for the remaining 10 parameters is not trivial, but the equations follow directly from nding P (E ) and Q(E ) and then satisfying the conditions in Eqs. 6.69 and 6.70. Using Eq. 6.73 to eliminate the 's we nd from Eq. 6.69 the four conditions
1+ 2 + 3 1+
= 1 2+ 3+ 4 = 1=2 3 1+ 4 2 ( 2 + 1 2 ) = 1=6 4 = 1=24 4 12
(1) (2) (3) (4)
(6.74)
These four relations guarantee that the ve terms in exactly match the rst 5 terms in the expansion of e h . To satisfy the condition that er = O(k5), we have to ful ll four more conditions 2 2 = 1=3 (3) 2 2+ 3 1+ 4 2 3+ 3 3+ 4 3 = 1=4 (4) 2 1 2 (6.75) 2 1 + 4 ( 2 2 + 2 2 ) = 1=12 (4) 3 1 (4) 3 1 1 + 4 2 ( 2 + 1 2 ) = 1=8 The number in parentheses at the end of each equation indicates the order that is the basis for the equation. Thus if the rst 3 equations in 6.74 and the rst equation in 6.75 are all satis ed, the resulting method would be thirdorder accurate. As discussed in Section 6.6.4, the fourth condition in Eq. 6.75 cannot be derived using the methodology presented here, which is based on a linear nonhomogenous representative ODE. A more general derivation based on a nonlinear ODE can be found in several books.7 There are eight equations in 6.74 and 6.75 which must be satis ed by the 10 unknowns. Since the equations are overdetermined, two parameters can be set arbitrarily. Several choices for the parameters have been proposed, but the most popular one is due to Runge. It results in the \standard" fourthorder RungeKutta method expressed in predictorcorrector form as 1 b un+1=2 = un + 2 hu0n b un+1=2 = un + 1 hu0n+1=2 ~ 2 un+1 = un + hu0n+1=2 ~ h i b un+1 = un + 1 h u0n + 2 u0n+1=2 + u0n+1=2 + u0n+1 ~ (6.76) 6
conditions for RungeKutta methods of up to third order.
7 The present approach based on a linear inhomogeneous equation provides all of the necessary
110
CHAPTER 6. TIMEMARCHING METHODS FOR ODE'S 9 > > = > >
Notice that this represents the simple sequence of conventional linear multistep methods referred to, respectively, as Euler Predictor Euler Corrector Leapfrog Predictor Milne Corrector
RK 4
One can easily show that both the Burstein and the MacCormack methods given by Eqs. 6.66 and 6.67 are secondorder RungeKutta methods, and thirdorder methods can be derived from Eqs. 6.72 by setting 4 = 0 and satisfying only Eqs. 6.74 and the rst equation in 6.75. It is clear that for orders one through four, RK methods of order k require k evaluations of the derivative function to advance the solution one time step. We shall discuss the consequences of this in Chapter 8. Higherorder RungeKutta methods can be developed, but they require more derivative evaluations than their order. For example, a fthorder method requires six evaluations to advance the solution one step. In any event, storage requirements reduce the usefulness of RungeKutta methods of order higher than four for CFD applications.
6.10 Implementation of Implicit Methods
We have presented a wide variety of timemarching methods and shown how to derive their ; relations. In the next chapter, we will see that these methods can have widely di erent properties with respect to stability. This leads to various tradeo s which must be considered in selecting a method for a speci c application. Our presentation of the timemarching methods in the context of a linear scalar equation obscures some of the issues involved in implementing an implicit method for systems of equations and nonlinear equations. These are covered in this Section.
6.10.1 Application to Systems of Equations
Consider rst the numerical solution of our representative ODE
u0 = u + ae t
(6.77)
using the implicit Euler method. Following the steps outlined in Section 6.2, we obtained (1 ; h)un+1 ; un = he
h
ae hn
(6.78)
6.10. IMPLEMENTATION OF IMPLICIT METHODS
Solving for un+1 gives 1 un+1 = 1 ; h (un + he h ae hn )
111 (6.79)
This calculation does not seem particularly onerous in comparison with the application of an explicit method to this ODE, requiring only an additional division. Now let us apply the implicit Euler method to our generic system of equations given by
~ 0 = A~ ; f (t) u u ~ ~ (I ; hA)~ n+1 ; ~ n = ;hf (t + h) u u
or
(6.80)
~ where ~ and f are vectors and we still assume that A is not a function of ~ or t. Now u u the equivalent to Eq. 6.78 is
(6.81) (6.82)
~ ~ n+1 = (I ; hA);1 ~ n ; hf (t + h)] u u
The inverse is not actually performed, but rather we solve Eq. 6.81 as a linear system of equations. For our onedimensional examples, the system of equations which must be solved is tridiagonal (e.g., for biconvection, A = ;aBp (;1 0 1)=2 x), and hence its solution is inexpensive, but in multidimensions the bandwidth can be very large. In general, the cost per time step of an implicit method is larger than that of an explicit method. The primary area of application of implicit methods is in the solution of sti ODE's, as we shall see in Chapter 8.
6.10.2 Application to Nonlinear Equations
Now consider the general nonlinear scalar ODE given by Application of the implicit Euler method gives
du = F (u t) dt
(6.83) (6.84)
un+1 = un + hF (un+1 tn+1)
This is a nonlinear di erence equation. As an example, consider the nonlinear ODE du + 1 u2 = 0 (6.85) dt 2
112
CHAPTER 6. TIMEMARCHING METHODS FOR ODE'S
solved using implicit Euler time marching, which gives 1 un+1 + h 2 u2 +1 = un (6.86) n which requires a nontrivial method to solve for un+1. There are several di erent approaches one can take to solving this nonlinear di erence equation. An iterative method, such as Newton's method (see below), can be used. In practice, the \initial guess" for this nonlinear problem can be quite close to the solution, since the \initial guess" is simply the solution at the previous time step, which implies that a linearization approach may be quite successful. Such an approach is described in the next Section.
6.10.3 Local Linearization for Scalar Equations
General Development
Let us start the process of local linearization by considering Eq. 6.83. In order to implement the linearization, we expand F (u t) about some reference point in time. Designate the reference value by tn and the corresponding value of the dependent variable by un. A Taylor series expansion about these reference quantities gives
! ! @F (u ; u ) + @F (t ; t ) F (u t) = F (un tn) + @u n n @t n n ! ! 1 @ 2 F (u ; u )2 + @ 2 F (u ; u )(t ; t ) + 2 @u2 n n n @u@t n !n 1 2 + 2 @ F (t ; tn)2 + @t2
n
(6.87)
On the other hand, the expansion of u(t) in terms of the independent variable t is
! ! @u + 1 (t ; t )2 @ 2 u + u(t) = un + (t ; tn) @t n @t2 n n 2
(6.88)
If t is within h of tn , both (t ; tn)k and (u ; un)k are O(hk ), and Eq. 6.87 can be written
! ! @F (u ; u ) + @F (t ; t ) + O(h2) F (u t) = Fn + @u n n @t n n
(6.89)
6.10. IMPLEMENTATION OF IMPLICIT METHODS
113
Notice that this is an expansion of the derivative of the function. Thus, relative to the order of expansion of the function, it represents a secondorderaccurate, locallylinear approximation to F (u t) that is valid in the vicinity of the reference station tn and the corresponding un = u(tn). With this we obtain the locally (in the neighborhood of tn) timelinear representation of Eq. 6.83, namely
! ! ! ! du = @F u + F ; @F u + @F (t ; t ) + O(h2) n n dt @u n @u n n @t n
(6.90)
Implementation of the Trapezoidal Method
As an example of how such an expansion can be used, consider the mechanics of applying the trapezoidal method for the time integration of Eq. 6.83. The trapezoidal method is given by 1 un+1 = un + 2 h Fn+1 + Fn] + hO(h2) (6.91)
where we write hO(h2) to emphasize that the method is second order accurate. Using Eq. 6.89 to evaluate Fn+1 = F (un+1 tn+1), one nds
un+1 = un + 1 h Fn + @F (un+1 ; un) + h @F + O(h2) + Fn 2 @u n @t n 2) +hO(h (6.92) Note that the O(h2) term within the brackets (which is due to the local linearization) is multiplied by h and therefore is the same order as the hO(h2) error from the Trapezoidal Method. The use of local time linearization updated at the end of each time step, and the trapezoidal time march, combine to make a secondorderaccurate numerical integration process. There are, of course, other secondorder implicit timemarching methods that can be used. The important point to be made here is that local linearization updated at each time step has not reduced the order of accuracy of a secondorder timemarching process. A very useful reordering of the terms in Eq. 6.92 results in the expression
1 un = hFn + 1 h2 @F (6.93) 1 ; 2 h @F @u n 2 @t n which is now in the delta form which will be formally introduced in Section 12.6. In many uid mechanic applications the nonlinear function F is not an explicit function
"
!
!
#
"
!#
!
114
CHAPTER 6. TIMEMARCHING METHODS FOR ODE'S
of t. In such cases the partial derivative of F (u) with respect to t is zero and Eq. 6.93 simpli es to the secondorder accurate expression
"
1 1 ; 2 h @F @u
!#
n
un = hFn
(6.94)
Notice that the RHS is extremely simple. It is the product of h and the RHS of the basic equation evaluated at the previous time step. In this example, the basic equation was the simple scalar equation 6.83, but for our applications, it is generally the spacedi erenced form of the steadystate equation of some uid ow problem. A numerical timemarching procedure using Eq. 6.94 is usually implemented as follows: ~ ~ 1. Solve for the elements of hFn, store them in an array say R, and save ~ n. u 2. Solve for the elements of the matrix multiplying ~ n and store in some approu priate manner making use of sparseness or bandedness of the matrix if possible. Let this storage area be referred to as B . 3. Solve the coupled set of linear equations
B ~n = R u ~ for ~ n. (Very seldom does one nd B ;1 in carrying out this step). u 4. Find ~ n+1 by adding ~ n to ~ n, thus u u u ~ n+1 = ~ n + ~ n u u u The solution for ~ n+1 is generally stored such that it overwrites the value of ~ n u u
and the process is repeated.
Implementation of the Implicit Euler Method
We have seen that the rstorder implicit Euler method can be written
un+1 = un + hFn+1 "
1 ; h @F @u
(6.95)
if we introduce Eq. 6.90 into this method, rearrange terms, and remove the explicit dependence on time, we arrive at the form
!#
n
un = hFn
(6.96)
6.10. IMPLEMENTATION OF IMPLICIT METHODS
115
We see that the only di erence between the implementation of the trapezoidal method 1 and the implicit Euler method is the factor of 2 in the brackets of the left side of Eqs. 6.94 and 6.96. Omission of this factor degrades the method in time accuracy by one order of h. We shall see later that this method is an excellent choice for steady problems.
Newton's Method
Consider the limit h ! 1 of Eq. 6.96 obtained by dividing both sides by h and setting 1=h = 0. There results or
; @F @u
!
n
un = Fn @F @u ! #;1
n
(6.97)
un+1 = un ;
"
Fn
(6.98)
This is the wellknown Newton method for nding the roots of a nonlinear equation F (u) = 0. The fact that it has quadratic convergence is veri ed by a glance at Eqs. 6.87 and 6.88 (remember the dependence on t has been eliminated for this case). By quadratic convergence, we mean that the error after a given iteration is proportional to the square of the error at the previous iteration, where the error is the di erence between the current solution and the converged solution. Quadratic convergence is thus a very powerful property. Use of a nite value of h in Eq. 6.96 leads to linear convergence, i.e., the error at a given iteration is some multiple of the error at the previous iteration. The reader should ponder the meaning of letting h ! 1 for the trapezoidal method, given by Eq. 6.94.
6.10.4 Local Linearization for Coupled Sets of Nonlinear Equations
In order to present this concept, let us consider an example involving some simple boundarylayer equations. We choose the FalknerSkan equations from classical boundarylayer theory. Our task is to apply the implicit trapezoidal method to the equations
d3f + f d2f + dt3 dt2
0 !21 @1 ; df A = 0 dt
(6.99)
Here f represents a dimensionless stream function, and is a scaling factor.
116
CHAPTER 6. TIMEMARCHING METHODS FOR ODE'S u1 = d f dt2
2
First of all we reduce Eq. 6.99 to a set of rstorder nonlinear equations by the transformations This gives the coupled set of three nonlinear equations u01 = F1 = ;u1 u3 ; 1 ; u2 2 0 = F =u u2 2 1 0 = F =u u3 3 2 and these can be represented in vector notation as
u2 = df dt
u3 = f
(6.100)
(6.101)
Now we seek to make the same local expansion that derived Eq. 6.90, except that this time we are faced with a nonlinear vector function, rather than a simple nonlinear scalar function. The required extension requires the evaluation of a matrix, called the Jacobian matrix.8 Let us refer to this matrix as A. It is derived from Eq. 6.102 by the following process
d~ = F (~ ) u ~ u dt
(6.102)
A = (aij ) = @Fi =@uj
For the general case involving a third order matrix this is
(6.103)
2 6 6 6 6 A=6 6 6 6 6 4
~ u The expansion of F (~ ) about some reference state ~ n can be expressed in a way u similar to the scalar expansion given by eq 6.87. Omitting the explicit dependency ~ ~ u on the independent variable t, and de ning F n as F (~ n), one has 9
2.2 vector for the second term, and tensor products for the terms of higher order.
@F1 @u1 @F2 @u1 @F3 @u1
@F1 @u2 @F2 @u2 @F3 @u2
@F1 @u3 @F2 @u3 @F3 @u3
3 7 7 7 7 7 7 7 7 7 5
(6.104)
8 Recall that we derived the Jacobian matrices for the twodimensional Euler equations in Section 9 The Taylor series expansion of a vector contains a vector for the rst term, a matrix times a
6.11. PROBLEMS
117
~ u ~ F (~ ) = F n + An ~ ; ~ n + O(h2) u u (6.105) where t ; tn and the argument for O(h2) is the same as in the derivation of Eq. 6.88.
Using this we can write the local linearization of Eq. 6.102 as
\constant" which is a locallylinear, secondorderaccurate approximation to a set of coupled nonlinear ordinary di erential equations that is valid for t tn + h. Any rst or secondorder timemarching method, explicit or implicit, could be used to integrate the equations without loss in accuracy with respect to order. The number of times, and the manner in which, the terms in the Jacobian matrix are updated as the solution proceeds depends, of course, on the nature of the problem. Returning to our simple boundarylayer example, which is given by Eq. 6.101, we nd the Jacobian matrix to be 2 ; u3 2 u2 ;u1 3 0 0 7 A=6 1 (6.107) 4 5 0 1 0 The student should be able to derive results for this example that are equivalent to those given for the scalar case in Eq. 6.93. Thus for the FalknerSkan equations the trapezoidal method results in 2 3 1 + h (u3)n ; h(u2)n h (u1)n 72 ( u1)n 3 2 ; (u1u3)n ; (1 ; u2)n 3 2 2 6 2 7 6 ;h (u1)n 74 6 5 4 5 1 0 76 ( u2)n 7 = h 6 2 4 5 ( u) (u2)n h 3n 0 ;2 1 We nd ~ n+1 from ~ n + ~ n, and the solution is now advanced one step. Reevaluate u u u the elements using ~ n+1 and continue. Without any iterating within a step advance, u the solution will be secondorderaccurate in time.
d~ = A ~ + F ; A ~ +O(h2) u ~ n n un nu dt  {z }
(6.106)
6.11 Problems
1. Find an expression for the nth term in the Fibonacci series, which is given by 1 1 2 3 5 8 : : : Note that the series can be expressed as the solution to a di erence equation of the form un+1 = un + un;1. What is u25 ? (Let the rst term given above be u0.)
118
CHAPTER 6. TIMEMARCHING METHODS FOR ODE'S
1 2. The trapezoidal method un+1 = un + 2 h(u0n+1 + u0n) is used to solve the representative ODE. (a) What is the resulting O E? (b) What is its exact solution? (c) How does the exact steadystate solution of the O E compare with the exact steadystate solution of the ODE if = 0? 3. The 2ndorder backward method is given by i 1h un+1 = 3 4un ; un;1 + 2hu0n+1 (a) Write the O E for the representative equation. Identify the polynomials P (E ) and Q(E ). (b) Derive the  relation. Solve for the roots and identify them as principal or spurious. (c) Find er and the rst two nonvanishing terms in a Taylor series expansion of the spurious root. (d) Perform a root trace relative to the unit circle for both di usion and convection. 4. Consider the timemarching scheme given by un+1 = un;1 + 23h (u0n+1 + u0n + u0n;1) (a) Write the O E for the representative equation. Identify the polynomials P (E ) and Q(E ). (b) Derive the ; relation. (c) Find er . 5. Find the di erence equation which results from applying the Gazdag predictorcorrector method (Eq. 6.65) to the representative equation. Find the  relation. 6. Consider the following timemarching method:
un+1=3 = un + hu0n=3 ~ un+1=2 = un + hu0n+1=3 =2 ~ un+1 = un + hu0n+1=2
6.11. PROBLEMS
119
Find the di erence equation which results from applying this method to the representative equation. Find the  relation. Find the solution to the di erence equation, including the homogeneous and particular solutions. Find er and er . What order is the homogeneous solution? What order is the particular solution? Find the particular solution if the forcing term is xed. 7. Write a computer program to solve the onedimensional linear convection equation with periodic boundary conditions and a = 1 on the domain 0 x 1. Use 2ndorder centered di erences in space and a grid of 50 points. For the initial condition, use
u(x 0) = e;0:5 (x;0:5)= ]2
with = 0:08. Use the explicit Euler, 2ndorder AdamsBashforth (AB2), implicit Euler, trapezoidal, and 4thorder RungeKutta methods. For the explicit Euler and AB2 methods, use a Courant number, ah= x, of 0.1 for the other methods, use a Courant number of unity. Plot the solutions obtained at t = 1 compared to the exact solution (which is identical to the initial condition). 8. Repeat problem 7 using 4thorder (noncompact) di erences in space. Use only 4thorder RungeKutta time marching at a Courant number of unity. Show solutions at t = 1 and t = 10 compared to the exact solution. 9. Using the computer program written for problem 7, compute the solution at t = 1 using 2ndorder centered di erences in space coupled with the 4thorder RungeKutta method for grids of 100, 200, and 400 nodes. On a loglog scale, plot the error given by
v uX u M (uj ; uexact)2 u j t M
j =1
where M is the number of grid nodes and uexact is the exact solution. Find the global order of accuracy from the plot. 10. Using the computer program written for problem 8, repeat problem 9 using 4thorder (noncompact) di erences in space. 11. Write a computer program to solve the onedimensional linear convection equation with in owout ow boundary conditions and a = 1 on the domain 0 x 1. Let u(0 t) = sin !t with ! = 10 . Run until a periodic steady state is reached which is independent of the initial condition and plot your solution
120
CHAPTER 6. TIMEMARCHING METHODS FOR ODE'S
compared with the exact solution. Use 2ndorder centered di erences in space with a 1storder backward di erence at the out ow boundary (as in Eq. 3.69) together with 4thorder RungeKutta time marching. Use grids with 100, 200, and 400 nodes and plot the error vs. the number of grid nodes, as described in problem 9. Find the global order of accuracy. 12. Repeat problem 11 using 4thorder (noncompact) centered di erences. Use a thirdorder forwardbiased operator at the in ow boundary (as in Eq. 3.67). At the last grid node, derive and use a 3rdorder backward operator (using nodes j ; 3, j ; 2, j ; 1, and j ) and at the second last node, use a 3rdorder backwardbiased operator (using nodes j ; 2, j ; 1, j , and j +1 see problem 1 in Chapter 3). 13. Using the approach described in Section 6.6.2, nd the phase speed error, erp, and the amplitude error, era , for the combination of secondorder centered differences and 1st, 2nd, 3rd, and 4thorder RungeKutta timemarching at a Courant number of unity. Also plot the phase speed error obtained using exact integration in time, i.e., that obtained using the spatial discretization alone. Note that the required roots for the various RungeKutta methods can be deduced from Eq. 6.69, without actually deriving the methods. Explain your results.
Í @Ì° 3448bDG£P9@h8ImP04@37PwXF44@aV0@PTVFjHY@E@o4@c1PIeFXG4EDC4A TAI 0 f 97 G 9W 9gg7 57 T 9727 A27 T2 F2 0 `YHGPFP9c@d7vGh3Ae0Sº76FjVIeT@xc8bs4S· ¯ sy ef1tI80DGe035VwPT4¹AXYP9ctpq4} 1X° ¯ syi@db4VpËVf1tIR FW T G w T7 0 G A27 ¥ 5 R F 9 R 0b { ¥ T7 w T 5 806GV035VwPT4¹AHh3fe0m80SP743Ahef6580P2u q@xP2q3280VGXGVF5 u txSz7XY39e0S87SE3gV0x3A80z7e0e5mV0@U4@PI8Fz7PÊeF6GV035ewu XT4A F F T G F 0 G F 27 F 0W 2 w G 57 T G G 57 T2 G F 60@Eeg4@wXf8Fwc8bdbXw4b X2448bDG HY41mSE8bheIXT@ee0@m~36TP04Sº7eFS@624432l3gcA 42 w7 T A27 0 G 0 G F F TAI 0 f 92I G7 G 97 GI É 9W 57q G 0A 0 0 0 ( wpA G F YW ¥ RI 5 R 0 G F cP23Fc4jVI85P2v} ¯ lyc@ÀVf1tIy48bP2d32e0eGHGeF5 u tÇzY7 u @w@z7È43ge0VGyV01Ç1~I u cP2Xq1@P2F 0WT 97 2 wT T 5I f5I G ( F 9Iq TAI 0 Æ 44eb6Go~ ® Â ½ÅÄ§ ¬ ¨ ² § ¶ 4@l EÃaÀ¨ q8I}4@37P04eb6Gh~65P04ymXG6FPwXF4jDid48bE@X04h@cDi0 A27 ªÂ Á ¿ ¾ G TA 9 AI 0 f 9p 9g 0 0 G 9g f7 5I¸ 0227 f 5 5 Gb F T F R 7 F 5 R 0 0( 0 G G AI 0 f ( F 57 1Eu63544Sq©VA@7 u 1tIeR1X(@87e5VGHYe9P5S½7x#P2¼ef1tI»aiPweF65432@i48bpeI#j4eb6Ghp4X24dbew@f DhPfeF4P2HY44@ 3A4X2@d7c@~XTxef1tIU6580VGSº7d48b4£j4eb6Gh#3ge0VG¹TD41~1E}1HY39@Xq8FDG3wleg6Te05 0 G 7 ( F 9gg7 YW 0 F GWI F 5 R G 9 0 GAI 0 f 02I 7 5I¸ 0 0 ² § ¶ E§² y¨ tz³E§² ® @· ¯ « µ ´ A2 4@7 « ª ¨ ±®X° ¯ V® § ¬ § y©y¦§ ¦ YW G 5 5 57 T2 G7p 0 0b { TAI 0 #6A8032V03TV043gV0e0@~41XI8FS46¥uV03T4}y64jc8b6Ghf ( F 57 0 GRI 2 G 9gg7 YW 0 G G 5 5 0 G f R G 0 0( 0 4P2cdbew@`f Xf8F£Si1PIeFSz7XwPF44SqrT}@XqeFSd732V03TV043gV0m4eb¤1eIt5h3Ae0S87654321T}vc8bG 32c8bh4@B£db@eIc4@xe06Ge065ewXT4PAh3fV0x48b§uT}BxPwVF@Ex4eb¡1eIt53Ae0S8765432@eT}@I 0 G A27 w7 5gg7 F F T 0 G YW ¢ T7W 0 G f R G 0 0( R T 5 R 0 G G 5 5 0 G 0g 0 0A A27 Gg7b T7 G7 f w T Vf1tIw@Xq8Fd732V03Te0c3ge0a4eb 6At@PI39@6q4io4@h~eT35e0c@E @l@D@3T60PwVF85Sdb4Vp8IG 65t06RV0P9X9F u §ij41PIeF6G3wp60e0 4@s§eb@l@w324c32l3g44X2e0@a6T30PwVF85Sd32P06Fh63Iw 5 0 2 0 T A27 GIW RI G 0A 0 0A F 57 G7 f G w 0 0 G F F u A 2 7 u G 0 A 0 0 0 G F w 2 w T 7 W 5 p 5 F c8bdbPw4b u P2BTgpc@BT}324462t6g4`A Xf8Fi8bHGF u PThe26530416I}PwVF@Ey4vrse1I u PTc8bG X23f4ebq1lP9390 @4w4v0 c4j3eTc3g60413Iu4P2eFVG3TV035e0X2q4S@87VG1lhPf~e0@V03T4}zP21lg F 0 G 2I A GI2 IA GpW G w2 w ( G F A27 G2 5Ig F 57 0b { G FI 3AjcdS87uc8beIo6A806GS8785VGSzY7 u @mPT 1PIeFcPp1VImeTXGySxef14w@eIEX2@q@4vS48bP24ebXGF u 0iC 0 f T 0 G G w G7 T 97 F 2 G 9 T F RI 5I2 0 f T F7 fIA G7 G F f R F F7 fIA F w 7 F 9W T 0W G F T F T T 7 T G 0 G 5 w G 1eIt5srXR P2Sq1c#P2@87VG6530qP2uX04@87VGelp8IoPA@e7mXTn3fe0VGVYVf35e0ue06T48b#`2m3eT@8I6GS8785VGS7 A27 G FIg 0iC RI T G F T FA RI F FW T F f72YA Fp F G 4S8TP21lk3Ajc@ihef35e0hP2g3Ae0VT43pVwPT46280fG@eXTs`YHGPFX94@d7vGB33TXwSEcPA44X2VG6Te035e0P2F RI 5 0 92I2 0 G RI T w 0 G 0 T2 w ( 0W 2I 0 fI 0bg 5 9 57 @T @374P2XF414Q4ebu@Qe0@37§c8by`2¤3AV0354XAeF416I4P26Fty1432h146244x@P743pXw8FvG@Eg 0 G 2I TA 0 0A A27 9 0 F2p 5I2 9g T 0 02 F FW T RI 2 F2 0A 9 0 0( c8b§@u4462t6g4sc@r@87VT351Hq44a14iX04hPfeFd35c8bXG6F4aPT`YXGPFX94@87VGUSQ@PI8FHG4EDC4B@87654321)' % $ # ! " $ & ©§¥ ¤¢ ¡ ¨ ¦ £
8YR8&©71¨QQ1W10T'7R8VUTRQ©P'7I'H©GED"¨CBA6§9871#'¨6'54¥3%©"¦210)#('&%©"¦!©§¥¤£ B # X ¨ ¨ ¨ # # S 73 F # #F & @ $ # $ # ¨ $ # ¨ ¨¦
36T4@P74§¢DiX04@3Iw 02 9g 9g f A27 ý 9g f w 0 G GIW F TAI 0 f Tp 57q RI FW T 0 G 0 F f7 0 G G F 0q2 4Ss iP04h16Iq4eb§8bSt§P2d4jc8b6Ghd41PIVF@SU@©`YHGPFX94@d7vGq4ebhcP2h@cDieIe32X0432@@3Iw GXF cEvP9X9F u 0 u 4Sw@XY691XqeF6G6wtVg3Te0· ¯ 4@§X° ¯ 6y 1tIR 4¡c@¼E@o3T4Xp@S321X(3F0 A2C A27 0 0 5 A27 T¥ 5 ÿ A27 ÿý T7 0 97q 0 V03T4ebh3A80E1P(VF3T4x0 u 3VT3580j@Edbe4@PI6qV04`2@`YXGPFX94@87VGp4eb4X2Vf35e06G4©eIe32P06FhceppV0@7 0 G G72 0A Gg7 w Tp F 5g FW T 0 G 0 F 0A G G w T 5 ¯ 4S©X° ¯ 3yþP2h3T60PwVF85S#4eb@q6T4Pp@S32@P(3F48b36Te0@374P2eFVG3TV035e0X2kSqþ`2 A27 T¥ F G7 f 0 G RI 0 97q 0 0 0 G T w ( G F Y27 f 3PTjvY@EP2@@qXfF Fi7 57 F(7 0 G (2 97 9 G A G 0 97q 0 ý 0 G I G7 F fI 2 0q2 w FA 0g 5I c8bwc1PI@mP0XFueIp432e0r6T4Pp@S32@P(3F¹0448bT u Eh3Ae0SEP2h14¹Aj1XI8F6G6w116ImXw4jXIeF65tr1E¸ ü 3XTjÇ@z7V05 Fi7 9 0 G7 02 0 G (2 97 9 G A G 0 97q 0 ý 0 G I G7 F fIA 2 Tp FA 5I @XqeFS3(chc8bq4@PI@iX0PF8I462806T4Pp@S32@P(3F¹044ebT u 4#3A80EP2h14©1XIeF4tXøcv1E¸ ü û( 9 R 0 G YW 57 f f T 0W G b(pI 0 G2 5Ig F 57 0b 4cP2F u PIX91tI48bp3A60PÊVF@qh4Vpdl8I141c32U@87VG1lhPfae0@ DY4}{ 33TX08FvG35l1eI4u`YXGXFP94@87VGÛSrXTeFXYSE@d4P2F u XIP91tI48bP2r3ThPfeFi@qd8I3Ae0V535t0DRe0~e0@41XI8FS876G3wlg 0g 5g FW TRI T 9727 ( 9 R 0 G F 0 G Y27 f G 5 57 T2 G 0 0 0b { F¸ T7 0 fI 0bg 0 G ( F F G2 w T 9W 5g 9 TYbg9 w 5g R iqe06T4}i1H°44X(h30V0Äc432h1432c4V03T4ebp4P2cP2@87@3Ih3fP0c1eI4B@37PwVFj4@z7Xw8F6GS874@I T f 0 T 0 97q 0 0 0 G 5 R 0g 0 0W 9p w G7 2 0q2 w FA 0g A27 2 T h4epe56G6wtVgdcPp@S321X(3Fd48b1tI6A80DG3wlc iwlXA416IS4b u 1PIeF6G6w113IXw4PIeF65to4@o1PIVF4tpR ºXR4d1tIwT}U694jhd41è1eIt5§441tIq0 u cÍ4h1XI8F6G6wjX2SP04h@cDid1E#63TP0eFVG65t1VI4g FA 5 R 0AI f 5pI f R A2p R 2 0 F 9g f7 0 5I¸ 0g 5 9@d732h@Ec4tp}P53Fc8b80SPfeFVG6TEzVGSz7P0~S vY13Ic8bxP2s@37u0 u 1vY@E1XI8FS87VGae0@7 µ 4@7 ª ¹R G 0 f7A2 R 0 G G7 0 T 9 G7 5 0 G F2 w 572 G T 5 A2 G7 F fIA G 5 0gg7 G 9 0 GRI 0g 5g 0 G 07 @e0SEX21ceIU@z7l4@6580eGº7a48bÈ@}6TP08FvG35l1eIcdc8b@@f 2 w T2 G 9gg7 9 w 5g T w F F 02I 0 0A 7 G T w 5 F 57 0W 2 @3741XI8FS37PwPFc4@r@z7Xw8FDG@874e0@37hdbPwcb u X2@c1q@Xq8FDG3wt0DR4o8Ie01XI39oXYP9VF@87e5XG4VW@plsSz7w 6f80vGjvYU80DGP04h16IvrSE8bÈ354cP2h3fe0ebXGF u cÍj41PIeF6G6wj~30V0j66f80VGvY4321X(3F}@Xq8FDG3wt0DR44@7 T T 9g f w 7 G7 G 0A F 5 7 2 0 T T T T 0 0 0 0A A2 806GX04h13Iheb@lx1zqEeE8ba3T60PweFe5S#8Is621Hq1plrX9u P9F u 1PIeFS87654PAVF41©SP04Sd7VGwPTú3f80vGVYT 9g f w GIW 0 7b G7 G G7 f G 0 F( 0W 2 G 0 T2I 9W T F T 7 F G YA T 7 0 w7 5gg7 2 G 9 T 7 F G7 0 G RI G7 0 27 0 F( mXRe0Sd7vG`Tc@3780VGh3T4dbSeI4c@1PIeFcPp@eI©ÃdbXw4b Serù@V4ebm@xe0SPf8FvG3T@h3T@Xq1eXGF G7 G 0( G27qA7 0AA7 G2 5Ig F 0 G T7b T T T 0 0 0 G RI T 972 ' 2 0 E8b1@87@Sc@o3A44S@d7vG1thPfa4eb~SEmh3fe0VGVYc321X(3Fde03Tc8bÈ@PTeFHY@Eyc ¯ 1XI8F6G6wj X2§6Ae0VT43pVwPT4vXTvPTc}dc4P26F432V035tXø430SEeg©4SehPf8Fec8b#8bSt©1462t6g4@z71PIeFXG4EDC4A F T FA F Fb { ( w 0 FA w7 T A27 0 G 0 G GIW 2I A 0 0A 2 w 2 F2 0 `YHGPFP9c@d7vGpÇDd®6T41eXwPF1leW65t XYX9@º76F3wleg3T 0 41XI8FSE6¥Ä@P78F32V035tXø4È@P78FVGSEx8I41XI8FS37PwPFc4@7 FW T 7 02I 9I 0gYb 9 0 T2 G7p 0 9 G 0 FA9 57g G T2 G 9gg 5pI F w T 9g T T Fp¥ GI2 F 2 G7 T 0 G T w F G F 0 I ÷ G7 f c1hP2yS304X2eFhP04hPfVF©eI80XGce@4uXTs1XI8FS48pXGVFp4ebpe0@37uPTc8bsX26351Dq0 u ~#tHiVF85SueTXGF R@¡6f80VGvY4321X(3Fh48b8I©6A80SP739e0hl VIPT@§32e0fG@Q@z7x· ¯ sy}S`YHGPFP9c@d7vG4}h ª @I 6f80G I T T 0 0 0 G G G 5 0W 97 RI 2 w ¥ RI FW T 0b { R rjvY432@P(3Fw48bo1 XY39V0P5eF624462t6g4eH° ¯ syr@x`YHGPFP9c@d7vGe4}· ¯ c@eH° ¯ 6yr@m8T´ vG@Eg T T 0 0 0 G 2I G 0 TA 0 0A ¥ RI FW T 0b { A27 T¥ RI 57 Tc13I46211h14x4eb@e1XIjzqE6btxc8bh1§XY69e0P5eF323AV0@EpV0@§`YXGPFX94@87VG~@e4@PI8FHG4EDC4e4v p 0 0(I fIb 0 G RI 5 F 7 0W 0 G 2I G 0 T7W 57 FW T RI T2 F2 0A 5p öçïô õô êç ò ç ð ê í çì êç ë êç éç È Ä©3óÀñ #ï hî©#©sèæ ±åpã ä × âÓÒ×Ü× Ô ÐÓ ÙÚ ßÝ Ü Ù ÙØ ÐÒ× Õ ÔÓÒ Ñ Ð Ï #rÄÇBÇávyràÛ}Þ3ÒÛrÚ4ÇÃ±ÖxBÄvmÎ SÌ°
ý ° ý ° ¨ ´ ´ ý E E E 0W Gb F f 57g Tp 0 0(I fIb 0 l}1X(vG@E416I4321@1cdc8bG R@)ef1tIl@z7Xw4`Yue44syS Sz780vG4P2j@X04S4 i1EBsdPI4x@4eA1@8Ip@PI8FE3¥X014X2eFT I 5 R9 Fg G 7 d ¥ RI A T F 9g f7 0 5I¸ w 9W 27 5I 7 G 2 G7p 0 9( 7 1eIt56T14@Edb©@PI8FE3¥h1XqeFSd762e03TV043gV0i48bB@¼ef@tIe48bh3f80vGVY1Hq8F6G6wt06R4xeI#3Ae0Sº769e05 f R 0(27 w 2 G7p 0 0 G G 5 5 0 G RI 5 R 0 G T T T 0 0A G G 60@Eegc321X(3FQ@s`2@PI8FDG3wV0wE8b©P2xH°4 8IpH°4Ç3ys@#XYX9@º76F3wleg6T§4S Íj4a41XI8F6G6wj w7 T 0 0 27 2 T G7 G F G T¥ 2I 9 0 0 A27 T2 0 X2U8THG4ep6Te0m48bq@I u e74v0 u PTc8bB1EB@30SEeg4621P(6FpP2r41XI8FcXp1eIrP56F48bÛSe0c8pSE48b6G3wlegcP2F F 9 5 0 G 2 5A F G 5I¸ w7 T 0 0 F T2 G 9 T 0 GRI 5 G72 0 G 0 T 0 u 66f80vGjvY432@P(3Fh1Hq8FDG3wt06Rc@±q4wE8bTU@o`YXGPFX94@87VGh48bQ4@d7vGeT654448I654eA1Q`2 T T T 0 0 0 0A 0 7b G7 G RI FW T 0 G A2 0A2p G 0 5I
2 1XIeF35e0XGVF35w `YHGPFP9c@d7vGp32V03544X2oc8bgX2P01VIs4zYº7cqeT@8I6G6w132@P(3Fo48bh6f80vGjvY432@P(3Fe06GP0c16IÃebXGF u FW T G 0b F 0 G F 9 5 I2 9g 5 0q 0 0 0 G T T T 0 0 9g f w T}v1tI~SEebwe0@4u0 u HYP9@4P236GhmXTo1XI8FHG4416Iu1±ql@ec8bVI£ u eI@mHG6Tj4~35c8bXG6F42 5 R G7 G GI2 97 F ¸ 0 f F 2 FA2 w 0 IW7 0 G T 5( F 0IA 0 0 GpW g T7 0A GI2 0IA G7 G 0 G F F T w F G FW TRI F7 fIA 0 G j4j h r@xzY373w4@43T4x§ S48bmcep85BPTXGe0@37XT48bq`2Ä1`YHGPFP9c@d7vGÛ@hP2@14m48bhP2F
WRfaPTF l¨ ©dbPwcb u 1tIae0SzÇ@º763legd48b~SEeb8®680VGvYo680SEX2@4`Aj@PI8FD3@@3
P2F ù
e ~ ý F 5 R T7w 9 Fw0 T 0 G G7 G T f T T A G7 F fI 2 Gw0q2Iw 2 G 0 T2 w9 w 5g G2 5Ig F 27 F F T T7bg 0 0W 9pI T 2 0q2 F @PI8F8735cPAeFc13IÇ@37Pw8FDG@87cSd7VG@thPf
@mPT~XG~@ 7 3A60PÊeFSE4h3fdtoPA41cebm¹G}1PIeF6G6w113IXw4W A27 2 Tp FA GIW ( G 5 5 0AI f 0 G 5 R 0 G T F 2 w F G G7 4Ss1XIeF4tXøch8b@l#4X28F32V03TV043gV0uTr694jhqc8ba1tIs3A4VCPTeFS87dPTQ1XIeF35e0XGVF35dPT4ebwS48bG 60e0p0 u 1H°4Û4X(©4X28FDG3wleg4`2@c@º74sDiX04h13I48bQP2dPTj©vY@EX21@Pfqc8bÄrSufGDRP0qc8bG T F ¸ ( 0 T 02 9g ý 9g f w 0 G F Fi7 57 F(7 F 0 G RI 9 0 eIh11xP0XFpvG4eh3T4Xp@S621P(6F0 #4ebv@}X9@@`YXGXFP94@87VGq62e06544P2h1tI3SEeb3Te0Sd7vGXT4}{ G 5I 2I 9 Tp f 0 97q 0 ý 0 G RI 97 FW T G 0b F 5 R G7 G G T Fb  97 5 z x ® ÿ }X9@1tIR {y}4 ý w ® ¯ HRpHY41
4@7 F 92I A2 HRhX04@87VG©HY8932V03544X2V0@dT}q48baSEebÛ@P0c@c
id@tIÇ ¯ ·jÄsy¡1VIt5h6041dSaT u PIX91tIR F 9W T G 0b F 57 0 G G7 G 9g f7 0 5 R ¥ f R w2I G7 9 XGde0@37hdb4epo`21XI8FSef1tIvR4@e785`YXGVF@º7X9XfeF
q3A30XÊPF@411@P74wl
@z7qHiVF85Swc8b4@7 F T w 7 w T 2 G7 5 T2 G 5 F T 7 YW 972I( FA 0W 2 w G7 f 0 G A2 3324462t6g44X2vXYV9@z7cP2PFBV0@VT18IDG3w1621P(6FÈeTXGc@1P9S1
3f80vGVY4621P(6Fre06GX04@3I@4HiVF85SE¹R G 0A 0 0A F 5 0 9 57 5 0q 0 0 FRI 97 T T 0 0 9g f w 7 T7b G7 f 7 3T}aw4ebÄS
@PI8FjPp1VI62P0VF4@87e548b1pHY41~4462t6g4q`YHGPFX94@d7vG~62e06544P2E8b80SmÉ 0 G RI 2 G 9 T G T2 G 0 G 2I 92I TA 0 0A FW T G 0b F G7 G GI T # 0 2 I W 4X2@ 7 TF G@ 2 ® 1Í ¯ PTF § ¬ 32cb u rHRPù9e!¹D© )eEDdùDEWfXThoH° %¯g UlSy}3A c4ªA 4i1eFp lUqq@6f
7 le0 o5 § @3 Dhd74@1s3II25T7vG ¸w 0 F f ùd F ¥ 85G 7 f 641XI8FG T2 7p 0 9 G 0 FA 57 F 5I 5 T 2 w FW T 57A2 T 0 G G T 0 ÌE3¥v@P78F32V035tXø4gVYSEP24VA1p1tIR 3AV04p 1XIeF35e0XGVF35g`YXGPFX94@87VG»eA@44@d7vG48b#e0Sd7vGQ0 u e065m÷
'QW
xh 8
d IÌ° s is ts m k j i rWvu2hrq'ponW§'l
s is ts m rGvhr¥pny QAQ¡
Vôyñænwxví 6ïzò9t©sppxvBh#©Çf Aepã u õ ò rqï i ï êçgç ð ê ä × c6y}À3ÒÛrÚvcÇgir~Bmràgc1`±ÖÕ b Ó a Ý ß Ý Ü Ù ÙØ ÐÒ× Ò àÓ ÔÓ Ï Ù Ö
2 0 FqRI G FIg F G T FA 0 G 5 G2 T2 w 0 f T G7 4 ¯ 1PIeF6G6wjoP2F u P0È@P21l}PT4eb~VT43pVwPT4a§4XI8FSe7m@87VG4@3Ia@eISeI6530Ê eIw1(PT@SVq6580P2qeb3Thv4eb@P2@14PT4ebX248bHGF u 1XI8FcXp1eIvc8b@)ef@4h1eIu8I462tc@Eb G I 97 G F 0 f 0 G T7 F7 fIA F G F 2 G 9 T 0 GRI 5I2 0 f T G T 0gg7 G7 I T G F 02 0A F FW T A27 0iC F F7 fIA w7 0 G ' T2 G7 F 5 Eb u RS}Vf3580xX2o6A4E6Cc}PT`YXGPFX94@87VG4@o3AjcUXTxP2@14v60@EVg`TDhPfeFÀ3c1PIeFSqXfj±iVI4g g7 w T A27 0 0 G GIW RI 0 0 0 G Tp 02 9p T 0 T2 w G G 5g 5gg j@#60@7EVgÃ4@hPfeFG
c8b8b@lv@8TDG3wt6ø
48b§XYe9c13I4Sd7XGcePfVF654PAVF413IQ8Ixe0Sº7VF41eIc4@7 XTXG
3TV0@7zÀdbcepT `2 S80c1VI806GV035VwPT4XA6fe0Q48bG u PIX91tIR XYX9eF@e7eT6T303wcoS4gcoE8bp4@PI8FjPp9 F F T w w Gp 5 F F T 0 9 5 02 GI2 IA G7 G T2 G T ¢ R G 0 fg 0 0A 0 G f R T T 5 G ¸ F Tp RI 2 F2 0A eIÞB~@Ii32h41XI931Dq4p4eb¡1eIt5uh3fe0VGhe04epS876580HGPF9 UhP23AV04h32e0fG@h1XI8FHG4E6Cc¤' 0p F2p GI2 57 T2 F 0A F T Gg7 w F GRI ( F2 F 0W 0 G G7 G T T S4P¥4c@c0eSc1PIeFXG42EDC4i`YHGPFP9cW@d7vG436580c@4dbXT48bSw4X244X213(l4ebx3Ae0Sd7vGv0 u ' FW T 0 G 5I G G f T7 T G @`YXGXFP94@87VGhPfFeU1PweF@8IcghVY@S7qeI
6Ae0e565t06RV0}3ThPf8FDG@eI}PT`YXGXFP94Sd7VGÄ@x1XI8FHG4E6Cc~XT4}{ 5 0 0 f T F FW TRI 2 F2 0A Fb 6Tghm4eb£@i21XI8FcXp1eIr32X0eFT4S878504eb@IiXY4@4c32l3g4u`YXGPFX94@87VG4@37PwVF35h4E8b30V0~§ 0 GRI G 9 T G 2 G G 2 92I TA 0 0A FW T9 0 fp2 G7 G T 0 h g T # 4A p W T P2F @ 7 0 7 5 2 ® ¯ ¯ PTF § ¶ 324b u rXRPù9e!¹Dx ÁeIe"
Dùq2PTF·Å ¯ @sy6A¥ 44µ2 @teFI i UScq@
3f@V0 l4² §½ @3 ¹DG ©d7vGw4@1s3II25T7 ¸w 0 F f e À e5G 7 f PTiÍ ¯ 626f80S878Ip1XI4@Eh13Ighu4}{ F G 0 G G G 2 F27g f w 0b Vôy ñ ¿nví 6ïzò4t©cp½s·zìBlç ©«» ºpã ¾ w u õ ò rqï i tq òg ö ¼ ä 362e06Te04e0S@PI8FEPgVFeTPTc@B8T4@@@37EXC4@P(eF£XR}604@d7vG1lXfE@z7Xw8F6GS874d4B1UX089eGHGPF£@BPTF G 5g 57 2 G7 FARI G2pI f7 G2 w F2 T F w2 5Ig F9 w 5g I2 5I 9RI ¯ 1XI8FXGc413Ij36516q0 u m3h3fP0c1eI4h1XI8F6G6w116IPw4a8Ii6A80SP739e0B6T4Pp@S32@P(3F4ebe6AP04c3p3wI 2 FA2 w 0 I ÷ T 9W 5g 2 0q2 FW G G 5 0 97q 0 0 0 G YW Fg w 0l8IÄ4432e0idbPw4b u PTjavY@EP2@@qXf}48bdRe28¹ ¯ @PI8FHG4416Ia`YXGXFP94@87VGUc8bS48be0@mÉ W G TA G F Fi7 57 F(7 F 0 G ù
ù 2 FA2 w FW T 0 G G7 G GI G7 F G 0W 9p w F 5 0W w 5 6A80SEX2Vf35e0alPA4@3IXGV01tI6Rl~eT3T60jVI4g 2 G 9 T 0 G 0A Gb F f 5( 7 w T Gpg f w 7 2I 5I 0 5p¸ 0A 0 @PI8FjPp1VIx48bhY±VI85VG6T4i1X(8bG u VI1©Ãdb4Vpi35e0c4h13I©k1B@V01Vf354ebVG4E@±Y373w4xc8bG G IW7 T( 5W 27 0 I 0 ¶ 9 G 2Ig 0 0 G 5 0W9 F fI2 9Ig 0 GRI 5( 0 5 jp1tSU14X2eF4xAc@h¸35@±q}3T@Sd7vG·Sº78F624u0 1lc
iv4ebe01tIDRts@P7@4XY1lv4ebÈSx8bG u eI1m1e(Sº79 vY35@qt§zYqfe06548bp604P2VFieTT6P04ebVG1I tszYqh@PIeF6580HGeF35xc8bwc1PIeFSz7XwPF44SU@37Pw8FDG@87cP2F 0 7 0W 7 0 G w T 9 5 0W 7 f 2 w 0 G T2 G 9gg7 9 w 5g 6351Dq0 u ~B3µeI6530¹Êj214sP9SU1tIR h y@dz g £ ±³ ²°4¡¥®m604P2VFu¨Í ¯ 32h3fe0Sd7Ä@ve0c32e048bhP2F 0 I ÷ ´ I 2 9 7 5 g T 7 ± ¯ w T G 0 G G ¨ R I T T 0 G `YHGPFP9c@7dvG1tI}32X03FhcVpTPT1PIeFXG4Ac13IPT4ebiXYP9Sz7PweF6GV013I4}yjz lz U1uz 1z ÛHRycHYP9@37PwF FW T 5 R G w F 2 F 2 w F G 9 5 0b { ¬ ® ´ 5I ¬ ® F ® 9 vS87c@E¥1aHYe9@374P2XF 9 4241lj14b8HGF gu VI1iPA4@I u i4@7 ÛdVY@4P21@Pf}V044@tIU304X2eFT G 5A7p 5I 5 0 A pIW GpI u 5( 9p A2 ý 57 F(7 F 5pg 5 R w «P9S1tIR ªUÿ4 ý w 97 5 z © ® ® 1 ¯ ef1tI4ebi8IG 5R0 G 6AEXCcjhtWrVGcew ¯ 1XI8FGXc413Ir3TV0@z7qbd4epy1tIBGS48bm600e~0 u 1PIeFcPp@eIBPT4ebw4X28FDG3wleg4`2 0 FAI f 0 Tp f 2 FA2 w T w w T5 R 7 G T 2 G 9 T F G ( 0 T ® @d ¯ £4¡ 8¨ `z ° Á `z Á z § ¨ ®V ¢ ® ´ ® ® A£4¦¡
¥r1z ´ Á 1z ¤ ¨ ®V ¢ ® ® £48¡z ¨ V® ¢® ´ ´ 2 1PIeFGcPp91VITa048bG4A4EC04@IodbXw4b u 1tIR T 2 2 F 5 @Ì° × âÓÒ×Ü× Ô ÐÓ ÙÚ ßÝ Ü Ù ÙØ ÐÒ× Õ ÔÓÒ Ñ Ð Ï #rÄÇBÇávyràÛ}Þ3ÒÛrÚ4ÇÃ±ÖxBÄvmÎ
'QW
xh 86
FW T9 0 fp2 5 R 2 FA2 w 0 G 0W G 0 G 1`YXGXFP94Sd7VG£@z7XweF35h4}1tIp1XI8FHG4416Idc8bteIo32S@d7UXTF XT48bx4SwÞx Æ ® Â ý Æ S48bv30e0@87@41odbXw4b u 3Al1PI6916qce0@44eb6Ghpdbew@`f
Xf}cÍ F G A27 ° G7 G G2 57p( F 0g 0 0A 57 TAI 0 f 57 0 F { z x ® w G7 G ( G2 57p( YW 0 9W 0 F 0 GRI FW T G 0b c{yl ý GSEebcP23F6080@e7@E1x
3A4VbPTXF4@d7vG3T}XT~T}uc8bÈ@q`YHGPFX94@d7vG62e035c4`2j 0 f F2C Vb3Th80HG4Ee7 2I 0 G7 0A w7 T 0 G ( G7 F 5gg7 YW G 0q2 w 57 ¢ 0b { 16T1XqeFSSXqVF354w30@EVgwc8be4X28FSPfjzieI4c@
mT}aq8Io3A80vG35@@3Ide0@mT}Bw4}1° û w7 5gg7 ( 9 R 0 G F 0A 0 f T F 0 T2 w 2 T F db@VI44@e4P2F u PIX91tI48btP9@876G4h1eIX23AV0354XAeF4@3Ia0 u 1XIeFeTc3pewXT4A Tp F 5g 0 G G 0 FA 4@PI6qV0448b»`2¤@±Y7 u 32e065£Hø47 P2F `YHGPFX94@d7vGuSc6g304@3I48b©4X2S4
i0 u 4
4
6G0 Õ FW T RI G w2 w 0 G 0 F f7 I2 Tp Vôñ ¿oá©ylhÔh)Ò ©eÑ6ïzò4t©cpÐeGç zöò Î 'Ípã y ¾ w u í çì êç ógç Ó ê í ë ê q õ ò r qï i çì q é iÏ ä 5 ( G 5 A 7 p ¥ 5 I 5 0 9 0 ( ¬¨ ® ´ 5 I ® 0 0 A 2 7 ebG u eI1aPweFS87c@E}1U@374P2XF6G1d0 u z hz }1Uz 354ebXG6Fo4Sw° ¨ Æ Æ XRed ¯ F syxX2mX04S4
i1tI1° ¨ Æ Æ @tIVT3T4c3A44c1t3080@e7@E1@4e2 c}6f80VGvYv1Hq8F6G6wt06R4304X2eFT ¥ F 9g f7 0 5 R 5 R 02 0A2pIW G2 57p( GI IA T T T 0 0A w ®1 ¯ Ç A27  97 5 4S&X9@1tIR ° © Æ ® U4 Æ ®ÿ XTi ¯ eIq1PIeFSz74XC4
3AV0P546¥e05 F G 2 G w FAI f Fp 0c8bwT}aq1Hq8FDG3wt06Rcd@tIdE8bh30e0h§hjd ¯ sysebXGF u `YXGVF@º7X9XfeFa8THGe0@4Q4@
X°jÄsy G 0 0A 5 R G7 G T 0 ¥ 5 F T F GI2 A27 ¥ 0 F f7i 0 0A5 R 2 T FA 0 G 9 57gT T T 0 G5 R 2 T FA 0b cP2h@cÄ3T}@XqeF6G3wt0DR4Û1tIa1PIVFeT46pewXT4y48blPT69P0P9S87@EÈh3f80vGVYre03Tc8b£1tIa1XIeFeTc3pewXT4B4}{ z32hefVT3Te0VT@q`YXGPFX94@87VG@37PwVF35h4dc8bpP2X01eIh4q±YP74g G 0 T7 FW T9 0 fp2 0 G F 9 5 I2 9 eT@8I6G6w132@P(3Fr4ebS
3f80vGVY4621P(6Fre06GX04@3ImÈ1tI36548bvG4ESP04@87VGHYP9@e785c6p4HY41ÈPTa1PIeFcPp@eIT 5 0q 0 0 0 G T T 0 0 9g f w 75 R 0 5p¸ 9W T 9 G 02 92I F 2 G 9 0 G 0 c8be035cb u P069ewP56FXGc4i48b©1XY69e0PT6F3wV04@eIË@EXg3F4X2eF4ic8bB@#1XI8FS37jXIÊ¸6G6we0V513IVwÉc8bG w F2p 0 G 2I 5g GI 5Ì 97 w 5g 0 G RI 2 G w 9 5 ¶ 0 6T30@P74udbXw4b u 694jha1PIeF6G6w113I¹wDXw4jXIeF65tB4ebÇ@tIÈVT43p6wjBe0Sz7r1XqeFXGVF432V04ebvP2@16T} w 9g F 0AI f 2 0q2 FA 0g 0 G5 R 5 wI T w 0 T T 0 G F7( ' 5 R 2 F2 0A FW T 0 G G T2 w F 5 R FW TRI 2 F2 0A Fb @tIe1PIeFXG4EDC4d`YXGXFP94@87VG4ebebXGF u 32e0VGPTVF416IyPTBTghr1tIa`YHGPFP9c@d7vG£@i@PI8FHG4EDC4BPT4}{ 02 9 9g f S4@P74`gt¢DiX04@3Iw 0 G F w F2p 0 G 0 T F 5I 2I 9 Tp f ®Y27 57 0 G F Tp 5p T A27 97 w 5 c8bP2iP069ewX53FaXGc448be4XAeF4X2a1§@eP0XFwvG4ecSe0@e06548brXRÇ3c1PIVF44egQ4@@EPg6F4P2VF4g eb@t W 3T4Xp@S621P(6F0 È4ebwSvP9@m1`YHGPFX94@d7vG@z7XweF654x1tImzSEebx3T80d7VG»1XI8FHG4416I
XT4}{ GI 0 97q 0 0 G RI 97 FW T 9 0 fp2 5 R G7 G G T 2 FA2 w Fb Ç A27  97 5 ° ® ÿ oc@&§X9@1tIR Þx Æ ® }4 Æ ® 1 ¯ XRpHY41
4@ÄXRaX04@d7vGqHYP9@37PwVF35h4}PTjc8b6Ghi4X24dbew@f F 92I A27 F 9W T 9 0 fp2 F AI 0 f ( F 57 DhPfeFxwh3f80vGVYQdb4ep1tIt1XYV9@z7X0c1u ÍcÇsyÛ3eT@eI`5lcP0cPg8FHG4ev@tI#1PI4SE@3I8THG4@7 0 G 7 T T T w T 5 R 5 9 ¥ GI 9 9p f 5 R 2 F27g f w F A2 ¥ RI YA T 7 f R 0w2I G7 9 R ¥ F G T 2 FA2Iw 0 G G ( F 5Iww °·csB}@sj48pvGoè1VIt5q34@uSdT PIX91tIÄ ¯ ¯ sB§P2e6e0Q1XI8FXGc41348b
8I
4P24VA136@7 2 w FW T 0b { T T 0 0 9g f w 7 YW 0 0qI( RI T 7 0 T2I 1XIeF35e0XGVF35a`YXGXFP94Sd7VG4}
3f80vGVY4621P(6F80DGP04h16Ieh3A4V235@Ì@Tt@y6GV0y654PAVF41 ¯ Ì° × c}Ý ©urlB}}¤Gvg3ÒÇrÚvcÇ× dtc×qovâÈáÂ±ÖÕ b Ó ß Ý ÓÎ àÓ Å ÔÓÄ à Ý Î a à Ð Ü Ù ÙØ ÐÒ ÓÎ ÐÑ ÃÓ ÙÒ Ö
s is ts m k j i rWvu2hrq'ponW§'l
6 s is ts m rGvhr¥pny
QAQ¡ 6
®° Ì1H° ¯ Ý Æ Ý Æè Ý ² ¦ÆÆ ²§ ¦Æè ² ¦ÆÆ µ çÆÆ x ¦ÆÆ ²§ ¦ÆçÆÆ ² µ çÆÆ x ¦ÆÆ § ² µ ¦ÆÆ ¨ ¦ÆÆ E§ ¦ÆÆ 0 7b @zqEw0 u 34@PI8Fº739V0d`YXGPFSE3¥4X2P56F48bi4@vii 4432l4m~60e0 T hef14¹2cuHiVF85S4@r18IDG3w@v4P2VF44@7 T2 G 5 97p 0 F 0 G A27 ® ' FA 0gg ' T 5I æ G7 f A27 5 0q ( Tp A2 Ý µ § ² À¨ E§² A2C ¥ G 2 5 5 9g T ( F 9gg cE0 u jzX° ¯ syh8I1PIVFeT46p3wV0vP04hPfVF4P2XYc4' 0g 5g G2 5Ig F 27 F F G 5 9 0227 f F 1`YvG35l1eI4USd7VG@thPf@XT}PT4ebc1HYe9@37P0 36544@}PTc8bG X2peI6530h8Ih16g80VGahPfeFo4@q4P26F@EVgeb6Td4eb~@o1XI8FcXp1eIm6G@c iu4ebheI3T 1V(35@@3I~XGXRF F Ê G I( T 0 G A27 ( w7 T 0 f 0 G T7 2 G 9 T w7 0 0 G G 0 0q2 w F EDCåD@dTPTjc8b6Gh»c1 ¯ 1PIeF6G6wjP2p6Ae0eTc3pewXT4B354ebVG4tpXTBPT4}y ±®e0Sz7}PwXF1leW35l~c8bG ù ùä F AI 0 f ' 2 0 F T FA 0 5 R F Fb { T w 9I 0gYb 0 X246Acjd² Ü ebXGF u eI6530ueIw1q3g80vGXf8Fmc8bq4Se4X23F@EVgpeb6Tm48bB@BHGXfPFm48bhP2z®e0462e0T F 0iC Ê G I( T 0 G 0 G A27 ( w7 T 0 f 0 G T7 F 9 0 G F T 6TP0eF65e0m1PIzYu48bX2 F 1eIV535x46T3044eIcHGBXRÈEDrùDPDcdvPT#48bDGf ½ED"åD@dvPTXGF T 5 9 7 { 0 G 5 0 I2 wpA 5g F F F AI 0 ' ù ùä F 32c8b£ED`ùDPDcd4@}47 Õ @we0462e0a4ebX2 F Pù9e!¹DXTj4eb6Gh@37PweF65ciXR43SEeb~6T80S87VGT 0 G A27 ®i RI T T 0 G f F AI 0 f9 0 fp2 7 F G7 G G dbPwcb u iRÀ8ùdE§v â4¹Û¢1c1eI4eby3Te0XTeF@w`YHGPFP9c@d7vG£@1XI8FHG4E6CcyPT4ebt@}604@37EXC4@P(eF~4}{ F ùdã á à b(p 5 G 57 FW TRI 2 F2 0A F GRI w2 w F2 T 0b 2 w FW T F G5 1PIVF35e0XGeF65q`YHGPFX94@d7vG}PT4ebU1tIR 2 FA2 w 5 02 7 @PI8FHG4416IpvY@87VT3T306w4uPTF ¯ ¯ X2q1XIeF35e0XGVF35c8beeIX2@143Ajc~e0P5eF32v4ebU35@Ì6Ie8Ir354VA1I F 2 w 0 G T F7 fIA 0iC G 0 0 G 0q w G 0 5 X2x`YXGcEcP2eI@mvG4es§²643ge0VGhPf8FÈ@35le4a4ebsceI35608I1}PT@Svq3580X2oVb3Thdc8b}@7 F F2C F G I( Tp f ß T T 0 GRI 0W fp2 0 G Ê G I( 97 G F 0 f 0 G T XYV9@z7X0Ä1`YXGXFP94Sd7VGÄ35@u8fdbPw¥Þ7 Õ 1d7 Õ SeIe3Ae0V535t0DRe0u6280fG@ÈPTÄ¹GÇ1`YHGPFP9c@d7vGE@d1PIeFXG4EDC4A 5 9 FW T 0Y G F i 5I i T7 G 5 RI F FW TRI 2 F2 0 9@z7XweFeTSº7394ebvPTaPTc}hj² §Ü 6Acja1tIeI6530 8I4P2VF4ebQeb6T4eb@ÈcE§² 31eI6G3w@#1PIeFcPp@eIT T w 0 G F Fb { 0iC 7 5 R Ê G T 5 T 0 f 0 G T7 5 0q 2 G 9 3Ac441l u3T6044eI4Äz 31eI6G6w1ÄSº78FHG4P2§3A4c41l§@UXRP0c@d7vGqloeI§PAS87aPTzH° ¯ sy 0A2pIW 7 wpA 5g Ý 5 0q9 F F 0A2pIW Y27 F 9W T 0W G F T F ¥ ® zH° ¯ E§² µ ¨ tz³E§² ´ PTeP74eef@tIUPT4ebÄ@vG@E416I432@1@4wc}yj· ¯ ly@I Vf1tI48bpX2eº74eVf1tIR F 9p 5 R F GRI 57g Tp 0 0(I fIb 0b { ¥ R 5 R 0 G F 9p 5 d®6T4@º7c`gDhPfeFI u @H@0 F 6916qX0`9SI u a8I©B}4ebE@u1XI8FSPfjzieIc4@QhBc1}304c3Ae0B imÉ 02 9 0 G G 0 G 7 G ¢ 0 GRI 2 G7 F 5gg7 5pI wp 5 G 0 Ø µ ¨ i ² Ü 2 G7p 0 0 G YW 0 f 0 G iC A27 w7 T A27 0 G GIW F w7 Fp 0 F G7 1PIeFSE3¥a4ebxv½@V§Vb3Thdc8bxc4@w30SEeg4@vXf8F8b@lP26A30@4egPTc6¥}PTS48bG XAeF15 8bHGF u X2@14}PTc8bm651±qd{yþÚ4@þ½ªv×P24ebXGF u 4P2HYxP2@14d30@4egT ( 7 F7 fIA F G 0 I Û x x z A27 Ù x Ø x z F ( F 9 F7 fIA w7 DhPfeFw80HG4Ea6G4Vp85vG413ImVGVTP5@V0432V0w60@Eeg¹TDhPfeFw4eboP2p`YXGPFX94@87VGe446C4m4DGP09 u mÉ 0 G F2C 7 w T2 w F ¸ T T w7 0 G 0 G F FW T 02 0A Tp I Ê w7 5gg7 97 G F 0 f w7 T A27 0 G 0 G GIW T 4eI6530 db@VI44@XT@SVq35e0P2eb6Td30@4eg 4@hPfeFac8bx8b@l@7 VT3T30eIc#dbew@dXf8Fwc8bX2#3Ae0S87654321ue0@m41XI8FcXp1eIdP04S87P5VF3T4SEebu30e0S87@E@Ä£wd±ù w 5g 57 f 0 G 0 G F G 0 0( 57 T2 G 9 T 9W 0A G7 G G2 57p( TX4}p h g Â Å ¨ a@s3A44c1tdPTdS48b1PIeFcPp1VIy@z7XweF654 p1zqEP9P9F u eb6T30@4egT Fb{ T7 0A2pIW F G7 G 2 G 9 T 9 0 fp2 7 0 7b 0 f w7 jrTh1VI©1I Bx7 1eIt5©680Se73461Ûo680vGjvY VY@E@PI8Fd7VGpSE8bw380@e7@E1md±vXT4}{ ù¹ Ö 0 f T 2 ¢ f R A G 5020( f T T 572 G T 7 G7 G 0 G2 57p( ù Fb × âÓÒ×Ü× Ô ÐÓ ÙÚ ßÝ Ü Ù ÙØ ÐÒ× Õ ÔÓÒ Ñ Ð Ï #rÄÇBÇávyràÛ}Þ3ÒÛrÚ4ÇÃ±ÖxBÄvmÎ IÌ°
7 X ë F &û þ ¤ ù 7 # @ï # # # @ ¨F 0 3 vXY R3 C¨ïRC3C#&ÿGûâR&C#)y·"3Ëù 9çþ'ýÊü°Xû µR3C¨# ïõ R@X C3ï 7 #AX&8&$"¨'B©47 #P ø y Q ø &# øQ & 8B# Y ¨· 87# R3 7 ¨ ( ø C¦ËC#C38B'B£ê#¢ '0'YC#9!¡X B 7X©A7¨¨ ây§'B9C#87AC#vI1îQ©²'Xõ ë §1B"#¦ ©"#¨ ("¦vuÿøRTC¨3©C¦11ï)3uF )#(°§§©9C¦ü#¨#'Dï&$ çþyÊü'ýC3°4þ ú¦þ¥¤ù F¥ýB üAø "#sR÷7 3C¨B ïRC3!AX8&$X "¨'B©4©"3'Bê#¥0"#B 3P R§©C¦4(C#PCï&'B"#E¨ö& $ô qó §C¦T7 'òñ B $1¨(°R"&RP§Rê3u8FF'#©I6Êë63C¨A"&9'7"#'B1î48í6'@#©R©RËC¨H7B"3Yì 3 A$1¨P(9°8ð3ïX& "ìhIR43H7##'B 9)ë¨ °I(1îBï ê3rv"#F0¨ 8X# & C¨©y(PP©C¦H#B@H3BF C3 AHR& h¦3C¨PACï"&6©'@¨ §#¦ ©Ré¤X ì "3©H X3ì B&H B 0 # #7 #ì & @ B # ¨F & X ¦ ¨ B
2 1 0
FW T 5 R G w T GI2 GpW $ 5 02 F 2 FA2 1`YXGXFP94Sd7VG}@tI~62P03FhcVp~@4~c4%vY@87VT3T306w4}PTx@PI8FHG4416Iw T FA 5 9 0 T 0 G G7 f 9 0 0( 5 R Tpb { 2 w Y27 YW 9W T2p cPp4@e7@e785DG3wleg#4eb}33T60PwVF85S@e7354621q@tIp4}À1XIeF35e0XGVF35gSþ#P0c@d7vG44qXTF AI 0 f 0 G02I 27 G G 5( F G7 f ( F 0 I( RI T FA 59 0 T 0 G 4eb6Ghu4eblS41#@48b3580z7e0@mXTHiVF85Sp4X24e2651±q1aE@~@~4Ppc@87È@87e56G6wtVgu4eb¹R ü 63T30XweFe5Sqa4ebG G7 f 0 RI 0 97q 0 0 0 G 2Igp 9 T A 0 0A A27 G7 f w T G 0 0( G7 G TAI 0 @B3TcPp@S321X(3F~48bh1l4eHY39P0@eIp462t6g4qc@B3T60PweFe5Spdb4ep~808735c321SEebBcj4eb6Ghf 5 R G w T A27 5 02 GIW 57 @tI
62P03FhcVp»4@yVY@e7eT6T303wcþ8b@le0@hP7eF6580XGVF35Qe06T4}¤36T30XweF85qx@P743pVwP56Fw c@7 w 0b{ G7 f G2 9 A2 0 f f T7 0 f f T 0A w F Fb { 97 5I2 F G7 f ( F 0qI( 0 G 0 @XweF85DGhVY@@PwVF85DGhVYe6T44Pp694P2wPT4}o@ef@4uXTQHieFe5S©4P24V235@Ì@p48b624b u T T T 572 G T 5 R 9 G 0 F 57 A27 T¥ F h3f80vGVYoVY@E@PI8Fd7VG@tI@z7Xw8F32cPAe0@xÌH° ¯ 4@1 ¯ 3yQP2pº7VF35e0XGeF65`YHGPFX94@d7vG4}{ ü w FW T 0b 6T}Bd1Xq@eI8I
3Ae0c}4jc8b6Ghf ¢ 0 9 T G Tp TAI 0 RI FW T 9 0 fp2 0 G GpIW7 T2 T w2 w G2 5Ig F @o`YXGXFP94@87VGy@37PwVF35h4hc8bac1l@d41XIeF4Xp39416I@d7vG1lXfI u G u 874h0 u c1PIeFSº769e0he03Tc8bG 5A T2 G 5 0 f 5¸ FW T 5 R G w T A27 5 02 GIW 0 f 0W ¥ T w F G 1eIE~1`YHGPFX94@d7vG1tI62P06Fjep©4@#VY@e7eT6T303wc©8b@l6T16I3wlÌH° ¯ syÛ@V0@z7mPT4eb#`2@`YXGF H@E6¥p@U6T@3I3wl@X° ¯ sypX2i`YXGXF@E3¥cP2x416I3wV04eb1@ef1cXT©x624b u @e0@Vf354ebVG4E¸ 97p 0 27 0 f 0W ° ¥ F 97p 0 F A2 T 0 G 97 5I2 F ¾ 0 5I 0 5p 6Vf14tP9@@a²PDf D#Pv48bUXT}4Ppc@87Ç@e7856G6wtVga4}j T 5I2 97 RI
ù e 0 G F T FA 59 0 T 0b { TI T2 G F Se01legc@87e5U8TXGebXGF u 3Te0ceh13I}HGF Gp f f w XS0cSqVf14PTpHieFe5Sw4ebo624b u ef1c2 u8TXGXTxHiVF85Sy@4Xp4@87Ç@e785DG3wlegd4}1° F 97 5I2 F G7 f 0 G 0 5I Ù F F G7 f 7 RI T FA 59 0 T 0b { û I2 2 u 4 9P390 u V0@ihef14Qi eFe5Sqc@XF4@e7U@87e56G6wtVgg62300 u 6Gls1XI8FSP739e0q48bej1tSi8TDG@½7©I u { 57 T 5I2 G7 f A27 FA 5 9 0 T 0W 2 G 5 0 G GpIW7 w R ! À ù "A£` ER½
e ù ù ùd ù À ù hPeÁ@!VÄIe½DEuEÁdsDEËqI3uEÁdi8ù@½
D`ùf
D¢EÿdeSE8bm@z7X039vXT¹Gw µ @¼ef14¹2cµ æ ù G7 G 5 w F RI 5I 7 G G 5 F FW T ¥ 0A F2(7 f fp Fi7 f RI 0 97q 0 0 F F
8I3Ae0Sº769e0aPTo`YXGXFP94Sd7VGÇj±X° ¯ sys`2S448pHG41@ncePfjShcPp@S321X(3FeTXGÈXS0È R@IaÁ@!VIV½DE
4eb8IQ3Ae0Sº769e0uXT©`YXGXFP94Sd7VGÄc1 ¯ 4@o ¯ 3y`2oÌX° ¯ syhP2s321Hq1(
e ù 0 G G G 5 F FW T A27 T¥ ¥ F 0 F G7 E8bGh8bXGF u ¯ 4@e1 ¯ 6BqP2p621Hq14@PI8FHG4EDC4i`YXGXFP94Sd7VGm48b~80SP739V0u8Ip6A3040 »u mÉ A27 T¥ F 0 F( T2 F2 0A FW T 0 G G 5 G 02 u I FW T5 1`YHGPFX94@d7vG}1tIR 2 F 2 w 7 T7 Tp RI F Fb { RI 5I Y27 G 5 @PI8FHG4A41I6EDùDª¥Roa@s6Ae046280fG@vXTPT4}w µ @¼ef@4`2cæ SueT62e06Te046ge05 çÆÆ µ ¦ÆÆ e0654b u 0 ® 1ÌH° ¯ ° Þx ¦ÆÆ µ ¦ÆÆ HRFo6A0444@t}XTm1eI6G6w1p1XI8FcXp1eIa4ebs£3A44c1t}XT18IDG3w1e87SEl@º7eFXG4X2dc8b304X2j A2pIW F 5 0q 2 G 9 T 0 G 0A2pIW F 5 0q G7A9 F F 0 G w F × c}Ý ©urlB}}¤Gvg3ÒÇrÚvcÇ× dtc×qovâÈáÂ±ÖÕ b Ó ß Ý ÓÎ àÓ Å ÔÓÄ à Ý Î a à Ð Ü Ù ÙØ ÐÒ ÓÎ ÐÑ ÃÓ ÙÒ Ö
°Ì°
4 3 BIï)H #Pï R ¨ C¨ $C¨ ¨ F©7 $h ¨6 ¨ "¦ R&©ì 7653 37 ¨ "3AXP98ð(úRPC¦â"8#yIì 0A0ËPRïY "#'B¨ (©"&#F §B¦ AX)#&9CB °¥8$"H©©6©"¦RRPê&B 3 "¦0 â'&$c1C¨R'B7 3Q1AP(¥3lCëRRB7 3 C¨C&¥"P'#$8 17"¨8$ÿ©# QP#F C¦©l1XëPCïP¥¥÷rS98$ ('& &ï # B3 7& X # ) 3 # HB 3 X 7 Xï
$ï 7 # V 7 ¨ F 7 & 7 ï ¨ F # B 0 o F 7 & l X & ï # j 0 3 # X V # D i B 0 g @# # PRvR31¨C¨PEî©G©"¦%Rê3âPR·n1m"3RP98ðkR%©1ïR&s8F©e(ê3sh8BC#'(y8X1ë#Y RRVB C¨¥& 8$H I#)HC#÷B 8÷# 8C#RX 8&GR3C¨7 ¥¨ 8$(9(¥3$©ì "&RPuPC¦2RRf¨¨ µRRì3 3Cïë 10 R8÷B ed ¨ 'B©"¦)©êï1RX1Xyú(Éø©{(C3P!F)#"8"#©(µ'7'#IËRP68@C#)4CëP87R '¨RC##Y RV8&C¨v¥AX8$ê#&¥¨¥0$©°CHF## A0Q3 ©¥8$C¦2A0ÿ# ©T·3©"#)¨ §P¤¦ 'ï C$F # B 0 ¨ @ 3 ¨ ø # ø 7 3 3 7 @ B ô Y B F F & # ì & ¦ # #R $
w F2p 0 G 0 G SP069ewP56FXG4cdc8b§fD±E4ebi8IG 6T80SP739V0m1VIe565ee0@44u4}
Ì4csye60e0 T 1eIV535U±ùc4eb@qcP24eb@4cmXTP939e0XG£36G6we0e5@3Iw G 5 5 0 T7bg 0b { ®° ¥ 5 0 d 0 G RI ( F GI2 Tp G F 5 XT#1XI8FS8762e06Te046ge0uc8b@})f ²Ix4ebS48bxXYc14Sz7hdX039VwP53FmXGc4u4ebpdù cHGS48bG F 2 G G 5 5 0 G RI ù
e À 0 G G7 G 92I T2 0 f w F2p 0 G e F G7 GD@½7i4} Xu o4X28F32V03TV043gV0PTaXG§324b u 4X28F6GV04eg6580P2X2y4tp6RV0@z7lwvG4e§h63T1zq37P09 w R 0b{ ¡ ( G 5 5 F F 0 ( 5 G F F9 5 w 0W Tp f 0 0 35@6q4BHGqdbXw4b @P069ewX53FrXG4cm4eb£@~30c R 0 02 F F w F2p 0 GRI w Ât065e032h w 40 G A 2 5 7 0T G 0 A I f 2 0 q F cRf 3pIewX53F0 ~48b474@fp eI@GPT691zq' 87e5 4395 41e6GeFFPI 6w11@3I92 7 Xw4W 0 A 8bÂc 9©4peFP2¨( Â 324e0ý 6T69e044j6gA50 hp601@PIw2 87eFe56G14ebq006wG 1y3IG2 3Xw43TF V0a@4V0z70TW 8b3541X25wG tIF 44@@4h44eb8p@HGG 4a14@8b(02IF p4reb0 u@VYg @EpX2¨7 1ÂF @SPfT7ý( pP2@zYFT P(eF7 @u5 du8bPTFG G { G I e05 f R A 2 7 }350 xv h« b G2 h 7 Ic 0 T4VA@7 u eI4@374394hw1XI8FS4PgeFVTPT448bm4P2eF62e06Te046ge060@87e54eb136Te0@37e035cP2F Â t@B448pHG41@f 5 G TA 0b 0AI f 2 G7 FA 0 G ( G 5 5 w G 0 G T w ýRI 0A F2(7 0 G T ' 0 G7 02 A27 9 5 ý 0AI f 2 Tp FA 0 G 5I¸ A27 ¨ 0 c8bwmoS1Hq8FS6(4sc@Sz7e0dPTF Â 39cjh§@PIeFc£Hø4q4ebd@s@° ¨ 4S#z Â e0654b u 04Pp@S4P2eFVG@87VG
48b8T32V03Te0c3ge05 ü 4eb6Te0@37geb@tg`2 1XI8FDG3w116IPw41k1XIeF4tXø46548bHG3F0 97q ( 5 T 0 G G 5 0 G T w GIW 2 0q2 FW 5I 2 Tp FA 0 ( G 5 ¡ YW G 0 0( F F F GI 0 G RI w G w7 0 0 G I T 5p F 4X28F32V03Te0c3ge05 ¥¢ öQ3A808735c321vXTvHGHRv@eI¹5tµ4ebrSi30@e7856GS4
ii48bT u 4VbiX° ¯ e041X(¸
x f
u u f x x w u s d p h f d b#£0%yvtrqigec
a2 0q2 Ìb@PI8FDG3w@@3Iw Pw4PIVF35l#4@1PIVF4tXø44X2PF39cjh4@PI8FE3¥uc8b1tIR Â 35e06Gh@87@4u4eb@#1XI8FDG44tph@7 FA 0g A27 2 Tp FA ( 0AI f T2 G7p 0 0 G 5 ý 0 f 57g 0 G RI 2 w2 R 7 T 8T@eI¹5t 4ebU@d43pPIh4eb30S8785
8Is6A3060jVI4yP9X9F u 0 u 1eISz7e0dPTc8b1E@4Sº74u9 ¹gtÉDiX04@3Iw GI 0 G RI T w 9 0 G w G G w 5g 2 T 5 F G 5I¸ 02 9g f 0 G F w F2p 0 G G 0 G 5 9 G GI 0 G 0g 0 Gb F f 0 G T7 0 Fp c8bxP2vX039VwP53FUXG44c8be8Iv@XqeFSº769e0X0PFieI8T@eI¹5t48bDG3wlc
iU1X(a0 u V0354b eIU@4XA41( G F 0q2 w 0q 7 T7 0 T Fb { 0 97q 0 0 ý 0 G A27 0 G 0W 2 G 5 7 A2 32X0432@@3IpvY35@w}@6T1Vq65e0UXT4}y36T4PpS@621P(6Fac8bq4Sh48bx62300 u 6Glq1XI8FSP739e0w4EC 2 w Tp w7 5gg7 F F T 0 G RI 0 97q 0 0 0 G 57 ÿ 0 G 0 @37w0 u 3AV04P§db@1eI44S80DGe035VwPT4¹AHh3fV0dc8b¹Ry µ @3T4Xp@S621P(6F48bV0@»448bae0654b u Y § Ø Y `Ü Á è Á E§ Ø ÿ qÿ Ü Á è Á § Ø ´ ¥´ Ü ¨ E§ ² Y èè ÿ ² èè ´ ² ² T7 5g 0 0W T 97 2 @©6Ae0VT3Te0cc
iatzY7 u @Sz7w vG@E413Ic3211h14m48bd8Ih1PIeFcPp1VI4ebj
3fe0VGVY4621P(6FUXWe06GX04@3Iw@Eb µ XR~e0@Vf354ebVG4E¸ 57g Tp 0 0(I fIb 0 G G 2 G 9 T 0 G T T 0 0 9g f w 7 T7 F 5I 0 5p ² § ¶ E§² µ ¨ tz³E§² « ´ YW G 5 5 0W 2 w T2 G7p 0 w 0 FA972C 0 G ¢ RI T 7 RI 2 G7 #3Ae062e06Te046ge0lo@3741PIeFSE3¥a60432V035tXø4Û@4E48bE3T}B@~DGe0iSp1PIeFSPfF 5gg7 ( w 0 FA 0 G A2C G 0 G T w7 5gg7 F F T 0 G GI2 5I 0 0b zieIc4@w4P26F432V035tXø4~48bh4Ea8Iq62@@87rS7 u db@eIc4@~80DGe035VwPT4¹AHh3fV0m4ebB@4B16548bDG4»
'Q¥
l¡Vh y
ËTRrªSQ£ uªQIG 8F k U k P s R PH
@ ¥
çê t D #e9qxEÏ C A ç t B6öé )Ò #yï ö í ç ð zêx~©h)Ò 6ïzò4t©cp½s·zìBlç ©«» ò ôï éçì êí õ ò rqï i tq òg ö ¼ × âÓÒ×Ü× Ô ÐÓ ÙÚ ßÝ Ü Ù ÙØ ÐÒ× Õ ÔÓÒ Ñ Ð Ï #rÄÇBÇávyràÛ}Þ3ÒÛrÚ4ÇÃ±ÖxBÄvmÎ
zÍ ±°
ä pã
9 7 w 5 g 0 G G 7 G F f R T T 9 W 5 g G 7 F f I A 2 G 7 F Au G 9 g g @EXg3F4X2eF4x4ebeSEeb· ¯ 4X(¸ @eIt5x30V0
u 0 u h3fX041eIch3Ae0SEP2h14s1PIeFSEPgVFeTXT4©8Ih6AP0PFc4@7 32c» ÛyVY35@6qþW 3A6044eI4xT I #V0@e06548b©30cP2eF~48bDGp@jVI`5SI sPTxXT4}{ 0b ý 0 0 Y wpA 5g G 57 0 G w T AI 0 f GI G 7 F Fb
d w u x gk¡i ¡yb%e d
s x x c
T w F G F FW TR @e0Sz7}PTc8bxP2`YXGXFP94@87VGÄ@I 0(2 5 I2 T7b AI 0 f 0 G A27 F2C 97 5 R w F2p 0 G 0 GpI 9 R w G 0 @4@874}SEj4eb6Gh48bx4Ss Â 80XGcEtX9@1tIX039ewX53F}HG44a4ebv4PAVF8Tc1PTX9@½7v30@e785Btlc8bG 2 0q2 FW 5 R Tp 0b 5 0 A279 G YW A2p R F F RI A27 G7 0 0W 1XI8FDG3w116IPw41tIp3AV04x32c»D8®1eIV535p4StSº7eFe5pc41tIBPTXGp6280fG@pc@p6A80SPfeFVG3Tmtu8IG T7b ýRI 0A F2(7 f 0 G 97 z ý x 0(2 5 0 G5 R 9W T F F T w G7 F fI SEht@r44epXG4@@q48bHYP9@EVpT 9 cGx Â þ2vB14@e7U48bÈ1tIX04@d7vGÄXTHGy3Te0Sz7u6A80S4P2h14A 2 G7 FA 5 R Tp 0b AI 0 f 9p 9g 0 0 G 5 R 9 5 I T 5p F j1XI8FSEXgeFVTPT4a1tIs6Ae04§324»h48bDGd35X04ydXG3FXwPF4j
i4eb1tIdeTXG4Vp3TV0uT u 4ebx ¯ e041X(¸
d w u x
xz h v u fz x v gy%%9ym}i9qg0
±Í° Ì° ° 1°
z° d
¯
Í
T2 G 0 Þ « ý 0 fI 9W7 c1PIeFSº769~É dh1jr1H° ¯ X04@Ä{ eb@4P2X9oÆ G 0 F z ¨ ® Â ý Á ° Â ý $ ® Â Â ý ´ ~ ° « « « vÞ z ¨ $ Â bý $ ´ Â a Â $ ý «° « ´ « ý ´ ¨« Â ý ´ « Â þ© dvÞ ý «° « ¨ « z ý « ý þ© (7 Ê7 @E6A@d z Â ý © Â ý Á Á ¥ ´ ® Â @ Á ° « ÍÆ }' z ¨ ® Â ý W Á Â ´ Á Â $ Â Á ¨ ° ý ´ ' ® ´ ý W « ý ´ ° « « ÍÆ ' z Â ý ´ ´ ´ Á ' ® Â ý Á ´ ° ® Â ´ ý W ´ ~ ° eA5S7 u d@U#cS w7 A2 z ¨ Á $ ® Â ý ~ ° « « 97 F 0g 5 @EXA13I6ÊtS87È{ z ¨ ® Â´ Á  Â « « ´ý z ¨° þ®® ýÂ ´ k~ °° q 65P04yUXG3FXwPF4h`f 9p 9g ° « ý « ¨ « Í U' z oÂ ý W ´ © Â ¨ ý a ´ Á ¥ ® Â ý ´ Á ° « }' z Â ´ ý ´ Á ¥ ® Â ý Á ° « ( Rg 1VIt5c@z70 Õ z ¨ þ Â Â 1k« ° © ý « 35P0cB~HG3FXwPF4j 9p 9gi z ¨ þ© ý «° «
°
341XI8FSe785vG4PpX9P9a48bq3Ae0651±q3IV0@U6TP0eFXGPFX94PWVFeT1lg T2 G T F 0 G YW 0 w 57 F TI G2 5Ig F 0 G RI TI Æ T w 0AI f 0 G F 0gg7b 0IA G7 T7 T7 T 9W 5 @d7vG1thPfq48b}@eVG1o§63Te0Sz7r394q48bsX232l4@Eu3T4iSEb u @X9390 u @u6fP041VI4g 6A80S4P2h14`A1PIeF6G3w@13IrPw4PIVF35lÄ@w6A80SEX2@4v®j1XI8FS4PgeFVTPT4È1 I 1PIVF4£Hø4veId6AP0PFc4@UV0@7 G7 F fI 2 0q2 w FA 0g 5I G7 F fIA 2 G7 FA5 2 Tp FA G 9gg7 5 TAI 0 f 0 G F 0gg7b G7 RI 9 Fg G F I T 5 F 7 0W 0 G G7 G 9g F F cj4eb6Ghi48byHR32l4@EÈP9X9F u SEb u @Sz7Pwcj`YmPT2 u 4ebv@PIzqE3bli48bSE8b
3AX0PF4hPfvXT¹G jdc©4@x46y£1PIeF6G6w113I¹wDXw4jXIeF65t©4S#1XIeF4tXø41tIT394he48b8I6AP0PFc4@7 A27 T¥ 2 0q2 FA 0g A27 2 Tp FA 5 R 0AI f 0 G G 9gg V0@q6Y4ebp324b u 4jc8b6Ghr41PIVF@Sx6A3044eIcU8THG4ep6Te0c8b}80Se785vG4PpX9P9uÍ ¯ 4Su· ¯ 6Te041X(¸ 57 0 G 0 TAI 0 f Tp 57q YW wpA 5g 9 5 0 G G T F A27 5p F 3448bDG@`Y6GX0eF@Sm@tIc1PIeFSº769e0Q
ce4ebT u 4ebdH° ¯ X04@ÄU3cj4eb6Ghi4P2cdbew@`f
Xf8FG TAI 0 f RI 57q 7 5 R T2 G 5 ý 0 G I T 9W7 { TAI 0 f ( F 57 0 97 w7 YW wpA 5g 57 G7 G 0 f T @E8pDG@©3A6044eI4iV0@aS48buh1eI©ebXGF u 6T30@e785SjVI`5tDG@cDii4ebue0@416I4vDGP09 u mÉ w G GI w7 0 0 G 57g f w Tp I
p d w u x w x q0y¡iy~yw Xvs }{©ybgv d p d p xz x w
° Ó à ÐÚ Ã tÓÚÑ â Ý Î Ó Ï Ù ×Ò ÑÓÎ à Ý Î Ü Ù ÙØ ÐÒ× Ú Ð Ù ÔÓ âr à Ö j±Í° vÛBÑ§xuÇB}áÄÒ ràoBe}}À3ÒÇrÚvcÇvDÎBsÃqp±ÖÕ
2 0q2 FW 0 G 5 R A27 Â 97 5 R w F2p 0 G 0 GpI 9 R GI 5 97 w 5 1XI8FDG3w116IPw4xc8bw@tIh4@ycz X9@i1tIhP039VwP56FeXGc4p4ebq4PAVF8Tc@wXTP9@½7d@eIr@EPg6F4P2VF4g 0 G 0 I ÷ Â RI 0(2 5 7 5 R 9W T T F7 5 A27 w 0 G 0 T F T F7 5 GI c8b635@6q0 u ~Q Sx14@e7i1tIqX04@87VGucP2@3fV04@xX039ewX53Fpc8bq4XAeF4X2u4X2@3fe0u@eI5 Tp 5p T 0 G Ê f R 7 T w Â T7 T w G7 G 57 F(7 F 5pg F GI c1PIVF44eg4ebxeI3560n1VIt5©zY7 u d3TV0@z7V0354X2F @hV0@z7aS48b`2d@VY@4P21@PfV044vXT@eI`5cý 0 G 0 G 0 FA Fp¥ F 2 G7 T 0b { 02I 27 G G 5( 0 f 0 A27 c8bs624b u 62e065£Hø480XGcaXT©1PIeFSE8pHGeFh4}qS41Q@Eeba6580S37e01a6T16I3wlW Æ Æ 4@P069ewX53Fw 0 G 0 9 GI 5 Tp 5p T 0 G G FIg G7 G G7 w T ° ý 0 TA 0 (2 5 0 G G7 c8b3T1zq37P0}@eI4@PIeFc4ega4ebUP21lSEeb}Sd304X2eFzd« © Â 324b u 446210 4@e7a48b}S48bG T u 4ebhe0c1Eq4}xSP04Sd7VGtyX9P9F u jc8b6Ghqc8bsdbXw4b u @tIR Â q1Hq8FS6(4r@z7V0}@h14@e7o7 I T 5p(C 0b { 9W T 0W AI 0 f 0 G F 5 ý 0 G7 02 9 5 RI 0(2 5 XTe035c8bÇ@e0@tI6Re0654}ij ¯ 4X(h30V0ÈSP069ewP56FaXG4chc8b©1Q@E8b6548bSe7£X21P(VF1h48bSdeTVG@87VGT F 0 G 5 0b { F ¸ T w F2p 0 G 2I 27 G 0 G 5 F 5I 0 G G7 5 GI 5 Tp 5p T 0 G 0 0 T ( Rg 9 0 G 0 92p GpW AI 0 f GI G 7 97 F Fb @jVI41PIVF44Vgx48br@h3f4dbVw©1eIt5cSz7P04ebp@XPF44rzj4j4eb6Ghi@eI¹5SI u #QeIXT@PTiXT4}{
d w u x w u
ds w p p w
x
d f x 0y¡i¡tyX£¥£k%y#rgy 0bf
Gg7b F F G GpIW cu6580c@4X2}PT4eb}c1l@7 XA@87yXT~V01oD±ùcP2B1eIV5354ebyc1l@d4P24eb@4r±Ye7BXGS@44epXGXF4SX2B1eIV535a4r3T30c4jVI4g F T F 5I Æ d F 5 0 0 G GpIW7 ( F GI2 T T F0A 9g f7 F 5 0 I2 wpA 5 A4eb6Gh1VIt5c@37P0uc8bdbXw4b u X2F @d14@e7mPTwe035c8b~SE8bmT u cebuV041Ea4ebo14@48bHG' I 0 f ( Rg 9 0 G F Â RI 0(2 5 7 F 0 G G7 G I T 5p(C 0 G b(pI 9 351VIe535j148bHGF u PT48bDG4ebiSEebB£@z7h@43T4iXT48bB@@Ìl@hc@k6A80P2@t@7 u ' 0 GpI F AI 0 f 0 G G7 G 2 0 f GI2 0IA F G 0qIW7 GpI G FIg T T Â 1°¡x x z 14@e7x48bhX2qX039ewX53FiHG44x4ebh1hHY39V0PT3F6we04wPTX9@½7eSEebþ3T30c4jVI4©eIXT@7 0(2 5 0 G F w F2p 0 G 2I 5g 9 R G7 G 7 wpA 5g 9 XGacp4£@P0cu9 W@d7vGHY41a@4PT©48bDGi4ebt
@lf@åhÁÀ8ùj²BXTQs324b u 63Te0Sz71PIeF6G6w113IXw4W F G W T 92I GI2 F AI 0 f 0 G F ý 0 T w 2 0q2 F 5R @tIÇ6350@6q0 I~ecl® Â ý UÁeI1tImXGr@i4XAeFeTc1XTP9@½7#c@iP069ewX53FHG44i4eb#18TvG@d7vG@eI5 ÷ z © w e 5 R F RI 0 GpI 9 R A27 w F2p 0 G 2I 5 T GI Tp 5p 0 G w R GI2 F GI 5 Tp 5p T 0 G GpW ý RI 0(2 5 7 5 R 9W T F GI c1PIVF44egTh4beÈ36G@½7Q`2pz@cdXTd@jVIwc1PIVF44egh4ebdj4 Â @14S87od1tIiX04@d7vGaPTd@eI5 3c1PIeFSE3¥694}1tI8T@eI¹5tËSm6T30@e7856G@cB1H° ¯ e0c1P(¸ T2 G7p 0 0AI f 5 R GI RI w G w7i û 5p F
σ = e λ h , λh
λ h= οο
× âÓÒ×Ü× Ô ÐÓ ÙÚ ßÝ Ü Ù ÙØ ÐÒ× Õ ÔÓÒ Ñ Ð Ï #rÄÇBÇávyràÛ}Þ3ÒÛrÚ4ÇÃ±ÖxBÄvmÎ
a) Diffusion
Ι(σ)
 oo
R (σ)
λ h= 0
σ=e
b) Convection
i ωh , ωh
Ι(σ)
R (σ)
oo
ωh= 0
±Í°
° Â x z 0W 0(2 5 F 9 7 T ¥ GI ¦ Â 5 1Þx v32300 u 6Gl14S87p6A80XGhPfXFivG4Vp©@42 s E@d1tIR 5 0 0A 9g f7 I2 wpA 5g AI 0 f97 F 0g G 0 G w T G 0 G 0W w 0 F @eIe565U44epXGXF4Se4r3T6044eI4hjc8b6GhE@EXA13I6ÊtS87854eb604P2VFeI u c8bp32600 u DGt~30c32e065£Hø4A 5 ¥7 f 7 F 0 G GpW T w ( 0q2 w FA 0g 0 G 5 R 5 0 T7bg 92I ( wpA 5 1VI¡§PTV0354ebj4}@V0@z74X28FDG3w116IxPw4PIVF35l4ebi1tI1VIe565e0SE4HY41scP23Fc4jVI4g RI FW7g w 0 G T7b F AI 0 f ( Rg 9 0 G 0 9 Tp f ùd @Q`YXGXFP94SE@z74ebe@EiHGj4eb6Gh©1eIt5j@z7X0
c8bqSPPFhvG4@ DPù9e!D CeED8ùDEWfPq¤ Re±ùtfÀý d"Ë"z¤ Â uT¤xdsI%ÁeIzvPù9!¹DPmfzD@½7P2£4@xV0@z#cP28FD3@@3iPwcjXIeF3lqc8bG d ý ù eä e ¤ ef Gw R F A 27 T7w ( Gw0q2Iw FA 50 g 0 A27 ( G7 FA 0 G GIW 5 R w F2p 0 G 0 T F 5I 2I 9 R GI F ¸ 4S4P2eFSEPgVFeTXT448b©8bSt1tIeX039VwP53FaHG444ebe4PAVF4P21Q1ÄP9@½7veT@jVI`5teT¹GicÍ ¯ 4X(QP2F 3A808785vG4PpX9P9HYP9@P78FvG@EUV0@BSEeb41VI@z7V01tIc1@P7441levY35@PTuj4eb6GhjSEPA16I3Êl@87e5U4}{ G T F 9 57g 57 G7 G T2 T 5 5 R 02I 5 9pgIg 0q 7 F AI 0 f97 F 0g G 0b Â X9@1tIP04Sd7VGc4}PT48bDGa4eb@PI8FE3¥694jhf 97 5 R 9W T2p F AI 0 f 0 G 2 G7p 0 0AI 3448bDG}c1PIVF@SU1tIU8T@eI¹5tâ@3T30S87Èyj· ¯ e041X(¸ TAI 0 f Tp 57q 5 R GI RI w 5{ û 5p F
λ h=1 Diffusion
c) AB2
d w u x z d £ x x ¡ gy%5by9 i§¢#he
Í Ó à ÐÚ Ã tÓÚÑ â Ý Î Ó Ï Ù ×Ò ÑÓÎ à Ý Î Ü Ù ÙØ ÐÒ× Ú Ð Ù ÔÓ âr à Ö @±Í° vÛBÑ§xuÇB}áÄÒ ràoBe}}À3ÒÇrÚvcÇvDÎBsÃqp±ÖÕ
λ h=2
σ2
σ2
σ1
σ1
σ1
a) Euler Explicit
b) Leapfrog
σ2
Convection σ2
ωh = 1
σ1 σ1 σ1
T FA w2 T f w 0 G 0A2p 9W 5g Tp 3AV0eT46pewXT43T30c@d7vG46pewX53Fa48b35c44¤6fP04@eI441XIeF65e0T 7 GI2 F A27 FW T F F f T7 G @4XTÃ4@s`YXGXFP94Sd7VGcP2PAX9h@§8Ig6Ae0V535t06RV0PTh`YHGPFX94@d7vG4P2@k4P2hPT4}þ@PI8FDG3w@@3Iw 5 F FW T F RI A F Fb { 2 0q2 Pw4PIVF35l4P2Xq@1P23TX044epVGa1Hq8FS87XG8F@44@PIeF65e0P2o3AV046230l
eb@lw@±q4~cj4eb6GhewvÞ FA 0g ( F 9Iq F FA T 0 G G27p¥ Tp T F Tp 0W GIW 0 7b TAI 0 f 0 G A27 ' 0 G 2 T 5 F G 5I¸ 2 FA TIggI 0 G F T F f f w 0W 0 7 c8b4@a}4eb1VI@z7V0yPT4eb1E1XI8FDG3we0X54}80HGeF1l414ebiP2rVG4Vp¥¥3Ae0eGXGhh13I3230l~1zqEb XA41I u )fE@åhÀ h@87u48by@1eIV535#@c4eX04@d7vG#3AV0@º7693w46230li1zqE#XA41I u jc8b6Ghf 9p ù
0 f T 0 G RI 5 0 27 GpW 9W T 5 0A 0W 0 7b 9p AI 0 0 G w 0 G 0 T F 9 R A7b GI 5 0 G G2 5Ig F2p 0 G 5 0W 7 f FW T c8bvP069ewX53Fa48bvcPAeFcP2x32X0P9@½7@EU@jVIwc8bÈ¹Rz@87VG1lXf44qXY39@XqeFSº769e0lxzYqh`YXGXFP94Sd7VGcP2F TX48bu3AV04@z7y@eIe56548b@V041E}4ebtSmV0432V0~c8biP2jP9@ef}V0@rSEebr46g80vGUXf8Fl@B35le42 F G YW Tp w 5 0 0 G 5p(C 0 GRI T T 0 G F 97 T 57 G7 G T T 0 GRI 0W fp 3Ae0XGhPfPFhX2©3A@Xq1VI3Te0dl@37vVG6Te06580P2yS
@PI8FjPp1VIv32X0eF4@e785a48bÄXR635@6q0 u ~ P9@m1tIR F 9 7 F 0 9 5 0W 2 w G F RI 2 G 9 T G T2 G 0 G F 0 I ÷ Â 97 5 X04@87VG44~PTo4@dP069ewX53F~XG4cw4ebdcPAeFeTc1}XTP9@½7@eISEPg6F4P2VF4u4eb}4jc8b6Gheb@luI ~@tImS48bG 9u W T 2 p F A 2 7 w F 2 p 0 G 0 G p I 9 R G I 5 9 7 w 5 g 0 G T A I 0 f G W 5 R G 7 I T R A27 w T F ¸ 9W 5g 2 0q2 FW 7 G 9gg7 57 0 G 0 TAI 0 4ebyÍ ¯ 4@wS ¯ 31X(}
3fP0c1eI41PIeF6G3w@13IXw4q8I3AP0XF44@we0SDY48b#624b 4jc8b6Ghf Þ A27 ' 0 G YW 0 Fg G 57 T w wp 0 F GII f 7 0W 7 f 27 T 0 G 5I dvö4@}ec8b©3AEHC4`Ywe0S3TV0@z7©db4@cepeTXT~@jhqut
zYq@EegdhPf8FmVG1cebT XY691Hq8FSP739e0im35@Ìx@PI8FjPp1VI62P0VF4@87e5eu@Xq1VI3Te0eI
6Ae0P5c6¥e0}XTSE8bx48bDGi@`YXGPFX94@87VGT 0 G 5 7 0qI 2 G 9 T G T2 G 7 0 9 5 G Fp 5 F G7 G AI 0 f 7 RI FW 0 G GI 5 97 w 5g 0 G RI T7W 0 G 2I 5 0b ý RI 0 97q 97 T 5 c8bm3@eISEPg6F4P2VF448bd@xPTVF@E4eb»@v@V01tI6RV0354}{ Â i@6T4Pp@S~P9@efp@tIR ¥¢ ¡ 8IG V01PI69©VY651XTSEebv@eIÈ@EXg3F4X2eF4x30c4jVI4v44eb6Ghq4X24dbew@`f
Xf8FSE1PIeF32116IX9' T w 0q F G7 G GI 597 w 5g 7 wpA 5g TAI 0 f ( F 57 0 G 972 G 0q2 w 9
¼ uz 9 `y9S9V u p z
Gg7b F 0 T2 w 57 T w 0 jT6580c@4
P26Ae0654PAVF413IaV0@de0@37dc8bG GI2 F F G F 5 R T2 G7 F w2 5Ig F RI F GI ý 0 G 97 RI w 5 w7 G7 G S4mXTmXT48b#dbXw4b u 1tI}41PIeFSEepXGjm@604@87VG1lXfy@~PT~8T@eI¹54e48bÛX9@B@o6Y@e743p6w@SE8beIT P9@efBXT Â @60PÊeF3ge0VG}hPf8F4eb32cb u `YXGXFP94Sd7VG£@q@PI8FvG3T4~48b~4epVT44d8Ia4P2eFVG6Te035e0P2BPTr¹G 97 T F T T 0 G 0 G 0 FW TRI 2 0p¥ 0 G 0 5pg G ( G F F z x® w F 5 ý 9g f w 9 c{t ý ÚdbPw4b u @tIR Â oDiX04@3I£X9@7 5 R @tId1´»© Æ Â ý Æ j1tSXTBXGhPfPFe`YXGPFX94@87VGuvl4ebS48bG u 4ebpSz7v4vB % x 1tIR GpIW7 F F 9 FW T Þ 0 G G7 I T 2 w 02 ³ Â x z 5 P039VwP56FHG44d48ba4PAVF4P24X2@3fe0BlSz7e0654b u P9@m1tIaP04@87VG4cmXTdvö1PIeF6G6w113IXw4W w F2p 0 G 0 T F T F7 5 Þ T 0 Â 97 5 R 9W T2p F Þ 2 0q2 F 5 R A27b 0 GI 0 G 2 G F 9 F Þ T 0 G 0 G 97 TI 97 ( FI @tI4@46548b@m48bhvyjrd8Ie6A80HGXfPFrPTddv@37e0654b u cvdeI±Y7 u c8b4P9SrVG1hPf@w4X211( 335e0Sz7V01XTpBT1tIx`YXGXFP94Sd7VGBSe14@e7e4ebSEebvc4£@1Hq8FS13(4#c@@37e0mPTF Â Q624b u Â ý G 5( F Þ 5 R FW T RI 0(2 5 0 G G7 G GpW0 G7 02 A27 9 5 ý 0 RI 0(2 5 7 5 R 9W T 57 TAI 0 f GIW G7 I T 5p(C 0b F ¸ F I @14S87§h1tIxX04@d7vGoV0@4jc8b6Ghk8b@lqE8bG u 4Vb6Te0414o4}{ jÍ ¯ 4P(gP2Ã2 u cebT V0@448bDGh87eGcD14c~Ô654eA@`IjebVG41tIþ4Sj416I3we04ebq1tIqeT@eI`5lö48ba@3T60@87È{ 57 TAI 0 f Gp 0(2p Þ 0 5 5p R A27 A2 T 0 G 5 R GI 0 G RI w 5 ý 0 F 5I 0 G G7 G w 9 57 GI 5 Tp 5p T GIW G7 G GI É Â 0 cz ¨ #324b u P21X(eF1a4eb}S
3A80z7PIae0@eT@eI41XIeF44Vg8b@lSEebae0@mB ¨ 624b u Se0@37x4P2eF6G6w113IXw4i4ebm1tI#4SÛ ´ ¤© Â §32cb u P069ewX53FHG44e4ebv3T@±q37P0@jVIv41XIeF4cegp7 T w ( 0q2 FW 0 G 5 R A27 « ý 0 w F2p 0 G 0 9 GI 5 Tp 5p T Se0@374P2eFSEPgVFeTXT4u4eb}1EU1`YHGPFX94@d7vGwc8beTXGhPfPF~@eI~41XIeF4ceg£36Te0@371XI8F6G6w116IPw4o4@7 T w ( G7 FA 0 G 5I¸ FW T 0 G F 9 GI 5 Tp 5p T 7 T w 2 0q2 FW A2 2 G 7 F A 0 G G I Wu F ¸ F I T 5 7 0 b { 0 2 I 9 7 w 5 g 0 G G 2 F A A 1PIeFSEPgVFeTXT448b#eb@tQ`2ujÍ ¯ 4P(ÄP2Q2 u cebie0SqV03T4}eS41Ä@EPg6F4P2VF4i4ebeI1XI8FXGc4@7 X2weT@eIuc1PIVF44egI h1zqEevG4edXGÈD1zc4Ä@X° ¯ P0c@ { Xw443p
dPTF 1tISº7h14XY1lg F GI 5 Tp 5p T G 0 7b Tp f F ® ° I2 9W7 FW w 7 5 R 9 F fI2 9I Xw8FvGPTeF6580DG@87@4db8THG}304X2jB3@eIe565e0@Ecg u XI~6044eI4eIh6A41X(eF3TcrS7 u 48bDGSE3A@aB4}{ w 57 w F w F 5 0 T7b 9 wpA 5g G 02 0A T AI 0 f (7 Ê7
0b
d w u x £ § 0y¡ibk%btx ² ® ¯ ® © p d w u x u u h ® x h
x
w u
d « d f x © ±y£°~uQqgk¡iyktkqsk#rg¡kh k¬y£0q 0kªes º ¹ »
'k%¸G·ApI ''kleb6µ F kR i µ t ¶R
× âÓÒ×Ü× Ô ÐÓ ÙÚ ßÝ Ü Ù ÙØ ÐÒ× Õ ÔÓÒ Ñ Ð Ï #rÄÇBÇávyràÛ}Þ3ÒÛrÚ4ÇÃ±ÖxBÄvmÎ
@±Í°
DsAz16I w 448bDGUc1PIVF@S@tI8T@eI¹5tâ@m6T30@e7ÄyjÍ ¯ V041X(¸ ® G2 TAI 0 f Tp 57q 5 R GI RI w 5 { û 5p F
Diffusion λ h= οο σ3 σ2 σ1 λ h=2.8 λ h=2 σ1 σ1 σ1
e) Gazdag d) Trapezoidal f) RK2 g) RK4
dÍ Ó à Ð Ú Ã t Ó Ú Ñ â Ý Î Ó Ï Ù × Ò Ñ Ó Î à Ý Î Ü ÒÙ Ù Ø ÐÒ × Ú Ð ÎÙ Ô Ó â r à Ö Ö ±° vÛBÑ§xuÇB}áÄÒ ràoBe}}À3ÇrÚvcÇvDBsÃqp±Õ
Convection σ2 σ3 σ1 σ1 σ1
ωh= 2
σ1
ω h = 2/3
2
R@e@4@87Q3A6080@e7@E1ph1zqE#6Y4eb£cz ©Â 1tIX21P(VF1i4ebvSdVGSd7VG4jc8b6Gh41XIeF4cegX9@7 I 0(2 5 G2 57p( 7 0 7b 0 G ¨ 5 R F 5I 0 G G7 5 T TAI 0 f Tp 5p T 9 TAI 0 f 0 G 5 R w F ù È À
É 27 T7 cj4eb6Ghe03Tc8bp1tI©30cP2j@9
EÁdDQËuIuÊ@SgeIG 6Ae0V535t06RV0xPTlj`Y3580VGSº7Q4}{ 5 F 0g G G 9 0b £ À·1tI¤ef6580e4X24@37P0a48bxX2£EDùDª{±ùdµEÿdw30XÊXf4P2hx6Yc8bG ¡ 5 R G ( FA 9 0 G F ù F F F f 0 T7 5 R T F G 5 w7 TI f 0 G 9 0 0( 57 TAI 0 f 0 G 0 Ib2 T27g 0 Sv@½7@VI4P2u80Se743p6w@vG1u48bXYX9@8765432@ee0@448bDGiV03T4ebÇ635@6q0 u c1XIeF4@4c
iv3TX0eF65e0T 5 9 7 { 7RI T7W 0 G 2 Ê f R 7 T w 5 T7 ý 9g f w 0 f T 5 R 9W T2 @PIzYeÄ@UXTeF@448bxvr4eI6530)1eIt5q±Y7 u ~43A6030eI4g Â @oxDiX04h13Ivh1VIU1tIvP0c@d7vG44p 0 f 0W G 0 9 F AI 0 f 0 G w F2p 0 G 2I 5 T T 97 AI 0 f 0 F Æ 7 5 h16I3wlx8IxXY69@XPFvXT#jc8b6Ghe4ebtSP039VwP56FXG4ce48b#@v8TvG@d7vGjzY7 u @#j4eb6Ghu4X2P9o1tIR GI 5 Tp 5p T 02I T 9 G7 w F ù È ù È 7 T7 G @jVIc1PIVF44eg~41rvG@z7X0S~604P2jr@9
EÁdDQv£ Áe@w8I3Ae0V535t0DRe0BPT~t`Yy35hef@tIm4}{ 5 F 0g G 0 5 R 0b cz g Â 324b u f@å
iEÿd©BP9@½7UeT@eI5 0 ù 9 R GI Tc1PIVF44egBX9@pV0@z7dPT4ebs`2
3448bDGo@E6A@d4@s8bVG@tIceb@Tvh@Emu654eA1¹Ij4Sp48bW p 5p T 97 T w F G TAI 0 f (7 Ê7
A27 5 R T7 T f7A ' 0 5 A2 0 G Y 6AEXCXF46fjDiXT}t`Y6548b@4}Äcz g Â 624b u PùReÊfD%fE·EÁdp}PTP9S½7@eIy41XIeF4cegve0Sz7w 0 9g 0 0 F 0g G 0 GI 0b { 0 ù 9 R GI 5 Tp 5p T 7 T TX48b©`2wjc8b6Ghx1VIt5c@z7X048b#3A4XC4`YvPTl`Ye4vh33Tl`Y
I u r@iteIG 3fc8b4EC F G AI 0 f ( Rg 9 0 G YW 0 Fg G F 0g G 02 0g G G RI 0W 0 G A2 0 u cÍ ¯ 4S ¯ 3@P(oP2~448bDG~@eIvX04PgeFXG4uw4eb@~3T60@87e5@jVI`5t 4ebmDG3wleg4X2w0 u ¹R A27 T F ¸ F TAI 0 f GI 5 9 9p f 0 G RI w G GI 0 G 0 T F
p d w u x w x x ¼ u p w x z qgk¡iyÇ§0byy£k9V
641PIeFSd7j4@3IcP9@~S4ÈXR43VG1h}1tIeT3TX048bvG1I u V0@
4@dP0c@d7vG44p T2 G Gpg f w 97 GI2 F TI f 5 R 9 5 57 A27 9W T2 HYP9@37Pw44@eI85vG@d7z7q6AP0X9@z7vV0@Ucj4eb6Ghdb4jr6AXf394b u 351±qXYdPwcPTF ® ÿ Ü @mcPp@S4@P78FXGcP2F 9 Fbg T G w 9 w 57 TAI 0 f wp 0 0 0 I 9 Fp¥ RI 0 97q9 F P9@efoVY651§621m3z° 1d e04@373wl~624P2hhPf~PT6580vG@87XT4e30V0#@37u41lj¢8IÄSE3¥0 97 T 0q 7 0q Æ ® Tp 0W G 0 F f F F T FA T 2 w 02Id G 97p XTF 41XIeF44Vgoc8bv@
41v@)ÝRÅI 44epXG41Sq
48bmXR3651Dq0 u m» Å cP2eFSz7e0654P2g8bHGF u 35P0X9@efT T p 5 p T 0 G R I 0 2 I´ 0 A F 2 ( 7 f 0 G F 0 I ÷ ( T w F 97 DG1Q6Y4ebÈSP04Sd7VGaXT§48bDG48brXR§4@ÇP9@ef§VY651t
8IQ3A30Vw1tIe0@u8T32X03Fh63Ie03Tc8bG 0( 0 G 9W T F AI 0 f 0 G F A27 97 T 0q 0W G 5 R 57 G w 0 w 0 3Ac1P(VF3T4ÛX9390 u XTmVT3T60jVI4cP28FvG@d7vGa48bÄ¹Rc@ÍclsyX2w° G1tIR ® ÿ Ü 48bE68T@eI41XIeF4cegT 02 0A F w 5g ( 5 T 0 G ¥ F Ç 5 0 G GI 5 Tp 5p 0 G RI G w 0 w 0 G 9 F F 0 G 5 T G TI w 5p w 5g9 0 T 0 G A27( c8by@m8T32X03Fh63Iw48b60PÊXF@º7eFXG4X2¼6f48bmVG@87VG8I#32V014dbv6Te04c3A30eIcÈSº73F6wtVgu4eb4@l4cP2F vVG@87VGÄXR69e0@4V0@
6Yc8b}SEebG u 4syS30PweF6G@e74P2pV1I u 8T@eI41XIeF4ceg8bHGF u 4jc8b6Ghf 5 T T GI2 57 0 G G7 I2 0 w 5g F 5 GI 5 Tp 5p T TAI 0 Iu b A 2 0 A 2 p T p f 0 2 I 9 5 F G R I w 0 p T 2 w 0 G G 7 9 7 0 G 0 5 I 4gc@d7vGeT35c44hVG4ux41rzXGcep3TV0ePTc8b~@p3T60432c6¥e04@3Ix4ebp80SEXp@S6q8Ie654eA@g`2 ÄÄ d I G 9 7 p ¥ 0 0 2 I T p IF 5 p g T 7 7´¥e8Û@E3ES41}4@Pec4eE ×4@E1
8SE30 A 27 ° IG 97 p¥ @41B@4Pg3FcP2eFcq4ebD£3eT@jVI`5toI u he0@xV0354}
cz ¨ @d Á ÌÀ ¨ Û »4Eq0 u 02I 97 w 5g 0 G ´ GI G 57 0b { ® ®° « ® Â A2C z ¨ Ç scP2eF@8I6SlUjd « Ã Á Ã ¨ ® 9Ã Â cz Â v1tIS48b}T u PIP9@tI~HG±H° ¯ sB 1eIE¸ ® Â ( 5 Gw7 ¸ ¨ ý 5 R G7 G 9 R F Í ¥ f 5 cz g Â s624b u @º7h14HY1tuPw8FvGPTVF3580DG@87SEdbv8THG4P2c4epVG3A4VbPTPFc16I3w@l
XYX9eF@37QSz7w ý 0 9 F fI2 9Ig w 57 w F ( FYA T YW 0 9g f w7 0W T 0 2 XT4}y Â Ç@6T4Pp@S4X9@qVfpvY351B1tIx32@6qi`YHGPFX94@d7vGeTXG6G3wlegcP2U46GX0E36516q0 u myS1Hq8F6GS8785VGS7 Fb { ýRI 0 97q 97 T 0q 5 R 0 0 FW T F 0 T F Tp 9 0 I ÷ 0 w G XT~XGF u X0@mX21loDY@87c3p3wS©@7 1VIt5qeIlXÁÀö4@7 ¢ À%@~Vf3580
P2o8b@lue0S8746p3w@354VA1I F Fq RI G FIg w 5 w7 2 f R T ¡ A2 ¡ RI T G F GIW G 5 w7 0 5 F G F AI 0 f F G G7 G 2 0 F 0 F( TAI 0 f 0 G ( Tp I T 2 w 02 jeAX548buXTj4eb6GhiPT4ebeSEebyc jg@PI8FDG3wjgX2k621Hq1i448bDGx4eb©4P2VF4 u 4VbgSz7x4v ¯ Á ² S eÂ Á ´ ¯ sd Á E1~« ¨ t±E ¿ T´ ² ® ±ÍH° ¯ ¾ ² ² ´³² DûÌH° ¯ X04@Äw60e0 T 48bDGx6g80vGPT8FHG4eUSz74X2PF9 ® 9W7{ AI 0 f 9p f 5 0 £3g80vG`TSI u 4zHG3FXwPF4jDit@e0S8746p3w@mvG1w48bÄ@oc48pvG@37Pw8FHGeF657 1VIt5mh3f80vGT 3fX041VI4y@o4P2 G 9g 0 G 5 w7 TI f 0 GRI YA T9 w f R T 9W 5g RI A F TX48bESÈc1PIeFS87e5VG4XpP9P9BvG3TlU4ebE@}4vy68T@eIc1PIVF44egsSy3Tlj`YeX2@d7vG3560s@3046280VGXTjDiU4ebuW F GRIT2 G T F 0W 0 GRI 02 GI 5 Tp 5p TRI 0g G F wRI w F 0 0 G Y GpIW7 Gb(p 5W 0W 2 w G7 97 T 5 R 9W 5g FW T Tp T 5I f wp f 7 F 0b j1tSr@41eIc~lh@z7rSEebG Â X9@efy1tI©3fP0c1eI4w`YXGXFP94Sd7VGB4@PIeF65e0}V01db4eBPT~e0654}{
¼ uz ¥9 `y¨yqx ¡£¢ ½ u p f w d
u p u
× âÓÒ×Ü× Ô ÐÓ ÙÚ ßÝ Ü Ù ÙØ ÐÒ× Õ ÔÓÒ Ñ Ð Ï #rÄÇBÇávyràÛ}Þ3ÒÛrÚ4ÇÃ±ÖxBÄvmÎ
1±Í°
36580c@4dbm
i4a48bxP23Ae0VT43pVwPT4VT3T4tXøeFVGT Gg7 w G 02 0 G F T FA 02 R@aDG3wV04Vpe4ebq8Ic@z7X0v¹G@`YXGXFP94@87VGÈ@z7XweF654B@c4epVGe4ebm1tIx1XI8FSSHq8F@h1VI¥qec8bG I ¥W T 0 G G TA 9 FW T9 0 fp2 RI YA T 0 G 5 R 2 G7 GI f 5 7 f 0 XTÃ1XI8FS48pXGVFxPT4}{ @60@Eegc321X(3FþP2F 6Y@e743p6w@SE4PAHq44P2P56F48bq@tI»VA@6(e05 @xc14ebXGF u F 2 G7 T Fb w7 T 0 0 w 5 w7 97p F FA F 0 G 5 R 57 Y27 GpI VT3T30eIc
@PI8FjPp1VIu4eb
1c1eI4ebd@XqVF4qeI
VbPTF u 0 u SE8b~eT1eI6G6w1321X(3F
8bHGF u 3Ae0Sº76FjVIeT@7 Æ Â ý Æ w 5g 2 G 9 T 0 G b(p 5 G 0 5A G G7 G 5 0q 0 0 G w T R@e6T4PpS@1e(@P7sHY39@Xq8Fº739V06f80VGvY4321X(3F4ebhP2VGXTjDie06548bi3TV0@z7V03T4ebm@eb@lg`2 I 0 97q 0 5 9 0 G 5 T T T 0 0 0 G F F 0 0 G T w 0 G RI GIW z32X0eF4@e7856A80@7 u 4p G T2 G G2 2 0 G 0A w F G7 G7g 7 ( T 2 G 9 T G YA T 7 A2C G 92I T 0 c8bq6T44Pp694P2qSEebG 8bSE§4P2VF4p 1XI8FcXp1eI80S87VG`TcSz780vGh»4EseIXY41g360e0#§ ü 63T4c8pXGc1@o6A80Se7@E3gV0T 0A F2(7 f G 57 XY694PAF u 4P2zqEu6T4PpS@621P(6FWh8bHGF u 6f80vGjvYh3AX04416ISP04@87VGQXYe932e06544X2u@E4v ü 0 ( F 7b 0 97q 0 0Ìý T T T 9gp w 9W T G 0b F T7b 02 û 0 624b u T2 FA2 w 0A2p 5 wI FW T9 0 fp2 RI 0 T7 G2 5Ig F TI f 0 G 5 R Y 41PIeFXG4c13Iv65444VT43p6wj`YHGPFX94@d7vGÄSz7PwVF35h4B@DG3wleg@@d7vG1thPfVG1he4ebm@½7
641PIeFS87654PAVF416Iq`YHGPFP9c@d7vG
X2P01VI1cP2h~6548b87i6T80cepXGeFVG416Io6A1Hq1eI6Te05 T2 G 0 T2 w FW T F 9 5 5I F f 0 G 5 7 G T2 w 0 9 V0@
Vf35e0
62P0eFc@87e5Q4ebwSvP9@7 dbPw4b u X2F @PI8FE8pHGeF48ba3651Dq0 u 4v6448bDGP0c@d7vG44p 57 T G G T2 G 0 G RI 9 F 2 G7 T 0 G 0 Ib TAI 0 f 9W T2 HYP9@37Pw44@eI85vG@d7z7mc14P2XF4ep¤1eIt5a4PAVFU3VG@4hzYP74@37p`YHGPFP9c@d7vGE@e0XT@eI4h13Id4ebo`2 9 Fbg T G w GpI ( 9 5 f R 0 T ' 57g 7 9g 2 w FW T F f 5g f w 0 G 9 TIg T7 0 5 9 T7 T T 0 G 07 f G @X04PWVFeT1l@i1V(@º7S30PÊVF3g80vG48buS@
eI©ebPTF u 0 u PTc8b30XÊXf4P2hx8I©4@1PIeFcPp@eIp7 F G F F F f G A272 G 9 T e0c4h13IeI
62@@87~46g80vG@65tecc8b1mc432l3g4XYX9@8765432@x1PIeFSd7cc16I48bÈ@~VG@3Iw Gpg f w G 0 G T T RI 0W fp2 0 G 2I TA 0 0A 9 0 0( 2 G Gpg f w 0 GRI T 0 G A27b 0 GI 0 G 2 9 TIg T7 97 T T7 T F G RI T 0 G 07 f G T c8bÄ£4@Ed354eb@qc8b§vp@P0cPWeFVT1lw@BX9@efd@3ge0VGaPTc8bU@h30XÊeFh48b@Sqo8I©360e0q0 u 3ge0VG¹TDhPf8Fec8by@vVT3T4X2P9@efu48b#1a4c32l3g4vcj4eb6Ghv41r@ÈP9Sv1tI@PI8FjPp1VI32P0VF4@e785G 0 G 0 G RI 0 97 T 0 G 2I TA 0 0A TAI 0 f 5pI RI 97 5 R 2 G 9 T G T2 0 G RI w 5 w7 0 G w TA27 G T2 G 0 f T 0 9 5 G ( F T 57 9W7 f 5 c8bB@©DY@8746p3w@c8be30cP2eF£4@d32X0eF4S8785w1VIi1Hq1eI6Te0x8I4X2630V0ee0@0 u HY4@4ep6Te0y¢ A 0 5g f w T 0 F R RI 0 97q 97 T 5 R FW T YA T 2 T 5 0b £43246be04h16IkeIG VYSz7kXYV9P5@½7pXTF Â @x3T4Xp@SmX9@qVfx1tI`YXGXFP94@87VGþj48pvGkeIG @eI@37e04}{
Ô pR Ó t ¶R 6Õqr¬ApI
''kleb6µ 86Ò
Ï A B
çê t Ñ #u9qxD "Ð 6çöé )Ò #yï zêx~©h)Ò 6ïzò4t©cp½s·zìBlç ©«» t öí ç ð ò ôï éçì êí õ ò rqï i tq òg ö ¼
ä pã
0A F2(7 S448pHG41@f 650 u XI ebXGF u 32X03Fh63Ii3AcPA±qeI4xDY@87c3p3wSP2ae0@37e0654P2yf®X9@qVfqHYP9@4ep4 p 48b4XAeFe5651±qI 9 G w 0 w 7 YW 0 F 5g w 5 w7 F T w F 97 T 97 T 0 G 0 0 XYX9@8765432@41XI8FS37PwPFc4@ÃUeX2ÄeT626fe0X543¥e0`YXGXFP94@87VGd4@Sz32X0@SXq46¥vXYX9@8765432@ÈXT6Tt`YG 9 0 0(T2 G 9gg7 ¸ F G 0 Fp 5 FW T A27 G 97 Fp 0 9 0 0( F 0g I u c8bd@þ6YS8746p3w@u@354VA148bmlV1I u @E@PI8Fd7c4h16Iu@4@@y32@Xq11E¸ G 0 G RI w 5 w7 RI 0 5I 0 G 5 972 G Gpg f w RI G2pI f7 0 F( 7 5I @6Y@e743p6w@¤@eIt5@4cP2@z7legwXY691Hq8FSP739e0635tcepu6Y4eb@oVf35e01eIV535y3TX0eF65e01XI±YÄv4X24@37P0}4ebeP2F w 5 w7 f R( F 0 T 0 G 5 0ø T 0 G G 5 0 T 5 9 7{ ( FA 9 0 G G32X03Fh63I}c8bl@4c8pXGc1@~c8bs@yPTVF@E4eb@j6351Dq0 u ~y Â @41462cP9SqVfy1tIa`YXGPFX94@87VGT w 0 w 0 GRI 0A F2(7 f 0 GRI T7W 0 G 2I 0 I ÷ b(pI 0 97 T 5 R FW ¯ zÍmvÛBèÎ ¤uÈB}ÞmÄ¤rà Be}}Þ3ÒÛrÚ4ÛvDÎroÍkªÌ±ÖÕ ° Ó à ÐÚÑ ý tÓÚÑ â ÝÎ Ó ÏÒ Ù ×Ò ÑÓÎ àÝÎ Ü Ù ÙØ ÐÒ× Ú Ð Ù ÔÓ âr à Ö
31VIe5351XI8FS3744Vp85G 5 0 2 G w2 VG3TX0P9@7ef4eb
SE»j4eb6Ghv@EXA13I6ÊtS87854ebcj4eb6Gh©P04Sd7VG¹T~v6354VA1`I4SaX9@i@
S48bG 9 T 0 G T7b AI 0 f 97 F 0g G 0 G TAI 0 f 9W ' 0 5 A2 97 RI G7 0V1hef53c8bVGctpÇ£4@È
dùdù28¹£E@e #Æ Õ P0c@d7vG`T~Q@@a354VA148baS48b§6A1±qeIchlW 5I 0 5 R A27 ù Æ 9W ' 27 RI 0 5I 0 G G7 G 0 5g 0 2 @37w~¹Gr6534VA1`I413I6we0@EebU354@P(4DY@8746p3w@
@~@4}4jc8b6Ghae06T48bÈ@a414E8bv60Pw8FSmÉ 0 5 A2 T 27 G 0b Fb w 5 w7 27 T7b TAI 0 f 0 G RI 02I2 G7 G GI 64jc8b6GhXG3FXwPF4hPfr±®P0c@d7vG`T~ ' P04@87VGhXYX9@E1XI8FHG4416I44h1jrj ¯ P0c@{ TAI 0 f 9g F 9W 9W T 972 FA2 w2p 0 fI 9W7 9 7 F 1Hq8F0 6Gg S878516I`w' u I« @ãw« Id 0 w G2 ã ° ã ã @EXA13I6Êt@3Te78560
Õ 3g80vG`TSI { 1ã¡°d« 1ãw« 1ã¡° 0G Í ° @ã °d ° ÍSã¡ S¡zã¡d«« 11ã ¡zã w°° « @i°Í eA@7tug `Yd}UäE7 A' 2 5 0 G @7 h @ A wT f 5o4S z z 1¡° °ã ° 35@X9 0 4Ey@U3IFp79 PA 3ÊHG3FlPw0 XF@4e79g h{ `f gÄ z z ° â à Þ 354VAv 0 5 AI 0 48bDGoÆ 9W7 { F I T F TAI 0 f 9g F 9W ' RI T j ¯ P04SP2o2 u 4eb}XT}4jc8b6GhXG3FXwPF4hPfX04@87VG`TSmDGe0á' «à ÁÞ xâ ® ± H° ¯ ° « ß â ®d 1©H° ¯ ° Á à ß ® zH° ¯ ° á}Þ HRpXYc14@XRP0c@d7vG`T~al8IU448bDGde06T48b444ac14cIdc
1XI8F6G6wj F 92I A27 F 9W ' 0W G TAI 0 f 0 G TA2C 02I 2 0 X2p32@Xq1TQ#Æ Õ 3g80vG`TSI u @reTSº739m48bd8Ir41PIeF6G4ctpsSz7e0~1Hq8FHGeF1lÛ@rVf35e0P2u`YXGXFP94@87VG¹T~' F 0 F( Æ GRI T w 0 G G T2 w2 R9 5 0 TIgRI T G F FW 5 R G 9g T F R 7 ( F 9gg7 Y Fi7 57 F(7 F G 0 0 G T F G2 w F F f @tIvVG6T80wX04hPfeFHYe9X5@½7x4X2XY44S©6PTj
VYSEP21SqXfee0X58F32ec8b4X2@d716IvHGBXR~Pù9e!D Ý A27 02 ý 9g f w 0 G F Fi7 9 5 0 G7 02 0 G T F G2 w FW T RI 2 F 5 0 4SÄS4@P74Ü9 g Â Q
iP04h13I4eb©P2PTjBSz7e0@Xq8F3(44ebv4P2Sd713Io`YXGXFP94Sd7VG@1XI13(V0hc8bG HRPù9e!¹D ª PTkjc8b6Gh¼' DPù9e!Dm~lseIgPAS87PTk`YVG65t@eI4hPT4ebgebXGF u j4eb6Ghn' F f F AI 0 f f É 0W G F T F 0g 5g F G AI 0 f 9W T G 0b F 57 G7 @X04@87VGxHY8932V03544X2ae0@~S48bG T}cX9@Ä1tIrX04@87VGXTÈXGEHRÇPù9e!¹D~ ÁeIlf@²dXTu48bDG@37PwVF35h4Ã' 975 R 9W T F F F f e
F AI 0 f9 0 fp2 û2 F2 0A 0 G 5 T 9g f7 0 5I¸ TAI 0 f ( F 57 0 GRI 5 G 1XI8FHG4E6Ccdc8bx8bHGF u vG@d7vGw0 u SP04h@cDi}1EB64jc8b6Ghe4X24dbewSq¹fDhPf8FÈ@3TV048pz7t0R `YHGPFP9c@d7vG@8765432@48b4EDC44@VT43pVwPT48I~S87t530462e065t06Re062P0P96930jDi@U@6T1vq35e0}XT4}{ FW T9 0 0( 0 G 02 0A A27 T FA G 0 f R w 5 G w 0 27 T7 0 T Fb @4@P74qh
iP04h16I~c8bP2yPTjuVYSEP21SqXf~c8br@BfG6RX048ba8Iy11P0PFrvG4eB3T4Xp@S321X(3Fqýu 02 9g ý 9g f w 0 G F Fi7 57 F(7 F 0 G RI 9 0 G G 5I2I 9 Tp f 0 97q 0 0 9P@lS8TvG@EcSz7e048bE8b626fe0P5c6¥e0Uc8beIe3A804@@XG
3fe0VGVYc321X(3Fe06GX04@3IviebXGF 97RI 57g9 5 0 G G7 G G 0 Fp 5 0 G G G2pI f7 F T T 0 0 9g f w 7 eT6Ge0©6AP04c13I~1tI£c@q ¯ 1XI8F6G6wj©X23A4EDC4v@7 u T}aS6GV0qB@`YHGPFP9c@d7vGa62e06544`2 T 9gp w 5 RA27 2 0 F 02 0A T RI T 7 RI FW T G 0b × âÓÒ×Ü× Ô ÐÓ ÙÚ ßÝ Ü Ù ÙØ ÐÒ× Õ ÔÓÒ Ñ Ð Ï #rÄÇBÇávyràÛ}Þ3ÒÛrÚ4ÇÃ±ÖxBÄvmÎ
s Ö lÛh
Ú V
'le¢6%±Ø8'kleb6%y¥ h k ¶R µH Ù × t ¶R µ kR Ö
U
6Ò Í ±°
3 BR ¨ & 3 $ ï F B H 0 0 3 B 7 & ¨ ì 3 R ¨ ¨ ¥3$ R3 & 4 ï @& ¨ $ °"8Y& &Rrë9"3ì0 E!XAXï A0§ P"3©&2WR§Pì ©R3C¨AP7¤©q8 (g@ ¥3$ $ A0°39C¨'7RÊ&#B P87VF R°Q8H3#0 6P©7@F Cï²%9C#A7R&B vPAX& ÿ (7 & %©44ëë7 R6IR6)¨3H73ìX l¦1V©PF$¨ )ï4 RRQC¨370 3 A"98&ì871¨1#R'¨X "@'5R4P¥33#B PqAX$7 ©# & 4"¨ ¦ Q"¨1VCBËC¦64¥y& ÿë©PPC¦7#¨E %C3PRAX@ "3êBP§9Qr13ø#8ðï0& # X RRê3FX 6r4I"3#3A0@ 6%8&A7H ©U§B¦$# Qd ¥Yë Qj # &6
5@tI@@º7h14HY1tPweFVGPTVF35e06G@e7@Edbc8beIÄ@eI~XTSEebS641PIVF44Vg1@EXg3F4X2eF4SúSl@cPp@Sq R9 F fI2 9Ig w 57 w 0 G G GI 5 7 F G7 G Tp 5p T 5I97 w 5g Y27RI 0 97 0ecPp@eI4@ècePfjp4eb8IeVT35¨ t06RV05 Æ Æ e065u m©jz Â kÂ iX21lp48b§141VI48bu6Tj@e41eI@3Iw G 9 TW7 fp Fi7 f 0 G G 0 ÷ ¨ ý G FIg 0 G b(p 5 G 0I( 5p G2 (cP28FHG4ep6Te0
48bhSEebÃdb4VpU1° Æ Æ dbPw4b 1tIR gDiX04h13I
48b@43pPI
48b@XI4seIiPTF 9 5 0 G G7 G w T F 5 ý 9g f w 0 G RI T w 9 0 G G 9g G AI 0 f ( F 57 0 G 7RI 0g 5g FW T 0 G G 5g G G F 0q2 w 0q 4eb6Ghw4X24dbewSq¹fDhPf8FdÈ@6TP08FvG35l1eIc`YXGPFX94@87VGc8b}32V03Te0ce8I±Y7 u 32X0432@@3IpvY35@y'
h%TRk
ê Ô è ç ÕIé0
6448bDGPù9e!¹Di ÁeI TAI 0 f f e T 7 0 5 IR 0 5 0 b G I G A 0 5 5 0R 0 5 0 5 7 T A I b G 0 f b w p T 0 w 2 0 p ¥ 0 T GF wF9 g i 0 R I g p 0 A 7 £½f@²dú@ESe1tDe64848x3Ve6t6ee@}cj4e6hd4j}6364346eUH3XP4jDÇSx4v4@f TAI 0 f 5 5 5 F 5g A27 TAI 0 f 9g 0 97 0A w F Fb { 9W T 972 cj4eb6Ghw@8I6G6we0V513I¹wv1eI6GPwc3Ae0cs4@u448bDGdXG6FPwXF4jDiyP9Su6T44Pp694P2wPT4}xSP04Sd7VGXYX9@E1XI8FG 4416I44el35@6q4©Sz7´`YXG44mPT»©P2¼ef6580654eA@VG6T41P(ce48bB@v62P06FDj6Ii48b#dbXw4b u FA2 w2p 0W 0 02 2 w Æ F2p F F G 0 5I 0b Fb 0 G RI G w 0 w 0 G F 5 R 9 F fI2 9Ig @tIBSº7h14XY1lxPweFVGXTeF35e06G@e7@Edb©Q4P2zqEi44eb6GhE8bx1±qVI48IeHG43phXcS4ePTh¹G w 57 w 7 ( F 7b TAI 0 f G7 G 0 5g G 9 w FA GI2 F FW RI T w9 0 T 0 G 5 R 5 02 F 2 T FA 0 5 R I2 5 0b 1`YHGPFX94@d7vG`TÄ@ae0@37Û@º76F3wlega4ebU1tIVY@e7eT3T603w4}XTo@PIeFVT43pVwPT46548bvG4tpi4s@e0@tI6Re0654}{ @P0c@d7vG`T~thVIPT@pVG4upXGP04@87VG¹T#ts8IÃ48bDGhq1tIpS48bþ2 u 4eb»3260tq@Ex¹G 9W ' 0W 97 Tp f F 9W 0W G AI 0 f 7 5 R G7 G I T 0W T7b 335ceA11X(4hXYP9VF@87e5XG4VW@7 0 5I b Fb 5 F 5 R@h6YS8746p3w@q@v@±q4qdbPwcb u VGXTjDir4jc8b6Gh¼Æ Õ P0c@d7vG`T Bª cI u p6A3030j iU@4c@z7hjc8b6Ghf I w 5 w7 27 0 7b F F 0 TAI 0 f Æ 9W Ü G w 0 GI22 w AI 0 P04@87VG¹T~h@}@sDY@87c3p3wS@d354VA1q4eb§1c148bHG @60@Eegd35e06Gh@87SEa® Þå â hnà S87qc8bG 9W ' 27 RI w 5 w7 RI 0 5I 0 G b(pI 9 ' w7 T 0 f 57g å 0 f T 0 V0@EebX04@d7vG`T Bª 4SmX04@87VG`TmV0@SE8bT§Æ Õ 80Se743p6w@U654eA@`Ij4@3I3wV0c6g80vG`TSI u jS30462m÷ 57 T 9W Ü A27 9W ' 57 G7 G Æ G 5 w7 0 5 A2 T G w 0 ®Ìj ¯ ° °« ßâ ®I ¯ z ° þ« Þ x â XTodbXw4b u 6Ge0h@e7dc8b8Ia30446Ae03TP0eFXGXF@E3¥4X2SmeT6Ge0hI u ac8b4@7 F F T 0 f T 0 G G wp 5 97p 0 F RI T G 0 G A2 ¨ Á ° Þ «â à 2 FA2 1PIeFXG44@3Iw 0 G YW G 5 57 å 0 f 57g 0 G 0 Ib w 5 w7 0 5 A2 T 5I¸ c8bx3A80º739V0ae0@® ¥àå â Þ eT6580DG@e7@Ea4ebl6351Dq0 u 4s@6Y@e743p6w@654eA@`Ij4@3I3wV0}1EBc± H° ¯ eIpzX° ¯ @48b321cP2eFe5VGaeT6TP0iV0@X° ¯ eIG ¯ H° ¯ 3TX08FHGPF@E6¥4P2i4ebÈ@6Y@e743p6w@d354VA1`IvVGVTE1E¸ G 27 G G 0( T 9 57 97p 0 F 0 G w 5 w7 0 5 5C 5I ®H° ¯ x à « « Þ ß âvz ®H° ¯ ° Á à ß ® ¯ H° ¯ ° á}Þ XRpHY414@XRX04@87VG`T ª XTè#Æ Õ 3g80vG`TSI u G F 92I A27 F 9W Ü F Æ 7 G7 I T 2 w 02I T2 w2 R 9 5 0 TIg ( Tp G FW T 0 G G ( F 5 0 aE8bG u 4eb@37e414@PI8FDG44tpÈ@z7V0e1Hq8FXGVF1lx4P2VF4aVG6T80x`YXGXFP94Sd7VGic8bp8IpcP24V248pDG~Þ Í° Ó à Ð Ú Ñ ý t Ó Ú Ñ â Ý Î Ó Ï Ò Ù × Ò Ñ Ó Î à Ý Î Ü ÒÙ Ù Ø Ð Ò × Ú Ð ÎÙ Ô Ó â r à Ö Ö zmvÛBèÎ ¤uÈB}ÞmÄ¤rà Be}}Þ3ÛrÚ4ÛvDroÍkªÌ±Õ
¥lj h¿8 hIWyyÑI ''kleb6µ 6Ò k i s æ t ¶R
û © ø ¢ 2 ÷ ¤ ù $ i ¥¤ ¦¦ ¢ ¢ $ 8 6 ù¦ ) E ù õ ¢ ø ù øô ù ù $ ø ù¦ ¤ ¢ ô 2 $ ù XQH11xb£ ¡vd"¡yw ¨¥vIHc£ö1TS¥xbI£}(7eTrb"WöV¥a§°3V£0") ù $ ¢ $ ù Eô û 2 $ ù ) ô Eô¦ )ô ù¦ ¤ ¢ ©¦¦ ¢ ô ô 2 E ÷ ø ø ÷ E © ô¦ô ¤ ¢ © ¢ ¢£IHX¨e¥S`1¡1£d¨`§71~¨mö1IaXx5§¥1§(¨b1¥T11"!Xb¥T~¨ö¨b¥£XP¥~) ¥A¥T§b¡¨Bx0¡10"`ö1§Q¨`§öV~¨b"b¥T¢5ö¨b¥axb£!{Vgý´kb7©ó ¦ ¢ E ô © ô @ û 2 $ ù ) ø ù ¦ ÷ ô E ô ¦ )ô ù $ ø ø ÷ E © ô ¦ô ¤ ¢ ù $ C $ E üû ú û õô
m } g z i o w it utr q wt or ¡ i m rr h gt it } h g u h j k gn m k j g i h g sR{ye"vvfyv¢"s1(0 qvHl"{qslslvBgyl(9&f
û ©¦ ù 4ô E
ù ø 2 ÷ ¨T¨(0Tù 0(£` y 1¥¢ ¥ PV ô ¢ © ø ¢ ô õ ¢ )ô ù $ ô ø
¢ ù 2 ÷¦ E ô ` û ú û õô ó ` 2 $ ù ) ¢ ÷ ' ù õ ÷ ø ù 2 ø ' $ ø ÷ ¨´yX1§~§m¢£¥P¨e9Wb1ö01§¢VGbb7©£013V£0"WaXVmb1ÛX¢B3£b" 2 ¢ ' 2 øô $ ù $ ¢ $ ø ¢¦ ÷ Eô ø ¢
ô ù Eô û û ú 2 ¢ dû ú û õô ó ù ù ` ü ` F ` õ ¢ 2 1I!3Xöbeb£IH§1X¨eH¨X¨e¥¢´°1Ie0`ö©r`(dmV1QHT¥¢ `1X{V¥U715bQ9&¥VVXXb¥R¥Xd(1 1` V¨bý{010ö`H0b§bù HX13V£0") õ ø
¢ ù¦ `û õû ù 2 ø ù $ ¦ ¢ ø ù 4 ù ` `û õû ù © E ù ô E
ù 2 ô $ ø ¢ $ 2 $ ù ô Eô¦
ù ¦ ¢ ø ù 4 $ õ ÷ $ ¦ û ) ù¦ ¤ ø
ô E ù 4 Eô ¤ ¦ ù 2 ) ù $ ø ù¦ ¤ ¢ ÷ ô ô 2 ¨X¨ö13§¥£`0ù bV£¨0ö1Xbc§(00T¨bf0b¡~gbQ"W71¥a11¨m¨Xc1¥¢ ô õô ø ù $ ø
ù E ù ô ¢ © ø ¢ ô õ ¢ )ô ù $ ø ¢
ù 2 ÷¦ E ô 2 $ ù ) ô $ ` ø ù 4 ù C §ö(eb"V0X3d{g¨¢{7X1§~§b£AIXQHP1Tb17T1¨~¡1£d§bTX0h( û¦ ¢ ô ô 2 E ô ô ` ø ø ù $ 2 ¢ ` © ø ¢ 2 ÷ ¤ ù $ i x ô © ô¦ô ¤ ¢ ô õ ø ù $ n¥1§(¨b1¥T§5¨t¥ù "0ù Xb7V¥H119b ¡vd qi ¨s5ö§¢¥af¥7¨Xù b!û û ù ¢¦
' 8 6 ù¦
) E ¦ ¢ $ ' ù¦ ù $ ô ø
ù ô
¢ ù ¦ E ù ø ÷ E ù $ Qb¥§13d71T5¥Hq07¢£l¥v§(Xe5bHWQ0(¨T10Ûb£0b!û þ 2 $ ù ) ô Eô¦
ù ø ø ÷ E © ô¦ô ¤ ¢ ¦¦ ¢ ¦ ¢ Eô
© ô ô © ¢ C C XV310"¨`§713d"!b£T5ö§¢¥a§¥f¥R¥T§b¡#¨¨B3US õ ô C ¦ ¦ ù $ @ 2 $ ù ) ø ù ¦ ÷ ô E ô¦
ù ù $ ø ø ÷ E © ô ¦ô ¤ ¢ ù $ C $ ¢ üû û õô b§Û(¨§"b£û 310"`ö1§¨`§713d¢£!"b£TP¨ö¨b¥£Xb£Q!(1X~ý´ú b7©ó
c
m } g z i o w it utr q p o k gn m k j g i h g sR{yxqvtsRPBgyl(9&f
û üû dû ô E ù ÿ ô ù 4ô 2 $ ù ) ' % ø ' ù ù $ ) ø 2 ù 4 ø ù 2 ô $ E ô $ C ` üû ú û õô bý´e´¥§edT´¨¨05õ 310"(¥3Xbb£{7T¨ô Xb¨Ëa§¢!V1ý´£b7©ó ô 2 ù ¢ ø ÷¦¦ô ù ø ¢ 2 $ ù ) ô Eô¦
)ô 2 ¢ ô Eô¦
ù $ ¤ ø ø ÷ E © ô¦ô ¤ ¢ ¦ ¢ Eô
© ¨rTey£e1ö¨§Û(QSV310"5X§71§"1¥¨`§713d"£¥D"Xb¥TW5ö¨b¥a¥U¨b¡ û ¡y yavt rqT w x w u s ipi h ¢ ¢ ` ù ø ù ø ù $ ` ô ø ù ù ø gISI("d(`b£fT¨ÇeX{d(ù cQö1IaXV1¨¨e0Xb¥PbY7§b!71¥£XW¨3V£0") b û ù¦ ¤ ¢ ÷ ô ô ` ø ù $ ù $ ù¦ô $ C ù¦ ¤ ¢ ô 2 $ ù ¦ ¢ Eô ø ù ) ÷ ù $ ø ÷ E ô $ ù 2ô ù ¢ $ Fû ô E ù ¥U¨(X"1%Q¢£b£TS§bR¥Q¢§(QbPIH&Gú §(00´ÿ {D!(§§¥"#BA09705¢ ) ø C ¦¦ @ û 8 6 ù 4ô õ
E f D
û 310"(&¢£#"!b£¨7§b¥£¡hbý´(¢ö©ó 2 $ ù ) ' % ù $ ø ø ÷ © ô¦ ô ¤ ¢ ÿ þ üû ú ù ø ÷ õ ô òñ ð
a) Euler Explicit
ï
í
× âÓÒ×Ü× Ô ÐÓ ÙÚ ßÝ Ü Ù ÙØ ÐÒ× Õ ÔÓÒ Ñ Ð Ï #rÄÇBÇávyràÛ}Þ3ÒÛrÚ4ÇÃ±ÖxBÄvmÎ
I( λ h)
î
ï
Stable
1
Unstable
R(λ h)
í
î
θ=0
Stable
b) Trapezoid Implicit
I λ h) (
Unstable
θ = 1/2 R(λ h)
I λ h) (
c) Euler Implicit Unstable 1 Stable
θ=1 R( λ h)
z !z°
û 2 $ ù ) ¢ ÷ ' ù õ ÷ ø ø ÷ © ô¦ô ¤ ¢ ÿ þ û ú ù ø ÷ õô `1¡1£dgaXVmb1Û!"!b¨ö¨b¥£¡Ø¢´(bö©ó
3.0 R(λ h) 2.0 1.0
û 2 $ ù ) ô Eô¦ ù ù ) ø ø ÷ © ô¦ô ¤ ¢ ÿ þ dû ú ø ÷ õô XV310"¨`§713dm""!¢£P¨7§b¥£¡Ø´G´ù böqó
ü RDº ® ª ² ³
¾ 6
½ ® ³ º ¹ ± ® ¼ µ ª ° ´ µ º ® ± ª ¹ ± · µ ° ³ ° ¶ ² µ ´ ³ ² ± ° ¯ ® ¬ ª § ¨§ lD!¸s»fQDW¸!TRQ1R fd!fP«©(¦
Stable Regions
RK1 RK2 RK3 RK4
I( λh) 3.0 2.0 1.0
û ¤ dû ú û õô ó ù ù 2 $ ù ) õ ø ¢ ù¦ ù $ ø ¢ $ ø ¢¦ô )ô © ø ù 4 ô © ø ¢ 2 ÷ ¤ © ô¦ô ¤ ¢ IG´¢böq`(v£¡10"g{VIUö¢£#"#yH£g£D§§"§(PXX#§sXHV1~5ö¨b¥a!£B@ è Ö 6 ø sp x ù¦ ¤ ¢ ô @ û û ¢F ê è 7cTb!C v ç" rHyw 71¥aQ¨mBQ¢F b1¥¢ ´}71¥s¢§°!(11¥¢ ê é ù $ û û ú ù¦ ¤ ¢ ô C $ 2
á × Ý yàÛ Ú Þ àÛ Ú ü Þ × UàÛ Ú ß 8 F Ü
Þ Ý Û × IHÚ × Ü Û Ö (HÚ
ø ¢ ø ù ô ù $ © I£`E¥!¨¢£P¤ ù 4ô õ 2 $ ù ) ù ¦ô ã ø ù 2 ø ' $ ü ù $ ô ø ù $ û û ú ù¦ ¤ ¢ ` æ û 2 $ ù ) ø ù 2 ø ' 2 ø 0¨Ç310"Wb¨§`bB3£I}b£Q¨mXb¥1e ´"71¥sVeb1vÅ¡1£dQXbX3XF ¦ ÷ ã ' ) ¢ 2 ù $ ô ù ù $ ù Î û æû û õô ó ô C $ ù ø ¢ ô $ ô ¢ ø ÷¦¦ô C ¥£¨V~IHQW¢£Q¨}(T¢£A¥W¢´G´ú b7©Ç§°!(1XW(§bA¥!1§(I£(X17§¨¨S# û ù¦ ¤ ¢ ©¦¦ ¢ ô ô 2 E ÷ ô ô ¢ $ ¢ ù ) ù 2 ô Eô¦ )ô ô 2 $ ù ) ¢ ù ÷ ¢ E ù ¤ ÷ öV¥a¨¨¥H¨e¨¢1TV1¨¨IHY¥U"W¥VWT1¨`§71~¨¨3V£0"¢7(VU0Eg1å
m } g z i o w it utr q wt or ¡ i m rr h gt it } h g u k gn m k j g i h g sl{yeqtvftvs1(0ä"vH9"{qslvBgyl(9&f
û
© ù ¦ô ã ù $ ù ø ¢ 2 $ ù ) ù $ $ ¤ ¢ $ ù Eô ù ¡b§¨&b£l¥(!V310"&ù TbsI¥yH£`§eI á × Þ Ý Û Ý yàÛ Ú Þ Ûà Ú Þ × UÛà Ú ß 8 F × IHÚ Ö × (HÚ Ü Ü Û
á Ý àÛ × IHÚ Þ × (HÚ Þ Ý Û Ü àÛ D8 × IHÚ Ö × (HÚ Ü Û ß
ù ù â ù ÷ 2 2 $ ù ) ¢ 2 0Xg£bb7¡1£dg1¥¢
2 $ ù ) ¦ ¢ 2ô
¢ ø
ù ' C ù $ ù ø ¢ © ø
ø
ô $ ù 4 ¢ $ ¢ $ 2 $ ù ) ø ù $ C ¡1£dIH§0Tù ¥£(0eXBIS¢£(vXXù Xb§bUH!IH!V310"!Xb¥~S# û ô ¢ ô $ ù¦ ù $ ø õ ô ©¦ 8 6 ø ù¦ ¤ ¢ ô 2 $ ù ) ¦ ¢ Eô ø ù ) ÷ ù $ 2 ¢ ô X§´y&§¢£#¥&q07by¢§¨YW"xöV¥a&¨y¡1£dD¥T§(`V~b£1¥&§´y¢ © ø ¢ ô õ ¢ )ô ù $ ô © ø ¢ 2 ÷ ¤ ù $ û ¤ üû ú û õô ó ô C $ ô ù ¢ E ô $ ø © ø ¢ 2 ÷ ¤ © ô¦ô ¤ ¢ H¨~§¢£A§rH119b0¥´¢ö©§Í!(1XD§v(¥TS¨b£A"gH119e¨ö¨b¥£X Ø ù $ bÙû × «Õa§b!x"«310"(¢b¥¢ù ¥U¥Ô`Tù b£#5ýn5Ó310"!¥H§TTù ¥e Ö % $ E ô $ C ø 2 $ ù ) ' % ù $ ¢ E ¦ ¢ ô E
ù $ `û ùû ô ` 2 $ ù ) ¦ ¢ 2 ô
¢ ø ù $ ô ù ¦ ¤ ¢ ©¦ ù ø ù $ ô ù 4¦ ù ) ù $ ù ø ¢ Ð Ï Î õ ô ¢ ø ù ù õ ù $ ù $ ¢£P§Ëö1¥£XÀ¨£0(X¢1§0¨0(Tb¢(7Ñ«b§(I£`bTÇ¢£ÒTb!C rHyw ö1IaX§ô sp x ù¦ ¤ ¢ ¢ $ 2 $ ù ) ¢ ù¦
) ¢ ù E ¢¦ E ù $ û ù¦ ¤ ¢ ÷ ©¦ ù ø ù $ ô ù ø ¢ õ ô ¢ ø õ ù ô ø ¢ © ù $ yH£¢¡1£dv¥W71~I1}§ô ¥§TxbWQö1¥£X1V~¨T(`b1¨}(v¢§ey£`e§}ù c0¢£ Ð Ï Î ù $ ø ù $ C ô õ ø ù ) ô ù¦ ¤ ¢ ù ø ¢ 2 $ ù ) ù ¦ ¤ ¢ ©¦ ¦ ¢ ô ô 2 E ÷ ¦¦ ¢ Ñ#bSù Xb!R1¨Xù h"X&¨Ø71¥£X(l1¡1£döV¥aQ5§¥H¥§e5b1011¨¥f¥ û ô õ ù p h Ë¤ ù $ ø ¢
ô ô ¢ © ø ¢ ô õ ¢ )ô ù $ 0§X(ø §TÍd¥ ¢£l¥XH¨§´y~XQH§Q¨¢£ ù ¢ E ù ù $ ¦¦ ¢ @ û úû ú û õô 0(¥T(0b£¥¨¥Y´´g¢ö©ó §ô !{Vx(£Xù ¢"¥¢£c£¨Û1¡10"71¥£X C $ ù ø ¢ © ø
ø
ù ) ¢ ù $ $ ô C 2 $ ù ) ù¦ ¤ ¢ ©¦¦ ¢ ô ô 2 E ÷ ô Eô¦
)ô ø ù $ ù ) ÿ p h Ë ¤ ¥ x sp¤ x Æ xi Ë ¤ Ë Ì ¤ xi w ¤ ¥ Ì xi 5§¥1§(¨b1¥T11B¨`§71~¨mXb¥g"3}û §TÍdV¡rvHd£dHVqx£¥~¥¥ÍHq7(¡qddh ¥i ¤i ¤ Ë Æ¤ Ê É È Ç Æ¤ x Æ Ä x Ã ù¦ ¤ ¢ ©¦¦ ¢ Eô ø ù ) ÷ ô 2 $ ù ) ù $ ¢ $ ¢ ù ) ô $ §q¥VqP¥ xSH"«dHfÅd¥d ö1¥£X£¨¨¥U¨(`1%W¨¡1£dvbWIHW1¥TW¨b
û 2 $ ù ) ô Eô¦
)ô ù¦ ¤ ¢ ©¦¦ ¢ ô ô 2 E `13V£0"S¨X¨ö1§m71¥£X¨¨¥H¨e¨¢1TV1÷ ¢£#"!b£¨7§b¥£¡h´´(¢ö©ó ù $ ø ø ÷ © ô¦ô ¤ ¢ ÿ þ úû ú ù ø ÷ õô
a) 2nd Order Backward Implicit
1.0 2.0 1.0 2.0 3.0 4.0
´ ® µ ´ · ´ ¯ ² ® ª° ³ Á ¹ · µ ° ³° ¶ ² µ ´ § ¦ ¯ ® µ º ² ¼ ¢sDÂAQR ¸!TRQ1À(¿DsQ±
I(λh)
Stable Outside
Unstable
R(λ h)
1.0 2.0
I(λh)
b) Adams Type
1.0
Unstable
Stable Outside
2.0
R( λh)
ü
ft g }n {a j z n x j ht f ft f n f g j g k j n r nt ht ft f n x n ht k j t x f g z g h xt xt i h w ft l jt r q g o n l k j i h g f e v¡ra}#}¡rCruu¡}~~}y#au¡{7Hyumu¡§ruDvuhsp©m#rrF#d hÛ u e e Ü Ú Ú p w u e vs rp i g Ö w Ü w vs rp i Ü i g fÖ UÛ e Ü
` ) ø ù ¦ ¢ ø ù ù õ ù $ ø ` ù E ô ÿ û
ê b ¦ y ê XXeI£X¢Tb!"V¥`1§´h§gé ù õ ¢ ø ù $ ô ùô¦ b ¦ y 2 ¢ ¦ ¢ ø ô y ù ø ù $ C ` ô ¢ ÷ X u x i b¥£¢£Y§T7öÅB1¥#¥Uù §»V(X¢!DÓ§(IH%tù aÓd£dx ¥ b¢£y¨eV7¢W§ô ù $ ô ÷¦ ¢ û ú ´n
w us e p i g e Ö d c b vtrqhhfAXWF Ú
$ Eô $ C ø ù 2 ÷ ô ô 2 a§¢!!`b11V§(¨b1¥TE ù $ õ ô 2
© ¤ 2 ù ô ) ø ù ù 2 ô ø ù C ¢ ù $ ` ù )ô ô © ¢ E ù 2 ø C ø õ ù 2 ÷ ô¦
) ¢ ô ¦¦ô ¢£b¨b1H¢Tb¨XX(0b§m`I1¥gbäa"§(¨~UU0bQÂ(X¢1e5ö1¥£¨k¨§b þ §(XT¢Yb£WWI 1¥ 0b£Ybö¥I¥§e5b§Ix¥¨b1S¥§eI1¢W0(§ô ô ù ÷ X ù $ I ¢ 2 ¢ $ ù ) ù $ ù ÷¦ ¢ 4 ¦ ¢ô ô ô ¢ ¢ Eô ) ø ¢ $ ¦ ¢ô ¢
¢ ù
) ù ù $ û ù 4 ¤ ¢ ù 4ô õ ô
) ÷ ¢ ù $ õ ô 2 ø E E ¢ ù E ¢
2 ¢ ù )ô ÷ $ õ ÷ ø $ ¢ÕTb¸Q(9IÒT¨ÍV§(V~V(Ib££yb¨bTTIX¥1Ò1¥"§e~V1¥b¢£ ¢ E ¢ ô $ ù 2 ¢ 2¦ ù
C U ù $ ô ô
S¦ ¢ Eô
© Q ¥£X10~I¨b£0(1÷ 1¥À§Tb¿{1Vb«¨¨9£TIU§¢3R¢ X{£ayH0P£bÎ ) ø ¢ ¢ 2 ù I ¢ ù
' 9 G E ' 6 A 9 8 64 2 0 ( ' % B¥HFDCB@)7531)¥&$ " # !
û ù ¤ x i u ¢ ¢ ù ) © ¤ ô ô ÷ ¤ ` ô ô 2 b¥ HdÍdÒd¥ !1¥T1v§Û5vVbVÓ§(¨b1¥TE © ô¦ô ¤ ¢ s Ë ¥ ¥ u ù¦ ¤ ¢ô¦ ø ©¦ øô ¢ ¢ ¢ ù 4 ø ù @ û ) ù¦ ¤ ø ¦ ¢ ø ù ù õ ù ø ) ø 2 $ ù ) 5ö§¢¥a d¥TaaaaÓx 71¥§öTù Í5öQvx¥0X`(B`~0ö1¥bl¥£`bTm("¡1£dx¢ ¥&T¢§bXS~bvb§(I¨eT"xb§bWIäTù 1y0eq¥§¨°¨eIT§711¥D¥T§(0¥bY@ ù ô $ ø C ù $ õ ô ¢ )ô ù ø ù 2ô ÷ õ ¢ ¢ 2 ÷ ù ô ô ô ¢ Eô¦ ¢ ¦ ¢ Eô E ¢ ø ¥ x wi¤i x w u s Ë x w h ui wi Ã Ë Æ x Ë ¨ ¥¤ x i u w u Ì xi s Ë Ã i¤ w u Ë ¥ w x Ã Ë û VyÍvq¥ya£¥3ydÍ¥TÍdHm¥¥V¥qHdÍdÕ(a£¡qy¥ytyva¡VvÓ¥¥VÆ ¨©¥Óqr7£¥ÅaÍ¡fÄ x§X1b¡¢~yH£gÑ!D¥¥W1¨eI~§´U¢1¥X10(`ù £by51 Ë x ip Ë Æ u i Æ Ð ¦ Î ù E ÷ 2 ø ¢ $ Ð Ï ¤ ô ¢ ) ô ø ¢ ù E ù ø ¢ô 2 ©¦ ù ô¦ ¢ Ì x i Ë ¥ sp ¤ u i ÿ û Ð Ï Î ù õ ¢ ù ¢ ô 2 ù ) ø ù ô ¢ 2 ù 2 ù 2 Tö711¥e¨ô ¡qy¡HrvdÍdv¤ cXÑ!I£XxeyÔb0XXe¨¥Y¥ä1Tù Tb&¥1&T1&¨ô 2 ¢ ø ¢ ø ô ÷ 2 ù ø ø ¢ E ©¦¦ ¢ ÷ ÷ ô ô ©¦ ¢ ¢ ô $ $ E ¢ ø ¢ ¢ ) ÷ ù 4 ø 1IfX£IXù s¨esV&Töô (T¨¨¥H(Vy¨q¨(¡¨IH¥s§bû a¥¢1¥WV¥VXBW ø ù ø ÷ ó ù $ ô ù ) ù $ E ¦ ¢ Eô ø ù ) ÷ ø ô ©¦ ¢ ¢ © ô¦ô ¤ ¢ ) ø ø ¢¦ ÷ ) ù $ ø ¢ © Xöô bHbs¨§T"TbaXV¥T§(`Vs"s¨(¡¨IH¥¨7§bIaXIþX"sQÔ1b¥9fX#bsvS
ý ñ ý ù ø ö û U0e¢üú
ù õ ñ ø ñ ÷ ö õ ó ð ò ñ ð ï î Ó0UUePô0ví
ë ì dPë
û 2 $ ù ) ô Eô¦
)ô ù¦ ¤ ¢ ©¦¦ ¢ ô ô 2 `1¡1£dS¨`§öV~¨mö1¥£X¨¨¥H¨e5b10E b#"!b¨ö¨b¥£¡Ø¢´(bö©ó ù $ ø ø ÷ © ô¦ô ¤ ¢ ÿ þ æû ú ù ø ÷ õô
a) AM3
F ü
´° ´ · ³ ² ª ² · µ ° ³° ¶ ² µ ´ ¯ ®° ¯ ¬ ¹ Á § ¦§ 3AlÅ!TRV¸D!sxÀ("(¦
3.0
2.0
Stable
1.0
R(λ h) I(λh)
2.0 3.0 1.0
Stable only on Imag Axis
b) Milne 4th Order
I(λh)
2.0 3.0 1.0
R(λ h)
ÉÈ Ç Å Ä ~¢Æ ¶qÃ Â À À x n  j h f j x j k f j hq g xt f gt ht  f g z j i h ½ ft n g  qÁ Wtv#r¿#vm´u}a¡{7r5¾u¼{aj uy²v¶}Wurt#m´~Wra#r¡rP³²v° t h »º¹ ¸ · n ft  j kt µ j k xt x x j f  j  f g
q±
û d ¦ 2 ¢ ¥VV¥y £ ` §I!"ö1¥£X1Vv¨BVTexb§bV¥ÔXX(s¥73V£0"¥1XHa§ÛUT(~5ö§¢¥aXöô bHó ¦¦ ¢ ø ù¦ ¤ ¢ ÷ ô
ù õ ô
¢¦ ø ù 4 2 $ ù ) Ð 2 ø ¢ $ Eô ` ù © ô¦ô ¤ ¢ ø ù ø ÷ ù $ © ¤ ¢ $ ` ù ø ù ø ù $ ` 2
ù b ¢£YxIHA¥("0(`b£D11Py£û ¡U¨¥x¨g &¢gIHeTöTg¨5ô a¨b!C ¥{ ¯ © ¢ C¦ ¢ ô ù ¢ $ ø ¢ ù¦ E ô $ Eô $ ) ø ® ® ¥ !× Þ Ø Ö Ø ù ø ¢ (gd ´ú û ¥¥3XSxbx´d ´yty#¥¥¡ø §ô ~¨ô ü ´yty#¥¨eV77§ô ´qty&1% ø C ù $ û û ú û X ¢ û ú û X ô ÷¦ ¢ û ú û X ÷ $
û ú d ´n é
Ö
¥
¬ 79B0E ¨ b ¦ y
¤
« 7ª ¥ ©
d ¦
b ¦ Ø ¦ §Þ £
ü
Ø
ø
á
w vs Ý e u
b Ø ¦ Þ £ Þ ¥ Ö w vs e ß u × Ý d ¦
ô ¢¦ ù ø ù $ ù 4ô õ û ú X û û X ô ÷ ô ¤ ÷ §(IÔ0(b!05}ü ´Eû 3£¨ô ¥ ´ú t3f¥7§(Ve5eXV1´ÿ û ú ü ´n á × Ý Ú Þ
hÛ aÛ Ú b ¥× £ Ü Ú ß Ø ¦ ¤Þ × Ú Ö × Ú yÛ Ý (Û Ü d ¦ aÛ
þ bü û ût3g¥ 05»1¥¢ b¥§edT´»§»T¢§(0e¥#&§b£X10e¢b§¢1¥ÔXX{!IY3V£0") X ¢ ù 4ô õ 2 û ü ô E ù ÿ ô 2 ù ô ù ) ¢ C ô $ û ù õ ô ¢¦ ø ù 4 2 $ ù Ð 2 ø ¢ $ E ô ¢ C I ô ¢ ÷ X ù ô ÷ ¢ô 2 ¦ ù 2 ) ù $ ô ¢ )ô ø ¢ ù E ø ù ¢ô 2 ù ô ÑX11a§Û¥x!(1¨eI1{¨(1E¢¢¡0b¡~#¢£Q¥§ey¨¢UXb1¥`1Tù XE¢bSe5bH ù $ ù¦ ) ¢ ù ø ¢ ¢ ø ù 2 û ù¦ ) ¢ ù © ¤ 2 ù ô ¢¦ ù ¤ ù ¤ ¢ E ø ÷ 2 ù E ø ù $ ¢£WöV~¥V£P¥`b§ô 1x`TöV~¥VÇ£0b§QÔ13d}ù X0E£¥Uxù bbT`3Xbxb
2 '¡ 0 ' A Fv§ y
! !
û 2 ù ô ¢ ô T£§ey£§}F ´¨¨e5b10HT`ed¥Hs¥ û ú ô ô 2 E ` ø ù ù ) ¢ ø ¢ ô ¢ ô ¤ ) E ù¦ ¤ ø C ù $ ô ` ¢ $ ù ø ù I ¢ ) ` © ô¦ô ¤ ¢ ø ô ô 2 E © ø ¢ ù E ù §(IH§¢WTg71§ô XXgb§kUI1£g(b÷ xP~5ö¨b¥a"¢¨e¨¢1T££XTX0b ¢ ¢ ` 2 2 $ ù ) ù 4ô © © 2 ù E ÷ 2 ø Ð ù $ ø ù 4¦ ) ù ¦ ¤ ø ù $ 2 ~¥!1I¢ ¡1£dÅT5õ ¥¢ ¤ 0X1¢3XbÑ¢£x"¨XY£g¨ô 0ö1¥bb£Å1¥¢
û ú F ´n
ê © ô¦ô ¤ ¢ ¦ ¢ Eô ø ù ) ÷ ø @5ö¨b¥a¥T§(`Vó á i g e ß ¢ $ ø ¢ ù¦ E ô ô IH#U7T§Û5b` Û Ö Û
i hRV g e i e ù E ô ` ù $ Ö hg pXV§(0b
þ ÷ $ ` © ¤ ù ø ø ù C $ Eô $ XV10(Tù bTù I7a§b!C i g XXebVHhHT1¥§(XTù b3dù e ) ø ù ù $ 2 ù C ` ô ø õ ô ô ¢ ) ù ø ù $ @ û ÷ 2 ù ø E ¢ ù ¤ ¢ E 2 b§b¨~T(vb30(0¥vE~¥UPV¥¢ `eXX0£x~0S¨ô aÛ DÚ 5e¥H%¢£ ) ø ù © ø ù 4 ù ) ) E © ô ¢ ÷ X ù $
´ ® µ ´ · ´ ¯ ² ® ª° ³ Á ¹ · µ ° ³° ¶ ² µ ´ § ¦ ¯ ® µ º ² ¼ ¢sDÂAQR ¸!TRQ1À(¿DsQ±
ü ü
û ú ú ´n
Ø b â £ ÚØ â
Ö Ú â
d â
ô ©¦ ¢ ¢ ô ÷ ¢ô 2 ø ô ¢ ÷ X ù ¦ ù 2 ) ù $ ø ù 2 §(¡5¥H¥¨(1E¢b"7¥§eyH%t00b3b£`b§ô 1 ù à û ò õ ý ñ ý û î áfsUß
Þ ì Pë
û ô ô 2 E $ E ø ù¦ ¤ ¢ ÷ © ¢ C¦ ¢ ô 2 $ ù ) ô $ ` üû ú û õô ó X1¨e5b10a1÷ "ö1¥£X1Vv¡U¨¥!¨¨¡10"§b11£b7©"£ õ ô 2 ø E E ¢ ` 2 ¢ $ E ø ¢ )' ù )ô ù $ ø 2 ù ÷ õ ô ¤ ô 2 $ ù ) ø ù¦ ÷ ô Eô¦ ù ù $ ¢ E ô $ ¢§bXT0¥¡1¥ea~¨e#b"0(1m¢§Xù ¨}310"AX71§A¨`§713db!ù ¥Uf§bW§ô ` ø ù 4 ù C û ù ÷¦ ¢ 4 ù õô ù © ø ¢ ô õ ¢ )ô ù ø ÷ ¢ $ ø ¢ ) ô $ ` ô 2 õ ô 2 ø E E ¢ 2 0XdI(DXT¢ö¥IT7XrH§¨m(b¢#¥H~5ô eI~¨b£3Q7©b1Tù 1W£Wb¨bTTI71¥¢ ¥ `d ac ¡¨" ´Et3¨©(ey&X¥HXb£H1IH`b£I&b#X£HR¥T(5eI`b c é Ü Û ô û ú û X ô ô ø ¢ ) ù E ¢ ù $ ` 2 ¢ $ ø ù $ ù $ Î û ø ¢ ¦ ¢ ù ø ù 4ô ¢ õ ù $ ô C ù ÷¦ ¢ 4 ù õô pp ø ù¦ ¤ ¢ ÷ ô 2 $ ù ) ô $ ` dû ú û õô ó ) ø ` 2 ¢ 2 $ ù ) õ ø ¢ ù¦ ù $ ¨ÛT¢ö¥IT7Xù q¥Ë "öV¥a11g§¡1£dR¨b£(3G´bö©·{(V¥310"{3¥U7¢£ © ¤ ÷ 2 ùô ø ¢ E õ ô ù ¤ ô õ ô $ E ø ¢ )' ù )ô ù $ ` ©¦ ø ¢ ù¦ û ø ¤ ) ÷ ù 4ô ¢ õ ù ¦ ¢ ù ø ø ¢ ø ¢ Vv0ö(ø Ueb§`ES§b§baX~d"§(Qb¡¥¨QUöS`Xù W1%¨(I`bÓ¥T(ù ~5ô eI~) ô $ ù ÷¦ ¢ 4 ù õô ù ù $ ¢ $ 2 ù C ô 2 ) ø ó û ùô E Ý ù E ø ¢¦ ¢ E ù 4ô ô ¢ ô ¨b£¥fTb7¥I0ö`Ûb£yH£g1HhYx5¢1Tù 1QþXtuT7X{TQÔ¥T#¨e5(A¨vÚ ¥ ù Xb!A c c ¡Y5((I~b£#0X1b¡¢P3V£0"ÑVHa¨vP§b¨X10(`ù £bX¥1 ø ù $ C Ü Û p Ú ô ø ¢ ) ù $ ù E ÷ 2 ø 2 $ ù ) Ð 2 ø ¢ $ Eô ô õ ô E ù ø ¢ô 2 ù E ¢ ù $ û ô ÷¦ E E ÷ ô 4 ¤ ø ù $ ¢ ø ù ù ø ¦ ÷ ø ù $ ` $ E ¢ ø ¢ ù ù ø E ô 2'ô ) ù $ 2 bQX1¨(17T1¥T1§Q`b£IT(0(bQ£¨V(Tù bÓa¥Xb1¥eed(XX§bB5"Tù eb¢1¥¢ ù Eô ø ¢ ) ¢¦ ÷ E øô E ô C ùô 4 ù $ TX¨((Ie¥Ô1`7X¥e§!07´~b£ X{&T71¥VdÇb¨bT`T(¢b~b¨¥VÍI!B@ ) ø ù¦ ) ¢ ù õ ô 2 ù E ù ø ù $ ù ô ) ¢ ù ù C û ù 4ô E ø ù ù ø ù ¢ô 2 ¢ ) ø ¢ 2 ù 4 ø ø ¢ ÷ ¤ ` ô ÷ E ô 5edTù `E!0(XE¢¢»{IT¨ô (VbH01¨(VX¨b2 ÷ ô 4 ø ø ÷ ¦¦ ¢ ô õ ù ¤ ù 4 ¢ $ ù C ù ) ¢ ù $ ©¦ E ¢ ù ô @ û ù 2ô E E ¢ ¢ 1§0ù b#bl¥9§¥7¨rb§ô 1÷ TXù UHhc¥b£¨£d¥Vd5¨ÛBAT¢§X0¥¥¥1§ô `I("d(`b£Ó0§ô 35¥H¥Xöô bHrb¢§mX(0"¥£QH¨ö¨b¥£Xgb"ä~IgX§1axbe`~0ö¦ ù ø ù ø ù $ ` ©¦ ¢ ¢ ø ù ø ÷ ó ù $ ô ø ù ù ) ¢ ø ¢ © ô¦ô ¤ ¢ ù $ ø ù Eô $ E ù $ û ) ù ' ¤ ø ¦ ù 2 ) ù $ 2 $ ù ) C ù $ ô ¢ Eô¦ ¢ © ¤ 2 ù ) ø ù Eô ø ¢ ) ¢¦ ÷ E øô E ù $ ¡bÓTb¡~b!V310"Sb£s¥v§(IU¨ö1V¥0X"#T`§((I~!IÔ1`ö`&¢£ ¥ aHr¥yHd¡5d ¢£eb¨b1H"Qù bbT`3Xbö1§ô eQ¢§öVT(0(0(b1÷ F0ö1¥Vd}S ¥ p Ë Ã x Ìi ù $ õ ô 2 ø ø ÷ 2 ù E ø ù¦ ) ¢ ù ô¦ ÷ 2 ù ù ù ø Ù ù¦ ) ¢ ù C ù $ 2 ¢ ` ) ù © ù $ ¢£1ITe¡Xe¢£¥ yda¥Hd¡5d I£Tù T(¢T(ø ´0t3¢¨ô vs Øb£¥(IUe¢£~¢§Xù ¤ w ¤ u Ã x Ìi ¢ ù ø ù û ú û X w u e ù $ ` ù ¢ E ù $ õ ô $ E ÷ ÿ û ¢¦ ÷ E øô E ô û û X ) ø ù $ ô ÷ ô ù ) ù $ E õ ô E ù ø ¢ô 2 ù $ ù $ C 2 ù ô a1´¥ÔVX7X!§t` ´ú ty¥¿"b¨V¢¨"TbaXb¨X10(`ù £b&¢£0b!0b§") 'ø ù ù 2 ` ø ¢ X(0b95ô eI~) b£AyH£¨Ø¨(¡¨IH¥e¨ö¨b¥£X`ö(¢H§r¨eV1÷ ¥b¨´5`b11vb × ù $ ¢ $ ô ô ©¦ ¢ ¢ © ô¦ô ¤ ¢ ø ùô ø ÷ ó ¢ ô ô ) ¢ õ ô ©¦ ø ù 2 ÷ ù $
2 ' 64 9 Ò FDv0
Ö Ò y7§FB7Í5Ô~v5v³ÑÐ Ó 0¡ G 6 94 Õ A Ò Ó A4 Ò 0¡ '
Î Ï !
û ø ù ù ) ¢ ø ¢ ù ø ù $ ù Eô $ X`ed¥HXù {b£l¥X¨1aE © ¢ ø ù ¦ ¤ ¢ ÷ ô ` ù © ô ¦ ô ¤ ¢ Ë ø ù ô ø ÷ ó ù $ © ¤ ` ¢ $ ô ¢ ) ô ø ¢ ù E ù ø ¢ ô 2 ' ù ô I!"ö1IaXV1§£TXT(~5ö¨b¥a#X7(bbP1UyH£P¨eI~§´U¢1¥XVT(`ù £be5bH ø ù $ ¢ ù 4 ¢ $ ù C ÷ $ û y 2 `b£I1¥PUHhe1%P0ÍV¥¢ ø ù ¦¦ ¢ ø `X1¥1A§¥" ³ ÌI1£U7T§¨ô a§b!C ¥{ ¯ ¢ $ ø ¢ ù¦ E ô $ Eô $ ) ø
b ¦ Ë b¦By§ô (p p d ¦ ¥
Ö
ù ¥U¨b£7@ ¢ E ô $
û ú ´n
b Ë á× ¥ Ý Ú ³× Ü Ú ß ¦ §¥ Ú Ö × Ú hÛ hÛ (Û Ü d ¦ hÛ
ô ¢ ÷ X ù ô E ù 4 Eô §(IH%t¨§(00T¨b¤ ¦ ù 2 ) ù $ ø ô ¢ )ô ø ¢ ù E ù ø ¢ô 2 ' ù ô ù $ ø ù 2 E ù¦ ) ¢ ù ø ù $ ¢ Tb¡¢b£x" §(I~§´Ub1I¢XVT(`ù £be5bH¢£`b§ô 1TöV~¥VsX¢£¥1IvQ d ü · ± ª ® µ ´° ´ ª ¹ WDs3Q± § Ê§ ¡´(¦
û ø' ù $ X¥¡p¢£ ~5Tùey¢Ôb0ù)§VTX`3XbWb¥¢¥U(X1dTù 1¨s£&(b(01÷ eX¥¡1eeyÔb0XXe¨Í1§ô ©¦ ô 2 ) ô 2 ù ù E ø ù 2 ¢ ù @ û E ô ù ø ÷ E ø ø' 6 ù ¢ô 2 ù ) ø ù ô ù Xb$Ó`¥`E1Tvx¦HÎ¥HH§b1¥ÙD¨(¥Heb£T`IdEÕ1§P1§}(`b ø ù ù ù û ¦ ¢ ô $ 2 ¢ Ï ¤ Eô ¢ ¤ ù $ ù ù C ù ¤ Ï Î ÷ ô 4 ¤ ô ù ø ù $ ò × Ü Ú Þ × Ý epð Þ × Ú ð ¥ çÖ × Ú ð Þ Úñ hÛ yÛ Ý (Û Ü
aÛ
)ø ù ù õ ¢øø ¢ ùø 2 ~XX(b¥(T(¨1¥¢ û ú ¥ ´n
d ¦
Ø b¦ £
ð
ù C û ¦ ù 4 ù¦ ù )ô ù ø ¢ô 2 C ¢ ù ÷¦ ¢ 4 ù õ ¢ ø ù 4 07¦ (X¨T0öQ¨e!T(`ù £bI¢ö¥I£`U¢ © ¤ 2 ù E ¢¦
ø ù ¤ ¢ ¨ô 0X¥§1Tù Ë0Xù ¥1$ btû 3§5eII5(X¢eXIHb£7§ô XX(l¥eT`&¢£a¨b!7§ô û ü X ô ù 4ô ¢ 4ô ø ù 2 ù E ¢
ù $ ) ø ù ¦ ¢ ø ù E ù $ $ Eô $ C
îï × Ü Ú Þ ìí aÛ × IÛ Ý Ú
Þ
× UÛ Ü
Ú
ë ê
b ¥ × Ý Ú ÍØ ¦ èé £ hÛ d ¦
Þ
Ú × IÛ Ý
Ö × (Û Ú Ü
þ ü ø ù ¢ $ ô 2 $ ù ) ¦ ù I ¢ ø ó' ø ó ÷ Ï ù $ 2 ù E ÷ 2 ø ô ù C ` ù¦ ) ¢ ù ø ó û 2 ø ù 2ô bÛX(V¥H¨¨¡10"ÓTQ5¥tXH£b£7T`1b¡e¨hV¥71¥VdHØ0Tù Xb¨(1¥TE ù © ù 4 ¢ $ ù C © ô¦ô ¤ô ¢ ` õ ô E ù ø ¢ô 2 ù E ¢ 2 ¢ ù )ô 2 ù ô ) ù E ÷ 2 ø ô ) ù $ dI1UHhg¨7§b¨(X9&V1¢§XVT(`ù £b`¥HvV¥Q"§(T´©X1b¡(§S§ô T¢£ ¥bf`~V£¨ô 7¥l¥g5ö§¢¥aIU§ô X"1%QbeI§1b§¢¥&£¡U¥~Qù ù Xb ù Î û ) $ ø õ¦ ¢ © ô¦ô ¤ ¢ ¦ ¢ E ø ù ) ÷ ù $ ù ¢¦ ÷ ô ¢ ) © ¢ C © ¢ ) ø ¢ ø ù $ £ û {Ø æ#¦ ç b ¦ ê d ù ) ù øô ÷ X ù ø ù $ © ¤ ù ô ù ù )ô ô 2 ù ô )ô¦ ©¦ ù ø ù 4 ø ù $ ¢ ø ø ¢ ù C ù ¢ E ô $ ø 2 00(ö¢{(b£&`§(0e"§(¨Te5¨ö¨0(X0ù QX¢£Iù gh&(IU§b"1¥¢ ù¦ ¤ ¢ ©¦¦ ¢ ô ô 2 E ` ø ù 4 ù C $ ` ô @ û Ð 6 ù 4ô ¢ õ ù ¦ ¢ ù ø ø ù¦ ¤ ¢ ô ô ù E 71¥£X¢5§¥1§(¨b1¥TDT`0h(1f0§B`Rv¨eyXb¥T("~71¥a&¨5s`1§ô æ ´§t3 û ú û X ù ¢ ø õ ù ô 2 $ ù ) ¢ ÷ ' ù õ ÷ ø ù 2 ø ' 2 (I£Xe¨£e¡10"QaXVmdb1ÛAXbX¡1 b(VI(XbTI¥¨1Tù s´HT"10b1 ù $ ù ÷ ` ù ø ÷ E ` 2¦ ÷ E b û ¢ ù ) ù $ ù õ ¢ ø ô ¤ô $ $ õ ÷ ù ø ¢ 2 ù ø ø ¢ E ù ø ù C ù )ô ù $ ¢ ù¦ ¤ ¢¦ô ¢ 4 ¢ $ E ¢ ø ¢ ©¦ b¥£(X5b§¢Vù £bVTv¢0öô (T¥1x(`I("§(x¢£I&öV¥Ô¨I(£a¥b1I¢51 ù $ ô ¢¦ ÷ E¦ ¢ E 2 ¢ $ ù $ ` ø ù 4 ù C $ ` © ô¦ô ¤ ¢ ¤ ø ¦¦ ¢ @ û © ô 4 ù¦ ù ) ù E ø ÷ ¢ ¢£"1§(IÔVX§IU1¥H!bVT`0h(1¥¥¨7§b¥1bV§¥~50ö"XÓ¥XXbXù ¤ (1 C ¢ E E ¢ ô $ 2 ¢ ` ô Eô 2 ù ø ø ù $ ¢ ù C ø 2 ø ¢ $ Eô © ¤ 2 ÷ ¢ C 2 $ ù ) ù $ ¥T0¥vW¨b£ 1¥¨ed§b0(bXbIUhW"äX11a§Û«ÅTù 1r¥#ä¡1£db û ¤ dû ú û õô ó ù ù ` ù ¢ 0IG´£b7©X(0T(IUE $ E ÷ ¦¦ ¢ ø ù¦ ¤ ¢ ÷ 2 $ ù ) õ ø ¢ ù¦ ù $ ô ø ' ÷ ô ø ÷ ù $ 2 ¢ ù 4ô ¢ õ ù 2 a1(Ó§¥"ö1IaXV1Û§ô 3V£0"{VIUöb7¨Û¥¡1§(¢bb£1¥5eIXb1¥¢ ¦ ¢ ø ¦¦ ¢ ù ø ¢ ¥Uù §§¥~( ` ' ` !¥¥3XVP¢£~`1§ô ~71¥£X11Ë¨¨¥U¨(X"1%§§b&yH£y0ö¥Td(ø ø' 6 ù $ ù E ù¦ ¤ ¢ ÷ ©¦¦ ¢ Eô ø ù ) ÷ ô ô $ ¢ $ 2 ù¦ ¢ ù 4 ù ø ù ¢ $ ô $ ô ©¦ ¢ ¢ ø ÷ Î `eV¥1¨b£«§r§ô ¡¨¥HIbÕû a~v¨eb£g"À0(1÷ 3V£0"c{3¥U7bÅ¨ÛC $ E ø ¢ ) ù )ô ù $ ø 2 ù 2 $ ù ) õ ø ¢ ù¦ ù $ $ ô
û ú æ ´n
å ®
b Ø ¦ Þ Ú ¥ c c vÛ £ ä
Ö Ú
ã
d
ä ã þ Ð Ï Î ) ù © ù $ 3áXÑ¥¿0eX¡¢£¥¢ ô ¢ ø ¢ ) ô 2 ù ù ø ù ¤ ¢ E ô 2 $ ù ) ù ø E ô 2'ô ) ù ¢ û ù¦ ¤ ¢ ÷ ù ¤ 2¦ ÷ ¥§eya¥1W5ô eI~¨¨0(XT(¢VBù IU¨310"!e0ù XX§bB5"T(#AöV¥a11vE~§1SC $ E ¢ ø ¢ ô $ ¢ $ ©¦ ù ¢ô 2 ù ) ) C I C ù C 2 ¢ © ¢ 2 ø ¢ ù¦ E ù ô ÷ X ô ù E E ù $ ` ø ù 4 a¥b1I§¢fI1£Q5T(IÔb0§ô (11(1#he1¥&UH¡£U7TSe¨¢³§uVT`1TSb£0X0ù '#(0!(1¥V¥#~¨7§b¥£X1¨qIU§ô X"1%A¥V0X1¥Te¢£¥"§eÓ1XHa¨vc@ C û C I ¢ C © ô¦ô ¤ ¢ ô ¦ ¢ E ø ù ) ÷ ù E E ù $ ` ù )ô Ð 2 ø ¢ $ Eô û úû ú ô E ù ÿ 2 $ ù ) ø ùô ø ÷ ó ù $ © ¤ © ô¦ô ¤ ¢ ô 2 ù ©¦ ¢ ¢ 2 ´´~¥§edT´¨¨ô 310"X7(bHbP5ö§¢¥a£¨¨T`¡¨¥HI1I¢ b£t3§ô û ü û X 2¡10"Q§¢£0eTù Tù beù QE¡!§bA¥1§(IH%t"b¨eIX(¨B"¡1£d¢T(¥9Xb $ ù ) ô $ 2 ù ø b û ù © ô $ ô ¢ ÷ X ù õ ô ¢ ø õ ù ô ø 2 $ ù ) ¢ 2 ù ø 2 ø ¢ $ Eô û ó ô C ù â ` E ¢ ô ` ù¦ ¤ ¢¦ô ¢ 4 ¢ ù ) ¢ E ù ¤ ø ù ÷ ) E ù ø ¤ ø ¢ ù © © ¢ X1QHa§ÛÓh§ÛdgÓ0Iv§tVé V( 71¥§§IUW"¥U0E`eVb0e("0ù T¥ã
ä
´ ® µ ´ · ´ ¯ ² ® ª° ³ Á ¹ · µ ° ³° ¶ ² µ ´ § ¦ ¯ ® µ º ² ¼ ¢sDÂAQR ¸!TRQ1À(¿DsQ±
ü
b ¦ d ¦
þ
ù ø ù $ (`b!C
´n Fû ú ¢ ô )ô¦ ù $ ù I ¢ I¨"§öyb£YQPa é Fû ú ´n
Ød â
ÚØ â Ø
þ ÿ£
¥ Ø b â £
ÚØ â
Ö Ú â
d â
2 ¢ ô ¢ ù ù 4 ¤ ¢ ù $ V¥c1¨(1¥1V(¥äb
2 ù b û VHy¢é Ì~c ¦ ý b ¦ d û ú û X ô ) ø ù ù $ ù E ¢¦ ù $ ô £5ÛC ¥ ´mt3¸¨~~XX(yb££X¥§1TÛ
ü b iii Þ Ú vvvi â 3¦ û Ø ¦ ( paap p d aÛ d Ø ¦ ( ¥ Ú w w w w b â Ø ¦ aÛ ü b û i w Þ Ú vi â y¦ ö¥ Ú w â aÛ aÛ d Ø ¦ Ö Þ Þ
b Ø ¦ × Ü Ú Þ ¥ DÚ ¥ × DÚ × Ý Ú aÛ × IÛ Ý (Û Ü aÛ ¥ d Ú ñ ¦
ù E ù ø ¢ô 2 ù E ¢
2 ¢ ù )ô 2 ù ô ) ù $ ø 2 0X10(`ù £bX¥HXV¥"§ev0¢©b#"71¥¢
p p aap ii Ú vvi â Ø ¦ d aÛ ò Ú i Ú aÛ â çÖ × IÛ Ý × (Û Ü
ù 4 ¢ $ ù C ù 4ô ¢ 4ô ø ù 2 ù )ô ù $ ø ó û¦¦ ¢ ) UH&h5eII5(X¢¨eb#Hun¨¥~ © ø ù 4 ù ) E ù ¤ ù ô $ ù ) ù $ ¢ õ ô 4¦ ©¦¦ ¢ ÷ E ¢ ù ø ¢ ù C ô ¢ ÷ X ù ¦ ¢ô ø ù ¢ô 2 ¦ ¢ô ø ¢ ù $ XT"T0E!X¨(0!bDIb§¨¥&¨§IHed¥!(hx¥§eyH%tb¥§eTù XE¢b3¥Ô(X1¢£ E ÷ ø E ø 2 ¢ ùô ø ø ¦ © ¢ ¢ û ú X ô ) ø ù ù $ 2 ¢ ù ù C ` ô $ ù 4 ø d1((X10Tù ¨V¥!T7(Xù #¨3UsWv§ô I ´£û 37¨5~XX(b£v1¥1VmI3T§¢£(Xbxs û Ï ¤ ¦ ¢ ô õô ø ù $ $ ô D¥1§7(b7¨ÛC dÓdd§d¥ s u x ¤ ¥i ¦ ù $ ô ô $ ø 2ô ¢ ù 4 ¢ $ ù C ù E ø ù $ ú x w Byau ¥¥§b#§§¢£"P¨HUHhX§ô bb
û`§¥§)V35 e¨e I§9XøX(¨ô¥¢§b¥9¥e¥§eyH§beTE qdÍr{# ¥b§ô 1}¡1£d#71¥£X ¦ ¢ô ©¦ ô ¢¦ ù ô õ ¢ ø õ ¢ â ô ¢ ô ¤ ) ¤i u ip ù ø ¢ õ ÷ 2 $ ù ) ù¦ ¤ ¢ © ¦ ¢ ô E ÷ 2 ÷ ù ¢ ù C ù ô ` ù ø ¤ 2 õ ) ù ¦ ø ù 4 ¤ ¢ ù $ 5¦§¥1§ô(¨b21¥T11¥¢11"4(1$hAXE1¨(y¥b÷ eù !£¡3Û¡£RTXù ¨1÷ Tù Ø(9¥Sb ÷§TÍ¤dWsrq¥yÍtq¥yaÓ¡ §ô ¡1£dbyH£#1IU"¨b p h Ë ¥ pp Ë x wi¤i x w u x 2 $ ù ) ù $ ¢ $ ¢ ù ) ô $ û ·õöê ð êxõb¥}ùb£¢¨ô ð ¥T(øÇ¥¢ø" Pù ya¨b!D¥f¨¥¥¡ôã ((`b é ù ¢ø $ ¦ ¢ ù © ê ø ¢ $ Eô $ C ¦¦ ¢ ø' ù ø ¢ ù ø ù $ ð Þ Ö gBq0E ð b ¦ y
b ¦ By Ø §( Ø ð ô
¥
ô Eô ¢ ø 2 ¢ ÷ X ù $ ô ÷¦ ù $ §m§(I£¢¥H%Hb£l¥v¨eV7b ¥ ¥ ¥ Þ é ÖA ð 9b¦ó{0E ð Ø 7 ð
ø
á
w vs Ý e u
Þ
w vs Wß ð u e
Þ
¥ ç7 ð Þ Ö × Ý 7 ð
ù E ¢ 2 ¢ ù¦ ô $ û û ú ô E ù ÿ ô 2 ù E ÷ 2 ø £XV!I!1¥Tö!§¢A´ú Ë¥§edT´¨0X1b¡(§ô ô ©¦ ¢ ¢ © ô¦ô ¤ ¢ ø ùô ø ÷ ó ù $ ¢ ù ) © ¤ ô ÷ ô $ © ø ¢ E © ¢ C ù¦
)ô ù $ §(¡5¥H¥¢¨ö¨b¥£XX7(b"b¥1IU"c¨V§bXø U7£cUXT71§(b
ú iü
· ± ª ® µ ´° ´ ª ¹ WDs3Q±
§ Ê§ ¡´(¦
ï é B Cî é
é ¥ èA
¥
Ö
9
c
ï B Cî
¥
5 é
¥
5 é à Ö HÚ
ã ã
¥
é
é ¥ èA
é
Ö Ù7 $ ô £5ÛC Ï Î ù $ ø ù 2ô &b£X¢§(Vû ý
5 ò ø ÷ î ð 6TAç4 2 ì 3Pë
d 9 @Þ Ú 7 8Ö Ú
û 11ø XsXe¨¢¥çÖ XV310"TQ5¥tXHÏ 2 ÷ ¤ ø ø ù © ø ¢ ø ô ¤ ø ¢ ¢ û 2 $ ù ) ¦ ù I ¢ ø ó' ø ó ÷ ' 2 1I¢ !"1¥§e5b107010e§(VT1¥d¥£b`T¥f¥X~~V´Ø¢´ö1If ø ô ô 2 E © E ù ô E 2 ¢ © E ¢ ø ÷ E E ¢ © ø ¢ ) ) ÷ ÿ þ Fû ú ù¦ ¤ ¢
1 ) b 0¦ %
d ¦
w ! £ # !
ù ø ù ø ù $ ("d(X¢ á w £ i ß ' % (&Ø ¨s¨1 ô ©¦ i ! Ö $"0eI~§´UbVQ # ! ù ¢ )ô ø
w ! £
# ! # Ö $! $ ô £5ÛC
i !
` ù ô ©¦¦ ¢ ô ô 2 T0e§(V~¨¨¥H¨e¨¢1
ù¦ ¤ ¢ ÿ ©¦¦ ¢ ô ô 2 E 71¥£¡¨¨¥H¨e¨¢1TV
ù ©¦ ) ø ô T(X§ô 1¨"¨b © E ù ø d1T(X§ô 1Hó
ù¦ ¤ ¢ ÿ ©¦¦ ¢ ô ô 2 ö1Ia¡s¨¨¥H¨e5b1
Ø w
õ ê d ¦
ê d #¦
¦ ù I ¢ø ó ø ó ÷ T¥££´XH£Ï
© ô¦ô ¤ ¢ ÿ ø ¨ö¨b¥£¡vHó
¢ ÷ ' ù õ ÷ ø ù 2ø ' 2 £VmdbVv!`bB31
û ù¦ ¤ ¢ ù $ 71¥abv§ô C $ 2 $ ù ) ¢ ÷ ' ù õ ÷ ø ù 2 ø ' 2 E ù ù $ ô ô 2 E © ô¦ô ¤ ¢ ù $ © ¤ 2 ù ) !(1c¡1£dv£Vmb1ÛXbX¡1T0(xbI§(¨b1¥T¨ö¨b¥£Xxb£0(§ô ¢ $ ¢ ù ø ù 4 ¢ Ù $ Eô $ C ù ù )ô ù $ ô ¢ ø E ¢ 2 ¢ ù yH£¥(`0ù ¥X1÷ ~¨ô a¨b!cTeW"§egb¨£(X1¥Tv£1¥Tö¦ ô $ ` ø ù 4 ù C w ¨b£lTX0h( þ û ù E ô 2 $ ù ) ù $ 2 ¢ ` d ¦ XTe§ô 1T&§¡10"Pb£1¥AXP§ô Ø ù $ ` ¢ E ô ¢ ) ù TbU¥£X10e1¨~T(ø i I1£yU¢ ¢ $ © ¢ C $ E ÷ ô a1(v¨é 1Ié yBA¢´ö1Ifv¨TX¨(~1(¨Í§(IH(¨(bD`TX¨(T(X ý b ¦ 2 ¢ ý d ¦ @ û Fû ú ù¦ ¤ ¢ ô 2 ù ô ø ¢ ) ) ÷ ô ô ¢ ÷ ô ù $ û ù ô ù ¦§¥~¨¢§bX§b¥I"¢00ù ¥ry(x#¡yd(cyÍv¥1 ¡ xHqqqAHdd§dVya7rgyU H¡ ¦ ¢ ) ©¦ õ ô $ ô ¢ 4 ø ù 4 Ãp w ¥ w ¤ ¤ w ¤ ¥ Ä x w i¤ Ë Æ¤ Æ¤i Ä ¤ x ¤ ¥i ¥ x w u sp w i x û ú û X ¢ E ¢ $ E ÷ @ û ô ¢ ÷ X ù ô ÷ ¢ô 2 ¢ ` ô ¢ ÷ X ù ù 4 ¢ ¤yÓe§i ¨¦¥£¡ É ù ¥Uca1(0¨eIH%{¨(1E¢¢Q¥1¨eI1{UC ¢ ¥ ´0t3 w x ¥ © §¢ ¤ ¢ w §ô ¢ $ © ¢ C ¨¡1£dP¢£yP¨Xc¨¨¥Hed¥7hy§(IH%tsb£AU¥£X1¥T1§~Tù ø i I1£yU¢ ô 2 $ ù ) ù $ © ¤ ù 4¦ ©¦¦ ¢ ÷ E ¢ ù C ô ¢ ÷ X ù ù $ ` ¢ E ô ¢ ) $ E ÷ ô a1(7¨ré V¥~é m¦ 5BVDû Íryd0dH¡Æ §ô Gú t3Dû ÍrydX¥Vù §ú ´Eû 3 ý b ¦ 2 ¢ ý d ô ÷ $ u ip w h ù s Fû û X u ip w h Ë Ë ô û ú X
´ ® µ ´ · ´ ¯ ² ® ª° ³ Á ¹ · µ ° ³° ¶ ² µ ´ § ¦ ¯ ® µ º ² ¼ ¢sDÂAQR ¸!TRQ1À(¿DsQ±
æ ü
` ù¦ ¤ ¢ '@ ø hö1¥£XBÍ ` ù¦ ¤ ¢ ' f 7 ` ù¦ ¤ ¢ ' ô @ û 2 $ ù ) ù $ ø ô ¢¦ ø ' 6 ù $ ù 4 ø ù Ï ` 2 ø ø ¥71¥aXgeIö1IaXB¨0¡1£d&b£"¥§eyÔTù yÏ1x¢£&¨ô XÅB1Tù (0E ô ù 2 ¨ÛT1 G¢t3§9äd1¥¡H1F¥!T¢ö¥IIH!x£1VsB&` ¢l¢0eISW¨Û@ dû û X ô c 2 ¢ ` % ` b ù ÷¦ ¢ 4 ¢ $ C ` @ ã ã â ù ' C ¢ ô `¨`§71~¨©s¨`§713dq¨uv¡1£d#¢£§u`bfIHÕR¢§1"s§¢£Qb¨(1310") ô Eô¦ )ô ø ô Eô¦ ù ô @ ` 2 $ ù ) ù $ ô ø ù 2 ø ¢ $ b û ¢¦ ÷ ) ø ô $ õ ô ÷ 2 $ ù õ ô $ E ø ¢ )' ù )ô ¢ ô ¢ ¤ ` ` ù 2 ô ¦ ¢ ø ) ù ù $ © ¤ X ù 2 ô ¦ ¢ô ¢ ù $ õ ô E ¢¦ ù ø © b§¢a~¨e#§QaVyfadbV§I£~0ebYdb1¨¥§eI1Ab#b§`¥Ô10(S b ¥ × × Ü Ú ¦ Ý Ú F
Ö
× Ü Ú w Þ Ú w Þ Ó× Ý Ú w W V Wü V W V
þ ù 4ô ¢ 4 ø ù 2 ø ¢ ø ô ¢ )ô ø ¢ ô ' F 2 ø ù ù E E ¢ ) E ù $ ¦¦ ¢ E ù û Q¨(II¨ô XbXX£W!"¨eI~§´U¢1¥!§9ByTù X(0X!d¥HTbÓ§¥TTÛ´d
$ õ ÷ ø $ ¢b X¢RIV310"£VmdbVvb££5Û!ù HDbbXb£ XbX ø ù 2 ø 2 $ ù ) ¢ ÷ ' ù õ ÷ ù $ $ ô C ø ¢
) û ü $ õ ÷ ø $ ø ù 2 ø ¥S13V£0"PX"VX¥#¥HQQb£"#b£TW¨ö¨b¥£Xb¥¨1v1¥b¨XXedbü 2 $ ù ) $ ø $ ¢ ' ) ¢ 2 ù $ ø ø ÷ E © ô¦ô ¤ ¢ ù $ ¦
2 ¢ ù ô ) ø ù ù Ï û
û ù¦ ¤ ¢ ô 2 $ ù ) ù $ $ Eô $ C ø Qö1¥£X¨¨¡1Ë 0"¢£7a§b!#" Û 3¥¢¥£b11VT¨(¡¨¥1¥P5ö¨b¥a × ù õ ¢ ø ù $ 2
` ô ©¦ ¢ ¢ © ô¦ô ¤ ¢ ø ù ø ÷ ó õ Xöô bH¨b§ô Uk¦ ç 8 ¨kTXù W1%l5"P¥bb£IY!(1F` Û ¶(`b!C û b ô ` ø ¤ ) ÷ â ó ø ¢ ø ÷ ù $ ¢ C I × ù ø ù $
û bü
× Þ Ú ¥ × Ü Ú ØÛ × Ý Û Ú Û Û
¥ × Þ × Ü Ú Û × Ý Û Ú Û
¥ Ú
Û
Ö × (Û Ú Ü
þ ù 4ô 2 $ ù X05õ 310") ¢ ø 2 b ¢ â C I ©¦ ù 2ô C ù $ ` ô ¢ ÷ X ù ô E ù 4 E ø ¢ ù ô¦ ù $ 2 ùô¦ ¢ ù $ b û PXb1Tù Ty&!{V¡¨0b§ÛbU¨eIH%{¨edT0ÓTb§7Db£#QT7ö11Ie0bÕ¢F û 2 $ ù 0¡10") b£#"{¥byUù TÛUv ù $ ø ù 4 ¤ ¢ ù $ ¢ ù E ù¦ E øô E ô ÷ ù $ öTXö`S5b1Bb£ û ¦ ù ) ¢ ù $ I§1"¥£b¥ C ¢ ø 2 2 ¢ ù ¢¦ ' ù¦ ) E ù $ ô ø ù $ ù E ¢ ø ù $ ¦ ¤ ¤ Ub1I¢¥Ô1B9¿d71Tb£x§e¥3XbÓ¥XI£eb£D¥¨AÌHv
û ù E ¢¦ ¢ ù¦ ¢ ø ù $ ù ÷¦ ¢ 4 ù ÷¦ ¤ ¢ ù $ ù ÷ ) E 2 ¢ é æû é `T`¥Ô1gXIUö!y~#£¥¡bsI0böI¥(Vö¥1¥beVb0V¥xG´ é X{x¢¥!¨¥IXX(§Ë¨~xPsû £T¢!¡1£d¢X"3¥#¥H ) ø d éû é ¦ ¢ 4 ø ù ô ô 8 ù I ¢ Ö 6 ù $ C 2 $ ù ) $ ø $ ¢ ' ) ¢ 2 ø ù 2 ø ' 2 `b31 bSI¥¡¿¢£I0böI¥IU§ô X"1%~¢£Igö1¥£7~e3b~¥t ù $ ø' ù $ ù ÷¦ ¢ 4 ¦ ¢ E ø ù ) ÷ ù $ ù¦ ¤ ¢ ¢ ù ÷ ) ¢
û
` 2 ÷ ¤ ù $ $ ô C ù E ¦ ÷ ø ø ÷ © ø ` © ô¦ô ¤ ¢ ø 8 2 ÷ 11V9ù Tb7¨Û!T(X§ô 10£¨V(Tù Ûbù 755ö¨b¥a"7!1V1¤ 2 ù E
ù ù $ ù ø ¢ ¢ $ C ` 2 $ ù ) ù ù ø $ ù $ ø ô ¢¦ ù ø ' 6 ù $ õ ô T(0Tù 3dbP(yH!01¡1£dv`(b7ù TbW"&V§(IÔ0(!Vb£b¨(SUv E
û ô ÷¦ © 2 ¢ ù E ¢ ù ù $ $ ô C ô ÷¦ 2 ù ÷
) E ù $ ù ø ¢
) 0¨eV773¥U(X0¥V¢£v£5ÛP§(VöX0eV¢~0b£(QH~¥ û
ù ù )ô é `1T(X"§(¢é "é ÕÅ1¥~1T(X"§(¢¥d "¢ü é Õ01T(XP"§( ø Ö 8 2 ¢
ù ù )ô é ø Ö 8 `
ù ù )ô é é d ø é Ö 8 `
ù ù )ô é é ¥¥&" G yA01T(Xg"§(v¥é " ¨ é Y(1qT0(¥T~`(bf§¥`1¡ ø Ö 8 ù ÷ ` ù ¢ E ù ù ø $ ¦¦ ¢ @ û 2 '3£d&¥X(I1¥0XöVel5X¨ö1§0X71§A¨`§713d5bb§ô 1é A~"§( $ ù ) I E ¢ ) ø E ¢ ã 2 ¢ ` ø ù¦ ÷ ô Eô¦
)ô ` ø ù¦ ÷ ô Eô¦
ù ù $ õ ÷ Ö d ù )ô ) ø QI { RP c c Ht Ú §(¨b1¥TR¥§e¨¢§b ¥{&(I£Xe¨£eb¡T~ge5(ø PHv G Ö é ô ô 2 E ¦ ¢ô ô ô ù $ ) ø ù ¢ ø õ ù ô ù 2 E ¢ ù ô b ¤
` ù )ô ô ù 4 ¢ $ ù ¤ ô ÷¦ Ï Î ù $ ù 2 C ` ô ÷¦ ù ¢ h¨ev§UH0Ev¥§e3öXÅ&b£T1¸(¨eV7(Ia ' © 2 ¢ ù ù $ ô ¢ $ b û ù õ ¢ I E ¢ ¦ ¢ E ø ù ) ÷ ¢ õ ÷ 7 ù ÷¦ ¢ 4 ù õô ù ù $ 2 ô ¢ V¥Teb£¨ÛIHÕQy¥HIU§ô X"1%Wxb§ô 1F¥T¢ö¥IT7XmbP1§qó t
ü
´ ® ³ ¶ ¹ ¯ º § D§ ¢#DÀE(¦
`¦ ¢ ) ¥~ø1ç×§§þX¨BD¢»T71¢`I1¥b»{#XXE¢b¨b£v!v§QÔ13f ô ô û æ ) ù¦ ¤ ø ø ù C ¢ ù $ ) ø ø ù ¢ô 2 ô $ © $ C ô ¢¦ ` © ô¦ ¤ ¢ ô ) © ¢ 2 ù C ¦¦ ¢ ø ¤ ) ÷ ¢ ø ÷ ¨7§ôbIaXQE§eI£V ¡X¥ø"PTh(§§IXù W1%!¥£¢ 1§´y~b§5IH!1¥ù Tb ) ÷ )ô ¢ ) ù $ ô ¢ $ C ` ù $ T(I#¿ûX0böI¢¥0ùö`'9 ùb£ Ù1H¡10"`ö1§¨`§713dù b£v"Ò¨eI§T(ø 2 ù ¢ ù ÷¦ 4 õô ù $ 2 ` 2 $ ù ) ø ù¦ ÷ ô Eô¦ ù $ ø ô ¢¦ ù ' ù $ 16¢b£cõb§ô (û q ¥~0öô eT¨ùb£ 75SV¥«§(Ia¥V 5ô eI~À0b1¥H«§ô e5(ø b ù ø ù $ ù 4ô õ 2 ¢ ô ¢ ø ¢ ) 2 ù 2 ¢ ¤ × ù ô Û yq ¥ HÚ Ö × (HÚ Û × Ü Û ä
` û ù û ô ` ) ø ø Eä ù 4 ' ô ø ¢ ) ô õ ô $ E ø ¢ ) ù ) ô ø ù ¦ ÷ ô E ô ¦
ù ¨Q¢"#£00Í¡©(ä (I7¨}bä §baX~¨eX71§!¨`§713dq¥ ô ¢ Eô¦
¢ ù $ ) ø ¦ ÷ ù ø $ Eô $ C ¦ Î ù $ ù ô b û ù ÷¦ ¢ 4 ù õô ù' 6 ù $ 2
û §(IU¨ö1V¥e¢£ ¥{£51(0(Ëa¨b!Åx§&b£(¨(ø AX0bö¥ITöX1&bVH1` ô 2 ù
õ ô û ) ø ø E ù 4 ' ô ø ¢ ) ô õ ô E ù ø ¢ô 2 ¦ ¢ô ¢
2 ø ¢ C I E ¢ ¤ ø ù 2 ø ' ø 5b10EVQQ¢§(p¨"l00B©(eyW¨b¨X10(`ù £b3IÔeyHxQ#¥HsXbXXX£
) ø ¦ ø $ Eô $ C Ð Ï Î {¨1÷ Tù a¨b!Ó¥ T(X¡X!be5(ø fH1V9¥ö(ÛbIx0e5XXù
) ù © ù $ ù ô b û © ø ¢ 2 ÷ ¤ $ õô ø ù $ ¢ 2 ù ô ) ø ¨Í¨e¨¢1TX111xDH1V9#q0ö¢£Iv¥§e5b10PH1190öVa§(7g¢ ô ô ô 2 E © ø ¢ 2 ÷ ¤ û © ø ¢ 2 ÷ ¤ ù¦ ù $ ¢ ô ô 2 E © ø ¢ 2 ÷ ¤ ù¦ $ Eô øô Ï ©¦
û ¨1VQl¢æ 0ö1¥b¨g¥e0Xù Uf5e¨ô #&¨Û§(00T9U¢§öA¢£X¢§(Vû ) ù¦ ¤ ø
ô ¢ 2 ù
ù 4 ¢ C ù 4ô
¢ $ ô C ô E ù 4 E ø ¢ ù ô¦ ù $ ø ù 2ô
û © ô¦ô ¤ ¢ ø ø ¤ ) ÷ ¢ ø ÷ ) ÷ )ô ¢ ) ù $ 2 ô ó 5ö§¢¥a"Xù W1%I£b»1¨´ybv1§©Øû 5e5(7£¨Ûv¨eI1{ù ù 4ô ô
$ ô C ô ¢ ÷ X ô E ù 4 E ø ¢ ù ô¦ ù $ 2 ùô¦
¢ õ ô $ E ø ¢ ) ù )ô ø ù¦ ÷ ô Eô¦
ù $ ô C 2 ù¦
÷ E õ ¨e000Ub¨ö&b£x£7T7ö11I~¢§baX"§(`ö1§¨`§713d¨£¨ÛT7110~b¨ô ' E ø ù ¢ô 2 2 ø ¢ C I E ¢ ¤ ø ù 2 ø ' ø
© ô¦ô ¤ ¢ ù $ ù ©¦ ¢ ¢ ` ©¦ ¢ ¢ ø ù ø ÷ ó õ ô û 1Tù XE¢by¥HXbXXb#¥¢5ö¨b¥a~¢£xX¡¨¥1¥lT§ô ¡¨¥HI&`öô bHË¢§(&¢æ û 2 $ ù ) ù ù $ ù ø ô ô 2 `1¡1£d(0b£l¥b"P¥§e5b10E © ô¦ô ¤ ¢ ù $ 2 ô 5ö§¢¥avb£yV§©ó h§b¡§ô Xù v(e1¨e¨¢1TyX111Pb V¥xT`1Tù XE¢b¥Ô( ` Eô 2 ø
ù ø ¢ ô ô 2 E © ø ¢ 2 ÷ ¤ ù $ 2 ¢ ù E ø ù ¢ô 2 ¦ ¢ô ' ¢
ù $ ø 2 ù ÷ ô õ ô E ù ø ù ¢ô {H¢¢£"Ò0(1§»b¨X10(XE¢¢2 0(X(TX~XbX¡1 ¨h¤ §bb¨eIX(¨ô 2 ùø ù ù E ø ù 2ø ' 2 ô ` Ï ô $ õ ô ¢ ø õ ù ¥ Ö (`b!C ø ù¦ ¤ ¢ ô ÷ ù ¤ 2¦ ÷ C 2 $ ù ) õ ô $ E ø ¢ )' ù )ô ô Eô¦
ù $ Eô $ b "71¥a5b(&E§1SQ1¡10"¢§baXBd"§(5X§713d~a¨bÕQû ù ø ù $
Ë
Ø b â
Ú Ø â Ö Ú â
þ Ï ¤ õ ô C ¦¦ ù $ ø ù 2ô û ÛD¢§Û(§¨"b£X¢§(V´ú
d â
û ô ©¦ ¢ ¢ ø ùô ø ÷ ó õ ô ÷ ¢ ù ù û ô X¨(¡¨IH¥!X7(b"b¨(1vITE0v0§( ' ¢ ÷ X ù ô ¢ô 2 ¦ ù 2 ) ù $ 2 ùô¦ ¢ 2 $ ù ) $ ø $ ¢ ' ) ¢ 2 ø ù 2 ø ' 2 {H%ty¥§(V÷ £bDTb¡~b££0ö711¥ä¡10"y£"VX¥#~IHQ&XbX3V b ù $ 2 ¢ ù E ù ø ù ¢ô 2 2 ù ø ù ù E ø ù 2 ø ' 2 1I0X10(XE¢¢vT(`eT`AXbX¡1 Ig§(IH§¢WTb"~0eX"§(ö1¥£X V) ô ¢ ô ¤ ) E ù $ ø ù ù )ô ù¦ ¤ ¢ ) ÷ ¥ '5´y~eb1HH95ô eI~£IXù ¥Ô(IHXeb¥Tb7¥I0ö`b~b¨(iû ô ¢ ) ù $ 2 ` ø ¢ ) ø ¢ ø ¦ ¢ô ¢ ù $ ù ÷¦ ¢ 4 ù õô ù ù $ õ ô Ö 8 6 $ ô C 2 ù ô ¢ ¤ ô ` ô ¢ ¦ ¢ ù ø ù 4ô ¢ õ ù ù $ E ù ø ù ô ø ÷ E © ô¦ô ¤ ¢ ù $ ù ø ù $ ¨Û0b§QaV&§qT§´yS¥T(s¨(IX¢Pb£&£00(XXe¨bTc5ö§¢¥ab(`b!C ô ù $ `û ùû ô ` 6 ù 4ô ¢ õ ù ¦ ¢ ù ø ø 8 ù ÷¦ ¢ 4 ù¦ ¤ ¢ ) ÷ )ô ¢ ) ù $ ¢ $ C $ ¨eb1¨QERW¨(IX¢lIU(Û"6 ¥¥b7¥I71¥£X»V§´y~&¢£IH£¸(1X
é ÙÖ ç 6 8
Þ
7 ç F 8 6
Þ
¥
Ø
2 $ ù ) $ ø $ ¢ §ô ¡1£d£"VI# ' ) ¢ 2 ø ù 2ø ' 2 ¥HQ!`bB31 b"¨eI§T(ø !V&b£yH£7T¨ô û ©((I£IXù ¥§eIHX ù $ ø ô ¢¦ ù ' 6 ù $ ¢ $ ù 4 ô ø ¢ ) ø ¢ ø ¦ ¢ô ¢ ù¢£l¥!Tb7¥I0ö`b£71HVV75b10EVQxb§ô h`1§(¨bVTXHV1§b¡§ô Xù $ ù ÷¦ ¢ 4 ù õô ù ù $ 2 ` üû ô 2 ù õ û ô ô 2 E © ø ¢ 2 ÷ ¤ Eô 2 ø $ ô C ô ¢ ÷ X ù ô ÷ ¢ô 2 ¦ ù 2 ) ù $ ô ù 4ô ¢ 4ô ø ù 2 ¦ ¢ô ¢ ù $ ô ¢ )ô ø ¨Ûä§(IH%t¨(1E¢¢#Tb¡bä§5eII5(X¢¥Ô(IH7b£Y¨eI~§´UbV¥¢ ù E ø ù ¢ô 2 2 ù ø ù ù E ø ù 2 ø ' 2 `1Tù XE¢b¢0(`eT``b31 ¢£gb§511¥Tb¨aV¥ 0e3ÀW¢£e5(ø ¢ ù $ õ ô ©¦ ¢ © ¤ 2 ù ô ¢ ¤ ) ù © Ï Î ù $ ù ô b û ´ ® µ ´ · ´ ¯ ² ® ª° ³ Á ¹ · µ ° ³° ¶ ² µ ´ § ¦ ¯ ® µ º ² ¼ ¢sDÂAQR ¸!TRQ1À(¿DsQ± é ¥d
Chapter 8 CHOICE OF TIMEMARCHING METHODS
In this chapter we discuss considerations involved in selecting a timemarching method for a speci c application. Examples are given showing how timemarching methods can be compared in a given context. An important concept underlying much of this discussion is sti ness, which is de ned in the next section.
8.1 Sti ness De nition for ODE's
8.1.1 Relation to Eigenvalues
The introduction of the concept referred to as \sti ness" comes about from the numerical analysis of mathematical models constructed to simulate dynamic phenomena containing widely di erent time scales. De nitions given in the literature are not unique, but fortunately we now have the background material to construct a de nition which is entirely su cient for our purposes. We start with the assumption that our CFD problem is modeled with su cient accuracy by a coupled set of ODE's producing an A matrix typi ed by Eq. 7.1. Any de nition of sti ness requires a coupled system with at least two eigenvalues, and the decision to use some numerical timemarching or iterative method to solve it. The di erence between the dynamic scales in physical space is represented by the di erence in the magnitude of the eigenvalues in eigenspace. In the following discussion we concentrate on the transient part of the solution. The forcing function may also be time varying in which case it would also have a time scale. However, we assume that this scale would be adequately resolved by the chosen timemarching method, and, since this part of the ODE has no e ect on the numerical stability of 149
150
CHAPTER 8. CHOICE OF TIMEMARCHING METHODS
Stable Region λh I(λ h)
1
R(λ h) Accurate Region
Figure 8.1: Stable and accurate regions for the explicit Euler method. the homogeneous part, we exclude the forcing function from further discussion in this section. Consider now the form of the exact solution of a system of ODE's with a complete eigensystem. This is given by Eq. 6.27 and its solution using a oneroot, timemarching method is represented by Eq. 6.28. For a given time step, the time integration is an approximation in eigenspace that is di erent for every eigenvector ~ m . In x many numerical applications the eigenvectors associated with the small j mj are well resolved and those associated with the large j mj are resolved much less accurately, if at all. The situation is represented in the complex h plane in Fig. 8.1. In this gure the time step has been chosen so that time accuracy is given to the eigenvectors associated with the eigenvalues lying in the small circle and stability without time accuracy is given to those associated with the eigenvalues lying outside of the small circle but still inside the large circle.
The whole concept of sti ness in CFD arises from the fact that we often do not need the time resolution of eigenvectors associated with the large j m j in the transient solution, although these eigenvectors must remain coupled into the system to maintain a high accuracy of the spatial resolution.
8.1. STIFFNESS DEFINITION FOR ODE'S
151
8.1.2 Driving and Parasitic Eigenvalues
For the above reason it is convenient to subdivide the transient solution into two parts. First we order the eigenvalues by their magnitudes, thus j 1j j 2j j Mj (8.1) Then we write p M Transient = X c e m t ~ + X c e m t ~ (8.2) xm xm m m Solution m=1 {z m=p+1  } {z }  Driving Parasitic This concept is crucial to our discussion. Rephrased, it states that we can separate our eigenvalue spectrum into two groups one 1 ! p] called the driving eigenvalues (our choice of a timestep and marching method must accurately approximate the time variation of the eigenvectors associated with these), and the other, p+1 ! M ], called the parasitic eigenvalues (no time accuracy whatsoever is required for the eigenvectors associated with these, but their presence must not contaminate the accuracy of the complete solution). Unfortunately, we nd that, although time accuracy requirements are dictated by the driving eigenvalues, numerical stability requirements are dictated by the parasitic ones.
8.1.3 Sti ness Classi cations
In particular we de ne the ratio and form the categories
The following de nitions are somewhat useful. An inherently stable set of ODE's is sti if
j pj
j Mj
Cr = j M j = j pj
Mildlysti Cr < 102 Stronglysti 103 < Cr < 105 Extremelysti 106 < Cr < 108 Pathologicallysti 109 < Cr It should be mentioned that the gaps in the sti category de nitions are intentional because the bounds are arbitrary. It is important to notice that these de nitions make no distinction between real, complex, and imaginary eigenvalues.
152
8.2 Relation of Sti ness to Space Mesh Size
CHAPTER 8. CHOICE OF TIMEMARCHING METHODS
Many ow elds are characterized by a few regions having high spatial gradients of the dependent variables and other domains having relatively low gradient phenomena. As a result it is quite common to cluster mesh points in certain regions of space and spread them out otherwise. Examples of where this clustering might occur are at a shock wave, near an airfoil leading or trailing edge, and in a boundary layer. One quickly nds that this grid clustering can strongly a ect the eigensystem of the resulting A matrix. In order to demonstrate this, let us examine the eigensystems of the model problems given in Section 4.3.2. The simplest example to discuss relates to the model di usion equation. In this case the eigenvalues are all real, negative numbers that automatically obey the ordering given in Eq. 8.1. Consider the case when all of the eigenvalues are parasitic, i.e., we are interested only in the converged steadystate solution. Under these conditions, the sti ness is determined by the ratio M = 1 . A simple calculation shows that 4 2 1 = ; 2 sin 2(M + 1) x
M
!
; 4x2
x 2=; 2
and the ratio is
; 4x2 sin2 2 = ; 4x2
4 =4 M +1 2 2
M= 1
The most important information found from this example is the fact that the sti ness of the transient solution is directly related to the grid spacing. Furthermore, in di usion problems this sti ness is proportional to the reciprocal of the space mesh size squared. For a mesh size M = 40, this ratio is about 680. Even for a mesh of this moderate size the problem is already approaching the category of strongly sti . For the biconvection model a similar analysis shows that j j=j j 1
M
x
1
Here the sti ness parameter is still spacemesh dependent, but much less so than for di usiondominated problems. We see that in both cases we are faced with the rather annoying fact that the more we try to increase the resolution of our spatial gradients, the sti er our equations tend to become. Typical CFD problems without chemistry vary between the mildly and strongly sti categories, and are greatly a ected by the resolution of a boundary layer since it is a di usion process. Our brief analysis has been limited to equispaced
x
8.3. PRACTICAL CONSIDERATIONS FOR COMPARING METHODS
153
problems, but in general the sti ness of CFD problems is proportional to the mesh intervals in the manner shown above where the critical interval is the smallest one in the physical domain.
8.3 Practical Considerations for Comparing Methods
We have presented relatively simple and reliable measures of stability and both the local and global accuracy of timemarching methods. Since there are an endless number of these methods to choose from, one can wonder how this information is to be used to pick a \best" choice for a particular problem. There is no unique answer to such a question. For example, it is, among other things, highly dependent upon the speed, capacity, and architecture of the available computer, and technology in uencing this is undergoing rapid and dramatic changes as this is being written. Nevertheless, if certain ground rules are agreed upon, relevant conclusions can be reached. Let us now examine some ground rules that might be appropriate. It should then be clear how the analysis can be extended to other cases. Let us consider the problem of measuring the e ciency of a time{marching method for computing, over a xed interval of time, an accurate transient solution of a coupled set of ODE's. The length of the time interval, T , and the accuracy required of the solution are dictated by the physics of the particular problem involved. For example, in calculating the amount of turbulence in a homogeneous ow, the time interval would be that required to extract a reliable statistical sample, and the accuracy would be related to how much the energy of certain harmonics would be permitted to distort from a given level. Such a computation we refer to as an event. The appropriate error measures to be used in comparing methods for calculating an event are the global ones, Era , Er and Er! , discussed in Section 6.6.5, rather than the local ones er , era , and erp discussed earlier. The actual form of the coupled ODE's that are produced by the semidiscrete approach is
~ u At every time step we must evaluate the function F (~ t) at least once. This function is usually nonlinear, and its computation usually consumes the major portion of the computer time required to make the simulation. We refer to a single calculation of the ~ u vector F (~ t) as a function evaluation and denote the total number of such evaluations by Fev .
d~ = F (~ t) u ~ u dt
154
8.4 Comparing the E ciency of Explicit Methods
As mentioned above, the e ciency of methods can be compared only if one accepts a set of limiting constraints within which the comparisons are carried out. The follow assumptions bound the considerations made in this Section: 1. The timemarch method is explicit. 2. Implications of computer storage capacity and access time are ignored. In some contexts, this can be an important consideration. 3. The calculation is to be timeaccurate, must simulate an entire event which takes a total time T , and must use a constant time step size, h, so that
CHAPTER 8. CHOICE OF TIMEMARCHING METHODS
8.4.1 Imposed Constraints
T = Nh
where N is the total number of time steps.
8.4.2 An Example Involving Di usion
Let the event be the numerical solution of
from t = 0 to T = ; ln(0:25) with u(0) = 1. Eq. 8.3 is obtained from our representative ODE with = ;1, a = 0. Since the exact solution is u(t) = u(0)e;t, this makes the exact value of u at the end of the event equal to 0.25, i.e., u(T ) = 0:25. To the constraints imposed above, let us set the additional requirement The error in u at the end of the event, i.e., the global error, must be < 0:5%. We judge the most e cient method as the one that satis es these conditions and has the fewest number of evaluations, Fev . Three methods are compared  explicit Euler, AB2, and RK4. First of all, the allowable error constraint means that the global error in the amplitude, see Eq. 6.48, must have the property:
du = ;u dt
(8.3)
Er < 0:005 eT
8.4. COMPARING THE EFFICIENCY OF EXPLICIT METHODS
Then, since h = T=N = ; ln(0:25)=N , it follows that
155
1 ; ( 1 (ln(:25)=N ))N =:25 < 0:005 where 1 is found from the characteristic polynomials given in Table 7.1. The results shown in Table 8.1 were computed using a simple iterative procedure. Method N h Fev Er 1 Euler 193 :00718 :99282 193 :001248 worst AB2 16 :0866 :9172 16 :001137 RK4 2 :6931 :5012 8 :001195 best Table 8.1: Comparison of timemarching methods for a simple dissipation problem. In this example we see that, for a given global accuracy, the method with the highest local accuracy is the most e cient on the basis of the expense in evaluating Fev . Thus the secondorder AdamsBashforth method is much better than the rstorder Euler method, and the fourthorder RungeKutta method is the best of all. The main purpose of this exercise is to show the (usually) great superiority of secondorder over rstorder timemarching methods.
8.4.3 An Example Involving Periodic Convection
Let us use as a basis for this example the study of homogeneous turbulence simulated by the numerical solution of the incompressible NavierStokes equations inside a cube with periodic boundary conditions on all sides. In this numerical experiment the function evaluations contribute overwhelmingly to the CPU time, and the number of these evaluations must be kept to an absolute minimum because of the magnitude of the problem. On the other hand, a complete event must be established in order to obtain meaningful statistical samples which are the essence of the solution. In this case, in addition to the constraints given in Section 8.4.1, we add the following:
~u The number of evaluations of F (~ t) is xed.
Under these conditions a method is judged as best when it has the highest global accuracy for resolving eigenvectors with imaginary eigenvalues. The above constraint has led to the invention of schemes that omit the function evaluation in the corrector step of a predictorcorrector combination, leading to the socalled incomplete
156
CHAPTER 8. CHOICE OF TIMEMARCHING METHODS
predictorcorrector methods. The presumption is, of course, that more e cient methods will result from the omission of the second function evaluation. An example is the method of Gazdag, given in Section 6.8. Basically this is composed of an AB2 predictor and a trapezoidal corrector. However, the derivative of the fundamental family is never found so there is only one evaluation required to complete each cycle. The  relation for the method is shown as entry 10 in Table 7.1. In order to discuss our comparisions we introduce the following de nitions: Let a kevaluation method be de ned as one that requires k evaluations of ~u F (~ t) to advance one step using that method's time interval, h. Let K represent the total number of allowable Fev . Let h1 be the time interval advanced in one step of a oneevaluation method. The Gazdag, leapfrog, and AB2 schemes are all 1evaluation methods. The second and fourth order RK methods are 2 and 4evaluation methods, respectively. For a 1evaluation method the total number of time steps, N , and the number of evaluations, K , are the same, one evaluation being used for each step, so that for these methods h = h1 . For a 2evaluation method N = K=2 since two evaluations are used for each step. However, in this case, in order to arrive at the same time T after K evaluations, the time step must be twice that of a one{evaluation method so h = 2h1. For a 4evaluation method the time interval must be h = 4h1 , etc. Notice that as k increases, the time span required for one application of the method increases. However, notice also that as k increases, the power to which 1 is raised to arrive at the nal destination decreases see the Figure below. This is the key to the true comparison of timemarch methods for this type of problem.
k=1 j k = 2 j 2h1 k=4 j 4h1
0
T
j j j
( h1 )]8 (2 h1)]4 (4 h1)]2
uN
Step sizes and powers of for kevaluation methods used to get to the same value of T if 8 evaluations are allowed. In general, after K evaluations, the global amplitude and phase error for kevaluation methods applied to systems with pure imaginary roots can be written1 Era = 1 ; j 1(ik!h1 )jK=k (8.4)
1
See Eqs. 6.38 and 6.39.
8.4. COMPARING THE EFFICIENCY OF EXPLICIT METHODS Er! = !T ; K tan;1 k
"
157 (8.5)
1 (ik!h1 )]imaginary 1 (ik!h1 )]real
#
Consider a convectiondominated event for which the function evaluation is very time consuming. We idealize to the case where = i! and set ! equal to one. The event must proceed to the time t = T = 10. We consider two maximum evaluation limits K = 50 and K = 100 and choose from four possible methods, leapfrog, AB2, Gazdag, and RK4. The rst three of these are oneevaluation methods and the last one is a fourevaluation method. It is not di cult to show that on the basis of local error (made in a single step) the Gazdag method is superior to the RK4 method in both amplitude and phase. For example, for !h = 0:2 the Gazdag method produces a j 1 j = 0:9992276 whereas for !h = 0:8 (which must be used to keep the number of evaluations the same) the RK4 method produces a j 1 j = 0:998324. However, we are making our comparisons on the basis of global error for a xed number of evaluations. First of all we see that for a oneevaluation method h1 = T=K . Using this, and the fact that ! = 1, we nd, by some rather simple calculations2 made using Eqs. 8.4 and 8.5, the results shown in Table 8.2. Notice that to nd global error the Gazdag root must be raised to the power of 50 while the RK4 root is raised only to the power of 50/4. On the basis of global error the Gazdag method is not superior to RK4 in either amplitude or phase, although, in terms of phase error (for which it was designed) it is superior to the other two methods shown.
K leapfrog AB2 Gazdag RK4 !h1 = :1 100 1:0 1:003 :995 :999 !h1 = :2 50 1:0 1:022 :962 :979
a. Amplitude, exact = 1.0. K leapfrog AB2 Gazdag RK4 !h1 = :1 100 ;:96 ;2:4 :45 :12 !h1 = :2 50 ;3:8 ;9:8 1:5 1:5 b. Phase error in degrees. Table 8.2: Comparison of global amplitude and phase errors for four methods.
The 1 root for the Gazdag method can be found using a numerical root nding routine to trace the three roots in the plane, see Fig. 7.3e.
2
158
CHAPTER 8. CHOICE OF TIMEMARCHING METHODS
Using analysis such as this (and also considering the stability boundaries) the RK4 method is recommended as a basic rst choice for any explicit timeaccurate calculation of a convectiondominated problem.
8.5 Coping With Sti ness
8.5.1 Explicit Methods
The ability of a numerical method to cope with sti ness can be illustrated quite nicely in the complex h plane. A good example of the concept is produced by studying the Euler method applied to the representative equation. The transient solution is un = (1 + h)n and the trace of the complex value of h which makes j1 + hj = 1 gives the whole story. In this case the trace forms a circle of unit radius centered at (;1 0) as shown in Fig. 8.1. If h is chosen so that all h in the ODE eigensystem fall inside this circle the integration will be numerically stable. Also shown by the small circle centered at the origin is the region of Taylor series accuracy. If some h fall outside the small circle but stay within the stable region, these h are sti , but stable. We have de ned these h as parasitic eigenvalues. Stability boundaries for some explicit methods are shown in Figs. 7.5 and 7.6. For a speci c example, consider the mildly sti system composed of a coupled twoequation set having the two eigenvalues 1 = ;100 and 2 = ;1. If uncoupled and evaluated in wave space, the time histories of the two solutions would appear as a rapidly decaying function in one case, and a relatively slowly decaying function in the other. Analytical evaluation of the time histories poses no problem since e;100t quickly becomes very small and can be neglected in the expressions when time becomes large. Numerical evaluation is altogether di erent. Numerical solutions, of course, depend upon ( m h)]n and no j m j can exceed one for any m in the coupled system or else the process is numerically unstable. Let us choose the simple explicit Euler method for the time march. The coupled equations in real space are represented by
u1(n) = c1 (1 ; 100h)nx11 + c2(1 ; h)nx12 + (PS )1 u2(n) = c1 (1 ; 100h)nx21 + c2(1 ; h)nx22 + (PS )2
(8.6)
We will assume that our accuracy requirements are such that su cient accuracy is obtained as long as j hj 0:1. This de nes a time step limit based on accuracy considerations of h = 0:001 for 1 and h = 0:1 for 2 . The time step limit based on stability, which is determined from 1, is h = 0:02. We will also assume that c1 = c2 = 1 and that an amplitude less than 0.001 is negligible. We rst run 66 time steps with h = 0:001 in order to resolve the 1 term. With this time step the
8.5. COPING WITH STIFFNESS
2
159
term is resolved exceedingly well. After 66 steps, the amplitude of the 1 term (i.e., (1 ; 100h)n) is less than 0.001 and that of the 2 term (i.e., (1 ; h)n) is 0.9361. Hence the 1 term can now be considered negligible. To drive the (1 ; h)n term to zero (i.e., below 0.001), we would like to change the step size to h = 0:1 and continue. We would then have a well resolved answer to the problem throughout the entire relevant time interval. However, this is not possible because of the coupled presence of (1 ; 100h)n, which in just 10 steps at h = 0:1 ampli es those terms by 109, far outweighing the initial decrease obtained with the smaller time step. In fact, with h = 0:02, the maximum step size that can be taken in order to maintain stability, about 339 time steps have to be computed in order to drive e;t to below 0.001. Thus the total simulation requires 405 time steps. Now let us reexamine the problem that produced Eq. 8.6 but this time using an unconditionally stable implicit method for the time march. We choose the trapezoidal method. Its behavior in the h plane is shown in Fig. 7.4b. Since this is also a oneroot method, we simply replace the Euler with the trapezoidal one and analyze the result. It follows that the nal numerical solution to the ODE is now represented in real space by 1 ; 50h nx + c u1(n) = c1 1 + 50h 11 2 ! 1 ; 50h nx + c u2(n) = c1 1 + 50h 21 2
!
8.5.2 Implicit Methods
1 ; 0:5h nx + (PS ) 1 1 + 0:5h 12 ! 1 ; 0:5h nx + (PS ) 2 1 + 0:5h 22
!
(8.7)
In order to resolve the initial transient of the term e;100t , we need to use a step size of about h = 0:001. This is the same step size used in applying the explicit Euler method because here accuracy is the only consideration and a very small step size must be chosen to get the desired resolution. (It is true that for the same accuracy we could in this case use a larger step size because this is a secondorder method, but that is not the point of this exercise). After 70 time steps the 1 term has amplitude less than 0.001 and can be neglected. Now with the implicit method we can proceed to calculate the remaining part of the event using our desired step size h = 0:1 without any problem of instability, with 69 steps required to reduce the amplitude of the second term to below 0.001. In both intervals the desired solution is secondorder accurate and well resolved. It is true that in the nal 69 steps one root is 1 ; 50(0:1)]= 1 + 50(0:1)] = 0:666 , and this has no physical meaning whatsoever. However, its in uence on the coupled solution is negligible at the end of the rst 70 steps, and, since (0:666 )n < 1, its in uence in the remaining 70
160
CHAPTER 8. CHOICE OF TIMEMARCHING METHODS
steps is even less. Actually, although this root is one of the principal roots in the system, its behavior for t > 0:07 is identical to that of a stable spurious root. The total simulation requires 139 time steps. It is important to retain a proper perspective on a problem represented by the above example. It is clear that an unconditionally stable method can always be called upon to solve sti problems with a minimum number of time steps. In the example, the conditionally stable Euler method required 405 time steps, as compared to about 139 for the trapezoidal method, about three times as many. However, the Euler method is extremely easy to program and requires very little arithmetic per step. For preliminary investigations it is often the best method to use for mildlysti di usion dominated problems. For re ned investigations of such problems an explicit method of second order or higher, such as AdamsBashforth or RungeKutta methods, is recommended. These explicit methods can be considered as e ective mildly sti stable methods. However, it should be clear that as the degree of sti ness of the problem increases, the advantage begins to tilt towards implicit methods, as the reduced number of time steps begins to outweigh the increased cost per time step. The reader can repeat the above example with 1 = ;10 000, 2 = ;1, which is in the stronglysti category. There is yet another technique for coping with certain sti systems in uid dynamic applications. This is known as the multigrid method. It has enjoyed remarkable success in many practical problems however, we need an introduction to the theory of relaxation before it can be presented.
8.5.3 A Perspective
8.6 Steady Problems
In Chapter 6 we wrote the O E solution in terms of the principal and spurious roots as follows:
un = c11 ( 1 )n ~ 1 + + cm1( m )n ~ m + + cM 1 ( M )n ~ M + P:S: 1x 1x 1x n~ + n~ + +c12 ( 1)2 x1 + cm2 ( m )2 xm + cM 2 ( M )n ~ M 2x n~ + n~ + +c13 ( 1)3 x1 + cm3 ( m )3 xm + cM 3 ( M )n ~ M 3x
+etc., if there are more spurious roots
(8.8)
When solving a steady problem, we have no interest whatsoever in the transient portion of the solution. Our sole goal is to eliminate it as quickly as possible. Therefore,
8.7. PROBLEMS
161
the choice of a timemarching method for a steady problem is similar to that for a sti problem, the di erence being that the order of accuracy is irrelevant. Hence the explicit Euler method is a candidate for steady di usion dominated problems, and the fourthorder RungeKutta method is a candidate for steady convection dominated problems, because of their stability properties. Among implicit methods, the implicit Euler method is the obvious choice for steady problems. When we seek only the steady solution, all of the eigenvalues can be considered to be parasitic. Referring to Fig. 8.1, none of the eigenvalues are required to fall in the accurate region of the timemarching method. Therefore the time step can be chosen to eliminate the transient as quickly as possible with no regard for time accuracy. For example, when using the implicit Euler method with local time linearization, Eq. 6.96, one would like to take the limit h ! 1, which leads to Newton's method, Eq. 6.98. However, a nite time step may be required until the solution is somewhat close to the steady solution.
8.7 Problems
1. Repeat the timemarch comparisons for di usion (Section 8.4.2) and periodic convection (Section 8.4.3) using 2nd and 3rdorder RungeKutta methods. 2. Repeat the timemarch comparisons for di usion (Section 8.4.2) and periodic convection (Section 8.4.3) using the 3rd and 4thorder AdamsBashforth methods. Considering the stability bounds for these methods (see problem 4 in Chapter 7) as well as their memory requirements, compare and contrast them with the 3rd and 4thorder RungeKutta methods. 3. Consider the di usion equation (with = 1) discretized using 2ndorder central di erences on a grid with 10 (interior) points. Find and plot the eigenvalues and the corresponding modi ed wavenumbers. If we use the explicit Euler timemarching method what is the maximum allowable time step if all but the rst two eigenvectors are considered parasitic? Assume that su cient accuracy is obtained as long as j hj 0:1. What is the maximum allowable time step if all but the rst eigenvector are considered parasitic?
162
CHAPTER 8. CHOICE OF TIMEMARCHING METHODS
Chapter 9 RELAXATION METHODS
In the past three chapters, we developed a methodology for designing, analyzing, and choosing timemarching methods. These methods can be used to compute the timeaccurate solution to linear and nonlinear systems of ODE's in the general form d~ = F (~ t) u ~u (9.1) dt which arise after spatial discretization of a PDE. Alternatively, they can be used to solve for the steady solution of Eq. 9.1, which satis es the following coupled system of nonlinear algebraic equations: ~u F (~ ) = 0 (9.2) In the latter case, the unsteady equations are integrated until the solution converges to a steady solution. The same approach permits a timemarching method to be used to solve a linear system of algebraic equations in the form A~ = ~ x b (9.3) To solve this system using a timemarching method, a time derivative is introduced as follows d~ = A~ ; ~ x x b (9.4) dt and the system is integrated in time until the transient has decayed to a su ciently low level. Following a timedependent path to steady state is possible only if all of the eigenvalues of the matrix A (or ;A) have real parts lying in the left halfplane. Although the solution ~ = A; ~ exists as long as A is nonsingular, the ODE given by x b Eq. 9.4 has a stable steady solution only if A meets the above condition.
1
163
164
CHAPTER 9. RELAXATION METHODS
The common feature of all timemarching methods is that they are at least rstorder accurate. In this chapter, we consider iterative methods which are not time accurate at all. Such methods are known as relaxation methods. While they are applicable to coupled systems of nonlinear algebraic equations in the form of Eq. 9.2, our analysis will focus on their application to large sparse linear systems of equations in the form Ab~ ; ~ b = 0 u f (9.5) where Ab is nonsingular, and the use of the subscript b will become clear shortly. Such systems of equations arise, for example, at each time step of an implicit timemarching method or at each iteration of Newton's method. Using an iterative method, we seek to obtain rapidly a solution which is arbitrarily close to the exact solution of Eq. 9.5, which is given by ~ 1 = A; ~ b u (9.6) b f
1
9.1 Formulation of the Model Problem
9.1.1 Preconditioning the Basic Matrix
It is standard practice in applying relaxation procedures to precondition the basic equation. This preconditioning has the e ect of multiplying Eq. 9.5 from the left by some nonsingular matrix. In the simplest possible case the conditioning matrix is a diagonal matrix composed of a constant D(b). If we designate the conditioning matrix by C , the problem becomes one of solving for ~ in u
CAb~ ; C~ b = 0 u f
Notice that the solution of Eq. 9.7 is
(9.7)
~ = CAb ]; C~ b = A; C ; C~ b = A; ~ b u f f b b f
1 1 1 1 1
(9.8)
which is identical to the solution of Eq. 9.5, provided C ; exists. In the following we will see that our approach to the iterative solution of Eq. 9.7 depends crucially on the eigenvalue and eigenvector structure of the matrix CAb, and, equally important, does not depend at all on the eigensystem of the basic matrix Ab . For example, there are wellknown techniques for accelerating relaxation schemes if the eigenvalues of CAb are all real and of the same sign. To use these schemes, the
9.1. FORMULATION OF THE MODEL PROBLEM
165
conditioning matrix C must be chosen such that this requirement is satis ed. A choice of C which ensures this condition is the negative transpose of Ab. For example, consider a spatial discretization of the linear convection equation using centered di erences with a Dirichlet condition on the left side and no constraint on the right side. Using a rstorder backward di erence on the right side (as in Section 3.6), this leads to the approximation
82 > 0 1 >6 > 0 6 1 <6 ;1 ;1 1 1 ~= 0 u 2 x >6 x 6 >6 ;1 0 >4 : ;2
3 2 u 7 6 ;0 a 7 6 7~ 6 7u+6 0 7 6 17 6 0 5 4
2 0
39 > 7> 7> 7= 7 7> 7> 5>
(9.9)
The matrix in Eq. 9.9 has eigenvalues whose imaginary parts are much larger than their real parts. It can rst be conditioned so that the modulus of each element is 1. This is accomplished using a diagonal preconditioning matrix
2 61 60 6 D = 2 x6 0 6 60 4
0 1 0 0 0 0
0 0 1 0 0
0 0 0 1 0
0 0 0 0
1 2
3 7 7 7 7 7 7 5
(9.10)
which scales each row. We then further condition with multiplication by the negative transpose. The result is
A = ;AT A =
2 1 1
2 0 6 ;1 1 1 6 0 6 6 ;1 0 6 6 ;1 4
1 0 ;1 ;1
32 0 7 6 ;1 1 1 76 0 76 76 ;1 0 76 176 ;1 54 3 7 7 7 17 7 17 5
1 0 ;1
3 7 7 7 7 7 17 5
1
2 0 6 ;1 ;2 1 1 6 0 0 6 = 6 1 0 ;2 0 6 6 1 0 ;2 4
1
(9.11)
1 ;2
166
1 1
CHAPTER 9. RELAXATION METHODS
1 1
If we de ne a permutation matrix P and carry out the process P T ;AT A ]P (which just reorders the elements of A and doesn't change the eigenvalues) we nd
2 60 60 6 60 6 60 4
1 0 0 0 1 0
0 0 0 1 0
0 1 0 0 0
0 7 6 ;1 0 1 76 76 0 7 6 0 ;2 0 1 76 7 6 1 0 ;2 0 1 7 6 76 176 76 76 054 1 0 ;2 1 7 6 54 0 1 1 ;2
32
32
0 1 0 0 0
0 0 0 1 0
0 0 0 0 1
0 0 1 0 0
17 07 7 07 7 07 5 0
3
1 ;1 which has all negative real eigenvalues, as given in Appendix B. Thus even when the basic matrix Ab has nearly imaginary eigenvalues, the conditioned matrix ;AT Ab b is nevertheless symmetric negative de nite (i.e., symmetric with negative real eigenvalues), and the classical relaxation methods can be applied. We do not necessarily recommend the use of ;AT as a preconditioner we simply wish to show that a broad b range of matrices can be preconditioned into a form suitable for our analysis.
2 1 6 ;2 ;2 1 6 1 6 1 ;2 1 =6 6 6 1 ;2 4
3 7 7 7 7 7 17 5
(9.12)
9.1.2 The Model Equations
Preconditioning processes such as those described in the last section allow us to prepare our algebraic equations in advance so that certain eigenstructures are guaranteed. In the remainder of this chapter, we will thoroughly investigate some simple equations which model these structures. We will consider the preconditioned system of equations having the form
A~ ; ~ = 0 f (9.13) where A is symmetric negative de nite. The symbol for the dependent variable has been changed to as a reminder that the physics being modeled is no longer time
2
A permutation matrix (de ned as a matrix with exactly one 1 in each row and column and has the property that P T = P ;1 ) just rearranges the rows and columns of a matrix. 2 We use a symmetric negative de nite matrix to simplify certain aspects of our analysis. Relaxation methods are applicable to more general matrices. The classical methods will usually converge if Ab is diagonally dominant, as de ned in Appendix A.
1
9.1. FORMULATION OF THE MODEL PROBLEM
1
167
accurate when we later deal with ODE formulations. Note that the solution of Eq. 9.13, ~ = A; ~ , is guaranteed to exist because A is nonsingular. In the notation of f Eqs. 9.5 and 9.7,
A = CAb and ~ = C~ b f f (9.14) The above was written to treat the general case. It is instructive in formulating the concepts to consider the special case given by the di usion equation in one dimension with unit di usion coe cient : @u = @ u ; g(x) @t @x This has the steadystate solution
2 2 2 2
(9.15)
@ u = g(x) (9.16) @x which is the onedimensional form of the Poisson equation. Introducing the threepoint central di erencing scheme for the second derivative with Dirichlet boundary conditions, we nd d~ = 1 B (1 ;2 1)~ + (bc) ; ~ u u ~ g (9.17) dt x ~ where (bc) contains the boundary conditions and ~ contains the values of the source g term at the grid nodes. In this case
2
Ab = 1 B (1 ;2 1) x ~ fb = ~ ; (bc) g ~ (9.18) Choosing C = x I , we obtain B (1 ;2 1)~ = ~ f (9.19) ~ where ~ = x fb. If we consider a Dirichlet boundary condition on the left side and f either a Dirichlet or a Neumann condition on the right side, then A has the form A=B 1~ 1 b
2 2 2
168
CHAPTER 9. RELAXATION METHODS
~ = ;2 ;2 b
;2 s]T
s = ;2 or ;1 (9.20) Note that s = ;1 is easily obtained from the matrix resulting from the Neumann boundary condition given in Eq. 3.24 using a diagonal conditioning matrix. A tremendous amount of insight to the basic features of relaxation is gained by an appropriate study of the onedimensional case, and much of the remaining material is devoted to this case. We attempt to do this in such a way, however, that it is directly applicable to two and threedimensional problems.
9.2 Classical Relaxation
9.2.1 The Delta Form of an Iterative Scheme
+1
We will consider relaxation methods which can be expressed in the following delta form: h i H ~ n ; ~ n = A~ n ; ~ f (9.21) where H is some nonsingular matrix which depends upon the iterative method. The matrix H is independent of n for stationary methods and is a function of n for nonstationary ones. The iteration count is designated by the subscript n or the superscript (n). The converged solution is designated ~ 1 so that
~ 1 = A; ~ f
1
(9.22)
9.2.2 The Converged Solution, the Residual, and the Error
+1 +1 1 1 1
Solving Eq. 9.21 for ~ n gives ~ n = I + H ; A]~ n ; H ; ~ = G~ n ; H ; ~ f f (9.23) where G I + H; A (9.24) Hence it is clear that H should lead to a system of equations which is easy to solve, or at least easier to solve than the original system. The error at the nth iteration is de ned as ~ n ~ n ; ~ 1 = ~ n ; A; ~ e f (9.25)
1 1
9.2. CLASSICAL RELAXATION
169
where ~ 1 was de ned in Eq. 9.22. The residual at the nth iteration is de ned as ~ n A~ n ; ~ r f (9.26) Multiply Eq. 9.25 by A from the left, and use the de nition in Eq. 9.26. There results the relation between the error and the residual A~ n ; ~ n = 0 e r (9.27) Finally, it is not di cult to show that ~ n = G~ n e e
+1
(9.28)
Consequently, G is referred to as the basic iteration matrix, and its eigenvalues, which we designate as m , determine the convergence rate of a method. In all of the above, we have considered only what are usually referred to as stationary processes in which H is constant throughout the iterations. Nonstationary processes in which H (and possibly C ) is varied at each iteration are discussed in Section 9.5.
9.2.3 The Classical Methods
Point Operator Schemes in One Dimension
Let us consider three classical relaxation procedures for our model equation B (1 ;2 1)~ = ~ f (9.29) as given in Section 9.1.2. The PointJacobi method is expressed in point operator form for the onedimensional case as n = 1 h n + n ;f i (9.30) j j j 2 j; This operator comes about by choosing the value of jn such that together with the old values of j; and j , the j th row of Eq. 9.29 is satis ed. The GaussSeidel method is n = 1h n + n ; f i (9.31) j j j 2 j; This operator is a simple extension of the pointJacobi method which uses the most recent update of j; . Hence the j th row of Eq. 9.29 is satis ed using the new values of j and j ; and the old value of j . The method of successive overrelaxation (SOR)
( +1) ( ) 1 ( ) +1 ( +1) 1 +1 ( +1) ( +1) 1 ( ) +1 1 1 +1
170
CHAPTER 9. RELAXATION METHODS
is based on the idea that if the correction produced by the GaussSeidel method tends to move the solution toward ~ 1, then perhaps it would be better to move further in this direction. It is usually expressed in two steps as h i ~j = 1 jn + jn ; fj 2 ; h i n = jn + ! ~j ; jn (9.32) j
( +1) 1 ( ) +1 ( +1) ( ) ( )
where ! generally lies between 1 and 2, but it can also be written in the single line n = ! n + (1 ; ! ) n + ! n ; ! f (9.33) j j 2 j; 2 j 2 j
( +1) ( +1) 1 ( ) ( ) +1
The General Form
The general form of the classical methods is obtained by splitting the matrix A in Eq. 9.13 into its diagonal, D, the portion of the matrix below the diagonal, L, and the portion above the diagonal, U , such that
A=L+D+U
(9.34)
Then the pointJacobi method is obtained with H = ;D, which certainly meets the criterion that it is easy to solve. The GaussSeidel method is obtained with H = ;(L + D), which is also easy to solve, being lower triangular.
9.3 The ODE Approach to Classical Relaxation
The particular type of delta form given by Eq. 9.21 leads to an interpretation of relaxation methods in terms of solution techniques for coupled rstorder ODE's, about which we have already learned a great deal. One can easily see that Eq. 9.21 results from the application of the explicit Euler timemarching method (with h = 1) to the following system of ODE's: ~ H d = A~ ; ~ f (9.35) dt This is equivalent to d~ = H ; C A ~ ; ~ = H ; A~ ; ~ ] fb f (9.36) b dt
1 1
9.3.1 The Ordinary Di erential Equation Formulation
9.3. THE ODE APPROACH TO CLASSICAL RELAXATION
1 1 1
171
In the special case where H ; A depends on neither ~ nor t, H ; ~ is also independent u f ; A are linearly independent, the solution can be of t, and the eigenvectors of H written as
~ = c e 1 t~ + + cM e M t~ M +~ 1 x} (9.37)  x {z error where what is referred to in timeaccurate analysis as the transient solution, is now referred to in relaxation analysis as the error. It is clear that, if all of the eigenvalues of H ; A have negative real parts (which implies that H ; A is nonsingular), then the system of ODE's has a steadystate solution which is approached as t ! 1, given by
1 1 1 1
~ 1 = A; ~ f
1
(9.38)
which is the solution of Eq. 9.13. We see that the goal of a relaxation method is to remove the transient solution from the general solution in the most e cient way possible. The eigenvalues are xed by the basic matrix in Eq. 9.36, the preconditioning matrix in 9.7, and the secondary conditioning matrix in 9.35. The eigenvalues are xed for a given h by the choice of timemarching method. Throughout the remaining discussion we will refer to the independent variable t as \time", even though no true time accuracy is involved. In a stationary method, H and C in Eq. 9.36 are independent of t, that is, they are not changed throughout the iteration process. The generalization of this in our approach is to make h, the \time" step, a constant for the entire iteration. Suppose the explicit Euler method is used for the time integration. For this method m = 1 + m h. Hence the numerical solution after n steps of a stationary relaxation method can be expressed as (see Eq. 6.28)
~ n = c ~ (1 + h)n + x
1 1 1
+ cm~ m (1 + mh)n + x {z error
+ cM ~ M (1 + x
n M h) } +~ 1 (9.39)
The initial amplitudes of the eigenvectors are given by the magnitudes of the cm . These are xed by the initial guess. In general it is assumed that any or all of the eigenvectors could have been given an equally \bad" excitation by the initial guess, so that we must devise a way to remove them all from the general solution on an equal basis. Assuming that H ; A has been chosen (that is, an iteration process has been decided upon), the only free choice remaining to accelerate the removal of the error terms is the choice of h. As we shall see, the three classical methods have all been conditioned by the choice of H to have an optimum h equal to 1 for a stationary iteration process.
1
172
CHAPTER 9. RELAXATION METHODS
9.3.2 ODE Form of the Classical Methods
The three iterative procedures de ned by Eqs. 9.30, 9.31 and 9.32 obey no apparent pattern except that they are easy to implement in a computer code since all of the data required to update the value of one point are explicitly available at the time of the update. Now let us study these methods as subsets of ODE as formulated in Section 9.3.1. Insert the model equation 9.29 into the ODE form 9.35. Then ~ H d = B (1 ;2 1)~ ; ~ f (9.40) dt As a start, let us use for the numerical integration the explicit Euler method 0 (9.41) n = n+h n with a step size, h, equal to 1. We arrive at H (~ n ; ~ n) = B (1 ;2 1)~ n ; ~ f (9.42)
+1 +1
It is clear that the best choice of H from the point of view of matrix algebra is ;B (1 ;2 1) since then multiplication from the left by ;B ; (1 ;2 1) gives the correct answer in one step. However, this is not in the spirit of our study, since multiplication by the inverse amounts to solving the problem by a direct method without iteration. The constraint on H that is in keeping with the formulation of the three methods described in Section 9.2.3 is that all the elements above the diagonal (or below the diagonal if the sweeps are from right to left) are zero. If we impose this constraint and further restrict ourselves to banded tridiagonals with a single constant in each band, we are led to 2 B (; ! 0)(~ n ; ~ n) = B (1 ;2 1)~ n ; ~ f (9.43) where and ! are arbitrary. With this choice of notation the three methods presented in Section 9.2.3 can be identi ed using the entries in Table 9.1.
1 +1
TABLE 9.1: VALUES OF and ! IN EQ. 9.43 THAT LEAD TO CLASSICAL RELAXATION METHODS
!
0 1 1
Method
Equation 6:2:3 6:2:4 6:2:5
1 PointJacobi 1 h i GaussSeidel 2= 1 + sin M + 1 Optimum SOR
9.4. EIGENSYSTEMS OF THE CLASSICAL METHODS
173
The fact that the values in the tables lead to the methods indicated can be veri ed by simple algebraic manipulation. However, our purpose is to examine the whole procedure as a special subset of the theory of ordinary di erential equations. In this light, the three methods are all contained in the following set of ODE's d~ = B ; (; 2 0) B (1 ;2 1)~ ; ~ f (9.44) dt ! and appear from it in the special case when the explicit Euler method is used for its numerical integration. The point operator that results from the use of the explicit Euler scheme is ! ! n = ! n + ! (h ; ) n ; (!h ; 1) n + !h n ; !h f (9.45) j j; j 2 j; 2 2 j 2 j
1 ( +1) ( +1) 1 ( ) 1 ( ) ( ) +1
This represents a generalization of the classical relaxation techniques.
9.4 Eigensystems of the Classical Methods
The ODE approach to relaxation can be summarized as follows. The basic equation to be solved came from some timeaccurate derivation Ab~ ; ~ b = 0 u f (9.46) This equation is preconditioned in some manner which has the e ect of multiplication by a conditioning matrix C giving A~ ; ~ = 0 f (9.47) An iterative scheme is developed to nd the converged, or steadystate, solution of the set of ODE's ~ H d = A~ ; ~ f (9.48) dt This solution has the analytical form ~ n = ~n + ~ 1 e (9.49) where ~ n is the transient, or error, and ~ 1 A; ~ is the steadystate solution. The e f three classical methods, PointJacobi, GaussSeidel, and SOR, are identi ed for the onedimensional case by Eq. 9.44 and Table 9.1.
1
174
CHAPTER 9. RELAXATION METHODS
1
Given our assumption that the component of the error associated with each eigenvector is equally likely to be excited, the asymptotic convergence rate is determined by the eigenvalue m of G ( I + H ; A) having maximum absolute value. Thus Convergence rate
j mjmax
m=1 2
1
M
(9.50)
In this section, we use the ODE analysis to nd the convergence rates of the three classical methods represented by Eqs. 9.30, 9.31, and 9.32. It is also instructive to inspect the eigenvectors and eigenvalues in the H ; A matrix for the three methods. This amounts to solving the generalized eigenvalue problem A~ m = m H~ m x x (9.51) for the special case 2 0)~ (9.52) ! xm The generalized eigensystem for simple tridigonals is given in Appendix B.2. The three special cases considered below are obtained with a = 1, b = ;2, c = 1, d = ; , e = 2=!, and f = 0. To illustrate the behavior, we take M = 5 for the matrix order. This special case makes the general result quite clear.
B (1 ;2 1)~ m = x
m B (;
9.4.1 The PointJacobi System
1 If = 0 and ! = 1 in Eq. 9.44, the ODE matrix H ; A reduces to simply B ( 1 ;1 2 ). 2 The eigensystem can be determined from Appendix B.1 since both d and f are zero. The eigenvalues are given by the equation m (9.53) m = ;1 + cos M +1 m = 1 2 ::: M The  relation for the explicit Euler method is m = 1 + m h. This relation can be plotted for any h. The plot for h = 1, the optimum stationary case, is shown in Fig. 9.1. For h < 1, the maximum j mj is obtained with m = 1, Fig. 9.2 and for h > 1, the maximum j m j is obtained with m = M , Fig. 9.3. Note that for h > 1:0 (depending on M ) there is the possibility of instability, i.e. j m j > 1:0. To obtain the optimal scheme we wish to minimize the maximum j m j which occurs when h = 1, j j = j M j and the best possible convergence rate is achieved:
1 1
j m jmax = cos M + 1
(9.54)
9.4. EIGENSYSTEMS OF THE CLASSICAL METHODS
1
175
For M = 40, we obtain j mjmax = 0:9971. Thus after 500 iterations the error content associated with each eigenvector is reduced to no more than 0.23 times its initial level. Again from Appendix B.1, the eigenvectors of H ; A are given by ~ j = (xj )m = sin j m x j = 1 2 ::: M (9.55) M +1 This is a very \wellbehaved" eigensystem with linearly independent eigenvectors and distinct eigenvalues. The rst 5 eigenvectors are simple sine waves. For M = 5, the eigenvectors can be written as
2 1=2 3 2 3 2 p3=2 3 p 7 p 7 6 3=2 7 6 1 7 6 3=2 7 6 6 0 7 6 ~ = 6 p 1 7 ~ = 6 p 7 ~ = 6 ;1 7 6 7 x 6 0 7 x 6 7 x 6 6 7 6 6 3=2 7 6 0 7 7 6 ; 3=2 7 4 5 4 5 5 4 p 7 1 1=2 ; 3=2 3 2 p 3 2 3 1= p =2 7 p2 7 6 ; 3=2 7 6 6 6 7 ~ 6 ; 3=2 7 ~ = 6 p0 7 x = 6 p 7 6 1 7 x 6 6 7 6 3=2 7 7 6 ; 3=2 7 5 4 p 5 4 1=2 ; 3=2
1 2 3 4 5
(9.56)
The corresponding eigenvalues are, from Eq. 9.53
1
2
3
= ;1 + 23 = ;0:134 = ;1 + 1 = ;0:5 2 = ;1 = ;1:0 = ;1 ; 1 = ;1:5 2
p
(9.57)
4
= ;1 ; 23 = ;1:866 From Eq. 9.39, the numerical solution written in full is
5
p
~ n ; ~ 1 = c 1 ; (1 ; 3 )h]n~ x 2 1 x + c 1 ; (1 ; 2 )h]n~
1 2 2
p
1
176
1.0
m=5
CHAPTER 9. RELAXATION METHODS
h = 1.0
m=1
Values are equal
σm
m=4 m=2
Μ=5
0.0 2.0
m=3
1.0
0.0
Figure 9.1: The
relation for PointJacobi, h = 1 M = 5. + c 1 ; (1
3 4
+ c 1 ; (1 + 1 )h]n~ 2 x p + c 1 ; (1 + 23 )h]n~ x
4 5
)h]n~ x
3
5
(9.58)
9.4.2 The GaussSeidel System
If and ! are equal to 1 in Eq. 9.44, the matrix eigensystem evolves from the relation
B (1 ;2 1)~ m = x
m B (;1
2 0)~ m x
(9.59)
1
which can be studied using the results in Appendix B.2. One can show that the H ; A matrix for the GaussSeidel method, AGS , is
AGS
B ; (;1 2 0)B (1 ;2 1) =
1
9.4. EIGENSYSTEMS OF THE CLASSICAL METHODS
Approaches 1.0
177
1.0
h < 1.0
h
0.0 2.0
De
cr
ea
h
si
ng
σm
ng
si
Figure 9.2: The
relation for PointJacobi, h = 0:9 M = 5.
Exceeds 1.0 at h
De cr ea
1.0
0.0
λ mh
∼ ∼
1.072 / Ustable h > 1.072
1.0
h > 1.0
ng si ea cr In
In
cr
ea
si
σm
ng
h
h
M = 5
0.0 2.0
1.0
0.0
λmh
Figure 9.3: The
relation for PointJacobi, h = 1:1 M = 5.
principal vectors. The equation for the nondefective eigenvalues in the ODE matrix is (for odd M ) + m ) m = 1 2 ::: M2 1 (9.61) m = ;1 + cos ( M +1 and the corresponding eigenvectors are given by j; m + ~ m = cos m x sin j M + 1 m = 1 2 ::: M2 1 (9.62) M +1 The  relation for h = 1, the optimum stationary case, is shown in Fig. 9.4. The m with the largest amplitude is obtained with m = 1. Hence the convergence rate is
2 1
CHAPTER 9. RELAXATION METHODS 2 ;1 1=2 3 1 6 0 ;3=4 1=2 7 2 6 7 6 0 1=8 ;3=4 1=2 7 3 6 7 6 7 6 0 1=16 1=8 ;3=4 1=2 7 4 6 7 (9.60) 6 0 1=32 1=16 1=8 ;3=4 1=2 7 5 6 7 6 . ... 6 .. 7 . . . . . . . . . . . . . . . 7 ... 4 5 0 1=2M M The eigenvector structure of the GaussSeidel ODE matrix is quite interesting. If M is odd there are (M + 1)=2 distinct eigenvalues with corresponding linearly independent eigenvectors, and there are (M ; 1)=2 defective eigenvalues with corresponding
178
Since this is the square of that obtained for the PointJacobi method, the error associated with the \worst" eigenvector is removed in half as many iterations. For M = 40, j m jmax = 0:9942. 250 iterations are required to reduce the error component of the worst eigenvector by a factor of roughly 0.23. The eigenvectors are quite unlike the PointJacobi set. They are no longer symmetrical, producing waves that are higher in amplitude on one side (the updated side) than they are on the other. Furthermore, they do not represent a common family for di erent values of M . The Jordan canonical form for M = 5 is
j m jmax = cos M + 1
2
(9.63)
3 2 h~ i 7 6 h~ i 7 6 6 7 2~ 37 ; A X =J = 6 7 6 X GS GS 1 6 7 6 6 ~ 77 6 6 1 77 4 4 55 ~
1 1 2 3 3 3
(9.64)
9.4. EIGENSYSTEMS OF THE CLASSICAL METHODS
1.0
179
h = 1.0
σm Μ=5
2 defective λ m 0.0 2.0 1.0
0.0
λmh
Figure 9.4: The
relation for GaussSeidel, h = 1:0 M = 5.
The eigenvectors and principal vectors are all real. For M = 5 they can be written
2 3 2 p 3 2 3 2 3 2 3 3=2 1=2 7 17 0 7 6 3=4 7 6 p3=4 7 607 6 2 7 6 0 7 6 6 7 6 6 6 0 7 ~ = 6 3=4 7 ~ = 6 p0 7 ~ = 6 0 7 ~ = 6 ;1 7 ~ = 6 4 7 (9.65) 6 7 x 6 7 x 6 7 x 6 7 x 6 7 x 6 6 6 7 6 7 6 7 6 9=16 7 7 6 ; 3=16 7 7 607 6 0 7 6 ;4 7 4 5 4 p 5 4 5 4 5 4 5 9=32 0 0 1 ; 3=32
1 2 3 4 5
The corresponding eigenvalues are = ;1=4 = ;3=4 ) = ;1 (4) Defective, linked to (5) Jordan block The numerical solution written in full is thus ~ n ; ~ 1 = c (1 ; h )n~ 4 x h x + c (1 ; 34 )n~
1 2 3 3 1 1 2 2
(9.66)
180
CHAPTER 9. RELAXATION METHODS " # n + c h n (1 ; h)n; + c h n(n ; 1) (1 ; h)n; ~ + c (1 ; h) x 1! 2! + c (1 ; h)n + c h n (1 ; h)n; ~ x
3 4 1 5 2 2 3 4 5
+ c (1 ; h)n~ x
5
1!
1
4
5
(9.67)
9.4.3 The SOR System
1
If = 1 and 2=! = x in Eq. 9.44, the ODE matrix is B ; (;1 x 0)B (1 ;2 1). One can show that this can be written in the form given below for M = 5. The generalization to any M is fairly clear. The H ; A matrix for the SOR method, ASOR B ; (;1 x 0)B (1 ;2 1), is
1 1
2 x 0 0 6 ;2;2x x ; x 0 6 1 6 ;2x + x x x 2x2x x 6 x + ; + x ; 2x x x 6 ;2x + x x ; 2x + x x ; 2x + x 6 x ; 2x 4 ;2 + x 1 ; 2x + x x ; 2x + x x ; 2x + x x
4 4 3 4 3 4 4 5 2 3 2 3 4 3 4 4 2 2 3 2 3 4 3 4 2 2 3 2 3 4
3
; 2x
0 0 0 x
4
4
3 7 7 7 7 (9.68) 7 7 5
Eigenvalues of the system are given by
m
+ = ;1 + !pm 2 zm
2
m = 1 2 :::M
2 2 1 2
(9.69)
where
zm = 4(1 ; !) + ! pm] = pm = cos m =(M + 1)]
2 2 2 1
If ! = 1, the system is GaussSeidel. If 4(1 ; !) + ! pm < 0 zm and m are complex. If ! is chosen such that 4(1 ; !) + ! p = 0 ! is optimum for the stationary case, and the following conditions hold: 1. Two eigenvalues are real, equal and defective. 2. If M is even, the remaining eigenvalues are complex and occur in conjugate pairs. 3. If M is odd, one of the remaining eigenvalues is real and the others are complex occurring in conjugate pairs.
9.4. EIGENSYSTEMS OF THE CLASSICAL METHODS
One can easily show that the optimum ! for the stationary case is and for ! = !opt
181 (9.70)
!opt = 2= 1 + sin M + 1 ~m x
m
= =
m;1 j ;1 m sin
2
where
m j M +1
(9.71)
Using the explicit Euler method to integrate the ODE's, m = 1 ; h + h m , and if h = 1, the optimum value for the stationary case, the  relation reduces to that shown in Fig. 9.5. This illustrates the fact that for optimum stationary SOR all the j m j are identical and equal to !opt ; 1. Hence the convergence rate is j m jmax = !opt ; 1 (9.72) !opt = 2= 1 + sin M + 1 For M = 40, j mjmax = 0:8578. Hence the worst error component is reduced to less than 0.23 times its initial value in only 10 iterations, much faster than both GaussSeidel and PointJacobi. In practical applications, the optimum value of ! may have to be determined by trial and error, and the bene t may not be as great. For odd M , there are two real eigenvectors and one real principal vector. The remaining linearly independent eigenvectors are all complex. For M = 5 they can be written 2 3 2 3 2 p3(1 3 2 3 1=2 7 ;6 7 1 7 p p )=2 7 6 1=2 7 6 6 3(1 i 2)=6 7 6 6 6 6 7~ 6 9 7~ 6 7 6 7~ 6 0 7 ~ = 6 1=3 7 x = 6 16 7 x = 6 p 7 x = 6 1=3 7 (9.73) 6 7 0p x 6 7 6 7 6 6 6 1=6 7 6 13 7 6 3(5 i 2)=54 7 7 6 0 7 7 4 5 4 5 4p 5 4 5 p 1=18 6 1=9 3(7 4i 2)=162 The corresponding eigenvalues are = ;2=3 (2) Defectivep linked to = ;(10 ; 2p2i)=9 = ;(10 + 2 2i)=9 = ;4=3 (9.74)
2 1 2 34 5 1 1 3 4 5
!opt p + i p ; p m= m 2 m
2 1 2
q
182
1.0
CHAPTER 9. RELAXATION METHODS
h = 1.0
2 real defective λ m
σm
2 complex λ m
1 real λ m
Μ=5
0.0 2.0 1.0
0.0
λmh
Figure 9.5: The
relation for optimum stationary SOR, M = 5, h = 1.
The numerical solution written in full is
~n ; ~1 = + + + +
c (1 ; 2h=3)n + c nh(1 ; 2h=3)n; ]~ x c (1 ; 2h=3)n~ x p c 1 ; (10 ; 2p2i)h=9]n~ x n~ c 1 ; (10 + 2 2i)h=9] x c (1 ; 4h=3)n~ x
1 2 1 2 3 2 3 4 5 4 5
1
(9.75)
9.5 Nonstationary Processes
In classical terminology a method is said to be nonstationary if the conditioning matrices, H and C , are varied at each time step. This does not change the steadystate solution A; ~ b , but it can greatly a ect the convergence rate. In our ODE b f approach this could also be considered and would lead to a study of equations with nonconstant coe cients. It is much simpler, however, to study the case of xed H and C but variable step size, h. This process changes the PointJacobi method to Richardson's method in standard terminology. For the GaussSeidel and SOR methods it leads to processes that can be superior to the stationary methods. The nonstationary form of Eq. 9.39 is
1
9.5. NONSTATIONARY PROCESSES
N ~ N = c ~ Y (1 + hn ) + x
1 1
183 + cm~ m x
M hn ) + ~ 1 N Y n=1
n=1
1
(1 +
m hn )
+
+ cM ~ M x
N Y n=1
(1 +
(9.76)
where the symbol stands for product. Since hn can now be changed at each step, the error term can theoretically be completely eliminated in M steps by taking hm = ;1= m , for m = 1 2 M . However, the eigenvalues m are generally unknown and costly to compute. It is therefore unnecessary and impractical to set hm = ;1= m for m = 1 2 : : : M . We will see that a few well chosen h's can reduce whole clusters of eigenvectors associated with nearby 's in the m spectrum. This leads to the concept of selectively annihilating clusters of eigenvectors from the error terms as part of a total iteration process. This is the basis for the multigrid methods discussed in Chapter 10. Let us consider the very important case when all of the m are real and negative (remember that they arise from a conditioned matrix so this constraint is not unrealistic for quite practical cases). Consider one of the error terms taken from
M N ~ N ~ N ; ~ 1 = X cm~ m Y (1 + e x m=1 n=1 m hn )
(9.77)
and write it in the form
cm~ m Pe( m ) cm~ m x x
N Y n=1
(1 +
m hn )
(9.78)
where Pe signi es an \Euler" polynomial. Now focus attention on the polynomial (Pe)N ( ) = (1 + h )(1 + h )
1 2
(1 + hN )
(9.79)
treating it as a continuous function of the independent variable . In the annihilation process mentioned after Eq. 9.76, we considered making the error exactly zero by taking advantage of some knowledge about the discrete values of m for a particular case. Now we pose a less demanding problem. Let us choose the hn so that the maximum value of (Pe)N ( ) is as small as possible for all lying between a and b such that b a 0. Mathematically stated, we seek
b
max a j(Pe)N ( )j = minimum
with(Pe)N (0) = 1
(9.80)
184
CHAPTER 9. RELAXATION METHODS
a; b TN a; b ! ; TN ; a; b a b
This problem has a well known solution due to Markov. It is 2 ;
!
(Pe)N ( ) = where
(9.81)
TN (y) = cos(N arccos y)
are the Chebyshev polynomials along the interval ;1 y 1 and
2 2
(9.82)
q q N N 1 TN (y) = 2 y + y ; 1 + 1 y ; y ; 1 (9.83) 2 are the Chebyshev polynomials for jyj > 1. In relaxation terminology this is generally referred to as Richardson's method, and it leads to the nonstationary step size choice given by ( " #) 1 = 1 ; ; + ( ; ) cos (2n ; 1) n = 1 2 :::N (9.84) b a h 2 b a 2N
n
Remember that all are negative real numbers representing the magnitudes of m in an eigenvalue spectrum. The error in the relaxation process represented by Eq. 9.76 is expressed in terms of a set of eigenvectors, ~ m , ampli ed by the coe cients cm Q(1 + m hn). With each x eigenvector there is a corresponding eigenvalue. Eq. 9.84 gives us the best choice of a series of hn that will minimize the amplitude of the error carried in the eigenvectors associated with the eigenvalues between b and a . As an example for the use of Eq. 9.84, let us consider the following problem: Minimize the maximum error associated with the eigenvalues in the interval ;2 ;1 using only 3 iterations. The three values of h which satisfy this problem are (9.85)
hn = 2= 3 ; cos (2n ; 1) 6
"
#!
(9.86)
9.5. NONSTATIONARY PROCESSES
and the amplitude of the eigenvector is reduced to (Pe) ( ) = T (2 + 3)=T (3)
3 3 3
185 (9.87) (9.88)
where
n p p o T (3) = 3 + 8] + 3 ; 8] =2 99
3 3 3
A plot of Eq. 9.87 is given in Fig. 9.6 and we see that the amplitudes of all the eigenvectors associated with the eigenvalues in the range ;2 ;1 have been reduced to less than about 1% of their initial values. The values of h used in Fig. 9.6 are
h = 4=(6 ; 3) h = 4=(6 ; p 0) h = 4=(6 + 3)
1 2 3
p
Return now to Eq. 9.76. This was derived from Eq. 9.37 on the condition that the explicit Euler method, Eq. 9.41, was used to integrate the basic ODE's. If instead the implicit trapezoidal rule
n+1
=
n+
is used, the nonstationary formula
1 h( 0 + 0 ) n 2 n
+1
(9.89)
would result. This calls for a study of the rational \trapezoidal" polynomial, Pt: 0 1 1 N B 1 + hn C Y (Pt )N ( ) = B 2 C (9.91) @ 1 A n 1 ; 2 hn under the same constraints as before, namely that
=1
0 1 M N 1+ h ~ N = X cm~ m Y B 2 n B x @ 1 m n 1 ; 2 hn
=1 =1
1 mC C + ~1 A
m
(9.90)
b
max a j(Pt)N ( )j = minimum , with (Pt)N (0) = 1
(9.92)
186
CHAPTER 9. RELAXATION METHODS
1 0.9 0.8 0.7 0.6
(Pe )3 (λ)
0.5 0.4 0.3 0.2 0.1 0 −2 −1.8 −1.6 −1.4 −1.2 −1
λ
−0.8
−0.6
−0.4
−0.2
0
0.02
0.015
0.01
0.005
(P ) (λ)
e 3
0
−0.005
−0.01
−0.015
−0.02 −2
−1.8
−1.6
−1.4
−1.2
−1
λ
−0.8
−0.6
−0.4
−0.2
0
Figure 9.6: Richardson's method for 3 steps, minimization over ;2
;1.
9.6. PROBLEMS !
187
The optimum values of h can also be found for this problem, but we settle here for the approximation suggested by Wachspress
n; = N ; 2 =; a n=1 2 N (9.93) b hn b This process is also applied to problem 9.85. The results for (Pt) ( ) are shown in Fig. 9.7. The error amplitude is about 1/5 of that found for (Pe) ( ) in the same interval of . The values of h used in Fig. 9.7 are h = 1 p h = 2 h = 2
( 1) ( 1) 3 3 1 2 3
9.6 Problems
1. Given a relaxation method in the form ~ H ~n = A~n ; f show that
1
~n = Gn ~ + (I ; Gn)A; f ~
0 1 1 2
where G = I + H ; A. 2. For a linear system of the form (A + A )x = b, consider the iterative method (I + A )~ = (I ; A )xn + b x (I + A )xn = (I ; A )~ + b x where is a parameter. Show that this iterative method can be written in the form H (xk ; xk ) = (A + A )xk ; b
1 2 2 +1 1 +1 1 2
Determine the iteration matrix G if = ;1=2. 3. Using Appendix B.2, nd the eigenvalues of H ; A for the SOR method with A = B (4 : 1 ;2 1) and ! = !opt. (You do not have to nd H ; . Recall that the eigenvalues of H ; A satisfy A~ m = mH~ m .) Find the numerical values, x x not just the expressions. Then nd the corresponding j m j values.
1 1 1
188
CHAPTER 9. RELAXATION METHODS
1 0.9 0.8 0.7 0.6
(P ) (λ)
t 3
0.5 0.4 0.3 0.2 0.1 0 −2 −1.8 −1.6 −1.4 −1.2 −1
λ
−0.8
−0.6
−0.4
−0.2
0
2
x 10
−3
1.5
1
0.5
(Pt )3 (λ)
0
−0.5
−1
−1.5
−2 −2
−1.8
−1.6
−1.4
−1.2
−1
λ
−0.8
−0.6
−0.4
−0.2
0
Figure 9.7: Wachspress method for 3 steps, minimization over ;2
;1.
9.6. PROBLEMS
189
4. Solve the following equation on the domain 0 x 1 with boundary conditions u(0) = 0, u(1) = 1:
@ u ; 6x = 0 @x For the initial condition, use u(x) = 0. Use secondorder centered di erences on a grid with 40 cells (M = 39). Iterate to steady state using (a) the pointJacobi method, (b) the GaussSeidel method, (c) the SOR method with the optimum value of !, and (d) the 3step Richardson method derived in Section 9.5. Plot the solution after the residual is reduced by 2, 3, and 4 orders of magnitude. Plot the logarithm of the L norm of the residual vs. the number of iterations. Determine the asymptotic convergence rate. Compare with the theoretical asymptotic convergence rate.
2 2 2
190
CHAPTER 9. RELAXATION METHODS
Chapter 10 MULTIGRID
The idea of systematically using sets of coarser grids to accelerate the convergence of iterative schemes that arise from the numerical solution to partial di erential equations was made popular by the work of Brandt. There are many variations of the process and many viewpoints of the underlying theory. The viewpoint presented here is a natural extension of the concepts discussed in Chapter 9.
10.1 Motivation
10.1.1 Eigenvector and Eigenvalue Identi cation with Space Frequencies Consider the eigensystem of the model matrix B (1 ;2 1). The eigenvalues and
eigenvectors are given in Sections 4.3.2 and 4.3.3, respectively. Notice that as the magnitudes of the eigenvalues increase, the spacefrequency (or wavenumber) of the corresponding eigenvectors also increase. That is, if the eigenvalues are ordered such that
j j j j
1 2 2
j Mj
(10.1)
then the corresponding eigenvectors are ordered from low to high space frequencies. This has a rational explanation from the origin of the banded matrix. Note that @ sin(mx) = ;m sin(mx) (10.2) @x and recall that 1 B (1 ;2 1)~ = X 1 D(~ ) X ; ~ (10.3) xx~ = x x 191
2 2 2 2 1
192
1
CHAPTER 10. MULTIGRID
where D(~ ) is a diagonal matrix containing the eigenvalues. We have seen that X ; ~ represents a sine transform, and X ~ , a sine synthesis. Therefore, the operation 1 D(~ ) represents the numerical approximation of the multiplication of the x appropriate sine wave by the negative square of its wavenumber, ;m . One nds that 1 M + 1 ;2 + 2 cos m ;m m << M (10.4) m= M +1 x Hence, the correlation of large magnitudes of m with high spacefrequencies is to be expected for these particular matrix operators. This is consistent with the physics of di usion as well. However, this correlation is not necessary in general. In fact, the complete counterexample of the above association is contained in the eigensystem 1 for B ( 2 1 1 ). For this matrix one nds, from Appendix B, exactly the opposite 2 behavior.
2 2 2 2 2
10.1.2 Properties of the Iterative Method
The second key motivation for multigrid is the following: Many iterative methods reduce error components corresponding to eigenvalues of large amplitude more e ectively than those corresponding to eigenvalues of small amplitude. This is to be expected of an iterative method which is time accurate. It is also true, for example, of the GaussSeidel method and, by design, of the Richardson method described in Section 9.5. The classical pointJacobi method does not share this property. As we saw in Section 9.4.1, this method produces the same value of j j for min and max . However, the property can be restored by using h < 1, as shown in Fig. 9.2. When an iterative method with this property is applied to a matrix with the above correlation between the modulus of the eigenvalues and the space frequency of the eigenvectors, error components corresponding to high space frequencies will be reduced more quickly than those corresponding to low space frequencies. This is the key concept underlying the multigrid process.
10.2 The Basic Process
First of all we assume that the di erence equations representing the basic partial di erential equations are in a form that can be related to a matrix which has certain
10.2. THE BASIC PROCESS
193
basic properties. This form can be arrived at \naturally" by simply replacing the derivatives in the PDE with di erence schemes, as in the example given by Eq. 3.27, or it can be \contrived" by further conditioning, as in the examples given by Eq. 9.11. The basic assumptions required for our description of the multigrid process are: 1. The problem is linear. 2. The eigenvalues, m , of the matrix are all real and negative. 3. The m are fairly evenly distributed between their maximum and minimum values. 4. The eigenvectors associated with the eigenvalues having largest magnitudes can be correlated with high frequencies on the di erencing mesh. 5. The iterative procedure used greatly reduces the amplitudes of the eigenvectors 1 associated with eigenvalues in the range between 2 j jmax and j jmax. These conditions are su cient to ensure the validity of the process described next. Having preconditioned (if necessary) the basic nite di erencing scheme by a procedure equivalent to the multiplication by a matrix C , we are led to the starting formulation C Ab ~ 1 ; ~ b ] = 0 f (10.5) where the matrix formed by the product CAb has the properties given above. In Eq. 10.5, the vector ~ b represents the boundary conditions and the forcing function, if f ~ 1 is a vector representing the desired exact solution. We start with some any, and initial guess for ~ 1 and proceed through n iterations making use of some iterative process that satis es property 5 above. We do not attempt to develop an optimum procedure here, but for clarity we suppose that the threestep Richardson method illustrated in Fig. 9.6 is used. At the end of the three steps we nd ~ , the residual, r where ~ = C Ab~ ; ~ b] r f (10.6) Recall that the ~ used to compute ~ is composed of the exact solution ~ 1 and the r ~ in such a way that error e A~ ; ~ = 0 e r (10.7) where A CAb (10.8)
194 If one could solve Eq. 10.7 for ~ then e ~1 = ~ ;~ e
CHAPTER 10. MULTIGRID
(10.9)
Thus our goal now is to solve for ~ . We can write the exact solution for ~ in terms of e e the eigenvectors of A, and the eigenvalues of the Richardson process in the form:
M= M ~ = X cm~ m Y ( m hn)] + X cm~ m Y ( mhn)] e x x
2 3 3
m=1
n=1
m=M=2+1

very low amplitude
{z
n=1
(10.10)
}
Combining our basic assumptions, we can be sure that the high frequency content of ~ has been greatly reduced (about 1% or less of its original value in the initial guess). e In addition, assumption 4 ensures that the error has been smoothed. Next we construct a permutation matrix which separates a vector into two parts, one containing the odd entries, and the other the even entries of the original vector (or any other appropriate sorting which is consistent with the interpolation approximation to be discussed below). For a 7point example
2 6 6 6 6 6 6 6 6 6 6 6 4
e e e e e e e
2 4 6 1 3 5 7
3 2 7 6 7 6 7 6 7 6 7 6 7=6 7 6 7 6 7 6 7 6 7 6 5 4
0 0 0 1 0 0 0
1 0 0 0 0 0 0
0 0 0 0 1 0 0
0 1 0 0 0 0 0
0 0 0 0 0 1 0
0 0 1 0 0 0 0
0 76 0 76 76 0 76 76 0 76 76 76 0 76 76 0 76 54 1
32
e e e e e e e
1 2 3 4 5 6 7
3 7 7 7 7 7 7 7 7 7 7 7 5
"~ #
ee ~ ~o = P e e
(10.11)
Multiply Eq. 10.7 from the left by P and, since a permutation matrix has an inverse which is its transpose, we can write PA P ; P ]~ = P ~ e r (10.12)
1
The operation PAP ; partitions the A matrix to form
1
2 3 A 1 A2 " ~ # " ~ # 6 7 ee = re 4 5 ~ ~o r A 3 A4 e o
(10.13)
Notice that
e r e A ~e + A ~o = ~e
1 2
(10.14)
10.2. THE BASIC PROCESS
195
It is important to notice that ea and eb represent errors on the boundaries where the error is zero if the boundary conditions are given. It is also important to notice that we are dealing with the relation between ~ and ~ so the original boundary conditions and e r forcing function (which are contained in ~ in the basic formulation) no longer appear f in the problem. Hence, no aliasing of these functions can occur in subsequent steps. Finally, notice that, in this formulation, the averaging of ~ is our only approximation, e no operations on ~ are required or justi ed. r If the boundary conditions are Dirichlet, ea and eb are zero, and one can write for the example case
2 1 0 0 16 1 1 0 A02 = 2 6 0 1 1 6 4 3 7 7 7 5
is an exact expression. At this point we make our one crucial assumption. It is that there is some connection between ~ e and ~ o brought about by the smoothing property of the Richardson e e relaxation procedure. Since the top half of the frequency spectrum has been removed, it is reasonable to suppose that the odd points are the average of the even points. For example 1 (e + e ) e 2 a 1 (e + e ) e 2 1 (e + e ) or ~ o = A0 ~ e e e (10.15) e 2 1 (e + e ) e b 2
1 2 3 2 4 5 4 6 2 7 6
(10.16)
0 0 1
With this approximation Eq. 10.14 reduces to A ~ e + A A0 ~ e = ~ e e e r
1 2 2
(10.17) (10.18) (10.19)
or
Ac~ e ; ~ e = 0 e r
where
Ac = A + A A0 ]
1 2 2
196
CHAPTER 10. MULTIGRID
The form of Ac, the matrix on the coarse mesh, is completely determined by the choice of the permutation matrix and the interpolation approximation. If the original A had been B (7 : 1 ;2 1), our 7point example would produce
2 ;2 6 6 6 6 6 6 ;1 = 6 PAP 6 1 6 6 6 1 6 6 4
;2
1 1
1
;2 ;2
1 1
1 1
1 1
;2
;2
;2
3 7 7 7 17 2A A 3 7 7 6 1 27 7=4 7 5 7 7 A3 A4 7 7 7 5
(10.20)
and Eq. 10.18 gives
z } 2 ;2 6 ;2 4
A1
;2
{ z } 3 2 1 1 7+6 1 1 5 4
A2
1 1
{ 3 7 5
2 z } 2 1 16 1 1 6 4 26 1 1
A
0
1
{ Ac 3 2 } z ;1 1=2 7 6 7 = 4 1=2 ;1 1=2 7 5
1=2 ;1
{ 3 7 (10.21) 5
If the boundary conditions are mixed DirichletNeumann, A in the 1D model equation is B (1 ~ 1) where ~ = ;2 ;2 ::: ;2 ;1]T . The eigensystem is given by b b Eq. B.19. It is easy to show that the high spacefrequencies still correspond to the eigenvalues with high magnitudes, and, in fact, all of the properties given in Section 10.1 are met. However, the eigenvector structure is di erent from that given in Eq. 9.55 for Dirichlet conditions. In the present case they are given by " !# (2m ; 1) xjm = sin j 2M + 1 m=1 2 M (10.22) and are illustrated in Fig. 10.1. All of them go through zero on the left (Dirichlet) side, and all of them re ect on the right (Neumann) side. For Neumann conditions, the interpolation formula in Eq. 10.15 must be changed. In the particular case illustrated in Fig. 10.1, eb is equal to eM . If Neumann conditions are on the left, ea = e . When eb = eM , the example in Eq. 10.16 changes to 2 1 0 03 7 16 A0 = 2 6 1 1 0 7 60 1 17 (10.23) 4 5 0 0 2
1 2
10.2. THE BASIC PROCESS
197
X
Figure 10.1: Eigenvectors for the mixed Dirichlet{Neumann case. The permutation matrix remains the same and both A and A in the partitioned matrix PAP ; are unchanged (only A is modi ed by putting ;1 in the lower right element). Therefore, we can construct the coarse matrix from
1 1 2 4
} z 2 ;2 6 ;2 4
A1
;2
{ 2 } 3 z 1 1 7+6 1 1 5 4
A2
1 1
{ 3 7 5
2 z } 2 1 61 1 16 4 26 1 1
A
0
2
{ Ac 3 2 } z ;1 1=2 7 6 7 = 4 1=2 ;1 7 5
1=2 ;1=2
{ 3 1=2 7 (10.24) 5
which gives us what we might have \expected." We will continue with Dirichlet boundary conditions for the remainder of this Section. At this stage, we have reduced the problem from B (1 ;2 1)~ = ~ on the e r 1 B (1 ;2 1)~e = ~ e on the next coarser mesh. Recall that our goal is ne mesh to 2 e r to solve for ~ , which will provide us with the solution ~ 1 using Eq. 10.9. Given ~ e e e computed on the coarse grid (possibly using even coarser grids), we can compute ~ o e using Eq. 10.15, and thus ~ . In order to complete the process, we must now determine e ~ e and ~ . the relationship between e e In order to examine this relationship, we need to consider the eigensystems of A and Ac: A = X X; Ac = Xc cXc; (10.25) For A = B (M : 1 ;2 1) the eigenvalues and eigenvectors are m j=1 2 M (10.26) ~ m = sin j m x m = ;2 1 ; cos m=1 2 M M +1 M +1
1 1
198
CHAPTER 10. MULTIGRID
Based on our assumptions, the most di cult error mode to eliminate is that with m = 1, corresponding to
~ = sin j = ;2 1 ; cos M + 1 x j=1 2 M (10.27) M +1 For example, with M = 51, = ;0:003649. If we restrict our attention to odd M , then Mc = (M ; 1)=2 is the size of Ac. The eigenvalue and eigenvector corresponding to m = 1 for the matrix Ac = 1 B (Mc 1 ;2 1) are 2 ( c) = ; 1 ; cos M2+ 1 (~ c) = sin j M2+ 1 j = 1 2 x Mc (10.28) For M = 51 (Mc = 25), we obtain ( c) = ;0:007291 = 1:998 . As M increases, ( c) approaches 2 . In addition, one can easily see that (~ c) coincides with ~ at x x every second point of the latter vector, that is, it contains the even elements of ~ . x Now let us consider the case in which all of the error consists of the eigenvector component ~ , i.e., ~ = ~ . Then the residual is x e x ~ = A~ = ~ r x x (10.29)
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
and the residual on the coarse grid is ~ e = (~ c) r x
1 1 1
1
(10.30)
since (~ c) contains the even elements of ~ . The exact solution on the coarse grid x x satis es ~ e = A; ~ e = Xc ; Xc; (~ c) e x c r c (10.31) 2 3 1 607 6 7 = Xc ; 6 .. 7 c 6 7 4 . 5 0 (10.32) 2 3 1=( c) 6 0 7 6 7 = Xc 6 .. 7 6 . 7 4 5 0 (10.33)
1 1 1 1 1 1 1 1 1
10.2. THE BASIC PROCESS
= ( c)
1 1
199 (~ c) x
1 1
1 (~ ) 2 xc
1
(10.34) (10.35)
Since our goal is to compute ~ = ~ , in addition to interpolating ~ e to the ne grid e x e (using Eq. 10.15), we must multiply the result by 2. This is equivalent to solving or 1A ~ = ~ e r 2 ce e (10.36)
1 B (M : 1 ;2 1)~ = ~ ee re (10.37) c 4 In our case, the matrix A = B (M : 1 ;2 1) comes from a discretization of the di usion equation, which gives
Ab = x B (M : 1 ;2 1) and the preconditioning matrix C is simply
2
(10.38)
C= x I
2
(10.39)
Applying the discretization on the coarse grid with the same preconditioning matrix as used on the ne grid gives, since xc = 2 x,
x C x B (Mc : 1 ;2 1) = x B (Mc : 1 ;2 1) = 1 B (Mc : 1 ;2 1) 4 c c
2 2 2
(10.40)
which is precisely the matrix appearing in Eq. 10.37. Thus we see that the process is recursive. The problem to be solved on the coarse grid is the same as that solved on the ne grid. The remaining steps required to complete an entire multigrid process are relatively straightforward, but they vary depending on the problem and the user. The reduction can be, and usually is, carried to even coarser grids before returning to the nest level. However, in each case the appropriate permutation matrix and the interpolation approximation de ne both the down and upgoing paths. The details of nding optimum techniques are, obviously, quite important but they are not discussed here.
200
10.3 A TwoGrid Process
CHAPTER 10. MULTIGRID
~ We now describe a twogrid process for the linear problem A ~ = f , which can be easily generalized to a process with an arbitrary number of grids due to the recursive nature of multigrid. Extension to nonlinear problems requires that both the solution and the residual be transferred to the coarse grid in a process known as full approximation storage multigrid.
1. Perform n iterations of the selected relaxation method on the ne grid, starting with ~ = ~n. Call the result ~ . This gives ~ = Gn1 ~n + (I ; Gn1 ) A; f ~ (10.41)
1 (1) 1 (1) 1 1 1
where
G = I + H; A (10.42) and H is de ned as in Chapter 9 (e.g., Eq. 9.21). Next compute the residual based on ~ : ~ ~ ~ ~ = A ~ ; f = AGn1 ~n + A (I ; Gn1 ) A; f ; f r ~ = AGn1 ~n ; AGn1 A; f (10.43)
1 1 1 1 1 (1) (1) (1) 1 1 1 1 1 1
2. Transfer (or restrict) ~ to the coarse grid: r ~ =R ~ r r (10.44) In our example in the preceding section, the restriction matrix is 2 3 0 1 0 0 0 0 0 R =6 0 0 0 1 0 0 0 7 (10.45) 4 5 0 0 0 0 0 1 0 that is, the rst three rows of the permutation matrix P in Eq. 10.11. This type of restriction is known as \simple injection." Some form of weighted restriction can also be used.
(1) (2) 2 (1) 1 2 1
3. Solve the problem A ~ = ~ on the coarse grid exactly: e r ~ = A; ~ e r
2 (2) (2) (2) 2 1 (2)
2
(10.46)
1 2
See problem 1 of Chapter 9. Note that the coarse grid matrix denoted A2 here was denoted Ac in the preceding section.
10.3. A TWOGRID PROCESS
2 2
201
Here A can be formed by applying the discretization on the coarse grid. In the 1 preceding example (eq. 10.40), A = 4 B (Mc : 1 ;2 1). It is at this stage that the generalization to a multigrid procedure with more than two grids occurs. If this is the coarsest grid in the sequence, solve exactly. Otherwise, apply the twogrid process recursively. 4. Transfer (or prolong) the error back to the ne grid and update the solution:
~n = ~ ; I ~ e
+1 (1)
1 (2) 2
(10.47)
3 7 7 7 7 7 7 7 7 7 7 7 5
In our example, the prolongation matrix is
2 6 6 6 6 6 1 I2 = 6 6 6 6 6 6 4
1=2 1 1=2 0 0 0 0
0 0 1=2 1 1=2 0 0
0 0 0 0 1=2 1 1=2
(10.48)
which follows from Eq. 10.15. Combining these steps, one obtains
~n = I ; I A; R A]Gn1 ~ n ; I ; I A; R A]Gn1 A; f + A; f ~ ~
+1 1 2 2 1 2 1 1 1 2 2 1 2 1 1 1 1
(10.49) (10.50)
Thus the basic iteration matrix is
I ; I A; R A]Gn1
1 2 2 1 2 1 1 1 1
The eigenvalues of this matrix determine the convergence rate of the twogrid process. The basic iteration matrix for a threegrid process is found from Eq. 10.50 by replacing A; with (I ; G )A; , where
2 2 3 2
G = I ; I A; R A ]Gn2
2 3 2 3 3 1 3 2 2 2 2 2 3 3 2 3
(10.51)
In this expression n is the number of relaxation steps on grid 2, I and R are the transfer operators between grids 2 and 3, and A is obtained by discretizing on grid 3. Extension to four or more grids proceeds in similar fashion.
202
10.4 Problems
CHAPTER 10. MULTIGRID
1. Derive Eq. 10.51. 2. Repeat problem 4 of Chapter 9 using a fourgrid multigrid method together with (a) the GaussSeidel method, (b) the 3step Richardson method derived in Section 9.5. Solve exactly on the coarsest grid. Plot the solution after the residual is reduced by 2, 3, and 4 orders of magnitude. Plot the logarithm of the L norm of the residual vs. the number of iterations. Determine the asymptotic convergence rate. Calculate the theoretical asymptotic convergence rate and compare.
2
Chapter 11 NUMERICAL DISSIPATION
Up to this point, we have emphasized the secondorder centereddi erence approximations to the spatial derivatives in our model equations. We have seen that a centered approximation to a rst derivative is nondissipative, i.e., the eigenvalues of the associated circulant matrix (with periodic boundary conditions) are pure imaginary. In processes governed by nonlinear equations, such as the Euler and NavierStokes equations, there can be a continual production of highfrequency components of the solution, leading, for example, to the production of shock waves. In a real physical problem, the production of high frequencies is eventually limited by viscosity. However, when we solve the Euler equations numerically, we have neglected viscous e ects. Thus the numerical approximation must contain some inherent dissipation to limit the production of highfrequency modes. Although numerical approximations to the NavierStokes equations contain dissipation through the viscous terms, this can be insu cient, especially at high Reynolds numbers, due to the limited grid resolution which is practical. Therefore, unless the relevant length scales are resolved, some form of added numerical dissipation is required in the numerical solution of the NavierStokes equations as well. Since the addition of numerical dissipation is tantamount to intentionally introducing nonphysical behavior, it must be carefully controlled such that the error introduced is not excessive. In this Chapter, we discuss some di erent ways of adding numerical dissipation to the spatial derivatives in the linear convection equation and hyperbolic systems of PDE's. 203
204
11.1 OneSided FirstDerivative Space Di erencing
We investigate the properties of onesided spatial di erence operators in the context of the biconvection model equation given by
CHAPTER 11. NUMERICAL DISSIPATION
@u = ;a @u @t @x
(11.1)
with periodic boundary conditions. Consider the following point operator for the spatial derivative term
;a( x u)j = 2;ax ;(1 + )uj;1 + 2 uj + (1 ; )uj+1] = ;a (;uj;1 + uj+1) + (;uj;1 + 2uj ; uj+1)] 2 x
(11.2)
The second form shown divides the operator into an antisymmetric component (;uj;1+ uj+1)=2 x and a symmetric component (;uj;1 + 2uj ; uj+1)=2 x. The antisymmetric component is the secondorder centered di erence operator. With 6= 0, the operator is only rstorder accurate. A backward di erence operator is given by = 1 and a forward di erence operator is given by = ;1. For periodic boundary conditions the corresponding matrix operator is
;a x = 2;ax Bp(;1 ; 2 1 ; )
The eigenvalues of this matrix are
m
= ;a x
m 1 ; cos 2M
m + i sin 2M
for m = 0 1 : : : M ; 1
If a is positive, the forward di erence operator ( = ;1) produces Re( m ) > 0, the centered di erence operator ( = 0) produces Re( m ) = 0, and the backward di erence operator produces Re( m) < 0. Hence the forward di erence operator is inherently unstable while the centered and backward operators are inherently stable. If a is negative, the roles are reversed. When Re( m) = 0, the solution will either grow or decay with time. In either case, our choice of di erencing scheme produces nonphysical behavior. We proceed next to show why this occurs.
11.2. THE MODIFIED PARTIAL DIFFERENTIAL EQUATION
11.2 The Modi ed Partial Di erential Equation
( xu)j = 2 1 x 42 x @u ; @x j
205
First carry out a Taylor series expansion of the terms in Eq. 11.2. We are lead to the expression
2
!
3 ! ! ! @ 2 u + x3 @ 3 u ; x4 @ 4 u + : : :5 x2 @x2 3 @x3 j 12 @x4 j j
We see that the antisymmetric portion of the operator introduces odd derivative terms in the truncation error while the symmetric portion introduces even derivatives. Substituting this into Eq. 11.1 gives
@u = ;a @u + a x @ 2 u ; a x2 @ 3 u + a x3 @ 4 u + : : : (11.3) @t @x 2 @x2 6 @x3 24 @x4 This is the partial di erential equation we are really solving when we apply the approximation given by Eq. 11.2 to Eq. 11.1. Notice that Eq. 11.3 is consistent with Eq. 11.1, since the two equations are identical when x ! 0. However, when we use a computer to nd a numerical solution of the problem, x can be small but it is not zero. This means that each term in the expansion given by Eq. 11.3 is excited to some degree. We refer to Eq. 11.3 as the modi ed partial di erential equation. We proceed next to investigate the implications of this concept. Consider the simple linear partial di erential equation @u = ;a @u + @ 2 u + @ 3 u + @ 4 u (11.4) @t @x @x2 @x3 @x4 Choose periodic boundary conditions and impose an initial condition u = ei x. Under these conditions there is a wavelike solution to Eq. 11.4 of the form u(x t) = ei xe(r+is)t provided r and s satisfy the condition r + is = ;ia ; 2 ; i 3 + 4 or r = ; 2 ( ; 2 ) s = ; (a + 2 ) The solution is composed of both amplitude and phase terms. Thus a ) u = e; 2 ({z; 2} ei x;({z+ 2 )t}] (11.5)   amplitude phase
206
CHAPTER 11. NUMERICAL DISSIPATION
It is important to notice that the amplitude of the solution depends only upon and , the coe cients of the even derivatives in Eq. 11.4, and the phase depends only on a and , the coe cients of the odd derivatives. If the wave speed a is positive, the choice of a backward di erence scheme ( = 1) produces a modi ed PDE with ; 2 > 0 and hence the amplitude of the solution decays. This is tantamount to deliberately adding dissipation to the PDE. Under the same condition, the choice of a forward di erence scheme ( = ;1) is equivalent to deliberately adding a destabilizing term to the PDE. By examining the term governing the phase of the solution in Eq. 11.5, we see that the speed of propagation is a + 2 . Referring to the modi ed PDE, Eq. 11.3 we have = ;a x2 =6. Therefore, the phase speed of the numerical solution is less than the actual phase speed. Furthermore, the numerical phase speed is dependent upon the wavenumber . This we refer to as dispersion. Our purpose here is to investigate the properties of onesided spatial di erencing operators relative to centered di erence operators. We have seen that the threepoint centered di erence approximation of the spatial derivative produces a modi ed PDE that has no dissipation (or ampli cation). One can easily show, by using the antisymmetry of the matrix di erence operators, that the same is true for any centered di erence approximation of a rst derivative. As a corollary, any departure from antisymmetry in the matrix di erence operator must introduce dissipation (or ampli cation) into the modi ed PDE. Note that the use of onesided di erencing schemes is not the only way to introduce dissipation. Any symmetric component in the spatial operator introduces dissipation (or ampli cation). Therefore, one could choose = 1=2 in Eq. 11.2. The resulting spatial operator is not onesided but it is dissipative. Biased schemes use more information on one side of the node than the other. For example, a thirdorder backwardbiased scheme is given by ( xu)j = 6 1 x (uj;2 ; 6uj;1 + 3uj + 2uj+1) = 1 (uj;2 ; 8uj;1 + 8uj+1 ; uj+2) 12 x + (uj;2 ; 4uj;1 + 6uj ; 4uj+1 + uj+2)]
(11.6)
The antisymmetric component of this operator is the fourthorder centered di erence operator. The symmetric component approximates x3 uxxxx=12. Therefore, this operator produces fourthorder accuracy in phase with a thirdorder dissipative term.
11.3. THE LAXWENDROFF METHOD
11.3 The LaxWendro Method
207
In order to introduce numerical dissipation using onesided di erencing, backward di erencing must be used if the wave speed is positive, and forward di erencing must be used if the wave speed is negative. Next we consider a method which introduces dissipation independent of the sign of the wave speed, known as the LaxWendro method. This explicit method di ers conceptually from the methods considered previously in which spatial di erencing and timemarching are treated separately. Consider the following Taylorseries expansion in time:
2 u(x t + h) = u + h @u + 1 h2 @ u + O(h3) @t 2 @t2
(11.7)
First replace the time derivatives with space derivatives according to the PDE (in this case, the linear convection equation @u + a @u = 0). Thus @t @x
@ 2 u = a2 @ 2 u @u = ;a @u (11.8) @t @x @t2 @x2 Now replace the space derivatives with threepoint centered di erence operators, giving !2 (n+1) = u(n) ; 1 ah (u(n) ; u(n) ) + 1 ah (u(n) ; 2u(n) + u(n) ) (11.9) uj j j +1 j j ;1 2 x j+1 j;1 2 x
This is the LaxWendro method applied to the linear convection equation. It is a fullydiscrete nitedi erence scheme. There is no intermediate semidiscrete stage. For periodic boundary conditions, the corresponding fullydiscrete matrix operator is 0 2 !3 ! 2 ! 31 1 4 ah + ah 25 1 ; ah 2 1 4; ah + ah 2 5A ~ ~ n+1 = Bp @ 2 x u un x x 2 x x The eigenvalues of this matrix are
! ah 2 1 ; cos 2 m ; i ah sin 2 m m =1; x M x M
for m = 0 1 : : : M ; 1
For j ah j 1 all of the eigenvalues have modulus less than or equal to unity and hence x the method is stable independent of the sign of a. The quantity j ah j is known as the x Courant (or CFL) number. It is equal to the ratio of the distance travelled by a wave in one time step to the mesh spacing.
208
CHAPTER 11. NUMERICAL DISSIPATION
The nature of the dissipative properties of the LaxWendro scheme can be seen by examining the modi ed partial di erential equation, which is given by @u + a @u = ; a ( x2 ; a2h2 ) @ 3 u ; a2 h ( x2 ; a2 h2) @ 4 u + : : : @t @x 6 @x3 8 @x4 This is derived by substituting Taylor series expansions for all terms in Eq. 11.9 and converting the time derivatives to space derivatives using Eq. 11.8. The two leading error terms appear on the right side of the equation. Recall that the odd derivatives on the right side lead to unwanted dispersion and the even derivatives lead to dissipation (or ampli cation, depending on the sign). Therefore, the leading error term in the LaxWendro method is dispersive and proportional to a ( x2 ; a2h2 ) @ 3 u = ; a x2 (1 ; C 2 ) @ 3 u ;6 n @x3 @x3 6 The dissipative term is proportional to 2 4 2 2 4 2 @ ; a8h ( x2 ; a2h2 ) @ u = ; a h8 x (1 ; Cn) @xu 4 @x4 This term has the appropriate sign and hence the scheme is truly dissipative as long as Cn 1. A closely related method is that of MacCormack. Recall MacCormack's timemarching method, presented in Chapter 6:
un+1 = un + hu0n ~ 1 ~ ~ (11.10) un+1 = 2 un + un+1 + hu0n+1] If we use rstorder backward di erencing in the rst stage and rstorder forward di erencing in the second stage,1 a dissipative secondorder method is obtained. For the linear convection equation, this approach leads to u(jn+1) = u(jn) ; ah (u(jn) ; u(jn)1) ~ ; x u(jn+1) = 1 u(jn) + u(jn+1) ; ah (~(jn+1) ; u(jn+1))] ~ (11.11) 2 x u +1 ~
which can be shown to be identical to the LaxWendro method. Hence MacCormack's method has the same dissipative and dispersive properties as the LaxWendro method. The two methods di er when applied to nonlinear hyperbolic systems, however.
1 Or viceversa for nonlinear problems, these should be applied alternately.
11.4. UPWIND SCHEMES
11.4 Upwind Schemes
209
In Section 11.1, we saw that numerical dissipation can be introduced in the spatial di erence operator using onesided di erence schemes or, more generally, by adding a symmetric component to the spatial operator. With this approach, the direction of the onesided operator (i.e., whether it is a forward or a backward di erence) or the sign of the symmetric component depends on the sign of the wave speed. When a hyperbolic system of equations is being solved, the wave speeds can be both positive and negative. For example, the eigenvalues of the ux Jacobian for the onedimensional Euler equations are u u + a u ; a. When the ow is subsonic, these are of mixed sign. In order to apply onesided di erencing schemes to such systems, some form of splitting is required. This is avoided in the LaxWendro scheme. However, as a result of their superior exibility, schemes in which the numerical dissipation is introduced in the spatial operator are generally preferred over the LaxWendro approach. Consider again the linear convection equation: @u + a @u = 0 (11.12) @t @x where we do not make any assumptions as to the sign of a. We can rewrite Eq. 11.12 as @u + (a+ + a; ) @u = 0 a = a 2 jaj @t @x + = a 0 and a; = 0. Alternatively, if a 0, then a+ = 0 and If a 0, then a a; = a 0. Now for the a+ ( 0) term we can safely backward di erence and for the a; ( 0) term forward di erence. This is the basic concept behind upwind methods, that is, some decomposition or splitting of the uxes into terms which have positive and negative characteristic speeds so that appropriate di erencing schemes can be chosen. In the next two sections, we present two splitting techniques commonly used with upwind methods. These are by no means unique. The above approach to obtaining a stable discretization independent of the sign of a can be written in a di erent, but entirely equivalent, manner. From Eq. 11.2, we see that a stable discretization is obtained with = 1 if a 0 and with = ;1 if a 0. This is achieved by the following point operator: ;a( xu)j = 2;1x a(;uj;1 + uj+1) + jaj(;uj;1 + 2uj ; uj+1)] (11.13) This approach is extended to systems of equations in Section 11.5. In this section, we present the basic ideas of uxvector and uxdi erence splitting. For more subtle aspects of implementation and application of such techniques to
210
CHAPTER 11. NUMERICAL DISSIPATION
nonlinear hyperbolic systems such as the Euler equations, the reader is referred to the literature on this subject.
11.4.1 FluxVector Splitting
Recall from Section 2.5 that a linear, constantcoe cient, hyperbolic system of partial di erential equations given by
@u + @f = @u + A @u = 0 (11.14) @t @x @t @x can be decoupled into characteristic equations of the form @wi + @wi = 0 (11.15) i @t @x where the wave speeds, i, are the eigenvalues of the Jacobian matrix, A, and the wi's are the characteristic variables. In order to apply a onesided (or biased) spatial di erencing scheme, we need to apply a backward di erence if the wave speed, i, is positive, and a forward di erence if the wave speed is negative. To accomplish this, let us split the matrix of eigenvalues, , into two components such that
= ++ ; where
+=
(11.16)
+j j ; = ;j j (11.17) 2 2 With these de nitions, + contains the positive eigenvalues and ; contains the negative eigenvalues. We can now rewrite the system in terms of characteristic variables as @w + @w = @w + + @w + ; @w = 0 (11.18) @t @x @t @x @x The spatial terms have been split into two components according to the sign of the wave speeds. We can use backward di erencing for the + @w term and forward @x di erencing for the ; @w term. Premultiplying by X and inserting the product @x X ;1X in the spatial terms gives
@Xw + @X +X ;1Xw + @X ;X ;1Xw = 0 @t @x @x
(11.19)
11.4. UPWIND SCHEMES
With the de nitions2
211
A+ = X +X ;1
and recalling that u = Xw, we obtain
A; = X ;X ;1
(11.20)
@u + @A+u + @A; u = 0 @t @x @x Finally the split ux vectors are de ned as f + = A+u
and we can write
(11.21) (11.22)
f ; = A; u
@u + @f + + @f ; = 0 (11.23) @t @x @x In the linear case, the de nition of the split uxes follows directly from the de nition of the ux, f = Au. For the Euler equations, f is also equal to Au as a result of their homogeneous property, as discussed in Appendix C. Note that f = f+ + f;
(11.24) Thus by applying backward di erences to the f + term and forward di erences to the f ; term, we are in e ect solving the characteristic equations in the desired manner. This approach is known as uxvector splitting. When an implicit timemarching method is used, the Jacobians of the split ux vectors are required. In the nonlinear case,
@f + 6= A+ @f ; 6= A; @u @u Therefore, one must nd and use the new Jacobians given by A++ = @f @u
+
(11.25)
A;; = @f @u
;
(11.26)
For the Euler equations, A++ has eigenvalues which are all positive, and A;; has all negative eigenvalues.
2 With these de nitions A+ has all positive eigenvalues, and A; has all negative eigenvalues.
212
CHAPTER 11. NUMERICAL DISSIPATION
11.4.2 FluxDi erence Splitting
Another approach, more suited to nitevolume methods, is known as uxdi erence splitting. In a nitevolume method, the uxes must be evaluated at cell boundaries. We again begin with the diagonalized form of the linear, constantcoe cient, hyperbolic system of equations @w + @w = 0 (11.27) @t @x The ux vector associated with this form is g = w. Now, as in Chapter 5, we consider the numerical ux at the interface between nodes j and j + 1, gj+1=2, as a ^ function of the states to the left and right of the interface, wL and wR, respectively. The centered approximation to gj+1=2, which is nondissipative, is given by 1 gj+1=2 = 2 (g(wL) + g(wR)) ^ (11.28) In order to obtain a onesided upwind approximation, we require ( (w ) (^i)j+1=2 = i(wi)L if i > 0 g (11.29) i i R if i < 0 where the subscript i indicates individual components of w and g. This is achieved with 1 1 (^i)j+1=2 = 2 i (wi)L + (wi)R ] + 2 j ij (wi)L ; (wi)R] g (11.30) or 1 gj+1=2 = 1 (wL + wR) + 2 j j (wL ; wR ) ^ (11.31) 2 Now, as in Eq. 11.19, we premultiply by X to return to the original variables and insert the product X ;1X after and j j to obtain 1 1 X gj+1=2 = 2 X X ;1X (wL + wR) + 2 X j jX ;1X (wL ; wR ) ^ (11.32) and thus 1 f^j+1=2 = 1 (fL + fR ) + 2 jAj (uL ; uR) (11.33) 2 where
jAj = X j jX ;1
(11.34)
11.5. ARTIFICIAL DISSIPATION
213
and we have also used the relations f = Xg, u = Xw, and A = X X ;1. In the linear, constantcoe cient case, this leads to an upwind operator which is identical to that obtained using uxvector splitting. However, in the nonlinear case, there is some ambiguity regarding the de nition of jAj at the cell interface j + 1=2. In order to resolve this, consider a situation in which the eigenvalues of A are all of the same sign. In this case, we would like our de nition of f^j+1=2 to satisfy
0 f f^j+1=2 = fL if all i 0s > 0 (11.35) R if all i s < 0 giving pure upwinding. If the eigenvalues of A are all positive, jAj = A if they are all negative, jAj = ;A. Hence satisfaction of Eq. 11.35 is obtained by the de nition 1 f^j+1=2 = 2 (fL + fR ) + 1 jAj+1=2j (uL ; uR) (11.36) 2 if Aj+1=2 satis es fL ; fR = Aj+1=2 (uL ; uR) (11.37) For the Euler equations for a perfect gas, Eq. 11.37 is satis ed by the ux Jacobian evaluated at the Roeaverage state given by p LuL + p R uR (11.38) uj+1=2 = p + p L R p H +p H L L Hj+1=2 = (11.39) p L+p R R R where u and H = (e + p)= are the velocity and the total enthalpy per unit mass, respectively.3
(
11.5 Arti cial Dissipation
We have seen that numerical dissipation can be introduced by using onesided differencing schemes together with some form of ux splitting. We have also seen that such dissipation can be introduced by adding a symmetric component to an antisymmetric (dissipationfree) operator. Thus we can generalize the concept of upwinding to include any scheme in which the symmetric portion of the operator is treated in such a manner as to be truly dissipative.
of this chapter.
3 Note that the ux Jacobian can be written in terms of u and H only see problem 6 at the end
214 For example, let
CHAPTER 11. NUMERICAL DISSIPATION
s a ( xu)j = ;uj+1 + 2uj ; uj;1 (11.40) ( x u)j = uj+1 ; uj;1 2 x 2 x a s Applying x = x + x to the spatial derivative in Eq. 11.15 is stable if i 0 and a s unstable if i < 0. Similarly, applying x = x ; x is stable if i 0 and unstable if i > 0. The appropriate implementation is thus a s (11.41) i x = i x + j ij x Extension to a hyperbolic system by applying the above approach to the characteristic variables, as in the previous two sections, gives a s (11.42) x (Au) = x (Au) + x (jAju) or a s (11.43) x f = x f + x (jAju) where jAj is de ned in Eq. 11.34. The second spatial term is known as arti cial dissipation. It is also sometimes referred to as arti cial di usion or arti cial viscosity. a s With appropriate choices of x and x, this approach can be related to the upwind approach. This is particularly evident from a comparison of Eqs. 11.36 and 11.43. s It is common to use the following operator for x
(11.44) x (uj;2 ; 4uj;1 + 6uj ; 4uj+1 + uj+2) where is a problemdependent coe cient. This symmetric operator approximates x3 uxxxx and thus introduces a thirdorder dissipative term. With an appropriate value of , this often provides su cient damping of high frequency modes without greatly a ecting the low frequency modes. For details of how this can be implemented for nonlinear hyperbolic systems, the reader should consult the literature. A more complicated treatment of the numerical dissipation is also required near shock waves and other discontinuities, but is beyond the scope of this book.
s ( xu)j =
11.6 Problems
1. A secondorder backward di erence approximation to a 1st derivative is given as a point operator by ( xu)j = 1 (uj;2 ; 4uj;1 + 3uj ) 2 x
11.6. PROBLEMS
215
(a) Express this operator in banded matrix form (for periodic boundary conditions), then derive the symmetric and skewsymmetric matrices that have the matrix operator as their sum. (See Appendix A.3 to see how to construct the symmetric and skewsymmetric components of a matrix.) (b) Using a Taylor table, nd the derivative which is approximated by the corresponding symmetric and skewsymmetric operators and the leading error term for each. 2. Find the modi ed wavenumber for the rstorder backward di erence operator. Plot the real and imaginary parts of x vs. x for 0 x . Using Fourier analysis as in Section 6.6.2, nd j j for the combination of this spatial operator with 4thorder RungeKutta time marching at a Courant number of unity and plot vs. x for 0 x . 3. Find the modi ed wavenumber for the operator given in Eq. 11.6. Plot the real and imaginary parts of x vs. x for 0 x . Using Fourier analysis as in Section 6.6.2, nd j j for the combination of this spatial operator with 4thorder RungeKutta time marching at a Courant number of unity and plot vs. x for 0 x . 4. Consider the spatial operator obtained by combining secondorder centered differences with the symmetric operator given in Eq. 11.44. Find the modi ed wavenumber for this operator with = 0 1=12 1=24, and 1=48. Plot the real and imaginary parts of x vs. x for 0 x . Using Fourier analysis as in Section 6.6.2, nd j j for the combination of this spatial operator with 4thorder RungeKutta time marching at a Courant number of unity and plot vs. x for 0 x . 5. Consider the hyperbolic system derived in problem 8 of Chapter 2. Find the matrix jAj. Form the plusminus split ux vectors as in Section 11.4.1. 6. Show that the ux Jacobian for the 1D Euler equations can be written in terms of u and H . Show that the use of the Roe average state given in Eqs. 11.38 and 11.39 leads to satisfaction of Eq. 11.37.
216
CHAPTER 11. NUMERICAL DISSIPATION
Chapter 12 SPLIT AND FACTORED FORMS
In the next two chapters, we present and analyze split and factored algorithms. This gives the reader a feel for some of the modi cations which can be made to the basic algorithms in order to obtain e cient solvers for practical multidimensional applications, and a means for analyzing such modi ed forms.
12.1 The Concept
Factored forms of numerical operators are used extensively in constructing and applying numerical methods to problems in uid mechanics. They are the basis for a wide variety of methods variously known by the labels \hybrid", \time split", and \fractional step". Factored forms are especially useful for the derivation of practical algorithms that use implicit methods. When we approach numerical analysis in the light of matrix derivative operators, the concept of factoring is quite simple to present and grasp. Let us start with the following observations: 1. Matrices can be split in quite arbitrary ways. 2. Advancing to the next time level always requires some reference to a previous one. 3. Time marching methods are valid only to some order of accuracy in the step size, h. Now recall the generic ODE's produced by the semidiscrete approach
d~ = A~ ; ~ u u f dt
217
(12.1)
218
CHAPTER 12. SPLIT AND FACTORED FORMS d~ = A + A ]~ ; ~ u u f dt
1 2 2
and consider the above observations. From observation 1 (arbitrary splitting of A) : (12.2) where A = A + A ] but A and A are not unique. For the time march let us choose the simple, rstorder, explicit Euler method. Then, from observation 2 (new data ~ n in terms of old ~ n): u u
1 2 1 1 +1
~ n = I + hA + hA ]~ n ; h~ + O(h ) u u f
+1 1 2 2 1 2 2 1 2
(12.3)
2
or its equivalent
~ n = h I + hA ] I + hA ] ; h A A i~ n ; h~ + O(h ) u u f
+1 2 1 2
Finally, from observation 3 (allowing us to drop higher order terms ;h A A ~ n): u (12.4) Notice that Eqs. 12.3 and 12.4 have the same formal order of accuracy and, in this sense, neither one is to be preferred over the other. However, their numerical stability can be quite di erent, and techniques to carry out their numerical evaluation can have arithmetic operation counts that vary by orders of magnitude. Both of these considerations are investigated later. Here we seek only to apply to some simple cases the concept of factoring.
+1 1 2 2
~ n = I + hA ] I + hA ]~ n ; h~ + O(h ) u u f
12.2 Factoring Physical Representations  Time Splitting
Suppose we have a PDE that represents both the processes of convection and dissipation. The semidiscrete approach to its solution might be put in the form where Ac and Ad are matrices representing the convection and dissipation terms, respectively and their sum forms the A matrix we have considered in the previous sections. Choose again the explicit Euler time march so that
d~ = A ~ + A ~ + (bc) u ~ cu du dt
(12.5)
~ ~ n = I + hAd + hAc]~ n + h(bc) + O(h ) u u
+1 2
(12.6)
1 Secondorder timemarching methods are considered later.
12.2. FACTORING PHYSICAL REPRESENTATIONS  TIME SPLITTING 219
Now consider the factored form
~n u
+1
Original Unfactored Terms Higher Order Terms and we see that Eq. 12.7 and the original unfactored form Eq. 12.6 have identical orders of accuracy in the time approximation. Therefore, on this basis, their selection is arbitrary. In practical applications equations such as 12.7 are often applied in a predictorcorrector sequence. In this case one could write
2
= =
~ I + hAd ] I + hAc]~ n + h(bc) u ~ u ~ I + hAd + hAc]~ n + h(bc) + h Ad Ac~ n + (bc) +O(h ) (12.7) u {z } 
2 2

{z
}
un ~ ~n u
+1 +1
= =
~ I + hAc]~ n + h(bc) u I + hAd]~n u
+1
(12.8)
Factoring can also be useful to form split combinations of implicit and explicit techniques. For example, another way to approximate Eq. 12.6 with the same order of accuracy is given by the expression
~n u
+1
Original Unfactored Terms where in this approximation we have used the fact that
1 2 2
= =
~ I ; hAd]; I + hAc]~ n + h(bc) u ~ I + hAd + hAc]~ n + h(bc) +O(h ) u  {z }
1 2
(12.9)
I ; hAd ]; = I + hAd + h Ad +
if h jjAdjj < 1, where jjAd jj is some norm of Ad ]. This time a predictorcorrector interpretation leads to the sequence
un ~ I ; hAd]~ n u
+1 +1
~ = I + hAc]~ n + h(bc) u = un ~
+1
(12.10)
The convection operator is applied explicitly, as before, but the di usion operator is now implicit, requiring a tridiagonal solver if the di usion term is central di erenced. Since numerical sti ness is generally much more severe for the di usion process, this factored form would appear to be superior to that provided by Eq. 12.8. However, the important aspect of stability has yet to be discussed.
stability, and a rstorder timemarch method is usually unsatisfactory.
2 We do not suggest that this particular method is suitable for use. We have yet to determine its
220
CHAPTER 12. SPLIT AND FACTORED FORMS un ; un = A u + A u + (bc) + O(h ) ~ c n d n h
+1 +1 2
We should mention here that Eq. 12.9 can be derived for a di erent point of view by writing Eq. 12.6 in the form Then
which is identical to Eq. 12.10.
~ I ; hAd ]un = I + hAc]un + h(bc)
+1
12.3.1 Mesh Indexing Convention
2 2
12.3 Factoring Space Matrix Operators in 2{D
Factoring is widely used in codes designed for the numerical solution of equations governing unsteady two and threedimensional ows. Let us study the basic concept of factoring by inspecting its use on the linear 2D scalar PDE that models di usion:
We begin by reducing this PDE to a coupled set of ODE's by di erencing the space derivatives and inspecting the resulting matrix operator. A clear description of a matrix nitedi erence operator in 2 and 3D requires some reference to a mesh. We choose the 3 4 point mesh shown in the Sketch 12.12. In this example Mx , the number of (interior) x points, is 4 and My , the number of (interior) y points is 3. The numbers 11 12 43 represent the location in the mesh of the dependent variable bearing that index. Thus u represents the value of u at j = 3 and k = 2.
3 32
@u = @ u + @ u @t @x @y
2 2
(12.11)
My k
1
13 23 33 43 12 22 32 42 11 21 31 41 1
(12.12)
j
Mx
Mesh indexing in 2D.
6 point mesh if the boundary points (labeled in the sketch) were included, but in these notes we describe the size of a mesh by the number of interior points.
3 This could also be called a 5
12.3. FACTORING SPACE MATRIX OPERATORS IN 2{D
221
12.3.2 Data Bases and Space Vectors
The dimensioned array in a computer code that allots the storage locations of the dependent variable(s) is referred to as a database. There are many ways to lay out a database. Of these, we consider only two: (1), consecutively along rows that are themselves consecutive from k = 1 to My , and (2), consecutively along columns that are consecutive from j = 1 to Mx. We refer to each row or column group as a space vector (they represent data along lines that are continuous in space) and label their sum with the symbol U . In particular, (1) and (2) above are referred to as xvectors and yvectors, respectively. The symbol U by itself is not enough to identify the structure of the database and is used only when the structure is immaterial or understood. To be speci c about the structure, we label a data{base composed of xvectors with U x , and one composed of yvectors with U y . Examples of the order of indexing for these space vectors are given in Eq. 12.16 part a and b.
( ) ( )
12.3.3 Data Base Permutations
U x = Pxy U y
( ) ( )
The two vectors (arrays) are related by a permutation matrix P such that and
U y = PyxU x
( ) 1
( )
(12.13)
where
Pyx = PT = P; xy xy Now consider the structure of a matrix nitedi erence operator representing 3point centraldi erencing schemes for both space derivatives in two dimensions. When the matrix is multiplying a space vector U , the usual (but ambiguous) representation is given by Ax y . In this notation the ODE form of Eq. 12.11 can be written
+ 4
If it is important to be speci c about the database structure, we use the notation Axx y or Axy y , depending on the data{base chosen for the U it multiplies. Examples are in Eq. 12.16 part a and b. Notice that the matrices are not the same although they represent the same derivative operation. Their structures are similar, however, and they are related by the same permutation matrix that relates U x to U y . Thus
( ) + ( ) + ( ) ( )
dU = A U + (bc) ~ x y dt
+
(12.14)
Axx y = Pxy Axy y Pyx
( ) + ( ) +
(12.15)
Ax+y and U , which are notations used in the special case of space vectors, are subsets of A and ~ , used in the previous sections. u
4 Notice that
222
2 6x 6 6 6 6 6 6 6 6 6 6o 6 6 (x) (x) Ax+y U = 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4
CHAPTER 12. SPLIT AND FACTORED FORMS x
j o x j o x x j o x j o
j j j j
j x j o o j x x j o o j x x j o o j x j j o j o j o j o j x j x x j x j x
3 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 o7 7 7 7 7 7 7 7 7 5 x7
;;
12 22 32 42 13 23 33 43
11 21 31 41
;;
a:Elements in 2dimensional, centraldi erence, matrix operator, Ax y , for 3 4 mesh shown in Sketch 12.12. Data base composed of My x{vectors stored in U x . Entries for x ! x, for y ! o, for both ! .
+ ( )
2 6o 6 6 6 6 6 6 6x 6 6 6 6 6 6 (y ) (y ) Ax+y U = 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4
j x j o j x j o j x j
o
j j j
j o j x j x j o o j x j x j o j x j j x j o j x j x j o o j x j x j o j j j j
+
j x j o j x j o j x j o
( )
3 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 x7 7 7 7 7 7 7 o5
;; ;; ;;
41 42 43 31 32 33 21 22 23
11 12 13
(12.16)
b: Elements in 2dimensional, centraldi erence, matrix operator, Ax y , for 3 4 mesh shown in Sketch 12.12. Data base composed of Mx y{vectors stored in U y . Entries for x ! x, for y ! o, for both ! .
12.3. FACTORING SPACE MATRIX OPERATORS IN 2{D
223
12.3.4 Space Splitting and Factoring
We are now prepared to discuss splitting in two dimensions. It should be clear that the matrix Axx y can be split into two matrices such that
( ) +
Axx y = Axx + Ayx
( ) + ( ) ( ) ( )
( )
(12.17)
where Axx and Ayx are shown in Eq. 12.22. Similarily
Axy y = Axy + Ayy
( ) + ( )
( )
(12.18)
where the split matrices are shown in Eq. 12.23. The permutation relation also holds for the split matrices so
Ayx = Pxy Ayy Pyx
( ) ( )
and
Axx = Pxy Axy Pyx
( ) ( )
The splittings in Eqs. 12.17 and 12.18 can be combined with factoring in the manner described in Section 12.2. As an example ( rstorder in time), applying the implicit Euler method to Eq. 12.14 gives
h i ~ Unx = Unx + h Axx + Ayx Unx + h(bc)
( ) +1 ( ) ( ) ( ) ( ) +1
or
h i ~ I ; hAxx ; hAyx Unx = Unx + h(bc) + O(h )
( ) ( ) ( ) +1 ( ) 2
(12.19) (12.20)
As in Section 12.2, we retain the same rst order accuracy with the alternative
h ih i ~ I ; hAxx I ; hAyx Unx = Unx + h(bc) + O(h )
( ) ( ) ( ) +1 ( ) 2
Write this in predictorcorrector form and permute the data base of the second row. There results
~ ~ I ; hAxx U x = Unx + h(bc) h i y ~ I ; hAyy Un = U y
( ) ( ) ( ) ( ) ( ) ( ) +1
h
i
(12.21)
224
CHAPTER 12. SPLIT AND FACTORED FORMS
Axx
( )
2 x 6x 6 6 6 6 6 6 6 6 6 6 6 6 (x) U =6 6 6 6 6 6 6 6 6 6 6 6 6 6 4
x x x x x x x x
j j j j j x x j x x x j x x x j x x
j j j j j j j j
Ayx
( )
6 6 6o 6 6 U (x) = 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4
j j j j 2 o j o 6 o j o 6 6 6 o j o 6 6 o j o 6 6 j o o j o o j o o j o j o j o j o j o
The splitting of Axx y .
( ) +
7 7 j o 7 7 j o 7 5 j o 7 j o
7 7 j o 7 7 j o 7 Ux 7 j o 7 7 j o7 7 7 7
j x x j x x x j x x j x x 3 j 7 j 7 7 7 j 7 7 j 7 7
7
3 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 5 x7
Ux
( )
(12.22)
( )
12.3. FACTORING SPACE MATRIX OPERATORS IN 2{D
225
Axy
( )
2 x 6 6 6 6 6 6 6 6x 6 6 6 6 6 6 U (y) = 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4
j x j x j x j x j x j
j j j
j x j x j x j x j x j x j x j x j j x j x j x j x j x j x j x j x j
Ayy
( )
6 6 6 6 6 6 6 6 6 (y ) U =6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4
j j j 2 o o j 6o o o j 6 6 o o j 6 6
j x j x j x j x j x j x 3 j j 7 j j 7 7 j j 7 7 j j j
7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 o5
3 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 x7 7 7 7 7 7 7 5
Uy
( )
(12.23)
j o o j j o o o j j o o j j j j j j j
j o o j j o o o j j o o j j j j
( ) +
Uy
( )
j o o j o o j o o
The splitting of Axy y .
226
12.4 SecondOrder Factored Implicit Methods
CHAPTER 12. SPLIT AND FACTORED FORMS
Secondorder accuracy in time can be maintained in a certain factored implicit methods. For example, apply the trapezoidal method to Eq. 12.14 where the derivative operators have been split as in Eq. 12.17 or 12.18. Let the data base be immaterial ~ and the (bc) be time invariant. There results 1 1 ~ I ; 2 hAx ; 1 hAy Un = I + 2 hAx + 1 hAy Un + h(bc) + O(h ) (12.24) 2 2 Factor both sides giving 1 I ; 1 hAx I ; 2 hAy ; 1 h AxAy Un 2 4 1 hA I + 1 hA ; 1 h A A U + h(bc) + O(h ) (12.25) ~ = I+ 2 x 2 y 4 x y n Then notice that the combination 1 h AxAy ](Un ; Un) is proportional to h since 4 the leading term in the expansion of (Un ; Un) is proportional to h. Therefore, we can write 1 ~ I ; 1 hAx I ; 1 hAy Un = I + 2 hAx I + 1 hAy Un + h(bc) + O(h )(12.26) 2 2 2 and both the factored and unfactored form of the trapezoidal method are secondorder accurate in the time march. An alternative form of this kind of factorization is the classical ADI (alternating direction implicit) method usually written 1 1 ~ I ; 2 hAx U = I + 2 hAy Un + 1 hFn 2 1 hA U + 1 hF + O(h ) 1 hA U (12.27) I;2 y n = I+2 x ~ 2 n
+1 3 2 +1 2 3 2 +1 3 +1 +1 3 5 +1 +1 3
For idealized commuting systems the methods given by Eqs. 12.26 and 12.27 di er only in their evaluation of a timedependent forcing term.
12.5 Importance of Factored Forms in 2 and 3 Dimensions
When the timemarch equations are sti and implicit methods are required to permit reasonably large time steps, the use of factored forms becomes a very valuable tool
5 A form of the Douglas or PeacemanRachford methods.
12.5. IMPORTANCE OF FACTORED FORMS IN 2 AND 3 DIMENSIONS 227
for realistic problems. Consider, for example, the problem of computing the time advance in the unfactored form of the trapezoidal method given by Eq. 12.24 1 ~ I ; 2 hAx y Un = I + 1 hAx y Un + h(bc) 2 Forming the right hand side poses no problem, but nding Un requires the solution of a sparse, but very large, set of coupled simultaneous equations having the matrix form shown in Eq. 12.16 part a and b. Furthermore, in real cases involving the Euler or NavierStokes equations, each symbol (o x ) represents a 4 4 block matrix with entries that depend on the pressure, density and velocity eld. Suppose we were to solve the equations directly. The forward sweep of a simple Gaussian elimination lls all of the 4 4 blocks between the main and outermost diagonal (e.g. between and o in Eq. 12.16 part b.). This must be stored in computer memory to be used to nd the nal solution in the backward sweep. If Ne represents the order of the small block matrix (4 in the 2D Euler case), the approximate memory requirement is (Ne My ) (Ne My ) Mx oating point words. Here it is assumed that My < Mx. If My > Mx, My and Mx would be interchanged. A moderate mesh of 60 200 points would require over 11 million words to nd the solution. Actually current computer power is able to cope rather easily with storage requirements of this order of magnitude. With computing speeds of over one giga op, direct solvers may become useful for nding steadystate solutions of practical problems in two dimensions. However, a threedimensional solver would require a memory of approximatly
+ +1 + +1 6 7 8
words and, for well resolved ow elds, this probably exceeds memory availability for some time to come. On the other hand, consider computing a solution using the factored implicit equation 12.25. Again computing the right hand side poses no problem. Accumulate the result of such a computation in the array (RHS ). One can then write the remaining terms in the twostep predictorcorrector form 1 ~ I ; 2 hAxx U x = (RHS ) x ~ I ; 1 hAyy Uny = U y (12.28) 2
( ) ( ) ( ) ( ) ( ) +1 ( )
Ne My Mz Mx
2 2 2
practical size the ll is mostly dense. 7 The lower band is also computed but does not have to be saved unless the solution is to be repeated for another vector. 8 One billion oatingpoint operations per second.
6 For matrices as small as those shown there are many gaps in this \ ll", but for meshes of
228
CHAPTER 12. SPLIT AND FACTORED FORMS
9
which has the same appearance as Eq. 12.21 but is secondorder time accurate. The rst step would be solved using My uncoupled block tridiagonal solvers . Inspecting the top of Eq. 12.22, we see that this is equivalent to solving My onedimensional ~ problems, each with Mx blocks of order Ne . The temporary solution U x would then y and an inspection of the bottom of Eq. 12.23 shows that the nal ~ be permuted to U step consists of solving Mx onedimensional implicit problems each with dimension My .
( ) ( )
12.6 The Delta Form
Clearly many ways can be devised to split the matrices and generate factored forms. One way that is especially useful, for ensuring a correct steadystate solution in a converged timemarch, is referred to as the \delta form" and we develop it next. Consider the unfactored form of the trapezoidal method given by Eq. 12.24, and ~ let the (bc) be time invariant: 1 1 1 ~ I ; 2 hAx ; 2 hAy Un = I + 2 hAx + 1 hAy Un + h(bc) + O(h ) 2 From both sides subtract 1 I ; 2 hAx ; 1 hAy Un 2 leaving the equality unchanged. Then, using the standard de nition of the di erence operator , Un = Un ; Un one nds h 1 1 ~ i I ; 2 hAx ; 2 hAy Un = h Ax y Un + (bc) + O(h ) (12.29) Notice that the right side of this equation is the product of h and a term that is identical to the right side of Eq. 12.14, our original ODE. Thus, if Eq. 12.29 converges, it is guaranteed to converge to the correct steadystate solution of the ODE. Now we can factor Eq. 12.29 and maintain O(h ) accuracy. We arrive at the expression h 1 ~ i (12.30) I ; 2 hAx I ; 1 hAy Un = h Ax y Un + (bc) + O(h ) 2 This is the delta form of a factored, 2ndorder, 2D equation.
+1 3 +1 + 3 2 + 3
replace the scalar ones, and matrix multiplications do not commute.
9 A block tridiagonal solver is similar to a scalar solver except that small block matrix operations
12.7. PROBLEMS
229
The point at which the factoring is made may not a ect the order of timeaccuracy, but it can have a profound e ect on the stability and convergence properties of a method. For example, the unfactored form of a rstorder method derived from the implicit Euler time march is given by Eq. 12.19, and if it is immediately factored, the factored form is presented in Eq. 12.20. On the other hand, the delta form of the unfactored Eq. 12.19 is
h ~ i I ; hAx ; hAy ] Un = h Ax y Un + (bc)
+ 10
and its factored form becomes
h ~ i I ; hAx] I ; hAy ] Un = h Ax y Un + (bc)
+
(12.31)
In spite of the similarities in derivation, we will see in the next chapter that the convergence properties of Eq. 12.20 and Eq. 12.31 are vastly di erent.
12.7 Problems
1. Consider the 1D heat equation:
@u = @ u @t @x
2 2
0 x 9
Let u(0 t) = 0 and u(9 t) = 0, so that we can simplify the boundary conditions. Assume that second order central di erencing is used, i.e., ( u) = 1 (u ; 2u + u )
xx j
The uniform grid has x = 1 and 8 interior points. (a) Space vector de nition i. What is the space vector for the natural ordering (monotonically increasing in index), u ? Only include the interior points. ii. If we reorder the points with the odd points rst and then the even points, write the space vector, u ? iii. Write down the permutation matrices,(P ,P ).
(1) (2) 12 21
x
2
j ;1
j
j +1
10 Notice that the only di erence between the O(h2 ) method given by Eq. 12.30 and the O(h) method given by Eq. 12.31 is the appearance of the factor 1 on the left side of the O(h2 ) method.
2
230
CHAPTER 12. SPLIT AND FACTORED FORMS
iv. The generic ODE representing the discrete form of the heat equation is
(1) 1 (1)
v. Applying implicit Euler time marching, write the delta form of the implicit algorithm. Comment on the form of the resulting implicit matrix operator. (b) System de nition In problem 1a, we de ned u u A A P , and P which partition the odd points from the even points. We can put such a partitioning to use. First de ne extraction operators 21 0 0 0 0 0 0 03 60 1 0 0 0 0 0 07 7 6 60 0 1 0 3 0 0 0 0 7 2I 7 6 7 6 0 7 6 60 0 0 1 0 0 0 07=4 7 Io =60 0 0 0 5 6 0 0 0 07 0 7 6 0 6 7 60 0 0 0 0 0 0 07 7 6 40 0 0 0 0 0 0 05 0 0 0 0 0 0 0 0 20 0 0 0 0 0 0 03 60 0 0 0 0 0 0 07 7 6 60 0 0 0 3 0 0 0 0 7 20 7 6 7 6 0 7 6 60 0 0 0 0 0 0 07=4 7 Ie =60 0 0 0 5 6 1 0 0 07 0 7 6 I 7 6 60 0 0 0 0 1 0 07 7 6 40 0 0 0 0 0 1 05 0 0 0 0 0 0 0 1
(1) (2) 1 2 12 21 ( ) 4 4 4 4 ( ) 4 4 4 4
du = A u + f dt Write down the matrix A . (Note f = 0, due to the boundary conditions) Next nd the matrix A such that du = A u dt Note that A can be written as 2 3 D UT 7 A =6 4 5 U D De ne D and U .
1 2 (2) 2 (2) 2 2
12.7. PROBLEMS
(2) ( ) ( ) (2) ( ) ( ) (2)
231
which extract the odd even points from u as follows: u o = I o u and ue =Ieu . i. Beginning with the ODE written in terms of u , de ne a splitting A = Ao + Ae, such that Ao operates only on the odd terms, and Ae operates only on the even terms. Write out the matrices Ao and Ae. Also, write them in terms of D and U de ned above. ii. Apply implicit Euler time marching to the split ODE. Write down the delta form of the algorithm and the factored delta form. Comment on the order of the error terms. iii. Examine the implicit operators for the factored delta form. Comment on their form. You should be able to argue that these are now trangular matrices (a lower and an upper). Comment on the solution process this gives us relative to the direct inversion of the original system.
(2) 2
232
CHAPTER 12. SPLIT AND FACTORED FORMS
Chapter 13 LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS
In Section 4.4 we introduced the concept of the representative equation, and used it in Chapter 7 to study the stability, accuracy, and convergence properties of timemarching schemes. The question is: Can we nd a similar equation that will allow us to evaluate the stability and convergence properties of split and factored schemes? The answer is yes  for certain forms of linear model equations. The analysis in this chapter is useful for estimating the stability and steadystate properties of a wide variety of timemarching schemes that are variously referred to as timesplit, fractionalstep, hybrid, and (approximately) factored. When these methods are applied to practical problems, the results found from this analysis are neither necessary nor su cient to guarantee stability. However, if the results indicate that a method has an instability, the method is probably not suitable for practical use.
13.1 The Representative Equation for Circulant Operators
Consider linear PDE's with coe cients that are xed in both space and time and with boundary conditions that are periodic. We have seen that under these conditions a semidiscrete approach can lead to circulant matrix di erence operators, and we discussed circulant eigensystems1 in Section 4.3. In this and the following section we assume circulant systems and our analysis depends critically on the fact that all circulant matrices commute and have a common set of eigenvectors.
1
See also the discussion on Fourier stability analysis in Section 7.7.
233
234 CHAPTER 13. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS Suppose, as a result of space di erencing the PDE, we arrive at a set of ODE's that can be written d~ = A ~ + A ~ ; ~ (t) u (13.1) apu b pu f dt where the subscript p denotes a circulant matrix. Since both matrices have the same set of eigenvectors, we can use the arguments made in Section 4.2.3 to uncouple the set and form the M set of independent equations
0 w1 = ( a + b)1 w1 ; g1 (t) ... 0 wm = ( a + b)m wm ; gm(t) ... 0 wM = ( a + b)M wM ; gM (t)
(13.2)
The analytic solution of the m'th line is
wm(t) = cm e( a+ b)mt + P:S:
Note that each a pairs with one, and only one2 , eigenvector. This suggests (see Section 4.4:
b
since they must share a common
The representative equation for split, circulant systems is du = + + + ]u + ae t a b c dt where a + b + c + are the sum of the eigenvalues in Aa Ab Ac share the same eigenvector.
(13.3) that
13.2 Example Analysis of Circulant Systems
13.2.1 Stability Comparisons of TimeSplit Methods
Consider as an example the linear convectiondi usion equation: @u + a @u = @ 2 u @t @x @x2
2
(13.4)
This is to be contrasted to the developments found later in the analysis of 2D equations.
13.2. EXAMPLE ANALYSIS OF CIRCULANT SYSTEMS
235
If the space di erencing takes the form d~ = ; a B (;1 0 1)~ + u u B (1 ;2 1)~ u (13.5) dt 2 x p x2 p the convection matrix operator and the di usion matrix operator, can be represented by the eigenvalues c and d, respectively, where (see Section 4.3.2): ( c)m = ia sin m x m (13.6) ( d)m = ; 4 2 sin2 2 x In these equations m = 2m =M , m = 0 1 M ; 1 , so that 0 m 2 . Using these values and the representative equation 13.4, we can analyze the stability of the two forms of simple timesplitting discussed in Section 12.2. In this section we refer to these as 1. the explicitimplicit Euler method, Eq. 12.10. 2. the explicitexplicit Euler method, Eq. 12.8. When applied to Eq. 13.4, the characteristic polynomial of this method is P (E ) = (1 ; h d )E ; (1 + h c) This leads to the principal root 1 + i ah sin m x = h sin2 m 1+4 2 2 x where we have made use of Eq. 13.6 to quantify the eigenvalues. Now introduce the dimensionless numbers Cn = ah Courant number x mesh Reynolds number R = a x and we can write for the absolute value of q 2 1 + Cn sin2 m j j= C m 1 + 4 R n sin2 2 0
m
1. The ExplicitImplicit Method
2
(13.7)
236 CHAPTER 13. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS
1.2 C = 2/ R ∆ n 0.8 C n C 0.8 n 1.2 C = R /2 n ∆ C = 2/ R n ∆
0.4
0.4
0
0
4 R∆ ExplicitImplicit
8
0
0
4 R∆ ExplicitExplicit
8
_ Figure 13.1: Stability regions for two simple timesplit methods. A simple numerical parametric study of Eq. 13.7 shows that the critical range of m for any combination of Cn and R occurs when m is near 0 (or 2 ). From this we nd that the condition on Cn and R that make j j 1 is 2 h i C 2 1 + Cn sin2 = 1 + 4 R n sin2 2 As ! 0 this gives the stability region 2 Cn < R which is bounded by a hyperbola and shown in Fig. 13.1.
2. The ExplicitExplicit Method
An analysis similar to the one given above shows that this method produces " # q 2 m 1 ; 4 Cn sin2 m 2 j j = 1 + Cn sin 0 m 2 R 2 Again a simple numerical parametric study shows that this has two critical ranges of m , one near 0, which yields the same result as in the previous example, and the
13.2. EXAMPLE ANALYSIS OF CIRCULANT SYSTEMS
237
other near 180o, which produces the constraint that 1 Cn < 2 R for R 2 The resulting stability boundary is also shown in Fig. 13.1. The totaly explicit, factored method has a much smaller region of stability when R is small, as we should have expected. Next let us analyze a more practical method that has been used in serious computational analysis of turbulent ows. This method applies to a ow in which there is a combination of di usion and periodic convection. The convection term is treated explicitly using the secondorder AdamsBashforth method. The di usion term is integrated implicitly using the trapezoidal method. Our model equation is again the linear convectiondi usion equation 13.4 which we split in the fashion of Eq. 13.5. In order to evaluate the accuracy, as well as the stability, we include the forcing function in the representative equation and study the e ect of our hybrid, timemarching method on the equation u0 = cu + du + ae t First let us nd expressions for the two polynomials, P (E ) and Q(E ). The characteristic polynomial follows from the application of the method to the homogeneous equation, thus 1 1 un+1 = un + 2 h c(3un ; un;1) + 2 h d (un+1 + un) This produces 3 P (E ) = (1 ; 1 h d)E 2 ; (1 + 2 h c + 1 h d)E + 1 h c 2 2 2 The form of the particular polynomial depends upon whether the forcing function is carried by the AB2 method or by the trapezoidal method. In the former case it is 1 Q(E ) = 2 h(3E ; 1) (13.8) and in the latter (13.9) Q(E ) = 1 h(E 2 + E ) 2
13.2.2 Analysis of a SecondOrder TimeSplit Method
238 CHAPTER 13. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS
Accuracy
From the characteristic polynomial we see that there are two {roots and they are given by the equation 3 1 + 2h c + 1h 2 =
s
d
The principal root follows from the plus sign and one can show 1 1 1 = 1 + ( c + d )h + 2 ( c + d )2 h2 + 4 3 + c 2 ; 2 d ; 3 h3 d d c c 1 1 From this equation it is clear that 6 3 = 6 ( c + d)3 does not match the coe cient of h3 in 1 , so er = O(h3) Using P (e h) and Q(e h ) to evaluate er in Section 6.6.3, one can show er = O(h3) using either Eq. 13.8 or Eq. 13.9. These results show that, for the model equation, the hybrid method retains the secondorder accuracy of its individual components.
1 1 + 3h c + 2h 2 2 1 ; 1h d 2
d
2
1 ; 2h c 1 ; 2 h
d
(13.10)
Stability
The stability of the method can be found from Eq. 13.10 by a parametric study of cn and R de ned in Eq. 13.7. This was carried out in a manner similar to that used to nd the stability boundary of the rstorder explicitimplicit method in Section 13.2.1. The results are plotted in Fig. 13.2. For values of R 2 this secondorder method has a much greater region of stability than the rstorder explicitimplicit method given by Eq. 12.10 and shown in Fig. 13.1.
13.3 The Representative Equation for SpaceSplit Operators
Consider the 2D model3 equations @u = @ 2 u + @ 2 u @t @x2 @y2
3
(13.11)
The extension of the following to 3D is simple and straightforward.
13.3. THE REPRESENTATIVE EQUATION FOR SPACESPLIT OPERATORS239
1.2
0.8
Cn
0.4
Stable
0
4
8 ∆
R
_ Figure 13.2: Stability regions for the secondorder timesplit method. and
@u + a @u + a @u = 0 (13.12) @t x @x y @y Reduce either of these, by means of spatial di erencing approximations, to the coupled set of ODE's: dU = A + A ]U + (bc) ~ (13.13) x y dt for the space vector U . The form of the Ax and Ay matrices for threepoint central di erencing schemes are shown in Eqs. 12.22 and 12.23 for the 3 4 mesh shown in Sketch 12.12. Let us inspect the structure of these matrices closely to see how we can diagonalize Ax + Ay ] in terms of the individual eigenvalues of the two matrices considered separately. First we write these matrices in the form 2 3 2 ~ 3 B b0 I ~1 I b A(xx) = 6 B 7 A(yx) = 6 ~;1 I ~0 I ~1 I 7 b b 5 4 5 4b ~;1 I ~0 I B b b
where B is a banded matrix of the form B (b;1 b0 b1 ). Now nd the block eigenvector
240 CHAPTER 13. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS matrix that diagonalizes B and use it to diagonalize A(xx). Thus 2 ;1 32 32 3 2 X B X 6 76 7=6 X ;1 B 76 X 4 54 54 5 4 ;1 X B X where 2 3
6 =6 6 4 3 7 5
1
2
set X
4 (x) is transparent to this transformation. That is, if we Notice that the matrix Ay
3
7 7 7 5
3 2 ~ 3 ~1 I b b0 I ~1 I b ~0 I ~1 I 7 X = 6 ~;1 I ~0 I ~1 I 7 b b 5 b b 5 4b ~;1 I ~0 I ~;1 I ~0 I b b b b One now permutes the transformed system to the yvector database using the permutation matrix de ned by Eq. 12.13. There results h i Pyx X ;1 A(xx) + A(yx) X Pxy = 3 2 3 2 ~ B 1 I 7 6 7 6 ~ 7 B 6 7+6 2 I 6 7 6 7 6 (13.14) ~ 4 5 4 B 7 3 I 5 ~ B 4 I ~ where B is the banded tridiagonal matrix B (~;1 ~0 ~1), see the bottom of Eq. 12.23. b b b ~ that diagonalize the B blocks. Let B diag(B ) and ~ ~ ~ Next nd the eigenvectors X ~ diag(X ) and form the second transformation ~ X 2~ 3 2~ 3 1 6 7 7 ~2 ~ ~~ 6 ~ ~ 7 ~=6 X ;1B X = 6 7 4 5 4 5 ~3 ~ This time, by the same argument as before, the rst matrix on the right side of Eq. 13.14 is transparent to the transformation, so the nal result is the complete diagonalization of the matrix Ax+y : h ih ih i ~ ~ X ;1 Pyx X ;1 A(xx+)y X Pxy X = 2 3 1I + ~ 6 7 6 7 2I + ~ 6 7 (13.15) 6 7 4 5 3I + ~ 4I + ~
diag(X )
2 ~ b0 X ;1 6 ~;1 4b
I I
13.3. THE REPRESENTATIVE EQUATION FOR SPACESPLIT OPERATORS241
It is important to notice that: The diagonal matrix on the right side of Eq. 13.15 contains every possible com~ bination of the individual eigenvalues of B and B . Now we are ready to present the representative equation for two dimensional systems. First reduce the PDE to ODE by some choice4 of space di erencing. This results in a spatially split A matrix formed from the subsets
A(xx) = diag(B )
~ A(yy) = diag(B )
(13.16)
~ where B and B are any two matrices that have linearly independent eigenvectors (this puts some constraints on the choice of di erencing schemes). Although Ax and Ay do commute, this fact, by itself, does not ensure the property of \all possible combinations". To obtain the latter property the structure of ~ the matrices is important. The block matrices B and B can be either circulant or noncirculant in both cases we are led to the nal result: The 2{D representative equation for model linear systems is du = + ]u + ae t x y dt where x and y are any combination of eigenvalues from Ax and Ay , a and are (possibly complex) constants, and where Ax and Ay satisfy the conditions in 13.16. Often we are interested in nding the value of, and the convergence rate to, the steadystate solution of the representative equation. In that case we set = 0 and use the simpler form du = + ]u + a (13.17) x y dt which has the exact solution a u(t) = ce( x+ y )t ; + (13.18) x y
We have used 3point central di erencing in our example, but this choice was for convenience only, and its use is not necessary to arrive at Eq. 13.15.
4
242 CHAPTER 13. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS
13.4 Example Analysis of 2D Model Equations
1. The stability. 2. The accuracy of the xed, steadystate solution. 3. The convergence rate to reach the steadystate.
In the following we analyze four di erent methods for nding a xed, steadystate solution to the 2D representative equation 13.16. In each case we examine
13.4.1 The Unfactored Implicit Euler Method
Consider rst this unfactored, rstorder scheme which can then be used as a reference case for comparison with the various factored ones. The form of the method is given by Eq. 12.19, and when it is applied to the representative equation, we nd (1 ; h from which
x ; h y )un+1
= un + ha
P (E ) = (1 ; h x ; h y )E ; 1 Q(E ) = h
giving the solution
(13.19)
un = c 1 ; h ; h ; x y Like its counterpart in the 1D case, this method:
1. Is unconditionally stable.
"
1
#n
a x+
y
2. Produces the exact (see Eq. 13.18) steadystate solution (of the ODE) for any h. 3. Converges very rapidly to the steadystate when h is large. Unfortunately, however, use of this method for 2D problems is generally impractical for reasons discussed in Section 12.5.
13.4. EXAMPLE ANALYSIS OF 2D MODEL EQUATIONS
243
Now apply the factored Euler method given by Eq. 12.20 to the 2D representative equation. There results (1 ; h x)(1 ; h y )un+1 = un + ha from which P (E ) = (1 ; h x)(1 ; h y )E ; 1 Q(E ) = h (13.20) giving the solution " #n 1 un = c (1 ; h )(1 ; h ) ; + a h x y x y; x y We see that this method: 1. Is unconditionally stable. 2. Produces a steady state solution that depends on the choice of h. 3. Converges rapidly to a steadystate for large h, but the converged solution is completely wrong. The method requires far less storage then the unfactored form. However, it is not very useful since its transient solution is only rstorder accurate and, if one tries to take advantage of its rapid convergence rate, the converged value is meaningless. Next apply Eq. 12.31 to the 2D representative equation. One nds (1 ; h x)(1 ; h y )(un+1 ; un) = h( xun + y un + a) which reduces to (1 ; h x)(1 ; h y )un+1 = 1 + h2 x y un + ha and this has the solution " #n 1 + h2 x y a un = c (1 ; h )(1 ; h ) ; + x y x y This method:
13.4.2 The Factored Nondelta Form of the Implicit Euler Method
13.4.3 The Factored Delta Form of the Implicit Euler Method
244 CHAPTER 13. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 1. Is unconditionally stable. 2. Produces the exact steadystate solution for any choice of h. 3. Converges very slowly to the steady{state solution for large values of h, since j j ! 1 as h ! 1. Like the factored nondelta form, this method demands far less storage than the unfactored form, as discussed in Section 12.5. The correct steady solution is obtained, but convergence is not nearly as rapid as that of the unfactored form. Finally consider the delta form of a secondorder timeaccurate method. Apply Eq. 12.30 to the representative equation and one nds 1 1 ; 2 h x 1 ; 1 h y (un+1 ; un) = h( xun + y un + a) 2 which reduces to 1 ; 1 h x 1 ; 1 h y un+1 = 1 + 1 h x 1 + 1 h y un + ha 2 2 2 2 and this has the solution 2 3n 1 + 1h x 1 + 1h y 7 6 a 2 2 7 ; un = c6 4 5 1 x+ y 1 ; 1h x 1 ; 2h y 2 This method: 1. Is unconditionally stable. 2. Produces the exact steady{state solution for any choice of h. 3. Converges very slowly to the steady{state solution for large values of h, since j j ! 1 as h ! 1. All of these properties are identical to those found for the factored delta form of the implicit Euler method. Since it is second order in time, it can be used when time accuracy is desired, and the factored delta form of the implicit Euler method can be used when a converged steadystate is all that is required.5 A brief inspection of eqs. 12.26 and 12.27 should be enough to convince the reader that the 's produced by those methods are identical to the produced by this method.
In practical codes, the value of h on the left side of the implicit equation is literally switched from h to 1 h. 2
5
13.4.4 The Factored Delta Form of the Trapezoidal Method
13.5. EXAMPLE ANALYSIS OF THE 3D MODEL EQUATION
13.5 Example Analysis of the 3D Model Equation
245
The arguments in Section 13.3 generalize to three dimensions and, under the conditions given in 13.16 with an A(zz) included, the model 3D cases6 have the following representative equation (with = 0): du = + + ]u + a (13.21) x y z dt Let us analyze a 2ndorder accurate, factored, delta form using this equation. First apply the trapezoidal method: un+1 = un + 1 h ( x + y + z )un+1 + ( x + y + z )un + 2a] 2 Rearrange terms: 1 1 1 ; 2 h( x + y + z ) un+1 = 1 + 2 h( x + y + z ) un + ha Put this in delta form: 1 ; 1 h( x + y + z ) un = h ( x + y + z )un + a] 2 Now factor the left side: 1 1 ; 2 h x 1 ; 1 h y 1 ; 1 h z un = h ( x + y + z )un + a] (13.22) 2 2 This preserves second order accuracy since the error terms 1 h2( 1 h3 x y + x z + y z ) un and 4 8 x y z are both O(h3). One can derive the characteristic polynomial for Eq. 13.22, nd the root, and write the solution either in the form 2 3n 1 h( + + ) + 1 h2 ( 1 h3 + + ); x y z 61 + 4 x y x z y z 8 x y z7 7 un = c6 2 4 5 1 h( + + ) + 1 h2( 1 h3 1 ; 2 x y z 4 x y + x z + y z) ; 8 x y z
; +a+ x y
6
z
(13.23)
Eqs. 13.11 and 13.12, each with an additional term.
246 CHAPTER 13. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS or in the form
2 6 un = c6 4
1 1 1 + 1 h x 1 + 2 h y 1 + 1 h z ; 4 h3 2 2 1h 1h 1 ; 2 x 1 ; 2 y 1 ; 1h z 2
3n x y z7 7 ; 5
x+ y+ z
a
(13.24)
It is interesting to notice that a Taylor series expansion of Eq. 13.24 results in 1 = 1 + h( x + y + z ) + 2 h2( x + y + z )2 (13.25) h i + 1 h3 3 + (2 y + 2 x) + 2 2 + 3 x y + 2 2 + 3 + 2 x 2 + 2 2 y + 3 + y y y y x x 4 z which veri es the second order accuracy of the factored form. Furthermore, clearly, if the method converges, it converges to the proper steadystate.7 With regards to stability, it follows from Eq. 13.23 that, if all the 's are real and negative, the method is stable for all h. This makes the method unconditionally stable for the 3D di usion model when it is centrally di erenced in space. Now consider what happens when we apply this method to the biconvection model, the 3D form of Eq. 13.12 with periodic boundary conditions. In this case, central di erencing causes all of the 's to be imaginary with spectrums that include both positive and negative values. Remember that in our analysis we must consider every possible combination of these eigenvalues. First write the root in Eq. 13.23 in the form 1+ = 1;i ; +i i ; +i
where , and are real numbers that can have any sign. Now we can always nd one combination of the 's for which , and are both positive. In that case since the absolute value of the product is the product of the absolute values
j
2 2 2 = (1 ; ) + ( + ) j (1 ; )2 + ( ; )2
>1
and the method is unconditionally unstable for the model convection problem. From the above analysis one would come to the conclusion that the method represented by Eq. 13.22 should not be used for the 3D Euler equations. In practical cases, however, some form of dissipation is almost always added to methods that are used to solve the Euler equations and our experience to date is that, in the presence of this dissipation, the instability disclosed above is too weak to cause trouble.
7
However, we already knew this because we chose the delta form.
13.6. PROBLEMS
13.6 Problems
247
1. Starting with the generic ODE, du = Au + f dt we can split A as follows: A = A1 + A2 + A3 + A4. Applying implicit Euler time marching gives un+1 ; un = A u + A u + A u + A u + f 1 n+1 2 n+1 3 n+1 4 n+1 h (a) Write the factored delta form. What is the error term? (b) Instead of making all of the split terms implicit, leave two explicit: un+1 ; un = A u + A u + A u + A u + f 1 n+1 2 n 3 n+1 4 n h Write the resulting factored delta form and de ne the error terms. (c) The scalar representative equation is du = ( + + + )u + a 1 2 3 4 dt For the fully implicit scheme of problem 1a, nd the exact solution to the resulting scalar di erence equation and comment on the stability, convergence, and accuracy of the converged steadystate solution. (d) Repeat 1c for the explicitimplicit scheme of problem 1b.
248 CHAPTER 13. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS
Appendix A USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA
A basic understanding of the fundamentals of linear algebra is crucial to our development of numerical methods and it is assumed that the reader is at least familar with this subject area. Given below is some notation and some of the important relations between matrices and vectors.
A.1 Notation
1. In the present context a vector is a vertical column or string. Thus 2 v1 3 6 7 ~ = 6 v..2 7 v 6 7 4 . 5 vm and its transpose ~ is the horizontal row vT ~ T = v1 v2 v3 : : : vm] ~ = v1 v2 v3 : : : vm ]T v v 2. A general m m matrix A can be written 2 a11 a12 a1m 3 6a a a 7 A = (aij ) = 6 21 22 . . . 2m 7 7 6 5 4 am1 am2 amm 249
250APPENDIX A. USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA 3. An alternative notation for A is h i A = ~1 ~2 : : : ~m a a a and its transpose AT is 2 ~T 3 a 6 ~1 7 6 aT 7 AT = 6 .2 7 6 7 6 .. 7 4 T5 ~m a
4. The inverse of a matrix (if it exists) is written A;1 and has the property that A;1A = AA;1 = I , where I is the identity matrix.
A.2 De nitions
1. A is symmetric if AT = A. 2. A is skewsymmetric or antisymmetric if AT = ;A. 3. A is diagonally dominant if aii Pj6=i jaij j i = 1 2 : : : m and aii > Pj6=i jaij j for at least one i. 4. A is orthogonal if aij are real and AT A = AAT = I 5. A is the complex conjugate of A. 6. P is a permutation matrix if P ~ is a simple reordering of ~ . v v 7. The trace of a matrix is Pi aii. 8. A is normal if AT A = AAT . 9. det A] is the determinant of A. 10. AH is the conjugate transpose of A, (Hermitian). 11. If b A= a d c then det A] = ad ; bc and d A;1 = det1 A] ;c ;b a
A.3. ALGEBRA
A.3 Algebra
We consider only square matrices of the same dimension. 1. A and B are equal if aij = bij for all i j = 1 2 : : : m. 2. A + (B + C ) = (C + A) + B , etc. 3. sA = (saij ) where s is a scalar. 4. In general AB 6= BA. 5. Transpose equalities: (A + B )T = AT + B T (AT )T = A (AB )T = B T AT 6. Inverse equalities (if the inverse exists): (A;1);1 = A (AB );1 = B ;1 A;1 (AT );1 = (A;1 )T
251
7. Any matrix A can be expressed as the sum of a symmetric and a skewsymmetric matrix. Thus: 1 A = 2 A + AT + 1 A ; AT 2
A.4 Eigensystems
1. The eigenvalue problem for a matrix A is de ned as A~ = ~ or A ; I ]~ = 0 x x x and the generalized eigenvalue problem, including the matrix B , as A~ = B~ or A ; B ]~ = 0 x x x 2. If a square matrix with real elements is symmetric, its eigenvalues are all real. If it is asymmetric, they are all imaginary.
252APPENDIX A. USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA 3. Gershgorin's theorem: The eigenvalues of a matrix lie in the complex plane in the union of circles having centers located by the diagonals with radii equal to the sum of the absolute values of the corresponding o diagonal row elements. 4. In general, an m m matrix A has n~ linearly independent eigenvectors with x n~ m and n distinct eigenvalues ( i) with n n~ m. x x 5. A set of eigenvectors is said to be linearly independent if a ~ m + b ~ n 6= ~ k x x x m 6= n 6= k for any complex a and b and for all combinations of vectors in the set. 6. If A posseses m linearly independent eigenvectors then A is diagonalizable, i.e.,
X ;1AX =
where X is a matrix whose columns are the eigenvectors, h i X = ~1 ~2 : : : ~m x x x and is the diagonal matrix
0 0 m If A can be diagonalized, its eigenvectors completely span the space, and A is said to have a complete eigensystem. 7. If A has m distinct eigenvalues, then A is always diagonalizable, and with each distinct eigenvalue there is one associated eigenvector, and this eigenvector cannot be formed from a linear combination of any of the other eigenvectors. 8. In general, the eigenvalues of a matrix may not be distinct, in which case the possibility exists that it cannot be diagonalized. If the eigenvalues of a matrix are not distinct, but all of the eigenvectors are linearly independent, the matrix is said to be derogatory, but it can still be diagonalized. 9. If a matrix does not have a complete set of linearly independent eigenvectors, it cannot be diagonalized. The eigenvectors of such a matrix cannot span the space and the matrix is said to have a defective eigensystem.
2 0 03 1 6 . . . .. 7 6 7 = 6 0 . .2 . . . 7 6 ... . . 0 7 4 5
A.4. EIGENSYSTEMS 2J 0 03 1 .7 6 . ;1 AS = 6 0 J2 . . . 7 6. . . .7 J =S 6 .. . . . . 0 7 4 5 0 0 Jk
253
10. Defective matrices cannot be diagonalized but they can still be put into a compact form by a similarity transform, S , such that
where there are k linearly independent eigenvectors and Ji is either a Jordan subblock or i. 11. A Jordan submatrix has the form
2 0 03 i 1 60 . . . ... 7 6 7 i 1 6 60 0 7 ... 0 7 Ji = 6 7 i 6. 6 .. 7 ... ... 1 7 4 5
0 0 0
i
12. Use of the transform S is known as putting A into its Jordan Canonical form. A repeated root in a Jordan block is referred to as a defective eigenvalue. For each Jordan submatrix with an eigenvalue i of multiplicity r, there exists one eigenvector. The other r ; 1 vectors associated with this eigenvalue are referred to as principal vectors. The complete set of principal vectors and eigenvectors are all linearly independent. 13. Note that if P is the permutation matrix
2 3 0 0 1 P = 60 1 07 4 5
1 0 0
P T = P ;1 = P
then
2 3 2 3 1 0 0 0 P ;1 6 0 17P = 61 07 4 5 4 5
0 0 0 1
14. Some of the Jordan subblocks may have the same eigenvalue. For example, the
254APPENDIX A. USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA matrix
22 66 64 6 6 6 6 6 6 6 6 6 6 6 4
1
1
1
3 17 5
1 1 1
1
1 2
1
2 3
3 7 7 7 7 7 7 7 7 7 7 7 7 7 5
is both defective and derogatory, having: 9 eigenvalues 3 distinct eigenvalues 3 Jordan blocks 5 linearly independent eigenvectors 3 principal vectors with 1 1 principal vector with 2
A.5 Vector and Matrix Norms
1. The spectral radius of a matrix A is symbolized by (A) such that (A) = j m jmax where
m
are the eigenvalues of the matrix A.
2. A pnorm of the vector ~ is de ned as v
0M 1 X pA1=p jjvjjp = @ jvj j
j =1
3. A pnorm of a matrix A is de ned as
Av jjAjjp = max jjjjvjjjjp x6=0
p
A.5. VECTOR AND MATRIX NORMS
255
4. Let A and B be square matrices of the same order. All matrix norms must have the properties
jjAjj 0 jjAjj = 0 implies A = 0 jjc Ajj = jcj jjAjj jjA + B jj jjAjj + jjB jj jjA B jj jjAjj jjB jj
5. Special pnorms are jjAjj1 = maxj=1
M
jjAjj1 = maxi=1 2 M PM jaij j maximum row sum j =1 where jjAjjp is referred to as the Lp norm of A.
6. In general (A) does not satisfy the conditions in 4, so in general (A) is not a true norm. 7. When A is normal, (A) is a true norm, in fact, in this case it is the L2 norm. 8. The spectral radius of A, (A), is the lower bound of all the norms of A.
jjAjj2 =
q
PM ja j maximum column sum i=1 ij
(A T A )
256APPENDIX A. USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA
Appendix B SOME PROPERTIES OF TRIDIAGONAL MATRICES
B.1 Standard Eigensystem for Simple Tridiagonals
In this work tridiagonal banded matrices are prevalent. It is useful to list some of their properties. Many of these can be derived by solving the simple linear di erence equations that arise in deriving recursion relations. Let us consider a simple tridiagonal matrix, i.e., a tridiagonal with constant scalar elements a,b, and c, see Section 3.4. If we examine the conditions under which the determinant of this matrix is zero, we nd (by a recursion exercise) det B (M : a b c)] = 0 if p m b + 2 ac cos M + 1 = 0 m=1 2 M
From this it follows at once that the eigenvalues of B (a b c) are p m m=1 2 M (B.1) m = b + 2 ac cos M +1 The righthand eigenvector of B (a b c) that is associated with the eigenvalue m satis es the equation B (a b c)~ m = m~ m x x (B.2) and is given by j;1 ~ m = (xj )m = a 2 sin j m x m=1 2 M (B.3) c M +1 257
258
APPENDIX B. SOME PROPERTIES OF TRIDIAGONAL MATRICES
These vectors are the columns of the righthand eigenvector matrix, the elements of which are j;1 a 2 sin jm j=1 2 M X = (xjm) = c (B.4) m=1 2 M M +1 Notice that if a = ;1 and c = 1, j;1 a 2 = ei j; 2 (B.5) c
( 1)
The lefthand eigenvector matrix of B (a b c) can be written ; c m 2 1 sin mj m=1 2 ; = 2 X j=1 2 M +1 a M +1 In this case notice that if a = ;1 and c = 1 ; c m 2 1 = e;i m; 2 a
1 ( 1)
M M
(B.6)
B.2 Generalized Eigensystem for Simple Tridiagonals
This system is de ned as follows 2 32 32 x 3 2e f x 3 b c 6a b c 76 x 7 6d e f 76 x 7 6 76 7 6 76 7 6 76 7 6 76 6 76 x 7 = 6 76 x 7 7 a b d e 6 . . . c 7 6 ... 7 6 . . . f 7 6 ... 7 6 76 7 6 76 7 4 54 5 4 54 5 a b xM d e xM In this case one can show after some algebra that det B (a ; d b ; e c ; f ] = 0 if q m b ; m e + 2 (a ; md)(c ; m f ) cos M + 1 = 0 m=1 2 If we de ne m m= M + 1 pm = cos m
1 2 3 1 2 3
(B.7)
M (B.8)
B.3. THE INVERSE OF A SIMPLE TRIDIAGONAL
eb ; 2(cd + af )pm + 2pm (ec ; fb)(ea ; bd) + (cd ; af )pm ] m= e ; 4fdpm The righthand eigenvectors are #j ; 1 " a ; m d 2 sin j ] m =1 2 M ~m = x m j =1 2 M c; f
2 2 2
259
2
q
m
These relations are useful in studying relaxation methods.
B.3 The Inverse of a Simple Tridiagonal
The inverse of B (a b c) can also be written in analytic form. Let DM represent the determinant of B (M : a b c)
DM det B (M : a b c)] De ning D to be 1, it is simple to derive the rst few determinants, thus D = 1 D = b D = b ; ac D = b ; 2abc (B.9) One can also nd the recursion relation DM = bDM ; ; acDM ; (B.10) Eq. B.10 is a linear O E the solution of which was discussed in Section 4.2. Its characteristic polynomial P (E ) is P (E ; bE + ac) and the two roots to P ( ) = 0 result in the solution 8" 9 p p " # < b + b ; 4ac #M 1 b ; b ; 4ac M = DM = p ; 2 2 b ; 4ac : M =0 1 2 (B.11) where we have made use of the initial conditions D = 1 and D = b. In the limiting case when b ; 4ac = 0, one can show that
0 0 1 2 3 2 3 1 2 2 2 +1 2 +1 2 2 0 1
b DM = (M + 1) 2
!M
b = 4ac
2
260
APPENDIX B. SOME PROPERTIES OF TRIDIAGONAL MATRICES
2 16 6 4 D4 6
Then for M = 4
B; =
1
D ;cD cD ;c D ;aD D D ;cD D c D a D ;aD D D D ;cD ;a D a D ;aD D
3 2 2 1 3 2 2 1 2 1 1 2 1 3 1 0 2 1 1 2 1 1 2 3
0
2
3 7 7 7 5
and for M = 5
B ;1 =
aD The general element dmn is
4 0
2 D4 6 ;aD3 1 6 a2 D 6 2 6 D5 6 ;a3 D1 4
cD ;c D c D D D ;cD D c D D ;c D ;aD D D D ;cD D c D a D D ;aD D D D ;cD ;a D aD ;aD D
3 2 2 3 1 3 1 2 2 1 4 1 1 3 0 2 1 2 2 2 2 1 2 1 3 1 1 2 2 1 3 1 2 3 4
;cD
3 7 17 7 2 7 7 5 3
Upper triangle:
m=1 2
M ;1
1
n=m+1 m+2
M
dmn = Dm; DM ;n(;c)n;m =DM
Diagonal:
n=m=1 2 M dmm = DM ; DM ;m =DM
1
Lower triangle:
m =n+1 n+2
M
1
n=1 2
M ;1
dmn = DM ;mDn; (;a)m;n=DM
B.4 Eigensystems of Circulant Matrices
B.4.1 Standard Tridiagonals
Consider the circulant (see Section 3.4.4) tridiagonal matrix
Bp(M : a b c )
(B.12)
B.4. EIGENSYSTEMS OF CIRCULANT MATRICES
The eigenvalues are
m
261
= b + (a + c) cos 2 m ; i(a ; c) sin 2 m M M
(2 )
m=0 1 2
The righthand eigenvector that satis es Bp(a b c)~ m = m~ m is x x ~ m = (xj )m = ei j m=M x j=0 1 M ;1 (B.14) p where i ;1, and the righthand eigenvector matrix has the form 2 m j =0 1 M ;1 X = (xjm) = eij M m =0 1 M ;1 The lefthand eigenvector matrix with elements x0 is 2 j m =0 1 M ;1 ; = (x0 ) = 1 e;im M X mj j =0 1 M ;1 M 1 Note that both X and X ; are symmetric and that X ; = M X , where X is the conjugate transpose of X .
1 1 1
M ;1 (B.13)
B.4.2 General Circulant Systems
Notice the remarkable fact that the elements of the eigenvector matrices X and X ; for the tridiagonal circulant matrix given by eq. B.12 do not depend on the elements a b c in the matrix. In fact, all circulant matrices of order M have the same set of linearly independent eigenvectors, even if they are completely dense. An example of a dense circulant matrix of order M = 4 is 2 b b b b 3 6 b b b b 7 6 7 6 7 (B.15) 4 b b b b 5 b b b b The eigenvectors are always given by eq. B.14, and further examination shows that the elements in these eigenvectors correspond to the elements in a complex harmonic analysis or complex discrete Fourier series. Although the eigenvectors of a circulant matrix are independent of its elements, the eigenvalues are not. For the element indexing shown in eq. B.15 they have the general form M; X bj ei jm=M = m
1 0 3 2 1 1 0 3 2 2 1 0 3 3 2 1 0 1 (2 )
of which eq. B.13 is a special case.
j =0
262
B.5 Special Cases Found From Symmetries
APPENDIX B. SOME PROPERTIES OF TRIDIAGONAL MATRICES
Consider a mesh with an even number of interior points such as that shown in Fig. B.1. One can seek from the tridiagonal matrix B (2M : a b a ) the eigenvector subset that has even symmetry when spanning the interval 0 x . For example, we seek the set of eigenvectors ~ m for which x
2 b 6a 6 6 6 6 6 6 6 4
a b a a ...
... a a b a b
32x 3 1 7 6 x2 7 76 7 76 . 7 76 . 7 76 . 7 76 . 7 76 . 7 6 . 7 54 5 a 7 6 x2 7
=
x
2 3 x1 6x 7 6 27 6 . 7 6 . 7 6 . 7 m6 . 7 6 . 7 6 . 7 6 7 4 x2 5
1
x
1
This leads to the subsystem of order M which has the form
b a 7 7 a b a 7 7 ... 7~ a ~ a)~ m = 7 xm = B (M : a b x 7 ... a 7 7 a b a 7 5 a b+a
2 6 6 6 6 6 6 6 6 6 6 4
3
x m~ m
(B.16)
By folding the known eigenvectors of B (2M : a b a) about the center, one can show from previous results that the eigenvalues of eq. B.16 are (2m ; 1) m = b + 2a cos 2M + 1
!
m=1 2
M
(B.17)
B.6. SPECIAL CASES INVOLVING BOUNDARY CONDITIONS
and the corresponding eigenvectors are ~ m = sin j (2m ; 1) x 2M + 1 j =1 2 M Imposing symmetry about the same interval but for a mesh with an odd number of points, see Fig. B.1, leads to the matrix 2 3 b a 6a b a 7 6 7 6 7 ... 6 7 a a B (M : ~ b a) = 6 ... a 7 6 7 6 7 6 4 5 a b a7 2a b By folding the known eigenvalues of B (2M ; 1 : a b a) about the center, one can show from previous results that the eigenvalues of eq. B.17 are (2m ; 1) m = b + 2a cos 2M and the corresponding eigenvectors are
!
263
Line of Symmetry
x =0
x=
j0 = 1 2 3 4 5 6 M0 j= 1 2 3 M a. An evennumbered mesh Line of Symmetry x =0 x= j0 = 1 2 3 4 5 M0 j= 1 2 3 M b. An odd{numbered mesh Figure B.1 { Symmetrical folds for special cases m=1 2 M
~ m = sin j (2m ; 1) x 2M
!
j=1 2
M
B.6 Special Cases Involving Boundary Conditions
We consider two special cases for the matrix operator representing the 3point central di erence approximation for the second derivative @ =@x at all points away from the boundaries, combined with special conditions imposed at the boundaries.
2 2
264
APPENDIX B. SOME PROPERTIES OF TRIDIAGONAL MATRICES
Note: In both cases
m =1 2 M j =1 2 M ;2 + 2 cos( ) = ;4 sin ( =2)
2
When the boundary conditions are Dirichlet on both sides,
2 6 6 6 6 6 6 4
;2
1 1 ;2 1 1 ;2 1 1 ;2 1 1 ;2
3 7 7 7 7 7 7 5
m ~m x
m = ;2 h+ 2 cos Mi+ 1 m = sin j M + 1
(B.18)
When one boundary condition is Dirichlet and the other is Neumann (and a diagonal preconditioner is applied to scale the last equation),
2 6 6 6 6 6 6 4
;2
1 1 ;2 1 1 ;2 1 1 ;2 1 1 ;1
3 7 7 7 7 7 7 5
~m x
m
= ;2 + 2 cos (2m ; 1) 2M + 1 (2m ; 1) = sin j 2M + 1
(B.19)
Appendix C THE HOMOGENEOUS PROPERTY OF THE EULER EQUATIONS
The Euler equations have a special property that is sometimes useful in constructing numerical methods. In order to examine this property, let us rst inspect Euler's theorem on homogeneous functions. Consider rst the scalar case. If F (u v) satis es the identity
F ( u v) = nF (u v) (C.1) for a xed n, F is called homogeneous of degree n. Di erentiating both sides with
respect to and setting = 1 (since the identity holds for all ), we nd
Consider next the theorem as it applies to systems of equations. If the vector F (Q) satis es the identity
u @F + v @F = nF (u v) @u @v
(C.2)
F ( Q) = nF (Q)
for a xed n, F is said to be homogeneous of degree n and we nd
"
(C.3)
@F Q = nF (Q) @q
265
#
(C.4)
266APPENDIX C. THE HOMOGENEOUS PROPERTY OF THE EULER EQUATIONS Now it is easy to show, by direct use of eq. C.3, that both E and F in eqs. 2.11 and 2.12 are homogeneous of degree 1, and their Jacobians, A and B , are homogeneous of degree 0 (actually the latter is a direct consequence of the former).1 This being the case, we notice that the expansion of the ux vector in the vicinity of tn which, according to eq. 6.105 can be written in general as,
E = En + An(Q ; Qn) + O(h2) F = Fn + Bn(Q ; Qn) + O(h2)
can be written
(C.5)
E = An Q + O(h2) F = BnQ + O(h2)
(C.6)
since the terms En ; AnQn and Fn ; BnQn are identically zero for homogeneous vectors of degree 1, see eq. C.4. Notice also that, under this condition, the constant term drops out of eq. 6.106. As a nal remark, we notice from the chain rule that for any vectors F and Q
@F (Q) = @F @Q = A @Q @x @Q @x @x We notice also that for a homogeneous F of degree 1, F = AQ and @F = A @Q + @A Q @x @x @x
Therefore, if F is homogeneous of degree 1,
" #
"
#
(C.7)
(C.8)
@A Q = 0 @x in spite of the fact that individually @A=@x] and Q are not equal to zero.
1
"
#
(C.9)
Note that this depends on the form of the equation of state. The Euler equations are homogeneous if the equation of state can be written in the form p = f ( ), where is the internal energy per unit mass.
Bibliography
1] D. Anderson, J. Tannehill, and R. Pletcher. Computational Fluid Mechanics and Heat Transfer. McGrawHill Book Company, New York, 1984. 2] C. Hirsch. Numerical Computation of Internal and External Flows, volume 1,2. John Wiley & Sons, New York, 1988.
267
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.