P. 1
134089628 Computational Fluid Dynamics

134089628 Computational Fluid Dynamics

|Views: 65|Likes:
Published by bari116
134089628 Computational Fluid Dynamics
134089628 Computational Fluid Dynamics

More info:

Published by: bari116 on May 26, 2013
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

11/15/2014

pdf

text

original

Fundamentals of Computational Fluid Dynamics

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 Navier-Stokes 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 Finite-Di erence Notation 3.2 Space Derivative Approximations . . . 3.3 Finite-Di 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 FINITE-DIFFERENCE 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 Higher-Order 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 SEMI-DISCRETE 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 Semi-Discrete Linear Forms . . . . . . . . . . 4.2.2 Single ODE's of First- and Second-Order . . . . . . . . . . . . 4.2.3 Coupled First-Order 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 One-Dimensional Examples . . . . . . . . . . . . . . . . . . . . . . 5.3.1 A Second-Order Approximation to the Convection Equation 5.3.2 A Fourth-Order Approximation to the Convection Equation 5.3.3 A Second-Order Approximation to the Di usion Equation . 5.4 A Two-Dimensional 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 FINITE-VOLUME 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 TIME-MARCHING METHODS FOR ODE'S

6.1 Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Converting Time-Marching Methods to O E's . . . . . . . . . . . . 6.3 Solution of Linear O E's With Constant Coe cients . . . . . . . . 6.3.1 First- and Second-Order Di erence Equations . . . . . . . . 6.3.2 Special Cases of Coupled First-Order 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 Time-Marching 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 One-Root Time-Marching Methods . . . . . . . . . . . . . . 6.6 Accuracy Measures of Time-Marching 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 Two-Step Linear Multistep Methods . . . . . . . . . . . . . 6.8 Predictor-Corrector Methods . . . . . . . . . . . . . . . . . . . . . . 6.9 Runge-Kutta 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

. . .1 Relation to -Eigenvalues . . . . . . . . . . . . . . . . . . . .2 Driving and Parasitic Eigenvalues .2 Relation of Sti ness to Space Mesh Size . . . .9 7. . .3 An Example Involving Periodic Convection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Numerical Stability of O E 's . . . . . 7. .1 The Criterion . . . . . . . . . . . . . .1. . . . . 8. . . .1 Stability for Large h.2 Some Examples . . . . . .1. . 8. . . . . . . . . . . . . 8. . . . . . . . . . . . . .4. . . . .6. . . 8.1 Imposed Constraints . . . . . . . . .3 A Perspective . .4. . . . . . . . . . . . . . . . . . . . . . . . 7. . . . . .5. . . . . . . . . . . . . . . . . . .3. Problems . . . . . . . . 8. A-Stable Methods . . . . . . . . . . . . . . . . . . . . . . 7. . . . . . . .4 Comparing the E ciency of Explicit Methods . . . . Numerical Stability Concepts in the Complex h Plane 7. . . . .3 Relation to Circulant Matrices . . . . . . . . .5. . . . . . . . . . . . . . . . . . Fourier Stability Analysis . .6. 149 149 149 151 151 152 153 154 154 154 155 158 158 159 160 160 161 . . . . . . Time-Space Stability and Convergence of O E's .6 Steady Problems .5 7. . . . . . . . . .7. . . . . .3 Sti ness Classi cations . . 7. . . . . . . . . 7.4 7. . . . . Consistency . . . . .4. . . . . .2 Stability for Small t . . . . . . . . . . . 8. 7. . . . . Numerical Stability Concepts in the Complex -Plane . . . . . . . . . . .1 The Basic Procedure . . 7. . . . . . . .7. . . . . .6 7. . . . . . . . .2 Implicit Methods . . .7 7. . . . .8 7. . . . . . . . . . . . . . . . . . . . . . . . 8. . . . . . . . . . . . . . . . . .1 Explicit Methods . . . . . . . . . . . . . . . . . . . . . . . . . .2 Complete Eigensystems . .3 Practical Considerations for Comparing Methods . . . 8. . . 7. . . . . . . . . .2 Unconditional Stability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7. . 8. . . . . . . . . . . . . . . . . . . . . . .5. . . . . . .3 7. . . . . . . . . . . . . 8. . . . . . . . 7. . . . . . . . . . . . .7. . . . . . . .3 Defective Eigensystems . . . . .5. . . . . . . . . . . . . . . . . . . 8. . . . . . . . . . . . . . . . . . . . . . .7 Problems . . . . . .1 Sti ness De nition for ODE's . . . . . . . . . .2 An Example Involving Di usion . . . .2. . . . . . . . . . . . . . . .6. . . . . . . . . . . . . 8. . . . . . . . . . . 8. . . . . . . 123 124 124 125 125 125 128 128 132 135 135 136 137 141 141 142 143 143 146 8 CHOICE OF TIME-MARCHING METHODS 8. . . . . . . . . . . . . .5. . . . . . . .1. . . 8. .5 Coping With Sti ness .3 Defective Eigensystems . . . . . . . . . . . . .3. . . . . . .3. 8. .3 Stability Contours in the Complex h Plane. . . . . . .1 -Root Traces Relative to the Unit Circle . 7. . . . .

.1 The Ordinary Di erential Equation Formulation . . The Lax-Wendro Method . . . . . . . .2. . . . . . . . . . . . . . . . . . .1. . . . . . . . . . . . . . 9. . . .1 Motivation . . . . . . . . . . . . 9. . 192 10. . . .2 Factoring Physical Representations | Time Splitting . . . . . . . . . . . . . . . . .9 RELAXATION METHODS 9. .3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 The Basic Process . . . . .6 Problems . . . . . . .2.3 The Classical Methods . . . . .5 Nonstationary Processes . . .3 The ODE Approach to Classical Relaxation . . . . . . . . . . . . 9. . . . . . . . . . . . . . .2 Properties of the Iterative Method . . . . .2 The Converged Solution. . .1. . . . . . . . . . . . .2.4 Eigensystems of the Classical Methods . . . . . . . . . . . . . . . . . . . . 220 217 . . . . . . . . . . . .3 The SOR System . . . .3 A Two-Grid Process . . . . . . . . . . . . . . . . . . . . . .2 Flux-Di erence Splitting . . .4. . . . . . . . 9. .2 11. . . . and the Error 9. . . . . .1 The Concept . . 9. . . . . . . .3 Factoring Space Matrix Operators in 2{D . . . . . . . . . . . . .6 Problems . . . 11. . . . . . .3. . . . . . . 217 12. . . . . . . . . . . . .4 One-Sided First-Derivative Space Di erencing The Modi ed Partial Di erential Equation . . . . . . . . . . . . . . . . . . . . . . . 191 10. . . . . . .2 Classical Relaxation . . . . . . . . . . . . . Upwind Schemes . . . . . . . . . . . . . . 9. . . . . . . . . . . . . . . . . . . . . . . . . . . . 9. . . . . . .2 ODE Form of the Classical Methods . . . . . . .2 The Model Equations .4. . . . . . . . . 9. the Residual. . . . . . . . . . 200 10. . . . . . . . . 202 11. . . . . . . . . .1 Formulation of the Model Problem . . . . 191 11 NUMERICAL DISSIPATION 203 204 205 207 209 210 212 213 214 12 SPLIT AND FACTORED FORMS 12. . .4. .1 11. . .1 Eigenvector and Eigenvalue Identi cation with Space Frequencies191 10.4 Problems . 9. . . . . . . . . . . . . . . . . . . . . 11. . . . .4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 The Delta Form of an Iterative Scheme . . . . . . . . . . . . . . . . . .1 Preconditioning the Basic Matrix . . . 9.1. . . . . . . . . . . .1. . . . 11. . . . . . . . . . . . . . . . . . . . 218 12. . . 192 10. . . . 9. . . .2 The Gauss-Seidel System . . . . . . . . .5 Arti cial Dissipation . . . . . . . .4. . . . . . . . . . . . . 9.1 Flux-Vector Splitting . . . . . . . . . 9. 163 164 164 166 168 168 168 169 170 170 172 173 174 176 180 182 187 10 MULTIGRID 10. . . . . . . . . .3 11. . . . . .1 The Point-Jacobi System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11. . . . . . . . . . . . . . . . . . 9.

. . . . . . 13. . . 13. .4.1 The Unfactored Implicit Euler Method . . The Inverse of a Simple Tridiagonal . . . . . . . . .2 The Factored Nondelta Form of the Implicit Euler Method 13. . . . . . . . 13. . . . Generalized Eigensystem for Simple Tridiagonals . . . . . . . . .1 Mesh Indexing Convention . . Second-Order Factored Implicit Methods . . .5 12. . . . . . 12. . . . . . . . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . . . .4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 The Representative Equation for Space-Split Operators . . . . . . . . . . . . . Problems . . . . . . . . . . . . . . . . . . . . . . 13. . . Vector and Matrix Norms . . . . . . . . . . . . . . . . . . . . . Algebra . . 12. . . . . . 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 .6 12. .12. . .2 General Circulant Systems . . 13. . Eigensystems . . . . . . . . . . . . . . . B. . . . . . . . .4 A. . . . .1 Stability Comparisons of Time-Split Methods .2 Data Bases and Space Vectors . . . . . . . . . .4 12. . . . . . . 257 . 13. . .6 Special Cases Involving Boundary Conditions . . . . . . . . .2 Analysis of a Second-Order Time-Split Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Eigensystems of Circulant Matrices . . . . . . . . . . . .4. . . . . . . . . . . . . . . . . .3 Data Base Permutations . . . . . .4. . . . . . . .3. . . . .2 Example Analysis of Circulant Systems . . . 220 221 221 223 226 226 228 229 13 LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 13. . .4 Example Analysis of 2-D Model Equations . . . .2 B.3 A. . . . . . . . . . . . . . .4 The Factored Delta Form of the Trapezoidal Method . 13. . . . .3 B. . . . .6 Problems . . . . . 12. . . . B. . . . . . . .4. . . . . . . . . . . . . . . . . . . . . . . . . . . .4 Notation .5 Special Cases Found From Symmetries . . . . . . . . . . . . . . . . . . B. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7 12.4 Space Splitting and Factoring . .2. .3 The Factored Delta Form of the Implicit Euler Method .2 A. . . . . . . . . . . . . . . . . .1 B. . . . . . . . .1 A. .1 The Representative Equation for Circulant Operators . . . . . . . . . .2. . .3. . A. . . . . .1 Standard Tridiagonals . . 13. . .5 B. . . . . . De nitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5 Example Analysis of the 3-D Model Equation . . . . . . . . .4. . . . . . . . . . . . . . . . . . B. . . . 13. . . . . 13. . . .3. Importance of Factored Forms in 2 and 3 Dimensions The Delta Form . . . . . . . . . . .

C THE HOMOGENEOUS PROPERTY OF THE EULER EQUATIONS265 .

the use of numerical methods to solve partial di erential equations introduces an approximation that. di usion. often have no analytic solution.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. can change the form of the basic partial di erential equations themselves. This. In the eld of aerodynamics. The principal such constraint was the demand for uni cation. but the authors believe the answer is a rmative and present this book as justi cation for that belief. These events are related to the action and interaction of phenomena such as dissipation. As we shall see in a later chapter. convection. The new equations. motivates the numerical solution of the associated partial di erential equations. in e ect. boundary layers. 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. As time went on and these attempts began to crystallize. Experience has shown that such is not the case. all of these phenomena are governed by the compressible Navier-Stokes equations. underlying constraints on the nature of the material began to form. Many of the most important aspects of these relations are nonlinear and. 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. and turbulence.Chapter 1 INTRODUCTION 1. At the same time it would seem to invalidate the use of linear algebra for the classi cation of the numerical methods. of course. slip surfaces. shock waves. The mathematical structure is the theory of linear algebra and the attendant eigenanalysis of linear systems. as a consequence. which 1 .

if used and interpreted properly. INTRODUCTION are the ones actually being solved by the numerical process.2 Background The eld of computational uid dynamics has a broad range of applicability. these di erences are usually referred to as truncation errors. as a consequence. There is nothing wrong. and algorithm development in general. The geometry may result from . It is safe to say. producing answers that represent little. However. as long as the error remains in some engineering sense \small". Thus methods are said to have a lot of \arti cial viscosity" or said to be highly dispersive. Mathematically. factorization. with identifying an error with a physical process. even if the errors are kept small enough that they can be neglected (for engineering purposes). the following sequence of steps generally must be followed in order to obtain a satisfactory solution. 1. they can lead to serious di culties. and the requirements of the simulation. the theory associated with the numerical analysis of uid mechanics was developed predominantly by scientists deeply interested in the physics of uid ow and. ow conditions. Since they are not precisely the same as the original equations. these errors are often identi ed with a particular physical phenomenon on which they have a strong e ect. and probably will. the resulting simulation can still be of little practical use if ine cient or inappropriate algorithms are used. including the geometry.2. are often referred to as the modi ed partial di erential equations. 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. they can. Independent of the speci c application under study. However. of course. This motivates studying the concepts of stability. Regardless of what the numerical errors are called. physical reality. convergence. and consistency.1 Problem Speci cation and Geometry Preparation The rst step involves the speci cation of the problem. these methods give very useful information. 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. This motivates studying the concepts of sti ness. 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. if any. On the other hand. All of these concepts we hope to clarify in this book. if their e ects are not thoroughly understood and controlled.2 CHAPTER 1. 1. for example. nor with deliberately directing an error to a speci c physical process.

or elements. symmetry boundaries. in a design context. it is always prudent to consider solving simpli ed forms of the Navier-Stokes equations when the simpli cations retain the physics which are essential to the goals of the simulation. Once the problem has been speci ed. and energy. Alternatively. must be selected. Boundary types which may be encountered include solid walls. a set of objectives and constraints must be speci ed. ii) the drag as well as the lift and pitching moment. the Euler equations require ow tangency to be enforced. The partial di erential equations resulting from these conservation laws are referred to as the Navier-Stokes equations. in the interest of e ciency. We concern ourselves here only with numerical methods requiring such a tessellation of the domain. momentum.1.2 Selection of Governing Equations and Boundary Conditions 1. while the Navier-Stokes equations require the no-slip condition. in ow and out ow boundaries.ow equations. and the thin-layer Navier-Stokes equations. For example. Possible simpli ed governing equations include the potential. 1. .3 Selection of Gridding Strategy and Numerical Method Next a numerical method and a strategy for dividing the ow domain into cells. BACKGROUND 3 measurements of an existing con guration or may be associated with a design study. etc.2. or iii) the details of the ow at some speci c location. which is known as a grid.2. However. or mesh. Instead. the turnaround time required. The boundary conditions which must be speci ed depend upon the governing equations. If necessary. Flow conditions might include. periodic boundaries. These may be steady or unsteady and compressible or incompressible.2. one may be interested in i) the lift and pitching moment only. the Reynolds number and Mach number for the ow over an airfoil. 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. an appropriate set of governing equations and boundary conditions must be selected. physical models must be chosen for processes which cannot be simulated within the speci ed constraints. at a solid wall. and the solution parameters of interest. for example. the Euler equations. The requirements of the simulation include issues such as the level of accuracy needed. no geometry need be supplied. It is generally accepted that the phenomena of importance to the eld of continuum uid dynamics are governed by the conservation of mass. The rst two of these requirements are often in con ict and compromise is necessary. 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. As an example of solution parameters of interest in computing the ow eld about an airfoil.

This step can require post-processing of the data. to develop. 1. INTRODUCTION Many di erent gridding strategies exist. or spectral methods.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 Navier-Stokes equations. analyzing. we present a foundation for developing. It is critical that the magnitude of both numerical and physical-model errors be well understood. and overlapping grids. Although the model equations are extremely simple and easy to solve.4 CHAPTER 1. including structured. These model equations isolate certain aspects of the physics contained in the complete set of equations. the results of the simulation must be assessed and interpreted. we can make use of much simpler expressions. the so-called \model" equations. Some of them are presented in the books by Anderson. and can be aided by sophisticated ow visualization tools and error estimation techniques. unstructured. nite-element. they have been carefully selected to be representative. and understand most numerical methods used to nd solutions for the complete compressible Navier-Stokes equations. 1. Fortunately. The numerical methods generally used in CFD can be classi ed as nite-di erence. hybrid. We believe that a thorough understanding of what happens when .4 Assessment and Interpretation of Results Finally. which are still evolving. Furthermore. of di culties and complexities that arise in realistic two. Tannehill. when used intelligently. the grid can be altered based on the solution in an approach known as solution-adaptive gridding. Here again. Many of these issues are not addressed in this book. 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. and understanding such methods. for example calculation of forces and moments. the use of nite-di erence methods is typically restricted to structured grids. nite-volume. analyze. The choices of a numerical method and a gridding strategy are strongly interdependent. with emphasis on nite-di erence and nite-volume methods for the Euler and Navier-Stokes equations. composite.and three-dimensional uid ow simulations. For example. and Pletcher 1] and Hirsch 2]. Instead we focus on numerical methods. Rather than presenting the details of the most advanced methods. the success of a simulation can depend on appropriate choices for the problem or class of problems of interest.2.

Bold type is reserved for real physical vectors.S.e.4 Notation The notation is generally explained as it is introduced. As a word of caution. 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.S. however. 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. proportional to) .1.. such as velocity. usually a vector A term of order (i. the variable u can denote a scalar Cartesian velocity component in the Euler and Navier-Stokes equations. For example. and a vector consisting of a collection of scalars in our presentation of hyperbolic systems. Otherwise. it should be noted that.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 Navier-Stokes equations. there are many aspects of practical problems which can only be understood in the context of the complete physical systems. k-D O( ) ~ bc Partial di erential equation Ordinary di erential equation Ordinary di erence equation Right-hand side Particular solution of an ODE or system of ODE's Fixed (time-invariant) steady-state solution k-dimensional space Boundary conditions. 1. 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. S. Some of the abbreviations used throughout the text are listed and de ned below. a scalar quantity in the linear convection and di usion equations. PDE ODE O E RHS P.

6 CHAPTER 1. INTRODUCTION .

The main focus. such as the Euler and Navier-Stokes equations and our model equations.1 Conservation Laws Conservation laws. and convergence. mass..g. They are linear partial di erential equations (PDE's) with coe cients that are constant in both space and time. The model equations we study have two properties in common. accuracy. will be on representative model equations. These equations contain many of the salient mathematical and physical features of the full Navier-Stokes equations.Chapter 2 CONSERVATION LAWS AND THE MODEL EQUATIONS We start out by casting our equations in the most general form. 2. The concepts of convection and di usion are prevalent in our development of numerical methods for computational uid dynamics. the integral conservation-law form. The Euler and Navier-Stokes equations are brie y discussed in this Chapter. stability. momentum. and the recurring use of these model equations allows us to develop a consistent framework of analysis for consistency. in particular. The equation is a statement of 7 .1) In this equation. Q is a vector containing the set of variables which are conserved. per unit volume. The equations are then recast into divergence form. the convection and di usion equations. and energy. which is useful in understanding the concepts involved in nite-volume schemes. though. Z V (t1 ) QdV + Z t2 I t1 S (t) n:FdSdt = Z t2 Z t1 V (t) PdV dt (2. e. can be written in the following integral form: Z V (t2 ) QdV . which is natural for nite-di erence schemes. and they represent phenomena of importance to the analysis of certain aspects of uid dynamic problems.

containing the ux of Q per unit area per unit time. 7 6 6 5 4 4 @u 3 @x 4 3 u @u @x + @T @x 7 7 7 7 7 5 (2. F is a set of vectors.2 and nd a solution for Q on that basis are referred to as nite-volume methods. On the other hand.3 and nd a solution for Q on that basis are referred to as nite-di erence methods.2 The Navier-Stokes and Euler Equations The Navier-Stokes equations form a coupled system of nonlinear PDE's describing the conservation of mass.3) @t where r: is the well-known divergence operator given. a partial derivative form of a conservation law can also be derived. is an area A(t) bounded by a closed contour C (t). Those methods which make various approximations of the derivatives in Eq.1 and 2.8 CHAPTER 2. or tensor. For a Newtonian uid in one dimension. 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 . they can be written as @Q + @E = 0 (2. 2. and z coordinate directions. In two dimensions. leading to @Q + r:F = P (2. 2. The divergence form of Eq. 2.6) .2) dt V (t) S (t) V (t) Those methods which make various numerical approximations of the integrals in Eqs. the region of space. 2.1 can be rewritten as d Z QdV + I n:FdS = Z PdV (2. and k are unit vectors in the x y. and P is the rate of production of Q per unit volume per unit time. If all variables are continuous in time. by ! @ @ @ (2. 2. respectively.2 is obtained by applying Gauss's theorem to the ux integral. momentum and energy for a uid.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 7 6 2 u +p 7 . The vector n is a unit vector normal to the surface pointing outward. t1 . then Eq.4) r: i @x + j @y + k @z : and i j. in Cartesian coordinates. or cell. Many of the advanced codes written for CFD applications are based on the nite-volume concept.

In two-dimensional Cartesian coordinates. Note that the convective uxes lead to rst derivatives in space. such as u and p. In order . p is the pressure. Thus the conservative form is appropriate for solving ows with features such as shock waves. Details can be found in Anderson. The steady solution to the one-dimensional Navier-Stokes equations must satisfy @E = 0 (2. This form of the equations is called conservation-law or conservative form. such as the ideal gas law. while the viscous and heat conduction terms involve second derivatives. Although non-conservative forms of the equations are analytically the same as the above form. the Euler equations are obtained. For such ows. T is the temperature.8) @t @x @y with 2 3 2 3 2 2 q1 u 3 v 3 6q 7 6 u7 6 6 u2 + p 7 uv 7 6 7 6 7 Q=6 6 27 7=6 6 7 7 E = 6 7 F = 6 2 7 (2. these can be written as @Q + @E + @F = 0 (2.2.9) 4 q3 5 4 v 5 4 4 v +p 5 uv 5 q4 e u(e + p) v ( e + p) where u and v are the Cartesian velocity components. . for example. 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=2.2. Later on we will make use of the following form of the Euler equations as well: @Q + A @Q + B @Q = 0 (2. Non-conservative forms can be obtained by expanding derivatives of products using the product rule or by introducing di erent dependent variables. or at least may be treated as such. is the coe cient of viscosity. Many ows of engineering interest are steady (time-invariant).7) @x If we neglect viscosity and heat conduction. THE NAVIER-STOKES AND EULER EQUATIONS 9 where is the uid density. u is the velocity. Tannehill. and Pletcher 1] and Hirsch 2]. with no interest in the transient portion of the solution. The ux vectors The matrices A = @Q @Q given above are written in terms of the primitive variables. they can lead to quite di erent numerical solutions in terms of shock strength and shock speed. These equations must be supplemented by relations between and and the uid state as well as an equation of state.10) @t @x @y @E and B = @F are known as the ux Jacobians. u. v. and p. e is the total energy per unit volume. we are often interested in the steady-state solution of the Navier-Stokes equations.

(u2 + v2)=2] from the ideal gas law. q2. 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 . where is the ratio of speci c heats. 2 q4 q 3 q1 2 q3 q1 . 2 q1 7 7 7 7 7 q3 q2 7 7 q1 7 7 7 5 3 2 q4 q2 .11) F 2 6 6 6 6 6 =6 6 6 6 6 6 4 3 2 F1 7 6 6 7 6 7 6 7 F2 7 6 6 7 6 7 = 6 7 6 6 F3 7 7 7 6 6 5 6 4 F4 q3 q3 q2 q1 ( .q q2 1 q1 a41 ! . ) q q2 1 q q3 1 0 (1 .1 0 a42 !2 a43 q2 q1 3 7 7 7 7 7 7 7 7 7 7 7 7 7 5 (2. cp=cv . .1 q2 + q3 q2 2 2 q1 2 q1 q1 q2 (2. .13) where a21 = . 1) e . 1)q4 + 2 q1 .12) We have assumed that the pressure satis es p = ( . q1 . 2q .10 CHAPTER 2. and q4 . + . CONSERVATION LAWS AND THE MODEL EQUATIONS to derive the ux Jacobian matrices.1 q2 7 7 2 q1 7 7 7 5 3 q3 2 q1 (2. 1 3 . as follows: 2 6 6 6 6 6 =6 6 6 6 6 6 4 E1 E2 E3 E4 E 2 3 6 6 7 6 7 6 7 6 7 6 7 6 7 7=6 6 7 6 7 6 7 7 6 5 6 6 4 ( 3 7 7 7 2 2 q2 q3 7 .1 q2 3 2 2 q1 3 7 7 7 7 7 7 7 7 2 . q3 . 1)q4 + 3. ) q q3 1 q2 q1 0 A= a21 q3 . 3 . we must rst write the ux vectors E and F in terms of the conservative variables. 2 q1 2 q2 q1 . 1 q3 2 .

)uv (2. THE NAVIER-STOKES AND EULER EQUATIONS a41 = ( 2 . )u (1 .16) Derivation of the two forms of B = @F=@Q is similar. The homogeneous property of the Euler equations is discussed in Appendix C.5. . 1) . ue a42 = e . . !2 3 5 q4 q1 ! q2 q1 ! 3 + q q1 (2.2. u2 2 2 a43 = (1 .2.( . Thus we also study linear hyperbolic systems of PDE's. aspects of convective phenomena associated with coupled systems of equations such as the Euler equations are important in developing numerical methods and boundary conditions. The eigenvalues of the ux Jacobian matrices are purely real. This motivates the choice of our two scalar model equations associated with the physics of convection and di usion. 1)4 ! 11 q2 q1 !3 3 + q q1 !2 !2 q4 .14) 3 7 7 7 7 7 7 7 7 7 7 7 5 0 1 0 0 0 a21 (3 . 1)u(u2 + v2) . The Navier-Stokes equations include both convective and di usive uxes. .uv a41 a21 = v a42 u a43 (2.15) u where a41 = ( . 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 2 q2 q1 !3 5. which are further discussed in Section 2. This is the de ning feature of hyperbolic systems of PDE's. Furthermore. 1 43 q2 a42 = q1 2 q1 ! ! q q 2 3 a43 = . 1 v2 . 1 (3u2 + v2 ) 2 . 3 . )v ( .

without special consideration of boundaries. This is referred to as the biconvection problem. the ow being simulated is periodic. At any given time. the scalar quantity u is given on one boundary. and the domain is in nite. the waveform travels through one boundary and reappears at the other boundary. It is easy to show by substitution that the exact solution to the linear convection equation is then u(x t) = u0(x . Hence the wave leaves the domain through the out ow boundary without distortion or re ection. 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. It is the simplest to study and serves to illustrate many of the basic properties of numerical methods applied to problems involving convection. Hence. With periodic boundary conditions.1 Di erential Form CHAPTER 2.3. as the convection problem.12 2. Now let us consider a situation in which the initial condition is given by u(x 0) = u0(x). This type of phenomenon is referred to. No boundary condition is speci ed at the opposite side. a real constant which may be positive or negative. the process continues forever without any . simply. while the right-hand boundary is the in ow boundary when a is negative. (2) In the other type. It represents most of the \usual" situations encountered in convecting systems.3 The Linear Convection Equation 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. In this case. 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 \out ow" boundary. Note that the left-hand boundary is the in ow boundary when a is positive. This is consistent in terms of the well-posedness of a 1st-order PDE. eventually returning to its initial position. corresponding to a wave entering the domain through this \in ow" boundary. what enters on one side of the domain must be the same as that which is leaving on the other. at) (2.17) @t @x Here u(x t) is a scalar quantity propagating with speed a. we pay a great deal of attention to it in the initial chapters.

must be an integer. the solution can be written as u(x t) = f (t)ei x (2. With such an initial condition. An arbitrary initial waveform can be produced by summing initial conditions of the form of Eq. 2. In order to satisfy the periodic boundary conditions. a.17. THE LINEAR CONVECTION EQUATION 13 change in the shape of the solution. 2.20) where the time dependence is contained in the complex function f (t). the wavenumber.2 Solution in Wave Space u(x 0) = M X m=1 fm (0)ei mx (2. are related by != a (2. we nd that f (t) satis es the following ordinary di erential equation (ODE) df = . The linear relation given by Eq. in a simulation. We now examine the biconvection problem in more detail. It is a measure of the number of wavelengths within the domain. and the phase speed. 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 (2.21) dt which has the solution f (t) = f (0)e. Preserving the shape of the initial condition u0(x) can be a di cult challenge for a numerical method.ia t (2. . most numerical methods introduce some dispersion that is. !.22) Substituting f (t) into Eq. one obtains 2.19. 2. Eq. As we shall see later. This means that the phase speed is the same for all wavenumbers.20 gives the following solution u(x t) = f (0)ei (x. and is the wavenumber.2.!t) (2.19) where f (0) is a complex constant.24) The relation between the frequency and the wavenumber is known as the dispersion relation.ia f (2. Substituting this solution into the linear convection equation.25) .3. waves with di erent wavenumbers travel at di erent speeds.at) = f (0)ei( x.24 is characteristic of wave propagation in a nondispersive medium. 2.23) where the frequency.3. For M modes.

g(x) = 0 @x2 (2.27. For example. 2.30) . this parabolic PDE governs the di usion of heat in one dimension.29) giving a steady state-solution which satis es @ 2 u . u must vary linearly with x at steady state such that the boundary conditions are satis ed. Other steady-state solutions are obtained if a source term g(x) is added to Eq.at) (2. Neumann (speci ed @u=@x). m=1 fm (0)ei m(x. the steady-state solution must satisfy @2u = 0 (2.27) @t @x2 where is a positive real constant. 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.23. the solution is obtained by summing solutions of the form of Eq.26) 2. Dirichlet (speci ed u). In contrast to the linear convection equation. the di usion equation has a nontrivial steady-state solution. as follows: @u = @ 2 u (2. Boundary conditions can be periodic. CONSERVATION LAWS AND THE MODEL EQUATIONS where the wavenumbers are often ordered such that 1 2 M . g(x) # (2.4 The Di usion Equation 2.27. 2. or mixed Dirichlet/Neumann.1 Di erential Form Di usive uxes are associated with molecular motion in a continuum uid. In the case of Eq. A simple linear model equation for a di usive process is Therefore. which is one that satis es the governing PDE with the partial derivative in time equal to zero.4. Since the wave equation is linear. with u representing the temperature.28) @x2 @u = @t " 2 @u @x2 . 2.14 CHAPTER 2.

THE DIFFUSION EQUATION 15 In two dimensions.33) where must be an integer in order to satisfy the boundary conditions. u( ) = ub.37 shows that high wavenumber components (large m) of the solution decay more rapidly than low wavenumber components. h(x) = ua + (ub .32 is elliptic.4. ua)x= .27 gives the following ODE for fm : dfm = .36) Substituting fm(t) into equation 2. 2 f (2.2. i. 2.. 2. Eq.31 is parabolic.37) The steady-state solution (t ! 1) is simply h(x). 2. Eq.34.e. and as Laplace's equation for zero g. g(x y) (2.2 Solution in Wave Space We now consider a series solution to Eq. It is clear that the steady-state solution is given by a linear function which satis es the boundary conditions. .35) m m dt and we nd m t fm(t) = fm(0)e.32) @x2 @y2 While Eq.27. 2. 2 (2. we obtain m=1 u(x t) = M X m=1 m t sin m x + h(x) fm (0)e. 2 (2. g(x y) = 0 (2. A solution of the form M X u(x t) = fm(t) sin mx + h(x) (2. consistent with the physics of di usion. 2. 2.31) @t @x2 @y2 where g(x y) is again a source term. Substituting this form into Eq. The corresponding steady equation is @ 2 u + @ 2 u .4. Let the initial condition be u(x 0) = M X m=1 fm(0) sin m x + h(x) (2. the di usion equation becomes " # @u = @ 2 u + @ 2 u . The latter is known as the Poisson equation for nonzero g.34) satis es the initial and boundary conditions. Let the domain be given by 0 x with boundary conditions u(0) = ua.

and noting that X and X . CONSERVATION LAWS AND THE MODEL EQUATIONS The Euler equations. we obtain =0 (2. The entries in the where f is the ux vector and A = @u ux Jacobian are @fi aij = @u (2.1u z }| { @ X .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. the equations have been decoupled into m scalar equations of the form @wi + @wi = 0 (2.41) where is a diagonal matrix containing the eigenvalues of A. 2. form a hyperbolic system of partial di erential equations.40) j The ux Jacobian for the Euler equations is derived in Section 2.5 Linear Hyperbolic Systems CHAPTER 2.1AX X .1 . such as the Maxwell equations describing the propagation of electromagnetic waves.16 2.2.1u. this equation can also be written in the form @u + @f = 0 (2.38 by X .42) @t + @x With w = X .8. and X is the matrix of right eigenvectors.1 Thus = X . postmultiplying A by the product XX . Premultiplying Eq.43) @t @x When written in this manner. 2. this can be rewritten as @w + @w = 0 (2.39) @t @x @f is the ux Jacobian matrix. Eq. . Many aspects of numerical methods for such systems can be understood by studying a one-dimensional constant-coe cient linear system of the form @u + A @u = 0 (2.1 are constants. are also of hyperbolic type.1AX (2. Such a system is hyperbolic if A is diagonalizable with real eigenvalues. Other systems of equations governing convection and wave propagation phenomena.1.44) i @t @x 1 @X .1u See Appendix A for a brief review of some basic relations and de nitions from linear algebra.

all of the wave speeds have the same sign. leading to three equations in the form of the linear convection equation. Therefore. we see that a hyperbolic system in the form of Eq. The signs of the eigenvalues of the matrix A are also important in determining suitable boundary conditions. The eigenvalues of the ux Jacobian matrix. Show that the 1-D 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 .6. where juj < c. which corresponds to in ow for these variables. The one-dimensional Euler equations can also be diagonalized. 2. c. the boundary conditions can be speci ed accordingly. characteristic variables associated with positive eigenvalues can be speci ed at the left boundary.6 Problems 1. are u u + c. corresponding to the fact that sound waves can travel upstream in a subsonic ow. and u . or qwave speeds. while u + c and u . Each characteristic variable satis es the linear convection equation with the speed given by the corresponding eigenvalue of A. Characteristic variables associated with negative eigenvalues can be speci ed at the right boundary.38 has a solution given by the superposition of waves which can travel in either the positive or negative directions and at varying speeds. The characteristic variables each satisfy the linear convection equation with the wave speed given by the corresponding eigenvalue. and c = p= is the local speed of sound. c are associated with sound waves. wave speeds of both positive and negative sign are present. Therefore. which is the inow boundary for these variables. While other boundary condition treatments are possible. PROBLEMS 17 The elements of w are known as characteristic variables. they must be consistent with this approach. Based on the above. In a subsonic ow. where u is the local uid velocity. although they remain nonlinear. where juj > c. While the scalar linear convection equation is clearly an excellent model equation for hyperbolic systems. 2.2. That is. The speed u is associated with convection of the uid. in a supersonic ow.1 7 M =6 4 0 u 5 0 p u . of course. we must ensure that our numerical methods are appropriate for wave speeds of arbitrary sign and possibly widely varying magnitudes.

33 with fm(0) = 1 for 1 m 3. respectively. i. Next consider a solution with boundary conditions ua = ub = 0. Plot the initial condition on the domain 0 x . j = 0 1 2 3.25. Plot the rst three basis functions used in constructing the exact solution to the di usion equation in Section 2.1. 2. 6. u2=2). 4. @x @y .4.e. p = ( . 2. 2. 7. Show that the two matrices M and A derived in questions 1 and 3. CONSERVATION LAWS AND THE MODEL EQUATIONS Assume an ideal gas. 2. Eq. Find its eigenvalues and compare with those obtained in question 2. (Hint: use M = 2 with 1 = 1 and 2 = .) Next consider the same initial condition de ned only at x = 2 j=4. 9.5). that is. in the form @U + A @U = 0 @t @x where U = @u=@x @u=@t]T . in the form of Eq. 3. 2. 1)(e . Given the initial condition u(x 0) = sin x de ned on 0 x 2 . Find the eigenvalues and eigenvectors of A. Derive the ux Jacobian matrix A = @E=@Q for the 1-D Euler equations resulting from the conservative variable formulation (Eq. nd the necessary values of fm (0).) 5. Find the eigenvalues and eigenvectors of the matrix M derived in question 1. Plot the solution at t = 1 with = 1. 8..2. 2. Write the classical wave equation @ 2 u=@t2 = c2 @ 2 u=@x2 as a rst-order system. write it in the form of Eq. The Cauchy-Riemann 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 + @u = 0 ! = . 1. are related by a similarity transform. Find the values of fm(0) required to reproduce the initial condition at these discrete points using M = 4 with m = m . fm (0) = 0 for m > 3. and initial conditions from Eq.18 CHAPTER 2. (Hint: make use of the matrix S = @Q=@R.31.2. Write the 2-D di usion equation.

For isentropic and homenthalpic ow..v q= u f = g = v v . (b) Determine the eigenvalues of the ux Jacobians.e. 1 = 1. (c) Determine the conditions (in terms of and u) under which the system is hyperbolic.6. (d) Are the above uxes homogeneous? (See Appendix C. the system is closed by the relation 1 u2 + v2 . has real eigenvalues. 2 .1 1 Note that the variables have been nondimensionalized.) . PROBLEMS 19 where ! = 0 for irrotational ow.u One approach to solving these equations is to add a time-dependent 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. . Combining the two PDE's.u . i.2. we have @f (q) + @g(q) = 0 @x @y where .

20 CHAPTER 2. CONSERVATION LAWS AND THE MODEL EQUATIONS .

 ¢¡ .

£ ¤ ¥§¦ ¨©   %  & &  $ ' !#" (    %  $ )'   0 £ 132)46587#79582A@CBEDGFDGFIH#HQPSRUTVDWBX582IYa`85cb8H6dW2IBE2I`eRI2IYfDGHgTVhpiqIRIBXhrqI5s@utv58RIdw795xhIHQy€HQPSRUTVDWBX582pY 4Q5¢2SDGT¢BX2wUT¢d‚DWBƒT¢yxhIH6dWBEbVTVDWBEb¢HQYv@CB„DGF…d‚HQYW†HQ4‡DˆDG5‰†5¢DGF‘Y‚UT¢46HCT¢2Ih…DGBE7#HV’”“2IH•4gT¢2–T¢pIdW5g—˜BX7#TVDGH DGFpHQYWH™YWBX7wRIyEDGT¢2IHQ5¢RIYWy„i9T¢2IhdDWFIHQ2eY‚58yEb¢H™DGFIHfdWH6YWRIy„DGBX2p`whIB„ghH6dWHQ2p4QHH6PxRITVDGBE582IYQ’jifyEDWHQdW2ITVDGB„b8HQy„i8t 582IHw4gT¢2kT¢pIdW5g—˜BX7#TVDGHuDGFpH…YWUTlDGBƒTVymhIH6dWB„b¢TlDGBEb¢HQY™nIdWY‚DQthDGFpHQdWH6‰oiId‚5˜hIRp4QBX2p`eTdY‚ixYfDGHQ7p5¢qˆ58dfr hIBE2UT¢d‚i–hpBEgsHQdWH62oDGBƒTVy†H6PSRUTVDGBE582IY6’jtuFIHfDWBX79HfhIH6dWB„b¢TlDGBEb¢HQYvT¢dWH™T¢IId‚5g—pBE7dTlDGHQh#2IH‡—xDQtpyXHQT¢hIBE2I`wDW5 Tkxzy|{d}‡~3{9¢€Gƒ‚xy|„x…†{d}‡x|‚U‡Qˆ9@CFIBE4ƒFrId‚5˜hIRp4QHQYwT‰Y‚H6Dw5¢qŠhIB„ghH6dWH62I4QH9HQPSRUTVDWBX582pYw@CFIBE4ƒF)4gTV2r‰†H YW5¢yEb8H6h‹’ŠtuFIBEYCBXYuDGFIHaT¢IpdW5oT¢4ƒF†HQ79IFUT¢Y‚BXŒ6HQhFpHQdWHV’Š132DGFIBEYf4ƒFUTVpDGH6dQtUDGFIHw4Q5¢2I4QH6pD5¢qmnU2IB„DGH‡r hIB„ghH6dWHQ2p4QHŽTVIIdW5g—˜BX7#TVDWBX582IYDW5UTVd‚DGBXT¢yjhIH6dWBEbVTVDWBEb¢HQYwBEYuIdWH6YWHQ2oDWHQh‹’†tuFIHQY‚HŽ4gT¢2r‰sH#T¢IIyEBXH6h HQB„DGFIH6dfDW5dYWITVDGBXT¢ymhpHQdWB„bVTVDGB„b8HQYf58dCDGBX79HwhIHQd‚BEbVTVDGB„b8H6YQ’™“Rpd™HQ79IFUTVYWBXYfBX2‰DGFIBEY•FUT¢˜DGHQd™BXY‘582 YWITVDGBXT¢yhhIH6dWB„b¢TlDGBEb¢HQY6’pDGBE7#HfhIHQd‚BEbVTVDGB„b8H6YuT¢dWHDWdWHQTVDGH6heBX2‰•FUT¢˜DGHQdC“˜’€”xDGdWTVDGH6`8BXH6Y•q•58duT¢pIyEixBX2p` DGFpHQYWHunU2IB„DGH–r3hIB„ghH6dWHQ2p4QHuT¢IpdW5g—˜BX7#TVDGBE582IY•@CBEyXyh‰†HuhIBXY‚4QRIY‚YWH6hBX2‰•FUTVpDGH6df—I’ i™yEyv5¢qjDWFIH‘7dTlDGHQd‚BƒT¢y˜‰†HQyE5s@BEYIdWH6YWH62SDWHQhABX2eTŠTVd‚DGH6YWBXT¢2eY‚ixY‚DWHQ7‰’w™H…H67#IFIT¢YWBEŒQH…DGFpH qšT¢4‡D‘DWFUTVDCPSRIB„DGHu`8HQ2pHQdGTVy›4QyƒTVYWYWH6Yf5Vqv79HQY‚FIHQY‘H‡—˜Id‚HQYWY‚HQheBE2‰`¢HQ2IH6dGT¢y›46RId‚bxBXyEBX2IHQT¢dŠ465˜58d‚hIBX2ITVDGH6Y BX2eœU‚xcž‡yŸW¢ vž3œpxƒ}4gTV2r‰†H9DWdGT¢2IYfq•58dW79HQhADG5†T†RI2pBEq•58d‚7¡ŠT¢d‚DWHQY‚BƒT¢27#H6YWF@CB„DGFHQPSRIBXY‚UT¢46HQh BX2oDWHQd‚bVT¢yEY‹BX2uTuY‚5Vr34QT¢yXyEHQh#ƒ‡l{™œ†¢xx£¢xšyŸ‡l„†¢ Vž3œpxƒ}‡tsT¢Y›Y‚FI5c@C2wBE2u¤vBX`8RpdWH”¥˜’E¦8’§tuFIHƒ‡l{™œ†¢xx£¢xšyŸ‡l„†¢  ž3œpxƒ}hBXY˜RI2IB„q•58dW7‰’¢TVyXy¢DWFIHv`¢HQ5879H6DWdWBX4”bVT¢d‚BƒTVDWBX582BEY§T¢‰IYW5¢dW‰†HQhwBE2oDG5Cb¢TVdWBƒTV‰IyXH”465˜H‡¨94QBXH62oDGYv5VqpDGFpH DGdWT¢2IYfq•58dW79HQhAHQPSRUTVDWBX582pYQ’e¤U58dDGFpBXYwdWHgTVYW582‹t§BX2A7…RI4ƒF5¢qŠDWFIH9q•5¢yXyX5c@CBE2I`T¢4Q46RIdGTV46iT¢2UT¢y„ixYWBXY @vHRIY‚HT¢2©HQPSRIBXY‚UT¢46HQhªŠT¢dfDGH6YWBƒTV2«Y‚ixYfDGHQ7 @CBEDWFI58RpD9‰†HQBE2I`rRI2phIRIyEi¬d‚HQYfDGdWBE46DWBEb8H‰58d9yX58Y‚BX2I` IdWT¢46DWBX4gTVy›T¢IIyEBX4QTVDGBE582‹’ ­®c¯ ° ±²”³±²µ´w¶e· ¸º¹g¶e¹6»v±f¼h½¾¹6¿)±Àj±¶eÁ‘± ÂÄÖ»v´a»”¹QÃ9¶ tuFIHY‚BX79IyXH6Y‚DŠ79HQY‚FBX2ob858y„bxBX2I`…‰†5¢DWFŽDGBX79HTV2IheYWUTV4QHuBXYvYWFI5c@C2‰BX2¤vBE`8RIdWH™¥p’   ’ji™2eBE2IYW†HQ4–r DGBE582Å5¢q‘DGFIBEY…nU`8RId‚HsH6dW79BEDGYwRIY‘DG5khIH6nI2IH‰DWFIHeDGH6dW79Y–T¢2Ih¬2I5¢DGTVDGBE582Å2IHQH6hIHQhÆDG5khIH6YW4Qd‚BX‰†H   ¦ .

 8  ǔȑɐʙËm̈ÍÏÎSЫÑvÒÔӑÒQËm̀Õz֐ÒÔєє̈͑̀Ówǔ̫ɐʈʈ͏×jؐÒfنə˧҇×uÓÚ y x ¤vBX`¢RIdWHu¥p’„¦8Û ÜjFoixYWBX4QT¢y˜T¢2Ih‰4Q5¢7#IR˜DƒTVDWBX582UTVyhYWIT¢4QH6YQ’ nU2IB„DGHhpBEgsHQdWH62I4QHuTVIIdW5g—˜BX7#TVDWBX582IY6’”132`8HQ2pHQdGTVyŸt˜DWFIHhIH6sH62IhIHQ2oDCbVT¢dWBXT¢‰IyEHQYQt˜Ý˜t˜q•58dŠH–—pT¢7#pyXH¢t T¢d‚H‘q•RI2p46DGBE582IYv5¢q‹DGFpH™BX2phIHQ†HQ2IhpHQ2oDŠbVT¢dWBXT¢‰IyEHQYjޖtpTV2IhdߘàGá†àƒâp’j¤U58djDWFIHfnUd‚Y‚D•YWH6b¢HQdWT¢yh4ƒFUTVpDGH6dWY @vH‘4Q582IY‚BXhIH6dIdWBE7dTVdWBXy„idDWFIHe¦6rÔãÏ4QT¢YWH‘ÝeäåÝmæ•ß˜àWނ疒u™ªFpHQ2k582Iy„i582pHwbVT¢dWBXT¢‰IyXHwBXY™hIHQ2I5VDGHQhht hIH6sH62IhIHQ2p4QH…582DGFpHw5VDGFIH6dfBEYCT¢YWY‚RI7#H6h‹’•tuFpHa79HQY‚FBX2phIH‡—eq•5¢dCßBXYw¢ é耢cžˆêUtëT¢2Ih‰DGFUTVDuq•5¢d ފBEYw¢ é耢sž•ìv’jtuFpHQ25¢2T¢2H6PxRpBXYWIT¢4QH6h`8d‚BXh ß Þ¡ä ä ßxíuä«êSî…ß Þ3ïaäðì‹î…Þ€äªì›ñ 暥p’E¦cç 暥p’   ç @CFIH6dWH‘î…ßkBXYuDWFIHwY‚UT¢4QBE2I`dBE2†ßAT¢2phî…ÞuDGFIHaYWUTV4QBX2p`dBX2Þ–tsT¢Y™Y‚FI5c@C2eBX2†¤vBX`8RpdWHw¥p’   ’‘òf5¢DWH DGFITVD™ñeäóî‘Þ•DGFId‚58RI`8FI5¢RpDg’vô˜TVDGH6dõT¢2Ih†öjT¢dWHuRIY‚HQhq•5¢d‘áŽT¢2Ih†â#BX2†T–YWBX79BXyXT¢dv@•Tgi8’•™ªFpHQ2 ìvà£ê8à–õsàƒöT¢d‚H†RpYWHQh©q•58d#5VDGFIH6d#IRpdW†58YWH6Yeæ•@CFpBX4ƒFºBEY#Y‚587#H‡DGBE7#H6Y92pHQ4QH6YWYWT¢d‚ipç‡tCyE5˜4QT¢yu4Q582oDWH‡—xD YWFp58RIyXh7#T¢÷¢HDWFIHu7#HQT¢2IBX2p`‘58‰ob˜BE58RIY6’ tuFIHw4Q582ob8H62oDGBX5¢2†q•58duY‚RI‰IYW46dWBEpDfT¢2Ih‰YWRI†HQd‚YW4Qd‚BXpD‘BE2IhIH‡—˜BE2I`–BXYCT¢Yuq•58yEyX5c@CYQÛ Ýmæ•Þ‹ø«õpñ†ç¡ä Ývæ|ߑø¬þî…ßsç¾ä Ýmæ|ߑø¬þî‘ߘà‚Þ˜ø«õpñ†ç¡ä Ýmæ‚ùúì9ø©õxû|ñëçˆäðÝUïgüIý Ýmæ‚ù ê™øÆþŽûXî‘ßsçˆäÿÝxí‚ü¡  ïcüIý¥¤ Ý£í‚ ¢¡ ü   暥p’ ¥8ç ò 5VDGBX46HDGFUTlDu@CFIHQ2‰RIYWH6hT¢yE582IHVtp‰†5¢DGFeDGFpHuDWBX79HT¢2IhY‚UT¢4QHwBX2phIBX46HQYuT¢psHQT¢d‘T¢YCT‘YWRI‰IY‚4Qd‚BXpDQt f ‰IRpD‘@CFpHQ2ÆRIYWH6h¬DW58`8H‡DGFIH6dQtjDGBE7#HŽBEY‘T¢y„@ŠTgixY–TYWRI†HQd‚YW46dWBX˜D9T¢2Ih¬BXY…RpYWRUT¢yEyEiHQ2I46yX58Y‚HQhÅ@CBEDWF UT¢d‚HQ2oDGFpHQYWH6YfDW59hpBXY‚DWBX2I`¢RIBXY‚F‰B„DŠq•dW5¢7 T¢2‰H‡—˜†582IHQ2oDQ’ .

ÎxЧ¦8кنÌvژș̀ډəәÖÑvÒÔÓfÒQËm̈՟֐ÒÔÑvє̈ÍfÌˆÓ…Ç”Ì Ów×aËIə˧҇×uÓ   ¥ ¨t Grid or Node Points n+1 n n-1 ∆x ∆t © ¤vBE`8RId‚H¥p’   Û j-2 j-1 j © j+1 © j+2 x ”˜UT¢46H‡rŸDGBX79H`8d‚BXhT¢d‚dGT¢2I`¢HQ79HQ2oDg’ ãHQd‚BEbVTVDGB„b8H6Y#TVdWHH‡—˜Id‚HQYWY‚HQh TV4Q4Q5¢dWhIBE2I`ADW5eDGFIHRIY‚RUT¢yC4Q582ob¢HQ2oDGBE582IYQ’©tuFSRIY#q•58d‘IT¢d‚DWBƒT¢y hIH6dWBEbVTVDWBEb¢HQYCBX2eY‚UT¢4QHw58dŠDWBX79H™@•HwRIY‚HaBE2SDWHQd‚4ƒFUT¢2I`8HQT¢‰Iy„i  Ýeä  Ý  ß à  Ýeä  Ý à  Þ ¥ Ýeä  Ý  ß  à H6DW4¢’ 暥p’ —oç ¤U58dŠDGFpHw5¢dWhIBE2UT¢d‚i#DGBE7#HhpHQdWB„bVTVDGB„b8HBX2eDWFIHuY‚DGRphpi5¢q”“uã Yu@•HuRIY‚H Ý!˜ä#" Ý Þ æš¥p’%$¢ç " 132DGFpBXYCDGH‡—xDY‚RI‰IY‚4QdWBEpDGY5¢2hIHQ†HQ2phIHQ2oDbVT¢dWBXT¢‰IyEHQYfT¢dWHa2IH6b¢HQdRIY‚HQhDG5dH‡—˜IdWH6YWYuhIH6dWBEbVTVDWBEb¢HQYQ’ tuFSRIY‘Ý  @CBEyXyh2p5¢D‘‰sHRIY‚HQh‰DG5–dWH6IdWH6YWHQ2oDfDGFIHnUd‚Y‚D‘hIHQd‚BEbVTVDWBEb8H5Vqv݉@CBEDGF‰dWH6YW†HQ4‡D‘DW59ߘ’ tuFIH™2p5¢DƒTVDWBX582‘q•58dˆhIB„ghH6dWH62I4QH™T¢IId‚5c—˜BE7dTVDWBX582pY”q•58yXyE5s@CY”DWFIH‘YGT¢79HfpFIBXyE58YW58pFSi…‰IRpDæ•@CBEDWF 582IH€H‡—˜46HQpDWBX582ëç›BED§BEY§2p5¢DvRp2IBXPSRIHV' ’ &viDWFIBXY›@vH•79HgT¢2uDGFUTlDmDWFIHvYfix7…‰s5¢) y (uBEY§RpYWHQhaDG5‘dWHQpdWHQY‚HQ2oD T‘hIBEgsHQd‚HQ2I46HwTVIIdW5g—˜BX7#TVDWBX582dDW59T–hpHQdWB„bVTVDGB„b8HuYWRI4ƒF‰DGFUTlDgtpq•58duH–—pT¢7#pyXH¢t ( 1023 à ( ¥102¥ 暥p’ “8ç ‰IRpDuDWFIHuIdWH64QBEYWHu2UTVDWRIdWH9æšT¢2Ih‰58dWhIH6dƒç•5¢q§DGFpHuT¢IId‚5c—˜BE7dTVDWBX582dBXYu2I5¢DC4QT¢dWd‚BXH6h‰BE2ŽDGFIHuYfi˜7w‰s58y (S’…“™DGFpHQd@•TcixYwT¢d‚H‘RIYWH6hkDW5hIH6DWHQdW79BX2pH…BEDWYIdWH64QBEYWH‘79HgT¢2IBE2I`I’utuFIH‘582pH‘H‡—˜4QH6pDGBE582kBXYDGFpH Y‚ix7…‰†58y˜îetI@CFIBE4ƒF†BEYuhIH6nU2pHQhY‚RI4ƒF†DWFUTVD î…Þ3ïaä Þ3ïcü5476rÞ3ïIà î…ßxíCäªßxí‚ü54'6rßxísà î‘ÝëïaäªÝUïcü5486ÅÝUï˜à H6DW4¢’ 暥p’%9¢ç ™ªFIH62‰DWFIHQd‚HaBEYu2I5–YWRI‰IY‚4Qd‚BXpDf582‰î…ÞŠ58dCî…ߘtUDGFIHuY‚UT¢4QBE2I`–BXYuRI2pBEq•58d‚7†’ .

  — ­®A@ ǔȑɐʙËm̈ÍÏÎSЫÑvÒÔӑÒQËm̀Õz֐ÒÔєє̈͑̀Ówǔ̫ɐʈʈ͏×jؐÒfنə˧҇×uÓÚ BDC´wÁ‘± ½ ±Àˆ¹FEw´w»j¹FE9± GHCICeÀjÃQP†¹SR ´a»j¹6Ã#¶e² iÆhIBEgsHQd‚HQ2I46HŠT¢IpdW5g—˜BX7#TVDGBE582u4gT¢2…‰†H€`8HQ2IH6dGTVDWHQh…5¢dmH‡b¢TVyXRUTVDWHQha‰Siw79HgT¢2pYv5¢qëT‘YWBE7#IyEHjtvTgixyX5¢d YWH6dWBEHQYŽH–—pIT¢2IYWBE582‹’ ¤ë5¢d#H–—ITV7#IyEH¢tu46582IYWBEhIHQd#Ýmæ•ß˜àWނç‘@CBEDGF©Þ‘np—˜HQh‹’ótuFpHQ2‹tuq•58yEyX5c@CBX2I`kDGFpH 2I5¢DGTVDGBE582e4Q5¢2Sb¢HQ2oDGBE582A`8BEb¢HQ2kBXT 2 jPSYQ’˜¥p’„¦…DG5‰¥p’ ¥pt‹ßä êSî…ß)T¢2phAÝmæ|ß#øðõ˜î‘ßsç™ä Ýmæ„êxî…ß#ø õpî…ßsçˆäÿÝxíWüIýsU ’ v—pIT¢2IhIBE2I`9DWFIHuyƒTVD‚DGH6d•DGH6dW7µT¢‰†58RpDuß`8BEb¢HQY ï  Ý  Ý ¦ ¦ ï  Ý  ÝxíWüIýCäðݘívø æzõpî…ßs¡ ç V ø æzõpî…ßsç V ø YaYaYcø ` æŸõpî…ßsç V ø YaYaYm暥pd ` ’ c8ç  X  ß  W í  ß ï W í ß W í ì b 7   ô›5˜4QT¢yohIB„ghH6dWHQ2p4QH€T¢IId‚5g—pBE7dTlDGBX5¢2IY‹DG5‘TC`8BEb¢HQ2wUT¢d‚DWBƒT¢y¢hIHQd‚BEbVTVDGB„b8Hˆ4gTV2a‰†Hjq•58dW79HQhuq•dW587 4Q5¢7…‰IBX2ITVDGBE582IYŠ5¢qmÝxíT¢2IhÝxíWüIýfq•58d‘õ#äfe–¦8à¥e   àSYaYaY„’ ¤U58duH‡—pT¢79IyXHVtI4Q5¢2IYWBEhIHQduDGFpHwtmTgixyX58duY‚HQdWBEHQY‘H‡—˜UT¢2pYWBX5¢2‰q•58duÝxí‚ü54ƒÛ ï  Ý  Ý ¦ ¦ ï  Ý æšî‘ßsç  V æšî…ßsç V Ýxí‚5 ü 4jäªÝxí”ø æšî…ßs¡ ç V ø ø YaYaYsø g ø YaYaY `  ßW í  ß  W í  ß ï W í ì8b   òf5c@ YWRI‰pDWdGT¢4‡D‘ݘí™T¢2Ih‰hIBEbxBXhpH‰Sieî‘߆DG5958‰pDGT¢BX2 yXBX2pHgT¢d 暥p’dh8ç  Ý Ý˜í‚ü5486)Ýxí äiV ø  ßpW í î‘ß   ¦ æšî…ßsç¡V  Ý  ß  W í ø`YaYaY æz¥p’„¦ ¡ ç tuFSRIYvDWFIHŠH‡—˜Id‚HQY‚YWBX5¢2‰æšÝxíWü54q6aÝxíQçsr¢î…ß#BEYvTfdWHgTVYW582UTV‰IyXHvT¢IId‚5g—pBE7dTlDGBX5¢2wq•58d1tvux w y T¢YvyE582I`TVY u í î…ßeBXYuY‚7dTVyXyŸ’ˆò‘H‡—xDf4Q582IY‚BXhIH6d‘DWFIHuYWUTV4QHwhIBEgsHQd‚HQ2I46HwT¢pIdW5g—˜BX7#TVDGBE582AæšÝxí‚5 ü 476ÅÝx¥ í €4fçrpæ   î…ßs燒 v—˜UT¢2IhðDGFIHeDGH6dW79YBX2 DGFIHk2SRI79HQdWTVDG58d‰T¢‰†58RpD9ê«T¢2IhðdWHQ`¢dW58RIªDGFpHAd‚HQY‚RIyEDeDW5Æq•58d‚7 DGFpH q•58yEyX5c@CBX2I`wHQPSRUTVDWBX582  Ý ÝxíWü54‚6)Ýxí¥€4 ä 6ƒV  ß W í   î…ß “ ¦ î‘ß  V „ Ý  ß „ W í ø  †Ý  ¦ î…ß…V  ß † W í8‡x‡x‡ ¦ V   ¡ æz¥p’„¦8¦sç ™ªFIH62H‡—˜Id‚HQY‚YWHQheBE2eDGFIBEY‘7#T¢2I2pHQdQtUB„DCBXYC4QyEHgT¢dŠDWFUTVDCDGFpHahIBEYW46dWH6DWHDGHQd‚7#YC582DGFpHayEH6q|DCYWBEhIHu5¢q DGFpHfH6PxRITVDGBE582#d‚HQId‚HQY‚HQ2oDuTnUdWYfD€hIHQd‚BEbVTVDGB„b8Hu@CBEDWFdTu4QH6d‚DGT¢BX2#T¢7958RI2oDˆ5¢q‹H6dWd‚58dj@CFIBX4ƒFdT¢I†HgT¢d‚Y 582¬DGFIHd‚BX`8FoD‘Y‚BXhIH5¢q‘DGFIHH6PxRIT¢yCYWBX`¢2‹’Å1£D9BXY–T¢yEYW5k4QyEHgT¢d…DWFUTVD‘DGFpH‰H6dWd‚58d–hIHQ†HQ2IhpY#582¬DGFpH `8d‚BXhrY‚UT¢46BX2I`DW5T46HQdfDƒT¢BE2r58d‚hIHQd6’tuFIHŽH6dWd‚58dwDWHQd‚7¾46582oDƒT¢BE2IBX2I`DGFpHd`8d‚BXhrY‚UT¢46BX2I`DW5DGFpH yX5c@vHQYfDjs5c@vHQdj`8B„b8H6YvDGFIHu5¢dWhIH6dv5¢qëDGFpHŠ7#H‡DGFI5xh‹’v¤ëd‚58ˆ 7 jP†’8¥p’E¦ ¡ tV@vHuYWH6HŠDGFUTVDmDGFIHuH–—˜IdWH6YWYWBE582 æšÝxí‚5 ü 4369ݘí6 ç rVî‘ß#BXYˆT™nIdWY‚Dfr358d‚hIHQdˆTVIIdW5g—˜BX7#TVDWBX582wDW‰ 5 tux w í y ’””˜BX79BXyXT¢dWy„i8‘ t jPë’S¥p’E¦¢¦uYWFI5c@CYjDGFUTlD u æšÝxí‚5 ü 4£6kÝxA í €4ԁ ç rIæ   î…ßsçvBXY•TwYWH64Q582Ihxr358d‚hIHQdCT¢pIdW5g—˜BX7#TVDGBE582#DW5‘TwnUd‚Y‚DuhIH6dWB„b¢TlDGBEb¢H¢’”tuFpHyƒTVD‚DGHQd BXYˆd‚H6q•HQd‚dWH6h#DG5aT¢YˆDGFIHCDGFpdWHQH–r3†58BX2oDˆ46HQ2oDGH6dWHQhehpBEgsHQdWH62I4QHfT¢IpdW5g—˜BX7#TVDGBE582‹toT¢2ph#582pHf5¢q|DWHQ29YWH6HQY DGFpHwY‚RI797dT¢dfiŽd‚HQYWRpyEDCId‚HQYWH62oDGHQhBX2eDGFIHq•58d‚7 V  Ý  ß W í ä ݘí‚ü54‚6ÅÝxíA€4 ø“’dæzî…ß  ç   î…ß æz¥p’„¦   ç .

ÎxÐ ÎxЫÑmÒÔÓfÒQËm̈՟֐ÒÔєÑvÌ€Í‘ÌˆÓ…Ç”Ì ­®ƒ­ ¸º¹g¶e¹‡»”±f¼h½ ×uʈ̈͑ə˕×uÍfÚ   $ ¹6¿Å±Àj±¶eÁC± ”•Cr±Àˆ´w»”Ã9Àˆ² –—˜–—s™ dfehgsikjml2gonqpsrp‚iutvpxwzy{psr|7j¡ehr} Ü H6dWFUTVIY™DGFpHa7958Y‚D‘46587#79582‰H‡—pT¢79IyXH6Yf5Vq”nU2IB„DGH‡r£hIBEgsHQd‚HQ2I46Hq•58dW7…RpyƒT¢Y‘T¢d‚HwDWFIHaDWFIdWH6H‡r3†58BE2SD m 4 4QH62oDGHQd‚HQh˜r£hIBEgsHQd‚HQ2I46H…T¢pIdW5g—˜BX7#TVDGBE582IY•q•5¢dŠDGFIHnUd‚Y‚DfT¢2IhYWH64Q582phhIH6dWBEbVTVDWBEb¢HQYQÛ V V  Ý  ß  W í  Ý ä  ß W í ä ¦ æšÝxí‚ü5476rÝ˜í¥€4fç›ø~’Žæzî…ß  ç î ß   … æz¥p’„¦c¥oç æz¥p’„¦g—Sç ¦ æ•Ý˜í‚ü5486   Ýxí”ø¬Ý˜í¥€4fç›ø~’Žæzî…ß  ç î ß  … tuFIH6YWHT¢d‚HfDWFIH™‰UTVYWBXYˆq•58djœU‡ly|„Uxˆˆ¢y }‡€G}‡„hƒ}‡ƒœU}‡€‚¢xԇl€GžˆY‚BX2I46HfDGFpH6i9`8BEb¢HT¢2dT¢IId‚5g—pBE7dTlDGBX5¢2–DW5 TahpHQdWB„bVTVDGB„b8HfTVD•5¢2IH™hIBEYW46dWH6DWH™†58BX2oDvBX2#Ta79HQY‚FŽBX29DGHQd‚7#Y€5¢q›Y‚RIdWd‚58RI2IhpBX2I`as58BE2oDGYQ ’ €‘5c@•H‡b8HQd6t 2IH6BEDGFpHQdw5¢q•DWFIHQY‚HdH‡—˜Id‚HQY‚YWBX5¢2IYaDWHQyXyjRpYaFI5c@ 5VDGFIH6da†58BX2oDWYwBE2ADWFIH97#H6YWF)TVdWH9hIBEgsHQd‚HQ2I46HQhr5¢d FI5c@ ‰†58RI2IhUTVd‚i4Q5¢2IhIBEDWBX582pY™T¢d‚HaH62pq•58d‚4QHQhh’™”xRI4ƒFeT¢hIhpBEDGBE582UT¢y›BE2pq•58d‚7dTVDWBX582ed‚HQPSRIBEdWHQY™T#7958dWH YW5¢IFIBXYfDGBE4gTVDWHQheq•58dW7wRIyƒTVDWBX582h’ –—˜–—o‚ ƒ |7j¡r!go„…lfg†n{p‚rp‚iutvpxwmy‡p‚r|7j¡eUr} ¦ æšÝxíWü5486   ݘívøÆÝxíA€4fç î ß  … •582IY‚BXhIH6duDGFIHudWH6yƒTVDWBX582 æˆ( ¥ ÝsçzíŠä æz¥p’„¦q$8ç @CFIBE4ƒFeBXYuTw†58BE2SD•hIBEgsHQd‚HQ2I46HT¢IpdW5g—˜BX7#TVDGBE582#DW5‘T…YWH64Q582IhhIH6dWB„b¢TlDGBEb¢H¢’jò‘5c@ÿyEH6D•RIYuhIHQd‚BEb¢HT {9¢xš€‡Š y ‰w58†HQdGTlDG58dCdWH6IdWH6YWH62SDGTVDGBE582†q•5¢dCDWFIHuYGT¢79HaT¢IId‚5c—˜BE7dTVDWBX582h’€•582IYWBEhIHQdCDGFpHwq•58Rpd‘†58BE2SD 79HQYWF@CB„DGF‰†58RI2phUT¢d‚i††58BX2oDGYTV D ‹‰T¢2IT h ŒuYWFI5c@C2k‰†HQyX5c@u’ò‘5¢DGBE4QHwDWFUTVDf@CFIHQ2e@vH‘YW†HgT¢÷‰5¢q  DWFIHa2SRI7w‰sH6d‘5Vq”†58BX2oDGYCBE2T–7#H6YW F Žptë@vHw7#HQT¢2‰DWFIHw2xRp7…‰sH6df5VqŠy|„Ux3}‡€‡yŸ‡l€•s58BE2oDGYCH‡—˜4QyERIhIBE2I` DGFpHw‰†58RI2phUT¢dWBEHQY6’ ßeä ê‘ä ‹ ¡ 6 ¦ ¦ 6   Y 6 ¥ Y ‘ 6 —  Œ ¤ë5¢RIdus5¢BX2oDu79HQYWFh’vî…߉䒐‚rI擑 øª¦sç ò 5c@ BE7#†58Y‚HŽãBXd‚BX4ƒFIyEH6Da‰s5¢RI2IhUT¢dfiÅ4Q5¢2IhIBEDWBX582pYQtjÝmæ ¡ ç‘ä ݕ”sà‘Ým斐˜ç9ä ݕ—uTV2Ih¬RIYWHeDGFpH f 4QH62oDGHQd‚HQhehIB„ghH6dWH62I4QHaT¢IId‚5g—pBE7dTlDGBX5¢2e`8BEb¢HQ2‰oi˜jPë’s¥˜’E¦q$–TVD‘H6b¢HQdfi†58BX2oD‘BX2DGFpHa79HQY‚F‹’•™H ™oš‡›œvždŸdŸ¡ a›£¢¤žd¥¦›7§“¨q›©“›sª£«¦¬‘ 1 a›s¢“žd¥¥­¥§“žd¥¦›'«F®›s¢¤­A§“«¦¢£©“¨q«¦¢“§“Ÿd¯¦° .

  “ ǔȑɐʙËm̈ÍÏÎSЫÑvÒÔӑÒQËm̀Õz֐ÒÔєє̈͑̀Ówǔ̫ɐʈʈ͏×jؐÒfنə˧҇×uÓÚ T¢d‚dWBEb¢HuTVDCDGFIHq•58RpduHQPSRUTVDWBX582IY6Û æ“( ¥ Ýsç 4 ä ä æ“( ¥ Ýsç ¦ î ß … ¦ î ß … ¦ î ß … ¦ î ß …     æšÝ!”k6   Ýv4‹ø¬Ý  ç  „ … ä ä æšÝ£486   Ý øÆÝ ç „  æšÝ æšÝ æ“( ¥ Ýsç æ“( ¥ Ýsç  6   Ý „ øÆÝ … ç „ 6   Ý … øÆÝ!—3ç çsr¢î…ß çsr¢î…ß çsr¢î…ß çsr¢î…ß æz¥p’„¦c“oç ™dWBEDWBX2I`…DWFIHQY‚HuHQPSRUTVDWBX582IYCBE2DWFIHu7#5¢dWHYWRp`8`8HQYfDGB„b8Huq•58dW7 æ“( ¥ Ýhç 4 æ“( ¥ Ýhç  æ“( ¥ Ýhç „ æ“( ¥ Ýhç ä ä ä ä æ æ æ æµÝ!” 6   Ý£4 ø Ý   Ý 4±6   Ý £  Ý  BEDwBEYw46yXHQT¢dDGFUTVDu@vHd4QT¢2rH–—ppdWHQY‚YwDGFpHQ7 BX2T‰b¢HQ4‡DG58dÔr37#TVDGd‚B„—eq•58d‚7†t§T¢2phAq•RIdfDGFIH6dQt§DGFUTVDwDGFpH dWH6YWRIy„DGBE2I`–7dTVDWdWB —ŽFUTVYfT…b¢HQd‚iY‚sH64QBXT¢y‹q•58dW7‰’”132oDGd‚5˜hpRI4QBE2I` Ý£4 ¶¸· … øÝ „ 6   Ý „ Ý „     æz¥p’„¦q98ç øÝ … 6   Ý … øÝ!— ² ä ݎ ³´´ T¢2Ih µ´ Ý „ ¹ Ý … Ý  ·· à t Œ¦² º y ä Ý!” ¸ ¶ ·· ¦ ¡ · ´³´ î…ß  ´µ ¡ ¹ Ý!— æz¥p’„¦acoç » ä @vHa4QT¢2dWH6@Cd‚BEDWH‡jP뒆¥p’„¦q9…T¢Y ¦ ´³ î ß  ´µ´ … 6   ¦ 6   ¦ ¦ 6   ¦ ¦ ¶ ·· · 6   ¦ ¹ æz¥p’„¦ahoç ² ä »Dݑ ² ø¼t ŒA² º y ( ¥ Ýe æz¥p’  ¢¡ ç t FIBXYH–—pT¢7#pyXH…BEyXyXRpY‚DGdWTVDGH6Y™TŽ7#TVDWdWB„—†hIB„ghH6dWH62I4QH‘58†HQdWTVDG5¢dQ’½ˆT¢4ƒFyEBX2IHa5¢qvTŽ7dTlDGdWB —‰hIBEqƒr u q•HQd‚HQ2I46H58†HQdWTVDG58dˆBEYv‰UTVYWHQhd582eTa†58BE2SD€hIB„ghH6dWH62I4QHf58sH6dGTVDW58dQtx‰IRpD€DGFpH™†58BX2oDv58sH6dGTVDW58dWY€RIY‚HQh q•dW5¢7 yXBE2IHŠDG5uyEBX2IHCT¢d‚Hu2I5¢D€2IH64QHQY‚YGT¢d‚BXy„i…DGFIHCYWT¢7#HV’j¤U58djH‡—pT¢79IyEH¢to‰†58RI2IhUTVd‚i‘46582IhIB„DGBX5¢2IYj7dTgi hIBE46DƒTlDGHDGFITVDŠDGFpHwyEBX2IH6YuTVDu5¢du2IHgT¢dŠDWFIHu‰s5VDWDG5¢7 58dŠDW585¢q˜DGFIHu7#TVDGd‚B„—#‰†H7#5xhIBEnIHQh‹’”132DGFpH H‡—xDWdWHQ79Hw4QT¢YWHa5¢qvDWFIHw7#TVDWdWB„—ŽhIBEgsHQd‚HQ2I46Hw58†HQdWTVDG5¢d‘d‚HQId‚HQY‚HQ2oDGBE2I`eT#Y‚sH646DWdGT¢y§7#H‡DGFI5xh‹tU2I5¢2IH 5¢qˆDGFpH‘yXBX2pHQYBEY™DGFIH‘YWT¢79H¢’wtuFIH–7dTVDWdWB —5¢sH6dGTVDW58dWYd‚HQId‚HQY‚HQ2oDGBE2I`‰DWFIH…DGFpdWHQH–r3†58BX2oD46HQ2oDGdWT¢y„r hIB„ghH6dWHQ2p4QH#T¢IId‚5c—˜BE7dTVDWBX582pYq•58dwT‰nUd‚Y‚DwT¢2IhrY‚HQ46582Ih)hIH6dWB„b¢TlDGBEb¢H‘@CBEDGFãBXd‚BX4ƒFIyEH6Du‰†58RI2IhIT¢d‚i 4Q5¢2IhIBEDWBX582pYu582‰T‘q•5¢RIdfr£s58BE2oDu7#H6YWF†T¢dWH .

ÎxÐ ÎxЫÑmÒÔÓfÒQËm̈՟֐ÒÔєÑvÌ€Í‘ÌˆÓ…Ç”Ì ×uʈ̈͑ə˕×uÍfÚ   9 ( ä ¦ î ß   … ³´´ ´µ ¡ 6 ¦ w ¡ 6 ¦ w ¦ ¡ 6 ¦ w ¦ ¶ ·· · ¡ ¦ ¹ à¥( ¥ ä ¦ ´³ î ß  ´µ´ … 6   ¦ 6   ¦ ¦ 6   ¦ ¦ ¶ ·· · 6   ¦ ¹ æz¥p’   ¦sç i™Y†TÅq•RId‚DWFIHQd‰H‡—pT¢79IyXHVt™d‚HQIyXT¢4QHkDGFIHkq•58RIdfDGFðyXBE2IHeBX2`jPë’f¥˜’E¦c“¬‰oi DGFIHkq•58yEyX5c@CBX2I`Ł†58BE2SD 58†HQdWTVDG58dŠq•5¢dCT9ò‘HQRI7#T¢2I2‰‰†58RI2IhUTVd‚ie46582IhIB„DGBX5¢2A柔˜H6Hw”xHQ46DWBX582†¥p’ “p’ ç‡Û æˆ( ¥ Ýsç … ä  Ý ¦ V 6 ¥ … î ß  ß W —   ¦ æ Ý 6rÝ ç • „ ¥ ‘ î ß  …   æz¥p’  8  ç @CFIH6dWHuDGFIHu‰†58RI2IhUTVd‚ie46582IhIB„DGBX5¢2eBXY V  Ý ä  ßhW ¦¾¡¿ V  Ý  ߁W — æz¥p’   ¥oç tuFIH62ºDGFIH†7dTVDWdWB —Æ58sH6dGTVDW58d–q•58ddTADGFId‚HQH‡r£s5¢BX2oDd46HQ2oDGdWT¢y„r£hIBEgsHQd‚HQ2I46BX2I`YW4ƒFIH67#HTVDdBE2oDGHQd‚BX58d †58BX2oDGY9T¢2phókžc}ƒƒ‡l„†ˆ¢~W‡l€‚ˆx}‡€‡œoœ†€GF ‡ ‰8y|{9¢xšyŸ‡l„eq•58d9TAò‘HQRI7#T¢2I2Æ4Q5¢2IhIBEDWBX582¬582¬DGFpH‰d‚BX`8FoD–BEY `8B„b8HQ2‰‰oi ¦ ´³ î ß  ´µ´ ‘ 6   ¦ ( ¥ ä 6   ¦ ¦ 6   r ¥   ¢ ¦ ¶¸·· · 6   r¢¥ ¦ ¹ æz¥p’   —Sç T¢d‚HT¢yEy‘Œ6HQd‚58YŽH‡—˜46HQpDŽq•5¢d#DGFp58YWHeTVyX582I`A‰UT¢2phIYd@CFIBE4ƒFªT¢dWH†4QyXRpY‚DGH6dWH6hªT¢dW58Rp2Ih«DGFpH4QH62oDGdGTVy hIBXT¢`8582UTVyŸ’m™Hu4gT¢yEy‹YWRp4ƒFT‘7dTlDGdWB —eT˜Àƒ¢„†ˆx}Wˆ9{9¢xš€–yŠ‰wT¢2ph‰BE2oDGdW5xhIRI46HDGFIHu2I5¢DGTVDGBE582 ˆT¢4ƒFr5¢q€DGFpHQYWH97#TVDGd‚B„—ehpBEgsHQdWH62I4QH–58†HQdWTVDG5¢dWYuBXYuTeYWPSRUTVdWH97dTVDWdWB —@CBEDWFkHQyXH67#H62oDGYDGFUTlD Áe椑 Û‹ëàÌsàúQçjä ³´´ ‹ ´´ ´´ ´µ Œ ‹ º ŒÄº Œ ’’ ’ ‹ º ¶¸·· ¦ ·· ·· · Œ ¹ ‘ ’’ ’ æz¥p’   $8ç @ FIH6dWHRIY‚H5¢q8‘ BX2#DGFIH™T¢dW`8Rp7#H62SD•BXYv58pDGBE582UT¢y†T¢2phŽDGFIHfBXyXyERIY‚DWdGTVDWBX582–BEY•`8B„b8HQ2Žq•5¢d•TwY‚BX79IyXH C xš€‡yšˆ¢yšc…x‡l„†¢ ‡7#TVDGd‚B„—uTVyEDGFp58RI`8FaT¢2oiu2SRI7…‰†HQd§5¢qI‰UTV2IhIY§BXY§Tf†58YWY‚BX‰IBEyXB„D3i8’›iÅDGd‚BXhIBXT¢`8582UTVy¢7#TVDGd‚B„— ² à ºQ ² à Œc ² 疒jtuFIHuT¢d‚`8RI79HQ2oDGY•q•5¢d @CBEDWFI58RpDŠ4Q582IYfDƒT¢2oDWYfTVyX582I`aDGFIH‰IT¢2IhIYC4gTV2e‰sHH–—ppdWHQY‚YWH6hT¢ Y ÁŽæ ‹ë Td‰UTV2IhIHQhe7#TVDWdWB„—T¢dWH‘TVyE@•TcixYf5˜hIhBX22SRI7…‰†HQd™T¢2IhDWFIH…46HQ2oDGdWT¢yv582pHe¢ é耢cž‘dWH6q•H6dWY™DW5dDGFpH 4QH62oDGdGTVy›hIBƒT¢`¢582UT¢yz’ .

  c ™rHu4QT¢2†2p5s@ ǔȑɐʙËm̈ÍÏÎSЫÑvÒÔӑÒQËm̀Õz֐ÒÔєє̈͑̀Ówǔ̫ɐʈʈ͏×jؐÒfنə˧҇×uÓÚ ² TVY  `8HQ2pHQdGTVyXBXŒ6H58RIdCId‚H6bxBX58RpY‘H–—pT¢7#pyXHQY6’ˆãH6nU2pBX2I` Ý ³´´ Ý£4 ¶ ·· · Ý ² ä ݍ ´´ Ý „ ´´ ’’’ ´µ Ý Å •  ·· ·· æz¥p’   “oç ¹ ² ‰oi @vHa4QT¢2‰T¢IId‚5c—˜BE7dTVDWHDGFIHuY‚HQ4Q5¢2IhhIH6dWB„b¢TlDGBEb¢H5¢q Ý ² ä ( ¥ Ýe ¦ ² ø¼t A ÁeæÔ¦8àx6   àg¦sç Ýw Œ ²º y î ß  … æz¥p’   98ç @CFIH6dWH~t ŒA² º y YfDƒT¢2phIYwq•58dwDGFIHdb¢HQ4‡DG58d‘FI5¢yXhIBE2I`‰DWFIHeãBEdWBX4ƒFpyXH6Da‰†58RI2IhUTVd‚i4Q582IhpBEDGBE582IYw582)DGFpH yXH‡q|D‘T¢2IhdWBE`8FoDCYWBEhIHQY6Û tA Œ ²º y ä ¦ ùúÝ!”sà ¡ àSYaYaY6à ¡ àGÝ!—Ÿû§Æ î ß  … æz¥p’   coç 1£q”@vHwId‚HQY‚4QdWBE‰sHwòfH6RI7dTV2I2‰†58RI2IhIT¢d‚i‰4Q5¢2IhIBEDWBX582pY‘5¢2†DGFpHad‚BX`8FoDfYWBXhpH¢t†T¢Y™BE2fjPSYQ’s¥p’   —ŽTV2Ih ¥p’  8  tp@vHunU2Ih ( ¥ Ý ² ä @CFIH6dWH ¦ ² Q ² à Œl ² Ç Áeæ ‹U à ¦sç Ýw ø t ŒA² º y î ß  … æz¥p’   hoç ‹² t¦ Œ ²º y Œ² ä ä ä ù ¦8àQ¦8àxYaYaY‡à   r¢¥Vû Æ ùŠ6   àx6   àx6   àxYaYaY6àx6   r¢¥Vû Æ ¦   î…ß ݕ”sà ¡ à ¡ àxYaYaY–à î ß ÉÈ ‘ ¥ V  Ý Æ  ß W —ˆÊ ò 5¢DWBX46H‘DWFUTVD€DGFIHf7dTVDWdWB —‘58†HQdWTVDG58d‚Yv`¢BEb8H62Ž‰oiˈPSY6’p¥p’   9wT¢2IhŽ¥˜’   hw4gT¢d‚d‚i–7958dWH‘BX2˜q•58dW7#Tlr f DGBE582eDGFUTV2‰DWFIHus58BE2oDC58sH6dGTVDW58du`8B„b8HQ2‰‰oiDjPë’s¥˜’E¦q$x’€132˜jPSYQ’†¥p’   9‘T¢2Ih†¥p’   hptpDGFpHa‰†58RI2IhIT¢d‚i 4Q5¢2IhIBEDWBX582pYfFUTgb¢H…‰sH6HQ2eRI2pBXPSRIHQy„i‰YW†HQ4QB„nUHQhkTV2IhBEDfBXYf4QyEHgT¢dCDGFUTlD™DGFpH…YGT¢79Hws58BE2oD58†HQdWTVDG58d FUT¢Y‘‰†HQHQ2T¢IIyEBXH6h†TlDfH‡b8H6d‚i‰s5¢BX2oDCBX2DGFIHnUH6yXh‰H‡—˜4QH6pDTlD‘DWFIHw‰s58Rp2IhUT¢d‚BXHQY6’vtuFpHwT¢‰pBXyXB„D3i#DW5 YW†HQ46BEq|iwBX2wDWFIHu7dTlDGdWB —ahIHQd‚BEbVTVDGB„b8HŠ5¢sH6dGTVDW58dvDGFpHuH‡—pT¢46Dˆ2ITVDGRId‚Hu5¢qUDGFpHCT¢IId‚5c—˜BE7dTVDWBX582wTVDjDGFpH bVT¢dWBE58RIYfs58BE2oDGY™BE2DGFpHwnUHQyEhBX2p4QyXRphIBX2I`9DGFIHw‰s58Rp2IhUT¢d‚BXHQY™sH6dW79BEDGYCDGFpH…RIYWH‘5VqˆPxRpBEDGHw`8HQ2IH6dGT¢y 4Q5¢2IY‚DWdWRI4‡DGBX5¢2IYu@CFIBX4ƒF@CBXyEy‹‰sHRpYWH6q•Rpy§yXTVDGH6duBX2e46582IYWBEhIHQdWTVDGBE582IYC5¢qmY‚DGT¢‰IBXyEBED3i¢’ ÌsÍ «¦§“›p§¤¨‘­¥§ÏÐ Î ž%©k­uъÒq¬‘ª†§“žd«¦¬Ó«¦Ñv§“ždÔk›1«¦¬‘ŸÕ¯q©ˆžd¬‘ª£›X›s­Fª˜¨Ï›£Ÿd›£Ôɛ£¬S§Éª†«¦¢¤¢¤›s©“®«F¬‘ q©'§¤«Q«¦¬q›1©ˆ®v›sª£ž¸ÖvªX©“®‘­¥§¤žd­¦Ÿ Ÿd«xªs­¥§¤žÕ«F¬Â° .

ÎxÐ ÎxЫÑmÒÔÓfÒQËm̈՟֐ÒÔєÑvÌ€Í‘ÌˆÓ…Ç”Ì ×uʈ̈͑ə˕×uÍfÚ   h ” BX2p4QHv@vH•7#T¢÷VHv46582IYWBEhIHQdWT¢‰IyEHvRIY‚HŠ5¢qp‰s5¢DWFw7#TVDGd‚B„—wTV2Ih…†58BX2oD§58†HQdGTlDG58d‚YQt¢B„D§BXY§BX79s58dfDƒT¢2oD ˜ DG59HQYfDƒT¢‰pyXBXY‚F‰T9d‚HQyƒTlDGBX5¢2e‰sH‡D3@•H6HQ2DWFIHQ7‰’vi †58BX2oD‘58†HQdWTVDG58duBEYC`8HQ2IH6dGT¢yEyEi#@CdWBED‚DGH62‰q•58duY‚5879H hIH6dWBEbVTVDWBEb¢H™TVDvDGFIHfdWH‡q•HQdWH62I4QH†58BE2SDvê…BX2#DGHQd‚7#Y€5¢q˜2IH6BX`8FS‰†58dWBE2I`ub¢TVyXRIH6Y•5¢q‹DWFIH™q•RI2p46DGBE582‹’j¤U58d H‡—pT¢79IyEH æ“(  Ýhçšíuä׋ ÝxíA€   ø~‹¡4fÝ˜í¥€4hø“ŒƒÝxíjø“ºx4fÝxíWü54 æz¥p’ ¥ ¡ ç 79BX`8FoDˆ‰†HCDWFIH‘s58BE2oDv5¢sH6dGTVDW58d”q•58d€TunIdWY‚DvhIH6dWBEbVTVDWBEb¢H¢’vtuFpH™4Q5¢dWdWH6YW†582IhIBE2I`w7dTVDWdWB —‘58†HQdWTVDG58d FUT¢YŠq•5¢duBEDGYŠT¢dW`8Rp7#H62SDWYuDGFIH465˜H‡¨#46BXH62SDWYC`8BEbxBX2p`…DGFIH@vHQBE`8FoDGYuDG5‘DGFpHb¢TVyXRIH6Yu5¢qmDWFIHq•RI2I4‡DGBE582 TVDwDGFIH9bVT¢dWBE58RIYwyX5x4gTVDWBX582IY6’ei ê¢r£YWFIB„q|DwBE2ADWFIHd†58BX2oDa58†HQdGTlDG58dw4Q58d‚dWH6YW†582IhIYwDW5T‰hIBƒT¢`¢582UT¢y YWFpBEq|DCBX2eDWFIHu7dTVDWdWB —Ž58†HQdWTVDG5¢dQ’jtuFSRIYCDGFIHu7#TVDGd‚B„—dHQPSRIBEbVT¢yEHQ2oDC5¢7 q jP†’†¥p’ ¥ ¡ BXY ò 5VDGHdDWFIHeT¢hIhpBEDGBE582r5Vq™T†ŒQHQd‚5eBX2rDWFIHŽnIq|DWFÅHQyXH67#H62oDw@CFIBE4ƒFÆ7#T¢÷VHQY…B„D…4QyEHgT¢dwDGFUTlDӌ9BEYwDGFpH f 4Q5xH6¨946BXHQ2oD‘5¢qmÝxís’ ² äfÁe戋  ¥ ( 5ݎ à ‹¡4‡à¥Œcॺa4ƒà ¡ ç Ý ² æz¥p’ ¥p¦sç –—˜–—˜– dfpsr!g†eXØQg†tك |7j¡r!g†tvp‚} tuFIHfT¢‰†5sb¢HfBEyXyERIY‚DWdGTVDWHQh‘4gTVYWHQY€BX2‘@CFIBE4ƒF9DWFIH‘‰s58Rp2IhUT¢dfi‘4Q582phIBEDWBX582IYˆTVdWHunp—˜HQh‹’j1£qsDGFIHC‰†58RI2Ihxr T¢dfie4Q582phIBEDWBX582IYwT¢d‚H…œU}‡€‡yŸ‡Qˆ¢yŸ‡t›DWFIH9q•5¢dW7 5¢qvDGFIH–7#TVDGd‚B„—58sH6dGTVDW58du4ƒFUT¢2I`¢HQYQ’e•582IYWBEhIHQduDGFpH HQBE`8FoD‚r£s58BE2oDv†HQd‚BX5xhIBX4€7#H6YWF–YWFp5s@C2–‰†HQyE5s@u’mtuFIBXYv4gT¢2‘H6BEDGFpHQd”‰†HŠId‚HQYWH62oDGHQh9582–T™yEBX2IHQT¢dv79HQY‚F @CBEDWF–dWHQ†HgTlDGHQh9HQ2oDWdWBXH6YQtS58dj7958dWH•YWRI`8`¢HQY‚DWBEb¢HQyEi…5829Tu46BXd‚4QRIyXT¢d”79HQYWF#T¢YjBE2#¤mBX`8RId‚Hu¥p’ ¥p’m™ªFpHQ2 DGFpH‘7#H6YWFeBXY™yƒT¢BEh58RpD582DGFIH‘†HQd‚BX79H6DWHQd™5Vq•TŽ46BXd‚4QyXHVt‹BED™hI5xHQYWÚ 2  Dud‚HgT¢yEyEi‰7dTlDWDGH6df@CFIH6dWH‘DGFpH 2SRI7…‰†HQd‚BX2I`uY‚DGT¢d‚DWY•TVY€yE582I`wT¢YˆBED  HQ2IhIs Y ŽwTlD€DWFIHCs5¢BX2oÚ D ۂRIY‚DvId‚HQ4QH6hIBX2p`aB„DGYjY‚DGT¢d‚DWBX2I`uyE5˜4QTVDGBE582‹’ ßeä ê‘ä YaYaY 6 9 6 c ¡ ¡ ¦ 6   ¦ 6 ¥ Y 6 — Y 6 $ Y 6 “ Y 6 9 Y 6 c Y    ¦ ‘ 6   YaYaY jBX`8FoDC†58BX2oDWYu582T‘yEBX2IHQT¢dŠ†HQdWBE5˜hpBX479HQY‚F‹’€î‘ßeä   srx‘ tuFIH‘7dTVDWdWB —…DGFITVDˆdWH6IdWH6YWHQ2oDWYŠhIBEgsHQd‚HQ2I46BX2I`uY‚4ƒFIHQ79HQY€q•58djYW4QT¢yƒT¢dˆH6PxRITVDGBE582IYˆ582#Tw†HQd‚BX5xhIBX4 7 HQYWFwBXY˜d‚H6q•HQd‚dWH6haDG5fT¢Y§TŠœU}‡€‡yŸ‡Qˆ¢yŸh7#TVDGd‚B„—s’˜i¬D3ixIBE4gT¢yS†HQd‚BX5xhIBX4vDWdWBEhIBƒT¢`¢582UT¢yV7dTVDWdWB —58†HQdWTVDG58d 9 @CBEDWF2I582SRI2IB„q•58dW7 HQ2oDGd‚BXHQY‘BXYu`¢BEb8H62eq•58dCT#“sr3†58BX2oDu79HQY‚F†‰oi ² à ºQ ² à Œs ² çjä ÁkÜoæz“–Û ‹† Œ 4 º S ‹¡4ތ  ´³´ ‹  ´´ ´´ ´µ´ ºx4 ‹ „ ‹ º„ Œ„ º Œ … … … ‹ † º† Œ† ¶ ·· ‹3Ý ¸ ·· ·· ·· ºAÝ ¹ Œ¥Ý æz¥p’ ¥   ç .

¥ ¡ ǔȑɐʙËm̈ÍÏÎSЫÑvÒÔӑÒQËm̀Õz֐ÒÔєє̈͑̀Ówǔ̫ɐʈʈ͏×jؐÒfنə˧҇×uÓÚ 1 8 2 7 3 6 5 ¤mBX`8RId‚Hu¥p’ ¥pÛ 4 jBX`8FoDC†58BX2oDGYu5¢2T‘4QBEdW46RIyƒT¢d•7#H6YWF‹’ –—˜–—ˆß àág†rtsâã†|iÉjmƒ |7j¡r!g£tvps} 132e`8H62IHQdWT¢yŸt›TVYYWFI5c@C2ABX2DWFIH…H–—ITV7#IyEH¢thDWFIH…H6yXHQ79HQ2oDWYuT¢yX582p`#DGFpH‘hIBƒT¢`¢582UT¢yEYf5¢qvTŽ†HQd‚BX5xhIBX4 7#TVDGd‚B„—uTVdWHˆ2I5¢D§4Q5¢2IY‚DGT¢2oDg' ’ €‘5c@•H‡b8HQd6tST‘Y‚sH64QBXT¢yoYWRI‰pYWH6Dv5¢qpT‘†HQd‚BX5xhIBX4j7#TVDGd‚B„—™BXY›DGFpHv46BXdW46RIyƒTV2SD 7#TVDGd‚B„—st‹q•58dW79HQhe@CFIH62DWFIH–HQyXH67#H62oDGYwT¢yX5¢2I`eDGFIH‘bVT¢dWBE58RIY‰UTV2IhIYaT¢dWH‘46582IY‚DGT¢2oDg’Ž•BXdW46RIyƒTV2SD 7#TVDGd‚BX4QH6YˆIyƒTgi9TwbxB„DƒT¢yUd‚58yXHCBX2958RIdvT¢2UTVyEixYWBEYQ’v™H‘@CBEyXyUFUTgb¢H™7…RI4ƒFd7#58d‚HuDG5wYWTgidTV‰s58R˜D€DWFIHQ7 yƒTlDGHQd6’jtuFIHu7#5¢Y‚DC`8H62IHQdWT¢y›4QBEdW4QRpyƒT¢2oDC7#TVDGd‚B„—d5¢qv5¢dWhIH6d‘—–BEY Œ¥äČS4åŒ  Œ „ Œ Œ¥ä±ŒS4åŒ  ´³´ „ Œ µ´  Œ „ Œ¥ä±Œx4 ŒS4æŒ  Œ „ ŒAä ¹ ¶¸·· · æz¥p’ ¥8¥oç òf5VDGBX46HwDWFUTVDfHgT¢4ƒFdW5c@ 5¢qˆT94QBEdW4QRpyƒT¢2oD‘7dTlDGdWB —eBEYfY‚FIBEq|DWHQhÅæ•YWH6H…¤vBE`8RIdWHw¥p’ ¥oçu582IHwHQyEHQ79HQ2oDfDW5 DGFpHŽdWBE`8FoDw5¢quDWFIHd5¢2IHeT¢‰†5cb8HŽB„Dg’‰tuFIHeYW†HQ46BƒT¢y€4gT¢Y‚HŽ5¢q‘T†DWdWBEhIBƒT¢`¢582UT¢y”46BXd‚4QRIyXT¢2oD…7#TVDGd‚B„—kBEY `8B„b8HQ2‰‰oi ÁkÜx擑 Û‹Uौcॺ6çˆä ³´´ ‹ ´´ ´´ ´µ º Œ Œ º º ‹ ’’ ’ ‹ Œ ¶ ·· ¦ ‹ ¸ ·· ·· · ’’ º ’ Œ ¹ ‘ æz¥p’ ¥¢—Sç ™ªFIHQ2#DGFIHfY‚DƒTV2IhUT¢d‚h#DGFpdWHQH–r3†58BX2oD€4QH62oDGdGTVy„r3hpBEgsHQdWH62I4QBE2I`aT¢IId‚5c—˜BE7dTVDWBX582pY”q•58d•TnUd‚Y‚DuTV2Ih .

ÎxÐ%çxÐóÇu×uәژËmÍÏè•Ç‘˧ÒÔÓ{é ֐ÒÔєє̈Íf̈ÓwÇvÒÔÓqé ڋǔÈfÌvنÌvÚ×uѬəÓëê ×uÍfÖÌˆÍ ¥p¦ Y H64Q582ph#hIH6dWB„b¢TlDGBEb¢H¢t8Y‚HQHXjPë’o¥p’   ¦8toT¢d‚HuRIYWH6h‘@CBEDGF‘†HQd‚BX5xhIBX4v‰s5¢RI2IhUT¢dfi…46582IhIB„DGBX5¢2IYQt¢DWFIH6iaDƒT¢÷VH W DGFpHuq•5¢dW7 ¡ 6 ¦ w ¦ æ“(  ç Ü ä ¦ î ß   … ³´´ ´µ ¦ ¡ 6 ¦ w 6w¦ ¡ ¡ ¦ 6a¦ ¸ ¶ ·· · ¦ ¹ ä ¦ ÁkÜxæ£6w¦8à ¡ àQ¦sç î ß   … T¢2Ih æ“( ¥ ç Ü ä ¦ ³´ î ß  ´µ´ … 6   ¦ ¦ 6   ¦ ¦ 6   ¦ ¦ ¦ ¶¸· ·· ä 6   ¦ ¹ ¦ ÁkÜxæf¦8àS6   àQ¦sç î ß  … æz¥p’ ¥3$8ç  yXHQT¢dWy„i8thDWFIHQY‚H–YW†HQ4QBXT¢yv4QT¢YWH6Yu5¢q€sH6dWBX5xhIBE4w58†HQdWTVDG58d‚YuT¢dWH…T¢yXY‚5e4QBXd‚4QRIyXT¢2oD58†HQdWTVDG58d‚YQ’aô˜TVDGH6d • 582)@vHeDƒT¢÷VHT¢hpbVT¢2oDƒT¢`¢HŽ5¢qCDGFIBEY…YW†HQ46BƒT¢yvIdW5¢sH6d‚D3i8’òf5VDGBX46HdDGFITVD…DGFpHQdWHTVdWHŽ2I5k‰†58RI2IhIT¢d‚i 4Q5¢2IhIBEDWBX582Žb¢HQ4‡DG58d‚YfY‚BX2I46HuDWFIBXYuBE2pq•58d‚7dTVDWBX582dBXYCT¢yEyhBX2oDWHQdWBE58dŠDG5‘DWFIHu7dTVDWdWBE4QHQYŠDWFIHQ79YWH6yEb¢HQYQ’ ­®£ì í Ã9¶e²m»jÀîeÁu»j¹g¶ð¾¹6¿Å±Àj±¶eÁ‘¹g¶ðï BÁ•³±R ±²ðÃÓñòG¾¶ôó ” Àv¼ ·±À –—ˆßõ—s™ öË|v÷õã£eUrgöË|øã†ps} tuFIHetmTgi˜yE58d…Y‚HQdWBEHQY‘H–—pIT¢2IYWBE582¬5¢qCq•RI2I4‡DGBX5¢2IY–T¢‰†58RpD–Tnp—˜H6hƁ†58BE2SD…IdW5cbxBXhIH6Y–T7#HQT¢2IYwq•5¢d 4Q5¢2IY‚DWdWRI4‡DGBX2p`¬nI2IBEDWH‡r3hpBEgsHQdWH62I4QHs58BE2oD‚r35¢sH6dGTVDW58dWY#5¢qaT¢2oi 58dWhpHQdQ’i YWBE7#pyXHTV2IhªY‚DGdWT¢BX`¢FSD q•58df@ŠT¢d‚h#@•Tgi9DW5a4QT¢dWdfi‘DGFIBEYv5¢RpD€BXYˆDG5w4Q582IYfDGd‚RI46DŠT  tmTgixyX58dˆDƒT¢‰pyXH¢¸ t Žu@CFIBX4ƒFŽ7#T¢÷VHQY€H‡—xDGH62IYWB„b8H RIY‚H5¢qDWFIHH–—pIT¢2IYWBE582º`8B„b8H62º‰oz i jP뒊¥˜d ’ cp’ªifYŽT¢2ºH–—pT¢7#pyXH¢t•4Q582IY‚BXhIH6ddtmT¢‰IyEH¥p’„¦†@CFpBX4ƒF dWH6IdWH6YWH62SDWYTrtmTcixyE58d#DGT¢‰IyXHq•58d#T¢2 T¢pIdW5g—˜BX7#TVDGBE582Æ5¢quTrY‚HQ4Q5¢2Ih hIHQd‚BEbVTVDWBEb8H†RIYWBE2I`DWFIdWH6H bVT¢yXRpHQY‰5¢qwDWFIHAq•Rp2I46DWBX582ÿ4QHQ2oDWHQdWH6h T¢‰†58RpDDGFIHA†58BX2oD†TVD‰@CFIBX4ƒFðDGFIHAhIHQd‚BEbVTVDWBEb8HABXYeDW5«‰†H H6bVT¢yERUTVDWHQh‹’ .

¥   ǔȑɐʙËm̈ÍÏÎSЫÑvÒÔӑÒQËm̀Õz֐ÒÔєє̈͑̀Ówǔ̫ɐʈʈ͏×jؐÒfنə˧҇×uÓÚ V Ýxí î‘ß  Y t u¦ú¤w y í u ú  Ý  ß  W í ø ¦ 戋SÝxí¥€4hø“ŒƒÝxíjø“º–ÝxíWü54fçˆä×ù î ß  … tvuxw y í u ‹{YSæ|rG¦sçhY 4 4 î…߁Y tvu¦ú¤w y í u ú ¦ î…ß  Y tvuAû¤w y í u û 4 ‹½Yxæ|rW¦sç „ Y Ý ºkYoæf¦sç „ Y Ý 4 î…ß „ Y tuAü˜ w y u ü í ‹‡Yxæ|rW¦sç … Y ºÉYSæf¦sç … Y 4 … … 4 î…ß … Y ‹{YQÝ˜í¥€4 ŒkYcÝxí ºkc Y Ýxí‚ü54 ‹ Œ º ºhYxæf¦sçhY 4 4 4 ‹‡Yxæƒrƒ¦sç  Y  4 ºkx Y æf¦sç  Y  tmT¢‰IyEH¥p’„¦8’¬tvTgixyX58d…DGT¢‰IyEHŽq•58d‘4QH62SDWHQd‚HQhº¥sr3†58BX2oD–ô˜T¢`8dWT¢2I`8BXT¢2¬T¢IpdW5g—˜BX7#TVDGBE582rDG5TeYWH64Q582Ih hpHQdWB„bVTVDGB„b8H¢’ t FIHDƒT¢‰IyEHeBXY‘46582IY‚DWdWRI4‡DGH6hºYW5kDGFUTlD#Y‚5879He5¢qfDWFIH†T¢yX`8H6‰IdGTBEY–YWBX79IyEBEnUH6h‹’)iŠD–DGFIHeDW58 u 5¢q€DGFIH‘DGT¢‰IyXHa@•H9YWH6H#TV2H–—ppdWHQY‚YWBE582A@CB„DGFATePSRIHQYfDGBX5¢27#T¢dW÷뒑tuFIBEYdWH6IdWH6YWHQ2oDWYw582IH–5¢q€DGFpH PSRIHQYfDGBE582IYjDGFUTlD€TY‚DGRphpi–5¢qëDGFIBEY”DƒT¢‰pyXHŠ4QT¢29T¢2IY‚@vHQd6’p2UTV7#H6yEi8tV@CFUTVDˆBXY”DWFIHuyX5x4gTVypHQd‚dW58dj4QT¢RIY‚HQh ‰oi–DGFIH‘RIYWHf5¢qhDWFIBXY€T¢pIdW5g—˜BX7#TVDGBE58! 2 ùeòf5VDGBX46H‘DWFUTVDvT¢yEyU5¢qhDWFIHfDWHQdW79YˆBX2–DGFIHfHQPSRUTVDWBX582#T¢I†HgTVd BX2T#4658yXRI792†TVDfDWFIHwyEH6q|D‘5¢q”DWFIHaDGT¢‰IyEHŽæšT¢yEDWFI58RI`8FhtUBX2†DWFIBXY‘4gTVYWH¢thî…ß  FUT¢Yf‰sH6HQ27…RpyEDGBEIyXBEHQh BX2oDW5HQT¢4ƒF«DGH6dW7 @CFIBE4ƒF«Y‚BX79IyXB„q•HQY…DGFpH‰DWHQdW79Y–DG5A‰sH‰IRpD#BX2oDG5kDGFpH‰DGT¢‰IyXHV’ 疒«tuFIH62º2I5¢DWBX46H DGFITVDCTVDuDWFIHFIHgTVh‰5VqmHQT¢4ƒF‰4Q58yERI7#2ŽDWFIHQd‚HuT¢I†HgT¢d‚YŠDGFIH4658797#582#qšT¢46DW58dŠDGFUTlDu5˜464QRId‚YCBX2ŽDGFpH H‡—˜UTV2IYWBE582†5VqvHQT¢4ƒF†DWHQdW7 T¢‰s5¢RpDuDGFIHu†58BX2oD€êUtpDWFUTVDCBXY6t î…ß ý Y‡V  ß ý W í  ý Ý õ#ä ¡ àg¦8à   àxYaYaY t FIHC4Q5¢yXRI792IY”DG5DWFIHCdWBX`¢FSD”5¢qsDWFIHCyXH‡q|DG7958Y‚Dj5¢2IH¢tSRI2IhpHQdjDGFIHCFIHQT¢hIBE2I`8YQto7#T¢÷VH‘Rp9DWFIHutvTgixyX5¢d u DƒTV‰IyXHV’'ˆT¢4ƒF–HQ2oDGdfi‘BXYmDGFIHu465˜H‡¨#46BXH62SDˆ5Vq†DWFIHŠDGH6dW7 TVD”DGFpHŠDG58‘5¢qëDWFIHu4Q58d‚dWH6YW†582IhIBE2I`4Q58yERI792 BX2†DGFIH…tmTgixyX58dY‚HQd‚BXHQY™H‡—˜UT¢2IY‚BX582e5Vq€DWFIHwDWHQdW7 DG5ŽDWFIH…yEH6q|D5VqjDGFIH…4658dWd‚HQY‚s582phIBX2I`ed‚5c@u’a¤U58d H‡—pT¢79IyEH¢t8DGFpHuyƒT¢YfDjdW5c@©BX2…DGFpHŠDƒT¢‰IyEHŠ4Q5¢dWdWH6YW†582IhIYˆDG5DWFIHutmTcixyE58djYWH6dWBXH6YjH‡—˜UT¢2pYWBX5¢295V£ q º–ÝxíW5 ü 4ƒÛ º–ݘí‚ü54 ä º‡Ýxívø“ºkYxæÔ¦sçhY  Ý  Ý ¦ ø~ºkYxæf¦cç  Y î…ß  Y‚V  ßpW í  ß  W í ¦    „Ý   …Ý ¦ ¦ øºkYxæf¦sç „ Y … î ß „ Y V ø“ºÉo Y æf¦sç … Y î…ß … Y V ø YaYaY«æz¥p’ ¥8“oç `  ß „ W í  ß … W í “   — ¦ î…ߘY‚V i tmTcixyE58d™DƒTV‰IyXH…BEYYWBE7#pyEi†TŽ4Q582ob¢HQ2IBEHQ2oDu@ŠTgie5¢qˆq•58d‚7#BE2I`dyEBX2IHQT¢d™46587…‰IBE2UTVDWBX582IYf5¢q€tvTgixyX5¢d YWH6dWBEHQYC582†T‘DWHQd‚7 ‰oieDGHQd‚7 ‰UT¢Y‚BXY6’ .

ÎxÐ%çxÐóÇu×uәژËmÍÏè•Ç‘˧ÒÔÓ{é ֐ÒÔєє̈Íf̈ÓwÇvÒÔÓqé ڋǔÈfÌvنÌvÚ×uѬəÓëê ×uÍfÖÌˆÍ ¥8¥  582IY‚BXhIH6dwDWFIH–YWRI7 5VqŠHgTV4ƒF5¢qvDGFIH6YWH94Q58yERI792IYQ’‘t§5†7#Tl—˜BX79BXŒ6HaDWFIH#5¢dWhIH6d5¢quT¢464QRIdWT¢46i • 5¢qmDGFIHw7#H‡DGFI5xh‹tU@vHwId‚5˜46HQHQh‰q•dW587µyEH6q|DCDG59dWBE`8FoDfT¢2Ih†q•5¢dW4QHVtë‰SiDWFIHapdW58†HQd‘4ƒFI58BE4QHw5¢qjTpt뉋t T¢2IhA4¢t›DWFIHQY‚H#Y‚RI7#YuDG5‰†H–ŒQHQd‚5I’#“2IH#4QT¢2AHgT¢Y‚BXyEiYWFI5c@ÏDWFUTVDDWFIH–YWRI79Yu5¢q•DWFIH‘nUdWYfDwDWFIdWH6H 4Q5¢yXRI792IYCT¢d‚HŒQHQd‚5#B„qm@vHuYGTVDWBXYfq|iŽDGFIHuH6PxRITVDGBE582 ‹ ¶· ³´µ 6w¦ ¡ ¦ ¹ ´µ³ Œ ¹ ä ¦ ¡ ¦ º tuFIHwYW58yERpDGBE582eBXYu`¢BEb8H62‰Siùý‹Uौcॺƒûsäˆ6wù ¦8àS6   àQ¦QûŸ’ ¦ ¦ ¦ ¶· ¡ ¶· ³´µ 6   ¹ ¡ tuFIHw4Q58yERI7#2pYŠDGFUTVDChp5#2I5VDCYWRI7§DW5#Œ6HQd‚5#46582IY‚DWBEDWRpDGHDWFIHuHQdWd‚58dQ’ ™HahIH6YWBE`82UTVDWHuDWFIHunUdWYfD™2I5¢2SbVT¢2pBXYWFpBX2I`9YWRI7 BED‘TVYŠDGFIHutmTcixyE58dCYWHQd‚BXH6Y‘H6dWd‚58dQ’ DW5d‰†H½þxÿ  t†T¢2Ih‰dWH6q•H6d‘DW5 132©DWFIBXY94gTVYWHðþxÿ  5x4Q4QRpdWYŽTlD9DWFIH†npq|DGF«4658yXRp7#2ÆBX2¬DGFIHDƒTV‰IyXHæ•q•5¢d9DWFIBXY–H‡—pT¢79IyEH†TVyXyuH6b¢HQ2 4Q5¢yXRI792IYŠ@CBXyEysbVT¢2IBEYWF‰oiY‚ix797#H‡DGd‚ipçuT¢2Ih582pHunI2IhIY þxÿ  ä ¦ ‹ ø î ß ¡    — …   —£¢ º î…ß …5V  …Ý  ß … W í ä î‘ß   … Ý V  ß … W í ¦   æz¥p’ ¥398ç òf5VDGHuDGFUTVDCDWFIH…î…ß  FIT¢Y™‰†HQH62hIB„b˜BEhIHQhDGFpdW58RI`¢F†DG5#7dTV÷¢HDGFIHaHQd‚dW58duDGH6dW7µ46582IYWBEY‚DWHQ2oDg’u™H FUTgb8HÛ‚RIY‚DuhpHQdWB„b8H6hŽDGFIH‘qšT¢7#BEyXBXT¢dj¥lr£s58BE2oD•46HQ2oDGdWT¢yhhIB„ghH6dWH62I4QBE2I`w†58BX2oD•58sH6dGTVDW58dˆq•58dŠTwY‚HQ46582Ih hIH6dWBEbVTVDWBEb¢H ! Ý ¦ 6 æ•Ý˜¥ í €4‚6   Ýxí”ø¬Ý˜í‚5 ü 4fçˆ× ä ’dæzî…ß  ç æz¥p’  ¥ coç V  ß  W í î…ß  t FIHtmTcixyE58d•DƒTV‰IyXH‘q•58duT…¥lr£s5¢BX2oDŠ‰UTV4ƒ÷S@•T¢d‚h˜r3hIB„ghH6dWH62I4QBE2I`‘58†HQdGTlDG58dŠd‚HQId‚HQY‚HQ2oDGBE2I`9TanUdWYfD u hIH6dWBEbVTVDWBEb¢HBXYuYWFp5s@C2†BX2tmT¢‰IyXHu¥˜’   ’ V Ýxí î…ßfY t ux w y í u  Ý  ß W í ø ¦ æ ‹ ÝxíA€ ø“‹¡4fÝxíA€4hø~ŒƒÝ˜í6çˆäfù –   î ß … î…ß  Y t ¦ u ú¤ w y u ú í î…ß „ Y t A u û˜ w y u û í î…ß … Y t A u ü˜ w y u ü í î‘߁Y t x u w y í u ¦ ‹  YcÝxí¥€  ‹¡47YcÝxí¥€4 Œkc Y Ýxí ‹  ‹ 4  Œ ‹  YSæƒr   çUY 4 4 ‹¡4'YSæƒrƒ¦sçUY 4 4 4 ‹  Yxæ|r   ç  Y 4 ‹ 48Yxæ|rƒ¦cç  Y  ¡  4 ‹  Yxæƒr   ç „ Y Ý 4 ‹ 4'YxæƒrG¦sç „ Y Ý ¡ 4 ‹  Yxæ|r   ç … Y 4 ‹ 47Yxæ|rW¦sç … Y  … ¡ … tmT¢‰IyEHk¥p’   ’ hpHQdWB„bVTVDGB„b8H¢’ tvTgixyX58dŽDGT¢‰IyEHq•58de‰UT¢4ƒ÷o@•T¢dWh ¥lr3†58BX2oDô˜T¢`¢dGT¢2I`¢BƒT¢2ªT¢pIdW5g—˜BX7#TVDGBE582«DW5©TÅnUd‚Y‚D .

¥¢— ǔȑɐʙËm̈ÍÏÎSЫÑvÒÔӑÒQËm̀Õz֐ÒÔєє̈͑̀Ówǔ̫ɐʈʈ͏×jؐÒfنə˧҇×uÓÚ DG5–Œ6HQdW59BEq tuFIBEYCDWBX79H™DGFpHunIdWY‚DCDWFIdWH6Hw4658yXRp7#2IYuY‚RI7 ³´µ   — ¦ ¦ ¦ ¦ ‹  ¶· ‹¡4 ä ¡ ¹ ´µ³ Œ ¹ ¡ ¦ ¶· ¡ ¶· ´µ³ ¦ ¹ ¡ @CFIBE4ƒF«`8B„b8HQYù¸‹ àˏ4‡à̃ûuä 6 ù ¦8àS6f—IàG¥VûŸ’«132©DGFpBXY94gT¢Y‚H‰DWFIH†q•58Rpd‚DGFÆ4Q5¢yXRI792©pdW5cbxBXhIH6Y#DGFpH   yXHQT¢hIBE2I`…DGd‚RI2I4gTlDGBX5¢2‰H6dWd‚58dŠDGH6dW7‰Û 4 þxÿ  ä ¦ î ß …   6 „ Ý c‹  ‹¡4 6 î‘ß „ V ä  ß „ W í “ “ ¢ î…ß  „ Ý V  ß „ W í ¥ æz¥p’ ¥hoç tuFSRIY–@•HFUTgb8H‰hIH6dWBEb¢HQh©TAY‚HQ46582Ih˜r£58dWhIH6d–‰UT¢4ƒ÷o@ŠTVdWh«hpBEgsHQdWH62I4QH‰T¢pIdW5g—˜BX7#TVDGBE582)5¢q™TnUdWYfD hIH6dWBEbVTVDWBEb¢H¢Û V –—ˆßõ—o‚  Ý  ß W í 6 ¦ æšÝxíA€ 6ŗ8Ýxí¥€4hø©¥¢Ýxí‡çjäf’Žæšî…ß  ç  î ß   … æz¥p’ — ¡ ç ¤×p8iupsr|ã£g¦¥v|7jg†ekiæe¨§Dlfg†n{p‚rp‚iutvp©ueUr£Çâã†|}    Ý V ø  ß   W í 132A`8HQ2IH6dGT¢yzt˜TehIBEgsHQd‚HQ2I46HdT¢pIdW5g—˜BX7#TVDGBE582DW5DWFIH–þŽDGFAhIHQd‚BEbVTVDWBEb8H–TlDu`8dWBEhA†58BE2SDuê4gT¢2A‰†H 4gTVY‚D‘BX2eDGH6dW79Yu5¢ q C ø …øð¦u2IH6BX`8FS‰†58dWBE2I`‘s5¢BX2oDGYCT¢Y  ¾ €‘Ü  ‹  Ýxí‚ü  äfþxÿ  æz¥p’ —I¦sç @CFIH6dWHdDWFIH ‹  T¢d‚Hd465˜H‡¨#46BXH62SDWYwDG5‰sH#hIH6DWHQdW79BX2pHQhrDWFIdW58Rp`8FrDGFpHŽRIYWHd5¢qutmTcixyE58dwtmT¢‰IyEHQYaDW5 Id‚5˜hIRp4QHeT¢pIdW5g—˜BX7#TVDGBE582IYu5¢q‘T`¢BEb8H62Å58dWhpHQdQ’•yXHQT¢dWy„iADWFIBXYaIdW5x4QH6YWY…4QT¢2ʼn†HdRIY‚HQhÅDG5‰nU2Ih q•58df@ŠT¢d‚h‹tS‰UT¢4ƒ÷o@•T¢dWh‹txYW÷VH6@vHQh‹tp58dj4QH62SDWdGT¢y끆58BX2oDj5¢sH6dGTVDW58dWYˆ5¢qhT¢2oi‘58d‚hIHQdjq•58dˆT¢2oi‘hpHQdWB„bVTVDGB„b8H¢’ 1£Dm4Q58RIyEhw‰†H•46587#pRpDGH6dvT¢R˜DG587#TVDGH6hwTV2IhwH–—˜DWHQ2IhpHQh‘DG5fFIBX`8FpHQd˜hIBX79HQ2pYWBX5¢2IYQ’ †58d‚H•BX79†58d‚DGT¢2oDgt FI5c@vH6b8H6dQtvBXY…DGFpHŽqšT¢46D…DWFUTVD–B„D…4gTV2Ɖ†HŽq•RIdfDGFIH6d‘`8HQ2IH6dGT¢yEBXŒ6HQh‹’A132¬58dWhIH6d…DG5ehI5eDWFIBXY6tjyXH6D…RIY T¢IpdW5oT¢4ƒF¬DGFpH†Y‚RI¡ ‰ ۂHQ4‡DdBE2ÆTAY‚yXBX`¢FSDWyEihIBEgsHQd‚HQ2oD–@ŠTgi¢t•DGFUTlD–BXY‘q•dW587¾DGFIH†58BX2oD–5¢q‘bxBXH6@§5¢q BX2oDWHQdW†58yXTVDGBE582Åq•58dW7…RpyƒT¢Y6’ÿtuFIH6YWHq•58d‚7…RIyXT¢Y9T¢dWH†hIBXY‚4QRIY‚YWHQhªBE2«7#T¢2oiÆDWH‡—xDGYd582º2SRI79HQdWBE4gT¢y T¢2UTVyEixYWBEYQ’ –—ˆßõ—˜– !Ó|#"hr|i$"Up¼|iuØ&%Ùpsr£ goj¡p'ÂiÉj¡psr!y{ehã£|7jg£ehiædfehã†÷õiue¨#g†|ã£} tuFIHaô˜T¢`8dGTV2I`8BƒTV2ŽBX2oDGH6dW†58yƒTlDGBX5¢2Žs5¢yEix2I5879BƒT¢y†BXYu`¢BEb8H62†‰oi Ýmæ•ßsçjä ý ¾ ä ( ‹Sý8æ•ßsçÔ݆ý æz¥p’ —   ç .

ÎxÐ%çxÐóÇu×uәژËmÍÏè•Ç‘˧ÒÔÓ{é ֐ÒÔєє̈Íf̈ÓwÇvÒÔÓqé ڋǔÈfÌvنÌvÚ×uѬəÓëê ×uÍfÖÌˆÍ ¥3$ @CFIH6dWH‡‹Sý8æ•ßsçuT¢d‚Hw†58y„i˜2p587#BXT¢yXY€BX2eße5VqvhIH6`8dWH6H0))’vtuFIHu46582IY‚DWdWRI4‡DGBE582†5VqvDGFpH‡‹oý8æ•ßsçŠ4gT¢2‰‰†H DƒTV÷¢HQ2‰q•dW587 DGFIHwYWBE7#IyEHuô˜T¢`8dGTV2I`8BƒTV2q•5¢dW7…RIyXTwq•58dCPSRUT¢hIdWTVDGBE4wBE2oDGHQd‚s5¢yƒTVDWBX582Aæ•58duH‡—xDGdWT¢†58yƒTlr DGBE582ëç€@CB„DGF‰2I582˜r£HQPSRIBEYWUT¢46HQh‰s5¢BX2oDGY æ•ß£476rßsçQæ|ß 6rßsç æ ß!äU6)ßsçQæ|ß 6)ßsç •   ø¬Ý£4 æ•ß£476)! ß ä‡çQæ|ß r 6 ß!ä‡ç æ•ß!U ä )ߣ4‚çQæ|ß 6)ߣ4Wç 6   æ•! ß äU6)ßsçQæ|ߣ4'6)ßsç øÝ  æ•! ß äU6)ß çQæ|£ ß 4'6)ß ç Ýmæ|ßsç€äªÝ!ä   æz¥p’ —o¥oç òf5VDGBX46HdDGFITVD…DGFpHe4Q5xH6¨94QBEHQ2oD–5¢q‘HgT¢4ƒFÆÝëýdBEY…582IHŽ@CFIH62Æß ä ßëý¢t€T¢2phŌQHQd‚5e@CFIHQ2¬ß©DGT¢÷¢H6Y T¢2oi†5¢DGFIH6dhIBEYW4Qd‚H6DWHwbVT¢yXRIHaBX2‰DGFIH…Y‚H6DQ’1£qj@•H…DGT¢÷¢HwDWFIHwnUd‚Y‚D™58dYWH64Q582phehIHQd‚BEbVTVDGB„b8Ha5¢qˆÝmæ•ßsç‡t BX79†58YWHdT¢2ÅHQPSRIBEYWUT¢46HQh¬79HQYWFhtvTV2IhÅH6bVT¢yERUTVDGH#DGFIH6YWHehIH6dWBEbVTVDWBEb¢HQY‘TVDwDGFIHT¢pIdW58pdWBƒTlDGHdhpBXYfr 4Qd‚H6DWHes58BE2oDgtv@vHdWHQhpHQdWB„b8HŽDWFIHdnU2IB„DGHŽhpBEgsHQdWH62I4QHTVIIdW5g—˜BX7#TVDWBX582IÉ Y ۂRIYfD‘IdWH6YWH62SDWHQh‹’)¤vBX2IB„DGH hIB„ghH6dWHQ2p4QH9YW4ƒFIH67#H6YaDGFITVDw4gT¢2A‰sH9hIH6dWBEb¢HQhAq•dW58å 7 jP뒧¥p’ —   T¢dWH–dWH‡q•HQd‚dWHQhDG5‰T¢Yaô§TV`8dGT¢2p`8BƒT¢2 T¢IpdW5g—˜BX7#TVDGBE582IYQ’ iª`8HQ2IH6dGT¢yEBXŒQTVDGBE582‘5¢qhDWFIH‘ô˜T¢`8dGTV2I`8BƒTV2#TVIIdW58T¢4ƒF#BEYˆ‰IdW58Rp`8FoD•TV‰s58R˜Dˆ‰Si–RpYWBX2p{ ` €‘HQdW79BEDWBƒT¢2 BX2oDWHQdW†58yXTVDGBE582‹’˜tm5a4Q582pY‚DGd‚RI46D•Tus58y„ix2I5879BƒT¢ySq•58dˆÝmæ•ßs燡 t €‘HQdW79BEDWHŠq•58dW7wRIyƒT¢YjRpYWHCb¢TVyXRIH6Y€5VqhDGFpH q•RI2I4‡DGBE582e¢„†ˆ…y|x|žwˆx}‡€–2 y 1l¢xz2 y 1¢4 } 3z6 ž 5uTVDv`8BEb¢HQ2d†58BE2SDWY•BE2#Y‚UT¢4QHV’v“RpdvBEyXyXRpY‚DGdWTVDGBE582‘BXYˆq•58dˆDGFIHf4gT¢Y‚H BX2k@CFIBE4ƒF)hIBXY‚4Qd‚H6DGH–bVT¢yXRpHQYw5¢qŠDGFpH9q•RI2p46DGBE582)TV2IhrB„DGYunUdWYfDwhIH6dWB„b¢TlDGBEb¢HdTVdWH9RIYWH6h‹t”Id‚5˜hIRp4QBX2p` DGFpHwH–—˜IdWH6YWYWBE582 Ývæ|ßsçjä  ‹oýoæ•ßsçÔÝëý€ø  Œƒýoæ•ßsç V  Ý  ß W ý æz¥p’ —8—Sç “ ‰ob˜BE58RIY‚yEi…FpBX`8FIH6dfr£58dWhIH6d”hIH6dWBEbVTVDWBEb¢HQY”4658RIyXh–‰†HuBX2I46yXRIhpHQh–T¢Y”DWFIHuIdW5¢‰IyXH67#Y”hpBX46DGTVDGHV’vi 46587–r  IyEH6DGHwhIBXY‚4QRIY‚YWBX5¢2e5¢qjDGFIH6YWH‘†58yEix2I5¢7#BXT¢yXYC4gTV2e‰sHwq•5¢RI2IheBX27dT¢2oi†dWH‡q•HQdWH62I4QH6Yu5822SRI79HQdWBE4gT¢y 79H6DGFp5˜hIY6tU‰IRpDCFIH6dWHu@•Hu2IH6HQh†582IyEi#DGFIHu46582I4QH6pDg’ tuFIHwpdWH6bxBE58RIY™H–—pT¢7#pyXHQY‘5¢q”T#tmTcixyE58duDƒT¢‰IyEHa46582IYfDGdWRp46DGH6h©s } ‰6œ† éyŸ‡y|x˜s5¢BX2oD‘hIBEgsHQd‚HQ2I46Hw58˜r HQdWTVDG5¢dWYq•dW587pô˜T¢`8dGTV2I`8BƒTV2BX2oDGH6dW†58yƒTlDGBX5¢2q•58dW7wRIyƒT¢Y6’–•582IY‚BXhIH6du2IH‡—xDuDGFIH–tmTgi˜yE58dDƒTV‰IyXHwq•5¢d T¢2Æy|{™œ† éyŸ‡y|x”YWIT¢4QH9hIBEgsHQd‚HQ2I46BX2I`Y‚4ƒFIHQ79H9q•5¢duTnUd‚Y‚DwhIH6dWBEbVTVDWBEb¢H–T¢dWBEYWBX2p`Žq•dW587pDWFIH–RIYWH95¢q•T €fH6dW79BEDWH‘BX2oDGH6dW†58yƒTVDWBX582‰q•58dW7wRIyƒTpq ’ &ŠT¢YWH6h5¢2 jP뒧¥p’ —8—It˜Te`8HQ2pHQdGTVyXBXŒQTVDGBE5825¢k q jPë’m¥˜’ —I¦–4QT¢2 BX2p4QyXRphIHhIHQd‚BEbVTVDWBEb8H6Y‘TVDC2IHQBE`8FS‰s5¢dWBX2p`‘s58BE2oDGYQtpBŸ’ H¢’„t    Ý Œ V ø   € 8   ß W ‚ íü  ¾ 9 7  ¾ €‘Ü  ‹  ÝxíWü  ä`þxÿ  æz¥p’ —Â$8ç T 2UTVyX58`85¢RIYmDW5QˆPë’o¥˜’ —8—p’mif29H–—pT¢7#pyXHvq•5¢dW7…RIyXT™BXYmBXyEyXRIYfDGdGTlDGHQhwTVD”DWFIHŠDG58‘5VqëtvT¢‰pyXHu¥p’ ¥p’7€fHQd‚H ¢ 2I5¢D‘582pyEiŽBEYŠDGFIHuhIH6dWB„b¢TlDGBEb¢HuTVD‘s5¢BX2oD€ê#d‚HQId‚HQYWH62oDGHQhhth‰IRpDfT¢yEYW5‘BE2I4QyERIhIHQh†T¢d‚HwhpHQdWB„bVTVDGB„b8HQYCTlD †58BX2oDGY”u ê 6ƦuTV2Ih–êCø«¦8t˜@CFpBX4ƒF‰T¢yXY‚5…7…RIYfDu‰†HH‡—˜UT¢2IhpHQhRIYWBE2I`‘tmTcixyE58dC”˜H6dWBXH6YuT¢‰†58RpDu†58BE2SD .

¥8“ ǔȑɐʙËm̈ÍÏÎSЫÑvÒÔӑÒQËm̀Õz֐ÒÔєє̈͑̀Ówǔ̫ɐʈʈ͏×jؐÒfنə˧҇×uÓÚ êU’atuFpBXYdWH6PSRIBXd‚HQYDGFpH…q•58yXyE5s@CBE2I`d`¢HQ2IH6dGT¢yEBXŒgTlDGBX5¢25¢qˆDGFIH‘tmTgixyX58dY‚HQd‚BXHQYH–—˜UT¢2IY‚BX582k`8B„b8HQ2kBE2 jPë’ë¥pd ’ c˜Û ï    Ý    Ý ¦ ï   V V æŸõ˜î‘ßsç äA@  ß   W í‚üIý  ß ï Ê  ß   E W D í ÈC ï B ¾ ä ì7b æz¥p’ —o“oç tuFIH‰hIHQd‚BEbVTVDWBEb8HŽDWHQd‚7#Y‘2I5c@§FUTgb8H‰4Q5xH6¨94QBEHQ2oDGYæ•DWFIH4Q5xH6¨94QBEHQ2oD9582ÆDWFIH9ꁆ58BX2oD–BEY…DƒT¢÷VHQ2 T¢Y–582IHDG5AYWBE7#pyXBEq|ikDWFIHTVyX`8H6‰IdGToça@CFpBX4ƒFÆ7…RIY‚D#‰†H‰hpH6DGH6dW79BX2IH6h©RpYWBX2p`ADWFIHtvTgixyX58d–tmT¢‰IyEH T¢IpdW5oT¢4ƒF†T¢YC58RpDWyXBE2IHQhe‰†HQyE5s@u’ " V  Ý ø  ßW íA€4 V Ýxí  Ý  Ý ø“þ V ø  ßW í  ßW W í ü54 î…ß  Y t ¦ u ú¤w y í u ú Yxæ|rW¦sçhY 44 " ¦ 戋SÝxí¥€4›ø~ŒƒÝ˜ívøzº–ݘí‚ü54fçjä2ù î ß … î…ß „ Y t A u û¤ w y u û í Yxæ|rW¦sç  Y 4 "  î…ß … Y t A u ü˜ w y u ü í 4 Yxæ|rƒ¦cç „ Y Ý " î…ß † Y t G u F¤w y í u F YxæƒrG¦cç … Y 4 " … î‘ߘY t uxw " u  y íA€4 u ! w y î‘ߘY tx u í î…ߘYqþ!tvux w y u í‚ü54 ‹‡c Y ÝxíA€4 ŒhYcÝxí ºkc Y ÝxíWü54 ‹ î…߁Y t x u w y u í " ¦ þ Œ º ‹{YSæ|rG¦sçhY 4 ºhYxæf¦sçhY 44 4 4 þ Yxæf¦cçhY 4 p 4 ‹‡Yxæƒrƒ¦sç  Y  4 ºkx Y æf¦sç  Y  4 þ YxæÔ¦sç  Y X 4 ‹½Yxæ|rW¦sç „ Y Ý ºkYxæÔ¦sç „ Y Ý 4 4 þ Yxæf¦sç „ Y Ý p 4 ‹‡Yxæ|rW¦sç … Y … 4 ºÉS Y æf¦sç … Y … 4 þ YSæf¦sç … Y X  …4 ‹{YSæ|rG¦sç † Y 4 ä  4 ºhYxæf¦sç † Y 4 ä  tmT¢‰IyEHu¥p’ ¥p’ˆtvTgixyX58dŠDGT¢‰IyEHq•58du4QH62SDWdGT¢y˜¥sr3†58BX2oDp€‘HQdW79BEDWBƒT¢2ŽT¢IId‚5c—˜BE7dTVDWBX582ŽDW59T‘nIdWY‚D‘hIH6dWB„b¢TlDGBEb¢H¢’ t§5#7#Tl—˜BX79BXŒ6HfDGFpHw5¢dWhIH6d‘5Vq”T¢464QRIdWT¢46i¢tI@•Hw7…RIYfD‘YGTVDWBXY‚q|i#DGFIHud‚HQyƒTlDGBX5¢2 ¦ 6 ¦ w ¦ 6 ¦ w ¦ ¡ ¡ ¡ ¡ ¦ ¦ ¦ ¦ ¦ ¦ ¡ ¦ ¡ ¶· ‹ ¶· · · ¦ ·· Œ ·· ´³   · ´´ º · ä — ¥ ¹ ¡ 6 ¦ w ¡ ¡ ¡ ³´´ ´´ ´µ 6   ¥ 6 — ‘ ´´µ ³´´ ´´ ´µ ¶ ·· ·· · ¹ "þ ¹ F TgbxBX2I`‰DGFIH#YW58yERpDGBE582«ù¸‹Uौcॺsà à¥þQû•ä U " 4Q5¢yXRI792eYWRI79YuDG5 … 4 ù ¥pà ¡ àx6f¥pàg¦8àg¦6ûŸ’IH‘2IhIH6dwDGFIH6YWH#4Q582IhpBEDGBE582IYuDGFIH#YWB —˜DWF î…ß … † Ý V  ß † W í ¦  V¡ þxÿ  ä æz¥p’ —Â98ç .

ÎxÐ%çxÐóÇu×uәژËmÍÏè•Ç‘˧ÒÔÓ{é ֐ÒÔєє̈Íf̈ÓwÇvÒÔÓqé ڋǔÈfÌvنÌvÚ×uѬəÓëê ×uÍfÖÌˆÍ ¥39 T¢2IhDGFpHw79H6DWFI5xh‰4QT¢2‰sHuH–—ppdWHQY‚YWH6hkT¢Y V  Ý  ß W íA€4 ø¬—V  Ý  ß W í ø V  Ý  ß W í‚ü54 6 ¥ æ†6fÝxí¥€4›ø¬ÝxíWü54fçˆäf’dæzî…ß…‡ç î ß ‘ æz¥p’ —3coç tuFIBEY‘BEYCT¢yXY‚5‘dWH‡q•HQdWd‚HQh‰DG59T¢Y‘TQPuoˆS}€ R q•58d‚7…RIyƒT˜’ –—ˆßõ—ˆß d r|tÚjg†tv|ãQT`yyãsg†tv|7jg†ehiie¨§Ddf|ØVp U ©ueUr£Çâã†|} 1£DfBXYC582IHuDGFIBE2I`–DG5d46582IYfDGdWRp46Df7#H‡DGFI5xhIY‘RIYWBE2I`#DWFIH{€‘HQd‚7#B„DGBƒTV2e4Q582I46HQpDT¢2Ih†PxRpBEDGHaT¢2I5VDGFIH6d DG5wBX79IyEHQ79HQ2oDvDWFIHQ7 BE2#Tu4Q5¢7#IR˜DGHQdˆ465˜hIHV’”132–DGFIHuq•58d‚7Ï5¢qhTw†58BX2oDˆ58†HQdWTVDG5¢djBEDˆBXYjpdW58‰UTV‰IyEi 2I5¢DCH‡bxBXhIH62SD‘TlDŠnUdWYf' D ۂRIY‚DCFI5c@fjP†’I¥p’ —c…4gT¢2‰†HuT¢IIyEBXHQhh’€f5c@vH6b8H6dQtëDGFIHY‚BEDWRUTVDGBE582ŽBEYŠPSRIBEDWH HgTVY‚idDW5…4Q5¢7#Id‚HQFIH62IheBEqh@•HH–—ppdWHQY‚YuDGFIHYWT¢79H™79H6DGFp5˜hŽBE2dDGFpH™q•58dW7 5¢q§T…7#TVDWdWB„—95¢sH6dGTVDW58dQ’ i ‰UT¢2IhpHQh‰7dTVDWdWB —Ž2I5VDƒTVDWBX582Žq•58p d jP뒆¥p’ 3 — c‘BXY ¦ “ ÁeæÔ¦8àG—Iàg¦cçs(  e ݲ ä ¦ ² ø ÁŽæ£6w¦8à ¡ àg¦cç Ýw î ß   ‘ t ŒA² º y æz¥p’ —3hoç BX2‰@CFIBX4ƒFkãBEdWBE4ƒFIyXH‡Df‰†58RI2phUT¢d‚i‰46582IhIB„DGBX5¢2IYfFITcb¢H…‰†HQHQ2eBE7#†58Y‚HQh‹’ „   TVDWFIHQ7#TVDWBX4gTVyXyEidDGFIBEYfBEY HQPSRIB„bVT¢yXH62SDuDW5 €4 ² ð (  ݍ ä “Uù¸ÁŽæf¦8àG—pàg¦sçÔû   ¦ ² Ç Áeæ£6w¦8à ¡ àQ¦sç Ýw ø t ŒA² º y î ß   … ¢ æz¥p’d$ ¡ ç @CFIBE4ƒF†4QT¢2†‰†HudWH6H‡—˜IdWH6YWY‚HQh‰oieDGFpH  IdWH6hIBX4‡DG58dÔr34Q5¢dWdWH646DW58dŽ–YWH6PSRIHQ2I46H Ý ²W ä ä ¦ ² Ç ÁŽæ£6w¦8à ¡ àg¦cç Ýw ø t ŒA² º y î ß   … €4 “fý ù ÁeæÔ¦8àG—Iàg¦cçÔû Ý ²W (  ݲ æz¥p’d$˜¦sç ™ BEDWFÅdWHQY‚sH646D–DG5IdGT¢4‡DGBE4gT¢yvBX79IyXH67#H62SDGTVDGBE582‹t§DGFIHŽ79HgTV2IBX2I`5¢qCDGFIHeId‚HQhIBE46DW58d…BX2DGFIBEY ª ² tChIB„ghH6dWH62I4QHB„Dgt YWH6PSRIHQ2I46HkYWFI58RpyXhº‰†H4QyEHgT¢d6’ 1£DdYWBX79Iy„i¬YGTgixYYX¬DƒTV÷¢HDGFpH†b8H646DW58dT¢dWdWTgi ݛ ²W ’™tuFIH T¢hIhDWFIH…‰†58RI2IhIT¢d‚i‰4Q5¢2IhIBEDWBX582pYQt‹TV2IhY‚DW58dWHaDGFIH…d‚HQY‚RIyEDBE2†DWFIH…BX2oDWHQdW79HQhpBƒTVDWH…T¢d‚dGTgi ݧ 79HgT¢2IBE2I`…5¢q§DGFpHYWHQ46582Ih‰dW5c@ÿBXYŠ7958dWH™YWRI‰pDWyXHVtUYWBX2p4QHBEDŠBXYŠhIH67dTV2IhIBX2p`…DGFIHH‡b¢TVyXRUTVDWBX582d5¢qmT¢2 BX2ob¢HQdWY‚Hw58†HQdWTVDG5¢dQtU‰IRpDfBEDCYfDGBEyXy‹4QT¢2‰†Hu`8BEb¢HQ2T–YWBE7#pyXHBX2oDWHQdWpdWH6DGTVDGBE582‹’€i™2†BX2ob8H6dWYWHw7#TVDGd‚B„— 58†HQdWTVDG58dwBX79IyEBXHQYDWFIHdY‚58yXR˜DGBX5¢25¢quT4Q58RpIyXH6hrYWH‡D‘5¢quyEBX2IHQT¢duHQPSRUTVDWBX582pYQ’‰tuFIHQY‚HŽ58†HQdGTlDG58d‚Y T¢d‚HCb¢HQdfi#4658797#582–BE2–nI2IBEDWH‘hpBEgsHQdWH62I4QHfT¢IpyXBX4QTVDGBE582IY6’mtuFIH‡i#T¢I†HgT¢dˆBX2–DGFpHCq•5¢dW7Ï5¢qh‰UT¢2IhpHQh 7#TVDGd‚BX4QH6Y#FITcbxBE2I`)TrY‚7dTVyXyu‰UT¢2php@CBXhpDWF‹tuBX2ÆDGFIBEY#4QT¢YWHTADGd‚BXhIBXT¢`8582IT¢yŸ’ºtuFpHH6bVT¢yERUTVDGBE582¬5¢q €4 ýù Áeæf¦8àW—Iàg¦sç3û BEYwq•58RI2Ihʼnoir7#HQT¢2IY‘5¢qfTDGd‚BXhIBXT¢`8582IT¢y  YW58y„b8H6 d ŽItj@CFIBE4ƒFÆBEY…YWBE7#pyXH–DG5k4Q5xhIH¢t H6¨946BXHQ2oD‘DG59dWRI2htsTV2Ih†@CBEhIHQy„ieRIYWH6h‹’Š132†`8HQ2IH6dGT¢yz! t €fH6dW79BEDGBXT¢258d‘ÜvT¢¨ h Hw ` T¢pIdW5g—˜BX7#TVDGBE582IYu4QT¢2 ‰†HwpdGT¢4‡DGBX4QT¢yh@CFIH62†DWFIH6ie4QT¢2‰‰sHuBX79IyEHQ79HQ2oDGH6he‰SidDGFIHuRIY‚Ha5¢qmH6¨946BXHQ2oD‘‰UT¢2phIHQh†YW58y„b8HQd‚YQ’ a¦b ¬§“¨‘žd©8ª£­¦©“›§¤¨q›¥¦›ªo§¤«¦¢‚ª£«¦¬S§¤­¦žÕ¬qžd¬dcɧ¤¨q›fev«¦Òq¬ q­A¢¤¯1ª£«¦¬‘ qž¸§¤žÕ«F¬‘©vœÚ«¦ÒqŸ% užÕ¬ª†Ÿdґ a›U¥¥­¦ŸÕÒq›©s«¦Ñgev«A§¤¨ e «¦§“¨$ e «FÒq¬‘ q­¦¢“žd›s©s° h Ðpi hrq ­A§s Ð ­¦¬‘  .

¥c ǔȑɐʙËm̈ÍÏÎSЫÑvÒÔӑÒQËm̀Õz֐ÒÔєє̈͑̀Ówǔ̫ɐʈʈ͏×jؐÒfنə˧҇×uÓÚ –—ˆßõ—6t w jvuupsrw% gx"¡uupsr€yÂw“rØpsr‚XtƒuupCxp‚} tmTcixyE58dŠDƒT¢‰pyXH¢’ˆ¤U58duH–—pT¢7#pyXH €fH6dW79BEDWBƒT¢2Aq•58dW79Yw5¢quDWFIHŽY‚HQ4Q5¢2Ih¬hIHQd‚BEbVTVDGB„b8HŽ4QT¢2¬T¢yEYW5‰†HŽHgTVYWBXy„iAhIHQd‚BEb8H6hʼnoi)79HgT¢2IYw5¢q‘T €4 ² å ( ¥ ݎ ä ¦   ù Áeæf¦¢àg¦ ¡ àg¦cçÔû ¦ ² ø t ŒA² º y ÁeæÔ¦8àx6   àg¦sç ݅ æz¥p’d$   ç ¢ î ß   … BXYh’Žæšî…ß … ç€TV2Ihd7#T¢÷¢H6Y•RIY‚H™5¢qh582Iy„i‘DGdWBEhIBƒTV`8582UT¢yp58sH6dGTVDWBX582pYQ’v1£DvYWFp58RIyXh#‰sH‘79HQ2oDGBE582IHQh9DGFUTlD DGFpHfY‚IyXBE2IH™TVIIdW5g—˜BX7#TVDWBX582–BXYˆ582pH‘q•58d‚7 5¢q˜TwÜvT¢¨ h H™ ` 7#TVDGd‚B„—–hIB„ghH6dWH62I4QHf58sH6dGTVDW58dQ’m1£D•BEYv`¢BEb8H62 ‰oi ² äÿ“‘ýù Áeæf¦¢àG—Iàg¦sç3û €4 ¦ Áeæf¦¢x ² ø t Œ¦² º y ( ¥ Ýe à 6   àg¦sç Ýw æz¥pd ’ $¢¥oç ¢   î…ß  ‰IRpDwBEDWYa58d‚hIHQda5¢qCT¢464QRIdWT¢46iBXYw582Iy„i ’Žæzî…ß  ç–f ’ €‘5c@ 7…RI4ƒFrDWFIBXYadWHQhpRI46DWBX582)BE2)T¢464QRIdWT¢46iABEY 5¢gsYWH‡DC‰Si#DGFIHBE2I4Qd‚HgT¢Y‚HQh`8yE58‰UT¢y†4Q582oDWBX2SRIBED3id‰IRIBEyEDvBX2oDG5–T…Y‚IyXBE2IH™nIDuBEYŠ2I5¢DŠ÷x2p5s@C2h’€™H2I5¢DWH DGFITVDaDWFIH9YWIyEBX2IH‘nIDa5¢quTnUdWYfDwhpHQdWB„bVTVDGB„b8H–BXYuBXhpHQ2oDGBE4gT¢yjDG5†T¢2oik5¢qŠDWFIH9H‡—˜IdWH6YWY‚BX582IYwBX2 ˆPSY6’ ¥p’ 3 — c…DG59¥pd ’ $˜¦8’ inI2UT¢y‹@v58dWh†58f 2 €‘HQd‚7#B„DGBƒTV2T¢IId‚5c—˜BE7dTVDWBX582pYQ’v•yXHQT¢dWy„iŽDGFIH‡iFUTgb8HaT¢2TVhpbVT¢2oDƒT¢`8Hu5cb8H6d ¥lr£s58BE2oD‘ô˜T¢`8dWT¢2I`8BXT¢2rY‚4ƒFIHQ79HQY–‰†HQ4gTVRIYWH5¢qCDGFpHQBXdaBX2I46dWHQT¢YWH6h©TV4Q4QRpdGT¢4‡i8’ €‘5c@•H‡b8HQd6tŠT7958dWH YWRp‰pDGyEH…s5¢BX2oD™BEYfDWFUTVD™DWFIH6i†`8H‡D™DGFpBXY™BE2I4Qd‚HgT¢Y‚H…BX2T¢4Q46RIdGT¢4‡iRpYWBX2p`ŽBX2pq•5¢dW7#TVDGBE582eDGFUTVD™BXYfY‚DGBEyXy yX5x4gTVyhDG5‘DGFpHw†58BE2SDu@CFpHQdWHuDGFpHwhpHQdWB„bVTVDGB„b8HQY‘T¢d‚Hu‰sH6BX2I`–H6bVT¢yERUTVDWHQh‹’ˆ132†T¢IIyEBX4gTlDGBX5¢2‹txDWFIBXYC4QT¢2 ‰†H™T¢h˜b¢TV2SDGT¢`8H658RIYvTVD€‰s58Rp2IhUT¢d‚BXHQY€T¢2phdBX2–DGFpH‘bxBX46BX2IB„D3i…5¢qhY‚DWHQHQd`8dGT¢hpBXHQ2oDWYQ’”1£DvBXYˆ58‰obxBX58RpYQtS5¢q 4Q5¢RIdWY‚H¢tpDWFUTVD€nIb8H†58BE2SDvYW4ƒFpHQ79HQYuRIY‚BX2I`wô§TV`8dGT¢2p`8BƒT¢2#T¢IpdW5g—˜BX7#TVDGBE582IYˆ4gT¢2d‰sHfhIHQd‚BEb8H6hŽDGFUTlD FUTgb8H‘DGFIHCYWT¢7#H‘58d‚hIHQdˆ5¢q‹T¢4Q46RIdGT¢4‡i9T¢YˆDGFIH‘7#H‡DGFI5xhIYˆ`8B„b8HQ29BXË 2 jPSYQ’˜¥p’  — cwTV2IhŽ¥pd ’ $   to‰pRpDvDWFIH6i @CBXyEyIFUTgb8H™Tu@CBEhIHQdˆY‚IdWHQT¢hd5¢qhYWUTV4QH™BE2IhIBE4QHQY6’v132#UT¢d‚DWBX46RIyƒT¢d6t8D3@•5wô˜T¢`8dGTV2I`8BƒTV2‘YW4ƒFIH67#H6Yv@CBEDWF DGFpHYGT¢79H™5¢dWhIH6dŠ5¢q§T¢464QRIdWT¢46iŽT¢dWH…æšFIH6dWH™@vHBX`82p58dWH‘DGFIHpdW58‰IyEHQ7 4Qd‚HgTVDWHQhe‰oidDWFIH‰†58RI2IhIT¢d‚i 4Q5¢2IhIBEDWBX582pYQtUT¢y„DGFI5¢RI`8FŽDGFpBXYuBXYu5¢2IHu5¢qmDWFIHuIdWBE2I4QBEUT¢y‹BEYWY‚RIHQYCBX2‰T¢pIyEixBX2p`–DWFIHQY‚HuYW4ƒFIH67#H6Yƒç‡Û  ݲ  ß  Ý ² ­®…„ 6 ¦ ¦   î…ß ² × ÁkÜxæf¦8àS61cpà ¡ à¥cpàx6w¦sç Ýd ä ’ðtWî…ß… y æz¥p’d$V—Sç  ß  6 ¦ ² ` ÁkÜxæ£6w¦8àg¦c“˜àx6™¥ ¡ àg¦c“pàS6a¦sç ݎ ä ’ðtWî…ß… y ¦   ‘ î ß  ¶´‰ˆ¦ó‰²”¹g² æz¥p’d$$8ç ¸eÃëîeÀ€¹6±À‡†óÀˆÀˆÃ9ÀiG 1 25¢dWhIH6dfDG59YWH6yXH646DT–nU2IB„DGH‡r£hIBEgsHQd‚HQ2I46HaY‚4ƒFIHQ79Haq•5¢d™T9`8BEb¢HQ2TVIIyXBE4gTVDWBX582582pHa7…RpY‚Df‰sHwT¢‰IyXH 3 DG5aT¢Y‚YWHQY‚YvDWFIH‘T¢4Q46RIdGT¢4‡i95¢qsDWFIH‘4gT¢2IhpBXhUTVDWHCYW4ƒFIH67#H6YQ’ˆtuFIHfT¢4Q46RIdGT¢4‡i–5¢q›TV2#58†HQdWTVDG5¢djBXYˆ5¢q|DGH62 H‡—˜Id‚HQY‚YWHQhrBX2DGH6dW79Y5¢q€DGFIH‘58d‚hIHQd5VqvDWFIH‘yXHQT¢hIBX2p`eHQdWd‚58d™DWHQdW7phIH6DWHQdW79BX2pHQheq•dW587pTŽtvTgixyX5¢d DƒTV‰IyXHV’v™ªFIBXyEH‘DWFIBXY€BXY•TwRIY‚H6q•RIys7#HQT¢YWRId‚H¢txBEDvIdW5cbxBXhpHQYuTaqšTVBXdWy„i‘yXBE7#B„DGH6h9hpHQYW46dWBEpDGBE582‹’j¤URIdfDGFIH6d BX2˜q•58dW7#TVDGBE582T¢‰s5¢RpD…DGFpHeHQdWd‚58d…‰†HQFUTgbxBX5¢d…5¢qfTnU2IB„DGH‡r£hIBEgsHQd‚HQ2I46HŽYW4ƒFpHQ79He4gT¢2¬‰†HŽ58‰pDGT¢BX2IH6h RIY‚BX2I`9¤U58RId‚BXHQduH6dWd‚58dCT¢2UT¢y„ixYWBXY6’ .

ÎxБSЫр×ÓèmÍCÒÔ̈Íó̈Íf͐×CÍóəÓfɓ’‚êˆÚxÒfÚ ¥h –—6tɗs™ T×yyã£g†tv|7jg†eki…j¡e ” | õyu|7jg†|ãDwzy{psr|7jeUr i™2–TVdW‰IB„DGdGTVd‚iasH6dWBX5xhIBE4€q•RI2p46DGBE582‘4gTV2‘‰sHuhIH64Q5879†58YWH6h‘BX2oDG5B„DGY”¤U58RpdWBXH6dv46587#†582IH62oDGYQt8@CFpBX4ƒF  T¢d‚H‘BE29DWFIHuq•58dW7 þ ‘• to@CFIHQd‚$ H –dBEY”DGFIHC@ŠTgb¢HQ2SRI7…‰†HQd6’€1£DvBXY”DWFIHQd‚H6q•58d‚Hf5Vq‹BX2oDWHQdWH6Y‚D€DG5uH‡—pT¢79BX2pH  FI5c@ @•H6yXyvTŽ`8BEb¢HQ2nU2IB„DGH–r3hIB„ghH6dWHQ2p4QH…58†HQdWTVDG5¢dT¢IpdW5g—˜BX7#TVDGH6Y™hIH6dWBEbVTVDWBEb¢HQY5¢U q þ —• ’™rHw@CBEyXy 4Q5¢2I4QH62SDWdGTVDWH¢t˜FIHQd‚H¢tp5¢2#nUd‚Y‚DŠhpHQdWB„bVTVDGB„b8HfT¢IId‚5c—˜BE7dTVDWBX582pYQtST¢y„DGFI58Rp`8F9DWFIHTV2UT¢yEixY‚BXY€BXY€HQPSRUT¢yEyEi T¢IpyXBX4QT¢‰IyEH™DG5–FIBE`8FIHQduhpHQdWB„bVTVDGB„b8HQY6’  tuFIHwH‡—pT¢46DCnUd‚Y‚D‘hIH6dWBEbVTVDWBEb¢H5¢ q þ ‘• BXY  þ ‘•  䙘6–þ  ß ‘•  æz¥p’d$¢“oç 1 qŠ@vHdTVIIyEi¢t§q•5¢dwH–—pT¢7#pyXH¢tmTYWHQ46582Ih˜r£58dWhpHQdw46HQ2oDGH6dWH6h)hIBEgsHQd‚HQ2I46Hd58†HQdWTVDG5¢duDW5†Ýxí‘ä#þ £ @CFIH6dWHußxíuä«êSî…ߘtI@•Hw`8H6D æˆ(  ݆çší ä ä ä ä ä ä ÝxíWü54‚6ÅÝxíA€4   î…ß ü 4“¤ ‘•…e  íW5 ‘• ¦d t þ ¢ 6 þ ‘•…e  çsþ æ“þ —•Ge  6 þ €   î‘ß  ¥ í €4“¤ ¢ ‘•…e —•  ‘•  æz¥p’d$98ç   î‘ß ¦ ù„æš4Q5¢Yƒ–ëî…ß9øV˜oYWBE2f–ëî…ßsç'6ªæ•4Q58Yƒ–Uî…ß 6g˜oY‚BX2f–Uî…ßsçÔû þ î ß   … 2 –Uî…ß  ˜ Y‚BXf þ ‘• î…ß ˜6–vh¥þ ‘•  @CFIH6dWHS– h BEY§DWFIH•795xhIBEnUH6hw@•Tgb8H62xRp7…‰sH6dQ’jtuFIHv7#5xhIB„nUHQhu@ŠTgb¢HQ2SRI7…‰†HQdjBEYmY‚52UT¢79HQha‰sH64gT¢RIY‚H BEDmT¢psHQT¢dWY˜@CFIHQd‚HvDWFIH€@•Tgb8H62xRp7…‰sH6dQti–‹tVT¢I†HgT¢d‚YmBE2wDWFIHvH–—pT¢46DvH‡—˜IdWH6YWY‚BX582‹’vtuFxRpYmDGFpHvhIH6`8dWH6H DG5f@CFIBX4ƒF‘DGFpHŠ7#5xhIB„nUHQhw@ŠTgb8H62SRI7…‰†HQd€T¢IId‚5g—pBE7dTlDGHQY§DGFIHŠT¢46DWRUT¢y˜@•Tgb8H62xRp7…‰sH6dˆBXY”T79HgTVYWRId‚H 5¢q§DGFIHwT¢464QRIdWT¢46i5VqmDWFIHwT¢IId‚5c—˜BE7dTVDWBX582h’ ¤U58dŽDGFIHeY‚HQ46582Ih˜r£58dWhIH6d†46HQ2oDGH6dWH6h hIB„ghH6dWHQ2p4QHk58†HQdGTlDG58ddDWFIHk7#5xhIB„nUHQhº@•Tgb8HQ2SRI7w‰sH6d†BEY `8B„b8HQ2‰‰oi –vhvä òf5VDGHDGFUTlDj– h T¢IpdW5g—˜BX7#TVDGH6Yf–‰DG5–YWH64Q582ph˜r358d‚hIHQdfT¢464QRIdWT¢46i¢tëT¢YCBXYŠDG59‰†HH‡—˜sH646DWHQh‹t†YWBE2I4QH Y‚BX2f–Uî…ß î…ß äk–Ó6 YWBE2f–ëî…ß î‘ß æz¥p’d$vcoç – „ î‘ß  “ ø ‡S‡x‡ jPSRUTVDGBE582ÿ¥p% ’ $c¬BEYIyX5¢D‚DGH6hðBE2 ¤mBX`8RId‚Hk¥p’ —ItT¢yE582I`Å@CBEDWFÿY‚BX79BXyXT¢d#d‚HQyXTVDGBE582IYŽq•58deDGFpHAYfDƒT¢2xr hUT¢d‚hdq•58RIdfDGF˜r£58dWhpHQd€4QHQ2oDWHQdWH6hhIBEgsHQd‚HQ2I46HYW4ƒFIH67#HT¢2IhdDWFIHfq•58Rpd‚DGFxr358d‚hIHQdvÜvT¢h¨H ` YW4ƒFIH67#HV’ˆtuFIH .

5 th 4 Pade 4 Central nd 1 0.— ¡ ǔȑɐʙËm̈ÍÏÎSЫÑvÒÔӑÒQËm̀Õz֐ÒÔєє̈͑̀Ówǔ̫ɐʈʈ͏×jؐÒfنə˧҇×uÓÚ 3 2.5 2 Central 0 0 0.5 2 2.5 1 1.5 κ*∆ x 2 th 1.5 3 κ∆ x ¤vBE`8RIdWHu¥˜’ —IÛ ‰5˜hIB„nUHQhe@•Tgb8HQ2SRI7w‰sH6dfq•58dŠbVT¢d‚BX58RIYuY‚4ƒFIHQ79HQY6’ H‡—˜Id‚HQY‚YWBX5¢2¬q•5¢dwDGFIHd7#5xhIBEnIHQhr@•Tgb8HQ2SRI7w‰sH6d9pdW5cbxBXhIH6Y…DGFIHeTV4Q4QRpdGT¢4‡ir@CBEDWF)@CFIBX4ƒF¬T`¢BEb8H62 @•Tcb¢HQ2SRI7…‰†HQdu46587#†582IH62oD5¢qjDGFpH…YW58yERpDGBE582BEY™dWH6YW58y„b8H6hkq•5¢dfDGFpH…HQ2oDGBEdWHw@•Tgb8H62xRp7…‰sH6dudGT¢2I`¢H TgbVT¢BXyXT¢‰IyXHfBX2‰T‘7#H6YWF‰5¢qvT–`¢BEb8H62†Y‚BXŒ6H¢t m ¡ l –ëî…ß l j’ 132†`¢HQ2IH6dGT¢yztpnU2IB„DGH‡r£hIBEgsHQd‚HQ2I46Hu58sH6dGTVDW58dWYu4QT¢2‰‰sH@CdWB„DWDWHQ2BX2eDGFpHuq•5¢dW7 æˆ(  çšíCä æ“(  çzí”øÿæ“(  7 çzí ” @CFIH6dWH†æ“(  çšíaBEYT¢2AT¢2oDGBEY‚ix7#79H6DWdWBE4w58†HQdWTVDG58dT¢2Ih©æ“(  7 çzíwBXYuTŽY‚ix797#H‡DGd‚BX4…58†HQdWTVDG5¢dQ’ … 1£qv@vH dWH6Y‚DWdWBX4‡D‘58RIduBX2oDWHQdWH6Y‚DCDG5–Y‚4ƒFIHQ79HQYfH‡—xDGH62IhIBX2p`9q•d‚587Ïê½6¬¥‘DW5wêø©¥˜tIDGFIH62 æ“(  Ýsçšíuä T¢2Ih æˆ(  7 ÝsçzíŠä ¦ ù äGÝxíjø î ß " … ” ” ¦ ù¸‹¡4Qæ•Ý˜í‚ü54‚6ÅÝxíA€4fç›ø~‹ æšÝxíWü 6)Ýxí¥€ ç›ø“‹ æ•Ý˜í‚ü 6)ÝxíA€ ç3û „ „ „    î ß … " 4‡æšÝxíWü54hø¬Ý˜í¥€4fç›ø æšÝxíWü øÆÝxí¥€ ç›ø æ•ÝxíWü „ øÆÝxí¥€ „ çÔû   "  " „ tuFIHw4Q58d‚dWH6YW†582IhIBE2I`–7#5xhIBEnIHQhŽ@•Tcb¢HQ2SRI7…‰†HQdfBXY ˜6–vh ä ø p b ¬X§¤›£¢¤Ô©5«¦Ñ)­kª†žd¢¤ª£ÒqŸd­¦¬S§5Ô­¥§¤¢“rž qp«¦®v›£¢˜­¥§¤«¦¢nsftx§“¨q›­¦¬F§¤žd©“¯xÔÉÔk›£§“¢¤ždª‚®‘­A¢“§£žd©Ú«Geq§¤­Ažd¬q› ъ¢¤«¦Ô™uvsxwEsCygz ­A¬ p§¤¨q›©ˆ¯xÔÉÔɛ†§¤¢“ž%ª'®‘­A¢“§£ÑŠ¢“«FÔ|uvsQ}~s y z i…{ ° ¦ ù ävø   æ 4x4658Yn–ëî…ß‘ø 4Q58Y   –Uî…ß…ø 4Q58Yë¥p–Uî…ßsç î ß " ‘ " "  " „ æ ‹4˜Y‚BX2j–Uî‘ß‘ø“‹ YWBE2   –Uî‘ß‘ø“‹ „ YWE B 2f¥o–ëî…ßsç   ˜ƒˆ  æz¥p’d$vhoç iG{ .

ÎxБSЫр×ÓèmÍCÒÔ̈Íó̈Íf͐×CÍóəÓfɓ’‚êˆÚxÒfÚ —I¦ ™ªFIH62wDWFIH”nU2IB„DGH‡r£hIBEgsHQd‚HQ2I46Hj58†HQdGTlDG58d‹BEY›T¢2oDGBEY‚ix797#H‡DGdWBE4uæ•4QHQ2oDWHQdWH6hëç‡tlDGFIHj795xhIBEnUH6h™@•Tcb¢HQ2SRI7‘r ‰†HQd•BEYvpRIdWH6yEi9dWHQT¢yŸ’§™ªFIHQ2ŽDWFIH™58†HQdWTVDG5¢dvBE2I4QyERIhIHQY•TaYfi˜7979H6DGd‚BX4‘4Q5879†582IHQ2oDQtxDWFIH™795˜hpBEnUH6h @•Tcb¢HQ2SRI7…‰†HQd#BXY–4Q5¢7#IyEH‡—stj@CBEDWFÆDWFIHBX7#T¢`8BX2IT¢d‚iA4Q5¢7#†582IH62SD9‰†HQBE2I`kHQ2oDGBEdWHQy„iÅHQd‚dW58d6’¬tuFIH q•58RIdfDGF˜r£58dWhpHQdCÜvTV¡ h Ha ` Y‚4ƒFIHQ79HuBXYC`8B„b8HQ2‰oi æˆ(  ݆çšíA€4‹øƗUæ“(  ݆çzíjøÿæ“(  ÝhçšíWü54”ä ¥ æšÝxí‚ü54‚6ÅÝ˜í¥€4fç î ß … tuFIHw7#5xhIB„nUHQhŽ@•Tgb8HQ2SRI7w‰sH6dfq•58dŠDWFIBXYuY‚4ƒFIHQ79HaYWTVDGBEY‚nUH6Y † ˜–£h¥þ € — •Ge @CFIBE4ƒF†`8B„b8H6Y  ø¬—o˜6–vhmø€˜–£hþ ‘•…e  ä ¥ æˆþ î ß … ‘•…e  6áþ € ‘•…e  ç ˜–£h•ä ¥ ˜oYWBE2f–Uî‘ß r æ   ø¬4Q58Yƒ–Uî…ßsç‚î…ß tuFIH‘795˜hIB„nUHQh‰@ŠTgb¢HQ2SRI7…‰†HQduId‚5sbxBEhIHQYuTdRpYWH6q•RpyvDG5x58y˜q•58dT¢Y‚YWH6YWYWBE2I`ehIB„ghH6dWHQ2p4QH–T¢IpdW5g—Sr BX7#TVDWBX582IY6’#132ADGFIH94Q5¢2SDWH‡—xD…5VqŠDGFIH9yXBE2IHgTVd4Q582ob8H646DWBX582)H6PSRUTVDGBE582‹t˜DGFIH#HQd‚dW58d‚Ya4QT¢2r‰†H#`¢BEb8H62 TdIFoixYWBE4gT¢y§BX2oDWHQdWpdWH6DGTVDGBE582‹’™•582IY‚BXhIH6d™582I46H‘T¢`oTVBX2DGFIHayXBX2pHgT¢d‘4Q582ob¢HQ46DWBX582eH6PSRUTVDGBE582BE2DGFpH q•58d‚7  Ý  Ý ø ‹ “ ä ¡  Þ  ß 582¬Thp587dTVBX2rH–—xDGHQ2phIBX2I`eq•d‚587 6$ DG‚ 5   ’ ƒCHQ4QT¢yXyˆq•dW5¢7ĔxHQ46DWBX582   ’ ¥p’   DWFUTVD–TY‚58yXR˜DGBX5¢2 BX2pBEDGBXTVDGH6hŽ‰oi‰T‘FUT¢dW79582IBE4™q•RI2I4‡DGBE582e@CBEDGFe@•Tgb8HQ2SRI7w‰sH6„ d –†BEY Ývæ|ߘàWނçj䆅”æ|ނçsþ @CFIH6dWH‡…”æ•Þ‚ç•YWTVDGBEY‚nUH6YCDWFIHa“ã ‘•  æz¥p’ “ ¡ ç " " … Þ ä¼6f˜˜‹i–£… ”˜58y„bxBX2I`…q•58df…væ•Þ‚çuT¢2IhYWRp‰IY‚DWBEDGR˜DGBX2p`9BE2oDG5ëjP†’류’ “ ¡ `8B„b8H6YCDWFIHuH‡—pT¢4‡D™Y‚58yXRpDWBX582TVY Ývæ|ߘàWނçj䆅”æ ¡ çsþ —• ¢  €¡”  ¤ 1£qjYWH64Q582ph˜r358d‚hIHQd‘4QH62oDGHQd‚HQhehIB„ghH6dWH62I4QH6Y™T¢d‚HwTVIIyXBEHQheDG59DGFIHwYWITVDGBXT¢yhDGH6dW7‰tUDGFIHuq•58yXyE5c@CBX2I` “ã©BEYu58‰pDƒTVBX2IH6hq•5¢f d …”æ•Þ‚ç–Û ¢ …‰äˆ6f˜˜‹i– h … " ˆsÍ «¦§“›§¤¨‘­¥§§¤›£¢¤Ô©U©“Ò‘ª˜¨Ï­¦©SuŠ‰¦‹ Ð zŒxŽ ™ ­A¢¤›¨‘­¦¬‘ aŸd›s ‰ex¯½Ÿd›†§“§“žd¬dc0‰¦‹ ÐQV2‘4’6“•”—– Œ™˜v‹ ©“¨qž¸Ñ §sžd¬S š° " Þ äˆ6f˜˜‹ …   YWBE2f–ëî…ß î‘ß ­¦¬‘ q›£¥¥­¦ŸÕґ­A§“žd¬dcu§“¨‘› .

—  

ǔȑɐʙËm̈ÍÏÎSЫÑvÒÔӑÒQËm̀Õz֐ÒÔєє̈͑̀Ówǔ̫ɐʈʈ͏×jؐÒfنə˧҇×uÓÚ DGFIHCY‚UTVDGBXT¢yëT¢IpdW5g—Sr æz¥p’ “p¦sç

” 58y„bxBX2I`DGFpBXYv“ãH‡—pT¢4‡DGy„i†æšYWBE2I4QHu@vHT¢d‚HC4Q582IY‚BXhIH6dWBE2I`DGFIH‘HQd‚dW58djq•dW5¢7 ˜ BX7#TVDWBX582Ž5¢2IyEipçŠT¢2IhY‚RI‰IY‚DWBEDWRpDGBE2I`–BX2oDG5ëjPë’s¥˜’ “ ¡ tp@•Hu58‰˜DƒT¢BE2 Ý 2xRp7#H6dWBX4QT¢y æ•ß˜à‚Þ‚ç€äk…”æ ¡ ç£þ 

‘•
¢ 

€¡”œ›  ¤

@CFIH6dWH‹ h BXY€DGFIHf2SRI7#H6dWBE4gT¢yvæš58d€795˜hIB„nUHQhU爁IFIT¢YWHY‚sH6HQh‹t˜@CFIBX4ƒFeBEYvd‚HQyƒTlDGHQh#DG5wDWFIH™795˜hpBEnUH6h @•Tcb¢HQ2SRI7…‰†HQdf‰oi

‹

‹
ä

h

ä

– h –

YWBE2“–Uî…ß ‹ –Uî…ß tuFIH€2xRp7#H6dWBX4QT¢y¢pFUT¢YWH€YW†HQH6haBEY‹DGFIHˆY‚sH6HQhaTVD›@CFIBX4ƒFaTuFUT¢d‚7#582pBX4vq•RI2p46DGBE582BXY›pdW58UTV`oTVDGH6h 2SRI79HQdWBE4gT¢yEyEi¢’ ”˜BE2I4QHf‹ h r‹ l ¦q•58d9DGFIBEYdH‡—pT¢79IyEH¢tŠDGFpHe2xRp7#H6dWBX4QT¢y‘YW58yERpDGBE582«Id‚58UT¢`8TVDGH6Y DG5x5eYWyE5c@CyEi8’‘”˜BX2I46Hë‹ h BXYuTŽq•RI2p46DGBE582e5¢q€DGFIH‘@•Tgb8HQ2SRI7w‰sH6dQtmDWFIH‘2SRI79HQdWBE4gT¢y”T¢IId‚5c—˜BE7dTVDWBX582 BX2oDWdW5xhIRI4QH6YdhIBEYW†HQd‚YWBX5¢2‹tuT¢y„DGFI58Rp`8FÆDWFIH5¢dWBX`¢BX2UT¢yC܈ ã ÏBEY#2I5¢2IhIBXY‚sH6dWY‚BEb8HV’ÿifYŽTrd‚HQY‚RIyEDQtuT @•Tcb¢H6q•58d‚7¡4Q582IY‚BXYfDGBX2p`†5¢qŠ7dT¢2oiehpBEgsHQdWH62oDa@•Tgb8HQ2SRI7w‰sH6d…4Q5879†582IHQ2oDWYaH‡b8HQ2oDWRUT¢yXy„iAyX58Y‚HQYuBEDWY 58d‚BX`8BE2UT¢yëq•58dW7‰’ ¤vBX`¢RIdWHu¥˜% ’ $™Y‚FI5c@CYˆDWFIHu2SRI79HQdWBE4gT¢y˜IFUT¢Y‚HCYW†HQHQh‘q•58dmDGFIHuY‚4ƒFIHQ79HQY€4Q582pYWBXhpHQdWH6h–IdWH‡b˜BE58RIY‚yEi8’ tuFIHŽ2SRI7w‰sH6dw5¢q•œU‡ly|„Ux|žuœU}‡€#è€r  1¢}‡ X}‡„x…¢x|‚ž æ ŸŸ  珉oik@CFIBE4ƒF¬T†`8B„b8HQ2@ŠTgb8HeBEYad‚HQY‚58yEb¢HQh)BEY `8B„b8HQ2ʼnoi   ‚ro–Uî…ߘ’tuFIHed‚HQY‚58yEbxBX2p`eH6¨94QBEHQ2I4‡iÅ5¢q‘TYW4ƒFpHQ79H‰4QT¢2¬‰sHdH‡—˜IdWH6YWY‚HQh«BE2)DGH6dW79Y 5¢q‘DGFI¡ H 0‡  dWH6PSRIBXd‚HQhÅDG5AIdW5xhIRI46HHQdWd‚58dWY‘‰†HQyE5c@µTeYW†HQ46BEnUH6h©yEH6b8H6yŸ’ŤU58d‘H–—pT¢7#pyXH¢tˆDGFpH YWH64Q582ph˜r358d‚hIHQdd4QHQ2oDWHQdWH6hªhIBEgsHQd‚HQ2I46HYW4ƒFIH67#Hed‚HQPSRIBEdWHQô Y c ¡ 0‡  DW5rId‚5˜hIRp4QHeT¢2 H6dWd‚58ddBE2 IFUTVYWH…Y‚sH6HQhe5¢qˆyEHQYWY‘DGFUTV2 ¡ ’„¦…sH6dW46HQ2oDg’tuFpÓ H $sr3†58BE2SDCq•58Rpd‚DGFxr358d‚hIHQdf4QHQ2oDWHQdWH6hkYW4ƒFIH67#H‘TV2Ih DGFpHŽq•58RId‚DWF˜r35¢dWhIH6d–ÜvT¢¨ h H ` Y‚4ƒFIHQ79HdWHQPSRIBEdWHeq ¦ $TV2Ihª¦ ¡ ŸŸ  dWHQY‚sH646DWBEb8H6yEi)DW5kT¢4ƒFIBEH6b8HDGFpH YGTV7#HH6dWdW5¢d‘yEH6b¢HQyŸ’ ¤UdW587 DGFIHq•58d‚7§5¢8 q jPë’ë¥p’ “p¦8t˜@vHuYWH6HuDWFUTVDuDWFIHdWHQT¢y‹UT¢dfDC5¢q§DGFIH795˜hpBEnUH6hŽ@ŠTgb8H62SRI7…‰†HQd yXHQT¢hIY9DG5ÅT¢2 H6dWd‚58ddBE2«DGFIHIdW58IT¢`oTVDWBX582©YW†HQHQhht‘@CFIBEyXHDGFIHBX7#T¢`8BX2IT¢d‚iŁIT¢d‚DeyEHgT¢hpY#DG5)T¢2 HQd‚dW58dvBX2dDWFIHT¢79IyEBEDGRphIHC5¢q›DGFIH™YW58yERpDGBE582‹’vtuFSRIYŠDWFIHTV2SDWBXYfi˜7979H6DGd‚BX4‘s5¢d‚DGBE582d5Vq›DGFIHY‚UTVDWBƒT¢y hIB„ghH6dWHQ2p4QHC58†HQdGTlDG58djhIH‡DGH6dW79BX2IH6Y”DGFIHCHQd‚dW58djBE2–YW†HQHQhdT¢2Ih‘DGFpH‘Yfix7#79H6DWdWBX4•s58dfDGBE582…DGFIHCH6dWdW5¢d BX2‰T¢79IyEBEDGRphIH¢’mtuFIBXYCBXYŠq•Rpd‚DGFpHQdChIBXY‚4QRIY‚YWH6hBX2†”˜H646DWBX582A¦8¦8’   ’

¤U58dŠDGFpHaTV‰s5cb8HuH–—ITV7#IyEH¢t

‹

h

­®£¢ ½

¹6¿Å±Àj±¶eÁC±

”…C±Àj´a»”Ã9Àˆ²

´w»‡¤

Ãëîe¶e·´wÀ€¹6±²

i™YhIBEYW46RIYWY‚HQh BE2ÿ”˜HQ4‡DGBX5¢2ÿ¥˜’ ¥p’   t™T¬7#TVDWdWB„—«hIB„ghH6dWHQ2p4QHk58†HQdGTlDG58dŽBE2I4Q5¢dW†58dGTVDWHQYe‰†5¢DGFªDGFpH hIB„ghH6dWHQ2p4QHŽTVIIdW5g—˜BX7#TVDWBX582kBX2ADGFIH#BX2oDGH6dWBE58du5¢qŠDGFpHdhI587#T¢BE2TV2IhrDGFITVD…TlDaDWFIHd‰†58RI2IhIT¢dWBEHQYQ’ 132DWFIBXYfYWHQ4‡DGBE582‹ts@vH…46582IYWBEhIHQdfDGFIHa‰s58Rp2IhUT¢dfi58†HQdWTVDG5¢dWY™2pHQHQhpHQheq•58d‘58RId™795xhIHQy§HQPSRUTVDWBX582pY q•58dC4Q5¢2Sb¢HQ4‡DGBX5¢2eT¢2Ih†hIBEgsRIYWBE582‹’€132†DWFIHa4QT¢YWHa5¢q”†HQd‚BX5xhIBX4‰†58RI2phUT¢d‚i46582IhIB„DGBE582IYQtU2p5dYW†HQ46BƒT¢y ‰†58RI2IhUTVd‚ie58†HQdWTVDG5¢dWYCT¢d‚Hwd‚HQPSRIBEdWHQhh’

ÎxЦ¥oЫ֐ÒÔÑvє̈Íf̈ӅǔÌó×Cʈ̈ÍCə˕×CÍfÚeəËV§×ÓèmÓf֑ə͑ÒỒÚ

—o¥

1.2

1

0.8

a* _ a

2 Central
th

nd

0.6

4 Central 4 Pade
th

0.4

0.2

0 0

0.5

1

1.5

2

2.5

3

κ ∆x

¤vBE`8RId‚Hu¥p’%$xÛ

òfRI79HQd‚BX4gTVyhIFUTVYWHwYW†HQHQh‰q•58dŠbVT¢dWBE58RIYuY‚4ƒFIHQ79HQY6’

–—¨—s™

öguup‚!ëgsiups|r2à ekiª©hp‚tÚjg†eki&«­¬Xâu|7jg£ehi

  ’ ¥ptëDGFpHw‰†58RI2IhUTVd‚i4Q5¢2IhIBEDWBX582pY‘q•58dCDGFpHwyXBE2IHgTVd‘46582ob8H646DGBE582H6PxRITVDGBE582 4gTV2d‰†HfH6BEDGFpHQdˆsH6dWBE5˜hIBE4u58dˆ5¢q›BE2pqU5c@Šr358R˜DWqU5c@«D3ix†H¢’j1329DGFIH‘yƒTVD‚DGH6d€4QT¢YWHVtpTaãBXd‚BX4ƒFIyEH6Dj‰†58RI2Ihxr T¢dfi4Q5¢2IhIBEDWBX582†BXY`¢BEb8H62ATVDDGFpH‘BX2pqU5c@‰†58RI2IhUTVd‚i8t‹@CFpBXyXHa2I5e4Q5¢2IhIBEDWBX582BXYY‚sH64QBEnIHQhATVDDGFpH 58RpD‚qU5c@ ‰†58RI2phUT¢d‚i¢q ’ €‘HQdWH…@vH9T¢YWY‚RI79H…DGFUTVD™DGFIH…@•Tgb8H9YW†HQHQh ‹BXY†58YWB„DGBEb¢H¢’‹DWFxRpYDGFIH‘yEH6q|D‚r FUT¢2phA‰†58RI2phUT¢d‚iBXY™DWFIH–BX2pqU5c@å‰s5¢RI2IhUT¢dfi8t˜T¢2phkDWFIH–dWBE`8FoD‚r3FIT¢2Ihe‰†58RI2IhUTVd‚iBEYDGFIH‘5¢RpDWqU5c@ ‰†58RI2IhUTVd‚i8’ˆtuFSRIYCDGFIHb¢HQ46DW58d‘5¢qvRp2I÷S2I5c@C2IYfBEY Ý ² äÏù Ý£4–àCÝ e

ƒCH6q•HQd‚dWBE2I`#DW5e”˜H646DGBE582

T¢2Ih‰Ý!äCBXYuYW†HQ46BEnUH6h‹’ •582IY‚BXhIH6d”nUdWYfD”DGFIH•BX2pqU5c@Ɖ†58RI2IhUTVd‚i8’m1£DjBXYm4QyEHgT¢dvDWFUTVD”T¢Y”yX5¢2I`T¢YvDWFIHŠBE2SDWHQd‚BX58d§hIBEgsHQd‚HQ2I46H T¢IpdW5g—˜BX7#TVDGBE582hp5˜H6Yw2I5¢DwH‡—xDGH62Ihʼn†H6i85¢2Ih¬Ý˜¥ í €4ƒtvDWFIHQ2r2I5YW†HQ46BƒT¢yˆDGd‚HgTVDW7#H62oD…BXYwdWH6PSRIBXd‚HQh q•58dŠDWFIBXYu‰†58RI2IhIT¢d‚i¢’v¤U58duH–—pT¢7#pyXH¢t˜@CBEDGFY‚HQ4Q5¢2Ih˜r35¢dWhIH6df46HQ2oDGH6dWHQh†hIBEgsHQd‚HQ2I46HQYC@•Hw58‰pDGT¢BX2 

à ‡S‡x‡ àCݕÅaû Æ

æz¥p’ “   ç

² ä »Dݑ ² ø ( £Ý
@CBEDWF ¡ ¦ ¡ ¦ ¦

t¦ Œ²º y 6‘Ý!ä ¶ ·· ·· 6w¦ ´³ ¡   î…ß ´´ ¡ ´µ ’’ ¹
’

æz¥p’ “8¥oç

» ä

6w¦
  î…ß

³´´ ´´µ

6w¦

¡ 6 ¦ w

¶ ·· ··
à ’’ ¹ ’

t ŒA² º y ä

æz¥p’ “¢—Sç

—8—

ǔȑɐʙËm̈ÍÏÎSЫÑvÒÔӑÒQËm̀Õz֐ÒÔєє̈͑̀Ówǔ̫ɐʈʈ͏×jؐÒfنə˧҇×uӐÚ

€f5c@vH6b¢HQdQt™BEq@vHARpYWHDGFpHq•58RId‚DWF˜r35¢dWhIH6dŽBX2oDGH6dWBE58dd58†HQdWTVDG5¢dŽ`8BEb¢HQ2 BE2gˆPë’u¥˜’%$V—ptCDWFIHQ2ªDGFpH T¢IpdW5g—˜BX7#TVDGBE582eTVD€êwäå¦udWHQPSRIBEdWH6YfT…bVT¢yERIH5¢qmÝxíA€ tI@CFpBX4ƒFBXYu58R˜DGYWBEhIHDGFpHwhp587dTVBX2‹’€‘HQ2I46H  ThIBEgsHQd‚HQ2oDw58sH6dGTVDW58duBXYudWH6PSRIBXd‚HQh)TlD™êä ¦–@CFIBX4ƒFH‡—xDGH62IhIYa582IyEiDG5#êëÿ 6 ¦8t§@CFpBXyXH‘FITcbxBE2I` DGFpHT¢IpdW58Id‚BƒTVDWHf5¢dWhIH6d•5¢q˜TV4Q4QRpdGT¢4‡i8’v”˜RI4ƒFT¢2#58†HQdGTlDG58d6t˜÷S2I5c@C2T¢Y•Td„U¢˜{d}‡€–yŸW¢ 5‡ À ‡l¢x„†ˆo¢€‡

žcƒ‚U}‡{d}‡t€4gT¢2©FUTgb8HTV2«58d‚hIHQd–5¢qT¢4Q46RIdGT¢4‡i¬@CFIBX4ƒF©BXY‡l„h}e‡l€‚ˆx}‡€e X‡lè•}‡€wDGFUTV2©DGFITVD95¢qDGFpH BX2oDWHQdWBE58d™Y‚4ƒFIHQ79H¢t˜TV2IheDGFIH…`¢yX58‰UTVyvT¢464QRIdWT¢46i@CBXyXy§HQPSRUTVyvDGFUTlD5¢qjDGFpH‘BX2oDGH6dWBX5¢d™YW4ƒFpHQ79H¢’…¤U58d H‡—pT¢79IyEH¢tp@CB„DGFŽq•58RIdfDGF˜r£58dWhpHQdŠ46HQ2oDGH6dWHQh†hIBEgsHQd‚HQ2I46HQY6tI@•Hu4gTV2‰RpYWHDGFIHq•5¢yXyX5c@CBE2I`aDWFIBXd‚h˜r358d‚hIHQd 58†HQdWTVDG58dCTlD€ê‘ä¦8Û æˆ(  Ýsç 4 ä ¦ æ†6   ݕä6¬¥¢Ý£4høƓ¢Ý 6ÅÝ ç „  “8î…ß æz¥p’ “3$8ç

@CFIBE4ƒFdBXYˆHQT¢YWBEyEi…hpHQdWB„b8H6hdRIY‚BX2I`aTwtmTgixyX58djDƒTV‰IyXHV’”tuFIH‘dWHQY‚RIyEDWBX2I`uhIB„ghH6dWH62I4QH‘58†HQdGTlDG58dˆFUT¢YˆDGFpH q•58d‚7§5¢7 q jP†’†¥p’ “8¥…@CBEDWF

» ä

6w¦
¦   î…ß

³´´ ´´µ

6f“ 6 c õ
¦

6õc

¦   ¡

6   c
¡

6a¦ c

¶¸··
¦

··
à

’’ ’

¹

t Œ¦² º y ä

6w¦
¦   î…ß

³´´ ´´µ

6‘—8Ý!ä ݕä
¡ ’’ ’

¶¸·· ·· ¹

æz¥p’ “8“oç

tuFIBEYfTVIIdW5g—˜BX7#TVDWBX582ŽBEYu`8yX5¢‰UT¢yXy„i–q•58RId‚DWF˜r35¢dWhIH6dCT¢4Q46RIdGTVDWH¢’ iŠDaDWFIH958RpDWqI5s@ ‰s5¢RI2IhUT¢dfi8tm2I5†‰†58RI2phUT¢d‚i4Q582IhpBEDGBE582kBXYuYW†HQ46BEnUH6h‹’d™H–7…RIY‚DwT¢IpdW5g—Sr BX7#TVDWH  5 Ý r  ß TVD…2I5˜hpD H ‘ @CBEDGFr2I5eBX2pq•5¢dW7#TVDGBE582rTV‰s58R˜D…• Ý Å5 ü 4–’ktuFSRIY…DGFIHeY‚HQ46582Ih˜r£58dWhIH6d 4QH62oDGHQd‚HQh˜r£hIBEgsHQd‚HQ2I46Ha5¢sH6dGTVDW58dQtx@CFIBX4ƒFdWH6PSRIBXd‚HQYuÝxíW5 ü 4GtI4gT¢2p2I5¢Du‰†HRIYWH6hTlDˆê…2 ä ‘’€i ‰UT¢4ƒ÷8r @•T¢dWhÆhIBEgsHQd‚HQ2I46Heq•58dW7…RpyƒT7…RpY‚D–‰†H‰RpYWHQhh’)™ªBEDGFÆTeYWH64Q582ph˜r358d‚hIHQd–BE2SDWHQd‚BX58d…5¢sH6dGTVDW58dQtjDGFpH q•58yEyX5c@CBX2I`anUdWYfD‚r35¢dWhIH6d‘‰IT¢4ƒ÷o@ŠT¢d‚h†q•58d‚7…RIyXT…4gT¢2‰†HwRpYWHQhhÛ æ“(  s Ý çÅ ä ¦ æšÝ•ÅÙ6)ݕҀ4‚ç î ß … æz¥p’ “398ç

tuFIBEY‘pdW5xhIRI4QH6Y‘T9hIB„ghH6dWH62I4QHu58†HQdWTVDG58dŠ@CB„DGF

» ä

´³´ 6w¦ 6w¦ ´ ´   î…ß ´´ ´´ µ´

¡

¦ ¡ 6 ¦ a ¡

¦ ¦

¶ ·· ·· ·· · 6  
  ¡ ¦ ’’ ’ 6w¦

·· à ¹

t¦ Œ ²º y ä

6a¦
  î…ß

³´´ ´´ ´´ ´µ

6‘Ý•ä ¶ ·· ·· ¡ ·· ¡ ·
’’ ’ ¡

æz¥p’ “coç

¹
dWH6PSRIBXd‚H

132dDWFIH4gTVYWH5¢q›Taq•58RIdfDGF˜r£58dWhIH6d•46HQ2oDGH6dWHQhBE2SDWHQd‚BX58dˆ58†HQdWTVDG58d€DGFIHfyƒT¢YfDvD3@v5‘dW5c@CYŠ5Vq » 795˜hIB„nU4gTlDGBX5¢2‹’

ÎxЦ¥oЫ֐ÒÔÑvє̈Íf̈ӅǔÌó×Cʈ̈ÍCə˕×CÍfÚeəËV§×ÓèmÓf֑ə͑ÒỒÚ

—Â$

i 2I5VDGFIH6daTVIIdW58T¢4ƒFeDG5eDGFIH‘hpH6b8H6yX58p7#H62SDw5¢qv‰†58RI2IhUTVd‚iY‚4ƒFIHQ79HQYwBXY™BX2DGH6dW79Y5¢qvYWUT¢46H ™ H‡—xDWdGT¢†58yƒTlDGBX5¢2‹’˜tuFIHvq•58yEyX5c@CBX2p`Šq•58dW7…RpyƒTCT¢yXyE5c@CY§Ý•Åü54hDG5‘‰sHvH‡—xDGdWT¢s5¢yƒTVDWHQhuq•dW587óDWFIHvBE2oDGHQd‚BX58d hUTVDGT…DG59T¢dW‰pBEDGdWT¢d‚id58dWhpHQdC582‰T¢2‰HQPSRIBXY‚UT¢46HQh`¢dWBXhhÛ æf¦p6V®

€4 Ü

ç ݕŏü54”ä ¡

æz¥p’ “hoç

@CFIH6dWH„® BEY€DWFIHY‚FIBEq|D•58sH6dGTVDW58d€hIH6nU2pHQh‰oi¯®wݘíŠäðÝxíWü54€TV2IhdDGFpH58dWhpHQd•5Vq›DGFIHTVIIdW5g—˜BX7#Tlr DGBE582BX° Y ë6º¦8’€¤U58dŠH–—pT¢7#pyXH¢t˜@CBEDGF~dä   @•Hw58‰pDGT¢BX2 æf¦p6  

®

€4

ø€®

€  çÔÝ5ŏü54”äÿݕŏü5486   ݕÅðøÆݕҀ4”ä ¡

æz¥p’d9 ¡ ç

@CFIBE4ƒF†`8B„b8H6YCDWFIHq•58yXyE5c@CBX2I`wnIdWY‚Dfr358d‚hIHQd‘T¢pIdW5g—˜BX7#TVDGBE582dDW59ݕŏü54–Û ݕŏü54jä   ݕÅ26ÅݕҀ4 æz¥p’d9˜¦sç

”˜RI‰pY‚DGB„DGRpDWBX2I`#DGFIBEYBX2oDG5#DGFIH‘Y‚HQ4Q5¢2Ih˜r35¢dWhIH6d4QHQ2oDWHQdWH6hhpBEgsHQdWH62I4QH‘5¢sH6dGTVDW58dT¢pIyXBEHQheTVD2I5˜hpH ‘ `8BEb¢HQY æ“(  s Ý çÅ ä ¦ æ•Ý5ŏü5486)ݕҀ4‚ç¡ä î ß   ‘ ä @CFIBE4ƒF†BEYuBXhIH62SDWBX4QT¢yhDG5ëjP뒆¥p’ “9˜’ ¦ æ   ݕÅÙ6)ݕҀ476ÅݕҀ4Ôç î ß   ‘ ¦ æ•5 Ý ÅÙ6ÅݕҀ4fç î…ß

æz¥p’d9   ç

–—¨—o‚

öguupÙl2gonÏâu}¡g£ehi&«­¬Xâu|7jg£ehi

132eYW5¢yEbxBX2I`…DWFIHhIB„ghRpYWBX5¢2ŽHQPSRUTVDWBX582Ž@vH7…RIYfDC4Q582pYWBXhpHQduãBEdWBX4ƒFpyXH6D•T¢2Ihò‘HQRI7#T¢2I2‰†58RI2IhIT¢d‚i 4Q5¢2IhIBEDWBX582pYQ’vtuFpHCDWdWHQTVDG79HQ2oDj5¢qhTuãBXd‚BX4ƒFIyEH6Dm‰s58Rp2IhUT¢dfi‘4Q582phIBEDWBX582…pdW5x4QHQH6hIY€T¢yX5¢2I`DGFIHuYWT¢79H yXBE2IHQY™T¢Y™DWFIH…BX2˜qU5c@剆58RI2phUT¢d‚i‰q•58d‘DGFIH…46582ob8H646DGBE582kHQPSRUTVDWBX582ehpBXYW46RIYWY‚HQhAT¢‰†5cb8H¢’™ªB„DGFDGFpH YWH64Q582ph˜r358d‚hIHQd‘4QH62oDGHQd‚HQh†BX2oDGH6dWBX5¢dŠ58†HQdGTlDG58d 

s æˆ( ¥ Ý çí ä

¦ æšÝxíWü5486   ݘívøÆÝxíA€4fç î ß  …

æz¥p’d9¢¥oç

2I5f7#5xhIBEnI4gTVDWBX582IY‹T¢dWH€dWH6PxRpBXdWH6h…2IHQT¢dm‰†58RI2phUT¢dWBEHQY6tVyEHgT¢hIBE2I`uDG5‘DGFIH€hIBEgsHQd‚HQ2I46H•58†HQdWTVDG5¢d§`¢BEb8H62 BX‰ 2 jPë’ë¥p’   9˜’ ¤U58dwTòfHQRp7dT¢2p2)‰s5¢RI2IhUT¢dfir46582IhIB„DGBE582‹t˜@•HeTVYWYWRp7#HdDWFUTVD  Ú Ý r  ß©BXYwYW†HQ46BEnUH6h¬TVDêAä ‘ øð¦8tpDWFUTVD‘BXY

V 

Ý  Ý äiV  ß W ŏü54  ß W —

æz¥p’d9V—Sç

—o“

ǔȑɐʙËm̈ÍÏÎSЫÑvÒÔӑÒQËm̀Õz֐ÒÔєє̈͑̀Ówǔ̫ɐʈʈ͏×jؐÒfنə˧҇×uÓÚ @CFIBE4ƒF†BEYuBX2eDGFIHq•58yEyX5c@CBX2p`wq•58dW7‰Û

tuFSRIYC@•HwhIHQY‚BX`82‰T¢2‰58†HQdWTVDG58dCTlDC2I5˜hpH{‘ æˆ( ¥ h Ý çÅ ä

¦ 擋oݕÅÿø~ŒƒÝ5Ҁ4fç›ø î ß  …

º
î…ß

V 

Ý 

ß W ŏü54

æz¥p’d9$8ç

@CFIH6dWH‡‹ëtŒstëTV2Ihfº™T¢dWHu4Q5¢2IY‚DGT¢2oDGYu@CFIBX4ƒF‰4gTV2†HQT¢YWBEyEi#‰sHuhIH‡DGHQd‚7#BE2IHQhRIY‚BX2I`9T–tvTgixyX5¢dŠDƒT¢‰IyEH¢t T¢YCYWFp5s@C2†BX2etmT¢‰IyEHa¥˜’ —I’

V 

Ý 

ß  W í
Ýxí

ø

¦ 戋SÝxí¥€4›ø~ŒƒÝ˜í‡ç›ø î ß  …

º
î‘ß

V 

Ý ä×ù  ß W í‚ü54 t u û w y í u û 4 ‹½Yxæ|rW¦sç „ Y Ý 4 ºkYoæf¦sç  Y 
î…ß „ Y

‹{YQÝ˜í¥€4 ŒkYcÝxí î‘ß YaºÉY t u¥w y W u í ü54

î‘ß  Y tu ú w y í u ú

t uxw y í u ‹{YSæ|rG¦sçhY 44

î…߁Y

t u ú w y í u ú
¦

î…ß  Y

t u ü w y í u ü ‹‡Yxæ|rW¦sç … Y 4 … 4 ºkx Y æÔ¦sç „ Y Ý

î…ß … Y

‹ Œ

º

‹‡Yxæƒrƒ¦sç  Y 4  4 ºkx Y æÔ¦sçhY 4

tvTV‰IyXHu¥p’ —I’jtmTgi˜yE58dŠDƒTV‰IyXHq•58dCò‘HQRI7#T¢2I2‰‰†58RI2IhUTVd‚ie46582IhIB„DGBX5¢2‹’ ”˜58y„bxBX2I`…q•58dp‹Ut!ŒctëT¢2Ih˜ºstp@•Hw58‰pDGT¢BX2ŽDWFIHq•58yXyE5c@CBX2I`…5¢sH6dGTVDW58dQÛ æˆ( ¥ s Ý çÅ ä ¦ æ   ݕҀ4‚6   ݕŅç›ø ¥8î…ß    ¥¢î‘ß

V 

Ý 

ß W ŏü54

æz¥p’d9¢“oç

@CFIBE4ƒFApdW5xhIRI4QH6YDGFIH‘hIB„ghH6dWH62I4QH‘58†HQdWTVDG5¢d`8BEb¢HQ2kBX2TjP뒛¥p’   hp’aòf5VDGBX46HwDGFITVDDGFpBXY58†HQdWTVDG58d BXYaYWH64Q582Ihxr358d‚hIHQd‘TV4Q4QRpdGTVDWH¢’132)DWFIHŽ4gTVYWHŽ5¢q‘T†2xRp7#H6dWBX4QT¢yvT¢IId‚5c—˜BE7dTVDWBX582kDG5TòfH6RI7dTV2I2 ‰†58RI2IhUTVd‚i4Q5¢2IhIBEDWBX582htpDGFIBEYuBXYC2IH64QHQY‚YGT¢dfi†DW5#5¢‰pDƒT¢BE2‰T9`8yE58‰UT¢yEyEi9YWH64Q582ph˜r358d‚hIHQdfT¢464QRIdWTVDGHuq•58dfr 7…RIyXTVDGBE582‹’mtuFIBXYC4Q5¢2SDWdGT¢YfDGYC@CBEDWFDWFIHu2SRI7#H6dWBE4gT¢y‹‰†58RI2phUT¢d‚iŽYW4ƒFIH67#H6YfhIH6YW46dWBX‰†HQh†IdWH‡bxBX58RIY‚yEi @CFIBE4ƒF†4QT¢2†‰†Hu582IHu58d‚hIHQduyE5s@vHQduDWFUT¢2DGFIHuBX2oDWHQdWBE58dŠY‚4ƒFIHQ79H¢’ ™rH–4QT¢2rTVyXYW5e5¢‰pDƒT¢BE2eDGFIH–58†HQdWTVDG58dBET 2 jP†’§¥pd ’ 9¢“ŽRIY‚BX2I`DGFpH‘YWUT¢46H#H–—xDGdGTVs58yXTVDGBE582BEhIHgTp’ •582IY‚BXhIH6d‘T–Y‚HQ46582Ih˜r£58dWhIH6df‰IT¢4ƒ÷o@ŠT¢d‚h˜r3hpBEgsHQdWH62I4QHaT¢IpdW5g—˜BX7#TVDGBE582T¢IIyEBXH6hTVDC2I5xhI‡ H ‘ øª¦8Û

V 

Ý 

ß W ŏü54
¦ ¥ È

ä

¦ æšÝ•Å‡€486ŗ8ݕŪø©¥¢Ý•Åü54‚ç›ø~’Žæzî…ß  ç î ß   …

æz¥p’d998ç

”˜58y„bxBX2I`…q•58duݕŏü54v`8B„b8HQY ݕŏü54jä —8ݕÅÙ6)ݕҀ4‹ø   î…ßðV 

Ý 

ß W ŏü54¤Ê

ø“’dæzî…ß „ ç

æz¥p’d9vcoç

Îxв±cЫʈ͏׳§ª’˜Ì€ÙÚ —Â9 ” RI‰pY‚DGB„DGRpDWBX2I`©DGFIBEYBX2oDG5ºDWFIH)Y‚HQ4Q5¢2Ih˜r35¢dWhIH6dA46HQ2oDGH6dWH6håhIB„ghH6dWH62I4QHÅ58†HQdWTVDG5¢d†q•58deTºY‚HQ46582Ih ˜ hIH6dWBEbVTVDWBEb¢HuT¢IIyEBXHQh‰TVDC2p5˜hIH½‘ `8BEb¢HQY æ“( ¥ s Ý çÅ ä ä ä ¦ æ ݕŏü5486   ݕŪøÆݕҀ4‚ç š î ß  … ¦ ¥¢Ý•Å‡€486œ¢Ý•ÅðøƗ8ݕÅ26ÅݕҀ4‹ø ¥8î…ß DÈ ¦ æ   Ý5Ҁ486   Ý5Ņç›ø ¥8î…ß    ¥8î…ß æz¥p’d9vhoç   î…ßðV  Ý V  Ý  ßUW ŏü54  ß W ŏü54¤Ê æz¥p’Õc ¡ ç @CFIBE4ƒF†BEYuBXhIH62SDWBX4QT¢yhDG5ëjP뒆¥p’%9V“p’ ­®G´ µ ÀjÃY¶·ˆQ±R ² ¦8’‘ãH6dWBEb¢H‘TdDGFpBXdWhxr358d‚hIHQd‘nU2IBEDWH‡r£hIBEgsHQd‚HQ2I46H‘T¢IId‚5g—pBE7dTlDGBX5¢2†DG5ŽTdnUdWYfDuhIHQd‚BEbVTVDGB„b8HaBX2DGFpH q•58d‚7 æ“(  Ýhçšíuä ¦ 擋oÝxíA€ ø~Œ–Ýxí¥€4høzº‡Ýxívø  î ß … " ÝxíWü54fç ¤vBE2IheDGFIHuyEHgT¢hIBE2I`‘HQd‚dW58dŠDWHQd‚7†’   ’‘ãH6dWBEb¢HuT–nI2IBEDWH‡r3hpBEgsHQdWH62I4QHuTVIIdW5g—˜BX7#TVDWBX582ŽDW5#T…nUd‚Y‚D‘hIH6dWBEbVTVDWBEb¢HBX2eDGFIHq•58d‚7 ‹†æ“(  ÝhçšíA€4høÿæ“(  Ýhçšíuä ¤vBE2IheDGFIHuyEHgT¢hIBE2I`‘HQd‚dW58dŠDWHQd‚7†’ ¦ 擌ƒÝxíA€4‹ø~º‡Ýxí”ø î ß … " ݘí‚ü54fç ¥p’fH‘YWBX2p`–T…—æšBE2SDWHQd‚BX58dG瀁†58BX2oDŠ79HQY‚F‹tp@Cd‚BEDGH™58RpDŠDWFIH—s¸ˆ—‘7dTVDWdWBE4QHQY•T¢2IhŽDWFIH‰s5¢RI2IhUT¢dfior 46582IhIB„DGBX5¢2†b8H646DW58dq•58dW79HQh‰SiRIYWBE2I`dDWFIH…YW4ƒFpHQ79H‘hIHQd‚BEb8H6hkBX2PSRIH6Y‚DGBE582   @CFIH62k‰s5¢DWF ÝT¢2Ih  5 Ý r  ßkTVdWH`8B„b8HQ2†TVD€ê‘ä ¡ TV2IhÝBXYu`¢BEb8H62TlD€ê‘2 ä $˜’ —I’fƒCHQ†HgTVD‘PSRIH6Y‚DGBE582   @CB„DGF " ä ¡ ’ $˜’‘ãH6dWBEb¢HuT–nI2IBEDWH‡r3hpBEgsHQdWH62I4QHuTVIIdW5g—˜BX7#TVDWBX582ŽDW5#T…DWFIBXd‚h‰hpHQdWB„bVTVDGB„b8HBX2eDWFIHq•58dW7 æˆ( ¥¥ ݆çšíuä ¦ 擋oÝxíA€ ø~Œ–Ýxí¥€4›ø~º‡Ýxívø  î ß „ … " ÝxíWü54hø“þgÝxí‚ü ç  ¤vBE2IheDGFIHuyEHgT¢hIBE2I`‘HQd‚dW58dŠDWHQd‚7†’ .

—3c ǔȑɐʙËm̈ÍÏÎSЫÑvÒÔӑÒQËm̀Õz֐ÒÔєє̈͑̀Ówǔ̫ɐʈʈ͏×jؐÒfنə˧҇×uÓÚ “p’‘ãH6dWBEb¢HuT‘4Q5879UT¢4‡Daæš58dCÜmT¢h¨Hc ` çŠnU2IB„DGH–r3hIB„ghH6dWHQ2p4QHuT¢IpdW5g—˜BX7#TVDGBE582dDG59T‘YWH64Q582phhIH6dWBEbVTlr DWBEb8HuBE2ŽDGFIHq•58d‚7 " æ“( ¥ ݆çzíA€4‹øÿæ“( ¥ ݆çzíjø~þSæ“( ¥ ݆çzíWü54jä ¦ 戋SÝxí¥€4›ø~ŒƒÝ˜ívø“º–ÝxíWü54‚ç î ß  … ¤vBE2IheDGFIHuyEHgT¢hIBE2I`‘HQd‚dW58dŠDWHQd‚7†’ 9˜’‘¤vBE2Ih#DWFIH795˜hpBEnUH6hd@ŠTgb¢HQ2SRI7…‰†HQduq•58dˆDGFIH™58sH6dGTVDW58d•hpHQdWB„b8H6h‰BE2dPSRIH6Y‚DGBE582e¦8’ˆÜjyX5VD•DGFpH d‚HgT¢y›TV2IheBX7#T¢`8BE2UT¢d‚i9UTVd‚DGY•5¢q#– h î…߆b˜Y6’£–Uî…ßq•58d ¡ml –Uî…ß l j  ’€•5879UT¢d‚HDGFIHd‚HgT¢y UTVd‚Du@CBEDWFeDGFUTVD‘58‰˜DƒT¢BE2IHQheq•dW5¢7 DGFpHuq•5¢RId‚DWF˜r358d‚hIHQdu46HQ2oDGH6dWH6h58†HQdGTlDG58dw戈  Pë’ë¥p’dV $ —S疒  cp’CifIIyXBE4gTVDWBX582d5¢qmDWFIHuYWH64Q582Ihxr3hIH6dWBEbVTVDWBEb¢Hw5¢sH6dGTVDW58dŠDG5‘DGFpHuq•Rp2I46DWBX582˜þ ‘• `8BEb¢HQY   þ ‘•   ä $ 6 –  þ ‘•  ß  ifIIyXBE4gTVDWBX582d5¢q”T–hIB„ghH6dWHQ2p4QHu58†HQdGTlDG58dŠq•58dŠDWFIHuYWH64Q582Ih†hIHQd‚BEbVTVDWBEb8H`¢BEb8H6Y æˆ( ¥ þ —• í e  çšíuäˆ6“– h  þ ‘•  DWFxRpYwhpH6nU2IBE2I`eDGFIH‘795xhIBEnUH6h@ŠTgb¢HQ2SRI7…‰†HQd‡– h q•5¢duTŽYWHQ46582IhAhIHQd‚BEbVTVDWBEb8H–TVIIdW5g—˜BX7#Tlr DWBX582‹’‰¤vBE2IhDWFIHŽ795˜hpBEnUH6hA@•Tgb8HQ2SRI7w‰sH6d–q•5¢dwDWFIHdY‚HQ4Q5¢2Ih˜r35¢dWhIH6d…4QH62SDWHQd‚HQh¬hIBEgsHQd‚HQ2I46H 58†HQdWTVDG5¢d–q•5¢d#TkYWH64Q582phºhIHQd‚BEbVTVDWBEb8HVtvDWFIH†2p582I4Q5¢7#UTV46D9q•58RId‚DWF˜r35¢dWhIH6d95¢sH6dGTVDW58d†“ æ jPë’ ¥pd ’ $$8ç–tIT¢2phŽDGFIH46587#IT¢46DŠq•5¢RId‚DWF˜r358d‚hIHQd•58sH6dGTVDW58d•hpHQdWB„b8H6h‰BE2ePSRIHQYfDGBX5¢2“p’ˆÜjyX5¢f D – h î…ß bxYQƒ ’ –Uî…ßq•58d m ¡ l –Uî…ß l j’ hp’‘¤vBE2Ih©DWFIH`8d‚BXh˜r£s5¢BX2oDGYÔr3†HQdfrŸ@ŠTgb¢HQyXH62I`¢DWF 晝0Ÿ ç…d‚HQPSRIBXd‚HQ79HQ2oDŽDW5ÅT¢4ƒFIBEH6b8HTrIFUT¢Y‚H Y sH6HQh‰HQdWd‚58duyXH6YWYŠDWFUT¢2 ¡ ’„¦usH6dW4QH62oD‘q•5¢dŠYWB —˜DWF˜r35¢dWhIH6du2I582I46587#IT¢46DCTV2Ih4Q5879UT¢4‡DC4QHQ2xr ‚ DWHQdWH6hTVIIdW5g—˜BX7#TVDWBX582IYvDG59T…nUdWYfD™hIH6dWB„b¢TlDGBEb¢H¢’ ¦ ¡ ’f•582pYWBXhpHQdvDWFIHfq•5¢yXyX5c@CBE2I`u582IH‡r£YWBEhIHQh#hIBEgsHQd‚HQ2I46BX2I`aYW4ƒFIH67#H6YQtp@CFIBE4ƒFeT¢d‚HfnUd‚Y‚D‚rGt˜Y‚HQ46582Ih˜rGt T¢2ph‰DWFIBXd‚h˜r358d‚hIHQd6tpd‚HQYW†HQ4‡DGB„b8HQy„i†Û æˆ(  ݆çšíuäÏæšÝxíU6ÅÝ˜í¥€4fçr¢î…ß æ“(  ݆çzíuäÏ暥¢Ýxíh6)—¢Ý˜í¥€4høÆÝxíA€ çrIæ   ‘ î ßsç  æ“(  Ýhçšíuä æÔ¦8¦gÝxíh6«¦ac¢ÝxíA€4hø~h¢ÝxíA€  6   ÝxíA€ „ çrIæz“¢î‘ßsç ¤ BE2Ih)DGFpHe7#5xhIBEnIHQh)@•Tcb¢HQ2SRI7…‰†HQd–q•58d…HQT¢4ƒFÆ5¢qCDGFIH6YWH‰YW4ƒFIH67#H6YQ’¬ÜjyX5¢DwDGFIHed‚HgT¢yuTV2Ih v BE7dT¢`¢BX2UT¢dfiUTVd‚DGYw5¢qf– h î…ßÆbxY6’#–Uî…ßÆq•5¢d ¡¹l –Uî‘ß l j’ãH6dWBEb¢H#DWFIHdD3@v5eyXHgTVhIBX2I` DWHQdW79YuBX2eDWFIHDGdWRp2I4gTVDWBX582H6dWdW5¢dCq•5¢duHgT¢4ƒF‰YW4ƒFIH67#HV’ .

these ODE's would be expressed in the form d~ u =F ~ (~ u t) (4.1) dt which includes all manner of nonlinear and time-dependent possibilities. ujn. 5 4 j = 2h x or h n n +un i ujn = ujn.3. On occasion. In the most general notation. Another strategy for constructing a nite-di erence approximation to a PDE is to approximate all the partial derivatives at once. As an example let us consider the model equation for di usion @u = @ u @t @x Using three-point central-di erencing schemes for both the time and space derivatives. x j 51 2 2 ( +1) ( 1) ( ) +1 ( ) 2 ( ) 1 ( +1) ( 1) 2 ( ) +1 ( ) ( ) 1 .1) which. but the rest of this chapter is devoted to a more specialized matrix notation described below. In the following chapters. we use this form. This generally leads to a point di erence operator (see Section 3.2) j j. we proceed with an analysis making considerable use of this concept.Chapter 4 THE SEMI-DISCRETE APPROACH One strategy for obtaining nite-di erence approximations to a PDE is to start by di erencing the space derivatives only. can be used for the time advance of the solution at any given point in the mesh. 2 u j j. in turn. which we refer to as the semi-discrete approach. without approximating the time derivative. u . 2 u (4. + 2h u . Di erencing the space derivatives converts the basic PDE into a set of coupled ODE's. we nd 2 n n +un 3 ujn .

It is a full discretization of the PDE. we can approximate the space derivatives with di erence operators and express the resulting ODE's with a matrix formulation. 4. There are a number of issues concerning the accuracy. In this case.2 with the time average of u at that point.3) which can be solved for u n and time advanced at the point j . 2 ( 1) ( ) 1 1 3 (4. 1 to the level n + 1. stability. + 2h 4ujn .2 and 4. Equation 4.1.52 CHAPTER 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 . and no semi-discrete form exists.2 is sometimes called Richardson's method of overlapping steps and Eq. there are subtle points to be made about using these methods to nd a numerical solution to the di usion equation. In this approach. namely (ujn + ujn. 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. 4. )=2. This results in the formula ( ) ( +1) ( 1) ujn ( +1) un = ujn. we shall separate the space di erence approximations from the time di erencing. and convergence of Eqs. 4. 2 @ j x ( 1) 2 ( ) +1 ( +1) 2 0 ( +1) + ujn.3 is referred to as the DuFort-Frankel method. As we shall see later on. 4. however.3 which we cannot comment on until we develop a framework for such investigations. Another possibility is to replace the value of ujn in the right hand side of Eq. this method has an intermediate semi-discrete form and can be analyzed by the methods discussed in the next few chapters. This is a simple and natural formulation when the ODE's are linear. the spatial and temporal discretizations are not separable. THE SEMI-DISCRETE APPROACH Clearly Eq. 4.1 Reduction of PDE's to ODE's 4. . that the spatial and temporal discretizations are separable. we reduce the governing PDE's to ODE's by discretizing the spatial terms and use the well-developed theory of ODE solutions to aid us in the development of an analysis of accuracy and stability. Thus. For the time being. Note.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. A n 5 + uj .

the 3-point central-di erencing approximation produces the ODE model given by d~ u = . using the 3-point central-di 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 .1. the elements of A depend on the solution ~ u and are usually derived by nding the Jacobian of a ux vector.5 are the model ODE's for di usion and biconvection of a scalar in one dimension. The vector ~ f (t) is usually determined by the boundary conditions and possibly source terms.5) dt 2 x p where the boundary condition vector is absent because the ow is periodic.4. a B (. 4. REDUCTION OF PDE'S TO ODE'S 53 Model ODE for Di usion For example.4) dt x ~ ) vector. that is. They are linear with coe cient matrices which are independent of x and t.4 and 4. It is used for the scalar convection model when the boundary conditions are periodic.6. In this case. Although the equations are nonlinear.1 0 1)~ u (4. Eqs. In general. 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 Navier-Stokes equations.2 1)~ u + (bc (4.1. with Dirichlet boundary conditions folded into the (bc 2 Model ODE for Biconvection The term biconvection was introduced in Section 2. 4. .~ f (t) dt (4. even the Euler and Navier-Stokes equations can be expressed in the form of Eq. In such cases the equations are nonlinear. 4.3.6) 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.2 The Generic Matrix Form The generic matrix form of a semi-discrete approximation is expressed by the equation d~ u = A~ u.

If A does depend explicitly on t.1 are said to be linear if F is linearly dependent on u (i. the general solution to Eq. 4. The ODE's represented by Eq. These ideas are developed in the following sections. the system of ODE's must be integrated using a time-marching method.2 Exact Solutions of Linear ODE's CHAPTER 4. This holds regardless of whether or ~. mI]~ xm = 0 (4. 4.6 can be written in terms of the eigenvalues and eigenvectors of A. 4. have the property that A~ xm = m~ xm or A . 4. which exist under certain conditions. 4.. if A does not depend explicitly on t. In order to analyze time-marching methods.6 in which the coe cient matrix. This will lead us to a representative scalar equation for use in analyzing time-marching methods. the general solution cannot be written whereas. m. . As we have already pointed out.54 4. not the forcing function.1 in time. if @F=@u = A where A is independent of u). f As we shall soon see. is independent of u.2. 4.e. An eigenvector. depends explicitly on t. the exact solution of Eq. when the ODE's are linear they can be expressed in a matrix notation as Eq.6 can be written. and its corresponding eigenvalue. and together they have the property that 1 2 X .8) where is a diagonal matrix whose elements are the eigenvalues of A. AX = 1 (4. we will make use of exact solutions of coupled systems of ODE's.7) The eigenvalues are the roots of the equation det A . THE SEMI-DISCRETE APPROACH In order to advance Eq. A. I] = 0 We form the right-hand eigenvector matrix of a complete system by lling its columns with the eigenvectors ~ xm : h ~ i X= ~ x x ::: ~ xM The inverse is the left-hand eigenvector matrix.1 Eigensystems of Semi-Discrete 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) . ~ xm .

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 7 J 6 7 6 7 . . 6 7 . J =6 7 6 7 Jm 4 5 ... and 2 3 m 1 1 ... 6 7 ... 6 7 m n 6 7 Jm = 6 ... 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 Second-Order
First-Order Equations
The simplest nonhomogeneous ODE of interest is given by the single, rst-order 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 steady-state solution if the right-hand 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 time-marching methods. This theme will be developed as we proceed.

56

CHAPTER 4. THE SEMI-DISCRETE 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;e t u(t) = u(0)e t + a 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.

Second-Order Equations
The homogeneous form of a second-order 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 First-Order ODE's
A Complete System
A set of coupled, rst-order, homogeneous equations is given by u0 = a u + a u u0 = a u + a u (4.14) which can be written " # 0 a a ~ ~ u = A~ u u = u u ]T 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 higher-order equation can be reduced to a coupled set of rst-order 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 SEMI-DISCRETE 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 u 6 4 u0 7 5=6 41 0 u 1
1 2 3

32 3 u 7 56 4u 7 5 u
1 2 3

if

a = u (0)
3

b = u (0)
2

1 u (0) c= 2
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, rst-order 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~ u = A~ u;~ f (t) (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 ~ u. There results ~ u = X ; AX X ; ~ X; d u ; X; ~ f (t) (4.21) dt Since A is independent of both ~ u and t, the elements in X ; and X are also independent of both ~ u and t, and Eq. 4.21 can be modi ed to
1 1 1 1 1 1 1 1 1

d X; ~ ;~ ;~ dt u = X u ; X f (t) ~ and ~ Finally, by introducing the new variables w g such that
1 1 1

~ = X; ~ w u
1

~ g(t) = X ; ~ f (t)
1

(4.22)

we reduce Eq. 4.20 to a new algebraic form

~ dw ~ ~ (4.23) dt = w ; g(t) 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 SEMI-DISCRETE APPROACH

4.23 are no longer coupled. They can be written line by line as a set of independent, single, rst-order 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) u(t) = X w =
M X m=1

wm(t)~ xm

(4.25)

where ~ xm is the m'th column of X , i.e., the eigenvector corresponding to m . We next focus on the very important subset of Eq. 4.20 when neither A nor ~ f has 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 u-system gives ~ ~ (t) u(t) = X w = = = =
M X m=1 M X m=1 M X m=1 M X m=1

wm(t)~ xm cm e m t ~ xm +
m=1 m M 1 X gm~ xm

cm e m t ~ xm + X

;1 X ;1~ f
1

cm e m t ~ xm + A; ~ f

|

Transient

{z

} |

Steady-state

{z

}

(4.26)

however. This permits us to say a great deal about such problems and serves as the basis for this section. In our application to uid dynamics.3. In particular. The second group is referred to classically as the particular solution or the particular integral. we reduce the partial di erential equations to a set of ordinary di erential equations represented by the generic form d~ u = A~ u.1 De nition . respectively.4.28) dt The dependent variable ~ u represents some physical quantity or quantities which relate to the problem of interest.27) 1 1 1 4.~ f (4. 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 most natural reference frame is the physical one. We say If a solution is expressed in terms of ~ u. That is.2. composed. the elements of A are independent of both u and t. We saw in Sections 4. we identify di erent mathematical reference frames (spaces) and view our solutions from within each. it is more instructive to refer to these groups as the transient and steady-state solutions. An alternative. we get di erent perspectives of the same solution. and this can add signi cantly to our understanding. another very useful frame. in the most general case.3 Real Space and Eigenspace Following the semi-discrete approach discussed in Section 4. How these relate to the numerical solutions of more practical problems depends upon the problem and.3. ~ f . as might be expected. it is said to be in real space.2 that pre. ~ f (4. but entirely equivalent. We begin by developing the concept of \spaces".1. to a much greater extent. we can develop some very important and fundamental concepts that underly the global properties of the numerical solutions to the model problems.and post-multiplication of A by the appropriate similarity matrices transforms A into a diagonal matrix. There is. of Jordan blocks or. form of the solution is ~ u(t) = c e 1 t ~ x + + cme m t ~ xm + + cM e M t ~ x M + A. REAL SPACE AND EIGENSPACE 1 61 Note that the steady-state solution is A. on the cleverness of the relator. In this way. in the 4.1 and 4. For the model problems on which we are focusing most of our attention.

and 2. the solutions in the two spaces are represented by X ~ u(t) = cm e m t ~ xm + X . The relations that transfer from one space to the other are: ~ = X. the transient portion of the solution in real space consists of a linear combination of contributions from each eigenvector. the ~ elements of w are linear combinations of all of the elements of ~ u. g which is an uncoupled set of rst-order ODE's that can be solved independently for ~ . X .3. 4.2 and. m 4.62 CHAPTER 4. ~ f 1 1 ~ ~ u = Xw ~ f = X~ g The elements of ~ u relate directly to the local physics of the problem. for simplicity. the transient portion of the solution in eigenspace provides the weighting of each eigenvector component of the transient solution in real space. using only complete systems for our examples. and individually they have no direct local physical interpretation. ~ f and wm(t) = cm e m t + 1 gm m=1 2 M 1 1 for real space and eigenspace. ~ w u ~ g = X.1. At this point we make two observations: 1. it is said to If a solution is expressed in terms of w be in eigenspace (often referred to as wave space). When the forcing function ~ f is independent of t. Following Section 4. we found that Eq. However. These model ODE's are presented in Section 4. of scalars. We say the dependent variable vector w ~ . THE SEMI-DISCRETE APPROACH simplest nondefective case.28 had the alternative form ~ dw ~ ~ dt = w . Equations for the eigenvalues of the simple . respectively.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.

Notice that the di usion eigenvalues are real and negative while those representing periodic convection are all pure imaginary. and x = 2 =M .4.ia sin 2m m=0 1 M . we present results for a simple 4-point mesh and then we give the general case. for the model biconvection equation . we nd for the model di usion equation with Dirichlet boundary conditions m . B (M : a b c) and Bp(M : a b c). from Section B. REAL SPACE AND EIGENSPACE 63 tridiagonals. 4. Consider Eq.1 (4. The right-hand 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 7 6 4 sin (x ) sin (2x ) sin (3x ) sin (4x ) 7 5 sin (x ) sin (2x ) sin (3x ) sin (4x ) The columns of the matrix are proportional to the eigenvectors.31) x is the modi ed wavenumber from Section 3.4.3 Eigenvectors of the Model Equations The Di usion Model Next we consider the eigenvectors of the two model equations. m = m. These follow as special cases from the results given in Appendix B.30) x m=0 1 = sin m M . are given in Appendix B. the model ODE's for di usion. First.1 (4.4 sin 2(M + 1) x and.3.3. From Section B.1 m = x M 2 2 2 where = . so in general the relation ~ u = Xw ~ can be written as 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. The interpretation of this result plays a very important role later in our stability analysis. 4.5.i m a m m=0 1 M . to help visualize the matrix structure.1. 2 + 2 cos M m = +1 x ! m m=1 2 M (4.4.29) = .32) . Recall that xj = j x = j =(M + 1).

.1 X m=0 (2 m=M ) j =0 1 m =0 1 M . 4. Similarly. we can write ~ u = Xw ~ as uj = wmeimxj j=0 1 (4. or left-hand eigenvector matrix X . For our model ODE. THE SEMI-DISCRETE APPROACH For the inverse.32 represents the sine synthesis that companions the sine transform given by Eq. Eq.34) .1 M .1 x = j 2 =M . In general w ~ = X. 1 The Biconvection Model Next consider the model ODE's for periodic convection.64 CHAPTER 4. Eq. we nd 1 2 sin (x ) 6 sin (2x ) 6 6 4 sin (3x ) 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. Eq. ~ ~ u = Xw is a sine synthesis from wave space back to real space. For the model di usion equation: ~ = X. In summary. the right-hand eigenvectors are given by ~ xm = ei j With xj = j M . ~ w u is a sine transform from real space to (sine) wave space.33. 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. ~ u gives wm = M X j =1 uj sin mxj m=1 2 M (4.33) In the eld of harmonic analysis. The coe cient matrices for these ODE's are always circulant. 4.1 M . 4.5.

we nd the following left-hand eigenvector matrix from Appendix B. i = e. i = 1 1 e e e 7 6 7 = 6 5 4 4 1 e. it is straightforward to establish the fact that the relation ~ ~ represents the Fourier synthesis of the variable w ~ back to ~ u = Xw u. =0 In general For any circulant system: ~ = X.27 becomes We can now combine the results of the previous sections to write the solutions of our model ODE's.u 7 5=X ~ 1 M . In summary. Eq. i = 1 e. x. uj (t) = where M X m=1 cme m t sin mxj + (A. f )j 1 j=1 2 M (4. i = 2 4 6 4 4 4 4 8 4 4 6 12 18 12 4 1 65 For a 4-point periodic mesh.3. i = .4. i = e. X 1 wm = M uj e.4 Solutions of the Model ODE's The Di usion Equation For the di usion equation.1 j This equation is identical to a discrete Fourier transform of the periodic dependent variable ~ u using an M -point sample between and including x = 0 and x = 2 . i = e. For circulant matrices.4: 1 2 3 4 4 4 4 32u 7 6 u 7 6 7 6 54u u 1 2 3 4 3 7 7 .35) m m = . ~ ~ u = Xw is a complex Fourier synthesis from wave space back to real space. 4. REAL SPACE AND EIGENSPACE 2w 6 w 6 6 4w w 3 21 1 1 1 7 6 .4 sin 2(M + 1) x 2 2 ! (4. i = e. i = .36) .imxj m=0 1 M .3. 1 4. ~ w u is a complex Fourier transform from real space to wave space.

31.41) with the modi ed wavenumber de ned in Eq. 2. THE SEMI-DISCRETE APPROACH (4. This di erence causes the various modes (or eigenvector components) to decay at rates which di er from the exact solution.39) We see that the solutions are identical except for the steady solution and the modi ed wavenumber in the transient term.66 CHAPTER 4.37) With the modi ed wavenumber de ned as 2 m x m = x sin 2 and using m = m. evaluated at the nodes of the grid: uj (t) = M X m=1 cm e.1 X m=0 cme. low wavenumber modes are accurately represented. We can write this ODE solution as uj (t) = M . while high wavenumber modes (if they have signi cant amplitudes) can have large errors.37.40) m = .1 (4. m 2 t sin m xj + h(xj ) j=1 2 M (4. m 2 t sin m xj + (A. f )j 1 j=1 2 M (4. The modi ed wavenumber is an approximation to the actual wavenumber.1 X m=0 cme m t ei mxj j=0 1 M . With conventional di erencing schemes. 4.1 (4.i m a (4. we can write the ODE solution as uj (t) = M X m=1 cme. The Convection Equation For the biconvection equation.38) This can be compared with the exact solution to the PDE. The di erence between the modi ed wavenumber and the actual wavenumber depends on the di erencing scheme and the grid resolution.i m at ei mxj j=0 1 M . Eq. we obtain uj (t) = where M .42) .

g (t) (4.4 The Representative Equation In Section 4. evaluated at the nodes of the grid: M .5.42 shows that the imaginary portion of the modi ed wavenumber produces nonphysical decay or growth in the numerical solution.1 (4. For such a purpose Eq. The role of m is clear it stands for some representative eigenvalue in the original A matrix. X uj (t) = fm (0)e. 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.26. Since the error in the phase speed depends on the wavenumber. We found the uncoupled solution to a set of ODE's in Section 4.4.44) m m m dt The goal in our analysis is to study typical behavior of general situations. solved there in their uncoupled form. the result is erroneous numerical dispersion. uncoupled equation and our conclusions will apply in general. This is helpful both in analyzing the accuracy of time-marching methods and in studying their stability. A typical member of the family is dwm = w .20 and 4. 4. In the case of non-centered di erencing. this leads to an error in the speed with which various modes are convected. m=0 4.23 express identical results but in terms of di erent groupings of the dependent variables. Our next objective is to nd a \typical" single ODE to analyze. The form of Eq. Eq. m As discussed in Section 3. THE REPRESENTATIVE EQUATION 1 67 and compare it to the exact solution of the PDE. . the modi ed wavenumber is complex. 2. 4.3. 4. while the actual phase speed is independent of the wavenumber. The importance of this concept resides in its message that we can analyze timemarching methods by applying them to a single. which are related by algebraic manipulation. since is real. topics which are covered in Chapters 6 and 7.4. discussed in Chapter 11.44 is not quite satisfactory.43) Once again the di erence appears through the modi ed wavenumber contained in . we pointed out that Eqs. However.i mat ei m xj j=0 1 M . and then returned as a coupled set to real space.2. not particular problems.

44 has the exact solution: k which can be used to evaluate all manner of time-marching methods.46) w(t) = ce t + ae . Write the semi-discrete form resulting from the application of second-order centered di erences to the following equation on the domain 0 x 1 with boundary conditions u(0) = 0. The exact solution of the representative ODE is (for 6= ): t (4. 3.68 CHAPTER 4. 4. 4. From this we can extract the k'th term and replace ik with .g(t) = ak eikt X ak eikt k ik .5 Problems 1. u(1) = 1: @u = @ u . For example X . Using this operator to approximate the spatial derivative in the linear convection equation. Find the entries in the boundary-condition vector. 3. in principle. Consider the application of the operator given in Eq. In such evaluations the parameters and must be allowed to take the worst possible combination of values that might occur in the ODE eigensystem. write the semi-discrete form obtained with periodic boundary conditions on a 5-point grid (M = 5). Write the resulting semi-discrete ODE form. This leads to w(t) = ce t + The Representative ODE dw = w + ae t dt (4.45) for which Eq. THE SEMI-DISCRETE APPROACH the question is: What should we use for gm(t) when the time dependence cannot be ignored? To answer this question. 6x @t @x 2 2 . we note that. Consider the nite-di erence operator derived in question 1 of Chapter 3. 2. one can express any one of the forcing terms gm(t) as a nite Fourier series.52 to the 1-D di usion equation with Dirichlet boundary conditions.

1 0 1)=(2 x) corresponding to the ODE form of the biconvection equation resulting from the application of second-order central di erencing on a 10-point grid.4. plot the exact solution of the di usion equation with = 1 at t = 1 with m = 1 and m = 3. The grid nodes are given by xj = j x j = 0 1 : : : 9. Compare with the exact solution of the PDE. For initial conditions u(x 0) = sin(mx) and boundary conditions u(0 t) = u( t) = 0. Using these.4. Note that the domain is 0 x 2 and x = 2 =10. 1 .Bp(10 . can be found from Appendix B. (Plot the solution at the grid nodes only. as well as the matrices X and X . Repeat for the initial condition u(x 0) = sin 2x. Consider a grid with 10 interior points spanning the domain 0 x . Consider the matrix A = . The eigenvalues of the above matrix A.37. PROBLEMS 69 4. compute and plot the ODE solution at t = 2 for the initial condition u(x 0) = sin x.5. 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. Calculate and plot the corresponding ODE solutions. 4.) Calculate the corresponding modi ed wavenumbers for the second-order centered operator from Eq. 5. .

70 CHAPTER 4. THE SEMI-DISCRETE APPROACH .

First.Chapter 5 FINITE-VOLUME METHODS In Chapter 3. we saw that the application of a nite-di erence approximation to the spatial derivatives in our model PDE's produces a coupled set of ODE's. In this Chapter.1 or Eq. of two advantages.2. While this property can usually be obtained using a nite-di erence formulation. we will show how similar semi-discrete forms can be derived using nite-volume approximations in space. they ensure that the discretization is conservative. and energy are conserved in a discrete sense. In Chapter 4. Finite-volume methods have become popular in CFD as a result. Second.1. Consistent with our emphasis on semi-discrete methods. we will study the latter form. As a result. it is obtained naturally from a nite-volume formulation. either in the form of Eq. primarily. i. Finite-volume methods are applied to the integral form of the governing equations. they can be applied on unstructured meshes consisting of arbitrary polyhedra in three dimensions or arbitrary polygons in two dimensions. 71 . we saw how to derive nite-di erence approximations to arbitrary derivatives. This will be followed by several examples which hopefully make these concepts clear. momentum. This increased exibility can be used to great advantage in generating grids about arbitrary geometries. mass. Next we will give our model equations in the form of Eq.1) We will begin by presenting the basic concepts which apply to nite-volume strategies. 2.e.. 5. which is d Z QdV + I n:FdS = Z PdV dt V (t) S (t) V (t) (5. 2. nitevolume methods do not require a coordinate transformation in order to be applied on irregular meshes.

which is a closed surface in three dimensions and a closed contour in two dimensions. Another approach known as ux-di erence splitting is described in Chapter 11. Thus after applying a timemarching method. Similarly. When these are used to calculate F(Q). This ux must then be integrated to nd the net ux through the boundary.1 Basic Concepts CHAPTER 5. . 5. FINITE-VOLUME METHODS The basic idea of a nite-volume 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. that is. Examining Eq.72 5.4) for a control volume which does not vary with time. Q can be represented within the cell by some piecewise approximation which produces the correct value of Q. we have updated values of the cell-averaged quantities Q. In order to evaluate the uxes. Thus the volume V in Eq. The basic elements of a nite-volume method are thus the following: 1 Time-marching methods will be discussed in the next chapter. In our examples. As we shall see in our examples. First. we see that several approximations must be made. This is a form of interpolation often referred to as reconstruction. The ux is required at the boundary of the control volume. which are a function of Q. each cell will have a di erent piecewise approximation to Q. 5. the ux will be discontinuous.1. they will generally produce di erent approximations to the ux at the boundary between two control volumes.3) d Q + I n:FdS = Z PdV V dt S V (5. Let us consider these approximations in more detail.2) at the next time step. Next a time-marching method1 can be applied to nd the value of Z V QdV (5. at the control-volume boundary. 5.1 can be written as 1 Z QdV V V (5. the source term P must be integrated over the control volume. we will consider only control volumes which do not vary with time.1 is that of a control volume whose shape is dependent on the nature of the grid. we note that the average value of Q in a cell with volume V is Q and Eq. A nondissipative scheme analogous to centered di erencing is obtained by taking the average of these two uxes.

Using this approximation. Advance the solution in time to obtain new values of Q.1 The Linear Convection Equation A two-dimensional form of the linear convection equation can be written as @u + a cos @u + a sin @u = 0 @t @x @y (5. Integrate the ux to nd the net ux through the control-volume boundary using some sort of quadrature. 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. Evaluate F(Q) at the boundary.6) 5.5) or. the following relation between rQ and Q is sometimes used: Z I rQdV = nQdS (5. These ideas should be clari ed by the examples in the remainder of this chapter.5. The one-dimensional form is recovered with = 0. MODEL EQUATIONS IN INTEGRAL FORM 73 1. .2. construct an approximation to Q(x y z) in each control volume. nd Q at the control-volume boundary. V S Z where the unit vector n points outward from the surface or contour.2. 3. This issue is discussed in Section 11. 4.7) This PDE governs a simple plane wave convecting the scalar quantity.2 Model Equations in Integral Form 5. in two dimensions. A rQdA = C nQdl I (5. 2. Since there is a distinct approximation to Q(x y z) in each control volume.2. two distinct values of the ux will generally be obtained at any point on the boundary between two control volumes. In order to include di usive uxes. u(x y t) with speed a along a straight line making an angle with respect to the x-axis. Given the value of Q for each control volume. The order of accuracy of the method is dependent on each of the approximations.4.

d Z udA = I n: i @u + j @u ds dt A @x @y C (5.3. x=2 xj+1=2 = xj + x=2 (5. i @x + j @y P = 0 Using these.15) ! to be the integral form of the two-dimensional di usion equation.2 gives the following integral form (5. Eq.ru ! @u @u = . We will use the following notation: xj.2. The nodes of the grid are located at xj = j x as usual.10) Since Q is a scalar.17) . x=2 to xj + x=2.11) where A is the area of the cell which is bounded by the closed contour C . Eq. Substituting these expressions into a twodimensional form of Eq. 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. as in the model equations. as shown in Fig. with 5. 2.8) (5. F is simply a vector. the two-dimensional linear convection equation is obtained from the general divergence form.16) 5. FINITE-VOLUME METHODS For unit speed a. Control volume j extends from xj .9) (5. We consider an equispaced grid with spacing x.74 CHAPTER 5. we nd (5.1=2 = xj .2 The Di usion Equation Q = u F = .14) (5. The integral form of the two-dimensional di usion equation with no source term and unit di usion coe cient is obtained from the general divergence form.1. 5. 2.3.12) (5.3 One-Dimensional Examples We restrict our attention to a scalar dependent variable u and a scalar ux f . 2.13) (5.

becomes uj + f .1=2 Now with = x . the cell-average value becomes Z xj +1=2 1 u (t) u(x t)dx j uj 1=2 = u(xj 1=2) x fj 1=2 = f (uj 1=2) (5. 5. f (5.3.11.23) x ddt j +1=2 j . 5. Hence the cell-average value and the value at the center of the cell di er by a term of second order. we can expand u(x) in Eq.5.3.20) j j +1=2 j . x=2 uj + @x j + 2 @x2 j + 6 @x3 j + : : : d 2 @2u ! 4 @4u ! x x = uj + 24 @x2 + 1920 @x4 + O( x6 ) (5. f = Pdx (5.22) where uj is the value at the center of the cell.1 A Second-Order Approximation to the Convection Equation In one dimension. With these de nitions.1=2 = 0 . the integral form of the linear convection equation. Eq.19) Z xj +1=2 d ( xu ) + f .18) (5. xj . ONE-DIMENSIONAL EXAMPLES j-1/2 ∆ x j+1/2 75 j-2 j-1 L R j L R j+1 j+2 Figure 5.19 in a Taylor series about xj (with t xed) to get 2 3 ! Z x=2 2 @2u ! 3 @3u ! @u 1 4 5 uj x .1=2 dt xj. uj = uj + O( x2) 5.1=2 or (5.1: Control volume in one dimension.21) j j and the integral form becomes xj.

giving fjL.1=2 = uj and cell j .24) Evaluating this at j + 1=2 gives L fjL+1=2 = f (uL (5.1=2 x xj+1=2 (5.1=2 j .1 j where f^ denotes a numerical ux which is an approximation to the exact ux.1. 5. we obtain uj + 1 (u + u ) .1 With periodic boundary conditions.1=2 = 2 (5.1=2 = uj. as shown in Fig.29 and 5. u ) = 0 (5.30) j .1=2 2 j.1 (5.1 j dt 2 j+1 j. 1 is the cell to the left of xj. Eq. 5.25) j +1=2) = uj +1=2 = uj 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 .27) .28) We have now accomplished the rst step from the list in Section 5. giving fjR (5. We choose a piecewise constant approximation to u(x) in each cell such that u(x) = uj xj. cell j is the cell to the right of xj. the discontinuity in the ux at the cell boundary is resolved by taking the average of the uxes on either side of the boundary.1=2.23.26) +1=2 = uj +1 Similarly. Substituting Eqs. this point operator produces the following semidiscrete form: d~ u = . 1 (u + u ) = x duj + 1 (u . 5. In this example. The cell to the right of xj +1=2.1 0 1)~ u (5. FINITE-VOLUME METHODS with f = u.76 CHAPTER 5.1=2 .30 into the integral form.32) dt 2 x p . which is cell j + 1.31) x ddt 2 j j+1 2 j. Thus 1 (f L + f R ) = 1 (u + u ) f^j+1=2 = 2 (5. gives fjR (5. 1 B (.1 we have de ned the uxes at the cell boundaries in terms of the cell-average data.29) j +1=2 j +1=2 2 j j+1 and 1 (f L + f R ) = 1 (u + u ) f^j.

b.35) the following values at the cell boundaries: 1 uL j +1=2 = 6 (2uj +1 + 5uj . uj . ONE-DIMENSIONAL EXAMPLES 77 This is identical to the expression obtained using second-order centered di erences.1 0 1) is relevant to nite-volume methods as well as nite-di erence methods.29 and 5. as in Eqs.3. except it is written in terms of the cell average ~ u.uj.1 + 26 24 With these values of a.5. we can conclude that the use of the average of the uxes on either side of the cell boundary. ~ u. x=2 u( )d = u x . Since the eigenvalues of Bp(.22ux 2 uj.36) . can lead to a nondissipative nite-volume method.3.3 x=2 1 j . uj+1 c = . rather than the nodal values.30.1 x Z x=2 .34) 1 These constraints lead to x Z 3 x=2 u( )d = uj+1 j + uj .1 0 1) are pure imaginary. b.1 a = uj+1 . and c are chosen to satisfy the following constraints: 1 Z . x=2 x=2 u( )d = uj (5.x (5.3.1 b = uj+12 . 5. xj . and c. The three parameters a.2 A Fourth-Order Approximation to the Convection Equation Let us replace the piecewise constant approximation in Section 5.1) uj . Hence our analysis and understanding of the eigensystem of the matrix Bp(.33) where is again equal to x . the piecewise quadratic approximation produces (5. 5.1 with a piecewise quadratic approximation as follows u( ) = a 2 + b + c (5.

while the second approach is more suited to extension to higher order accuracy.3. we again use the average of the uxes on either side of the boundary to obtain f (uL ) + f (uR f^j+1=2 = 1 j +1 = 2 j +1=2 )] 2 1 (.1 j .u + 7u + 7u .1) 1 uR j +1=2 = 6 (.2) 12 (5.2 12 j+2 This is a fourth-order approximation to the integral form of the equation.uj+1 + 7uj + 7uj.8 0 8 . the following semi-discrete form is obtained: d~ u = .1=2 = 6 (2uj + 5uj . Recalling that f = u. u ) (5.2) (5.1.1=2 = 6 (.uj +2 + 5uj +1 + 2uj ) 1 uL j .1)~ u (5.uj +1 + 5uj + 2uj .37) (5. Eq.40) = 12 j +2 j +1 j j .78 CHAPTER 5.1=2 = 2 j .39) using the notation de ned in Section 5. 5.1=2 j . With periodic boundary conditions.1 and 1 f (uL ) + f (uR )] f^j. 5. uj . 1 B (1 . FINITE-VOLUME METHODS 1 uR j .42) j +1 j .1 . gives uj + 1 (.1 .23. The rst approach is simpler to extend to multidimensions.43) dt 12 x p This is a system of ODE's governing the evolution of the cell-average data. 8u + u ) = 0 x ddt (5.38) (5.1=2 = 1 (.3. uj. as can be veri ed using Taylor series expansions (see question 1 at the end of this chapter). In this section. we describe two approaches to deriving a nite-volume approximation to the di usion equation.u + 8u .41) Substituting these expressions into the integral form.3 A Second-Order Approximation to the Di usion Equation .

47) j +1=2 @x j+1=2 x j+1 j Similarly. ONE-DIMENSIONAL EXAMPLES uj + f . with Dirichlet boundary conditions.1=2 = 0 Zb 79 (5. Eq.50) j +1 This provides a semi-discrete nite-volume approximation to the di usion equation.3.2 1)~ ~c u+ b 2 f^j.44) (5.1 j (5.5.@u=@x.1) dt x j . d~ u = 1 B (1 .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 . we can write ! @u f^ = .33 we have dt x @u = @u = 2a + b @x @ (5. Eq. the integral form of the di usion equation. f x ddt j +1=2 j .48) (5. we use a piecewise quadratic approximation as in Section 5.2. u ) (5. u(a) @x x xj @x x j+1 j From Eq. and we see that the properties of the matrix B (1 . 5.16.ru = . 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 second-order accuracy. = . Hence.46) @u dx = u(b) . 5. 5. u ) (5.51) . 5.49) (5. we obtain x duj = 1 (u . From Eq. Eq.45) In one dimension.44.1=2 = .22.3. 2u + u ) or. For our second approach. Substituting these into the integral form. 5.2 1) are relevant to the study of nite-volume methods as well as nite-di erence methods. 1x (uj . uj. to second-order. becomes with f = . 1 (u . Also.

1=2 = . This produces duj = 1 (u .2.35. The resulting semi-discrete form with periodic boundary conditions is d~ u = 1 B (1 . uj . FINITE-VOLUME METHODS j +1=2 j +1=2 with a and b given in Eq.3. and A. 5.1 j j +1 dt x2 p 5. As in Section 5.52) (5.1=2 = fj .54) which is identical to Eq. x (uj . 1 ` = p 3 p A = 3 2 3 `2 ` = 2 (5. With f = . 1 (u . is the length of the sides of the triangles. 1 L fjR . as shown in Figure 5. 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 control volumes are regular hexagons with area A. The following relations hold between `.@u=@x.57) .4 A Two-Dimensional Example The above one-dimensional examples of nite-volume approximations obscure some of the practical aspects of such methods.80 CHAPTER 5. and ` is the length of the sides of the hexagons.49.2 1)~ u (5. this gives f R = f L = . u ) x j+1 j (5. Thus our nal example is a nite-volume approximation to the two-dimensional linear convection equation on a grid consisting of regular triangles. The nodal data are stored at the vertices of the triangles formed by the grid.1. 5.56) A 3 The two-dimensional form of the conservation law is d Z QdA + I n:Fdl = 0 dt A C (5. .1 ) dt x2 j . 2u + u ) (5.53) Notice that there is no discontinuity in the ux at the cell boundary.55) which is written entirely in terms of cell-average data.

. 3 j)=2 pi (i + p 3 j)=2 (.i + 3 j)=2 . A TWO-DIMENSIONAL EXAMPLE c b 2 3 81 p d 4 1 l ∆ 0 5 a e f Figure 5.2. 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.1. pi (. Side 0 1 2 3 4 5 Outward Normal n (i . respectively.4. Outward normals. where we have ignored the source term.i .1. 5.2: Triangular grid. Since these sides are all straight. as shown in Figure 5. 3 j)=2 p Table 5.5. see Fig. The contour in the line integral is composed of the sides of the hexagon.2. A list of the normals for the mesh orientation shown is given in Table 5. i and j are unit normals along x and y.

`=2 u ( )d (5. cos p (.59) Then. 5. we have " Z # 5 1=2 d Z u dA + X n (i cos + j sin ) ` u(z)dz = 0 (5.62) 2 .60) dt A . cos + 3 sin )=2 .61) . That is.60. 3 sin )=2 p Table 5.82 CHAPTER 5. Taking the average of the ux on either side of each edge of the hexagon gives for edge 1: Z 1=2 Z u + u p a dz = up + ua u (z)dz = (5. 5. we have for side n Z Fdl = n (i cos + j sin ) Z `=2 .`=2 Z `=2 Z 1=2 are no numerical approximations in Eq. see Eq. Making the change of variable z = =`. the approximation to the ux integral becomes trivial.1=2 2 1 A u dA = Aup (5. Side 0 1 2 3 4 5 n (i cos + j sin ) (cos .1=2 =0 The values of n (i cos + j sin ) are given by the expressions in Table 5. Z and the piecewise-constant approximation u = up over the entire hexagon. Introducing the cell average. Weights of ux integrals. the integrated time rate of change of the integral of u over the area of the hexagon is known exactly. There .2.2.58) where is a length measured from the middle of a side . 3 sin )=2 cos p (cos + p 3 sin )=2 (.60.11. the two-dimensional linear convection equation. in terms of u and the hexagon area A.1=2 u(z)dz (5. cos . 5. one has the expression u( )d = ` . FINITE-VOLUME METHODS For Eq. if the integrals in the equation are evaluated exactly.

5. PROBLEMS Similarly. we obtain uf u (z)dz = up + 2 0 up + ` (2 cos )(u . 5.5 Problems 1.65) (5.2. uf )] = 0 (5. uf )] = 0 (5. u ) A ddt b e a d 2 p +(. Use Taylor series to verify that Eq. Find the semi-discrete ODE form governing the cell-average data resulting from the use of a linear approximation in developing a nite-volume method for the linear convection equation. we have for the other ve edges: Z 83 ub u (z)dz = up + 2 2 uc u (z)dz = up + 2 3 (5.66) (5.67) Z Z ud u (z)dz = up + 2 4 ue u (z)dz = up + 2 5 Z Z Substituting these into Eq. that this is a second-order approximation to the integral form of the two-dimensional linear convection equation. cos + 3 sin )(uc .60. along with the expressions in Table 5. Use the following linear approximation: u( ) = a + b . 5. cos + 3 sin )(uc . u ) a d b e dt 3 p +(.68) or dup + 1 (2 cos )(u . 5.69) The reader can verify. 2.5.23.42 is a fourth-order approximation to Eq.64) (5. using Taylor series expansions. u ) + (cos + p3 sin )(u .5.63) (5. u ) + (cos + p3 sin )(u .

1 a = uj+12 .84 where b = uj and CHAPTER 5.3.3. 4. Using the rst approach given in Section 5. derive a nite-volume approximation to the spatial terms in the two-dimensional di usion equation on a square grid. 3.x and use the average ux at the cell interface. Repeat question 3 for a grid consisting of equilateral triangles. . FINITE-VOLUME METHODS uj.

topics which are covered in the next two chapters. or directly using Gaussian elimination or some variation thereof. one can consider a time-dependent path to the steady state and use a time-marching method to integrate the unsteady form of the equations until the solution is su ciently close to the steady solution.). the goal is simply to remove the transient portion of the solution as quickly as possible timeaccuracy is not required. time-marching methods for ODE's. Alternatively. This produces an iterative method in which a coupled system of linear algebraic equations must be solved at each iteration. 85 d~ u =F ~ (~ u t) dt (6. The subject of the present chapter. one can consider the use of Newton's method.10. When using a time-marching method to compute steady ows. is thus relevant to both steady and unsteady ow problems. which will be discussed in Chapter 9. These can be solved iteratively using relaxation methods. some sort of iterative method is required to obtain a solution.1) . which is widely used for nonlinear algebraic equations (See Section 6. For a steady ow problem. This motivates the study of stability and sti ness.Chapter 6 TIME-MARCHING METHODS FOR ODE'S After discretizing the spatial derivatives in the governing PDE's (such as the NavierStokes equations).3. spatial discretization leads to a coupled system of nonlinear algebraic equations in the form ~ (~ F u) = 0 (6. For example.2) As a result of the nonlinearity of these equations. we obtain a coupled system of nonlinear ODE's in the form These can be integrated in time using a time-marching method to obtain a timeaccurate solution to an unsteady ow problem.

the reader should recall the arguments made in Chapter 4 to justify the study of a scalar ODE. We then face a further task concerning the numerical stability of the resulting methods.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 ~. They are both commonly used in the literature on ODE's. TIME-MARCHING METHODS FOR ODE'S Application of a spatial discretization to a PDE produces a coupled system of ODE's. Combining this notation with Eq. but we postpone such considerations to the next chapter. we need only consider the scalar case Although we use u to represent the dependent variable.1u0n.1 0 un .4) un+1 = f 1 u0n+1 0u0n . In this chapter we will present some basic theory of linear O E's which closely parallels that for linear ODE's. we reduce our PDE to a set of coupled ODE's represented in general by Eq.1 un. we developed exact solutions to our model PDE's and ODE's. and. 6. using this theory. 4. for the purpose of this chapter.1 du = u0 = F (u t) dt (6. The methods we study are to be applied to linear or nonlinear ODE's. For these we use the notation ~n+ = F (~ u ~0n+ = F un+ tn + h) The choice of u0 or F to express the derivative in a scheme is arbitrary.1. 6.3 to some given accuracy. we introduced the convention that the n subscript. always points to a discrete time value. or the (n) superscript. However.3) . and h represents the time interval t. but the methods themselves are formed by linear combinations of the dependent variable and its derivative at various time intervals.86 CHAPTER 6. where accuracy can be measured either in a local or a global sense. In Chapter 2. They are represented conceptually by (6.1 Notation Using the semi-discrete approach. rather than w. etc. 6. Our rst task is to nd numerical approximations that can be used to carry out the time integration of Eq. we will develop exact solutions for the model O E's arising from the application of timemarching methods to the model ODE's. In earlier chapters. u. Application of a time-marching method to an ODE produces an ordinary di erence equation (O E ).

The material presented in Chapter 4 develops a basis for evaluating such methods by introducing the concept of the representative equation u ~n+1 = un + hu0n 1 u +u ~0n+1] un+1 = 2 n ~n+1 + hu (6. which is a fully-discrete method. that is. the new predicted solution is also a function of the time derivative at the new time level. An explicit method is one in which the new predicted solution is only a function of known data. implicit methods require more complicated strategies to solve for un+1 than explicit methods. 6.1. For an implicit method.1 respectively. u0n+1. the rst and third of these are examples of explicit methods. These methods are simple recipes for the time advance of a function in terms of its value and the value of its derivative.1 for a method using two previous time levels.4. for example. we can analytically convert MacCormack's method. The second is implicit and referred to as the implicit (or backward) Euler method. these methods can be constructed to give a local Taylor series accuracy of any order.2 Converting Time-Marching Methods to O E's Examples of some very common forms of methods used for time-marching general ODE's are: un+1 = un + hu0n un+1 = un + hu0n+1 and (6.6. un and un. u0n. will be presented in Section 11. As we shall see. for systems of ODE's and nonlinear problems. The value of this equation arises from the fact that. u0n.5) (6. We refer to them as the explicit Euler method and the MacCormack predictor-corrector method. u.7) du = u0 = u + ae t (6. The method commonly referred to as .6) According to the conditions presented under Eq. by applying a time-marching method. CONVERTING TIME-MARCHING METHODS TO O E'S 87 With an appropriate choice of the 0s and 0s.8) dt written here in terms of the dependent variable. 6. and therefore the time advance is simple. at given time intervals. The methods are said to be explicit if 1 = 0 and implicit otherwise.3 1 Here we give only MacCormack's time-marching method.2.

Eq.9) Eq.8. (1 + h)un = ahe hn 1 h(n+1) 1 (1 + h)~ un+1 + un+1 . 6. 1 u (6.11 is obtained by noting that u ~0n+1 = F (~ un+1 tn + h) = u ~n+1 + ae h(n+1) (6. As another example.11 is identical to Eq. to Eq. 6. 6. The second line in Eq.9. and just as powerful as.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. 6.7 is simply the explicit Euler method.9 is a linear O E . since the predictor step in Eq. un = he h ae hn (6. Eq. TIME-MARCHING METHODS FOR ODE'S such a linear ODE into a linear O E . h)un+1 . expressed in terms of the dependent variable un and the independent variable n. This is demonstrated by repeating some of the developments in Section 4.2 n = ahe 2 2 which is a coupled set of linear O E's with constant coe cients.12) Now we need to develop techniques for analyzing these di erence equations so that we can compare the merits of the time-marching methods that generated them.88 CHAPTER 6.2. the predictor-corrector sequence. We next consider examples of this conversion process and then go into the general theory on solving O E's.5. (1 + h)un = hae hn (6. Apply the simple explicit Euler scheme. to Eq. we nd un+1 = un + h un+1 + ae h(n+1) or (1 . applying the implicit Euler method.11) . 6.10) As a nal example. 6. . 6. 6.8. the theory of ODE's. There results un+1 = un + h( un + ae hn) or un+1 . with constant coe cients. Eq. gives u ~n+1 . but for di erence rather than di erential equations.6. The latter are subject to a whole body of analysis that is similar in many respects to. 6. 6. 6. Note that the rst line in Eq.7.

complex parameters. The exact solution of Eq.and Second-Order Di erence Equations First-Order Equations The simplest nonhomogeneous O E of interest is given by the single rst-order equation un+1 = un + abn (6. 6. is not a function of either n or u. In terms of the initial value of u it can be written Just as in the development of Eq.13) where . The independent variable is now n rather than t. thus b bn = bn+ = E bn . (b = ). one can readily show that the solution of the defective case. n Second-Order Equations The homogeneous form of a second-order di erence equation is given by un+2 + a1 un+1 + a0un = 0 (6.1 First.10. n. in general. a. SOLUTION OF LINEAR O E'S WITH CONSTANT COEFFICIENTS 89 6.6. un+1 = un + a n is h i un = u0 + an . and since the equations are linear and have constant coe cients.1 n This can all be easily veri ed by substitution. n un = c1 n + bab . and b are. 4. un = u0 n + a b b .3.3.13 is where c1 is a constant determined by the initial conditions. we use for O E's the Instead of the di erential operator D dt 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.14) d used for ODE's.

17) which can also be written ~ un+1 = C~ un (c11 .17 can be written u1 (n) = C . 6. this time having the form P (E ) = det C .14 can now be re-expressed in an operational notion as (E 2 + a1E + a0)un = 0 (6. 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. The -roots are found from P ( ) = det (c11c . ) (c c12 =0 21 22 . E I ]. 6.3. The fact that Eq. Thus Eq. TIME-MARCHING METHODS FOR ODE'S where can be any fraction or irrational number. ) . Again we are led to a characteristic polynomial. 6. 6. 2 . 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. rst-order. They are found by solving the equation P ( ) = 0. Eq. E I ]~ un = 0 c21 (c22 . linear homogeneous di erence equations have the form u(1n+1) = c11u(1n) + c12 u(2n) u(2n+1) = c21u(1n) + c22 u(2n) h iT ~ un = u(1n) u(2n) c12 (6. and refer to them as the -roots.14. E ) u2 which must be zero for all u1 and u2. . In the analysis of O E's. E ) 11 c12 C= c c21 c22 The operational form of Eq.2 Special Cases of Coupled First-Order Equations A Complete System Coupled.16) where c1 and c2 depend upon the initial conditions.14 for all c1 c2 and n should be veri ed by substitution. we label these roots 1 . 6. 6. there are just two roots and in terms of them the solution can be written un = c1( 1 )n + c2( 2 )n (6. etc.90 CHAPTER 6. its roots determine the solution to the O E.15) which must be zero for all un.16 is a solution to Eq.15 is known as the operational form of Eq. In the simple example given above. 6.

the solution of Eq. rst-order ordinary di erence equations. 1]un = h E ae hn (6.17 is 2 X ~ un = ck ( k )n~ xk k=1 where ck are constants determined by the initial conditions.4.21) " # " 1 # . For example. 6. 1 .18) 6. produced by applying a time-marching method to the representative equation. (1 + h)]un = h ae hn (1 .2 for defective ODE's.2 n 2 2 E . The solution of O E's with defective eigensystems follows closely the logic in Section 4. 1) . these equations can be written E .9 to 6.6. SOLUTION OF THE REPRESENTATIVE O E'S 91 Obviously the k are the eigenvalues of C and.1 The Operational Form and its Solution Examples of the nonhomogeneous.20) (6.4. h)E .4 Solution of the Representative O E's 6.2.2. 6. 2 un = u0 + u ^0n + u0 2 n (6. linear. one can show that the solution to A Defective System 2 3 2 u n+1 6 ^n+1 7 4u 5=6 41 un+1 1 32 3 un 7 7 ^n 5 56 4u un is n un = u h0 i u ^n = u ^0 + u0n . ~ if x are its eigenvectors. following the logic of Section 4.(1 + h) # " u ~ =h 1 E ae hn 1 (1 + h)E E . are given by Eqs. Using the displacement operator.1 n " # n ( n . E .19) (6.11. 1 u .

22 can be expressed as K h X un = ck ( k )n + ae hn Q(e h) (6. the ratio Q(E )=P (E ) can be found by Kramer's rule.2 Examples of Solutions to Time-Marching O E's As examples of the use of Eqs. Keep in mind that for methods such as in Eq.19. TIME-MARCHING 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 (6.45.23) P (e ) k=1 where k are the K roots of the characteristic polynomial. P ( ) = 0.24) and the solution of its representative O E follows immediately from Eq. The general solution of Eq.23. we have P (E ) = E . 1 . We can express in terms of Eq. 6. 6. Eq. as would be the case for Eq. h Q(E ) = h (6. Eq.19 to 6. h)E . Notice also. we have P (E ) = (1 . one for un and u ~n and we are usually only interested in the nal solution un.22) which is produced by applying time-marching methods to the representative ODE. the important subset of this solution which occurs when = 0. 6. When determinants are involved in the construction of P (E ) and Q(E ).25) .1. 4. For the explicit Euler method.23: un = c1(1 + h)n + ae hn For the implicit Euler method. 6. representing a time invariant particular solution. 6. respectively. we derive the solutions of Eqs. The terms P (E ) and Q(E ) are polynomials in E referred to as the characteristic polynomial and the particular polynomial. 6.21.21 there are multiple (two in this case) solutions.22 and 6. 6. 6. or a steady state. In such a case K X Q(1) un = ck ( k )n + a P (1) k=1 6.20. h h (6.22 all manner of standard time-marching methods having multiple time steps and various types of intermediate predictor-corrector families.21. Eq.92 CHAPTER 6.4. 6. 1 Q(E ) = hE e h.

Eq. Relation We have now been introduced to two basic kinds of roots. 6. First we make use of the semi-discrete approach to nd a system of ODE's and then express its solution in the form of Eq.21. h. (1 + h ) 2 h2 P (E ) = det . and there results # " E . one solves for the nal family un (one can also nd a solution for the intermediate family u ~). h.1.1 Establishing the Relation 6. one can write n u(t) = c1 e 1 h ~ x1 + + cm e mh n ~ xm + + cM e Mh n ~ xM + P:S: (6. There is a fundamental relation between the two which can be used to identify many of the essential properties of a time-march method.5 The .1.5. RELATION so 93 In the case of the coupled predictor-corrector equations. The former are the eigenvalues of the A matrix in the ODE's found by space di erencing the original PDE. the -roots and the -roots.26) 2 h2 h e . h)e h . Remembering that t = nh. THE . 2 which has only one nontrivial root ( = 0 is simply a shift in the reference index). 4.6. h (1 . 1 .27) . and the latter are the roots of the characteristic polynomial in a representative O E found by applying a time-marching method to the representative ODE. 1 (1 E 1 2 2 + h)E 2 hE The -root is found from 1 2 h2 = 0 P( ) = . The complete solution can therefore be written 1h e h + 1 + h n 1 2 2 hn 2 un = c1 1 + h + 2 h + ae (6. 1 (1 + h)E E . 1 6.27. h . 1 = E E . This relation is rst demonstrated by developing it for the explicit Euler method. 1 2 2 2 h = 1 hE (E + 1 + h) Q(E ) = det . 1 2 he h 1 n + ae hn un = c1 1 .5.

. Suppose. So if we use the Euler method for the time advance of the ODE's. TIME-MARCHING 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. instead of the Euler method. 2 hE . For one of these we nd m q 1+ h + m 1 = 1 + mh + 2 = 2 h2 . Comparing Eq. 6.28. which is de ned by un+1 = un. . 6. Since the value of e h can be expressed in terms of the series 1 2h2 + 1 3h3 + e h =1+ h+ 2 6 1 n hn + +n ! the truncated expansion = 1 + h is a reasonable3 approximation for small enough h. 3 The error is O( 2 h2 ).27 and Eq. will be discussed in Section 6. 1 4 h4 + m 2 2 mh (6. which is given by = 1 + h.3.30) Now we notice that each produces two -roots. 1 + mh .4. The other root.5. one -root.29) Applying Eq. 2 mh m .31) (6. we use the leapfrog method for the time advance. we have the characteristic polynomial P (E ) = E 2 . so that for every the must satisfy the relation m . the solution2 of the resulting O E is un = c1( 1 )n ~ x1 + + cm ( m )n ~ xm + + cM ( M )n ~ xM + P:S: (6.1 + 2hu0n (6.29.28) where the cm and the ~ xm in the two equations are identical and m = (1 + m h). 2 Based on Section 4. 6. 1 = 0 2 (6.8 to Eq. we see a correspondence between m and e m h. 1. 6.94 CHAPTER 6.32) mh 8 m h 3 3 This q is 2an2 approximation to e m with an error O( h ).

Now consider the solution obtained using the same method with a time step h=2. Therefore the error at the end of the simulation is reduced by a factor of four.5. However. Note that a second-order approximation to a derivative written in the form ( t u)n = 1 (un+1 . 6. relation for the leapfrog method produces two -roots for each . while the second-order time-marching method which results from this approximation. Consider a solution obtained at a given time T using a second-order time-marching method with a time step h. consistent with a second-order approximation. and designate it ( m )1 .6. RELATION 95 6.6. Thus the principal root is an approximation to e h up to O hk . We refer to the root that has the above property as the principal -root. 4.34. This arises simply because of our notation for the time-marching 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. 6.5. The above property can be stated regardless of the details of the timemarching method.2 The Principal -Root Based on the above we make the following observation: Application of the same time-marching method to all of the equations in a coupled system linear ODE's in the form of Eq.33) 1 1 2 2 k k k +1 = 1 + h + 2 h + + k! h + O h where k is the order of the time-marching method.1) (6. Since the error per time step is O(h3).30 that the .1 + 2hu0n (6.34) 2h has a leading truncation error which is O(h2). One of these we identi ed as the principal root which always . which is the leapfrog method: un+1 = un. 6. THE .3 Spurious -Roots We saw from Eq. un. twice as many time steps are required to reach the time T .35) has a leading truncation error O(h3). The following example makes this clear. knowing only that its leading error is O hk+1 . always produces one -root for every -root that satis es the relation (6. this is reduced by a factor of eight (considering the leading term only).5.

However. after some nite time the amplitude of the error associated with the spurious roots is even smaller then when it was initialized.36 must be initialized by some starting procedure. if there are more spurious roots (6.96 CHAPTER 6. In fact.. Following again the message of Section 4. No matter whether a -root is principal or spurious. The initial vector ~ u0 does not provide enough data to initialize all of the coe cients. if there is one spurious root to a method. the . . For example. 1 or earlier. TIME-MARCHING METHODS FOR ODE'S has the property given in 6. generation of spurious roots does not. in itself. The other is referred to as a spurious -root and designated ( m )2 .28 must be modi ed. 1 or earlier to advance the solution from time level n to n + 1. many very accurate methods in practical use for integrating some forms of ODE's have spurious roots. 6.1 and thus cannot be started using only ~ un. Presumably (i.33. the solution for the O E in the form shown in Eq.4 One-Root Time-Marching Methods There are a number of time-marching methods that produce only one -root for each -root. We refer to them as one-root methods. are spurious. If a time-marching method produces spurious -roots. To express this fact we use the notation = ( h). It should be mentioned that methods with spurious roots are not self starting. they play virtually no role in accuracy analysis. make a method inferior..e. Such roots originate entirely from the numerical approximation of the time-marching method and have nothing to do with the ODE being solved.36) Spurious roots arise if a method uses data from time level n . For example. the leapfrog method requires ~ un. They are also called one-step methods. That is. if one starts the method properly) the spurious coe cients are all initialized with very small magnitudes. all of the coe cients (cm )2 in Eq. Thus while spurious roots must be considered in stability analysis.5. and presumably the magnitudes of the spurious roots themselves are all less than one (see Chapter 7).4. All spurious roots are designated ( m )k where k = 2 3 . except for the principal one. 6. Then the presence of spurious roots does not contaminate the answer. This results because methods which produce spurious roots require data from time level n . 6. we have un = c11 ( 1 )n x1 + + cm1( m )n xm + + cM 1 ( M )n xM + P:S: 1~ 1~ 1~ n n +c12 ( 1)2 ~ x1 + + cm2 ( m )2 ~ xm + + cM 2 ( M )n xM 2~ n n n +c13 ( 1)3 ~ x1 + + cm3 ( m )3 ~ xm + + cM 3 ( M )3 ~ xM +etc. relation produced by a time-marching scheme can result in multiple -roots all of which. In general. it is always some algebraic function of the product h.

A popular method having this property. so that all the constants.6. They have the signi cant advantage of being self-starting which carries with it the very useful property that the time-step interval can be changed at will throughout the marching process. the so-called -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. The only error made by introducing the time marching is the error that makes in approximating e h. and matrices are identical except the ~ u and the terms inside the parentheses on the right hand sides. as we shall see in Chapter 8. vectors.2. relation is . the trapezoidal ( = 2 implicit Euler methods ( = 1). these equations are identical. Three one-root methods were analyzed in Section 6.4. Notice that when t and n = 0. h It is instructive to compare the exact solution to a set of ODE's (with a complete eigensystem) having time-invariant forcing terms with the exact solution to the O E's for one-root methods. )u0n + u0n+1 6.4. )h = 1 +1(1 . Its .6. respectively. It is usually used as the basis for establishing the order of a method. For example. is given by the formula h i un+1 = un + h (1 .6.6 Accuracy Measures of Time-Marching Methods 6. ACCURACY MEASURES OF TIME-MARCHING METHODS 97 1 ). These are ~ u(t) = c1 e 1 h n ~ x1 + + cm e m h n ~ xm + + cM e M h n ~ xM + A. It is quite common to judge a time-marching method on the basis of results found from a Taylor table.1 Local and Global Error Measures There are two broad categories of errors that can be used to derive and evaluate timemarching methods. This is a global error.1~ f (6. and the The -method represents the explicit Euler ( = 0).1~ f ~ un = c1( 1 )n ~ x1 + + cm( m )n ~ xm + + cM ( M )n ~ xM + A. However. This is a local error such as that found from a Taylor table analysis. One is the error made in each time step.37) respectively. see Section 3. It is useful for comparing methods. it is of no use in: . a Taylor series analysis is a very limited tool for nding the more subtle properties of a numerical time-marching method.

Our error measures are based on the di erence between the exact solution to the representative ODE. TIME-MARCHING METHODS FOR ODE'S nding spurious roots. is to some extent arbitrary. However. and to study them we make use of the material developed in the previous sections of this chapter. including only the contribution from the principal root. a necessary condition for the choice should be that the measure can be used consistently for all methods. The order of the method is the last power of h matched exactly. given by t u(t) = ce t + ae . Therefore. The latter three of these are of concern to us here. We designate this by er and make the following de nition er e h. (6. . either local or global.6.2 Local Accuracy of the Transient Solution (er Transient error er! ) The particular choice of an error measure.98 CHAPTER 6. a very natural local error measure for the transient solution is the value of the di erence between solutions based on these two roots. analyzing the particular solution of predictor-corrector combinations.relation we saw that all time-marching methods produce a principal -root for every -root that exists in a set of linear ODE's. which can be written as h un = c1 ( 1)n + ae hn Q(e h ) P (e ) (6. This is similar to the error found from a Taylor table. nding the global error. 1 The leading error term can be found by expanding in a Taylor series and choosing the rst nonvanishing term.39) j j 6. evaluating numerical stability and separating the errors in phase and amplitude. In the discussion of the .38) and the solution to the representative O E's.

The above expression for er! can be normalized to give the error in the phase speed. Then the numerical method must produce a principal -root that is complex and expressible in the form (6. ( 1)2 r + ( 1 )i er! !h .ia .41) Amplitude and phase errors are important measures of the suitability of time-marching methods for convection and wave propagation phenomena.1 (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. tan.5 can be extended to the combination of a spatial discretization and a time-marching method applied to the linear convection equation. We have found these to be given by 1 P:S:(ODE) = ae t ( . The approach to error analysis described in Section 3.6. ) . while a negative value corresponds to phase lead (the numerical phase speed is too large). we have h = . is found using = .6.1 ( 1 )i=( 1 )r )] and the local error in phase can be de ned as (6. Such can indeed be the case when we study the equations governing periodic convection which produces harmonic motion. Introducing the Courant number. For such cases it is more meaningful to express the error in terms of amplitude and phase.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. A positive value of erp corresponds to phase lag (the numerical phase speed is too small). !h Cn x 6.iCn x. The principal root.a . ACCURACY MEASURES OF TIME-MARCHING METHODS 99 Amplitude and Phase Error Suppose a eigenvalue is imaginary. j 1 j = 1 . Thus one can obtain values of the principal root over the range 0 x for a given value of the Courant number. Let = i! where ! is a real number representing a frequency. that is q 2 era = 1 . as follows . Cn = ah= x.6. where is the modi ed wavenumber of the spatial discretization.42) er = er! = 1 + tan ( 1)i =( 1)r )] p where ! = . 1( h).

and it is necessary that the advertised order of accuracy be valid for the nonlinear cases as well as for the linear ones. is equal to zero. For a measure of the local error in the particular solution we introduce the de nition er P:S:(O E) .44) where h( . and for this case. 6. many numerical methods generate an er that is also zero. The algebra involved in nding the order of er can be quite tedious.45) (6.43) The multiplication by h converts the error from a global measure to a local one.4 Time Accuracy For Nonlinear Applications In practice.43 can be written in terms of the characteristic and particular polynomials as o er = c .46) (6. TIME-MARCHING METHODS FOR ODE'S h P:S:(O E) = ae t Q(e h) P (e ) respectively. 1 h P:S: (ODE ) ( ) (6. If the forcing function is independent of time. However. time-marching methods are usually applied to nonlinear ODE's.P e h o (6. this order is quite important in determining the true order of a time-marching method by the process that has been outlined. 6. ) co = h lim !0 P e h The value of co is a method-dependent constant that is often equal to one. n ( . Eq.47) . More precisely.6. a time-marching 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. In order to determine the leading error term. )Q e h .100 and CHAPTER 6. An illustration of this is given in the section on Runge-Kutta methods. so that the order of er and er are consistent. 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.

the analysis based on a linear nonhomogeneous ODE produces the appropriate conditions for the majority of time-marching methods used in CFD. Let T be the xed time of the event and h be the chosen step size. ( 1 ( h))N (6. These are given by Era = 1 . 1 = !T .5 Global Accuracy In contrast to the local error measures which have just been discussed. to derive all of the necessary conditions for the fourth-order Runge-Kutta method presented later in this chapter the derivation must be performed for a nonlinear ODE.6. N tan ( 1)i =( 1)r ] !# (6.6.1 ( 1 )r . These are useful when we come to the evaluation of time-marching methods for speci c purposes. we are more concerned with the global error in amplitude and phase. We refer to such a computation as an \event". 6. is N. This subject is covered in Chapter 8 after our introduction to stability in Chapter 7.48) Global error in amplitude and phase If the event is periodic. For example. Suppose we wish to compute some time-accurate phenomenon over a xed interval of time using a constant time step. tan.49) ( 1 )i Er! N !h . we can also de ne global error measures. and q " 2 ( 1 )2 r + ( 1 )i N (6.6. ACCURACY MEASURES OF TIME-MARCHING METHODS 101 The reader should be aware that this is not su cient.50) . 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 . However. Then the required number of time steps.

K k Fn+k (6. The methods are said to be linear because the 's and 's are independent of u and n.4.51 is applied to the representative equation. We shall not spend much time on development or design of these methods.K 1 0 1 X k k E Aun = h@ k=1. They are explicit if 1 = 0 and implicit otherwise. TIME-MARCHING 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. and they are said to be K -step because K time-levels of data are required to marching the solution one time-step.51) where the notation for F is de ned in Section 6. since most of them have historic origins from a wide variety of disciplines and applications.102 CHAPTER 6.52) . we introduce classes of methods along with their associated error analysis. 6. In the subsequent sections.K 1 k hn k E A( un + ae ) (6. 6.7. It can be measured by Er ( . and the result is expressed in operational form. Eq. ) Qeh .1. one nds 0 1 @ X k=1. 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.1 P eh 6.7 Linear Multistep Methods In the previous sections. 6. When Eq.8.1 The General Formulation When applied to the nonlinear ODE all linear multistep methods can be expressed in the general form 1 X k=1. h.K k un+k = h du = u0 = F(u t) dt 1 X k=1. 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.

7.2)1 .1.1 .(.1 1 . Taylor table for the Adams-Moulton three-step linear multistep method.4.1 .2 1 .51 if X X X k = 0 and k = (K + k .(. to be of any value in time accuracy. can be designed using the Taylor table approach of Section 3.2 .5.2)3 . .1 1 . We can also agree that. that is ! (1 + h) as h ! 0.2u0n. 6.2 u0n.1 2 .54 can be generated as 1 1 1 un+1 1 1 2 6 24 . a method should at least be rst-order accurate. LINEAR MULTISTEP METHODS 103 We recall from Section 6.51 can be multiplied by an arbitrary constant. The Adams-Moulton family is obtained from Eq. 1) k Since both sides of Eq.2 (6.(. The condition referred to as consistency simply means that ! 1 as h ! 0.un .2 .1 u0n.53) 1=1 0 = . these methods are often \normalized" by requiring X k=1 Under this condition co = 1 in Eq.2 .44.1 .h 0 u0n .h .54) A Taylor table for Eq. un h u0n h2 u00 n h3 u000 n h4 u0000 n . 11 6 .1 . and it is certainly a necessary condition for the accuracy of any time marching method.2 2 6 Table 6. 6. 6.1 . 6.1 u0n.1 6 . that approximates e h through the order of the method.2 that a time-marching method when applied to the representative equation must provide a -root.2)0 . referred to as Adams-Bashforth (explicit) and AdamsMoulton (implicit).(.0 1 1 .2) (6.6. labeled 1.1 + . Two well-known families of them.1 k=0 The Adams-Bashforth family has the same 's with the additional constraint that 1 = 0.h .2)2 . k k k k 6.2 Examples There are many special explicit and implicit forms of linear multistep methods.h 1 u0n+1 .51 with k = . The three-step Adams-Moulton method can be written in the following form un+1 = un + h( 1u0n+1 + 0 u0n + . .7. 6. 12 . One can show that these conditions are met by any method represented by Eq.

1 + 2hun+1 h 0 i 2nd-order Backward h 0 0 = un + 12 5un+1 + 8un .5=24 .58) 0 = 23=12 .4 5 4 . un. some of which are very common in CFD applications.1 = . One can verify that the Adams type schemes given below satisfy Eqs.1 + 5u0n. 6.1 AB2 h i h 23u0n .56) 1 = 9=24 0 = 19=24 . un.104 CHAPTER 6. 16u0n. resulting in (6.16=12 . is given below together with identifying names that are sometimes associated with them.2 = un + 12 AB3 Implicit Methods un+1 un+1 un+1 un+1 4 Recall from Section 6.55 and 6.2 .2 = 1=24 which produces a method which is fourth-order accurate.4 7 17 6 7 6 7 6 7 0 7 6 7 6 7 = 6 (6.2 that a kth-order time-marching method has a leading truncation error term which is O(hk+1 ).4 . TIME-MARCHING METHODS FOR ODE'S This leads to the linear system 2 1 1 1 1 32 3 213 1 6 6 6 2 0 .5. A list of simple methods. Explicit Methods un+1 un+1 un+1 un+1 = un + hu0n Euler 0 = un.1 5 = 4 1 7 5 0 3 12 1 .57 up to the order of the method.57) 4 0 .1 5 4 1 7 5 4 0 .55) 4 3 0 3 12 5 4 .2 = 5=12 This is the third-order Adams-Bashforth method.1 = . In the following material AB(n) and AM(n) are used as abbreviations for the (n)th order Adams-Bashforth and (n)th order Adams-Moulton methods.4 With 1 = 0 one obtains 2 32 3 2 3 1 1 1 1 0 6 7 6 7 6 (6.1 AM3 .1 + 2hhun Leapfrog i 1 0 0 = un + 2 h 3un . = un + hu0n Implicit Euler h+10 0 i 1 = un Trapezoidal (AM2) h + 2 h un + un+1 0 i 1 = 3 4un .2 .32 1 .2 giving (6.2 to solve for the 's. un.

This limits the interest in multistep methods to about two time levels.e. Notice that the Adams methods have = 0.6.2..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 6.7. un. 6. which corresponds to . can be written as h i (1 + )un+1 = (1 + 2 )un . 6. that is at least rst-order accurate.2=9 0 1=2 . The most general two-step linear multistep method (i.59 is given in Table 6.1=2.51. Methods with = . =3 = 6 . the methods are 2nd-order accurate) if 1 '= .' = .5=6 .1=2 0 0 0 0 . 6.1=2 .1=6 .1 (6.1=2 .3 Two-Step 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.and two-step methods.1=4 .51. ' 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 . + ')u0n . One can show after a little algebra that both er and er are reduced to 0(h3) (i. see Eq. Some linear one. LINEAR MULTISTEP METHODS 105 6.1=6 0 .7. A list of methods contained in Eq. 6..e. 6.2. which corresponds to 0 = 0 in Eq. K=2 in Eq.1 = 0 in Eq.59) Clearly the methods are explicit if = 0 and implicit otherwise.59. are known as Milne methods.1=3 .51). Finally a unique fourth-order method is found by setting = .1=3 0 1=12 .1] + h u0n+1 + (1 . +2 The class of all 3rd-order methods is determined by imposing the additional constraint =2 .1=2 0 .1=2 .5 6 1. 'u0n.

it is obvious from Eq.61 that + =1 =1 2 which provides two equations for three unknowns.60) where the parameters and are arbitrary parameters to be determined.61) (6.62) Considering only local accuracy. where measures of e ciency are discussed in the next two chapters. 1 . Their use in solving ODE's is relatively easy to illustrate and understand. . therefore. 6. For the method to be second-order accurate both er and er must be O(h3). following the example leading to Eq.106 6. A simple one-predictor.44 that the same conditions also make it O(h3). 6.4. to the following observations.8 Predictor-Corrector Methods CHAPTER 6. Their use in solving PDE's can be much more subtle and demands concepts5 which have no counterpart in the analysis of ODE's. Q(E ) = E h E + + h] 2 h2 i (6. fractional-step. and usually the nal family has a higher Taylor-series order of accuracy than the intermediate ones. that the predictor-corrector sequence is a second-order accurate method for any . each of which is referred to as a family in the solution process. one is led. 1 u0n (6. ( + ) h .26. u ~n+ = un + hu0n 1h 1 u un+1 = un + 2 ~0n+ + 2 . TIME-MARCHING METHODS FOR ODE'S There are a wide variety of predictor-corrector schemes created and used for a variety of purposes.6. and hybrid methods. Predictor-corrector methods constructed to time-march linear or nonlinear ODE's are composed of sequences of linear multistep methods. There may be many families in the sequence. one-corrector example is given by u ~n+ = un + h hu0n i un+1 = un + h u ~0n+ + u0n (6. but it is not di cult to show using Eq. by following the discussion in Section 6.63) 5 Such as alternating direction. 6. It is easy to show. The situation for er requires some algebra. Their use is motivated by ease of application and increased e ciency. For this to hold for er . that h P (E ) = E E . One concludes. One can analyze this sequence by applying it to the representative equation and using the operational techniques outlined in Section 6.

64) Some simple.67) (6.65) 2h u The Burstein method. 6. is u ~n+1 = un + hu0n 1 u +u un+1 = 2 ~0n+1] n ~n+1 + hu Note that MacCormack's method can also be written as u ~n+1 = un + hu0n 0 ~0 ] un+1 = un + 1 n+1 2 h un + u from which it is clear that it is obtained from Eq. presented earlier in this chapter. referred to as Runge-Kutta methods.63 with = 1.66) (6. nally. speci c. RUNGE-KUTTA METHODS 107 A classical predictor-corrector sequence is formed by following an Adams-Bashforth predictor of any order with an Adams-Moulton corrector having an order one higher. .1 (6.9. which we discuss in Chapter 8. MacCormack's method. 6 Although implicit and multi-step Runge-Kutta methods exist. second-order accurate methods are given below. The Adams-BashforthMoulton sequence for k = 3 is h i u ~n+1 = un + 1 h 3u0n . If the order of the corrector is (k).6 that produce just one -root for each -root such that ( h) corresponds explicit Runge-Kutta methods here. u0n. The Gazdag method. is h 0 0 i u ~n+1 = un + 1 h 3~ u . u0n.1 2 i h h5~ un+1 = un + 12 u0n+1 + 8u0n .i1 un+1 = un + 1 ~0n + u ~0n+1 (6. we refer to these as ABM(k) methods.9 Runge-Kutta Methods There is a special subset of predictor-corrector methods.u ~ 2 h n n. 6. The order of the combination is then equal to the order of the corrector.63 with = 1=2 is u ~n+1=2 = un + 1 hu0n 2 un+1 = un + hu ~0n+1=2 and. (6. obtained from Eq.68) 6.6. we will consider only single-step.

. We present it below in some detail. but. and 2 . 6. First of all.69) = 1 + h + 1 2h2 + + 1 k hk 2 k! It is not particularly di cult to build this property into a method. 1. the method must further satisfy the constraint that er = O(hk+1) (6. the principal (and only) -root is given by (6. 6. Thus. as we pointed out in Section 6.6.108 CHAPTER 6. Thus for a Runge-Kutta method of order k (up to 4th order).69 and 6.72) un+1 = un + 1hu0n + 2hu ~0n+ 1 + 4hu0n+ 2 Appearing in Eqs. TIME-MARCHING METHODS FOR ODE'S to the Taylor series expansion of e h out through the order of the method and then truncates. u(tn) = 1 k1 + 2k2 + 3k3 + 4k4 (6.71 is bn+ = un + hu0n u b0n+ u ~n+ 1 = un + 1 hu0n + 1hu b0n+ + 2hu un+ 2 = un + 2 hu0n + 2hu ~0n+ 1 b0n+ + 3hu (6. The most widely publicized Runge-Kutta process is the one that leads to the fourth-order method. it is not su cient to guarantee k'th order accuracy for the solution of u0 = F (u t) or for the representative equation. 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) .70) and this is much more di cult. They must satisfy the relations = 1 = 1+ 1 (6.70.72 are a total of 13 parameters which are to be determined such that the method is fourth-order according to the requirements in Eqs.4.71 and 6.73) 2 = 2+ 2+ 2 . a scheme entirely equivalent to 6.71) However. we prefer to present it using predictor-corrector notation. are not arbitrary. the choices for the time samplings. To ensure k'th order accuracy.

but the most popular one is due to Runge.74 and 6.75 are all satis ed.76) n+1=2 n+1=2 n+1 6 n conditions for Runge-Kutta methods of up to third order. To satisfy the condition that er = O(k5). the resulting method would be third-order accurate.6.74 and the rst equation in 6. 7 The present approach based on a linear inhomogeneous equation provides all of the necessary . we have to ful ll four more conditions 2+ 4 2 = 1=3 (3) 2 2+ 3 1 2 3 3 3 + + = 1 = 4 (4) 2 3 1 4 2 (6. As discussed in Section 6. It results in the \standard" fourth-order Runge-Kutta method expressed in predictor-corrector form as bn+1=2 = un + 1 hu0n u 2 b0n+1=2 u ~n+1=2 = un + 1 hu 2 un+1 = un + hu ~0n+1=2 h0 0 0 0 i b un+1 = un + 1 h u + 2 u + u ~ + u (6. 6.75 which must be satis ed by the 10 unknowns.73 to eliminate the 's we nd from Eq. Thus if the rst 3 equations in 6.75 cannot be derived using the methodology presented here. 6. Using Eq.6.69 and 6.70. 6.74) These four relations guarantee that the ve terms in exactly match the rst 5 terms in the expansion of e h . two parameters can be set arbitrarily. which is based on a linear nonhomogenous representative ODE.4. Since the equations are overdetermined.75) 2 2 2 (4) 3 1 + 4 ( 2 + 1 2 ) = 1=12 (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.69 the four conditions 1+ 2 + 3 1+ = 1 2+ 3+ 4 = 1=2 3 1+ 4 2 ( + ) 4 2 1 2 = 1=6 = 1=24 4 12 (1) (2) (3) (4) (6. the fourth condition in Eq.9. but the equations follow directly from nding P (E ) and Q(E ) and then satisfying the conditions in Eqs.7 There are eight equations in 6. RUNGE-KUTTA METHODS 109 The algebra involved in nding algebraic equations for the remaining 10 parameters is not trivial. A more general derivation based on a nonlinear ODE can be found in several books. Several choices for the parameters have been proposed. 6.

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. Our presentation of the time-marching 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.110 CHAPTER 6. respectively.74 and the rst equation in 6.67 are second-order Runge-Kutta methods. 6. relations.10. 6. we will see that these methods can have widely di erent properties with respect to stability.75. RK methods of order k require k evaluations of the derivative function to advance the solution one time step. In any event.1 Application to Systems of Equations Consider rst the numerical solution of our representative ODE u0 = u + ae t (6. Following the steps outlined in Section 6. 6. we obtained (1 . TIME-MARCHING METHODS FOR ODE'S 9 > > = > > Notice that this represents the simple sequence of conventional linear multistep methods referred to. 6. We shall discuss the consequences of this in Chapter 8.66 and 6. storage requirements reduce the usefulness of Runge-Kutta methods of order higher than four for CFD applications.72 by setting 4 = 0 and satisfying only Eqs. Higher-order RungeKutta methods can be developed. It is clear that for orders one through four. In the next chapter. These are covered in this Section. un = he h ae hn (6. and third-order methods can be derived from Eqs.10 Implementation of Implicit Methods We have presented a wide variety of time-marching methods and shown how to derive their . a fth-order method requires six evaluations to advance the solution one step. h)un+1 .77) using the implicit Euler method.2.78) . This leads to various tradeo s which must be considered in selecting a method for a speci c application. but they require more derivative evaluations than their order. For example. 6.

aBp (.81) (6. the system of equations which must be solved is tridiagonal (e. the cost per time step of an implicit method is larger than that of an explicit method. but rather we solve Eq.1 0 1)=2 x). hA)~ un+1 .10. consider the nonlinear ODE du + 1 u2 = 0 (6. For our one-dimensional examples. h n 111 (6.10.84) un+1 = un + hF (un+1 tn+1) This is a nonlinear di erence equation. The primary area of application of implicit methods is in the solution of sti ODE's.83) (6. hf The inverse is not actually performed.85) dt 2 . as we shall see in Chapter 8. but in multidimensions the bandwidth can be very large. A = . Now let us apply the implicit Euler method to our generic system of equations given by ~(t) ~ u0 = A~ u.f ~(t + h) (I . In general.78 is (6.2 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. for biconvection. IMPLEMENTATION OF IMPLICIT METHODS Solving for un+1 gives 1 (u + he h ae hn ) un+1 = 1 .1 ~ un .79) This calculation does not seem particularly onerous in comparison with the application of an explicit method to this ODE. 6. hA).g. ~ un = .81 as a linear system of equations. 6. Now the equivalent to Eq. As an example. 6. and hence its solution is inexpensive. requiring only an additional division..80) ~ are vectors and we still assume that A is not a function of ~ where ~ u and f u or t.hf or (6.6.82) ~(t + h)] ~ un+1 = (I .

tn) n n 2F ! 2F ! 1 @ @ + 2 @u2 (u . un)(t . t )2 + +2 n @t2 n (6. Designate the reference value by tn and the corresponding value of the dependent variable by un.83. TIME-MARCHING METHODS FOR ODE'S solved using implicit Euler time marching. we expand F (u t) about some reference point in time. A Taylor series expansion about these reference quantities gives ! ! @F @F F (u t) = F (un tn) + @u (u . 6. tn) @t2 + n n (6. can be used. un)2 + @u@t (u . which gives 1 u2 = u un+1 + h 2 (6. tn) n !n 2 1 @ F (t . tn)k and (u . such as Newton's method (see below). tn) @t + 2 (t . tn) + O(h2) n n (6. There are several di erent approaches one can take to solving this nonlinear di erence equation. the expansion of u(t) in terms of the independent variable t is ! 2u ! @u 1 @ 2 u(t) = un + (t . 6. In practice. An iterative method. 6.87) On the other hand. which implies that a linearization approach may be quite successful. un) + @t (t . In order to implement the linearization.87 can be written ! ! @F @F F (u t) = Fn + @u (u . Such an approach is described in the next Section. un)k are O(hk ).112 CHAPTER 6.86) n n+1 which requires a nontrivial method to solve for un+1. the \initial guess" for this nonlinear problem can be quite close to the solution.88) If t is within h of tn . since the \initial guess" is simply the solution at the previous time step.89) .10. and Eq.3 Local Linearization for Scalar Equations General Development Let us start the process of local linearization by considering Eq. un) + @t (t . both (t .

6. The use of local time linearization updated at the end of each time step. namely ! ! ! ! du = @F u + F .83.90) Implementation of the Trapezoidal Method As an example of how such an expansion can be used.10.6. In many uid mechanic applications the nonlinear function F is not an explicit function " ! ! # " !# ! .83. other second-order implicit timemarching methods that can be used. it represents a second-order-accurate. Thus. @F u + @F (t . locally-linear approximation to F (u t) that is valid in the vicinity of the reference station tn and the corresponding un = u(tn). 6.6. There are. 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 second-order time-marching process. t ) + O(h2) n n dt @u n @u n n @t n (6. IMPLEMENTATION OF IMPLICIT METHODS 113 Notice that this is an expansion of the derivative of the function. 6.89 to evaluate Fn+1 = F (un+1 tn+1). consider the mechanics of applying the trapezoidal method for the time integration of Eq. and the trapezoidal time march. With this we obtain the locally (in the neighborhood of tn) time-linear representation of Eq. of course.91) where we write hO(h2) to emphasize that the method is second order accurate. u ) + h @F + O(h2) + F un+1 = un + 1 h F n+ n 2 @u n n+1 n @t n +hO(h2) (6. The trapezoidal method is given by 1 h F + F ] + hO(h2) un+1 = un + 2 n+1 n (6. one nds @F (u . A very useful reordering of the terms in Eq. combine to make a second-order-accurate numerical integration process. relative to the order of expansion of the function.93) 1.92) 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. 6. Using Eq. 2 n = hFn + h @u n 2 @t n which is now in the delta form which will be formally introduced in Section 12.92 results in the expression 1 2 @F 1 h @F u (6.

Implementation of the Implicit Euler Method We have seen that the rst-order implicit Euler method can be written un+1 = un + hFn+1 " 1 .1 in carrying out this step). Let this storage area be referred to as B . and save ~ 1. 4. TIME-MARCHING METHODS FOR ODE'S of t.93 simpli es to the second-order accurate expression " 1 h @F 1. 6. store them in an array say R ~ . 2 @u !# n un = hFn (6.94) Notice that the RHS is extremely simple. Solve for the elements of hF un.96) . (Very seldom does one nd B . Find ~ un+1 by adding ~ un to ~ un. it is generally the space-di erenced form of the steady-state equation of some uid ow problem. we arrive at the form !# n un = hFn (6. but for our applications. 3. the basic equation was the simple scalar equation 6. A numerical time-marching procedure using Eq.95) if we introduce Eq. In such cases the partial derivative of F (u) with respect to t is zero and Eq. It is the product of h and the RHS of the basic equation evaluated at the previous time step.83. In this example. Solve for the elements of the matrix multiplying ~ un and store in some appropriate manner making use of sparseness or bandedness of the matrix if possible. Solve the coupled set of linear equations ~ B ~ un = R for ~ un. rearrange terms.90 into this method. thus ~ un+1 = ~ un + ~ un The solution for ~ un+1 is generally stored such that it overwrites the value of ~ un and the process is repeated.114 CHAPTER 6. 6. and remove the explicit dependence on time.94 is usually implemented as follows: ~n. 2. 6. h @F @u (6.

88 (remember the dependence on t has been eliminated for this case).97) un+1 = un . and is a scaling factor. Newton's Method Consider the limit h ! 1 of Eq.10.96 leads to linear convergence. the error at a given iteration is some multiple of the error at the previous iteration.10. We shall see later that this method is an excellent choice for steady problems. 6. There results or .96 obtained by dividing both sides by h and setting 1=h = 0. 6. IMPLEMENTATION OF IMPLICIT METHODS 115 We see that the only di erence between the implementation of the trapezoidal method 1 in the brackets of the left side of and the implicit Euler method is the factor of 2 Eqs.99) Here f represents a dimensionless stream function. The fact that it has quadratic convergence is veri ed by a glance at Eqs. 6. 6. let us consider an example involving some simple boundary-layer equations.96. We choose the Falkner-Skan equations from classical boundary-layer theory.94.98) This is the well-known Newton method for nding the roots of a nonlinear equation F (u) = 0.87 and 6.. By quadratic convergence.94 and 6. Quadratic convergence is thus a very powerful property. where the error is the di erence between the current solution and the converged solution. .6. @F @u ! n un = Fn @F @u ! #. we mean that the error after a given iteration is proportional to the square of the error at the previous iteration.e.4 Local Linearization for Coupled Sets of Nonlinear Equations In order to present this concept. 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. 6.1 n (6. i. 6. given by Eq. The reader should ponder the meaning of letting h ! 1 for the trapezoidal method. Omission of this factor degrades the method in time accuracy by one order of h. Use of a nite value of h in Eq. " Fn (6.

87.101) Now we seek to make the same local expansion that derived Eq. @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. 1 . TIME-MARCHING METHODS FOR ODE'S f u1 = d dt2 2 First of all we reduce Eq.100) (6.102 by the following process d~ u =F ~ (~ u) dt (6. It is derived from Eq. 6. and de ning F un).u1 u3 . u2 2 0 u2 = F2 = u1 u03 = F3 = u2 and these can be represented in vector notation as u2 = df dt u3 = f (6.116 CHAPTER 6.103) 2 6 6 6 6 A=6 6 6 6 6 4 ~ (~ The expansion of F u) about some reference state ~ un can be expressed in a way similar to the scalar expansion given by eq 6. 6. one has 9 2. 6. called the Jacobian matrix.90.102) A = (aij ) = @Fi =@uj For the general case involving a third order matrix this is (6.2 vector for the second term. except that this time we are faced with a nonlinear vector function. Omitting the explicit dependency ~ n as F ~ (~ on the independent variable t.104) 8 Recall that we derived the Jacobian matrices for the two-dimensional Euler equations in Section 9 The Taylor series expansion of a vector contains a vector for the rst term. rather than a simple nonlinear scalar function.99 to a set of rst-order nonlinear equations by the transformations This gives the coupled set of three nonlinear equations u01 = F1 = . The required extension requires the evaluation of a matrix. and tensor products for the terms of higher order.8 Let us refer to this matrix as A. a matrix times a .

Re-evaluate ~ the elements using un+1 and continue. the solution will be second-order-accurate in time. 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. and the manner in which. 6. Thus for the Falkner-Skan equations the trapezoidal method results in 2 h (u3)n .2 1 We nd ~ un+1 from ~ un + ~ un. Using this we can write the local linearization of Eq. could be used to integrate the equations without loss in accuracy with respect to order. Find an expression for the nth term in the Fibonacci series.1. tn and the argument for O(h2) is the same as in the derivation of Eq.~ un + O(h2) (6. explicit or implicit. the terms in the Jacobian matrix are updated as the solution proceeds depends. u2) 3 1 + 1n7 6 1 3n 2 2 6 7 2n7 7 6 6 h ( u ) ( u ) 7 = h 6 4 5 4 5 .11 Problems 1. Returning to our simple boundary-layer example.106) 6. Without any iterating within a step advance. 1 0 2 n 1 n 2 4 5 ( u) ( u ) h 3n 2n 0 .107) 4 1 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. What is u25 ? (Let the rst term given above be u0.93. An~ dt | {z } (6. second-order-accurate approximation to a set of coupled nonlinear ordinary di erential equations that is valid for t tn + h.101. we nd the Jacobian matrix to be 2 3 . d~ u =A ~ ~ un +O(h2) n u + F n . 6.6. h(u2)n h (u1)n 32 ( u ) 3 2 .) . u 2 u2 .u1 3 0 0 7 A=6 (6. PROBLEMS 117 ~ (~ ~ n + An ~ F u) = F u. (u u ) . of course.102 as \constant" which is a locally-linear.or second-order time-marching method. and the solution is now advanced one step.105) where t . 6. The number of times. on the nature of the problem. 6.11. which is given by Eq. (1 . Any rst.88.

(b) Derive the .118 CHAPTER 6. (d) Perform a -root trace relative to the unit circle for both di usion and convection. (c) Find er and the rst two nonvanishing terms in a Taylor series expansion of the spurious root. The trapezoidal method un+1 = un + 2 n+1 n sentative ODE. relation.1 + 23h (u0n+1 + u0n + u0n. 5. (b) Derive the . Identify the polynomials P (E ) and Q(E ). Consider the time-marching scheme given by un+1 = un. 6.1 n+1 (a) Write the O E for the representative equation. Consider the following time-marching method: u ~n+1=3 = un + hu0n=3 un+1=2 = un + hu ~0n+1=3 =2 un+1 = un + hu0n+1=2 .65) to the representative equation. Find the . Find the di erence equation which results from applying the Gazdag predictorcorrector method (Eq. Solve for the -roots and identify them as principal or spurious.1) (a) Write the O E for the representative equation. 6. 4. TIME-MARCHING METHODS FOR ODE'S 1 h(u0 + u0 ) is used to solve the repre2. (c) Find er .relation. Identify the polynomials P (E ) and Q(E ). (a) What is the resulting O E? (b) What is its exact solution? (c) How does the exact steady-state solution of the O E compare with the exact steady-state solution of the ODE if = 0? 3.relation. The 2nd-order backward method is given by 1 h4u . u + 2hu0 i un+1 = 3 n n.

ah= x.6. Use only 4th-order Runge-Kutta time marching at a Courant number of unity. implicit Euler. 7. compute the solution at t = 1 using 2nd-order centered di erences in space coupled with the 4th-order Runge-Kutta method for grids of 100. What order is the homogeneous solution? What order is the particular solution? Find the particular solution if the forcing term is xed. of 0. On a log-log scale. Find the global order of accuracy from the plot. Find er and er . Write a computer program to solve the one-dimensional linear convection equation with periodic boundary conditions and a = 1 on the domain 0 x 1.0:5)= ]2 with = 0:08. and 400 nodes. 2nd-order Adams-Bashforth (AB2). 8. PROBLEMS 119 Find the di erence equation which results from applying this method to the representative equation. Using the computer program written for problem 7. 11. Run until a periodic steady state is reached which is independent of the initial condition and plot your solution . Use 2nd-order centered di erences in space and a grid of 50 points. For the initial condition. Find the solution to the di erence equation. Use the explicit Euler. and 4th-order Runge-Kutta methods. plot the error given by v u M u X uexact u j )2 t (uj . Using the computer program written for problem 8. use u(x 0) = e. including the homogeneous and particular solutions. 9. Repeat problem 7 using 4th-order (noncompact) di erences in space. 10. repeat problem 9 using 4th-order (noncompact) di erences in space.relation. use a Courant number. trapezoidal. Show solutions at t = 1 and t = 10 compared to the exact solution.0:5 (x.M j =1 where M is the number of grid nodes and uexact is the exact solution. 200.1 for the other methods. Write a computer program to solve the one-dimensional linear convection equation with in ow-out ow boundary conditions and a = 1 on the domain 0 x 1. use a Courant number of unity. Let u(0 t) = sin !t with ! = 10 . Plot the solutions obtained at t = 1 compared to the exact solution (which is identical to the initial condition).11. Find the . For the explicit Euler and AB2 methods.

derive and use a 3rd-order backward operator (using nodes j . and the amplitude error. use a 3rd-order backwardbiased operator (using nodes j . 3rd. j . 2. 12. erp. TIME-MARCHING METHODS FOR ODE'S compared with the exact solution. and j +1 see problem 1 in Chapter 3). nd the phase speed error. j . 2. 2nd. Use grids with 100.69) together with 4th-order Runge-Kutta time marching. for the combination of second-order centered differences and 1st. Use a third-order forward-biased operator at the in ow boundary (as in Eq. and 4th-order Runge-Kutta time-marching at a Courant number of unity. Repeat problem 11 using 4th-order (noncompact) centered di erences. without actually deriving the methods. j . and 400 nodes and plot the error vs. Also plot the phase speed error obtained using exact integration in time. 6. 3. era .e..67). Explain your results.69. 3.2. 200.120 CHAPTER 6. 13. j . Use 2nd-order centered di erences in space with a 1st-order backward di erence at the out ow boundary (as in Eq. At the last grid node. i. 1. Find the global order of accuracy. and j ) and at the second last node. the number of grid nodes. 1. as described in problem 9. . Using the approach described in Section 6. 3.6. Note that the required -roots for the various Runge-Kutta methods can be deduced from Eq. that obtained using the spatial discretization alone.

  ¡ ¢¤£ ¥§¦©¨         &  $  !  #" $  % ')(1032406587@9BA40DCE24FHG8FPI@2QISRUTVG87@W4FX9PFXG`YFPTa2406FXG8bc035dTeFPfhg49X0i24I15ap424FHq1035VT87@9r7@2cAsA406gt0624A4TuI@2§G8bc0 gE7@5vG8FXw3p49P7@5xg4b40624I1fh0324I12yWt06FP24(€w6I124TeFXA4035V03A‚¤ƒ`2yG8bc0§w37@Te0QI@RuGeb40Q24I1249XFP24037@5…„‡†‡ˆ‰‘T’I@R FP2“Ge035e06TVG”FX2€•4p4FPA–AcY—2E7Sf˜FXw3T3™BTvGd7@W4FX9PFHG`YsFXTeI@RfG8062€A4FPTVw3p4TVTe03AgFP2hGe035efhTiI@RCcij03AkglI1FP2“G8T”7S24A 7SGVG8587Sw6G8I@5eT3mƒ`2#G8b406Te0uGe035Vf˜T‡7˜TVY—TVGe03fnFXTmTe7@FPAoG8IpWl0eTVG87@W49X0uFP27qw3065VG87@FP2#AcI1fq7SFP2…FXRr™sRt5eI1f b4FP2’G8b47SGvA4I@fq7@FX2‚™ETeI@f˜0w24I15efxISRyFXGeTmTVI19PpcGeFPI12…FPTm7@9 u 7zY—T‡7SGVG8587Sw6G806AoGeI˜G8bc0uT87Sf˜0dCcij03A uglI1FXGeFP2“ Gz|{}b403TV0‡7@5e0~FPfhglI15VG87@2“G7S24AqFX2€Ge035V03TVGeFP24(uw3I124w603gcGeT3™jW4pcG 0vA4Iw24I@GA u 039P9lI12qGeb403f‚FX2 G8bcFPT u I15eƒsr„vp45yWE7@TVFPw}w6I124w3065e2hFPT u FHG8biG8FXf˜0…„`A406gt0624A4032“G„‡†‡ˆ}u ‰†TB7@2cAp„‡‡”ˆg ‰‘TBFP2 u b4FPwdb”G8bc0 w3I—06ˆhw6FP032“Gdf˜7SG85VFPw306Ta7@5e0”FX24A403gl032cA4032“GwI@RWlI@Geb§‰s7@24A‹Š…Œ|Te060p—03w6GeFPI12‹Ž4ji‘§0 u FX9P9’5V06Rt065 G8I”TVp4wdbf˜7SG85VFPw603T7@T”“D•—–@•™˜—šœ›l–@…žœŸbE7@gcGe035eT~Žh7@24Ao iA406q@039PI@gt06A…Geb40a5e03gc5e03TV032“Gd7œG8FXq@0wRtI15Vf˜T I@R”„‡†‡ˆ}‰†Te(@032406587SGe03A–Rt5eI1f¡Geb40xWE7@TVFPwx¢B†‡ˆ}‰†TuW€Y§G8b40xTV03fhFPA4FXTew65e06Ge0x7@g4gc5eI“7@wdb£™B7@24AhG8bc032 G8bc0v„‡‡”ˆ}‰†T(1032406587SGe03AhRt5eI1f¤Geb40m5V03g45V03TV032“Gd7SGeFXq@0‡„‡†‡ˆ}‰†TrW“Yq7Sg4g49PFXwz7SGeFPI12iISR£G8FXf˜0…„`f˜7@5ewdbcFP24( fh06G8bcIjA4T6y{}b403TV0u06¥—p47SG8FXI124T~7@5e0‡5V03g45V03TV032“G806A#W“Y ¦y‰ § ‰˜ § ­ ŠV® ­™¯ X°±® ¦ Š©¨yª § « ¬| 7@24A ‰E§ ²z³t´ ¨yµ ‰E§ ²‡« ¶ § ² ­™¯ ·@® 5e06Tegl03wDG8FXq@039HY1}¸EI15~7”I1240D„¹TVGe03g#fh06Geb4IjA£™4G8b40d9º7SGVG8065URtI15efxFXT~I@WcGd7@FX2403A…W“Y…7@g4g49HY—FP24(˜7”GeFPfh0D„ f˜7@5ewdb4FX24(pfh06Geb4IjA#GeIpG8b40i(@0324065eFPwi„a†‡ˆ»RtI15ef¼FP2#7xR½7SFP5e9HY’TVGe587@FX(1b“GeRtI15 u 7@5VA©fq7S24240356u¸EI15 0Dic7@fhg49X0@™EG8b40d0Dijg49XFPw6FXGmˆyp49P065~fh06Geb4I—A9P037@A4T}G8Iq¾ ¨À¿aÁÃÂEª ™l7@24A ¶ § ² ¨ ħ¬ ­½Å  ®…~Æo06Geb4I—A4T FP2“q@I19Xq—FP2c(G u I~I15Çf˜I15V0yTVGe03g4TÈwz7@2w7@9 u 7zY—TÇWt0 u 5eFHGeGe032dFP2‡G8b40yRtI15VfÀI@Rcˆy¥l ¯ }W€YvFP2“G85VIjA4pcw3FP2c( 240 u Ac03gl0324A4062“G‡q@7S5eFº7SW49P06T3~ÉmI@Ge0e7@9XTeIhG8bE7SGmI1249HY…f˜0DG8b4I—A4T”FX2 u b4FXwdbG8bc0wG8FXf˜0w7@24ATegE7@w60 A4FXTew35V06GeFPÊz7œG8FPI@24TU7@5V0mGe5e0z7œG803AxTV03gE7S587SGe039XY˜wz7S2xWt0 u 5VFXGVG8032qFP2x7@2qFP2“G8065efh03A4FP7SG80mTe03fhFH„¹A4FPTVw35V06G80 RtI15VfËTVp4wdb€7@Tiˆy¥s ¯ X°1…{}b40qRtpc9P9XY“„¹A4FPTVw35e0DG80”RtI15ef…™’ˆy¥s ¯ ·7S24AsG8bc0x7@TeTVIjw6Fº7SGe03AhTvGd7@WcFP9PFHG`Y A40DCE24FXGeFPI12cT7@24Ao7@2E7@9HYjTVFPT7@5V0a7@g4g49XFPw37@W49P0mG8Ih7@9P9£f˜0DG8b4I—A4T3 °Ì@Í .

°ÌSŽ ΒÏmДÑvÒÄÓBÔxÕ±ÖÃ×ÇÒ4ДØÙrÚÛÙ3ҐÜÞÝ}߀ÚÛÙrà”ÓyÐvÔá×ÇÜB×ÇÒÄÓrâ#× ãpä±å æèç‡ésç‡ê©ë#ç‡ê©ìçîíhê ï’ð#ç ñÀò3ó”ç‡ê©ôÄõ…ôÈï’ç‡ö „vp45eA40DCE24FHG8FPI@24TeI@R~TVG87@W4FX9PFXG`Y§7@5V0pWE7@TV03A€032“GeFP5e069XY§I12hG8bc0xWt06bE7zqjFXI15eI@RGeb40xb4I1fhI1(1062403I1pcT gE7@5vG8´ TmI@Rrˆy¥—T6 ¯ H°e7@2cA ¯ ·—”{}b40eTvGd7@WcFP9PFHG`YxI@Rrˆy¥s ¯ H°eA406gt0624A4T‡062€GeFP5V039XY…I12oG8b40w03FP(@0324TvYjTr„ G806f I@R ª h{}b40˜TvGd7@WcFP9PFHG`YISR}ˆy¥s ¯ ·xwz7@2QI@RfGe032§7@9PTVI…Wl0h5e039P7SG806A©G8I’G8b40h03FX(10324TvY—TVG806f¡I@R FXGeTuf˜7SG85VFHit#÷~I u 0Dq10356™FX2sG8bcFPTuw37@Te0pGeb40pTVFXG8p47SG8FXI12sFXTu24I@Ge¥—pcFXG80˜TeI©TVFPfhg49P0hTeFX24w30S™yFP2hI1pc5 7@g4gc9PFPw37SG8FXI124TG8IxgE7S5VG8FP7@9ÈA4FXøt035V032“G8FP7@9Ä06¥€pE7SG8FXI124T ­ 03Tegl03w6Fº7@9X9XY…b“Y—gt065eWlI19PFXweI12406Td®D™Ç7pTvGd7@WcFP9PFHG`Y A40DCE24FXGeFPI12wz7@2‹A406gt0624A‹I12©WtISG8b#G8bc0eG8FPfh0e7S24A©TegE7Sw30”A4FXøt035V0324w6FP24(cd{}bcFPTvFXTvA4FPTVw3p4TVTe06A§FX2 j06w6G8FXI12 ¯ †Žcy'”2E7@9HY—TeFPTI@RÈG8bc03Te0d03FX(1032cTVY—TVGe03fhTmbE7ST~Geb40aFPfhgtI15vGd7@2“G7SA4A403Ao7@AcqS7@2“G87@(10‡G8bE7œG FXGe(1FXq@03Th7@2€03TvG8FPf˜7SGe0xI@RmGeb40‹V–@•rùe7SG u b4FXwdbÃ7©TeI@9PpcGeFPI12€7@gcg45eI“7Swdb403Th7‹TVG8037@AcY“„`TvGd7SGe0’FXRm7 TVY—TvG803fúFPTwTVGd7SW49P0S©ŸI124TVFPA406587SGeFPI12 u FP9X9rWl0p(1FHq1062sG8I#f˜7SGe5eFPw603TaG8bE7œGebE7zq10xWlI@Gebhw3I1fhg49X06G80 7@24A”A40DRt03wDG8FXq@0}03FX(10324TvY—TVG806f˜T6™jTV030~j06w6GeFPI12”Ž4j†Íc™ u FXGeb”7‡5e03fhFP2cA4035ÈG8bE7SGr7vw6I1fhg49P0DG80UTvYjTvG806f wz7S2sWl0p7@5VW4FXGe587@5VFP9XYow39XI1Te0”G8I7oA40DRt03wDG8FXq@0qI12c0@™|FX2 u bcFPwdbhw37@Te0˜g4587@wDG8FXwz7@9r7@g4gc9PFPw37SG8FXI124T‡w37@2 f˜7@ƒ@0‡G8bc0dgc5eI1gl035vG8FP06T}I@RÈG8b40a9º7œGeG80657@g4gl0z7@5UGeI˜AcI1f˜FX2E7SGe0@ ƒ¹R ª 7@24A µ 7@5e0aTVG87SG8FXI12E7@5vY1™ u 0uw37@2‚™sFP2xG8bc03I15vY…7SG~9P0z7STVGz™E06TVGeFPf˜7SG80‡G8bc03FP5}Rtp42cAE7@fh032“Gd7@9 g45VI1gt065VGeFP03T6#¸EI15d0Dic7@fhg49P0S™|FX2–j06w6G8FXI12hŽ4†Íc 0qRtI1p424A§Rt5eI1fèI1p45dfhIjA4069U„‡†‡ˆ}‰‘TwRtI15dA4FXRº„ Rtp4TVFPI12o7@24Aogt065eFPI—A4FXw‡w3I12“q106w6GeFPI12 u b47SG”w6I1p49XAu Wl0w0…icgl03wDG806ARtI15G8b40d03FX(1032“qS7@9Ppc0dTVgt06w6Ge5ep4fhT I@Rg4587Sw6G8FXwz7@9|g4b“YjTVFPw37@9Bg45eI1Wc9P03fhT‡w3I@2€G87@FP2cFP24(pGeb403TV0˜g4bc0324I1fh03247c™ÄTV030h¸|FX(4‚Ž4H°1i{}b406Te0q0…i€„ gl03w6G87SG8FXI124T7@5e0~5e0DRt035V5e03A˜G8Idfq7@2“YiGeFPfh03TrFP2”G8b40RtI19P9XI u FP24(d7@2E7S9XY—TeFXTrISRÛTVG87@W4FP9XFXG`Yug45eI1gl035vG8FX03T3 {}b40DY…7@5e0aFPfhglI15VG87@2“GU032cI1p4(1b’G8I”Wl0dTVp4fhfq7@5VFPÊ603A’W“YpG8b40‡RtI19X9PI u FP2c(4û ¸ I15vAcFXøtp4TeFXI12©A4I1fhFP2E7œG803A#•4I u T‡Geb40˜ý4„¹03FP(@032“qS7@9Pp406T”G80624A‹G8I’9PFX0i7@9PI@24(qG8bc0h2c03(“7SGeFXq@0 ü 5VE z0 7@9Ç7œijFXT3 ¸ I15rgt065eFXIjA4FXwmw6I12“q106w6G8FXI12j„¹A4I1fhFP2E7SGe03Ah•EI u TG8b40‡ý4„¹03FP(@032“qS7@9Pp406TrGe0324ApGeIu9XFP0m7@9PI12c(wG8bc0 ü FXE f 7@(@FP2E7@5vY˜7œijFPT3 q ƒ`2þfq7S2€YkFX2€Ge035V03TVGeFP24(–w37@Te06T3™G8b4003FP(@032“qS7@9Pp406TqI@R‡Geb40#f˜7SG85VFPw603ThFP2þˆy¥€T3 ¯ X°©7S24A ¯  7 5V0pTepcˆhw6FP032“GeGeI©A406Ge035Vf˜FX240˜Geb40pTVG87@W4FX9PFXG`Y@ƒ`2–g45V06q—FPI@p4TewdbE7@gjG8035VT3™ u 0xA403TVFP(12E7œG803AhGeb403TV0 @ 03FX(1032“qS7@9Xp403To7@TýEÿ¼7@2cA¡ 4ÿ RtI15…ˆy¥—T6 ¯ X°§7@24A ¯ ·—™‡5e03TVgt06w6GeFXq1069XY@™w7S24A u 0 u FX9P9vCE2cA FXG w3I@2€q@0324FX032“GeGeI0Dic7@fhFP2c0hGeb40qTvGd7@W4FX9PFHG`Y©I@RUqS7@5VFPI1p4Tdfh06G8bcIjA4TdFP2§WtISG8b§Geb40qw6I1fhg49P0…isý€7S24A w3I@f˜g49X0D¢ i  §g49P7@2406T3 £¥¤§¦©¨¨!"¦©#%$'&()#¨102"¦©#%$¥34#'56$'¨1#879§@6&ACB"¨EDGF©#'HI#'7PF©Q3R7TSU#8VR#'7T¨10W¨1Q¨QX1¨7©#8&RB8Y .

÷ 065e0 u 0QTvGd7SGe0#G8b40‹TvGd7@24A47@5eA»TVG87@W4FX9PFXG`Ygw35VFXGe035eFXI12 p4TV03A RtI15pI15VA4FP2E7S5VYgA4FXøt035V032“G8FP7@9v03¥€pE7̄ m G8FXI124T6 ¸sI@5˜7h“D•—–@•™˜—šœ›l–@…ž’fq7œG85eF i ª ™}ˆy¥l ¯ H°oFXT’˜f›WdEùDdùD›E•)e ž©“D•¹–!f9ePù”FHRr™ u bc032 ¬ § FPT ­™¯ Í1® w3I124TvGd7@2“G3™ ‰o § 5e06fq7@FX24TUWlI1p424Ac03A#7STUŠ%g h  ÉmISG80‡G8bE7œGFP24b4065e062€G~TvGd7@W4FX9PFHG`YqA406gt0624A4T~I1249HYpI12’G8b40‡Ge587@24TVFP062€GTVI19PpjG8FPI@2…ISRÄGeb40w„a†‡ˆ}‰‘T3 ¹ƒ RE7f˜7SG85VFHi‡b47@T|7w3I@f˜g49X06Ge0r06FP(10624TVY—TvG803f…™17S9P91I@RcFXGeTÈ06FP(1062“q103wDG8I15VT’7@5V0B9PFP2c0z7@5V9XYvFX24A406gt0624A4032“G3™ 7@24A’G8bc0wf˜7SG85VFHiqwz7@2…Wl0wA4FP7@(1I1247@9PFXÊ303AqW€Y…7”TeFXf˜FX9º7@5VFXG`Y”G85e7@24TvRtI15ef˜7SG8FXI12‚’ƒ`2oTep4wdb7hw37@Te0dFXG RtI19X9PI u Ta7SGdI124w60hRt5VI1f¡ˆy¥sÄ j· ¯ ™ÇRtI@5d0…ic7@f˜gc9P0@™ÛGebE7SGaG8b40q„‡†‡ˆ}‰†Td7@5V0˜FX24b4035V032“G89HY©TVG87@W49X0hFHR 7@24A…I1249HYpFHR w ­ ý4ÿ}y ® x{z RtI157@9X} 9 | ­™¯ †Ž“® {}b4FXT˜TvGd7SGe03T˜GebE7SG3™RtI15hFP24b4065e062€GqTVG87@W4FP9XFXG`Y@™7@9X9}I@RvGeb40#ý 06FP(1062€qS7@9Xp403Thfep4TvGp9XFP0xI12‚™I15hGeI G8bc0q9P0DRfGuISRr™ÄG8bc0qFPf˜7@(1FX2E7@5vY©7œijFPTdFP2QG8b40˜w3I1fhg49X0Disý–g49º7@2c0@’ƒ`24Tegl03wDG8FX24(©¸|FX(4ÛŽ4H°1™ u 0pTe060 G8b47SGwGeb4FPTdw35VFXGe035eFXI12QFPTdT87SGeFPTVC403AsRtI15aG8bc0qfhIjA4069r„‡†‡ˆ‰‘Tu5V03g45V03TV032“G8FX24(#WlI@G8bhAcFXøtp4TeFXI12s7S24A W4FXw3I12“q106w6GeFPI12‚}ƒ¹GmTebcI1p49PAoWt0d03fhg4bE7STeFPÊ603A ­ 7@T~FXG~FPTm7@2…FPfhgtI@5VGd7S2€Ggc587@wDG8FPw37@9Çw3I12cTeFPAc03587œG8FPI@2 FP2…w3I@2€q@03wDG8FPI@2j„`A4I@f˜FX2E7SG806AoTvY—TVG806f˜T8®GebE7SG~G8b40dTegl03w6Fº7@9Çwz7STe0aRtI15 u bcFPwdb©ý ¨~l€ FPTa˜f‚ › Re„ƒW…—†ù … FP2hG8b40mA4I1f˜7@FP2hI@RÛTvGd7@WcFP9PFHG`Y1Äƒ`2qG8b4FXTw37@Te0”FXGFPTBG85epc0mG8b47SG ‰x Š g h ™jW4pjG § A4I—03T24I@GA403w37zYx7@Tr‡ 2406FXG8bc035~A4Ij06T”FHGm(@5eI u ™£TVI˜G8bc0e7@WlI±q10uw6I124A4FHG8FXI12oFXTmfh06G3¸|FP247@9P9HY u 0u24I@Ge0wGebE7SG~RtI15v„‡†‡ˆ}‰‘T w6I1f˜gc9P06Ge0’03FP(@0324TvYjTvG806f˜ThG8b40o03FP(@032“q106w6G8I@5eTqgc9º7zY€24Is5VI19P0’FX2gG8bc0oFX24b4035V032“GpTvGd7@WcFP9PFHG`Y uw35VFXFXGeGebà 035eFXI12‚ ƒ`2QI15eA4065‡G8I’p424A4065eTvGd7@24AQG8b40hTVG87@W4FX9PFXG`YoI@RU„‡†‡ˆ‰‘T‡G8bE7œGwb47±q@0˜Ac06Rt03wDG8FHq10h03FP(@0324TvYjTvG806f˜T6™ u 0 FP2cTegl03w6GG8b40”2E7SG8pc5e0”ISRÛG8b406FP5rTeI19XpcG8FXI124TrFP2p06FP(10624TegE7Sw30@B¸EI15BG8bcFPT u 0vA45e7 u I@2qG8b40m5e06Tep49HG8TUFX2 j06w6G8FXI124TaŽ4j†Í…7S24A§06Tegl03w6Fº7@9X9XY#I@2sˆy¥€T3ÇŽ4H”° “pG8I…Ž4H”° •xFP2©G8bE7œGwTV03wDG8FPI@2‚”ƒ`2s7@2Q03FX(1032cTegE7@w60 5e069º7SGe03A#GeIxA406Rt06w6G8FHq10”TVY—TvG803fhT‡G8b40eRtI@5ef¼I@RBG8b40i5V03g45V03Te062“Gd7SGeFXq10h03¥€pE7œG8FPI@2©wdbE7@24(106T‡Rt5eI1f 7 TeFX24(19X0”03¥€pE7œG8FPI@2pG8I”— 7 –@I15eA47@2xW49PI—wdƒsB¸EI150…i47Sf˜g49X0@™jFP24TvG80z7SA…ISR|ˆy¥s4Ž4‘ Ž de7uG`Y—g4FXwz7@9lRtI15ef)I@R G8bc0dbcI1f˜I@(1032406I1p4TgE7@5vGf˜FX(1b“G}Wl0 ˜™  žEŸ ˜™ ‰‚´ žEŸ ‰ ›´ žEŸ ˜™ ý ‰ ° ý › œ š ‰ ›   ¨ š ° ý   š ‰‰ œ   ƒ‚…„ƒ‚‰ˆ Š¡Q‹AQŒ‚–”…Žynm”¥p™r˜hsvtGs–‘ri’s ƒ‚…„ƒ‚…„ ‘‰—™i€jlk'§–Wnmo”'pqr˜hs2tusv–Wri€s ƒ‚…„ƒ‚8† ‡‰ˆhx‘“’•”…–WQ’•”'—™˜ ±Õ Ö1`cÖgÙràmϔÓBÔ~ÓràiÒg×ÇÒcÐvØÙrÚÛÙ3ҐÜÀÝ}ßyÝ6a‡Ócb × ãpäAe fÇê©ð#çhgBçvêxïpipïqsr©ò9tzò6ï‚õ ívuxwnæñ€yVô °ÌId .

° @  Ì Î’ÏmДÑvÒÄÓBÔxÕ±ÖÃ×ÇÒ4ДØÙrÚÛÙ3ҐÜÞÝ}߀ÚÛÙrà”ÓyÐvÔá×ÇÜB×ÇÒÄÓrâ#× RtI15 u b4FXwdboI@240‡CE24A4TG8b40aTVI19PpcGeFPI12 ‰‚´ ­ ŠV® ¨ ‰‚´ ­ z“®8¡4¢£ ‰ œ ­ ŠV® ¨ ¤ ‰ œ ­ z1® Á ‰‚´ ­ z1®rŠ…¥¦¡4¢A£ ‰  ­ ŠV® ¨ § ‰  ­ z“® Á ‰ œ ­ z“®`Š Á ° ‰‚´ ­ z“®`Š œ8¨ ¡4¢£ ­™¯  d@® ƒ`24Tegl03wDG8FX24(wGeb4FPTBTeI@9PpcGeFPI12‚™ u 0~Te060mG8b47SGBRtI15yTep4wdbqwz7@TV03Trw3I124AcFXG8FXI12 ¯ ‘ŽwfepcTVGrWt0fhIjAcFXCE06A G8IiGeb40‡RtI15ef w ­ ý4ÿUƒ ® ©ªz RtI157S9P« 9 | ­™¯  1® TeFX24w30URtI@5’g4p4 5V0}FPf˜7@(1FP247@5VYdýۙ“‰ œ 7@24Ai‰  u I@p49PAi(15VI ugu FHG8b4I1pjG’WlI1p424A ­ 9XFP24037@5e9HYaI15’¥€pE7@Ac587SGv„ œ FPw37@9P9HYc®yFHRۉ ­ z1®­¨ ¬ zuI15U‰‚´ ­ z“®l¨ ¬ zjy{}b403I15V06GeFPwz7S9P9XYiGeb4FPTw3I12cA4FXGeFPI12˜FPTTVpcˆhw3FX032“G}RtI15TvGd7@WcFP9PFHG`Y FP2hG8b40”Te032cTe0vI@Rč€Gd7SGe03fh032“G ¯ †ÍuTVFP24w60m¥Š ®4¡°¯²± ³‰± £ g zd7@Ty‡ Š g h RtI15U7S9P9s24I12j„¹Ê3065eµ I ´3B÷~I u 0Dq10356™ FP2–gc587@wDG8FPw37@9U7Sg4g49PFXwz7SGeFPI12cTwG8bc0xw35eFHG8065eFPI@2hfq7zYsWt0 u I15VGeb49P06TeTiTVFP24w60pG8b4065e0’fq7zY§Wt0’7q@035vY 9º7S5e(10m(15eI u G8bxISRÈGeb40vglI19XY—24I@f˜FP7@9sWt0DRtI15e0”Geb40v0…icglI124062“G8Fº7S· 9 ¶vGd7Sƒ@03T}I±q@03†5 ¸h7@2cAxW45eFX24(1TU7SWtI1pjG G8bc0xA403w37±Y@€¸Ep45VGeb4035Vf˜I15V0@™BI12k7©w3I1fhg4pcGe035iTVp4wdbÃ7©(15VI u G8b€f˜FX(1b“GiA406TVG85VI±YhG8b40xTVI19PpjG8FPI@2 g45VIjw603TeT~Wl06RtI15V0‡FXGw3I@p49PA’Wl0aGe035Vf˜FX2E7SG806A‚ ÉmI@Ge0G8b47SG’G8bc0UTVG87@W4FP9XFXG`Yaw6I124A4FHG8FPI@2 ¯ † ˜8ù ¹2Re„ƒ‘…—ùd“‚G8b40}FXfq7@(@FP2E7@5vYa7œijFPT u b4FPwdbiGe0324A4TÄG8I‡Wl0 I—w3w3pcg4FP06A˜W“YeGeb4003FP(@032“qS7@9Pp406TB5e039P7SG806AiG8IaW4FPw6I12“q106w6G8FXI12hg45eI1Wc9P03fhT3|÷mI u 06q10653™jw3I124AcFXG8FXI12 ¯ †  FPTBI@R£9PFHGeG89X0UI15B24Idg4587Sw6G8FXwz7@9EFXf˜glI15vGd7@24w60}FXR£TeFP(@24FXCEw37@2“G7@f˜I@p42“G8TBI@R‚AcFPTeTVFPgE7œG8FPI@2˜7S5e0g45e06Te062€G3 ãpä º »«¼©ö l ç gBòz· ì qst½ip ï qcr©4ò tzò6ï‚õ v í unw¿¾ ñ yVô {}b40u„‡‡hˆgw3I1fhgE7@24FXI12pG8I˜—G87SG806f˜062“G ¯ ÍiFPT ¸sI@5w7©“D•¹–@•½˜—šœ›l–@…ž”fq7SGe5eF i ™’ˆy¥s ¯ ·’FPT”›ƒ2ÀqùD…˜"e–IeÁe žx“D•¹–!f9ePù‡FXRr™ u b4032 ¶ § FPT ­™¯  ¯ ® w3I124TvGd7@2“G3™ ‰4§ ²”5V03f˜7@FP2cTUWtI@p4µ 24A406A#7@T Å g h  ‘§0~TV030G8bE7œG2€p4fh035VFPw37@94TVG87@W4FX9PFXG`YuA403gl032cA4TI@249XYiI@2”Geb40G8587S24TeFX032“GrTeI19XpcG8FXI12iI@R£Geb40m„‡‡hˆg‰†T6 {}b4FXT~Ac06CE24FHG8FXI12xI@RÄTVGd7SW4FP9XFXG`Y˜FPT}TeI@f˜0DG8FPfh03T}5V06Rt065e5e06A…GeIq7ST7@TVY—fhgcG8I@GeFPw‡I15UGeFPfh0‡TVG87@W4FP9XFXG`Y@ '”T 0vTvGd7SGe03Ax7œGGeb40”Wl03(1FX2424FX24(wISR‚G8b4FXTwdb47@gcG80653™4TvGd7@WcFP9PFHG`YiA40DCE24FXGeFPI12cT7S5e0”2cI@Gpc24FP¥€p40S '¤Ac06CE24u FHG8FXI12hI@RfGe032hp4TV03A€FP2hŸU¸’† 9PFHG806587SGep45e0hTVGe03fhTuRt5eI1f¡Geb40pA40Dq1039XI1g4fh032“GiI@R~¢B†‡ˆÞTeIœ„ 9PpjG8FPI@24TpG8bE7œGoAcIg24ISGo2c03w306TeTe7@5eFX9XY RtI19X9PI u G8b40QTe06f˜FXA4FPTVw35V06G80‹5VI1pcG80S ƒ`2 TepcwdbÀwz7@TV03T…FXG’FXT 7@g4gc5eI1g45VFº7SGe0xG8IQw3I124TVFPA4065˜TVFPfepc9XGd7S2403I1pcTe9XY§G8b40…06øt03wDG8T˜I@RvWlI@G8b€G8bc0…GeFPfh0’7@24AÃTVgE7@w60#7@gj„ g45VI±ijFXfq7SGeFPI12cT3’'ÀGeFPfh0D„`TVgE7@w60vA4I1f˜7@FP2xFXTUCcij03Ao7@24A’TVG87@W4FX9PFXG`Y˜FPT}Ac06CE2406AoFX2xG8035Vf˜T}ISR u bE7œG bE7@gcgt0624TG8IuTeI1fh0”24I@5ef)I@R‚G8bc0vTeI19XpcG8FXI12 u FHG8b4FX2˜Geb4FPTA4I1f˜7@FP2˜7@TGeb40vfh03TebqFP2“G8065VqS7@9PT(1IwGeI Ê3065eI˜7SGTeI@f˜0aw3I@24TVG87@2“Gm5e7SG8FXI4|‘§0aA4FPTVw3p4TVT~Geb4FPT}glI1FP2“GI@RÈq—FP0 u FP2oj06w6GeFPI12 ¯ ‘Ž4 ƒ‚‰ˆ6‚8† ‡‰ˆhx‘“’•”…–WQ’•”'—™˜ .

Ÿ I124TVFPA4065y7”TV06GyI@Rt„‡‡”ˆ€‰†T’(@IÌq@035V2403AhW“Ye7”w6I1fhg49P0DG8006FP(10624TVY—TvG803f…’{}b40TVGd7SW4FP9XFXG`Yaw35VFXGe035eFXI12‚™  7@w6w3I15VA4FP24(…G8I…G8b40˜w3I124AcFXG8FXI12QTe06GeFP2§ˆB¥s ¯  ¯ ™ÄRtI19X9PI u Td7SGuI@24w30qRt5VI1fè7oTvG8p4AjYsI@R}ˆB¥s| c·°“ 7@24A‹FHG8T‡w3I@f˜gE7S24FPI12#RtI@5vfep49HG8FPgc9P0c l„`5eI—I@GeT3™Ûˆy¥sÇ c†Í1 c”Ÿ9X0z7@5V9XY1™tRtI15‡Tep4wdbQTVY—TvG803fhTw7xGeFPfh0D„ f˜7@5ewdb4FX24(ifh06G8bcIjA’FPT}2€p4fh035VFPw37@9P9HYqTvGd7@W49X0aFXRÄ7@24A…I1249HYpFXR Æ ­  4ÿ}® ® Æ xÞ° RtI157@9X§ 9 |&7@2co AÇ ­™¯  “1® {}b4FXT…w6I124A4FHG8FXI12»TVGd7œG803TxGebE7SGz™mRtI15x2€p4f˜065eFXwz7@9‡TvGd7@W4FX9PFHG`Y1™m7@9P9vISRwGeb4È 0   06FP(1062€qS7@9Xp403T ­ WtI@Geb g45VFP24w6FPgE7@9’7@24AQTeg4p45VFPI1pcT3™ÇFXRrG8b4065e0˜7@5e0”7S2€Yc®‡fep4TvGw9XFP0eI@2§I15aFX24TeFXA40eG8b40”p42cFXGaw3FX5ew69P0iFP2‹G8bc0 w3I@f˜g49X0D¢ i  t„`g49P7@240S {}b4FPTBA40DCE24FHG8FPI@2iI@R£TvGd7@WcFP9PFHG`YaRtI15r„‡‡hˆg‰†TBFPTyw6I124TVFPTVGe032“G FXGeb”Geb40TVG87@W4FP9XFXG`YdA40DCE24FXGeFPI12eRtI@5 „‡†‡ˆ}‰†T6|'”(17@FP2vGeb40TV0324TVFXGeFXq10rwz7STe0BIjw6w3p45VTÈRtI@5ÇGeb40Bgt065eFXIjA4FXwD„¹u w3I12“q106w6GeFPI12afhIjA4069 u b4FXwdbug49P7@w306T G8bc0ɶVw3I15V5e06w6Gʸ’9XIjw37SG8FXI12#I@RBG8bc0ig45eFX24w3FXgE7@9Ë •Ì—5eI—I@G‡g45V03w6FPTe069XYI12©G8b40ip42cFXG‡w6FP5ew69P0 u bc035e0”G8bc0 TeI@9PpcGeFPI12aFPTÈI1249HY‡2406pcG85e7@9P9HY‡TVG87@W49P0S|¸Ep45vG8b40653™œRtI15È7mw3I@f˜g49X06Ge0r06FP(10624TVY—TvG803f…™SGeb40r03FP(@032“q106w6G8I@5eT g49P7±Yq24Ih5eI19X0‡FP2pG8bc0d2€p4fh035VFPw37@9‚TVG87@W4FX9PFXG`Yq7@TVTe03TVTefh032“Gz {}b40BA4FXTew3pcTeTeFXI12aRtI15£G8bc03Te0rTVY—TvG803fhTÈgE7@587S9P9P069PTlG8b40yA4FXTew6p4TeTVFPI12aRtI15ÛA40DRt03w6GeFXq@0„‡†‡ˆ}‰‘T3Äˆ’ic7@fhFP2c0 ˆy¥sÄ jX”° “…7@24AQ24I@Ge0˜FHG8TaTeFXf˜FX9º7@5VFXG`Y u FXGebsˆy¥s ¯  djh‘§0hTe030hG8bE7œGdRtI@5dAc06Rt03wDG8FHq10q„a‡”ˆ}‰‘TwG8bc0 5e06¥€p4FP5V03A…f˜I—A4FXC4wz7SGeFPI12qGeI ¯  “iFXT Æ ­  4ÿU® ® Æ © ° RtI157@9X‚ 9 |&7S24 AÇ ­™¯  •1® TeFX24w30”A406Rt06w6G8FHq10vTvY—TVG806f˜T}AcIeœ 24I@G(1pE7@5e7@2“G8030”WtI1pc24A403Ac2403TVTRtI@5 Æ   Æ ¨ °1™—RtI150…i47Sf˜g49X0mFX2xˆy¥s ¯ deFXR Æ   Æ ¨ °w7S24Ao06FXGeb4035‰ ­ z“®UI15}‰‚´ ­ z“®h¨ ¬ z u 0d(106G9XFP24037@5UI15}¥€pE7@Ac587SGeFPwa(15eI u Geb‚ ãp'ä Í Î òzö G ç Ïi’ é qeìÐ ç ip ï qcr©4ò tzò6ï‚Ñ õ qeê©ë Ò)íhÔ ê Óhl ç gyó”ç‡ê©ìçᗠí uow¿¾ € ñ yVô Õ 06G…p4T…24I u 0…i47Sf˜FX240©G8b40‹w3I@24w306gcGISRuTvGd7@W4FX9PFHG`Y FP2 7–A4FHø£065e032“G u 7±Y@¤ƒ`2»G8b40‹g45V06q—FPI@p4T A4FXTew3pcTeTeFXI12 u 0aw3I@24TeFXA4035V03AFX2’TeI1fh0‡A406G87@FP9tG8b40‡RtI19X9PI u FP24(e7@g4g45VI“7@wdb‚û °1{}b40w¢B†‡ˆ}‰†Tm7@5e0dw3I@2€q@035vG803AoG8Iq„a†‡ˆ}‰‘TmW“Y…7@gcg45eIzijFPf˜7SG8FX24(eG8bc0wTVgE7@w30wA4035VFXqS7SGeFXq106TI12 7eCE24FHG80‡fh03TVb‚ jƒ`24bc035e062“G”TvGd7@W4FX9PFHG`YqI@RÈG8bc0u„‡†‡ˆ}‰‘T~FXT}03TvGd7@W49XFPTVb403A…W“Yx(1pE7@5e7@2“G80603FP2c(”GebE7SG G w ­ ýly ® x{zc Íc{}FXf˜0…„`f˜7@5ewdbpfh06Geb4I—A4T7@5e0‡Ac06q1069PI1gl03A u b4FXwdbo(1p47@587@2“Ge030vG8b47SG Æ   ­ ý  ® Æ xÞ°w7S24AxG8b4FXT FXTUGd7@ƒS032oGeI”Wt0‡G8bc0dw6I124A4FHG8FXI12pRtI15}2€p4fh035eFXwz7@9£TVGd7SW4FP9XFXG`Y1 ƒ‚‰ˆ6‚‰ˆ Š¡Q‹AQŒ‚–”…Žynm”¥p™r˜hsvtGs–‘ri’s ƒ‚‰ˆ6‚…„ ‘‰—™i€jlk'§–Wnmo”'pqr˜hs2tusv–Wri€s Õ±Ö—ÖáÒÈÙvâoÓq××cÑtÐdÎ’Ó ×ÇÒcÐv؁ÙrÚÇÙ3ҐÜgÐvàGa¤Î}Ý}à—Ä’ÓBÔlŐÓràuÎ’Ó©Ýß Ý}‡”Ócb × °Ì ¯ .

° I“ Ì Î’ÏmДÑvÒÄÓBÔxÕ±ÖÃ×ÇÒ4ДØÙrÚÛÙ3ҐÜÞÝ}߀ÚÛÙrà”ÓyÐvÔá×ÇÜB×ÇÒÄÓrâ#× {}b4FXTv…—š±ùd“m(1pE7@5e7@2“G8030wG8bE7SG‡7pTVGd7œG8FPI@2E7@5VY…TvYjTvG806fo™Û(10624035e7SG806A©Rt5eI1f 7x¢B†‡ˆ I12©TVI1fh0TÖr¹jù†… Teg47@w30˜f˜06Teb‚™ u FP9P9|bE7zq10p7…2€p4f˜065eFXwz7@9yTVI19PpcGeFPI12‹G8b47SGdFPTdWtI1pc24A403As7@TaŠ ¨ Å Â g h p{}b4FXT …—š±ùd“w›£šœ•Ä(@pE7@587S2€Ge030uGebE7SG”A403TVFP587SW49P0dTeI19XpcG8FXI124Tm7@5e0u(1032406587SGe03A#FX2G8bc0wG8FXf˜0df˜7@5ewdb#gc5eI—w303TVT 7@TWlI@G8bxG8bc0aGeFPfh0‡7@24A…Teg47@w30df˜06Teb’FP2“Ge035VqS7@9XT7@g4g45VI“7@wdb…Ê3065eI4 ÉmI u 9P0DG”p4TmA406C4240eTVG87@W4FX9PFXG`YpFP2oGeb40wGeFPfh0D„¹TegE7@w60wTV0324TV0@”¸|FP5VTVGmw3I124TvG85Vp4w6Ga7”CE24FHG80wGeFPfh0D„ Teg47@w30dA4 I1f˜7@FP2x9HY—FP24( u FXGeb4FP2×zvxªØx½Ùþ7@24AÚz—xþŠyx{ÛdŸI±q1065mG8bcFPT}A4I1f˜7@FX2 u FHG8b…7”(1œ 5eFXA G8b47SGFPT}06¥—pcFPTeg47@w306AFP2’WlI@G8b’G8FXf˜0v7@24A’TegE7Sw30w7@24A’CcixG8bc0dfh03TVb§V–@•½˜—švW“YxGeb40a03¥€pE7SGeFPI12 ‡iŠ ܲ ¨ µ ‡Ø Ém0…i—GB5e03Acp4w30}I1pc5B„‡‡hˆQ7@g4gc5eIzijFPf˜7SG8FXI12uI@REGeb40}¢B†‡ˆ©G8Ia7”G u IS„`9X06q1069 ­ F—‘0@H™@G u I‡GeFPfh0D„`gc9º7@2406Td® RtI15Vfep49º7eFX2pG8b40‡RtI15Vf I@R|ˆy¥l ¯ ·jy{}bc0wb4I@f˜I1(@032406I1p4TgE7@5vGI@RÄGeb4FPTURtI@5efep49P7eFPT ‰E§ ²z³t´ ¨ µ ‰E§ ² ­—¯ H”° z“® ˆy¥s ¯ H”° z’FPTaT87SFPA§GeIoWl0qTvGd7@Wc9P0”FXRU7@2“Y§WlI1p42cA403A§FP24FHG8Fº7S9Äq106w6GeI153™ • ‰ Ýz™Äg45eI—A4p4w603Tu7…WlI1p424Ac03A TeI@9PpcGeFPI12#q@03w6GeI153™ ‰E§ ²c™È7@T‡Geb40”f˜06TebQTeb45VFP24ƒ€T‡G8I…Ê3065eI’RtI15a7’Cjic06A ܧ ²jh{}bcFPTaFPTvGeb40”w39º7STeTeFXwz7@9 A40DCE24FXGeFPI12dI@RETvGd7@WcFP9PFHG`Y1Çƒ¹GÄFPTÈI@RfG8062u5e0DRt035V5e03AeGeI”7ST Õ 7œidI15 Õ 7œi€„¥ÞFPwdb“G8fuY@035ÄTVGd7SW4FP9XFXG`Y1ÄŸ9X0z7@5V9XY 7@T}G8bc0dfh03TVboFX2“G8035vqS7@9PT}(1I”G8I”Ê6035eIc™sGeb40a2€p4feWl035I@RÈG8FPfh0‡TVGe03g4T6² ™ ߧ™sfep4TvGm(@I”GeI”FP2cCE2cFXG`YxFX2 I15VA4035rG8Iew6IÌq@035UGeb40v032“GeFP5e0~Ccij03A’A4I1f˜7@FX2‚™€TeIeG8bc0”w35VFXGe035eFXI12qFX2 ¯  ¯ FPT7u2406w303TVT87@5vYpw6I124A4FHG8FPI@2 RtI15UGeb4FPT}TvGd7@W4FX9PFHG`Yqw65eFXGe035VFPI12‚ {}b40~TeFP(@24FXCEw37@24w60}I@RtGeb4FPTyAc06CE24FHG8FXI12”I@R£TvGd7@WcFP9PFHG`Yw7@5eFXTe03TyGeb45eI1pc(1¢ b àÛ4– ¹âá v “ ã§dEùdšœ8Rù Ài™ u bcFPwdb TVG87SG806T~GebE7SG3™4FXR|7i2—pcf˜065eFPw37@9‚fh06Geb4IjA’FXT”“D•¹! – f9ePù ­ FX2’Geb40aTe0624Te0wI@R Õ 7œi4®}7@24‰ A dšœ›c“D˜P“D•`ùD›E•£G8bc032 FXG”FPv T dšœ › ä@ùD" å—ùD›E•½”' f˜0DG8b4I—A#FPv T dšœ›c“D˜P“D•rùD›E•ÈFXRBFHG”gc5eI—A4p4w306T‡24Ix035V5eI15 ­ FX2G8b40u{|7zY—9PI15mTe065eFP06T Te0624Te0z®FP2hG8b40m9PFXf˜FHGB7@TBG8b40mf˜06TebpTVgE7@w3FX24(e7S24AqG8bc0mG8FXf˜0TvG803gq(1IwGeIuÊ3065eI ­ u FXGeb Ü ²dCjic06A‚™4FX2 G8bc0~b“Y—gl035eWlI19XFPw}wz7STe0±®…y{}b4FPTBFXT’Rtp45VGeb4035BA4FXTew3pcTeTe06ApFP2˜j06w6GeFPI12 ¯ 1 “c’'»fh06G8bcIjA”FPT T dšœ › ä@ùDC å—ùD›E• FXR|FXG~w3I@2€q@035V(103T”GeIhGeb40u0…ic7@w6GmTeI19XpcG8FXI12o7@T~Geb40df˜06TebTVgE7@w6FP24(q7@24AoGeFPfh0aTVG806g(1IhG8IhÊ3065eIpFX2 G8bcFPT}f˜7@24240653  Ÿ9P037@5e9HY1™cGeb4FPT}FXT7@2’FPfhgtI@5VGd7S2€GUg45eI1gl035vG`Y1 '”g4gc9XY—FP24(”TVFPfhg49P0v5V03w6p45eTVFPI12’G8Ihˆy¥s ¯ X”° zj™ u 0‡CE2cA ‰E§ ² ¨Àµ ² • ‰§ Ý 7@24A”p4TVFP24(vq@03wDG8I15r7@24A”f˜7SG85VFHu i æc„¹24I15efhT ­ Te060~'mg4gl0324A4F ii'v®’7@24AiG8b406FP5’FX2403¥€pE7S9PFXG`Yd5V039º7œG8FPI@24T3™ w 0 E b z 7 @ q 0 u Æ¦Æ ‰E ‰ § Ý Æ¦Æ x ÆçÆ µ ² ÆçƔè‘Æ¦Æ ² ‰ § Ý Æ¦Æ x ÆçÆ µ Æ¦Æ ² è‘Æ¦Æ ² ‰ § Ý Æ¦Æ ­—¯ H°1°Ì® § ² Æ¦Æ ¨ Æ¦Æ µ ² • ¨§ B"C&AïPC¨$‰3™ ¨&R‚&AC¦H©PHP#‡BC3(@yH©3”BCF”¨&#8"#XvrF”0ê¨13(îIB§ ¦9)¨ë43RHI7h#'B"#8ì°ð93(ïPX1&A¨$Ë H©7 B"3RY ì©X#'@68í4&ƒF”¨1î#'B"#'79B"&AC¨3™@6&Êë6ìI#‡7©#'#†F”#8Fu0ê3RB§HP&RB"&Rì°3(X1¨$ H©B"3Réñ'ì©ò¤§ X7T #'¦©@6   h ¦ ° ï C¨  R 3 C¦©&™#q74ó‚ ô‘õö X$8¨E"#&A'X!B‰C&3RCïXPï”BCC¨#(3R÷47s C¦©ù"#Aø‚ 3Rü¥B‰ý¦Fú ù‰°ûÊü'ýçþ9ü§¨§ï°)#†Fuø¨1C¦Tÿø‚3™#87("¨1# B"#'¦²Xë· F” ¨1YîIv #8BC&A#8C#79'Q@ y #†&Aø7©¨¨X7¡X! 9BC'#'Y£ ¢Q#8BC#Ëø‚# B"74#¥ï©0ê@ƒ #'B‚#'"3B"ƒ © ï ƒ @ ' # " B  ¨ P   û þ ¥ ¤„4þâC3yC¦©##'D”&($…)3(X1ï©C¨3R7u3R0v"¦©§ õ © ¨Q § B ‚ ø â # ê 0 ' # B‚C3yC¦P# ¨$8&AXC3RXï”C¨3R7µù‰°ûÊü'ýç þ ¤„9þË"3·&y)C#†&RF”ëGÿ‰&C#‡C3RXï”C¨3R7vY .

±Õ Ö—ÖáÒÈÙvâoÓq××cÑtÐdÎ’Ó ×ÇÒcÐv؁ÙrÚÇÙ3ҐÜgÐvàGa¤Î}Ý}à—Ä’ÓBÔlŐÓràuÎ’Ó©Ýß Ý}‡”Ócb × °Ì°• jFX24w30‡G8bc0dFX24FXGeFº7@9lAE7SG87eq103wDG8I15FXT}WtI1pc24A403A£™sGeb40aTeI19XpcG8FXI12pq106w6GeI15mFXT}WtI@p424A406AoFHR Æ¦Æ µ Æ¦Æ xÞ° ­—¯ H°Ì1® uRtI1b45}06Tv5eGd0 7@W4ƦÆFXµ 9PFHG`ÆçY1Æ  5e06g45e06Te062€GeTu7S2€Y æc„`24I@5ef¼I@R µ w{}b4FPT‡FXTvI@RfG8062‹p4Te06As7@Ta7oR“ ƒ¥ªD˜—ùD›E•|w6I124A4FHG8FPI@2 ÉmI u»u 0”240306ApG8Iu5V039P7SG80~G8b40mTVGd7SW4FP9XFXG`YiA40DCE24FHG8FPI@24T(1FHq1062pFP2qˆB¥€T6 ¯ 1 “e7@24A ¯  • u FXG8bhG8bE7œG (1FHq1032sFP2hˆy¥s ¯ X°Ì—oƒ`2€ˆy¥€T3 ¯ “o7@24A ¯ 1•c™ÄTVGd7SW4FP9XFXG`Y©FXTu5e069º7SGe03AQG8IGeb40…“EùD•½V–IeV–!…@˜Áƒ““aI@R F™†0SX™ÈFXGeT‡03FX(1032“qS7@9Ppc0hISRf˜7œijFPfepcfnf˜7@(124FHG8p4A40S”ƒ`2sˆy¥s ¯ X°±j™ÇTVGd7SW4FP9XFXG`YoFPTa5e069º7SGe03A‹G8I…7 µ „¹™È æc 24I15ef¼I@R µ wƒ¹G‡FXTvw39X0z7@5mG8bE7SGeÿ• dE¢ ù D…˜f•`ùD…˜½––@8ùiÁ• dEùu“3I – Àq ù ËdEùD›sÁ• dEùu“ Eù D•½eI – eĝV! – …@Á˜ ƒ€“e˜P“h– •½R ƒE ù  `›£šœA À"! { u I©R½7@wDG8Ti7SWtI1pjGeG8b40q5e039P7SG8FXI12sWl06G u 03062gTVgt06w6Ge587@9U5e7@A4FXF7@2cA–fq7SGe5eF iQ24I15efhTi7@5V0 u 039P9 ƒ€24I u 2‚û °1{}b40dTegl03wDG85e7@9Ç587@A4FXp4TI@Ry7”f˜7SG85VFHixFXTFXG8u T Ù œ 2cI15ef u b4062oGeb40wf˜7SGe5eFHipFPT24I15Vfq7S9—™cF™†0SX™ FHG}w3I1fhfepcGe03T u FXGeb’FXG8TUGe587@2cTeglI1Te0S j{}b40aTVgt06w6G85e7@9Ç587@AcFPp4T}FXTUG8b4v 0 eP# š ùD fDP š ƒ—² › …aI@R’7@9P9t24I15Vf˜T6 ¸Ep45VGeb4035Vf˜I@5e0@™ u b4062x¾©FXT2cI15ef˜7@9—™1Geb40”TV03w6I124AxFP2c03¥€pE7@9XFXG`YiFX2pˆy¥s ¯ X°@°”Wl03w3I@f˜06TU7@2p06¥€pE7@9H„ FXG`Y@”ƒ`2#Geb4FPTmwz7@TV0@™Ûˆy¥s ¯ H°Ì˜Wl03w6I1f˜06T”WlI@G8b©2c03w306TeTe7@5VY#7@24A©Tepjˆ˜w6FP062€G”RtI15”TvGd7@W4FX9PFHG`Y1~¸E5eI1f G8bc03Te0h5e069º7SGeFPI12cT u 0hA4587 u G u IFXf˜glI15vGd7@2“G‡w6I124w39Xp4TeFXI124Td7@WlI1pcGaG8bc0h2€p4fh035VFPw37@9yTVG87@W4FP9XFXG`YoI@R fh06G8bcIjA4T}pcTe03A…G8I”TeI@9Xq10d¢B†‡ˆ}‰†T6 0”TvGd7@W4FX9PFHG`Yw65eFXGe035VFº7pFP2Qˆy¥€T3 ¯ 1 “’7@24A ¯ H°Ìx7@5e0”FPAc032“G8FXwz7@9|RtI@5‡TVGd7œG8FPI@2E7@5VYoTVY—TvG803fhT ü {}b4 G8b40p(@IÌq@035V24FP24(©f˜7SGe5eFHiQFXTu24I@5ef˜7@9—o{}b4FPTwFP24w69Pp4A406TeTVY—fhf˜0DG85VFPw@™’7@TVY—fhf˜0DG85eFXw@™ u7@2cb4A 062– w6FP5Vw3p49P7@2“Gxfq7œG85eFXw306T3¤{}b406Te0Qw35VFXG8065eFP7h7@5e0‹WlI@G8bþ2c03w306TeTe7@5VYy7@24A»TVpcˆhw3FP062“G…RtI@5 fh06Geb4IjAcTBGebE7SG(1032c03587œG80~Tep4wdbpf˜7SGe5eFPw603TB7@2cAqA406gt0624ApTeI@9P039HYep4glI12hG8b40~03FX(1032“qS7@9Ppc03TBI@R Geb40afq7SGe5eFXw303T6 R’Geb40uTVgt06w6Ge587@9È587@AcFPp4T~I@R~–@›Eža(1I±q1065e24FX24(pf˜7SG85VFHi’FXTm(@5e0z7œG8035G8b47@2#I1240S™lGeb40ufh06Geb4I—A ü ƒ¹FXTq p424TvGd7@Wc9P0#W“Yþ7S2€Ygw35VFXGe035eFXI12‚À{}b€p4TpRtI@5q(10624035e7@9”f˜7SG85VFPw603T3™}Geb40#Tegl03wDG85e7@9”5e7@A4FPpcT w6I124A4FHG8FPI@2xFPT}2406w303TVT87@5v% Y $‡W4pcG~24I@G~TVpcˆhw3FP062“G~RtI@5}TVGd7SW4FP9XFXG`Y1 r $¥÷PCïì© Pï” &RX1Xël C¦¨PQ#6 7© #8"¨$'17P#8"$¥CC¨&R3RBC7h ël¨$¥ 3(7© 7P3AF”Q$ ¨1C¨'3RBC7c ¨$'â "¦P0ê3R&B ‡3R"¦ïP ©B #6H©)HI #8B"H° $¥CB39‰)&A#†X‘§B"¦©&(#'F”B"¨#RïPY Ë3A00) ìI#yX#8"âC¦P&R7ú3(B#8ð9ïP&AX‘"3102 S 9 ¥ 8 "¦ © ” F  ¨ ÿ  1 ¨ C¨  R &  X © ï 35476 &(' .

°±Íz ãpä¥@ ƒ‚Fƒ‚8† ΒÏmДÑvÒÄÓBÔxÕ±ÖÃ×ÇÒ4ДØÙrÚÛÙ3ҐÜÞÝ}߀ÚÛÙrà”ÓyÐvÔá×ÇÜB×ÇÒÄÓrâ#× »«¼©ö çlgBòzì·qst½ipïqcr©4 ò tzò6ï‚õ Ò)íhê©ì~ç‡éï’ôxòzê ïyð#ç Ò)í˜ö ö é t6B çA C ÏEDxt9qeê#ç GIHQP ‘»b40DG8b4065I15B24I@GBGeb40mTV03fhFH„¹A4FPTVw35e0DG80~7@g4gc5eI“7@wdb u 7STrG87@ƒ@062qG8IaCE24AhG8b40~A4FXøt035V0324w6FP24(w7@g4gc5eIzi€„ FPf˜7SGeFPI12pISR’7iTe0DG~I@R’¢B†‡ˆ}‰‘T3™EG8b40‡CE247@9ÛA4FXøt035V0324w60a03¥€pE7SGeFPI124Tw37@2oWl0‡5e06g45e06Te062€Ge03A#W“Y ‰E§ ²z³t´ ¨ µ ‰E§ ²‡« ¶ § ² ¸Ep45VGeb4035Vf˜I@5e0~FXR µ bE7@T7ww3I@f˜g49X06GeX0 WU06FP(10624TVY—TVGe03f…™jGeb40”TVI19PpcGeFPI12hG8IdG8b40mb4I1fhI1(1032c03I1p4TgE7@5vG wz7S27@9 u 7zY—TWt0a0…icgc5e03TVTe06A©7@T ‰E§ ² ¨ Ü ¥´   ²´ ؂§ ´ Á è”è”è Á Ü q ÿ   ÿ² ØE§ ÿ Á è”è”è Á Ü`Y   Y² Ø § Y 065e0aG8b4­ 0  4ÿ»7@5V0‡G8b40‡06FP(1062€qS7@9Xp403TI@R µ yƒ¹R|G8bc0dTV03fhFH„¹A4FPTVw35e0DG80‡7Sg4g45eI17@wdb§˜P“p4TV03A‚™ u 0ww37@2 uCE24b4A’ 7w5e039P7SG8FXI12qWl06G u 03062xG8b4h 0  ‹7S24AqG8bc0aý’06FP(1062“q@7S9Pp406T3y{}b4FXTUTe065Vq106T7@T}7wq@035vYpw3I@2€q@0324FX032“G (1p4FXA40‡7@TUGeI u b4035V0 u 0af˜FX(1b“GU0…icgl03wDGG8b40­ t„¹5eI—I@G8TGeIi9PFX0”5e069º7SGeFXq@0vG8IeG8bc0‡p424FXGUw3FP5Vw39X0vFP2xG8bc0 w3I@f˜g49X0DÉ i  t„¹g49º7S240@’¸EI15‡G8bcFPTd5e0z7STeI12 u 0 u FX9P9yg45VIjw60306AsG8I…G8587Sw30˜Geb40h9PI—w3p4TdI@RUGeb40  t„¹5eI—I@G8T 7@T”7hRtp424wDG8FXI12#I@R’Geb40ug47@587@fh06Ge035”ý  RtI15G8bc0u03¥€pE7œG8FPI@24T”fhIjAc039PFX24(”A4FXøtp4TVFPI127@24A#gl035VFPI—A4FPw w3I@2€q@03wDG8FPI@b 2 á ¸ FX(1p45e0 ¯ X°iTVb4I u T‡G8b40i0…i47Sw6G‡G85e7@w30iISRrGeb40µ t„¹5eI—I@GvFHRFHGvFXTv(1032c03587œG803AQW“Yö¡ ¢¥ˆ 5e03gc5e03TV032“G8FX24( | 03FHG8b4065”A4FXøtp4TeFXI12kI15”W4FPw6I12“q103wDG8FXI12‚ ƒ`2gWtI@Gebgw37@Te06T˜Geb40 ü 5e03gc5e03TV032“G8T˜G8b40…TVG87@5VGeFP24(‹qS7@9Pp40 z#7S24“ A   ¨ °@’¸sI@5dGeb40qA4FHø£pcTeFPI@2§fhIjAc039—™|ý  FPTd5e0z7S97@2cAs2406(“7SG8FHq10So'mTwG8bc0 uf˜b47@06(15e240 FHG8– ¨ p4A40BI@Rtý  FP2cw35e037@Te06T3™1Geb40Ge587@w605e06g45e06Te062€GeFP24(mG8b40A4FPTVTeFPg47SG8FXI12wfhI—A4039€b4037@A4T|GeI u 7@5VA4T G8bc0dI@5eFP(@FP2p7ST”ý Âp g «hh r„v2xG8b40aI@Geb4035}b47@24A‚™4RtI15G8b40aW4FXw3I12“q106w6GeFPI12’fhIjA4069—™4ý Âp{ ¨ FPTu7@9 u 7zY—TeFPf˜7@(1FX2E7@5VY@’'”TuGeb40pf˜7@(124FHG8p4A40hI@R ‰ FX24w35V0z7@TV03T3™yGeb40˜Ge587@w60p5e06g45e06Te032“GeFP24¨ (©€9G8‰bc 0 W4FXw3I12“q106w6GeFPI12˜f˜I—A40394Ge587zq1069PT7@5eI@p424A”G8b40~w3FX5ew3pcfhRt065e032cw30~I@R£Geb40mpc24FXGrw3FX5ew69P0@™ u b4FXwdbqFHGB2406q@035 9P037zq103T6h‘§0”fep4TvGwWl0”wz7@5V06Rtp49yFX2‹FP2“G8065eg45V06G8FX24(   u b4032§FXGaFPT‡5V03g45V03TV032“G8FX24o ( ¡X’‘ ˆ ”{}b40iR½7@wDG G8b47SG”FHc G e ˜—ùd“”šœ›pGeb40up42cFXGmw3FP5Vw39X0dfh0z7S24T”I12c9XYxG8b47SG”Geb40xI – À“²e ˜f)• ƒ‘…—ù}I@R’G8bc0u5e06g45e06Te062€G87SG8FXI12#FXT w3I@5e5e06w6G3™£FXGGe039P9XTmpcT”24I@Geb4FP2c(qI@R’Geb4” 0 dc–œ“±ùU035V5eI15 ­ Te060eˆy¥s‚ c‘Ž4°Ì®…{}b40ug4b47@Te0e065e5VI15m5V039P7SG806T G8IiGeb4• 0 Eš±“D˜f•™˜—šœ›§šœ›”G8bc0dpc24FXGw6FP5ew69P0S cedgfihqprdtsvuwyx€x%0‚£fƒu„u†…#‚bf‡x ª—u—‡– ‡ ’£R‡ŒQsSPªrkTR˖”…Žy –‘— –ˆhVU¡˜l”…– ‘ ”¥’ŒQk' #· Cï©3A ì 02—ÿ#8C¦© $¥‡3 AC0‚H°F©#†#¥$¥0ê¨&A#8$¥XvC¨$8&R VRC#ƒ#R#'Y ˜¨  R#87PCë4)C#8@6¦P&RËìI#'#'7µ&(F©F”B"#8")#†F!Y™ôPBC3(@{7©3øÉ3(7úø‚#yø¨1XX•3R@ƒ¨10êï©B)"¦©#'B F ¨C$'–™eïP¤§ ” d "¦©B8)¨÷‘ 3R¨17T  ¨ Q  3Rïµ #u#8HP$¥&RC¨B"VR&R#8@y 쩨$¥3(79V(#8$¥C¨0 3RëR7G $8&RXC¨˜#8fR8#R÷”÷2BCC¦#†P)HI X1ë(#'Y C#8Bhgs0ê3(Bei©D”#8FsV&RX1ï©#†%3R0kj—#8ð9ïP&RX•"3mln1·&R7PFoâ0ê3RB%"¦©#GF©¨EîïPC¨1¨3R7v&R7PF .

Õ±Öqp€ÖÃàsr|âÓBԐÙDÎ|Ðvڀ×ÇÒcÐv؁ÙrÚÇÙ3ҐÜÀÎ}Ý}àeΒÓBєÒ|×oÙrà ÒÄϔÓáÎ}ݐâÑBÚÇÓutx §Ã—ÑBÚÛÐvà”Ó °±Íj° É I u 9P0DGvp4T‡w6I1f˜g47@5e0uGeb40i0Dic7@wDG­ t„`5VIjISG”G85e7@w306T u FXGeb©TeI1fh0uG8b47SGa7@5V0ig45eI—A4p4w603A‹W“Y©7@wDG8pE7@9 m G8FXf˜0…„`f˜7@5ewdbcFP24(ifh06Geb4IjAcT3U{Ä7@W49X0 ¯ H°dTeb4I u T”Geb40eýc„… Q5e069º7SGeFPI12cTRtI@5m7”qS7@5eFX06G`Y’I@R|f˜0DG8b4I—A4T3 ¸|FX(1p45e06T ¯ ·u7S24A ¯ †ÍuFP9X9Pp4TvG85e7SG80}G8bc0”5e06Tep49HG8TUgc5eI—A4p4w306AxW“Y˜qS7@5VFPI1p4Trfh06G8bcIjA4T u b4032pGeb406Yq7@5V0 7@g4gc9PFP06AG8I˜G8b40efhI—A4039|„‡†‡ˆ‰‘T”RtI15”A4FXøtp4TeFXI12#7S24A©gt065eFXIjA4FXwD„¹w3I12“q106w6GeFPI12‚™£ˆy¥—T6‚Ž4‘Žx7@24A©Žcdj ƒ¹G‡FXTvFPfhg49PFX03A…G8bE7SG”G8b40iWl03bE7zq—FPI@5vTeb4I u 2‹FPTmG`YjgcFPwz7S9’I@R u bE7SG u FX9P9ÈbE7@g4gl032‹FHRyG8b40ifh06Geb4IjAcT 7@5V0U7@g4gc9PFP06AdGeIvA4FHø£p4TVFPI12—„ ­ I15ÈA4FPTVTeFPg47SG8FXI12j„v®‚A4I@f˜FX2E7SG806AwI@5Ägl035VFPI—A4FPwrw3I12“q@03w6GeFPI12—„`A4I1fhFP247SG806A g45VI1W49P06f˜Tu7@T u 039X97@T u bE7SGiA4I—03TubE7@g4gl032–FX2sG8b40qf˜I—A4039rwz7STe03T6§ÆoI1TVGeI@R}G8b40qFPfhgtI15vGd7@2“G glI1TeTVFPW4FX9PFXGeFP06TU7@5V0‡w3I±q1065e03AW€YqG8b40aFP9X9Pp4TvG85e7SG8FXI124T3 °  ©«þ°«€ý ¨ z ˆ ijg49PFXw3FHG~ˆBpc9P035 |    œœ «k1ý   © « þ° ¨ z ´ Õ 0z7@gcRt5VI1( Í    « ­ ° Á œ…œ  ý  ®¥  Áœ œ ý ´ ¨ z ' €} • Ž   « ­ ° Á ´ œ ý  ®¥  Á ´ œa ý   ©« ´ W œ ý Âo¨ z ' €UÍ • d  q™°«ký ´  ®|«þ° ¨ z ´ ƒ`fhg49PFX3w FXGUˆyp49P065     œ ­ °~« œ œ ý  ®|« ­ ° Á œ ý ´  ® ¨ z {È587Sgt60 Ê3I1FXAE7@9 $ ý | ® «   z S  c 2 # A „‚€U@ 7 wdƒ u 7S5eA   ~ ° « ¯ œ­    Á  ¨ ´ “   œ ­ °~« ´ W œ ý ´   ®’« ­ ° ´ „ '‡ÆÍ œ ý ®'  Á ´ ´ œ ý ¨ z ´ œ Á ƒ œÂ •   « ­ ° ý Á œ œ $W ý   '®   Á ´ ´ œ ý  ­ ° Á Wœ ý  ® ¨ z '•€}ÆÍ °”z    « ­ ° ÁÁ @´ œý  ˜ ®¥´   œ Á œ œ ý   ©« œ ý ¨ z …d7@Ê6AE7@( °1°  ©«þ°«€ý  « œ´ ý œ  œ ¨ z´   ´ Þv†d °Ì  ©«þ°«€ý « ý « ý « ýb$ $ z Þv†‡Ž °±Í   œ ­ °~« ´ ý   ®|« œ $ ý    ‹« a ­ ° Á ´ ý  œ $ ® ¨  z ¨ ÆoFX9P240‡Ž@Geb {Ä7@W49X0 ¯ H°1rjI1fh0dý˜«  ÉÞ~069º7SGeFPI12cT ¸ FX(1p45e0 ¯ xTeb4I u Tu5V03TVp49XGeTdRtI15‡Geb40˜0…ijg49PFXw3FXGdˆyp49X035df˜0DG8b4I—A‚h‘»b4032§p4Te06AsRtI15aA4FPTVTeFXgE7SG8FXI12j„ | A4I1fhFP247SG806Auwz7STe03TyFHG’FXTÄTvGd7@W49X0RtI15ÈG8b40U5e7@24(10B„v2xþý  xGzc ­9Ž TVpE7@9P9HY‡G8b40fq7@(@24FXGep4A40rI@RtýhbE7ST G8IuWt0m03TVGeFPf˜7SG806Ap7@2cApI@RfG8062pFXGFPTBRtI1p42cApW“Y˜Ge5eFº7S9t7S24Ap035V5eI158®D|‘»bc032xp4TV03ApRtI15W4FPw6I12“q103wDG8FXI12 G8bcl 0  t„BG85e7@w30vR½7@9X9PTI1pcG8TVFPA40vGeb40ap424FHG}w3FX5ew39X0”RtI157@9X9tCE2cFXG80  ™s7S24AxG8b40‡fh06Geb4IjA’bE7ST}24I”587@24(@0 I@RÄTVG87@W4FP9XFXG`Y˜FP2xG8bcFPT}wz7STe0@  ‡cex%‚bxys‘…¡d ’“Œxiu¡wkdg ‚0‡ˆv“gxqz9‰Šfi‰‹u}vmhyz9x%…„Œx%uwydg v“g‚bwyx©z{x|p}d svpXd wyx~wyxiu¡wyd0qp {}b4FXTxFPT’7sG u IS„`5VIjI@Gpf˜0DG8b4I—A‚™~TeFP2cw30©G8b4065e0‹7@5V0#G u I  ’‰‘Txg45eI—A4p4w603A W“Yþ06q@035VYyýہ ‘»bc032 7@g4gc9PFP06AhG8I©A4FXTeTVFPgE7SGeFPI12sA4I1fhFP2E7SGe03Ahgc5eI1W49X03fhT u 0…TV030xRt5eI@f ¸|FX(4 ¯ ·GebE7SGeGeb40xg45eFX24w3FXgE7@9 .

λh a) Diffusion .°±Í“ Ι(σ) ΒÏmДÑvÒÄÓBÔxÕ±ÖÃ×ÇÒ4ДØÙrÚÛÙ3ҐÜÞÝ}߀ÚÛÙrà”ÓyÐvÔá×ÇÜB×ÇÒÄÓrâ#× Ι(σ) λ h= 0 ωh= 0 λ h= οο R (σ) R (σ) σ = e λ h . ωh oo ¸ FP(1pc5e0 ¯ H°1ûBˆ|ic7@w6GG85e7@w306TmISRË t„¹5eI—I@G8TRtI15}f˜I—A4069‚03¥€pE7SGeFPI12cT3 | 5eI—I@GdFPTaTvGd7@W49X0iRtI15d7o587S24(10”I@Rý  ™|W4pjGdGeb40hTeg4p45VFPI1pcTw5VIjI@GdFXTd2cI@Gzpƒ`2QR½7@w6G3™ÈGeb40hTeg4p45VFPI1pcT 5eI—I@G‡TvGd7@5vG8T‡I12#Geb40ip424FHG‡w3FX5ew69P0i7@2cA#R½7@9P9XT”I1pcGeTeFXA40iI@RrFXGmRtI15”–IeÁeUw ­ ý  ®l©zce÷~I u 06q@0356™ÇRtI@5 W4FXw3I12“q106w6GeFPI12‹wz7STe03T6™ u b4032sýQFXB T ²ƒj8ù”˜ÁÀh–”å@˜f›l–@…žœ™tGeb40if˜0DG8b4I—A©FPT”24I@GaI1249HYTvGd7@Wc9P0@™£W4pcGaFXG 7@9XTeI©g45VIjA4pcw303T”7 þGebE7SGeR½7@9X9PTwg45e06w3FPTV039HYhI12hGeb40xp424FHGiw3FX5ew39X0qFX2hG8b40x5e7@24(10 z x ‰ x¡°1 '”T u 7@T”gtI@FP2“G806AkI@pcGh7@WlIÌq@0@™BG8b4FXTiA4I—03T”24I@G”fh0z7@2£™BGebE7SGiGeb40’f˜0DG8b4I—A–FPT u FHG8b4I1pj G”035e5VI153 '”9HG8b4I@p4(1boGeb40aCE(1p45V0uTebcI u TmG8bE7SG~G8bc035e0wFPTm7˜5e7@24(10dI@R ‰ FX2 u b4FXwdbG8bc0u9P037@gcRt5VI1(˜fh06Geb4I—A g45VIjA4pcw303Tr24Ia035V5eI15BFX2˜7Sf˜g49XFXGep4A40@™SFXGBTe7±Y—Tr24I@Geb4FP24(d7@WlI1pcGyGeb40035V5eI15BFP” 2 dc–œ“±ùD’ÆoI15V0~FXTyT87@FXA 7@WlI1pcG}Geb4FPT}FX2Ÿb47@gcG806u 5 “c {}b4FXTiFPT”7@9XTeIQ7#G u IS„¹5eI—I@Gifh06Geb4IjA€W4pjGz™rp4249PFXƒ@0pGeb40’9P0z7SgcRt5eI1(©TVwdb403fh0@™rG8b40xTVg4p45VFPI1p4T”5VIjI@G TVG87@5VGeTd7SG‡G8b40hI15VFP(1FX2‚™£5e7SG8b4065‡G8bE7@2QI12©G8bc0hpc24FXGaw6FP5ew69P0S™ÈTV030h¸|FX(4 ¯ ji{}b4065e06RtI@5e0@™ÇG8bc035e0”FXT 7o5e7@24(10hI@R}5V0z7@9r2406(“7SG8FHq10qý  RtI@5 u b4FXwdbsG8bc0qfh06G8bcIjA u FP9X9yWt0˜TVGd7SW49P0Sx{}b40qCE(1pc5e0hTeb4I u T G8b47SG}G8b40a5e7@24(1œ 0‡0624A4T u b4032ý  © «d‡ ° Ÿz”TeFX24w30d7SG}GebE7SGglI1FP2“GUGeb40aTeg4pc5eFPI@p4T5eI—I@G}9P037zq103TG8bc0 Æ Æ w3FX5ew69P0”7@24A   Wl03w6I1f˜06Ta(15e037SG8065aGebE7@2QI1240Sq{}b40hTeFHG8pE7SGeFPI12©FXTa¥—pcFXG80”A4FHø£065e062€G u b4062sG8bc0 ýc„`5eI—I@G‡FXTvg4p45V0”FPf˜7@(1FP247@5VY@dƒ`2‹G8b47SGawz7@TV0h7@T ‰ FX24w35V0z7@TV03Td7 u 7zY©Rt5VI1fnÊ6035eIxGeb40”Teg4p45VFPI1pcT 5eI—I@Gu5e03f˜7@FX24TuFX24TeFXA40qG8bc0pw3FX5ew39X0x7@24A–5V03f˜7@FP2cTuTVG87@W49X0qRtI15i75e7@24(10xISR ‰ Q÷~I u 06q@0356™G8bc0 g45VFP24w6FPgE7@9r5eI—I@GdR½7@9P9XTwI@pcG8TVFPA40qGeb40pp42cFXGew6FP5Vw39P0hRtI15i7@9X9 ‰ž    zc™y7@24AhRtI@5wG8bc0xW4FPw6I12“q103wDG8FXI12 fb‡•0x|f‡d –yg—˜r…#yx%…š™šk‚£w›p¥—œ‚£pXwys‘dt…u¡wžŒxiu¡wyd0 b) Convection .oo σ=e i ωh .

Õ±Öqp€ÖÃàsr|âÓBԐÙDÎ|Ðvڀ×ÇÒcÐv؁ÙrÚÇÙ3ҐÜÀÎ}Ý}àeΒÓBєÒ|×oÙrà ÒÄϔÓáÎ}ݐâÑBÚÇÓutx §Ã—ÑBÚÛÐvà”Ó °±Í@Í σ1 σ1 λ h=-2 a) Euler Explicit ωh = 1 σ2 σ1 σ2 σ1 b) Leapfrog σ2 λ h=-1 σ1 σ2 σ1 c) AB2 ¸ FX(1p45e0 ¯ ·jûy{È587Sw303TI@Râ t„¹5eI—I@G8TURtI15UqS7@5VFPI1pcT}f˜0DG8b4I—A4T3 | fhIjA4069‚03¥€pE7œG8FPI@2’Geb40af˜0DG8b4I—A’FPT}p42cTVGd7SW49P0‡RtI157@9X9   e‡¡h…#‚¢x§x|£id ‰9y‚bz5Œx%uwydg Diffusion Convection { b40UGe587@gl03Ê6I1FPAE7S9jfh06Geb4IjAuFPT’7”q@035vYeglI1g4p49P7@5|I12c0RtI15|5V0z7@TVI124T’GebE7SGB7@5V0UgE7@5vG8FP7@9P9HY‡FP9X9Pp4TvG8587œG803A } FP2Q¸|FX(4 ¯ Íciƒ¹GeT­ t„`5VIjI@GeTvR½7@9P9ÄI12QI15‡FP24TVFPA40eGeb40”p424FHGaw3FP5Vw39X0eRtI15‡WtISG8b©G8b40”A4FXTeTVFPgE7SGeFP24(’7S24A G8bc0qgl035eFXIjAcFPwiw3I@2€q@03wDG8FP2c(#wz7@TV0x7@24A£™|FP2‹R½7@wDGz™˜f•m˜P“”“D•¹–!f9ePùv¤zšœ˜–IeÁe%䜖Ie„ƒsùd“xšT¤uý  ¤zšœ"Ëd—˜"d ýÀ˜f•t“±Rù e ¤q˜P“’˜fW › dEùD8ùD›EC• e ž“D•—! – f9ePùD –@p4TvGh9PFPƒS0qG8bc0…9X0z7@gjRt5eI1(©fh06Geb4IjA€FHGibE7@TeGeb40’wz7@gE7SW4FP9XFXG`YQI@R g45VIjA4pcw3FP2c(sI1249HY–g4bE7STe0065e5VI15˜RtI15iGeb40gl035VFPI—A4FPwxw6I12“q103wDG8FX24(–wz7@TV0@™}W4pjG˜Geb4035V0FPT˜7§f˜¡ 7 ¥VI15 A4FHø£065e032cw30~Wt0DG u 06032pG8bc0G u IeTVFP24w60Geb40G8587Sgt06Ê3I1FXAE7@9Efh06G8bcIjAhg45eI—A4p4w603Tr24Ie7Sf˜g49XFXGep4A40U065e5eI@5 RtI15d–@›Ež ‰s G ¥Vp4TvG”7i9XFPfhFXG806Ap587S24(10‡Wl06G u 0303 2 zvx ‰ xÞ°1  ¦ 24I@© .

°±Í@Ž xt‡§š‚b£%k‚b’“Œxiu¡wyd0 ΒÏmДÑvÒÄÓBÔxÕ±ÖÃ×ÇÒ4ДØÙrÚÛÙ3ҐÜÞÝ}߀ÚÛÙrà”ÓyÐvÔá×ÇÜB×ÇÒÄÓrâ#× {}b40B…a7@Ê3AE7S(‡f˜0DG8b4I—A u 7STrAc03TeFX(12406AhGeI‡g45eI—A4p4w60~9XI u gcbE7@Te0065e5eI@53BjFX24w30}FHG8T’wdb47@587@wDG8065eFPTvG8FXw glI19XY—24I1fhFº7S9|RtI1­ 5   FPTd7…w3p4W4FXw ­ {|7@Wc9P0 ¯ X°@™Ä24Ic°”z1®D™ÈFXGdfep4TvGebE7zq10hG ITeg4p45VFPI1pcTu5eI—I@GeTwFX2 7@A4AcFXG8FXI12GeI’Geb40ig45VFP24w6FPgE7@9ÄI1240Se{}b403TV0q7S5e0iTebcI u 2QFP2‹¸ÄFP(4 ¯ Íjuƒ`2Qu WtI@Geb#G8b40”A4FXTeTVFPgE7SGeFPI12 7@24AoW4FPw6I12“q106w6G8FXI12w37@Te06T3™£7˜Tegcp45eFXI1p4T~5eI—I@G~9PFPfhFXGeTG8bc0wTvGd´ 7@W4FX9PFHG`Y1U¸EI15}Geb40uA4FXTeTVFPgE7SGeFP24(”w37@Te0S™ 7pTegcp45eFXI1p4Tv5VIj I@G”9P037±q@03TvGeb40ep424FHG‡w3FX5ew69P0 u bc032§ý  ©¤« œ ™Û7S24A#RtI15mGeb40iW4FXw3I12“q106w6GeFP24(xw37@Te0S™ œ u b4062 ‰¨     BÉmI@Ge0aGebE7SGWlI@G8b’TVg4p45eFXI1p4T5eI—I@GeT7@5e0a9PI—wz7œG803A…7SG}Geb40aI15eFX(1FP2 u b4032#ý ¨ zc {È587@w603T’I@RaG8b40ö l„`5eI—I@GeTqRtI15qGeb40‹Te03w6I124Aj„˜7S24AþRtI1p45VGebj„`I@5eA4065ÔÞ~pc24(10D„†‡pcGeG87hf˜0DG8b4I—A4T’7@5V0 TebcI u 2ÃFP2g¸|FP(4 ¯ Íj {}b40oC4(1p45e06T˜TVb4I u G8bE7œGqWlI@G8bkfh06G8bcIjA4T˜7@5V0oTvGd7@W49X0xRtI15h7§587S24(10’I@R ý b4062Qý FPTm5e037@9’7@2cA#2403(17SG8FHq10@™£W4pcGvGebE7SG”Geb40e5e7@24(10eISRBTVGd7SW4FP9XFXG`YxRtI1T 5 ÞB†‡ŽpFXT”(15V0z7SGe0353™ (1I1 FX24u (w7@9PfhI1ÂTVGr7S9P94G8bc0 u 7±Y”GeId„v—“c™ u b4065e037@TƒÞv†ddFPTr9PFXf˜FHG806AeG8Id„rjy„v2hG8b40mI@G8b4065b47@24A˜RtI@5 W4FXw3I12“q106w6GeFPI1ö 2 Þv†d”FXTmpc24TVG87@W49P0aRtI15m7@9P9 ‰ ™ u b4065e0z7Sl T ÞB†‡Ž˜5e03f˜7@FX24TFP24TVFPA40aG8b40dp424FHG”w6FP5Vw39P0 0 Þv†‡ŽuTVG87@W4FX9PFXG`Ye9PFPfhFXGBFXT7SWtI1pjG Æ ý Â Æ ©»´  Ÿ1“dRtI@5 RtI1™ 5 z—x ‰ x %  ³ —B„v240vwz7S2pTeb4I u G8b47SGGeb4l 7@9X9£w3I@f˜g49X0Dioý  RtI@5 u b4FPwdÚ b w ­ ýtƒ ® x{zc ¹ƒ GrFPTBFP2“Ge035e06TVGeFP24(aG8Idg4p45VTep40~G8b40~¥€p403TvG8FPI@2qI@R£TVGd7SW4FP9XFXG`Y u bc032˜Geb40G8FPfh0}TVGe03g˜TeFPÊ60@™  ™—FXTBTef˜7@9P9 TeI˜G8bE7SG”7@w6w3p45e7@w6YoI@RB7@9X9ÛG8b40eý4„¹5eI—I@G8T~FPT~I@RyFXf˜glI15VG87@24w60@mjFXGepE7SGeFPI124T}RtI15 u b4FXwdb#G8b4FXTmFXTm24ISG G8bc0dw37@Te0d7@5V0aw3I124TVFPA4065e06AFP2…Ÿb47@gcG8065T“j ' 9X9|w6I12“q1032“GeFPI12E7S9|G8FXf˜0…„`f˜7@5ewdb4FX24(qfh06Geb4I—A4Tvg45VIjA4pcw30”7xg45eFX24w3FXgE7@9È5eI—I@GvGebE7SG‡FXT”q1065VY©w69PI1TV0 ” G8I ¡ ¢¥ˆ RtI@5pTef˜7@9P9~qS7@9Pp406T’I@Riý   {}b4035V06RtI15V0@™vI@2»Geb40‹WE7@TVFPTxI@RdG8b40‹g45VFP24w6FPgE7S9”5eI—I@G3™mG8bc0 TVG87@W4FX9PFXG`Y˜I@R|7if˜0DG8b4I—AxG8bE7SGFXT}5e06¥—pcFP5e06A…GeI”5e03TVI19Xq@0u7eGe587@24TVFP062€GTVI19PpjG8FPI@2xIÌq@035m7i5e039P7SG8FHq1069XY TebcI15VGmG8FPfh0dTegE7@2fq7zY…Wt0u7qfhIjI@G~FXTeTepc0@‡—p4wdb©wz7@TV03T‡7S5e0wG`Y—g4FHCE03A©W“Y’G8bc0e• ' €}˜7@24ö A Þv†d fh06G8bcIjA4T b4062#G8b40DY7S5e0w7@g4g49XFP03A’G8Iq7˜W4FXw3I12“q@03w6GeFPI12g45eI1Wc9P03f…}¸|FX(1T3 ¯ Sww7@24A ¯ ÍœRyTeb4I G8b47SGmRtI@5~Wlu I@Gebfh06G8bcIjA4T}Geb40ug45VFP24w6FPgE7S9‚5eI—I@GR½7@9P9XT}I1pcGeTeFPAc0dGeb40wpc24FXG~w3FX5ew69P0d7@24AoFPT~p424TVG87@W4u9X0 RtI15m7@9P9 ‰ ÷~I u 06q@0356™‚FXRÄG8b40aG85e7@24TeFX032“GvTVI19PpjG8FPI@2…ISRyFP2“G8065e06TVGvw37@2Wl0d5e03TVI19Xq@03A©FX27h9PFPfhFXGe03A 2€p4feWl035BI@RlG8FXf˜0UTvG806g4TrGebE7SGr7@5V0}Tef˜7@9P9jFP2iG8bc0~TV0324TV0mISRtGeb40}CE(1p45V0@™“G8b40065e5eI@5ywz7@p4TV03A˜W€YuG8b4FXT FP2cTVGd7SW4FP9XFXG`Yhfq7zYxWl0‡5e069º7SGeFXq@039XYqp424FXf˜glI15VG87@2“Gz’ƒ¹RÈG8bc0w5VIjI@GUbE7@A’R½7@9P9X032xFP2cTeFPAc0vG8b40aw3FX5ew69P0vG8bc0 fh06G8bcIjA u I1p49XA#bE7zq10iWl03062‹A403w69º7@5V03A#TvGd7@W49X0eW4pcG‡7@2#035V5eI15”I@RyG8b40uT87@fh0 Àh”– å@›E˜f)• ƒ‘…—ù u I1p49XA bE7zq10~Wl03032˜w3I1fhfhFXGeGe03A‚™¥¥Vp4TVGrFP2iGeb40I1g4glI1TeFHG80}A4FX5e03wDG8FXI12‚’¸EI15’Geb4FPTy5V0z7@TVI12˜Geb40'•€}a7@24A”G8bc0 Þv†wefh06Geb4IjAcT~b47±q@0wWlI@Geb…Wl03062p4TV03AoFP2’Te065eFPI@p4T¥—p47@2“G8FXG87SG8FHq10aTVGep4A4FX03TFP2“q1I@9Xq—FP24(”gl035VFPI—A4FPw w3I@2€q@03wDG8FPI@2‚þ{}b4FPThƒ€FP24AkI@R”FP24TvGd7@W4FX9PFHG`YhFPT”5V06Rt035V5e06AgG8I§7@Thf˜FX9PA€FP2cTVGd7SW4FP9XFXG`Ys7@24AÃFXT”24I@G˜7 Te065eFXI1p4Tg45eI@W49P06f¤p424Ac035G8b40aw3FX5ew6p4f˜TvGd7@2cw303TA4FXTew6p4TeTV03A‚ ŒV‰Šz9S‰9–yp`u‚  ‰Šz9‰‹u¼ se©ª’k‡•0x|f‡d –q0—‚£–y¬«kd hk…u¡wg—˜r…#kx|…s­šhq–k’tx‡—®€hku†u†‚yŒxiu¡wkdgqpQ©u­š®~¯°‚£–y±­š®€² ƒ‚Fƒ‚…„ 6 µ –bRe¶l”'k'”…–Itp‹A—™’·µGi¸R%k'k„¹»º .

8 Diffusion ωh= 2 2 ¸|FX(1p45V0 ¯ Íjûy{Ä5e7@w306TmI@Râ t„¹5eI—I@G8TRtI@5qS7@5VFPI1pcTUf˜0DG8b4I—A4T ­ w6I12“Gz‰‘As®D Convection .Õ±Öqp€ÖÃàsr|âÓBԐÙDÎ|Ðvڀ×ÇÒcÐv؁ÙrÚÇÙ3ҐÜÀÎ}Ý}àeΒÓBєÒ|×oÙrà ÒÄϔÓáÎ}ݐâÑBÚÇÓutx §Ã—ÑBÚÛÐvà”Ó °±Íd σ1 σ1 λ h=.οο σ3 σ2 σ1 d) Trapezoidal σ2 σ1 ω h = 2/3 σ3 e) Gazdag σ1 σ1 λ h=-2 f) RK2 σ1 σ1 g) RK4 λ h=-2.

°±Í1  ½ ‚¢u‚£pu…¡d xqwy‰Šf¨‰Š–yp`u‚  ‰9zŠ‰‹u¥¼ ΒÏmДÑvÒÄÓBÔxÕ±ÖÃ×ÇÒ4ДØÙrÚÛÙ3ҐÜÞÝ}߀ÚÛÙrà”ÓyÐvÔá×ÇÜB×ÇÒÄÓrâ#× {}b4065e0~FPTB7‡fep4wdb˜f˜I15V0}Te065eFPI@p4TBTVGd7SW4FP9XFXG`Ywg45eI1Wc9P03f©RtI15yTef˜7@9X9  GebE7SGrwz7@2hWl0~Wc5eI1p4(@b€Gr7SWtI1pjG W“YuGeb40U0DijFXTVG80624w30I@Rsw6035vGd7@FX2eG`Yjgl03TyISRsTeg4p45VFPI1pcT’5eI—I@G8T6y„v240}I@REGeb40UWl03TvGBFP9P9Xp4TVGe587SGeFPI12cTÈISREG8b4FXT ƒ€FP24AoI@Ryg45VI1W49X03f TvG803fhTmRt5VI1f 7˜w65eFHG8FPw37@9‚TvG8p4AcYoI@RÄG8b40wf˜I1TvGm7@w3w6p4587SGe0@™t0Dijg49PFXw3FHGz™4G u IS„`TvG803g£™ 9PFX240z7S5Ufep49HG8FPTvG806gxf˜0DG8b4I—A ­ Te060w{Ä7@W49X0 ¯ H°Ì®Dû ‰E²±³t´ ¨ «~Ž1‰E² Á dœ‰s² ¯ ´ Á  e ­—¯ H°±Í“®  ¾ S‰ ›² Á ‰ ›² ¯ T´ ¿ „v240xwz7S2gTVb4I u ™p4TVFP24(©Geb40xf˜0DG8b4I—A4Ti(1FHq1062kFX2gj03wDG8FPI@2g j c™yGebE7SGeGeb4FPTifh06Geb4IjA€FXTuG8b4FX5eAj„ I15VA4035”7@w3w6p4587SGe0uWlI@G8boFP2…G8035Vf˜T~I@% R ¡†À 7@24ö A ¡†ÀXÁ—™lTeIqRt5VI1f 7@2©7Sw3w3pc587@wDYoglI1FX2€GmI@R’q—FX0 FXG~FXT 7SGVG8587Sw6G8FHq10Sy÷mI u 06q10653™E9X06Gp4TUFP2cTegl03w6G¢ FXGeTTvGd7@W4FX9PFHG`Yi06q@032xRtI15Bq1035vYpTVfq7@9X94qS7@9Pp406TI@RÇý  u y{}b4FXT wz7S2Q037@TeFX9XY…Wl0”7@w3w6I1f˜gc9PFPTVb403AW€YTVGep4AcY—FP24(’FHG8TvwdbE7S587@wDG8035Vœ FPTvG8FPwugtI19HY—24I1fhFº7@9 u b4062sý  g zc ¸E5eI1f ˆB¥s ¯ H°±Í˜FHG~RtI@9P9PI u T}G8b47SGRtI15vý ’¨ zc™  ­9à ® ¨ à Á Ž à « djU¸l7Sw6G8I@5eFP2c(s ­  Ç® ¨ z CE24» A  ­  Û® ¨ ­œ  ‹«À°Ì® ­   Á d@® ¨ zc…{}b4035V0x7@5e0hG u o I  t„`5VIjI@GeT3‡ Œ  £´D™|Geb40qgc5eFP2cw3FPg47@9BI1240@™ u03¥€0q pE7S9‚G8I…°1™E7@24× A   ™E7”Teg4pc5eFPI@p4T}I1240S™E03¥€pE7@9ÛG8Ie¥„ d´Ä7Ä ƒ`2gI@5eA4065eG8I‹06qS7@9XpE7SG80pGeb40xw3I@24Te06¥—pc0324w603TpI@R~G8bcFPTe5V03Tepc9XGz™rI1240xfup4TVGhp424Ac035eTvGd7@2cAgb4I u fh06G8bcIjA4T u FHG8b’Tegcp45eFXI1p4T5eI—I@G8T u I15VƒpFP2’g45e7@w6GeFPw30Sy‘s0‡ƒ€24I u GebE7SG}G8bc06Y…7@5V0‡24I@GTe069XRÄTVG87@5VGv„ FP2c(4™l7@24AGeb40uTVgt06w3Fº7S9Ègc5eI—w303Acp45e06Tvwdb4I1TV032#G8I˜TVG87@5VGmG8b406f¼FX24FXGeFº7@9XFPÊ60”G8b40ww3I—06ˆhw3FX032“G8TmI@RyG8bc0 Tegcp45eFXI1p4T5eI—I@G8T6™EG8b40 Ü ÿ ® RtI1G 5 Ç   °wFX2ˆy¥sl cÍ@ cƒ¹RÄG8b40aTvGd7@5vG8FP2c(˜g45VIjw603TVTmFXT u 039X9ÛA403TVFP(12c03A G8bc03Te0˜w3I—06ˆhw3FX032“G8Tu7@5e0”RtI15Vw303AQG8I…Wt0”q1065VY§Tef˜7@9P9™™Ç7@24A§FXRrG8b40˜f˜0DG8b4I—A§FXTaTVGd7SW49P0S™ÈGeb406YQ(10DG Tef˜7@9X9P035 u FHG8bgFP24w65e0z7STeFP2c( Å »÷mI u 0Dq10653™FXRmG8b40…fq7S(124FXGep4A40’)´ I@ÝvR”Å I1240…I@RvG8bc0oTVg4p45eFXI1p4— T   FXT 03¥€pE7S9ÄG8¢ I dj™lI1240uw37@2#TV030eA4FXT87@TvG8065”FPT~FPfhfhFP24062€G~Wl03w37@p4Te0 ­ d1® ”° z‡Æ3mˆ’q1062§7”q1065VYoTef˜7@9P9 FP2cFXG8FP7@94qS7@9Ppc0mI@R Ü ÿ ® FPT¥—pcFPwdƒ€9XY˜I±q1035 u b4039Xf˜06A‚rjp4wdb’fh06Geb4IjAcTU7@5V0vwz7@9X9P06Aqwz7œGd7@TvG85eI@g4b4FPw37@9P9HY p424TvGd7@Wc9P0d7@24A…7@5V0 u I15vG8b49X03TeTRtI15}fhI1TVG3™4FXRÈ24ISG~7@9P9™™cw3I@f˜g4pjGd7SGeFPI124T6 ¹ƒ R u 0wFX24Tegl03wDGmGeb40  t„`5VIjI@GGe587@w603T~I@R|Geb40wfup49XGeFPg49X0v5eI—I@G~f˜0DG8b4I—A4T~FP2o¸|FP(@T3 ¯ ˜7S24A ¯ Íc™ u 0 CE24A‹G8bc03f GeI’Wt0iI@RrG u I…G`Y—gl03T3h„v240eG`Y—gl0”FPTvG`Y—g4FXC403A‹W“Y#G8b40”9X0z7@gcRt5VI1(xfh06G8bcIjA‚wƒ`2©G8b4FXT wz7STe07vTegcp45eFXI1p4Ty5eI—I@G’R½7S9P9PT}šœ›p•ÁdEù·ƒ—›E˜f•%D˜fÊRePù u b4062  g zcÄ{}b40I@G8b4065|G`Y—gt0}FXT’0Dij06f˜g49XFXCE06A W“Y‹G8b40pS24Aj„¹I15eA4065u'mAE7@fhTvT„ €7@TebcRtI@5VG8bs7@24› A …d7@Ê6AE7@(of˜0DG8b4I—A4T3…ƒ`2sGeb4FPTdwz7@TV0p7@9X9BTeg4p45VFPI1pcT 5eI—I@GeTUR½7@9P9Bšœ›©ÿ• dEùišœ…˜ å@˜f› u b4032  g zc {}b40mRtI@5efh035yG`Y—gt0~FPTB5e0DRt035V5e03A˜G8Iw7@T7ȘÁe ›£ùvÈQùD•ÁdEš9…@rjFP24w60~7SG9X0z7@TvGrI1240~Teg4p45VFPI1pcT5VIjI@G RtI15”7’ÆoFP9X240ufh06Geb4IjA#7@9 u 7zYjT‡TvGd7@5vG8TvI@2#G8b40epc24FXG‡w6FP5Vw39P0S™tGeb40efh06G8bcIjA#FXTv9PFXƒ@069XYxG8IxWl03w6I1fh0 p424TvGd7@Wc9P0vRtI15UTVI1fh0vw3I1fhg49X0Dixýo7@T  g45eI—w30603A4T~7 u 7±YqRt5eI1f)Ê3065eI4r„v2xG8b40‡W47@TeFXTUI@RÄ7e{|7zY—9PI@5 Te065eFX03Tv0…icg47@24TeFXI12‚™‚bcI u 06q@0356™ÇGeb403TV0if˜0DG8b4I—A4T”7@5e0e(@032406587@9X9XY’G8b40uf˜I1TvG‡7@w6w3p45e7SG80uFP24TVI@R½7@5v7ST G8bc06YxfhFP24FXf˜FXÊ30wÿ• dEµ ù dš±{ù ªD˜—ùD›E•£FX2xG8b40a9P037@A4FX24(eG8065ef¤RtI1· 5 ¡†À £  {}b40Q9º7SGVG8035˜G`Yjgl0‹FPTx5V06Rt035V5e06A GeIg7ST’7@Ê 2 Éu…“I – ÀuË “ ÈQùDÁ• dE9š …@‚jFP2cw30©RtI15pG8bc03Te0‹fh06Geb4IjAcT 7@9X9|Tegcp45eFXI1p4T‡fh06G8bcIjA4T”TVGd7S5VGd7SGvGeb40iI15VFP(1FX2RtI15 ©¨ zc™£Geb406Y#bE7zq10h7p(1pE7@5e7@2“G80603AQ587@24(@0eI@R ŒV‰Šz9–kx€‚£–yž™šy‚bw›pu¼0x§xÇwyxiu¡wkdgqp .

±Õ ֪̓ÖkàÍr|âoÓrԐÙDÎ|Ðvڀ×ÛÒ4ДØÙrÚÛÙ3ҐÜÞÎ}Ý}àeΒÓBєÒ|ׅÙrà¤ÒÄÏmÓÞÎ}ݐâÑBÚÈÓut¤ý ÎèÑBÚÛÐvàmӐ°zÍ ¯ TVG87@W4FX9PFXG`YaRtI15yTVfq7S9P9c0624I1p4(@b  y÷~I u 0Dq10356™jI@2”Geb40WE7@TVFPTyI@RsG8bc0~f˜7@(12cFXG8pcA40I@RlG8bc0}w3I—06ˆhw3FX032“G FP2eGeb40}9P037@A4FX24(v{Ä7±Y—9XI15’Te065eFX03Ty035V5eI15’Ge035Vfo™@Geb406YuTepcøt0356™€5e039P7SG8FHq1069XYwTegl0z7@ƒ€FP2c(4™@Rt5eI@f¤7@w6w3p45e7@w6Y@ ¸EI15’7–(1FXq@032y7@f˜I@p42“G’I@Ruw6I1fhg4pcGd7œG8FPI@2E7@9 I15Vƒl™mG8b40‹I15VA4035’I@Ru7@w3w6p4587Sw6YþI@RdG8bc0‹G I G`Y—gt06T|FXTÈ(@032406587@9X9XYv06¥€p4FXqS7@9X032“Gz™S7@24AdTVG87@W4FP9XFXG`Y”5e03u¥€p4FX5e06f˜062€GeTÄFX2eŸU¸|†Ã7@g4gc9PFPw37SG8FXI124T‚(@032406587@u 9X9XY I±q1065e5eFXA40‡G8b40 ­ p4Tep47@9P9HYqTVfq7@9X9f®yFP24w65e037@Te0aFP2’7Sw3w3pc587@wDYxg45eI±q—FPAc03A’W“Y’7iw3I—06ˆhw3FX032“G u FXGeb…9XI u 065 f˜7@(124FHG8p4A40S »«¼©ö l ç gBòzì·qst½ipïqcr©ò4tzò6ï‚õ Ò)íhê©ì~ç‡éï’ôxòzê ïyð#ç Ò)í˜ö éöt6çBA Ð"Ñ Dxt9quê#ç ƒ‚‘Ò6‚8† µ6b – Re¶l”'k'”…–Itp‹A—™¬ ’ ÓrR‡’pqÕ  Ô6‚ ãpäÏ { b40‹5e037@TeI@2 GeIkTvG8p4AjYþTVG87@W4FP9XFXG`Y€RtI15xTVfq7@9X9mqS7@9Xp403TxI@R  FXTpR½7@FP5V9XYk0z7STVY GeIkw6I1fhg45e06b40324A£ } ¢y5e06Tep4f˜7@W49HY u 0”7@5e0eTV0306ƒ—FX24(’G8Ix5e06TeI19Hq10iTVI1f˜0wG8587S24TeFX032“Gd7@24A‚™£TeFP2cw30eG8bc0”7@w3w6p4587@wDY©I@RBG8bc0 G85e7@24TVFP032“G‡TVI19PpjG8FPI@2RtI15v7S9P9ÈI@RrI1p45vfh06Geb4IjAcTvA403gl032cA4TaI12#G8b40uTef˜7@9P9X2403TVTvI@RyG8bc0eG8FPfh0D„¹TVGe03g‚™ Te0603ƒ©G8Iofq7Sƒ@0”G8b40hTeFXÊ30hI@RUG8bcFPTaTVGe03g–7@TdTef˜7@9X9B7@TwglI1TVTeFPWc9P0@p„v2§G8bc0qI@Geb4035dbE7@24A£™ÄG8bc0 uw3I@0q TVG~I@RÈG8b40‡w6I1f˜gcpcGd7SGeFPI12x(@032406587@9X9XY˜A403gl0324AcTmI12’G8bc0‡2—pcfeWt065I@R|TvG806g4T~G87@ƒ@062…GeI”w3I1fhg4pcGe0 7pTeI@9PpcGeFPI12‚™‚7@24A©G8IxfhFP24FXf˜FXÊ30‡G8bcFPT 0 u FPTeb©GeI…f˜7@ƒS0uG8b40”TvG803g‹TVFPÊ30”7ST‡9º7@5V(10i7@T‡glI1TeTVFPW49X0@ ƒ`2oGeb40dw3I1fhg45eI@f˜FXTe0@™ETvGd7@WcFP9PFHG`Ypw37@2ug49P7zY’ 7hg47@5VG3U'”TVFPA40aRt5eI1f¤5ep49XFP24(”I1pcGmwz7œGd7@TvG85eI@g4b4FPw37@9P9HY p424TvGd7@Wc9P0‹f˜0DG8b4I—A4T6™vb4I u 0Dq10653™aG8b40‹TeFHG8pE7œG8FPI@2 FX2 u b4FPwdb 7@9P9vISRwGeb40QGe587@2cTeFP062“G…Ge035Vf˜T…7@5V0 5e06TeI19Hq106Aow6I124TVGeFXGepcG806T”7i587œG8b4065~fhFP2cI155VI19P0‡FX2…TvGd7@WcFP9PFHG`Yqw6I124TVFPA406587SGeFPI124T6 €Y…R½7@5mGeb40efhI1TVG”FPfhgtI15vGd7@2“G”7@Tegl03wDG‡I@RB2€p4fh035VFPwz7S9ÄTvGd7@W4FX9PFHG`Y’Ijw6w3p45VT‡p424A4065vw3I12cA4FXGeFPI124T u b4062‚û 0˜bE7@TuFX24b4065e032“Ge9XYQTVG87@W49P0S™|w6I1p4g49X03AhTvYjTvG806f˜T u FHG8bhW ý ̗06FP(1062“q@7S9Pp406TubE7zq—FP24( u FPA4069XY ü TV„v0324 gE7@5e7SG806Aof˜7@(12cFXG8pcA403T6 ‘ 0#Te0603ƒgI1249XY€GeIsCE24A»7hTvG80z7SAcY“„`TVG87SG80TeI19XpcG8FXI12 p4TVFP24(–7§gE7SG8b GebE7SGqFP24w69Pp4A406TqG8bc0 ü § p 2 u 7@2“G806AG85e7@24TeFX032“Gz 4 ƒ`2gWlI@Geb–I@RmGeb403TV0’wz7@TV03TiG8b4065e0’0DijFXTVG”FP2hGeb40’03FX(10324TvY—TVG806f˜T”5V039º7œG8FXq@039HYs9P7@5e(10˜q@7S9Pp406TeI@R Æ ý Â Æ 7@TeTVIjw6Fº7SGe03A u FHG8b…03FX(1032“q106w6GeI15eT~G8bE7SG u 0 u FPTVb…GeIqA45VFXq@0dGeb45eI1pc(1b…Geb40uTVI19PpjG8FPI@2…gc5eI—w303TVT Geb4I1pcGx7@2“Y 5e06(“7@5VA»RtI@5qG8b406FP5˜FP24A4FHq—FPA4pE7S9”7@w6w3p45e7@w6Y FP2þ03FX(1032cTegE7@w60@ {}b4FPTxTVFXG8p47SG8FXI12ÃFXT uG8bcFX0e fq7 ¥VI15”fhI@G8FHqS7SG8FXI12xRtI15mGeb40eTVGep4AcYI@RB2€p4f˜065eFXwz7@9ÈTVG87@W4FP9XFXG`Y@”ƒ¹Gv9X0z7@AcT”G8IqGeb40eTVp4| W ¥V03wDGaI@R TVGeFXøt2403TVTA4FPTVw3p4TVTe03AFP2xG8b40a240…i—Gmwdb47@gcG80653 šœ .

ƒ`24b4065e062€GaTvGd7@WcFP9PFHG`Y’I@RB7qTV06G‡ISR„a†‡ˆ}‰‘T u 7@TvA40DCE2403A‹FX2©j06w6G8FXI12 ¯ q7@2cA‚™£RtI15~w3I1pcg49P06A©Te06GeT 7vw3I@f˜g49X06Ge003FX(1032cTVY—TVGe03f…™“FXG|7@f˜I@p42“G803AeGeI”G8bc0U5e06¥—pcFP5e06f˜062“G’G8bE7œG’G8b405e0z7S9cgE7@5vG8T’ISRl7@9P9 uýqFX03GeFXbi (1032“qS7@9Xp403TBfep4TvGr9PFP0I12‚™€I15yG8IaG8b409X06RfGBI@Rr™“G8bc0~FXfq7S(1FP2E7S5VYu7œijFPTyFP2”G8bc0~w6I1fhg49P0…ihýqg49P7@240@ {}b4FXT}Te035vq106T”7@TU7@2’0Dijw3069P9P062“G5e06Rt065e0624w30‡Rt587Sf˜0~G8I”A4FPTVw3p4TVT7@24A’A40DCE240‡G8b40‡(@032406587@9‚TvGd7@WcFP9PFHG`Y Rt0z7œG8p45V03TI@RÈG8FPfh0D„¹fq7S5ewdb4FX24(efh06G8bcIjA4T6B¸EI15}0Dic7@fhg49P0S™ u 0wTvGd7@5vG u FHG8bxG8bc0dAc06CE24FHG8FXI12‚û 'Ã2€p4fh035VFPw37@9€f˜0DG8b4I—AuFX™ T ƒ—‚ › dšœ² › …@˜f•™˜—šœ›lI – eÁe ž~“D•¹! – f9ePùÇFHREFXGÈFXT|TVG87@W49X0rRtI15Ä7@9X9c„‡†‡ˆ}‰‘T G8b47SG~7@5e0aFX24b4035V032“G89HYxTVG87@W49X0@ 'nfh06Geb4IjA FXGebgGeb4FPThg45eI@gt065VG`YkFPT”T87SFPAgGeIsWl~ 0 Ém—“D•—! – f9ePùD '¼fh06G8bcIjAkFPT ª Ü —“D•¹! – f9ePù”FHR G8bc0h5V03(1FXI12‹I@RTVGdu 7SW4FP9XFXG`Yow3I12“Gd7SFP24TvGeb40”2403(“7œG8FXq@0”5e0z7S9B7œijFPT‡FP2©Geb40”w3I1fhg49P0…iQý  “ g49P7@240S™Ä7S24A – f9ePù~FXRBFHGvw6I12“Gd7@FX24T”G8bc0e032“G8FX5e0eFXfq7S(1FP2E7S5VY…7œijFPT6ˆ  €Y©7Sg4g49XY—FX24(x7˜R½7@FX5e9HY’TeFPfhg49X0wG806TVGvRtI@5 Ý —“D•—! '~„¹TVG87@W4FP9XFXG`YuFP2˜Ge035Vf˜TrI@RÛglI1TeFHG8FHq10~5e0z7S9sRtpc24w6GeFPI124TrG8IdG8b40mw39º7STeTrI@R‚G u IS„`TvG803g Õ Æ#ÆQ‰†T(1FXq@032pFX2 j06w6G8FXI12… c dI•c™4I12c0aC424A4TG8b406Te0df˜0DG8b4I—A4TUG8I”Wl0a'~„`TvGd7@Wc9P0‡FXR|7@24A’I12c9XYpFHR ° Þ}ßáà ­—¯ H°zŽ€® Á ° ° d1® ­—¯ H© â ß « ° ­—¯ H°± “® â x Þ Á à«  'áTe0DGmISR|'„`TVG87@W49X0‡FPfhg49PFXw3FXGfh06G8bcIjA4T}FXT}Teb4I u 2oFP2’{|7SW49P0 ¯ j Æo0DG8b4I—A „v5VA4035 Þ à â °  zz zz ƒ`fh g47@9Xgl FPw303FHGU ˆyFPAEp47@9X039 5 ° ° ã1 ¡ Ä { e 5 3 Ê @ I  ° °¡zã1 «d¡ z ãSŽ S24'” AoAE„ä €U7@wdƒ u 7@5eA  Í¡ i ãSŽ ° @ 7 h f } T ` G — Y t g 0 °°¡ã@ Í «w °°ã1   «d °°¡ã@ ¡ Í { IS„`TvG803g… T gt06Ê3I1FXAE7@9  Õ 06G8035e7@ ã1 w « ã1 d « ã1 d‡ãI“ «w ° ã@  «”‡  ãI• u '„`w6I12“G8587Sw6G8FHq10  { 7@Wc9P0 ¯ jrjI1fh0‡p424w6I124A4FHG8FXI12E7@9X9XYhTVG87@W49P0 ­ '~„`TvGd7@Wc9P0±®rFPfhg49PFXw3FXGfh06G8bcIjA4T6 | ÉmISG8FPw60vG8bE7œG24I1240aI@RÈG8b406Te0afh06G8bcIjA4T}b47@T~7@2…7@w3w6p4587@wDY’b4FP(@b4035UGebE7@2’Te06w3I124A—„`I15VA40356rƒ¹G~w37@2 Wl0hgc5eI±q106A§G8b47SGaG8b40”I15VA4035aI@R7@2Q'~„`TvGd7@Wc9P0 Õ Æ#Æ e–@›E›£šœ•‡8ù ¹2dùd†ù ……Š• šœ™È7@24A£™ÇRtpc5VG8bc035efhI15V0 G8b47SG…I@Ri7@9X9aS24A—„`I15VA40356™v'~„¹TVGd7SW49P0©fh06Geb4IjAcT’Geb40‹G8587Sgt06Ê3I1FXAE7@9vfh06Geb4IjA»bE7ST…Geb40‹Tef˜7@9P9X03TVG G85Vp424w37SG8FXI12’035e5VI153 ƒ‚‘Ò6‚…„ °±Í“ U ˜hŒ—™˜•Ö­”…–‘”¥—y˜•R%kšµ6–bRe¶l”'k8”‰–tØהٱH%µ6–¢Re¶lk'VÚ –‘ˆh—ÛÖls ΒÏmДÑvÒÄÓBÔxÕ±ÖÃ×ÇÒ4ДØÙrÚÛÙ3ҐÜÞÝ}߀ÚÛÙrà”ÓyÐvÔá×ÇÜB×ÇÒÄÓrâ#× .

ձ֪̓ÖkàÍr|âoÓrԐÙDÎ|Ðvڀ×ÛÒ4ДØÙrÚÛÙ3ҐÜÞÎ}Ý}àeΒÓBєÒ|ׅÙrà¤ÒÄÏmÓÞÎ}ݐâÑBÚÈÓut¤ý ÎèÑBÚÛÐvàmӐ°zÍ• Þ~0DG8p45V24FP2c(pG8IpG8bc0iTVGd7SW4FP9XFXG`YxG806TVGap4TVFP24(xglI1TVFXG8FHq10e5V0z7@9ÈRtp424wDG8FPI@24T‡I1240ew37@2‹Teb4I u G8bE7œGa7 G u IS„`TvG803g Õ Æ#ÆèFXT “ ª Ü „`TVG87@W49X0‡FXR|7@24A’I1249HYpFXR ° Þ}ßáà ­—¯ H° ¯ ® Á ° ­—¯ H”° ““® â ß «  zvx Þ « à ­—¯ H”° •“® ¸EI15”CE5VTVGv„`I15VA4035d7@w6w3p45e7@w6Y@™ÈGeb40iFP2406¥€pE7@9PFHG8FX03T ¯ H° ¯ GeI ¯ X”° •’7@5V0i9P06TeTaTVGe5eFP2c(1032“G‡G8b47@2 ¯ X°zŽpGeI ¯ H°± cB¸EI15}TV03w3I@24Aj„`I@5eA4065”7@w6w3p45e7@w6Y@™sb4I u 0Dq10356™lGeb40agE7@5e7@f˜0DG8065eT Š­ Þ|å â å¥à ®7@5e0a5V039º7œG803A’W“YxG8bc0 w3I@24A4FXGeFPI12 ° à â « Þ ¨ Á  7@24A’G8bc0aG u IhTe06GeTmISR|FX2403¥€pE7@9XFXGeFP03T5e06A4p4w30aG8I”G8bc0dTe7@fh0‡Te06G u b4FXwdboFXT  z“® ­—¯ I â x  Þ «þ° ° ­—¯ j°Ì® â ß «  ÷m0624w30S™jG u IS„`TvG806g‚™cTV03w3I@24Aj„`I@5eA4065U7@w6w3p45e7SG80 Õ ÆƧ‰†TG8bE7SG7@5V0m'„`TVG87@W49X0m7S24A B „`TvGd7@W49X0TebE7@5V0 ª Ü G8bc0qT87Sf˜0 n­ àhå â åÞ ®agE7S587@fh06Ge035dTegE7@w60@…'”9HG8b4I1pc(1b§Geb40qI15VA4035dI@R7Sw3w3pc587@wDYsI@R}7@2h'~„¹TVG87@W49P0 fh06G8bcIjAhwz7@2c24I@GU0…ijw30306ApG u Ic™ B „`TvGd7@Wc9P0 Õ ÆƼfh06G8bcIjA4Tr0DijFXTVG u bcFPwdbqb47±q@0v7@2q7@w3w6p4587Sw6YhI@R 7@5VW4FXGe587@5VFP9XYhb4FX(1b’I15eAc0353 ª Ü ƒ¹GxbE7@TqWt06032»Teb4I 2þG8b47SGpRtI15q7hf˜0DG8b4I—AÃG8IsWt0#ƒ—„¹TVG87@W49P0FXGpfup4TVGp7@9PTVIhWt0'~„`TvGd7@Wc9P0@ {}b4065e06RtI@5e0@™s24IiRtp45vG8b406u 5A4FPTVw3p4TVTeFPI@2oFXT}2403w603TeTe7@5VY’RtI15UGeb40aTegl03w6Fº7@9Ûw37@Te0aI@R㗄`TvGd7@W4FX9PFHG`Y1 ƒ¹GhFPTe24ISG˜AcFXˆhw3p49HGeG8I‹g45VI±q10xG8bE7œG˜fh06Geb4I—A4TibE7zq—FP24(Q7©wdbE7@5e7@w6Ge035eFXTVGeFPwxglI19XY—24I1fhFº7S9BRtI@5 FXwdb#G8b40iw6Ij0Dˆ˜w6FP062€GvI@RBG8b40ebcFP(1b406TVG‡I@5eA4065”G8065ef¼FP2©ˆ»FPTmp424FXG`´ Y Ƈwz7S2©2406q@035‡Wl0ep424w6I124A4F „ uG8FXI1b42E 7@9X9XYTVGd7SW49P0Sx{}b4FPTwFP24w69Pp4A406Tu7S9P9y0Dijg49XFPw6FXGdf˜0DG8b4I—A4Tu7@24Asgc5e03AcFPw6GeI15v„¹w3I15V5e06w6G8I@5wfh06Geb4IjAcT f˜7@A40vp4gxISRÇ0Dijg49PFXw3FHGUTe06¥€p40324w603T6}jp4wdb’fh06Geb4IjAcT}7@5e0”5e06Rt065e5V03AxG8I4™—G8b4065e0DRtI15e0S™E7@ú T dšœ² › …@˜f•½˜—šœ£ › – eÁe ži“D•¹! I – f9ePùf˜0DG8b4I—A4T6 'yq@035vYpw3I@2€q@0324FX032“G u 7±Y˜G8Iegc5e03TV032“G}G8bc0”TVG87@W4FX9PFXG`Y”gc5eI1gl035vG8FP06TI@RÈ7dG8FPfh0D„¹fq7S5ewdb4FX24(wfh06Geb4I—A FPTiGeIsg49XI@G”G8b40…9PI—w3p4T”I@R”G8b40…w3I1fhg49X0Digý  RtI15 u b4FPwdb Æ   Æ ¨ °1™UTVp4wdbÃGebE7SGhG8b40…5e06Tep49HG8FP2c( w3I@2€GeI1p45e(@Ij06TuG8b45VI1p4(1b§G8b40pglI1FX2€Giý Âk¨ zj©÷m065e0 Æ   Æ 5V06Rt035VTeG8IGeb40pf˜7œijFPfepcfè7@W4TeI@9PpcGe0 qS7@9Ppc0I@Rl7S2€ú Y  ’™Sg45eFX24w3FXgE7@9€I15|TVg4p45VFPI1p4T6™SGebE7SG’FXT’7~5eI—I@GÄGeI”G8bc0wdbE7@5e7@w6Ge035VFPTVGeFPwgtI19HY—24I1fhFº7@9@RtI@5 jƒgQ8 ƒ‚‘Ò6‚‰ˆ 6 µ –bRe¶l”'k'”…–Itё“—y˜y–W—Iæh’s ”8˜¿–‘ˆh ‘‰—™i jlk¥0çéèIÔÕê kTR%˜h™‚ ë d BU$8&A76ìI#r@6&RF”#r#8ð9ïP&AXPC3%ïP7©¨Eÿëyì4ë6&ËCB"¨1V4¨&AXP7P3RB"@ƒ&RX1¨˜ì8&AC¨3R7 4 F©¨1V4¨)¨3R76ì4ëƒ&%$¥3(7Pÿ‰&A79²¨7PF”#8H°#87PF”#'79‚3A0 ¥ &R$¥B"r $¥3(@qYQ 쩨¤§7P¦©&A#% C¨3RH©7PB"34 §3A3R0W020êH©3RB"XX13”3F”øQïP§$…‰0êB"§3R3A@ 0&AC¦XPX!#‡ ¨E‰0  393RC¦"°8&Y Q"¦©#q$¥34#'56$'¨1#879"Q3R0•)ïP$‰¦l&6HI3RXë47©3R@ƒ¨&AXv&RBC#%Cï©@6Q3R0VÊ&RBC¨3Rï° .

°zŽ!z ï I( λ h) Unstable ΒÏmДÑvÒÄÓBÔxÕ±ÖÃ×ÇÒ4ДØÙrÚÛÙ3ҐÜÞÝ}߀ÚÛÙrà”ÓyÐvÔá×ÇÜB×ÇÒÄÓrâ#× î I (λ h) í I (λ h) Stable 1 Stable Stable Unstable 1 ï R(λ h) ð θ=0 î R(λ h) ñò θ = 1/2 Unstable í R( λ h) θ=1 a) Euler Explicit b) Trapezoid Implicit c) Euler Implicit ó©ôöõ‡÷¢ø(ùˆú´ûýübþhÿ¡ £¢¥¤bô§¦7ô¨ © £ƒ÷bø!"‡ø# £$¢ù&%(')"ù0 $132 û ¢”õ‡ô54‡ù076980ûA@B #"¥¦§¦§(C!D{ø) ÿ´ù0E0 (ô§ ú|ûGF& $H¢I PbùQ(ô§2¢ùQ¥R $bô§SET £ƒ÷bø £$¢ùQ%÷1)"ùXø(ô¨EU¢¥¦ )"ù0 £$V3› 2 ô¨WX £¢¥¤17 ¦ ùC!b $ ô§7 ¦ ùY $bùP¥ $bùXø0`eô¨ ô¨”÷1VX a¢I¤1¦öùQûcb ùšø(ùd{ùXøe Çô¨ T`f £$bù`ø(ùd"‡ø(ùI`S¢Ig¢ hTiqpri tsvuawyx €wy¡‚ û ƒ„©¡b … ô¨EU¢¥¦†‡ a¢¥b ¤ ô¨ö ¦ ô5 ©WET¥ˆ ‡  ÷bøX„"‡øD¤‰¥ £" $ ùd3…17 ¦ ô§` E ô¨ „¢¥1" 2 ô§)‘…17 ¦ ô§X E ô5 ’" ) ù0 $132VSQ ¢ ø(ùÛô§¦¨ö ¦ ÷1‡e   ø£¢ye   ùTr 2 ô¨ ó©ô7õbû£ú´ûýü1`VC!¢ $ ô§EaË $ ô¨“b 2 ùXø ô¨ƒ 4 ùT27{ø)” £b $ ù•b  ù– ' øX3¥ —%('" ) ù0 $132 õ‡ô5‡ 4 ù0¨  ô¨€  ÿ´ùTEde   ô§¥™´ ˜ ûe´ d ûýübû “D•—– f• E• f&g‰h9i(g‰jlkymgnBg†kPopRqsrttvutqixwyoi{z|gR}sm ó©ô7õbû ú´ûýü¢~X$1(C!Q £$bùX £¢¥¤bô¨¦öô¨ ©PET £ƒ÷bø€"‡ø! £$¢ù”ùd3…1¦7ô§E`ô¨ €§÷1¦öù`ø€)"ù0 $132 û“@‚ £$bù"¦§¦¨(CÛô§bõ  CS‘C¢U©3B¨ ô  ¨ ô # ©¡…b§ ô ET¢¥¦R¥f¢¥¦§¦ƒ‡ a¢¥¤¢§ ô ¦ö5 ô  ©‘ET £ƒ÷b ø !"‡ød ù 3…1¦7§ ô E`¨ ô  “)"0 ù  $132VXþ „ û!ƒ“$bù•E0 £‡÷bøÛù01ET¦¨(ù0Q¢W…Hbô5 eù•…‰‡øX (ô§v¥l £$¢ù¦7ù0q ‡'$H¢¥¦5|ET)‘…1¦7ùd†6‰83'…1¦§¢¥bùQû ‡ û!ƒ“$bùˆø ùXõ‡ô¨7¥f‡ a¢¥¤¢ô§¦öô5 ©sô¨ iqx divˆ¡‰  $bù•¤9ƒ÷112H¢ƒø‡©`†¢¥V27 $bùXø ù0"‡ø ù¥`tô¨ 5ô§“ET¥12bô¨ (ô§1¢¥¦nû Š€(Chù04ƒùXøT`ƒ $bô§€)„ùd £$1¡2~ô¨1ET¦7÷12bùT€1P…H¢QøX ‹IA £$bùô§)~¢ƒõ‡ô§1¢ƒøX©7¢{¢ô¨gŒ{ùd3EXù0…V "‡ø€ $bùeƒø(ôöõƒô§†Ž ¢¥12cXš  ô¨m   ô¨•  ÷11‡ a¢¥¤17 ¦ ùW"‡øQ b $ ùg)~¡b 2 ù0¦fb ¤ ô¨ET‡ 4 ù0E0(   ô§cb … øX¤1ö ¦ ù0€ ) û•€¦¨ £$V‡  ÷bõ$ ù0‡ 4 ù`ø£¢¥§ ¦ ù–3…1ö ¦ ô¨X E ô¨  )"ù0 £$V321“X$Hƒ ¢ ø ù• b $ ô§b 2 ù0…H` E ôöù01E0©‘{ Œ ùƒûýõbû¨`V‹’ ‡ `1“€” ‡ Ž–`1(  ùd‡ 4 ùXø¢¥¦R¥ b $ ùXøX—2V‘V¥ &9 Œ ùQû’õbû5`17 ¦ ùU¢¥…V{øX‡  õ1` • ¢¥–T2HQ ¢ õ1`ƒ“—”•FV`ƒ“—m ” üˆŽd`‰(  ù`ùró©ôöõ`  û0ú|ûed¢I1° 2 ú´û—¢ ˜ û˜—¥e   ô¨X E ùô¨‚…Hƒ ¢ ø‡e   ô¨X E ÷1¦§ƒ ¢ ø— $H¢I € £b $ ùe b $ ôöøX23'!¢I12 "‡÷bø‡ £$3'B‡  ø¢ 2 ùXø“Û “ ÷1b  õ‡ù–'€m ” ÷V X a¢W" ) ù0 £$V3210£ ` ó©ô7õbûbú|ûG˜V¢ ` ô§1E0ö ¦ ÷1b 2 ù‹¢W…9‡  ø‡e   ô¨P¥| £¢ $ ùmô§)~ƒ ¢ õ‡ô§1ƒ ¢ øX©‘¢y´  ô¨  ÷V “ P™ „š—›¥œ ¢¥12™™ ‡ˆ ‡yœ £ ‡ ` ø(ù0… ùTE0(   ô¨ƒ 4 ùT¦¨ƒ © û f&g‰h9i(g‰jlkymgnBg†kvjlhsug‰hl}stqi{t"g‰hlžHrvrqŸ m0i(ž1¡sr"o¢tvwyqfrvtvut"iewyoi{z|gR}sm ó©ô7õbûkú´ûýüEg$V{C! £$bùx‡ a¢¥¤bô¨¦öô5 ©¢ET¥ˆ ‡÷bø"‡ø‹ $bù„ô¨)~…V¦öô§E`ô¨ Q§÷1¦öù`ø)"ù0 $1¡20ûx@B ô¨‹ ©¡…bô§ET¢¥¦A¥ )~¢¥©PX £¢¥¤b¨ ô ¦ö¨ ô  ©~ET¥ˆ ‡÷bX ø !"ƒø1 ÷ 1ET¥12b¨ ô  (§ ô 1¢¥¦§¦5©xX a¢I¤1¦öùm¨ ô )~…1¦7§ ô E`¨ ô  „)„d ù  £$1¡21`S û ˜—¥ e¨ ô EX‹ ù  $H¢I “ £$¢ù )"0 ù  £$V32°§ ô ‡ a¢¥¤V¦öW ù "‡‹ ø  $bùrT ù  eô7ø(ù}£ ø ¢Ibõ‡x ù ¥SET)‘…1¦ö– ù £6‰8c $H¢I v¢¥¦¨¦ wy¡¤"¥divˆ¡‰  £$bx ù ¤‰‡1 ÷ 12H¢QX ø ©‡û c“ .

0 RK1 R(λ h) -3.0 ó©ôöõƒ÷bø(ù”ú´û—˜¢þØÿ¡ £¢¥¤bô¨¦öô¨ © ‡÷bø!"‡ø!“Û÷1bõ‡ù–'€”m÷V X a¢g)„ùd £$1¡21`û .0 Stable Regions RK4 RK3 2.0 RK2 1.0 -1.0 -2.¦(§©¨§«ªP¬f­™®„¯!°d±f²‹³ ´Rµ1²Q¶°€³R°Tµ!·¸±“¹“ªW±’®DºQµf´†°€ª»µs¼€®¸±“¹!­‚ºD³l®’½ 6‰¾ ºD³R²‹ª€® „ ü „ óqôöõ‡÷bø ùˆú´ûGd´þØÿ¡ £¢¥¤bô§¦7ô¨ ©P £‡÷¢ø!"‡ø“)"ùmùd3…1¦7ô§E`ô¨ “)"ù0 $132VXû I( λh) 3.

0 2.0 Stable Outside 1.0 4.0 1.0 Unstable R(λ h) Unstable 1.0 1.0 Stable Outside 2.0 2.0 R( λh) a) 2nd Order Backward Implicit b) Adams Type ó©ôöõ‡÷¢ø(ùˆú´û˜ú´þhÿ¡ £¢¥¤bô§¦7ô¨ © £ƒ÷bø!"‡ø# £$¢ù ‡ ÷1VET12¢ô¨ eô¨H¢¥¦¨¦¨©‘X £¢¥¤1¦7ùmô§)‘…1¦öô¨EXô¨ S)"ù0 £$V321`û ƒ“$bô¨W)„ùT¢¥1W $H¢I W $bùv)„ùd £$1¡2›ô¨W%÷1)„ù`ø(ô¨EU¢¥¦¨¦¨©£X £¢¥¤1¦öù ‰dÃ¥‰dxÅÄfÆH‰dx«¤"ÆH‰ÇSȋÉxÊ ¥P¤qÆVË¥¤iq¤i§¥ hd‰diqx¡Ì†¥(‰ˆ™¤w7iqxH¤‰ÍÌ¥‚Ë¥¤‰~Ë¥‚£‰xiqxVÆH‰d‚£‰dxH¤vprs†¡xV¥d¤ÍËhTp§‰ û}ÿ3)"ùg¥ $bùXømô¨)~…1¦7ô§E`ô¨ B÷11ET¥12bô¨ (ô§1¢¥¦§¦5© X £¢¥¤17 ¦ ù‘" ) ù0 $1¡21Û C ô¨ £$c £¢ $ ù‘¢¥" ) ù‘¢ … ø… ùXø‡ ©£ƒ ¢ ø(ùx$V{C! ô§ ó©ôöõ¢ûgú´û˜ú´û‘@Y¢¥¦¨¦’¥„ £b $ ù0(  ùET¢¥(  ù0  £¢ $ ù”ô¨)Q ¢ õ‡ô§HQ ¢ øX©~¢y´  ô§  ô¨“…Hƒ ¢ øX —¥l £¢ $ ù ¥d¤ÍËhTp§‰ ø(ùXõƒô§0  û ˜€¥ f¢¥¦¨‡ ¦ ÷11E01b 2 ô5e   ô§¥H¢¥¦§¦5©Q‡ a¢¥¤Vö ¦ ù„„ ) ùd £$1¡21lƒ ¢ ø(ù„X £¢¥¤17 ¦ ùØô¨&X" ) ùhø ùXõ‡ô¨1RC!b $ ùXø ùS b $ ù#΍ύ“ÐÑ  £¢ $ ù0©cƒ ¢ ø ù}ô§e   ù`õ‡ø£¢ye   ô§¢  õvƒ ¢ ø(ù}ô¨1b $ ù`ø(ùT ¦¨~ © ÷V1X £¢¥¤1ö ¦ ùQûWƒ“b $ ùxET¦§¢¥ ô§} E ù–1¢I)~…17 ¦ ùW¥¢v„ ) ùd £$1¡2¢ £$H¢y  ô§X a¢I¤1ö ¦ ù wyxHprs C!b $ ùTÒ £¢ $ ùÇõ‡ùTb  ù`ø£¢I(   ô§b  õ«ÎÏ“ÐÑ7ƒ ¢ ø(ù¢ b $ ùT)‘(  ù0¦¨‡ 4 ù0š  ô§1¢ $ ùXø(ù0 £¦¨©ÀX £¢¥¤1ö ¦ ùËô§P £¢ $ ù  eø¢¥… ùT–Tƒ  ô§2H¢¥¦!" ) ù0 $132Ó5 ` ônûýùƒû5`# £b $ ù™… ùT` E ôÔ¢¥¦—EU¢¥ ù¢¥‹ b $ ù¢%('" ) ù0 $132«"‡øxC!b $ ô§Ea$Õ%«Ö × ûك“b $ ù Ø X £¢¥b ¤ ô¨ö ¦ ô¨ ©e¤9ƒ  ÷112Hƒ ¢ ø‡©g"‡øA £b $ ô¨SET¢¥(  ùvô§DX$1(C!Í  ô§„  ó©ôöõ¢û|ú´û’ü¥0 ¤ û’ƒ“b $ ù—¤9ƒ  ÷112Hƒ ¢ ø‡r © ô§A £¢ $ ù“ô§)~ƒ ¢ õ‡ô¨Hƒ ¢ ø‡© ¢y´  ô§&¢¥12‘ £b $ ù~|  ÷V„ ) ù`ø(ô§ET¢¥¦D„ ) ùd £$1¡y 2 ô¨&‡ a¢¥¤Vö ¦ ùx"‡øW6‰8Y¦¨| © ô§¢  õ™y‡  ø ™ b $ ù7 ¦ ù0q &¥# £¢ $ ô§&¢y´  ô§X  û ƒ#CS~¥ b $ ùXø!" ) ù0 $132V! $H¢I !$H¢Uƒ 4 ù b $ ô§“b … øX… ùXøX ©vƒ ¢ ø(ù £¢ $ ù• CSI'BXe   ù0…™(   ø£¢¥… ùT–0‡  ô§2H¢I¦ƒ„ ) ùd £$1¡2 ¢¥12™¢g)„ùd £$1¡272b÷bù £‘âgùXù0 ÚHÛ(Ü × Ö ÚHÛIÝ ×ƒÞ 8Dß ÚHÛ( à Ü × Þ ÚHÛI à Ý ×€á ‡ 8 ß Ú àÛUÜ × Þ Ú àÛ Þ Ú àÛyÝ ×€á F ˜€I eô§E`ù‹ £$H¢y €¤‰¥ $†Is $bùT ù&)"ù0 $132V!¢ƒø(ù•¥l £$bù&ã€ô¨¦§bù‹ ©¡… ƒ ù û ÚHÛ(Ü × Ö ÚHÛIÝ ×ƒÞ f&g‰h9i(g‰jlkymgnBg†kvug‰hl}stqi{t"g9h|žHrvr"Ÿäm0i(ž1¡srvo‚ttw‘qfrvttutqiewyoi{zlgƒ}sm å ÷1‡ gE ƒ ¤ ù0EUƒ ¢ ÷V(  ù7¢¢" ) ù0 £$V3› 2 ô¨”  ô¨)~…17 ¦ ô§` E ô¨ •21|  ùTWV¥ W)"ùU¢¥Y $H¢I ”ô¨ ô¨”÷1VET12¢ô¨ eô¨H¢¥¦¨¦¨©‡ a¢¥¤V¦öùƒû ƒ#CS¨ô¨¦§¦7÷1X (ø£¢I (ô§1!¥A $bô§‹¢ƒø(ùWX$1(C!°ô§Çó©ô7õbû ú´ûGæ´û΍¢ùW¥A £$¢ùT(ù}ô¨Q £$¢ùWQ2H¢I)~€'‡ã™‡÷V¦¨ £¥ F øX23'‡ ƒ  øXb 2 ùXøQ„ ) ùd £$1¡2Ōv1b  ûeæV`ƒƒs¢¥¤17 ¦ ù"ú´û „  Ž ûe€1¥ $bùXømô¨Q £$bù}üI £$3'B‡ø2bù`øã€ô§¦¨bùW)"ù0 $132Çõƒô¨4‡ù0 ¤©P £¢ $ ù…‰‡  ô¨ˆ !¥E … ù`ø£¢I ‡  ø ÚHÛ(Ü × Ö ÚHÛIÝ ×ƒÞ F „ 8 ß Ú àU Û Ü × Þ ü Ú àÛ Þ Ú ày Û Ý  ×á ¢¥12‚$1(C!°  ô§¢ƒs¢¥¤17 ¦ ù}ú´û „ ¢¥€1b  û „ ¢ F ûQ@Bm   ô¨Q‡ a¢¥¤17 ¦ ù wyxHprs "‡ø6‚Öç™ œvè C!b $ ùTcé7ê è ê  ¢ F û @B £!‡ a¢¥¤bô¨¦öô5 ©~¤‰‡÷1V2H¢ƒøX©sô§#4ƒùXøX©P(ô§)"ô§¦§¢ƒøD £g £$H¢y #"‡ø# £$¢ù¦öùU¢I…V{ø‡õg)"ù0 $1¡2£Œv(ù`ù”óqôöõbû¢ú´ûGdI¤†Žû .0 3.„ ü ‡ I(λh) ±’¼€²Qº‹µs®D¯¿¦(§À´|µ1²Q¶°€³R°Tµ!·¸¹“Á ³R°€ªQ®A²‹¯Â´|·D´|µs®„­¢´ I(λh) 2.

¦(§"¦(§ÀÁ„¹x¬s¯!°€®D¯¸´|µV²‹¶°€³R°Tµ!·Å²‹ª—²‹³l·A´3°‡´ I(λh) 3.0 2.0 -1.0 Stable 1.0 -2.0 R(λ h) R(λ h) a) AM3 b) Milne 4th Order ó©ôöõƒ÷bø(ù”ú´û—æ¢þØÿ¡ £¢¥¤bô¨¦öô¨ © ‡÷bø!"‡ø# $bù ‡ E012bô5 eô¨H¢¥¦¨¦¨©‘X £¢¥¤1¦öùmô¨)~…V¦öô§E`ô¨ S)„ùd £$1¡21`û ëPìdë ívî‘ïð„ñ0òðôóPõ’öe÷ñUøUñ0õÓù úüû¢öeø0ù†ý’ñUý ’S©v¢ƒøs $bù#)‘X f…9¥…b÷1¦Ô¢Qøs"‡øX)þI†X a¢I¤bô§¦7ô¨ ©¢¥H¢I¦¨©¡(ô¨s"‡øs|÷V)„ù`ø(ô§ET¢¥¦VXEa$bùT)"ùT§ô¨s $bùóH‡÷bø ôöùXø ‡ø|4WB ˜ ùX÷V)¢¥VW¢¥…1¢ … ø¢¥Ea$ û’ƒ“b $ ô§s¢¥H¢I¦¨©¡(  ô¨q  ô¨y  ÷V(  ÷H¢¥¦¨¦¨©•ETƒ ¢ ø(ø ôöùT2&‡  ÷V se…‰‡  ô¨ˆ s… ùXø¢I £‡  øXf¢I12 ¥ H‰¡ Ë  yiqx¡Ì ô¨ e¢¥…1…17 ô¨ &21|  ùT&1¥ &b 2 ùT… ùT12ä¥Y¢¥›  ô¨e   ùXøX„ ) ù0b 2 ôÔ¢ye   ùxX £ƒ ¢ õ‡ùI—΍ύ!ÐÑX  ûcÿ ¤v‚diÍud¤vprs™ ¦ ôöùT ‚¨ 1¦5©y ‘b 2 ô£¢ ù`ø(ù01X E ù†¢¥…1¢ … øU´  ô§)~¢Ie   ô¨1W¥¥¤Dύ!ÐÑg £$H¢y ~¢ … ø¡b 2 ÷1X E ù™§ Î ¦x“З ÄfÆ¡iÍuaÆÅË¥‚£‰7priqxӉË¥© ÆV˥剑xÓwv ¥ VË¡ua‰vwy‚‘¤t i ‰‘ÃyË¥‚–syiqx¡Ì£uaw( ‰ Õudi͉dxH¤q ¥ ¨‹Ë¥x‰ˆ†ÆV˥åm ‰ H‰d‚–iÍwTˆ¥iÍuhdwy3x‰ˆË¥‚–s£uawyx‰ˆ¥iq¤viÍwyxV¥ û @Yb … ø¢¥E0(   ô§ET¢¥¦D¢¥…1…17 ¦ ô§ET¢Ie   ô¨°  ô¨”   ô§¥q eù0y  ÷1 ùT2ä¢IW€ ¢ õƒ÷bô§b 2 ùx"‡øˆùT‡e   ô¨)¢I(   ô§b  õv b $ ù~CS‡  øX b $ ô§¢  ùT&¥ ¢x„ ) ùd £$1¡2†"ƒø€)‘‡  ø(ùmõƒùTb  ù`ø£¢¥¦lb … ø¥¤1ö ¦ ù0)~`  û@B —(  ù`øX‡ 4 ù0¢¥€¢xv¢Qôöø¦5Í © ø ùTö ¦ ô§¢¥¤17 ¦ ù xӉaua‰a¥a¥TË¥‚ds ‡ a¢¥¢ ¤ ô§ö ¦ ô5 ©  Òudi͉dxH¤ ¥b ET¥1b 2 ô¨(   ô§Ó`Vb ¤ ÷Vv   ô5Û   ô§“¤©v1‘„ ) ùT¢¥1!¢ ¥d  ùƒû  ! #" $&%¥')(10325476)8@9BAC6D'FEHG¥9B' Î bù™ £¢PIƒù02H¢y a¢£{øX) ¢RQ€ ©3…¢ô§EU¢I¦TS£…9‡ô¨ „ô¨« $bùVU1{C¿…bùT¦§2À¢¥12 ÷1(ù0 £$bô¨¢I~E01X £¢¥  ‹  £$¢ø‡÷bõ¥$1‡÷V ~ eô§)"ù™¢¥12Ò…1¢¥EXù¢IETETƒø2bô¨bõy £‘ £$bù¢I(÷V)~…V (ô§VÍõ‡ô¨4ƒùTÒ¢I¤9(4‡ùQû¸ƒ“$bùTÕ¢ù ô§)‘…‰(  ù0W¢¢…1¢Ie   ô§¢¥¦S$1ƒ ¢ ø)‘b  ô¨E¢¥x¢I›  ô§b  ô5e   ô§¢¥¦’4I¢¥ö ¦ ÷bùY £b $ ù„ ) ù0$ ¢¥12 ¢IW  IˆW £b $ ùYX|÷¢ùTX(   ô§ þ bžô§¦¨k ¦ ô¨ £€¢¥)‘…1ö ¦ ô5e   ÷1¢ 2 ùõ‡øX(C‡  øQb 2 ù0EU¢U~ © ô¨™(   ô§" ) ùa`䃓b $ ùg¢¥1‡I C ù`ømô§—b 2 ù0(   ùXøX„ ) ô¨b  ùT2¢¤©™…H1b 2 ô¨b  õ £¢ $ ù ET¥1b 2 ô¨(   ô§V  ÷11b 2 ù`ø!C!¢ $ ô§Ea$ Ú ŒbFcWdXŽAÖfehghiqpretsvuw Œnú´û ‡‡ Ž x ‰d‚£‰dxÓua‰ ùtX%÷H¢I (ô§Ó`DC!$¢ùXø(ùVy»ô§ø ùU¢¥¦#¢¥12€yB¦bŦöô7ùTô§Y £$¢ùšø£¢¥bõƒù ô§W¢¢¦7÷V eô¨y £¢ $bù ˆ¥i ‹ égê‚yƒ¦„‚ b ꆅ§ûhÿ´§ ô 1E`ù¥`V"ƒø! $bùˆõ‡ùT¢ùXø£¢I¦ƒ eùXøX)™` ÛUܐ‰’‘ ܐ‰—– i ‘ r ‰ – i pre svu ”™– w p Ûh‘ p e v Ö e g — s u ‡ w ܕ”˜– w ‘ ‚ ڈ“ ‡ ܕ” Öfe g ‡ i ڈ“ ‡ d#eFfrgh’irj#kmlˆn©opgqsrhtujvlDtufrw§h’i¡tuxmtuxyh’gHz7g{f¡x’tu|aj#kyt}h~nx~nf„t}f¡th€tun‚rnuƒrjC„rk€g…¡}j#†‡gfˆnf„t}fa‰Šfrth€j‹x’„¡nzvj |ag{†  n t}fŒ .0 Stable only on Imag Axis 2.0 -3.0 1.0 „ üF I(λh) 3.

„ ü‡ü ±’¼€²Qº‹µs®D¯¿¦(§À´|µ1²Q¶°€³R°Tµ!·¸¹“Á ³R°€ªQ®A²‹¯Â´|·D´|µs®„­¢´  £$¢ùX%÷H¢¥ eô5 © ÚD“ ‡ Ûa‘ ô¨SE0)‘)~x £ù04ƒùXøX©‘ eù`ø) ¢¥V2PEU¢¥~¤Eù—v¢¥E0 ‡ø(ù02‡÷3 †û’@ $bùvø(ùT)~¢ƒô¨bô§bõ ùd3b … ø ùTX(  ô§¥1T`HChù…HV2† $bù eùXøX)Že g – i C!$bô§Ea$7CIùˆø ùT…bø ùT(ù0 ‹¤©’`1 $|÷VXþ V‘Rehg – i ƒ“$bù0ƒ`†(ô§VEXùpe ghi Ö Û ße g – i Ö  Û `bô5 Ûô§“ET¦7ùU¢ƒø# $H¢I  ‚ á ó†ƒø“|÷V)„ù`ø(ô§ET¢¥¦ƒ‡ a¢¥¤bô¨¦öô5 ©@’ “’ˆê „ Œnú´û ‡ FŽ ¢¥12Å £$bù†…bø¥¤1¦öù0) ô¨g £YX¦¨4ƒù†"‡øx £$¢ù”’ÐÑ…bøX32¢÷1EXù02 ¤© ¢¥© õ‡ô54‡ùTÅ)„ùd £$1¡2 ¢I12ƒ`!¢¥~¢ bù0X E ùTX£ƒ ¢ ø‡©£ET1¢ 2 ô¨ eô¨¢"‡ø‡ a¢¥¤bô¨¦öô5 ©`ƒ)~¢PIƒùx ÷bø(ùg £$1¢I U`kô§ $bùgC‡øXX …‰ ô§¤1¦7ùgET)W¤¢ô§H¢I (ô§ ¥s…Hƒ ¢ ø¢¥„ ) ùde   ù`øT`HE01b 2 ô5e   ô¨¨  ú´û ‡ } F ô§“£¢ye   ô§‡£ … ùT• 2 •û  ! !– —˜A“™š'œ›ž‹0y™ Ÿ§¡v'F2 ƒ $bùx…bøX3E`ùT2b÷bø ùxEU¢¥£¤Eù0X ¤ ù}ùd3…1¦Ô¢Qô§bù02£¤©Çù–V¢¥)~…V¦öùT`ûx1 ô§2bù`ø¢¥•¢P…£ø‡ ù–V¢¥)~…V¦öùW £$¢ù “ …Hbô5 eùS2bô¢¢EùXø ùT1E`ù“¢¥…1…bøXU¢ô¨)¢y eô§¥• Q £$¢ù#)~¡2bù0¦¡2¢ô¢¢E÷1(ô¨”ù{X|÷1¢I eô¨„Iˆ1(C!x¢¥’“Ûô§Ea$1¢ƒø21XƒÐÑ ) ù0 £$V32YI!{ƒ " 4 ùXøX¦Ô¢¥…1¢ … ô§b  õ¢‡e   ù0…1X  û£ƒ“b $ ô§&C#¢¥e„ ) ù0ˆ(   ô§¢  ùT» 2 ô§»  ÿ´ùTEde   ô§¥›  übû „ ¢¥1» 2 õ‡ô5‡ 4 ù0 ¢¥g3  Xtû übû „ þ Ú “‡ ‡ ¦„d Û(Ü × ‘ Ö ÛyÝ ‘ Ûh‘ ß Ûa‘ ‡ Ûa‘ Ú “‡ × ¤ Þ £ ¦„b Ø Ú “ ‡ Ü ×™¥ Ú “ ‡ Þ Ú “ ‡ Ý × á ‡ ¦„d ß e svu – w ‡ Þ e Ý svu – w ހ£ ¦„b Ø ¥ á Œnú´û ‡ üˆŽ ÿ´÷1¤VX eô5 e÷V (ô§7¥f3Xtû ú´û ‡¥‡ ô¨ £‘3X ûEú´û ‡ ü}õ‡ô54‡ù0! $bùˆø(ù0¦Ô¢I (ô§ ‚Ö‡ Ý × ‡ø  Ø Þ§¦ ¦b Ø © ü £ ¦„d Œ„ ¥ E0‚yB¦„b9Ž7¨  ‡ ¤ ª7« ¬ ¥ „ Ö é Œnú´û ‡ dŽ ƒ“$%÷1&yXtûqú´û ‡‡ ô§•¢7¦7÷V eô¨¥#yXtûyú´û ‡ ü ô¨~ ô§•¢ ø¡¥ ¥#yXtûyú´û ‡ d´ûxƒ“$bùx CS€øX3¥ ¥ ú´û ‡ dgƒ ¢ ø(ù  Ø Ö ™  Ø „ ×!­ ¥‹® ® Þ {ø¥) C!b $ ô¨Ea$ ô5   ô¨gETö ¦ ùTƒ ¢ øe $H¢I g¢  ù&’ “’gô¨x¢¥¦¨C¢U©¡¯ „ û£byùP…112ƒ`D £$bù`ø(ù0"ƒø(ù¥`A $H¢I x¤©Y £$¢ù óH‡  ÷bø ôöùXø“‡ a¢¥¢ ¤ ô§ö ¦ ô5 ©~(   ùT‡ U`†Û “ ô§Ea$Hƒ ¢ øX21¥ƒÐ——" ) ù0 £$V327¥s(4‡ùXøX¦Ô¢¥…V…bô§bõx‡ eùT…VBô¨v÷V1X £¢¥¤1¦öù"‡ø!¢I¦§¦ ` £ y‚¢¥VV 2 ¦„d¥û ° ±²q³…Pgƒrf¡|rj#|afrjWx’x~t´xmk€j#µtƒrtuk’jW|„t}f¶n˜·ƒ¸Š¹vº²»yh’tu† j‹|ag{†¼ntuf¾½5h’irjz7g{f¡|at}h’tugfˆt´xmgqvh’j#f¿„rk€j#x’jvfth’jW|nxÀ ÁƒÀq v ÃqĶŠƢÇ~È—É Œ .

¦(§´Ê¡§ ±“¹“ªQ´3°‡´|µs®DªW±€· „ üd  v¢I1¥ £$¢ùXøsù–V¢¥)~…V¦öù†ET1 ô§2bù`ø £$¢ù™…Hbô5 eù–'2bô£¢ ù`ø(ùTVEXù¢¢I…1…bøU´ô§)~¢I (ô§ "‡øx £$bù¢)‘¡2bùT¦ Q ¤bô¨ET4‡ù0E0 (ô§¨ùtX%÷H¢I (ô§ Ú “‡ @7 £$bô¨“EU¢¥ ù ™Ö ¦ d „ Û(Ü × ‘ Ö Ûh‘ Ûh‘ Ûh‘ Ú “ ‡ ¥§‡ Ë ¦„b ß Ú “ ‡ Ü ×³¥ Ú “ ‡ Ý × á Œnú´û ‡ ˜Ž „ ¥ ¦„d p œ p( ô§‹yB¦„b Ë ¦„b {ø¥) C!$bô§Ea$ ô¨ ”ô§•ET¦7ùU¢ƒø• £$1¢I Ì’ “’³¯ „ "‡ø¢¥¦§¦A1¥1–Xù`ø ¢¥V2Íy0ûPƒ“$%÷1eChù$H¢U4ƒùP¢¥1I £$bù`ø …Hb5 ô  eù–'2bô£¢ ù`ø(ùTVEXù•¢¥…1…¢øU´ô§)~¢I eô¨P £$H¢y U`1¤©P $bù”ó†ƒ÷bø(ô7ùXø#Ë ‡ a¢¥¤bô¨¦öô5 ©~ (ùTX T`£ô§÷1VX a¢I¤1¦öù‹"‡ø!¢Iˆ© Ea$1‡¨ ô EX• ù ¥l £$bù{ø ùXù•…H¢ƒø¢¥)„ùd eù`øXû  ! ÏÎ ÐÑ'³¡v0˜Ò54vA~ÓÔÒ5AŽÕÍ479B6FG§¡70yÓ Ò†Ö 0˜Ò•9ƒ4v6D'F2 ƒ“$bùv÷112bù`ø¦5©´ô¨bõ¢¥ ÷1)‘…V eô¨rô§‘¢óH‡÷¢ø(ôöù`ø’X £¢¥¤bô¨¦öô¨ ©e¢¥H¢I¦¨©¡(ô¨Øô¨’ £$H¢y A £$bù‹× )~¢I eø ô59`2bù0 (ùXø€' )"ô§b  ù02†C!b $ ù0‚ £¢ $ ù&b 2 ô£¢ ù`ø(ù01EXô¨bõXEa$bùT)"ù”ô¨—…¢÷V “ô¨™ $bù"ƒø)¿¥’yXtû ú´û ‡ `tô§!EXô7øEX÷V¦Ô¢¥ †ûÿ´÷1Ea$ ¤ ùXô§¢  õ~ £¢ $ ùeEU¢I(  ù¥`‰ £b $ ùØe svuw ô¨¢3Xtû0ú´û ‡‡ ø(ùT…¢ø(ùT ùT £•¢I ‰di5Ì¡‰dxHÃ¥‰aud¤wy‚ ¥’ £$¢ùeX©¡‡ eùT)†`ƒ¢I12‚ £$¢ù  CS}  ùdV¢¥)‘…1ö ¦ ù0F  Ù ÷1‡ “b … ø(ù0(  ù0ˆ(   ùT2†‡  ÷V ö ¦ ô§¢  ùQ¢e ô§)‘…1ö ¦ ù—b … øX3` E ùTb 2 ÷bø ùQ"‡ø„…H1b 2 ô¨b  õe £¢ $ ù ‰di5Ì¡‰dxHÃyË¥prH‰a¥ ¥  £¢ $ ù&` E ôöø` E ÷1¦Ô¢Iˆ !)~¢I(   ø(ô§` E ùT#"‡øX„ ) ù02™¤©†¢¥…V…1ö ¦ ô¨EU¢I(   ô§v¥s £b $ ù CS" ) ù0 $132V! ‘ b $ ù)~¡b 2 ùT¦Ób … ø¤¡' ¦ ù0)~` ö  ûeƒ“b $ ùxEa$1ƒ  ô§X E ùgI~ä"‡ø‹ b $ ùgX £¢¥b ¤ ô¨ö ¦ ô¨ ©†…HQ ¢ ø£¢¥" ) ù0(   ùXømô§¢ b $ ùróH‡  ÷bø ôöùXø‹¢¥H¢¥¦5©3 ô§0`Ó £b $ ù`ø(ùd"‡ø(ùI` ô§1¥ —¢¥†¢¥E0X E ô§¢ 2 ùT†   ûA@BÛ   ô¨5  ùdV¢¥Ed £¦¨© £b $ ù¢¥„ ) ùch C ù•$H¢U‡ 4 ù•¤ ùXùT ÷1 ô§b  õrô¨7¢¥¦§¦9¥l‡÷bø#b … ø ù0| 4 ô§ƒ  ÷1 2 ô¨X b E ÷V(  ô¨10`Hb ¤ ÷V €ƒ ¢ ø(ø ô¨ƒ 4 ùT2™¢I “{ø)»¢‘¢ 2 ô¢¢EùXø(ù0 !E … ù`ø… ùTEde   ô5‡ 4 ùƒû @B!CIùÍù–V¢¥„ ) ô¨b  ù~ b $ ù¢ … ø(ùT` E ùTb 2 ô¨b  õÇùdV¢¥)‘…17 ¦ ùT&{øX)  £b $ ù~´ 4 ô7ù0C!…‰‡  ô§ e¥“EXô7ø` E ÷1¦Ô¢¥ e)¢I(   ø(ô¨X E ùT ¢¥12¢ b $ ùe ùT" ) ô5'Bb 2 ô§XX E ø(ùde   ùe¢¥…1b … øX¢¥Ea$ƒ`Ó b $ ùø ùT(  ÷V¦¨ £Qb … ø(ù0(  ùTˆ   ø¢I £b $ ù`øQ¤| 4 ô§ƒ  ÷1‹ET¥1ET7 ¦ ÷1(  ô¨1X  ûQƒ“b $ ù …1¢¥X E ùb 2 ô£¢ ù`ø(ù01X E ô¨b  õô§Pv “ ô¨Ea$Hƒ ¢ ø2VƒÐÑ“" ) ù0 £$V32P¢ … ø¡b 2 ÷1X E ù0# £b $ ù)~¢I(   ø(ô5Y  ړpۓܡŒ „ c ‡ c „ ŽAC!b $ ùXø ù ¥ Úvô¨A¢‹…‰(  ô5e   ô¨ƒ 4 ù#ET¢¥¦ÔQ ¢ ø’ET|  ù{ݑX E ô7ùT†   ûuótøX)þQ…1… ùT1¢ 2 ô5x’Yh C ù“…H12g £$H¢y ’ £b $ ùÛù`ôöõ‡ù0ˆ4I¢¥7 ¦ ÷bùTf¥† £b $ ô¨ )~¢Ie   ø ô5~ƒ ¢ ø ù•ø(ùT¢¥¦Ób  ù`õ¢I(   ô¨‡ 4 ù€%  ÷1)W¤ ùXø`  ûSö ¦ ùUQ ¢ ø¦¨©¥`¡ b $ ùQ(   ô§" ) ùd')~ƒ ¢ øXEab $ ô§b  õ”ô§SE ¤ ù`ô§b  õeEUƒ ¢ ø ø(ôöù02v‡  ÷V ¤ˆ©  £¢ $ ù„7 ¦ ùU¢¥…3{ø‡  õ“" ) ù0 $132¢¥V2ƒ`({ø· ) ó©ôöõbû†ú´ûGd3`( £b $ ô¨R„ ) ùd £$1¡ˆ 2 ô§g  ÷11‡ a¢¥¤Vö ¦ ù’"‡ø Ë¥pqp ùXô7õ‡ùT4I¢¥ö ¦ ÷¢ùTƒÛ C ô¨ $  ù`õ¢Ie b   ô5‡ 4 ù”ø(ùT¢¥¦R…Hƒ ¢ ø‡ £X  û#΋† b $ ù&I £b $ ù`ø€$H¢I12ƒ`H £b $ ù•X…H¢¥X E ù&)¢ye   ø(ô©š  ô¨™3  XtûEú´û ‡ " ˜ ô¨˜  ۓܡŒ „ ca• é c „ Žd` ¥ ¢¥127¢IETETƒ  øb 2 ô¨b  õW £W€…1… ùT1b 2 ô©7’Q`3 £b $ ô¨)~¢Ie   ø ô5~$H¢¥#¢ … ÷bø(ùmô¨)ƒ ¢ õƒô§Hƒ ¢ ø‡r © ùXô7õ‡ùT4I¢¥ö ¦ ÷¢ùTX  ûDŠ—(I C ùd‡ 4 ùXø0` ô§W b $ ô§fEU¢¥ ù! b $ ùùd3…17 ¦ ô§` E ô¨ A§  ÷17 ¦ ùXøA" ) ù0 $13} 2 ô¨’¤ ùXô§¢  õm÷1(  ù02‘"‡ø’ b $ ù#e   ô¨„ ) ù–')~ƒ ¢ øEa$e¢¥12ƒ`¡¢¥E0ET‡  øXb 2 ô§¢  õ  £"  ó©ô7õbû£ú|û’ü1`1 b $ ô§“" ) ù0 $1¡¨ 2 ô¨!¢¥¦¨C¢U©¡v  ÷V1X £¢¥¤1ö ¦ ù"‡ø“ ÷1Ea$‚E01b 2 ô5e   ô¨1X  û ëPìÞ ß”î‘ûý’ñUýsõfòûáà“ù 1 ô§2bù`ø“ £$bù•)‘32bù0¦0ùtX%÷H¢y eô§¥7"ƒø“2bô¢¢E÷1(ô¨†¢¥H¢¥¦5©¡(ô§ â Ú Ö â d â Ø Ú £ â b Ø Œnú´û ‡ úŽ .

„ ü˜ ±’¼€²Qº‹µs®D¯¿¦(§À´|µ1²Q¶°€³R°Tµ!·¸¹“Á ³R°€ªQ®A²‹¯Â´|·D´|µs®„­¢´ ア¥©†©‡ùT¢ƒø‹¤ ù0"ƒø(ùeE0)‘…b÷V eù`ø€¤Eù0EU¢¥)"ùW¢U4I¢ƒô§¦§¢¥¤1¦7ù‘Œ „(›V„ éV`tô§™v¢IE0 –Ž–`ÓâgùdCÛô§•óhûӓÛô§Ea$H¢Qø21X …bøX…9¥(  ùT2¢¢„ ) ùd £$1¡2™"ƒøBô¨ˆ (ùXõ‡ø¢I eô¨bõ"ùtX%÷H¢I (ô§1—¥A $bô§! ©¡…EùQû—b ùe…bø ùT ùT eù02£$¢ô§Q)"ù0 $1¡2 ô§†3  Xtû£übû ‡ ¢I12‚¢IH¢¥¦¨©¡` – ùT¨ 2 ô¨ £“‡ a¢¥¢ ¤ ô§ö ¦ ô5 ©¤©P b $ ù”óH‡  ÷bø(ô7ùXø“" ) ù0 $132 ô¨¨ÿ´ùTEd eô§¥~ú´û˜ú´û @cv “ ô¨Ea$Hƒ ¢ øX21ÓЗ—e   ô§" ) ù¥`† £¢ $ ùeET¥1X E ù0…V •¥A%÷1" ) ùXø ô§EU¢Iq ¦ ô¨1X £¢¥b ¤ ô§7 ¦ ô¨ ©~C#¢¥V¥   Iˆ1(C!0  û•Š—(C#' ù0ƒ 4 ùXø0` £b $ ùSET1` E ùT…Vu   ô§³  X|÷¢ô¨e   ùSET7 ¦ ùUƒ ¢ ø| £¡2H¢U©&¢¥12eh C ù#1(1 C Iˆ1(C ô§)‘„ ) ù0b 2 ôÔ¢I(   ùT¦5©Q £$1¢I f¢ $ ô§’¢I…1b … ø¢¥Ea$ CS‡  ÷1¦§2~E ¤ ùv÷11‡ a¢¥¤Vö ¦ ùƒûA€#¢(  ùT" ) ô5'Bb 2 ô§XX E ø ù0e   ù!" ) ù0 $13„ 2 ô¨ „EU¢I¤ ùBù–V¢ … ø(ùTX(  ù0¨ 2 ô¨)~¢Ie   ø ô5W1¥ a¢ye   ô§¥ ¢¥“ £¢ $ ù‡©¡Xe   ù0)¿¥’Εύ“ÐÑX  þáã3ä ã ä ä Ú Ö d £ ÛvŒ „ c ‡ c „ Ž Ú Þ ¥ ¦„b Ø Œ ®å Ž Œnú´û ‡ æŽ CÛô¨ $Å $bù™¦7ùU¢¥…3{ø‡õc)"ù0 £$V32 ÷1(ù02À"‡øg £$bù† eô¨)„ùv)~¢ƒøEa$ ûÕΕ÷bø¢IH¢¥¦¨©¡ ô§rô§« £$bô¨$1¢¥…V eù`ø ø(ùd‡ 4 ùT¢¥ö ¦ ù02y £$H¢y & b $ ô§ˆ  ô§%  ÷1" ) ùXø(ô¨EU¢¥¦¨¦¨©Ë÷11X £¢¥¤1¦7ù~ ô§1E`ù~ £$¢ùP6V'øX3¥ •¥!’Œ „ ` ' ‡ ` „ Ž¢ƒø(ù~¢¥¦§¦§ø ùU¢¥¦ ¢¥12™b  ùXõ¢Ie   ô5‡ 4 ù¢¥12† £b $ ùb … ÷¢ø(ô§ƒ  ÷1‹  ‰ ' ø¡¥Û   ô¨7 b $ ù¦öùU¢I…V{ø‡õ‘)"ù0 £$V32 ô§Û÷1VX a¢I¤1¦öù•"‡ø—¢¥¦§¦Ó(÷1Ea$ EU¢I(  ùT0`†(  ùXùó©ô7õbû£ú´ûGdI0 ¤ û ƒ“b $ ù†„ ) ùd £$1¡2äC#¢¥r  ÷1 ùT2Ťˆ©«Û “ ô§Ea$1ƒ ¢ ø21Xä"‡øWh C ùU¢I b $ ùXø‘b … ø(ù0b 2 ô§Ede   ô¨ƒ`„¢¥12  £b $ ô¨Wv¢¥E0 ‘ET¢¥ 1(C ¤ ù—¢X‡  ÷bøXX E ù—¥ÓX" ) ù“ö ¦ ù0| 4 ô5 ‡ © û’@~¢¥¦§¦Vb … ø¤1¢¥b ¤ ô§7 ¦ ô¨ ©¥`¥$1(h C ù0‡ 4 ù`øT`V b $ ù!$H¢¥12‘EU¢I¦§X E ÷V¦Ô¢I(   ô§1“Œ" £¢ $ ù 1¦5©¢¢I…1b … ø¢¥Ea$£¢(4Iƒ ¢ ô¨¦Ô¢¥¤Vö ¦ ù&¢I ‹ £¢ $ ùx(   ô§" ) ù(Ž“I C ù`ø(ùx1¥ ETƒ ¢ ø(ø ôöù02¢v¢ƒø•ùTV‡  ÷bõ$ £ ù–V¢ $ ô§b ¤ ô5 X(   ø£¢¥b  õƒù …1b $ ù01" ) ùTH´ ¢ ûsb ù“ET‡  ÷1¦¨2ƒ`¥I†ET‡  ÷bøX(  ùI| ` ÷V(  ù b $ ù ‡ 12¡'‡  øXb 2 ùXøAÛ “ ÷1b  õƒùd'€m ” ÷V X a¢Q" ) ù0 $1¡2e £•  ô¨e   ùXõƒø£¢I(  ù  Xtû§ú´û ‡ æ™ ô§1` 3 E ùsô5”   ô¨&‡ a¢¥¤17 ¦ ù~"‡ø”ø(ùT¢¥¦„b  ùXõ¢ye   ô¨ƒ 4 ùv6RЗ`  û™@B”   ô§0`f$1(h C ù0‡ 4 ù`øT`DET¥1b 2 ô¨(   ô§1¢¥¦§¦5©¢X £¢¥¤17 ¦ù ¢¥12™"‡ø— b $ ô§—EU¢I(  ù&h C ùgƒ ¢ ø ùø¢I £¢ $ ùXøQ ù0ƒ 4 ùXø(ù0¦¨©™ö ¦ ô¨„ ) ô5e   ùTš 2 ô¨™(   ô§" ) ù‡e   ù0…(  ô§` – ù&¤©† £b $ ùø(ù{X|÷¢ôöø(ù0„ ) ù0  Ø ç Œ‡  ¦„d#æ ê ¦„b { Ž û £ ƒ“b $ ùXø ùƒ ¢ ø ùQ)~¢¥©‘C¢U©¡ £&)¢¥¢  ô§b … ÷1¦§¢Ie   ù“ b $ ùQ%  ÷1" ) ùXø ô§EU¢I¦‰‡ a¢¥¢ ¤ ô§ö ¦ ô5 ©g¥l¢¥7 ¦ õ‡  ø ô¨ £$V)~`  ûf΋b  ù€¥  £¢ $ ùT) ô§S   ô§(   ø¡b 2 ÷1X E ù€„ ) ô©´  ùT2(   ô§" ) ùQ¢¥V2v…H¢¥` E ù‹b 2 ô£¢ ù`ø(ùTVX E ô§¢  õ1`V¢&…9X(  ô¨b ¤ ô§7 ¦ ô¨ ©gh C ù‹$H¢U‡ 4 ù1I ‡ © ùd  ET¥1(  ô¨b 2 ùXø ùT0 2 ûØóH‡  øùdV¢¥)‘…17 ¦ ù¥`Vh C ùô¨e   ø¡b 2 ÷1` E ùT27 £b $ ù• Ï ÷£óH‡  ø‡ X ' ótø¢¥5  IQùT¦Ó" ) ù0 $1¡¨ 2 ô¨™$H¢¥…V(   ùXøÛübþ Û(Ü ‘ ÛIÝ ‘ ڈ“ ‡ × Ö ˆ Ú “‡ × Þ ÛUÜ ‘ ÛIÝ ‘ ‡ ¦„d Ú ˆ ‡ × Þ ˆ Ú ‡ × Ûh‘ Ûa‘ “ “ £ ‡“êë Ø èé ڈ“ ‡ Ý × ¥ ¦b Í ‡ íì Þ Úˆ“ ‡ Ü ×—îï ô§7C!$bô¨Ea$™ £$¢ù&E`ùT eø¢¥¦l (ùXøX) ô§7 £$bù…H¢IEXùe2¢ùXø(ô54I¢I eô54‡ùˆô§™3X ûtübû ‡ $1¢¥€¤ ùXù0Ëø ùT…1¦§¢¥EXù02‚¤© ô¨  ¢U4‡ù`£ ø ¢ƒõ‡• ù 4I¢¥¦ö÷¢• ù ¢I “ C‘2bô£¢ ù`ø(ùT ! eô¨)„ùQ¦öù04ƒùT¦¨Xû„˜€(C ¦7ù0  ð ‘ ‡ £ ¦„d ¦b Ø Œnú´û ‡¥› Ž ¢¥12¨ø(ùT¢ƒø(ø¢¥bõ‡ù‹ (ùXøX)~ ‘ Ûa‘ Œ„ Ž Û(Ü × ‘ ç Ö Œ„ Ž Ú “ ‡ ÛyÝ × ‘ ðpñeÚ “ ‡ Ûh ‡ ð Ú Ý Ü ×€ò “ Þ ð Ú “‡ Þ Þ ¥ × ƒ“$bù`ø(ù}ô§€1P¤4|ô§ƒ÷1ÎÏÕE ¤ ùd I C ù`ùT £b $ ùe¤H¢¥(ô¨E„¤DύÙ¢¥12‚ $bô§—…HH¢¥¦’ÎH¦xûƒŠvùT1E`ù¥`Ó $bùXø ù ô§—1Íô¨ eùXøX)„ù02bôÔ¢y eùe61‘ ' ø¡¥ —Xe   ø ÷1E0(   ÷bø(ù& £s  ô¨1… ùTEd†   û€@1X(   ùU¢¥2¢¥b  ùWb … øX3` E ùXùT2V•  ô§)‘„ ) ù0b 2 ôÔ¢ye   ùT¦5©~   £$¢ùp‰'ø¡¥ X  û .

¦(§´Ê¡§ ±“¹“ªQ´3°‡´|µs®DªW±€· „ üiú ƒ $bù‘(ô§)‘…1¦7ùTX C¢U©c £7EU¢ƒø øX© $bô§•‡÷V ”ô¨¤ˆ©c)"ùU¢I1•¥ $bù"ó†ƒ÷bø(ô7ùXøX £¢¥¤bô¨¦öô¨ ©¢¢¥H¢¥¦5©¡(ô§ “ ô§ (ø¡2b÷1EXù02€ô¨€ÿ´ùTEd eô§¥Ëú|û ú´ûAƒ“$¢ô§!¦öùT¢¥21!¢I !VEXù• £ Œ„ Ž7‚ÖçŒ „ Ž7 Ý × ßWe svu – w e Ý svu – w Þ ð Þ ð Þ ¥ ð á ‡ø Œ„ Ž7 Ø Œ„ ŽAÖ é ‡ 𠏍E0{Œ’󕦄b9Ž Þ ð ¥ ¥ ¥ 𠃓$bù¦7÷V eô¨v¥l £$bùHX%÷H¢¥¢ 2 ø£¢I (ô§Emô§ Þ ð ƒ“$bù`ø(ù‘¢ƒø(ù ‡ ãô‰'ø¡¥ ‹¢¥¦¨¦f¥DC!$bô¨Ea$y¢ƒø ùPê „ "ƒø¢¥©Çø(ùT¢¥¦ ð ô¨¢ £$bù}ø¢¥bõ‡ùxé†ê ð êöõ·û ‰ ÷ ƒ“$bô¨—)"ùU¢I1# £$H¢y — $bù•)„ùd £$1¡2 ô§ ¡xÓuawyx‰ˆ¥iq¤tiÍwyx‰Ë¥pqprsW¥d¤ÍËhTp§ ƒ“$bùS¢¥¤9(4ƒùØø ùT ÷1¦¨ | ùXùT)‘R £¡Ûõ3¡2‹ £!¤ ù’ eø ÷bù¥`y(  ô¨1X E ùAh C ù„$1¢(4ƒù„"ƒ÷112¢¥”÷11ET¥12bô¨ (ô§1¢¥¦§¦5© X £¢¥¤1¦7ù#)„ùd £$1¡2}÷1 ô§bõ•¢¥ ‰#ø{ù‰priÍudiq¤ ET)e¤bô§H¢ye   ô§¥e¥9â|ƒ ¢ õ‡ø¢¥b  õ‡ô§¢¥”  ô¨ˆ(   ùXøX…9¦§¢Ie   ô¨e…‰¦5©3V„ ) ô§¢¥¦§`  û ƒ“$bù•…bø ô§EXù‹Chù•$H¢U4‡ù…H¢ƒô¨2P"‡ø £$¢ô§ô§# $bù¦§¥“¥ uawyxBú ¥di§¥d¤‰dxÓuds CÛô¨ $7 $bù•‡ø(ô7õ‡ô§1¢¥¦ˆ¤Dύû ƒsxb … øX(‡ 4 ù‹ £¢ $ ô§T`3CIùmù–V…1¢¥12v £$bù‹ (ùXøX)~5ô¨73X û£ú´û ‡I› ô§v¢Wƒs¢U©3¦¨‡ø# ùXø(ô7ùT!¢¥V2¨ø ùTE01X (ø(÷1Ed   £$¢ù—…1¢ƒøX (ôÔ¢¥¦32bô¢¢EùXø ùT eô§¢¥¦bùtX%÷H¢y eô§¥xChù€¢ƒø(ù!¢¥Ed e÷H¢I¦§¦¨©&¥¦¨4|ô§bõ•¢ID $bù!)„ù0$‘(ô¨–Xù!¤Eù0ET)"ùT’4ƒùXø‡© )~¢¥¦¨n ¦ ûuóH‡  ø# b $ ùe   ô¨)„ù‹2¢ùXø(ô54I¢I eô54‡ù‹Chù&$H¢U4ƒù ‡ ¦„d ñ Ú “ ‡ „ Û(Ü × ‘ ¥ Ú “‡ ‚Ö ð E0qyB¦bg™  „ „ ¥ ð Ø (ô§ Ø yB¦„b ÛIÝ × ‘ ç Ö Œ â Ú Ž ‡ Ûa‘ i “ Þ ò ˜ „ ¦„d Ø Œ ⠎ Ûa‘ ivivi Ú “ ‡ Þ papap ¢¥127"‡ø# $bù•)„ô©¢ù02v eô§)"ù¢¥V2™X…H¢¥EXù•2bô£¢ ù`ø(ù01EXù0 Ûa‘ Û(Ü ‘ ÛIÝ ‘ Ûa‘ ڈ“ ‡ Ý ˜ Ú “‡ × ¥ D Ú “‡ × Þ ˆ Ú “‡ Ü × × ¥ D ¦b Ø Ö Œâ ¦„d Ø Ž ‡ Ûa‘ Œâ Ú Ú w w “ ¥öû „  ivi ¦ byü „ ¦„b Ø Œ ⠎ ‡ Ûa‘ Ú “ ¥  w  w  w w „(‡ ¦„d Ø „ ¦„d Ø Œâ Ž Ûa‘ ivivivi Ú “ ‡ û ¦„b3ü „(‡ Ž ‡ Ûa‘ “ Þ Þ papap Œnú´û—FéŽ “ÛùT…1¦§¢¥EXù£ £$bùy (ùXøX)~~ô¨¸3Xtûmú´û ‡¥› CÛô5 £$ ¦„d c~¦„bÌý ¢ é û’y b ù…HV2 â Ú Ö â d  $bù䢥¤‰(4‡ù›ù–V…1¢¥1(ô¨1c¢¥V2 Ø â Ø Ú â dØ  a¢PIQùY £$bùy¦öô§)"ô¨ †¢I Œnú´û—F „ Ž â Ø Ú £ â b Ø ¥ ‘ þ £ÿþ ¦ d „ ¦„b C!$bù`ø(ù .

„ üæ ±’¼€²Qº‹µs®D¯¿¦(§À´|µ1²Q¶°€³R°Tµ!·¸¹“Á ³R°€ªQ®A²‹¯Â´|·D´|µs®„­¢´ s ùH‰d‚0hdwypriÍu ûDƒ“$|÷VBô5 ¦„dmý é~¢¥V2”¦„bžý érô¨7(÷1Ea$  X ûEú´û ‡ úô§ ùVË¥‚XËhdwypriÍu ûD3Xtû ú|ûGF „ ô§ Æ¡ 3 – ¢™C¢U©y £$1¢I  – i ø ùT)~¢ƒô§1ET¥1X £¢¥ U`A £$bùsùtX%÷H¢I (ô§yChù7¢¥Ed e÷H¢¥¦¨¦¨©cX¦¨4ƒùP¤©y £$¢ùP)„ùd £$1¡2›ô¨ w ¢U4‡ùù{X|÷1¢I eô¨ƒ`†1¥ Q¢‘2¢ô¢¢E÷1(ô¨€ù{X%÷H¢I eô¨0û“@™(÷1Ea$c¢‘EU¢¥ ù É 3Xtû0ú´û ‡¥› ô§€¢‘C ¡£¢¥¤¦¢¨§© i§¥exÓwy¤ ¡xHi Uwy ‚ gprs7uawyxV¥di§¥d¤€‰dxH¤„ÄAiq¤qƂ¤qÆH‰x‰ ¡ 1Ë¥¤viÍwyxcĉ•¥(‰d¤—wy¡¤#¤€wx¥(wyprÃ¥‰ ù0‡ 4 ù0¢"‡ø“4I¢¥bô§X$bô§¢õ¦¨©)~¢¥¦§¦ X(   ùT…†(  ô¨X – ùT`  ûDƒ“b $ ù•(  ô¨(   ÷H¢I(   ô§Í  ô¨“(  ÷1)‘)~ƒ ¢ ø(ô¨X – ùTš 2 ô¨vƒf¢I¤1ö ¦ ù”ú´û—¢ F ûA@B˜¦„dyý 鑢I1ž 2 ¦bý  é ô¨v(  ÷1Ea$ ¢™C¢U©y £$1¢I  – – i ø(ùT)~ƒ ¢ ô¨1eE01X £¢¥ U`’ b $ ùT Ø ô§  P— Œ ¦dXŽ–`A¢¥12‘ £b $ ùP" ) ù0 $1¡› 2 ô§&ET1 ô§‡e   ùTX   û þ  w  Š€(h C ù0ƒ 4 ùXøT`l £b $ ô¨ö ¦ ùT¢¥21‹ £†¢vET¥1X(   ø£ƒ ¢ ô¨ˆ  b $ ùge   ô§" ) ùW‡e   ùT…cC!b $ ô¨Ea$ ô¨~  Ù ÷1X ¢¥• ù0‡ 4 ù`ø(ù¢¥‹ £$H¢y  ô§)‘…‰(  ù02¤ˆ©‚ £b $ ùxX £¢¥b ¤ ô¨ö ¦ ô¨ ©™ET¥1b 2 ô¨(   ô§I„ b $ ùx(  ù0ET12¡'‡  øXb 2 ùXøÛ “ ÷1b  õ‡ù–'€m ” ÷V  £¢v„ ) ùd £$1¡2c$1(C! ô§v b $ ù a¢¥¤17 ¦ ùƒû ‡ 123'B‡ø2bù`ø!“v÷Vbõ‡ùd'”m÷V  £¢ ¦d#ê – w  Ø  óH‡ø—1 ô§X (ùT1Ed© óH‡øvÿ¡ £¢¥¤bô¨¦öô¨ © ϕ÷£óH‡øX ´ó£ø£¢¥•IƒùT¦ ¦„dê‡õ 12bô5 eô¨H¢¥¦¨¦¨©sÿ¡ a¢I¤1¦öù —bô¨"ƒø)‘¦¨©1 ô§X (ùT  CÛô5 £$ !$# Ö ! # ! i €VET12¢ô¨ eô¨H¢¥¦¨¦¨©„ÿ¡ £¢¥¤1¦7ù 12¢ô¨ eô¨H¢¥¦¨¦¨©~V(ô§‡ eù0ˆ T` Q…V…bøU´ô§)~¢I eù0"!$# Ö ! i 1¦¨s © ô¨ Ø %(' ß – i & £ – w á ƒ“¢ $ ùXø(ùd"‡ø(ù ¦„d % ! # £ ! w £ ! w  ¦b0)  1 ƒf¢I¤1¦öù”ú´û—F¢þØÿ´÷V)~)~¢ƒøX©¥f¢¥ETE`÷bø£¢¥Ed©†¢¥12†ETV(ô§‡ eù01E0©7E012bô5 eô§¥1"‡ø!“—” ‡ ¢I12 Ï ÷€óH‡ •  ø‡ X ' ótø¢¥5  IQùT¦ƒ" ) ù0 $132VX  û ' Ö碥†ƒ ¢ ø¢ ¤ ô¨e   øƒ ¢ øXs © ùXø ø‡  ø“¤‰‡  ÷112 û ëPì32 4çðAî‘÷øTò65 ý ã 㠄 û€V(ô§2¢ùXø“ £$bù&΍ύ ÚHà Ö CÛô5 £$ 7ÙÖ Ú Ö87 Ú @ Þ 9 d ¥ èA é ¥ „ é „ é „ ¥ é š5„ ¥ „ „ é š5„ ¥ „ C îB „ ï c 9 Ö èA é ¥ é C îB é ï „ .

¦(§EDˆ§ÀºD¯¹“¶#³|®„­¢´ „ ü › Œt¢Ž q ó ô§12P $bùmùXô7õ‡ùT4I¢¥¦ö÷¢ùT“¥F7÷1 ô§bõx¢W%÷1)"ùXø ô§EU¢I¦ƒ…H¢¥EIy¢ƒõ‡ùQû’bÕ$H¢I Ûô¨ £$bù‡ eùT¢¥2V©' ‡ a¢I (ù•¦7÷V eô¨ƒ`™Š—(C¸21|ùT“ £$bù&΍ύÅX¦ö÷3 eô§¥v¤Eù0$H¢U4‡ù”ô§v eô¨)„ùh` Œv¤HŽPb ( ø ô5 eùg¢~ET¡2bùe £šô¨ eùXõƒø£¢I (ù&{ø¥)  $bùô§¢ô¨ eô§¢¥¦RET¥12bô¨ (ô§ Ú Œt鎓ÖHG „ c „ c „PIRQ {ø)   ô§" ( ) ù~dAÖ é“÷1 ô§bõ‹ $bù5ùd3…1¦7ô§E`ô¨ A§÷1¦7ùXø0`‡ô§)‘…1¦öô¨EXô5 le÷V¦öùXø0`ˆ¢¥12‘ã‚¢IE(‡øX)¢¥EI&)„ùd £$3' ¡21`  û@‘¢¥¦§¦f b $ ø(ù`ù~ET¢¥(  ù0Tq ` ÷1(  ù8YÖ é š¨„ "‡ø „ éé¥év(   ô§" ) ùgX(   ùT…10`A8yÖ”é šG‡ "‡ø&d¥é¥é   ô§" ( ) ùPX(   ùT…10`“8ÕÖ é š ü¢"‡ø ‡ d¥é¢(   ô§" ) ùX(   ùT…1~¢¥12Å8ÕÖ „š é‚"‡ø „ éé¢(   ô§" ) ùX(   ùT…1`  û ¥)~…HQ ¢ ø(ù £b $ ù•E0)~¢ … ÷Ve   ù02™Xö ¦ ÷V(   ô§PÛ C ô5 £$v £¢ $ ù”ù–V¢¥E0 —X(   ùU¢¥23©77 ¦ ÷Ve   ô¨0  û ŒvEUŽS—(ô¨bõ £$bù†6V'!žø(ù0¦Ô¢I (ô§V&"‡øW $bùT ù7 $bø(ù`ùv)„ùd £$1¡210`„C!$H¢y ‘¢ƒø(ùP $bùšùd3… ùTE0 (ùT2 ¤‰‡  ÷1V21!™87"‡ø“‡ a¢¥b ¤ ô¨ö ¦ ô5 5 © `7“  ø ù‹©‡÷bøÛø ùT(÷V¦¨ £“E01 ô§X (ùT !CÛô¨ $7 $bùT ù•¤9‡÷V121` ‡ û Œt¢Ž†¥)~…b÷3 eù~¢7 £¢¥¤1¦öùgI £$¢ù~%÷1)"ùXø ô§EU¢I¦’4¥¢I¦ö÷bù0I £$¢ù¿‰'ø¡¥ IS $bù ‡ 123'ƒø2bù`ø €2H¢¥)‘‡'’#¢¥$3"‡øX $¢„ ) ùd £$1¡2™C!¢ $ ùT£6‚Ö œ û—ƒs¢PIƒùx8~ô¨Ëô§ (ùXøX4I¢¥¦¨!¥„é¢û—édx{øX) é  ´ é ûGæéx¢¥V2†E0)‘b … ÷Ve   ù b $ ù¢¥¤1¥ö ¦ ÷V(   ù‹4¥¢Iö ¦ ÷bù0—Is b $ ùˆø¡¥ £# ~¢y !ö ¦ ùU¢IX €˜g…1¦Ô¢¥` E ùT`û Œv¤HŽÌ¤A¦¨¥ D £$bù— eø£¢IEXù—¥Ó $bù“øX3¥ eô§x £$bù—ET)‘…1¦7ùd¿9'B…1¦Ô¢¥¢ù€¢I122bø¢UC  £$bùB÷1bô5 SE`ôöøXET¦öù ¥† b $ ù•£¢¥" ) ù…1¦§I†   û ŒvEUŽ†“ÛùT… ùU¢y — $bù¢¥¤‰{4ƒù"ƒø# £$bù“—” ‡ )"ù0 $1¡20û F¢û“bÕ$bù0e¢I…1…1¦öô7ùT2Q # £$bùD¦7ô§bùT¢ƒøÓE04‡ùTEd eô¨ù{X%÷H¢I eô¨ƒ`U $bù’CÛô§2bù0¦¨©I¡V{C!&â|¢yTˆb ùT12bøXP¢ ) ù0 $132 õ‡ô5‡ " 4 ù0X  þ Ú “ Û(Ü × Ö Û Ú “ ¥ × Œ Û ‡ Û Ú “Ü × ¥ „ Û Ú “Ý ׎ Þ ‡ „ × ÛØ Œ Ú Û Ü “ × ¥ ‡ Ú “Û Þ Ú “Û Ý Ž × C!$bù`ø(ù¶× Û `FIˆ1(C!Y¢I‹ £$bù‡÷bø¢¥ PŒ"‡ø5ó’âlŽ‹%÷1)W¤ ùXøT`kô¨ 8 ç ¦„bkûU€ ô§bõ¨óH‡÷bø ôöùXø ‡ a¢¥b ¤ ô¨ö ¦ ô5 ©P¢¥1¢¥¦¨©¡(ô¨T`V…112† $bùˆø£¢¥¢õ‡ù¥3× Û "‡ø#C!$bô§Ea$7 £$¢ù)"ù0  Ë $1¡2¨ô¨“X £¢¥¤1¦öùQû übû—Ï•ùd eùXøX)„ô¨bù€¢¥12v…1¦¨¥  $bù‹X £¢¥¤bô¨¦öô¨ ©WET £ƒ÷bø#"‡ø„ £$bùQQ2H¢¥)‘‡'’#¢¥X$V"‡øX $P)"ù0 £$V321S¥ ‡X ø 2bùXø „  £$bX ø ‡÷b õ $„übD û )‘…H¢ƒø ! ù CÛ5 ô  £$‘ £$b€ ù “vV ÷ bõ‡d ù '”mV ÷   £¢)"0 ù  $132V„IR‡ ø 2¢ùXø „  $b ø ‡÷¢ õ $ übû d´û—“ÛùTET¢¥¦§¦Ó $bù•ET)‘…H¢¥Ed !EXù0ˆ (ùXø ùT2Fy'B…9ƒô§ !¢¥…1…¢øU´ô§)~¢I eô¨"‡ø!¢W…£øXX —2bùXø ô¨4I¢I (ô¨4‡ùQþ ŒWV Ú Ž “ Ý ü†ŒWV Ú Ž “ ŒV Ú Ž“Ü Ö W × Þ Ó Þ × w w w F ¦b ŒÚ “Ü Ž × ¥ Ú “Ý × ’S©“ø(ù0…1¦Ô¢¥E`ô§bõ# $bùA…1¢I eô§¢¥¦ƒô¨12bùdYX€¤©‹ $bù’ eù0)~…‰‡ø£¢I¦ƒô§V2bùda`f`y¤V a¢Qô§¢# eô¨)„ù–')~¢ƒøEa$¢ô§bõ )"ù0 $13” 2 ÷1(  ô¨b  õQ £¢ $ ô§s"ƒø ) ÷1¦§¢ ¢ ûRbÕ$H¢I f‡  øb 2 ù`øuô§| £¢ $ ù#„ ) ùd £$1¡ƒ 2 `v@u  ô¨q   ùd3…17 ¦ ô§` E ô¨ s‡  ø©ô¨)~…17 ¦ ô§` E ô¨  ` @Û  ô¨ —¢W CSI'‡e   ù0…¢âlã¢ã `&@BsV`1 £xC!$H¢I “4I¢¥ö ¦ ÷¢ùT!¥Fb1`H%¡`†¢¥1d 2 cä9 Œ ô§†3  Xtû†¢ ˜ ûGd › Ž21|  ùTÛ  ô¨  E0‡  ø(ø ùT…‰1B 2 `ŕ Ï ùXø ô¨ƒ 4 ù& £¢ $ ùx61Ï ' yø ùT¦Ô¢ye   ô§¥†"‡ø— £b $ ù&„ ) ùd £$1¡0 2 û€@•  ô¨ €—'BX a¢I¤1ö ¦ ùIe ` 7gfX'‡ a¢¥¤17 ¦ ù¥` ‡ø“@Í'BX £¢¥¤1ö ¦ ùh` .

„ d¥é ±’¼€²Qº‹µs®D¯¿¦(§À´|µ1²Q¶°€³R°Tµ!·¸¹“Á ³R°€ªQ®A²‹¯Â´|·D´|µs®„­¢´ ˜¢û“b ( ø ô5 eù‹ £$¢ùW΍ύÀ‡©3‡ eù0) ¥¤V a¢ƒô¨bùT2†¤©†¢¥…1…1¦5©|ô§bõg £$¢ù ‡ 123'ƒø2bù`ø—E`ùT eù`ø(ù02¢2bô¢¢EùXø ùT1E`ù ¢¥…V…bøU´ô§)~¢I eô¨Y £ $bù7…H¢I (ôÔ¢¥¦“2¢ùXø(ô54I¢I eô54‡ùšô§ä $bù™)‘¡2bùT¦#2¢ô¢¢E÷1(ô¨“ùtX%÷H¢I (ô§äCÛô¨ $ … ùXø ô§¡b 2 ô§E‹¤‰‡  ÷1V2Hƒ ¢ øX©ETVb 2 ô¨(   ô§1`  ûh€ ô§b  õxQ…VE … ù01b 2 ô57“ ’ û’üV`V…H127 £b $ ùˆù`ôöõ‡ù0ˆ4I¢¥7 ¦ ÷bùT!¥l £¢ $ ù X…H¢Ie   ô§¢¥¦ƒ… ùXø¢I £ƒ  ø“)¢I(   ø(ô© û • ô¨ƒ 4 ùT7 £$H¢y “ £b $ ù&6V! '  ø(ùT¦§¢Ie   ô¨"‡ø“ b $ ù ‡ 123'B‡  øb 2 ù`ø!Q2H¢¥)‘‡' ’#¢I$V"‡ø‡ £$†„ ) ùd £$1¡2 ô§  Ø ¥ Œ„ Þ F6‰8 ç ‡ Ž7 Þ 6‰8 ç ‡ ÖÙé  $1(C¸ £$H¢I — £$¢ù&)~¢y´ô§)÷V)»X £¢¥¤1¦7ù4I¢¥¦7÷bù•¥¥’ 6‰8˜’ˆ"‡øÛø(ùU¢I¦l¢ùXõ¢I (ô¨4ƒùW6R`EôŠû’ùQû¨`1 $bùe…‰‡ô¨ˆ  X C!$bù`ø(ù $bù‡ a¢¥¤¢ô§¦öô5 ©cET ‡÷bø”ô¨ eùXøX(ù0E0 £& £$bùP¢ùXõ¢I (ô¨4ƒùsø(ùT¢¥¦S¢y´ô§T`qô§&¤V a¢Qô§bù02‘CÛô¨ $ 6‰8™Ö  õ~ b $ ù”ù`ôöõ‡ù0ˆ4I¢¥7 ¦ ÷bùT—¥’ b $ ùeX…H¢I(   ôÔ¢¥¦|… ùXø¢I £‡  ø—)~¢Ie   ø ô59`H…H12‚ b $ ùe)~¢y´  ô5' „ ûi—(ô¨b ¥ ) ÷V) X £¢¥¤1ö  ¦ ù“(   ô§" ) ù“Xe   ù0…~"‡ø’ b $ ù—ET)W¢ ¤ ô§H¢I(   ô§gI ‡ 12¡'‡  øXb 2 ùXøA` E ùTe   ù`ø(ùT2v¢ 2 ô¢¢EùXø(ù01X E ù0¢I12  b $ ù ‡ V23'‡  øXb 2 ùXø&Q2H¢I)~€'€’#¢¥X$V"‡ø‡ £$y" ) ù0 $1¡2䢥…1…17 ¦ ôöù02‘ £† £b $ ù)~¡b 2 ùT¦Db 2 ô£¢ ÷V(  ô§¥y  ùtX%÷H¢{'  (ô§0  û’v “ ù0E … ùT¢Iv   ÷1(  ô¨b  õ"ó†ƒ  ÷bø(ô7ùXø!¢¥H¢I¦¨©¡(  ô¨X  û ú´û€V(ô§2¢ùXø“ £$bù"¦¨¦§(CÛô§¢õ¤DύÛþ â Ú Ö â d œ â Ø Ú â b Ø C!$bù`ø(ù œ Ö  „ ûQbÕ$bô¨Ea$~ùd3…1¦7ô§EXô5 € (ô§)"ùd'B)¢ƒøXEa$bô§¢õ‘)"ù0 $1¡21QCS‡÷1¦§2‚¤Eù&(÷bô5 a¢¥¤1¦7ù"ƒø ¥ ô¨ˆ(   ùXõ‡ø¢I eô¨bõ‘ $bô§ ¤„ύ“`hô¨ ‡ 12¡'‡øX2bùXø~EXùT (ùXø(ù02 2¢ô¢¢EùXø(ù01EXô¨bõ»ô§„÷1(ù02Ò"‡ø £$¢ù¢…H¢{'   ôÔ¢¥¦„b ( 2 ô¢¢EùXø ùT1` E ùTx¢¥V2  $bùP¤‰‡÷1121¢ƒøX©yET12¢ô¨ eô¨1e¢ƒø(ùv… ùXø ô§¡2bô§Eh` ó©ô§V2y £$bùv‡ a¢¥¤¢ô§¦öô5 © E01b 2 ô5e   ô§¥P"‡ø“b  ù•¥l £b $ ù0(  ù„ ) ùd £$1¡21`  û æ¢û&—(ô§¢õËóH‡÷bø ôöù`ø&¢IH¢¥¦¨©¡ ô§T`l¢¥1¢¥¦¨©¡–Xùx £$¢ù~‡ a¢¥¤bô¨¦öô5 ©¢¥#…bøX ‡'‡øX2bùXø¤H¢¥EIC¢ƒø2y2bô¢¢EùXø ùT1E–' ô¨b  õ~E0‡  ÷1…17 ¦ ùT2†CÛô¨ £$¨ùd3…1¦7ô§E`ô¨ —§÷1¦öù`ø— (ô§)"ù•)¢ƒøXEa$bô§¢õ~¢I…1…1¦öô7ùT27 £x £$bù&¦öô¨bùU¢ƒø“E04‡ù0E0 eô¨ ù{X|÷1¢Ie   ô¨vÛ C ô¨ £$7…‰(  ô5e   ô5‡ 4 ù ûØó©ô§12v b $ ù•)¢y´  ô¨ ) ÷1)»‡  ÷bø£¢Iˆ —%  ÷1)W¤ ùXø“"‡ø“‡ a¢¥¢ ¤ ô§ö ¦ ô5 ‡ © û › û€V(ô§2¢ùXøƒ £$¢ùA¦öô§¢ùU¢ƒø9ET4‡ù0E0 (ô§•CÛô¨ $&¢#…‰ ô¨ eô54‡ùfC¢U4‡ù„… ùXù02e¢¥gô¨…bø¥¤1¦öù0) æ¢ûlQ…V…1¦¨© ¢g• Ï ô7ø(ô§Ea$Vö ¦ ù0 ¤9ƒ  ÷112Hƒ ¢ ø‡©PE01b 2 ô5e   ô§¥v¢I “ £¢ $ ùö ¦ ù0q #¤9‡÷V12H¢ƒø‡©‡ûD˜€x¤‰‡÷1121¢ƒøX©ET12¢ô¨ eô¨Íô¨ … ùXøX„ ) ô5 e   ù02x¢I ’ b $ ùÛø(ôöõ¥$ˆ ’¤9‡  ÷V12Hƒ ¢ ø‡‡ © ûfb ø(ô5e   ù b $ ù!X©¡X(   ùT) ¥Ó΍ύ“З’C!b $ ô¨Ea„ $ ø ùT ÷1¦¨ ’{ø) … øXX ‡'‡ £  øXb 2 ùXøs¤H¢¥ E IC#Q ¢ ø2x…H¢ye   ôÔ¢I¦3b 2 ô£¢ ù`ø(ù01X E ô¨b  õ“ô¨W)¢ye   ø(ô©ˆ'B‡ 4 ù0E0 ‡  øl"ƒø¨ ) ûp—(ô§¢  õQQ…VE … ù01b 2 ô5 ’ û „ `1…HV2† b “ $ ù&61‘ ' ùXôöõƒùT4I¢¥ö ¦ ÷bù0X  ûAb ø(ô¨(   ù‹ £b $ ù&§ Î ¦xÅC!b $ ô¨EaË $ ø(ù0(  ÷1¦5 £“{ø¥)  £¢ $ ùe¢¥…V…1ö ¦ ô¨EU¢I(   ô§ ¥qùd3…17 ¦ ô§` E ô¨ !§  ÷17 ¦ ùXø“e   ô¨„ ) ù‹)~ƒ ¢ øXEab $ ô§b  ä õ}ô¨7)¢I(   ø( 4 ù0ä E0 £ƒ  ø#"‡ø)†¢ ` ôŠû’ùQû¨` ä ô©¡'͇ Ö × ÚHÛ ä ÚHÛ(Ü × ‚ ¥ qyÛ b ø(ô5 eùž× ô§«¤H¢¥12bù02À)~¢I eø ô5 V¥ a¢I (ô§«¢¥V2Sõ‡ô54‡ù7 £$bù¨ùT eø ôöù0~¥ q û(€ ô§bõc £$bù¢61'  ø(ùT¦§¢I eô¨Ò"‡øv £$bù ùd3…1¦7ô§E`ô¨ ™§  ÷1ö ¦ ù`ø†" ) ù0 $1¡2ƒ`‹…H12Ù £b $ ù 9 ' ù`ôöõ‡ù04¥¢I¦ö÷bù0Xû¿’#¢I(ùT2   $bùT ù¥`1C!$H¢I 5ô§ $bù‹)~¢y´ô§)÷1) ‡  ÷¢ø£¢¥ !%  ÷1)W¤ ùXø“¢I¦§¦§(h C ùT2P"‡ø¢¥X©¡)‘…V £Ie   ô§EQX a¢Ib ¤ ô§¦7ô¨ ©•` f3…1¦Ô¢Qô§vC!$©v £$bô¨“2bô¢¢EùXøX#{ø)» b $ ù¢¥1‡I C ù`ø— ‘¢ … ø¤17 ¦ ùT)»¢ æ ûDB Š ô¨ˆX   þ§ô§˜  ×ç1ƒ  ø)~¢¥ ¦` Ë .

1 Sti ness De nition for ODE's 8. 7. 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.1. However.Chapter 8 CHOICE OF TIME-MARCHING METHODS In this chapter we discuss considerations involved in selecting a time-marching method for a speci c application. Examples are given showing how time-marching methods can be compared in a given context. In the following discussion we concentrate on the transient part of the solution. Any de nition of sti ness requires a coupled system with at least two eigenvalues. and the decision to use some numerical time-marching or iterative method to solve it. De nitions given in the literature are not unique. The forcing function may also be time varying in which case it would also have a time scale. we assume that this scale would be adequately resolved by the chosen time-marching method. The di erence between the dynamic scales in physical space is represented by the di erence in the magnitude of the eigenvalues in eigenspace.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. but fortunately we now have the background material to construct a de nition which is entirely su cient for our purposes. An important concept underlying much of this discussion is sti ness. since this part of the ODE has no e ect on the numerical stability of 149 . which is de ned in the next section. 8. and.

the homogeneous part. Consider now the form of the exact solution of a system of ODE's with a complete eigensystem. the time integration is an approximation in eigenspace that is di erent for every eigenvector ~ xm . 6.27 and its solution using a one-root. 8. 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. . 6. although these eigenvectors must remain coupled into the system to maintain a high accuracy of the spatial resolution. 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.150 CHAPTER 8. we exclude the forcing function from further discussion in this section.1: Stable and accurate regions for the explicit Euler method. if at all. This is given by Eq. For a given time step. The situation is represented in the complex h plane in Fig. In 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. CHOICE OF TIME-MARCHING METHODS Stable Region λh I(λ h) 1 R(λ h) Accurate Region Figure 8.28.1. timemarching method is represented by Eq.

it states that we can separate our eigenvalue spectrum into two groups one 1 ! p] called the driving eigenvalues (our choice of a time-step and marching method must accurately approximate the time variation of the eigenvectors associated with these). p+1 ! M ]. First we order the eigenvalues by their magnitudes. . Unfortunately. but their presence must not contaminate the accuracy of the complete solution).1. Rephrased. and the other. although time accuracy requirements are dictated by the driving eigenvalues. called the parasitic eigenvalues (no time accuracy whatsoever is required for the eigenvectors associated with these. 8. numerical stability requirements are dictated by the parasitic ones.1) Then we write p M X Transient = X t m ~ (8. we nd that. An inherently stable set of ODE's is sti if j pj j Mj Cr = j M j = j pj Mildly-sti Cr < 102 Strongly-sti 103 < Cr < 105 Extremely-sti 106 < Cr < 108 Pathologically-sti 109 < Cr It should be mentioned that the gaps in the sti category de nitions are intentional because the bounds are arbitrary. STIFFNESS DEFINITION FOR ODE'S 151 8. thus j 1j j 2j j Mj (8. and imaginary eigenvalues.3 Sti ness Classi cations In particular we de ne the ratio and form the categories The following de nitions are somewhat useful.8.2 Driving and Parasitic Eigenvalues For the above reason it is convenient to subdivide the transient solution into two parts.1. complex.1. It is important to notice that these de nitions make no distinction between real.2) cm e xm + cme m t ~ xm Solution m =1 m = p +1 | {z } {z } | Driving Parasitic This concept is crucial to our discussion.

in di usion problems this sti ness is proportional to the reciprocal of the space mesh size squared.. One quickly nds that this grid clustering can strongly a ect the eigensystem of the resulting A matrix.1. For a mesh size M = 40. 2 and the ratio is . we are interested only in the converged steady-state solution. and are greatly a ected by the resolution of a boundary layer since it is a di usion process. In order to demonstrate this.152 8. Our brief analysis has been limited to equispaced x . but much less so than for di usion-dominated problems. Under these conditions. this ratio is about 680.2. In this case the eigenvalues are all real. The simplest example to discuss relates to the model di usion equation. and in a boundary layer.e. Consider the case when all of the eigenvalues are parasitic. 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. near an airfoil leading or trailing edge. let us examine the eigensystems of the model problems given in Section 4. 8. Even for a mesh of this moderate size the problem is already approaching the category of strongly sti . 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. A simple calculation shows that 4 2 1 = . As a result it is quite common to cluster mesh points in certain regions of space and spread them out otherwise. negative numbers that automatically obey the ordering given in Eq. Examples of where this clustering might occur are at a shock wave. 4x2 x 2=. the sti er our equations tend to become. 2 sin 2(M + 1) x M ! . the sti ness is determined by the ratio M = 1 . For the biconvection model a similar analysis shows that j j=j j 1 M x 1 Here the sti ness parameter is still space-mesh dependent.2 Relation of Sti ness to Space Mesh Size CHAPTER 8. Furthermore. CHOICE OF TIME-MARCHING 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. 4x2 sin2 2 = . i.3. Typical CFD problems without chemistry vary between the mildly and strongly sti categories.

Era .3 Practical Considerations for Comparing Methods We have presented relatively simple and reliable measures of stability and both the local and global accuracy of time-marching methods. and technology in uencing this is undergoing rapid and dramatic changes as this is being written. among other things. and the accuracy would be related to how much the energy of certain harmonics would be permitted to distort from a given level. The actual form of the coupled ODE's that are produced by the semi-discrete approach is ~ (~ At every time step we must evaluate the function F u t) at least once. relevant conclusions can be reached. the time interval would be that required to extract a reliable statistical sample. it is. Let us consider the problem of measuring the e ciency of a time{marching method for computing.8. It should then be clear how the analysis can be extended to other cases. Since there are an endless number of these methods to choose from. This function is usually nonlinear. highly dependent upon the speed. and architecture of the available computer. over a xed interval of time. discussed in Section 6.5.6. 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. d~ u =F ~ (~ u t) dt . era . The length of the time interval. T . one can wonder how this information is to be used to pick a \best" choice for a particular problem. For example. 8.3. PRACTICAL CONSIDERATIONS FOR COMPARING METHODS 153 problems. We refer to a single calculation of the ~ (~ vector F u t) as a function evaluation and denote the total number of such evaluations by Fev . The appropriate error measures to be used in comparing methods for calculating an event are the global ones. For example. in calculating the amount of turbulence in a homogeneous ow. and the accuracy required of the solution are dictated by the physics of the particular problem involved. There is no unique answer to such a question. Let us now examine some ground rules that might be appropriate. and its computation usually consumes the major portion of the computer time required to make the simulation. Nevertheless. capacity. an accurate transient solution of a coupled set of ODE's. if certain ground rules are agreed upon. and erp discussed earlier. rather than the local ones er . Er and Er! . Such a computation we refer to as an event.

i.4 Comparing the E ciency of Explicit Methods As mentioned above.. and must use a constant time step size. We judge the most e cient method as the one that satis es these conditions and has the fewest number of evaluations.2 An Example Involving Di usion Let the event be the numerical solution of from t = 0 to T = . the e ciency of methods can be compared only if one accepts a set of limiting constraints within which the comparisons are carried out.48. Eq. 8. In some contexts.4. this makes the exact value of u at the end of the event equal to 0. To the constraints imposed above.e. ln(0:25) with u(0) = 1.. must simulate an entire event which takes a total time T . 2. u(T ) = 0:25.1. must be < 0:5%. and RK4. this can be an important consideration. a = 0.3) Er < 0:005 eT . must have the property: du = . so that CHAPTER 8.e. Since the exact solution is u(t) = u(0)e. let us set the additional requirement The error in u at the end of the event. First of all. The follow assumptions bound the considerations made in this Section: 1. i. Implications of computer storage capacity and access time are ignored. the allowable error constraint means that the global error in the amplitude.u dt (8. CHOICE OF TIME-MARCHING METHODS 8.25.t. Fev .3 is obtained from our representative ODE with = .154 8. Three methods are compared | explicit Euler. 3. 6. see Eq. 8.1 Imposed Constraints T = Nh where N is the total number of time steps. The calculation is to be time-accurate. the global error. h. AB2.4. The time-march method is explicit.

1: Comparison of time-marching methods for a simple dissipation problem. The results shown in Table 8. The main purpose of this exercise is to show the (usually) great superiority of second-order over rst-order time-marching methods. since h = T=N = . it follows that 155 1 . COMPARING THE EFFICIENCY OF EXPLICIT METHODS Then. The above constraint has led to the invention of schemes that omit the function evaluation in the corrector step of a predictor-corrector combination. leading to the so-called incomplete .1 were computed using a simple iterative procedure.4. In this numerical experiment the function evaluations contribute overwhelmingly to the CPU time. in addition to the constraints given in Section 8. 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. a complete event must be established in order to obtain meaningful statistical samples which are the essence of the solution.4. the method with the highest local accuracy is the most e cient on the basis of the expense in evaluating Fev . Thus the second-order Adams-Bashforth method is much better than the rstorder Euler method.4.1.3 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 Navier-Stokes equations inside a cube with periodic boundary conditions on all sides. we add the following: ~ (~ The number of evaluations of F u t) is xed. ln(0:25)=N . and the number of these evaluations must be kept to an absolute minimum because of the magnitude of the problem. Under these conditions a method is judged as best when it has the highest global accuracy for resolving eigenvectors with imaginary eigenvalues. In this case. 8. On the other hand. In this example we see that.1. for a given global accuracy.8. ( 1 (ln(:25)=N ))N =:25 < 0:005 where 1 is found from the characteristic polynomials given in Table 7. and the fourth-order Runge-Kutta method is the best of all.

that more e cient methods will result from the omission of the second function evaluation. in this case. The presumption is. CHOICE OF TIME-MARCHING METHODS predictor-corrector methods. However. the time step must be twice that of a one{evaluation method so h = 2h1. 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 k-evaluation methods used to get to the same value of T if 8 evaluations are allowed. The second and fourth order RK methods are 2. of course.38 and 6. so that for these methods h = h1 . The . the global amplitude and phase error for kevaluation methods applied to systems with pure imaginary -roots can be written1 Era = 1 .8. However.1. and the number of evaluations. in order to arrive at the same time T after K evaluations. the power to which 1 is raised to arrive at the nal destination decreases see the Figure below. N . one evaluation being used for each step. The Gazdag. after K evaluations. An example is the method of Gazdag. This is the key to the true comparison of time-march methods for this type of problem. In general. Notice that as k increases. and AB2 schemes are all 1-evaluation methods.156 CHAPTER 8. . given in Section 6. respectively.4) 1 See Eqs. 6. Basically this is composed of an AB2 predictor and a trapezoidal corrector.relation for the method is shown as entry 10 in Table 7. However. For a 1evaluation method the total number of time steps. In order to discuss our comparisions we introduce the following de nitions: Let a k-evaluation method be de ned as one that requires k evaluations of ~ (~ F u t) to advance one step using that method's time interval.39. K . etc. notice also that as k increases. For a 2-evaluation method N = K=2 since two evaluations are used for each step. j 1(ik!h1 )jK=k (8. the time span required for one application of the method increases. the derivative of the fundamental family is never found so there is only one evaluation required to complete each cycle. h. For a 4-evaluation method the time interval must be h = 4h1 . Let h1 be the time interval advanced in one step of a one-evaluation method. Let K represent the total number of allowable Fev . leapfrog. are the same.and 4-evaluation methods.

K tan k " 157 (8. The rst three of these are one-evaluation methods and the last one is a four-evaluation method.4. and RK4. 2 . Phase error in degrees.3e.3:8 . For example. We idealize to the case where = i! and set ! equal to one. the results shown in Table 8. exact = 1.1 Er! = !T . 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. Gazdag.8. 7. 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. The event must proceed to the time t = T = 10. leapfrog. Using this.2. Table 8. see Fig. However. Amplitude. by some rather simple calculations2 made using Eqs. On the basis of global error the Gazdag method is not superior to RK4 in either amplitude or phase.2: Comparison of global amplitude and phase errors for four methods. we nd. although.2:4 :45 :12 !h1 = :2 50 . 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 one-evaluation method h1 = T=K .4 and 8. AB2. K leapfrog AB2 Gazdag RK4 !h1 = :1 100 1:0 1:003 :995 :999 !h1 = :2 50 1:0 1:022 :962 :979 a.9:8 1:5 1:5 b. in terms of phase error (for which it was designed) it is superior to the other two methods shown.:96 .5. COMPARING THE EFFICIENCY OF EXPLICIT METHODS . 8. We consider two maximum evaluation limits K = 50 and K = 100 and choose from four possible methods.5) 1 (ik!h1 )]imaginary 1 (ik!h1 )]real # Consider a convection-dominated event for which the function evaluation is very time consuming. 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. The 1 root for the Gazdag method can be found using a numerical root nding routine to trace the three roots in the -plane.0. and the fact that ! = 1. K leapfrog AB2 Gazdag RK4 !h1 = :1 100 .

We have de ned these h as parasitic eigenvalues. is h = 0:02. If h is chosen so that all h in the ODE eigensystem fall inside this circle the integration will be numerically stable.6) We will assume that our accuracy requirements are such that su cient accuracy is obtained as long as j hj 0:1. 8. Numerical solutions. 100h)nx11 + c2(1 . The coupled equations in real space are represented by u1(n) = c1 (1 . 100h)nx21 + c2(1 .6. the time histories of the two solutions would appear as a rapidly decaying function in one case.5 Coping With Sti ness 8. consider the mildly sti system composed of a coupled two-equation set having the two eigenvalues 1 = . The time step limit based on stability.1. of course. We will also assume that c1 = c2 = 1 and that an amplitude less than 0. We rst run 66 time steps with h = 0:001 in order to resolve the 1 term. 8. these h are sti .1 Explicit Methods The ability of a numerical method to cope with sti ness can be illustrated quite nicely in the complex h plane. and a relatively slowly decaying function in the other. 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. Analytical evaluation of the time histories poses no problem since e. h)nx22 + (PS )2 (8. which is determined from 1. CHOICE OF TIME-MARCHING 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 time-accurate calculation of a convection-dominated problem.5.5 and 7. If some h fall outside the small circle but stay within the stable region. but stable.158 CHAPTER 8. Stability boundaries for some explicit methods are shown in Figs. If uncoupled and evaluated in wave space.1. Also shown by the small circle centered at the origin is the region of Taylor series accuracy. This de nes a time step limit based on accuracy considerations of h = 0:001 for 1 and h = 0:1 for 2 .100 and 2 = .001 is negligible.1 0) as shown in Fig. With this time step the . For a speci c example.100t quickly becomes very small and can be neglected in the expressions when time becomes large. 7. 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. Numerical evaluation is altogether di erent. In this case the trace forms a circle of unit radius centered at (. h)nx12 + (PS )1 u2(n) = c1 (1 . Let us choose the simple explicit Euler method for the time march.

with h = 0:02.001 and can be neglected.8.. its in uence in the remaining 70 . In both intervals the desired solution is second-order accurate and well resolved.001. We choose the trapezoidal method. Since this is also a oneroot method. It follows that the nal numerical solution to the ODE is now represented in real space by n 1 . 100h)n. 50(0:1)]= 1 + 50(0:1)] = 0:666 .6 but this time using an unconditionally stable implicit method for the time march. After 66 steps. 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. which in just 10 steps at h = 0:1 ampli es those terms by 109. the maximum step size that can be taken in order to maintain stability..e. we would like to change the step size to h = 0:1 and continue. but that is not the point of this exercise). below 0. the amplitude of the 1 term (i. we simply replace the Euler with the trapezoidal one and analyze the result. far outweighing the initial decrease obtained with the smaller time step. 50 h u1(n) = c1 1 + 50h x11 + c2 !n 1 . about 339 time steps have to be computed in order to drive e. 0:5h nx + (PS ) 1 1 + 0:5h 12 ! 1 . Thus the total simulation requires 405 time steps. 100h)n) is less than 0. 8.e. We would then have a well resolved answer to the problem throughout the entire relevant time interval. its in uence on the coupled solution is negligible at the end of the rst 70 steps. However.7) In order to resolve the initial transient of the term e. In fact.100t . It is true that in the nal 69 steps one -root is 1 . 50 h u2(n) = c1 1 + 50h x21 + c2 ! 8. Now let us re-examine the problem that produced Eq. After 70 time steps the 1 term has amplitude less than 0.4b. this is not possible because of the coupled presence of (1 .e. since (0:666 )n < 1. (It is true that for the same accuracy we could in this case use a larger step size because this is a second-order method. and this has no physical meaning whatsoever. 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. To drive the (1 . (1 . Its behavior in the h plane is shown in Fig. (1 . 0:5h nx + (PS ) 2 1 + 0:5h 22 ! (8. Hence the 1 term can now be considered negligible. 7. h)n term to zero (i.5. and. we need to use a step size of about h = 0:001.. COPING WITH STIFFNESS 2 159 term is resolved exceedingly well.001 and that of the 2 term (i. However.5.9361.001.2 Implicit Methods 1 . h)n) is 0. with 69 steps required to reduce the amplitude of the second term to below 0.t to below 0.001).

is recommended. For preliminary investigations it is often the best method to use for mildly-sti di usion dominated problems. the Euler method is extremely easy to program and requires very little arithmetic per step. It is important to retain a proper perspective on a problem represented by the above example. it should be clear that as the degree of sti ness of the problem increases. Our sole goal is to eliminate it as quickly as possible.1. Therefore. about three times as many.8) When solving a steady problem. 2 = . It has enjoyed remarkable success in many practical problems however. The reader can repeat the above example with 1 = . There is yet another technique for coping with certain sti systems in uid dynamic applications. These explicit methods can be considered as e ective mildly sti stable methods. In the example.5. the conditionally stable Euler method required 405 time steps. Actually. we need an introduction to the theory of relaxation before it can be presented.3 A Perspective 8.10 000. we have no interest whatsoever in the transient portion of the solution. if there are more spurious roots (8. It is clear that an unconditionally stable method can always be called upon to solve sti problems with a minimum number of time steps. . However. as the reduced number of time steps begins to outweigh the increased cost per time step. as compared to about 139 for the trapezoidal method. such as Adams-Bashforth or Runge-Kutta methods.160 CHAPTER 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 x1 + + cm1( m )n xm + + cM 1 ( M )n xM + P:S: 1~ 1~ 1~ n n +c12 ( 1)2 ~ x1 + + cm2 ( m )2 ~ xm + + cM 2 ( M )n xM 2~ n n n +c13 ( 1)3 ~ x1 + + cm3 ( m )3 ~ xm + + cM 3 ( M )3 ~ xM +etc. its behavior for t > 0:07 is identical to that of a stable spurious root. For re ned investigations of such problems an explicit method of second order or higher. 8. although this root is one of the principal roots in the system. which is in the strongly-sti category. This is known as the multigrid method. However. the advantage begins to tilt towards implicit methods. CHOICE OF TIME-MARCHING METHODS steps is even less. The total simulation requires 139 time steps..

Eq.4.7. What is the maximum allowable time step if all but the rst eigenvector are considered parasitic? . Hence the explicit Euler method is a candidate for steady di usion dominated problems. Referring to Fig.3) using 2nd.4. when using the implicit Euler method with local time linearization. When we seek only the steady solution. 2. Therefore the time step can be chosen to eliminate the transient as quickly as possible with no regard for time accuracy. 6. a nite time step may be required until the solution is somewhat close to the steady solution.98. Find and plot the eigenvalues and the corresponding modi ed wavenumbers.and 3rd-order Runge-Kutta methods.4.4.96. For example. the implicit Euler method is the obvious choice for steady problems. Repeat the time-march comparisons for di usion (Section 8. Repeat the time-march comparisons for di usion (Section 8. compare and contrast them with the 3rd. Eq.and 4th-order Adams-Bashforth methods. Considering the stability bounds for these methods (see problem 4 in Chapter 7) as well as their memory requirements.2) and periodic convection (Section 8.1.3) using the 3rd. because of their stability properties. which leads to Newton's method.2) and periodic convection (Section 8. none of the eigenvalues are required to fall in the accurate region of the time-marching method.and 4th-order Runge-Kutta methods. and the fourth-order Runge-Kutta method is a candidate for steady convection dominated problems. all of the eigenvalues can be considered to be parasitic. 8. 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 0:1.7 Problems 1. the di erence being that the order of accuracy is irrelevant. Consider the di usion equation (with = 1) discretized using 2nd-order central di erences on a grid with 10 (interior) points. PROBLEMS 161 the choice of a time-marching method for a steady problem is similar to that for a sti problem. 6.8. However. 3. Among implicit methods. one would like to take the limit h ! 1. 8.

CHOICE OF TIME-MARCHING METHODS .162 CHAPTER 8.

a time derivative is introduced as follows d~ x = A~ x . the unsteady equations are integrated until the solution converges to a steady solution. Although the solution ~ x = A. These methods can be used to compute the time-accurate solution to linear and nonlinear systems of ODE's in the general form d~ u =F ~ (~ u t) (9. they can be used to solve for the steady solution of Eq. Alternatively.1. analyzing.4) dt and the system is integrated in time until the transient has decayed to a su ciently low level.4 has a stable steady solution only if A meets the above condition. 1 163 . we developed a methodology for designing. the ODE given by Eq. and choosing time-marching methods.2) In the latter case. ~ b exists as long as A is nonsingular.Chapter 9 RELAXATION METHODS In the past three chapters. which satis es the following coupled system of nonlinear algebraic equations: ~ (~ F u) = 0 (9. 9. Following a time-dependent path to steady state is possible only if all of the eigenvalues of the matrix A (or .1) dt which arise after spatial discretization of a PDE. 9.3) To solve this system using a time-marching method. The same approach permits a time-marching method to be used to solve a linear system of algebraic equations in the form A~ x =~ b (9.A) have real parts lying in the left half-plane.~ b (9.

This preconditioning has the e ect of multiplying Eq. While they are applicable to coupled systems of nonlinear algebraic equations in the form of Eq. 9. we consider iterative methods which are not time accurate at all. equally important. Using an iterative method. there are well-known techniques for accelerating relaxation schemes if the eigenvalues of CAb are all real and of the same sign. In this chapter.7 depends crucially on the eigenvalue and eigenvector structure of the matrix CAb. does not depend at all on the eigensystem of the basic matrix Ab . and the use of the subscript b will become clear shortly. b C C f b = Ab f b 1 1 1 1 1 (9. C~ f b = A.5 from the left by some nonsingular matrix. our analysis will focus on their application to large sparse linear systems of equations in the form Ab~ u.~ ~ u = CAb ]. for example. 9. If we designate the conditioning matrix by C . and. we seek to obtain rapidly a solution which is arbitrarily close to the exact solution of Eq. In the simplest possible case the conditioning matrix is a diagonal matrix composed of a constant D(b).1 Formulation of the Model Problem 9.7 is .8) which is identical to the solution of Eq.1. C~ fb = 0 Notice that the solution of Eq.164 CHAPTER 9. To use these schemes. ~ . 9. Such methods are known as relaxation methods.~ fb = 0 (9. at each time step of an implicit time-marching method or at each iteration of Newton's method. 9. the . exists. Such systems of equations arise. (9.7) (9.5. RELAXATION METHODS The common feature of all time-marching methods is that they are at least rstorder accurate. In the following we will see that our approach to the iterative solution of Eq. which is given by ~ ~ u1 = A. For example.6) b fb 1 9. the problem becomes one of solving for ~ u in CAb~ u .1 Preconditioning the Basic Matrix It is standard practice in applying relaxation procedures to precondition the basic equation.5. 9.2. provided C .5) where Ab is nonsingular. 9.

2 0 6 6 1 0 .1 0 1 7 6 7 6 .1 0 1 6 6 6 0 .1 32 0 1 7 6 7 6 .2 3 2 .1 3 7 7 7 17 7 17 5 1 0 .6).2 .1 0 > :4 . The result is A = .9) The matrix in Eq. 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. It can rst be conditioned so that the modulus of each element is 1.1 4 1 0 .1 3 7 7 7 7 7 17 5 1 2 . this leads to the approximation 82 > 0 1 > 6 > .1 0 1 u = x 6 2 x> 6 > .1 0 1 6 1 <6 ~ 6 .9. Using a rst-order backward di erence on the right side (as in Section 3.2 0 1 =6 1 0 .1.1 .2 4 1 (9. 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.AT A = 2 1 1 2 0 1 6 6 . 9.1 0 6 6 .1 0 1 6 6 .9 has eigenvalues whose imaginary parts are much larger than their real parts.10) which scales each row. We then further condition with multiplication by the negative transpose.ua 7 6 7 6 0 7 6 7 ~ 6 0 u + 7 6 7 6 15 4 0 2 39 > > 7 > 7 = 7 7 7 > 7 5> > 0 (9. This is accomplished using a diagonal preconditioning matrix 2 1 6 6 60 0 D = 2 x6 6 6 40 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 (9.11) 1 .1 0 7 6 7 6 154 .

12) 9.166 1 1 CHAPTER 9. as de ned in Appendix A.2 1 7 56 4 56 4 0 1 1 . Thus even when the basic matrix Ab has nearly imaginary eigenvalues. symmetric with negative real eigenvalues). 2 .2 0 1 7 6 7 6 7 6 7 1 7 6 1 0 . RELAXATION METHODS 1 1 If we de ne a permutation matrix P and carry out the process P T .2 1 =6 6 1 .13) where A is symmetric negative de nite. we will thoroughly investigate some simple equations which model these structures.1 0 1 7 6 6 7 6 07 0 .2 1 6 6 1 ..2 0 1 7 6 6 07 1 0 .AT A ]P (which just reorders the elements of A and doesn't change the eigenvalues) we nd 2 0 6 6 0 6 6 0 6 6 40 1 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 7 6 . In the remainder of this chapter. 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 .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.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. We do not necessarily recommend the use of . The classical methods will usually converge if Ab is diagonally dominant.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 .2 1 6 6 1 .2 4 3 7 7 7 7 7 17 5 (9. and the classical relaxation methods can be applied. ~ f =0 (9.AT b as a preconditioner we simply wish to show that a broad range of matrices can be preconditioned into a form suitable for our analysis.1. the conditioned matrix . as given in Appendix B. We will consider the preconditioned system of equations having the form A~ . 1 .AT b Ab is nevertheless symmetric negative de nite (i.e.1 which has all negative real eigenvalues.

14) The above was written to treat the general case.7. FORMULATION OF THE MODEL PROBLEM 1 167 accurate when we later deal with ODE formulations. A = CAb and ~ f = C~ fb (9. (b (9.17) dt x ~c) contains the boundary conditions and ~ where (b g contains the values of the source term at the grid nodes. we obtain B (1 .2 1) x ~b = ~ ~c) f g .2 1)~ ~c) . Note that the solution of Eq. If we consider a Dirichlet boundary condition on the left side and where ~ f= xf either a Dirichlet or a Neumann condition on the right side. 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 steady-state solution 2 2 2 2 (9. we nd d~ u = 1 B (1 .16) @x which is the one-dimensional form of the Poisson equation.18) Choosing C = x I .15) @ u = g(x) (9.13. Introducing the threepoint central di erencing scheme for the second derivative with Dirichlet boundary conditions. 9. is guaranteed to exist because A is nonsingular. In this case 2 Ab = 1 B (1 . ~ f . In the notation of Eqs.2 1)~ = ~ f (9.19) ~b.1. ~ u + (b g (9. ~ = A.5 and 9. 9. then A has the form A=B 1~ b1 2 2 2 .9.

H . The iteration count is designated by the subscript n or the superscript (n). and much of the remaining material is devoted to this case.1 is easily obtained from the matrix resulting from the Neumann boundary condition given in Eq.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 . RELAXATION METHODS ~ b = . 9.24) Hence it is clear that H should lead to a system of equations which is easy to solve.24 using a diagonal conditioning matrix.21 for ~ n gives ~ n = I + H . We attempt to do this in such a way. H . ~ f (9. The error at the nth iteration is de ned as ~ en ~ n . The converged solution is designated ~ 1 so that ~ 1 = A. 3. A (9.22) 9. 9.1 (9.2. and the Error +1 +1 1 1 1 Solving Eq. A]~ n .25) 1 1 .20) Note that s = .168 CHAPTER 9. The matrix H is independent of n for stationary methods and is a function of n for nonstationary ones. that it is directly applicable to two.21) where H is some nonsingular matrix which depends upon the iterative method.2 . ~ f = G~ n .2 Classical Relaxation 9. or at least easier to solve than the original system. ~ 1 = ~ n .2. ~ f 1 (9. A. ~ n = A~ n . ~ f (9.2 .23) where G I + H. however.2 The Converged Solution. the Residual. A tremendous amount of insight to the basic features of relaxation is gained by an appropriate study of the one-dimensional case.2 or .2 s]T s = . ~ f (9.and three-dimensional problems.

9. which we designate as m . There results the relation between the error and the residual A~ en . and its eigenvalues.29 is satis ed using the new values of j and j . 9.3 The Classical Methods Point Operator Schemes in One Dimension Let us consider three classical relaxation procedures for our model equation B (1 . Nonstationary processes in which H (and possibly C ) is varied at each iteration are discussed in Section 9.27) Finally. ~ rn = 0 (9.2. and use the de nition in Eq.22. CLASSICAL RELAXATION 169 where ~ 1 was de ned in Eq.f i (9. This operator comes about by choosing the value of jn such that together with the old values of j. 9. the j th row of Eq. This operator is a simple extension of the point-Jacobi method which uses the most recent update of j. it is not di cult to show that ~ en = G~ en +1 (9.29) as given in Section 9.1. f i (9. The residual at the nth iteration is de ned as ~ rn A~ n .30) j j j 2 j.29 is satis ed. The Gauss-Seidel method is n = 1h n + n . 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.5. and the old value of j . G is referred to as the basic iteration matrix.26. ~ f (9.9.26) Multiply Eq.2.25 by A from the left. determine the convergence rate of a method. Hence the j th row of Eq. The Point-Jacobi method is expressed in point operator form for the one-dimensional case as n = 1 h n + n . 9. 9. The method of successive overrelaxation (SOR) ( +1) ( ) 1 ( ) +1 ( +1) 1 +1 ( +1) ( +1) 1 ( ) +1 1 1 +1 .2. and j . 9.28) Consequently.31) j j j 2 j.2 1)~ = ~ f (9. .

35) dt = A . 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.34) Then the point-Jacobi method is obtained with H = . being lower triangular. the portion of the matrix below the diagonal. about which we have already learned a great deal.21 leads to an interpretation of relaxation methods in terms of solution techniques for coupled rst-order ODE's. ~ f] (9. U .D. such that A=L+D+U (9.170 CHAPTER 9. h j ij n = jn + ! ~j . C A ~ . A~ . ! f (9.33) j j 2 j. One can easily see that Eq. ! ) n + ! n . ~ f b = H . It is usually expressed in two steps as h n .3 The ODE Approach to Classical Relaxation The particular type of delta form given by Eq. The Gauss-Seidel method is obtained with H = . but it can also be written in the single line n = ! n + (1 .36) b dt 1 1 9. 9. then perhaps it would be better to move further in this direction. f This is equivalent to d~ = H .1 The Ordinary Di erential Equation Formulation .f i ~j = 1 jn + 2 . L. and the portion above the diagonal. jn (9.32) j ( +1) 1 ( ) +1 ( +1) ( ) ( ) where ! generally lies between 1 and 2. 9. D.3. 9.(L + D). which certainly meets the criterion that it is easy to solve. 9.21 results from the application of the explicit Euler time-marching method (with h = 1) to the following system of ODE's: ~ ~ ~ Hd (9.13 into its diagonal. RELAXATION METHODS is based on the idea that if the correction produced by the Gauss-Seidel method tends to move the solution toward ~ 1. which is also easy to solve.

In general it is assumed that any or all of the eigenvectors could have been given an equally \bad" excitation by the initial guess. 9. even though no true time accuracy is involved.37) | x + {z+ cM e xM} + 1 error where what is referred to in time-accurate analysis as the transient solution. The eigenvalues are xed by the basic matrix in Eq.36 are independent of t. that is. As we shall see.3. In a stationary method. Suppose the explicit Euler method is used for the time integration. an iteration process has been decided upon). Hence the numerical solution after n steps of a stationary relaxation method can be expressed as (see Eq. if all of the eigenvalues of H . the preconditioning matrix in 9. they are not changed throughout the iteration process. A is nonsingular). A have negative real parts (which implies that H . 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. THE ODE APPROACH TO CLASSICAL RELAXATION 1 1 1 171 In the special case where H . the solution can be written as M t~ ~ ~ = c e 1 t~ (9.39) The initial amplitudes of the eigenvectors are given by the magnitudes of the cm . It is clear that. ~ f 1 (9. the only free choice remaining to accelerate the removal of the error terms is the choice of h. 9. Assuming that H . 9. H . so that we must devise a way to remove them all from the general solution on an equal basis. ~ f is also independent . Throughout the remaining discussion we will refer to the independent variable t as \time".38) which is the solution of Eq. A depends on neither ~ u nor t.13. The generalization of this in our approach is to make h. H and C in Eq. a constant for the entire iteration. These are xed by the initial guess. and the eigenvectors of H A are linearly independent.36.28) n ~n = c ~ | x (1 + h) + 1 1 1 n + cm~ xm (1 + {z mh) + error + cM ~ xM (1 + n M h) } +~ 1 (9. 6. A has been chosen (that is. 1 . is now referred to in relaxation analysis as the error. 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. For this method m = 1 + m h. the \time" step.7.35. of t. and the secondary conditioning matrix in 9. given by 1 1 1 1 ~ 1 = A. then the system of ODE's has a steady-state solution which is approached as t ! 1. The eigenvalues are xed for a given h by the choice of time-marching method.9.

Insert the model equation 9. ~ f (9. 1 +1 TABLE 9.2 1)~ n .2. this is not in the spirit of our study.B (1 .30.32 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. ! (9. If we impose this constraint and further restrict ourselves to banded tridiagonals with a single constant in each band. since multiplication by the inverse amounts to solving the problem by a direct method without iteration.3.2 1) .3 is that all the elements above the diagonal (or below the diagonal if the sweeps are from right to left) are zero.2.3 can be identi ed using the entries in Table 9. Then ~ ~ ~ Hd (9. The constraint on H that is in keeping with the formulation of the three methods described in Section 9.2 1) gives the correct answer in one step.43 THAT LEAD TO CLASSICAL RELAXATION METHODS ! 0 1 1 Method Equation 6:2:3 6:2:4 6:2:5 1 Point-Jacobi 1 h i Gauss-Seidel 2= 1 + sin M + 1 Optimum SOR . ~ n) = B (1 . However. equal to 1. (1 .35.41) n = n+h n with a step size. Now let us study these methods as subsets of ODE as formulated in Section 9. we are led to 2 0)(~ . 9. let us use for the numerical integration the explicit Euler method 0 (9.1. h. 9.43) n n n f where and ! are arbitrary. ~ B (.2 1)~ .172 CHAPTER 9.1.29 into the ODE form 9. With this choice of notation the three methods presented in Section 9. ~ ) = B (1 .3. RELAXATION METHODS 9. f As a start. 9.31 and 9.B . We arrive at H (~ n .2 1) since then multiplication from the left by .40) dt = B (1 .42) +1 +1 It is clear that the best choice of H from the point of view of matrix algebra is .2 ODE Form of the Classical Methods The three iterative procedures de ned by Eqs.1: VALUES OF and ! IN EQ.

46) This equation is preconditioned in some manner which has the e ect of multiplication by a conditioning matrix C giving A~ . !h f (9. and ~ 1 A. are identi ed for the one-dimensional case by Eq.47) An iterative scheme is developed to nd the converged. The point operator that results from the use of the explicit Euler scheme is ! ! n = ! n + ! (h . solution of the set of ODE's ~ ~ . or steady-state.49) where ~ en is the transient.45) j j. (. 1 . 9.~ Hd = A f (9. 1) n + !h n . 9. our purpose is to examine the whole procedure as a special subset of the theory of ordinary di erential equations. ~ f (9. and SOR.~ fb = 0 (9.44) dt ! and appear from it in the special case when the explicit Euler method is used for its numerical integration. The three classical methods. 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. (!h . Point-Jacobi. However. ) n . or error.2 1)~ .44 and Table 9.1. the three methods are all contained in the following set of ODE's d~ = B . Gauss-Seidel. j 2 j.48) dt This solution has the analytical form ~n = ~ en + ~ 1 (9. ~ f =0 (9.4 Eigensystems of the Classical Methods The ODE approach to relaxation can be summarized as follows. 2 0) B (1 .4.9. ~ f is the steady-state solution. 2 2 j 2 j 1 ( +1) ( +1) 1 ( ) 1 ( ) ( ) +1 This represents a generalization of the classical relaxation techniques. The basic equation to be solved came from some time-accurate derivation Ab~ u. In this light.

9. Fig.32.174 CHAPTER 9. 9. A matrix for the three methods. the maximum j m j is obtained with m = M .2 1)~ xm = m B (. This amounts to solving the generalized eigenvalue problem A~ xm = m H~ xm (9. Note that for h > 1:0 (depending on M ) there is the possibility of instability.2.1 2 ).30.3. and f = 0. j j = j M j and the best possible convergence rate is achieved: 1 1 j m jmax = cos M + 1 (9. is shown in Fig.50) In this section. This relation can be plotted for any h. the ODE matrix H . we take M = 5 for the matrix order. j m j > 1:0. 9. i. e = 2=!.52) ! xm The generalized eigensystem for simple tridigonals is given in Appendix B. To illustrate the behavior.54) .44. the maximum j mj is obtained with m = 1. This special case makes the general result quite clear. . For h < 1.51) for the special case 2 0)~ (9. A reduces to simply B ( 1 2 . b = . Fig. the asymptotic convergence rate is determined by the eigenvalue m of G ( I + H .4.53) m = . The eigensystem can be determined from Appendix B.1 The Point-Jacobi System 1 If = 0 and ! = 1 in Eq. 9.1. The three special cases considered below are obtained with a = 1. B (1 . The eigenvalues are given by the equation m (9.e. 9.2.31. A) having maximum absolute value. To obtain the optimal scheme we wish to minimize the maximum j m j which occurs when h = 1.1 + cos M +1 m = 1 2 ::: M The . the optimum stationary case. d = .1 since both d and f are zero.relation for the explicit Euler method is m = 1 + m h. It is also instructive to inspect the eigenvectors and eigenvalues in the H . RELAXATION METHODS 1 Given our assumption that the component of the error associated with each eigenvector is equally likely to be excited. 9. 9. we use the ODE analysis to nd the convergence rates of the three classical methods represented by Eqs. c = 1. and 9. The plot for h = 1.2 and for h > 1. Thus Convergence rate j mjmax m=1 2 1 M (9.

1 + 1 2 = . 3=2 1 2 3 4 5 (9.1 + 23 = . 3=2 7 7 6 7 6 ~ ~ 7 6 x =6 0 x = 1 7 6 7 6 p 7 p 7 6 6 3=2 5 5 4 p 4 .1 . The rst 5 eigenvectors are simple sine waves. 3 )h]n~ x 2 1 )h]n~ x + c 1 . Thus after 500 iterations the error content associated with each eigenvector is reduced to no more than 0. ~ 1 = c 1 . 2 1 2 2 p 1 . we obtain j mjmax = 0:9971. For M = 5. 3=2 7 .0:5 = .1 7 x = 6 p1 7 x = 6 p 7 6 6 6 4 3=2 7 5 4 0 7 5 5 4 .4. 9.p3=2 7 1 1=2 .1:0 = . 9.23 times its initial level.39. the numerical solution written in full is 5 p ~ n .9. from Eq. A are given by m ~ xj = (xj )m = sin j M j = 1 2 ::: M (9.1:866 From Eq.55) +1 This is a very \well-behaved" eigensystem with linearly independent eigenvectors and distinct eigenvalues. (1 . 3=2 3 2 p 3 2 3=2 1=2 p p 6 6 7 6 7 6 . the eigenvectors can be written as 2 1=2 3 2 3 2 p3=2 3 1 7 p p 6 7 6 7 6 6 6 6 3=2 7 3=2 7 0 7 6 7 6 7 7 6 ~ ~ ~ 6 7 6 7 6 0 7 x = 6 . EIGENSYSTEMS OF THE CLASSICAL METHODS 1 175 For M = 40.57) 4 = . the eigenvectors of H .53 1 2 3 = . 23 = .1:5 p (9. (1 .1 .1.1 = . 1 2 = . 3=2 7 1=2 .0:134 = . Again from Appendix B.56) The corresponding eigenvalues are.

One can show that the H .0 0.0 -2.176 1. h = 1 M = 5.2 The Gauss-Seidel System If and ! are equal to 1 in Eq. (. AGS . 9.44.59) 1 which can be studied using the results in Appendix B. the matrix eigensystem evolves from the relation B (1 . RELAXATION METHODS h = 1.0 Figure 9.4.0 m=5 CHAPTER 9. + c 1 . is AGS B .1: The relation for Point-Jacobi.0 m=1 Values are equal σm m=4 m=2 Μ=5 0.2 1) = 1 .2 1)~ xm = m B (. (1 + 23 )h]n~ x 4 5 )h]n~ x 3 5 (9. A matrix for the Gauss-Seidel method. (1 3 4 n~ ) h ] x + c 1 .1 2 0)B (1 . (1 + 1 2 p + c 1 .0 m=3 -1.58) 9.2.1 2 0)~ xm (9.

Exceeds 1.0 h < 1.0 -2.2: The relation for Point-Jacobi.072 1.0 h > 1.0 h 0.3: The relation for Point-Jacobi.0 0.0 De cr ea h si ng σm ng si Figure 9.0 -1. h = 0:9 M = 5.0 ng si ea cr In In cr ea si σm ng h h M = 5 0. .0 -2.4.9.0 λ mh ∼ ∼ 1. h = 1:1 M = 5.0 at h De cr ea -1.0 177 1.072 / Ustable h > 1.0 λmh Figure 9. EIGENSYSTEMS OF THE CLASSICAL METHODS Approaches 1.0 0.

. . . . They are no longer symmetrical. The m with the largest amplitude is obtained with m = 1. . and there are (M . 3 = 4 1 = 2 5 6 7 6 7 . . .63) 3 2 h~ i 7 6 h~ i 7 6 7 6 7 6 2 3 .3=4 1=2 2 6 7 6 7 6 7 0 1=8 . . 0 1=2M M The eigenvector structure of the Gauss-Seidel ODE matrix is quite interesting.64) . If M is odd there are (M + 1)=2 distinct eigenvalues with corresponding linearly independent eigenvectors. The eigenvectors are quite unlike the Point-Jacobi set.61) m = .3=4 1=2 3 6 7 6 7 0 1 = 16 1 = 8 .4. . .62) +1 M +1 2 The . . 7 6 ~ 1 X AGS X = JGS = 6 7 7 6 6 7 6 6 7 5 4 4 ~ 1 57 ~ 1 1 2 3 3 3 (9. they do not represent a common family for di erent values of M . For M = 40.principal vectors.1 + cos ( M +1 2 and the corresponding eigenvectors are given by j. The Jordan canonical form for M = 5 is j m jmax = cos M + 1 2 (9. .1 1=2 3 1 6 7 0 . RELAXATION METHODS 2 . m m M +1 ~ xm = cos M sin j m = 1 2 : : : (9. The equation for the nondefective eigenvalues in the ODE matrix is (for odd M ) M +1 m ) m = 1 2 : : : (9.23. the optimum stationary case. . Furthermore. . 250 iterations are required to reduce the error component of the worst eigenvector by a factor of roughly 0. . 9. j m jmax = 0:9942.60) 6 7 0 1 = 32 1 = 16 1 = 8 .relation for h = 1. . 6 . 4 . . Hence the convergence rate is 2 1 CHAPTER 9. is shown in Fig. 3 = 4 1 = 2 4 6 7 (9.7 5 . .. 1)=2 defective eigenvalues with corresponding 178 Since this is the square of that obtained for the Point-Jacobi method. the error associated with the \worst" eigenvector is removed in half as many iterations.. producing waves that are higher in amplitude on one side (the updated side) than they are on the other.

34 1 2 3 3 1 1 2 2 (9.9.65) 6 6 6 6 6 4 9=16 7 5 4 . EIGENSYSTEMS OF THE CLASSICAL METHODS 1.66) .0 179 h = 1. ~ 1 = c (1 . For M = 5 they can be written 2 3 2 p 3 2 3 2 3 2 3 3 = 2 1 = 2 1 0 0 7 p 6 7 6 7 6 7 6 7 6 6 6 6 6 6 3=4 7 3=4 7 07 2 7 0 7 6 7 6 7 6 7 6 7 6 7 ~ 6 7 ~ 6 7 ~ 6 7 ~ 6 7 ~ 6 x = 6 3=4 7 x = 6 p0 7 x = 6 0 7 x = 6 . The eigenvectors and principal vectors are all real.3=4 ) = . h = 1:0 M = 5.4: The relation for Gauss-Seidel.0 -2. linked to (5) Jordan block The numerical solution written in full is thus ~ n .p3=16 7 5 407 5 4 0 7 5 4 .0 λmh Figure 9.0 0. h )n~ 4 x h )n~ x + c (1 .0 σm Μ=5 2 defective λ m 0.4.4 7 5 9=32 0 0 1 . 3=32 1 2 3 4 5 The corresponding eigenvalues are = .1 (4) Defective.1=4 = .1 7 x = 6 4 7 7 (9.0 -1.

2x + x x . !) + ! pm] = pm = cos m =(M + 1)] 2 2 2 1 If ! = 1. + c (1 . n .69) where zm = 4(1 . 2 x + x x . The H . the ODE matrix is B . h) + c h 2! (1 . (. 2x + x .4.2 1). The generalization to any M is fairly clear. h) + c h 1! (1 . !) + ! p = 0 ! is optimum for the stationary case. and the following conditions hold: 1. h)n. 2 x x x 6 6 x . . A matrix for the SOR method.180 CHAPTER 9. 3. is 1 1 2 . ~ x 3 4 1 5 2 2 3 4 5 + c (1 . h)n~ x 5 1! 1 4 5 (9. If M is odd.44. 1) n n . Two eigenvalues are real. 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 .1 + !pm 2 2 m = 1 2 :::M 2 2 1 2 (9. the system is Gauss-Seidel. 2x 4 .1 x 0)B (1 . 2 x + x x .3 The SOR System 1 If = 1 and 2=! = x in Eq. If M is even.2x x 0 0 6 . If ! is chosen such that 4(1 . 2x x 0 6 1 6 6 . 2x 0 0 0 x 4 4 3 7 7 7 7 (9. 9. 2x + x x .1 x 0)B (1 .2 1). !) + ! pm < 0 zm and m are complex. one of the remaining eigenvalues is real and the others are complex occurring in conjugate pairs.2 + x 1 . 2x + x x . the remaining eigenvalues are complex and occur in conjugate pairs. ASOR B . 2. One can show that this can be written in the form given below for M = 5. h)n + c h n (1 . equal and defective.2x + x x .2x + x x . (.68) 7 7 5 Eigenvalues of the system are given by m + zm = . h) ~ x + c (1 . If 4(1 . RELAXATION METHODS " # n n ( n .67) 9.

9. 1. j mjmax = 0:8578. the . the optimum value for the stationary case. Hence the convergence rate is j m jmax = !opt . For odd M . and if h = 1. there are two real eigenvectors and one real principal vector.9. the optimum value of ! may have to be determined by trial and error.4. Hence the worst error component is reduced to less than 0.4=3 (9.5.70) !opt = 2= 1 + sin M + 1 ~ xm m = = m. much faster than both GaussSeidel and Point-Jacobi. The remaining linearly independent eigenvectors are all complex. 2p2i)=9 = . In practical applications. This illustrates the fact that for optimum stationary SOR all the j m j are identical and equal to !opt .73) 6 7 6 7 6 7 6 7 p p 6 7 6 7 6 7 6 7 4 1=6 5 4 13 5 4 p 3(5 ip 2)=54 5 4 0 5 1=18 6 1=9 3(7 4i 2)=162 The corresponding eigenvalues are = .1 j .2=3 (2) Defectivep linked to = .71) Using the explicit Euler method to integrate the ODE's.23 times its initial value in only 10 iterations. 1 (9.74) 2 1 2 34 5 1 1 3 4 5 !opt p + i p . h + h m . For M = 5 they can be written 2 3 2 3 2 p3(1 3 2 3 )=2 1 = 2 .1 m sin 2 where m j M +1 (9.relation reduces to that shown in Fig.(10 + 2 2i)=9 = .72) !opt = 2= 1 + sin M + 1 For M = 40.(10 . and the bene t may not be as great. 6 1 p p 6 6 6 6 6 7 6 7 6 7 6 7 3(1 i 2)=6 7 1=2 7 9 7 0 7 6 7 6 7 6 7 6 7 ~ 7 ~ 6 7 ~ 6 7 ~ 6 7 1 = 3 16 0 1 = 3 x =6 x = x = x = (9. EIGENSYSTEMS OF THE CLASSICAL METHODS One can easily show that the optimum ! for the stationary case is and for ! = !opt 181 (9. m = 1 . p m= m 2 m 2 1 2 q .

9. h = 1.0 0. This process changes the Point-Jacobi method to Richardson's method in standard terminology. b f b .75) 9. The numerical solution written in full is ~n .182 1. It is much simpler. The nonstationary form of Eq. (10 + 2 2i)h=9] ~ x c (1 . but it can greatly a ect the convergence rate. RELAXATION METHODS h = 1.0 -2. For the Gauss-Seidel and SOR methods it leads to processes that can be superior to the stationary methods.0 λmh Figure 9. M = 5. 4h=3)n~ x 1 2 1 2 3 2 3 4 5 4 5 1 (9.5: The relation for optimum stationary SOR.5 Nonstationary Processes In classical terminology a method is said to be nonstationary if the conditioning matrices. 2h=3)n. 2p2i)h=9]n~ x n c 1 . to study the case of xed H and C but variable step size. however. (10 . In our ODE approach this could also be considered and would lead to a study of equations with nonconstant coe cients.0 -1.39 is 1 .0 2 real defective λ m σm 2 complex λ m 1 real λ m Μ=5 0. 2h=3)n~ x p c 1 . 2h=3)n + c nh(1 . ]~ x c (1 . h. ~1 = + + + + c (1 .0 CHAPTER 9. are varied at each time step. This does not change the steady~ state solution A. H and C .

1= m . Consider one of the error terms taken from M N X Y ~ eN ~ N .78) where Pe signi es an \Euler" polynomial.76. In the annihilation process mentioned after Eq. Now we pose a less demanding problem. We will see that a few well chosen h's can reduce whole clusters of eigenvectors associated with nearby 's in the m spectrum. NONSTATIONARY PROCESSES N Y ~N = c ~ x (1 + hn ) + 1 1 183 + cm~ xm M hn ) + ~ 1 N Y n=1 n=1 1 (1 + m hn ) + + cM ~ xM N Y n=1 (1 + (9. Mathematically stated. This leads to the concept of selectively annihilating clusters of eigenvectors from the error terms as part of a total iteration process.79) treating it as a continuous function of the independent variable .1= m for m = 1 2 : : : M .9. we considered making the error exactly zero by taking advantage of some knowledge about the discrete values of m for a particular case. This is the basis for the multigrid methods discussed in Chapter 10. 9.76) where the symbol stands for product. the error term can theoretically be completely eliminated in M steps by taking hm = . 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). for m = 1 2 M . 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. we seek b max a j(Pe)N ( )j = minimum with(Pe)N (0) = 1 (9.5. However.80) . Since hn can now be changed at each step. the eigenvalues m are generally unknown and costly to compute. It is therefore unnecessary and impractical to set hm = . ~ 1 = cm~ xm (1 + m=1 n=1 m hn ) (9. Now focus attention on the polynomial (Pe)N ( ) = (1 + h )(1 + h ) 1 2 (1 + hN ) (9.77) and write it in the form cm~ xm Pe( m ) cm~ xm N Y n=1 (1 + m hn ) (9.

9. cos (2n . 1) n = 1 2 :::N (9. and it leads to the nonstationary step size choice given by ( " #) 1 = 1 . The three values of h which satisfy this problem are 1) hn = 2= 3 .1 y 1 and 2 2 (9. 9. ampli ed by the coe cients cm (1 + m hn). It is 2 .82) 1 y + qy .86) . a b This problem has a well known solution due to Markov. + ( . .1 using only 3 iterations. b TN . As an example for the use of Eq. a.84. 9. The error in the relaxation process represented by Eq. b TN a. xm .2 .76 is expressed in terms Q ~ of a set of eigenvectors. qy .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. RELAXATION METHODS a.184 CHAPTER 9.81) TN (y) = cos(N arccos y) are the Chebyshev polynomials along the interval .85) " #! (9. 6 (9. 1 N + 1 y . Eq. ) cos (2n . b ! . In relaxation terminology this is generally referred to as Richardson's method.83) 2 are the Chebyshev polynomials for jyj > 1. With each eigenvector there is a corresponding eigenvalue. ! (Pe)N ( ) = where (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 . 1 N TN (y) = 2 (9. let us consider the following problem: Minimize the maximum error associated with the eigenvalues in the interval .

This was derived from Eq. namely that =1 0 1 M N B 1 + hn Y 2 ~ N = X cm~ xm B @ m n 1. 2 hn under the same constraints as before.89) would result.87) (9.6 and we see that the amplitudes of all the eigenvectors associated with the eigenvalues in the range . 9. was used to integrate the basic ODE's.41.88) where n p p o T (3) = 3 + 8] + 3 .5. 9.2 . 8] =2 99 3 3 3 A plot of Eq. 1 2 hn =1 =1 1 mC C A + ~1 m (9.9. This calls for a study of the rational \trapezoidal" polynomial. 9. If instead the implicit trapezoidal rule n+1 = n+ is used. 9.92) . 3) h = 4=(6 .90) b max a j(Pt)N ( )j = minimum .91) @ A 1 n 1 . Eq. 9.37 on the condition that the explicit Euler method.87 is given in Fig. with (Pt)N (0) = 1 (9. The values of h used in Fig.1 have been reduced to less than about 1% of their initial values.6 are h = 4=(6 . Pt: 0 1 1 N Y B 1 + 2 hn C C (Pt )N ( ) = B (9. the nonstationary formula 1 h( 0 + 0 ) n 2 n +1 (9.76. 9. p 0) h = 4=(6 + 3) 1 2 3 p Return now to Eq. NONSTATIONARY PROCESSES and the amplitude of the eigenvector is reduced to (Pe) ( ) = T (2 + 3)=T (3) 3 3 3 185 (9.

015 −0.7 0. RELAXATION METHODS 1 0.6 −1.6 −0.2 −1 λ −0.6 −1.01 0.2 0 0.005 −0.9 0.015 0. minimization over .8 0.4 −0.8 −0.005 (P ) (λ) e 3 0 −0.3 0.01 −0.8 −0.1.4 −1.2 .1 0 −2 −1.8 −1.4 −0.4 0.2 −1 λ −0.2 0.6 −0. .6 (Pe )3 (λ) 0.4 −1.186 CHAPTER 9.6: Richardson's method for 3 steps.02 −2 −1.02 0.2 0 Figure 9.8 −1.5 0.

The values of h used in Fig.93) b hn b This process is also applied to problem 9.85. xk ) = (A + A )xk . 1 1 1 . f ~ 0 1 1 2 where G = I + H . Recall that the eigenvalues of H .7 are h = 1 p h = 2 h = 2 ( 1) ( 1) 3 3 1 2 3 9. consider the iterative method (I + A )~ x = (I . . Show that this iterative method can be written in the form H (xk . Using Appendix B. A. f show that 1 ~n = Gn ~ + (I .) Find the numerical values. 3. = N . A )~ x+ b where is a parameter. The error amplitude is about 1/5 of that found for (Pe) ( ) in the same interval of . 9. A for the SOR method with A = B (4 : 1 .6 Problems 1. not just the expressions. A satisfy A~ xm = mH~ xm . PROBLEMS ! 187 The optimum values of h can also be found for this problem. Given a relaxation method in the form ~ H ~n = A~n . Then nd the corresponding j m j values. The results for (Pt) ( ) are shown in Fig.2. b 1 2 2 +1 1 +1 1 2 Determine the iteration matrix G if = . For a linear system of the form (A + A )x = b. 2 =.1=2. 9. but we settle here for the approximation suggested by Wachspress n. a n=1 2 N (9.7. (You do not have to nd H .6.2 1) and ! = !opt.9. Gn)A. 2. A )xn + b (I + A )xn = (I . nd the eigenvalues of H .

6 −1.8 0.8 −0. RELAXATION METHODS 1 0.2 −1 λ −0.5 −2 −2 −1.8 −1.188 CHAPTER 9.2 0 2 x 10 −3 1.4 −0.1.7: Wachspress method for 3 steps.2 .6 (P ) (λ) t 3 0.9 0.6 −0.2 0 Figure 9.5 1 0.4 −1.6 −0.6 −1.5 (Pt )3 (λ) 0 −0. .3 0.5 −1 −1.8 −0.4 −1.2 −1 λ −0. minimization over .8 −1.5 0.4 −0.7 0.1 0 −2 −1.2 0.4 0.

5. u(1) = 1: @ u . use u(x) = 0. the number of iterations. Plot the solution after the residual is reduced by 2. and (d) the 3-step Richardson method derived in Section 9. Determine the asymptotic convergence rate. Iterate to steady state using (a) the point-Jacobi method.6. 6x = 0 @x For the initial condition. (c) the SOR method with the optimum value of !. Use second-order centered di erences on a grid with 40 cells (M = 39). Solve the following equation on the domain 0 x 1 with boundary conditions u(0) = 0. PROBLEMS 189 4. Plot the logarithm of the L -norm of the residual vs. and 4 orders of magnitude.9. 3. (b) the Gauss-Seidel method. Compare with the theoretical asymptotic convergence rate. 2 2 2 .

190 CHAPTER 9. RELAXATION METHODS .

1) then the corresponding eigenvectors are ordered from low to high space frequencies.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. That is. The eigenvalues and eigenvectors are given in Sections 4. ~ (10. This has a rational explanation from the origin of the banded matrix.1. There are many variations of the process and many viewpoints of the underlying theory.2) @x and recall that 1 B (1 .2 and 4.2 1). the space-frequency (or wavenumber) of the corresponding eigenvectors also increase.3) xx~ = x x 191 2 2 2 2 1 .3. Notice that as the magnitudes of the eigenvalues increase.3. The viewpoint presented here is a natural extension of the concepts discussed in Chapter 9. 10.1 Motivation 10.2 1)~ = X 1 D(~ ) X . respectively.3. Note that @ sin(mx) = .m sin(mx) (10.1 Eigenvector and Eigenvalue Identi cation with Space Frequencies Consider the eigensystem of the model matrix B (1 . if the eigenvalues are ordered such that j j j j 1 2 2 j Mj (10.

m . error components corresponding to high space frequencies will be reduced more quickly than those corresponding to low space frequencies. It is also true. the correlation of large magnitudes of m with high space-frequencies is to be expected for these particular matrix operators. and X ~ . the opera~ tion 1 x D( ) represents the numerical approximation of the multiplication of the appropriate sine wave by the negative square of its wavenumber.m m << M (10. However. . 10. a sine synthesis. the complete counterexample of the above association is contained in the eigensystem 1 1 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 correlation is not necessary in general.1. ~ represents a sine transform. The classical point-Jacobi method does not share this property.4. 9. for example. this method produces the same value of j j for min and max .2. We have seen that X . This is the key concept underlying the multigrid process.4) m= M +1 x Hence. One nds that 1 M + 1 . of the Gauss-Seidel method and. of the Richardson method described in Section 9. In fact. 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. This is consistent with the physics of di usion as well. by design. Therefore. 2 2 2 2 2 10.5. exactly the opposite for B ( 2 2 behavior.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 . MULTIGRID where D(~ ) is a diagonal matrix containing the eigenvalues. However. from Appendix B. as shown in Fig. For this matrix one nds.2 + 2 cos m .192 1 CHAPTER 10.1. the property can be restored by using h < 1. As we saw in Section 9. This is to be expected of an iterative method which is time accurate.

6 is used. 3. the vector ~ f b represents the boundary conditions and the forcing function. of the matrix are all real and negative. The eigenvectors associated with the eigenvalues having largest magnitudes can be correlated with high frequencies on the di erencing mesh. The iterative procedure used greatly reduces the amplitudes of the eigenvectors 1 j jmax and j jmax. The problem is linear. 9.5) where the matrix formed by the product CAb has the properties given above. and 1 is a vector representing the desired exact solution.7) where A CAb (10.6) Recall that the ~ used to compute ~ r is composed of the exact solution ~ 1 and the error ~ e in such a way that A~ e . ~ f b] (10. as in the examples given by Eq. The basic assumptions required for our description of the multigrid process are: 1. THE BASIC PROCESS 193 basic properties. or it can be \contrived" by further conditioning.5.10. This form can be arrived at \naturally" by simply replacing the derivatives in the PDE with di erence schemes. associated with eigenvalues in the range between 2 These conditions are su cient to ensure the validity of the process described next.27.~ r=0 (10. The m are fairly evenly distributed between their maximum and minimum values. as in the example given by Eq.11. the residual. Having preconditioned (if necessary) the basic nite di erencing scheme by a procedure equivalent to the multiplication by a matrix C . m . if ~ any.8) . We do not attempt to develop an optimum procedure here. we are led to the starting formulation C Ab ~ 1 . The eigenvalues. 2. At the end of the three steps we nd ~ r. In Eq. We start with some initial guess for ~ 1 and proceed through n iterations making use of some iterative process that satis es property 5 above. 3. 5. but for clarity we suppose that the three-step Richardson method illustrated in Fig. ~ f b] = 0 (10.2. 4. where ~ r = C Ab~ . 10. 9.

partitions the A matrix to form 1 2 3 # "~ # A 1 A2 " ~ e r 6 7 e 4 5 ~ = ~e ro A 3 A4 e o (10.14) . 10. MULTIGRID (10. For a 7-point 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 5 6 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 6 07 7 6 7 0 76 6 6 07 7 6 7 6 7 0 76 6 07 56 4 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 = P~ e ~ eo (10.7 from the left by P and.~ e CHAPTER 10. assumption 4 ensures that the error has been smoothed. P ]~ e = P~ r (10. 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).12) 1 The operation PAP . we can write PA P .13) Notice that eo = ~ re ee + A ~ A~ 1 2 (10. In addition. one containing the odd entries. Next we construct a permutation matrix which separates a vector into two parts.7 for ~ e then ~1 = ~ .10) } Combining our basic assumptions.9) Thus our goal now is to solve for ~ e. and the eigenvalues of the Richardson process in the form: M= M X Y X Y ~ e = cm~ xm ( m hn)] + c m~ xm ( mhn)] 2 3 3 m=1 n=1 m=M=2+1 | very low amplitude {z n=1 (10.194 If one could solve Eq.11) Multiply Eq. We can write the exact solution for ~ e in terms of the eigenvectors of A. since a permutation matrix has an inverse which is its transpose. 10. we can be sure that the high frequency content of ~ e has been greatly reduced (about 1% or less of its original value in the initial guess).

If the boundary conditions are Dirichlet. ea and eb are zero.14 reduces to A~ ee + A A0 ~ ee = ~ re 1 2 2 (10. it is reasonable to suppose that the odd points are the average of the even points. It is that there is some connection between ~ ee and ~ eo brought about by the smoothing property of the Richardson relaxation procedure. For example 1 (e + e ) e 2 a 1 (e + e ) e 2 1 (e + e ) or ~ eo = A0 ~ ee (10. and one can write for the example case 2 1 0 0 6 1 1 1 0 A02 = 2 6 6 40 1 1 3 7 7 7 5 is an exact expression.19) or Ac~ ee . Since the top half of the frequency spectrum has been removed. ~ re = 0 where Ac = A + A A0 ] 1 2 2 .18) (10.10. Hence. in this formulation.15) e 2 1 (e + e ) e b 2 1 2 3 2 4 5 4 6 2 7 6 (10. notice that.16) 0 0 1 With this approximation Eq. Finally. the averaging of ~ e is our only approximation. 10.17) (10. 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. no operations on ~ r are required or justi ed. It is also important to notice that we are dealing with the relation between ~ e and ~ r so the original boundary conditions and forcing function (which are contained in ~ f in the basic formulation) no longer appear in the problem. At this point we make our one crucial assumption.2. no aliasing of these functions can occur in subsequent steps.

2 ::: .20) and Eq.18 gives z }| 2 . 10.1 are met.2 .2 .16 changes to 2 1 0 03 16 1 1 07 6 7 A0 = 2 6 (10.2 .2 . the example in Eq. 10.23) 40 1 17 5 0 0 2 1 2 .2 6 6 6 6 6 6 .2 1 1 1 1 1 1 .1.2 3 7 7 7 2 3 17 7 A1 A2 7 7 7 7 =6 4 5 7 7 A3 A4 7 7 7 5 (10. and. 1 PAP = 6 6 6 1 6 6 1 6 6 4 . 2 6 .55 for Dirichlet conditions. 1 1 =2 7 6 7 7 5 = 4 1=2 . ea = e . If the original A had been B (7 : 1 . MULTIGRID The form of Ac. All of them go through zero on the left (Dirichlet) side. 10. 10. the matrix on the coarse mesh. 9. In the particular case illustrated in Fig. the interpolation formula in Eq.1 { 3 7 5 (10. all of the properties given in Section 10. For Neumann conditions. It is easy to show that the high space-frequencies still correspond to the eigenvalues with high magnitudes.1. B. 10. 1) xjm = sin j 2M + 1 m=1 2 M (10.1]T . our 7-point example would produce 2 . the eigenvector structure is di erent from that given in Eq.196 CHAPTER 10. In the present case they are given by " !# (2 m .21) If the boundary conditions are mixed Dirichlet-Neumann.22) and are illustrated in Fig. If Neumann conditions are on the left. A in the 1-D model equation is B (1 ~ b 1) where ~ b = .2 4 A1 . However.1 1=2 1=2 . is completely determined by the choice of the permutation matrix and the interpolation approximation. eb is equal to eM . The eigensystem is given by Eq.19.2 .2 { 2 z }| 3 1 1 7 5+6 4 1 1 A2 1 1 { 3 7 5 2 z 2 }| 1 16 1 1 6 6 24 1 1 A 0 1 { Ac 3 }| z 2 .2 1).15 must be changed. in fact. and all of them re ect on the right (Neumann) side. When eb = eM .2 1 1 1 .

" We will continue with Dirichlet boundary conditions for the remainder of this Section. Therefore.1 in the lower right element).2 1)~ e=~ r on the 1 ~ ~ ne mesh to 2 B (1 . The permutation matrix remains the same and both A and A in the partitioned matrix PAP .2.2 1 . In order to examine this relationship. THE BASIC PROCESS 197 X Figure 10.1 1=2 .2 1) the eigenvalues and eigenvectors are m m j=1 2 M (10. 10. (10.1: Eigenvectors for the mixed Dirichlet{Neumann case. 1 1 = 2 7 6 7 7 5 = 4 1=2 . we have reduced the problem from B (1 . which will provide us with the solution ~ 1 using Eq. cos m = sin j m = 1 2 M M +1 M +1 1 1 .9. and thus ~ e. we must now determine ~ the relationship between ee and ~ e.10.2 4 A1 . 2 6 . Given ~ ee computed on the coarse grid (possibly using even coarser grids). At this stage.1=2 { 3 1=2 7 5 (10.2 { z }| 3 2 1 1 7 5+6 4 1 1 A2 1 1 { 3 7 5 2 z 2 }| 1 6 16 1 1 6 4 1 1 2 A 0 2 { Ac 3 }| z 2 . In order to complete the process.24) which gives us what we might have \expected.26) ~ x m = .25) For A = B (M : 1 . 10. we can compute ~ eo using Eq. we can construct the coarse matrix from 1 1 2 4 }| z 2 . Recall that our goal is to solve for ~ e.15. we need to consider the eigensystems of A and Ac: A = X X. Ac = Xc cXc. are unchanged (only A is modi ed by putting .2 1)ee = re on the next coarser mesh.

32) 2 3 1=( c) 6 0 7 6 7 6 = Xc 6 . MULTIGRID Based on our assumptions. 7 4 .31) 2 3 1 6 07 6 7 .e. with M = 51. xc) c ~ c Xc (10. re = Xc . In addition. the most di cult error mode to eliminate is that with m = 1. then Mc = (M .2 1 . The exact solution on the coarse grid satis es . Now let us consider the case in which all of the error consists of the eigenvector component ~ x . 1)=2 is the size of Ac. one can easily see that (~ xc) coincides with ~ x at every second point of the latter vector. 7 5 0 (10.. 7 5 0 (10. = . that is..2 1) are ( c) = . ( c) approaches 2 .28) For M = 51 (Mc = 25).27) For example. As M increases. corresponding to ~ = . cos M2+ 1 (~ xc) = sin j M2+ 1 j = 1 2 Mc (10. 1 . ~ e=~ x . 6 = Xc c 6 . Then the residual is ~ r = A~ x = ~ x (10. i.. The eigenvalue and eigenvector corresponding to m = 1 for the matrix Ac = 1 2 B (Mc 1 .0:003649.0:007291 = 1:998 . cos M + 1 x = sin j M + 1 j=1 2 M (10. we obtain ( c) = . it contains the even elements of ~ x.30) since (~ xc) contains the even elements of ~ x .198 CHAPTER 10.33) 1 1 1 1 1 1 1 1 1 . 7 4 . If we restrict our attention to odd M . (~ ~ ee = A.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 ~ re = (~ xc) 1 1 1 1 (10.

40) which is precisely the matrix appearing in Eq. in addition to interpolating ~ ee to the ne grid (using Eq. which gives Ab = x B (M : 1 .15).2. obviously. The remaining steps required to complete an entire multigrid process are relatively straightforward.37) c 4 In our case. THE BASIC PROCESS = ( c) 1 1 199 (~ xc ) 1 1 1 (~ 2 xc ) 1 (10. The problem to be solved on the coarse grid is the same as that solved on the ne grid. and usually is. quite important but they are not discussed here. in each case the appropriate permutation matrix and the interpolation approximation de ne both the down.2 1) C x B (Mc : 1 . The details of nding optimum techniques are.2 1) comes from a discretization of the di usion equation.2 1) = x c c 4 c c 2 2 2 (10. However. carried to even coarser grids before returning to the nest level. but they vary depending on the problem and the user. Thus we see that the process is recursive. since xc = 2 x.36) 1 B (M : 1 . we must multiply the result by 2.34) (10.2 1) and the preconditioning matrix C is simply 2 (10.39) Applying the discretization on the coarse grid with the same preconditioning matrix as used on the ne grid gives. This is equivalent to solving or 1A ~ e =~ r 2 ce e (10.37.2 1) = 1 B (M : 1 . 10.10.2 1)~ ee = ~ re (10. x B (M : 1 .and up-going paths. 10.35) Since our goal is to compute ~ e=~ x .38) C= x I 2 (10. . the matrix A = B (M : 1 . The reduction can be.

f ~.42) and H is de ned as in Chapter 9 (e. Transfer (or restrict) ~ r to the coarse grid: ~ r =R ~ r (10.g. starting with ~ = ~n.44) In our example in the preceding section.f ~ ~ r = A~ . Note that the coarse grid matrix denoted A2 here was denoted Ac in the preceding section. 10.21)." Some form of weighted restriction can also be used. Solve the problem A ~ e =~ r on the coarse grid exactly: ~ e = A. A (10. f ~ = AGn1 ~n . the restriction matrix is 2 3 0 1 0 0 0 0 0 R =6 (10. 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. Gn1 ) A. Gn1 ) A. the rst three rows of the permutation matrix P in Eq. (1) (2) 2 (1) 1 2 1 3. 1. f ~ (10. .41) 1 (1) 1 (1) 1 1 1 where G = I + H.3 A Two-Grid Process CHAPTER 10. This gives ~ = Gn1 ~n + (I . MULTIGRID ~.46) 1 2 See problem 1 of Chapter 9.200 10. ~ r 2 (2) (2) (2) 2 1 (2) 2 (10. f (10.11. Call the result ~ .43) 1 1 1 1 1 (1) (1) (1) 1 1 1 1 1 1 2. which can be easily We now describe a two-grid process for the linear problem A ~ = f generalized to a process with an arbitrary number of grids due to the recursive nature of multigrid.45) 40 0 0 1 0 0 07 5 0 0 0 0 0 1 0 that is. This type of restriction is known as \simple injection.. AGn1 A. Eq. 9. Perform n iterations of the selected relaxation method on the ne grid. Next compute the residual based on ~ : ~ = AGn1 ~n + A (I .

I ~ e +1 (1) 1 (2) 2 (10. A TWO-GRID PROCESS 2 2 201 Here A can be formed by applying the discretization on the coarse grid. R A ]Gn2 2 3 2 3 3 1 3 2 2 2 2 2 3 3 2 3 (10. 4. A = 4 generalization to a multigrid procedure with more than two grids occurs.50 by replacing A.10. I A. apply the two-grid process recursively. 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. Combining these steps.15.51) In this expression n is the number of relaxation steps on grid 2. f ~ +1 1 2 2 1 2 1 1 1 2 2 1 2 1 1 1 1 (10. Otherwise. with (I . In the 1 B (Mc : 1 . I and R are the transfer operators between grids 2 and 3. R A]Gn1 A. Transfer (or prolong) the error back to the ne grid and update the solution: ~n = ~ . I A. If this is the coarsest grid in the sequence.47) 3 7 7 7 7 7 7 7 7 7 7 7 5 In our example.2 1). I A. 10. G )A.40). one obtains ~n = I .3. Extension to four or more grids proceeds in similar fashion. I . I A.49) (10.50) Thus the basic iteration matrix is I .48) which follows from Eq. 10. R A]Gn1 1 2 2 1 2 1 1 1 1 The eigenvalues of this matrix determine the convergence rate of the two-grid process. It is at this stage that the preceding example (eq. solve exactly. and A is obtained by discretizing on grid 3. R A]Gn1 ~ n . where 2 2 3 2 G = I . . 10. The basic iteration matrix for a three-grid process is found from Eq. . f ~ + A.

Determine the asymptotic convergence rate. (b) the 3-step Richardson method derived in Section 9. Solve exactly on the coarsest grid. the number of iterations. and 4 orders of magnitude. 2 .202 10. 3.51.4 Problems CHAPTER 10. Plot the solution after the residual is reduced by 2. Calculate the theoretical asymptotic convergence rate and compare. Derive Eq. 10. Repeat problem 4 of Chapter 9 using a four-grid multigrid method together with (a) the Gauss-Seidel method. 2. Plot the logarithm of the L -norm of the residual vs. MULTIGRID 1.5.

this can be insu cient. Thus the numerical approximation must contain some inherent dissipation to limit the production of high-frequency modes. In this Chapter. we have neglected viscous e ects. due to the limited grid resolution which is practical. we have emphasized the second-order centered-di erence approximations to the spatial derivatives in our model equations. In a real physical problem.Chapter 11 NUMERICAL DISSIPATION Up to this point. to the production of shock waves. such as the Euler and Navier-Stokes equations. especially at high Reynolds numbers. leading. there can be a continual production of high-frequency components of the solution. However. Although numerical approximations to the Navier-Stokes equations contain dissipation through the viscous terms. Since the addition of numerical dissipation is tantamount to intentionally introducing nonphysical behavior. for example.. the production of high frequencies is eventually limited by viscosity. when we solve the Euler equations numerically. In processes governed by nonlinear equations. the eigenvalues of the associated circulant matrix (with periodic boundary conditions) are pure imaginary. i. 203 . unless the relevant length scales are resolved. We have seen that a centered approximation to a rst derivative is nondissipative. some form of added numerical dissipation is required in the numerical solution of the Navier-Stokes equations as well. Therefore. 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. it must be carefully controlled such that the error introduced is not excessive.e.

uj+1)] 2 x (11.1. Consider the following point operator for the spatial derivative term .2) The second form shown divides the operator into an antisymmetric component (.(1 + )uj.1) produces Re( m ) > 0. )uj+1] = . uj+1)=2 x. ) The eigenvalues of this matrix are m = .1 .a x = 2. For periodic boundary conditions the corresponding matrix operator is .1 + 2uj .1 + uj+1) + (.a( x u)j = 2. In either case. Hence the forward di erence operator is inherently unstable while the centered and backward operators are inherently stable. our choice of di erencing scheme produces nonphysical behavior. With 6= 0.1) with periodic boundary conditions. cos 2M m + i sin 2M for m = 0 1 : : : M .uj.uj. We proceed next to show why this occurs.1 + 2 uj + (1 .a @u @t @x (11. 1 If a is positive. NUMERICAL DISSIPATION @u = .a x m 1 .ax Bp(. the forward di erence operator ( = . the roles are reversed. . and the backward di erence operator produces Re( m) < 0. the solution will either grow or decay with time.1 + 2uj .ax . When Re( m) = 0.204 11.uj.1+ uj+1)=2 x and a symmetric component (. A backward di erence operator is given by = 1 and a forward di erence operator is given by = . The antisymmetric component is the second-order centered di erence operator.1 One-Sided First-Derivative Space Di erencing We investigate the properties of one-sided spatial di erence operators in the context of the biconvection model equation given by CHAPTER 11. If a is negative. 2 1 . the centered di erence operator ( = 0) produces Re( m ) = 0.uj. the operator is only rst-order accurate.a (.

Substituting this into Eq. when we use a computer to nd a numerical solution of the problem.3 is consistent with Eq. 2 ) i x. i 3 + 4 or r = .4 of the form u(x t) = ei xe(r+is)t provided r and s satisfy the condition r + is = . THE MODIFIED PARTIAL DIFFERENTIAL EQUATION 11.ia . 11. x can be small but it is not zero.3 as the modi ed partial di erential equation. We are lead to the expression 2 ! 3 2u ! 3 @3u ! 4 @4u ! @ x x x2 @x2 + 3 @x3 .2.(a+ 2 )t] u=e (11.1. 11. (a + 2 ) The solution is composed of both amplitude and phase terms.1. 12 @x4 + : : :5 j 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. Consider the simple linear partial di erential equation @u = .a @u + @ 2 u + @ 3 u + @ 4 u (11. 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. 2 ( . 11. Notice that Eq. We refer to Eq.11.3 is excited to some degree. 2 .1 gives @u = .a @u + a x @ 2 u .2 The Modi ed Partial Di erential Equation ( xu)j = 2 1 x 42 x @u @x j . 2 ) s = . 205 First carry out a Taylor series expansion of the terms in Eq. since the two equations are identical when x ! 0. 11. 11.5) | {z } e | {z } amplitude phase .2 to Eq. However. Under these conditions there is a wave-like solution to Eq. 11. This means that each term in the expansion given by Eq.4) @t @x @x2 @x3 @x4 Choose periodic boundary conditions and impose an initial condition u = ei x.2. 2 ( . Thus . 11. We proceed next to investigate the implications of this concept. 11. 11.

any departure from antisymmetry in the matrix di erence operator must introduce dissipation (or ampli cation) into the modi ed PDE. that the same is true for any centered di erence approximation of a rst derivative. NUMERICAL DISSIPATION It is important to notice that the amplitude of the solution depends only upon and . Eq. 8uj. If the wave speed a is positive. One can easily show. 4uj. Therefore. 4uj+1 + uj+2)] (11. the choice of a forward di erence scheme ( = . By examining the term governing the phase of the solution in Eq.1) is equivalent to deliberately adding a destabilizing term to the PDE.2 . Therefore. 11. Biased schemes use more information on one side of the node than the other. the coe cients of the even derivatives in Eq. the choice of a backward di erence scheme ( = 1) produces a modi ed PDE with . 6uj. 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). Furthermore. The resulting spatial operator is not one-sided but it is dissipative. This we refer to as dispersion. 2 > 0 and hence the amplitude of the solution decays. the numerical phase speed is dependent upon the wavenumber . one could choose = 1=2 in Eq. 11. 11. we see that the speed of propagation is a + 2 . Our purpose here is to investigate the properties of one-sided spatial di erencing operators relative to centered di erence operators.1 + 6uj . 11.1 + 8uj+1 . by using the antisymmetry of the matrix di erence operators.a x2 =6. uj+2) 12 x + (uj. this operator produces fourth-order accuracy in phase with a third-order dissipative term.2.2 . and the phase depends only on a and . As a corollary. a third-order backward-biased scheme is given by ( xu)j = 6 1 x (uj. Therefore. The symmetric component approximates x3 uxxxx=12.1 + 3uj + 2uj+1) = 1 (uj. Under the same condition. Note that the use of one-sided di erencing schemes is not the only way to introduce dissipation.5. Any symmetric component in the spatial operator introduces dissipation (or ampli cation).206 CHAPTER 11. . For example. the phase speed of the numerical solution is less than the actual phase speed.6) The antisymmetric component of this operator is the fourth-order centered di erence operator. Referring to the modi ed PDE. This is tantamount to deliberately adding dissipation to the PDE.2 .3 we have = .4. the coe cients of the odd derivatives.

2uj + uj . the corresponding fully-discrete matrix operator is 0 2 !23 !2 2 !2 31 1 ah ah ah 1 ah ah ~ un+1 = Bp @ 2 4 x + x 5 1 . Next we consider a method which introduces dissipation independent of the sign of the wave speed. For periodic boundary conditions.8) @t @x @t2 @x2 Now replace the space derivatives with three-point centered di erence operators. 1 For j ah to unity and hence x j 1 all of the eigenvalues have modulus less than or equalah the method is stable independent of the sign of a. It is equal to the ratio of the distance travelled by a wave in one time step to the mesh spacing. 2 x (uj+1 . It is a fully-discrete nite-di erence scheme. known as the Lax-Wendro method.7) First replace the time derivatives with space derivatives according to the PDE (in @u this case.a @u (11. x + x 5A ~ un The eigenvalues of this matrix are !2 ah 1 . .1 ) This is the Lax-Wendro method applied to the linear convection equation. This explicit method di ers conceptually from the methods considered previously in which spatial di erencing and time-marching are treated separately. backward di erencing must be used if the wave speed is positive. x 2 4.3 The Lax-Wendro Method 207 In order to introduce numerical dissipation using one-sided di erencing. THE LAX-WENDROFF METHOD 11.1) + 2 x (u(jn +1 . i ah sin 2 m m =1. Thus @ 2 u = a2 @ 2 u @u = . cos 2 m .3. The quantity j x j is known as the Courant (or CFL) number. the linear convection equation @u @t + a @x = 0). There is no intermediate semi-discrete stage. giving !2 1 ah 1 ah ( n ) ( n ) ) (n) (n) ( n +1) ( n ) (11.11.9) uj = uj . uj. x M x M for m = 0 1 : : : M . Consider the following Taylor-series expansion in time: 1 h2 @ 2 u + O(h3) + u(x t + h) = u + h @u @t 2 @t2 (11. and forward di erencing must be used if the wave speed is negative.

depending on the sign). The two methods di er when applied to nonlinear hyperbolic systems. this approach leads to (n) (n) u ~(jn+1) = u(jn) . Cn) @x3 The dissipative term is proportional to 2 4u a2 h x2 (1 . ah (~ (n+1) ~(n+1) )] u(jn+1) = 1 u ~ (11. a8h ( x2 . a ( x2 . a2 h2) @ 4 u + : : : @t @x 6 @x3 8 @x4 This is derived by substituting Taylor series expansions for all terms in Eq. a2h2 ) @ = .8.10) un+1 = 2 n ~n+1 + hu If we use rst-order backward di erencing in the rst stage and rst-order forward di erencing in the second stage. Hence MacCormack's method has the same dissipative and dispersive properties as the Lax-Wendro method. 6 ( x .11) j j j 2 x uj+1 . 6 (1 .208 CHAPTER 11. a2h2 ) @ 3 u . . Therefore. a h ) @x3 = . a2 h ( x2 . Recall that the odd derivatives on the right side lead to unwanted dispersion and the even derivatives lead to dissipation (or ampli cation. NUMERICAL DISSIPATION The nature of the dissipative properties of the Lax-Wendro scheme can be seen by examining the modi ed partial di erential equation. these should be applied alternately. presented in Chapter 6: u ~n+1 = un + hu0n 1 u +u ~0n+1] (11. however. 11. A closely related method is that of MacCormack. uj. the leading error term in the Lax-Wendro method is dispersive and proportional to 3u 2 3u a @ a x @ 2 2 2 2 . 11. The two leading error terms appear on the right side of the equation. 1 Or vice-versa for nonlinear problems. For the linear convection equation.1 a dissipative second-order method is obtained.9 and converting the time derivatives to space derivatives using Eq. which is given by @u + a @u = . C 2 ) @ 4 u . n @x4 @x4 8 This term has the appropriate sign and hence the scheme is truly dissipative as long as Cn 1. ah x (uj .1) (n) + u (n+1) . Recall MacCormack's timemarching method. u which can be shown to be identical to the Lax-Wendro method.

then a+ = 0 and a.1x a(.11. 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. the wave speeds can be both positive and negative. For example.4 Upwind Schemes 209 In Section 11. we saw that numerical dissipation can be introduced in the spatial di erence operator using one-sided di erence schemes or. We can rewrite Eq.uj.12) @t @x where we do not make any assumptions as to the sign of a. If a 0.a( xu)j = 2.5. some form of splitting is required. This is avoided in the Lax-Wendro scheme. The above approach to obtaining a stable discretization independent of the sign of a can be written in a di erent. as a result of their superior exibility. we present the basic ideas of ux-vector and ux-di erence splitting. more generally. This is the basic concept behind upwind methods.12 as @u + (a+ + a.e.13) This approach is extended to systems of equations in Section 11. but entirely equivalent. This is achieved by the following point operator: . then a = a 0 and a = 0. With this approach. we see that a stable discretization is obtained with = 1 if a 0 and with = . ( 0) term forward di erence. In the next two sections. 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. When a hyperbolic system of equations is being solved. In this section. uj+1)] (11.1 if a 0. From Eq.2. 11. these are of mixed sign. ) @u = 0 a jaj a = @t @x 2 + . that is.4. In order to apply one-sided di erencing schemes to such systems. 11. = a 0. For more subtle aspects of implementation and application of such techniques to . we present two splitting techniques commonly used with upwind methods.1 + 2uj . When the ow is subsonic. Alternatively.uj. These are by no means unique. manner. UPWIND SCHEMES 11. Now for the a+ ( 0) term we can safely backward di erence and for the a. the direction of the one-sided operator (i. if a 0. schemes in which the numerical dissipation is introduced in the spatial operator are generally preferred over the Lax-Wendro approach. a.1 + uj+1) + jaj(.1. However.. Consider again the linear convection equation: @u + a @u = 0 (11. the eigenvalues of the ux Jacobian for the onedimensional Euler equations are u u + a u . by adding a symmetric component to the spatial operator.

17) 2 2 With these de nitions. Premultiplying by X and inserting the product @x .5 that a linear. where += (11. A.j j (11.1Xw + @X . constant-coe cient. @w = 0 (11. and a forward di erence if the wave speed is negative. hyperbolic system of partial di erential equations given by @u + @f = @u + A @u = 0 (11. NUMERICAL DISSIPATION nonlinear hyperbolic systems such as the Euler equations. @w term. 1 X X in the spatial terms gives @Xw + @X +X . and the wi's are the characteristic variables.14) @t @x @t @x can be decoupled into characteristic equations of the form @wi + @wi = 0 (11.4.X .1 Flux-Vector Splitting Recall from Section 2. 11. we need to apply a backward di erence if the wave speed. = . the reader is referred to the literature on this subject. contains the negative eigenvalues. To accomplish this. let us split the matrix of eigenvalues.16) +j j . i.19) . into two components such that = ++ . We can use backward di erencing for the + @w @x term and forward di erencing for the .1Xw = 0 @t @x @x (11.18) @t @x @t @x @x The spatial terms have been split into two components according to the sign of the wave speeds. In order to apply a one-sided (or biased) spatial di erencing scheme. are the eigenvalues of the Jacobian matrix.15) i @t @x where the wave speeds. We can now rewrite the system in terms of characteristic variables as @w + @w = @w + + @w + . .210 CHAPTER 11. + contains the positive eigenvalues and . i. is positive.

we are in e ect solving the characteristic equations in the desired manner.X .22) f . the Jacobians of the split ux vectors are required.21) (11. we obtain A. = A.24) Thus by applying backward di erences to the f + term and forward di erences to the f .. A++ has eigenvalues which are all positive. When an implicit time-marching method is used. f is also equal to Au as a result of their homogeneous property.4. one must nd and use the new Jacobians given by A++ = @f @u + (11. (11.11. as discussed in Appendix C.1 and recalling that u = Xw.25) A. has all negative eigenvalues. 2 With these de nitions A+ has all positive eigenvalues. term. 6= A. u @u + @f + + @f . f = Au. u = 0 @t @x @x Finally the split ux vectors are de ned as f + = A+u and we can write (11. . and A. In the nonlinear case. and A. For the Euler equations. = @f @u . Note that f = f+ + f.20) @u + @A+u + @A. the de nition of the split uxes follows directly from the de nition of the ux.. = 0 (11. @u @u Therefore.26) For the Euler equations. has all negative eigenvalues. (11. This approach is known as ux-vector splitting. @f + 6= A+ @f .1 (11. UPWIND SCHEMES With the de nitions2 211 A+ = X +X .23) @t @x @x In the linear case. = X .

29) i wi )R if i < 0 where the subscript i indicates individual components of w and g. constant-coe cient.27) @t @x The ux vector associated with this form is g = w.1X (w + w ) + 1 X j jX .212 CHAPTER 11.1X after and j j to obtain 1 X X . This is achieved with 1 (w ) + (w ) ] + 1 j j (w ) .34) . as a function of the states to the left and right of the interface.32) L R L R 2 and thus 1 f^j+1=2 = 1 ( f (11. we consider the numerical ux at the interface between nodes j and j + 1.31) L + wR ) + j j (wL . as in Eq. is given by 1 (g(w ) + g(w )) g ^j+1=2 = 2 (11. which is nondissipative. wL and wR.30) i i L i R i R 2 i iL or 1 g ^j+1=2 = 1 ( w (11.2 Flux-Di erence Splitting Another approach. We again begin with the diagonalized form of the linear. as in Chapter 5. the uxes must be evaluated at cell boundaries. more suited to nite-volume methods. Now. wR ) 2 2 Now. uR ) 2 2 where jAj = X j jX .28) L R In order to obtain a one-sided upwind approximation. w ) Xg ^j+1=2 = 2 (11. The centered approximation to gj+1=2. In a nite-volume method. is known as ux-di erence splitting. respectively.1X (w .1 (11. we require ( (wi)L if i > 0 (^ gi)j+1=2 = i( (11. we premultiply by X to return to the original variables and insert the product X .19. NUMERICAL DISSIPATION 11. hyperbolic system of equations @w + @w = 0 (11. g ^j+1=2. 11.4.33) L + fR ) + jAj (uL . (w ) ] (^ gi)j+1=2 = 2 (11.

11.36) L R 2 j+1=2 L R if Aj+1=2 satis es fL . 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. In order to resolve this. of this chapter. If the eigenvalues of A are all positive.5 Arti cial Dissipation We have seen that numerical dissipation can be introduced by using one-sided 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 (dissipation-free) operator. Eq.11. In the linear. jAj = . jAj = A if they are all negative.39) p L+p R R where u and H = (e + p)= are the velocity and the total enthalpy per unit mass. u ) f^j+1=2 = 2 (11. ARTIFICIAL DISSIPATION 213 and we have also used the relations f = Xg. and A = X X . However.1. respectively. 11. In this case.3 ( 11.A.38) uj+1=2 = p + p L R p H +p H L L R Hj+1=2 = (11.35) 0 R if all i s < 0 giving pure upwinding. uR) (11. consider a situation in which the eigenvalues of A are all of the same sign. this leads to an upwind operator which is identical to that obtained using ux-vector splitting. Hence satisfaction of Eq. u = Xw.37 is satis ed by the ux Jacobian evaluated at the Roe-average state given by p LuL + p R uR (11. there is some ambiguity regarding the de nition of jAj at the cell interface j + 1=2. we would like our de nition of f^j+1=2 to satisfy fL if all i0s > 0 f^j+1=2 = f (11.37) For the Euler equations for a perfect gas. constant-coe cient case. 3 Note that the ux Jacobian can be written in terms of u and H only see problem 6 at the end . in the nonlinear case. fR = Aj+1=2 (uL .5.35 is obtained by the de nition 1 (f + f ) + 1 jA j (u .

4uj+1 + uj+2) where is a problem-dependent coe cient. applying x = x x i > 0. This is particularly evident from a comparison of Eqs. s is stable if i 0 and unstable if unstable if i < 0. It is also sometimes referred to as arti cial di usion or arti cial viscosity. uj .uj +1 + 2uj . The appropriate implementation is thus a s (11. 11.36 and 11.34.41) i x = i x + j ij x Extension to a hyperbolic system by applying the above approach to the characteristic variables. A second-order backward di erence approximation to a 1st derivative is given as a point operator by ( xu)j = 1 (uj. as in the previous two sections. this often provides su cient damping of high frequency modes without greatly a ecting the low frequency modes.2 .1 (x (11. 11.214 For example. s u) = (x j 11.2 . With an appropriate value of . s It is common to use the following operator for x (11. Similarly. This symmetric operator approximates x3 uxxxx and thus introduces a third-order dissipative term.15 is stable if i 0 and Applying x = x x a . a and s . this approach can be related to the upwind With appropriate choices of x x approach. 4uj. A more complicated treatment of the numerical dissipation is also required near shock waves and other discontinuities. 11.43.43) x f = x f + x (jAju) where jAj is de ned in Eq. For details of how this can be implemented for nonlinear hyperbolic systems. let CHAPTER 11. The second spatial term is known as arti cial dissipation. NUMERICAL DISSIPATION s u) = .42) x (Au) = x (Au) + x (jAju) or a s (11.1 a u) = uj +1 .1 + 3uj ) 2 x .44) x (uj. 4uj.1 + 6uj . but is beyond the scope of this book.40) (x j j 2 x 2 x a + s to the spatial derivative in Eq. gives a s (11.6 Problems 1. the reader should consult the literature. uj .

11. 11.3 to see how to construct the symmetric and skew-symmetric components of a matrix.2. Using Fourier analysis as in Section 6.44. 3. x for 0 x .6. (See Appendix A. Find the modi ed wavenumber for this operator with = 0 1=12 1=24.6. Plot the real and imaginary parts of x vs. 11. Plot the real and imaginary parts of x vs. Using Fourier analysis as in Section 6. nd j j for the combination of this spatial operator with 4th-order Runge-Kutta time marching at a Courant number of unity and plot vs. .11.1. x for 0 x . nd j j for the combination of this spatial operator with 4th-order Runge-Kutta time marching at a Courant number of unity and plot vs. Using Fourier analysis as in Section 6.2. Find the matrix jAj. PROBLEMS 215 (a) Express this operator in banded matrix form (for periodic boundary conditions). then derive the symmetric and skew-symmetric matrices that have the matrix operator as their sum.4. x for 0 x .6.) (b) Using a Taylor table. x for 0 x . Find the modi ed wavenumber for the operator given in Eq.6.37. 6. 11.2. Form the plus-minus split ux vectors as in Section 11.39 leads to satisfaction of Eq. Consider the hyperbolic system derived in problem 8 of Chapter 2. Show that the ux Jacobian for the 1-D Euler equations can be written in terms of u and H . 4. and 1=48. 2. Show that the use of the Roe average state given in Eqs. 5.6. nd j j for the combination of this spatial operator with 4th-order Runge-Kutta time marching at a Courant number of unity and plot vs. Consider the spatial operator obtained by combining second-order centered differences with the symmetric operator given in Eq. Find the modi ed wavenumber for the rst-order backward di erence operator. x for 0 x .38 and 11. nd the derivative which is approximated by the corresponding symmetric and skew-symmetric operators and the leading error term for each. Plot the real and imaginary parts of x vs. x for 0 x .

216 CHAPTER 11. NUMERICAL DISSIPATION .

When we approach numerical analysis in the light of matrix derivative operators. 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. Advancing to the next time level always requires some reference to a previous one. Factored forms are especially useful for the derivation of practical algorithms that use implicit methods. Matrices can be split in quite arbitrary ways. Time marching methods are valid only to some order of accuracy in the step size. 12. 3.1 The Concept Factored forms of numerical operators are used extensively in constructing and applying numerical methods to problems in uid mechanics.~ f dt 217 (12.Chapter 12 SPLIT AND FACTORED FORMS In the next two chapters. Let us start with the following observations: 1. and a means for analyzing such modi ed forms.1) . and \fractional step". They are the basis for a wide variety of methods variously known by the labels \hybrid". Now recall the generic ODE's produced by the semi-discrete approach d~ u = A~ u. \time split". h. 2. the concept of factoring is quite simple to present and grasp. we present and analyze split and factored algorithms.

h~ f + O(h ) +1 2 1 2 Finally. their numerical stability can be quite di erent. The semi-discrete approach to its solution might be put in the form where Ac and Ad are matrices representing the convection and dissipation terms.2 Factoring Physical Representations | Time Splitting Suppose we have a PDE that represents both the processes of convection and dissipation. respectively and their sum forms the A matrix we have considered in the previous sections. rst-order.2) where A = A + A ] but A and A are not unique.3) 2 or its equivalent h i ~ un = I + hA ] I + hA ] . h A A ~ un . 12. and techniques to carry out their numerical evaluation can have arithmetic operation counts that vary by orders of magnitude. +1 1 2 2 ~ un = I + hA ] I + hA ]~ un . from observation 2 (new data ~ un in terms of old ~ un): 1 2 1 1 +1 ~ un = I + hA + hA ]~ un .3 and 12. from observation 3 (allowing us to drop higher order terms . h~ f + O(h ) +1 1 2 2 1 2 2 1 2 (12.h A A ~ un): (12. From observation 1 (arbitrary splitting of A) : (12. Both of these considerations are investigated later.5) ~ ) + O (h ) ~ un = I + hAd + hAc]~ un + h(bc +1 2 (12. explicit Euler method. Choose again the explicit Euler time march so that d~ u =A~ ~ ~ c u + Ad u + (bc) dt (12. in this sense.4 have the same formal order of accuracy and.218 CHAPTER 12. . h~ f + O(h ) 12. However.~ f dt 1 2 2 and consider the above observations. SPLIT AND FACTORED FORMS d~ u = A + A ]~ u. neither one is to be preferred over the other. For the time march let us choose the simple. Here we seek only to apply to some simple cases the concept of factoring.6) 1 Second-order time-marching methods are considered later.4) Notice that Eqs. Then.

9) I . FACTORING PHYSICAL REPRESENTATIONS | TIME SPLITTING 219 Now consider the factored form ~ un +1 Original Unfactored Terms Higher Order Terms and we see that Eq.12. and a rst-order time-march method is usually unsatisfactory.8. 12. hAd]. but the di usion operator is now implicit. where jjAd jj is some norm of Ad ].8) Factoring can also be useful to form split combinations of implicit and explicit techniques. For example. stability. 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. This time a predictor-corrector interpretation leads to the sequence u ~n I . 12.7 are often applied in a predictor-corrector sequence. their selection is arbitrary.6 with the same order of accuracy is given by the expression ~ un +1 Original Unfactored Terms where in this approximation we have used the fact that 1 2 2 = = ~) I . requiring a tridiagonal solver if the di usion term is central di erenced. We have yet to determine its . I + hAc]~ un + h(bc ~ ) +O(h ) I + hAd + h A ]~ u + h(bc | {z c n } 1 2 (12. hAd ]. 2 We do not suggest that this particular method is suitable for use. In practical applications equations such as 12. In this case one could write 2 = = ~) I + hAd ] I + hAc]~ un + h(bc ~ ) +O(h ) (12.6 have identical orders of accuracy in the time approximation. another way to approximate Eq.7) ~ ) + h Ad Ac~ un + (bc I + hAd + h A ]~ u + h(bc {z c n } | 2 2 | {z } u ~n ~ un +1 +1 = = ~) I + hAc]~ un + h(bc I + hAd]~ un +1 (12. However.7 and the original unfactored form Eq. Therefore. the important aspect of stability has yet to be discussed. as before. = I + hAd + h Ad + if h jjAdjj < 1. on this basis. 12. hAd]~ un +1 +1 ~) = I + hAc]~ un + h(bc = u ~n +1 (12. 12.10) The convection operator is applied explicitly.2.

12) j Mx Mesh indexing in 2-D.10. un = A u + A u + (bc ~ ) + O(h ) c n d n h +1 +1 2 We should mention here that Eq. 3 32 @u = @ u + @ u @t @x @y 2 2 (12. A clear description of a matrix nite-di erence operator in 2.3 Factoring Space Matrix Operators in 2{D Factoring is widely used in codes designed for the numerical solution of equations governing unsteady two. We choose the 3 4 point mesh shown in the Sketch 12. The numbers 11 12 43 represent the location in the mesh of the dependent variable bearing that index. is 4 and My .1 Mesh Indexing Convention 2 2 12.9 can be derived for a di erent point of view by writing Eq. ~) I . 12. Let us study the basic concept of factoring by inspecting its use on the linear 2-D 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. 6 point mesh if the boundary points (labeled in the sketch) were included. 12.12. the number of (interior) y points is 3. SPLIT AND FACTORED FORMS un . Thus u represents the value of u at j = 3 and k = 2.220 CHAPTER 12.3. 3 This could also be called a 5 . the number of (interior) x points. hAd ]un = I + hAc]un + h(bc +1 12.6 in the form Then which is identical to Eq.and three-dimensional ows. In this example Mx . but in these notes we describe the size of a mesh by the number of interior points.11) My k 1 13 23 33 43 12 22 32 42 11 21 31 41 1 (12. 12.and 3-D requires some reference to a mesh.

12. Examples are in Eq. 12.16 part a and b. Examples of the order of indexing for these space vectors are given in Eq. 12. FACTORING SPACE MATRIX OPERATORS IN 2{D 221 12. ( ) ( ) 12. we consider only two: (1). respectively.16 part a and b. consecutively along columns that are consecutive from j = 1 to Mx. we label a data{base composed of x-vectors with U x . Notice that the matrices are not the same although they represent the same derivative operation.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 data-base. In particular. Of these. we use the notation Axx y or Axy y .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. 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 . used in the previous sections. Their structures are similar. (1) and (2) above are referred to as xvectors and y-vectors.12. When the matrix is multiplying a space vector U . To be speci c about the structure. Thus ( ) + ( ) + ( ) ( ) dU = A U + (bc ~) x y dt + (12. consecutively along rows that are themselves consecutive from k = 1 to My . and one composed of y-vectors with U y .14) Axx y = Pxy Axy y Pyx ( ) + ( ) + (12. 4 Notice that .3.15) Ax+y and U . and (2). depending on the data{base chosen for the U it multiplies.3.3. The symbol U by itself is not enough to identify the structure of the data-base and is used only when the structure is immaterial or understood. There are many ways to lay out a data-base.13) where . and they are related by the same permutation matrix that relates U x to U y . Pyx = PT xy = Pxy Now consider the structure of a matrix nite-di erence operator representing 3point central-di erencing schemes for both space derivatives in two dimensions.11 can be written + 4 If it is important to be speci c about the data-base structure. the usual (but ambiguous) representation is given by Ax y . In this notation the ODE form of Eq. are subsets of A and ~ u. which are notations used in the special case of space vectors. however.

matrix operator. central-di erence. central-di erence.16) b: Elements in 2-dimensional. 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 . for 3 4 mesh shown in Sketch 12... 12 22 32 42 13 23 33 43 11 21 31 41 . Data base composed of My x{vectors stored in U x . Entries for x ! x. for both ! . Ax y .222 2 6 x 6 6 6 6 6 6 6 6 6 6 o 6 6 A(xx+)y U (x) = 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4 CHAPTER 12. for 3 4 mesh shown in Sketch 12. a:Elements in 2-dimensional. + ( ) 2 6 o 6 6 6 6 6 6 6 x 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 . for y ! o. Data base composed of Mx y{vectors stored in U y . Entries for x ! x.12. Ax y . for both ! .12. 41 42 43 31 32 33 21 22 23 11 12 13 (12. for y ! o.. . .. .. matrix operator.

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 ( rst-order 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 ~ ) + O(h ) I ; hAxx ; hAyx Unx = Unx + h(bc
( ) ( ) ( ) +1 ( ) 2

(12.19) (12.20)

As in Section 12.2, we retain the same rst order accuracy with the alternative
h ih i ~ ) + O(h ) I ; hAxx I ; hAyx Unx = Unx + h(bc
( ) ( ) ( ) +1 ( ) 2

Write this in predictor-corrector form and permute the data base of the second row. There results

~) ~ x = Unx + h(bc I ; hAxx U h i y ~y I ; hAyy Un = U
( ) ( ) ( ) ( ) ( ) ( ) +1

h

i

(12.21)

224

CHAPTER 12. SPLIT AND FACTORED FORMS

Axx

( )

2 x 6 x 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 6 o 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 6 x 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 6 o 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 Second-Order Factored Implicit Methods

CHAPTER 12. SPLIT AND FACTORED FORMS

Second-order 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 ~ ) be time invariant. There results and the (bc 1 hA ; 1 hA U = I + 1 hA + 1 hA U + h(bc ~ ) + O(h ) (12.24) I;2 x 2 y n 2 x 2 y n Factor both sides giving 1 1 I;1 hA x I ; hAy ; h Ax Ay Un 2 2 4 1 1 1 ~ ) + O(h ) (12.25) = I + hAx I + hAy ; h AxAy Un + h(bc 2 2 4 Then notice that the combination 1 4 h AxAy ](Un ; Un) is proportional to h since the leading term in the expansion of (Un ; Un) is proportional to h. Therefore, we can write 1 1 1 ~ I;1 hA x I ; hAy Un = I + hAx I + hAy Un + h(bc) + O (h )(12.26) 2 2 2 2 and both the factored and unfactored form of the trapezoidal method are second-order accurate in the time march. An alternative form of this kind of factorization is the classical ADI (alternating direction implicit) method usually written 1 hA U 1 hA U + 1 hF I;2 I+2 x ~ = y n 2 n 1 hA U 1 hA = I + (12.27) I;1 y Un x ~ + hFn + O (h ) 2 2 2
+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 time-dependent forcing term.

12.5 Importance of Factored Forms in 2 and 3 Dimensions
When the time-march 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 Peaceman-Rachford 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 hA U = I + 1 hA U + h(bc ~) I;2 x y n 2 x y n 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 Navier-Stokes 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 2-D 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 steady-state solutions of practical problems in two dimensions. However, a three-dimensional 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 two-step predictor-corrector form 1 hA x U ~ x = (RHS ) x I;2 x y y ~y I;1 (12.28) 2 hAy Un = U
( ) ( ) ( ) ( ) ( ) +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 oating-point operations per second.

6 For matrices as small as those shown there are many gaps in this \ ll", but for meshes of

2 (12. Thus. 12.22. 2-D equation. 1 hA U = hhA U + (bc ~ )i + O(h ) (12. for ensuring a correct steady-state solution in a converged time-march.2 x y n x y n 2 This is the delta form of a factored. 1 hA U I. and matrix multiplications do not commute. Now we can factor Eq.2 x 2 y n 2 x 2 y n From both sides subtract 1 hA .2 x 2 y n leaving the equality unchanged. SPLIT AND FACTORED FORMS 9 which has the same appearance as Eq.21 but is second-order time accurate. using the standard de nition of the di erence operator . and ~ ) be time invariant: let the (bc 1 hA .30) I. Inspecting the top of Eq. Consider the unfactored form of the trapezoidal method given by Eq. 12. if Eq. 12. 2nd-order. 12.29 converges.6 The Delta Form Clearly many ways can be devised to split the matrices and generate factored forms. Un one nds 1 hA . The rst step would be solved using My uncoupled block tridiagonal solvers . 12.29 and maintain O(h ) accuracy.228 CHAPTER 12. we see that this is equivalent to solving My one-dimensional ~ x would then problems. ( ) ( ) 12. it is guaranteed to converge to the correct steady-state solution of the ODE. +1 3 +1 + 3 2 + 3 replace the scalar ones.23 shows that the nal be permuted to U step consists of solving Mx one-dimensional implicit problems each with dimension My . each with Mx blocks of order Ne . Then.14. 12. our original ODE.24. 1 hA U = I + 1 hA + 1 hA U + h(bc ~ ) + O(h ) I. One way that is especially useful. We arrive at the expression 1 hA I . is referred to as the \delta form" and we develop it next. 9 A block tridiagonal solver is similar to a scalar solver except that small block matrix operations . Un = Un . The temporary solution U y ~ and an inspection of the bottom of Eq.29) x y n x y n 2 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. 1 hA U = hhA U + (bc ~ )i + O(h ) I.

12. i. hAy ] Un = h Ax y Un + (bc + 10 and its factored form becomes h ~ )i I . u ? iii. 12. hAx] I . and if it is immediately factored. 12.1 j j +1 10 Notice that the only di erence between the O(h2 ) method given by Eq.. 12. the unfactored form of a rst-order method derived from the implicit Euler time march is given by Eq. (1) (2) 12 21 x 2 j .20 and Eq. we will see in the next chapter that the convergence properties of Eq.12.P ). Assume that second order central di erencing is used. hAx . the delta form of the unfactored Eq. If we reorder the points with the odd points rst and then the even points. ii. but it can have a profound e ect on the stability and convergence properties of a method.e. 2 . 12. ( u) = 1 (u .(P . Write down the permutation matrices. For example. 2u + u ) xx j The uniform grid has x = 1 and 8 interior points.7. write the space vector.31 is the appearance of the factor 1 on the left side of the O(h2 ) method. so that we can simplify the boundary conditions. hAy ] Un = h Ax y Un + (bc + (12.20. 12. Consider the 1-D heat equation: @u = @ u @t @x 2 2 0 x 9 Let u(0 t) = 0 and u(9 t) = 0. PROBLEMS 229 The point at which the factoring is made may not a ect the order of time-accuracy. 12. On the other hand. (a) Space vector de nition i. What is the space vector for the natural ordering (monotonically increasing in index).30 and the O(h) method given by Eq. 12.31 are vastly di erent.19.19 is h ~ )i I .31) In spite of the similarities in derivation. the factored form is presented in Eq. u ? Only include the interior points.7 Problems 1.

(b) System de nition In problem 1a. write the delta form of the implicit algorithm. We can put such a partitioning to use. we de ned u u A A P . First de ne extraction operators 21 0 0 0 0 0 0 03 6 0 0 0 07 0 1 0 0 7 6 7 6 2 3 0 0 0 0 0 0 1 0 7 6 7 6 I 0 7 6 0 0 0 07=6 0 0 0 1 7 Io =6 5 7 4 6 0 0 0 0 0 0 0 0 7 6 0 0 6 7 6 0 0 0 07 7 60 0 0 0 40 0 0 0 0 0 0 05 0 0 0 0 0 0 0 0 20 0 0 0 0 0 0 03 6 0 0 0 0 0 0 0 07 7 6 7 6 2 3 0 0 0 0 0 0 0 0 7 6 7 6 0 0 7 6 0 0 0 0 0 0 0 07=6 7 Ie =6 4 5 7 6 0 0 0 0 1 0 0 0 7 6 0 I 7 6 6 0 1 0 07 7 60 0 0 0 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 . SPLIT AND FACTORED FORMS iv. Applying implicit Euler time marching. Comment on the form of the resulting implicit matrix operator. (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 T3 U D 7 A =6 4 5 U D De ne D and U . The generic ODE representing the discrete form of the heat equation is (1) 1 (1) v.230 CHAPTER 12. 1 2 (2) 2 (2) 2 2 . and P which partition the odd points from the even points.

such that Ao operates only on the odd terms. PROBLEMS (2) ( ) ( ) (2) ( ) ( ) (2) 231 which extract the odd even points from u as follows: u o = I o u and ue =Ieu .12. (2) 2 . ii. iii.7. Also. Comment on the solution process this gives us relative to the direct inversion of the original system. Write out the matrices Ao and Ae. de ne a splitting A = Ao + Ae. Comment on the order of the error terms. write them in terms of D and U de ned above. Examine the implicit operators for the factored delta form. You should be able to argue that these are now trangular matrices (a lower and an upper). Write down the delta form of the algorithm and the factored delta form. and Ae operates only on the even terms. Comment on their form. Beginning with the ODE written in terms of u . Apply implicit Euler time marching to the split ODE. i.

232 CHAPTER 12. SPLIT AND FACTORED FORMS .

fractional-step. However.4 we introduced the concept of the representative equation. and we discussed circulant eigensystems1 in Section 4. hybrid. We have seen that under these conditions a semi-discrete approach can lead to circulant matrix di erence operators. accuracy.7. the method is probably not suitable for practical use. The analysis in this chapter is useful for estimating the stability and steady-state properties of a wide variety of time-marching schemes that are variously referred to as time-split. and (approximately) factored. and used it in Chapter 7 to study the stability. and convergence properties of timemarching schemes.3. When these methods are applied to practical problems. 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.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. 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.Chapter 13 LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS In Section 4. the results found from this analysis are neither necessary nor su cient to guarantee stability. 233 . 13. 1 See also the discussion on Fourier stability analysis in Section 7. if the results indicate that a method has an instability.

as a result of space di erencing the PDE.3) that 13.2. g (t) wm a b m m m . and only one2 .4) This is to be contrasted to the developments found later in the analysis of 2-D equations. . eigenvector.4: b since they must share a common The representative equation for split.1) p u + Ab p~ a dt where the subscript p denotes a circulant matrix. 0 = ( + ) w . Since both matrices have the same set of eigenvectors. 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. This suggests (see Section 4. 0 = ( + ) w .2 Example Analysis of Circulant Systems 13.2. g (t) wM a b M M M (13. we can use the arguments made in Section 4.3 to uncouple the set and form the M set of independent equations 0 = ( + ) w .234 CHAPTER 13. g (t) w1 a b 1 1 1 . we arrive at a set of ODE's that can be written d~ u =A ~ ~ u .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..1 Stability Comparisons of Time-Split Methods Consider as an example the linear convection-di usion equation: @u + a @u = @ 2 u @t @x @x2 2 (13. (13.. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS Suppose. f (t) (13...

EXAMPLE ANALYSIS OF CIRCULANT SYSTEMS 235 If the space di erencing takes the form d~ u = . Eq. 13.5) dt 2 x p x the convection matrix operator and the di usion matrix operator.1 0 1)~ u + 2 Bp(1 .2): ( c)m = ia x sin m m (13. 2. Using these values and the representative equation 13. 13. 4 2 sin2 2 x In these equations m = 2m =M .3. The Explicit-Implicit Method 2 (13. a B (. respectively. where (see Section 4. 1 .8.7) . the characteristic polynomial of this method is P (E ) = (1 . When applied to Eq.4. In this section we refer to these as 1. 12. the explicit-implicit Euler method. we can analyze the stability of the two forms of simple time-splitting discussed in Section 12. 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 sin2 m 1 + Cn j j= Cn sin2 m 1 + 4R 2 0 m 1. can be represented by the eigenvalues c and d.2. Eq.13.2 1)~ u (13. (1 + h c) This leads to the principal root 1 + i ah x sin m = 1 + 4 h 2 sin2 m 2 x where we have made use of Eq. so that 0 m 2 .4.10. h d )E . 12. m = 0 1 M . the explicit-explicit Euler method.2.6) ( d)m = .6 to quantify the eigenvalues.

and the .2 C = 2/ R ∆ n 0. From this we nd that the condition on Cn and R that make j j 1 is 2 h 2 sin2 i = 1 + 4 Cn sin2 1 + Cn R 2 As ! 0 this gives the stability region 2 Cn < R which is bounded by a hyperbola and shown in Fig. 4 sin j j = 1 + Cn 0 m 2 R 2 Again a simple numerical parametric study shows that this has two critical ranges of m . one near 0. 13. 2.8 n 1. Figure 13.2 C = R /2 n ∆ C = 2/ R n ∆ 0.4 0 0 4 R∆ Explicit-Implicit 8 0 0 4 R∆ Explicit-Explicit 8 _ tability regions for two simple time-split methods.8 C n C 0. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 1.4 0.1: S A simple numerical parametric study of Eq.1.7 shows that the critical range of m for any combination of Cn and R occurs when m is near 0 (or 2 ). 13. The Explicit-Explicit Method An analysis similar to the one given above shows that this method produces " # q C n m 2 2 2 sin m 1 .236 CHAPTER 13. which yields the same result as in the previous example.

In the former case it is 1 h(3E . The characteristic polynomial follows from the application of the method to the homogeneous equation. 1 d )E . thus 1 h (3u . The di usion term is integrated implicitly using the trapezoidal method. which produces the constraint that 1R Cn < 2 for R 2 The resulting stability boundary is also shown in Fig.2 Analysis of a Second-Order Time-Split Method .1 2 d n+1 n This produces 3 1 1 2 h P (E ) = (1 . This method applies to a ow in which there is a combination of di usion and periodic convection. In order to evaluate the accuracy. as well as the stability. The totaly explicit. The convection term is treated explicitly using the second-order Adams-Bashforth method. 13. 13. Our model equation is again the linear convection-di usion equation 13.5.13.8) and in the latter h(E 2 + E ) (13. as we should have expected. we include the forcing function in the representative equation and study the e ect of our hybrid.9) Q(E ) = 1 2 13. time-marching method on the equation u0 = cu + du + ae t First let us nd expressions for the two polynomials.1.4 which we split in the fashion of Eq. Next let us analyze a more practical method that has been used in serious computational analysis of turbulent ows. (1 + h c + h d )E + h c 2 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. P (E ) and Q(E ). 1) Q(E ) = 2 (13.2. EXAMPLE ANALYSIS OF CIRCULANT SYSTEMS 237 other near 180o.2. factored method has a much smaller region of stability when R is small. u ) + 1 h (u + u ) un+1 = un + 2 c n n.

10 and shown in Fig.2. for the model equation.11) The extension of the following to 3-D is simple and straightforward.6. .7.3 The Representative Equation for Space-Split Operators Consider the 2-D model3 equations @u = @ 2 u + @ 2 u @t @x2 @y2 3 (13. For values of R 2 this second-order method has a much greater region of stability than the rst-order explicit-implicit method given by Eq.9.2. 13. 13. 13. These results show that. 12. 2 .238 CHAPTER 13. 2 d (13. 13. so er = O(h3) Using P (e h) and Q(e h ) to evaluate er in Section 6.10 by a parametric study of cn and R de ned in Eq.1. 13. 13. one can show er = O(h3) using either Eq. 3 h3 c d 1 = 1 + ( c + d )h + 2 ( c + d )2 h2 + 4 3 d c d c 1 3 = 1 ( c + d)3 does not match the coe cient From this equation it is clear that 6 6 of h3 in 1 . 1 1+ 3 h c+ h 2 2 1 2 1 .1. 2h c 1 . 13. 2h d d 2 1h . This was carried out in a manner similar to that used to nd the stability boundary of the rst-order explicit-implicit method in Section 13. The results are plotted in Fig.3.10) Stability The stability of the method can be found from Eq. 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 3h + 1h 1+ 2 c 2 = s d The principal -root follows from the plus sign and one can show 1 1 + 2 . the hybrid method retains the second-order accuracy of its individual components.8 or Eq.

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.8 Cn 0. The form of the Ax and Ay matrices for three-point central di erencing schemes are shown in Eqs.22 and 12. First we write these matrices in the form 2 3 2 ~ 3 ~ B b I b I 0 1 A(xx) = 6 B 7 A(yx) = 6 b.2 0. Now nd the block eigenvector . 12.1 b0 b1 ).23 for the 3 4 mesh shown in Sketch 12.12) @t x @x y @y Reduce either of these.3.4 Stable 0 4 8 ∆ R _ Figure 13.1 I ~ b0 I where B is a banded matrix of the form B (b.12. and @u + a @u + a @u = 0 (13.13. by means of spatial di erencing approximations.1 I ~ b0 I ~ b1 I 7 4 5 4~ 5 ~ B b. to the coupled set of ODE's: dU = A + A ]U + (bc ~) (13. THE REPRESENTATIVE EQUATION FOR SPACE-SPLIT OPERATORS239 1.13) x y dt for the space vector U .2: Stability regions for the second-order time-split method.

Thus 2 .14 is transparent to the transformation.1 B 7 X 4 54 54 5 4 . 12.1 I b0 I b.13.23. the rst matrix on the right side of Eq.1 6 b .1 I ~ ~ ~ ~ ~ b. 1 X B X where 2 3 6 =6 6 4 3 7 5 1 2 set X 4 ( x ) Notice that the matrix Ay is transparent to this transformation.1 4~ I I . 12.15) 6 7 4 5 3I + ~ 4I + ~ diag(X ) 2 ~ b0 X .240 CHAPTER 13. see the bottom of Eq. 13. by the same argument as before.1B ~X ~ =6 ~=6 X 6 4 5 ~ 7 4 5 ~ 3 ~ This time.1 I b0 I One now permutes the transformed system to the y-vector data-base using the permutation matrix de ned by Eq. There results h i Pyx X . if we 3 7 7 7 5 3 2 ~ 3 ~ b1 I b0 I ~ b1 I 6b ~ b0 I ~ b1 I 7 b0 I ~ b1 I 7 5 X = 4~ 5 .1 Pyx X . so the nal result is the complete diagonalization of the matrix Ax+y : h ih ) ih i ~ .1 A(xx) + A(yx) X Pxy = 3 2 3 2 ~ B 1 I 6 7 6 7 ~ 6 7 I B 6 7 2 6 7 6 7+6 (13. ~ ~ ~ diag(B ~ ) and Next nd the eigenvectors X that diagonalize the B blocks.1 ~ b0 ~ b1). Let B ~ ~ X diag(X ) and form the second transformation 2~ 3 2~ 3 1 6 7 ~ 7 7 ~2 ~ .1 32 32 3 2 X B X 6 76 6 7=6 X . That is.14) 7 ~ 4 5 4 B 3 I 5 ~ B 4 I ~ is the banded tridiagonal matrix B (~ where B b.1 A(xx+ ~ = X y X Pxy X 2 3 ~ I + 1 6 7 6 7 2I + ~ 6 7 (13. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS matrix that diagonalizes B and use it to diagonalize A(xx).

3. 4 .16) ~ are any two matrices that have linearly independent eigenvectors (this where B and B puts some constraints on the choice of di erencing schemes).15. THE REPRESENTATIVE EQUATION FOR SPACE-SPLIT OPERATORS241 It is important to notice that: The diagonal matrix on the right side of Eq. and where Ax and Ay satisfy the conditions in 13.15 contains every possible com~.16. 13. Although Ax and Ay do commute. In that case we set = 0 and use the simpler form du = + ]u + a (13. To obtain the latter property the structure of ~ can be either circulant or the matrices is important. and its use is not necessary to arrive at Eq. the steady-state solution of the representative equation. + (13.13. bination of the individual eigenvalues of B and B Now we are ready to present the representative equation for two dimensional systems. does not ensure the property of \all possible combinations". and the convergence rate to. a and are (possibly complex) constants. by itself. 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. The block matrices B and B 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 . 13. but this choice was for convenience only.18) x y We have used 3-point central di erencing in our example. Often we are interested in nding the value of.17) x y dt which has the exact solution a u(t) = ce( x+ y )t . this fact.

3. h from which x .4. steady-state solution. " 1 #n a x+ y 2. we nd (1 . h y )E . . Produces the exact (see Eq. 13.16. The accuracy of the xed. Is unconditionally stable.19.4 Example Analysis of 2-D Model Equations 1. 2. and when it is applied to the representative equation.5. h x . The stability.1 The Unfactored Implicit Euler Method Consider rst this unfactored. h . The form of the method is given by Eq. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 13. In the following we analyze four di erent methods for nding a xed.18) steady-state solution (of the ODE) for any h. use of this method for 2-D problems is generally impractical for reasons discussed in Section 12. this method: 1. Converges very rapidly to the steady-state when h is large. x y Like its counterpart in the 1-D case. Unfortunately. 1 Q(E ) = h giving the solution (13. h . 12.19) un = c 1 . however.242 CHAPTER 13. steady-state solution to the 2-D representative equation 13. h y )un+1 = un + ha P (E ) = (1 . rst-order scheme which can then be used as a reference case for comparison with the various factored ones. 3. The convergence rate to reach the steady-state. In each case we examine 13.

13. 12. Converges rapidly to a steady-state for large h.4.4. Produces a steady state solution that depends on the choice of h. the converged value is meaningless. but the converged solution is completely wrong. h x)(1 . There results (1 .20 to the 2-D representative equation. Is unconditionally stable. h x)(1 . + a x y x y . h y )(un+1 . h ) . 2. h x)(1 . if one tries to take advantage of its rapid convergence rate. h )(1 . it is not very useful since its transient solution is only rst-order accurate and. Next apply Eq.20) giving the solution " #n 1 un = c (1 . The method requires far less storage then the unfactored form.2 The Factored Nondelta Form of the Implicit Euler Method 13.h x y We see that this method: 1.3 The Factored Delta Form of the Implicit Euler Method . h y )E .31 to the 2-D representative equation. EXAMPLE ANALYSIS OF 2-D MODEL EQUATIONS 243 Now apply the factored Euler method given by Eq. However. h )(1 . un) = h( xun + y un + a) which reduces to (1 . h ) . 12. h y )un+1 = un + ha from which P (E ) = (1 . 1 Q(E ) = h (13. h x)(1 . 3.4. + x y x y This method: 13. h y )un+1 = 1 + h2 x y un + ha and this has the solution " #n 2 x y 1 + h a un = c (1 . One nds (1 .

it can be used when time accuracy is desired. but convergence is not nearly as rapid as that of the unfactored form.27 should be enough to convince the reader that the 's produced by those methods are identical to the produced by this method. 12. Is unconditionally stable. 1 h x 1. h y (un+1 .30 to the representative equation and one nds 1 1h 1. Like the factored nondelta form. Converges very slowly to the steady{state solution for large values of h. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 1. Is unconditionally stable. 5 13. un = c6 4 5 1 x+ y 1. Converges very slowly to the steady{state solution for large values of h. Since it is second order in time. Apply Eq. h y 2 2 This method: 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 1 h 3n 1h 1 + 1 + 6 a 2 x 2 y 7 7 . In practical codes.5 A brief inspection of eqs.4.26 and 12. Finally consider the delta form of a second-order time-accurate method. the value of h on the left side of the implicit equation is literally switched from h to 1 2 h. 12. since j j ! 1 as h ! 1.5. this method demands far less storage than the unfactored form. un) = h( x un + y un + a) 2 which reduces to 1 . 1 h x 1 . as discussed in Section 12. 2 x 1 . All of these properties are identical to those found for the factored delta form of the implicit Euler method. Produces the exact steady-state solution for any choice of h.4 The Factored Delta Form of the Trapezoidal Method . 3. since j j ! 1 as h ! 1. The correct steady solution is obtained. and the factored delta form of the implicit Euler method can be used when a converged steady-state is all that is required.244 CHAPTER 13. 3. 2. 2. Produces the exact steady{state solution for any choice of h.

under the conditions given in 13. nd the root.5 Example Analysis of the 3-D Model Equation 245 The arguments in Section 13.16 with an A(zz) included.22. 2 h( x + y + z ) + 4 h ( x y + x z + y z ) .5.23) Eqs. +a+ x y 6 z (13. the model 3-D cases6 have the following representative equation (with = 0): du = + + ]u + a (13.21) x y z dt Let us analyze a 2nd-order accurate. factored.11 and 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). 1 h( x + y + z ) un = h ( x + y + z )un + a] 2 Now factor the left side: 1h 1 1 1.13. delta form using this equation. 8 h x y z .3 generalize to three dimensions and. 2 x y z n+1 2 x y z n Put this in delta form: 1 .12. 2 x 1 . h z un = h ( x + y + z )un + a] (13. 13. One can derive the characteristic polynomial for Eq. 8 h x y z 7 7 un = c6 4 5 1 1 1 2 3 1 . and write the solution either in the form 2 3n 1 1 1 2 3 6 1 + 2 h( x + y + z ) + 4 h ( x y + x z + y z ) . h y 1 . 13. each with an additional term. First apply the trapezoidal method: un+1 = un + 1 h ( + + )u + ( x + y + z )un + 2a] 2 x y z n+1 Rearrange terms: 1 h( + + ) u = 1 + 1 h( + + ) u + ha 1. EXAMPLE ANALYSIS OF THE 3-D MODEL EQUATION 13. .

24 results in 1 h2( + + )2 = 1 + h( x + y + z ) + 2 (13. it follows from Eq. the instability disclosed above is too weak to cause trouble. 13. 2h y 1 . if the method converges. 5 x+ y+ z a (13. +i where . h 2 2 2 4 1 1 1 1 . if all the 's are real and negative. 13. the 3-D form of Eq.7 With regards to stability. in the presence of this dissipation. 13. In practical cases. In this case. clearly. and are real numbers that can have any sign. 13.23 in the form 1+i . however.246 CHAPTER 13. Furthermore.25) x y z 3 h 3 + (2 y + 2 x ) + 2 2 + 3 x y + 2 2 + 3 + 2 x 2 + 2 2 y + 3 i + h + 1 y y y y x x 4 z which veri es the second order accuracy of the factored form.23 that. )2 + ( . the method is stable for all h. 13.22 should not be used for the 3-D Euler equations. 7 However. Remember that in our analysis we must consider every possible combination of these eigenvalues. In that case since the absolute value of the product is the product of the absolute values j 2 + ( + )2 (1 . LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS or in the form 2 6 un = c6 4 1 1 1 3 1+ 1 h x 1+ h y 1+ h z . 2h x 1 .24) It is interesting to notice that a Taylor series expansion of Eq.12 with periodic boundary conditions. we already knew this because we chose the delta form. )2 >1 and the method is unconditionally unstable for the model convection problem. central di erencing causes all of the 's to be imaginary with spectrums that include both positive and negative values. and are both positive. First write the root in Eq. ) 2 j = (1 . From the above analysis one would come to the conclusion that the method represented by Eq. 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.i . Now we can always nd one combination of the 's for which . This makes the method unconditionally stable for the 3-D di usion model when it is centrally di erenced in space. Now consider what happens when we apply this method to the biconvection model. +i =1 . . 2h z 3n x y z7 7 . it converges to the proper steady-state.

(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. What is the error term? (b) Instead of making all of the split terms implicit.6.13. (d) Repeat 1c for the explicit-implicit scheme of problem 1b. and accuracy of the converged steady-state solution. convergence. du = Au + f dt we can split A as follows: A = A1 + A2 + A3 + A4. . PROBLEMS 13. 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. leave two explicit: un+1 . Starting with the generic ODE.6 Problems 247 1. 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.

248 CHAPTER 13. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS .

. Given below is some notation and some of the important relations between matrices and vectors.... In the present context a vector is a vertical column or string.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. A. 7 5 vm and its transpose ~ vT is the horizontal row ~ ~ vT = v1 v2 v3 : : : vm] v = v1 v2 v3 : : : vm ]T 2. A general m m matrix A can be written 2 a11 a12 a1m 3 6 a21 a22 a2m 7 7 A = (aij ) = 6 7 6 . 5 4 am1 am2 amm 249 .1 Notation 1. Thus 2 v1 3 6 v2 7 7 ~ v=6 6 4 .

6.T 7 5 ~ am 4.A. The inverse of a matrix (if it exists) is written A. 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. USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA 3.2 De nitions 1. 2. 9. (Hermitian).1 = I .1 and has the property that A. c a .1A = AA.250APPENDIX A.1 = det1 A] . If b A= a c d then det A] = ad . 4. det A] is the determinant of A. bc and d . 7. 3. A is the complex conjugate of A. 11. An alternative notation for A is h ~ i A= ~ a1 a2 : : : ~ am and its transpose AT is 2 ~T 3 a1 7 6 6 7 ~ aT 2 7 AT = 6 6 7 6 4 . 8.. A is skew-symmetric or antisymmetric if AT = . 10.b A. A is normal if AT A = AAT . where I is the identity matrix. A is symmetric if AT = A. A is orthogonal if aij are real and AT A = AAT = I 5. AH is the conjugate transpose of A.. The trace of a matrix is Pi aii. A. P is a permutation matrix if P ~ v is a simple reordering of ~ v.

4. 2.1 )T 251 7.3. sA = (saij ) where s is a scalar. Any matrix A can be expressed as the sum of a symmetric and a skew-symmetric matrix. The eigenvalue problem for a matrix A is de ned as A~ x= ~ x or A .3 Algebra We consider only square matrices of the same dimension. they are all imaginary. Inverse equalities (if the inverse exists): (A. In general AB 6= BA. I ]~ x=0 and the generalized eigenvalue problem.1 = B .4 Eigensystems 1.1 = A (AB ). 1. its eigenvalues are all real. . AT A= 2 2 A. Transpose equalities: (A + B )T = AT + B T (AT )T = A (AB )T = B T AT 6. If it is asymmetric. If a square matrix with real elements is symmetric.1 A. including the matrix B . A and B are equal if aij = bij for all i j = 1 2 : : : m. 3. 5.1).A. ALGEBRA A.1 = (A. A + (B + C ) = (C + A) + B . Thus: 1 A + AT + 1 A .1 (AT ). etc. B ]~ x=0 2. as A~ x = B~ x or A .

In general. then A is always diagonalizable. 8.1AX = where X is a matrix whose columns are the eigenvectors. USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA 3. 2 0 03 1 6 6 0 2 ... X . and A is said to have a complete eigensystem. If a matrix does not have a complete set of linearly independent eigenvectors. . 6 . A set of eigenvectors is said to be linearly independent if a ~ xm + b ~ xn 6= ~ xk m 6= n 6= k for any complex a and b and for all combinations of vectors in the set.252APPENDIX A. and with each distinct eigenvalue there is one associated eigenvector. If A posseses m linearly independent eigenvectors then A is diagonalizable. 6. . and this eigenvector cannot be formed from a linear combination of any of the other eigenvectors. in which case the possibility exists that it cannot be diagonalized.e. the eigenvalues of a matrix may not be distinct. If A has m distinct eigenvalues. its eigenvectors completely span the space. 7 7 7 =6 .. . . but it can still be diagonalized. The eigenvectors of such a matrix cannot span the space and the matrix is said to have a defective eigensystem. h ~ i X= ~ x1 x2 : : : ~ xm and is the diagonal matrix 0 0 m If A can be diagonalized. . In general. 9. the matrix is said to be derogatory. . i.. . 5. . If the eigenvalues of a matrix are not distinct. 0 7 5 . but all of the eigenvectors are linearly independent. 7. it cannot be diagonalized. 4 . 4. an m m matrix A has n~ x linearly independent eigenvectors with n~ n~ x m and n distinct eigenvalues ( i ) with n x m. 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.

EIGENSYSTEMS 2J 0 03 1 6 .A. . Ji = 6 7 . 1 vectors associated with this eigenvalue are referred to as principal vectors. 4. A repeated root in a Jordan block is referred to as a defective eigenvalue.. 6 . 7 6 7 6 7 . . ... For each Jordan submatrix with an eigenvalue i of multiplicity r.. For example. 13. . there exists one eigenvector. . 1 6 7 J = S AS = 6 . .. .4.. 5 0 0 0 i 12. The complete set of principal vectors and eigenvectors are all linearly independent. Some of the Jordan subblocks may have the same eigenvalue. S . . 11. . . . 1 7 4 .. 07 5 0 0 Jk 253 10. the . The other r . 0 0 0 i 6 7 . Note that if P is the permutation matrix 2 3 0 0 1 P =6 40 1 07 5 1 0 0 P T = P .. . . Defective matrices cannot be diagonalized but they can still be put into a compact form by a similarity transform. A Jordan submatrix has the form 2 0 03 i 1 6 6 7 0 i 1 . . Use of the transform S is known as putting A into its Jordan Canonical form. . such that where there are k linearly independent eigenvectors and Ji is either a Jordan subblock or i.1 = P then 2 3 2 3 1 0 0 0 P . 7 6 7 0 J 2 .1 6 17 07 40 5P = 6 41 5 0 0 0 1 14. .

2. 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.5 Vector and Matrix Norms 1. A p-norm of a matrix A is de ned as jjAvjjp jjAjjp = max x6=0 jjv jj p . A p-norm of the vector ~ v is de ned as 0M 11=p X jjvjjp = @ jvj jpA j =1 3. USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA matrix 22 6 6 4 6 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.254APPENDIX A. having: 9 eigenvalues 3 distinct eigenvalues 3 Jordan blocks 5 linearly independent eigenvectors 3 principal vectors with 1 1 principal vector with 2 A.

8. The spectral radius of A. in fact. so in general (A) is not a true norm. 7. is the lower bound of all the norms of A. Special p-norms are jjAjj1 = maxj=1 M jjAjj1 = maxi=1 2 M PM maximum row sum j =1 jaij j where jjAjjp is referred to as the Lp norm of A. (A) is a true norm. in this case it is the L2 norm. (A). When A is normal. 6. VECTOR AND MATRIX NORMS 255 4. jjAjj2 = q PM ja j maximum column sum i=1 ij (A T A ) . 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 5.5. Let A and B be square matrices of the same order. In general (A) does not satisfy the conditions in 4.A.

256APPENDIX A. USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA .

1) m = b + 2 ac cos M +1 The right-hand eigenvector of B (a b c) that is associated with the eigenvalue m satis es the equation B (a b c)~ xm = m~ xm (B.1 Standard Eigensystem for Simple Tridiagonals In this work tridiagonal banded matrices are prevalent. It is useful to list some of their properties. If we examine the conditions under which the determinant of this matrix is zero. Let us consider a simple tridiagonal matrix. a tridiagonal with constant scalar elements a. and c. i. see Section 3.e.1 a m ~ xm = (xj )m = c 2 sin j M m=1 2 M (B.4.b. Many of these can be derived by solving the simple linear di erence equations that arise in deriving recursion relations.3) +1 257 .2) and is given by j.Appendix B SOME PROPERTIES OF TRIDIAGONAL MATRICES B.. we nd (by a recursion exercise) det B (M : a b c)] = 0 if p m b + 2 ac cos M m=1 2 M +1 =0 From this it follows at once that the eigenvalues of B (a b c) are p m m=1 2 M (B.

. md)(c . 6 7 6 7 6 7 6 .2 Generalized Eigensystem for Simple Tridiagonals This system is de ned as follows 2 32 32 x 3 2e f x 3 b c 6 7 6 6 7 6 x 7 x 7 a b c d e f 6 7 6 7 6 7 6 7 6 7 6 7 6 7 6 7 6 7 6 7 6 7 6 7 x x a b d e = 6 7 6 7 6 7 6 7 . 1 .1 and c = 1. j. f ] = 0 if q m b . e c . SOME PROPERTIES OF TRIDIAGONAL MATRICES These vectors are the columns of the right-hand eigenvector matrix. m e + 2 (a . .258 APPENDIX B. X = M + 1 a 2 sin M j=1 2 +1 In this case notice that if a = . 5 4 ..5) c ( 1) The left-hand eigenvector matrix of B (a b c) can be written m. . c m2 2 =e a 1 ( 1) M M (B. m f ) cos M m=1 2 +1 =0 If we de ne m m= M + 1 pm = cos m 1 2 3 1 2 3 (B..1 a jm j=1 2 M X = (xjm) = c 2 sin M (B. the elements of which are j.1 and c = 1 .8) . f 5 4 .i m.6) B.1 c 2 mj m=1 2 . d b .1 a 2 = ei j. c 5 4 . 2 (B. . 7 4 5 a b xM d e xM In this case one can show after some algebra that det B (a .4) m = 1 2 M +1 Notice that if a = .7) M (B. .

fb)(ea . b . bE + ac) and the two roots to P ( ) = 0 result in the solution 8" p p " #M 9 < b + b . (B. . 4ac #M = 1 b .B. 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 . it is simple to derive the rst few determinants.3. 2abc (B. In the limiting case when b .3 The Inverse of a Simple Tridiagonal The inverse of B (a b c) can also be written in analytic form.9) One can also nd the recursion relation DM = bDM . Its characteristic polynomial P (E ) is P (E . 2 2 b . Let DM represent the determinant of B (M : a b c) DM det B (M : a b c)] De ning D to be 1. f j =1 2 M 2 2 2 259 2 q m These relations are useful in studying relaxation methods. 2(cd + af )pm + 2pm (ec . acDM . 4ac : M =0 1 2 (B. 4ac = 0. 4fdpm The right-hand eigenvectors are #j .10) Eq. B. THE INVERSE OF A SIMPLE TRIDIAGONAL eb . thus D = 1 D = b D = b . af )pm ] m= e . d m =1 2 M m ~ sin j m ] xm = c . 1 " 2 a . 4 ac DM = p .11) where we have made use of the initial conditions D = 1 and D = b. ac D = b .10 is a linear O E the solution of which was discussed in Section 4. B.2. bd) + (cd .

c D .aD D D D . DM .a D aD .mDn.m =DM Diagonal: n=m=1 2 M dmm = DM .a D a D .4.cD D c D a D .c D c D D D .cD D c D D .cD cD .aD D D D .cD 3 7 17 7 2 7 7 5 3 Upper triangle: m=1 2 M .n=DM B.1 Standard Tridiagonals Consider the circulant (see Section 3.4.1 dmn = DM .1 1 n=m+1 m+2 M dmn = Dm.cD D c D a D D .1 = aD The general element dmn is 4 0 2 D4 6 .a)m.a D1 cD . SOME PROPERTIES OF TRIDIAGONAL MATRICES 2 16 6 4 D4 6 Then for M = 4 B.c)n.m =DM 1 Lower triangle: m =n+1 n+2 M 1 n=1 2 M .aD D D . = 1 D .n(.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 .12) .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 .aD D D D .cD .4 Eigensystems of Circulant Matrices B.4) tridiagonal matrix Bp(M : a b c ) (B.c D . DM .260 APPENDIX B.cD . (.aD3 16 6 a2 D 6 3 2 D5 6 4 .

13 is a special case.14. and the right-hand eigenvector matrix has the form 2 m j =0 1 M . 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.1 X = (xjm) = eij M m =0 1 M .1 1 X . B. even if they are completely dense. B.1. for the tridiagonal circulant matrix given by eq.12 do not depend on the elements a b c in the matrix.4.15 they have the general form M . where X is the Note that both X and X . and further examination shows that the elements in these eigenvectors correspond to the elements in a complex harmonic analysis or complex discrete Fourier series.4. 1 1 1 M . the eigenvalues are not. EIGENSYSTEMS OF CIRCULANT MATRICES The eigenvalues are m 261 = b + (a + c) cos 2 m .15) 4 b b b b 5 b b b b The eigenvectors are always given by eq. are symmetric and that X .13) B. For the element indexing shown in eq. B. all circulant matrices of order M have the same set of linearly independent eigenvectors. = M conjugate transpose of X .1 The left-hand eigenvector matrix with elements x0 is . j =0 .im 2 j 1 m =0 1 M .1 (B. 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.1 (B. i(a . c) sin 2 m M M (2 ) m=0 1 2 The right-hand eigenvector that satis es Bp(a b c)~ xm = m~ xm is ~ xm = (xj )m = ei j m=M j=0 1 M . In fact. Although the eigenvectors of a circulant matrix are independent of its elements.14) p where i . 0 M X = (xmj ) = M e j =0 1 M .B.2 General Circulant Systems Notice the remarkable fact that the elements of the eigenvector matrices X and X . B.1 .

B. .1. 6 7 7 . 1) m = b + 2a cos 2M + 1 ! m=1 2 M (B. 7 76 6 a 5 4 x2 7 5 = x 2 3 x1 6 x2 7 6 7 6 . 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 . .16) By folding the known eigenvectors of B (2M : a b a) about the center. 7 6 7 7 6 7 7 6 7 ..... SOME PROPERTIES OF TRIDIAGONAL MATRICES Consider a mesh with an even number of interior points such as that shown in Fig.. B..5 Special Cases Found From Symmetries APPENDIX B.17) . 7 6 6 7 7 m6 . a a b a b 32x 3 1 6 7 7 x 6 2 7 7 6 7 . For example. one can show from previous results that the eigenvalues of eq.262 B. a 7 7 a b a 7 5 a b+a 2 6 6 6 6 6 6 6 6 6 6 4 3 xm m~ (B. we seek the set of eigenvectors ~ xm for which 2 b 6 a 6 6 6 6 6 6 6 4 a b a a ...16 are (2m .. 7 6 7 6 6 7 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 ~ ~ ~ 7 B (M : a b a)xm = xm = 7 . 7 .

.. a a b a) = 6 B (M : ~ . 1) ~ xm = sin j (2 2M + 1 j =1 2 M Imposing symmetry about the same interval but for a mesh with an odd number of points. B. B. 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. a 7 6 7 6 4 5 a b a7 2a b By folding the known eigenvalues of B (2M .B.17 are (2m .6 Special Cases Involving Boundary Conditions We consider two special cases for the matrix operator representing the 3-point central di erence approximation for the second derivative @ =@x at all points away from the boundaries. An odd{numbered mesh Figure B. combined with special conditions imposed at the boundaries. An even-numbered mesh Line of Symmetry x =0 x= j0 = 1 2 3 4 5 M0 j= 1 2 3 M b. see Fig. 2 2 . SPECIAL CASES INVOLVING BOUNDARY CONDITIONS and the corresponding eigenvectors are m . 6 7 .1 { Symmetrical folds for special cases m=1 2 M . leads to the matrix 2 3 b a 6 7 a b a 6 7 6 7 .6. 1 : a b a) about the center. one can show from previous results that the eigenvalues of eq.1. 6 7 . 1) ~ xm = sin j (2m 2M ! j=1 2 M B.

2 1 1 .4 sin ( =2) 2 When the boundary conditions are Dirichlet on both sides. 1) = .2 h+ 2 cos M m i+ 1 = sin j M +1 (B.2 1 1 . SOME PROPERTIES OF TRIDIAGONAL MATRICES Note: In both cases m =1 2 M j =1 2 M . 2 6 6 6 6 6 6 4 .19) .2 + 2 cos( ) = .18) When one boundary condition is Dirichlet and the other is Neumann (and a diagonal preconditioner is applied to scale the last equation).1 3 7 7 7 7 7 7 5 ~ xm m m . 2 6 6 6 6 6 6 4 .264 APPENDIX B.2 + 2 cos (2 2M + 1 m .2 1 1 .2 1 1 .2 1 1 .2 1 1 .2 3 7 7 7 7 7 7 5 m ~ xm m = . 1) = sin j (2 2M + 1 (B.2 1 1 .2 1 1 .

In order to examine this property. Consider rst the scalar case.3) @F Q = nF (Q) @q 265 # (C. Di erentiating both sides with respect to and setting = 1 (since the identity holds for all ). let us rst inspect Euler's theorem on homogeneous functions.1) for a xed n. If F (u v) satis es the identity F ( u v) = nF (u v) (C.4) .2) F ( Q) = nF (Q) for a xed n. we nd Consider next the theorem as it applies to systems of equations. F is said to be homogeneous of degree n and we nd " (C. F is called homogeneous of degree n.Appendix C THE HOMOGENEOUS PROPERTY OF THE EULER EQUATIONS The Euler equations have a special property that is sometimes useful in constructing numerical methods. If the vector F (Q) satis es the identity u @F + v @F = nF (u v) @u @v (C.

3. 6.7) (C. BnQn are identically zero for homogeneous vectors of degree 1.5) E = An Q + O(h2) F = BnQ + O(h2) (C.266APPENDIX C.8) @A Q = 0 @x in spite of the fact that individually @A=@x] and Q are not equal to zero. that both E and F in eqs. F = AQ and @F = A @Q + @A Q @x @x @x Therefore. E = En + An(Q .11 and 2. 6. the constant term drops out of eq. A and B .12 are homogeneous of degree 1. As a nal remark.106. and their Jacobians. Notice also that. Qn) + O(h2) can be written (C. C. 2. AnQn and Fn . Qn) + O(h2) F = Fn + Bn(Q . THE HOMOGENEOUS PROPERTY OF THE EULER EQUATIONS Now it is easy to show. under this condition. by direct use of eq.4. where is the internal energy per unit mass. see eq. " # " # (C.6) since the terms En .105 can be written in general as. according to eq. 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. if F is homogeneous of degree 1. C. .1 This being the case. The Euler equations are homogeneous if the equation of state can be written in the form p = f ( ). we notice that the expansion of the ux vector in the vicinity of tn which. 1 " # (C. are homogeneous of degree 0 (actually the latter is a direct consequence of the former).9) Note that this depends on the form of the equation of state.

Hirsch.Bibliography 1] D. J. 1984. volume 1. Numerical Computation of Internal and External Flows. Computational Fluid Mechanics and Heat Transfer. Anderson. Tannehill. McGraw-Hill Book Company. and R. Pletcher. John Wiley & Sons. New York. 2] C. 1988. 267 .2. New York.

You're Reading a Free Preview

Download
scribd
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->