Business Advertising Branding Business Management Business Ethics Careers, Jobs & Employment Customer Service Marketing Networking Network

Marketing Pay-Per-Click Advertising Presentation Public Relations Sales Sales Management Sales Telemarketing Sales Training Small Business Strategic Planning Entrepreneur Negotiation Tips Team Building Top Quick Tips Internet & Businesses Online Affiliate Revenue Blogging, RSS & Feeds Domain Name E-Book E-commerce Email Marketing Ezine Marketing Ezine Publishing Forums & Boards Internet Marketing Online Auction Search Engine Optimization (SEO) Spam Blocking Streaming Audio & Online Music Traffic Building Video Streaming Web Design Web Development Web Hosting Web Site Promotion Finance Credit Currency Trading Debt Consolidation Debt Relief Loan Insurance Investing Mortgage Refinance Personal Finance Real Estate Taxes Stocks & Mutual Fund Structured Settlements Leases & Leasing Wealth Building

Communications Broadband Internet Mobile & Cell Phone VOIP Video Conferencing Satellite TV Reference & Education Book Reviews College & University Psychology Science Articles Food & Drinks Coffee Cooking Tips Recipes & Food and Drink Wine & Spirits Home & Family Crafts & Hobbies Elder Care Holiday Home Improvement Home Security Interior Design & Decorating Landscaping & Gardening Babies & Toddler Pets Parenting Pregnancy News & Society Dating Divorce Marriage & Wedding Political Relationships Religion Sexuality Computers & Technology Computer Hardware Data Recovery & Computer Backup Game Internet Security Personal Technology Software Arts & Entertainment Casino & Gambling Humanities Humor & Entertainment Language Music & MP3 Philosophy Photography Poetry Shopping & Product Reviews Book Reviews Fashion & Style

Health & Fitness Acne Aerobics & Cardio Alternative Medicine Beauty Tips Depression Diabetes Exercise & Fitness Fitness Equipment Hair Loss Medicine Meditation Muscle Building & Bodybuilding Nutrition Nutritional Supplements Weight Loss Yoga Recreation and Sport Fishing Golf Martial Arts Motorcycle Self Improvement & Motivation Attraction Coaching Creativity Dealing with Grief & Loss Finding Happiness Get Organized - Organization Leadership Motivation Inspirational Positive Attitude Tips Goal Setting Innovation Spirituality Stress Management Success Time Management Writing & Speaking Article Writing Book Marketing Copywriting Public Speaking Writing Travel & Leisure Aviation & Flying Cruising & Sailing Outdoors Vacation Rental Cancer Breast Cancer Mesothelioma & Asbestos Cancer

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. . . . 85 7 STABILITY OF LINEAR SYSTEMS 121 122 123 123 123 . . . . . . . . . . . 6. . . . . .2 Converting Time-Marching Methods to O E's . . .2 Complete Eigensystems . .10. . . . . . . .1 Establishing the Relation . . .4 Local Linearization for Coupled Sets of Nonlinear Equations 6. .6. . . . . . . . . . .1 The Criterion . . .8 Predictor-Corrector Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6. .7 Linear Multistep Methods . . . . 6. . . . . . 6.1 The Operational Form and its Solution . . 6. . . . . . . . . 6. . . . . . . 6. . . . . . . . . . . . . . . . . . . . . . . . .5. 6. . . . . . . 6. . . . . . . . . . . . . . 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 6.3 Spurious -Roots . . .6. . . . . . . .6 Accuracy Measures of Time-Marching Methods . . . . . . . . . 7. . . . . . . . . . . . . .1 Local and Global Error Measures . . . . . 6. .4. . . . . . . . . . . . . .5. .6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Notation . 6. . . . 6. . . 6. . . . . . . . . . . . .2 Application to Nonlinear Equations . . . . . . . . . . . . . . . . . . .11 Problems . . . . .5 Global Accuracy . . . . . . . . . . . . . .1 Dependence on the Eigensystem 7.5 The . . . . . . . . . . .1 The General Formulation . . . . . . . . . . . . . .10. 6. . . . . . . . . . . . . . . . . . . . . .7. . . . 6. . . .3 Local Linearization for Scalar Equations . . . .4 Solution of the Representative O E's . . . . . . . . . . . . . . . . . . .2 Special Cases of Coupled First-Order Equations . 6. . . . . . . . . . . . . . . . . . . . . . . . 7. . . . . . . . . . . . 6. . . .and Second-Order Di erence Equations .10 Implementation of Implicit Methods .5 Problems . . . . . .4. . . . . . . 6. 6. . . 6. 6. . . . . .9 Runge-Kutta Methods . . . . . . . . . . . . . . . . . . . . .2 Inherent Stability of ODE's . . . . . .5.3 Two-Step Linear Multistep Methods . . . . . . . .6. 6. . . .10. . . . . . . . .3 Solution of Linear O E's With Constant Coe cients . . . 7.1 First. . . . . . .7.3. . . . . Relation . . . .2. . . .2 Examples of Solutions to Time-Marching O E's . . . . . . . . . .2. . . . .10. . .3 Local Accuracy of the Particular Solution (er ) . . . . .2 The Principal -Root . . . . . . 6.6. . . . . . . . . . . . . . . . . . . . . . . . . . 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Local Accuracy of the Transient Solution (er j j er! ) . . . . . 6.7. . . . . 6. .3. 6. . . . . . . . . . . .4 One-Root Time-Marching Methods .2 Examples . . .1 Application to Systems of Equations . . . . . . . . . . . . . . . 6. . . . . . . . . .5. . . . 6. . . .4 Time Accuracy For Nonlinear Applications . . . .

. .3 Stability Contours in the Complex h Plane. . . .4 Comparing the E ciency of Explicit Methods . . . 7. . . . . . . . . . . . . . . . . . . .2. .2 Some Examples . . . . . . . .5. . . . . . . . . . . . . . . . . . 7. . . . . . . . . . . . . . . . . . . . . . . . . . 8. . . . . . . . . . . . . . . . . . . . . . . . . .6. . . . . . . . . . . . . . . . . .8 7. 7.1. Numerical Stability of O E 's . .3.2 Implicit Methods . . . .5 Coping With Sti ness . .2 Driving and Parasitic Eigenvalues . . Consistency . . . . .6 Steady Problems . . . . . . . . . . . . . . . . . . .9 7. . . . 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. . . . .7 7.1 Imposed Constraints . . . 7. . . . . . . . . 8. . .7. . . .6 7. . . . . 8. . . 7. .7. .7. .1 The Criterion . . . . . . 8. . . . . . . . .3 Relation to Circulant Matrices . . . . .1 Explicit Methods . . . . . . . . . . . . . . . . 8. . . . . . . . . . . . . . . . . . . . . . . .6. . . . . . . . . . . . . . . . . . . . . . .4. . . . . . . . . . . . . . . . . 8. . . . .5. . . . . . . . . Time-Space Stability and Convergence of O E's . . . . . . . . . . .3 A Perspective . . . . . . . .2 An Example Involving Di usion .3. . . .5 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . Numerical Stability Concepts in the Complex -Plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7. . . . . . .3 7. . . . . . . . . . . . .1. . .4. . 8. . . . . . . . . . 7. . 8. . . . . . . 7. . . . . . . . . . . . 149 149 149 151 151 152 153 154 154 154 155 158 158 159 160 160 161 . . .2 Complete Eigensystems .3 Sti ness Classi cations . . . . . . .2 Relation of Sti ness to Space Mesh Size . . . . 8. . .5. . . 8.7. . . . . . . . . . .3 Practical Considerations for Comparing Methods . . . . .3 Defective Eigensystems . . . . . .3 An Example Involving Periodic Convection . . . . . . . 7. . . .2 Unconditional Stability. . . . . . . . . . . . .4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Stability for Large h.7 Problems . .1 Relation to -Eigenvalues . . . . . . . 8. . . . . . . . . . . . . . . . . . .2 Stability for Small t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5. . . . . . .3 Defective Eigensystems . . 7. . . . . . Problems . . . . . . . 8. . . . . . . . . .4 7. . . . . . . . . . . . . . Fourier Stability Analysis . . .5. . . . . 8. . . . . . . . 8. . . . .1. . Numerical Stability Concepts in the Complex h Plane 7. . . . . . . . . . . . .1 Sti ness De nition for ODE's . . . . . . . . . . . . . . . . . 8. . . . . .1 The Basic Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6. . . . .1 -Root Traces Relative to the Unit Circle .3. . . . . A-Stable Methods . . .

. . . . .2 Factoring Physical Representations | Time Splitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.5 Arti cial Dissipation . . . . . . . . . . . . .9 RELAXATION METHODS 9. . . . . . .4 One-Sided First-Derivative Space Di erencing The Modi ed Partial Di erential Equation . . . . 9.3 The SOR System . . . 9. 9. . . . . . . . . . . . . .2 The Converged Solution. . . . . . . . . . . . . . . . .1 Preconditioning the Basic Matrix . . .3 11.4. . . . .3 Factoring Space Matrix Operators in 2{D . . . . . . . . 9. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 The Classical Methods . . . . . . . . . . . . . . . 9. . . . . . . . . . . . . 9. . . . . . . . . . .3 The ODE Approach to Classical Relaxation . . . . . . . .6 Problems . . .1. . . . . . . . . . . . . . . . . . . . . . 11. . . . . .1 The Ordinary Di erential Equation Formulation . . . . . . . . . . . . . . . 202 11. . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 11. . . . . . . . . . . . . . . . 163 164 164 166 168 168 168 169 170 170 172 173 174 176 180 182 187 10 MULTIGRID 10. . .2 Flux-Di erence Splitting . . . . . 9. . . and the Error 9. . 218 12.4 Problems .4. . 192 10. .4 Eigensystems of the Classical Methods . . .2 11. . . . . . . . .5 Nonstationary Processes . . . . . . . . . . . . . . . . . . . . . . . . . . 9. . . . . . . Upwind Schemes . . . . .3. . . .2. . . . . the Residual. . . . . . . . . . . . . .1. . . .1 Flux-Vector Splitting . . . .4. . . . 9. .2 The Basic Process .3. . . . . . . . . . . . 191 10. . . . 192 10. . . . . . . . . . . . . 9. . . . . . . . . . . . . . . . . 11. . . . .1 The Point-Jacobi System . . . . . . . . .2 ODE Form of the Classical Methods . . .1 Formulation of the Model Problem . . . . . . . . . .3 A Two-Grid Process . . . . . . . . . . . .2. . . . . . . . . . . . .6 Problems . . 191 11 NUMERICAL DISSIPATION 203 204 205 207 209 210 212 213 214 12 SPLIT AND FACTORED FORMS 12. . . . . . . . . . . . . . . . 9. .2. . . 9. . . . . . . . . . . . . . . . 11. . . . . . . . . . . . . . . . . . .1 Eigenvector and Eigenvalue Identi cation with Space Frequencies191 10. . . The Lax-Wendro Method . . . . . . . . . . . . .1. . . . . . . . 220 217 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11. . . . 200 10. .4. . . . . . . . . . . . . . . . . . . 217 12. . . 9. . . .4. . . . .1 The Delta Form of an Iterative Scheme . . . .2 The Model Equations . . . .1 Motivation . . . . . . . . . . . .2 Properties of the Iterative Method . . . . . . . . . . . . .2 The Gauss-Seidel System . . . . . . . . . .2 Classical Relaxation . . . . . . . . . . . . . . . .1 The Concept . . . . . . . . . . . . . . . . .1. . . . . . . .

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

C THE HOMOGENEOUS PROPERTY OF THE EULER EQUATIONS265 .

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

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

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

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

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. 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. the variable u can denote a scalar Cartesian velocity component in the Euler and Navier-Stokes equations. however. As a word of caution. usually a vector A term of order (i.S.S.e. it should be noted that. such as velocity. 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.4 Notation The notation is generally explained as it is introduced. Some of the abbreviations used throughout the text are listed and de ned below. there are many aspects of practical problems which can only be understood in the context of the complete physical systems.4. S. a scalar quantity in the linear convection and di usion equations. Otherwise. and a vector consisting of a collection of scalars in our presentation of hyperbolic systems. 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) . Bold type is reserved for real physical vectors. PDE ODE O E RHS P. 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.1.. For example. 1.

6 CHAPTER 1. INTRODUCTION .

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

1 can be rewritten as d Z QdV + I n:FdS = Z PdV (2. the region of space. 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 . momentum and energy for a uid.4) r: i @x @y @z and i j.6) .3 and nd a solution for Q on that basis are referred to as nite-di erence methods. On the other hand. in Cartesian coordinates. or cell. leading to @Q + r:F = P (2. respectively. In two dimensions. 2. 2. Many of the advanced codes written for CFD applications are based on the nite-volume concept. 2. and k are unit vectors in the x y. t1 . then Eq. 2. For a Newtonian uid in one dimension. a partial derivative form of a conservation law can also be derived. or tensor. 2. The divergence form of Eq.2 The Navier-Stokes and Euler Equations The Navier-Stokes equations form a coupled system of nonlinear PDE's describing the conservation of mass. and P is the rate of production of Q per unit volume per unit time.5) @t @x with 2 3 2 3 u 3 2 0 6 6 Q=6 6 6 4 e 7 7 u7 7 7 5 E 6 6 =6 6 6 4 u(e + p) 7 6 7 6 2+p 7.2 is obtained by applying Gauss's theorem to the ux integral. by ! @ +j @ +k @ : (2.2) dt V (t) S (t) V (t) Those methods which make various numerical approximations of the integrals in Eqs. is an area A(t) bounded by a closed contour C (t). The vector n is a unit vector normal to the surface pointing outward.6 u 7 6 7 6 5 4 4 @u 3 @x 4 3 u @u + @x @T @x 7 7 7 7 7 5 (2. Those methods which make various approximations of the derivatives in Eq.3) @t where r: is the well-known divergence operator given. If all variables are continuous in time. and z coordinate directions.1 and 2.2 and nd a solution for Q on that basis are referred to as nite-volume methods.8 CHAPTER 2. F is a set of vectors. they can be written as @Q + @E = 0 (2. containing the ux of Q per unit area per unit time.

p is the pressure. e is the total energy per unit volume. . This form of the equations is called conservation-law or conservative form. The steady solution to the one-dimensional Navier-Stokes equations must satisfy @E = 0 (2.8) @t @x @y with 2 3 2 3 2 2 q1 u 3 v 3 6 7 6 7 6 2+ 7 6 7 Q = 6 q2 7 = 6 u 7 E = 6 u uv p 7 F = 6 v2uv p 7 6 7 6 7 6 7 6 (2. In order . THE NAVIER-STOKES AND EULER EQUATIONS 9 where is the uid density. Tannehill. and Pletcher 1] and Hirsch 2]. Later on we will make use of the following form of the Euler equations as well: @Q + A @Q + B @Q = 0 (2. they can lead to quite di erent numerical solutions in terms of shock strength and shock speed. and is the thermal conductivity. u is the velocity. such as u and p. such as the ideal gas law. Although non-conservative forms of the equations are analytically the same as the above form. for example. with no interest in the transient portion of the solution.7) @x If we neglect viscosity and heat conduction. we are often interested in the steady-state solution of the Navier-Stokes equations. Note that the convective uxes lead to rst derivatives in space. while the viscous and heat conduction terms involve second derivatives.9) 4 q3 5 4 v 5 4 5 4 5 + 7 q4 e u(e + p) v ( e + p) where u and v are the Cartesian velocity components. Thus the conservative form is appropriate for solving ows with features such as shock waves. For such ows. In two-dimensional Cartesian coordinates. T is the temperature.10) @t @x @y @E @F The matrices A = @Q and B = @Q are known as the ux Jacobians. the Euler equations are obtained.2. these can be written as @Q + @E + @F = 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. v. Many ows of engineering interest are steady (time-invariant). These equations must be supplemented by relations between and and the uid state as well as an equation of state. and p. 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. The ux vectors given above are written in terms of the primitive variables. u. Details can be found in Anderson.2.

q3 . 2 q4 q 3 q1 2 q3 q1 . (u2 + v2)=2] from the ideal gas law. q2 .1 q2 2 3 2 q1 3 7 7 7 7 7 7 7 7 2 7 . 2 q1 2 q2 q1 . cp=cv .1 q2 + q3 2 2 2 q1 2 q1 q1 q2 (2. as follows: 2 6 6 6 6 6 =6 6 6 6 6 6 4 E1 E2 E3 E4 E 2 3 6 7 6 7 6 7 6 7 6 7 6 7=6 7 6 7 6 7 6 7 6 7 6 5 6 6 4 ( 3 7 7 7 2 2 7 3. 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 . and q4 . ) q3 q1 q2 q1 0 A= a21 q .11) F 2 6 6 6 6 6 =6 6 6 6 6 6 4 3 2 F1 7 6 7 6 7 6 6 F2 7 6 7 6 7=6 7 6 7 6 F3 7 6 7 6 7 6 5 6 4 F4 q3 q3 q2 q1 ( . 1)q4 + 2 q1 2 q1 7 7 7 7 q3 q2 7 7 q1 7 7 7 5 3 2q q4 q2 . CONSERVATION LAWS AND THE MODEL EQUATIONS to derive the ux Jacobian matrices.13) where a21 = . q2 q1 q1 3 a41 ! . 1) e . 2q . q1 .1 q3 7 . . 3 . + . we must rst write the ux vectors E and F in terms of the conservative variables. .1 0 a42 !2 a43 q2 q1 3 7 7 7 7 7 7 7 7 7 7 7 7 7 5 (2. q2.1 q2 7 2 q1 7 7 7 5 3 q3 2 q1 (2. where is the ratio of speci c heats.10 CHAPTER 2. ) q2 q1 q q3 1 0 (1 . 1 q3 2 . 1)q4 + 3.12) We have assumed that the pressure satis es p = ( .

aspects of convective phenomena associated with coupled systems of equations such as the Euler equations are important in developing numerical methods and boundary conditions.2. 1) . 1)4 ! 11 q2 q1 !3 + q3 q1 2 !2 !2 q2 q1 !3 5. . Thus we also study linear hyperbolic systems of PDE's.14) 3 7 7 7 7 7 7 7 7 7 7 7 5 0 1 0 0 0 a21 (3 . 1) q q 1 1 and in terms of the primitive variables as 2 6 6 6 6 6 A=6 6 6 6 6 6 4 + q3 q1 (2. The eigenvalues of the ux Jacobian matrices are purely real. )uv (2.15) u where a41 = ( .uv a41 a21 = v a42 u a43 (2. THE NAVIER-STOKES AND EULER EQUATIONS a41 = ( 2 . 1 43 q2 a42 = q1 2 q1 ! ! q2 q3 a43 = . 1 v2 .( . . The Navier-Stokes equations include both convective and di usive uxes. 1)u(u2 + v2) . ue a42 = e . )v ( . This motivates the choice of our two scalar model equations associated with the physics of convection and di usion. . 3 . 1 (3u2 + v2 ) 2 .16) Derivation of the two forms of B = @F=@Q is similar. which are further discussed in Section 2.2. )u (1 . The homogeneous property of the Euler equations is discussed in Appendix C. Furthermore. u2 2 2 a43 = (1 . This is the de ning feature of hyperbolic systems of PDE's.5. !2 3 5 q4 q1 ! q2 q1 ! q4 .

3 The Linear Convection Equation 2. while the right-hand boundary is the in ow boundary when a is negative. the ow being simulated is periodic. Hence. simply. 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. Now let us consider a situation in which the initial condition is given by u(x 0) = u0(x). eventually returning to its initial position. This is consistent in terms of the well-posedness of a 1st-order PDE. the scalar quantity u is given on one boundary. the \out ow" boundary. and the domain is in nite. (2) In the other type. we pay a great deal of attention to it in the initial chapters. corresponding to a wave entering the domain through this \in ow" boundary. At any given time. as the convection problem. the process continues forever without any . 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. With periodic boundary conditions. a real constant which may be positive or negative. the waveform travels through one boundary and reappears at the other boundary. Hence the wave leaves the domain through the out ow boundary without distortion or re ection. at) (2. Note that the left-hand boundary is the in ow boundary when a is positive. It is easy to show by substitution that the exact solution to the linear convection equation is then u(x t) = u0(x . This type of phenomenon is referred to.3.12 2. It represents most of the \usual" situations encountered in convecting systems. what enters on one side of the domain must be the same as that which is leaving on the other.1 Di erential Form CHAPTER 2. No boundary condition is speci ed at the opposite side. without special consideration of boundaries.17) @t @x Here u(x t) is a scalar quantity propagating with speed a. This is referred to as the biconvection problem.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. It is the simplest to study and serves to illustrate many of the basic properties of numerical methods applied to problems involving convection.

ia t (2. a.22) Substituting f (t) into Eq.ia f (2. the solution can be written as u(x t) = f (t)ei x (2. and the phase speed.2. Preserving the shape of the initial condition u0(x) can be a di cult challenge for a numerical method. THE LINEAR CONVECTION EQUATION 13 change in the shape of the solution. must be an integer. Substituting this solution into the linear convection equation. We now examine the biconvection problem in more detail.3. and is the wavenumber. This means that the phase speed is the same for all wavenumbers. Eq. are related by != a (2. Let the domain be given by 0 x 2 . in a simulation. .24 is characteristic of wave propagation in a nondispersive medium. For M modes. As we shall see later. It is a measure of the number of wavelengths within the domain. the wavenumber.19) where f (0) is a complex constant.23) where the frequency. 2. With such an initial condition.17. We restrict our attention to initial conditions in the form u(x 0) = f (0)ei x (2.24) The relation between the frequency and the wavenumber is known as the dispersion relation. most numerical methods introduce some dispersion that is.!t) (2.20 gives the following solution u(x t) = f (0)ei (x. !.20) where the time dependence is contained in the complex function f (t).2 Solution in Wave Space u(x 0) = M X m=1 fm (0)ei mx (2. one obtains 2. An arbitrary initial waveform can be produced by summing initial conditions of the form of Eq. 2. The linear relation given by Eq. 2. waves with di erent wavenumbers travel at di erent speeds.3.25) . 2.19.21) dt which has the solution f (t) = f (0)e. In order to satisfy the periodic boundary conditions. we nd that f (t) satis es the following ordinary di erential equation (ODE) df = .at) = f (0)ei( x.

2. In contrast to the linear convection equation. In the case of Eq.27. with u representing the temperature. CONSERVATION LAWS AND THE MODEL EQUATIONS where the wavenumbers are often ordered such that 1 2 M .26) 2. the steady-state solution must satisfy @2u = 0 (2. the solution is obtained by summing solutions of the form of Eq.30) . Since the wave equation is linear. or mixed Dirichlet/Neumann. Other steady-state solutions are obtained if a source term g(x) is added to Eq. 2. as follows: @u = @ 2 u (2. For example.27. Neumann (speci ed @u=@x).23.4.27) @t @x2 where is a positive real constant. 2. Dirichlet (speci ed u). g(x) # (2.29) giving a steady state-solution which satis es @ 2 u . which is one that satis es the governing PDE with the partial derivative in time equal to zero.1 Di erential Form Di usive uxes are associated with molecular motion in a continuum uid. g(x) = 0 @x2 (2. the di usion equation has a nontrivial steady-state solution. this parabolic PDE governs the di usion of heat in one dimension.28) @x2 @u = @t " 2 @u @x2 . Boundary conditions can be periodic. A simple linear model equation for a di usive process is Therefore. u must vary linearly with x at steady state such that the boundary conditions are satis ed.at) (2. giving u(x t) = M X Dispersion and dissipation resulting from a numerical approximation will cause the shape of the solution to change from that of the original waveform. m=1 fm (0)ei m(x.14 CHAPTER 2.4 The Di usion Equation 2.

4. Let the initial condition be u(x 0) = M X m=1 fm(0) sin m x + h(x) (2. 2 f (2.4. ua)x= . 2 t (2. THE DIFFUSION EQUATION 15 In two dimensions. The corresponding steady equation is @ 2 u + @ 2 u . 2. 2. h(x) = ua + (ub .33) where must be an integer in order to satisfy the boundary conditions.31) @t @x2 @y2 where g(x y) is again a source term. 2 t sin m x + h(x) (2.34.31 is parabolic. i. The latter is known as the Poisson equation for nonzero g. 2. the di usion equation becomes " # @u = @ 2 u + @ 2 u . 2.27.36) Substituting fm(t) into equation 2.37 shows that high wavenumber components (large m) of the solution decay more rapidly than low wavenumber components.e. and as Laplace's equation for zero g.2 Solution in Wave Space We now consider a series solution to Eq. 2. g(x y) = 0 (2.35) m m dt and we nd m fm(t) = fm(0)e.32 is elliptic. g(x y) (2. consistent with the physics of di usion. . Let the domain be given by 0 x with boundary conditions u(0) = ua. Eq.27 gives the following ODE for fm : dfm = .. A solution of the form M X u(x t) = fm(t) sin mx + h(x) (2. It is clear that the steady-state solution is given by a linear function which satis es the boundary conditions. we obtain m=1 u(x t) = M X m=1 m fm (0)e.32) @x2 @y2 While Eq. 2.2.37) The steady-state solution (t ! 1) is simply h(x). Substituting this form into Eq.34) satis es the initial and boundary conditions. u( ) = ub. Eq.

Other systems of equations governing convection and wave propagation phenomena. . such as the Maxwell equations describing the propagation of electromagnetic waves. The entries in the ux Jacobian are @f aij = @ui (2. this can be rewritten as @w + @w = 0 (2. the equations have been decoupled into m scalar equations of the form @wi + @wi = 0 (2. 2. and X is the matrix of right eigenvectors.1u.1AX (2.38) @t @x where u = u(x t) is a vector of length m and A is a real m m matrix. we obtain =0 (2. this equation can also be written in the form @u + @f = 0 (2.38 by X . form a hyperbolic system of partial di erential equations.39) @t @x @f where f is the ux vector and A = @u is the ux Jacobian matrix. CONSERVATION LAWS AND THE MODEL EQUATIONS The Euler equations. postmultiplying A by the product XX .8.1u z }| { @ X . For conservation laws.1 are constants.1u See Appendix A for a brief review of some basic relations and de nitions from linear algebra. are also of hyperbolic type.1 . 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 Thus = X . Eq.16 2. and noting that X and X .5 Linear Hyperbolic Systems CHAPTER 2.2.1. 2. Such a system is hyperbolic if A is diagonalizable with real eigenvalues.43) @t @x When written in this manner.44) i @t @x 1 @X .1AX X .42) @t + @x With w = X . Premultiplying Eq.41) where is a diagonal matrix containing the eigenvalues of A.40) j The ux Jacobian for the Euler equations is derived in Section 2.

all of the wave speeds have the same sign. in a supersonic ow. In a subsonic ow.6 Problems 1. While the scalar linear convection equation is clearly an excellent model equation for hyperbolic systems. of course. leading to three equations in the form of the linear convection equation. The speed u is associated with convection of the uid. The characteristic variables each satisfy the linear convection equation with the wave speed given by the corresponding eigenvalue. Based on the above. wave speeds of both positive and negative sign are present.1 7 4 5 0 p u . PROBLEMS 17 The elements of w are known as characteristic variables. 2. c are associated with sound waves. c. we must ensure that our numerical methods are appropriate for wave speeds of arbitrary sign and possibly widely varying magnitudes. or wave speeds. while u + c and u . and u . which is the inow boundary for these variables. the boundary conditions can be speci ed accordingly. The signs of the eigenvalues of the matrix A are also important in determining suitable boundary conditions. That is. we see that a hyperbolic system in the form of Eq. where u is the local uid velocity. Each characteristic variable satis es the linear convection equation with the speed given by the corresponding eigenvalue of A. While other boundary condition treatments are possible. Therefore. 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 M = 6 0 u . are q u u + c. corresponding to the fact that sound waves can travel upstream in a subsonic ow. they must be consistent with this approach. The one-dimensional Euler equations can also be diagonalized.2. The eigenvalues of the ux Jacobian matrix.38 has a solution given by the superposition of waves which can travel in either the positive or negative directions and at varying speeds. Characteristic variables associated with negative eigenvalues can be speci ed at the right boundary. where juj > c. 2.6. Therefore. which corresponds to in ow for these variables. although they remain nonlinear. and c = p= is the local speed of sound. where juj < c. characteristic variables associated with positive eigenvalues can be speci ed at the left boundary.

are related by a similarity transform. (Hint: make use of the matrix S = @Q=@R. that is. fm (0) = 0 for m > 3.25. Plot the rst three basis functions used in constructing the exact solution to the di usion equation in Section 2. 9.e.) Next consider the same initial condition de ned only at x = 2 j=4. 2. 2. j = 0 1 2 3. Find the eigenvalues and eigenvectors of A. Given the initial condition u(x 0) = sin x de ned on 0 x 2 . 2. Find its eigenvalues and compare with those obtained in question 2. 2. Find the eigenvalues and eigenvectors of the matrix M derived in question 1. 7. Plot the initial condition on the domain 0 x .18 CHAPTER 2. Plot the solution at t = 1 with = 1. CONSERVATION LAWS AND THE MODEL EQUATIONS Assume an ideal gas. Derive the ux Jacobian matrix A = @E=@Q for the 1-D Euler equations resulting from the conservative variable formulation (Eq.2. write it in the form of Eq. 3. 4. in the form of Eq. nd the necessary values of fm (0). Write the classical wave equation @ 2 u=@t2 = c2 @ 2 u=@x2 as a rst-order system. (Hint: use M = 2 with 1 = 1 and 2 = . 8. 2.1. Write the 2-D di usion equation.5)..31. 1.33 with fm(0) = 1 for 1 m 3. @x + @u = 0 @y . 6. 2.2. and initial conditions from Eq. u2=2). respectively. 1)(e . Show that the two matrices M and A derived in questions 1 and 3. Next consider a solution with boundary conditions ua = ub = 0. 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 ! = . p = ( . i. Find the values of fm(0) required to reproduce the initial condition at these discrete points using M = 4 with m = m . in the form @U + A @U = 0 @t @x where U = @u=@x @u=@t]T .4.) 5.

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

20 CHAPTER 2. CONSERVATION LAWS AND THE MODEL EQUATIONS .

¡ ¢  .

¦   H ‰B d 4 Y H h 5 D h H h H H 2 2 5B D T D 5 2 h 2 T Y 7 d H D H F D H 2 n H h 5 D Y R Y DB 7 d H  H d R ` n YB F D q 5 2 5B †X‚QW6IkGÆQI6QIÅ8EGVG¢I¬I¢–9W6GeIW‰II6IkG‘IwGE9W6s‚I8U…EIG‘¢Å8EGD r 4 H  Y 2B 2 i ’ ¥ H d R `B ¤ 2B 2 @ 5 F Y YB H 4 T  Y h 2 T H 7B D F D 5 ‰ ` 2B by 5 b 2B F Y H 7 D Y Hy  7B Y H F –Q†WIEe™j’   p™WI8EvX‰CcIWvXuQVUWeIV9XGŽW¢†…IXx„88oX‚Q9Š‚6XI9X‚Iut ¶ à ¹ » ´ » à 9Q”av–Ä ± Á ¶ ± À ± ¿ ¹ ½ ¼ ± » ¹ ¶ ¹ ‘ej)6¾hfv6egº¸ · ¶ ´ ² ± ³ ² ewµ”± ° ¯ ® c­ ’ 2 5B D T 4By   T y T 4B D 4 T d ‹8EGVQXEII¢›VgXW6¢WI ` 2B Y 5y d 5 H bB D 4B d D Y H d iy R h 2 R ` 2B H ‰ D R 5 F DB IX‚8X98‰8EW6EWGfQ‚¬EIIpIrIEQ†9p8IWEC@ QGfx‚«VƒW6Gf¢ŠªQ6¢U‚XISQ©¢‚Iv@ 7 H DY i Y 2 TB Y H Dd T  h H 4 T  YB R P H 2 T H Y R H YB Y iy T 2 T i 4 T d R 4 4 T ` 2B @ 5yy 5q H F D q 5 F 4 R 7 2B t 2 5 Y T H d YB F D d 5 ¤ ’ Y 2 5B D T R P H h H 7 d 5q Y 2 T d XWx„¢U¢6VGI6Q¢IECcXX¢•9IWŠ¢ƒI…AX§‹8WVgWwXpG8UeQp8XWVUSQAQ9W8•fI¢WGD H F D q 5 Y D 2 HB 4 ¨ H 5 4 Hy ‰ TB d T b 5 D 2B h H ‰ d 5 Y ‰ T YB 2 5B D TB d T b 4B d D H 7 5 H ` H F D yy T ’ 7 d 5qB 2 R YB }   œ pGpVvGo6XQ9‡˜6”XIVƒWV¢CGoEwQ†W¢WI¢§E8XWVƒ‚¢V”XWW698Q¢vIW¢XV¢‰W8•„II˜Xhƒxp3ž  ¢†lŸš¢£x†™lƒIu§8E˜”Wp8XvuEwCcI‚›¢s‡ƒxp3V¢†lŸš¢£x†™lƒ#QEX¢Q3V‚uuX‹E¢V‚QWoXB  „ ‡yx x ¢ œ { ‡  H F t ’ ¦’ ¥ H d R `B ¤ 2B 2 @ 5 F Y Y T t }   œž    „ ‡yx x ¢ œ { ‡  h Hyy T 4r 5 Y T 2B Yy T b d H D 2 h H 4 T  YB R P H F DB @ F Y H 7 2 TB Y H D d T  7 d 5qB 2 R T 5 D h H 7 d 5q Y 2 T d D H ‰ 2 T 4 }   œ ž    y ž  ‚ œ 2 Q6¢U‚XISQG„CW6#¢ƒ‚QW‚¢Š¡‚8•EpI††GAQ9W8•fI¢GW9†rVgƒxp3v¢WŸ‡cxUeXB Y H D T 2B h d 5 5 4 d T H 2ByB b d R 4 y T d H 2 H ` 2B h H Y Y H d  — H Y H F Y H 7 q 5 Y H Y Y Ty 4 y T d H 2 H ` H DB R P D T F D D 4 T 6GVIXI‚8˜6Š¢QIXEXx‚I6›¢G6IQ¢‰EeQ‚WQ‚I˜‡‘QI‚Q9vVf6WWVƒQ›VGQpQ8uG„ISCVUW‘‡¢šq H F D H ŒB Y T F  7 H H ™ ’ 7 H D Y i Y 2 TB Y H D d T  T 2B h H D 2 H Y H d  YB @ 5y H ‰ y TB d H D T 7 H F D q 5 yy pG…QEW¢II#6…w‰QW‚x‚e¢XW6G‚VŠeXAQWS6W6WIEsEQ†˜¢ƒ‚QGld‘IWj¢vE™i ’ — d H D  T F  2B h H Y Y R 4 YB h H ‰ yyB @ Y 2 5B D T 7B — 5 d   T H 4 2 H d H gB h r H DB 2 n H Y H F If6GpVU•‰X6W‚IQ‚XIu†hXEC•I8EGV#X˜gWpI¢uQpQW6h„I3–G„IUuWQpGD ` 2B iy   T d 5q Y HB ` H D T d D ” ’ “ d H D  T F  2B h H D T H d D H d T Y H bB D T bB d H h H 7B D ’ Y H bB D T bB d H h y TB D T  pXxEIp¢u8••6X86GVWGx€˜CQG˜¢U•‰Xe6GVQWWW¢u68„GVVE‚QIf#EGp6Q¢EGl¢„W6Ih¢XGVIWY 2 5 YB d H D  T F  YB F D 2B YB Y T F  7 H d R “ ’ Y H bB D T bB d H h H 7B D d 5 Y H bB D T bB d H h y TB D T  Y 5 D d H F DB 8‘X™QG˜¢U•EIG‰XfXWVUI9Q™p™Q68„GVVE‚QIw9XGC8fQ8„GVV„WQpm¢XGVIWdWf6IG„QH h HBy   T H ‰ 2 T 4 H Y H F t ’ h H D 2 H Y H d  YB Y H bB D T bB d H h y TB D d T  5 D Y 2 5B D T 7B — 5 d   T H 4 2 H d H gB 6XEII¢#sr¢gŽ‚QIu†‹QWoQW6WIuEwQ¢EWVVEW6Ij¢XG‚VUWI8XWV#X˜gWIIVŽQpQW6h„Ih r H DB 2 n q 5 D  H 4 2 5 4 H F D t d H D  T F 4 YB F D 2 1 ’ H d H F h H ŒB Y T F  7 H F 4 T 5 d   T H F D YB YB F t ’ h H by 5 ‡G„IUm¢p6QI¢QwIGUQ6GpVUƒfEIG3ŠVWQpQ6X‚¢UI9Q†ƒ¢oWpI¢aIGuXCEIuŠ‹68E¢WY H ‰ 2 T 4 F 4B F @ Y 2 5B D T R P H H 4 2 H d H gB h q 5 D H Y T Y H 4 R h 5 d  F 4B F @ ˆ ‡ ‚x } { … „y ‚  €  { ~ } {yx †rVg)ƒEICwp8XWVUSQ9QI6W6h„IŠ¢w6‚‰wQQpI˜‚IrƒEIC9QU|‡d†x|xƒG¢93‡d|zkT 5 D ` 2B h T Hy t D — H 2 h H D T 7B — 5 d   T H d T Y H bB D T bB d H h H 7B D H F t ’ Y 2 5B D T R P H y TB D 2 H d H gB h i d T 2B WwIEI¢QXpQx‡I#QGldEpg‚II¢™W¢vQ¢EGl¢„W6If9XWfIuj6I8EGVUS6†VƒGo6WQsEp–‚¢UEIh r d 5 q 5 7 H D Y i Y T ` 2B 4 R h 5 d  i ‰ H d H F D t D Y d n Y H bB D T bB d H h y TB D T  Y H F D H D T 7B — 5 d   T 2 T 4 H 2 f8ˆ¢pQGfx‚depXQpI˜‚Io6WQpGhQ‚WI™Q¢EGl¢„W6ImVƒGlUW…pGuGV#X˜gWIp¢k¢gwI85 t8„Q8„GVIWQWEfjQI8EGVIx6QpQW6h„IwpXG„IW6WfIG™¢E8‚eQIWdI¢9„WI¢QI¢GEIwXW™WQpGD iy H bB D T 2 d H Dy i ’ Y 2 5B D T R P H H 4 2 H d H gB h ` 2B Dy R Y H d H F D H by 5 Y 2 H F D h 2 T iy Y R 5 H 2 T Dy R 7B Y H Y H F H D T 7B — 5 d   T 2 T 4 H 2 “ ’ H 7B D h 2 T H 4 T  Y F D 5 ‰ 5 D D 4 H  Y H d F DB @ Y H bB D T bB d H h y TB D d T  2B T D 2 5 GV#X˜gWIp¢–¢g•I”V#EG…I¢C6¢U‚‘G¢†Gˆ‡Q†WQ‚…G„CvQ¢EWVVEW6Ix¢ƒW‚¢UwX¢GS¢Q4 Y 2 5B D T R P H y H h 5 7 d R 5 t @ 5 q hB R q i h T H D Y 2 R ` 2B 2 d H b 5 ` Y 2 5B D T R P H H F D F DB @ 2 5 7 7 5 4 2 p8XWVUSQ€QIx9wI8vusIrXIIpVgGfIIeIEIW68c8aI8XWVUSQ#IGGECA89#86)31 0 £   ' ) $  %     ( " #! ' $ &   &    %   © ¨ ¦ §¥ ¤ £ .

’ D 2 H 2 5  — H 2 QoQI8†˜‡‰¢T ¢W•Š„‰‚XI¢IXW‚Xp9Wf6WQpGoQ‚¢U 7 5 dq DB F YB R ` 2B D YB h 5 D Y H Y H F D 2 H d T F DB @ h H Y 5y 4 2 H iyy T R Y R YB h 2 T D B d 4 Y d H  R Y T Y i T @y T YB H 7B D t d H F D H ` 5 D h H Y R 2 H F @ D R WECÅQ‚8X6IQEE¢UWp…X¬I¢9˜XW6W‚Q†IW–xgŠ„¢‘EŽ#EGjQ6IG‡88W¬6WIÆQpC‘pI‰ t D B d 4 Y ‰ R Y T Y T d T H   T Y H 4B h 2B H 4 T  Y h 2 T H 7B D H F D F D 5 ‰ t H 2 5y T h H Y R 2 H F @ D T F D H 4B D 5 QpX‚Q‚IIW‘C¢‘¢Qsp¢uQ6XIpXwQ¢U‚I¢9XWupGeG¢†pVI8E¢6WI‰QICulUG6XGVfò ä ç ñ õ ø Þà ß î þ ø ßæ ¡†p«˜‚˜‘¬‘|mÝ ä ç ß î þ ø ßæ ¾s…¬‘|vÝ ä ç ñ õ ø Þæ ¡†p«‹•mÝ ç ’ ¥ 8¥ pšæ   ü ¢í ¡‚£Ý ¤ý ü ¥Icï   ü í Ý ä ç ß îû þ ø ù æ ¡‚xÿˆs‘XŽÆ™ê ‚mÝ ý ü ï Ý ä ç ñ û õ ø ìù æ IgUðˆë|x©9ú‚mÝ Û Y @ 5yy 5q Y T YB ` 2B — H h 2B D B d 4 Y d H  R Y h 2 T D B d 4 Y ‰ R Y d 5q 2 5B D 2 H b 2 5 4 H F QCcXE8•u¢CX–IE˜‡IIE‘pX‚QW‚Q†IW‰I¢fpEW6WII‚u8•†¢XGo68o8QwIut ’ Y R 5B b ‰ 5 ` 2B 2 T H 7 H F D H ÷ T 7 hy R 5 F 6I8E˜o8‘pXI¢Q#uIW¢¢#XI8pWY Dx‡Wo8Qu¢Q˜EC‡p‚¢WW6QQp96#EG‡#8‚#EºƒXpC•e6W8†WpI#6IGV#8•©QWp†‚¢ƒs–8£vì — H D 2 5 4 y T 4 5y t ç i d T Y Y H 4 H 2 Y H 7B D H 7 5 Y YB F 4B F @æ Y H Y 5  d R  d H F D 5 d 5q h H Y R H d T ö à õ à êà 2 H F ™ ’ i T @ d TyB 7B Y T 2B â h 2 T á d 5q h H Y R H d T ö h 2 T õ d H D T ô ’ D R 5 F ` R 5 d F D Þ î ä ñ D T F Qpª•8g•v¢XX9XW–†X#†I¢Ž‘¢•Q‚IuW¢j†I¢6GV˜vgp¢I8I8‚IG•‘óe™VIGD HD 5 ò W¢f‘’   pwWp8Xv†XeCcI‚™¢s–XdpXQVUWaIGu…p¢A†EdIEQ¢U‚wIWuXk…‘W6IC@ ’ ¥ H d R `B ¤ 2B 2 @ 5 F Y Y T t Þ 2B ` 2B 4 T  Y H F D Þ î h 2 T ß 2B ` 2B 4 T  Y H F D YB ß î H d H F ç   pšæ ’ ¥ ç ¦’ ¥ cEpšæ ñ ì ä Þ î ì ä ï ›ª€…‹ða3Þ ß î ê ä í …S«uxß ä ä ¡Þ ß hB d ` h H 4 T  YB R P H 2 T 2 5 2 H F t ’ ì ž   è   YB X‚86Q¢IWXpx6¢¢Qpujv•s¢€é¢wEŠÞ d 5q D T F D h 2 T t ê ž   è   YB ß d 5q — H h 2B F Y H 7 H F t ’ h H 7 R Y Y T YB d H F D 5 H F D 2 5 H 4 2 H h 2 H  H ¢•uVUG‰I¢ëUˆc¢€é¢wXC¢•e‡IpX‚Q9apu•‹6#I‚W¢CEf6IGVwpG8…QpQII6s6Ih t h H D 5 2 H h YB Hy ‰ TB d T b H 2 5 iy 2 5 2 H F ™ ’ ç Þ à ßæ Ý ä Ý H Y T 4 ã r ¦ H F D iyB d T 7B d  d H hB Y 2 5 4 H hQGVIQI™XwXI¢XW¢Vwp8„I8kQpªu–‚W˜•måe‘W¢QÏÔ6eIWd„XWVdEWI6IX‚I8Q‘v@ Y d H D  T F 4 y T d H b H Y D Y d n H F D d 5 ¤ ’ â à á à ß h 2 T t Þ Y Hy ‰ TB d T b D 2 H h 2 H  H h 2B H F D q 5 Y 2 5B D 4 2 Rq H d W6GpVUƒh¢WQ¢6W•‚‚UfIWj8Ujpƒ†G˜dIVp–jQEI¢XW¢VŠoQpIQ†QIpX™pG‹¢vI8EG6pI•‘‚¢T t Hy  7 T — H d 5q t Ý t Y Hy ‰ TB d T b D 2 H h 2 H  H h H F D ty T d H 2 H ` 2 1 ’ Y 2 5B D T 7B — 5 d   T H 4 2 H d H gB h H DB 2 ¢Xp#¢p–Š8•˜˜˜QQEI¢XW¢VCoQII6s6IIW˜ŸVGQpQ83”6I8XWV#X˜gWIIVuQI6WQsEpG„IUn ’ Y H 4 T  Y y T 2 5B D T D R  7 5 4 h 2 T y T 4B Y i F Q6Q¢IWhVU8XWVƒ˜I#¢Q‰I¢˜¢QXWxojÜ Û ¦’ ¥ H d R `B 8„puWI¢Xv¤ Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u‡§™†fjˆˆ«”w€‘ˆ””Ԑz€mQ‘Ôv«Sψm™‘”Ç y x   8  .

’ 7 d 5qB 2 R YB ` 2B 4 T  Y H F D t ß î d 5 Þ î 2 5 D B d 4 Y ‰ R Y 5 2 YB H d H F D 2 H F †‚8•EpIuX–IEQ¢U‚uIGU˜…C8Š…‰8fpX‚Q‚IIW–IuEa‚QIW‰6Iª™ ç 9’ ¥ ¢%pšæ ’ 4D ¢W6H à ï Ý 6 4 ü ï Ý ä ï Ý ˜UÅ85cUªaë‘î à í ß 6 4 üí ß ä í ß sxr'5‚xªCx…î à ïÞ 6 4 ü ï I3r75c3Þ ä ïÞ a3…î D T F D F 4 R Y h H 2 n H h YB F 4B F @ t î y 5 ‰ 7 i VUW†ƒI‚QpU6IuE†ƒEICIe˜8†…x‚Y H F D YB 2 5B D  H 4 — H H 2 5 H F t ’ ` 2B 2 T H 7 H YB 4 H d  Y DB H 2B 7 d H D H h 5 D h H Y R H d T Y i T @ d H F D “ ’ pGXk8EGp6Q˜‡‘p8‘IuuIIEI¢g9‘WEQ6WIWE…pX9WQW6IWk6WI‘‚¢wxc•QpG™…S( y 5 ‰ 7 i Y H F D 2B h HB d d T 4 D 5 2 YB 2 5B D T 7B — 5 d   T H F D q 5 ç d H h d 5 h 2 Tæ H d R D T 2 H YB 4 H d  H F D D R 8sw˜fuIGŽE‰6X‚W¢QC¢IuXd8XWVdE˜c‚II¢upG§¢•ƒ6IW8‰I¢š9WIWVUuWEQ6WIuIWupI‰ ç ’ ¥ 8“ pšæ   0  ¥21¥ ( à 321 (   0 t Hy  7 T — H d 5q t D T F D F 4 R Y H bB D T bB d H h T 5 D 2 5B D T 7B — 5 d   T H 4 2 H d H gB h ¢Xp#¢p–u8•pglUG‰ƒIWu8„GVV„WQp–9Wd8XWV#X˜gWIIVw6IQ‚QsEI‘T D 2 H Y H d  H d 5 D h H Y R YB ( y 5 ‰ 7 i Y H F D D T F D 2 T H 7 H @ YB F D i & ’ H R PB 2 R D 5 2 YB DB ç 2 5B D  H 4 — H H 2 oQ‚QWpQW‘GaQWp§Eu)¢s…xfvIWmlUGu¢g9•v›XIWv'VISXIpv¢p§E§E›ë8XWpQ6˜‡€I85 F DB @æ D R ‰ i F  5 Y 5yB F  H 7 T Y H F D Y @ 5yy 5q Y 2 5B D T 7B — 5 d   T H 4 2 H d H gB h d 5q 2 5B D T D 5 2 H F WEC•pI…Sp8W8EXIpf9¢G‘IW”CsEX8•”p8XWVdE˜c‚II¢™QI6W6h„Iˆ8•‘8XWVƒ¢p™Iut ’ ß 5 D D 4 H  Y H d F DB @ Ý q 5 H bB D T bB d H h D Y d n H F D D 2 H Y H d  H d 5 D h H Y R H ‰ D 5 2 yyB ˜9W‘‡Q†W6W‰GEC‰vV8EWVVE‚QI‘‚‚UIGfoQW6WI6W–G‰Q‚Is‘¢phXEC@  ‘ISut Ý Y R F ’ Y H bB D T bB d H h Y Y H d  — H 5 D h H Y R d H b H 2 H d T Y Hy ‰ TB d T b D 2 H h 2 H  H h 2 5 Y D B d 4 Y ‰ R Y D — H D YB F D 2 QQ¢EWVVEW6IuW6WI˜‡dGQ‚IQ¢6IaW¢fQEI¢XW¢VoQIpQ†QI¢GpEWQ‚II‚x‡GCXpG31 ç $’ ¥ ¢%pšæ Þ Ý " " ä #˜!Ý H Y R H @   㠓 q 5 i h R D Y H F D 2B H bB D T bB d H h H 7B D i d T 2B h d 5 H F D d 5 ‚Iu•uY u”¢ppG‚uIWeX8„GVV„WQp#EG#‚¢UEIW¢wpGŠ8U¤ ç ’ ¥ o— pšæ ’ 4D ¢W6H à  ß  Ý   ä   eÝ ¥ à Þ  Ý  ä  eÝ  à ß  Ý  ä  eÝ  iy ‰ T H ` 2 T F 4 d H D 2B H Y R H @ H 7B D d 5 H 4 T  Y 2B Y H bB D T bB d H „I¢Q8I¢Uƒ‚QWSEa‚Iw•™9XWŠ8wQ¢U‚eXCQ¢EWVVEW6Ih y TB D d T  d 5q Y R F t ’ Y 2 5B D 2 H b 2 5 4 y T R Y R H F D 5 D ` 2B h d 5 4 4 ¢ƒW‚¢I‘8•#ISu©QI8EGoQ¢o8QC¢U‚IIGeWAIEIW¢QQVT Q‚WQ‚I˜‡WV#68„GVVE‚Qã h H Y Y H d  — H H d T Y H bB D T bB d H ’ D 2 H 7 H ` 2 T d d T hB d ` H 7B Dr H 4 T  goQ9Q¢I¢G‚¢X‚89XGŸ‡6¢U˜” Û   p‚I8Ev¤ ’ ¥ H d R `B © © ¥   Ó × Ò Ë É Ë × u‡§™IawÓ Ì Ç Ó Ì Í Ì Ñ Ñ Ò ÖÕ Ì Ë Ò Ó Ò Ñ Ö Ó É Ú Ì È Ú Ì Ù Ð ¦Ð ”…ˆfˆ”vԐŸˆmQfÔv™™‰€™˜v†º8§xÎ n-1 n+1 n t j-2 ∆t j-1 j ∆x j+1 j+2 Grid or Node Points x © ¨ .

ç   „pzæ ¦’ ¥ ç  …zd“ø ß îæ ’ 4 €í Ý 6 4 üí Axł5‚˜Ý ß …î   ä í W ß  V Ý  7 d 5q H F D 2B h H D 2 H Y H d  Dy R Y H d i d T 7 7 R Y H F ‚8•IGeXQGo6WQ‚ICEpWQ‚Žf¢d9I‚wpGD Y H H Y 2 H Dq 5 H 2 5 h 2 T t 2 5B D T 7B — 5 d   T H 4 2 H d H gB h h H d H D 2 H 4 D 2B 5 r H H d F D H F D Y T 5 D h H d d Hq H d Y Q6W9QW|¢fp8#p¢o‹8EGV#X˜gWpI¢fQI6WQsEpeQW6GoQ6ˆoX8†3–QWpGCIGˆ¢aG#6W‚Q•6‚ˆXB d H D D Ty H F t ’ H bB D T bB d H h D Y d n T 5 D 2 5B D T 7B — 5 d   T d H h d 5 r h 2 5 4 H Y T YB ç ß QG‚Vƒpu”¢¢EGl¢„W6Iu‚‚Uw‘W#8EGV#X˜gWIp¢CQI‚83xI8Q6Ww•Xvs…î   IÔAxk£5‚xšæ æ rç4 € í Ý 6 4 üí Ý u D T F D Y @ 5 F Y ¦ ¦’ ¥ ’ P  t iy d TyB 7B ” lUGjCcIWu¢EpSëj‘8„W¢XX9X˜”’ y í x‰Ww8XWV#X˜gWIIVˆQI‚83f‚WI™ˆX#‘V6˜935‚xšæ w u t 5 D 2 5 B D T 7 B — 5 d   T d H h d 5 r D Y d n T Y B ß î r ç í Ý 6 4 ü í Ý 2 5B Y Y H d  — H H F D D T F D H H Y H @ 8EWW6WI˜–uIGmVUGŠ6WuvVt ¡ Ep8†jˆ8‚ëv‹xIG‡#ŠpGë¢v6IW¢uIGv68„8jQvcsjfQvcXy ¦’ ¥ ’ P  7 5 d ¤ ’ h 5 F D H 7 H F D q 5 d H h d 5 H F D Y H bB ` d H @ 5  D Y H @ 5 H F D 5 D ` 2B 4 T  Y hB d ` H F D ` 2B 2B T D 2 5 4 7 d H D d 5 d d H H F t ’ d H h d 5 2B T D d H 4 T 5 D ` 2B 4 T  Y hB d pGWIX6¢U‚rX‚8dpGIXIE¢ƒo86¾‚QWw8‚W6ŽIu6QI‚8rE¢ƒfQ6WIX6¢U‚rX‚8` H F D 2 5 Y h 2 H  H h d 5 d d H H F D D T F D d T Hy 4 5 Yy T YB D1 ’ 2 `B Y y T R P H H F D q 5 H hB Y D F `B d H F D 2 pG¬8#pIQ†QI–8‚W6‰pG‘VUW…¢gEQkWE¢–X9£Å‹¢XWC¢Ix6IG‘¢IX‚‘o8X‚IG¬85 Y d T H   T F 4B F @ d 5 d d H q 5 D 2 R 5 7 T 2B T D d H 4 T F DB @ H bB D T bB d H h D Y d n T D 2 H Y H d  H d 2 5B D T R P H H F ‚¢g†I¢dƒXICj8‚W6‹¢ˆoI89¢#X¢G‚6QudWECu8„GVVE‚QI€fWUuoQ‚Q‚IQ‚#8EGVIx6fpGD q 5 H hB Y Dq Hy H F D 2 5 Y 7 d H D H D H d 4 YB h H F D D T F D d T Hy 4 YB DB t d H 2 2 T 7 YB F D 2B h H Y Y H d  — H 2 H F ¢uIEWC|6EapG8C#‚QGW6W6WEIapGCVUWŠ¢gEQCXC„UQQpI¢#‘EIGeEeQW‚Q‚I˜‡6Iª™ ç ¦ ¦’ ¥ s8„pzæ ‡x‡x‡8í W † ß  ¡ V  ¦ V … ß …î Ý  ¦ † ø í W „ ß  “ V  ‘î ß Ý  ¦ „ ä í W ß  ß …î   V ƒ6 Ý  4 €í Ý 6 4 üí ¥x)‚5WxÝ 2 5B D T R P H ` 2B @ 5yy 5 8XWVUSQwIXCcXE8•q H F pGD ‚8•ÆWeEI‚Q‚ApGªI8W¢QWðI¢«9p8†¢‰8GVWQ9ISkIGD X9W6GeIGðI¢U˜v 7 d 5q 5 D Dy R Y H d H F D  R 5 d ` H d h 2 T ê D R 5 ‰ T d 5 D T d H 7 R 2 H F 2B Y 7 d H D H F D h 2 T  — ’‡s…î   pf¥xÅ75‚xšA8EGV#X˜gWIp¢w6IQ‚QsEIwQVUWuIW‘6IX‚I8Qfx‡‘ˆŸXVd‚uXe…î ç ß æ r ç 4 € í Ý 6 4 ü í Ýæ 2 5B D T 7B — 5 d   T H 4 2 H d H gB h H 4 T  Y H F D d H hB Y 2 5 4 D — H ò ’yy T 7 Y YB ß u y w u t d 5 q 2 5 B D T 7 B — 5 d   T H y ‰ T 2 5 Y T H d T Y B ß î r ç í Ý 6 4 ü í Ý æ 2 5 B Y Y H d  — H H F D Y R F Y T ` 2 5y Y VI8Ev¢T í xv18•w¢XGldEpg‚II¢vXIVU8WVgWfvE#…¢sQxaq5Wxš‰¢XW‚Q‚I˜‡ŠIWvISut ç ¡ „pzæ ¦’ ¥ Y Y Y aa`ø í W  ß    V ç ß î ¡s…šæ Ý   ¦ ø í Wpß  ß ‘î V iä í Ý 6 4 üí x)85‚˜Ý Ý  @ 5 cfò 2B T D ‰ 5 5 D ß î i ‰ H hB bB h h 2 T í Ý D 4 T d D ‰ R X¢Gp89G†‘eSpXxEI‰I¢™˜‘‡¢GWpIWY ç h’ ¥ 8dpšæ d T H 2B ¢gpXXy í W í  í W b ï ß  V s…šæ 8ì saagø W ß  V  s‘šæ   ø ß  ¡s…šæ ”xªj5‚xÝ Y Y Y aa`ø ø Y Y Y V ç ß î ø í Ý ä 4 üí ç ß î ç ß î ¦ ¦ Ý  Ý  ï Ý   ï Û 4 ü í Ý d 5q 2 5B Y 2 T  — H Y HB d H Y d 5y i T t H F D d H hB Y 2 5 4 t Hy  7 T — H d 5 ƒ5‚xu8•‰¢XWp¢U˜‡‘QEWQ‚u8XxgmwpGuQIEWI¢QIVXI9¢p‡u8U¤ ’Y Y Y „aaSà   ¥8–f#‘8•fIWxI¢xm¢ŠI8EGVIXI…¢Q4 e à ¦ e ä õ d 5q ý ü í Ý h 2 T í Ý q 5 Y 2 5B D T 2B ‰ 7 5 7 5 dq h H 7 d 5q H ‰ 2 T 4 H bB D T bB d H h y TB D d T  2 H bB ` T 5 D Y 2 5B D T 7B — 5 d   T H 4 2 H d H gB h y T 4 5 8W•uQ9W8•j†aVgˆ8„GVVE‚QI¢¢ƒW‚¢UwQ¢E8C‘G‹I¢XGldEpg‚II¢€QpQW6h„Io¢Q˜›ô í W í  í W b ï ß  V s…pŸæ 7ì caa`ø W ß  V  s…pzæ   ø Xß  ¡s…pzæ v˜ðCIWxÝ ç8dpšmaa`ø c’ ¥ æ Y Y Y ç ß î õ ø Y Y Y ç ß î õ V ç ß î õ ø í Ý ä ý üí Ý  ï ¦ Ý  ¦ Ý   ï Y H bB ` ß D R 5 ‰ T 7 d H D d H D D Ty H F D ` 2B h 2 T  —  ’ ý ü í Ý ä ç ß î Q¢E8up8†¢µW6G•6G‚VƒuIW9IEII¢IpvUsIWxÿˆs…põ ø ß î êæ #…x„mÝ ™s‘˜ð#|mAp¢)…Sê ‹p¥ p‰G…„p˜QSjTXkQ¢E8A8EGoQ¢S¢Qe8EGVG¢I2 ä ç ß î õ ø ßæ Ý h 2 T ß î ä ß t ’ ¥ 5 D ¦’ ¥ ’ Y P  2B 2 H bB ` 2 5B D 2 H b 2 5 4 2 5B D T D 5 H F D ` 2B @ 5yy 5q t 2 H F t ’ h H — n Þ F DB @ ç Þ à ßæ Ý d H hB Y 2 5 4 t Hy  7 T — H d 5 pGkIXCcXE8•u‹Qpuó‹Q˜p‘©GEC‘‚W˜•m#QIEWI86u¢EI#VI–#¢ë¤ ‹8EWI¢Ip–ŽQEW6WY ’ 2 5B Y 2 T  — H Y HB d H d 5y i T t Hy  7B Y T q 5 Y 2 T H 7 i ‰ h H D T Ry T b H d 5 h H D T d H 2 H ` H ‰ 2 T 4 2 5B D T 7B — 5 d   T H 4 2 H d H gB h ¢XxgvjEI#EW‘ë¢vp¢g9wSaQWVUXV¢‡m¢…QWVG6IQ8€†…¢gu8EGV#X˜gWpI¢Š6IQ‚QsEIÆi ² ¶ à ¹ » e#6ja´ R ¹ P à À C C S†QjeIHG ± E ¹ » ´ E ¹ À 9FjwwFˆ± ½ ± Á ´ C ‘wDB @® A­ —   Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u‡§™†fjˆˆ«”w€‘ˆ””Ԑz€mQ‘Ôv«Sψm™‘”Ç .

° ¯Ÿ § ¢ « ¨ © ¢ « § ­ ¢ › ® « › ¥ž § ­ ¥ž ¢ ›     ¬ « ª › © › ¨ § › ¥ž ¢ ›   ŸŸž œ › š ¦d““¦q“£¦“A¤sF'¦d“¥¥d“sa1‘¦£s“q“7¦d¤£a¡ddv‡o™ H ™ ’ F Y H 7 H F D 2B D 2B 5  i d H b H D T $ ¦’ ¥ ’ P  i ‰ 2 H bB ` 2 5B D T 7B — 5 d   T H 4 2 H d H gB h h H d H D 2 H •‹‚Q9apGX‘oX8†fQ¢6‘V–qE˜sëj˜oQ¢E8e¢XGldEpg‚II¢aQI6W6h„IeQ‚QGo6Q4 H F D HY R h 2 T — pGeWI¬IVu•Ý 9˜–m‘s•Ý ‘ç ¡ mjQp8XWEII¢QÅf¢UII¢sa6EIƒX‚XŽ‚8†#EB cfò ä ç æ Ý à ” ä æ Ý t Y 2 5B DB h 2 5 4 i d T h 2 R 5 ‰ D Hy F 4B dB ã H Y 5  7 @ 5 ç ¦ sªø ‘æ r  ä ß î ’ F Y H 7 D 2B 5  d R 5 “I‚’‰…vhWQ9uoX¢suI¢ë¤  Œ — 6 ‘ Y ¥ 6 Y   6 ¦ ¦ 6 ¡ ‹ ä ‘ê ä eß ’ Y HB d T h 2 R 5 ‰ H F 6QEW¢UpI8†wpGD ` 2B h Ry 4 — H Y D 2B 5  € ‡y € }x „y q 5 d H ‰ 7 R 2 H F D 2 T H 7 H @ t Ž F Y H 7 T 2B Y D 2B 5  q 5 d H ‰ 7 R 2 H F IEIIEQ˜‡CGoE8s•lŸ‡‡3U|ŠVf6s…pxwIW‰¢Q#wvëpW6#–ECGoX8†”V‘6swISaIWD  q¢‰¢g†W‘veQICfVUWwQEG¢‘ucXQ†kCcIWuTI¢‰VGoX8††‚¢UpI8†G„CWQ97 5 ÷ T H  Y H @ 2 H F @ D T F D H 4B D 5 ò ’ @ 5y H ‰ 2 @ 5 F Y Œ h 2 T ‹ D T Y D 2B 5  i d T h 2 R 5 ‰ F DB @ F Y H DSE8†‘p8•wpGCQIEWI8•€h8XWVdE˜c‚II¢a9¢GuIWC¢•†8EGVGS6W6WI6WC8GlGQ†8wŠ‡š¢9{ 2B 5  d R 5q H F D d H hB Y 2 5  ’ 2 5B D T 7B — 5 d   T H 7 T Y H F D d 5q 2 5B D T D 2 H Y H d  H d d 5 D T d H  5 ‰y €x  T H bB d H h Y R D Hy @ 5 ò ’ H bB D T bB d H h h 2 5 4 H Y T 5 D 2 5B D T 7B — 5 d   T H 4 2 H d H gB h D 2B 5  T YB F 4B F ¢E‚QIuI•6Eÿc‘j¢¢EGl¢„W6II8Q6W…‘W#8EGV#X˜gWpI¢6IQ‚QsEI•SE8†wuXeƒEIC@ ç $ ¦’ ¥ 8q„pzæ ç4 € í Ý ø í fAxÆv˜Ý   85Wxšæ 6 4 üí Ý  ß…î ¦ ä íç  ( ŠzsÝ ¥ ˆæ 2 5B D Ty H d H F D d H hB Y 2 5 8XWVƒ6WuIGu6IX‚I8• } r e j | r p y w p t i p r p ng l „g r j U¡7‚‡mxvu‚‚{†f…o!¡7| ƒ ‚— –— o˜– ’ 2 5B D Ty R 7 d 5q h H D T 4B D YB F  5 h8XWVƒIwW8•eQWVgEGfXII¢WY H d 5 7 T Y H dB R P H d 2 5B D T 7 d 5q 2B y T 2 5B DB h h T F 4 R ” ’ h H 4 d 5q 2 H H d T Y 2 5B DB h 2 5 4 i d T h 2 R 5 W89#™QWEISQ‚e8XWVd‚8•pE›¢U8EGEpI¢eƒIx™hQQ‚8•p6a‚¢™p8XWEII¢Q‚VUII8†‰ cIF @ 5 d 5 h H 4 2 H d H gB h H d T F Y H 7 H F D 2B Y D 2B 5  d H F D ¢rQ6IQ‚QsEI9WV)W6#9IWAEwWoX8†a6IGV5 cIapjXQWaI¢XW‚Q‚I˜‡d‚QIW•¢wQpGE6I2 @ 5 F Y R yy H D Y 2 5B Y Y H d  — H H Y H F D q 5 d H F DB H t6Q8‡•c‘QGoE8saIXpII8‚WI‚›¢€#‚QG9XŽ‚Q9a#XvoX8†™W6W6WEI™I¢•Vf8„GVV„WQpaT d H b H @ 5 € ’ Y D 2B 5  ` 2B h 2 R 5 d d R Y q 5 Y 7 d H D 2B F Y H 7 T 2B D 2B 5  H D H d 4 YB h H 2 5 D T H bB D T bB d H h 5 D 2 5B D T 7B — 5 d   T 2 T H bB ` i H F D H 4 2B Y ž € ‡ x  € } œ ‡ }  „ } € } y ˆ x „y ‡ œ d 5q YB Y T ‰ H F D H d T H Y H F W–¢XGldEpg‚II¢d¢¢E896pGf6IX‚ˆGlÔ¢‚‡Uƒƒh‡G‡ ¢ˆU|lUj8•ˆXWVU™IWf‚¢W6Iut ç — ¦’ ¥ Sg„pzæ ç ¥ ¦’ ¥ oc„pzæ ç  …zŽ~›f¥˜r75‚xšæ ß îæ ’ ø ç4 € í Ý 6 4 üí Ý í W  ß   ߅î ç  ߅îz排~ø›çf4€¥í˜Ý¬ø”íxÝ   6845ü‚형æ Ý ä V ¦ Ý   í W ß  V Ý  ß…î   ¦ ä Û Y H bB D T bB d H h h 2 5 4 H Y h 2 T D Y d n H F D d 5q Y 2 5B D T 7B — 5 d   T H 4 2 H d H gB hr h H d H D 2 H 4 QQ¢EWVVEW6Ip8Q6WI¢f‚‚UIGŠ¢••I8EGV#X˜gWIp¢…6IQ‚QsEI£˜Q‚QGo6Q4 D 2B 5  r H H d F D H F D H d T Y Ty R 7 d 5q H 4 2 H d H gB hr H DB 2 n q 5 Y Hy  7 T — H 2 5 7 7 5 4 D Y 5 7 H F D Y  T F d H SE8†3‡6WIWaIWw‚¢‘¢ƒp…W8•6IQ‚QsEI£‡G„IU”Vf6XI9¢p‡‰89#86‘‚89apG™IVUW6mÜ } r e j | r p y w p t i p r p ng l j ig e h¡7s{zxvu‚sqo2mkshfd ² À à » ´ À ± C ˆ9”wˆr•” ± Á ¶ ± À ± ¿ CejÅ6¹ ½ ¼ ± » ¹ ¶ ¹ hf”‡egº¸ ™— –— s˜– ­ ® ƒ­ $   Ú Í × Ë É Í Ì Ê fu•™‘ˆˆu× Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Ð ”…ˆ‘€v”ԐŸˆmQfÔm«xÎ xÎ .

H d T F Y H 7 D 2B 5 r d R 5q T 2 5 Y 2 5B DB h 2 5 W¢†W6#uoE8s£fI¢•‘‰8up8XWEII¢Q4 i d T h 2 R 5 ‰ D Hy F 4B dB ã F DB @ H bB D T bB d H h h 2 5 4 H Y h 2 T D Y d n T d 5q Y 2 5B D T 7B — 5 d   T H 4 2 H d H gB ‚¢III8†u6EIƒX‚XGEC‘¢EGl¢„W6I)I86Q‚rI¢w‚‚U‰w8•p8XWVdE˜c‚II¢#QpQW6h„Ih ry T d D 2 H 4 D 2B 5  r H H d F D H F D ` 2B D 2 H Y H d  H d Y d 5 D T d H  5 B d D T 7 H F t ’ H 7 T Y H F D YB Y H 2By H F D q „¢WGoQ6oX8†3–QWpG…IW‰IEGoQ‚Q‚IQ‚W8WVG6s¢— WWVd–Iuw¢9¢W‘IG™EQpXX‘pGˆ¢5 H 2 5 2 t h 5 F D H 7 y T d D 4 H  Y T ` 2B D 2 H Y H d  H d d 5 D T d H  5 H 4 2 H d H gB h —B d D T 7 H F D q 5 H Y T 4 H 7 H d D — I¢IU‹xIG‡#§¢GW66s‚#eIEGoQ‚Q‚IQ‚‘¢GVWQ†8w6IQ‚QsEIŽ„WWV#wIWv¢aW¢Qw9QWWx‡H H F D 2 1 ’ h H nB h 5 7 H ‰ —B d D T 7 H F D q 5  5 D d pG3”‹QIEIx#†#„‚GV#uIG˜¢8WŠ85 ¢GWVsuIWŠ¢gIu¢uVu6IXEwpGŠVIGGlƒ6EIh 7 5 D D 5 ‰ H F D d T H 2 d 5 D T Y H 2By H F D D T F D H D T D 4B i T 7 Y 2 5B DB h 2 5 4 i d T h 2 R 5 ‰ t Hy  7 T — H d 5 ¤ ’ H 7 T Y H F D iyB d T Y Y H 4 H 2 D 5 2 H d T H 2By 5 D H 2B 7 5 d gdjI¢XG„II86‘‚VUII8†o¢EI9¢p‡j8UjV#¢WCIG…„X‚¢G‚QQ6I€¢Iu‚¢CIXEuGŠIEXy ¢W•q h H Y R Y d 5 D T d H  5 D 2B 5  H F D D R ‰ t d 5 D T d H  5 H 4 2 H d H gB h D 2B 5  T 2 5 h H Y T ‰ YB d 5 D T d H  5 H 4 2 H d H Q‚I€W8WVG6s8voX8†™pG€pIxQ8WVG6s8fQI6W6h„I€SE8†ae8dQWVUvEˆ8GVWQ†86IQ‚Q•q rqB h B d D T 7 T q 5 H 2By F 4 T  ’ d 5 D T d H  5 H 4 2 H d H gB h —B d D T 7 T Y H D T d D Y RyyB Hy  7 T — H YB F ƒEI‰— WGldŽv¢aIXEƒ¢ˆ½Q¢GVWQ†8‘QI6W6h„I†„WWV#Ž™6GVWG‚pXXE…Xp#¢p–XIut ¡ ’ ¥ ç ¢  pzæ y AŒ ¼‘D» e² Ý ¥ ( º² t ø ²Ý ä  Y T 9 ¦’ ¥ ’ P  H DB d @ H d 2 T 4 H ¢…q„p†ëj‡WE‚C6W¢Qav@   6 ¦ ¦ ¦ ç h ¦’ ¥ oa„pzæ ¹ · ·· ¶ — !Ý ¹ · ¡ ¡   6 ¦ ¦   6 ¦ ³´´   6 ´µ  ß…î ¦ ä » h 2 I¢T ç c ¦’ ¥ oa„pzæ ·· ” ¸¶ !Ý ³´´ ´µ  ß…î ¦ ä y ¦Œ t º² à · ·· 4 ¸¶ £Ý … Ý ¹ „ ´µ Ý ´  Ý ³´ ä Ž² Ý ` 2B 4 R h 5 d D 2 1 ’ 7 d 5q y TB 4 H  Y i d H b T Y T F B d D T 7 ` 2B Dy R Y H IEQIp˜‚Go3”‰W8•‹¢XQ6s‚‚Q¢…fVUŽ— WWVd–IEG„IW6Wd H F D D T F D t d H F D d Rq h 2 T t 7 d 5q —B d D T 7 r d 5 D 4 H b T 2 pGwVUG§Q6IGfI•Ap¢§†‚8•e„‚GV#3Ô8G‡Q¢‰XB QpGw‚QWpp–r¢QdvuVUG¢QX6wEwEB 7 H F D Y Y H d  — H 2 T 4 H @ D T F D d T Hy 4 YB D — Ý !ø … … ç 9 ¦’ ¥ 8q„pzæ Ý   6 Ý ø ß î r  …¢sç ß î r  …¢sç ß î r  …¢sç ß î r  …¢sç „ Ý „ Ý   6 „ ø Ý  Ý æ æ ä æ ä ä ä  Ý   ±£Ý 6 4  ø Ý 4 £Ý   6 … ä ä ä ä ” Ý !µæ … ç  hÝ ¥ „ ç  hÝ ¥  ç  hÝ ¥ 4 hÝ ¥ ç  (“æ (“æ (“æ ( “æ 7 d 5q H bB D Y H ` ` R Y H d 5 7 H F D 2B Y 2 5B D T R P H H Y H F D ` 2B DB d W8•u8„GfQ88pWW¢#uIWECI8XWVUSQu‚QIW…IXWEW™ ç “ ¦’ ¥ oc„pzæ ç— Ý 3!Æø …   „ Ý  Ý 6 šæ ç  ( sÝ ¥ “æ ç  ( sÝ ¥ “æ „  ç Æø Ý ç …  Ý  „ ç ÝÆø Ý   6 šæ  Ý   8£šæ  6 4 Ý „  ¬‹vÝ   Ý ø 4 6 ” Ý k!šæ  ç  ( sÝ ¥ “æ ß…î ¦ ß…î ¦ ß…î ¦ ß…î ¦ 4 sÝ ¥ “æ ç  ( Û Y 2 5B D T R P H d R 5q H F D D T H bB d d 6I8XWVUSQup8•IGCVu¢EW‚¢T Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u‡§™†fjˆˆ«”w€‘ˆ””Ԑz€mQ‘Ôv«Sψm™‘”Ç “   .

’y T 2 5 ` TB h y T d D 2 H z¢U8¢¢ƒI›VGGo6Q4 H F D 5 D Y d Hq H d ž   è   H 2 5 y T d D 2 H 4 H F D h 2 T d H ‰ 7 R 2 2B h h 5 Y i T @y T H d T —B d D T 7 h H h 2 T ‰ pGdW™W6•6W‘c¢€é¢ep8v¢WGoQ6…IWI¢™Q†…ISXI˜fxc•EV‘W¢„WWV#eQIIVUdT d¢••GoQ9I8‚¢uIuj–Q² º c²Œ ë² ‹ Ž¢6W‚QWpp–seVgCII¢IIGaI8XVfWo¢ƒfI8QŠp8IWEC@ 5q Y D 2 H 7 R ` d T H F t ’ ç à à æ Á Y T h H Y Y H d  — H H ‰ 2 T 4 Y h 2 T ‰ H F D ` 2 5y T Y D 2 T D Y 2 5 4 D R 5 F DB —B d D T 7 y T 2 5 ` TB hB d D i ’ i DByB ‰B Y Y 5  T YB Y h 2 T ‰ q 5 d H ‰ 7 R 2 i 2 T F ` R 5 F Dy T —B d D T 7    „ ‡ … y ˆy € „‚GV#¢VU88¢XIX‚Gś83„XEIX‚W8†f§X§IIVUI¢§Q†…ISuo¢a8I8pGEVu„‚GV#‡¢†lxcš¢š‡šx Hy  7B Y T d 5q 2 H bB ` YB 2 5B D T d D Y RyyB H F D h 2 T y T 2 5B D  5 YB D 2 H 7 R ` d T H F D 2 XI9X‚w•¢•ŽQ8„8•E–8XWVGW‚IEXXfIGŽp¢†¢U8EGp8vX•S6#p8W¢™IG#XB 8¢‚IW6IC@ ‘ q 5 HY R HdH F ’ ç ’ ¥ 8$   pzæ ’’ ‘ ¹ Œ · ·· ¦ º ’’ ’ ‹ ·· Œ ·· ¸¶ º Č º ‹ Œ ´´µ ´´ ´´ ‹ ³´ ä ç º à Œ à ‹ jQÃsÃëÂÛ ‘æ ¤eÁ 2 5B D T D 5 2 H F D H 4 R h 5 d D 2B h 2 T ‰y €x  { ˆ } ˆ „  À T B d D T 7 T F 4 R Y yy T 4 H ™ ’y T 2 5 ` TB 8EGVG¢IuIG6IIxWGoE‰p¢wŠ–š¢99Wx†¢ƒ˜e— WGld‘ƒpW‹E¢gumŸVU88¢XIh yVGGo6QpG«Ip8W¢ª6W6G‚pXQ†W¢ªƒEICdIp¢UAI8XVeW8pG#¢•ŽpQ6˜‡Ž8‚Q6‘E¢‚¢T T d D 2 H 4 H F D h 2 R 5 d T h H d H D Y Ry 4 H d T F 4B F @ Y h 2 T ‰ ` 2 5y T H Y 5 F D d 5q D  H 4 — H Y 5 d H Œ yy T H d D T F D Y D 2 H 7 Hy H F DB @ B d D T 7 H d T R P Y T YB Y d 5 D T d H  5 H 4 2 H d H gB h —B d D T 7 H Y H F D q 5 F 4 T lUGGo6#6XQkWEC— WWVd9WVUSWeuXuW¢GVWQ†8–QI6WQsEpe„‚GV#9WQpG€¢rƒ¢ˆ ç ’ ¥ S—   pzæ ¹ ¦ ¥ ¢r   6 ¦ ¥¢r     6 ¦ ¦ · ·· ¸¶   6 ¦ ³´´   6 ´µ  ß‘î ¦ ä ¥ (  i ‰ 2 H bB o‰Q8„8` YB D F `B d H F D 2 5 2 5B DB h 2 5 4 2 2 T 7 R H ò T d 5q „ ‡yx  {y ‰ ‡ € œ œ  € } ˆ € ‡ ~ ˆ „ ‡  } ž  h 2 T Y D 2B 5 E–o8X‚‰pG¬8¬8XWEII¢QÆI¢#IQ‘A98•elŸš¢9|8FG†o‡‡x‚lW¢†lƒƒckóp¢9GoX8† d 5B d H D 2B D T H 7 H F 4 Y ` 2B 4 2 H d H gB hry T d D 2 H 4 D 2B 5 r H H d F D T d 5q d 5 D T d H  5 B d D T 7 H F D 2 H F 8X‚QGoEdV#6IƒWIX6IQ‚QsEI£„¢WGoQ6doX¢s£‡Q‚IGAd8•–8WVG6s8Ɨ WWVd†IGº6Iut ç ’ ¥ o¥   pzæ — Wß  V Ý  ä ¿¡¾¦ Whß  V Ý  YB 2 5B DB h 2 5 4 i d T h 2 R 5 ‰ H F D H d H F Xe¢XG„II86e‚VUII8†uIGuW6IC@   ’ ¥ ç 8  pzæ ç ß ß „ r6 … •æ  ‘î ¥ 6 — W ß  V …î ¥ Ý Ý ¦   Ý  ¦   ä … ç  ( sÝ ¥ ˆæ D 2B 5  ` 2B @ 5yy 5q H F SE8†ÅIXCcXE8•kIGD ۇç p“ p†8XW6Qxw6˜ŸA¢XG„II86e‚VUII8†‰I¢#IQ‘9C¢•Š8GVWQ†85 ’ ’ ¥ 2 5B D 4 H ” H H ”æ 2 5B DB h 2 5 4 i d T h 2 R 5 ‰ 2 2 T 7 R H ò T d 5q d 5 D T d H  i ‰ “ ¦’ ¥ ’ P  2B H 2By F D d R 5q H F D H 4 Ty  H d t Hy  7 T — H d H F D d Rq T Y o¬cE˜fëj`XeIEXðGfI8•kIGkQ¢XIQ‚™VXI9¢p‡‰QIW‚I•Å†™i ¦ ¦   6 ¦ ¦ ç ’ ¥ s¦   pzæ ¹ ¦   6 · ·· ¶   6 ¦   6 ´µ  ß…î ³´´ ¦ ä ¥ ¥à  ( ¹ ¦ ¡ ¦ ¦w6 ¡ ¦ · ·· ¶ ¦w6 ¡ ´µ ¦w6 ¡ ³´´ ß…î   ¦ ä  ( 9   Ú Í × Ë É Í Ì Ê fu•™‘ˆˆu× Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Ð ”…ˆ‘€v”ԐŸˆmQfÔm«xÎ xÎ .

° ¬ «ž § ­ ª « ÂFÕ¤¥sxdŸ Ÿ ­ ž § ­ ® © ª Ö ž ª › ® © › ¬ « « § ©   ¬ « ® © › ¢ ¢ « ª § ¬ › Ô › Ÿ › ¨ ª ­ › › ª ¬ ž © ¯ Ÿ ¬ « › Ô ž § Ñ « ¬ « ž § ª ¬ Ò Ñ ­ © ÎÐ § ­ ¨ § › § Í ¦d¤¥‘“Xv¸£svˆ1q¦Q¤'q‘F“s¤¤¦†ÉS£É£d£Ï˜FsX£‘dˆqՑ¦1kd“v¦Ó¦d“†‘qŠuk%ž Ï¥‘¤p“¦« sÌ ’ iDByB ‰ T D Y q 5 Y 2 5B D T d H hB Y 2 5 4 2B d H D Ty y Rq H Y R H ‰ yyB @ F 4B F @ Y 2 5B D 4 R d D Y 2 5 ¢3EEXI¢G‚m¢CI8EGVWQIEWI86eXu6GVX§p•6Wps‹EXCƒXICuI¢XG‡IWW‚I¢Q4 y¢G6IQ8wGEpxˆV‘WI…pGCGE9W6s™QX‚¢UIp8swIG9IXIpXQpXEQUwpGE™GoE8sfI8EW¢Vb T d H 2 H ` H DB R P q 5 H Y R H F D Y DB 7 d H  Y HB d T h 2 R 5 ‰ H F D ` 2B h Ry 4 2B hy H n H F D 2B Y D 2B 5  Y R 5B d T H F D D T 2 5B D T 7B — 5 d   T H F D q 5 H d R D T 2 D 4 T — H H F D d 5 D T d H  5 H bB D T bB d H h B d D T 7 H F D 2B iqB 4 H  pGjVw8XWVdE˜c‚II¢CpGU¢u‚IGVIˆ6¢p‡upGv8WVG6s¢Š8„GVVE‚QIa— WGlduIWwXw|E6Q†WY 5 D i DByB ‰ T H F t ’ Y HB d T h 2 R 5 ‰ H F D D T D  H 4 — H hy H n H F D 2B D 2B 5  i d H b H D T h HBy   T 2 H H ‰ Y T W#3„XXp¢wpuv6QX‚¢UIp8swIW‘lp6Q˜‡‰X6UIGXCoX¢s‰‚68‡fl†6XEII¢QQ†‘¢UF d 5 D T d H  5 D 2B 5  H 7 T Y H F D D T F D d T Hy 4 YB DB h 2 T h H nB 4 H  Y iy H R PB 2 R 2 H H ‰ H b T F Y 2 5B DB h 2 5 8GVWQ†8oE8sw9¢G…pG™lUGC¢gEQfXfEIVkQU„QQ†W‰„QISXpIeQ6s…¢gUfp8XWEII¢Q4 id T h 2 R 5 ‰ H F D t ’ ¥ h 2 T ‚¢III8†apGpph   p†I¢‘9   p†QSj˜3€xqE˜sëjDo‰Q8„8u8WVG6s8CoE8suIW‰VUGe8EGD ’ ¥ ’ Y P  2 1 ’ $ ¦’ ¥ ’ P  i ‰ 2 H bB ` d 5 D T d H  5 D 2B 5  H F D 2 T F D 2 5B r T 7 d 5q 2B H d 5 7 i d d T 4 l#W8•˜X‘W89–‚‚¢gwh   ˜ŽI¢w9   pp6SˆËoŽ68E¢v‚8GVWQ†8‘— WWVdfIG€VUW‘6XW¢fò ’ ¥ h 2 T ’ ¥ ’ Y P  i ‰ 2 H bB ` Y d 5 D T d H  5 B d D T 7 H F D D T F D H 4B D 5 Æ Êˆ— W ß  V Ý  È ß É ‘î ¥ à Y Y Y –aaxà ¡ à ¡ s•Ý à ” ß …î   ¦ ä ä ä y ¦Œ t º² ²Œ ²‹ Hd H F W6IC@ Æ ûV¢r   6xà6YaYaxà   xà   xà   Šù ¥ Y 6 6 6 Æ V¢r   ‡aax8Q8¦ ù û ¥ à Y Y Y à ¦ à ç ’ ¥ oh   pzæ y AŒ Çw² Ý sQl²Œ U² ‹ eÁ º² t ø ç ¦ à à æ  ß…î ¦ ä  ² Ý ¥ ( h 2 n H @   ’ IUuvpt 8  p¥ h 2 T IVŽ—   psQSjfE™¢†¢pXWfo8X‚apG†¢‘p8XWEII¢Q‰‚¢III8†IVdI6fwsEWQ‚Q‚Iwv”£1 ’ ¥ ’ Y P  2B Y T t H hB Y D F `B d H F D 2 5 Y 2 5B DB h 2 5 4 i d T h 2 R 5 ‰ 2 2 T 7 R H ò H ‰B d 4 Y H d  H @ q ç ’ ¥ oc   pzæ Æû— Ý à Y Y Y à ” §Ÿ!Gà ¡ 6aaSà ¡ s!úÝù  ß…î ¦ ä y AŒ t º² Û Y H hB Y D F `B d h 2 T Dq H 6QIEWCo8EWI¢‘|‡Xy H F D 2 5 Y 2 5B DB h 2 5 4 i d T h 2 R 5 ‰ D Hy F 4B dB ã H F D ` 2B hy 5 F d 5 D 4 H b H F D d 5q Y h 2 T D pG)8wI8EGEpI8Q‚VUII8†a6XpƒXWEeIW‰IEIX¢I‘8G‡Q¢dIGw8•wIp¢ƒfY y AŒ ~W6IC@ º² t H d H F ç ’ ¥ 89   pzæ y AŒ ¼w² Ý sgà   x8ÔeÁ º² t ø ç ¦ 6 à ¦æ  ß…î ¦ ä  e² Ý ¥ ( i ‰ o² Ý ¢¢EGl¢„W6II¢QQ‚uIGWVdE˜c‚II¢‰¢Qav@ q 5 H bB D T bB d H h h 2 5 4 H Y H F D H D T 7B — 5 d   T 2 T 4 H ç ’ ¥ o“   pzæ ·· ·· Å ¹ •Ý ´µ ’ ´ ’’ ´ · „ Ý ´´  Ý ³´´ 4 ·· ¶ £Ý ä ² Ý Y T ` 2B 2 n H 㠒 Y Hy  7 T — H Y R 5B b H d  d R 5 H ŒBy T d H 2 H @ 5 2 2 T 4 H  V² Ý IXpU6ˆ6QXp#¢p–‘p8Xx6‚ICI86XXVGQpQ8` sp†¢Qur™ Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u‡§™†fjˆˆ«”w€‘ˆ””Ԑz€mQ‘Ôv«Sψm™‘”Ç c   .

Ý ¥Œ ¹ Ý Aº ·· ·· ·· ·· Ý ¸¶ 3‹ † ‹ †Œ †º … ç   ¥ pzæ ’ ¥ … … Œ º ‹ „ ‹ „Œ „º 4 xº ´µ ´´ ´´  ‹ ´´  Þ¡‹ ³´ Œ 4  º SŒ 4 ä ç à à Û “æ Ü jQ² º s²Œ †² ‹ –zokÁ i ‰ F Y H 7 D 2B 5 r “ T d 5q 2 H bB ` YB Y HB d D 2 o†‚Q9uoX8†3s#C8•e68E¢uX‘QX‚GoQH W8•„IIS8IWEC@ 7 d 5qB 2 R 2 5 2 F DB d 5 D T d H  5 B d D T 7 y T 2 5 ` TB hB d D 4B h 5B d H  y T 4B  i D i ’ —B d D T 7 y ˆ ‡y € } œ T Y T 5 D h H d d Hq H d YB F Y H 8GVWQ†8— WWVdV¢U8¢¢ƒIEWWvXIxX‚Q†S¢gEIx3¬˜s„‚GV#hŸ¢QŸ‡‡UŠ§¢fGa6W‚Q•6‚˜XwWQ97 4B h 5B d H  T 2 5 Y 2 5B D T R P H d Ty T 4 Y d 5q Y H 7 H F 4 Y ` 2B 4 2 H d H gB h Y D 2 H Y H d  H d D T F D B d D T 7 H F XIxX‚Q†w#8ˆI8EGVIx6ˆ¢ƒ¢QWj8•€Q9QIƒ‚uIX6IQ‚QsEIŠWoQW6WI6WˆVIG…— WWVd‘Iut ‘ r xs   e‘€‹‚Q9Xp˜EWQ†Š¢QIXE‘8uWoX8†Co8Xj ä ß î ’ F Y H 7 4B h 5B d H  d T H 2By T 2 5 Y D 2B 5  D F `B 6   Y Y aaY ‘    ¦ Y c 6 Y 9 6 Y “ 6 Y $ 6 Y — 6 Y ¥ 6 ¦   6 ¦ ¡ ¡ c 6 9 6 Y Y aaY ä ‘ê ä eß ’‹8EGVQ˜EuIXW‚¢G‚jG„apXI6QQ‚Iv‚I‚ÚoX¢sCIW€lwsIIQH  Eˆ¢wI8E€V•W‚¢G‚uIX‚Q†…IS2 2 5B D T 4 5y ` 2B D d T D Y Y DB ` 2B h H 4 H d  D Y R Û D 2B 5  H F D D T Ž Y h 2 DB Y T ` 2 5y Y T Y D d T D Y ` 2B d H ‰ 7 R H F D H d H F @ d H D D T 7 iyy T H d  2 Y H 5 h DB t Hy 4 dB 4 T q 5 d H D H 7B d H  H F D 2 5 D R 5 hB Ty YB F Y H 7 H F pG‘W6ICf6GWld‰EE¢g‚uD ÚWQxI™E‹VXQ‚X6Ž•V™QW69X‚Q†‘IG8p8E¢ƒ™XeW6#‘pGD 2 H F ™ ’ ’ ¥ H d R `B ¤ 2B Y T F Y H 7 d Ty R 4 dB 4 T 2 5 iy H bB D Y H ` ` R Y H d 5 7 d 5 t Y HB d D 2 H h H D T H  H d F DB Qpªmp¥ pu‚I8Xm#Ej¢#WQ9”¢XIQ‚X6u98…EQ¢EW‚Q¢8IW•W89j8SQ6XWWoQ9QGlg†QW–WEC@ F Y H 7 d T H 2By T 2 5 h H D 2 H Y H d  H ‰ d H F DB H 2 T 4 YB F t ’ @ 5y H ‰ 2 @ 5 F Y F Y H 7 4B h 5B d H  D 2B 5 r D F `B ‚Q9v¢QIXE™–89QGo6WQ‚IŠ†”QpGE6‘¢gvXIumusEQ†–CspW–W6#€XIxX‚Q†voE8s£‚o8EQH H F D d H hB Y 2 5  ’ Y H ` 2 T F 4 d 5 D T d H  5 —B d D T 7 H F D q pGuQIEWI8•eQQ¢I¢Uƒu8WVG6s8„‚GV#–IGv¢5 W¢•9IW›‡Ÿ¢QŸ‡‡U…‚¢wI8XWEIp8Qef¢T 7 d 5q H F D t y ˆ ‡y € } œ H d T Y 2 5B DB h 2 5 4 i d r h 2 R 5 ‰ H F D q1 ’ h H — n H d T Y 2 5B DB h 2 5 4 i d T h 2 R 5 ‰ H F D F 4B F @ 2B Y H Y T 4 h H D T d D Y RyyB H b 5 ‰ T H F xII8†CIGs£j‹Q˜puWVˆI8XWEIp8Q‘f¢UIp8s‘IW9ƒEIC‘X€QWVg‘QWVGW‚IEXEf¢s†¢fIut } p tg r j ‚v†!¡7| ƒ tg Ø eg r p نQX†!sfd –— –— ˜˜– ’ í Ý q 5 D 2 HB 4 ¨ H 5 sxm¢‘oQX696xQ4 H F D YB Œ D T F D d T Hy 4 DB Y H ÷ T 7 F 4B F @ D 2 H 7 Hy H F Dq n H F D 2B 5 d H Œ T q 5 2 5B DB h h T H F D H D 5 pGwE9ÓlUGw¢gEQ…„…QV¢#ƃEICwo6#6XQÅW|IŽIWrXe‚QQ†™Vr8EGEpI¢eIWdGVfò ç ¦ ’ ¥ sp¥ pzæ à 4 º à Œ à 4 ‹ ‹ æ Á ä ²Ý ² Ý ç ¡ ƒa¥c¥‡¡¥à  ˆefŽ5 ( Y XB ¡ ¥ p††j7¢CoQE¢VEISQd„‚GV#uIGCISujQ¢GVWQ†8Ž— WWVduIWeXC|EpWY ’ ¥ ’ P  q 5 D 2 Hy T bB R P H —B d D T 7 H F D Y R F t ’ d 5 D T d H  5 B d D T 7 H F D 2B DqB F y¢U8¢¢ƒI‰WwII8†W6W‚8Qw8GlGQ†8aoX8†dIWAEw|„IW£¢ê e6I8XWVgxXwI8EW¢V9IGwVT T 2 5 ` TB h T 5 D Y h 2 5  Y H d d 5 4 d 5 D T d H  5 D 2B 5  H F D 2B DqB F Yr i ’ Y 2 5B D T 4 5y Y R 5B d T b H F D D 2 5B D 4 2 Rq H F D q 5 Y H Ry T b H F D 5 D Y D F `B H @ H F D ` 2B bB ` Y D 2 HB 4 ¨ H 5 4 H F D Y D 2 H 7 R ` d T Y DB d 5q Y T 8EG‡II•IWm¢u6IXV¢pG‘GuGo8EQvIG…pXxE8CWS6X6#‡˜6IGuWS6#p8W¢ŠGEu¢•Š¢UF d 5 D T d H  5 B d D T 7 ` 2B h 2 5  Y H d d 5 4 H F t ’ H bB D T bB d H h D Y d n T d 5q d 5 D T d H  5 D 2B 5  H F D H ‰ D F `B 8GVWQ†8‘— WWVdwIEII8†W6WW¢Q™puv¢¢EWVVEW6Iv‚WIu€8•”8WVG6s¢voE8s‘IWC†ˆo8X97 ç ¡ ¥ pzæ ’ ¥ 4 üí Ý4 º ø í ݌ ø 4 €í Ý4 ‹ 5Wxfx“jxƒ“h¥˜f¡~ø  AxÝ  ×ušhÝ  “æ €í ‹ ä íç ( Hy  7 T — EI9¢p‡H d8Uj‹8EG6pI•™IW‹¢•6IXV¢uIEW8†S8X6I˜¢€#‚QG#X…vSE8†QI6WQ•‡WfIGvV™¢EWVVEW6Ih 5 ¤ ’ 2 5B D 4 2 Rq H F D q 5 Y H Ry T b ` 2B d 5 ‰ F `B H 2 q 5 Y 7 d H D 2B ê D 2B 5  H 4 2 H d Hq H d H F D D T H bB D T bB d H H 7 5 Y d 5q 2 H D DB d @ iyy T d H 2 H ` YB d 5 D T d H  5 D 2B 5 98‚u8•‰6G‚EWC#EE¢G6IQ8CEu8GVWQ†8‘oX8† v‰QIWQ6•3‡se¢XGlƒQ‚9‰‚XXp¢ƒfQ9GD i ’ 7 H F D 2 H H @D H ‰ 2 5B D Ty H d T F YBy ‰ T DY H 5 D 2 T D d 5  7B YB DB t Y d 5 D T d H  5 D 2B 5  h 2 T —B d D T 7 F D 5 ‰ q 5 H Y R Hy ‰ T d H hB Y 2 5 4 H ÷ T 7 H @ H 4 2B o¢ƒf8s9X§X§„¢Q‚8GlGQ†8§oX8†…IVw„‚GV#wW¢sp¢Š‚IvEI¢WQIEWI86vV¢#•vvQpX˜” h   Ú Í × Ë É Í Ì Ê fu•™‘ˆˆu× Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Ð ”…ˆ‘€v”ԐŸˆmQfÔm«xÎ xÎ .

h 2 T D Y d n T d 5q Y 2 5B D T 7B — 5 d   T ` 2B 4 2 H d H gB h ry T d D 2 H 4 D 2B 5  r H H d F D h d T h 2 T D Y H F D 2 H F IVu‚‚U•8•”p8XWVdE˜c‚II¢aIEQI6WQsEp3„VGGo6Q€oX8†3–QWpG#‚¢UIVƒ‚fIG#QIª™ ’ ç — ’ ¥ S¢¥ pzæ ’’ ‘ ¹ Œ · ·· ¦ º Œ ‹ ‹ º º Œ º Œ ·· ’’ ’ ·· ¸¶ ‹ ´´µ ´´ ´´ ‹ ³´ ä ç º à Œ à ‹ ˆ6¥c¥UÂÛ ‘æ Ü “xkÁ i ‰ 2 H bB o‰Q8„8` YB —B d D T 7 D 2 Ty R 4 dB 4 y T 2 5 ` TB hB d D T q 5 H Y T 4 y TB 4 H  Y H F t ’ DB H b 5 ‰ T H 2 5 H F D q 5 D F `B d H F Ek„‚GV#…o¢XIQ‚X6”¢U8¢¢ƒIEWW†‘¢Ž‚¢g€¢ƒ6Q†WeIu‰g„Ž8c†¢eI¢dIWu¢wo8EWŽpGD 5 D D 2 H 7 Hy H H 2 5 ç ’ ¥ H d R `B ¤ H H Yæ h H DqB F Y YB B d D T 7 D 2 Ty R 4 dB 4 T q WfoQ9QEQwI8uo¥ pwWI8Ev…6W•ÅQW|EI‚fEe— WGld‘o¢ƒpQWEQ9ˆ¢5 cWƒ¢gfVUWw6XGVfò @ 5 d F 4 T H D T F D H 4B D 5 ç ¥ ’ ¥ o8¥ pzæ ä AŒ „ Œ  æSŒ Œ 4 ¹ 4 Œ ä x±¥Œ „ Œ  Œ ´´µ Œ 4 Œ ä „ Œ ³´ · åS±¥Œ ·· ¸¶ „ Œ  åSÄ¥Œ Œ 4 Œ ä YB — d H h d 5 q 5 —B d D T 7 D 2 Ty R 4 dB 4 y T d H 2 H ` D Y 5 7 H F t ’ d H D T E–‘6IW¢v¢d„‚GV#Co¢ƒpQWEQ›¢WQI68C‚¢#uIuj6QGlƒy 7 H F D D R 5 ‰ T i T Y 5 D H d 5 7 F 4 R 7 H b T F yyB @ H ™ ’ YB Y iy T 2 T d R 5 2B Hy 5 d y T DB b T i Ty  Y H 4B d D T QIW€˜8sVdgWwGu‚8#dƒI…™¢gUUXEC‘vQEWxEVU¢vI89XCX8‚U¢ƒ„xw9gƒIˆ6QX‚GV#7 DSVƒI6WX•Žgo¢G‚I86‘W¢aIIVUI8EW¢V‘IGeI¢X¢wGo6#6XQ–IW6ICeQ9W8•‹s„‚GV#7 2 Ty R 4 dB  ’ D 2 T D Y 2 5 4 H d T Y h 2 T ‰ Y R 5B d T b H F D ` 2 5y T Y D 2 H 7 Hy H H F D 2 H F @ h H 7 d 5q t —B d D T DSVƒI6WX6vpG›X™„‚GV#jXIxX‚Q†‘p¢v6WpIWo¢XQ6s‚‘S6Q8‡•c‘'go¢G‚I¢Q§¢IˆWVu„‚GV#7 2 Ty R 4 dB 4 H F D YB —B d D T 7 4B h 5B d H  T q 5 D H Y ‰ R Y y TB 4 H  Y T t d H b H @ 5 € ’ D 2 T D Y 2 5 4 D 5 2 H d T —B d D T 4B h 5B d H  T q 5 Yy T 2 5 ` TB h H F D ` 2 5y T Y D 2 H 7 Hy H H F D t Hy  7 T — H H F D 2B 2 @ 5 F Y Y T ty T d H 2 H ` 2 XIxX‚Q†Žv¢fE¢U8¢¢ƒI‘pG#p8X¢uWoQ9QX6…IWh¢EI#VI–…IWXACcIWV›Ÿ¢WQI68e31 } p tg r j sv£!¡7| ’ F Y H 7 d Ty R 4 dB 4 T 2 5 Y D 2B 5  D F `B ‹W6#•¢ƒI6WEQ‘¢uGoX8†Co8Xj ƒ j i |ã â t rg mɁ†s†áà ߗ –— ˆ˜– Û ’ ¥ H d R `B p¥ pu‚I8Xm¤ Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u‡§™†fjˆˆ«”w€‘ˆ””Ԑz€mQ‘Ôv«Sψm™‘”Ç 7 6 8 1 ¡ ¥ 5 4 2 3 .

’ h H D T Ry T b ‹QWVUE¢V6H H ‰ 5 D YB H bB D T bB d H h H F D F 4B F @ D T D 2B 5  H F D D R 5 ‰ †«WeXA8EWVVE‚QIAIGðƒXIC‰V†oX8†AIGp8†¢T 6WQWoQQÿ8XW6Ip•AIWw¢‰QpX¢Vb h H d H D 2 H 4 2 5B D 4 2 Rq H F D q 5 Y H Ry T H H d F D ` 2B Y R H bB D T bB d H 6WIWIEWI†8EWVVE‚QIh I¢QQ‚ru¢Æ8EGV#X˜gWIp¢T ¢#8•XI¢G#8ExcmrWS6W6WI6Wd h 2 5 4 H Y T q 5 2 5B D T 7B — 5 d   2 T d 5q Hy ‰ T D d 5y i T t T Y D 2 H Y H d  H F 4B F @ ¦’ ¥ Hy ‰ T t d H hB Y 2 5 4 t Hy  7 T — H 2 T Y i ’ c’ ¥ ’ P  i ‰ 2 H bB ` 2 5B Y 2 T  — H H F D q 5 H Y ƒXpC†„pEI¢md6IX‚I8Q•¢Xp#¢p–º¢Žfªpd˜Šëjzoº68„8º8EWI¢Ip–IW¢‚IR H bB Y 2 H D — H Y H ÷ T 7 F 4 B F @ Žt H y ‰ T D d 5 y i T 8„WI6Gx‡€QV¢#ŽƒXICu¸¢Xp¢ƒˆ8Xxgmt  Š6I‚GfI8QwGˆX€p¢vEIG‘fW¢QaW9g•#‚¢Šf8•q T D 4 R d D Y 2 5 4 5 D YB D R 5 YB F D i d d T 4 5 D i T @ h d T @ d 5 D F `B T d D Y h 2 T Hy  7B S¢X¢WG‚ªIVXp#EWY QQpW85 o¢a¢#W8WVG6s¢3‚oE8sQI6WQsEp3‡WEII¬pXG‡IWW‚I¢Q4 i ’ d H hd i 2 T q 5 Y d 5 D T d H  5r D 2B 5  H 4 2 H d H gB hr H DB 2 n ` 2B D 4 R d D Y 2 5 d 5q Y 2 T H 7 T Y H hB b 5 d  D 2B 5  h H — n T D R 5 ‰ T Y 2 5B D 4 2 Rq q 5 2 5B Y 2 T  — H Y HB d H Y d 5y i T t H F ¢•wI¢Q#–6IXxcWI…SE8†Æ6˜p–p8†¢–I¢XG‡II•C¢¬8EWI¢Ip–‘QEWQ‚…8E˜gmeIut } pã ø | ö r eã ÷ | s†ËgU£õvËö À ± · ¼ vÀ ™— ߗ sõˆ– ” ó ¶ G ñ à ² ô¾òÓ𐱠R ± ³ Á •B ï ¶ ¹ Á ¶ ± À ± ¿¹ ½ ï ¶ ¹ » Á î À » ² ¶ ðg‘ejÅ6¾¡ðgjuejme9à í ì® £­ ’ Y H by H Y 7 H F D Y H 4B d D T 7 H F D 5 D d 5B d H D 2B yy T YB 2 5B D T 7 d 5q 2B YB F D H 4 2B Y Y d 5 D 4 H b 2 5B DB h 2 5 QQ¢E6W9QIWŠQQEWWVduIW‘GŠ8EWQWoXhE¢CXd8XWVd‚8•pEuXIWu6IX‚f‚8G‡Q¢Ž8XWEII¢Q4 i d T h 2 R 5 ‰ 5 2 H d T H d H F D D T F D H 4B D 5 ò ’ iD d H  5 d  y TB 4 H  Y YB F D q 5 H ` T D 2 T b h T H ÷ T D H @ 2 ‚¢III8†kIŽWVWQpG…VIGd6XGVf83‚6s¢WIv¢ƒ6Q†W…EIGC¢Ž¢¢ƒo¢Vp¢V¢ƒev)85 d H D T ô ’ Y d 5 D T d H  5 D 2 Ty R 4 dB 4 5 Yy T H d T Y d 5 D T d H  5 4B h 5B d H  q 5 Y H Y T 4 y TB 4 H  Y H Y H F D t iy d T Hy 6GV˜aQ‚8GVWQ†8o¢XIQ‚XQe‚X¢…W¢u‚8GVWQ†8wEIxXW6s€¢u6W¢Qv¢XQQ†W–‚QIWh8„W¢QX•   6 ¦ ¦ ¦ ¦ ¦ ¦   6 ¦ ç $ ’ ¥ 83¥ pzæ  ߅î çsQà   6Sà8¦fæxÜkÁ ¦ ¦ ä ¹ · ·· ¸¶ ¦   6 ³´´   6 ´µ  ß…î ¦ ä Ü ç ¥ “æ  ( h 2 I¢T ¡ ¦ ¦ ç ¦ à ¦ 6æ Ü sQà ¡ 8w£xkÁ ß…î   ¦ ä ¹ · ¦w6 ¡ ¦ ¦ ·· ¦ ¸¶ a6 ¦w6 ¡ ´µ ¦w6 ¡ ³´´ ß…î   ¦ ä Ü ç  “æ ( 7d 5q H F W¢•upGD H ÷ T D i H F D t Y 2 5B DB h 2 5 4 i d T h 2 R 5 ‰ 4B h 5B d H  F DB @ h H Y R H d T t ’ ¥ ’ P  H H Y t H bB D T bB d H h h 2 5 4 H V¢ƒa6IW¢QI¢XG„II86…f¢UII¢svXIxX‚Q†‘GEC‘6WIu‚¢o8¦   poëjXQ‚8¢¢EGl¢„W6I#p8Q6WY ¦ p¥ Í Ì Ö Í ˆfu× ê Ó É Ñ × Ú Ì Ù Ì È Ç ë™¬uv†vf”‹Ú qÔvwˆfˆ””ÔÖ {Ô§‘•Ïm˜™uuóx%xÎ é ÓÒ Ç Ó Ì Í Ì Ñ ÑÒ é Ó Ò Ë Ç è Í Ë Ú Ó × Ç Ð çÐ .

’ YB Y T 6X‚¢U‰ ‚QGeo‰ ‚QW‘†8CQEW6WY 7d H D i 7 d H D T 2 5 Y HB d H d 5y i T t q 5 Y 2 5B D T 2B ‰ 7 5 4 d T H 2By ` 2B 7 d 5q q 5 i T @ D 2 HB 2 H b 2 5 4 T iy  7B Y YB Hy ‰ T D d 5y i T ¢Xxgv€¢fI8XWVUEI…86™¢QIXEdIE#‚8•ˆ¢egŠuoQEIQ¢o8QŽ†Ep#EWE…XIVƒ™8Excmt i í W … ß  ço“8¥ ’p¥zæ«YaYaY`ø ß V Y … …î Ý …  í W  ß  V ‚Y  …î ß Ý   ¦ í W „ ß  —   “ Y … çs¦fæoYɺ“ø ß ç ¦æ Y º V Y „ …î Y „ sfxkø ¦ Ý  ¦ „ í Wpß    ¦ V Y ß ‚˜…î hsÔxk“vx‡º Y ç ¦æ Y º ø í Ý Ý  ¦ Y  cfxk~ø ç ¦æ Y º ä 4 üí Ý 5‚˜–º ۃ5Wx–£V9¢XWp¢U˜‡j6XW6Wj8ExcmuIWGˆII8†W6WW¢QŠEI¢ƒŠpG…X©cWjf¢ƒupG8¢EI9¢p‡H 4 ü í Ý º q 5 2 5B Y 2 T  — H Y HB d H Y d 5y i T t H F D 5 D Y h 2 5  Y H d d 5 4 Hy ‰ T D H F D 2B @ 5 d D Y Ty H F D t Hy  7 T — d8Uauc‚eIXIp8s‚Q‚W86…IGjV|6E…IWŽGD WQWwIW€Ve8X‚I¢U˜‡™QX‚Q‚8Xxgm…IG†XB 5 ¤ ’ @ 5 d ` 2B h 2 5  Y H d d 5 4 H F D q 5 Dq Hy H F D 5 7 d H D H F D q 5 2 5B Y 2 T  — H Y HB d H Y d 5y i T t H F D 2 2 7 Ry 5 4 ` 2B h 2 5  Y H d d 5 4 H F D q 5  5 D H F D D 9IE8QIEII8†W6W‚8QuIW뢑8GŠpG”VT W6GŠIW†VˆS6X6#‡˜6uIGmX‘fGoQ–ƒ¢ˆ'VXIVƒD 7 d H D H F D q 5 D 2 HB 4 ¨ H 5 4 H F D YB id D 2 H F 4 T  ’ Hy ‰ T d 5y i T t H F D  R H ÷ T 7 t Y ` 2B h T H F H F D d H h 2 R t H 2 5 D Y 5 7 Dq Hy H F D q 5 D F `B d H F D 5 D Y 2 7 Ry 5 4 H F ¢XxgvuIW9p‘V¢#oQ8IEI¢QICIGjQpIIS¢I¢j‚89G|‡XCIWs¢”S¢XWCIWG”I9IX¢QCIut Y Y Y à ¦ aaxà   8gà ¡ #õ ä í W ý ß  V ‡Y …î ß ý ý  Ý t YB D T F D t ê D 2B 5  H F D D R 5 ‰ 6XCVUWpU€oX8†uIGup¢s¢T WQW†ƒ¢QvV†8EWIVU˜‡H 7 d H D F 4 T H q 5 2 5B Y 2 T  — H F D 2B Y d R 4 4 5 D T F D d 5 D 4 Tq 2 5 7 7 5 4 H F D Y d T H   T H d H F D 2 7 Ry 5 4 F 4 T H q 5 h T H F H F D D T D T F pGŽXC‚IQ6˜ulUGŠ8W6¢š#8#986IGŠ‚¢g†I¢u‚QIWŽ#IE8Q‰ƒ¢QmV‰VgIIWuVCVIGD H 4B D 5 2 2 H F t ’ ’ Hy ‰ T D H F D 5 D 2B D R  H ‰ 5 D Y 7 d H D H F D Y HqBy  7B Y F 4B F 6XW¢Iº6Iu«–ç VXI¢G‰pGkGoX#pI‰sAG–9WQW‰pG…Q•„XI9X‚«ƒEIC@ W6G«ƒ¢QWoXB 7d H D F 4 T H 5D 2 h HBy B Dy R 7 2 H H ‰ Y T QEXIEGEp…Q6sf¢UF  …h¢WVg‘XIW†XUh8I8IWE¢šŽEI¢GaIW”¢‘|6EwIWfV†9IX86#XB ß î t H Y T 4 YB F D 2B t F ` R 5 F Dy Tæ Hy ‰ T D H F D q 5 Dq Hy H F D D T 2 7 Ry 5 4 T 2 dVg†I¢#8XWVUSQfIG–Xˆ9WQWfIWh¢UE¢vVUW‘6XGVfe!8EGV#X˜gWIp¢€XIWh¢fWI‘IG–o‰ T H   T 2 5B D T R P H H F D 2B Y 7 d H D H F D q 5 yy T D T F D H 4B D 5 ò ù 2 5B D T 7B — 5 d   T YB F D q 5 H Y R H F D i h H Y R T 4 d 5 d d H y T 4 5y H F D YB D T F @ t iy H 7 T 2 ’ d H @ Y 2 T 2 T 4 Hy ‰ T D YB F D q 5 i h R D Y T D T F D Y 2 5B D Y H R Q‚I¢Qj8W‚QpVgxXuIW”XˆVUCV8E6#VUp6Qv‚I¢9¢QŠXp¢ƒ”EIG뢖ppG‚€lUGjI8EGfQISP HpG€¢–I8wWoQW6WI6WEIu‘ëW¢#¢XGfQISeAG„CA8EW‚QWpp–V#6W9•aXI¢G‘IG€¢5 F D q 5 H 2 5 Y D 2 H Y H d  H d YB F t ’ ÷ d T 7 2 5B D Y H R P T F DB @ 2 5B Y Y H d  — H 2 T H H Y H @ Hy ‰ T D H F D q  5 D H F D D i ’ h H nBy  7B Y YB T d ‰ H `y T H F D q 5 H 7 5 Y D T F D 5 Y h H D 4 R d D Y 2 5 4 YB Hy ‰ T D H F 8WeIG–Š)‹6UEEI9XW–EGI68X¢†IWf¢e98‚#lUGkWº6G‡IWW‚I86‘XeEI¢ƒIut ’ H bB D T bB d H ¢8„GVV„WQph h 2 5 4 H Y T 5 D 2 5B D T 7B — 5 d   T 2 TB ` 2 T d ` T ô D 2B 5  r ¥ h H d H D 2 H 4 d 5q Hy ‰ T D d 5y i T t ’ ¦’ ¥ Hy ‰ T I8Q6WeGr8EGV#X˜gWpI¢¬¢X8I¢W8¢˜–oX8†3sºQ‚QWS6Q‘8•ŽEI¢G…8Xxgv¬8„pEI¢mt … … 4 4 Y … sfSɺ ç ¦æ Y 4 4 Y … sW|x‡‹ ç ¦ræ Y Ý Y „ sfokº ç ¦æ Y 4  4  Y  sfxkº ç ¦æ Y 4 4 Y  sƒƒx‡‹ ç ¦ ræ Y ¦ 4 hsfxhº Y ç ¦æ Y º Œ 4 hsG|S{‹ Y ç ¦ ræ Y Ý Y „ sW|x½‹ ç ¦ræ Y ‹ 4 üí Ý Y 5‚xckº í Ý Y xckŒ 4 €í Ý Y ¥˜Q{‹ í y ú ¤¦uu t Y  ‘î w ú ß í y ü ˜Aut wüu Y … …î ß í y û ¤Auvt w û u Y „ …î ß í y ú ¤¦uvt w ú u Y  …î ß í y xuvt w u Y ß …î í xÝ ù ä ç4 üí ݺ ø í ݌ ø 4 €í Ý ‹ ׈f5Wx–“jxƒ“h¥xSˆæ  ß…î ¦ ø í W  ß  V Ý   Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u‡§™†fjˆˆ«”w€‘ˆ””Ԑz€mQ‘Ôv«Sψm™‘”Ç   ¥ .

D Y d n T 5 D 2 5B D T 7B — 5 d   T 2 TB ` 2 T d ` T ô D 2B 5 r ‚‚UÅ©W«8EGV#X˜gWIp¢ª¢ƒ¢I¢G¢¢˜oX8†3l¥ h d T @ ÷ 4 T ‰ d 5q Hy ‰ T D d 5y i T W¢•oƒ¢Ue8•EI¢GŽ8Xxgvt ’¢8„GVV„WQph H bB D T bB d H ’   pkEI¢mt ’ ¥ Hy ‰ T … …4  4 Y … sW|x7¡‹ ç ¦ræ Y 4 Y … ç   |xY  ‹ ræ 4 4Ý Ý Y „ sGƒx'¡‹ ç ¦ ræ Y 4 Y „ ç   ƒxY  ‹ ræ 4 í y û u w û ˜Au t Y „ …î ß 4  Y  cƒ|x8¡‹ ç ¦ ræ Y 4 Y  ç   |xY  ‹ ræ 4 4 UsƒƒS'¡‹ Y ç ¦ ræ Y 4 44 Uç   ƒSY  ‹ Y ræ ¦ Œ 4‹  ‹ í Ý Y xckŒ 4 €í Ý Y 4 ¥xc7¡‹  ¥xcY  ‹ €í Ý í y xuu t f…î w Y ß í y ü u wü ˜Au t Y … …î ß í y ú u w ú ¤¦u t Y  …î ß í y xuu t w Y ß ‘î í xÝ í W ß  ß …î ù ä çí ݌ ø 4 €í Ý4 ‹ fˆ6˜ƒ~hAxf¡“ø  AxÝ  –æ €í ‹ ø V Ý  ¦ ’   ˜uXI¢mX†CspWuX¢EWVVEW6Ih ’ ¥ Hy ‰ T t 2B 2 @ 5 F Y YB H bB D T bB d H D Y d n T ` 2B D 2 H Y H d  H d d 5 D T d H  5 ` 2B 4 2 H d H gB h r h d T @ ÷ 4 T ‰ D 2B 5 r ¥ T d 5q Hy ‰ T D d 5y i T t H F fWUa9IEGoQ‚Q‚IQ‚Š8GlGQ†8‘IEQI6W6h„I3˜‚¢•SƒVUŠoX¢s£l…u8•‘XIVƒ•8ExcmIut í W  ß   …î ß V ço¥ pzæ c ’ ¥ ç  …zd׈f5‚˜¬”xÝ   ‚¥˜•æ ß îæ ’ ä ç4 üí Ý ø í 6 4 €í Ý 6 Ý ! ¦  H bB D T bB d H ¢EWVVEW6Ih h 2 5 4 H Y T d 5q d 5 D T d H  5 D 2B 5  ` 2B 4 2 H d H gB h y T d D 2 H 4 D 2B 5 r ¥ d TByB 7 Tq H F D h H bB d H h D Y R Û H b T I86Q‚wŠ8•ˆ8WVG6s8•oX8†wIEQI6W6h„Ih¢WGoQ6•oE8s£lj¢XXE#¢š‘IGŽ68„WQpu‚I‚8gUF H ™ ’ D 2 H D YB Y 2 5 4 7 d H D d 5 d d H H F D H ÷ T 7 5 D F ` R 5 d F D h H hB bB h 2 H H ‰ Y T ugoQW‚EWI86µW6Gu8W‚QaIG¢Vd#G†¢I8WpGQIE˜„I6Q†™¢IF  ……IWCVUGuGVfò ß î H FD D T F D H D 5 ç 9 ’ ¥ 83¥ pzæ í W … ß    ¦ V ß Ý …   ‘î ä í W … ß  V ß 5… …î Ý …  ¢ £— º   ø —    ¡ ß…î ¦ ‹ ä  xþ ÿ Y h 2 n H 2 5 h 2 T ç i d D H 7 7 i Y i ‰ F YB 2 T b yyB @ Y 2 7 Ry 5 IIIup8I¢up‚G‡#9x‚oWEI¢VsEXCŠI9IX¢Q4 2 H b H yy T Hy  7 T — H YB F D d 5qæ Hy ‰ T D H F D 2B 2 7 Ry 5 4 F Dq n H F D D T Y d R 4 4 Q¢6uXV†EI9¢p‡–XIW9¢••XIVƒIG¬XÆ#pX86«G|p†IW9lŽWpQQx5  xðWVg9XIW©31 ÿ þ H Y T 4 YB F D 2 5D d Hq Hd h 2 T ÿ þ H ‰ 5 W‘6•6W‰I¢†t  x½†dWD ’ d 5 d d H Y HB d H Y d 5y i T t H F D Y T D Q8‚W6‘6X‚QWC8ExcmuIGŠV‘EB 7 R Y ` 2B F YB 2 T b 2 5 2 D Y d n H F D H D T 2 `B Y H h H IW9IXpWXp¢VS¢I™fWUuIWuWVU8EW6Ia™ ’ d 5 d d H H F D H D R DB D Y 2 5 4 5 d H Œ 5 D 7 R Y D 5 2 5 h D T F D Y 2 7 Ry 5 4 H F Q8‚WQuIWGpWEW‚I86#‚Q6#W§IWCVI#pCVUGŠp#IE8QwIut ¹   6 ¡ ³´µ ·¶ ¡ ’û ¦ ŸQQà   S8¦ wˆsƒ¥c¥UýS68E¢uXe8EGpE8WwIut 6 à ù 6 ä û º à Œ à ‹ù i ‰ 2 H b B ` Y B 2 5 B D R y 5 Y H F ¡ ¦ ¹ º ³´µ ¹ ¦ ¡ w6 ³´µ ä Œ ¦ ¦ ·¶ ‹ ·¶ ¦ ¦ ¦ 2 5B D T R P H H F D iq YB D T Y H @ qB 5 d H Œ H d T Y 2 7 Ry 5 8EGVIx6uIGŽ|fXWVGuvm„#‚QQ‚¢CI9IX¢Q4 H H d F D D Y d n H F D q 5 Y 7 R Y H F D D T F D @ 5 F Y iyB Y T H 2 T 4 H 2 “ ’ 5 d H Œ H ‰ 5 D Y 7 R Y H Y H F D t 4 h 2 6WIWwfWU‘IW•¢u9IW–IWVUWÏcIWEX‚¢gA¢Q#I#I‚QQ–†Gu#I‚#‚QIW›¢AI¢T t‹ëpj¢wQE8Iƒ‘Q†8WpaIWSëVQW¢•†I¢fo8EW9GC|6Eµ8W•‰QQ6˜‚IwvU‹xIG‡#wIGm¢5 ‰ t T q 5 H 4B 5 F 4 d H  5 d  H F D i ‰ t H 4 d 5q h 2 T D F `B d 5 D Dq Hy 7 5 dq h H H 4 5 d  H @ t h 5 F D H 7 H F D q i 4 T d R 4 4 T q 5 d H h d 5 H F D H ŒB 7B — T 7 5 t ’ Y 2 7 Ry 5 4 H Y H F D q 5 F 4 T H q 6¢WIQ6¢u¢6IW¢#IWa6X9X˜l#†§‘QI9IE8Q9W6IGv¢ƒVgŠV5 IW–IWw6IX‚I8• 7 R Y H F D d H hB Y 2 5 ¥ 8¥ Í Ì Ö Í ˆfu× ê Ó É Ñ × Ú Ì Ù Ì È Ç ë™¬uv†vf”‹Ú qÔvwˆfˆ””ÔÖ {Ô§‘•Ïm˜™uuóx%xÎ é ÓÒ Ç Ó Ì Í Ì Ñ ÑÒ é Ó Ò Ë Ç è Í Ë Ú Ó × Ç Ð çÐ .

ç   — pzæ ’ ¥ ý Ýç ßæ ý †Ôs•8S‹ ä ¾ý ( ä ç ßæ js•mÝ i ‰ 2 H bB ` YB y TB 7 5 2 iy 5  2 5B D Ty 5  d H D 2B 2 TB ` 2 T d ` T ô H F o†68E¢uX†¢ƒ98IxE¢sŽ¢XGlƒ8†W6GoXŽVƒ8IVG8¢˜aIut }ã |g  e i ÷ã e d i eg j |ã e y r p j i ' p j £†#¨uõ†hfæh£7£h{!s¡É¡og  r p % Ø i | p " i | r " | £sÙ&u¼U$h#Ó! –— ߗ ˜õˆ– ’ YB Y iy T 2 QEWxEVU¢T y T 4B d H 7 R 2 2 5 Y D — H D i 2 T 7 2B h H Y Y R 4 YB h H d T Y Ty R 7 d 5q H Y H F t ’ Y Ty R 7 d 5q 2 5B D Ty 5  d H D 2 ¢gEWQ9ISº8dGx‡WÆo¢#«EªQW‚IQ‚XI†W¢9¢XI…‚8•W6Iuÿ6¢ƒp…W8•Å8EGVX8†WQWoXB q 5 @ HB b q 5 D 2B 5  H F D 7 5 dq YB D T F D t i T @ D 2 H d H gB h iy D F `By Y T 2B D 4 H Û ‰ R Y H F D F 4 T 5 d   ¢§6Xx‘¢–oX8†IG¾8W•‘X–lUG•¢gŠ–oQ‚QsEIEWS¢XX‚AÆEd‡Q‚¡I‚†pG¬ƒ¢oWpI¢T Y R D Hy t YB F D 5 h 5 D d H h d 5 2 1 ’ h H ŒBy T d H 2 H ` d H F D d Rq H ‰ 2 T 4 DB D T F D D 4 Tq H F D YB t d H b H @ 5 I…6Xj6XIWeIeG…6IW8¬3A‹Q6XE¢G6IQ8‘6IGfI•Ž†ÆVg…„–VUW…6¢šŽpG…XvQ686vcIF tgo¢G‚8†9X•‚8† QI¢XWpQ9XI˜Qp8XIfG‘QpIQW˜–wIVw6GV#8G˜¢v6Gpp#86•†wEI8Qm£1 D 2 T D d 5  7B H d 5 ’ Y 2 5B Y 2 H 7B h d H F `B F 5 D h H h 2 H D — H h 2 T h H D T 7 5 D R T d H D R  7 5 4 H ‰ hy R 5 4 D ’¢8„GVV„WQp‘o¢ˆ8•jQI‚8‘o¢h¢ˆW8WVG6s¢joX8†ë¢GWS6Qj8p‹Qv6VWx‹W¢•oƒ¢US‹‚¢Šf8•q H bB D T bB d H h i 2 T d 5q d H h d 5 i 2 T q 5 Y d 5 D T d H  5 D 2B 5  y T d D 2 H 4 d 5 t h H @ H ÷ Y t h d T @ ÷ 4 T ‰ t h d T @ d 5 h 2 n 5 D h H Y R H ‰ 2 T 4 Y Y H 4 5 d  YB F D iy d T Hy  ’ d H h d 5 2 H bB ` T q 5 Y 2 5B D T 7B — 5 d   T H 4 R h 5 d IU‰GÅQ‚Id†Å¢Q…W6QxWIaXIWA„W¢QX•QQpW8Å68E¢‘¢uI8EGV#X˜gWIp¢eQpI˜‚I 5 D Y Hy ‰ T t d 5y i T t q 5 H Y R H F D F ` R 5 d F D h H 2B 7 d H D H h H ‰ 5 D Y D 2 HB 4 ¨ H 5 4 H d WaQEI¢mw8Excmu¢dWIŽpGr8p8WIWrQpX9WQW6I#sGwWS6X6#‡˜6d‚¢T  ‹ IWdW6IC@ H FD Hd H F ç ¦ ’ ¥ sI— pzæ ÿ þ  xfä  üí ‚xÝ  ‹ Ü ‘€ ¾   ø í W ß    V Ý Y T Y D 2B 5  ` 2B d 5 ‰ F `B H 2 ¦ ø  ø  q 5 Y 7 d H D 2B D Y T ¢CGoX¢s‘IEW8†S8X6Iuð…C¢u9W6GeX‘‚Vg4 H ‰ 2 T 4 ê D 2B 5  hB d ` D T H bB D T bB d H h F D þ H F D 5 D 2 5B D T 7B — 5 d   T H 4 2 H d H gB h T ty T d H 2 H ` 2 †A¢guSE8†AEW8ul–8EWVVE‚QIAGŽ–IWW8EGV#X˜gWIp¢d6IQ‚QsEIe˜z¢G6IQ8A31    } |ã â  r e © p t i p r p ng l § e i eg j | ¥gã | r p i p †Ç£Uuvu‚‚{†fD¨æk†7v¦£su8פ ç ¡ — pzæ ’ ¥ ç  …šŽfj‡x¢©h¥x8Å6 ß îæ ’ ä ç í Ý ¥ ø 4 € í Ý — ß  Axšæ …î   6 í W ß  V €í Ý ¦ Ý  ‚— ߗ oõˆ– Û H bB D T bB d H ¢¢EWVVEW6Ih D Y d n T q 5 2 5B D T 7B — 5 d   T H 4 2 H d H gB h h d T @ ÷ 4 T ‰ d H h d 5r h 2 5 4 H Y T h H bB d H h H b T F H @ Y R F fWU™¢)8EGV#X˜gWIp¢‰QI6WQsEp«WVŠoƒ¢U–6IW8£˜I86Q‚A©Q¢EW6I‰8gU•–ISut ç h ’ ¥ o¥ pzæ í W „ ß  ¥ V Ý   …î ß „ ä í W „ ß  ¢ “ ß V „ ‘î Ý  4 ¡‹ „ 6  6 ‹ c 4 “   ß…î ¦ ä  xþ ÿ Û ‰7 d H D d 5 d d H 2 5 B D T 4 2 R d D ` 2 B h T H W6GŠ8‚W6‰¢XG lgII‚G…IEI¢QXy H F D Y H hB b 5 d  2 7 Ry 5 4 F D d R 5q H F D H Y T 4 YB F D 21 ’û ¥ à — 6 à pG#6IXxcWp©9IX¢QÆG‚p8•†IW‰‚¢g9XpG©3«ŸVGIfS8¦ ù 6 uƒÃ‡Ãà ¸Q8„8«ƒEIC@ ä û Œ à 4 ‹ ‹ù Y H bB ` F 4B F ¹ ·¶ ¡ ¡ ´ ¦ ³µ ¦ ¦ ¦ ¦ ¡ ¹ Œ ¹ ¡ ä ¡‹ ³µ 4 ´  ·¶ ‹ ·¶ ¦ qB 5 d H Œ 5 E9WQ6–GD —   ³´µ 7 R Y Y 2 7 Ry 5 4 H H d F D D Y d n H F D H 7B D YB F I‚uI#pX86w6WIWC‚WIupG™9XWCEIut — ¢¥ Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u‡§™†fjˆˆ«”w€‘ˆ””Ԑz€mQ‘Ôv«Sψm™‘”Ç .

DSE8†up8†¢u6XW6˜C8Excm‘IEWIQpI¢U˜‡†ufI……‚X¢‰ƒXpC˜8«C–IVuÆu”GoX8† 2B 5  D R 5 ‰ T Y HB d H ” d 5y i T t ` 2B Y R h H h 2 T  — H H ‰ D Y R 7 5 Yy T F 4B F @ t ¦ ø ê h 2 T ¦ 6 ê Y D 2B 5 DlCQ8„GVV„WQpw‚¢†QIIEQIE‘WE¢fpIhhQGo6WQ‚IQ‚#€oX¢s‘Vu¢EGl¢„W6IuIGŠEŽEp8‘¢I2 T Y H bB D T bB d H h H d T h H h Ry 4 2B 5 Yy T D R ‰ t h H D 2 H Y H d  H d ê D 2B 5  D T H bB D T bB d H h H F D YB iy 2 5 D 5 H d H € ’ ’ ¥ Hy ‰ T t q 5  5 D H F D D T h H D T d D Y RyyB YB Ty R 7 d 5q Hy  7 T — H 2 i ’ — ’ ¥ ’ P  5 D Y R 5 ` 5y T 2 ‚Qf7p¥ puXp¢vëV‘8GŠIW”VwQGlGGfIXEXmX™XI…W¢•vXp#¢p–9fmp8— ˜oëˆQWmI¢88XVU¢T ç $ ’ ¥ 8— pzæ ÿ þ  x`ä  üí WxÝ  ‹ Ü ‘€ ¾   ø  ü ‚í W ß  8 9€ ¾    V Œ  Ý    7 t’ ’B t Y D 2B 5  ` 2B d 5 ‰ F `B H 2 D T Y H bB D T bB d H h H h Ry 4 2 „¢H ŸpQGoE8s‘pXW¢sS8EQICV‘68EWVVE‚QIIpXQpXB 2 T 4 ¦ ’ ¥ ’ P  q 5 2 5B D T ŒBy T d H 2 H ` T t — ’ ¥ ’ P ¢Q–I— ˜mëjk¢8EGVQXXVGQpQ8e˜I8— p§ëj ¢6W¢ŠqpƒIwW8•‰8XWVƒ8†W6GoX‘WE9W6f€ 2 5 h H Y T & ’ Ty R 7 d 5q 2 5B D Ty 5  d H D 2B H DB 7 d H T q 5 H Y R H F D 7 5 dq ` 2B YB d T H bB D T bB d H h D Y d n T d 5q H 7 H F 4 Y ` 2B 4 2 H d H gB h H 4 T  Y xy  y  œ {y 2 •¢9WI–IWp8W•ŽpXWEW¢–¢EWVVEW6Iw‚‚Uu¢•99QIƒ‚IX6IQ‚QsEI9Q¢IW”|‡Ÿé†™|Æ¢T d 5q Hy ‰ T D d 5y i T t H F D D — H 2 d H hB Y 2 5  ’ Y Ty R 7 d 5q 2 5B D Ty 5  d H D 2B 2 TB ` 2 T d ` T ô 7 5 dq Y d 5 D T d ¢•wXIVƒ8E˜gm–IGux‡Iu6IX‚I8•–6¢ƒIwW8•¢XGlƒ8†W6GoXVƒ8IVG8¢˜p8W•W¢GVWQH r  5 H 4 2 H d H gB h D 2B 5  xy  y  œ ‰ } h H D 4 R d D Y 2 5 4 Hy ‰ T D d 5y i T t T q 5 Y Hy  7 T — H Y R 5B b H d  H F ˜8w6IQ‚QsEI‘oX¢s˜|‡Ÿé†6s©6G6pWGfI86aEI¢ƒu8Excm#”¢‘QXp#¢p–™I8Ex6WpwIut ’ D  H 4 2 5 4 H F D iy 2 5 h H H 2 H @ H d H F D R ‰ t Y h 5 F D H gp6QI86uIG#EI8†Q6Iu•uW6ICpIU6I˜pG697 y¢gEWQ9IS8u6QI6WQ•‡W†o¢dXeII¢•wseVgCX¢X#¢IxE8†‘W6IGj¢e¢XW‚IQ‚XIwG6EI T 4B d H 7 R 2 2 5 Y H 4 2 H d Hq H d i 2 T 7 2B h 2 R 5q H ‰ 2 T 4 Yy TB 7 5 2 iy 5  H Y H F D q 5 2 5B Y Y R 4 YB h H D Hy r 7 5 i ’ H D T D 4B h Y 7 Hy ‰ 5 d  H F D Y T h H h Ry 4 2B H ‰ hy R 5 4 Y H bB D T bB d H h d H h d 5r d H F `B F iy Y R 5B b ‰ –864 vVGVG6Xp”#6XI¢WIuIW”¢–QpIX6IXu†–XI86”Q¢EWVVEW6I”6IW8£f6I8Xp…E‚I8E˜o“ ç — ’ ¥ S8— pzæ ý W ß  ç ßæ ý V s•oƒŒ Ý   ø ý Ýç ßæ ý €ëÔs•oo‹  ä ç ßæ js|vÝ 2 5B Y Y H d  — H H F 8EWW6WI˜–wpGD ` 2B 4 R h 5 d  t h H Y R H d T H bB D T bB d H h D Y d n Y DB h 2 T 2 5B D 4 2 Rq H F D q 5 Y H Ry T b H D H d 4 YB h F 4B F @ 2 pXQpI˜‚I”‹6WI9WVd¢EGl¢„W6IwfWUuG„rIV)8EG6pI•9pGŠ¢wQpX¢V–G6‚Q‚XI)ƒEICkXB H Y T 4 H F D d 5q YB 2 5B D T d D Y RyyB d R “ ’ H 4 T  Y 2B Y D 2B 5  2 H bB ` D T 5ž3 } 1yx  1y € } ˆ žxy ˆ „  2 5B D 4 2 R ‚¢gfIGˆ8•ˆX‘8EGVWG‚pXXEvpvVQ¢U‚#E•WSE8†dQ¢E8vVu6z4¢2z¢l2–‡xw||…†¢e8EG‡II•q H F D q 5 Y H Ry T b H Y R Y Ty R 7 d 5q H DB 7 d H € t ç ßæ Ý d 5q y TB 7 5 2 iy 5  T D 4 R d D Y 2 5 4 5 t ’ 2 5B D Ty 5  d H D 2 pGhV€6IXV¢CWpj¢ƒIwW8•ŠWE9WQ‘¡‡s•mˆ8•S¢ƒ98Ix„8su•6I‚G‚p8Qam˜‹8EGVX8†WQWoXB 2 TB DB 7 d H € ` 2B Y R i ‰ D R 5 ‰ T D F ` R 5 d ‰ YB F 4 T 5 d   T 2 TB ` 2 T d ` T ô H F D q 5 2 5B D T ŒBy T d H 2 H ` ¢ƒWE9WQ‘{pXWp–Sˆ˜8sV•o8p8WIˆE#ƒ¢8WIIV#Vƒ8IVG8¢˜‘IWh¢‘8EGVQXE¢G6IQ8ªi ’ Y 2 5B D T 7B — 5 d   QI8EGV#X˜gWpI¢T 2 TB ` 2 T d ` T ô Y T 5 D h H dd Hq H d H d ¢ƒ8p¢G8V§a¢‰GQW‚Q•‡W–W¢T   — p§ëjå8W•AQ¢EW6I9sA¢gwVIGa6#6IƒW9QpQW6h„Ih ’ ¥ ’ P  7 5 dq h H bB d H h H ‰ 2 T 4 D T F D Y H 7 H F 4 Y H 4 2 H d H gB H DB 2B ¤ ’ h H D 2 H Y H d  D Y R Û Y 2 5B D T 7B — 5 d   T H 4 2 H d H gB h H DB 2 n H F D H bB d H h H d H @ t D 2B 5  H D H d G„IXv)‹QWS6W6WI‘fI‚ÉI8XWV#X˜gWIIVQI6WQsEpŽG„IUdIWŽ8„WQpQWvvgoE8seW6‚Q4 r YB h H D TB d  5 d   T H F D D T Y H bB D T bB d H h H Y H F D H D T Ry T b H h 2 T t F Y H 7 h H 4 T  YB R P H 2 T H Y 5  7 fXpdGlƒWp8WIp¢IGwV‘Q¢EWVVEW6IeW6IG#GVUE¢V6ÅIVvhWQ9¬Q6¢UWEISQÅ¢dW8†9XB t‡s•mˆ¢a8„GVVE‚QIep8Q6W8™‚‚UwIWw¢¢G…•j£Q6‚…IG‰XaIX¢VwW6‚QWEI6IG¢†o¢T ç ßæ Ý q 5 H bB D T bB d H h h 2 5 4 H Y d 5 D Y d n H F D H ÷ T D H @ q1 ’ D H Y H F D 2B H Ry T b H D H d 4 YB h d H F D 5 i 2 Y H ÷ T D ß 2 H F @ 5d H Œ h 2 T t ý 6¢¢G©¬QICe‚QQÅp¢€¢ëß ä Æ6ICŽI8…Edëƃ¢g‘¢–oQEQ96xQepG…VIGd6XGVfò ß 2 H F @ H 2 5 YB ý Ý F 4 T H q 5 D 2 HB 4 ¨ H 5 4 H F D D T F D H 4B D 5 ç ¥ ’ ¥ oo— pzæ ç )'£|Qç )U!•æ  ß 6 4 ßæ ß 6 ä ß Ý ø ç ß 6 4 ßæ ç ß 6 ä ß s)'£|Qs)U!•æ  ßæ ç4 ß 6 ä ß  ßæ ç ä ß 6 4 ß çW£)6 |Q‚£)U!•æ 4 ß ç ä ß ‡!r6 |Q‡!)7£•æ 4 Ý ä Ý ä ç ßæ  ß æ ç ß 6 ä ß £¬ø sr6  |Qsr7£•æ !ª€s|mÝ ç ß s)6 |Qs)U!•æ ç ß ßæ ç ß 6 4 ß   Y D 2B 5  h H 4 T  YB R P Hr 2 5 2 F DB @ ç 2 5B GoX¢s‰Q6¢UWEISQ£˜8I‰G„C€ë8EGD rlƒ8†¢WGx‡u8•A8XWVƒ¢s‚QGoEwEGVWI¢USC8•wXI…W¢•Vƒ8IVG8¢˜uEI#EWwIGD 8W•‰Q¢VƒD Ty 5  T d D — H d 5æ 2 5B D Ty 5  d H D 2B 4B D T d h T R P d 5q Ty R 7 d 5q 2 TB ` 2 T d ` T ô Hy  7B Y H F 7 5dq 2 H ÷ T H ‰ 2 T 4 ç ßæ ý ‹ H F D q 5 2 5B D 4 R d D Y 2 5 4 H F t ’ ) H H d ` H h q 5 ß 2B Yy TB 7 5 2 iy 5  H d T ç ßæ ý ‹ H d H F †‰¢gŠs•8o‡pGvV†8EG‡IWW‚I86uIuv)06W86IvVeeX€X¢X#8p˜„8†w‚¢us•8S‡W6IC@ $ 3¥ Í Ì Ö Í ˆfu× ê Ó É Ñ × Ú Ì Ù Ì È Ç ë™¬uv†vf”‹Ú qÔvwˆfˆ””ÔÖ {Ô§‘•Ïm˜™uuóx%xÎ é ÓÒ Ç Ó Ì Í Ì Ñ ÑÒ é Ó Ò Ë Ç è Í Ë Ú Ó × Ç Ð çÐ .

ç 9 ’ ¥ 8— pzæ í W † ß  ¡ V  ¦ V Ý  … …î ß † ä  xþ ÿ F D B Y H F D Y 2 5B DB h 2 5 4 H Y H F D d H h 2 H ’û ¦ à ¦ à ¥ 6 à W˜— W#IGuI8EGEpI8Q#W6IGw6II‘IŸ6g8gpfxà ¡ p¥ ù 4 … ¦ ¦ ¦ ¦ ¦ 5 D Y 7 R Y 2 7 Ry 5 Gu9IWe9IX¢Q4 " ä û þ à º à Œ à ‹ù 2 5B D Ry 5 Y H F D ` 2 B b T •Q¥à s¥c¥U¸«8EGpE8W#IG‰IXxgUF ¡ ¡ ¡ ¦ ¡ · ¹ ¡ ´µ ´ ¦ ·· w6 ¡ ´´ ¡ ·· ¶ ¡ ³´ ·   ·· ¦ ¹ ¥ — ´µ ´´ ³´´ þ ¹ " ´µ ´´ ä · º ´ ³´ ·· Œ ·· ¶ ‹ ·· ¶ ¡ — ‘6 ¥   6 ¦ ¡ ¦ ¦w6 ¦ ¦w6 ¦ 2 5B D Ty H d H F D iq YB D T Y DY R 7 H @ t i 4 T d R 4 4 T q 5 d H h d 5 H F D H ŒB 7B — T 7 5 ¢XGlƒQ‚uIG#|‚XWVG‘fI…w•I¢6¢WIQ6¢”V‘6IW¢wpGf6X9X˜l##§t ’ H bB D T bB d H h D Y d n T 5 D 2 5B D T 7B — 5 d   T 2 TB DB 7 d H € D 2B 5  r ¥ y T d D 2 H 4 d 5q Hy ‰ T D d 5y i T t ’ ’ ¥ Hy ‰ T ¢¢EGl¢„W6I‘‚WI‘9WŽ8XWVdE˜c‚II¢Ž¢ƒWE9WQ‘poX8†3s˜¢GWS6Qu8•EI¢GŠ8Xxgvˆp¥ puEI¢mt … … 4 4 4 Y … sfSɺ ç ¦æ Y Ý ç ¦æ Y 4 Y „ sÔxkº Ý Y sW|x½‹ ç ¦ræ Y 4 „ Y  sÔxXþ ç ¦æ Y 4 4  4  Y  sfxkº ç ¦æ Y Y  sƒƒx‡‹ ç ¦ ræ Y 4 hcfxpþ Y ç ¦æ Y 4 4 4 hsfxhº 4 Y ç ¦æ Y º Œ 4 hsG|S{‹ Y ç ¦ ræ Y þ ¦ ä  4 Y † sfxhº ç ¦æ Y 4 ä  4 Y sG|S{‹ ç ¦ ræ Y 4…  † Y … sfSXþ ç ¦æ Y 4 Y … sW|x‡‹ ç ¦ræ Y Ý Y „ sfxpþ ç ¦æ Y ‹ " í xuu t y w Y ß …î ø í xÝ … " …ç ¦ ræ 4 Y cGƒxY í y F u w F ¤Gu t Y † …î ß Ý Y cƒ|xY " ç ¦ ræ 4 „ í y ü u wü ˜Au t Y … …î ß  " ç ¦ræ 4 Y  sW|xY í y û u w û ¤Au t Y „ …î ß ß…î ¦ ø 4 ç ¦ræ " 4 YhsW|xY í y ú u w ú ¤¦u t Y  …î ß 4 üí Ý Y 5Wxckº í Ý Y xchŒ 4 €í Ý Y Axc‡‹ 4 ü y w u t þ Y ß 5‚í xuv!q˜…î í !xut ˜‘î y w u Y ß 4 € w Aí y xuu t " ˜‘î Y ß ù ä ç4 üí ݺ ø í ݌ ø 4 €í Ý ‹ 2jf5‚˜–zv˜ƒ~›¥xSˆæ 45üWí Wß  í ß  W þ V “ø V Ý  Ý  4€Aí Wß  " V Ý  ’ @ 5y H ‰ h H 2By D R 5 Y T F 4 T 5 d   usEQ†eQIEXWp8C¢†ƒ¢oWpI¢T Hy ‰ T t d 5y i T t H F D ` 2B Y R h H 2B 7 d H D H h H ‰ D Y R 7 F 4B F @ ç T d ‰ H `y T H F D iqBy  7B Y 5 D H 2 5 Y EI¢m–8XxgvIWApXWp©6IX9W6G6p‰†#‚I…ƃXpCaoGI68XVIWk|EXp#EWAGI8–¢T 2 H ÷ T D YB D 2B 5  ê H F D 2 5 D 2 HB 4 ¨ H 5 4 H F Dæ Y D 2 HB 4 ¨ H 5 4 H b T F @ 5 2 Y 7 d H D H bB D T bB d H h H F QV¢ƒ…E–oX8†9IWÆ89oQEQ96xQIW•GoQEQ96xQ‰8gU§cI‘#‚QWŽ8EWVVE‚QI‰Iut ç “ ’ ¥ oo— pzæ í Ê DEW b ä ¾B ï   ß  V ï ß  s‘˜Ÿæ 7ì CÈ ç ß î õ  ï ¦ ï  Ý    @ Aä ý ü I‚í W ß    V Ý    Û c’ ¥ ’ P ˜dpëëj 2B 2 H bB ` 2 5B Y 2 T  — H Y HB d H Y d 5y i T t H F D q 5 2 5B D T ŒBy T d H 2 H ` ` 2B @ 5yy 5q H F D Y H dB R P H d YB F t ’ EkQ8„8k8X‚I¢U˜–QX‚Q‚8Xxgm‘IGˆ¢¢XGlgXE¢G6IQ¢dIECsEX8•…pGQ‚XIS6WXpuaUê Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u‡§™†fjˆˆ«”w€‘ˆ””Ԑz€mQ‘Ôv«Sψm™‘”Ç “ 8¥ .

° © ›ž ¢ ­   ¬ Ò « e ¨ § « e § q i ssd“¦q‘qF$“¦sA­ rh pÐ   ¬ ‘¦­ Ð ¤Avg¦sqÕ¦¥Ua‘d†Õu%q¦Úv‘FÕ¤¸q‘¦£1¤Aqq¦vfq¤ÉddqÕ¦¤S¦£‚¦¤o¦q¤“¦£8d‘“¬ ¦a ¨ § « e Ñ « © › ҟ ­ ¥ ›   ҟ ª ¬ž  Ÿ Ò « œ © ¬ «ž §ž   ¬ « ª ¯ ¢ ­   ¬ Ò « e › ¨ § c ¬ž ¬ž ­ § ¬ « ª ¢ « § ª › ¥ › ¨ § › © ­ ª ©ž ¨ § b h ’ Y d H by 5 Y h H h 2 T ‰ D 2 HB 4 ¨ H q 5 H Y R H F D i ‰ h H D 2 H 7 Hy  7B H ‰ 2 T 4 i H F D 2 H F @ y T 4B D 4 T d  H Q‚Q8„8W†QIp¢U‘oQX696m¢a‚IuIGdSe6GoQ9QEI9Xus‰¢Qe6IW†6ICh¢QXG‡¢Gpw†‰ 2 T 4 Y 2 5 B D T 7 B — 5 d   T `H h T Ü d 5 2 T B D B 7 d H € t y T d H 2 H ` 2 1 ’ h H Y R i y H h B @ h 2 T t 2 R d 5 D D 2 H B 4 ¨ ¢QuI8EGV#X˜gWIp¢w¨¢v‘8¢XGE9W6f!z¢G6IQ8†3Š‹6WIe„QIEC†IVshIW9G‘oQX696H t H h 5 4 5 D Hy  7B Y YB F 4B F @ t Ž d H by 5 ¢IxQkG–Xp#EW…EƃEICjI68„8WY  ¢I88¢XIX‚Gf¢‘I¢Q#roÅII8•wEB y T 2 5 ` TB hB d D T q 5 Y 2 T H 7 i ‰ h 2 R 5q Y ûç ¦ à —à ¦æ Á 3sgIW8feýù 4 € q 5 2 5B D T Ry T b H H F t ’y T 2 5 ` TB hB d D T H Y T 4 YB F D 2B t F D hB @ h 2 T ‰ yy T 7 Y T ` 2B b T F Y H 4B d D T ¢¬8EGVUE¢V6puºŸ¢I88¢XIX‚GAW¢Q#EIGÆXu‹WpXCpp¢UuXVd‚r)IExcI#6QX‚GV#7 h H h 2 T ‰ q 5 7 d 5q H F D 2B d T H   T i H F t ’ Y 2 5B D T 4By   T H 4 2 H d H gB h H DB 2 n 2B 2 5 7 7 5 4 i d H b H d QpI¢Uh¢ÏW¢•CpG–Xˆ¢g†I¢#‡Ium6I8EGVQXXpI¢fQI6WQsEp‘WEII–E–8#986#fQ¢C‚¢T Y d 5 D T d H  5 H Y H F t ’ Y 2 5B D T R P H d T H 2By q 5 D H Y h Hy  R 5 4 T q 5 2 5B D Ry 5 Y H F D Y HBy  7B d 5 D T d H  ‚8GlGQ†8Ž‚QIu‰Qp8XWVUSQu¢QIXEu¢‘‡Wr6XIp8Qu¢¢XG˜X8‚dIWQXEI9Xw8GVWQ†85 —B d D T 7 H Y d H b 2B 2 i ’ 2 5B D T D H d  d H D 2B Hy  7B Y T 2 H bB ` H ‰ 2 T 4 yyB D Y DB D R ‰ t d 5 D T d H  5 H Y d H b 2 „‚GV#wWW68oX†™€‹8EGVG6WpWQWoXXp#EW–Q¢E8u†¢Q‹XEGfCEfpIUQ¢GVWQ†8w‚WQ¢oXB 2 T q 5 2 5B D T Ry T b H H F D ` 2B h 2 T 7 H h YB DB H 4 2B Y t Hy D ‰ R Y H d 5 7 YB @ 5 d h 2 5 4 H Y H F D q 5 ` 2B 2 T H ¢m¢d8XWVUXV¢‡IG…pXIIVd6IŠXŠEQpXWUVXWpIW™W89ŠXÿcW‰I86QWpG§¢…IEI¢g97 H F t ’ Iu™§²W Ý gG‚¢…WVƒpQ9WQWoX…IW†EEI‚Q‚…IGaW8W‚IV‹Qp8XWEII¢Q‰‚¢III8†…IWI¢T i T d d T H D TB h H 7 d H D 2B H F D 2B Dy R Y H d H F D H d 5 D Y h 2 T t Y 2 5B DB h 2 5 4 i d T h 2 R 5 ‰ H F D h h tg„QI6W6h„IC›² Ý gWW¢8W668†pG¢Vƒ¬YxgG¬„I9XWd£1 6¢gEQ†ºXp8IWk6IQIS6WY DB H 4 2 H d H gB h t i T d d T d 5 D 4 H b H F D H ÷ T D X Y i T Y iy  7B Y D ’ d T Hy 4 H ‰ hy R 5 F Y H 4 2 H R P H YB F D 2B d 5 D 4B h H d  H F D q 5 ` 2B 2 T H 7 H F D t 2 5B D T D 2 H 7 Hy  7B y T 4B D 4 T d  5 D D 4 H  Y H d F DB EIGX…8W6EIQ‚IeIGC¢IXIVg9ŽIG§‹8EGVGS6#6XI9Xv¢gEG‡¢GIG–66s‚QWÅWEª™ ç ¦ $’ ¥ s˜dpzæ ä ä ²Ý  ( ²W Ý ûç ¦ à — à ¦æ ù W² Ý € ÔcgIG8ÔeýÁf“ 4 ß   y AŒ Çw² Ý cgà ¡ 8w£ŽÁ …î º² t ø ç ¦ à ¦ 6 æ ¦ H 4 2 H R P H Y Ž d 5 D 4 H d d 5 4rd 5 D 4B h H d 6IQIS6W–8W66WW¢Q3Ô8G‡XI6WI  pGeoQ‚W6WI˜‡6Wu††¢Q†ƒEIC@ H F D i ‰ h H Y Y H d  — H H d H ‰ 2 T 4 F 4B F ç ¡ dpzæ $’ ¥ ¢ ß    y AŒ Çw² Ý sQà ¡ 8w£eÁ …î º² t ø ç ¦ à ¦ 6 æ û ç ¦ à — à ¦ æ Áù “ ä ÔsgpG8fŽ¸Uð² Ý  ( 4 € ¦ 5 D D 2 Hy T bB R P WuS6X¢V„ISQH YB YB F D iyy T 4B D T 7 H F D T EfEIGdEXVgXWV#QIWV „ ‹Q‚8†#EeQQ†…¢cIfI¢XG„II86‰‚¢UpI8†f‡XIƒEWEkƒXIC‰XB ’ h H Y 5  7B 2 H H ‰ H b T F Y 2 5B DB h 2 5 4 i d T h 2 R 5 ‰ D Hy F 4B dB ã F 4B F @ 2 ç h ’ ¥ o3— pzæ y AŒ t º² ø ç ¦ à ¦ 6æ w² Ý cgà ¡ 8w£ŽÁ ß‘î   ¦ ä e² Ý  scgIG8ÔeÁ (ç ¦ à —à ¦æ ¦ “ YB c ’ ¥ ’ P  d 5q 2 5B D T D 5 2 B d D T 7 h H h 2 T X‘3— p†ëjp8•Ž8XWVƒVIŽ— WWVd‰QpI¢U‰ i ’Q8WVG6s¢9„WWV#…§¢5 W8•™pGdEŽ˜pG69™9¢WIGu‚QWpp–•hEeI6IQ‚I#¢Q…Wd‚VgH d 5 D T d H  5 —B d D T 7 T q 7 d 5q H F D 2B h 5 F D H 7 H 7 T Y H F D Y Y H d  — H H @ qB h 2 H F H d  7 5 4 5 D i Y T H DB R P YB 2 5B D T R DB Y H F D t d H b H @ 5 € ’ h HBy   T H ‰ 2 T 4 c ’ ¥ ’ P  @ 5 F D Y R Û D Y d n D T D 2 H hB b H D 5 WEISŠEŽ8EGVUWE‚IGëQ686vcfhQXEII¢u†¢g…— pI†jfcIC‚I‚'fWUŠl‘S6IXx‡C¢I2 iy ‰ T ‰ 5 d  YB DB d 5 D T d H  5 D 2B 5  T q 5 7 d 5q H F D 21 ’ H h 5 4 d H D R  7 5 4 T 2 7 H F D D 2 H 7 Hy  7B 5 EIVU8WpjXˆEj¢GVWQ†8ˆoX8†wh¢Ï‚8•uIG–3”VI˜6ˆQG˜I#¢Qu#EB QIWvoQ9QEI9XwGD d H F D 5 2 T H DB R P h 2 T D  H 4 2 5 4 2 TB DB 7 d H € H F D ` 2B Y R Y h 5 F D H 7 D 4 R d D Y 2 5 4 5 D ` 2B F D H 2 5 YB D 6IGVI¢aGEpx†I¢pQ6I8QeVƒG„#‚Q‘{IW#IEWI‘IxIG‡#f6pWGfI86dG–IEIGuI8CXf£1 } | ã â  r e © Up Ø | d § e i e g j | t g ã y y T ã | t g j t | †Ç£UuVfD¨ih†7v†s`Qv†Úr d ߗ ߗ ˆõˆ– ’ T y R 7 d 5 q R} ˆ  P T Y T 5 D h H d d H q H d 5 Y y T Y B Y B F ˜ƒI…‚8•€SouQ‘¢9G‰Q‚WQ•‡W‘‚X¢CE‘EIut ç c ’ ¥ o3— pzæ ç … ß îæ ’ ä ç4 üí Ý ø 4 € í Ý 6 ‡…zdfˆf5Wx¬›¥xf†æ ß‘î ¥ 6 45ü‚í W ß  V Ý  ø í W ß  4 € Aí W ß  V — ¬ø V Ý  Ý  Y T h HYY Hd  — H H ‰ 2 T 4 h 5 FD H 7 H F D h 2 ¢k6W‚QWpp–us¢Q‰xIW69wpGI¢T 9 3¥ Í Ì Ö Í ˆfu× ê Ó É Ñ × Ú Ì Ù Ì È Ç ë™¬uv†vf”‹Ú qÔvwˆfˆ””ÔÖ {Ô§‘•Ïm˜™uuóx%xÎ é ÓÒ Ç Ó Ì Í Ì Ñ ÑÒ é Ó Ò Ë Ç è Í Ë Ú Ó × Ç Ð çÐ .

’ YB Y iy T 2 T d 5 d d H d HB d R 5 ¤ ` 2B Y 6XWx„¢U¢C8‚W6uQX‚I8U9IX‚IR h H 2B T D ‰ 5 H ‰ 2 T 4 H 7 H F 4 Y H 4 2 H d H gB hr H DB 2 n T q 5 d 5B b T F H ‰ d 5 d d H H F D D R 5 ‰ T 2 5B D T 7 d 5q 2 6IX¢Gp8Ž†¬¢ge9QpƒWŽ6IQ‚QsEI£‡G„IUf¢…¢XxgUQ†…8‚WQepG…p¢s¢8EGV#W8•˜XB d6IGfIUj‹8EGpEW6WQp96G„#EX‘„WXVšauQpXxcWIvEx¢‚IW¢Q#sI•6‚Iw•X€XIW‘EXIªvVXIVƒD H F D d R ¤ ’ 2 5B D B d 4 Y H h h H DB 7By iy dB Tq T Y H hB b 5 d  DB t H d R Y T H 7 y Rq H Y R T YB YB F D HyB F ™ ’ Hy ‰ T d¢XxgvŽp8W•eQpX9WQW6IpWQW™8‚WQepXI¢QX‘IWvVQI‚8‘IG€¢9W6GXrQW‚Q‚I˜‡H 5y i T t T 7 5 dq h H 2B 7 d H D H h 7 d H D d 5 d d H ` 2B h T Hy H F D q 5 d H h d 5 H F D q 5 Y 7 d H D 2B h H Y Y H d  — 2 H Dq 5 YB d 5 D T d H  5 2 T q 5 i 4 T d R 4 4 T H F t ’ Y H 7 H F 4 Y H D T hB h 2 T 4 H F D q 5 i 4 T d R 4 4 T H F D Y Y H Y Y T 5 6G|¢ˆXj¢GVWQ†8#V›¢–‡¢GI6Q¢fIuˆQ6#6IƒWCWVUXpI¢g‘IWs¢9‡¢GI6Q¢‘IWv‚QW‚¢aGD Hy ‰ T H ‰ D Y R 7 H 2 5 2 5B D T 4By   T 2 H bB ` T d 5q H 7 H F 4 Y H 4 2 H d H gB hr H DB 2 n T D 4 Hy H Y 5 D d H h d 5 2 XI¢wsf‚p…ap88XWVgEXIIVQ¢E89™¢•a9QIƒ‚a6IQ‚QsEI£‡G„IU–66X6W9Gf6IW¢31 ² ¹ ² ó ˆ ´ g”‰¦‰¶ ç $ $’ ¥ 8dpzæ G À à À À † À ±¹ À î à i9ˆˆó‡6€eëe¸ 6 ß  „® …­  ß y …Wð`Ž² Ý saSpcgà ¡ ™x˜cg8w£xkÁ ‘î   ¦ … ß ît ’ ä ç ¦ 6 à “ ¦ ¥ 6 à “ ¦ à ¦ 6æ Ü ¦ ß   y …Wð×d² Ý swxp¥à ¡ p1S8fxkÁ …î ¦ … ß ît ’ ä ç ¦ 6 à c à c 6 à ¦æ Ü ¦  ß   ² Ý  ç — $’ ¥ SVdpzæ 6 ²Ý  Û ç Y H 7 H F 4 Y H Y H F D ` 2B iy   T 2B Y H R Y YB y T B 4 2B d  H F D q 5 H 2 5 YB YB F D F ` R 5 F Dy T t Y 2 5B DB h 2 5 ‡ƒ6#6IƒWu‚QIW–pXxEIp¢‰XCQI‚WE‹¢UEQIEWIuIWm¢uI¢uXuXpGŽ8I¢IG„¢UQp8XWEII¢Q4 id T h 2 R 5 ‰ H FD i ‰ h HD T Hd ‚¢III8†IWdoeQWVg‚Q4 QEI8WpIG‘W8p8Xv™W6Iš…W¢Ž6¢WIQ6¢§¢Š6IW¢™9¢GpGD 7 Hy ‰ 5 d  H F D H d 5 2 `B H @ H d H Fæ H d T i 4 T d R 4 4 T q 5 d H h d 5 H 7 T Y H F F DB @ Y H 7 H F 4 Y 2 TB ` 2 T d ` T ô 5 @D t d Ty R 4B D d T  21 ’ Y H 4B h 2B H 4 T  Y q 5 h T H d  Y d H hB @ T H b T F yyB WECv6#6IƒW‘Vƒ8IVG8¢˜w•386¢ƒI6XW‚¢U#3v6QQEIIE™QVUWh¢d¢QWI‚ˆQIECu™8gUIEXC@ i H FD D R ‰ 6IWvppot   dpŽIVw— p˜QSjËX9Q8„8ˆIxIG‡#‘IGˆ¢9‡¢GI6Q¢‹¢ˆQI‚8‘#¢WCIG‘8gUF $’ ¥ h 2 T c ’ ¥ ’ Y P  2B 2 H bB ` Y h 5 F D H 7 H F D Y T i 4 T d R 4 4 T q 5 d H h d 5 H 7 T Y H F D H b T D T F D h H bB d H h H ‰ 2 T 4 Y 2 5B D T 7B — 5 d   T 2 TB ` 2 T d ` T ô ` 2B Y R Y H 7 H F 4 Y D 2B 5  H b n D T F D t H Y d R 5 lUGŽ68E‚QIfsd¢gˆI8EGV#X˜gWpI¢#¢ƒ8p¢G8V§wIX‚IuQ9QpƒWvSE8†8I€VUWp¢‚WI¢Q4 q 5 t Y R 5B b ‰ 5 YB D1 ’ Y D 2 HB h T d `  H H D Y q 5 i DB 2B 4B b H F D 2B h 2 T Y HB d T h 2 R 5 ‰ D T Y R 5 H ` T D 2 T b h T H ¢SQp8Xxo8ˆXv£”QWoQXp¢G8dQQW‚h¢…3„IX6Xx‘pG–Xdp¢€QX‚¢UIp8s€VvI868¢GSV¢˜¢™†‰ 2 T 4 YB F D t 2 5B D T 4By   T 2 1 ’ h H D T Ry T b H ` 2B H ‰ H d T Y H bB D T bB d H h H F D H d H F @ D 2B 5  H F D 5 D y T 4 5 ¢QCXIWx‹¢XGlgXEII¢†3ˆ‹QWVUE¢V6–IX6su‚¢‘Q8„GVV„WQpwpGuWQpCuSE8†wpG‘GhVgxXy yyB D Y YB D T F D 2 5B D T 7 d 5q 2B ` 2B Y R i 4 T d R 4 4 T 2B H Y T H d 4 2B YB F D D H ` i H F D D T F D YB D 2B 5  Hy D ‰ R XEG‚fX™VUGe8EGV#W¢•pXŽpXWp‡¢GI6Q¢X…‚¢g‚QIE™XpG™‡8†6IW™VUWfE™oX¢s…EGppWY Hd 5 7 T t d H b H @ 5 W89Š6Q8‡•c‘€ 8‡¢GpQQV©6W¢QW6IXaXQpGC¢WIVgQ†–Q9QIƒ‚r¢X8I¢W8¢˜‘oE8s£l¥ ’ i 4 T d R 4 4 T h H Y T H d 4 2B dB H F D q 5 H Y R T 4 H ‰ Y H 7 H F 4 Y 2 TB ` 2 T d ` T ô D 2B 5 r d H b 5 H ` T D 2 T b h T 2 T H b T F i H F D iy d T Hy  ’ Y 2 5B D T 7B — 5 d   T 2 TB DB 7 d H € 2 5 h d 5 @ y T 2 n 68cu8¢ƒo¢VpV¢a8gU‡IGŽ„W¢QX•vQp8XWVdE˜c‚II¢VƒG„#‚Q‘f8†W8v‹¢UIi ’ ¦ $’ ¥ 5 D c ’ 8˜dp9G…3— p¥ ’6Sˆ XwI8X‚W6WI˜‡9IWŠ¢ko¢†Gj¢gEGoQpXuX–8„GVV„WQpwfWUu¢aI‘IXEIW9IWaVIGD Y P 2B Y 2 5B Y Y H d  — H H F D q 5 i 2 T 5 D y T 4B D 2 H hB YB H bB D T bB d H h D Y d n T q 5 D n H 2By  Y H F D D T F H D 5 2 H ™ ’ 2 @ 5 2 ÷ D 5 2 YB D n H 2By  Y T 5 D 2B DyB R ‰ iDB R 2B D 2 5 4 y T ‰ 5y ` h H Y T H d 4 2B H F D i ‰ D H Y g W¢I€hCspxŠ¢IŠEuI™IEXI‚…–GoXvEEIId3EISXWo8Q†¢U8E8Q‚¢g‚QIEIG#SC‡Ws¢5 YB i 4 T d R 4 4 T 2B 2 5B D 4 R h H d YB F D F 4 R EA6¢WIQ6¢)E)8XW6IpQWaXIWrƒI…7 c‘f–ç  …zŽ’ „I8wX6¢WIQ6¢C¢aQI‚8aWEwpI‰ @ 5 € ’ ß îæ iy 2 5 YB i 4 T d R 4 4 T q 5 d H h d 5 Y DB D R  ß   ¢ y ¦Œ t w² Ý sgà   x¢feÁ …î ç ¥ $’ ¥ o¢dpzæ º² ø ç ¦ 6 à ¦æ û ç ¦ à — à ¦ æ Áù “ ä  3sgIG¢feý‘ÿe² Ý ¥ ( 4 € ¦ i o‰ 2 H b B ` Y B D 1 ’ d 5 D T d H  5 H 4 2 H d H g B h — B d D T 7 `H h T Ü T q 68E¢vE•£mQ8WVG6s8fQI6W6h„I–„‚GV#™¨¢vw˜¢5 ‚8•‘p8ˆX–8XWV#X˜gWIIV™IEXI‚fpGD 7 d 5q H 2 5 YB 2 5B D T 7B — 5 d   T H 2By  Y H F D T F D h H 2 5B D 2 H 7 H ‰ hy R 5 F Y D1 ’ Y 2 5B D T d H  5 y T 2 5 ` TB hB d D iy 2 5 q 5 H Y R Y H ÷ T 7 h 2 T lUG9QI8EGoQ9‘s#XI8pWv£vQp8XWVG6s8p¢U88VƒIEWG‘„I8h¢™‚I•6¢¢#dIV€ç … …šŽhXB ß îæ ’ Y  ß   ¢ y AŒ t …² Ý sgà   x8ÔeÁ …î ç   dpzæ $’ ¥ º² ø ç ¦ 6 à ¦æ ûç ¦ ¦ à ¦æ Ôcgà ¡ g¢feÁ ù   厲 Ý ¥ ( ¦ ä  4 € ¦ Hy  7 T — H d 5 ¤ ’ Hy ‰ T D d 5y i T Xp#¢p–u8Uˆ¢Xp¢ƒŠ8Excmt T q 5 Y 2 T H 7 i ‰ h H bB d H h iyB Y T H H ‰ 5 Yy T 2 T 4 H bB D T bB d H h h 2 5 4 H Y H F D q 5 Y 7 d 5q 2 TB DB 7 d H ‘¢wI¢g9)oÅ68E‚QIA„XWVgŽ†WE¢¬¢QŽ8„GVVE‚QI¬I¢QQ‚ŽIWu¢w9W8•A¢ƒWE9W6f€ } p  p u t  r p Ø r w y r p u " ‚xCuƒX‚s“€su¡xg Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u‡§™†fjˆˆ«”w€‘ˆ””Ԑz€mQ‘Ôv«Sψm™‘”Ç % r p u wsuvj w t— ߗ 6õˆ– c ¥ .

H F t ’ H 7 H F 4 Y `H h T Ü d H h d 5 r F D d R 5 q H F D h 2 T H 7 H F 4 Y H 4 2 H d H g B h h H d H D 2 H 4 d H h d 5 r F D d R 5 q h d T IuˆV#6IƒW¨¢vvQI‚83xG‚p8•fIWdI¢#6IƒW6IQ‚QsEI6WQWoQQ€QpW8£˜GfI8•d‚¢Uh r 2 T D Y H F D d 5q Y 2 5B D Ty H d d TyB 7B Y F DB @ ` 2 5y T t ’ ¥ H d R `B x¢ƒfApGe8•ŽI8EGVXQ‚#¢XX9X‚ÿWECÅI8E¢I— pk‚I8Xm¤ Eð6G‚¢XIE¬%pÿ8EGVUSj 2B h H D D 5y  YB c $’ ¥ 2 5B D T R P ‡ ‡ xS‡ ø “ ß  ‘î „ – 6 – Ókä ß …î ß î – 2B …UfX‚Y H 4 2B Y t h H D 4 H  — H H ‰ 5 D YB Y T t i 4 T d R 4 4 T d H h d 5r h 2 5 4 H Y 5 D – Y H D T 7B — 5 d   QIEW†‹QW66s˜‡†9GŠXC¢ë¢6¢WIQ6¢fQI‚83˜p8Q6W–G‰f6GV#X˜gWpI¢T ç c $’ ¥ ovdpzæ ß ‘î ß î – 2B …ëfEWY ä h vv– h jlUGGVfò – DT FD HD5 i ‰ 2 H bB o‰Q8„8` YE†6swISQ8g•ºQU„Ix#kIWd8GlGQ†8kQpQW6h„Ih 6W6GoQ6†6IW8£˜I86Q‚eIGŽ8U¤ B d H ‰ 7 R 2 H b T @ h H nB h 5 7 H F D d 5 D T d H  5 H 4 2 H d H gB h Hd H D 2 H 4 d H hd 5r h 2 5 4 HY H F D d 5 ’ 2 5B D T 7B — 5 d   T H F D q 5 i 4 T d R 4 4 T H F D q h8XWVdE˜c‚II¢wIWmV6¢WIQ6¢wIG§¢5 H‚IWVg9”Xˆ6s…px68g•˜¢UW6¢ŠIG§QGldEpg‚II¢€Q†…IS68gŠwQU„Ix#ŠpG‘ƒXICfGD d R Y T H 7 T YB d H ‰ 7 R 2 H b T @ y T R D 4 T H F D Y H D T 7B — 5 d   T d H ‰ 7 R 2 H b T @ h H nB h 5 7 H F D F 4B F @ 5 H6W86IvpGmpxuv‹8X‚W6WI˜‡v6¢p–vIWwEm‚¢g†I¢V‹iQ6s…px68g•€IWv‚QIC˜W¢Qsp¢mEB H d ` H h H F D Y R F t ’ 2 5B Y Y H d  — H D 4 T — H H F D 2B Y d T H   T t – t d H ‰ 7 R 2 H b T @ H F D H d H F @ Y d T H   T D H Y R T 4 H ‰ h H 7 T 2 5 Y YB d H ‰ 7 R 2 H b T @ h H nB h 5 7 H F t ’ d H ‰ 7 R 2 H b T @ h H nB h 5 7 H F D Y ‚I¢g6saQ9¢U‚mEjQ†…ISQ¢gŠuQU„Ix#vIujQ6s…px68g•w6UEIx9•IW§EB h SW6IC@ – HdH F ç 9 $’ ¥ 8dpzæ   ‘•  þh – ‘• ¥v6˜ ä ä ä ä ‘• þ Ôs…UfX‚og6 …Uƒ8Q•ª's…ëfEWoV9…냢Qš„ù ûç ß î – 2B Y ˜ ß î – Y 5 4æ 6 ç ß î – 2B Y ˜ ø ß î – Y 5 4æ ß ‘î   ߅î þ ˜ ߅U–fX‚Y î 2B ß…î   ¦  —• þ sç  e …‘• ß ‘î   € þ 6  ¤“45Wí  ü ß …î   ¢ e G—• e …‘• þ “æ þ ä ¢ ¤4 € “¥í  e …‘• þ 6 4 €í Ý 6 4 üí Axł5WxÝ t ä íç ( š†Ý  ˆæ d ¦ ‘• þ ä í Ý 5 D d 5 D T d H  5 H 4 2 H d H gB h h H d H D 2 H 4 d H h d 5r h 2 5 4 H Y T t Hy  7 T — H d 5q t iy   T H @ q #‘x†Wu¢GVWQ†8d6IQ‚QsEI)6W6GoQ6wQpW8£˜I86QWm¢Xp#¢p–w¢•§¢EIIVdvŠ£1  þ –˜ ‘• 6™ä ‘•  ß  þ  D H ` H @ t ß î ê ä í ß Hd H F 68w•I˜…S«uxuW6IC@ ç “ $’ ¥ o¢dpzæ Y XB • þ q 5 H b B D T b B d H h D Y d n D 4 T — H H F  ‘ ¢¢EWVVEW6I‘‚‚UC6¢p‡wIut ’ Y H bB D T bB d H h d H F `B F 5 D Hy ‰ T 4By   6Q8„GVV„WQpuQI8EI–G™EI¢QXXpI¢T iyy T R P H YB YB Y iy T 2 T H F D F ` R 5 F Dy T t Y 2 5B D T 7B — 5 d   T H bB D T bB d H h D Y d n 2 5 t H d H F t H D T d D 2 H 4 2 5 EE¢USQ€X€X‚xE¢UVIW98p8IG„¢SQp8XWVdE˜c‚II¢f8„GVV„WQpŠ‚‚U#¢p¢‚QI˜¢WVGWS6QI¢Q4 yXECwr’ — U¢Q¢EWVVEW6I™6GV#X˜gWpI¢¢GVWQ†8…QpQW6h„I3–G„IUQ¢E8ŽvX6•@ cIF yB @ H ™ þ q 5 Y H bB D T bB d H h Y H D T 7B — 5 d   T d 5 D T d H  5 H 4 2 H d H gB h r H DB 2 n 2 H bB ` T yy H @ 5 • HpX9¢p‡uG€‚6WQWoX‹Vf‚8•6‚QIW”Xv£€6Q†…ISQ¢gŠCIG”Ed$‚QICot ‘ þ W8•uIW9E‘‚¢T 2B 7 T — H 5 D D Y H d H D 2B q 5 H d 5q H d H F D YB D1 ’ d H ‰ 7 R 2 H b T @ H F D YB – H d H F @ • 7 d 5q H F D 2B H d  F 4B F @ t Y D 2 H 2 5  7 5 4 d HB d R 5 ¤ Y DB 5 D 2B h H Y 5  7 5 4 H h H ‰ 2 T 4 2 5B D 4 2 Rq 4B h 5B d H  i d T d DB ‰ d T 2 ƒXpC8QGo6I8†#86v6XWp8U”G„GoX‘6W8†98Q6Ius‘Vg‘8EG6pI•€EIxXW6sa‚VGG„IWV–™i r e j | r p y w ã |g j | y  U7s{zD†7uõ”| h ¥ e j i eg j | tgã y y ¡…k†7v†£×T ™— t— sÉ6– Ú Ò Ú ê ’ É Ó É Í × Í Í Ì Í Ì Ò Í è × Ñ Ð Ð fxˆ‚“f™óCfˆóˆÔCmӀ«S‘xÎ .

{ …i s u ž §› › { z y s w u Ô « Ñ   › ž ­ § e « ž § ¢ ­ ® ž ¢ § › Ô Ô ¯ ž § ¬ ­ › ¨ § t s ¢ ° « § ¢ ›y ® ~} rq“ § |Ô 5« § Š¢Ñ ¦ § Ÿ ¢ Ò ­ ¤® d¢†ª “¢ ­ †Ñ ÉÔ Ô Ô ¯ ¢ © ¤§ q § b   ¬ Gi gCExvs™¦¤Š¢qd¬A¤qGÚd©£“A‘‚dª¤“£kÉx“d©¤F¦q“xfn¦¤¥˜z­ £v¦pQ¤vs¢ ¥F“S£d­“qA£‘'%ªk¤)¦5Éx¤ˆ£X¤¬¨ pp A­ « ž ­ Ô ¬ ª ž « © › ç h $’ ¥ ovdpzæ  ø ä ç ß î – ¥ 2B ‹ ø ß î s…ëofEWY „ “‘‘U–   EWY “‘‘UjX‚˜ˆƒ˜   2B ‹ ø ß î – 2B Y 4 ‹æ " ß  " ‘…ën86x4 " æ   vä " ù ‘î ç ß î – ¥ Y 5 s…Upë8Q4 „ ø ß î ……U–   8Q4 Y 5 ø ß î – Y 5 4 ø ¦ h – v6˜ ä íç ( ŠzsÝ 7  ˆæ h 2 I¢T YB d H ‰ 7 R 2 H b T @ h H nB h 5 7 ` 2B h 2 5  Y H d d 5 4 H F XfQ†…ISQ¢c•ŽQIEIx#–IEII8†W6W‚8QwIut " û € í Ý Ôç „ ¥xÆø „ Wx•æ „ üí Ý ø ›ç ¥xÆø €í Ý  ß  Wxšæ  " ›f¥˜¬h5Wxš‡4 " jxGä " ù …î üí Ý ø ç4 € í Ý ø 4 üí Ýæ ø í Ý ¦ û € í Ý 3ç Ax)6 „ ß „ ‚˜•æ „ “›ç  ¥x)6  Wxšæ  ~›fAxł5‚˜•Q¡¸ù …î üí Ý ‹ ø €í Ý üí Ý ‹ ø ç4 € í Ý 6 4 üí Ýæ 4 ‹ ¦ ä íç ( ušsÝ  “æ ” 2 H F D t ¥ ø ê 5 D ¥ 6 ê 7 5 dq ` 2B h 2 H D — H Y H 7 H F 4 Y 5 D D Y H d H D 2B d R 5 D 4B d D Y H 6IGI˜©wW‘¬½Ï8‚•9pXII6Gx‡fQ9QIƒ‚–GC‚6WQWoXuI8‘‡XWW‚6Wd H @ q vv£1 … Q¢GVWQ†8…X‚G‡#9x‚ŽuXwzç 7  “©I¢8GVWQ†8wEWW69#x‚EGo¢A¢Eašç  “†W6IC@ ’ d 5 D T d H  5 4B d D H 7 7 i Y T YB í (æ h 2 T d 5 D T d H  5 4B d D H 7 7 i YB D 2 T 2 T YB í (æ H d H F ” í (æ ø í ( zç 7  “ÿ”zç  “æ ” ä í ( Cšç  ˆæ 7 d 5q H F D 2B 2 H D DB d @ H ‰ 2 T 4 Y d 5 D T d H  5 H 4 2 H d H gB hr H DB 2 n ty T d H 2 H ` 2 W¢•upGeXQWW„WCs‰¢QuW8WVG6s8u6IQ‚QsEI£‡G„IUpz¢G6IQ¢†31 l t H ŒB Y 2 H bB ` T q 5 F Y H 7 T 2B Hy ‰ TyB T b ’ j l …ë– m¡ ¢6X‚†68E¢–v¢‰W6#‘‰XfXI¢XX¢VgT ß î H ` 2 T d d H ‰ 7 R 2 H b T @ H dB D 2 H H F D d 5q h H by 5 Y H d YB 2 5B D Ry 5 Y H F D q 5 D 2 H 2 5  7 5 4 d H ‰ 7 R 2 H b T ¢I¢Gu6s…px68g•wWEGoQ…pGf¢•k68„8W6W™E8EGpE8W…pGj¢o6I8†#86uQ†…ISQ¢c•@ 2 H bB ` T F 4B F @ F DB @ i 4 T d R 4 4 T H F D Y H hB b 5 d  d H ‰ 7 R 2 H b T @ h H nB h 5 7 H F D d 5q 2 5B Y Y H d  — 68E¢¬ƒXIC)WECr‡¢GpQQVeIG…6IXxcWp96swISQ8g•rQIEIx#dIGw¢•¬¢XW‚Q‚I˜‡H ’ Y H 7 H F 4 Y Y R 5B d T b d 5q d H ‰ 7 R 2 H b T @ h H nB h 5 6Q9QIƒ‚uI8X‚¢VŠ8•f6swISQ8g•eQU„I˜‰ 0 0 0.5 2.5 1 2 3 2 Central nd κ∆ x 1.5 3 ¡ — .5 1 Û ’ ¥ H d R `B I— ˜uWI8Ev¤ Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u‡§™†fjˆˆ«”w€‘ˆ””Ԑz€mQ‘Ôv«Sψm™‘”Ç κ*∆ x 0.5 th 4 Central 4 Pade th 2 2.5 1.

› ¨ § c ¬ž § ­ ҟ ­ ¥ ›   ¬ ‘“udd“A‘Õ¦¥£q‘¦­ ° š ¬ž ў ¨ Sds§ ¸q“© —–”•6’42VQÐ “ ‘  ‹ ‰ c ž§§ Ÿ ¯ e   Ÿ   ¬ ­ ¨ ›¢ Ž Œ ‹ ‰u © ­ ¨ª Ò© © Ô¢ ›§ § ­ ¨§ ›§ Í ¦0dd¬““†d›½x‰sd›a‘¦‘¤A­ ™ xz Ð ¦ŠS¦Ï˜‘“U¤£¤¥‘¤“¦« sˆ ‹v˜™Œ " ß ‘î   … h i˜fˆ‰… ¢ …ëfEWY ˜fˆä Þ … " – ‹˜ 6 ä ‹˜ 6 ß î – 2B Û ç Þæ … d 5q h H 2B T D ‰ 5 YB  㠖‚•”f¢•6IXVƒp8uE©“ ` 2B @ 5yy 5q H F D t 7 d H D y TB D T  Y H F D 5 D h HBy   T H d T Y H 4 2 H d H gB h h H d H D 2 H 4 d H h d 5 r h 2 5 4 H Y q IXCcEX8•uIGU‰W6Gh¢XGVIWwIG9GeQEXIIVw‚¢™6QI6W6h„IeQ‚QGo6Q‘QI‚83˜p8Q6Wj£1 ¤  ¡€  ” ¢ —• þ æ … ä çÞà ßæ sç ¡ ”†j‚W˜|vÝ Þ Y T 2 5B D Ry 5 Y D 4 T — H H F D Y H bB V8XWpX8‚™‡¢p‡uIWC68„8` ¡ “ ˜ë†jëGoE9pXG˜GEW‚IpWI¢u‚•vf8•…IXx„8˜” ’ ¥ ’ P  5 D 2B ` 2B D R DB D Y ‰ R Y h 2 T ç Þæ … d 5q ` 2B by 5 … – ‹˜ 6 £i˜f¼ä ç ¡ “ pzæ ’ ¥ " … "  㠓 H F D Y H n YB D T Y ç Þæ … H d H F aIWC6U‚EGVW•‚•”‡W6IC@  ‘• þçÞæ … ä çÞà ßæ s‚|”†j‚W˜|vÝ YB – d H ‰ 7 R 2 H b T @ F DB @ 2 5B D 4 2 Rq 4B 2 5 7 d T F T i ‰ h H D TB DB 2 E†„6swISQ8g•eGECe8EG‡II•™EI89W¢U‘‰oŽ6GVXGEpXB 2 5B D Ry 5 Y T D T F ¢XG˜X8‚–VUWD   p¥ ’   8XW6QxÄ¢W•ˆX¢QQC’ ‚GD $6 8‚•eIXIpQGx–rXVd8p¬85 ’ 2 5B D 4 H ” 7 5 dq yy T 4 H ƒ  5  7 5 dq ` 2B h 2 H D — H 2B T 7 5 h T 2 ß  Þ  ¡ ä ‹ “ø Ý  Ý  7d 5 ‚8•q H F D 2B 2 5B D T R P H 2 5B D 4 H b 2 5 4 d T H 2By H F D 2B T ` T H 4 2 5 d H hB Y 2 5  ’ 2 5B D T D H d  d H D 2B y T 4B Y i F  pGE8EGVUS6e8XW6Q¢o8Q‘¢gpXXaIGXVo¢‘6I8™6IX‚I8•™‹8EGVG6WpWQWoX§¢gEWxoIdT 2 H bB ` H ‰ 2 T 4 Y d 5 d d H H F D t 2 5B D T R P H 2 5B D 4 H b 2 5 4 d T H 2By H F D q 5 D — H D 2 5 4 H F D 21 ’ Y 2 5B D T 7 68E¢#†r¢Qa‚8W‚Q#IG˜‹8EGVUS6)8XW668o8QVgIEX9IGŠV…x‡WS¢Q9IGA3#6I8XWV#XB r — 5 d   T H 4 2 H d H gB h ` 2B Y Y H Y Y T d 5q y 5 5 D y Rq H Y R T Y H hB b 5 d  d H ‰ 7 R 2 H b T @ h H nB h 5 7 H F SgWpI¢–QpQW6h„IeIEWW6W‚¢8•˜8xGvp•6WpduQIExs‚IuQ†…ISQ¢gŠ‰QU„I˜9‘Iut ߅î‚çs…Uƒ8Q¬ø   æ ß î – Y 5 4 ß î – 2B Y ˜ ‘UfEWor¥ ç ä h – •£˜ Y H bB ` F 4B F 68„8†ƒEIC@  e …‘• € þ á6  e …‘• þ ˆæ ß…î ¥ ä  e þh –˜ ø h –˜ — …‘• £€mv6o¬ø  G—• e € þh – ¥£˜ Y H n YB D T Y H 7 H F 4 Y YB F D d 5q d H ‰ 7 R 2 H b T @ h H nB h 5 7 H F † 6U‚EGVWa9QIƒ‚uXIWŠ8•f6swISQ8g•ŽQU„Ix#wIut ç4 € í Ý 6 4 üí Ý f¥˜Å‚5‚xšæ ß…î ¥ ä 4 üíç (æ ø íç (æ — ø 4 € íç ( ”5WšhÝ  “ÿjz†Ý  “UƋAš†Ý  ˆæ i ‰ 2 H b B ` Y B H 7 H F 4 Y `H h T Ü d H h d 5 r F D d R 5 oQ8„8CXu9QIƒ‚a¡VvCQpW8£˜GfI8•q H F t ’ d 5 d d H iy H dB D 2 H ` 2B H ‰ D 2 H 2 5  7 5 4 i d T 2B ` T 7B H F D F DB @ t — Hy  7 5 4 YB d H ‰ 7 R 2 H b T Iu¬68W‚QńQWEGoQkIEQ†9S6I8†#¢QA‚¢IX8¢#XIWÆWECjs‡EI#¢Q–X#Q†…ISQ¢c•@ h H nB h 5 7 H F D t D 2 H 2 5  7 5 4 4B d D H 7 7 i Y T Y H h Ry 4 2B d 5 D T d H  5 H F D 2 H F ™ ’y T H d iy H d R  YB d H 6UEp˜9™IWxQoQI8†98Q‘X‚G699˜fa•QIIEQIEv¢GVWQ†8™IWŽQIª§Ÿ¢QW9E6WIpvE•Q†‰ r 7 R 2 H b T @ h H nB h 5 7 H F D t ç h H d H D 2 H 4æ 4B d D H 7 7 i YB D 2 T YB d 5 D T d H  5 H 4 2 H d H gB hr H DB 2 n H F D 2 H F ‘ISQ¢c•™6UEIx9jIGl‡ë6WQWoQQ•uEWG‡#9x‚EGo¢›E‹8GlGQ†8j6IQ‚QsEI£‡G„IU”IWw6Iª™ ¦ I— Ú Ò Ú ê ’ É Ó É Í × Í Í Ì Í Ì Ò Í è × Ñ Ð Ð fxˆ‚“f™óCfˆóˆÔCmӀ«S‘xÎ .

’ h H dB R P H d H d T Y d 5 D T d H  5 i d T h 2 R 5 hQWEISQ‚w‚¢CW¢GVWQ†8e‚VUII8†‰ y TB 4 H  Y 5 2 t Y 2 5B DB h 2 5 4 i d T h 2 R 5 ‰ 4B h 5B d H  q 5 H Y T 4 H F D 2 1 ’ 2 5B Y R gB h h 2 T 2 5B D 4 H b 2 5 4 d 5 ¢ƒ6Q†WdpUQI8EG„II86‚¢UpI8†XIxX‚Q†”¢aW¢QaIW†3€‹8EWIsEI†I¢e¢XG‡Q¢S¢QC8•q Y 2 5B D T R P H y H h 5 7 d R 5 d 5q h H h H H 2 Y d 5 D T d H  5 i d T h 2 R 5 ‰ H F D d H hB Y 2 5 4 H @ t 2 5B D 4 H Y YB F D 2 p8XWVUSQ§QIx9™I8‘8•eQpQQp™W¢GVWQ†8f¢UIp8saIGfQIEWI86…vs‹8EG‡QWfXIW31 ’QQEW¢III8†dIWal…VIGrIVE¢#8IdpGŠ¢u8EW6GoX#IGAXk8XWV#X˜gWIIVŽQpQW6h„Ih Y HB d T h 2 R 5 ‰ H F D D T D T F D h 2 T 2B T 7 5 h H F D q 5 d 5B d H D 2B H F D 2B 2 5B D T 7B — 5 d   T H 4 2 H d H gB H F D F D 5 ‰ Y H D T d 5  d 5 4 2B d 5 D T d H  5 H 4 2 H d H gB h —B d D T 7 T pGªG¢†eQWVG8†W¢QIEŽ8GlGQ†8kQpQW6h„I«„WWV#¬™t   p¥ ˜ÿ¢XG‡Q˜ÿEB Q‚WI6WEI™i ’ ’ ¥ 2 5B D 4 H ” 2 h H Y Y R 4 YB h Y ² ± ¹ À ´ · ¶ î 6€weeëà ¤ » ‡w´ ² À à » ´ À ± C ˆ9”aj…” ± Á ¶ ± À ± ¿ CejÅ6¹ ½ ¢® £­ ’   88A8XW66˜†X6W‚IQ‚XICQpG‚p•ŠXCXIum¢IpGEEI9¢‰XB ’ ¦ ¦ 2 5B D 4 H ” 2B h H Y Y R 4 YB h d H F D d Rq YB YB F t ’ H h R DBy  7 T 2 d 5 d d H H F D 2 5B D d 5  4B d D H 7 7 i Y H F D h 2 T h H H  Y 2B d 5 d d H H F D Y H 2B 7 d H D H h d 5 D T d H  5 H 4 2 H d H gB ¢WW6CIG…8EGf8s•XWW69#xf‘pG‘I¢dQQ†W–Ej8W‚QCIG”6IX9W6G‡Ij8GlGQ†8CQpQW6h„Ih y TB D T  Y H F D q 5 2 5B D d 5  4B d D H 7 7 i YB D 2 T H F D Y R F t ’ 2 5B D Ry 5 Y H F D q 5 H h R DBy  7 T H F D 2B d 5 d d ¢ƒWVU‚IG›Vd8EG‚¢s‘X‚G699˜fXWSVIWŠISuv‹8EGpE8W™IG›¢CIpGEEI9¢IWdXv8W‚QH 2 T 5 D Y h T Hy D d T  i d T 2B ` T 7B H F D HyB F @ t h H H  Y 2 5B D T ` T  5 d  H F D 2B d 5 d d 2 T 5 D Y h T H ¢)G#p¢gEe‚¢Ił¢IX8¢#XIGXEIC‘hQQ†W©8XWVo¢I8WIIG«Ed8‚W6H ¢ÅG9I¢QXy d H ‰ 7 R 2 H b T @ h H nB h 5 7 H F D q 5 D d T  y T H d H F D D T F D H H Y H @ t ¦ ’ ¥ ’ P  q 5 7 d 5q H F Q†…IS68gŠŽ6UEp˜9IG§¢Cf¢U‹¢QWIWuVUWu6Wuv˜8p“ pëëj8¢§‚8•IGD 8WU¤ 7 5d ’y H b Hy d 5 d d H H 7 T ŸQ¢6E‘¢WW6#VGY H F D H b HB F 4 T 5 D iy H bB D 4 H  Y H pG86EIƒ¢kW)E68EW66s‚QWd ŸŸ ¡ ªIVqeWEISQW9QIƒ‚¨¢v–6IW¢3˜W‚I8•ŽpGD    ¦ h 2 T $ ¦ H d B R P H d H 7 H F 4 Y `H h T Ü d H h d 5 r F D d R 5 q H F h 2 T H 7 H F 4 Y h H d H D 2 H 4 d H h d 5r F D d R 5q D 2B 5 r $ H F t ’ D 2 H 4 d H  ¦ IV‘#6IƒWk6WQWoQQfQI‚83xG‚p8•CSE8†3sÓpugoQ6W6s…„’ ¡ VUG‘WQEˆ¢eQ6s‚…WVUI 2 T F D Y Y Hy q 5 h H H  Y H Y T F 2B d 5 d d 2 T H 4 R h 5 d  5 Ed8‚W6H ¢eQpI˜‚IrWD    ‡0 ¡ ôQWEISQ‚e#6IƒW6IQ‚QsEIª6WQWoQQdQI‚83˜p8Q6WY c Y H dB R P H d H 7 H F 4 Y H 4 2 H d H gB h h H d H D 2 H 4 d H h d 5r h 2 5 4 H H F D t Hy  7 T — H d 5 ¤ ’y H b Hy h H nB 4 H  Y T @ 5y H ‰ Y d 5 d d H H 4 R h 5 d  5 D h H dB R P H pGˆ¢Xp#¢p–‘8Uş686E©6UE6Q†WeµcEQ†‘W8‚WQ6IIxWIAGÅQ‚XIS6Wd     H FD q ‡0¡IG‘¢5 Y 7 d H D 2B h H Y Y H d  — H H ‰ 2 T 4 H 7 H F 4 Y T q 5 i 4 2 HB 4 ¨ H ` 2B by 5 Y H d H F t ’ ß î – r 9W6G)E«Q‚W6WI˜‡ds¬¢Q‰9QpƒW‘¢Å‡IQEQ96epXxE8‚Q‚eIu˜…Uo‚   oÅQ8„8` i ‰ 2 H bB YB H by 5 Y H d YB H b T @ 2 H bB ` T F 4B F @ i ‰ E)hQ¢E8‚Q‚aEe8gŠQ8„8†¬ƒEICkoç ŸŸž|¢x‡X‡¢r€#‡Uu|U|lU•¢w6swISŽIut    æ ‚x … „ }  } 1  è € } œ žx „y ‡ œ q 5 d H ‰ 7 R 2 H F ’ iy Y 5B b H d h H d H hB Y 2 5 4 Y H 7 H F 4 Y H F D d 5q h H H  Y H Y T F  y T 4B d H 7 R 2 H F D Y @ 5 F Y $’ ¥ H d R `B 8E‚IR8E˜‡WI–6WQpXWp8Q€Q9QIƒ‚uIGm8•‘QQ†WC‚¢UI˜¢gEWQ9ISuIWˆCcI‚™%˜uWI¢Xv¤ ’ 7 d 5q y T 2B `B d ‰W8•ë¢UE8X‚85 Y DB Y Y 5y iy T R D H b H Y 2 H 2 5  7 5 4 d H ‰ 7 R 2 H b T @ D 2 H d H gB h i 2 T 7 q 5 ` 2B D YB Y 2 5 4 7 d 5q H b T WEuQH‚8XA„Xy¢UWo2Q8‡aWDoQI8†98Q…6swISQ8g•ao6WQsEpeo¢dŠ¢†pXGfX‚I8Q¡‚8•6¢c•@ T t Dy R H d T uQEI‚YQ‚rŽYfÿ’V8E‚W6Hs‚XII¢5I#EψC¢UX¢XW¢IWÆ8p8IG„¢u‹¢XW‚Q†WEId6QIIxWWoXB i H bB Y d  YB h 2 2 YB  ã Ü y T 2B `B d 5 H F D F ` R 5 F Dy T t 2 5B Y d H  YB h Y H 4 R h 5 d D 2 2 5B D T 7 — 5 d  T y T B d H 7 2 H F t d ‰ 7 R 2 H b T @ H F D q 5 2 5B D 4 2 Rq T Y 8XWVdEB˜c‚II¢”¢g4EWQ9IRS‘IWDmQ6HswISQ8g•‘IG€¢e8EG6pI•ŽuXB h ë6IX˜‘8ECcEWexGD ‹ H 4 2B ” ’ iy @ 5y Y 5 5 Y6GV8¢U85‚I«85EGpE85W‘¢QXWd6#px2eHpGDŠ¢EI9¢p‡dEIG98•¦ l r h fQIE˜” ¢EE¢gEWQ9IS2 H D T ` T  d  2 B D Ry Y y T 4B H 7 R F t Hy  7 T — H YB F D d 5q ‹ ‹ H 4 2B ’ iyy T 4B d H 7 R h D T ` T  5  YB 5B D 4 Rq 4B 2 7 d T 6HGVoVU8Wdp›X28EG6p2I•vXp85#‚¢UFuTaFƒXBIC›VaQ6s‚ˆIG‹Ea6Q†W€W¢Up¢¢QXW6#px€Iut 4 F @ D T h H H  Y H F D YB h H H  Y H Y T F  y T 4B d H 7 R 2 H F ß î …U– ‹ ä h‹ ß î – 2B …U“EWY t Hy  7 T — H H b 5 ‰ T H F D d 5 ¢EI#VI–u8csVapGŠ8U¤ i ‰ d H ‰ 7 R 2 H b T ofQ†…ISQ¢c•@ h H nB h 5 7 H F D 5 D h H D Ty H d YB F 4B F @ t h H H  Y H Y T F  ç h H nB h 5 7 d 5æ y T 4B d H 7 R 2 H F D Y 6UEp˜9™IWwG#QGlƒQ‚vEeƒXIC˜‹Q6s‚W¢IIˆUQU„I˜9€8šv¢gEW6#ISfIG€XB h W6IC@ ‹ HdH F ç ¦ ’ ¥ sp“ pzæ r — 5 d   T y TB D T  Y H F SgWpI¢ë¢XGVU‚CIGD – h – ä h ‹ ‹ ¤  œ¡€  ›” ¢ ‘• þ æ … ä ç Þ à ß y T 4B d H 7 R £ç ¡ ”k€‚‚˜•æ ¢QXW6#px2 Ý 2B T D ‰ 5 H @ E¢ƒ˜8u•pt ¡ “ ˜sëjëGoX–IEGpWEW‚II‚I¢ŠpEI¢Ž8XWV#XB ’ ¥ ’ P  5 D 2B ` 2B D R DB D Y ‰ R Y h 2 T ç iy 2 5 2 5B D T 7 7 5 dq d 5 d d H H F D ` 2B d H hB Y 2 5 4 H d T H @ H 4 2B Yæ iy D 4 T — H  㠓 YB F D ` 2B by 5 ¢W•j8W‚Q‘IGIEW6IX‚I8QC‚¢vuQIEWš†„G‡¢p‡vXpGIXx„8˜”   — Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u‡§™†fjˆˆ«”w€‘ˆ””Ԑz€mQ‘Ôv«Sψm™‘”Ç .

8 1.5 3 .6 0.4 0.¹ ç — ’ ¥ S¢“ pzæ ’ ’’ ¡ ¡ ´´µ ³´´ ¹ ’’’ à ·· ß…î   ¦ w6 ä y AŒ t º² ä Ý ·· ¶ !‘6 ç ¥ ’ ¥ o8“ pzæ ·· ¦ ·· ¶ y ¦Œ t º² ø ²Ý ‘D» ¦ ¡ ¦ ¦w6 ¡ ´´µ ¦ w6 ¡ ³´´ ß…î   ¦ w6 ä » F DB WEC@ ä ²Ý £ ( 2B T D ‰ 5 H @ Y H 4 2 H d H gB h h H d H D 2 H 4 d H h d 5r h 2 5 4 H Y F DB @ t Hy  7 T — H d 5 ¤ ’ i d T h 2 R 5 ‰ YB F D d 5 X¢Gp8w•CQ6IQ‚QsEI†QW6GoQ6f6IW¢3˜I¢QQ‚GEC˜¢Xp#¢p–u8Uv¢‚¢III8†uXIWŠ8•q h H dB R P H d YB D 2 H 7 D T H d D y TB 4 H  Y 5 2 2 H F D t 4 € í Ý h 2 5 i H ‰ h 2 H D — H D 5 2 Y H 5 h 2 5B D T 7B — 5 d   Q‚XIS6WwX…o6#WVg‚Gˆ¢ƒ6Q†WIrQIWvƒ¥˜¬I¢86†ÅI6Gx‡w¢Iw6˜p8EGV#X˜gWpI¢T H 4 2 H d H gB h d 5B d H D 2B H F D Y T ` 2 5y Y T D T F D d T Hy 4 YB D1 ’ i d T h 2 R 5 ‰ @ 5 q 2B H F D D Y d n d H hB Y 2 5 6IQ‚QsEI§8X‚QWSEŠIWv¢I¢X”¢”VUWv¢gEQmXj£m8‚VUII8†ÆcUpX•IG”fWU”6IX‚I8• ’ h H nB 4 H  Y YB ä Ý h 2 ‹6UE6Q†WuXC!‰I¢T ç   “ pzæ ’ ¥ Æ a•Cà û Å Ý ‡ ‡ xS‡ à  C–£Ý Ïe² Ý Ý à 4 ù ä YB Y 2 @ 5 2 ÷ 2 R q 5 d 5 D 4 H b H F D Y R F t ’ i d T h 2 R 5 EfICcISIpv¢‘8W6Q¢IGCISuˆ8‚VUII8†‰ @ 5 q D R 5 H F D YB i d T h 2 R 5 ‰ h 2 T Fr D F `B d H F D h 2 T t i d T h 2 R 5 ‰ @ 5 q 2B H F D YB i d T h 2 R 5 ‰ h 2 T cUWp¢‘IGE‚VUII8†eI¢I3‚o8EW–IWkp¢˜8f¢UII¢såcUpX–IW™X‚¢UpI8†Ap¢UF r‚|6E‘IGpxW‹¢¢EG„W8†X‹ QQ†W98g•…IG™VUG…9I‚W¢9v…WQ‘q¢‚¢UpI8†‰ cU‚p85 Dq Hy H F D Y R F D ’ H bB DB Y 5  YB h H H Y H b T @ H F D D T F D H 7 RYY T H @ Hd H € ’ id T h 2 R 5 @ 5 qD R H F D D T h H nB 4 H  Y YB 2 5B DB h 2 5 4 5 2 HyB F @ t i d T h 2 R 5 ‰ @ 5 q 2B H F D D T 2 H bB ` YB 2 5B DB h 2 5 4 i d pGVAQIEQ6s‚X8XWEII¢QeIaXXpC‹8‚VUII8†cUpX‘pGVA68E¢X†8XWEII¢Qf¢T rxII8†j6EIƒX‚XapVW¢Q€6G‚Vƒ‘IG93j¢†x3«cUW˜83ŠcUpE›¢ˆ8uEI˜EW6sˆQpGE6f†dVg4 h 2 R 5 ‰ D Hy F 4B dB ã T t H Y T 4 d H D D Ty H F D 21 ’ H  iD @ 5 q D R 5r @ 5 q 2B q 5 d 5 4B h 5B d H  d H F DB H H ‰ 2 T 2 5B D T R P H 2 5B D 4 H b 2 5 4 d T H 2By H F D d 5q Y 2 5B DB h 2 5 4 i d T h 2 R 5 ‰ H F D t 8EGVIx68EG668o86‘VgIEXwpGC8•‘p8XWEII¢Q‚VUII8†wpGëp¥ ’   8EG66˜eW#IEW‚Q•6Cƒ 2 5B D 4 H ” 5 D ` 2B d d Hq H i eg j | â ¬ « i eg j t p © i h£7uX­&k†Ú‚hªke 2susë‚ugö à r | p ig ! p u ’ Y H 7 H F 4 Y Y R 5B d T b d 5q h H H  Y H Y T F  y T 4B d H 7 R 6Q9QIƒ‚uI8EW¢VŠ8•‰QQ†WwWVUIhVgX‚Q9Ifò 0 0 0.2 1 2 Central nd 4 Central th κ ∆x 4 Pade 1.5 th 2 2.2 0.5 1 ™— ¨— s– Û $’ ¥ H d R `B x%pu‚I8Ev¤ ¥ o— Ú Ì Ò Í É Ö Ó è × § Ë É Ú Í × Ë É Í Ì Ê × Ì Ç Ó Ì Í Ì Ñ Ñ Ò Ö Ð ¥Ð €Ô‘™‘fmӏV™efC•™CˆˆC󔅈fˆ”vԐ«o¦xÎ a* _ a 0.

H dB R P H ‚XIS6Wd q 5 Y @ 5 d 5 @D D Y Ty H F D d 5 D T d H  5 d 5B d H D 2B h H d H D 2 H 4 d H h d 5r F D d R 5q T q 5 H Y T 4 H F D 2 » VŠCcW‘v3vf¢ƒfIG€8GVWQ†8ˆ8X‚QWSEQW6GoQ6•6IW8£˜GfI8•a›¢WVgIWd31 ’ 2 5B D T 4 nB h 5 ‹¢XGlgU„I˜97 ¹ ç c ’ ¥ o“ pzæ · ·· ·· ’ ’’ ¡ ¡ ¡ ´µ ´´ ´´ ³´´ ¹ ß…î   ¦ a6 ä y ¦Œ t º² à ·· · ·· ·· ·· ¶ ¦   ¡   6 ´µ ´´ ¦ ¡ ¦ w6 ’’ ’ ¦ ´´ ´´ ¦ w6 ¡ ä Ý ·· ¶ •‘6 ¦a6 ¡ ¦ ³´´ ß…î   ¦ w6 ä » F DB @ d 5 D T d H  5 H 4 2 H d H gB h T Y H 4 R h 5 d  YB F G„CŠ8GVWQ†8uQI6W6h„I9‘6QIIxWp‘EIut ç 9 ’ ¥ 83“ pzæ ç4 € Å Ý 6 Å Ý ‚‡•)ٕšæ ß…î ¦ ä Å sÝ  “æ ç ( Û h H Y R H ‰ 2 T 4 Ty R 7 d 5q h d T @ ÷ 4 T ‰ d H h d 5r D Y d n ` 2B @ 5yy 5 hQWpw†¢g…XI…‚8•†‚¢Šoƒ¢I‘6IW¢3‚fWUaIXCcXE8•q HpGjQ85WVG6Hs¢…8X‚dQWSE–QI‚83˜p8Q6WeÆGEª)hQWp‰†–‚p…ƒp…W8•e6IQ‚QsEIÆW¢•@ F D t d D T d  5 d 5B H D 2B d H h d 5r h 2 5 4 H Y T F DB ™ ’ h H Y R H ‰ D Y R 7 Ty R 7 d 5q H 4 2 H d H gB h h d T r ÷ T 8ƒ4¢U‰ i€‘2…ˆlT6WIR†u¢Ip¢gIG5WxuQ‚XIS6WƒXICxQ8WVG6s¢a6IQ‚QsEI£˜Q‚QGo6Q4 ’ ä ê D h H Y H ‰ D 5 2 2 T 4 t 4 ü í Ý Y H dB R P H d F 4B F @ t d 5 D T d H  5 H 4 2 H d H gB hr h H d H D 2 H d H h 5r h 5 4 H Y F D Y R t 4 Å Ý D R 5 ‰ T 2 5B D T 7 d 5q 2B 5 2 F DB 6IWd8£˜I286Q‚eHIG…ISFuk’–5ü•…˜8sVr8EGV#W¢•pXeIrGEC@ Dp˜I…VT ß  5Ý  WV#XB ‘ H h 5 2 D r HD T 7 r — 5d  T Y R 7 SgWpI¢wD‚I…–Hd‹6UnE6QH†WYuXBk8EGEpI8Q‚¢UpI8††Im8f¢UII¢s‰ sIWp89IWaŠi ™ ’ h H B 4  Y 2 5B DB h 2 5 4 i d T h 2 R 5 ‰ 5 2 t i d T h 2 R 5 @ 5 qD R 5 H FD D ’ H D T d R 4 4 T d H h d 5 r F D d R 5q iyy T ‰ 5y ` YB 2 5B D T 7B — 5 d   T YB F ¢WVGI6Q¢C6IW¢3˜W‚I8•–„X¢U¢X8uEŽ8XWV#X˜gWIIVfEIut ’ ¹ à ç “ ’ ¥ o8“ pzæ ä y ¦Œ t º² ¹ ’’ ´´µ …î   ¦ ß ¡ ³´´ w6 ¦ ä •Ý ·· ·· ä Ý — ¸¶ !8‘6 ·· ¦ ’ ’ ’ ¦ a6 c ¡   6 ·· ¸¶ c ¡   ¦ c õ6 ¦ ´´µ c õ6 “ f6 ³´´ ß…î   ¦ ¦ w6 ä » F DB @ ¥ ’ ¥ ’ P  q 5 7 d 5 WEC…8“ p††j7¢§‚8•q H F D Y T F d 5 D T d H  5 H 4 2 H d H gB h ` 2B Dy R Y H d H F t ’ Hy ‰ T D d 5y i T t T ` 2B Y R h H bB d H h iyB Y T H YB F 4B F pGˆ¢Uˆ8GlGQ†8‘QI6W6h„IuIXWEI‚QW‘Iu”VXIVƒj8XxgmwaIX‚Id68„WQp…EEW¢QˆXdƒEIC@ ç $ ’ ¥ 83“ pzæ ç ß î „ Å6  ¢Æh£¢¬•Ý   †æ …8“ Ý Ý “ ø 4 Ý ¥ 6 ä 6 ¦ ä 4 sÝ  ˆæ ç ( Û ¦ ä ê D T d 5 D Td H  8‘€lC8GVWQ†85 d H h d 5 r h dB F D ` 2B @ 5yy 5q H F D H Y R 2 T 4 H @ t Y H 4 2 H d H gB h h H d H D 2 H 4 d H h d 5r F D d R 5q F DB @ t Hy  7 T — QI‚83˜‚XIWaIECcXX¢•IGWp‰Vgu•I6Q6IQ‚QsEI†QW6GoQ6ŠQpW8£˜GfI8•ŽG„Cp¢EI9¢p‡H d8U…¢9QpƒW™¢XW6GoX‘pGj¢lUGvVUSQ§XXC6¢WIQ6¢vVU8X¢…IGeIV˜¢9QIƒ‚™8EWQWoXB 5 ¤ ’ H 7 H F 4 Y d 5B d H D 2B H F D q 5 D T F D y T R P H yyB @ i 4 T d R 4 4 T y T ‰ 5y ` H F D h 2 T t H 7 H F 4 Y d 5B d H D 2 H F D q 5 D T F D 2 T F D € } è ‡  € } ˆ € ‡ } „ ‡ YB F 4B F @ i 4 T d R 4 4 T q 5 d H h d 5 2 T H b T F 2 T 4 t } { } ‚  pG¢9VIG©VUGw‡•lXe‡x‚lehlX©ƒXIC¬‡¢GI6Q¢¢–QI‚8«V8gU©¢g€‡d‡Uƒcž ‡¢o†xl‡5¢WŸ–‡d˜Ud•¢CcIS˜68GlGQ†8#¢ƒI˜v8‡¢GpQQV˜¢•6IW¢fWVƒ‚I8WpI¢pGD €  ˆ „ ¢ ‡ À    y € } { ¢ „ T Y T 2 @ 5 2 ÷ t d 5 D T d H  5 2 T F 4 R ” ’ i 4 T d R 4 4 T q 5 d H h d 5 H D TBd  5 d   T H F ` 2B b T F HyB F @ t ¦ 6 ê 5 D iy 2 5 Y h 2 H D — H F 4B F @ IExcI‘XXpC§8ÿë#GEI8aII6Gx‡ƒXIC–¦ ™l)Q‚XIS6WuXu8WVG6s8woQ‚QsEIT ä ê D T h H dB R P H d YB d 5 D T d H  5 D 2 H d H gB h  € í Ý q 5 H Ry T b T Y H dB R P H d ¦ ä ê D T 2 5B D T 7B — 5 d   H 4 2 H € ’ 2B T 7 5 h H F D H hB Y D R 5 YB F 4B F @ 6IQ‘‹XVd8pwpGIEWG˜8uXƒXpCIt Axm¢IE¢V…f6WEISQWuåw€Ve8EGV#X˜gWpI¢T H F D 2 H F D t — $’ ¥ ’ P  2 2 H bB ` d 5 D T d H  5 d 5B d H D 2B d H h d 5 r F D d R 5q H F D H Y R H @ qB t d H b H @ 5 pGªQIWCpV%˜uëˆgEB Q¢E8Ž¢GVWQ†8d8EW6GoXŽ6IW¢3˜W‚I8•pGWpAvE™QQ¢6vcf€ Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u‡§™†fjˆˆ«”w€‘ˆ””Ԑz€mQ‘Ôv«Sψm™‘”Ç — 8— .

ç — 9’ ¥ SVdpzæ — W ß  4 ü 5Å W ß  V iä V Ý  Ý  YB D T F D t ¦ X‘VUWp8ðø ‘ ä ê D T h H nB 4 H  Y YB AV¬6UE6Q†WwX©ß  ÚÝ  VUWd#pWWVe•˜‹8EG„II86rf¢UII¢s)p¢dpQfw8U¤ r D T F D H 7 R Y Y T H @ t 2 5B DB h 2 5 4 i d T h 2 R 5 ‰ 2 2 T 7 R H ò T d 5 ’ ’ ¥ ’ P  2 ˜9   pëëj‰XB 2 H bB ` d 5 D T d H  5 H 4 2 H d H gB h H F D 5 D ` 2B h T Hy t Y HB d T h 2 R 5 ‰ d T H 2 h H dB R P H d H d T Y 2 5B D T 4 nB h 5 7 5 68E¢§¢GVWQ†8•6IQ‚QsEI€IG‘GuIEI¢gEV6QEW¢UpI8†m¢QI…6WXpx6W€W¢‹I8XWVgIEIx#fI2 ç ¥ 9’ ¥ o¢dpzæ ç4 € í Ý ø í fAxÆv˜Ý   85Wxšæ 6 4 üí Ý  ß…î ¦ ä í sÝ ¥ ˆæ ç  ( d 5 D T d H  5 d 5B d H D 2B h H d H D 2 H 4 d H h d 5r h 2 5 4 H 8GlGQ†8Š¢XW6GoX†Q‚QGo6Q‘QI‚83˜p8Q6WY HpGG„ª¢8c†¢AQ‚WI6WXpe8XWVUSQk8EG668o86…IG‘8•‰‚¢UpI8†åcU˜X…IW™¢™QIEXy F D F DB ™ ’ H b 5 ‰ T h H Y Y R 4 YB h 2 5B D T R P H 2 5B D 4 H b 2 5 4 H F D d 5q i d T h 2 R 5 ‰ @ 5 q 2B H F D Y T Y H 2B H 7 T Y H F D ` 2 5y T Y h H H 4 5 d  2 5B DB h 2 5 4 i d T h 2 R 5 ‰ D Hy F 4B dB ã T q 5 D 2 H 7 D T H d D H F t ’ Y 2 5B DB h 2 5 9¢WuIGI¢X¢€I6QQxWp…8XWEIp8Q‘f¢UIp8sm6EIƒX‚Xuh¢joQ9GVQWWCpuvQp8XWEII¢Q4 i d T h 2 R 5 ‰ 2 2 T 7 R H ò h 2 T D Hy F 4B dB ã d H hB Y 2 5 4 D Y R 7 H @ 2 5B D T R P H 2 5B Y R gB h H F D ` 2B by 5 Y 2 ‚¢III8†I¢#IQ‘I¢•6XpƒXWEuQpXWp8QCfI…vŽ8XWVUSQŽ¢XWph„IIW…IXxE¢We31 i eg j | â ¬ « i eg } â ng l p u h£7uX­&h£¡uÏo2Ùugö ‚— ¨— o– ’ 9 ’ ¥ ’ P  5 D y T 4B D 2 H hB YB F 4B F ˜“ p†ëjëGh¢QXWS6IXuE†ƒEIC@ ç   dpzæ 9’ ¥ ä ä ç4 € Å Ý 6 4 ü Å Ý ¡‚‡•)855•æ ß‘î   ¦ ä ç4 € Å Ý 6 4 € Å Ý 6 Å Ô‡•Å7‡•)Ù•Ý   æ ß …î çf4€‡Å•ÝÅ6ÙÅ5Ý•æ ¦ ß‘î   ¦ Å sÝ  “æ ç ( Y H bB Q¢E8` ‘ H h 5 2 D T h HBy   T d 5 D T d H  5 H 4 2 H d H gB h h H d H D 2 H 4 d H h d 5 r h 2 5 4 H Y H F D 5 D 2B YB F D ` 2B D R DB D Y ‰ R p˜IVeQEXIp¢8WVG6s¢‘QI6WQsEp6WQWoQQ6IW¢3˜I¢QQ‚‘IG#GoXEIG#IXWpG„G‚pI˜” ç ¦ 9’ ¥ s˜dpzæ 4 € Å Ý 6 Å ‡•Å2•Ý   j5•Ý ä 4 ü Å Û 4 ü Å Ý 5 D 2 5B D T 7B — 5 d   T d H h d 5 r D Y d n ` 2B @ 5yy 5q H F D Y H bB ` F 4B F –5•9Wd8EGV#X˜gWIp¢‘QI‚83f‚WIwIXCcEX8•IWC68„8†ƒEIC@ ç ¡ dpzæ 9’ ¥ ¡ ä 4 € Å Ý ø Å ”‡•Æð•Ý   85•ÿ”55Ôç  6 4 ü Å Ý ä 4 ü Å Ý € ® €ø 4 € ®   pfæ 6 ¦ 2B T D ‰ 5 H X¢Gp8w•@   d~GEC˜¢Xp#¢p–Š8U€8ºë°X8EGD ä  F DB @ t Hy  7 T — H d 5 ¤ ’ ¦ 6  YB 2 5B r T 7B — 5 d   T H F D q 5 d H h d 5 H F D h 2 T 4 ü í Ý ä í Ý ® i ‰ h H 2 n H h d 5 D T d H  5 DqB F Y H F D Y ® H d H F l#X˜gWIIVIG›V•QpW8pGdIV€5WxðŠ˜w¯oQpU6I€8WVG6s8•|EI‚IW€EB „W6IC@ ç h ’ ¥ o“ pzæ ¡ ä 4 ü Å ”5•Ý ç Ü € 4 ® 6 ¦ Vpfæ Û hB d ` h H 4 T  YB R P H 2 T 2 5 d H h d 5 i d T d DB ‰ d T 5 D T D T hXW¢Q6¢U‚XISQ‰¢‰8CQpW8d‚¢WGEpW¢9G…GVUh d8X‚QGoEvIWó8W•uQWVƒ¢s¢WGx‡vs‘Gh5•§CcEX¢Cƒp…W8•ŠpXCcXE8•vIu˜‹¢XGlƒ8†¢GWx‡H 5B d H D 2B H F D 7 5 dq h H D Ty 5  T d D — H H ‰ 5 D 4 ü Å Ý Y @ 5yy T Ty R 7 d 5q ` 2B @ 5yy 5q H F t ’ 2 5B D Ty 5  T d D — H 4 T  Y q 5 Y 7 d H D 2B YB Y H 7 H F 4 Y i d T h 2 R 5 ‰ q 5 D 2 H 7  5y H b H h H F D 5 D F 4 T 5 d   T d H F D 5 2 6¢UWv¢9W6GX™XwQ9QIƒ‚‚VUII8†v¢wS6#p8X686p‘IGeGeƒ¢8WIIVa6IGVI™i $ — Ú Ì Ò Í É Ö Ó è × § Ë É Ú Í × Ë É Í Ì Ê × Ì Ç Ó Ì Í Ì Ñ Ñ Ò Ö Ð ¥Ð €Ô‘™‘fmӏV™efC•™CˆˆC󔅈fˆ”vԐ«o¦xÎ .

ç c 9’ ¥ ovdpzæ ç „ …zd“ø ß îæ ’ ʤ45üÅ W ß  V ß ð…î   Ý  ø 4 € Å Ý 6 Å Ý ‹‡•)ٕ8— È ¥ ¦ ä 4 ü Å j5•Ý Y H bB ` 4 ü Å Ý d 5q ` 2B by 5 Q8„8v5•u8•…IXx„8˜” ç 9 9’ ¥ 8dpzæ ç  …zŽ~›‚5•¢©ª•8Å8‡•šæ ß îæ ’ ø ç4 ü Å Ý ¥ ø Å Ý — 6 4 € Å Ý ß…î   ¦ ä 45üÅ W ß  V Ý  Û ¦ 8ªø ‡IxICV6XEII¢8EGV#X˜gWpI¢aQI6WQsEp3˜‚¢Šoƒ¢If6IW8£˜I86Q‚–‘6IX‚I8• ‘ H h 5 2 D T h HBy   T 2 5B D T 7B — 5 d   T H 4 2 H d H gB hr h d T @ ÷ 4 T ‰ d H h d 5r h 2 5 4 H Y T d H hB Y 2 5 ’pgIE8EGVX8sVGGx–#6¢UW‘pGIX‚IŽ¢dp§†jTE8GVWQ†8–IGeE¢ƒp¢eWXVr¢Q–r™ T H hB 2 5B D Ty 5  T d D — H H 4 T  Y H F D ` 2B Y R “ 9’ ¥ ’ P  2B d 5 D T d H  5 H F D 2B T D ‰ 5 5 Yy T 2 T 4 H ’ H 7 H F 4 Y d 5B d H D 2B H F D 2 T F D d H @ 5y d H h d 5 H 2 5 H ‰ 2 T 4 F 4B F ¢9QIƒ‚Š8EWQWoXuIG¢UWuQvsEuQI‚8uI8u††¢Q†ƒEIC@ iy Y R 5B b H d  h H ‰B d 4 Y H h Y H 7 H F 4 Y i d T h 2 R 5 ‰ y T 4B d H 7 R 2 H F D F DB @ Y D Y T d D 2 5 4 YB F t ’ 2 5B D Ty R E‚I8Xx‡WI†Q†XW6W6If6#6IƒWŽ‚¢UpI8†‹¢gEW6#ISuIWWECCGf¢GWS¢QCXIum‹8EGVXI…7 rf8•uGVWIQ6¢fQI‚83˜p8Q6W9EE¢U8E89‰E¢ƒp¢#W†f¢G‚QQ6ICXuEIGph8XWEII¢Q‚VUII8†‰ d 5q H D T d R 4 4 T d H h d 5r h 2 5 4 H Y iyy T ‰ 5y ` T 2B T D ‰ 5 5 D i d T Y Y H 4 H 2 YB YB F D t 2 5B DB h 2 5 4 i d T h 2 R 5 2 2 T 7 R H ò T 5 D 2 5B D T 7B — 5 d   T y T 4B d H 7 R 2 T q 5 H Y T 4 H F D 21 ’ H D T d R 4 4 T d H h d 5r h 2 5 4 H Y Y IVdI6fGk8XWVdE˜c‚II¢v¢QXW6#px†‘¢ŽWVgŽIW)3¢WVGpQQV‘QI‚83xI8Q6WaXB d 5 D T d H  5 YB F D D T F D H 4B D 5 ò ’ ’ ¥ ’ P  2B 2 H bB ` d 5 D T d H  5 H 4 2 H d H gB h H F D Y H 4 R h 5 d  F 4B F 8GVWQ†8XpGVIGw6XGVfaph   p›ëjTXkQ¢E8¢GVWQ†8‘QI6W6h„I‘IG6QIIxWpAƒEIC@ ç “ 9’ ¥ o¢dpzæ 45üÅ W ß  V Ý  ߑ¢¥ î   ø ç Å ›…•Ý   ‚‡•Ý   æ 6 4 € Å  ߅8¥ î ¦ ä Å sÝ ¥ ˆæ ç  ( Û d 5 D T d H  5 ` 2B @ 5yy 5q H F D 2B T D ‰ 5 H @ t º h 2 T t Œ t ‹ d 5q ` 2B by 5 Q8WVG6s¢…IXCcEX8•IWŽX¢Gp8w•ps˜I¢ëc!Up8•…IXx„8˜” ’ 2 5B DB h 2 5 4 i d T h 2 R 5 ‰ 2 2 T 7 R H ò d 5q Hy ‰ T D d 5y i T t ’ ’ ¥ Hy ‰ T ‹¢XG„II86e‚VUII8†‰I¢#IQ‘C8•XIVƒŠ8E˜gmjI— puXIVvt 4 4  Y  sfokº ç ¦æ Y º 4 Y ç ¦ ræ Y 4 hsG|S{‹ í y xuu t w Y ß …î u Ý Y sÔxkº ç ¦æ Y 4 „ … …ç ¦ræ Y 4 Y sW|x‡‹ í y ü w ü uu t Y … …î ß Ý Y „ sW|x½‹ ç ¦ræ Y 4 hsÔxkº Y ç ¦æ Y 4  ç ¦ ræ Y 4 Y  sƒƒx‡‹ ¦ í y ú w ú uu t Y  …î ß ß ‘î Œ ‹ 45Wí y ¥u t ÉaY ‘î ü w Y º ß í Ý Y xckŒ 4 €í Ý Y ¥˜Q{‹ í y ú w ú ut Y  ‘î ß u í y û w û uu t Y „ …î ß í xÝ ù ×ä 45ü‚í W ß  V Ý  º ø ç í Ý Œ ø 4 € í Ý ‹ ›‡˜ƒ~›¥xSˆæ  ß…î ¦ ø í W  ß  V Ý   ’ ’ ¥ Hy ‰ T t 2B 2 @ 5 F Y Y I— ˜aEI¢meX†CspWC¢T t Hy ‰ T D d 5y i T t T ` 2B Y R h H 2B 7 d H D H h H ‰ iyB Y T H 2 T 4 F 4B F @ Y D 2 T D Y 2 5 4 H d T º h 2 T t Œ t ‹ H d H F ¢EI¢ƒŠ¢Xxgv–9IX‚IQIE#‚QG‡Ius#EEW¢Q†Vg‰ƒXICuGo¢G‚I¢QuW¢™fIVësë‡W6IC@ ç $ 9’ ¥ 8dpzæ ß …î 45üÅ W ß  V Ý  º ø ç4 € Å Ý Œ ø Å Ý ‹ ›f‡5ƒ~ÿ•o“æ  ß…î ¦ ä Å hÝ ¥ ˆæ ç  ( Û 7 d 5q ` 2B @ 5yy 5q H F D 2B YB F 4B F ‰W8•wpXCcXE8•IGeXuE†ƒEIC@ ‘ H h 5 2 D T d 5 D T d H  5 2 T 2 `B Y H h H @ Y R F {p˜IClC8GVWQ†8‰¢‰8X‚QIw•CISut “ o— Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u‡§™†fjˆˆ«”w€‘ˆ””Ԑz€mQ‘Ôv«Sψm™‘”Ç .

’ 7 d H D d 5 d d H ` 2B h T Hy H F D h 2B †‚QWŠ8W‚Q‘IEI¢gEuIGeIEv¤ ç ‚xg“h5WxÝ üí Ý þ ø 4 üí  " ø í Ý º ø 4 € í Ý Œ vx‡~›¥x–~ø ß  Axo“æ „ …î €í Ý ‹ ¦ ä íç   ( uš†Ý ¥¥ ˆæ 7 d 5q H F D 2B H bB D T bB d H h h dB F D T 5 D 2 5B D T 7B — 5 d   T H 4 2 H d H gB h r H DB 2 n T H bB d H 㠒 W8•IWeX8„GVV„WQp‰‚XIW…#WŽ8XWV#X˜gWIIVuQI6WQsEp3‡WEII–u¢EW6‘˜$ ’ ¡ ä " F DB G„C@   8EG‚6IS‘Vg†QCfI— 2 5B D Y H R P D T H  H ƒ ’ ’ $ ä ê D T 2 H bB ` YB Ý h 2 ˜2‘€l68E¢uXIVT ¡ ‘€V†Q8„8WVkß  5Ý  I¢Ý ä ê D T 2 H bB ` H d T r h 2 T FD 5 ‰ 2 H F W¢sk6IC@   8EG‚6ISXk68E‚QI‘9QpƒW…IWdIEWISQ9W8•8W668†¢XG„II864 2 5B D Y H R P 2B h H bB d H h H 7 H F 4 Y H F D ` 2B Y R i ‰ h H 7 d 5q d 5 D 4 H b 2 5B DB h 2 5 r i d T h 2 R 5 ‰ H F D h 2 T Y H 4B d D T 7 — ¸ — H F D D R 5 H DB d @ t F Y H 7 D 2B 5  ç d 5B d H D 2Bæ — T ` 2B Y H ’ of¢UII¢sIWŽI¢•QQEWWVd‘ˆsIWŠp8™GE‚Cp‹‚Q9ŠoX8†€G8X‚QWSEš…–pXW‘fp¥ ’ 7 d H D d 5 d d H ` 2B h T Hy H F D h 2B †‚QWŠ8W‚Q‘IEI¢gEuIGeIEv¤ ç4 üí f5‚˜Ý " ø í Ý º ø 4 € í Ý Œ ”x‡~‹Axƒ“æ ß…î ¦ ä íç (æ ø 4 € íç (æ ušhÝ  “ÿhAšhÝ  “†‹ 7 d 5q H F D 2B H bB D T bB d H h D Y d n T 5 D 2 5B D T 7B — 5 d   T H 4 2 H d H gB hr H DB 2 n T H bB d H 㠂8•IGeX¢EWVVEW6I‘‚‚U…#WŽ8XWV#X˜gWIIVuQI6WQsEp3‡WEII–u¢EW6‘’   ’ 7 d H D d 5 d d H ` 2B h T Hy H F D h 2B †‚QWŠ8W‚Q‘IEI¢gEuIGeIEv¤ ç4 üí f5WxÝ " ø í Ý º ø 4 € í Ý Œ vx‡zh¥x–~ø ß  Axo“æ …î €í Ý ‹ ¦ ä íç ( ušhÝ  “æ 7d 5 ‚8•q H F D 2B H bB D T bB d H h D Y d n T 5 D 2 5B D T 7B — 5 d   T H 4 2 H d H gB hr H DB 2 n d H h d 5 r h dB F D T H bB d H 㠒 pGXa8„GVVE‚QIufWUdŽG†¢XGldEpg‚II¢‘6IQ‚QsEI£‡WEIU‘QI‚83xWXpGd‘¢EW6‘8¦ ² R ± ˆ ¶ à Q·YjÀ µ ´® G­ ’ “ 9’ ¥ ’ P  5 D y T 4B D 2 H hB YB F 4B F pV%p†ëjëGh¢QXWS6IXuE†ƒEIC@ ç ¡ Õpzæ c’ ¥ 45üÅ WUß  V Ý  ߅8¥ î   ø ç Å ›…5Ý   8‡5Ý   æ 6 4 € Å  ߅8¥ î ¦ ä ä ä ʤ45üÅ W ß  V ß ð…î   Ý  ç h 9’ ¥ ovdpzæ ÈD …8¥ ß î ø‹4€‡Å•ÝÅ62ŕÝ8—Æøðŕݢ“Å684€‡Å•¢¥ Ý ¦  ß…î ¦ ç4 € Å Ý ø Å ‚‡•Æª•Ý   6 4 ü Å Ý 85•šæ Å sÝ ¥ “æ ç  ( Y H bB Q¢E8` ½I˜pCV‰QXEII¢u¢EWVVEW6Ih ‘ H h 5 2 D T h HBy   T H bB D T bB d H h 2 5 4 H Y T d 5q d 5 D T d H  5 H 4 2 H d H gB h h H d H D 2 H 4 d H h d 5 r h 2 5 4 H Y H F D 5 D 2B YB F D ` 2B D R DB D Y ‰ R I86Q‚ºe8•†¢GVWQ†8ÅQI6W6h„Iå6W6GoQ6A6IW¢3˜I¢QQ‚)IWºGoXEIG©IXWpG„G‚pI˜” 9 — Ú Ù Ì ’ § × Í Ê Ð ±Ð €˜ª³ˆ«c²xÎ .

’ H 7 H F 4 Y F 4 T H d 5q d 5 d d H 2 5B D T 4 2 R d D H F D 2B Y 7 d H V#6IƒW‰ƒ¢gu¢•C¢WW68XWVgIpWGIWeXu9WQWD l d 5q ß î – ’ Y b ß – q 5 Y D d T  id T 2B ` T 7 `IXIVgXev3dIW#¢EW6j l ‘U– ¹¡ ¢•Æ…U#6xƅî h f¢wG‚VUf¢UX¢¢dEB 2B h T Hy 5 @D H F D H bB d H 㠒 ß î h 2 T y T H d H F D D 5y Ü ’ Y H 7 H F 4 Y H Y H F D q 5 F 4 T H d 5q d H ‰ 7 R 2 H b T @ h H nB h 5 7 H F D h 2B IVu¢g‚eIGw¢Xj¬Q6#6IƒW‰W6IGC¢Æƒ¢Q…8•–Q†…ISQ¢c•)QIEIx#epG)IEv¤  €í Ý h ø 4 €í Ý c ¦ 6 í Ý ¦ ¦ ç ß î “æ r s‘¢zIç „ AxÝ   6 Ax¢~hAx¢a«hxg8Ôæ €í ä íç ( ušhÝ  “æ  €í Ý ø 4 €í Ý — 6 í Ý ¥æ ä íç ( ç ß s‘î   Iç AxÆh¥˜¢)hx¢šÏuz†Ý  “æ æ r ß î r ç4 € í Ý 6 í Ýæ ä íç ( …¢f¥˜ÅUxšÏuš†Ý  ˆæ Û iy H bB D 4 H  Y H d t d H h d 5 r h dB F D h 2 †„Q8„G‡Q†WQ‚p6QI‚83˜‚XIW‰p¢T t r h 2 5 4 H Y t r D Y d n H d T F 4B F @ t Y H 7 H F 4 Y ` 2B 4 2 H d H gB h h H hB Yr H 2 5 ` 2B @ 5yy 5q H F D d H hB Y 2 5  G˜I86Q‚˜G‚‚‚Uf‚¢eƒEICpQ6#6IƒWaIX6IQ‚QsEI#QIEW£‡I8uIECcXX¢•fIWvQpXWp8•f’ ¡ ¦ ’ H bB D T bB d H h D Y d n T 5 D Y 2 5B D T 7B — 5 d   T h H d H ¢¢EGl¢„W6I™fWU…9GvI8XWV#X˜gWIIV6WQWD r 2 H 4 D 4 T  7 5 4 h 2 T D 4 T  7 5 4 2 5 2 d H hd 5r FD BY d 5q D 2 H 4d H  ¦ xQQC‡¢U98QIVC6¢I#86I8Iu6IW¢3˜W˜— WŠ¢•‘o6QW6su„’ ¡ ¢UWŠW6Xu8‚WQ‰Q6s‚Y 2 T F D Y Y Hy d 5 d d H h H H  H Y T F  T H b HB F 4 T 5 D D 2 H 7 H dB R P H d ç     F D ` 2 Hy H b T @r d H r Y D 2B 5 r hB d ` H F D h 2B ¤ ’ ‚¢UIr86EIƒ¢ÅWŽoQ9Q‚XISQ‚…Ÿ0™æ W¢I6XQ¢gŠŸfQ†3ÔGoX¢s£˜X‚8IW©IEv‘ph l d 5q ß î – ’ Y ’ j l …U– m¡ 8•…UƒQxb ß î ß …î h f¢Xjˆp¢XGfQISeE‰68„WQp•8WVG6s8•QI‚83˜W‚I¢•Š6¢I#86IGŽp¢I–8dp¥ – D 5y Ü ’ “ 2 5B D Y H R P 2B h H bB d H h d 5 D T d H  5 d H h d 5 r F D d R 5 q D 4 T  7 5 4 H F D h 2 T t ç $ $’ ’ëj“†8WVG6s¢96IW¢3˜W‚I8•96VU#¢QI8p†IWvV8EWVVE‚QIºp8Q6Wk#¢•–¢GVWQ†85 P æ d 5 D T d H  5 d H h d 5r F D d R 5q D 4 T  7 5 4 2 5 2 H F D t H bB D T bB d H h h 2 5 4 H Y T d 5q d 5 D T d H  H 4 2 H d H gB h h H d H D 2 H 4 d H h d 5r h 2 5 4 H Y H F D d 5q d H ‰ 7 R 2 H b T @ h H nB h 5 7 H F D h 2B ¤ ’ 2 5B 6IQ‚QsEI¬Q‚QWS6Q…6IW¢3˜I¢QQ‚dIWw¢•–6swISQ8g•A6UEp˜9ŽIWIEv‰‹8XWD r T 7B — 5 d   T H bB D T bB d H h h 2 5 4 H Y T d 5 l#X˜gWIIV–8EWVVE‚QIAI86QWŽu¢•q h ‡Q†…ISQ¢gŠ6UEIx9‘IGeIEIU6pwpxWD – d H ‰ 7 R 2 H b T @ h H nB h 5 7 H F D ` 2B 2 n H h Y R F  – 6 ä í ‘• þ  h “ˆušç  e í —• þ ¥ ˆæ  ( Y H bB ` H bB D T bB d H h h 2 5 4 H Y H F D d 5q d 5 D T d H  5 H 4 2 H d H gB h T q 5 2 5B D T 4By   68E¢8EWVVE‚QI†I8Q6WuIWŠ8•Š8GlGQ†8uQpQW6h„I–”¢d8XWVgEXIIfi Y H bB Q¢E8`  ‘•  ß  – • þ $6 ä ‘• þ    ‘   þ 2 5B D 4 2 Rq H F D 5 D d 5 D T d H  5 H bB D T bB d H hr h 2 5 4 H Y H F D q 5 2 5B D T 4By   i ’ ˜8XW6Ip•upG‘GŠ8WVG6s¢w¢EWVVEW6I3xI8Q6WuIWm¢d8XWVgEXIIfCpc ’ ç — $’ ¥ ’ P  æ d 5 D T d H  5 h H d H D 2 H 4 d H h d 5 r F D d R 5 q H F –SVdpë눈w8GlGQ†86W6GoQ6uQI‚83˜W‚I¢•upGD ¢W•eQIE¢ƒ˜8‘VUGeWECu‚VU 7 5 dq h H 2B T D ‰ 5 D T F D F DB @ D d T l d 5q ß î – ’ Y b ß – q 5 Y D d T  i d T 2B ` T 7B h 2 T y T H y T Hd H F D Hd T  7 5  ’ ¢g‚IG‚¢U98•€j l …U– m¡ 8•…U£6˜†…î h #¢•G‚VU9‚¢UE8¢#XeIV›¢g‚d ß î H F D D 5y Ü ’ ¦ 2 5B D Y H R P 2B h H bB d H h d 5 D T d H  5 H F D d 5q d H ‰ 7 R 2 H b T @ h H nB h 5 7 H F D h 2B ¤ ’ pG•VXjˆ8e8EG‚6ISdE‰68„WQp•8WVG6s8™IGˆ8•uQ†…ISQ¢gŠd6UEp˜9IW#IEv‘˜9 ’ 7 d H D d 5 d d H ` 2B h T Hy H F D h 2B †‚QWŠ8W‚Q‘IEI¢gEuIGeIEv¤ ç4 üí Ý º ø í Ý Œ ø 4 € í Ý ‹ ‚5Wx–“v˜ƒ~›¥xSˆæ  ß…î ¦ ä 4 üíç  (æ þ ø íç  (æ ø 4 € íç  ( j5Wz†Ý ¥ “S~jz†Ý ¥ “ÿ‹Az†Ý ¥ “æ " 7 d 5q H F D 2B H bB ‚8•IGŽEu8EWD r T b B d H h h 2 5 4 H Y T 5 D 2 5 B D T 7 B — 5 d   T H 4 2 H d H g B h r H D B 2 n ç `H h T Ü d 5 æ D 4 T  7 5 4 T H b B d H 㠒 lVEW6Ip8Q6W‘9Gd8EGV#X˜gWpI¢uQpQW6h„I3–G„IUŠc¨¢mC8ša‡¢U98Q‘u¢EW6‘p“ Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u‡§™†fjˆˆ«”w€‘ˆ””Ԑz€mQ‘Ôv«Sψm™‘”Ç c 3— .

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

the spatial and temporal discretizations are not separable. This is a simple and natural formulation when the ODE's are linear. we shall separate the space di erence approximations from the time di erencing. 4. that the spatial and temporal discretizations are separable. Another possibility is to replace the value of ujn in the right hand side of Eq. 4. there are subtle points to be made about using these methods to nd a numerical solution to the di usion equation. 2 @ j x ( 1) 2 ( ) +1 ( +1) 2 0 ( +1) + ujn.3 is referred to as the DuFort-Frankel method. In this case.1.2 and 4. In this approach. In these simple cases. 4. A n 5 + uj . As we shall see later on. 1 to the level n + 1.1 The Model ODE's First let us consider the model PDE's for di usion and biconvection described in Chapter 2.3 which we cannot comment on until we develop a framework for such investigations. It is a full discretization of the PDE. namely (ujn + ujn. and convergence of Eqs.3) which can be solved for u n and time advanced at the point j . 4. There are a number of issues concerning the accuracy. Equation 4. 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. stability. Note. This results in the formula ( ) ( +1) ( 1) ujn ( +1) un = ujn.1 Reduction of PDE's to ODE's 4. 2 ( 1) ( ) 1 1 3 (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 .52 CHAPTER 4. and no semi-discrete form exists. Thus.2 is sometimes called Richardson's method of overlapping steps and Eq. 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. )=2. 4.2 with the time average of u at that point. this method has an intermediate semi-discrete form and can be analyzed by the methods discussed in the next few chapters. . we can approximate the space derivatives with di erence operators and express the resulting ODE's with a matrix formulation. however. For the time being. + 2h 4ujn . THE SEMI-DISCRETE APPROACH Clearly Eq.

They are linear with coe cient matrices which are independent of x and t. 2 Model ODE for Biconvection The term biconvection was introduced in Section 2.2 The Generic Matrix Form The generic matrix form of a semi-discrete approximation is expressed by the equation d~ = A~ . 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. It is used for the scalar convection model when the boundary conditions are periodic. a B (.1 0 1)~ u u (4. In this case.4. The vector ~ (t) is usually determined f by the boundary conditions and possibly source terms.3. In such cases the equations are nonlinear.4 and 4. . 4. even the Euler and Navier-Stokes equations can be expressed in the form of Eq. ~ (t) u u f dt (4. that is.2 1)~ + (bc) u ~ (4. the 3-point central-di erencing approximation produces the ODE model given by d~ = . 4. 4. REDUCTION OF PDE'S TO ODE'S 53 Model ODE for Di usion For example. Eqs.6.4) dt x ~ with Dirichlet boundary conditions folded into the (bc) vector.1.5) dt 2 x p where the boundary condition vector is absent because the ow is periodic. the elements of A depend on the solution ~ and u are usually derived by nding the Jacobian of a ux vector. In general.1. 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 .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.5 are the model ODE's for di usion and biconvection of a scalar in one dimension. Although the equations are nonlinear.

If A does depend explicitly on t.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) . m. 4. The ODE's represented by Eq.54 4. the general solution to Eq. 4. f . we will make use of exact solutions of coupled systems of ODE's. the system of ODE's must be integrated using a time-marching method. This will lead us to a representative scalar equation for use in analyzing time-marching methods.6 can be written in terms of the eigenvalues and eigenvectors of A. . I] = 0 We form the right-hand eigenvector matrix of a complete system by lling its columns with the eigenvectors ~ m: x h i X = ~ ~ : : : ~M x x x The inverse is the left-hand eigenvector matrix.8) where is a diagonal matrix whose elements are the eigenvalues of A. is independent of u. In order to analyze time-marching methods. 4. 4.. As we shall soon see. the general solution cannot be written whereas. if A does not depend explicitly on t. mI]~ m = 0 x (4. have the property that A~ m = m~ m or x x A . if @F=@u = A where A is independent of u). These ideas are developed in the following sections. A. This holds regardless of whether or ~ not the forcing function. the exact solution of Eq. which exist under certain conditions.6 in which the coe cient matrix. 4. 4.1 in time. depends explicitly on t. AX = 1 (4.6 can be written. THE SEMI-DISCRETE APPROACH In order to advance Eq. and together they have the property that 1 2 X . As we have already pointed out.1 are said to be linear if F is linearly dependent on u (i.2 Exact Solutions of Linear ODE's CHAPTER 4. An eigenvector.7) The eigenvalues are the roots of the equation det A .e. when the ODE's are linear they can be expressed in a matrix notation as Eq.2. and its x corresponding eigenvalue. ~ m .

This theme will be developed as we proceed. and an eigenvalue with multiplicity n within a Jordan block is said to be a defective eigenvalue. i. be transformed to a diagonal set of blocks.. 4. and is homogeneous if the forcing function is zero. . all of which can be complex numbers. rst-order equation du = u + ae t (4. Defective systems play a role in numerical stability analysis..9 displays many of the fundamental properties and issues involved in the construction and study of most popular time-marching methods.e.2 Single ODE's of First. a. i. 4 5 m n The matrix J is said to be in Jordan canonical form.9) dt where . if = 0. the numerical analysis of Eq. . if a = 0.. and has a general solution because does not depend on t.... there exists some S which transforms any matrix A such that S .and Second-Order First-Order Equations The simplest nonhomogeneous ODE of interest is given by the single.. Although it is quite simple. however. The equation is linear because does not depend on u. some of which may be scalars (see Appendix A). 6 7 J =6 7 6 7 Jm 4 5 . and 2 3 m 1 6 7 1 . m n =6 Jm 6 7 . In general. and it is said to be defective. 1 7 . 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. 6 7 .. It has a steady-state solution if the right-hand side is independent of t. . 1 1 2 ( ) 4.. AS = J where 2J 3 6 J 7 6 7 6 7 . It can..2.e.4. and are scalars.2..

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). t u(t) = u(0)e t + a e . In terms of the initial value of u.10) u(t) = u(0) + at]e t As we shall soon see. since the roots of these polynomials determine the solutions of the equations. 4.11) dt dt where a and a are complex constants.11. we often label these roots in order of increasing modulus as . Using this limiting device. u(t) = c e 1 1 t+ ae t .9 is. . m . Characteristic polynomials are fundamental to the analysis of both ODE's and O E's. where c is a constant determined by the initial conditions. for 6= . For ODE's. 4. 4. . solving. e The interesting question can arise: What happens to the solution of Eq. this solution is required for the analysis of defective systems.56 CHAPTER 4. we nd that the solution to du = u + ae t dt is given by (4. and then taking the limit as ! 0. it can be written t. Second-Order Equations The homogeneous form of a second-order equation is given by d u + a du + a u = 0 (4.9 when = ? This is easily found by setting = + . 1 2 . THE SEMI-DISCRETE APPROACH The exact solution of Eq. 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.

In our simple example.4. these are indeed solutions to Eq.15) By substitution.11. homogeneous equations is given by u0 = a u + a u u0 = a u + a u (4. 4. by setting u = u0 u =u we nd Eq. and .2.14) which can be written " # 0 a a ~ = A~ ~ = u u ]T u u u A = (aij ) = a a Consider the possibility that a solution is represented by u = c x e 1t + c x e 2t u = c x e 1t + c x e 2t (4. M .14. One nds the result c e 1t( + a + a ) + c e 2t( + a + a ) which is identically zero for all c . rst-order.3 Coupled First-Order ODE's A Complete System A set of coupled.17) which is a subset of Eq. EXACT SOLUTIONS OF LINEAR ODE'S 1 2 57 . a u u0 = u (4. 4. They are found by solving the equation P ( ) = 0. c .12. 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 . The proof of this is simple and is found by substituting Eq.2. and t if and only if the 's satisfy Eq.11 is given by u(t) = c e 1 t + c e 2 t (4.13 into Eq.12) and the solution to Eq.14 if and only if " #" # " # " #" # " # a a x = x a a x = x (4. 2 1 0 1 2 1 2 1 2 1 1 1 0 2 2 2 1 2 0 1 2 4. determined from P( ) = + a + a = 0 (4.a u .13) where c and c are constants determined from initial conditions. 4.11 can be written u0 = .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. 4. Thus. 4. 4. there would be two roots. 4.

16 with A = . 4. one can solve the top equation rst.18) whose solution is not obvious. A Defective System If A is defective. THE SEMI-DISCRETE APPROACH 1 2 A Derogatory System Eq. 4. then it can be represented by the Jordan canonical form " # " #" # u0 = 0 u u0 1 u 1 2 1 2 (4.15 is still a solution to Eq. In this case 0 0 1 = 0 1 0 and 0 0 0 = 1 0 1 provide such a solution. substituting this result into the second equation. in this case. one nds du = u + u (0)e t dt which is identical in form to Eq. 4. .10 and has the solution 1 1 2 2 1 u (t) = u (0) + u (0)t]e t 2 2 1 From this the reader should be able to verify that u (t) = a + bt + ct e t 3 2 is a solution to 2 03 2 6 u0 7 = 6 1 4u 5 4 u0 1 1 2 3 32 3 76u 7 54u 5 u 1 2 3 if a = u (0) 3 b = u (0) 2 1 c = 2 u (0) 1 (4.14 if = = . giving u (t) = u (0)e t . 4. Then.19) The general solution to such defective systems is left as an exercise. However. provided two linearly independent vectors exist to satisfy Eq. This is the case where A has a complete set of eigenvectors and is not defective.58 CHAPTER 4.

nonhomogeneous. and Eq. and insert the identity combination XX .4 General Solution of Coupled ODE's with Complete Eigensystems Let us consider a set of coupled. we exclude defective systems. X . X .20 from the left by X . In the following.20 and 4. ~ = X . du = X . EXACT SOLUTIONS OF LINEAR ODE'S 59 4. = I between A and ~ . but because they are only of limited interest in the general development of our theory. X . ~ . ~ (t) u f (4.1. to a diagonal matrix having diagonal elements which are the eigenvalues of A. ~ (t) ~ g (4. Now let us multiply Eq.2. see Section 4. and X are also indepenu dent of both ~ and t. ~ .23 are expressing exactly the same equality. AX X .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. Represent them by the equation d~ = A~ . However. not because they cannot be analyzed (the example at the conclusion of the previous section proves otherwise).2.21 can be modi ed to u 1 1 1 1 1 1 1 1 1 d X .23) dt It is important at this point to review the results of the previous paragraph. 4. 4. There results u ~ X . 1 . ~ (t) u u f (4.2.4. by introducing the new variables w and ~ such that g 1 1 1 ~ w = X. ~ u 1 ~ (t) = X . the elements in X . ~ (t) u f dt u ~ Finally. rst-order ODE's with constant coe cients which might have been derived by space di erencing a set of PDE's.22) we reduce Eq.20 to a new algebraic form ~ dw = w .21) dt Since A is independent of both ~ and t. ~ (t) g f 1 (4. 4. and X . 4. linear. Notice that Eqs. this regrouping is crucial for the solution process because Eqs. The only di erence between them was brought about by algebraic manipulations which regrouped the variables.

gM (t) (4.25) where ~ m is the m'th column of X .23 and 4. ~ x f | Transient {z } | Steady-state {z } (4.. 4. rst-order equations. THE SEMI-DISCRETE APPROACH 4. gm (t) M wM .. the eigenvector corresponding to m .24 are also time invariant and the solution to any line in Eq.1~ f 1 cm e m t ~ m + A.23 are no longer coupled.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. In such a case. Transforming back to the u-system gives ~ (t) = X w(t) ~ u = = = = M X m=1 M X m=1 M X m=1 M X m=1 wm(t)~ m x cm e m t ~ m + x m=1 m M X 1 ~ gm xm cm e m t ~ m + X x .24) For any given set of gm (t) each of these equations can be solved separately and then recoupled. They can be written line by line as a set of independent. using the inverse of the relations given in Eqs. 0 wm = . 4. the gm in Eqs. g (t) 1 1 m wm . 4.20 when neither A nor ~ has f any explicit dependence on t. x We next focus on the very important subset of Eq.1 X .. single. thus w0 = ..26) .e. 0 wM = 1 1 w . i..22: ~ (t) = X w(t) ~ u = M X m=1 wm(t)~ m x (4. 4.60 CHAPTER 4.

The second group is referred to classically as the particular solution or the particular integral. respectively. In our application to uid dynamics.28) dt The dependent variable ~ represents some physical quantity or quantities which relate u to the problem of interest.4.and post-multiplication of A by the appropriate similarity matrices transforms A into a diagonal matrix. on the cleverness of the relator. but entirely equivalent. in the most general case. An alternative. ~ . we get di erent perspectives of the same solution. In this way. however. We saw in Sections 4. ~ u u f (4. ~ u x x x f (4. as might be expected. For the model problems on which we are focusing most of our attention. In particular.1. This permits us to say a great deal about such problems and serves as the basis for this section. we can develop some very important and fundamental concepts that underly the global properties of the numerical solutions to the model problems. we identify di erent mathematical reference frames (spaces) and view our solutions from within each. REAL SPACE AND EIGENSPACE 1 61 Note that the steady-state solution is A.3. form of the solution is ~ (t) = c e 1 t ~ + + cme m t ~ m + + cM e M t ~ M + A. f The rst group of terms on the right side of this equation is referred to classically as the complementary solution or the solution of the homogeneous equations. to a much greater extent. composed. we reduce the partial di erential equations to a set of ordinary di erential equations represented by the generic form d~ = A~ .3. The most natural reference frame is the physical one.1 and 4.2. We begin by developing the concept of \spaces". the elements of A are independent of both u and t. it is more instructive to refer to these groups as the transient and steady-state solutions. in the 4.3 Real Space and Eigenspace Following the semi-discrete approach discussed in Section 4. and this can add signi cantly to our understanding.1 De nition . There is. That is. another very useful frame. it is said u to be in real space. How these relate to the numerical solutions of more practical problems depends upon the problem and. We say If a solution is expressed in terms of ~ .27) 1 1 1 4. of Jordan blocks or.2 that pre.

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

B (M : a b c) and Bp(M : a b c).30) = sin m x m=0 1 M .5.3 Eigenvectors of the Model Equations The Di usion Model Next we consider the eigenvectors of the two model equations.3.3. Consider Eq. the model ODE's for di usion.4. Recall that xj = j x = j =(M + 1).ia sin 2m m=0 1 M .29) = 2(M + 1) x and. First. are given in Appendix B. 4.i m a m m=0 1 M .1.2 + 2 cos Mm+ 1 m = x ! m . we present results for a simple 4-point mesh and then we give the general case. to help visualize the matrix structure.31) x is the modi ed wavenumber from Section 3.4. m = m. Notice that the di usion eigenvalues are real and negative while those representing periodic convection are all pure imaginary.1 (4. These follow as special cases from the results given in Appendix B.32) . so in general the relation ~ = X w can be written as u ~ 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 uj = M X m=1 wm sin mxj j=1 2 M (4. From Section B. The interpretation of this result plays a very important role later in our stability analysis. we nd for the model di usion equation with Dirichlet boundary conditions .1 m = x M 2 2 2 where = . and x = 2 =M . 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 6 sin (x ) sin (2x ) sin (3x ) sin (4x ) 7 7 4 5 sin (x ) sin (2x ) sin (3x ) sin (4x ) The columns of the matrix are proportional to the eigenvectors. REAL SPACE AND EIGENSPACE 63 tridiagonals. from Section B.4 sin m=1 2 M (4.1 (4.4. for the model biconvection equation . 4.

Eq.32 represents the sine synthesis that companions the sine transform given by Eq.64 CHAPTER 4.1 M . 4. For our model ODE. we can write ~ = X w as u ~ uj = wmeimxj j=0 1 (4. ~ ~ u gives wm = M X j =1 uj sin mxj m=1 2 M (4. In summary. ~ = Xw ~ u is a sine synthesis from wave space back to real space. ~ is a sine transform from real space to (sine) wave u space. The coe cient matrices for these ODE's are always circulant. we nd 1 2 sin (x ) 6 sin (2x ) 6 6 sin (3x ) 4 sin (4x ) 1 1 1 1 sin (x ) sin (2x ) sin (3x ) sin (4x ) 2 2 2 2 sin (x ) sin (2x ) sin (3x ) sin (4x ) 3 3 3 3 sin (x ) 3 sin (2x ) 7 7 5 sin (3x ) 7 sin (4x ) 4 4 4 4 1 The rows of the matrix are proportional to the eigenvectors. the right-hand eigenvectors are given by ~ m = ei j x With xj = j M .5. In general w = X . For the model di usion equation: ~ w = X . Eq. 1 The Biconvection Model Next consider the model ODE's for periodic convection.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. . 4. Similarly.33) In the eld of harmonic analysis.1 x = j 2 =M . Eq. 4. THE SEMI-DISCRETE APPROACH For the inverse. 4.34) .1 X m=0 (2 m=M ) j =0 1 m =0 1 M .33. or left-hand eigenvector matrix X .1 M .

uj (t) = where M X m=1 cme m t sin mxj + (A.i = . i = 1 e 2 4 6 4 4 4 4 8 4 4 6 12 18 12 4 1 65 For a 4-point periodic mesh. REAL SPACE AND EIGENSPACE 2w 6w 6 6w 4 w 3 21 1 1 1 7 1 6 1 e. ~ 7 u 5 1 X 1 M . Eq.4 Solutions of the Model ODE's The Di usion Equation For the di usion equation. i = 7= 6 .3. ~ is a complex Fourier transform from real space to u wave space. u For circulant matrices.i = . 4.4: 1 2 3 4 4 4 4 32u 76u 76 76u 54 u 1 2 3 4 3 7 7 = X.4 sin 2(M + 1) x 2 2 ! (4. u e. f )j 1 j=1 2 M (4. we nd the following left-hand eigenvector matrix from Appendix B. ~ ~ u u =0 In general For any circulant system: ~ w = X .4. i = 7 4 61 e 5 4 e e . x.1 j j This equation is identical to a discrete Fourier transform of the periodic dependent variable ~ using an M -point sample between and including x = 0 and x = 2 .36) . i = e . it is straightforward to establish the fact that the relation ~ = X w represents the Fourier synthesis of the variable w back to ~ . i = e. i = e. 1 4. i = e.27 becomes We can now combine the results of the previous sections to write the solutions of our model ODE's.35) m m = . ~ = Xw ~ u is a complex Fourier synthesis from wave space back to real space.3.imxj wm = M m=0 1 M . In summary.

37. m 2 t sin m xj + h(xj ) j=1 2 M (4.i m a (4.39) We see that the solutions are identical except for the steady solution and the modi ed wavenumber in the transient term.1 (4.42) . The di erence between the modi ed wavenumber and the actual wavenumber depends on the di erencing scheme and the grid resolution. 4. We can write this ODE solution as uj (t) = M .37) With the modi ed wavenumber de ned as 2 m x m = x sin 2 and using m = m. Eq.1 X m=0 cme. f )j 1 j=1 2 M (4.i m at ei mxj j=0 1 M . This di erence causes the various modes (or eigenvector components) to decay at rates which di er from the exact solution. evaluated at the nodes of the grid: uj (t) = M X m=1 cm e. The modi ed wavenumber is an approximation to the actual wavenumber. With conventional di erencing schemes. THE SEMI-DISCRETE APPROACH (4.38) This can be compared with the exact solution to the PDE. we can write the ODE solution as uj (t) = M X m=1 cme.31.41) with the modi ed wavenumber de ned in Eq.1 X m=0 cme m t ei mxj j=0 1 M .1 (4. m 2 t sin m xj + (A. 2.40) m = . while high wavenumber modes (if they have signi cant amplitudes) can have large errors. The Convection Equation For the biconvection equation. low wavenumber modes are accurately represented.66 CHAPTER 4. we obtain uj (t) = where M .

the result is erroneous numerical dispersion. . 4. evaluated at the nodes of the grid: M. In the case of non-centered di erencing. solved there in their uncoupled form. not particular problems.23 express identical results but in terms of di erent groupings of the dependent variables.26.4. The role of m is clear it stands for some representative eigenvalue in the original A matrix. and then returned as a coupled set to real space. X uj (t) = fm (0)e. A typical member of the family is dwm = w .i mat ei m xj j=0 1 M .1 (4.3. the modi ed wavenumber is complex.5. m=0 4.20 and 4.4 The Representative Equation In Section 4. As discussed in Section 3. Eq.44) m m m dt The goal in our analysis is to study typical behavior of general situations.42 shows that the imaginary portion of the modi ed wavenumber produces nonphysical decay or growth in the numerical solution. 4.43) Once again the di erence appears through the modi ed wavenumber contained in . For such a purpose Eq. 4. This is helpful both in analyzing the accuracy of time-marching methods and in studying their stability. g (t) (4.4. The form of Eq. while the actual phase speed is independent of the wavenumber. 2. uncoupled equation and our conclusions will apply in general.44 is not quite satisfactory.2. this leads to an error in the speed with which various m modes are convected. since is real. which are related by algebraic manipulation. However. we pointed out that Eqs. topics which are covered in Chapters 6 and 7. THE REPRESENTATIVE EQUATION 1 67 and compare it to the exact solution of the PDE. We found the uncoupled solution to a set of ODE's in Section 4. discussed in Chapter 11. 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. The importance of this concept resides in its message that we can analyze timemarching methods by applying them to a single. Our next objective is to nd a \typical" single ODE to analyze. Since the error in the phase speed depends on the wavenumber.

This leads to w(t) = ce t + The Representative ODE dw = w + ae t dt (4. Consider the application of the operator given in Eq. u(1) = 1: @u = @ u . Write the resulting semi-discrete ODE form. The exact solution of the representative ODE is (for 6= ): t (4. 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.46) w(t) = ce t + ae . 3. Using this operator to approximate the spatial derivative in the linear convection equation. 3.44 has the exact solution: k which can be used to evaluate all manner of time-marching methods.52 to the 1-D di usion equation with Dirichlet boundary conditions. in principle. 6x @t @x 2 2 . write the semi-discrete form obtained with periodic boundary conditions on a 5-point grid (M = 5).68 CHAPTER 4.g(t) = ak eikt X ak eikt k ik .5 Problems 1. Consider the nite-di erence operator derived in question 1 of Chapter 3. Find the entries in the boundary-condition vector. we note that. For example X . 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) = 0.45) for which Eq. 4. one can express any one of the forcing terms gm(t) as a nite Fourier series. 4. From this we can extract the k'th term and replace ik with . In such evaluations the parameters and must be allowed to take the worst possible combination of values that might occur in the ODE eigensystem. 2.

Consider a grid with 10 interior points spanning the domain 0 x . The grid nodes are given by xj = j x j = 0 1 : : : 9. Using these. PROBLEMS 69 4. Note that the domain is 0 x 2 and x = 2 =10. (Plot the solution at the grid nodes only.) Calculate the corresponding modi ed wavenumbers for the second-order centered operator from Eq.Bp(10 . Calculate and plot the corresponding ODE solutions. The eigenvalues of the above matrix A. . plot the exact solution of the di usion equation with = 1 at t = 1 with m = 1 and m = 3.4.5. Compare with the exact solution of the PDE.37. 1 . For initial conditions u(x 0) = sin(mx) and boundary conditions u(0 t) = u( t) = 0.4. compute and plot the ODE solution at t = 2 for the initial condition u(x 0) = sin x. 5. can be found from Appendix B.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. Repeat for the initial condition u(x 0) = sin 2x. 4. Consider the matrix A = . 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. as well as the matrices X and X .

THE SEMI-DISCRETE APPROACH .70 CHAPTER 4.

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

A nondissipative scheme analogous to centered di erencing is obtained by taking the average of these two uxes.2) at the next time step. This ux must then be integrated to nd the net ux through the boundary. The ux is required at the boundary of the control volume. that is. Similarly. Thus the volume V in Eq. the source term P must be integrated over the control volume. In our examples. which are a function of Q. which is a closed surface in three dimensions and a closed contour in two dimensions. Next a time-marching method1 can be applied to nd the value of Z V QdV (5. Q can be represented within the cell by some piecewise approximation which produces the correct value of Q. 5. we will consider only control volumes which do not vary with time. . 5. When these are used to calculate F(Q).1. Let us consider these approximations in more detail.3) I Z d V dt Q + S n:FdS = V PdV (5. we have updated values of the cell-averaged quantities Q. In order to evaluate the uxes. Another approach known as ux-di erence splitting is described in Chapter 11. we see that several approximations must be made. each cell will have a di erent piecewise approximation to Q. the ux will be discontinuous.4) for a control volume which does not vary with time. First.1 Basic Concepts CHAPTER 5. they will generally produce di erent approximations to the ux at the boundary between two control volumes. 5.1 can be written as 1 Z QdV V V (5. As we shall see in our examples. The basic elements of a nite-volume method are thus the following: 1 Time-marching methods will be discussed in the next chapter.1 is that of a control volume whose shape is dependent on the nature of the grid. 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.72 5. This is a form of interpolation often referred to as reconstruction. Thus after applying a timemarching method. at the control-volume boundary. Examining Eq. we note that the average value of Q in a cell with volume V is Q and Eq.

This issue is discussed in Section 11.1 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 (5.2. Integrate the ux to nd the net ux through the control-volume boundary using some sort of quadrature.7) This PDE governs a simple plane wave convecting the scalar quantity.5. The one-dimensional form is recovered with = 0. 3. Using this approximation. These ideas should be clari ed by the examples in the remainder of this chapter.6) 5. The order of accuracy of the method is dependent on each of the approximations. Since there is a distinct approximation to Q(x y z) in each control volume.2 Model Equations in Integral Form 5.2. 4. V S Z where the unit vector n points outward from the surface or contour. A rQdA = C nQdl I (5.5) or. Advance the solution in time to obtain new values of Q. the following relation between rQ and Q is sometimes used: Z I rQdV = nQdS (5. In order to include di usive uxes. construct an approximation to Q(x y z) in each control volume. nd Q at the control-volume boundary.4. two distinct values of the ux will generally be obtained at any point on the boundary between two control volumes. in two dimensions. u(x y t) with speed a along a straight line making an angle with respect to the x-axis. 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. 2.2. MODEL EQUATIONS IN INTEGRAL FORM 73 1. Given the value of Q for each control volume. . Evaluate F(Q) at the boundary.

3. The nodes of the grid are located at xj = j x as usual. We consider an equispaced grid with spacing x.2 gives the following integral form (5. as shown in Fig. F is simply a vector.10) Since Q is a scalar. 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. d Z udA = I n: i @u + j @u ds dt A @x @y C (5. Eq. i @x @y P = 0 Using these. we nd (5. 5. 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. x=2 to xj + x=2. 2. the two-dimensional linear convection equation is obtained from the general divergence form.2.ru ! @u + j @u = .9) (5. 2.2 The Di usion Equation Q = u F = . Control volume j extends from xj .8) (5. with 5. We will use the following notation: xj.12) (5.3. x=2 xj+1=2 = xj + x=2 (5.11) where A is the area of the cell which is bounded by the closed contour C . FINITE-VOLUME METHODS For unit speed a. as in the model equations. Substituting these expressions into a twodimensional form of Eq.14) (5.74 CHAPTER 5.13) (5.15) ! to be the integral form of the two-dimensional di usion equation.17) .16) 5. 2.3 One-Dimensional Examples We restrict our attention to a scalar dependent variable u and a scalar ux f .1. Eq.1=2 = xj .

1=2 = 0 . With these de nitions.23) x ddtj + fj+1=2 .20) j j +1=2 dt j .19) Zx d ( xu ) + f . 5.1=2 or (5. Eq.1=2 Now with = x .3.5.1=2 = x j+1=2 Pdx (5. x=2 j @x j 2 @x2 j 6 @x3 j ! ! x2 @ 2 u + x4 @ 4 u + O( x6 ) = uj + 24 @x2 (5.18) (5. becomes u (5.1: Control volume in one dimension.3. 5.19 in a Taylor series about xj (with t xed) to get 2 3 ! ! ! 1 Z x=2 4u + @u + 2 @ 2 u + 3 @ 3 u + : : :5 d uj x . uj = uj + O( x2) 5. 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. the cell-average value becomes 1 Z xj+1=2 u(x t)dx u (t) j uj 1=2 = u(xj 1=2) x fj 1=2 = f (uj 1=2) (5.22) where uj is the value at the center of the cell. Hence the cell-average value and the value at the center of the cell di er by a term of second order. fj.1 A Second-Order Approximation to the Convection Equation In one dimension. the integral form of the linear convection equation. fj.11.21) 4 j 1920 @x j and the integral form becomes xj. we can expand u(x) in Eq. xj .

29 and 5. the discontinuity in the ux at the cell boundary is resolved by taking the average of the uxes on either side of the boundary. as shown in Fig. giving fjR 1=2 = uj (5. The cell to the right of xj +1=2.1 0 1)~ u u (5. Thus 1 f^j+1=2 = 2 (fjL+1=2 + fjR =2) = 1 (uj + uj+1) (5. cell j is the cell to the right of xj. gives fjR =2 = uj+1 (5.30 into the integral form.76 CHAPTER 5. 1 is the cell to the left of xj. We choose a piecewise constant approximation to u(x) in each cell such that u(x) = uj xj. and cell j .32) dt 2 x p .1=2 .1 + uj ) = x ddtj + 1 (uj+1 .1=2.1=2 = 2 (fjL.1=2 x xj+1=2 (5.1 we have de ned the uxes at the cell boundaries in terms of the cell-average data.24) Evaluating this at j + 1=2 gives fjL+1=2 = f (uL+1=2) = uL+1=2 = uj (5. 5. uj.1.1 + uj ) (5. 1 B (. Eq.1=2 = uj.1) = 0 (5. 5. we obtain u 1 u x ddtj + 2 (uj + uj+1) .1=2 + fjR 1=2) = 1 (uj. FINITE-VOLUME METHODS with f = u. In this example. 2 where f^ denotes a numerical ux which is an approximation to the exact ux.1 (5.27) . which is cell j + 1. Substituting Eqs.29) +1 2 and 1 f^j.26) +1 Similarly. giving fjL.23. 1 (uj.30) .31) 2 2 With periodic boundary conditions.25) j j where the L indicates that this approximation to fj+1=2 is obtained from the approximation to u(x) in the cell to the left of xj+1=2 . 5. this point operator produces the following semidiscrete form: d~ = .28) We have now accomplished the rst step from the list in Section 5.

35) the following values at the cell boundaries: uL+1=2 = 1 (2uj+1 + 5uj .uj. the piecewise quadratic approximation produces (5.5.1) j 6 c = .22ux2+ uj.1 with a piecewise quadratic approximation as follows u( ) = a 2 + b + c (5.3. and c are chosen to satisfy the following constraints: 1 Z . as in Eqs. ONE-DIMENSIONAL EXAMPLES 77 This is identical to the expression obtained using second-order centered di erences.1 + 26uj .1 u b = uj+12 . The three parameters a. uj+1 24 With these values of a.29 and 5.3. we can conclude that the use of the average of the uxes on either side of the cell boundary.30. and c.1 0 1) is relevant to nite-volume methods as well as nite-di erence methods.34) 1 These constraints lead to x Z 3 x=2 u( )d = uj+1 j a = uj+1 . u u Hence our analysis and understanding of the eigensystem of the matrix Bp(.36) .33) where is again equal to x . ~ . uj.3.1 x Z x=2 . x=2 u( )d = u x . xj . x=2 x=2 u( )d = uj (5. 5. Since the eigenvalues of Bp(.1 0 1) are pure imaginary. 5.3 x=2 1 j . b. b. except it is written in terms of the cell average ~ .x j.1 (5. rather than the nodal values. can lead to a nondissipative nite-volume method.2 A Fourth-Order Approximation to the Convection Equation Let us replace the piecewise constant approximation in Section 5.

With periodic boundary conditions.uj+1 + 5uj + 2uj.8 0 8 .23.1)~ u u (5.uj+2 + 5uj+1 + 2uj ) j 6 1 uL.2) 12 (5. The rst approach is simpler to extend to multidimensions.1 + uj.uj+2 + 8uj+1 . uj.1=2 = 1 (.37) (5. gives u 1 x ddtj + 12 (. the following semi-discrete form is obtained: d~ = .1=2 = 6 (2uj + 5uj.42) This is a fourth-order approximation to the integral form of the equation. we describe two approaches to deriving a nite-volume approximation to the di usion equation. uj.uj+2 + 7uj+1 + 7uj .78 CHAPTER 5.40) = 12 (.1 .1 .2) j (5. Eq.1) and 1 f^j. 1 B (1 . 5.uj+1 + 7uj + 7uj.3. while the second approach is more suited to extension to higher order accuracy.1) j 6 uR+1=2 = 1 (.3 A Second-Order Approximation to the Di usion Equation . as can be veri ed using Taylor series expansions (see question 1 at the end of this chapter).1. Recalling that f = u.3.39) using the notation de ned in Section 5.2) = 0 (5.1=2) + f (uR.41) Substituting these expressions into the integral form.1=2)] j j = 1 (.1=2 = 2 f (uL. we again use the average of the uxes on either side of the boundary to obtain f^j+1=2 = 1 f (uL+1=2) + f (uR+1=2)] j 2 j 1 (5. In this section. 5. uj. FINITE-VOLUME METHODS uR. 8uj.43) dt 12 x p This is a system of ODE's governing the evolution of the cell-average data.38) (5.

@u=@x. u ) (5. with Dirichlet boundary conditions.1 j (5.2. the integral form of the di usion equation. Also. Eq.49) (5. and we see that the properties of the matrix B (1 . 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.44) (5.16. Substituting these into the integral form.1=2 = 0 Zb 79 (5. = . u ) (5.48) (5. 5. 5.22. ONE-DIMENSIONAL EXAMPLES u x ddtj + fj+1=2 . fj. u(a) @x x xj @x x j+1 j From Eq. Eq.3.47) j +1=2 @x j+1=2 x j+1 j Similarly. we can write ! @u f^ = . 2u + u ) or. to second-order. 5. 5. 5.44.50) j +1 This provides a semi-discrete nite-volume approximation to the di usion equation.2 1) are relevant to the study of nite-volume methods as well as nite-di erence methods.33 we have dt x @u = @u = 2a + b @x @ (5. becomes with f = . Hence. 1x (uj .2 1)~ + bc u u ~ 2 f^j.1=2 = .1) dt x j .46) @u dx = u(b) . we use a piecewise quadratic approximation as in Section 5. uj.3.5. From Eq.45) In one dimension.51) . For our second approach. 1 (u . Eq.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 obtain x duj = 1 (u . d~ = 1 B (1 .ru = .

5. 1 (u .55) which is written entirely in terms of cell-average data. 1x (uj . and ` is the length of the sides of the hexagons. 2u + u ) (5. With f = . and A. 5. dt x2 j . .1.1 j j +1 dt x2 p 5.53) Notice that there is no discontinuity in the ux at the cell boundary. Thus our nal example is a nite-volume approximation to the two-dimensional linear convection equation on a grid consisting of regular triangles. fjR 1=2 = fjL. is the length of the sides of the triangles. this gives f R = f L = .56) A 3 The two-dimensional form of the conservation law is d Z QdA + I n:Fdl = 0 dt A C (5.3.@u=@x.2 1)~ u u (5.1) .1=2 = . The following relations hold between `. 1 ` = p 3 p A = 3 2 3 `2 ` = 2 (5.35. The nodal data are stored at the vertices of the triangles formed by the grid. The resulting semi-discrete form with periodic boundary conditions is d~ = 1 B (1 . we use a piecewise constant approximation in each control volume and the ux at the control volume boundary is the average of the uxes obtained on either side of the boundary. u ) x j+1 j (5.49. As in Section 5. uj.80 CHAPTER 5.2. as shown in Figure 5.57) . FINITE-VOLUME METHODS j +1=2 j +1=2 with a and b given in Eq. The control volumes are regular hexagons with area A. This produces duj = 1 (u .4 A Two-Dimensional Example The above one-dimensional examples of nite-volume approximations obscure some of the practical aspects of such methods.54) which is identical to Eq.52) (5.

A TWO-DIMENSIONAL EXAMPLE c b 2 3 81 p d 4 1 l ∆ 0 5 a e f Figure 5. Side 0 1 2 3 4 5 Outward Normal n (i .2.i p (. .2. 3 j)=2 pi (i + p j)=2 3 (.i + 3 j)=2 .5.2: Triangular grid. The contour in the line integral is composed of the sides of the hexagon. Since these sides are all straight.4. as shown in Figure 5. 3 j)=2 p Table 5. see Fig. Outward normals.i . 5.1. 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. where we have ignored the source term. A list of the normals for the mesh orientation shown is given in Table 5. respectively.1. i and j are unit normals along x and y.

60. That is. cos p (. 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.2.1=2 (5. Weights of ux integrals. in terms of u and the hexagon area A. cos .11. 5. FINITE-VOLUME METHODS For Eq. Side 0 1 2 3 4 5 n (i cos + j sin ) (cos .59) Then.58) where is a length measured from the middle of a side . if the integrals in the equation are evaluated exactly. Making the change of variable z = =`. see Eq.`=2 u ( )d (5.2. 5. we have " # 5 d Z u dA + X n (i cos + j sin ) ` Z 1=2 u(z)dz = 0 (5.82 CHAPTER 5.1=2 u(z)dz (5. cos + 3 sin )=2 .62) 2 1 A u dA = Aup (5.1=2 =0 The values of n (i cos + j sin ) are given by the expressions in Table 5.61) . Taking the average of the ux on either side of each edge of the hexagon gives for edge 1: Z up + ua Z 1=2 dz = up + ua u(z)dz = 2 .60) dt A . 3 sin )=2 p Table 5.60. the approximation to the ux integral becomes trivial. There . the integrated time rate of change of the integral of u over the area of the hexagon is known exactly. the two-dimensional linear convection equation. Z and the piecewise-constant approximation u = up over the entire hexagon. Introducing the cell average. one has the expression u( )d = ` . 5. 3 sin )=2 cos p (cos + p sin )=2 3 (.

64) (5. ud) + (cos + 3 sin )(ub .65) (5. using Taylor series expansions.42 is a fourth-order approximation to Eq.69) The reader can verify. that this is a second-order approximation to the integral form of the two-dimensional linear convection equation.5. PROBLEMS Similarly. uf )] = 0 (5. cos + 3 sin )(uc .60. ue) p +(.68) or dup + 1 (2 cos )(u . 5.23.67) Z Z u(z)dz = up + ud 2 4 u(z)dz = up + ue 2 5 Z Z Substituting these into Eq. we obtain u(z)dz = up + uf 2 0 p u ` A ddtp + 2 (2 cos )(ua .2. u ) a d b e dt 3 p +(.66) (5. 2. 5. uf )] = 0 (5. Use the following linear approximation: u( ) = a + b .5.63) (5. u ) + (cos + p3 sin )(u . cos + 3 sin )(uc . 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. 5.5 Problems 1. we have for the other ve edges: Z 83 u(z)dz = up + ub 2 2 u(z)dz = up + uc 2 3 (5. along with the expressions in Table 5. 5. Use Taylor series to verify that Eq.

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

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

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

These methods are simple recipes for the time advance of a function in terms of its value and the value of its derivative. will be presented in Section 11. The value of this equation arises from the fact that. 6. that is.7) du = u0 = u + ae t (6.1 for a method using two previous time levels. the new predicted solution is also a function of the time derivative at the new time level. CONVERTING TIME-MARCHING METHODS TO O E'S 87 With an appropriate choice of the 0s and 0s. The second is implicit and referred to as the implicit (or backward) Euler method.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. which is a fully-discrete method. we can analytically convert MacCormack's method.4. u0n.3 1 Here we give only MacCormack's time-marching method. u0n. The methods are said to be explicit if 1 = 0 and implicit otherwise. We refer to them as the explicit Euler method and the MacCormack predictor-corrector method. 6.8) dt written here in terms of the dependent variable. at given time intervals. these methods can be constructed to give a local Taylor series accuracy of any order. by applying a time-marching method. For an implicit method.6) According to the conditions presented under Eq. the rst and third of these are examples of explicit methods. and therefore the time advance is simple. un and un. for example.1 respectively.2.5) (6. The method commonly referred to as .1. u0n+1. implicit methods require more complicated strategies to solve for un+1 than explicit methods. The material presented in Chapter 4 develops a basis for evaluating such methods by introducing the concept of the representative equation un+1 = un + hu0n ~ 1 ~ ~ un+1 = 2 un + un+1 + hu0n+1] (6. for systems of ODE's and nonlinear problems.6. An explicit method is one in which the new predicted solution is only a function of known data. u. As we shall see.

11) . 6.9 is a linear O E . Eq. un = he h ae hn (6. 6. 6.8. This is demonstrated by repeating some of the developments in Section 4. Eq. the predictor-corrector sequence.6. 6.12) 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.7. expressed in terms of the dependent variable un and the independent variable n.5. We next consider examples of this conversion process and then go into the general theory on solving O E's. TIME-MARCHING METHODS FOR ODE'S such a linear ODE into a linear O E .8. As another example. to Eq. The latter are subject to a whole body of analysis that is similar in many respects to. The second line in Eq. the theory of ODE's.10) As a nal example.11 is identical to Eq. with constant coe cients.9.88 CHAPTER 6. . (1 + h)un = ahe hn ~ 1 u (6.2. 6. 1 un = 1 ahe h(n+1) 2 2 which is a coupled set of linear O E's with constant coe cients. and just as powerful as. to Eq. since the predictor step in Eq. h)un+1 .7 is simply the explicit Euler method. 6. gives un+1 . 2 (1 + h)~n+1 + un+1 .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. we nd un+1 = un + h un+1 + ae h(n+1) or (1 . Note that the rst line in Eq. 6. Eq. 6. 6.9) Eq. There results un+1 = un + h( un + ae hn) or un+1 . (1 + h)un = hae hn (6. but for di erence rather than di erential equations.11 is obtained by noting that u0n+1 = F (~n+1 tn + h) ~ u = un+1 + ae h(n+1) ~ (6. Apply the simple explicit Euler scheme. 6. applying the implicit Euler method.

one can readily show that the solution of the defective case. In terms of the initial value of u it can be written Just as in the development of Eq. and since the equations are linear and have constant coe cients. complex parameters.6. un+1 = un + a n is h i un = u0 + an . The exact solution of Eq. 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.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. in general. 6. SOLUTION OF LINEAR O E'S WITH CONSTANT COEFFICIENTS 89 6. and b are.3. un = u0 n + a b b .10.3. n un = c1 n + bab .1 n This can all be easily veri ed by substitution.13) where . (b = ). The independent variable is now n rather than t. n.1 First. is not a function of either n or u. we use for O E's the di erence operator E (commonly referred to as the displacement or shift operator) and de ned formally by the relations un+1 = Eun un+k = E k un Further notice that the displacement operator also applies to exponents. a.13 is where c1 is a constant determined by the initial conditions.14) d Instead of the di erential operator D dt used for ODE's. thus b bn = bn+ = E bn . 4.

14 can now be re-expressed in an operational notion as (E 2 + a1E + a0)un = 0 (6.14 for all c1 c2 and n should be veri ed by substitution. 6. 6.14.17) which can also be written ~ n+1 = C~ n u u (c11 . In the simple example given above.15) which must be zero for all un. 6. etc. E I ]~ = 0 un c21 (c22 . E ) C = c11 c12 c21 c22 The operational form of Eq. Eq. The fact that Eq. Thus Eq. 6. 6.16 is a solution to Eq. rst-order. In the analysis of O E's. . TIME-MARCHING METHODS FOR ODE'S where can be any fraction or irrational number. this time having the form P (E ) = det C . 2 . ) (c c12 ) = 0 21 22 . 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. linear homogeneous di erence equations have the form u(1n+1) = c11u(1n) + c12 u(2n) u(2n+1) = c21u(1n) + c22 u(2n) ~ n = hu(1n) u(2n)iT u c12 (6. there are just two roots and in terms of them the solution can be written un = c1( 1 )n + c2( 2 )n (6. we label these roots 1 . They are found by solving the equation P ( ) = 0.3. 6.17 can be written u1 (n) = C .2 Special Cases of Coupled First-Order Equations A Complete System Coupled. The -roots are found from P ( ) = det (c11c . its roots determine the solution to the O E.16) where c1 and c2 depend upon the initial conditions. . E ) u2 which must be zero for all u1 and u2. 6. Again we are led to a characteristic polynomial. E I ].15 is known as the operational form of Eq.90 CHAPTER 6. 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. and refer to them as the -roots.

one can show that the solution to A Defective System 2 3 2 un+1 7 6 6 un+1 5 = 4 1 4^ un+1 1 32 3 7 6 un 7 ^ 5 4 un 5 un is un = u0 n h i un = u0 + u0n . h)E . following the logic of Section 4.1 + u n(n .11. E .2. 1) . ~ are its eigenvectors.21) " .18) 6.4 Solution of the Representative O E's 6.2 for defective ODE's. 6. rst-order ordinary di erence equations. 1]un = h E ae hn (6.1 n ^ ^ " # .4. For example.(1 + h) # " u # = h " 1 # ae hn ~ 1E 1 (1 + h)E E .4. linear.19) (6.20) (6. are given by Eqs. the solution of Eq. Using the displacement operator.2 2 2 E .2 un = u0 + u0n ^ 0 2 n (6. 6.1 The Operational Form and its Solution Examples of the nonhomogeneous.2.9 to 6. The solution of O E's with defective eigensystems follows closely the logic in Section 4. 1 u n . produced by applying a time-marching method to the representative equation. (1 + h)]un = h ae hn (1 . these equations can be written E . SOLUTION OF THE REPRESENTATIVE O E'S 91 Obviously the k are the eigenvalues of C and.6.17 is if x 2 ~ n = X ck ( k )n~ k u x k=1 where ck are constants determined by the initial conditions.

Eq. Notice also. Eq.4.22) which is produced by applying time-marching methods to the representative ODE. 1 .23. 4. the ratio Q(E )=P (E ) can be found by Kramer's rule. 6. 6. We can express in terms of Eq. we have P (E ) = (1 . h)E .23: un = c1(1 + h)n + ae hn For the implicit Euler method. 6. The terms P (E ) and Q(E ) are polynomials in E referred to as the characteristic polynomial and the particular polynomial.21.19. Eq. as would be the case for Eq.25) . one for un and un and we are usually only interested in the nal solution un.21.1.2 Examples of Solutions to Time-Marching O E's As examples of the use of Eqs. 6.92 CHAPTER 6.45.23) P (e ) k=1 where k are the K roots of the characteristic polynomial.19 to 6. respectively. 6. 6. representing a time invariant particular solution.24) and the solution of its representative O E follows immediately from Eq. 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. Keep in mind that for methods such as in Eq. When determinants are involved in the construction of P (E ) and Q(E ). 1 Q(E ) = hE e h. P ( ) = 0. For the explicit Euler method. h Q(E ) = h (6. 6. or a steady state.22 all manner of standard time-marching methods having multiple time steps and various types of intermediate predictor-corrector families. The general solution of Eq.21 there are multiple (two in this case) solutions.22 can be expressed as K h X un = ck ( k )n + ae hn Q(e h) (6. h h (6.20. we have P (E ) = E . In such a case K X Q(1) un = ck ( k )n + a P (1) k=1 6. 6. ~ the important subset of this solution which occurs when = 0. 6. we derive the solutions of Eqs.22 and 6.

h . one can write n u(t) = c1 e 1 h ~ 1 + x + cm e mh n ~m + x + cM e Mh n ~ M + P:S: (6. h)e h . 1 2h2 P (E ) = det . 1 2 2 2 h Q(E ) = det . The complete solution can therefore be written 1 h 1 2h2 n + ae hn 2 h e + 1 + h un = c1 1 + h + 2 (6. 1 6.1 Establishing the Relation 6. 1 2h2 2 he h 1 n un = c1 1 .5 The . This relation is rst demonstrated by developing it for the explicit Euler method. Remembering that t = nh. 2 2 h2 = 0 which has only one nontrivial root ( = 0 is simply a shift in the reference index).26) e h . Relation We have now been introduced to two basic kinds of roots. one solves for the nal family un (one can also nd a solution for the intermediate family u). 1 (1 + h)E E . Eq. 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.27. RELATION so 93 In the case of the coupled predictor-corrector equations. 1 (1 E h)E 1 hE = 1 hE (E + 1 + h) 2 2 + 2 The -root is found from 1 P( ) = . 1 . 1 . 4. 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.5. There is a fundamental relation between the two which can be used to identify many of the essential properties of a time-march method. the -roots and the -roots. THE .6. 1 .(1 + h) = E E . and there ~ results # " E .5. The former are the eigenvalues of the A matrix in the ODE's found by space di erencing the original PDE. 6. h . h + ae hn (1 . h .21.27) x .

3 The error is O( 2 h2 ). instead of the Euler method. 2 hE . we see a correspondence between m and e m h.29) Applying Eq. the solution2 of the resulting O E is un = c1( 1 )n ~ 1 + x + cm ( m )n ~ m + x + cM ( M )n ~ M + P:S: x (6. So if we use the Euler method for the time advance of the ODE's. 6. The other root. . For one of these we nd m q h+ 1+ m 1 = 1 + mh + 2 = 2 h2 . TIME-MARCHING METHODS FOR ODE'S where for the present we are not interested in the form of the particular solution (P:S:). 1. 1 = 0 2 (6.29.1 + 2hu0n (6.3.30) Now we notice that each produces two -roots. m 2 Based on Section 4. Now the explicit Euler method produces for each -root. 6. 1 4 h4 + m 2 h2 m (6. we have the characteristic polynomial P (E ) = E 2 .28.4.28) where the cm and the ~ m in the two equations are identical and m = (1 + m h). which is de ned by un+1 = un. Suppose. so that for every the must satisfy the relation m . q 1 + 2 h2 . x Comparing Eq.31) (6. which is given by = 1 + h. will be discussed in Section 6.94 CHAPTER 6. 6.27 and Eq. 2 mh m . Since the value of e h can be expressed in terms of the series 1 e h = 1 + h + 2 2h2 + 1 3h3 + 6 1 + n! nhn + the truncated expansion = 1 + h is a reasonable3 approximation for small enough h. 6. we use the leapfrog method for the time advance.32) mh 8 m This is an approximation to e m h with an error O( 3h3).5. one -root.8 to Eq. .

while the second-order time-marching method which results from this approximation. un.34. always produces one -root for every -root that satis es the relation (6. Now consider the solution obtained using the same method with a time step h=2. which is the leapfrog method: un+1 = un.1) (6. relation for the leapfrog method produces two -roots for each . this is reduced by a factor of eight (considering the leading term only). Therefore the error at the end of the simulation is reduced by a factor of four. 4. Consider a solution obtained at a given time T using a second-order time-marching method with a time step h. One of these we identi ed as the principal root which always .5.2 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. RELATION 95 6. 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.5.1 + 2hu0n (6. 6. We refer to the root that has the above property as the principal -root. twice as many time steps are required to reach the time T .3 Spurious -Roots We saw from Eq. THE . knowing only that its leading error is O hk+1 . However. The following example makes this clear. Thus the principal root is an approximation to e h up to O hk . consistent with a second-order approximation.5. 6. and designate it ( m )1 .6.6.30 that the .34) 2h has a leading truncation error which is O(h2). The above property can be stated regardless of the details of the timemarching method. Note that a second-order approximation to a derivative written in the form ( t u)n = 1 (un+1 . 6.33) 1 2h2 + + 1 k hk + O hk+1 =1+ h+ 2 k! where k is the order of the time-marching method.35) has a leading truncation error O(h3). Since the error per time step is O(h3).

28 must be modi ed. The initial vector ~ 0 u does not provide enough data to initialize all of the coe cients. 1 or earlier. the .96 CHAPTER 6. if one starts the method properly) the spurious coe cients are all initialized with very small magnitudes.1 and thus cannot be started using u only ~ n. Then the presence of spurious roots does not contaminate the answer. they play virtually no role in accuracy analysis. generation of spurious roots does not. u Presumably (i.4 One-Root Time-Marching Methods There are a number of time-marching methods that produce only one -root for each -root. Following again the message of Section 4. except for the principal one. For example.36 must be initialized by some starting procedure. In general. and presumably the magnitudes of the spurious roots themselves are all less than one (see Chapter 7). 1 or earlier to advance the solution from time level n to n + 1. relation produced by a time-marching scheme can result in multiple -roots all of which. if there is one spurious root to a method. This results because methods which produce spurious roots require data from time level n . all of the coe cients (cm )2 in Eq. The other is referred to as a spurious -root and designated ( m )2 . in itself. That is.36) Spurious roots arise if a method uses data from time level n . If a time-marching method produces spurious -roots. 6. the leapfrog method requires ~ n.5.4.. we have un = c11 ( 1 )n ~ 1 + + cm1( m )n ~ m + + cM 1 ( M )n ~ M + P:S: 1x 1x 1x n~ + n~ + +c12 ( 1)2 x1 + cm2 ( m )2 xm + cM 2 ( M )n ~ M 2x n~ + n~ + +c13 ( 1)3 x1 + cm3 ( m )3 xm + cM 3 ( M )n ~ M 3x +etc. To express this fact we use the notation = ( h). For example. after some nite time the amplitude of the error associated with the spurious roots is even smaller then when it was initialized. Thus while spurious roots must be considered in stability analysis. 6. They are also called one-step methods. However. if there are more spurious roots (6.. All spurious roots are designated ( m )k where k = 2 3 . TIME-MARCHING METHODS FOR ODE'S has the property given in 6. it is always some algebraic function of the product h. . are spurious. In fact. the solution for the O E in the form shown in Eq. No matter whether a -root is principal or spurious. We refer to them as one-root methods. Such roots originate entirely from the numerical approximation of the time-marching method and have nothing to do with the ODE being solved. 6. make a method inferior.33. It should be mentioned that methods with spurious roots are not self starting.e. many very accurate methods in practical use for integrating some forms of ODE's have spurious roots.

It is usually used as the basis for establishing the order of a method. However. 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. vectors. )u0n + u0n+1 6. One is the error made in each time step. the so-called -method.6. the trapezoidal ( = 2 ). For example. h h . 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.6. and the implicit Euler methods ( = 1). A popular method having this property. and matrices are identical except the ~ and the terms inside u the parentheses on the right hand sides.1~ u x x x f ~ n = c1( 1 )n ~ 1 + + cm( m )n ~ m + + cM ( M )n ~ M + A. It is quite common to judge a time-marching method on the basis of results found from a Taylor table. 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. ACCURACY MEASURES OF TIME-MARCHING METHODS 97 1 The -method represents the explicit Euler ( = 0).4.4. is given by the formula h i un+1 = un + h (1 . respectively. it is of no use in: . The only error made by introducing the time marching is the error that makes in approximating e h.1 Local and Global Error Measures There are two broad categories of errors that can be used to derive and evaluate timemarching methods. Notice that when t and n = 0. This is a local error such as that found from a Taylor table analysis.2. It is useful for comparing methods.6. a Taylor series analysis is a very limited tool for nding the more subtle properties of a numerical time-marching method. This is a global error. relation is ) = 1 +1(1 . These are ~ (t) = c1 e 1 h n ~ 1 + + cm e m h n ~ m + + cM e M h n ~ M + A. as we shall see in Chapter 8. these equations are identical. Three one-root methods were analyzed in Section 6.37) respectively. so that all the constants. see Section 3.1~ u x x x f (6.6 Accuracy Measures of Time-Marching Methods 6. Its .

. (6. nding the global error.2 Local Accuracy of the Transient Solution (er Transient error er! ) The particular choice of an error measure. The order of the method is the last power of h matched exactly. including only the contribution from the principal root. This is similar to the error found from a Taylor table.39) j j 6. In the discussion of the . and to study them we make use of the material developed in the previous sections of this chapter. We designate this by er and make the following de nition er e h. analyzing the particular solution of predictor-corrector combinations. TIME-MARCHING METHODS FOR ODE'S nding spurious roots. 1 The leading error term can be found by expanding in a Taylor series and choosing the rst nonvanishing term. either local or global. However. is to some extent arbitrary. given by t u(t) = ce t + ae . a necessary condition for the choice should be that the measure can be used consistently for all methods.98 CHAPTER 6.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 ) (6.38) and the solution to the representative O E's. evaluating numerical stability and separating the errors in phase and amplitude. a very natural local error measure for the transient solution is the value of the di erence between solutions based on these two roots.6. The latter three of these are of concern to us here. Our error measures are based on the di erence between the exact solution to the representative ODE. Therefore.

Such can indeed be the case when we study the equations governing periodic convection which produces harmonic motion. For such cases it is more meaningful to express the error in terms of amplitude and phase. The approach to error analysis described in Section 3. tan. Cn = ah= x. A positive value of erp corresponds to phase lag (the numerical phase speed is too small).a . as follows . Let = i! where ! is a real number representing a frequency. that is q era = 1 .1 ( 1 )i=( 1 )r )] and the local error in phase can be de ned as (6. while a negative value corresponds to phase lead (the numerical phase speed is too large). Then the numerical method must produce a principal -root that is complex and expressible in the form (6.1 (6. we have h = . is found using = . We have found these to be given by 1 P:S:(ODE) = ae t ( . ) .42) er = er! = 1 + tan ( 1)i =( 1)r )] p where ! = . ACCURACY MEASURES OF TIME-MARCHING METHODS 99 Amplitude and Phase Error Suppose a eigenvalue is imaginary. Thus one can obtain values of the principal root over the range 0 x for a given value of the Courant number. ( 1)2 + ( 1 )2 i r er! !h .ia .41) Amplitude and phase errors are important measures of the suitability of time-marching methods for convection and wave propagation phenomena. Introducing the Courant number. 1( h).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. j 1 j = 1 .iCn x.5 can be extended to the combination of a spatial discretization and a time-marching method applied to the linear convection equation.6. !h Cn x 6.6.6. where is the modi ed wavenumber of the spatial discretization. The above expression for er! can be normalized to give the error in the phase speed.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. The principal root.

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. 6. TIME-MARCHING METHODS FOR ODE'S h P:S:(O E) = ae t Q(e h) P (e ) respectively. If the forcing function is independent of time. More precisely. and it is necessary that the advertised order of accuracy be valid for the nonlinear cases as well as for the linear ones.6. For a measure of the local error in the particular solution we introduce the de nition er P:S: h P:S:(O E) .43 can be written in terms of the characteristic and particular polynomials as er = co . 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. so that the order of er and er are consistent. this order is quite important in determining the true order of a time-marching method by the process that has been outlined.P e h o (6.44) co = h!0 h( . many numerical methods generate an er that is also zero. 1 (ODE ) ( ) (6. Eq. and for this case.46) (6. time-marching methods are usually applied to nonlinear ODE's. In order to determine the leading error term. An illustration of this is given in the section on Runge-Kutta methods. 6. )Q e h .45) (6.47) . However. where n ( .100 and CHAPTER 6. is equal to zero.43) The multiplication by h converts the error from a global measure to a local one.h ) lim P e The value of co is a method-dependent constant that is often equal to one.4 Time Accuracy For Nonlinear Applications In practice. The algebra involved in nding the order of er can be quite tedious.

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. given by the relation T = Nh Global error in the transient A natural extension of er to cover the error in an entire event is given by Er e T .1 ( 1) i 1r . we are more concerned with the global error in amplitude and phase. This subject is covered in Chapter 8 after our introduction to stability in Chapter 7. tan.1 ( ) =( ) ] = !T . and q " ( 1 )2 + ( 1)2 i r N (6.6.50) . N tan 1i 1r !# (6. ACCURACY MEASURES OF TIME-MARCHING METHODS 101 The reader should be aware that this is not su cient.6. ( 1 ( h))N (6. For example. Then the required number of time steps.48) Global error in amplitude and phase If the event is periodic.5 Global Accuracy In contrast to the local error measures which have just been discussed.49) ( ) Er! N !h . Suppose we wish to compute some time-accurate phenomenon over a xed interval of time using a constant time step. is N. we can also de ne global error measures. the analysis based on a linear nonhomogeneous ODE produces the appropriate conditions for the majority of time-marching methods used in CFD. However. These are given by Era = 1 .6. These are useful when we come to the evaluation of time-marching methods for speci c purposes. We refer to such a computation as an \event". Let T be the xed time of the event and h be the chosen step size.

1.K 1 0 1 X k k E Aun = h@ k=1.7 Linear Multistep Methods In the previous sections. and they are said to be K -step because K time-levels of data are required to marching the solution one time-step.52) . we introduce classes of methods along with their associated error analysis. 6. Eq. h. They are explicit if 1 = 0 and implicit otherwise. The methods are said to be linear because the 's and 's are independent of u and n. 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.K 1 k hn k E A( un + ae ) (6. TIME-MARCHING METHODS FOR ODE'S Global error in the particular solution Finally. ) Qeh . we have developed the framework of error analysis for time advance methods and have randomly introduced a few methods without addressing motivational.51 is applied to the representative equation. 6.K k Fn+k (6.1 P eh 6. one nds 0 1 @ X k=1. the global error in the particular solution follows naturally by comparing the solutions to the ODE and the O E.7. When Eq.4.K k un+k = h du = u0 = F(u t) dt 1 X k=1.102 CHAPTER 6. It can be measured by Er ( . and the result is expressed in operational form. 6.8. In the subsequent sections. since most of them have historic origins from a wide variety of disciplines and applications.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. developmental or design issues.51) where the notation for F is de ned in Section 6. We shall not spend much time on development or design of these methods.

un h u0n h2 u00 n h3 u000 n h4 u0000 n .6.h .2 . One can show that these conditions are met by any method represented by Eq.1 u0n.1 .1 k=0 The Adams-Bashforth family has the same 's with the additional constraint that 1 = 0.4. 1) k Since both sides of Eq. referred to as Adams-Bashforth (explicit) and AdamsMoulton (implicit).44.54) A Taylor table for Eq. Two well-known families of them.1 + . and it is certainly a necessary condition for the accuracy of any time marching method. to be of any value in time accuracy.2)3 .1. can be designed using the Taylor table approach of Section 3.2 (6.51 with k = . 6. that approximates e h through the order of the method.1 . these methods are often \normalized" by requiring X k=1 Under this condition co = 1 in Eq.2 1 6 Table 6.2 .1 u0n. 6. 11 6 .53) 1=1 0 = .5.2 Examples There are many special explicit and implicit forms of linear multistep methods.1 6 . labeled 1. The three-step Adams-Moulton method can be written in the following form un+1 = un + h( 1u0n+1 + 0 u0n + .54 can be generated as 1 1 1 un+1 1 1 2 6 24 .1 .0 1 1 . .1 1 . that is ! (1 + h) as h ! 0.51 can be multiplied by an arbitrary constant.2u0n.(.(. 6. 12 .(. We can also agree that. .un . The Adams-Moulton family is obtained from Eq.h 0 u0n . Taylor table for the Adams-Moulton three-step linear multistep method.2 that a time-marching method when applied to the representative equation must provide a -root. LINEAR MULTISTEP METHODS 103 We recall from Section 6. 6.2 u0n. k k k k 6.7.1 1 .51 if X X X k = 0 and k = (K + k . 6.1 2 . The condition referred to as consistency simply means that ! 1 as h ! 0.1 .2)2 .(.2 .1 .2) (6. a method should at least be rst-order accurate.2)0 .h 1 u0n+1 .7.h .2)1 .2 2 .

2 = 1=24 which produces a method which is fourth-order accurate.1 + 2hhun Leapfrog 1 h 3u0n .4 7 6 0 7 6 1 7 6 7 6 7 6 3 0 3 12 7 6 76 (6. TIME-MARCHING METHODS FOR ODE'S This leads to the linear system 2 1 1 1 1 32 3 213 1 6 2 0 .1 7 = 6 1 7 5 4 5 4 0 . resulting in (6.58) 0 = 23=12 .55 and 6.4 7 6 .57 up to the order of the method.4 .5=24 .2 .2 to solve for the 's.2 .1 + 5u0n. u0n.1 5 = 4 1 5 (6.56) 1 = 9=24 0 = 19=24 .5.2 that a kth-order time-marching method has a leading truncation error term which is O(hk+1 ). Explicit Methods un+1 un+1 un+1 un+1 = un + hu0n Euler 0 = un. u0n. 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.16=12 . un. One can verify that the Adams type schemes given below satisfy Eqs.1 = .2 giving (6. some of which are very common in CFD applications.55) 4 5 4 .104 CHAPTER 6. is given below together with identifying names that are sometimes associated with them. = un + hu0n+1 Implicit Euler 1 hhu0n + u0n+1i = un + 2 Trapezoidal (AM2) 1 h4un .57) 4 54 0 3 12 1 .1 = .1i = un + 12 AM3 . 16u0n. A list of simple methods.4 With 1 = 0 one obtains 2 32 3 2 3 1 1 1 0 7 617 6 0 . 6.2 = 5=12 This is the third-order Adams-Bashforth method.1 + 2hu0n+1i =3 2nd-order Backward h h5u0n+1 + 8u0n .32 1 .2 AB3 Implicit Methods un+1 un+1 un+1 un+1 4 Recall from Section 6.1i = un + 2 h 0 i AB2 h = un + 12 23un .

6.51. K=2 in Eq.1=2 .6.5=6 . The most general two-step linear multistep method (i. which corresponds to 0 = 0 in Eq. un.1=6 . 'u0n. . LINEAR MULTISTEP METHODS 105 6.1=2 0 0 0 0 . + ')u0n .1] + h u0n+1 + (1 .e.2=9 0 1=2 .1=2. see Eq.1=3 .2. which corresponds to .1=3 0 1=12 .1 (6. +2 The class of all 3rd-order methods is determined by imposing the additional constraint =2 .' = ..2.1=2 .1=4 .1=2 0 .51. =3 = 6 .5 6 1 Finally a unique fourth-order method is found by setting = .e. Some linear one. the methods are 2nd-order accurate) if 1 '= . are known as Milne methods. One can show after a little algebra that both er and er are reduced to 0(h3) (i. Notice that the Adams methods have = 0.59) Clearly the methods are explicit if = 0 and implicit otherwise.7.. that is at least rst-order accurate.7.1=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 6.51). can be written as h i (1 + )un+1 = (1 + 2 )un . ' 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 . Methods with = . 6.59 is given in Table 6. This limits the interest in multistep methods to about two time levels. 6.1=2 . A list of methods contained in Eq. 6.3 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.1=6 0 . 6.and two-step methods.1 = 0 in Eq.59.

Their use in solving ODE's is relatively easy to illustrate and understand.62) Considering only local accuracy.106 6. For this to hold for er .6. 6. One concludes. ( + ) h . 6. TIME-MARCHING METHODS FOR ODE'S There are a wide variety of predictor-corrector schemes created and used for a variety of purposes. One can analyze this sequence by applying it to the representative equation and using the operational techniques outlined in Section 6. each of which is referred to as a family in the solution process. it is obvious from Eq. fractional-step. one-corrector example is given by un+ = un + h 0n ~ hu i un+1 = un + h u0n+ + u0n ~ (6. following the example leading to Eq. therefore. one is led.4. Q(E ) = E h E + + h] 2 h2 i (6. It is easy to show. Their use is motivated by ease of application and increased e ciency. where measures of e ciency are discussed in the next two chapters. that h P (E ) = E E . 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. 1 . 1 u0n ~ (6. There may be many families in the sequence. to the following observations. A simple one-predictor.8 Predictor-Corrector Methods CHAPTER 6.44 that the same conditions also make it O(h3). and usually the nal family has a higher Taylor-series order of accuracy than the intermediate ones.26. but it is not di cult to show using Eq. and hybrid methods. by following the discussion in Section 6. un+ = un + hu0n ~ 1 un+1 = un + 2 h 1 u0n+ + 2 .60) where the parameters and are arbitrary parameters to be determined. Predictor-corrector methods constructed to time-march linear or nonlinear ODE's are composed of sequences of linear multistep methods.63) 5 Such as alternating direction. that the predictor-corrector sequence is a second-order accurate method for any .61) (6. For the method to be second-order accurate both er and er must be O(h3).61 that + =1 =1 2 which provides two equations for three unknowns. . The situation for er requires some algebra.

66) (6. is un+1 = un + hu0n ~ 1 un+1 = 2 un + un+1 + hu0n+1] ~ ~ Note that MacCormack's method can also be written as un+1 = un + hu0n ~ un+1 = un + 1 h u0n + u0n+1] ~ 2 from which it is clear that it is obtained from Eq. The Gazdag method. MacCormack's method. 6.9. u0n.68) 6. presented earlier in this chapter. The Adams-BashforthMoulton sequence for k = 3 is i 1 h un+1 = un + 2 h 3u0n .64) Some simple.6 that produce just one -root for each -root such that ( h) corresponds explicit Runge-Kutta methods here.1 ~ i hh u un+1 = un + 12 5~0n+1 + 8u0n . speci c. 6. we refer to these as ABM(k) methods. (6. which we discuss in Chapter 8. is 1 hu ~ i un+1 = un + 2 h 3~0n . .1 ~ h i un+1 = un + 1 h u0n + u0n+1 ~ ~ (6.63 with = 1=2 is 1 un+1=2 = un + 2 hu0n ~ un+1 = un + hu0n+1=2 ~ and.67) (6. referred to as Runge-Kutta methods. If the order of the corrector is (k).65) 2 The Burstein method. u0n. 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.9 Runge-Kutta Methods There is a special subset of predictor-corrector methods. 6 Although implicit and multi-step Runge-Kutta methods exist.6. second-order accurate methods are given below.1 (6.63 with = 1. nally. The order of the combination is then equal to the order of the corrector. u0n. we will consider only single-step. obtained from Eq.

70) and this is much more di cult. as we pointed out in Section 6. . TIME-MARCHING METHODS FOR ODE'S to the Taylor series expansion of e h out through the order of the method and then truncates. Thus for a Runge-Kutta method of order k (up to 4th order). They must satisfy the relations = 1 = 1+ 1 (6. The most widely publicized Runge-Kutta process is the one that leads to the fourth-order method.71 and 6. 6. Thus. a scheme entirely equivalent to 6. the choices for the time samplings. we prefer to present it using predictor-corrector notation. 6. the principal (and only) -root is given by (6. u(tn) = 1 k1 + 2k2 + 3k3 + 4k4 (6.69) = 1 + h + 1 2h2 + + 1 k hk 2 k! It is not particularly di cult to build this property into a method. First of all. the method must further satisfy the constraint that er = O(hk+1) (6. To ensure k'th order accuracy. We present it below in some detail. but.72) un+1 = un + 1hun 2 bn+ 3 ~n+ 1 + 4 hu0n+ 2 Appearing in Eqs.70.6. it is not su cient to guarantee k'th order accuracy for the solution of u0 = F (u t) or for the representative equation.69 and 6.73) 2 = 2+ 2+ 2 . are not arbitrary. 1. 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) .4.71) However.108 CHAPTER 6.72 are a total of 13 parameters which are to be determined such that the method is fourth-order according to the requirements in Eqs. and 2 .71 is b un+ = un + hu0n b un+ 1 = un + 1 hu0n + 1hu0n+ ~ b un+ 2 = un + 2 hu0n + 2hu0n+ + 2hu0n+ 1 ~ 0 + hu0 + hu0 (6.

two parameters can be set arbitrarily.7 There are eight equations in 6. To satisfy the condition that er = O(k5).76) 6 conditions for Runge-Kutta methods of up to third order.69 the four conditions 1+ 2 + 3 1+ = 1 2+ 3+ 4 = 1=2 3 1+ 4 2 ( 2 + 1 2 ) = 1=6 4 = 1=24 4 12 (1) (2) (3) (4) (6. RUNGE-KUTTA METHODS 109 The algebra involved in nding algebraic equations for the remaining 10 parameters is not trivial. but the equations follow directly from nding P (E ) and Q(E ) and then satisfying the conditions in Eqs. 7 The present approach based on a linear inhomogeneous equation provides all of the necessary .6.73 to eliminate the 's we nd from Eq. 6.69 and 6. It results in the \standard" fourth-order Runge-Kutta method expressed in predictor-corrector form as 1 b un+1=2 = un + 2 hu0n b un+1=2 = un + 1 hu0n+1=2 ~ 2 un+1 = un + hu0n+1=2 ~ h i b un+1 = un + 1 h u0n + 2 u0n+1=2 + u0n+1=2 + u0n+1 ~ (6. we have to ful ll four more conditions 2 2 = 1=3 (3) 2 2+ 3 1+ 4 2 3+ 3 3+ 4 3 = 1=4 (4) 2 1 2 (6. Since the equations are overdetermined. which is based on a linear nonhomogenous representative ODE. 6.70. Several choices for the parameters have been proposed. As discussed in Section 6.75) 2 1 + 4 ( 2 2 + 2 2 ) = 1=12 (4) 3 1 (4) 3 1 1 + 4 2 ( 2 + 1 2 ) = 1=8 The number in parentheses at the end of each equation indicates the order that is the basis for the equation. but the most popular one is due to Runge. Thus if the rst 3 equations in 6.75 which must be satis ed by the 10 unknowns.6.74) These four relations guarantee that the ve terms in exactly match the rst 5 terms in the expansion of e h . the fourth condition in Eq. Using Eq.75 are all satis ed.75 cannot be derived using the methodology presented here.74 and the rst equation in 6.4. 6.9. A more general derivation based on a nonlinear ODE can be found in several books.74 and 6. the resulting method would be third-order accurate. 6.

we will see that these methods can have widely di erent properties with respect to stability. It is clear that for orders one through four. Following the steps outlined in Section 6.66 and 6. Higher-order RungeKutta methods can be developed. and third-order methods can be derived from Eqs.1 Application to Systems of Equations Consider rst the numerical solution of our representative ODE u0 = u + ae t (6.10 Implementation of Implicit Methods We have presented a wide variety of time-marching methods and shown how to derive their .2. storage requirements reduce the usefulness of Runge-Kutta methods of order higher than four for CFD applications. respectively. We shall discuss the consequences of this in Chapter 8. but they require more derivative evaluations than their order. 6. In any event.74 and the rst equation in 6. 6.10. 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. These are covered in this Section. un = he h ae hn (6. we obtained (1 . h)un+1 .72 by setting 4 = 0 and satisfying only Eqs.78) . a fth-order method requires six evaluations to advance the solution one step. 6.75. RK methods of order k require k evaluations of the derivative function to advance the solution one time step. TIME-MARCHING METHODS FOR ODE'S 9 > > = > > Notice that this represents the simple sequence of conventional linear multistep methods referred to. In the next chapter. For example. 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. relations. This leads to various tradeo s which must be considered in selecting a method for a speci c application.110 CHAPTER 6. 6. 6.77) using the implicit Euler method.67 are second-order Runge-Kutta methods.

hf (t + h) u u or (6.83) (6. hA)~ n+1 .85) dt 2 . For our one-dimensional examples.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. 6. for biconvection.81) (6. h (un + he h ae hn ) 111 (6.1 0 1)=2 x). and hence its solution is inexpensive. hA). In general. f (t) u u ~ ~ (I . IMPLEMENTATION OF IMPLICIT METHODS Solving for un+1 gives 1 un+1 = 1 .80) ~ where ~ and f are vectors and we still assume that A is not a function of ~ or t. as we shall see in Chapter 8.aBp (. A = . but rather we solve Eq.78 is (6. Now u u the equivalent to Eq.82) ~ ~ n+1 = (I . 6.79) This calculation does not seem particularly onerous in comparison with the application of an explicit method to this ODE.. consider the nonlinear ODE du + 1 u2 = 0 (6.10. hf (t + h)] u u The inverse is not actually performed.84) un+1 = un + hF (un+1 tn+1) This is a nonlinear di erence equation. As an example. the cost per time step of an implicit method is larger than that of an explicit method. The primary area of application of implicit methods is in the solution of sti ODE's.1 ~ n .81 as a linear system of equations.10.6. 6. Now let us apply the implicit Euler method to our generic system of equations given by ~ 0 = A~ . the system of equations which must be solved is tridiagonal (e. but in multidimensions the bandwidth can be very large. requiring only an additional division.g. ~ n = .

TIME-MARCHING METHODS FOR ODE'S solved using implicit Euler time marching. u )2 + @ 2 F (u . 6. t ) F (u t) = F (un tn) + @u n n @t n n ! ! 1 @ 2 F (u .87) On the other hand. 6.88) If t is within h of tn .86) n which requires a nontrivial method to solve for un+1.112 CHAPTER 6. both (t . Designate the reference value by tn and the corresponding value of the dependent variable by un.87 can be written ! ! @F (u . An iterative method. There are several di erent approaches one can take to solving this nonlinear di erence equation. un)k are O(hk ).10. A Taylor series expansion about these reference quantities gives ! ! @F (u . tn)2 + @t2 n (6. can be used. tn)k and (u . we expand F (u t) about some reference point in time.83. the \initial guess" for this nonlinear problem can be quite close to the solution. t )2 @ 2 u + u(t) = un + (t . which gives 1 un+1 + h 2 u2 +1 = un (6. tn) @t n @t2 n n 2 (6.89) . u )(t . and Eq. u ) + @F (t . which implies that a linearization approach may be quite successful. 6.3 Local Linearization for Scalar Equations General Development Let us start the process of local linearization by considering Eq. Such an approach is described in the next Section. In order to implement the linearization. the expansion of u(t) in terms of the independent variable t is ! ! @u + 1 (t . t ) + O(h2) F (u t) = Fn + @u n n @t n n (6. u ) + @F (t . such as Newton's method (see below). since the \initial guess" is simply the solution at the previous time step. t ) + 2 @u2 n n n @u@t n !n 1 2 + 2 @ F (t . In practice.

un) + h @F + O(h2) + Fn 2 @u n @t n 2) +hO(h (6. There are. Using Eq. and the trapezoidal time march.6. 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. one nds un+1 = un + 1 h Fn + @F (un+1 . 6. 2 h @F @u n 2 @t n which is now in the delta form which will be formally introduced in Section 12. locally-linear approximation to F (u t) that is valid in the vicinity of the reference station tn and the corresponding un = u(tn).91) where we write hO(h2) to emphasize that the method is second order accurate.10. relative to the order of expansion of the function.90) Implementation of the Trapezoidal Method As an example of how such an expansion can be used. 6.6.89 to evaluate Fn+1 = F (un+1 tn+1).83. 6. @F u + @F (t .92 results in the expression 1 un = hFn + 1 h2 @F (6. 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. A very useful reordering of the terms in Eq. Thus.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. namely ! ! ! ! du = @F u + F . consider the mechanics of applying the trapezoidal method for the time integration of Eq.83. With this we obtain the locally (in the neighborhood of tn) time-linear representation of Eq.93) 1 . The trapezoidal method is given by 1 un+1 = un + 2 h Fn+1 + Fn] + hO(h2) (6. it represents a second-order-accurate. The use of local time linearization updated at the end of each time step. combine to make a second-order-accurate numerical integration process. of course. In many uid mechanic applications the nonlinear function F is not an explicit function " ! ! # " !# ! . other second-order implicit timemarching methods that can be used.

u 2. Solve for the elements of the matrix multiplying ~ n and store in some approu priate manner making use of sparseness or bandedness of the matrix if possible. and remove the explicit dependence on time.90 into this method.83. it is generally the space-di erenced form of the steady-state equation of some uid ow problem. 6. Solve the coupled set of linear equations B ~n = R u ~ for ~ n.1 in carrying out this step).95) if we introduce Eq. In such cases the partial derivative of F (u) with respect to t is zero and Eq. store them in an array say R. Solve for the elements of hFn. we arrive at the form !# n un = hFn (6.114 CHAPTER 6.94 is usually implemented as follows: ~ ~ 1. 3. 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 . It is the product of h and the RHS of the basic equation evaluated at the previous time step. Find ~ n+1 by adding ~ n to ~ n. but for our applications. 6. Let this storage area be referred to as B .94) Notice that the RHS is extremely simple. h @F @u (6. TIME-MARCHING METHODS FOR ODE'S of t. u 4.93 simpli es to the second-order accurate expression " 1 1 . A numerical time-marching procedure using Eq. thus u u u ~ n+1 = ~ n + ~ n u u u The solution for ~ n+1 is generally stored such that it overwrites the value of ~ n u u and the process is repeated.96) . rearrange terms. 6. and save ~ n. (Very seldom does one nd B . In this example. the basic equation was the simple scalar equation 6. 2 h @F @u !# n un = hFn (6.

Omission of this factor degrades the method in time accuracy by one order of h.97) un+1 = un . we mean that the error after a given iteration is proportional to the square of the error at the previous iteration. There results or .94 and 6. 6. Our task is to apply the implicit trapezoidal method to the equations d3f + f d2f + dt3 dt2 0 !21 @1 . Quadratic convergence is thus a very powerful property. We choose the Falkner-Skan equations from classical boundary-layer theory. 6.96 obtained by dividing both sides by h and setting 1=h = 0. By quadratic convergence.10. Use of a nite value of h in Eq. @F @u ! n un = Fn @F @u ! #. given by Eq.4 Local Linearization for Coupled Sets of Nonlinear Equations In order to present this concept. let us consider an example involving some simple boundary-layer equations. The reader should ponder the meaning of letting h ! 1 for the trapezoidal method.6.87 and 6. Newton's Method Consider the limit h ! 1 of Eq. 6. 6. i.88 (remember the dependence on t has been eliminated for this case). df A = 0 dt (6. The fact that it has quadratic convergence is veri ed by a glance at Eqs.99) Here f represents a dimensionless stream function.. IMPLEMENTATION OF IMPLICIT METHODS 115 We see that the only di erence between the implementation of the trapezoidal method 1 and the implicit Euler method is the factor of 2 in the brackets of the left side of Eqs. 6.96. and is a scaling factor.e.98) This is the well-known Newton method for nding the roots of a nonlinear equation F (u) = 0. We shall see later that this method is an excellent choice for steady problems.1 n (6.96 leads to linear convergence. the error at a given iteration is some multiple of the error at the previous iteration.94. . where the error is the di erence between the current solution and the converged solution. 6. " Fn (6.10.

2 vector for the second term.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. called the Jacobian matrix.101) Now we seek to make the same local expansion that derived Eq.90. rather than a simple nonlinear scalar function. It is derived from Eq.103) 2 6 6 6 6 A=6 6 6 6 6 4 ~ u The expansion of F (~ ) about some reference state ~ n can be expressed in a way u similar to the scalar expansion given by eq 6. 6. TIME-MARCHING METHODS FOR ODE'S u1 = d f dt2 2 First of all we reduce Eq.116 CHAPTER 6. 1 . 6.102 by the following process d~ = F (~ ) u ~ u dt (6. and de ning F n as F (~ n). except that this time we are faced with a nonlinear vector function.8 Let us refer to this matrix as A. The required extension requires the evaluation of a matrix.87. u2 2 0 = F =u u2 2 1 0 = F =u u3 3 2 and these can be represented in vector notation as u2 = df dt u3 = f (6. a matrix times a .100) (6. Omitting the explicit dependency ~ ~ u on the independent variable t. and tensor products for the terms of higher order.u1 u3 .99 to a set of rst-order nonlinear equations by the transformations This gives the coupled set of three nonlinear equations u01 = F1 = .102) A = (aij ) = @Fi =@uj For the general case involving a third order matrix this is (6. @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. one has 9 2. 6.

PROBLEMS 117 ~ u ~ F (~ ) = F n + An ~ .102 as \constant" which is a locally-linear.105) where t . Find an expression for the nth term in the Fibonacci series. What is u25 ? (Let the rst term given above be u0. The number of times. (u1u3)n . we nd the Jacobian matrix to be 2 . and the manner in which.11.107) 4 5 0 1 0 The student should be able to derive results for this example that are equivalent to those given for the scalar case in Eq.or second-order time-marching method.) . u the solution will be second-order-accurate in time. Thus for the Falkner-Skan equations the trapezoidal method results in 2 3 1 + h (u3)n . tn and the argument for O(h2) is the same as in the derivation of Eq. d~ = A ~ + F . 6. A ~ +O(h2) u ~ n n un nu dt | {z } (6. of course. ~ n + O(h2) u u (6. Returning to our simple boundary-layer example. u2)n 3 2 2 6 2 7 6 . 6.2 1 We nd ~ n+1 from ~ n + ~ n.6.101. 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.93. on the nature of the problem.h (u1)n 74 6 5 4 5 1 0 76 ( u2)n 7 = h 6 2 4 5 ( u) (u2)n h 3n 0 . Any rst. (1 . and the solution is now advanced one step. Without any iterating within a step advance. 6.u1 3 0 0 7 A=6 1 (6. which is given by Eq.1.88. u3 2 u2 . could be used to integrate the equations without loss in accuracy with respect to order. 6.11 Problems 1.106) 6. Using this we can write the local linearization of Eq. second-order-accurate approximation to a set of coupled nonlinear ordinary di erential equations that is valid for t tn + h. explicit or implicit. h(u2)n h (u1)n 72 ( u1)n 3 2 . Re-evaluate u u u the elements using ~ n+1 and continue. the terms in the Jacobian matrix are updated as the solution proceeds depends.

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

Let u(0 t) = sin !t with ! = 10 .0:5)= ]2 with = 0:08. PROBLEMS 119 Find the di erence equation which results from applying this method to the representative equation. Use only 4th-order Runge-Kutta time marching at a Courant number of unity. Use 2nd-order centered di erences in space and a grid of 50 points. 200. 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. For the explicit Euler and AB2 methods. plot the error given by v uX u M (uj . Find the global order of accuracy from the plot. ah= x.11. implicit Euler. 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. Find the . Use the explicit Euler. of 0. Plot the solutions obtained at t = 1 compared to the exact solution (which is identical to the initial condition). use u(x 0) = e. For the initial condition. What order is the homogeneous solution? What order is the particular solution? Find the particular solution if the forcing term is xed. Using the computer program written for problem 7. Run until a periodic steady state is reached which is independent of the initial condition and plot your solution . Show solutions at t = 1 and t = 10 compared to the exact solution. 7. uexact)2 u j t M j =1 where M is the number of grid nodes and uexact is the exact solution. and 400 nodes. use a Courant number of unity. use a Courant number. 8.relation. including the homogeneous and particular solutions. 2nd-order Adams-Bashforth (AB2). On a log-log scale. Using the computer program written for problem 8. and 4th-order Runge-Kutta methods. trapezoidal. 10. repeat problem 9 using 4th-order (noncompact) di erences in space. 11. Repeat problem 7 using 4th-order (noncompact) di erences in space. Find the solution to the di erence equation. Find er and er . 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 100.0:5 (x.6.1 for the other methods. 9.

i.67). Use 2nd-order centered di erences in space with a 1st-order backward di erence at the out ow boundary (as in Eq.69. the number of grid nodes. 6. Repeat problem 11 using 4th-order (noncompact) centered di erences. for the combination of second-order centered differences and 1st.2. 2.69) together with 4th-order Runge-Kutta time marching.120 CHAPTER 6. Note that the required -roots for the various Runge-Kutta methods can be deduced from Eq. Explain your results. without actually deriving the methods. Find the global order of accuracy. 1. j . derive and use a 3rd-order backward operator (using nodes j . and j +1 see problem 1 in Chapter 3). and j ) and at the second last node. Use a third-order forward-biased operator at the in ow boundary (as in Eq. j . that obtained using the spatial discretization alone. 3.. 200. . 3.6.e. At the last grid node. 2. erp. Using the approach described in Section 6. and 4th-order Runge-Kutta time-marching at a Courant number of unity. era . Use grids with 100. 3. j . 3rd. 13. as described in problem 9. nd the phase speed error. use a 3rd-order backwardbiased operator (using nodes j . and 400 nodes and plot the error vs. TIME-MARCHING METHODS FOR ODE'S compared with the exact solution. j . 1. 2nd. Also plot the phase speed error obtained using exact integration in time. and the amplitude error. 12.

͏ @Ì° 34—48bDG˜£P9@h8ImP04@37PwXF44@aV0@PTVFjHY@E@o4@c1PIeFXG4EDC4A TAI 0 f 97 G 9W 9gg7 57 T 9727 A27 T2 F2 0 `YHGPFP9c@d7vGh3Ae0Sº76FjVIeT@xc8bs4S· ¯ sy’…ef1tI”80DGe035VwPT4¹A“XYP9ctpq4}…1X° ¯ syi@€db4VpËVf1tIR FW T G w T7 0 G A27 ¥ ˆ ™ 5 R F „ 9 R 0b {  ¥ ˆ T7 w T 5 806GV035VwPT4¹AH„h3fe0m80SP743Ahef6580“P2u q@xP2q3280VGXGVF5 u txSz7˜XY39e0S87SE3gV0x3A80œz7e0e5mV0@U4@PI8Fœz7PÊeF6GV035ewu XT4A F F T G F 0 G F 27 F 0W 2 w G 57 T G G 57 T2 G F 60@Eeg4@w˜Xf8Fwc8bdbXw4b X2”4—48bDG˜…HY41mSE8bheIXT@ee0@m~36TP04Sº7eFS@‡“624432l3gcA 42 w7 T A27 0 G 0 G F F TAI 0 f 92I G7 G 97 GI É 9W 57q G 0A 0 0 0 ( wpA G F YW ¥ ˆRI 5 R 0 G F cP23Fc4jVI85“P2v€} ¯ lyc@ÀVf1tIy48b‡P2d32e0eGHGeF5 u tǗzY7 u @w@z7È43ge0VGyV01˜Ç1~I u cP2—Xq1@“P2F 0WT 97 2 wT T 5I f5I G ( F 9Iq TAI 0 Æ 4—4eb6Go~…®  ½Åħ­ ¬ ¨ ² § ¶ 4@l™ EÃaÀ¨ q8I}4@37P0—4eb6Gh~65P04ymXG6FPwXF4jDid48bE@X04h@cDi0 A27 ªÂ Á ¿ ¾ G TA 9 AI 0 f 9p ˆ 9g 0 0 G™ 9g f7 5I¸ 0227 f 5 5 Gb F T F R 7 F 5 R ˆ † „ 0 0( 0 G G AI 0 f ( F 57 1Eu63544Sq©VA@7 u 1tIeR“1X(@87e5VG’HYe9P5S½7x#P2¼ef1tI»‡aiPweF65432@i48bpeI#j4eb6Ghp4X24dbew@˜f D„hPfeF”˜4P2—HY44@…“…3A4X2@d7c@~XTxef1tIU6580VGSº7d48b4£j4eb6Gh#3ge0VG¹TD„41”~1E}1HY39@Xq8FDG3wleg6Te05 0 G 7 ( F 9gg7 YW 0 F GWI F 5 R G 9 0 G™AI 0 f 02I 7 5I¸  0 0 ² § ¶ ‡E§² ‰ y¨ tz³E§² ‰ ® @· ™¯­ « µ ´ A2 4@7 ­ « ª ¨ ±®X° ™¯­ V®Š |§ ¬ ˜§ ‰ y©y¦‰§Š ¦ YW G 5 5 57 T2 G7p 0 0b { TAI 0 “#6A80“32V03TV043gV0‡e0@~41XI8FS4—6¥uV03T4}y64jc8b6Ghf ( F 57 0 GRI 2 G 9gg7 YW ‰ ˆ † „ 0 G G 5 5 0 G f R G 0 0( ‰ ˆ ‡ „ 0 4P2cdbew@˜`f…„˜Xf8F£Si1PIeFSz7XwPF44Sq“r†T}‡‡‡@XqeFSd7“32V03TV043gV0m4eb¤1eIt5h3Ae0S87654321†T}”‡vc8bG 32c8bh4@B£db@“eIc4@xe06Ge065ewXT4PAh3fV0x48b§€u†T}‡BxPwVF@Ex4eb¡1eIt5–3Ae0S8765432@e†T}‡‡”@I 0 G A27 ™ w7 5gg7 F F T 0 G YW ‰ ˆ † ¢ T7W 0 G f R G 0 0( ‰ ˆ † „ R T 5 R 0 G G 5 5 0 G 0g 0 0A   A27 Ž Gg7b Ÿ  –› • “ T7 G7 f w T ˜Vf1tIw@Xq8Fœd7“32V03Te0c3ge0a4eb…6At@PI39@6q4io4@h~eT35e0c@Eœ…ž@lœ—š™˜@—–D•”@3T60PwVF85S˜db4Vp”8IG 65t06RV0’P9X9F u §ij4‹1PIeF6G3w—p60e0|…Œ‹4@s§eb@l@w“324c32l3g44X2”e0@a6T30PwVF85S˜d“32P06Fh6ˆ—3Iw 5 0 ‘  Ž 2 0 T Š A27 ‰ GIW RI G 0A 0 0A F 57 G7 f G w 0 0 G F F ‰ u ˆ ‡ „ A 2 7 ‰u ˆ † „ G 0 A 0 0 0 G F w 2 w T 7 W 5 p „  5 F c8b”dbPw4b u P2B‘Tg”‡pc@B†T}‡‡“324462t6g4`A…„˜Xf8Fi8bHGF u PThe26530416I}PwVF@Ey4vrseƒ1I u PTc8bG X2‚3f4ebq1lP9390 @4w4v0 c4j3™eTc3g60413Iu4P2eFVG3TV035e0€X2q4S“@87VG1lhPf~e0@‡V03T4}|z“P21lg F 0 G 2I A GI2 IA GpW G w2 w ( G F A27 G2 5Ig F 57 0b { G FI 3Ajcd˜S87uc8b˜eIo6A806GS8785VGS‡—zY7 u @mPT…1PIeFcPp1VImeTXGySxef14w˜@eIE‚X2@q@4vS48b’P24ebXGF u 0iC 0 f T 0 G G w G7 T 97 F 2 G 9 T F RI 5I2 0 f T™ F7 fIA G7 G F f R F F7 fIA F w 7 F 9W T 0W G F T F T T 7 T G 0 G ƒ 5 w G 1eIt5sr™XR…P2Sq1c#P2@87VG6530qP2uX04@87VGelp8IoPA@e7mXTn3fe0VG—VY˜‡˜Vf35e0ue06T48b#`2m3eT@8I6GS8785VGS7 A27 G FIg 0iC RI T G F T FA RI F FW T F f72YA Fp• F G 4S”8T“P21lk3Ajc@ihef35e0hP2g3Ae0VT43pVwPT4€6280fG@eXTs`YHGPFX94@d7vGB3™3TXw˜SE—c–PA44€X2”VG6Te035e0“P2F RI ‰ ˆ † „ 5 0 92I2 0 G RI T w 0 G ƒ  0 T2 w ( 0W 2I 0 fI 0bg 5 9 57 @’‘T‡‡…@374P2XF414Q4ebu@Qe0@37§c8by`2¤‚3AV0354XAeF416I€4P26Fty1432h146244x@P743pXw8FvG@Eg 0 G 2I TA 0 0A A27 9 0 F2p 5I2 9g T 0 02 F FW T RI 2 F2 0A 9 0 0( c8b§@u4462t6g4sc@r@87VT351Hq44a14iX04hPfeFd35c8bXG6F4aPT`YXGPFX94@87VGUSQ@PI8FHG4EDC4B@87654321)' %  $ #  ! "  $  &             ©§¥ ¤¢ ¡   ¨ ¦ £ .

8YR8&©71¨QQ1W10T'7R8VUTRQ©P'7I'H©GED"¨CBA6§9871#'¨6'54¥3%©"¦210)#('&%©"¦!©§¥¤£ B # X ¨ ¨ ¨ # # S 73 F # #F & @ $ # $ # ¨  $ #  ¨ ¨¦ 36T4@P74§¢DiX04˜@3Iw 02 9g   9g f A27 ý 9g f w 0 G GIW F TAI 0 f Tp 57q RI FW T 0 G 0 F f7 0 G G F 0q2 4S€s…iP04h16Iq4eb§8bSt§P2d4jc8b6Ghd41PIVF@SU@©`YHGPFX94@d7vGq4ebhcP2h@cDieIe“32X0432@€@3Iw GXF cEvP9X9F u 0 u 4Sw@XY691XqeF6G6wtVg3Te0‡—· ¯ 4@§X° ¯ 6—y…1tIR 4¡c@¼E@o3T4Xp@S“321X(3F0 A2C A27 ™ 0 0 5 ™ A27 T¥ ˆ 5 ÿ  A27 ÿý T7 0 97q 0 V03T4ebh3A80œE1P(VF3T4x0 u 3™VT3580j@Edbe4@PI—6qV04–`2@`YXGPFX94@87VGp4eb˜4X2˜Vf35e06G4©eIe“32P06FhceppV0@7 0 G G72 0A Gg7 w Tp F 5g ƒ  FW T 0 G 0 F 0A G G w ˆ T 5  ¯ 4S©X° ¯ 3€yþP2h3T60PwVF85S˜#4eb‡@q6T4Pp@S“32@P(3F48b3™6Te0@37–4P2eFVG3TV035e0€X2k€Sqþ`2ƒ A27 T¥ ˆ F G7 f 0 G RI 0 97q 0 0 0 G T w ( G F Y27 f 3PTjœ˜vY@EP2@@qXfF Fi7 57 F(7 0 G (2 97 9 G A G 0 97q 0 „ý 0 G I• G7 F fI „2 0q2 w FA 0g 5I c8bwc1PI@mP0XFueIp432e0r6T4Pp@S“32@P(3F¹04‡48bT u Eh3Ae0SEP2h14¹Aj1XI8F6G6w1“16ImXw4jXIeF65tr1E¸ ü 3XTjœÇ@z7V05 Fi7 9 0 G7 02 0 G (2 97 9 G A G 0 97q 0 „ý 0 G I• G7 F fIA 2 Tp FA 5I @XqeFS“3(chc8bq4@PI@iX0PF’8I‹46280”6T4Pp@S“32@P(3F¹04˜4eb‡T u 4#3A80œEP2h14©1XIeF4tXøcv1E¸ ü û( 9 R 0 G YW 57 f f T 0W G b(pI 0 G2 5Ig F 57 0b 4cP2F u PIX91tI‡48bp“’3A60PÊVF@qh4Vpdl”8I’141c32U“@87VG1lhPfae0@…DY4}{ 33TX08FvG35l1eI4u`YXGXFP94@87VGÛSrXTeF—XYSE@d4P2F u XIP91tI48b”P2r3ThPfeFi“@qd8I˜3Ae0V535t0DRe0~e0@41XI8FS876G3wlg 0g 5g FW TRI T 9727 ( 9 R 0 G F 0 G Y27 f G 5 57 T2 G 0 „ 0 0b {  Ž F¸ T™7 0 fI 0bg 0 G ( F F G2 w T 9W 5g 9 TYbg9 w 5g R €…iqe06T4}i1H°4‚4X(|h30V0Äc432h1432c4˜V03T4ebp4P2cP2@87€@3I‡h3fP0c1eI4B@37PwVFj“4|@z7Xw8F6GS874@I T f 0 T 0 97q 0 0 0 G 5 R 0g 0 0W 9p w G7 2 0q2 w FA 0g A27 2 T h4epe56G6wtVgdcPp@S“321X(3Fd48b1tI6A80DG3wlc…iwlXA416I”S4b u 1PIeF6G6w1“13I‡Xw4—PIeF65to4@o1PIVF4tpR º„XR4d1tIw‘T}‡‡U694jhd41è1eIt5§441tIq0 u c†Í4h1XI8F6G6wj–X2|SP04h@cDid1E#63TP0eFVG65t1VI4g FA 5 R ‰ ˆ † „ 0AI f 5pI f R A2p R  Ž 2 0 F ™ 9g f7 0 5I¸ 0g 5 9@d7“32h@Ec4tp}P53Fc8b‡80S˜PfeFVG6TEz™VGSz7P0~S…vY13Ic8bxP2s‚@37u0 u 1vY@E1XI8FS87VGae0@7 µ 4@7 ª ¹Rƒ G 0 f7A2 R 0 G G7 0 T 9 G7 5 0 G F™2 w ™ 572 G T 5 A2  G7 F fIA G 5 0gg7 G 9 0 GRI 0g 5g 0 G 0ƒ7 @e0SEX2˜1c˜eIU@z7l4@6580eGœº7a48bÈ@}6TP08FvG35l1eIcdc8b‡@@˜f 2 w T2 G 9gg7 9 w 5g T w F F ™02I 0 0A 7 G T w 5 F 57 0W 2 @37‡41XI8FS37PwPFc4@r@z7Xw8FDG@874˜e0@37hdbPwcb u X2|@c1q@Xq8FDG3wt0DR4o8I”e01XI39oXYP9VF@87e5XG4VW@plsSz7w 6f80vGjvYU80DGP04h16IvrSE8bÈ354cP2h3fe0‡”ebXGF u c†Íj4”1PIeF6G6wj~30V0j6™˜6f80VG—vY4321X(3F}@Xq8FDG3wt0DR4”4@7 T T 9g f w 7 G7 G 0A F 5 7 ™  Ž 2 0 T T T T 0 0 0 0A A2 806GX04h13Iheb@lx1zqEeœE8ba3T60PweFe5S˜#8Is621Hq1plrX9u P9F u 1PIeFS87654PAVF41©SP04Sd7VGwPTú3f80vG—VYT 9g f w GIW 0 7b G7 G G7 f G 0 F( 0W 2 G 0 T2I Ÿ  9W T F T 7 F G „YA T 7 0 w7 5gg7 2 G 9 T 7 F G7 •  0 G RI G7 0 27 0 F( mXR’e0Sd7vG`T“c@3780VG‹h3T4dbS“eI4c@€1PIeFcPp@eI©ÃdbXw4b Serù@V–‹4ebm@xe0S˜Pf8FvG3T€@h3T@Xq1eXGF G7 G 0( G27qA7 0AA7 G2 5Ig F 0 G T7b T T T 0 0 0 G RI T 972 '  2 0 œE8b‡1@87“@Sc@o3A44S“@d7vG1thPfa4eb~SEmh3fe0VG—VYc321X(3Fde03Tc8bÈ@PTeF—HY@E”yc†Ž ¯ 1XI8F6G6wj X2§6Ae0VT43pVwPT4vXTvPTc}dc4P26F432V035tXø4”30SEeg©4SehPf8Fec8b#8bSt©1‹462t6g4‹@z71PIeFXG4EDC4A F T FA F Fb { ( w 0 FA w7 T A27 0 G 0 G GIW 2I A 0 0A 2 w 2 F2 0 `YHGPFP9c@d7vGpÇD™d®6T41eXwPF1leW65t—“…XYX9@º76F3wleg3T ­0 41XI8FSE€6¥Ä@P78F“32V035tXø4È@P78FVGSEx8I41XI8FS37PwPFc4@7 FW T 7 02I 9I 0gYb 9 0 T2 G7p 0 9 G 0 FA9 57g G T2 G 9gg 5pI F ™ w T 9g T T Fp¥ GI2 F 2 G7 T 0 G T w F G F 0 I ÷  G7 f c1hP2yS304X2eFhP04hPfVF©eI˜80XGc—e@4uXTs1XI8FS48pXGVFp4ebpe0@37uPTc8bsX26™351Dq0 u ~#tHiVF85S˜ueTXGF R@¡6f80VG—vY4321X(3Fh48b’8I©6A80SP739e0hl…VIPT@§32e0fG@Q@z7x· ¯ sy}S`YHGPFP9c@d7vG˜4}h ª @I 6f80G I T T 0 0 0 G G G 5 0W 97 RI 2 w ¥ ˆ RI FW T 0b { R „rjvY432@P(3Fw48bo1…XY39V0P5eF€62‡4462t6g4eH° ¯ syr@x`YHGPFP9c@d7vGe4}”—· ¯ c@eH° ¯ 6—yr@m8T´ vG@Eg T T 0 0 0 G 2I G 0 TA 0 0A ¥ ˆ RI FW T 0b {  A27 T¥ ˆ RI 57 Tc13I46211h14x4eb@e1XIjzqE6btxc8bh1§XY69e0P5eF“32€3AV0@EpV0@§`YXGPFX94@87VG~@e4@PI8FHG4EDC4e4v„ p 0 0(I fIb 0 G RI 5 F 7 0W 0 G 2I G 0 T7W 57 FW T RI T2 F2 0A 5p öçïô õô êç ò ç ð ê í çì êç ë êç éç ‡’È…Ä©‡”3óÀñ #’ï h‡#©‡s‡èæ ±åpã ä × âÓÒ×Ü× Ô ÐÓ ÙÚ ßÝ Ü Ù ÙØ ÐÒ× Õ ÔÓÒ Ñ Ð Ï #rÄÇBÇávy”ràۀ}ސ3ÒÛrÚ”4ÇñÖxBÄv”m’Î Ž SÌ° .

› ‰  ‰ ý ° ‰ š   ý ° š ¨   ´ œ› ‰ š ´œ ý ˜™ EŸž ‰ ˜™ › EŸž ‚‰ ˜™ EŸž 0W Gb F f 57g Tp 0 0(I fIb 0 l}“1X(˜vG@E416I4321@˜1cdc8bG R@)ef1tIl@z7Xw4—`Yue‘Ž44sy|S…Sz780vG4P2j@X04˜S4…i1EBsdƒ—PI4x@4eA1@—”8Ip@PI8FœE€3¥”X014X2eFT I 5 R9 Fg G 7 d Ž¥ ˆ RI A T F™ 9g f7 0 5I¸  w 9W 27 5I– 7 G 2 G7p 0 9( 7 1eIt5‡6T14@Edb©@PI8FœE€3¥h1XqeFSd7“62e03TV043gV0i48bB@¼ef@tIe48b‡h3f80vG—VY”1Hq8F6G6wt06R4xeI#3Ae0Sº769e05 f R 0(27 w 2 G7p 0 0 G G 5 5 0 G RI 5 R 0 G T T T 0 0A G G 60@Eegc321X(3FQ@s`2”‚@PI8FDG3wV0wœE8b©P2x”•H°4…8Ip”“H°4Ç3€ys@#XYX9@º76F3wleg6T§4S…†Íj4a41XI8F6G6wj w7 T 0 0 27 ƒ 2 T G7 G F Ž G Ž T¥ ˆ 2I 9 0 0 A27  Ž T2 0 X2U8THG4ep6Te0m48bq@I u e74v0 u PTc8bB1EB@30SEeg4621P(6FpP2r41XI8FcXp1eIrP56F48bÛS”e0c8pSE”48b6G3wlegcP2F F 9 5 0 G 2 5A F G 5I¸  w7 T 0 0 F T2 G 9 T 0 GRI 5 G72 0 G 0 T 0 u 6™˜6f80vGjvY432@P(3Fh1Hq8FDG3wt06Rc˜@±q4wœE8b‡‘T‡‡U@o`YXGPFX94@87VGh48bQ4@d7vGeT65444’8I‡654eA1Q`2ƒ T T T 0 0 0 0A 0 7b G7 G ‰ ˆ † „ RI FW T 0 G A2 0A2p G 0 5I 2 ‚1XIeF35e0XGVF35w `YHGPFP9c@d7vGp“32V03544X2oc8bgX2’P01VIs4€zYº7cqeT@8I6G6w1“32@P(3Fo48bh˜6f80vGjvY432@P(3F’e06GP0c˜16IÃebXGF u FW T G 0b F 0 G F 9 5 I2 9g 5 0q 0 0 0 G T T T 0 0 9g f w ‘T}‡‡v1tI~SEebwe0@4u0 u HYP9@4P2|36GhmXTo1XI8FHG4416Iu1±ql@ec8b˜VI£™ u eI@mHG”6Tj4~35c8bXG6F42 ‰ ˆ † „ 5 R G7 G GI2 97 F ¸ 0 f F 2 FA2 w 0 IW7 0 G T 5( F 0IA 0 0 GpW g Š T7 0A GI2 0IA G7 G 0 G F F T w F G ƒ  FW TRI F7 fIA 0 G j4j™ h ‡r@xzY373w4@43T—4x§ ‰ S48bmcep85BPTXG”e0@37XT48bq`2Ä1`YHGPFP9c@d7vGÛ@hP2@˜14m48bhP2F †…—W„ƒR‚f›aPTF l¨ ©dbPwcb u 1tIae0SzÇ@º763legd48b~SEeb8®˜680VG—vYo680SEX2˜@4`Aj@PI8FD3@€@3…P2F ù… e ˜ € ~ ý F 5 R T7w 9 Fw0 T 0 G G7 G T f T T A G7 F fI „2 Gw0q2Iw 2 G 0 T2 w9 w 5g G2 5Ig F 27 F F T T7bg 0 0W 9pI T ƒ 2 0q2 F @PI8Fœ8735cPAeFc13IÇ@37Pw8FDG@87c€Sd7VG@thPf…@mPT~XG~@ ­7 3A60PÊeFSE4h3fdtoPA41cebm¹G}‚1PIeF6G6w1“13IXw4W A27 2 Tp FA GIW ( G 5 5 ‰ ˆ † „ 0AI f 0 G 5 R 0 G T F 2 w F G G7 4Ss1XIeF4tXøch8b@l#4X28F“32V03TV043gV0u‘T‡‡r694jhqc8ba1tIs3A4VCPTeFS87dPTQ1XIeF35e0XGVF35dPT4ebwS48bG 60e0p0 u 1H°4Û4X(|©4X28FDG3wleg4`2’@c@º74–sDiX04h13I˜48bQP2dPTjœ©vY@EX21@˜Pfqc8bÄr™SufGDRP0qc8bG T ™ Ž F ¸ ( 0 T ƒ 02 9g ý 9g f w 0 G F Fi7 57 F(7 F 0 G RI 9 0 eIh1‚1xP0XFpvG4eh3T4Xp@S€621P(6F0 #4ebv@}X9@@`YXGXFP94@87VGq€62e06544P2h1tI3™SEeb˜3Te0Sd7vG˜XT4}{ G 5I ™2I 9 Tp f 0 97q 0 ý 0 G RI 97 ™ FW T G 0b F 5 R G7 G G T Fb | 97 5 z x ® ÿ }X9@1tIR {y}4 ­ý w ® “†Ž ™¯­ HRpHY41…4@7 F 92I A2 HRhX04@87VG©HY89“32V03544X2˜V0@d†T}‡‡q48baSEebÛ@P0c˜@c…id@tIǙ ¯ ·jÄsy¡1VIt5h6041dSaT u PIX91tIR F 9W T G 0b F 57 ‰ ˆ † „ 0 G G7 G™ 9g f7 0 5 R   ¥ ˆ f R w2I G7 9 XGde0@37hdb4epo`2’‚1XI8FS˜ef1tIvR4@e785”`YXGVF@º7X9˜XfeF”…€q3A30XÊPF@411@P74wl…@z7qHiVF85S˜wc8b’4@7 F T w 7 w T ƒ 2 G7 5 T2 G 5 F T 7 YW 972I( FA 0W 2 w G7 f 0 G A2 3™“324462t6g44X2vXYV9@z7cP2PFBV0@’VT18IDG3w1“621P(6FÈeTXGc@1P9S1…3f80vG—VY4621P(6Fre06GX04˜@3I|@4‡HiVF85S˜E¹Rƒ G 0A 0 0A F 5 0 9 57 5 0q 0 0 FRI 97™ T T 0 0 9g f w 7 T7b G7 f 7 3‘T}‡aw4ebÄS…@PI8FjPp1VI€62P0VF4@87e5‡48b’1pHY41~4462t6g4q`YHGPFX94@d7vG~€62e06544P2œE8b‡80SmÉ ‰ ˆ † „ 0 G RI 2 G 9 T G T2 G 0 G 2I 92I TA 0 0A FW T G 0b F G7 G GI  Š T # 0 2 I W 4X2@ 7 TF G@ 2 ® 1Í ™¯­ PTF § ¬ 32cb u r™HR”Pù9e!¹–D•©ž )eED›dùDEWf›’XThoH° %¯g UlSy}3Aˆ ™ c4ªA 4i1eFp lœUq’q@6f…ž7 le0– œo™˜—š5› § ‰ @3™• —–D•“hd7˜4@1s3II25T7vG ¸w 0 F f “ • ùd ˜ F ¥ 85G 7 f “ 641XI8FG T2 „7p 0 9 G 0 FA 57 F 5I 5 T 2 w FW T 57A2 T 0 G G T 0 ÌE€3¥v@P78F“32V035tXø4gVYSEP24VA1p1tIR 3AV04p 1XIeF35e0XGVF35g`YXGPFX94@87VG»eA@44@d7vG‹48b#e0Sd7vGQ0 u e065m÷ ˜ ”’ ”’ ‘  ˆ ™'—•QW…–•“xh‰‡ 8†ƒ…„ƒ ‚ ‚ d IÌ° s i–s ts ˜ ” m – k j i — €rWvu2hrq'ponW§'l€™‰‘ …„ƒ…„ƒ ‚ ‚ s i–s ts ˜ ” m  ”–Œ ‹ ’r‘Gvhr™¥pny…Ž‚QAQ¡Š ‰ˆƒ…„ƒ ‚ ‚ Vôy€ñænwxví ‚6ïzò9t©sppxvBh#©Çf Aepã u õ ò rqï i ï êçgç ð ê ä × c‡6y}À3ÒÛrÚvcÇgir~B”mràgc1`±ÖÕ b Ó a Ý ß Ý Ü Ù ÙØ ÐÒ× Ò àÓ ÔÓ Ï Ù Ö .

 2 0 FqRI G FIg F G T FA 0 ‘  G 5 G2 T2 w 0 f T G7 4‘Ž ¯ 1PIeF6G6wjoP2F u P0—È@“P21l}PT4eb~VT43pVwPT4a§|4XI8FSe7m“@87VG4@3Ia˜@eIS˜eI6530Ê eIw1(PT@SVq6580“P2qeb3Thv4eb@˜P2@˜14PT4eb˜X248bHGF u 1XI8FcXp1eIvc8b‚@)ef@4”h1eIu8I462tc@Eb G I 97 G F 0 f 0 G T7 F7 fIA F G F 2 G 9 T 0 GRI 5I2 0 f T G T 0gg7 G7 I T G F 02 0A F FW T A27 0iC F F7 fIA w7 0 G ' T2 G7 F 5 œEb u RS}˜Vf3580xX2o6A4E6Cc}PT˜`YXGPFX94@87VG’4@o3AjcUXTxP2@˜14v60@EVg`TD„hPfeFÀ’3c1PIeFSqXfj±iVI4g „g7 w T A27 0 0 G GIW RI 0 0 0 G Tp 02 9p T 0 T2 w G G 5g 5gg j@#60@7EVgÃ4@’hPfeFG…c8b€8b@lv@˜8TDG3wt6ø…48b§XYe9c13I4Sd7XGcePfVF˜654PAVF413IQ8Ixe0Sº7VF41eIc4@7 XT’XG…3TV0@7zÀdbcepT `2ƒ S80c1VI‹806GV035VwPT4XA˜6fe0Q48bG u PIX91tIR XYX9eF@e7eT6T303wcoS4gcoœE8bp4@PI8FjPp9 F F T w w  Gp 5 F F T 0 9 5 02 GI2 IA G7 G T2 G „ T ˆ † ¢ R G 0 fg 0 0A 0 G f R T T 5 G † ¸ Ÿ F Tp RI 2 F2 0A œeIއB~@Ii“32h41XI931Dq4p4eb¡1eIt5uh3fe0VGhe04epS876580HGPF9 ’UhP2€3AV04h32e0fG@h1XI8FHG4E6Cc¤' 0p F2p GI2 57 T2 F 0A F T Gg7 w F GRI ( F2 F 0W 0 G G7 G T T S4€P¥4c@c”0eSc1PIeFXG42EDC4i`YHGPFP9cW@d7vG43™6580c@4dbXT48b‚Sw4X244X213(l”4ebœx3Ae0Sd7vGv0 u ”'  FW T 0 G 5I G G f T7 T G @`YXGXFP94@87VG‡hPfFeU1‡PweF@8Icgh—VY@S7qeI…6Ae0e565t06RV0}3ThPf8FDG˜@eI}PT˜`YXGXFP94Sd7VGÄ@x1XI8FHG4E6Cc~XT4}{ 5 0 0 f T F FW TRI 2 F2 0A Fb 6†T‰gh‡m4eb£@i21XI8FcXp1eIr“32X0eFT4S878504eb”@IiXY4@4c32l3g4u`YXGPFX94@87VG4@37PwVF35h4€œE8b30V0~§‘ ˆ ‡ „ 0 GRI G 9 T G 2 G G 2 92I TA 0 0A FW T9 0 fp2 G7 G T 0  h g T # 4A p W T P2F ˜ ” @ 7 0 7 5 2 ® ¯ ™¯­ PTF § ¶ 324b u r™XR‡Pù9e!¹–D•xž ÁeIe–"…˜Dùq2›”PTF’·Å ¯ @sy6A¥ ’™44µ2 @teFI i UScq@”…ž3f@V0 lœ4—š²› §½˜‰ @3™• ¹–“D•G ©d7vGw4@1s3II25T7 ¸w 0 F f “ e Àƒ  ˆ e5G 7 f – “ PTiÍ ¯ “62˜6f80S87—˜8Ip1XI4@Eh13Igh‡u4}{ F G 0 G G G 2 F27g f w ˆ ‡ „ 0b Vôy ñ ¿nví ‚6ïzò4t©cp½s·zìBlç ©«» ºpã ¾ w u õ ò rqï i tq òg ö ¼ ä 3€62e06Te04e0S˜@PI8FœEPgVFeTPTc‚@B8T“4@˜@“@37EXC4@P(eF£XR}604@d7vG1l˜XfE@z7Xw8F6GS874d4B1UX089eGHGPF£@BPTF G 5g 57 2 G7 FARI G2pI f7 G2 w F2 T F w2 5Ig F9 w 5g I2 5I 9RI †  ¯ 1XI8FXGc413Ij3™6516q0 u m|3h3fP0c1eI4h1XI8F6G6w1“16IPw4a8Ii6A80SP739e0B6T4Pp@S“32@P(3F4ebe“˜6AP04c3p3w—I 2 FA2 w 0 I ÷ T 9W 5g 2 0q2 FW G G 5 0 97q 0 0 0 G YW Fg w 0l‡8IÄ4432e0idbPw4b u PTjœavY@EP2@@qXf}48b‚d“—‘„ƒRe28¹˜†  ¯ @PI8FHG4416Ia`YXGXFP94@87VGUc8b’S48be0@mÉ W G TA G F Fi7 57 F(7 F 0 G ù…  ù 2 FA2 w FW T 0 G G7 G GI  G7 F G 0W 9p w F 5 0W w 5 ‚6A80SEX2˜Vf35e0al’PA4@3IXG‡V01tI6Rl~eT3T60jVI4g 2 G 9 T 0 G 0A Gb F f 5( 7 w T Gpg f w 7 2I ™ 5I 0 5p¸  0A 0 @PI8FjPp1VIx48bhY±VI85VG6T4i“1X(˜€8bG u VI1©Ãdb4Vpi35e0c4h13I©k1B@V01˜Vf354ebVG4E€@±Y373w4xc8bG G IW7 T( 5W 27 0 I 0ƒ ¶ 9 G 2Ig 0 0 G 5 0W9 F fI2 9Ig 0 GRI 5( 0 5 jp1tSU14X2eF4xAc@h†¸35@±q}3T@Sd7vG·Sº78F“624u0 1lc…iv4eb”e01tIDRts@P7˜@4—XY1lv4ebÈSx8bG u eI1m1e(Sº79 vY35@q’t§zYqf’e06548bp604P2VFieTT6P04ebVG1I tszYqh@PIeF6580HGeF35xc8bwc1PIeFSz7XwPF44SU@37Pw8FDG@87c–P2F 0 7 0W 7 0 G w T 9 5 0W 7 f 2 w 0 G T2 G 9gg7 9 w 5g 6™351Dq0 u ~B3µeI6530¹Êj214sP9SU1tIR h ‡y@dz g £ ‰±³ ²°4¡¥®m604P2VFu¨†Í ¯ “32h3fe0Sd7€Ä@ve0c32e0”48bhP2F 0 I ÷ ´ „ I 2 9 7 5 g Š T 7 ± ¯ Š w T G 0 G G ¨ R I T T 0 G `YHGPFP9c@7dvG1tI}“32X03FhcVpTPT˜1PIeFXG4Ac13IPT4ebiXYP9Sz7PweF6GV013I4}yjz l“z ­ ‚U1uz ­1z ­ ÛHRycHYP9@37PwF FW T 5 R G w ˆ F 2 F 2 w F G 9 5 0b {  ¬ ® ´‰ 5I ¬ ® ‰ F ® 9 v„S87c@E€¥’1aHYe9@374P2XF ­9 4241l’j14b8HGF gu VI1iPA4@I u  i4@7 œ “ÛdVY@4P21@˜Pf}V04œ4’@tIU304X2eFT G 5A7p 5I 5 0 A pIW GpI u 5( 9p ‰ A2 ‰™ý 57 F(7 F 5pg 5 R w |«P9S1tIR ªƒUÿ4 ­ý w 97 5 z © ® ® 1  ™¯­ ef1tI‡4ebi8IG 5R0 G 6AEXCcjhtWrVGcew‘Ž ¯ 1XI8FGXc413Ir3TV0@z7qbd4epy1tIBGS48bm600e~0 u ‚1PIeFcPp@eIBPT4ebw4X28FDG3wleg4`2ƒ 0 FAI f 0 Tp f 2 FA2 w T w w T5 R 7 G T ™2 G 9 T F G ( 0 T ® @d ™¯­ £4¡ 8¨œ Š`“z ­ ‚‰ ° Á `Š“z ­ œ ‰ Á “z ­  ‰ § ¨ ®V ­Š  ‰ ¢ ® ´ ® ® A£4¦¡…¥rŠ1z ­ ´‚‰ Á 1z ­ œ ‰ ¤ ¨ ®V ­Š œ ‰ ¢ ® ® £48¡“z ­ ‚‰ ¨ V® ­Š ‚‰ ¢® ´ ´ 2 1PIeFGcPp91VITa048bG4A4EC‡04@IodbXw4b u 1tIR T 2 2 F 5   @Ì° × âÓÒ×Ü× Ô ÐÓ ÙÚ ßÝ Ü Ù ÙØ ÐÒ× Õ ÔÓÒ Ñ Ð Ï #rÄÇBÇávy”ràۀ}ސ3ÒÛrÚ”4ÇñÖxBÄv”m’Î ˜ ”’ ”’ ‘  ˆ ™'—•QW…–•“xh‰‡ 8†6‰ˆƒ ‚ ‚ .

 FW T9 0 fp2 5 R 2 FA2 w 0 G 0W G 0ƒ G 1`YXGXFP94Sd7VG£@z7XweF35h4€}1tIp1XI8FHG4416Idc8b‡t”eIo32S@d7UXTF XT48bx4SwÞx Æ ®  ­ý   Æ S48bv30e0“@87@41odbXw4b u 3Al1PI6916qc‡e0@4—4eb6Ghpdbew@˜`f…„˜Xf}cÍ F G A27 ° G7 G G2 57p( F 0g 0 0A 57 TAI 0 f 57 0 F {  z x ® w G7 G ( G2 57p( YW 0 9W 0 F ‰ ˆ † „ 0 GRI FW T G 0b ƒ  c{yl ­ý GSEeb”cP23F6080“@e7@E1x“…3A4VbPTXF4@d7vG3T}XT~‘T}‡‡uc8bÈ@q`YHGPFX94@d7vG”“62e035c4`2j  0 f F2C ‚Vb3Th‡80HG4Ee7 2I 0 G7 0A w7 T 0 G ( G7 F 5gg7 YW ‰ ˆ † „ G 0q2 w 57 ‰ ˆ † ¢ 0b {  16T1XqeFSSXqVF354w30@EVgwc8be4X28FS˜PfjzieI4c@…“m‘T}‡aq8Io3A80vG35@€@3Ide0@m†T}‡Bw4}1° û w7 5gg7 ( 9 R 0 G F 0A 0 f T F 0 T2 w 2 T F ‚db@“VI44@e4P2F u PIX91tI‡48btP9@876G4‡h1eI’X23AV0354XAeF4@3Ia0 u 1XIeFeTc3pewXT4A Tp F 5g 0 G ƒ  G 0 FA 4@PI—6qV04‹48b»`2¤@±Y7 u “32e065£Hø4–7 P2F `YHGPFX94@d7vGuSc6g304@3I‹48b©4X2˜S4…i0 u 4…4…6G0 Õ FW T RI G w2 w 0 G 0 F f7 I2 Tp Vô€ñ ¿o—ᐩ‡”ylhÔh)Ò ©eт6ïzò4t©cpАe’Gç zöò Î 'Ípã y ¾ w u í çì êç ógç Ó ê í ë ê q õ ò r qï i çì q é iÏ ä  5 ( G 5 A 7 p ¥ 5 I 5 0 9 0 ( ¬¨ ® ´ ‰ 5 I ® ‰ 0 0 A 2 7 ‚ebG u eI1aPweFS87c@E€}1U@374P2XF6G1d0 u z h“z ­ ‚}1U“z ­ 354ebXG6Fo4Sw° ¨ Æ   Æ XRed ¯ F syxX2mX04˜S4…i1tI—1° ¨ Æ   Æ @tIVT3T4c3A44c1t”3080“@e7@E1@4eœ2 c}˜6f80VG—vYv1Hq8F6G6wt06R4”304X2eFT ¥ ˆ F 9g f7 0 5 R™ 5 R 02 0A2pIW G2 57p( GI IA T T T 0 0A w ®1• ™¯­ Ç A27 | 97 5 4S&‚X9@1tIR ° © Æ ® U4  ­ Æ ®ÿ XTi“ ¯ eIq1PIeFSz74XC4—˜…3AV0P54€6¥e05 F G 2 G w FAI f Fp 0c8bw‘T}”aq1Hq8FDG3wt06Rcd@tIdœE8bh30e0h§hjd ¯ sysebXGF u `YXGVF@º7X9˜XfeFa8THG˜e0@4Q4@…”“X°jÄsyˆ G ‰ ˆ ‡ „ 0 0A 5 R G7 G T 0 ‘  ¥ ˆ 5 F T F GI2 A27  ¥ 0 F f7i ˆ ‰ ˆ † „ 0 0A5 R 2 T FA 0 G 9 57gT T T 0 G5 R 2 T FA 0b cP2h@c’Ä3‘T}‡‡@XqeF6G3wt0DR4Û1tIa1PIVFeT46pewXT4y48blPT69P0P9S87@EÈh3f80vG—VYre03Tc8b£1tIa1XIeFeTc3pewXT4B4}{ z“32hefVT3Te0VT@q`YXGPFX94@87VG‚@37PwVF35h4€dc8bpP2‡X01eIh4q±YP74g G 0 T7 FW T9 0 fp2 0 G F 9 5 I2 9 eT@8I6G6w1“32@P(3Fr4ebS…3f80vG—VY4621P(6Fre06GX04˜@3ImÈ1tIœ3™6548bvG4E|SP04@87VG‡HYP9@e785c6p4‡HY41ÈPTa1PIeFcPp@eIT 5 0q 0 0 0 G™ T T 0 0 9g f w 75 R 0 5p¸  9W T 9 G 02 92I F 2 G 9 0 G 0 c8b”e035cb u P069ewP56F‡XGc4i48b©1XY69e0PT6F3wV04‡@—eI—•Ë@EXg3F4X2eF4ic8bB@#1XI8FS37jXI’ʸ6G6we0V513IVwÉc8bG w F2p 0 G 2I 5g GI 5Ì  97 w 5g 0 G RI 2 G w 9 5 ¶ 0 6T30@P74udbXw4b u 694jha1PIeF6G6w1“13I¹wD„Xw4jXIeF65tB4ebÇ@tIÈVT43p6wjBe0Sz7r1XqeFXGVF432V04ebvP2@1”|6†T}‡‡„ w 9g F 0AI f 2 0q2 FA 0g 0 G5 R 5 wI T w 0 T T 0 G F7( ' ‰ ˆ † 5 R 2 F2 0A FW T 0 G G T2 w F ‰ ˆ ‡ „ 5 R FW TRI 2 F2 0A Fb @tIe1PIeFXG4EDC4d`YXGXFP94@87VG4eb”ebXGF u “32e0VGPTVF416IyPTB†Tgh‡r1tIa`YHGPFP9c@d7vG£@i@PI8FHG4EDC4BPT4}{ 02 9 „  9g f S4@P74`gt¢DiX04˜@3Iw 0 G F w F2p 0 G 0 T F 5I 2I 9 Tp f ®Y27 57 0 G F Tp 5p T A27 97 w 5 c8b‹P2iP069ewX53FaXGc4”48be4XAeF4X2a1§@eP0XFwvG4e‡c€S”e0@˜e06548brXRÇ3™c1PIVF44egQ4@’@EPg6F4P2VF4g eb@t ­W 3T4Xp@S€621P(6F0 È4ebwSvP9@m1`YHGPFX94@d7vG‡@z7XweF65˜4€x1tImz™SEebx3T80œd7VG»1XI8FHG4416I…XT4}{ GI 0 97q 0   0 G RI 97 ™ FW T 9 0 fp2 5 R G7 G G T 2 FA2 w Fb Ç A27 | 97 5 ° ® ÿ oc@&§X9@1tIR Þx Æ ® }4  ­ Æ ® 1“ ™¯­ XRpHY41…4@ÄXRaX04@d7vGqHYP9@37PwVF35h4€}PT’jc8b6Ghi4X24dbew@˜f F 92I A27 F 9W T 9 0 fp2 F AI 0 f ( F 57 D„hPfeFxwh3f80vG—VYQdb4ep‡1tIt1XYV9@z7X0”c1u  †ÍcÇsyÛ3™eT@—eI`5lcP0cPg8FHG4ev@tI#1PI4SE˜@3I‡8THG‹4@7 0 G 7 T T T w T 5 R™ 5 9 Ÿ   ¥ ˆ GI „  9 9p f 5 R 2 F27g f w F A2 “    ¥ ˆ RI YA T 7 f R 0w2I G7 9 R  ¥ ˆ F G T 2 FA2Iw 0 G G ( F 5Iww °·c|sB}@sj48pvGoè1VIt5q34@uSdT PIX91tIę ¯ ¯ sB§P2e6e0Q1XI8FXGc413˜48b…8I…4P24VA136@7 ™2 w FW T 0b {  T T 0 0 9g f w 7 YW 0 0qI( ‰ ˆ ‡ „RI T 7 0 T2I ‚1XIeF35e0XGVF35a`YXGXFP94Sd7VG4}’…3f80vG—VY4621P(6F80DGP04h16I”e“h3A4V235@Ì@’†T€”‡t@y6GV0”y654PAVF41Ÿ ¯ Ì°  × c”}Ý ©’urlB’—}}¤Gvg3ÒÇrځvcÇ× ’dtc—×qovâÈá—±ÖÕ b Ó ‡ ß Ý ÓÎ àÓ Å ÔÓÄ à Ý Î a à Ð Ü Ù ÙØ ÐÒ ÓÎ ÐÑ ÃÓ ÙÒ Ö s i–s ts ˜ ” m – k j i — €rWvu2hrq'ponW§'l€™‰‘ …„6‰ˆƒ ‚ ‚ s i–s ts ˜ ” m  ”–Œ ‹ ’r‘Gvhr™¥pny…Ž‚QAQ¡Š ‰ˆ6‰ˆƒ ‚ ‚ .

®° Ì1H° —¯­ Ý Æ Ý Æè Ý ² ¦ÆÆ ²§ ‰ ¦Æ‘è ² ¦ÆÆ µ çÆÆ x ¦ÆÆ ²§ ‰ ¦Æ‘”çÆÆ ² µ çÆÆ x ¦ÆÆ •§ ‰ ² µ ¦ÆÆ ¨ ¦ÆÆ E§ ‰ ¦ÆÆ 0 7b @zqEw0 u ™34@PI8Fœº739V0d`YXGPFSE€3¥4X2’P56F48bi4@’vii 4432l4m~60e0 ­T hef14¹2cuHiVF85S˜”4@r18IDG3w@v4P2VF4”4@7 T2 G 5 97p 0 F 0 G A27 ® ' FA 0gg ' T 5I „æ G7 f A27 5 0q ( Tp A2 Ý µ •§ ‰ ² À¨ E§² ‰ A2C ™ ¥ˆ G 2 5 5 9g T ( F 9gg cE‡0 u j”zX° ¯ syh8I’1PIVFeT46p3wV0vP04hPfVF”4P2—XYc4”'  0g 5g G2 5Ig F 27 F F G™ 5 9 0227 f F 1`YvG35l1eI4U€Sd7VG@thPf’@XT}PT4ebc1HYe9@37P0Ÿ 36544@˜}PTc8bG X2peI6530h8Ih16g80VGahPfeFo4@q4P26F@EVgeb6T˜d4eb~@o1XI8FcXp1eIm6G@c…iu4ebheI”3T 1V(35@€@3I~XG|XRF F Ê G I( T 0 G A27 ( w7 T 0 f 0 G T7 2 G 9 T w7 0 0 G G 0 0q2 w F •ED›—CåD@œdšTPT”jc8b6Gh»’c1“ ¯ 1PIeF6G6wj˜P2p6Ae0eTc3pewXT4B354ebVG4tp’XTBPT4}y…±®e0Sz7}PwXF1leW35l—“~c8bG ù ùä›  F AI 0 f '  2 0 F T FA 0 5 R F Fb { T w 9I 0gYb 0 X24‚6Acjd² Ü ebXGF u ­ eI6530ueIw1q3g80vG˜Xf8Fmc8bq4Se4X23F@EVgpeb6T˜m48bB@BHG˜XfPFm48bhP2z®e0462e0T F™ 0iC Ê G I( T 0 G 0 G A27 ( w7 T 0 f 0 G T7 F 9 0 G F T 6TP0eF65e0m1PI—zY|u48bX2 ­F 1eIV535x4‡6T3044—eIc”HGBXRÈED›rùD•P“D˜cœdšvPT#—48bDG˜f ”½ED›—"åD@œdšvPT”XGF T 5 9 7 { 0 G 5 0 I2 wpA 5g F F• “›  F AI 0 ' • ù ùä›  F 32c8b£ED›`ùD•P“D˜cœdš‰4@}4œ7 Õ @we0462e0a4eb’X2 ­F Pù9e!¹–D•”XT’j4eb6Gh‚@37PweF65˜c—i|XR43™SEeb~6T80S87VGT 0 G• “›  A27 ®i RI T T 0 G f “ F AI 0 f9 0 fp2 7 F G7 G G dbPwcb u iRÀ8ùœdšE§v“ â4¹Û¢1c1eI4eby3Te0XTeF@w`YHGPFP9c@d7vG£@”1XI8FHG4E6CcyPT4ebt@}604@37EXC4@P(eF~4}{ F ™  ùdã á –à b(p 5 G 57 FW TRI 2 F2 0A F GRI w2 w F2 T 0b 2 w FW T F G5 ‚1PIVF35e0XGeF65q`YHGPFX94@d7vG}PT4ebU1tIR 2 FA2 w 5 02 7 @PI8FHG4416IpvY@87VT3T306w4uPTF ¯ ¯ X2q1XIeF35e0XGVF35”c8beeI€‚X2@˜14’3Ajc~e0P5eF“32v4ebU35@Ì6Ie8Ir354VA1I F 2 w 0 G T™ F7 fIA 0iC G 0 0 G 0q w G 0 5 X2x`YXGcEcP2”eI”@mvG4es§²6™43ge0VG‡hPf8FÈ@35le4€a4ebsceI3560”8I”1}PT@Svq3580“X2oVb3Thdc8b}@7 F F2C F G I( Tp f™ ß T T 0 GRI 0W fp2 0 G™ Ê G I( 97 G F 0 f 0 G T XYV9@z7X0Ä1`YXGXFP94Sd7VGÄ35@u8f“dbPw¥Þ€œ7 Õ 1dœ7 Õ S”eIe3Ae0V535t0DRe0u6280fG@ÈPTĹGÇ1`YHGPFP9c@d7vGE@d1PIeFXG4EDC4A 5 9 Ÿ  FW T 0Y G F „i 5I i T7 G 5 RI F ƒ FW TRI 2 F2 0 9@z7XweFeTSº739”4ebvPTaPTc}hj² §Ü 6Acj’a1tI’eI6530…8I‡€4P2VF4ebQeb6T˜”4eb‡@ÈcE§² ‰ 3™1eI6G3w@#1PIeFcPp@eIT T w 0 G F Fb {  0iC 7 5 R Ê G Tƒ 5 T 0 f 0 G T7™ 5 0q 2 G 9 3Ac441l…u3T6044—eI4Äz™•‰ 3™1eI6G6w1ÄSº78FHG4P2§3A4c41l§“@UXR”P0c@d7vGqloeI§PAS87aPT’”zH° ¯ syˆ 0A2pIW 7 wpA 5g Ý 5 0q9 F F 0A2pIW Y27 F 9W T 0W G F T F ¥ ® “”zH° —¯­ E§² ‰ µ ¨ tz³E§² ‰ ´ PTeP74eef@tIUPT4ebÄ@vG@E416I432@1˜@4wc}yj· ¯ ly|@I Vf1tI‡48bpX2eº74eVf1tIR F 9p 5 R F GRI 57g Tp 0 0(I fIb 0b {  ¥ ˆ R 5 R 0 G F 9p 5 d®6T4@º7c`gD„hPfeF‡I u @H™@‘0— ­F 6916qX0`9SI u ”a8I©‡B}4ebE@u1XI8FS˜PfjzieIc4@Qh‡Bc1}304c3Ae0B—…imÉ 02 9 0 G G  0 „ G 7 G ˆ † ¢ 0 GRI 2 G7 F 5gg7 ˆ ‡ „ 5pI wp 5 G 0 Ø µ‡ ¨ Š i‡ ² Ü 2 G7p 0 0 G YW •  0 f 0 G iC A27 w7 T A27 0 G GIW F w7 Fp 0 F G7 1PIeFSE€3¥a4ebx“v—š½˜@V–§Vb3Thdc8bxc’4@w30SEeg’4@v˜Xf8F’8b@l’P26A30@4egPTc—6¥}PTS48bG XAeF1œ5 ”…8bHGF u X2@˜14}PTc8bm651±qd{yþ—Ú4@þ½ªv×P24ebXGF u 4P2—HYxP2@˜14d30@4egT ( 7 F7 fIA F G 0 I Ÿ  Û x Š x z A27 Ù x Ø x z F ( F 9 F7 fIA w7 „DhPfeFw80HG4E”a6G4Vp85vG413ImVGVTP5|”@V0432V0w60@Eeg¹TD„hPfeFw4eboP2p`YXGPFX94@87VGe446C4m4”DGP09 u mÉ 0 G F2C 7 w T2 w F ¸  T T w7 0 G 0 G F FW T 02 0A Tp I  Ê w7 5gg7 97 G F 0 f w7 T A27 0 G 0 G GIW T 4eI6530…db@“VI44@XT@SVq35e0“P2’eb6T˜d30@4eg…4@‡hPfeFac8bx8b@l@7 VT3T30—eIc#dbew@˜d˜Xf8Fwc8bX2#3Ae0S87654321ue0@m41XI8FcXp1eIdP04S87P5VF3T4”SEebu30e0€S87@E@Ĝ£wd“±ù—… w 5g 57 f 0 G 0 G F G 0 0( 57 T2 G 9 T 9W 0A G7 G G2 57p( •š› š TX4}p h g  Š¨ a@s3A44c1tdPTdS48b‹1PIeFcPp1VIy@z7XweF65˜4€…p1zqE|P9P9F u ‚eb6T˜˜30@4egT Fb{ Š T7 0A2pIW F G7 G 2 G 9 T 9 0 fp2 7 0 7b ™ 0 f w7 †…jrTh1VI©1I ‡Bx7 1eIt5©680Se73461Ûo680vGjvY…VY@E@PI8Fœd7VGp‡SE8bw380“@e7@E1md“±—vXT4}{ ù¹ Ö 0 f T 2 ˆ † ¢ f R A G 5020( ™ f T T 572 G T 7 G7 G 0 G2 57p( ùš… Fb × âÓÒ×Ü× Ô ÐÓ ÙÚ ßÝ Ü Ù ÙØ ÐÒ× Õ ÔÓÒ Ñ Ð Ï #rÄÇBÇávy”ràۀ}ސ3ÒÛrÚ”4ÇñÖxBÄv”m’Î “ IÌ° 7 X   ë F  &û þ ¤ ù 7 # @ï # # # @  ¨F 0 3 vXY R3 C¨”ïRC3‡C#‰&ÿGû”âR†&C#)y·"3Ëù 9„çþ'ýÊü°Xû ‰µR3C¨”# ‘ïõ R@X C3ï 7 #AX&8&$"¨'Bƒ©47 #P ø y Q ø †&# ‚øQ & 8B# Y ”¨· 87# R3 ‚7 ¨ ( ø  ‚C¦ËC#C38B‚'B£ê#¢ '0'YC#9!¡X B 7X©A7¨¨ ây§'B9C#87AC#vI1îQ©²'Xõ ë §1B"#¦ ©"#¨ ("¦v™uÿøRTC¨3©C¦11ï)3u†F…)#(°”§§©9C¦ü#¨#'Dï&$  çþyÊü'ýC3°4‰þ „ú¦þ¥¤ù ‰F¥ýB ü‚Aø "#sR÷7 3C¨”B ïRC3!AX8&$X "¨'Bƒ©4™©ƒ"3‚'Bê#¥0"#B 3P R§©C¦4(C#PCï‰&'B"#E¨ö& ™ ‚$ô qó §C¦T7 'òñ B $1¨(°R"&RP§Rê3u8F”†F'#©‡I6Êë6™3C¨A"&9'7"#'B148í6'@#©R©RËC¨H7B"3Yì 3 A$1¨P(9°8ð3ïX& "ìhIR43H7##'B 9)ë¨ °I(1îBï ê3”rv"#F0¨ 8X”# & C¨”©y(‡PP©C¦H#B@H3BF C3 AH‚R& h‰¦3C¨PACï"&6©'@¨ §#¦ ©Ré¤X ì "3©H X3ì B&H B 0 # #7 #ì & @ B # ¨F &  X ¦ ¨ B .

2 1 0  FW T 5 R G w ˆ T GI2 GpW $ 5 02 F 2 FA2 1`YXGXFP94Sd7VG}@tI~“62P03FhcVp~@4~c4‡%vY@87VT3T306w4}PTx@PI8FHG4416Iw T FA 5 9 0 T 0 G G7 f 9 0 0( 5 R Tpb { 2 w Y27 YW 9W T2p cPp4@e7”@e785DG3wleg#4eb}3™3T60PwVF85S˜”@e7354621q@tIp4€}À‚1XIeF35e0XGVF35g€Sþ“#P0c@d7vG44qXTF AI 0 f 0 G™02I 27 G G 5( F G7 f ( F 0 I( ž›– RI T FA 59 0 T 0 G —4eb6Ghu4eblS41#@48b3580œz7e0@mXT’HiVF85S˜p4X24e2651±q1aE@~@~4Ppc@87È@87e56G6wtVgu4eb’¹Rƒ ü 63T30XweFe5Sqa4ebG G7 f 0 RI 0 97q 0 0 0 G 2Igp 9 T A 0 0A A27 G7 f w T G 0 0( G7 G TAI 0 @B3TcPp@S“321X(3F~48bh1l4eHY39P0@eIp462t6g4qc@B3T60PweFe5S˜pdb4ep~80œ8735c321SEebBcj4eb6Ghf 5 R G w ˆ T A27 5 02 GIW 57 @tI…“62P03FhcVp»4@yVY@e7eT6T303wcþ8b@l‹e0@hP7eF6580XGVF35Qe06T4}¤36T30XweF85œqx“@P743pVwP56Fw c@7 w 0b{ G7 f G2 9 A2 ™ 0 f f T7 ™ 0 f f T 0A w F Fb { 97 5I2 F G7 f ( F 0qI( 0 G 0 @XweF85DG˜h—VY@’@PwVF85DG˜h—VYe6T44Pp694P2wPT4}o—@˜ef@4uXTQHieFe5S˜©4P24V235@Ì@p48b–624b u T T T 572 G T 5 R 9 G 0 F 57  A27 T¥ ˆ F h3f80vG—VYoVY@E@PI8Fœd7VG‡@tI|@z7Xw8F“32cPA”e0@xÌH° ¯ 4@’1“ ¯ 3€yQP2pº7VF35e0XGeF65`YHGPFX94@d7vG”4}{ ü w FW T 0b 6†T}‡Bd1Xq@eI”8I…3Ae0c}4jc8b6Ghf ‰ ˆ †¢ 0 9 T G Tp TAI 0 RI FW T 9 0 fp2 0 G GpIW7 T2 T w2 w G2 5Ig F @o`YXGXFP94@87VGy@37PwVF35h4€hc8bac1l@d41XIeF4Xp39416I‡“@d7vG1l˜XfI u G u 874h0 u c1PIeFSº769e0he03Tc8bG 5A T2 G 5 0 f 5¸  FW T 5 R G w ˆ T A27 5 02 GIW 0 f 0W  ¥ ˆ™ T w F G ƒ  1eIE~1`YHGPFX94@d7vG”1tI”€62P06F˜jep©4@#VY@e7eT6T303wc©8b@l”6T˜16I3wl˜ÌH° ¯ syÛ@V0@z7mPT4eb#`2”@`YXGF H„@E€6¥p@U6T˜@3I3wl”@X° ¯ sypX2i`YXGXF@E€3¥cP2x416I3wV0”4eb1—™@˜ef1cXT©x624b u @e0@˜Vf354ebVG4E¸ 97p 0 27 0 f 0W ° ¥ ˆ F 97p 0 F A2 T 0 G 97 5I2 F ¾ 0 ™ 5I 0 5p 6˜Vf14tP9@’@a²—PDšf D#Pšv48bUXT}4Ppc@87Ç@e7856G6wtVga4}j T 5I2 97 RI …›ƒ ù  e 0 G F T FA 59 0 T 0b {   TI T2 G F Se01legc@87e5U8TXG’ebXGF u 3Te0ceh13I}HGF Gp f f w ™XS†0™c—™SqVf14PTpHieFe5S˜w4ebo624b u ef1c2 œ u8TXGXTxHiVF85S˜”y@4Xp4@87Ç@e785DG3wlegd4}1°  F 97 5I2 F G7 f 0 G 0 5I Ù F F G7 f 7 RI T FA 59 0 T 0b {  û I2 ‚2 u 4€ƒ 9P390 u V0@ihef14Qi eFe5Sq–c@XF4@e7U@87e56G6wtVgg62300 u 6Gls1XI8FSP739e0q48bej1tSi8TDG@½7©I u { 57 T 5I2 G7 f A27 FA 5 9 0 T 0W 2 G 5 0 G GpIW7 w R ! À  š ›  ù ƒ  "Aœ£` ER½• – ˜ “ ˜… e ù “ ù • ùd  ù À “ ù • – hP“e€Áƒ@!V–ÄIe–½D•EuEÁdsD›EËqI3–uEÁdi8ù@½–…˜D`ùf•…˜D¢EÿdeSE8bm@z7X039vXT‡¹Gw µ @¼ef14¹2cµ æ  ù • G7 G 5 w F ƒ RI 5I „ 7 G G 5 F FW T ™ ¥ ˆ ƒ 0A F2(7 f fp Fi7 f RI 0 97q 0 0 F  F …8I‹3Ae0Sº769e0aPTo`YXGXFP94Sd7VGÇj±X° ¯ sys`2”S448pHG41@˜ncePfjœ˜ShcPp@S“321X(3F‡eTXGÈX™S†0™È™ R@Ia“Áƒ@!V–IV–½D•E…4eb8IQ3Ae0Sº769e0uXT©`YXGXFP94Sd7VGÄc1• ¯ 4@o“ ¯ 3€y€`2o—ÌX° ¯ syhP2s321Hq1( “ ˜…  e ù “ 0 G G G 5 F FW T™ A27 T¥ ˆ ƒ  ¥ ˆ F 0 F G7 œE8bGh8bXGF u • ¯ 4@e1“ ¯ 6€BqP2p621Hq14@PI8FHG4EDC4i`YXGXFP94Sd7VGm48b~80SP739V0u8Ip6A304”0 »u mÉ A27 T¥ ˆ F 0 F( T2 F2 0A FW T 0 G G 5 G 02 u I  FW T5 1`YHGPFX94@d7vG}1tIR 2 F 2 w  “ 7 T7 Tp RI F Fb { RI 5I „ Y27 G 5 @PI8FHG4A41I6|•ED›—ùD˜ª¥Rƒoa@s6Ae04‹6280fG@vXT‡PT4}w µ @¼ef@4`2cæ €SueT€62e06Te046ge05 çÆÆ µ ¦ÆÆ e0654b u 0 ® 1ÌH° —¯­ ° Þx ¦ÆÆ µ ¦ÆÆ HRFo6A0444@t}XTm1eI6G6w1p1XI8FcXp1eIa4ebs£3A44c1t}XT18IDG3w1e87SEl@º7eFXG4X2dc8b‡304X2j A2pIW F 5 0q 2 G 9 T 0 G™ 0A2pIW F 5 0q G7A9 F F 0 G w F × c”}Ý ©’urlB’—}}¤Gvg3ÒÇrځvcÇ× ’dtc—×qovâÈá—±ÖÕ b Ó ‡ ß Ý ÓÎ àÓ Å ÔÓÄ à Ý Î a à Ð Ü Ù ÙØ ÐÒ ÓÎ ÐÑ ÃÓ ÙÒ Ö • °Ì° 4 3 BIï)H #Pï  R ¨ C¨ $C¨ ¨ F©7 $h ¨6 ¨ "¦ R&©ì 7653 37 ¨ "3‘AXP98ð(úRPC¦â"8#yIì 0A0ËPRïY "#”'B¨ (©"&#F §‘†B¦ AX)#‰&9CB °¥8$"H©©6©"¦R‡RPê&B 3 "¦0 â'&$c1C¨R'B7 3Q1”AP(¥3lCëRRB7 3 C¨C&¥"P'#$8 17"¨8$ÿ©# ”QP#F C¦©l1Xë”PCïP¥¥÷rS98$ ('& &ï # B3 7& X # ) 3 # HB 3 X 7 Xï .

$ï 7 # V 7 ¨ F 7 & 7  ï ¨ F # B 0 o F 7 & l X & ï # j 0 3 # X V # D i B 0 g @# # PRvR31¨C¨PEî©G©"¦%Rê3âPR·n1m"3•RP98ð—kR%†©1ïR&s8F”©e(ê3sh8BC#'(y8X1ë#Y RRVB C¨¥& 8$H I#)H†C#”÷B 8÷# 8C#RX 8&GR3C¨7 ¥¨ 8$(9(¥3$©ì "&RPuPC¦2RR˜f¨¨ µRRì3 3Cïë 10 R‘8÷B ed™ ¨ 'B©"¦)©êï1ƒR•X1Xy‚ú(Éø©{(C3P™!†F)#"8"#”©(µ'7'#IËRP68@C#)4CëP87R˜ '¨RƒC##Y RV8&C¨v¥AX8$ê#&¥¨¥0†$©°‚CHF## A0Q‡3 ©¥8$C¦2A0—ÿ# ©T·3©"#)¨ §P¤¦ 'ï– C$”F # B 0 ¨ @ 3 ¨ ø # ø 7 3 3 7 @ B ô Y B F F & # ì  & ¦    # #R $  w F2p 0 G ›š › š 0 G SP069ewP56FXG4cdc8b”œ§œ—š™˜f•D˜±“E•4ebi8IG 6T80SP739V0m1VIe565ee0@44u4}…Ì4‘Žc‚sye60e0 ­T 1eIV535U±ùœc”4eb’@qcP24eb@4”cmXTP939e0XG£3™6G6we0e5@3Iw G 5 5 0 T7bg 0b { ®°  ¥ ˆ 5 0 “–d 0 G RI ( F GI2 Tp G F 5 XT#1XI8FS87€62e06Te046ge0uc8b’@}—‘)ƒf•˜ ²“Ix4eb”S48bxXYc1”4Sz7hdX039VwP53FmXGc4u4ebpœ”d“—ù˜ cHG”S48bG F 2 G G 5 5 0 G RI ù… e À– 0 G G7 G 92I T2 0 f w F2p 0 G ›š e F G7 GD@½7i4}” ˆ ’‘Xu o4X28F“32V03TV043gV0‡PTaXG§324b u   4X28F6GV04eg6580“P2‹X2y4tp6RV0@z7”lwvG4e”§h63T1zq37P09 w R 0b{ ¡ ( G 5 5 F F 0 ( 5 G F F9 5 w 0W Tp f 0 ‘ 0 35@6q4BHGqdbXw4b @P069ewX53FrXG4cm4eb£@~30c R 0 02 F F ™ w F2p 0 GRI w Ât065e032h w 40 G A 2 5 7 0T G 0 A I f 2 0 q F ‰ cRf 3pIewX53F0 ~48b”474@fp eI@GPT691zq' 87e5 4395 4—˜˜1e6GeFFPI 6w1“1@3I92 7 Xw4W 0 A 8bÂc 9‰©4peFP2¨(  “324e0ý —™6T69e044j6gA50 hp601@PIw2 87eFe56G˜14“ebq006wG 1y3IG2 3™Xw43TF V0a@4V0z70TW 8b3541X25wG tIF 4‚4@@4h44eb8p@HGG 4a14@8b(02IF ˜p4reb„0 u”’@VYg @EpX2¨7 1ÂF @”˜SPfT7ý( pP2—@zYFT P(eF7 @u5 du8bPTFG €G { G I e05 ’ f R ™ A 2 7 }350 xv h« b G2 h 7 Ic 0 T4VA@7 u eI|4@374€394—hw1XI8FS4PgeFVTPT448bm4P2eF€62e06Te046ge060@87e54eb13™6Te0@37e035cP2F  t@B448pHG41@˜f 5 G TA 0b 0AI f 2 G7 FA 0 G ( G 5 5 w G 0 G T w ýRI 0A F2(7 0 G T ' 0 G7 02 A27 9 5 ý 0AI f 2 Tp FA 0 G 5I¸    A27 ¨ 0 c8bwmoS1Hq8FS“6(4sc@Sz7e0dPTF  |—™39cjh§@PIeFc£Hø4q4ebd@s’@° ¨ “4S#z – e0654b u 04Pp@S‹4P2eFVG@87VG…48b˜8T“32V03Te0c3ge05 ü 4eb˜6Te0@37geb@tg`2ƒ ‚1XI8FDG3w1“16IPw4”1k1XIeF4tXø4”6548bHG3F0 97q ( 5 T 0 G G 5 0 G T w GIW 2 0q2 FW 5I 2 Tp FA 0 ( G 5 ¡ YW G 0 0( F F F GI „  0 G RI w G w7 0 0 G I T 5p F 4X28F“32V03Te0c3ge05 ¥ˆ¢ ö“Q3A80œ8735c321vXTvHGHRv@—eI¹5tµ4ebrSi30@e785‡6GS4…ii48b‡T u 4VbiX° ¯ e041X(|¸ x f ‚ … u u f ‚  x x w u s d p h f d ‡b#†„ƒ£0%€yvtrqigec a2 0q2 Ìb@PI8FDG3w@€@3Iw Pw4—PIVF35l#4@1PIVF4tXø4”4X2PF39cjh”4@PI8FœE€3¥uc8b1tIR  ”35e06Gh@87@4u4eb’@#1XI8FDG44tph”@7 FA 0g A27 2 Tp FA ( 0AI f T2 G7p 0 0 G 5 ý 0 f 57g 0 G RI 2 w2 R 7 T 8T@—eI¹5t  4ebU@d43p—PIh4eb˜30S8785…8Is6A3060jVI4yP9X9F u 0 u 1eISz7e0dPTc8b‡1E’@4Sº74u9 ¹gtÉDiX04˜@3Iw GI „ 0 G RI T w 9 0 G w G G w 5g 2 T 5 F G 5I¸ 02 „  9g f 0 G F w F2p 0 G G 0 G 5 9 G GI „  0 G 0g 0 Gb F f 0 G T7 0 Fp c8bxP2vX039VwP53FUXG44‡c8be8Iv@XqeFSº769e0”X0PFieI8T@—eI¹5t­48bDG3wlc…iU“1X(˜a0 u V0354b eIU@‡4XA41( G F 0q2 w 0q 7 T7 0 T Fb { 0 97q 0 0 ý 0 G A27   0 G 0W 2 G 5 7 A2 “32X0432@€@3IpvY35@w}@6T1Vq65e0UXT4}y36T4PpS@“621P(6F’ac8bq4S‹h48bx62300 u 6Glq1XI8FSP739e0w’4EC 2 w ™ Tp ˜ w7 5gg7 F F T 0 G ƒ RI 0 97q 0 0 0 G 57 ÿ  0 G 0 @37w0 u ‚3AV04P“§db@1eI44S‡80DGe035VwPT4¹AH„h3fV0dc8b|¹Ry µ @3T4Xp@S€621P(6F‡48b‡V0@»4­48bae0654b u Y § Ø Y   `Ü Á ””è Á E§ Ø ÿ qÿ Ü Á ””è Á ‚§ Ø ´ ¥´ Ü ¨ E§ ‰ ² Y èè ÿ ²  èè ´ ²  ² T7 5g 0 0W T 97 2 @©6Ae0VT3Te0cc…iat—zY7 u @Sz7w vG@E413Ic3211h14m48bd8Ih1PIeFcPp1VI”4ebj…3fe0VG—VY4621P(6FUXWe06GX04˜@3Iw@Eb µ XR~e0@˜Vf354ebVG4E¸ 57g Tp 0 0(I fIb 0 G G 2 G 9 T 0 G™ T T 0 0 9g f w 7 T7 F 5I 0 5p ² § ¶ ‡E§² ‰ µ ¨ tz³E§² ‰ « ´ YW G 5 5 0W 2 w T2 G7p 0 w 0 FA972C 0 G ‰ ˆ † ¢ RI T 7 RI 2 G7 “#3Ae0€62e06Te046ge0‡lo@3741PIeFSE€3¥a60432V035tXø4Û@4E‡48bE3™‘T}‡B’@~DGe0i’Sp1PIeFS˜PfF „ 5gg7 ( w 0 FA 0 G A2C G 0ƒ G T w7 5gg7 F F T 0 G GI2 5I 0 0b €zieIc4@w4P26F432V035tXø4~48bh4Ea8Iq62@@87rS7 u db@“eIc4@~80DGe035VwPT4¹AH„h3fV0m4ebB@4B16548bDG4»‘ 'Q¥’” ‘ …–l¡Vh– ‘– y…ŽËTRrªSQ‡£’ ‡ ‡uªQIG 8†ƒFƒ kŒ ” ˜ U  ˆ —  ”– k P sŒ R – — — PH ‚ ‚ @ ¥ çê t D #e9qxEÏ C A ç t B6öé ˜)Ò #yï ö í ç ð zêx’‡~©h)Ò ‚6ïzò4t©cp½s·zìBlç ©«» ò ôï éçì êí õ ò rqï i tq òg ö ¼ × âÓÒ×Ü× Ô ÐÓ ÙÚ ßÝ Ü Ù ÙØ ÐÒ× Õ ÔÓÒ Ñ Ð Ï #rÄÇBÇávy”ràۀ}ސ3ÒÛrÚ”4ÇñÖxBÄv”m’Î zÍ ±° ä pã .

9 7 w 5 g 0 G G 7 G  F f R T T 9 W 5 g G 7 F f I A 2 G 7 F Au G 9 g g @EXg3F4X2eF4x4ebeSEeb· ¯ 4X(|¸ @eIt5x30V0…u 0 u h3fX041eIch3Ae0SEP2h14s1PIeFSEPgVFeTXT4©8Ih6AP0PFc4@7 32c»‘ ÛyVY35@6qþ“W 3A6044—eI4x‘T’  I #V0@‹e06548b©30cP2eF~‚—48bDG˜p@jVI`5SI s’PTxXT4}{ 0b ý 0 0 Y wpA 5g ‰ G 57 0 G w T ™AI 0 f GI „ G 7 F Fb  d w u x Œ gk¡i“’ ¡‘yb%e‡  d …s x ‚ x c  T w F G F FW TR @e0Sz7}PTc8bxP2˜`YXGXFP94@87VGÄ@I 0(2 5 I2 T7b AI 0 f 0 G A27 F2C 97 5 R w F2p 0 G 0 GpI 9 R w G „  0 @4@87”4}SE’j4eb6Gh‡48bx4Ss™  80XGcEtX9@1tI”X039ewX53F}HG44a4ebv4PAVF8Tc1PTX9@½7v30@e785Btlc8bG 2 0q2 FW 5 R Tp 0b ‘ 5 0 A279 G YW A2p R F F RI A27 G7 0 0W 1XI8FDG3w1“16IPw41tIp3AV04x32c»|D8®1eIV535p4StSº7eFe5˜“pc41tIBPTXGp6280fG@pc@p6A80S˜PfeFVG3Tmtu8IG T7b ýRI 0A F2(7 f 0 G 97 z ý x „ 0(2 5 0 G5 R 9W T F F T w G7 F fI SEht@r44epXG4@@q48b‡HYP9@EVpT 9Ž­ cGx  þ2vB14@e7U48bÈ1tIX04@d7vGÄXT’HGy3Te0Sz7u6A80S4P2h14A „2 G7 FA 5 R Tp 0b ‘ AI 0 f 9p ˆ 9g 0 0 G 5 R 9 5 I T 5p F j1XI8FSEXgeFVTPT4a1tIs6Ae04§324»h‚—48bDG˜d35X04ydXG3FXwPF4j…i˜4eb‡1tIdeTXG4Vp3TV0uT u 4ebx ¯ e041X(|¸  d w u x Œ … xz h v u‰ f‰z x  v ‡ gy%„%9ym}‹iŠ9qg“ˆ0‚ ±Í°  Ì° ° 1° z”° • “ d ¯   Í T2 G 0 Þ « ý 0 fI  9W7 c1PIeFSº769~É  ˜dh1jr1H° ¯ X04@Ä{ eb@‡4P2X9oÆ GŽ 0 F z ¨ ®  ý  Á ­° ‹   ý $ |®   ý ´ ~ ­° œ   « « « Ž† ‡vÞ z ¨ $  bý $ ´  a  $ ý «° « ´ œ « ý ´ ¨«  ý ´œ «  €þ©  † dvÞ ý «° « ¨  «  z œ œ ý œ « ý €þ©  (7 Ê7 @E6A@d… z  ý œ ©   ý œ œ Á œ Á ¥ ´ ®  @ Á ­° «   Íƀ }•' z ¨ ®  ý Wœ Á  ´ Á  $ ˜Â Á ¨ ­° ý œ ´ ' ® ƒ œ ´  ý W œ œ « ý œ ´ ­° « «    ÍÆ ‡' z  ý œ ´ ´ ´ Á ' ®  œ ý œ „Á ´ ­° ’®   ´ ý œ W ´ ~ ­° œ   eA5S7 u dƒ@U‚#cS w7 € „ A2 z ¨  Á   $ |®  ý  ~ ­° œ   « « 97 F 0g 5 @EXA13I6ÊtS87È{ z ¨ ® ´ Á |  « œ « ´ý œ z ¨­° þ®|® ýœ ´œ k~ ™°­° q   65P04yUXG3FXwPF4h`fƒ 9pˆ 9g ° « ý  « ¨ « ̀ U•' z o ý œ W ´ ©   ¨ ý aœ ´ Á ¥ ®  ý œ ´ Á ­° «   € }•' z  ´ ý ´œ œÁ ¥ ®  ý …œœ Á ­° «    ( Rg 1VIt5c@z70 Õ z ¨ þ     1k« œœ   ° © ý « 35P0cB~HG3FXwPF4j|ˆ 9pˆ 9gi z ¨ €þ©  ý «° « °  Ž 341XI8FSe785vG4PpX9P9a48bq€3Ae0651±q3I‡V0@U6TP0eFXGPFX94PWVFeT1lg T2 G T F 0 G YW 0 w 57 F TI G2 5Ig F 0 G RI TI Æ T w 0AI f 0 G F 0gg7b 0IA G7 T7 T7 T 9W 5 “@d7vG1thPfq48b}@eVG1o§63Te0Sz7r394—˜q48bsX2–32l4@Eu3T—4iSEb u @X9390 u @u˜6fP041VI4g 6A80S4P2h14`A—1PIeF6G3w@“13IrPw4—PIVF35lÄ@w6A80SEX2˜@4‚v®j1XI8FS4PgeFVTPT4È1 ­I —1PIVF4£Hø4veId6AP0PFc4@UV0@7 G7 F fI „2 0q2 w FA 0g 5I G7 F fIA „2 G7 FA5 „2 Tp FA G 9gg7 5 TAI 0 f 0 G F 0gg7b G7 RI 9 Fg G F I T 5 F 7 0W 0 G G7 G 9g F F cj4eb6Ghi48byHR‹32l4@EÈP9X9F u SEb u @’Sz7Pwcj`YmPT‹2 u 4ebv@PI—zqE3bli48b”SE8b…3AX0PF4hPfvXT‡¹Gƒ jdc©4@x‘Ž4‚6—y£‚1PIeF6G6w1“13I¹wD„Xw4jXIeF65t©4S#1XIeF4tXø4”1tI”‘T‡‡|394—he48b˜8I6AP0PFc4@7 Ž A27 Ž T¥ ˆ™2 0q2 FA 0g A27 2 Tp FA 5 R ‰ ˆ † „ 0AI f 0 G G 9gg V0@q6Y4ebp324b u 4jc8b6Ghr41PIVF@S˜“x6A3044—eIcU8THG4ep6Te0”c8b}80Se785vG4PpX9P9u†Í ¯ 4Su· ¯ 6Te041X(|¸ 57 0 G 0 TAI 0 f Tp 57q YW wpA 5g 9 5 0 G G T F A27 5p F 34—48bDG˜|@’`Y6GX0eF@S”m@tIc1PIeFSº769e0Q… ce4eb”T u 4ebdH° ¯ X04@ÄU3cj4eb6Ghi4P2cdbew@˜`f…„˜Xf8FG TAI 0 f RI 57q 7 5 R T2 G 5 „ý 0 G I T 9W7 { TAI 0 f ( F 57 0 97 w7 YW wpA 5g 57 G7 G 0 f T @E8pDG@©“‹3A6044—eI4iV0@aS48buh1eI©ebXGF u 6T30@e785”SjVI`5t­DG@cDii4ebue0@4˜16I‡4vDGP09 u mÉ w G GI „  w7 0 0 G 57g f w Tp I p  d w u x w x q0y¡iy~yw Xvs }|{©ybg“v d p d p xz x w ‚  ° Ó à ÐÚ Ã  tÓÚÑ â Ý Î Ó Ï Ù ×Ò ÑÓÎ à Ý Î Ü Ù ÙØ ÐÒ× Ú Ð Ù ÔÓ âr à Ö j±Í° ”vÛB—ѧxuÇB}á”ÄÒ rào|”B’e}}À3ÒÇrځvcǀv|DΐB|sÀqp±ÖÕ .

λh λ h= οο × âÓÒ×Ü× Ô ÐÓ ÙÚ ßÝ Ü Ù ÙØ ÐÒ× Õ ÔÓÒ Ñ Ð Ï #rÄÇBÇávy”ràۀ}ސ3ÒÛrÚ”4ÇñÖxBÄv”m’Î a) Diffusion Ι(σ) .2 0q2 FW 0 G 5 R A27 ™    97 5 R w F2p 0 G 0 GpI 9 R GI 5 97 w 5 1XI8FDG3w1“16IPw4xc8bw@tIh4@ycz ž‰ X9@i1tIhP039VwP56FeXGc4p4ebq4PAVF8Tc@wXTP9@½7d@—eIr@EPg6F4P2VF4g 0 G 0 I ÷  RI 0(2 5 7 5 R 9W T T F7 5 A27 w 0 G 0 T F T F7 5 GI c8b6™35@6q0 u ~Q ‰ Sx14@e7i1tIqX04@87VGucP2@˜3fV0–4@xX039ewX53Fpc8bq4XAeF4X2u4X2@˜3fe0u@—eI5 Tp 5p T 0 G Ê f R 7 T w  T7 T w G7 G ƒ  57 F(7 F 5pg F GI „ c1PIVF44eg”4ebxeI3560n1VIt5©zY7 u d3TV0@z7V0354X2F ‰ @hV0@z7aS48b‹`2d@VY@4P21@˜Pf”V044vXT‡@—eI`5cý 0 G 0 G 0 FA Fp¥ F 2 G7 T 0b { 02I 27 G G 5( 0 f 0 A27 c8bs624b u €62e065£Hø4”80XGc—aXT©1PIeFSE8pHGeFh4}qS41Q@Eeba6580S37e01a6T˜16I3wlW Æ   Æ 4@”P069ewX53Fw 0 G 0 9 GI 5 Tp 5p T 0 G G FIg G7 G G7 w T Ÿ° ý 0 TA 0 (2 5 0 G G7 c8b3T1zq37P0}@—eI4@PIeFc4ega4ebU“P21lSEeb}Sd304X2eF”z‡d« ©  324b u 4462‡1œ0 4@e7a48b}S48bG T u 4ebhe0c1Eq4}xSP04Sd7VG˜tyX9P9F u jc8b6Ghqc8bsdbXw4b u @tIR  q1Hq8FS“6(4r@z7V0}@h14@e7o7 I T 5p(C 0b {  9W T 0W AI 0 f 0 G F 5 ý 0 G7 02 9 5 RI 0(2 5 XT”e035c8bÇ@e0@tI6Re0654}ij ¯ 4X(|h30V0ÈSP069ewP56FaXG4chc8b©1Q@E8b‡6548bSe7£‚X21P(VF1h48b‡SdeTVG@87VGT F 0 G™ 5 0b {   F ¸ T™ w F2p 0 G 2I 27 G 0 G 5™ F 5I 0 G G7 5 GI 5 Tp 5p T 0 G ™0 0 T ( Rg 9 0 G 0 92p GpW AI 0 f GI „ G 7 97 F Fb @jVI”41PIVF44Vgx48br@h3f4dbVw©1eIt5cSz7P0’4ebp@XƒPF44rz™j4€j4eb6Ghi@—eI¹5SI u #QeIXT@”PTiXT4}{  d w u x Œ w u … ds w p ‚ œ — p w ‚  ™ … x  … ˜ —  d f x • ‡ 0y¡iž¡t‘yX£¥›£kšš%y#rgy– ‡|0bf “ Gg7b Ÿ F F G GpIW cu6580c@4X2}PT4eb}c1l@7 XA@87yXT~V01o’D±ùœc”P2B1eIV5354ebyc1l@d4P24eb@4r—±Ye7BXGS@44epXGXF4˜S˜X2B1eIV535a4r3T30c4jVI4g F T F 5I Æ “–d F 5 0 0 G GpIW7 ( F GI2 T T F™0A 9g f7 F 5 0 I2 wpA 5 A—4eb6Gh˜1VIt5c@37P0uc8bdbXw4b u X2F ‰ @d14@e7˜mPTwe035c8b~SE8bmT u cebuV041Ea4ebo14@48bHG”' I 0 f ( Rg 9 0 G F  RI 0(2 5 7 F 0 G G7 G I T 5p(C 0 G b(pI 9 351VIe535”j148bHGF u PT–—48bDG˜’4ebiSEebB£@z7h”@4”3T—4iXT48bB@@Ìl@hc@k6A80“P2@t”@7 u ”' 0 GpI F AI 0 f 0 G G7 G ™2 0 f GI2 0IA F G ™0qIW7 GpI G FIg T T   1°¡x ‰ x z 14@e7x48bhX2qX039ewX53FiHG44x4ebh1hHY39V0PT3F6we04wPTX9@½7eSEebþ”3T30c4jVI4©eIXT@7 0(2 5 0 G F w F2p 0 G 2I 5g 9 R G7 G   7 wpA 5g 9 XGacp4£@P0cu9 W@d7vGHY41a@4”PT©—48bDG˜i4ebtœ…ž@lf›@”åhÁÀ”8ùj²BXTQs324b u 6™3Te0Sz7‹1PIeF6G6w1“13IXw4W F G W™ T 92I GI2 F AI 0 f 0 G™ – ˜ – ˜ ƒ F ý 0 T w 2 0q2 F 5R @tIÇ6™350@6q0 I~ecl®  ­ý UÁeI”1tImXGr@i4XAeFeTc1”XTP9@½7#c@iP069ewX53F‡HG44i4eb#1‡8TvG@d7vG‡@—eI5 ÷ z © w e– 5 R F RI 0 GpI 9 R A27 w F2p 0 G 2I 5 T GI Tp 5p 0 G w R ƒ GI2 F GI 5 Tp 5p T 0 G GpW ý RI 0(2 5 7 5 R 9W T F GI c1PIVF44egTh4beș36G@½7Q`2pz@cdXTd@jVIwc1PIVF44egh4ebdj4|™  @”14S87od1tIiX04@d7vGaPTd@—eI5 3c1PIeFSE€3¥‚694—˜}1tI8T@—eI¹5tËSm6T30@e7856G@c|B1H° ¯ e0c1P(|¸ T2 G7p 0 0AI f 5 R GI „  RI w G w7i ˆ û 5p F σ = e λ h . ωh Ι(σ) R (σ) oo ωh= 0  “±Í° .oo R (σ) λ h= 0 σ=e b) Convection i ωh .

°  x z 0W 0(2 5 F 9 7 T ¥ GI ¦  ž›– 5 1Þx ‰ v32300 u 6Gl‡14S87p6A80XGhPfXFi”vG4Vp©@42 s‰ E@d1tIR 5 0 0A 9g f7 I2 wpA 5g AI 0 f97 F 0g G 0 G w T G 0 G 0W w 0 F @eIe565U44epXGXF4˜Se4r3T6044—eI4hjc8b6GhE@EXA13I6ÊtS87854eb604P2VFeI u c8bp32600 u DGt~30c32e065£Hø4A 5 ¥7 f 7 F 0 G GpW ™ T w ( 0q2 w FA 0g 0 G 5 R 5 0 T7bg 92I ( wpA 5 1VI¡˜§˜PTV0354eb˜j4}@V0@z7–4X28FDG3w1“16IxPw4—PIVF35l4ebi1tI˜1VIe565e0SE4–HY41scP23Fc4jVI4g RI FW7g w 0 G T7b F AI 0 f ( Rg 9 0 G 0 9 Tp f “ • ùd ˜ ˜ ˜ @Q`YXGXFP94SE@z7’4ebe@EiHG€j4eb6Gh©1eIt5j@z7X0…c8bqSPƒPFhvG4@– DPù9e!—–D•ž CeED›8ùDEWf›’P“q¤ Re±ùt“f•Àý d"—Ë"œz¤  uT¤xd“s„ƒIœ%ÁeI˜œzvPù9!¹–D”P“mf•zD@½7‹P2|£4@xV0@z#cP28FD3@€@3iPwcjXIeF3lqc8bG ˜d  š ý š ù e–ä e– š¤ ef •“ ˜ ˜ ™Gw R F ™ A 27 T7w ( Gw0q2Iw FA 50 g 0 A27 ( G7 FA 0 G GIW 5 R w F2p 0 G 0 T F 5I 2I 9 R GI „  ƒ   F ¸ 4S’4P2eFSEPgVFeTXT4”48b©8bSt‡1tIeX039VwP53FaHG44”4ebe4PAVF4P2‡1Q1ÄP9@½7veT@jVI`5t­eT¹GicÍ ¯ 4X(|QP2F 3A80œ8785vG4PpX9P9‡HYP9@P78FvG@EUV0@BSEeb’41VI@z7V0|1tIc1|@P7441levY35@”’PTuj4eb6GhjSEPA16I3Êl@87e5U4}{ G T F 9 57g 57 G7 G T2 T 5 5 R 02I 5 9pgIg 0q 7 F AI 0 f97 F 0g G 0b   X9@1tI‡P04Sd7VGc4}PT’—48bDG˜a4eb’@PI8FœE€3¥‚694jhf 97 5 R 9W T2p F AI 0 f 0 G 2 G7p 0 0AI 34—48bDG˜}c1PIVF@SU1tIU8T@—eI¹5tâ@3T30S87Èyj· ¯ e041X(|¸ TAI 0 f Tp 57q 5 R GI „  RI w 5{ û 5p F λ h=-1 Diffusion c) AB2  d w u x Œ z ‚  d £ x x ‚ … ¡ ‡ gy%5by9‰ i|§¢#he Í Ó à ÐÚ Ã  tÓÚÑ â Ý Î Ó Ï Ù ×Ò ÑÓÎ à Ý Î Ü Ù ÙØ ÐÒ× Ú Ð Ù ÔÓ âr à Ö @±Í° ”vÛB—ѧxuÇB}á”ÄÒ rào|”B’e}}À3ÒÇrځvcǀv|DΐB|sÀqp±ÖÕ λ h=-2 σ2 σ2 σ1 σ1 σ1 a) Euler Explicit b) Leapfrog σ2 Convection σ2 ωh = 1 σ1 σ1 σ1 .

 T FA w2 T f w 0 G 0A2p 9W 5g Tp ‚3AV0eT46pewXT43T30c@d7vG˜46pewX53Fa48b35c44¤6fP04@eI441XIeF65e0T 7 GI2 F A27 FW T F F f T7 G ˜@4”XTÃ4@s`YXGXFP94Sd7VGcP2€PAX9˜h@§8Ig6Ae0V535t06RV0”PTh`YHGPFX94@d7vG4P2”@k4P2€hPT4}þ‚@PI8FDG3w@€@3Iw 5 F FW T F RI A Fƒ Fb { 2 0q2 Pw4—PIVF35l”4P2—Xq@1“P23TX044epVGa1Hq8FS87XG8F“@4—4@PIeF65e0’P2o3AV046230l…eb@lw@±q4~cj4eb6GhewvÞ FA 0g ( F 9Iq F FA T 0 G G27p¥ Tp T F Tp 0W GIW 0 7b TAI 0 f  † 0 G A27  € ' 0 G 2 T 5 F G 5I¸ 2 FA TIggI 0 G F T ™ F f f w 0W 0 7 c8b”4@a}•4eb˜1VI@z7V0yPT4eb’1E’‚1XI8FDG3we0X54}80HGeF1l414ebiP2rVG4Vp¥¥‚3Ae0eGXGhh13I˜3230l~1zqEb XA41I u —‘)ƒf•E@”åhÀ h@87u48by@”1eIV535#@‡c4eX04@d7vG#3AV0@º7693w4‹6230li1zqE#XA41I u jc8b6Ghf 9p ù… ˜› – 0 f T 0 G RI 5 0 27 GpW 9W T 5 0A 0W 0 7b 9p AI 0 0 G w 0 G 0 T F 9 R A7b GI 5 0 G ƒ G2 5Ig F2p 0 G 5 0W 7 f FW T c8bvP069ewX53Fa48bvcPAeFcP2x32X0P9@½7’@EU@jVIwc8bȹR’z“@87VG1l˜Xf44qXY39@XqeFSº769e0‡lxzYqh`YXGXFP94Sd7VGcP2F TX48bu€˜3AV04@z7y@eIe56548b“@V041E}4ebtSmV0432V0~c8biP2jP9@˜ef}V0@rSEebr46g80vGU˜Xf8Fl@B35le4€2 F G YW Tp w 5 0 0 G™ 5p(C 0 GRI T T 0 G F 97 T 57 G7 G T T 0 GRI 0W fp 3Ae0XGhPfPFhX2©3A@Xq1VI3Te0dl@37vVG6Te06580“P2yS…@PI8FjPp1VIv“32X0eF4@e785a48bÄXR‚6™35@6q0 u ~ ‰ P9@m1tIR F 9 7 F 0 9 5 0W 2 w G F RI 2 G 9 T G T2 G 0 G F 0 I ÷  97 5 X04@87VG44~PTo4@dP069ewX53F~XG4cw4ebdcPAeFeTc1}XTP9@½7@—eI‚SEPg6F4P2VF4u4eb}4jc8b6Gheb@luI ~@tImS48bG 9u W T 2 p F A 2 7 w F 2 p 0 G 0 G p I 9 R G I 5 9 7 w 5 g 0 G T A I 0 f G W 5 R G 7 I T R A27 w T F ¸  9W 5g 2 0q2 FW 7 G 9gg7 57 0 G 0 TAI 0 4ebyœÍ ¯ 4@wS ¯ 31X(|}…3fP0c1eI41PIeF6G3w@“13IXw4˜q8I’3AP0XF44@we0SDY48b#624b 4jc8b6Ghf  † Þ A27  € ' 0 G YW 0 Fg G 57 T w wp 0 F GII f 7 0W 7 f 27 T 0 G 5I dvö4@˜}•ec8b’“©3AEHC4—`Ywe0S‡3TV0@z7©db4—‡@cepeTXT~@jhqut…zYq@EegdhPf8FmVG1cebT XY691Hq8FSP739e0im35@Ìx@PI8FjPp1VI€62P0VF4@87e5eu@Xq1VI3Te0”eI…6Ae0P5c—6¥e0}XTSE8bx—48bDG˜i|@˜`YXGPFX94@87VGT 0 G 5 7 0qI 2 G 9 T G T2 G 7 0 9 5 G Fp 5 F G7 G AI 0 f 7 RI FW 0 G GI 5 97 w 5g 0 G RI T7W 0 G 2I ™ 5 0b ý RI 0 97q 97 T 5 c8bm3™@—eI”SEPg6F4P2VF4‹48bd@xPTVF@E‹4eb»@v@V01tI6RV0354}{   i@’6T4Pp@S~P9@˜efp@tIR ¥ˆ¢ ¡ 8IG V01PI69©VY651”XT‡SEebv@—eIÈ@EXg3F4X2eF4x”30c4jVI4v4—4eb6Ghq4X24dbew@˜`f…„˜Xf8F|SE1PIeF“321“16I|X9”' T w 0q F G7 G GI 597 w 5g 7 wpA 5g TAI 0 f ( F 57 0 G 972 G 0q2 w 9 ¼ u‰z ‹9Š‰  `y9S9ŠVŒ ‚ u p –‰ z‰ “ Gg7b Ÿ F 0 T2 w 57 T w 0 jT6580c@4…P26Ae0654PAVF413IaV0@de0@37dc8bG GI2 F F G F 5 R T2 G7 F  w2 5Ig F RI F GI „ý 0 G 97 RI w 5 w7 G7 G S4mXTmXT48b#dbXw4b u 1tI}41PIeFSEepXGjm@604@87VG1l˜Xfy@~PT~8T@—eI¹54e48bÛX9@B@o6Y@e743p6w@”SE8b˜eIT P9@˜efBXT—™  @60PÊeF˜3ge0VG}hPf8F4eb˜32cb u `YXGXFP94Sd7VG£@q@PI8FvG3T4€~48b~4epVT44d8Ia4P2eFVG6Te035e0“P2BPTr¹Gƒ 97 T F ™ T T 0 G 0 G 0 FW TRI 2 0p¥ 0 G 0 5pg G ( G F F z x® w F 5 ý 9g f w 9 c{ƒt ­ý ÚdbPw4b u @tIR  oDiX04˜@3I£X9@7 5 R Ÿ @tId1“´»© Æ Â ý Æ j1tSXTBXGhPfPFe`YXGPFX94@87VGu‡vl4ebS48bG u 4ebpSz7v4vB— % x ‰ —™1tIR GpIW7 F F 9 FW T Ž † Þ 0 G G7 I T 2 w 02 „  ³  x z 5 P039VwP56F”HG44d48ba4PAVF4P24X2@˜3fe0˜‡BlSz7e0654b u ™ ‰ P9@m1tIaP04@87VG4cmXT”dvö1PIeF6G6w1“13IXw4W w F2p 0 G 0 T F T F7 5 Ž † Þ T 0  97 5 R 9W T2p F  † Þ 2 0q2 F 5 R A27b 0 GI 0 G 2 „  „ G F 9 F  † Þ T 0 ™  „ G 0 G 97 TI 97 ( FI @tI˜4@46548b@m48bhvyjrd8Ie6A80HG˜XfPFrPTddvƒ@37e0654b u c“—vdeI”±Y7 u c8b4P9SrVG1hPf@w4X211( ™335e0Sz7V01”XTp‡BT1tIx`YXGXFP94Sd7VGBSe14@e7e4eb”SEebvc4£@1Hq8FS13(4#c@’@37e0mPTF  Q624b u  ý G 5( F Ž † Þ 5 R FW T RI 0(2 5 0 G G7 G GpW™0 G7 02 A27 9 5 ý 0 RI 0(2 5 7 5 R 9W T 57 TAI 0 f GIW G7 I T 5p(C 0b   F ¸ F I @’14S87§h1tIxX04@d7vGoV0@˜4jc8b6Ghk8b@lqœE8bG u 4Vb˜6Te0414o4}{ jÍ ¯ 4P(|gP2Ã2 u cebT V0@’4—48bDG˜h87eGc‡†D„14c~Ô654eA@`IjebVG41tIþ4S˜j416I3we0‹4ebq1tIqeT@—eI`5lö48ba@’3T60@87È{ 57 TAI 0 f Gp 0(2p Þ 0 5 „ 5p R A27 „A2 T 0 G 5 R GI „  0 G RI w 5  ý 0 F 5I 0 G G7 G w 9 57 GI 5 Tp 5p T GIW G7 G GI É    0 cz ¨ #324b u P21X(eF1a4eb}S…3A80œz7—PIae0@eT@—eI41XIeF44Vg’8b@lSEebae0@mB  ¨‰ 624b u ™Se0@37x4P2eF6G6w1“13IXw4i4ebm1tI#4Sۙ ´œ ¤©  §32cb u P069ewX53F‡HG44e4ebv3T@±q37P0”@jœVIv41XIeF4cegp7 T w ( 0q2 FW 0 G 5 R A27 « ý 0 w F2p 0 G 0 9 GI 5 Tp 5p T ™Se0@37”4P2eFSEPgVFeTXT4u4eb}1EU1`YHGPFX94@d7vGwc8beTXGhPfPF~@—eI~41XIeF4ceg˜£3™6Te0@371XI8F6G6w1“16IPw4o4@7 T w ( G7 FA 0 G 5I¸  FW T 0 G F 9 GI 5 Tp 5p T 7 T w 2 0q2 FW A2 2 G 7 F A 0 G G I Wu ƒ   F ¸ F I T 5 7 0 b {  0 2 I 9 7 w 5 g 0 G G 2 F A A 1PIeFSEPgVFeTXT4”48b#eb@tQ`2ujÍ ¯ 4P(ċP2Q2 u cebie0SqV03T4}eS41Ä@EPg6F4P2VF4i4eb’eI1XI8FXGc4@7 X2weT@—eIuc1PIVF44egI h1zqEevG4edXGÈD™1”zc4Ä@X° ¯ P0c@| ­{ Xw443p…dPTF ­1tI|Sº7h14—XY1lg F GI 5 Tp 5p T G 0 7b Tp f F ® ° I2™ 9W7 FW w 7   5 R 9 F fI2 9I Xw8FvGPTeF6580DG@87@4db’8THG}304X2jB3@eIe565e0@Ecg u XI~6044—eI4‡eIh6A41X(eF3TcrS7 u —48bDG˜‡SE3A@aB4}{ w 57 w F w F 5 0 T7b 9 wpA 5g G 02 0A T AI 0 f (7 Ê7 … 0b  d w u x Œ ’ ‚  £ ‚ § ‡ 0y¡i“bk%bštx ² ® ­  – ‚ ¯ ® ­ © p  d w u x Œ ‚ u u h ® — x ’ – h ­ … x  … ˜ — w u … d «  – ‚ —  d f x • ‡ ’© €š±y£°~šuQqgk¡iy††k€‡tkqšs|k#rg¡kh k¬y£0q– ‡|0kªes º ¹ »„ 'k%¸G·™A—pI…–'”'kleb6µ …„ƒFƒ kR i µ ’ ‹ t ” ¶R– ‚ ‚ × âÓÒ×Ü× Ô ÐÓ ÙÚ ßÝ Ü Ù ÙØ ÐÒ× Õ ÔÓÒ Ñ Ð Ï #rÄÇBÇávy”ràۀ}ސ3ÒÛrÚ”4ÇñÖxBÄv”m’Î Ž @±Í° .

8 λ h=-2 σ1 σ1 σ1 e) Gazdag d) Trapezoidal f) RK2 g) RK4 dÍ Ó à Ð Ú Ã   t Ó Ú Ñ â Ý Î Ó Ï Ù × Ò Ñ Ó Î à Ý Î Ü ÒÙ Ù Ø ÐÒ × Ú Ð ÎÙ Ô Ó â r à Ö Ö ±° ”vÛB—ѧxuÇB}á”ÄÒ rào|”B’e}}À3Çrځvcǀv|DB|sÀqp±Õ Convection σ2 σ3 σ1 σ1 σ1 ωh= 2 σ1 ω h = 2/3 2 .Ds‘Az‰“16I ­w 4—48bDG˜Uc1PIVF@S@tI8T@—eI¹5tâ@m6T30@e7ÄyjÍ ¯ V041X(|¸ ® G2 TAI 0 f Tp 57q 5 R GI „  RI w 5 { û 5p F Diffusion λ h=.οο σ3 σ2 σ1 λ h=-2.

R@e@4@87Q3A6080“@e7@E1ph1zqE#6Y4eb£cz ©Â 1tIX21P(VF1i4ebvSdVGSd7VG”4jc8b6Gh‡41XIeF4ceg|X9@7 I 0(2 5 G2 57p( 7 0 7b 0 G™ ¨ 5 R F 5I 0 G G7 5 T TAI 0 f Tp 5p T 9 TAI 0 f 0 G 5 R w F  š ù È “ À–… É 27 T7 cj4eb6Gh‹e03Tc8bp1tI©30cP2j‚@9…EÁdD•QËuI“uÊ@’SgeIG 6Ae0V535t06RV0xPT‹lj`Y˜3580VGSº7Q4}{ 5 F 0g G G 9 0b  £ †À·1tI¤ef6580e4X24@37P0a48bxX2£ED›—ùD˜ª{±ùdšµEÿdw30XʘXf4P2hx6Yc8bG ¡ 5 R G ( FA 9 0 G F•   ù • F F F f 0 T7 5 R T F G 5 w7 TI f 0 G 9 0 0( 57 TAI 0 f 0 G 0 Ib™2 T27g 0 Sv@½7@VI4P2u80Se743p6w@‡vG1˜u48b’XYX9@8765432@ee0@”4—48bDG˜iV03T4ebÇ6™35@6q0 u c‚‚1XIeF4@4c…iv3TX0eF65e0T 5 9 7 { 7RI T7W 0 G 2 „  Ê f R 7 T w 5 T7 ý 9g f w 0 f T 5 R 9W T2 @PI—zY|eÄ@UXTeF@4‡48bxvr4eI6530)1eIt5q±Y7 u ~43A6030—eI4g  @oxDiX04h13Ivh1VIU1tIvP0c@d7vG44p 0 f 0W G 0 9 F AI 0 f 0 G™ w F2p 0 G 2I 5 T T 97 AI 0 f 0 F Æ 7 5 h16I3wlx8IxXY69@XƒPFvXT#jc8b6Ghe4ebtSP039VwP56F‡XG4ce48b#@v8TvG@d7vG‡jzY7 u @#j4eb6Ghu4X2P9o’”1tIR GI 5 Tp 5p T 02I T 9 G7 w F  š ù È ù ˜ È 7 T7 G @jVIc1PIVF44eg~41rvG@z7X0S~604P2jr@9…EÁdD•Qv£› Áe@w8I˜3Ae0V535t0DRe0BPT~t—`Yy35hef@tIm4}{ 5 F 0g G 0 5 R 0b  cz g  324b u f›@å …˜œiEÿd©œBP9@½7UeT@—eI5 0 ˜ š ù • ›š 9 R GI Tc1PIVF44egBX9@pV0@z7dPT4ebs`2…34—48bDG˜o@E6A@d›4@s8bVG@tIceb@T€v„h@Emu654eA1¹Ij4Sp48b‹“W p 5p T 97 T w F G ƒ TAI 0 f (7 Ê7 … A27 5 R T7 T f7A ' 0 5 „A2 0 G Y 6AEXCXF4˜6fjDi’XT}t—`Y|6548b@4}Äcz g  624b u PùReʁfD˜%f•E—·EÁdpœ}PTP9S½7’@—eIy41XIeF4cegve0Sz7w 0 9g 0 0 F 0g G 0 GI 0b { 0  ˜›ƒ ù • ›š 9 R GI 5 Tp 5p T 7 T TX48b©`2w‚jc8b6Ghx1VIt5c@z7X0”48b#“‹3A4XC4—`YvPT”l—`Ye4vh33Tl—`Y…I u r@it’eIG 3fc8b‹4EC F G ƒ AI 0 f ( Rg 9 0 G YW 0 Fg G F 0g G 02 „ 0g G G RI 0W 0 G A2 0 u cÍ ¯ 4S˜ ¯ 3@P(|oP2~4—48bDG˜~@—eIvX04PgeFXG4uw4eb|@~3T60@87e5@jVI`5t  4ebmDG3wleg4X2w0 u ¹Rƒ ™ A27 T F ¸ F TAI 0 f GI 5 9 9p f 0 G RI w G GI „ 0 G 0 T F p  d w u x w x x ¼ u p w ‚  ™  – ‚ x –z‰ qgk¡iyǧ0›byšžy£€k9ŠVŒ 641PIeFSd7j4˜@3Ic™™P9@~S4ÈXR43™VG1h}1tIeT3TX048bvG1I u V0@…4@dP0c@d7vG44p T2 G Gpg f w 97 GI2 F TI f 5 R 9 5 57 A27 9W T2 HYP9@37Pw44@eI85vG@d7œz7q6AP0X9@z7vV0@Ucj4eb6Gh’db4jr‚6A˜Xf394b u 351±q˜XY€dƒPwc—PTF ® ÿ Ü @mcPp@S4@P78FXGcP2F 9 Fbg T G w 9 w 57 TAI 0 f wp  0 0 0 I 9 Fp¥ RI 0 97q9 F P9@˜efoVY651”§621’m3‡”z° 1d ­ e04@373wl~€624P2hhPf~PT”6580vG@87XT4e30V0#@37u41lj¢8IÄSE€3¥0 97 T 0q 7 0q ˆ Æ ® Tp 0W G 0 F f F F T FA T 2 w 02I™d G 97p XTF —41XIeF44Vgoc8bv@…41”v@)ÝRÅI ’44epXG41Sq…48bmXR3™651Dq0 u m» Å cP2eFSz7e0654P2g8bHGF u 35P0X9@˜efT   T p 5 p T 0 G R I 0 2 I´ 0 A F 2 ( 7 f 0 G F 0 I ÷ ( T w F 97 DG1Q6Y4ebÈSP04Sd7VGaXT§—48bDG˜˜48brXR§4@Ǚ™P9@˜ef§VY651”t…8IQ3A30Vw1tI”e0@u8T“32X03Fh6ˆ—3I˜e03Tc8bG 0( 0 G™ 9W T F AI 0 f 0 G F A27 97 T 0q 0W G 5 R 57 G w 0 w 0 3Ac1P(VF3T4ÛX9390 u XTmVT3T60jVI4˜cP28FvG@d7vGa48bĹRc@ÍclsyX2w°   G1tIR ® ÿ Ü 48bE6™8T@—eI41XIeF4cegT 02 0A F w 5g ( 5 T 0 G ƒ    ¥ ˆ F Ç 5 0 G GI 5 Tp 5p 0 G RI G w 0 w 0 G 9 F F 0 G 5 T G TI w 5p w 5g9 0 T 0 G A27™( c8by@m8T“32X03Fh6ˆ—3Iw48b”60PÊXF@º7eFXG4X2¼6f48bmVG@87VG˜8I#32V014dbv6Te04c3A30—eIcÈSº73F6wtVgu4eb4@l4cP2F v„VG@87VGÄXR69e0@4‡V0@…6Yc8b}SEebG u 4€‡syS30PweF6G@e74’P2pVƒ1I u 8T@—eI41XIeF4ceg’8bHGF u 4jc8b6Ghf 5 T T GI2 57 0 G G7 I2ƒ 0 ‘  w 5g F 5 GI 5 Tp 5p T TAI 0 Iu b A 2 0 A 2 p T p f 0 2 I 9 5 F G R I w 0 p T 2 w 0 G G 7 9 7 0 G 0 5 I 4gc@d7vGeT35c44hVG4ux41rz™XGcep3TV0ePTc8b~@p3T60432c—6¥e04@3Ix4ebp80SEXp@S6q‹8Ie654eA@g`2ƒ ÄÄ d „ I G 9 7 p ¥ 0 ™ 0 2 I T p IF 5 p g T 7 7´¥e8Û@E€3ES41}4@Pec4e”E™ ×4@E1…8‚SE€30   A 27 ™° IG 97 p¥ ™@41B@4Pg3FcP2eFcq4eb|D™£‡3ŒeT@jVI`5toI u he0@xV0354}…cz ¨ @d Á   ­ ÌÀ‹  œ­ ¨ Û  ­ »4Eq0 u 02I 97 w 5g 0 G ´  GI „  G 57 0b {  ® ®° « ®  A2C z ¨ Ç  ­ scP2eF@8I6SlUjd « à Ž Á à ¨ ® 9í  cz ’ v1tIS48b}T u PIP9@tI~HG˜±H° ¯ sBˆ 1eIE¸ ®  ( 5 Gw7 ¸  ™ ¨ ý 5 R G7 G 9 R F Í ¥ f 5 cz g  s624b u @º7h14—HY1tuPw8FvGPTVFœ3580DG@87SEdbv8THG’4P2—c4epVG€3A4VbPTPFc˜16I3w@”l…XYX9eF@37QSz7w ý 0 9 F fI2 9Ig w 57 w F ( FYA T YW 0 9g f w7 0W T 0 2 XT4}y  Ç@6T4Pp@S4X9@qVfpvY351B1tIx32@6qi`YHGPFX94@d7vGeTXG6G3wlegcP2U46GX0E3™6516q0 u myS1Hq8F6GS8785VGS7 Fb { ýRI 0 97q 97 T 0q 5 R 0 0 FW T F 0 T F Tp 9 0 I ÷ 0 w G XT~XGF u X0—’@m€X21loDY@87c3p3wS©@7 1VIt5qeIl—XÁ†Àö4@7 ¢ †À%@~˜Vf3580…P2o8b@lue0S8746p3w@”354VA1I F Fq RI G FIg w 5 w7 2 f R T™ ¡ A2 ¡ RI T G F GIW G 5 w7 0 5 „ F G F AI 0 f F G G7 G ™   2 0 F 0 F( TAI 0 f 0 G ( Tp I T 2 w 02 jeAX548buXT€j4eb6GhiPT4ebeSEebyc jg@PI8FDG3wjgX2k621Hq1i4—48bDG˜x4eb©4P2VF4™ u 4VbgSz7x4v„ ¯ Á ›² S e  Á ´ ¯ sœd Á E1~« ¨ t±E‰ ¿ T´ ›² ‰ ® “±ÍH° —¯­ ‰¾ ²‰ ²‰Ž ´³² DûÌH° ¯ X04@Äw60e0 ­T —48bDG˜x6g80vGPT8FHG4eUSz74X2PF9 ® 9W7{ AI 0 f 9p f 5 0 ™£3g80vG`TSI u 4z™HG3FXwPF4jDit@e0S8746p3w@mvG1˜w48bÄ@oc48pvG‚@37Pw8FHGeF65˜7 1VIt5mh3f80vGT 3fX041VI4y@o4P2€ƒ „ G 9g 0™ G 5 w7 TI f 0 GRI YA T9 w f R T 9W 5g RI A F TX48bESÈc1PIeFS87e5VG4XpP9P9BvG3TlU4ebE@}4vy68T@—eI’c1PIVF44egsSy3Tlj`YeX2@d7vG3560s@3046280VGXTjDiU4ebu“W F GRIT2 G T F 0W 0 GRI 02 „ GI 5 Tp 5p TRI 0g G F wRI w F 0 0 G Y GpIW7 Gb(p 5W 0W 2 w G7 97 T 5 R 9W 5g FW T Tp T 5I f wp f 7 F 0b j1tSr€@41eIc~lh@z7rSEebG  X9@˜efy1tI©3fP0c1eI4w`YXGXFP94Sd7VGB4@PIeF65e0}V01˜˜db4e‡BPT~e0654}{ ¼ u‰z ¥‹Š9‰  `yŠ¨Šyqx ¡£¢‚ ½ ‚ u p –‰ f‰ w d … u p ‚ u × âÓÒ×Ü× Ô ÐÓ ÙÚ ßÝ Ü Ù ÙØ ÐÒ× Õ ÔÓÒ Ñ Ð Ï #rÄÇBÇávy”ràۀ}ސ3ÒÛrÚ”4ÇñÖxBÄv”m’Î   1±Í° .

36580c@4dbm—…i4a48bxP23Ae0VT43pVwPT4VT3T4tXøeFVGT Gg7 w G 02 0 G F T FA 02 R@aDG3wV0|4Vpe4ebq8I”c@z7X0v¹G”@`YXGXFP94@87VGÈ@z7XweF65˜4€B@c4epVGe4ebm1tIx1XI8FSSHq8F@h”1VI¥qec8bG I ¥W T 0 G G TA 9 ƒ  FW T9 0 fp2 RI YA T 0 G 5 R 2 G7 GI f 5 7 f 0 XTÃ1XI8FS48pXGVFxPT4}{ @60@Eegc321X(3FþP2F 6Y@e743p6w@”SE4PA—Hq44P2˜P56F48bq@tI»VA@“6(e05 “@xc14ebXGF u F 2 G7 T Fb  w7 T 0 0 w 5 w7 97p F FA F 0 G 5 R 57 Y27 GpI VT3T30—eIc…@PI8FjPp1VIu4eb…1c1eI4ebd@XqVF4qeI…VbPTF u 0 u SE8b~eT1eI6G6w1“321X(3F…8bHGF u 3Ae0Sº76FjVIeT@7 Æ Â ý Æ w 5g 2 G 9 T 0 G b(p 5 G 0 5A G G7 G 5 0q 0 0 G w T R@e6T4PpS@˜1e(@P7sHY39@Xq8Fœº739V0”˜6f80VG—vY4321X(3F’4ebhP2”VGXTjDi’e06548bi3TV0@z7’V03T4ebm@–eb@lg`2ƒ I 0 97q 0 5 9 0 G 5 T T T 0 0 0 G F F 0 0 G T w 0 G RI GIW z“32X0eF4@e7856A80“@7 u 4p G T2 G G2 2 0 G 0A w F G7 G7g 7 ( T 2 G 9 T G „YA T 7 A2C G 92I T 0 c8bq6T44Pp694P2qSEebG 8bSE§–4P2VF4p 1XI8FcXp1eI80S87VG`T“cSz780vGh»4EseI€XY41g3ƒ60e0#§‘ ü 63T4c8pXGc1@˜o6A80Se7@E3gV0T 0A F2(7 f G 57 XY694PAF u 4P2—zqEu6T4PpS@“621P(6F—Wh8bHGF u ˜6f80vGjvYh3AX04416I|SP04@87VGQXYe9“32e06544X2u@E˜4v„ ü 0 ( F 7b 0 97q 0 0Ìý T T T 9gp w ™ 9W T G 0b F T7b 02 û 0 ‚624b u T2 FA2 w 0A2p 5 wI FW T9 0 fp2 RI 0 T7 G2 5Ig F TI f 0 G 5 R Y 41PIeFXG4c13Iv65444‡VT43p6wj’`YHGPFX94@d7vGÄSz7PwVF35h4€B@‡DG3wleg@”“@d7vG1thPf”VG1he4ebm@½7…€ 641PIeFS87654PAVF416Iq`YHGPFP9c@d7vG…X2‡P01VI1cP2h~6548bœ87i”6T80cepXGeFVG416Io6A1Hq1eI6Te05 T2 G 0 T2 w FW T F 9 5 5I F f 0 G 5 7 G T2 w 0 9 V0@…˜Vf35e0…“62P0eFc@87e5Q4ebwSvP9@7 dbPw4b u X2F @PI8FœE8pHGeF‹48ba3™651Dq0 u 4v6™4—48bDG˜‹P0c@d7vG44p 57 T G G T2 G 0 G RI 9 F 2 G7 T 0 G 0 Ib TAI 0 f 9W T2 HYP9@37Pw44@eI85vG@d7œz7mc1”4P2XF4ep¤1eIt5a4PAVF”U3VG@4h’zYP74@37p`YHGPFP9c@d7vGE@e0XT˜@eI4h13Id4ebo`2ƒ 9 Fbg T G w GpI ( 9 5 f R 0 T ' 57g 7 9g 2 w FW T™ F f 5g f w 0 G  9 TIg T7 0 5 9 T7 T T 0 G 0ƒ7 f G @X04PWVFeT1l‡@i1V(@º7‡S”30PÊVF‹3g80vG”48buS@˜…eI©ebPTF u 0 u PTc8b‡30XʘXf4P2hx8I©4@‚‚1PIeFcPp@eIp7 F G F F F f G A27™2 G 9 T e0c4h13I”eI…62@@87~46g80vG|@65tec—‡c8b’1mc432l3g4˜XYX9@8765432@x1PIeFSd7cc˜16I‡48bÈ@~VG@3Iw Gpg f w G 0ƒ G T T RI 0W fp2 0 G 2I TA 0 0A 9 0 0( 2 G Gpg f w 0 GRI T 0 G ™A27b 0 GI 0 G 2 „  9 TIg T7 97 T T7 T F G RI T 0 G 0ƒ7 f G T c8bÄ£4@Ed354eb@qc8b§vp@P0cPWeFVT1lw@BX9@˜efd@–3ge0VGaPTc8bU@h30XÊeFh48b”@Sqo8I©3ƒ60e0q0 u ™‚3ge0VG¹TD„hPf8Fec8by@vVT3T4X2P9@˜efu48b#1a4c32l3g4vcj4eb6Ghv41r@ÈP9Sv1tI@PI8FjPp1VI‡“32P0VF4@e785G 0 G 0 G RI 0 97 T 0 G 2I TA 0 0A TAI 0 f 5pI RI 97 5 R 2 G 9 T G T2 0 G RI w 5 w7 0 G w T™A27 G T2 G 0 f T 0 9 5 G ( F T 57 9W7 f 5 c8bB@©DY@8746p3w@”c8be30cP2eF£‚4@d“32X0eF4S8785w˜1VIi1Hq1eI6Te0x8I’4X2—6ƒ30V0ee0@”0 u HY4@˜4ep6Te0y¢ A 0 5g f w T 0 F R RI 0 97q 97 T 5 R FW T YA T 2 T 5 0b £43246be04h16IkeIG VYSz7kXYV9P5@½7pXTF  @x3T4Xp@SmX9@qVfx1tI€`YXGXFP94@87VGþj48pvGkeIG @eI@37e0‹4}{ ‚Ô p’R Ó ’ ‹ t ” ¶R– ‚ ‚ 6Õq‡r¬™A—pI…–'”'kleb6µ 8†6‘Òƒ Ï  A B  œš çê t Ñ #u9qxD "Ð 6çöé ˜)Ò #yï zêx’‡~©h)Ò ‚6ïzò4t©cp½s·zìBlç ©«» t öí ç ð ò ôï éçì êí õ ò rqï i tq òg ö ¼ ä p㠁0A F2(7 S448pHG41@˜f 650 u XI…ebXGF u “32X03Fh6ˆ—3Ii’“’3AcPA—±qeI4xDY@87c3p3wS’P2ae0@37e0654P2yf®X9@qVfqHYP9@4ep4 ­p 48b‡4XAeFe5651±qI 9 G w 0 w 7 YW 0 F 5g w 5 w7 F T w F 97 T 97 T 0 G 0 0 XYX9@8765432@‚41XI8FS37PwPFc4@Ã|UeX2ÄeT€62˜6fe0X54€3¥e0”`YXGXFP94@87VGd4@Sz™“32X0@SXq4€6¥vXYX9@8765432@ÈXT|6Tt—`YG 9 0 0(T2 G 9gg7 † ¸ Ÿ F G 0 Fp 5 FW T A27 G 97 Fp 0 9 0 0( F 0g I u ‹c8bd@þ6YS8746p3w@u@’354VA1‹48bmlVƒ1I u @E@PI8Fœd7c4h16Iu@’“4@˜@y32@Xq1–’1E¸ G 0 G RI w 5 w7 RI 0 5I 0 G ™ 5 972 G Gpg f w RI G2pI f7 0 F( 7 5I @6Y@e743p6w@¤@eIt5@4cP2€@z7legwXY691Hq8FSP739e0€6™35tcepu6Y4eb@oVf35e0’1eIV535y3TX0eF65e0’1XI—±YÄv4X24@37P0}4ebeP2F w 5 w7 f R™( Fƒ 0 T 0 G 5 0ø T 0 G™ G 5 0 T 5 9 7{ ( FA 9 0 G G“32X03Fh6ˆ—3I}c8bl@4c8pXGc1@˜~c8bs@yPTVF@E4eb”@j6™351Dq0 u ~y  @41462cP9SqVfy1tIa`YXGPFX94@87VGT w 0 w 0 GRI 0A F2(7 f 0 GRI T7W 0 G 2I 0 I ÷ b(pI 0 97 T 5 R FW ¯ z͐mvÛBèÎ ¤uÈB}ÞmĤrà…|”B’e}}ސ3ÒÛrÚ”4ۀv|Dΐro|Ík“ªÌ±ÖÕ ° Ó à ÐÚÑ ý tÓÚÑ â ÝÎ Ó ÏÒ Ù ×Ò ÑÓÎ àÝÎ Ü Ù ÙØ ÐÒ× Ú Ð Ù ÔÓ âr à Ö .

31VIe535’1XI8FS3744Vp85G 5 0 2 G w2 VG3TX0P9@7˜ef‹4eb…SE»j4eb6Ghv@EXA13I6ÊtS8785‹4eb’cj4eb6Gh©P04Sd7VG¹T~v6™354VA1`I—4SaX9@i@…S48bG 9 T 0 G T7b AI 0 f 97 F 0g G 0 G TAI 0 f 9W „ ' 0 5 „A2 97 RI G7 0V1hef53c8bVGctpÇ£4@ȜŠ…†…dùdù28¹‡œ£E@e– #Æ Õ P0c@d7vG`T~Q@@a354VA1”48baS48b§6A1±qeIchlW 5I 0 5 R ™A27 ™š •  ù •š›› Æ 9W „ ' 27 RI 0 5I 0 G G7 G 0 5g 0 2 @37w~¹Gƒr6534VA1`I—413I6we0’@EebU354@P(4’DY@8746p3w@…@~@4}4jc8b6Ghae06T48bÈ@a414œE8bv60Pw8FSmÉ 0 5 „A2 T 27 G 0b Fb w 5 w7 27 T7b TAI 0 f 0 G RI 02I2 G7 G GI 64jc8b6GhXG3FXwPF4hPfr±®P0c@d7vG`T~ ­' P04@87VGhXYX9@E1XI8FHG4416I44‡h1jrj ¯ P0c@|{ TAI 0 f 9g F 9W „ 9W T 972 FA2 w2p 0 fI  9W7  9 7 F 1Hq8F0 6Gg S8785“16I`w' u I‡”« @ãw« I‡d 0 w G2 „ •ã   ° “ã 㠏 @EXA13I6Êt@3Te78560…Õ 3g80vG`TSI { 1ã¡°d« 1ãw« 1ã¡° 0G „ ͏Ž ° @ã °d  ° ÍSã¡ S¡zã¡d«« 11㏏ ¡zã w°° « @i°Í Ž  eA@7t—ug `Ydƒ}UäE„7 ”A' 2 5 0 G @7 h€ @ A wT f 5o4S z z 1¡°  °ã ° 35@X9 0 4Ey@U3IˆFp79 PA 3ÊHG3FlPw0 XF@4e79g h{ `fƒ gÄ z z ° â à Þ 354VAv„ 0 5 AI 0 —48bDGoÆ  9W7 { F I T F TAI 0 f 9g F 9W „ ' RI T j ¯ P04S|’P2o2 u 4eb}XT}4jc8b6GhXG3FXwPF4hPf‡X04@87VG`T|SmDGe0á'  «à ÁÞ xâ ® “± H° —¯­ ° « ß â ®d 1©H° —¯­ ° Á à ß ® €zŽH° —¯­ ° á}Þ HRpXYc1’4@|XR‡P0c@d7vG`T~al”8IU4—48bDG˜de06T48b444ac14cIdc…1XI8F6G6wj F 92I A27 F 9W „ ' 0W G TAI 0 f 0 G TA2C 02I™•   2 0 X2p32@Xq1†TQ#Æ Õ 3g80vG`TSI u ‚@reTSº739m48bd8Ir41PIeF6G4ctpsSz7e0~1Hq8FHGeF1lÛ@r˜Vf35e0˜P2u`YXGXFP94@87VG¹T~' F 0 F( ‰ Æ „ GRI T w 0 G G T2 w2 R9 5 0 TIgRI T G F FW „ 5 R G 9g T F R 7 ( F 9gg7 Y € Fi7 57 F(7 F G 0 0 G T F G2 w F F f @tIvVG6T80wX04hPfeF’HYe9X5@½7˜x4X2—XY44S©ˆ6PTjœ…VYSEP21SqXfee0X58F“32ec8b”4X2@d7“16IvHGBXR~Pù9e!—–D•—“ Ý A27 ™02 ý 9g f w 0 G F Fi7 9 5 0 G7 02 0 G T F G2 w FW T RI 2 F 5 0 4SÄS4@P74Ü9 g  Q…iP04h13I”4eb©P2‡PTjœBSz7e0”@Xq8Fœ“3(4”4ebv4P2Sd7“13Io`YXGXFP94Sd7VG@‹1XI13(V0hc8bG HR”Pù9e!¹–D•—“ “ª PTkjc8b6Gh¼' DPù9e!—–D•—“m~lseIgPAS87”PTk`YVG65t@eI4hPT4ebgebXGF u j4eb6Ghn' F f F AI 0 f f  É 0W G F T F 0g 5g F G AI 0 f  9W T G 0b F 57 G7 @X04@87VGxHY89“32V03544X2ae0@~S48bG ‘T}‡‡cX9@Ä1tIrX04@87VG|XTÈXGEHRÇPù9e!¹–D•~ž ÁeIlœ—š™˜f•@²œdš‚—™XTu—48bDG˜€@37PwVF35h4€Ã' ‰ ˆ † „ 975 R 9W T F F F f “ e–› ˜…› ›ƒ F AI 0 f9 0 fp2 û2 F2 0A 0 G 5 T ™ 9g f7 0 5I¸ TAI 0 f ( F 57 0 GRI 5 G ‚1XI8FHG4E6Ccdc8bx8bHGF u vG@d7vGw0 u SP04h@cDi}1EB64jc8b6Ghe4X24dbewSq¹fD„hPf8FÈ@3TV048pœz7t0R `YHGPFP9c@d7vG‚@8765432@‡48b‡4EDC4’4@VT43pVwPT4”8I~˜S87t5‡30462e065t06Re0“62P0P96930jDi’@U@”6T1vq35e0}XT4}{ FW T9 0 0( 0 G 02 0A A27 T FA G 0 f R w 5 G w 0 27 T7 0 T Fb @4@P74qh…iP04h16I~c8b”P2yPTjœuVYSEP21SqXf~c8b“r™@BfG6RX048ba8Iy1€‚1P0PFrvG4eB3T4Xp@S“321X(3Fqýu 02 9g ý 9g f w 0 G F Fi7 57 F(7 F 0 G RI 9 0 G G 5I™2I 9 Tp f 0 97q 0 0 9P@lS’8TvG@EcSz7e048b’œE8b’“62˜6fe0P5c—6¥e0Uc8b”eIe3A80“4@˜@|XG“…3fe0VG—VYc321X(3Fe06GX04˜@3IviebXGF 97RI 57g9 5 0 G G7 G G 0 Fp 5 0 G G G2pI f7 F™ T T 0 0 9g f w 7 eT6Ge0©6AP04c13I~1tI£‚c@q ¯ 1XI8F6G6wj©X2‹3A4EDC4v@7 u ‘T}‡aS‡6GV0qB@’`YHGPFP9c@d7vGa€62e06544`2ƒ T 9gp w 5 R™A27 2 0 F 02 0A T ‰ ˆ † „ RI T 7 RI FW T G 0b × âÓÒ×Ü× Ô ÐÓ ÙÚ ßÝ Ü Ù ÙØ ÐÒ× Õ ÔÓÒ Ñ Ð Ï #rÄÇBÇávy”ràۀ}ސ3ÒÛrÚ”4ÇñÖxBÄv”m’Î s Ö — ˆ– lÛh‘ Ú V 'le¢6%±”؉–8”'kleb6š%•y¥—‘…–­•™h˜ k ¶R– µH Ù × t ” ¶R– µ kR ˜ ” ” Ö ˜ —Œ U …„‚6‘Òƒ ‚ “Í ±° .

3 BR ¨ & 3 $ ï F B H 0 0 3 B 7 & ¨ ì  3 R ” ¨  ¨ ¥3$ R3ƒ & 4  ï @& ¨ $ °"8Y& &Rrë9"3ì‰0 E!XAXï A0§‰…P””"3©&2WR§Pì ©R3C¨AP7¤©q8 (g@ ¥3$ $ A0°‚39C¨'7R”Ê&#B P87VF R°Q8H3#0 6P©7@F Cï²%9C#A7R‰&B vPAX& ÿ (7 & ƒ%©44ëë7 R6IR6)¨3H73ìX l‰¦1V©PF$¨ )ï4 •RRQC¨370 3 A"98&˜ì871¨1#R'¨X ƒ"@'5R4P¥33#B PqAX$7 ©# & 4"¨ ¦ Q"¨1VCBËC¦64¥y& ÿ뇩PPC¦7#¨E %C3PRAX@ "3êBP§9Qr13ø#8ðï0& ”# X RRê3FX 6r4I"3#3A0@ 6%8&A7H ©U§B¦$# Qd ¥Yë Qƒj # &6 5@tI@@º7h14—HY1tPweFVGPTVF35e06G@e7@Edbc8b”eIÄ@—eI~’XT’SEebS6™41PIVF44Vg|1€@EXg3F4X2eF4S’ú€Sl@cPp@Sq R9 F fI2 9Ig w 57 w 0 G G GI 5 7 F G7 G Tp 5p T 5I97 w 5g™   Y27RI 0 97 0ecPp@eI4@ècePfjœ˜p4eb8IeVT35¨ t06RV05 Æ   Æ e065u m©jz  k i€X21lp48b§141VI48bu6Tj@e41eI€@3Iw G 9 TW7 fp Fi7 f 0 G G 0 ÷  ¨ ý G FIg 0 G b(p 5 G 0I( 5p G2 (cP28FHG4ep6Te0…48bhSEebÃdb4VpU1° Æ   Æ dbPw4b 1tIR gDiX04h13I…48b”@”43p—PI…48b”@XI4seIiPTF 9 5 0 G G7 G w T ™ F 5 ý 9g f w 0 G RI T w 9 0 G G 9g G AI 0 f ( F 57 0 G 7RI 0g 5g FW T 0 G G 5g G G F 0q2 w 0q —4eb6Ghw4X24dbewSq¹fD„hPf8FdÈ@6TP08FvG35l1eIc”`YXGPFX94@87VG”c8b}“32V03Te0ce8I˜±Y7 u “32X0432@€@3IpvY35@y' ‚˜ ™h%TRk ê Ô è ç ÕIé0 64—48bDG˜Pù9e!¹–D•iž ÁeI– TAI 0 f f “ e  › š˜•˜ … › š  T 7 ™ 0 5 IR 0 5 0 b G ™ I G A 0 5 5 0R 0 5 0 5 7 T A I b G 0 f b w p   T 0 w 2 0 p ¥ 0 T GF wF9 g i 0 R I g p 0 A 7 £œ—½f@²œdú@ESe1tDe648—48x3Ve6t6e”e@}cj4e6h’d4j}636434€6eUH3XP4jDÇSx4v4@˜f TAI 0 f 5 5 5 F 5g A27 TAI 0 f 9g 0 97 0A w F Fb {  9W T 972 cj4eb6Ghw@8I6G6we0V513I¹wv„1eI6GPwc3Ae0cs4@u4—48bDG˜dXG6FPwXF4jDiyP9Su6T44Pp694P2wPT4}xSP04Sd7VGXYX9@E1XI8FG „ 4416I44el‡35@6q4©Sz7‡´`YXG44mPT»©P2¼ef6580”654eA@‡VG6T41P(ce48bB@v€62P06F˜Dˆj6Ii48b#dbXw4b u FA2 w2p 0W 0 02 2 w Æ F2p F ˆ F G 0 5I 0b Fb 0 G RI G w 0 w 0 G F 5 R 9 F fI2 9Ig @tIBSº7h14—XY1lxPweFVGXTeF35e06G@e7@Edb©Q4P2—zqEi4—4eb6Gh˜œE8bx1±qVI4‹8IeHG43phXˆc˜S4ePTh¹Gƒ w 57 w 7 ( F 7b TAI 0 f G7 G 0 5g G 9 w FA GI2 F  FW ƒRI T w9 0 T 0 G 5 R 5 02 F 2 T FA 0 5 R I2™ 5 0b 1`YHGPFX94@d7vG`T—„Ä@ae0@37Û@º76F3wlega4ebU1tI’VY@e7eT3T603w4}XTo@PIeFVT43pVwPT46548bvG4tpi4s@e0@tI6Re0654}{ @P0c@d7vG`T~thVIPT@pVG4upXGP04@87VG¹T—„#ts8I×48bDG˜hq1tIpS48bþ2 u 4eb»3260tq@Ex¹Gƒ 9W „ ' 0W 97 Tp f F 9W ƒ 0W G AI 0 f 7 5 R G7 G I T 0W T7b 335ceA1’1X(4hXYP9VF@87e5XG4VW@7 0 5I b Fb 5 F 5 R@h6YS8746p3w@q@v@±q4qdbPwcb u VGXTjDir4jc8b6Gh¼Æ Õ P0c@d7vG`T„ Bª cI u p6A3030j…iU@4c@z7hjc8b6Ghf I w 5 w7 27 0 7b F F 0 TAI 0 f Æ 9W Ü ™ G w 0 GI22 w AI 0 P04@87VG¹T~h@}@sDY@87c3p3wS@d354VA1q4eb§1c148bHG”…@60@Eegd35e06Gh@87SEa® Þå â hnà­ ˜S87qc8bG 9W „ ' 27 RI w 5 w7 RI 0 5I 0 G b(pI 9 '  w7 T 0 f 57g å 0 f T 0 V0@EebX04@d7vG`T„ Bª 4SmX04@87VG`TmV0@SE8b†T§Æ Õ 80Se743p6w@U654eA@`Ij4@3I3wV0c‚6g80vG`TSI u jS30462m÷ 57 T 9W Ü A27 9W „ ' 57 G7 G ‰ Æ G 5 w7 0 5 „A2 T™ „ G™ w 0  ®Ìj —¯­ ° °« ßâ ®“I —¯­ z ° þ« Þ  x â XTodbXw4b u 6Ge0‡h@e7dc8b”8Ia30446Ae03TP0eFXGXF@E€3¥4X2|SmeT6Ge0hI u ac8b’4@7 F F T 0 f T 0 G G wp 5 97p 0 F RI T G 0 G A2 ¨  Á ° Þ «â à 2 FA2 1PIeFXG44@3Iw 0 G YW G 5 57 å 0 f 57g 0 G 0 Ib™ w 5 w7 0 5 „A2 T 5I¸  c8bx“’3A80œº739V0ae0@® ¥àå â |ŠÞ­ eT6580DG˜@e7@Ea4ebl6™351Dq0 u 4s@6Y@e743p6w@”654eA@`Ij4@3I3wV0}1EBc± H° ¯ eIpzŽX° ¯ @48b‡“321cP2eFe5VGaeT6TP0iV0@’”•X° ¯ eIG ¯ H° ¯ 3TX08FHGPF@E€6¥4P2i4ebÈ@6Y@e743p6w@d354VA1`Iv„VGVTE”1E¸ G 27 G G 0( T 9 57 97p 0 F 0 G™ w 5 w7 0 5 5C 5I ®“”•H° —¯­ x à « « Þ ß âvz ®“”“H° —¯­ ° Á à ß ® ¯ H° —¯­ ° á}Þ XRpHY41’4@|XR‡X04@87VG`T„ “ª XTè#Æ Õ 3g80vG`TSI u G F 92I A27 F 9W Ü F Æ „ 7 G7 I T 2 w 02I T2 w2 R 9 5 0 TIg ( Tp G FW T 0 G G ( F 5 0 aœE8bG u 4eb‹@37e41‡4@PI8FDG44tpÈ@z7V0e1Hq8FXGVF1lx4P2VF4aVG6T80x`YXGXFP94Sd7VGic8bp8IpcP24V248pDG~Þ •Í° Ó à Ð Ú Ñ ý t Ó Ú Ñ â Ý Î Ó Ï Ò Ù × Ò Ñ Ó Î à Ý Î Ü ÒÙ Ù Ø Ð Ò × Ú Ð ÎÙ Ô Ó â r à Ö Ö zmvÛBèÎ ¤uÈB}ÞmĤrà…|”B’e}}ސ3ÛrÚ”4ۀv|Dro|Ík“ªÌ±Õ ¥lj ™‰‘ h‘¿8˜” hIWyy“ÑI…–'”'kleb6µ ‰ˆ6‘Òƒ k i—  ˆ– s’ 旖˜— ‘ t ” ¶R– ‚ ‚ .

û © ø ¢ 2  ÷  ¤ ù $ ‰ ˆi ¥¤  ¦¦ ¢   ¢ $   8 6  ù¦ … )  E  ù õ  ¢ ø ù øô    ù ù $   ø  ù¦ ¤ ¢    ô 2  $   ù ‡XQH11‡‰xb£  ¡vd"¡yw ¨¥vIHc‰£–ö1‘TS¥x‡bI£}(7eTrb‹‡"WöV¥a‡§°3V£0") ù $     ¢ $   ù Eô    ˜ û  2  $   ù )  ô Eô¦ … )ô ù¦ ¤ ¢    ©¦¦ ¢  ô  ô 2   E  ÷ ø   ø ÷      E ©  ô¦ô ¤ ¢    ©  ¢ ¢£“IH‹X¨e¥—S`1¡1£d„„¨`§71~¨mö1IaXx5§¥1§(¨b1¥T11ƒ"!Xb‡ˆ¥T~¨ö¨b¥£XP¥~) ¥A¥T§b¡‹¨Bx0¡10"`ö1§Q¨`§öV~¨„b‹‡"b‡ˆ¥T¢5ö¨b¥a‡xb£!{Vgý´kb7©ó  ¦ ¢ E ô … ©   ô   @ û 2  $   ù ) ø ù ¦ ÷   ô E ô ¦ … )ô ù $   ø  ø ÷      E ©  ô ¦ô ¤ ¢    ù $    C  $  E üû ú û õô m } g z i o w it utr q wt or ¡ ž i m Ÿrr ž h gt it } h g u h j k gn m k j g i h g sR|{ye"vvfyv¢"s1(0 qvHl‰"{qsl‰slv†Bgyl‰(9‰&f û ©¦ ù 4ô   E …  ù ø œ  ™ 2  œ ›š ™      ÷ ƒ¨Tƒ¨(0Tù 0(£` y‡ ˆ‡ ™1¥¢ ¥—„ P“V‡ ô  ¢ © ø ¢ ô õ ¢ )ô ù $     ô   ø  … ¢ ù 2 ÷¦ E ô ` ˜û ú û õô ó `  2  $   ù ) ¢     ÷ ” ' ù õ  ÷ “ ø ù 2 ø  ' $   ø ÷  ¨´y‘Xƒ1§‡ƒ~§m¢£|¥P¨e‡‡9W‹b1ö01§¢VG|bb7©£013V£0"WaXVm€–‡b1ۓX¢‡B3£‡b‡" 2  ¢ ' 2 øô $   ù $     ¢ $   ø ¢¦ ÷ Eô   ø ¢ … ô ù Eô    ˜ û ˜û ú 2  ¢ dû ú û  õô ó ù ù  ` Ž ü ” “ ` F ” “ ` õ ¢ 2 – 1I!3Xöbeb£€IH—ƒ§1X¨e‡ƒH‚¨X¨e¥—¢—´°1Ie|0`ö©r`(‰dˆm—ƒV•—ƒ1QHT¥¢ • `1‡X{V¥U715b’Q9&¥V‘V—XXb¥R¥X‡d(1–Ž ‡ €1` ‡ ‹V¨býƒ{‘010ö`H0b§b•ù ƒHX“13V£0") õ  ø … ¢ ù¦ `û õû ù Œ      2  ø ù $    ¦ ¢ ø ù 4 ù  ` ” “ ’  `û õû ù Œ © E  ù ô E … ù 2  ô $   ø ¢ $   2  $   ù  ô Eô¦ …  ù ¦ ¢ ø ù 4  $ õ ÷  $  ¦  û ) ù¦ ¤  ø …  ô   E ù 4   Eô ¤ ¦ ù 2  ) ù $   ø  ù¦ ¤ ¢     ÷ ô  ô   2  ¨X¨ö13–§¥£`‡0ù b‡V£¨€•€0ö1Xbc§(00‡T¨bf0b¡~gbQ‡"W71¥a‡11•¨m¨šXc1¥¢ Ž ô õô ø  ù $   ø    … ù E  ùŒ ô  ¢ © ø ¢ ô õ ¢ )ô ù $      ø ¢ …    ù 2 ÷¦ E ô 2  $   ù ) ô $   ` ø ù 4 ù C  †§ƒö(ƒeb€‡"V0X3d{g¨¢{7Xƒ1§‡ƒ~§b£AI‹XQHP1€Tb17T1¨~¡1£d„€§bƒTXƒ0h(€Š û¦ ¢  ô  ô 2   E ô  ô ` ø  ø ù $   2  ¢ ` © ø ¢ 2  ÷  ¤ ù $ ‰ ˆi x ô ©  ô¦ô ¤ ¢      ô õ ø ù $ ƒ n¥1§(¨b1¥T“§5¨t¥ù ‡"0ù Xb7V¥†‡ƒH11ƒ9•b  ¡vd qi ¨s5ö§¢¥a‡f¥7¨‡Xù ˆb“!û ‡ û ù  ¢¦ … ' 8 6  ù¦ … )  E ¦ ¢ $ '   ù¦ ù $     ô   ø  … ù  ô  … ¢  ù  ¦ E  ù ø ÷      E ù $ ƒ Qb¥§13‰†d71‘T|5¥H‡q07¢£l¥v§(X‡‰•e5bHWQ0(¨T10Ûb‡£0•b“!û „ þ  2  $   ù )  ô Eô¦ …  ù ø   ø ÷      E ©  ô¦ô ¤ ¢    ¦¦ ¢  ¦ ¢ Eô … ©   ô  ô  © ¢ C  C XV310"“¨`§713d‡"!bƒ£T‘5ö§¢¥a‡ƒ§¥f¥R¥T§b¡#¨¨B3U‘S  õ  ô C ¦ ¦  ù $    @ 2  $   ù ) ø ù ¦ ÷   ô E ô¦ …  ù ù $   ø  ø ÷      E ©  ô ¦ô ¤ ¢    ù $    C  $  ¢ üû û õô b§Û(¨§"b£‚“û 310"€`ö1§€¨`§713d”¢£!‡"€bƒ£TP¨ö¨b¥£Xb£Q!(1X~ý´ú b7©ó “ c m } g z i o w it utr q p o k gn m k j g i h g sR|{yxqvtsRP†Bgyl‰(9‰&f û üû dû ˜   ô   E ù ÿ ô  ù 4ô 2  $   ù ) ' %     ø ' ù   ù $   )  ø 2 ù 4 ø ù 2 ô $ E ô $ C ` üû ú û õô bý´e´™¥§edT´€¨¨0‡5‡õ 310"(—¥3X–b•b£”{7Tƒ¨ô Xb“¨Ëa§¢!V1ý´£b7©ó ô 2 ù   ¢ ø    ÷¦¦ô ù ø ¢  2  $   ù )  ô Eô¦ … )ô 2  ¢  ô Eô¦ …  ù $    ¤ ø   ø ÷      E ©  ô¦ô ¤ ¢    ¦ ¢ Eô … © ¨rTey£e‡1ö¨§Û(QSV310"’5X§71‘§"1¥„¨`§713d"£¥‰D‡"„Xb‡ˆ¥TW5ö¨b¥a‡†¥U¨b¡„ƒ û ¡y€ yavt rqT ‚  w x w u s ipi h ¢  ¢ ` ù ø  ù ø ù $   `  ô    ø ù ù ø gISI(‡"d(`b£fT¨ÇeX{d(šù cQö1IaXV1”¨¨e0Xb¥PbY7§b!71¥£XW¨›3V£0") b û ù¦ ¤ ¢     ÷ ô  ô ` ø ù $    ù $     ù¦ô $ C ù¦ ¤ ¢    ô 2  $   ù ¦ ¢ Eô ø ù ) ÷  ù $   ø ÷      E ô $    ù 2ô  ù       ¢ $   Fû  ô   E ù ¥U¨(X"1%Q¢£bƒ£TS§bR¥Q¢§(QbPIH&G|ú §(00´ÿ {D!(§§¥"#BA0970‡5‡”¢ )  ø  C ¦¦   @ û 8 6  ù 4ô õ • E f• —–D•“ û 310"(&¢£#‡"!bƒ£¨7§b¥£¡hbý´ˆ(¢‡ö©ó 2  $   ù ) ' % ù $   ø   ø ÷       ©  ô¦ ô ¤ ¢   ÿ þ üû ú ù ø ÷ õ ô òñ ð a) Euler Explicit ï í × âÓÒ×Ü× Ô ÐÓ ÙÚ ßÝ Ü Ù ÙØ ÐÒ× Õ ÔÓÒ Ñ Ð Ï #rÄÇBÇávy”ràۀ}ސ3ÒÛrÚ”4ÇñÖxBÄv”m’Î I( λ h) î ï Stable 1 Unstable R(λ h) í î θ=0 Stable b) Trapezoid Implicit I λ h) ( Unstable θ = 1/2 R(λ h) I λ h) ( c) Euler Implicit Unstable 1 Stable θ=1 R( λ h) zŽ !z° .

0 R(λ h) -2.0 1.0 .û  2  $   ù ) ¢     ÷ ” ' ù õ  ÷ “ ø   ø ÷       ©  ô¦ô ¤ ¢   ÿ þ ˜û ú ù ø ÷ õô `1¡1£d„gaXVm€–‡b1Û!‡"!b‡¨ö¨b¥£¡Ø¢—´”(bƒö©ó -3.0 -1.0 û  2  $   ù )  ô Eô¦ …  ù ù )   ø   ø ÷       ©  ô¦ô ¤ ¢   ÿ þ dû ú ø ÷ õô XV310"“¨`§713dm"“‡"!¢‡£P¨7§b¥£¡Ø´G´ˆù b‡öqó „ ü „ €‹RDº ® ª ² ³ ¾ ‰6 ½ ® ³ º ­ ¹ ± ® ¼ µ ª ° ´ µ º ® ± ª ¹ ± · µ ° ³ ° ¶ ² µ ´ ³ ² ± ° ¯ ® ­ ¬ ª § ¨§ ’lD‚!“¸€s»€†fQD’W““¸!TR€Q1R ‹fd!„™fP«©(¦ Stable Regions RK1 RK2 RK3 RK4 I( λh) 3.0 2.

0 1.û Ž ¤ dû ú û õô ó ù ù  Œ 2  $   ù ) õ  ø … ¢ ù¦ ù $   ø    ¢ $      ø ¢¦ô )ô  © ø ù 4 ô © ø ¢ 2  ÷  ¤ ©  ô¦ô ¤ ¢       †IG´¢böq”`(v£¡10"g‡{VIUö¢£#‡"#yH£g£Dƒ§§"§(PXXƒ#§sXƒHV1‡‰~5ö¨b¥a‡!£B@ è ™ Ö 6 ø  sp x ù¦ ¤ ¢    ô   @ û û ¢F  ê è 7cTb!C vœ 炍‡" rHyw 71¥a‡Q¨mBQ¢F „ b1€¥¢ „ ´}71¥s¢§°!(1‚1¥¢ ê é  ù $ û    û ú ù¦ ¤ ¢ ƒ ô  C  $  2  á × Ý yàÛ Ú Þ àÛ Ú ü Þ × UàÛ Ú ß 8 F Ü „ Þ Ý Û ƒ× IHÚ × Ü Û Ö (HÚ ø    ¢ ø ù …    ô  … ù $   © ‡I£`E¥!ˆ¨‡‰¢£P¤  ù 4ô õ 2  $   ù ) ù ¦ô ã ø ù 2 ø ' $   ü ù $   ô ø ù $      û û ú ù¦ ¤ ¢ ƒ ` æ û  Œ 2  $   ù ) ø ù 2 ø ' 2 ø 0‡¨ƒÇ310"Wb¨§€`b‡B3£I}b£Q¨mXb¥1€eŽ „ ´"71¥sƒVeb1vÅ¡1£d„QXbX‡3XƒF    ¦ ÷  ã '  ) ¢ 2  ù $   ô ù  ù $    ù  Î û æû û õô ó ô  C  $  ù ø ¢ ô $      ô   ¢ ø    ÷¦¦ô  C ¥£¨V‡™‡€~IHQW¢£Q¨}(T¢£A¥W¢´G´ú b7©Ç§°!(1XW(ƒ‹§bA¥!1§(I£(X17§¨¨S#ƒ û ù¦ ¤ ¢    ©¦¦ ¢  ô  ô 2   E  ÷ ô  ô   ¢ $    ¢ ù )      ù  2  ô Eô¦ … )ô ô 2  $   ù ) ¢ ù  ÷ ¢ E ù ¤    ÷ ƒöV¥a‡¨¨¥H¨e¨¢1TV1”¨¨”IHY¥U"W¥VWT|1•¨`§71~¨”¨›3V£0"¢7(VƒU0Eg‡1ƒå m } g z i o w it utr q wt or ¡ ž i m Ÿrr ž h gt it } h g u k gn m k j g i h g sƒl{yeqtvf‘t‚vs1(0ä"vH|9"{qsl‰v†Bgyl‰(9‰&f û … ©  ù ¦ô ã ù $    ù ø ¢  2  $   ù )  ù $    $    ¤   ¢ $   ù Eô    ƒù ¡‹b§¨€&b£l¥•(ƒ!V310"&ù TbsI†¥‰€yH£‹`§eI€˜ á €× Þ Ý Û Ý yàÛ Ú Þ Ûà Ú Þ × UÛà Ú ß 8 F ƒ× IHÚ Ö × (HÚ Ü Ü Û ‡ á Ý àÛ €× IHÚ Þ × (HÚ Þ Ý Û Ü àÛ D8 ƒ× IHÚ Ö × (HÚ Ü Û ß  ù ù â   ù ÷ 2 2  $  ù ) ¢ 2  0Xg‘£bb7¡1£d„g™1¥¢ 2  $   ù ) ¦ ¢ 2ô  – … ¢ ø   … ù   '  C  ù $   ù ø ¢ ©  ø …  ø … ô $   ù 4 ¢ $   ¢ $    2  $   ù ) ø ù $     C ¡1£d„ƒIH§‡0Tù ¥£(™0eXBIS•¢£(ƒvXXù Xb“§bƒUH!IH!V310"!Xb¥~S#ƒ û ô  ¢ ô $      ù¦ ù $      ø    õ ô ©¦ 8 6 ø  ù¦ ¤ ¢    ô 2  $   ù ) ¦ ¢ Eô ø ù ) ÷  ù $   2  ¢ ô  X§´y&§¢£#¥&q07b™‡y™¢§|¨Y‰W‡"xöV¥a‡&¨y¡1£d„D¥T§(`„V|~b£‘1¥&§´y¢ © ø ¢ ô õ ¢ )ô ù $   ô © ø ¢ 2  ÷  ¤ ù $ ƒ û ¤ üû ú û õô ó ô  C  $  ô ù  ¢ E ô $   ø  © ø ¢ 2  ÷  ¤ ©  ô¦ô ¤ ¢   ‡ƒH¨‡ƒ~§“¢£A§r‡ƒH11ƒ9—b“’0¥’´|¢ö©„§Í!(1XD§v(¥TS¨b£A‡"g‡ƒH11ƒ9e¨ö¨b¥£X Ø ù $ ƒ b“Ùû × «Õa§b!x‡"«310"(¢b‹¥¢ù ¥U—¥Ô`Tù ™b£#5ƒýn5Ó310"!¥H§ƒTTù ¥e  Ö % $ E ô $ C ø  2  $   ù ) ' % ù $     ¢ E ¦ ¢ ô E …  ù $   `û ùû ô ` 2  $   ù ) ¦ ¢ 2 ô  – … ¢ ø ù $    ô ù ¦ ¤ ¢    ©¦    ù ø ù $  ô  ù 4¦ ù  ) ù $   ù ø ¢ Ð  Ï Î õ  ô   ¢ ø ù  ù õ ù $    ù $ ¢£P§Ëö1¥£XÀ¨£0(X¢1§š0‡¨0(‘Tb¢(ƒ7ѓ«b§(I£`bT‡Ç¢£ÒTb!C rHyw ö1IaX§ô sp x ù¦ ¤ ¢       ¢ $   2  $   ù ) ¢  ù¦ … ) ¢  ù E   ¢¦ E ù $ ƒ û ù¦ ¤ ¢     ÷ ©¦    ù ø ù $ ô ù ø ¢ õ ô   ¢ ø õ ù   ô ø ¢ © ù $ yH£¢¡1£d„v¥W71~I1–}§ô ¥§Txb“WQö1¥£X1V~¨T(`b1¨}(ƒv¢§ey£‡`e§}ù ƒc0¢£  Ð  Ï Î ù $   ø ù $ C   ô õ ø ù )   ô ù¦ ¤ ¢    ù ø ¢  2  $   ù ) ù ¦ ¤ ¢    ©¦ ¦ ¢   ô  ô 2   E  ÷ ¦¦ ¢    ѓ#bSù Xb!R1¨‡Xù h"X&¨Ø71¥£X„(ƒl1¡1£d„„öV¥a‡Q5§¥H¥§e5b1011‡¨¥f¥€˜ û  ô õ ù ‰p h ˤ ù $      ø ¢ … ô ô  ¢ © ø ¢ ô õ ¢ )ô ù $ 0§ƒX(ø §TÍd¥ ¢£l¥—XƒH“¨§´y~XQH§‡Q¨”¢£   ù  ¢ E ù  ù $    ¦¦ ¢  @ û úû ú û õô 0(¥T(0b£„¥’¨¥Y‘´˜´g¢ö©ó §ô !{Vx(ƒ£‡Xù ¢‘"¥‘¢£c£¨Û1¡10"‘71¥£X   C  $  ù ø ¢ ©  ø …  ø … ù ) ¢  ù $   $  ô C  2  $   ù ) ù¦ ¤ ¢   ©¦¦ ¢  ô  ô 2   E  ÷  ô Eô¦ … )ô ø ù $    ù )  ÿ ‰p h Ë ¤ ¥ x  sp¤ x ‰ ‚ ‰ Æ xi ‰ ‚ Ë ‰ ¤ Ë ‚ Ì ‰ ¤ xi w ¤ ˆ ‰ ¥  Ì xi ‰ 5§¥1§(¨b1¥T11B¨`§71~¨mXb¥g"3}û §TÍdV¡†rvHd£dHVqx£¥~¥¥ÍHq7™(†¡qddh ¥i ¤i ¤ Ë Æ¤ Ê É È Ç ‰ Ƥ x ‰ Æ Ä x ‰ à ù¦ ¤ ¢    ©¦¦ ¢ Eô ø ù ) ÷  ô 2  $   ù ) ù $     ¢ $     ¢ ù ) ô $ §q¥VqP¥ x‹SH"«dHfÅd¥d‰ ö1¥£X£¨¨¥U¨(`„1%W¨›¡1£d„vbWIHW1¥T„W¨b“ƒ û  2  $   ù )  ô Eô¦ … )ô ù¦ ¤ ¢    ©¦¦ ¢  ô  ô 2   E  `13V£0"S¨X¨ö1‘§m71¥£X‘¨¨¥H¨e¨¢1TV1÷ ‡ ¢£#‡"!bƒ£¨7§b¥£¡h´˜´ˆ(¢‡ö©ó ù $   ø   ø ÷       ©  ô¦ô ¤ ¢   ÿ þ úû ú ù ø ÷ õô a) 2nd Order Backward Implicit 1.0 R( λh) ‡ ü „ .0 2.0 2.0 4.0 2.0 I(λh) b) Adams Type 1.0 Unstable Stable Outside 2.0 3.0 ´ ­ ® µ ´ · ´ ¯ ² ® ª° ³ Á ¹ · µ ° ³° ¶ ² µ ´ § ¦ ¯ ® µ º ² ¼ ¢„s|D|‹AQ€R “¸!TR€Q1|À(¿Ds‹Q€’± I(λh) Stable Outside Unstable R(λ h) 1.

0 1.0 R(λ h) I(λh) 2.0 3.0 R(λ h) .0 Stable only on Imag Axis b) Milne 4th Order I(λh) 2.0 -2.Œ ft † g }n {a| j z n „ x j ht f ‰ ft f n f g † j … g k „ j ƒ n r  nt ht ft f n x n ht k j |t x f g z g h xt xt i h w ft l jt r q g o n l k j i h g f e v¡’‹€rŠa}„ˆ‡#}¡€rCru‚u€¡}„~~}y#au’¡{7H’yumu¡’§ruDvuhsp©ˆm#r’rF#d ‘ hÛ ‡ u e ‡ e ” ‡“ •Ü ˆÚ ‡“ ˆÚ p w – u e ™” vs rp i —‰ g ‚Ö ‘ w ˜•Ü w vs rp ‘ i —Ü i g fÖ ’UÛ – e – ” –‰ ‘‰ Ü ` ) ø ù   ¦ ¢ ø ù  ù õ ù $   ø  ` ù E ô ÿ û … ê b ¦ y ê ™XXeƒI£X¢T‡ˆb!ƒ"V¥`1§´h§†‚„ƒ‚gé ù õ  ¢ ø ù $   ô  ùô¦ b ¦ y 2  ¢ ¦ ¢ ø ô y ù ø ù $ C `  ô   ¢ ÷ X ‰ u x ‰ ‚ ‰ i ƒb¥£š¢£Y§T7öŁB€1¥#¥Uù §»V(X¢!DÓ§(IH%tù aÓd£d‹x ¥ˆ b¢£y¨eV7¢W§ô ù $       ô   ÷¦   ¢  ‡ û ú Ž ‡ ´nŒ w us e p i g e Ö Ž d c b vtrqhhfAXWFŒ Ú $ Eô $ C ø ù 2  ÷   ô  ô 2   a§¢!!`b11V§(¨b1¥TE ù $   õ ô 2  … © ¤ 2 ù ô ) ø ù   ù 2 ô ø ù C   ¢ ù $ ƒ ` ù )ô   ô © ¢ E ù 2 ø  C  ø õ ù 2 ÷  ô¦ … ) ¢   ô ¦¦ô ¢£b¨b1H™¢Tb¨„XX(0b—§m`I‡1¥gb“äa"§(™¨~UU0bQ‡Â(X‡¢1e5ö1‘¥€£¨k¨§žb þ §(XT¢|Yb£WˆWI 1¥ 0„b£Ybö¥I’¥§e5b§›Ix¥¨b‘ƒ1S¥§eI1¢W0(‰‘§ô  ô    ù ÷ X ù $    I  ¢ 2  ¢ $  ù ) ù $     ù ÷¦ ¢ 4 ¦ ¢ô  ô ô  ¢  ¢ Eô   ) ø ¢ $ ¦ ¢ô   ¢ …  ¢  ù   … ) ù    ù $ ƒ û ù 4  ¤ ¢  ù 4ô õ   ô   … ) ÷   ¢ ù $      õ ô 2 ø  E E ¢ ù E ¢ …  2  ¢ ù )ô     ÷  $ õ ÷  ø $ ¢ÕTb“¸Q‡(9IÒTƒ¨‡ÍV§(V~V(Ib£‘£yb¨bƒTTIX¥1Ò1¥™"§e~V‡1¥b‡¢£     ¢      E  ¢ ô $    ù  2  ¢ 2¦ ù … C  U ù $   ô   ô  … S¦ ¢ Eô … ©  Q ¥£X10~I¨b£0(1÷ 1¥À§Tb¿{1Vb«¨„¨‡9£TIU§¢3€R¢ X{£ayH0ƒP£™b‹Î ) ø ¢  ¢ 2  ù I ¢  ù  ' 9 G E ' 6 A 9 8 64 2 0 ( ' % B¥HFDCB@)7531)¥&$ "  # !  û ù  ¤ x ‰i u   ¢   ¢ ù )   © ¤ ô  ô   ÷ ¤ `  ô  ô 2   ƒb¥ HdÍdÒd¥ !1¥T„‘1v“§Û5vVbVÓ§(¨b1¥TE ©  ô¦ô ¤ ¢   s ‚ Ë ¥ ¥ ‰ u ‰ ù¦ ¤ ¢ô¦ ø ©¦ øô ¢ ¢  ¢  ù 4 ø ù   @ û  ) ù¦ ¤  ø … ¦ ¢ ø ù  ù õ ù ø  ) ø  2  $   ù ) 5ö§¢¥a‡ d¥TaaaaÓx 71¥§öTù Í5öQvx€¥0‡X`(—B`~0ö1¥bl¥£`bTƒm(‡‘€ƒ"†¡1£d„x¢ ¥&T¢§bX‡S~bvb§(I¨e‡Tˆ‡"xb§bƒ€WIäTù 1y0eq¥§”¨°¨eIT§711¥D¥T§(0¥bY@    ù ô $   ø  C ù $   õ ô   ¢ )ô    ù ø  ù 2ô ÷ õ ¢  ¢ 2  ÷  ù    ô  ô  ô   ¢ Eô¦ … … ¢ ¦ ¢ Eô   E ¢ ø …   ¥ x wi¤i ˆ x w u s ‚ Ë ˆ x  w h ui ˆ wi ‚ ‰  ‰ Ã Ë Æ ˆ x Ë ¨ ¥¤ x ‰i u ‰ w u Ì xi s ‚ Ë Ã ‰ i¤ ‚ w ‰ u Ë ¥ w x ‰ Ã Ë û VyÍvq¥‰ya£–¥‰3ydÍ¥T͖dHm¥¥V†‰¥‹qHdÍdÕ(a£¡qy–¥y‘t‘yva¡Vvӑ¥¥VÆ ¨©¥Óqr7£¥ÅaÍ¡fÄ —“x§™X1b¡¢~yH£gÑ!D¥¥W1¨eI~§´U¢1¥†X10(`ù £b‘y51 ‚ Ë ‰ x ip ‰ ‚ Ë Æ u i Æ Ð  ¦ Î ù E ÷ 2  ø …   ¢ $   Ð  Ï ¤    ô   ¢ ) ô   ø … … ¢ ù E  ù ø ¢ô 2    ©¦   ù ô¦ … … ¢   Ì x i   Ë ‰  ¥ sp ¤ u i ‚ ÿ û Ð  Ï Î  ù õ ¢    ù   ¢ ô 2 ù ) ø ù   ô  ¢   2  … ù 2      ù  2   Tö711¥e¨ô ¡qy¡H™rvdÍdv¤ cXÑ!—I‡ƒ£XxeyÔb0„XXe¨›¥Y¥ä1Tù Tb&¥1&T|1&¨ô 2  ¢  ø    ¢ ø …    ô  …     ÷  2 ù ø ø ¢ E ©¦¦ ¢ ÷  ÷ ô ô  ©¦ ¢  ¢ ô $ ƒ $ E ¢  ø … … ¢   ¢ ) ÷ ù ˜   4 ø 1IfX‡£IXù sˆ¨‡‰esV‡&Töô (ƒT•¨¨¥H(Vy¨q¨(¡¨IH¥s§b“’û a¥¢1¥WV¥VXBW|‡ ø ù ø ÷  ó ù $   ô  ù ) ù $ E  ¦ ¢ Eô ø ù ) ÷  ø  ô  ©¦ ¢  ¢ © ô¦ô ¤ ¢     ) ø  ø ¢¦ ÷ …  …     ) ù $   ø ¢ © Xöô b‡Hbs¨§T"TbaXV¥T§(`„V|s‡"s¨(¡¨IH¥¨7§bIaX†IþX‡"sQÔ1b¥9fX‘#bsƒvS’ ý ñ ý ù ø ö û U’†0e¢üú ù õ ñ ø ñ ÷ ö õ ó ð ò ñ ð ï î Ó0UUe’Pô0„‘ví ë ì dPë û  2  $   ù )  ô Eô¦ … )ô ù¦ ¤ ¢    ©¦¦ ¢  ô  ô 2   `1¡1£d„S¨`§öV~¨mö1¥£X‘¨¨¥H¨e5b10E ‡ b#‡"!b‡¨ö¨b¥£¡Ø¢—´”(bƒö©ó ù $   ø   ø ÷       ©  ô¦ô ¤ ¢   ÿ þ æû ú ù ø ÷ õô a) AM3 F ü „ ´° ´ · ³ ² ª ² · µ ° ³° ¶ ² µ ´ ¯ ®° ¯ ¬ ¹ Á § ¦§ ‡3Al‹—‹Å!TR€‹V|¸D€!sx„À("(¦ -3.0 Stable -1.0 1.0 3.

ÉÈ Ç Å Ä Œ —~¢Æ ¶qà  À À x n | j h f j x j k „ f j hq g xt f gt ht | f g z j i h ½ ft n † g | qƒÁ W’tv’#€r¿#’vm´ˆu’}a¡{7r’5¾u¼{a‹j u’y²vŠƒ˜¶}„W’urt#€m´~’Wra#r¡rP³²v° †t h »º¹ ¸ · n ft | j kt ƒ µ j k xt x x j f | j | f ƒ g … q± û d ¦ 2  ¢ ¥„VV¥‚y £ ` §I!‡"ö1¥£X1Vv¨BVTe‡xb§bV¥ÔXX‡(s¥73V£0"——ƒ¥1XƒHa§Û†U‡T(~5ö§¢¥a‡“Xöô b‡Hó ¦¦ ¢ ø  ù¦ ¤ ¢     ÷ ô  … ù    õ ô … … ¢¦ ø ù 4   2  $   ù ) Ð    2 ø ¢ $ Eô “ `    ù   ©  ô¦ô ¤ ¢    ø ù ø ÷  ù $   © ¤   ¢ $   ` ù ø  ù ø ù $   ` 2  … ù b ¢£YxIHA¥(ƒ"0(`b£Dƒ11Py£û „ ¡U¨¥x¨g“ &¢gIHeƒTöTg¨5ô a¨b!C ¥{ ¯  © ¢ C¦ ¢ ô ’ ’ ù     ¢ $   ø ¢ ù¦ E ô   $ Eô $ ) ø ® ® ‹¥ ­ !× „ Þ Ø  ™ Ö Ø  ù ø ¢ (ƒgd ‡ ´ú û ¥¥3X€Sxb“x´d ‡ ´yty#¥¥¡ø •§ô ~¨ô ü ‡ ´yty#¥¨eV77•§ô ‡ ´qty&1%“ƒ     ø  C  ù $ ƒ û û ú û X       ¢   û ú û X    ô   ÷¦   ¢  ‡ û ú û X   ÷ $ Ž û ú d ‡ ´nŒ é Ö „ ¥ ¬  79„B‚0E ¨Ž b ¦ y   ¤ ‡ « 7ª ¥ © „Œ d „¦ b ¦ Ø ¦ §Þ £ ü Ø  ø ‡ á w – vs Ý e u b Ø „¦ €Þ £ Þ ‡ ¥  Ö w – vs e ß u × Ý ‡‚ d „¦ ‡  ô   ¢¦ ù ø ù $    ù 4ô õ û ú X      ‡ û û X    ô   ÷  ô    ¤ ÷ §(IÔ0(ˆb!0‡5‡}ü ‡ ´Eû 3‘£¨ô ¥‡ ´ú t3f¥7§(Ve5eXV1´ÿ Ž û ú ˆü ‡ ´nŒ á × Ý ‡“ Ú Þ ‘ hÛ ‘ aÛ ‡“ Ú ‡ b ¥™× £ Ü ‡ “ Ú ß Ø „¦ ¤Þ × ‡ “ Ú Ö × ‡ “ Ú ‘ yÛ Ý ‘ (Û Ü d „¦ ‡ ‘ aÛ þ „ bü û ût3g¥ 0‡5‡»1¥¢ „ b›¥§edT´»§»T¢§(ˆ0„e¥#&§b“£X10e‡¢b§¢1¥ÔXXƒ{!IY3V£0") X   ¢  ù 4ô õ 2  û ü  ô   E ù ÿ ô 2 ù  ô    ù )  ¢ C ô $ ƒ û  … ù    õ ô … … ¢¦ ø ù 4   2  $   ù Ð    2 ø ¢ $ E ô “  ¢  C   I  ô   ¢ ÷ X ù  ô  ÷ ¢ô 2 ¦ ù 2  ) ù $        ô   ¢ )ô   ø … … ¢ ù E  ø ù ¢ô 2 ù  ô  уX1ƒ1a§Û’¥x!(1ˆ„¨eI1|{”¨(1E¢¢¡0b¡~#¢£Q•¥§ey¨¢UXb1¥“`1Tù XE¢bSe5bH… ù $   ù¦ … ) ¢  ù   ø … ¢  ¢ ø ù 2     û  ù¦ … ) ¢  ù © ¤ 2 ù ô ¢¦ …  ù ¤    ù ¤  ¢ E ø ÷ 2 ù E  ø … ù $ ¢£WöV~¥V–‡£P•¥`b§ô 1x`TöV~¥V–Ç£0b§QÔ13d}ù X0E£¥Uxù bbT`3Xbxb“ƒ 2 '¡ Ÿ ™ 0  › ' ™ A Fv§ y‹žœš“˜— –  ! !  û • 2 ù … ô   ¢  ô •T£‡§ey£“§}F ‡ ´¨¨e5b10HT`ed„¥ƒHs¥ û ú  ô  ô 2   E `  ø ù   ù ) ¢ ø ¢ …  ô   ¢ ô ¤ )  E ù¦ ¤    …    ø  C ù $   ô `   ¢ $   ù ø  ù I ¢ ) ` ©  ô¦ô ¤ ¢    ø   ô  ô 2   E © ø ¢   ù E ù §(IH§¢WTg71§ô ‰XX‡gb§kUI1£g(b÷ xƒP~ƒ5ö¨b¥a‡‡"¢¨e¨¢1T£‡ƒ£XTX0b ¢  ¢ ` 2  2  $   ù )  ù 4ô ©  © 2 ù E ÷ 2  ø … Ð  ù $   ø  ù 4¦       ) ù ¦ ¤  ø … ù $   2  ~¥!ƒ1I¢ ¡1£d„ÅT‡5‡õ ¥¢ ¤ 0X1¢3Xbђ”¢£x‡"†ƒ¨XY£g¨ô 0ö1¥b†b£Å1¥¢ Ž û ú F ‡ ´nŒ „ ê ’ ’ ©  ô¦ô ¤ ¢    ¦ ¢ Eô ø ù ) ÷  ø  ˆ“ @5ö¨b¥a‡ƒ¥T§(`„V|“ƒ†ó á i g e ß   ¢ $   ø ¢ ù¦ E ô  ô IH#ƒU7T“§Û5b` Û ‚Ö  Û – i – hRV g e ‘ i e ù E ô  `  ù $ Ö hg pXV§(†ƒ0b“ƒ þ  ÷ $   `  © ¤    ù  ø … ø ù C $ Eô $ XV|1’‹0(Tù bTù ˆI7a§b!C i – g ŽXXeb†VHhHT1¥§(XTù b3dù e ) ø ù   ù $   2  … ù C `   ô   ø …  õ ô ô ¢ ) ù ø ù $   @ û   ÷  2 ù ø    E ¢ ù ¤  ¢ E 2  b§b¨ƒ~T(vb’†3‡0(‡0¥v—E~¥UPV¥¢ `e‘XXƒ0£x~‘0S¨ô aÛ DÚ 5e¥H%¢£  ) ø ù   © ø ù 4 ù      ) )  E  ‘ ‡“ ©   ô    ¢ ÷ X ù $ ´ ­ ® µ ´ · ´ ¯ ² ® ª° ³ Á ¹ · µ ° ³° ¶ ² µ ´ § ¦ ¯ ® µ º ² ¼ ¢„s|D|‹AQ€R “¸!TR€Q1|À(¿Ds‹Q€’± ü ‡ü „ .

Ž û ú ú ‡ ´nŒ Ø b â £ ÚØ â Ö Ú â d â ô  ©¦ ¢  ¢  ô  ÷ ¢ô 2 ø   ô   ¢ ÷ X ù ¦ ù 2  ) ù $   ø ù 2    §(¡5¥H¥†¨(1E¢b“ƒ"7¥§eyH%t00b3‘•b£“`b§ô 1 ù à û ò õ ý ñ ý û î “áfsU’‘”ß Þ ì Pë û   ô  ô 2   E $ E  ø  ù¦ ¤ ¢     ÷  © ¢ C¦ ¢ ô 2  $   ù ) ô $   ` üû ú û õô ó  X1¨e5b10‚a1÷ “‡"ö1¥£X1Vv¡U¨¥!¨¨¡10"“§b11’|£b7©"£  õ ô 2 ø  E E ¢ ` 2  ¢ $ E ø ¢ )' ù )ô   ù $   ø  2 ù  ÷ õ ô ¤ ô 2  $   ù ) ø ù¦ ÷   ô Eô¦ …  ù ù $    ¢ E ô $    ¢§bX‡T0¥¡ƒ1¥eaƒ~–„¨e#b’‡"‘0(1m¢§Xù ’¨}310"AX71§A¨`§713db!ù ¥Uf§bW§ô ` ø ù 4 ù C  Š û  ù ÷¦ ¢ 4  ù õô ù © ø ¢ ô õ ¢ )ô ù ø ÷ …  ¢ $  ø   ¢ ) ô $   ` ’ ô 2  … …     õ ô 2 ø  E E ¢ 2  0X‡dI(—DXT¢ö¥IT‡7Xr‡ƒH§ƒƒ¨m(b¢#¥H~5ô eI~¨b£3Q7©b1Tù 1€W£Wb¨bƒTTI71¥¢ ¥ `dŽ „ •ac „ ¡“˜¨"˜ ‡ ´Et3™¨š©(ey&X¥HX•b£Hƒ1IH€`b£I&b†‹#X£‡ƒHR¥T(”‡5eI`b c é Œ Ü Û ô û ú û X  ô ô ø   ¢ ) ù E ¢ …  ù $   ` 2  ¢ $ ø ù $    ù $    Î û    ø ¢ … ¦ ¢ ù ø ù 4ô   ¢ õ ù $  ô C  ù ÷¦ ¢ 4  ù õô pp ø  ù¦ ¤ ¢     ÷ ô 2  $   ù ) ô $   ` dû ú û õô ó )  ø ` 2  ¢ 2  $   ù ) õ  ø … ¢ ù¦ ù $ ¨ÛƒT¢ö¥IT‡7Xù q¥Ë ‡"’öV¥a‡11g§ˆ¡1£d„R¨b£(3G´†bö©·{(ƒV¥310"“‡{3¥U7„¢£  © ¤   ÷  2 ùô ø ¢ E õ ô ù ¤ ô õ ô $ E ø ¢ )' ù )ô   ù $   ` ©¦ ø ¢ ù¦  û  ø ¤ ) ÷  ù 4ô   ¢ õ ù  ¦ ¢ ù ø ø ¢  ø   ¢ ˆV‡v0ö(ø ƒUeb§`ES§”b§baXƒ~d"§(Qb¡¥¨QUöS`Xù W1%€‡¨(I`bÓ¥T(•ù ƒ~5ô eI~) ô $     ù ÷¦ ¢ 4  ù õô ù ù $     ¢ $   2  … ù C ’ ô 2  … …  )  ø ó û    ùô E Ý ù  E ø ¢¦ ¢ E  ù 4ô  ô   … ¢ ô ¨b£†¥fTb7¥Iˆ0‡ö`Ûb£’yH£g1H“hYx5¢1Tù 1QþXtu†T7X‘{|T’QÔ¥T#ƒ¨e5(‰‹A¨vÚ ¥ ù Xb!AŽ „ c ‡ c „ ¡““Y5((I~b£#0X1b¡¢P3V£0"“уVƒHa¨vP§b¨X10(`ù £bX¥1 ø ù $ C Œ Ü Û p Ú ô ø   ¢ ) ù $    ù E ÷ 2  ø … 2  $   ù ) Ð    2 ø ¢ $ Eô “  ô õ ô E  ù ø ¢ô 2 ù E ¢ … ù $ ƒ û   ô  ÷¦ E   E  ÷ ô 4 ¤  ø ù $   ¢ ø    ù  ù ø …   ¦ ÷  ø ù $   ` $ E ¢  ø … … ¢ ù   ù ø E ô 2'ô )  ù $   2  b“QX1¨(17T1¥T‹1ƒ§|Q`b£IˆT(0(bQ£¨V(Tù bӃa¥Xb1¥eed(XX§bB5"Tù eb¢1¥¢  ù Eô ø   ¢ )    ¢¦ ÷ E øô E    ô  … C ùô 4 ù $ TX¨((Ie¥Ô1`7X“¥e§‡‰!07´~b£  X{&T71‘¥VdÇb¨bT`T(¢b~b¨„¥V–ÍI!B@ )  ø  ù¦ … ) ¢  ù õ ô 2 ù E ù ø … ù $   ù ô ) ¢  ù ù C û ù 4ô   E …  ø ù …    ù ø ù ¢ô 2 ¢ )  ø   ¢ 2 ù 4 ø ø ¢   ÷ ¤ `   ô   ÷ E ô ƒ‡5edTù `E!0(XE¢¢‘»{“I™Tƒ¨ô (ƒ€VbH01¨(VX¨b2  ÷ ô 4 ø … ø ÷   ¦¦ ¢ ô õ    ù ¤ ù 4 ¢ $ ù C  ù ) ¢  ù $   ©¦   E ¢  ù ô  @ û    ù 2ô E E ¢  ¢      1ƒ§|0ù b#b‡l¥9§¥7¨rb§ô 1÷ TXù •‡UH•hc„¥b£¨£d¥Vd5¨ÛBA†T¢§X0¥†¥—¥1§ô `I(‡"d(`b£Ó0§ô 35¥H¥‹Xöô b‡Hrb¢§mX(0"¥£QH†¨ö¨b¥£Xgb‹‡"ä~IgX§ƒ1axb“e`~0ö¦ ù ø  ù ø ù $   `   ©¦ ¢  ¢ ø ù ø ÷  ó ù $   ô ø ù   ù ) ¢ ø ¢ … ©  ô¦ô ¤ ¢    ù $   ø    ù Eô  $ E ù $ ƒ û  ) ù ' ¤  ø … ¦ ù 2  ) ù $       2  $   ù )  C  ù $     ô   ¢ Eô¦ … … ¢ © ¤ 2 ù ) ø   ù Eô ø   ¢ )    ¢¦ ÷ E øô E ù $ ¡bÓTb¡~b‘!V310"Sb£s¥v§(IU¨ö1V¥†™0„X‡"#T`§((I~!ˆIÔ1`ö`&¢£  ¥ aHr¥yHd¡5d‰ ¢£eb¨b1H„‡"Qù bbT`3Xb—ö1‘§ô eQ¢§öV‡†T(ˆ0(0(b“‡1÷ F0ö1‘¥Vd}S  ¥ ‰ p Ë Ã x ‰ Ìi ù $   õ ô 2  … ø  ø ÷ 2 ù E  ø … ù¦ … )  ¢ ù ô¦   ÷  2 ù    ù  ù ø …    Ù  ù¦ … ) ¢  ù  C ù $   2  ¢ ` ) ù   ©  ù $   ¢£‚1Iƒ†Te‡¡Xe¢£’¥ yda¥Hd¡5d‰ I•£Tù T(¢T(ø ‡ ´0t3¢¨ô vs Øb£‰¥(IUe¢£~¢§Xù ¤ ‚ w ¤ u ‰ à x ‰ Ìi  ¢      ù ø … ù ‡ û ú û X   w u e ù $   ` ù  ¢ E ù $   õ ô $ E ÷ ÿ û    ¢¦ ÷ E øô E ô û û X   ) ø  ù $   ô   ÷ … ô ù ) ù $ E  õ ô E  ù ø ¢ô 2 ù $    ù $ C 2 ù ô a1´†¥ÔVX7X!§t` ‡ ´ú ty’¥¿ƒ"b™¨“V¢—¨”"TbaXb¨X10(`ù £b&¢£‚0b!†0b§") 'ø ù  ù 2 `  ø  ¢ €X(0b95ô eI~) ‹b£AyH£’¨Ø¨(¡¨IH¥e¨ö¨b¥£X’`ö(¢‡H‘§r¨eV‘1÷ ¥b¨´5`b11vb“ƒ × ù $     ¢ $   ô ô  ©¦ ¢  ¢ ©  ô¦ô ¤ ¢    ø ùô ø ÷  ó ¢ ô  ô   … )   ¢ õ ô ©¦ ø ù 2  ÷ ù $ 2 ' 64 9 Ò FDvƒ•˜0 Ö †Ò y7§FB7͎5Ô~v5˜v³ÑÐ Ó 0¡ G 6 94 Õ A Ò Ó A4 Ò 0¡ ' Î  Ï !  û  ø ù   ù ) ¢ ø ¢ … ù ø ù $    ù Eô  $ X`ed„¥ƒH•Xù {b£l¥•X¨‡1aE ©  ¢ ø  ù ¦ ¤ ¢     ÷  ô `    ù   ©   ô ¦ ô ¤ ¢   Ë ø ù ô ø ÷  ó ù $   © ¤ `   ¢ $     ô   ¢ ) ô   ø … … ¢ ù E  ù ø ¢ ô 2 ' ù   ô  ˆI!‡"‹ö1IaXV1§£TXT(~5ö¨b¥a‡#X7(bƒ†”bP1UyH£P¨eI~§´U¢1¥•XVT(`ù £b–e5bH… ø ù $     ¢ ù 4 ¢ $ ù C  ÷ $ ƒ û y 2  `b£I1¥PƒUHhe1%“P0ÍV¥¢  ø ù –    ¦¦ ¢ ø  `X1¥1A§¥‡" „ ³“ ÌI1£•ƒU7T•§”¨ô a§b!C ¥{ ¯ ’ ’   ¢ $   ø ¢ ù¦ E ô   $ Eô $ ) ø b „¦ Ë b„¦By‹§ô (p œ p  d „¦ ¥ „ Ö ™ ù ¥U“¨b£7@  ¢ E ô $    Ž û ú ˜ ‡ ´nŒ b ˇ á× ¥ Ý ‡ “ Ú ³× Ü ‡ “ Ú ß „¦ §¥ ‡ “ Ú Ö × ‡ “ Ú ‘ hÛ ‘ hÛ ‘ (Û Ü d „¦ ‘ hÛ  ô   ¢ ÷ X ù  ô   E ù 4   Eô §(IH%t¨§(00‡T¨b¤ ¦ ù 2  ) ù $   ø   ô   ¢ )ô   ø … … ¢ ù E  ù ø ¢ô 2 ' ù  ô  … ù $   ø ù 2    E ù¦ … ) ¢  ù ø ù $     ¢  Tb¡‘¢b£x‡" §(I~§´Ub1I¢XVT(`ù £b–e5bH™¢£`b§ô 1T†öV~¥V–sX¢£¥1IvQ d ü „ · ± ª ® µ ´° ´ ª ¹ €WDs|‡3Q““± § ʧ ¡´(¦ .

û      ø'  ù $ X¥¡‰p¢£   ~5Tùey¢Ôb0ù„‘)§•VTX`3XbWb¥¢¥U(X1€†dTù 1¨s£&(b(01÷ eX—¥¡‘1eeyÔb0„XXe¨Í1—§ô ©¦   ô 2 ) ô 2 ù ù E  ø … ù   2 ¢ ù     @ û   E …  ô    ù ø ÷   E ø        ø' 6 ù   ¢ô 2 ù ) ø ù   ô    ù Xb$Ó`¥`E1TvŠƒx¦HΒ¥HH—§b‚1¥ÙD„¨(¥Heb£T`IdEՍ1ƒ§|P1€§}(`b“ƒ ø ù   ù  ù û ¦ ¢  … ô $   2  ¢  Ï ¤ Eô  ¢ ¤ ù $    ù ù C  ù ¤  Ï Î  ÷ ô 4 ¤    ô ù ø ù $ ò €× Ü ‡ “ Ú Þ × Ý ‡ “ epð Þ × ‡ “ Ú Ž ð ¥ „ çÖ × ‡ “ Ú Ž ð Þ „ Œ Úñ Œ ‘ hÛ ‘ yÛ Ý ‘ (Û Ü ‘ aÛ  )ø ù  ù õ  ¢øø ¢ ùø 2  ~XX(‹‡b¥(ƒT(¨1¥¢ › û ú Ž ¥‡ ´nŒ d „¦ Ø b¦ £ ‡ ‘ ð   ù C  ˜ û ¦ ù 4 ù¦ ù )ô      ù ø ¢ô 2  C     ¢ ù ÷¦ ¢ 4 ù õ ¢ ø ù 4 07¦ (€„X¨Tƒ0öQ„¨e!T(`ù £b‘“I•¢ö¥I•‡ƒ£`‡U¢    © ¤ 2 ù E ¢¦ … ø  ù ¤  ¢ ¨ô ‚0X¥§1Tù Ë0Xù €¥1$ ‡ btû 3™§ˆ‡5eII5(X¢eXIHb£7§ô XX(l¥eT`&¢£™a¨b!7§ô û ü X  ô ù 4ô   ¢ 4ô ø ù 2 ù E ¢ …  ù $    ) ø ù   ¦ ¢ ø    ù E ù $   $ Eô $ C  îï ‡ “ —× Ü ˆÚ Þ ìí ‘ aÛ ‘ × IÛ Ý ‡“ ˆÚ ‡ Þ ‘ × UÛ Ü ‡“ ˆÚ 듇 ê b ¥ × Ý ˆÚ ÍØ ¦ ‡ “ èé £ ‘ hÛ d „¦ ‡ Þ ‡“ ˆÚ ‘ × IÛ Ý ‡“ Ö ‘ × (Û ˆÚ Ü þ ü ø ù   … ¢ $  ô 2  $   ù ) ¦ ù I  ¢ ø ó'   ø  ó ÷ Ï ù $   2 ù E ÷ 2  ø   ô ù C ` ù¦ … ) ¢  ù ø  ó û 2 ø ù 2ô    bÛX(V¥H™¨¨¡10"ÓTQ5¥tX‡‡H£•b£7T`1b¡e¨hV¥71‘¥Vd‡HØ0Tù Xb¨(1¥TE   ù ©     ù 4 ¢ $ ù C ©  ô¦ô ¤ô    … ¢ ` õ ô E  ù ø ¢ô 2 ù E ¢ …  2  ¢ ù )ô   2 ù ô ) ù E ÷ 2  ø   ô     ) ù $ d‡I1‡UH‹hg¨7§b¨(X9&V1¢§XVT(`ù £b‹`¥HvV¥Q"§(T´©„€X1b¡(§S§ô T¢£  ¥€b‹f`~V£¨ô ‡7¥l¥g5ö§¢¥a‡‰IU§ô X"1%Qb“eI§1b§¢¥&£¡U‘¥~Qù ƒù Xb“ƒ  ù  Î û  ) $   ø  õ¦ ¢  ©  ô¦ô ¤ ¢    ¦ ¢ E ø ù ) ÷  ù $   ù   ¢¦ ÷ …ô  ¢ )     © ¢ C ©  ¢ ) ø ¢ ø ù $ £ û Ž ‡ Œ {Ø „æ#„¦ ç b ¦ ê d    ù ) ù øô ÷ X ù ø ù $   © ¤ ù –ô  … ù    ù )ô   ô 2 ù  ô )ô¦ ©¦ ù ø ù 4  ø ù $   ¢ ø ø ¢ ù C ù  ¢ E ô $   ø  2  0„0(ö¢|{(b£†&`§(0e‡"§(™¨šTe5„¨ö™¨0(Xƒ0ù QX¢£Iù ƒgh&(IU—§b—‡"™1¥¢ ù¦ ¤ ¢    ©¦¦ ¢  ô  ô 2   E ` ø ù 4 ù C  $ ` ô   @ û Ð 6 ù 4ô   ¢ õ ù  ¦ ¢ ù ø ø  ù¦ ¤ ¢    ô  ô ù E   71¥£X¢5§¥1§(¨b1¥TDT`‡0h(1f0§”B™`—Rvƒ¨eyXb„¥T(”‡"~71¥a‡&¨”5s`1§ô ™æ ‡ ´§t3 û ú û X ù   ¢ ø õ ù   ô    2  $   ù ) ¢     ÷ ” ' ù õ  ÷ “ ø ù 2 ø ' 2 (I£ƒXe¨•£e¡10"QaXVm€dƒb1ÛAXbX‡¡1 ‡ b(V|I(Xb‡T†I¥ƒ¨1‡T“ù s´HT"10b1… ù $   ù  ÷ ` ù  ø ÷  E  ` 2¦ ÷  E b û ¢  ù )   ù $ ù õ  ¢ ø     ô ¤ô $     $ õ ÷   ù ø ¢ 2 ù ø ø ¢ E     ù ø ù C Ž ù )ô   ù $     ¢ ù¦ ¤ ¢¦ô ¢ 4 ¢ $ E ¢  ø … … ¢ ©¦  ƒb¥£(X5b§¢V–ù £b‡VT•ƒv¢0öô (ƒT¥1x(`I“("§(x¢£‹I&öV¥Ô¨ƒI(£a¥b1I¢51 ù $  Œ   ô   ¢¦ ÷ E¦ ¢ E 2  ¢ $ ù $   ` ø ù 4 ù C  $ ` ©  ô¦ô ¤ ¢ ¤  ø … ¦¦ ¢  @ û ©  ô 4 ù¦ ù )    ù E ø ÷   ¢ ¢£"“1§(IÔVX§IU‘1¥H!bVT`‡0h(1¥¥¨7§b¥1bV§¥~’‡5|0ö“"XÓ¥—XXb‡X—ù ¤ (1 C   ¢ E   E ¢ ô $   2  ¢ `  ô   Eô 2 ù ø … ø ù $   ¢ ù C ø     2 ø ¢ $ Eô “ © ¤ 2  ÷  ¢ C 2  $   ù ) ù $ ¥T‘0¥vW¨b£ 1¥„ƒ¨ed§b0(b‘XbIUhW‡"äX1ƒ1a§Û«ˆÅTù 1r¥#ä¡1£d„†b“ƒ û ¤ dû ú û õô ó ù ù  `  ù  ¢ 0IG´£b7©X(†0T(IUE $ E ÷  ¦¦ ¢ ø  ù¦ ¤ ¢     ÷  2  $   ù ) õ  ø … ¢ ù¦ ù $   ô     ø '   ÷ ô ø ÷ …  ù $   2  ¢ ù 4ô   ¢ õ ù  2  a1(Ó§¥—‡"•ö1IaXV1Û§ô 3V£0"‘‡{VIUöb7¨Û¥¡‰‹1ƒ§(¢bb£†1¥‡5eIXb™1¥¢ ¦ ¢ ø ¦¦ ¢ ù ø ¢ ¥Uù §§¥~(ƒŽ „ ` ‡ ' ` „ !¥•¥3XVP¢£~`1§ô ~71¥£X11˨¨¥U¨(X"1%§ˆ§b&yH£y0ö¥T‡d(ø Œ ’       ø' 6 ù $   ù E   ù¦ ¤ ¢     ÷ ©¦¦ ¢ Eô ø ù ) ÷  ô ô $     ¢ $   2 ù¦ ¢ ù 4 ù ø ù   … ¢ $  ô $   ô   ©¦ ¢  ¢ ø ÷ Î `eV¥1¨b£«§r§ô ¡¨¥HIb•Õû aƒ~v„¨e†b£g‡"À0(1÷ 3V£0"c‡{3¥U7™bŨÛC $ E ø ¢ ) ù )ô   ù $   ø  2 ù  2  $   ù ) õ  ø … ¢ ù¦ ù $   $  ô Ž û ú æ ‡ ´nŒ Ž å ® Œ b Ø „¦ Þ Ú „ ‡ ¥ „ Ž c c vÛ £ Œ ä Ö Ú ã d ä ã þ Ð  Ï Î  ) ù    ©  ù $    3áXѓ•’¥¿0eX¡‡¢£“¥¢  ô   ¢      ø   ¢ ) ô 2 ù   ù ø …  ù ¤  ¢ E  ô 2  $   ù ) ù   ø E ô 2'ô ) ù  ¢   û ù¦ ¤ ¢     ÷ ù ¤ 2¦ ÷  ¥§eya¥1W5ô eI~¨¨0(XT(¢V–Bù IU„¨„310"!e0ù XX§bB5"T(#€AƒöV¥a‡11vE~§1‡SC $ E ¢  ø … … ¢ ô $   ¢ $   ©¦ ù   ¢ô 2 ù ) ) C   I C   ù C 2  ¢ © ¢ 2    ø ¢ ù¦ E ù  ô ÷ X ô   … ù E   E ù $   ` ø ù 4 a¥b1I’§¢fI1£Q5T(IÔb0„‘§ô (1ˆ1(1#he1¥&UH¡£|ƒU7TSe¨¢|³§uVT`1TSb£0Xƒ0ù '#(—•0!(1ˆ¥V¥#~¨7§b¥£X1¨qIU§ô X"1%A¥•V0X1¥Te¢£†¥"§e——Ó1XƒHa¨vc@ C  Š û  C   I      ¢ C ©  ô¦ô ¤ ¢    ô ¦ ¢ E ø ù ) ÷     … ù E   E ù $   ` ù )ô   Ð    2 ø ¢ $ Eô “  û úû ú  ô   E ù ÿ  2  $   ù ) ø ùô ø ÷  ó ù $   © ¤ ©  ô¦ô ¤ ¢      ô 2 ù – ©¦ ¢  ¢ 2  ´˜´~¥§edT´¨¨ô 310"“X7(b‡H”bP5ö§¢¥a‡“£¨¨T`¡¨¥HI‚1I¢ ‡ b£t3†§ô û ü û X   2¡10"Q§¢£0eTù Tù beù —QE¡!§bA¥—1§(IH%t"b¨eI‡X(ˆ¨Bƒ"™¡1£d„¢T(¥9Xb…  $   ù ) ô $ 2 ù     ø … b û ù … ©  ô $      ô   ¢ ÷ X ù õ ô   ¢ ø õ ù   ô ø  2  $   ù ) ¢ 2 ù   …  ø    2 ø ¢ $ Eô “ û ó ô C ù â ` Ž   E ¢ ô ` „ › Œ ù¦ ¤ ¢¦ô ¢ 4 ¢ ù ) ¢ E ù ¤  ø ù   ÷ … )  E ù ø  ¤  ø ¢ ù © ©  ¢ X1QHa§ÛÓh•§ÛdgӖ–0Iv™§tVé V(„ ‘71¥§§ƒIUW"¥U0E€`eVb‘0e(ƒ"0ù ‹ƒT‡†¥‚ã ä ´ ­ ® µ ´ · ´ ¯ ² ® ª° ³ Á ¹ · µ ° ³° ¶ ² µ ´ § ¦ ¯ ® µ º ² ¼ ¢„s|D|‹AQ€R “¸!TR€Q1|À(¿Ds‹Q€’± ˜ ü „ .

b „¦ d „¦ ‘ þ ù ø ù $ (`b!C Ž „ —´nŒ Fû ú  ¢  ô )ô¦ ù $   ù I ¢ I†¨"§öyb£YQPa  Ž é Fû ú —´nŒ Ød â ÚØ â Ø þ ÿ£ ¥ Ø b â £ ÚØ â Ö Ú â d â 2  ¢   ô   ¢ …  ù ù 4  ¤ ¢ ù $ V¥c1¨(1¥1V–›‡(‰¥äb  2  … ù b û VHy’¢é ̄~c „¦ ý b ¦ d › û ú û X  ô  ) ø ù   ù $   ù E ¢¦ … ù $  ô £5ÛC ¥‡ ´mt3¸¨~~XX(yb££X¥§1Tۓ ‡ ü b iii Þ ‡ “ Ú vvvi ⠌ 3„¦ û Ø „¦ („ paap p d ‘ aÛ Ž „ d Ø „¦ ‡ („ ¥ ‡ “ Ú w w w w b ⠌ Ø „¦ ‘ aÛ Ž „ ü b û i w Þ ‡ “ Ú vi ⠌ y„¦ ö¥ ‡ “ Ž Ú w ⠌ ‘ aÛ Ž ‘ aÛ d Ø „¦ Ö Þ Þ b Ø ¦ × Ü ˆÚ Þ ¥ ‡“ ‡“ DÚ ¥ × DÚ ˜× Ý ˆÚ ‡“ ‡“ ‘ aÛ ‘ × IÛ Ý ‘ (Û Ü ‘ aÛ ¥ d ‡ “ Ú ñ „¦ ‡  ù E  ù ø ¢ô 2 ù E ¢ …  2  ¢ ù )ô   2 ù ô ) ù $   ø  2  0X10(`ù £b•X¥HX™V¥"§ev0¢©„•b#‡"71¥¢ p p aap ii ‡ “ Ú vvi ⠌ Ø „¦ ˜ d ‘ aÛ Ž „ ò ‡“ Ú i Œ ‡“ Ú ‘ aÛ Ž â çÖ ‘ × IÛ Ý ‘ × (Û Ü „ ù 4 ¢ $ ù C ù 4ô   ¢ 4ô ø ù 2 ù )ô   ù $   ø  ó û¦¦ ¢ ) ƒUH&h‹‡5eII5(X¢‹„¨eb#‡Hun¨¥~ © ø ù 4  ù )  E ù ¤ ù –ô  $  ù ) ù $    ¢ õ ô 4¦   ©¦¦ ¢ ÷   E ¢ ù ø ¢ ù C  ô   ¢ ÷ X ù ¦ ¢ô    ø ù ¢ô 2 ¦ ¢ô   ø ¢ … ù $ ‡Xƒ’T"T0E!X¨(‘0„!bDI•b§|¨¥&¨§IHed¥!(ƒ€hx¥§eyH%tb¥§eTù XE¢b3¥Ô(Xƒ1—¢£    E ÷ ø      E ø 2  ¢  ùô ø  ø ¦ © ¢ ƒ ¢  › û ú X  ô  ) ø ù   ù $   2  ¢ …  ù ù C ` ô $   ù 4  ø …  d1((X10Tù ¨V¥!T7(Xù #‡¨3UsWv§ô I‡ ´£û 37¨5~XX(‹b£v1¥1V–mI3T§¢£‹‡(Xbxsƒ û  Ï ¤ ¦ ¢ ô õô ø  ù $   $  ô Dˆ¥1§‡7(‡•b7¨ÛC dÓdd§d¥ s u x ‰¤ ¥i    ¦ ù $   ô ô $   ø  2ô ¢ … ù 4 ¢ $ ù C ù E ø … ù $ ú x w Byau ¥“¥§b#§§¢£‡"P¨ƒH‡UH•h‹X§ô b•b“ƒ û`§¥§„)V35‰…e¨e I§9XøX(ˆ¨ô”¥¢§‡b¥‡ƒ|9¥e¥§eyH§beTE qdÍr‰{#‰ ¥•b§ô 1}¡1£d„#71¥£X ¦ ¢ô  ©¦   ô ¢¦  … ù     ô õ  ¢ ø õ ¢ â   ô   ¢ ô ¤ )  ¤i u ip ù ø  ¢ õ   ÷ 2  $   ù ) ù¦ ¤ ¢   © ¦ ¢    ô   E ÷  2  ÷ ù ¢ ù C ù ô  ` ù ø   ¤    2   õ      ) ù   ¦  ø ù 4  ¤ ¢ ù $ 5¦§¥1§ô(¨b21¥T11”¥¢11ƒ"„ƒ4(1$„hAXE1¨(y¥b÷ e’ù !£‹¡3Û¡£R‘TXù |¨1÷ Tù ؃(9¥Sb“ƒ ÷§TͤdWsrq¥‰yÍtq¥‰yaÓ¡ §ô ¡1£d„•b—yH£#1IU"—¨b“ƒ ‰p h Ë ¥ pp Ë x wi¤i ˆ x w u x  2  $   ù ) ù $     ¢ $     ¢ ù ) ô $ û ·õöê ð ê†x‡õb¥}ùb£¢¨ô ð ¥T(øÇ¥¢øƒ" „ Pù ƒya¨b!D¥f¨¥‹¥¡‰ô㠇 (ƒ‘(`b“ƒ é ù  ¢ø $  ¦ ¢ ù ©  ê ø ¢ $ Eô $ C  ¦¦ ¢      ø'  ù ø ¢ ù ø ù $ ð Þ „ Ö ‚ „  gBq0E ð ™ b ¦ y   b ¦ „By Ø §( Ø ð ô ¥ ô Eô   ¢ ø 2 ¢ ÷ X ù $     ô   ÷¦   ù $ §m§(I£¢¥H%Hb£l¥v¨eV7b“ƒ ¥ ¥ ¥ Þ „ „Œ é ÖAŽ ð Ž9b„¦•ó’Œ{0E ð ‡  Œ Ø 7Ž ð ø ‡ á w – vs Ý e u Þ w – vs Wß ð u e Þ ¥ „ ç‚7Ž ð Þ „ Œ  Œ Ö  × Ý 7Ž ð    ù E     ¢  2 ¢ ù¦ ô $ ƒ û û ú  ô   E ù ÿ ô 2 ù E ÷ 2  ø    £•XV!I!1¥Tö!§¢“A´ú |Ë¥§edT´€¨€0X1b¡(§ô ô  ©¦ ¢  ¢ ©  ô¦ô ¤ ¢    ø ùô ø ÷  ó ù $      ¢ ù ) © ¤ ô   ÷  ô $   © ø ¢ E    © ¢ C    ù¦ … )ô  ù $ §(¡5¥H¥¢¨ö¨b¥£XX7(bƒ†"b¥•1IU"cˆ¨”V‡•§bXø ƒU7£cUXT71‘§(‘b“ƒ ú iü „ · ± ª ® µ ´° ´ ª ¹ €WDs|‡3Q““± § ʧ ¡´(¦ .

ï é B Cî é é ¥ èA ¥ „ Ö 9 c ï „ B Cî „ „ ¥ „ 5š é „ ¥ „ 5š é à Ö HÚ ã ã ¥ „ é „ é ¥ èA é „ Ö Ù7 $  ô £5ÛC  Ï Î ù $   ø ù 2ô     &b£“X¢§(V€û „ ý 5 ò ø ÷ î ð 6T‘Aç4 2 ì 3Pë d 9 @Þ Ú 7 8Ö Ú û 11‡‰“‡ø XsXƒe¨¢ƒ†¥çÖ XV310"ƒTQ5¥tX‡‡H€•Ï 2  ÷  ¤ ø  ø ù © ø ¢ ø  ô ¤ ø ¢  ¢ û  2  $   ù ) ¦ ù I  ¢ ø ó'   ø  ó ÷ ' 2  1I¢ ‡ —!‡"1¥§e5b107010e‡§(VT†1¥†d¥£b`T¥f¥Xƒ~~V´Ø¢—´”ö1Ifƒ ” “ ø    ô  ô 2   E © E  ù   ô    E 2  ¢ © E ¢ ø ÷ E E ¢  © ø ¢ ) ) ÷ ÿ þ Fû ú ù¦ ¤ ¢ 1  ) b 0¦ % d „¦  w ! £ # ! ù ø  ù ø ù $ (‡"d(X¢“ƒ á w – £ i – ß ' % (&Ø ¨s¨1 ô ©¦  i ! Ö $"0eI~§´UbVQ # !  ù   ¢ )ô   ø … …  w ! £ # ! # Ö $! $  ô £5ÛC i ! `    ù   ô     ©¦¦ ¢  ô  ô 2   Tˆ0e‡§(V~¨¨¥H¨e¨¢1 ù¦ ¤ ¢   ÿ ©¦¦ ¢  ô  ô 2   E  71¥£¡„¨¨¥H¨e¨¢1TV€    ù        ©¦ ) ø  ô  T(X§ô 1¨‘ƒ"¨b— © E  ù      ø  d1T(X§ô 1—‡Hó ù¦ ¤ ¢   ÿ ©¦¦ ¢  ô  ô 2   ö1Ia¡s¨¨¥H¨e5b1  Ø  w – õ ê d ‡„¦ ê d #¦ ¦ ù I  ¢ø ó  ø  ó ÷ Tƒ•¥££´X‡H£•Ï © ô¦ô ¤ ¢   ÿ ø  ¨ö¨b¥£¡v‡Hó ¢   ÷ ”' ù õ  ÷ “ ø ù 2ø ' 2 £Vmd‡bVv!`b‡B31 ‡ û ù¦ ¤ ¢   ù $    ƒ71¥abv§ô  C  $  2  $   ù ) ¢     ÷ ” ' ù õ  ÷ “ ø ù 2 ø ' 2   E ù  ù $     ô  ô 2   E © ô¦ô ¤ ¢    ù $   © ¤ 2 ù   … ) !(1c¡1£d„v£Vm€–‡b1ÛXbX‡¡1T0(xb„I§(¨b1¥T™¨ö¨b¥£Xxb£‚ˆ0(‰‘§ô   ¢ $    ¢ ù ø ù 4   ¢    Ù  $ Eô $ C … ù    ù )ô   ù $       ô ¢ ø      E ¢     2 ¢ ù yH£‹¥(`‡0ù •¥X1÷ ~¨ô a¨b!cTe‡W"§egbˆ¨ƒ£(X1¥Tv†£‹1¥Tö¦   ô $   ` ø ù 4 ù C  w ¨b£lTXƒ0h(€Š þ û    ù      E ô 2  $   ù ) ù $   2  ¢ ` Ž d ¦Œ   XTe‡§ô 1T&§›¡10"Pb£‘1¥A–X—P§ô Ø  ù $   `    ¢      E  ô ¢ ) ù Tb’U¥£X10e1¨ƒ~T(ø i – – I1£yU™¢   ¢ $  © ¢ C $ E ÷  ô a1(v¨é ž1I‘é y„˜BA¢—´”ö1Ifv¨šTX¨(ƒ~‘1(“¨Í§(IH(¨(•b“D`TX¨(†T(X ý b ¦ 2  ¢ ý d ¦ @ û Fû ú ù¦ ¤ ¢ ƒ ô 2 ù –ô ø ¢ ) ) ÷  ô  ô   ¢ ÷  ô  ù $ ƒ û  ù –ô  … ù   ¦§¥~¨¢§bX§b¥I“‡"¢0‡0ù ¥ry(x€#¡y—d(•cyÍv¥1 ¡ xHq‚qqA„Hd€d§dVya7rgyU H¡ ¦ ¢ )  ©¦ õ ô $ ô  ¢ 4 ø   ù 4 ‰ Ãp w ¥ w ¤ ¤  w ¤ ‰ ¥ ‰ Ä x w i¤ Ë ‰ ‰ Ƥ Ƥi Ä ¤ x ‰ ¤ ¥i ¥ x w u sp  ‚ w i x › û ú û X  ¢ E ¢ $ E ÷  @ û  ô   ¢ ÷ X ù  ô  ÷ ¢ô 2 ¢     `  ô   ¢ ÷ X ù ù 4 ¢ ¤yÓe§i ¨¦¥£¡ É ù ¥U‘ca1(™“0¨eIH%{€¨(1E¢¢‘Q¥1†ƒ¨eI1|{‡UC ¢  ¥‡ ´0t3 w x ¥ © §¢ ¤ ¢ w ‘€§ô   ¢ $  © ¢ C ¨›¡1£d„P¢£yPƒ¨Xc¨¨¥Hed¥7hy§(IH%tsb£AU¥£X1¥T1§ƒ~Tù ø i – – I1£yU™¢ ô 2  $   ù ) ù $   © ¤ ù 4¦   ©¦¦ ¢ ÷   E ¢ ù C  ô   ¢ ÷ X ù ù $   `    ¢      E  ô ¢ ) $ E ÷  ô a1(7¨ré ž„”V¥~é m„¦ 5BV|“Dû Íryd0dH¡Æ §ô „ G|ú t3Dû ÍrydX¥Vù §ú ‡ ´Eû 3 ý b ¦ 2  ¢ ý d ô  ÷ $ ƒ u ip w h ‚ ‰ ù s  Fû û X  u ip w h Ë ‚ Ë ô û ú X ´ ­ ® µ ´ · ´ ¯ ² ® ª° ³ Á ¹ · µ ° ³° ¶ ² µ ´ § ¦ ¯ ® µ º ² ¼ ¢„s|D|‹AQ€R “¸!TR€Q1|À(¿Ds‹Q€’± æ ü „ .

` ù¦ ¤ ¢   '@ ø hö1¥£XB͓‡ ` ù¦ ¤ ¢   ' f 7 ` ù¦ ¤ ¢   '   ô @ û 2  $   ù ) ù $   ø   ô   ¢¦ ø ' 6 ù $   ù 4 ø ù Ï ` 2   …  ø ø  ¥71¥a‡XgeIö1IaXB—€¨•€0¡1£d„&b£—‡"†¥§eyÔTù yÏ1x¢£&ƒ¨ô X•ÅB1‰Tù (‡0E  ô  ù  2 ¨ÛT|1Ž › G¢†t3†§9äd1¥†¡H1F¥!T¢ö¥I“IH!x£1VsB&` ¢l¢0e‡ISW—¨Û@ dû ˜ û X  ô Œ c 2  ¢ ` % ` b   ù ÷¦ ¢ 4   ¢ $ C    `   @ ã ã â … ù   '  C  ¢  ô  `¨`§71~¨©‡s¨`§713dq¨uvƒ¡1£d„#¢£|§u`b‡fIHÕR¢§1ƒ"s§¢£Qb¨(1”310")  ô Eô¦ … )ô ø   ô Eô¦ …  ù  ô @ ` 2  $   ù ) ù $   ô ø ù 2 ø    ¢ $ b û ¢¦ ÷ ) ø  ô $   õ ô  ÷ 2  $   ù õ ô $ E ø ¢ )' ù )ô   ¢ ô ¢   ¤  ` `  ù 2 ô ¦ ¢ ø  … ) ù   ù $   © ¤ X  ù 2 ô ¦ ¢ô   ¢ …  ù $   õ ô E ¢¦ … ù ø © b§¢aƒ~–„¨e#§QaVyfadbV§ƒI£‡‰~0e’b‹€Ydb1¨ƒ¥§eI1Ab#b§`¥Ô10(“S’ b ¥ × × Ü “ Ú Œ ¦ Ž ݓ Ú F Ö × Ü“ Ú w Þ “ Ú w Þ Ó× Ý “ Ú w Ž WŒ V Ž W†ü VŒ Ž WŒ V þ ù 4ô   ¢ 4 ø ù 2    ø … ¢ ø   ô   ¢ )ô   ø … … ¢   ô  …' F 2 ø ù    ù E   E ¢ … )  E ù $   ¦¦ ¢ E ù “ û Q‡¨(II¨ô Xb—XX£W!‡"¨eI~§´U¢1¥!§ƒ9ByTù X(ˆ0X!d¥H‘T•bÓ§¥TTۗ´d $ õ ÷ ø $ ¢‡b  „ X¢‡RI„V310"£Vmd‡bVv€b£‘£5Û!ù ƒH‘Db„b‡Xb£  „ XbX‡ ø ù 2 ø    2  $   ù ) ¢     ÷ ” ' ù õ  ÷ “ ù $   $  ô C ø ¢ … )   û ü $ õ ÷  ø $ ø ù 2 ø ¥S13V£0"PX‡"VX¥#‡‘¥HQQb£„‡"#bƒ£TW¨ö¨b¥£X‹b¥¨1v1¥€b¨„XXed•—bü   2  $   ù ) $   ø  $  ¢ ’'  ) ¢ 2  ù $   ø   ø ÷      E © ô¦ô ¤ ¢    ù $     ¦ … 2  ¢ ù ô ) ø ù   ù Ï û û ù¦ ¤ ¢    ô 2  $   ù ) ù $   $ Eô $ C ø  Qö1¥£X“¨¨¡1Ë 0"¢£7a§b!#‡" Û 3¥‡¢¥£ˆb†11VT¨(¡¨¥1¥P5ö¨b¥a‡ ×  ù õ  ¢ ø ù $   2  … ` ô  ©¦ ¢  ¢ ©  ô¦ô ¤ ¢   ø ù ø ÷  ó õ   Xöô b‡H¨b§ô €Uk„¦ ç 8 ¨kTXù W1%‹l’5‡"P¥b‡b£‹IY!(1ˆF` Û ¶(`b!C û b ô ` ø ¤ ) ÷  Ž â ó  ø Œ    ¢ ø ÷   ù $    ¢  C   I × ù ø ù $ û bü × Þ “ Ú ¥ × Ü “ Ú Œ ØÛ × ‡ Ž Ý Û“ Ú Û ‡ Û „ ¥ × Þ × Ü“ ڌ Û × ‡ Ž Ý Û“ Ú Û „ ¥ “ Ú Û Ö × (Û “ Ú Ü þ  ù 4ô 2  $   ù X0‡5‡õ 310") ¢  ø 2  b   ¢ â  C   I ©¦ ù 2ô C ù $   `  ô   ¢ ÷ X ù  ô   E ù 4   E ø ¢ ù ô¦ ù $      2 ùô¦ … … ¢  ù $ b û PXb1Tù ˆTy|&!{V¡¨0b§Û’bUƒ¨eIH%{¨edT‡0ӃTb§7Db£#QT7ö11Ie0bՓ¢F û 2  $   ù 0¡10") ‡ —b£#ƒ"ƒ{‰¥b—yUù TۆUvŒ ” “ ù $  ø  ù 4  ¤ ¢ ù $    ¢ … ù “ Ž E ù¦ E øô E  ô  ÷ ù $ öTXö`S5b1Bb£  û   ¦ … ù ) ¢  ù $    †I§1"¥£•b†¥ C ¢ ø 2 2  ¢ ù  ¢¦ …'   ù¦ … )  E ù $   ô      ø ù $    ù E ¢ ø   ù $     ¦ ¤ Ž ¤ Ub1I€¢¥Ô1B9¿d71‘T—b£x§e¥3X“bÓ¥—XI£e—b£D¥¨AÌHvŒ û  ù E ¢¦ … ˜    ¢ ù¦   ¢         ø ù $     ù ÷¦ ¢ 4 ù   ÷¦   ¤ ¢ ù $   ù   ÷ … )  E 2  ¢ é æû é  `T`¥Ô1g€XIUö!y~#£¥¡ˆbsI—0böI¥‹(Vö¥1¥beVb‘0†V¥xG´  é X{x—¢„¥!¨¥IXX(§Ë¨~xƒPs—û œ ‚£T¢!™¡1£d„¢X‡"3¥#‡‘¥H€ )  ø d éû é  ¦ ¢ 4 ø ù   ô ô 8 ù I ¢ ƒ Ö 6  ù $ C 2  $  ù ) $ ø  $ ¢ ’' ) ¢ 2 ø ù 2 ø ' 2 `bƒ31 ‡ bSI¥¡‰¿¢£I0böI¥’IU§ô X"1%~¢£Igö1¥£7~e3b~¥†tŒ ù $         ø'  ù $     ù ÷¦ ¢ 4 ¦ ¢ E ø ù ) ÷  ù $    ù¦ ¤ ¢   ¢ ù   ÷ … )   Ž ¢ û ‡ `  2  ÷  ¤  ù $   $  ô C    ù       E   ¦ ÷  ø ø ÷  © ø  ` ©  ô¦ô ¤ ¢    ø  8    2  ÷  11V‡9•ù Tb7¨Û!T(X§ô 10“£¨V(Tù Ûb‡‹ù “755ö¨b¥a‡“‡"7™!1V1‡‰¤ 2 ù   E …  ù ù $   ù ø ¢   ¢ $ C `  2  $   ù ) ù ù ø $    ù $   ø    ô   ¢¦ ù ø ' 6 ù $   õ ô  T(0Tù 3dšbP(ƒ‘yH!„01¡1£d„v`(b7ù TbW‡"&V§(IÔ0(ž!V†b£b¨(—SUvŒ Ž E û  ô   ÷¦   © 2 ¢ ù      E ¢  ù ù $   $  ô C  ô   ÷¦   2 ù   ÷ … )  E ù $   ù ø ¢ … )  0¨eV773¥U(X—0¥V–”¢£v£5ÛP§(VöX™0eV¢~0•b£(QH~¥ š û  … ù    ù )ô   é `1T(X"§(¢é „ ‡"‚é „ ÕÅ1¥~1T(X"§(¢¥d ‡ ‡"¢ü š é Փ01T(XP"§(  ø  Ö 8 2  ¢  … ù    ù )ô   é ø  Ö 8 `  … ù    ù )ô é é d ø  ‡ é Ö 8 `  … ù    ù )ô   é é ¥¥&‡" Gš ”yA01T(Xg"§(v¥é „ ‡" ¨š é Y(1qT0(¥T~`(bf§¥‘`1¡ ø  „ Ö 8 ù  ÷ `  ù  ¢ E ù ù ø $   ¦¦ ¢ @ û  2 '3£d„&¥X‡(I‚‘1¥ˆ0XöVel5X¨ö1‘§‡0X71§A¨`§713d5b‹b§ô 1“é A~"§(  $   ù ) I E ¢ ) ø   E ¢ ã 2  ¢ ` ø ù¦ ÷   ô Eô¦ … )ô ` ø ù¦ ÷   ô Eô¦ …  ù ù $   õ   ÷ Ö d ù )ô )  ø QI { RP„ c „ c „ H“tŒ Ú §(¨b1¥TR¥§e¨¢§b  ¥{&(I£ƒXe¨š£eb¡T~ge5(ø PHvŒ G Ö Ž é  ô  ô 2   E ¦ ¢ô  ô ô ù $ )  ø ù   ¢ ø õ ù   ô    ù 2  E ¢ ù  ô b Ž ¤ ` ù )ô   ô ù 4 ¢ $ ù ¤  ô   ÷¦    Ï Î ù $    ù  2 C  Š `  ô   ÷¦   ù   ¢   h„¨ev§”‡UH0Ev¥§e3öXō&b£“T|1¸(—™ƒ¨eV7•(Ia‡ ' © 2 ¢ ù    ù $   ô   ¢ $ b û ù õ ¢ I E ¢ … ¦ ¢ E ø ù ) ÷  ¢ õ   ÷ 7   ù ÷¦ ¢ 4  ù õô ù ù $   2 ô Ž ¢ V¥Te‡b£¨ÛIHՒQ‡ƒy¥HƒIU§ô X"1%Wxb§ô 1F¥“T¢ö¥IT‡7XmbP1§qó tŒ › ü „ ´ ­ ® ³ ¶ ¹ ¯ º § D§ ¢„|#“DÀˆE(¦ .

`¦ ¢ )   ¥~øƒ1çט§§þXˆ¨BD¢»T71¢‘—`I‡1¥b»{#XXE¢b“¨b£v!v§QÔ13f ô   ô Š û æ ) ù¦ ¤  ø …    ø ù C   ¢ ù $   )  ø  ø ù ¢ô 2 ô $   © $ C ô ¢¦ …  ` © ô¦ ¤ ¢    ô      ) © ¢  2 ù C ¦¦ ¢ ø ¤ ) ÷     ¢ ø ÷  •¨7§ôbIaXQE§eI£V…‘¡X¥ø‡"PTh(§§I“Xù W1%!¥£¢‡ 1§´y~‹b§5IH!1¥ù Tb  ) ÷ )ô  ¢ ) ù $   ô   ¢ $ C `  ù $  T(I#¿ûX0böI¢¥0ù‡ö`'9 ùb£ Ù1H‹ƒ¡10"†`ö1§™¨`§713dù b£v‡"Ò¨eI§T(ø  2 ù ¢ ’  ù ÷¦ 4  õô ù $ 2  … ` 2  $   ù ) ø ù¦ ÷   ô Eô¦ …  ù $   ø   ô   ¢¦ ù ' ù $    16¢b£cõb§ô €(û q ¥~0öô eT¨ùb£ 7‡5‡SV¥«§(Ia¥V 5ô eI~À0b1¥H«§ô že5(ø b  ù ø   ù $ ù 4ô õ 2  ¢  ô   ¢      ø   ¢ ) 2 ù 2  ¢ ¤  × ù  ô Û yq ¥ HÚ ‚Ö × (HÚ Û × Ü Û ä ` û ù û ô ` ) ø  ø    Eä ù 4 '  ô ø   ¢ )  ô õ  ô $ E ø ¢ ) ù ) ô   ø ù ¦ ÷    ô E ô ¦ …  ù ¨Q’Š¢†‡"#ƒ£00‡Í¡©(ä (I7¨}bä §baXƒ~‹„¨e“X71§!¨`§713dq¥  ô   ¢ Eô¦ … … ¢ ù $ )  ø   ¦ ÷  ù ø $ Eô $ C  ¦ Î ù $   ù  ô b û  ù ÷¦ ¢ 4  ù õô ù' 6 ù $   2  … û §(IU¨ö1V¥e¢£  ¥{“£51(0(Ëa¨b!Åx§&b£‹(¨(ø AX0bö¥ITƒöX‘1&b†VH1` „ “’ ô 2  ù … …  õ ô  û ) ø  ø    E ù 4 ' ô ø   ¢ ) ô õ ô E  ù ø ¢ô 2 ¦ ¢ô   ¢ …  2 ø ¢ C I E ¢ ¤ ø ù 2 ø  '    ø 5b10EVQQ¢§(—p¨ƒ"l‡00‡Bˆ©(eyW¨“b¨X10(`ù £b3IÔeyHxQ#¥HsXbX‡‡XX£… )  ø   ¦  ø $ Eô $ C Ð  Ï Î {’¨1÷ Tù „a¨b!’—“Ó¥ T(X¡X!be5(ø f‡‡ƒH1V‡9’ˆ¥ö(Ûb’Ix0e5„XXù … ) ù    ©  ù $   ù  ô b û © ø ¢ 2  ÷  ¤   $ õô ø ù $     ¢ 2 ù    ô ) ø ¨Í¨e¨¢1TXƒ111‡‰x€D‡‡ƒH1V‡9#q0ö¢£“Iv¥§e5b10P‡ƒH11ƒ90öVa§(7•g¢ ô  ô  ô 2   E © ø ¢ 2  ÷  ¤  ˜ û © ø ¢ 2  ÷  ¤   ù¦ ù $     ¢  ô  ô 2   E © ø ¢ 2  ÷  ¤   ù¦ $ Eô øô Ï ©¦ … …  û ¨1VQl¢æ 0ö1¥b¨g¥e0Xù „‡Uf‡5e¨ô ‰#&¨Û•§(00‡T9ƒU¢§öA¢£ƒX¢§(V€û › ) ù¦ ¤  ø … ô  ¢ 2 ù …  ù 4 ¢ C ù 4ô     … ¢ $  ô C  ô   E ù 4   E ø ¢ ù ô¦ ù $   ø ù 2ô     û ©  ô¦ô ¤ ¢    ø  ø ¤ ) ÷     ¢ ø ÷   ) ÷ )ô  ¢ ) ù $   2 ô ó ‡5ö§¢¥a‡“‡"“Xù W1%—ˆI£b‡»1¨´y•bv1§©Øû ‡5e5(‰7£¨Ûv¨eI1|{ù ù 4ô  ô   … $  ô C  ô   ¢ ÷ X  ô   E ù 4   E ø ¢ ù ô¦ ù $      2 ùô¦ … … ¢ õ ô $ E ø ¢ ) ù )ô   ø ù¦ ÷   ô Eô¦ …  ù $  ô C 2 ù¦ … ÷  E õ  ¨e00‡0“ƒUb¨ö&b£x£7T7ö11I~¢§baXƒ•"§(—`ö1§—¨`§713d¨£¨Û†T711‡0~b¨ô ' E  ø ù ¢ô 2 2 ø ¢ C I E ¢ ¤ ø ù 2 ø  '    ø …  ©  ô¦ô ¤ ¢    ù $   ù – ©¦ ¢  ¢ `   ©¦ ¢  ¢ ø ù ø ÷  ó õ ô  û –1Tù XE¢byƒ¥HXbX‡‡Xb#¥¢5ö¨b¥a‡~¢£xX¡¨¥1¥lT§ô ¡¨¥HI&`öô b‡HË¢§(—&¢æ û  2  $   ù ) ù  ù $    ù   ø   ô  ô 2   `1¡1£d„(0b£l¥•b“‡"P¥§e5b10E ©  ô¦ô ¤ ¢    ù $   2 ô 5ö§¢¥a‡vb£yV§©ó h§b¡§ô Xù v(ƒe1¨e¨¢1TyXƒ111‡‰Pb V¥xT`1Tù XE¢b„¥Ô(  ` Eô 2  ø … ù ø ¢   ô  ô 2   E © ø ¢ 2  ÷  ¤ ù $   2  ¢  ù E  ø ù ¢ô 2 ¦ ¢ô ' ¢ …  ù $   ø  2 ù  ÷ ô õ ô E  ù ø ù ¢ô {H¢¢£‡"Ò0(1„§»b¨X10(XE¢¢2 0(X(TX~XbX‡¡1 ‡ ¨h“„¤ §b‘b¨eI‡X(ˆ¨ô 2 ùø ù   ù E ø ù 2ø ' 2 ô `  Ï ô $   õ ô   ¢ ø õ ù    ¥  Ö œ (`b!C ø  ù¦ ¤ ¢  ô ÷  ù ¤ 2¦ ÷  C  2  $   ù ) õ ô $ E ø ¢ )' ù )ô    ô Eô¦ …  ù $ Eô $ b ƒ"71¥a5b(&E‚§1‡SQ1¡10"‘¢§baXƒBd"§(€5X§713d~a¨bÕQû „ ù ø ù $ Ë Ø b â Ú Ø â œ Ö Ú â þ  Ï ¤ õ ô C ¦¦  ù $   ø ù 2ô     û ۍD¢§Û(§¨"b£“X¢§(V€´ú d â û ô  ©¦ ¢  ¢ ø ùô ø ÷  ó õ ô  ÷   ¢ ù … ù “ û  ô X¨(¡¨IH¥!X7(bƒ†"b¨(1vITE0v’0§(  ' ¢ ÷ X ù  ô  ¢ô 2 ¦ ù 2  ) ù $      2 ùô¦ … … ¢ 2  $   ù ) $   ø  $  ¢ ’ '  ) ¢ 2  ø ù 2 ø  ' 2 {H%ty¥§(V÷ £bDTb¡~b£†£‘0ö711¥ä¡10"y£‡‡"VX¥#€€~IHQ&XbX‡3V ‡ b  ù $ 2  ¢  ù E  ù ø ù ¢ô 2 2 ù ø ù    ù E ø ù 2 ø ' 2 1I0X10(XE¢¢vT(`eT`AXbX‡¡1 ‡ Ig§(IH§¢WT—b’‡"~0eX“"§(“ö1¥£X V)   ô   ¢ ô ¤ )  E ù $   ø  … ù    ù )ô   ù¦ ¤ ¢   ) ÷ ¥ '5´y~eb‚1HH95ô eI~—‡£IXù |¥Ô(IHXeb’¥—Tb7¥Iˆ0‡ö`”b~b¨(—iû „ ô  ¢ ) ù $   2  … `  ø   ¢ ) ø    ¢ ø …  ¦ ¢ô   ¢ …  ù $     ù ÷¦ ¢ 4  ù õô ù ù $   õ ô  Ö 8 ™‰6 $  ô C 2 ù ô ¢   ¤  ô ` ô  ¢ ¦ ¢ ù ø ù 4ô   ¢ õ ù  ù $      E ù  ø ù   ô ø ÷      E ©  ô¦ô ¤ ¢    ù $   ù ø ù $ ¨Û‘0b§QaV&§qT§´yS¥T(sƒ¨(IX¢Pb£&£00(XXe¨”b‡Tc5ö§¢¥a‡b(`b!C   ô  … ù $   `û ùû ô ` 6 ù 4ô   ¢ õ ù  ¦ ¢ ù ø ø  ’ 8 ’  ù ÷¦ ¢ 4 ù¦ ¤ ¢    ) ÷ )ô  ¢ ) ù $     ¢ $   C  $ ˆ¨‡‰eb1¨Q’ŠERWƒ¨(IX¢lIU(ۇ"ˆ˜‰6 ¥¥•b7¥I71¥£X»V§´y~&¢£—IH£¸(1X é ÙÖ ‡ ç ‰6 8 Þ  7Ž ‡ ç ‰F 8 6 Þ „Œ ¥ Ø   2  $  ù ) $ ø  $ ¢ §ô ¡1£d„†£‡‡"VI#’ ' ) ¢ 2  ø ù 2ø ' 2 ‡‘¥HQ!`b‡B31 ‡ b“‡"¨eI§T(ø !V&b£“yH£7Tƒ¨ô • û ©((I“ƒ£IXù ƒ¥§eIHX ù $   ø   ô   ¢¦ ù ' 6 ù $     ¢ $    ù 4 ô ø   ¢ ) ø    ¢ ø …  ¦ ¢ô   ¢ … ù¢£l¥!Tb7¥Iˆ0‡ö`ˆb£71HVV’“75b10EVQxb§ô €h`1§(¨bVTXƒHV1‡‰‹§b¡§ô Xù … $     ù ÷¦ ¢ 4  ù õô ù ù $   2  … ` üû ’ ô 2  ù … …  õ   û   ô  ô 2   E © ø ¢ 2  ÷  ¤ Eô 2  ø $  ô C  ô   ¢ ÷ X ù  ô  ÷ ¢ô 2 ¦ ù 2  ) ù $   ô ù 4ô   ¢ 4ô ø ù 2 ¦ ¢ô   ¢ …  ù $       ô   ¢ )ô   ø … … ¨Ûä§(IH%t“¨(1E¢¢#Tb¡‘™b䧚‡5eII5(X¢“¥Ô(IH7b£Y¨eI~§´UbV¥¢ ù E  ø ù ¢ô 2 2 ù ø ù    ù E ø ù 2 ø ' 2 `1Tù XE¢b¢0(`eT`—`bƒ31 ‡ ¢£gb§|511¥††Tb¨ƒaV¥ 0e‡3‡ÀW¢£‹e5(ø “¢˜ ù $   õ ô ©¦ … … ¢ © ¤ 2 ù ô ¢   ¤ ) ù    ©   Ï Î ù $   ù  ô b û ´ ­ ® µ ´ · ´ ¯ ² ® ª° ³ Á ¹ · µ ° ³° ¶ ² µ ´ § ¦ ¯ ® µ º ² ¼ ¢„s|D|‹AQ€R “¸!TR€Q1|À(¿Ds‹Q€’± é ¥d „ .

In the following discussion we concentrate on the transient part of the solution. The forcing function may also be time varying in which case it would also have a time scale. The di erence between the dynamic scales in physical space is represented by the di erence in the magnitude of the eigenvalues in eigenspace. 7. However. we assume that this scale would be adequately resolved by the chosen time-marching method.1. and the decision to use some numerical time-marching or iterative method to solve it.1 Sti ness De nition for ODE's 8.1. and. Examples are given showing how time-marching methods can be compared in a given context. but fortunately we now have the background material to construct a de nition which is entirely su cient for our purposes.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. Any de nition of sti ness requires a coupled system with at least two eigenvalues. 8. which is de ned in the next section. An important concept underlying much of this discussion is sti ness. De nitions given in the literature are not unique. since this part of the ODE has no e ect on the numerical stability of 149 . 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 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.

1: Stable and accurate regions for the explicit Euler method. 8. For a given time step. 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. .27 and its solution using a one-root.1. In this gure the time step has been chosen so that time accuracy is given to the eigenvectors associated with the eigenvalues lying in the small circle and stability without time accuracy is given to those associated with the eigenvalues lying outside of the small circle but still inside the large circle. In x many numerical applications the eigenvectors associated with the small j mj are well resolved and those associated with the large j mj are resolved much less accurately. The situation is represented in the complex h plane in Fig. This is given by Eq. the homogeneous part.28. 6. if at all. we exclude the forcing function from further discussion in this section. Consider now the form of the exact solution of a system of ODE's with a complete eigensystem. although these eigenvectors must remain coupled into the system to maintain a high accuracy of the spatial resolution.150 CHAPTER 8. 6. the time integration is an approximation in eigenspace that is di erent for every eigenvector ~ m . CHOICE OF TIME-MARCHING METHODS Stable Region λh I(λ h) 1 R(λ h) Accurate Region Figure 8. 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).2 Driving and Parasitic Eigenvalues For the above reason it is convenient to subdivide the transient solution into two parts. It is important to notice that these de nitions make no distinction between real. we nd that.3 Sti ness Classi cations In particular we de ne the ratio and form the categories The following de nitions are somewhat useful. and imaginary eigenvalues. First we order the eigenvalues by their magnitudes. complex.8. although time accuracy requirements are dictated by the driving eigenvalues. 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. Unfortunately.1. but their presence must not contaminate the accuracy of the complete solution). 8. Rephrased.1. . p+1 ! M ]. numerical stability requirements are dictated by the parasitic ones.1. and the other.2) xm xm m m Solution m=1 {z m=p+1 | } {z } | Driving Parasitic This concept is crucial to our discussion. called the parasitic eigenvalues (no time accuracy whatsoever is required for the eigenvectors associated with these.1) Then we write p M Transient = X c e m t ~ + X c e m t ~ (8. STIFFNESS DEFINITION FOR ODE'S 151 8. thus j 1j j 2j j Mj (8.

As a result it is quite common to cluster mesh points in certain regions of space and spread them out otherwise. and in a boundary layer. 2 sin 2(M + 1) x M ! . this ratio is about 680. 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. but much less so than for di usion-dominated problems. In this case the eigenvalues are all real.2 Relation of Sti ness to Space Mesh Size CHAPTER 8. i. Our brief analysis has been limited to equispaced x . in di usion problems this sti ness is proportional to the reciprocal of the space mesh size squared. we are interested only in the converged steady-state solution. One quickly nds that this grid clustering can strongly a ect the eigensystem of the resulting A matrix. Consider the case when all of the eigenvalues are parasitic. 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. A simple calculation shows that 4 2 1 = . 8. Furthermore.152 8. 2 and the ratio is . near an airfoil leading or trailing edge. In order to demonstrate this.e. the sti er our equations tend to become. For a mesh size M = 40. 4x2 sin2 2 = . Examples of where this clustering might occur are at a shock wave. let us examine the eigensystems of the model problems given in Section 4. The simplest example to discuss relates to the model di usion equation. Under these conditions. 4x2 x 2=.1. Typical CFD problems without chemistry vary between the mildly and strongly sti categories. 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.3. 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. the sti ness is determined by the ratio M = 1 . negative numbers that automatically obey the ordering given in Eq.2.. and are greatly a ected by the resolution of a boundary layer since it is a di usion process.

8. it is. and technology in uencing this is undergoing rapid and dramatic changes as this is being written. Nevertheless. Er and Er! . The length of the time interval. Such a computation we refer to as an event. and erp discussed earlier. and architecture of the available computer. Let us now examine some ground rules that might be appropriate. if certain ground rules are agreed upon. T . capacity. era . Let us consider the problem of measuring the e ciency of a time{marching method for computing. For example. There is no unique answer to such a question. relevant conclusions can be reached. over a xed interval of time.6. in calculating the amount of turbulence in a homogeneous ow. The actual form of the coupled ODE's that are produced by the semi-discrete approach is ~ u At every time step we must evaluate the function F (~ t) at least once.5. rather than the local ones er . and the accuracy required of the solution are dictated by the physics of the particular problem involved. Era . The appropriate error measures to be used in comparing methods for calculating an event are the global ones.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. It should then be clear how the analysis can be extended to other cases. the time interval would be that required to extract a reliable statistical sample. 8. Since there are an endless number of these methods to choose from. one can wonder how this information is to be used to pick a \best" choice for a particular problem. discussed in Section 6. We refer to a single calculation of the ~ u vector F (~ t) as a function evaluation and denote the total number of such evaluations by Fev . 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. and its computation usually consumes the major portion of the computer time required to make the simulation. an accurate transient solution of a coupled set of ODE's.3. This function is usually nonlinear. d~ = F (~ t) u ~ u dt . highly dependent upon the speed. For example. PRACTICAL CONSIDERATIONS FOR COMPARING METHODS 153 problems. 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.

Implications of computer storage capacity and access time are ignored. Eq. 8.25. 6. i.4.3 is obtained from our representative ODE with = . First of all. and must use a constant time step size.3) Er < 0:005 eT .e. i. In some contexts.t. u(T ) = 0:25. must simulate an entire event which takes a total time T . AB2. Three methods are compared | explicit Euler. 8. this can be an important consideration.e. The follow assumptions bound the considerations made in this Section: 1.48. We judge the most e cient method as the one that satis es these conditions and has the fewest number of evaluations.4. The calculation is to be time-accurate.4 Comparing the E ciency of Explicit Methods As mentioned above. see Eq. ln(0:25) with u(0) = 1.. the e ciency of methods can be compared only if one accepts a set of limiting constraints within which the comparisons are carried out.1. 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. 2. h. To the constraints imposed above. and RK4. The time-march method is explicit.. this makes the exact value of u at the end of the event equal to 0. Fev .2 An Example Involving Di usion Let the event be the numerical solution of from t = 0 to T = . the allowable error constraint means that the global error in the amplitude. a = 0.154 8. CHOICE OF TIME-MARCHING METHODS 8.u dt (8. the global error.1 Imposed Constraints T = Nh where N is the total number of time steps. 3. must have the property: du = . so that CHAPTER 8. must be < 0:5%.

Under these conditions a method is judged as best when it has the highest global accuracy for resolving eigenvectors with imaginary eigenvalues.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. 8. In this example we see that.4.1: Comparison of time-marching methods for a simple dissipation problem.4.1. it follows that 155 1 . leading to the so-called incomplete . 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. The results shown in Table 8. ln(0:25)=N . a complete event must be established in order to obtain meaningful statistical samples which are the essence of the solution. The main purpose of this exercise is to show the (usually) great superiority of second-order over rst-order time-marching methods. we add the following: ~u The number of evaluations of F (~ t) is xed. since h = T=N = . for a given global accuracy.1 were computed using a simple iterative procedure. The above constraint has led to the invention of schemes that omit the function evaluation in the corrector step of a predictor-corrector combination. In this numerical experiment the function evaluations contribute overwhelmingly to the CPU time.1. and the number of these evaluations must be kept to an absolute minimum because of the magnitude of the problem. the method with the highest local accuracy is the most e cient on the basis of the expense in evaluating Fev .8. and the fourth-order Runge-Kutta method is the best of all.4. Thus the second-order Adams-Bashforth method is much better than the rstorder Euler method. in addition to the constraints given in Section 8. ( 1 (ln(:25)=N ))N =:25 < 0:005 where 1 is found from the characteristic polynomials given in Table 7. On the other hand. In this case. COMPARING THE EFFICIENCY OF EXPLICIT METHODS Then.

For a 1evaluation method the total number of time steps.4) 1 See Eqs. The . and AB2 schemes are all 1-evaluation methods. one evaluation being used for each step. are the same. Basically this is composed of an AB2 predictor and a trapezoidal corrector. Notice that as k increases. in order to arrive at the same time T after K evaluations. j 1(ik!h1 )jK=k (8. However. the time step must be twice that of a one{evaluation method so h = 2h1. notice also that as k increases. the global amplitude and phase error for kevaluation methods applied to systems with pure imaginary -roots can be written1 Era = 1 .8. in this case. The Gazdag.and 4-evaluation methods.relation for the method is shown as entry 10 in Table 7. 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 derivative of the fundamental family is never found so there is only one evaluation required to complete each cycle.39. given in Section 6. However. K . Let K represent the total number of allowable Fev . h. Let h1 be the time interval advanced in one step of a one-evaluation method. CHOICE OF TIME-MARCHING METHODS predictor-corrector methods. that more e cient methods will result from the omission of the second function evaluation. The presumption is. For a 2-evaluation method N = K=2 since two evaluations are used for each step. leapfrog. The second and fourth order RK methods are 2.38 and 6.156 CHAPTER 8. However. of course. and the number of evaluations. so that for these methods h = h1 . respectively. For a 4-evaluation method the time interval must be h = 4h1 . etc.1. An example is the method of Gazdag. In general. after K evaluations. . This is the key to the true comparison of time-march methods for this type of problem. 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 ~u F (~ t) to advance one step using that method's time interval. 6. the time span required for one application of the method increases. the power to which 1 is raised to arrive at the nal destination decreases see the Figure below. N .

COMPARING THE EFFICIENCY OF EXPLICIT METHODS Er! = !T . 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. and RK4.5. we nd.8. AB2. exact = 1.4. Amplitude.5) 1 (ik!h1 )]imaginary 1 (ik!h1 )]real # Consider a convection-dominated event for which the function evaluation is very time consuming. 7. However. 2 . Notice that to nd global error the Gazdag root must be raised to the power of 50 while the RK4 root is raised only to the power of 50/4.0.2:4 :45 :12 !h1 = :2 50 . see Fig. we are making our comparisons on the basis of global error for a xed number of evaluations. Using this. K leapfrog AB2 Gazdag RK4 !h1 = :1 100 . Gazdag. We consider two maximum evaluation limits K = 50 and K = 100 and choose from four possible methods.9:8 1:5 1:5 b.3e. First of all we see that for a one-evaluation method h1 = T=K .3:8 . Phase error in degrees. although. 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.:96 . in terms of phase error (for which it was designed) it is superior to the other two methods shown. and the fact that ! = 1. Table 8. We idealize to the case where = i! and set ! equal to one. On the basis of global error the Gazdag method is not superior to RK4 in either amplitude or phase. The event must proceed to the time t = T = 10. by some rather simple calculations2 made using Eqs. K tan. 8. leapfrog. K leapfrog AB2 Gazdag RK4 !h1 = :1 100 1:0 1:003 :995 :999 !h1 = :2 50 1:0 1:022 :962 :979 a.2.1 k " 157 (8. the results shown in Table 8. The 1 root for the Gazdag method can be found using a numerical root nding routine to trace the three roots in the -plane.2: Comparison of global amplitude and phase errors for four methods. For example.4 and 8. The rst three of these are one-evaluation methods and the last one is a four-evaluation method.

8. h)nx22 + (PS )2 (8. We rst run 66 time steps with h = 0:001 in order to resolve the 1 term. For a speci c example. h)nx12 + (PS )1 u2(n) = c1 (1 . is h = 0:02.158 CHAPTER 8. Let us choose the simple explicit Euler method for the time march.100t quickly becomes very small and can be neglected in the expressions when time becomes large. 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. The time step limit based on stability. which is determined from 1. 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. Also shown by the small circle centered at the origin is the region of Taylor series accuracy. If some h fall outside the small circle but stay within the stable region.6) We will assume that our accuracy requirements are such that su cient accuracy is obtained as long as j hj 0:1.6. 8. A good example of the concept is produced by studying the Euler method applied to the representative equation. 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 = . these h are sti . If uncoupled and evaluated in wave space. 7. 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. With this time step the . The coupled equations in real space are represented by u1(n) = c1 (1 .5. and a relatively slowly decaying function in the other. but stable.5 and 7.5 Coping With Sti ness 8. Analytical evaluation of the time histories poses no problem since e. the time histories of the two solutions would appear as a rapidly decaying function in one case. 100h)nx11 + c2(1 . of course. Stability boundaries for some explicit methods are shown in Figs. consider the mildly sti system composed of a coupled two-equation set having the two eigenvalues 1 = . In this case the trace forms a circle of unit radius centered at (.1. Numerical evaluation is altogether di erent.1 Explicit Methods The ability of a numerical method to cope with sti ness can be illustrated quite nicely in the complex h plane.1 0) as shown in Fig. If h is chosen so that all h in the ODE eigensystem fall inside this circle the integration will be numerically stable. Numerical solutions. We will also assume that c1 = c2 = 1 and that an amplitude less than 0. 100h)nx21 + c2(1 . We have de ned these h as parasitic eigenvalues.1.001 is negligible.

8.5. COPING WITH STIFFNESS
2

159

term is resolved exceedingly well. After 66 steps, the amplitude of the 1 term (i.e., (1 ; 100h)n) is less than 0.001 and that of the 2 term (i.e., (1 ; h)n) is 0.9361. Hence the 1 term can now be considered negligible. To drive the (1 ; h)n term to zero (i.e., below 0.001), we would like to change the step size to h = 0:1 and continue. We would then have a well resolved answer to the problem throughout the entire relevant time interval. However, this is not possible because of the coupled presence of (1 ; 100h)n, which in just 10 steps at h = 0:1 ampli es those terms by 109, far outweighing the initial decrease obtained with the smaller time step. In fact, with h = 0:02, the maximum step size that can be taken in order to maintain stability, about 339 time steps have to be computed in order to drive e;t to below 0.001. Thus the total simulation requires 405 time steps. Now let us re-examine the problem that produced Eq. 8.6 but this time using an unconditionally stable implicit method for the time march. We choose the trapezoidal method. Its behavior in the h plane is shown in Fig. 7.4b. Since this is also a oneroot method, we simply replace the Euler with the trapezoidal one and analyze the result. It follows that the nal numerical solution to the ODE is now represented in real space by 1 ; 50h nx + c u1(n) = c1 1 + 50h 11 2 ! 1 ; 50h nx + c u2(n) = c1 1 + 50h 21 2
!

8.5.2 Implicit Methods

1 ; 0:5h nx + (PS ) 1 1 + 0:5h 12 ! 1 ; 0:5h nx + (PS ) 2 1 + 0:5h 22
!

(8.7)

In order to resolve the initial transient of the term e;100t , we need to use a step size of about h = 0:001. This is the same step size used in applying the explicit Euler method because here accuracy is the only consideration and a very small step size must be chosen to get the desired resolution. (It is true that for the same accuracy we could in this case use a larger step size because this is a second-order method, but that is not the point of this exercise). After 70 time steps the 1 term has amplitude less than 0.001 and can be neglected. Now with the implicit method we can proceed to calculate the remaining part of the event using our desired step size h = 0:1 without any problem of instability, with 69 steps required to reduce the amplitude of the second term to below 0.001. In both intervals the desired solution is second-order accurate and well resolved. It is true that in the nal 69 steps one -root is 1 ; 50(0:1)]= 1 + 50(0:1)] = 0:666 , and this has no physical meaning whatsoever. However, its in uence on the coupled solution is negligible at the end of the rst 70 steps, and, since (0:666 )n < 1, its in uence in the remaining 70

160

CHAPTER 8. CHOICE OF TIME-MARCHING METHODS

steps is even less. Actually, although this root is one of the principal roots in the system, its behavior for t > 0:07 is identical to that of a stable spurious root. The total simulation requires 139 time steps. It is important to retain a proper perspective on a problem represented by the above example. It is clear that an unconditionally stable method can always be called upon to solve sti problems with a minimum number of time steps. In the example, the conditionally stable Euler method required 405 time steps, as compared to about 139 for the trapezoidal method, about three times as many. However, the Euler method is extremely easy to program and requires very little arithmetic per step. For preliminary investigations it is often the best method to use for mildly-sti di usion dominated problems. For re ned investigations of such problems an explicit method of second order or higher, such as Adams-Bashforth or Runge-Kutta methods, is recommended. These explicit methods can be considered as e ective mildly sti stable methods. However, it should be clear that as the degree of sti ness of the problem increases, the advantage begins to tilt towards implicit methods, as the reduced number of time steps begins to outweigh the increased cost per time step. The reader can repeat the above example with 1 = ;10 000, 2 = ;1, which is in the strongly-sti category. There is yet another technique for coping with certain sti systems in uid dynamic applications. This is known as the multigrid method. It has enjoyed remarkable success in many practical problems however, we need an introduction to the theory of relaxation before it can be presented.

8.5.3 A Perspective

8.6 Steady Problems
In Chapter 6 we wrote the O E solution in terms of the principal and spurious roots as follows:

un = c11 ( 1 )n ~ 1 + + cm1( m )n ~ m + + cM 1 ( M )n ~ M + P:S: 1x 1x 1x n~ + n~ + +c12 ( 1)2 x1 + cm2 ( m )2 xm + cM 2 ( M )n ~ M 2x n~ + n~ + +c13 ( 1)3 x1 + cm3 ( m )3 xm + cM 3 ( M )n ~ M 3x
+etc., if there are more spurious roots

(8.8)

When solving a steady problem, we have no interest whatsoever in the transient portion of the solution. Our sole goal is to eliminate it as quickly as possible. Therefore,

8.7. PROBLEMS

161

the choice of a time-marching method for a steady problem is similar to that for a sti problem, the di erence being that the order of accuracy is irrelevant. Hence the explicit Euler method is a candidate for steady di usion dominated problems, and the fourth-order Runge-Kutta method is a candidate for steady convection dominated problems, because of their stability properties. Among implicit methods, the implicit Euler method is the obvious choice for steady problems. When we seek only the steady solution, all of the eigenvalues can be considered to be parasitic. Referring to Fig. 8.1, none of the eigenvalues are required to fall in the accurate region of the time-marching method. Therefore the time step can be chosen to eliminate the transient as quickly as possible with no regard for time accuracy. For example, when using the implicit Euler method with local time linearization, Eq. 6.96, one would like to take the limit h ! 1, which leads to Newton's method, Eq. 6.98. However, a nite time step may be required until the solution is somewhat close to the steady solution.

8.7 Problems
1. Repeat the time-march comparisons for di usion (Section 8.4.2) and periodic convection (Section 8.4.3) using 2nd- and 3rd-order Runge-Kutta methods. 2. Repeat the time-march comparisons for di usion (Section 8.4.2) and periodic convection (Section 8.4.3) using the 3rd- 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, compare and contrast them with the 3rd- and 4th-order Runge-Kutta methods. 3. Consider the di usion equation (with = 1) discretized using 2nd-order central di erences on a grid with 10 (interior) points. Find and plot the eigenvalues and the corresponding modi ed wavenumbers. If we use the explicit Euler timemarching method what is the maximum allowable time step if all but the rst two eigenvectors are considered parasitic? Assume that su cient accuracy is obtained as long as j hj 0:1. What is the maximum allowable time step if all but the rst eigenvector are considered parasitic?

162

CHAPTER 8. CHOICE OF TIME-MARCHING METHODS

Chapter 9 RELAXATION METHODS
In the past three chapters, we developed a methodology for designing, analyzing, and choosing time-marching methods. These methods can be used to compute the time-accurate solution to linear and nonlinear systems of ODE's in the general form d~ = F (~ t) u ~u (9.1) dt which arise after spatial discretization of a PDE. Alternatively, they can be used to solve for the steady solution of Eq. 9.1, which satis es the following coupled system of nonlinear algebraic equations: ~u F (~ ) = 0 (9.2) In the latter case, the unsteady equations are integrated until the solution converges to a steady solution. The same approach permits a time-marching method to be used to solve a linear system of algebraic equations in the form A~ = ~ x b (9.3) To solve this system using a time-marching method, a time derivative is introduced as follows d~ = A~ ; ~ x x b (9.4) dt and the system is integrated in time until the transient has decayed to a su ciently low level. Following a time-dependent path to steady state is possible only if all of the eigenvalues of the matrix A (or ;A) have real parts lying in the left half-plane. Although the solution ~ = A; ~ exists as long as A is nonsingular, the ODE given by x b Eq. 9.4 has a stable steady solution only if A meets the above condition.
1

163

provided C . In the simplest possible case the conditioning matrix is a diagonal matrix composed of a constant D(b).1.1 Preconditioning the Basic Matrix It is standard practice in applying relaxation procedures to precondition the basic equation.5. and the use of the subscript b will become clear shortly. To use these schemes. there are well-known techniques for accelerating relaxation schemes if the eigenvalues of CAb are all real and of the same sign. 9. ~ b u (9. 9. 9. Such systems of equations arise. ~ b u f f b b f 1 1 1 1 1 (9. the . C~ b = A.7 is (9. 9. we consider iterative methods which are not time accurate at all. If we designate the conditioning matrix by C . RELAXATION METHODS The common feature of all time-marching methods is that they are at least rstorder accurate. which is given by ~ 1 = A. and. for example.1 Formulation of the Model Problem 9.8) which is identical to the solution of Eq.6) b f 1 9.2. C~ b = A. 9. the problem becomes one of solving for ~ in u CAb~ . we seek to obtain rapidly a solution which is arbitrarily close to the exact solution of Eq. For example. In this chapter.7) ~ = CAb ]. our analysis will focus on their application to large sparse linear systems of equations in the form Ab~ . 9. Using an iterative method. C .164 CHAPTER 9. In the following we will see that our approach to the iterative solution of Eq. at each time step of an implicit time-marching method or at each iteration of Newton's method. ~ b = 0 u f (9. equally important. C~ b = 0 u f Notice that the solution of Eq. exists.5 from the left by some nonsingular matrix.5) where Ab is nonsingular. Such methods are known as relaxation methods. This preconditioning has the e ect of multiplying Eq. does not depend at all on the eigensystem of the basic matrix Ab .7 depends crucially on the eigenvalue and eigenvector structure of the matrix CAb. While they are applicable to coupled systems of nonlinear algebraic equations in the form of Eq.5.

6). This is accomplished using a diagonal preconditioning matrix 2 61 60 6 D = 2 x6 0 6 60 4 0 1 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 2 3 7 7 7 7 7 7 5 (9.1 32 0 7 6 .11) 1 .10) which scales each row. FORMULATION OF THE MODEL PROBLEM 165 conditioning matrix C must be chosen such that this requirement is satis ed.1 .2 0 6 6 1 0 .1 . It can rst be conditioned so that the modulus of each element is 1.1.1 4 1 0 . For example. this leads to the approximation 82 > 0 1 >6 > 0 6 1 <6 . 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.2 1 1 6 0 0 6 = 6 1 0 .1 0 >4 : .9.AT A = 2 1 1 2 0 6 .1 0 6 6 .1 .1 1 1 76 0 76 76 .0 a 7 6 7~ 6 7u+6 0 7 6 17 6 0 5 4 2 0 39 > 7> 7> 7= 7 7> 7> 5> (9.1 1 1 ~= 0 u 2 x >6 x 6 >6 . The result is A = .1 1 1 6 0 6 6 . A choice of C which ensures this condition is the negative transpose of Ab.2 3 2 u 7 6 .9 has eigenvalues whose imaginary parts are much larger than their real parts.2 .2 4 1 (9. Using a rst-order backward di erence on the right side (as in Section 3.9) The matrix in Eq. 9.1 3 7 7 7 7 7 17 5 1 2 0 6 . We then further condition with multiplication by the negative transpose.1 0 76 176 .1 54 3 7 7 7 17 7 17 5 1 0 .

and the classical relaxation methods can be applied.2 1 6 1 6 1 . as de ned in Appendix A.1 ) just rearranges the rows and columns of a matrix.AT as a preconditioner we simply wish to show that a broad b range of matrices can be preconditioned into a form suitable for our analysis.2 1 =6 6 6 1 .e.2 .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 . Relaxation methods are applicable to more general matrices. RELAXATION METHODS 1 1 If we de ne a permutation matrix P and carry out the process P T . symmetric with negative real eigenvalues).2 0 1 76 7 6 1 0 .. as given in Appendix B.2 4 3 7 7 7 7 7 17 5 (9. In the remainder of this chapter. the conditioned matrix .166 1 1 CHAPTER 9. 1 .AT Ab b is nevertheless symmetric negative de nite (i.1 0 1 76 76 0 7 6 0 . 2 We use a symmetric negative de nite matrix to simplify certain aspects of our analysis.1.1 which has all negative real eigenvalues.2 0 1 7 6 76 176 76 76 054 1 0 . Thus even when the basic matrix Ab has nearly imaginary eigenvalues. ~ = 0 f (9. 2 1 6 .12) 9. we will thoroughly investigate some simple equations which model these structures. We do not necessarily recommend the use of . 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 . The classical methods will usually converge if Ab is diagonally dominant.2 1 7 6 54 0 1 1 .AT A ]P (which just reorders the elements of A and doesn't change the eigenvalues) we nd 2 60 60 6 60 6 60 4 1 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 7 6 .13) where A is symmetric negative de nite.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. We will consider the preconditioned system of equations having the form A~ .

5 and 9. g(x) @t @x This has the steady-state solution 2 2 2 2 (9. ~ .2 1) x ~ fb = ~ . If we consider a Dirichlet boundary condition on the left side and f either a Dirichlet or a Neumann condition on the right side. 9.2 1)~ + (bc) .16) @x which is the one-dimensional form of the Poisson equation. In the notation of f Eqs. ~ = A. (bc) g ~ (9.18) Choosing C = x I . Note that the solution of Eq. is guaranteed to exist because A is nonsingular. A = CAb and ~ = C~ b f f (9.13. 9.19) ~ where ~ = x fb. we obtain B (1 .17) dt x ~ where (bc) contains the boundary conditions and ~ contains the values of the source g term at the grid nodes. then A has the form A=B 1~ 1 b 2 2 2 . 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 .14) The above was written to treat the general case.1.7.2 1)~ = ~ f (9. FORMULATION OF THE MODEL PROBLEM 1 167 accurate when we later deal with ODE formulations. In this case 2 Ab = 1 B (1 .15) @ u = g(x) (9. we nd d~ = 1 B (1 . Introducing the threepoint central di erencing scheme for the second derivative with Dirichlet boundary conditions.9. ~ u u ~ g (9.

RELAXATION METHODS ~ = . 9. ~ f f (9. The converged solution is designated ~ 1 so that ~ 1 = A. or at least easier to solve than the original system.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 . ~ f 1 (9. ~ n = A~ n . ~ = G~ n .21 for ~ n gives ~ n = I + H .22) 9.2 b . and the Error +1 +1 1 1 1 Solving Eq. A (9.2 Classical Relaxation 9. A.2. The error at the nth iteration is de ned as ~ n ~ n .24) Hence it is clear that H should lead to a system of equations which is easy to solve.2 . that it is directly applicable to two.1 (9. the Residual.25) 1 1 . 9. H . ~ e f (9. H . The matrix H is independent of n for stationary methods and is a function of n for nonstationary ones. and much of the remaining material is devoted to this case.2 or .24 using a diagonal conditioning matrix. however. A tremendous amount of insight to the basic features of relaxation is gained by an appropriate study of the one-dimensional case. We attempt to do this in such a way.2. A]~ n .168 CHAPTER 9.2 s]T s = .and three-dimensional problems.23) where G I + H.1 is easily obtained from the matrix resulting from the Neumann boundary condition given in Eq. The iteration count is designated by the subscript n or the superscript (n). 3.21) where H is some nonsingular matrix which depends upon the iterative method. ~ 1 = ~ n . ~ f (9.20) Note that s = .2 The Converged Solution.

25 by A from the left. CLASSICAL RELAXATION 169 where ~ 1 was de ned in Eq. This operator comes about by choosing the value of jn such that together with the old values of j.3 The Classical Methods Point Operator Schemes in One Dimension Let us consider three classical relaxation procedures for our model equation B (1 . and use the de nition in Eq.29 is satis ed using the new values of j and j .22. f i (9. In all of the above. . and j .1.29) as given in Section 9.31) j j j 2 j. and its eigenvalues.9. This operator is a simple extension of the point-Jacobi method which uses the most recent update of j. The Point-Jacobi method is expressed in point operator form for the one-dimensional case as n = 1 h n + n . 9. 9. we have considered only what are usually referred to as stationary processes in which H is constant throughout the iterations.2. 9.28) Consequently. the j th row of Eq.26. The residual at the nth iteration is de ned as ~ n A~ n . Hence the j th row of Eq. Nonstationary processes in which H (and possibly C ) is varied at each iteration are discussed in Section 9.29 is satis ed.26) Multiply Eq. There results the relation between the error and the residual A~ n .2. determine the convergence rate of a method. 9.2.f i (9. it is not di cult to show that ~ n = G~ n e e +1 (9. ~ n = 0 e r (9. The method of successive overrelaxation (SOR) ( +1) ( ) 1 ( ) +1 ( +1) 1 +1 ( +1) ( +1) 1 ( ) +1 1 1 +1 .30) j j j 2 j. and the old value of j . G is referred to as the basic iteration matrix.5. 9. ~ r f (9. 9. The Gauss-Seidel method is n = 1h n + n . which we designate as m .27) Finally.2 1)~ = ~ f (9.

~ f (9.3. A~ . then perhaps it would be better to move further in this direction. 9.170 CHAPTER 9. D.35) dt This is equivalent to d~ = H . The Gauss-Seidel method is obtained with H = .32) j ( +1) 1 ( ) +1 ( +1) ( ) ( ) where ! generally lies between 1 and 2.33) j j 2 j.1 The Ordinary Di erential Equation Formulation . 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. fj 2 . about which we have already learned a great deal. It is usually expressed in two steps as h i ~j = 1 jn + jn . being lower triangular. h i n = jn + ! ~j .21 leads to an interpretation of relaxation methods in terms of solution techniques for coupled rst-order ODE's. RELAXATION METHODS is based on the idea that if the correction produced by the Gauss-Seidel method tends to move the solution toward ~ 1. and the portion above the diagonal.D. which is also easy to solve. the portion of the matrix below the diagonal.34) Then the point-Jacobi method is obtained with H = . which certainly meets the criterion that it is easy to solve. jn (9.21 results from the application of the explicit Euler time-marching method (with h = 1) to the following system of ODE's: ~ H d = A~ . ! f (9. but it can also be written in the single line n = ! n + (1 . L. 9. ~ = H .36) b dt 1 1 9. 9. One can easily see that Eq. U .3 The ODE Approach to Classical Relaxation The particular type of delta form given by Eq. such that A=L+D+U (9. 9.(L + D). ! ) n + ! n . C A ~ .13 into its diagonal. ~ ] fb f (9.

and the eigenvectors of H written as ~ = c e 1 t~ + + cM e M t~ M +~ 1 x} (9. A has been chosen (that is. a constant for the entire iteration. These are xed by the initial guess.36. For this method m = 1 + m h. In general it is assumed that any or all of the eigenvectors could have been given an equally \bad" excitation by the initial guess. that is.3.39) The initial amplitudes of the eigenvectors are given by the magnitudes of the cm .36 are independent of t. ~ f 1 (9. 6. even though no true time accuracy is involved. The eigenvalues are xed by the basic matrix in Eq. The eigenvalues are xed for a given h by the choice of time-marching method.38) which is the solution of Eq. the solution can be of t.7. the only free choice remaining to accelerate the removal of the error terms is the choice of h. so that we must devise a way to remove them all from the general solution on an equal basis. THE ODE APPROACH TO CLASSICAL RELAXATION 1 1 1 171 In the special case where H .13. A is nonsingular). Assuming that H . given by 1 1 1 1 ~ 1 = A. Hence the numerical solution after n steps of a stationary relaxation method can be expressed as (see Eq. 1 . if all of the eigenvalues of H . A depends on neither ~ nor t. H and C in Eq. 9. The generalization of this in our approach is to make h. 9.37) | x {z error where what is referred to in time-accurate analysis as the transient solution. Throughout the remaining discussion we will refer to the independent variable t as \time". the \time" step.9. the preconditioning matrix in 9.35. 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. is now referred to in relaxation analysis as the error. It is clear that. H . and the secondary conditioning matrix in 9. In a stationary method. an iteration process has been decided upon). 9. 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. then the system of ODE's has a steady-state solution which is approached as t ! 1. As we shall see. Suppose the explicit Euler method is used for the time integration. they are not changed throughout the iteration process. ~ is also independent u f . A are linearly independent. A have negative real parts (which implies that H .28) ~ n = c ~ (1 + h)n + |x 1 1 1 + cm~ m (1 + mh)n + x {z error + cM ~ M (1 + x n M h) } +~ 1 (9.

Now let us study these methods as subsets of ODE as formulated in Section 9.29 into the ODE form 9.172 CHAPTER 9. RELAXATION METHODS 9. ~ f (9. 9.43) where and ! are arbitrary. 1 +1 TABLE 9. ~ n) = B (1 .1: VALUES OF and ! IN EQ.2.2 ODE Form of the Classical Methods The three iterative procedures de ned by Eqs.40) dt As a start.B . h.30.3 can be identi ed using the entries in Table 9. We arrive at H (~ n .1.35. ! 0)(~ n . 9.B (1 . 9. With this choice of notation the three methods presented in Section 9.42) +1 +1 It is clear that the best choice of H from the point of view of matrix algebra is .1. However. since multiplication by the inverse amounts to solving the problem by a direct method without iteration. this is not in the spirit of our study. Then ~ H d = B (1 . The constraint on H that is in keeping with the formulation of the three methods described in Section 9. ~ f (9.2 1) since then multiplication from the left by .3 is that all the elements above the diagonal (or below the diagonal if the sweeps are from right to left) are zero. let us use for the numerical integration the explicit Euler method 0 (9. Insert the model equation 9.3.2 1)~ . we are led to 2 B (.2 1)~ n .2 1) gives the correct answer in one step.2.31 and 9. If we impose this constraint and further restrict ourselves to banded tridiagonals with a single constant in each band. (1 . equal to 1. ~ n) = B (1 .41) n = n+h n with a step size.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 .3.2 1)~ n .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. ~ f (9.

1) n + !h n . In this light. are identi ed for the one-dimensional case by Eq. The point operator that results from the use of the explicit Euler scheme is ! ! n = ! n + ! (h .47) An iterative scheme is developed to nd the converged. ) n . our purpose is to examine the whole procedure as a special subset of the theory of ordinary di erential equations. ~ is the steady-state solution.44 and Table 9. or steady-state. j 2 j.49) where ~ n is the transient. Gauss-Seidel. The e f three classical methods. (. the three methods are all contained in the following set of ODE's d~ = B . 1 .2 1)~ .46) This equation is preconditioned in some manner which has the e ect of multiplication by a conditioning matrix C giving A~ .48) dt This solution has the analytical form ~ n = ~n + ~ 1 e (9.44) dt ! and appear from it in the special case when the explicit Euler method is used for its numerical integration. ~ = 0 f (9. or error. !h f (9.4 Eigensystems of the Classical Methods The ODE approach to relaxation can be summarized as follows. 9. 9. solution of the set of ODE's ~ H d = A~ . and SOR. (!h .45) j j. The basic equation to be solved came from some time-accurate derivation Ab~ . However. 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. Point-Jacobi. 2 0) B (1 .9. ~ f (9.4.1. ~ f (9. and ~ 1 A. ~ b = 0 u f (9. 2 2 j 2 j 1 ( +1) ( +1) 1 ( ) 1 ( ) ( ) +1 This represents a generalization of the classical relaxation techniques.

This relation can be plotted for any h. Fig.2.174 CHAPTER 9.54) . Thus Convergence rate j mjmax m=1 2 1 M (9. the optimum stationary case. . 9. d = . the maximum j m j is obtained with m = M .30. It is also instructive to inspect the eigenvectors and eigenvalues in the H .e. we use the ODE analysis to nd the convergence rates of the three classical methods represented by Eqs. B (1 . RELAXATION METHODS 1 Given our assumption that the component of the error associated with each eigenvector is equally likely to be excited. e = 2=!. For h < 1. 9. 2 The eigensystem can be determined from Appendix B. A matrix for the three methods. 9.1 2 ).2 and for h > 1.44.53) m = . i. This amounts to solving the generalized eigenvalue problem A~ m = m H~ m x x (9. The plot for h = 1. 9.1 since both d and f are zero. we take M = 5 for the matrix order.50) In this section. 9. The eigenvalues are given by the equation m (9. A) having maximum absolute value. The three special cases considered below are obtained with a = 1.52) ! xm The generalized eigensystem for simple tridigonals is given in Appendix B. the ODE matrix H .2 1)~ m = x m B (.51) for the special case 2 0)~ (9.3.2.31.1.relation for the explicit Euler method is m = 1 + m h. Note that for h > 1:0 (depending on M ) there is the possibility of instability. A reduces to simply B ( 1 .4. 9. and 9. 9. b = .1 The Point-Jacobi System 1 If = 0 and ! = 1 in Eq.32. j j = j M j and the best possible convergence rate is achieved: 1 1 j m jmax = cos M + 1 (9. To illustrate the behavior. j m j > 1:0. c = 1. is shown in Fig. This special case makes the general result quite clear.1 + cos M +1 m = 1 2 ::: M The . To obtain the optimal scheme we wish to minimize the maximum j m j which occurs when h = 1. and f = 0. Fig. the maximum j mj is obtained with m = 1. the asymptotic convergence rate is determined by the eigenvalue m of G ( I + H .

(1 . EIGENSYSTEMS OF THE CLASSICAL METHODS 1 175 For M = 40.1 .1 = .1:5 2 p (9. we obtain j mjmax = 0:9971. the eigenvectors can be written as 2 1=2 3 2 3 2 p3=2 3 p 7 p 7 6 3=2 7 6 1 7 6 3=2 7 6 6 0 7 6 ~ = 6 p 1 7 ~ = 6 p 7 ~ = 6 . 9.39. 23 = . ~ 1 = c 1 . 3=2 7 ~ = 6 p0 7 x = 6 p 7 6 1 7 x 6 6 7 6 3=2 7 7 6 .56) The corresponding eigenvalues are. Thus after 500 iterations the error content associated with each eigenvector is reduced to no more than 0.9. the eigenvectors of H . 3=2 1 2 3 4 5 (9.1.1 7 6 7 x 6 0 7 x 6 7 x 6 6 7 6 6 3=2 7 6 0 7 7 6 . (1 . 2 )h]n~ 1 2 2 p 1 . 9. 3 )h]n~ x 2 1 x + c 1 .1:866 From Eq. 3=2 7 5 4 p 5 4 1=2 . For M = 5.57) 4 = . 3=2 3 2 p 3 2 3 1= p =2 7 p2 7 6 .1 .55) M +1 This is a very \well-behaved" eigensystem with linearly independent eigenvectors and distinct eigenvalues.0:5 2 = . the numerical solution written in full is 5 p ~ n .23 times its initial level.1:0 = .53 1 2 3 = . 1 = .1 + 23 = .4. A are given by ~ j = (xj )m = sin j m x j = 1 2 ::: M (9.0:134 = . 3=2 7 6 6 6 7 ~ 6 .1 + 1 = . from Eq. 3=2 7 4 5 4 5 5 4 p 7 1 1=2 . Again from Appendix B. The rst 5 eigenvectors are simple sine waves.

176
1.0
m=5

CHAPTER 9. RELAXATION METHODS
h = 1.0
m=1

Values are equal

σm
m=4 m=2

Μ=5
0.0 -2.0
m=3

-1.0

0.0

Figure 9.1: The

relation for Point-Jacobi, h = 1 M = 5. + c 1 ; (1
3 4

+ c 1 ; (1 + 1 )h]n~ 2 x p + c 1 ; (1 + 23 )h]n~ x
4 5

)h]n~ x

3

5

(9.58)

9.4.2 The Gauss-Seidel System
If and ! are equal to 1 in Eq. 9.44, the matrix eigensystem evolves from the relation

B (1 ;2 1)~ m = x

m B (;1

2 0)~ m x

(9.59)
1

which can be studied using the results in Appendix B.2. One can show that the H ; A matrix for the Gauss-Seidel method, AGS , is

AGS

B ; (;1 2 0)B (1 ;2 1) =
1

9.4. EIGENSYSTEMS OF THE CLASSICAL METHODS
Approaches 1.0

177

1.0

h < 1.0
h

0.0 -2.0

De

cr

ea
h

si
ng

σm

ng
si

Figure 9.2: The

relation for Point-Jacobi, h = 0:9 M = 5.
Exceeds 1.0 at h

De cr ea

-1.0

0.0

λ mh

∼ ∼

1.072 / Ustable h > 1.072

1.0

h > 1.0
ng si ea cr In

In

cr

ea

si

σm

ng

h

h

M = 5
0.0 -2.0

-1.0

0.0

λmh

Figure 9.3: The

relation for Point-Jacobi, h = 1:1 M = 5.

principal vectors. The equation for the nondefective eigenvalues in the ODE matrix is (for odd M ) + m ) m = 1 2 ::: M2 1 (9.61) m = ;1 + cos ( M +1 and the corresponding eigenvectors are given by j; m + ~ m = cos m x sin j M + 1 m = 1 2 ::: M2 1 (9.62) M +1 The - relation for h = 1, the optimum stationary case, is shown in Fig. 9.4. The m with the largest amplitude is obtained with m = 1. Hence the convergence rate is
2 1

CHAPTER 9. RELAXATION METHODS 2 ;1 1=2 3 1 6 0 ;3=4 1=2 7 2 6 7 6 0 1=8 ;3=4 1=2 7 3 6 7 6 7 6 0 1=16 1=8 ;3=4 1=2 7 4 6 7 (9.60) 6 0 1=32 1=16 1=8 ;3=4 1=2 7 5 6 7 6 . ... 6 .. 7 . . . . . . . . . . . . . . . 7 ... 4 5 0 1=2M M The eigenvector structure of the Gauss-Seidel ODE matrix is quite interesting. If M is odd there are (M + 1)=2 distinct eigenvalues with corresponding linearly independent eigenvectors, and there are (M ; 1)=2 defective eigenvalues with corresponding
178

Since this is the square of that obtained for the Point-Jacobi method, the error associated with the \worst" eigenvector is removed in half as many iterations. For M = 40, j m jmax = 0:9942. 250 iterations are required to reduce the error component of the worst eigenvector by a factor of roughly 0.23. The eigenvectors are quite unlike the Point-Jacobi set. They are no longer symmetrical, producing waves that are higher in amplitude on one side (the updated side) than they are on the other. Furthermore, they do not represent a common family for di erent values of M . The Jordan canonical form for M = 5 is

j m jmax = cos M + 1

2

(9.63)

3 2 h~ i 7 6 h~ i 7 6 6 7 2~ 37 ; A X =J = 6 7 6 X GS GS 1 6 7 6 6 ~ 77 6 6 1 77 4 4 55 ~
1 1 2 3 3 3

(9.64)

9.4. EIGENSYSTEMS OF THE CLASSICAL METHODS
1.0

179

h = 1.0

σm Μ=5

2 defective λ m 0.0 -2.0 -1.0

0.0

λmh

Figure 9.4: The

relation for Gauss-Seidel, h = 1:0 M = 5.

The eigenvectors and principal vectors are all real. For M = 5 they can be written

2 3 2 p 3 2 3 2 3 2 3 3=2 1=2 7 17 0 7 6 3=4 7 6 p3=4 7 607 6 2 7 6 0 7 6 6 7 6 6 6 0 7 ~ = 6 3=4 7 ~ = 6 p0 7 ~ = 6 0 7 ~ = 6 ;1 7 ~ = 6 4 7 (9.65) 6 7 x 6 7 x 6 7 x 6 7 x 6 7 x 6 6 6 7 6 7 6 7 6 9=16 7 7 6 ; 3=16 7 7 607 6 0 7 6 ;4 7 4 5 4 p 5 4 5 4 5 4 5 9=32 0 0 1 ; 3=32
1 2 3 4 5

The corresponding eigenvalues are = ;1=4 = ;3=4 ) = ;1 (4) Defective, linked to (5) Jordan block The numerical solution written in full is thus ~ n ; ~ 1 = c (1 ; h )n~ 4 x h x + c (1 ; 34 )n~
1 2 3 3 1 1 2 2

(9.66)

180

CHAPTER 9. RELAXATION METHODS " # n + c h n (1 ; h)n; + c h n(n ; 1) (1 ; h)n; ~ + c (1 ; h) x 1! 2! + c (1 ; h)n + c h n (1 ; h)n; ~ x
3 4 1 5 2 2 3 4 5

+ c (1 ; h)n~ x
5

1!

1

4

5

(9.67)

9.4.3 The SOR System
1

If = 1 and 2=! = x in Eq. 9.44, the ODE matrix is B ; (;1 x 0)B (1 ;2 1). One can show that this can be written in the form given below for M = 5. The generalization to any M is fairly clear. The H ; A matrix for the SOR method, ASOR B ; (;1 x 0)B (1 ;2 1), is
1 1

2 x 0 0 6 ;2;2x x ; x 0 6 1 6 ;2x + x x x 2x2x x 6 x + ; + x ; 2x x x 6 ;2x + x x ; 2x + x x ; 2x + x 6 x ; 2x 4 ;2 + x 1 ; 2x + x x ; 2x + x x ; 2x + x x
4 4 3 4 3 4 4 5 2 3 2 3 4 3 4 4 2 2 3 2 3 4 3 4 2 2 3 2 3 4

3

; 2x

0 0 0 x

4

4

3 7 7 7 7 (9.68) 7 7 5

Eigenvalues of the system are given by
m

+ = ;1 + !pm 2 zm

2

m = 1 2 :::M
2 2 1 2

(9.69)

where

zm = 4(1 ; !) + ! pm] = pm = cos m =(M + 1)]
2 2 2 1

If ! = 1, the system is Gauss-Seidel. If 4(1 ; !) + ! pm < 0 zm and m are complex. If ! is chosen such that 4(1 ; !) + ! p = 0 ! is optimum for the stationary case, and the following conditions hold: 1. Two eigenvalues are real, equal and defective. 2. If M is even, the remaining eigenvalues are complex and occur in conjugate pairs. 3. If M is odd, one of the remaining eigenvalues is real and the others are complex occurring in conjugate pairs.

2p2i)=9 = . For odd M .(10 + 2 2i)=9 = . h + h m .5. This illustrates the fact that for optimum stationary SOR all the j m j are identical and equal to !opt .6 7 1 7 p p )=2 7 6 1=2 7 6 6 3(1 i 2)=6 7 6 6 6 6 7~ 6 9 7~ 6 7 6 7~ 6 0 7 ~ = 6 1=3 7 x = 6 16 7 x = 6 p 7 x = 6 1=3 7 (9. The remaining linearly independent eigenvectors are all complex.4. EIGENSYSTEMS OF THE CLASSICAL METHODS One can easily show that the optimum ! for the stationary case is and for ! = !opt 181 (9. p m= m 2 m 2 1 2 q . much faster than both GaussSeidel and Point-Jacobi. 1.70) !opt = 2= 1 + sin M + 1 ~m x m = = m.(10 .4=3 (9. Hence the worst error component is reduced to less than 0.2=3 (2) Defectivep linked to = . For M = 5 they can be written 2 3 2 3 2 p3(1 3 2 3 1=2 7 .74) 2 1 2 34 5 1 1 3 4 5 !opt p + i p . and the bene t may not be as great.73) 6 7 0p x 6 7 6 7 6 6 6 1=6 7 6 13 7 6 3(5 i 2)=54 7 7 6 0 7 7 4 5 4 5 4p 5 4 5 p 1=18 6 1=9 3(7 4i 2)=162 The corresponding eigenvalues are = .71) Using the explicit Euler method to integrate the ODE's.9. j mjmax = 0:8578. the optimum value of ! may have to be determined by trial and error. there are two real eigenvectors and one real principal vector.relation reduces to that shown in Fig. the optimum value for the stationary case.72) !opt = 2= 1 + sin M + 1 For M = 40. 9. Hence the convergence rate is j m jmax = !opt .1 m sin 2 where m j M +1 (9.1 j . m = 1 . the . In practical applications. 1 (9.23 times its initial value in only 10 iterations. and if h = 1.

2h=3)n + c nh(1 . h = 1. 2p2i)h=9]n~ x n~ c 1 . The numerical solution written in full is ~n .0 λmh Figure 9. H and C . The nonstationary form of Eq.39 is 1 .0 2 real defective λ m σm 2 complex λ m 1 real λ m Μ=5 0. (10 .5: The relation for optimum stationary SOR. 2h=3)n. This process changes the Point-Jacobi method to Richardson's method in standard terminology. are varied at each time step. h. (10 + 2 2i)h=9] x c (1 .182 1.5 Nonstationary Processes In classical terminology a method is said to be nonstationary if the conditioning matrices. It is much simpler. RELAXATION METHODS h = 1. however. 9. ~1 = + + + + c (1 . ~ b .0 -1. This does not change the steadystate solution A. M = 5.0 -2.75) 9. to study the case of xed H and C but variable step size. but it can greatly a ect the convergence rate. 2h=3)n~ x p c 1 .0 0.0 CHAPTER 9. ]~ x c (1 . In our ODE b f approach this could also be considered and would lead to a study of equations with nonconstant coe cients. For the Gauss-Seidel and SOR methods it leads to processes that can be superior to the stationary methods. 4h=3)n~ x 1 2 1 2 3 2 3 4 5 4 5 1 (9.

Mathematically stated. This is the basis for the multigrid methods discussed in Chapter 10. Let us consider the very important case when all of the m are real and negative (remember that they arise from a conditioned matrix so this constraint is not unrealistic for quite practical cases). for m = 1 2 M . Now we pose a less demanding problem. the eigenvalues m are generally unknown and costly to compute. It is therefore unnecessary and impractical to set hm = . In the annihilation process mentioned after Eq. 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. 9. ~ 1 = X cm~ m Y (1 + e x m=1 n=1 m hn ) (9. the error term can theoretically be completely eliminated in M steps by taking hm = . However. we considered making the error exactly zero by taking advantage of some knowledge about the discrete values of m for a particular case. NONSTATIONARY PROCESSES N ~ N = c ~ Y (1 + hn ) + x 1 1 183 + cm~ m x M hn ) + ~ 1 N Y n=1 n=1 1 (1 + m hn ) + + cM ~ M x N Y n=1 (1 + (9. we seek b max a j(Pe)N ( )j = minimum with(Pe)N (0) = 1 (9.80) . Now focus attention on the polynomial (Pe)N ( ) = (1 + h )(1 + h ) 1 2 (1 + hN ) (9. This leads to the concept of selectively annihilating clusters of eigenvectors from the error terms as part of a total iteration process.5. We will see that a few well chosen h's can reduce whole clusters of eigenvectors associated with nearby 's in the m spectrum. Since hn can now be changed at each step.1= m .9.78) where Pe signi es an \Euler" polynomial.76) where the symbol stands for product. Consider one of the error terms taken from M N ~ N ~ N .79) treating it as a continuous function of the independent variable .76.1= m for m = 1 2 : : : M .77) and write it in the form cm~ m Pe( m ) cm~ m x x N Y n=1 (1 + m hn ) (9.

cos (2n . As an example for the use of Eq.1 y 1 and 2 2 (9.85) hn = 2= 3 .84. 1 (9. y . ampli ed by the coe cients cm Q(1 + m hn). With each x eigenvector there is a corresponding eigenvalue. and it leads to the nonstationary step size choice given by ( " #) 1 = 1 . 9. ! (Pe)N ( ) = where (9. In relaxation terminology this is generally referred to as Richardson's method. b a b This problem has a well known solution due to Markov. a. ~ m .76 is expressed in terms of a set of eigenvectors.86) . .184 CHAPTER 9. 1) n = 1 2 :::N (9. ) cos (2n . It is 2 . 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 .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. 1) 6 " #! (9.2 . Eq.1 using only 3 iterations. TN . The three values of h which satisfy this problem are (9.83) 2 are the Chebyshev polynomials for jyj > 1.82) q q N N 1 TN (y) = 2 y + y . 9. b ! . + ( . 1 + 1 y . b TN a. let us consider the following problem: Minimize the maximum error associated with the eigenvalues in the interval .81) TN (y) = cos(N arccos y) are the Chebyshev polynomials along the interval . The error in the relaxation process represented by Eq. RELAXATION METHODS a.

89) would result. 8] =2 99 3 3 3 A plot of Eq. If instead the implicit trapezoidal rule n+1 = n+ is used. was used to integrate the basic ODE's.1 have been reduced to less than about 1% of their initial values. the nonstationary formula 1 h( 0 + 0 ) n 2 n +1 (9. NONSTATIONARY PROCESSES and the amplitude of the eigenvector is reduced to (Pe) ( ) = T (2 + 3)=T (3) 3 3 3 185 (9.88) where n p p o T (3) = 3 + 8] + 3 . 2 hn =1 =1 1 mC C + ~1 A m (9. p 0) h = 4=(6 + 3) 1 2 3 p Return now to Eq. This was derived from Eq.90) b max a j(Pt)N ( )j = minimum . 9.87) (9. 9.37 on the condition that the explicit Euler method.91) @ 1 A n 1 .5.6 are h = 4=(6 . 3) h = 4=(6 . This calls for a study of the rational \trapezoidal" polynomial.6 and we see that the amplitudes of all the eigenvectors associated with the eigenvalues in the range . 9.2 . The values of h used in Fig.92) . 9.76. namely that =1 0 1 M N 1+ h ~ N = X cm~ m Y B 2 n B x @ 1 m n 1 .9. 2 hn under the same constraints as before. with (Pt)N (0) = 1 (9.41. 9.87 is given in Fig. Eq. 9. Pt: 0 1 1 N B 1 + hn C Y (Pt )N ( ) = B 2 C (9.

8 −1.8 −1. RELAXATION METHODS 1 0.2 −1 λ −0.015 0.2 0.6 −0.4 −0.2 0 Figure 9.4 0.01 −0.186 CHAPTER 9.2 .8 −0.6 (Pe )3 (λ) 0.4 −1.8 −0.1. .6 −0.4 −0.01 0. minimization over .015 −0.2 −1 λ −0.6: Richardson's method for 3 steps.6 −1.5 0.4 −1.005 −0.9 0.005 (P ) (λ) e 3 0 −0.3 0.02 −2 −1.02 0.8 0.1 0 −2 −1.2 0 0.6 −1.7 0.

Using Appendix B. f show that 1 ~n = Gn ~ + (I . 3. The values of h used in Fig.7 are h = 1 p h = 2 h = 2 ( 1) ( 1) 3 3 1 2 3 9. 9. Show that this iterative method can be written in the form H (xk . For a linear system of the form (A + A )x = b. Gn)A. A )xn + b x (I + A )xn = (I . A )~ + b x where is a parameter. (You do not have to nd H .1=2.93) b hn b This process is also applied to problem 9. The results for (Pt) ( ) are shown in Fig. Given a relaxation method in the form ~ H ~n = A~n . b 1 2 2 +1 1 +1 1 2 Determine the iteration matrix G if = . The error amplitude is about 1/5 of that found for (Pe) ( ) in the same interval of . Recall that the eigenvalues of H . 9.9.2.2 1) and ! = !opt. Then nd the corresponding j m j values. nd the eigenvalues of H . A satisfy A~ m = mH~ m . 2 =. A.6 Problems 1. f ~ 0 1 1 2 where G = I + H . 1 1 1 . = N .) Find the numerical values.6.85.7. 2. but we settle here for the approximation suggested by Wachspress n. consider the iterative method (I + A )~ = (I . xk ) = (A + A )xk . x x not just the expressions. PROBLEMS ! 187 The optimum values of h can also be found for this problem. a n=1 2 N (9. A for the SOR method with A = B (4 : 1 . .

2 0 Figure 9.8 −1.8 0.6 −1.2 −1 λ −0.4 −1.6 −0.2 .4 −0.2 −1 λ −0.4 0.188 CHAPTER 9. .1.5 0.5 −1 −1.3 0.5 1 0.6 −0.2 0.6 (P ) (λ) t 3 0.7: Wachspress method for 3 steps.6 −1.8 −0.5 −2 −2 −1.1 0 −2 −1.4 −1.8 −1.7 0.8 −0.2 0 2 x 10 −3 1.5 (Pt )3 (λ) 0 −0. minimization over .9 0.4 −0. RELAXATION METHODS 1 0.

PROBLEMS 189 4. 6x = 0 @x For the initial condition. Plot the solution after the residual is reduced by 2. Solve the following equation on the domain 0 x 1 with boundary conditions u(0) = 0. Iterate to steady state using (a) the point-Jacobi method.6. the number of iterations. Use second-order centered di erences on a grid with 40 cells (M = 39). (b) the Gauss-Seidel method. Compare with the theoretical asymptotic convergence rate. Plot the logarithm of the L -norm of the residual vs.5.9. use u(x) = 0. Determine the asymptotic convergence rate. u(1) = 1: @ u . (c) the SOR method with the optimum value of !. and (d) the 3-step Richardson method derived in Section 9. and 4 orders of magnitude. 2 2 2 . 3.

RELAXATION METHODS .190 CHAPTER 9.

1 Eigenvector and Eigenvalue Identi cation with Space Frequencies Consider the eigensystem of the model matrix B (1 . That is. if the eigenvalues are ordered such that j j j j 1 2 2 j Mj (10.2 1). the space-frequency (or wavenumber) of the corresponding eigenvectors also increase.1) then the corresponding eigenvectors are ordered from low to high space frequencies.2 and 4. This has a rational explanation from the origin of the banded matrix.1. ~ (10. Notice that as the magnitudes of the eigenvalues increase. Note that @ sin(mx) = .3. 10.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.3) xx~ = x x 191 2 2 2 2 1 .2) @x and recall that 1 B (1 .m sin(mx) (10.1 Motivation 10.2 1)~ = X 1 D(~ ) X . The eigenvalues and eigenvectors are given in Sections 4. There are many variations of the process and many viewpoints of the underlying theory. respectively.3. The viewpoint presented here is a natural extension of the concepts discussed in Chapter 9.3.

this method produces the same value of j j for min and max .2 Properties of the Iterative Method The second key motivation for multigrid is the following: Many iterative methods reduce error components corresponding to eigenvalues of large amplitude more e ectively than those corresponding to eigenvalues of small amplitude. This is the key concept underlying the multigrid process. This is consistent with the physics of di usion as well.1. the operation 1 D(~ ) represents the numerical approximation of the multiplication of the x appropriate sine wave by the negative square of its wavenumber. However. the complete counterexample of the above association is contained in the eigensystem 1 for B ( 2 1 1 ). 2 2 2 2 2 10. In fact. the property can be restored by using h < 1.m m << M (10. This is to be expected of an iterative method which is time accurate.4. by design. One nds that 1 M + 1 . from Appendix B.2 + 2 cos m .5. of the Richardson method described in Section 9.4) m= M +1 x Hence. MULTIGRID where D(~ ) is a diagonal matrix containing the eigenvalues.192 1 CHAPTER 10. 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. 10.2 The Basic Process First of all we assume that the di erence equations representing the basic partial di erential equations are in a form that can be related to a matrix which has certain . The classical point-Jacobi method does not share this property.1. and X ~ . error components corresponding to high space frequencies will be reduced more quickly than those corresponding to low space frequencies. a sine synthesis. for example. 9. We have seen that X .2. ~ represents a sine transform. this correlation is not necessary in general. of the Gauss-Seidel method and. . exactly the opposite 2 behavior. the correlation of large magnitudes of m with high space-frequencies is to be expected for these particular matrix operators. as shown in Fig.m . However. Therefore. It is also true. For this matrix one nds. As we saw in Section 9.

6) Recall that the ~ used to compute ~ is composed of the exact solution ~ 1 and the r ~ in such a way that error e A~ . ~ b ] = 0 f (10. In Eq. if f ~ 1 is a vector representing the desired exact solution. as in the example given by Eq. The eigenvectors associated with the eigenvalues having largest magnitudes can be correlated with high frequencies on the di erencing mesh. ~ b] r f (10. 5.5) where the matrix formed by the product CAb has the properties given above.10. of the matrix are all real and negative. the residual. we are led to the starting formulation C Ab ~ 1 . 9.27. At the end of the three steps we nd ~ . The m are fairly evenly distributed between their maximum and minimum values. The iterative procedure used greatly reduces the amplitudes of the eigenvectors 1 associated with eigenvalues in the range between 2 j jmax and j jmax. 9. or it can be \contrived" by further conditioning. 3. The eigenvalues. ~ = 0 e r (10. Having preconditioned (if necessary) the basic nite di erencing scheme by a procedure equivalent to the multiplication by a matrix C . m . We do not attempt to develop an optimum procedure here. r where ~ = C Ab~ . 10. 4.2. as in the examples given by Eq. and initial guess for ~ 1 and proceed through n iterations making use of some iterative process that satis es property 5 above. 2.8) .6 is used. We start with some any. These conditions are su cient to ensure the validity of the process described next. This form can be arrived at \naturally" by simply replacing the derivatives in the PDE with di erence schemes. the vector ~ b represents the boundary conditions and the forcing function. but for clarity we suppose that the three-step Richardson method illustrated in Fig. THE BASIC PROCESS 193 basic properties.5.7) where A CAb (10. The problem is linear. 3. The basic assumptions required for our description of the multigrid process are: 1.11.

we can write PA P .11) Multiply Eq. 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 6 5 4 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 76 0 76 76 0 76 76 0 76 76 76 0 76 76 0 76 54 1 32 e e e e e e e 1 2 3 4 5 6 7 3 7 7 7 7 7 7 7 7 7 7 7 5 "~ # ee ~ ~o = P e e (10. and the eigenvalues of the Richardson process in the form: M= M ~ = X cm~ m Y ( m hn)] + X cm~ m Y ( mhn)] e x x 2 3 3 m=1 n=1 m=M=2+1 | very low amplitude {z n=1 (10. since a permutation matrix has an inverse which is its transpose. we can be sure that the high frequency content of ~ has been greatly reduced (about 1% or less of its original value in the initial guess).13) Notice that e r e A ~e + A ~o = ~e 1 2 (10. assumption 4 ensures that the error has been smoothed.7 from the left by P and. We can write the exact solution for ~ in terms of e e the eigenvectors of A.194 If one could solve Eq.12) 1 The operation PAP . 10. 10.9) Thus our goal now is to solve for ~ .~ e CHAPTER 10.7 for ~ then e ~1 = ~ . partitions the A matrix to form 1 2 3 A 1 A2 " ~ # " ~ # 6 7 ee = re 4 5 ~ ~o r A 3 A4 e o (10. 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). e In addition.14) . P ]~ = P ~ e r (10. one containing the odd entries. Next we construct a permutation matrix which separates a vector into two parts. MULTIGRID (10.10) } Combining our basic assumptions.

2. r If the boundary conditions are Dirichlet. 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. the averaging of ~ is our only approximation.10. and one can write for the example case 2 1 0 0 16 1 1 0 A02 = 2 6 0 1 1 6 4 3 7 7 7 5 is an exact expression. in this formulation. Since the top half of the frequency spectrum has been removed.17) (10.14 reduces to A ~ e + A A0 ~ e = ~ e e e r 1 2 2 (10.19) or Ac~ e . notice that. It is that there is some connection between ~ e and ~ o brought about by the smoothing property of the Richardson e e relaxation procedure.18) (10.15) e 2 1 (e + e ) e b 2 1 2 3 2 4 5 4 6 2 7 6 (10.16) 0 0 1 With this approximation Eq. no aliasing of these functions can occur in subsequent steps. It is also important to notice that we are dealing with the relation between ~ and ~ so the original boundary conditions and e r forcing function (which are contained in ~ in the basic formulation) no longer appear f in the problem. Hence. At this point we make our one crucial assumption. e no operations on ~ are required or justi ed. Finally. it is reasonable to suppose that the odd points are the average of the even points. ~ e = 0 e r where Ac = A + A A0 ] 1 2 2 . 10. ea and eb are zero. For example 1 (e + e ) e 2 a 1 (e + e ) e 2 1 (e + e ) or ~ o = A0 ~ e e e (10.

However.1.19.196 CHAPTER 10.2 . When eb = eM . For Neumann conditions.2 { z }| 3 2 1 1 7+6 1 1 5 4 A2 1 1 { 3 7 5 2 z }| 2 1 16 1 1 6 4 26 1 1 A 0 1 { Ac 3 2 }| z . the interpolation formula in Eq.1. 10. B.2 . all of the properties given in Section 10. our 7-point example would produce 2 .22) and are illustrated in Fig. It is easy to show that the high space-frequencies still correspond to the eigenvalues with high magnitudes. the example in Eq.2 3 7 7 7 17 2A A 3 7 7 6 1 27 7=4 7 5 7 7 A3 A4 7 7 7 5 (10. and.2 6 6 6 6 6 6 . is completely determined by the choice of the permutation matrix and the interpolation approximation.2 6 . the matrix on the coarse mesh.2 ::: . ea = e .1 { 3 7 (10. A in the 1-D model equation is B (1 ~ 1) where ~ = . In the particular case illustrated in Fig. MULTIGRID The form of Ac.1 1=2 7 5 1=2 .2 . in fact. 9. 10.18 gives z }| 2 . 1) xjm = sin j 2M + 1 m=1 2 M (10.1 1=2 7 6 7 = 4 1=2 .2 4 A1 .23) 4 5 0 0 2 1 2 . If Neumann conditions are on the left.15 must be changed. If the original A had been B (7 : 1 .16 changes to 2 1 0 03 7 16 A0 = 2 6 1 1 0 7 60 1 17 (10.21) 5 If the boundary conditions are mixed Dirichlet-Neumann.20) and Eq.2 1 1 1 1 1 1 .1]T . The eigensystem is given by b b Eq.2 1 1 1 . 10. eb is equal to eM .2 . 10.1 = 6 PAP 6 1 6 6 6 1 6 6 4 .2 . 10. and all of them re ect on the right (Neumann) side. the eigenvector structure is di erent from that given in Eq.1 are met.55 for Dirichlet conditions. In the present case they are given by " !# (2m . All of them go through zero on the left (Dirichlet) side.2 1).

1 in the lower right element).1 7 5 1=2 .26) ~ m = sin j m x m = .2 { 2 }| 3 z 1 1 7+6 1 1 5 4 A2 1 1 { 3 7 5 2 z }| 2 1 61 1 16 4 26 1 1 A 0 2 { Ac 3 2 }| z ." We will continue with Dirichlet boundary conditions for the remainder of this Section. The permutation matrix remains the same and both A and A in the partitioned matrix PAP . cos m=1 2 M M +1 M +1 1 1 . 10.10. 10.25) For A = B (M : 1 . the relationship between e e In order to examine this relationship. THE BASIC PROCESS 197 X Figure 10.2 4 A1 .2.9. Therefore. Recall that our goal is ne mesh to 2 e r to solve for ~ .1=2 { 3 1=2 7 (10. which will provide us with the solution ~ 1 using Eq.15.2 1) the eigenvalues and eigenvectors are m j=1 2 M (10. we have reduced the problem from B (1 .2 1 . Given ~ e e e computed on the coarse grid (possibly using even coarser grids). we must now determine e ~ e and ~ . we can construct the coarse matrix from 1 1 2 4 }| z 2 .2 1)~e = ~ e on the next coarser mesh. we can compute ~ o e using Eq.24) 5 which gives us what we might have \expected.1 1=2 7 6 7 = 4 1=2 . Ac = Xc cXc.2 1)~ = ~ on the e r 1 B (1 . In order to complete the process. are unchanged (only A is modi ed by putting . At this stage. (10.2 6 . and thus ~ . we need to consider the eigensystems of A and Ac: A = X X.1: Eigenvectors for the mixed Dirichlet{Neumann case.

31) 2 3 1 607 6 7 = Xc .2 1) are 2 ( c) = . If we restrict our attention to odd M .33) 1 1 1 1 1 1 1 1 1 . cos M2+ 1 (~ c) = sin j M2+ 1 j = 1 2 x Mc (10. ~ e = Xc . ~ = ~ ..2 1 .198 CHAPTER 10. i. corresponding to ~ = sin j = . we obtain ( c) = .28) For M = 51 (Mc = 25). 7 4 5 0 (10. = . x Now let us consider the case in which all of the error consists of the eigenvector component ~ .32) 2 3 1=( c) 6 0 7 6 7 = Xc 6 . Xc. cos M + 1 x j=1 2 M (10. The exact solution on the coarse grid x x satis es ~ e = 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 ~ e = (~ c) r x 1 1 1 1 (10.0:003649. (~ c) e x c r c (10. In addition.e. that is. one can easily see that (~ c) coincides with ~ at x x every second point of the latter vector.27) M +1 For example. 1 . 7 c 6 7 4 .. then Mc = (M . ( c) approaches 2 . As M increases. the most di cult error mode to eliminate is that with m = 1. 7 6 .30) since (~ c) contains the even elements of ~ . it contains the even elements of ~ . with M = 51.. The eigenvalue and eigenvector corresponding to m = 1 for the matrix Ac = 1 B (Mc 1 . Then the residual is x e x ~ = A~ = ~ r x x (10.0:007291 = 1:998 . 5 0 (10. 6 . 1)=2 is the size of Ac. MULTIGRID Based on our assumptions.

which gives Ab = x B (M : 1 . in addition to interpolating ~ e to the ne grid e x e (using Eq.and up-going paths. THE BASIC PROCESS = ( c) 1 1 199 (~ c) x 1 1 1 (~ ) 2 xc 1 (10. the matrix A = B (M : 1 . in each case the appropriate permutation matrix and the interpolation approximation de ne both the down. The problem to be solved on the coarse grid is the same as that solved on the ne grid.2 1) = 1 B (Mc : 1 .2 1) comes from a discretization of the di usion equation. we must multiply the result by 2.2 1) = x B (Mc : 1 .37) c 4 In our case.37.15). This is equivalent to solving or 1A ~ = ~ e r 2 ce e (10.40) which is precisely the matrix appearing in Eq. and usually is. x C x B (Mc : 1 . quite important but they are not discussed here.2 1) 4 c c 2 2 2 (10. 10. 10.38) C= x I 2 (10. However. The details of nding optimum techniques are. since xc = 2 x.2 1) and the preconditioning matrix C is simply 2 (10.2.36) 1 B (M : 1 .2 1)~ = ~ ee re (10. Thus we see that the process is recursive.10.35) Since our goal is to compute ~ = ~ .34) (10. The reduction can be. The remaining steps required to complete an entire multigrid process are relatively straightforward.39) Applying the discretization on the coarse grid with the same preconditioning matrix as used on the ne grid gives. obviously. carried to even coarser grids before returning to the nest level. but they vary depending on the problem and the user. .

Next compute the residual based on ~ : ~ ~ ~ ~ = A ~ . 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. f ~ (10. 1. .11.3 A Two-Grid Process CHAPTER 10. ~ e r 2 (2) (2) (2) 2 1 (2) 2 (10. Call the result ~ . (1) (2) 2 (1) 1 2 1 3. A (10. MULTIGRID ~ We now describe a two-grid process for the linear problem A ~ = f . f r ~ = AGn1 ~n . the restriction matrix is 2 3 0 1 0 0 0 0 0 R =6 0 0 0 1 0 0 0 7 (10. f = AGn1 ~n + A (I .45) 4 5 0 0 0 0 0 1 0 that is. Perform n iterations of the selected relaxation method on the ne grid. This type of restriction is known as \simple injection. Solve the problem A ~ = ~ on the coarse grid exactly: e r ~ = A. This gives ~ = Gn1 ~n + (I . AGn1 A.g." Some form of weighted restriction can also be used. Transfer (or restrict) ~ to the coarse grid: r ~ =R ~ r r (10.200 10. f .46) 1 2 See problem 1 of Chapter 9.41) 1 (1) 1 (1) 1 1 1 where G = I + H.43) 1 1 1 1 1 (1) (1) (1) 1 1 1 1 1 1 2. starting with ~ = ~n. 10. Gn1 ) A.. which can be easily generalized to a process with an arbitrary number of grids due to the recursive nature of multigrid.44) In our example in the preceding section. 9. the rst three rows of the permutation matrix P in Eq.21). Note that the coarse grid matrix denoted A2 here was denoted Ac in the preceding section.42) and H is de ned as in Chapter 9 (e. Eq. f (10. Gn1 ) A.

10.3. A TWO-GRID PROCESS
2 2

201

Here A can be formed by applying the discretization on the coarse grid. In the 1 preceding example (eq. 10.40), A = 4 B (Mc : 1 ;2 1). It is at this stage that the generalization to a multigrid procedure with more than two grids occurs. If this is the coarsest grid in the sequence, solve exactly. Otherwise, apply the two-grid process recursively. 4. Transfer (or prolong) the error back to the ne grid and update the solution:

~n = ~ ; I ~ e
+1 (1)

1 (2) 2

(10.47)
3 7 7 7 7 7 7 7 7 7 7 7 5

In our example, the prolongation matrix is
2 6 6 6 6 6 1 I2 = 6 6 6 6 6 6 4

1=2 1 1=2 0 0 0 0

0 0 1=2 1 1=2 0 0

0 0 0 0 1=2 1 1=2

(10.48)

which follows from Eq. 10.15. Combining these steps, one obtains

~n = I ; I A; R A]Gn1 ~ n ; I ; I A; R A]Gn1 A; f + A; f ~ ~
+1 1 2 2 1 2 1 1 1 2 2 1 2 1 1 1 1

(10.49) (10.50)

Thus the basic iteration matrix is

I ; I A; R A]Gn1
1 2 2 1 2 1 1 1 1

The eigenvalues of this matrix determine the convergence rate of the two-grid process. The basic iteration matrix for a three-grid process is found from Eq. 10.50 by replacing A; with (I ; G )A; , where
2 2 3 2

G = I ; I A; R A ]Gn2
2 3 2 3 3 1 3 2 2 2 2 2 3 3 2 3

(10.51)

In this expression n is the number of relaxation steps on grid 2, I and R are the transfer operators between grids 2 and 3, and A is obtained by discretizing on grid 3. Extension to four or more grids proceeds in similar fashion.

202

10.4 Problems

CHAPTER 10. MULTIGRID

1. Derive Eq. 10.51. 2. Repeat problem 4 of Chapter 9 using a four-grid multigrid method together with (a) the Gauss-Seidel method, (b) the 3-step Richardson method derived in Section 9.5. Solve exactly on the coarsest grid. Plot the solution after the residual is reduced by 2, 3, and 4 orders of magnitude. Plot the logarithm of the L -norm of the residual vs. the number of iterations. Determine the asymptotic convergence rate. Calculate the theoretical asymptotic convergence rate and compare.
2

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

204

11.1 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. NUMERICAL DISSIPATION

@u = ;a @u @t @x

(11.1)

with periodic boundary conditions. Consider the following point operator for the spatial derivative term

;a( x u)j = 2;ax ;(1 + )uj;1 + 2 uj + (1 ; )uj+1] = ;a (;uj;1 + uj+1) + (;uj;1 + 2uj ; uj+1)] 2 x

(11.2)

The second form shown divides the operator into an antisymmetric component (;uj;1+ uj+1)=2 x and a symmetric component (;uj;1 + 2uj ; uj+1)=2 x. The antisymmetric component is the second-order centered di erence operator. With 6= 0, the operator is only rst-order accurate. A backward di erence operator is given by = 1 and a forward di erence operator is given by = ;1. For periodic boundary conditions the corresponding matrix operator is

;a x = 2;ax Bp(;1 ; 2 1 ; )
The eigenvalues of this matrix are
m

= ;a x

m 1 ; cos 2M

m + i sin 2M

for m = 0 1 : : : M ; 1

If a is positive, the forward di erence operator ( = ;1) produces Re( m ) > 0, the centered di erence operator ( = 0) produces Re( m ) = 0, and the backward di erence operator produces Re( m) < 0. Hence the forward di erence operator is inherently unstable while the centered and backward operators are inherently stable. If a is negative, the roles are reversed. When Re( m) = 0, the solution will either grow or decay with time. In either case, our choice of di erencing scheme produces nonphysical behavior. We proceed next to show why this occurs.

11.2. THE MODIFIED PARTIAL DIFFERENTIAL EQUATION

11.2 The Modi ed Partial Di erential Equation
( xu)j = 2 1 x 42 x @u ; @x j

205

First carry out a Taylor series expansion of the terms in Eq. 11.2. We are lead to the expression

2

!

3 ! ! ! @ 2 u + x3 @ 3 u ; x4 @ 4 u + : : :5 x2 @x2 3 @x3 j 12 @x4 j j

We see that the antisymmetric portion of the operator introduces odd derivative terms in the truncation error while the symmetric portion introduces even derivatives. Substituting this into Eq. 11.1 gives

@u = ;a @u + a x @ 2 u ; a x2 @ 3 u + a x3 @ 4 u + : : : (11.3) @t @x 2 @x2 6 @x3 24 @x4 This is the partial di erential equation we are really solving when we apply the approximation given by Eq. 11.2 to Eq. 11.1. Notice that Eq. 11.3 is consistent with Eq. 11.1, since the two equations are identical when x ! 0. However, when we use a computer to nd a numerical solution of the problem, x can be small but it is not zero. This means that each term in the expansion given by Eq. 11.3 is excited to some degree. We refer to Eq. 11.3 as the modi ed partial di erential equation. We proceed next to investigate the implications of this concept. Consider the simple linear partial di erential equation @u = ;a @u + @ 2 u + @ 3 u + @ 4 u (11.4) @t @x @x2 @x3 @x4 Choose periodic boundary conditions and impose an initial condition u = ei x. Under these conditions there is a wave-like solution to Eq. 11.4 of the form u(x t) = ei xe(r+is)t provided r and s satisfy the condition r + is = ;ia ; 2 ; i 3 + 4 or r = ; 2 ( ; 2 ) s = ; (a + 2 ) The solution is composed of both amplitude and phase terms. Thus a ) u = e; 2 ({z; 2} ei x;({z+ 2 )t}] (11.5) | | amplitude phase

11.2. we see that the speed of propagation is a + 2 . this operator produces fourth-order accuracy in phase with a third-order dissipative term.5. the choice of a forward di erence scheme ( = . This is tantamount to deliberately adding dissipation to the PDE. If the wave speed a is positive.a x2 =6.3 we have = . NUMERICAL DISSIPATION It is important to notice that the amplitude of the solution depends only upon and . one could choose = 1=2 in Eq. 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. 6uj. Therefore. Therefore.2 . the coe cients of the even derivatives in Eq. For example.1 + 8uj+1 . the choice of a backward di erence scheme ( = 1) produces a modi ed PDE with . Referring to the modi ed PDE. This we refer to as dispersion. by using the antisymmetry of the matrix di erence operators. Under the same condition. .2 . 8uj. Any symmetric component in the spatial operator introduces dissipation (or ampli cation). 11.1 + 3uj + 2uj+1) = 1 (uj. a third-order backward-biased scheme is given by ( xu)j = 6 1 x (uj. the phase speed of the numerical solution is less than the actual phase speed. the numerical phase speed is dependent upon the wavenumber . Therefore. Eq. Furthermore. By examining the term governing the phase of the solution in Eq.206 CHAPTER 11.6) The antisymmetric component of this operator is the fourth-order centered di erence operator. the coe cients of the odd derivatives. uj+2) 12 x + (uj. 2 > 0 and hence the amplitude of the solution decays. Our purpose here is to investigate the properties of one-sided spatial di erencing operators relative to centered di erence operators. The resulting spatial operator is not one-sided but it is dissipative. 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). and the phase depends only on a and . One can easily show. 4uj+1 + uj+2)] (11. Note that the use of one-sided di erencing schemes is not the only way to introduce dissipation. The symmetric component approximates x3 uxxxx=12.1) is equivalent to deliberately adding a destabilizing term to the PDE. 11. 11. As a corollary. Biased schemes use more information on one side of the node than the other. 4uj.1 + 6uj .4.2 .

the corresponding fully-discrete matrix operator is 0 2 !3 ! 2 ! 31 1 4 ah + ah 25 1 . Consider the following Taylor-series expansion in time: 2 u(x t + h) = u + h @u + 1 h2 @ u + O(h3) @t 2 @t2 (11. u(n) ) + 1 ah (u(n) .1 2 x j+1 j.1 2 x This is the Lax-Wendro method applied to the linear convection equation. ah + ah 2 5A ~ ~ n+1 = Bp @ 2 x u un x x 2 x x The eigenvalues of this matrix are ! ah 2 1 .a @u (11. i ah sin 2 m m =1.8) @t @x @t2 @x2 Now replace the space derivatives with three-point centered di erence operators. backward di erencing must be used if the wave speed is positive. known as the Lax-Wendro method. Thus @t @x @ 2 u = a2 @ 2 u @u = .3 The Lax-Wendro Method 207 In order to introduce numerical dissipation using one-sided di erencing. It is equal to the ratio of the distance travelled by a wave in one time step to the mesh spacing. There is no intermediate semi-discrete stage. THE LAX-WENDROFF METHOD 11. the linear convection equation @u + a @u = 0). The quantity j ah j is known as the x Courant (or CFL) number. and forward di erencing must be used if the wave speed is negative. 2u(n) + u(n) ) (11. This explicit method di ers conceptually from the methods considered previously in which spatial di erencing and time-marching are treated separately. ah 2 1 4.9) uj j j +1 j j . x M x M for m = 0 1 : : : M . cos 2 m .7) First replace the time derivatives with space derivatives according to the PDE (in this case. 1 ah (u(n) . .3. 1 For j ah j 1 all of the eigenvalues have modulus less than or equal to unity and hence x the method is stable independent of the sign of a. It is a fully-discrete nite-di erence scheme.11. giving !2 (n+1) = u(n) . For periodic boundary conditions. Next we consider a method which introduces dissipation independent of the sign of the wave speed.

11) 2 x u +1 ~ which can be shown to be identical to the Lax-Wendro method. ah (u(jn) . a2h2 ) @ 3 u = . a2h2 ) @ 3 u . . a2 h ( x2 . a x2 (1 . 11. this approach leads to u(jn+1) = u(jn) . u(jn)1) ~ .6 n @x3 @x3 6 The dissipative term is proportional to 2 4 2 2 4 2 @ . a ( x2 . 1 Or vice-versa for nonlinear problems. the leading error term in the Lax-Wendro method is dispersive and proportional to a ( x2 . Recall MacCormack's timemarching method. depending on the sign). 11. presented in Chapter 6: un+1 = un + hu0n ~ 1 ~ ~ (11. Therefore. ah (~(jn+1) . a2 h2) @ 4 u + : : : @t @x 6 @x3 8 @x4 This is derived by substituting Taylor series expansions for all terms in Eq. Recall that the odd derivatives on the right side lead to unwanted dispersion and the even derivatives lead to dissipation (or ampli cation. Cn) @xu 4 @x4 This term has the appropriate sign and hence the scheme is truly dissipative as long as Cn 1.9 and converting the time derivatives to space derivatives using Eq. a8h ( x2 .10) un+1 = 2 un + un+1 + hu0n+1] If we use rst-order backward di erencing in the rst stage and rst-order forward di erencing in the second stage. a2h2 ) @ u = .8. The two methods di er when applied to nonlinear hyperbolic systems. Hence MacCormack's method has the same dissipative and dispersive properties as the Lax-Wendro method. a h8 x (1 . these should be applied alternately. A closely related method is that of MacCormack. The two leading error terms appear on the right side of the equation. however. x u(jn+1) = 1 u(jn) + u(jn+1) .1 a dissipative second-order method is obtained.208 CHAPTER 11. u(jn+1))] ~ (11. 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. C 2 ) @ 3 u . which is given by @u + a @u = . For the linear convection equation.

= 0.1 + 2uj .uj. the direction of the one-sided operator (i. In order to apply one-sided di erencing schemes to such systems. ( 0) term forward di erence. The above approach to obtaining a stable discretization independent of the sign of a can be written in a di erent. schemes in which the numerical dissipation is introduced in the spatial operator are generally preferred over the Lax-Wendro approach. 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. some form of splitting is required. then a+ = 0 and If a 0. that is. From Eq. 11. In this section. ) @u = 0 a = a 2 jaj @t @x + = a 0 and a. if a 0. we saw that numerical dissipation can be introduced in the spatial di erence operator using one-sided di erence schemes or. by adding a symmetric component to the spatial operator.12) @t @x where we do not make any assumptions as to the sign of a. When the ow is subsonic. Now for the a+ ( 0) term we can safely backward di erence and for the a.5. For more subtle aspects of implementation and application of such techniques to .11. UPWIND SCHEMES 11. we see that a stable discretization is obtained with = 1 if a 0 and with = . For example. the wave speeds can be both positive and negative. With this approach. Alternatively.uj. manner. Consider again the linear convection equation: @u + a @u = 0 (11..1x a(. This is the basic concept behind upwind methods. some decomposition or splitting of the uxes into terms which have positive and negative characteristic speeds so that appropriate di erencing schemes can be chosen. In the next two sections. However. a. 11.e. these are of mixed sign.1 + uj+1) + jaj(.13) This approach is extended to systems of equations in Section 11.12 as @u + (a+ + a.a( xu)j = 2. more generally.1 if a 0.1. uj+1)] (11.4.2.4 Upwind Schemes 209 In Section 11. the eigenvalues of the ux Jacobian for the onedimensional Euler equations are u u + a u . These are by no means unique. but entirely equivalent. = a 0. then a a. When a hyperbolic system of equations is being solved. This is achieved by the following point operator: . This is avoided in the Lax-Wendro scheme. we present two splitting techniques commonly used with upwind methods. We can rewrite Eq. as a result of their superior exibility. we present the basic ideas of ux-vector and ux-di erence splitting.

Premultiplying by X and inserting the product @x X .18) @t @x @t @x @x The spatial terms have been split into two components according to the sign of the wave speeds. We can now rewrite the system in terms of characteristic variables as @w + @w = @w + + @w + . i. In order to apply a one-sided (or biased) spatial di erencing scheme. contains the negative eigenvalues.19) . @w term. where += (11.1Xw + @X . + contains the positive eigenvalues and . A.17) 2 2 With these de nitions. is positive. @w = 0 (11. the reader is referred to the literature on this subject. we need to apply a backward di erence if the wave speed.1Xw = 0 @t @x @x (11.5 that a linear. = . constant-coe cient. hyperbolic system of partial di erential equations given by @u + @f = @u + A @u = 0 (11.4. and the wi's are the characteristic variables. i. into two components such that = ++ . . and a forward di erence if the wave speed is negative. let us split the matrix of eigenvalues.1 Flux-Vector Splitting Recall from Section 2.X .14) @t @x @t @x can be decoupled into characteristic equations of the form @wi + @wi = 0 (11.15) i @t @x where the wave speeds. 11. are the eigenvalues of the Jacobian matrix.16) +j j .1X in the spatial terms gives @Xw + @X +X .j j (11.210 CHAPTER 11. NUMERICAL DISSIPATION nonlinear hyperbolic systems such as the Euler equations. We can use backward di erencing for the + @w term and forward @x di erencing for the . To accomplish this.

UPWIND SCHEMES With the de nitions2 211 A+ = X +X . we are in e ect solving the characteristic equations in the desired manner.1 (11. has all negative eigenvalues. When an implicit time-marching method is used. the Jacobians of the split ux vectors are required. @f + 6= A+ @f . 6= A. A++ has eigenvalues which are all positive.. and A. In the nonlinear case.X . This approach is known as ux-vector splitting. = A.11.25) A.22) f . 2 With these de nitions A+ has all positive eigenvalues. f = Au. f is also equal to Au as a result of their homogeneous property.26) For the Euler equations. u @u + @f + + @f . (11. and A.23) @t @x @x In the linear case. .4.1 and recalling that u = Xw.24) Thus by applying backward di erences to the f + term and forward di erences to the f . one must nd and use the new Jacobians given by A++ = @f @u + (11. = @f @u . (11. term.20) @u + @A+u + @A. For the Euler equations. the de nition of the split uxes follows directly from the de nition of the ux. as discussed in Appendix C. Note that f = f+ + f.. = 0 (11. we obtain A. = X . u = 0 @t @x @x Finally the split ux vectors are de ned as f + = A+u and we can write (11. has all negative eigenvalues. @u @u Therefore.21) (11.

wL and wR.1X (wL + wR) + 2 X j jX . is known as ux-di erence splitting. which is nondissipative. We again begin with the diagonalized form of the linear. respectively. we require ( (w ) (^i)j+1=2 = i(wi)L if i > 0 g (11. as in Chapter 5. is given by 1 gj+1=2 = 2 (g(wL) + g(wR)) ^ (11. (wi)R] g (11. we consider the numerical ux at the interface between nodes j and j + 1. 11.30) or 1 gj+1=2 = 1 (wL + wR) + 2 j j (wL .27) @t @x The ux vector associated with this form is g = w.1X (wL .1 (11. This is achieved with 1 1 (^i)j+1=2 = 2 i (wi)L + (wi)R ] + 2 j ij (wi)L . wR ) ^ (11. hyperbolic system of equations @w + @w = 0 (11.32) and thus 1 f^j+1=2 = 1 (fL + fR ) + 2 jAj (uL . uR) (11. wR ) ^ (11. we premultiply by X to return to the original variables and insert the product X . more suited to nite-volume methods.29) i i R if i < 0 where the subscript i indicates individual components of w and g.31) 2 Now.4. as in Eq. gj+1=2.19.28) In order to obtain a one-sided upwind approximation.34) . Now. NUMERICAL DISSIPATION 11.1X after and j j to obtain 1 1 X gj+1=2 = 2 X X .2 Flux-Di erence Splitting Another approach.33) 2 where jAj = X j jX . The centered approximation to gj+1=2. as a ^ function of the states to the left and right of the interface.212 CHAPTER 11. In a nite-volume method. the uxes must be evaluated at cell boundaries. constant-coe cient.

In this case. Eq.37) For the Euler equations for a perfect gas.11. ARTIFICIAL DISSIPATION 213 and we have also used the relations f = Xg. uR) (11. u = Xw. constant-coe cient case. 11. Hence satisfaction of Eq. We have also seen that such dissipation can be introduced by adding a symmetric component to an antisymmetric (dissipation-free) operator. 11.35) R if all i s < 0 giving pure upwinding. this leads to an upwind operator which is identical to that obtained using ux-vector splitting. we would like our de nition of f^j+1=2 to satisfy 0 f f^j+1=2 = fL if all i 0s > 0 (11. 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. consider a situation in which the eigenvalues of A are all of the same sign. respectively.A.1.39) p L+p R R R where u and H = (e + p)= are the velocity and the total enthalpy per unit mass.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.37 is satis ed by the ux Jacobian evaluated at the Roe-average state given by p LuL + p R uR (11. and A = X X .3 ( 11. If the eigenvalues of A are all positive.38) uj+1=2 = p + p L R p H +p H L L Hj+1=2 = (11. of this chapter. uR) (11. jAj = A if they are all negative.35 is obtained by the de nition 1 f^j+1=2 = 2 (fL + fR ) + 1 jAj+1=2j (uL . fR = Aj+1=2 (uL .36) 2 if Aj+1=2 satis es fL . there is some ambiguity regarding the de nition of jAj at the cell interface j + 1=2. In order to resolve this. 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.5. In the linear. However. jAj = .

a s With appropriate choices of x and x.15 is stable if i 0 and a s unstable if i < 0. s It is common to use the following operator for x (11.6 Problems 1.41) i x = i x + j ij x Extension to a hyperbolic system by applying the above approach to the characteristic variables. This symmetric operator approximates x3 uxxxx and thus introduces a third-order dissipative term.uj+1 + 2uj . 11. NUMERICAL DISSIPATION s a ( xu)j = . let CHAPTER 11. uj.2 .42) x (Au) = x (Au) + x (jAju) or a s (11. 11.43. 11. A second-order backward di erence approximation to a 1st derivative is given as a point operator by ( xu)j = 1 (uj. x is stable if i 0 and unstable if i > 0.34. this approach can be related to the upwind approach. It is also sometimes referred to as arti cial di usion or arti cial viscosity. The second spatial term is known as arti cial dissipation. For details of how this can be implemented for nonlinear hyperbolic systems. With an appropriate value of .1 + 3uj ) 2 x .214 For example. but is beyond the scope of this book.44) x (uj. The appropriate implementation is thus a s (11. the reader should consult the literature. uj. this often provides su cient damping of high frequency modes without greatly a ecting the low frequency modes. Similarly. 4uj.43) x f = x f + x (jAju) where jAj is de ned in Eq. as in the previous two sections. 4uj+1 + uj+2) where is a problem-dependent coe cient.1 + 6uj . applying x = x .1 (11.1 2 x 2 x a s Applying x = x + x to the spatial derivative in Eq.40) ( x u)j = uj+1 . A more complicated treatment of the numerical dissipation is also required near shock waves and other discontinuities. s ( xu)j = 11.2 .36 and 11. 4uj. This is particularly evident from a comparison of Eqs. gives a s (11.

1.6. Show that the ux Jacobian for the 1-D Euler equations can be written in terms of u and H . 11. Form the plus-minus split ux vectors as in Section 11.2. 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. Plot the real and imaginary parts of x vs. Show that the use of the Roe average state given in Eqs. x for 0 x . x for 0 x . 4. Using Fourier analysis as in Section 6. PROBLEMS 215 (a) Express this operator in banded matrix form (for periodic boundary conditions).39 leads to satisfaction of Eq. (See Appendix A. 11.44. then derive the symmetric and skew-symmetric matrices that have the matrix operator as their sum.37. Plot the real and imaginary parts of x vs. Consider the spatial operator obtained by combining second-order centered differences with the symmetric operator given in Eq. 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. Find the modi ed wavenumber for this operator with = 0 1=12 1=24. . 5. Plot the real and imaginary parts of x vs.3 to see how to construct the symmetric and skew-symmetric components of a matrix. Find the modi ed wavenumber for the rst-order backward di erence operator. x for 0 x .6.2. Find the matrix jAj.) (b) Using a Taylor table. 6.11. 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. 2. x for 0 x . Using Fourier analysis as in Section 6.2.4. Consider the hyperbolic system derived in problem 8 of Chapter 2. Using Fourier analysis as in Section 6. Find the modi ed wavenumber for the operator given in Eq.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. and 1=48. 3. 11.6.6. x for 0 x .

NUMERICAL DISSIPATION .216 CHAPTER 11.

Matrices can be split in quite arbitrary ways. They are the basis for a wide variety of methods variously known by the labels \hybrid".1) . and a means for analyzing such modi ed forms. h. Factored forms are especially useful for the derivation of practical algorithms that use implicit methods. Now recall the generic ODE's produced by the semi-discrete approach d~ = A~ . we present and analyze split and factored algorithms. When we approach numerical analysis in the light of matrix derivative operators.1 The Concept Factored forms of numerical operators are used extensively in constructing and applying numerical methods to problems in uid mechanics. 2. 3. Let us start with the following observations: 1. 12. ~ u u f dt 217 (12. Advancing to the next time level always requires some reference to a previous one. 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. \time split". and \fractional step". the concept of factoring is quite simple to present and grasp.Chapter 12 SPLIT AND FACTORED FORMS In the next two chapters. Time marching methods are valid only to some order of accuracy in the step size.

Both of these considerations are investigated later. from observation 3 (allowing us to drop higher order terms . . Then. +1 1 2 2 ~ n = I + hA ] I + hA ]~ n . From observation 1 (arbitrary splitting of A) : (12.4 have the same formal order of accuracy and.5) ~ ~ n = I + hAd + hAc]~ n + h(bc) + O(h ) u u +1 2 (12. h~ + O(h ) u u f +1 2 1 2 Finally. from observation 2 (new data ~ n in terms of old ~ n): u u 1 2 1 1 +1 ~ n = I + hA + hA ]~ n . SPLIT AND FACTORED FORMS d~ = A + A ]~ .2) where A = A + A ] but A and A are not unique. and techniques to carry out their numerical evaluation can have arithmetic operation counts that vary by orders of magnitude. h~ + O(h ) u u f 12. For the time march let us choose the simple.h A A ~ n): u (12. neither one is to be preferred over the other. h~ + O(h ) u u f +1 1 2 2 1 2 2 1 2 (12. 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. However. their numerical stability can be quite di erent.3 and 12.2 Factoring Physical Representations | Time Splitting Suppose we have a PDE that represents both the processes of convection and dissipation. Choose again the explicit Euler time march so that d~ = A ~ + A ~ + (bc) u ~ cu du dt (12. h A A i~ n . 12.3) 2 or its equivalent ~ n = h I + hA ] I + hA ] . in this sense. Here we seek only to apply to some simple cases the concept of factoring. explicit Euler method.6) 1 Second-order time-marching methods are considered later. rst-order.218 CHAPTER 12. respectively and their sum forms the A matrix we have considered in the previous sections. ~ u u f dt 1 2 2 and consider the above observations.4) Notice that Eqs.

and a rst-order time-march method is usually unsatisfactory. However. = I + hAd + h Ad + if h jjAdjj < 1. the important aspect of stability has yet to be discussed.7 are often applied in a predictor-corrector sequence. as before. FACTORING PHYSICAL REPRESENTATIONS | TIME SPLITTING 219 Now consider the factored form ~n u +1 Original Unfactored Terms Higher Order Terms and we see that Eq. on this basis. I + hAc]~ n + h(bc) u ~ I + hAd + hAc]~ n + h(bc) +O(h ) u | {z } 1 2 (12. requiring a tridiagonal solver if the di usion term is central di erenced. Therefore.7 and the original unfactored form Eq. hAd]~ n u +1 +1 ~ = I + hAc]~ n + h(bc) u = un ~ +1 (12. stability.6 with the same order of accuracy is given by the expression ~n u +1 Original Unfactored Terms where in this approximation we have used the fact that 1 2 2 = = ~ I . where jjAd jj is some norm of Ad ]. 12.10) The convection operator is applied explicitly. this factored form would appear to be superior to that provided by Eq.9) I . For example.8) Factoring can also be useful to form split combinations of implicit and explicit techniques. We have yet to determine its . hAd]. hAd ]. In practical applications equations such as 12.6 have identical orders of accuracy in the time approximation. 12. 12.7) u {z } | 2 2 | {z } un ~ ~n u +1 +1 = = ~ I + hAc]~ n + h(bc) u I + hAd]~n u +1 (12.12.2. 12. but the di usion operator is now implicit. Since numerical sti ness is generally much more severe for the di usion process. their selection is arbitrary. another way to approximate Eq.8. This time a predictor-corrector interpretation leads to the sequence un ~ I . 2 We do not suggest that this particular method is suitable for use. In this case one could write 2 = = ~ I + hAd ] I + hAc]~ n + h(bc) u ~ u ~ I + hAd + hAc]~ n + h(bc) + h Ad Ac~ n + (bc) +O(h ) (12.

the number of (interior) y points is 3.and three-dimensional ows. The numbers 11 12 43 represent the location in the mesh of the dependent variable bearing that index. un = A u + A u + (bc) + O(h ) ~ c n d n h +1 +1 2 We should mention here that Eq. 3 This could also be called a 5 .12. We choose the 3 4 point mesh shown in the Sketch 12. In this example Mx .and 3-D requires some reference to a mesh. 12.1 Mesh Indexing Convention 2 2 12.11) My k 1 13 23 33 43 12 22 32 42 11 21 31 41 1 (12. the number of (interior) x points.6 in the form Then which is identical to Eq.3. 3 32 @u = @ u + @ u @t @x @y 2 2 (12. 12. ~ I .10. but in these notes we describe the size of a mesh by the number of interior points.220 CHAPTER 12. Thus u represents the value of u at j = 3 and k = 2. 6 point mesh if the boundary points (labeled in the sketch) were included. 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.9 can be derived for a di erent point of view by writing Eq.12) j Mx Mesh indexing in 2-D. hAd ]un = I + hAc]un + h(bc) +1 12. 12. is 4 and My . SPLIT AND FACTORED FORMS un . 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.

16 part a and b. 12. however.3. 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 . consecutively along columns that are consecutive from j = 1 to Mx. and one composed of y-vectors with U y . 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. we use the notation Axx y or Axy y . Notice that the matrices are not the same although they represent the same derivative operation.14) Axx y = Pxy Axy y Pyx ( ) + ( ) + (12.16 part a and b. When the matrix is multiplying a space vector U . FACTORING SPACE MATRIX OPERATORS IN 2{D 221 12. Examples are in Eq.15) Ax+y and U . respectively. and (2). depending on the data{base chosen for the U it multiplies.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. we consider only two: (1).11 can be written + 4 If it is important to be speci c about the data-base structure.13) where Pyx = PT = P. ( ) ( ) 12. Of these. Examples of the order of indexing for these space vectors are given in Eq. 12. xy xy Now consider the structure of a matrix nite-di erence operator representing 3point central-di erencing schemes for both space derivatives in two dimensions. In this notation the ODE form of Eq. u 4 Notice that .12. are subsets of A and ~ . we label a data{base composed of x-vectors with U x . and they are related by the same permutation matrix that relates U x to U y . Thus ( ) + ( ) + ( ) ( ) dU = A U + (bc) ~ x y dt + (12. Their structures are similar. used in the previous sections. (1) and (2) above are referred to as xvectors and y-vectors. 12.3. In particular. consecutively along rows that are themselves consecutive from k = 1 to My . There are many ways to lay out a data-base.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. To be speci c about the structure.3. the usual (but ambiguous) representation is given by Ax y . which are notations used in the special case of space vectors.

Ax y .. 12 22 32 42 13 23 33 43 11 21 31 41 .. + ( ) 2 6o 6 6 6 6 6 6 6x 6 6 6 6 6 6 (y ) (y ) Ax+y U = 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4 j x j o j x j o j x j o j j j j o j x j x j o o j x j x j o j x j j x j o j x j x j o o j x j x j o j j j j + j x j o j x j o j x j o ( ) 3 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 x7 7 7 7 7 7 7 o5 . for y ! o. Ax y . 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 . matrix operator. . Entries for x ! x.12. ..222 2 6x 6 6 6 6 6 6 6 6 6 6o 6 6 (x) (x) Ax+y U = 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4 CHAPTER 12. for both ! . a:Elements in 2-dimensional. Entries for x ! x. matrix operator. central-di erence. Data base composed of Mx y{vectors stored in U y . 41 42 43 31 32 33 21 22 23 11 12 13 (12. . central-di erence..16) b: Elements in 2-dimensional. Data base composed of My x{vectors stored in U x . for 3 4 mesh shown in Sketch 12.. for 3 4 mesh shown in Sketch 12.12. for both ! . for y ! o.

2. hAyy Un = U y ( ) ( ) ( ) ( ) ( ) ( ) +1 h i (12.22. The permutation relation also holds for the split matrices so Ayx = Pxy Ayy Pyx ( ) ( ) and Axx = Pxy Axy Pyx ( ) ( ) The splittings in Eqs.17 and 12. applying the implicit Euler method to Eq. hAyx Unx = Unx + h(bc) + O(h ) ( ) ( ) ( ) +1 ( ) 2 (12.4 Space Splitting and Factoring We are now prepared to discuss splitting in two dimensions. 12.3.3.14 gives h i ~ Unx = Unx + h Axx + Ayx Unx + h(bc) ( ) +1 ( ) ( ) ( ) ( ) +1 or h i ~ I .12.23.19) (12. hAyx Unx = Unx + h(bc) + O(h ) ( ) ( ) ( ) +1 ( ) 2 Write this in predictor-corrector form and permute the data base of the second row. FACTORING SPACE MATRIX OPERATORS IN 2{D 223 12. 12.17) where Axx and Ayx are shown in Eq. 12.2.21) . There results ~ ~ I .20) As in Section 12. As an example ( rst-order in time). hAxx . hAxx U x = Unx + h(bc) h i y ~ I . hAxx I .18) where the split matrices are shown in Eq. Similarily Axy y = Axy + Ayy ( ) + ( ) ( ) (12. 12.18 can be combined with factoring in the manner described in Section 12. we retain the same rst order accuracy with the alternative h ih i ~ I . It should be clear that the matrix Axx y can be split into two matrices such that ( ) + Axx y = Axx + Ayx ( ) + ( ) ( ) ( ) ( ) (12.

SPLIT AND FACTORED FORMS Axx ( ) 2 x 6x 6 6 6 6 6 6 6 6 6 6 6 6 (x) U =6 6 6 6 6 6 6 6 6 6 6 6 6 6 4 x x x x x x x x j j j j j x x j x x x j x x x j x x j j j j j j j j Ayx ( ) 6 6 6o 6 6 U (x) = 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4 j j j j 2 o j o 6 o j o 6 6 6 o j o 6 6 o j o 6 6 j o o j o o j o o j o j o j o j o j o The splitting of Axx y .22) ( ) .224 CHAPTER 12. ( ) + 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.

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 . FACTORING SPACE MATRIX OPERATORS IN 2{D 225 Axy ( ) 2 x 6 6 6 6 6 6 6 6x 6 6 6 6 6 6 U (y) = 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4 j x j x j x j x j x j j j j j x j x j x j x j x j x j x j x j j x j x j x j x j x j x j x j x j Ayy ( ) 6 6 6 6 6 6 6 6 6 (y ) U =6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4 j j j 2 o o j 6o o o j 6 6 o o j 6 6 j x j x j x j x j x j x 3 j j 7 j j 7 7 j j 7 7 j j j 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 o5 3 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 x7 7 7 7 7 7 7 5 Uy ( ) (12.3.12. .

the use of factored forms becomes a very valuable tool 5 A form of the Douglas or Peaceman-Rachford methods. 2 hAy . .26 and 12. 1 h A A U + h(bc) + O(h ) (12.2 y n = I+2 x ~ 2 n +1 3 2 +1 2 3 2 +1 3 +1 +1 3 5 +1 +1 3 For idealized commuting systems the methods given by Eqs. 12.226 12. 12. 1 hAy Un = I + 2 hAx + 1 hAy Un + h(bc) + O(h ) (12. 1 hAx I .14 where the derivative operators have been split as in Eq. 1 hAy Un = I + 2 hAx I + 1 hAy Un + h(bc) + O(h )(12. 2 hAx . An alternative form of this kind of factorization is the classical ADI (alternating direction implicit) method usually written 1 1 ~ I . For example.25) ~ = I+ 2 x 2 y 4 x y n Then notice that the combination 1 h AxAy ](Un .27) I. Let the data base be immaterial ~ and the (bc) be time invariant.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.18.4 Second-Order Factored Implicit Methods CHAPTER 12.27 di er only in their evaluation of a time-dependent forcing term. SPLIT AND FACTORED FORMS Second-order accuracy in time can be maintained in a certain factored implicit methods. we can write 1 ~ I . 2 hAx U = I + 2 hAy Un + 1 hFn 2 1 hA U + 1 hF + O(h ) 1 hA U (12.26) 2 2 2 and both the factored and unfactored form of the trapezoidal method are second-order accurate in the time march.24) 2 2 Factor both sides giving 1 I . apply the trapezoidal method to Eq. Therefore. 1 hAx I .17 or 12. 1 h AxAy Un 2 4 1 hA I + 1 hA . There results 1 1 ~ I . 12. Un) is proportional to h since 4 the leading term in the expansion of (Un . Un) is proportional to h. 12.

the problem of computing the time advance in the unfactored form of the trapezoidal method given by Eq.g. a three-dimensional solver would require a memory of approximatly + +1 + +1 6 7 8 words and. One can then write the remaining terms in the two-step predictor-corrector form 1 ~ I . between and o in Eq. 6 For matrices as small as those shown there are many gaps in this \ ll". Suppose we were to solve the equations directly.24 1 ~ I . direct solvers may become useful for nding steady-state solutions of practical problems in two dimensions. in real cases involving the Euler or Navier-Stokes equations. set of coupled simultaneous equations having the matrix form shown in Eq. 1 hAyy Uny = U y (12. Consider. density and velocity eld. 2 hAxx U x = (RHS ) x ~ I .12.16 part a and b. each symbol (o x ) represents a 4 4 block matrix with entries that depend on the pressure. this probably exceeds memory availability for some time to come. On the other hand. This must be stored in computer memory to be used to nd the nal solution in the backward sweep. 12. for example. Actually current computer power is able to cope rather easily with storage requirements of this order of magnitude. 8 One billion oating-point operations per second.). If My > Mx. Again computing the right hand side poses no problem.28) 2 ( ) ( ) ( ) ( ) ( ) +1 ( ) Ne My Mz Mx 2 2 2 practical size the ll is mostly dense. 2 hAx y Un = I + 1 hAx y Un + h(bc) 2 Forming the right hand side poses no problem. 12. but very large. IMPORTANCE OF FACTORED FORMS IN 2 AND 3 DIMENSIONS 227 for realistic problems. A moderate mesh of 60 200 points would require over 11 million words to nd the solution. but nding Un requires the solution of a sparse. Furthermore.5.25. 12. consider computing a solution using the factored implicit equation 12. The forward sweep of a simple Gaussian elimination lls all of the 4 4 blocks between the main and outermost diagonal (e. but for meshes of . Accumulate the result of such a computation in the array (RHS ). If Ne represents the order of the small block matrix (4 in the 2-D Euler case). 7 The lower band is also computed but does not have to be saved unless the solution is to be repeated for another vector. However. With computing speeds of over one giga op. Here it is assumed that My < Mx. for well resolved ow elds. the approximate memory requirement is (Ne My ) (Ne My ) Mx oating point words. My and Mx would be interchanged.16 part b.

24. 9 A block tridiagonal solver is similar to a scalar solver except that small block matrix operations . Un = Un . 2 hAx . and matrix multiplications do not commute. 2-D equation. 12.21 but is second-order time accurate. is referred to as the \delta form" and we develop it next.14. +1 3 +1 + 3 2 + 3 replace the scalar ones. 12.30) I . if Eq. Then. Consider the unfactored form of the trapezoidal method given by Eq. Un one nds h 1 1 ~ i I . 2 hAy Un = I + 2 hAx + 1 hAy Un + h(bc) + O(h ) 2 From both sides subtract 1 I . Now we can factor Eq. 12. using the standard de nition of the di erence operator . it is guaranteed to converge to the correct steady-state solution of the ODE.29 converges. Inspecting the top of Eq. 1 hAy Un = h Ax y Un + (bc) + O(h ) 2 This is the delta form of a factored. 2 hAx . 1 hAy Un 2 leaving the equality unchanged. for ensuring a correct steady-state solution in a converged time-march. We arrive at the expression h 1 ~ i (12. we see that this is equivalent to solving My one-dimensional ~ problems. and ~ let the (bc) be time invariant: 1 1 1 ~ I . SPLIT AND FACTORED FORMS 9 which has the same appearance as Eq. each with Mx blocks of order Ne . The rst step would be solved using My uncoupled block tridiagonal solvers . One way that is especially useful.23 shows that the nal ~ be permuted to U step consists of solving Mx one-dimensional implicit problems each with dimension My . 2nd-order.6 The Delta Form Clearly many ways can be devised to split the matrices and generate factored forms. ( ) ( ) 12.228 CHAPTER 12. The temporary solution U x would then y and an inspection of the bottom of Eq. 2 hAy Un = h Ax y Un + (bc) + O(h ) (12. 12. 12. our original ODE.22. 2 hAx I . 2 hAx .29 and maintain O(h ) accuracy. 12.29) Notice that the right side of this equation is the product of h and a term that is identical to the right side of Eq. Thus. 12.

u ? iii.P ). u ? Only include the interior points.7 Problems 1.. hAy ] Un = h Ax y Un + (bc) + 10 and its factored form becomes h ~ i I . i. 2 . For example. PROBLEMS 229 The point at which the factoring is made may not a ect the order of time-accuracy. but it can have a profound e ect on the stability and convergence properties of a method. 12.19 is h ~ i I .e. hAy ] Un = h Ax y Un + (bc) + (12. (a) Space vector de nition i. 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. If we reorder the points with the odd points rst and then the even points.20. ( u) = 1 (u . 2u + u ) xx j The uniform grid has x = 1 and 8 interior points. 12. write the space vector.31 is the appearance of the factor 1 on the left side of the O(h2 ) method. the factored form is presented in Eq.30 and the O(h) method given by Eq. 12. 12. hAx] I . we will see in the next chapter that the convergence properties of Eq. What is the space vector for the natural ordering (monotonically increasing in index). the delta form of the unfactored Eq. ii. Assume that second order central di erencing is used. (1) (2) 12 21 x 2 j . Write down the permutation matrices. and if it is immediately factored. the unfactored form of a rst-order method derived from the implicit Euler time march is given by Eq. On the other hand.20 and Eq. so that we can simplify the boundary conditions. 12.1 j j +1 10 Notice that the only di erence between the O(h2 ) method given by Eq.7.31 are vastly di erent. 12.31) In spite of the similarities in derivation. hAx .(P .12. 12.19. 12.

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

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

SPLIT AND FACTORED FORMS .232 CHAPTER 12.

13. hybrid. When these methods are applied to practical problems. We have seen that under these conditions a semi-discrete approach can lead to circulant matrix di erence operators.4 we introduced the concept of the representative equation.Chapter 13 LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS In Section 4. and used it in Chapter 7 to study the stability. the method is probably not suitable for practical use. 233 . fractional-step. 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. However. and convergence properties of timemarching schemes. and (approximately) factored. and we discussed circulant eigensystems1 in Section 4. the results found from this analysis are neither necessary nor su cient to guarantee stability.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. if the results indicate that a method has an instability.7. 1 See also the discussion on Fourier stability analysis in Section 7.3. In this and the following section we assume circulant systems and our analysis depends critically on the fact that all circulant matrices commute and have a common set of eigenvectors. accuracy. 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.

~ (t) u (13.4: b since they must share a common The representative equation for split..2 Example Analysis of Circulant Systems 13. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS Suppose.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. eigenvector. 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.4) This is to be contrasted to the developments found later in the analysis of 2-D equations.3) that 13. we arrive at a set of ODE's that can be written d~ = A ~ + A ~ .. g1 (t) ..3 to uncouple the set and form the M set of independent equations 0 w1 = ( a + b)1 w1 . Since both matrices have the same set of eigenvectors. . (13. 0 wm = ( a + b)m wm .2. and only one2 .234 CHAPTER 13. as a result of space di erencing the PDE. 0 wM = ( a + b)M wM .2.1) apu b pu f dt where the subscript p denotes a circulant matrix. gm(t) . we can use the arguments made in Section 4.. This suggests (see Section 4.1 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. gM (t) (13.

5) dt 2 x p x2 p the convection matrix operator and the di usion matrix operator. 12.6) ( d)m = .2): ( c)m = ia sin m x m (13. h d )E . 2.6 to quantify the eigenvalues.1 0 1)~ + u u B (1 . 1 . 13. m = 0 1 M . respectively.4. (1 + h c) This leads to the principal root 1 + i ah sin m x = h sin2 m 1+4 2 2 x where we have made use of Eq. When applied to Eq.2 1)~ u (13. the explicit-implicit Euler method.8. In this section we refer to these as 1.10. where (see Section 4. so that 0 m 2 . can be represented by the eigenvalues c and d. 13. 4 2 sin2 2 x In these equations m = 2m =M .7) . the characteristic polynomial of this method is P (E ) = (1 . Using these values and the representative equation 13. the explicit-explicit Euler method. a B (. we can analyze the stability of the two forms of simple time-splitting discussed in Section 12.3. Now introduce the dimensionless numbers Cn = ah Courant number x mesh Reynolds number R = a x and we can write for the absolute value of q 2 1 + Cn sin2 m j j= C m 1 + 4 R n sin2 2 0 m 1. EXAMPLE ANALYSIS OF CIRCULANT SYSTEMS 235 If the space di erencing takes the form d~ = .13. The Explicit-Implicit Method 2 (13. Eq.4.2. Eq.2. 12.

which yields the same result as in the previous example. 13.1. The Explicit-Explicit Method An analysis similar to the one given above shows that this method produces " # q 2 m 1 .2 C = 2/ R ∆ n 0. 4 Cn sin2 m 2 j j = 1 + Cn sin 0 m 2 R 2 Again a simple numerical parametric study shows that this has two critical ranges of m . 2. one near 0. A simple numerical parametric study of Eq.8 C n C 0. 13.1: Stability regions for two simple time-split methods.236 CHAPTER 13. and the .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 _ Figure 13.4 0. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 1.8 n 1.7 shows that the critical range of m for any combination of Cn and R occurs when m is near 0 (or 2 ). From this we nd that the condition on Cn and R that make j j 1 is 2 h i C 2 1 + Cn sin2 = 1 + 4 R n sin2 2 As ! 0 this gives the stability region 2 Cn < R which is bounded by a hyperbola and shown in Fig.

as well as the stability.2 Analysis of a Second-Order Time-Split Method . (1 + 2 h c + 1 h d)E + 1 h c 2 2 2 The form of the particular polynomial depends upon whether the forcing function is carried by the AB2 method or by the trapezoidal method.2. The totaly explicit. P (E ) and Q(E ). 13.1. 1) (13. time-marching method on the equation u0 = cu + du + ae t First let us nd expressions for the two polynomials. thus 1 1 un+1 = un + 2 h c(3un .9) Q(E ) = 1 h(E 2 + E ) 2 13. Next let us analyze a more practical method that has been used in serious computational analysis of turbulent ows. un. This method applies to a ow in which there is a combination of di usion and periodic convection. EXAMPLE ANALYSIS OF CIRCULANT SYSTEMS 237 other near 180o. The convection term is treated explicitly using the second-order Adams-Bashforth method.5. 1 h d)E 2 .13.1) + 2 h d (un+1 + un) This produces 3 P (E ) = (1 . The di usion term is integrated implicitly using the trapezoidal method.2. factored method has a much smaller region of stability when R is small. which produces the constraint that 1 Cn < 2 R for R 2 The resulting stability boundary is also shown in Fig. as we should have expected. Our model equation is again the linear convection-di usion equation 13.4 which we split in the fashion of Eq. we include the forcing function in the representative equation and study the e ect of our hybrid. The characteristic polynomial follows from the application of the method to the homogeneous equation. 13. In order to evaluate the accuracy.8) and in the latter (13. In the former case it is 1 Q(E ) = 2 h(3E .

LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS Accuracy From the characteristic polynomial we see that there are two {roots and they are given by the equation 3 1 + 2h c + 1h 2 = s d The principal -root follows from the plus sign and one can show 1 1 1 = 1 + ( c + d )h + 2 ( c + d )2 h2 + 4 3 + c 2 . 13. The results are plotted in Fig.3 The Representative Equation for Space-Split Operators Consider the 2-D model3 equations @u = @ 2 u + @ 2 u @t @x2 @y2 3 (13.7.10 by a parametric study of cn and R de ned in Eq.10) Stability The stability of the method can be found from Eq. one can show er = O(h3) using either Eq.3.2. 2 d .238 CHAPTER 13.1. 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. so er = O(h3) Using P (e h) and Q(e h ) to evaluate er in Section 6. 2 h d (13. the hybrid method retains the second-order accuracy of its individual components. 13. 1 1 + 3h c + 2h 2 2 1 .1. 3 h3 d d c c 1 1 From this equation it is clear that 6 3 = 6 ( c + d)3 does not match the coe cient of h3 in 1 .8 or Eq.11) The extension of the following to 3-D is simple and straightforward. 13. 13. 1h d 2 d 2 1 . for the model equation. 13. . 12.6. 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. These results show that.10 and shown in Fig. 13.9. 2h c 1 .2. 13.

23 for the 3 4 mesh shown in Sketch 12.2: Stability regions for the second-order time-split method.13) x y dt for the space vector U .1 I ~0 I B b b where B is a banded matrix of the form B (b. THE REPRESENTATIVE EQUATION FOR SPACE-SPLIT OPERATORS239 1.4 Stable 0 4 8 ∆ R _ Figure 13.12.1 I ~0 I ~1 I 7 b b 5 4 5 4b ~.3.1 b0 b1 ).22 and 12. Let us inspect the structure of these matrices closely to see how we can diagonalize Ax + Ay ] in terms of the individual eigenvalues of the two matrices considered separately. First we write these matrices in the form 2 3 2 ~ 3 B b0 I ~1 I b A(xx) = 6 B 7 A(yx) = 6 ~.8 Cn 0. to the coupled set of ODE's: dU = A + A ]U + (bc) ~ (13. Now nd the block eigenvector .12) @t x @x y @y Reduce either of these.2 0. by means of spatial di erencing approximations. and @u + a @u + a @u = 0 (13. 12. The form of the Ax and Ay matrices for three-point central di erencing schemes are shown in Eqs.13.

14) ~ 4 5 4 B 7 3 I 5 ~ B 4 I ~ where B is the banded tridiagonal matrix B (~. Thus 2 . b b b ~ that diagonalize the B blocks. 13. so the nal result is the complete diagonalization of the matrix Ax+y : h ih ih i ~ ~ X .1 A(xx) + A(yx) X Pxy = 3 2 3 2 ~ B 1 I 7 6 7 6 ~ 7 B 6 7+6 2 I 6 7 6 7 6 (13.1 32 32 3 2 X B X 6 76 7=6 X .1 I ~0 I b b b b One now permutes the transformed system to the y-vector data-base using the permutation matrix de ned by Eq. Let B diag(B ) and ~ ~ ~ Next nd the eigenvectors X ~ diag(X ) and form the second transformation ~ X 2~ 3 2~ 3 1 6 7 7 ~2 ~ ~~ 6 ~ ~ 7 ~=6 X .1 Pyx X .1 B 76 X 4 54 54 5 4 .1 6 ~.1 ~0 ~1). LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS matrix that diagonalizes B and use it to diagonalize A(xx).1 X B X where 2 3 6 =6 6 4 3 7 5 1 2 set X 4 (x) is transparent to this transformation.23.1 A(xx+)y X Pxy X = 2 3 1I + ~ 6 7 6 7 2I + ~ 6 7 (13.15) 6 7 4 5 3I + ~ 4I + ~ diag(X ) 2 ~ b0 X .1 I ~0 I ~.240 CHAPTER 13.14 is transparent to the transformation. if we Notice that the matrix Ay 3 7 7 7 5 3 2 ~ 3 ~1 I b b0 I ~1 I b ~0 I ~1 I 7 X = 6 ~.13.1B X = 6 7 4 5 4 5 ~3 ~ This time. There results h i Pyx X . That is. 12. 12. see the bottom of Eq.1 4b I I . by the same argument as before. the rst matrix on the right side of Eq.1 I ~0 I ~1 I 7 b b 5 b b 5 4b ~.

but this choice was for convenience only.3. and where Ax and Ay satisfy the conditions in 13. by itself. Now we are ready to present the representative equation for two dimensional systems.16. In that case we set = 0 and use the simpler form du = + ]u + a (13.13.18) x y We have used 3-point central di erencing in our example. a and are (possibly complex) constants.15. The block matrices B and B can be either circulant or noncirculant in both cases we are led to the nal result: The 2{D representative equation for model linear systems is du = + ]u + ae t x y dt where x and y are any combination of eigenvalues from Ax and Ay . this fact.16) ~ where B and B are any two matrices that have linearly independent eigenvectors (this puts some constraints on the choice of di erencing schemes). does not ensure the property of \all possible combinations". and the convergence rate to. 13. To obtain the latter property the structure of ~ the matrices is important.15 contains every possible com~ bination of the individual eigenvalues of B and B . + (13. THE REPRESENTATIVE EQUATION FOR SPACE-SPLIT OPERATORS241 It is important to notice that: The diagonal matrix on the right side of Eq. Although Ax and Ay do commute. 13. and its use is not necessary to arrive at Eq. First reduce the PDE to ODE by some choice4 of space di erencing. Often we are interested in nding the value of. 4 . the steady-state solution of the representative equation.17) x y dt which has the exact solution a u(t) = ce( x+ y )t . This results in a spatially split A matrix formed from the subsets A(xx) = diag(B ) ~ A(yy) = diag(B ) (13.

Unfortunately. and when it is applied to the representative equation. we nd (1 . steady-state solution to the 2-D representative equation 13. The form of the method is given by Eq. Is unconditionally stable. " 1 #n a x+ y 2.4 Example Analysis of 2-D Model Equations 1. rst-order scheme which can then be used as a reference case for comparison with the various factored ones. h from which x . 1 Q(E ) = h giving the solution (13. Converges very rapidly to the steady-state when h is large. h .4. steady-state solution. In the following we analyze four di erent methods for nding a xed. 12. h x . In each case we examine 13. h .242 CHAPTER 13. however. The stability.19) un = c 1 .19. 3.5. The accuracy of the xed.18) steady-state solution (of the ODE) for any h. The convergence rate to reach the steady-state. Produces the exact (see Eq. . LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 13.1 The Unfactored Implicit Euler Method Consider rst this unfactored. use of this method for 2-D problems is generally impractical for reasons discussed in Section 12. x y Like its counterpart in the 1-D case. 2. h y )un+1 = un + ha P (E ) = (1 . 3. h y )E .16. 13. this method: 1.

However. h )(1 . it is not very useful since its transient solution is only rst-order accurate and. Next apply Eq. Is unconditionally stable. if one tries to take advantage of its rapid convergence rate. 12. but the converged solution is completely wrong.2 The Factored Nondelta Form of the Implicit Euler Method 13.4.3 The Factored Delta Form of the Implicit Euler Method . One nds (1 . Produces a steady state solution that depends on the choice of h.20 to the 2-D representative equation. h x)(1 .4. h ) .20) giving the solution " #n 1 un = c (1 . x y We see that this method: 1. Converges rapidly to a steady-state for large h.13. h y )(un+1 . The method requires far less storage then the unfactored form. 12. h y )E . h )(1 .31 to the 2-D representative equation. h x)(1 . + x y x y This method: 13. h y )un+1 = 1 + h2 x y un + ha and this has the solution " #n 1 + h2 x y a un = c (1 . 2. + a h x y x y. There results (1 .4. h ) . h x)(1 . EXAMPLE ANALYSIS OF 2-D MODEL EQUATIONS 243 Now apply the factored Euler method given by Eq. h y )un+1 = un + ha from which P (E ) = (1 . h x)(1 . 3. un) = h( xun + y un + a) which reduces to (1 . the converged value is meaningless. 1 Q(E ) = h (13.

26 and 12. it can be used when time accuracy is desired.4. 1 h y (un+1 . Since it is second order in time. 3. 1h x 1 . 1 h x 1 . as discussed in Section 12. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 1. the value of h on the left side of the implicit equation is literally switched from h to 1 h.30 to the representative equation and one nds 1 1 .4 The Factored Delta Form of the Trapezoidal Method . 2 5 13. since j j ! 1 as h ! 1. but convergence is not nearly as rapid as that of the unfactored form. this method demands far less storage than the unfactored form. Like the factored nondelta form. 2 h x 1 . Converges very slowly to the steady{state solution for large values of h. Is unconditionally stable. 12.244 CHAPTER 13. Is unconditionally stable. Produces the exact steady-state solution for any choice of h. 3. and the factored delta form of the implicit Euler method can be used when a converged steady-state is all that is required. 1 h y un+1 = 1 + 1 h x 1 + 1 h y un + ha 2 2 2 2 and this has the solution 2 3n 1 + 1h x 1 + 1h y 7 6 a 2 2 7 . 2h y 2 This method: 1.5 A brief inspection of eqs. Converges very slowly to the steady{state solution for large values of h. 2.5. All of these properties are identical to those found for the factored delta form of the implicit Euler method. In practical codes. Finally consider the delta form of a second-order time-accurate method. The correct steady solution is obtained. since j j ! 1 as h ! 1. Produces the exact steady{state solution for any choice of h. un) = h( xun + y un + a) 2 which reduces to 1 . 2. 12. un = c6 4 5 1 x+ y 1 .27 should be enough to convince the reader that the 's produced by those methods are identical to the produced by this method. Apply Eq.

1 h( x + y + z ) un = h ( x + y + z )un + a] 2 Now factor the left side: 1 1 . and write the solution either in the form 2 3n 1 h( + + ) + 1 h2 ( 1 h3 + + ). +a+ x y 6 z (13. 1 h z un = h ( x + y + z )un + a] (13. 13.21) x y z dt Let us analyze a 2nd-order accurate. One can derive the characteristic polynomial for Eq. delta form using this equation. EXAMPLE ANALYSIS OF THE 3-D MODEL EQUATION 13. 8 x y z .12.3 generalize to three dimensions and.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). First apply the trapezoidal method: un+1 = un + 1 h ( x + y + z )un+1 + ( x + y + z )un + 2a] 2 Rearrange terms: 1 1 1 . the model 3-D cases6 have the following representative equation (with = 0): du = + + ]u + a (13. factored. 13. nd the root. under the conditions given in 13.5 Example Analysis of the 3-D Model Equation 245 The arguments in Section 13. 2 h( x + y + z ) un+1 = 1 + 2 h( x + y + z ) un + ha Put this in delta form: 1 .23) Eqs. x y z 61 + 4 x y x z y z 8 x y z7 7 un = c6 2 4 5 1 h( + + ) + 1 h2( 1 h3 1 .11 and 13. each with an additional term. 2 x y z 4 x y + x z + y z) . 1 h y 1 .16 with an A(zz) included.5.22. . 2 h x 1 .13.

and are real numbers that can have any sign.246 CHAPTER 13. In this case. Furthermore. )2 >1 and the method is unconditionally unstable for the model convection problem. 13. 2 y 1 . 13. if the method converges. In that case since the absolute value of the product is the product of the absolute values j 2 2 2 = (1 . 13.i . In practical cases. First write the root in Eq. Now we can always nd one combination of the 's for which . and are both positive. clearly. 1h z 2 3n x y z7 7 . Remember that in our analysis we must consider every possible combination of these eigenvalues. This makes the method unconditionally stable for the 3-D di usion model when it is centrally di erenced in space. ) + ( + ) j (1 . it follows from Eq. 4 h3 2 2 1h 1h 1 .23 that. in the presence of this dissipation. we already knew this because we chose the delta form. 13. 13. 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. 2 x 1 . the 3-D form of Eq. 7 However. the method is stable for all h. it converges to the proper steady-state. if all the 's are real and negative.12 with periodic boundary conditions. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS or in the form 2 6 un = c6 4 1 1 1 + 1 h x 1 + 2 h y 1 + 1 h z . +i where . 5 x+ y+ z a (13. .24) It is interesting to notice that a Taylor series expansion of Eq. however.25) h i + 1 h3 3 + (2 y + 2 x) + 2 2 + 3 x y + 2 2 + 3 + 2 x 2 + 2 2 y + 3 + y y y y x x 4 z which veri es the second order accuracy of the factored form.23 in the form 1+ = 1.24 results in 1 = 1 + h( x + y + z ) + 2 h2( x + y + z )2 (13.7 With regards to stability. +i i . From the above analysis one would come to the conclusion that the method represented by Eq. )2 + ( .22 should not be used for the 3-D Euler equations. the instability disclosed above is too weak to cause trouble. central di erencing causes all of the 's to be imaginary with spectrums that include both positive and negative values. Now consider what happens when we apply this method to the biconvection model.

Applying implicit Euler time marching gives un+1 . and accuracy of the converged steady-state solution. convergence. nd the exact solution to the resulting scalar di erence equation and comment on the stability. 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. du = Au + f dt we can split A as follows: A = A1 + A2 + A3 + A4. (d) Repeat 1c for the explicit-implicit scheme of problem 1b. .13. PROBLEMS 13. Starting with the generic ODE. (c) The scalar representative equation is du = ( + + + )u + a 1 2 3 4 dt For the fully implicit scheme of problem 1a. What is the error term? (b) Instead of making all of the split terms implicit.6 Problems 247 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. leave two explicit: un+1 .6.

248 CHAPTER 13. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS

Appendix A USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA
A basic understanding of the fundamentals of linear algebra is crucial to our development of numerical methods and it is assumed that the reader is at least familar with this subject area. Given below is some notation and some of the important relations between matrices and vectors.

A.1 Notation
1. In the present context a vector is a vertical column or string. Thus 2 v1 3 6 7 ~ = 6 v..2 7 v 6 7 4 . 5 vm and its transpose ~ is the horizontal row vT ~ T = v1 v2 v3 : : : vm] ~ = v1 v2 v3 : : : vm ]T v v 2. A general m m matrix A can be written 2 a11 a12 a1m 3 6a a a 7 A = (aij ) = 6 21 22 . . . 2m 7 7 6 5 4 am1 am2 amm 249

250APPENDIX A. USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA 3. An alternative notation for A is h i A = ~1 ~2 : : : ~m a a a and its transpose AT is 2 ~T 3 a 6 ~1 7 6 aT 7 AT = 6 .2 7 6 7 6 .. 7 4 T5 ~m a

4. The inverse of a matrix (if it exists) is written A;1 and has the property that A;1A = AA;1 = I , where I is the identity matrix.

A.2 De nitions
1. A is symmetric if AT = A. 2. A is skew-symmetric or antisymmetric if AT = ;A. 3. A is diagonally dominant if aii Pj6=i jaij j i = 1 2 : : : m and aii > Pj6=i jaij j for at least one i. 4. A is orthogonal if aij are real and AT A = AAT = I 5. A is the complex conjugate of A. 6. P is a permutation matrix if P ~ is a simple reordering of ~ . v v 7. The trace of a matrix is Pi aii. 8. A is normal if AT A = AAT . 9. det A] is the determinant of A. 10. AH is the conjugate transpose of A, (Hermitian). 11. If b A= a d c then det A] = ad ; bc and d A;1 = det1 A] ;c ;b a

A.3. ALGEBRA

A.3 Algebra
We consider only square matrices of the same dimension. 1. A and B are equal if aij = bij for all i j = 1 2 : : : m. 2. A + (B + C ) = (C + A) + B , etc. 3. sA = (saij ) where s is a scalar. 4. In general AB 6= BA. 5. Transpose equalities: (A + B )T = AT + B T (AT )T = A (AB )T = B T AT 6. Inverse equalities (if the inverse exists): (A;1);1 = A (AB );1 = B ;1 A;1 (AT );1 = (A;1 )T

251

7. Any matrix A can be expressed as the sum of a symmetric and a skew-symmetric matrix. Thus: 1 A = 2 A + AT + 1 A ; AT 2

A.4 Eigensystems
1. The eigenvalue problem for a matrix A is de ned as A~ = ~ or A ; I ]~ = 0 x x x and the generalized eigenvalue problem, including the matrix B , as A~ = B~ or A ; B ]~ = 0 x x x 2. If a square matrix with real elements is symmetric, its eigenvalues are all real. If it is asymmetric, they are all imaginary.

252APPENDIX A. USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA 3. Gershgorin's theorem: The eigenvalues of a matrix lie in the complex plane in the union of circles having centers located by the diagonals with radii equal to the sum of the absolute values of the corresponding o -diagonal row elements. 4. In general, an m m matrix A has n~ linearly independent eigenvectors with x n~ m and n distinct eigenvalues ( i) with n n~ m. x x 5. A set of eigenvectors is said to be linearly independent if a ~ m + b ~ n 6= ~ k x x x m 6= n 6= k for any complex a and b and for all combinations of vectors in the set. 6. If A posseses m linearly independent eigenvectors then A is diagonalizable, i.e.,

X ;1AX =
where X is a matrix whose columns are the eigenvectors, h i X = ~1 ~2 : : : ~m x x x and is the diagonal matrix

0 0 m If A can be diagonalized, its eigenvectors completely span the space, and A is said to have a complete eigensystem. 7. If A has m distinct eigenvalues, then A is always diagonalizable, and with each distinct eigenvalue there is one associated eigenvector, and this eigenvector cannot be formed from a linear combination of any of the other eigenvectors. 8. In general, the eigenvalues of a matrix may not be distinct, in which case the possibility exists that it cannot be diagonalized. If the eigenvalues of a matrix are not distinct, but all of the eigenvectors are linearly independent, the matrix is said to be derogatory, but it can still be diagonalized. 9. If a matrix does not have a complete set of linearly independent eigenvectors, it cannot be diagonalized. The eigenvectors of such a matrix cannot span the space and the matrix is said to have a defective eigensystem.

2 0 03 1 6 . . . .. 7 6 7 = 6 0 . .2 . . . 7 6 ... . . 0 7 4 5

. . . The other r . 7 6. 1 7 4 5 0 0 0 i 12.. 7 6 7 i 1 6 60 0 7 . there exists one eigenvector. Note that if P is the permutation matrix 2 3 0 0 1 P = 60 1 07 4 5 1 0 0 P T = P . 0 7 Ji = 6 7 i 6. the . . .A. 13. 11. For each Jordan submatrix with an eigenvalue i of multiplicity r. A Jordan submatrix has the form 2 0 03 i 1 60 .7 J =S 6 ..1 6 0 17P = 61 07 4 5 4 5 0 0 0 1 14. . EIGENSYSTEMS 2J 0 03 1 . . 0 7 4 5 0 0 Jk 253 10.1 = P then 2 3 2 3 1 0 0 0 P . . S . Some of the Jordan subblocks may have the same eigenvalue.. Use of the transform S is known as putting A into its Jordan Canonical form. . 7 .1 AS = 6 0 J2 ...7 6 . .4. 6 . The complete set of principal vectors and eigenvectors are all linearly independent. . . Defective matrices cannot be diagonalized but they can still be put into a compact form by a similarity transform. A repeated root in a Jordan block is referred to as a defective eigenvalue. . . For example.. such that where there are k linearly independent eigenvectors and Ji is either a Jordan subblock or i.. 1 vectors associated with this eigenvalue are referred to as principal vectors.. ..

A p-norm of a matrix A is de ned as Av jjAjjp = max jjjjvjjjjp x6=0 p . USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA matrix 22 66 64 6 6 6 6 6 6 6 6 6 6 6 4 1 1 1 3 17 5 1 1 1 1 1 2 1 2 3 3 7 7 7 7 7 7 7 7 7 7 7 7 7 5 is both defective and derogatory.254APPENDIX A. A p-norm of the vector ~ is de ned as v 0M 1 X pA1=p jjvjjp = @ jvj j j =1 3.5 Vector and Matrix Norms 1. The spectral radius of a matrix A is symbolized by (A) such that (A) = j m jmax where m are the eigenvalues of the matrix A. 2. having: 9 eigenvalues 3 distinct eigenvalues 3 Jordan blocks 5 linearly independent eigenvectors 3 principal vectors with 1 1 principal vector with 2 A.

Special p-norms are jjAjj1 = maxj=1 M jjAjj1 = maxi=1 2 M PM jaij j maximum row sum j =1 where jjAjjp is referred to as the Lp norm of A. Let A and B be square matrices of the same order.A. In general (A) does not satisfy the conditions in 4. in fact. 6. is the lower bound of all the norms of A. 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. (A). 7. The spectral radius of A. (A) is a true norm. VECTOR AND MATRIX NORMS 255 4. so in general (A) is not a true norm.5. When A is normal. in this case it is the L2 norm. 8.

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

. Many of these can be derived by solving the simple linear di erence equations that arise in deriving recursion relations.4. see Section 3.1 Standard Eigensystem for Simple Tridiagonals In this work tridiagonal banded matrices are prevalent.Appendix B SOME PROPERTIES OF TRIDIAGONAL MATRICES B.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)~ m = m~ m x x (B.b. we nd (by a recursion exercise) det B (M : a b c)] = 0 if p m b + 2 ac cos M + 1 = 0 m=1 2 M From this it follows at once that the eigenvalues of B (a b c) are p m m=1 2 M (B.1 ~ m = (xj )m = a 2 sin j m x m=1 2 M (B.3) c M +1 257 . i.2) and is given by j.e. and c. a tridiagonal with constant scalar elements a. Let us consider a simple tridiagonal matrix. If we examine the conditions under which the determinant of this matrix is zero. It is useful to list some of their properties.

. j.2 Generalized Eigensystem for Simple Tridiagonals This system is de ned as follows 2 32 32 x 3 2e f x 3 b c 6a b c 76 x 7 6d e f 76 x 7 6 76 7 6 76 7 6 76 7 6 76 6 76 x 7 = 6 76 x 7 7 a b d e 6 . c m 2 1 = e. c m 2 1 sin mj m=1 2 . 2 (B. SOME PROPERTIES OF TRIDIAGONAL MATRICES These vectors are the columns of the right-hand eigenvector matrix. 7 6 .7) M (B.1 a 2 sin jm j=1 2 M X = (xjm) = c (B. f 7 6 . .5) c ( 1) The left-hand eigenvector matrix of B (a b c) can be written ...i m. d b .8) . 2 a 1 ( 1) M M (B.6) B.. . m f ) cos M + 1 = 0 m=1 2 If we de ne m m= M + 1 pm = cos m 1 2 3 1 2 3 (B.1 and c = 1.4) m=1 2 M M +1 Notice that if a = . .258 APPENDIX B. . md)(c . = 2 X j=1 2 M +1 a M +1 In this case notice that if a = . e c .1 a 2 = ei j.1 and c = 1 . 7 6 76 7 6 76 7 4 54 5 4 54 5 a b xM d e xM In this case one can show after some algebra that det B (a . f ] = 0 if q m b . the elements of which are j. c 7 6 . m e + 2 (a .

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

cD D c D D .a D aD .mDn.12) . = 1 D .c D c D D D .1 = aD The general element dmn is 4 0 2 D4 6 .cD cD .aD D D D .cD .m =DM Diagonal: n=m=1 2 M dmm = DM .n=DM B.m =DM 1 Lower triangle: m =n+1 n+2 M 1 n=1 2 M . (. SOME PROPERTIES OF TRIDIAGONAL MATRICES 2 16 6 4 D4 6 Then for M = 4 B.4 Eigensystems of Circulant Matrices B.cD .4) tridiagonal matrix Bp(M : a b c ) (B.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 .aD D D D .aD D D .4.260 APPENDIX B.1 dmn = DM .a3 D1 4 cD . DM .aD D D D .a D a D .c D .n(.cD D c D a D .1 1 n=m+1 m+2 M dmn = Dm.aD3 1 6 a2 D 6 2 6 D5 6 .4.c D . DM .a)m.aD D 3 2 2 3 1 3 1 2 2 1 4 1 1 3 0 2 1 2 2 2 2 1 2 1 3 1 1 2 2 1 3 1 2 3 4 .cD D c D a D D .cD 3 7 17 7 2 7 7 5 3 Upper triangle: m=1 2 M .1 Standard Tridiagonals Consider the circulant (see Section 3.c)n.

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 1 1 M .1 The left-hand eigenvector matrix with elements x0 is 2 j m =0 1 M .13 is a special case.1 M 1 Note that both X and X .1. and the right-hand eigenvector matrix has the form 2 m j =0 1 M . the eigenvalues are not. c) sin 2 m M M (2 ) m=0 1 2 The right-hand eigenvector that satis es Bp(a b c)~ m = m~ m is x x ~ m = (xj )m = ei j m=M x j=0 1 M .im M X mj j =0 1 M . = M X . 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. and further examination shows that the elements in these eigenvectors correspond to the elements in a complex harmonic analysis or complex discrete Fourier series.4.4.1 .1 (B. j =0 . B. are symmetric and that X . B.14) p where i .14. B.13) B.1 (B.15) 4 b b b b 5 b b b b The eigenvectors are always given by eq.12 do not depend on the elements a b c in the matrix. where X is the conjugate transpose of X . In fact. i(a . EIGENSYSTEMS OF CIRCULANT MATRICES The eigenvalues are m 261 = b + (a + c) cos 2 m .2 General Circulant Systems Notice the remarkable fact that the elements of the eigenvector matrices X and X .1 X = (xjm) = eij M m =0 1 M . B.B. = (x0 ) = 1 e. all circulant matrices of order M have the same set of linearly independent eigenvectors.15 they have the general form M. Although the eigenvectors of a circulant matrix are independent of its elements. for the tridiagonal circulant matrix given by eq. even if they are completely dense. For the element indexing shown in eq.

7 76 . B. 7 76 .... 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 .. 7 76 . 7 m6 .. 7 76 . 7 6 .. 7 6 .16) By folding the known eigenvectors of B (2M : a b a) about the center. 7 6 7 4 x2 5 1 x 1 This leads to the subsystem of order M which has the form b a 7 7 a b a 7 7 .16 are (2m . 1) m = b + 2a cos 2M + 1 ! m=1 2 M (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 x m~ m (B. For example.1. a a b a b 32x 3 1 7 6 x2 7 76 7 76 .5 Special Cases Found From Symmetries APPENDIX B..262 B. we seek the set of eigenvectors ~ m for which x 2 b 6a 6 6 6 6 6 6 6 4 a b a a . 7 54 5 a 7 6 x2 7 = x 2 3 x1 6x 7 6 27 6 .. B. 7 6 . 7 6 . .17) . 7 6 . one can show from previous results that the eigenvalues of eq. 7~ a ~ a)~ m = 7 xm = B (M : a b x 7 . SOME PROPERTIES OF TRIDIAGONAL MATRICES Consider a mesh with an even number of interior points such as that shown in Fig.

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. 2 2 . 1) x 2M + 1 j =1 2 M Imposing symmetry about the same interval but for a mesh with an odd number of points. a 7 6 7 6 7 6 4 5 a b a7 2a b By folding the known eigenvalues of B (2M . one can show from previous results that the eigenvalues of eq. combined with special conditions imposed at the boundaries. see Fig. 6 7 a a B (M : ~ b a) = 6 .B.17 are (2m . An odd{numbered mesh Figure B. B.1.6 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. SPECIAL CASES INVOLVING BOUNDARY CONDITIONS and the corresponding eigenvectors are ~ m = sin j (2m .. leads to the matrix 2 3 b a 6a b a 7 6 7 6 7 . 1) x 2M ! j=1 2 M B. 1 : a b a) about the center. B..6.1 { Symmetrical folds for special cases m=1 2 M ~ m = sin j (2m .. An even-numbered mesh Line of Symmetry x =0 x= j0 = 1 2 3 4 5 M0 j= 1 2 3 M b..

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

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

5) E = An Q + O(h2) F = BnQ + O(h2) (C. where is the internal energy per unit mass. As a nal remark. Qn) + O(h2) can be written (C. and their Jacobians. AnQn and Fn .7) (C. C. The Euler equations are homogeneous if the equation of state can be written in the form p = f ( ).4.6) since the terms En . 6. by direct use of eq. .11 and 2. " # " # (C.1 This being the case. we notice that the expansion of the ux vector in the vicinity of tn which. C. 1 " # (C.9) Note that this depends on the form of the equation of state.8) @A Q = 0 @x in spite of the fact that individually @A=@x] and Q are not equal to zero. 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. BnQn are identically zero for homogeneous vectors of degree 1. A and B . according to eq. 6. under this condition. 2.266APPENDIX C. THE HOMOGENEOUS PROPERTY OF THE EULER EQUATIONS Now it is easy to show.106. the constant term drops out of eq. E = En + An(Q . if F is homogeneous of degree 1. F = AQ and @F = A @Q + @A Q @x @x @x Therefore. see eq. are homogeneous of degree 0 (actually the latter is a direct consequence of the former). that both E and F in eqs.105 can be written in general as. Qn) + O(h2) F = Fn + Bn(Q .12 are homogeneous of degree 1.3. Notice also that.

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

Sign up to vote on this title
UsefulNot useful