Ref Af6 Fundamentals of CFD | Eigenvalues And Eigenvectors | Navier–Stokes Equations

Fundamentals of Computational Fluid Dynamics

Harvard Lomax and Thomas H. Pulliam NASA Ames Research Center David W. Zingg University of Toronto Institute for Aerospace Studies August 26, 1999

Contents
1 INTRODUCTION
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.1 Problem Speci cation and Geometry Preparation . . . . . . 1.2.2 Selection of Governing Equations and Boundary Conditions 1.2.3 Selection of Gridding Strategy and Numerical Method . . . 1.2.4 Assessment and Interpretation of Results . . . . . . . . . . . 1.3 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4 Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Conservation Laws . . . . . . . . . . . . 2.2 The Navier-Stokes and Euler Equations . 2.3 The Linear Convection Equation . . . . 2.3.1 Di erential Form . . . . . . . . . 2.3.2 Solution in Wave Space . . . . . . 2.4 The Di usion Equation . . . . . . . . . . 2.4.1 Di erential Form . . . . . . . . . 2.4.2 Solution in Wave Space . . . . . . 2.5 Linear Hyperbolic Systems . . . . . . . . 2.6 Problems . . . . . . . . . . . . . . . . . . 3.1 Meshes and Finite-Di erence Notation 3.2 Space Derivative Approximations . . . 3.3 Finite-Di erence Operators . . . . . . 3.3.1 Point Di erence Operators . . . 3.3.2 Matrix Di erence Operators . . 3.3.3 Periodic Matrices . . . . . . . . 3.3.4 Circulant Matrices . . . . . . . iii . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1

1 2 2 3 3 4 4 5

2 CONSERVATION LAWS AND THE MODEL EQUATIONS

7 8 12 12 13 14 14 15 16 17

7

3 FINITE-DIFFERENCE APPROXIMATIONS
. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

21

21 24 25 25 25 29 30

3.4 Constructing Di erencing Schemes of Any Order . . . . . 3.4.1 Taylor Tables . . . . . . . . . . . . . . . . . . . . 3.4.2 Generalization of Di erence Formulas . . . . . . . 3.4.3 Lagrange and Hermite Interpolation Polynomials 3.4.4 Practical Application of Pade Formulas . . . . . . 3.4.5 Other Higher-Order Schemes . . . . . . . . . . . . 3.5 Fourier Error Analysis . . . . . . . . . . . . . . . . . . . 3.5.1 Application to a Spatial Operator . . . . . . . . . 3.6 Di erence Operators at Boundaries . . . . . . . . . . . . 3.6.1 The Linear Convection Equation . . . . . . . . . 3.6.2 The Di usion Equation . . . . . . . . . . . . . . . 3.7 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

31 31 34 35 37 38 39 39 43 44 46 47

4 THE SEMI-DISCRETE APPROACH

4.1 Reduction of PDE's to ODE's . . . . . . . . . . . . . . . . . . . . . . 4.1.1 The Model ODE's . . . . . . . . . . . . . . . . . . . . . . . . 4.1.2 The Generic Matrix Form . . . . . . . . . . . . . . . . . . . . 4.2 Exact Solutions of Linear ODE's . . . . . . . . . . . . . . . . . . . . 4.2.1 Eigensystems of Semi-Discrete Linear Forms . . . . . . . . . . 4.2.2 Single ODE's of First- and Second-Order . . . . . . . . . . . . 4.2.3 Coupled First-Order ODE's . . . . . . . . . . . . . . . . . . . 4.2.4 General Solution of Coupled ODE's with Complete Eigensystems 4.3 Real Space and Eigenspace . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 De nition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2 Eigenvalue Spectrums for Model ODE's . . . . . . . . . . . . . 4.3.3 Eigenvectors of the Model Equations . . . . . . . . . . . . . . 4.3.4 Solutions of the Model ODE's . . . . . . . . . . . . . . . . . . 4.4 The Representative Equation . . . . . . . . . . . . . . . . . . . . . . 4.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 Basic Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Model Equations in Integral Form . . . . . . . . . . . . . . . . . . . 5.2.1 The Linear Convection Equation . . . . . . . . . . . . . . . 5.2.2 The Di usion Equation . . . . . . . . . . . . . . . . . . . . . 5.3 One-Dimensional Examples . . . . . . . . . . . . . . . . . . . . . . 5.3.1 A Second-Order Approximation to the Convection Equation 5.3.2 A Fourth-Order Approximation to the Convection Equation 5.3.3 A Second-Order Approximation to the Di usion Equation . 5.4 A Two-Dimensional Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

51
52 52 53 54 54 55 57 59 61 61 62 63 65 67 68 72 73 73 74 74 75 77 78 80

5 FINITE-VOLUME METHODS

71

5.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

83 86 87 88 89 90 91 91 92 93 93 95 95 96 97 97 98 99 100 101 102 102 103 105 106 107 110 110 111 112 115 117

6 TIME-MARCHING METHODS FOR ODE'S

6.1 Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Converting Time-Marching Methods to O E's . . . . . . . . . . . . 6.3 Solution of Linear O E's With Constant Coe cients . . . . . . . . 6.3.1 First- and Second-Order Di erence Equations . . . . . . . . 6.3.2 Special Cases of Coupled First-Order Equations . . . . . . . 6.4 Solution of the Representative O E's . . . . . . . . . . . . . . . . 6.4.1 The Operational Form and its Solution . . . . . . . . . . . . 6.4.2 Examples of Solutions to Time-Marching O E's . . . . . . . 6.5 The ; Relation . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.5.1 Establishing the Relation . . . . . . . . . . . . . . . . . . . . 6.5.2 The Principal -Root . . . . . . . . . . . . . . . . . . . . . . 6.5.3 Spurious -Roots . . . . . . . . . . . . . . . . . . . . . . . . 6.5.4 One-Root Time-Marching Methods . . . . . . . . . . . . . . 6.6 Accuracy Measures of Time-Marching Methods . . . . . . . . . . . 6.6.1 Local and Global Error Measures . . . . . . . . . . . . . . . 6.6.2 Local Accuracy of the Transient Solution (er j j er! ) . . . 6.6.3 Local Accuracy of the Particular Solution (er ) . . . . . . . 6.6.4 Time Accuracy For Nonlinear Applications . . . . . . . . . . 6.6.5 Global Accuracy . . . . . . . . . . . . . . . . . . . . . . . . 6.7 Linear Multistep Methods . . . . . . . . . . . . . . . . . . . . . . . 6.7.1 The General Formulation . . . . . . . . . . . . . . . . . . . . 6.7.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.7.3 Two-Step Linear Multistep Methods . . . . . . . . . . . . . 6.8 Predictor-Corrector Methods . . . . . . . . . . . . . . . . . . . . . . 6.9 Runge-Kutta Methods . . . . . . . . . . . . . . . . . . . . . . . . . 6.10 Implementation of Implicit Methods . . . . . . . . . . . . . . . . . . 6.10.1 Application to Systems of Equations . . . . . . . . . . . . . 6.10.2 Application to Nonlinear Equations . . . . . . . . . . . . . . 6.10.3 Local Linearization for Scalar Equations . . . . . . . . . . . 6.10.4 Local Linearization for Coupled Sets of Nonlinear Equations 6.11 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1 Dependence on the Eigensystem 7.2 Inherent Stability of ODE's . . 7.2.1 The Criterion . . . . . . 7.2.2 Complete Eigensystems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

85

7 STABILITY OF LINEAR SYSTEMS

121

122 123 123 123

. . . . . . . . . . . . . . . . . . . . . .3. . .7. . . . . . . . . . . . . . . . Numerical Stability of O E 's . . . . . . . . . . . . . . .5 Coping With Sti ness . . . . . . . . Numerical Stability Concepts in the Complex -Plane .1 -Root Traces Relative to the Unit Circle .3 Relation to Circulant Matrices . 7. . . . . . . 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Implicit Methods . . . . . . . 8. . . . A-Stable Methods . . . 149 149 149 151 151 152 153 154 154 154 155 158 158 159 160 160 161 .8 7. . . . .7.2 An Example Involving Di usion . . . .1. .3. . .2 Relation of Sti ness to Space Mesh Size . . . . . . . . . . . . . . . . . . . . . 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. . . . 8. . . . . . . 8.7 7. .6. . . .3 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1. . . . . . . . . . . . . . .1 The Basic Procedure . . . . .3 Stability Contours in the Complex h Plane.3 Defective Eigensystems . .2 Complete Eigensystems . . . . 8.5. . . . . . . . . . . . . . 7.4. . . . . . . . . . . . . . 7. . . . . . 7. . . . . . . . . . .1 Explicit Methods . . . .5 7. . . . . . . . . . . . . . . .1.2 Driving and Parasitic Eigenvalues . . . . . . . . .3 An Example Involving Periodic Convection . . . . . . . . . . . . . . . .3 A Perspective . . . . . Problems . . . . . . . . . . . . . . . . . . . .5. . . . . . . . . . . . . . . . . . . . . . . . . . . . 8. . . . 8. . . . . . . . . . . . . .9 7. .1 The Criterion . . . . 8. . .1 Relation to -Eigenvalues . . . . . . . . . . . . . . . . . . . . . . . . . 7. . . . . . . . . . . . . . . . . Time-Space Stability and Convergence of O E's . . . . . . . . . . . . .1 Imposed Constraints . . . . . .5. . . . .6. . . . . . . . 8. . . . . .2 Stability for Small t . . . . . . . . . . . . . . . . . . . . . . . Consistency . . . . . . . . . . . . . . . . . . . .6 7. . . . . . . 8. . . . . . . . . . .4 Comparing the E ciency of Explicit Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8. . 7. .6. . . . . . . .7 Problems . . . .2. . . . . .2 Some Examples . . . .3 Defective Eigensystems . . . . . . . . 8. . . .7. . . . .1 Sti ness De nition for ODE's . . . . .5. .4. . . . . . . . . . . .7. . . . . . . . .3. . . . . . . . . . . . . 8. . . . . . . . . . . . . . Fourier Stability Analysis . .6 Steady Problems . . . . . . . . . . .5. .3 Sti ness Classi cations . .4 7. . . . . . . . . 8. 8. . . 7. . . . . . . . .1 Stability for Large h. . 7. . . . . .3 Practical Considerations for Comparing Methods . . . . . . . . . . . . . . . Numerical Stability Concepts in the Complex h Plane 7. . . . . . . . .2 Unconditional Stability. . . . . . 8. .4. . . . . . . . . . .

. . . . . . . . . . . The Lax-Wendro Method . . .1 The Delta Form of an Iterative Scheme . . . . . . . . . . . . . .1 Formulation of the Model Problem . . . . . . . . . . . . . . . . . . 200 10. . . . . . . . . . . 11. . . . . . .3 The SOR System . . . . . . . . . . . . . . . . . . . . . . . .2 ODE Form of the Classical Methods . . . . . . . . . . . . . . . . . . . 9. . . . . . . . . . . . . . 9. . . 202 11. . . .2 The Gauss-Seidel System .6 Problems . . . . . . . . . . . . . . .1 The Ordinary Di erential Equation Formulation . . . . . . . .2. . . .1 The Concept . .1 Eigenvector and Eigenvalue Identi cation with Space Frequencies191 10. . . .2 11. . . . . . . . . . . . . .2 The Basic Process . . . . . . . . . . . . . . . . .5 Nonstationary Processes . . . 11. . . . . . .1. . . . . 191 10. . . . . . . . .4. .3. . . . . . . . . 9. . . . . . . .4. . . . . .5 Arti cial Dissipation . .3 11. . . . . . . . . . . . 192 10. . . .2 Flux-Di erence Splitting . . . . . . .1 Preconditioning the Basic Matrix . . . . . . . . . 9. . . . . . . . . . .3. 9. . . . . . . . . . . . . .3 Factoring Space Matrix Operators in 2{D . . . . . . . . . . . . . .3 A Two-Grid Process . .2 The Converged Solution. . . and the Error 9. . . . .4. . . . . . . . . . 192 10. . .1. . .4 Eigensystems of the Classical Methods . . 9. . . . . . . . . . . . . . . . . . . . . . . .1 11. . .2. . . . . . . . 220 217 . . . . . . . . . . 9. . . . . . . 9. . . . . . Upwind Schemes . . . . . . . . . . 191 11 NUMERICAL DISSIPATION 203 204 205 207 209 210 212 213 214 12 SPLIT AND FACTORED FORMS 12. . . 11. . . the Residual.1 The Point-Jacobi System . . . . . . . . . . . . . . . . . . . . . 9. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9. . . . . . . . . . . . . . . .4 One-Sided First-Derivative Space Di erencing The Modi ed Partial Di erential Equation . . . . . . . . . . . . . . . . . . . . . .1. . . . . . . . 9. . . . . . . . . . . . . 9. . . . 217 12. . . . . . 163 164 164 166 168 168 168 169 170 170 172 173 174 176 180 182 187 10 MULTIGRID 10. . . . . . . . .1 Flux-Vector Splitting . . . .4. . . 11. 218 12. . . . . . . . . . . .9 RELAXATION METHODS 9. . . .2 Classical Relaxation . . .2 Properties of the Iterative Method . . . . . . . .2. . . . . .1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . .2 The Model Equations .3 The Classical Methods . . . . . . . . . . . . .4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . .6 Problems . . . . . .2 Factoring Physical Representations | Time Splitting . . . . . . . . . . .1. . . . . . . 9.4. . . . . . . . . .3 The ODE Approach to Classical Relaxation . . 9. . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . .3. . . Algebra . . . . . . . . . . . . . . . . . . . .2 Example Analysis of Circulant Systems . . . . . . . .4. . . . .4 The Factored Delta Form of the Trapezoidal Method . . . . . .1 A. . . . . . . . . . . . . . . . . . . .2. . . . . . .4. . . . . 13. . .12. . .6 Special Cases Involving Boundary Conditions .2 Analysis of a Second-Order Time-Split Method . . . . . . . . . . . .1 Mesh Indexing Convention . . . .3 The Factored Delta Form of the Implicit Euler Method . . . . . . . . . . . . . . . . . . . . . . . . .3 Data Base Permutations .1 The Unfactored Implicit Euler Method . . . . . . . . .4.4 A. . . . . . . . . . . . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5 Special Cases Found From Symmetries . 220 221 221 223 226 226 228 229 13 LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 13. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. . . . . . . . . . . . . . . . . . . . . .6 Problems . . . .5 Example Analysis of the 3-D Model Equation . . .2 The Factored Nondelta Form of the Implicit Euler Method 13. . . .4. . 12. . 233 233 234 234 237 238 242 242 243 243 244 245 247 249 250 251 251 254 257 258 259 260 260 261 262 263 A USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA 249 B SOME PROPERTIES OF TRIDIAGONAL MATRICES Standard Eigensystem for Simple Tridiagonals .5 12.3.2 A. .1 Standard Tridiagonals .2. . . . B. . . . . . . . . . . . . . . . . . . . . . 13. . . . . . B. . . . . . .2 General Circulant Systems . . . . . . . . . . .4 Space Splitting and Factoring . . . . . . . . . . . . . . . . . . . . . Generalized Eigensystem for Simple Tridiagonals . . . . . . 12. . . . .3 The Representative Equation for Space-Split Operators . . . . . . . A. . . Importance of Factored Forms in 2 and 3 Dimensions The Delta Form . . . De nitions . . . B. . . . . . . . . . . . . . . . . . . 13. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 Example Analysis of 2-D Model Equations .4. . . . . . .5 B. . . . . . . . . 13. . . . . . . . . . . . . . . . . . . . Problems . . .6 12. Eigensystems . . . .3 A. . . . . .2 B. . . . . . 13. . . . . . . . . . . . . . . . .3. 257 .1 The Representative Equation for Circulant Operators . . . . . . . . . . . Second-Order Factored Implicit Methods . . . . . . . . . . . . . 13. Eigensystems of Circulant Matrices .4. . . . . 12. .4 Notation .1 B. . . . . . . . . . . . . . 13. . . 13. . . . . . The Inverse of a Simple Tridiagonal . . . . . .2 Data Bases and Space Vectors . . .3 B. 13. . . . . . .7 12. . . . . . . . . . . . .1 Stability Comparisons of Time-Split Methods . . . .4 12. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Vector and Matrix Norms . . 13. . . . . . . .

C THE HOMOGENEOUS PROPERTY OF THE EULER EQUATIONS265 .

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

This motivates studying the concepts of sti ness. and consistency. as a consequence. Independent of the speci c application under study.2.1 Problem Speci cation and Geometry Preparation The rst step involves the speci cation of the problem. 1. if their e ects are not thoroughly understood and controlled. producing answers that represent little. with identifying an error with a physical process. physical reality. of course. Thus methods are said to have a lot of \arti cial viscosity" or said to be highly dispersive. these errors are often identi ed with a particular physical phenomenon on which they have a strong e ect. as long as the error remains in some engineering sense \small". factorization. including the geometry. INTRODUCTION are the ones actually being solved by the numerical process. the following sequence of steps generally must be followed in order to obtain a satisfactory solution. if any. and probably will. the resulting simulation can still be of little practical use if ine cient or inappropriate algorithms are used. However. convergence. Since they are not precisely the same as the original equations. However. they can lead to serious di culties. they can. The geometry may result from . nor with deliberately directing an error to a speci c physical process. 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. these methods give very useful information. are often referred to as the modi ed partial di erential equations. ow conditions. the theory associated with the numerical analysis of uid mechanics was developed predominantly by scientists deeply interested in the physics of uid ow and. It is safe to say. and algorithm development in general. This motivates studying the concepts of stability. 1. these di erences are usually referred to as truncation errors. if used and interpreted properly. There is nothing wrong.2 CHAPTER 1. On the other hand. Mathematically. 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 the requirements of the simulation. for example. that most numerical methods in practical use for solving the nondissipative Euler equations create a modi ed partial di erential equation that produces some form of dissipation.2 Background The eld of computational uid dynamics has a broad range of applicability. even if the errors are kept small enough that they can be neglected (for engineering purposes). All of these concepts we hope to clarify in this book.

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

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

4. NOTATION 5 numerical approximations are applied to the model equations is a major rst step in making con dent and competent use of numerical approximations to the Euler and Navier-Stokes equations. 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 use of a vector consisting of a collection of scalars should be apparent from the context and is not identi ed by any special notation. S. it should be noted that. As a word of caution.e. 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.. 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.4 Notation The notation is generally explained as it is introduced. the variable u can denote a scalar Cartesian velocity component in the Euler and Navier-Stokes equations. Bold type is reserved for real physical vectors. proportional to) . and a vector consisting of a collection of scalars in our presentation of hyperbolic systems. Some of the abbreviations used throughout the text are listed and de ned below.1. Otherwise. a scalar quantity in the linear convection and di usion equations. For example. usually a vector A term of order (i. such as velocity. PDE ODE O E RHS P. there are many aspects of practical problems which can only be understood in the context of the complete physical systems. 1.S. however.S.

INTRODUCTION .6 CHAPTER 1.

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

3 and nd a solution for Q on that basis are referred to as nite-di erence methods. F is a set of vectors. 2. is an area A(t) bounded by a closed contour C (t). 2. The vector n is a unit vector normal to the surface pointing outward. respectively.2) dt V (t) S (t) V (t) Those methods which make various numerical approximations of the integrals in Eqs. For a Newtonian uid in one dimension. then Eq. 2. The divergence form of Eq. or tensor. 2. t1 . or cell. and P is the rate of production of Q per unit volume per unit time. leading to @Q + r:F = P (2.1 can be rewritten as d Z QdV + I n:FdS = Z PdV (2. by ! @ +j @ +k @ : (2. they can be written as @Q + @E = 0 (2. containing the ux of Q per unit area per unit time.2 is obtained by applying Gauss's theorem to the ux integral.3) @t where r: is the well-known divergence operator given.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. in Cartesian coordinates. and k are unit vectors in the x y.6) . 2. On the other hand. a partial derivative form of a conservation law can also be derived.8 CHAPTER 2. Many of the advanced codes written for CFD applications are based on the nite-volume concept. and z coordinate directions. 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 .2 The Navier-Stokes and Euler Equations The Navier-Stokes equations form a coupled system of nonlinear PDE's describing the conservation of mass. momentum and energy for a uid.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. If all variables are continuous in time.2 and nd a solution for Q on that basis are referred to as nite-volume methods.4) r: i @x @y @z and i j.1 and 2. the region of space. In two dimensions. Those methods which make various approximations of the derivatives in Eq.

. while the viscous and heat conduction terms involve second derivatives. e is the total energy per unit volume. v. In order .10) @t @x @y @E @F The matrices A = @Q and B = @Q are known as the ux Jacobians. Later on we will make use of the following form of the Euler equations as well: @Q + A @Q + B @Q = 0 (2. Details can be found in Anderson. and Pletcher 1] and Hirsch 2].9) 4 q3 5 4 v 5 4 5 4 5 + 7 q4 e u(e + p) v ( e + p) where u and v are the Cartesian velocity components. 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. p is the pressure. we are often interested in the steady-state solution of the Navier-Stokes equations. The ux vectors given above are written in terms of the primitive variables. Tannehill. and p. u is the velocity.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. The steady solution to the one-dimensional Navier-Stokes equations must satisfy @E = 0 (2. Although non-conservative forms of the equations are analytically the same as the above form. T is the temperature. these can be written as @Q + @E + @F = 0 (2.7) @x If we neglect viscosity and heat conduction. Thus the conservative form is appropriate for solving ows with features such as shock waves. Note that the convective uxes lead to rst derivatives in space. is the coe cient of viscosity. 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 Euler equations are obtained. Non-conservative forms can be obtained by expanding derivatives of products using the product rule or by introducing di erent dependent variables. for example. they can lead to quite di erent numerical solutions in terms of shock strength and shock speed. u.2. THE NAVIER-STOKES AND EULER EQUATIONS 9 where is the uid density.2. and is the thermal conductivity. such as the ideal gas law. In two-dimensional Cartesian coordinates. For such ows. with no interest in the transient portion of the solution. or at least may be treated as such. This form of the equations is called conservation-law or conservative form. such as u and p.

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. 2 q4 q 3 q1 2 q3 q1 .13) where a21 = . ) q3 q1 q2 q1 0 A= a21 q . + . q1 . From this it follows that the ux Jacobian of E can be written in terms of the conservative variables as 2 6 6 6 6 6 @Ei = 6 6 6 @qj 6 6 6 6 6 4 0 1 (3 . q2 . 1)q4 + 3. q2 q1 q1 3 a41 ! . .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 ( . where is the ratio of speci c heats. 1)q4 + 2 q1 2 q1 7 7 7 7 q3 q2 7 7 q1 7 7 7 5 3 2q q4 q2 . cp=cv . . (u2 + v2)=2] from the ideal gas law. we must rst write the ux vectors E and F in terms of the conservative variables.10 CHAPTER 2.1 0 a42 !2 a43 q2 q1 3 7 7 7 7 7 7 7 7 7 7 7 7 7 5 (2. 2q . q3 . q2. and q4 .1 q2 + q3 2 2 2 q1 2 q1 q1 q2 (2.1 q3 7 . 1 q3 2 .12) We have assumed that the pressure satis es p = ( . 3 . 1) e . CONSERVATION LAWS AND THE MODEL EQUATIONS to derive the ux Jacobian matrices. 2 q1 2 q2 q1 .1 q2 2 3 2 q1 3 7 7 7 7 7 7 7 7 2 7 .1 q2 7 2 q1 7 7 7 5 3 q3 2 q1 (2. ) q2 q1 q q3 1 0 (1 .

3 . )u (1 . which are further discussed in Section 2. 1)u(u2 + v2) . . 1 v2 . The Navier-Stokes equations include both convective and di usive uxes. Thus we also study linear hyperbolic systems of PDE's. )uv (2. !2 3 5 q4 q1 ! q2 q1 ! q4 .14) 3 7 7 7 7 7 7 7 7 7 7 7 5 0 1 0 0 0 a21 (3 .16) Derivation of the two forms of B = @F=@Q is similar.5. . Furthermore.( .uv a41 a21 = v a42 u a43 (2. This motivates the choice of our two scalar model equations associated with the physics of convection and di usion. 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. aspects of convective phenomena associated with coupled systems of equations such as the Euler equations are important in developing numerical methods and boundary conditions. The eigenvalues of the ux Jacobian matrices are purely real. 1 43 q2 a42 = q1 2 q1 ! ! q2 q3 a43 = .2. . ue a42 = e .15) u where a41 = ( . u2 2 2 a43 = (1 . 1) . The homogeneous property of the Euler equations is discussed in Appendix C. 1 (3u2 + v2 ) 2 . This is the de ning feature of hyperbolic systems of PDE's. )v ( . 1)4 ! 11 q2 q1 !3 + q3 q1 2 !2 !2 q2 q1 !3 5. THE NAVIER-STOKES AND EULER EQUATIONS a41 = ( 2 .2.

Now let us consider a situation in which the initial condition is given by u(x 0) = u0(x). At any given time. corresponding to a wave entering the domain through this \in ow" boundary. while the right-hand boundary is the in ow boundary when a is negative.17) @t @x Here u(x t) is a scalar quantity propagating with speed a.3 The Linear Convection Equation 2. a real constant which may be positive or 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. at) (2. as the convection problem. With periodic boundary conditions. This is consistent in terms of the well-posedness of a 1st-order PDE. the ow being simulated is periodic. The manner in which the boundary conditions are speci ed separates the following two phenomena for which this equation is a model: (1) In one type. the waveform travels through one boundary and reappears at the other boundary. Note that the left-hand boundary is the in ow boundary when a is positive. we pay a great deal of attention to it in the initial chapters. simply. It is easy to show by substitution that the exact solution to the linear convection equation is then u(x t) = u0(x . Hence the wave leaves the domain through the out ow boundary without distortion or re ection. the scalar quantity u is given on one boundary. the \out ow" boundary. what enters on one side of the domain must be the same as that which is leaving on the other. No boundary condition is speci ed at the opposite side. (2) In the other type. It represents most of the \usual" situations encountered in convecting systems.1 Di erential Form CHAPTER 2. Hence.12 2. without special consideration of boundaries. eventually returning to its initial position. the process continues forever without any . and the domain is in nite. 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 (2.18) The initial waveform propagates unaltered with speed jaj to the right if a is positive and to the left if a is negative. This is referred to as the biconvection problem. This type of phenomenon is referred to.3.

!. We restrict our attention to initial conditions in the form u(x 0) = f (0)ei x (2.ia t (2.3.20) where the time dependence is contained in the complex function f (t). Substituting this solution into the linear convection equation.25) . Preserving the shape of the initial condition u0(x) can be a di cult challenge for a numerical method. most numerical methods introduce some dispersion that is.20 gives the following solution u(x t) = f (0)ei (x.23) where the frequency.24) The relation between the frequency and the wavenumber is known as the dispersion relation. .3. This means that the phase speed is the same for all wavenumbers. Let the domain be given by 0 x 2 . waves with di erent wavenumbers travel at di erent speeds. In order to satisfy the periodic boundary conditions.22) Substituting f (t) into Eq.21) dt which has the solution f (t) = f (0)e. and is the wavenumber. are related by != a (2.19. must be an integer. the solution can be written as u(x t) = f (t)ei x (2. in a simulation. We now examine the biconvection problem in more detail. 2. 2. Eq. As we shall see later.at) = f (0)ei( x. The linear relation given by Eq. With such an initial condition. It is a measure of the number of wavelengths within the domain.2. we nd that f (t) satis es the following ordinary di erential equation (ODE) df = . THE LINEAR CONVECTION EQUATION 13 change in the shape of the solution. one obtains 2. the wavenumber. and the phase speed.17.19) where f (0) is a complex constant.ia f (2. a. 2.24 is characteristic of wave propagation in a nondispersive medium.2 Solution in Wave Space u(x 0) = M X m=1 fm (0)ei mx (2. For M modes. 2. An arbitrary initial waveform can be produced by summing initial conditions of the form of Eq.!t) (2.

the di usion equation has a nontrivial steady-state solution. this parabolic PDE governs the di usion of heat in one dimension. Since the wave equation is linear. 2.27. m=1 fm (0)ei m(x. 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. Boundary conditions can be periodic.1 Di erential Form Di usive uxes are associated with molecular motion in a continuum uid.28) @x2 @u = @t " 2 @u @x2 .14 CHAPTER 2. g(x) # (2. which is one that satis es the governing PDE with the partial derivative in time equal to zero.at) (2. 2.29) giving a steady state-solution which satis es @ 2 u . A simple linear model equation for a di usive process is Therefore. g(x) = 0 @x2 (2.4 The Di usion Equation 2. Neumann (speci ed @u=@x). the solution is obtained by summing solutions of the form of Eq. In the case of Eq. Other steady-state solutions are obtained if a source term g(x) is added to Eq. as follows: @u = @ 2 u (2.23. CONSERVATION LAWS AND THE MODEL EQUATIONS where the wavenumbers are often ordered such that 1 2 M .26) 2. or mixed Dirichlet/Neumann. with u representing the temperature.27) @t @x2 where is a positive real constant. Dirichlet (speci ed u). 2. For example.30) .27.4. In contrast to the linear convection equation. the steady-state solution must satisfy @2u = 0 (2. u must vary linearly with x at steady state such that the boundary conditions are satis ed.

2.4. Let the initial condition be u(x 0) = M X m=1 fm(0) sin m x + h(x) (2. 2.e.37 shows that high wavenumber components (large m) of the solution decay more rapidly than low wavenumber components.27. 2.27 gives the following ODE for fm : dfm = .31) @t @x2 @y2 where g(x y) is again a source term. consistent with the physics of di usion.35) m m dt and we nd m fm(t) = fm(0)e. Substituting this form into Eq. 2. The latter is known as the Poisson equation for nonzero g. 2 t (2. and as Laplace's equation for zero g.2 Solution in Wave Space We now consider a series solution to Eq.32 is elliptic.36) Substituting fm(t) into equation 2. . i.37) The steady-state solution (t ! 1) is simply h(x).34) satis es the initial and boundary conditions.31 is parabolic. A solution of the form M X u(x t) = fm(t) sin mx + h(x) (2. the di usion equation becomes " # @u = @ 2 u + @ 2 u .2. 2. ua)x= . THE DIFFUSION EQUATION 15 In two dimensions. Eq. 2 f (2. Eq.34.. we obtain m=1 u(x t) = M X m=1 m fm (0)e.32) @x2 @y2 While Eq. Let the domain be given by 0 x with boundary conditions u(0) = ua. 2. g(x y) = 0 (2.33) where must be an integer in order to satisfy the boundary conditions. h(x) = ua + (ub . u( ) = ub. It is clear that the steady-state solution is given by a linear function which satis es the boundary conditions. 2 t sin m x + h(x) (2.4. The corresponding steady equation is @ 2 u + @ 2 u . g(x y) (2.

1 Thus = X . CONSERVATION LAWS AND THE MODEL EQUATIONS The Euler equations.1AX (2.39) @t @x @f where f is the ux vector and A = @u is the ux Jacobian matrix.42) @t + @x With w = X . postmultiplying A by the product XX .38 by X .2.38) @t @x where u = u(x t) is a vector of length m and A is a real m m matrix.44) i @t @x 1 @X . 2. Such a system is hyperbolic if A is diagonalizable with real eigenvalues.8.5 Linear Hyperbolic Systems CHAPTER 2. Premultiplying Eq. .1u z }| { @ X . Other systems of equations governing convection and wave propagation phenomena. Eq.43) @t @x When written in this manner. form a hyperbolic system of partial di erential equations.16 2. we obtain =0 (2.1 . 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. and X is the matrix of right eigenvectors.1u. the equations have been decoupled into m scalar equations of the form @wi + @wi = 0 (2. 2. such as the Maxwell equations describing the propagation of electromagnetic waves. this equation can also be written in the form @u + @f = 0 (2. are also of hyperbolic type.1u See Appendix A for a brief review of some basic relations and de nitions from linear algebra.40) j The ux Jacobian for the Euler equations is derived in Section 2.1 are constants. and noting that X and X . The entries in the ux Jacobian are @f aij = @ui (2.1AX X . For conservation laws.41) where is a diagonal matrix containing the eigenvalues of A. this can be rewritten as @w + @w = 0 (2.1.

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

. Find the eigenvalues and eigenvectors of the matrix M derived in question 1. nd the necessary values of fm (0).2.5). 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 = ( .) 5. Given the initial condition u(x 0) = sin x de ned on 0 x 2 . (Hint: make use of the matrix S = @Q=@R. j = 0 1 2 3. CONSERVATION LAWS AND THE MODEL EQUATIONS Assume an ideal gas. are related by a similarity transform.25. write it in the form of Eq. u2=2).31. i. 4. 2. 1)(e . respectively.18 CHAPTER 2.4. Plot the initial condition on the domain 0 x . Plot the rst three basis functions used in constructing the exact solution to the di usion equation in Section 2. Find its eigenvalues and compare with those obtained in question 2. Write the 2-D di usion equation. @x + @u = 0 @y .e. (Hint: use M = 2 with 1 = 1 and 2 = . Derive the ux Jacobian matrix A = @E=@Q for the 1-D Euler equations resulting from the conservative variable formulation (Eq. and initial conditions from Eq. Find the eigenvalues and eigenvectors of A. 2. Eq. 1. Show that the two matrices M and A derived in questions 1 and 3. in the form of Eq. fm (0) = 0 for m > 3. 9. Next consider a solution with boundary conditions ua = ub = 0. in the form @U + A @U = 0 @t @x where U = @u=@x @u=@t]T . 7. Find the values of fm(0) required to reproduce the initial condition at these discrete points using M = 4 with m = m . 8. 3. that is.33 with fm(0) = 1 for 1 m 3. Write the classical wave equation @ 2 u=@t2 = c2 @ 2 u=@x2 as a rst-order system. 2. 6.1.2. 2.) Next consider the same initial condition de ned only at x = 2 j=4. 2. Plot the solution at t = 1 with = 1. 2.

e. i..1 1 Note that the variables have been nondimensionalized. (d) Are the above uxes homogeneous? (See Appendix C.u v . . Combining the two PDE's. (c) Determine the conditions (in terms of and u) under which the system is hyperbolic.6.2. PROBLEMS 19 where ! = 0 for irrotational ow. For isentropic and homenthalpic ow. the system is closed by the relation = 1 .) . 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. (b) Determine the eigenvalues of the ux Jacobians. we have @f (q) + @g(q) = 0 @x @y where v q= u f = . 1 u2 + v2 . 1 2 .v u g = . has real eigenvalues.

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 th 4 Central 4 Pade th 2 2.5 3 ¡ — .5 1.5 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 1 2 3 2 Central nd κ∆ x 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€weeëÃ

¤ » ‡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

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 0.5 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

0.2

0.4

0.6

0.8

1.2

1

2 Central

nd

4 Central

th

κ ∆x
4 Pade
1.5
th

2

2.5

3

H dB R P H ‚XIS6Wd

q 5 Y @ 5 d 5 @D D Y Ty H F D d 5 D T d H  5 d 5B d H D 2B h H d H D 2 H 4 d H h d 5r F D d R 5q T q 5 H Y T 4 H F D 2 » 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— .

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

1. stability.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 . 4. Note. It is a full discretization of the PDE.1 The Model ODE's First let us consider the model PDE's for di usion and biconvection described in Chapter 2.1 Reduction of PDE's to ODE's 4. For the time being. 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. . 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.2 is sometimes called Richardson's method of overlapping steps and Eq.2 with the time average of u at that point. 2 ( 1) ( ) 1 1 3 (4. and no semi-discrete form exists. we can approximate the space derivatives with di erence operators and express the resulting ODE's with a matrix formulation. This results in the formula ( ) ( +1) ( 1) ujn ( +1) un = ujn. 4. 4.2 and 4. THE SEMI-DISCRETE APPROACH Clearly Eq. the spatial and temporal discretizations are not separable. we shall separate the space di erence approximations from the time di erencing. namely (ujn + ujn. 4. Equation 4. A n 5 + uj . + 2h 4ujn . however.3 which we cannot comment on until we develop a framework for such investigations.3) which can be solved for u n and time advanced at the point j . this method has an intermediate semi-discrete form and can be analyzed by the methods discussed in the next few chapters.3 is referred to as the DuFort-Frankel method. there are subtle points to be made about using these methods to nd a numerical solution to the di usion equation. that the spatial and temporal discretizations are separable. In this case. There are a number of issues concerning the accuracy. Thus.52 CHAPTER 4. As we shall see later on. Another possibility is to replace the value of ujn in the right hand side of Eq. and convergence of Eqs. 2 @ j x ( 1) 2 ( ) +1 ( +1) 2 0 ( +1) + ujn. 1 to the level n + 1. This is a simple and natural formulation when the ODE's are linear. In these simple cases. )=2. 4. In this approach.

1 0 1)~ u u (4. The vector ~ (t) is usually determined f by the boundary conditions and possibly source terms.4) dt x ~ with Dirichlet boundary conditions folded into the (bc) vector.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. Although the equations are nonlinear. 4. 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~ . In such cases the equations are nonlinear. 4. In general.4.2 1)~ + (bc) u ~ (4. ~ (t) u u f dt (4. the 3-point central-di erencing approximation produces the ODE model given by d~ = . In this case. 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 . . that is. It is used for the scalar convection model when the boundary conditions are periodic.6. 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. even the Euler and Navier-Stokes equations can be expressed in the form of Eq.4 and 4. REDUCTION OF PDE'S TO ODE'S 53 Model ODE for Di usion For example. 4.1.5) dt 2 x p where the boundary condition vector is absent because the ow is periodic. Eqs.3.5 are the model ODE's for di usion and biconvection of a scalar in one dimension.1. They are linear with coe cient matrices which are independent of x and t. a B (. the elements of A depend on the solution ~ and u are usually derived by nding the Jacobian of a ux vector.

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) . The ODE's represented by Eq. In order to analyze time-marching methods. mI]~ m = 0 x (4.6 in which the coe cient matrix. m. the general solution cannot be written whereas. 4. This will lead us to a representative scalar equation for use in analyzing time-marching methods. 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. As we shall soon see. 4. 4. f . A. . These ideas are developed in the following sections. ~ m .7) The eigenvalues are the roots of the equation det A ..e. if @F=@u = A where A is independent of u). when the ODE's are linear they can be expressed in a matrix notation as Eq. 4.8) where is a diagonal matrix whose elements are the eigenvalues of A. This holds regardless of whether or ~ not the forcing function. the general solution to Eq. and together they have the property that 1 2 X . 4. As we have already pointed out.2 Exact Solutions of Linear ODE's CHAPTER 4.6 can be written. An eigenvector.54 4. have the property that A~ m = m~ m or x x A . THE SEMI-DISCRETE APPROACH In order to advance Eq. we will make use of exact solutions of coupled systems of ODE's. depends explicitly on t.1 in time.2. AX = 1 (4. the system of ODE's must be integrated using a time-marching method. 4.1 are said to be linear if F is linearly dependent on u (i. if A does not depend explicitly on t.6 can be written in terms of the eigenvalues and eigenvectors of A. is independent of u. which exist under certain conditions. and its x corresponding eigenvalue. the exact solution of Eq. If A does depend explicitly on t.

4.2. EXACT SOLUTIONS OF LINEAR ODE'S

55

Defective Systems

If an M M matrix does not have a complete set of linearly independent eigenvectors, it cannot be transformed to a diagonal matrix of scalars, and it is said to be defective. It can, however, be transformed to a diagonal set of blocks, some of which may be scalars (see Appendix A). In general, there exists some S which transforms any matrix A such that S ; AS = J where 2J 3 6 J 7 6 7 6 7 ... 6 7 J =6 7 6 7 Jm 4 5 ... and 2 3 m 1 6 7 1 ... 6 7 ... m n =6 Jm 6 7 . . . 1 7 ... 4 5 m n The matrix J is said to be in Jordan canonical form, and an eigenvalue with multiplicity n within a Jordan block is said to be a defective eigenvalue. Defective systems play a role in numerical stability analysis.
1 1 2 ( )

4.2.2 Single ODE's of First- and Second-Order
First-Order Equations
The simplest nonhomogeneous ODE of interest is given by the single, rst-order equation du = u + ae t (4.9) dt where , a, and are scalars, all of which can be complex numbers. The equation is linear because does not depend on u, and has a general solution because does not depend on t. It has a steady-state solution if the right-hand side is independent of t, i.e., if = 0, and is homogeneous if the forcing function is zero, i.e., if a = 0. Although it is quite simple, the numerical analysis of Eq. 4.9 displays many of the fundamental properties and issues involved in the construction and study of most popular time-marching methods. This theme will be developed as we proceed.

56

CHAPTER 4. THE SEMI-DISCRETE APPROACH
The exact solution of Eq. 4.9 is, for 6= ,

u(t) = c e
1 1

t+

ae t

;

where c is a constant determined by the initial conditions. In terms of the initial value of u, it can be written
t; t u(t) = u(0)e t + a e ; e

The interesting question can arise: What happens to the solution of Eq. 4.9 when = ? This is easily found by setting = + , solving, and then taking the limit as ! 0. Using this limiting device, we nd that the solution to

du = u + ae t dt
is given by

(4.10)

u(t) = u(0) + at]e t As we shall soon see, this solution is required for the analysis of defective systems.

Second-Order Equations
The homogeneous form of a second-order equation is given by

d u + a du + a u = 0 (4.11) dt dt where a and a are complex constants. Now we introduce the di erential operator D such that d D dt
2 2 1 0 1 0

and factor u(t) out of Eq. 4.11, giving
2

(D + a D + a ) u(t) = 0
1 0

The polynomial in D is referred to as a characteristic polynomial and designated P (D). Characteristic polynomials are fundamental to the analysis of both ODE's and O E's, since the roots of these polynomials determine the solutions of the equations. For ODE's, we often label these roots in order of increasing modulus as , , , m ,
1 2

4.2. EXACT SOLUTIONS OF LINEAR ODE'S
1 2

57

, M . They are found by solving the equation P ( ) = 0. In our simple example, there would be two roots, and , determined from P( ) = + a + a = 0 (4.12) and the solution to Eq. 4.11 is given by u(t) = c e 1 t + c e 2 t (4.13) where c and c are constants determined from initial conditions. The proof of this is simple and is found by substituting Eq. 4.13 into Eq. 4.11. One nds the result c e 1t( + a + a ) + c e 2t( + a + a ) which is identically zero for all c , c , and t if and only if the 's satisfy Eq. 4.12.
2 1 0 1 2 1 2 1 2 1 1 1 0 2 2 2 1 2 0 1 2

4.2.3 Coupled First-Order ODE's
A Complete System
A set of coupled, rst-order, homogeneous equations is given by u0 = a u + a u u0 = a u + a u (4.14) which can be written " # 0 a a ~ = A~ ~ = u u ]T u u u A = (aij ) = a a Consider the possibility that a solution is represented by u = c x e 1t + c x e 2t u = c x e 1t + c x e 2t (4.15) By substitution, these are indeed solutions to Eq. 4.14 if and only if " #" # " # " #" # " # a a x = x a a x = x (4.16) a a x x a a x x Notice that a higher-order equation can be reduced to a coupled set of rst-order equations by introducing a new set of dependent variables. Thus, by setting u = u0 u =u we nd Eq. 4.11 can be written u0 = ;a u ; a u u0 = u (4.17) which is a subset of Eq. 4.14.
1 2 11 21 1 1 12 22 2 2 1 2 11 21 12 22 1 2 1 1 11 21 2 2 12 22 11 12 11 21 21 22 1 11 21 11 12 22 12 22 21 2 12 22 1 2 1 1 1 1 0 2 2

58

CHAPTER 4. THE SEMI-DISCRETE APPROACH
1 2

A Derogatory System
Eq. 4.15 is still a solution to Eq. 4.14 if = = , provided two linearly independent vectors exist to satisfy Eq. 4.16 with A = . In this case 0 0 1 = 0 1 0 and 0 0 0 = 1 0 1

provide such a solution. This is the case where A has a complete set of eigenvectors and is not defective.

A Defective System
If A is defective, then it can be represented by the Jordan canonical form

"

# " #" # u0 = 0 u u0 1 u
1 2 1 2

(4.18)

whose solution is not obvious. However, in this case, one can solve the top equation rst, giving u (t) = u (0)e t . Then, substituting this result into the second equation, one nds du = u + u (0)e t dt which is identical in form to Eq. 4.10 and has the solution
1 1 2 2 1

u (t) = u (0) + u (0)t]e t
2 2 1

From this the reader should be able to verify that

u (t) = a + bt + ct e t
3 2

is a solution to

2 03 2 6 u0 7 = 6 1 4u 5 4 u0 1
1 2 3

32 3 76u 7 54u 5 u
1 2 3

if

a = u (0)
3

b = u (0)
2

1 c = 2 u (0)
1

(4.19)

The general solution to such defective systems is left as an exercise.

4.2. EXACT SOLUTIONS OF LINEAR ODE'S

59

4.2.4 General Solution of Coupled ODE's with Complete Eigensystems
Let us consider a set of coupled, nonhomogeneous, linear, rst-order ODE's with constant coe cients which might have been derived by space di erencing a set of PDE's. Represent them by the equation

d~ = A~ ; ~ (t) u u f (4.20) dt Our assumption is that the M M matrix A has a complete eigensystem and can be transformed by the left and right eigenvector matrices, X ; and X , to a diagonal matrix having diagonal elements which are the eigenvalues of A, see Section 4.2.1. Now let us multiply Eq. 4.20 from the left by X ; and insert the identity combination XX ; = I between A and ~ . There results u ~ X ; du = X ; AX X ; ~ ; X ; ~ (t) u f (4.21) dt Since A is independent of both ~ and t, the elements in X ; and X are also indepenu dent of both ~ and t, and Eq. 4.21 can be modi ed to u
1 1 1 1 1 1 1 1 1

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

~ w = X; ~ u
1

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

(4.22)

we reduce Eq. 4.20 to a new algebraic form

~ dw = w ; ~ (t) ~ g (4.23) dt It is important at this point to review the results of the previous paragraph. Notice that Eqs. 4.20 and 4.23 are expressing exactly the same equality. The only di erence between them was brought about by algebraic manipulations which regrouped the variables. However, this regrouping is crucial for the solution process because Eqs.
In the following, we exclude defective systems, not because they cannot be analyzed (the example at the conclusion of the previous section proves otherwise), but because they are only of limited interest in the general development of our theory.
1

60

CHAPTER 4. THE SEMI-DISCRETE APPROACH

4.23 are no longer coupled. They can be written line by line as a set of independent, single, rst-order equations, thus

w0 = ... 0 wm = ... 0 wM =
1

1

w ; g (t)
1 1

m wm ; gm (t) M wM

; gM (t)

(4.24)

For any given set of gm (t) each of these equations can be solved separately and then recoupled, using the inverse of the relations given in Eqs. 4.22: ~ (t) = X w(t) ~ u =
M X m=1

wm(t)~ m x

(4.25)

where ~ m is the m'th column of X , i.e., the eigenvector corresponding to m . x We next focus on the very important subset of Eq. 4.20 when neither A nor ~ has f any explicit dependence on t. In such a case, the gm in Eqs. 4.23 and 4.24 are also time invariant and the solution to any line in Eq. 4.24 is w (t) = c e m t + 1 g
m m m m

where the cm are constants that depend on the initial conditions. Transforming back to the 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

;1 X ;1~ f
1

cm e m t ~ m + A; ~ x f

|

Transient

{z

} |

Steady-state

{z

}

(4.26)

1. We saw in Sections 4. ~ u u f (4. respectively.1 De nition . we identify di erent mathematical reference frames (spaces) and view our solutions from within each.27) 1 1 1 4. For the model problems on which we are focusing most of our attention. however. The second group is referred to classically as the particular solution or the particular integral. This permits us to say a great deal about such problems and serves as the basis for this section.2. as might be expected. but entirely equivalent.and post-multiplication of A by the appropriate similarity matrices transforms A into a diagonal matrix.3 Real Space and Eigenspace Following the semi-discrete approach discussed in Section 4. REAL SPACE AND EIGENSPACE 1 61 Note that the steady-state solution is A. ~ . we reduce the partial di erential equations to a set of ordinary di erential equations represented by the generic form d~ = A~ . it is more instructive to refer to these groups as the transient and steady-state solutions. The most natural reference frame is the physical one. to a much greater extent. We say If a solution is expressed in terms of ~ . composed. in the 4. the elements of A are independent of both u and t. An alternative. of Jordan blocks or. That is. In this way. There is. 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. How these relate to the numerical solutions of more practical problems depends upon the problem and.2 that pre. it is said u to be in real space. and this can add signi cantly to our understanding. in the most general case. In our application to uid dynamics. We begin by developing the concept of \spaces".3. on the cleverness of the relator. another very useful frame.4.28) dt The dependent variable ~ represents some physical quantity or quantities which relate u to the problem of interest.1 and 4.3. In particular. we can develop some very important and fundamental concepts that underly the global properties of the numerical solutions to the model problems. we get di erent perspectives of the same solution. ~ u x x x f (4.

When the forcing function ~ is independent of t. Equations for the eigenvalues of the simple . the transient portion of the solution in eigenspace provides the weighting of each eigenvector component of the transient solution in real space. However. At this point we make two observations: 1. and 2. of scalars. We say ~ If a solution is expressed in terms of w. THE SEMI-DISCRETE APPROACH simplest nondefective case. These model ODE's are presented in Section 4. we found that Eq. ~ g f 1 1 ~ = Xw ~ u ~ = X~ f g The elements of ~ relate directly to the local physics of the problem. the transient portion of the solution in real space consists of a linear combination of contributions from each eigenvector. ~ ~ g dt which is an uncoupled set of rst-order ODE's that can be solved independently for ~ the dependent variable vector w. for simplicity. using only complete systems for our examples.1.2 and. X .62 CHAPTER 4. ~ u x f and wm(t) = cm e m t + 1 gm m=1 2 M 1 1 for real space and eigenspace. the u ~ are linear combinations of all of the elements of ~ . m 4. 4. and individually elements of w u they have no direct local physical interpretation. respectively.3.28 had the alternative form ~ dw = w .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. it is said to be in eigenspace (often referred to as wave space). the solutions in the two spaces f are represented by ~ (t) = X cm e m t ~ m + X . The relations that transfer from one space to the other are: ~ w = X. ~ u ~ = X. Following Section 4.

from Section B. The interpretation of this result plays a very important role later in our stability analysis.32) .4.3 Eigenvectors of the Model Equations The Di usion Model Next we consider the eigenvectors of the two model equations. B (M : a b c) and Bp(M : a b c).ia sin 2m m=0 1 M .1. These follow as special cases from the results given in Appendix B. are given in Appendix B. From Section B.3. for the model biconvection equation .4. m = 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.5. to help visualize the matrix structure.4 sin m=1 2 M (4.i m a m m=0 1 M .4. we nd for the model di usion equation with Dirichlet boundary conditions . Notice that the di usion eigenvalues are real and negative while those representing periodic convection are all pure imaginary.3.1 (4. we present results for a simple 4-point mesh and then we give the general case. First. REAL SPACE AND EIGENSPACE 63 tridiagonals. Recall that xj = j x = j =(M + 1). 4.31) x is the modi ed wavenumber from Section 3. 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 (4.30) = sin m x m=0 1 M . 4.1 (4. the model ODE's for di usion. Consider Eq.29) = 2(M + 1) x and. and x = 2 =M .1 m = x M 2 2 2 where = .2 + 2 cos Mm+ 1 m = x ! m .

4. 1 The Biconvection Model Next consider the model ODE's for periodic convection. In general w = X . Similarly. 4.34) . or left-hand eigenvector matrix X . Eq. . THE SEMI-DISCRETE APPROACH For the inverse.64 CHAPTER 4. Eq.33.32 represents the sine synthesis that companions the sine transform given by Eq. we can write ~ = X w as u ~ uj = wmeimxj j=0 1 (4. the right-hand eigenvectors are given by ~ m = ei j x With xj = j M . Eq.1 X m=0 (2 m=M ) j =0 1 m =0 1 M . ~ = Xw ~ u is a sine synthesis from wave space back to real space.1 M . For the model di usion equation: ~ w = X . 4.1 x = j 2 =M . For our model ODE. 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.1 M . ~ ~ u gives wm = M X j =1 uj sin mxj m=1 2 M (4. ~ is a sine transform from real space to (sine) wave u space.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( ) = 0.5. In summary.33) In the eld of harmonic analysis. The coe cient matrices for these ODE's are always circulant. 4.

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

39) We see that the solutions are identical except for the steady solution and the modi ed wavenumber in the transient term.1 X m=0 cme m t ei mxj j=0 1 M . f )j 1 j=1 2 M (4. The di erence between the modi ed wavenumber and the actual wavenumber depends on the di erencing scheme and the grid resolution.i m a (4.1 (4. m 2 t sin m xj + h(xj ) j=1 2 M (4.38) This can be compared with the exact solution to the PDE. This di erence causes the various modes (or eigenvector components) to decay at rates which di er from the exact solution. we can write the ODE solution as uj (t) = M X m=1 cme.66 CHAPTER 4.37.37) With the modi ed wavenumber de ned as 2 m x m = x sin 2 and using m = m. 2. we obtain uj (t) = where M .40) m = . We can write this ODE solution as uj (t) = M .1 (4. 4. 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.31. The Convection Equation For the biconvection equation. With conventional di erencing schemes. THE SEMI-DISCRETE APPROACH (4.41) with the modi ed wavenumber de ned in Eq.i m at ei mxj j=0 1 M .1 X m=0 cme.42) . low wavenumber modes are accurately represented. Eq. m 2 t sin m xj + (A. while high wavenumber modes (if they have signi cant amplitudes) can have large errors.

4. not particular problems. 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. which are related by algebraic manipulation.44 is not quite satisfactory.42 shows that the imaginary portion of the modi ed wavenumber produces nonphysical decay or growth in the numerical solution. topics which are covered in Chapters 6 and 7.44) m m m dt The goal in our analysis is to study typical behavior of general situations. The importance of this concept resides in its message that we can analyze timemarching methods by applying them to a single. discussed in Chapter 11.2. X uj (t) = fm (0)e. However. this leads to an error in the speed with which various m modes are convected. THE REPRESENTATIVE EQUATION 1 67 and compare it to the exact solution of the PDE. This is helpful both in analyzing the accuracy of time-marching methods and in studying their stability. For such a purpose Eq.43) Once again the di erence appears through the modi ed wavenumber contained in .5. g (t) (4. Since the error in the phase speed depends on the wavenumber. uncoupled equation and our conclusions will apply in general. 4. while the actual phase speed is independent of the wavenumber.20 and 4. since is real. . The form of Eq. Our next objective is to nd a \typical" single ODE to analyze. Eq. m=0 4.i mat ei m xj j=0 1 M . and then returned as a coupled set to real space. A typical member of the family is dwm = w . the modi ed wavenumber is complex.4.3. we pointed out that Eqs. The role of m is clear it stands for some representative eigenvalue in the original A matrix. We found the uncoupled solution to a set of ODE's in Section 4. 2.1 (4.4.4 The Representative Equation In Section 4.26. the result is erroneous numerical dispersion. solved there in their uncoupled form.23 express identical results but in terms of di erent groupings of the dependent variables. In the case of non-centered di erencing. As discussed in Section 3. evaluated at the nodes of the grid: M. 4.

From this we can extract the k'th term and replace ik with . Find the entries in the boundary-condition vector. The exact solution of the representative ODE is (for 6= ): t (4. 4. This leads to w(t) = ce t + The Representative ODE dw = w + ae t dt (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. write the semi-discrete form obtained with periodic boundary conditions on a 5-point grid (M = 5). Using this operator to approximate the spatial derivative in the linear convection equation.45) for which Eq. 2. u(1) = 1: @u = @ u . 4. Consider the nite-di erence operator derived in question 1 of Chapter 3. Consider the application of the operator given in Eq.52 to the 1-D di usion equation with Dirichlet boundary conditions. Write the resulting semi-discrete ODE form. 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. In such evaluations the parameters and must be allowed to take the worst possible combination of values that might occur in the ODE eigensystem. 3. one can express any one of the forcing terms gm(t) as a nite Fourier series. 3.46) w(t) = ce t + ae .68 CHAPTER 4. in principle. we note that.5 Problems 1.g(t) = ak eikt X ak eikt k ik . 6x @t @x 2 2 .44 has the exact solution: k which can be used to evaluate all manner of time-marching methods.

4. Compare with the exact solution of the PDE. can be found from Appendix B.37. Note that the domain is 0 x 2 and x = 2 =10. 5.) Calculate the corresponding modi ed wavenumbers for the second-order centered operator from Eq. Repeat for the initial condition u(x 0) = sin 2x.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. Calculate and plot the corresponding ODE solutions. plot the exact solution of the di usion equation with = 1 at t = 1 with m = 1 and m = 3. compute and plot the ODE solution at t = 2 for the initial condition u(x 0) = sin x.4. . Consider the matrix A = . Using these. For initial conditions u(x 0) = sin(mx) and boundary conditions u(0 t) = u( t) = 0.4. 1 . as well as the matrices X and X . PROBLEMS 69 4. 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. The grid nodes are given by xj = j x j = 0 1 : : : 9.Bp(10 . The eigenvalues of the above matrix A. Consider a grid with 10 interior points spanning the domain 0 x .5. (Plot the solution at the grid nodes only.

THE SEMI-DISCRETE APPROACH .70 CHAPTER 4.

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

they will generally produce di erent approximations to the ux at the boundary between two control volumes. In our examples. Another approach known as ux-di erence splitting is described in Chapter 11. that is. . 5.2) at the next time step. Examining Eq. Similarly. Q can be represented within the cell by some piecewise approximation which produces the correct value of Q. This is a form of interpolation often referred to as reconstruction.4) for a control volume which does not vary with time. each cell will have a di erent piecewise approximation to Q. First. In order to evaluate the uxes.1 is that of a control volume whose shape is dependent on the nature of the grid.1. 5.1 can be written as 1 Z QdV V V (5. The ux is required at the boundary of the control volume.72 5. we have updated values of the cell-averaged quantities Q. Let us consider these approximations in more detail. we will consider only control volumes which do not vary with time. we note that the average value of Q in a cell with volume V is Q and Eq. at the control-volume boundary. which are a function of Q. When these are used to calculate F(Q). This ux must then be integrated to nd the net ux through the boundary. we see that several approximations must be made. 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. 5. The basic elements of a nite-volume method are thus the following: 1 Time-marching methods will be discussed in the next chapter.3) I Z d V dt Q + S n:FdS = V PdV (5. which is a closed surface in three dimensions and a closed contour in two dimensions. A nondissipative scheme analogous to centered di erencing is obtained by taking the average of these two uxes. Next a time-marching method1 can be applied to nd the value of Z V QdV (5.1 Basic Concepts CHAPTER 5. Thus the volume V in Eq. Thus after applying a timemarching method. As we shall see in our examples. the source term P must be integrated over the control volume. the ux will be discontinuous.

4. Advance the solution in time to obtain new values of Q. In order to include di usive uxes. u(x y t) with speed a along a straight line making an angle with respect to the x-axis. MODEL EQUATIONS IN INTEGRAL FORM 73 1. the following relation between rQ and Q is sometimes used: Z I rQdV = nQdS (5. two distinct values of the ux will generally be obtained at any point on the boundary between two control volumes. V S Z where the unit vector n points outward from the surface or contour.2 Model Equations in Integral Form 5. Integrate the ux to nd the net ux through the control-volume boundary using some sort of quadrature. These ideas should be clari ed by the examples in the remainder of this chapter. Evaluate F(Q) at the boundary. 3.7) This PDE governs a simple plane wave convecting the scalar quantity. Using this approximation.2. nd Q at the control-volume boundary. in two dimensions. construct an approximation to Q(x y z) in each control volume.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. A rQdA = C nQdl I (5.6) 5. The order of accuracy of the method is dependent on each of the approximations. This issue is discussed in Section 11.5) or. The one-dimensional form is recovered with = 0. Since there is a distinct approximation to Q(x y z) in each control volume.5. Given the value of Q for each control volume. 2.2. Apply some strategy for resolving the discontinuity in the ux at the controlvolume boundary to produce a single value of F(Q) at any point on the boundary. . 4.

as in the model equations.9) (5. 2.74 CHAPTER 5. We will use the following notation: xj. FINITE-VOLUME METHODS For unit speed a.17) .3. we nd (5. with 5.3.2 gives the following integral form (5.ru ! @u + j @u = .8) (5.14) (5.1. Eq. d Z udA = I n: i @u + j @u ds dt A @x @y C (5. The nodes of the grid are located at xj = j x as usual.12) (5. 2. We consider an equispaced grid with spacing x.11) where A is the area of the cell which is bounded by the closed contour C . 2. with Q = u F = iu cos + ju sin P = 0 d Z udA + I n:(iu cos + ju sin )ds = 0 dt A C (5.10) Since Q is a scalar. as shown in Fig. i @x @y P = 0 Using these. F is simply a vector.1=2 = xj .2. 5. Control volume j extends from xj .13) (5.3 One-Dimensional Examples We restrict our attention to a scalar dependent variable u and a scalar ux f . Eq. x=2 xj+1=2 = xj + x=2 (5. x=2 to xj + x=2.15) ! to be the integral form of the two-dimensional di usion equation.16) 5. Substituting these expressions into a twodimensional form of Eq. the two-dimensional linear convection equation is obtained from the general divergence form.2 The Di usion Equation Q = u F = . The integral form of the two-dimensional di usion equation with no source term and unit di usion coe cient is obtained from the general divergence form.

1=2 = x j+1=2 Pdx (5.3.18) (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) (5.1=2 = 0 . the integral form of the linear convection equation. we can expand u(x) in Eq. Hence the cell-average value and the value at the center of the cell di er by a term of second order. becomes u (5.22) where uj is the value at the center of the cell.1=2 or (5.1: Control volume in one dimension. uj = uj + O( x2) 5.20) j j +1=2 dt j . 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 5.23) x ddtj + fj+1=2 . Eq.21) 4 j 1920 @x j and the integral form becomes xj.1=2 Now with = x . 5. x=2 j @x j 2 @x2 j 6 @x3 j ! ! x2 @ 2 u + x4 @ 4 u + O( x6 ) = uj + 24 @x2 (5.5. 5. xj .1 A Second-Order Approximation to the Convection Equation In one dimension.19) Zx d ( xu ) + f .11. With these de nitions. fj.3.19 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 . fj.

and cell j . the discontinuity in the ux at the cell boundary is resolved by taking the average of the uxes on either side of the boundary.28) We have now accomplished the rst step from the list in Section 5.29 and 5. we obtain u 1 u x ddtj + 2 (uj + uj+1) .26) +1 Similarly. gives fjR =2 = uj+1 (5.76 CHAPTER 5. 5. 1 (uj.1=2.1.1 0 1)~ u u (5. as shown in Fig.1 + uj ) (5.1 we have de ned the uxes at the cell boundaries in terms of the cell-average data.1) = 0 (5.27) . The cell to the right of xj +1=2.29) +1 2 and 1 f^j. We choose a piecewise constant approximation to u(x) in each cell such that u(x) = uj xj. 5.31) 2 2 With periodic boundary conditions. cell j is the cell to the right of xj.1=2 + fjR 1=2) = 1 (uj. Substituting Eqs. 1 is the cell to the left of xj.24) Evaluating this at j + 1=2 gives fjL+1=2 = f (uL+1=2) = uL+1=2 = uj (5.1=2 x xj+1=2 (5. In this example. uj.30) .1 (5.25) j j where the L indicates that this approximation to fj+1=2 is obtained from the approximation to u(x) in the cell to the left of xj+1=2 .30 into the integral form. 5. giving fjL. 1 B (. 2 where f^ denotes a numerical ux which is an approximation to the exact ux.32) dt 2 x p .1 + uj ) = x ddtj + 1 (uj+1 .1=2 = 2 (fjL.1=2 = uj. FINITE-VOLUME METHODS with f = u. Thus 1 f^j+1=2 = 2 (fjL+1=2 + fjR =2) = 1 (uj + uj+1) (5. giving fjR 1=2 = uj (5. this point operator produces the following semidiscrete form: d~ = .23. which is cell j + 1.1=2 . Eq.

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

8 0 8 .uj+2 + 5uj+1 + 2uj ) j 6 1 uL.2) 12 (5.2) = 0 (5. the following semi-discrete form is obtained: d~ = . 8uj.3.2) j (5.1 . 1 B (1 .43) dt 12 x p This is a system of ODE's governing the evolution of the cell-average data. 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 (5.38) (5. uj.1 .uj+2 + 8uj+1 .1) and 1 f^j.1=2 = 2 f (uL.42) This is a fourth-order approximation to the integral form of the equation. With periodic boundary conditions.uj+1 + 7uj + 7uj. 5. as can be veri ed using Taylor series expansions (see question 1 at the end of this chapter).1=2 = 6 (2uj + 5uj.1 + uj.23.1) j 6 uR+1=2 = 1 (. while the second approach is more suited to extension to higher order accuracy. we describe two approaches to deriving a nite-volume approximation to the di usion equation.uj+1 + 5uj + 2uj. FINITE-VOLUME METHODS uR. uj.41) Substituting these expressions into the integral form.uj+2 + 7uj+1 + 7uj . Recalling that f = u.1=2)] j j = 1 (. The rst approach is simpler to extend to multidimensions.78 CHAPTER 5.1=2) + f (uR.3.1)~ u u (5.3 A Second-Order Approximation to the Di usion Equation .39) using the notation de ned in Section 5. 5. Eq.1=2 = 1 (.37) (5.40) = 12 (. In this section. gives u 1 x ddtj + 12 (.1. uj.

From Eq. we obtain x duj = 1 (u . Eq. d~ = 1 B (1 . with Dirichlet boundary conditions. we use a piecewise quadratic approximation as in Section 5. and we see that the properties of the matrix B (1 .44.49) (5. becomes with f = .48) (5.33 we have dt x @u = @u = 2a + b @x @ (5.47) j +1=2 @x j+1=2 x j+1 j Similarly. Hence.46) @u dx = u(b) .45) In one dimension. Eq. 5. 5. 1 (u .2 1) are relevant to the study of nite-volume methods as well as nite-di erence methods. = . 2u + u ) or.22. the integral form of the di usion equation.51) . u ) (5. 5. uj. Also.@u=@x.2. Eq.ru = .1 j (5. For our second approach. 5. to second-order. ONE-DIMENSIONAL EXAMPLES u x ddtj + fj+1=2 . u(a) @x x xj @x x j+1 j From Eq. fj. we can write ! @u f^ = .3. 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.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 .44) (5.1=2 = . Substituting these into the integral form.50) j +1 This provides a semi-discrete nite-volume approximation to the di usion equation. 1x (uj .1) dt x j .5.2 1)~ + bc u u ~ 2 f^j.16.1=2 = 0 Zb 79 (5. 5.3. u ) (5.

uj. 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 . fjR 1=2 = fjL. 1 (u . dt x2 j .80 CHAPTER 5.1 j j +1 dt x2 p 5. and ` is the length of the sides of the hexagons. 2u + u ) (5. This produces duj = 1 (u .2 1)~ u u (5.1=2 = .4 A Two-Dimensional Example The above one-dimensional examples of nite-volume approximations obscure some of the practical aspects of such methods.57) . The control volumes are regular hexagons with area A. 1x (uj . 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. The following relations hold between `. 5. 1 ` = p 3 p A = 3 2 3 `2 ` = 2 (5.35. 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.55) which is written entirely in terms of cell-average data. is the length of the sides of the triangles.@u=@x.54) which is identical to Eq. With f = .49.53) Notice that there is no discontinuity in the ux at the cell boundary. As in Section 5.3. u ) x j+1 j (5.2. Thus our nal example is a nite-volume approximation to the two-dimensional linear convection equation on a grid consisting of regular triangles. and A.1) . as shown in Figure 5. 5. FINITE-VOLUME METHODS j +1=2 j +1=2 with a and b given in Eq.52) (5. .1.

. 3 j)=2 p Table 5.2: Triangular grid.1.i p (. A list of the normals for the mesh orientation shown is given in Table 5. A TWO-DIMENSIONAL EXAMPLE c b 2 3 81 p d 4 1 l ∆ 0 5 a e f Figure 5. 3 j)=2 pi (i + p j)=2 3 (. i and j are unit normals along x and y. where we have ignored the source term. the unit normals can be taken outside the integral and the ux balance is given by 5 d Z Q dA + X n dt A =0 Z Fdl = 0 where indexes a side of the hexagon. as shown in Figure 5.5.1. The contour in the line integral is composed of the sides of the hexagon. Side 0 1 2 3 4 5 Outward Normal n (i . Since these sides are all straight.i .i + 3 j)=2 . 5.2.4. respectively.2. see Fig. Outward normals.

2. the integrated time rate of change of the integral of u over the area of the hexagon is known exactly.60. 5. 5. 3 sin )=2 p Table 5.60.59) Then.1=2 u(z)dz (5.58) where is a length measured from the middle of a side . Z and the piecewise-constant approximation u = up over the entire hexagon.`=2 u ( )d (5. cos .`=2 Z `=2 Z 1=2 are no numerical approximations in Eq. in terms of u and the hexagon area A. cos + 3 sin )=2 . cos p (. we have for side n Z Fdl = n (i cos + j sin ) Z `=2 . if the integrals in the equation are evaluated exactly. That is. Making the change of variable z = =`. 3 sin )=2 cos p (cos + p sin )=2 3 (.11. see Eq. one has the expression u( )d = ` . FINITE-VOLUME METHODS For Eq.1=2 =0 The values of n (i cos + j sin ) are given by the expressions in Table 5. Introducing the cell average. the approximation to the ux integral becomes trivial. 5. Side 0 1 2 3 4 5 n (i cos + j sin ) (cos .1=2 (5. There . 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 . we have " # 5 d Z u dA + X n (i cos + j sin ) ` Z 1=2 u(z)dz = 0 (5.62) 2 1 A u dA = Aup (5. Weights of ux integrals. the two-dimensional linear convection equation.82 CHAPTER 5.61) .60) dt A .2.

along with the expressions in Table 5. cos + 3 sin )(uc . Use the following linear approximation: u( ) = a + b .5 Problems 1. ue) p +(. that this is a second-order approximation to the integral form of the two-dimensional linear convection equation.67) Z Z u(z)dz = up + ud 2 4 u(z)dz = up + ue 2 5 Z Z Substituting these into Eq. 5. u ) + (cos + p3 sin )(u .5. Use Taylor series to verify that Eq.69) The reader can verify. uf )] = 0 (5.66) (5. 5. we have for the other ve edges: Z 83 u(z)dz = up + ub 2 2 u(z)dz = up + uc 2 3 (5.60.2. uf )] = 0 (5.68) or dup + 1 (2 cos )(u . 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. u ) a d b e dt 3 p +(. cos + 3 sin )(uc .5. 5. using Taylor series expansions.63) (5.65) (5. we obtain u(z)dz = up + uf 2 0 p u ` A ddtp + 2 (2 cos )(ua . 5.23. ud) + (cos + 3 sin )(ub . PROBLEMS Similarly.42 is a fourth-order approximation to Eq. 2.64) (5.

1 and use the average ux at the cell interface.3.84 where b = uj and CHAPTER 5. . 4.3. 3. FINITE-VOLUME METHODS u a = uj+12 . Repeat question 3 for a grid consisting of equilateral triangles.x j. Using the rst approach given in Section 5. derive a nite-volume approximation to the spatial terms in the two-dimensional di usion equation on a square grid.

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

and. rather than w.1 un. we need only consider the scalar case Although we use u to represent the dependent variable. we will develop exact solutions for the model O E's arising from the application of timemarching methods to the model ODE's.3 to some given accuracy. or the (n) superscript. u.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. but we postpone such considerations to the next chapter.1 du = u0 = F (u t) dt (6. etc. We then face a further task concerning the numerical stability of the resulting methods.1u0n. 6. where accuracy can be measured either in a local or a global sense. TIME-MARCHING METHODS FOR ODE'S Application of a spatial discretization to a PDE produces a coupled system of ODE's. They are represented conceptually by (6. For these we use the notation ~ ~ u0n+ = Fn+ = F (~n+ tn + h) ~ u The choice of u0 or F to express the derivative in a scheme is arbitrary. They are both commonly used in the literature on ODE's. the reader should recall the arguments made in Chapter 4 to justify the study of a scalar ODE. The methods we study are to be applied to linear or nonlinear ODE's.3) . 4. Our rst task is to nd numerical approximations that can be used to carry out the time integration of Eq.1 0 un . we introduced the convention that the n subscript.1 Notation Using the semi-discrete approach. for the purpose of this chapter. However. always points to a discrete time value. Combining this notation with Eq. 6. but the methods themselves are formed by linear combinations of the dependent variable and its derivative at various time intervals. 6.86 CHAPTER 6. we developed exact solutions to our model PDE's and ODE's. In Chapter 2. and h represents the time interval t. In earlier chapters. In this chapter we will present some basic theory of linear O E's which closely parallels that for linear ODE's.1. we reduce our PDE to a set of coupled ODE's represented in general by Eq. using this theory. Application of a time-marching method to an ODE produces an ordinary di erence equation (O E ).4) un+1 = f 1 u0n+1 0u0n .

by applying a time-marching method. we can analytically convert MacCormack's method. for example. the rst and third of these are examples of explicit methods. For an implicit method.1 respectively.6) According to the conditions presented under Eq. the new predicted solution is also a function of the time derivative at the new time level.6. The value of this equation arises from the fact that. u0n+1.3 1 Here we give only MacCormack's time-marching method. will be presented in Section 11. These methods are simple recipes for the time advance of a function in terms of its value and the value of its derivative.1 for a method using two previous time levels. u. The second is implicit and referred to as the implicit (or backward) Euler method.4.2 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. at given time intervals. 6. these methods can be constructed to give a local Taylor series accuracy of any order. which is a fully-discrete method. An explicit method is one in which the new predicted solution is only a function of known data. The methods are said to be explicit if 1 = 0 and implicit otherwise.1. u0n. u0n. that is. CONVERTING TIME-MARCHING METHODS TO O E'S 87 With an appropriate choice of the 0s and 0s. and therefore the time advance is simple.5) (6.8) dt written here in terms of the dependent variable. As we shall see. un and un.2. 6. The material presented in Chapter 4 develops a basis for evaluating such methods by introducing the concept of the representative equation un+1 = un + hu0n ~ 1 ~ ~ un+1 = 2 un + un+1 + hu0n+1] (6.7) du = u0 = u + ae t (6. for systems of ODE's and nonlinear problems. We refer to them as the explicit Euler method and the MacCormack predictor-corrector method. The method commonly referred to as . implicit methods require more complicated strategies to solve for un+1 than explicit methods.

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. There results un+1 = un + h( un + ae hn) or un+1 . 6.9. to Eq.88 CHAPTER 6. 6. 1 un = 1 ahe h(n+1) 2 2 which is a coupled set of linear O E's with constant coe cients.6. 6.2.11 is identical to Eq. h)un+1 .10) As a nal example. Note that the rst line in Eq.7 is simply the explicit Euler method. un = he h ae hn (6. 6. Eq. the predictor-corrector sequence.11 is obtained by noting that u0n+1 = F (~n+1 tn + h) ~ u = un+1 + ae h(n+1) ~ (6. TIME-MARCHING METHODS FOR ODE'S such a linear ODE into a linear O E . gives un+1 . . 2 (1 + h)~n+1 + un+1 . Apply the simple explicit Euler scheme. We next consider examples of this conversion process and then go into the general theory on solving O E's. the theory of ODE's. The latter are subject to a whole body of analysis that is similar in many respects to. The second line in Eq. (1 + h)un = ahe hn ~ 1 u (6.7. 6. since the predictor step in Eq. 6. applying the implicit Euler method.11) .8.8. 6.9) Eq. This is demonstrated by repeating some of the developments in Section 4. 6.3 Solution of Linear O E's With Constant Coe cients The techniques for solving linear di erence equations with constant coe cients is as well developed as that for ODE's and the theory follows a remarkably parallel path. 6. with constant coe cients. and just as powerful as. to Eq. Eq. (1 + h)un = hae hn (6. but for di erence rather than di erential equations.9 is a linear O E . 6. Eq.5. As another example. 6. we nd un+1 = un + h un+1 + ae h(n+1) or (1 . expressed in terms of the dependent variable un and the independent variable n.

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. 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. one can readily show that the solution of the defective case.13 is where c1 is a constant determined by the initial conditions. a. 4. n un = c1 n + bab . The exact solution of Eq.and Second-Order Di erence Equations First-Order Equations The simplest nonhomogeneous O E of interest is given by the single rst-order equation un+1 = un + abn (6.13) where . n. and b are. 6. un = u0 n + a b b . complex parameters.6. SOLUTION OF LINEAR O E'S WITH CONSTANT COEFFICIENTS 89 6. and since the equations are linear and have constant coe cients. (b = ).14) d Instead of the di erential operator D dt used for ODE's. In terms of the initial value of u it can be written Just as in the development of Eq. thus b bn = bn+ = E bn .3.10.1 n This can all be easily veri ed by substitution.1 First. in general. un+1 = un + a n is h i un = u0 + an . is not a function of either n or u.3. The independent variable is now n rather than t.

. TIME-MARCHING METHODS FOR ODE'S where can be any fraction or irrational number.14.3. ) (c c12 ) = 0 21 22 . etc. E I ]~ = 0 un c21 (c22 . Thus Eq.16 is a solution to Eq. rst-order. 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. E ) u2 which must be zero for all u1 and u2.14 for all c1 c2 and n should be veri ed by substitution. In the analysis of O E's. 6.90 CHAPTER 6. They are found by solving the equation P ( ) = 0. The -roots are found from P ( ) = det (c11c . 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.17 can be written u1 (n) = C . The fact that Eq. 6. The roles of D and E are the same insofar as once they have been introduced to the basic equations the value of u(t) or un can be factored out. E ) C = c11 c12 c21 c22 The operational form of Eq. 6. 6.15) which must be zero for all un. and refer to them as the -roots. we label these roots 1 .17) which can also be written ~ n+1 = C~ n u u (c11 . Eq. In the simple example given above. this time having the form P (E ) = det C . 6.2 Special Cases of Coupled First-Order Equations A Complete System Coupled. there are just two roots and in terms of them the solution can be written un = c1( 1 )n + c2( 2 )n (6. Again we are led to a characteristic polynomial. . E I ].16) where c1 and c2 depend upon the initial conditions. its roots determine the solution to the O E.14 can now be re-expressed in an operational notion as (E 2 + a1E + a0)un = 0 (6. 6. 2 .15 is known as the operational form of Eq. 6.

19) (6. produced by applying a time-marching method to the representative equation. 6. 6.21) " .17 is if x 2 ~ n = X ck ( k )n~ k u x k=1 where ck are constants determined by the initial conditions.4. rst-order ordinary di erence equations. h)E .4 Solution of the Representative O E's 6. one can show that the solution to A Defective System 2 3 2 un+1 7 6 6 un+1 5 = 4 1 4^ un+1 1 32 3 7 6 un 7 ^ 5 4 un 5 un is un = u0 n h i un = u0 + u0n . (1 + h)]un = h ae hn (1 .1 n ^ ^ " # .(1 + h) # " u # = h " 1 # ae hn ~ 1E 1 (1 + h)E E .1 + u n(n . are given by Eqs. E . following the logic of Section 4.2 for defective ODE's.2. linear. ~ are its eigenvectors. the solution of Eq.2 un = u0 + u0n ^ 0 2 n (6.4. 1]un = h E ae hn (6.18) 6. these equations can be written E . Using the displacement operator. 1) . SOLUTION OF THE REPRESENTATIVE O E'S 91 Obviously the k are the eigenvalues of C and. For example.20) (6. The solution of O E's with defective eigensystems follows closely the logic in Section 4.9 to 6.6.1 The Operational Form and its Solution Examples of the nonhomogeneous.2. 1 u n .2 2 2 E .11.

21.2 Examples of Solutions to Time-Marching O E's As examples of the use of Eqs.23) P (e ) k=1 where k are the K roots of the characteristic polynomial.21 there are multiple (two in this case) solutions. The general solution of Eq. 6. h Q(E ) = h (6.4. respectively. 6.21. 4. Eq. For the explicit Euler method. 6. 6.23. Eq.23: un = c1(1 + h)n + ae hn For the implicit Euler method. 6. h)E . In such a case K X Q(1) un = ck ( k )n + a P (1) k=1 6. Keep in mind that for methods such as in Eq. 6. Notice also.25) . We can express in terms of Eq.1. we derive the solutions of Eqs. When determinants are involved in the construction of P (E ) and Q(E ). one for un and un and we are usually only interested in the nal solution un. ~ the important subset of this solution which occurs when = 0. 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. 6. 6. P ( ) = 0. 1 Q(E ) = hE e h. as would be the case for Eq.22 and 6. h h (6.22) which is produced by applying time-marching methods to the representative ODE. we have P (E ) = E .45.24) and the solution of its representative O E follows immediately from Eq. we have P (E ) = (1 . 6. representing a time invariant particular solution.22 all manner of standard time-marching methods having multiple time steps and various types of intermediate predictor-corrector families.20.22 can be expressed as K h X un = ck ( k )n + ae hn Q(e h) (6. or a steady state. 1 .19.19 to 6. The terms P (E ) and Q(E ) are polynomials in E referred to as the characteristic polynomial and the particular polynomial. the ratio Q(E )=P (E ) can be found by Kramer's rule.92 CHAPTER 6.

1 (1 E h)E 1 hE = 1 hE (E + 1 + h) 2 2 + 2 The -root is found from 1 P( ) = . Relation We have now been introduced to two basic kinds of roots. 1 . 1 (1 + h)E E . The complete solution can therefore be written 1 h 1 2h2 n + ae hn 2 h e + 1 + h un = c1 1 + h + 2 (6.27.5.21.5 The . h . 4. one solves for the nal family un (one can also nd a solution for the intermediate family u). 6.26) e h . Eq. 1 2h2 P (E ) = det . h + ae hn (1 .1 Establishing the Relation 6. 1 . This relation is rst demonstrated by developing it for the explicit Euler method. and there ~ results # " E . THE . 1 2h2 2 he h 1 n un = c1 1 . h . There is a fundamental relation between the two which can be used to identify many of the essential properties of a time-march method. RELATION so 93 In the case of the coupled predictor-corrector equations. 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.27) x . 1 . 1 6. The former are the eigenvalues of the A matrix in the ODE's found by space di erencing the original PDE.6. Remembering that t = nh. the -roots and the -roots. 2 2 h2 = 0 which has only one nontrivial root ( = 0 is simply a shift in the reference index). h)e h .5. 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. 1 2 2 2 h Q(E ) = det .(1 + h) = E E . 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. h .

6.29) Applying Eq. which is de ned by un+1 = un. 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 have the characteristic polynomial P (E ) = E 2 . q 1 + 2 h2 .8 to Eq.29.94 CHAPTER 6. we use the leapfrog method for the time advance.3. instead of the Euler method. TIME-MARCHING METHODS FOR ODE'S where for the present we are not interested in the form of the particular solution (P:S:). so that for every the must satisfy the relation m .28) where the cm and the ~ m in the two equations are identical and m = (1 + m h).1 + 2hu0n (6. So if we use the Euler method for the time advance of the ODE's. 2 hE .5. 1. 6. 1 = 0 2 (6.28.31) (6. one -root. Now the explicit Euler method produces for each -root. For one of these we nd m q h+ 1+ m 1 = 1 + mh + 2 = 2 h2 . . x Comparing Eq.30) Now we notice that each produces two -roots. we see a correspondence between m and e m h. m 2 Based on Section 4. 1 4 h4 + m 2 h2 m (6. 6. will be discussed in Section 6. which is given by = 1 + h. 2 mh m . Suppose.32) mh 8 m This is an approximation to e m h with an error O( 3h3).4. 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. The other root.27 and Eq. 3 The error is O( 2 h2 ). .

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. However.5. twice as many time steps are required to reach the time T . 6. this is reduced by a factor of eight (considering the leading term only). We refer to the root that has the above property as the principal -root. consistent with a second-order approximation.1 + 2hu0n (6. always produces one -root for every -root that satis es the relation (6. The following example makes this clear. One of these we identi ed as the principal root which always . and designate it ( m )1 .30 that the . Since the error per time step is O(h3). Consider a solution obtained at a given time T using a second-order time-marching method with a time step h. The above property can be stated regardless of the details of the timemarching method.5. Thus the principal root is an approximation to e h up to O hk .3 Spurious -Roots We saw from Eq. un.6. relation for the leapfrog method produces two -roots for each . 6.6. which is the leapfrog method: un+1 = un.33) 1 2h2 + + 1 k hk + O hk+1 =1+ h+ 2 k! where k is the order of the time-marching method.1) (6. THE . Now consider the solution obtained using the same method with a time step h=2.5. 4. Therefore the error at the end of the simulation is reduced by a factor of four. knowing only that its leading error is O hk+1 . while the second-order time-marching method which results from this approximation.34.2 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 6.35) has a leading truncation error O(h3). Note that a second-order approximation to a derivative written in the form ( t u)n = 1 (un+1 .34) 2h has a leading truncation error which is O(h2). 6.

make a method inferior. However. For example.5.28 must be modi ed.36) Spurious roots arise if a method uses data from time level n . TIME-MARCHING METHODS FOR ODE'S has the property given in 6.36 must be initialized by some starting procedure.4.. This results because methods which produce spurious roots require data from time level n .4 One-Root Time-Marching Methods There are a number of time-marching methods that produce only one -root for each -root. they play virtually no role in accuracy analysis.e.96 CHAPTER 6.1 and thus cannot be started using u only ~ n. the leapfrog method requires ~ n. To express this fact we use the notation = ( h). if one starts the method properly) the spurious coe cients are all initialized with very small magnitudes. It should be mentioned that methods with spurious roots are not self starting. The initial vector ~ 0 u does not provide enough data to initialize all of the coe cients. u Presumably (i. In general. If a time-marching method produces spurious -roots. we have un = c11 ( 1 )n ~ 1 + + cm1( m )n ~ m + + cM 1 ( M )n ~ M + P:S: 1x 1x 1x n~ + n~ + +c12 ( 1)2 x1 + cm2 ( m )2 xm + cM 2 ( M )n ~ M 2x n~ + n~ + +c13 ( 1)3 x1 + cm3 ( m )3 xm + cM 3 ( M )n ~ M 3x +etc. if there are more spurious roots (6. No matter whether a -root is principal or spurious. 1 or earlier. it is always some algebraic function of the product h. In fact. relation produced by a time-marching scheme can result in multiple -roots all of which. except for the principal one. Then the presence of spurious roots does not contaminate the answer. 6. many very accurate methods in practical use for integrating some forms of ODE's have spurious roots. They are also called one-step methods. Thus while spurious roots must be considered in stability analysis. Following again the message of Section 4. are spurious. all of the coe cients (cm )2 in Eq. 6. after some nite time the amplitude of the error associated with the spurious roots is even smaller then when it was initialized. All spurious roots are designated ( m )k where k = 2 3 . For example. We refer to them as one-root methods. That is.. the . if there is one spurious root to a method. . 1 or earlier to advance the solution from time level n to n + 1. 6.33. and presumably the magnitudes of the spurious roots themselves are all less than one (see Chapter 7). Such roots originate entirely from the numerical approximation of the time-marching method and have nothing to do with the ODE being solved. generation of spurious roots does not. in itself. the solution for the O E in the form shown in Eq. The other is referred to as a spurious -root and designated ( m )2 .

For example.6. However.37) respectively. vectors. respectively. It is quite common to judge a time-marching method on the basis of results found from a Taylor table. the so-called -method. it is of no use in: . see Section 3. relation is ) = 1 +1(1 . A popular method having this property. is given by the formula h i un+1 = un + h (1 . 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. One is the error made in each time step.6. Three one-root methods were analyzed in Section 6.1 Local and Global Error Measures There are two broad categories of errors that can be used to derive and evaluate timemarching methods. It is usually used as the basis for establishing the order of a method.1~ u x x x f (6. It is useful for comparing methods.6. They have the signi cant advantage of being self-starting which carries with it the very useful property that the time-step interval can be changed at will throughout the marching process. Its . This is a global error. 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. so that all the constants. as we shall see in Chapter 8. Notice that when t and n = 0. This is a local error such as that found from a Taylor table analysis. these equations are identical.4.6 Accuracy Measures of Time-Marching Methods 6. )u0n + u0n+1 6. These are ~ (t) = c1 e 1 h n ~ 1 + + cm e m h n ~ m + + cM e M h n ~ M + A. ACCURACY MEASURES OF TIME-MARCHING METHODS 97 1 The -method represents the explicit Euler ( = 0).2. the trapezoidal ( = 2 ). and matrices are identical except the ~ and the terms inside u the parentheses on the right hand sides. and the implicit Euler methods ( = 1). a Taylor series analysis is a very limited tool for nding the more subtle properties of a numerical time-marching method.4. The only error made by introducing the time marching is the error that makes in approximating e h. h h .1~ u x x x f ~ n = c1( 1 )n ~ 1 + + cm( m )n ~ m + + cM ( M )n ~ M + A.

98 CHAPTER 6. Our error measures are based on the di erence between the exact solution to the representative ODE. evaluating numerical stability and separating the errors in phase and amplitude.39) j j 6. The order of the method is the last power of h matched exactly. and to study them we make use of the material developed in the previous sections of this chapter. The latter three of these are of concern to us here. including only the contribution from the principal root.2 Local Accuracy of the Transient Solution (er Transient error er! ) The particular choice of an error measure. . This is similar to the error found from a Taylor table. which can be written as h un = c1 ( 1)n + ae hn Q(e h ) P (e ) (6. Therefore.relation we saw that all time-marching methods produce a principal -root for every -root that exists in a set of linear ODE's. given by t u(t) = ce t + ae . 1 The leading error term can be found by expanding in a Taylor series and choosing the rst nonvanishing term. 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. However. is to some extent arbitrary. analyzing the particular solution of predictor-corrector combinations. either local or global. TIME-MARCHING METHODS FOR ODE'S nding spurious roots.38) and the solution to the representative O E's. In the discussion of the . nding the global error. We designate this by er and make the following de nition er e h.6. a necessary condition for the choice should be that the measure can be used consistently for all methods.

Such can indeed be the case when we study the equations governing periodic convection which produces harmonic motion. 1( h). !h Cn x 6. we have h = .1 ( 1 )i=( 1 )r )] and the local error in phase can be de ned as (6.5 can be extended to the combination of a spatial discretization and a time-marching method applied to the linear convection equation. where is the modi ed wavenumber of the spatial discretization.6. j 1 j = 1 . Introducing the Courant number. ( 1)2 + ( 1 )2 i r er! !h .ia . A positive value of erp corresponds to phase lag (the numerical phase speed is too small). Thus one can obtain values of the principal root over the range 0 x for a given value of the Courant number.3 Local Accuracy of the Particular Solution (er ) The numerical error in the particular solution is found by comparing the particular solution of the ODE with that for the O E.a .6. ACCURACY MEASURES OF TIME-MARCHING METHODS 99 Amplitude and Phase Error Suppose a eigenvalue is imaginary. while a negative value corresponds to phase lead (the numerical phase speed is too large). For such cases it is more meaningful to express the error in terms of amplitude and phase. Let = i! where ! is a real number representing a frequency.41) Amplitude and phase errors are important measures of the suitability of time-marching methods for convection and wave propagation phenomena. Cn = ah= x. We have found these to be given by 1 P:S:(ODE) = ae t ( . The approach to error analysis described in Section 3. Then the numerical method must produce a principal -root that is complex and expressible in the form (6. is found using = . as follows . tan. that is q era = 1 .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. The above expression for er! can be normalized to give the error in the phase speed.42) er = er! = 1 + tan ( 1)i =( 1)r )] p where ! = .iCn x. ) .6.1 (6.

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

For example.49) ( ) Er! N !h . However. 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 . N tan 1i 1r !# (6. we are more concerned with the global error in amplitude and phase.5 Global Accuracy In contrast to the local error measures which have just been discussed. This subject is covered in Chapter 8 after our introduction to stability in Chapter 7.1 ( ) =( ) ] = !T . Then the required number of time steps. These are useful when we come to the evaluation of time-marching methods for speci c purposes.6. Suppose we wish to compute some time-accurate phenomenon over a xed interval of time using a constant time step. ACCURACY MEASURES OF TIME-MARCHING METHODS 101 The reader should be aware that this is not su cient.6. the analysis based on a linear nonhomogeneous ODE produces the appropriate conditions for the majority of time-marching methods used in CFD. we can also de ne global error measures. is N. 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. Let T be the xed time of the event and h be the chosen step size. These are given by Era = 1 . ( 1 ( h))N (6.48) Global error in amplitude and phase If the event is periodic. 6. tan. We refer to such a computation as an \event".1 ( 1) i 1r .6.50) . and q " ( 1 )2 + ( 1)2 i r N (6.

and they are said to be K -step because K time-levels of data are required to marching the solution one time-step. In the subsequent sections.102 CHAPTER 6. and the result is expressed in operational form. we introduce classes of methods along with their associated error analysis. 6.1.K k un+k = h du = u0 = F(u t) dt 1 X k=1.51) where the notation for F is de ned in Section 6. TIME-MARCHING METHODS FOR ODE'S Global error in the particular solution Finally. Eq. When Eq. one nds 0 1 @ X k=1. They are explicit if 1 = 0 and implicit otherwise. ) Qeh . 6. developmental or design issues. We shall not spend much time on development or design of these methods.1 P eh 6. It can be measured by Er ( . 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. the global error in the particular solution follows naturally by comparing the solutions to the ODE and the O E.K k Fn+k (6. 6. we have developed the framework of error analysis for time advance methods and have randomly introduced a few methods without addressing motivational.1 The General Formulation When applied to the nonlinear ODE all linear multistep methods can be expressed in the general form 1 X k=1. h.4. since most of them have historic origins from a wide variety of disciplines and applications.K 1 0 1 X k k E Aun = h@ k=1.51 is applied to the representative equation.52) .7 Linear Multistep Methods In the previous sections.8. The methods are said to be linear because the 's and 's are independent of u and n.K 1 k hn k E A( un + ae ) (6.7.

un h u0n h2 u00 n h3 u000 n h4 u0000 n . 11 6 . 12 .h 1 u0n+1 .53) 1=1 0 = .4. Two well-known families of them.2 Examples There are many special explicit and implicit forms of linear multistep methods. 6.h .un . that approximates e h through the order of the method. 1) k Since both sides of Eq.1 6 .1 1 .1 u0n.1 .2 that a time-marching method when applied to the representative equation must provide a -root.1 .1 k=0 The Adams-Bashforth family has the same 's with the additional constraint that 1 = 0. 6.1 + .6.1 2 . 6. that is ! (1 + h) as h ! 0.2) (6. The three-step Adams-Moulton method can be written in the following form un+1 = un + h( 1u0n+1 + 0 u0n + .51 with k = .7.1.1 .5. labeled 1.2)2 . 6. The condition referred to as consistency simply means that ! 1 as h ! 0. can be designed using the Taylor table approach of Section 3. k k k k 6. these methods are often \normalized" by requiring X k=1 Under this condition co = 1 in Eq.h 0 u0n .2 . a method should at least be rst-order accurate. and it is certainly a necessary condition for the accuracy of any time marching method.2)1 .(.1 .2 (6.h . The Adams-Moulton family is obtained from Eq. to be of any value in time accuracy.2 .7. One can show that these conditions are met by any method represented by Eq. Taylor table for the Adams-Moulton three-step linear multistep method.51 if X X X k = 0 and k = (K + k .2 .1 u0n. .2 1 6 Table 6.2 u0n.(. 6. referred to as Adams-Bashforth (explicit) and AdamsMoulton (implicit).1 1 .51 can be multiplied by an arbitrary constant.(.2 2 .2)3 . .44.0 1 1 .2)0 .54 can be generated as 1 1 1 un+1 1 1 2 6 24 . We can also agree that.54) A Taylor table for Eq.1 . LINEAR MULTISTEP METHODS 103 We recall from Section 6.(.2u0n.

un.1 = .55 and 6.4 .5=24 .4 With 1 = 0 one obtains 2 32 3 2 3 1 1 1 0 7 617 6 0 . is given below together with identifying names that are sometimes associated with them.58) 0 = 23=12 . u0n. = un + hu0n+1 Implicit Euler 1 hhu0n + u0n+1i = un + 2 Trapezoidal (AM2) 1 h4un .2 .1i = un + 12 AM3 . some of which are very common in CFD applications.1 + 2hhun Leapfrog 1 h 3u0n .2 = 5=12 This is the third-order Adams-Bashforth method.55) 4 5 4 .1 + 2hu0n+1i =3 2nd-order Backward h h5u0n+1 + 8u0n . u0n.4 7 6 0 7 6 1 7 6 7 6 7 6 3 0 3 12 7 6 76 (6. 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.56) 1 = 9=24 0 = 19=24 .32 1 . 16u0n.2 = 1=24 which produces a method which is fourth-order accurate.57 up to the order of the method.1 + 5u0n.2 to solve for the 's.1 5 = 4 1 5 (6.1i = un + 2 h 0 i AB2 h = un + 12 23un .2 AB3 Implicit Methods un+1 un+1 un+1 un+1 4 Recall from Section 6.4 7 6 .16=12 . A list of simple methods. One can verify that the Adams type schemes given below satisfy Eqs.2 giving (6.5.2 . 6.104 CHAPTER 6. resulting in (6. Explicit Methods un+1 un+1 un+1 un+1 = un + hu0n Euler 0 = un.2 that a kth-order time-marching method has a leading truncation error term which is O(hk+1 ).1 7 = 6 1 7 5 4 5 4 0 .57) 4 54 0 3 12 1 .1 = . 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=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. Methods with = . Notice that the Adams methods have = 0. 6. ' 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 . which corresponds to 0 = 0 in Eq.59 is given in Table 6. The most general two-step linear multistep method (i.59.59) Clearly the methods are explicit if = 0 and implicit otherwise. the methods are 2nd-order accurate) if 1 '= .5 6 1 Finally a unique fourth-order method is found by setting = .1 = 0 in Eq. that is at least rst-order accurate.1=3 0 1=12 .and two-step methods.1] + h u0n+1 + (1 . Some linear one.51.1=2 0 0 0 0 .1=2 . 6..e. + ')u0n .1 (6. A list of methods contained in Eq. +2 The class of all 3rd-order methods is determined by imposing the additional constraint =2 . 'u0n.1=4 . =3 = 6 .2. see Eq.1=6 .7.1=2.51.1=2 . 6. which corresponds to . LINEAR MULTISTEP METHODS 105 6.1=6 0 . un.2. This limits the interest in multistep methods to about two time levels. can be written as h i (1 + )un+1 = (1 + 2 )un . 6.1=3 .51). are known as Milne methods.1=2 .5=6 .' = .e.6.1=2 0 .2=9 0 1=2 . 6.7. K=2 in Eq. . One can show after a little algebra that both er and er are reduced to 0(h3) (i..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.

One concludes.4. It is easy to show. 1 u0n ~ (6. 6. One can analyze this sequence by applying it to the representative equation and using the operational techniques outlined in Section 6. where measures of e ciency are discussed in the next two chapters. that h P (E ) = E E . Q(E ) = E h E + + h] 2 h2 i (6.106 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. The situation for er requires some algebra.61 that + =1 =1 2 which provides two equations for three unknowns. fractional-step.26.8 Predictor-Corrector Methods CHAPTER 6. ( + ) h . 6. to the following observations.63) 5 Such as alternating direction. Their use in solving ODE's is relatively easy to illustrate and understand.60) where the parameters and are arbitrary parameters to be determined. 1 . TIME-MARCHING METHODS FOR ODE'S There are a wide variety of predictor-corrector schemes created and used for a variety of purposes. each of which is referred to as a family in the solution process. un+ = un + hu0n ~ 1 un+1 = un + 2 h 1 u0n+ + 2 . For this to hold for er . Predictor-corrector methods constructed to time-march linear or nonlinear ODE's are composed of sequences of linear multistep methods. by following the discussion in Section 6. and hybrid methods.62) Considering only local accuracy.6. A simple one-predictor. that the predictor-corrector sequence is a second-order accurate method for any . one-corrector example is given by un+ = un + h 0n ~ hu i un+1 = un + h u0n+ + u0n ~ (6. Their use is motivated by ease of application and increased e ciency. . and usually the nal family has a higher Taylor-series order of accuracy than the intermediate ones.44 that the same conditions also make it O(h3). therefore. There may be many families in the sequence. 6. following the example leading to Eq. it is obvious from Eq. but it is not di cult to show using Eq. one is led. For the method to be second-order accurate both er and er must be O(h3).61) (6.

RUNGE-KUTTA METHODS 107 A classical predictor-corrector sequence is formed by following an Adams-Bashforth predictor of any order with an Adams-Moulton corrector having an order one higher.1 ~ i hh u un+1 = un + 12 5~0n+1 + 8u0n . 6. 6.63 with = 1=2 is 1 un+1=2 = un + 2 hu0n ~ un+1 = un + hu0n+1=2 ~ and. The Gazdag method.1 ~ h i un+1 = un + 1 h u0n + u0n+1 ~ ~ (6. 6 Although implicit and multi-step Runge-Kutta methods exist.6 that produce just one -root for each -root such that ( h) corresponds explicit Runge-Kutta methods here. 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. u0n. . obtained from Eq. The order of the combination is then equal to the order of the corrector. speci c. The Adams-BashforthMoulton sequence for k = 3 is i 1 h un+1 = un + 2 h 3u0n . u0n. If the order of the corrector is (k).66) (6.9. u0n. MacCormack's method. (6. nally.68) 6.67) (6.1 (6.63 with = 1.6. second-order accurate methods are given below. which we discuss in Chapter 8.64) Some simple. we will consider only single-step. presented earlier in this chapter. referred to as Runge-Kutta methods.9 Runge-Kutta Methods There is a special subset of predictor-corrector methods. we refer to these as ABM(k) methods. is 1 hu ~ i un+1 = un + 2 h 3~0n .65) 2 The Burstein method.

108 CHAPTER 6.72) un+1 = un + 1hun 2 bn+ 3 ~n+ 1 + 4 hu0n+ 2 Appearing in Eqs. as we pointed out in Section 6.4.6. 6.69) = 1 + h + 1 2h2 + + 1 k hk 2 k! It is not particularly di cult to build this property into a method. To ensure k'th order accuracy.70) and this is much more di cult. the method must further satisfy the constraint that er = O(hk+1) (6.70. a scheme entirely equivalent to 6.71) However. 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) . we prefer to present it using predictor-corrector notation. We present it below in some detail. 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. . 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 is b un+ = un + hu0n b un+ 1 = un + 1 hu0n + 1hu0n+ ~ b un+ 2 = un + 2 hu0n + 2hu0n+ + 2hu0n+ 1 ~ 0 + hu0 + hu0 (6.71 and 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. the choices for the time samplings.73) 2 = 2+ 2+ 2 . the principal (and only) -root is given by (6. u(tn) = 1 k1 + 2k2 + 3k3 + 4k4 (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. Thus for a Runge-Kutta method of order k (up to 4th order).69 and 6. First of all. and 2 . but. 1. are not arbitrary. Thus.

A more general derivation based on a nonlinear ODE can be found in several books. two parameters can be set arbitrarily. 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. Thus if the rst 3 equations in 6. but the equations follow directly from nding P (E ) and Q(E ) and then satisfying the conditions in Eqs. RUNGE-KUTTA METHODS 109 The algebra involved in nding algebraic equations for the remaining 10 parameters is not trivial. 6.75 cannot be derived using the methodology presented here.73 to eliminate the 's we nd from Eq.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 ~ (6.76) 6 conditions for Runge-Kutta methods of up to third order.75 which must be satis ed by the 10 unknowns. 7 The present approach based on a linear inhomogeneous equation provides all of the necessary . 6. 6.74 and 6. To satisfy the condition that er = O(k5).74) These four relations guarantee that the ve terms in exactly match the rst 5 terms in the expansion of e h .74 and the rst equation in 6. Using Eq. As discussed in Section 6.69 and 6. the resulting method would be third-order accurate.70.4.75 are all satis ed.7 There are eight equations in 6. but the most popular one is due to Runge. Several choices for the parameters have been proposed. the fourth condition in Eq.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. Since the equations are overdetermined. 6. which is based on a linear nonhomogenous representative ODE.9.75) 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.6.

6. 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. TIME-MARCHING METHODS FOR ODE'S 9 > > = > > Notice that this represents the simple sequence of conventional linear multistep methods referred to. and third-order methods can be derived from Eqs.75.10 Implementation of Implicit Methods We have presented a wide variety of time-marching methods and shown how to derive their . 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. we obtained (1 .10. In any event. For example.66 and 6. storage requirements reduce the usefulness of Runge-Kutta methods of order higher than four for CFD applications. h)un+1 .1 Application to Systems of Equations Consider rst the numerical solution of our representative ODE u0 = u + ae t (6.74 and the rst equation in 6.110 CHAPTER 6.77) using the implicit Euler method. It is clear that for orders one through four. These are covered in this Section. respectively.72 by setting 4 = 0 and satisfying only Eqs. a fth-order method requires six evaluations to advance the solution one step. 6. but they require more derivative evaluations than their order. 6. un = he h ae hn (6. Following the steps outlined in Section 6.78) .67 are second-order Runge-Kutta methods. In the next chapter. we will see that these methods can have widely di erent properties with respect to stability. 6. We shall discuss the consequences of this in Chapter 8. Higher-order RungeKutta methods can be developed. relations. This leads to various tradeo s which must be considered in selecting a method for a speci c application. 6.2. RK methods of order k require k evaluations of the derivative function to advance the solution one time step.

hf (t + h) u u or (6. but in multidimensions the bandwidth can be very large. h (un + he h ae hn ) 111 (6. hA)~ n+1 .84) un+1 = un + hF (un+1 tn+1) This is a nonlinear di erence equation.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. 6.78 is (6. Now u u the equivalent to Eq.82) ~ ~ n+1 = (I . 6.10.83) (6.80) ~ where ~ and f are vectors and we still assume that A is not a function of ~ or t. The primary area of application of implicit methods is in the solution of sti ODE's. and hence its solution is inexpensive.10.79) This calculation does not seem particularly onerous in comparison with the application of an explicit method to this ODE. the cost per time step of an implicit method is larger than that of an explicit method.1 0 1)=2 x). for biconvection. IMPLEMENTATION OF IMPLICIT METHODS Solving for un+1 gives 1 un+1 = 1 . but rather we solve Eq. consider the nonlinear ODE du + 1 u2 = 0 (6. ~ n = . hf (t + h)] u u The inverse is not actually performed. In general.85) dt 2 . Now let us apply the implicit Euler method to our generic system of equations given by ~ 0 = A~ .81 as a linear system of equations. requiring only an additional division. 6. as we shall see in Chapter 8. hA). A = . For our one-dimensional examples. the system of equations which must be solved is tridiagonal (e.6. As an example.g..aBp (. f (t) u u ~ ~ (I .1 ~ n .81) (6.

6. since the \initial guess" is simply the solution at the previous time step. 6. There are several di erent approaches one can take to solving this nonlinear di erence equation. and Eq. t ) F (u t) = F (un tn) + @u n n @t n n ! ! 1 @ 2 F (u .87 can be written ! ! @F (u . Designate the reference value by tn and the corresponding value of the dependent variable by un. such as Newton's method (see below).89) . u )2 + @ 2 F (u .83. we expand F (u t) about some reference point in time. un)k are O(hk ). which gives 1 un+1 + h 2 u2 +1 = un (6. which implies that a linearization approach may be quite successful. u )(t . can be used.86) n which requires a nontrivial method to solve for un+1. Such an approach is described in the next Section.10.87) On the other hand. the \initial guess" for this nonlinear problem can be quite close to the solution. t )2 @ 2 u + u(t) = un + (t . In practice.112 CHAPTER 6. 6. tn)k and (u . tn) @t n @t2 n n 2 (6. A Taylor series expansion about these reference quantities gives ! ! @F (u . the expansion of u(t) in terms of the independent variable t is ! ! @u + 1 (t . tn)2 + @t2 n (6.3 Local Linearization for Scalar Equations General Development Let us start the process of local linearization by considering Eq. both (t . An iterative method. u ) + @F (t . t ) + O(h2) F (u t) = Fn + @u n n @t n n (6.88) If t is within h of tn . TIME-MARCHING METHODS FOR ODE'S solved using implicit Euler time marching. In order to implement the linearization. t ) + 2 @u2 n n n @u@t n !n 1 2 + 2 @ F (t . u ) + @F (t .

other second-order implicit timemarching methods that can be used. of course. 6. With this we obtain the locally (in the neighborhood of tn) time-linear representation of Eq.89 to evaluate Fn+1 = F (un+1 tn+1). Using Eq. @F u + @F (t . namely ! ! ! ! du = @F u + F . consider the mechanics of applying the trapezoidal method for the time integration of Eq. 2 h @F @u n 2 @t n which is now in the delta form which will be formally introduced in Section 12. and the trapezoidal time march.92 results in the expression 1 un = hFn + 1 h2 @F (6. In many uid mechanic applications the nonlinear function F is not an explicit function " ! ! # " !# ! .83. There are. one nds un+1 = un + 1 h Fn + @F (un+1 . A very useful reordering of the terms in Eq.6.6.83.93) 1 . 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.10. IMPLEMENTATION OF IMPLICIT METHODS 113 Notice that this is an expansion of the derivative of the function. t ) + O(h2) n n dt @u n @u n n @t n (6. locally-linear approximation to F (u t) that is valid in the vicinity of the reference station tn and the corresponding un = u(tn). The use of local time linearization updated at the end of each time step.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. Thus.90) Implementation of the Trapezoidal Method As an example of how such an expansion can be used. it represents a second-order-accurate. 6. un) + h @F + O(h2) + Fn 2 @u n @t n 2) +hO(h (6. The trapezoidal method is given by 1 un+1 = un + 2 h Fn+1 + Fn] + hO(h2) (6. 6. relative to the order of expansion of the function. combine to make a second-order-accurate numerical integration process.91) where we write hO(h2) to emphasize that the method is second order accurate. 6.

90 into this method.83. TIME-MARCHING METHODS FOR ODE'S of t.1 in carrying out this step). u 4. rearrange terms. and remove the explicit dependence on time. it is generally the space-di erenced form of the steady-state equation of some uid ow problem. 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. store them in an array say R. we arrive at the form !# n un = hFn (6. but for our applications. It is the product of h and the RHS of the basic equation evaluated at the previous time step. 6.94 is usually implemented as follows: ~ ~ 1.94) Notice that the RHS is extremely simple.96) .114 CHAPTER 6. 2 h @F @u !# n un = hFn (6. u 2. In this example. and save ~ n. Find ~ n+1 by adding ~ n to ~ n. the basic equation was the simple scalar equation 6. A numerical time-marching procedure using Eq. (Very seldom does one nd B . 6.95) if we introduce Eq. Let this storage area be referred to as B . 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 .93 simpli es to the second-order accurate expression " 1 1 . h @F @u (6. In such cases the partial derivative of F (u) with respect to t is zero and Eq. 6. Solve for the elements of hFn. Solve the coupled set of linear equations B ~n = R u ~ for ~ n. 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.

where the error is the di erence between the current solution and the converged solution. 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. The fact that it has quadratic convergence is veri ed by a glance at Eqs. let us consider an example involving some simple boundary-layer equations. By quadratic convergence. There results or . df A = 0 dt (6. @F @u ! n un = Fn @F @u ! #. Omission of this factor degrades the method in time accuracy by one order of h.1 n (6.98) This is the well-known Newton method for nding the roots of a nonlinear equation F (u) = 0. and is a scaling factor. We shall see later that this method is an excellent choice for steady problems..99) Here f represents a dimensionless stream function. given by Eq. The reader should ponder the meaning of letting h ! 1 for the trapezoidal method.4 Local Linearization for Coupled Sets of Nonlinear Equations In order to present this concept.e.96.87 and 6.94.88 (remember the dependence on t has been eliminated for this case).6. 6. 6. We choose the Falkner-Skan equations from classical boundary-layer theory.10.96 obtained by dividing both sides by h and setting 1=h = 0. Our task is to apply the implicit trapezoidal method to the equations d3f + f d2f + dt3 dt2 0 !21 @1 . 6. Use of a nite value of h in Eq.97) un+1 = un . 6. Quadratic convergence is thus a very powerful property.94 and 6.10.96 leads to linear convergence. " Fn (6. 6. i. . Newton's Method Consider the limit h ! 1 of Eq. 6. the error at a given iteration is some multiple of the error at the previous iteration. we mean that the error after a given iteration is proportional to the square of the error at the previous iteration.

100) (6.116 CHAPTER 6.104) 8 Recall that we derived the Jacobian matrices for the two-dimensional Euler equations in Section 9 The Taylor series expansion of a vector contains a vector for the rst term. rather than a simple nonlinear scalar function.87. and de ning F n as F (~ n). 6. called the Jacobian matrix.102 by the following process d~ = F (~ ) u ~ u dt (6. It is derived from Eq.8 Let us refer to this matrix as A. 1 . 6.99 to a set of rst-order nonlinear equations by the transformations This gives the coupled set of three nonlinear equations u01 = F1 = . @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. except that this time we are faced with a nonlinear vector function. The required extension requires the evaluation of a matrix. 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 (6.103) 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.u1 u3 .2 vector for the second term. and tensor products for the terms of higher order.102) A = (aij ) = @Fi =@uj For the general case involving a third order matrix this is (6. one has 9 2. TIME-MARCHING METHODS FOR ODE'S u1 = d f dt2 2 First of all we reduce Eq.101) Now we seek to make the same local expansion that derived Eq. 6. Omitting the explicit dependency ~ ~ u on the independent variable t.90. a matrix times a .

Any rst.102 as \constant" which is a locally-linear. Without any iterating within a step advance. of course. u3 2 u2 .or second-order time-marching method. u the solution will be second-order-accurate in time. the terms in the Jacobian matrix are updated as the solution proceeds depends. Using this we can write the local linearization of Eq.2 1 We nd ~ n+1 from ~ n + ~ n. which is given by Eq. 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. u2)n 3 2 2 6 2 7 6 .107) 4 5 0 1 0 The student should be able to derive results for this example that are equivalent to those given for the scalar case in Eq. 6.6. Re-evaluate u u u the elements using ~ n+1 and continue. 6. ~ n + O(h2) u u (6.93.106) 6. h(u2)n h (u1)n 72 ( u1)n 3 2 . 6. on the nature of the problem. tn and the argument for O(h2) is the same as in the derivation of Eq.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 . Find an expression for the nth term in the Fibonacci series. and the solution is now advanced one step. second-order-accurate approximation to a set of coupled nonlinear ordinary di erential equations that is valid for t tn + h. (u1u3)n .1. could be used to integrate the equations without loss in accuracy with respect to order. d~ = A ~ + F .101.11 Problems 1.88. (1 . 6. Thus for the Falkner-Skan equations the trapezoidal method results in 2 3 1 + h (u3)n . A ~ +O(h2) u ~ n n un nu dt | {z } (6.105) where t . Returning to our simple boundary-layer example. The number of times.u1 3 0 0 7 A=6 1 (6. explicit or implicit. and the manner in which. PROBLEMS 117 ~ u ~ F (~ ) = F n + An ~ .11. What is u25 ? (Let the rst term given above be u0. we nd the Jacobian matrix to be 2 .) .

(c) Find er . The 2nd-order backward method is given by i 1h un+1 = 3 4un . Identify the polynomials P (E ) and Q(E ). Consider the time-marching scheme given by un+1 = un. un. (d) Perform a -root trace relative to the unit circle for both di usion and convection. (a) What is the resulting O E? (b) What is its exact solution? (c) How does the exact steady-state solution of the O E compare with the exact steady-state solution of the ODE if = 0? 3. relation.118 CHAPTER 6.65) to the representative equation. Find the .relation. (c) Find er and the rst two nonvanishing terms in a Taylor series expansion of the spurious root. 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 .1 + 2hu0n+1 (a) Write the O E for the representative equation. Solve for the -roots and identify them as principal or spurious. 6. 5. Find the di erence equation which results from applying the Gazdag predictorcorrector method (Eq.1) (a) Write the O E for the representative equation.relation. (b) Derive the . 4. The trapezoidal method un+1 = un + 2 h(u0n+1 + u0n) is used to solve the representative ODE.1 + 23h (u0n+1 + u0n + u0n. Identify the polynomials P (E ) and Q(E ). TIME-MARCHING METHODS FOR ODE'S 1 2. 6. (b) Derive the .

trapezoidal. and 400 nodes. 7. 200. Use 2nd-order centered di erences in space and a grid of 50 points. Plot the solutions obtained at t = 1 compared to the exact solution (which is identical to the initial condition). Let u(0 t) = sin !t with ! = 10 . For the explicit Euler and AB2 methods. Using the computer program written for problem 8. 10. including the homogeneous and particular solutions. repeat problem 9 using 4th-order (noncompact) di erences in space. implicit Euler. PROBLEMS 119 Find the di erence equation which results from applying this method to the representative equation. On a log-log scale. use u(x 0) = e.0:5)= ]2 with = 0:08. 11.11.1 for the other methods. Use the explicit Euler. Find the solution to the di erence equation. and 4th-order Runge-Kutta methods. ah= x. plot the error given by v uX u M (uj . of 0. 8. uexact)2 u j t M j =1 where M is the number of grid nodes and uexact is the exact solution. Repeat problem 7 using 4th-order (noncompact) di erences in space. Show solutions at t = 1 and t = 10 compared to the exact solution. Run until a periodic steady state is reached which is independent of the initial condition and plot your solution . 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. use a Courant number of unity. Use only 4th-order Runge-Kutta time marching at a Courant number of unity. 2nd-order Adams-Bashforth (AB2). Using the computer program written for problem 7. use a Courant number. 9. Find er and er . What order is the homogeneous solution? What order is the particular solution? Find the particular solution if the forcing term is xed. Find the global order of accuracy from the plot.0:5 (x. 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. 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 100.6. Find the .relation. For the initial condition.

1. Use grids with 100. 3. 2. i.6. Use 2nd-order centered di erences in space with a 1st-order backward di erence at the out ow boundary (as in Eq.69. era . 6. and 4th-order Runge-Kutta time-marching at a Courant number of unity. nd the phase speed error. and j +1 see problem 1 in Chapter 3). for the combination of second-order centered differences and 1st. Find the global order of accuracy.. and the amplitude error. j . 2. . 200. TIME-MARCHING METHODS FOR ODE'S compared with the exact solution. Use a third-order forward-biased operator at the in ow boundary (as in Eq. and 400 nodes and plot the error vs.2. 3rd. At the last grid node. use a 3rd-order backwardbiased operator (using nodes j .69) together with 4th-order Runge-Kutta time marching. that obtained using the spatial discretization alone. 1. Repeat problem 11 using 4th-order (noncompact) centered di erences. and j ) and at the second last node. j . without actually deriving the methods. erp. 3. 13.e.120 CHAPTER 6.67). Also plot the phase speed error obtained using exact integration in time. j . 12. the number of grid nodes. derive and use a 3rd-order backward operator (using nodes j . 2nd. Explain your results. Note that the required -roots for the various Runge-Kutta methods can be deduced from Eq. 3. Using the approach described in Section 6. as described in problem 9. j .

͏ @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° .

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=-2.8 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 .

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 -1.0 1.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 R(λ h) -2.0 2.0 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 .

0 I(λh) b) Adams Type 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 2.0 2.0 Unstable Stable Outside 2.0 R( λh) ‡ ü „ .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 3.0 4.0 1.0 2.

0 R(λ h) .0 1.0 1.0 Stable only on Imag Axis b) Milne 4th Order I(λh) 2.0 -2.0 3.0 3.0 R(λ h) I(λh) 2.0 Stable -1.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.

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 „ .

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

CHOICE OF TIME-MARCHING METHODS Stable Region λh I(λ h) 1 R(λ h) Accurate Region Figure 8.27 and its solution using a one-root.1: Stable and accurate regions for the explicit Euler method. The situation is represented in the complex h plane in Fig. we exclude the forcing function from further discussion in this section. For a given time step. the time integration is an approximation in eigenspace that is di erent for every eigenvector ~ m .28. if at all.150 CHAPTER 8. timemarching method is represented by Eq. 6. the homogeneous part. 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. although these eigenvectors must remain coupled into the system to maintain a high accuracy of the spatial resolution. 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. Consider now the form of the exact solution of a system of ODE's with a complete eigensystem.1. 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. . 8. 6. This is given by Eq.

complex. It is important to notice that these de nitions make no distinction between real. and the other. although time accuracy requirements are dictated by the driving eigenvalues.1.2 Driving and Parasitic Eigenvalues For the above reason it is convenient to subdivide the transient solution into two parts. 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.1.8.2) xm xm m m Solution m=1 {z m=p+1 | } {z } | Driving Parasitic This concept is crucial to our discussion. 8. we nd that. called the parasitic eigenvalues (no time accuracy whatsoever is required for the eigenvectors associated with these. Rephrased. STIFFNESS DEFINITION FOR ODE'S 151 8. thus j 1j j 2j j Mj (8. First we order the eigenvalues by their magnitudes.1. but their presence must not contaminate the accuracy of the complete solution). numerical stability requirements are dictated by the parasitic ones. . Unfortunately.1) Then we write p M Transient = X c e m t ~ + X c e m t ~ (8. p+1 ! M ]. and imaginary eigenvalues. 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).3 Sti ness Classi cations In particular we de ne the ratio and form the categories The following de nitions are somewhat useful.

i. Under these conditions.1. 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. For a mesh size M = 40. 2 and the ratio is . 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. In this case the eigenvalues are all real. 2 sin 2(M + 1) x M ! . near an airfoil leading or trailing edge. 4x2 sin2 2 = . Typical CFD problems without chemistry vary between the mildly and strongly sti categories. One quickly nds that this grid clustering can strongly a ect the eigensystem of the resulting A matrix. Furthermore. In order to demonstrate this. the sti er our equations tend to become.e. 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. 8. 4x2 x 2=. 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. Consider the case when all of the eigenvalues are parasitic. 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. A simple calculation shows that 4 2 1 = . we are interested only in the converged steady-state solution. Our brief analysis has been limited to equispaced x . this ratio is about 680. Even for a mesh of this moderate size the problem is already approaching the category of strongly sti . and are greatly a ected by the resolution of a boundary layer since it is a di usion process.152 8.. the sti ness is determined by the ratio M = 1 .2 Relation of Sti ness to Space Mesh Size CHAPTER 8.2. The simplest example to discuss relates to the model di usion equation. let us examine the eigensystems of the model problems given in Section 4. negative numbers that automatically obey the ordering given in Eq.3. but much less so than for di usion-dominated problems. in di usion problems this sti ness is proportional to the reciprocal of the space mesh size squared. Examples of where this clustering might occur are at a shock wave.

3. T . PRACTICAL CONSIDERATIONS FOR COMPARING METHODS 153 problems. an accurate transient solution of a coupled set of ODE's. and technology in uencing this is undergoing rapid and dramatic changes as this is being written. if certain ground rules are agreed upon. Such a computation we refer to as an event. 8. Nevertheless. There is no unique answer to such a question. 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. d~ = F (~ t) u ~ u dt . It should then be clear how the analysis can be extended to other cases. 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. Let us now examine some ground rules that might be appropriate. Since there are an endless number of these methods to choose from. 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 . and architecture of the available computer. Er and Er! . rather than the local ones er . era . one can wonder how this information is to be used to pick a \best" choice for a particular problem. over a xed interval of time. the time interval would be that required to extract a reliable statistical sample. it is. in calculating the amount of turbulence in a homogeneous ow. Era . and erp discussed earlier. relevant conclusions can be reached. For example.6.3 Practical Considerations for Comparing Methods We have presented relatively simple and reliable measures of stability and both the local and global accuracy of time-marching methods. and the accuracy required of the solution are dictated by the physics of the particular problem involved. 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. This function is usually nonlinear. capacity. The length of the time interval. For example. and its computation usually consumes the major portion of the computer time required to make the simulation. Let us consider the problem of measuring the e ciency of a time{marching method for computing. discussed in Section 6. The appropriate error measures to be used in comparing methods for calculating an event are the global ones. highly dependent upon the speed.8.

3 is obtained from our representative ODE with = . The follow assumptions bound the considerations made in this Section: 1..e.. First of all. and RK4. must be < 0:5%. Eq. 6.4. ln(0:25) with u(0) = 1. 8. a = 0. Three methods are compared | explicit Euler. AB2. this makes the exact value of u at the end of the event equal to 0.u dt (8.3) Er < 0:005 eT . 2. Implications of computer storage capacity and access time are ignored. In some contexts.4 Comparing the E ciency of Explicit Methods As mentioned above.4. let us set the additional requirement The error in u at the end of the event. The calculation is to be time-accurate.t.1. the allowable error constraint means that the global error in the amplitude. 3. this can be an important consideration. Fev . must simulate an entire event which takes a total time T . and must use a constant time step size. To the constraints imposed above. so that CHAPTER 8.154 8. u(T ) = 0:25. the global error. We judge the most e cient method as the one that satis es these conditions and has the fewest number of evaluations.48. see Eq. i. the e ciency of methods can be compared only if one accepts a set of limiting constraints within which the comparisons are carried out. must have the property: du = .25.2 An Example Involving Di usion Let the event be the numerical solution of from t = 0 to T = . CHOICE OF TIME-MARCHING METHODS 8. 8. The time-march method is explicit. Since the exact solution is u(t) = u(0)e. h. i.e.1 Imposed Constraints T = Nh where N is the total number of time steps.

COMPARING THE EFFICIENCY OF EXPLICIT METHODS Then.1: Comparison of time-marching methods for a simple dissipation problem. The above constraint has led to the invention of schemes that omit the function evaluation in the corrector step of a predictor-corrector combination.4. the method with the highest local accuracy is the most e cient on the basis of the expense in evaluating Fev . ln(0:25)=N .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. and the fourth-order Runge-Kutta method is the best of all. Under these conditions a method is judged as best when it has the highest global accuracy for resolving eigenvectors with imaginary eigenvalues. The results shown in Table 8. In this numerical experiment the function evaluations contribute overwhelmingly to the CPU time. a complete event must be established in order to obtain meaningful statistical samples which are the essence of the solution. In this example we see that. Thus the second-order Adams-Bashforth method is much better than the rstorder Euler method.4.1. In this case. for a given global accuracy. 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. and the number of these evaluations must be kept to an absolute minimum because of the magnitude of the problem. leading to the so-called incomplete . The main purpose of this exercise is to show the (usually) great superiority of second-order over rst-order time-marching methods.4. in addition to the constraints given in Section 8.8. since h = T=N = . 8. we add the following: ~u The number of evaluations of F (~ t) is xed. On the other hand. it follows that 155 1 .1 were computed using a simple iterative procedure. ( 1 (ln(:25)=N ))N =:25 < 0:005 where 1 is found from the characteristic polynomials given in Table 7.1.

the global amplitude and phase error for kevaluation methods applied to systems with pure imaginary -roots can be written1 Era = 1 . The second and fourth order RK methods are 2. N . However. For a 2-evaluation method N = K=2 since two evaluations are used for each step. However. Notice that as k increases. j 1(ik!h1 )jK=k (8. and AB2 schemes are all 1-evaluation methods.38 and 6. given in Section 6. 6. so that for these methods h = h1 . the power to which 1 is raised to arrive at the nal destination decreases see the Figure below. notice also that as k increases.4) 1 See Eqs. the time step must be twice that of a one{evaluation method so h = 2h1. respectively. the derivative of the fundamental family is never found so there is only one evaluation required to complete each cycle.156 CHAPTER 8. This is the key to the true comparison of time-march methods for this type of problem. CHOICE OF TIME-MARCHING METHODS predictor-corrector methods. For a 4-evaluation method the time interval must be h = 4h1 . 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. K . that more e cient methods will result from the omission of the second function evaluation. leapfrog. Let K represent the total number of allowable Fev . after K evaluations.1.and 4-evaluation methods. The Gazdag. However. For a 1evaluation method the total number of time steps. in order to arrive at the same time T after K evaluations.8. are the same. the time span required for one application of the method increases. The . in this case. 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. In general. etc.relation for the method is shown as entry 10 in Table 7. and the number of evaluations. The presumption is. one evaluation being used for each step. . h. Basically this is composed of an AB2 predictor and a trapezoidal corrector. An example is the method of Gazdag. of course.39. Let h1 be the time interval advanced in one step of a one-evaluation method.

5. leapfrog. K tan. Amplitude.4. and the fact that ! = 1. 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. We consider two maximum evaluation limits K = 50 and K = 100 and choose from four possible methods. K leapfrog AB2 Gazdag RK4 !h1 = :1 100 . First of all we see that for a one-evaluation method h1 = T=K . in terms of phase error (for which it was designed) it is superior to the other two methods shown. The rst three of these are one-evaluation methods and the last one is a four-evaluation method.3e. 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. 7.3:8 . by some rather simple calculations2 made using Eqs. On the basis of global error the Gazdag method is not superior to RK4 in either amplitude or phase. For example. AB2. we nd.4 and 8. Using this.5) 1 (ik!h1 )]imaginary 1 (ik!h1 )]real # Consider a convection-dominated event for which the function evaluation is very time consuming. see Fig. The 1 root for the Gazdag method can be found using a numerical root nding routine to trace the three roots in the -plane. 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. K leapfrog AB2 Gazdag RK4 !h1 = :1 100 1:0 1:003 :995 :999 !h1 = :2 50 1:0 1:022 :962 :979 a. However. 2 . COMPARING THE EFFICIENCY OF EXPLICIT METHODS Er! = !T .1 k " 157 (8. the results shown in Table 8. we are making our comparisons on the basis of global error for a xed number of evaluations. 8. although. Gazdag.:96 .0. We idealize to the case where = i! and set ! equal to one. exact = 1. Phase error in degrees. Table 8. and RK4.9:8 1:5 1:5 b.2.2:4 :45 :12 !h1 = :2 50 .8. The event must proceed to the time t = T = 10.2: Comparison of global amplitude and phase errors for four methods.

In this case the trace forms a circle of unit radius centered at (. If h is chosen so that all h in the ODE eigensystem fall inside this circle the integration will be numerically stable. is h = 0:02.6. h)nx12 + (PS )1 u2(n) = c1 (1 . CHOICE OF TIME-MARCHING METHODS Using analysis such as this (and also considering the stability boundaries) the RK4 method is recommended as a basic rst choice for any explicit time-accurate calculation of a convection-dominated problem. For a speci c example. If some h fall outside the small circle but stay within the stable region. Numerical solutions. 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.001 is negligible.5 Coping With Sti ness 8.1.100 and 2 = . We have de ned these h as parasitic eigenvalues. The coupled equations in real space are represented by u1(n) = c1 (1 .158 CHAPTER 8. of course. 7. depend upon ( m h)]n and no j m j can exceed one for any m in the coupled system or else the process is numerically unstable. Also shown by the small circle centered at the origin is the region of Taylor series accuracy. 100h)nx11 + c2(1 . 8. the time histories of the two solutions would appear as a rapidly decaying function in one case. Analytical evaluation of the time histories poses no problem since e. If uncoupled and evaluated in wave space. and a relatively slowly decaying function in the other.1 Explicit Methods The ability of a numerical method to cope with sti ness can be illustrated quite nicely in the complex h plane.5 and 7. The time step limit based on stability.1.100t quickly becomes very small and can be neglected in the expressions when time becomes large. h)nx22 + (PS )2 (8. consider the mildly sti system composed of a coupled two-equation set having the two eigenvalues 1 = . 8.6) We will assume that our accuracy requirements are such that su cient accuracy is obtained as long as j hj 0:1. these h are sti . With this time step the . which is determined from 1. We will also assume that c1 = c2 = 1 and that an amplitude less than 0. Numerical evaluation is altogether di erent. Stability boundaries for some explicit methods are shown in Figs. Let us choose the simple explicit Euler method for the time march. 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 .5. We rst run 66 time steps with h = 0:001 in order to resolve the 1 term.1 0) as shown in Fig. but stable. 100h)nx21 + c2(1 .

the maximum step size that can be taken in order to maintain stability.4b.. We would then have a well resolved answer to the problem throughout the entire relevant time interval. After 66 steps.e. 50h nx + c u1(n) = c1 1 + 50h 11 2 ! 1 . 8. we would like to change the step size to h = 0:1 and continue.001.001 and can be neglected. we need to use a step size of about h = 0:001.e. which in just 10 steps at h = 0:1 ampli es those terms by 109. this is not possible because of the coupled presence of (1 . In fact. h)n) is 0. COPING WITH STIFFNESS 2 159 term is resolved exceedingly well. It follows that the nal numerical solution to the ODE is now represented in real space by 1 .001). with h = 0:02.5. and this has no physical meaning whatsoever. In both intervals the desired solution is second-order accurate and well resolved. since (0:666 )n < 1. but that is not the point of this exercise).9361. we simply replace the Euler with the trapezoidal one and analyze the result. After 70 time steps the 1 term has amplitude less than 0. with 69 steps required to reduce the amplitude of the second term to below 0.5. 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.001. Thus the total simulation requires 405 time steps. about 339 time steps have to be computed in order to drive e.. 50h nx + c u2(n) = c1 1 + 50h 21 2 ! 8. and. (1 .t to below 0. Hence the 1 term can now be considered negligible. (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. 100h)n. 100h)n) is less than 0.2 Implicit Methods 1 .001 and that of the 2 term (i. below 0. Since this is also a oneroot method. It is true that in the nal 69 steps one -root is 1 . its in uence in the remaining 70 . Its behavior in the h plane is shown in Fig. However. To drive the (1 . 7. h)n term to zero (i. 0:5h nx + (PS ) 1 1 + 0:5h 12 ! 1 . far outweighing the initial decrease obtained with the smaller time step.e. 0:5h nx + (PS ) 2 1 + 0:5h 22 ! (8. However. 50(0:1)]= 1 + 50(0:1)] = 0:666 . the amplitude of the 1 term (i. We choose the trapezoidal method..7) In order to resolve the initial transient of the term e. its in uence on the coupled solution is negligible at the end of the rst 70 steps.6 but this time using an unconditionally stable implicit method for the time march.100t . Now let us re-examine the problem that produced Eq. (1 . 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.8.

8. as the reduced number of time steps begins to outweigh the increased cost per time step. For preliminary investigations it is often the best method to use for mildly-sti di usion dominated problems. it should be clear that as the degree of sti ness of the problem increases. 2 = .3 A Perspective 8. its behavior for t > 0:07 is identical to that of a stable spurious root. about three times as many. which is in the strongly-sti category. the advantage begins to tilt towards implicit methods. It is important to retain a proper perspective on a problem represented by the above example. This is known as the multigrid method. we have no interest whatsoever in the transient portion of the solution. CHOICE OF TIME-MARCHING METHODS steps is even less. It has enjoyed remarkable success in many practical problems however. The reader can repeat the above example with 1 = . is recommended. such as Adams-Bashforth or Runge-Kutta methods.10 000. the Euler method is extremely easy to program and requires very little arithmetic per step. These explicit methods can be considered as e ective mildly sti stable methods. There is yet another technique for coping with certain sti systems in uid dynamic applications.1. However.160 CHAPTER 8. Therefore.. . Our sole goal is to eliminate it as quickly as possible. For re ned investigations of such problems an explicit method of second order or higher. However. although this root is one of the principal roots in the system. The total simulation requires 139 time steps. as compared to about 139 for the trapezoidal method. we need an introduction to the theory of relaxation before it can be presented.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. Actually.5.8) When solving a steady problem. the conditionally stable Euler method required 405 time steps. if there are more spurious roots (8. In the 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.

For example.4. the implicit Euler method is the obvious choice for steady problems. one would like to take the limit h ! 1. Eq. Repeat the time-march comparisons for di usion (Section 8. all of the eigenvalues can be considered to be parasitic. Therefore the time step can be chosen to eliminate the transient as quickly as possible with no regard for time accuracy. Consider the di usion equation (with = 1) discretized using 2nd-order central di erences on a grid with 10 (interior) points.1. However. 8. and the fourth-order Runge-Kutta method is a candidate for steady convection dominated problems.4. PROBLEMS 161 the choice of a time-marching method for a steady problem is similar to that for a sti problem. Find and plot the eigenvalues and the corresponding modi ed wavenumbers.2) and periodic convection (Section 8. Among implicit methods. If we use the explicit Euler timemarching method what is the maximum allowable time step if all but the rst two eigenvectors are considered parasitic? Assume that su cient accuracy is obtained as long as j hj 0:1.7. Considering the stability bounds for these methods (see problem 4 in Chapter 7) as well as their memory requirements. the di erence being that the order of accuracy is irrelevant.and 4th-order Adams-Bashforth methods. compare and contrast them with the 3rd. 6. a nite time step may be required until the solution is somewhat close to the steady solution. Hence the explicit Euler method is a candidate for steady di usion dominated problems. When we seek only the steady solution.4.4. 6.7 Problems 1.2) and periodic convection (Section 8. none of the eigenvalues are required to fall in the accurate region of the time-marching method. 3.8. Referring to Fig. Eq. What is the maximum allowable time step if all but the rst eigenvector are considered parasitic? .3) using the 3rd. because of their stability properties.and 3rd-order Runge-Kutta methods. which leads to Newton's method. 8. 2.and 4th-order Runge-Kutta methods. Repeat the time-march comparisons for di usion (Section 8. when using the implicit Euler method with local time linearization.98.3) using 2nd.96.

162 CHAPTER 8. CHOICE OF TIME-MARCHING METHODS .

which satis es the following coupled system of nonlinear algebraic equations: ~u F (~ ) = 0 (9.Chapter 9 RELAXATION METHODS In the past three chapters. 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. analyzing. 9. and choosing time-marching methods. Alternatively. we developed a methodology for designing. Following a time-dependent path to steady state is possible only if all of the eigenvalues of the matrix A (or .1) dt which arise after spatial discretization of a PDE.A) have real parts lying in the left half-plane.4) dt and the system is integrated in time until the transient has decayed to a su ciently low level. the ODE given by x b Eq. Although the solution ~ = A. 1 163 . 9. they can be used to solve for the steady solution of Eq.1. ~ exists as long as A is nonsingular. 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. a time derivative is introduced as follows d~ = A~ .2) In the latter case.4 has a stable steady solution only if A meets the above condition.3) To solve this system using a time-marching method. ~ x x b (9.

C~ b = A. This preconditioning has the e ect of multiplying Eq.1 Preconditioning the Basic Matrix It is standard practice in applying relaxation procedures to precondition the basic equation. To use these schemes. provided C . does not depend at all on the eigensystem of the basic matrix Ab . In the simplest possible case the conditioning matrix is a diagonal matrix composed of a constant D(b). C~ b = 0 u f Notice that the solution of Eq. 9.1 Formulation of the Model Problem 9. In the following we will see that our approach to the iterative solution of Eq.5) where Ab is nonsingular. and the use of the subscript b will become clear shortly. ~ b u f f b b f 1 1 1 1 1 (9. 9. which is given by ~ 1 = A. we seek to obtain rapidly a solution which is arbitrarily close to the exact solution of Eq.7 is (9.7 depends crucially on the eigenvalue and eigenvector structure of the matrix CAb. 9. 9. equally important.6) b f 1 9. 9. there are well-known techniques for accelerating relaxation schemes if the eigenvalues of CAb are all real and of the same sign. Such systems of equations arise. our analysis will focus on their application to large sparse linear systems of equations in the form Ab~ .2.1. RELAXATION METHODS The common feature of all time-marching methods is that they are at least rstorder accurate. and. If we designate the conditioning matrix by C . ~ b u (9.5 from the left by some nonsingular matrix. For example. the problem becomes one of solving for ~ in u CAb~ . at each time step of an implicit time-marching method or at each iteration of Newton's method. While they are applicable to coupled systems of nonlinear algebraic equations in the form of Eq. 9. Such methods are known as relaxation methods.7) ~ = CAb ].164 CHAPTER 9. In this chapter. ~ b = 0 u f (9. C . for example. the . Using an iterative method.5. C~ b = A.5.8) which is identical to the solution of Eq. exists. we consider iterative methods which are not time accurate at all.

We then further condition with multiplication by the negative transpose.9.2 .9) The matrix in Eq.1 32 0 7 6 .AT A = 2 1 1 2 0 6 .2 3 2 u 7 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. The result is A = .1 . FORMULATION OF THE MODEL PROBLEM 165 conditioning matrix C must be chosen such that this requirement is satis ed.1 0 6 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.1.2 0 6 6 1 0 .1 0 76 176 .11) 1 . For example.1 4 1 0 . It can rst be conditioned so that the modulus of each element is 1.9 has eigenvalues whose imaginary parts are much larger than their real parts.6).1 .10) which scales each row.1 3 7 7 7 7 7 17 5 1 2 0 6 . this leads to the approximation 82 > 0 1 >6 > 0 6 1 <6 . A choice of C which ensures this condition is the negative transpose of Ab.1 0 >4 : .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 1 76 0 76 76 . 9. Using a rst-order backward di erence on the right side (as in Section 3.1 54 3 7 7 7 17 7 17 5 1 0 .2 1 1 6 0 0 6 = 6 1 0 .1 1 1 ~= 0 u 2 x >6 x 6 >6 .2 4 1 (9.1 1 1 6 0 6 6 .

We do not necessarily recommend the use of .AT as a preconditioner we simply wish to show that a broad b range of matrices can be preconditioned into a form suitable for our analysis..1. ~ = 0 f (9. as de ned in Appendix A.2 1 6 1 6 1 .1 which has all negative real eigenvalues.13) where A is symmetric negative de nite. 2 1 6 .1 0 1 76 76 0 7 6 0 . The symbol for the dependent variable has been changed to as a reminder that the physics being modeled is no longer time 2 A permutation matrix (de ned as a matrix with exactly one 1 in each row and column and has the property that P T = P . 2 We use a symmetric negative de nite matrix to simplify certain aspects of our analysis. as given in Appendix B.2 4 3 7 7 7 7 7 17 5 (9.2 0 1 7 6 76 176 76 76 054 1 0 . RELAXATION METHODS 1 1 If we de ne a permutation matrix P and carry out the process P T . we will thoroughly investigate some simple equations which model these structures.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 . the conditioned matrix .2 0 1 76 7 6 1 0 . 1 . The classical methods will usually converge if Ab is diagonally dominant. Thus even when the basic matrix Ab has nearly imaginary eigenvalues. and the classical relaxation methods can be applied.166 1 1 CHAPTER 9.AT A ]P (which just reorders the elements of A and doesn't change the eigenvalues) we nd 2 60 60 6 60 6 60 4 1 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 7 6 .1 ) just rearranges the rows and columns of a matrix.2 .2 1 7 6 54 0 1 1 . symmetric with negative real eigenvalues).12) 9.AT Ab b is nevertheless symmetric negative de nite (i. In the remainder of this chapter.2 1 =6 6 6 1 . Relaxation methods are applicable to more general matrices. We will consider the preconditioned system of equations having the form A~ .e.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.

5 and 9. A = CAb and ~ = C~ b f f (9. (bc) g ~ (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.2 1)~ = ~ f (9. ~ u u ~ g (9. ~ . 9. FORMULATION OF THE MODEL PROBLEM 1 167 accurate when we later deal with ODE formulations.15) @ u = g(x) (9. g(x) @t @x This has the steady-state solution 2 2 2 2 (9. is guaranteed to exist because A is nonsingular.17) dt x ~ where (bc) contains the boundary conditions and ~ contains the values of the source g term at the grid nodes. Note that the solution of Eq. In this case 2 Ab = 1 B (1 . In the notation of f Eqs.16) @x which is the one-dimensional form of the Poisson equation. 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 .9. then A has the form A=B 1~ 1 b 2 2 2 .7. ~ = A. 9.2 1)~ + (bc) . we nd d~ = 1 B (1 .14) The above was written to treat the general case.1.18) Choosing C = x I .13. Introducing the threepoint central di erencing scheme for the second derivative with Dirichlet boundary conditions. we obtain B (1 .19) ~ where ~ = x fb.

A.25) 1 1 .1 is easily obtained from the matrix resulting from the Neumann boundary condition given in Eq.2 The Converged Solution.2 b . 9.2 s]T s = .24) Hence it is clear that H should lead to a system of equations which is easy to solve.and three-dimensional problems. H .20) Note that s = .23) where G I + H.2. The converged solution is designated ~ 1 so that ~ 1 = A. however. ~ 1 = ~ n . The error at the nth iteration is de ned as ~ n ~ n . ~ f 1 (9. and the Error +1 +1 1 1 1 Solving Eq. H .2 or . A]~ n .2 . We attempt to do this in such a way.168 CHAPTER 9. The matrix H is independent of n for stationary methods and is a function of n for nonstationary ones. 3.1 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 .2. A tremendous amount of insight to the basic features of relaxation is gained by an appropriate study of the one-dimensional case. that it is directly applicable to two.24 using a diagonal conditioning matrix.2 Classical Relaxation 9. ~ n = A~ n . and much of the remaining material is devoted to this case.21 for ~ n gives ~ n = I + H . The iteration count is designated by the subscript n or the superscript (n). 9. A (9. or at least easier to solve than the original system. the Residual. RELAXATION METHODS ~ = .1 (9. ~ = G~ n . ~ f f (9.22) 9.21) where H is some nonsingular matrix which depends upon the iterative method. ~ f (9. ~ e f (9.

The residual at the nth iteration is de ned as ~ n A~ n . In all of the above.26.9. Nonstationary processes in which H (and possibly C ) is varied at each iteration are discussed in Section 9. and the old value of j .25 by A from the left.2. Hence the j th row of Eq.2. ~ r f (9. and use the de nition in Eq.30) j j j 2 j.f i (9. 9. The Point-Jacobi method is expressed in point operator form for the one-dimensional case as n = 1 h n + n . The method of successive overrelaxation (SOR) ( +1) ( ) 1 ( ) +1 ( +1) 1 +1 ( +1) ( +1) 1 ( ) +1 1 1 +1 .27) Finally. 9. CLASSICAL RELAXATION 169 where ~ 1 was de ned in Eq.22. and j . and its eigenvalues. This operator is a simple extension of the point-Jacobi method which uses the most recent update of j.29 is satis ed. 9.26) Multiply Eq. This operator comes about by choosing the value of jn such that together with the old values of j. There results the relation between the error and the residual A~ n . 9. which we designate as m .3 The Classical Methods Point Operator Schemes in One Dimension Let us consider three classical relaxation procedures for our model equation B (1 . ~ n = 0 e r (9. The Gauss-Seidel method is n = 1h n + n . .5.29 is satis ed using the new values of j and j . G is referred to as the basic iteration matrix. f i (9.28) Consequently. the j th row of Eq.2 1)~ = ~ f (9. determine the convergence rate of a method. we have considered only what are usually referred to as stationary processes in which H is constant throughout the iterations.29) as given in Section 9.1.31) j j j 2 j. 9. 9. it is not di cult to show that ~ n = G~ n e e +1 (9.2.

~ f (9. h i n = jn + ! ~j . It is usually expressed in two steps as h i ~j = 1 jn + jn .13 into its diagonal. L. but it can also be written in the single line n = ! n + (1 .D. A~ .3 The ODE Approach to Classical Relaxation The particular type of delta form given by Eq. 9. One can easily see that Eq. 9.(L + D). which certainly meets the criterion that it is easy to solve. jn (9. and the portion above the diagonal.21 leads to an interpretation of relaxation methods in terms of solution techniques for coupled rst-order ODE's.170 CHAPTER 9. U . such that A=L+D+U (9.1 The Ordinary Di erential Equation Formulation . being lower triangular.34) Then the point-Jacobi method is obtained with H = .32) j ( +1) 1 ( ) +1 ( +1) ( ) ( ) where ! generally lies between 1 and 2.3. fj 2 . C A ~ . 9.35) dt This is equivalent to d~ = H . ! ) n + ! n .33) j j 2 j. 9. The Gauss-Seidel method is obtained with H = . 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.36) b dt 1 1 9. which is also easy to solve. about which we have already learned a great deal.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~ . then perhaps it would be better to move further in this direction. ~ ] fb f (9. ! f (9. D. ~ = H . RELAXATION METHODS is based on the idea that if the correction produced by the Gauss-Seidel method tends to move the solution toward ~ 1. the portion of the matrix below the diagonal.

9. H and C in Eq. Hence the numerical solution after n steps of a stationary relaxation method can be expressed as (see Eq. The eigenvalues are xed by the basic matrix in Eq. A have negative real parts (which implies that H . given by 1 1 1 1 ~ 1 = A. Assuming that H . and the eigenvectors of H written as ~ = c e 1 t~ + + cM e M t~ M +~ 1 x} (9. is now referred to in relaxation analysis as the error. an iteration process has been decided upon).13. so that we must devise a way to remove them all from the general solution on an equal basis. The generalization of this in our approach is to make h. 9. ~ is also independent u f . A has been chosen (that is.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. In general it is assumed that any or all of the eigenvectors could have been given an equally \bad" excitation by the initial guess.36 are independent of t. and the secondary conditioning matrix in 9. they are not changed throughout the iteration process. the three classical methods have all been conditioned by the choice of H to have an optimum h equal to 1 for a stationary iteration process. the solution can be of t. A is nonsingular). In a stationary method. Suppose the explicit Euler method is used for the time integration. the preconditioning matrix in 9. the \time" step. a constant for the entire iteration. As we shall see. 1 . These are xed by the initial guess.7. 6.38) which is the solution of Eq. Throughout the remaining discussion we will refer to the independent variable t as \time". H . if all of the eigenvalues of H . The eigenvalues are xed for a given h by the choice of time-marching method. ~ f 1 (9. A are linearly independent. 9.37) | x {z error where what is referred to in time-accurate analysis as the transient solution. A depends on neither ~ nor t. For this method m = 1 + m h.39) The initial amplitudes of the eigenvectors are given by the magnitudes of the cm . THE ODE APPROACH TO CLASSICAL RELAXATION 1 1 1 171 In the special case where H .3.36. 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. 9. the only free choice remaining to accelerate the removal of the error terms is the choice of h. then the system of ODE's has a steady-state solution which is approached as t ! 1. that is. even though no true time accuracy is involved.35. It is clear that.

h.42) +1 +1 It is clear that the best choice of H from the point of view of matrix algebra is . ~ f (9.41) n = n+h n with a step size.3.35. we are led to 2 B (. Now let us study these methods as subsets of ODE as formulated in Section 9.2 ODE Form of the Classical Methods The three iterative procedures de ned by Eqs. 9.B .2 1) since then multiplication from the left by .2.1. ~ n) = B (1 . ~ f (9. However. since multiplication by the inverse amounts to solving the problem by a direct method without iteration.1. Then ~ H d = B (1 . 9.1: VALUES OF and ! IN EQ.40) dt As a start.2 1)~ . We arrive at H (~ n .172 CHAPTER 9.31 and 9. (1 .3 can be identi ed using the entries in Table 9.32 obey no apparent pattern except that they are easy to implement in a computer code since all of the data required to update the value of one point are explicitly available at the time of the update. let us use for the numerical integration the explicit Euler method 0 (9.43 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 . this is not in the spirit of our study. 9. ! 0)(~ n .30.2 1)~ n .43) where and ! are arbitrary. equal to 1. ~ n) = B (1 .3 is that all the elements above the diagonal (or below the diagonal if the sweeps are from right to left) are zero.2.2 1)~ n . If we impose this constraint and further restrict ourselves to banded tridiagonals with a single constant in each band.29 into the ODE form 9.3. 1 +1 TABLE 9.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) gives the correct answer in one step. Insert the model equation 9. RELAXATION METHODS 9. With this choice of notation the three methods presented in Section 9.

1) n + !h n .9.49) where ~ n is the transient. or steady-state. and ~ 1 A. !h f (9. 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. In this light. (.48) dt This solution has the analytical form ~ n = ~n + ~ 1 e (9. are identi ed for the one-dimensional case by Eq. j 2 j. 1 . and SOR. However. our purpose is to examine the whole procedure as a special subset of the theory of ordinary di erential equations. 9. 2 2 j 2 j 1 ( +1) ( +1) 1 ( ) 1 ( ) ( ) +1 This represents a generalization of the classical relaxation techniques. ~ = 0 f (9. ~ is the steady-state solution. ~ b = 0 u f (9.2 1)~ . Gauss-Seidel.4. (!h .44 and Table 9.45) j j. or error.4 Eigensystems of the Classical Methods The ODE approach to relaxation can be summarized as follows.46) This equation is preconditioned in some manner which has the e ect of multiplication by a conditioning matrix C giving A~ . ~ f (9. the three methods are all contained in the following set of ODE's d~ = B . ~ f (9. The e f three classical methods. 2 0) B (1 . ) n . The basic equation to be solved came from some time-accurate derivation Ab~ . solution of the set of ODE's ~ H d = A~ . 9. Point-Jacobi. 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.44) dt ! and appear from it in the special case when the explicit Euler method is used for its numerical integration.1.

174 CHAPTER 9. For h < 1. d = . The three special cases considered below are obtained with a = 1. the optimum stationary case.relation for the explicit Euler method is m = 1 + m h. the maximum j m j is obtained with m = M . 9. e = 2=!. is shown in Fig.1.44. A) having maximum absolute value. c = 1. 2 The eigensystem can be determined from Appendix B. This amounts to solving the generalized eigenvalue problem A~ m = m H~ m x x (9. j m j > 1:0.4. and 9.54) .3. The plot for h = 1. and f = 0.30.1 2 ).50) In this section.1 since both d and f are zero. To illustrate the behavior. b = . we use the ODE analysis to nd the convergence rates of the three classical methods represented by Eqs. RELAXATION METHODS 1 Given our assumption that the component of the error associated with each eigenvector is equally likely to be excited. 9.52) ! xm The generalized eigensystem for simple tridigonals is given in Appendix B.2 and for h > 1. the maximum j mj is obtained with m = 1. To obtain the optimal scheme we wish to minimize the maximum j m j which occurs when h = 1. we take M = 5 for the matrix order.1 The Point-Jacobi System 1 If = 0 and ! = 1 in Eq. A matrix for the three methods. 9. the ODE matrix H .51) for the special case 2 0)~ (9. 9.31. 9. 9. i.2 1)~ m = x m B (.32.e. . Fig. A reduces to simply B ( 1 . j j = j M j and the best possible convergence rate is achieved: 1 1 j m jmax = cos M + 1 (9. 9. This special case makes the general result quite clear. It is also instructive to inspect the eigenvectors and eigenvalues in the H . Fig.1 + cos M +1 m = 1 2 ::: M The .53) m = . The eigenvalues are given by the equation m (9. This relation can be plotted for any h. B (1 . Note that for h > 1:0 (depending on M ) there is the possibility of instability. the asymptotic convergence rate is determined by the eigenvalue m of G ( I + H . Thus Convergence rate j mjmax m=1 2 1 M (9.2.2.

23 = . 3=2 7 5 4 p 5 4 1=2 .57) 4 = . 9. 3 )h]n~ x 2 1 x + c 1 .1:866 From Eq. 3=2 7 6 6 6 7 ~ 6 . A are given by ~ j = (xj )m = sin j m x j = 1 2 ::: M (9. EIGENSYSTEMS OF THE CLASSICAL METHODS 1 175 For M = 40.1 = . 3=2 7 ~ = 6 p0 7 x = 6 p 7 6 1 7 x 6 6 7 6 3=2 7 7 6 . the numerical solution written in full is 5 p ~ n .1 + 1 = . 9. 3=2 7 4 5 4 5 5 4 p 7 1 1=2 . 3=2 3 2 p 3 2 3 1= p =2 7 p2 7 6 .4.1 .56) The corresponding eigenvalues are. ~ 1 = c 1 .1. Thus after 500 iterations the error content associated with each eigenvector is reduced to no more than 0.1 + 23 = .53 1 2 3 = . 3=2 1 2 3 4 5 (9. Again from Appendix B. we obtain j mjmax = 0:9971.1:0 = .1 .0:5 2 = . from Eq.23 times its initial level. (1 .9.0:134 = .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 . 2 )h]n~ 1 2 2 p 1 .55) M +1 This is a very \well-behaved" eigensystem with linearly independent eigenvectors and distinct eigenvalues.39. the eigenvectors can be written as 2 1=2 3 2 3 2 p3=2 3 p 7 p 7 6 3=2 7 6 1 7 6 3=2 7 6 6 0 7 6 ~ = 6 p 1 7 ~ = 6 p 7 ~ = 6 . 1 = . the eigenvectors of H . For M = 5. (1 .1:5 2 p (9. The rst 5 eigenvectors are simple sine waves.

0 m=1 Values are equal σm m=4 m=2 Μ=5 0.58) 9. (. RELAXATION METHODS h = 1. the matrix eigensystem evolves from the relation B (1 .59) 1 which can be studied using the results in Appendix B.0 m=5 CHAPTER 9.1 2 0)~ m x (9.176 1. + c 1 .44.0 m=3 -1.0 -2. is AGS B . AGS .2.2 1)~ m = x m B (. One can show that the H .0 Figure 9.4. A matrix for the Gauss-Seidel method. (1 3 4 + c 1 .1 2 0)B (1 . h = 1 M = 5.1: The relation for Point-Jacobi. (1 + 23 )h]n~ x 4 5 )h]n~ x 3 5 (9.2 The Gauss-Seidel System If and ! are equal to 1 in Eq. (1 + 1 )h]n~ 2 x p + c 1 .0 0. 9.2 1) = 1 .

072 1.0 -1. Exceeds 1.0 De cr ea h si ng σm ng si Figure 9.0 -2.0 177 1.0 0.2: The relation for Point-Jacobi.072 / Ustable h > 1. .0 -2.0 h 0.0 at h De cr ea -1. h = 1:1 M = 5.0 h > 1.0 h < 1.4.0 0.0 λ mh ∼ ∼ 1. EIGENSYSTEMS OF THE CLASSICAL METHODS Approaches 1. h = 0:9 M = 5.0 λmh Figure 9.9.0 ng si ea cr In In cr ea si σm ng h h M = 5 0.3: The relation for Point-Jacobi.

. m + ~ m = cos m x sin j M + 1 m = 1 2 ::: M2 1 (9. . 4 5 0 1=2M M The eigenvector structure of the Gauss-Seidel ODE matrix is quite interesting. . .63) 3 2 h~ i 7 6 h~ i 7 6 6 7 2~ 37 . is shown in Fig. .3=4 1=2 7 2 6 7 6 0 1=8 . . .1 + cos ( M +1 and the corresponding eigenvectors are given by j. .principal vectors. 7 .23. they do not represent a common family for di erent values of M . For M = 40. . .. 9. the error associated with the \worst" eigenvector is removed in half as many iterations.62) M +1 The . .61) m = . j m jmax = 0:9942.3=4 1=2 7 3 6 7 6 7 6 0 1=16 1=8 .64) .4. The Jordan canonical form for M = 5 is j m jmax = cos M + 1 2 (9.1 1=2 3 1 6 0 . Hence the convergence rate is 2 1 CHAPTER 9. . producing waves that are higher in amplitude on one side (the updated side) than they are on the other. The m with the largest amplitude is obtained with m = 1. They are no longer symmetrical.. . .3=4 1=2 7 5 6 7 6 . If M is odd there are (M + 1)=2 distinct eigenvalues with corresponding linearly independent eigenvectors.. RELAXATION METHODS 2 . .relation for h = 1. . Furthermore. and there are (M .. 250 iterations are required to reduce the error component of the worst eigenvector by a factor of roughly 0. The eigenvectors are quite unlike the Point-Jacobi set. 1)=2 defective eigenvalues with corresponding 178 Since this is the square of that obtained for the Point-Jacobi method. 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. The equation for the nondefective eigenvalues in the ODE matrix is (for odd M ) + m ) m = 1 2 ::: M2 1 (9. 6 . the optimum stationary case. 7 .60) 6 0 1=32 1=16 1=8 .3=4 1=2 7 4 6 7 (9.

0 0. linked to (5) Jordan block The numerical solution written in full is thus ~ n .0 σm Μ=5 2 defective λ m 0.1 (4) Defective.1 7 ~ = 6 4 7 (9. 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 .9. EIGENSYSTEMS OF THE CLASSICAL METHODS 1.0 -2. h )n~ 4 x h x + c (1 . ~ 1 = c (1 . 3=32 1 2 3 4 5 The corresponding eigenvalues are = .4: The relation for Gauss-Seidel.0 λmh Figure 9.3=4 ) = . 34 )n~ 1 2 3 3 1 1 2 2 (9. The eigenvectors and principal vectors are all real.4. 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 .0 -1. h = 1:0 M = 5.66) .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 .0 179 h = 1.1=4 = .

(.2x + x x . the remaining eigenvalues are complex and occur in conjugate pairs. and the following conditions hold: 1. h)n. equal and defective. 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 . (. ASOR B . 3. 2x + x x . x 0 6 1 6 . 9. the system is Gauss-Seidel. is 1 1 2 x 0 0 6 .180 CHAPTER 9. If ! is chosen such that 4(1 .67) 9. h)n. !) + ! pm] = pm = cos m =(M + 1)] 2 2 2 1 If ! = 1. 2x x x 6 . !) + ! pm < 0 zm and m are complex. RELAXATION METHODS " # n + c h n (1 . The H .2. One can show that this can be written in the form given below for M = 5. !) + ! p = 0 ! is optimum for the stationary case. + c h n(n .1 + !pm 2 zm 2 m = 1 2 :::M 2 2 1 2 (9. the ODE matrix is B . If M is even. If M is odd. The generalization to any M is fairly clear. 2x 4 . .69) where zm = 4(1 . one of the remaining eigenvalues is real and the others are complex occurring in conjugate pairs. 2x + x x . h) x 1! 2! + c (1 . Two eigenvalues are real.2x + x x x 2x2x x 6 x + .68) 7 7 5 Eigenvalues of the system are given by m + = . 2x 0 0 0 x 4 4 3 7 7 7 7 (9. A matrix for the SOR method. 1) (1 . 2x + x 6 x . ~ + c (1 .2 1). h)n~ x 5 1! 1 4 5 (9.1 x 0)B (1 . If 4(1 . ~ x 3 4 1 5 2 2 3 4 5 + c (1 . h)n. 2. h)n + c h n (1 . 2x + x x .2x x .4. + x .2 1).3 The SOR System 1 If = 1 and 2=! = x in Eq.2 + x 1 .1 x 0)B (1 .44.

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 = . h + h m . j mjmax = 0:8578. In practical applications. This illustrates the fact that for optimum stationary SOR all the j m j are identical and equal to !opt . 1.2=3 (2) Defectivep linked to = .9.72) !opt = 2= 1 + sin M + 1 For M = 40. 2p2i)=9 = .1 j . For M = 5 they can be written 2 3 2 3 2 p3(1 3 2 3 1=2 7 . the optimum value for the stationary case.5. Hence the convergence rate is j m jmax = !opt . and the bene t may not be as great. Hence the worst error component is reduced to less than 0.(10 + 2 2i)=9 = .70) !opt = 2= 1 + sin M + 1 ~m x m = = m. 1 (9.71) Using the explicit Euler method to integrate the ODE's.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 (9.23 times its initial value in only 10 iterations. much faster than both GaussSeidel and Point-Jacobi. there are two real eigenvectors and one real principal vector. 9. and if h = 1. m = 1 .relation reduces to that shown in Fig.1 m sin 2 where m j M +1 (9.4=3 (9. For odd M . p m= m 2 m 2 1 2 q .(10 .4. the optimum value of ! may have to be determined by trial and error. the . The remaining linearly independent eigenvectors are all complex.74) 2 1 2 34 5 1 1 3 4 5 !opt p + i p . EIGENSYSTEMS OF THE CLASSICAL METHODS One can easily show that the optimum ! for the stationary case is and for ! = !opt 181 (9.

39 is 1 . ~ b . 9. ~1 = + + + + c (1 .0 λmh Figure 9. It is much simpler. h. RELAXATION METHODS h = 1.0 -1.5 Nonstationary Processes In classical terminology a method is said to be nonstationary if the conditioning matrices. The nonstationary form of Eq.5: The relation for optimum stationary SOR. For the Gauss-Seidel and SOR methods it leads to processes that can be superior to the stationary methods. In our ODE b f approach this could also be considered and would lead to a study of equations with nonconstant coe cients. This does not change the steadystate solution A. but it can greatly a ect the convergence rate. The numerical solution written in full is ~n . This process changes the Point-Jacobi method to Richardson's method in standard terminology. 2h=3)n~ x p c 1 . are varied at each time step. H and C . 4h=3)n~ x 1 2 1 2 3 2 3 4 5 4 5 1 (9.0 CHAPTER 9. to study the case of xed H and C but variable step size. (10 + 2 2i)h=9] x c (1 . 2h=3)n + c nh(1 .75) 9. ]~ x c (1 .0 -2. 2p2i)h=9]n~ x n~ c 1 . however. M = 5.0 0.0 2 real defective λ m σm 2 complex λ m 1 real λ m Μ=5 0. h = 1.182 1. 2h=3)n. (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). However. Now we pose a less demanding problem.80) .9. for m = 1 2 M .76. we considered making the error exactly zero by taking advantage of some knowledge about the discrete values of m for a particular case. It is therefore unnecessary and impractical to set hm = . the eigenvalues m are generally unknown and costly to compute. ~ 1 = X cm~ m Y (1 + e x m=1 n=1 m hn ) (9.76) where the symbol stands for product. Consider one of the error terms taken from M N ~ N ~ N . Mathematically stated. This leads to the concept of selectively annihilating clusters of eigenvectors from the error terms as part of a total iteration process. In the annihilation process mentioned after Eq. Now focus attention on the polynomial (Pe)N ( ) = (1 + h )(1 + h ) 1 2 (1 + hN ) (9.5. 9.1= 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. 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 0.79) treating it as a continuous function of the independent variable . This is the basis for the multigrid methods discussed in Chapter 10.78) where Pe signi es an \Euler" polynomial. the error term can theoretically be completely eliminated in M steps by taking hm = . 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 + (9.1= m . We will see that a few well chosen h's can reduce whole clusters of eigenvectors associated with nearby 's in the m spectrum. we seek b max a j(Pe)N ( )j = minimum with(Pe)N (0) = 1 (9. Since hn can now be changed at each step.

! (Pe)N ( ) = where (9. In relaxation terminology this is generally referred to as Richardson's method. 9. a.86) .184 CHAPTER 9. Eq.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.85) hn = 2= 3 . and it leads to the nonstationary step size choice given by ( " #) 1 = 1 .84.1 y 1 and 2 2 (9. 1 + 1 y .1 using only 3 iterations. 9. TN . b ! . b a b This problem has a well known solution due to Markov. . ampli ed by the coe cients cm Q(1 + m hn). 9. The three values of h which satisfy this problem are (9. With each x eigenvector there is a corresponding eigenvalue.2 . b TN a. 1) 6 " #! (9. let us consider the following problem: Minimize the maximum error associated with the eigenvalues in the interval . RELAXATION METHODS a. cos (2n .81) TN (y) = cos(N arccos y) are the Chebyshev polynomials along the interval . ~ m . 1) n = 1 2 :::N (9. As an example for the use of Eq. 1 (9. ) cos (2n . y . It is 2 . The error in the relaxation process represented by Eq.82) q q N N 1 TN (y) = 2 y + y .83) 2 are the Chebyshev polynomials for jyj > 1. + ( .76 is expressed in terms of a set of eigenvectors.

2 .87) (9. 3) h = 4=(6 . 8] =2 99 3 3 3 A plot of Eq.76. 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 (9.9.37 on the condition that the explicit Euler method. If instead the implicit trapezoidal rule n+1 = n+ is used. with (Pt)N (0) = 1 (9.6 and we see that the amplitudes of all the eigenvectors associated with the eigenvalues in the range . 2 hn under the same constraints as before.6 are h = 4=(6 . 9. 9. This was derived from Eq. Eq.92) .90) b max a j(Pt)N ( )j = minimum . p 0) h = 4=(6 + 3) 1 2 3 p Return now to Eq. Pt: 0 1 1 N B 1 + hn C Y (Pt )N ( ) = B 2 C (9. 9.1 have been reduced to less than about 1% of their initial values. 9.88) where n p p o T (3) = 3 + 8] + 3 .5.41. 2 hn =1 =1 1 mC C + ~1 A m (9. namely that =1 0 1 M N 1+ h ~ N = X cm~ m Y B 2 n B x @ 1 m n 1 .87 is given in Fig. 9. was used to integrate the basic ODE's. 9.89) would result. The values of h used in Fig.91) @ 1 A n 1 . This calls for a study of the rational \trapezoidal" polynomial.

6 −0.8 0.2 0.6 −0.015 0.5 0.01 −0.2 0 0.6 −1.005 (P ) (λ) e 3 0 −0.2 .7 0. RELAXATION METHODS 1 0.1 0 −2 −1.02 0.186 CHAPTER 9.4 0.005 −0. .4 −0.02 −2 −1.8 −0.8 −1.4 −1.015 −0.4 −0. minimization over .01 0.1.2 −1 λ −0.2 −1 λ −0.8 −0.8 −1.6 (Pe )3 (λ) 0.2 0 Figure 9.6: Richardson's method for 3 steps.3 0.6 −1.4 −1.9 0.

9. b 1 2 2 +1 1 +1 1 2 Determine the iteration matrix G if = . consider the iterative method (I + A )~ = (I .2. 2. 2 =. A. a n=1 2 N (9. Then nd the corresponding j m j values.6. PROBLEMS ! 187 The optimum values of h can also be found for this problem. A )xn + b x (I + A )xn = (I . Gn)A. = N .7.7 are h = 1 p h = 2 h = 2 ( 1) ( 1) 3 3 1 2 3 9. x x not just the expressions. A )~ + b x where is a parameter. 3. nd the eigenvalues of H .6 Problems 1. (You do not have to nd H . Show that this iterative method can be written in the form H (xk . but we settle here for the approximation suggested by Wachspress n. Recall that the eigenvalues of H .85. The error amplitude is about 1/5 of that found for (Pe) ( ) in the same interval of . 9. f show that 1 ~n = Gn ~ + (I .) Find the numerical values. f ~ 0 1 1 2 where G = I + H . The results for (Pt) ( ) are shown in Fig. For a linear system of the form (A + A )x = b. .1=2. 1 1 1 . xk ) = (A + A )xk . The values of h used in Fig. A for the SOR method with A = B (4 : 1 .2 1) and ! = !opt. 9.93) b hn b This process is also applied to problem 9. Using Appendix B. Given a relaxation method in the form ~ H ~n = A~n . A satisfy A~ m = mH~ m .

4 −1.1.2 0 Figure 9.1 0 −2 −1.2 −1 λ −0.7 0.5 1 0.2 −1 λ −0.8 −0.9 0.2 0.188 CHAPTER 9.8 −1.6 −0.5 −2 −2 −1.8 −1.6 (P ) (λ) t 3 0.4 −0.3 0.7: Wachspress method for 3 steps.6 −0. .5 −1 −1.4 −1.4 −0.6 −1. RELAXATION METHODS 1 0.8 0.5 0.5 (Pt )3 (λ) 0 −0.4 0.6 −1. minimization over .8 −0.2 0 2 x 10 −3 1.2 .

(c) the SOR method with the optimum value of !. 6x = 0 @x For the initial condition. the number of iterations. 3. Iterate to steady state using (a) the point-Jacobi method.5.9. Use second-order centered di erences on a grid with 40 cells (M = 39). and 4 orders of magnitude. Plot the solution after the residual is reduced by 2. and (d) the 3-step Richardson method derived in Section 9.6. u(1) = 1: @ u . PROBLEMS 189 4. Compare with the theoretical asymptotic convergence rate. Solve the following equation on the domain 0 x 1 with boundary conditions u(0) = 0. (b) the Gauss-Seidel method. Plot the logarithm of the L -norm of the residual vs. use u(x) = 0. Determine the asymptotic convergence rate. 2 2 2 .

190 CHAPTER 9. RELAXATION METHODS .

10. Note that @ sin(mx) = .3.2 and 4. This has a rational explanation from the origin of the banded matrix.1.2 1)~ = X 1 D(~ ) X .1) then the corresponding eigenvectors are ordered from low to high space frequencies. Notice that as the magnitudes of the eigenvalues increase.2 1). if the eigenvalues are ordered such that j j j j 1 2 2 j Mj (10. That is.1 Motivation 10.2) @x and recall that 1 B (1 . The eigenvalues and eigenvectors are given in Sections 4.3. ~ (10. respectively.3.1 Eigenvector and Eigenvalue Identi cation with Space Frequencies Consider the eigensystem of the model matrix B (1 . The viewpoint presented here is a natural extension of the concepts discussed in Chapter 9.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.m sin(mx) (10.3) xx~ = x x 191 2 2 2 2 1 . the space-frequency (or wavenumber) of the corresponding eigenvectors also increase. There are many variations of the process and many viewpoints of the underlying theory.

1. As we saw in Section 9. error components corresponding to high space frequencies will be reduced more quickly than those corresponding to low space frequencies. 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.4) m= M +1 x Hence. Therefore. of the Gauss-Seidel method and.192 1 CHAPTER 10.m m << M (10. this correlation is not necessary in general. the complete counterexample of the above association is contained in the eigensystem 1 for B ( 2 1 1 ). 10. for example. We have seen that X .2 + 2 cos m . It is also true. However. For this matrix one nds. The classical point-Jacobi method does not share this property. exactly the opposite 2 behavior. 2 2 2 2 2 10. as shown in Fig. from Appendix B. This is consistent with the physics of di usion as well. this method produces the same value of j j for min and max . In fact.1.2. the correlation of large magnitudes of m with high space-frequencies is to be expected for these particular matrix operators. One nds that 1 M + 1 .m . ~ represents a sine transform. and X ~ . the operation 1 D(~ ) represents the numerical approximation of the multiplication of the x appropriate sine wave by the negative square of its wavenumber. 9. . a sine synthesis. of the Richardson method described in Section 9. the property can be restored by using h < 1.2 Properties of the Iterative Method The second key motivation for multigrid is the following: Many iterative methods reduce error components corresponding to eigenvalues of large amplitude more e ectively than those corresponding to eigenvalues of small amplitude. This is the key concept underlying the multigrid process.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 . by design.5. However. This is to be expected of an iterative method which is time accurate. MULTIGRID where D(~ ) is a diagonal matrix containing the eigenvalues.4.

7) where A CAb (10. 9. At the end of the three steps we nd ~ .5. but for clarity we suppose that the three-step Richardson method illustrated in Fig. 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. THE BASIC PROCESS 193 basic properties. We do not attempt to develop an optimum procedure here. 5.11. Having preconditioned (if necessary) the basic nite di erencing scheme by a procedure equivalent to the multiplication by a matrix C . 9. as in the example given by Eq.27.5) where the matrix formed by the product CAb has the properties given above. 4. of the matrix are all real and negative. In Eq. if f ~ 1 is a vector representing the desired exact solution. The eigenvalues.10.8) . as in the examples given by Eq.6 is used. These conditions are su cient to ensure the validity of the process described next. We start with some any. ~ b ] = 0 f (10.2. the vector ~ b represents the boundary conditions and the forcing function. ~ = 0 e r (10. The eigenvectors associated with the eigenvalues having largest magnitudes can be correlated with high frequencies on the di erencing mesh. or it can be \contrived" by further conditioning. ~ b] r f (10.6) Recall that the ~ used to compute ~ is composed of the exact solution ~ 1 and the r ~ in such a way that error e A~ . r where ~ = C Ab~ . the residual. 3. This form can be arrived at \naturally" by simply replacing the derivatives in the PDE with di erence schemes. The m are fairly evenly distributed between their maximum and minimum values. 10. The basic assumptions required for our description of the multigrid process are: 1. 2. 3. we are led to the starting formulation C Ab ~ 1 . The problem is linear. m . and initial guess for ~ 1 and proceed through n iterations making use of some iterative process that satis es property 5 above.

11) Multiply Eq.7 for ~ then e ~1 = ~ . 10.13) Notice that e r e A ~e + A ~o = ~e 1 2 (10.194 If one could solve Eq.12) 1 The operation PAP . since a permutation matrix has an inverse which is its transpose.10) } Combining our basic assumptions.14) . 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. MULTIGRID (10.7 from the left by P and. assumption 4 ensures that the error has been smoothed.~ e CHAPTER 10.9) Thus our goal now is to solve for ~ . P ]~ = P ~ e r (10. 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). one containing the odd entries. 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. We can write the exact solution for ~ in terms of e e the eigenvectors of A. e In addition. Next we construct a permutation matrix which separates a vector into two parts. we can write PA P . 10. 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).

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.14 reduces to A ~ e + A A0 ~ e = ~ e e e r 1 2 2 (10. 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. It is that there is some connection between ~ e and ~ o brought about by the smoothing property of the Richardson e e relaxation procedure. the averaging of ~ is our only approximation. At this point we make our one crucial assumption. 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.15) e 2 1 (e + e ) e b 2 1 2 3 2 4 5 4 6 2 7 6 (10. no aliasing of these functions can occur in subsequent steps. r If the boundary conditions are Dirichlet.16) 0 0 1 With this approximation Eq. in this formulation.2. Hence. notice that. 10. e no operations on ~ are required or justi ed. it is reasonable to suppose that the odd points are the average of the even points. Finally.19) or Ac~ e . Since the top half of the frequency spectrum has been removed.18) (10. For example 1 (e + e ) e 2 a 1 (e + e ) e 2 1 (e + e ) or ~ o = A0 ~ e e e (10. ea and eb are zero.17) (10. ~ e = 0 e r where Ac = A + A A0 ] 1 2 2 .10.

1 { 3 7 (10.22) and are illustrated in Fig.23) 4 5 0 0 2 1 2 . However. ea = e . the matrix on the coarse mesh. the example in Eq.55 for Dirichlet conditions. and. the eigenvector structure is di erent from that given in Eq. 10.2 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. All of them go through zero on the left (Dirichlet) side. the interpolation formula in Eq. A in the 1-D model equation is B (1 ~ 1) where ~ = .2 6 6 6 6 6 6 .18 gives z }| 2 . and all of them re ect on the right (Neumann) side. In the particular case illustrated in Fig.15 must be changed.2 1).2 1 1 1 . In the present case they are given by " !# (2m .1.2 ::: . in fact.19. our 7-point example would produce 2 . When eb = eM .1. 9. 1) xjm = sin j 2M + 1 m=1 2 M (10. 10. 10.2 1 1 1 1 1 1 . The eigensystem is given by b b Eq. 10.1 1=2 7 6 7 = 4 1=2 . 10.196 CHAPTER 10.2 .2 .1 are met. all of the properties given in Section 10. If the original A had been B (7 : 1 .2 . For Neumann conditions.1]T .2 6 . It is easy to show that the high space-frequencies still correspond to the eigenvalues with high magnitudes.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 .21) 5 If the boundary conditions are mixed Dirichlet-Neumann.1 = 6 PAP 6 1 6 6 6 1 6 6 4 .20) and Eq.2 4 A1 .1 1=2 7 5 1=2 . MULTIGRID The form of Ac.2 .16 changes to 2 1 0 03 7 16 A0 = 2 6 1 1 0 7 60 1 17 (10. eb is equal to eM . is completely determined by the choice of the permutation matrix and the interpolation approximation. B. If Neumann conditions are on the left.2 .

THE BASIC PROCESS 197 X Figure 10.2 6 . At this stage.1 7 5 1=2 . we have reduced the problem from B (1 . we must now determine e ~ e and ~ .24) 5 which gives us what we might have \expected. we need to consider the eigensystems of A and Ac: A = X X.1=2 { 3 1=2 7 (10. cos m=1 2 M M +1 M +1 1 1 .2 1)~e = ~ e on the next coarser mesh. The permutation matrix remains the same and both A and A in the partitioned matrix PAP .2 1 . Therefore.2 1)~ = ~ on the e r 1 B (1 .25) For A = B (M : 1 . (10.9.10.2. In order to complete the process. and thus ~ . Ac = Xc cXc. Recall that our goal is ne mesh to 2 e r to solve for ~ . are unchanged (only A is modi ed by putting .15.2 1) the eigenvalues and eigenvectors are m j=1 2 M (10. Given ~ e e e computed on the coarse grid (possibly using even coarser grids). 10." We will continue with Dirichlet boundary conditions for the remainder of this Section.26) ~ m = sin j m x m = . we can construct the coarse matrix from 1 1 2 4 }| z 2 .1 1=2 7 6 7 = 4 1=2 .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 . which will provide us with the solution ~ 1 using Eq. we can compute ~ o e using Eq.1 in the lower right element).1: Eigenvectors for the mixed Dirichlet{Neumann case. 10.2 4 A1 . the relationship between e e In order to examine this relationship.

it contains the even elements of ~ . one can easily see that (~ c) coincides with ~ at x x every second point of the latter vector. that is. Then the residual is x e x ~ = A~ = ~ r x x (10. = . i. then Mc = (M . 5 0 (10. ~ e = Xc .27) M +1 For example. with M = 51.33) 1 1 1 1 1 1 1 1 1 . 7 6 . If we restrict our attention to odd M . 1)=2 is the size of Ac.28) For M = 51 (Mc = 25).31) 2 3 1 607 6 7 = Xc .. The exact solution on the coarse grid x x satis es ~ e = A. The eigenvalue and eigenvector corresponding to m = 1 for the matrix Ac = 1 B (Mc 1 .e.0:003649. 7 4 5 0 (10..0:007291 = 1:998 .2 1 .. (~ c) e x c r c (10. ( c) approaches 2 . 1 . 6 . ~ = ~ . x Now let us consider the case in which all of the error consists of the eigenvector component ~ . corresponding to ~ = sin j = . cos M2+ 1 (~ c) = sin j M2+ 1 j = 1 2 x Mc (10. Xc. MULTIGRID Based on our assumptions. 7 c 6 7 4 .2 1) are 2 ( c) = .30) since (~ c) contains the even elements of ~ .198 CHAPTER 10. As M increases. the most di cult error mode to eliminate is that with m = 1.32) 2 3 1=( c) 6 0 7 6 7 = Xc 6 . cos M + 1 x j=1 2 M (10. In addition.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. we obtain ( c) = .

This is equivalent to solving or 1A ~ = ~ e r 2 ce e (10.34) (10. and usually is. which gives Ab = x B (M : 1 . in each case the appropriate permutation matrix and the interpolation approximation de ne both the down.2. However.2 1) = x B (Mc : 1 .40) which is precisely the matrix appearing in Eq.37) c 4 In our case.and up-going paths. since xc = 2 x. 10. the matrix A = B (M : 1 . Thus we see that the process is recursive.37. but they vary depending on the problem and the user. The remaining steps required to complete an entire multigrid process are relatively straightforward. The problem to be solved on the coarse grid is the same as that solved on the ne grid.10. The reduction can be.15).35) Since our goal is to compute ~ = ~ .36) 1 B (M : 1 . we must multiply the result by 2.2 1) 4 c c 2 2 2 (10. x C x B (Mc : 1 .2 1) comes from a discretization of the di usion equation. obviously. . in addition to interpolating ~ e to the ne grid e x e (using Eq.39) Applying the discretization on the coarse grid with the same preconditioning matrix as used on the ne grid gives. The details of nding optimum techniques are. THE BASIC PROCESS = ( c) 1 1 199 (~ c) x 1 1 1 (~ ) 2 xc 1 (10. carried to even coarser grids before returning to the nest level.2 1)~ = ~ ee re (10. quite important but they are not discussed here. 10.38) C= x I 2 (10.2 1) and the preconditioning matrix C is simply 2 (10.2 1) = 1 B (Mc : 1 .

41) 1 (1) 1 (1) 1 1 1 where G = I + H. 9. the restriction matrix is 2 3 0 1 0 0 0 0 0 R =6 0 0 0 1 0 0 0 7 (10. Extension to nonlinear problems requires that both the solution and the residual be transferred to the coarse grid in a process known as full approximation storage multigrid.46) 1 2 See problem 1 of Chapter 9. Eq. f .g.44) In our example in the preceding section. ~ e r 2 (2) (2) (2) 2 1 (2) 2 (10. Gn1 ) A. Gn1 ) A. f (10.21). Perform n iterations of the selected relaxation method on the ne grid.43) 1 1 1 1 1 (1) (1) (1) 1 1 1 1 1 1 2." Some form of weighted restriction can also be used. 1. 10.. f r ~ = AGn1 ~n . the rst three rows of the permutation matrix P in Eq. (1) (2) 2 (1) 1 2 1 3. f = AGn1 ~n + A (I .42) and H is de ned as in Chapter 9 (e. which can be easily generalized to a process with an arbitrary number of grids due to the recursive nature of multigrid. This type of restriction is known as \simple injection.200 10. A (10.3 A Two-Grid Process CHAPTER 10. starting with ~ = ~n.45) 4 5 0 0 0 0 0 1 0 that is. . AGn1 A. MULTIGRID ~ We now describe a two-grid process for the linear problem A ~ = f . This gives ~ = Gn1 ~n + (I . Transfer (or restrict) ~ to the coarse grid: r ~ =R ~ r r (10. Next compute the residual based on ~ : ~ ~ ~ ~ = A ~ .11. Solve the problem A ~ = ~ on the coarse grid exactly: e r ~ = A. Call the result ~ . f ~ (10. Note that the coarse grid matrix denoted A2 here was denoted Ac in the preceding section.

.3. 10. Otherwise. apply the two-grid process recursively. I A. 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 (10.50) Thus the basic iteration matrix is I . In the 1 preceding example (eq. R A ]Gn2 2 3 2 3 3 1 3 2 2 2 2 2 3 3 2 3 (10. . R A]Gn1 1 2 2 1 2 1 1 1 1 The eigenvalues of this matrix determine the convergence rate of the two-grid process. 10.10. and A is obtained by discretizing on grid 3. I ~ e +1 (1) 1 (2) 2 (10.2 1). I A. with (I . 10. I A. Extension to four or more grids proceeds in similar fashion. The basic iteration matrix for a three-grid process is found from Eq. 4.49) (10. I A. I . Combining these steps. I and R are the transfer operators between grids 2 and 3. Transfer (or prolong) the error back to the ne grid and update the solution: ~n = ~ . R A]Gn1 ~ n . solve exactly. where 2 2 3 2 G = I . one obtains ~n = I . f ~ ~ +1 1 2 2 1 2 1 1 1 2 2 1 2 1 1 1 1 (10. If this is the coarsest grid in the sequence. f + A.47) 3 7 7 7 7 7 7 7 7 7 7 7 5 In our example. A TWO-GRID PROCESS 2 2 201 Here A can be formed by applying the discretization on the coarse grid. A = 4 B (Mc : 1 .40). G )A.15. It is at this stage that the generalization to a multigrid procedure with more than two grids occurs. R A]Gn1 A.48) which follows from Eq.50 by replacing A.51) In this expression n is the number of relaxation steps on grid 2.

Solve exactly on the coarsest grid.5. 3. Repeat problem 4 of Chapter 9 using a four-grid multigrid method together with (a) the Gauss-Seidel method. 2. (b) the 3-step Richardson method derived in Section 9. Determine the asymptotic convergence rate. 2 . 10. Derive Eq.4 Problems CHAPTER 10.202 10. the number of iterations. and 4 orders of magnitude. MULTIGRID 1. Calculate the theoretical asymptotic convergence rate and compare.51. Plot the logarithm of the L -norm of the residual vs. Plot the solution after the residual is reduced by 2.

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

(1 + )uj. .ax Bp(. )uj+1] = . ) The eigenvalues of this matrix are m = .1 + 2 uj + (1 . cos 2M m + i sin 2M for m = 0 1 : : : M . the forward di erence operator ( = . We proceed next to show why this occurs. NUMERICAL DISSIPATION @u = .1.1 .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.uj. If a is negative.1) produces Re( m ) > 0. our choice of di erencing scheme produces nonphysical behavior. With 6= 0.a x m 1 .1 + uj+1) + (.a( x u)j = 2.a x = 2.ax . When Re( m) = 0. 2 1 . and the backward di erence operator produces Re( m) < 0.1) with periodic boundary conditions. the centered di erence operator ( = 0) produces Re( m ) = 0. For periodic boundary conditions the corresponding matrix operator is . 1 If a is positive.2) The second form shown divides the operator into an antisymmetric component (. uj+1)] 2 x (11.uj. the roles are reversed.1+ uj+1)=2 x and a symmetric component (. In either case. uj+1)=2 x. The antisymmetric component is the second-order centered di erence operator. the solution will either grow or decay with time.uj. Hence the forward di erence operator is inherently unstable while the centered and backward operators are inherently stable.204 11. Consider the following point operator for the spatial derivative term .a (.1 + 2uj .a @u @t @x (11. the operator is only rst-order accurate.uj.1 + 2uj . A backward di erence operator is given by = 1 and a forward di erence operator is given by = .

2 .1. i 3 + 4 or r = .1. 11. 2 ({z.5) | | amplitude phase . a x2 @ 3 u + a x3 @ 4 u + : : : (11. 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. We are lead to the expression 2 ! 3 ! ! ! @ 2 u + x3 @ 3 u .ia . 11. 11.a @u + a x @ 2 u .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. x can be small but it is not zero. Consider the simple linear partial di erential equation @u = . Thus a ) u = e. Notice that Eq. 11. 11.2 to Eq. since the two equations are identical when x ! 0. when we use a computer to nd a numerical solution of the problem. 11.2 The Modi ed Partial Di erential Equation ( xu)j = 2 1 x 42 x @u . (a + 2 ) The solution is composed of both amplitude and phase terms.a @u + @ 2 u + @ 3 u + @ 4 u (11.2. However. 2} ei x.1 gives @u = . We refer to Eq.11. @x j 205 First carry out a Taylor series expansion of the terms in Eq.4) @t @x @x2 @x3 @x4 Choose periodic boundary conditions and impose an initial condition u = ei x. 11.3 is excited to some degree. 11.({z+ 2 )t}] (11. THE MODIFIED PARTIAL DIFFERENTIAL EQUATION 11.2. 2 ) s = . 11. Substituting this into Eq. We proceed next to investigate the implications of this concept.4 of the form u(x t) = ei xe(r+is)t provided r and s satisfy the condition r + is = . Under these conditions there is a wave-like solution to Eq. This means that each term in the expansion given by Eq.3 is consistent with Eq. 2 ( .3 as the modi ed partial di erential equation.

2 > 0 and hence the amplitude of the solution decays. the coe cients of the odd derivatives. Under the same condition. one could choose = 1=2 in Eq.4.1 + 3uj + 2uj+1) = 1 (uj. If the wave speed a is positive. 11.2 . 6uj. This is tantamount to deliberately adding dissipation to the PDE. Referring to the modi ed PDE.2 . Therefore. The symmetric component approximates x3 uxxxx=12. the choice of a forward di erence scheme ( = .6) The antisymmetric component of this operator is the fourth-order centered di erence operator.1 + 6uj . The resulting spatial operator is not one-sided but it is dissipative. By examining the term governing the phase of the solution in Eq. Our purpose here is to investigate the properties of one-sided spatial di erencing operators relative to centered di erence operators. this operator produces fourth-order accuracy in phase with a third-order dissipative term.2 . the coe cients of the even derivatives in Eq.2. 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). 11.1 + 8uj+1 . One can easily show. This we refer to as dispersion. Eq. Note that the use of one-sided di erencing schemes is not the only way to introduce dissipation. . by using the antisymmetry of the matrix di erence operators. 11.206 CHAPTER 11. Furthermore. 8uj. 4uj+1 + uj+2)] (11. we see that the speed of propagation is a + 2 . Any symmetric component in the spatial operator introduces dissipation (or ampli cation). Therefore. As a corollary. uj+2) 12 x + (uj. Therefore. any departure from antisymmetry in the matrix di erence operator must introduce dissipation (or ampli cation) into the modi ed PDE.1) is equivalent to deliberately adding a destabilizing term to the PDE.3 we have = . that the same is true for any centered di erence approximation of a rst derivative.5. a third-order backward-biased scheme is given by ( xu)j = 6 1 x (uj. For example. the phase speed of the numerical solution is less than the actual phase speed. 11.a x2 =6. the numerical phase speed is dependent upon the wavenumber . Biased schemes use more information on one side of the node than the other. and the phase depends only on a and . the choice of a backward di erence scheme ( = 1) produces a modi ed PDE with . NUMERICAL DISSIPATION It is important to notice that the amplitude of the solution depends only upon and . 4uj.

3 The Lax-Wendro Method 207 In order to introduce numerical dissipation using one-sided di erencing. the linear convection equation @u + a @u = 0). 2u(n) + u(n) ) (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. 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. known as the Lax-Wendro method. 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. the corresponding fully-discrete matrix operator is 0 2 !3 ! 2 ! 31 1 4 ah + ah 25 1 . There is no intermediate semi-discrete stage. backward di erencing must be used if the wave speed is positive. It is a fully-discrete nite-di erence scheme. THE LAX-WENDROFF METHOD 11. Thus @t @x @ 2 u = a2 @ 2 u @u = . u(n) ) + 1 ah (u(n) .8) @t @x @t2 @x2 Now replace the space derivatives with three-point centered di erence operators. i ah sin 2 m m =1.11. x M x M for m = 0 1 : : : M . This explicit method di ers conceptually from the methods considered previously in which spatial di erencing and time-marching are treated separately.9) uj j j +1 j j .1 2 x This is the Lax-Wendro method applied to the linear convection equation.1 2 x j+1 j. and forward di erencing must be used if the wave speed is negative. 1 For j ah j 1 all of the eigenvalues have modulus less than or equal to unity and hence x the method is stable independent of the sign of a. The quantity j ah j is known as the x Courant (or CFL) number.7) First replace the time derivatives with space derivatives according to the PDE (in this case. . ah 2 1 4. It is equal to the ratio of the distance travelled by a wave in one time step to the mesh spacing. cos 2 m . 1 ah (u(n) .3.

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. a8h ( x2 . a2h2 ) @ 3 u = .11) 2 x u +1 ~ which can be shown to be identical to the Lax-Wendro method. Recall MacCormack's timemarching method. which is given by @u + a @u = .1 a dissipative second-order method is obtained. . the leading error term in the Lax-Wendro method is dispersive and proportional to a ( x2 . a2h2 ) @ 3 u . u(jn)1) ~ .6 n @x3 @x3 6 The dissipative term is proportional to 2 4 2 2 4 2 @ . The two methods di er when applied to nonlinear hyperbolic systems. a2 h ( x2 . For the linear convection equation. these should be applied alternately. ah (~(jn+1) . x u(jn+1) = 1 u(jn) + u(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. 1 Or vice-versa for nonlinear problems. 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. ah (u(jn) . Cn) @xu 4 @x4 This term has the appropriate sign and hence the scheme is truly dissipative as long as Cn 1. a h8 x (1 . a x2 (1 . 11. u(jn+1))] ~ (11. C 2 ) @ 3 u . Hence MacCormack's method has the same dissipative and dispersive properties as the Lax-Wendro method.208 CHAPTER 11. A closely related method is that of MacCormack. Therefore. 11. a ( x2 . however.9 and converting the time derivatives to space derivatives using Eq. this approach leads to u(jn+1) = u(jn) . a2h2 ) @ u = .8. Recall that the odd derivatives on the right side lead to unwanted dispersion and the even derivatives lead to dissipation (or ampli cation. depending on the sign). presented in Chapter 6: un+1 = un + hu0n ~ 1 ~ ~ (11. The two leading error terms appear on the right side of the equation.

uj+1)] (11. we present the basic ideas of ux-vector and ux-di erence splitting.1 + uj+1) + jaj(. With this approach. a.1 + 2uj .12) @t @x where we do not make any assumptions as to the sign of a. then a+ = 0 and If a 0.a( xu)j = 2. The above approach to obtaining a stable discretization independent of the sign of a can be written in a di erent.uj. as a result of their superior exibility. by adding a symmetric component to the spatial operator. 11. whether it is a forward or a backward di erence) or the sign of the symmetric component depends on the sign of the wave speed.1. However.e. Alternatively. = 0.5. some form of splitting is required.4 Upwind Schemes 209 In Section 11.. For example. we saw that numerical dissipation can be introduced in the spatial di erence operator using one-sided di erence schemes or. we present two splitting techniques commonly used with upwind methods.4. but entirely equivalent. This is achieved by the following point operator: . This is avoided in the Lax-Wendro scheme. In the next two sections. Now for the a+ ( 0) term we can safely backward di erence and for the a. we see that a stable discretization is obtained with = 1 if a 0 and with = . When the ow is subsonic. manner. 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.1x a(. then a a. 11.uj. the wave speeds can be both positive and negative. that is. From Eq. These are by no means unique. This is the basic concept behind upwind methods. In order to apply one-sided di erencing schemes to such systems. the direction of the one-sided operator (i. more generally. Consider again the linear convection equation: @u + a @u = 0 (11. if a 0. In this section. these are of mixed sign.1 if a 0.13) This approach is extended to systems of equations in Section 11. = a 0. ) @u = 0 a = a 2 jaj @t @x + = a 0 and a. When a hyperbolic system of equations is being solved. We can rewrite Eq. ( 0) term forward di erence.2. UPWIND SCHEMES 11.11. For more subtle aspects of implementation and application of such techniques to . schemes in which the numerical dissipation is introduced in the spatial operator are generally preferred over the Lax-Wendro approach. the eigenvalues of the ux Jacobian for the onedimensional Euler equations are u u + a u .12 as @u + (a+ + a.

contains the negative eigenvalues. @w term. = . To accomplish this.19) .18) @t @x @t @x @x The spatial terms have been split into two components according to the sign of the wave speeds. hyperbolic system of partial di erential equations given by @u + @f = @u + A @u = 0 (11.1 Flux-Vector Splitting Recall from Section 2. We can now rewrite the system in terms of characteristic variables as @w + @w = @w + + @w + . A. @w = 0 (11. Premultiplying by X and inserting the product @x X . where += (11.15) i @t @x where the wave speeds. are the eigenvalues of the Jacobian matrix. In order to apply a one-sided (or biased) spatial di erencing scheme. is positive.5 that a linear.1Xw = 0 @t @x @x (11. i. + contains the positive eigenvalues and . NUMERICAL DISSIPATION nonlinear hyperbolic systems such as the Euler equations. .j j (11. let us split the matrix of eigenvalues.1Xw + @X .X . into two components such that = ++ .17) 2 2 With these de nitions. and the wi's are the characteristic variables.14) @t @x @t @x can be decoupled into characteristic equations of the form @wi + @wi = 0 (11.1X in the spatial terms gives @Xw + @X +X .4. i. the reader is referred to the literature on this subject. and a forward di erence if the wave speed is negative. We can use backward di erencing for the + @w term and forward @x di erencing for the . 11.210 CHAPTER 11.16) +j j . constant-coe cient. we need to apply a backward di erence if the wave speed.

and A. This approach is known as ux-vector splitting. u @u + @f + + @f .. = X . @u @u Therefore. UPWIND SCHEMES With the de nitions2 211 A+ = X +X . one must nd and use the new Jacobians given by A++ = @f @u + (11.11. u = 0 @t @x @x Finally the split ux vectors are de ned as f + = A+u and we can write (11.26) For the Euler equations. has all negative eigenvalues. Note that f = f+ + f.25) A. the de nition of the split uxes follows directly from the de nition of the ux. In the nonlinear case.21) (11. . we obtain A. we are in e ect solving the characteristic equations in the desired manner. and A. has all negative eigenvalues. @f + 6= A+ @f .23) @t @x @x In the linear case.22) f . A++ has eigenvalues which are all positive. (11. = @f @u . = 0 (11. 2 With these de nitions A+ has all positive eigenvalues. f is also equal to Au as a result of their homogeneous property.1 and recalling that u = Xw. term. 6= A. f = Au. the Jacobians of the split ux vectors are required. (11. When an implicit time-marching method is used.. as discussed in Appendix C.4.1 (11. = A.20) @u + @A+u + @A.X .24) Thus by applying backward di erences to the f + term and forward di erences to the f . For the Euler equations.

In a nite-volume method. more suited to nite-volume methods.33) 2 where jAj = X j jX .1 (11. as in Chapter 5.19. we premultiply by X to return to the original variables and insert the product X . 11. constant-coe cient.28) In order to obtain a one-sided upwind approximation. This is achieved with 1 1 (^i)j+1=2 = 2 i (wi)L + (wi)R ] + 2 j ij (wi)L . is given by 1 gj+1=2 = 2 (g(wL) + g(wR)) ^ (11. Now.212 CHAPTER 11. uR) (11.1X after and j j to obtain 1 1 X gj+1=2 = 2 X X . respectively. wL and wR.30) or 1 gj+1=2 = 1 (wL + wR) + 2 j j (wL . as in Eq. The centered approximation to gj+1=2. (wi)R] g (11. is known as ux-di erence splitting.4. NUMERICAL DISSIPATION 11. We again begin with the diagonalized form of the linear. which is nondissipative. we require ( (w ) (^i)j+1=2 = i(wi)L if i > 0 g (11. we consider the numerical ux at the interface between nodes j and j + 1.29) i i R if i < 0 where the subscript i indicates individual components of w and g. as a ^ function of the states to the left and right of the interface.34) . wR ) ^ (11.32) and thus 1 f^j+1=2 = 1 (fL + fR ) + 2 jAj (uL . the uxes must be evaluated at cell boundaries. hyperbolic system of equations @w + @w = 0 (11.1X (wL + wR) + 2 X j jX . gj+1=2. wR ) ^ (11.31) 2 Now.27) @t @x The ux vector associated with this form is g = w.1X (wL .2 Flux-Di erence Splitting Another approach.

and A = X X . However. ARTIFICIAL DISSIPATION 213 and we have also used the relations f = Xg.A. jAj = A if they are all negative. 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. In this case.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.5. 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. respectively. constant-coe cient case.36) 2 if Aj+1=2 satis es fL . in the nonlinear case. In order to resolve this. Eq. uR) (11.3 ( 11. fR = Aj+1=2 (uL .39) p L+p R R R where u and H = (e + p)= are the velocity and the total enthalpy per unit mass.37 is satis ed by the ux Jacobian evaluated at the Roe-average state given by p LuL + p R uR (11. If the eigenvalues of A are all positive. 11. In the linear. We have also seen that such dissipation can be introduced by adding a symmetric component to an antisymmetric (dissipation-free) operator. 11.1. there is some ambiguity regarding the de nition of jAj at the cell interface j + 1=2.37) For the Euler equations for a perfect gas. uR) (11. of this chapter.35) R if all i s < 0 giving pure upwinding.38) uj+1=2 = p + p L R p H +p H L L Hj+1=2 = (11.35 is obtained by the de nition 1 f^j+1=2 = 2 (fL + fR ) + 1 jAj+1=2j (uL . 3 Note that the ux Jacobian can be written in terms of u and H only see problem 6 at the end . Hence satisfaction of Eq.11. consider a situation in which the eigenvalues of A are all of the same sign. jAj = . u = Xw.

42) x (Au) = x (Au) + x (jAju) or a s (11. 4uj. A more complicated treatment of the numerical dissipation is also required near shock waves and other discontinuities. 11.36 and 11. For details of how this can be implemented for nonlinear hyperbolic systems.44) x (uj. as in the previous two sections. With an appropriate value of .43) x f = x f + x (jAju) where jAj is de ned in Eq.40) ( x u)j = uj+1 . s ( xu)j = 11.2 . This symmetric operator approximates x3 uxxxx and thus introduces a third-order dissipative term.uj+1 + 2uj . s It is common to use the following operator for x (11. The appropriate implementation is thus a s (11. but is beyond the scope of this book. uj.41) i x = i x + j ij x Extension to a hyperbolic system by applying the above approach to the characteristic variables. applying x = x .214 For example. a s With appropriate choices of x and x. The second spatial term is known as arti cial dissipation.6 Problems 1. this approach can be related to the upwind approach.1 + 3uj ) 2 x . NUMERICAL DISSIPATION s a ( xu)j = . x is stable if i 0 and unstable if i > 0. 11.15 is stable if i 0 and a s unstable if i < 0. 11.2 . 4uj.1 2 x 2 x a s Applying x = x + x to the spatial derivative in Eq.34. gives a s (11. A second-order backward di erence approximation to a 1st derivative is given as a point operator by ( xu)j = 1 (uj.1 + 6uj . uj. This is particularly evident from a comparison of Eqs. this often provides su cient damping of high frequency modes without greatly a ecting the low frequency modes. Similarly.43. the reader should consult the literature. 4uj+1 + uj+2) where is a problem-dependent coe cient.1 (11. It is also sometimes referred to as arti cial di usion or arti cial viscosity. let CHAPTER 11.

. Plot the real and imaginary parts of x vs. Find the modi ed wavenumber for this operator with = 0 1=12 1=24. 2. Consider the spatial operator obtained by combining second-order centered differences with the symmetric operator given in Eq.2.11. 4.37. Using Fourier analysis as in Section 6.6. (See Appendix A.6. PROBLEMS 215 (a) Express this operator in banded matrix form (for periodic boundary conditions). Using Fourier analysis as in Section 6. 11. Form the plus-minus split ux vectors as in Section 11. Show that the ux Jacobian for the 1-D Euler equations can be written in terms of u and H . x for 0 x . 11. x for 0 x . 3.38 and 11. Plot the real and imaginary parts of x vs.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. Find the modi ed wavenumber for the operator given in Eq. 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. Show that the use of the Roe average state given in Eqs. x for 0 x . Find the matrix jAj.6.2. 11. Consider the hyperbolic system derived in problem 8 of Chapter 2.1. 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. x for 0 x . Using Fourier analysis as in Section 6. 6.39 leads to satisfaction of Eq. x for 0 x .4.6. nd the derivative which is approximated by the corresponding symmetric and skew-symmetric operators and the leading error term for each. 5. then derive the symmetric and skew-symmetric matrices that have the matrix operator as their sum. Find the modi ed wavenumber for the rst-order backward di erence operator.2. 11. and 1=48.44. x for 0 x . Plot the real and imaginary parts of x vs.3 to see how to construct the symmetric and skew-symmetric components of a matrix.) (b) Using a Taylor table.

NUMERICAL DISSIPATION .216 CHAPTER 11.

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

neither one is to be preferred over the other. in this sense. ~ u u f dt 1 2 2 and consider the above observations. 12. SPLIT AND FACTORED FORMS d~ = A + A ]~ .h A A ~ n): u (12. from observation 2 (new data ~ n in terms of old ~ n): u u 1 2 1 1 +1 ~ n = I + hA + hA ]~ n .3) 2 or its equivalent ~ n = h I + hA ] I + hA ] . Both of these considerations are investigated later.4) Notice that Eqs.3 and 12.4 have the same formal order of accuracy and. h~ + O(h ) u u f +1 2 1 2 Finally. However. Here we seek only to apply to some simple cases the concept of factoring. Choose again the explicit Euler time march so that d~ = A ~ + A ~ + (bc) u ~ cu du dt (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. and techniques to carry out their numerical evaluation can have arithmetic operation counts that vary by orders of magnitude. their numerical stability can be quite di erent. h~ + O(h ) u u f 12. +1 1 2 2 ~ n = I + hA ] I + hA ]~ n . explicit Euler method.2 Factoring Physical Representations | Time Splitting Suppose we have a PDE that represents both the processes of convection and dissipation. from observation 3 (allowing us to drop higher order terms .6) 1 Second-order time-marching methods are considered later.218 CHAPTER 12.5) ~ ~ n = I + hAd + hAc]~ n + h(bc) + O(h ) u u +1 2 (12.2) where A = A + A ] but A and A are not unique. h~ + O(h ) u u f +1 1 2 2 1 2 2 1 2 (12. respectively and their sum forms the A matrix we have considered in the previous sections. Then. For the time march let us choose the simple. . From observation 1 (arbitrary splitting of A) : (12. rst-order. h A A i~ n .

10) The convection operator is applied explicitly. = I + hAd + h Ad + if h jjAdjj < 1. However.7 are often applied in a predictor-corrector sequence. hAd]~ n u +1 +1 ~ = I + hAc]~ n + h(bc) u = un ~ +1 (12. 2 We do not suggest that this particular method is suitable for use. the important aspect of stability has yet to be discussed. 12.8) Factoring can also be useful to form split combinations of implicit and explicit techniques. 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. this factored form would appear to be superior to that provided by Eq. For example. Therefore.2.7) u {z } | 2 2 | {z } un ~ ~n u +1 +1 = = ~ I + hAc]~ n + h(bc) u I + hAd]~n u +1 (12. In practical applications equations such as 12. 12. on this basis. and a rst-order time-march method is usually unsatisfactory. hAd].9) I . where jjAd jj is some norm of Ad ].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 . requiring a tridiagonal solver if the di usion term is central di erenced. as before. This time a predictor-corrector interpretation leads to the sequence un ~ I . 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. stability. another way to approximate Eq.8.7 and the original unfactored form Eq. I + hAc]~ n + h(bc) u ~ I + hAd + hAc]~ n + h(bc) +O(h ) u | {z } 1 2 (12. but the di usion operator is now implicit. hAd ]. 12.6 have identical orders of accuracy in the time approximation. Since numerical sti ness is generally much more severe for the di usion process.12. 12. We have yet to determine its . their selection is arbitrary.

3 32 @u = @ u + @ u @t @x @y 2 2 (12. 12. Let us study the basic concept of factoring by inspecting its use on the linear 2-D scalar PDE that models di usion: We begin by reducing this PDE to a coupled set of ODE's by di erencing the space derivatives and inspecting the resulting matrix operator. 12. the number of (interior) y points is 3.12) j Mx Mesh indexing in 2-D. hAd ]un = I + hAc]un + h(bc) +1 12. is 4 and My .12. 12.220 CHAPTER 12.and 3-D requires some reference to a mesh. 3 This could also be called a 5 .1 Mesh Indexing Convention 2 2 12.10. ~ I .3 Factoring Space Matrix Operators in 2{D Factoring is widely used in codes designed for the numerical solution of equations governing unsteady two. In this example Mx . A clear description of a matrix nite-di erence operator in 2.6 in the form Then which is identical to Eq.9 can be derived for a di erent point of view by writing Eq. We choose the 3 4 point mesh shown in the Sketch 12. 6 point mesh if the boundary points (labeled in the sketch) were included.11) My k 1 13 23 33 43 12 22 32 42 11 21 31 41 1 (12. SPLIT AND FACTORED FORMS un .and three-dimensional ows. The numbers 11 12 43 represent the location in the mesh of the dependent variable bearing that index.3. but in these notes we describe the size of a mesh by the number of interior points. the number of (interior) x points. un = A u + A u + (bc) + O(h ) ~ c n d n h +1 +1 2 We should mention here that Eq. Thus u represents the value of u at j = 3 and k = 2.

16 part a and b.3. and (2).16 part a and b. depending on the data{base chosen for the U it multiplies. the usual (but ambiguous) representation is given by Ax y . There are many ways to lay out a data-base. Examples of the order of indexing for these space vectors are given in Eq. Their structures are similar.14) Axx y = Pxy Axy y Pyx ( ) + ( ) + (12.13) where Pyx = PT = P. FACTORING SPACE MATRIX OPERATORS IN 2{D 221 12. consecutively along columns that are consecutive from j = 1 to Mx. Of these.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. and they are related by the same permutation matrix that relates U x to U y . (1) and (2) above are referred to as xvectors and y-vectors. 12. consecutively along rows that are themselves consecutive from k = 1 to My . In particular. 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.11 can be written + 4 If it is important to be speci c about the data-base structure.12. we label a data{base composed of x-vectors with U x . ( ) ( ) 12.3. u 4 Notice that . used in the previous sections. however. we use the notation Axx y or Axy y . When the matrix is multiplying a space vector U . are subsets of A and ~ . 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 .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). Thus ( ) + ( ) + ( ) ( ) dU = A U + (bc) ~ x y dt + (12. 12.15) Ax+y and U . 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. Examples are in Eq. To be speci c about the structure. 12. respectively. which are notations used in the special case of space vectors. Notice that the matrices are not the same although they represent the same derivative operation. In this notation the ODE form of Eq.3.

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. 12 22 32 42 13 23 33 43 11 21 31 41 . for both ! . Entries for x ! x. central-di erence.12. Ax y . + ( ) 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 .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 y ! o.. . Entries for x ! x.12.16) b: Elements in 2-dimensional. matrix operator.. Data base composed of My x{vectors stored in U x . 41 42 43 31 32 33 21 22 23 11 12 13 (12. for both ! . for y ! o. Ax y . . a:Elements in 2-dimensional.. for 3 4 mesh shown in Sketch 12. for 3 4 mesh shown in Sketch 12.. . Data base composed of Mx y{vectors stored in U y . central-di erence.

12.3. FACTORING SPACE MATRIX OPERATORS IN 2{D

223

12.3.4 Space Splitting and Factoring
We are now prepared to discuss splitting in two dimensions. It should be clear that the matrix Axx y can be split into two matrices such that
( ) +

Axx y = Axx + Ayx
( ) + ( ) ( ) ( )

( )

(12.17)

where Axx and Ayx are shown in Eq. 12.22. Similarily

Axy y = Axy + Ayy
( ) + ( )

( )

(12.18)

where the split matrices are shown in Eq. 12.23. The permutation relation also holds for the split matrices so

Ayx = Pxy Ayy Pyx
( ) ( )

and

Axx = Pxy Axy Pyx
( ) ( )

The splittings in Eqs. 12.17 and 12.18 can be combined with factoring in the manner described in Section 12.2. As an example ( rst-order in time), applying the implicit Euler method to Eq. 12.14 gives
h i ~ Unx = Unx + h Axx + Ayx Unx + h(bc)
( ) +1 ( ) ( ) ( ) ( ) +1

or
h i ~ I ; hAxx ; hAyx Unx = Unx + h(bc) + O(h )
( ) ( ) ( ) +1 ( ) 2

(12.19) (12.20)

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

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

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

h

i

(12.21)

224

CHAPTER 12. SPLIT AND FACTORED FORMS

Axx

( )

2 x 6x 6 6 6 6 6 6 6 6 6 6 6 6 (x) U =6 6 6 6 6 6 6 6 6 6 6 6 6 6 4

x x x x x x x x

j j j j j x x j x x x j x x x j x x

j j j j j j j j

Ayx

( )

6 6 6o 6 6 U (x) = 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4

j j j j 2 o j o 6 o j o 6 6 6 o j o 6 6 o j o 6 6 j o o j o o j o o j o j o j o j o j o
The splitting of Axx y .
( ) +

7 7 j o 7 7 j o 7 5 j o 7 j o

7 7 j o 7 7 j o 7 Ux 7 j o 7 7 j o7 7 7 7

j x x j x x x j x x j x x 3 j 7 j 7 7 7 j 7 7 j 7 7
7

3 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 5 x7

Ux

( )

(12.22)

( )

12.3. FACTORING SPACE MATRIX OPERATORS IN 2{D

225

Axy

( )

2 x 6 6 6 6 6 6 6 6x 6 6 6 6 6 6 U (y) = 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4

j x j x j x j x j x j

j j j

j x j x j x j x j x j x j x j x j j x j x j x j x j x j x j x j x j

Ayy

( )

6 6 6 6 6 6 6 6 6 (y ) U =6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4

j j j 2 o o j 6o o o j 6 6 o o j 6 6

j x j x j x j x j x j x 3 j j 7 j j 7 7 j j 7 7 j j j
7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 o5

3 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 x7 7 7 7 7 7 7 5

Uy

( )

(12.23)

j o o j j o o o j j o o j j j j j j j

j o o j j o o o j j o o j j j j
( ) +

Uy

( )

j o o j o o j o o

The splitting of Axy y .

226

12.4 Second-Order Factored Implicit Methods

CHAPTER 12. SPLIT AND FACTORED FORMS

Second-order accuracy in time can be maintained in a certain factored implicit methods. For example, apply the trapezoidal method to Eq. 12.14 where the derivative operators have been split as in Eq. 12.17 or 12.18. Let the data base be immaterial ~ and the (bc) be time invariant. There results 1 1 ~ I ; 2 hAx ; 1 hAy Un = I + 2 hAx + 1 hAy Un + h(bc) + O(h ) (12.24) 2 2 Factor both sides giving 1 I ; 1 hAx I ; 2 hAy ; 1 h AxAy Un 2 4 1 hA I + 1 hA ; 1 h A A U + h(bc) + O(h ) (12.25) ~ = I+ 2 x 2 y 4 x y n Then notice that the combination 1 h AxAy ](Un ; Un) is proportional to h since 4 the leading term in the expansion of (Un ; Un) is proportional to h. Therefore, we can write 1 ~ I ; 1 hAx I ; 1 hAy Un = I + 2 hAx I + 1 hAy Un + h(bc) + O(h )(12.26) 2 2 2 and both the factored and unfactored form of the trapezoidal method are second-order accurate in the time march. An alternative form of this kind of factorization is the classical ADI (alternating direction implicit) method usually written 1 1 ~ I ; 2 hAx U = I + 2 hAy Un + 1 hFn 2 1 hA U + 1 hF + O(h ) 1 hA U (12.27) I;2 y n = I+2 x ~ 2 n
+1 3 2 +1 2 3 2 +1 3 +1 +1 3 5 +1 +1 3

For idealized commuting systems the methods given by Eqs. 12.26 and 12.27 di er only in their evaluation of a time-dependent forcing term.

12.5 Importance of Factored Forms in 2 and 3 Dimensions
When the time-march equations are sti and implicit methods are required to permit reasonably large time steps, the use of factored forms becomes a very valuable tool
5 A form of the Douglas or Peaceman-Rachford methods.

12.5. IMPORTANCE OF FACTORED FORMS IN 2 AND 3 DIMENSIONS 227
for realistic problems. Consider, for example, the problem of computing the time advance in the unfactored form of the trapezoidal method given by Eq. 12.24 1 ~ I ; 2 hAx y Un = I + 1 hAx y Un + h(bc) 2 Forming the right hand side poses no problem, but nding Un requires the solution of a sparse, but very large, set of coupled simultaneous equations having the matrix form shown in Eq. 12.16 part a and b. Furthermore, in real cases involving the Euler or Navier-Stokes equations, each symbol (o x ) represents a 4 4 block matrix with entries that depend on the pressure, density and velocity eld. Suppose we were to solve the equations directly. The forward sweep of a simple Gaussian elimination lls all of the 4 4 blocks between the main and outermost diagonal (e.g. between and o in Eq. 12.16 part b.). This must be stored in computer memory to be used to nd the nal solution in the backward sweep. If Ne represents the order of the small block matrix (4 in the 2-D Euler case), the approximate memory requirement is (Ne My ) (Ne My ) Mx oating point words. Here it is assumed that My < Mx. If My > Mx, My and Mx would be interchanged. A moderate mesh of 60 200 points would require over 11 million words to nd the solution. Actually current computer power is able to cope rather easily with storage requirements of this order of magnitude. With computing speeds of over one giga op, direct solvers may become useful for nding steady-state solutions of practical problems in two dimensions. However, a three-dimensional solver would require a memory of approximatly
+ +1 + +1 6 7 8

words and, for well resolved ow elds, this probably exceeds memory availability for some time to come. On the other hand, consider computing a solution using the factored implicit equation 12.25. Again computing the right hand side poses no problem. Accumulate the result of such a computation in the array (RHS ). One can then write the remaining terms in the two-step predictor-corrector form 1 ~ I ; 2 hAxx U x = (RHS ) x ~ I ; 1 hAyy Uny = U y (12.28) 2
( ) ( ) ( ) ( ) ( ) +1 ( )

Ne My Mz Mx
2 2 2

practical size the ll is mostly dense. 7 The lower band is also computed but does not have to be saved unless the solution is to be repeated for another vector. 8 One billion oating-point operations per second.

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

23 shows that the nal ~ be permuted to U step consists of solving Mx one-dimensional implicit problems each with dimension My .14.29) 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.6 The Delta Form Clearly many ways can be devised to split the matrices and generate factored forms. our original ODE. 12. 2 hAx .29 and maintain O(h ) accuracy. SPLIT AND FACTORED FORMS 9 which has the same appearance as Eq. 2nd-order. we see that this is equivalent to solving My one-dimensional ~ problems. ( ) ( ) 12. 2 hAx . 12. 2 hAx . is referred to as the \delta form" and we develop it next. Consider the unfactored form of the trapezoidal method given by Eq. 9 A block tridiagonal solver is similar to a scalar solver except that small block matrix operations .30) I . 12. Thus. 2-D equation. 12.21 but is second-order time accurate. Inspecting the top of Eq. Then. We arrive at the expression h 1 ~ i (12. 2 hAy Un = h Ax y Un + (bc) + O(h ) (12. each with Mx blocks of order Ne . it is guaranteed to converge to the correct steady-state solution of the ODE. Un one nds h 1 1 ~ i I . Now we can factor Eq. +1 3 +1 + 3 2 + 3 replace the scalar ones.22. 1 hAy Un = h Ax y Un + (bc) + O(h ) 2 This is the delta form of a factored. using the standard de nition of the di erence operator . 12.228 CHAPTER 12. if Eq.29 converges. The rst step would be solved using My uncoupled block tridiagonal solvers . 1 hAy Un 2 leaving the equality unchanged.24. 2 hAy Un = I + 2 hAx + 1 hAy Un + h(bc) + O(h ) 2 From both sides subtract 1 I . One way that is especially useful. and matrix multiplications do not commute. 12. 2 hAx I . for ensuring a correct steady-state solution in a converged time-march. Un = Un . and ~ let the (bc) be time invariant: 1 1 1 ~ I . The temporary solution U x would then y and an inspection of the bottom of Eq. 12.

ii.31 are vastly di erent. and if it is immediately factored.30 and the O(h) method given by Eq. 12. the delta form of the unfactored Eq. (1) (2) 12 21 x 2 j . 12.19. 12. write the space vector.31) In spite of the similarities in derivation.P ). the factored form is presented in Eq. Assume that second order central di erencing is used. What is the space vector for the natural ordering (monotonically increasing in index). hAx . hAy ] Un = h Ax y Un + (bc) + 10 and its factored form becomes h ~ i I . the unfactored form of a rst-order method derived from the implicit Euler time march is given by Eq. hAx] I . 12.(P . Write down the permutation matrices. u ? Only include the interior points. (a) Space vector de nition i.. u ? iii. 2u + u ) xx j The uniform grid has x = 1 and 8 interior points. 2 .1 j j +1 10 Notice that the only di erence between the O(h2 ) method given by Eq.19 is h ~ i I .7. On the other hand. i. PROBLEMS 229 The point at which the factoring is made may not a ect the order of time-accuracy.7 Problems 1. but it can have a profound e ect on the stability and convergence properties of a method. so that we can simplify the boundary conditions. ( u) = 1 (u .20. 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. 12. hAy ] Un = h Ax y Un + (bc) + (12. 12.12. we will see in the next chapter that the convergence properties of Eq.31 is the appearance of the factor 1 on the left side of the O(h2 ) method. 12.20 and Eq. 12.e. For example. If we reorder the points with the odd points rst and then the even points.

(b) System de nition In problem 1a. We can put such a partitioning to use. SPLIT AND FACTORED FORMS iv. Comment on the form of the resulting implicit matrix operator. and P which partition the odd points from the even points. 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 . Applying implicit Euler time marching. 1 2 (2) 2 (2) 2 2 .230 CHAPTER 12. The generic ODE representing the discrete form of the heat equation is (1) 1 (1) v. write the delta form of the implicit algorithm. we de ned u u A A P . (Note f = 0. 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 .

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

232 CHAPTER 12. SPLIT AND FACTORED FORMS .

233 . 1 See also the discussion on Fourier stability analysis in Section 7. 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. if the results indicate that a method has an instability. We have seen that under these conditions a semi-discrete approach can lead to circulant matrix di erence operators. and (approximately) factored. the method is probably not suitable for practical use. and we discussed circulant eigensystems1 in Section 4.7.4 we introduced the concept of the representative equation. and convergence properties of timemarching schemes.Chapter 13 LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS In Section 4. 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. hybrid. the results found from this analysis are neither necessary nor su cient to guarantee stability. 13.3. 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. fractional-step. However. When these methods are applied to practical problems. and used it in Chapter 7 to study the stability. accuracy.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.

. gM (t) (13. gm(t) .234 CHAPTER 13.4) This is to be contrasted to the developments found later in the analysis of 2-D equations.4: b since they must share a common The representative equation for split.3) that 13. g1 (t) . This suggests (see Section 4. 0 wM = ( a + b)M wM .. and only one2 .. eigenvector. Since both matrices have the same set of eigenvectors. . as a result of space di erencing the PDE.2 Example Analysis of Circulant Systems 13.2. 0 wm = ( a + b)m wm .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. ~ (t) u (13.2. 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. we arrive at a set of ODE's that can be written d~ = A ~ + A ~ .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. we can use the arguments made in Section 4.. (13.1) apu b pu f dt where the subscript p denotes a circulant matrix.3 to uncouple the set and form the M set of independent equations 0 w1 = ( a + b)1 w1 . LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS Suppose.

1 . 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 1.2): ( c)m = ia sin m x m (13. the explicit-explicit Euler method.3.1 0 1)~ + u u B (1 . 4 2 sin2 2 x In these equations m = 2m =M . where (see Section 4. Using these values and the representative equation 13. a B (.5) dt 2 x p x2 p the convection matrix operator and the di usion matrix operator. EXAMPLE ANALYSIS OF CIRCULANT SYSTEMS 235 If the space di erencing takes the form d~ = . (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. m = 0 1 M . the explicit-implicit Euler method. 12.2 1)~ u (13. 12. respectively. Eq.2. When applied to Eq. The Explicit-Implicit Method 2 (13. 2. In this section we refer to these as 1.7) . Eq. the characteristic polynomial of this method is P (E ) = (1 .4. 13. 13.13.10.2.6) ( d)m = . can be represented by the eigenvalues c and d.8.4. we can analyze the stability of the two forms of simple time-splitting discussed in Section 12. h d )E . so that 0 m 2 .6 to quantify the eigenvalues.

and the . one near 0.8 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 ).4 0 0 4 R∆ Explicit-Implicit 8 0 0 4 R∆ Explicit-Explicit 8 _ Figure 13. 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. The Explicit-Explicit Method An analysis similar to the one given above shows that this method produces " # q 2 m 1 . A simple numerical parametric study of Eq.2 C = R /2 n ∆ C = 2/ R n ∆ 0. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 1.4 0. which yields the same result as in the previous example. 2.8 C n C 0.1: Stability regions for two simple time-split methods. 13. 13.236 CHAPTER 13.1. 4 Cn sin2 m 2 j j = 1 + Cn sin 0 m 2 R 2 Again a simple numerical parametric study shows that this has two critical ranges of m .2 C = 2/ R ∆ n 0.

The di usion term is integrated implicitly using the trapezoidal method.4 which we split in the fashion of Eq. factored method has a much smaller region of stability when R is small. (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. The convection term is treated explicitly using the second-order Adams-Bashforth method.9) Q(E ) = 1 h(E 2 + E ) 2 13.1) + 2 h d (un+1 + un) This produces 3 P (E ) = (1 .1. This method applies to a ow in which there is a combination of di usion and periodic convection.5. Next let us analyze a more practical method that has been used in serious computational analysis of turbulent ows. P (E ) and Q(E ). time-marching method on the equation u0 = cu + du + ae t First let us nd expressions for the two polynomials. 1) (13.8) and in the latter (13. un. as well as the stability. we include the forcing function in the representative equation and study the e ect of our hybrid.2. The totaly explicit. which produces the constraint that 1 Cn < 2 R for R 2 The resulting stability boundary is also shown in Fig. EXAMPLE ANALYSIS OF CIRCULANT SYSTEMS 237 other near 180o. 13. 1 h d)E 2 . The characteristic polynomial follows from the application of the method to the homogeneous equation. thus 1 1 un+1 = un + 2 h c(3un . In order to evaluate the accuracy. 13.2. In the former case it is 1 Q(E ) = 2 h(3E . Our model equation is again the linear convection-di usion equation 13. as we should have expected.2 Analysis of a Second-Order Time-Split Method .13.

10 and shown in Fig.1.11) The extension of the following to 3-D is simple and straightforward. one can show er = O(h3) using either Eq.1. 2 d . 13. .2. 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. 2h c 1 . 13.9. so er = O(h3) Using P (e h) and Q(e h ) to evaluate er in Section 6. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS Accuracy From the characteristic polynomial we see that there are two {roots and they are given by the equation 3 1 + 2h c + 1h 2 = s d The principal -root follows from the plus sign and one can show 1 1 1 = 1 + ( c + d )h + 2 ( c + d )2 h2 + 4 3 + c 2 . 2 h d (13.238 CHAPTER 13. 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.2. 1 1 + 3h c + 2h 2 2 1 . 1h d 2 d 2 1 . the hybrid method retains the second-order accuracy of its individual components. These results show that.7. for the model equation. The results are plotted in Fig. 13.3. 13. 13.10) Stability The stability of the method can be found from Eq. 13. 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 .10 by a parametric study of cn and R de ned in Eq.8 or Eq.3 The Representative Equation for Space-Split Operators Consider the 2-D model3 equations @u = @ 2 u + @ 2 u @t @x2 @y2 3 (13.

The form of the Ax and Ay matrices for three-point central di erencing schemes are shown in Eqs. and @u + a @u + a @u = 0 (13.13) x y dt for the space vector U . 12.2 0.4 Stable 0 4 8 ∆ R _ Figure 13.23 for the 3 4 mesh shown in Sketch 12.3. THE REPRESENTATIVE EQUATION FOR SPACE-SPLIT OPERATORS239 1. Now nd the block eigenvector .8 Cn 0. to the coupled set of ODE's: dU = A + A ]U + (bc) ~ (13.2: Stability regions for the second-order time-split method. by means of spatial di erencing approximations.12) @t x @x y @y Reduce either of these. 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.13.12.1 I ~0 I B b b where B is a banded matrix of the form B (b.1 I ~0 I ~1 I 7 b b 5 4 5 4b ~. First we write these matrices in the form 2 3 2 ~ 3 B b0 I ~1 I b A(xx) = 6 B 7 A(yx) = 6 ~.1 b0 b1 ).22 and 12.

the rst matrix on the right side of Eq. That is. b b b ~ that diagonalize the B blocks.1 I ~0 I ~1 I 7 b b 5 b b 5 4b ~.1 I ~0 I ~.1B X = 6 7 4 5 4 5 ~3 ~ This time. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS matrix that diagonalizes B and use it to diagonalize A(xx).1 6 ~. by the same argument as before. 12. if we Notice that the matrix Ay 3 7 7 7 5 3 2 ~ 3 ~1 I b b0 I ~1 I b ~0 I ~1 I 7 X = 6 ~.1 ~0 ~1). 13. see the bottom of Eq.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.14 is transparent to the transformation.15) 6 7 4 5 3I + ~ 4I + ~ diag(X ) 2 ~ b0 X .240 CHAPTER 13. 12. 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 4b I I . so the nal result is the complete diagonalization of the matrix Ax+y : h ih ih i ~ ~ X .1 Pyx X .1 A(xx+)y X Pxy X = 2 3 1I + ~ 6 7 6 7 2I + ~ 6 7 (13.13.1 B 76 X 4 54 54 5 4 .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. There results h i Pyx X . Thus 2 .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.23.14) ~ 4 5 4 B 7 3 I 5 ~ B 4 I ~ where B is the banded tridiagonal matrix B (~.1 32 32 3 2 X B X 6 76 7=6 X .

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 . First reduce the PDE to ODE by some choice4 of space di erencing. by itself. this fact. 13. THE REPRESENTATIVE EQUATION FOR SPACE-SPLIT OPERATORS241 It is important to notice that: The diagonal matrix on the right side of Eq. Now we are ready to present the representative equation for two dimensional systems. Often we are interested in nding the value of. and where Ax and Ay satisfy the conditions in 13.15.3. 13.15 contains every possible com~ bination of the individual eigenvalues of B and B . Although Ax and Ay do commute. + (13. This results in a spatially split A matrix formed from the subsets A(xx) = diag(B ) ~ A(yy) = diag(B ) (13. To obtain the latter property the structure of ~ the matrices is important. In that case we set = 0 and use the simpler form du = + ]u + a (13.16) ~ where B and B are any two matrices that have linearly independent eigenvectors (this puts some constraints on the choice of di erencing schemes). does not ensure the property of \all possible combinations". and its use is not necessary to arrive at Eq. 4 .13.18) x y We have used 3-point central di erencing in our example. the steady-state solution of the representative equation.16. and the convergence rate to. a and are (possibly complex) constants.17) x y dt which has the exact solution a u(t) = ce( x+ y )t . but this choice was for convenience only.

steady-state solution. The form of the method is given by Eq. h y )un+1 = un + ha P (E ) = (1 . h from which x . h . h y )E . 2. The accuracy of the xed.19. x y Like its counterpart in the 1-D case. The convergence rate to reach the steady-state. In the following we analyze four di erent methods for nding a xed.4.18) steady-state solution (of the ODE) for any h. h x . rst-order scheme which can then be used as a reference case for comparison with the various factored ones. 13. Converges very rapidly to the steady-state when h is large.16. 3.5. Unfortunately.4 Example Analysis of 2-D Model Equations 1. Is unconditionally stable. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 13. 3. In each case we examine 13. The stability. .19) un = c 1 . this method: 1. we nd (1 . 12. use of this method for 2-D problems is generally impractical for reasons discussed in Section 12. 1 Q(E ) = h giving the solution (13. and when it is applied to the representative equation.1 The Unfactored Implicit Euler Method Consider rst this unfactored. steady-state solution to the 2-D representative equation 13. however.242 CHAPTER 13. h . " 1 #n a x+ y 2. Produces the exact (see Eq.

However. 3. + a h x y x y.3 The Factored Delta Form of the Implicit Euler Method . 12. x y We see that this method: 1.4. h y )un+1 = 1 + h2 x y un + ha and this has the solution " #n 1 + h2 x y a un = c (1 . The method requires far less storage then the unfactored form. it is not very useful since its transient solution is only rst-order accurate and. Next apply Eq. but the converged solution is completely wrong. 2. h y )un+1 = un + ha from which P (E ) = (1 .20 to the 2-D representative equation. h ) . h x)(1 . h y )E . + x y x y This method: 13.13. h x)(1 .4. h x)(1 . h ) . h )(1 . There results (1 . h x)(1 .4. h y )(un+1 . EXAMPLE ANALYSIS OF 2-D MODEL EQUATIONS 243 Now apply the factored Euler method given by Eq. un) = h( xun + y un + a) which reduces to (1 .20) giving the solution " #n 1 un = c (1 . Produces a steady state solution that depends on the choice of h.2 The Factored Nondelta Form of the Implicit Euler Method 13.31 to the 2-D representative equation. 1 Q(E ) = h (13. Converges rapidly to a steady-state for large h. h )(1 . the converged value is meaningless. if one tries to take advantage of its rapid convergence rate. 12. One nds (1 . Is unconditionally stable.

and the factored delta form of the implicit Euler method can be used when a converged steady-state is all that is required.30 to the representative equation and one nds 1 1 . Produces the exact steady{state solution for any choice of h. 12. as discussed in Section 12. since j j ! 1 as h ! 1. Produces the exact steady-state solution for any choice of h. Finally consider the delta form of a second-order time-accurate method.4. The correct steady solution is obtained. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 1. 1 h y un+1 = 1 + 1 h x 1 + 1 h y un + ha 2 2 2 2 and this has the solution 2 3n 1 + 1h x 1 + 1h y 7 6 a 2 2 7 . 12. 2 h x 1 . but convergence is not nearly as rapid as that of the unfactored form. this method demands far less storage than the unfactored form. Converges very slowly to the steady{state solution for large values of h.27 should be enough to convince the reader that the 's produced by those methods are identical to the produced by this method. Since it is second order in time.4 The Factored Delta Form of the Trapezoidal Method . 1 h x 1 . 2. since j j ! 1 as h ! 1. 2. un) = h( xun + y un + a) 2 which reduces to 1 . un = c6 4 5 1 x+ y 1 . 3. 2h y 2 This method: 1. Converges very slowly to the steady{state solution for large values of h. Apply Eq.244 CHAPTER 13. All of these properties are identical to those found for the factored delta form of the implicit Euler method. it can be used when time accuracy is desired. Is unconditionally stable.26 and 12. the value of h on the left side of the implicit equation is literally switched from h to 1 h. Like the factored nondelta form. 1h x 1 . 2 5 13.5 A brief inspection of eqs. 1 h y (un+1 . In practical codes. 3. Is unconditionally stable.5.

2 h x 1 . delta form using this equation. and write the solution either in the form 2 3n 1 h( + + ) + 1 h2 ( 1 h3 + + ).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).11 and 13.22. . 2 x y z 4 x y + x z + y z) . 2 h( x + y + z ) un+1 = 1 + 2 h( x + y + z ) un + ha Put this in delta form: 1 .3 generalize to three dimensions and. 1 h z un = h ( x + y + z )un + a] (13. under the conditions given in 13. 1 h( x + y + z ) un = h ( x + y + z )un + a] 2 Now factor the left side: 1 1 . each with an additional term. EXAMPLE ANALYSIS OF THE 3-D MODEL EQUATION 13. 8 x y z . 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 .12. 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 .5. 13.21) x y z dt Let us analyze a 2nd-order accurate.13. 13. +a+ x y 6 z (13. the model 3-D cases6 have the following representative equation (with = 0): du = + + ]u + a (13. factored.5 Example Analysis of the 3-D Model Equation 245 The arguments in Section 13.16 with an A(zz) included. One can derive the characteristic polynomial for Eq.23) Eqs. nd the root. 1 h y 1 .

and are both positive. 5 x+ y+ z a (13. 7 However. 13. +i i . Furthermore. Now we can always nd one combination of the 's for which . )2 >1 and the method is unconditionally unstable for the model convection problem. )2 + ( . if all the 's are real and negative. In that case since the absolute value of the product is the product of the absolute values j 2 2 2 = (1 . 13.24) It is interesting to notice that a Taylor series expansion of Eq.12 with periodic boundary conditions. From the above analysis one would come to the conclusion that the method represented by Eq. 13. 13.25) h i + 1 h3 3 + (2 y + 2 x) + 2 2 + 3 x y + 2 2 + 3 + 2 x 2 + 2 2 y + 3 + y y y y x x 4 z which veri es the second order accuracy of the factored form. This makes the method unconditionally stable for the 3-D di usion model when it is centrally di erenced in space.i . ) + ( + ) j (1 . and are real numbers that can have any sign. Now consider what happens when we apply this method to the biconvection model. First write the root in Eq. if the method converges.23 that. central di erencing causes all of the 's to be imaginary with spectrums that include both positive and negative values. In this case. 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 y 1 .24 results in 1 = 1 + h( x + y + z ) + 2 h2( x + y + z )2 (13. . LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS or in the form 2 6 un = c6 4 1 1 1 + 1 h x 1 + 2 h y 1 + 1 h z . it follows from Eq. 4 h3 2 2 1h 1h 1 . the 3-D form of Eq. 13.246 CHAPTER 13. 1h z 2 3n x y z7 7 . however.7 With regards to stability. clearly.22 should not be used for the 3-D Euler equations. Remember that in our analysis we must consider every possible combination of these eigenvalues. in the presence of this dissipation. 2 x 1 . it converges to the proper steady-state. +i where . the method is stable for all h.23 in the form 1+ = 1. we already knew this because we chose the delta form. the instability disclosed above is too weak to cause trouble. In practical cases.

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

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

. A. A general m m matrix A can be written 2 a11 a12 a1m 3 6a a a 7 A = (aij ) = 6 21 22 . . Given below is some notation and some of the important relations between matrices and vectors. Thus 2 v1 3 6 7 ~ = 6 v.1 Notation 1.2 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. In the present context a vector is a vertical column or string. .Appendix A USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA A basic understanding of the fundamentals of linear algebra is crucial to our development of numerical methods and it is assumed that the reader is at least familar with this subject area. 2m 7 7 6 5 4 am1 am2 amm 249 .

det A] is the determinant of A. A is skew-symmetric or antisymmetric if AT = . A is orthogonal if aij are real and AT A = AAT = I 5. If b A= a d c then det A] = ad . AH is the conjugate transpose of A. 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 .b a . 2. A is normal if AT A = AAT .2 De nitions 1.A. 9. (Hermitian). A is symmetric if AT = A. 11. A is the complex conjugate of A.. 6.1 = det1 A] . 7 4 T5 ~m a 4.c . The trace of a matrix is Pi aii. v v 7.250APPENDIX A. USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA 3.1 = I .2 7 6 7 6 . 10. 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. P is a permutation matrix if P ~ is a simple reordering of ~ . The inverse of a matrix (if it exists) is written A. A. where I is the identity matrix.1 and has the property that A. 4. bc and d A. 3. 8.1A = AA.

Transpose equalities: (A + B )T = AT + B T (AT )T = A (AB )T = B T AT 6.1). A + (B + C ) = (C + A) + B . Inverse equalities (if the inverse exists): (A. The eigenvalue problem for a matrix A is de ned as A~ = ~ or A . Thus: 1 A = 2 A + AT + 1 A . If it is asymmetric.1 = (A.1 A. they are all imaginary.1 )T 251 7. 2.3. . 1. If a square matrix with real elements is symmetric.4 Eigensystems 1. In general AB 6= BA. ALGEBRA A.A.1 = B . I ]~ = 0 x x x and the generalized eigenvalue problem.3 Algebra We consider only square matrices of the same dimension. A and B are equal if aij = bij for all i j = 1 2 : : : m. B ]~ = 0 x x x 2. AT 2 A.1 = A (AB ). 4. Any matrix A can be expressed as the sum of a symmetric and a skew-symmetric matrix. etc. sA = (saij ) where s is a scalar. 3.1 (AT ). its eigenvalues are all real. 5. including the matrix B . as A~ = B~ or A .

and this eigenvector cannot be formed from a linear combination of any of the other eigenvectors. If the eigenvalues of a matrix are not distinct. 6.. 7 6 . h i X = ~1 ~2 : : : ~m x x x and is the diagonal matrix 0 0 m If A can be diagonalized..e. . . its eigenvectors completely span the space. . . 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. the eigenvalues of a matrix may not be distinct. X . . . If A has m distinct eigenvalues.. . If a matrix does not have a complete set of linearly independent eigenvectors. 0 7 4 5 . an m m matrix A has n~ linearly independent eigenvectors with x n~ m and n distinct eigenvalues ( i) with n n~ m. Gershgorin's theorem: The eigenvalues of a matrix lie in the complex plane in the union of circles having centers located by the diagonals with radii equal to the sum of the absolute values of the corresponding o -diagonal row elements. If A posseses m linearly independent eigenvectors then A is diagonalizable. In general. 4. USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA 3.. and A is said to have a complete eigensystem. but it can still be diagonalized. 2 0 03 1 6 . i. 7 6 7 = 6 0 .2 . In general. . and with each distinct eigenvalue there is one associated eigenvector. 9. then A is always diagonalizable. it cannot be diagonalized. in which case the possibility exists that it cannot be diagonalized.252APPENDIX A. x x 5. 7. The eigenvectors of such a matrix cannot span the space and the matrix is said to have a defective eigensystem.1AX = where X is a matrix whose columns are the eigenvectors. the matrix is said to be derogatory. 8. but all of the eigenvectors are linearly independent.

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

2. 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. having: 9 eigenvalues 3 distinct eigenvalues 3 Jordan blocks 5 linearly independent eigenvectors 3 principal vectors with 1 1 principal vector with 2 A.5 Vector and Matrix Norms 1. A p-norm of a matrix A is de ned as Av jjAjjp = max jjjjvjjjjp x6=0 p . A p-norm of the vector ~ is de ned as v 0M 1 X pA1=p jjvjjp = @ jvj j j =1 3. 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.

In general (A) does not satisfy the conditions in 4. 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.5. VECTOR AND MATRIX NORMS 255 4. in fact. Let A and B be square matrices of the same order. 8. is the lower bound of all the norms of A. in this case it is the L2 norm. (A). The spectral radius of A. When A is normal. 7. (A) is a true norm. so in general (A) is not a true norm. jjAjj2 = q PM ja j maximum column sum i=1 ij (A T A ) . 6.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.

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

1 ~ m = (xj )m = a 2 sin j m x m=1 2 M (B.. Many of these can be derived by solving the simple linear di erence equations that arise in deriving recursion relations. Let us consider a simple tridiagonal matrix.e.1 Standard Eigensystem for Simple Tridiagonals In this work tridiagonal banded matrices are prevalent.2) and is given by j.Appendix B SOME PROPERTIES OF TRIDIAGONAL MATRICES B. i. we nd (by a recursion exercise) det B (M : a b c)] = 0 if p m b + 2 ac cos M + 1 = 0 m=1 2 M From this it follows at once that the eigenvalues of B (a b c) are p m m=1 2 M (B.1) m = b + 2 ac cos M +1 The 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. see Section 3.3) c M +1 257 . a tridiagonal with constant scalar elements a. It is useful to list some of their properties. If we examine the conditions under which the determinant of this matrix is zero. and c.4.b.

m e + 2 (a .1 a 2 = ei j. SOME PROPERTIES OF TRIDIAGONAL MATRICES These vectors are the columns of the right-hand eigenvector matrix.i m. 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. . . f ] = 0 if q m b . . d b .258 APPENDIX B. md)(c . c 7 6 .1 a 2 sin jm j=1 2 M X = (xjm) = c (B. f 7 6 .7) M (B. 2 (B..6) B. c m 2 1 = e.5) c ( 1) The left-hand eigenvector matrix of B (a b c) can be written .4) m=1 2 M M +1 Notice that if a = . = 2 X j=1 2 M +1 a M +1 In this case notice that if a = .8) . . e c ..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 . j.1 and c = 1... 7 6 .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 . the elements of which are j. c m 2 1 sin mj m=1 2 . 2 a 1 ( 1) M M (B.

2. 4fdpm The right-hand eigenvectors are #j .10) Eq. THE INVERSE OF A SIMPLE TRIDIAGONAL eb . 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 . m d 2 sin j ] m =1 2 M ~m = x m j =1 2 M c. 4ac = 0. Let DM represent the determinant of B (M : a b c) DM det B (M : a b c)] De ning D to be 1.11) where we have made use of the initial conditions D = 1 and D = b. b . B.10 is a linear O E the solution of which was discussed in Section 4. bE + ac) and the two roots to P ( ) = 0 result in the solution 8" 9 p p " # < b + b .B. ac D = b . fb)(ea . 4ac #M 1 b . thus D = 1 D = b D = b . it is simple to derive the rst few determinants. bd) + (cd .3 The Inverse of a Simple Tridiagonal The inverse of B (a b c) can also be written in analytic form. . acDM . 4ac : M =0 1 2 (B. In the limiting case when b . Its characteristic polynomial P (E ) is P (E .9) One can also nd the recursion relation DM = bDM . 1 " a . af )pm ] m= e . 2(cd + af )pm + 2pm (ec .3. B. 2abc (B. 4ac M = DM = p . 2 2 b . f 2 2 2 259 2 q m These relations are useful in studying relaxation methods. (B.

cD D c D a D .c D .n(.a)m.4) tridiagonal matrix Bp(M : a b c ) (B. DM . (.aD D D .c D c D D D . SOME PROPERTIES OF TRIDIAGONAL MATRICES 2 16 6 4 D4 6 Then for M = 4 B.c D .1 dmn = DM .a D a D .m =DM Diagonal: n=m=1 2 M dmm = DM .cD 3 7 17 7 2 7 7 5 3 Upper triangle: m=1 2 M .cD cD . DM .12) .aD D 3 2 2 3 1 3 1 2 2 1 4 1 1 3 0 2 1 2 2 2 2 1 2 1 3 1 1 2 2 1 3 1 2 3 4 .aD D D D .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 .m =DM 1 Lower triangle: m =n+1 n+2 M 1 n=1 2 M .aD D D D .mDn.cD .aD3 1 6 a2 D 6 2 6 D5 6 .1 = aD The general element dmn is 4 0 2 D4 6 .1 Standard Tridiagonals Consider the circulant (see Section 3. = 1 D .n=DM B.cD D c D a D D .4 Eigensystems of Circulant Matrices B.a3 D1 4 cD .4.a D aD .aD D D D .cD .c)n.cD D c D D .260 APPENDIX B.1 1 n=m+1 m+2 M dmn = Dm.4.

1 . 1 1 1 M . = M X . B. and the right-hand eigenvector matrix has the form 2 m j =0 1 M . i(a . B.1 M 1 Note that both X and X . are symmetric and that X .B. B. In fact.14. 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.2 General Circulant Systems Notice the remarkable fact that the elements of the eigenvector matrices X and X . Although the eigenvectors of a circulant matrix are independent of its elements. = (x0 ) = 1 e. all circulant matrices of order M have the same set of linearly independent eigenvectors.13) B. B.1 (B.1 (B. EIGENSYSTEMS OF CIRCULANT MATRICES The eigenvalues are m 261 = b + (a + c) cos 2 m .14) p where i .4.1 X = (xjm) = eij M m =0 1 M . For the element indexing shown in eq. where X is the conjugate transpose of X .im M X mj j =0 1 M . X bj ei jm=M = m 1 0 3 2 1 1 0 3 2 2 1 0 3 3 2 1 0 1 (2 ) of which eq.15 they have the general form M. the eigenvalues are not. and further examination shows that the elements in these eigenvectors correspond to the elements in a complex harmonic analysis or complex discrete Fourier series. j =0 .1.15) 4 b b b b 5 b b b b The eigenvectors are always given by eq.4.12 do not depend on the elements a b c in the matrix. for the tridiagonal circulant matrix given by eq. 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 .13 is a special case.1 The left-hand eigenvector matrix with elements x0 is 2 j m =0 1 M . even if they are completely dense.

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

1 { Symmetrical folds for special cases m=1 2 M ~ m = sin j (2m .1. 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. 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 .17 are (2m .6. 1) x 2M ! j=1 2 M B. An even-numbered mesh Line of Symmetry x =0 x= j0 = 1 2 3 4 5 M0 j= 1 2 3 M b. 6 7 a a B (M : ~ b a) = 6 . see Fig.B. An odd{numbered mesh Figure B.. one can show from previous results that the eigenvalues of eq. leads to the matrix 2 3 b a 6a b a 7 6 7 6 7 . B... combined with special conditions imposed at the boundaries. 1 : a b a) about the center.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 . B. 2 2 ..

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

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

1 This being the case. the constant term drops out of eq. Qn) + O(h2) can be written (C. under this condition.266APPENDIX C. according to eq. by direct use of eq. we notice from the chain rule that for any vectors F and Q @F (Q) = @F @Q = A @Q @x @Q @x @x We notice also that for a homogeneous F of degree 1. E = En + An(Q . where is the internal energy per unit mass. if F is homogeneous of degree 1. 2. THE HOMOGENEOUS PROPERTY OF THE EULER EQUATIONS Now it is easy to show. The Euler equations are homogeneous if the equation of state can be written in the form p = f ( ). 6. BnQn are identically zero for homogeneous vectors of degree 1.5) E = An Q + O(h2) F = BnQ + O(h2) (C.7) (C. that both E and F in eqs. AnQn and Fn . F = AQ and @F = A @Q + @A Q @x @x @x Therefore.11 and 2.4. . " # " # (C. and their Jacobians. 6. see eq. 1 " # (C.105 can be written in general as. Qn) + O(h2) F = Fn + Bn(Q .9) Note that this depends on the form of the equation of state. Notice also that. are homogeneous of degree 0 (actually the latter is a direct consequence of the former). C.8) @A Q = 0 @x in spite of the fact that individually @A=@x] and Q are not equal to zero. A and B .3. As a nal remark.6) since the terms En . we notice that the expansion of the ux vector in the vicinity of tn which.12 are homogeneous of degree 1.106. C.

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

Sign up to vote on this title
UsefulNot useful