Fundamentals of Computational Fluid Dynamics

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

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

1

1 2 2 3 3 4 4 5

2 CONSERVATION LAWS AND THE MODEL EQUATIONS

7 8 12 12 13 14 14 15 16 17

7

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

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

21

21 24 25 25 25 29 30

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

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

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

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

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

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

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

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

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

4 THE SEMI-DISCRETE APPROACH

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

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

5 FINITE-VOLUME METHODS

71

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

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

6 TIME-MARCHING METHODS FOR ODE'S

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

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

85

7 STABILITY OF LINEAR SYSTEMS

121

122 123 123 123

. . . . . . . . . . . . . . . . . . . . . .1 The Criterion . 8. . . . . . . . . . . . . . . . . . . . . . . . . . .1 Sti ness De nition for ODE's . . . . . . . . . . . . . . . . . . . . . . .6 Steady Problems . . . . . . . . . . . . .7. . . . . . . . . . . 8. . . . . . . .1 Relation to -Eigenvalues . . . . . . . . . . . . . . . . . . . . . . . . . . . .5. . . . 8. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 7. . . . 8. . . . . . . . . . . .3 Defective Eigensystems . . Problems . . . . . . . . . . . . . . . .3. . . . . .7. . . . . . . . . . . . 8. 149 149 149 151 151 152 153 154 154 154 155 158 158 159 160 160 161 . . . .4. . . . . . 8. . . . . . . . . . . . . . . .5. . . . . . 7. . . . .3. . . . . . . 7. . . . . . . . .3 Stability Contours in the Complex h Plane. . . . . . . . . .7 Problems . . . . . . Consistency . .3 7. . . . . . . . 8. . . . . . . . . . .1 Stability for Large h. . . . . . . . . . . . . . . Numerical Stability Concepts in the Complex -Plane . . . . . . . . . .7. . . . . .7 7. . . . . 7. 8. . . . . . . . . . . . . . . . . . . Numerical Stability Concepts in the Complex h Plane 7. .4.3 An Example Involving Periodic Convection . . . . . . . . . . . . . .6. . . . . . . 8. . . . . . . . . .1. .5 Coping With Sti ness .2 Some Examples . . . . . . . . . . Fourier Stability Analysis . 7. . 7. . . .8 7. . . . . . . 8.3 Relation to Circulant Matrices . . . . . . . . . . . . . . . . . . . . . . .6 7. . . .9 7. . . . . . . . . . . . Numerical Stability of O E 's . . . . . . . . . 7. . . . . . . .5. 7. . .2 Driving and Parasitic Eigenvalues . .4. . . . . . . . . . . . . . . . . . . . . . . .3 Defective Eigensystems . . . . . . . Time-Space Stability and Convergence of O E's . . . . . 8. . . . . . .2. . . . .1 Imposed Constraints .7. . . . . . . . . . . . . . . .1. .1 The Basic Procedure . . .2 Complete Eigensystems . . . . . . .3 A Perspective .5. . . . . . . . . . . . . . .3 Practical Considerations for Comparing Methods . . . . . .1 -Root Traces Relative to the Unit Circle . . . . . . . . . . . 8. . . . . . . . 8. 7. . . . . . .2 Unconditional Stability. . . . . . . . .6. . . . . . 8. . . . . . . . . . . . . .6. . 7. . . .3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8. . . . . . . . . . A-Stable Methods . . . .2 Relation of Sti ness to Space Mesh Size .5 7. . . . . . . . .1. . . . . . . .2 Stability for Small t . . . . . .2 Implicit Methods . . . . . .5. .4 Comparing the E ciency of Explicit Methods . . .3 Sti ness Classi cations . . . . . . . . . . . . . . . .2 An Example Involving Di usion . . . . . . . 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 8.1 Explicit Methods . . . . . . . . . . . . . . . 7.

4. . . .2 Classical Relaxation . . . . . . . . . . . . . . . .6 Problems . . . . . . . .3 Factoring Space Matrix Operators in 2{D . . . 9. . . . . 11. 200 10. . . 9. . . . . . . . . . . . . . . . .2 The Basic Process . . 217 12. . . . . . . . . . . . . .5 Nonstationary Processes . . .2 11. . . . . . . . The Lax-Wendro Method . . . .3 The ODE Approach to Classical Relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9. . . . . . . . . .6 Problems . . . .2. .1 The Ordinary Di erential Equation Formulation . . . .3. . . . . . . . . . . . . . . . . . . . . . . . . . .1. . . . . . . . . . . . . . . . .3 The Classical Methods . . . . . . . . . . . .3 A Two-Grid Process . . . . . . . . . .2 The Gauss-Seidel System . . . .9 RELAXATION METHODS 9. . . . . . 192 10. . . . . . . . . . . . . the Residual. . . . . . . 9. . . . . . . . . . . . . . . . . . . . . . . . . 9. . . 192 10. . 11. . . . . . . 9. Upwind Schemes . . . . . . . . . . . .1. . . . . . . . . . . . . . . . . 9. . . 11. . . .2 The Converged Solution. . . . . . . . . . . . .1 The Delta Form of an Iterative Scheme . . . . . . . . . . . . . . . . . . . . . 202 11. . . 163 164 164 166 168 168 168 169 170 170 172 173 174 176 180 182 187 10 MULTIGRID 10. . . . . . . . . . 9. . . . . .1. . . . . . . . . . . . . . . . . . . . . . .2.4. .1 Formulation of the Model Problem . . . .4 Eigensystems of the Classical Methods . . . . . . . . . . .1 11. . . . . . . . 191 10. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 The Concept . . 9.4. . . . . . . . . 220 217 . . . . . . . . . . . . . . 11. .2 Properties of the Iterative Method . . . . . . . . . .3.1 Eigenvector and Eigenvalue Identi cation with Space Frequencies191 10. 191 11 NUMERICAL DISSIPATION 203 204 205 207 209 210 212 213 214 12 SPLIT AND FACTORED FORMS 12. . . . 218 12. . . . . .2 Factoring Physical Representations | Time Splitting . . . . . . . . . . . . . 9. . . . . . . . . . . . . 9. . . .2 ODE Form of the Classical Methods . . . . . . . . . . . .4 Problems . . . . . .1 Flux-Vector Splitting . . . . . . . . . .5 Arti cial Dissipation . . . . . . . . 9. .2 Flux-Di erence Splitting . . . 9. . . . . . . . . . . . . . . . . 9. . . . . . . . . . . . . . . . . . . . .1. . .2. . . . . .1 The Point-Jacobi System . . . . . .2 The Model Equations .4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Motivation . . . . . . . . . . . . . . .4 One-Sided First-Derivative Space Di erencing The Modi ed Partial Di erential Equation . . . .3 11. . . .4. . . . . . . . . . . . . . .3 The SOR System . . . . . .1 Preconditioning the Basic Matrix . . . . . . . . . . . . . . . . . . . . . . . . and the Error 9. . . . . . . . . .

. . . . .2 General Circulant Systems . . . .4 Space Splitting and Factoring . . . .1 The Unfactored Implicit Euler Method . 13. . . . . Generalized Eigensystem for Simple Tridiagonals . . . . A. . . . . . . 12. . . . . . . . . . Algebra . . . . . . . . . . . . . . . . .2 A. . . .1 Stability Comparisons of Time-Split Methods . . . . . . . . . . . . . . . . . . . . .2. . . . . . . . .5 12. . . . . .5 Example Analysis of the 3-D Model Equation . . . . . . . . . . . . . De nitions . . . .3 The Representative Equation for Space-Split Operators . . . . . . . . Importance of Factored Forms in 2 and 3 Dimensions The Delta Form . . . . . The Inverse of a Simple Tridiagonal . . . . 257 .1 A. .3 A. 13. . . . . . . . .2 The Factored Nondelta Form of the Implicit Euler Method 13. . . . . . . . . . . . . . . . . . .4. . .3 B. . . . . . . . . . . . 13. . . . . . . . . . . . 13. . . . . . . . . .2 B. . . . . . . . . . Problems . . . .5 B. . . . . . . . . . . . . . . . . . . . .1 Mesh Indexing Convention . .4. 13. . . . . . . . . . . . . . .3. B. . . . .2 Example Analysis of Circulant Systems . . .6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Vector and Matrix Norms . 12. . . . . . . . . . .1 The Representative Equation for Circulant Operators . . . . . . . . 13. . . . . 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 . . . . . . . . . . 12. . . . . . . . . . . . . . . .5 Special Cases Found From Symmetries . . . . . . . . . . . . . . . . . . . B. . . . . . .3. . . . . . . . . .2 Analysis of a Second-Order Time-Split Method . . .4. . . .12. . . . . . . . . Eigensystems of Circulant Matrices . . . . . . . . . . . 13. .3 The Factored Delta Form of the Implicit Euler Method . .4 12. . . . . . . . . . . . . . Eigensystems . . . . . . . . . .2 Data Bases and Space Vectors .2. . . . .6 12. . . . . . . . . . . . .3 Data Base Permutations . . . . . .4. . . . .4. . . . . . . . 220 221 221 223 226 226 228 229 13 LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 13. . . .4 A. . . . . . . . . . . .7 12. . . . . . . 13. . . . . . . . . . . . . B. . . . . . . . . . . . . . . . . . . . . . . 13. . . . . . . . . . . 13. . . . . . . . . . . . . . .1 B. . . Second-Order Factored Implicit Methods . . .4 The Factored Delta Form of the Trapezoidal Method . . . . . . . . . .1 Standard Tridiagonals . . .4 Notation . . . . . . . . B. . . . .3. .4 Example Analysis of 2-D Model Equations . . . . . . . .6 Special Cases Involving Boundary Conditions . . . . . . . . . . . . . . . .4.3. . . .

C THE HOMOGENEOUS PROPERTY OF THE EULER EQUATIONS265 .

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

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

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

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

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. a scalar quantity in the linear convection and di usion equations. however. it should be noted that. 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. 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. As a word of caution. Bold type is reserved for real physical vectors. the variable u can denote a scalar Cartesian velocity component in the Euler and Navier-Stokes equations.S. Otherwise.e.S. proportional to) . usually a vector A term of order (i. 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. 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. 1.1. such as velocity.. Some of the abbreviations used throughout the text are listed and de ned below. For example. S.4. and a vector consisting of a collection of scalars in our presentation of hyperbolic systems.4 Notation The notation is generally explained as it is introduced.

6 CHAPTER 1. INTRODUCTION .

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

4) r: i @x @y @z and i j. they can be written as @Q + @E = 0 (2. and P is the rate of production of Q per unit volume per unit time.1 and 2. is an area A(t) bounded by a closed contour C (t). 2.3 and nd a solution for Q on that basis are referred to as nite-di erence methods. momentum and energy for a uid. In two dimensions. For a Newtonian uid in one dimension. and k are unit vectors in the x y. t1 . 2.5) @t @x with 2 3 2 3 u 3 2 0 6 6 Q=6 6 6 4 e 7 7 u7 7 7 5 E 6 6 =6 6 6 4 u(e + p) 7 6 7 6 2+p 7. On the other hand. 2.2 and nd a solution for Q on that basis are referred to as nite-volume methods.1 can be rewritten as d Z QdV + I n:FdS = Z PdV (2. or cell. respectively. 2. The vector n is a unit vector normal to the surface pointing outward. then Eq.2 is obtained by applying Gauss's theorem to the ux integral. in Cartesian coordinates.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. containing the ux of Q per unit area per unit time.6) . by ! @ +j @ +k @ : (2.3) @t where r: is the well-known divergence operator given. F is a set of vectors. a partial derivative form of a conservation law can also be derived. leading to @Q + r:F = P (2.2 The Navier-Stokes and Euler Equations The Navier-Stokes equations form a coupled system of nonlinear PDE's describing the conservation of mass. The divergence form of Eq. the region of space.2) dt V (t) S (t) V (t) Those methods which make various numerical approximations of the integrals in Eqs.8 CHAPTER 2. 2. If all variables are continuous in time. and z coordinate directions. or tensor. Many of the advanced codes written for CFD applications are based on the nite-volume concept. 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 . Those methods which make various approximations of the derivatives in Eq.

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

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. where is the ratio of speci c heats.13) where a21 = . From this it follows that the ux Jacobian of E can be written in terms of the conservative variables as 2 6 6 6 6 6 @Ei = 6 6 6 @qj 6 6 6 6 6 4 0 1 (3 . ) q2 q1 q q3 1 0 (1 . q3 .10 CHAPTER 2. q2 . 2q . q1 .1 q3 7 .1 q2 + q3 2 2 2 q1 2 q1 q1 q2 (2. . . q2 q1 q1 3 a41 ! . cp=cv .1 q2 2 3 2 q1 3 7 7 7 7 7 7 7 7 2 7 . 2 q4 q 3 q1 2 q3 q1 . 1 q3 2 .1 0 a42 !2 a43 q2 q1 3 7 7 7 7 7 7 7 7 7 7 7 7 7 5 (2. 3 . and q4 . q2. + . (u2 + v2)=2] from the ideal gas law. 2 q1 2 q2 q1 . 1) e .1 q2 7 2 q1 7 7 7 5 3 q3 2 q1 (2. CONSERVATION LAWS AND THE MODEL EQUATIONS to derive the ux Jacobian matrices. we must rst write the ux vectors E and F in terms of the conservative variables. 1)q4 + 3.11) 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 ( . ) q3 q1 q2 q1 0 A= a21 q .12) We have assumed that the pressure satis es p = ( . 1)q4 + 2 q1 2 q1 7 7 7 7 q3 q2 7 7 q1 7 7 7 5 3 2q q4 q2 .

3 . which are further discussed in Section 2. )u (1 . )v ( . 1 v2 . . 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. )uv (2. The homogeneous property of the Euler equations is discussed in Appendix C. 1)4 ! 11 q2 q1 !3 + q3 q1 2 !2 !2 q2 q1 !3 5.( .2. Furthermore. . . THE NAVIER-STOKES AND EULER EQUATIONS a41 = ( 2 .5.14) 3 7 7 7 7 7 7 7 7 7 7 7 5 0 1 0 0 0 a21 (3 . u2 2 2 a43 = (1 .15) u where a41 = ( . 1 (3u2 + v2 ) 2 . This is the de ning feature of hyperbolic systems of PDE's. 1) .16) Derivation of the two forms of B = @F=@Q is similar. Thus we also study linear hyperbolic systems of PDE's.2. 1)u(u2 + v2) . This motivates the choice of our two scalar model equations associated with the physics of convection and di usion. aspects of convective phenomena associated with coupled systems of equations such as the Euler equations are important in developing numerical methods and boundary conditions. ue a42 = e . !2 3 5 q4 q1 ! q2 q1 ! q4 . The Navier-Stokes equations include both convective and di usive uxes.uv a41 a21 = v a42 u a43 (2. The eigenvalues of the ux Jacobian matrices are purely real. 1 43 q2 a42 = q1 2 q1 ! ! q2 q3 a43 = .

without special consideration of boundaries. 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. and the domain is in nite. a real constant which may be positive or negative. the \out ow" boundary.12 2. while the right-hand boundary is the in ow boundary when a is negative. This is consistent in terms of the well-posedness of a 1st-order PDE.18) The initial waveform propagates unaltered with speed jaj to the right if a is positive and to the left if a is negative. With periodic boundary conditions. At any given time. the ow being simulated is periodic. simply. In this case. The manner in which the boundary conditions are speci ed separates the following two phenomena for which this equation is a model: (1) In one type. eventually returning to its initial position.3 The Linear Convection Equation 2. at) (2. This is referred to as the biconvection problem. corresponding to a wave entering the domain through this \in ow" boundary. It is easy to show by substitution that the exact solution to the linear convection equation is then u(x t) = u0(x . It represents most of the \usual" situations encountered in convecting systems. Now let us consider a situation in which the initial condition is given by u(x 0) = u0(x). we pay a great deal of attention to it in the initial chapters. It is the simplest to study and serves to illustrate many of the basic properties of numerical methods applied to problems involving convection.3.1 Di erential Form CHAPTER 2. This type of phenomenon is referred to. as the convection problem. (2) In the other type. No boundary condition is speci ed at the opposite side.17) @t @x Here u(x t) is a scalar quantity propagating with speed a. the scalar quantity u is given on one boundary. what enters on one side of the domain must be the same as that which is leaving on the other. Hence. 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. Hence the wave leaves the domain through the out ow boundary without distortion or re ection. the process continues forever without any .

With such an initial condition.20 gives the following solution u(x t) = f (0)ei (x. We restrict our attention to initial conditions in the form u(x 0) = f (0)ei x (2.2.ia f (2.!t) (2. and the phase speed. .17. the wavenumber.19) where f (0) is a complex constant. we nd that f (t) satis es the following ordinary di erential equation (ODE) df = . We now examine the biconvection problem in more detail. An arbitrary initial waveform can be produced by summing initial conditions of the form of Eq.24) The relation between the frequency and the wavenumber is known as the dispersion relation. 2. in a simulation.25) . Eq. the solution can be written as u(x t) = f (t)ei x (2. are related by != a (2. 2.2 Solution in Wave Space u(x 0) = M X m=1 fm (0)ei mx (2. must be an integer.24 is characteristic of wave propagation in a nondispersive medium. and is the wavenumber. As we shall see later.at) = f (0)ei( x. THE LINEAR CONVECTION EQUATION 13 change in the shape of the solution. !.20) where the time dependence is contained in the complex function f (t). 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. For M modes.3. 2. a. It is a measure of the number of wavelengths within the domain. In order to satisfy the periodic boundary conditions.23) where the frequency.3. 2. The linear relation given by Eq. This means that the phase speed is the same for all wavenumbers.21) dt which has the solution f (t) = f (0)e. waves with di erent wavenumbers travel at di erent speeds.19. Let the domain be given by 0 x 2 .ia t (2. Substituting this solution into the linear convection equation. one obtains 2.22) Substituting f (t) into Eq.

Boundary conditions can be periodic.27) @t @x2 where is a positive real constant.27. the steady-state solution must satisfy @2u = 0 (2. giving u(x t) = M X Dispersion and dissipation resulting from a numerical approximation will cause the shape of the solution to change from that of the original waveform. or mixed Dirichlet/Neumann. with u representing the temperature.27. In the case of Eq. 2. Other steady-state solutions are obtained if a source term g(x) is added to Eq.1 Di erential Form Di usive uxes are associated with molecular motion in a continuum uid.4. this parabolic PDE governs the di usion of heat in one dimension.4 The Di usion Equation 2. CONSERVATION LAWS AND THE MODEL EQUATIONS where the wavenumbers are often ordered such that 1 2 M .30) . which is one that satis es the governing PDE with the partial derivative in time equal to zero. 2. 2. Dirichlet (speci ed u).23. u must vary linearly with x at steady state such that the boundary conditions are satis ed.28) @x2 @u = @t " 2 @u @x2 . m=1 fm (0)ei m(x.29) giving a steady state-solution which satis es @ 2 u . as follows: @u = @ 2 u (2. g(x) = 0 @x2 (2.26) 2. A simple linear model equation for a di usive process is Therefore. In contrast to the linear convection equation. g(x) # (2. Since the wave equation is linear.14 CHAPTER 2. Neumann (speci ed @u=@x). the di usion equation has a nontrivial steady-state solution. the solution is obtained by summing solutions of the form of Eq. For example.at) (2.

2 t (2.36) Substituting fm(t) into equation 2.32 is elliptic. . 2. g(x y) (2.4. 2. THE DIFFUSION EQUATION 15 In two dimensions. A solution of the form M X u(x t) = fm(t) sin mx + h(x) (2. 2. consistent with the physics of di usion. g(x y) = 0 (2.31) @t @x2 @y2 where g(x y) is again a source term. Let the domain be given by 0 x with boundary conditions u(0) = ua. 2 f (2.2. and as Laplace's equation for zero g. 2. 2 t sin m x + h(x) (2. Let the initial condition be u(x 0) = M X m=1 fm(0) sin m x + h(x) (2.37) The steady-state solution (t ! 1) is simply h(x).37 shows that high wavenumber components (large m) of the solution decay more rapidly than low wavenumber components. Eq.27 gives the following ODE for fm : dfm = .33) where must be an integer in order to satisfy the boundary conditions. The latter is known as the Poisson equation for nonzero g. Substituting this form into Eq. ua)x= . 2. The corresponding steady equation is @ 2 u + @ 2 u . Eq. the di usion equation becomes " # @u = @ 2 u + @ 2 u .32) @x2 @y2 While Eq. 2. u( ) = ub.2 Solution in Wave Space We now consider a series solution to Eq.4.31 is parabolic.e.34.34) satis es the initial and boundary conditions. It is clear that the steady-state solution is given by a linear function which satis es the boundary conditions. i. h(x) = ua + (ub .35) m m dt and we nd m fm(t) = fm(0)e.27.. we obtain m=1 u(x t) = M X m=1 m fm (0)e.

39) @t @x @f where f is the ux vector and A = @u is the ux Jacobian matrix.1. CONSERVATION LAWS AND THE MODEL EQUATIONS The Euler equations. and noting that X and X . this equation can also be written in the form @u + @f = 0 (2.38 by X .1AX (2.1u See Appendix A for a brief review of some basic relations and de nitions from linear algebra.1u.38) @t @x where u = u(x t) is a vector of length m and A is a real m m matrix. Other systems of equations governing convection and wave propagation phenomena.1 Thus = X . .41) where is a diagonal matrix containing the eigenvalues of A.42) @t + @x With w = X . Eq.1u z }| { @ X .43) @t @x When written in this manner.1AX X . For conservation laws. Such a system is hyperbolic if A is diagonalizable with real eigenvalues. such as the Maxwell equations describing the propagation of electromagnetic waves. The entries in the ux Jacobian are @f aij = @ui (2. are also of hyperbolic type. and X is the matrix of right eigenvectors. we obtain =0 (2. form a hyperbolic system of partial di erential equations.1 .8. the equations have been decoupled into m scalar equations of the form @wi + @wi = 0 (2. postmultiplying A by the product XX .2. 2. this can be rewritten as @w + @w = 0 (2.1 are constants. 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 (2.44) i @t @x 1 @X . Premultiplying Eq. 2.5 Linear Hyperbolic Systems CHAPTER 2.40) j The ux Jacobian for the Euler equations is derived in Section 2.16 2.

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

31. and initial conditions from Eq. i. 2. 2. 4. 6.4.33 with fm(0) = 1 for 1 m 3. Next consider a solution with boundary conditions ua = ub = 0. Eq. Find the eigenvalues and eigenvectors of A. 9. Plot the solution at t = 1 with = 1. 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 ! = .2. write it in the form of Eq. that is. nd the necessary values of fm (0). Find its eigenvalues and compare with those obtained in question 2.2.5). fm (0) = 0 for m > 3. 2. Given the initial condition u(x 0) = sin x de ned on 0 x 2 . 1)(e . 2. Write the classical wave equation @ 2 u=@t2 = c2 @ 2 u=@x2 as a rst-order system. Find the eigenvalues and eigenvectors of the matrix M derived in question 1. 3. 2. (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. Plot the rst three basis functions used in constructing the exact solution to the di usion equation in Section 2. are related by a similarity transform. 8. in the form @U + A @U = 0 @t @x where U = @u=@x @u=@t]T . @x + @u = 0 @y .25. 1.e. u2=2). CONSERVATION LAWS AND THE MODEL EQUATIONS Assume an ideal gas. p = ( . Plot the initial condition on the domain 0 x .1.) 5.) Next consider the same initial condition de ned only at x = 2 j=4. Write the 2-D di usion equation. j = 0 1 2 3.18 CHAPTER 2.. (Hint: make use of the matrix S = @Q=@R. respectively. Show that the two matrices M and A derived in questions 1 and 3. in the form of Eq. 2. 7. Find the values of fm(0) required to reproduce the initial condition at these discrete points using M = 4 with m = m .

the system is closed by the relation = 1 . i. PROBLEMS 19 where ! = 0 for irrotational ow. 1 2 .) . One approach to solving these equations is to add a time-dependent term and nd the steady solution of the following equation: @q + @f + @g = 0 @t @x @y (a) Find the ux Jacobians of f and g with respect to q. has real eigenvalues.2. we have @f (q) + @g(q) = 0 @x @y where v q= u f = . For isentropic and homenthalpic ow.1 1 Note that the variables have been nondimensionalized. . 1 u2 + v2 . (d) Are the above uxes homogeneous? (See Appendix C. (b) Determine the eigenvalues of the ux Jacobians.v u g = .u v . (c) Determine the conditions (in terms of and u) under which the system is hyperbolic. Combining the two PDE's..e.6.

CONSERVATION LAWS AND THE MODEL EQUATIONS .20 CHAPTER 2.

¡ ¢  .

¦   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 .

5 2.{ …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 th 4 Central 4 Pade th 2 2.5 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 1.5 3 ¡ — .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— .

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

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

The vector ~ (t) is usually determined f by the boundary conditions and possibly source terms.2 The Generic Matrix Form The generic matrix form of a semi-discrete approximation is expressed by the equation d~ = A~ . Eqs. the 3-point central-di erencing approximation produces the ODE model given by d~ = . It is used for the scalar convection model when the boundary conditions are periodic. ~ (t) u u f dt (4. 4.5 are the model ODE's for di usion and biconvection of a scalar in one dimension.6.3. 2 Model ODE for Biconvection The term biconvection was introduced in Section 2. that is. In this case. even the Euler and Navier-Stokes equations can be expressed in the form of Eq. 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 .4.5) dt 2 x p where the boundary condition vector is absent because the ow is periodic. a B (.1 0 1)~ u u (4. REDUCTION OF PDE'S TO ODE'S 53 Model ODE for Di usion For example. 4. Although the equations are nonlinear.2 1)~ + (bc) u ~ (4.1. the elements of A depend on the solution ~ and u are usually derived by nding the Jacobian of a ux vector. .4 and 4. In general.4) dt x ~ with Dirichlet boundary conditions folded into the (bc) vector.1. In such cases the equations are nonlinear. 4. 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.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. They are linear with coe cient matrices which are independent of x and t.

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

4.2. EXACT SOLUTIONS OF LINEAR ODE'S

55

Defective Systems

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

4.2.2 Single ODE's of First- and 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 De nition . We say If a solution is expressed in terms of ~ . How these relate to the numerical solutions of more practical problems depends upon the problem and. We begin by developing the concept of \spaces". f The rst group of terms on the right side of this equation is referred to classically as the complementary solution or the solution of the homogeneous equations. The second group is referred to classically as the particular solution or the particular integral. to a much greater extent.3 Real Space and Eigenspace Following the semi-discrete approach discussed in Section 4. it is said u to be in real space.4.1. This permits us to say a great deal about such problems and serves as the basis for this section. and this can add signi cantly to our understanding. respectively. of Jordan blocks or. another very useful frame. The most natural reference frame is the physical one.3.2 that pre.2.28) dt The dependent variable ~ represents some physical quantity or quantities which relate u to the problem of interest. we identify di erent mathematical reference frames (spaces) and view our solutions from within each. In particular. There is. we can develop some very important and fundamental concepts that underly the global properties of the numerical solutions to the model problems. it is more instructive to refer to these groups as the transient and steady-state solutions. We saw in Sections 4. in the most general case. we get di erent perspectives of the same solution. In this way. REAL SPACE AND EIGENSPACE 1 61 Note that the steady-state solution is A.27) 1 1 1 4. That is.3. on the cleverness of the relator. the elements of A are independent of both u and t. we reduce the partial di erential equations to a set of ordinary di erential equations represented by the generic form d~ = A~ .and post-multiplication of A by the appropriate similarity matrices transforms A into a diagonal matrix.1 and 4. composed. but entirely equivalent. as might be expected. form of the solution is ~ (t) = c e 1 t ~ + + cme m t ~ m + + cM e M t ~ M + A. ~ u u f (4. ~ u x x x f (4. An alternative. however. in the 4. ~ . In our application to uid dynamics. For the model problems on which we are focusing most of our attention.

At this point we make two observations: 1. Following Section 4.2 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. THE SEMI-DISCRETE APPROACH simplest nondefective case. The relations that transfer from one space to the other are: ~ w = X. using only complete systems for our examples. it is said to be in eigenspace (often referred to as wave space).1. Equations for the eigenvalues of the simple . we found that Eq.62 CHAPTER 4. ~ u ~ = X. ~ g f 1 1 ~ = Xw ~ u ~ = X~ f g The elements of ~ relate directly to the local physics of the problem. respectively.2 and. However. X . 4.28 had the alternative form ~ dw = w . ~ u x f and wm(t) = cm e m t + 1 gm m=1 2 M 1 1 for real space and eigenspace. and individually elements of w u they have no direct local physical interpretation. for simplicity. of scalars.3. the solutions in the two spaces f are represented by ~ (t) = X cm e m t ~ m + X . We say ~ If a solution is expressed in terms of w. the transient portion of the solution in real space consists of a linear combination of contributions from each eigenvector. and 2. ~ ~ g dt which is an uncoupled set of rst-order ODE's that can be solved independently for ~ the dependent variable vector w. m 4. When the forcing function ~ is independent of t. These model ODE's are presented in Section 4. the u ~ are linear combinations of all of the elements of ~ . the transient portion of the solution in eigenspace provides the weighting of each eigenvector component of the transient solution in real space.

and x = 2 =M .31) x is the modi ed wavenumber from Section 3. 4. m = m. First. 4. The right-hand eigenvector matrix X is given by 2 sin (x ) sin (2x ) sin (3x ) sin (4x ) 3 6 sin (x ) sin (2x ) sin (3x ) sin (4x ) 7 6 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.4. Recall that xj = j x = j =(M + 1). for the model biconvection equation .1 m = x M 2 2 2 where = .3. From Section B. are given in Appendix B.1. Notice that the di usion eigenvalues are real and negative while those representing periodic convection are all pure imaginary. These follow as special cases from the results given in Appendix B. from Section B.29) = 2(M + 1) x and.4 sin m=1 2 M (4.4.1 (4.2 + 2 cos Mm+ 1 m = x ! m .30) = sin m x m=0 1 M .5.4.32) .1 (4. The interpretation of this result plays a very important role later in our stability analysis.3 Eigenvectors of the Model Equations The Di usion Model Next we consider the eigenvectors of the two model equations. to help visualize the matrix structure. Consider Eq.i m a m m=0 1 M . 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 (4.3. we nd for the model di usion equation with Dirichlet boundary conditions . B (M : a b c) and Bp(M : a b c).ia sin 2m m=0 1 M . the model ODE's for di usion. REAL SPACE AND EIGENSPACE 63 tridiagonals. we present results for a simple 4-point mesh and then we give the general case.

the right-hand eigenvectors are given by ~ m = ei j x With xj = j M .5. 4. Eq. THE SEMI-DISCRETE APPROACH For the inverse. ~ = Xw ~ u is a sine synthesis from wave space back to real space.1 M .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. Eq.64 CHAPTER 4. 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. For our model ODE.1 M .33) In the eld of harmonic analysis. ~ ~ u gives wm = M X j =1 uj sin mxj m=1 2 M (4. 4. or left-hand eigenvector matrix X . 1 The Biconvection Model Next consider the model ODE's for periodic convection. The coe cient matrices for these ODE's are always circulant. 4.1 X m=0 (2 m=M ) j =0 1 m =0 1 M . we can write ~ = X w as u ~ uj = wmeimxj j=0 1 (4. Eq. .34) . Similarly.33.1 x = j 2 =M .32 represents the sine synthesis that companions the sine transform given by Eq. For the model di usion equation: ~ w = X . In summary. In general w = X . 4. ~ is a sine transform from real space to (sine) wave u space.

i = e. x. i = 7= 6 . we nd the following left-hand eigenvector matrix from Appendix B. i = e . Eq. f )j 1 j=1 2 M (4. ~ is a complex Fourier transform from real space to u wave space.27 becomes We can now combine the results of the previous sections to write the solutions of our model ODE's.i = .4: 1 2 3 4 4 4 4 32u 76u 76 76u 54 u 1 2 3 4 3 7 7 = X.imxj wm = M m=0 1 M .35) m m = . REAL SPACE AND EIGENSPACE 2w 6w 6 6w 4 w 3 21 1 1 1 7 1 6 1 e.3. ~ ~ u u =0 In general For any circulant system: ~ w = X .i = . u e. 1 4.36) .4.1 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 . 4. i = e.4 Solutions of the Model ODE's The Di usion Equation For the di usion equation. i = 7 4 61 e 5 4 e e . i = e. it is straightforward to establish the fact that the relation ~ = X w represents the Fourier synthesis of the variable w back to ~ . u For circulant matrices. ~ = Xw ~ u is a complex Fourier synthesis from wave space back to real space. In summary.3. 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. uj (t) = where M X m=1 cme m t sin mxj + (A.4 sin 2(M + 1) x 2 2 ! (4. ~ 7 u 5 1 X 1 M .

Eq.1 (4.1 (4.41) with the modi ed wavenumber de ned in Eq. 2.39) We see that the solutions are identical except for the steady solution and the modi ed wavenumber in the transient term. 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. We can write this ODE solution as uj (t) = M .66 CHAPTER 4. f )j 1 j=1 2 M (4. we obtain uj (t) = where M .37) With the modi ed wavenumber de ned as 2 m x m = x sin 2 and using m = m. This di erence causes the various modes (or eigenvector components) to decay at rates which di er from the exact solution. m 2 t sin m xj + (A. The Convection Equation For the biconvection equation. low wavenumber modes are accurately represented. while high wavenumber modes (if they have signi cant amplitudes) can have large errors. evaluated at the nodes of the grid: uj (t) = M X m=1 cm e. THE SEMI-DISCRETE APPROACH (4. With conventional di erencing schemes.1 X m=0 cme m t ei mxj j=0 1 M .42) .40) m = . 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.i m at ei mxj j=0 1 M .37. we can write the ODE solution as uj (t) = M X m=1 cme.1 X m=0 cme. The modi ed wavenumber is an approximation to the actual wavenumber. 4.31.

solved there in their uncoupled form. The role of m is clear it stands for some representative eigenvalue in the original A matrix.4. However. The form of Eq. 4. m=0 4. Eq. we pointed out that Eqs. topics which are covered in Chapters 6 and 7. uncoupled equation and our conclusions will apply in general. not particular problems.3.43) Once again the di erence appears through the modi ed wavenumber contained in .4 The Representative Equation In Section 4. which are related by algebraic manipulation. 4. 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.4.44) m m m dt The goal in our analysis is to study typical behavior of general situations. This is helpful both in analyzing the accuracy of time-marching methods and in studying their stability. As discussed in Section 3.23 express identical results but in terms of di erent groupings of the dependent variables.5. 2. Since the error in the phase speed depends on the wavenumber.20 and 4. the modi ed wavenumber is complex.1 (4.2. g (t) (4.i mat ei m xj j=0 1 M . 4. For such a purpose Eq. since is real. and then returned as a coupled set to real space. this leads to an error in the speed with which various m modes are convected. The importance of this concept resides in its message that we can analyze timemarching methods by applying them to a single.42 shows that the imaginary portion of the modi ed wavenumber produces nonphysical decay or growth in the numerical solution. discussed in Chapter 11.26. We found the uncoupled solution to a set of ODE's in Section 4. while the actual phase speed is independent of the wavenumber. evaluated at the nodes of the grid: M. . A typical member of the family is dwm = w . THE REPRESENTATIVE EQUATION 1 67 and compare it to the exact solution of the PDE. In the case of non-centered di erencing.44 is not quite satisfactory. Our next objective is to nd a \typical" single ODE to analyze. the result is erroneous numerical dispersion. X uj (t) = fm (0)e.

46) w(t) = ce t + ae .44 has the exact solution: k which can be used to evaluate all manner of time-marching methods. Find the entries in the boundary-condition vector. one can express any one of the forcing terms gm(t) as a nite Fourier series.52 to the 1-D di usion equation with Dirichlet boundary conditions. write the semi-discrete form obtained with periodic boundary conditions on a 5-point grid (M = 5). This leads to w(t) = ce t + The Representative ODE dw = w + ae t dt (4. Write the resulting semi-discrete ODE form. 6x @t @x 2 2 . For example X . u(1) = 1: @u = @ u .5 Problems 1. In such evaluations the parameters and must be allowed to take the worst possible combination of values that might occur in the ODE eigensystem. Consider the nite-di erence operator derived in question 1 of Chapter 3. The exact solution of the representative ODE is (for 6= ): t (4. 4. 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.g(t) = ak eikt X ak eikt k ik . From this we can extract the k'th term and replace ik with . 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. 4.45) for which Eq. in principle. Using this operator to approximate the spatial derivative in the linear convection equation.68 CHAPTER 4. Consider the application of the operator given in Eq. we note that. 2. 3. 3.

as well as the matrices X and X .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. Using these.4.37. Consider the matrix A = . For initial conditions u(x 0) = sin(mx) and boundary conditions u(0 t) = u( t) = 0. can be found from Appendix B. Consider a grid with 10 interior points spanning the domain 0 x . plot the exact solution of the di usion equation with = 1 at t = 1 with m = 1 and m = 3. The grid nodes are given by xj = j x j = 0 1 : : : 9. 5. Note that the domain is 0 x 2 and x = 2 =10. Repeat for the initial condition u(x 0) = sin 2x. (Plot the solution at the grid nodes only. compute and plot the ODE solution at t = 2 for the initial condition u(x 0) = sin x.Bp(10 . Calculate and plot the corresponding ODE solutions. PROBLEMS 69 4.4. 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. Compare with the exact solution of the PDE.5. The eigenvalues of the above matrix A.) Calculate the corresponding modi ed wavenumbers for the second-order centered operator from Eq. 1 .

70 CHAPTER 4. THE SEMI-DISCRETE APPROACH .

Finite-volume methods are applied to the integral form of the governing equations. and energy are conserved in a discrete sense. we saw how to derive nite-di erence approximations to arbitrary derivatives.1 or Eq. First. we saw that the application of a nite-di erence approximation to the spatial derivatives in our model PDE's produces a coupled set of ODE's. In Chapter 4. In this Chapter. Finite-volume methods have become popular in CFD as a result.2.1) We will begin by presenting the basic concepts which apply to nite-volume strategies.e. Second. 71 .Chapter 5 FINITE-VOLUME METHODS In Chapter 3. they ensure that the discretization is conservative. 2.. mass. Consistent with our emphasis on semi-discrete methods. As a result. which is d Z QdV + I n:FdS = Z PdV dt V (t) S (t) V (t) (5. momentum. Next we will give our model equations in the form of Eq. While this property can usually be obtained using a nite-di erence formulation. nitevolume methods do not require a coordinate transformation in order to be applied on irregular meshes. This increased exibility can be used to great advantage in generating grids about arbitrary geometries. we will study the latter form. primarily. i. it is obtained naturally from a nite-volume formulation.1. they can be applied on unstructured meshes consisting of arbitrary polyhedra in three dimensions or arbitrary polygons in two dimensions. 2. 5. This will be followed by several examples which hopefully make these concepts clear. of two advantages. we will show how similar semi-discrete forms can be derived using nite-volume approximations in space. either in the form of Eq.

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

6) 5. Integrate the ux to nd the net ux through the control-volume boundary using some sort of quadrature.5) or.2. 2. Since there is a distinct approximation to Q(x y z) in each control volume. Evaluate F(Q) at the boundary. These ideas should be clari ed by the examples in the remainder of this chapter. the following relation between rQ and Q is sometimes used: Z I rQdV = nQdS (5.7) This PDE governs a simple plane wave convecting the scalar quantity.2. A rQdA = C nQdl I (5. This issue is discussed in Section 11. Given the value of Q for each control volume.2 Model Equations in Integral Form 5. 3.1 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. Advance the solution in time to obtain new values of Q. construct an approximation to Q(x y z) in each control volume. The order of accuracy of the method is dependent on each of the approximations.5. V S Z where the unit vector n points outward from the surface or contour.2.4. in two dimensions. . nd Q at the control-volume boundary. 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. In order to include di usive uxes. The one-dimensional form is recovered with = 0. MODEL EQUATIONS IN INTEGRAL FORM 73 1. two distinct values of the ux will generally be obtained at any point on the boundary between two control volumes. u(x y t) with speed a along a straight line making an angle with respect to the x-axis. 4. Using this approximation.

d Z udA = I n: i @u + j @u ds dt A @x @y C (5. with Q = u F = iu cos + ju sin P = 0 d Z udA + I n:(iu cos + ju sin )ds = 0 dt A C (5. Substituting these expressions into a twodimensional form of Eq. 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.16) 5.3 One-Dimensional Examples We restrict our attention to a scalar dependent variable u and a scalar ux f .3. The nodes of the grid are located at xj = j x as usual. FINITE-VOLUME METHODS For unit speed a. 2. with 5. Eq. Control volume j extends from xj .74 CHAPTER 5. 2.2.8) (5.1=2 = xj .12) (5.2 gives the following integral form (5. the two-dimensional linear convection equation is obtained from the general divergence form. x=2 xj+1=2 = xj + x=2 (5. as shown in Fig. We consider an equispaced grid with spacing x. we nd (5.1. We will use the following notation: xj.ru ! @u + j @u = . i @x @y P = 0 Using these. 5.11) where A is the area of the cell which is bounded by the closed contour C . as in the model equations.2 The Di usion Equation Q = u F = .13) (5. F is simply a vector.3. x=2 to xj + x=2.14) (5.9) (5.17) . 2. Eq.15) ! to be the integral form of the two-dimensional di usion equation.10) Since Q is a scalar.

uj = uj + O( x2) 5. becomes u (5. With these de nitions.1=2 = x j+1=2 Pdx (5. Eq.1=2 Now with = x . the integral form of the linear convection equation.11.21) 4 j 1920 @x j and the integral form becomes xj. fj.22) where uj is the value at the center of the cell.5.1: Control volume in one dimension.3. 5.18) (5.3.1=2 or (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. xj . ONE-DIMENSIONAL EXAMPLES j-1/2 ∆ x j+1/2 75 j-2 j-1 L R j L R j+1 j+2 Figure 5.23) x ddtj + fj+1=2 .20) j j +1=2 dt j .1 A Second-Order Approximation to the Convection Equation In one dimension.19 in a Taylor series about xj (with t xed) to get 2 3 ! ! ! 1 Z x=2 4u + @u + 2 @ 2 u + 3 @ 3 u + : : :5 d uj x . x=2 j @x j 2 @x2 j 6 @x3 j ! ! x2 @ 2 u + x4 @ 4 u + O( x6 ) = uj + 24 @x2 (5. 5.1=2 = 0 . we can expand u(x) in Eq. fj. Hence the cell-average value and the value at the center of the cell di er by a term of second order.19) Zx d ( xu ) + f .

1 + uj ) = x ddtj + 1 (uj+1 . The cell to the right of xj +1=2. as shown in Fig.23.25) 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 .1 we have de ned the uxes at the cell boundaries in terms of the cell-average data.24) Evaluating this at j + 1=2 gives fjL+1=2 = f (uL+1=2) = uL+1=2 = uj (5. giving fjL.1 (5.1=2 = 2 (fjL.1 + uj ) (5. 2 where f^ denotes a numerical ux which is an approximation to the exact ux. We choose a piecewise constant approximation to u(x) in each cell such that u(x) = uj xj.1=2 + fjR 1=2) = 1 (uj. this point operator produces the following semidiscrete form: d~ = . 5. we obtain u 1 u x ddtj + 2 (uj + uj+1) . Thus 1 f^j+1=2 = 2 (fjL+1=2 + fjR =2) = 1 (uj + uj+1) (5.31) 2 2 With periodic boundary conditions. Substituting Eqs. 1 is the cell to the left of xj.1=2. In this example.1=2 . 1 B (.1=2 = uj. uj. Eq. FINITE-VOLUME METHODS with f = u. gives fjR =2 = uj+1 (5. the discontinuity in the ux at the cell boundary is resolved by taking the average of the uxes on either side of the boundary.28) We have now accomplished the rst step from the list in Section 5. which is cell j + 1. 5.32) dt 2 x p .1 0 1)~ u u (5.30 into the integral form. cell j is the cell to the right of xj. and cell j .29 and 5.1=2 x xj+1=2 (5. 1 (uj.26) +1 Similarly.29) +1 2 and 1 f^j. 5.30) . giving fjR 1=2 = uj (5.1) = 0 (5.1.27) .76 CHAPTER 5.

3. 5.1 + 26uj . rather than the nodal values.1 with a piecewise quadratic approximation as follows u( ) = a 2 + b + c (5. we can conclude that the use of the average of the uxes on either side of the cell boundary.uj. xj .29 and 5.3 x=2 1 j .1 0 1) are pure imaginary. uj.2 A Fourth-Order Approximation to the Convection Equation Let us replace the piecewise constant approximation in Section 5.3. 5.1 0 1) is relevant to nite-volume methods as well as nite-di erence methods. except it is written in terms of the cell average ~ .30. can lead to a nondissipative nite-volume method. uj+1 24 With these values of a. b. b.1) j 6 c = .1 (5.x j. the piecewise quadratic approximation produces (5. and c.1 u b = uj+12 . x=2 x=2 u( )d = uj (5. u u Hence our analysis and understanding of the eigensystem of the matrix Bp(.34) 1 These constraints lead to x Z 3 x=2 u( )d = uj+1 j a = uj+1 .35) the following values at the cell boundaries: uL+1=2 = 1 (2uj+1 + 5uj . as in Eqs. Since the eigenvalues of Bp(. and c are chosen to satisfy the following constraints: 1 Z . ONE-DIMENSIONAL EXAMPLES 77 This is identical to the expression obtained using second-order centered di erences.22ux2+ uj. The three parameters a.3.5.36) . x=2 u( )d = u x .1 x Z x=2 .33) where is again equal to x . ~ .

38) (5.1 . Eq.uj+2 + 5uj+1 + 2uj ) j 6 1 uL. 5. the following semi-discrete form is obtained: d~ = .1) j 6 uR+1=2 = 1 (. while the second approach is more suited to extension to higher order accuracy.37) (5.1) and 1 f^j.uj+1 + 5uj + 2uj.uj+1 + 7uj + 7uj. uj. FINITE-VOLUME METHODS uR. In this section. The rst approach is simpler to extend to multidimensions.1=2)] j j = 1 (.1.1 + uj. uj.1=2) + f (uR.3.41) Substituting these expressions into the integral form.1=2 = 2 f (uL.2) 12 (5. With periodic boundary conditions. gives u 1 x ddtj + 12 (.1=2 = 1 (.uj+2 + 8uj+1 . we describe two approaches to deriving a nite-volume approximation to the di usion equation.42) This is a fourth-order approximation to the integral form of the equation. 8uj.2) j (5.23.43) dt 12 x p This is a system of ODE's governing the evolution of the cell-average data.1)~ u u (5.39) using the notation de ned in Section 5. 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 (5.1=2 = 6 (2uj + 5uj.2) = 0 (5.78 CHAPTER 5. uj.1 . Recalling that f = u.3.8 0 8 .40) = 12 (. 1 B (1 . as can be veri ed using Taylor series expansions (see question 1 at the end of this chapter). 5.uj+2 + 7uj+1 + 7uj .3 A Second-Order Approximation to the Di usion Equation .

2u + u ) or. 1 (u . fj. Eq.47) j +1=2 @x j+1=2 x j+1 j Similarly.1=2 = 0 Zb 79 (5. u ) (5.2 1) are relevant to the study of nite-volume methods as well as nite-di erence methods.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 . with Dirichlet boundary conditions. 5.48) (5.@u=@x.50) j +1 This provides a semi-discrete nite-volume approximation to the di usion equation.16.1) dt x j . ONE-DIMENSIONAL EXAMPLES u x ddtj + fj+1=2 . u(a) @x x xj @x x j+1 j From Eq. Hence.2 1)~ + bc u u ~ 2 f^j. we obtain x duj = 1 (u .44.46) @u dx = u(b) .1=2 = . 5.51) .ru = . From Eq. 5. 5. 1x (uj .3.2. the integral form of the di usion equation. becomes with f = . to second-order.5. uj. Eq. Eq. 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. = .22. Also. we can write ! @u f^ = . Substituting these into the integral form. For our second approach. 5.44) (5.49) (5.33 we have dt x @u = @u = 2a + b @x @ (5. we use a piecewise quadratic approximation as in Section 5. d~ = 1 B (1 . u ) (5.3. and we see that the properties of the matrix B (1 .45) In one dimension.1 j (5.

1) . .49. The control volumes are regular hexagons with area A. With f = . 1 ` = p 3 p A = 3 2 3 `2 ` = 2 (5. u ) x j+1 j (5. we use a piecewise constant approximation in each control volume and the ux at the control volume boundary is the average of the uxes obtained on either side of the boundary. this gives f R = f L = .2 1)~ u u (5. is the length of the sides of the triangles.1 j j +1 dt x2 p 5.2. 1 (u . as shown in Figure 5.1=2 = .52) (5. dt x2 j . The following relations hold between `.3. fjR 1=2 = fjL. 2u + u ) (5.57) . uj.4 A Two-Dimensional Example The above one-dimensional examples of nite-volume approximations obscure some of the practical aspects of such methods. 1x (uj . The nodal data are stored at the vertices of the triangles formed by the grid. FINITE-VOLUME METHODS j +1=2 j +1=2 with a and b given in Eq. As in Section 5.54) which is identical to Eq. This produces duj = 1 (u . Thus our nal example is a nite-volume approximation to the two-dimensional linear convection equation on a grid consisting of regular triangles.@u=@x. and ` is the length of the sides of the hexagons.1. 5.55) which is written entirely in terms of cell-average data.35. 5. The resulting semi-discrete form with periodic boundary conditions is d~ = 1 B (1 . and A.56) A 3 The two-dimensional form of the conservation law is d Z QdA + I n:Fdl = 0 dt A C (5.53) Notice that there is no discontinuity in the ux at the cell boundary.80 CHAPTER 5.

1.i p (. 3 j)=2 pi (i + p j)=2 3 (. A list of the normals for the mesh orientation shown is given in Table 5.2. 5. 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.i + 3 j)=2 .5. A TWO-DIMENSIONAL EXAMPLE c b 2 3 81 p d 4 1 l ∆ 0 5 a e f Figure 5.4. Since these sides are all straight. The contour in the line integral is composed of the sides of the hexagon. respectively. Side 0 1 2 3 4 5 Outward Normal n (i . .1. i and j are unit normals along x and y.2.2: Triangular grid. see Fig. as shown in Figure 5.i . 3 j)=2 p Table 5. Outward normals. where we have ignored the source term.

we have " # 5 d Z u dA + X n (i cos + j sin ) ` Z 1=2 u(z)dz = 0 (5. There . 5.2. 5.1=2 (5.61) . Weights of ux integrals. 5. Z and the piecewise-constant approximation u = up over the entire hexagon. 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 . Introducing the cell average. if the integrals in the equation are evaluated exactly.11. see Eq.60) dt A . Side 0 1 2 3 4 5 n (i cos + j sin ) (cos .1=2 u(z)dz (5. 3 sin )=2 p Table 5. Making the change of variable z = =`.2. one has the expression u( )d = ` . cos .`=2 u ( )d (5. 3 sin )=2 cos p (cos + p sin )=2 3 (.60.60. in terms of u and the hexagon area A.59) Then.`=2 Z `=2 Z 1=2 are no numerical approximations in Eq.62) 2 1 A u dA = Aup (5. the approximation to the ux integral becomes trivial.1=2 =0 The values of n (i cos + j sin ) are given by the expressions in Table 5. FINITE-VOLUME METHODS For Eq. That is. cos + 3 sin )=2 .58) where is a length measured from the middle of a side . we have for side n Z Fdl = n (i cos + j sin ) Z `=2 .82 CHAPTER 5. cos p (. the integrated time rate of change of the integral of u over the area of the hexagon is known exactly. the two-dimensional linear convection equation.

23.66) (5.2. 5. using Taylor series expansions. PROBLEMS Similarly.60. ue) p +(. Find the semi-discrete ODE form governing the cell-average data resulting from the use of a linear approximation in developing a nite-volume method for the linear convection equation. we have for the other ve edges: Z 83 u(z)dz = up + ub 2 2 u(z)dz = up + uc 2 3 (5. 5. along with the expressions in Table 5. uf )] = 0 (5.5. Use the following linear approximation: u( ) = a + b .65) (5. 2. 5.5.42 is a fourth-order approximation to Eq. Use Taylor series to verify that Eq.67) Z Z u(z)dz = up + ud 2 4 u(z)dz = up + ue 2 5 Z Z Substituting these into Eq. u ) + (cos + p3 sin )(u .69) The reader can verify. cos + 3 sin )(uc . cos + 3 sin )(uc . ud) + (cos + 3 sin )(ub .64) (5. we obtain u(z)dz = up + uf 2 0 p u ` A ddtp + 2 (2 cos )(ua . uf )] = 0 (5.5 Problems 1. that this is a second-order approximation to the integral form of the two-dimensional linear convection equation.63) (5. u ) a d b e dt 3 p +(.68) or dup + 1 (2 cos )(u . 5.

FINITE-VOLUME METHODS u a = uj+12 .3. 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. 4. Repeat question 3 for a grid consisting of equilateral triangles.84 where b = uj and CHAPTER 5.1 and use the average ux at the cell interface.3.x j. 3. .

topics which are covered in the next two chapters. spatial discretization leads to a coupled system of nonlinear algebraic equations in the form ~u F (~ ) = 0 (6.Chapter 6 TIME-MARCHING METHODS FOR ODE'S After discretizing the spatial derivatives in the governing PDE's (such as the NavierStokes equations). 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.10. which will be discussed in Chapter 9. is thus relevant to both steady and unsteady ow problems. some sort of iterative method is required to obtain a solution. The subject of the present chapter. the goal is simply to remove the transient portion of the solution as quickly as possible timeaccuracy is not required. When using a time-marching method to compute steady ows. These can be solved iteratively using relaxation methods. Alternatively. 85 d~ = F (~ t) u ~u dt (6.3. This motivates the study of stability and sti ness.1) . or directly using Gaussian elimination or some variation thereof. which is widely used for nonlinear algebraic equations (See Section 6. one can consider the use of Newton's method. time-marching methods for ODE's. For example. 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.2) As a result of the nonlinearity of these equations. For a steady ow problem.). This produces an iterative method in which a coupled system of linear algebraic equations must be solved at each iteration.

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

1 respectively. the rst and third of these are examples of explicit methods. u0n. these methods can be constructed to give a local Taylor series accuracy of any order. The value of this equation arises from the fact that. we can analytically convert MacCormack's method. 6. which is a fully-discrete method. For an implicit method. 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] (6.5) (6. u0n. un and un.8) dt written here in terms of the dependent variable.6) According to the conditions presented under Eq. u. by applying a time-marching method.4. and therefore the time advance is simple. at given time intervals. CONVERTING TIME-MARCHING METHODS TO O E'S 87 With an appropriate choice of the 0s and 0s.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. for systems of ODE's and nonlinear problems.6. implicit methods require more complicated strategies to solve for un+1 than explicit methods. We refer to them as the explicit Euler method and the MacCormack predictor-corrector method.1. u0n+1. for example. These methods are simple recipes for the time advance of a function in terms of its value and the value of its derivative.7) du = u0 = u + ae t (6. that is.2. the new predicted solution is also a function of the time derivative at the new time level. The second is implicit and referred to as the implicit (or backward) Euler method. As we shall see. will be presented in Section 11.3 1 Here we give only MacCormack's time-marching 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. 6. The method commonly referred to as .1 for a method using two previous time levels.

TIME-MARCHING METHODS FOR ODE'S such a linear ODE into a linear O E . gives un+1 . (1 + h)un = ahe hn ~ 1 u (6. since the predictor step in Eq.7. un = he h ae hn (6.11 is identical to Eq. Apply the simple explicit Euler scheme. Note that the rst line in Eq.11 is obtained by noting that u0n+1 = F (~n+1 tn + h) ~ u = un+1 + ae h(n+1) ~ (6. 6.9) Eq. 6. 6.2. 6.11) . (1 + h)un = hae hn (6. 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. 6.7 is simply the explicit Euler method. 6. 6. expressed in terms of the dependent variable un and the independent variable n. Eq. This is demonstrated by repeating some of the developments in Section 4. to Eq. The latter are subject to a whole body of analysis that is similar in many respects to. h)un+1 . to Eq.88 CHAPTER 6. We next consider examples of this conversion process and then go into the general theory on solving O E's. 2 (1 + h)~n+1 + un+1 . but for di erence rather than di erential equations. Eq.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. 6.6. and just as powerful as.10) As a nal example. The second line in Eq. the predictor-corrector sequence. we nd un+1 = un + h un+1 + ae h(n+1) or (1 . 1 un = 1 ahe h(n+1) 2 2 which is a coupled set of linear O E's with constant coe cients. Eq. the theory of ODE's. with constant coe cients.8. There results un+1 = un + h( un + ae hn) or un+1 . 6.5. .8.9. As another example. 6.9 is a linear O E . applying the implicit Euler method.

n. in general. In terms of the initial value of u it can be written Just as in the development 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 is where c1 is a constant determined by the initial conditions. un = u0 n + a b b .3. (b = ). and b are.10. thus b bn = bn+ = E bn . 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.13) where . a. and since the equations are linear and have constant coe cients.6. n un = c1 n + bab .1 First.1 n This can all be easily veri ed by substitution. n Second-Order Equations The homogeneous form of a second-order di erence equation is given by un+2 + a1 un+1 + a0un = 0 (6.3. The exact solution of Eq. is not a function of either n or u.14) d Instead of the di erential operator D dt used for ODE's. 4. SOLUTION OF LINEAR O E'S WITH CONSTANT COEFFICIENTS 89 6. un+1 = un + a n is h i un = u0 + an . 6. one can readily show that the solution of the defective case. complex parameters. The independent variable is now n rather than t.

In the analysis of O E's.2 Special Cases of Coupled First-Order Equations A Complete System Coupled. 6. 6. Thus Eq. E I ]~ = 0 un c21 (c22 .14.17 can be written u1 (n) = C . The fact that Eq.14 for all c1 c2 and n should be veri ed by substitution. E ) u2 which must be zero for all u1 and u2. 6. there are just two roots and in terms of them the solution can be written un = c1( 1 )n + c2( 2 )n (6. etc. E I ]. 6. 2 .16 is a solution to Eq. its roots determine the solution to the O E.3. 6. . 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 (6.15) which must be zero for all un. They are found by solving the equation P ( ) = 0. rst-order.90 CHAPTER 6. Eq. ) (c c12 ) = 0 21 22 .17) which can also be written ~ n+1 = C~ n u u (c11 . Again we are led to a characteristic polynomial. 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. TIME-MARCHING METHODS FOR ODE'S where can be any fraction or irrational number.14 can now be re-expressed in an operational notion as (E 2 + a1E + a0)un = 0 (6. this time having the form P (E ) = det C . 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.16) where c1 and c2 depend upon the initial conditions. 6. The -roots are found from P ( ) = det (c11c . and refer to them as the -roots. E ) C = c11 c12 c21 c22 The operational form of Eq.15 is known as the operational form of Eq. we label these roots 1 . 6. In the simple example given above. .

1]un = h E ae hn (6.21) " . E .19) (6.18) 6.17 is if x 2 ~ n = X ck ( k )n~ k u x k=1 where ck are constants determined by the initial conditions.2.4.1 n ^ ^ " # . the solution of Eq. h)E . rst-order ordinary di erence equations. these equations can be written E . For example.2 2 2 E . (1 + h)]un = h ae hn (1 .4. Using the displacement operator.1 + u n(n .2 un = u0 + u0n ^ 0 2 n (6.20) (6.9 to 6. produced by applying a time-marching method to the representative equation. linear. ~ are its eigenvectors. 1) .6. The solution of O E's with defective eigensystems follows closely the logic in Section 4.11.2. following the logic of Section 4. 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 .2 for defective ODE's.1 The Operational Form and its Solution Examples of the nonhomogeneous.(1 + h) # " u # = h " 1 # ae hn ~ 1E 1 (1 + h)E E . are given by Eqs. SOLUTION OF THE REPRESENTATIVE O E'S 91 Obviously the k are the eigenvalues of C and. 6.4 Solution of the Representative O E's 6. 1 u n . 6.

2 Examples of Solutions to Time-Marching O E's As examples of the use of Eqs.20. 6. 6.19.23. We can express in terms of Eq. we have P (E ) = E . When determinants are involved in the construction of P (E ) and Q(E ).4.1. Eq. 1 Q(E ) = hE e h.21 there are multiple (two in this case) solutions. The terms P (E ) and Q(E ) are polynomials in E referred to as the characteristic polynomial and the particular polynomial. Notice also. or a steady state. 6.22 and 6. 1 . h Q(E ) = h (6. P ( ) = 0. one for un and un and we are usually only interested in the nal solution un. representing a time invariant particular solution. For the explicit Euler method.21.92 CHAPTER 6. Eq. we derive the solutions of Eqs. 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. respectively. as would be the case for Eq. 4.24) and the solution of its representative O E follows immediately from Eq.23: un = c1(1 + h)n + ae hn For the implicit Euler method. 6. ~ the important subset of this solution which occurs when = 0. h)E . 6. 6. h h (6.23) P (e ) k=1 where k are the K roots of the characteristic polynomial. the ratio Q(E )=P (E ) can be found by Kramer's rule.21. we have P (E ) = (1 . The general solution of Eq. In such a case K X Q(1) un = ck ( k )n + a P (1) k=1 6. 6. Eq.22 all manner of standard time-marching methods having multiple time steps and various types of intermediate predictor-corrector families. 6.45.19 to 6.25) . Keep in mind that for methods such as in Eq.22) which is produced by applying time-marching methods to the representative ODE.22 can be expressed as K h X un = ck ( k )n + ae hn Q(e h) (6.

This relation is rst demonstrated by developing it for the explicit Euler method. 1 2h2 P (E ) = det . 4.21.5 The . 1 . 1 2h2 2 he h 1 n un = c1 1 . 1 . 6. 2 2 h2 = 0 which has only one nontrivial root ( = 0 is simply a shift in the reference index). h + ae hn (1 . There is a fundamental relation between the two which can be used to identify many of the essential properties of a time-march method.(1 + h) = E E . h . 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. Relation We have now been introduced to two basic kinds of roots. 1 6. 1 2 2 2 h Q(E ) = det . 1 .6. and there ~ results # " E . 1 (1 + h)E E . h)e h . one solves for the nal family un (one can also nd a solution for the intermediate family u). h .5. the -roots and the -roots. one can write n u(t) = c1 e 1 h ~ 1 + x + cm e mh n ~m + x + cM e Mh n ~ M + P:S: (6.27. The former are the eigenvalues of the A matrix in the ODE's found by space di erencing the original PDE.27) x .5. 1 (1 E h)E 1 hE = 1 hE (E + 1 + h) 2 2 + 2 The -root is found from 1 P( ) = . h . 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. 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. Eq.1 Establishing the Relation 6.26) e h . THE . Remembering that t = nh.

94 CHAPTER 6. 6. 2 mh m . we see a correspondence between m and e m h. TIME-MARCHING METHODS FOR ODE'S where for the present we are not interested in the form of the particular solution (P:S:).4. m 2 Based on Section 4.29) Applying Eq. Now the explicit Euler method produces for each -root. we have the characteristic polynomial P (E ) = E 2 .31) (6. Suppose. 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. which is de ned by un+1 = un.8 to Eq. we use the leapfrog method for the time advance. So if we use the Euler method for the time advance of the ODE's. which is given by = 1 + h. so that for every the must satisfy the relation m . will be discussed in Section 6.27 and Eq. 6. x Comparing Eq. 1 = 0 2 (6.5. The other root. 1 4 h4 + m 2 h2 m (6.1 + 2hu0n (6. 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 (6.29.3. 2 hE . 1. instead of the Euler method.28) where the cm and the ~ m in the two equations are identical and m = (1 + m h).32) mh 8 m This is an approximation to e m h with an error O( 3h3). 3 The error is O( 2 h2 ). .28. For one of these we nd m q h+ 1+ m 1 = 1 + mh + 2 = 2 h2 .30) Now we notice that each produces two -roots. one -root. . 6. 6. q 1 + 2 h2 .

Therefore the error at the end of the simulation is reduced by a factor of four.34) 2h has a leading truncation error which is O(h2). Since the error per time step is O(h3). 4.35) has a leading truncation error O(h3).33) 1 2h2 + + 1 k hk + O hk+1 =1+ h+ 2 k! where k is the order of the time-marching method. RELATION 95 6. We refer to the root that has the above property as the principal -root. and designate it ( m )1 . Note that a second-order approximation to a derivative written in the form ( t u)n = 1 (un+1 . However.5. which is the leapfrog method: un+1 = un. 6. The following example makes this clear. THE .30 that the . un.6. while the second-order time-marching method which results from this approximation.5.1 + 2hu0n (6. twice as many time steps are required to reach the time T .5. One of these we identi ed as the principal root which always .1) (6. relation for the leapfrog method produces two -roots for each . Thus the principal root is an approximation to e h up to O hk . 6. knowing only that its leading error is O hk+1 . this is reduced by a factor of eight (considering the leading term only). This arises simply because of our notation for the time-marching method in which we have multiplied through by h to get an approximation for the function un+1 rather than the derivative as in Eq. 6.6. consistent with a second-order approximation.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.3 Spurious -Roots We saw from Eq. always produces one -root for every -root that satis es the relation (6. The above property can be stated regardless of the details of the timemarching method.34. Consider a solution obtained at a given time T using a second-order time-marching method with a time step h. Now consider the solution obtained using the same method with a time step h=2.

They are also called one-step methods. That is. except for the principal one. TIME-MARCHING METHODS FOR ODE'S has the property given in 6. it is always some algebraic function of the product h.1 and thus cannot be started using u only ~ n. All spurious roots are designated ( m )k where k = 2 3 . For example. If a time-marching method produces spurious -roots. 1 or earlier to advance the solution from time level n to n + 1. 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. However. This results because methods which produce spurious roots require data from time level n . Then the presence of spurious roots does not contaminate the answer. make a method inferior. the solution for the O E in the form shown in Eq. Thus while spurious roots must be considered in stability analysis. Such roots originate entirely from the numerical approximation of the time-marching method and have nothing to do with the ODE being solved. 6.36 must be initialized by some starting procedure. and presumably the magnitudes of the spurious roots themselves are all less than one (see Chapter 7).. they play virtually no role in accuracy analysis.28 must be modi ed.33.. For example. many very accurate methods in practical use for integrating some forms of ODE's have spurious roots. the .96 CHAPTER 6. In general. No matter whether a -root is principal or spurious.5. The other is referred to as a spurious -root and designated ( m )2 . if one starts the method properly) the spurious coe cients are all initialized with very small magnitudes. To express this fact we use the notation = ( h). all of the coe cients (cm )2 in Eq.4 One-Root Time-Marching Methods There are a number of time-marching methods that produce only one -root for each -root.36) Spurious roots arise if a method uses data from time level n . if there are more spurious roots (6. if there is one spurious root to a method. after some nite time the amplitude of the error associated with the spurious roots is even smaller then when it was initialized. 6. in itself. The initial vector ~ 0 u does not provide enough data to initialize all of the coe cients. We refer to them as one-root methods. are spurious. relation produced by a time-marching scheme can result in multiple -roots all of which. generation of spurious roots does not. It should be mentioned that methods with spurious roots are not self starting.4. . the leapfrog method requires ~ n. Following again the message of Section 4. u Presumably (i. In fact. 6. 1 or earlier.e.

1~ u x x x f ~ n = c1( 1 )n ~ 1 + + cm( m )n ~ m + + cM ( M )n ~ M + A.1 Local and Global Error Measures There are two broad categories of errors that can be used to derive and evaluate timemarching methods. A popular method having this property. h h .2.4. It is useful for comparing methods. see Section 3. respectively. Its .6. it is of no use in: . ACCURACY MEASURES OF TIME-MARCHING METHODS 97 1 The -method represents the explicit Euler ( = 0). relation is ) = 1 +1(1 . One is the error made in each time step. 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. as we shall see in Chapter 8. For example.6. However.6. and matrices are identical except the ~ and the terms inside u the parentheses on the right hand sides. This is a global error. so that all the constants. The only error made by introducing the time marching is the error that makes in approximating e h.1~ u x x x f (6. )u0n + u0n+1 6. these equations are identical. a Taylor series analysis is a very limited tool for nding the more subtle properties of a numerical time-marching method. This is a local error such as that found from a Taylor table analysis. The other is the error determined at the end of a given event which has covered a speci c interval of time composed of many time steps. the so-called -method.37) respectively. and the implicit Euler methods ( = 1). These are ~ (t) = c1 e 1 h n ~ 1 + + cm e m h n ~ m + + cM e M h n ~ M + A. It is usually used as the basis for establishing the order of a method. It is quite common to judge a time-marching method on the basis of results found from a Taylor table. It is instructive to compare the exact solution to a set of ODE's (with a complete eigensystem) having time-invariant forcing terms with the exact solution to the O E's for one-root methods. is given by the formula h i un+1 = un + h (1 . Notice that when t and n = 0. vectors.6 Accuracy Measures of Time-Marching Methods 6.4. the trapezoidal ( = 2 ). Three one-root methods were analyzed in Section 6.

1 The leading error term can be found by expanding in a Taylor series and choosing the rst nonvanishing term. is to some extent arbitrary. which can be written as h un = c1 ( 1)n + ae hn Q(e h ) P (e ) (6. a very natural local error measure for the transient solution is the value of the di erence between solutions based on these two roots. given by t u(t) = ce t + ae . TIME-MARCHING METHODS FOR ODE'S nding spurious roots. . and to study them we make use of the material developed in the previous sections of this chapter. (6. evaluating numerical stability and separating the errors in phase and amplitude. In the discussion of the .6.relation we saw that all time-marching methods produce a principal -root for every -root that exists in a set of linear ODE's. either local or global. a necessary condition for the choice should be that the measure can be used consistently for all methods. The latter three of these are of concern to us here.38) and the solution to the representative O E's.2 Local Accuracy of the Transient Solution (er Transient error er! ) The particular choice of an error measure. analyzing the particular solution of predictor-corrector combinations. However. The order of the method is the last power of h matched exactly. Our error measures are based on the di erence between the exact solution to the representative ODE.98 CHAPTER 6.39) j j 6. Therefore. This is similar to the error found from a Taylor table. including only the contribution from the principal root. We designate this by er and make the following de nition er e h. nding the global error.

iCn x.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. Thus one can obtain values of the principal root over the range 0 x for a given value of the Courant number. !h Cn x 6. Let = i! where ! is a real number representing a frequency. ( 1)2 + ( 1 )2 i r er! !h . For such cases it is more meaningful to express the error in terms of amplitude and phase. while a negative value corresponds to phase lead (the numerical phase speed is too large). is found using = . that is q era = 1 . where is the modi ed wavenumber of the spatial discretization. tan.6. j 1 j = 1 . Then the numerical method must produce a principal -root that is complex and expressible in the form (6. Such can indeed be the case when we study the equations governing periodic convection which produces harmonic motion.ia . we have h = . The above expression for er! can be normalized to give the error in the phase speed. A positive value of erp corresponds to phase lag (the numerical phase speed is too small). Introducing the Courant number. The approach to error analysis described in Section 3. Cn = ah= x. as follows .6.3 Local Accuracy of the Particular Solution (er ) The numerical error in the particular solution is found by comparing the particular solution of the ODE with that for the O E. We have found these to be given by 1 P:S:(ODE) = ae t ( .a . ACCURACY MEASURES OF TIME-MARCHING METHODS 99 Amplitude and Phase Error Suppose a eigenvalue is imaginary. The principal root.6. ) .1 ( 1 )i=( 1 )r )] and the local error in phase can be de ned as (6. 1( h).1 (6.41) Amplitude and phase errors are important measures of the suitability of time-marching methods for convection and wave propagation phenomena.42) er = er! = 1 + tan ( 1)i =( 1)r )] p where ! = .5 can be extended to the combination of a spatial discretization and a time-marching method applied to the linear convection equation.

and it is necessary that the advertised order of accuracy be valid for the nonlinear cases as well as for the linear ones. so that the order of er and er are consistent. where n ( . However. 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. is equal to zero.h ) lim P e The value of co is a method-dependent constant that is often equal to one.47) .6. a time-marching method is said to be of order k if er = c1 ( h)k1+1 er = c2 ( h)k2+1 where k = smallest of(k1 k2) (6.46) (6. TIME-MARCHING METHODS FOR ODE'S h P:S:(O E) = ae t Q(e h) P (e ) respectively. 6.P e h o (6.44) co = h!0 h( . many numerical methods generate an er that is also zero. 1 (ODE ) ( ) (6.4 Time Accuracy For Nonlinear Applications In practice. and for this case. In order to determine the leading error term.45) (6. Eq. )Q e h . If the forcing function is independent of time. The algebra involved in nding the order of er can be quite tedious. time-marching methods are usually applied to nonlinear ODE's.100 and CHAPTER 6. this order is quite important in determining the true order of a time-marching method by the process that has been outlined.43 can be written in terms of the characteristic and particular polynomials as er = co . For a measure of the local error in the particular solution we introduce the de nition er P:S: h P:S:(O E) . 6.43) The multiplication by h converts the error from a global measure to a local one. An illustration of this is given in the section on Runge-Kutta methods. More precisely.

we can also de ne global error measures. However. Let T be the xed time of the event and h be the chosen step size. N tan 1i 1r !# (6.1 ( 1) i 1r .50) . 6. we are more concerned with the global error in amplitude and phase. These are given by Era = 1 . Then the required number of time steps.49) ( ) Er! N !h . For example. We refer to such a computation as an \event".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 7.48) Global error in amplitude and phase If the event is periodic.6. 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. ACCURACY MEASURES OF TIME-MARCHING METHODS 101 The reader should be aware that this is not su cient. 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 . These are useful when we come to the evaluation of time-marching methods for speci c purposes. is N. the analysis based on a linear nonhomogeneous ODE produces the appropriate conditions for the majority of time-marching methods used in CFD.6.6. tan. ( 1 ( h))N (6. and q " ( 1 )2 + ( 1)2 i r N (6.1 ( ) =( ) ] = !T . Suppose we wish to compute some time-accurate phenomenon over a xed interval of time using a constant time step.

since most of them have historic origins from a wide variety of disciplines and applications.K 1 k hn k E A( un + ae ) (6. h. one nds 0 1 @ X k=1.51 is applied to the representative equation. We shall not spend much time on development or design of these methods.K 1 0 1 X k k E Aun = h@ k=1.4.102 CHAPTER 6. 6. developmental or design issues.7 Linear Multistep Methods In the previous sections.51) where the notation for F is de ned in Section 6. In the subsequent sections. When Eq. The methods are said to be linear because the 's and 's are independent of u and n.K k un+k = h du = u0 = F(u t) dt 1 X k=1. They are explicit if 1 = 0 and implicit otherwise. Eq. 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. It can be measured by Er ( . and they are said to be K -step because K time-levels of data are required to marching the solution one time-step.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 P eh 6. we introduce classes of methods along with their associated error analysis.8. TIME-MARCHING METHODS FOR ODE'S Global error in the particular solution Finally.7. ) Qeh . the global error in the particular solution follows naturally by comparing the solutions to the ODE and the O E.52) .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. 6.1. and the result is expressed in operational form.

Taylor table for the Adams-Moulton three-step linear multistep method.2 2 .4.h 0 u0n .2 (6. The three-step Adams-Moulton method can be written in the following form un+1 = un + h( 1u0n+1 + 0 u0n + .2) (6.7.h .51 can be multiplied by an arbitrary constant.1 u0n.51 with k = .1 .51 if X X X k = 0 and k = (K + k .1 .2 1 6 Table 6.2 Examples There are many special explicit and implicit forms of linear multistep methods.(.1 . 6.2 that a time-marching method when applied to the representative equation must provide a -root.2 u0n.2 . The condition referred to as consistency simply means that ! 1 as h ! 0.1 1 . a method should at least be rst-order accurate. un h u0n h2 u00 n h3 u000 n h4 u0000 n . . One can show that these conditions are met by any method represented by Eq.(. can be designed using the Taylor table approach of Section 3. 6.2)2 .2)0 . that approximates e h through the order of the method. 12 .(. 6. these methods are often \normalized" by requiring X k=1 Under this condition co = 1 in Eq.0 1 1 . Two well-known families of them.2)1 .5. 11 6 .1 k=0 The Adams-Bashforth family has the same 's with the additional constraint that 1 = 0.44.h .1 + .1.53) 1=1 0 = .1 1 . labeled 1.2 .1 6 .h 1 u0n+1 . 6. LINEAR MULTISTEP METHODS 103 We recall from Section 6. We can also agree that.54) A Taylor table for Eq.(.1 . The Adams-Moulton family is obtained from Eq.2u0n. . referred to as Adams-Bashforth (explicit) and AdamsMoulton (implicit).un .7. that is ! (1 + h) as h ! 0.6.1 .2 . to be of any value in time accuracy.1 2 .54 can be generated as 1 1 1 un+1 1 1 2 6 24 . 1) k Since both sides of Eq. k k k k 6. and it is certainly a necessary condition for the accuracy of any time marching method.1 u0n.2)3 . 6.

16=12 .1i = un + 2 h 0 i AB2 h = un + 12 23un . un. 16u0n.57) 4 54 0 3 12 1 .2 giving (6.2 that a kth-order time-marching method has a leading truncation error term which is O(hk+1 ).1i = un + 12 AM3 .2 = 1=24 which produces a method which is fourth-order accurate.4 7 6 0 7 6 1 7 6 7 6 7 6 3 0 3 12 7 6 76 (6.1 = .2 AB3 Implicit Methods un+1 un+1 un+1 un+1 4 Recall from Section 6.1 + 2hhun Leapfrog 1 h 3u0n .57 up to the order of the method. A list of simple methods. Explicit Methods un+1 un+1 un+1 un+1 = un + hu0n Euler 0 = un.2 . TIME-MARCHING METHODS FOR ODE'S This leads to the linear system 2 1 1 1 1 32 3 213 1 6 2 0 .58) 0 = 23=12 . 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.4 .32 1 .2 = 5=12 This is the third-order Adams-Bashforth method.2 .55 and 6. resulting in (6.1 + 2hu0n+1i =3 2nd-order Backward h h5u0n+1 + 8u0n .5.5=24 . u0n. some of which are very common in CFD applications.2 to solve for the 's. = un + hu0n+1 Implicit Euler 1 hhu0n + u0n+1i = un + 2 Trapezoidal (AM2) 1 h4un .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.1 5 = 4 1 5 (6.56) 1 = 9=24 0 = 19=24 . u0n. One can verify that the Adams type schemes given below satisfy Eqs.104 CHAPTER 6.55) 4 5 4 .4 7 6 .1 = .1 7 = 6 1 7 5 4 5 4 0 .1 + 5u0n.

.1=2 .2.e. Some linear one.1=2 . 6.7. +2 The class of all 3rd-order methods is determined by imposing the additional constraint =2 . Methods with = . LINEAR MULTISTEP METHODS 105 6. The most general two-step linear multistep method (i.and two-step methods. un.1=6 Method Euler Implicit Euler Trapezoidal or AM2 2nd Order Backward Adams type Lees Type Two{step trapezoidal A{contractive Leapfrog AB2 Most accurate explicit Third{order implicit AM3 Milne Order 1 1 2 2 2 2 2 2 2 2 3 3 3 4 Table 6.51).5=6 .51. can be written as h i (1 + )un+1 = (1 + 2 )un . This limits the interest in multistep methods to about two time levels. are known as Milne methods.1=6 0 .59 is given in Table 6.1=3 . see Eq. Notice that the Adams methods have = 0.1=2 .51.e. K=2 in Eq. the methods are 2nd-order accurate) if 1 '= .1=2 0 .1=2.2=9 0 1=2 .1 = 0 in Eq.59) Clearly the methods are explicit if = 0 and implicit otherwise. which corresponds to .1=4 . ' 1=2 1 3=4 1=3 1=2 5=9 0 0 0 1=3 5=12 1=6 0 1 0 0 0 1=2 0 .1=6 .5 6 1 Finally a unique fourth-order method is found by setting = .7. which corresponds to 0 = 0 in Eq. that is at least rst-order accurate. 6. + ')u0n .. One can show after a little algebra that both er and er are reduced to 0(h3) (i.1] + h u0n+1 + (1 .59.2. A list of methods contained in Eq. . 6.1 (6.1=2 0 0 0 0 .6. 6. =3 = 6 .1=3 0 1=12 .' = .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. 'u0n. 6.

60) where the parameters and are arbitrary parameters to be determined.44 that the same conditions also make it O(h3). by following the discussion in Section 6. One concludes. therefore. Their use in solving PDE's can be much more subtle and demands concepts5 which have no counterpart in the analysis of ODE's.61) (6.61 that + =1 =1 2 which provides two equations for three unknowns. One can analyze this sequence by applying it to the representative equation and using the operational techniques outlined in Section 6. and usually the nal family has a higher Taylor-series order of accuracy than the intermediate ones. each of which is referred to as a family in the solution process. fractional-step. There may be many families in the sequence. to the following observations. it is obvious from Eq. 6. Their use in solving ODE's is relatively easy to illustrate and understand.106 6. and hybrid methods. following the example leading to Eq.8 Predictor-Corrector Methods CHAPTER 6. Predictor-corrector methods constructed to time-march linear or nonlinear ODE's are composed of sequences of linear multistep methods. A simple one-predictor. one is led.4.6. 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. but it is not di cult to show using Eq.26. .62) Considering only local accuracy. Q(E ) = E h E + + h] 2 h2 i (6. where measures of e ciency are discussed in the next two chapters. It is easy to show. un+ = un + hu0n ~ 1 un+1 = un + 2 h 1 u0n+ + 2 . 6. For this to hold for er . that the predictor-corrector sequence is a second-order accurate method for any . 6. The situation for er requires some algebra. ( + ) h .63) 5 Such as alternating direction. that h P (E ) = E E . TIME-MARCHING METHODS FOR ODE'S There are a wide variety of predictor-corrector schemes created and used for a variety of purposes. 1 u0n ~ (6. 1 . For the method to be second-order accurate both er and er must be O(h3).

The Gazdag method. . which we discuss in Chapter 8. u0n.65) 2 The Burstein method. we refer to these as ABM(k) methods. presented earlier in this chapter.6. nally. we will consider only single-step.9 Runge-Kutta Methods There is a special subset of predictor-corrector methods. second-order accurate methods are given below.63 with = 1=2 is 1 un+1=2 = un + 2 hu0n ~ un+1 = un + hu0n+1=2 ~ and. u0n. (6. 6. If the order of the corrector is (k). The order of the combination is then equal to the order of the corrector. obtained from Eq. is 1 hu ~ i un+1 = un + 2 h 3~0n .1 (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. MacCormack's method.68) 6.9.66) (6. u0n. referred to as Runge-Kutta methods.67) (6.63 with = 1. The Adams-BashforthMoulton sequence for k = 3 is i 1 h un+1 = un + 2 h 3u0n . speci c. 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.1 ~ i hh u un+1 = un + 12 5~0n+1 + 8u0n . 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.64) Some simple. 6.1 ~ h i un+1 = un + 1 h u0n + u0n+1 ~ ~ (6.

u(tn) = 1 k1 + 2k2 + 3k3 + 4k4 (6. as we pointed out in Section 6. 1. They must satisfy the relations = 1 = 1+ 1 (6. 6. we prefer to present it using predictor-corrector notation. the principal (and only) -root is given by (6.69 and 6. We present it below in some detail. and 2 . First of all. 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) . To ensure k'th order accuracy. . TIME-MARCHING METHODS FOR ODE'S to the Taylor series expansion of e h out through the order of the method and then truncates.73) 2 = 2+ 2+ 2 .70.6.72 are a total of 13 parameters which are to be determined such that the method is fourth-order according to the requirements in Eqs. 6. the method must further satisfy the constraint that er = O(hk+1) (6.69) = 1 + h + 1 2h2 + + 1 k hk 2 k! It is not particularly di cult to build this property into a method.70) and this is much more di cult. the choices for the time samplings.71) However.4. Thus for a Runge-Kutta method of order k (up to 4th order).72) un+1 = un + 1hun 2 bn+ 3 ~n+ 1 + 4 hu0n+ 2 Appearing in Eqs. but.108 CHAPTER 6.71 is b un+ = un + hu0n b un+ 1 = un + 1 hu0n + 1hu0n+ ~ b un+ 2 = un + 2 hu0n + 2hu0n+ + 2hu0n+ 1 ~ 0 + hu0 + hu0 (6. a scheme entirely equivalent to 6. Thus. The most widely publicized Runge-Kutta process is the one that leads to the fourth-order method. are not arbitrary.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.

4. 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. the fourth condition in Eq.73 to eliminate the 's we nd from Eq. RUNGE-KUTTA METHODS 109 The algebra involved in nding algebraic equations for the remaining 10 parameters is not trivial. but the most popular one is due to Runge.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. 6.6.69 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) (6.70. two parameters can be set arbitrarily. 6.69 and 6.76) 6 conditions for Runge-Kutta methods of up to third order.75 which must be satis ed by the 10 unknowns. Thus if the rst 3 equations in 6. 7 The present approach based on a linear inhomogeneous equation provides all of the necessary . but the equations follow directly from nding P (E ) and Q(E ) and then satisfying the conditions in Eqs. 6. To satisfy the condition that er = O(k5).9. which is based on a linear nonhomogenous representative ODE. Using Eq.74 and the rst equation in 6. A more general derivation based on a nonlinear ODE can be found in several books.74 and 6.6. the resulting method would be third-order accurate. 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. Several choices for the parameters have been proposed. 6. Since the equations are overdetermined. As discussed in Section 6.7 There are eight equations in 6.75 cannot be derived using the methodology presented here.74) These four relations guarantee that the ve terms in exactly match the rst 5 terms in the expansion of e h .75 are all satis ed.

6. 6.1 Application to Systems of Equations Consider rst the numerical solution of our representative ODE u0 = u + ae t (6. we will see that these methods can have widely di erent properties with respect to stability. we obtained (1 . For example.78) .10 Implementation of Implicit Methods We have presented a wide variety of time-marching methods and shown how to derive their . RK methods of order k require k evaluations of the derivative function to advance the solution one time step. TIME-MARCHING METHODS FOR ODE'S 9 > > = > > Notice that this represents the simple sequence of conventional linear multistep methods referred to.66 and 6.67 are second-order Runge-Kutta methods. 6. This leads to various tradeo s which must be considered in selecting a method for a speci c application. storage requirements reduce the usefulness of Runge-Kutta methods of order higher than four for CFD applications. and third-order methods can be derived from Eqs. relations.75. Higher-order RungeKutta methods can be developed.110 CHAPTER 6. It is clear that for orders one through four. un = he h ae hn (6.10. but they require more derivative evaluations than their order. 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. In any event.2. Following the steps outlined in Section 6. h)un+1 .72 by setting 4 = 0 and satisfying only Eqs. Our presentation of the time-marching methods in the context of a linear scalar equation obscures some of the issues involved in implementing an implicit method for systems of equations and nonlinear equations. respectively. a fth-order method requires six evaluations to advance the solution one step. We shall discuss the consequences of this in Chapter 8. 6.74 and the rst equation in 6. 6. These are covered in this Section. In the next chapter.77) using the implicit Euler method.

IMPLEMENTATION OF IMPLICIT METHODS Solving for un+1 gives 1 un+1 = 1 .79) This calculation does not seem particularly onerous in comparison with the application of an explicit method to this ODE. In general.81 as a linear system of equations.10. For our one-dimensional examples. hf (t + h)] u u The inverse is not actually performed.82) ~ ~ n+1 = (I .81) (6.78 is (6.hf (t + h) u u or (6. requiring only an additional division.83) (6.g.80) ~ where ~ and f are vectors and we still assume that A is not a function of ~ or t.aBp (. A = . The primary area of application of implicit methods is in the solution of sti ODE's. f (t) u u ~ ~ (I .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. hA).84) un+1 = un + hF (un+1 tn+1) This is a nonlinear di erence equation. Now let us apply the implicit Euler method to our generic system of equations given by ~ 0 = A~ . consider the nonlinear ODE du + 1 u2 = 0 (6. but rather we solve Eq. for biconvection.1 0 1)=2 x). Now u u the equivalent to Eq. h (un + he h ae hn ) 111 (6. but in multidimensions the bandwidth can be very large. hA)~ n+1 . 6. as we shall see in Chapter 8.6..1 ~ n . the cost per time step of an implicit method is larger than that of an explicit method. 6.85) dt 2 . ~ n = . and hence its solution is inexpensive. the system of equations which must be solved is tridiagonal (e. 6. As an example.10.

which implies that a linearization approach may be quite successful. Designate the reference value by tn and the corresponding value of the dependent variable by un.86) n which requires a nontrivial method to solve for un+1.10. Such an approach is described in the next Section. TIME-MARCHING METHODS FOR ODE'S solved using implicit Euler time marching. u ) + @F (t .87) On the other hand. tn)k and (u . the expansion of u(t) in terms of the independent variable t is ! ! @u + 1 (t . tn)2 + @t2 n (6.83. 6.3 Local Linearization for Scalar Equations General Development Let us start the process of local linearization by considering Eq.89) . t )2 @ 2 u + u(t) = un + (t . In practice. we expand F (u t) about some reference point in time. such as Newton's method (see below).88) If t is within h of tn . A Taylor series expansion about these reference quantities gives ! ! @F (u . can be used. and Eq. t ) + 2 @u2 n n n @u@t n !n 1 2 + 2 @ F (t . There are several di erent approaches one can take to solving this nonlinear di erence equation. An iterative method. tn) @t n @t2 n n 2 (6. In order to implement the linearization.87 can be written ! ! @F (u . t ) F (u t) = F (un tn) + @u n n @t n n ! ! 1 @ 2 F (u . t ) + O(h2) F (u t) = Fn + @u n n @t n n (6. since the \initial guess" is simply the solution at the previous time step. un)k are O(hk ). u )2 + @ 2 F (u .112 CHAPTER 6. which gives 1 un+1 + h 2 u2 +1 = un (6. the \initial guess" for this nonlinear problem can be quite close to the solution. 6. u ) + @F (t . both (t . u )(t . 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). t ) + O(h2) n n dt @u n @u n n @t n (6. In many uid mechanic applications the nonlinear function F is not an explicit function " ! ! # " !# ! . consider the mechanics of applying the trapezoidal method for the time integration of Eq. @F u + @F (t . The use of local time linearization updated at the end of each time step.90) Implementation of the Trapezoidal Method As an example of how such an expansion can be used. relative to the order of expansion of the function.6. it represents a second-order-accurate. There are. combine to make a second-order-accurate numerical integration process.83. A very useful reordering of the terms in Eq.91) where we write hO(h2) to emphasize that the method is second order accurate.83. 6. un) + h @F + O(h2) + Fn 2 @u n @t n 2) +hO(h (6. With this we obtain the locally (in the neighborhood of tn) time-linear representation of Eq. Thus.93) 1 .89 to evaluate Fn+1 = F (un+1 tn+1). namely ! ! ! ! du = @F u + F . Using Eq. other second-order implicit timemarching methods that can be used. 2 h @F @u n 2 @t n which is now in the delta form which will be formally introduced in Section 12. The trapezoidal method is given by 1 un+1 = un + 2 h Fn+1 + Fn] + hO(h2) (6.10.6.92) Note that the O(h2) term within the brackets (which is due to the local linearization) is multiplied by h and therefore is the same order as the hO(h2) error from the Trapezoidal Method. and the trapezoidal time march. 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. 6. 6. IMPLEMENTATION OF IMPLICIT METHODS 113 Notice that this is an expansion of the derivative of the function. of course. 6. one nds un+1 = un + 1 h Fn + @F (un+1 .92 results in the expression 1 un = hFn + 1 h2 @F (6.

6. In such cases the partial derivative of F (u) with respect to t is zero and Eq.93 simpli es to the second-order accurate expression " 1 1 . 3. h @F @u (6. u 2.114 CHAPTER 6. Solve the coupled set of linear equations B ~n = R u ~ for ~ n. 6.83. and save ~ n.1 in carrying out this step). 6. it is generally the space-di erenced form of the steady-state equation of some uid ow problem.94 is usually implemented as follows: ~ ~ 1. the basic equation was the simple scalar equation 6. rearrange terms. In this example. (Very seldom does one nd B . we arrive at the form !# n un = hFn (6. Find ~ n+1 by adding ~ n to ~ 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. store them in an array say R.95) if we introduce Eq. A numerical time-marching procedure using Eq. It is the product of h and the RHS of the basic equation evaluated at the previous time step. and remove the explicit dependence on time.90 into this method. u 4. 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. Let this storage area be referred to as B . Solve for the elements of hFn.94) Notice that the RHS is extremely simple. but for our applications. TIME-MARCHING METHODS FOR ODE'S of t. 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 .96) . 2 h @F @u !# n un = hFn (6.

We shall see later that this method is an excellent choice for steady problems.99) Here f represents a dimensionless stream function. @F @u ! n un = Fn @F @u ! #. df A = 0 dt (6. The reader should ponder the meaning of letting h ! 1 for the trapezoidal method.e. 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.88 (remember the dependence on t has been eliminated for this case). Omission of this factor degrades the method in time accuracy by one order of h. let us consider an example involving some simple boundary-layer equations.96 obtained by dividing both sides by h and setting 1=h = 0.98) This is the well-known Newton method for nding the roots of a nonlinear equation F (u) = 0. 6. 6.. Our task is to apply the implicit trapezoidal method to the equations d3f + f d2f + dt3 dt2 0 !21 @1 . By quadratic convergence.96 leads to linear convergence.10. Newton's Method Consider the limit h ! 1 of Eq.4 Local Linearization for Coupled Sets of Nonlinear Equations In order to present this concept. where the error is the di erence between the current solution and the converged solution. 6. Use of a nite value of h in Eq. We choose the Falkner-Skan equations from classical boundary-layer theory. There results or .96. . The fact that it has quadratic convergence is veri ed by a glance at Eqs. 6.6.10. " Fn (6.94 and 6. 6. 6.1 n (6. and is a scaling factor. Quadratic convergence is thus a very powerful property.94. given by Eq.87 and 6.97) un+1 = un . we mean that the error after a given iteration is proportional to the square of the error at the previous iteration. the error at a given iteration is some multiple of the error at the previous iteration. i.

87. called the Jacobian matrix.102 by the following process d~ = F (~ ) u ~ u dt (6.8 Let us refer to this matrix as A.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 .101) Now we seek to make the same local expansion that derived Eq. 6.116 CHAPTER 6. Omitting the explicit dependency ~ ~ u on the independent variable t.2 vector for the second term. 6. and de ning F n as F (~ n). TIME-MARCHING METHODS FOR ODE'S u1 = d f dt2 2 First of all we reduce Eq.100) (6.102) A = (aij ) = @Fi =@uj For the general case involving a third order matrix this is (6. rather than a simple nonlinear scalar function.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.99 to a set of rst-order nonlinear equations by the transformations This gives the coupled set of three nonlinear equations u01 = F1 = . one has 9 2. 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. It is derived from Eq.90. 1 . The required extension requires the evaluation of a matrix. @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. and tensor products for the terms of higher order. 6. except that this time we are faced with a nonlinear vector function. a matrix times a .

) . (1 . the terms in the Jacobian matrix are updated as the solution proceeds depends. second-order-accurate approximation to a set of coupled nonlinear ordinary di erential equations that is valid for t tn + h. Returning to our simple boundary-layer example. 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. The number of times.6. What is u25 ? (Let the rst term given above be u0. 6.101. on the nature of the problem.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. A ~ +O(h2) u ~ n n un nu dt | {z } (6. of course. Without any iterating within a step advance. ~ n + O(h2) u u (6.h (u1)n 74 6 5 4 5 1 0 76 ( u2)n 7 = h 6 2 4 5 ( u) (u2)n h 3n 0 . 6.106) 6.102 as \constant" which is a locally-linear. explicit or implicit. d~ = A ~ + F . Find an expression for the nth term in the Fibonacci series.or second-order time-marching method. Any rst.93.u1 3 0 0 7 A=6 1 (6.105) where t . Thus for the Falkner-Skan equations the trapezoidal method results in 2 3 1 + h (u3)n . could be used to integrate the equations without loss in accuracy with respect to order. u3 2 u2 . 6. u the solution will be second-order-accurate in time.2 1 We nd ~ n+1 from ~ n + ~ n. Re-evaluate u u u the elements using ~ n+1 and continue. and the solution is now advanced one step. tn and the argument for O(h2) is the same as in the derivation of Eq. u2)n 3 2 2 6 2 7 6 . h(u2)n h (u1)n 72 ( u1)n 3 2 .1. PROBLEMS 117 ~ u ~ F (~ ) = F n + An ~ .11 Problems 1. Using this we can write the local linearization of Eq. and the manner in which. we nd the Jacobian matrix to be 2 .88.11. (u1u3)n . which is given by Eq.

Solve for the -roots and identify them as principal or spurious. (b) Derive the . (c) Find er and the rst two nonvanishing terms in a Taylor series expansion of the spurious root. Identify the polynomials P (E ) and Q(E ). 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 .relation.1 + 2hu0n+1 (a) Write the O E for the representative equation. Find the di erence equation which results from applying the Gazdag predictorcorrector method (Eq. (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. (c) Find er . un. Find the . 6. 4.65) to the representative equation. The 2nd-order backward method is given by i 1h un+1 = 3 4un . 5.relation. Consider the time-marching scheme given by un+1 = un. (b) Derive the .1 + 23h (u0n+1 + u0n + u0n. Identify the polynomials P (E ) and Q(E ). 6.1) (a) Write the O E for the representative equation. relation.118 CHAPTER 6. The trapezoidal method un+1 = un + 2 h(u0n+1 + u0n) is used to solve the representative ODE. (d) Perform a -root trace relative to the unit circle for both di usion and convection. TIME-MARCHING METHODS FOR ODE'S 1 2.

implicit Euler. Find er and er . 2nd-order Adams-Bashforth (AB2). 11. 200. For the explicit Euler and AB2 methods.11.0:5)= ]2 with = 0:08. Use only 4th-order Runge-Kutta time marching at a Courant number of unity. of 0. and 400 nodes. Using the computer program written for problem 7. Let u(0 t) = sin !t with ! = 10 . 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. repeat problem 9 using 4th-order (noncompact) di erences in space. Show solutions at t = 1 and t = 10 compared to the exact solution. uexact)2 u j t M j =1 where M is the number of grid nodes and uexact is the exact solution. For the initial condition. trapezoidal. ah= x. 8. Using the computer program written for problem 8. Run until a periodic steady state is reached which is independent of the initial condition and plot your solution .6. and 4th-order Runge-Kutta methods. 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. On a log-log scale. Find the . including the homogeneous and particular solutions. PROBLEMS 119 Find the di erence equation which results from applying this method to the representative equation. use a Courant number of unity. Repeat problem 7 using 4th-order (noncompact) di erences in space.relation. 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. 9. use u(x 0) = e. 7. use a Courant number. Use the explicit Euler. Find the solution to the di erence equation. Find the global order of accuracy from the plot. 10. Plot the solutions obtained at t = 1 compared to the exact solution (which is identical to the initial condition). What order is the homogeneous solution? What order is the particular solution? Find the particular solution if the forcing term is xed. Use 2nd-order centered di erences in space and a grid of 50 points.1 for the other methods.0:5 (x. plot the error given by v uX u M (uj .

At the last grid node. 6. 12.e. Use grids with 100. Use 2nd-order centered di erences in space with a 1st-order backward di erence at the out ow boundary (as in Eq. 200. Using the approach described in Section 6. Also plot the phase speed error obtained using exact integration in time. as described in problem 9. 2. and 4th-order Runge-Kutta time-marching at a Courant number of unity.120 CHAPTER 6. j . that obtained using the spatial discretization alone. erp. 13. i. without actually deriving the methods. 2.. and j +1 see problem 1 in Chapter 3). . j . 3. 1. era . 3rd. Repeat problem 11 using 4th-order (noncompact) centered di erences. use a 3rd-order backwardbiased operator (using nodes j . j . the number of grid nodes. 2nd. Use a third-order forward-biased operator at the in ow boundary (as in Eq. and j ) and at the second last node. derive and use a 3rd-order backward operator (using nodes j . j . and 400 nodes and plot the error vs.69. 3.2. nd the phase speed error.6. for the combination of second-order centered differences and 1st. and the amplitude error.69) together with 4th-order Runge-Kutta time marching.67). Explain your results. 3. Find the global order of accuracy. Note that the required -roots for the various Runge-Kutta methods can be deduced from Eq. TIME-MARCHING METHODS FOR ODE'S compared with the exact solution. 1.

͏ @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 .

oo R (σ) λ h= 0 σ=e b) Convection i ωh .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° . 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) .

°  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° .

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 .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=.

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 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.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 .0 R(λ h) -2.0 -1.0 2.0 1.

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

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

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

In this gure the time step has been chosen so that time accuracy is given to the eigenvectors associated with the eigenvalues lying in the small circle and stability without time accuracy is given to those associated with the eigenvalues lying outside of the small circle but still inside the large circle. CHOICE OF TIME-MARCHING METHODS Stable Region λh I(λ h) 1 R(λ h) Accurate Region Figure 8.150 CHAPTER 8. the time integration is an approximation in eigenspace that is di erent for every eigenvector ~ m . In x many numerical applications the eigenvectors associated with the small j mj are well resolved and those associated with the large j mj are resolved much less accurately. 6. although these eigenvectors must remain coupled into the system to maintain a high accuracy of the spatial resolution. .27 and its solution using a one-root.1: Stable and accurate regions for the explicit Euler method.28. 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. 6. This is given by Eq. if at all. Consider now the form of the exact solution of a system of ODE's with a complete eigensystem. timemarching method is represented by Eq. For a given time step. 8. the homogeneous part. The situation is represented in the complex h plane in Fig. we exclude the forcing function from further discussion in this section.1.

Rephrased. but their presence must not contaminate the accuracy of the complete solution). p+1 ! M ]. called the parasitic eigenvalues (no time accuracy whatsoever is required for the eigenvectors associated with these. and the other. First we order the eigenvalues by their magnitudes.1.2) xm xm m m Solution m=1 {z m=p+1 | } {z } | Driving Parasitic This concept is crucial to our discussion. It is important to notice that these de nitions make no distinction between real. and imaginary eigenvalues.1. . complex. 8. STIFFNESS DEFINITION FOR ODE'S 151 8. thus j 1j j 2j j Mj (8.2 Driving and Parasitic Eigenvalues For the above reason it is convenient to subdivide the transient solution into two parts.1.8. 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. although time accuracy requirements are dictated by the driving eigenvalues. Unfortunately. 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).1) Then we write p M Transient = X c e m t ~ + X c e m t ~ (8. numerical stability requirements are dictated by the parasitic ones.3 Sti ness Classi cations In particular we de ne the ratio and form the categories The following de nitions are somewhat useful. we nd that.

this ratio is about 680. In order to demonstrate this. 8. 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. we are interested only in the converged steady-state solution.e. Consider the case when all of the eigenvalues are parasitic. Typical CFD problems without chemistry vary between the mildly and strongly sti categories. For the biconvection model a similar analysis shows that j j=j j 1 M x 1 Here the sti ness parameter is still space-mesh dependent.2. Even for a mesh of this moderate size the problem is already approaching the category of strongly sti . let us examine the eigensystems of the model problems given in Section 4. One quickly nds that this grid clustering can strongly a ect the eigensystem of the resulting A matrix. 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. As a result it is quite common to cluster mesh points in certain regions of space and spread them out otherwise. 2 sin 2(M + 1) x M ! . 2 and the ratio is . A simple calculation shows that 4 2 1 = .1. and in a boundary layer. in di usion problems this sti ness is proportional to the reciprocal of the space mesh size squared. the sti er our equations tend to become. the sti ness is determined by the ratio M = 1 . i. For a mesh size M = 40.152 8. Our brief analysis has been limited to equispaced x . but much less so than for di usion-dominated problems. 4x2 sin2 2 = . near an airfoil leading or trailing edge. 4x2 x 2=..2 Relation of Sti ness to Space Mesh Size CHAPTER 8. 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. In this case the eigenvalues are all real. Under these conditions. negative numbers that automatically obey the ordering given in Eq. and are greatly a ected by the resolution of a boundary layer since it is a di usion process.3. The simplest example to discuss relates to the model di usion equation. Furthermore. Examples of where this clustering might occur are at a shock wave.

The appropriate error measures to be used in comparing methods for calculating an event are the global ones. highly dependent upon the speed. Er and Er! . and its computation usually consumes the major portion of the computer time required to make the simulation. This function is usually nonlinear. It should then be clear how the analysis can be extended to other cases. among other things. Era . and erp discussed earlier. and the accuracy required of the solution are dictated by the physics of the particular problem involved.5. in calculating the amount of turbulence in a homogeneous ow.6. relevant conclusions can be reached. capacity. and architecture of the available computer.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. d~ = F (~ t) u ~ u dt . era . 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. Since there are an endless number of these methods to choose from. Let us consider the problem of measuring the e ciency of a time{marching method for computing. For example. and the accuracy would be related to how much the energy of certain harmonics would be permitted to distort from a given level. rather than the local ones er . 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. the time interval would be that required to extract a reliable statistical sample. Nevertheless. T . discussed in Section 6. 8. 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 .8.3. There is no unique answer to such a question. an accurate transient solution of a coupled set of ODE's. PRACTICAL CONSIDERATIONS FOR COMPARING METHODS 153 problems. Such a computation we refer to as an event. The length of the time interval. it is. one can wonder how this information is to be used to pick a \best" choice for a particular problem. For example. over a xed interval of time. Let us now examine some ground rules that might be appropriate. if certain ground rules are agreed upon. and technology in uencing this is undergoing rapid and dramatic changes as this is being written.

Fev .25.3 is obtained from our representative ODE with = . u(T ) = 0:25.u dt (8.3) Er < 0:005 eT .t. this makes the exact value of u at the end of the event equal to 0. must have the property: du = . To the constraints imposed above. and must use a constant time step size. h.e.48. 3. The calculation is to be time-accurate. must be < 0:5%. The follow assumptions bound the considerations made in this Section: 1. 6..154 8. Three methods are compared | explicit Euler. 8. i.4 Comparing the E ciency of Explicit Methods As mentioned above. Since the exact solution is u(t) = u(0)e. the allowable error constraint means that the global error in the amplitude. Implications of computer storage capacity and access time are ignored. CHOICE OF TIME-MARCHING METHODS 8.4. this can be an important consideration. let us set the additional requirement The error in u at the end of the event. the e ciency of methods can be compared only if one accepts a set of limiting constraints within which the comparisons are carried out.1. First of all. i.4.e. In some contexts.2 An Example Involving Di usion Let the event be the numerical solution of from t = 0 to T = .. 8. 2. The time-march method is explicit. and RK4.1 Imposed Constraints T = Nh where N is the total number of time steps. We judge the most e cient method as the one that satis es these conditions and has the fewest number of evaluations. so that CHAPTER 8. Eq. ln(0:25) with u(0) = 1. a = 0. see Eq. AB2. must simulate an entire event which takes a total time T . the global error.

Under these conditions a method is judged as best when it has the highest global accuracy for resolving eigenvectors with imaginary eigenvalues. The above constraint has led to the invention of schemes that omit the function evaluation in the corrector step of a predictor-corrector combination. Thus the second-order Adams-Bashforth method is much better than the rstorder Euler method. In this numerical experiment the function evaluations contribute overwhelmingly to the CPU time. ln(0:25)=N . we add the following: ~u The number of evaluations of F (~ t) is xed. On the other hand.4. COMPARING THE EFFICIENCY OF EXPLICIT METHODS Then. The results shown in Table 8. In this case. a complete event must be established in order to obtain meaningful statistical samples which are the essence of the solution. in addition to the constraints given in Section 8. since h = T=N = . Method N h Fev Er 1 Euler 193 :00718 :99282 193 :001248 worst AB2 16 :0866 :9172 16 :001137 RK4 2 :6931 :5012 8 :001195 best Table 8. The main purpose of this exercise is to show the (usually) great superiority of second-order over rst-order time-marching methods. leading to the so-called incomplete . In this example we see that.1 were computed using a simple iterative procedure. for a given global accuracy. the method with the highest local accuracy is the most e cient on the basis of the expense in evaluating Fev .1. 8.1: Comparison of time-marching methods for a simple dissipation problem.8. and the fourth-order Runge-Kutta method is the best of all. it follows that 155 1 .4.1.3 An Example Involving Periodic Convection Let us use as a basis for this example the study of homogeneous turbulence simulated by the numerical solution of the incompressible Navier-Stokes equations inside a cube with periodic boundary conditions on all sides. ( 1 (ln(:25)=N ))N =:25 < 0:005 where 1 is found from the characteristic polynomials given in Table 7. and the number of these evaluations must be kept to an absolute minimum because of the magnitude of the problem.4.

leapfrog. However. Basically this is composed of an AB2 predictor and a trapezoidal corrector. one evaluation being used for each step. The second and fourth order RK methods are 2.4) 1 See Eqs. However. Let h1 be the time interval advanced in one step of a one-evaluation method. after K evaluations. In general. For a 2-evaluation method N = K=2 since two evaluations are used for each step. and AB2 schemes are all 1-evaluation methods. etc. 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 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.39.38 and 6. are the same. in this case. An example is the method of Gazdag. respectively. given in Section 6. and the number of evaluations. so that for these methods h = h1 . the time span required for one application of the method increases. in order to arrive at the same time T after K evaluations.1. that more e cient methods will result from the omission of the second function evaluation. the time step must be twice that of a one{evaluation method so h = 2h1. 6.8.156 CHAPTER 8. of course. the derivative of the fundamental family is never found so there is only one evaluation required to complete each cycle. The . K . However. For a 1evaluation method the total number of time steps. the global amplitude and phase error for kevaluation methods applied to systems with pure imaginary -roots can be written1 Era = 1 . The presumption is. Let K represent the total number of allowable Fev . the power to which 1 is raised to arrive at the nal destination decreases see the Figure below. N . . Notice that as k increases.and 4-evaluation methods. The Gazdag. j 1(ik!h1 )jK=k (8. notice also that as k increases. This is the key to the true comparison of time-march methods for this type of problem. h. For a 4-evaluation method the time interval must be h = 4h1 . CHOICE OF TIME-MARCHING METHODS predictor-corrector methods.relation for the method is shown as entry 10 in Table 7.

COMPARING THE EFFICIENCY OF EXPLICIT METHODS Er! = !T . K leapfrog AB2 Gazdag RK4 !h1 = :1 100 .2: Comparison of global amplitude and phase errors for four methods. by some rather simple calculations2 made using Eqs. 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. 2 .1 k " 157 (8. and the fact that ! = 1. in terms of phase error (for which it was designed) it is superior to the other two methods shown. see Fig. and RK4.2. On the basis of global error the Gazdag method is not superior to RK4 in either amplitude or phase. Phase error in degrees. We consider two maximum evaluation limits K = 50 and K = 100 and choose from four possible methods. 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. 7. although. the results shown in Table 8. Gazdag. However.9:8 1:5 1:5 b. The event must proceed to the time t = T = 10. The 1 root for the Gazdag method can be found using a numerical root nding routine to trace the three roots in the -plane. We idealize to the case where = i! and set ! equal to one.5) 1 (ik!h1 )]imaginary 1 (ik!h1 )]real # Consider a convection-dominated event for which the function evaluation is very time consuming.0. leapfrog. The rst three of these are one-evaluation methods and the last one is a four-evaluation method. exact = 1.:96 .4. Table 8. 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.3e. For example. we nd. K tan. AB2. we are making our comparisons on the basis of global error for a xed number of evaluations. First of all we see that for a one-evaluation method h1 = T=K . K leapfrog AB2 Gazdag RK4 !h1 = :1 100 1:0 1:003 :995 :999 !h1 = :2 50 1:0 1:022 :962 :979 a.8.3:8 . 8. Amplitude.2:4 :45 :12 !h1 = :2 50 . Using this.4 and 8.5.

With this time step the . h)nx22 + (PS )2 (8. Stability boundaries for some explicit methods are shown in Figs.6.5.100 and 2 = .1 0) as shown in Fig. If h is chosen so that all h in the ODE eigensystem fall inside this circle the integration will be numerically stable. In this case the trace forms a circle of unit radius centered at (.5 and 7.001 is negligible. 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. We will also assume that c1 = c2 = 1 and that an amplitude less than 0.100t quickly becomes very small and can be neglected in the expressions when time becomes large. h)nx12 + (PS )1 u2(n) = c1 (1 . these h are sti . Numerical evaluation is altogether di erent. We rst run 66 time steps with h = 0:001 in order to resolve the 1 term. which is determined from 1. Let us choose the simple explicit Euler method for the time march. but stable. A good example of the concept is produced by studying the Euler method applied to the representative equation. If uncoupled and evaluated in wave space.1. This de nes a time step limit based on accuracy considerations of h = 0:001 for 1 and h = 0:1 for 2 . Also shown by the small circle centered at the origin is the region of Taylor series accuracy. the time histories of the two solutions would appear as a rapidly decaying function in one case. 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. If some h fall outside the small circle but stay within the stable region. 8.1 Explicit Methods The ability of a numerical method to cope with sti ness can be illustrated quite nicely in the complex h plane. For a speci c example. and a relatively slowly decaying function in the other. 100h)nx11 + c2(1 . 100h)nx21 + c2(1 . consider the mildly sti system composed of a coupled two-equation set having the two eigenvalues 1 = . is h = 0:02. of course. depend upon ( m h)]n and no j m j can exceed one for any m in the coupled system or else the process is numerically unstable.158 CHAPTER 8. We have de ned these h as parasitic eigenvalues.1.6) We will assume that our accuracy requirements are such that su cient accuracy is obtained as long as j hj 0:1. 7.5 Coping With Sti ness 8. The coupled equations in real space are represented by u1(n) = c1 (1 . The time step limit based on stability. 8. Analytical evaluation of the time histories poses no problem since e.

However.5. the maximum step size that can be taken in order to maintain stability.e.2 Implicit Methods 1 .t to below 0.e. In both intervals the desired solution is second-order accurate and well resolved.. (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. After 66 steps. h)n term to zero (i. with h = 0:02. this is not possible because of the coupled presence of (1 . It is true that in the nal 69 steps one -root is 1 . After 70 time steps the 1 term has amplitude less than 0.001.9361.5.001. Its behavior in the h plane is shown in Fig. h)n) is 0. However. 50(0:1)]= 1 + 50(0:1)] = 0:666 . and this has no physical meaning whatsoever. Hence the 1 term can now be considered negligible. its in uence in the remaining 70 . 7. 50h nx + c u2(n) = c1 1 + 50h 21 2 ! 8. below 0. which in just 10 steps at h = 0:1 ampli es those terms by 109.6 but this time using an unconditionally stable implicit method for the time march. but that is not the point of this exercise). 100h)n) is less than 0. 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. we would like to change the step size to h = 0:1 and continue. we simply replace the Euler with the trapezoidal one and analyze the result. Now let us re-examine the problem that produced Eq. We would then have a well resolved answer to the problem throughout the entire relevant time interval.001 and that of the 2 term (i. since (0:666 )n < 1. 0:5h nx + (PS ) 2 1 + 0:5h 22 ! (8.001 and can be neglected. Now with the implicit method we can proceed to calculate the remaining part of the event using our desired step size h = 0:1 without any problem of instability. To drive the (1 .7) In order to resolve the initial transient of the term e. the amplitude of the 1 term (i. far outweighing the initial decrease obtained with the smaller time step.. Thus the total simulation requires 405 time steps. with 69 steps required to reduce the amplitude of the second term to below 0. 8.100t . we need to use a step size of about h = 0:001. 50h nx + c u1(n) = c1 1 + 50h 11 2 ! 1 .001).4b.e..8. (1 . It follows that the nal numerical solution to the ODE is now represented in real space by 1 . (1 . 100h)n. its in uence on the coupled solution is negligible at the end of the rst 70 steps. and. 0:5h nx + (PS ) 1 1 + 0:5h 12 ! 1 . COPING WITH STIFFNESS 2 159 term is resolved exceedingly well. We choose the trapezoidal method. In fact. Since this is also a oneroot method. about 339 time steps have to be computed in order to drive e.

the conditionally stable Euler method required 405 time steps.3 A Perspective 8. its behavior for t > 0:07 is identical to that of a stable spurious root. It is important to retain a proper perspective on a problem represented by the above example.160 CHAPTER 8. Therefore. For preliminary investigations it is often the best method to use for mildly-sti di usion dominated problems. This is known as the multigrid method. .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. However. In the example. There is yet another technique for coping with certain sti systems in uid dynamic applications. about three times as many. as compared to about 139 for the trapezoidal method. which is in the strongly-sti category. we have no interest whatsoever in the transient portion of the solution.10 000.1. CHOICE OF TIME-MARCHING METHODS steps is even less. although this root is one of the principal roots in the system. It is clear that an unconditionally stable method can always be called upon to solve sti problems with a minimum number of time steps. such as Adams-Bashforth or Runge-Kutta methods..5. if there are more spurious roots (8. we need an introduction to the theory of relaxation before it can be presented. the advantage begins to tilt towards implicit methods. Our sole goal is to eliminate it as quickly as possible. The reader can repeat the above example with 1 = .8) When solving a steady problem. Actually. it should be clear that as the degree of sti ness of the problem increases. It has enjoyed remarkable success in many practical problems however. as the reduced number of time steps begins to outweigh the increased cost per time step. the Euler method is extremely easy to program and requires very little arithmetic per step. 2 = . However. The total simulation requires 139 time steps. is recommended. For re ned investigations of such problems an explicit method of second order or higher. These explicit methods can be considered as e ective mildly sti stable methods. 8.

When we seek only the steady solution.1.and 3rd-order Runge-Kutta methods. compare and contrast them with the 3rd. Among implicit methods. 6. Considering the stability bounds for these methods (see problem 4 in Chapter 7) as well as their memory requirements. PROBLEMS 161 the choice of a time-marching method for a steady problem is similar to that for a sti problem.4. Repeat the time-march comparisons for di usion (Section 8.3) using 2nd. 6. 8. 2.4.7. one would like to take the limit h ! 1.3) using the 3rd.7 Problems 1. Find and plot the eigenvalues and the corresponding modi ed wavenumbers.and 4th-order Runge-Kutta methods.4.and 4th-order Adams-Bashforth methods. However. 3. the di erence being that the order of accuracy is irrelevant. What is the maximum allowable time step if all but the rst eigenvector are considered parasitic? . Eq. and the fourth-order Runge-Kutta method is a candidate for steady convection dominated problems. because of their stability properties. Consider the di usion equation (with = 1) discretized using 2nd-order central di erences on a grid with 10 (interior) points. a nite time step may be required until the solution is somewhat close to the steady solution. 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. Referring to Fig.2) and periodic convection (Section 8. Therefore the time step can be chosen to eliminate the transient as quickly as possible with no regard for time accuracy. Hence the explicit Euler method is a candidate for steady di usion dominated problems. all of the eigenvalues can be considered to be parasitic. Eq.4. which leads to Newton's method. 8. the implicit Euler method is the obvious choice for steady problems.8. For example. Repeat the time-march comparisons for di usion (Section 8.96. when using the implicit Euler method with local time linearization. none of the eigenvalues are required to fall in the accurate region of the time-marching method.2) and periodic convection (Section 8.98.

CHOICE OF TIME-MARCHING METHODS .162 CHAPTER 8.

2) In the latter case. which satis es the following coupled system of nonlinear algebraic equations: ~u F (~ ) = 0 (9. 1 163 . the ODE given by x b Eq. we developed a methodology for designing. Alternatively. ~ exists as long as A is nonsingular.4) dt and the system is integrated in time until the transient has decayed to a su ciently low level.A) have real parts lying in the left half-plane.1. they can be used to solve for the steady solution of Eq.Chapter 9 RELAXATION METHODS In the past three chapters. analyzing. a time derivative is introduced as follows d~ = A~ . These methods can be used to compute the time-accurate solution to linear and nonlinear systems of ODE's in the general form d~ = F (~ t) u ~u (9.3) To solve this system using a time-marching method. 9. Following a time-dependent path to steady state is possible only if all of the eigenvalues of the matrix A (or .4 has a stable steady solution only if A meets the above condition. and choosing time-marching methods. ~ x x b (9. 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. Although the solution ~ = A. the unsteady equations are integrated until the solution converges to a steady solution. 9.1) dt which arise after spatial discretization of a PDE.

our analysis will focus on their application to large sparse linear systems of equations in the form Ab~ . If we designate the conditioning matrix by C . Such methods are known as relaxation methods. ~ b u f f b b f 1 1 1 1 1 (9. ~ b u (9. does not depend at all on the eigensystem of the basic matrix Ab .164 CHAPTER 9.1 Preconditioning the Basic Matrix It is standard practice in applying relaxation procedures to precondition the basic equation.5) where Ab is nonsingular. 9. provided C .5. which is given by ~ 1 = A.1 Formulation of the Model Problem 9. C . at each time step of an implicit time-marching method or at each iteration of Newton's method. Such systems of equations arise. ~ b = 0 u f (9. the problem becomes one of solving for ~ in u CAb~ . C~ b = 0 u f Notice that the solution of Eq. we seek to obtain rapidly a solution which is arbitrarily close to the exact solution of Eq. RELAXATION METHODS The common feature of all time-marching methods is that they are at least rstorder accurate. 9. While they are applicable to coupled systems of nonlinear algebraic equations in the form of Eq. In the following we will see that our approach to the iterative solution of Eq. and. Using an iterative method.8) which is identical to the solution of Eq. we consider iterative methods which are not time accurate at all.7 depends crucially on the eigenvalue and eigenvector structure of the matrix CAb.5.1. 9. and the use of the subscript b will become clear shortly. 9.7) ~ = CAb ]. 9.5 from the left by some nonsingular matrix. the . C~ b = A. This preconditioning has the e ect of multiplying Eq. for example.7 is (9. In this chapter. In the simplest possible case the conditioning matrix is a diagonal matrix composed of a constant D(b). 9. For example. C~ b = A. To use these schemes.2. equally important. exists. there are well-known techniques for accelerating relaxation schemes if the eigenvalues of CAb are all real and of the same sign.6) b f 1 9.

6).1 1 1 ~= 0 u 2 x >6 x 6 >6 .11) 1 .AT A = 2 1 1 2 0 6 .9 has eigenvalues whose imaginary parts are much larger than their real parts.1 0 >4 : .2 4 1 (9.1 .10) which scales each row.2 3 2 u 7 6 . It can rst be conditioned so that the modulus of each element is 1.1 . consider a spatial discretization of the linear convection equation using centered di erences with a Dirichlet condition on the left side and no constraint on the right side.2 0 6 6 1 0 . We then further condition with multiplication by the negative transpose.2 .1 32 0 7 6 . 9.9) The matrix in Eq.1 0 6 6 . Using a rst-order backward di erence on the right side (as in Section 3.1 1 1 76 0 76 76 . this leads to the approximation 82 > 0 1 >6 > 0 6 1 <6 .1 . A choice of C which ensures this condition is the negative transpose of Ab. FORMULATION OF THE MODEL PROBLEM 165 conditioning matrix C must be chosen such that this requirement is satis ed.2 1 1 6 0 0 6 = 6 1 0 .1 54 3 7 7 7 17 7 17 5 1 0 .9. The result is A = .1 3 7 7 7 7 7 17 5 1 2 0 6 .1. This is accomplished using a diagonal preconditioning matrix 2 61 60 6 D = 2 x6 0 6 60 4 0 1 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 2 3 7 7 7 7 7 7 5 (9.1 0 76 176 .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. For example.1 4 1 0 .1 1 1 6 0 6 6 .

2 1 6 ..1 0 1 76 76 0 7 6 0 .2 0 1 76 7 6 1 0 . as given in Appendix B.2 1 =6 6 6 1 . Relaxation methods are applicable to more general matrices. 1 . 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 .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.AT Ab b is nevertheless symmetric negative de nite (i. The classical methods will usually converge if Ab is diagonally dominant. we will thoroughly investigate some simple equations which model these structures.2 4 3 7 7 7 7 7 17 5 (9. the conditioned matrix .2 32 32 0 1 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 1 0 0 17 07 7 07 7 07 5 0 3 1 . 2 We use a symmetric negative de nite matrix to simplify certain aspects of our analysis.1 ) just rearranges the rows and columns of a matrix.e.2 1 6 1 6 1 . We do not necessarily recommend the use of . RELAXATION METHODS 1 1 If we de ne a permutation matrix P and carry out the process P T . as de ned in Appendix A. In the remainder of this chapter.2 0 1 7 6 76 176 76 76 054 1 0 . ~ = 0 f (9. We will consider the preconditioned system of equations having the form A~ .166 1 1 CHAPTER 9.1.2 The Model Equations Preconditioning processes such as those described in the last section allow us to prepare our algebraic equations in advance so that certain eigenstructures are guaranteed.2 .1 which has all negative real eigenvalues.2 1 7 6 54 0 1 1 .12) 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 . symmetric with negative real eigenvalues).13) where A is symmetric negative de nite. and the classical relaxation methods can be applied. Thus even when the basic matrix Ab has nearly imaginary eigenvalues.

In the notation of f Eqs.7. It is instructive in formulating the concepts to consider the special case given by the di usion equation in one dimension with unit di usion coe cient : @u = @ u .14) The above was written to treat the general case. g(x) @t @x This has the steady-state solution 2 2 2 2 (9. ~ u u ~ g (9.2 1) x ~ fb = ~ . then A has the form A=B 1~ 1 b 2 2 2 .1. 9.9.19) ~ where ~ = x fb. ~ .2 1)~ + (bc) . we nd d~ = 1 B (1 . 9.16) @x which is the one-dimensional form of the Poisson equation. A = CAb and ~ = C~ b f f (9. (bc) g ~ (9. FORMULATION OF THE MODEL PROBLEM 1 167 accurate when we later deal with ODE formulations.15) @ u = g(x) (9.2 1)~ = ~ f (9. ~ = A. If we consider a Dirichlet boundary condition on the left side and f either a Dirichlet or a Neumann condition on the right side. is guaranteed to exist because A is nonsingular. Introducing the threepoint central di erencing scheme for the second derivative with Dirichlet boundary conditions.5 and 9.18) Choosing C = x I .13. In this case 2 Ab = 1 B (1 . we obtain B (1 . Note that the solution of Eq.17) dt x ~ where (bc) contains the boundary conditions and ~ contains the values of the source g term at the grid nodes.

2 Classical Relaxation 9. We attempt to do this in such a way.24 using a diagonal conditioning matrix. A (9. ~ = G~ n . H .21) where H is some nonsingular matrix which depends upon the iterative method. A. ~ e f (9. the Residual.25) 1 1 .and three-dimensional problems. ~ f (9.2 b . A tremendous amount of insight to the basic features of relaxation is gained by an appropriate study of the one-dimensional case. however. A]~ n .2.1 is easily obtained from the matrix resulting from the Neumann boundary condition given in Eq. 3. The iteration count is designated by the subscript n or the superscript (n). The matrix H is independent of n for stationary methods and is a function of n for nonstationary ones.21 for ~ n gives ~ n = I + H .1 The Delta Form of an Iterative Scheme +1 We will consider relaxation methods which can be expressed in the following delta form: h i H ~ n . RELAXATION METHODS ~ = .22) 9. ~ f f (9. The error at the nth iteration is de ned as ~ n ~ n .1 (9.2 s]T s = . and the Error +1 +1 1 1 1 Solving Eq. 9.20) Note that s = . ~ n = A~ n . and much of the remaining material is devoted to this case. that it is directly applicable to two.2. The converged solution is designated ~ 1 so that ~ 1 = A. or at least easier to solve than the original system.23) where G I + H.2 .2 or . ~ f 1 (9. 9.2 The Converged Solution.24) Hence it is clear that H should lead to a system of equations which is easy to solve. H .168 CHAPTER 9. ~ 1 = ~ n .

which we designate as m . 9. 9.26) Multiply Eq.27) Finally. 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 . we have considered only what are usually referred to as stationary processes in which H is constant throughout the iterations. The Gauss-Seidel method is n = 1h n + n . 9. The residual at the nth iteration is de ned as ~ n A~ n . determine the convergence rate of a method. CLASSICAL RELAXATION 169 where ~ 1 was de ned in Eq.26.29) as given in Section 9. the j th row of Eq. ~ r f (9. and use the de nition in Eq. This operator is a simple extension of the point-Jacobi method which uses the most recent update of j. G is referred to as the basic iteration matrix.25 by A from the left.9. and its eigenvalues.2.29 is satis ed using the new values of j and j .31) j j j 2 j. it is not di cult to show that ~ n = G~ n e e +1 (9. Hence the j th row of Eq. In all of the above.29 is satis ed.2. f i (9.f i (9. The Point-Jacobi method is expressed in point operator form for the one-dimensional case as n = 1 h n + n .22. . and the old value of j . 9.2.28) Consequently.30) j j j 2 j.1. Nonstationary processes in which H (and possibly C ) is varied at each iteration are discussed in Section 9.3 The Classical Methods Point Operator Schemes in One Dimension Let us consider three classical relaxation procedures for our model equation B (1 . The method of successive overrelaxation (SOR) ( +1) ( ) 1 ( ) +1 ( +1) 1 +1 ( +1) ( +1) 1 ( ) +1 1 1 +1 .5. 9.2 1)~ = ~ f (9. and j . ~ n = 0 e r (9. 9.

13 into its diagonal.D.34) Then the point-Jacobi method is obtained with H = . such that A=L+D+U (9. h i n = jn + ! ~j . A~ . It is usually expressed in two steps as h i ~j = 1 jn + jn . then perhaps it would be better to move further in this direction. which is also easy to solve. The Gauss-Seidel method is obtained with H = .170 CHAPTER 9.3. ~ f (9. ~ ] fb f (9. 9. the portion of the matrix below the diagonal. ~ = H . jn (9. being lower triangular. U . fj 2 .36) b dt 1 1 9. RELAXATION METHODS is based on the idea that if the correction produced by the Gauss-Seidel method tends to move the solution toward ~ 1.3 The ODE Approach to Classical Relaxation The particular type of delta form given by Eq. 9. but it can also be written in the single line n = ! n + (1 . 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.35) dt This is equivalent to d~ = H . C A ~ .33) j j 2 j.(L + D). which certainly meets the criterion that it is easy to solve. ! ) n + ! n .1 The Ordinary Di erential Equation Formulation . and the portion above the diagonal. 9. ! f (9. One can easily see that Eq. D. about which we have already learned a great deal. 9.21 leads to an interpretation of relaxation methods in terms of solution techniques for coupled rst-order ODE's.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~ . L.32) j ( +1) 1 ( ) +1 ( +1) ( ) ( ) where ! generally lies between 1 and 2.

given by 1 1 1 1 ~ 1 = A. the preconditioning matrix in 9. an iteration process has been decided upon). The eigenvalues are xed for a given h by the choice of time-marching method. A is nonsingular). so that we must devise a way to remove them all from the general solution on an equal basis. 6. In a stationary method. 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.35. A depends on neither ~ nor t. a constant for the entire iteration. then the system of ODE's has a steady-state solution which is approached as t ! 1. 9. that is.13. Throughout the remaining discussion we will refer to the independent variable t as \time".36 are independent of t. and the eigenvectors of H written as ~ = c e 1 t~ + + cM e M t~ M +~ 1 x} (9. For this method m = 1 + m h.28) ~ n = c ~ (1 + h)n + |x 1 1 1 + cm~ m (1 + mh)n + x {z error + cM ~ M (1 + x n M h) } +~ 1 (9. the \time" step. A are linearly independent. if all of the eigenvalues of H . H . the solution can be of t. ~ f 1 (9. even though no true time accuracy is involved. In general it is assumed that any or all of the eigenvectors could have been given an equally \bad" excitation by the initial guess. The eigenvalues are xed by the basic matrix in Eq. Assuming that H . H and C in Eq. Hence the numerical solution after n steps of a stationary relaxation method can be expressed as (see Eq.3.37) | x {z error where what is referred to in time-accurate analysis as the transient solution. THE ODE APPROACH TO CLASSICAL RELAXATION 1 1 1 171 In the special case where H . A have negative real parts (which implies that H . As we shall see. These are xed by the initial guess. 1 . they are not changed throughout the iteration process. A has been chosen (that is. Suppose the explicit Euler method is used for the time integration. It is clear that. The generalization of this in our approach is to make h. is now referred to in relaxation analysis as the error. and the secondary conditioning matrix in 9.38) which is the solution of Eq. 9.9. the three classical methods have all been conditioned by the choice of H to have an optimum h equal to 1 for a stationary iteration process. ~ is also independent u f .36.7.39) The initial amplitudes of the eigenvectors are given by the magnitudes of the cm .

RELAXATION METHODS 9.3.2 1) since then multiplication from the left by .2. since multiplication by the inverse amounts to solving the problem by a direct method without iteration. h. The constraint on H that is in keeping with the formulation of the three methods described in Section 9.42) +1 +1 It is clear that the best choice of H from the point of view of matrix algebra is . ~ n) = B (1 .41) n = n+h n with a step size.29 into the ODE form 9.2 1)~ .B . this is not in the spirit of our study. ~ n) = B (1 . equal to 1.30.2 1)~ n .32 obey no apparent pattern except that they are easy to implement in a computer code since all of the data required to update the value of one point are explicitly available at the time of the update. We arrive at H (~ n . With this choice of notation the three methods presented in Section 9.40) dt As a start. let us use for the numerical integration the explicit Euler method 0 (9. 9. Now let us study these methods as subsets of ODE as formulated in Section 9.B (1 .1. 9. 9. ! 0)(~ n . (1 .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 . If we impose this constraint and further restrict ourselves to banded tridiagonals with a single constant in each band. However.2. we are led to 2 B (. ~ f (9. Insert the model equation 9.172 CHAPTER 9. Then ~ H d = B (1 . ~ f (9.1: VALUES OF and ! IN EQ. ~ f (9.35.2 1)~ n .1.2 ODE Form of the Classical Methods The three iterative procedures de ned by Eqs.43) where and ! are arbitrary.3.3 is that all the elements above the diagonal (or below the diagonal if the sweeps are from right to left) are zero.31 and 9.2 1) gives the correct answer in one step.3 can be identi ed using the entries in Table 9. 1 +1 TABLE 9.

~ f (9. Gauss-Seidel. ~ b = 0 u f (9. However. the three methods are all contained in the following set of ODE's d~ = B . solution of the set of ODE's ~ H d = A~ . and ~ 1 A.4. The basic equation to be solved came from some time-accurate derivation Ab~ . ) n . In this light.4 Eigensystems of the Classical Methods The ODE approach to relaxation can be summarized as follows. and SOR.2 1)~ . j 2 j. 9. 9. or steady-state.44 and Table 9. The point operator that results from the use of the explicit Euler scheme is ! ! n = ! n + ! (h .46) This equation is preconditioned in some manner which has the e ect of multiplication by a conditioning matrix C giving A~ . (. are identi ed for the one-dimensional case by Eq.45) j j. 1) n + !h n . or error.1. our purpose is to examine the whole procedure as a special subset of the theory of ordinary di erential equations. EIGENSYSTEMS OF THE CLASSICAL METHODS 173 The fact that the values in the tables lead to the methods indicated can be veri ed by simple algebraic manipulation. !h f (9. (!h .44) dt ! and appear from it in the special case when the explicit Euler method is used for its numerical integration. 2 0) B (1 . ~ is the steady-state solution. 1 . The e f three classical methods. 2 2 j 2 j 1 ( +1) ( +1) 1 ( ) 1 ( ) ( ) +1 This represents a generalization of the classical relaxation techniques.49) where ~ n is the transient.47) An iterative scheme is developed to nd the converged. ~ f (9.9.48) dt This solution has the analytical form ~ n = ~n + ~ 1 e (9. ~ = 0 f (9. Point-Jacobi.

To obtain the optimal scheme we wish to minimize the maximum j m j which occurs when h = 1. i.2. 9. Thus Convergence rate j mjmax m=1 2 1 M (9.1.2 and for h > 1.3. 9. the optimum stationary case. This special case makes the general result quite clear. j j = j M j and the best possible convergence rate is achieved: 1 1 j m jmax = cos M + 1 (9.1 2 ). the maximum j m j is obtained with m = M .52) ! xm The generalized eigensystem for simple tridigonals is given in Appendix B. j m j > 1:0.1 + cos M +1 m = 1 2 ::: M The . B (1 . This relation can be plotted for any h. 9. For h < 1.31.174 CHAPTER 9.50) In this section. and 9.53) m = . RELAXATION METHODS 1 Given our assumption that the component of the error associated with each eigenvector is equally likely to be excited.44.relation for the explicit Euler method is m = 1 + m h. 9.1 since both d and f are zero.51) for the special case 2 0)~ (9. 9. Note that for h > 1:0 (depending on M ) there is the possibility of instability.54) . is shown in Fig. The plot for h = 1. the asymptotic convergence rate is determined by the eigenvalue m of G ( I + H .4.e. Fig. the maximum j mj is obtained with m = 1. we use the ODE analysis to nd the convergence rates of the three classical methods represented by Eqs. A matrix for the three methods. To illustrate the behavior.2 1)~ m = x m B (.30. we take M = 5 for the matrix order.32. 9. This amounts to solving the generalized eigenvalue problem A~ m = m H~ m x x (9. . c = 1. A) having maximum absolute value. The three special cases considered below are obtained with a = 1. and f = 0. It is also instructive to inspect the eigenvectors and eigenvalues in the H . A reduces to simply B ( 1 .1 The Point-Jacobi System 1 If = 0 and ! = 1 in Eq. 9. The eigenvalues are given by the equation m (9.2. b = . the ODE matrix H . d = . 2 The eigensystem can be determined from Appendix B. Fig. e = 2=!.

3=2 1 2 3 4 5 (9.1:5 2 p (9.1 7 6 7 x 6 0 7 x 6 7 x 6 6 7 6 6 3=2 7 6 0 7 7 6 . Again from Appendix B. Thus after 500 iterations the error content associated with each eigenvector is reduced to no more than 0.1 . For M = 5. 3=2 7 6 6 6 7 ~ 6 . we obtain j mjmax = 0:9971.56) The corresponding eigenvalues are. ~ 1 = c 1 . from Eq. the eigenvectors of H .1:866 From Eq. the numerical solution written in full is 5 p ~ n .1 + 23 = . (1 . 3=2 7 5 4 p 5 4 1=2 . 3=2 7 4 5 4 5 5 4 p 7 1 1=2 .1 + 1 = . 3=2 7 ~ = 6 p0 7 x = 6 p 7 6 1 7 x 6 6 7 6 3=2 7 7 6 . 1 = . 9.55) M +1 This is a very \well-behaved" eigensystem with linearly independent eigenvectors and distinct eigenvalues.0:134 = .9. 23 = .23 times its initial level. EIGENSYSTEMS OF THE CLASSICAL METHODS 1 175 For M = 40.1 .4. A are given by ~ j = (xj )m = sin j m x j = 1 2 ::: M (9.57) 4 = . 3 )h]n~ x 2 1 x + c 1 . 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 . 2 )h]n~ 1 2 2 p 1 . 9.1 = .53 1 2 3 = .1.1:0 = . 3=2 3 2 p 3 2 3 1= p =2 7 p2 7 6 . The rst 5 eigenvectors are simple sine waves.0:5 2 = .39. (1 .

RELAXATION METHODS h = 1.0 Figure 9. the matrix eigensystem evolves from the relation B (1 .0 m=3 -1. 9.58) 9.1: The relation for Point-Jacobi. One can show that the H .2.1 2 0)B (1 . (1 3 4 + c 1 . is AGS B .0 m=1 Values are equal σm m=4 m=2 Μ=5 0. + c 1 .59) 1 which can be studied using the results in Appendix B.176 1.4.0 -2.2 1) = 1 .44.1 2 0)~ m x (9. AGS .2 1)~ m = x m B (.0 m=5 CHAPTER 9.0 0. (. h = 1 M = 5. (1 + 1 )h]n~ 2 x p + c 1 . A matrix for the Gauss-Seidel method. (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.

072 / Ustable h > 1.0 h < 1.0 -2.0 λmh Figure 9.072 1.2: The relation for Point-Jacobi.9.0 0.0 λ mh ∼ ∼ 1. h = 0:9 M = 5.0 De cr ea h si ng σm ng si Figure 9. EIGENSYSTEMS OF THE CLASSICAL METHODS Approaches 1.0 h 0.0 ng si ea cr In In cr ea si σm ng h h M = 5 0.0 h > 1. h = 1:1 M = 5.0 at h De cr ea -1.4.0 -1.0 0.3: The relation for Point-Jacobi. Exceeds 1. .0 177 1.0 -2.

.1 1=2 3 1 6 0 . and there are (M . 7 . producing waves that are higher in amplitude on one side (the updated side) than they are on the other. is shown in Fig. 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. . .relation for h = 1.. .. RELAXATION METHODS 2 . Hence the convergence rate is 2 1 CHAPTER 9.61) m = . 4 5 0 1=2M M The eigenvector structure of the Gauss-Seidel ODE matrix is quite interesting. They are no longer symmetrical.3=4 1=2 7 5 6 7 6 .. The Jordan canonical form for M = 5 is j m jmax = cos M + 1 2 (9. .3=4 1=2 7 4 6 7 (9. . If M is odd there are (M + 1)=2 distinct eigenvalues with corresponding linearly independent eigenvectors. The eigenvectors are quite unlike the Point-Jacobi set. . . The equation for the nondefective eigenvalues in the ODE matrix is (for odd M ) + m ) m = 1 2 ::: M2 1 (9. For M = 40.3=4 1=2 7 2 6 7 6 0 1=8 . j m jmax = 0:9942. . 6 .62) M +1 The . .63) 3 2 h~ i 7 6 h~ i 7 6 6 7 2~ 37 . m + ~ m = cos m x sin j M + 1 m = 1 2 ::: M2 1 (9. 1)=2 defective eigenvalues with corresponding 178 Since this is the square of that obtained for the Point-Jacobi method. 250 iterations are required to reduce the error component of the worst eigenvector by a factor of roughly 0.4. . the optimum stationary case.. 9. . .60) 6 0 1=32 1=16 1=8 ..3=4 1=2 7 3 6 7 6 7 6 0 1=16 1=8 .64) . Furthermore.1 + cos ( M +1 and the corresponding eigenvectors are given by j. The m with the largest amplitude is obtained with m = 1.principal vectors. 7 . they do not represent a common family for di erent values of M . the error associated with the \worst" eigenvector is removed in half as many iterations.23. . .

0 -2. h )n~ 4 x h x + c (1 . For M = 5 they can be written 2 3 2 p 3 2 3 2 3 2 3 3=2 1=2 7 17 0 7 6 3=4 7 6 p3=4 7 607 6 2 7 6 0 7 6 6 7 6 6 6 0 7 ~ = 6 3=4 7 ~ = 6 p0 7 ~ = 6 0 7 ~ = 6 . ~ 1 = c (1 . 3=16 7 7 607 6 0 7 6 . 34 )n~ 1 2 3 3 1 1 2 2 (9. linked to (5) Jordan block The numerical solution written in full is thus ~ n . 3=32 1 2 3 4 5 The corresponding eigenvalues are = .4: The relation for Gauss-Seidel.4 7 4 5 4 p 5 4 5 4 5 4 5 9=32 0 0 1 .9.66) .0 179 h = 1. h = 1:0 M = 5.1 (4) Defective. The eigenvectors and principal vectors are all real.1=4 = .0 σm Μ=5 2 defective λ m 0.1 7 ~ = 6 4 7 (9.65) 6 7 x 6 7 x 6 7 x 6 7 x 6 7 x 6 6 6 7 6 7 6 7 6 9=16 7 7 6 .3=4 ) = .0 λmh Figure 9. EIGENSYSTEMS OF THE CLASSICAL METHODS 1.0 -1.4.0 0.

A matrix for the SOR method.1 x 0)B (1 . + x . 1) (1 .2 1). ~ x 3 4 1 5 2 2 3 4 5 + c (1 . equal and defective.2x + x x x 2x2x x 6 x + . and the following conditions hold: 1. 2x 0 0 0 x 4 4 3 7 7 7 7 (9.1 x 0)B (1 . 2x 4 . If M is even. RELAXATION METHODS " # n + c h n (1 . 2x + x x . 2x + x x . !) + ! pm < 0 zm and m are complex. one of the remaining eigenvalues is real and the others are complex occurring in conjugate pairs.67) 9. + c h n(n . 2x x x 6 . the remaining eigenvalues are complex and occur in conjugate pairs. h)n. 2x + x x .2 1). .4. the system is Gauss-Seidel. One can show that this can be written in the form given below for M = 5.68) 7 7 5 Eigenvalues of the system are given by m + = .44. h)n~ x 5 1! 1 4 5 (9. !) + ! pm] = pm = cos m =(M + 1)] 2 2 2 1 If ! = 1. the ODE matrix is B . 2x + x 6 x . (. If 4(1 .180 CHAPTER 9. The generalization to any M is fairly clear.2 + x 1 . 3. !) + ! p = 0 ! is optimum for the stationary case. 9. h) x 1! 2! + c (1 .3 The SOR System 1 If = 1 and 2=! = x in Eq. (.2x x . If M is odd. h)n.2x + x x . 2. is 1 1 2 x 0 0 6 . If ! is chosen such that 4(1 . 2x + x x 4 4 3 4 3 4 4 5 2 3 2 3 4 3 4 4 2 2 3 2 3 4 3 4 2 2 3 2 3 4 3 .1 + !pm 2 zm 2 m = 1 2 :::M 2 2 1 2 (9. x 0 6 1 6 .69) where zm = 4(1 .2. ~ + c (1 . Two eigenvalues are real. h)n. h)n + c h n (1 . ASOR B . The H .

For M = 5 they can be written 2 3 2 3 2 p3(1 3 2 3 1=2 7 . there are two real eigenvectors and one real principal vector. 1. the optimum value of ! may have to be determined by trial and error.5.1 j .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 = .1 m sin 2 where m j M +1 (9.relation reduces to that shown in Fig. p m= m 2 m 2 1 2 q .23 times its initial value in only 10 iterations. In practical applications.(10 + 2 2i)=9 = . 9.9.4.71) Using the explicit Euler method to integrate the ODE's. much faster than both GaussSeidel and Point-Jacobi.2=3 (2) Defectivep linked to = . 1 (9. the . j mjmax = 0:8578. the optimum value for the stationary case.(10 . This illustrates the fact that for optimum stationary SOR all the j m j are identical and equal to !opt .72) !opt = 2= 1 + sin M + 1 For M = 40.74) 2 1 2 34 5 1 1 3 4 5 !opt p + i p . and the bene t may not be as great.70) !opt = 2= 1 + sin M + 1 ~m x m = = m. EIGENSYSTEMS OF THE CLASSICAL METHODS One can easily show that the optimum ! for the stationary case is and for ! = !opt 181 (9. The remaining linearly independent eigenvectors are all complex.4=3 (9. h + h m .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. 2p2i)=9 = . Hence the convergence rate is j m jmax = !opt . For odd M . and if h = 1. m = 1 . Hence the worst error component is reduced to less than 0.

H and C . h. 2h=3)n~ x p c 1 . ~1 = + + + + c (1 . For the Gauss-Seidel and SOR methods it leads to processes that can be superior to the stationary methods.75) 9.0 -2.0 -1. RELAXATION METHODS h = 1. 4h=3)n~ x 1 2 1 2 3 2 3 4 5 4 5 1 (9.182 1.5: The relation for optimum stationary SOR.0 CHAPTER 9. (10 .5 Nonstationary Processes In classical terminology a method is said to be nonstationary if the conditioning matrices. This does not change the steadystate solution A. This process changes the Point-Jacobi method to Richardson's method in standard terminology.0 0.39 is 1 . h = 1. 2h=3)n + c nh(1 . 9. In our ODE b f approach this could also be considered and would lead to a study of equations with nonconstant coe cients. 2p2i)h=9]n~ x n~ c 1 . but it can greatly a ect the convergence rate. are varied at each time step. 2h=3)n.0 λmh Figure 9. The nonstationary form of Eq.0 2 real defective λ m σm 2 complex λ m 1 real λ m Μ=5 0. to study the case of xed H and C but variable step size. M = 5. (10 + 2 2i)h=9] x c (1 . The numerical solution written in full is ~n . ~ b . ]~ x c (1 . It is much simpler. however.

In the annihilation process mentioned after Eq. we considered making the error exactly zero by taking advantage of some knowledge about the discrete values of m for a particular case. We will see that a few well chosen h's can reduce whole clusters of eigenvectors associated with nearby 's in the m spectrum. Mathematically stated.5. 9. This is the basis for the multigrid methods discussed in Chapter 10. 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. Since hn can now be changed at each step.1= m for m = 1 2 : : : M . Now focus attention on the polynomial (Pe)N ( ) = (1 + h )(1 + h ) 1 2 (1 + hN ) (9.1= m .9. the eigenvalues m are generally unknown and costly to compute. Now we pose a less demanding problem. Consider one of the error terms taken from M N ~ N ~ N . we seek b max a j(Pe)N ( )j = minimum with(Pe)N (0) = 1 (9.76.76) where the symbol stands for product.78) where Pe signi es an \Euler" polynomial. 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. the error term can theoretically be completely eliminated in M steps by taking hm = .77) and write it in the form cm~ m Pe( m ) cm~ m x x N Y n=1 (1 + m hn ) (9.79) treating it as a continuous function of the independent variable . for m = 1 2 M . This leads to the concept of selectively annihilating clusters of eigenvectors from the error terms as part of a total iteration process. 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). It is therefore unnecessary and impractical to set hm = . However. ~ 1 = X cm~ m Y (1 + e x m=1 n=1 m hn ) (9.80) .

1 (9.76 is expressed in terms of a set of eigenvectors. cos (2n . + ( . TN . 9.82) q q N N 1 TN (y) = 2 y + y . The three values of h which satisfy this problem are (9. Eq.184 CHAPTER 9.84) b a h 2 b a 2N n Remember that all are negative real numbers representing the magnitudes of m in an eigenvalue spectrum. a. ampli ed by the coe cients cm Q(1 + m hn). RELAXATION METHODS a.86) . b ! .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 . With each x eigenvector there is a corresponding eigenvalue. 1) n = 1 2 :::N (9.83) 2 are the Chebyshev polynomials for jyj > 1. y . ! (Pe)N ( ) = where (9. 9. It is 2 .85) hn = 2= 3 . 1) 6 " #! (9.2 . 9. The error in the relaxation process represented by Eq. 1 + 1 y .84. . As an example for the use of Eq. let us consider the following problem: Minimize the maximum error associated with the eigenvalues in the interval . ~ m . and it leads to the nonstationary step size choice given by ( " #) 1 = 1 .81) TN (y) = cos(N arccos y) are the Chebyshev polynomials along the interval .1 using only 3 iterations. b TN a. ) cos (2n . b a b This problem has a well known solution due to Markov.1 y 1 and 2 2 (9. In relaxation terminology this is generally referred to as Richardson's method.

2 hn =1 =1 1 mC C + ~1 A m (9. 2 hn under the same constraints as before.88) where n p p o T (3) = 3 + 8] + 3 .92) . Eq. 3) h = 4=(6 . Pt: 0 1 1 N B 1 + hn C Y (Pt )N ( ) = B 2 C (9. 9. NONSTATIONARY PROCESSES and the amplitude of the eigenvector is reduced to (Pe) ( ) = T (2 + 3)=T (3) 3 3 3 185 (9.6 and we see that the amplitudes of all the eigenvectors associated with the eigenvalues in the range .6 are h = 4=(6 .87 is given in Fig.37 on the condition that the explicit Euler method. 9. 9. p 0) h = 4=(6 + 3) 1 2 3 p Return now to Eq.5.9. 8] =2 99 3 3 3 A plot of Eq.41. with (Pt)N (0) = 1 (9.90) b max a j(Pt)N ( )j = minimum .91) @ 1 A n 1 .87) (9. was used to integrate the basic ODE's.89) would result. the nonstationary formula 1 h( 0 + 0 ) n 2 n +1 (9. 9.1 have been reduced to less than about 1% of their initial values. This was derived from Eq.2 . 9. If instead the implicit trapezoidal rule n+1 = n+ is used.76. The values of h used in Fig. 9. This calls for a study of the rational \trapezoidal" polynomial. namely that =1 0 1 M N 1+ h ~ N = X cm~ m Y B 2 n B x @ 1 m n 1 .

2 0.2 −1 λ −0.4 −0.4 0.8 0.6 −0.2 0 Figure 9.6 (Pe )3 (λ) 0.8 −0.2 −1 λ −0.015 −0.01 −0. RELAXATION METHODS 1 0.6: Richardson's method for 3 steps.7 0.186 CHAPTER 9.1.01 0. .6 −0.4 −0.8 −1.8 −0.015 0. minimization over .4 −1.5 0.005 −0.9 0.6 −1.1 0 −2 −1.02 0.2 0 0.005 (P ) (λ) e 3 0 −0.2 .3 0.8 −1.4 −1.02 −2 −1.6 −1.

xk ) = (A + A )xk . A for the SOR method with A = B (4 : 1 . A satisfy A~ m = mH~ m . Recall that the eigenvalues of H . A )xn + b x (I + A )xn = (I .2 1) and ! = !opt. but we settle here for the approximation suggested by Wachspress n.6. A )~ + b x where is a parameter. Using Appendix B. A. = N .1=2. 9. b 1 2 2 +1 1 +1 1 2 Determine the iteration matrix G if = . nd the eigenvalues of H .9. 2.93) b hn b This process is also applied to problem 9. x x not just the expressions.) Find the numerical values.6 Problems 1. a n=1 2 N (9. . The values of h used in Fig. 2 =. For a linear system of the form (A + A )x = b. f ~ 0 1 1 2 where G = I + H . PROBLEMS ! 187 The optimum values of h can also be found for this problem.7. 1 1 1 . Gn)A. f show that 1 ~n = Gn ~ + (I . The results for (Pt) ( ) are shown in Fig. The error amplitude is about 1/5 of that found for (Pe) ( ) in the same interval of .85. 3. consider the iterative method (I + A )~ = (I . Show that this iterative method can be written in the form H (xk .2.7 are h = 1 p h = 2 h = 2 ( 1) ( 1) 3 3 1 2 3 9. 9. Then nd the corresponding j m j values. Given a relaxation method in the form ~ H ~n = A~n . (You do not have to nd H .

8 −1.7: Wachspress method for 3 steps.9 0.6 −0.6 −1.8 0.4 −0.7 0.1.6 −0.8 −0.4 −0.2 0 2 x 10 −3 1.5 −2 −2 −1.4 −1.2 −1 λ −0.5 −1 −1.2 .4 −1. . minimization over .2 0 Figure 9.8 −1.4 0.5 0.5 1 0.6 −1.6 (P ) (λ) t 3 0.8 −0. RELAXATION METHODS 1 0.1 0 −2 −1.2 0.188 CHAPTER 9.2 −1 λ −0.3 0.5 (Pt )3 (λ) 0 −0.

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

190 CHAPTER 9. RELAXATION METHODS .

That is.3.3.1.3.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.2 1).3) xx~ = x x 191 2 2 2 2 1 . 10. There are many variations of the process and many viewpoints of the underlying theory. This has a rational explanation from the origin of the banded matrix. if the eigenvalues are ordered such that j j j j 1 2 2 j Mj (10. Notice that as the magnitudes of the eigenvalues increase. ~ (10.1 Motivation 10. the space-frequency (or wavenumber) of the corresponding eigenvectors also increase.2 and 4.1 Eigenvector and Eigenvalue Identi cation with Space Frequencies Consider the eigensystem of the model matrix B (1 . The eigenvalues and eigenvectors are given in Sections 4.2) @x and recall that 1 B (1 .2 1)~ = X 1 D(~ ) X .m sin(mx) (10.1) then the corresponding eigenvectors are ordered from low to high space frequencies. respectively. Note that @ sin(mx) = . The viewpoint presented here is a natural extension of the concepts discussed in Chapter 9.

It is also true. exactly the opposite 2 behavior. One nds that 1 M + 1 . As we saw in Section 9. 10. a sine synthesis. the correlation of large magnitudes of m with high space-frequencies is to be expected for these particular matrix operators.4) m= M +1 x Hence. this method produces the same value of j j for min and max . However. as shown in Fig.1. Therefore. MULTIGRID where D(~ ) is a diagonal matrix containing the eigenvalues. by design.5. However. the complete counterexample of the above association is contained in the eigensystem 1 for B ( 2 1 1 ). This is consistent with the physics of di usion as well. the property can be restored by using h < 1. of the Gauss-Seidel method and. 2 2 2 2 2 10.2 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. ~ represents a sine transform.m . For this matrix one nds. this correlation is not necessary in general. of the Richardson method described in Section 9. We have seen that X .1.m m << M (10. error components corresponding to high space frequencies will be reduced more quickly than those corresponding to low space frequencies. and X ~ . for example. This is the key concept underlying the multigrid process. 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. 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. from Appendix B. This is to be expected of an iterative method which is time accurate.192 1 CHAPTER 10.2.2 + 2 cos m . The classical point-Jacobi method does not share this property. In fact.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 . .

~ b ] = 0 f (10. m . The problem is linear. r where ~ = C Ab~ .6 is used. In Eq. The eigenvectors associated with the eigenvalues having largest magnitudes can be correlated with high frequencies on the di erencing mesh. but for clarity we suppose that the three-step Richardson method illustrated in Fig.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~ . 9. as in the example given by Eq. the vector ~ b represents the boundary conditions and the forcing function. of the matrix are all real and negative. The basic assumptions required for our description of the multigrid process are: 1. This form can be arrived at \naturally" by simply replacing the derivatives in the PDE with di erence schemes. or it can be \contrived" by further conditioning.5) where the matrix formed by the product CAb has the properties given above. 2. ~ b] r f (10. 5. 9. as in the examples given by Eq. 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. 10. 3. if f ~ 1 is a vector representing the desired exact solution. We do not attempt to develop an optimum procedure here. 3. THE BASIC PROCESS 193 basic properties.7) where A CAb (10.11.8) . and initial guess for ~ 1 and proceed through n iterations making use of some iterative process that satis es property 5 above. At the end of the three steps we nd ~ .27. We start with some any. The eigenvalues. The m are fairly evenly distributed between their maximum and minimum values.5. we are led to the starting formulation C Ab ~ 1 .2. ~ = 0 e r (10.10. Having preconditioned (if necessary) the basic nite di erencing scheme by a procedure equivalent to the multiplication by a matrix C . 4. the residual. These conditions are su cient to ensure the validity of the process described next.

since a permutation matrix has an inverse which is its transpose.11) Multiply Eq. 10.10) } Combining our basic assumptions. 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. P ]~ = P ~ e r (10.7 from the left by P and. Next we construct a permutation matrix which separates a vector into two parts. 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. 10.14) . assumption 4 ensures that the error has been smoothed.12) 1 The operation PAP . 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).9) Thus our goal now is to solve for ~ . 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 (10.7 for ~ then e ~1 = ~ . 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. we can write PA P .13) Notice that e r e A ~e + A ~o = ~e 1 2 (10.194 If one could solve Eq.~ e CHAPTER 10. e In addition.

in this formulation. 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.16) 0 0 1 With this approximation Eq. 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.18) (10. 10.17) (10. notice that. r If the boundary conditions are Dirichlet. For example 1 (e + e ) e 2 a 1 (e + e ) e 2 1 (e + e ) or ~ o = A0 ~ e e e (10. and one can write for the example case 2 1 0 0 16 1 1 0 A02 = 2 6 0 1 1 6 4 3 7 7 7 5 is an exact expression.10. e no operations on ~ are required or justi ed. Hence.15) e 2 1 (e + e ) e b 2 1 2 3 2 4 5 4 6 2 7 6 (10. the averaging of ~ is our only approximation. no aliasing of these functions can occur in subsequent steps. It is that there is some connection between ~ e and ~ o brought about by the smoothing property of the Richardson e e relaxation procedure. Since the top half of the frequency spectrum has been removed. ea and eb are zero. Finally.19) or Ac~ e . At this point we make our one crucial assumption. it is reasonable to suppose that the odd points are the average of the even points. ~ e = 0 e r where Ac = A + A A0 ] 1 2 2 .14 reduces to A ~ e + A A0 ~ e = ~ e e e r 1 2 2 (10.2.

2 6 6 6 6 6 6 . the matrix on the coarse mesh.21) 5 If the boundary conditions are mixed Dirichlet-Neumann.1 1=2 7 5 1=2 . B.15 must be changed.18 gives z }| 2 . ea = e .2 4 A1 . the example in Eq. If Neumann conditions are on the left.1]T .1 { 3 7 (10.2 .55 for Dirichlet conditions.2 { z }| 3 2 1 1 7+6 1 1 5 4 A2 1 1 { 3 7 5 2 z }| 2 1 16 1 1 6 4 26 1 1 A 0 1 { Ac 3 2 }| z .1.2 6 . 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. eb is equal to eM . the interpolation formula in Eq. the eigenvector structure is di erent from that given in Eq.1. is completely determined by the choice of the permutation matrix and the interpolation approximation.2 1 1 1 1 1 1 . However. our 7-point example would produce 2 . A in the 1-D model equation is B (1 ~ 1) where ~ = .2 . All of them go through zero on the left (Dirichlet) side. In the particular case illustrated in Fig. In the present case they are given by " !# (2m . The eigensystem is given by b b Eq. 10. 10. If the original A had been B (7 : 1 .1 1=2 7 6 7 = 4 1=2 . 1) xjm = sin j 2M + 1 m=1 2 M (10. MULTIGRID The form of Ac. 9.19. and all of them re ect on the right (Neumann) side.16 changes to 2 1 0 03 7 16 A0 = 2 6 1 1 0 7 60 1 17 (10.2 ::: .2 1 1 1 . all of the properties given in Section 10.1 = 6 PAP 6 1 6 6 6 1 6 6 4 .23) 4 5 0 0 2 1 2 .2 . 10. When eb = eM .196 CHAPTER 10.20) and Eq.2 . and.22) and are illustrated in Fig.2 . in fact. For Neumann conditions.2 1).1 are met. 10. It is easy to show that the high space-frequencies still correspond to the eigenvalues with high magnitudes.

(10. Recall that our goal is ne mesh to 2 e r to solve for ~ . and thus ~ .2 1 .2." We will continue with Dirichlet boundary conditions for the remainder of this Section.1 7 5 1=2 . are unchanged (only A is modi ed by putting .9.2 1)~e = ~ e on the next coarser mesh. Ac = Xc cXc. 10.15. cos m=1 2 M M +1 M +1 1 1 .1: Eigenvectors for the mixed Dirichlet{Neumann case.26) ~ m = sin j m x m = . At this stage.1 1=2 7 6 7 = 4 1=2 . we have reduced the problem from B (1 .2 6 . we can construct the coarse matrix from 1 1 2 4 }| z 2 .24) 5 which gives us what we might have \expected.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 . Therefore.10. we need to consider the eigensystems of A and Ac: A = X X. Given ~ e e e computed on the coarse grid (possibly using even coarser grids). we must now determine e ~ e and ~ .1 in the lower right element). the relationship between e e In order to examine this relationship. which will provide us with the solution ~ 1 using Eq. The permutation matrix remains the same and both A and A in the partitioned matrix PAP .2 4 A1 .2 1)~ = ~ on the e r 1 B (1 . we can compute ~ o e using Eq. In order to complete the process.2 1) the eigenvalues and eigenvectors are m j=1 2 M (10. 10.1=2 { 3 1=2 7 (10. THE BASIC PROCESS 197 X Figure 10.25) For A = B (M : 1 .

(~ c) e x c r c (10. cos M2+ 1 (~ c) = sin j M2+ 1 j = 1 2 x Mc (10..31) 2 3 1 607 6 7 = Xc . one can easily see that (~ c) coincides with ~ at x x every second point of the latter vector.. it contains the even elements of ~ . we obtain ( c) = . ( c) approaches 2 .32) 2 3 1=( c) 6 0 7 6 7 = Xc 6 .0:003649. ~ = ~ . Xc. = .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. x Now let us consider the case in which all of the error consists of the eigenvector component ~ . cos M + 1 x j=1 2 M (10. then Mc = (M .30) since (~ c) contains the even elements of ~ .e. i. Then the residual is x e x ~ = A~ = ~ r x x (10. 1 . 7 6 .2 1 . As M increases.198 CHAPTER 10.33) 1 1 1 1 1 1 1 1 1 . 1)=2 is the size of Ac.28) For M = 51 (Mc = 25). 7 4 5 0 (10. 5 0 (10. MULTIGRID Based on our assumptions..27) M +1 For example. that is.2 1) are 2 ( c) = . corresponding to ~ = sin j = . 6 . 7 c 6 7 4 . If we restrict our attention to odd M . the most di cult error mode to eliminate is that with m = 1. In addition. The eigenvalue and eigenvector corresponding to m = 1 for the matrix Ac = 1 B (Mc 1 . The exact solution on the coarse grid x x satis es ~ e = A. ~ e = Xc . with M = 51.0:007291 = 1:998 .

39) Applying the discretization on the coarse grid with the same preconditioning matrix as used on the ne grid gives. 10. since xc = 2 x. .2 1) = x B (Mc : 1 . in each case the appropriate permutation matrix and the interpolation approximation de ne both the down. but they vary depending on the problem and the user. the matrix A = B (M : 1 . However. x C x B (Mc : 1 .35) Since our goal is to compute ~ = ~ .40) which is precisely the matrix appearing in Eq.37.2 1) and the preconditioning matrix C is simply 2 (10.10.38) C= x I 2 (10.2 1)~ = ~ ee re (10. THE BASIC PROCESS = ( c) 1 1 199 (~ c) x 1 1 1 (~ ) 2 xc 1 (10. The remaining steps required to complete an entire multigrid process are relatively straightforward.36) 1 B (M : 1 . This is equivalent to solving or 1A ~ = ~ e r 2 ce e (10. 10.2 1) comes from a discretization of the di usion equation. quite important but they are not discussed here. which gives Ab = x B (M : 1 .34) (10. The reduction can be. we must multiply the result by 2.and up-going paths. Thus we see that the process is recursive.2 1) 4 c c 2 2 2 (10. obviously.37) c 4 In our case. The details of nding optimum techniques are. and usually is. carried to even coarser grids before returning to the nest level.15).2 1) = 1 B (Mc : 1 . The problem to be solved on the coarse grid is the same as that solved on the ne grid. in addition to interpolating ~ e to the ne grid e x e (using Eq.2.

42) and H is de ned as in Chapter 9 (e. f . f (10. Gn1 ) A.g. 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. starting with ~ = ~n. . Next compute the residual based on ~ : ~ ~ ~ ~ = A ~ . Eq. f r ~ = AGn1 ~n .11. (1) (2) 2 (1) 1 2 1 3.200 10.. f ~ (10.46) 1 2 See problem 1 of Chapter 9.41) 1 (1) 1 (1) 1 1 1 where G = I + H. MULTIGRID ~ We now describe a two-grid process for the linear problem A ~ = f .44) In our example in the preceding section." Some form of weighted restriction can also be used. Transfer (or restrict) ~ to the coarse grid: r ~ =R ~ r r (10. This type of restriction is known as \simple injection. which can be easily generalized to a process with an arbitrary number of grids due to the recursive nature of multigrid. This gives ~ = Gn1 ~n + (I . A (10. ~ e r 2 (2) (2) (2) 2 1 (2) 2 (10.45) 4 5 0 0 0 0 0 1 0 that is. f = AGn1 ~n + A (I . Note that the coarse grid matrix denoted A2 here was denoted Ac in the preceding section. 1. the rst three rows of the permutation matrix P in Eq. AGn1 A. Perform n iterations of the selected relaxation method on the ne grid.3 A Two-Grid Process CHAPTER 10. 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. Call the result ~ . Solve the problem A ~ = ~ on the coarse grid exactly: e r ~ = A. Gn1 ) A.21).43) 1 1 1 1 1 (1) (1) (1) 1 1 1 1 1 1 2.

10. Extension to four or more grids proceeds in similar fashion. Otherwise. I A. G )A. 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. . A TWO-GRID PROCESS 2 2 201 Here A can be formed by applying the discretization on the coarse grid. I ~ e +1 (1) 1 (2) 2 (10. I and R are the transfer operators between grids 2 and 3. and A is obtained by discretizing on grid 3. R A]Gn1 ~ n .2 1). Transfer (or prolong) the error back to the ne grid and update the solution: ~n = ~ .40).3. 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. with (I .47) 3 7 7 7 7 7 7 7 7 7 7 7 5 In our example.48) which follows from Eq. A = 4 B (Mc : 1 .10. The basic iteration matrix for a three-grid process is found from Eq. 10. one obtains ~n = I . f ~ ~ +1 1 2 2 1 2 1 1 1 2 2 1 2 1 1 1 1 (10. solve exactly. 4.49) (10. where 2 2 3 2 G = I . Combining these steps. I A. f + A. I A.51) In this expression n is the number of relaxation steps on grid 2.50 by replacing A. R A]Gn1 A.50) Thus the basic iteration matrix is I . It is at this stage that the generalization to a multigrid procedure with more than two grids occurs. 10. . R A ]Gn2 2 3 2 3 3 1 3 2 2 2 2 2 3 3 2 3 (10. I . If this is the coarsest grid in the sequence. In the 1 preceding example (eq.15.

5. the number of iterations. Repeat problem 4 of Chapter 9 using a four-grid multigrid method together with (a) the Gauss-Seidel method. 2 . Derive Eq. and 4 orders of magnitude. 10.4 Problems CHAPTER 10. Plot the solution after the residual is reduced by 2. Determine the asymptotic convergence rate. MULTIGRID 1. 3. 2. Calculate the theoretical asymptotic convergence rate and compare. (b) the 3-step Richardson method derived in Section 9.202 10.51. Plot the logarithm of the L -norm of the residual vs. Solve exactly on the coarsest grid.

the eigenvalues of the associated circulant matrix (with periodic boundary conditions) are pure imaginary. unless the relevant length scales are resolved. there can be a continual production of high-frequency components of the solution.e. We have seen that a centered approximation to a rst derivative is nondissipative. some form of added numerical dissipation is required in the numerical solution of the Navier-Stokes equations as well. such as the Euler and Navier-Stokes equations. 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. leading. Although numerical approximations to the Navier-Stokes equations contain dissipation through the viscous terms. to the production of shock waves. In this Chapter. However. 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. especially at high Reynolds numbers. we have neglected viscous e ects. we have emphasized the second-order centered-di erence approximations to the spatial derivatives in our model equations. 203 .. In processes governed by nonlinear equations. i. the production of high frequencies is eventually limited by viscosity. In a real physical problem. Therefore. due to the limited grid resolution which is practical. when we solve the Euler equations numerically.Chapter 11 NUMERICAL DISSIPATION Up to this point. this can be insu cient. Since the addition of numerical dissipation is tantamount to intentionally introducing nonphysical behavior. for example.

(1 + )uj.204 11.1 + 2 uj + (1 .ax .1 .2) The second form shown divides the operator into an antisymmetric component (.a x m 1 . cos 2M m + i sin 2M for m = 0 1 : : : M .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. the solution will either grow or decay with time.uj. For periodic boundary conditions the corresponding matrix operator is .uj.1.a @u @t @x (11. ) The eigenvalues of this matrix are m = .uj. the roles are reversed. The antisymmetric component is the second-order centered di erence operator. With 6= 0. Consider the following point operator for the spatial derivative term . Hence the forward di erence operator is inherently unstable while the centered and backward operators are inherently stable.a x = 2. our choice of di erencing scheme produces nonphysical behavior.a( x u)j = 2. )uj+1] = .1 + 2uj . In either case.1 + 2uj . .1+ uj+1)=2 x and a symmetric component (. the forward di erence operator ( = . A backward di erence operator is given by = 1 and a forward di erence operator is given by = .1) produces Re( m ) > 0. When Re( m) = 0. uj+1)] 2 x (11.1) with periodic boundary conditions.a (. 1 If a is positive. NUMERICAL DISSIPATION @u = . We proceed next to show why this occurs.uj. the centered di erence operator ( = 0) produces Re( m ) = 0. uj+1)=2 x.ax Bp(. the operator is only rst-order accurate. and the backward di erence operator produces Re( m) < 0. 2 1 .1 + uj+1) + (. If a is negative.

11. 11.3 is excited to some degree.5) | | amplitude phase . 11. a x2 @ 3 u + a x3 @ 4 u + : : : (11. Consider the simple linear partial di erential equation @u = .1. 11. 2} ei x. 2 .ia . (a + 2 ) The solution is composed of both amplitude and phase terms.3 as the modi ed partial di erential equation. 2 ) s = .({z+ 2 )t}] (11.2.1 gives @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.4) @t @x @x2 @x3 @x4 Choose periodic boundary conditions and impose an initial condition u = ei x. 11. i 3 + 4 or r = . x can be small but it is not zero. We are lead to the expression 2 ! 3 ! ! ! @ 2 u + x3 @ 3 u . since the two equations are identical when x ! 0. 2 ( . 11.a @u + @ 2 u + @ 3 u + @ 4 u (11. However. 11.11. Under these conditions there is a wave-like solution to Eq.4 of the form u(x t) = ei xe(r+is)t provided r and s satisfy the condition r + is = . @x j 205 First carry out a Taylor series expansion of the terms in Eq.a @u + a x @ 2 u . Notice that Eq.2 to Eq.3) @t @x 2 @x2 6 @x3 24 @x4 This is the partial di erential equation we are really solving when we apply the approximation given by Eq.2. 11. 2 ({z. We refer to Eq. Thus a ) u = e. when we use a computer to nd a numerical solution of the problem. Substituting this into Eq. We proceed next to investigate the implications of this concept. THE MODIFIED PARTIAL DIFFERENTIAL EQUATION 11. This means that each term in the expansion given by Eq.2 The Modi ed Partial Di erential Equation ( xu)j = 2 1 x 42 x @u .3 is consistent with Eq.1.

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

cos 2 m . giving !2 (n+1) = u(n) .1 2 x This is the Lax-Wendro method applied to the linear convection equation.3. the linear convection equation @u + a @u = 0).a @u (11. u(n) ) + 1 ah (u(n) . 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 .11. It is a fully-discrete nite-di erence scheme. Thus @t @x @ 2 u = a2 @ 2 u @u = . . 1 ah (u(n) . The quantity j ah j is known as the x Courant (or CFL) number. This explicit method di ers conceptually from the methods considered previously in which spatial di erencing and time-marching are treated separately.7) First replace the time derivatives with space derivatives according to the PDE (in this case. 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. 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.3 The Lax-Wendro Method 207 In order to introduce numerical dissipation using one-sided di erencing. the corresponding fully-discrete matrix operator is 0 2 !3 ! 2 ! 31 1 4 ah + ah 25 1 . 2u(n) + u(n) ) (11.8) @t @x @t2 @x2 Now replace the space derivatives with three-point centered di erence operators. There is no intermediate semi-discrete stage. x M x M for m = 0 1 : : : M .9) uj j j +1 j j . backward di erencing must be used if the wave speed is positive. ah 2 1 4. and forward di erencing must be used if the wave speed is negative. THE LAX-WENDROFF METHOD 11. Next we consider a method which introduces dissipation independent of the sign of the wave speed. It is equal to the ratio of the distance travelled by a wave in one time step to the mesh spacing. i ah sin 2 m m =1. For periodic boundary conditions. known as the Lax-Wendro method.1 2 x j+1 j.

Therefore. 11. Recall that the odd derivatives on the right side lead to unwanted dispersion and the even derivatives lead to dissipation (or ampli cation.8. A closely related method is that of MacCormack. which is given by @u + a @u = . 11. ah (u(jn) . C 2 ) @ 3 u . presented in Chapter 6: un+1 = un + hu0n ~ 1 ~ ~ (11. 1 Or vice-versa for nonlinear problems. Hence MacCormack's method has the same dissipative and dispersive properties as the Lax-Wendro method. u(jn+1))] ~ (11. a x2 (1 . . a2h2 ) @ 3 u .9 and converting the time derivatives to space derivatives using Eq. ah (~(jn+1) . this approach leads to u(jn+1) = u(jn) . a2h2 ) @ 3 u = .1 a dissipative second-order method is obtained. Cn) @xu 4 @x4 This term has the appropriate sign and hence the scheme is truly dissipative as long as Cn 1. the leading error term in the Lax-Wendro method is dispersive and proportional to a ( x2 . a2 h2) @ 4 u + : : : @t @x 6 @x3 8 @x4 This is derived by substituting Taylor series expansions for all terms in Eq. u(jn)1) ~ . however. a8h ( x2 . a2h2 ) @ u = . depending on the sign).208 CHAPTER 11. these should be applied alternately. The two methods di er when applied to nonlinear hyperbolic systems.6 n @x3 @x3 6 The dissipative term is proportional to 2 4 2 2 4 2 @ . For the linear convection equation.11) 2 x u +1 ~ which can be shown to be identical to the Lax-Wendro method. 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. a ( x2 . x u(jn+1) = 1 u(jn) + u(jn+1) . Recall MacCormack's timemarching method.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. a h8 x (1 . a2 h ( x2 . The two leading error terms appear on the right side of the equation.

these are of mixed sign. In the next two sections. schemes in which the numerical dissipation is introduced in the spatial operator are generally preferred over the Lax-Wendro approach.1.1 + uj+1) + jaj(.uj. 11. We can rewrite Eq.11. a. we saw that numerical dissipation can be introduced in the spatial di erence operator using one-sided di erence schemes or. When a hyperbolic system of equations is being solved. Alternatively. more generally. The above approach to obtaining a stable discretization independent of the sign of a can be written in a di erent.1 if a 0. the wave speeds can be both positive and negative. ( 0) term forward di erence.. we see that a stable discretization is obtained with = 1 if a 0 and with = . if a 0. uj+1)] (11. These are by no means unique. When the ow is subsonic.e. then a a. In order to apply one-sided di erencing schemes to such systems. as a result of their superior exibility. 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. = a 0. ) @u = 0 a = a 2 jaj @t @x + = a 0 and a. by adding a symmetric component to the spatial operator. we present two splitting techniques commonly used with upwind methods. that is.uj. the eigenvalues of the ux Jacobian for the onedimensional Euler equations are u u + a u . Consider again the linear convection equation: @u + a @u = 0 (11.4. then a+ = 0 and If a 0. = 0.4 Upwind Schemes 209 In Section 11. This is avoided in the Lax-Wendro scheme. the direction of the one-sided operator (i. but entirely equivalent.2. This is the basic concept behind upwind methods. we present the basic ideas of ux-vector and ux-di erence splitting.12 as @u + (a+ + a. For more subtle aspects of implementation and application of such techniques to . In this section.12) @t @x where we do not make any assumptions as to the sign of a. For example.13) This approach is extended to systems of equations in Section 11. some form of splitting is required. 11. However.a( xu)j = 2. UPWIND SCHEMES 11.1 + 2uj .5. Now for the a+ ( 0) term we can safely backward di erence and for the a. This is achieved by the following point operator: . From Eq. With this approach.1x a(. 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. manner.

We can now rewrite the system in terms of characteristic variables as @w + @w = @w + + @w + . = .14) @t @x @t @x can be decoupled into characteristic equations of the form @wi + @wi = 0 (11. the reader is referred to the literature on this subject. is positive. i.j j (11.19) . NUMERICAL DISSIPATION nonlinear hyperbolic systems such as the Euler equations.16) +j j . contains the negative eigenvalues. Premultiplying by X and inserting the product @x X .4. A.1X in the spatial terms gives @Xw + @X +X . hyperbolic system of partial di erential equations given by @u + @f = @u + A @u = 0 (11. + contains the positive eigenvalues and .18) @t @x @t @x @x The spatial terms have been split into two components according to the sign of the wave speeds. and a forward di erence if the wave speed is negative. @w = 0 (11.17) 2 2 With these de nitions. we need to apply a backward di erence if the wave speed. let us split the matrix of eigenvalues. @w term. are the eigenvalues of the Jacobian matrix. i. and the wi's are the characteristic variables. 11. To accomplish this. constant-coe cient. .1Xw + @X .210 CHAPTER 11. We can use backward di erencing for the + @w term and forward @x di erencing for the .1Xw = 0 @t @x @x (11. In order to apply a one-sided (or biased) spatial di erencing scheme.1 Flux-Vector Splitting Recall from Section 2. where += (11.X .5 that a linear.15) i @t @x where the wave speeds. into two components such that = ++ .

the de nition of the split uxes follows directly from the de nition of the ux. @u @u Therefore. f = Au. When an implicit time-marching method is used. f is also equal to Au as a result of their homogeneous property.1 (11. the Jacobians of the split ux vectors are required.X . = 0 (11. For the Euler equations. 6= A. = A.4.24) Thus by applying backward di erences to the f + term and forward di erences to the f .20) @u + @A+u + @A.21) (11. u @u + @f + + @f .11.22) f . term. = X . . and A.1 and recalling that u = Xw. (11. one must nd and use the new Jacobians given by A++ = @f @u + (11. = @f @u . has all negative eigenvalues. 2 With these de nitions A+ has all positive eigenvalues.26) For the Euler equations. we obtain A. (11.25) A.23) @t @x @x In the linear case. as discussed in Appendix C. we are in e ect solving the characteristic equations in the desired manner. has all negative eigenvalues. u = 0 @t @x @x Finally the split ux vectors are de ned as f + = A+u and we can write (11... A++ has eigenvalues which are all positive. This approach is known as ux-vector splitting. and A. Note that f = f+ + f. In the nonlinear case. UPWIND SCHEMES With the de nitions2 211 A+ = X +X . @f + 6= A+ @f .

which is nondissipative. gj+1=2. 11. respectively. hyperbolic system of equations @w + @w = 0 (11. wR ) ^ (11.4. more suited to nite-volume methods. The centered approximation to gj+1=2. We again begin with the diagonalized form of the linear.1 (11.32) and thus 1 f^j+1=2 = 1 (fL + fR ) + 2 jAj (uL . as in Chapter 5.1X after and j j to obtain 1 1 X gj+1=2 = 2 X X .1X (wL + wR) + 2 X j jX .29) i i R if i < 0 where the subscript i indicates individual components of w and g. we consider the numerical ux at the interface between nodes j and j + 1.30) or 1 gj+1=2 = 1 (wL + wR) + 2 j j (wL .34) . Now. constant-coe cient. is given by 1 gj+1=2 = 2 (g(wL) + g(wR)) ^ (11.2 Flux-Di erence Splitting Another approach. wR ) ^ (11. as a ^ function of the states to the left and right of the interface. uR) (11. wL and wR.33) 2 where jAj = X j jX . as in Eq.28) In order to obtain a one-sided upwind approximation. In a nite-volume method. the uxes must be evaluated at cell boundaries.31) 2 Now.212 CHAPTER 11. NUMERICAL DISSIPATION 11. we premultiply by X to return to the original variables and insert the product X . is known as ux-di erence splitting.19. we require ( (w ) (^i)j+1=2 = i(wi)L if i > 0 g (11. This is achieved with 1 1 (^i)j+1=2 = 2 i (wi)L + (wi)R ] + 2 j ij (wi)L .1X (wL .27) @t @x The ux vector associated with this form is g = w. (wi)R] g (11.

this leads to an upwind operator which is identical to that obtained using ux-vector splitting. jAj = .37) For the Euler equations for a perfect gas. 11.5. However. In order to resolve this. fR = Aj+1=2 (uL . 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. 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. jAj = A if they are all negative. 11.35) R if all i s < 0 giving pure upwinding. there is some ambiguity regarding the de nition of jAj at the cell interface j + 1=2.38) uj+1=2 = p + p L R p H +p H L L Hj+1=2 = (11. uR) (11. u = Xw. consider a situation in which the eigenvalues of A are all of the same sign. Hence satisfaction of Eq.A.37 is satis ed by the ux Jacobian evaluated at the Roe-average state given by p LuL + p R uR (11. In the linear.11. and A = X X . We have also seen that such dissipation can be introduced by adding a symmetric component to an antisymmetric (dissipation-free) operator.1.35 is obtained by the de nition 1 f^j+1=2 = 2 (fL + fR ) + 1 jAj+1=2j (uL .3 ( 11. ARTIFICIAL DISSIPATION 213 and we have also used the relations f = Xg.39) p L+p R R R where u and H = (e + p)= are the velocity and the total enthalpy per unit mass. 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. 3 Note that the ux Jacobian can be written in terms of u and H only see problem 6 at the end . of this chapter.36) 2 if Aj+1=2 satis es fL . Eq. If the eigenvalues of A are all positive. uR) (11. in the nonlinear case. constant-coe cient case. respectively.

43.214 For example.44) x (uj.34. uj. 11. The appropriate implementation is thus a s (11. 4uj.1 (11. let CHAPTER 11. A more complicated treatment of the numerical dissipation is also required near shock waves and other discontinuities. NUMERICAL DISSIPATION s a ( xu)j = . as in the previous two sections. The second spatial term is known as arti cial dissipation.41) i x = i x + j ij x Extension to a hyperbolic system by applying the above approach to the characteristic variables. gives a s (11. this approach can be related to the upwind approach. Similarly.15 is stable if i 0 and a s unstable if i < 0. 4uj+1 + uj+2) where is a problem-dependent coe cient. 11. a s With appropriate choices of x and x. s It is common to use the following operator for x (11. 11.6 Problems 1. This symmetric operator approximates x3 uxxxx and thus introduces a third-order dissipative term.uj+1 + 2uj .36 and 11.42) x (Au) = x (Au) + x (jAju) or a s (11. 4uj.43) x f = x f + x (jAju) where jAj is de ned in Eq.1 + 3uj ) 2 x . this often provides su cient damping of high frequency modes without greatly a ecting the low frequency modes. With an appropriate value of .2 . s ( xu)j = 11. x is stable if i 0 and unstable if i > 0.1 2 x 2 x a s Applying x = x + x to the spatial derivative in Eq. A second-order backward di erence approximation to a 1st derivative is given as a point operator by ( xu)j = 1 (uj. applying x = x . uj. but is beyond the scope of this book.1 + 6uj . For details of how this can be implemented for nonlinear hyperbolic systems. the reader should consult the literature.2 . It is also sometimes referred to as arti cial di usion or arti cial viscosity. This is particularly evident from a comparison of Eqs.40) ( x u)j = uj+1 .

2. 11. Consider the hyperbolic system derived in problem 8 of Chapter 2. 2. Show that the ux Jacobian for the 1-D Euler equations can be written in terms of u and H . x for 0 x .2. Plot the real and imaginary parts of x vs. x for 0 x .) (b) Using a Taylor table. Show that the use of the Roe average state given in Eqs.6.37. and 1=48. 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.6. nd the derivative which is approximated by the corresponding symmetric and skew-symmetric operators and the leading error term for each. Find the matrix jAj. then derive the symmetric and skew-symmetric matrices that have the matrix operator as their sum. 11. Plot the real and imaginary parts of x vs. . Using Fourier analysis as in Section 6. 11. 5.44. Form the plus-minus split ux vectors as in Section 11. Find the modi ed wavenumber for the rst-order backward di erence operator. Using Fourier analysis as in Section 6. x for 0 x .3 to see how to construct the symmetric and skew-symmetric components of a matrix.2. PROBLEMS 215 (a) Express this operator in banded matrix form (for periodic boundary conditions).38 and 11. Consider the spatial operator obtained by combining second-order centered differences with the symmetric operator given in Eq. 3. x for 0 x .4.11. x for 0 x . 11. x for 0 x . (See Appendix A.6. 6. Find the modi ed wavenumber for the operator given in Eq. Using Fourier analysis as in Section 6. nd j j for the combination of this spatial operator with 4th-order Runge-Kutta time marching at a Courant number of unity and plot vs. Find the modi ed wavenumber for this operator with = 0 1=12 1=24.6. 4. Plot the real and imaginary parts of x vs.6.39 leads to satisfaction of 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.1.

NUMERICAL DISSIPATION .216 CHAPTER 11.

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

However. From observation 1 (arbitrary splitting of A) : (12. 12. h A A i~ n . from observation 3 (allowing us to drop higher order terms . from observation 2 (new data ~ n in terms of old ~ n): u u 1 2 1 1 +1 ~ n = I + hA + hA ]~ n . 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. h~ + O(h ) u u f +1 1 2 2 1 2 2 1 2 (12. . Choose again the explicit Euler time march so that d~ = A ~ + A ~ + (bc) u ~ cu du dt (12. explicit Euler method.4 have the same formal order of accuracy and.5) ~ ~ n = I + hAd + hAc]~ n + h(bc) + O(h ) u u +1 2 (12.6) 1 Second-order time-marching methods are considered later. respectively and their sum forms the A matrix we have considered in the previous sections. h~ + O(h ) u u f +1 2 1 2 Finally. For the time march let us choose the simple.4) Notice that Eqs.3 and 12. +1 1 2 2 ~ n = I + hA ] I + hA ]~ n .h A A ~ n): u (12.2 Factoring Physical Representations | Time Splitting Suppose we have a PDE that represents both the processes of convection and dissipation.3) 2 or its equivalent ~ n = h I + hA ] I + hA ] . neither one is to be preferred over the other. and techniques to carry out their numerical evaluation can have arithmetic operation counts that vary by orders of magnitude. h~ + O(h ) u u f 12.218 CHAPTER 12. SPLIT AND FACTORED FORMS d~ = A + A ]~ . in this sense.2) where A = A + A ] but A and A are not unique. their numerical stability can be quite di erent. Both of these considerations are investigated later. rst-order. ~ u u f dt 1 2 2 and consider the above observations. Then. Here we seek only to apply to some simple cases the concept of factoring.

7 and the original unfactored form Eq. hAd]~ n u +1 +1 ~ = I + hAc]~ n + h(bc) u = un ~ +1 (12. I + hAc]~ n + h(bc) u ~ I + hAd + hAc]~ n + h(bc) +O(h ) u | {z } 1 2 (12. 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 ) (12.6 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 . this factored form would appear to be superior to that provided by Eq. However.2.8. This time a predictor-corrector interpretation leads to the sequence un ~ I . hAd ]. their selection is arbitrary. on this basis. Since numerical sti ness is generally much more severe for the di usion process. but the di usion operator is now implicit. hAd]. 2 We do not suggest that this particular method is suitable for use. FACTORING PHYSICAL REPRESENTATIONS | TIME SPLITTING 219 Now consider the factored form ~n u +1 Original Unfactored Terms Higher Order Terms and we see that Eq. 12.9) I .6 have identical orders of accuracy in the time approximation. 12.8) Factoring can also be useful to form split combinations of implicit and explicit techniques.7) u {z } | 2 2 | {z } un ~ ~n u +1 +1 = = ~ I + hAc]~ n + h(bc) u I + hAd]~n u +1 (12. requiring a tridiagonal solver if the di usion term is central di erenced. We have yet to determine its . In practical applications equations such as 12. Therefore.12. the important aspect of stability has yet to be discussed. and a rst-order time-march method is usually unsatisfactory. where jjAd jj is some norm of Ad ]. 12. 12. For example. = I + hAd + h Ad + if h jjAdjj < 1. as before.10) The convection operator is applied explicitly. another way to approximate Eq.7 are often applied in a predictor-corrector sequence. stability.

10. the number of (interior) y points is 3. SPLIT AND FACTORED FORMS un . 12. but in these notes we describe the size of a mesh by the number of interior points.9 can be derived for a di erent point of view by writing Eq.12) j Mx Mesh indexing in 2-D. the number of (interior) x points.3. A clear description of a matrix nite-di erence operator in 2.11) My k 1 13 23 33 43 12 22 32 42 11 21 31 41 1 (12. The numbers 11 12 43 represent the location in the mesh of the dependent variable bearing that index. ~ I .and three-dimensional ows.and 3-D requires some reference to a mesh.3 Factoring Space Matrix Operators in 2{D Factoring is widely used in codes designed for the numerical solution of equations governing unsteady two. 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. un = A u + A u + (bc) + O(h ) ~ c n d n h +1 +1 2 We should mention here that Eq.1 Mesh Indexing Convention 2 2 12. 12.220 CHAPTER 12. 12. 6 point mesh if the boundary points (labeled in the sketch) were included. is 4 and My . 3 32 @u = @ u + @ u @t @x @y 2 2 (12. In this example Mx .6 in the form Then which is identical to Eq. 3 This could also be called a 5 . We choose the 3 4 point mesh shown in the Sketch 12. hAd ]un = I + hAc]un + h(bc) +1 12. Thus u represents the value of u at j = 3 and k = 2.12.

depending on the data{base chosen for the U it multiplies. we consider only two: (1).3. we use the notation Axx y or Axy y .12. ( ) ( ) 12.3 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. 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. used in the previous sections. To be speci c about the structure.13) where Pyx = PT = P. and (2). we label a data{base composed of x-vectors with U x . consecutively along rows that are themselves consecutive from k = 1 to My . Thus ( ) + ( ) + ( ) ( ) dU = A U + (bc) ~ x y dt + (12. consecutively along columns that are consecutive from j = 1 to Mx. Of these. 12. We refer to each row or column group as a space vector (they represent data along lines that are continuous in space) and label their sum with the symbol U . however.16 part a and b. the usual (but ambiguous) representation is given by Ax y .15) Ax+y and U .16 part a and b. 12.11 can be written + 4 If it is important to be speci c about the data-base structure. When the matrix is multiplying a space vector U . and they are related by the same permutation matrix that relates U x to U y .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. respectively. are subsets of A and ~ . u 4 Notice that .3.14) Axx y = Pxy Axy y Pyx ( ) + ( ) + (12. Notice that the matrices are not the same although they represent the same derivative operation. 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. FACTORING SPACE MATRIX OPERATORS IN 2{D 221 12. Examples of the order of indexing for these space vectors are given in Eq. 12. Their structures are similar.3. There are many ways to lay out a data-base. In particular. and one composed of y-vectors with U y . which are notations used in the special case of space vectors. Examples are in Eq. (1) and (2) above are referred to as xvectors and y-vectors. In this notation the ODE form of Eq.

central-di erence. for y ! o. matrix operator. Data base composed of Mx y{vectors stored in U y . for 3 4 mesh shown in Sketch 12. for both ! . matrix operator.. central-di erence. .12. Data base composed of My x{vectors stored in U x . . SPLIT AND FACTORED FORMS x j o x j o x x j o x j o j j j j j x j o o j x x j o o j x x j o o j x j j o j o j o j o j x j x x j x j x 3 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 o7 7 7 7 7 7 7 7 7 5 x7 .12.. Entries for x ! x..16) b: Elements in 2-dimensional. for y ! o.222 2 6x 6 6 6 6 6 6 6 6 6 6o 6 6 (x) (x) Ax+y U = 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4 CHAPTER 12. for both ! . Entries for x ! x. Ax y . 41 42 43 31 32 33 21 22 23 11 12 13 (12.. for 3 4 mesh shown in Sketch 12.. + ( ) 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 . 12 22 32 42 13 23 33 43 11 21 31 41 . a:Elements in 2-dimensional. . Ax y .

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

each with Mx blocks of order Ne .14. 2nd-order. 12. our original ODE. Thus. 12. 12. is referred to as the \delta form" and we develop it next. and matrix multiplications do not commute.23 shows that the nal ~ be permuted to U step consists of solving Mx one-dimensional implicit problems each with dimension My . 2 hAx .29 converges. 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. Then. 12. 1 hAy Un = h Ax y Un + (bc) + O(h ) 2 This is the delta form of a factored. 2 hAy Un = I + 2 hAx + 1 hAy Un + h(bc) + O(h ) 2 From both sides subtract 1 I . if Eq. 12. The rst step would be solved using My uncoupled block tridiagonal solvers .24.6 The Delta Form Clearly many ways can be devised to split the matrices and generate factored forms. ( ) ( ) 12.22. 12. 2 hAx I . 2-D equation. for ensuring a correct steady-state solution in a converged time-march. SPLIT AND FACTORED FORMS 9 which has the same appearance as Eq. +1 3 +1 + 3 2 + 3 replace the scalar ones. Consider the unfactored form of the trapezoidal method given by Eq. 1 hAy Un 2 leaving the equality unchanged.30) I . Now we can factor Eq.21 but is second-order time accurate. 2 hAx .228 CHAPTER 12.29 and maintain O(h ) accuracy. 2 hAx . it is guaranteed to converge to the correct steady-state solution of the ODE. One way that is especially useful. 12. 9 A block tridiagonal solver is similar to a scalar solver except that small block matrix operations . Un one nds h 1 1 ~ i I . using the standard de nition of the di erence operator . We arrive at the expression h 1 ~ i (12. we see that this is equivalent to solving My one-dimensional ~ problems. 2 hAy Un = h Ax y Un + (bc) + O(h ) (12. Inspecting the top of Eq.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. Un = Un .

ii. 12. PROBLEMS 229 The point at which the factoring is made may not a ect the order of time-accuracy.19. 12. the delta form of the unfactored Eq.31 is the appearance of the factor 1 on the left side of the O(h2 ) method. 12.P ). (a) Space vector de nition i. 12.30 and the O(h) method given by Eq. u ? iii. ( u) = 1 (u .20. For example. 12. Consider the 1-D heat equation: @u = @ u @t @x 2 2 0 x 9 Let u(0 t) = 0 and u(9 t) = 0.e.12. the factored form is presented in Eq. write the space vector. 2u + u ) xx j The uniform grid has x = 1 and 8 interior points. hAy ] Un = h Ax y Un + (bc) + 10 and its factored form becomes h ~ i I . 12. we will see in the next chapter that the convergence properties of Eq. What is the space vector for the natural ordering (monotonically increasing in index).(P .19 is h ~ i I .7. Assume that second order central di erencing is used.20 and Eq. but it can have a profound e ect on the stability and convergence properties of a method.31 are vastly di erent. 12.31) In spite of the similarities in derivation.7 Problems 1. hAx] I . and if it is immediately factored. (1) (2) 12 21 x 2 j . If we reorder the points with the odd points rst and then the even points. the unfactored form of a rst-order method derived from the implicit Euler time march is given by Eq. 12.1 j j +1 10 Notice that the only di erence between the O(h2 ) method given by Eq. i. 2 . so that we can simplify the boundary conditions. hAx . u ? Only include the interior points. On the other hand. Write down the permutation matrices.. hAy ] Un = h Ax y Un + (bc) + (12.

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

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

232 CHAPTER 12. SPLIT AND FACTORED FORMS .

and (approximately) factored. if the results indicate that a method has an instability. accuracy. and we discussed circulant eigensystems1 in Section 4. 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. We have seen that under these conditions a semi-discrete approach can lead to circulant matrix di erence operators.7. the method is probably not suitable for practical use. hybrid. 233 . fractional-step. When these methods are applied to practical problems. and used it in Chapter 7 to study the stability. However.Chapter 13 LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS In Section 4. 13. 1 See also the discussion on Fourier stability analysis in Section 7. and convergence properties of timemarching schemes.3. the results found from this analysis are neither necessary nor su cient to guarantee stability.1 The Representative Equation for Circulant Operators Consider linear PDE's with coe cients that are xed in both space and time and with boundary conditions that are periodic.4 we introduced the concept of the representative equation. 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. 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.

we arrive at a set of ODE's that can be written d~ = A ~ + A ~ . as a result of space di erencing the PDE..2.3 to uncouple the set and form the M set of independent equations 0 w1 = ( a + b)1 w1 . we can use the arguments made in Section 4.1) apu b pu f dt where the subscript p denotes a circulant matrix. 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. 0 wm = ( a + b)m wm . ~ (t) u (13. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS Suppose. 0 wM = ( a + b)M wM . and only one2 . . (13.4: b since they must share a common The representative equation for split.2.. This suggests (see Section 4.3) that 13. Since both matrices have the same set of eigenvectors.234 CHAPTER 13. gM (t) (13. g1 (t) .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 (13.4) This is to be contrasted to the developments found later in the analysis of 2-D equations. gm(t) .2) The analytic solution of the m'th line is wm(t) = cm e( a+ b)mt + P:S: Note that each a pairs with one. eigenvector.2 Example Analysis of Circulant Systems 13...

2 1)~ u (13.10. 12. where (see Section 4.3.6 to quantify the eigenvalues. so that 0 m 2 . 2.6) ( d)m = .8.2): ( c)m = ia sin m x m (13. 12. Eq. we can analyze the stability of the two forms of simple time-splitting discussed in Section 12. 4 2 sin2 2 x In these equations m = 2m =M .4. Now introduce the dimensionless numbers Cn = ah Courant number x mesh Reynolds number R = a x and we can write for the absolute value of q 2 1 + Cn sin2 m j j= C m 1 + 4 R n sin2 2 0 m 1. the explicit-implicit Euler method. In this section we refer to these as 1. the characteristic polynomial of this method is P (E ) = (1 . 13. 1 . Eq. The Explicit-Implicit Method 2 (13. m = 0 1 M .2. 13. EXAMPLE ANALYSIS OF CIRCULANT SYSTEMS 235 If the space di erencing takes the form d~ = .13. When applied to Eq. respectively.5) dt 2 x p x2 p the convection matrix operator and the di usion matrix operator. (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. can be represented by the eigenvalues c and d.7) . a B (. h d )E .4.2. the explicit-explicit Euler method. Using these values and the representative equation 13.1 0 1)~ + u u B (1 .

one near 0.2 C = 2/ R ∆ n 0. 4 Cn sin2 m 2 j j = 1 + Cn sin 0 m 2 R 2 Again a simple numerical parametric study shows that this has two critical ranges of m .1: Stability regions for two simple time-split methods. 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.4 0. 13.2 C = R /2 n ∆ C = 2/ R n ∆ 0. and the .1. 13. which yields the same result as in the previous example. 2.4 0 0 4 R∆ Explicit-Implicit 8 0 0 4 R∆ Explicit-Explicit 8 _ Figure 13. A simple numerical parametric study of Eq.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 ). LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 1.236 CHAPTER 13.8 C n C 0. The Explicit-Explicit Method An analysis similar to the one given above shows that this method produces " # q 2 m 1 .

13.4 which we split in the fashion of Eq. thus 1 1 un+1 = un + 2 h c(3un . factored method has a much smaller region of stability when R is small.2.9) Q(E ) = 1 h(E 2 + E ) 2 13. 13. The di usion term is integrated implicitly using the trapezoidal method. un. Next let us analyze a more practical method that has been used in serious computational analysis of turbulent ows. as well as the stability.8) and in the latter (13. 13. P (E ) and Q(E ).1. In the former case it is 1 Q(E ) = 2 h(3E .1) + 2 h d (un+1 + un) This produces 3 P (E ) = (1 . which produces the constraint that 1 Cn < 2 R for R 2 The resulting stability boundary is also shown in Fig.2 Analysis of a Second-Order Time-Split Method . EXAMPLE ANALYSIS OF CIRCULANT SYSTEMS 237 other near 180o. time-marching method on the equation u0 = cu + du + ae t First let us nd expressions for the two polynomials. This method applies to a ow in which there is a combination of di usion and periodic convection. 1) (13. The characteristic polynomial follows from the application of the method to the homogeneous equation. The totaly explicit. 1 h d)E 2 . as we should have expected. In order to evaluate the accuracy. we include the forcing function in the representative equation and study the e ect of our hybrid.5.2. Our model equation is again the linear convection-di usion equation 13. The convection term is treated explicitly using the second-order Adams-Bashforth method. (1 + 2 h c + 1 h d)E + 1 h c 2 2 2 The form of the particular polynomial depends upon whether the forcing function is carried by the AB2 method or by the trapezoidal method.

10 and shown in Fig. the hybrid method retains the second-order accuracy of its individual components. 2 h d (13.10) Stability The stability of the method can be found from Eq. .8 or Eq. 2 d . so er = O(h3) Using P (e h) and Q(e h ) to evaluate er in Section 6. 1 1 + 3h c + 2h 2 2 1 .11) The extension of the following to 3-D is simple and straightforward.2. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS Accuracy From the characteristic polynomial we see that there are two {roots and they are given by the equation 3 1 + 2h c + 1h 2 = s d The principal -root follows from the plus sign and one can show 1 1 1 = 1 + ( c + d )h + 2 ( c + d )2 h2 + 4 3 + c 2 . 13.7. 13. These results show that. 13. For values of R 2 this second-order method has a much greater region of stability than the rst-order explicit-implicit method given by Eq.10 by a parametric study of cn and R de ned in Eq.3.9. 1h d 2 d 2 1 . 2h c 1 . 13.1. The results are plotted in Fig. 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. 13.3 The Representative Equation for Space-Split Operators Consider the 2-D model3 equations @u = @ 2 u + @ 2 u @t @x2 @y2 3 (13. for the model equation. 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 . 13. one can show er = O(h3) using either Eq.6.2. 12. 13.1.238 CHAPTER 13.

1 b0 b1 ). and @u + a @u + a @u = 0 (13.13. The form of the Ax and Ay matrices for three-point central di erencing schemes are shown in Eqs. 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 ~.23 for the 3 4 mesh shown in Sketch 12.12) @t x @x y @y Reduce either of these.3.22 and 12.12. Let us inspect the structure of these matrices closely to see how we can diagonalize Ax + Ay ] in terms of the individual eigenvalues of the two matrices considered separately.2: Stability regions for the second-order time-split method. Now nd the block eigenvector .8 Cn 0.1 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 ~. to the coupled set of ODE's: dU = A + A ]U + (bc) ~ (13.13) x y dt for the space vector U . THE REPRESENTATIVE EQUATION FOR SPACE-SPLIT OPERATORS239 1.2 0. 12.4 Stable 0 4 8 ∆ R _ Figure 13. by means of spatial di erencing approximations.

12. see the bottom of Eq.1 A(xx) + A(yx) X Pxy = 3 2 3 2 ~ B 1 I 7 6 7 6 ~ 7 B 6 7+6 2 I 6 7 6 7 6 (13.1 ~0 ~1). 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 ~. 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 .240 CHAPTER 13.1 4b I I .1 I ~0 I ~.14) ~ 4 5 4 B 7 3 I 5 ~ B 4 I ~ where B is the banded tridiagonal matrix B (~.15) 6 7 4 5 3I + ~ 4I + ~ diag(X ) 2 ~ b0 X . 12. That is. the rst matrix on the right side of Eq. There results h i Pyx X . LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS matrix that diagonalizes B and use it to diagonalize A(xx).1 Pyx X .1 6 ~. b b b ~ that diagonalize the B blocks.1B X = 6 7 4 5 4 5 ~3 ~ This time. by the same argument as before. Thus 2 . 13.1 32 32 3 2 X B X 6 76 7=6 X .23.13.1 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.1 B 76 X 4 54 54 5 4 .1 I ~0 I ~1 I 7 b b 5 b b 5 4b ~. so the nal result is the complete diagonalization of the matrix Ax+y : h ih ih i ~ ~ X .14 is transparent to the transformation.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.1 A(xx+)y X Pxy X = 2 3 1I + ~ 6 7 6 7 2I + ~ 6 7 (13.

this fact.18) x y We have used 3-point central di erencing in our example. a and are (possibly complex) constants.15. First reduce the PDE to ODE by some choice4 of space di erencing. 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). and the convergence rate to. THE REPRESENTATIVE EQUATION FOR SPACE-SPLIT OPERATORS241 It is important to notice that: The diagonal matrix on the right side of Eq. the steady-state solution of the representative equation. 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 . by itself. + (13. 13.3.17) x y dt which has the exact solution a u(t) = ce( x+ y )t .16. but this choice was for convenience only. This results in a spatially split A matrix formed from the subsets A(xx) = diag(B ) ~ A(yy) = diag(B ) (13. Now we are ready to present the representative equation for two dimensional systems. 4 .13. and its use is not necessary to arrive at Eq. Although Ax and Ay do commute. 13. does not ensure the property of \all possible combinations". To obtain the latter property the structure of ~ the matrices is important.15 contains every possible com~ bination of the individual eigenvalues of B and B . and where Ax and Ay satisfy the conditions in 13. Often we are interested in nding the value of.

The accuracy of the xed.18) steady-state solution (of the ODE) for any h. The convergence rate to reach the steady-state. Converges very rapidly to the steady-state when h is large. 3. h from which x . Unfortunately.16. " 1 #n a x+ y 2. The form of the method is given by Eq. h .19) un = c 1 .242 CHAPTER 13. steady-state solution.1 The Unfactored Implicit Euler Method Consider rst this unfactored. h y )un+1 = un + ha P (E ) = (1 . 12. In each case we examine 13. Is unconditionally stable. this method: 1. 1 Q(E ) = h giving the solution (13. The stability. rst-order scheme which can then be used as a reference case for comparison with the various factored ones. we nd (1 . In the following we analyze four di erent methods for nding a xed. h x . x y Like its counterpart in the 1-D case. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 13. use of this method for 2-D problems is generally impractical for reasons discussed in Section 12.4. steady-state solution to the 2-D representative equation 13. Produces the exact (see Eq. 3. however. 2.19.4 Example Analysis of 2-D Model Equations 1. h .5. h y )E . . and when it is applied to the representative equation. 13.

12. One nds (1 .4. x y We see that this method: 1.3 The Factored Delta Form of the Implicit Euler Method . h x)(1 .4. it is not very useful since its transient solution is only rst-order accurate and. h )(1 . There results (1 . h y )E .20) giving the solution " #n 1 un = c (1 . Produces a steady state solution that depends on the choice of h. h ) . Converges rapidly to a steady-state for large h. However. h y )un+1 = 1 + h2 x y un + ha and this has the solution " #n 1 + h2 x y a un = c (1 . Is unconditionally stable.20 to the 2-D representative equation. h x)(1 . + a h x y x y. h x)(1 . 1 Q(E ) = h (13. 2. but the converged solution is completely wrong.2 The Factored Nondelta Form of the Implicit Euler Method 13. if one tries to take advantage of its rapid convergence rate. un) = h( xun + y un + a) which reduces to (1 . + x y x y This method: 13. The method requires far less storage then the unfactored form.4.31 to the 2-D representative equation. h x)(1 . Next apply Eq. EXAMPLE ANALYSIS OF 2-D MODEL EQUATIONS 243 Now apply the factored Euler method given by Eq. 12. h y )un+1 = un + ha from which P (E ) = (1 . 3. h )(1 . h ) .13. h y )(un+1 . the converged value is meaningless.

the value of h on the left side of the implicit equation is literally switched from h to 1 h. this method demands far less storage than the unfactored form. and the factored delta form of the implicit Euler method can be used when a converged steady-state is all that is required. 2h y 2 This method: 1.5. 1 h x 1 . 2. Finally consider the delta form of a second-order time-accurate method. The correct steady solution is obtained. 12. Converges very slowly to the steady{state solution for large values of h. Is unconditionally stable. 1 h y (un+1 . un = c6 4 5 1 x+ y 1 . All of these properties are identical to those found for the factored delta form of the implicit Euler method. 3. 12. but convergence is not nearly as rapid as that of the unfactored form. In practical codes. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 1. 2 5 13. since j j ! 1 as h ! 1. Apply Eq. Is unconditionally stable.4 The Factored Delta Form of the Trapezoidal Method . 1h x 1 . 1 h y un+1 = 1 + 1 h x 1 + 1 h y un + ha 2 2 2 2 and this has the solution 2 3n 1 + 1h x 1 + 1h y 7 6 a 2 2 7 .4. Produces the exact steady-state solution for any choice of h. 2. Like the factored nondelta form.244 CHAPTER 13. Produces the exact steady{state solution for any choice of h.30 to the representative equation and one nds 1 1 .5 A brief inspection of eqs. 2 h x 1 . Since it is second order in time. 3. it can be used when time accuracy is desired. as discussed in Section 12.26 and 12. since j j ! 1 as h ! 1. un) = h( xun + y un + a) 2 which reduces to 1 .27 should be enough to convince the reader that the 's produced by those methods are identical to the produced by this method. Converges very slowly to the steady{state solution for large values of h.

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).3 generalize to three dimensions and. under the conditions given in 13. 2 h x 1 .21) x y z dt Let us analyze a 2nd-order accurate. the model 3-D cases6 have the following representative equation (with = 0): du = + + ]u + a (13. 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 . EXAMPLE ANALYSIS OF THE 3-D MODEL EQUATION 13. 2 x y z 4 x y + x z + y z) . +a+ x y 6 z (13. 13.12. 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 .16 with an A(zz) included. nd the root. 2 h( x + y + z ) un+1 = 1 + 2 h( x + y + z ) un + ha Put this in delta form: 1 .23) Eqs.22. 13.13. 1 h y 1 . and write the solution either in the form 2 3n 1 h( + + ) + 1 h2 ( 1 h3 + + ). . each with an additional term.5. 8 x y z . One can derive the characteristic polynomial for Eq.11 and 13.5 Example Analysis of the 3-D Model Equation 245 The arguments in Section 13. 1 h z un = h ( x + y + z )un + a] (13. 1 h( x + y + z ) un = h ( x + y + z )un + a] 2 Now factor the left side: 1 1 . factored. delta form using this equation.

and are both positive. 2 x 1 . the method is stable for all h. 4 h3 2 2 1h 1h 1 .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.246 CHAPTER 13. Now we can always nd one combination of the 's for which . This makes the method unconditionally stable for the 3-D di usion model when it is centrally di erenced in space. ) + ( + ) j (1 . however. 13. some form of dissipation is almost always added to methods that are used to solve the Euler equations and our experience to date is that. 13.12 with periodic boundary conditions. From the above analysis one would come to the conclusion that the method represented by Eq. clearly. In that case since the absolute value of the product is the product of the absolute values j 2 2 2 = (1 . )2 + ( . )2 >1 and the method is unconditionally unstable for the model convection problem. the instability disclosed above is too weak to cause trouble. +i i . 7 However. 13. In practical cases. central di erencing causes all of the 's to be imaginary with spectrums that include both positive and negative values. 13.23 that. if the method converges. Furthermore. it converges to the proper steady-state. 1h z 2 3n x y z7 7 . Now consider what happens when we apply this method to the biconvection model. +i where .24) It is interesting to notice that a Taylor series expansion of Eq. it follows from Eq. 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 . .24 results in 1 = 1 + h( x + y + z ) + 2 h2( x + y + z )2 (13. and are real numbers that can have any sign. Remember that in our analysis we must consider every possible combination of these eigenvalues. we already knew this because we chose the delta form. 2 y 1 . the 3-D form of Eq. 5 x+ y+ z a (13.i . if all the 's are real and negative.7 With regards to stability. in the presence of this dissipation.22 should not be used for the 3-D Euler equations.23 in the form 1+ = 1. First write the root in Eq. 13. In this case.

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

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

In the present context a vector is a vertical column or string. A general m m matrix A can be written 2 a11 a12 a1m 3 6a a a 7 A = (aij ) = 6 21 22 . 2m 7 7 6 5 4 am1 am2 amm 249 .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 2.1 Notation 1. . A. Given below is some notation and some of the important relations between matrices and vectors.. .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. Thus 2 v1 3 6 7 ~ = 6 v.

P is a permutation matrix if P ~ is a simple reordering of ~ . A. A is diagonally dominant if aii Pj6=i jaij j i = 1 2 : : : m and aii > Pj6=i jaij j for at least one i. 8. A is symmetric if AT = A.250APPENDIX A. A is orthogonal if aij are real and AT A = AAT = I 5. A is normal if AT A = AAT . AH is the conjugate transpose of A. 11. where I is the identity matrix.1 = det1 A] . bc and d A.b a . 9. The trace of a matrix is Pi aii. 4. 2.1 and has the property that A. v v 7. The inverse of a matrix (if it exists) is written A.A. det A] is the determinant of A. A is skew-symmetric or antisymmetric if AT = .1 = I . 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 . USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA 3. 6. 3.. 10.c . (Hermitian).2 De nitions 1. A is the complex conjugate of A.2 7 6 7 6 . If b A= a d c then det A] = ad . 7 4 T5 ~m a 4.1A = AA.

sA = (saij ) where s is a scalar.1 A. 1. In general AB 6= BA. including the matrix B . Transpose equalities: (A + B )T = AT + B T (AT )T = A (AB )T = B T AT 6. . If a square matrix with real elements is symmetric.1 = A (AB ). The eigenvalue problem for a matrix A is de ned as A~ = ~ or A .4 Eigensystems 1. Any matrix A can be expressed as the sum of a symmetric and a skew-symmetric matrix.1 = (A. as A~ = B~ or A . Thus: 1 A = 2 A + AT + 1 A . A + (B + C ) = (C + A) + B . AT 2 A. Inverse equalities (if the inverse exists): (A.1 = B . B ]~ = 0 x x x 2.A. ALGEBRA A.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. 3. 5. I ]~ = 0 x x x and the generalized eigenvalue problem.1 )T 251 7. 4. they are all imaginary. its eigenvalues are all real. 2.1 (AT ).1). etc.3. If it is asymmetric.

. h i X = ~1 ~2 : : : ~m x x x and is the diagonal matrix 0 0 m If A can be diagonalized. 7. If the eigenvalues of a matrix are not distinct... If a matrix does not have a complete set of linearly independent eigenvectors.2 . 9. the matrix is said to be derogatory. . 7 6 .252APPENDIX A. A set of eigenvectors is said to be linearly independent if a ~ m + b ~ n 6= ~ k x x x m 6= n 6= k for any complex a and b and for all combinations of vectors in the set. 6. but all of the eigenvectors are linearly independent. 4. but it can still be diagonalized.1AX = where X is a matrix whose columns are the eigenvectors. . 2 0 03 1 6 . and this eigenvector cannot be formed from a linear combination of any of the other eigenvectors. its eigenvectors completely span the space. . . The eigenvectors of such a matrix cannot span the space and the matrix is said to have a defective eigensystem. . . in which case the possibility exists that it cannot be diagonalized.. and A is said to have a complete eigensystem. an m m matrix A has n~ linearly independent eigenvectors with x n~ m and n distinct eigenvalues ( i) with n n~ m. the eigenvalues of a matrix may not be distinct. If A has m distinct eigenvalues. 8. If A posseses m linearly independent eigenvectors then A is diagonalizable.e. it cannot be diagonalized. then A is always diagonalizable. . . 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. 7 6 7 = 6 0 . In general. and with each distinct eigenvalue there is one associated eigenvector. X . USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA 3. In general. i. x x 5. 0 7 4 5 .

.1 6 0 17P = 61 07 4 5 4 5 0 0 0 1 14. such that where there are k linearly independent eigenvectors and Ji is either a Jordan subblock or i. .A. S . . 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 . Use of the transform S is known as putting A into its Jordan Canonical form. 7 6. EIGENSYSTEMS 2J 0 03 1 . 1 7 4 5 0 0 0 i 12. Some of the Jordan subblocks may have the same eigenvalue. The other r . 13. . 11...7 J =S 6 . 0 7 Ji = 6 7 i 6. . The complete set of principal vectors and eigenvectors are all linearly independent. 7 6 7 i 1 6 60 0 7 .. . . . 7 . the . . 0 7 4 5 0 0 Jk 253 10. 1 vectors associated with this eigenvalue are referred to as principal vectors.1 = P then 2 3 2 3 1 0 0 0 P . .. . For example. For each Jordan submatrix with an eigenvalue i of multiplicity r.. there exists one eigenvector.. Defective matrices cannot be diagonalized but they can still be put into a compact form by a similarity transform.. A Jordan submatrix has the form 2 0 03 i 1 60 . .7 6 .4.. 6 .1 AS = 6 0 J2 . . A repeated root in a Jordan block is referred to as a defective eigenvalue. ...

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. A p-norm of a matrix A is de ned as Av jjAjjp = max jjjjvjjjjp x6=0 p . having: 9 eigenvalues 3 distinct eigenvalues 3 Jordan blocks 5 linearly independent eigenvectors 3 principal vectors with 1 1 principal vector with 2 A. The spectral radius of a matrix A is symbolized by (A) such that (A) = j m jmax where m are the eigenvalues of the matrix A. 2.254APPENDIX A. A p-norm of the vector ~ is de ned as v 0M 1 X pA1=p jjvjjp = @ jvj j j =1 3.5 Vector and Matrix Norms 1.

8. Let A and B be square matrices of the same order. The spectral radius of A. Special p-norms are jjAjj1 = maxj=1 M jjAjj1 = maxi=1 2 M PM jaij j maximum row sum j =1 where jjAjjp is referred to as the Lp norm of A. (A) is a true norm. (A).A. in fact. When A is normal. so in general (A) is not a true norm. All matrix norms must have the properties jjAjj 0 jjAjj = 0 implies A = 0 jjc Ajj = jcj jjAjj jjA + B jj jjAjj + jjB jj jjA B jj jjAjj jjB jj 5.5. jjAjj2 = q PM ja j maximum column sum i=1 ij (A T A ) . 7. 6. is the lower bound of all the norms of A. in this case it is the L2 norm. In general (A) does not satisfy the conditions in 4. VECTOR AND MATRIX NORMS 255 4.

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

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

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

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

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

j =0 .15) 4 b b b b 5 b b b b The eigenvectors are always given by eq.13) B.B. and further examination shows that the elements in these eigenvectors correspond to the elements in a complex harmonic analysis or complex discrete Fourier series. B.4. i(a . B.4. EIGENSYSTEMS OF CIRCULANT MATRICES The eigenvalues are m 261 = b + (a + c) cos 2 m .1 The left-hand eigenvector matrix with elements x0 is 2 j m =0 1 M . Although the eigenvectors of a circulant matrix are independent of its elements. For the element indexing shown in eq.1 (B. In fact. 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 .1 M 1 Note that both X and X . even if they are completely dense.14.1 X = (xjm) = eij M m =0 1 M . and the right-hand eigenvector matrix has the form 2 m j =0 1 M .1. X bj ei jm=M = m 1 0 3 2 1 1 0 3 2 2 1 0 3 3 2 1 0 1 (2 ) of which eq. B.12 do not depend on the elements a b c in the matrix.im M X mj j =0 1 M .15 they have the general form M. the eigenvalues are not.1 . for the tridiagonal circulant matrix given by eq.14) p where i . all circulant matrices of order M have the same set of linearly independent eigenvectors. = (x0 ) = 1 e. 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. B.2 General Circulant Systems Notice the remarkable fact that the elements of the eigenvector matrices X and X .1 (B. are symmetric and that X . 1 1 1 M . = M X .13 is a special case. where X is the conjugate transpose of X .

. 7 6 .. For example.. a a b a b 32x 3 1 7 6 x2 7 76 7 76 .. 7 6 . 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.. . B.5 Special Cases Found From Symmetries APPENDIX B. 7 76 . 7 76 .262 B. SOME PROPERTIES OF TRIDIAGONAL MATRICES Consider a mesh with an even number of interior points such as that shown in Fig. B..17) . 7 76 .1.16 are (2m . 7 6 . 7 54 5 a 7 6 x2 7 = x 2 3 x1 6x 7 6 27 6 . 7 6 . 7 m6 . 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 . 1) m = b + 2a cos 2M + 1 ! m=1 2 M (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 . one can show from previous results that the eigenvalues of eq.. 7~ a ~ a)~ m = 7 xm = B (M : a b x 7 .16) By folding the known eigenvectors of B (2M : a b a) about the center. 7 6 . 7 6 7 4 x2 5 1 x 1 This leads to the subsystem of order M which has the form b a 7 7 a b a 7 7 . 7 76 ..

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

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

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

Notice also that. where is the internal energy per unit mass. 1 " # (C. according to eq.7) (C. under this condition. Qn) + O(h2) F = Fn + Bn(Q .105 can be written in general as.5) E = An Q + O(h2) F = BnQ + O(h2) (C. " # " # (C.1 This being the case.9) Note that this depends on the form of the equation of state. BnQn are identically zero for homogeneous vectors of degree 1.4. 6. by direct use of eq. As a nal remark. . C.106. see eq.8) @A Q = 0 @x in spite of the fact that individually @A=@x] and Q are not equal to zero. A and B .266APPENDIX C. and their Jacobians.6) since the terms En . the constant term drops out of eq.12 are homogeneous of degree 1. are homogeneous of degree 0 (actually the latter is a direct consequence of the former). Qn) + O(h2) can be written (C. if F is homogeneous of degree 1. 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 . C. AnQn and Fn . we notice that the expansion of the ux vector in the vicinity of tn which. that both E and F in eqs.3. The Euler equations are homogeneous if the equation of state can be written in the form p = f ( ). F = AQ and @F = A @Q + @A Q @x @x @x Therefore. 6. THE HOMOGENEOUS PROPERTY OF THE EULER EQUATIONS Now it is easy to show.11 and 2. 2.

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

Sign up to vote on this title
UsefulNot useful