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

. . . . . . . . . . . . . . 8. . . . 7. .3 7. . . . . 8.3 Relation to Circulant Matrices . . . . . . . .7. . .3. . . . . . . . . . . . . . . . . . . . . . .4 Comparing the E ciency of Explicit Methods . . . . . . . . . . . . . . . . . . . .7 7. . Consistency . . . . . . . . . . . . . . . . .3 Sti ness Classi cations . . . . . . . . . . . . . . . . . . . . .3 An Example Involving Periodic Convection . . . . . . . .6. . . . . . . . . . . . . .5. . . . . . . . . . . . . . .2 Implicit Methods . . Numerical Stability Concepts in the Complex h Plane 7. . . . . . . .5 7. . . . . . . . . . . 7. . . . . . . . . . . . . . . . . . . . . . . . . .2 Stability for Small t . . . . . . . . . . . . . . . . . 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. . . . 8. .4 7. . . . . .6. . . .6.2 Complete Eigensystems . . . . . . . . . . . . . . . . . 8. . . . . . . . . .2 Some Examples . . . . . . . . . . . . .1 The Criterion . . . . . . Fourier Stability Analysis . . . . . . . .3 Practical Considerations for Comparing Methods . . . . . . . . . . . . . . . .2 An Example Involving Di usion . . . . . . . . . . . . . . . . . 8. . . . . . . . . .5.3 Defective Eigensystems . . . .5.9 7. . . . . . . . . . . . . . . . 8. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7. A-Stable Methods . . . 7. . . . . . . . . . . . . . . . . . . . . . . . . 8. . . . . . 8. . . . . . . . . .7. . . . . . . . . . . . . . . . . 8. . . .3 Stability Contours in the Complex h Plane. 149 149 149 151 151 152 153 154 154 154 155 158 158 159 160 160 161 . . . . . . . . . 8. . . . . . . Numerical Stability Concepts in the Complex -Plane . . . . . . . . .4. . . . . . . . . . . . . . . . . . . . . 7. . . . .6 7. . .7. . .3 Defective Eigensystems . . . . .2 Unconditional Stability. . . . . 7. . . . . . . .3. . . . .6 Steady Problems . . . . . . . . . . . . . 7. . . .4. . . . . . . 8. . . . . . . . . .2. 7. .1 Sti ness De nition for ODE's . . .2 Driving and Parasitic Eigenvalues . . . . .5. . . . . . . . . . .7. . . . . . . . . . . . 7. . . . .5 Coping With Sti ness . . . . . . . . . . . . . . . . . . . . . . . . . . . 8. . . Numerical Stability of O E 's . . . . . . . . . .1 Relation to -Eigenvalues . 8. . . . . . 8. . . . . . . . . . . . . . . 7. . . .1 -Root Traces Relative to the Unit Circle . . . . . . .1. . . . . . . . . . . . . . . Problems . . .1 Imposed Constraints . . . . . . .1. . . . . 8. . .3.1 Explicit Methods . . . . . . . . . .2 Relation of Sti ness to Space Mesh Size .1 The Basic Procedure . . . . .4. . . Time-Space Stability and Convergence of O E's . . . . . . . . . . . . . . . . .7 Problems . . . .1. . .3 A Perspective . . . . . . . . . . . . .5.8 7.1 Stability for Large h. . . .

. . . . . . . . . . . . . . . . . . . . . . . . . 9. . . . . . . . .3. . . . Upwind Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1. . . . . . . . . . . . . . . . . .3 The ODE Approach to Classical Relaxation . . 9. .1 Motivation . . .2 Properties of the Iterative Method .1 Eigenvector and Eigenvalue Identi cation with Space Frequencies191 10. . . . . . . . . . . . . . . . . . . . . . . . 9. . . . . . . . . . . . . .3 Factoring Space Matrix Operators in 2{D . . . . . . . . . . . . . 191 11 NUMERICAL DISSIPATION 203 204 205 207 209 210 212 213 214 12 SPLIT AND FACTORED FORMS 12. . . . . .1 The Point-Jacobi System . . . . . . . . . . . . . . . . . . . . . .2 The Gauss-Seidel System . . . . . . The Lax-Wendro Method . . . . . . . .3 A Two-Grid Process . . . . .6 Problems . . . . . .2 Flux-Di erence Splitting . . . . . . . the Residual. . 220 217 . . . . . . . . . . . . . . . . . . .3. . . . . . . . . . . . . . . . .6 Problems . . . . . . . . . . . . . . 163 164 164 166 168 168 168 169 170 170 172 173 174 176 180 182 187 10 MULTIGRID 10. 9. . . . . . . . . . . .1 11. . . . . . . . . . . . . and the Error 9. . . . . . . . . . . . . . . . . . . . . . . . 192 10. . . . . . . . . . . . . 9. . . . 9.2.3 The Classical Methods . . .2 Classical Relaxation . . . . . . . . . . . . . . .4 Problems . . . . . . . . . . . . . 11. . . . . . . . . . . . . . . . .9 RELAXATION METHODS 9. . . . . . 9. . . . . . . . . . . 11. . . 11. . . . . . . . . 9. . .4. . .4. . . . . . .4.4. 202 11.4 One-Sided First-Derivative Space Di erencing The Modi ed Partial Di erential Equation . . .4 Eigensystems of the Classical Methods . . . . . . . . . . . . . .1. . . . . . . .1 The Delta Form of an Iterative Scheme . 11. .2 Factoring Physical Representations | Time Splitting . . 191 10. . 217 12. . . . . . . . . . . . .2. . . . . . . .2 The Converged Solution. . . . . . . . . . . . . . . . . . . . . .5 Nonstationary Processes . . . . . . 218 12. . . . . . . . . . . . . 192 10. . . .5 Arti cial Dissipation . .1 Preconditioning the Basic Matrix . . . .4. . . 9. . . . .1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 ODE Form of the Classical Methods . . .2 The Model Equations . . . . . . . . . . . . . . . . . . . . . . 9. . 9. . . . . . . .3 11. . . . . . . . . . . . . . . . . . . . . . . . . . . . 9. . . . . . . . . . . .2. . . . . . .1 The Concept . . . . . . . 9. . . . . . . . . .1 The Ordinary Di erential Equation Formulation . . . . .1 Formulation of the Model Problem . . . . . . . . . . . . . . . . . . . . . . 200 10.3 The SOR System . . . . . . . . . . .2 The Basic Process . . . . . .1. . .2 11. . . . .1 Flux-Vector Splitting . 9. . . .

. . Second-Order Factored Implicit Methods . . . . . B. .3. . . . . . . . . . . . . .2 Example Analysis of Circulant Systems . . . . . A. . . . . . . .2 The Factored Nondelta Form of the Implicit Euler Method 13. . . . . . . . . .3. . . . . . . . . . . . .1 Stability Comparisons of Time-Split Methods . . . . . . . . . . The Inverse of a Simple Tridiagonal .12.5 Example Analysis of the 3-D Model Equation . . . . . . . . .1 B. . . . Vector and Matrix Norms . . . . . . . .2. . . . . . . . . . B. . . . . . . . . .4 A. 13. . De nitions . . . .3. . . . . . . . . . . . . . . . 220 221 221 223 226 226 228 229 13 LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 13. . . . . . . . .2 B. . . . . . . . . .2 Data Bases and Space Vectors . .3. . . . . .3 B. . . . . . . . . . . . . . . . . . . . .6 Problems . . . . . . . . . . . . . 13. . . . .2 A. . . . .1 The Unfactored Implicit Euler Method . . . . . . . . . . . . . . . . . B. . . . . .4 12. . . . . 233 233 234 234 237 238 242 242 243 243 244 245 247 249 250 251 251 254 257 258 259 260 260 261 262 263 A USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA 249 B SOME PROPERTIES OF TRIDIAGONAL MATRICES Standard Eigensystem for Simple Tridiagonals . .2. . . . . . . . . .4 The Factored Delta Form of the Trapezoidal Method . .4 Space Splitting and Factoring . . . . . . . . 13. . . . . . . . . . . . . . . . . . . .5 B. . . . Eigensystems of Circulant Matrices . . . . . .4 Notation . .4 Example Analysis of 2-D Model Equations . . . . . . . . . . . . . . . . . . . . .3 The Representative Equation for Space-Split Operators . .4. 13. . .3 A. . . . . . . . . . . . . 13. . . . . 12. . . . . Importance of Factored Forms in 2 and 3 Dimensions The Delta Form . 257 . . . .5 Special Cases Found From Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13. . . .5 12. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 The Factored Delta Form of the Implicit Euler Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 A. . . . Generalized Eigensystem for Simple Tridiagonals . .4. . . . . . . . . . . . B. . . .3 Data Base Permutations . . . . . .1 Standard Tridiagonals . . . . . . . . . . . . . . . .4. . . . 13.1 Mesh Indexing Convention . . . . .4. . 13. Problems . 13. . . . . . . . . . . .2 Analysis of a Second-Order Time-Split Method . . . . . . . . 12. . . . . . . . . . . . . . . . . . .2 General Circulant Systems . Eigensystems . .7 12. . . . . . . . . . . . . . .4. . . . . . 13. . . . . . . . . . . . . .6 12. . . . . . . . . . . Algebra . . . . . . . . . .4. . . . . . . . . . . . . . .1 The Representative Equation for Circulant Operators . . . . . . . . . . . 12. . . . . . . . . . .6 Special Cases Involving Boundary Conditions . . . . . .

C THE HOMOGENEOUS PROPERTY OF THE EULER EQUATIONS265 .

The principal such constraint was the demand for uni cation.Chapter 1 INTRODUCTION 1.1 Motivation The material in this book originated from attempts to understand and systemize numerical solution techniques for the partial di erential equations governing the physics of uid ow. as a consequence. This. often have no analytic solution. In the eld of aerodynamics. in e ect. 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. all of these phenomena are governed by the compressible Navier-Stokes equations. The mathematical structure is the theory of linear algebra and the attendant eigenanalysis of linear systems. convection. can change the form of the basic partial di erential equations themselves. underlying constraints on the nature of the material began to form. di usion. As we shall see in a later chapter. of course. which 1 . and turbulence. boundary layers. motivates the numerical solution of the associated partial di erential equations. The ultimate goal of the eld of computational uid dynamics (CFD) is to understand the physical events that occur in the ow of uids around and within designated objects. As time went on and these attempts began to crystallize. The new equations. but the authors believe the answer is a rmative and present this book as justi cation for that belief. These events are related to the action and interaction of phenomena such as dissipation. Many of the most important aspects of these relations are nonlinear and. the use of numerical methods to solve partial di erential equations introduces an approximation that. At the same time it would seem to invalidate the use of linear algebra for the classi cation of the numerical methods. Experience has shown that such is not the case. slip surfaces. shock waves.

these errors are often identi ed with a particular physical phenomenon on which they have a strong e ect.2 Background The eld of computational uid dynamics has a broad range of applicability.2 CHAPTER 1. are often referred to as the modi ed partial di erential equations. the resulting simulation can still be of little practical use if ine cient or inappropriate algorithms are used. including the geometry. if any. This motivates studying the concepts of stability. On the other hand. if used and interpreted properly. Since they are not precisely the same as the original equations. The geometry may result from . they can lead to serious di culties. the theory associated with the numerical analysis of uid mechanics was developed predominantly by scientists deeply interested in the physics of uid ow and. Thus methods are said to have a lot of \arti cial viscosity" or said to be highly dispersive. All of these concepts we hope to clarify in this book.2. nor with deliberately directing an error to a speci c physical process. as long as the error remains in some engineering sense \small". This motivates studying the concepts of sti ness. Regardless of what the numerical errors are called. and consistency. 1. for example. these di erences are usually referred to as truncation errors. 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. producing answers that represent little. as a consequence. Independent of the speci c application under study. convergence. the following sequence of steps generally must be followed in order to obtain a satisfactory solution. these methods give very useful information. However. INTRODUCTION are the ones actually being solved by the numerical process. There is nothing wrong. However. if their e ects are not thoroughly understood and controlled. factorization. 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. even if the errors are kept small enough that they can be neglected (for engineering purposes). 1.1 Problem Speci cation and Geometry Preparation The rst step involves the speci cation of the problem. physical reality. with identifying an error with a physical process. and probably will. and the requirements of the simulation. they can. It is safe to say. 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. and algorithm development in general. ow conditions. of course. Mathematically.

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

the so-called \model" equations. This step can require post-processing of the data. and Pletcher 1] and Hirsch 2]. nite-volume. the grid can be altered based on the solution in an approach known as solution-adaptive gridding.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. we can make use of much simpler expressions. Rather than presenting the details of the most advanced methods. Many of these issues are not addressed in this book. to develop. For example.2. composite. 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. Here again. INTRODUCTION Many di erent gridding strategies exist. 1.4 Assessment and Interpretation of Results Finally. the success of a simulation can depend on appropriate choices for the problem or class of problems of interest. for example calculation of forces and moments. Furthermore. and overlapping grids. the use of nite-di erence methods is typically restricted to structured grids. These model equations isolate certain aspects of the physics contained in the complete set of equations. It is critical that the magnitude of both numerical and physical-model errors be well understood. or spectral methods. hybrid. the results of the simulation must be assessed and interpreted. nite-element. We believe that a thorough understanding of what happens when . unstructured. which are still evolving.4 CHAPTER 1. Fortunately. when used intelligently. 1. we present a foundation for developing.and three-dimensional uid ow simulations. they have been carefully selected to be representative. Some of them are presented in the books by Anderson. The numerical methods generally used in CFD can be classi ed as nite-di erence. Tannehill. analyzing. The choices of a numerical method and a gridding strategy are strongly interdependent. with emphasis on nite-di erence and nite-volume methods for the Euler and Navier-Stokes equations. and can be aided by sophisticated ow visualization tools and error estimation techniques. Instead we focus on numerical methods. including structured. and understanding such methods. and understand most numerical methods used to nd solutions for the complete compressible Navier-Stokes equations. analyze. of di culties and complexities that arise in realistic two. Although the model equations are extremely simple and easy to solve.

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

INTRODUCTION .6 CHAPTER 1.

though.g. 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. mass. such as the Euler and Navier-Stokes equations and our model equations. and convergence. the integral conservation-law form. which is natural for nite-di erence schemes. The equations are then recast into divergence form. Q is a vector containing the set of variables which are conserved. and energy. The equation is a statement of 7 . accuracy. These equations contain many of the salient mathematical and physical features of the full Navier-Stokes equations. in particular. per unit volume. can be written in the following integral form: Z V (t2 ) QdV . the convection and di usion equations. The Euler and Navier-Stokes equations are brie y discussed in this Chapter.. and they represent phenomena of importance to the analysis of certain aspects of uid dynamic problems. e.1) In this equation.Chapter 2 CONSERVATION LAWS AND THE MODEL EQUATIONS We start out by casting our equations in the most general form. The main focus.1 Conservation Laws Conservation laws. will be on representative model equations. which is useful in understanding the concepts involved in nite-volume schemes. momentum. The model equations we study have two properties in common. The concepts of convection and di usion are prevalent in our development of numerical methods for computational uid dynamics. stability. 2. They are linear partial di erential equations (PDE's) with coe cients that are constant in both space and time.

or cell.2) dt V (t) S (t) V (t) Those methods which make various numerical approximations of the integrals in Eqs.8 CHAPTER 2. in Cartesian coordinates.4) r: i @x @y @z and i j.2 is obtained by applying Gauss's theorem to the ux integral.1 and 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. 2.6) . 2. t1 . or tensor. In two dimensions. The divergence form of Eq. momentum and energy for a uid. Those methods which make various approximations of the derivatives in Eq.1 can be rewritten as d Z QdV + I n:FdS = Z PdV (2. and z coordinate directions. and k are unit vectors in the x y. respectively. 2. is an area A(t) bounded by a closed contour C (t). For a Newtonian uid in one dimension. F is a set of vectors.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. If all variables are continuous in time. On the other hand.3 and nd a solution for Q on that basis are referred to as nite-di erence methods. 2. Many of the advanced codes written for CFD applications are based on the nite-volume concept.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. by ! @ +j @ +k @ : (2. 2. The vector n is a unit vector normal to the surface pointing outward. the region of space.2 and nd a solution for Q on that basis are referred to as nite-volume methods. containing the ux of Q per unit area per unit time. a partial derivative form of a conservation law can also be derived.3) @t where r: is the well-known divergence operator given. 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 . and P is the rate of production of Q per unit volume per unit time. leading to @Q + r:F = P (2. then Eq. they can be written as @Q + @E = 0 (2.

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

1)q4 + 3.10 CHAPTER 2. 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 ( . q2. 1)q4 + 2 q1 2 q1 7 7 7 7 q3 q2 7 7 q1 7 7 7 5 3 2q q4 q2 . q2 . 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. and q4 . we must rst write the ux vectors E and F in terms of the conservative variables. cp=cv .1 q2 + q3 2 2 2 q1 2 q1 q1 q2 (2. ) q3 q1 q2 q1 0 A= a21 q . . + .1 q2 2 3 2 q1 3 7 7 7 7 7 7 7 7 2 7 . 2 q1 2 q2 q1 . . 1 q3 2 . 2 q4 q 3 q1 2 q3 q1 . (u2 + v2)=2] from the ideal gas law.1 q3 7 . where is the ratio of speci c heats.1 0 a42 !2 a43 q2 q1 3 7 7 7 7 7 7 7 7 7 7 7 7 7 5 (2. ) q2 q1 q q3 1 0 (1 . 2q . 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 . q3 . q2 q1 q1 3 a41 ! . CONSERVATION LAWS AND THE MODEL EQUATIONS to derive the ux Jacobian matrices.12) We have assumed that the pressure satis es p = ( . q1 . 1) e .1 q2 7 2 q1 7 7 7 5 3 q3 2 q1 (2.13) where a21 = .

( . 1 (3u2 + v2 ) 2 .2.uv a41 a21 = v a42 u a43 (2. 1) . aspects of convective phenomena associated with coupled systems of equations such as the Euler equations are important in developing numerical methods and boundary conditions. THE NAVIER-STOKES AND EULER EQUATIONS a41 = ( 2 . The homogeneous property of the Euler equations is discussed in Appendix C. 1 v2 .15) u where a41 = ( . )v ( . 1)u(u2 + v2) . The eigenvalues of the ux Jacobian matrices are purely real. .5.16) Derivation of the two forms of B = @F=@Q is similar.2.14) 3 7 7 7 7 7 7 7 7 7 7 7 5 0 1 0 0 0 a21 (3 . 1) q q 1 1 and in terms of the primitive variables as 2 6 6 6 6 6 A=6 6 6 6 6 6 4 + q3 q1 (2. u2 2 2 a43 = (1 . This is the de ning feature of hyperbolic systems of PDE's. 1)4 ! 11 q2 q1 !3 + q3 q1 2 !2 !2 q2 q1 !3 5. )u (1 . The Navier-Stokes equations include both convective and di usive uxes. ue a42 = e . 1 43 q2 a42 = q1 2 q1 ! ! q2 q3 a43 = . !2 3 5 q4 q1 ! q2 q1 ! q4 . Thus we also study linear hyperbolic systems of PDE's. 3 . This motivates the choice of our two scalar model equations associated with the physics of convection and di usion. )uv (2. Furthermore. . which are further discussed in Section 2. .

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

3.22) Substituting f (t) into Eq. 2.17. We restrict our attention to initial conditions in the form u(x 0) = f (0)ei x (2.!t) (2.20 gives the following solution u(x t) = f (0)ei (x.19. 2.at) = f (0)ei( x. and the phase speed.3.25) . !. As we shall see later.21) dt which has the solution f (t) = f (0)e. we nd that f (t) satis es the following ordinary di erential equation (ODE) df = . the wavenumber. and is the wavenumber. In order to satisfy the periodic boundary conditions. Eq.23) where the frequency. 2. waves with di erent wavenumbers travel at di erent speeds. most numerical methods introduce some dispersion that is. THE LINEAR CONVECTION EQUATION 13 change in the shape of the solution. are related by != a (2.19) where f (0) is a complex constant. .20) where the time dependence is contained in the complex function f (t).24) The relation between the frequency and the wavenumber is known as the dispersion relation. the solution can be written as u(x t) = f (t)ei x (2. We now examine the biconvection problem in more detail. Substituting this solution into the linear convection equation. Preserving the shape of the initial condition u0(x) can be a di cult challenge for a numerical method. 2.2. must be an integer.ia t (2.24 is characteristic of wave propagation in a nondispersive medium. a.2 Solution in Wave Space u(x 0) = M X m=1 fm (0)ei mx (2. in a simulation. Let the domain be given by 0 x 2 . This means that the phase speed is the same for all wavenumbers. For M modes.ia f (2. It is a measure of the number of wavelengths within the domain. With such an initial condition. An arbitrary initial waveform can be produced by summing initial conditions of the form of Eq. one obtains 2. The linear relation given by Eq.

2.30) . Other steady-state solutions are obtained if a source term g(x) is added to Eq. In contrast to the linear convection equation. the solution is obtained by summing solutions of the form of Eq. or mixed Dirichlet/Neumann. In the case of Eq. g(x) # (2.26) 2.27) @t @x2 where is a positive real constant.14 CHAPTER 2. as follows: @u = @ 2 u (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.4 The Di usion Equation 2. For example. Dirichlet (speci ed u). A simple linear model equation for a di usive process is Therefore.at) (2. this parabolic PDE governs the di usion of heat in one dimension.27.28) @x2 @u = @t " 2 @u @x2 . with u representing the temperature. which is one that satis es the governing PDE with the partial derivative in time equal to zero. Since the wave equation is linear. the steady-state solution must satisfy @2u = 0 (2. the di usion equation has a nontrivial steady-state solution.27. 2.4. g(x) = 0 @x2 (2. Neumann (speci ed @u=@x). 2. CONSERVATION LAWS AND THE MODEL EQUATIONS where the wavenumbers are often ordered such that 1 2 M . m=1 fm (0)ei m(x.1 Di erential Form Di usive uxes are associated with molecular motion in a continuum uid.29) giving a steady state-solution which satis es @ 2 u .23. Boundary conditions can be periodic. u must vary linearly with x at steady state such that the boundary conditions are satis ed.

.4. 2 t (2. Let the domain be given by 0 x with boundary conditions u(0) = ua. i.e.37 shows that high wavenumber components (large m) of the solution decay more rapidly than low wavenumber components. Substituting this form into Eq.27. Eq. h(x) = ua + (ub .37) The steady-state solution (t ! 1) is simply h(x).2 Solution in Wave Space We now consider a series solution to Eq. 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. The corresponding steady equation is @ 2 u + @ 2 u . 2 f (2. 2 t sin m x + h(x) (2. The latter is known as the Poisson equation for nonzero g. It is clear that the steady-state solution is given by a linear function which satis es the boundary conditions. Let the initial condition be u(x 0) = M X m=1 fm(0) sin m x + h(x) (2. consistent with the physics of di usion.35) m m dt and we nd m fm(t) = fm(0)e. Eq. u( ) = ub. and as Laplace's equation for zero g. 2. g(x y) = 0 (2.33) where must be an integer in order to satisfy the boundary conditions.4.31 is parabolic. we obtain m=1 u(x t) = M X m=1 m fm (0)e.2.27 gives the following ODE for fm : dfm = .31) @t @x2 @y2 where g(x y) is again a source term.34) satis es the initial and boundary conditions. the di usion equation becomes " # @u = @ 2 u + @ 2 u . g(x y) (2. 2.32 is elliptic. 2. ua)x= . 2.32) @x2 @y2 While Eq. 2.34.36) Substituting fm(t) into equation 2.

38 by X .1u. 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. Other systems of equations governing convection and wave propagation phenomena. Eq.1.1 are constants. such as the Maxwell equations describing the propagation of electromagnetic waves.16 2. postmultiplying A by the product XX . we obtain =0 (2. 2. form a hyperbolic system of partial di erential equations.43) @t @x When written in this manner. .1AX X . Premultiplying Eq.1 Thus = X .42) @t + @x With w = X .39) @t @x @f where f is the ux vector and A = @u is the ux Jacobian matrix. and X is the matrix of right eigenvectors. CONSERVATION LAWS AND THE MODEL EQUATIONS The Euler equations.5 Linear Hyperbolic Systems CHAPTER 2. this equation can also be written in the form @u + @f = 0 (2.38) @t @x where u = u(x t) is a vector of length m and A is a real m m matrix.1 . and noting that X and X .1u See Appendix A for a brief review of some basic relations and de nitions from linear algebra.1u z }| { @ X . are also of hyperbolic type. Such a system is hyperbolic if A is diagonalizable with real eigenvalues. For conservation laws.2. this can be rewritten as @w + @w = 0 (2.1AX (2. the equations have been decoupled into m scalar equations of the form @wi + @wi = 0 (2. The entries in the ux Jacobian are @f aij = @ui (2.44) i @t @x 1 @X .41) where is a diagonal matrix containing the eigenvalues of A.8. 2.40) j The ux Jacobian for the Euler equations is derived in Section 2.

which corresponds to in ow for these variables. 2. Each characteristic variable satis es the linear convection equation with the speed given by the corresponding eigenvalue of A. or wave speeds. 2. characteristic variables associated with positive eigenvalues can be speci ed at the left boundary. Therefore. Show that the 1-D Euler equations can be written in terms of the primitive variables R = u p]T as follows: @R + M @R = 0 @t @x where 2 3 u 0 M = 6 0 u . corresponding to the fact that sound waves can travel upstream in a subsonic ow. in a supersonic ow. The eigenvalues of the ux Jacobian matrix. The one-dimensional Euler equations can also be diagonalized.6 Problems 1. PROBLEMS 17 The elements of w are known as characteristic variables. The speed u is associated with convection of the uid. In a subsonic ow. while u + c and u . where juj > c. Characteristic variables associated with negative eigenvalues can be speci ed at the right boundary. where u is the local uid velocity. That is. we see that a hyperbolic system in the form of Eq. c. and u . While other boundary condition treatments are possible. are q u u + c. wave speeds of both positive and negative sign are present. The characteristic variables each satisfy the linear convection equation with the wave speed given by the corresponding eigenvalue. which is the inow boundary for these variables. the boundary conditions can be speci ed accordingly. of course. The signs of the eigenvalues of the matrix A are also important in determining suitable boundary conditions.6. c are associated with sound waves. leading to three equations in the form of the linear convection equation.1 7 4 5 0 p u . although they remain nonlinear. Based on the above. where juj < c.2. we must ensure that our numerical methods are appropriate for wave speeds of arbitrary sign and possibly widely varying magnitudes. they must be consistent with this approach.38 has a solution given by the superposition of waves which can travel in either the positive or negative directions and at varying speeds. and c = p= is the local speed of sound. Therefore. While the scalar linear convection equation is clearly an excellent model equation for hyperbolic systems. all of the wave speeds have the same sign.

and initial conditions from Eq. Find the eigenvalues and eigenvectors of A. 4. (Hint: make use of the matrix S = @Q=@R. in the form of Eq. Find the eigenvalues and eigenvectors of the matrix M derived in question 1. 2. j = 0 1 2 3. Write the 2-D di usion equation.. Plot the solution at t = 1 with = 1. nd the necessary values of fm (0). (Hint: use M = 2 with 1 = 1 and 2 = .e.2.33 with fm(0) = 1 for 1 m 3.31. respectively.5).1. 9. i. that is. Given the initial condition u(x 0) = sin x de ned on 0 x 2 . Write the classical wave equation @ 2 u=@t2 = c2 @ 2 u=@x2 as a rst-order system. Derive the ux Jacobian matrix A = @E=@Q for the 1-D Euler equations resulting from the conservative variable formulation (Eq. 8.4.18 CHAPTER 2. Next consider a solution with boundary conditions ua = ub = 0. write it in the form of Eq. 1. 2. u2=2). Plot the initial condition on the domain 0 x . Find the values of fm(0) required to reproduce the initial condition at these discrete points using M = 4 with m = m . 2. fm (0) = 0 for m > 3. 2.) 5. Plot the rst three basis functions used in constructing the exact solution to the di usion equation in Section 2. 2.2. Find its eigenvalues and compare with those obtained in question 2. p = ( . 2. 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 ! = . CONSERVATION LAWS AND THE MODEL EQUATIONS Assume an ideal gas. Show that the two matrices M and A derived in questions 1 and 3. 7. in the form @U + A @U = 0 @t @x where U = @u=@x @u=@t]T . 6. 3.) Next consider the same initial condition de ned only at x = 2 j=4. 1)(e . @x + @u = 0 @y . are related by a similarity transform.25. Eq.

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

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 1 2 3 2 Central nd κ∆ x 1.5 3 ¡ — .5 1.5 1 Û ’ ¥ H d R `B I— ˜uWI8Ev¤ Ú Ó × Ò Ë É ÙÒ Ø × Í Ê Ê É Ì Ç Ó Ì Í Ì Ñ ÑÒ ÖÕ Ì Ë Ò ÓÒ Ñ Ð Î Í Ì Ë Ê É È u‡§™†fjˆˆ«”w€‘ˆ””Ԑz€mQ‘Ôv«Sψm™‘”Ç κ*∆ x 0.5 th 4 Central 4 Pade th 2 2.5 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.

› ¨ § 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— .

In the most general notation.1) which. This generally leads to a point di erence operator (see Section 3. Di erencing the space derivatives converts the basic PDE into a set of coupled ODE's. which we refer to as the semi-discrete approach. ujn. On occasion. 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.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. without approximating the time derivative. can be used for the time advance of the solution at any given point in the mesh. x 51 2 2 ( +1) ( 1) ( ) +1 ( ) 2 ( ) 1 ( +1) ( 1) 2 ( ) +1 ( ) ( ) 1 . 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. + 2h ujn . we use this form. 2ujn + ujn 5 .3. = 4 2h x or h i ujn = ujn. uj . we nd 2 n 3 ujn .1) dt which includes all manner of nonlinear and time-dependent possibilities. but the rest of this chapter is devoted to a more specialized matrix notation described below. 2ujn + ujn (4.2) . in turn. we proceed with an analysis making considerable use of this concept. In the following chapters.

and convergence of Eqs. we shall separate the space di erence approximations from the time di erencing. Another possibility is to replace the value of ujn in the right hand side of Eq. 2 ( 1) ( ) 1 1 3 (4. however. This is a simple and natural formulation when the ODE's are linear.3 which we cannot comment on until we develop a framework for such investigations. For the time being. This results in the formula ( ) ( +1) ( 1) ujn ( +1) un = ujn.3 is referred to as the DuFort-Frankel method. THE SEMI-DISCRETE APPROACH Clearly Eq. 4. We introduce these methods here only to distinguish between methods in which the temporal and spatial terms are discretized separately and those for which no such separation is possible. In these simple cases. 4. )=2. the spatial and temporal discretizations are not separable. 2 @ j x ( 1) 2 ( ) +1 ( +1) 2 0 ( +1) + ujn. 4. stability.2 and 4. we can approximate the space derivatives with di erence operators and express the resulting ODE's with a matrix formulation. namely (ujn + ujn. Thus.1. 4.3) which can be solved for u n and time advanced at the point j .2 is a di erence equation which can be used at the space point j to advance the value of u from the previous time levels n and n .52 CHAPTER 4. 1 to the level n + 1. There are a number of issues concerning the accuracy.1 Reduction of PDE's to ODE's 4.2 with the time average of u at that point. In this case. Note. As we shall see later on. 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. and no semi-discrete form exists. . Equation 4. this method has an intermediate semi-discrete form and can be analyzed by the methods discussed in the next few chapters. + 2h 4ujn .1 The Model ODE's First let us consider the model PDE's for di usion and biconvection described in Chapter 2. there are subtle points to be made about using these methods to nd a numerical solution to the di usion equation. that the spatial and temporal discretizations are separable. A n 5 + uj . It is a full discretization of the PDE.2 is sometimes called Richardson's method of overlapping steps and Eq. 4. In this approach.

3. 4. REDUCTION OF PDE'S TO ODE'S 53 Model ODE for Di usion For example. 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. .1. ~ (t) u u f dt (4. using the 3-point central-di erencing scheme to represent the second derivative in the scalar PDE governing di usion leads to the following ODE di usion model d~ = u B (1 .6) Note that the elements in the matrix A depend upon both the PDE and the type of di erencing scheme chosen for the space terms.6. The vector ~ (t) is usually determined f by the boundary conditions and possibly source terms.4.5) dt 2 x p where the boundary condition vector is absent because the ow is periodic.2 1)~ + (bc) u ~ (4. 2 Model ODE for Biconvection The term biconvection was introduced in Section 2.1 0 1)~ u u (4.5 are the model ODE's for di usion and biconvection of a scalar in one dimension.1.4) dt x ~ with Dirichlet boundary conditions folded into the (bc) vector.2 The Generic Matrix Form The generic matrix form of a semi-discrete approximation is expressed by the equation d~ = A~ . In such cases the equations are nonlinear. 4.4 and 4. Eqs. In this case. It is used for the scalar convection model when the boundary conditions are periodic. In general. even the Euler and Navier-Stokes equations can be expressed in the form of Eq. They are linear with coe cient matrices which are independent of x and t. the 3-point central-di erencing approximation produces the ODE model given by d~ = . a B (. 4. the elements of A depend on the solution ~ and u are usually derived by nding the Jacobian of a ux vector. that is. Although the equations are nonlinear.

which exist under certain conditions. 4. 4.6 can be written in terms of the eigenvalues and eigenvectors of A. 4. .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) . A.6 in which the coe cient matrix. This holds regardless of whether or ~ not the forcing function.2 Exact Solutions of Linear ODE's CHAPTER 4. and together they have the property that 1 2 X . is independent of u. and its x corresponding eigenvalue. if @F=@u = A where A is independent of u). THE SEMI-DISCRETE APPROACH In order to advance Eq. 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.2.e. AX = 1 (4. the exact solution of Eq. 4. An eigenvector. f ..1 in time. m. the general solution cannot be written whereas. the general solution to Eq. we will make use of exact solutions of coupled systems of ODE's. when the ODE's are linear they can be expressed in a matrix notation as Eq. mI]~ m = 0 x (4. In order to analyze time-marching methods.8) where is a diagonal matrix whose elements are the eigenvalues of A.54 4. As we have already pointed out.1 are said to be linear if F is linearly dependent on u (i. the system of ODE's must be integrated using a time-marching method. 4.7) The eigenvalues are the roots of the equation det A . if A does not depend explicitly on t. The ODE's represented by Eq. If A does depend explicitly on t. have the property that A~ m = m~ m or x x A . This will lead us to a representative scalar equation for use in analyzing time-marching methods. ~ m . As we shall soon see. depends explicitly on t. These ideas are developed in the following sections.6 can be written. 4.

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)

That is. This permits us to say a great deal about such problems and serves as the basis for this section. of Jordan blocks or. another very useful frame. the elements of A are independent of both u and t. REAL SPACE AND EIGENSPACE 1 61 Note that the steady-state solution is A.4. it is more instructive to refer to these groups as the transient and steady-state solutions. however. as might be expected. In particular. In this way.2.2 that pre. on the cleverness of the relator.3. we get di erent perspectives of the same solution. but entirely equivalent.3 Real Space and Eigenspace Following the semi-discrete approach discussed in Section 4. and this can add signi cantly to our understanding. we identify di erent mathematical reference frames (spaces) and view our solutions from within each. composed. in the 4. in the most general case.1 and 4. For the model problems on which we are focusing most of our attention. ~ u u f (4. to a much greater extent.27) 1 1 1 4. we reduce the partial di erential equations to a set of ordinary di erential equations represented by the generic form d~ = A~ . we can develop some very important and fundamental concepts that underly the global properties of the numerical solutions to the model problems. The second group is referred to classically as the particular solution or the particular integral. An alternative. We saw in Sections 4. it is said u to be in real space. ~ . There is.1.and post-multiplication of A by the appropriate similarity matrices transforms A into a diagonal matrix. 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.3. We begin by developing the concept of \spaces". ~ u x x x f (4.1 De nition . How these relate to the numerical solutions of more practical problems depends upon the problem and. form of the solution is ~ (t) = c e 1 t ~ + + cme m t ~ m + + cM e M t ~ M + A. respectively. We say If a solution is expressed in terms of ~ .28) dt The dependent variable ~ represents some physical quantity or quantities which relate u to the problem of interest. The most natural reference frame is the physical one. In our application to uid dynamics.

The relations that transfer from one space to the other are: ~ w = X. and 2. X . ~ u x f and wm(t) = cm e m t + 1 gm m=1 2 M 1 1 for real space and eigenspace. the u ~ are linear combinations of all of the elements of ~ . THE SEMI-DISCRETE APPROACH simplest nondefective case. and individually elements of w u they have no direct local physical interpretation. the transient portion of the solution in eigenspace provides the weighting of each eigenvector component of the transient solution in real space. respectively. the transient portion of the solution in real space consists of a linear combination of contributions from each eigenvector. m 4. At this point we make two observations: 1.62 CHAPTER 4. However.1. When the forcing function ~ is independent of t. the solutions in the two spaces f are represented by ~ (t) = X cm e m t ~ m + X . of scalars. ~ u ~ = X.2 Eigenvalue Spectrums for Model ODE's It is instructive to consider the eigenvalue spectrums of the ODE's formulated by central di erencing the model equations for di usion and biconvection. These model ODE's are presented in Section 4. 4. for simplicity. ~ ~ g dt which is an uncoupled set of rst-order ODE's that can be solved independently for ~ the dependent variable vector w.3.28 had the alternative form ~ dw = w . Equations for the eigenvalues of the simple . we found that Eq. We say ~ If a solution is expressed in terms of w. it is said to be in eigenspace (often referred to as wave space). ~ g f 1 1 ~ = Xw ~ u ~ = X~ f g The elements of ~ relate directly to the local physics of the problem. Following Section 4. using only complete systems for our examples.2 and.

the model ODE's for di usion.5. Recall that xj = j x = j =(M + 1).3.4 sin m=1 2 M (4. from Section B. and x = 2 =M .1.1 m = x M 2 2 2 where = . Consider Eq.4.4. we nd for the model di usion equation with Dirichlet boundary conditions .2 + 2 cos Mm+ 1 m = x ! m .1 (4. REAL SPACE AND EIGENSPACE 63 tridiagonals.3. are given in Appendix B. Notice that the di usion eigenvalues are real and negative while those representing periodic convection are all pure imaginary.1 (4.i m a m m=0 1 M . to help visualize the matrix structure. we present results for a simple 4-point mesh and then we give the general case.ia sin 2m m=0 1 M . The right-hand eigenvector matrix X is given by 2 sin (x ) sin (2x ) sin (3x ) sin (4x ) 3 6 sin (x ) sin (2x ) sin (3x ) sin (4x ) 7 6 6 sin (x ) sin (2x ) sin (3x ) sin (4x ) 7 7 4 5 sin (x ) sin (2x ) sin (3x ) sin (4x ) The columns of the matrix are proportional to the eigenvectors. These follow as special cases from the results given in Appendix B. From Section B. 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. 4. m = m.3 Eigenvectors of the Model Equations The Di usion Model Next we consider the eigenvectors of the two model equations. B (M : a b c) and Bp(M : a b c).4. for the model biconvection equation .32) . The interpretation of this result plays a very important role later in our stability analysis.29) = 2(M + 1) x and.30) = sin m x m=0 1 M . First. 4.31) x is the modi ed wavenumber from Section 3.

Eq. 4. In general w = X . For the model di usion equation: ~ w = X . or left-hand eigenvector matrix X . ~ is a sine transform from real space to (sine) wave u space.1 M . Eq. ~ ~ u gives wm = M X j =1 uj sin mxj m=1 2 M (4.33. THE SEMI-DISCRETE APPROACH For the inverse. The coe cient matrices for these ODE's are always circulant. 1 The Biconvection Model Next consider the model ODE's for periodic convection.32 represents the sine synthesis that companions the sine transform given by Eq. In summary.64 CHAPTER 4.1 x = j 2 =M .34) . 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. ~ = Xw ~ u is a sine synthesis from wave space back to real space. 4.5.33 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( ) = 0.1 M . 4. For our model ODE. Eq.33) In the eld of harmonic analysis. the right-hand eigenvectors are given by ~ m = ei j x With xj = j M . .1 X m=0 (2 m=M ) j =0 1 m =0 1 M . Similarly. we can write ~ = X w as u ~ uj = wmeimxj j=0 1 (4.

REAL SPACE AND EIGENSPACE 2w 6w 6 6w 4 w 3 21 1 1 1 7 1 6 1 e. ~ is a complex Fourier transform from real space to u wave space.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 . ~ 7 u 5 1 X 1 M . ~ = Xw ~ u is a complex Fourier synthesis from wave space back to real space. we nd the following left-hand eigenvector matrix from Appendix B. u For circulant matrices.35) m m = . uj (t) = where M X m=1 cme m t sin mxj + (A. Eq. it is straightforward to establish the fact that the relation ~ = X w represents the Fourier synthesis of the variable w back to ~ .3.4. In summary.27 becomes We can now combine the results of the previous sections to write the solutions of our model ODE's.imxj wm = M m=0 1 M . i = e.4: 1 2 3 4 4 4 4 32u 76u 76 76u 54 u 1 2 3 4 3 7 7 = X.4 Solutions of the Model ODE's The Di usion Equation For the di usion equation. i = e . x. i = e.4 sin 2(M + 1) x 2 2 ! (4. i = 7= 6 . 1 4. 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.36) . ~ ~ u u =0 In general For any circulant system: ~ w = X . i = 7 4 61 e 5 4 e e .3. f )j 1 j=1 2 M (4. i = e.i = . u e.i = . 4.

i m a (4.39) We see that the solutions are identical except for the steady solution and the modi ed wavenumber in the transient term. we can write the ODE solution as uj (t) = M X m=1 cme. low wavenumber modes are accurately represented. while high wavenumber modes (if they have signi cant amplitudes) can have large errors. m 2 t sin m xj + h(xj ) j=1 2 M (4. evaluated at the nodes of the grid: uj (t) = M X m=1 cm e. we obtain uj (t) = where M .1 (4.40) m = . 2.38) This can be compared with the exact solution to the PDE.1 (4. Eq. The Convection Equation For the biconvection equation.41) with the modi ed wavenumber de ned in Eq.66 CHAPTER 4. This di erence causes the various modes (or eigenvector components) to decay at rates which di er from the exact solution. We can write this ODE solution as uj (t) = M .42) . The modi ed wavenumber is an approximation to the actual wavenumber. With conventional di erencing schemes.1 X m=0 cme. THE SEMI-DISCRETE APPROACH (4. m 2 t sin m xj + (A. The di erence between the modi ed wavenumber and the actual wavenumber depends on the di erencing scheme and the grid resolution. f )j 1 j=1 2 M (4.1 X m=0 cme m t ei mxj j=0 1 M .31.37.i m at ei mxj j=0 1 M .37) With the modi ed wavenumber de ned as 2 m x m = x sin 2 and using m = m. 4.

23 express identical results but in terms of di erent groupings of the dependent variables.44) m m m dt The goal in our analysis is to study typical behavior of general situations. The form of Eq. the result is erroneous numerical dispersion. the modi ed wavenumber is complex.42 shows that the imaginary portion of the modi ed wavenumber produces nonphysical decay or growth in the numerical solution. As discussed in Section 3. since is real. not particular problems. We found the uncoupled solution to a set of ODE's in Section 4.4 The Representative Equation In Section 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. Eq. In the case of non-centered di erencing.20 and 4. solved there in their uncoupled form.5. 2. and then returned as a coupled set to real space. X uj (t) = fm (0)e. uncoupled equation and our conclusions will apply in general. For such a purpose Eq. topics which are covered in Chapters 6 and 7. The role of m is clear it stands for some representative eigenvalue in the original A matrix. 4. A typical member of the family is dwm = w . discussed in Chapter 11. This is helpful both in analyzing the accuracy of time-marching methods and in studying their stability. evaluated at the nodes of the grid: M.1 (4. while the actual phase speed is independent of the wavenumber. this leads to an error in the speed with which various m modes are convected.4. THE REPRESENTATIVE EQUATION 1 67 and compare it to the exact solution of the PDE. g (t) (4.2.43) Once again the di erence appears through the modi ed wavenumber contained in .4.44 is not quite satisfactory. which are related by algebraic manipulation. However. 4. m=0 4.26. . we pointed out that Eqs.3. Our next objective is to nd a \typical" single ODE to analyze. Since the error in the phase speed depends on the wavenumber.i mat ei m xj j=0 1 M . The importance of this concept resides in its message that we can analyze timemarching methods by applying them to a single.

3. Consider the nite-di erence operator derived in question 1 of Chapter 3. Find the entries in the boundary-condition vector.45) for which Eq. 6x @t @x 2 2 . in principle.5 Problems 1. The exact solution of the representative ODE is (for 6= ): t (4. For example X . In such evaluations the parameters and must be allowed to take the worst possible combination of values that might occur in the ODE eigensystem. 3. write the semi-discrete form obtained with periodic boundary conditions on a 5-point grid (M = 5).52 to the 1-D di usion equation with Dirichlet boundary conditions.46) w(t) = ce t + ae . 2. 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. one can express any one of the forcing terms gm(t) as a nite Fourier series.68 CHAPTER 4. 4. Consider the application of the operator given in Eq. Using this operator to approximate the spatial derivative in the linear convection equation. we note that. This leads to w(t) = ce t + The Representative ODE dw = w + ae t dt (4.g(t) = ak eikt X ak eikt k ik .44 has the exact solution: k which can be used to evaluate all manner of time-marching methods. u(1) = 1: @u = @ u . 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. Write the resulting semi-discrete ODE form. From this we can extract the k'th term and replace ik with .

.5. The eigenvalues of the above matrix A. (Plot the solution at the grid nodes only. 1 . For initial conditions u(x 0) = sin(mx) and boundary conditions u(0 t) = u( t) = 0. as well as the matrices X and X . 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. compute and plot the ODE solution at t = 2 for the initial condition u(x 0) = sin x. PROBLEMS 69 4. Calculate and plot the corresponding ODE solutions. 5. Repeat for the initial condition u(x 0) = sin 2x. Using these. Compare with the exact solution of the PDE.) Calculate the corresponding modi ed wavenumbers for the second-order centered operator from Eq. can be found from Appendix B. plot the exact solution of the di usion equation with = 1 at t = 1 with m = 1 and m = 3.4. Note that the domain is 0 x 2 and x = 2 =10. 4.4.Bp(10 . Consider the matrix A = .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. The grid nodes are given by xj = j x j = 0 1 : : : 9. Consider a grid with 10 interior points spanning the domain 0 x .37.

70 CHAPTER 4. THE SEMI-DISCRETE APPROACH .

Chapter 5 FINITE-VOLUME METHODS In Chapter 3. of two advantages. either in the form of Eq..1. which is d Z QdV + I n:FdS = Z PdV dt V (t) S (t) V (t) (5.1 or Eq. In Chapter 4. Finite-volume methods have become popular in CFD as a result. primarily. mass. 5. they ensure that the discretization is conservative. 2. 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.e.1) We will begin by presenting the basic concepts which apply to nite-volume strategies. First. Finite-volume methods are applied to the integral form of the governing equations. 71 . Consistent with our emphasis on semi-discrete methods. we saw how to derive nite-di erence approximations to arbitrary derivatives. This increased exibility can be used to great advantage in generating grids about arbitrary geometries. and energy are conserved in a discrete sense. 2. 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. In this Chapter. As a result. it is obtained naturally from a nite-volume formulation. they can be applied on unstructured meshes consisting of arbitrary polyhedra in three dimensions or arbitrary polygons in two dimensions. we will show how similar semi-discrete forms can be derived using nite-volume approximations in space. Second. Next we will give our model equations in the form of Eq. i. momentum. we will study the latter form.2. This will be followed by several examples which hopefully make these concepts clear.

This ux must then be integrated to nd the net ux through the boundary. . 5. we note that the average value of Q in a cell with volume V is Q and Eq. which are a function of Q. Examining Eq. In order to evaluate the uxes. each cell will have a di erent piecewise approximation to Q. When these are used to calculate F(Q). Q can be represented within the cell by some piecewise approximation which produces the correct value of Q.1 can be written as 1 Z QdV V V (5. Thus the volume V in Eq.1 is that of a control volume whose shape is dependent on the nature of the grid.1 Basic Concepts CHAPTER 5. Let us consider these approximations in more detail. we will consider only control volumes which do not vary with time. they will generally produce di erent approximations to the ux at the boundary between two control volumes. the ux will be discontinuous. 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 have updated values of the cell-averaged quantities Q. In our examples. at the control-volume boundary. that is.4) for a control volume which does not vary with time. which is a closed surface in three dimensions and a closed contour in two dimensions. the source term P must be integrated over the control volume. A nondissipative scheme analogous to centered di erencing is obtained by taking the average of these two uxes. Thus after applying a timemarching method. As we shall see in our examples.72 5.2) at the next time step. Another approach known as ux-di erence splitting is described in Chapter 11. Next a time-marching method1 can be applied to nd the value of Z V QdV (5. First. This is a form of interpolation often referred to as reconstruction. we see that several approximations must be made.1. The basic elements of a nite-volume method are thus the following: 1 Time-marching methods will be discussed in the next chapter. 5. 5.3) I Z d V dt Q + S n:FdS = V PdV (5. The ux is required at the boundary of the control volume. Similarly.

The order of accuracy of the method is dependent on each of the approximations. 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.2. V S Z where the unit vector n points outward from the surface or contour.5. A rQdA = C nQdl I (5.5) or. This issue is discussed in Section 11.2 Model Equations in Integral Form 5.7) This PDE governs a simple plane wave convecting the scalar quantity. In order to include di usive uxes.2. Advance the solution in time to obtain new values of Q. construct an approximation to Q(x y z) in each control volume. These ideas should be clari ed by the examples in the remainder of this chapter. 2.1 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. Evaluate F(Q) at the boundary. Since there is a distinct approximation to Q(x y z) in each control volume. 3. in two dimensions. 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. u(x y t) with speed a along a straight line making an angle with respect to the x-axis. Given the value of Q for each control volume. . The one-dimensional form is recovered with = 0. nd Q at the control-volume boundary. Integrate the ux to nd the net ux through the control-volume boundary using some sort of quadrature. the following relation between rQ and Q is sometimes used: Z I rQdV = nQdS (5.6) 5.4. 4. Using this approximation.2.

3. x=2 to xj + x=2.16) 5.1=2 = xj .3 One-Dimensional Examples We restrict our attention to a scalar dependent variable u and a scalar ux f .13) (5. We consider an equispaced grid with spacing x. Eq.10) Since Q is a scalar. Substituting these expressions into a twodimensional form of Eq.12) (5.1. 2.17) . we nd (5. The integral form of the two-dimensional di usion equation with no source term and unit di usion coe cient is obtained from the general divergence form.2 The Di usion Equation Q = u F = . FINITE-VOLUME METHODS For unit speed a. We will use the following notation: xj.74 CHAPTER 5. as shown in Fig. the two-dimensional linear convection equation is obtained from the general divergence form.15) ! to be the integral form of the two-dimensional di usion equation. 2.3.ru ! @u + j @u = . with 5. Control volume j extends from xj . 5.8) (5. 2. d Z udA = I n: i @u + j @u ds dt A @x @y C (5. i @x @y P = 0 Using these.9) (5.2 gives the following integral form (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 (5.2.11) where A is the area of the cell which is bounded by the closed contour C .14) (5. F is simply a vector. as in the model equations. Eq. The nodes of the grid are located at xj = j x as usual. x=2 xj+1=2 = xj + x=2 (5.

fj.19) Zx d ( xu ) + f .1=2 = 0 . 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. With these de nitions. Hence the cell-average value and the value at the center of the cell di er by a term of second order. the cell-average value becomes 1 Z xj+1=2 u(x t)dx u (t) j uj 1=2 = u(xj 1=2) x fj 1=2 = f (uj 1=2) (5.1=2 Now with = x .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 .3. becomes u (5. uj = uj + O( x2) 5.5.1 A Second-Order Approximation to the Convection Equation In one dimension. xj .1=2 = x j+1=2 Pdx (5. fj. we can expand u(x) in Eq.18) (5.20) j j +1=2 dt j . the integral form of the linear convection equation.3. Eq. 5. 5.23) x ddtj + fj+1=2 .11. x=2 j @x j 2 @x2 j 6 @x3 j ! ! x2 @ 2 u + x4 @ 4 u + O( x6 ) = uj + 24 @x2 (5.21) 4 j 1920 @x j and the integral form becomes xj.22) where uj is the value at the center of the cell.1=2 or (5.1: Control volume in one dimension.

this point operator produces the following semidiscrete form: d~ = . gives fjR =2 = uj+1 (5. 1 is the cell to the left of xj. as shown in Fig. 1 (uj.30) . 2 where f^ denotes a numerical ux which is an approximation to the exact ux. Thus 1 f^j+1=2 = 2 (fjL+1=2 + fjR =2) = 1 (uj + uj+1) (5. The cell to the right of xj +1=2. 5.28) We have now accomplished the rst step from the list in Section 5.30 into the integral form.24) Evaluating this at j + 1=2 gives fjL+1=2 = f (uL+1=2) = uL+1=2 = uj (5. uj.1 + uj ) (5.1 + uj ) = x ddtj + 1 (uj+1 .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 . In this example. cell j is the cell to the right of xj. Substituting Eqs. 5. FINITE-VOLUME METHODS with f = u.1 0 1)~ u u (5. 5.1 (5.26) +1 Similarly. giving fjL.31) 2 2 With periodic boundary conditions.1 we have de ned the uxes at the cell boundaries in terms of the cell-average data.1=2 + fjR 1=2) = 1 (uj. we obtain u 1 u x ddtj + 2 (uj + uj+1) .1=2 = 2 (fjL. which is cell j + 1. giving fjR 1=2 = uj (5.1=2 x xj+1=2 (5. and cell j .76 CHAPTER 5.32) dt 2 x p .1=2 .23.29 and 5.29) +1 2 and 1 f^j. Eq.1) = 0 (5.1.1=2 = uj. the discontinuity in the ux at the cell boundary is resolved by taking the average of the uxes on either side of the boundary. We choose a piecewise constant approximation to u(x) in each cell such that u(x) = uj xj.1=2. 1 B (.27) .

1 with a piecewise quadratic approximation as follows u( ) = a 2 + b + c (5.36) .2 A Fourth-Order Approximation to the Convection Equation Let us replace the piecewise constant approximation in Section 5. the piecewise quadratic approximation produces (5.29 and 5.34) 1 These constraints lead to x Z 3 x=2 u( )d = uj+1 j a = uj+1 . as in Eqs. u u Hence our analysis and understanding of the eigensystem of the matrix Bp(.3. x=2 x=2 u( )d = uj (5. ONE-DIMENSIONAL EXAMPLES 77 This is identical to the expression obtained using second-order centered di erences. x=2 u( )d = u x .1 (5. and c are chosen to satisfy the following constraints: 1 Z . The three parameters a. and c.uj. Since the eigenvalues of Bp(.33) where is again equal to x .3.x j. b. we can conclude that the use of the average of the uxes on either side of the cell boundary.1 x Z x=2 . ~ .1 u b = uj+12 . xj .1 0 1) are pure imaginary.3 x=2 1 j . uj+1 24 With these values of a. except it is written in terms of the cell average ~ .35) the following values at the cell boundaries: uL+1=2 = 1 (2uj+1 + 5uj . rather than the nodal values.1 0 1) is relevant to nite-volume methods as well as nite-di erence methods.22ux2+ uj.1 + 26uj . b. 5. uj.1) j 6 c = .30.5. can lead to a nondissipative nite-volume method.3. 5.

uj+2 + 5uj+1 + 2uj ) j 6 1 uL. we again use the average of the uxes on either side of the boundary to obtain f^j+1=2 = 1 f (uL+1=2) + f (uR+1=2)] j 2 j 1 (5.uj+1 + 5uj + 2uj.1) and 1 f^j. 5. the following semi-discrete form is obtained: d~ = .40) = 12 (.uj+2 + 8uj+1 . FINITE-VOLUME METHODS uR. gives u 1 x ddtj + 12 (. With periodic boundary conditions. In this section. uj.1 + uj.1) j 6 uR+1=2 = 1 (. 5.41) Substituting these expressions into the integral form.1=2 = 2 f (uL.42) This is a fourth-order approximation to the integral form of the equation.1=2) + f (uR. uj.1 .1 .1=2 = 1 (. uj.1)~ u u (5. Recalling that f = u.uj+2 + 7uj+1 + 7uj .78 CHAPTER 5.38) (5.43) dt 12 x p This is a system of ODE's governing the evolution of the cell-average data.37) (5.23.8 0 8 . while the second approach is more suited to extension to higher order accuracy. 8uj. The rst approach is simpler to extend to multidimensions.1=2 = 6 (2uj + 5uj.1. 1 B (1 .2) 12 (5.3. Eq.3 A Second-Order Approximation to the Di usion Equation .2) j (5. we describe two approaches to deriving a nite-volume approximation to the di usion equation. as can be veri ed using Taylor series expansions (see question 1 at the end of this chapter).3.uj+1 + 7uj + 7uj.39) using the notation de ned in Section 5.1=2)] j j = 1 (.2) = 0 (5.

3.44. u ) (5. Eq.ru = .16. we use a piecewise quadratic approximation as in Section 5. the integral form of the di usion equation. Hence.1 j (5.33 we have dt x @u = @u = 2a + b @x @ (5. 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. becomes with f = .51) . 5.2. and we see that the properties of the matrix B (1 . u ) (5.3. = .5. uj.1) dt x j . 2u + u ) or. ONE-DIMENSIONAL EXAMPLES u x ddtj + fj+1=2 .44) (5.22. For our second approach.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 . d~ = 1 B (1 . to second-order.2 1)~ + bc u u ~ 2 f^j. fj. we obtain x duj = 1 (u . 1 (u . Substituting these into the integral form. 5. 1x (uj . 5. From Eq.45) In one dimension.47) j +1=2 @x j+1=2 x j+1 j Similarly.1=2 = . u(a) @x x xj @x x j+1 j From Eq.46) @u dx = u(b) .2 1) are relevant to the study of nite-volume methods as well as nite-di erence methods. with Dirichlet boundary conditions. Eq.50) j +1 This provides a semi-discrete nite-volume approximation to the di usion equation.@u=@x. Also. 5. 5.1=2 = 0 Zb 79 (5.48) (5. Eq. we can write ! @u f^ = .49) (5.

FINITE-VOLUME METHODS j +1=2 j +1=2 with a and b given in Eq. With f = .56) A 3 The two-dimensional form of the conservation law is d Z QdA + I n:Fdl = 0 dt A C (5. u ) x j+1 j (5. and ` is the length of the sides of the hexagons.54) which is identical to Eq. we use a piecewise constant approximation in each control volume and the ux at the control volume boundary is the average of the uxes obtained on either side of the boundary. The following relations hold between `. 1 (u . 1x (uj . This produces duj = 1 (u .49. this gives f R = f L = .52) (5. fjR 1=2 = fjL. The nodal data are stored at the vertices of the triangles formed by the grid. As in Section 5. The control volumes are regular hexagons with area A.53) Notice that there is no discontinuity in the ux at the cell boundary.2.3.1=2 = .2 1)~ u u (5. . Thus our nal example is a nite-volume approximation to the two-dimensional linear convection equation on a grid consisting of regular triangles. as shown in Figure 5.1 j j +1 dt x2 p 5. 5.57) .@u=@x.1) . The resulting semi-discrete form with periodic boundary conditions is d~ = 1 B (1 . 2u + u ) (5. is the length of the sides of the triangles. uj.80 CHAPTER 5. and A. dt x2 j . 1 ` = p 3 p A = 3 2 3 `2 ` = 2 (5.4 A Two-Dimensional Example The above one-dimensional examples of nite-volume approximations obscure some of the practical aspects of such methods. 5.35.55) which is written entirely in terms of cell-average data.1.

see Fig. i and j are unit normals along x and y.2. as shown in Figure 5. . where we have ignored the source term. Side 0 1 2 3 4 5 Outward Normal n (i . Since these sides are all straight. 3 j)=2 p Table 5. The contour in the line integral is composed of the sides of the hexagon.2.i p (. Outward normals. 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.5.4. respectively.1. A TWO-DIMENSIONAL EXAMPLE c b 2 3 81 p d 4 1 l ∆ 0 5 a e f Figure 5.2: Triangular grid. A list of the normals for the mesh orientation shown is given in Table 5. 3 j)=2 pi (i + p j)=2 3 (. 5.1.i .i + 3 j)=2 .

3 sin )=2 p Table 5. the two-dimensional linear convection equation. That is. FINITE-VOLUME METHODS For Eq.60.1=2 (5.59) Then. one has the expression u( )d = ` .60) dt A . Weights of ux integrals.1=2 =0 The values of n (i cos + j sin ) are given by the expressions in Table 5. the approximation to the ux integral becomes trivial. cos p (.60. 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 .11. 5.61) . 5. in terms of u and the hexagon area A.82 CHAPTER 5.`=2 u ( )d (5.2. if the integrals in the equation are evaluated exactly. There .58) where is a length measured from the middle of a side . 5. the integrated time rate of change of the integral of u over the area of the hexagon is known exactly. Introducing the cell average.62) 2 1 A u dA = Aup (5.1=2 u(z)dz (5.`=2 Z `=2 Z 1=2 are no numerical approximations in Eq. 3 sin )=2 cos p (cos + p sin )=2 3 (.2. Making the change of variable z = =`. cos + 3 sin )=2 . we have for side n Z Fdl = n (i cos + j sin ) Z `=2 . see Eq. Z and the piecewise-constant approximation u = up over the entire hexagon. Side 0 1 2 3 4 5 n (i cos + j sin ) (cos . cos . we have " # 5 d Z u dA + X n (i cos + j sin ) ` Z 1=2 u(z)dz = 0 (5.

along with the expressions in Table 5. using Taylor series expansions. cos + 3 sin )(uc . cos + 3 sin )(uc .68) or dup + 1 (2 cos )(u .67) Z Z u(z)dz = up + ud 2 4 u(z)dz = up + ue 2 5 Z Z Substituting these into Eq.5 Problems 1.66) (5.69) The reader can verify. Use Taylor series to verify that Eq. ud) + (cos + 3 sin )(ub . we obtain u(z)dz = up + uf 2 0 p u ` A ddtp + 2 (2 cos )(ua .5.42 is a fourth-order approximation to Eq. u ) + (cos + p3 sin )(u .23. we have for the other ve edges: Z 83 u(z)dz = up + ub 2 2 u(z)dz = up + uc 2 3 (5.65) (5. uf )] = 0 (5. 5.63) (5. 2. 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.2. 5. u ) a d b e dt 3 p +(. 5.5. ue) p +(. PROBLEMS Similarly. that this is a second-order approximation to the integral form of the two-dimensional linear convection equation.64) (5. uf )] = 0 (5. 5.60. Use the following linear approximation: u( ) = a + b .

FINITE-VOLUME METHODS u a = uj+12 . .3. Repeat question 3 for a grid consisting of equilateral triangles.3.1 and use the average ux at the cell interface. 4. derive a nite-volume approximation to the spatial terms in the two-dimensional di usion equation on a square grid.84 where b = uj and CHAPTER 5. Using the rst approach given in Section 5.x j. 3.

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

and. 6. but the methods themselves are formed by linear combinations of the dependent variable and its derivative at various time intervals. TIME-MARCHING METHODS FOR ODE'S Application of a spatial discretization to a PDE produces a coupled system of ODE's. we reduce our PDE to a set of coupled ODE's represented in general by Eq. They are both commonly used in the literature on ODE's. In Chapter 2. In earlier chapters. where accuracy can be measured either in a local or a global sense. Application of a time-marching method to an ODE produces an ordinary di erence equation (O E ). the reader should recall the arguments made in Chapter 4 to justify the study of a scalar ODE.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 developed exact solutions to our model PDE's and ODE's. 6. we need only consider the scalar case Although we use u to represent the dependent variable.1 un. using this theory.1 du = u0 = F (u t) dt (6. but we postpone such considerations to the next chapter. In this chapter we will present some basic theory of linear O E's which closely parallels that for linear ODE's. 4.1 Notation Using the semi-discrete approach.1. 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. for the purpose of this chapter. and h represents the time interval t. 6.86 CHAPTER 6. The methods we study are to be applied to linear or nonlinear ODE's. Our rst task is to nd numerical approximations that can be used to carry out the time integration of Eq. we will develop exact solutions for the model O E's arising from the application of timemarching methods to the model ODE's. etc.1u0n. we introduced the convention that the n subscript. u.3) . always points to a discrete time value. rather than w. Combining this notation with Eq. We then face a further task concerning the numerical stability of the resulting methods. or the (n) superscript. However.3 to some given accuracy. They are represented conceptually by (6.4) un+1 = f 1 u0n+1 0u0n .1 0 un .

The method commonly referred to as .4.1 respectively. by applying a time-marching method. The value of this equation arises from the fact that. which is a fully-discrete method.3 1 Here we give only MacCormack's time-marching method. CONVERTING TIME-MARCHING METHODS TO O E'S 87 With an appropriate choice of the 0s and 0s. The second is implicit and referred to as the implicit (or backward) Euler method.1 for a method using two previous time levels. An explicit method is one in which the new predicted solution is only a function of known data. As we shall see. 6.7) du = u0 = u + ae t (6. these methods can be constructed to give a local Taylor series accuracy of any order.5) (6. For an implicit method. The methods are said to be explicit if 1 = 0 and implicit otherwise. the new predicted solution is also a function of the time derivative at the new time level. u. implicit methods require more complicated strategies to solve for un+1 than explicit methods. and therefore the time advance is simple. 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. u0n.6) According to the conditions presented under Eq. These methods are simple recipes for the time advance of a function in terms of its value and the value of its derivative. for systems of ODE's and nonlinear problems. at given time intervals. 6.1.6. the rst and third of these are examples of explicit methods.2.8) dt written here in terms of the dependent variable. for example. that is.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. will be presented in Section 11. u0n. We refer to them as the explicit Euler method and the MacCormack predictor-corrector method. un and un. u0n+1. we can analytically convert MacCormack's method.

and just as powerful as. Apply the simple explicit Euler scheme. gives un+1 . but for di erence rather than di erential equations. to Eq.7.11 is obtained by noting that u0n+1 = F (~n+1 tn + h) ~ u = un+1 + ae h(n+1) ~ (6. 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. 6.8. Eq. 6. 6. since the predictor step in Eq.3 Solution of Linear O E's With Constant Coe cients The techniques for solving linear di erence equations with constant coe cients is as well developed as that for ODE's and the theory follows a remarkably parallel path. 6. with constant coe cients.7 is simply the explicit Euler method. This is demonstrated by repeating some of the developments in Section 4. the predictor-corrector sequence. Note that the rst line in Eq.10) As a nal example. 6. (1 + h)un = ahe hn ~ 1 u (6. The second line in Eq.9 is a linear O E . 1 un = 1 ahe h(n+1) 2 2 which is a coupled set of linear O E's with constant coe cients.2. There results un+1 = un + h( un + ae hn) or un+1 . un = he h ae hn (6. applying the implicit Euler method. 6. As another example. expressed in terms of the dependent variable un and the independent variable n. to Eq. h)un+1 .9) Eq.9. 6. 6. 6. . The latter are subject to a whole body of analysis that is similar in many respects to.5. (1 + h)un = hae hn (6. We next consider examples of this conversion process and then go into the general theory on solving O E's. Eq. 6.88 CHAPTER 6.11) . TIME-MARCHING METHODS FOR ODE'S such a linear ODE into a linear O E . the theory of ODE's.8.6. we nd un+1 = un + h un+1 + ae h(n+1) or (1 . 2 (1 + h)~n+1 + un+1 .11 is identical to Eq. 6.

is not a function of either n or u. The independent variable is now n rather than t.1 n This can all be easily veri ed by substitution.1 First. In terms of the initial value of u it can be written Just as in the development of Eq. un+1 = un + a n is h i un = u0 + an . in general. 4.14) d Instead of the di erential operator D dt used for ODE's. n un = c1 n + bab . (b = ). un = u0 n + a b b . The exact solution of Eq. SOLUTION OF LINEAR O E'S WITH CONSTANT COEFFICIENTS 89 6. n Second-Order Equations The homogeneous form of a second-order di erence equation is given by un+2 + a1 un+1 + a0un = 0 (6. and b are.13 is where c1 is a constant determined by the initial conditions. we use for O E's the di erence operator E (commonly referred to as the displacement or shift operator) and de ned formally by the relations un+1 = Eun un+k = E k un Further notice that the displacement operator also applies to exponents. n.10. and since the equations are linear and have constant coe cients.3.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 (6.3.6.13) where . 6. one can readily show that the solution of the defective case. a. complex parameters. thus b bn = bn+ = E bn .

15) which must be zero for all un.2 Special Cases of Coupled First-Order Equations A Complete System Coupled.90 CHAPTER 6. E I ]~ = 0 un c21 (c22 .14 can now be re-expressed in an operational notion as (E 2 + a1E + a0)un = 0 (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 (6.17) which can also be written ~ n+1 = C~ n u u (c11 . 6. and refer to them as the -roots. TIME-MARCHING METHODS FOR ODE'S where can be any fraction or irrational number.14. Thus Eq. 6. etc. E I ]. In the simple example given above. 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. ) (c c12 ) = 0 21 22 . E ) C = c11 c12 c21 c22 The operational form of Eq. The -roots are found from P ( ) = det (c11c . 6. rst-order.17 can be written u1 (n) = C . 6. . The roles of D and E are the same insofar as once they have been introduced to the basic equations the value of u(t) or un can be factored out. They are found by solving the equation P ( ) = 0. we label these roots 1 . In the analysis of O E's.14 for all c1 c2 and n should be veri ed by substitution.15 is known as the operational form of Eq.16) where c1 and c2 depend upon the initial conditions.3. 2 . there are just two roots and in terms of them the solution can be written un = c1( 1 )n + c2( 2 )n (6. E ) u2 which must be zero for all u1 and u2. its roots determine the solution to the O E. The fact that Eq. . 6. 6. Again we are led to a characteristic polynomial. Eq.16 is a solution to Eq. 6. this time having the form P (E ) = det C .

1) . 6. E . For example.1 + u n(n .4.(1 + h) # " u # = h " 1 # ae hn ~ 1E 1 (1 + h)E E . SOLUTION OF THE REPRESENTATIVE O E'S 91 Obviously the k are the eigenvalues of C and.11.19) (6.18) 6.1 n ^ ^ " # .1 The Operational Form and its Solution Examples of the nonhomogeneous. linear. rst-order ordinary di erence equations.2 un = u0 + u0n ^ 0 2 n (6. are given by Eqs. 1]un = h E ae hn (6. produced by applying a time-marching method to the representative equation. 1 u n . 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. 6.20) (6. h)E . Using the displacement operator.2.21) " . following the logic of Section 4. these equations can be written E .6. ~ are its eigenvectors.17 is if x 2 ~ n = X ck ( k )n~ k u x k=1 where ck are constants determined by the initial conditions. the solution of Eq. (1 + h)]un = h ae hn (1 . The solution of O E's with defective eigensystems follows closely the logic in Section 4.4.2.2 2 2 E .9 to 6.4 Solution of the Representative O E's 6.

representing a time invariant particular solution. we have P (E ) = (1 . we have P (E ) = E .4. Eq.23) P (e ) k=1 where k are the K roots of the characteristic polynomial.22 can be expressed as K h X un = ck ( k )n + ae hn Q(e h) (6.2 Examples of Solutions to Time-Marching O E's As examples of the use of Eqs. 6. 6.1. we derive the solutions of Eqs. ~ the important subset of this solution which occurs when = 0. 6.25) .45.21. 6. as would be the case for Eq.22 all manner of standard time-marching methods having multiple time steps and various types of intermediate predictor-corrector families. The terms P (E ) and Q(E ) are polynomials in E referred to as the characteristic polynomial and the particular polynomial.20. 6. When determinants are involved in the construction of P (E ) and Q(E ). 1 .19. 4. 6. 6. P ( ) = 0.19 to 6. 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 (6.92 CHAPTER 6. We can express in terms of Eq. Keep in mind that for methods such as in Eq. h)E . For the explicit Euler method.22) which is produced by applying time-marching methods to the representative ODE. 6. Eq.21 there are multiple (two in this case) solutions. the ratio Q(E )=P (E ) can be found by Kramer's rule. 1 Q(E ) = hE e h. h Q(E ) = h (6. 6. The general solution of Eq. respectively. h h (6. one for un and un and we are usually only interested in the nal solution un.23.21. In such a case K X Q(1) un = ck ( k )n + a P (1) k=1 6. Eq.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. Notice also. or a steady state.22 and 6.

h .(1 + h) = E E .27. and there ~ results # " E .1 Establishing the Relation 6. 1 2 2 2 h Q(E ) = det . The former are the eigenvalues of the A matrix in the ODE's found by space di erencing the original PDE. h .26) e h .5.5 The . There is a fundamental relation between the two which can be used to identify many of the essential properties of a time-march method. 6. Remembering that t = nh. 1 . 1 2h2 P (E ) = det . h)e h . 1 .6. 4.21. THE . Eq. the -roots and the -roots.27) x . 2 2 h2 = 0 which has only one nontrivial root ( = 0 is simply a shift in the reference index). 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. 1 (1 E h)E 1 hE = 1 hE (E + 1 + h) 2 2 + 2 The -root is found from 1 P( ) = . RELATION so 93 In the case of the coupled predictor-corrector equations. one solves for the nal family un (one can also nd a solution for the intermediate family u). 1 2h2 2 he h 1 n un = c1 1 . This relation is rst demonstrated by developing it for the explicit Euler method. h . 1 (1 + h)E E . h + ae hn (1 . 1 6. 1 . The complete solution can therefore be written 1 h 1 2h2 n + ae hn 2 h e + 1 + h un = c1 1 + h + 2 (6.5. one can write n u(t) = c1 e 1 h ~ 1 + x + cm e mh n ~m + x + cM e Mh n ~ M + P:S: (6. 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.

6.27 and Eq. we use the leapfrog method for the time advance. one -root. 3 The error is O( 2 h2 ). Suppose.28) where the cm and the ~ m in the two equations are identical and m = (1 + m h). 2 hE .94 CHAPTER 6.3. 2 mh m . which is given by = 1 + h.28.29. TIME-MARCHING METHODS FOR ODE'S where for the present we are not interested in the form of the particular solution (P:S:). we see a correspondence between m and e m h. instead of the Euler method. 6. m 2 Based on Section 4. 6. Since the value of e h can be expressed in terms of the series 1 e h = 1 + h + 2 2h2 + 1 3h3 + 6 1 + n! nhn + the truncated expansion = 1 + h is a reasonable3 approximation for small enough h. 6. So if we use the Euler method for the time advance of the ODE's.8 to Eq. we have the characteristic polynomial P (E ) = E 2 .1 + 2hu0n (6.31) (6. 1 = 0 2 (6. 1.29) Applying Eq.32) mh 8 m This is an approximation to e m h with an error O( 3h3). For one of these we nd m q h+ 1+ m 1 = 1 + mh + 2 = 2 h2 . Now the explicit Euler method produces for each -root. x Comparing Eq. The other root. so that for every the must satisfy the relation m .4. .30) Now we notice that each produces two -roots. . which is de ned by un+1 = un. 1 4 h4 + m 2 h2 m (6. will be discussed in Section 6.5. q 1 + 2 h2 . 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.

35) has a leading truncation error O(h3). One of these we identi ed as the principal root which always .5. and designate it ( m )1 . 4. always produces one -root for every -root that satis es the relation (6.2 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. twice as many time steps are required to reach the time T .1 + 2hu0n (6. which is the leapfrog method: un+1 = un.3 Spurious -Roots We saw from Eq. this is reduced by a factor of eight (considering the leading term only). relation for the leapfrog method produces two -roots for each . consistent with a second-order approximation. un. 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. while the second-order time-marching method which results from this approximation.33) 1 2h2 + + 1 k hk + O hk+1 =1+ h+ 2 k! where k is the order of the time-marching method.34. 6. 6. We refer to the root that has the above property as the principal -root. Since the error per time step is O(h3). However.6. This arises simply because of our notation for the time-marching method in which we have multiplied through by h to get an approximation for the function un+1 rather than the derivative as in Eq. The following example makes this clear.34) 2h has a leading truncation error which is O(h2).1) (6.5.5. THE . RELATION 95 6. Thus the principal root is an approximation to e h up to O hk . knowing only that its leading error is O hk+1 . Therefore the error at the end of the simulation is reduced by a factor of four. 6. The above property can be stated regardless of the details of the timemarching method.30 that the .6. Note that a second-order approximation to a derivative written in the form ( t u)n = 1 (un+1 .

if there is one spurious root to a method.1 and thus cannot be started using u only ~ n. However. except for the principal one. 1 or earlier to advance the solution from time level n to n + 1.. They are also called one-step methods. To express this fact we use the notation = ( h). many very accurate methods in practical use for integrating some forms of ODE's have spurious roots.36 must be initialized by some starting procedure. they play virtually no role in accuracy analysis. All spurious roots are designated ( m )k where k = 2 3 . The other is referred to as a spurious -root and designated ( m )2 . 1 or earlier.4. That is. 6. generation of spurious roots does not. The initial vector ~ 0 u does not provide enough data to initialize all of the coe cients. in itself. if one starts the method properly) the spurious coe cients are all initialized with very small magnitudes. It should be mentioned that methods with spurious roots are not self starting. ..36) Spurious roots arise if a method uses data from time level n . Thus while spurious roots must be considered in stability analysis. TIME-MARCHING METHODS FOR ODE'S has the property given in 6. We refer to them as one-root methods. No matter whether a -root is principal or spurious. 6. For example. make a method inferior. and presumably the magnitudes of the spurious roots themselves are all less than one (see Chapter 7). Following again the message of Section 4. 6. the solution for the O E in the form shown in Eq. For example. u Presumably (i. If a time-marching method produces spurious -roots. Then the presence of spurious roots does not contaminate the answer. relation produced by a time-marching scheme can result in multiple -roots all of which. the leapfrog method requires ~ n. Such roots originate entirely from the numerical approximation of the time-marching method and have nothing to do with the ODE being solved.96 CHAPTER 6.5.e.33.4 One-Root Time-Marching Methods There are a number of time-marching methods that produce only one -root for each -root. after some nite time the amplitude of the error associated with the spurious roots is even smaller then when it was initialized. 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. In general. In fact. all of the coe cients (cm )2 in Eq. the . are spurious. it is always some algebraic function of the product h. This results because methods which produce spurious roots require data from time level n . if there are more spurious roots (6.28 must be modi ed.

and the implicit Euler methods ( = 1).4.1~ u x x x f (6.1 Local and Global Error Measures There are two broad categories of errors that can be used to derive and evaluate timemarching methods. For example.1~ u x x x f ~ n = c1( 1 )n ~ 1 + + cm( m )n ~ m + + cM ( M )n ~ M + A. it is of no use in: .6. However. Its . the trapezoidal ( = 2 ). It is useful for comparing methods.2. Three one-root methods were analyzed in Section 6. h h . The only error made by introducing the time marching is the error that makes in approximating e h. It is quite common to judge a time-marching method on the basis of results found from a Taylor table. as we shall see in Chapter 8. is given by the formula h i un+1 = un + h (1 . This is a local error such as that found from a Taylor table analysis. the so-called -method. It is usually used as the basis for establishing the order of a method.6. and matrices are identical except the ~ and the terms inside u the parentheses on the right hand sides. This is a global error. see Section 3.4. One is the error made in each time step.6 Accuracy Measures of Time-Marching Methods 6. ACCURACY MEASURES OF TIME-MARCHING METHODS 97 1 The -method represents the explicit Euler ( = 0). )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. vectors. These are ~ (t) = c1 e 1 h n ~ 1 + + cm e m h n ~ m + + cM e M h n ~ M + A.6. so that all the constants. 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. relation is ) = 1 +1(1 . 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. Notice that when t and n = 0. A popular method having this property. respectively. 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.37) respectively.

a necessary condition for the choice should be that the measure can be used consistently for all methods. a very natural local error measure for the transient solution is the value of the di erence between solutions based on these two roots.39) j j 6.38) and the solution to the representative O E's. However. This is similar to the error found from a Taylor table. The latter three of these are of concern to us here. either local or global. TIME-MARCHING METHODS FOR ODE'S nding spurious roots. In the discussion of the .relation we saw that all time-marching methods produce a principal -root for every -root that exists in a set of linear ODE's. is to some extent arbitrary.98 CHAPTER 6. and to study them we make use of the material developed in the previous sections of this chapter. We designate this by er and make the following de nition er e h. .2 Local Accuracy of the Transient Solution (er Transient error er! ) The particular choice of an error measure. 1 The leading error term can be found by expanding in a Taylor series and choosing the rst nonvanishing term.6. which can be written as h un = c1 ( 1)n + ae hn Q(e h ) P (e ) (6. The order of the method is the last power of h matched exactly. (6. Therefore. including only the contribution from the principal root. Our error measures are based on the di erence between the exact solution to the representative ODE. analyzing the particular solution of predictor-corrector combinations. given by t u(t) = ce t + ae . evaluating numerical stability and separating the errors in phase and amplitude. nding the global error.

1( h).1 ( 1 )i=( 1 )r )] and the local error in phase can be de ned as (6. The principal root. A positive value of erp corresponds to phase lag (the numerical phase speed is too small).a .ia . Cn = ah= x. tan. The above expression for er! can be normalized to give the error in the phase speed. ( 1)2 + ( 1 )2 i r er! !h . Such can indeed be the case when we study the equations governing periodic convection which produces harmonic motion. while a negative value corresponds to phase lead (the numerical phase speed is too large).6. we have h = . We have found these to be given by 1 P:S:(ODE) = ae t ( . Let = i! where ! is a real number representing a frequency. For such cases it is more meaningful to express the error in terms of amplitude and phase. Thus one can obtain values of the principal root over the range 0 x for a given value of the Courant number. ACCURACY MEASURES OF TIME-MARCHING METHODS 99 Amplitude and Phase Error Suppose a eigenvalue is imaginary. !h Cn x 6.40) 1 = r + i i ei!h From this it follows that the local error in amplitude is measured by the deviation of j 1 j from unity. that is q era = 1 .5 can be extended to the combination of a spatial discretization and a time-marching method applied to the linear convection equation. The approach to error analysis described in Section 3.6.iCn x.42) er = er! = 1 + tan ( 1)i =( 1)r )] p where ! = . Introducing the Courant number. ) . 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. j 1 j = 1 .41) Amplitude and phase errors are important measures of the suitability of time-marching methods for convection and wave propagation phenomena. is found using = . Then the numerical method must produce a principal -root that is complex and expressible in the form (6. where is the modi ed wavenumber of the spatial discretization.1 (6.

and for this case. a time-marching method is said to be of order k if er = c1 ( h)k1+1 er = c2 ( h)k2+1 where k = smallest of(k1 k2) (6. In order to determine the leading error term. where n ( . many numerical methods generate an er that is also zero. 1 (ODE ) ( ) (6.P e h o (6. 6. 6.47) .100 and CHAPTER 6.45) (6. An illustration of this is given in the section on Runge-Kutta methods. 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. Eq.4 Time Accuracy For Nonlinear Applications In practice.43 can be written in terms of the characteristic and particular polynomials as er = co . The algebra involved in nding the order of er can be quite tedious. )Q e h . this order is quite important in determining the true order of a time-marching method by the process that has been outlined.h ) lim P e The value of co is a method-dependent constant that is often equal to one. If the forcing function is independent of time. and it is necessary that the advertised order of accuracy be valid for the nonlinear cases as well as for the linear ones. so that the order of er and er are consistent. More precisely. However.43) The multiplication by h converts the error from a global measure to a local one. time-marching methods are usually applied to nonlinear ODE's.44) co = h!0 h( . For a measure of the local error in the particular solution we introduce the de nition er P:S: h P:S:(O E) . is equal to zero.6. TIME-MARCHING METHODS FOR ODE'S h P:S:(O E) = ae t Q(e h) P (e ) respectively.46) (6.

we are more concerned with the global error in amplitude and phase.1 ( 1) i 1r . These are useful when we come to the evaluation of time-marching methods for speci c purposes.6. Let T be the xed time of the event and h be the chosen step size. ( 1 ( h))N (6. 6. given by the relation T = Nh Global error in the transient A natural extension of er to cover the error in an entire event is given by Er e T .6.48) Global error in amplitude and phase If the event is periodic. and q " ( 1 )2 + ( 1)2 i r N (6. We refer to such a computation as an \event". is N.50) .6. Then the required number of time steps. we can also de ne global error measures. tan. ACCURACY MEASURES OF TIME-MARCHING METHODS 101 The reader should be aware that this is not su cient.49) ( ) Er! N !h . the analysis based on a linear nonhomogeneous ODE produces the appropriate conditions for the majority of time-marching methods used in CFD.5 Global Accuracy In contrast to the local error measures which have just been discussed. Suppose we wish to compute some time-accurate phenomenon over a xed interval of time using a constant time step. However.1 ( ) =( ) ] = !T . 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. These are given by Era = 1 . This subject is covered in Chapter 8 after our introduction to stability in Chapter 7. N tan 1i 1r !# (6. For example.

7.K k un+k = h du = u0 = F(u t) dt 1 X k=1.8. 6. The methods are said to be linear because the 's and 's are independent of u and n.7 Linear Multistep Methods In the previous sections. ) Qeh . since most of them have historic origins from a wide variety of disciplines and applications. 6.1 The General Formulation When applied to the nonlinear ODE all linear multistep methods can be expressed in the general form 1 X k=1.K 1 k hn k E A( un + ae ) (6. one nds 0 1 @ X k=1. Eq. 6. It can be measured by Er ( .52) . and the result is expressed in operational form. They are explicit if 1 = 0 and implicit otherwise.51) where the notation for F is de ned in Section 6. In the subsequent sections. h.K 1 0 1 X k k E Aun = h@ k=1.51 is applied to the representative equation. 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. developmental or design issues. We shall not spend much time on development or design of these methods. the global error in the particular solution follows naturally by comparing the solutions to the ODE and the O E. we have developed the framework of error analysis for time advance methods and have randomly introduced a few methods without addressing motivational. When Eq.4. TIME-MARCHING METHODS FOR ODE'S Global error in the particular solution Finally.102 CHAPTER 6. we introduce classes of methods along with their associated error analysis.1. 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 (6.1 P eh 6.

51 if X X X k = 0 and k = (K + k . can be designed using the Taylor table approach of Section 3.2 . The Adams-Moulton family is obtained from Eq.2 . referred to as Adams-Bashforth (explicit) and AdamsMoulton (implicit).51 with k = . One can show that these conditions are met by any method represented by Eq.2 u0n. 6.un .h 0 u0n .4. The three-step Adams-Moulton method can be written in the following form un+1 = un + h( 1u0n+1 + 0 u0n + .53) 1=1 0 = . Taylor table for the Adams-Moulton three-step linear multistep method. 1) k Since both sides of Eq. k k k k 6. Two well-known families of them. un h u0n h2 u00 n h3 u000 n h4 u0000 n .1 2 . 6.0 1 1 .2 . labeled 1. LINEAR MULTISTEP METHODS 103 We recall from Section 6.7.1 .1 6 .54) A Taylor table for Eq. .(.2 1 6 Table 6.(. 6.2)3 .h 1 u0n+1 .1. The condition referred to as consistency simply means that ! 1 as h ! 0.1 . We can also agree that.2)0 .1 k=0 The Adams-Bashforth family has the same 's with the additional constraint that 1 = 0.2 2 .1 . 6. to be of any value in time accuracy.1 . that approximates e h through the order of the method.h .44.2 that a time-marching method when applied to the representative equation must provide a -root.2)2 .(. and it is certainly a necessary condition for the accuracy of any time marching method.51 can be multiplied by an arbitrary constant.2 Examples There are many special explicit and implicit forms of linear multistep methods.2 (6.7.5. that is ! (1 + h) as h ! 0.(.h .1 . 6.54 can be generated as 1 1 1 un+1 1 1 2 6 24 .1 1 . 11 6 . these methods are often \normalized" by requiring X k=1 Under this condition co = 1 in Eq. . 12 .1 + .2u0n.1 u0n. a method should at least be rst-order accurate.1 u0n.6.1 1 .2)1 .2) (6.

32 1 .55) 4 5 4 . 6.57 up to the order of the method.1 7 = 6 1 7 5 4 5 4 0 . 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.5=24 .4 With 1 = 0 one obtains 2 32 3 2 3 1 1 1 0 7 617 6 0 .1 = .1i = un + 12 AM3 .58) 0 = 23=12 . 16u0n.2 to solve for the 's. = un + hu0n+1 Implicit Euler 1 hhu0n + u0n+1i = un + 2 Trapezoidal (AM2) 1 h4un .55 and 6.2 .1 5 = 4 1 5 (6.1i = un + 2 h 0 i AB2 h = un + 12 23un . u0n. resulting in (6.4 . TIME-MARCHING METHODS FOR ODE'S This leads to the linear system 2 1 1 1 1 32 3 213 1 6 2 0 .104 CHAPTER 6. Explicit Methods un+1 un+1 un+1 un+1 = un + hu0n Euler 0 = un.16=12 . some of which are very common in CFD applications. A list of simple methods.2 = 1=24 which produces a method which is fourth-order accurate. One can verify that the Adams type schemes given below satisfy Eqs. un.1 + 2hhun Leapfrog 1 h 3u0n .1 + 5u0n. is given below together with identifying names that are sometimes associated with them.2 that a kth-order time-marching method has a leading truncation error term which is O(hk+1 ).1 = .5.2 AB3 Implicit Methods un+1 un+1 un+1 un+1 4 Recall from Section 6.1 + 2hu0n+1i =3 2nd-order Backward h h5u0n+1 + 8u0n .4 7 6 0 7 6 1 7 6 7 6 7 6 3 0 3 12 7 6 76 (6.56) 1 = 9=24 0 = 19=24 .2 giving (6.57) 4 54 0 3 12 1 .2 = 5=12 This is the third-order Adams-Bashforth method.4 7 6 .2 . u0n.

and two-step methods. un.1 (6. A list of methods contained in Eq. =3 = 6 . LINEAR MULTISTEP METHODS 105 6.59) Clearly the methods are explicit if = 0 and implicit otherwise. . Some linear one. +2 The class of all 3rd-order methods is determined by imposing the additional constraint =2 .1 = 0 in Eq.2.e.7. 'u0n. can be written as h i (1 + )un+1 = (1 + 2 )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. + ')u0n . ' 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 . 6.1=3 0 1=12 .1] + h u0n+1 + (1 . Methods with = .e.1=3 . which corresponds to . the methods are 2nd-order accurate) if 1 '= .1=6 0 .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.2.2=9 0 1=2 . which corresponds to 0 = 0 in Eq.5 6 1 Finally a unique fourth-order method is found by setting = .51.7.1=4 .59 is given in Table 6. One can show after a little algebra that both er and er are reduced to 0(h3) (i.1=2 .' = .. 6.59.1=2 .51.51). that is at least rst-order accurate.6.1=2.1=2 0 . 6.1=2 . see Eq. K=2 in Eq. This limits the interest in multistep methods to about two time levels..1=6 .5=6 . 6. Notice that the Adams methods have = 0.1=2 0 0 0 0 . are known as Milne methods. The most general two-step linear multistep method (i. 6.

therefore. and usually the nal family has a higher Taylor-series order of accuracy than the intermediate ones. The situation for er requires some algebra.61) (6. For the method to be second-order accurate both er and er must be O(h3). Their use is motivated by ease of application and increased e ciency. but it is not di cult to show using Eq.4. 6.26. that h P (E ) = E E . un+ = un + hu0n ~ 1 un+1 = un + 2 h 1 u0n+ + 2 . 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. and hybrid methods. Q(E ) = E h E + + h] 2 h2 i (6. One concludes. A simple one-predictor. For this to hold for er . Predictor-corrector methods constructed to time-march linear or nonlinear ODE's are composed of sequences of linear multistep methods. it is obvious from Eq. Their use in solving ODE's is relatively easy to illustrate and understand.60) where the parameters and are arbitrary parameters to be determined.8 Predictor-Corrector Methods CHAPTER 6. ( + ) h . Their use in solving PDE's can be much more subtle and demands concepts5 which have no counterpart in the analysis of ODE's. following the example leading to Eq. one is led. 6.61 that + =1 =1 2 which provides two equations for three unknowns.106 6.44 that the same conditions also make it O(h3).6. There may be many families in the sequence. where measures of e ciency are discussed in the next two chapters. to the following observations. . It is easy to show. one-corrector example is given by un+ = un + h 0n ~ hu i un+1 = un + h u0n+ + u0n ~ (6.63) 5 Such as alternating direction. each of which is referred to as a family in the solution process. fractional-step. One can analyze this sequence by applying it to the representative equation and using the operational techniques outlined in Section 6. 1 . that the predictor-corrector sequence is a second-order accurate method for any .62) Considering only local accuracy. 6. by following the discussion in Section 6.

63 with = 1=2 is 1 un+1=2 = un + 2 hu0n ~ un+1 = un + hu0n+1=2 ~ and. . u0n. obtained from Eq. u0n.64) Some simple.9 Runge-Kutta Methods There is a special subset of predictor-corrector methods. u0n. is 1 hu ~ i un+1 = un + 2 h 3~0n .63 with = 1. 6. 6. The Gazdag method. nally.6.65) 2 The Burstein method. The Adams-BashforthMoulton sequence for k = 3 is i 1 h un+1 = un + 2 h 3u0n . If the order of the corrector is (k). The order of the combination is then equal to the order of the corrector.1 ~ i hh u un+1 = un + 12 5~0n+1 + 8u0n .67) (6.6 that produce just one -root for each -root such that ( h) corresponds explicit Runge-Kutta methods here. referred to as Runge-Kutta methods. which we discuss in Chapter 8. speci c.9.1 (6. (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. we will consider only single-step. second-order accurate methods are given below.1 ~ h i un+1 = un + 1 h u0n + u0n+1 ~ ~ (6.68) 6.66) (6. 6 Although implicit and multi-step Runge-Kutta methods exist. 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. we refer to these as ABM(k) methods. presented earlier in this chapter.

69 and 6.70) and this is much more di cult. u(tn) = 1 k1 + 2k2 + 3k3 + 4k4 (6.4. the method must further satisfy the constraint that er = O(hk+1) (6. 1. the choices for the time samplings.72) un+1 = un + 1hun 2 bn+ 3 ~n+ 1 + 4 hu0n+ 2 Appearing in Eqs.70. 6. The most widely publicized Runge-Kutta process is the one that leads to the fourth-order method. We present it below in some detail. but.71 and 6. First of all. a scheme entirely equivalent to 6.71 is b un+ = un + hu0n b un+ 1 = un + 1 hu0n + 1hu0n+ ~ b un+ 2 = un + 2 hu0n + 2hu0n+ + 2hu0n+ 1 ~ 0 + hu0 + hu0 (6. it is not su cient to guarantee k'th order accuracy for the solution of u0 = F (u t) or for the representative equation.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. To ensure k'th order accuracy. 6. .73) 2 = 2+ 2+ 2 .6.108 CHAPTER 6. Thus.69) = 1 + h + 1 2h2 + + 1 k hk 2 k! It is not particularly di cult to build this property into a method. They must satisfy the relations = 1 = 1+ 1 (6. we prefer to present it using predictor-corrector notation. are not arbitrary.71) However. TIME-MARCHING METHODS FOR ODE'S to the Taylor series expansion of e h out through the order of the method and then truncates. 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) . the principal (and only) -root is given by (6. as we pointed out in Section 6. Thus for a Runge-Kutta method of order k (up to 4th order). and 2 .

6. Using Eq. 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. 7 The present approach based on a linear inhomogeneous equation provides all of the necessary .74) These four relations guarantee that the ve terms in exactly match the rst 5 terms in the expansion of e h . but the most popular one is due to Runge. 6.7 There are eight equations in 6. To satisfy the condition that er = O(k5). As discussed in Section 6.69 and 6.74 and the rst equation in 6.76) 6 conditions for Runge-Kutta methods of up to third order.69 the four conditions 1+ 2 + 3 1+ = 1 2+ 3+ 4 = 1=2 3 1+ 4 2 ( 2 + 1 2 ) = 1=6 4 = 1=24 4 12 (1) (2) (3) (4) (6. the fourth condition in Eq. RUNGE-KUTTA METHODS 109 The algebra involved in nding algebraic equations for the remaining 10 parameters is not trivial.73 to eliminate the 's we nd from Eq.75 are all satis ed. which is based on a linear nonhomogenous representative ODE. 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 (6.74 and 6. Thus if the rst 3 equations in 6. 6.75 cannot be derived using the methodology presented here.75 which must be satis ed by the 10 unknowns. A more general derivation based on a nonlinear ODE can be found in several books. two parameters can be set arbitrarily. the resulting method would be third-order accurate.6. 6. Since the equations are overdetermined.9. but the equations follow directly from nding P (E ) and Q(E ) and then satisfying the conditions in Eqs.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. 6.4. Several choices for the parameters have been proposed.70.

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. storage requirements reduce the usefulness of Runge-Kutta methods of order higher than four for CFD applications.2. un = he h ae hn (6. 6.10 Implementation of Implicit Methods We have presented a wide variety of time-marching methods and shown how to derive their . respectively.66 and 6. This leads to various tradeo s which must be considered in selecting a method for a speci c application. 6.10. 6.77) using the implicit Euler method. RK methods of order k require k evaluations of the derivative function to advance the solution one time step. 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.1 Application to Systems of Equations Consider rst the numerical solution of our representative ODE u0 = u + ae t (6. relations. we will see that these methods can have widely di erent properties with respect to stability. These are covered in this Section. Following the steps outlined in Section 6.74 and the rst equation in 6. 6. For example. Higher-order RungeKutta methods can be developed. 6.75.67 are second-order Runge-Kutta methods. In the next chapter.72 by setting 4 = 0 and satisfying only Eqs. We shall discuss the consequences of this in Chapter 8. and third-order methods can be derived from Eqs. In any event.78) . we obtained (1 . but they require more derivative evaluations than their order.110 CHAPTER 6. h)un+1 . It is clear that for orders one through four. a fth-order method requires six evaluations to advance the solution one step. TIME-MARCHING METHODS FOR ODE'S 9 > > = > > Notice that this represents the simple sequence of conventional linear multistep methods referred to.

for biconvection.6.hf (t + h) u u or (6. hA).80) ~ where ~ and f are vectors and we still assume that A is not a function of ~ or t. requiring only an additional division. and hence its solution is inexpensive.1 ~ n . f (t) u u ~ ~ (I .79) This calculation does not seem particularly onerous in comparison with the application of an explicit method to this ODE. ~ n = . the cost per time step of an implicit method is larger than that of an explicit method. 6. hA)~ n+1 .83) (6. but in multidimensions the bandwidth can be very large. IMPLEMENTATION OF IMPLICIT METHODS Solving for un+1 gives 1 un+1 = 1 . The primary area of application of implicit methods is in the solution of sti ODE's.1 0 1)=2 x).84) un+1 = un + hF (un+1 tn+1) This is a nonlinear di erence equation.81 as a linear system of equations.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 (6.10. as we shall see in Chapter 8. Now let us apply the implicit Euler method to our generic system of equations given by ~ 0 = A~ .78 is (6. 6. Now u u the equivalent to Eq. As an example. h (un + he h ae hn ) 111 (6.. the system of equations which must be solved is tridiagonal (e. In general. A = .g.10.aBp (. For our one-dimensional examples.82) ~ ~ n+1 = (I .81) (6. hf (t + h)] u u The inverse is not actually performed.85) dt 2 . 6. consider the nonlinear ODE du + 1 u2 = 0 (6. but rather we solve Eq.

6. A Taylor series expansion about these reference quantities gives ! ! @F (u . Such an approach is described in the next Section.3 Local Linearization for Scalar Equations General Development Let us start the process of local linearization by considering Eq. the \initial guess" for this nonlinear problem can be quite close to the solution. which implies that a linearization approach may be quite successful. and Eq. tn) @t n @t2 n n 2 (6. An iterative method. TIME-MARCHING METHODS FOR ODE'S solved using implicit Euler time marching. t ) + 2 @u2 n n n @u@t n !n 1 2 + 2 @ F (t . 6. There are several di erent approaches one can take to solving this nonlinear di erence equation. since the \initial guess" is simply the solution at the previous time step.87) On the other hand. both (t . Designate the reference value by tn and the corresponding value of the dependent variable by un. In practice. can be used. such as Newton's method (see below). In order to implement the linearization.89) . tn)2 + @t2 n (6. the expansion of u(t) in terms of the independent variable t is ! ! @u + 1 (t . 6.86) n which requires a nontrivial method to solve for un+1. un)k are O(hk ).83. tn)k and (u . t ) + O(h2) F (u t) = Fn + @u n n @t n n (6. u ) + @F (t . u )2 + @ 2 F (u . we expand F (u t) about some reference point in time. u ) + @F (t . t ) F (u t) = F (un tn) + @u n n @t n n ! ! 1 @ 2 F (u . t )2 @ 2 u + u(t) = un + (t .10.112 CHAPTER 6. which gives 1 un+1 + h 2 u2 +1 = un (6. u )(t .88) If t is within h of tn .87 can be written ! ! @F (u .

83. @F u + @F (t .90) Implementation of the Trapezoidal Method As an example of how such an expansion can be used.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. consider the mechanics of applying the trapezoidal method for the time integration of Eq. 6. one nds un+1 = un + 1 h Fn + @F (un+1 . 6.6. it represents a second-order-accurate. There are. 6. un) + h @F + O(h2) + Fn 2 @u n @t n 2) +hO(h (6. and the trapezoidal time march. t ) + O(h2) n n dt @u n @u n n @t n (6. other second-order implicit timemarching methods that can be used. The trapezoidal method is given by 1 un+1 = un + 2 h Fn+1 + Fn] + hO(h2) (6. 2 h @F @u n 2 @t n which is now in the delta form which will be formally introduced in Section 12.10. In many uid mechanic applications the nonlinear function F is not an explicit function " ! ! # " !# ! .93) 1 . IMPLEMENTATION OF IMPLICIT METHODS 113 Notice that this is an expansion of the derivative of the function. With this we obtain the locally (in the neighborhood of tn) time-linear representation of Eq. A very useful reordering of the terms in Eq. 6. relative to the order of expansion of the function.83. locally-linear approximation to F (u t) that is valid in the vicinity of the reference station tn and the corresponding un = u(tn). of course.89 to evaluate Fn+1 = F (un+1 tn+1). namely ! ! ! ! du = @F u + F . Thus.6.91) where we write hO(h2) to emphasize that the method is second order accurate.92 results in the expression 1 un = hFn + 1 h2 @F (6. The important point to be made here is that local linearization updated at each time step has not reduced the order of accuracy of a second-order time-marching process. Using Eq. The use of local time linearization updated at the end of each time step. combine to make a second-order-accurate numerical integration process.

94) Notice that the RHS is extremely simple. rearrange terms. u 2. TIME-MARCHING METHODS FOR ODE'S of t.83. 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 . 6. store them in an array say R. but for our applications. we arrive at the form !# n un = hFn (6.95) if we introduce Eq.93 simpli es to the second-order accurate expression " 1 1 . 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. and save ~ n.94 is usually implemented as follows: ~ ~ 1. It is the product of h and the RHS of the basic equation evaluated at the previous time step.114 CHAPTER 6. Let this storage area be referred to as B . In this example. u 4. (Very seldom does one nd B .1 in carrying out this step). 3. Solve the coupled set of linear equations B ~n = R u ~ for ~ n. 6. In such cases the partial derivative of F (u) with respect to t is zero and Eq. and remove the explicit dependence on time. the basic equation was the simple scalar equation 6.96) . 6.90 into this method. it is generally the space-di erenced form of the steady-state equation of some uid ow problem. A numerical time-marching procedure using Eq. Find ~ n+1 by adding ~ n to ~ n. h @F @u (6. 2 h @F @u !# n un = hFn (6. Solve for the elements of hFn. 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.

" Fn (6. 6. and is a scaling factor. let us consider an example involving some simple boundary-layer equations.94 and 6. we mean that the error after a given iteration is proportional to the square of the error at the previous iteration. Use of a nite value of h in Eq. where the error is the di erence between the current solution and the converged solution.96 leads to linear convergence. 6. Our task is to apply the implicit trapezoidal method to the equations d3f + f d2f + dt3 dt2 0 !21 @1 . 6.1 n (6.96 obtained by dividing both sides by h and setting 1=h = 0. Omission of this factor degrades the method in time accuracy by one order of h.97) un+1 = un . We shall see later that this method is an excellent choice for steady problems.96.99) Here f represents a dimensionless stream function.88 (remember the dependence on t has been eliminated for this case). given by Eq. The fact that it has quadratic convergence is veri ed by a glance at Eqs. .10. We choose the Falkner-Skan equations from classical boundary-layer theory.10.. i.6.4 Local Linearization for Coupled Sets of Nonlinear Equations In order to present this concept. 6. df A = 0 dt (6. the error at a given iteration is some multiple of the error at the previous iteration. By quadratic convergence.87 and 6. 6. Newton's Method Consider the limit h ! 1 of Eq. @F @u ! n un = Fn @F @u ! #. There results or . 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. Quadratic convergence is thus a very powerful property. 6.e.98) This is the well-known Newton method for nding the roots of a nonlinear equation F (u) = 0. The reader should ponder the meaning of letting h ! 1 for the trapezoidal method.94.

and tensor products for the terms of higher order. and de ning F n as F (~ n).u1 u3 . TIME-MARCHING METHODS FOR ODE'S u1 = d f dt2 2 First of all we reduce Eq.116 CHAPTER 6. 6.102 by the following process d~ = F (~ ) u ~ u dt (6.8 Let us refer to this matrix as A. Omitting the explicit dependency ~ ~ u on the independent variable t. 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 (6.99 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 2.102) A = (aij ) = @Fi =@uj For the general case involving a third order matrix this is (6.104) 8 Recall that we derived the Jacobian matrices for the two-dimensional Euler equations in Section 9 The Taylor series expansion of a vector contains a vector for the rst term.87.90. The required extension requires the evaluation of a matrix. 6. 6.100) (6. 1 . except that this time we are faced with a nonlinear vector function. @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. called the Jacobian matrix. It is derived from Eq. rather than a simple nonlinear scalar function. a matrix times a .101) Now we seek to make the same local expansion that derived Eq.103) 2 6 6 6 6 A=6 6 6 6 6 4 ~ u The expansion of F (~ ) about some reference state ~ n can be expressed in a way u similar to the scalar expansion given by eq 6.2 vector for the second term.

106) 6. the terms in the Jacobian matrix are updated as the solution proceeds depends. Any rst. ~ n + O(h2) u u (6. we nd the Jacobian matrix to be 2 .11 Problems 1.93. What is u25 ? (Let the rst term given above be u0.1. Without any iterating within a step advance. Find an expression for the nth term in the Fibonacci series. A ~ +O(h2) u ~ n n un nu dt | {z } (6. u3 2 u2 . (u1u3)n .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 . PROBLEMS 117 ~ u ~ F (~ ) = F n + An ~ . of course. explicit or implicit. Returning to our simple boundary-layer example.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. and the solution is now advanced one step. h(u2)n h (u1)n 72 ( u1)n 3 2 . 6. d~ = A ~ + F .u1 3 0 0 7 A=6 1 (6.) . The number of times. on the nature of the problem.102 as \constant" which is a locally-linear. which is given by Eq.or second-order time-marching method.11. and the manner in which. u2)n 3 2 2 6 2 7 6 .88.6.101. Re-evaluate u u u the elements using ~ n+1 and continue. u the solution will be second-order-accurate in time.2 1 We nd ~ n+1 from ~ n + ~ n. could be used to integrate the equations without loss in accuracy with respect to order.105) where t . 6. (1 . tn and the argument for O(h2) is the same as in the derivation of Eq. which is given by 1 1 2 3 5 8 : : : Note that the series can be expressed as the solution to a di erence equation of the form un+1 = un + un. 6. Using this we can write the local linearization of Eq. second-order-accurate approximation to a set of coupled nonlinear ordinary di erential equations that is valid for t tn + h. Thus for the Falkner-Skan equations the trapezoidal method results in 2 3 1 + h (u3)n .

(b) Derive the . (c) Find er . 5.118 CHAPTER 6. TIME-MARCHING METHODS FOR ODE'S 1 2.1) (a) Write the O E for the representative equation. relation. The 2nd-order backward method is given by i 1h un+1 = 3 4un .1 + 2hu0n+1 (a) Write the O E for the representative equation. 4.relation. Identify the polynomials P (E ) and Q(E ). 6. Consider the time-marching scheme given by un+1 = un. un.65) to the representative equation. 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 . (c) Find er and the rst two nonvanishing terms in a Taylor series expansion of the spurious root.relation.1 + 23h (u0n+1 + u0n + u0n. (b) Derive the . Find the . Find the di erence equation which results from applying the Gazdag predictorcorrector method (Eq. The trapezoidal method un+1 = un + 2 h(u0n+1 + u0n) is used to solve the representative ODE. 6. Identify the polynomials P (E ) and Q(E ). Solve for the -roots and identify them as principal or spurious. (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. (d) Perform a -root trace relative to the unit circle for both di usion and convection.

11. of 0. use u(x 0) = e. Using the computer program written for problem 7. ah= x.1 for the other methods. For the explicit Euler and AB2 methods.6. trapezoidal. use a Courant number. and 4th-order Runge-Kutta methods. Find er and er . including the homogeneous and particular solutions. 10. 8.0:5)= ]2 with = 0:08. Let u(0 t) = sin !t with ! = 10 . Repeat problem 7 using 4th-order (noncompact) di erences in space. implicit Euler. 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. and 400 nodes. On a log-log scale. Using the computer program written for problem 8. Find the solution to the di erence equation. 2nd-order Adams-Bashforth (AB2). Use the explicit Euler. Run until a periodic steady state is reached which is independent of the initial condition and plot your solution . Find the . 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. uexact)2 u j t M j =1 where M is the number of grid nodes and uexact is the exact solution. plot the error given by v uX u M (uj . Find the global order of accuracy from the plot. Plot the solutions obtained at t = 1 compared to the exact solution (which is identical to the initial condition). 200. Use 2nd-order centered di erences in space and a grid of 50 points. Use only 4th-order Runge-Kutta time marching at a Courant number of unity. What order is the homogeneous solution? What order is the particular solution? Find the particular solution if the forcing term is xed.relation. 11. repeat problem 9 using 4th-order (noncompact) di erences in space. Show solutions at t = 1 and t = 10 compared to the exact solution.0:5 (x. use a Courant number of unity. 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. 7. PROBLEMS 119 Find the di erence equation which results from applying this method to the representative equation. For the initial condition.

12. 2nd. Using the approach described in Section 6. j . Find the global order of accuracy. the number of grid nodes. . 6. Repeat problem 11 using 4th-order (noncompact) centered di erences. 200. j . erp. 1. 13.2. Explain your results. and j ) and at the second last node. 2. j . use a 3rd-order backwardbiased operator (using nodes j . without actually deriving the methods. Use grids with 100. Also plot the phase speed error obtained using exact integration in time.120 CHAPTER 6. i. as described in problem 9. TIME-MARCHING METHODS FOR ODE'S compared with the exact solution.69) together with 4th-order Runge-Kutta time marching. j . and 4th-order Runge-Kutta time-marching at a Courant number of unity. 3. At the last grid node. nd the phase speed error. and the amplitude error. Use 2nd-order centered di erences in space with a 1st-order backward di erence at the out ow boundary (as in Eq. 3. 1. 3rd. and 400 nodes and plot the error vs. derive and use a 3rd-order backward operator (using nodes j .69. Note that the required -roots for the various Runge-Kutta methods can be deduced from Eq. that obtained using the spatial discretization alone.. 2.6. era . Use a third-order forward-biased operator at the in ow boundary (as in Eq.e. for the combination of second-order centered differences and 1st. and j +1 see problem 1 in Chapter 3).67). 3.

͏ @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 .

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

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

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

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

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

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

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

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

0 2.0 R( λh) ‡ ü „ .0 I(λh) b) Adams Type 1.0 2.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 Unstable Stable Outside 2.0 4.0 3.0 1.0 ´ ­ ® µ ´ · ´ ¯ ² ® ª° ³ Á ¹ · µ ° ³° ¶ ² µ ´ § ¦ ¯ ® µ º ² ¼ ¢„s|D|‹AQ€R “¸!TR€Q1|À(¿Ds‹Q€’± I(λh) Stable Outside Unstable R(λ h) 1.

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

Examples are given showing how time-marching methods can be compared in a given context.1. However. Any de nition of sti ness requires a coupled system with at least two eigenvalues. and the decision to use some numerical time-marching or iterative method to solve it. which is de ned in the next section. since this part of the ODE has no e ect on the numerical stability of 149 . we assume that this scale would be adequately resolved by the chosen time-marching method. An important concept underlying much of this discussion is sti ness.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. De nitions given in the literature are not unique. and.1 Sti ness De nition for ODE's 8. 8. but fortunately we now have the background material to construct a de nition which is entirely su cient for our purposes. 7.1 Relation to -Eigenvalues The introduction of the concept referred to as \sti ness" comes about from the numerical analysis of mathematical models constructed to simulate dynamic phenomena containing widely di erent time scales.1. 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. In the following discussion we concentrate on the transient part of the solution. The di erence between the dynamic scales in physical space is represented by the di erence in the magnitude of the eigenvalues in eigenspace. The forcing function may also be time varying in which case it would also have a time scale.

the time integration is an approximation in eigenspace that is di erent for every eigenvector ~ m .1. 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. This is given by Eq.27 and its solution using a one-root. 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.1: Stable and accurate regions for the explicit Euler method. CHOICE OF TIME-MARCHING METHODS Stable Region λh I(λ h) 1 R(λ h) Accurate Region Figure 8. timemarching method is represented by Eq. 8. if at all. although these eigenvectors must remain coupled into the system to maintain a high accuracy of the spatial resolution. The situation is represented in the complex h plane in Fig.28. the homogeneous part. 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. For a given time step. 6. . Consider now the form of the exact solution of a system of ODE's with a complete eigensystem.150 CHAPTER 8. 6. we exclude the forcing function from further discussion in this section.

Unfortunately. 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. thus j 1j j 2j j Mj (8. . numerical stability requirements are dictated by the parasitic ones. called the parasitic eigenvalues (no time accuracy whatsoever is required for the eigenvectors associated with these.1. and the other. although time accuracy requirements are dictated by the driving eigenvalues.8.1. and imaginary eigenvalues. 8. First we order the eigenvalues by their magnitudes. p+1 ! M ]. It is important to notice that these de nitions make no distinction between real. complex. STIFFNESS DEFINITION FOR ODE'S 151 8. 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).1.1) Then we write p M Transient = X c e m t ~ + X c e m t ~ (8.2 Driving and Parasitic Eigenvalues For the above reason it is convenient to subdivide the transient solution into two parts. but their presence must not contaminate the accuracy of the complete solution).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. Rephrased.2) xm xm m m Solution m=1 {z m=p+1 | } {z } | Driving Parasitic This concept is crucial to our discussion.

The simplest example to discuss relates to the model di usion equation. 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. 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.1. i. the sti ness is determined by the ratio M = 1 . and are greatly a ected by the resolution of a boundary layer since it is a di usion process.e. Furthermore. In order to demonstrate this. 2 and the ratio is . Even for a mesh of this moderate size the problem is already approaching the category of strongly sti . near an airfoil leading or trailing edge. Examples of where this clustering might occur are at a shock wave. and in a boundary layer. 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. For a mesh size M = 40. As a result it is quite common to cluster mesh points in certain regions of space and spread them out otherwise. this ratio is about 680. 4x2 x 2=. we are interested only in the converged steady-state solution.3. 4x2 sin2 2 = . let us examine the eigensystems of the model problems given in Section 4.152 8.. but much less so than for di usion-dominated problems.2. the sti er our equations tend to become. Typical CFD problems without chemistry vary between the mildly and strongly sti categories. in di usion problems this sti ness is proportional to the reciprocal of the space mesh size squared. One quickly nds that this grid clustering can strongly a ect the eigensystem of the resulting A matrix. A simple calculation shows that 4 2 1 = . Our brief analysis has been limited to equispaced x .2 Relation of Sti ness to Space Mesh Size CHAPTER 8. 2 sin 2(M + 1) x M ! . negative numbers that automatically obey the ordering given in Eq. 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. Under these conditions. In this case the eigenvalues are all real. Consider the case when all of the eigenvalues are parasitic. 8.

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

The time-march method is explicit.1. 6. so that CHAPTER 8. ln(0:25) with u(0) = 1. let us set the additional requirement The error in u at the end of the event. a = 0. and must use a constant time step size.4. the allowable error constraint means that the global error in the amplitude. Three methods are compared | explicit Euler.2 An Example Involving Di usion Let the event be the numerical solution of from t = 0 to T = . must simulate an entire event which takes a total time T . this makes the exact value of u at the end of the event equal to 0.154 8.u dt (8. Eq. In some contexts.t. this can be an important consideration.. h. see Eq. must be < 0:5%. Implications of computer storage capacity and access time are ignored. CHOICE OF TIME-MARCHING METHODS 8. the global error. 2. AB2. The calculation is to be time-accurate. i.3 is obtained from our representative ODE with = . Fev . and RK4. To the constraints imposed above. 3.1 Imposed Constraints T = Nh where N is the total number of time steps. The follow assumptions bound the considerations made in this Section: 1. Since the exact solution is u(t) = u(0)e.4 Comparing the E ciency of Explicit Methods As mentioned above.e.e. the e ciency of methods can be compared only if one accepts a set of limiting constraints within which the comparisons are carried out. must have the property: du = . u(T ) = 0:25.. We judge the most e cient method as the one that satis es these conditions and has the fewest number of evaluations. First of all.25. i.48. 8.3) Er < 0:005 eT . 8.4.

1 were computed using a simple iterative procedure. 8. and the fourth-order Runge-Kutta method is the best of all.4. the method with the highest local accuracy is the most e cient on the basis of the expense in evaluating Fev .1: Comparison of time-marching methods for a simple dissipation problem. ( 1 (ln(:25)=N ))N =:25 < 0:005 where 1 is found from the characteristic polynomials given in Table 7. In this example we see that.1. it follows that 155 1 . Under these conditions a method is judged as best when it has the highest global accuracy for resolving eigenvectors with imaginary eigenvalues.4. Thus the second-order Adams-Bashforth method is much better than the rstorder Euler method. leading to the so-called incomplete .1. 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. In this numerical experiment the function evaluations contribute overwhelmingly to the CPU time. The results shown in Table 8.8. we add the following: ~u The number of evaluations of F (~ t) is xed. for a given global accuracy. in addition to the constraints given in Section 8. ln(0:25)=N . and the number of these evaluations must be kept to an absolute minimum because of the magnitude of the problem. In this case. since h = T=N = . The main purpose of this exercise is to show the (usually) great superiority of second-order over rst-order time-marching methods. The above constraint has led to the invention of schemes that omit the function evaluation in the corrector step of a predictor-corrector combination. a complete event must be established in order to obtain meaningful statistical samples which are the essence of the solution.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. COMPARING THE EFFICIENCY OF EXPLICIT METHODS Then.4. On the other hand.

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 this case. The second and fourth order RK methods are 2. and the number of evaluations. The . respectively. For a 4-evaluation method the time interval must be h = 4h1 . Let h1 be the time interval advanced in one step of a one-evaluation method. one evaluation being used for each step. CHOICE OF TIME-MARCHING METHODS predictor-corrector methods. the derivative of the fundamental family is never found so there is only one evaluation required to complete each cycle. j 1(ik!h1 )jK=k (8. However. notice also that as k increases. An example is the method of Gazdag. . the time step must be twice that of a one{evaluation method so h = 2h1. are the same. the time span required for one application of the method increases. 6.39. Let K represent the total number of allowable Fev . that more e cient methods will result from the omission of the second function evaluation. given in Section 6. N . leapfrog. K . the power to which 1 is raised to arrive at the nal destination decreases see the Figure below. of course. and AB2 schemes are all 1-evaluation methods. the global amplitude and phase error for kevaluation methods applied to systems with pure imaginary -roots can be written1 Era = 1 . However. Notice that as k increases. The Gazdag. 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. Basically this is composed of an AB2 predictor and a trapezoidal corrector. etc. For a 2-evaluation method N = K=2 since two evaluations are used for each step. after K evaluations. This is the key to the true comparison of time-march methods for this type of problem.1. For a 1evaluation method the total number of time steps. in order to arrive at the same time T after K evaluations.and 4-evaluation methods.156 CHAPTER 8. The presumption is. In general.8. However. so that for these methods h = h1 . h.38 and 6.4) 1 See Eqs.relation for the method is shown as entry 10 in Table 7.

For example.2. Notice that to nd global error the Gazdag root must be raised to the power of 50 while the RK4 root is raised only to the power of 50/4.2:4 :45 :12 !h1 = :2 50 . by some rather simple calculations2 made using Eqs. Gazdag. K leapfrog AB2 Gazdag RK4 !h1 = :1 100 1:0 1:003 :995 :999 !h1 = :2 50 1:0 1:022 :962 :979 a. we are making our comparisons on the basis of global error for a xed number of evaluations. 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.2: Comparison of global amplitude and phase errors for four methods. in terms of phase error (for which it was designed) it is superior to the other two methods shown. Table 8. On the basis of global error the Gazdag method is not superior to RK4 in either amplitude or phase. and RK4. Phase error in degrees. 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. and the fact that ! = 1.3:8 . we nd.5) 1 (ik!h1 )]imaginary 1 (ik!h1 )]real # Consider a convection-dominated event for which the function evaluation is very time consuming. 2 .1 k " 157 (8. The 1 root for the Gazdag method can be found using a numerical root nding routine to trace the three roots in the -plane.0. K tan. exact = 1.3e. Amplitude.:96 . 7. the results shown in Table 8. The event must proceed to the time t = T = 10. The rst three of these are one-evaluation methods and the last one is a four-evaluation method.5. 8.8. However. see Fig. COMPARING THE EFFICIENCY OF EXPLICIT METHODS Er! = !T . K leapfrog AB2 Gazdag RK4 !h1 = :1 100 . First of all we see that for a one-evaluation method h1 = T=K . We consider two maximum evaluation limits K = 50 and K = 100 and choose from four possible methods.4. Using this. AB2. We idealize to the case where = i! and set ! equal to one. leapfrog.9:8 1:5 1:5 b.4 and 8. although.

100t quickly becomes very small and can be neglected in the expressions when time becomes large. We rst run 66 time steps with h = 0:001 in order to resolve the 1 term. Let us choose the simple explicit Euler method for the time march. If h is chosen so that all h in the ODE eigensystem fall inside this circle the integration will be numerically stable. 8. With this time step the .5.001 is negligible. For a speci c example. and a relatively slowly decaying function in the other.6. We have de ned these h as parasitic eigenvalues. If some h fall outside the small circle but stay within the stable region.1. This de nes a time step limit based on accuracy considerations of h = 0:001 for 1 and h = 0:1 for 2 . which is determined from 1. 8.1. In this case the trace forms a circle of unit radius centered at (. 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. but stable.100 and 2 = .1 0) as shown in Fig. Stability boundaries for some explicit methods are shown in Figs.5 and 7. If uncoupled and evaluated in wave space. The coupled equations in real space are represented by u1(n) = c1 (1 . 100h)nx21 + c2(1 . A good example of the concept is produced by studying the Euler method applied to the representative equation. The transient solution is un = (1 + h)n and the trace of the complex value of h which makes j1 + hj = 1 gives the whole story.158 CHAPTER 8. 100h)nx11 + c2(1 . consider the mildly sti system composed of a coupled two-equation set having the two eigenvalues 1 = .5 Coping With Sti ness 8. The time step limit based on stability. Numerical solutions.6) We will assume that our accuracy requirements are such that su cient accuracy is obtained as long as j hj 0:1. We will also assume that c1 = c2 = 1 and that an amplitude less than 0. these h are sti . h)nx22 + (PS )2 (8. depend upon ( m h)]n and no j m j can exceed one for any m in the coupled system or else the process is numerically unstable. Also shown by the small circle centered at the origin is the region of Taylor series accuracy. of course. Numerical evaluation is altogether di erent. Analytical evaluation of the time histories poses no problem since e. 7. is h = 0:02.1 Explicit Methods The ability of a numerical method to cope with sti ness can be illustrated quite nicely in the complex h plane. the time histories of the two solutions would appear as a rapidly decaying function in one case. h)nx12 + (PS )1 u2(n) = c1 (1 .

In fact. and.e. with 69 steps required to reduce the amplitude of the second term to below 0. we would like to change the step size to h = 0:1 and continue. However. After 70 time steps the 1 term has amplitude less than 0...001 and that of the 2 term (i. since (0:666 )n < 1. (1 . COPING WITH STIFFNESS 2 159 term is resolved exceedingly well. It is true that in the nal 69 steps one -root is 1 . h)n) is 0.100t . 50h nx + c u1(n) = c1 1 + 50h 11 2 ! 1 . 0:5h nx + (PS ) 1 1 + 0:5h 12 ! 1 .. 100h)n) is less than 0. the maximum step size that can be taken in order to maintain stability. the amplitude of the 1 term (i. Hence the 1 term can now be considered negligible. Now let us re-examine the problem that produced Eq. To drive the (1 . 8.4b. 50(0:1)]= 1 + 50(0:1)] = 0:666 . far outweighing the initial decrease obtained with the smaller time step.001). and this has no physical meaning whatsoever. It follows that the nal numerical solution to the ODE is now represented in real space by 1 .e. Its behavior in the h plane is shown in Fig. We choose the trapezoidal method. we need to use a step size of about h = 0:001. its in uence on the coupled solution is negligible at the end of the rst 70 steps. Thus the total simulation requires 405 time steps. h)n term to zero (i.5.9361. (1 . its in uence in the remaining 70 .6 but this time using an unconditionally stable implicit method for the time march. 7. We would then have a well resolved answer to the problem throughout the entire relevant time interval. 50h nx + c u2(n) = c1 1 + 50h 21 2 ! 8. 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.001.001 and can be neglected.e.7) In order to resolve the initial transient of the term e. (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. 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. but that is not the point of this exercise).5. In both intervals the desired solution is second-order accurate and well resolved. 100h)n. 0:5h nx + (PS ) 2 1 + 0:5h 22 ! (8. about 339 time steps have to be computed in order to drive e.t to below 0. which in just 10 steps at h = 0:1 ampli es those terms by 109.2 Implicit Methods 1 . with h = 0:02. However. After 66 steps. this is not possible because of the coupled presence of (1 . below 0. Since this is also a oneroot method.001.8. we simply replace the Euler with the trapezoidal one and analyze the result.

although this root is one of the principal roots in the system.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. CHOICE OF TIME-MARCHING METHODS steps is even less.5. It is important to retain a proper perspective on a problem represented by the above example. Therefore. Our sole goal is to eliminate it as quickly as possible. it should be clear that as the degree of sti ness of the problem increases. we need an introduction to the theory of relaxation before it can be presented.8) When solving a steady problem. The reader can repeat the above example with 1 = . the conditionally stable Euler method required 405 time steps. For preliminary investigations it is often the best method to use for mildly-sti di usion dominated problems. the Euler method is extremely easy to program and requires very little arithmetic per step. about three times as many. However. is recommended. 2 = . the advantage begins to tilt towards implicit methods. For re ned investigations of such problems an explicit method of second order or higher.. which is in the strongly-sti category. There is yet another technique for coping with certain sti systems in uid dynamic applications. In the example. its behavior for t > 0:07 is identical to that of a stable spurious root. Actually. if there are more spurious roots (8. These explicit methods can be considered as e ective mildly sti stable methods. such as Adams-Bashforth or Runge-Kutta methods. This is known as the multigrid method. as the reduced number of time steps begins to outweigh the increased cost per time step. we have no interest whatsoever in the transient portion of the solution. 8. .10 000. It has enjoyed remarkable success in many practical problems however.3 A Perspective 8. as compared to about 139 for the trapezoidal method. The total simulation requires 139 time steps.160 CHAPTER 8.1. It is clear that an unconditionally stable method can always be called upon to solve sti problems with a minimum number of time steps.

8. a nite time step may be required until the solution is somewhat close to the steady solution.and 4th-order Runge-Kutta methods. When we seek only the steady solution. the di erence being that the order of accuracy is irrelevant.3) using 2nd. 3. Eq. Hence the explicit Euler method is a candidate for steady di usion dominated problems.7 Problems 1.2) and periodic convection (Section 8.4. because of their stability properties. Referring to Fig.and 3rd-order Runge-Kutta methods. 8. the implicit Euler method is the obvious choice for steady problems.1. when using the implicit Euler method with local time linearization. 6.4.7. all of the eigenvalues can be considered to be parasitic. Consider the di usion equation (with = 1) discretized using 2nd-order central di erences on a grid with 10 (interior) points. 6.3) using the 3rd.4.96. compare and contrast them with the 3rd.2) and periodic convection (Section 8. What is the maximum allowable time step if all but the rst eigenvector are considered parasitic? . Among implicit methods. If we use the explicit Euler timemarching method what is the maximum allowable time step if all but the rst two eigenvectors are considered parasitic? Assume that su cient accuracy is obtained as long as j hj 0:1.98. Find and plot the eigenvalues and the corresponding modi ed wavenumbers.4. none of the eigenvalues are required to fall in the accurate region of the time-marching method. Therefore the time step can be chosen to eliminate the transient as quickly as possible with no regard for time accuracy. Repeat the time-march comparisons for di usion (Section 8. Eq. 2. 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.8.and 4th-order Adams-Bashforth methods. However. and the fourth-order Runge-Kutta method is a candidate for steady convection dominated problems. For example. which leads to Newton's method. Repeat the time-march comparisons for di usion (Section 8. one would like to take the limit h ! 1.

162 CHAPTER 8. CHOICE OF TIME-MARCHING METHODS .

3) To solve this system using a time-marching method.4 has a stable steady solution only if A meets the above condition.A) have real parts lying in the left half-plane. 9. Although the solution ~ = A. the ODE given by x b Eq.1. analyzing. which satis es the following coupled system of nonlinear algebraic equations: ~u F (~ ) = 0 (9. Alternatively. ~ exists as long as A is nonsingular.2) In the latter case. ~ x x b (9. they can be used to solve for the steady solution of Eq. and choosing time-marching methods.4) dt and the system is integrated in time until the transient has decayed to a su ciently low level. 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. the unsteady equations are integrated until the solution converges to a steady solution. The same approach permits a time-marching method to be used to solve a linear system of algebraic equations in the form A~ = ~ x b (9.1) dt which arise after spatial discretization of a PDE. a time derivative is introduced as follows d~ = A~ .Chapter 9 RELAXATION METHODS In the past three chapters. Following a time-dependent path to steady state is possible only if all of the eigenvalues of the matrix A (or . 1 163 . 9. we developed a methodology for designing.

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

9 has eigenvalues whose imaginary parts are much larger than their real parts.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> (9.6).2 4 1 (9.1 . For example.1 32 0 7 6 .1 4 1 0 . It can rst be conditioned so that the modulus of each element is 1.1 .1 0 76 176 .2 3 2 u 7 6 .1 1 1 ~= 0 u 2 x >6 x 6 >6 .2 0 6 6 1 0 .2 1 1 6 0 0 6 = 6 1 0 .1 0 6 6 .10) which scales each row.1 1 1 76 0 76 76 . We then further condition with multiplication by the negative transpose.1 54 3 7 7 7 17 7 17 5 1 0 . consider a spatial discretization of the linear convection equation using centered di erences with a Dirichlet condition on the left side and no constraint on the right side.1 1 1 6 0 6 6 . this leads to the approximation 82 > 0 1 >6 > 0 6 1 <6 .1 0 >4 : . FORMULATION OF THE MODEL PROBLEM 165 conditioning matrix C must be chosen such that this requirement is satis ed.9. 9. 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 (9.11) 1 .1.1 3 7 7 7 7 7 17 5 1 2 0 6 . Using a rst-order backward di erence on the right side (as in Section 3. A choice of C which ensures this condition is the negative transpose of Ab.9) The matrix in Eq.1 .2 .AT A = 2 1 1 2 0 6 . The result is A = .

2 .2 1 =6 6 6 1 .2 0 1 76 7 6 1 0 . the conditioned matrix . RELAXATION METHODS 1 1 If we de ne a permutation matrix P and carry out the process P T ..e.166 1 1 CHAPTER 9. 2 1 6 . 2 We use a symmetric negative de nite matrix to simplify certain aspects of our analysis.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.12) 9.2 4 3 7 7 7 7 7 17 5 (9. We will consider the preconditioned system of equations having the form A~ . and the classical relaxation methods can be applied.1 which has all negative real eigenvalues. Relaxation methods are applicable to more general matrices. as given in Appendix B.AT as a preconditioner we simply wish to show that a broad b range of matrices can be preconditioned into a form suitable for our analysis.1.13) where A is symmetric negative de nite.2 1 7 6 54 0 1 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 .1 0 1 76 76 0 7 6 0 . symmetric with negative real eigenvalues). We do not necessarily recommend the use of . In the remainder of this chapter.1 ) just rearranges the rows and columns of a matrix. 1 . Thus even when the basic matrix Ab has nearly imaginary eigenvalues. as de ned in Appendix A. The classical methods will usually converge if Ab is diagonally dominant.2 0 1 7 6 76 176 76 76 054 1 0 . we will thoroughly investigate some simple equations which model these structures.AT Ab b is nevertheless symmetric negative de nite (i.2 1 6 1 6 1 .2 32 32 0 1 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 1 0 0 17 07 7 07 7 07 5 0 3 1 .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 . ~ = 0 f (9.

9. Note that the solution of Eq.15) @ u = g(x) (9.2 1) x ~ fb = ~ . ~ = A. is guaranteed to exist because A is nonsingular. 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 .19) ~ where ~ = x fb. In the notation of f Eqs.2 1)~ + (bc) . ~ u u ~ g (9. we nd d~ = 1 B (1 . then A has the form A=B 1~ 1 b 2 2 2 . A = CAb and ~ = C~ b f f (9.7. ~ .18) Choosing C = x I . 9. we obtain B (1 . 9. Introducing the threepoint central di erencing scheme for the second derivative with Dirichlet boundary conditions.1. (bc) g ~ (9.17) dt x ~ where (bc) contains the boundary conditions and ~ contains the values of the source g term at the grid nodes.14) The above was written to treat the general case.16) @x which is the one-dimensional form of the Poisson equation. g(x) @t @x This has the steady-state solution 2 2 2 2 (9. FORMULATION OF THE MODEL PROBLEM 1 167 accurate when we later deal with ODE formulations.5 and 9.2 1)~ = ~ f (9. In this case 2 Ab = 1 B (1 .13. If we consider a Dirichlet boundary condition on the left side and f either a Dirichlet or a Neumann condition on the right side.

A]~ n . that it is directly applicable to two.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 .22) 9. A tremendous amount of insight to the basic features of relaxation is gained by an appropriate study of the one-dimensional case. The converged solution is designated ~ 1 so that ~ 1 = A.1 is easily obtained from the matrix resulting from the Neumann boundary condition given in Eq.21) where H is some nonsingular matrix which depends upon the iterative method. the Residual.2 Classical Relaxation 9. ~ = G~ n . The error at the nth iteration is de ned as ~ n ~ n .2 .21 for ~ n gives ~ n = I + H .20) Note that s = . and the Error +1 +1 1 1 1 Solving Eq. 3.2.2 or . A (9. ~ n = A~ n .and three-dimensional problems. RELAXATION METHODS ~ = .2 The Converged Solution. ~ e f (9. The iteration count is designated by the subscript n or the superscript (n). H . ~ 1 = ~ n . 9.23) where G I + H.2 s]T s = .25) 1 1 .1 (9. or at least easier to solve than the original system.2 b . A.2. however. ~ f (9. H . ~ f 1 (9. 9. We attempt to do this in such a way. and much of the remaining material is devoted to this case.168 CHAPTER 9.24) Hence it is clear that H should lead to a system of equations which is easy to solve.24 using a diagonal conditioning matrix. The matrix H is independent of n for stationary methods and is a function of n for nonstationary ones. ~ f f (9.

26) Multiply Eq.2 1)~ = ~ f (9.1.27) Finally. This operator comes about by choosing the value of jn such that together with the old values of j.31) j j j 2 j.29) as given in Section 9. In all of the above.29 is satis ed. and its eigenvalues. Hence the j th row of Eq. 9. 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. The Point-Jacobi method is expressed in point operator form for the one-dimensional case as n = 1 h n + n .3 The Classical Methods Point Operator Schemes in One Dimension Let us consider three classical relaxation procedures for our model equation B (1 .29 is satis ed using the new values of j and j .25 by A from the left. it is not di cult to show that ~ n = G~ n e e +1 (9. 9. we have considered only what are usually referred to as stationary processes in which H is constant throughout the iterations. f i (9. The residual at the nth iteration is de ned as ~ n A~ n . which we designate as m .2. 9. determine the convergence rate of a method. CLASSICAL RELAXATION 169 where ~ 1 was de ned in Eq.30) j j j 2 j. and the old value of j . The Gauss-Seidel method is n = 1h n + n . . 9.26. 9.2. and use the de nition in Eq.2.f i (9. the j th row of Eq. Nonstationary processes in which H (and possibly C ) is varied at each iteration are discussed in Section 9.5. ~ r f (9. There results the relation between the error and the residual A~ n . 9. The method of successive overrelaxation (SOR) ( +1) ( ) 1 ( ) +1 ( +1) 1 +1 ( +1) ( +1) 1 ( ) +1 1 1 +1 . and j . ~ n = 0 e r (9.28) Consequently.9.22.

about which we have already learned a great deal.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. C A ~ . 9. 9. fj 2 . ~ f (9. ! f (9.170 CHAPTER 9.35) dt This is equivalent to d~ = H . It is usually expressed in two steps as h i ~j = 1 jn + jn .32) j ( +1) 1 ( ) +1 ( +1) ( ) ( ) where ! generally lies between 1 and 2.1 The Ordinary Di erential Equation Formulation .(L + D). and the portion above the diagonal. ! ) n + ! n . the portion of the matrix below the diagonal.21 leads to an interpretation of relaxation methods in terms of solution techniques for coupled rst-order ODE's. such that A=L+D+U (9. 9. which is also easy to solve. L.13 into its diagonal.D.33) j j 2 j.3 The ODE Approach to Classical Relaxation The particular type of delta form given by Eq. One can easily see that Eq. h i n = jn + ! ~j . jn (9.3. but it can also be written in the single line n = ! n + (1 . which certainly meets the criterion that it is easy to solve.34) Then the point-Jacobi method is obtained with H = . D. 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. ~ = H . A~ . being lower triangular.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~ . The Gauss-Seidel method is obtained with H = . ~ ] fb f (9. U . then perhaps it would be better to move further in this direction. 9.

H . the preconditioning matrix in 9.35. 9.36 are independent of t. 9. The generalization of this in our approach is to make h. For this method m = 1 + m h. A is nonsingular). A depends on neither ~ nor t. even though no true time accuracy is involved. and the eigenvectors of H written as ~ = c e 1 t~ + + cM e M t~ M +~ 1 x} (9. so that we must devise a way to remove them all from the general solution on an equal basis. These are xed by the initial guess. Throughout the remaining discussion we will refer to the independent variable t as \time".13. that is. Suppose the explicit Euler method is used for the time integration. 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. It is clear that.28) ~ n = c ~ (1 + h)n + |x 1 1 1 + cm~ m (1 + mh)n + x {z error + cM ~ M (1 + x n M h) } +~ 1 (9.39) The initial amplitudes of the eigenvectors are given by the magnitudes of the cm . 6. ~ f 1 (9. A has been chosen (that is. H and C in Eq. an iteration process has been decided upon). 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. they are not changed throughout the iteration process. the solution can be of t.37) | x {z error where what is referred to in time-accurate analysis as the transient solution. A have negative real parts (which implies that H . In a stationary method. The eigenvalues are xed for a given h by the choice of time-marching method. if all of the eigenvalues of H .9.7. and the secondary conditioning matrix in 9. ~ is also independent u f . Hence the numerical solution after n steps of a stationary relaxation method can be expressed as (see Eq. As we shall see. then the system of ODE's has a steady-state solution which is approached as t ! 1. 9. In general it is assumed that any or all of the eigenvectors could have been given an equally \bad" excitation by the initial guess. 1 . a constant for the entire iteration. given by 1 1 1 1 ~ 1 = A. The eigenvalues are xed by the basic matrix in Eq.38) which is the solution of Eq. A are linearly independent. THE ODE APPROACH TO CLASSICAL RELAXATION 1 1 1 171 In the special case where H . the only free choice remaining to accelerate the removal of the error terms is the choice of h.3. is now referred to in relaxation analysis as the error.36. the \time" step. Assuming that H .

Insert the model equation 9.29 into the ODE form 9.42) +1 +1 It is clear that the best choice of H from the point of view of matrix algebra is . RELAXATION METHODS 9.35.172 CHAPTER 9.2 1)~ . 1 +1 TABLE 9. ~ n) = B (1 . ~ f (9. this is not in the spirit of our study.3 is that all the elements above the diagonal (or below the diagonal if the sweeps are from right to left) are zero. ~ f (9. ~ f (9.2. equal to 1. 9.2 1) since then multiplication from the left by .3. With this choice of notation the three methods presented in Section 9.2 1)~ n .B .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 are led to 2 B (.2 1) gives the correct answer in one step.3 can be identi ed using the entries in Table 9. Now let us study these methods as subsets of ODE as formulated in Section 9. The constraint on H that is in keeping with the formulation of the three methods described in Section 9.2 ODE Form of the Classical Methods The three iterative procedures de ned by Eqs.1.43) where and ! are arbitrary.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 . Then ~ H d = B (1 . ~ n) = B (1 . We arrive at H (~ n .31 and 9.3.2. h. ! 0)(~ n . 9.40) dt As a start.B (1 . let us use for the numerical integration the explicit Euler method 0 (9.2 1)~ n .1.1: VALUES OF and ! IN EQ. 9. since multiplication by the inverse amounts to solving the problem by a direct method without iteration.30. If we impose this constraint and further restrict ourselves to banded tridiagonals with a single constant in each band.41) n = n+h n with a step size. (1 . However.

2 2 j 2 j 1 ( +1) ( +1) 1 ( ) 1 ( ) ( ) +1 This represents a generalization of the classical relaxation techniques. ~ f (9. or steady-state.4 Eigensystems of the Classical Methods The ODE approach to relaxation can be summarized as follows. 1) n + !h n .44) dt ! and appear from it in the special case when the explicit Euler method is used for its numerical integration. !h f (9.1. 9. However. solution of the set of ODE's ~ H d = A~ . ~ = 0 f (9. or error. The basic equation to be solved came from some time-accurate derivation Ab~ . are identi ed for the one-dimensional case by Eq. and ~ 1 A.47) An iterative scheme is developed to nd the converged. ~ is the steady-state solution.2 1)~ . ~ b = 0 u f (9. 1 .45) j j. j 2 j. (!h . our purpose is to examine the whole procedure as a special subset of the theory of ordinary di erential equations. ) n . 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.48) dt This solution has the analytical form ~ n = ~n + ~ 1 e (9.44 and Table 9.9. In this light. (. Gauss-Seidel.4. Point-Jacobi.49) where ~ n is the transient. 9.46) This equation is preconditioned in some manner which has the e ect of multiplication by a conditioning matrix C giving A~ . The point operator that results from the use of the explicit Euler scheme is ! ! n = ! n + ! (h . The e f three classical methods. and SOR. the three methods are all contained in the following set of ODE's d~ = B . 2 0) B (1 . ~ f (9.

c = 1.50) In this section. To obtain the optimal scheme we wish to minimize the maximum j m j which occurs when h = 1. the ODE matrix H .32. the maximum j m j is obtained with m = M . A) having maximum absolute value.2.44. and f = 0. e = 2=!. 9. This amounts to solving the generalized eigenvalue problem A~ m = m H~ m x x (9. Thus Convergence rate j mjmax m=1 2 1 M (9. RELAXATION METHODS 1 Given our assumption that the component of the error associated with each eigenvector is equally likely to be excited. i. The plot for h = 1. For h < 1. j j = j M j and the best possible convergence rate is achieved: 1 1 j m jmax = cos M + 1 (9. The three special cases considered below are obtained with a = 1.30. Note that for h > 1:0 (depending on M ) there is the possibility of instability.4.1 + cos M +1 m = 1 2 ::: M The .1 since both d and f are zero.1 2 ). the asymptotic convergence rate is determined by the eigenvalue m of G ( I + H .51) for the special case 2 0)~ (9. is shown in Fig.53) m = . It is also instructive to inspect the eigenvectors and eigenvalues in the H . The eigenvalues are given by the equation m (9. 9. . A reduces to simply B ( 1 .1. 9.52) ! xm The generalized eigensystem for simple tridigonals is given in Appendix B.3.174 CHAPTER 9. 9. A matrix for the three methods. Fig. we use the ODE analysis to nd the convergence rates of the three classical methods represented by Eqs. d = . B (1 . b = . Fig. and 9.54) . This relation can be plotted for any h.1 The Point-Jacobi System 1 If = 0 and ! = 1 in Eq.e. we take M = 5 for the matrix order. j m j > 1:0. To illustrate the behavior.2 1)~ m = x m B (. the optimum stationary case. 9. the maximum j mj is obtained with m = 1.2.2 and for h > 1. 9. 2 The eigensystem can be determined from Appendix B.relation for the explicit Euler method is m = 1 + m h.31. 9. This special case makes the general result quite clear.

we obtain j mjmax = 0:9971. 3=2 7 ~ = 6 p0 7 x = 6 p 7 6 1 7 x 6 6 7 6 3=2 7 7 6 .0:5 2 = .1 . 9. EIGENSYSTEMS OF THE CLASSICAL METHODS 1 175 For M = 40.56) The corresponding eigenvalues are. 2 )h]n~ 1 2 2 p 1 . The rst 5 eigenvectors are simple sine waves. 9. For M = 5.1 .9.39. A are given by ~ j = (xj )m = sin j m x j = 1 2 ::: M (9. 23 = . 3=2 7 6 6 6 7 ~ 6 . ~ 1 = c 1 .1 + 1 = . the eigenvectors of H . (1 .1 + 23 = .1. 3=2 7 4 5 4 5 5 4 p 7 1 1=2 .53 1 2 3 = .1:5 2 p (9. (1 .1 7 6 7 x 6 0 7 x 6 7 x 6 6 7 6 6 3=2 7 6 0 7 7 6 . from Eq.1 = . 3=2 1 2 3 4 5 (9.23 times its initial level.0:134 = . 3=2 3 2 p 3 2 3 1= p =2 7 p2 7 6 .4.55) M +1 This is a very \well-behaved" eigensystem with linearly independent eigenvectors and distinct eigenvalues. Again from Appendix B. 3 )h]n~ x 2 1 x + c 1 .1:866 From Eq. 1 = . Thus after 500 iterations the error content associated with each eigenvector is reduced to no more than 0. 3=2 7 5 4 p 5 4 1=2 . 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 . the numerical solution written in full is 5 p ~ n .57) 4 = .1:0 = .

h = 1 M = 5.2. One can show that the H .59) 1 which can be studied using the results in Appendix B. RELAXATION METHODS h = 1. (.2 1)~ m = x m B (. (1 + 23 )h]n~ x 4 5 )h]n~ x 3 5 (9.0 -2. + c 1 .4.0 Figure 9.58) 9.176 1.0 m=1 Values are equal σm m=4 m=2 Μ=5 0.0 m=5 CHAPTER 9. (1 3 4 + c 1 . is AGS B .2 The Gauss-Seidel System If and ! are equal to 1 in Eq.1: The relation for Point-Jacobi.1 2 0)~ m x (9. (1 + 1 )h]n~ 2 x p + c 1 .1 2 0)B (1 . the matrix eigensystem evolves from the relation B (1 .0 0.0 m=3 -1. A matrix for the Gauss-Seidel method. 9.44. AGS .2 1) = 1 .

0 0.3: The relation for Point-Jacobi.072 1.0 at h De cr ea -1. h = 0:9 M = 5. EIGENSYSTEMS OF THE CLASSICAL METHODS Approaches 1. .0 177 1. Exceeds 1.0 λ mh ∼ ∼ 1.0 -2.0 h < 1.0 0.0 ng si ea cr In In cr ea si σm ng h h M = 5 0.9.0 λmh Figure 9.4.0 -2.0 h > 1.0 -1.2: The relation for Point-Jacobi.0 h 0.0 De cr ea h si ng σm ng si Figure 9. h = 1:1 M = 5.072 / Ustable h > 1.

. 9.principal vectors. j m jmax = 0:9942. RELAXATION METHODS 2 .1 + cos ( M +1 and the corresponding eigenvectors are given by j. and there are (M .. They are no longer symmetrical. The m with the largest amplitude is obtained with m = 1.. . .1 1=2 3 1 6 0 . 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. . . . the error associated with the \worst" eigenvector is removed in half as many iterations. .23.3=4 1=2 7 4 6 7 (9. ..relation for h = 1. m + ~ m = cos m x sin j M + 1 m = 1 2 ::: M2 1 (9. is shown in Fig. Furthermore. 7 . . the optimum stationary case. . The Jordan canonical form for M = 5 is j m jmax = cos M + 1 2 (9. 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. Hence the convergence rate is 2 1 CHAPTER 9. producing waves that are higher in amplitude on one side (the updated side) than they are on the other. .61) m = .3=4 1=2 7 3 6 7 6 7 6 0 1=16 1=8 .. 6 . they do not represent a common family for di erent values of M . .3=4 1=2 7 5 6 7 6 .4. . The equation for the nondefective eigenvalues in the ODE matrix is (for odd M ) + m ) m = 1 2 ::: M2 1 (9. .60) 6 0 1=32 1=16 1=8 .62) M +1 The .63) 3 2 h~ i 7 6 h~ i 7 6 6 7 2~ 37 . 7 .64) . The eigenvectors are quite unlike the Point-Jacobi set.3=4 1=2 7 2 6 7 6 0 1=8 .. 4 5 0 1=2M M The eigenvector structure of the Gauss-Seidel ODE matrix is quite interesting. For M = 40. . If M is odd there are (M + 1)=2 distinct eigenvalues with corresponding linearly independent eigenvectors.

4: The relation for Gauss-Seidel.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 . 34 )n~ 1 2 3 3 1 1 2 2 (9.9.0 179 h = 1. ~ 1 = c (1 .1=4 = . The eigenvectors and principal vectors are all real. linked to (5) Jordan block The numerical solution written in full is thus ~ n .3=4 ) = .4.66) .0 -1. EIGENSYSTEMS OF THE CLASSICAL METHODS 1.0 0.1 (4) Defective. 3=32 1 2 3 4 5 The corresponding eigenvalues are = .0 -2.0 λmh Figure 9. For M = 5 they can be written 2 3 2 p 3 2 3 2 3 2 3 3=2 1=2 7 17 0 7 6 3=4 7 6 p3=4 7 607 6 2 7 6 0 7 6 6 7 6 6 6 0 7 ~ = 6 3=4 7 ~ = 6 p0 7 ~ = 6 0 7 ~ = 6 .4 7 4 5 4 p 5 4 5 4 5 4 5 9=32 0 0 1 . h )n~ 4 x h x + c (1 . 3=16 7 7 607 6 0 7 6 .0 σm Μ=5 2 defective λ m 0.1 7 ~ = 6 4 7 (9. h = 1:0 M = 5.

2x 0 0 0 x 4 4 3 7 7 7 7 (9.4.3 The SOR System 1 If = 1 and 2=! = x in Eq. 3. !) + ! p = 0 ! is optimum for the stationary case. ASOR B .1 + !pm 2 zm 2 m = 1 2 :::M 2 2 1 2 (9. the remaining eigenvalues are complex and occur in conjugate pairs.2 + x 1 . (. !) + ! pm] = pm = cos m =(M + 1)] 2 2 2 1 If ! = 1. 2x + x x .2x + x x x 2x2x x 6 x + . If M is even. x 0 6 1 6 .2 1). h)n~ x 5 1! 1 4 5 (9.2. The generalization to any M is fairly clear. .44. and the following conditions hold: 1. ~ x 3 4 1 5 2 2 3 4 5 + c (1 . ~ + c (1 . 2x 4 .1 x 0)B (1 . h) x 1! 2! + c (1 . h)n.2x + x x . + x . If M is odd.67) 9. (. h)n + c h n (1 . is 1 1 2 x 0 0 6 . If ! is chosen such that 4(1 .69) where zm = 4(1 . h)n. RELAXATION METHODS " # n + c h n (1 . h)n. the ODE matrix is B . the system is Gauss-Seidel. one of the remaining eigenvalues is real and the others are complex occurring in conjugate pairs.180 CHAPTER 9. + c h n(n .1 x 0)B (1 . Two eigenvalues are real. equal and defective. 2x + x x . The H . If 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 . 2x x x 6 . 2x + x x . A matrix for the SOR method. 1) (1 . One can show that this can be written in the form given below for M = 5.2 1). !) + ! pm < 0 zm and m are complex. 2x + x 6 x .2x x .68) 7 7 5 Eigenvalues of the system are given by m + = . 9. 2.

23 times its initial value in only 10 iterations. j mjmax = 0:8578.74) 2 1 2 34 5 1 1 3 4 5 !opt p + i p . In practical applications. much faster than both GaussSeidel and Point-Jacobi. 1.71) Using the explicit Euler method to integrate the ODE's.5. and the bene t may not be as great. the . the optimum value for the stationary case. p m= m 2 m 2 1 2 q .2=3 (2) Defectivep linked to = . For odd M .(10 .(10 + 2 2i)=9 = . m = 1 . 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. and if h = 1.relation reduces to that shown in Fig. 1 (9. 2p2i)=9 = .4. Hence the worst error component is reduced to less than 0.73) 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 = . This illustrates the fact that for optimum stationary SOR all the j m j are identical and equal to !opt .9. The remaining linearly independent eigenvectors are all complex. EIGENSYSTEMS OF THE CLASSICAL METHODS One can easily show that the optimum ! for the stationary case is and for ! = !opt 181 (9. For M = 5 they can be written 2 3 2 3 2 p3(1 3 2 3 1=2 7 .1 j .72) !opt = 2= 1 + sin M + 1 For M = 40.70) !opt = 2= 1 + sin M + 1 ~m x m = = m. Hence the convergence rate is j m jmax = !opt . the optimum value of ! may have to be determined by trial and error. 9.1 m sin 2 where m j M +1 (9.4=3 (9. there are two real eigenvectors and one real principal vector.

]~ x c (1 .0 0. The nonstationary form of Eq.75) 9.5 Nonstationary Processes In classical terminology a method is said to be nonstationary if the conditioning matrices.0 λmh Figure 9.0 -2. 2p2i)h=9]n~ x n~ c 1 .5: The relation for optimum stationary SOR. For the Gauss-Seidel and SOR methods it leads to processes that can be superior to the stationary methods. are varied at each time step. 9. but it can greatly a ect the convergence rate. The numerical solution written in full is ~n . h.0 -1. It is much simpler.0 2 real defective λ m σm 2 complex λ m 1 real λ m Μ=5 0. 2h=3)n + c nh(1 .0 CHAPTER 9. M = 5.182 1.39 is 1 . 2h=3)n. In our ODE b f approach this could also be considered and would lead to a study of equations with nonconstant coe cients. This does not change the steadystate solution A. RELAXATION METHODS h = 1. h = 1. ~ b . (10 + 2 2i)h=9] x c (1 . to study the case of xed H and C but variable step size. H and C . (10 . This process changes the Point-Jacobi method to Richardson's method in standard terminology. 2h=3)n~ x p c 1 . 4h=3)n~ x 1 2 1 2 3 2 3 4 5 4 5 1 (9. however. ~1 = + + + + c (1 .

we considered making the error exactly zero by taking advantage of some knowledge about the discrete values of m for a particular case.9. Now we pose a less demanding problem. Let us choose the hn so that the maximum value of (Pe)N ( ) is as small as possible for all lying between a and b such that b a 0. 9. In the annihilation process mentioned after Eq. ~ 1 = X cm~ m Y (1 + e x m=1 n=1 m hn ) (9. 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). This leads to the concept of selectively annihilating clusters of eigenvectors from the error terms as part of a total iteration process.5.76. Consider one of the error terms taken from M N ~ N ~ N .76) where the symbol stands for product. Since hn can now be changed at each step. 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. Now focus attention on the polynomial (Pe)N ( ) = (1 + h )(1 + h ) 1 2 (1 + hN ) (9. we seek b max a j(Pe)N ( )j = minimum with(Pe)N (0) = 1 (9. Mathematically stated.79) treating it as a continuous function of the independent variable .1= m for m = 1 2 : : : M . for m = 1 2 M . This is the basis for the multigrid methods discussed in Chapter 10.77) and write it in the form cm~ m Pe( m ) cm~ m x x N Y n=1 (1 + m hn ) (9. We will see that a few well chosen h's can reduce whole clusters of eigenvectors associated with nearby 's in the m spectrum. the eigenvalues m are generally unknown and costly to compute. the error term can theoretically be completely eliminated in M steps by taking hm = . However.78) where Pe signi es an \Euler" polynomial.1= m . It is therefore unnecessary and impractical to set hm = .80) .

Eq. ) cos (2n . 1) n = 1 2 :::N (9. a. 9. TN . . As an example for the use of Eq. y .1 using only 3 iterations. + ( .1 y 1 and 2 2 (9. In relaxation terminology this is generally referred to as Richardson's method. 1 + 1 y .83) 2 are the Chebyshev polynomials for jyj > 1. b TN a. It is 2 . ! (Pe)N ( ) = where (9.82) q q N N 1 TN (y) = 2 y + y .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 . RELAXATION METHODS a. 9. The three values of h which satisfy this problem are (9. The error in the relaxation process represented by Eq. 1 (9. b a b This problem has a well known solution due to Markov.184 CHAPTER 9. 1) 6 " #! (9. ~ m .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. ampli ed by the coe cients cm Q(1 + m hn).81) TN (y) = cos(N arccos y) are the Chebyshev polynomials along the interval . With each x eigenvector there is a corresponding eigenvalue.2 . b ! . and it leads to the nonstationary step size choice given by ( " #) 1 = 1 .85) hn = 2= 3 .76 is expressed in terms of a set of eigenvectors. let us consider the following problem: Minimize the maximum error associated with the eigenvalues in the interval .84.86) . 9. cos (2n .

6 are h = 4=(6 .88) where n p p o T (3) = 3 + 8] + 3 . with (Pt)N (0) = 1 (9.41. 9. 9. 3) h = 4=(6 . 8] =2 99 3 3 3 A plot of Eq. 9. 9. namely that =1 0 1 M N 1+ h ~ N = X cm~ m Y B 2 n B x @ 1 m n 1 .9. the nonstationary formula 1 h( 0 + 0 ) n 2 n +1 (9. Pt: 0 1 1 N B 1 + hn C Y (Pt )N ( ) = B 2 C (9.87 is given in Fig.6 and we see that the amplitudes of all the eigenvectors associated with the eigenvalues in the range .5.87) (9. p 0) h = 4=(6 + 3) 1 2 3 p Return now to Eq.92) .89) would result. The values of h used in Fig. This was derived from Eq.90) b max a j(Pt)N ( )j = minimum . 9. Eq. This calls for a study of the rational \trapezoidal" polynomial. 2 hn under the same constraints as before. 2 hn =1 =1 1 mC C + ~1 A m (9.91) @ 1 A n 1 . If instead the implicit trapezoidal rule n+1 = n+ is used. NONSTATIONARY PROCESSES and the amplitude of the eigenvector is reduced to (Pe) ( ) = T (2 + 3)=T (3) 3 3 3 185 (9. was used to integrate the basic ODE's.76. 9.2 .37 on the condition that the explicit Euler method.1 have been reduced to less than about 1% of their initial values.

9 0.6: Richardson's method for 3 steps.4 −0.2 .8 0.015 −0.005 (P ) (λ) e 3 0 −0.6 −0.2 0 0.6 −1.4 −0.8 −1.186 CHAPTER 9. minimization over . RELAXATION METHODS 1 0.2 −1 λ −0.2 0 Figure 9.2 −1 λ −0.6 (Pe )3 (λ) 0. .6 −1.02 −2 −1.2 0.5 0.01 −0.4 0.1 0 −2 −1.4 −1.1.7 0.8 −0.4 −1.01 0.015 0.6 −0.005 −0.02 0.3 0.8 −1.8 −0.

Using Appendix B. Recall that the eigenvalues of H .85. A.6 Problems 1. a n=1 2 N (9. Given a relaxation method in the form ~ H ~n = A~n . 1 1 1 . For a linear system of the form (A + A )x = b. Show that this iterative method can be written in the form H (xk . The error amplitude is about 1/5 of that found for (Pe) ( ) in the same interval of . Gn)A. x x not just the expressions.9.2.7 are h = 1 p h = 2 h = 2 ( 1) ( 1) 3 3 1 2 3 9.7. = N . 2. f show that 1 ~n = Gn ~ + (I . (You do not have to nd H . A )xn + b x (I + A )xn = (I . f ~ 0 1 1 2 where G = I + H . A satisfy A~ m = mH~ m . The values of h used in Fig. nd the eigenvalues of H .) Find the numerical values. A for the SOR method with A = B (4 : 1 .6. The results for (Pt) ( ) are shown in Fig. Then nd the corresponding j m j values. PROBLEMS ! 187 The optimum values of h can also be found for this problem.2 1) and ! = !opt. 9.93) b hn b This process is also applied to problem 9.1=2. A )~ + b x where is a parameter. consider the iterative method (I + A )~ = (I . 2 =. 3. xk ) = (A + A )xk . . 9. but we settle here for the approximation suggested by Wachspress n. b 1 2 2 +1 1 +1 1 2 Determine the iteration matrix G if = .

6 −1.5 1 0.5 −1 −1. RELAXATION METHODS 1 0.2 −1 λ −0.4 −1.4 0.2 .188 CHAPTER 9.9 0.1 0 −2 −1.8 −1.2 0 Figure 9.6 −1.3 0.5 0.4 −0.8 −0.2 0.7: Wachspress method for 3 steps.7 0.4 −0. . minimization over .2 0 2 x 10 −3 1.6 (P ) (λ) t 3 0.8 −0.2 −1 λ −0.5 (Pt )3 (λ) 0 −0.8 0.6 −0.8 −1.6 −0.1.5 −2 −2 −1.4 −1.

Compare with the theoretical asymptotic convergence rate. the number of iterations. u(1) = 1: @ u . 3. Determine the asymptotic convergence rate. use u(x) = 0. (b) the Gauss-Seidel method. PROBLEMS 189 4. and 4 orders of magnitude. (c) the SOR method with the optimum value of !. and (d) the 3-step Richardson method derived in Section 9. Plot the solution after the residual is reduced by 2.9. Iterate to steady state using (a) the point-Jacobi method.5. Plot the logarithm of the L -norm of the residual vs. 2 2 2 . Solve the following equation on the domain 0 x 1 with boundary conditions u(0) = 0. Use second-order centered di erences on a grid with 40 cells (M = 39).6. 6x = 0 @x For the initial condition.

RELAXATION METHODS .190 CHAPTER 9.

3) xx~ = x x 191 2 2 2 2 1 .m sin(mx) (10. 10. Note that @ sin(mx) = .2 and 4. That is. the space-frequency (or wavenumber) of the corresponding eigenvectors also increase.2) @x and recall that 1 B (1 .1 Eigenvector and Eigenvalue Identi cation with Space Frequencies Consider the eigensystem of the model matrix B (1 .3.1 Motivation 10.3.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. There are many variations of the process and many viewpoints of the underlying theory. The eigenvalues and eigenvectors are given in Sections 4. This has a rational explanation from the origin of the banded matrix.3.2 1)~ = X 1 D(~ ) X . Notice that as the magnitudes of the eigenvalues increase.1) then the corresponding eigenvectors are ordered from low to high space frequencies. The viewpoint presented here is a natural extension of the concepts discussed in Chapter 9. if the eigenvalues are ordered such that j j j j 1 2 2 j Mj (10.1. ~ (10. respectively.2 1).

This is the key concept underlying the multigrid process.2. 10. However.2 + 2 cos m .4) m= M +1 x Hence. as shown in Fig. 9. We have seen that X . The classical point-Jacobi method does not share this property.1. It is also true. In fact. this method produces the same value of j j for min and max . Therefore.m . For this matrix one nds. for example. However.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 . of the Richardson method described in Section 9. This is to be expected of an iterative method which is time accurate.1. by design. As we saw in Section 9. this correlation is not necessary in general. 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.4.192 1 CHAPTER 10. MULTIGRID where D(~ ) is a diagonal matrix containing the eigenvalues. and X ~ . the operation 1 D(~ ) represents the numerical approximation of the multiplication of the x appropriate sine wave by the negative square of its wavenumber. . One nds that 1 M + 1 .2 Properties of the Iterative Method The second key motivation for multigrid is the following: Many iterative methods reduce error components corresponding to eigenvalues of large amplitude more e ectively than those corresponding to eigenvalues of small amplitude. the complete counterexample of the above association is contained in the eigensystem 1 for B ( 2 1 1 ). ~ represents a sine transform. exactly the opposite 2 behavior. the correlation of large magnitudes of m with high space-frequencies is to be expected for these particular matrix operators. the property can be restored by using h < 1. a sine synthesis.5.m m << M (10. of the Gauss-Seidel method and. This is consistent with the physics of di usion as well. from Appendix B. error components corresponding to high space frequencies will be reduced more quickly than those corresponding to low space frequencies. 2 2 2 2 2 10.

7) where A CAb (10. 2. 9.11.8) . 4.6) Recall that the ~ used to compute ~ is composed of the exact solution ~ 1 and the r ~ in such a way that error e A~ . we are led to the starting formulation C Ab ~ 1 .10.5. r where ~ = C Ab~ . 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. ~ = 0 e r (10.2. The eigenvalues. The eigenvectors associated with the eigenvalues having largest magnitudes can be correlated with high frequencies on the di erencing mesh. as in the example given by Eq. but for clarity we suppose that the three-step Richardson method illustrated in Fig.27.5) where the matrix formed by the product CAb has the properties given above. ~ b] r f (10. 9. the residual. These conditions are su cient to ensure the validity of the process described next. if f ~ 1 is a vector representing the desired exact solution. Having preconditioned (if necessary) the basic nite di erencing scheme by a procedure equivalent to the multiplication by a matrix C . The problem is linear. as in the examples given by Eq. ~ b ] = 0 f (10. We start with some any. 3. The m are fairly evenly distributed between their maximum and minimum values. In Eq. or it can be \contrived" by further conditioning. of the matrix are all real and negative. the vector ~ b represents the boundary conditions and the forcing function. THE BASIC PROCESS 193 basic properties. We do not attempt to develop an optimum procedure here. 5. m .6 is used. 10. 3. The basic assumptions required for our description of the multigrid process are: 1. At the end of the three steps we nd ~ . and initial guess for ~ 1 and proceed through n iterations making use of some iterative process that satis es property 5 above. This form can be arrived at \naturally" by simply replacing the derivatives in the PDE with di erence schemes.

9) Thus our goal now is to solve for ~ . e In addition. MULTIGRID (10.14) .13) Notice that e r e A ~e + A ~o = ~e 1 2 (10. 10. P ]~ = P ~ e r (10. and the eigenvalues of the Richardson process in the form: M= M ~ = X cm~ m Y ( m hn)] + X cm~ m Y ( mhn)] e x x 2 3 3 m=1 n=1 m=M=2+1 | very low amplitude {z n=1 (10.12) 1 The operation PAP . we can write PA P . 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 (10.10) } Combining our basic assumptions.~ e CHAPTER 10. assumption 4 ensures that the error has been smoothed. and the other the even entries of the original vector (or any other appropriate sorting which is consistent with the interpolation approximation to be discussed below). For a 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. since a permutation matrix has an inverse which is its transpose. one containing the odd entries. 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).11) Multiply Eq. We can write the exact solution for ~ in terms of e e the eigenvectors of A. 10.7 from the left by P and.194 If one could solve Eq. Next we construct a permutation matrix which separates a vector into two parts.7 for ~ then e ~1 = ~ .

it is reasonable to suppose that the odd points are the average of the even points. r If the boundary conditions are Dirichlet. e no operations on ~ are required or justi ed. 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.10. ea and eb are zero. ~ e = 0 e r where Ac = A + A A0 ] 1 2 2 . the averaging of ~ is our only approximation.19) or Ac~ e . notice that. It is that there is some connection between ~ e and ~ o brought about by the smoothing property of the Richardson e e relaxation procedure. Since the top half of the frequency spectrum has been removed.14 reduces to A ~ e + A A0 ~ e = ~ e e e r 1 2 2 (10.18) (10. Hence. 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. For example 1 (e + e ) e 2 a 1 (e + e ) e 2 1 (e + e ) or ~ o = A0 ~ e e e (10.16) 0 0 1 With this approximation Eq. no aliasing of these functions can occur in subsequent steps. 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.17) (10.2. Finally. At this point we make our one crucial assumption.15) e 2 1 (e + e ) e b 2 1 2 3 2 4 5 4 6 2 7 6 (10.

19.1 1=2 7 5 1=2 .1]T .2 1 1 1 .2 1 1 1 1 1 1 . 10. and all of them re ect on the right (Neumann) side. 9. B. For Neumann conditions.2 6 6 6 6 6 6 . 10.2 .2 .1.18 gives z }| 2 .2 . is completely determined by the choice of the permutation matrix and the interpolation approximation.21) 5 If the boundary conditions are mixed Dirichlet-Neumann. eb is equal to eM . the matrix on the coarse mesh. If the original A had been B (7 : 1 .22) and are illustrated in Fig. 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 (10.2 . However. When eb = eM . the interpolation formula in Eq. A in the 1-D model equation is B (1 ~ 1) where ~ = .1 { 3 7 (10.20) and Eq.1 are met.196 CHAPTER 10. our 7-point example would produce 2 .1. All of them go through zero on the left (Dirichlet) side. In the present case they are given by " !# (2m . all of the properties given in Section 10.2 4 A1 .1 1=2 7 6 7 = 4 1=2 . the eigenvector structure is di erent from that given in Eq.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 . in fact. the example in Eq. The eigensystem is given by b b Eq.23) 4 5 0 0 2 1 2 . 1) xjm = sin j 2M + 1 m=1 2 M (10. If Neumann conditions are on the left.1 = 6 PAP 6 1 6 6 6 1 6 6 4 . MULTIGRID The form of Ac. and. ea = e .2 ::: .2 6 .2 1).2 . 10. In the particular case illustrated in Fig. It is easy to show that the high space-frequencies still correspond to the eigenvalues with high magnitudes.16 changes to 2 1 0 03 7 16 A0 = 2 6 1 1 0 7 60 1 17 (10. 10.15 must be changed.55 for Dirichlet conditions.

we can construct the coarse matrix from 1 1 2 4 }| z 2 . THE BASIC PROCESS 197 X Figure 10. cos m=1 2 M M +1 M +1 1 1 .2 1 ." We will continue with Dirichlet boundary conditions for the remainder of this Section. the relationship between e e In order to examine this relationship.2 4 A1 .2 1)~e = ~ e on the next coarser mesh. we need to consider the eigensystems of A and Ac: A = X X. we must now determine e ~ e and ~ .2 1) the eigenvalues and eigenvectors are m j=1 2 M (10. Recall that our goal is ne mesh to 2 e r to solve for ~ .1 1=2 7 6 7 = 4 1=2 .1 in the lower right element).10.2 { 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 .24) 5 which gives us what we might have \expected. Therefore. we can compute ~ o e using Eq.2. which will provide us with the solution ~ 1 using Eq.2 1)~ = ~ on the e r 1 B (1 . Given ~ e e e computed on the coarse grid (possibly using even coarser grids). and thus ~ .26) ~ m = sin j m x m = . (10. are unchanged (only A is modi ed by putting .15. 10. In order to complete the process. we have reduced the problem from B (1 .1 7 5 1=2 .25) For A = B (M : 1 . The permutation matrix remains the same and both A and A in the partitioned matrix PAP .9.1=2 { 3 1=2 7 (10. At this stage.2 6 .1: Eigenvectors for the mixed Dirichlet{Neumann case. Ac = Xc cXc. 10.

31) 2 3 1 607 6 7 = Xc .27) M +1 For example. ~ = ~ . x Now let us consider the case in which all of the error consists of the eigenvector component ~ . it contains the even elements of ~ .0:007291 = 1:998 . 6 . the most di cult error mode to eliminate is that with m = 1.198 CHAPTER 10. (~ c) e x c r c (10. cos M + 1 x j=1 2 M (10. If we restrict our attention to odd M . 7 4 5 0 (10. cos M2+ 1 (~ c) = sin j M2+ 1 j = 1 2 x Mc (10.e. we obtain ( c) = . corresponding to ~ = sin j = . The eigenvalue and eigenvector corresponding to m = 1 for the matrix Ac = 1 B (Mc 1 .33) 1 1 1 1 1 1 1 1 1 . 1)=2 is the size of Ac.0:003649.30) since (~ c) contains the even elements of ~ . Then the residual is x e x ~ = A~ = ~ r x x (10. ( c) approaches 2 . Xc. 7 c 6 7 4 . 7 6 .28) For M = 51 (Mc = 25). MULTIGRID Based on our assumptions. ~ e = Xc . 1 . that is. = . The exact solution on the coarse grid x x satis es ~ e = A. In addition. 5 0 (10.. then Mc = (M .2 1) are 2 ( c) = .2 1 . with M = 51. one can easily see that (~ c) coincides with ~ at x x every second point of the latter vector.32) 2 3 1=( c) 6 0 7 6 7 = Xc 6 ..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.. As M increases. i.

the matrix A = B (M : 1 . which gives Ab = x B (M : 1 .2 1) = 1 B (Mc : 1 . carried to even coarser grids before returning to the nest level.2. and usually is. This is equivalent to solving or 1A ~ = ~ e r 2 ce e (10.38) C= x I 2 (10.2 1) 4 c c 2 2 2 (10. However. obviously. since xc = 2 x. The details of nding optimum techniques are.40) which is precisely the matrix appearing in Eq.37) c 4 In our case.15). The reduction can be.2 1)~ = ~ ee re (10.10.2 1) = x B (Mc : 1 . x C x B (Mc : 1 .and up-going paths. in addition to interpolating ~ e to the ne grid e x e (using Eq.39) Applying the discretization on the coarse grid with the same preconditioning matrix as used on the ne grid gives. THE BASIC PROCESS = ( c) 1 1 199 (~ c) x 1 1 1 (~ ) 2 xc 1 (10. 10.35) Since our goal is to compute ~ = ~ . The problem to be solved on the coarse grid is the same as that solved on the ne grid. Thus we see that the process is recursive.34) (10. . but they vary depending on the problem and the user.36) 1 B (M : 1 . in each case the appropriate permutation matrix and the interpolation approximation de ne both the down.2 1) and the preconditioning matrix C is simply 2 (10. 10.37.2 1) comes from a discretization of the di usion equation. The remaining steps required to complete an entire multigrid process are relatively straightforward. quite important but they are not discussed here. we must multiply the result by 2.

1. Perform n iterations of the selected relaxation method on the ne grid. Eq.43) 1 1 1 1 1 (1) (1) (1) 1 1 1 1 1 1 2. starting with ~ = ~n. f (10. Transfer (or restrict) ~ to the coarse grid: r ~ =R ~ r r (10.45) 4 5 0 0 0 0 0 1 0 that is. Gn1 ) A. Solve the problem A ~ = ~ on the coarse grid exactly: e r ~ = A.. Next compute the residual based on ~ : ~ ~ ~ ~ = A ~ . the restriction matrix is 2 3 0 1 0 0 0 0 0 R =6 0 0 0 1 0 0 0 7 (10. 9. the rst three rows of the permutation matrix P in Eq.3 A Two-Grid Process CHAPTER 10.200 10. Note that the coarse grid matrix denoted A2 here was denoted Ac in the preceding section. which can be easily generalized to a process with an arbitrary number of grids due to the recursive nature of multigrid. This type of restriction is known as \simple injection. f r ~ = AGn1 ~n . f . MULTIGRID ~ We now describe a two-grid process for the linear problem A ~ = f . A (10. Call the result ~ .46) 1 2 See problem 1 of Chapter 9.41) 1 (1) 1 (1) 1 1 1 where G = I + H. 10.11. This gives ~ = Gn1 ~n + (I . ~ e r 2 (2) (2) (2) 2 1 (2) 2 (10.21). AGn1 A. Extension to nonlinear problems requires that both the solution and the residual be transferred to the coarse grid in a process known as full approximation storage multigrid. (1) (2) 2 (1) 1 2 1 3. f ~ (10." Some form of weighted restriction can also be used. f = AGn1 ~n + A (I . .44) In our example in the preceding section.g. Gn1 ) A.42) and H is de ned as in Chapter 9 (e.

solve exactly.2 1). R A ]Gn2 2 3 2 3 3 1 3 2 2 2 2 2 3 3 2 3 (10. The basic iteration matrix for a three-grid process is found from Eq. 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. R A]Gn1 A. 10. Transfer (or prolong) the error back to the ne grid and update the solution: ~n = ~ . and A is obtained by discretizing on grid 3.50 by replacing A. I A. A TWO-GRID PROCESS 2 2 201 Here A can be formed by applying the discretization on the coarse grid. R A]Gn1 ~ n . 10. . 10. I and R are the transfer operators between grids 2 and 3. I A.10. where 2 2 3 2 G = I . I A. A = 4 B (Mc : 1 . one obtains ~n = I . I . Otherwise. It is at this stage that the generalization to a multigrid procedure with more than two grids occurs.15.49) (10. I ~ e +1 (1) 1 (2) 2 (10. f ~ ~ +1 1 2 2 1 2 1 1 1 2 2 1 2 1 1 1 1 (10.48) which follows from Eq. G )A. If this is the coarsest grid in the sequence. 4. Extension to four or more grids proceeds in similar fashion. Combining these steps. . I A. R A]Gn1 1 2 2 1 2 1 1 1 1 The eigenvalues of this matrix determine the convergence rate of the two-grid process. with (I .50) Thus the basic iteration matrix is I .40).3.47) 3 7 7 7 7 7 7 7 7 7 7 7 5 In our example. f + A. In the 1 preceding example (eq.51) In this expression n is the number of relaxation steps on grid 2. apply the two-grid process recursively.

10. Plot the solution after the residual is reduced by 2.5. Solve exactly on the coarsest grid. Derive Eq.4 Problems CHAPTER 10. and 4 orders of magnitude.51. 2 . Repeat problem 4 of Chapter 9 using a four-grid multigrid method together with (a) the Gauss-Seidel method.202 10. Calculate the theoretical asymptotic convergence rate and compare. MULTIGRID 1. 2. the number of iterations. Plot the logarithm of the L -norm of the residual vs. 3. (b) the 3-step Richardson method derived in Section 9. Determine the asymptotic convergence rate.

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

a( x u)j = 2. We proceed next to show why this occurs. A backward di erence operator is given by = 1 and a forward di erence operator is given by = .uj.1) produces Re( m ) > 0.uj. )uj+1] = . uj+1)=2 x.1 + 2 uj + (1 .1 + 2uj . cos 2M m + i sin 2M for m = 0 1 : : : M . When Re( m) = 0.1. uj+1)] 2 x (11.1 . Hence the forward di erence operator is inherently unstable while the centered and backward operators are inherently stable. the centered di erence operator ( = 0) produces Re( m ) = 0. NUMERICAL DISSIPATION @u = .(1 + )uj.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 11.1+ uj+1)=2 x and a symmetric component (. For periodic boundary conditions the corresponding matrix operator is . Consider the following point operator for the spatial derivative term .ax . the forward di erence operator ( = . If a is negative. the roles are reversed. 1 If a is positive.ax Bp(. our choice of di erencing scheme produces nonphysical behavior. 2 1 .a @u @t @x (11.a (. the operator is only rst-order accurate.uj. The antisymmetric component is the second-order centered di erence operator. With 6= 0. In either case.1 + uj+1) + (. the solution will either grow or decay with time.1) with periodic boundary conditions.1 + 2uj .2) The second form shown divides the operator into an antisymmetric component (.a x = 2. ) The eigenvalues of this matrix are m = .a x m 1 .204 11. . and the backward di erence operator produces Re( m) < 0.uj.

2 The Modi ed Partial Di erential Equation ( xu)j = 2 1 x 42 x @u .3 is consistent with Eq. 11.({z+ 2 )t}] (11.2 to Eq. 2 . This means that each term in the expansion given by Eq. 2 ) s = .4 of the form u(x t) = ei xe(r+is)t provided r and s satisfy the condition r + is = . 2 ( .a @u + @ 2 u + @ 3 u + @ 4 u (11.4) @t @x @x2 @x3 @x4 Choose periodic boundary conditions and impose an initial condition u = ei x. 11. when we use a computer to nd a numerical solution of the problem. However. 2} ei x. Substituting this into 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.1.1. x can be small but it is not zero.3 as the modi ed partial di erential equation. 11. Notice that Eq. 11.a @u + a x @ 2 u . 11. 11.11. Consider the simple linear partial di erential equation @u = .5) | | amplitude phase . Thus a ) u = e.ia . i 3 + 4 or r = . We are lead to the expression 2 ! 3 ! ! ! @ 2 u + x3 @ 3 u . Under these conditions there is a wave-like solution to Eq. 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.2. @x j 205 First carry out a Taylor series expansion of the terms in Eq.1 gives @u = . We refer to Eq. a x2 @ 3 u + a x3 @ 4 u + : : : (11. 11. We proceed next to investigate the implications of this concept.2.3 is excited to some degree. THE MODIFIED PARTIAL DIFFERENTIAL EQUATION 11. 2 ({z. 11. (a + 2 ) The solution is composed of both amplitude and phase terms. since the two equations are identical when x ! 0.

By examining the term governing the phase of the solution in Eq. The resulting spatial operator is not one-sided but it is dissipative. Any symmetric component in the spatial operator introduces dissipation (or ampli cation). This is tantamount to deliberately adding dissipation to the PDE. this operator produces fourth-order accuracy in phase with a third-order dissipative term.206 CHAPTER 11.2 .1 + 8uj+1 . the coe cients of the even derivatives in Eq.4. the numerical phase speed is dependent upon the wavenumber . 4uj. 4uj+1 + uj+2)] (11. 11. NUMERICAL DISSIPATION It is important to notice that the amplitude of the solution depends only upon and . 11. As a corollary.6) The antisymmetric component of this operator is the fourth-order centered di erence operator. 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). a third-order backward-biased scheme is given by ( xu)j = 6 1 x (uj. For example. by using the antisymmetry of the matrix di erence operators. 6uj.2.5. One can easily show. Furthermore. Under the same condition. 11.1) is equivalent to deliberately adding a destabilizing term to the PDE.1 + 6uj .2 . 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.a x2 =6. The symmetric component approximates x3 uxxxx=12. Referring to the modi ed PDE. Therefore. the phase speed of the numerical solution is less than the actual phase speed. one could choose = 1=2 in Eq. Therefore. Eq. Therefore. the coe cients of the odd derivatives. This we refer to as dispersion. the choice of a forward di erence scheme ( = . If the wave speed a is positive.1 + 3uj + 2uj+1) = 1 (uj. Note that the use of one-sided di erencing schemes is not the only way to introduce dissipation. any departure from antisymmetry in the matrix di erence operator must introduce dissipation (or ampli cation) into the modi ed PDE. 2 > 0 and hence the amplitude of the solution decays.3 we have = . we see that the speed of propagation is a + 2 . and the phase depends only on a and . . 8uj.2 . the choice of a backward di erence scheme ( = 1) produces a modi ed PDE with . that the same is true for any centered di erence approximation of a rst derivative. 11. Biased schemes use more information on one side of the node than the other.

3 The Lax-Wendro Method 207 In order to introduce numerical dissipation using one-sided di erencing. THE LAX-WENDROFF METHOD 11. It is a fully-discrete nite-di erence scheme. 2u(n) + u(n) ) (11. the linear convection equation @u + a @u = 0).9) uj j j +1 j j . There is no intermediate semi-discrete stage. The quantity j ah j is known as the x Courant (or CFL) number. backward di erencing must be used if the wave speed is positive.8) @t @x @t2 @x2 Now replace the space derivatives with three-point centered di erence operators. ah 2 1 4.1 2 x j+1 j. Thus @t @x @ 2 u = a2 @ 2 u @u = .7) First replace the time derivatives with space derivatives according to the PDE (in this case. giving !2 (n+1) = u(n) . This explicit method di ers conceptually from the methods considered previously in which spatial di erencing and time-marching are treated separately. 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 .1 2 x This is the Lax-Wendro method applied to the linear convection equation. Next we consider a method which introduces dissipation independent of the sign of the wave speed. known as the Lax-Wendro method. .3. u(n) ) + 1 ah (u(n) . cos 2 m . 1 ah (u(n) . 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. x M x M for m = 0 1 : : : M .11. and forward di erencing must be used if the wave speed is negative. i ah sin 2 m m =1.a @u (11. For periodic boundary conditions. 1 For j ah j 1 all of the eigenvalues have modulus less than or equal to unity and hence x the method is stable independent of the sign of a. It is equal to the ratio of the distance travelled by a wave in one time step to the mesh spacing. the corresponding fully-discrete matrix operator is 0 2 !3 ! 2 ! 31 1 4 ah + ah 25 1 .

ah (u(jn) . which is given by @u + a @u = . a2 h2) @ 4 u + : : : @t @x 6 @x3 8 @x4 This is derived by substituting Taylor series expansions for all terms in Eq.1 a dissipative second-order method is obtained. 11. these should be applied alternately. The two leading error terms appear on the right side of the equation. a x2 (1 . Therefore. a2h2 ) @ u = . A closely related method is that of MacCormack. a2 h ( x2 .208 CHAPTER 11. presented in Chapter 6: un+1 = un + hu0n ~ 1 ~ ~ (11. C 2 ) @ 3 u . a2h2 ) @ 3 u . The two methods di er when applied to nonlinear hyperbolic systems. u(jn+1))] ~ (11.11) 2 x u +1 ~ which can be shown to be identical to the Lax-Wendro method. a8h ( x2 . a h8 x (1 .6 n @x3 @x3 6 The dissipative term is proportional to 2 4 2 2 4 2 @ . a2h2 ) @ 3 u = . u(jn)1) ~ . For the linear convection equation. Hence MacCormack's method has the same dissipative and dispersive properties as the Lax-Wendro method. 1 Or vice-versa for nonlinear problems.8. 11. this approach leads to u(jn+1) = u(jn) . the leading error term in the Lax-Wendro method is dispersive and proportional to a ( x2 .10) un+1 = 2 un + un+1 + hu0n+1] If we use rst-order backward di erencing in the rst stage and rst-order forward di erencing in the second stage. Recall that the odd derivatives on the right side lead to unwanted dispersion and the even derivatives lead to dissipation (or ampli cation.9 and converting the time derivatives to space derivatives using Eq. 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. however. . Recall MacCormack's timemarching method. a ( x2 . depending on the sign). Cn) @xu 4 @x4 This term has the appropriate sign and hence the scheme is truly dissipative as long as Cn 1. ah (~(jn+1) . x u(jn+1) = 1 u(jn) + u(jn+1) .

This is the basic concept behind upwind methods.1x a(. 11. manner. These are by no means unique. With this approach. but entirely equivalent. the wave speeds can be both positive and negative.1.1 + 2uj .11. uj+1)] (11. For example.2. = 0.a( xu)j = 2.uj.13) This approach is extended to systems of equations in Section 11. we present two splitting techniques commonly used with upwind methods. Now for the a+ ( 0) term we can safely backward di erence and for the a.1 if a 0. Alternatively. In the next two sections. 11.12 as @u + (a+ + a. In order to apply one-sided di erencing schemes to such systems. these are of mixed sign. then a+ = 0 and If a 0. the eigenvalues of the ux Jacobian for the onedimensional Euler equations are u u + a u . ) @u = 0 a = a 2 jaj @t @x + = a 0 and a. the direction of the one-sided operator (i. When a hyperbolic system of equations is being solved. schemes in which the numerical dissipation is introduced in the spatial operator are generally preferred over the Lax-Wendro approach. as a result of their superior exibility.uj.1 + uj+1) + jaj(. by adding a symmetric component to the spatial operator. 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. if a 0. we present the basic ideas of ux-vector and ux-di erence splitting. a. that is.5. This is achieved by the following point operator: . The above approach to obtaining a stable discretization independent of the sign of a can be written in a di erent. From Eq. For more subtle aspects of implementation and application of such techniques to . more generally. When the ow is subsonic. In this section. 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. This is avoided in the Lax-Wendro scheme. UPWIND SCHEMES 11..4 Upwind Schemes 209 In Section 11. ( 0) term forward di erence. some form of splitting is required. we saw that numerical dissipation can be introduced in the spatial di erence operator using one-sided di erence schemes or. However.4. = a 0. Consider again the linear convection equation: @u + a @u = 0 (11. We can rewrite Eq. we see that a stable discretization is obtained with = 1 if a 0 and with = .e.12) @t @x where we do not make any assumptions as to the sign of a. then a a.

the reader is referred to the literature on this subject. @w = 0 (11. @w term.X . A.15) i @t @x where the wave speeds.1 Flux-Vector Splitting Recall from Section 2. into two components such that = ++ .j j (11. is positive.16) +j j .4. let us split the matrix of eigenvalues. We can now rewrite the system in terms of characteristic variables as @w + @w = @w + + @w + . 11. and a forward di erence if the wave speed is negative.14) @t @x @t @x can be decoupled into characteristic equations of the form @wi + @wi = 0 (11. and the wi's are the characteristic variables. To accomplish this. NUMERICAL DISSIPATION nonlinear hyperbolic systems such as the Euler equations. In order to apply a one-sided (or biased) spatial di erencing scheme. + contains the positive eigenvalues and .19) .1X in the spatial terms gives @Xw + @X +X .210 CHAPTER 11.5 that a linear. = . Premultiplying by X and inserting the product @x X . we need to apply a backward di erence if the wave speed.17) 2 2 With these de nitions.18) @t @x @t @x @x The spatial terms have been split into two components according to the sign of the wave speeds. . are the eigenvalues of the Jacobian matrix. constant-coe cient. We can use backward di erencing for the + @w term and forward @x di erencing for the . i. hyperbolic system of partial di erential equations given by @u + @f = @u + A @u = 0 (11. where += (11.1Xw + @X . contains the negative eigenvalues.1Xw = 0 @t @x @x (11. i.

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

The centered approximation to gj+1=2.31) 2 Now. gj+1=2. constant-coe cient.27) @t @x The ux vector associated with this form is g = w.4.1X (wL + wR) + 2 X j jX . as in Eq.28) In order to obtain a one-sided upwind approximation. We again begin with the diagonalized form of the linear. This is achieved with 1 1 (^i)j+1=2 = 2 i (wi)L + (wi)R ] + 2 j ij (wi)L . (wi)R] g (11.1 (11. which is nondissipative.1X after and j j to obtain 1 1 X gj+1=2 = 2 X X . 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. we consider the numerical ux at the interface between nodes j and j + 1.19. more suited to nite-volume methods.30) or 1 gj+1=2 = 1 (wL + wR) + 2 j j (wL . wR ) ^ (11. Now. respectively. we require ( (w ) (^i)j+1=2 = i(wi)L if i > 0 g (11. the uxes must be evaluated at cell boundaries.33) 2 where jAj = X j jX .34) .2 Flux-Di erence Splitting Another approach. 11.1X (wL . In a nite-volume method. uR) (11.29) i i R if i < 0 where the subscript i indicates individual components of w and g. is given by 1 gj+1=2 = 2 (g(wL) + g(wR)) ^ (11. as a ^ function of the states to the left and right of the interface. wR ) ^ (11. wL and wR. hyperbolic system of equations @w + @w = 0 (11.212 CHAPTER 11.32) and thus 1 f^j+1=2 = 1 (fL + fR ) + 2 jAj (uL . as in Chapter 5.

jAj = . Eq. fR = Aj+1=2 (uL . In order to resolve this. 11.35 is obtained by the de nition 1 f^j+1=2 = 2 (fL + fR ) + 1 jAj+1=2j (uL . However. ARTIFICIAL DISSIPATION 213 and we have also used the relations f = Xg. Hence satisfaction of Eq.35) R if all i s < 0 giving pure upwinding. uR) (11.11.37) For the Euler equations for a perfect gas.1. uR) (11. In this case.3 ( 11. 11.5 Arti cial Dissipation We have seen that numerical dissipation can be introduced by using one-sided differencing schemes together with some form of ux splitting.37 is satis ed by the ux Jacobian evaluated at the Roe-average state given by p LuL + p R uR (11.39) p L+p R R R where u and H = (e + p)= are the velocity and the total enthalpy per unit mass. If the eigenvalues of A are all positive. We have also seen that such dissipation can be introduced by adding a symmetric component to an antisymmetric (dissipation-free) operator.5. In the linear. respectively. in the nonlinear case. and A = X X .36) 2 if Aj+1=2 satis es fL . constant-coe cient case.38) uj+1=2 = p + p L R p H +p H L L Hj+1=2 = (11. consider a situation in which the eigenvalues of A are all of the same sign. of this chapter. jAj = A if they are all negative. 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. 3 Note that the ux Jacobian can be written in terms of u and H only see problem 6 at the end . u = Xw. 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. this leads to an upwind operator which is identical to that obtained using ux-vector splitting. there is some ambiguity regarding the de nition of jAj at the cell interface j + 1=2.A.

NUMERICAL DISSIPATION s a ( xu)j = .1 + 6uj . 11.214 For example. With an appropriate value of .40) ( x u)j = uj+1 . applying x = x . The second spatial term is known as arti cial dissipation.1 + 3uj ) 2 x .1 (11.36 and 11.uj+1 + 2uj . this approach can be related to the upwind approach. uj.43) x f = x f + x (jAju) where jAj is de ned in Eq. gives a s (11.43. 11. uj. let CHAPTER 11. It is also sometimes referred to as arti cial di usion or arti cial viscosity.2 . 4uj.41) i x = i x + j ij x Extension to a hyperbolic system by applying the above approach to the characteristic variables. 11.44) x (uj. A second-order backward di erence approximation to a 1st derivative is given as a point operator by ( xu)j = 1 (uj. s ( xu)j = 11.34. s It is common to use the following operator for x (11. as in the previous two sections. this often provides su cient damping of high frequency modes without greatly a ecting the low frequency modes. the reader should consult the literature. This is particularly evident from a comparison of Eqs.2 . Similarly. A more complicated treatment of the numerical dissipation is also required near shock waves and other discontinuities.42) x (Au) = x (Au) + x (jAju) or a s (11.15 is stable if i 0 and a s unstable if i < 0. 4uj.6 Problems 1. 4uj+1 + uj+2) where is a problem-dependent coe cient. but is beyond the scope of this book. For details of how this can be implemented for nonlinear hyperbolic systems. a s With appropriate choices of x and x. x is stable if i 0 and unstable if i > 0. The appropriate implementation is thus a s (11. This symmetric operator approximates x3 uxxxx and thus introduces a third-order dissipative term.1 2 x 2 x a s Applying x = x + x to the spatial derivative in Eq.

Find the matrix jAj. 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. 2. x for 0 x . nd j j for the combination of this spatial operator with 4th-order Runge-Kutta time marching at a Courant number of unity and plot vs.2. Plot the real and imaginary parts of x vs. 4. 11. Find the modi ed wavenumber for this operator with = 0 1=12 1=24. Consider the hyperbolic system derived in problem 8 of Chapter 2. 6. 3. 11. Using Fourier analysis as in Section 6. Find the modi ed wavenumber for the operator given in Eq.2.6.38 and 11. x for 0 x . nd j j for the combination of this spatial operator with 4th-order Runge-Kutta time marching at a Courant number of unity and plot vs. 11.4.6.37. nd the derivative which is approximated by the corresponding symmetric and skew-symmetric operators and the leading error term for each. Consider the spatial operator obtained by combining second-order centered differences with the symmetric operator given in Eq.6.3 to see how to construct the symmetric and skew-symmetric components of a matrix. PROBLEMS 215 (a) Express this operator in banded matrix form (for periodic boundary conditions). Show that the use of the Roe average state given in Eqs. then derive the symmetric and skew-symmetric matrices that have the matrix operator as their sum. x for 0 x . 11. .11.2. Plot the real and imaginary parts of x vs. Show that the ux Jacobian for the 1-D Euler equations can be written in terms of u and H . Form the plus-minus split ux vectors as in Section 11. 5. Using Fourier analysis as in Section 6. x for 0 x .) (b) Using a Taylor table. (See Appendix A.6.6. and 1=48. x for 0 x .44. Find the modi ed wavenumber for the rst-order backward di erence operator.1. Plot the real and imaginary parts of x vs. x for 0 x .39 leads to satisfaction of Eq.

NUMERICAL DISSIPATION .216 CHAPTER 11.

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

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

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. Since numerical sti ness is generally much more severe for the di usion process. We have yet to determine its . the important aspect of stability has yet to be discussed. another way to approximate Eq.8) Factoring can also be useful to form split combinations of implicit and explicit techniques. 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 . 12. However. stability.7 are often applied in a predictor-corrector sequence. = I + hAd + h Ad + if h jjAdjj < 1.6 have identical orders of accuracy in the time approximation. I + hAc]~ n + h(bc) u ~ I + hAd + hAc]~ n + h(bc) +O(h ) u | {z } 1 2 (12.8. requiring a tridiagonal solver if the di usion term is central di erenced.10) The convection operator is applied explicitly. This time a predictor-corrector interpretation leads to the sequence un ~ I . where jjAd jj is some norm of Ad ].7) u {z } | 2 2 | {z } un ~ ~n u +1 +1 = = ~ I + hAc]~ n + h(bc) u I + hAd]~n u +1 (12. hAd ].9) I . and a rst-order time-march method is usually unsatisfactory. hAd]. In practical applications equations such as 12. 12. on this basis. Therefore. 2 We do not suggest that this particular method is suitable for use. their selection is arbitrary. For example.12.2. 12. hAd]~ n u +1 +1 ~ = I + hAc]~ n + h(bc) u = un ~ +1 (12. 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.7 and the original unfactored form Eq. as before. but the di usion operator is now implicit. this factored form would appear to be superior to that provided by Eq.

10. 12.12. 12.3.1 Mesh Indexing Convention 2 2 12.and three-dimensional ows. 3 32 @u = @ u + @ u @t @x @y 2 2 (12.6 in the form Then which is identical to Eq.9 can be derived for a di erent point of view by writing Eq. but in these notes we describe the size of a mesh by the number of interior points. 3 This could also be called a 5 . hAd ]un = I + hAc]un + h(bc) +1 12. 6 point mesh if the boundary points (labeled in the sketch) were included. We choose the 3 4 point mesh shown in the Sketch 12.3 Factoring Space Matrix Operators in 2{D Factoring is widely used in codes designed for the numerical solution of equations governing unsteady two. un = A u + A u + (bc) + O(h ) ~ c n d n h +1 +1 2 We should mention here that Eq. 12. is 4 and My . SPLIT AND FACTORED FORMS un .11) My k 1 13 23 33 43 12 22 32 42 11 21 31 41 1 (12.12) j Mx Mesh indexing in 2-D. the number of (interior) y points is 3.and 3-D requires some reference to a mesh.220 CHAPTER 12. ~ I . The numbers 11 12 43 represent the location in the mesh of the dependent variable bearing that index. 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. A clear description of a matrix nite-di erence operator in 2. the number of (interior) x points. In this example Mx . Thus u represents the value of u at j = 3 and k = 2.

we use the notation Axx y or Axy y . There are many ways to lay out a data-base. ( ) ( ) 12. used in the previous sections. respectively. Examples are in Eq. we consider only two: (1).13) where Pyx = PT = P. Thus ( ) + ( ) + ( ) ( ) dU = A U + (bc) ~ x y dt + (12.3. Examples of the order of indexing for these space vectors are given in Eq.12. (1) and (2) above are referred to as xvectors and y-vectors.15) Ax+y and U . 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.3. which are notations used in the special case of space vectors. 12.2 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.16 part a and b. 12. In particular.16 part a and b. depending on the data{base chosen for the U it multiplies. Notice that the matrices are not the same although they represent the same derivative operation. and they are related by the same permutation matrix that relates U x to U y . FACTORING SPACE MATRIX OPERATORS IN 2{D 221 12. and (2). the usual (but ambiguous) representation is given by Ax y . 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 . and one composed of y-vectors with U y . Of these. consecutively along rows that are themselves consecutive from k = 1 to My . Their structures are similar. 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. consecutively along columns that are consecutive from j = 1 to Mx. In this notation the ODE form of Eq. 12. however. When the matrix is multiplying a space vector U .11 can be written + 4 If it is important to be speci c about the data-base structure. u 4 Notice that .3 Data Base Permutations U x = Pxy U y ( ) ( ) The two vectors (arrays) are related by a permutation matrix P such that and U y = PyxU x ( ) 1 ( ) (12. To be speci c about the structure.14) Axx y = Pxy Axy y Pyx ( ) + ( ) + (12.3. we label a data{base composed of x-vectors with U x . are subsets of A and ~ .

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

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

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. We arrive at the expression h 1 ~ i (12. and ~ let the (bc) be time invariant: 1 1 1 ~ I . and matrix multiplications do not commute. 12. 12. 2 hAy Un = I + 2 hAx + 1 hAy Un + h(bc) + O(h ) 2 From both sides subtract 1 I . Consider the unfactored form of the trapezoidal method given by Eq. if Eq. Inspecting the top of Eq. The rst step would be solved using My uncoupled block tridiagonal solvers . our original ODE. 2 hAx . we see that this is equivalent to solving My one-dimensional ~ problems. 12. for ensuring a correct steady-state solution in a converged time-march. 9 A block tridiagonal solver is similar to a scalar solver except that small block matrix operations . each with Mx blocks of order Ne . SPLIT AND FACTORED FORMS 9 which has the same appearance as Eq.228 CHAPTER 12.6 The Delta Form Clearly many ways can be devised to split the matrices and generate factored forms.21 but is second-order time accurate. using the standard de nition of the di erence operator .23 shows that the nal ~ be permuted to U step consists of solving Mx one-dimensional implicit problems each with dimension My . 12. Un = Un .29 and maintain O(h ) accuracy. Then. it is guaranteed to converge to the correct steady-state solution of the ODE. +1 3 +1 + 3 2 + 3 replace the scalar ones. 2-D equation. 1 hAy Un = h Ax y Un + (bc) + O(h ) 2 This is the delta form of a factored. 2nd-order.22. Thus. 2 hAx I . One way that is especially useful.24.14. 2 hAy Un = h Ax y Un + (bc) + O(h ) (12. 12. 12. The temporary solution U x would then y and an inspection of the bottom of Eq. Un one nds h 1 1 ~ i I . is referred to as the \delta form" and we develop it next.30) I . 12. ( ) ( ) 12. Now we can factor Eq. 2 hAx .29 converges. 1 hAy Un 2 leaving the equality unchanged. 2 hAx .

hAx] I . Write down the permutation matrices. hAy ] Un = h Ax y Un + (bc) + 10 and its factored form becomes h ~ i I . the factored form is presented in Eq. u ? Only include the interior points. Assume that second order central di erencing is used.31 are vastly di erent. 12. hAy ] Un = h Ax y Un + (bc) + (12.e. we will see in the next chapter that the convergence properties of Eq. u ? iii.30 and the O(h) method given by Eq. write the space vector. 12.12. 12. i. 2u + u ) xx j The uniform grid has x = 1 and 8 interior points.7 Problems 1. What is the space vector for the natural ordering (monotonically increasing in index). 12. hAx .19. 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. (1) (2) 12 21 x 2 j . If we reorder the points with the odd points rst and then the even points. ii.(P . but it can have a profound e ect on the stability and convergence properties of a method.P ).19 is h ~ i I . PROBLEMS 229 The point at which the factoring is made may not a ect the order of time-accuracy. 12.. On the other hand.20. (a) Space vector de nition i. the delta form of the unfactored Eq.20 and Eq. ( u) = 1 (u .7. the unfactored form of a rst-order method derived from the implicit Euler time march is given by Eq. and if it is immediately factored. 12. so that we can simplify the boundary conditions.1 j j +1 10 Notice that the only di erence between the O(h2 ) method given by Eq. 12.31 is the appearance of the factor 1 on the left side of the O(h2 ) method. For example.31) In spite of the similarities in derivation. 12. 2 .

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

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

SPLIT AND FACTORED FORMS .232 CHAPTER 12.

7.3. 13. accuracy. and (approximately) factored. When these methods are applied to practical problems. and used it in Chapter 7 to study the stability. fractional-step. if the results indicate that a method has an instability. However. the method is probably not suitable for practical use. 1 See also the discussion on Fourier stability analysis in Section 7. In this and the following section we assume circulant systems and our analysis depends critically on the fact that all circulant matrices commute and have a common set of eigenvectors.Chapter 13 LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS In Section 4. hybrid. 233 . We have seen that under these conditions a semi-discrete approach can lead to circulant matrix di erence operators. and convergence properties of timemarching schemes. The analysis in this chapter is useful for estimating the stability and steady-state properties of a wide variety of time-marching schemes that are variously referred to as time-split. and we discussed circulant eigensystems1 in Section 4. the results found from this analysis are neither necessary nor su cient to guarantee stability.4 we introduced the concept of the representative equation.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. The question is: Can we nd a similar equation that will allow us to evaluate the stability and convergence properties of split and factored schemes? The answer is yes | for certain forms of linear model equations.

.1) apu b pu f dt where the subscript p denotes a circulant matrix. ~ (t) u (13. gm(t) .234 CHAPTER 13. 0 wm = ( a + b)m wm . Since both matrices have the same set of eigenvectors.3) that 13.2 Example Analysis of Circulant Systems 13. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS Suppose.2. g1 (t) ..2.4) This is to be contrasted to the developments found later in the analysis of 2-D equations.2) The analytic solution of the m'th line is wm(t) = cm e( a+ b)mt + P:S: Note that each a pairs with one. we can use the arguments made in Section 4.4: b since they must share a common The representative equation for split. we arrive at a set of ODE's that can be written d~ = A ~ + A ~ .3 to uncouple the set and form the M set of independent equations 0 w1 = ( a + b)1 w1 .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. 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. gM (t) (13. and only one2 .. . This suggests (see Section 4. as a result of space di erencing the PDE. eigenvector. 0 wM = ( a + b)M wM .. (13.

Eq. h d )E . 13.8. In this section we refer to these as 1. (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. we can analyze the stability of the two forms of simple time-splitting discussed in Section 12. Using these values and the representative equation 13.13. a B (. Eq.6) ( d)m = . respectively.2): ( c)m = ia sin m x m (13. 2. 1 .2. 12.4. the characteristic polynomial of this method is P (E ) = (1 . 13. The Explicit-Implicit Method 2 (13. EXAMPLE ANALYSIS OF CIRCULANT SYSTEMS 235 If the space di erencing takes the form d~ = .2.2 1)~ u (13.3. Now introduce the dimensionless numbers Cn = ah Courant number x mesh Reynolds number R = a x and we can write for the absolute value of q 2 1 + Cn sin2 m j j= C m 1 + 4 R n sin2 2 0 m 1.1 0 1)~ + u u B (1 .4. 12. so that 0 m 2 . the explicit-implicit Euler method. the explicit-explicit Euler method.10. can be represented by the eigenvalues c and d.5) dt 2 x p x2 p the convection matrix operator and the di usion matrix operator.7) .6 to quantify the eigenvalues. 4 2 sin2 2 x In these equations m = 2m =M . When applied to Eq. where (see Section 4. m = 0 1 M .

2 C = 2/ R ∆ n 0.1.4 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. and the . 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.236 CHAPTER 13.2 C = R /2 n ∆ C = 2/ R n ∆ 0. which yields the same result as in the previous example. 2. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 1. The Explicit-Explicit Method An analysis similar to the one given above shows that this method produces " # q 2 m 1 . 13. one near 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. 13.8 C n C 0.7 shows that the critical range of m for any combination of Cn and R occurs when m is near 0 (or 2 ).4 0.

2 Analysis of a Second-Order Time-Split Method .1. thus 1 1 un+1 = un + 2 h c(3un . time-marching method on the equation u0 = cu + du + ae t First let us nd expressions for the two polynomials. The characteristic polynomial follows from the application of the method to the homogeneous equation. un. 1 h d)E 2 . Next let us analyze a more practical method that has been used in serious computational analysis of turbulent ows. Our model equation is again the linear convection-di usion equation 13. we include the forcing function in the representative equation and study the e ect of our hybrid.8) and in the latter (13. This method applies to a ow in which there is a combination of di usion and periodic convection. The di usion term is integrated implicitly using the trapezoidal method. P (E ) and Q(E ). 1) (13. factored method has a much smaller region of stability when R is small. In order to evaluate the accuracy.13.9) Q(E ) = 1 h(E 2 + E ) 2 13. as we should have expected. The totaly explicit.2. which produces the constraint that 1 Cn < 2 R for R 2 The resulting stability boundary is also shown in Fig. 13. (1 + 2 h c + 1 h d)E + 1 h c 2 2 2 The form of the particular polynomial depends upon whether the forcing function is carried by the AB2 method or by the trapezoidal method. In the former case it is 1 Q(E ) = 2 h(3E . 13.1) + 2 h d (un+1 + un) This produces 3 P (E ) = (1 .2. The convection term is treated explicitly using the second-order Adams-Bashforth method. EXAMPLE ANALYSIS OF CIRCULANT SYSTEMS 237 other near 180o.5.4 which we split in the fashion of Eq. as well as the stability.

The results are plotted in Fig. 1 1 + 3h c + 2h 2 2 1 .7.2. 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 . 12.3. so er = O(h3) Using P (e h) and Q(e h ) to evaluate er in Section 6. 2 h d (13.10 and shown in Fig. 2h c 1 . 13. 1h d 2 d 2 1 .1. 13. 13.10 by a parametric study of cn and R de ned in Eq. 13.11) The extension of the following to 3-D is simple and straightforward.9.8 or Eq. one can show er = O(h3) using either Eq.1. This was carried out in a manner similar to that used to nd the stability boundary of the rst-order explicit-implicit method in Section 13.6.10) Stability The stability of the method can be found from Eq.3 The Representative Equation for Space-Split Operators Consider the 2-D model3 equations @u = @ 2 u + @ 2 u @t @x2 @y2 3 (13. 13. 13. .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 . the hybrid method retains the second-order accuracy of its individual components.238 CHAPTER 13. 2 d . These results show that. 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. 13.

12) @t x @x y @y Reduce either of these.4 Stable 0 4 8 ∆ R _ Figure 13.1 I ~0 I B b b where B is a banded matrix of the form B (b. to the coupled set of ODE's: dU = A + A ]U + (bc) ~ (13.12.2: Stability regions for the second-order time-split method. by means of spatial di erencing approximations. 12.13) x y dt for the space vector U .1 b0 b1 ). First we write these matrices in the form 2 3 2 ~ 3 B b0 I ~1 I b A(xx) = 6 B 7 A(yx) = 6 ~.8 Cn 0.22 and 12. and @u + a @u + a @u = 0 (13.1 I ~0 I ~1 I 7 b b 5 4 5 4b ~. Now nd the block eigenvector .2 0.3. THE REPRESENTATIVE EQUATION FOR SPACE-SPLIT OPERATORS239 1.23 for the 3 4 mesh shown in Sketch 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.13. The form of the Ax and Ay matrices for three-point central di erencing schemes are shown in Eqs.

1 A(xx+)y X Pxy X = 2 3 1I + ~ 6 7 6 7 2I + ~ 6 7 (13.23.1 X B X where 2 3 6 =6 6 4 3 7 5 1 2 set X 4 (x) is transparent to this transformation. by the same argument as before. 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 .13. b b b ~ that diagonalize the B blocks.14) ~ 4 5 4 B 7 3 I 5 ~ B 4 I ~ where B is the banded tridiagonal matrix B (~.1 32 32 3 2 X B X 6 76 7=6 X . 12. if we Notice that the matrix Ay 3 7 7 7 5 3 2 ~ 3 ~1 I b b0 I ~1 I b ~0 I ~1 I 7 X = 6 ~.1 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. so the nal result is the complete diagonalization of the matrix Ax+y : h ih ih i ~ ~ X .1 Pyx X .1 4b I I .1 6 ~. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS matrix that diagonalizes B and use it to diagonalize A(xx).240 CHAPTER 13.1 I ~0 I ~.14 is transparent to the transformation.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. That is. 12. Thus 2 . 13.15) 6 7 4 5 3I + ~ 4I + ~ diag(X ) 2 ~ b0 X . the rst matrix on the right side of Eq.1 B 76 X 4 54 54 5 4 .1B X = 6 7 4 5 4 5 ~3 ~ This time. see the bottom of Eq.1 ~0 ~1). There results h i Pyx X .1 I ~0 I ~1 I 7 b b 5 b b 5 4b ~.

This results in a spatially split A matrix formed from the subsets A(xx) = diag(B ) ~ A(yy) = diag(B ) (13. THE REPRESENTATIVE EQUATION FOR SPACE-SPLIT OPERATORS241 It is important to notice that: The diagonal matrix on the right side of Eq. and the convergence rate to. by itself.17) x y dt which has the exact solution a u(t) = ce( x+ y )t .13. a and are (possibly complex) constants. Often we are interested in nding the value of. Now we are ready to present the representative equation for two dimensional systems.3. Although Ax and Ay do commute. 13. To obtain the latter property the structure of ~ the matrices is important.15. but this choice was for convenience only. 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 .18) x y We have used 3-point central di erencing in our example.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). the steady-state solution of the representative equation. 4 . and its use is not necessary to arrive at Eq. 13. this fact. In that case we set = 0 and use the simpler form du = + ]u + a (13.16. does not ensure the property of \all possible combinations".15 contains every possible com~ bination of the individual eigenvalues of B and B . + (13. and where Ax and Ay satisfy the conditions in 13. First reduce the PDE to ODE by some choice4 of space di erencing.

242 CHAPTER 13.4. however. " 1 #n a x+ y 2. 3. Produces the exact (see Eq. 12.5. 2. steady-state solution. h y )E . h y )un+1 = un + ha P (E ) = (1 . and when it is applied to the representative equation. The form of the method is given by Eq. The stability. The accuracy of the xed. . h from which x . use of this method for 2-D problems is generally impractical for reasons discussed in Section 12.4 Example Analysis of 2-D Model Equations 1.16. Converges very rapidly to the steady-state when h is large. steady-state solution to the 2-D representative equation 13. h . In each case we examine 13. rst-order scheme which can then be used as a reference case for comparison with the various factored ones. 3. Is unconditionally stable. 13. h . The convergence rate to reach the steady-state.18) steady-state solution (of the ODE) for any h. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 13. this method: 1.19) un = c 1 . we nd (1 . x y Like its counterpart in the 1-D case. In the following we analyze four di erent methods for nding a xed. 1 Q(E ) = h giving the solution (13.1 The Unfactored Implicit Euler Method Consider rst this unfactored.19. Unfortunately. h x .

h y )(un+1 . The method requires far less storage then the unfactored form.20 to the 2-D representative equation. One nds (1 . x y We see that this method: 1. There results (1 .2 The Factored Nondelta Form of the Implicit Euler Method 13. h x)(1 . h )(1 . 12. + x y x y This method: 13. h y )E . However. h x)(1 . it is not very useful since its transient solution is only rst-order accurate and.20) giving the solution " #n 1 un = c (1 . Is unconditionally stable.13. 12. h y )un+1 = 1 + h2 x y un + ha and this has the solution " #n 1 + h2 x y a un = c (1 .4. h )(1 . h ) . the converged value is meaningless.4.4.31 to the 2-D representative equation. h x)(1 . 3. Next apply Eq. 2. if one tries to take advantage of its rapid convergence rate.3 The Factored Delta Form of the Implicit Euler Method . h x)(1 . Produces a steady state solution that depends on the choice of h. 1 Q(E ) = h (13. but the converged solution is completely wrong. un) = h( xun + y un + a) which reduces to (1 . h ) . + a h x y x y. EXAMPLE ANALYSIS OF 2-D MODEL EQUATIONS 243 Now apply the factored Euler method given by Eq. h y )un+1 = un + ha from which P (E ) = (1 . Converges rapidly to a steady-state for large h.

All of these properties are identical to those found for the factored delta form of the implicit Euler method. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 1. Converges very slowly to the steady{state solution for large values of h. 2 h x 1 . Is unconditionally stable. since j j ! 1 as h ! 1. Since it is second order in time. Produces the exact steady{state solution for any choice of h. it can be used when time accuracy is desired.30 to the representative equation and one nds 1 1 .244 CHAPTER 13. 3. un) = h( xun + y un + a) 2 which reduces to 1 . Produces the exact steady-state solution for any choice of h.4. Like the factored nondelta form. The correct steady solution is obtained.5. 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 . as discussed in Section 12.4 The Factored Delta Form of the Trapezoidal Method . 3.27 should be enough to convince the reader that the 's produced by those methods are identical to the produced by this method. Apply Eq. but convergence is not nearly as rapid as that of the unfactored form. Converges very slowly to the steady{state solution for large values of h. Finally consider the delta form of a second-order time-accurate method. 1 h x 1 .26 and 12. and the factored delta form of the implicit Euler method can be used when a converged steady-state is all that is required. 1h x 1 . 2. 2 5 13. 2h y 2 This method: 1. 2. the value of h on the left side of the implicit equation is literally switched from h to 1 h. Is unconditionally stable. this method demands far less storage than the unfactored form. 12. since j j ! 1 as h ! 1. In practical codes.5 A brief inspection of eqs. 1 h y (un+1 . 12. un = c6 4 5 1 x+ y 1 .

1 h y 1 .22.21) x y z dt Let us analyze a 2nd-order accurate. 1 h z un = h ( x + y + z )un + a] (13. the model 3-D cases6 have the following representative equation (with = 0): du = + + ]u + a (13. 2 h x 1 .5 Example Analysis of the 3-D Model Equation 245 The arguments in Section 13. 8 x y z . . 2 x y z 4 x y + x z + y z) . +a+ x y 6 z (13. each with an additional term. delta form using this equation. under the conditions given in 13. EXAMPLE ANALYSIS OF THE 3-D MODEL EQUATION 13.13.5.11 and 13. 1 h( x + y + z ) un = h ( x + y + z )un + a] 2 Now factor the left side: 1 1 . nd the root.16 with an A(zz) included.23) Eqs. 2 h( x + y + z ) un+1 = 1 + 2 h( x + y + z ) un + ha Put this in delta form: 1 . and write the solution either in the form 2 3n 1 h( + + ) + 1 h2 ( 1 h3 + + ).12.22) 2 2 This preserves second order accuracy since the error terms 1 h2( 1 h3 x y + x z + y z ) un and 4 8 x y z are both O(h3). One can derive the characteristic polynomial for Eq. factored.3 generalize to three dimensions and. 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 . 13. 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 . 13.

if all the 's are real and negative. This makes the method unconditionally stable for the 3-D di usion model when it is centrally di erenced in space. the instability disclosed above is too weak to cause trouble. Now consider what happens when we apply this method to the biconvection model. it converges to the proper steady-state. the 3-D form of Eq. Now we can always nd one combination of the 's for which . . 2 y 1 . )2 >1 and the method is unconditionally unstable for the model convection problem. 4 h3 2 2 1h 1h 1 . 2 x 1 . )2 + ( .7 With regards to stability. In practical cases. and are both positive. 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. Remember that in our analysis we must consider every possible combination of these eigenvalues. 13. ) + ( + ) j (1 . 1h z 2 3n x y z7 7 .12 with periodic boundary conditions. 7 However. if the method converges. 13. we already knew this because we chose the delta form.24 results in 1 = 1 + h( x + y + z ) + 2 h2( x + y + z )2 (13.24) It is interesting to notice that a Taylor series expansion of Eq. First write the root in Eq. From the above analysis one would come to the conclusion that the method represented by Eq. In that case since the absolute value of the product is the product of the absolute values j 2 2 2 = (1 . central di erencing causes all of the 's to be imaginary with spectrums that include both positive and negative values. +i where .23 that.i . In this case. 13. it follows from Eq. the method is stable for all h.246 CHAPTER 13.25) h i + 1 h3 3 + (2 y + 2 x) + 2 2 + 3 x y + 2 2 + 3 + 2 x 2 + 2 2 y + 3 + y y y y x x 4 z which veri es the second order accuracy of the factored form. in the presence of this dissipation. Furthermore. 13. 5 x+ y+ z a (13. and are real numbers that can have any sign. however. clearly.22 should not be used for the 3-D Euler equations.23 in the form 1+ = 1. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS or in the form 2 6 un = c6 4 1 1 1 + 1 h x 1 + 2 h y 1 + 1 h z . +i i .

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

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

2m 7 7 6 5 4 am1 am2 amm 249 . Given below is some notation and some of the important relations between matrices and vectors. A.. In the present context a vector is a vertical column or string. . 5 vm and its transpose ~ is the horizontal row vT ~ T = v1 v2 v3 : : : vm] ~ = v1 v2 v3 : : : vm ]T v v 2. Thus 2 v1 3 6 7 ~ = 6 v. A general m m matrix A can be written 2 a11 a12 a1m 3 6a a a 7 A = (aij ) = 6 21 22 .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. .1 Notation 1.2 7 v 6 7 4 .

The trace of a matrix is Pi aii. AH is the conjugate transpose of A. A is the complex conjugate of A. v v 7. (Hermitian). The inverse of a matrix (if it exists) is written A.1A = AA.b a . A is normal if AT A = AAT . det A] is the determinant of A. 4. where I is the identity matrix. A is orthogonal if aij are real and AT A = AAT = I 5. A is skew-symmetric or antisymmetric if AT = . 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 .A.1 and has the property that A.c . If b A= a d c then det A] = ad .2 De nitions 1. 10.1 = I .250APPENDIX A. 9. 7 4 T5 ~m a 4. 2. 6. A.2 7 6 7 6 . USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA 3. P is a permutation matrix if P ~ is a simple reordering of ~ . 8. 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. bc and d A. 3. 11..1 = det1 A] . A is symmetric if AT = A.

1. 5. 2.1 = B . they are all imaginary. Transpose equalities: (A + B )T = AT + B T (AT )T = A (AB )T = B T AT 6.1 A. .1 )T 251 7. A + (B + C ) = (C + A) + B .A. 4. B ]~ = 0 x x x 2. ALGEBRA A. The eigenvalue problem for a matrix A is de ned as A~ = ~ or A . I ]~ = 0 x x x and the generalized eigenvalue problem.1). including the matrix B . as A~ = B~ or A . Any matrix A can be expressed as the sum of a symmetric and a skew-symmetric matrix. 3. If a square matrix with real elements is symmetric.3. If it is asymmetric. sA = (saij ) where s is a scalar.1 (AT ).3 Algebra We consider only square matrices of the same dimension.1 = A (AB ). AT 2 A. its eigenvalues are all real. In general AB 6= BA. A and B are equal if aij = bij for all i j = 1 2 : : : m. etc.4 Eigensystems 1. Inverse equalities (if the inverse exists): (A.1 = (A. Thus: 1 A = 2 A + AT + 1 A .

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

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

5 Vector and Matrix Norms 1. having: 9 eigenvalues 3 distinct eigenvalues 3 Jordan blocks 5 linearly independent eigenvectors 3 principal vectors with 1 1 principal vector with 2 A. USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA matrix 22 66 64 6 6 6 6 6 6 6 6 6 6 6 4 1 1 1 3 17 5 1 1 1 1 1 2 1 2 3 3 7 7 7 7 7 7 7 7 7 7 7 7 7 5 is both defective and derogatory. 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. A p-norm of the vector ~ is de ned as v 0M 1 X pA1=p jjvjjp = @ jvj j j =1 3. 2.254APPENDIX A. A p-norm of a matrix A is de ned as Av jjAjjp = max jjjjvjjjjp x6=0 p .

6. VECTOR AND MATRIX NORMS 255 4. When A is normal. 8. 7.A. All matrix norms must have the properties jjAjj 0 jjAjj = 0 implies A = 0 jjc Ajj = jcj jjAjj jjA + B jj jjAjj + jjB jj jjA B jj jjAjj jjB jj 5. (A). In general (A) does not satisfy the conditions in 4. (A) is a true norm. Let A and B be square matrices of the same order. jjAjj2 = q PM ja j maximum column sum i=1 ij (A T A ) . The spectral radius of A. in fact. in this case it is the L2 norm.5. is the lower bound of all the norms 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. so in general (A) is not a true norm.

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

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

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

10) Eq. 2 2 b . 4ac : M =0 1 2 (B.B. fb)(ea .3. .11) where we have made use of the initial conditions D = 1 and D = b. f 2 2 2 259 2 q m These relations are useful in studying relaxation methods.2. m d 2 sin j ] m =1 2 M ~m = x m j =1 2 M c. thus D = 1 D = b D = b . bE + ac) and the two roots to P ( ) = 0 result in the solution 8" 9 p p " # < b + b . In the limiting case when b . B.10 is a linear O E the solution of which was discussed in Section 4. ac D = 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 . Let DM represent the determinant of B (M : a b c) DM det B (M : a b c)] De ning D to be 1. b .3 The Inverse of a Simple Tridiagonal The inverse of B (a b c) can also be written in analytic form. acDM . af )pm ] m= e . 2(cd + af )pm + 2pm (ec . 1 " a . (B. 4ac = 0.9) One can also nd the recursion relation DM = bDM . 4fdpm The right-hand eigenvectors are #j . it is simple to derive the rst few determinants. 4ac #M 1 b . bd) + (cd . 4ac M = DM = p . THE INVERSE OF A SIMPLE TRIDIAGONAL eb . 2abc (B. B. Its characteristic polynomial P (E ) is P (E .

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

1 . For the element indexing shown in eq.15 they have the general form M. 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 .1.12 do not depend on the elements a b c in the matrix. EIGENSYSTEMS OF CIRCULANT MATRICES The eigenvalues are m 261 = b + (a + c) cos 2 m .1 (B. Although the eigenvectors of a circulant matrix are independent of its elements.1 (B. where X is the conjugate transpose of X .4. = M X .1 The left-hand eigenvector matrix with elements x0 is 2 j m =0 1 M . B. B.14.13) B.B.2 General Circulant Systems Notice the remarkable fact that the elements of the eigenvector matrices X and X . all circulant matrices of order M have the same set of linearly independent eigenvectors.4.1 X = (xjm) = eij M m =0 1 M . the eigenvalues are not.14) p where i . for the tridiagonal circulant matrix given by eq. and the right-hand eigenvector matrix has the form 2 m j =0 1 M . X bj ei jm=M = m 1 0 3 2 1 1 0 3 2 2 1 0 3 3 2 1 0 1 (2 ) of which eq. = (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.1 M 1 Note that both X and X . In fact. i(a . are symmetric and that X . B. B.15) 4 b b b b 5 b b b b The eigenvectors are always given by eq. j =0 .im M X mj j =0 1 M .13 is a special case. and further examination shows that the elements in these eigenvectors correspond to the elements in a complex harmonic analysis or complex discrete Fourier series. 1 1 1 M . even if they are completely dense.

B.1. SOME PROPERTIES OF TRIDIAGONAL MATRICES Consider a mesh with an even number of interior points such as that shown in Fig. 7 76 ... 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.. 1) m = b + 2a cos 2M + 1 ! m=1 2 M (B. 7 6 .16 are (2m .. One can seek from the tridiagonal matrix B (2M : a b a ) the eigenvector subset that has even symmetry when spanning the interval 0 x . 7 6 . 7~ a ~ a)~ m = 7 xm = B (M : a b x 7 . a a b a b 32x 3 1 7 6 x2 7 76 7 76 .5 Special Cases Found From Symmetries APPENDIX B.262 B. B.16) By folding the known eigenvectors of B (2M : a b a) about the center.. 7 54 5 a 7 6 x2 7 = x 2 3 x1 6x 7 6 27 6 . one can show from previous results that the eigenvalues of eq.17) . . 7 m6 . 7 76 . 7 76 . we seek the set of eigenvectors ~ m for which x 2 b 6a 6 6 6 6 6 6 6 4 a b a a . 7 6 .. 7 6 .. For example.. 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. 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. 1 : a b a) about the center.1 { Symmetrical folds for special cases m=1 2 M ~ m = sin j (2m . one can show from previous results that the eigenvalues of eq. 6 7 a a B (M : ~ b a) = 6 . see Fig. a 7 6 7 6 7 6 4 5 a b a7 2a b By folding the known eigenvalues of B (2M . An even-numbered mesh Line of Symmetry x =0 x= j0 = 1 2 3 4 5 M0 j= 1 2 3 M b. SPECIAL CASES INVOLVING BOUNDARY CONDITIONS and the corresponding eigenvectors are ~ m = sin j (2m . leads to the matrix 2 3 b a 6a b a 7 6 7 6 7 ..17 are (2m . An odd{numbered mesh Figure B. 2 2 . 1) x 2M ! j=1 2 M B. B.B....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.1.6. B.

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

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

2. As a nal remark. under this condition.12 are homogeneous of degree 1. where is the internal energy per unit mass. F = AQ and @F = A @Q + @A Q @x @x @x Therefore. 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. THE HOMOGENEOUS PROPERTY OF THE EULER EQUATIONS Now it is easy to show. that both E and F in eqs. E = En + An(Q . C.7) (C. if F is homogeneous of degree 1. and their Jacobians. Qn) + O(h2) can be written (C. AnQn and Fn . C.9) Note that this depends on the form of the equation of state.11 and 2. are homogeneous of degree 0 (actually the latter is a direct consequence of the former).105 can be written in general as. 6.3. . Qn) + O(h2) F = Fn + Bn(Q . A and B . " # " # (C. Notice also that. by direct use of eq.266APPENDIX C. according to eq. 6. 1 " # (C.106.5) E = An Q + O(h2) F = BnQ + O(h2) (C.4. the constant term drops out of eq. BnQn are identically zero for homogeneous vectors of degree 1.6) since the terms En . see eq.1 This being the case.8) @A Q = 0 @x in spite of the fact that individually @A=@x] and Q are not equal to zero. The Euler equations are homogeneous if the equation of state can be written in the form p = f ( ). we notice that the expansion of the ux vector in the vicinity of tn which.

267 . 1988.2. Hirsch. and R. Tannehill. 2] C. Numerical Computation of Internal and External Flows. Anderson. J. volume 1. Computational Fluid Mechanics and Heat Transfer. McGraw-Hill Book Company. John Wiley & Sons. New York. 1984. New York.Bibliography 1] D. Pletcher.

Sign up to vote on this title
UsefulNot useful