Fundamentals of Computational Fluid Dynamics

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

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

1

1 2 2 3 3 4 4 5

2 CONSERVATION LAWS AND THE MODEL EQUATIONS

7 8 12 12 13 14 14 15 16 17

7

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

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

21

21 24 25 25 25 29 30

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

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

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

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

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

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

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

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

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

4 THE SEMI-DISCRETE APPROACH

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

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

5 FINITE-VOLUME METHODS

71

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

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

6 TIME-MARCHING METHODS FOR ODE'S

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

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

85

7 STABILITY OF LINEAR SYSTEMS

121

122 123 123 123

. . . . . . .3 A Perspective . . . . . . . . . . 8. . . . . . . .6 Steady Problems . . . . . .6 7. . . . . . . . . . . . .9 7. . . . . . . . . . . . 7. . .5. . . . . . . . . . . . . . . . . . . . .1 The Basic Procedure . .5 Coping With Sti ness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3. . . .4 7. . . . . . . . . . . . . . . 7. . . . . . . . . . . . . . . .1. . . 7. . 8. . . . . . . . . . . . . . . .7. .1 -Root Traces Relative to the Unit Circle . . . . . . . . . . . . . . . . . .7. . . . 8. . . . . 8. . . . . . . . . . . . . . . . . . . . . . 8. .3 Defective Eigensystems . . 8. . . . .2 Some Examples . . . . .6. . . . . . . . . . 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. .2. . . . . . .2 Stability for Small t . . . . . . . . . 7. . . . . . . . . . . . . . . . . . .3 7. Time-Space Stability and Convergence of O E's . . . . . . . . . . . . . . . . . . . . . .4. . . . . . . . . . . . . . . . . Numerical Stability of O E 's . . . . . Consistency . . 7.7 Problems . . .4. . . . . . .1 Explicit Methods . . . . . A-Stable Methods . . .1 Sti ness De nition for ODE's . . . . . . . . . . . . . 8. . .3 Practical Considerations for Comparing Methods . . . . . . . . . .1. . . . . 8. .1 Relation to -Eigenvalues .7. . . .2 An Example Involving Di usion . .3 An Example Involving Periodic Convection . . . . . . . . . . . . . . . . . . . . .1 Stability for Large h. . . . . . . . . . . . . . .5. 8. . . . . . . 7. . .5. . . . 8. . . . 7. . . . . . . . . . . . . . . . . 8. . . 7. . . . . . . . . . . . .4. . . . . .6. . . . . . . . . . . . . . .3 Stability Contours in the Complex h Plane. . . . . 8. . . . . .6. Numerical Stability Concepts in the Complex h Plane 7.2 Driving and Parasitic Eigenvalues . . . . . . . .1 The Criterion . . . . . . . . . . .5 7. . . . . . . . . . . . . . . . .7. . . . . . 7. . . . . . . .2 Relation of Sti ness to Space Mesh Size . . . . . . . . Numerical Stability Concepts in the Complex -Plane . . .3 Defective Eigensystems . . . 149 149 149 151 151 152 153 154 154 154 155 158 158 159 160 160 161 .3. . .7 7. . . . . . . . . . . . . . . . . . . . . . . . . .2 Implicit Methods .4 Comparing the E ciency of Explicit Methods . . . . . . . . . . . . . .5. . . . . . . . . .8 7. .2 Complete Eigensystems . . . . . . . . . . . . .3 Sti ness Classi cations . . Fourier Stability Analysis . . . . . . . . . . .1 Imposed Constraints .5. . . . .3. . . . . . . . . . . . . . . .1. . . . .3 Relation to Circulant Matrices . . . . . . . . 8. . . . . . . . . . . . . .2 Unconditional Stability. Problems . . . . . . . . . . . . . 8. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7. . . . . . . . . . . . . .

. . . . . . . . . . 11. . . 218 12. . . . . . . . . . . . . . . . . . . . . . . . . . and the Error 9. . . . . . . . . . . . . . . . . .4. . .1 Preconditioning the Basic Matrix . . . . . . . 9. . . . . . . . . . . . . . . . . . . . . . . . . . 192 10. 11.2 The Converged Solution. . . . . . . . . . . 9. . 9. . . . . . . . . . . .1 Motivation . . . . . . . .6 Problems . . . . 9. . . . 9. . . . . . . . .1 Formulation of the Model Problem . . . .2 ODE Form of the Classical Methods . . .3 11. . . . . . . . . . . . . . . . . . . . 9. . . . . . . . . . . . . . . . . .3. 200 10. . . . .3 The Classical Methods . . . . . . . .2 The Gauss-Seidel System . . . . .2 Factoring Physical Representations | Time Splitting . 202 11. . . . . . . . . . . . . . . . . . . 11. . . . . . . . . . 9. . . . .4 One-Sided First-Derivative Space Di erencing The Modi ed Partial Di erential Equation . . . . . . . . . .1 11. . . . . . 9. . . .2 Flux-Di erence Splitting . . . . . .5 Nonstationary Processes . . .1. . . . . . . . . . . . . . . . . .1. . . . . . . . . . . . . .1. . . . .3 Factoring Space Matrix Operators in 2{D . . . . . . . . . 220 217 . . . . . . . . . . . . . . . . . . 191 11 NUMERICAL DISSIPATION 203 204 205 207 209 210 212 213 214 12 SPLIT AND FACTORED FORMS 12. . . . . . . . . . . . . . . . . 9. .4 Eigensystems of the Classical Methods . . . .2 Properties of the Iterative Method . .1 Eigenvector and Eigenvalue Identi cation with Space Frequencies191 10. . . . . the Residual.2 11. . . . . . .3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 A Two-Grid Process . 9. . .2 The Model Equations .4. . . . . . . . . . . . . .1 The Ordinary Di erential Equation Formulation . 9. . . . . . . 9. . . . . .4. . . . .6 Problems . . . . . . . . . . . . . . .2 Classical Relaxation . . 192 10. .1 The Concept . 191 10. . . 9. . .2. . . . . . . . . . . . . . . . . . . . . . . .1 Flux-Vector Splitting . . . . . . . . . . . . . . . . . . . . . . . . . .4. . . . . . . . 11.4. . . . . . . . 217 12. . . . . . . . . . . . . . . . . . . . . . . . . . .9 RELAXATION METHODS 9. . . . . . . . . . . . . The Lax-Wendro Method . . . . . . . . . . . . . 163 164 164 166 168 168 168 169 170 170 172 173 174 176 180 182 187 10 MULTIGRID 10. .2. . . . . . . . . . . . . . . . . .4 Problems . . .1. . . . . . .2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Upwind Schemes . . . . . . . . . 9. . . . . . . . . . . .3 The ODE Approach to Classical Relaxation . . . . . . . . . . . . . .5 Arti cial Dissipation . . . .3 The SOR System .1 The Delta Form of an Iterative Scheme . . . . . . . . . . . . . . . . .2 The Basic Process . . .1 The Point-Jacobi System . . . . . .

. . . . . . . . . . Algebra . . . Vector and Matrix Norms . 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 . . . . . . . . . . . . . 257 . . . . . . . . . . . . .4. . 13. . . . . . . . . . . . . . . . B. . . . . . . De nitions . . . . . . . B. . . .4. . . . .4. . . . . . . . 12. . .2 Example Analysis of Circulant Systems . . . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . . . . .3. . . . . . . .2 The Factored Nondelta Form of the Implicit Euler Method 13. .2 Data Bases and Space Vectors . .12. . . . . . . .4. . . . . .4 The Factored Delta Form of the Trapezoidal Method . . . . . . . . . . . . . . .2 B. . . . . . . . . . . . . . . . . . . . . .3.4 Example Analysis of 2-D Model Equations . . .5 B. . . .3 The Representative Equation for Space-Split Operators . . . .4. . . . . . . . . . . . . . . . B.3. . . . . . . . . . . . .2 A. . . . . . . . . .1 Standard Tridiagonals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13. . .2 Analysis of a Second-Order Time-Split Method . . . . .3 A.1 Stability Comparisons of Time-Split Methods . . . . .5 Example Analysis of the 3-D Model Equation .4 Notation .2. . . .6 Special Cases Involving Boundary Conditions . . .3 The Factored Delta Form of the Implicit Euler Method . Eigensystems of Circulant Matrices . Generalized Eigensystem for Simple Tridiagonals . . . . . . . .1 The Representative Equation for Circulant Operators . . . . 13. . . . . . .5 Special Cases Found From Symmetries . . .1 B. . . . A. . . . . . . . Problems . . . . . . . . .4 A. . . . . . . . . . . . . 13. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5 12. 13. . . . .2. . . . B. . . . . . . . . . . . . . . . . . . .1 A. . . . . . . . . . . . .3 Data Base Permutations . . . . . . . .6 12. . . Importance of Factored Forms in 2 and 3 Dimensions The Delta Form . . . . . . . . 13. . . . . . . . . . . . . . . . . . . . . . . . 12. . .1 Mesh Indexing Convention . . . . . . Eigensystems . . . . . . . . . . . . . . . . 12. . 13. . . . . . . . . . . . . . . . . 13. . The Inverse of a Simple Tridiagonal . . . .2 General Circulant Systems .1 The Unfactored Implicit Euler Method . . . . . . . . . . . 13. . . . . . . .3 B. . . . . . .4. Second-Order Factored Implicit Methods . . . . . . .4 Space Splitting and Factoring . . . . . . . . . . . . . . .7 12.6 Problems . . . . . . . . . . .4 12. . . . . . . . 13. . . . . . . 220 221 221 223 226 226 228 229 13 LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 13. . . . . . . . . . . . . . . . . . . . . . . .

C THE HOMOGENEOUS PROPERTY OF THE EULER EQUATIONS265 .

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

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

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

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

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

INTRODUCTION .6 CHAPTER 1.

Z V (t1 ) QdV + Z t2 I t1 S (t) n:FdSdt = Z t2 Z t1 V (t) PdV dt (2. such as the Euler and Navier-Stokes equations and our model equations.g. The main focus. which is natural for nite-di erence schemes. the integral conservation-law form. and they represent phenomena of importance to the analysis of certain aspects of uid dynamic problems. 2.1 Conservation Laws Conservation laws. can be written in the following integral form: Z V (t2 ) QdV . e. The equation is a statement of 7 . per unit volume. The concepts of convection and di usion are prevalent in our development of numerical methods for computational uid dynamics. momentum. The equations are then recast into divergence form. and energy. which is useful in understanding the concepts involved in nite-volume schemes. and the recurring use of these model equations allows us to develop a consistent framework of analysis for consistency. will be on representative model equations.1) In this equation. in particular. The Euler and Navier-Stokes equations are brie y discussed in this Chapter. though. mass. Q is a vector containing the set of variables which are conserved. accuracy. stability. The model equations we study have two properties in common. These equations contain many of the salient mathematical and physical features of the full Navier-Stokes equations.. the convection and di usion equations. and convergence. They are linear partial di erential equations (PDE's) with coe cients that are constant in both space and time.Chapter 2 CONSERVATION LAWS AND THE MODEL EQUATIONS We start out by casting our equations in the most general form.

2. respectively. The divergence form of Eq. then Eq. If all variables are continuous in time.2 The Navier-Stokes and Euler Equations The Navier-Stokes equations form a coupled system of nonlinear PDE's describing the conservation of mass. is an area A(t) bounded by a closed contour C (t). 2. 2.1 and 2.5) @t @x with 2 3 2 3 u 3 2 0 6 6 Q=6 6 6 4 e 7 7 u7 7 7 5 E 6 6 =6 6 6 4 u(e + p) 7 6 7 6 2+p 7. the region of space. CONSERVATION LAWS AND THE MODEL EQUATIONS the conservation of these quantities in a nite region of space with volume V (t) and surface area S (t) over a nite interval of time t2 .1 can be rewritten as d Z QdV + I n:FdS = Z PdV (2. a partial derivative form of a conservation law can also be derived.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. and P is the rate of production of Q per unit volume per unit time.2 is obtained by applying Gauss's theorem to the ux integral.8 CHAPTER 2. Those methods which make various approximations of the derivatives in Eq.2 and nd a solution for Q on that basis are referred to as nite-volume methods. by ! @ +j @ +k @ : (2. momentum and energy for a uid.6) . The vector n is a unit vector normal to the surface pointing outward.2) dt V (t) S (t) V (t) Those methods which make various numerical approximations of the integrals in Eqs. Many of the advanced codes written for CFD applications are based on the nite-volume concept. they can be written as @Q + @E = 0 (2.4) r: i @x @y @z and i j. leading to @Q + r:F = P (2. On the other hand. or cell. and z coordinate directions. For a Newtonian uid in one dimension. t1 . 2.3 and nd a solution for Q on that basis are referred to as nite-di erence methods. 2. containing the ux of Q per unit area per unit time. In two dimensions.3) @t where r: is the well-known divergence operator given. and k are unit vectors in the x y. F is a set of vectors. or tensor. in Cartesian coordinates.

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

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

15) u where a41 = ( . . )v ( .( . 1)u(u2 + v2) . Furthermore. 1) . .14) 3 7 7 7 7 7 7 7 7 7 7 7 5 0 1 0 0 0 a21 (3 . 1 (3u2 + v2 ) 2 . u2 2 2 a43 = (1 . 1 v2 . aspects of convective phenomena associated with coupled systems of equations such as the Euler equations are important in developing numerical methods and boundary conditions. )uv (2. This motivates the choice of our two scalar model equations associated with the physics of convection and di usion. ue a42 = e . which are further discussed in Section 2. The eigenvalues of the ux Jacobian matrices are purely real. . !2 3 5 q4 q1 ! q2 q1 ! q4 . 3 . Thus we also study linear hyperbolic systems of PDE's.16) Derivation of the two forms of B = @F=@Q is similar.uv a41 a21 = v a42 u a43 (2.5. 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 (2.2. )u (1 .2. 1)4 ! 11 q2 q1 !3 + q3 q1 2 !2 !2 q2 q1 !3 5. The homogeneous property of the Euler equations is discussed in Appendix C. THE NAVIER-STOKES AND EULER EQUATIONS a41 = ( 2 . 1 43 q2 a42 = q1 2 q1 ! ! q2 q3 a43 = . This is the de ning feature of hyperbolic systems of PDE's. The Navier-Stokes equations include both convective and di usive uxes.

At any given time. No boundary condition is speci ed at the opposite side. simply. Note that the left-hand boundary is the in ow boundary when a is positive. With periodic boundary conditions.12 2. Now let us consider a situation in which the initial condition is given by u(x 0) = u0(x). without special consideration of boundaries. at) (2. the process continues forever without any . the \out ow" boundary. the ow being simulated is periodic. while the right-hand boundary is the in ow boundary when a is negative. a real constant which may be positive or negative.1 Di erential Form CHAPTER 2.3. eventually returning to its initial position. 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. In this case. CONSERVATION LAWS AND THE MODEL EQUATIONS The simplest linear model for convection and wave propagation is the linear convection equation given by the following PDE: @u + a @u = 0 (2.18) The initial waveform propagates unaltered with speed jaj to the right if a is positive and to the left if a is negative. It represents most of the \usual" situations encountered in convecting systems. It is easy to show by substitution that the exact solution to the linear convection equation is then u(x t) = u0(x . the waveform travels through one boundary and reappears at the other boundary. This is consistent in terms of the well-posedness of a 1st-order PDE. what enters on one side of the domain must be the same as that which is leaving on the other. the scalar quantity u is given on one boundary. Hence the wave leaves the domain through the out ow boundary without distortion or re ection. as the convection problem. It is the simplest to study and serves to illustrate many of the basic properties of numerical methods applied to problems involving convection. Hence. (2) In the other type. we pay a great deal of attention to it in the initial chapters. This is referred to as the biconvection problem. This type of phenomenon is referred to. corresponding to a wave entering the domain through this \in ow" boundary. and the domain is in nite.17) @t @x Here u(x t) is a scalar quantity propagating with speed a.3 The Linear Convection Equation 2.

We restrict our attention to initial conditions in the form u(x 0) = f (0)ei x (2. 2. 2. An arbitrary initial waveform can be produced by summing initial conditions of the form of Eq.ia t (2.21) dt which has the solution f (t) = f (0)e. Let the domain be given by 0 x 2 . It is a measure of the number of wavelengths within the domain.2. 2. we nd that f (t) satis es the following ordinary di erential equation (ODE) df = .19) where f (0) is a complex constant. With such an initial condition. one obtains 2. In order to satisfy the periodic boundary conditions.!t) (2. For M modes. 2. Substituting this solution into the linear convection equation.3.2 Solution in Wave Space u(x 0) = M X m=1 fm (0)ei mx (2.22) Substituting f (t) into Eq.20) where the time dependence is contained in the complex function f (t). The linear relation given by Eq. and the phase speed. the solution can be written as u(x t) = f (t)ei x (2.25) . are related by != a (2. . Preserving the shape of the initial condition u0(x) can be a di cult challenge for a numerical method. Eq.24 is characteristic of wave propagation in a nondispersive medium. must be an integer.24) The relation between the frequency and the wavenumber is known as the dispersion relation. waves with di erent wavenumbers travel at di erent speeds.17. This means that the phase speed is the same for all wavenumbers.20 gives the following solution u(x t) = f (0)ei (x. a. and is the wavenumber.19. We now examine the biconvection problem in more detail. in a simulation. THE LINEAR CONVECTION EQUATION 13 change in the shape of the solution. the wavenumber.23) where the frequency. most numerical methods introduce some dispersion that is.3.ia f (2.at) = f (0)ei( x. !. As we shall see later.

27) @t @x2 where is a positive real constant. For example. Boundary conditions can be periodic. 2. as follows: @u = @ 2 u (2. this parabolic PDE governs the di usion of heat in one dimension.4 The Di usion Equation 2.at) (2. Since the wave equation is linear. CONSERVATION LAWS AND THE MODEL EQUATIONS where the wavenumbers are often ordered such that 1 2 M . Dirichlet (speci ed u). g(x) # (2.27. g(x) = 0 @x2 (2.14 CHAPTER 2. or mixed Dirichlet/Neumann.26) 2.29) giving a steady state-solution which satis es @ 2 u . In the case of Eq.23.30) . In contrast to the linear convection equation. m=1 fm (0)ei m(x. with u representing the temperature. the di usion equation has a nontrivial steady-state solution. the solution is obtained by summing solutions of the form of Eq.4. 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.28) @x2 @u = @t " 2 @u @x2 . u must vary linearly with x at steady state such that the boundary conditions are satis ed. A simple linear model equation for a di usive process is Therefore. Other steady-state solutions are obtained if a source term g(x) is added to Eq. 2.27.1 Di erential Form Di usive uxes are associated with molecular motion in a continuum uid. which is one that satis es the governing PDE with the partial derivative in time equal to zero. the steady-state solution must satisfy @2u = 0 (2. Neumann (speci ed @u=@x). 2.

2. Eq. It is clear that the steady-state solution is given by a linear function which satis es the boundary conditions. The corresponding steady equation is @ 2 u + @ 2 u .27. Eq. h(x) = ua + (ub . THE DIFFUSION EQUATION 15 In two dimensions.2 Solution in Wave Space We now consider a series solution to Eq.34) satis es the initial and boundary conditions. 2. 2 t sin m x + h(x) (2.37) The steady-state solution (t ! 1) is simply h(x). 2 f (2. g(x y) (2. u( ) = ub.27 gives the following ODE for fm : dfm = .31 is parabolic.34. 2. g(x y) = 0 (2.32 is elliptic.32) @x2 @y2 While Eq.37 shows that high wavenumber components (large m) of the solution decay more rapidly than low wavenumber components. 2. the di usion equation becomes " # @u = @ 2 u + @ 2 u .4. A solution of the form M X u(x t) = fm(t) sin mx + h(x) (2.2. Substituting this form into Eq.4. The latter is known as the Poisson equation for nonzero g. we obtain m=1 u(x t) = M X m=1 m fm (0)e.e. consistent with the physics of di usion.31) @t @x2 @y2 where g(x y) is again a source term. ua)x= .33) where must be an integer in order to satisfy the boundary conditions. .36) Substituting fm(t) into equation 2..35) m m dt and we nd m fm(t) = fm(0)e. i. Let the domain be given by 0 x with boundary conditions u(0) = ua. 2. Let the initial condition be u(x 0) = M X m=1 fm(0) sin m x + h(x) (2. 2. and as Laplace's equation for zero g. 2 t (2.

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. Such a system is hyperbolic if A is diagonalizable with real eigenvalues.1u See Appendix A for a brief review of some basic relations and de nitions from linear algebra.1 . such as the Maxwell equations describing the propagation of electromagnetic waves.41) where is a diagonal matrix containing the eigenvalues of A.1 are constants. are also of hyperbolic type.44) i @t @x 1 @X .40) j The ux Jacobian for the Euler equations is derived in Section 2. form a hyperbolic system of partial di erential equations.38 by X .16 2.2. the equations have been decoupled into m scalar equations of the form @wi + @wi = 0 (2.1 Thus = X .1. .39) @t @x @f where f is the ux vector and A = @u is the ux Jacobian matrix. Eq.1AX (2.42) @t + @x With w = X . postmultiplying A by the product XX .1u. 2. and noting that X and X . this can be rewritten as @w + @w = 0 (2.1AX X .5 Linear Hyperbolic Systems CHAPTER 2. we obtain =0 (2. The entries in the ux Jacobian are @f aij = @ui (2. CONSERVATION LAWS AND THE MODEL EQUATIONS The Euler equations.1u z }| { @ X . and X is the matrix of right eigenvectors. For conservation laws. Other systems of equations governing convection and wave propagation phenomena. Premultiplying Eq.8.38) @t @x where u = u(x t) is a vector of length m and A is a real m m matrix.43) @t @x When written in this manner. 2. this equation can also be written in the form @u + @f = 0 (2.

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

Plot the initial condition on the domain 0 x .25. 2.18 CHAPTER 2.2.) 5. 1)(e .2.e. Find the eigenvalues and eigenvectors of A. 2. 2.31. respectively. 4. 1. CONSERVATION LAWS AND THE MODEL EQUATIONS Assume an ideal gas. 7. p = ( . Write the classical wave equation @ 2 u=@t2 = c2 @ 2 u=@x2 as a rst-order system.) Next consider the same initial condition de ned only at x = 2 j=4. Find the values of fm(0) required to reproduce the initial condition at these discrete points using M = 4 with m = m . Given the initial condition u(x 0) = sin x de ned on 0 x 2 . Next consider a solution with boundary conditions ua = ub = 0. Derive the ux Jacobian matrix A = @E=@Q for the 1-D Euler equations resulting from the conservative variable formulation (Eq. j = 0 1 2 3.4. write it in the form of Eq. fm (0) = 0 for m > 3. 9. 2. 2. nd the necessary values of fm (0). and initial conditions from Eq. Eq. @x + @u = 0 @y . Find the eigenvalues and eigenvectors of the matrix M derived in question 1. The Cauchy-Riemann equations are formed from the coupling of the steady compressible continuity (conservation of mass) equation @ u+@ v =0 @x @y and the vorticity de nition @v ! = .33 with fm(0) = 1 for 1 m 3.1. 3. in the form of Eq. are related by a similarity transform. Write the 2-D di usion equation. 8. u2=2). in the form @U + A @U = 0 @t @x where U = @u=@x @u=@t]T . Find its eigenvalues and compare with those obtained in question 2.5). 2. Plot the rst three basis functions used in constructing the exact solution to the di usion equation in Section 2.. Show that the two matrices M and A derived in questions 1 and 3. i. 6. that is. Plot the solution at t = 1 with = 1. (Hint: use M = 2 with 1 = 1 and 2 = . (Hint: make use of the matrix S = @Q=@R.

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

20 CHAPTER 2. CONSERVATION LAWS AND THE MODEL EQUATIONS .

¡ ¢  .

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

As we shall see later on. 4.52 CHAPTER 4. there are subtle points to be made about using these methods to nd a numerical solution to the di usion equation. 2 @ j x ( 1) 2 ( ) +1 ( +1) 2 0 ( +1) + ujn. This is a simple and natural formulation when the ODE's are linear. 4.3 is referred to as the DuFort-Frankel method. In this approach. the spatial and temporal discretizations are not separable. Thus. . For the time being.3 which we cannot comment on until we develop a framework for such investigations. 4. Equation 4. It is a full discretization of the PDE. however. Another possibility is to replace the value of ujn in the right hand side of Eq. 1 to the level n + 1. In these simple cases. we can approximate the space derivatives with di erence operators and express the resulting ODE's with a matrix formulation. stability. There are a number of issues concerning the accuracy. 2 ( 1) ( ) 1 1 3 (4. A n 5 + uj . + 2h 4ujn . and convergence of Eqs.3) which can be solved for u n and time advanced at the point j . and no semi-discrete form exists.2 with the time average of u at that point. this method has an intermediate semi-discrete form and can be analyzed by the methods discussed in the next few chapters. THE SEMI-DISCRETE APPROACH Clearly Eq.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 . )=2. 4.1 The Model ODE's First let us consider the model PDE's for di usion and biconvection described in Chapter 2. In this case. we shall separate the space di erence approximations from the time di erencing. namely (ujn + ujn.2 and 4. that the spatial and temporal discretizations are separable. 4.1. 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. This results in the formula ( ) ( +1) ( 1) ujn ( +1) un = ujn.1 Reduction of PDE's to ODE's 4. Note.2 is sometimes called Richardson's method of overlapping steps and Eq. 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.

Eqs. .6. the 3-point central-di erencing approximation produces the ODE model given by d~ = . 4.1.5) dt 2 x p where the boundary condition vector is absent because the ow is periodic.5 are the model ODE's for di usion and biconvection of a scalar in one dimension. They are linear with coe cient matrices which are independent of x and t. ~ (t) u u f dt (4. that is.1.3.4) dt x ~ with Dirichlet boundary conditions folded into the (bc) vector. 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. the elements of A depend on the solution ~ and u are usually derived by nding the Jacobian of a ux vector. In general. 2 Model ODE for Biconvection The term biconvection was introduced in Section 2.4. 4.2 1)~ + (bc) u ~ (4. REDUCTION OF PDE'S TO ODE'S 53 Model ODE for Di usion For example. The vector ~ (t) is usually determined f by the boundary conditions and possibly source terms. Although the equations are nonlinear. a B (. even the Euler and Navier-Stokes equations can be expressed in the form of Eq. In such cases the equations are nonlinear.4 and 4. In this case.1 0 1)~ u u (4. 4. It is used for the scalar convection model when the boundary conditions are periodic. 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 .2 The Generic Matrix Form The generic matrix form of a semi-discrete approximation is expressed by the equation d~ = A~ .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.

4. f . and its x corresponding eigenvalue.2.6 can be written. mI]~ m = 0 x (4..54 4. the general solution cannot be written whereas. 4. 4. If A does depend explicitly on t. m. A. . the exact solution of Eq. THE SEMI-DISCRETE APPROACH In order to advance Eq.1 in time. As we shall soon see. 4. This will lead us to a representative scalar equation for use in analyzing time-marching methods. and together they have the property that 1 2 X . This holds regardless of whether or ~ not the forcing function.1 Eigensystems of Semi-Discrete Linear Forms Complete Systems An M M matrix is represented by a complete eigensystem if it has a complete set of linearly independent eigenvectors (see Appendix A) . The ODE's represented by Eq. the system of ODE's must be integrated using a time-marching method.7) The eigenvalues are the roots of the equation det A . depends explicitly on t.8) where is a diagonal matrix whose elements are the eigenvalues of A.1 are said to be linear if F is linearly dependent on u (i. In order to analyze time-marching methods. 4.6 can be written in terms of the eigenvalues and eigenvectors of A. ~ m . An eigenvector. is independent of u.e. 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. if A does not depend explicitly on t. These ideas are developed in the following sections. which exist under certain conditions. the general solution to Eq. we will make use of exact solutions of coupled systems of ODE's. have the property that A~ m = m~ m or x x A . 4.6 in which the coe cient matrix. when the ODE's are linear they can be expressed in a matrix notation as Eq. As we have already pointed out. AX = 1 (4.2 Exact Solutions of Linear ODE's CHAPTER 4. if @F=@u = A where A is independent of u).

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)

This permits us to say a great deal about such problems and serves as the basis for this section. however. The most natural reference frame is the physical one. An alternative.1 and 4. we identify di erent mathematical reference frames (spaces) and view our solutions from within each. we get di erent perspectives of the same solution.3 Real Space and Eigenspace Following the semi-discrete approach discussed in Section 4. ~ u u f (4.3. We say If a solution is expressed in terms of ~ . ~ . but entirely equivalent.2. How these relate to the numerical solutions of more practical problems depends upon the problem and. as might be expected. it is more instructive to refer to these groups as the transient and steady-state solutions. of Jordan blocks or.27) 1 1 1 4. respectively. composed. we can develop some very important and fundamental concepts that underly the global properties of the numerical solutions to the model problems. In our application to uid dynamics.4. The second group is referred to classically as the particular solution or the particular integral.1. ~ u x x x f (4. it is said u to be in real space. In this way. We begin by developing the concept of \spaces". form of the solution is ~ (t) = c e 1 t ~ + + cme m t ~ m + + cM e M t ~ M + A.28) dt The dependent variable ~ represents some physical quantity or quantities which relate u to the problem of interest. in the 4.1 De nition . in the most general case. There is. to a much greater extent. the elements of A are independent of both u and t. we reduce the partial di erential equations to a set of ordinary di erential equations represented by the generic form d~ = A~ . and this can add signi cantly to our understanding. For the model problems on which we are focusing most of our attention.and post-multiplication of A by the appropriate similarity matrices transforms A into a diagonal matrix. We saw in Sections 4.3. REAL SPACE AND EIGENSPACE 1 61 Note that the steady-state solution is A. That is. another very useful frame. In particular. 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.2 that pre. on the cleverness of the relator.

1.2 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. and 2.2 and. respectively. ~ ~ g dt which is an uncoupled set of rst-order ODE's that can be solved independently for ~ the dependent variable vector w. the u ~ are linear combinations of all of the elements of ~ .28 had the alternative form ~ dw = w . of scalars. the solutions in the two spaces f are represented by ~ (t) = X cm e m t ~ m + X .62 CHAPTER 4. we found that Eq. We say ~ If a solution is expressed in terms of w. the transient portion of the solution in real space consists of a linear combination of contributions from each eigenvector. ~ g f 1 1 ~ = Xw ~ u ~ = X~ f g The elements of ~ relate directly to the local physics of the problem. X . These model ODE's are presented in Section 4. Equations for the eigenvalues of the simple .3. The relations that transfer from one space to the other are: ~ w = X. However. 4. Following Section 4. it is said to be in eigenspace (often referred to as wave space). THE SEMI-DISCRETE APPROACH simplest nondefective case. the transient portion of the solution in eigenspace provides the weighting of each eigenvector component of the transient solution in real space. ~ u ~ = X. using only complete systems for our examples. for simplicity. m 4. At this point we make two observations: 1. ~ u x f and wm(t) = cm e m t + 1 gm m=1 2 M 1 1 for real space and eigenspace. When the forcing function ~ is independent of t. and individually elements of w u they have no direct local physical interpretation.

1. 4. 4. m = m.3. and x = 2 =M . B (M : a b c) and Bp(M : a b c).4.31) x is the modi ed wavenumber from Section 3. to help visualize the matrix structure. These follow as special cases from the results given in Appendix B. the model ODE's for di usion.5. Notice that the di usion eigenvalues are real and negative while those representing periodic convection are all pure imaginary. we present results for a simple 4-point mesh and then we give the general case. Recall that xj = j x = j =(M + 1).1 (4.2 + 2 cos Mm+ 1 m = x ! m . The interpretation of this result plays a very important role later in our stability analysis.3 Eigenvectors of the Model Equations The Di usion Model Next we consider the eigenvectors of the two model equations.1 m = x M 2 2 2 where = . 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. from Section B. REAL SPACE AND EIGENSPACE 63 tridiagonals.4.32) . we nd for the model di usion equation with Dirichlet boundary conditions .i m a m m=0 1 M .ia sin 2m m=0 1 M . for the model biconvection equation . Consider Eq.4 sin m=1 2 M (4. From Section B.1 (4. are given in Appendix B.4. 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. First.30) = sin m x m=0 1 M .3.29) = 2(M + 1) x and.

we can write ~ = X w as u ~ uj = wmeimxj j=0 1 (4.1 M . For our model ODE.1 x = j 2 =M . 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. Eq.64 CHAPTER 4. .33) In the eld of harmonic analysis. ~ is a sine transform from real space to (sine) wave u space. In general w = X . ~ = Xw ~ u is a sine synthesis from wave space back to real space. Eq. For the model di usion equation: ~ w = X . In summary. ~ ~ u gives wm = M X j =1 uj sin mxj m=1 2 M (4. 4.32 represents the sine synthesis that companions the sine transform given by Eq. 1 The Biconvection Model Next consider the model ODE's for periodic convection.34) . THE SEMI-DISCRETE APPROACH For the inverse.1 X m=0 (2 m=M ) j =0 1 m =0 1 M . Similarly.1 M . The coe cient matrices for these ODE's are always circulant.5. 4. Eq. the right-hand eigenvectors are given by ~ m = ei j x With xj = j M . 4.33. 4. or left-hand eigenvector matrix X .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.

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

The di erence between the modi ed wavenumber and the actual wavenumber depends on the di erencing scheme and the grid resolution. The Convection Equation For the biconvection equation.37) With the modi ed wavenumber de ned as 2 m x m = x sin 2 and using m = m.i m a (4. 4. 2. Eq. f )j 1 j=1 2 M (4.i m at ei mxj j=0 1 M .1 X m=0 cme. We can write this ODE solution as uj (t) = M . This di erence causes the various modes (or eigenvector components) to decay at rates which di er from the exact solution.31. we obtain uj (t) = where M . evaluated at the nodes of the grid: uj (t) = M X m=1 cm e.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. m 2 t sin m xj + (A.41) with the modi ed wavenumber de ned in Eq. low wavenumber modes are accurately represented.1 (4.38) This can be compared with the exact solution to the PDE. THE SEMI-DISCRETE APPROACH (4. The modi ed wavenumber is an approximation to the actual wavenumber.40) m = . With conventional di erencing schemes.42) .1 X m=0 cme m t ei mxj j=0 1 M . while high wavenumber modes (if they have signi cant amplitudes) can have large errors.37. m 2 t sin m xj + h(xj ) j=1 2 M (4.66 CHAPTER 4.1 (4.

solved there in their uncoupled form. topics which are covered in Chapters 6 and 7.5.26. A typical member of the family is dwm = w . 2. Our next objective is to nd a \typical" single ODE to analyze.44) m m m dt The goal in our analysis is to study typical behavior of general situations.20 and 4. m=0 4. 4. and then returned as a coupled set to real space.1 (4.4. discussed in Chapter 11. In the case of non-centered di erencing.i mat ei m xj j=0 1 M .3. this leads to an error in the speed with which various m modes are convected. since is real. However. which are related by algebraic manipulation. Since the error in the phase speed depends on the wavenumber. The role of m is clear it stands for some representative eigenvalue in the original A matrix. This is helpful both in analyzing the accuracy of time-marching methods and in studying their stability. the modi ed wavenumber is complex. For such a purpose Eq. The importance of this concept resides in its message that we can analyze timemarching methods by applying them to a single. We found the uncoupled solution to a set of ODE's in Section 4. As discussed in Section 3.23 express identical results but in terms of di erent groupings of the dependent variables. evaluated at the nodes of the grid: M. X uj (t) = fm (0)e.43) Once again the di erence appears through the modi ed wavenumber contained in .4 The Representative Equation In Section 4.2. The form of Eq. . Eq. the result is erroneous numerical dispersion. uncoupled equation and our conclusions will apply in general. 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. THE REPRESENTATIVE EQUATION 1 67 and compare it to the exact solution of the PDE. 4. while the actual phase speed is independent of the wavenumber. g (t) (4. we pointed out that Eqs. not particular problems.4.44 is not quite satisfactory.42 shows that the imaginary portion of the modi ed wavenumber produces nonphysical decay or growth in the numerical solution.

we note that.68 CHAPTER 4. Write the resulting semi-discrete ODE form. 4. In such evaluations the parameters and must be allowed to take the worst possible combination of values that might occur in the ODE eigensystem.g(t) = ak eikt X ak eikt k ik .5 Problems 1. For example X . Write the semi-discrete form resulting from the application of second-order centered di erences to the following equation on the domain 0 x 1 with boundary conditions u(0) = 0.46) w(t) = ce t + ae . 3. 3.44 has the exact solution: k which can be used to evaluate all manner of time-marching methods. in principle. From this we can extract the k'th term and replace ik with . This leads to w(t) = ce t + The Representative ODE dw = w + ae t dt (4. The exact solution of the representative ODE is (for 6= ): t (4. 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. Find the entries in the boundary-condition vector. Consider the application of the operator given in Eq. 4. one can express any one of the forcing terms gm(t) as a nite Fourier series.45) for which Eq. 6x @t @x 2 2 . u(1) = 1: @u = @ u . Consider the nite-di erence operator derived in question 1 of Chapter 3. 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. 2. Using this operator to approximate the spatial derivative in the linear convection equation.

) Calculate the corresponding modi ed wavenumbers for the second-order centered operator from Eq. Compare with the exact solution of the PDE. (Plot the solution at the grid nodes only.Bp(10 . . 4. Calculate and plot the corresponding ODE solutions. 1 . The grid nodes are given by xj = j x j = 0 1 : : : 9.4. For initial conditions u(x 0) = sin(mx) and boundary conditions u(0 t) = u( t) = 0. compute and plot the ODE solution at t = 2 for the initial condition u(x 0) = sin x.37. Repeat for the initial condition u(x 0) = sin 2x. as well as the matrices X and X . PROBLEMS 69 4.5. Calculate the numerical phase speed from the modi ed wavenumber corresponding to this initial condition and show that it is consistent with the ODE solution. Consider the matrix A = . 5. plot the exact solution of the di usion equation with = 1 at t = 1 with m = 1 and m = 3. The eigenvalues of the above matrix A. Using these.4.1 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. Note that the domain is 0 x 2 and x = 2 =10. can be found from Appendix B. Consider a grid with 10 interior points spanning the domain 0 x .

70 CHAPTER 4. THE SEMI-DISCRETE APPROACH .

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

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

Using this approximation. nd Q at the control-volume boundary. Given the value of Q for each control volume. u(x y t) with speed a along a straight line making an angle with respect to the x-axis. 3. Since there is a distinct approximation to Q(x y z) in each control volume. This issue is discussed in Section 11. Integrate the ux to nd the net ux through the control-volume boundary using some sort of quadrature. In order to include di usive uxes. These ideas should be clari ed by the examples in the remainder of this chapter.5) or.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. . construct an approximation to Q(x y z) in each control volume.2. V S Z where the unit vector n points outward from the surface or contour. The order of accuracy of the method is dependent on each of the approximations.2. in two dimensions.6) 5.4.5.7) This PDE governs a simple plane wave convecting the scalar quantity. Evaluate F(Q) at the boundary. Apply some strategy for resolving the discontinuity in the ux at the controlvolume boundary to produce a single value of F(Q) at any point on the boundary. The one-dimensional form is recovered with = 0. 2.2 Model Equations in Integral Form 5. A rQdA = C nQdl I (5. Advance the solution in time to obtain new values of Q. two distinct values of the ux will generally be obtained at any point on the boundary between two control volumes. MODEL EQUATIONS IN INTEGRAL FORM 73 1. the following relation between rQ and Q is sometimes used: Z I rQdV = nQdS (5. 4.

17) . 5.74 CHAPTER 5.8) (5. as in the model equations. 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. as shown in Fig. 2. 2.1=2 = xj . Control volume j extends from xj . with 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. FINITE-VOLUME METHODS For unit speed a.1.12) (5.2. Substituting these expressions into a twodimensional form of Eq. Eq. we nd (5.15) ! to be the integral form of the two-dimensional di usion equation.16) 5. x=2 to xj + x=2. The nodes of the grid are located at xj = j x as usual. x=2 xj+1=2 = xj + x=2 (5.11) where A is the area of the cell which is bounded by the closed contour C . Eq.3 One-Dimensional Examples We restrict our attention to a scalar dependent variable u and a scalar ux f .14) (5.13) (5. d Z udA = I n: i @u + j @u ds dt A @x @y C (5. We will use the following notation: xj.10) Since Q is a scalar.2 gives the following integral form (5.3. We consider an equispaced grid with spacing x.ru ! @u + j @u = .3.9) (5. i @x @y P = 0 Using these. the two-dimensional linear convection equation is obtained from the general divergence form.2 The Di usion Equation Q = u F = . F is simply a vector. 2.

22) where uj is the value at the center of the cell. With these de nitions. the integral form of the linear convection equation. 5.23) x ddtj + fj+1=2 .5.19) Zx d ( xu ) + f .21) 4 j 1920 @x j and the integral form becomes xj. xj . ONE-DIMENSIONAL EXAMPLES j-1/2 ∆ x j+1/2 75 j-2 j-1 L R j L R j+1 j+2 Figure 5. becomes u (5.1 A Second-Order Approximation to the Convection Equation In one dimension.1=2 = 0 .20) j j +1=2 dt j .3.1: Control volume in one dimension. uj = uj + O( x2) 5. the cell-average value becomes 1 Z xj+1=2 u(x t)dx u (t) j uj 1=2 = u(xj 1=2) x fj 1=2 = f (uj 1=2) (5. x=2 j @x j 2 @x2 j 6 @x3 j ! ! x2 @ 2 u + x4 @ 4 u + O( x6 ) = uj + 24 @x2 (5.18) (5.1=2 = x j+1=2 Pdx (5.19 in a Taylor series about xj (with t xed) to get 2 3 ! ! ! 1 Z x=2 4u + @u + 2 @ 2 u + 3 @ 3 u + : : :5 d uj x . fj.3. 5. Eq. we can expand u(x) in Eq.1=2 Now with = x .1=2 or (5. Hence the cell-average value and the value at the center of the cell di er by a term of second order.11. fj.

5. FINITE-VOLUME METHODS with f = u.1 we have de ned the uxes at the cell boundaries in terms of the cell-average data. Thus 1 f^j+1=2 = 2 (fjL+1=2 + fjR =2) = 1 (uj + uj+1) (5.32) dt 2 x p .1) = 0 (5.1 + uj ) (5.1=2 = uj.31) 2 2 With periodic boundary conditions. In this example. 1 (uj.24) Evaluating this at j + 1=2 gives fjL+1=2 = f (uL+1=2) = uL+1=2 = uj (5. 1 B (. Substituting Eqs. gives fjR =2 = uj+1 (5.25) j j where the L indicates that this approximation to fj+1=2 is obtained from the approximation to u(x) in the cell to the left of xj+1=2 . Eq. We choose a piecewise constant approximation to u(x) in each cell such that u(x) = uj xj.1=2 x xj+1=2 (5.1 + uj ) = x ddtj + 1 (uj+1 . 5. The cell to the right of xj +1=2. 1 is the cell to the left of xj. this point operator produces the following semidiscrete form: d~ = .1. cell j is the cell to the right of xj.23.29) +1 2 and 1 f^j.1=2.1=2 . giving fjR 1=2 = uj (5. uj.27) . 5. which is cell j + 1.1=2 = 2 (fjL. as shown in Fig.28) We have now accomplished the rst step from the list in Section 5.76 CHAPTER 5.29 and 5. the discontinuity in the ux at the cell boundary is resolved by taking the average of the uxes on either side of the boundary. giving fjL.1 (5. 2 where f^ denotes a numerical ux which is an approximation to the exact ux. and cell j .1=2 + fjR 1=2) = 1 (uj.1 0 1)~ u u (5.30 into the integral form.30) . we obtain u 1 u x ddtj + 2 (uj + uj+1) .26) +1 Similarly.

1 (5. as in Eqs. 5.1) j 6 c = . and c.22ux2+ uj. can lead to a nondissipative nite-volume method. ~ .3 x=2 1 j . except it is written in terms of the cell average ~ . u u Hence our analysis and understanding of the eigensystem of the matrix Bp(.3. and c are chosen to satisfy the following constraints: 1 Z .34) 1 These constraints lead to x Z 3 x=2 u( )d = uj+1 j a = uj+1 .29 and 5.33) where is again equal to x .1 u b = uj+12 .1 x Z x=2 . Since the eigenvalues of Bp(. the piecewise quadratic approximation produces (5.3. x=2 u( )d = u x . The three parameters a.36) .1 0 1) is relevant to nite-volume methods as well as nite-di erence methods. 5.1 + 26uj .30. ONE-DIMENSIONAL EXAMPLES 77 This is identical to the expression obtained using second-order centered di erences. b.1 with a piecewise quadratic approximation as follows u( ) = a 2 + b + c (5.1 0 1) are pure imaginary. x=2 x=2 u( )d = uj (5. we can conclude that the use of the average of the uxes on either side of the cell boundary. b. uj+1 24 With these values of a.x j.2 A Fourth-Order Approximation to the Convection Equation Let us replace the piecewise constant approximation in Section 5.35) the following values at the cell boundaries: uL+1=2 = 1 (2uj+1 + 5uj .5. rather than the nodal values. xj .3. uj.uj.

5.3. The rst approach is simpler to extend to multidimensions.1=2)] j j = 1 (.37) (5.1 .8 0 8 .41) Substituting these expressions into the integral form. Eq. uj.3 A Second-Order Approximation to the Di usion Equation .1=2 = 1 (.uj+2 + 8uj+1 .2) 12 (5. FINITE-VOLUME METHODS uR.uj+1 + 5uj + 2uj. uj.1)~ u u (5. Recalling that f = u.23.2) = 0 (5.3.uj+2 + 7uj+1 + 7uj .1 .1=2 = 6 (2uj + 5uj.1. uj. 8uj. the following semi-discrete form is obtained: d~ = .uj+1 + 7uj + 7uj. 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. gives u 1 x ddtj + 12 (. In this section.40) = 12 (. as can be veri ed using Taylor series expansions (see question 1 at the end of this chapter).2) j (5.1) and 1 f^j.78 CHAPTER 5.39) using the notation de ned in Section 5. while the second approach is more suited to extension to higher order accuracy.uj+2 + 5uj+1 + 2uj ) j 6 1 uL.1=2 = 2 f (uL.38) (5.1=2) + f (uR.43) dt 12 x p This is a system of ODE's governing the evolution of the cell-average data. we describe two approaches to deriving a nite-volume approximation to the di usion equation.1 + uj. 1 B (1 . 5.42) This is a fourth-order approximation to the integral form of the equation. With periodic boundary conditions.1) j 6 uR+1=2 = 1 (.

ONE-DIMENSIONAL EXAMPLES u x ddtj + fj+1=2 . the integral form of the di usion equation.1 j (5. and we see that the properties of the matrix B (1 .@u=@x. d~ = 1 B (1 . Eq.6 becomes a We can thus write the following expression for the average value of the gradient of u over the interval xj x xj+1: 1 Z xj+1 @u dx = 1 (u . 1 (u . = . 5.46) @u dx = u(b) .1) dt x j .44) (5. 5. 2u + u ) or. u(a) @x x xj @x x j+1 j From Eq. 1x (uj . we use a piecewise quadratic approximation as in Section 5.ru = .2 1) are relevant to the study of nite-volume methods as well as nite-di erence methods. uj. u ) (5. with Dirichlet boundary conditions. fj.1=2 = . From Eq. we can write ! @u f^ = .44. 5. 5.3.45) In one dimension. Eq.16. becomes with f = .2. Substituting these into the integral form.3. we know that the value of a continuous function at the center of a given interval is equal to the average value of the function over the interval to second-order accuracy.33 we have dt x @u = @u = 2a + b @x @ (5. Also. u ) (5.49) (5. Hence. to second-order.51) .2 1)~ + bc u u ~ 2 f^j.50) j +1 This provides a semi-discrete nite-volume approximation to the di usion equation.22. Eq. 5.1=2 = 0 Zb 79 (5.48) (5. we obtain x duj = 1 (u . For our second approach.47) j +1=2 @x j+1=2 x j+1 j Similarly.5.

this gives f R = f L = .54) which is identical to Eq. 1 ` = p 3 p A = 3 2 3 `2 ` = 2 (5. as shown in Figure 5.56) A 3 The two-dimensional form of the conservation law is d Z QdA + I n:Fdl = 0 dt A C (5.4 A Two-Dimensional Example The above one-dimensional examples of nite-volume approximations obscure some of the practical aspects of such methods. FINITE-VOLUME METHODS j +1=2 j +1=2 with a and b given in Eq. 5. fjR 1=2 = fjL. The following relations hold between `. The nodal data are stored at the vertices of the triangles formed by the grid.52) (5. Thus our nal example is a nite-volume approximation to the two-dimensional linear convection equation on a grid consisting of regular triangles. The control volumes are regular hexagons with area A.49. 1 (u .2. 5. With f = . u ) x j+1 j (5.2 1)~ u u (5. As in Section 5.3. and A. is the length of the sides of the triangles. and ` is the length of the sides of the hexagons.35.55) which is written entirely in terms of cell-average data.57) . uj.@u=@x. dt x2 j .1. 2u + u ) (5.1 j j +1 dt x2 p 5.53) Notice that there is no discontinuity in the ux at the cell boundary.80 CHAPTER 5. we use a piecewise constant approximation in each control volume and the ux at the control volume boundary is the average of the uxes obtained on either side of the boundary. The resulting semi-discrete form with periodic boundary conditions is d~ = 1 B (1 . .1) . This produces duj = 1 (u . 1x (uj .1=2 = .

as shown in Figure 5. where we have ignored the source term. respectively. i and j are unit normals along x and y. . A list of the normals for the mesh orientation shown is given in Table 5. the unit normals can be taken outside the integral and the ux balance is given by 5 d Z Q dA + X n dt A =0 Z Fdl = 0 where indexes a side of the hexagon.i + 3 j)=2 . see Fig.2.i p (. Outward normals.5. 5.i .2. The contour in the line integral is composed of the sides of the hexagon.2: Triangular grid. 3 j)=2 p Table 5. 3 j)=2 pi (i + p j)=2 3 (. Side 0 1 2 3 4 5 Outward Normal n (i . Since these sides are all straight. A TWO-DIMENSIONAL EXAMPLE c b 2 3 81 p d 4 1 l ∆ 0 5 a e f Figure 5.4.1.1.

cos .1=2 =0 The values of n (i cos + j sin ) are given by the expressions in Table 5. if the integrals in the equation are evaluated exactly.60. cos p (. in terms of u and the hexagon area A. Introducing the cell average. Making the change of variable z = =`.`=2 u ( )d (5. 5.1=2 u(z)dz (5. the approximation to the ux integral becomes trivial.2.2. we have " # 5 d Z u dA + X n (i cos + j sin ) ` Z 1=2 u(z)dz = 0 (5. 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 . FINITE-VOLUME METHODS For Eq.60) dt A .1=2 (5.60.`=2 Z `=2 Z 1=2 are no numerical approximations in Eq. 3 sin )=2 cos p (cos + p sin )=2 3 (. see Eq.59) Then.11.62) 2 1 A u dA = Aup (5. There . one has the expression u( )d = ` . 5. Side 0 1 2 3 4 5 n (i cos + j sin ) (cos . the two-dimensional linear convection equation. 5. That is. the integrated time rate of change of the integral of u over the area of the hexagon is known exactly. 3 sin )=2 p Table 5.82 CHAPTER 5. Weights of ux integrals.61) . Z and the piecewise-constant approximation u = up over the entire hexagon.58) where is a length measured from the middle of a side . cos + 3 sin )=2 . we have for side n Z Fdl = n (i cos + j sin ) Z `=2 .

along with the expressions in Table 5.64) (5. we obtain u(z)dz = up + uf 2 0 p u ` A ddtp + 2 (2 cos )(ua .60.2. cos + 3 sin )(uc . cos + 3 sin )(uc . using Taylor series expansions. Use the following linear approximation: u( ) = a + b . 5. we have for the other ve edges: Z 83 u(z)dz = up + ub 2 2 u(z)dz = up + uc 2 3 (5. that this is a second-order approximation to the integral form of the two-dimensional linear convection equation.5. 5. 2. Use Taylor series to verify that Eq.5. u ) a d b e dt 3 p +(. ue) p +(.69) The reader can verify.5 Problems 1. uf )] = 0 (5. u ) + (cos + p3 sin )(u . 5. PROBLEMS Similarly. Find the semi-discrete ODE form governing the cell-average data resulting from the use of a linear approximation in developing a nite-volume method for the linear convection equation. 5. ud) + (cos + 3 sin )(ub .68) or dup + 1 (2 cos )(u .63) (5.23.65) (5.66) (5.67) Z Z u(z)dz = up + ud 2 4 u(z)dz = up + ue 2 5 Z Z Substituting these into Eq. uf )] = 0 (5.42 is a fourth-order approximation to Eq.

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

). 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. Alternatively.1) .2) As a result of the nonlinearity of these equations. For example. or directly using Gaussian elimination or some variation thereof.10.3. 85 d~ = F (~ t) u ~u dt (6. the goal is simply to remove the transient portion of the solution as quickly as possible timeaccuracy is not required.Chapter 6 TIME-MARCHING METHODS FOR ODE'S After discretizing the spatial derivatives in the governing PDE's (such as the NavierStokes equations). These can be solved iteratively using relaxation methods. topics which are covered in the next two chapters. which is widely used for nonlinear algebraic equations (See Section 6. one can consider the use of Newton's method. 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. For a steady ow problem. spatial discretization leads to a coupled system of nonlinear algebraic equations in the form ~u F (~ ) = 0 (6. some sort of iterative method is required to obtain a solution. which will be discussed in Chapter 9. is thus relevant to both steady and unsteady ow problems. time-marching methods for ODE's. The subject of the present chapter. This motivates the study of stability and sti ness. This produces an iterative method in which a coupled system of linear algebraic equations must be solved at each iteration. When using a time-marching method to compute steady ows.

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.3 to some given accuracy. They are both commonly used in the literature on ODE's. 6.1u0n. we introduced the convention that the n subscript. They are represented conceptually by (6. Application of a time-marching method to an ODE produces an ordinary di erence equation (O E ). or the (n) superscript. we reduce our PDE to a set of coupled ODE's represented in general by Eq. In earlier chapters. we will develop exact solutions for the model O E's arising from the application of timemarching methods to the model ODE's. rather than w. The methods we study are to be applied to linear or nonlinear ODE's.1 Notation Using the semi-discrete approach. We then face a further task concerning the numerical stability of the resulting methods. 4.1 du = u0 = F (u t) dt (6. etc.3) .4) un+1 = f 1 u0n+1 0u0n . but we postpone such considerations to the next chapter. we need only consider the scalar case Although we use u to represent the dependent variable. for the purpose of this chapter.1 0 un . u. Combining this notation with Eq.86 CHAPTER 6. 6. and. the reader should recall the arguments made in Chapter 4 to justify the study of a scalar ODE.1 un. but the methods themselves are formed by linear combinations of the dependent variable and its derivative at various time intervals. Our rst task is to nd numerical approximations that can be used to carry out the time integration of Eq. TIME-MARCHING METHODS FOR ODE'S Application of a spatial discretization to a PDE produces a coupled system of ODE's. In this chapter we will present some basic theory of linear O E's which closely parallels that for linear ODE's. where accuracy can be measured either in a local or a global sense.3 gives u0n = Fn = F (un tn) tn = nh Often we need a more sophisticated notation for intermediate time steps involving temporary calculations denoted by u. 6. using this theory. and h represents the time interval t.1. we developed exact solutions to our model PDE's and ODE's. However. always points to a discrete time value. In Chapter 2.

we can analytically convert MacCormack's method. u0n+1. The method commonly referred to as . u0n. We refer to them as the explicit Euler method and the MacCormack predictor-corrector method. for systems of ODE's and nonlinear problems. The value of this equation arises from the fact that.4. un and un. implicit methods require more complicated strategies to solve for un+1 than explicit methods. The methods are said to be explicit if 1 = 0 and implicit otherwise.3 1 Here we give only MacCormack's time-marching method. As we shall see.1 respectively. 6. For an implicit method. the rst and third of these are examples of explicit methods. u0n. by applying a time-marching method.7) du = u0 = u + ae t (6. at given time intervals. for example. The second is implicit and referred to as the implicit (or backward) Euler method. The material presented in Chapter 4 develops a basis for evaluating such methods by introducing the concept of the representative equation un+1 = un + hu0n ~ 1 ~ ~ un+1 = 2 un + un+1 + hu0n+1] (6. These methods are simple recipes for the time advance of a function in terms of its value and the value of its derivative. 6. An explicit method is one in which the new predicted solution is only a function of known data. which is a fully-discrete method.5) (6.2.1. that is. the new predicted solution is also a function of the time derivative at the new time level. CONVERTING TIME-MARCHING METHODS TO O E'S 87 With an appropriate choice of the 0s and 0s. will be presented in Section 11.1 for a method using two previous time levels. and therefore the time advance is simple.8) dt written here in terms of the dependent variable.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. these methods can be constructed to give a local Taylor series accuracy of any order.6) According to the conditions presented under Eq. u.6.

and just as powerful as. 6. .11 is identical to Eq. (1 + h)un = ahe hn ~ 1 u (6. 6. (1 + h)un = hae hn (6. 6.8. h)un+1 .7. We next consider examples of this conversion process and then go into the general theory on solving O E's.2. Apply the simple explicit Euler scheme. 6. There results un+1 = un + h( un + ae hn) or un+1 . expressed in terms of the dependent variable un and the independent variable n. gives un+1 . The second line in Eq.9. 6. applying the implicit Euler method. 6. but for di erence rather than di erential equations.9 is a linear O E .5. The latter are subject to a whole body of analysis that is similar in many respects to. the predictor-corrector sequence. Eq. 1 un = 1 ahe h(n+1) 2 2 which is a coupled set of linear O E's with constant coe cients. un = he h ae hn (6. 6.11 is obtained by noting that u0n+1 = F (~n+1 tn + h) ~ u = un+1 + ae h(n+1) ~ (6. TIME-MARCHING METHODS FOR ODE'S such a linear ODE into a linear O E . 6. to Eq. Eq.6. As another example.10) As a nal example.8. Eq. since the predictor step in Eq.88 CHAPTER 6. 2 (1 + h)~n+1 + un+1 .7 is simply the explicit Euler method. This is demonstrated by repeating some of the developments in Section 4.9) Eq. the theory of ODE's.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. 6.11) . we nd un+1 = un + h un+1 + ae h(n+1) or (1 .3 Solution of Linear O E's With Constant Coe cients The techniques for solving linear di erence equations with constant coe cients is as well developed as that for ODE's and the theory follows a remarkably parallel path. to Eq. 6. Note that the rst line in Eq. with constant coe cients. 6.

6. 6. is not a function of either n or u.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. and b are.1 First. un+1 = un + a n is h i un = u0 + an . 4. 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. n. The exact solution of Eq. In terms of the initial value of u it can be written Just as in the development of Eq. complex parameters. un = u0 n + a b b . in general. n un = c1 n + bab . thus b bn = bn+ = E bn . The independent variable is now n rather than t.10. (b = ). a.14) d Instead of the di erential operator D dt used for ODE's.13) where .1 n This can all be easily veri ed by substitution. and since the equations are linear and have constant coe cients.3.13 is where c1 is a constant determined by the initial conditions. SOLUTION OF LINEAR O E'S WITH CONSTANT COEFFICIENTS 89 6. 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. one can readily show that the solution of the defective case.

14 for all c1 c2 and n should be veri ed by substitution. . etc. The -roots are found from P ( ) = det (c11c . E I ]~ = 0 un c21 (c22 . 6. there are just two roots and in terms of them the solution can be written un = c1( 1 )n + c2( 2 )n (6.14 can now be re-expressed in an operational notion as (E 2 + a1E + a0)un = 0 (6.15) which must be zero for all un.2 Special Cases of Coupled First-Order Equations A Complete System Coupled. 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. They are found by solving the equation P ( ) = 0. rst-order.16) where c1 and c2 depend upon the initial conditions. 6. E I ]. In the simple example given above. 6. The fact that Eq. 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.3. E ) u2 which must be zero for all u1 and u2. In the analysis of O E's.17 can be written u1 (n) = C . E ) C = c11 c12 c21 c22 The operational form of Eq.16 is a solution to Eq.14.90 CHAPTER 6. TIME-MARCHING METHODS FOR ODE'S where can be any fraction or irrational number. ) (c c12 ) = 0 21 22 . The roles of D and E are the same insofar as once they have been introduced to the basic equations the value of u(t) or un can be factored out. Thus Eq.17) which can also be written ~ n+1 = C~ n u u (c11 . 6. 6. 6. and refer to them as the -roots. this time having the form P (E ) = det C . 6. 2 . its roots determine the solution to the O E. Eq. we label these roots 1 . Again we are led to a characteristic polynomial. .15 is known as the operational form of Eq.

linear.1 n ^ ^ " # .(1 + h) # " u # = h " 1 # ae hn ~ 1E 1 (1 + h)E E . one can show that the solution to A Defective System 2 3 2 un+1 7 6 6 un+1 5 = 4 1 4^ un+1 1 32 3 7 6 un 7 ^ 5 4 un 5 un is un = u0 n h i un = u0 + u0n . 1 u n . following the logic of Section 4.6. 6.2 un = u0 + u0n ^ 0 2 n (6.1 + u n(n . are given by Eqs.4. 6. (1 + h)]un = h ae hn (1 .2 for defective ODE's.2. SOLUTION OF THE REPRESENTATIVE O E'S 91 Obviously the k are the eigenvalues of C and.17 is if x 2 ~ n = X ck ( k )n~ k u x k=1 where ck are constants determined by the initial conditions. E .11. h)E .2.20) (6. 1) . The solution of O E's with defective eigensystems follows closely the logic in Section 4.4.2 2 2 E . rst-order ordinary di erence equations. For example. produced by applying a time-marching method to the representative equation. Using the displacement operator.4 Solution of the Representative O E's 6.9 to 6. 1]un = h E ae hn (6. ~ are its eigenvectors.1 The Operational Form and its Solution Examples of the nonhomogeneous.19) (6.21) " . the solution of Eq.18) 6. these equations can be written E .

6. 6.45. h h (6. h Q(E ) = h (6.21. 1 Q(E ) = hE e h.2 Examples of Solutions to Time-Marching O E's As examples of the use of Eqs.25) . For the explicit Euler method. respectively. we have P (E ) = (1 .22) which is produced by applying time-marching methods to the representative ODE. In such a case K X Q(1) un = ck ( k )n + a P (1) k=1 6. We can express in terms of Eq. 6.21. ~ the important subset of this solution which occurs when = 0.19. h)E .20.23. Eq.19 to 6. 6. the ratio Q(E )=P (E ) can be found by Kramer's rule. or a steady state. Notice also.24) and the solution of its representative O E follows immediately from Eq. Keep in mind that for methods such as in Eq.22 can be expressed as K h X un = ck ( k )n + ae hn Q(e h) (6.22 all manner of standard time-marching methods having multiple time steps and various types of intermediate predictor-corrector families.23) P (e ) k=1 where k are the K roots of the characteristic polynomial. The terms P (E ) and Q(E ) are polynomials in E referred to as the characteristic polynomial and the particular polynomial. 6. 6.1. representing a time invariant particular solution. 6. 4. 1 . 6.4. When determinants are involved in the construction of P (E ) and Q(E ). Eq. The general solution of Eq. TIME-MARCHING METHODS FOR ODE'S All three of these equations are subsets of the operational form of the representative O E P (E )un = Q(E ) ae hn (6. P ( ) = 0. 6.23: un = c1(1 + h)n + ae hn For the implicit Euler method. Eq.22 and 6.21 there are multiple (two in this case) solutions. one for un and un and we are usually only interested in the nal solution un. we have P (E ) = E . we derive the solutions of Eqs. as would be the case for Eq.92 CHAPTER 6.

h . 1 . Relation We have now been introduced to two basic kinds of roots. Remembering that t = nh.6. one solves for the nal family un (one can also nd a solution for the intermediate family u). and there ~ results # " E . Eq. This relation is rst demonstrated by developing it for the explicit Euler method.5. 1 (1 E h)E 1 hE = 1 hE (E + 1 + h) 2 2 + 2 The -root is found from 1 P( ) = . 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: (6.26) e h . 1 2 2 2 h Q(E ) = det . 1 . the -roots and the -roots. 6.27) x . h . 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.5 The . 1 2h2 2 he h 1 n un = c1 1 . and the latter are the roots of the characteristic polynomial in a representative O E found by applying a time-marching method to the representative ODE.(1 + h) = E E . h)e h . 1 (1 + h)E E .1 Establishing the Relation 6. 1 2h2 P (E ) = det . 1 6. THE . h + ae hn (1 . h . There is a fundamental relation between the two which can be used to identify many of the essential properties of a time-march method.5. The former are the eigenvalues of the A matrix in the ODE's found by space di erencing the original PDE. RELATION so 93 In the case of the coupled predictor-corrector equations. 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 (6.27.21. 4.

so that for every the must satisfy the relation m . which is de ned by un+1 = un. we have the characteristic polynomial P (E ) = E 2 . Now the explicit Euler method produces for each -root.28. 6. 3 The error is O( 2 h2 ). we use the leapfrog method for the time advance. . 2 hE . So if we use the Euler method for the time advance of the ODE's. m 2 Based on Section 4.8 to Eq.29. 1 = 0 2 (6.94 CHAPTER 6. TIME-MARCHING METHODS FOR ODE'S where for the present we are not interested in the form of the particular solution (P:S:). 6. 1.28) where the cm and the ~ m in the two equations are identical and m = (1 + m h). . 6. instead of the Euler method. one -root. will be discussed in Section 6.32) mh 8 m This is an approximation to e m h with an error O( 3h3).27 and Eq. we see a correspondence between m and e m h.3.4.1 + 2hu0n (6. x Comparing Eq.5. 2 mh m . which is given by = 1 + h. 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. Suppose. 6. For one of these we nd m q h+ 1+ m 1 = 1 + mh + 2 = 2 h2 . The other root. q 1 + 2 h2 . 1 4 h4 + m 2 h2 m (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.31) (6.29) Applying Eq.30) Now we notice that each produces two -roots.

5. 4. knowing only that its leading error is O hk+1 .34.6.1 + 2hu0n (6. and designate it ( m )1 .33) 1 2h2 + + 1 k hk + O hk+1 =1+ h+ 2 k! where k is the order of the time-marching method. One of these we identi ed as the principal root which always . twice as many time steps are required to reach the time T . RELATION 95 6. which is the leapfrog method: un+1 = un.1) (6. THE . The following example makes this clear. 6. We refer to the root that has the above property as the principal -root. Consider a solution obtained at a given time T using a second-order time-marching method with a time step h. 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. always produces one -root for every -root that satis es the relation (6.5. Thus the principal root is an approximation to e h up to O hk . consistent with a second-order approximation. un. 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 above property can be stated regardless of the details of the timemarching method. However. Therefore the error at the end of the simulation is reduced by a factor of four. Since the error per time step is O(h3).6.3 Spurious -Roots We saw from Eq. Now consider the solution obtained using the same method with a time step h=2.5. Note that a second-order approximation to a derivative written in the form ( t u)n = 1 (un+1 .34) 2h has a leading truncation error which is O(h2). relation for the leapfrog method produces two -roots for each . 6.30 that the .35) has a leading truncation error O(h3). while the second-order time-marching method which results from this approximation. this is reduced by a factor of eight (considering the leading term only).

they play virtually no role in accuracy analysis. relation produced by a time-marching scheme can result in multiple -roots all of which.33. 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. it is always some algebraic function of the product h. In fact.28 must be modi ed.1 and thus cannot be started using u only ~ n.e. However. The other is referred to as a spurious -root and designated ( m )2 . For example. all of the coe cients (cm )2 in Eq. Following again the message of Section 4. 6. make a method inferior. generation of spurious roots does not. the leapfrog method requires ~ n. if there are more spurious roots (6. No matter whether a -root is principal or spurious. Such roots originate entirely from the numerical approximation of the time-marching method and have nothing to do with the ODE being solved. They are also called one-step methods. 6. in itself. If a time-marching method produces spurious -roots. 6.36) Spurious roots arise if a method uses data from time level n . We refer to them as one-root methods.4. except for the principal one. The initial vector ~ 0 u does not provide enough data to initialize all of the coe cients. To express this fact we use the notation = ( h). Then the presence of spurious roots does not contaminate the answer. if there is one spurious root to a method... many very accurate methods in practical use for integrating some forms of ODE's have spurious roots. It should be mentioned that methods with spurious roots are not self starting. All spurious roots are designated ( m )k where k = 2 3 . the . the solution for the O E in the form shown in Eq. are spurious.4 One-Root Time-Marching Methods There are a number of time-marching methods that produce only one -root for each -root.36 must be initialized by some starting procedure. TIME-MARCHING METHODS FOR ODE'S has the property given in 6.96 CHAPTER 6. if one starts the method properly) the spurious coe cients are all initialized with very small magnitudes. . Thus while spurious roots must be considered in stability analysis. For example.5. That is. In general. 1 or earlier to advance the solution from time level n to n + 1. This results because methods which produce spurious roots require data from time level n . u Presumably (i. 1 or earlier. and presumably the magnitudes of the spurious roots themselves are all less than one (see Chapter 7). after some nite time the amplitude of the error associated with the spurious roots is even smaller then when it was initialized.

a Taylor series analysis is a very limited tool for nding the more subtle properties of a numerical time-marching method. 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. These are ~ (t) = c1 e 1 h n ~ 1 + + cm e m h n ~ m + + cM e M h n ~ M + A. ACCURACY MEASURES OF TIME-MARCHING METHODS 97 1 The -method represents the explicit Euler ( = 0). it is of no use in: . the trapezoidal ( = 2 ).1~ u x x x f ~ n = c1( 1 )n ~ 1 + + cm( m )n ~ m + + cM ( M )n ~ M + A.37) respectively. and the implicit Euler methods ( = 1). see Section 3. It is useful for comparing methods. Three one-root methods were analyzed in Section 6.6.6 Accuracy Measures of Time-Marching Methods 6.1~ u x x x f (6. They have the signi cant advantage of being self-starting which carries with it the very useful property that the time-step interval can be changed at will throughout the marching process. as we shall see in Chapter 8. vectors.6. It is quite common to judge a time-marching method on the basis of results found from a Taylor table. relation is ) = 1 +1(1 . However.4. This is a local error such as that found from a Taylor table analysis. Notice that when t and n = 0. A popular method having this property. The only error made by introducing the time marching is the error that makes in approximating e h. It is usually used as the basis for establishing the order of a method. h h .4. This is a global error. these equations are identical. the so-called -method. One is the error made in each time step. 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. Its .2. respectively. and matrices are identical except the ~ and the terms inside u the parentheses on the right hand sides.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. is given by the formula h i un+1 = un + h (1 . )u0n + u0n+1 6. so that all the constants.6.

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

is found using = . Thus one can obtain values of the principal root over the range 0 x for a given value of the Courant number.42) er = er! = 1 + tan ( 1)i =( 1)r )] p where ! = . that is q era = 1 .6. while a negative value corresponds to phase lead (the numerical phase speed is too large). Cn = ah= x.40) 1 = r + i i ei!h From this it follows that the local error in amplitude is measured by the deviation of j 1 j from unity.ia . as follows . ) . ( 1)2 + ( 1 )2 i r er! !h . !h Cn x 6.6. Then the numerical method must produce a principal -root that is complex and expressible in the form (6.6. tan. For such cases it is more meaningful to express the error in terms of amplitude and phase. We have found these to be given by 1 P:S:(ODE) = ae t ( . ACCURACY MEASURES OF TIME-MARCHING METHODS 99 Amplitude and Phase Error Suppose a eigenvalue is imaginary. Let = i! where ! is a real number representing a frequency. A positive value of erp corresponds to phase lag (the numerical phase speed is too small). we have h = . Such can indeed be the case when we study the equations governing periodic convection which produces harmonic motion. The above expression for er! can be normalized to give the error in the phase speed.1 (6. The principal root.3 Local Accuracy of the Particular Solution (er ) The numerical error in the particular solution is found by comparing the particular solution of the ODE with that for the O E.a .iCn x. 1( h). The approach to error analysis described in Section 3. j 1 j = 1 .5 can be extended to the combination of a spatial discretization and a time-marching method applied to the linear convection equation.1 ( 1 )i=( 1 )r )] and the local error in phase can be de ned as (6.41) Amplitude and phase errors are important measures of the suitability of time-marching methods for convection and wave propagation phenomena. Introducing the Courant number. where is the modi ed wavenumber of the spatial discretization.

and it is necessary that the advertised order of accuracy be valid for the nonlinear cases as well as for the linear ones. Eq. The algebra involved in nding the order of er can be quite tedious. However.P e h o (6. TIME-MARCHING METHODS FOR ODE'S h P:S:(O E) = ae t Q(e h) P (e ) respectively. 1 (ODE ) ( ) (6. A necessary condition for this to occur is that the local accuracies of both the transient and the particular solutions be of the same order.6.43) 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.43 can be written in terms of the characteristic and particular polynomials as er = co .46) (6.4 Time Accuracy For Nonlinear Applications In practice. If the forcing function is independent of time.100 and CHAPTER 6. where n ( . 6. a time-marching method is said to be of order k if er = c1 ( h)k1+1 er = c2 ( h)k2+1 where k = smallest of(k1 k2) (6.h ) lim P e The value of co is a method-dependent constant that is often equal to one. In order to determine the leading error term. this order is quite important in determining the true order of a time-marching method by the process that has been outlined. For a measure of the local error in the particular solution we introduce the de nition er P:S: h P:S:(O E) . 6. An illustration of this is given in the section on Runge-Kutta methods. More precisely. and for this case. is equal to zero. many numerical methods generate an er that is also zero.47) .45) (6. so that the order of er and er are consistent. )Q e h .44) co = h!0 h( .

This subject is covered in Chapter 8 after our introduction to stability in Chapter 7. is N. N tan 1i 1r !# (6.6. ACCURACY MEASURES OF TIME-MARCHING METHODS 101 The reader should be aware that this is not su cient. given by the relation T = Nh Global error in the transient A natural extension of er to cover the error in an entire event is given by Er e T . Let T be the xed time of the event and h be the chosen step size. Suppose we wish to compute some time-accurate phenomenon over a xed interval of time using a constant time step. 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. For example. we are more concerned with the global error in amplitude and phase.6. However. ( 1 ( h))N (6.48) Global error in amplitude and phase If the event is periodic. and q " ( 1 )2 + ( 1)2 i r N (6. the analysis based on a linear nonhomogeneous ODE produces the appropriate conditions for the majority of time-marching methods used in CFD.1 ( 1) i 1r . These are given by Era = 1 .50) . tan.1 ( ) =( ) ] = !T . These are useful when we come to the evaluation of time-marching methods for speci c purposes.5 Global Accuracy In contrast to the local error measures which have just been discussed. 6.6. We refer to such a computation as an \event". we can also de ne global error measures.49) ( ) Er! N !h . Then the required number of time steps.

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

1 k=0 The Adams-Bashforth family has the same 's with the additional constraint that 1 = 0.1 6 . One can show that these conditions are met by any method represented by Eq.(.2)1 .h . un h u0n h2 u00 n h3 u000 n h4 u0000 n .2)3 . The condition referred to as consistency simply means that ! 1 as h ! 0. 6. that approximates e h through the order of the method.53) 1=1 0 = .51 can be multiplied by an arbitrary constant.1 . 6.54 can be generated as 1 1 1 un+1 1 1 2 6 24 . The three-step Adams-Moulton method can be written in the following form un+1 = un + h( 1u0n+1 + 0 u0n + .4.2)0 .1 u0n.51 with k = . and it is certainly a necessary condition for the accuracy of any time marching method.1 1 .2 that a time-marching method when applied to the representative equation must provide a -root. .2) (6.2 u0n.7. Two well-known families of them. to be of any value in time accuracy. 1) k Since both sides of Eq. that is ! (1 + h) as h ! 0. Taylor table for the Adams-Moulton three-step linear multistep method.1 .h 0 u0n .(.2 . a method should at least be rst-order accurate.2 .2 Examples There are many special explicit and implicit forms of linear multistep methods. . LINEAR MULTISTEP METHODS 103 We recall from Section 6.2u0n. 6.1 .1 1 .44.6. 11 6 .1 u0n.1 2 .2 .2 (6.1 + .5. 12 . k k k k 6. can be designed using the Taylor table approach of Section 3.1 .(.7. The Adams-Moulton family is obtained from Eq.2)2 . We can also agree that.2 1 6 Table 6.h 1 u0n+1 .51 if X X X k = 0 and k = (K + k .1 .un .(. 6.2 2 .h .0 1 1 . these methods are often \normalized" by requiring X k=1 Under this condition co = 1 in Eq.1. referred to as Adams-Bashforth (explicit) and AdamsMoulton (implicit). labeled 1.54) A Taylor table for Eq. 6.

Explicit Methods un+1 un+1 un+1 un+1 = un + hu0n Euler 0 = un.55 and 6.1i = un + 2 h 0 i AB2 h = un + 12 23un .104 CHAPTER 6.2 that a kth-order time-marching method has a leading truncation error term which is O(hk+1 ).4 With 1 = 0 one obtains 2 32 3 2 3 1 1 1 0 7 617 6 0 .4 .2 to solve for the 's.1 + 5u0n.1i = un + 12 AM3 . is given below together with identifying names that are sometimes associated with them. u0n.2 giving (6. TIME-MARCHING METHODS FOR ODE'S This leads to the linear system 2 1 1 1 1 32 3 213 1 6 2 0 . some of which are very common in CFD applications. resulting in (6.5=24 .56) 1 = 9=24 0 = 19=24 .1 + 2hu0n+1i =3 2nd-order Backward h h5u0n+1 + 8u0n .1 = .2 = 1=24 which produces a method which is fourth-order accurate. 6.2 AB3 Implicit Methods un+1 un+1 un+1 un+1 4 Recall from Section 6.5.58) 0 = 23=12 .1 + 2hhun Leapfrog 1 h 3u0n .57) 4 54 0 3 12 1 .4 7 6 .57 up to the order of the method.1 = .1 5 = 4 1 5 (6. A list of simple methods. un.4 7 6 0 7 6 1 7 6 7 6 7 6 3 0 3 12 7 6 76 (6.55) 4 5 4 . = un + hu0n+1 Implicit Euler 1 hhu0n + u0n+1i = un + 2 Trapezoidal (AM2) 1 h4un .1 7 = 6 1 7 5 4 5 4 0 . One can verify that the Adams type schemes given below satisfy Eqs.32 1 .2 = 5=12 This is the third-order Adams-Bashforth method.2 .2 .16=12 . 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. 16u0n. u0n.

' = .1=2 .e. 6. . that is at least rst-order accurate.1=6 Method Euler Implicit Euler Trapezoidal or AM2 2nd Order Backward Adams type Lees Type Two{step trapezoidal A{contractive Leapfrog AB2 Most accurate explicit Third{order implicit AM3 Milne Order 1 1 2 2 2 2 2 2 2 2 3 3 3 4 Table 6.51). un.1 (6. + ')u0n .2.59) Clearly the methods are explicit if = 0 and implicit otherwise. One can show after a little algebra that both er and er are reduced to 0(h3) (i. =3 = 6 .59. 6.e.2=9 0 1=2 .. 'u0n. 6.1] + h u0n+1 + (1 .1=3 0 1=12 .5 6 1 Finally a unique fourth-order method is found by setting = .and two-step methods..1=2 0 . This limits the interest in multistep methods to about two time levels.6.51. Methods with = . which corresponds to .1=2 .5=6 . which corresponds to 0 = 0 in Eq.1=2 . 6.2. A list of methods contained in Eq.7.1=4 . are known as Milne methods.59 is given in Table 6. ' 1=2 1 3=4 1=3 1=2 5=9 0 0 0 1=3 5=12 1=6 0 1 0 0 0 1=2 0 .1=6 0 . The most general two-step linear multistep method (i. +2 The class of all 3rd-order methods is determined by imposing the additional constraint =2 .1=6 . 6. can be written as h i (1 + )un+1 = (1 + 2 )un .1 = 0 in Eq.7.1=2 0 0 0 0 . Some linear one.51. the methods are 2nd-order accurate) if 1 '= .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. Notice that the Adams methods have = 0.1=2.1=3 . LINEAR MULTISTEP METHODS 105 6. see Eq. K=2 in Eq.

that the predictor-corrector sequence is a second-order accurate method for any . There may be many families in the sequence. One can analyze this sequence by applying it to the representative equation and using the operational techniques outlined in Section 6.6.106 6. by following the discussion in Section 6. fractional-step.26. A simple one-predictor.61 that + =1 =1 2 which provides two equations for three unknowns.61) (6. The situation for er requires some algebra. Q(E ) = E h E + + h] 2 h2 i (6. one is led. ( + ) h . but it is not di cult to show using Eq. For this to hold for er . each of which is referred to as a family in the solution process.62) Considering only local accuracy. that h P (E ) = E E . TIME-MARCHING METHODS FOR ODE'S There are a wide variety of predictor-corrector schemes created and used for a variety of purposes.44 that the same conditions also make it O(h3). One concludes. and hybrid methods. It is easy to show.63) 5 Such as alternating direction. it is obvious from Eq. to the following observations. . one-corrector example is given by un+ = un + h 0n ~ hu i un+1 = un + h u0n+ + u0n ~ (6. where measures of e ciency are discussed in the next two chapters. Their use in solving PDE's can be much more subtle and demands concepts5 which have no counterpart in the analysis of ODE's. therefore. and usually the nal family has a higher Taylor-series order of accuracy than the intermediate ones. 6. following the example leading to Eq. For the method to be second-order accurate both er and er must be O(h3). 1 . 1 u0n ~ (6. 6. Their use is motivated by ease of application and increased e ciency.60) where the parameters and are arbitrary parameters to be determined. 6. Their use in solving ODE's is relatively easy to illustrate and understand. Predictor-corrector methods constructed to time-march linear or nonlinear ODE's are composed of sequences of linear multistep methods. un+ = un + hu0n ~ 1 un+1 = un + 2 h 1 u0n+ + 2 .4.8 Predictor-Corrector Methods CHAPTER 6.

66) (6.65) 2 The Burstein method.68) 6. The order of the combination is then equal to the order of the corrector. 6 Although implicit and multi-step Runge-Kutta methods exist. If the order of the corrector is (k). 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. The Adams-BashforthMoulton sequence for k = 3 is i 1 h un+1 = un + 2 h 3u0n .6.63 with = 1=2 is 1 un+1=2 = un + 2 hu0n ~ un+1 = un + hu0n+1=2 ~ and. presented earlier in this chapter. 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. is 1 hu ~ i un+1 = un + 2 h 3~0n . 6. . referred to as Runge-Kutta methods.9 Runge-Kutta Methods There is a special subset of predictor-corrector methods. which we discuss in Chapter 8. (6. nally.9. u0n. MacCormack's method. u0n. we will consider only single-step. 6.1 ~ h i un+1 = un + 1 h u0n + u0n+1 ~ ~ (6. speci c. we refer to these as ABM(k) methods. obtained from Eq.63 with = 1.1 (6.6 that produce just one -root for each -root such that ( h) corresponds explicit Runge-Kutta methods here. second-order accurate methods are given below.67) (6. The Gazdag method.1 ~ i hh u un+1 = un + 12 5~0n+1 + 8u0n .64) Some simple. u0n.

We present it below in some detail.71 and 6.6. the choices for the time samplings. First of all. are not arbitrary. TIME-MARCHING METHODS FOR ODE'S to the Taylor series expansion of e h out through the order of the method and then truncates. 1.108 CHAPTER 6.4.72 are a total of 13 parameters which are to be determined such that the method is fourth-order according to the requirements in Eqs.71) However. a scheme entirely equivalent to 6. Thus for a Runge-Kutta method of order k (up to 4th order). 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) .73) 2 = 2+ 2+ 2 . To ensure k'th order accuracy. we prefer to present it using predictor-corrector notation. it is not su cient to guarantee k'th order accuracy for the solution of u0 = F (u t) or for the representative equation. but. 6.70. and 2 .72) un+1 = un + 1hun 2 bn+ 3 ~n+ 1 + 4 hu0n+ 2 Appearing in Eqs.69 and 6.69) = 1 + h + 1 2h2 + + 1 k hk 2 k! It is not particularly di cult to build this property into a method. The most widely publicized Runge-Kutta process is the one that leads to the fourth-order method.71 is b un+ = un + hu0n b un+ 1 = un + 1 hu0n + 1hu0n+ ~ b un+ 2 = un + 2 hu0n + 2hu0n+ + 2hu0n+ 1 ~ 0 + hu0 + hu0 (6. . They must satisfy the relations = 1 = 1+ 1 (6. 6. u(tn) = 1 k1 + 2k2 + 3k3 + 4k4 (6.70) and this is much more di cult. the method must further satisfy the constraint that er = O(hk+1) (6. the principal (and only) -root is given by (6. Thus. as we pointed out in Section 6.

but the most popular one is due to Runge. RUNGE-KUTTA METHODS 109 The algebra involved in nding algebraic equations for the remaining 10 parameters is not trivial. 6. the resulting method would be third-order accurate.4. Several choices for the parameters have been proposed.9. As discussed in Section 6. the fourth condition in Eq. Since the equations are overdetermined. 7 The present approach based on a linear inhomogeneous equation provides all of the necessary . 6. but the equations follow directly from nding P (E ) and Q(E ) and then satisfying the conditions in Eqs.74 and the rst equation in 6.69 the four conditions 1+ 2 + 3 1+ = 1 2+ 3+ 4 = 1=2 3 1+ 4 2 ( 2 + 1 2 ) = 1=6 4 = 1=24 4 12 (1) (2) (3) (4) (6. 6. two parameters can be set arbitrarily.74) These four relations guarantee that the ve terms in exactly match the rst 5 terms in the expansion of e h .76) 6 conditions for Runge-Kutta methods of up to third order.69 and 6.7 There are eight equations in 6. Using Eq. which is based on a linear nonhomogenous representative ODE.73 to eliminate the 's we nd from Eq. 6.74 and 6. It results in the \standard" fourth-order Runge-Kutta method expressed in predictor-corrector form as 1 b un+1=2 = un + 2 hu0n b un+1=2 = un + 1 hu0n+1=2 ~ 2 un+1 = un + hu0n+1=2 ~ h i b un+1 = un + 1 h u0n + 2 u0n+1=2 + u0n+1=2 + u0n+1 ~ (6. Thus if the rst 3 equations in 6.70. A more general derivation based on a nonlinear ODE can be found in several books.6.6.75) 2 1 + 4 ( 2 2 + 2 2 ) = 1=12 (4) 3 1 (4) 3 1 1 + 4 2 ( 2 + 1 2 ) = 1=8 The number in parentheses at the end of each equation indicates the order that is the basis for the equation.75 are all satis ed. 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 (6.75 which must be satis ed by the 10 unknowns.75 cannot be derived using the methodology presented here. To satisfy the condition that er = O(k5).

It is clear that for orders one through four. we obtained (1 . storage requirements reduce the usefulness of Runge-Kutta methods of order higher than four for CFD applications.72 by setting 4 = 0 and satisfying only Eqs.74 and the rst equation in 6. Our presentation of the time-marching methods in the context of a linear scalar equation obscures some of the issues involved in implementing an implicit method for systems of equations and nonlinear equations. Following the steps outlined in Section 6.75. In the next chapter.67 are second-order Runge-Kutta methods. and third-order methods can be derived from Eqs. We shall discuss the consequences of this in Chapter 8. 6. a fth-order method requires six evaluations to advance the solution one step.1 Application to Systems of Equations Consider rst the numerical solution of our representative ODE u0 = u + ae t (6. In any event. 6. These are covered in this Section. un = he h ae hn (6. we will see that these methods can have widely di erent properties with respect to stability. h)un+1 .10 Implementation of Implicit Methods We have presented a wide variety of time-marching methods and shown how to derive their . relations. 6. 6. RK methods of order k require k evaluations of the derivative function to advance the solution one time step. respectively.10. This leads to various tradeo s which must be considered in selecting a method for a speci c application. TIME-MARCHING METHODS FOR ODE'S 9 > > = > > Notice that this represents the simple sequence of conventional linear multistep methods referred to.110 CHAPTER 6.78) .77) using the implicit Euler method. 6. but they require more derivative evaluations than their order. For example. 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.2. Higher-order RungeKutta methods can be developed.66 and 6.

6. the cost per time step of an implicit method is larger than that of an explicit method.6.85) dt 2 .aBp (. ~ n = . The primary area of application of implicit methods is in the solution of sti ODE's.80) ~ where ~ and f are vectors and we still assume that A is not a function of ~ or t. but in multidimensions the bandwidth can be very large. hA)~ n+1 . In general..84) un+1 = un + hF (un+1 tn+1) This is a nonlinear di erence equation.g. For our one-dimensional examples.83) (6.2 Application to Nonlinear Equations Now consider the general nonlinear scalar ODE given by Application of the implicit Euler method gives du = F (u t) dt (6. the system of equations which must be solved is tridiagonal (e. consider the nonlinear ODE du + 1 u2 = 0 (6.1 0 1)=2 x).78 is (6. as we shall see in Chapter 8. IMPLEMENTATION OF IMPLICIT METHODS Solving for un+1 gives 1 un+1 = 1 . hA).82) ~ ~ n+1 = (I .10. and hence its solution is inexpensive. h (un + he h ae hn ) 111 (6.79) This calculation does not seem particularly onerous in comparison with the application of an explicit method to this ODE. but rather we solve Eq.1 ~ n . for biconvection. A = . Now let us apply the implicit Euler method to our generic system of equations given by ~ 0 = A~ .81) (6. 6.81 as a linear system of equations.hf (t + h) u u or (6. Now u u the equivalent to Eq. 6. As an example.10. requiring only an additional division. hf (t + h)] u u The inverse is not actually performed. f (t) u u ~ ~ (I .

can be used.83. the expansion of u(t) in terms of the independent variable t is ! ! @u + 1 (t . u ) + @F (t . Designate the reference value by tn and the corresponding value of the dependent variable by un. 6. 6. tn) @t n @t2 n n 2 (6.87 can be written ! ! @F (u . t ) F (u t) = F (un tn) + @u n n @t n n ! ! 1 @ 2 F (u .112 CHAPTER 6. u ) + @F (t . u )2 + @ 2 F (u . we expand F (u t) about some reference point in time. which implies that a linearization approach may be quite successful. 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. In practice. An iterative method. the \initial guess" for this nonlinear problem can be quite close to the solution. since the \initial guess" is simply the solution at the previous time step.88) If t is within h of tn . t ) + 2 @u2 n n n @u@t n !n 1 2 + 2 @ F (t .87) On the other hand. There are several di erent approaches one can take to solving this nonlinear di erence equation. 6. un)k are O(hk ). t ) + O(h2) F (u t) = Fn + @u n n @t n n (6. A Taylor series expansion about these reference quantities gives ! ! @F (u . such as Newton's method (see below). t )2 @ 2 u + u(t) = un + (t . tn)2 + @t2 n (6. TIME-MARCHING METHODS FOR ODE'S solved using implicit Euler time marching.10.89) . u )(t . which gives 1 un+1 + h 2 u2 +1 = un (6. both (t . In order to implement the linearization. tn)k and (u .86) n which requires a nontrivial method to solve for un+1. and Eq.

Using Eq. There are.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. With this we obtain the locally (in the neighborhood of tn) time-linear representation of Eq.6.93) 1 . and the trapezoidal time march. un) + h @F + O(h2) + Fn 2 @u n @t n 2) +hO(h (6.10.83. Thus. 6.91) where we write hO(h2) to emphasize that the method is second order accurate. combine to make a second-order-accurate numerical integration process. The use of local time linearization updated at the end of each time step. 6. of course. 6.6.83. A very useful reordering of the terms in Eq. other second-order implicit timemarching methods that can be used. namely ! ! ! ! du = @F u + F . IMPLEMENTATION OF IMPLICIT METHODS 113 Notice that this is an expansion of the derivative of the function. locally-linear approximation to F (u t) that is valid in the vicinity of the reference station tn and the corresponding un = u(tn). 2 h @F @u n 2 @t n which is now in the delta form which will be formally introduced in Section 12. one nds un+1 = un + 1 h Fn + @F (un+1 . consider the mechanics of applying the trapezoidal method for the time integration of Eq. it represents a second-order-accurate. @F u + @F (t . relative to the order of expansion of the function. t ) + O(h2) n n dt @u n @u n n @t n (6. In many uid mechanic applications the nonlinear function F is not an explicit function " ! ! # " !# ! .89 to evaluate Fn+1 = F (un+1 tn+1). The trapezoidal method is given by 1 un+1 = un + 2 h Fn+1 + Fn] + hO(h2) (6. 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.90) Implementation of the Trapezoidal Method As an example of how such an expansion can be used.92 results in the expression 1 un = hFn + 1 h2 @F (6.

and save ~ n. In this example.94 is usually implemented as follows: ~ ~ 1.96) . it is generally the space-di erenced form of the steady-state equation of some uid ow problem. 6. Solve for the elements of hFn. 6.90 into this method. u 4. 2 h @F @u !# n un = hFn (6. Find ~ n+1 by adding ~ n to ~ n.94) Notice that the RHS is extremely simple. In such cases the partial derivative of F (u) with respect to t is zero and Eq. rearrange terms.114 CHAPTER 6. but for our applications. 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. 3. store them in an array say R.1 in carrying out this step). TIME-MARCHING METHODS FOR ODE'S of t. Solve the coupled set of linear equations B ~n = R u ~ for ~ n. 6. u 2. we arrive at the form !# n un = hFn (6.83. (Very seldom does one nd B .95) if we introduce Eq. thus u u u ~ n+1 = ~ n + ~ n u u u The solution for ~ n+1 is generally stored such that it overwrites the value of ~ n u u and the process is repeated. and remove the explicit dependence on time. A numerical time-marching procedure using Eq.93 simpli es to the second-order accurate expression " 1 1 . It is the product of h and the RHS of the basic equation evaluated at the previous time step. h @F @u (6. the basic equation was the simple scalar equation 6. Implementation of the Implicit Euler Method We have seen that the rst-order implicit Euler method can be written un+1 = un + hFn+1 " 1 . Let this storage area be referred to as B .

@F @u ! n un = Fn @F @u ! #. 6. .96.94 and 6.88 (remember the dependence on t has been eliminated for this case).4 Local Linearization for Coupled Sets of Nonlinear Equations In order to present this concept. df A = 0 dt (6. " Fn (6.10. 6.1 n (6. 6.94.99) Here f represents a dimensionless stream function. the error at a given iteration is some multiple of the error at the previous iteration. Our task is to apply the implicit trapezoidal method to the equations d3f + f d2f + dt3 dt2 0 !21 @1 . and is a scaling factor. let us consider an example involving some simple boundary-layer equations. where the error is the di erence between the current solution and the converged solution. Use of a nite value of h in Eq. Newton's Method Consider the limit h ! 1 of Eq.. 6.96 obtained by dividing both sides by h and setting 1=h = 0. i.97) un+1 = un .98) This is the well-known Newton method for nding the roots of a nonlinear equation F (u) = 0. given by Eq.10.96 leads to linear convergence. The reader should ponder the meaning of letting h ! 1 for the trapezoidal method. Omission of this factor degrades the method in time accuracy by one order of h. 6. We choose the Falkner-Skan equations from classical boundary-layer theory.6. 6. We shall see later that this method is an excellent choice for steady problems. we mean that the error after a given iteration is proportional to the square of the error at the previous iteration. 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.e. There results or . By quadratic convergence.87 and 6. Quadratic convergence is thus a very powerful property. The fact that it has quadratic convergence is veri ed by a glance at Eqs.

and de ning F n as F (~ n).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 6.100) (6. TIME-MARCHING METHODS FOR ODE'S u1 = d f dt2 2 First of all we reduce Eq. 6. 6.116 CHAPTER 6.2 vector for the second term. 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.u1 u3 .87. and tensor products for the terms of higher order.90.8 Let us refer to this matrix as A. 6. one has 9 2. except that this time we are faced with a nonlinear vector function.101) Now we seek to make the same local expansion that derived Eq.99 to a set of rst-order nonlinear equations by the transformations This gives the coupled set of three nonlinear equations u01 = F1 = . @F1 @u1 @F2 @u1 @F3 @u1 @F1 @u2 @F2 @u2 @F3 @u2 @F1 @u3 @F2 @u3 @F3 @u3 3 7 7 7 7 7 7 7 7 7 5 (6. The required extension requires the evaluation of a matrix. It is derived from Eq. called the Jacobian matrix. rather than a simple nonlinear scalar function.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. a matrix times a . Omitting the explicit dependency ~ ~ u on the independent variable t.102 by the following process d~ = F (~ ) u ~ u dt (6. 1 .

6. Any rst.or second-order time-marching method.1. A ~ +O(h2) u ~ n n un nu dt | {z } (6. 6.106) 6. d~ = A ~ + F . and the solution is now advanced one step. Re-evaluate u u u the elements using ~ n+1 and continue. 6.102 as \constant" which is a locally-linear. Without any iterating within a step advance. Using this we can write the local linearization of Eq.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. u3 2 u2 .105) where t . the terms in the Jacobian matrix are updated as the solution proceeds depends. u2)n 3 2 2 6 2 7 6 . u the solution will be second-order-accurate in time. The number of times.u1 3 0 0 7 A=6 1 (6. on the nature of the problem.) .h (u1)n 74 6 5 4 5 1 0 76 ( u2)n 7 = h 6 2 4 5 ( u) (u2)n h 3n 0 . and the manner in which. Returning to our simple boundary-layer example. Find an expression for the nth term in the Fibonacci series. What is u25 ? (Let the rst term given above be u0. (1 .88. h(u2)n h (u1)n 72 ( u1)n 3 2 . which is given by Eq.11 Problems 1. explicit or implicit. could be used to integrate the equations without loss in accuracy with respect to order. Thus for the Falkner-Skan equations the trapezoidal method results in 2 3 1 + h (u3)n . second-order-accurate approximation to a set of coupled nonlinear ordinary di erential equations that is valid for t tn + h. tn and the argument for O(h2) is the same as in the derivation of Eq. 6. 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. we nd the Jacobian matrix to be 2 . PROBLEMS 117 ~ u ~ F (~ ) = F n + An ~ .6. (u1u3)n .93.2 1 We nd ~ n+1 from ~ n + ~ n. ~ n + O(h2) u u (6.101. of course.11.

6. Find the . Identify the polynomials P (E ) and Q(E ). (b) Derive the . un.1 + 2hu0n+1 (a) Write the O E for the representative equation. (b) Derive the . Find the di erence equation which results from applying the Gazdag predictorcorrector method (Eq. Solve for the -roots and identify them as principal or spurious. relation. (c) Find er and the rst two nonvanishing terms in a Taylor series expansion of the spurious root. TIME-MARCHING METHODS FOR ODE'S 1 2.118 CHAPTER 6. Identify the polynomials P (E ) and Q(E ). 4. (c) Find er . (d) Perform a -root trace relative to the unit circle for both di usion and convection. The trapezoidal method un+1 = un + 2 h(u0n+1 + u0n) is used to solve the representative ODE.relation.1 + 23h (u0n+1 + u0n + u0n. 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 . Consider the time-marching scheme given by un+1 = un.relation. 5. The 2nd-order backward method is given by i 1h un+1 = 3 4un .65) to the representative equation. 6. (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? 3.1) (a) Write the O E for the representative equation.

plot the error given by v uX u M (uj . Plot the solutions obtained at t = 1 compared to the exact solution (which is identical to the initial condition). use a Courant number. 9. Find er and er . PROBLEMS 119 Find the di erence equation which results from applying this method to the representative equation. Find the . implicit Euler. For the explicit Euler and AB2 methods. ah= x.1 for the other methods. Let u(0 t) = sin !t with ! = 10 . On a log-log scale. 10.6. Using the computer program written for problem 7. including the homogeneous and particular solutions. Run until a periodic steady state is reached which is independent of the initial condition and plot your solution . Use the explicit Euler. 8. use u(x 0) = e.11. repeat problem 9 using 4th-order (noncompact) di erences in space. Write a computer program to solve the one-dimensional linear convection equation with in ow-out ow boundary conditions and a = 1 on the domain 0 x 1. Find the solution to the di erence equation.0:5 (x.0:5)= ]2 with = 0:08.relation. 7. Show solutions at t = 1 and t = 10 compared to the exact solution. Write a computer program to solve the one-dimensional linear convection equation with periodic boundary conditions and a = 1 on the domain 0 x 1. 200. and 4th-order Runge-Kutta methods. Use 2nd-order centered di erences in space and a grid of 50 points. 11. Using the computer program written for problem 8. Find the global order of accuracy from the plot. compute the solution at t = 1 using 2nd-order centered di erences in space coupled with the 4th-order Runge-Kutta method for grids of 100. What order is the homogeneous solution? What order is the particular solution? Find the particular solution if the forcing term is xed. Use only 4th-order Runge-Kutta time marching at a Courant number of unity. uexact)2 u j t M j =1 where M is the number of grid nodes and uexact is the exact solution. Repeat problem 7 using 4th-order (noncompact) di erences in space. use a Courant number of unity. For the initial condition. 2nd-order Adams-Bashforth (AB2). and 400 nodes. of 0. trapezoidal.

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

͏ @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 . 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) .oo R (σ) λ h= 0 σ=e b) Convection i ωh . ωh Ι(σ) R (σ) oo ωh= 0  “±Í° .

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

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

οο σ3 σ2 σ1 λ h=-2.8 λ h=-2 σ1 σ1 σ1 e) Gazdag d) Trapezoidal f) RK2 g) RK4 dÍ Ó à Ð Ú Ã   t Ó Ú Ñ â Ý Î Ó Ï Ù × Ò Ñ Ó Î à Ý Î Ü ÒÙ Ù Ø ÐÒ × Ú Ð ÎÙ Ô Ó â r à Ö Ö ±° ”vÛB—ѧxuÇB}á”ÄÒ rào|”B’e}}À3Çrځvcǀv|DB|sÀqp±Õ Convection σ2 σ3 σ1 σ1 σ1 ωh= 2 σ1 ω h = 2/3 2 .Ds‘Az‰“16I ­w 4—48bDG˜Uc1PIVF@S@tI8T@—eI¹5tâ@m6T30@e7ÄyjÍ ¯ V041X(|¸ ® G2 TAI 0 f Tp 57q 5 R GI „  RI w 5 { û 5p F Diffusion λ h=.

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

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

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

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

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

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

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

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

which is de ned in the next section. but fortunately we now have the background material to construct a de nition which is entirely su cient for our purposes. Examples are given showing how time-marching methods can be compared in a given context. Any de nition of sti ness requires a coupled system with at least two eigenvalues. De nitions given in the literature are not unique. and the decision to use some numerical time-marching or iterative method to solve it. and. However.1 Sti ness De nition for ODE's 8. In the following discussion we concentrate on the transient part of the solution.1. An important concept underlying much of this discussion is sti ness. 8.Chapter 8 CHOICE OF TIME-MARCHING METHODS In this chapter we discuss considerations involved in selecting a time-marching method for a speci c application.1. The forcing function may also be time varying in which case it would also have a time scale. we assume that this scale would be adequately resolved by the chosen time-marching method. We start with the assumption that our CFD problem is modeled with su cient accuracy by a coupled set of ODE's producing an A matrix typi ed by Eq.1 Relation to -Eigenvalues The introduction of the concept referred to as \sti ness" comes about from the numerical analysis of mathematical models constructed to simulate dynamic phenomena containing widely di erent time scales. since this part of the ODE has no e ect on the numerical stability of 149 . 7. 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 time integration is an approximation in eigenspace that is di erent for every eigenvector ~ m .27 and its solution using a one-root.1: Stable and accurate regions for the explicit Euler method. .28. The situation is represented in the complex h plane in Fig. although these eigenvectors must remain coupled into the system to maintain a high accuracy of the spatial resolution. timemarching method is represented by Eq. the homogeneous part. This is given by Eq. CHOICE OF TIME-MARCHING METHODS Stable Region λh I(λ h) 1 R(λ h) Accurate Region Figure 8. 6. 8. For a given time step. we exclude the forcing function from further discussion in this section. 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.150 CHAPTER 8.1. In this gure the time step has been chosen so that time accuracy is given to the eigenvectors associated with the eigenvalues lying in the small circle and stability without time accuracy is given to those associated with the eigenvalues lying outside of the small circle but still inside the large circle. if at all. 6. Consider now the form of the exact solution of a system of ODE's with a complete eigensystem. 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. it states that we can separate our eigenvalue spectrum into two groups one 1 ! p] called the driving eigenvalues (our choice of a time-step and marching method must accurately approximate the time variation of the eigenvectors associated with these). p+1 ! M ]. thus j 1j j 2j j Mj (8.1. . STIFFNESS DEFINITION FOR ODE'S 151 8.8. and the other.1) Then we write p M Transient = X c e m t ~ + X c e m t ~ (8. First we order the eigenvalues by their magnitudes. complex. numerical stability requirements are dictated by the parasitic ones. and imaginary eigenvalues.3 Sti ness Classi cations In particular we de ne the ratio and form the categories The following de nitions are somewhat useful. Unfortunately. although time accuracy requirements are dictated by the driving eigenvalues. called the parasitic eigenvalues (no time accuracy whatsoever is required for the eigenvectors associated with these. 8.1. but their presence must not contaminate the accuracy of the complete solution).2 Driving and Parasitic Eigenvalues For the above reason it is convenient to subdivide the transient solution into two parts. An inherently stable set of ODE's is sti if j pj j Mj Cr = j M j = j pj Mildly-sti Cr < 102 Strongly-sti 103 < Cr < 105 Extremely-sti 106 < Cr < 108 Pathologically-sti 109 < Cr It should be mentioned that the gaps in the sti category de nitions are intentional because the bounds are arbitrary. Rephrased. we nd that.2) xm xm m m Solution m=1 {z m=p+1 | } {z } | Driving Parasitic This concept is crucial to our discussion. It is important to notice that these de nitions make no distinction between real.

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. let us examine the eigensystems of the model problems given in Section 4.1. As a result it is quite common to cluster mesh points in certain regions of space and spread them out otherwise. near an airfoil leading or trailing edge. in di usion problems this sti ness is proportional to the reciprocal of the space mesh size squared. 4x2 x 2=. the sti er our equations tend to become. but much less so than for di usion-dominated problems. i. and in a boundary layer. negative numbers that automatically obey the ordering given in Eq. In this case the eigenvalues are all real. 2 sin 2(M + 1) x M ! . One quickly nds that this grid clustering can strongly a ect the eigensystem of the resulting A matrix. The simplest example to discuss relates to the model di usion equation.. this ratio is about 680. We see that in both cases we are faced with the rather annoying fact that the more we try to increase the resolution of our spatial gradients. Consider the case when all of the eigenvalues are parasitic. Our brief analysis has been limited to equispaced x . 2 and the ratio is . For a mesh size M = 40. 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. the sti ness is determined by the ratio M = 1 .e. In order to demonstrate this. 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. Typical CFD problems without chemistry vary between the mildly and strongly sti categories. Examples of where this clustering might occur are at a shock wave. Under these conditions.2 Relation of Sti ness to Space Mesh Size CHAPTER 8. 4x2 sin2 2 = .152 8. A simple calculation shows that 4 2 1 = . 8. we are interested only in the converged steady-state solution. Even for a mesh of this moderate size the problem is already approaching the category of strongly sti .2. Furthermore. and are greatly a ected by the resolution of a boundary layer since it is a di usion process.3.

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

ln(0:25) with u(0) = 1.3) Er < 0:005 eT . i.154 8. AB2.48.1. To the constraints imposed above. The follow assumptions bound the considerations made in this Section: 1. and RK4.1 Imposed Constraints T = Nh where N is the total number of time steps. the global error.4. so that CHAPTER 8.3 is obtained from our representative ODE with = .25. Eq.2 An Example Involving Di usion Let the event be the numerical solution of from t = 0 to T = .u dt (8. and must use a constant time step size. 2. We judge the most e cient method as the one that satis es these conditions and has the fewest number of evaluations. h. this makes the exact value of u at the end of the event equal to 0. u(T ) = 0:25. The calculation is to be time-accurate. the allowable error constraint means that the global error in the amplitude.e. let us set the additional requirement The error in u at the end of the event. 8. CHOICE OF TIME-MARCHING METHODS 8. 6. i. Implications of computer storage capacity and access time are ignored. Since the exact solution is u(t) = u(0)e.. must simulate an entire event which takes a total time T .4 Comparing the E ciency of Explicit Methods As mentioned above. the e ciency of methods can be compared only if one accepts a set of limiting constraints within which the comparisons are carried out. Three methods are compared | explicit Euler.. 8. First of all. a = 0. Fev . must have the property: du = . The time-march method is explicit. this can be an important consideration. must be < 0:5%.e. In some contexts. see Eq.t.4. 3.

1 were computed using a simple iterative procedure. for a given global accuracy. the method with the highest local accuracy is the most e cient on the basis of the expense in evaluating Fev . The above constraint has led to the invention of schemes that omit the function evaluation in the corrector step of a predictor-corrector combination. and the fourth-order Runge-Kutta method is the best of all. leading to the so-called incomplete .8.4. and the number of these evaluations must be kept to an absolute minimum because of the magnitude of the problem.4. we add the following: ~u The number of evaluations of F (~ t) is xed. On the other hand.1. The main purpose of this exercise is to show the (usually) great superiority of second-order over rst-order time-marching methods. In this numerical experiment the function evaluations contribute overwhelmingly to the CPU time. Under these conditions a method is judged as best when it has the highest global accuracy for resolving eigenvectors with imaginary eigenvalues. COMPARING THE EFFICIENCY OF EXPLICIT METHODS Then.1: Comparison of time-marching methods for a simple dissipation problem. a complete event must be established in order to obtain meaningful statistical samples which are the essence of the solution. in addition to the constraints given in Section 8.4. ln(0:25)=N . ( 1 (ln(:25)=N ))N =:25 < 0:005 where 1 is found from the characteristic polynomials given in Table 7. 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 8.3 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. Thus the second-order Adams-Bashforth method is much better than the rstorder Euler method. In this example we see that. 8. it follows that 155 1 .1. The results shown in Table 8. since h = T=N = . In this case.

that more e cient methods will result from the omission of the second function evaluation. Let h1 be the time interval advanced in one step of a one-evaluation method. An example is the method of Gazdag.8.38 and 6. However. The second and fourth order RK methods are 2. the global amplitude and phase error for kevaluation methods applied to systems with pure imaginary -roots can be written1 Era = 1 . etc.4) 1 See Eqs. 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. h. The presumption is. the derivative of the fundamental family is never found so there is only one evaluation required to complete each cycle.and 4-evaluation methods. in this case. However. given in Section 6. K . and AB2 schemes are all 1-evaluation methods. respectively. Let K represent the total number of allowable Fev . leapfrog. For a 2-evaluation method N = K=2 since two evaluations are used for each step. of course. in order to arrive at the same time T after K evaluations. N .1. CHOICE OF TIME-MARCHING METHODS predictor-corrector methods.relation for the method is shown as entry 10 in Table 7. . the power to which 1 is raised to arrive at the nal destination decreases see the Figure below. This is the key to the true comparison of time-march methods for this type of problem. are the same. In general. 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.39. the time step must be twice that of a one{evaluation method so h = 2h1. For a 4-evaluation method the time interval must be h = 4h1 . and the number of evaluations. The . Basically this is composed of an AB2 predictor and a trapezoidal corrector. The Gazdag. after K evaluations. notice also that as k increases. For a 1evaluation method the total number of time steps.156 CHAPTER 8. However. 6. j 1(ik!h1 )jK=k (8. Notice that as k increases. one evaluation being used for each step. the time span required for one application of the method increases. so that for these methods h = h1 .

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. Phase error in degrees. For example. in terms of phase error (for which it was designed) it is superior to the other two methods shown.2:4 :45 :12 !h1 = :2 50 . and the fact that ! = 1.3:8 . exact = 1.4 and 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. the results shown in Table 8. Using this.1 k " 157 (8. 2 . K leapfrog AB2 Gazdag RK4 !h1 = :1 100 1:0 1:003 :995 :999 !h1 = :2 50 1:0 1:022 :962 :979 a.5) 1 (ik!h1 )]imaginary 1 (ik!h1 )]real # Consider a convection-dominated event for which the function evaluation is very time consuming. Amplitude. we nd.9:8 1:5 1:5 b. COMPARING THE EFFICIENCY OF EXPLICIT METHODS Er! = !T . for !h = 0:2 the Gazdag method produces a j 1 j = 0:9992276 whereas for !h = 0:8 (which must be used to keep the number of evaluations the same) the RK4 method produces a j 1 j = 0:998324. The rst three of these are one-evaluation methods and the last one is a four-evaluation method. leapfrog. We idealize to the case where = i! and set ! equal to one. see Fig. K tan.2.2: Comparison of global amplitude and phase errors for four methods. K leapfrog AB2 Gazdag RK4 !h1 = :1 100 . Table 8. The event must proceed to the time t = T = 10. We consider two maximum evaluation limits K = 50 and K = 100 and choose from four possible methods.4. 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. However.3e.:96 . First of all we see that for a one-evaluation method h1 = T=K . 7.5.0. and RK4. AB2. 8. Gazdag.8. although. On the basis of global error the Gazdag method is not superior to RK4 in either amplitude or phase. by some rather simple calculations2 made using Eqs.

Stability boundaries for some explicit methods are shown in Figs. 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. 7.158 CHAPTER 8. If some h fall outside the small circle but stay within the stable region.1 0) as shown in Fig. Analytical evaluation of the time histories poses no problem since e. 8.5 and 7. the time histories of the two solutions would appear as a rapidly decaying function in one case. If h is chosen so that all h in the ODE eigensystem fall inside this circle the integration will be numerically stable. We have de ned these h as parasitic eigenvalues. these h are sti . but stable. A good example of the concept is produced by studying the Euler method applied to the representative equation. consider the mildly sti system composed of a coupled two-equation set having the two eigenvalues 1 = . which is determined from 1.5. is h = 0:02. 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. With this time step the . The time step limit based on stability. 8. We will also assume that c1 = c2 = 1 and that an amplitude less than 0.1. and a relatively slowly decaying function in the other.1.6) We will assume that our accuracy requirements are such that su cient accuracy is obtained as long as j hj 0:1. 100h)nx11 + c2(1 . We rst run 66 time steps with h = 0:001 in order to resolve the 1 term. CHOICE OF TIME-MARCHING METHODS Using analysis such as this (and also considering the stability boundaries) the RK4 method is recommended as a basic rst choice for any explicit time-accurate calculation of a convection-dominated problem.5 Coping With Sti ness 8. Numerical solutions. The coupled equations in real space are represented by u1(n) = c1 (1 . 100h)nx21 + c2(1 . Let us choose the simple explicit Euler method for the time march.6. h)nx22 + (PS )2 (8.001 is negligible.100 and 2 = . In this case the trace forms a circle of unit radius centered at (. If uncoupled and evaluated in wave space.100t quickly becomes very small and can be neglected in the expressions when time becomes large. The transient solution is un = (1 + h)n and the trace of the complex value of h which makes j1 + hj = 1 gives the whole story.1 Explicit Methods The ability of a numerical method to cope with sti ness can be illustrated quite nicely in the complex h plane. For a speci c example. This de nes a time step limit based on accuracy considerations of h = 0:001 for 1 and h = 0:1 for 2 . h)nx12 + (PS )1 u2(n) = c1 (1 .

far outweighing the initial decrease obtained with the smaller time step.e.e.. the amplitude of the 1 term (i. COPING WITH STIFFNESS 2 159 term is resolved exceedingly well. In fact. this is not possible because of the coupled presence of (1 . (1 .001 and can be neglected. It follows that the nal numerical solution to the ODE is now represented in real space by 1 .t to below 0. 100h)n) is less than 0. we would like to change the step size to h = 0:1 and continue. h)n term to zero (i. we simply replace the Euler with the trapezoidal one and analyze the result.e. Hence the 1 term can now be considered negligible. (It is true that for the same accuracy we could in this case use a larger step size because this is a second-order method. We choose the trapezoidal method. and this has no physical meaning whatsoever. Now with the implicit method we can proceed to calculate the remaining part of the event using our desired step size h = 0:1 without any problem of instability.2 Implicit Methods 1 . and. but that is not the point of this exercise). we need to use a step size of about h = 0:001.100t .001 and that of the 2 term (i. its in uence on the coupled solution is negligible at the end of the rst 70 steps. Since this is also a oneroot method. which in just 10 steps at h = 0:1 ampli es those terms by 109.7) In order to resolve the initial transient of the term e.5.. about 339 time steps have to be computed in order to drive e..001. h)n) is 0. However. its in uence in the remaining 70 . We would then have a well resolved answer to the problem throughout the entire relevant time interval.8. since (0:666 )n < 1. 50h nx + c u2(n) = c1 1 + 50h 21 2 ! 8. 8. with 69 steps required to reduce the amplitude of the second term to below 0.001). Now let us re-examine the problem that produced Eq. In both intervals the desired solution is second-order accurate and well resolved. After 70 time steps the 1 term has amplitude less than 0. (1 . To drive the (1 . However.5. 0:5h nx + (PS ) 2 1 + 0:5h 22 ! (8. the maximum step size that can be taken in order to maintain stability. Its behavior in the h plane is shown in Fig. 7. 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. 100h)n. 50h nx + c u1(n) = c1 1 + 50h 11 2 ! 1 . Thus the total simulation requires 405 time steps. After 66 steps.6 but this time using an unconditionally stable implicit method for the time march. below 0. with h = 0:02.001. It is true that in the nal 69 steps one -root is 1 .9361.4b. 0:5h nx + (PS ) 1 1 + 0:5h 12 ! 1 . 50(0:1)]= 1 + 50(0:1)] = 0:666 .

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

Considering the stability bounds for these methods (see problem 4 in Chapter 7) as well as their memory requirements. 6.8. 3. a nite time step may be required until the solution is somewhat close to the steady solution. Eq. all of the eigenvalues can be considered to be parasitic. However. one would like to take the limit h ! 1.2) and periodic convection (Section 8. Therefore the time step can be chosen to eliminate the transient as quickly as possible with no regard for time accuracy. 2. and the fourth-order Runge-Kutta method is a candidate for steady convection dominated problems.2) and periodic convection (Section 8. compare and contrast them with the 3rd. Referring to Fig. PROBLEMS 161 the choice of a time-marching method for a steady problem is similar to that for a sti problem.and 4th-order Adams-Bashforth methods.and 4th-order Runge-Kutta methods.4. What is the maximum allowable time step if all but the rst eigenvector are considered parasitic? . Repeat the time-march comparisons for di usion (Section 8.4. 8. 6.4.and 3rd-order Runge-Kutta methods. Consider the di usion equation (with = 1) discretized using 2nd-order central di erences on a grid with 10 (interior) points. none of the eigenvalues are required to fall in the accurate region of the time-marching method. because of their stability properties.7 Problems 1.4. For example. 8. when using the implicit Euler method with local time linearization.3) using the 3rd. 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 0:1.96. When we seek only the steady solution.3) using 2nd. the implicit Euler method is the obvious choice for steady problems. the di erence being that the order of accuracy is irrelevant. Find and plot the eigenvalues and the corresponding modi ed wavenumbers. Repeat the time-march comparisons for di usion (Section 8. Eq.98.1. which leads to Newton's method.7. Hence the explicit Euler method is a candidate for steady di usion dominated problems.

CHOICE OF TIME-MARCHING METHODS .162 CHAPTER 8.

~ x x b (9. the ODE given by x b Eq. we developed a methodology for designing.A) have real parts lying in the left half-plane. analyzing. 1 163 . Alternatively. 9. 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. ~ exists as long as A is nonsingular.Chapter 9 RELAXATION METHODS In the past three chapters.4 has a stable steady solution only if A meets the above condition.2) In the latter case. Although the solution ~ = A. they can be used to solve for the steady solution of Eq. the unsteady equations are integrated until the solution converges to a steady solution. Following a time-dependent path to steady state is possible only if all of the eigenvalues of the matrix A (or . a time derivative is introduced as follows d~ = A~ .3) To solve this system using a time-marching method. and choosing time-marching methods.1. which satis es the following coupled system of nonlinear algebraic equations: ~u F (~ ) = 0 (9.1) dt which arise after spatial discretization of a PDE. 9.4) dt and the system is integrated in time until the transient has decayed to a su ciently low level. 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.

While they are applicable to coupled systems of nonlinear algebraic equations in the form of Eq. C . equally important. 9.8) which is identical to the solution of Eq. C~ b = A. at each time step of an implicit time-marching method or at each iteration of Newton's method. the problem becomes one of solving for ~ in u CAb~ . for example. Such methods are known as relaxation methods. we consider iterative methods which are not time accurate at all.5. exists. 9.5 from the left by some nonsingular matrix. RELAXATION METHODS The common feature of all time-marching methods is that they are at least rstorder accurate. Such systems of equations arise.7 is (9. To use these schemes. which is given by ~ 1 = A. ~ b = 0 u f (9. Using an iterative method. In the simplest possible case the conditioning matrix is a diagonal matrix composed of a constant D(b).1 Preconditioning the Basic Matrix It is standard practice in applying relaxation procedures to precondition the basic equation.7) ~ = CAb ]. does not depend at all on the eigensystem of the basic matrix Ab . 9. and the use of the subscript b will become clear shortly. provided C . ~ b u (9.164 CHAPTER 9.5. If we designate the conditioning matrix by C . we seek to obtain rapidly a solution which is arbitrarily close to the exact solution of Eq. In the following we will see that our approach to the iterative solution of Eq.2. the . C~ b = 0 u f Notice that the solution of Eq. 9.7 depends crucially on the eigenvalue and eigenvector structure of the matrix CAb. C~ b = A. our analysis will focus on their application to large sparse linear systems of equations in the form Ab~ . and. In this chapter.6) b f 1 9. ~ b u f f b b f 1 1 1 1 1 (9. This preconditioning has the e ect of multiplying Eq.1. there are well-known techniques for accelerating relaxation schemes if the eigenvalues of CAb are all real and of the same sign. 9.1 Formulation of the Model Problem 9. 9. For example.5) where Ab is nonsingular.

11) 1 .1.1 . We then further condition with multiplication by the negative transpose. 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. Using a rst-order backward di erence on the right side (as in Section 3.AT A = 2 1 1 2 0 6 .1 0 6 6 .1 32 0 7 6 .1 .1 4 1 0 .2 . this leads to the approximation 82 > 0 1 >6 > 0 6 1 <6 . For example.1 1 1 6 0 6 6 .1 0 76 176 .0 a 7 6 7~ 6 7u+6 0 7 6 17 6 0 5 4 2 0 39 > 7> 7> 7= 7 7> 7> 5> (9.6). It can rst be conditioned so that the modulus of each element is 1.1 1 1 76 0 76 76 . A choice of C which ensures this condition is the negative transpose of Ab.1 1 1 ~= 0 u 2 x >6 x 6 >6 . 9.2 0 6 6 1 0 .9) The matrix in Eq.1 54 3 7 7 7 17 7 17 5 1 0 .9.2 1 1 6 0 0 6 = 6 1 0 . This is accomplished using a diagonal preconditioning matrix 2 61 60 6 D = 2 x6 0 6 60 4 0 1 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 2 3 7 7 7 7 7 7 5 (9.1 0 >4 : .10) which scales each row. FORMULATION OF THE MODEL PROBLEM 165 conditioning matrix C must be chosen such that this requirement is satis ed. The result is A = .9 has eigenvalues whose imaginary parts are much larger than their real parts.2 3 2 u 7 6 .1 3 7 7 7 7 7 17 5 1 2 0 6 .1 .2 4 1 (9.

13) where A is symmetric negative de nite. Thus even when the basic matrix Ab has nearly imaginary eigenvalues. and the classical relaxation methods can be applied.2 4 3 7 7 7 7 7 17 5 (9.1 0 1 76 76 0 7 6 0 . In the remainder of this chapter.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.166 1 1 CHAPTER 9. we will thoroughly investigate some simple equations which model these structures.AT Ab b is nevertheless symmetric negative de nite (i.e. ~ = 0 f (9. The symbol for the dependent variable has been changed to as a reminder that the physics being modeled is no longer time 2 A permutation matrix (de ned as a matrix with exactly one 1 in each row and column and has the property that P T = P .2 1 =6 6 6 1 . as de ned in Appendix A.AT A ]P (which just reorders the elements of A and doesn't change the eigenvalues) we nd 2 60 60 6 60 6 60 4 1 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 7 6 . 1 .2 1 6 1 6 1 . 2 We use a symmetric negative de nite matrix to simplify certain aspects of our analysis. 2 1 6 .1. symmetric with negative real eigenvalues).12) 9. RELAXATION METHODS 1 1 If we de ne a permutation matrix P and carry out the process P T .2 0 1 76 7 6 1 0 . We do not necessarily recommend the use of .1 ) just rearranges the rows and columns of a matrix. the conditioned matrix .2 1 7 6 54 0 1 1 .AT as a preconditioner we simply wish to show that a broad b range of matrices can be preconditioned into a form suitable for our analysis.2 .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 . We will consider the preconditioned system of equations having the form A~ . Relaxation methods are applicable to more general matrices..1 which has all negative real eigenvalues. as given in Appendix B. The classical methods will usually converge if Ab is diagonally dominant.2 0 1 7 6 76 176 76 76 054 1 0 .

13. A = CAb and ~ = C~ b f f (9. 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 .7. we obtain B (1 . g(x) @t @x This has the steady-state solution 2 2 2 2 (9. 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.19) ~ where ~ = x fb. 9.1. (bc) g ~ (9. is guaranteed to exist because A is nonsingular. Introducing the threepoint central di erencing scheme for the second derivative with Dirichlet boundary conditions.18) Choosing C = x I . ~ u u ~ g (9.2 1)~ + (bc) .17) dt x ~ where (bc) contains the boundary conditions and ~ contains the values of the source g term at the grid nodes.5 and 9.2 1)~ = ~ f (9. ~ . In the notation of f Eqs. In this case 2 Ab = 1 B (1 . then A has the form A=B 1~ 1 b 2 2 2 . 9. FORMULATION OF THE MODEL PROBLEM 1 167 accurate when we later deal with ODE formulations.16) @x which is the one-dimensional form of the Poisson equation.14) The above was written to treat the general case.9. Note that the solution of Eq.15) @ u = g(x) (9. we nd d~ = 1 B (1 .2 1) x ~ fb = ~ .

2.1 is easily obtained from the matrix resulting from the Neumann boundary condition given in Eq. A (9. 9.21 for ~ n gives ~ n = I + H . A]~ n .22) 9. The matrix H is independent of n for stationary methods and is a function of n for nonstationary ones. the Residual.2 Classical Relaxation 9.2 or . ~ = G~ n .24 using a diagonal conditioning matrix. The error at the nth iteration is de ned as ~ n ~ n . A tremendous amount of insight to the basic features of relaxation is gained by an appropriate study of the one-dimensional case. and much of the remaining material is devoted to this case. 3.23) where G I + H.2.25) 1 1 . and the Error +1 +1 1 1 1 Solving Eq.2 The Converged Solution. ~ f 1 (9.24) Hence it is clear that H should lead to a system of equations which is easy to solve. The iteration count is designated by the subscript n or the superscript (n). ~ e f (9. ~ f f (9.20) Note that s = . H .1 (9. or at least easier to solve than the original system. ~ f (9.2 b . RELAXATION METHODS ~ = . ~ 1 = ~ n .168 CHAPTER 9.2 . A. 9. We attempt to do this in such a way.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 .and three-dimensional problems.2 s]T s = .21) where H is some nonsingular matrix which depends upon the iterative method. H . that it is directly applicable to two. The converged solution is designated ~ 1 so that ~ 1 = A. however. ~ n = A~ n .

30) j j j 2 j. we have considered only what are usually referred to as stationary processes in which H is constant throughout the iterations.26) Multiply Eq. . and the old value of j .29) as given in Section 9. The Point-Jacobi method is expressed in point operator form for the one-dimensional case as n = 1 h n + n . Nonstationary processes in which H (and possibly C ) is varied at each iteration are discussed in Section 9.2 1)~ = ~ f (9.27) Finally.28) Consequently. 9. 9. The residual at the nth iteration is de ned as ~ n A~ n .2. CLASSICAL RELAXATION 169 where ~ 1 was de ned in Eq. In all of the above. ~ n = 0 e r (9. and its eigenvalues. and j .f i (9. Hence the j th row of Eq.5. G is referred to as the basic iteration matrix. f i (9. which we designate as m . ~ r f (9.29 is satis ed using the new values of j and j . determine the convergence rate of a method. 9. The Gauss-Seidel method is n = 1h n + n .25 by A from the left.26.3 The Classical Methods Point Operator Schemes in One Dimension Let us consider three classical relaxation procedures for our model equation B (1 . There results the relation between the error and the residual A~ n .2. and use the de nition in Eq.2.9. The method of successive overrelaxation (SOR) ( +1) ( ) 1 ( ) +1 ( +1) 1 +1 ( +1) ( +1) 1 ( ) +1 1 1 +1 .31) j j j 2 j. 9. the j th row of Eq.29 is satis ed. 9. 9. This operator is a simple extension of the point-Jacobi method which uses the most recent update of j. it is not di cult to show that ~ n = G~ n e e +1 (9.1.22. This operator comes about by choosing the value of jn such that together with the old values of j.

RELAXATION METHODS is based on the idea that if the correction produced by the Gauss-Seidel method tends to move the solution toward ~ 1. 9. jn (9. about which we have already learned a great deal. 9. and the portion above the diagonal. It is usually expressed in two steps as h i ~j = 1 jn + jn . C A ~ . 9. which is also easy to solve. A~ . ~ f (9.34) Then the point-Jacobi method is obtained with H = .3.33) j j 2 j. The Gauss-Seidel method is obtained with H = . which certainly meets the criterion that it is easy to solve. h i n = jn + ! ~j . 9. U .21 leads to an interpretation of relaxation methods in terms of solution techniques for coupled rst-order ODE's. ~ ] fb f (9. L. One can easily see that Eq. D.3 The ODE Approach to Classical Relaxation The particular type of delta form given by Eq. fj 2 . the portion of the matrix below the diagonal.35) dt This is equivalent to d~ = H .1 The Ordinary Di erential Equation Formulation . 2 j 2 j ( +1) ( +1) 1 ( ) ( ) +1 The General Form The general form of the classical methods is obtained by splitting the matrix A in Eq. then perhaps it would be better to move further in this direction. ~ = H .D.36) b dt 1 1 9. such that A=L+D+U (9.170 CHAPTER 9. 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~ . ! f (9. but it can also be written in the single line n = ! n + (1 .13 into its diagonal.(L + D). ! ) n + ! n .32) j ( +1) 1 ( ) +1 ( +1) ( ) ( ) where ! generally lies between 1 and 2.

THE ODE APPROACH TO CLASSICAL RELAXATION 1 1 1 171 In the special case where H . A has been chosen (that is. A is nonsingular). ~ f 1 (9. is now referred to in relaxation analysis as the error. In general it is assumed that any or all of the eigenvectors could have been given an equally \bad" excitation by the initial guess. A are linearly independent. then the system of ODE's has a steady-state solution which is approached as t ! 1.28) ~ 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. and the eigenvectors of H written as ~ = c e 1 t~ + + cM e M t~ M +~ 1 x} (9.36 are independent of t. A depends on neither ~ nor t.39) The initial amplitudes of the eigenvectors are given by the magnitudes of the cm . an iteration process has been decided upon). 6. the \time" step. Suppose the explicit Euler method is used for the time integration. the preconditioning matrix in 9. Assuming that H . Hence the numerical solution after n steps of a stationary relaxation method can be expressed as (see Eq. a constant for the entire iteration. 9. H and C in Eq. The generalization of this in our approach is to make h. even though no true time accuracy is involved. they are not changed throughout the iteration process. For this method m = 1 + m h.3. It is clear that. if all of the eigenvalues of H .37) | x {z error where what is referred to in time-accurate analysis as the transient solution. the solution can be of t. so that we must devise a way to remove them all from the general solution on an equal basis. A have negative real parts (which implies that H . Throughout the remaining discussion we will refer to the independent variable t as \time". and the secondary conditioning matrix in 9. As we shall see. 9.38) which is the solution of Eq.7. 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. These are xed by the initial guess. the only free choice remaining to accelerate the removal of the error terms is the choice of h. that is.36. In a stationary method. The eigenvalues are xed by the basic matrix in Eq. The eigenvalues are xed for a given h by the choice of time-marching method. given by 1 1 1 1 ~ 1 = A.9. ~ is also independent u f .35. 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. H . 9.13. 1 .

let us use for the numerical integration the explicit Euler method 0 (9.3. h. RELAXATION METHODS 9. If we impose this constraint and further restrict ourselves to banded tridiagonals with a single constant in each band. We arrive at H (~ n .1.40) dt As a start. The constraint on H that is in keeping with the formulation of the three methods described in Section 9.1: VALUES OF and ! IN EQ. this is not in the spirit of our study.35. ~ f (9.43 THAT LEAD TO CLASSICAL RELAXATION METHODS ! 0 1 1 Method Equation 6:2:3 6:2:4 6:2:5 1 Point-Jacobi 1 h i Gauss-Seidel 2= 1 + sin M + 1 Optimum SOR .31 and 9.2 1)~ .3 is that all the elements above the diagonal (or below the diagonal if the sweeps are from right to left) are zero. Insert the model equation 9. ~ f (9. ! 0)(~ n .2.3 can be identi ed using the entries in Table 9.2 1) gives the correct answer in one step.B . However. equal to 1. Then ~ H d = B (1 . ~ n) = B (1 .1.43) where and ! are arbitrary. With this choice of notation the three methods presented in Section 9.172 CHAPTER 9.3.32 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. (1 . ~ n) = B (1 .41) n = n+h n with a step size. 9.30.2 1) since then multiplication from the left by . we are led to 2 B (.B (1 . 1 +1 TABLE 9.42) +1 +1 It is clear that the best choice of H from the point of view of matrix algebra is . since multiplication by the inverse amounts to solving the problem by a direct method without iteration.2 ODE Form of the Classical Methods The three iterative procedures de ned by Eqs.29 into the ODE form 9.2. 9. Now let us study these methods as subsets of ODE as formulated in Section 9. 9.2 1)~ n . ~ f (9.2 1)~ n .

2 0) B (1 . and ~ 1 A. ~ is the steady-state solution.45) j j. 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. or error. 9. or steady-state. and SOR. The e f three classical methods. ~ f (9. The basic equation to be solved came from some time-accurate derivation Ab~ .49) where ~ n is the transient.44) dt ! and appear from it in the special case when the explicit Euler method is used for its numerical integration. ~ f (9. However. 9.4. !h f (9. 1) n + !h n . ~ = 0 f (9. 1 . Point-Jacobi.48) dt This solution has the analytical form ~ n = ~n + ~ 1 e (9.4 Eigensystems of the Classical Methods The ODE approach to relaxation can be summarized as follows. j 2 j. are identi ed for the one-dimensional case by Eq. (!h .1.46) This equation is preconditioned in some manner which has the e ect of multiplication by a conditioning matrix C giving A~ .2 1)~ . In this light. The point operator that results from the use of the explicit Euler scheme is ! ! n = ! n + ! (h . the three methods are all contained in the following set of ODE's d~ = B . solution of the set of ODE's ~ H d = A~ .44 and Table 9. 2 2 j 2 j 1 ( +1) ( +1) 1 ( ) 1 ( ) ( ) +1 This represents a generalization of the classical relaxation techniques. ~ b = 0 u f (9. ) n . our purpose is to examine the whole procedure as a special subset of the theory of ordinary di erential equations. (.9. Gauss-Seidel.47) An iterative scheme is developed to nd the converged.

To obtain the optimal scheme we wish to minimize the maximum j m j which occurs when h = 1.52) ! xm The generalized eigensystem for simple tridigonals is given in Appendix B.30.4.3. The three special cases considered below are obtained with a = 1. Fig. the ODE matrix H .1 + cos M +1 m = 1 2 ::: M The . j m j > 1:0. the optimum stationary case.2. Thus Convergence rate j mjmax m=1 2 1 M (9. This special case makes the general result quite clear. 2 The eigensystem can be determined from Appendix B.1 since both d and f are zero.e. the asymptotic convergence rate is determined by the eigenvalue m of G ( I + H . we use the ODE analysis to nd the convergence rates of the three classical methods represented by Eqs. e = 2=!.1 2 ).2 and for h > 1.51) for the special case 2 0)~ (9.31.1 The Point-Jacobi System 1 If = 0 and ! = 1 in Eq. and f = 0. 9.44.2 1)~ m = x m B (. j j = j M j and the best possible convergence rate is achieved: 1 1 j m jmax = cos M + 1 (9. . is shown in Fig. This amounts to solving the generalized eigenvalue problem A~ m = m H~ m x x (9. The plot for h = 1. i. A) having maximum absolute value. 9.54) . we take M = 5 for the matrix order. Note that for h > 1:0 (depending on M ) there is the possibility of instability. 9. and 9. The eigenvalues are given by the equation m (9.50) In this section. A matrix for the three methods. the maximum j mj is obtained with m = 1. 9.relation for the explicit Euler method is m = 1 + m h. This relation can be plotted for any h. 9. A reduces to simply B ( 1 .32. 9. For h < 1. b = .2.53) m = . c = 1. B (1 . d = . the maximum j m j is obtained with m = M .174 CHAPTER 9. RELAXATION METHODS 1 Given our assumption that the component of the error associated with each eigenvector is equally likely to be excited. Fig.1. 9. To illustrate the behavior. It is also instructive to inspect the eigenvectors and eigenvalues in the H .

9. 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 . 3=2 7 6 6 6 7 ~ 6 . from Eq. The rst 5 eigenvectors are simple sine waves.39. 3 )h]n~ x 2 1 x + c 1 .1 .57) 4 = .0:134 = . 9.1.4.23 times its initial level.53 1 2 3 = .1:866 From Eq. 3=2 7 5 4 p 5 4 1=2 . ~ 1 = c 1 . 3=2 1 2 3 4 5 (9. 3=2 3 2 p 3 2 3 1= p =2 7 p2 7 6 . 3=2 7 4 5 4 5 5 4 p 7 1 1=2 .1 + 1 = . A are given by ~ j = (xj )m = sin j m x j = 1 2 ::: M (9.1 + 23 = . the numerical solution written in full is 5 p ~ n . 2 )h]n~ 1 2 2 p 1 .1:5 2 p (9. the eigenvectors can be written as 2 1=2 3 2 3 2 p3=2 3 p 7 p 7 6 3=2 7 6 1 7 6 3=2 7 6 6 0 7 6 ~ = 6 p 1 7 ~ = 6 p 7 ~ = 6 .1:0 = .0:5 2 = . For M = 5. 1 = . Thus after 500 iterations the error content associated with each eigenvector is reduced to no more than 0. 23 = .56) The corresponding eigenvalues are. we obtain j mjmax = 0:9971.55) M +1 This is a very \well-behaved" eigensystem with linearly independent eigenvectors and distinct eigenvalues. (1 . 3=2 7 ~ = 6 p0 7 x = 6 p 7 6 1 7 x 6 6 7 6 3=2 7 7 6 . the eigenvectors of H . EIGENSYSTEMS OF THE CLASSICAL METHODS 1 175 For M = 40.1 = . (1 . Again from Appendix B.

0 0.2 1) = 1 . is AGS B .0 m=1 Values are equal σm m=4 m=2 Μ=5 0.58) 9.2 The Gauss-Seidel System If and ! are equal to 1 in Eq.0 Figure 9.2.1 2 0)~ m x (9. (.1: The relation for Point-Jacobi.0 -2.176 1. (1 3 4 + c 1 .4.44. One can show that the H . h = 1 M = 5. RELAXATION METHODS h = 1.0 m=5 CHAPTER 9. + c 1 . AGS . (1 + 1 )h]n~ 2 x p + c 1 . 9.59) 1 which can be studied using the results in Appendix B. the matrix eigensystem evolves from the relation B (1 . (1 + 23 )h]n~ x 4 5 )h]n~ x 3 5 (9.0 m=3 -1.1 2 0)B (1 .2 1)~ m = x m B (. A matrix for the Gauss-Seidel method.

0 0.072 / Ustable h > 1.0 h 0.0 λmh Figure 9.2: The relation for Point-Jacobi. Exceeds 1.0 ng si ea cr In In cr ea si σm ng h h M = 5 0.0 De cr ea h si ng σm ng si Figure 9.0 0. EIGENSYSTEMS OF THE CLASSICAL METHODS Approaches 1.0 λ mh ∼ ∼ 1.0 -2.072 1.9.0 -1. . h = 0:9 M = 5. h = 1:1 M = 5.0 h > 1.0 h < 1.0 -2.0 177 1.0 at h De cr ea -1.3: The relation for Point-Jacobi.4.

250 iterations are required to reduce the error component of the worst eigenvector by a factor of roughly 0.. and there are (M .64) . is shown in Fig. 6 . . RELAXATION METHODS 2 .. j m jmax = 0:9942. 4 5 0 1=2M M The eigenvector structure of the Gauss-Seidel ODE matrix is quite interesting.1 + cos ( M +1 and the corresponding eigenvectors are given by j. Furthermore. they do not represent a common family for di erent values of M . ..3=4 1=2 7 2 6 7 6 0 1=8 . . The m with the largest amplitude is obtained with m = 1. the optimum stationary case.61) m = .. The equation for the nondefective eigenvalues in the ODE matrix is (for odd M ) + m ) m = 1 2 ::: M2 1 (9. The Jordan canonical form for M = 5 is j m jmax = cos M + 1 2 (9.1 1=2 3 1 6 0 . the error associated with the \worst" eigenvector is removed in half as many iterations. . producing waves that are higher in amplitude on one side (the updated side) than they are on the other. Hence the convergence rate is 2 1 CHAPTER 9.3=4 1=2 7 3 6 7 6 7 6 0 1=16 1=8 . 1)=2 defective eigenvalues with corresponding 178 Since this is the square of that obtained for the Point-Jacobi method.3=4 1=2 7 4 6 7 (9. 9.principal vectors. .4. If M is odd there are (M + 1)=2 distinct eigenvalues with corresponding linearly independent eigenvectors. . m + ~ m = cos m x sin j M + 1 m = 1 2 ::: M2 1 (9.. 7 .60) 6 0 1=32 1=16 1=8 . The eigenvectors are quite unlike the Point-Jacobi set. . . 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. . They are no longer symmetrical.63) 3 2 h~ i 7 6 h~ i 7 6 6 7 2~ 37 . .23. . .3=4 1=2 7 5 6 7 6 . .relation for h = 1. 7 . . . For M = 40.62) M +1 The .

0 σm Μ=5 2 defective λ m 0. The eigenvectors and principal vectors are all real. EIGENSYSTEMS OF THE CLASSICAL METHODS 1.0 -2.3=4 ) = . ~ 1 = c (1 .1 7 ~ = 6 4 7 (9. linked to (5) Jordan block The numerical solution written in full is thus ~ n .9.4: The relation for Gauss-Seidel.0 0.4 7 4 5 4 p 5 4 5 4 5 4 5 9=32 0 0 1 .66) .1 (4) Defective.0 -1.0 179 h = 1. For M = 5 they can be written 2 3 2 p 3 2 3 2 3 2 3 3=2 1=2 7 17 0 7 6 3=4 7 6 p3=4 7 607 6 2 7 6 0 7 6 6 7 6 6 6 0 7 ~ = 6 3=4 7 ~ = 6 p0 7 ~ = 6 0 7 ~ = 6 .1=4 = . 34 )n~ 1 2 3 3 1 1 2 2 (9.4. 3=16 7 7 607 6 0 7 6 . h = 1:0 M = 5. h )n~ 4 x h x + c (1 . 3=32 1 2 3 4 5 The corresponding eigenvalues are = .0 λmh Figure 9.65) 6 7 x 6 7 x 6 7 x 6 7 x 6 7 x 6 6 6 7 6 7 6 7 6 9=16 7 7 6 .

2x 0 0 0 x 4 4 3 7 7 7 7 (9. (. One can show that this can be written in the form given below for M = 5. the ODE matrix is B .2 + x 1 . the remaining eigenvalues are complex and occur in conjugate pairs. h)n.2. 3.4. equal and defective.1 x 0)B (1 . h)n + c h n (1 . (. If M is odd. one of the remaining eigenvalues is real and the others are complex occurring in conjugate pairs. !) + ! pm] = pm = cos m =(M + 1)] 2 2 2 1 If ! = 1. 2. is 1 1 2 x 0 0 6 . If M is even.2x x . If ! is chosen such that 4(1 . 2x + x x . h) x 1! 2! + c (1 .69) where zm = 4(1 . A matrix for the SOR method. 2x 4 . 2x + x x . 9. + x . The generalization to any M is fairly clear.180 CHAPTER 9. ~ x 3 4 1 5 2 2 3 4 5 + c (1 . + c h n(n . h)n.1 x 0)B (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 . !) + ! pm < 0 zm and m are complex. Two eigenvalues are real.1 + !pm 2 zm 2 m = 1 2 :::M 2 2 1 2 (9. h)n.2x + x x x 2x2x x 6 x + . x 0 6 1 6 . ~ + c (1 . the system is Gauss-Seidel.3 The SOR System 1 If = 1 and 2=! = x in Eq. and the following conditions hold: 1. If 4(1 .2 1). 2x x x 6 . RELAXATION METHODS " # n + c h n (1 . . 1) (1 .68) 7 7 5 Eigenvalues of the system are given by m + = . ASOR B . h)n~ x 5 1! 1 4 5 (9. The H .44.2x + x x . 2x + x 6 x .67) 9. !) + ! p = 0 ! is optimum for the stationary case. 2x + x x .2 1).

EIGENSYSTEMS OF THE CLASSICAL METHODS One can easily show that the optimum ! for the stationary case is and for ! = !opt 181 (9.5.(10 . much faster than both GaussSeidel and Point-Jacobi. The remaining linearly independent eigenvectors are all complex. Hence the worst error component is reduced to less than 0. h + h m . Hence the convergence rate is j m jmax = !opt . For M = 5 they can be written 2 3 2 3 2 p3(1 3 2 3 1=2 7 . 1.73) 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 = . and if h = 1.1 j .1 m sin 2 where m j M +1 (9.9. 1 (9.(10 + 2 2i)=9 = .relation reduces to that shown in Fig.72) !opt = 2= 1 + sin M + 1 For M = 40.4=3 (9.6 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. m = 1 .23 times its initial value in only 10 iterations. 2p2i)=9 = . there are two real eigenvectors and one real principal vector. the . This illustrates the fact that for optimum stationary SOR all the j m j are identical and equal to !opt . For odd M . In practical applications. the optimum value of ! may have to be determined by trial and error. and the bene t may not be as great. p m= m 2 m 2 1 2 q .74) 2 1 2 34 5 1 1 3 4 5 !opt p + i p .71) Using the explicit Euler method to integrate the ODE's.4. the optimum value for the stationary case. j mjmax = 0:8578.2=3 (2) Defectivep linked to = .70) !opt = 2= 1 + sin M + 1 ~m x m = = m. 9.

2p2i)h=9]n~ x n~ c 1 . h.182 1. ]~ x c (1 . H and C . This does not change the steadystate solution A. 9. are varied at each time step. 2h=3)n.5: The relation for optimum stationary SOR. The numerical solution written in full is ~n . 2h=3)n + c nh(1 . however. (10 + 2 2i)h=9] x c (1 . RELAXATION METHODS h = 1.0 -1. The nonstationary form of Eq.75) 9.0 -2. It is much simpler.0 CHAPTER 9. ~1 = + + + + c (1 . h = 1. 2h=3)n~ x p c 1 .39 is 1 .5 Nonstationary Processes In classical terminology a method is said to be nonstationary if the conditioning matrices.0 λmh Figure 9. but it can greatly a ect the convergence rate. This process changes the Point-Jacobi method to Richardson's method in standard terminology. ~ b .0 2 real defective λ m σm 2 complex λ m 1 real λ m Μ=5 0. For the Gauss-Seidel and SOR methods it leads to processes that can be superior to the stationary methods.0 0. M = 5. 4h=3)n~ x 1 2 1 2 3 2 3 4 5 4 5 1 (9. In our ODE b f approach this could also be considered and would lead to a study of equations with nonconstant coe cients. (10 . to study the case of xed H and C but variable step size.

~ 1 = X cm~ m Y (1 + e x m=1 n=1 m hn ) (9. However. for m = 1 2 M .79) treating it as a continuous function of the independent variable . This leads to the concept of selectively annihilating clusters of eigenvectors from the error terms as part of a total iteration process. the error term can theoretically be completely eliminated in M steps by taking hm = . we considered making the error exactly zero by taking advantage of some knowledge about the discrete values of m for a particular case.78) where Pe signi es an \Euler" polynomial.76. We will see that a few well chosen h's can reduce whole clusters of eigenvectors associated with nearby 's in the m spectrum. Now focus attention on the polynomial (Pe)N ( ) = (1 + h )(1 + h ) 1 2 (1 + hN ) (9. Let us choose the hn so that the maximum value of (Pe)N ( ) is as small as possible for all lying between a and b such that b a 0. Consider one of the error terms taken from M N ~ N ~ N . we seek b max a j(Pe)N ( )j = minimum with(Pe)N (0) = 1 (9.80) . Now we pose a less demanding problem. 9. the eigenvalues m are generally unknown and costly to compute. This is the basis for the multigrid methods discussed in Chapter 10.76) where the symbol stands for product. It is therefore unnecessary and impractical to set hm = . In the annihilation process mentioned after Eq.1= m for m = 1 2 : : : M . 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).1= m . Since hn can now be changed at each step.77) and write it in the form cm~ m Pe( m ) cm~ m x x N Y n=1 (1 + m hn ) (9. 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. Mathematically stated.9.5.

y . let us consider the following problem: Minimize the maximum error associated with the eigenvalues in the interval .85) hn = 2= 3 . As an example for the use of Eq. b a b This problem has a well known solution due to Markov. The error in the relaxation process represented by Eq. b TN a. cos (2n . + ( .2 . 1) 6 " #! (9. RELAXATION METHODS a. ~ m . ampli ed by the coe cients cm Q(1 + m hn). 1) n = 1 2 :::N (9.84 gives us the best choice of a series of hn that will minimize the amplitude of the error carried in the eigenvectors associated with the eigenvalues between b and a .81) TN (y) = cos(N arccos y) are the Chebyshev polynomials along the interval . ) cos (2n . TN . 9.82) q q N N 1 TN (y) = 2 y + y .76 is expressed in terms of a set of eigenvectors.86) . 9.1 y 1 and 2 2 (9. and it leads to the nonstationary step size choice given by ( " #) 1 = 1 . In relaxation terminology this is generally referred to as Richardson's method. Eq. It is 2 . ! (Pe)N ( ) = where (9. 1 + 1 y . The three values of h which satisfy this problem are (9. With each x eigenvector there is a corresponding eigenvalue. . 1 (9. b ! .184 CHAPTER 9.83) 2 are the Chebyshev polynomials for jyj > 1.84) b a h 2 b a 2N n Remember that all are negative real numbers representing the magnitudes of m in an eigenvalue spectrum.1 using only 3 iterations. a. 9.84.

9. 9. 2 hn under the same constraints as before.9. This calls for a study of the rational \trapezoidal" polynomial. 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.87) (9. 9. p 0) h = 4=(6 + 3) 1 2 3 p Return now to Eq. 8] =2 99 3 3 3 A plot of Eq. 3) h = 4=(6 . 9.92) .41.91) @ 1 A n 1 . the nonstationary formula 1 h( 0 + 0 ) n 2 n +1 (9.90) b max a j(Pt)N ( )j = minimum . 2 hn =1 =1 1 mC C + ~1 A m (9.37 on the condition that the explicit Euler method. If instead the implicit trapezoidal rule n+1 = n+ is used.1 have been reduced to less than about 1% of their initial values. Eq. with (Pt)N (0) = 1 (9.6 are h = 4=(6 .87 is given in Fig. Pt: 0 1 1 N B 1 + hn C Y (Pt )N ( ) = B 2 C (9. The values of h used in Fig.6 and we see that the amplitudes of all the eigenvectors associated with the eigenvalues in the range . namely that =1 0 1 M N 1+ h ~ N = X cm~ m Y B 2 n B x @ 1 m n 1 .88) where n p p o T (3) = 3 + 8] + 3 .2 .89) would result. This was derived from Eq.5. 9.76. 9.

4 −1.8 −1.005 (P ) (λ) e 3 0 −0.2 0 Figure 9.4 −0.4 0.02 0. RELAXATION METHODS 1 0.6 −0.6 −1.01 −0.6 (Pe )3 (λ) 0.2 −1 λ −0.2 −1 λ −0.8 −1.6: Richardson's method for 3 steps. .4 −0.8 0.015 −0.1 0 −2 −1.3 0.2 0 0.1.5 0.2 .186 CHAPTER 9.9 0.8 −0.7 0.2 0.8 −0.4 −1.01 0. minimization over .015 0.005 −0.6 −0.6 −1.02 −2 −1.

consider the iterative method (I + A )~ = (I . The values of h used in Fig.7 are h = 1 p h = 2 h = 2 ( 1) ( 1) 3 3 1 2 3 9.93) b hn b This process is also applied to problem 9. (You do not have to nd H . xk ) = (A + A )xk .85. Gn)A. .7. 2 =. A for the SOR method with A = B (4 : 1 . nd the eigenvalues of H . The error amplitude is about 1/5 of that found for (Pe) ( ) in the same interval of . A. A satisfy A~ m = mH~ m . For a linear system of the form (A + A )x = b.) Find the numerical values. 2. a n=1 2 N (9. PROBLEMS ! 187 The optimum values of h can also be found for this problem. The results for (Pt) ( ) are shown in Fig. Show that this iterative method can be written in the form H (xk . but we settle here for the approximation suggested by Wachspress n. Given a relaxation method in the form ~ H ~n = A~n . A )xn + b x (I + A )xn = (I . A )~ + b x where is a parameter. Using Appendix B. b 1 2 2 +1 1 +1 1 2 Determine the iteration matrix G if = . Then nd the corresponding j m j values. = N . 9. 9.6. 3.1=2.9. f show that 1 ~n = Gn ~ + (I . f ~ 0 1 1 2 where G = I + H . x x not just the expressions.2 1) and ! = !opt. Recall that the eigenvalues of H . 1 1 1 .6 Problems 1.2.

1.2 0.6 −1.1 0 −2 −1. .4 0.7: Wachspress method for 3 steps.2 −1 λ −0.8 0.3 0.8 −0.2 0 2 x 10 −3 1.2 . minimization over .2 −1 λ −0.5 −2 −2 −1. RELAXATION METHODS 1 0.4 −0.8 −1.4 −0.5 (Pt )3 (λ) 0 −0.6 (P ) (λ) t 3 0.5 −1 −1.9 0.4 −1.6 −0.5 1 0.5 0.7 0.4 −1.8 −0.2 0 Figure 9.188 CHAPTER 9.8 −1.6 −0.6 −1.

(c) the SOR method with the optimum value of !. u(1) = 1: @ u . Plot the logarithm of the L -norm of the residual vs.9. Solve the following equation on the domain 0 x 1 with boundary conditions u(0) = 0. Determine the asymptotic convergence rate. 3. Use second-order centered di erences on a grid with 40 cells (M = 39). 2 2 2 .5. 6x = 0 @x For the initial condition. PROBLEMS 189 4.6. Plot the solution after the residual is reduced by 2. use u(x) = 0. (b) the Gauss-Seidel method. and (d) the 3-step Richardson method derived in Section 9. Compare with the theoretical asymptotic convergence rate. the number of iterations. and 4 orders of magnitude. Iterate to steady state using (a) the point-Jacobi method.

RELAXATION METHODS .190 CHAPTER 9.

2 1).3) xx~ = x x 191 2 2 2 2 1 . There are many variations of the process and many viewpoints of the underlying theory. the space-frequency (or wavenumber) of the corresponding eigenvectors also increase.2 1)~ = X 1 D(~ ) X . respectively.3. The viewpoint presented here is a natural extension of the concepts discussed in Chapter 9. Notice that as the magnitudes of the eigenvalues increase.2 and 4.Chapter 10 MULTIGRID The idea of systematically using sets of coarser grids to accelerate the convergence of iterative schemes that arise from the numerical solution to partial di erential equations was made popular by the work of Brandt.3. if the eigenvalues are ordered such that j j j j 1 2 2 j Mj (10. 10. That is.3.1) then the corresponding eigenvectors are ordered from low to high space frequencies.m sin(mx) (10.1 Motivation 10. ~ (10. Note that @ sin(mx) = . This has a rational explanation from the origin of the banded matrix. The eigenvalues and eigenvectors are given in Sections 4.2) @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 .1.

10.192 1 CHAPTER 10. As we saw in Section 9. the property can be restored by using h < 1.2 Properties of the Iterative Method The second key motivation for multigrid is the following: Many iterative methods reduce error components corresponding to eigenvalues of large amplitude more e ectively than those corresponding to eigenvalues of small amplitude.2 + 2 cos m .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 .m .m m << M (10. When an iterative method with this property is applied to a matrix with the above correlation between the modulus of the eigenvalues and the space frequency of the eigenvectors. of the Gauss-Seidel method and. This is to be expected of an iterative method which is time accurate. In fact. error components corresponding to high space frequencies will be reduced more quickly than those corresponding to low space frequencies. The classical point-Jacobi method does not share this property.1. this method produces the same value of j j for min and max . However. and X ~ . For this matrix one nds. We have seen that X . Therefore. MULTIGRID where D(~ ) is a diagonal matrix containing the eigenvalues. exactly the opposite 2 behavior.2.4) m= M +1 x Hence. of the Richardson method described in Section 9. a sine synthesis.5. This is the key concept underlying the multigrid process. the operation 1 D(~ ) represents the numerical approximation of the multiplication of the x appropriate sine wave by the negative square of its wavenumber. This is consistent with the physics of di usion as well. 9. However. It is also true.4. the complete counterexample of the above association is contained in the eigensystem 1 for B ( 2 1 1 ). by design. from Appendix B. One nds that 1 M + 1 .1. for example. this correlation is not necessary in general. as shown in Fig. 2 2 2 2 2 10. . ~ represents a sine transform. the correlation of large magnitudes of m with high space-frequencies is to be expected for these particular matrix operators.

We start with some any.5) where the matrix formed by the product CAb has the properties given above. The eigenvalues. or it can be \contrived" by further conditioning.27. This form can be arrived at \naturally" by simply replacing the derivatives in the PDE with di erence schemes. ~ b ] = 0 f (10. 5. the vector ~ b represents the boundary conditions and the forcing function. m .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~ . The m are fairly evenly distributed between their maximum and minimum values. The iterative procedure used greatly reduces the amplitudes of the eigenvectors 1 associated with eigenvalues in the range between 2 j jmax and j jmax. In Eq. 10. 9. 3. r where ~ = C Ab~ . 3. 9. 2. we are led to the starting formulation C Ab ~ 1 . The problem is linear.5. of the matrix are all real and negative. the residual. but for clarity we suppose that the three-step Richardson method illustrated in Fig.10. as in the example given by Eq.8) .11. if f ~ 1 is a vector representing the desired exact solution. At the end of the three steps we nd ~ . Having preconditioned (if necessary) the basic nite di erencing scheme by a procedure equivalent to the multiplication by a matrix C . The basic assumptions required for our description of the multigrid process are: 1. ~ b] r f (10.7) where A CAb (10. These conditions are su cient to ensure the validity of the process described next. as in the examples given by Eq. We do not attempt to develop an optimum procedure here.6 is used. and initial guess for ~ 1 and proceed through n iterations making use of some iterative process that satis es property 5 above. 4. The eigenvectors associated with the eigenvalues having largest magnitudes can be correlated with high frequencies on the di erencing mesh. THE BASIC PROCESS 193 basic properties.2. ~ = 0 e r (10.

10. For a 7-point example 2 6 6 6 6 6 6 6 6 6 6 6 4 e e e e e e e 2 4 6 1 3 5 7 3 2 7 6 7 6 7 6 7 6 7 6 7=6 7 6 7 6 7 6 7 6 7 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 (10.7 from the left by P and. Next we construct a permutation matrix which separates a vector into two parts. and the eigenvalues of the Richardson process in the form: M= M ~ = X cm~ m Y ( m hn)] + X cm~ m Y ( mhn)] e x x 2 3 3 m=1 n=1 m=M=2+1 | very low amplitude {z n=1 (10. MULTIGRID (10.12) 1 The operation PAP . 10. we can be sure that the high frequency content of ~ has been greatly reduced (about 1% or less of its original value in the initial guess). assumption 4 ensures that the error has been smoothed. one containing the odd entries.11) Multiply Eq.~ e CHAPTER 10.10) } Combining our basic assumptions. we can write PA P . P ]~ = P ~ e r (10.9) Thus our goal now is to solve for ~ .7 for ~ then e ~1 = ~ . 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). We can write the exact solution for ~ in terms of e e the eigenvectors of A.194 If one could solve Eq. 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. e In addition.13) Notice that e r e A ~e + A ~o = ~e 1 2 (10.14) . since a permutation matrix has an inverse which is its transpose.

It is that there is some connection between ~ e and ~ o brought about by the smoothing property of the Richardson e e relaxation procedure. THE 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. e no operations on ~ are required or justi ed. Hence. Since the top half of the frequency spectrum has been removed. ea and eb are zero. in this formulation. 10. it is reasonable to suppose that the odd points are the average of the even points. For example 1 (e + e ) e 2 a 1 (e + e ) e 2 1 (e + e ) or ~ o = A0 ~ e e e (10. no aliasing of these functions can occur in subsequent steps.2.19) or Ac~ e .15) e 2 1 (e + e ) e b 2 1 2 3 2 4 5 4 6 2 7 6 (10. ~ e = 0 e r where Ac = A + A A0 ] 1 2 2 . the averaging of ~ is our only approximation. Finally.18) (10.14 reduces to A ~ e + A A0 ~ e = ~ e e e r 1 2 2 (10. At this point we make our one crucial assumption. r If the boundary conditions are Dirichlet. 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.16) 0 0 1 With this approximation Eq.17) (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. notice that.10.

196 CHAPTER 10. the eigenvector structure is di erent from that given in Eq. all of the properties given in Section 10. in fact.1 1=2 7 6 7 = 4 1=2 . It is easy to show that the high space-frequencies still correspond to the eigenvalues with high magnitudes. our 7-point example would produce 2 .23) 4 5 0 0 2 1 2 .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 . B. In the particular case illustrated in Fig.2 6 6 6 6 6 6 . 10.1 are met. The eigensystem is given by b b Eq. If the original A had been B (7 : 1 . For Neumann conditions. the matrix on the coarse mesh.2 .55 for Dirichlet conditions.2 4 A1 .2 1 1 1 .2 .1 = 6 PAP 6 1 6 6 6 1 6 6 4 .2 1). When eb = eM .2 .1 1=2 7 5 1=2 .1.19. 9.16 changes to 2 1 0 03 7 16 A0 = 2 6 1 1 0 7 60 1 17 (10. and.2 6 . All of them go through zero on the left (Dirichlet) side. MULTIGRID The form of Ac. A in the 1-D model equation is B (1 ~ 1) where ~ = .2 1 1 1 1 1 1 .20) and Eq. 10.18 gives z }| 2 . If Neumann conditions are on the left.1]T . 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. ea = e . 1) xjm = sin j 2M + 1 m=1 2 M (10. and all of them re ect on the right (Neumann) side. the example in Eq. However. In the present case they are given by " !# (2m .2 .1 { 3 7 (10.2 ::: .2 . 10. the interpolation formula in Eq.21) 5 If the boundary conditions are mixed Dirichlet-Neumann.22) and are illustrated in Fig.15 must be changed. 10. is completely determined by the choice of the permutation matrix and the interpolation approximation.1. eb is equal to eM .

2 1)~ = ~ on the e r 1 B (1 . Given ~ e e e computed on the coarse grid (possibly using even coarser grids). THE BASIC PROCESS 197 X Figure 10.24) 5 which gives us what we might have \expected. The permutation matrix remains the same and both A and A in the partitioned matrix PAP . In order to complete the process. Therefore.1: Eigenvectors for the mixed Dirichlet{Neumann case.26) ~ m = sin j m x m = . Recall that our goal is ne mesh to 2 e r to solve for ~ . we can compute ~ o e using Eq. cos m=1 2 M M +1 M +1 1 1 . (10.2 1 .15. we must now determine e ~ e and ~ . and thus ~ .2 4 A1 . which will provide us with the solution ~ 1 using Eq. are unchanged (only A is modi ed by putting .1 1=2 7 6 7 = 4 1=2 . Ac = Xc cXc.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 .2 6 . 10.2.1 in the lower right element). the relationship between e e In order to examine this relationship. we need to consider the eigensystems of A and Ac: A = X X." We will continue with Dirichlet boundary conditions for the remainder of this Section.9.25) For A = B (M : 1 . 10. we have reduced the problem from B (1 .2 1)~e = ~ e on the next coarser mesh.1 7 5 1=2 . we can construct the coarse matrix from 1 1 2 4 }| z 2 . At this stage.2 1) the eigenvalues and eigenvectors are m j=1 2 M (10.1=2 { 3 1=2 7 (10.10.

. cos M + 1 x j=1 2 M (10. Xc. we obtain ( c) = . corresponding to ~ = sin j = . cos M2+ 1 (~ c) = sin j M2+ 1 j = 1 2 x Mc (10. = . (~ c) e x c r c (10. 1 . As M increases. 5 0 (10. 1)=2 is the size of Ac. 7 6 .0:003649. x Now let us consider the case in which all of the error consists of the eigenvector component ~ .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 (10.27) M +1 For example. that is...31) 2 3 1 607 6 7 = Xc . In addition. ~ e = Xc . Then the residual is x e x ~ = A~ = ~ r x x (10. If we restrict our attention to odd M .28) For M = 51 (Mc = 25).2 1) are 2 ( c) = .2 1 . one can easily see that (~ c) coincides with ~ at x x every second point of the latter vector.0:007291 = 1:998 . The exact solution on the coarse grid x x satis es ~ e = A. 6 .32) 2 3 1=( c) 6 0 7 6 7 = Xc 6 .e.33) 1 1 1 1 1 1 1 1 1 . with M = 51. The eigenvalue and eigenvector corresponding to m = 1 for the matrix Ac = 1 B (Mc 1 . i.198 CHAPTER 10. the most di cult error mode to eliminate is that with m = 1. MULTIGRID Based on our assumptions. 7 c 6 7 4 . ( c) approaches 2 . then Mc = (M . it contains the even elements of ~ . 7 4 5 0 (10.30) since (~ c) contains the even elements of ~ . ~ = ~ .

2 1)~ = ~ ee re (10. .37) c 4 In our case.2 1) comes from a discretization of the di usion equation. However. in each case the appropriate permutation matrix and the interpolation approximation de ne both the down.34) (10.39) Applying the discretization on the coarse grid with the same preconditioning matrix as used on the ne grid gives. we must multiply the result by 2. since xc = 2 x. but they vary depending on the problem and the user. This is equivalent to solving or 1A ~ = ~ e r 2 ce e (10. carried to even coarser grids before returning to the nest level. and usually is. THE BASIC PROCESS = ( c) 1 1 199 (~ c) x 1 1 1 (~ ) 2 xc 1 (10.35) Since our goal is to compute ~ = ~ .2 1) = x B (Mc : 1 .37.2 1) 4 c c 2 2 2 (10. in addition to interpolating ~ e to the ne grid e x e (using Eq. Thus we see that the process is recursive.and up-going paths. x C x B (Mc : 1 . quite important but they are not discussed here.38) C= x I 2 (10.10.36) 1 B (M : 1 .2 1) and the preconditioning matrix C is simply 2 (10. which gives Ab = x B (M : 1 . 10. the matrix A = B (M : 1 . The details of nding optimum techniques are. 10.2. The reduction can be.15).2 1) = 1 B (Mc : 1 . The problem to be solved on the coarse grid is the same as that solved on the ne grid. The remaining steps required to complete an entire multigrid process are relatively straightforward.40) which is precisely the matrix appearing in Eq. obviously.

(1) (2) 2 (1) 1 2 1 3. Solve the problem A ~ = ~ on the coarse grid exactly: e r ~ = A.21). Transfer (or restrict) ~ to the coarse grid: r ~ =R ~ r r (10. This gives ~ = Gn1 ~n + (I . f r ~ = AGn1 ~n . . 9.11. 10. f (10. which can be easily generalized to a process with an arbitrary number of grids due to the recursive nature of multigrid.44) In our example in the preceding section. A (10. Perform n iterations of the selected relaxation method on the ne grid. ~ e r 2 (2) (2) (2) 2 1 (2) 2 (10." Some form of weighted restriction can also be used. f .43) 1 1 1 1 1 (1) (1) (1) 1 1 1 1 1 1 2. Note that the coarse grid matrix denoted A2 here was denoted Ac in the preceding section. Gn1 ) A.46) 1 2 See problem 1 of Chapter 9. Eq.3 A Two-Grid Process CHAPTER 10. Extension to nonlinear problems requires that both the solution and the residual be transferred to the coarse grid in a process known as full approximation storage multigrid. f ~ (10. AGn1 A. the rst three rows of the permutation matrix P in Eq. This type of restriction is known as \simple injection.200 10. 1. the restriction matrix is 2 3 0 1 0 0 0 0 0 R =6 0 0 0 1 0 0 0 7 (10. starting with ~ = ~n. f = AGn1 ~n + A (I . Gn1 ) A..45) 4 5 0 0 0 0 0 1 0 that is. Call the result ~ .41) 1 (1) 1 (1) 1 1 1 where G = I + H. MULTIGRID ~ We now describe a two-grid process for the linear problem A ~ = f .42) and H is de ned as in Chapter 9 (e.g. Next compute the residual based on ~ : ~ ~ ~ ~ = A ~ .

R A]Gn1 A. Extension to four or more grids proceeds in similar fashion.3.51) In this expression n is the number of relaxation steps on grid 2. with (I . 10. apply the two-grid process recursively. Combining these steps. . A = 4 B (Mc : 1 . I . R A ]Gn2 2 3 2 3 3 1 3 2 2 2 2 2 3 3 2 3 (10. It is at this stage that the generalization to a multigrid procedure with more than two grids occurs. G )A. I A. where 2 2 3 2 G = I .50 by replacing A.49) (10.40).2 1).15. f + 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. f ~ ~ +1 1 2 2 1 2 1 1 1 2 2 1 2 1 1 1 1 (10.50) Thus the basic iteration matrix is I .10.48) which follows from Eq. I A. 4. If this is the coarsest grid in the sequence. Otherwise. 10. I and R are the transfer operators between grids 2 and 3. one obtains ~n = I . . Transfer (or prolong) the error back to the ne grid and update the solution: ~n = ~ . A TWO-GRID PROCESS 2 2 201 Here A can be formed by applying the discretization on the coarse grid. I A. I A. and A is obtained by discretizing on grid 3. R A]Gn1 ~ n . In the 1 preceding example (eq. I ~ e +1 (1) 1 (2) 2 (10. The basic iteration matrix for a three-grid process is found from Eq. solve exactly.47) 3 7 7 7 7 7 7 7 7 7 7 7 5 In our example. 10. 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.

Determine the asymptotic convergence rate. 10. and 4 orders of magnitude. 3. Derive Eq. Repeat problem 4 of Chapter 9 using a four-grid multigrid method together with (a) the Gauss-Seidel method.5. the number of iterations. Plot the logarithm of the L -norm of the residual vs. Solve exactly on the coarsest grid.202 10. 2 .51. (b) the 3-step Richardson method derived in Section 9. 2. MULTIGRID 1. Plot the solution after the residual is reduced by 2.4 Problems CHAPTER 10. Calculate the theoretical asymptotic convergence rate and compare.

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

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

2. However.4) @t @x @x2 @x3 @x4 Choose periodic boundary conditions and impose an initial condition u = ei x.({z+ 2 )t}] (11. Thus a ) u = e.2 The Modi ed Partial Di erential Equation ( xu)j = 2 1 x 42 x @u .2.ia . THE MODIFIED PARTIAL DIFFERENTIAL EQUATION 11. 11. We are lead to the expression 2 ! 3 ! ! ! @ 2 u + x3 @ 3 u . 2} ei x. Substituting this into Eq. Under these conditions there is a wave-like solution to Eq. Notice that Eq. x can be small but it is not zero.3 is consistent with Eq. We proceed next to investigate the implications of this concept. a x2 @ 3 u + a x3 @ 4 u + : : : (11.11. 11.1 gives @u = . @x j 205 First carry out a Taylor series expansion of the terms in Eq. when we use a computer to nd a numerical solution of the problem.5) | | amplitude phase . 11. 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. 11. 11. 2 ({z.a @u + @ 2 u + @ 3 u + @ 4 u (11. 2 ( . 2 ) s = . i 3 + 4 or r = .4 of the form u(x t) = ei xe(r+is)t provided r and s satisfy the condition r + is = . This means that each term in the expansion given by Eq. Consider the simple linear partial di erential equation @u = . 2 . We refer to Eq. 11.3 is excited to some degree. 11.1.3) @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. (a + 2 ) The solution is composed of both amplitude and phase terms. since the two equations are identical when x ! 0.3 as the modi ed partial di erential equation.1.2 to Eq. 11.a @u + a x @ 2 u .

5.2 .1 + 6uj . Therefore. This is tantamount to deliberately adding dissipation to the PDE.1 + 3uj + 2uj+1) = 1 (uj.6) The antisymmetric component of this operator is the fourth-order centered di erence operator. . that the same is true for any centered di erence approximation of a rst derivative. one could choose = 1=2 in Eq. Note that the use of one-sided di erencing schemes is not the only way to introduce dissipation.1 + 8uj+1 . any departure from antisymmetry in the matrix di erence operator must introduce dissipation (or ampli cation) into the modi ed PDE. Under the same condition. the coe cients of the odd derivatives. 11. Our purpose here is to investigate the properties of one-sided spatial di erencing operators relative to centered di erence operators.4. and the phase depends only on a and . Therefore. 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 x2 =6. This we refer to as dispersion. Referring to the modi ed PDE. 4uj+1 + uj+2)] (11. 8uj.3 we have = . the choice of a forward di erence scheme ( = . 2 > 0 and hence the amplitude of the solution decays. uj+2) 12 x + (uj. NUMERICAL DISSIPATION It is important to notice that the amplitude of the solution depends only upon and . As a corollary.206 CHAPTER 11. the coe cients of the even derivatives in Eq. the numerical phase speed is dependent upon the wavenumber . the choice of a backward di erence scheme ( = 1) produces a modi ed PDE with . a third-order backward-biased scheme is given by ( xu)j = 6 1 x (uj. Any symmetric component in the spatial operator introduces dissipation (or ampli cation).2 . For example. If the wave speed a is positive. Therefore. Biased schemes use more information on one side of the node than the other. By examining the term governing the phase of the solution in Eq. this operator produces fourth-order accuracy in phase with a third-order dissipative term. 11. 11. the phase speed of the numerical solution is less than the actual phase speed. we see that the speed of propagation is a + 2 .1) is equivalent to deliberately adding a destabilizing term to the PDE.2 . 4uj. by using the antisymmetry of the matrix di erence operators. 6uj. Eq. Furthermore. The resulting spatial operator is not one-sided but it is dissipative. The symmetric component approximates x3 uxxxx=12. 11. One can easily show.2.

the corresponding fully-discrete matrix operator is 0 2 !3 ! 2 ! 31 1 4 ah + ah 25 1 . giving !2 (n+1) = u(n) .1 2 x j+1 j. .8) @t @x @t2 @x2 Now replace the space derivatives with three-point centered di erence operators.7) First replace the time derivatives with space derivatives according to the PDE (in this case. backward di erencing must be used if the wave speed is positive. 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. and forward di erencing must be used if the wave speed is negative.9) uj j j +1 j j . the linear convection equation @u + a @u = 0). u(n) ) + 1 ah (u(n) . THE LAX-WENDROFF METHOD 11.a @u (11. The quantity j ah j is known as the x Courant (or CFL) number. For periodic boundary conditions.3 The Lax-Wendro Method 207 In order to introduce numerical dissipation using one-sided di erencing. Consider the following Taylor-series expansion in time: 2 u(x t + h) = u + h @u + 1 h2 @ u + O(h3) @t 2 @t2 (11.1 2 x This is the Lax-Wendro method applied to the linear convection equation. known as the Lax-Wendro method. There is no intermediate semi-discrete stage. Thus @t @x @ 2 u = a2 @ 2 u @u = . x M x M for m = 0 1 : : : M .11.3. 2u(n) + u(n) ) (11. 1 ah (u(n) . ah + ah 2 5A ~ ~ n+1 = Bp @ 2 x u un x x 2 x x The eigenvalues of this matrix are ! ah 2 1 . cos 2 m . It is equal to the ratio of the distance travelled by a wave in one time step to the mesh spacing. This explicit method di ers conceptually from the methods considered previously in which spatial di erencing and time-marching are treated separately. i ah sin 2 m m =1. Next we consider a method which introduces dissipation independent of the sign of the wave speed. It is a fully-discrete nite-di erence scheme. ah 2 1 4.

u(jn+1))] ~ (11. 11.208 CHAPTER 11. Recall that the odd derivatives on the right side lead to unwanted dispersion and the even derivatives lead to dissipation (or ampli cation. . a2h2 ) @ 3 u = .11) 2 x u +1 ~ which can be shown to be identical to the Lax-Wendro method. which is given by @u + a @u = . Cn) @xu 4 @x4 This term has the appropriate sign and hence the scheme is truly dissipative as long as Cn 1. a8h ( x2 . Recall MacCormack's timemarching method. x u(jn+1) = 1 u(jn) + u(jn+1) . a h8 x (1 . the leading error term in the Lax-Wendro method is dispersive and proportional to a ( x2 . A closely related method is that of MacCormack. depending on the sign). this approach leads to u(jn+1) = u(jn) . a2 h ( x2 . 1 Or vice-versa for nonlinear problems. u(jn)1) ~ . these should be applied alternately. 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. For the linear convection equation. a x2 (1 . ah (~(jn+1) . ah (u(jn) . a ( x2 . Therefore. a2h2 ) @ u = . The two methods di er when applied to nonlinear hyperbolic systems.8. Hence MacCormack's method has the same dissipative and dispersive properties as the Lax-Wendro method.6 n @x3 @x3 6 The dissipative term is proportional to 2 4 2 2 4 2 @ .9 and converting the time derivatives to space derivatives using Eq. presented in Chapter 6: un+1 = un + hu0n ~ 1 ~ ~ (11. 11. The two leading error terms appear on the right side of the equation. 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.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. however. a2h2 ) @ 3 u . C 2 ) @ 3 u .

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. This is achieved by the following point operator: . However.5. This is avoided in the Lax-Wendro scheme. we present the basic ideas of ux-vector and ux-di erence splitting. We can rewrite Eq. 11. as a result of their superior exibility.1 if a 0. we present two splitting techniques commonly used with upwind methods. we saw that numerical dissipation can be introduced in the spatial di erence operator using one-sided di erence schemes or.. Now for the a+ ( 0) term we can safely backward di erence and for the a. more generally. In this section. these are of mixed sign. but entirely equivalent.4. Consider again the linear convection equation: @u + a @u = 0 (11.1 + uj+1) + jaj(. When a hyperbolic system of equations is being solved.1 + 2uj .2. by adding a symmetric component to the spatial operator. then a a.uj. In the next two sections.1.uj. manner. the wave speeds can be both positive and negative.13) This approach is extended to systems of equations in Section 11. When the ow is subsonic. UPWIND SCHEMES 11. we see that a stable discretization is obtained with = 1 if a 0 and with = . = 0. ) @u = 0 a = a 2 jaj @t @x + = a 0 and a. 11.1x a(. then a+ = 0 and If a 0. that is. uj+1)] (11. For example. 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 the basic concept behind upwind methods. ( 0) term forward di erence. For more subtle aspects of implementation and application of such techniques to .12) @t @x where we do not make any assumptions as to the sign of a. In order to apply one-sided di erencing schemes to such systems.4 Upwind Schemes 209 In Section 11. some form of splitting is required.12 as @u + (a+ + a. the direction of the one-sided operator (i.e.a( xu)j = 2.11. a. From Eq. These are by no means unique. With this approach. The above approach to obtaining a stable discretization independent of the sign of a can be written in a di erent. the eigenvalues of the ux Jacobian for the onedimensional Euler equations are u u + a u . schemes in which the numerical dissipation is introduced in the spatial operator are generally preferred over the Lax-Wendro approach. if a 0. Alternatively. = a 0.

where += (11.210 CHAPTER 11.19) . we need to apply a backward di erence if the wave speed. To accomplish this.15) i @t @x where the wave speeds.1 Flux-Vector Splitting Recall from Section 2. @w term.4. @w = 0 (11.1X in the spatial terms gives @Xw + @X +X . is positive.5 that a linear. + contains the positive eigenvalues and . . constant-coe cient. A. are the eigenvalues of the Jacobian matrix. In order to apply a one-sided (or biased) spatial di erencing scheme. = . contains the negative eigenvalues.j j (11. Premultiplying by X and inserting the product @x X . We can now rewrite the system in terms of characteristic variables as @w + @w = @w + + @w + .X . i.14) @t @x @t @x can be decoupled into characteristic equations of the form @wi + @wi = 0 (11. 11. NUMERICAL DISSIPATION nonlinear hyperbolic systems such as the Euler equations.18) @t @x @t @x @x The spatial terms have been split into two components according to the sign of the wave speeds. let us split the matrix of eigenvalues. hyperbolic system of partial di erential equations given by @u + @f = @u + A @u = 0 (11.16) +j j . and the wi's are the characteristic variables.1Xw = 0 @t @x @x (11. i. the reader is referred to the literature on this subject.1Xw + @X . We can use backward di erencing for the + @w term and forward @x di erencing for the . into two components such that = ++ .17) 2 2 With these de nitions. and a forward di erence if the wave speed is negative.

the Jacobians of the split ux vectors are required.4. one must nd and use the new Jacobians given by A++ = @f @u + (11.1 (11. as discussed in Appendix C. For the Euler equations.X . A++ has eigenvalues which are all positive.11. and A. we are in e ect solving the characteristic equations in the desired manner. u = 0 @t @x @x Finally the split ux vectors are de ned as f + = A+u and we can write (11. has all negative eigenvalues. Note that f = f+ + f. @u @u Therefore. In the nonlinear case. UPWIND SCHEMES With the de nitions2 211 A+ = X +X . @f + 6= A+ @f .21) (11.20) @u + @A+u + @A. = X . 2 With these de nitions A+ has all positive eigenvalues.. . term. This approach is known as ux-vector splitting. has all negative eigenvalues. f is also equal to Au as a result of their homogeneous property. (11. and A. = A.23) @t @x @x In the linear case.26) For the Euler equations. 6= A. When an implicit time-marching method is used.25) A.24) Thus by applying backward di erences to the f + term and forward di erences to the f . we obtain A.22) f . the de nition of the split uxes follows directly from the de nition of the ux. f = Au. = @f @u . u @u + @f + + @f .1 and recalling that u = Xw. = 0 (11.. (11.

31) 2 Now.2 Flux-Di erence Splitting Another approach. we premultiply by X to return to the original variables and insert the product X .33) 2 where jAj = X j jX . NUMERICAL DISSIPATION 11. Now. as in Chapter 5.4.1X after and j j to obtain 1 1 X gj+1=2 = 2 X X .30) or 1 gj+1=2 = 1 (wL + wR) + 2 j j (wL . gj+1=2. the uxes must be evaluated at cell boundaries. we require ( (w ) (^i)j+1=2 = i(wi)L if i > 0 g (11. (wi)R] g (11. The centered approximation to gj+1=2.32) and thus 1 f^j+1=2 = 1 (fL + fR ) + 2 jAj (uL . wL and wR. This is achieved with 1 1 (^i)j+1=2 = 2 i (wi)L + (wi)R ] + 2 j ij (wi)L .1X (wL . wR ) ^ (11.29) i i R if i < 0 where the subscript i indicates individual components of w and g. hyperbolic system of equations @w + @w = 0 (11.27) @t @x The ux vector associated with this form is g = w. 11.212 CHAPTER 11. wR ) ^ (11.28) In order to obtain a one-sided upwind approximation. is given by 1 gj+1=2 = 2 (g(wL) + g(wR)) ^ (11. we consider the numerical ux at the interface between nodes j and j + 1. as in Eq. which is nondissipative. constant-coe cient.1X (wL + wR) + 2 X j jX . is known as ux-di erence splitting. In a nite-volume method.1 (11.34) . as a ^ function of the states to the left and right of the interface. more suited to nite-volume methods.19. respectively. We again begin with the diagonalized form of the linear. uR) (11.

uR) (11.35) R if all i s < 0 giving pure upwinding. 11. In the linear. In this case. In order to resolve this. respectively. there is some ambiguity regarding the de nition of jAj at the cell interface j + 1=2. u = Xw.11. Hence satisfaction of Eq.5. consider a situation in which the eigenvalues of A are all of the same sign. constant-coe cient case.37) For the Euler equations for a perfect gas. 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.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. in the nonlinear case. and A = X X . this leads to an upwind operator which is identical to that obtained using ux-vector splitting.A. jAj = A if they are all negative. Eq. 11. jAj = .1.35 is obtained by the de nition 1 f^j+1=2 = 2 (fL + fR ) + 1 jAj+1=2j (uL . fR = Aj+1=2 (uL . We have also seen that such dissipation can be introduced by adding a symmetric component to an antisymmetric (dissipation-free) operator. 3 Note that the ux Jacobian can be written in terms of u and H only see problem 6 at the end . 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. If the eigenvalues of A are all positive.38) uj+1=2 = p + p L R p H +p H L L Hj+1=2 = (11. However. ARTIFICIAL DISSIPATION 213 and we have also used the relations f = Xg.3 ( 11.37 is satis ed by the ux Jacobian evaluated at the Roe-average state given by p LuL + p R uR (11. Thus we can generalize the concept of upwinding to include any scheme in which the symmetric portion of the operator is treated in such a manner as to be truly dissipative.36) 2 if Aj+1=2 satis es fL . of this chapter.

applying x = x .214 For example. 4uj. but is beyond the scope of this book.43.uj+1 + 2uj . gives a s (11. uj. Similarly. 4uj.1 (11. With an appropriate value of .1 + 6uj .43) x f = x f + x (jAju) where jAj is de ned in Eq.44) x (uj.42) x (Au) = x (Au) + x (jAju) or a s (11. 11. this approach can be related to the upwind approach. This symmetric operator approximates x3 uxxxx and thus introduces a third-order dissipative term. this often provides su cient damping of high frequency modes without greatly a ecting the low frequency modes. A second-order backward di erence approximation to a 1st derivative is given as a point operator by ( xu)j = 1 (uj. a s With appropriate choices of x and x. NUMERICAL DISSIPATION s a ( xu)j = . The appropriate implementation is thus a s (11. the reader should consult the literature.2 .1 2 x 2 x a s Applying x = x + x to the spatial derivative in Eq. x is stable if i 0 and unstable if i > 0.41) i x = i x + j ij x Extension to a hyperbolic system by applying the above approach to the characteristic variables. uj.40) ( x u)j = uj+1 . 11. It is also sometimes referred to as arti cial di usion or arti cial viscosity. s ( xu)j = 11. 11.36 and 11. For details of how this can be implemented for nonlinear hyperbolic systems. s It is common to use the following operator for x (11. as in the previous two sections. A more complicated treatment of the numerical dissipation is also required near shock waves and other discontinuities. 4uj+1 + uj+2) where is a problem-dependent coe cient. This is particularly evident from a comparison of Eqs. let CHAPTER 11.34.2 .1 + 3uj ) 2 x .6 Problems 1.15 is stable if i 0 and a s unstable if i < 0. The second spatial term is known as arti cial dissipation.

2. then derive the symmetric and skew-symmetric matrices that have the matrix operator as their sum. Find the modi ed wavenumber for the operator given in Eq. nd j j for the combination of this spatial operator with 4th-order Runge-Kutta time marching at a Courant number of unity and plot vs.44.6. Show that the ux Jacobian for the 1-D Euler equations can be written in terms of u and H . nd the derivative which is approximated by the corresponding symmetric and skew-symmetric operators and the leading error term for each. Form the plus-minus split ux vectors as in Section 11.37. Find the matrix jAj.2. x for 0 x . 11.3 to see how to construct the symmetric and skew-symmetric components of a matrix. x for 0 x .6. 5. Show that the use of the Roe average state given in Eqs. Plot the real and imaginary parts of x vs. 3. and 1=48.11. Using Fourier analysis as in Section 6.6. . 2. Consider the spatial operator obtained by combining second-order centered differences with the symmetric operator given in Eq. Plot the real and imaginary parts of x vs. x for 0 x .) (b) Using a Taylor table. 11.6.6. x for 0 x . Using Fourier analysis as in Section 6. Consider the hyperbolic system derived in problem 8 of Chapter 2. x for 0 x . PROBLEMS 215 (a) Express this operator in banded matrix form (for periodic boundary conditions). Find the modi ed wavenumber for this operator with = 0 1=12 1=24.2. Using Fourier analysis as in Section 6. 11.4.1. 6. 4. Find the modi ed wavenumber for the rst-order backward di erence operator. x for 0 x . (See Appendix A. nd j j for the combination of this spatial operator with 4th-order Runge-Kutta time marching at a Courant number of unity and plot vs. 11. Plot the real and imaginary parts of x vs. 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.39 leads to satisfaction of Eq.38 and 11.

NUMERICAL DISSIPATION .216 CHAPTER 11.

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

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

requiring a tridiagonal solver if the di usion term is central di erenced. and a rst-order time-march method is usually unsatisfactory. stability. but the di usion operator is now implicit. hAd]~ n u +1 +1 ~ = I + hAc]~ n + h(bc) u = un ~ +1 (12. 12.7 are often applied in a predictor-corrector sequence. 12. 12. I + hAc]~ n + h(bc) u ~ I + hAd + hAc]~ n + h(bc) +O(h ) u | {z } 1 2 (12. on this basis.9) I . their selection is arbitrary. We have yet to determine its .6 have identical orders of accuracy in the time approximation.8. = I + hAd + h Ad + if h jjAdjj < 1. Since numerical sti ness is generally much more severe for the di usion process. In practical applications equations such as 12.2.12. hAd].6 with the same order of accuracy is given by the expression ~n u +1 Original Unfactored Terms where in this approximation we have used the fact that 1 2 2 = = ~ I . where jjAd jj is some norm of Ad ]. this factored form would appear to be superior to that provided by Eq. 12.7) u {z } | 2 2 | {z } un ~ ~n u +1 +1 = = ~ I + hAc]~ n + h(bc) u I + hAd]~n u +1 (12. another way to approximate Eq. as before. 2 We do not suggest that this particular method is suitable for use.8) Factoring can also be useful to form split combinations of implicit and explicit techniques. 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 ) (12.10) The convection operator is applied explicitly. Therefore. This time a predictor-corrector interpretation leads to the sequence un ~ I . However.7 and the original unfactored form Eq. hAd ]. the important aspect of stability has yet to be discussed. 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. For example.

is 4 and My . un = A u + A u + (bc) + O(h ) ~ c n d n h +1 +1 2 We should mention here that Eq. but in these notes we describe the size of a mesh by the number of interior points. 12.and three-dimensional ows.1 Mesh Indexing Convention 2 2 12.3.11) My k 1 13 23 33 43 12 22 32 42 11 21 31 41 1 (12. 6 point mesh if the boundary points (labeled in the sketch) were included. A clear description of a matrix nite-di erence operator in 2. In this example Mx .9 can be derived for a di erent point of view by writing Eq. hAd ]un = I + hAc]un + h(bc) +1 12.10. 12. the number of (interior) x points. SPLIT AND FACTORED FORMS un . The numbers 11 12 43 represent the location in the mesh of the dependent variable bearing that index. 3 This could also be called a 5 .12) j Mx Mesh indexing in 2-D.3 Factoring Space Matrix Operators in 2{D Factoring is widely used in codes designed for the numerical solution of equations governing unsteady two. Let us study the basic concept of factoring by inspecting its use on the linear 2-D scalar PDE that models di usion: We begin by reducing this PDE to a coupled set of ODE's by di erencing the space derivatives and inspecting the resulting matrix operator. ~ I .6 in the form Then which is identical to Eq. We choose the 3 4 point mesh shown in the Sketch 12. Thus u represents the value of u at j = 3 and k = 2.and 3-D requires some reference to a mesh. 12. 3 32 @u = @ u + @ u @t @x @y 2 2 (12.12.220 CHAPTER 12. the number of (interior) y points is 3.

( ) ( ) 12. consecutively along rows that are themselves consecutive from k = 1 to My . Examples are in Eq. Of these. To be speci c about the structure. There are many ways to lay out a data-base. In particular.3.16 part a and b.3. respectively. Examples of the order of indexing for these space vectors are given in Eq. and one composed of y-vectors with U y . are subsets of A and ~ . consecutively along columns that are consecutive from j = 1 to Mx. 12.15) Ax+y and U .16 part a and b.14) Axx y = Pxy Axy y Pyx ( ) + ( ) + (12. Thus ( ) + ( ) + ( ) ( ) dU = A U + (bc) ~ x y dt + (12. we use the notation Axx y or Axy y . In this notation the ODE form of Eq.3 Data Base Permutations U x = Pxy U y ( ) ( ) The two vectors (arrays) are related by a permutation matrix P such that and U y = PyxU x ( ) 1 ( ) (12. We refer to each row or column group as a space vector (they represent data along lines that are continuous in space) and label their sum with the symbol U .11 can be written + 4 If it is important to be speci c about the data-base structure. Notice that the matrices are not the same although they represent the same derivative operation. FACTORING SPACE MATRIX OPERATORS IN 2{D 221 12. Their structures are similar.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. u 4 Notice that . used in the previous sections. and they are related by the same permutation matrix that relates U x to U y . which are notations used in the special case of space vectors. 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. (1) and (2) above are referred to as xvectors and y-vectors. 12. depending on the data{base chosen for the U it multiplies. 12.12. the usual (but ambiguous) representation is given by Ax y . however. When the matrix is multiplying a space vector U . and (2). The symbol U by itself is not enough to identify the structure of the data-base and is used only when the structure is immaterial or understood. we label a data{base composed of x-vectors with U x .13) where Pyx = PT = P. we consider only two: (1).

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

30) I . it is guaranteed to converge to the correct steady-state solution of the ODE.14.29 converges. Un one nds h 1 1 ~ i I . Then. 2 hAy Un = I + 2 hAx + 1 hAy Un + h(bc) + O(h ) 2 From both sides subtract 1 I . ( ) ( ) 12. 2 hAx .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. is referred to as the \delta form" and we develop it next. 12. Consider the unfactored form of the trapezoidal method given by Eq. The temporary solution U x would then y and an inspection of the bottom of Eq.23 shows that the nal ~ be permuted to U step consists of solving Mx one-dimensional implicit problems each with dimension My .22. One way that is especially useful. 1 hAy Un = h Ax y Un + (bc) + O(h ) 2 This is the delta form of a factored. 1 hAy Un 2 leaving the equality unchanged. 2 hAx . 2nd-order.21 but is second-order time accurate. Un = Un .29 and maintain O(h ) accuracy.228 CHAPTER 12. 9 A block tridiagonal solver is similar to a scalar solver except that small block matrix operations .6 The Delta Form Clearly many ways can be devised to split the matrices and generate factored forms. Thus. 2 hAx . 12. The rst step would be solved using My uncoupled block tridiagonal solvers . 2 hAx I . Inspecting the top of Eq. using the standard de nition of the di erence operator . 12. we see that this is equivalent to solving My one-dimensional ~ problems. 2-D equation. +1 3 +1 + 3 2 + 3 replace the scalar ones. 2 hAy Un = h Ax y Un + (bc) + O(h ) (12. and matrix multiplications do not commute. 12. Now we can factor Eq. 12. and ~ let the (bc) be time invariant: 1 1 1 ~ I .24. for ensuring a correct steady-state solution in a converged time-march. 12. our original ODE. We arrive at the expression h 1 ~ i (12. 12. each with Mx blocks of order Ne . if Eq. SPLIT AND FACTORED FORMS 9 which has the same appearance as Eq.

hAy ] Un = h Ax y Un + (bc) + (12. (a) Space vector de nition i. Consider the 1-D heat equation: @u = @ u @t @x 2 2 0 x 9 Let u(0 t) = 0 and u(9 t) = 0. the unfactored form of a rst-order method derived from the implicit Euler time march is given by Eq. (1) (2) 12 21 x 2 j . i. 12. and if it is immediately factored. write the space vector.e. u ? Only include the interior points. hAx] I . 12. If we reorder the points with the odd points rst and then the even points. 2u + u ) xx j The uniform grid has x = 1 and 8 interior points. What is the space vector for the natural ordering (monotonically increasing in index). ii.. 12.20. u ? iii. On the other hand. 2 . 12.1 j j +1 10 Notice that the only di erence between the O(h2 ) method given by Eq. the delta form of the unfactored Eq. hAy ] Un = h Ax y Un + (bc) + 10 and its factored form becomes h ~ i I .19. but it can have a profound e ect on the stability and convergence properties of a method.30 and the O(h) method given by Eq.31) In spite of the similarities in derivation.31 are vastly di erent. hAx .12. Write down the permutation matrices. we will see in the next chapter that the convergence properties of Eq. 12. ( u) = 1 (u .20 and Eq. 12. so that we can simplify the boundary conditions.7. For example. PROBLEMS 229 The point at which the factoring is made may not a ect the order of time-accuracy.P ). 12.(P . the factored form is presented in Eq.7 Problems 1. Assume that second order central di erencing is used.31 is the appearance of the factor 1 on the left side of the O(h2 ) method.19 is h ~ i I . 12.

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

(2) 2 . i. You should be able to argue that these are now trangular matrices (a lower and an upper). Comment on their form.7.12. de ne a splitting A = Ao + Ae. PROBLEMS (2) ( ) ( ) (2) ( ) ( ) (2) 231 which extract the odd even points from u as follows: u o = I o u and ue =Ieu . and Ae operates only on the even terms. iii. write them in terms of D and U de ned above. Apply implicit Euler time marching to the split ODE. Comment on the solution process this gives us relative to the direct inversion of the original system. Also. Write out the matrices Ao and Ae. Beginning with the ODE written in terms of u . ii. Comment on the order of the error terms. Write down the delta form of the algorithm and the factored delta form. such that Ao operates only on the odd terms. Examine the implicit operators for the factored delta form.

232 CHAPTER 12. SPLIT AND FACTORED FORMS .

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

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

can be represented by the eigenvalues c and d. the explicit-implicit Euler method. a B (. we can analyze the stability of the two forms of simple time-splitting discussed in Section 12. the characteristic polynomial of this method is P (E ) = (1 . where (see Section 4. the explicit-explicit Euler method.7) .3.2 1)~ u (13. 4 2 sin2 2 x In these equations m = 2m =M .2.13. m = 0 1 M . 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. respectively. 12.10. 13. 12. Eq. 1 . EXAMPLE ANALYSIS OF CIRCULANT SYSTEMS 235 If the space di erencing takes the form d~ = .2.1 0 1)~ + u u B (1 . The Explicit-Implicit Method 2 (13.2): ( c)m = ia sin m x m (13. h d )E . so that 0 m 2 .6 to quantify the eigenvalues. 13.5) dt 2 x p x2 p the convection matrix operator and the di usion matrix operator. In this section we refer to these as 1. Eq. (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. 2. When applied to Eq. Using these values and the representative equation 13.4.4.8.6) ( d)m = .

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. and the . A simple numerical parametric study of Eq. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 1.8 C n C 0.8 n 1. 4 Cn sin2 m 2 j j = 1 + Cn sin 0 m 2 R 2 Again a simple numerical parametric study shows that this has two critical ranges of m .236 CHAPTER 13. one near 0.4 0.1.2 C = R /2 n ∆ C = 2/ R n ∆ 0.4 0 0 4 R∆ Explicit-Implicit 8 0 0 4 R∆ Explicit-Explicit 8 _ Figure 13. 13. The Explicit-Explicit Method An analysis similar to the one given above shows that this method produces " # q 2 m 1 . 2. 13.1: Stability regions for two simple time-split methods. which yields the same result as in the previous example.7 shows that the critical range of m for any combination of Cn and R occurs when m is near 0 (or 2 ).2 C = 2/ R ∆ n 0.

1 h d)E 2 . Our model equation is again the linear convection-di usion equation 13. P (E ) and Q(E ). which produces the constraint that 1 Cn < 2 R for R 2 The resulting stability boundary is also shown in Fig. The di usion term is integrated implicitly using the trapezoidal method. The totaly explicit. un. (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.5.2 Analysis of a Second-Order Time-Split Method .13.1. 13. thus 1 1 un+1 = un + 2 h c(3un . In order to evaluate the accuracy. In the former case it is 1 Q(E ) = 2 h(3E .2. as we should have expected. This method applies to a ow in which there is a combination of di usion and periodic convection. EXAMPLE ANALYSIS OF CIRCULANT SYSTEMS 237 other near 180o.8) and in the latter (13. we include the forcing function in the representative equation and study the e ect of our hybrid. time-marching method on the equation u0 = cu + du + ae t First let us nd expressions for the two polynomials.2. The characteristic polynomial follows from the application of the method to the homogeneous equation. factored method has a much smaller region of stability when R is small. 1) (13. as well as the stability.1) + 2 h d (un+1 + un) This produces 3 P (E ) = (1 .9) Q(E ) = 1 h(E 2 + E ) 2 13. 13. The convection term is treated explicitly using the second-order Adams-Bashforth method.4 which we split in the fashion of Eq. Next let us analyze a more practical method that has been used in serious computational analysis of turbulent ows.

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 . the hybrid method retains the second-order accuracy of its individual components. 13.7. 13. For values of R 2 this second-order method has a much greater region of stability than the rst-order explicit-implicit method given by Eq. 2 h d (13. 13.11) The extension of the following to 3-D is simple and straightforward.238 CHAPTER 13. 1 1 + 3h c + 2h 2 2 1 . . so er = O(h3) Using P (e h) and Q(e h ) to evaluate er in Section 6. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS Accuracy From the characteristic polynomial we see that there are two {roots and they are given by the equation 3 1 + 2h c + 1h 2 = s d The principal -root follows from the plus sign and one can show 1 1 1 = 1 + ( c + d )h + 2 ( c + d )2 h2 + 4 3 + c 2 . 13. for the model equation.2.10 and shown in Fig. 13.8 or Eq.10 by a parametric study of cn and R de ned in Eq.9. 1h d 2 d 2 1 .1.6. The results are plotted in Fig. 13. This was carried out in a manner similar to that used to nd the stability boundary of the rst-order explicit-implicit method in Section 13. 13.3 The Representative Equation for Space-Split Operators Consider the 2-D model3 equations @u = @ 2 u + @ 2 u @t @x2 @y2 3 (13.10) Stability The stability of the method can be found from Eq.1. 2 d . 2h c 1 . one can show er = O(h3) using either Eq.3. These results show that. 12.2.

3.12) @t x @x y @y Reduce either of these.8 Cn 0. The form of the Ax and Ay matrices for three-point central di erencing schemes are shown in Eqs.2: Stability regions for the second-order time-split method.22 and 12.1 I ~0 I ~1 I 7 b b 5 4 5 4b ~. 12.2 0.4 Stable 0 4 8 ∆ R _ Figure 13.23 for the 3 4 mesh shown in Sketch 12.1 I ~0 I B b b where B is a banded matrix of the form B (b.13. and @u + a @u + a @u = 0 (13. 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. to the coupled set of ODE's: dU = A + A ]U + (bc) ~ (13. 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 ~. THE REPRESENTATIVE EQUATION FOR SPACE-SPLIT OPERATORS239 1.1 b0 b1 ).12.13) x y dt for the space vector U . Now nd the block eigenvector . by means of spatial di erencing approximations.

13. so the nal result is the complete diagonalization of the matrix Ax+y : h ih ih i ~ ~ X .1 A(xx+)y X Pxy X = 2 3 1I + ~ 6 7 6 7 2I + ~ 6 7 (13.14 is transparent to the transformation.1B X = 6 7 4 5 4 5 ~3 ~ This time.14) ~ 4 5 4 B 7 3 I 5 ~ B 4 I ~ where B is the banded tridiagonal matrix B (~.15) 6 7 4 5 3I + ~ 4I + ~ diag(X ) 2 ~ b0 X .1 X B X where 2 3 6 =6 6 4 3 7 5 1 2 set X 4 (x) is transparent to this transformation.1 Pyx X .1 A(xx) + A(yx) X Pxy = 3 2 3 2 ~ B 1 I 7 6 7 6 ~ 7 B 6 7+6 2 I 6 7 6 7 6 (13. Let B diag(B ) and ~ ~ ~ Next nd the eigenvectors X ~ diag(X ) and form the second transformation ~ X 2~ 3 2~ 3 1 6 7 7 ~2 ~ ~~ 6 ~ ~ 7 ~=6 X .1 6 ~. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS matrix that diagonalizes B and use it to diagonalize A(xx).1 4b I I . 12. see the bottom of Eq.1 ~0 ~1). That is. the rst matrix on the right side of Eq.1 32 32 3 2 X B X 6 76 7=6 X . b b b ~ that diagonalize the B blocks. by the same argument as before.240 CHAPTER 13.23. Thus 2 .1 I ~0 I ~. 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.13.1 B 76 X 4 54 54 5 4 . There results h i Pyx X .1 I ~0 I ~1 I 7 b b 5 b b 5 4b ~. 12.

and where Ax and Ay satisfy the conditions in 13. This results in a spatially split A matrix formed from the subsets A(xx) = diag(B ) ~ A(yy) = diag(B ) (13. 13. the steady-state solution of the representative equation.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). + (13. In that case we set = 0 and use the simpler form du = + ]u + a (13. by itself. and its use is not necessary to arrive at Eq. does not ensure the property of \all possible combinations". First reduce the PDE to ODE by some choice4 of space di erencing. To obtain the latter property the structure of ~ the matrices is important. Now we are ready to present the representative equation for two dimensional systems.17) x y dt which has the exact solution a u(t) = ce( x+ y )t . Although Ax and Ay do commute. 13.15 contains every possible com~ bination of the individual eigenvalues of B and B . THE REPRESENTATIVE EQUATION FOR SPACE-SPLIT OPERATORS241 It is important to notice that: The diagonal matrix on the right side of Eq. Often we are interested in nding the value of.3.15. and the convergence rate to. but this choice was for convenience only.16. this fact.18) x y We have used 3-point central di erencing in our example.13. 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 . a and are (possibly complex) constants. 4 .

however. Unfortunately. " 1 #n a x+ y 2.1 The Unfactored Implicit Euler Method Consider rst this unfactored.16. In the following we analyze four di erent methods for nding a xed. h . 12. In each case we examine 13. we nd (1 . LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 13. rst-order scheme which can then be used as a reference case for comparison with the various factored ones. and when it is applied to the representative equation. 3. 13. Produces the exact (see Eq.5. 1 Q(E ) = h giving the solution (13. h from which x . Converges very rapidly to the steady-state when h is large. 3. use of this method for 2-D problems is generally impractical for reasons discussed in Section 12. The convergence rate to reach the steady-state. steady-state solution. h y )E .4. steady-state solution to the 2-D representative equation 13. h x .18) steady-state solution (of the ODE) for any h. x y Like its counterpart in the 1-D case. Is unconditionally stable. The stability. h y )un+1 = un + ha P (E ) = (1 . The form of the method is given by Eq. 2. this method: 1.4 Example Analysis of 2-D Model Equations 1. h . .242 CHAPTER 13.19) un = c 1 .19. The accuracy of the xed.

h ) . 12. Is unconditionally stable. 12.20) giving the solution " #n 1 un = c (1 . Next apply Eq.31 to the 2-D representative equation. if one tries to take advantage of its rapid convergence rate. The method requires far less storage then the unfactored form. Produces a steady state solution that depends on the choice of h. h y )un+1 = un + ha from which P (E ) = (1 . h y )un+1 = 1 + h2 x y un + ha and this has the solution " #n 1 + h2 x y a un = c (1 . h y )E . Converges rapidly to a steady-state for large h. + x y x y This method: 13. One nds (1 . h x)(1 .3 The Factored Delta Form of the Implicit Euler Method .13. h x)(1 .4.2 The Factored Nondelta Form of the Implicit Euler Method 13. EXAMPLE ANALYSIS OF 2-D MODEL EQUATIONS 243 Now apply the factored Euler method given by Eq. x y We see that this method: 1. h x)(1 . it is not very useful since its transient solution is only rst-order accurate and. h x)(1 . un) = h( xun + y un + a) which reduces to (1 . 1 Q(E ) = h (13.20 to the 2-D representative equation. There results (1 . 2. but the converged solution is completely wrong. + a h x y x y. h ) .4. 3. h y )(un+1 . However. h )(1 . the converged value is meaningless.4. h )(1 .

Apply Eq.5 A brief inspection of eqs. Converges very slowly to the steady{state solution for large values of h. since j j ! 1 as h ! 1. un = c6 4 5 1 x+ y 1 . 1 h y (un+1 . In practical codes.26 and 12. the value of h on the left side of the implicit equation is literally switched from h to 1 h. Like the factored nondelta form. it can be used when time accuracy is desired. Is unconditionally stable. 12. as discussed in Section 12. 3. Produces the exact steady-state solution for any choice of h. All of these properties are identical to those found for the factored delta form of the implicit Euler method. but convergence is not nearly as rapid as that of the unfactored form. 12. LINEAR ANALYSIS OF SPLIT AND FACTORED FORMS 1. 2 h x 1 .4. 3. 2. The correct steady solution is obtained. 2h y 2 This method: 1. 2.27 should be enough to convince the reader that the 's produced by those methods are identical to the produced by this method. since j j ! 1 as h ! 1. and the factored delta form of the implicit Euler method can be used when a converged steady-state is all that is required. un) = h( xun + y un + a) 2 which reduces to 1 .4 The Factored Delta Form of the Trapezoidal Method .5.30 to the representative equation and one nds 1 1 . 1 h y un+1 = 1 + 1 h x 1 + 1 h y un + ha 2 2 2 2 and this has the solution 2 3n 1 + 1h x 1 + 1h y 7 6 a 2 2 7 . Is unconditionally stable. 1 h x 1 . 2 5 13. 1h x 1 . Since it is second order in time. this method demands far less storage than the unfactored form. Finally consider the delta form of a second-order time-accurate method. Produces the exact steady{state solution for any choice of h. Converges very slowly to the steady{state solution for large values of h.244 CHAPTER 13.

+a+ x y 6 z (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 .23) Eqs.12. nd the root. 13.21) x y z dt Let us analyze a 2nd-order accurate. under the conditions given in 13. 2 h x 1 . 1 h( x + y + z ) un = h ( x + y + z )un + a] 2 Now factor the left side: 1 1 .22. 8 x y z . factored.5 Example Analysis of the 3-D Model Equation 245 The arguments in Section 13. One can derive the characteristic polynomial for Eq.13. and write the solution either in the form 2 3n 1 h( + + ) + 1 h2 ( 1 h3 + + ). 1 h y 1 . 2 x y z 4 x y + x z + y z) .5. the model 3-D cases6 have the following representative equation (with = 0): du = + + ]u + a (13.16 with an A(zz) included.11 and 13.3 generalize to three dimensions and. 2 h( x + y + z ) un+1 = 1 + 2 h( x + y + z ) un + ha Put this in delta form: 1 . 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 .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). 13. 1 h z un = h ( x + y + z )un + a] (13. EXAMPLE ANALYSIS OF THE 3-D MODEL EQUATION 13. each with an additional term. delta form using this equation. .

and are both positive. 13. 13. First write the root in Eq.23 in the form 1+ = 1. 13.23 that. the 3-D form of Eq.24 results in 1 = 1 + h( x + y + z ) + 2 h2( x + y + z )2 (13. 2 x 1 .22 should not be used for the 3-D Euler equations.12 with periodic boundary conditions. we already knew this because we chose the delta form. 7 However. 13. it follows from Eq. 13.246 CHAPTER 13. some form of dissipation is almost always added to methods that are used to solve the Euler equations and our experience to date is that. +i where . 4 h3 2 2 1h 1h 1 . In this case. 2 y 1 . In practical cases. however. 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 . and are real numbers that can have any sign. Now consider what happens when we apply this method to the biconvection model. .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. central di erencing causes all of the 's to be imaginary with spectrums that include both positive and negative values.i . if all the 's are real and negative. )2 >1 and the method is unconditionally unstable for the model convection problem.24) It is interesting to notice that a Taylor series expansion of Eq. Now we can always nd one combination of the 's for which . +i i . 5 x+ y+ z a (13. it converges to the proper steady-state. Remember that in our analysis we must consider every possible combination of these eigenvalues.7 With regards to stability. From the above analysis one would come to the conclusion that the method represented by Eq. Furthermore. the method is stable for all h. clearly. This makes the method unconditionally stable for the 3-D di usion model when it is centrally di erenced in space. ) + ( + ) j (1 . if the method converges. In that case since the absolute value of the product is the product of the absolute values j 2 2 2 = (1 . 1h z 2 3n x y z7 7 . in the presence of this dissipation. )2 + ( . the instability disclosed above is too weak to cause trouble.

PROBLEMS 13. un = A u + A u + A u + A u + f 1 n+1 2 n 3 n+1 4 n h Write the resulting factored delta form and de ne the error terms. (d) Repeat 1c for the explicit-implicit scheme of problem 1b. convergence.6 Problems 247 1. leave two explicit: un+1 . du = Au + f dt we can split A as follows: A = A1 + A2 + A3 + A4. Starting with the generic ODE. Applying implicit Euler time marching gives un+1 . (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.6. What is the error term? (b) Instead of making all of the split terms implicit.13. 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.

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

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

where I is the identity matrix.c .1 = I . P is a permutation matrix if P ~ is a simple reordering of ~ .1 and has the property that A. A. A is orthogonal if aij are real and AT A = AAT = I 5.. A is the complex conjugate of A. bc and d A. 11. The trace of a matrix is Pi aii. The inverse of a matrix (if it exists) is written A. v v 7. A is symmetric if AT = A.250APPENDIX A. 4. 6. 7 4 T5 ~m a 4. A is skew-symmetric or antisymmetric if AT = .1 = det1 A] . 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. 2. 3. (Hermitian). If b A= a d c then det A] = ad . 9.1A = AA. A is normal if AT A = AAT . USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA 3.A. AH is the conjugate transpose of A.b a . det A] is the determinant of A.2 De nitions 1. An alternative notation for A is h i A = ~1 ~2 : : : ~m a a a and its transpose AT is 2 ~T 3 a 6 ~1 7 6 aT 7 AT = 6 .2 7 6 7 6 . 10.

they are all imaginary. If it is asymmetric.4 Eigensystems 1. I ]~ = 0 x x x and the generalized eigenvalue problem. 5.1 (AT ).3 Algebra We consider only square matrices of the same dimension. as A~ = B~ or A . sA = (saij ) where s is a scalar. its eigenvalues are all real. Transpose equalities: (A + B )T = AT + B T (AT )T = A (AB )T = B T AT 6. Inverse equalities (if the inverse exists): (A. 4.1 A. Any matrix A can be expressed as the sum of a symmetric and a skew-symmetric matrix.1). ALGEBRA A. A + (B + C ) = (C + A) + B .1 = (A.1 )T 251 7. 1. The eigenvalue problem for a matrix A is de ned as A~ = ~ or A .3. B ]~ = 0 x x x 2. 2.A. In general AB 6= BA.1 = A (AB ). A and B are equal if aij = bij for all i j = 1 2 : : : m. If a square matrix with real elements is symmetric. . including the matrix B . Thus: 1 A = 2 A + AT + 1 A . AT 2 A. etc. 3.1 = B .

If the eigenvalues of a matrix are not distinct. USEFUL RELATIONS AND DEFINITIONS FROM LINEAR ALGEBRA 3.. In general. an m m matrix A has n~ linearly independent eigenvectors with x n~ m and n distinct eigenvalues ( i) with n n~ m.e. 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. . and with each distinct eigenvalue there is one associated eigenvector.. and A is said to have a complete eigensystem. 4. 7 6 . If A posseses m linearly independent eigenvectors then A is diagonalizable. 8. 6. . . .2 . h i X = ~1 ~2 : : : ~m x x x and is the diagonal matrix 0 0 m If A can be diagonalized. i. the eigenvalues of a matrix may not be distinct. . the matrix is said to be derogatory.. X . x x 5. 7.1AX = where X is a matrix whose columns are the eigenvectors. 9. 7 6 7 = 6 0 . but it can still be diagonalized. 2 0 03 1 6 . . but all of the eigenvectors are linearly independent. .. 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. In general. its eigenvectors completely span the space. then A is always diagonalizable. If a matrix does not have a complete set of linearly independent eigenvectors. . and this eigenvector cannot be formed from a linear combination of any of the other eigenvectors. The eigenvectors of such a matrix cannot span the space and the matrix is said to have a defective eigensystem. 0 7 4 5 . it cannot be diagonalized. in which case the possibility exists that it cannot be diagonalized.

. 0 7 4 5 0 0 Jk 253 10. 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 .1 AS = 6 0 J2 . there exists one eigenvector.. For example.. 1 7 4 5 0 0 0 i 12.7 6 . . . 7 6... . The complete set of principal vectors and eigenvectors are all linearly independent.7 J =S 6 . . 13. the . . 7 .1 = P then 2 3 2 3 1 0 0 0 P . EIGENSYSTEMS 2J 0 03 1 ..4..A. such that where there are k linearly independent eigenvectors and Ji is either a Jordan subblock or i. . . For each Jordan submatrix with an eigenvalue i of multiplicity r.. 0 7 Ji = 6 7 i 6. . 7 6 7 i 1 6 60 0 7 . ... Some of the Jordan subblocks may have the same eigenvalue. 6 . S . A repeated root in a Jordan block is referred to as a defective eigenvalue..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 . . 11. Use of the transform S is known as putting A into its Jordan Canonical form. The other r . . 1 vectors associated with this eigenvalue are referred to as principal vectors. Defective matrices cannot be diagonalized but they can still be put into a compact form by a similarity transform. .

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

A. in this case it is the L2 norm. 8. VECTOR AND MATRIX NORMS 255 4. When A is normal. Let A and B be square matrices of the same order. (A). so in general (A) is not a true norm. (A) is a true norm. 7. In general (A) does not satisfy the conditions in 4. The spectral radius of A. Special p-norms are jjAjj1 = maxj=1 M jjAjj1 = maxi=1 2 M PM jaij j maximum row sum j =1 where jjAjjp is referred to as the Lp norm of A. jjAjj2 = q PM ja j maximum column sum i=1 ij (A T A ) . All matrix norms must have the properties jjAjj 0 jjAjj = 0 implies A = 0 jjc Ajj = jcj jjAjj jjA + B jj jjAjj + jjB jj jjA B jj jjAjj jjB jj 5. is the lower bound of all the norms of A. 6.5. in fact.

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

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. i.2) and is given by j. Many of these can be derived by solving the simple linear di erence equations that arise in deriving recursion relations.b..4.1 ~ m = (xj )m = a 2 sin j m x m=1 2 M (B.3) c M +1 257 .1 Standard Eigensystem for Simple Tridiagonals In this work tridiagonal banded matrices are prevalent. It is useful to list some of their properties. see Section 3.e.Appendix B SOME PROPERTIES OF TRIDIAGONAL MATRICES B. a tridiagonal with constant scalar elements a.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. Let us consider a simple tridiagonal matrix. and c. If we examine the conditions under which the determinant of this matrix is zero.

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

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

cD .a3 D1 4 cD . DM .n(.c D .1 Standard Tridiagonals Consider the circulant (see Section 3.4) tridiagonal matrix Bp(M : a b c ) (B. = 1 D .mDn.cD D c D D .a D a D .aD D D .4.aD D D D .m =DM Diagonal: n=m=1 2 M dmm = DM .c)n.1 1 n=m+1 m+2 M dmn = Dm.1 dmn = DM .aD D D D .260 APPENDIX B.cD 3 7 17 7 2 7 7 5 3 Upper triangle: m=1 2 M .cD D c D a D D .a)m. SOME PROPERTIES OF TRIDIAGONAL MATRICES 2 16 6 4 D4 6 Then for M = 4 B.cD D c D a D .1 = aD The general element dmn is 4 0 2 D4 6 .n=DM B.aD D 3 2 2 1 3 2 2 1 2 1 1 2 1 3 1 0 2 1 1 2 1 1 2 3 0 2 3 7 7 7 5 and for M = 5 B . (.c D c D D D .4.a D aD .cD cD .c D .4 Eigensystems of Circulant Matrices B.cD . DM .m =DM 1 Lower triangle: m =n+1 n+2 M 1 n=1 2 M .12) .aD3 1 6 a2 D 6 2 6 D5 6 .aD D 3 2 2 3 1 3 1 2 2 1 4 1 1 3 0 2 1 2 2 2 2 1 2 1 3 1 1 2 2 1 3 1 2 3 4 .aD D D D .

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 .B. B.15) 4 b b b b 5 b b b b The eigenvectors are always given by eq. even if they are completely dense. B.4. all circulant matrices of order M have the same set of linearly independent eigenvectors. = (x0 ) = 1 e. are symmetric and that X .14.1 (B. the eigenvalues are not.1 The left-hand eigenvector matrix with elements x0 is 2 j m =0 1 M .12 do not depend on the elements a b c in the matrix. where X is the conjugate transpose of X .15 they have the general form M.1 X = (xjm) = eij M m =0 1 M . j =0 . i(a . EIGENSYSTEMS OF CIRCULANT MATRICES The eigenvalues are m 261 = b + (a + c) cos 2 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.13 is a special case.4.13) B. 1 1 1 M .im M X mj j =0 1 M .1 M 1 Note that both X and X .1. In fact. = M X . and the right-hand eigenvector matrix has the form 2 m j =0 1 M .1 (B.1 . An example of a dense circulant matrix of order M = 4 is 2 b b b b 3 6 b b b b 7 6 7 6 7 (B.2 General Circulant Systems Notice the remarkable fact that the elements of the eigenvector matrices X and X . for the tridiagonal circulant matrix given by eq. B. B. For the element indexing shown in eq.14) p where i . and further examination shows that the elements in these eigenvectors correspond to the elements in a complex harmonic analysis or complex discrete Fourier series. Although the eigenvectors of a circulant matrix are independent of its elements.

B.5 Special Cases Found From Symmetries APPENDIX B. 7 6 .. For example. 7 76 .17) . 7 76 . 7~ a ~ a)~ m = 7 xm = B (M : a b x 7 .. 7 76 . .. 7 m6 . 1) m = b + 2a cos 2M + 1 ! m=1 2 M (B.. 7 6 .. 7 54 5 a 7 6 x2 7 = x 2 3 x1 6x 7 6 27 6 . 7 6 . SOME PROPERTIES OF TRIDIAGONAL MATRICES Consider a mesh with an even number of interior points such as that shown in Fig. 7 6 . 7 6 . One can seek from the tridiagonal matrix B (2M : a b a ) the eigenvector subset that has even symmetry when spanning the interval 0 x . 7 76 .262 B. a 7 7 a b a 7 5 a b+a 2 6 6 6 6 6 6 6 6 6 6 4 3 x m~ m (B.16 are (2m .. B. one can show from previous results that the eigenvalues of eq.1. 7 6 7 4 x2 5 1 x 1 This leads to the subsystem of order M which has the form b a 7 7 a b a 7 7 ...16) By folding the known eigenvectors of B (2M : a b a) about the center. a a b a b 32x 3 1 7 6 x2 7 76 7 76 . 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 .

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

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

4) . let us rst inspect Euler's theorem on homogeneous functions. 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. F is said to be homogeneous of degree n and we nd " (C.3) @F Q = nF (Q) @q 265 # (C. Di erentiating both sides with respect to and setting = 1 (since the identity holds for all ). we nd Consider next the theorem as it applies to systems of equations.1) for a xed n.2) F ( Q) = nF (Q) for a xed n. If F (u v) satis es the identity F ( u v) = nF (u v) (C. If the vector F (Q) satis es the identity u @F + v @F = nF (u v) @u @v (C. F is called homogeneous of degree n. In order to examine this property.

we notice that the expansion of the ux vector in the vicinity of tn which. see eq.5) E = An Q + O(h2) F = BnQ + O(h2) (C. Qn) + O(h2) F = Fn + Bn(Q . " # " # (C. As a nal remark. by direct use of eq.12 are homogeneous of degree 1. BnQn are identically zero for homogeneous vectors of degree 1. Qn) + O(h2) can be written (C.3. are homogeneous of degree 0 (actually the latter is a direct consequence of the former). Notice also that. C. THE HOMOGENEOUS PROPERTY OF THE EULER EQUATIONS Now it is easy to show. under this condition. the constant term drops out of eq.1 This being the case. C.8) @A Q = 0 @x in spite of the fact that individually @A=@x] and Q are not equal to zero.7) (C.11 and 2. 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. AnQn and Fn . if F is homogeneous of degree 1. .4.106. where is the internal energy per unit mass. 1 " # (C. F = AQ and @F = A @Q + @A Q @x @x @x Therefore. according to eq. and their Jacobians.266APPENDIX C.105 can be written in general as.9) Note that this depends on the form of the equation of state. 2. 6.6) since the terms En . that both E and F in eqs. A and B . E = En + An(Q . The Euler equations are homogeneous if the equation of state can be written in the form p = f ( ). 6.

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

Sign up to vote on this title
UsefulNot useful