# Numerical Analysis

**Using MATLAB® and Excel®
**

Third Edition

Steven T. Karris

Orchard Publications www.orchardpublications.com

Numerical Analysis

Using MATLAB® and Excel® Third Edition

Students and working professionals will find Numerical Analysis Using MATLAB® and Excel®, Third Edition, to be a concise and easy-to-learn text. It provides complete, clear, and detailed explanations of the principal numerical analysis methods and well known functions used in science and engineering. These are illustrated with many real-world examples.

This text includes the following chapters and appendices: • Introduction to MATLAB • Root Approximations • Sinusoids and Complex Numbers • Matrices and Determinants • Review of Differential Equations • Fourier, Taylor, and Maclaurin Series • Finite Differences and Interpolation • Linear and Parabolic Regression • Solution of Differential Equations by Numerical Methods • Integration by Numerical Methods • Difference Equations • Partial Fraction Expansion • The Gamma and Beta Functions • Orthogonal Functions and Matrix Factorizations • Bessel, Legendre, and Chebyshev Polynomials • Optimization Methods • Difference Equations in Discrete-Time Systems • Introduction to Simulink • Ill-Conditioned Matrices Each chapter contains numerous practical applications supplemented with detailed instructions for using MATLAB and/or Excel to obtain quick solutions. Steven T. Karris is the president and founder of Orchard Publications, has undergraduate and graduate degrees in electrical engineering, and is a registered professional engineer in California and Florida. He has more than 35 years of professional engineering experience and more than 30 years of teaching experience as an adjunct professor, most recently at UC Berkeley, California.

Orchard Publications Visit us on the Internet www.orchardpublications.com or email us: info@orchardpublications.com

19 0 1 ISBN-13: 978-1-934404-04-1 9 0 7 ISBN-10: 1-934404-04-7

$60.00 USA

Numerical Analysis

Using MATLAB® and Excel®

Third Edition

Steven T. Karris

Orchard Publications www.orchardpublications.com

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright ” 2007 Orchard Publications. All rights reserved. Printed in the United States of America. No part of this publication may be reproduced or distributed in any form or by any means, or stored in a data base or retrieval system, without the prior written permission of the publisher. Direct all inquiries to Orchard Publications, info@orchardpublications.com Product and corporate names are trademarks or registered trademarks of the Microsoft™ Corporation and The MathWorks™, Inc. They are used only for identification and explanation, without intent to infringe.

**Library of Congress Cataloging-in-Publication Data
**

Library of Congress Control Number: 2007922100 Copyright TX 5-589-152

ISBN-13: 978-1-934404-04-1 ISBN-10: 1-934404-04-7

Disclaimer

The author has made every effort to make this text as complete and accurate as possible, but no warranty is implied. The author and publisher shall have neither liability nor responsibility to any person or entity with respect to any loss or damages arising from the information contained in this text.

Preface

Numerical analysis is the branch of mathematics that is used to find approximations to difficult problems such as finding the roots of non−linear equations, integration involving complex expressions and solving differential equations for which analytical solutions do not exist. It is applied to a wide variety of disciplines such as business, all fields of engineering, computer science, education, geology, meteorology, and others. Years ago, high−speed computers did not exist, and if they did, the largest corporations could only afford them; consequently, the manual computation required lots of time and hard work. But now that computers have become indispensable for research work in science, engineering and other fields, numerical analysis has become a much easier and more pleasant task. This book is written primarily for students/readers who have a good background of high−school algebra, geometry, trigonometry, and the fundamentals of differential and integral calculus.* A prior knowledge of differential equations is desirable but not necessary; this topic is reviewed in Chapter 5. One can use Fortran, Pascal, C, or Visual Basic or even a spreadsheet to solve a difficult problem. It is the opinion of this author that the best applications programs for solving engineering problems are 1) MATLAB which is capable of performing advanced mathematical and engineering computations, and 2) the Microsoft Excel spreadsheet since the versatility offered by spreadsheets have revolutionized the personal computer industry. We will assume that the reader has no prior knowledge of MATLAB and limited familiarity with Excel. We intend to teach the student/reader how to use MATLAB via practical examples and for detailed explanations he/she will be referred to an Excel reference book or the MATLAB User’s Guide. The MATLAB commands, functions, and statements used in this text can be executed with either MATLAB Student Version 12 or later. Our discussions are based on a PC with Windows XP platforms but if you have another platform such as Macintosh, please refer to the appropriate sections of the MATLAB’s User Guide that also contains instructions for installation. MATLAB is an acronym for MATrix LABoratory and it is a very large computer application which is divided to several special application fields referred to as toolboxes. In this book we will be using the toolboxes furnished with the Student Edition of MATLAB. As of this writing, the latest release is MATLAB Student Version Release 14 and includes SIMULINK which is a

* These topics are discussed in Mathematics for Business, Science, and Technology, Third Edition, ISBN 0−9709511− 0−8. This text includes probability and other advanced topics which are supplemented by many practical applications using Microsoft Excel and MATLAB.

software package used for modeling, simulating, and analyzing dynamic systems. SIMULINK is not discussed in this text; the interested reader may refer to Introduction to Simulink with Engineering Applications, ISBN 0−9744239−7−1. Additional information including purchasing the software may be obtained from The MathWorks, Inc., 3 Apple Hill Drive, Natick, MA 01760−2098. Phone: 508 647−7000, Fax: 508 647−7001, e−mail: info@mathwork.com and web site http://www.mathworks.com. The author makes no claim to originality of content or of treatment, but has taken care to present definitions, statements of physical laws, theorems, and problems. Chapter 1 is an introduction to MATLAB. The discussion is based on MATLAB Student Version 5 and it is also applicable to Version 6. Chapter 2 discusses root approximations by numerical methods. Chapter 3 is a review of sinusoids and complex numbers. Chapter 4 is an introduction to matrices and methods of solving simultaneous algebraic equations using Excel and MATLAB. Chapter 5 is an abbreviated, yet practical introduction to differential equations, state variables, state equations, eigenvalues and eigenvectors. Chapter 6 discusses the Taylor and Maclaurin series. Chapter 7 begins with finite differences and interpolation methods. It concludes with applications using MATLAB. Chapter 8 is an introduction to linear and parabolic regression. Chapters 9 and 10 discuss numerical methods for differentiation and integration respectively. Chapter 11 is a brief introduction to difference equations with a few practical applications. Chapters 12 is devoted to partial fraction expansion. Chapters 13, 14, and 15 discuss certain interesting functions that find wide application in science, engineering, and probability. This text concludes with Chapter 16 which discusses three popular optimization methods. New to the Third Edition This is an extensive revision of the first edition. The most notable changes are the inclusion of Fourier series, orthogonal functions and factorization methods, and the solutions to all end−of− chapter exercises. It is in response to many readers who expressed a desire to obtain the solutions in order to check their solutions to those of the author and thereby enhancing their knowledge. Another reason is that this text is written also for self−study by practicing engineers who need a review before taking more advanced courses such as digital image processing. The author has prepared more exercises and they are available with their solutions to those instructors who adopt this text for their class. Another change is the addition of a rather comprehensive summary at the end of each chapter. Hopefully, this will be a valuable aid to instructors for preparation of view foils for presenting the material to their class. The last major change is the improvement of the plots generated by the latest revisions of the MATLAB® Student Version, Release 14. Orchard Publications Fremont, California www.orchardpublications.com info@orchardpublications.com

**Table of Contents 1 Introduction to MATLAB
**

1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 1.10 1.11 1.12 1.13

1−1

Command Window.................................................................................................1−1 Roots of Polynomials...............................................................................................1−3 Polynomial Construction from Known Roots ........................................................1−4 Evaluation of a Polynomial at Specified Values .....................................................1−5 Rational Polynomials ..............................................................................................1−8 Using MATLAB to Make Plots..............................................................................1−9 Subplots.................................................................................................................1−18 Multiplication, Division and Exponentiation.......................................................1−19 Script and Function Files......................................................................................1−26 Display Formats ....................................................................................................1−31 Summary ...............................................................................................................1−33 Exercises................................................................................................................1−37 Solutions to End−of−Chapter Exercises ...............................................................1−38

MATLAB Computations: Entire chapter

2

Root Approximations 2.1 2.2 2.3 2.4 2.5 2.6

2−1

Newton’s Method for Root Approximation...........................................................2−1 Approximations with Spreadsheets........................................................................2−7 The Bisection Method for Root Approximation .................................................2−19 Summary...............................................................................................................2−27 Exercises ...............................................................................................................2−28 Solutions to End−of−Chapter Exercises ...............................................................2−29

MATLAB Computations: Pages 2−2 through 2−7, 2−14, 2−21 through 2−23, 2−29 through 2−34 Excel Computations: Pages 2−8 through 2−19, 2−24 through 2−26

3

Sinusoids and Phasors 3.1 3.2 3.3 3.4 3.5 3.6 3.7

3−1

Alternating Voltages and Currents ........................................................................ 3−1 Characteristics of Sinusoids.................................................................................... 3−2 Inverse Trigonometric Functions ......................................................................... 3−10 Phasors.................................................................................................................. 3−10 Addition and Subtraction of Phasors ................................................................... 3−11 Multiplication of Phasors...................................................................................... 3−12 Division of Phasors ............................................................................................... 3−13

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

i

3.8 3.9 3.10 3.11

Exponential and Polar Forms of Phasors ..............................................................3−13 Summary ...............................................................................................................3−24 Exercises................................................................................................................3−27 Solutions to End−of−Chapter Exercises................................................................3−28

MATLAB Computations: Pages 3−15 through 3−23, 3−28 through 3−31 Simulink Modeling: Pages 3−16 through 3−23

4

Matrices and Determinants 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 4.10 4.11 4.12 4.13 4.14

4−1

Matrix Definition.....................................................................................................4−1 Matrix Operations ...................................................................................................4−2 Special Forms of Matrices........................................................................................4−5 Determinants ...........................................................................................................4−9 Minors and Cofactors ............................................................................................4−13 Cramer’s Rule ........................................................................................................4−18 Gaussian Elimination Method...............................................................................4−20 The Adjoint of a Matrix ........................................................................................4−22 Singular and Non−Singular Matrices ....................................................................4−22 The Inverse of a Matrix.........................................................................................4−23 Solution of Simultaneous Equations with Matrices ..............................................4−25 Summary ................................................................................................................4−32 Exercises ................................................................................................................4−36 Solutions to End−of−Chapter Exercises ................................................................4−38

MATLAB Computations: Pages 4−3, 4−5 through 4−8, 4−10, 4−12, 4−3, 4−5, 4−19 through 4−20, 4−24, 4−26, 4−28, 4−30, 4−38, 4−41, 4−43 Excel Computations: Pages 4−28 through 4−29, 4−42 through 4−43

5

Differential Equations, State Variables, and State Equations 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 5.10 5.11 5.12

5−1

Simple Differential Equations..................................................................................5−1 Classification............................................................................................................5−2 Solutions of Ordinary Differential Equations (ODE) .............................................5−6 Solution of the Homogeneous ODE ...................................................................... 5−8 Using the Method of Undetermined Coefficients for the Forced Response ........ 5−10 Using the Method of Variation of Parameters for the Forced Response ............. 5−20 Expressing Differential Equations in State Equation Form.................................. 5−24 Solution of Single State Equations ....................................................................... 5−27 The State Transition Matrix ................................................................................ 5−28 Computation of the State Transition Matrix ...................................................... 5−30 Eigenvectors.......................................................................................................... 5−38 Summary .............................................................................................................. 5−42

ii

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

5.13 Exercises ............................................................................................................... 5−47 5.14 Solutions to End−of−Chapter Exercises............................................................... 5−49 MATLAB Computations: Pages 5−11, 5−13 through 5−14, 5−16 through 5−17, 5−19, 5−23, 5−33 through 5−35, 5−37, 5−49 through 5−53, 5−55 Fourier, Taylor, and Maclaurin Series 6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 6.9 6.10 6.11 6.12 6.13 6−1 Wave Analysis ........................................................................................................6−1 Evaluation of the Coefficients ...............................................................................6−2 Symmetry ...............................................................................................................6−7 Waveforms in Trigonometric Form of Fourier Series .........................................6−12 Alternate Forms of the Trigonometric Fourier Series .........................................6−25 The Exponential Form of the Fourier Series .......................................................6−29 Line Spectra .........................................................................................................6−33 Numerical Evaluation of Fourier Coefficients .....................................................6−36 Power Series Expansion of Functions ..................................................................6−40 Taylor and Maclaurin Series ................................................................................6−41 Summary ..............................................................................................................6−48 Exercises ..............................................................................................................6−51 Solutions to End−of−Chapter Exercises ..............................................................6−53

6

MATLAB Computations: Pages 6−35, 6−45, 6−58 through 6−61 Excel Computations: Pages 6−37 through 6−39

7

Finite Differences and Interpolation 7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.8 7.9 7.10 7.11

7−1

Divided Differences ............................................................................................... 7−1 Factorial Polynomials ............................................................................................. 7−6 Antidifferences ................................................................................................... 7−12 Newton’s Divided Difference Interpolation Method ......................................... 7−15 Lagrange’s Interpolation Method ........................................................................ 7−17 Gregory−Newton Forward Interpolation Method .............................................. 7−19 Gregory−Newton Backward Interpolation Method ........................................... 7−21 Interpolation with MATLAB ............................................................................. 7−24 Summary ............................................................................................................. 7−39 Exercises ............................................................................................................. 7−44 Solutions to End−of−Chapter Exercises ............................................................. 7−45

MATLAB Computations: Pages 7−8 through 7−9, 7−13 through 7−15, 7−26 through 7−38, 7−45 through 7−46, 7−48, 7−50, 7−52 Excel Computations: Pages 7−17 through 7−19, 7−22 through 7−25, 7−49, 7−52

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

iii

8

Linear and Parabolic Regression 8.1 8.2 8.3 8.4 8.5 8.6 8.7

8−1

Curve Fitting ..........................................................................................................8−1 Linear Regression ...................................................................................................8−2 Parabolic Regression ..............................................................................................8−7 Regression with Power Series Approximations ....................................................8−14 Summary ..............................................................................................................8−24 Exercises ...............................................................................................................8−26 Solutions to End−of−Chapter Exercises ...............................................................8−28

MATLAB Computations: Pages 8−11 through 8−14, 8−17 through 8−23, 8−30 through 8−34 Excel Computations: Pages 8−5 through 8−10, 8−15 through 8−19, 8−28 through 8−32

9

Solution of Differential Equations by Numerical Methods 9.1 9.2 9.3 9.4 9.5 9.6 9.7

9−1

Taylor Series Method ............................................................................................ 9−1 Runge−Kutta Method ............................................................................................ 9−5 Adams’ Method ................................................................................................... 9−13 Milne’s Method .................................................................................................... 9−15 Summary .............................................................................................................. 9−17 Exercises .............................................................................................................. 9−20 Solutions to End−of−Chapter Exercises .............................................................. 9−21

MATLAB Computations: Pages 9−5, 9−9 through 9−12, 9−21 through 9−23 Excel Computations: Page 9−2, 9−14, 9−22 through 9−26

10

Integration by Numerical Methods 10.1 10.2 10.3 10.4 10.5

10−1

The Trapezoidal Rule .......................................................................................... 10−1 Simpson’s Rule ..................................................................................................... 10−6 Summary ............................................................................................................ 10−14 Exercises ............................................................................................................ 10−15 Solution to End−of−Chapter Exercises .............................................................. 10−16

MATLAB Computations: Pages 10−3 through 10−6, 10−9 through 10−13, 10−16, 10−18 through 10−21 Excel Computations: Pages 10−10, 10−19 through 10−21

11

Difference Equations

11−1

11.1 Introduction ......................................................................................................... 11−1 11.2 Definition, Solutions, and Applications .............................................................. 11−1 11.3 Fibonacci Numbers .............................................................................................. 11−7

iv

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

11.4 Summary .............................................................................................................11−11 11.5 Exercises ............................................................................................................. 11−13 11.6 Solutions to End−of−Chapter Exercises .............................................................11−14

12

Partial Fraction Expansion 12.1 12.2 12.3 12.4 12.5

12−1

Partial Fraction Expansion ..................................................................................12−1 Alternate Method of Partial Fraction Expansion ..............................................12−13 Summary ............................................................................................................12−19 Exercises ............................................................................................................12−22 Solutions to End−of−Chapter Exercises ............................................................12−23

MATLAB Computations: Pages 12−3 through 12−5, 12−9 through 12−12, 12−16 through 12-18, 12−23 through 12−28

13

The Gamma and Beta Functions and Distributions 13.1 13.2 13.3 13.4 13.5 13.6 13.7

13−1

The Gamma Function .........................................................................................13−1 The Gamma Distribution ..................................................................................13−16 The Beta Function .............................................................................................13−17 The Beta Distribution ........................................................................................13−20 Summary ............................................................................................................13−22 Exercises ............................................................................................................13−24 Solutions to End−of−Chapter Exercises ............................................................13−25

MATLAB Computations: Pages 13−3, 13−5, 13−10, 13−19, 13−25 Excel Computations: Pages 13−5, 13−10, 13−16 through 13−17, 13−19, 13−21

14

Orthogonal Functions and Matrix Factorizations 14.1 14.2 14.3 14.4 14.5 14.6 14.7 14.8 14.9 14.10 14.11

14−1

Orthogonal Functions ......................................................................................14−1 Orthogonal Trajectories ...................................................................................14−2 Orthogonal Vectors ..........................................................................................14−4 The Gram−Schmidt Orthogonalization Procedure ..........................................14−7 The LU Factorization .......................................................................................14−9 The Cholesky Factorization ............................................................................14−23 The QR Factorization .....................................................................................14−25 Singular Value Decomposition .......................................................................14−28 Summary .........................................................................................................14−30 Exercises .........................................................................................................14−32 Solutions to End−of−Chapter Exercises .........................................................14−34

MATLAB Computations: Pages 14−8 through 14−9, 14−11 through 14−29, 14−36, 14−38 through 14−39

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

v

15

Bessel, Legendre, and Chebyshev Functions 15.1 15.2 15.3 15.4 15.5 15.6 15.7

15−1

The Bessel Function ............................................................................................15−1 Legendre Functions ...........................................................................................15−10 Laguerre Polynomials .........................................................................................15−21 Chebyshev Polynomials .....................................................................................15−22 Summary ............................................................................................................15−27 Exercises .............................................................................................................15−32 Solutions to End−of−Chapter Exercises ............................................................15−33

MATLAB Computations: Pages 15−3 through 15−4, 15−6, 15−9, 14−19 through 15−22, 15−25, 15−33, 15−35 through 15−37 Excel Computations: Pages 15−5, 15−9

16

Optimization Methods 16.1 16.2 16.3 16.4 16.5 16.6

16−1

Linear Programming ........................................................................................... 16−1 Dynamic Programming ........................................................................................16−4 Network Analysis ...............................................................................................16−14 Summary ............................................................................................................16−19 Exercises .............................................................................................................15−20 Solutions to End−of−Chapter Exercises ............................................................15−22

MATLAB Computations: Pages 16−3 Excel Computations: Pages 16−4, 16−23, 16−25 through 16−27

A B C

Difference Equations in Discrete−Time Systems

A−1

A.1 Recursive Method for Solving Difference Equations........................................... A−1 A.2 Method of Undetermined Coefficients ................................................................A−1 MATLAB Computations: Pages A−4, A−7, A−9 Introduction to Simulink® B.1 B.2 B−1

Simulink and its Relation to MATLAB ............................................................... B−1 Simulink Demos .................................................................................................. B−20

MATLAB Computations and Simulink Modeling: Entire Appendix B Ill-Conditioned Matrices C−1

C.1 The Norm of a Matrix ...........................................................................................C−1 C.2 Condition Number of a Matrix .............................................................................C−2 C.3 Hilbert Matrices ....................................................................................................C−3

vi

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

MATLAB Computations: Pages C−1. Third Edition Copyright © Orchard Publications vii . C−4 through C−5 References Index R−1 IN−1 Numerical Analysis Using MATLAB® and Excel®.

Several examples are provided with detailed explanations. and return to the command screen to execute the program as explained below. and making plots. t and T are two different characters in MATLAB language. access to MATLAB’s Editor/ Debugger. and a right justified horizontal bar will denote the end of the example. Third Edition Copyright © Orchard Publications 1−1 . we see the toolbar on top of the command screen and the prompt EDU>>. a left justified horizontal bar will denote the beginning of an example. for example.m. that is. we will use the following conventions: Click: Click the left button of the mouse Courier Font: Screen displays Helvetica Font: User inputs at MATLAB’s command window prompt EDU>>* Helvetica Bold: MATLAB functions Bold Italic: Important terms and facts. To use the Editor/Debugger: 1. we choose New and click on M−File.1 Command Window To distinguish the screen displays from the user commands. myfile01. MATLAB now waits for a new command from the user. It is a good * EDU>> is the MATLAB prompt in the Student Version. We must save our program with a file name which starts with a letter. Also. it distinguishes between upper− and lower−case letters. From the File menu on the toolbar. Throughout this text. These bars will not be shown whenever an example begins at the top of a page or at the bottom of a page. 1. notes.m. Numerical Analysis Using MATLAB® and Excel®. finding the roots of a polynomial. save it. Important! MATLAB is case sensitive. We can use the Editor/Debugger to write our program. The files that we create are saved with the file name we use and the extension . important terms and MATLAB functions. procedures for naming and saving the user generated files. when one example follows immediately after a previous example. or open a previously saved file. the right justified bar will be omitted.Chapter 1 Introduction to MATLAB T his chapter is an introduction of the basic MATLAB commands and functions. comment lines. Thus. and file names When we first start MATLAB. This prompt is displayed also after execution of a command. This takes us to the Editor Window where we can type our script (list of statements) for a new file.

3. but leave only the EDU>> prompt on the upper left of the screen. We can do this periodically to become familiar with them. variables and equations which we have already used. A comment can be typed on the same line as the function or command or as a separate line. 2. and MATLAB returns to the command window.Chapter 1 Introduction to MATLAB practice to save the script in a file name that is descriptive of our script content. and equations without exiting. such as 3 – 4i or 3 – 4j . In other words. we type quit or exit. For example. The command help matlab iofun will display input/output information. we should use the clear command.q) % performs multiplication of polynomials p and q % The next statement performs partial fraction expansion of p(x) / q(x) are both correct. All text after the % (percent) symbol is interpreted by MATLAB as a comment line and thus it is ignored during the execution of a program.m extension at the EDU>> prompt. The clc command clears the screen but MATLAB still remembers all values. Third Edition Copyright © Orchard Publications . we type help matlab graphics. For instance. For example. We should also use a separate disk to backup our files. Whenever we want to return to the command window. to get information on graphics. we click on the Close button. If we have saved our file in drive a or any other drive. we type the file name without the . When we are done and want to leave MATLAB. Henceforth. we may exit the Editor/Debugger window by clicking on Exit Editor/Debugger of the File menu. the statements conv(p. To execute a program. it will be understood that each input command is typed after the EDU>> prompt and followed by the <enter> key. For instance. We can use either i . We can also get help from the Help pull− down menu. we can type help followed by any topic from the displayed menu. One of the most powerful features of MATLAB is the ability to do computations involving complex numbers. if the script performs some matrix operations. Once the script is written and saved as an m−file. But if we want to clear all previous values. or j to denote the imaginary part of a complex number. To get help with other MATLAB topics. variables. we must make sure that it is added it to the desired directory in MATLAB’s search path. it is like exiting MATLAB and starting it again.m or any other similar name. then. we ought to name and save that file as matrices01. The MATLAB User’s Guide provides more information on this topic. we can use the clc command. we type demo and we see the MATLAB Demos menu. if we want MATLAB to retain all previously entered commands. The MATLAB User’s Guide contains numerous help topics. To appreciate MATLAB’s capabilities. we press <enter> and observe the execution and the values obtained from it. This command erases everything. the statement z=3−4j 1−2 Numerical Analysis Using MATLAB® and Excel®.

if the imaginary part is a function or variable such as cos ( x ) . We can find the roots of any polynomial with the roots(p) function where p is a row vector containing the polynomial coefficients in descending order. The elements a i of this vector are the coefficients of the polynomial in descending order. and the roots as a column vector.0000 We observe that MATLAB displays the polynomial coefficients as a row vector.2 Roots of Polynomials In MATLAB. and the roots as roots_ p1. Third Edition Copyright © Orchard Publications 1−3 . We must include terms whose coefficients are zero.1 Find the roots of the polynomial p 1 ( x ) = x – 10x + 35x – 50x + 24 4 3 2 (1.0000 3. Example 1. p1=[1 −10 35 −50 24] % Specify the coefficients of p1(x) p1 = 1 -10 35 -50 24 roots_ p1=roots(p1) % Find the roots of p1(x) roots_p1 = 4. We have denoted the polynomial as p1. Numerical Analysis Using MATLAB® and Excel®.0000 2. that is.0000 .0000 1. we must use the multiplication sign. a multiplication (*) sign between 4 and j was not necessary because the complex number consists of numerical constants. a polynomial is expressed as a row vector of the form [ a n a n – 1 a 2 a 1 a 0 ] . 1.Roots of Polynomials displays z = 3. we must type cos(x)*j or j*cos(x). However.4.1) Solution: The roots are found with the following two statements.0000i In the example above.

say r3 . Compute the coefficients of this polynomial. and two complex roots.Chapter 1 Introduction to MATLAB Example 1. Of course. we find the coefficients with the poly(r) function as shown below. with the given roots as elements of this vector.7428 -1.3 Polynomial Construction from Known Roots We can compute the coefficients of a polynomial from a given set of roots with the poly(r) function where r is a row vector containing the roots. Third Edition Copyright © Orchard Publications . therefore. and the roots of this polynomial as roots_ p2.3366 + 1. p2=[1 −7 0 16 25 52] p2 = 1 -7 0 16 25 52 roots_ p2=roots(p2) roots_ p2 = 6. 1. 2.3 It is known that the roots of a polynomial are 1. The roots are found with the statements below where we have defined the polynomial as p2. then.3366 . 3. the conjugate of a complex number A = a + jb is A∗ = a – jb 1−4 Numerical Analysis Using MATLAB® and Excel®.5014 2. complex roots always occur in complex conjugate* pairs.2) Solution: There is no cube term. and 4 .5711 -0. we must enter zero as its coefficient. Solution: We first define a row vector.3202i -0.2 Find the roots of the polynomial p 2 ( x ) = x – 7x + 16x + 25x + 52 5 4 2 (1. Example 1.1.3202i The result indicates that this polynomial has three real roots. * By definition.

Solution: We form a row vector.0000i poly_r4 = 1 14 100 340 5 4 499 3 246 2 Therefore. the polynomial is p 4 ( x ) = x + 14x + 100x + 340x + 499x + 246 (1. – 3.1. Numerical Analysis Using MATLAB® and Excel®.4 It is known that the roots of a polynomial are – 1. say r4 .x) function evaluates a polynomial p ( x ) at some specified value of the independent variable x .0000 .0000 Column 5 -4. with the given roots.Evaluation of a Polynomial at Specified Values r3=[1 2 3 4] % Specify the roots of the polynomial r3 = 1 2 3 4 poly_r3=poly(r3) % Find the polynomial coefficients poly_r3 = 1 -10 35 -50 24 We observe that these are the coefficients of the polynomial p 1 ( x ) of Example 1.0000 + 5. Example 1. Find the coefficients of this polynomial.5. r4=[ −1 −2 −3 4+5j 4−5j ] r4 = Columns 1 through 4 -1.0000i poly_r4=poly(r4) -4.3) 1. 4 + j5.4 Evaluation of a Polynomial at Specified Values The polyval(p. Third Edition Copyright © Orchard Publications 1−5 . and we find the polynomial coefficients with the poly(r) function as shown below.0000 -2. and 4 – j5 . – 2.0000 -3.

No semicolon is used here % because we want the answer to be displayed p 5 ( x ) = x – 3x + 5x – 4x + 3x + 2 6 5 3 2 (1.Chapter 1 Introduction to MATLAB Example 1.b) function.b) − multiplies two polynomials a and b [q.4) val_minus3 = 1280 Other MATLAB functions used with polynomials are the following: conv(a. Example 1. Solution: p1=[1 −3 0 5 7 9]. % These are the coefficients % The semicolon (. -8 34 18 -24 -74 -88 78 166 174 108 1−6 Numerical Analysis Using MATLAB® and Excel®. % val_minus3=polyval(p5.6 Let p 1 = x – 3x + 5x + 7x + 9 p 2 = 2x – 8x + 4x + 10x + 12 6 4 2 5 4 2 (1.p2) p1p2 = 2 -6 Therefore. Third Edition Copyright © Orchard Publications .5) Compute the product p 1 ⋅ p 2 with the conv(a.d) −divides polynomial c by polynomial d and displays the quotient q and remainder r. polyder(p) − produces the coefficients of the derivative of a polynomial p. Solution: p5=[1 −3 0 5 −4 3 2].r]=deconv(c.) after the right bracket suppresses the display of the row vector % that contains the coefficients of p5.5 Evaluate the polynomial at x = – 3 . p1p2=conv(p1. p2=[2 0 −8 0 4 10 12]. −3)% Evaluate p5 at x=−3.

Commas will display the results whereas semicolons will suppress the display.p4) 6 5 2 7 5 3 (1. the quotient q ( x ) and remainder r ( x ) are q ( x ) = 0. [q.Evaluation of a Polynomial at Specified Values p 1 Þ p 2 = 2x 11 – 6x 5 10 – 8x + 34x + 18x – 24x 4 3 2 9 8 7 6 – 74x – 88x + 78x + 166x + 174x + 108 We can write MATLAB statements in one line if we separate them by commas or semicolons. Example 1.7) der_p5 = 12 Therefore. Solution: p3=[1 0 −3 0 5 7 9].5 Example 1.7 Let p 3 = x – 3x + 5x + 7x + 9 p 4 = 2x – 8x + 4x + 10x + 12 Compute the quotient p 3 ⁄ p 4 using the deconv(p.6) q = 0.8 Let p 5 = 2x – 8x + 4x + 10x + 12 Compute the derivative dp 5 ⁄ dx using the polyder(p) function. p4=[2 −8 0 0 4 10 12].q) function. Solution: p5=[2 0 −8 0 4 10 12]. der_p5=polyder(p5) 6 4 2 r ( x ) = 4x – 3x + 3x + 2x + 3 5 4 2 (1. 0 -32 0 8 10 Numerical Analysis Using MATLAB® and Excel®. Third Edition Copyright © Orchard Publications 1−7 .r]=deconv(p3.5000 r = 0 4 -3 0 3 2 3 Therefore.

2108 + 0.4922i -0.= ---------------------------------------------------------------------6 4 2 p den 2x – 8x + 4x + 10x + 12 (1.9870i 2. Example 1.4922 ) ⋅ ( x + 1. we have the factored form p num = ( x – 2.3370 .% Do not display num and den coefficients roots_num=roots(num).9961i roots_den = 1.9304 -1. that is.0712i -0.9870i -1.9961 ) ⋅ ( x + 0.9961i 1.9870 ) ⋅ ( x + 0. Solution: num=[1 −3 0 5 7 9].0000 ) 1−8 Numerical Analysis Using MATLAB® and Excel®. using the roots(p) function.1633 ) ⋅ ( x + 0.6760 + 0.5 Rational Polynomials Rational Polynomials are those which can be expressed in ratio form.4922 ) ⋅ ( x – 1.3370 + j0. the complex roots occur in complex conjugate pairs. For the numerator.3370 – j0.3370 + 0.9870 ) ⋅ ( x + 1. Third Edition Copyright © Orchard Publications .Chapter 1 Introduction to MATLAB dp 5 ⁄ dx = 12x – 32x + 4x + 8x + 10 5 3 2 1. we have p den = ( x – 1.= ------------------------------------------------------------------------------------------------------------------------------------------m m–1 m–2 Den ( x ) am x + am – 1 x + am – 2 x + … + a1 x + a0 n n–1 n–2 (1.4186 + 1.4186 + j1.9304 ) ⋅ ( x + 0.0.0000 As expected.4922i -0. den=[2 0 −8 0 4 10 12].0712i -0.6760 + j0.6760 .6760 – j0. roots_den=roots(den) % Display num and den roots roots_num = 2.9 Let 5 4 2 p num x – 3x + 5x + 7x + 9 R ( x ) = ----------.2108 – j 0.9) Express the numerator and denominator in factored form.0. as bn x + bn – 1 x + bn – 2 x + … + b1 x + b0 ) R ( x ) = Num ( x .2108 + j0.0712 ) ⋅ ( x + 1.0712 ) ⋅ ( x – 2.4186 – j1.1633 -1.0. We can find the roots of the numerator and denominator with the roots(p) function as before.1.2108 .8) where some of the terms in the numerator and/or denominator may be zero.4186 .9961 ) and for the denominator.

4186 −1.0712i) % Form the product of the 1st set of complex conjugates ans = 6.1.3370−0.0186 1. that is. Here. while the product of the roots is equal to the constant term c .Using MATLAB to Make Plots We can also express the numerator and denominator of this rational function as a combination of linear and quadratic factors. Numerical Analysis Using MATLAB® and Excel®.9870i) ans = 1. so we will use the English letter w instead. while the amplitude was held constant. This is a very easy task with the MATLAB plot(x. (2. that is. where the radian frequency ω (radians/second) of the applied voltage was varied from 300 to 3000 in steps of 100 radians/second. Z versus radian frequency ω . Example 1. x 1 ⋅ x 2 = c .4186+1.9961i)*(−0. x is the horizontal axis (abscissa) and y is the vertical axis (ordinate).0712i)*(2. The ammeter readings were then recorded for each frequency. we want to plot a set of ordered pairs.2108+0.3370+0. Accordingly.1058 (1. We recall that in a quadratic equation of the form x 2 + bx + c = 0 whose roots are x 1 and x 2 .4922i)*(1.6 Using MATLAB to Make Plots Quite often.9870i)*(−0. then we multiply the complex conjugate roots to obtain the constant term c using MATLAB as indicated below. that is. we form the coefficient b by addition of the complex conjugate roots and this is done by inspection.10 Consider the electric circuit of Figure 1. Solution: We cannot type ω (omega) in the MATLAB command window.6760−0.9961i) % Form the product of the 2nd set of complex conjugates ans = 1. Third Edition Copyright © Orchard Publications 1−9 . – ( x 1 + x 2 ) = b .1. the negative sum of the roots is equal to the coefficient b of the x term.6760+0.4922i) ans = 3.y) command which plots y versus x . Plot the magnitude of the impedance.9971 (−0.2108−0.0512 (−0. The magnitude of the impedance Z was computed as Z = V ⁄ A and the data were tabulated in Table 1.

.437 222.111 ω (rads/s) 1700 1800 1900 2000 2100 2200 2300 2400 2500 2600 2700 2800 2900 3000 Z Ohms 90. Also. or a row vector is too long to fit in one line.432 38.1.1 Table for Example 1.481 58.052 145.428 48.. then pressing <enter> to start a new line. % Use 4 periods to continue 1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500. and continue to enter data. as mentioned before.10 TABLE 1.513 62. it can be continued to the next line by typing three or more periods.032 187.938 469. Electric circuit for Example 1..Chapter 1 Introduction to MATLAB A R1 R2 C V L Figure 1. This is illustrated below for the data of w and z.182 40...184 97.589 71.588 67.056 1014..182 436.603 81. we use the semicolon (.122 42.) to suppress the display of numbers which we do not care to see on the screen.286 44. 1−10 Numerical Analysis Using MATLAB® and Excel®. Third Edition Copyright © Orchard Publications .83 266.611 51.088 73.751 120.665 140.353 103.10 ω (rads/s) 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 Z Ohms 39.717 46.845 If a statement.339 52. The data are entered as follows: w=[300 400 500 600 700 800 900 1000 1100 1200 1300 1400.240 54.

90.182 40.2.182 436.938 469. we click on the Window pull−down menu. and remains in effect unless canceled or overridden by the operator. Third Edition Copyright © Orchard Publications 1−11 . The command grid toggles them.665 140. We can make the above.432 38. MATLAB displays the plot on MATLAB’s graph screen. When this command is executed.z).122 42. and we select Figure. we use the plot(x. that is. we select MATLAB Command Window...240 54.286 44.032 187.111.. or any plot.481 58. For this example.353 103. The grid off command removes the grid. To plot z ( y – axis ) versus w ( x – axis ).611 51.. and we press <enter>. 1200 1000 800 600 400 200 0 0 500 1000 1500 2000 2500 3000 Figure 1.y) command. we simply type w or z. Numerical Analysis Using MATLAB® and Excel®. 1014. This plot is shown in Figure 1.603 81.052 145. * Default is a particular value for a variable or condition that is assigned automatically by an operating system. Of course.789 71.588 67..339 52. To return to the command window. 48. we use plot(w..513 62. To see the graph again.830 266.104 97. we press any key.437 222.468. or from the Window pull−down menu.Using MATLAB to Make Plots 2600 2700 2800 2900 3000]. % Use semicolon to suppress display of these numbers % z=[39. The default* is off.056..10 This plot is referred to as the amplitude frequency response of the circuit..088 73. Plot of impedance z versus frequency ω for Example 1. if we want to see the values of w or z or both.751 120. more presentable with the following commands: grid on: This command adds grid lines to the plot. changes from off to on or vice versa.2.717 46.845]..

title(‘string’): This command adds a line of the text string (label) at the top of the plot. We must remember. Modified frequency response plot of Figure 1. xlabel('w in rads/sec').2. Likewise. except that the x −axis is represented as a log scale. % Replaces the plot(w. Throughout this text.y) command is similar to the plot(x. it will be understood that log is the common (base 10) logarithm. voltage. Radian Frequency').y) command uses logarithmic scales for both axes.y) command that is similar to the plot(x.y) command. The loglog(x. If the y −axis represents power. Radian Frequency 10 w in rads/sec 3 10 4 Figure 1. our plot is as shown in Figure 1.Chapter 1 Introduction to MATLAB box off: This command removes the box (the solid lines which enclose the plot). grid. whereas the common logarithm is expressed as log10(x). xlabel(‘string’) and ylabel(‘string’) are used to label the x − and y −axis respectively. The command box toggles them. the logarithm to the base 2 is expressed as log2(x). except that the y −axis is represented as a log scale. however.z).3. 1200 1000 800 |Z| in Ohms 600 400 200 0 2 10 Magnitude of Impedance vs.3. and the x −axis as a linear scale. 1−12 Numerical Analysis Using MATLAB® and Excel®. and the y −axis as a linear scale.y) command. The default is on. Third Edition Copyright © Orchard Publications . the semilogy(x. and box on restores the box. by adding the following statements: semilogx(w. We can use the semilogx(x. or current.z) command title('Magnitude of Impedance vs. A review of the decibel unit follows. and the y −axis in dB (decibels) scale. Let us now redraw the plot with the above options. Likewise. the function log(x) in MATLAB is the natural logarithm. ylabel('|Z| in Ohms') After execution of these commands. the x −axis of the frequency response is more often shown in a logarithmic scale. and ln is the natural (base e) logarithm. The amplitude frequency response is usually represented with the x −axis in a logarithmic scale.

we say that an amplifier has 10 dB power gain.5 Likewise. the dB values for voltage and current ratios become Numerical Analysis Using MATLAB® and Excel®. we can estimate other values. Thus. 000 60 dB represents a power ratio of 1.25 3 dB represents a power ratio of approximately 2 7 dB represents a power ratio of approximately 5 From these. Third Edition Copyright © Orchard Publications 1−13 . 000 Also. 000.= I Z -----Z 2 2 n if we let Z = 1 . 10 dB represents a power ratio of 10 10n dB represents a power ratio of 10 It is very useful to remember that: 20 dB represents a power ratio of 100 30 dB represents a power ratio of 1. 27 dB = 20 dB + 7 dB and this is equivalent to a power ratio of approximately 100 × 5 = 500 Using the relations y = log x = 2 log x and 2 P = V.25 then.10) Therefore. If the gain (or loss) is 0 dB the output is equal to the input. 1 dB represents a power ratio of approximately 1. voltage. or current) can be expressed in decibels (dB). 4 dB = 3 dB + 1 dB and since 3 dB ≅ power ratio of 2 and 1 dB ≅ power ratio of 1. 4 dB ≅ ratio of ( 2 × 1. For instance. or a transmission line has a power loss of 7 dB (or gain – 7 dB ).25 ) = ratio of 2. P out dB = 10 log --------P in (1. By definition.Using MATLAB to Make Plots The ratio of any two values of the same quantity (power.

The first line of the script below has the form linspace(first_value. MATLAB has no default color.95. there is no default marker.65. and colors for our plots. markers. For instance.x.y.Chapter 1 Introduction to MATLAB V out dB v = 10 log --------V in 2 and V out = 20 log --------V in (1. we add the relation dB=20*log10(v).x. % turn grid and axes box on text(0. y=sin(x). and string is the label which we want to place at that location. plot(x.v).2.y. it starts with blue and cycles through the first seven colors listed in Table 1. 'sin(x)'). The command text(x.4.z) command with semilogx(w. we can move the cross−hair to the position where we want our label to begin. An alternate command uses the colon notation and has the format x=first: increment: last This format specifies the increments between points but not the number of data points. These.65. 0. The command gtext(‘string’) switches to the current Figure Window. We will illustrate its use with the following example which plots a 3−phase sinusoidal waveform.u. and displays a cross−hair which can be moved around with the mouse. last_value. 60). or a combination of these.12) I out 2 I out dB i = 10 log ------.’string’) is similar to gtext(‘string’). and we press <enter>. where s is a character string containing one or more characters shown on the three columns of Table 1. number_of_values) This command specifies the number of data points but not the increments between data points.75. we did not specify line styles. 0.11) (1. MATLAB allows us to specify various line types. text(2. text(4.s) command. Third Edition Copyright © Orchard Publications .= 20 log ------I in I in To display the voltage v in a dB scale on the y – axis . plot symbols.2 for each additional line in the plot. It places a label on a plot in some specific location specified by x and y. % pi is a built−in function in MATLAB. In our previous examples. we can use the command gtext(‘Impedance |Z| versus Frequency’). The script for the 3−phase plot is as follows: x=linspace(0.dB). and colors.65.y. 'sin(x+2*pi/3)'). box on. However. % we could have used x=0:0.02: 2)*pi instead.85. 'sin(x+4*pi/3)') These three waveforms are shown on the same plot of Figure 1. using the mouse. u=sin(x+2*pi/3). % The x−axis must be specified for each function grid on. and this will place a cross−hair in the Figure window. Then. v=sin(x+4*pi/3).02*pi:2*pi or x = (0: 0. Also. can be added with the plot(x. no markers are 1−14 Numerical Analysis Using MATLAB® and Excel®. and we replace the semilogx(w. 2*pi. 0.

and markets used in MATLAB Symbol b g r c m y k w Color blue green red cyan magenta yellow black white Symbol Marker point circle x−mark plus star square diamond triangle down triangle up triangle left triangle right pentagram hexagram Symbol Line Style solid line dotted line dash−dot line dashed line . Third Edition Copyright © Orchard Publications 1−15 . 1 sin(x) 0.'rs−').5 -1 0 1 2 3 4 5 6 7 Figure 1.y. The default line is the solid line. For additional information we can type help plot in MATLAB’s command screen.'rs') plots a red square at each data point. −− ⁄ Ÿ < > p h For example. o x + * s d − : −. If we want to connect the data points with a solid line. colors. and plot(x. Three−phase waveforms TABLE 1. Numerical Analysis Using MATLAB® and Excel®.5 sin(x+2*pi/3) sin(x+4*pi/3) 0 -0.2 Styles.4.Using MATLAB to Make Plots drawn unless they are selected. we must type plot(x. the command plot(x. but does not draw any line because no line was selected.'m*:') plots a magenta dotted line with a star at each data point.y.y.

x= −10: 0. and dot exponentiation where these operations are preceded by a dot (period).x3.^3+x+3.*x. where x .z).y. yn.z3.5. z= −2.z2.13) The three−dimensional plot is shown in Figure 1. Page 1−19.y3. y= x. Example 1. These operations will be explained in Section 1.y2. that is. These strings are the same as those of the two−dimensional plots.x2.s2.*y. marker symbol. and zn are vectors or matrices.5. plot3(x.5: 10. Third Edition Copyright © Orchard Publications .. y . and z .z1.z) plots a line in 3−space through the points whose coordinates are the elements of x . y .y1.11 * This statement uses the so called dot multiplication. The command plot3(x. they are drawn on two axes.^2−1. dot division.11 Plot the function z = – 2x + x + 3y – 1 Solution: We arbitrarily choose the interval (length) shown with the script below. 1−16 Numerical Analysis Using MATLAB® and Excel®.. The general format is plot3(x1. MATLAB has also a three−dimensional (three−axes) capability and this is discussed next.y. and z are three vectors of the same length. or line style.) where xn.s1.s3.8. Three dimensional plot for Example 1. and sn are strings specifying color.Chapter 1 Introduction to MATLAB The plots which we have discussed thus far are two−dimensional. 3000 2000 1000 0 -1000 -2000 10 5 0 -5 -10 -10 -5 0 5 10 Figure 1.. grid % Length of vector x % Length of vector y must be same as x % Vector z is function of both x and y* 3 2 (1.

connected with a solid line. j ) } . contour(Z. we must first generate the X and Y matrices which consist of repeated rows and columns over the range of the variables x and y .z. This must be done for each plot. Third Edition Copyright © Orchard Publications 1−17 . Example 1. we can use the zlabel(‘string’) command in addition to the xlabel(‘string’) and ylabel(‘string’). The three−dimensional text(x. and exponentiation. y. We observe that x corresponds to the columns of Z . h ) The three−dimensional plot is created with the following MATLAB script where. draws contour lines for n levels. as in the previous example.y) function which creates the matrix X whose rows are copies of the vector x. the vertices of the mesh lines are the triples { x ( j ). y ( i ).Y]=meshgrid(x.z) commands.Using MATLAB to Make Plots The command plot3(x. in the second line we have used the dot multiplication. Likewise. Numerical Analysis Using MATLAB® and Excel®. Solution: The volume of the cone is a function of both the radius r and the height h . To produce a mesh plot of a function of two variables. grid on and box off are the default states.12 The volume V of a right circular cone of radius r and height h is given by 2 V = 1 πr h -3 (1. in a three−dimensional plot we can set the limits of all three axes with the axis([xmin xmax ymin ymax zmin zmax]) command.z) command displays a three−dimensional plot.’string’) command will place string beginning at the co−ordinate ( x.z) command with two vector argum e n t s . n ] = size ( Z ) . Z ( i. z ) on the plot. We can generate the matrices X and Y with the [X.y. As mentioned in the footnote of the previous page.8. we can set the limits of the x − and y − axes with the axis([xmin xmax ymin ymax]) command.'bd−') will display the plot in blue diamonds. y ) .y) or plot3(x. In a three−dimensional plot. T h e s e m u s t b e d e f i n e d a s length ( x ) = n a n d length ( y ) = m w h e r e [ m. The mesh(x. say z = f ( x.y.y. We can also use the mesh(x. Page 1−19. and the matrix Y whose columns are copies of the vector y. V = f ( r. Another command. that is. this topic will be explained in Section 1.14) Plot the volume of the cone as r and h vary on the intervals 0 ≤ r ≤ 4 and 0 ≤ h ≤ 6 meters.y. It must be placed after the plot(x.n).z. In this case. In a two−dimensional plot.y. For three−dimensional plots. division. or on the same line without first executing the plot command. and y corresponds to the rows of Z .

n. volume (cubic meters)'). 0: 6). MATLAB can generate very sophisticated and impressive three−dimensional plots. 1. Volume of Right Circular Cone z-axis. We will illustrate the use of the subplot(m.^ 2 . and p . Volume of a right circular cone.H]=meshgrid(0: 4. H. V) xlabel('x−axis.5. 1−18 Numerical Analysis Using MATLAB® and Excel®.7 Subplots MATLAB can display up to four windows of different plots on the Figure window using the command subplot(m.Chapter 1 Introduction to MATLAB [R. The possible combinations are shown in Figure 1. n . No spaces or commas are required between the three integers m .6. and the plot of Figure 1. division and exponentiation that follows.* R . mesh(R. This command divides the window into an m × n matrix of plotting areas and chooses the pth area to be active. box on The three−dimensional plot of Figure 1. % Creates R and H matrices from vectors r and h V=(pi . radius r (meters)'). altitude h (meters) x-axis. Third Edition Copyright © Orchard Publications . shows how the volume of the cone increases as the radius and height are increased.7. title('Volume of Right Circular Cone').p). This.n. are rudimentary. volume (cubic meters) 150 100 50 0 6 4 2 0 0 1 2 3 4 y-axis. altitude h (meters)'). The MATLAB User’s manual contains more examples.6. radius r (meters) Figure 1.* H) ./ 3. ylabel('y−axis.p) command following the discussion on multiplication. zlabel('z−axis.

the arrays [ a b c … ] . 6. and exponentiation. They are explained in the following paragraphs. and exponentiation. and thus are called row vectors. We recall that the elements of a row vector are separated by spaces. is to write it first as a row vector. To distinguish between row and column vectors. the elements of a column vector must be separated by semicolons. An easier way to construct a column vector. it is called a column vector. division.Multiplication. a column vector can be written either as b=[−1.7. and then transpose it into a column vector. division. 11] b = -1 3 6 11 b=[−1 3 6 11]' b = -1 3 Numerical Analysis Using MATLAB® and Excel®. Third Edition Copyright © Orchard Publications 1−19 . 3. division. 11] or as b=[−1 3 6 11]' MATLAB produces the same display with either format as shown below. Possible subpot arrangements in MATLAB 1. such a those that contained the coefficients of polynomials. and the element−by−element multiplication. Division and Exponentiation 111 Full Screen 211 212 221 222 212 221 223 211 223 224 222 224 221 223 122 Default 121 122 222 224 121 Figure 1.8 Multiplication. b=[−1. These are the matrix multiplication. In Section 1. and exponentiation. 3. If an array has one column and multiple rows.2. Division and Exponentiation MATLAB recognizes two types of multiplication. consisted of one row and multiple columns. 6. MATLAB uses the single quotation character (¢) to transpose a vector. Thus.

MATLAB expects 1−20 Numerical Analysis Using MATLAB® and Excel®. multiplication of the row vector A by the column vector B . as indicated by the transpose operator (′). 1. Here. We observe that A is defined as a row vector whereas B is defined as a column vector. A*B = [ a 1 b 1 + a 2 b 2 + a 3 b 3 + … + a n b n ] = sin gle value For example. A*B % No transpose operator (‘) here When these statements are executed. if and A = [1 2 3 4 5] B = [ – 2 6 – 3 8 7 ]' the matrix multiplication A*B produces the single value 68. MATLAB displays the following message: ??? Error using ==> * Inner matrix dimensions must agree. B=[ −2 6 −3 8 7]'. Third Edition Copyright © Orchard Publications . because we have used the matrix multiplication operator (*) in A*B. A*B % Observe transpose operator (‘) in B (B.15) ans = 68 Now. B=[−2 6 −3 8 7]. Here.Chapter 1 Introduction to MATLAB 6 11 We will now define Matrix Multiplication and Element−by−Element multiplication. Then. let us suppose that both A and B are row vectors. Matrix Multiplication (multiplication of row by column vectors) Let A = [ a1 a2 a3 … an ] and B = [ b 1 b 2 b 3 … b n ]' be two vectors. that is. and we attempt to perform a row−by− row multiplication with the following MATLAB statements. A∗ B = 1 × ( – 2 ) + 2 × 6 + 3 × ( – 3 ) + 4 × 8 + 5 × 7 = 68 and this is verified with the MATLAB script A=[1 2 3 4 5]. A=[1 2 3 4 5]. is performed with the matrix multiplication operator (*).

) is never required with the plus (+) and minus (−) operators. multiplication of the row vector C by the row vector D is performed with the dot multiplication operator (.^) are used for element− by−element division and exponentiation. Third Edition Copyright © Orchard Publications 1−21 . 2.*D % Vectors C and D must have % same number of elements % We observe that this is a dot multiplication ans = -2 12 -9 32 35 Similarly. Here.∗ D = [ c 1 d 1 c2 d2 c3 d3 … cn dn ] (B. There is no space between the dot and the multiplication symbol. It recognizes that B is a row vector. Accordingly. Element−by−Element Multiplication (multiplication of a row vector by another row vector) Let C = [ c1 c2 c3 … cn ] and D = [ d1 d2 d3 … dn ] be two row vectors. Numerical Analysis Using MATLAB® and Excel®. and warns us that we cannot perform this multiplication using the matrix multiplication operator (*). the division (/) and exponentiation (^) operators.16) This product is another row vector with the same number of elements./) and dot exponentiation (. we must perform this type of multiplication with a different operator. not a row vector. D=[−2 6 −3 8 7]. C. C.∗ D = 1 × ( – 2 ) 2 × 6 3 × ( – 3 ) 4 × 8 5 × 7 = – 2 12 – 9 32 35 Check with MATLAB: C=[1 2 3 4 5].) and the multiplication (*). As an example. Note: A dot (. let and C = [1 2 3 4 5] D = [ –2 6 –3 8 7 ] Dot multiplication of these two row vectors produce the following result. as the elements of C and D . division ( /). whereas dot division (.Multiplication. are used for matrix division and exponentiation. Thus. and exponentiation (^) operators. We must remember that no space is allowed between the dot (.*). as illustrated with the examples above. This operator is defined below. Division and Exponentiation vector B to be a column vector. C.

* t) . Solution: The MATLAB script for this example is as follows: t=0: 0.* exp(−3 . xlabel('t').* exp(−4 .17) in the 0 ≤ t ≤ 5 seconds interval. ymin and ymax. plot(t. This is because the last term (the rational fraction) of the given expression. MATLAB would have displayed the following message: Warning: Divide by zero.y). % Define t−axis in 0.01 increments y=3 . To avoid division by zero. 6 4 y=f(t) 2 0 -2 Plot for Example 1.13 0 1 2 t 3 4 5 Figure 1. and exponentiation.2 × 10 – 16 . This command must be placed after the plot command and must be repeated for each plot. ylabel('y=f(t)'). xmax. It will be used with the next example. The following example illustrates the use of the dot multiplication. and also MATLAB’s capability of 1−22 Numerical Analysis Using MATLAB® and Excel®. which is a number approximately equal to 2. grid.Chapter 1 Introduction to MATLAB Example 1.* t) . say as – 3 ≤ t ≤ 3 . There are no commas between these four arguments.* sin(2 . Third Edition Copyright © Orchard Publications .* t)−2 .01: 5.8.8 shows the plot for this example. defined the time interval starting with a negative value equal to or less than – 1 .13 Write the MATLAB script that produces a simple plot for the waveform defined as y = f ( t ) = 3e –4 t 2 cos 5t – 2e –3 t t sin 2t + -----------t+1 (1. is divided by zero when t = – 1 .* t) + t .^2 . title('Plot for Example 1.13 Had we./ (t+1). The command axis([xmin xmax ymin ymax]) scales the current plot to the values specified by the arguments xmin.13') Figure 1. the eps number. we use the special MATLAB function eps. division. Plot for Example 1.* cos(5 . in the example above. the axis([xmin xmax ymin ymax]) command.

Use the subplot command to display these functions on four windows on the same graph.9. 1 y=(sinx)2 1 z=(cosx)2 0. title('v=(sinx)^2/(cosx)^2'). Third Edition Copyright © Orchard Publications 1−23 .5 0 0 2 2 4 2 6 0 0 2 4 6 w=(sinx) *(cosx) 0. title('y=(sinx)^2'). subplot(222). % Interval with 100 data points % add eps to avoid division by zero % upper left of four subplots % upper right of four subplots % lower left of four subplots % lower right of four subplots These subplots are shown in Figure 1.14 Numerical Analysis Using MATLAB® and Excel®.2 400 v=(sinx)2/(cosx)2 200 0.1 0 0 2 4 6 0 0 2 4 6 Figure 1. z=(cos(x) .18) in the interval 0 ≤ x ≤ 2π using 100 data points.y)./ (z+eps). axis([0 2*pi 0 1]).w). axis([0 2*pi 0 1]).100). Division and Exponentiation displaying up to four windows of different plots.v). v = sin 2x ⁄ cos 2x (1.^ 2).z). plot(x. y=(sin(x) .* z. plot(x. subplot(224). 2*pi.5 0. w = sin 2x ⋅ cos 2x. subplot(223). axis([0 2*pi 0 400]).14 Plot the functions y = sin 2x.9. title('z=(cosx)^2'). Example 1. Subplots for the functions of Example 1. plot(x. plot(x. Solution: The MATLAB script to produce the four subplots is as follows: x=linspace(0. title('w=(sinx)^2*(cosx)^2').3]). axis([0 2*pi 0 0. z = cos 2x. subplot(221).^ 2).Multiplication. v=y . w=y .

Plot Re { Z } (the real part of the impedance Z ) versus frequency ω . c. and the round(n) function that rounds a number to its nearest integer. theta is the angle in radians./ (10 + j . Solution: The MATLAB script below computes the real and imaginary parts of Z ab that is.19) 1−24 Numerical Analysis Using MATLAB® and Excel®. Example 1.^ 6 . for simplicity.* (0. It also produces a polar plot (part c). w=0: 1: 2000. and the angle(z) functions that compute the absolute value (magnitude) and phase angle of the complex quantity z = x + iy = r –θ .* 10 . We will introduce the real(z) and imag(z) functions which display the real and imaginary parts of the complex quantity z = x + iy.10.Chapter 1 Introduction to MATLAB The next example illustrates MATLAB’s capabilities with imaginary numbers.1 H Z ab Figure 1.15 Consider the electric circuit of Figure 1. Electric circuit for Example 1. and plots these as two separate graphs (parts a & b).^5.1 .r) function that produces a plot in polar coordinates.* w −10./ (w+eps)))). and capacitance. inductance. 10 – j ( 10 ⁄ w ) Z ab = Z = 10 + ------------------------------------------------------5 10 + j ( 0.10./ (w+eps)) . where r is the magnitude. Third Edition Copyright © Orchard Publications . % Define interval with one radian interval z=(10+(10 . a 10 Ω 10 Ω 10 μF 0. Plot the impedance Z versus frequency ω in polar coordinates. the impedance Z ab as a function of the radian frequency ω can be computed from the following expression. denoted as z . the abs(z).^ 4 −j . 4 6 (1. b.1 w – 10 ⁄ w ) a. Plot Im { Z } (the imaginary part of the impedance Z ) versus frequency ω . We will also use the polar(theta.15 b With the given values of resistance.

grid.12. plot(w. ylabel('Real part of Z'). plot(w. grid. % % The next five statements (next two lines) compute and plot Im{z} imag_part=imag(z). and 1.13 respectively. Third Edition Copyright © Orchard Publications 1−25 . 1200 1000 800 Real part of Z 600 400 200 0 0 200 400 600 800 1000 1200 radian frequency w 1400 1600 1800 2000 Figure 1. ylabel('Polar Plot of Z'). Division and Exponentiation % % The first five statements (next two lines) compute and plot Re{z} real_part=real(z). imaginary.11.% Computes the phase angle of impedance Z polar(theta. 1. ylabel('Imaginary part of Z').% Angle is the first argument grid.Multiplication.% Rounds |Z| to read polar plot easier theta=angle(z). The real.rndz). xlabel('radian frequency w').% Computes |Z| rndz=round(abs(z)).15 Numerical Analysis Using MATLAB® and Excel®.imag_part). xlabel('radian frequency w'). and polar plots are shown in Figures 1.11. Plot for the real part of Z in Example 1.real_part). % The last six statements (next six lines) below produce the polar plot of z mag=abs(z).

Both types are referred to as m−files since both require the .Chapter 1 Introduction to MATLAB 600 400 Imaginary part of Z 200 0 -200 -400 -600 0 200 400 600 800 1000 1200 radian frequency w 1400 1600 1800 2000 Figure 1. the script for each of the examples we discussed earlier.12. make up a script file.15 90 120 150 Polar Plot of Z 1500 1000 60 30 500 180 0 210 240 270 300 330 Figure 1. Third Edition Copyright © Orchard Publications . Plot for the imaginary part of Z in Example 1.9 Script and Function Files MATLAB recognizes two types of files: script files and function files. a script file is one which was generated and saved as an m−file with an editor such as the MATLAB’s 1−26 Numerical Analysis Using MATLAB® and Excel®.m extension. fast. Thus.13.15 clearly illustrates how powerful. 1. accurate. A script file consists of two or more built−in functions such as those we have discussed thus far. Polar plot of Z in Example 1. and flexible MATLAB is.15 Example 1. Generally.

5 Solution: We first plot this function to observe the approximate zeros. MATLAB searches for a value near a point where the function f changes sign.2 ) + 0. it is MATLAB’s response to an undefined expression such as 0 ⁄ 0 . We use function files for repetitive tasks. To save it. The function name and file name must be the same.5 ≤ x ≤ 1. Important: We must remember that we use roots(p) to find the roots of polynomials only.m. the function file consisting of the two lines below function y = myfunction(x) y=x . NaN (Not−a−Number) is not a function.x) tries to find a zero of a function of one variable.^ 3 + cos(3 . which we mentioned earlier.– 10 2 2 ( x – 0.1 ) + 0. followed by the output argument. For example.04 in the interval – 1.20) Numerical Analysis Using MATLAB® and Excel®. the equal sign ( = ). The function fzero(f. A function file is a user−defined function using MATLAB. Example 1. The string fcn must be the name of an m−file function or a string with variable x . maxima and minima of the function 1 1 f ( x ) = ----------------------------------------.1 and 1.Script and Function Files Editor/Debugger.+ ----------------------------------------. fplot(fcn. but the file name must have the extension . and minima using the following script: (1. We will use the following MATLAB functions with the next example. and the input argument enclosed in parentheses.16 Find the zeros. or returns NaN if the search fails. We can avoid division by zero using the eps number. from the File menu of the command window. where f is a string containing the name of a real−valued function of a single real variable. such as those in Examples 1.lims) − plots the function specified by the string fcn between the x−axis limits specified by lims = [xmin xmax]. maxima. we choose New and click on M−File. ∞ ⁄ ∞ .m. Using lims = [xmin xmax ymin ymax] also controls the y−axis limits. This takes us to the Editor Window where we type these two lines and we save it as myfunction. or inability to produce a result as described on the next paragraph.2. and returns that value. The first line of a function file must contain the word function.01 ( x – 1. Third Edition Copyright © Orchard Publications 1−27 .* x) is a function file and must be saved.

2). MATLAB appends the extension . grid This plot is shown in Figure 1.Chapter 1 Introduction to MATLAB x=−1. Next.y).14./ ((x−1. As expected. approximately. To save this file. we can use the fplot(fcn. we type the script above and we save it as funczero01.5 -1 -0.01: 1.14 which was obtained with the plot(x. y min = – 34 .14. we define and save f(x) as the funczero01. Now.15. and when the Editor Window appears.m function m−file with the following script: function y=funczero01(x) % Finding the zeros of the function shown below y=1/((x−0. Plot for Example 1.m to it. y=1.1 where. [−1. this plot is identical to the plot of Figure 1.01)−1/((x−1.5]).2)^2+0.04)−10. 100 80 60 40 20 0 -20 -40 -1.2 and x = 0.5 1. Third Edition Copyright © Orchard Publications .5 0 0. from the File drop menu on the Command Window.1)^2+0.14. y max = 90 . we choose New./ ((x−0.16 using the plot command The roots (zeros) of this function appear to be in the neighborhood of x = – 0.5 1 1. and the minimum occurs at approximately x = 1.^ 2 + 0.^ 2 + 0.lims) command to plot f ( x ) as follows: fplot('funczero01'. approximately.y) command as shown in Figure 1. 1−28 Numerical Analysis Using MATLAB® and Excel®.5.2 where.3 . The maximum occurs at approximately x = 0.1). plot(x. grid The plot is shown in Figure 1.01) −1.04) −10.5 Figure 1.5: 0.

Script and Function Files 100 80 60 40 20 0 -20 -40 -1.1)^2+0.16 using the fplot command We will use the fzero(f.20) more precisely. From elementary calculus. x1).. Thus. Plot for Example 1. Third Edition Copyright © Orchard Publications 1−29 .5 1 1.x1..5 Figure 1. x2= fzero('funczero01'. This can be visualized by flipping the plot of a function f ( x ) upside−down. we recall that the maxima or minima of a function y = f ( x ) can be found by setting the first derivative of a function equal to zero and solving for the independent variable x .5 -1 -0..1919 and r2= 0.x2) and with this function we could compute both a minimum of some function f ( x ) or a maximum of f ( x ) since a maximum of f ( x ) is equal to a minimum of – f ( x ) .01)−1/((x−1.3788 The earlier MATLAB versions included the function fmin(f. The MATLAB script below will accomplish this.2)^2+0. For this example we use the diff(x) function which produces the approximate derivative of a function.2). fprintf(' and r2= %3.3). −0. This function is no longer used in MATLAB and thus we will compute the maxima and minima from the derivative of the given function.15. zmin=diff(ymin) zmin = -1/((x-1/10)^2+1/100)^2*(2*x-1/5)+1/((x-6/5)^2+1/25)^2*(2*x-12/5) When the command Numerical Analysis Using MATLAB® and Excel®. fprintf('The roots (zeros) of this function are r1= %3. x1= fzero('funczero01'.5 0 0.x) function to compute the roots of f ( x ) in Equation (1. 0.4f'.04)−10.4f \n'. we use the following MATLAB script: syms x ymin zmin. x2) MATLAB displays the following: The roots (zeros) of this function are r1= -0. ymin=1/((x−0.

04)+10.16.1) ^ 2 + 0.01:1.5:0.04) −10 ymin = -34.6585 . Third Edition Copyright © Orchard Publications . ymax=−1.3437i ans = 0.2) ^ 2 + 0. grid and the plot is shown in Figure 1. that is.5.3437i ans = 1. ymax=−(1/((x−0.1)^2+0. plot(x. To find the value of y corresponding to this value of x.Chapter 1 Introduction to MATLAB solve(zmin) is executed./((x−1.2).5 1 1.2)^2+0.5 -1 -0.^2+0.1812 We can find the maximum value from – f ( x ) whose plot is produced with the script x=−1.01) −1 / ((x−1.04)−10). zmax=diff(ymax) 1−30 Numerical Analysis Using MATLAB® and Excel®. MATLAB displays a very long expression which when copied at the command prompt and executed.ymax).5 0 0.16 Next we compute the first derivative of – f ( x ) and we solve for x to find the value where the maximum of ymax occurs. This is accomplished with the MATLAB script below.01)+1. we substitute it into f ( x ) ./((x−0.2012.2012 above is the value of x at which the function y has its minimum value as we observe also in the plot of Figure 1.16.2012 The real value 1.^2+0. ymin=1 / ((x−0. x=1. produces the following: ans = 0.5 Figure 1. syms x ymax zmax. Plot of – f ( x ) for Example 1. 40 20 0 -20 -40 -60 -80 -100 -1.0.1).6585 + 0.01)−1/((x−1.15.

MATLAB displays a very long expression which when copied at the command prompt and executed.2) ^ 2 + 0. The format displayed has nothing to do with the accuracy in the computations. FORMAT may be used to switch between different output display formats as follows: Numerical Analysis Using MATLAB® and Excel®. or in short floating point format with four decimals if it a fractional number.0999.3437i ans = 1. x=0. MATLAB performs all computations with accuracy up to 16 decimal places.2000 1.6585 . Third Edition Copyright © Orchard Publications 1−31 .10 Display Formats MATLAB displays the results on the screen in integer format without decimals if the result is an integer number.16. The output format can changed with the format command.6585 + 0.Display Formats zmax = 1/((x-1/10)^2+1/100)^2*(2*x-1/5)-1/((x-6/5)^2+1/25)^2*(2*x-12/5) solve(zmax) When the command solve(zmax) is executed. Accordingly. All computations in MATLAB are done in double precision. we execute the following script to obtain the value of ymin .04) −10 ymax = 89.0999 which is consistent with the plots of Figures 1.01) −1 / ((x−1. produces the following: ans = 0.3437i ans = 0.0. % Using this value find the corresponding value of ymax ymax=1 / ((x−0.2012 ans = 0.0999 From the values above we choose x = 0. The available formats can be displayed with the help format command as follows: help format FORMAT Set output format.15 and 1.1) ^ 2 + 0.

.33333333333334 16 digits format short e 3. negative and zero elements. format short 33. FORMAT + The symbols +. Same as SHORT. SHORT Scaled fixed point format with 5 digits. FORMAT LOOSE Puts the extra line-feeds back in.and blank are printed for positive. FORMAT LONG G Best of fixed or floating point format with 15 digits. Third Edition Copyright © Orchard Publications . Spacing: FORMAT COMPACT Suppress extra line-feeds.33 two decimal digits format + only + or − or zero are printed format rat 100/3 rational approximation 1−32 Numerical Analysis Using MATLAB® and Excel®.3333e+01 Four decimal digits plus exponent format short g 33. Imaginary parts are ignored. FORMAT RAT Approximation by ratio of small integers. LONG E Floating point format with 15 digits. LONG Scaled fixed point format with 15 digits.3335 Four decimal digits (default) format long 33. SHORT G Best of fixed or floating point format with 5 digits. FORMAT HEX Hexadecimal format. SHORT E Floating point format with 5 digits. Some examples with different format displays age given below. FORMAT BANK Fixed format for dollars and cents.Chapter 1 Introduction to MATLAB FORMAT FORMAT FORMAT FORMAT FORMAT FORMAT Default.333 Better of format short or format short e format bank 33.

the command help matlab\iofun will display input/output information. • The conv(a. A comment can be typed on the same line as the function or command or as a separate line. We must include terms whose coefficients are zero. we type demo and we see the different topics. or j to denote the imaginary part of the complex number. Third Edition Copyright © Orchard Publications 1−33 . • The polyder(p) function produces the coefficients of the derivative of a polynomial p. Whenever we want to return to the command window.11 Summary • We can get help with MATLAB topics by typing help followed by any topic available. • In MATLAB.Summary 1. The elements a i of this vector are the coefficients of the polynomial in descending order. Commas will display the results whereas semicolons will suppress the display. • The [q. • The MATLAB Demos menu displays MATLAB’s capabilities. Numerical Analysis Using MATLAB® and Excel®. • We type quit or exit when we are done and want to leave MATLAB. • We find the roots of any polynomial with the roots(p) function where p is a row vector containing the polynomial coefficients in descending order. variables and equations which we have already used. • All text after the % (percent) symbol is interpreted by MATLAB as a comment line and thus it is ignored during the execution of a program.b) function multiplies the polynomials a and b. a polynomial is expressed as a row vector of the form [ a n a n – 1 a 2 a 1 a 0 ] .r]=deconv(c.d) function divides polynomial c by polynomial d and displays the quotient q and remainder r. and equations without exiting.x) function evaluates a polynomial p ( x ) at some specified value of the independent variable x . we click on the Close button. • We use the clear command if we want to clear all previous values. • For computations involving complex numbers we can use either i . and help matlab graphics will display help on graphics. • We can write MATLAB statements in one line if we separate them by commas or semicolons. • The polyval(p. variables. • The clc command clears the screen but MATLAB still remembers all values. • We can compute the coefficients of a polynomial from a given set of roots with the poly(r) function where r is a row vector containing the roots. To access it. For example.

and displays a cross−hair which can be moved around with the mouse. • If a statement. except that the y −axis is represented as a log scale.= ------------------------------------------------------------------------------------------------------------------------------------------m m–1 m–2 Den ( x ) am x + am – 1 x + am – 2 x + … + a1 x + a0 where some of the terms in the numerator and/or denominator may be zero. is normally expressed in decibels (dB) and by definition. • The command linspace(first_value. and continue to enter data.y) command is similar to the plot(x. Normally. number_of_values) specifies the number of data points but not the increments between data points. Likewise. This format specifies the increments between points but not the number of data points. and the y −axis as a linear scale. and y is the vertical axis (ordinate). it can be continued to the next line by typing three or more periods. that is.y) command uses logarithmic scales for both axes. The command text(x. The loglog(x. as bn x + bn – 1 x + bn – 2 x + … + b1 x + b0 ) R ( x ) = Num ( x . or a row vector is too long to fit in one line. we can also use the zlabel(‘string’) command.y) command. xlabel(‘string’). • The ratio of any two values of the same quantity. and string is the label which we want to place at that location. This command plots y versus x where x is the horizontal axis (abscissa).y) to make two−dimensional plots.y. last_value. the logarithm to the base 2 is expressed as log2(x). and ylabel(‘string’). title(‘string’).y) command. • The semilogx(x. For a three−dimensional plot. except that the x −axis is represented as a log scale. • We use the MATLAB command plot(x. Likewise. P out dB = 10 log --------P in • The command gtext(‘string’) switches to the current Figure Window. Third Edition Copyright © Orchard Publications . whereas the common logarithm is expressed as log10(x). An alternate command uses the colon notation and has the format x=first: increment: last. n n–1 n–2 1−34 Numerical Analysis Using MATLAB® and Excel®. box. the semilogy(x.y) command is similar to the plot(x. and the x −axis as a linear scale. we express the numerator and denominator of a rational function as a combination of linear and quadratic factors. • We can make a two−dimensional plot more presentable with the commands grid.Chapter 1 Introduction to MATLAB • Rational Polynomials are those which can be expressed in ratio form. • The function log(x) in MATLAB is the natural logarithm.’string’) is similar to gtext(‘string’). then pressing <enter> to start a new line. it places a label on a plot in some specific location specified by x and y. typically power.

the vertices of the mesh lines are the triples { x ( j ).z) command displays a three−dimensional plot. Similarly. the division (/) and exponentiation (^) operators.y) function which creates the matrix X whose rows are copies of the vector x.Summary • MATLAB has no default color. We can generate the matrices X and Y with the [X./) and dot exponentiation (.n).y. and z .z. it starts with blue and cycles through seven colors. we use the special MATLAB function eps. Numerical Analysis Using MATLAB® and Excel®. y ) .2 × 10 – 16 . whereas element−by−element multiplication is performed with the dot multiplication operator (. We observe that x corresponds to the columns of Z .z) command plots a line in 3−space through the points whose coordinates are the elements of x . which is a number approximately equal to 2. Another command. and y corresponds to the rows of Z . n ] = size ( Z ) . • With MATLAB. y ( i ). and z are three vectors of the same length. no markers are drawn unless they are selected. The three − dimensional text(x.n. To produce a mesh plot of a function of two variables. say z = f ( x.*). y.y) or plot3(x. • The command axis([xmin xmax ymin ymax]) scales the current plot to the values specified by the arguments xmin. Also. • To avoid division by zero. we can set the limits of the x − and y −axes with the axis([xmin xmax ymin ymax]) command.’string’) command will place string beginning at the co−ordinate ( x. whereas dot division (. there is no default marker. in a three−dimensional plot we can set the limits of all three axes with the axis([xmin xmax ymin ymax zmin zmax]) command. • In a two−dimensional plot. There are no commas between these four arguments. • The mesh(x. where x. We can also use the mesh(x. Likewise.Y]=meshgrid(x. ymin and ymax. • The plot3(x.y. y .y. Z ( i.z) commands. • MATLAB can display up to four windows of different plots on the Figure window using the command subplot(m. This must be done for each plot. y. This command divides the window into an m × n matrix of plotting areas and chooses the pth area to be active.p). This command must be placed after the plot command and must be repeated for each plot. contour(Z. The default line is the solid line. draws contour lines for n levels.y. It must be placed after the plot(x. Third Edition Copyright © Orchard Publications 1−35 . j ) } .z) command with two vector arguments. xmax. we must first generate the X and Y matrices which consist of repeated rows and columns over the range of the variables x and y .y. matrix multiplication (multiplication of row by column vectors) is performed with the matrix multiplication operator (*). are used for matrix division and exponentiation.^) are used for element−by−element division and exponentiation. These must be defined as length ( x ) = n and length ( y ) = m where [ m. or on the same line without first executing the plot command. z ) on the plot. and the matrix Y whose columns are copies of the vector y. In this case.

the equal sign ( = ). The string fcn must be the name of an m−file function or a string with variable x . Thus. We use function files for repetitive tasks. It uses specifiers to indicate where and in which format the values would be displayed and printed. Third Edition Copyright © Orchard Publications . and if %e is used. • MATLAB displays the results on the screen in integer format without decimals if the result is an integer number. a script file is one which was generated and saved as an m−file with an editor such as the MATLAB’s Editor/ Debugger. the values will be displayed and printed in fixed decimal format.array) command used above displays and prints both text and arrays. Both types are referred to as m−files. The first line of a function file must contain the word function. and the input argument enclosed in parentheses. The function name and file name must be the same. • The MATLAB fplot(fcn.x2) function minimizes a function of one variable. • The MATLAB fprintf(format. where r is the magnitude. Generally. A function file is a user−defined function using MATLAB. but the file name must have the extension .Chapter 1 Introduction to MATLAB • The real(z) and imag(z) functions display the real and imaginary parts of the complex quantity z = x + iy. the values will be displayed and printed in scientific notation format. The polar(theta. and the abs(z). It attempts to return a value of x where f ( x ) is minimum in the interval x 1 < x < x 2 . A script file consists of two or more built−in functions. and the angle(z) functions compute the absolute value (magnitude) and phase angle of the complex quantity z = x + iy = r –θ . • MATLAB recognizes two types of files: script files and function files. The string f contains the name of the function to be minimized. and theta is the angle in radians. followed by the output argument. 1−36 Numerical Analysis Using MATLAB® and Excel®. Using lims = [xmin xmax ymin ymax] also controls the y −axis limits.lims) command plots the function specified by the string fcn between the x −axis limits specified by lims = [xmin xmax].x1. • The MATLAB fmin(f. or in short floating point format with four decimals if it a fractional number.r) function produces a plot in polar coordinates.m. MATLAB performs all computations with accuracy up to 16 decimal places. if %f is used. The format displayed has nothing to do with the accuracy in the computations. With these commands only the real part of each parameter is processed.

p ( y ) = y + 7y + 19y + 25y + 16y + 4 2.3132 5 4 3 2 3 2 b.7146 + j0. Use MATLAB to evaluate the polynomials below at the specified values.9 Plot V C (in dB scale) versus ω (in common log scale) and label the axes appropriately. the applied voltage V S was kept constant and the voltage V C across the capacitor was measured and recorded at several frequencies as shown on the table below.5 1200 102. a.25 b. p ( x ) = x + 8x + 10x + 4 b.12 Exercises 1.000 and three roots at x = – 3.000 3.3 1500 96.0 1300 100.5 1000 104.8 600 98.8 1600 94. R1 R2 VS C L VC 5 4 3 2 3 2 Capacitor voltage versus radian frequency ω VC ω VC 500 88. p ( x ) = x + 8x + 10x + 4 at x = 1.75 4. Third Edition Copyright © Orchard Publications 1−37 . Two roots at x = – 2. In the electric circuit below.Exercises 1. Use MATLAB to derive the polynomials having the following roots: a.5708 – 0.9 1100 103.3132 – 0. – 6.7146 – j 0. p ( y ) = y + 7y + 19y + 25y + 16y + 4 at y = – 3.9 1400 98. Use MATLAB to compute the roots of the following polynomials: a. Numerical Analysis Using MATLAB® and Excel®.7 800 104.4 700 103.9 900 105.

make an honest effort to find the solutions to the exercises without first looking at the solutions that follow.13 Solutions to End−of−Chapter Exercises Dear Reader: The remaining pages on this chapter contain the solutions to the exercises. Refer to the solutions as a last resort and rework those exercises at a later date. For the exercises that you are uncertain. for your benefit. 1−38 Numerical Analysis Using MATLAB® and Excel®. You must. review this chapter and try again. You should follow this practice with the rest of the exercises of this book.Chapter 1 Introduction to MATLAB 1. It is recommended that first you go through and work out those you feel that you know. Third Edition Copyright © Orchard Publications .

0000 2 8.3132j −0.3132i -0. 1.0000 -2. roots(Py) ans = -2. Pv=[1 8 10 4].7146+0.0. r2=[−2 −2 −3 −3 −3].Solutions to End−of−Chapter Exercises 1. −3.0000 . r1=[−6.7146−0.6904 Numerical Analysis Using MATLAB® and Excel®.7146 . a.0000 9.0000 -1.0000 p ( x ) = x + 8x + 10x + 4 b. a.0000 -1.9531 b.7146 + 0. Pw=[1 7 19 25 16 4].3132j]. poly_r1=poly(r1) poly_r1 = 3 1. value=polyval(Pw.5708 -0.9997 4.75) value = -63.5708 −0. roots(Px) b.0000i -1.25) value = 30.0000 + 0.3132i Py=[1 7 19 25 16 4]. Px=[1 8 10 4].0. value=polyval(Pv. poly_r2=poly(r2) poly_r2 = 1 5 13 4 67 3 171 2 216 108 p ( z ) = z + 13z + 67z + 171z + 216z + 108 3. ans = -6.0000i 2. a. Third Edition Copyright © Orchard Publications 1−39 .

. title('Magnitude of Vc vs.5 2 10 Magnitude of Vc vs.dB).. Vc=[88.Chapter 1 Introduction to MATLAB 4.9 105.9 96.9]. ylabel('|Vc| in volts') 40. grid.5 103 104.5 40 |Vc| in volts 39. xlabel('w in rads/sec').8 103.3 104. dB=20*log10(Vc). w 10 w in rads/sec 3 10 4 1−40 Numerical Analysis Using MATLAB® and Excel®.7 98..4 100.8 102. Third Edition Copyright © Orchard Publications . semilogx(w.5 94.5 39 38. w=[5 6 7 8 9 10 11 12 13 14 15 16]*100.9 98. w').

y y = f(x ) • • Tangent line (slope) to the curve y = f ( x ) at point { x 1. the slope (first derivative) at x = x 1 is y – f ( x1 ) f ' ( x 1 ) = -------------------x – x1 y – f ( x1 ) = f ' ( x1 ) ( x – x 1 ) (2.1) The slope crosses the x – axis at x = x 2 and y = 0 .2) and in general. Third Edition Copyright © Orchard Publications 2−1 .1). f ( x 1 ) } x ( x 2. 0 ) lies on the slope line. 2. Newton’s method for approximating real roots of a function We assume that the slope is neither zero nor infinite. This is an iterative (repetitive procedure) method and it is derived with the aid of Figure 2. Since this point [ x 2.1 Newton’s Method for Root Approximation Newton’s (or Newton−Raphson) method can be used to approximate the roots of any linear or non−linear equation of any degree.3) Numerical Analysis Using MATLAB® and Excel®. By substitution. 0 ) Figure 2. 0 – f ( x1 ) = f ' ( x1 ) ( x2 – x1 ) f ( x1 ) x 2 = x 1 – -------------f ' ( x1 ) (2. f ( xn ) x n + 1 = x n – -------------f ' ( xn ) (2. Several examples are presented to illustrate practical solutions using MATLAB and Excel spreadsheets.Chapter 2 Root Approximations T his chapter is an introduction to Newton’s and bisection methods for approximating roots of linear and non−linear equations.1. it satisfies (2.1. Then. f ( x 2 ) ] = ( x 2. f ( x 1 ) } { x 1.

X. grid % Specifies 100 values between -4 and 4 % The dot exponentiation is a must 2−2 Numerical Analysis Using MATLAB® and Excel®. See the Language Reference Guide or a C manual for complete details.\t. These specifications involve the character %. The following description was extracted from the help fprintf function. c. plot(x. We recall that the polyder(p) function displays the row vector whose values are the coefficients of the first derivative of the polynomial p.^ 2 − 5.x0) We’ve used the fprintf command in Chapter 1. The following two lines of script will display the graph of the given equation in the interval – 4 ≤ x ≤ 4 . backspace.x0)/polyval(q. optional width and precision fields. E. let us review it again. and writes it to the file associated with file identifier fid and contains C language conversion specifications. u.y). g. Knowing the polynomial p and the first approximation x 0 . x=linspace(−4. we must start with a reasonable approximation of the root value. and formfeed characters respectively. The special formats \n. this can best be done by plotting f ( x ) versus x with the familiar statements below. In all cases. We will now introduce some new MATLAB functions and review some which are discussed in Chapter 1. optional subtype specifier. This can be done easily with a simple plot using MATLAB or a spreadsheet. Therefore. tab. Solution: As a first step. and waits for an input from the user. 4.4) to find out where it crosses the x – axis . f.Chapter 2 Root Approximations Example 2. carriage return. The polyval(p. and conversion characters d. q=polyder(p) x1=x0−polyval(p. To apply Newton’s method. we can use the following script for the next approximation x 1 . e. i. under control of the specified format string. x. We must enclose f ( x ) = x2 – 5 (2. we can compute the next iteration for approximating a root with Newton’s method using these functions. We start with MATLAB and will discuss the steps for using a spreadsheet afterwards. y=x .\r.4) the text in single quotation marks. G.1 Use Newton’s method to approximate the positive root of to four decimal places. Therefore.\f can be used to produce linefeed. we plot the curve of (2. we will use it many more times. o. Use \\ to produce a backslash character and %% to produce the percent character. and s. optional flags. input(‘string’): It displays the text string.\b. Third Edition Copyright © Orchard Publications . It formats the data in the real part of matrix A (and in any additional matrix arguments). 100).x) function evaluates the polynomial p at some value x.

x0=input('Enter starting value: ').6). 15 10 5 0 -5 -4 -2 0 2 4 Figure 2. and we compute the next value x 1 as 2 f ( x0 ) ( –1 ) (2) – 5 x 1 = x 0 – -------------.= 2 – -----------------.25 f ' ( x0 ) 2(2) 4 (2. and Numerical Analysis Using MATLAB® and Excel®.0625 = 2. where the interval may not be so obvious.2361 4. so we take x 0 = 2 as our first approximation.5) The second approximation yields 2 f ( x1 ) --------------------------------------x 2 = x 1 – -------------.25 – ( 2. % Inserts a blank line % % The next function displays the value of x1 in decimal format as indicated % by the specifier %9. q=polyder(p). we expect this value to be a value between 2 and 3 .x0)/polyval(q. i.xxxxxx.= 2 – ---------.Newton’s Method for Root Approximation We chose this interval because the given equation asks for the square root of 5 .5 2 ( 2. and then redefine the interval. with 9 digits where 6 of these digits % are to the right of the decimal point such as xxx. The plot is shown in Figure 2. observe the x – axis crossings. we can choose a larger interval.. For other functions. Third Edition Copyright © Orchard Publications 2−3 .2.x0).= 2.e.5) and (2. the curve shows one crossing between x = 2 and x = 3 . fprintf('\n').= 2.25 ) – 5 = 2. % Approximation of a root of a polynomial function p(x) % Do not forget to enclose the coefficients in brackets [ ] p=input('Enter coefficients of p(x) in descending order: ').6) We will use the following MATLAB script to verify (2.25 – 0. Plot for the curve of Example 2.6f.1 As expected.2.25 ) f ' ( x1 ) (2. % Calculates the derivative of p(x) x1=x0−polyval(p.

6).* x) is a function file and must be saved as myfunction. The following lines show MATLAB’s inquiries and our responses (inputs) for the first two approximations. We use function files for repetitive tasks. Third Edition Copyright © Orchard Publications ... approximation \n'). the equal sign (=). We recall that a function file is a user− defined function using MATLAB. end where the commands . in the second line are executed as long as all elements in expression of the first line are true.. In Chapter 1 we discussed script files and function files..m We will use the while end loop. whose general form is while expression commands . % Inserts another blank line % fprintf('Rerun the program using this value as your next.. the function file consisting of the two lines below function y = myfunction(x) y=x .236111 We observe that this approximation is in close agreement with (2.. The function name and file name must be the same but the file name must have the extension . Enter coefficients of P(x) in descending order: [1 0 −5] Enter starting value: 2 The next approximation is: 2.250000 Rerun the program using this value as your next approximation Enter polynomial coefficients in descending order: [1 0 −5] Enter starting value: 2. We will also be using the following commands: 2−4 Numerical Analysis Using MATLAB® and Excel®.25 The next approximation is: 2.Chapter 2 Root Approximations % \n prints a blank line before printing x1 fprintf('The next approximation is: %9. and the input argument enclosed in parentheses..m. For example. x1) fprintf('\n').6f \n'.^ 3 + cos(3 . The first line of a function file must contain the word function followed by the output argument.

the text is displayed. x=linspace(−pi. The plot shows that one real root is approximately at x = – 1 . we choose New and click on M−File. This takes us to the Editor Window where we type the following three lines and we save it as funcnewt01. Next. Third Edition Copyright © Orchard Publications 2−5 .* cos(x). function y=funcnewt01(x) % Approximating roots with Newton's method y=x . pi.round(pi)) ans = 3 where the format script %d specifies an integer.* x + 3 + sin(x) − x .m. we generate the function funcnewt01 and we save it as an m−file.142 where the format script %4. Example 2.A): Formats the data in the real part of matrix A under control of the specified format string. Likewise. and disp(‘volts’) displays volts. if v = 12 .3f specifies a fixed format of 4 digits where 3 of these digits are allocated to the fractional part.^ 2 + 4 .3f'. sprintf('%d'. We generate the plot with the script below. For example.7) to four decimal places using Newton’s method.^ 2 + 4 . we sketch the curve to find out where the curve crosses the x – axis . Solution: As a first step. If x is a string.3. plot(x.* x + 3 + sin(x) − x .Newton’s Method for Root Approximation disp(x): Displays the array x without printing the array name. y=x . 100).2 Approximate one real root of the non−linear equation f ( x ) = x + 4x + 3 + sin x – x cos x 2 (2. To save it.* cos(x). disp(v) displays 12. so we will use this value as our first approximation. sprintf(format. For example. Numerical Analysis Using MATLAB® and Excel®. from the File menu of the command window. sprintf('%4.y). grid The plot is shown in Figure 2.pi) ans = 3.

fprimex = funcnewt02(x). we choose New and click on M−File. however. and to introduce the diff(s) function.2 We also need the first derivative of y. but correct % answer will be displayed if they are used. y1=diff(y) % Dot operations are not necessary with % symbolic expressions.3. and we save it as m−file. Our script for finding the next approximation with Newton’s method follows. This is y' = 2x + 4 + x sin x The computation of the derivative for this example was a simple task. just as a check. syms x y = x^2+4*x+3+sin(x)−x*cos(x). The syms function is used to define one or more symbolic expressions. function y=funcnewt02(x) % Finding roots by Newton's method % The following is the first derivative of the function defined as funcnewt02 y=2 . x = input('Enter starting value: '). fx = funcnewt01(x). we can let MATLAB do the differentiation. xnext = x−fx/fprimex. Third Edition Copyright © Orchard Publications .* x + 4 + x . To save it. This takes us to the Editor Window where we type these two lines and we save it as funcnewt02. we generate the function funcnewt02. x = xnext.* sin(x).Chapter 2 Root Approximations 30 20 10 0 -10 -4 -3 -2 -1 0 1 2 3 4 Figure 2. 2−6 Numerical Analysis Using MATLAB® and Excel®. % Find the derivative of y y1 = 2*x+4+x*sin(x) Now.m. Plot for the equation of Example 2. This function performs differentiation of symbolic expressions. from the File menu of the command window.

A cubic equation of the form y + py + qy + r = 0 3 2 (2.8) (2.9) can be reduced to the simpler form x + ax + b = 0 3 where * We will illustrate our examples with Excel.1=yes)'). x)) while input('Next approximation? (<enter>=no.1=yes) <enter> We can also use the fzero(f. Numerical Analysis Using MATLAB® and Excel®.895225 Next approximation? (<enter>=no. xnext=x−fx/fprimex. x)) end.x) function.1=yes)1 -0. fx=funcnewt01(x).894010 Next approximation? (<enter>=no. will be those of Excel. Enter starting value: −1 First approximation is: -0. The string f contains the name of a real−valued function of a single real variable. we will go through several examples to illustrate the procedure of using a spreadsheet such as Excel* to approximate the real roots of linear and non−linear equations. Henceforth. As we recall. This function tries to find a zero of a function of one variable.2 Approximations with Spreadsheets In this section. disp(sprintf('%9. and Quattro can also be used. disp(sprintf('Next approximation is x = %9.Approximations with Spreadsheets fx = funcnewt01(x). it is included here for convenience. all spreadsheet commands and formulas that we will be using. We recall that there is a standard procedure for finding the roots of a cubic equation. Third Edition Copyright © Orchard Publications 2−7 .6f \n'. x)) MATLAB produces the following result with – 1 as a starting value.6f \n'. MATLAB searches for a value near a point where the function f changes sign and returns that value. fprimex = funcnewt02(x). x=xnext. 2. It was introduced in Chapter 1. although others such as Lotus 1−2−3. or returns NaN if the search fails. disp(sprintf('First approximation is x = %9.6f \n'. fprimex=funcnewt02(x).

– ------------. Note: In our subsequent discussion we will omit the word cell and the key <enter>. q . and so on will be understood to be cell B3.+ ----.< 0 the roots will be real and unequal 4 27 a b If ---. Then. cell C11.+ ------------.– 3 2 2 A+B A–B x 3 = – ------------.12) (2.+ ----. Thus B3. Third Edition Copyright © Orchard Publications .10) For the solution it is convenient to let 3 A = a –b b ----. Also.+ ----4 27 2 2 3 a –b b B = – ----. This procedure is illustrated with the examples that follow. we can use a spreadsheet to define a range of x values with small increments and compute the corresponding values of y = f ( x ) .11) Then.Chapter 2 Root Approximations p x = y + -3 1 2 a = -.14) 2−8 Numerical Analysis Using MATLAB® and Excel®.+ ---.( 2p – 9pq + 27r ) 27 (2. after an entry has been made. the values of x for which the cubic equation of (2. it will be understood that the <enter> key was pressed. For real roots.( 3q – p ) 3 1 3 b = ----. we can plot y versus x to observe the values of x that make f ( x ) = 0 . and so on.– 3 2 2 (2.> 0 one root will be real and the other two complex conjugates 4 27 a b If ---. we should not use a spreadsheet to find the roots of a cubic equation by substitution in the above formulas. then b. Therefore. However.13) If the coefficients p . Example 2. spreadsheets do not. C11.+ ---.a If ---.3 Compute the roots of the polynomial using Excel.+ ----4 27 2 3 2 3 (2. unless we know that the roots are all real.= 0 there will be three real roots with at least two equal 4 27 2 3 2 3 2 3 While MATLAB handles complex numbers very well. y = f ( x ) = x – 7x + 16x – 12 3 2 (2.11) is equal to zero are x1 = A + B A+B A–B x 2 = – ------------. and r are real.+ ----. we can use a spreadsheet to find the real root since in any cubic equation there is at least one real root.

we select A2. We select a cell by moving the thick hollow white cross pointer to the desired cell and we click. a selected cell is surrounded by a heavy border. we select B1. We click on the Fixed Decimal check box to place a check mark and we choose 2 as the number of decimal places.000 . we click on Edit on the main taskbar. Next. that is.05 increments up to 5.00 in A2. there is a small black square. then we release the mouse button. Next. from the Tools drop menu of the Menu bar. we want to copy this formula to the range B3:B102 (the colon : means B3 through B102). We observe that. we can use the Copy and Paste icons of the taskbar. we drag it down to A102. the background on the monitor has changed from white to black.05. Alternately. We will use values in 0. We observe that after pressing the <enter> key.00 . We observe that this number is displayed exactly as it was typed. and to make all values look uniform. If it does not appear on the spreadsheet. We click on Edit. this becomes the next selected cell. We type 0.00 to 5. To continue. * To use this feature. we click on letter A on top of column A. that is.05 in A3. as we drag the fill handle. we choose Options and we click on the Edit tab. We observe that this value is displayed just as 0. and we click on Copy. In an Excel worksheet. a pop−up note shows the cell entry for the last value in the range. we drag the mouse down to A3 so that both these two cells are highlighted. This is the value of f ( x ) when x = 0. Next. For this example. We click. Next.00. and we release the mouse button. we can make it visible by performing the sequential steps Tools>Options. A2 will have a white background but A3 will have a black background. we type = A2^3-7*A2^2 + 16*A2-2 where A2 represents the first value of x = 0. We repeat these steps for Column B and we choose 3 decimal places. that is. Column A now contains 100 values of x from 0. without decimals. We observe that on the lower right corner of A3. With B2 still selected. then on Paste and we observe that this range is now filled with the values of f ( x ) .00 Next. We will now use the AutoFill* feature to fill−in the other values of x in Column A. we click and holding the mouse left button down. and we observe that this range is now highlighted.Approximations with Spreadsheets Solution: We start with a blank worksheet. We observe that B2 displays the value – 12. we first select A1 and we type x. we release the mouse button. and the numbers in Column B will be fixed with three decimal places. and place a check mark on the Drag and Drop setting. hold down the mouse button. In B2. We select the range B3:B102 with the mouse. We will enter more values in column A.00 in increments of 0. select the Edit tab. we type the equation formula with the = sign in front of it. Third Edition Copyright © Orchard Publications 2−9 . We observe that the entire column is now highlighted. and we type f(x). we point the mouse to the fill handle and we observe that the mouse pointer appears as a small cross. we type 0. When properly done. we highlight cells A2 and A3. Then. this is called the fill handle. the next cell moves downwards to A2. Numerical Analysis Using MATLAB® and Excel®. all numbers that we will type in Column A will be fixed numbers with two decimal places.

Then.422 0. We click on the Patterns tab to select it.0 Maximum: 5. on the Format axis menu. we select Number from the Category column. A chart similar to the one on Figure 2.45 -6.803 0.000 0.126 0.40 -6. and we delete it by pressing the Delete key.072 0. We click on OK to return to the graph. and on the Chart type column we click on XY (Scatter). and we click on the small box next to it (the box with the hand icon). From the Chart menu box we select Value (Y) axis and we click on the small box next to it (the 2−10 Numerical Analysis Using MATLAB® and Excel®. Next.0 Minor unit: 0.00 -12.3.05 -11. that is. Note: The Chart menu appears on the main taskbar and on the box below it.215 0. only when the graph box is selected.5 We click on the Number tab.625 0. and we click on Low on the Tick mark labels (lower right box). Third Edition Copyright © Orchard Publications . we click on Toolbars. This is done with the following steps: 1. and to see more precisely the x – axis crossing(s). the roots of f ( x ) . Then.25 -8. we select Value (X) axis. We will modify this plot to make it more presentable.35 -7.10 -10.Chapter 2 Root Approximations To plot f ( x ) versus x . The Chart menu appears in two places. we select any font.4 appears. that is.50 -5.151 f(x) 20 15 10 5 0 -5 -10 -15 0 1 2 3 4 5 6 Figure 2. We click on the Font tab.469 0. and we type 0 in the Decimal places box.217 0. Size 9.754 0. when it is enclosed in black square handles.30 -7. and we place a check mark on Chart.4. on the main taskbar and below it in a box where next to it is another small box with the hand icon. we click on the Scale tab and we make the following entries: Minimum: 0. we see it enclosed in six black square handles.0 Major unit: 1. From the Chart menu box (below the main taskbar). Then.55 -5. 3. We click on the Series 1 box to select it. We click anywhere inside the graph box. From the View menu. Plot of the equation of Example 2. x f(x) 0. From the displayed charts. we choose the one on top of the right side (the smooth curves without connection points). and Finish. 2.15 -9. we click on the Chart Wizard icon of the Standard Toolbar.20 -9.656 0. Regular style. we click on Next. Next.

select any font.0 Major unit: 0. we click at any point near it and we observe that several black square handles appear along the curve. Now. Then. it is highly recommended that this file is saved with any name.05 We click on the Number tab. and on the Chart Options. We click on OK to return to the graph. we select Plot Area and we observe that the gray background of the plot area is surrounded by black square handles. Finally. and we click on Outside on the Major tick mark type (upper right box). On the Format axis menu. To make the line of the curve f ( x ) thicker. We click on the Font tab.xls is the default extension for file names saved in Excel. We click on Gridlines. The plot area on the chart now appears on white background.0 Maximum: 1. The given polynomial has two roots at x = 2 . and on the Area side of the Patterns tab. We click on the Patterns tab. 7. and on the Patterns tab. We will follow the same procedure for generating the graphs of the other examples which follow. and the third root is x = 3 . and we select 2 in the Decimal places box.25 Minor unit: 0. We click on the box next to it (the box with the hand icon). The plot now resembles the one shown in Figure 2. we select Number from the Category column. Size 9. 4. we click on the Scale tab. Series 1 appears on the Chart menu box. From the Weight selections we select the first of the thick lines. We click on Chart on the main taskbar. Third Edition Copyright © Orchard Publications 2−11 . therefore. to change Chart Area square corners to round.5 where we have shown partial lists of x and f ( x ) . We click on the small box next to it.xls where. From the Chart menu box below the main task bar. we will change the background of the plot area from gray to white. we click on the white square which is immediately below the gray box. we click on the Titles tab and we make the following entries: Chart title: f(x) = the given equation (or whatever we wish) Value (X) axis: x (or whatever we wish) Value (Y) axis: y=f(x) (or whatever we wish) 5. and on the Patterns tab we place a check mark on the Round corners box. 6. Regular style.Approximations with Spreadsheets box with the hand icon). Numerical Analysis Using MATLAB® and Excel®. we place check marks on Major gridlines of both Value (X) axis and Value (Y) axis. and we make the following entries: Minimum: −1. we select Chart Area from the Chart menu. say poly01.

30 -7.00 -2.3.283 0.7x2 + 16x .80 -3. When this is done.95 -0.000 0.15 -9.xls. or any other name.516 0. Solution: To save lots of unnecessary work.803 0.05 -0.65 -4.009 3.xls.05 -11.25 -0.95 -0.40 -6.50 -5. we arbitrarily y = f ( x ) = 3x – 2x + 6x – 8 5 3 (2.00 0.25 -8.55 -5. Modified plot of the equation of Example 2.003 2.055 2.75 0.541 0.20 -9.00 0. Since we do not know where a real root is in the x−axis interval. Example 2.45 -6.656 0.000 f(x) = x3 .168 0. and save it with another name such as poly02.704 0. poly01.00 0 1 2 3 4 5 f(x) x x f(x) x f(x) 1. For this example.011 2.50 0. we perform the following steps: 1.000 3.90 -0.75 -1.5.625 0. that is.072 0. Third Edition Copyright © Orchard Publications .081 1.xls.002 3.217 0.887 0.00 0.00 -12.xls (or any other file name that was assigned to it).15) 2−12 Numerical Analysis Using MATLAB® and Excel®.843 0. the spreadsheet of the previous example still exists as poly01.85 -2. the highest power of the polynomial is 5 (odd number).Chapter 2 Root Approximations x f(x) 0.126 0.469 0.4 Find a real root of the polynomial using Excel.10 0. Next.215 0. we choose the Save as option.60 -4.00 -0.90 -0.045 Roots 2.50 -0.70 -3. we save it as poly02. and from the File drop down menu.xls.121 f(x) =0 at x=2 (double root) and at x=3 Figure 2.10 -10.422 0.25 0.90 -2. then.05 0.35 -7.10 -0.75 -3.151 0.000 2. and since we know that complex roots occur in conjugate pairs.260 1. we invoke (open) the spreadsheet of the previous example.754 0.12 1. This is done by first opening the file poly01. we expect that this polynomial will have at least one real root.95 -2.

4 This table shows that f ( x ) changes sign somewhere in the interval from x = 1 and x = 2 . and the plot shows that the curve crosses the x−axis somewhere between x = 1 and x = 2. From the Chart menu box (below the main taskbar) we select Value (X) axis. Our spreadsheet now shows that there is a sign change from B3 to B4.00 10. and so on.000 298052.00 2.00 1.6. we perform Steps 2 through 4 below.000 -1.05. and we observe the square handles surrounding it.000 -175751. Partial table for Example 2.01 Numerical Analysis Using MATLAB® and Excel®. Using the AutoFill feature.00 f(x) -298068. To obtain a good approximation of the real root in that interval.000 175735. We observe that the chart has changed shape to conform to the new data. The Chart menu on the main taskbar and the Chart menu box below it. i. On the View menu.000 Sign Change Figure 2. we enter −10 and −9 in A2 and A3 respectively. We do this by highlighting the range A23:B102.Approximations with Spreadsheets choose the interval – 10 ≤ x ≤ 10 . A2 contains 1.05 and x = 1. Let us then redefine our interval of the x values as 1 ≤ x ≤ 2 in increments of 0. We select the graph box by clicking inside it.6. we fill−in the range A4:A22.1 Major unit: 0. 2. are now displayed.000 -8. A part of the table is shown in Figure 2.0 Maximum: 1. we click on the Scale tab and make the following entries: Minimum: 1. Now we select B2 where we enter the formula for the given equation. Then. We must now delete all rows starting with 23 and downward. Next. and B (values of f ( x ) )..00 -9.e. to get better approximations.10 . reveal some useful information. and we have the interval from −10 to 10 in increments of 1. When this is done A1 contains 1.000 84. we click on Toolbars and place a check mark on Chart. and we click on the small box next to it (the box with the hand). x -10. Third Edition Copyright © Orchard Publications 2−13 . and we press the Delete key. and thus we expect that a real root exists between x = 1.05.00.00 0. Columns A (values of x).02 Minor unit: 0.00 9. Columns A and B now contain values of x and f ( x ) respectively. =3*A2^5−2*A2^3+6*A2−8 We copy this formula to B3:B22. on the Format axis menu.

on the Format axis menu.965 73.0 Major unit: 0.75 1. we suspect that the others are complex conjugates.692 24.631 11.00 0.892 3.447 63.Chapter 2 Root Approximations 3.20 f(x) = − 0.031 20.10 1.00 1.970 41.4 Since no other roots are indicated on the plot.545 8.06 f(x) -1.865 29.000 f(x) = 3x 5 . Then.90 1. We click on the Titles tab and make the following entries: Chart title: f(x) = the given equation (or whatever we wish) Value (X) axis: x (or whatever we wish) Value (Y) axis: y=f(x) (or whatever we wish) Our spreadsheet now should look like the one in Figure 2.35 1.95 2. x 1.832 17.50 1.60 1.55 1.20 1.892 3.50 f(x) 0. and we click on the small box next to it.00 1.1 4.05 1.85 1.10 1.186 0.209 Figure 2.000 -0.08 1.45 1.770 1.749 6.05 Real Root between 1.30 1. roots_p=roots(p) 2−14 Numerical Analysis Using MATLAB® and Excel®.5 Minor unit: 0.8 1.455 84. we click on the Scale tab and make the following entries: Minimum: −1.00 1.06 1.40 1.770 1.605 34.2x 3 + 6x .007 at x = 1.00 f(x) -1.70 1.0 Maximum: 1. Graph for Example 2.209 4.25 1.80 1.15 1.047 13.06.000 -0.04 1.15 1. From the Chart menu we select Value (Y) axis.02 1.00 1. We confirm this with MATLAB as follows: p = [ 3 0 −2 0 6 −8].00 -0.7 and we see that one real root is approximately 1.50 -1.823 55.10 x x 1.021 47. Third Edition Copyright © Orchard Publications .186 0.65 1.7.

8212i 0. we expect two more real roots.8 where the curve crosses the x – axis . such as poly03. Solution: We invoke (open) the spreadsheet of one of the last two examples..xls.16) and we copy this formula to B3:B62. We observe that Columns A and B contain the following sign changes (only a part of the table is shown): x 1. Approximate values of these roots can also be observed on the plot of Figure 2. Third Edition Copyright © Orchard Publications 2−15 . and save it with another name.30 interval.20 2. If we redefine the x – axis range as 1 to 2.0604 Example 2.xls.90 in A2 and A3 respectively.1415 0.xls or poly02.194 Sign Change Sign Change We observe two sign changes.20 ≤ x ≤ 1. that is. we fill−in the range A4:A72 and thus we have the interval from −1 to 6 in increments of 0.9476i 0.30 2.20 ≤ x ≤ 2. i.e.1415 -1. Numerical Analysis Using MATLAB® and Excel®.8212i 0.30 and x = 2.30 f(x) 0. poly01. Next. Using the AutoFill feature. this is found by substitution of zero into the given equation. Therefore. we arbitrarily choose the interval – 1 ≤ x ≤ 6 .Approximations with Spreadsheets roots_p = -1. =COS(2*A2)+SIN(2*A2)+A2−1 y = f ( x ) = cos 2x + sin 2x + x – 1 + + - 0. we will find that the other two roots are approximately x = 1.5. one in the 1. Then. we enter −1. There is a root at x = 0 . we select B2 and we enter the formula for the given equation.5 Compute the real roots of the trigonometric function using Excel. Since we do not know where real roots (if any) are in the x−axis interval.24 .041 -0.6113 0.9476i (2.059 0.10.30 interval and the other in the 2.20 1.138 -0.6113 1.00 and −0.

90 f(x) -3. we can use Goal Seek to set the dependent variable y to the desired value (goal) and from it.60 0.00 1. Third Edition Copyright © Orchard Publications .20 1.325 -3.70 -0. From the Tools drop menu.194 Figure 2.90 -0.30 f(x) 0.30 0.059 0.855 0. Graph for Example 2.138 (or 1.60 -0.00 0.041 -0.Chapter 2 Root Approximations x -1.50 0.690 0. To illustrate the Goal Seek feature.668 -0. These indicate that for 2−16 Numerical Analysis Using MATLAB® and Excel®.40 0. When this is done properly.5 We can obtain more accurate approximations using Excel’s Goal Seek feature. We copy range A24:B24 (or A25:B25) to two blank cells.515 -2.5.894 0. and when the Goal Seek dialog box appears. we will use it to find better approximations for the non−zero roots of Example 2.882 0.1 6 4 f(x) 2 0 -2 -4 -1 0 1 2 3 4 5 6 x Real Root at Real Root between Real Root between x 0. we click on Goal Seek.30 2.101 -2. Numbers tab.10 0. we will observe the changes in J1 and K1.510 0. In the last three examples our goal was to find the values of x for which y = f ( x ) = 0 .30 and −0.814 0.50 -0. but we do not know the input value which satisfies that result.829 -2. We increase the accuracy of Columns J and K to 5 decimal places by clicking on Format. We use Goal Seek when we know the desired result of a single formula.20 2. we make the following entries: Set cell: K1 To value: 0 By changing cell: J1 3.30 -0.10 0. find the value of the independent variable x which satisfies that goal.80 -0.20 0.40 -0.170 -1.20 and K1 contains 0.00 -0.20 -0.279 0.000 0.80 0.421 -1.319 0.647 f(x) = cos2x + sin2x + x .039 -0. Thus.801 -1.770 0.70 0. We do this with the following steps: 1.8.138 -0. say J1 and K1. if we have the function y = f ( x ) .000 0.041 if range A25:B25 was copied). so that J1 contains 1. 2. Cells.

9.000. For repetitive tasks.041).6 Construct a template (model spreadsheet). Third Edition Copyright © Orchard Publications 2−17 . use it to compute a root of the polynomial y = f ( x ) = x – 6x + 5x – 4x + 3x – 2x + x – 15 7 6 5 4 3 2 (2. We observe that other points are also displayed as the mouse is moved at different points near the curve. −0.00020 . such as finding the roots of polynomials. We will not discuss this method here. 2. Next. We repeat the above steps for the next root near x = 2. We observe now that A24 displays 1. in the By changing cell we enter A24 and we click on OK. * There exists a numerical procedure. with Excel. known as Bairstow’s method. and when the cross symbol appears.5. This is illustrated with the next example. we drag it towards the x−axis to change its value. y = f ( x ) = 0.30. it is prudent to construct a template (model spreadsheet) with the appropriate formulas and then enter the coefficients of the polynomial to find its real roots*. in the To value box we enter 0. Numerical Analysis Using MATLAB® and Excel®.27647 . We click anywhere near the curve. and we observe that five handles (black square boxes) are displayed along different points on the curve.22515 . (1.Approximations with Spreadsheets x = 1. 1.00002 . Solution: 1. We will illustrate the procedure with the chart of Example 2.17) given that one real root lies in the 4 ≤ x ≤ 6 interval. is with a chart such as those we’ve created for the last three examples. Another method of using the Goal Seek feature. 4. We begin with a blank spreadsheet and we make the entries shown in Figure 2. near the x = 1. that we can use to find the complex roots of a polynomial with real coefficients.20 . and we verify that for x = 2.30 point. Then. The Goal Seek dialog box then appears where the Set cell shows B24. A square box appears and displays Series 1. We point the mouse at the curve where it intersects the x−axis. it can be found in advanced numerical analysis textbooks. then. Example 2.30 point.28 and B24 displays 0. we click on the handle near the x = 1. which uses Newton’s method to approximate a real root of any polynomial with real coefficients up to the seventh power. y = f ( x ) = 0.

We observe that there is no change in the values of E16 and F16. We also enter the polynomial coefficients. we save it with a different name. say Example_2_6. Then. say template. 3. therefore. and the coefficients of the first derivative. Third Edition Copyright © Orchard Publications .xls.10. we take the average 5 and we enter it in A16. The formula in B16 of Figure 2. Model spreadsheet for finding real roots of polynomials. and in B16 we type the formula =A16-($A$7*A16^7+$B$7*A16^6+$C$7*A16^5+$D$7*A16^4 +$E$7*A16^3+$F$7*A16^2+$G$7*A16^1+$H$7)/ ($B$12*A16^6+$C$12*A16^5+$D$12*A16^4+$E$12*A16^3 +$F$12*A16^2+$G$12*A16^1+$H$12) The use of the dollar sign ($) is explained in Paragraph 4 below.Chapter 2 Root Approximations A 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 B C D E F G H Spreadsheet for finding approximations of the real roots of polynomials up the 7th power by Newton's Method. we terminate the approximation steps there. 2. and the coefficients of the first derivative in Rows 7 and 12 respectively.9. We observe that B16 now displays #DIV/0! (this is a warning that some value is being divided by zero). Next. Powers of x and corresponding coefficients of given polynomial p(x) Enter coefficients of p(x) in Row 7 x6 x5 x4 x7 x3 x2 x Constant Coefficients of the derivative p'(x) Enter coefficients of p'(x) in Row 12 x5 x4 x6 x3 x2 x Constant Approximations: xn+1 = xn − p(xn)/p'(xn) Initial (x0) 1st (x1) 2nd (x2) 3rd (x3) 4th (x4) 5th (x5) 6th (x6) 7th (x7) Figure 2. This value is our first (initial) approximation. but this will change once we enter the polynomial coefficients. is the familiar Newton’s formula which also appears in Row 14.9 with a name. 2−18 Numerical Analysis Using MATLAB® and Excel®.xls. Since we are told that one real root is between 4 and 6. we copy B16 to C16:F16 and the spreadsheet now appears as shown in the spreadsheet of Figure 2. We save the spreadsheet of Figure 2.10.

Spreadsheet for Example 2.16507 5.6. at the end. we press the function key F4. and so on. The contents of a relative cell changes when it is copied from one location to another. B16 is a relative cell. All cells in the formula of B16. Thus. changes to B16. except A16. and $B$16 is an absolute cell. when copied to the next column to the right. We can easily convert a relative cell to absolute or vice versa. Third Edition Copyright © Orchard Publications 2−19 . by first placing the cursor in front. Powers of x and corresponding coefficients of given polynomial p(x) Enter coefficients of p(x) in Row 7 x6 x5 x4 x7 1 -6 5 -4 x3 3 x2 -2 x 1 Constant -15 Coefficients of the derivative p'(x) Enter coefficients of p'(x) in Row 12 x5 x4 x6 7 -36 25 x3 -16 x2 9 x -4 Constant 1 Approximations: xn+1 = xn − p(xn)/p'(xn) Initial (x0) 1st (x1) 2nd (x2) 3rd (x3) 4th (x4) 5th (x5) 6th (x6) 7th (x7) 5. In this example. We can now use this template with any other polynomial by just entering the coefficients of the new polynomial in row 7 and the coefficients of its derivative in Row 12.The Bisection Method for Root Approximation A 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 B C D E F G H Spreadsheet for finding approximations of the real roots of polynomials up the 7th power by Newton's Method. we observe the successive approximations in Row 16. or between the letters and numbers of the cell. have dollar signs ($) in front of the column letter. changes to C16. then. absolute so that the formula of B16 can be copied to C16.163190 5. without changing its value.163194 5. 2. The relative cell A16. The value of an absolute cell does not change when it is copied from one position to another. A cell that is not absolute is said to be relative cell.163190 Figure 2. * This is a step−by−step problem−solving procedure. recursive computational procedure for solving a problem in a finite number of steps.0 5. 4.20409 5. D16 and so on. and in front of the row number. especially an established.10. when copied to the next column. Numerical Analysis Using MATLAB® and Excel®. except A16.3 The Bisection Method for Root Approximation The Bisection (or interval halving) method is an algorithm* for locating the real roots of a function. These cells are said to be absolute. we made all cells. then.

the above process is repeated until the root we are seeking has a specified tolerance. Sketches to illustrate the bisection method when f ( x 1 ) and f ( x m ) have same sign 2. Sketches to illustrate the bisection method when f ( x 1 ) and f ( x m ) have opposite signs After making the appropriate substitution. f ( x 1 ) and f ( x m ) have opposite signs and thus their product is negative f ( x 1 ) and f ( x m ) have opposite signs and thus their product is negative • • x•1 x m x 2 x1 • • • xm x2 Figure 2. To terminate the iterations. or f ( x 1 ) < 0 and f ( x 2 ) > 0 . If any of these two conditions is satisfied. either f ( x 1 ) > 0 and f ( x 2 ) < 0 . f ( x m ) ⋅ f ( x 1 ) < 0 . In this case.12. the following decisions are made: 1. we replace x 2 with x m . specify a tolerance on the error of f ( x ) 2−20 Numerical Analysis Using MATLAB® and Excel®. that is.11. we replace x 1 with x m . f ( x 1 ) and f ( x m ) are both positive and thus their product is positive • x•1 x•m x 2 f ( x 1 ) and f ( x m ) are both negative and thus their product is positive • • x•1 x m x 2 Figure 2. This indicates that x m and x 1 are on the left side of the x−axis crossing as shown in Figure 2. that is. In this case. Then. their product will be negative. that is.18) Knowing x m . If f ( x m ) and f ( x 1 ) have the same sign. so that f ( x 1 ) and f ( x 2 ) have opposite signs. we can compute the midpoint xm of the interval x 1 ≤ x ≤ x 2 with x1 + x2 x m = ---------------2 (2.12. their product will be positive. we either: a. This indicates that x m and x 2 are on the right side of the x−axis crossing as in Figure 2.11. we can find f ( x m ) . Third Edition Copyright © Orchard Publications . say x 1 and x 2 . If f ( x m ) and f ( x 1 ) have opposite signs. specify a number of iterations b. f ( x m ) ⋅ f ( x 1 ) > 0 .Chapter 2 Root Approximations The objective is to find two values of x.

evaluate all commands down to the end command. the loop ends and any commands after the end are computed and displayed as commanded. 6 of which will be to the right of the decimal point. a. function y= funcbisect01(x).* x . and using a for end loop MATLAB program b.* x − 8. k = 16 .m.. We will discuss other special formats as they appear in subsequent examples. by specifying 16 iterations. y = 3 . with a total of 9 digits. that is.fm) line. % We must not forget to type the semicolon at the end of the line above. 6 of which will be to the right of the decimal point. %9. and %d for integer format. The special format \n specifies a linefeed. the statement for k = 1:16 says for k = 1.. we can type help fprintf. The for end loop allows a group of functions to be repeated a fixed and predetermined number of times. and using a while end loop MATLAB program Solution: This is the same polynomial as in Example 2.6f and %13. % otherwise our script will fill the screen with values of y On the script below.4.19) a. …. end Before we write the program script.The Bisection Method for Root Approximation We will illustrate the Bisection Method with examples using both MATLAB and Excel. After the k = 16 iteration.^ 5 − 2 . it prints everything specified up to that point and starts a new line. For more information.6f \n'. xm. by specifying 0.7 Use the Bisection Method with MATLAB to approximate one of the roots of by y = f ( x ) = 3x – 2x + 6x – 8 5 3 (2. the first specifies that the value of xm must be expressed in decimal format also called fixed point format.* x . Here.^ 3 + 6 .00001 tolerance for f ( x ) . k = 2. Likewise. %s for string format. we must define a function assigned to the given polynomial and save it as a function m−file.6f are referred to as format specifiers or format scripts. Some other specifiers are %e for scientific format.6f %13. We will define this function as funcbisect01 and will save it as funcbisect01. Let us also review the meaning of the fprintf('%9. fm must be expressed in decimal format with a total of 13 digits. Example 2. Numerical Analysis Using MATLAB® and Excel®. Third Edition Copyright © Orchard Publications 2−21 . The syntax is: for x = array commands.

xm. The while end loop evaluates a group of commands an indefinite number of times.060364 -0.046875 -0.000380 1. x1=1.031250 -0. Third Edition Copyright © Orchard Publications .Chapter 2 Root Approximations The script for the first part of Example 2. fm is f(xm) disp('-------------------------') % insert line under xm and fm for k=1:16.062500 0.001844 1.749023 1.. The syntax is: while expression commands.032885 1.060303 -0. % We know this interval from Example 2.103195 1.002604 1.058594 -0.000102 We observe that the values are displayed with 6 decimal places as we specified.506944 1.6f %13.054688 -0. Figure 2.031250 1.308441 1.060410 0..006289 1. fm=funcbisect01(xm).060059 -0. but for the integer part unnecessary leading zeros are not displayed.059570 -0.000176 1. x2=2.060394 -0.500000 17. else x1=xm.038318 1. b.060425 0.fm) % Prints xm and fm on same line.241184 1.6f \n'. f2=funcbisect01(x2). xm=(x1+x2) / 2.060547 0.7 is given below. end end When this program is executed. f1=funcbisect01(x1). fprintf('%9.6 disp(' xm fm') % xm is the average of x1 and x2. 2−22 Numerical Analysis Using MATLAB® and Excel®. MATLAB displays the following: xm fm ------------------------1.4.125000 1.000732 1. if (f1*fm<0) x2=xm.015168 1.250000 4.

MATLAB displays the following: xm fm ------------------------1. end end When this program is executed. There is no need to create another function m−file. fprintf('%9. disp('-------------------------') while (abs(x1-x2)>2*tol). The script should be written so that eventually a false condition is reached and the loop then terminates.060547 0.062500 0.002604 1. Now we type and execute the following while end loop program.001844 1.The Bisection Method for Root Approximation end The commands between while and end are executed as long as all elements in expression are true.032885 1.054688 -0.000032 1.00001.000102 1. tol=0.308441 1. if (f1*fm<0). f2=funcbisect01(x2).031250 -0.000176 1.6f \n'.000037 1. we will use the same as in part a.6f %13.125000 1. fm=funcbisect01(xm).250000 4.000380 1.006289 1. f1=funcbisect01(x1).060059 -0.059570 -0.058594 -0.060404 -0.060425 0. x2=2.046875 -0.038318 1.060394 -0. xm. x2=xm.103195 1.060402 -0.060406 0.500000 17.241184 1. xm=(x1+x2)/2.506944 1.060410 0.749023 1. disp(' xm fm').000732 1.000003 Numerical Analysis Using MATLAB® and Excel®.060364 -0.015168 1.060303 -0.031250 1.fm). Third Edition Copyright © Orchard Publications 2−23 . else x1=xm. x1=1.

value_if_false is the value that is returned if logical_test is false . otherwise assign the number 1200. 800. A13: 2 B13: 3 C13: =(A13+B13)/2 2 5 within 2−24 Numerical Analysis Using MATLAB® and Excel®.6 so we can use it with any polynomial equation. =IF(D22<E22. =IF(M8<>N17. If logical_test is false and value_if_false is omitted.5 . We start with a blank spreadsheet and we make the entries in rows 1 through 12 as shown in Figure 2. Third Edition Copyright © Orchard Publications . Value_if_true can be another formula. Value_if_false can be another formula. Example 2. K7*12. The average of these values is x m = 2. assign the number 800. use the value in K7 multiplied by 12. 1200):If the value in D22 is less than the value of E22. If logical_test is true and value_if_true is omitted. true is returned. We will create a template as we did in Example 2.13. use the value in A15. Solution: Finding the square root of 5 is equivalent to finding the roots of x – 5 = 0 . B15):If the value in C11 is greater than or equal to 1500. we make the following entries in rows 13 and 14. This requires repeated use of the IF function which has the following syntax.8 Use the bisection method with an Excel spreadsheet to approximate the value of 0. We expect the positive root to be in the 2 < x < 3 interval so we assign x 1 = 2 and x 2 = 3 .00001 accuracy. otherwise use the value in L8 divided by 24. These statements may be clarified with the following examples.Chapter 2 Root Approximations Next.value_if_true. L8/24):If the value in M8 is not equal to the value in N17. =IF(C11>=1500.A15. otherwise use the value in B15.value_if_false) where logical_test: any value or expression that can be evaluated to true or false. =IF(logical_test. value_if_true: the value that is returned if logical_test is true. false is returned. Now. we will use an Excel spreadsheet to construct a template that approximates a real root of a function with the bisection method.

we highlight A14:F14.00000 1.00000 0. B13) We copy C13 into C14 and we verify that C14: =(A14+B14)/2 Next. Partial spreadsheet for Example 2. Then. Finally. We place the cursor on D14 and from the Edit menu we click on Paste. C13. we place the cursor on A15. from the Edit menu. from the Edit menu we click on Copy.14.00000 0. The square root of 5 accurate to six decimal places is shown on C30 in the spreadsheet of Figure 2.00000 0. C13. we highlight D13:F13 and on the Edit menu we click on Copy.13.8 D13: =$A$9*A13^7+$B$9*A13^6+$C$9*A13^5+$D$9*A13^4 +$E$9*A13^3+$F$9*A13^2+$G$9*A13^1+$H$9*A13^0 E13: =$A$9*C13^7+$B$9*C13^6+$C$9*C13^5+$D$9*C13^4 +$E$9*C13^3+$F$9*C13^2+$G$9*C13^1+$H$9*C13^0 F13: =D13*E13 A14: =IF(A14=A13.The Bisection Method for Root Approximation A 1 2 3 4 5 6 7 8 9 10 11 12 B C D E F G H Spreadsheet for finding approximations of the real roots of polynomials using the Bisection method Equation: y = f(x) = x − 5 = 0 Powers of x and corresponding coefficients of given polynomial f(x) Enter coefficients of f(x) in Row 9 2 x 7 x 6 x 5 x 4 x 3 x 2 x 0 Constant -5 0. we click on Paste and we observe the values in A15:F30. B13) B14: =IF(A14=A13. and holding the mouse left button. we highlight the range A15:A30.00000 0. We verify that the numbers on D14:F14 are as shown on the spreadsheet of Figure 2.00000 x1 x2 xm (x1+x2)/2 f(x1) f(xm) f(x1)f(xm) Figure 2. Third Edition Copyright © Orchard Publications 2−25 . Numerical Analysis Using MATLAB® and Excel®.14.

00000 0.23535 2.23607 2.00320 -0.25000 2.21875 2.00000 0. Entire spreadsheet for Example 2.23602 2.00102 -0.12500 2.18750 2.23438 2.00000 -1.00000 0.00003 f(xm) 1.00001 f(x1)f(xm) -1.23608 2.02740 0.48438 -0.00000 0.23438 2.00006 -0.00058 -0.00000 0.24219 2.21875 2.23584 2.00047 -0.00007 -0.00000 2.23633 2.00002 0.48438 0.21484 -0.23633 2.00000 0.07715 -0.00020 -0.00000 0.00006 0.23828 2.50000 2.00757 0.23608 2.06250 0.00047 -0.23606 f(x1) -1.00116 -0.25000 -0.00990 0.23605 2.23608 2.00000 0.00000 1.00757 -0.23633 2.18750 2.23606 2.00000 2.23606 x2 3.23535 2.23438 2.00007 -0.23438 2.23605 2.00000 0.10406 0.48438 -0.00320 -0.8 2−26 Numerical Analysis Using MATLAB® and Excel®.Chapter 2 Root Approximations A 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 B C D E F G H Spreadsheet for finding approximations of the real roots of polynomials using the Bisection method 2 Equation: y = f(x) = x − 5 = 0 Powers of x and corresponding coefficients of given polynomial f(x) Enter coefficients of f(x) in Row 9 x 7 x 6 x 5 x 4 x 3 x 2 x 0 Constant -5 0.23608 2.21484 -0.00000 -1.07715 -0.01657 0.00000 0.25000 2.00102 -0.14.00001 0.25000 0.00757 -0.23596 2.23828 2.00003 -0.00000 x1 2.23633 2.00006 -0.00000 2.24219 2.50000 2.00757 -0.00000 -0.00020 -0.00021 -0.00757 -0.00000 Figure 2.23607 xm (x1+x2)/2 2.23607 2.00000 2.25000 2.12500 2.25000 2.25000 2.23584 2.23602 2. Third Edition Copyright © Orchard Publications .23438 2.00102 0.23605 2.00000 0.23584 2.23608 2.23596 2.06250 -0.25000 2.00000 -0.

find the value of the independent variable x which satisfies that goal. • We can use an Excel spreadsheet to construct a template that approximates a real root of a function with the bisection method. this can best be done by plotting f ( x ) versus x . that is. but we do not know the input value which satisfies that result. such as finding the roots of polynomials. we can compute the midpoint xm of the interval x 1 ≤ x ≤ x 2 with x1 + x2 x m = ---------------2 • We can use the Bisection Method with MATLAB to approximate one of the roots by specifying a number of iterations using a for end or by specifying a tolerance using a while end loop program. or f ( x 1 ) < 0 and f ( x 2 ) > 0 . We use Goal Seek when we know the desired result of a single formula. Thus.Summary 2. The objective is to find two values of x. • For approximating real roots we can use Excel’s Goal Seek feature. It uses the formula f ( xn ) x n + 1 = x n – -------------f ' ( xn ) To apply Newton’s method. we must begin with a reasonable approximation of the root value. • The Bisection (or interval halving) method is an algorithm for locating the real roots of a function.value_if_false) Numerical Analysis Using MATLAB® and Excel®. if we have the function y = f ( x ) . so that f ( x 1 ) and f ( x 2 ) have opposite signs. we can use Goal Seek to set the dependent variable y to the desired value (goal) and from it. • We can use a spreadsheet to approximate the real roots of linear and non−linear equations but to approximate all roots (real and complex conjugates) it is advisable to use MATLAB. In all cases.4 Summary • Newton’s (or Newton−Raphson) method can be used to approximate the roots of any linear or non−linear equation of any degree. say x 1 and x 2 . it is prudent to construct a template (model spreadsheet) with the appropriate formulas and then enter the coefficients of the polynomial to find its real roots.value_if_true. • The MATLAB the while end loop evaluates a group of statements an indefinite number of times and thus can be effectively used for root approximation. Third Edition Copyright © Orchard Publications 2−27 . This requires repeated use of the IF function which has the =IF(logical_test. If any of these two conditions is satisfied. either f ( x 1 ) > 0 and f ( x 2 ) < 0 . • For repetitive tasks.

use Newton’s method to estimate the root of f ( x ) = 0 that lies between a and b . f 2 ( x ) = 4 a = 1 b = 2 a = 2 b = 4 2x + 1 – x + 4 Hint: Start with x 0 = ( a + b ) ⁄ 2 2. a.5 using MATLAB.Chapter 2 Root Approximations 2. Repeat Example 2. Use MATLAB to sketch the graph y = f ( x ) for each of the following functions. and verify from the graph that f ( a ) and f ( b ) . Then. f 1 ( x ) = x + x – 3 b. Third Edition Copyright © Orchard Publications . where a and b defined below. Hint: Use the procedure of Example 2. 3. have opposite signs.5 Exercises 1. Repeat Exercise 1 above using the Bisection method.2 2−28 Numerical Analysis Using MATLAB® and Excel®.

= 1.Solutions to End-of-Chapter Exercises 2.1 – -------------------.5 2 From the plot above we see that the positive root lies between x = 1 and x = 1.= 1.5 -1 -0. Third Edition Copyright © Orchard Publications 2−29 .3241i 0. x=−2:0. f1x=x. roots(pa) ans = -1.5 0 0.6 Solutions to End-of-Chapter Exercises 1.= 1.1 – -----------------------------------.324 4 ( 1.= 1.164 3 7.1 – 3 ( – 0.1443 + 1.25 so we choose a = 1 and b = 1. We compute the next value x 1 as 4 f ( x0 ) ( 1.= 1.05:2.169 3 f ' ( x0 ) 6.25 so we take x 0 = 1.0365 ( 1.1 ) + 1. plot(x.169 – -----------------------------------------------. grid 16 14 12 10 8 6 4 2 0 -2 -4 -2 -1.1 as our first approximation.1640 Numerical Analysis Using MATLAB® and Excel®.f1x).^4+x−3. a.1 ) + 1 The second approximation yields 4 f( x1 ) 0.1443 .3241i 1.5 1 1.169 – 3 x 2 = x 1 – -------------.39 f ' ( x1 ) 4 ( 1.169 ) + 1 Check with MATLAB: pa=[1 0 0 1 −3].436 ) x 1 = x 0 – -------------.= 1.4526 0.169 – --------------.169 ) + 1.1.

solve(f2x) 7 – 7 = 0.⋅ ( 2x + 1 ) – 1 ⁄ 2 ⋅ 2 – -.⋅ f 2 ( x ) = -. ans = 3 2−30 Numerical Analysis Using MATLAB® and Excel®. 0. To compute the next value x 1 we first need to find the first derivative of f 2 ( x ) . We rewrite it as f2 ( x ) = 2x + 1 – x + 4 = ( 2x + 1 ) 1⁄2 1⁄2 – (x + 4) Then. 1 1 d 1 1 ----.5 0 -0. We also observe that since f ( x 0 ) = there was no need to find the first derivative f ' ( x0 ) . f2x=sqrt(2.= 3 – ----------------------------------------------. Third Edition Copyright © Orchard Publications . the real root is exactly x = 3 .*x+1)−sqrt(x+4).Chapter 2 Root Approximations b.= 3 f ' ( x0 ) 1 ⁄ 7 – 1 ⁄ (2 7) 1 ⁄ (2 7) and Thus. plot(x. x=−5:0. Check with MATLAB: syms x.5 -1 -1.f2x).⋅ ( x + 4 ) – 1 ⁄ 2 ⋅ 1 = ------------------. grid Warning: Imaginary parts of complex X and/or Y arguments ignored. f2x=sqrt(2.*x+1)−sqrt(x+4).= 3 – --------------------.05:5.5 -2 -5 -4 -3 -2 -1 0 1 2 3 4 5 From the plot above we see that the positive root is very close to x = 3 and so we take x 0 = 3 as our first approximation.– ------------------2 2 dx 2x + 1 2 x + 4 f( x0 ) 2×3+1– 3+4 0 x 1 = x 0 – -------------.

6f \n'.250000 0.014045 1.125000 -0.^ 4 +x − 3. After saving this file as exercise2.163574 -0.162109 -0.m function y= exercise2(x). We will use the while end loop MATLAB program and specify a tolerance of 0.163086 -0.Solutions to End-of-Chapter Exercises 2. else x1=xm. fprintf('%9. end end MATLAB displays the following: xm fm ------------------------1.176041 1. x2=2.003367 1. xm=(x1+x2) / 2.562500 1. Numerical Analysis Using MATLAB® and Excel®. fm=exercise2(xm).164063 0. f1=exercise2(x1).057803 1. a. We will define this function as exercise2 and will save it as exercise2. fm is f(xm) disp('-------------------------') % insert line under xm and fm for k=1:12.500000 3.156250 -0. we execute the following program: x1=1.00001.fm)% Prints xm and fm on same line.056411 1.160156 -0.000200 1.6f %13. if (f1*fm<0) x2=xm. We need to redefine the function m−file because the function in part (b) is not the same as in part a.273193 1.m.001584 b.171875 0.187500 0.028229 1. Before we write the program script.163818 -0. xm.006930 1. we must define a function assigned to the given polynomial and save it as a function m − file. y = x . Third Edition Copyright © Orchard Publications 2−31 . We will use the for end loop MATLAB program and specify 12 iterations. f2=exercise2(x2). % x1=a and x2=b disp(' xm fm') % xm is the average of x1 and x2.691406 1.

000061 0. tol=0. xm.002344 0.028125 0. we execute the following program: x1=2.1. After saving this file as exercise2.000369 3. % If we specify x1=a=2 and x2=b=4. x2=4.000011 0. else x1=xm.998047 -0.001182 3.011733 2.*x+1)−sqrt(x+4).000002 2−32 Numerical Analysis Using MATLAB® and Excel®.999994 -0. y = sqrt(2.999658 -0. if (f1*fm<0).000005 3.993750 -0.650000 -0.000037 2.010938 0.000195 0.m. disp('-------------------------') while (abs(x1-x2)>2*tol).925000 -0.000014 3.014289 3.200000 0.3.999121 -0.002065 3. the program % will not display any values because xm=(x1+x2)/2 = 3 = answer disp(' xm fm'). end end When this program is executed.6f \n'.062500 0. fm=exercise2(xm). Third Edition Copyright © Orchard Publications .000012 2.00001.999927 -0. xm=(x1+x2)/2.000001 3.068779 2. f1=exercise2(x1).fm).005299 3.000443 2. x2=xm.000166 2.6f %13.Chapter 2 Root Approximations function y= exercise2(x).037013 2. fprintf('%9. MATLAB displays the following: xm fm ------------------------3.000027 0.000065 2. f2=exercise2(x2).

*x)+sin(2. syms x. y = f ( x ) = cos 2x + sin 2x + x – 1 We use the following script to plot this function. This is because when solving equations of periodic functions.Solutions to End-of-Chapter Exercises 3.*x)+x−1. To find a good approximation of the second root that lies between x = 2 and x = 3 . The second value (2) is not exactly correct as we can see from the plot. plot(x. From Example 2. Third Edition Copyright © Orchard Publications 2−33 . function y=exercise3(x) % Finding roots by Newton's method using MATLAB y=cos(2. y=cos(2*x)+sin(2*x)+x−1.*x)+sin(2. there are an infinite number of solutions and MATLAB restricts its search for solutions to a limited range near zero and returns a non−unique subset of solutions. solve(y) ans = [0] [2] The first value (0) is correct as it can be seen from the plot above and also verified by substitution of x = 0 into the given function. grid 6 4 2 0 -2 -4 -6 -8 -5 -4 -3 -2 -1 0 1 2 3 4 5 Let us find out what a symbolic solution gives. y=cos(2.y).*x)+x−1.5. we write and save the function files exercise3 and exercise3der as defined below. x=−5:0.02:5. function y=exercise3der(x) Numerical Analysis Using MATLAB® and Excel®.

x = xnext.6f \n'. fx=exercise3(x).2295 and this is consistent with the value shown on the plot. fprimex=exercise3der(x). we write and execute the following program and we find that the second root is x = 2. disp(sprintf('Next approximation is x = %9. x)) while input('Next approximation? (<enter>=no. xnext = x−fx/fprimex. xnext=x−fx/fprimex.*x)+1.1=yes)'). x)) Enter starting value: 3 First approximation is x = 2.*sin(2. fx = exercise3(x). disp(sprintf('%9. fprimex = exercise3der(x). x = input('Enter starting value: ').*x)+2. fx = exercise3(x).Chapter 2 Root Approximations % Finding roots by Newton's method % The following is the first derivative of % the function defined as exercise3 y=−2. Third Edition Copyright © Orchard Publications . x=xnext. Now.6f \n'. disp(sprintf('First approximation is x = %9.*cos(2.229485 2−34 Numerical Analysis Using MATLAB® and Excel®. fprimex = exercise3der(x).6f \n'. x)) end.

Phasors which are rotating vectors in terms of complex numbers. Voltage or Current Voltage or Current 1 2 0 . Numerical Analysis Using MATLAB® and Excel®.5 -1 -1 0 2 T 4 6 8 1 0 1 2 -2 0 2 T 4 6 8 1 0 1 2 Voltage or Current Time Voltage or Current Time T T Figure 3. Voltage and current relationships are expressed in sinusoidal terms.1 Alternating Voltages and Currents The waveforms shown in Figure 3.Chapter 3 Sinusoids and Phasors T his chapter is an introduction to alternating current waveforms.1 may represent alternating currents or voltages. Examples of alternating voltages and currents Thus an alternating current (AC) is defined as a periodic current whose average value over a period is zero.1. Third Edition Copyright © Orchard Publications 3−1 . are also introduced and their relationships to sinusoids are derived. 3.5 1 0 Time 0 Time -0 . The characteristics of sinusoids are discussed and the frequency. Also. Stated differently. phase angle. and period are defined. the average of the positive and negative values over a period is zero. an alternating current alternates between positive and negative values at regularly recurring intervals of time.

then ωT = 2π or ----T = 2π ω (3. and let f ( t ) = A sin t where A is the amplitude of this function. which is a circle with radius of one unit. Two main reasons for studying sinusoids are: (1) many physical phenomena such as electric machinery produce (nearly) sinusoidal voltages and currents and (2) by Fourier analysis. We observe that when the phasor has completed a cycle (one revolution). is used to express the number of cycles per second. the cos ωt and sin ω t functions are generated from the projections of the phasor on the horizontal and vertical axis respectively.2 Characteristics of Sinusoids Consider the sine waveform shown in Figure 3. Thus. if it takes one second to complete one cycle (one revolution around the unit circle). that is. we say 3−2 Numerical Analysis Using MATLAB® and Excel®. Generation of a sinusoid by rotation of a phasor At the completion of one cycle.1) The term frequency in Hertz. commonly known as angular or radian frequency.2. denoted as Hz . it has traveled 2π radians or 360° degrees. our subsequent discussion will be restricted to sine or cosine waveforms and these are referred to as sinusoids. 3. can be represented by a sum of sinusoids. or any other unit. and then repeats itself to form another cycle. Third Edition Copyright © Orchard Publications . if we let the phasor (rotating vector) travel around the unit circle with an angular velocity ω .Chapter 3 Sinusoids and Phasors As shown in Figure 3. A = 1 as shown.2.1. f (t) Voltage or Current Sine Wavef orm π ⁄ 2 ( 90° ) Phasor Direction of rotation 0 ( 0° ) 2π ( 360° ) A π ( 180° ) ω A = 1 0 π π /2 3π /2 2π −A 3π ⁄ 2 ( 270° ) Time Figure 3. where f ( t ) may represent either a voltage or a current function. Thus. such as the square and sawtooth waveforms on the previous page. any periodic waveform which is not a sinusoid. t = T (one period). Unless otherwise stated. the period T of an alternating current or voltage is the smallest value of time which separates recurring values of the alternating waveform. and since ω is the angular velocity. A sinusoid (sine or cosine function) can be constructed graphically from the unit circle.

4 shows three sinusoids which are outof-phase. that two sinusoids are out-of-phase by a phase angle other than 90° .5 -1 – V max 0 1 2 3 T 4 5 6 7 Figure 3. Since the cosine and sine functions are usually known in terms of degrees or radians. V max1 0. Thus.5 cos ω t sin ωt π⁄4 π⁄2 π 3π ⁄ 2 2π ωt 0 -0. the two sinusoids are said to be in-phase. we see that the cosine function will be the same as the sine function if the latter is shifted to the left by π ⁄ 2 radians. we say that the sine function lags (is behind) the cosine function by π ⁄ 2 radians or 90° . Numerical Analysis Using MATLAB® and Excel®.2) The frequency f is often referred to as the cyclic frequency to distinguish it from the radian frequency ω . It is possible. two sinusoids of different frequencies can never be in phase. Another common expression is that the cosine and sine functions are out-of-phase by 90° . we obtain the sine waveform. We must remember that when we say that one sinusoid leads or lags another sinusoid.3. Figure 3. Likewise. of course.3. and i ( t ) = I max sin ω t are plotted as shown in Figure 3. it is convenient to plot sinusoids versus ωt (radians) rather that time t . Plot of the cosine and sine functions By comparing the sinusoidal waveforms of Figure 3.3. Third Edition Copyright © Orchard Publications 3−3 .Characteristics of Sinusoids that the frequency is 1 Hz or one cycle per second.1) we have 1 f = --T or ω = 2π f (3. if we shift the cosine function to the right by π ⁄ 2 radians or 90° . or 90° . Obviously. we say that the cosine function leads (is ahead of) the sine function by π ⁄ 2 radians or 90° . The frequency is denoted by the letter f and in terms of the period T and (3. v ( t ) = V max cos ωt . in this case. If the phase angle between them is 0° degrees. or there is a phase angle of 90° between the cosine and sine functions. For example. these are of the same frequency.

or both as sine functions. and that the sine function lags the cosine by 90° .25 sin ( ωt – ϕ ) 1. it is acceptable to express as v ( t ) = 100 sin ( 2000πt – π ⁄ 6 ) v ( t ) = 100 sin ( 2000πt – 30° ) since the subtraction inside the parentheses needs not to be performed.5 -1 -1. and – sin x = sin ( x ± 180 ° ) and sin x = cos ( x – 90 ° ) 3−4 Numerical Analysis Using MATLAB® and Excel®. When two sinusoids are to be compared in terms of their phase difference. Third Edition Copyright © Orchard Publications .75 sin ( ωt + θ ) 0 ωt ϕ -2 0 2 4 6 8 Figure 3.5 -4 sin ωt θ 0. Then. Example 3. and should also be written with positive amplitudes.4. We should remember also that a negative amplitude implies 180° phase shift.5 0 -0. Out-of-phase sinusoids It is convenient to express the phase angle in degrees rather than in radians in a sinusoidal function.Chapter 3 Sinusoids and Phasors 1.5 1 0. these must first be written either both as cosine functions. For example.1 Find the phase difference between the sinusoids i 1 = 120 cos ( 100πt – 30 ° ) and i 2 = – 6 sin ( 100πt – 30 ° ) Solution: We recall that the minus (−) sign indicates a ± 180° phase shift.

Then.3 deg r r π radians Figure 3.4) (3.5. there are 2π or 6.707 4 2 π 1 cos 60° = cos -. provide the following relations and formulas for quick reference.= -. Definition of radian As shown in Figure 3.866 6 2 π 2 cos 45° = cos -.6) (3. and it is denoted by the symbol ω . In our subsequent discussion.= 0. Some useful trigonometric relations are given below for quick reference.7) Numerical Analysis Using MATLAB® and Excel®.≈ 57. we will be using several trigonometric identities.283… radians in 360° degrees. therefore. whose radius is r units in length. or i 1 lags i 2 by 90° . we see that i 2 leads i 1 by 90° . 1 radian = 57.5. the radian is a circular angle subtended by an arc equal in length to the radius of the circle.Characteristics of Sinusoids i 2 = 6 sin ( 100πt – 210 ° ) = 6 sin ( 100πt + 150 ° ) = 6 cos ( 100πt + 150 ° – 90 ° ) = 6 cos ( 100πt + 60 ° ) and comparing i 2 with i 1 .= -----.5 3 2 (3.= 0. therefore. cos 0° = cos 360° = cos 2π = 1 π 3 cos 30° = cos -.3° 2π (3. derivatives and integrals involving trigonometric functions.5) (3. Let us also review the definition of a radian and its relationship to degrees with the aid of Figure 3.= -----.3) The angular velocity is expressed in radians per second. a rotating vector that completes n revolutions per second.= 0. We. 360° 1 radian = ---------. has an angular velocity ω = 2πn radians per second. The circumference of a circle is 2πr units.5. Then. Third Edition Copyright © Orchard Publications 3−5 .

15) (3.= 0.= -.= 0.= 0.= 0.12) (3.= -----.11) (3.= 0 2 2π –1 cos 120° = cos ----.= -----.13) (3.Chapter 3 Sinusoids and Phasors π cos 90° = cos -.5 3 11π cos 330° = cos -------.= ----.9) (3.23) 3−6 Numerical Analysis Using MATLAB® and Excel®.866 3 2 (3.5 3 2 ------------cos 150° = cos 5π = – 3 = – 0.17) (3.22) (3.10) (3.707 2 4 --------cos 240° = cos 4π = – 1 = – 0.= – 0.707 4 2 π 1 sin 60 ° = sin -.866 6 sin 0 ° = sin 360 ° = sin 2 π = 0 --sin 30 ° = sin π = 1 = 0.866 2 6 cos 180° = cos π = – 1 7π – 3 cos 210° = cos ----.18) (3.5 2 6 π 2 sin 45 ° = sin -. Third Edition Copyright © Orchard Publications .19) (3.16) (3.= 0.14) (3.866 6 2 ------------cos 225° = cos 5π = – 2 = – 0.= – 0.= --------.20) (3.= 1 2 2π 3 sin 120 ° = sin ----.5 2 3 ----cos 270° = cos 3π = 0 2 5π cos 300° = cos ----.8) (3.21) (3.866 3 2 π sin 90 ° = sin -.

= ----.25) (3.5 6 2 5π – 2 sin 225 ° = sin ----.= – 0.38) (3.42) Numerical Analysis Using MATLAB® and Excel®.866 2 3 ----sin 270 ° = sin 3π = – 1 2 5π – 3 sin 300 ° = sin ----.5 6 2 cos ( – θ ) = cos θ cos ( 90° + θ ) = – sin θ cos ( 180° – θ ) = – cos θ sin ( – θ ) = – sin θ sin ( 90° + θ ) = cos θ sin ( 180° – θ ) = sin θ sin θ tan θ = ----------cos θ cos θ 1 cot θ = ----------.= ----.= --------.35) (3.28) (3.39) (3.27) (3.= – 0.37) (3.= 0.26) (3.30) (3.41) (3.34) (3.40) (3. Third Edition Copyright © Orchard Publications 3−7 .= ---------sin θ tan θ 1 sec θ = ----------cos θ 1 csc θ = ---------sin θ tan ( 90° + θ ) = – cot θ (3.33) (3.= – 0.29) (3.707 4 2 ------------sin 240 ° = sin 4π = – 3 = – 0.31) (3.36) (3.32) (3.= – 0.24) (3.5 6 2 sin 180 ° = sin π = 0 7π –1 sin 210 ° = sin ----.= --------.866 3 2 11π –1 sin 330 ° = sin -------.Characteristics of Sinusoids 5π 1 sin 150 ° = sin ----.= -.

55) (3.44) (3.Chapter 3 Sinusoids and Phasors tan ( 180° – θ ) = – tan θ cos ( θ + φ ) = cos θ cos φ – sin θ sin φ cos ( θ – φ ) = cos θ cos φ + sin θ sin φ sin ( θ + φ ) = sin θ cos φ + cos θ sin φ sin ( θ – φ ) = sin θ cos φ – cos θ sin φ tan θ + tan φ tan ( θ + φ ) = ------------------------------1 – tan θ tan φ tan θ – tan φtan ( θ – φ ) = ------------------------------1 + tan θ tan φ cos θ + sin θ = 1 cos 2θ = cos θ – sin θ sin 2 θ = 2 sin θ cos θ 2 tan θ tan 2θ = --------------------2 1 – tan θ 1 2 cos θ = -.58) (3.49) (3.43) (3.47) (3.54) (3.( 1 + cos 2θ ) 2 1 2 sin θ = -.50) (3.46) (3.57) (3.cos ( θ – φ ) 2 2 1 -cos θ sin φ = -.52) (3.56) (3.cos ( θ – φ ) – 1 cos ( θ + φ ) 2 2 2 2 2 2 (3.51) (3.cos ( θ + φ ) + -.48) (3.( 1 – cos 2θ ) 2 1 1 cos θ cos φ = -.sin ( θ + φ ) + -.sin ( θ + φ ) – 1 sin ( θ – φ ) 2 2 1 1 sin θ cos φ = -.59) 3−8 Numerical Analysis Using MATLAB® and Excel®. Third Edition Copyright © Orchard Publications .53) (3.sin ( θ – φ ) 2 2 1 -sin θ sin φ = -.45) (3.

64) tan -. a β c γ b α Figure 3.e Numerical Analysis Using MATLAB® and Excel®.6 be any triangle.( α – β ) 2 a – b = --------------------------------------1 a+b tan -.= ---------------------------1 b+c tan -.68) (3.63) 1 tan -.sin ax + c a 1 ax +c a dx = -. are used extensively in engineering.( β – γ ) 2 2 2 2 b–c ---------.60) (3. General triangle Then. 1 tan -. a b c ---------.65) (3.66) (3.( γ + α ) 2 The following differential and integral trigonometric and exponential functions.= ---------------------------1 c+a (3. by the law of cosines. Third Edition Copyright © Orchard Publications 3−9 . by the law of sines.( γ – α ) and by the law of tangents.67) (3.69) (3.62) (3.= ---------.61) (3.70) ∫ sin ax dx ∫ cos a x dx ∫e ax = – 1 cos ax + c -a 1 = -.6.( β + γ ) 2 2 c–a ---------.( α + β ) 2 c = a + b – 2ab cos γ 1 tan -.= --------sin α sin β sin γ a = b + c – 2bc cos α b = a + c – 2ac cos β 2 2 2 2 2 2 (3.Characteristics of Sinusoids Let Figure 3. dv d ( sin v ) = cos v ----dx dx dv d ( cos v ) = – sin v ----dx dx d ( e v ) = e v dv ----dx dx (3.

the square root of minus one is denoted as i . and if z = tan u . These are called Inverse Trigonometric Functions. Also. then u = tan –1z . A fourth 90° rotation returns the vector to its original position.7. θ = 60° 3. In this case. then v = sin –1w .5 = θ Solution: Here.4 Phasors In the language of mathematics. * A more appropriate nomenclature for the real and imaginary axes would be the axis of the cosines and the axis of the sines respectively.* An imaginary number is the product of a real number. if it is given that a vector A has the direction along the right side of the x -axis as shown in Figure 3.7). the imaginary axis is just as important as the real axis. another multiplication of the new vector jA by j will produce another 90° counterclockwise direction. Essentially. 2 3 4 3−10 Numerical Analysis Using MATLAB® and Excel®. the vector A has rotated 180° and its new value now is – A .5. j = 1 . and the rotating vector A is referred to as a phasor. j = – j . j is an operator that produces a 90° counterclockwise rotation to any vector to which it is applied as a multiplying factor. we denote i as j to avoid confusion with current i . Thus. we conclude that j = – 1 . we want to find the angle θ given that its cosine is 0. Thus.3 Inverse Trigonometric Functions The notation cos –1y or arc cos y is used to denote an angle whose cosine is y . but it has been rotated counterclockwise by 90° . From (3. i = – 1 .5 . by the operator j . multiplication of this vector by the operator j will result in a new vector jA whose magnitude remains the same. Note: In our subsequent discussion. that is. Therefore. r is a real number and jr is an imaginary number. Example 3. Similarly. if w = sin v . When this vector is rotated by another 90° for a total of 270° . we will designate the x -axis (abscissa) as the real axis. and the y -axis (ordinate) as the imaginary axis with the understanding that the “imaginary” axis is just as “real” as the real axis.2 Find the angle θ if cos –10. if y = cos x . and thus its value is again A . cos 60° = 0. In the electrical engineering field.Chapter 3 Sinusoids and Phasors 3. In other words. Third Edition Copyright © Orchard Publications . Thus. then x = cos –1y . its value becomes j ( – A ) = – jA . Therefore. say r .

and b = Im { A } the imaginary part of A .7. the number A = a + jb where a and b are both real numbers. When written as A = a + jb . then and A + B = (a + c) + j(b + d) A – B = (a – c) + j(b – d ) Example 3. and an imaginary component equal to the sum of the imaginary components. 3. Find A + B and A – B Solution: Numerical Analysis Using MATLAB® and Excel®. By definition. we change the signs of the components of the subtrahend and we perform addition. if A = a + jb and B = c + jd . For subtraction. are equal if and only if their real parts are equal and also their imaginary parts are equal. Thus.Addition and Subtraction of Phasors jA y j ( jA ) = j 2 A = – A A x 2 j ( – jA ) = – j A = A j ( – A ) = j 3 A = – jA Figure 3.5 Addition and Subtraction of Phasors The sum of two phasors has a real component equal to the sum of the real components. is a complex number. Since in engineering we use complex quantities as phasors. A = B if and only if a = c and b = d . henceforth any complex number will be referred to as a phasor. For example. Third Edition Copyright © Orchard Publications 3−11 . Thus. two phasors A and B where A = a + jb and B = c + jd .3 It is given that A = 3 + j4 . and B = 4 – j2 . a = Re { A } and b = Im { A } where Re { A } denotes real part of A . it is said to be expressed in rectangular form. The j operator A complex number is the sum (or difference) of a real number and an imaginary number. Then.

Chapter 3 Sinusoids and Phasors A + B = (3 + j4) + (4 – j2) = (3 + 4) + j(4 – 2) = 7 + j2 A – B = (3 + j4) – (4 – j2) = (3 – 4) + j(4 + 2) = – 1 + j6 3. Find A∗ Solution: The conjugate of the phasor A has the same real component. Then. Example 3. Find A ⋅ B Solution: A ⋅ B = ( 3 + j4 ) ⋅ ( 4 – j2 ) = 12 – j6 + j16 – j 2 8 = 20 + j10 The conjugate of a phasor. but the imaginary component has opposite sign. the result is a real number. if A = a + jb and B = c + jd . if A = a + jb . denoted as A∗ .71) Example 3.6 Multiplication of Phasors Phasors are multiplied using the rules of elementary algebra. and making use of the fact that j 2 = – 1 . and with an imaginary component of opposite sign.5 It is given that A = 3 + j5 .4 It is given that A = 3 + j4 and B = 4 – j2 . Thus. it follows that A ⋅ B = ac + jad + jbc – b d = ( ac – bd ) + j ( ad + bc ) (3. then A ⋅ B = ( a + jb ) ⋅ ( c + jd ) = ac + jad + jbc + j 2 bd and since j 2 = – 1 . then A∗ = a – j b . A∗ = 3 – j 5 If a phasor A is multiplied by its conjugate. if A = a + jb . Third Edition Copyright © Orchard Publications . then A ⋅ A∗ = ( a + jb ) ( a – jb ) = a 2 – jab + jab – j 2 b 2 = a + b 2 2 3−12 Numerical Analysis Using MATLAB® and Excel®. is another phasor with the same real component. Thus. Thus.

it is desirable to obtain the quotient separated into a real part and an imaginary part. A B∗ a + jb ( a + jb ) ( c – jd ) ( ac + bd ) + j ( bc – ad ) A --.72).= ----------------------------------------------------2 2 B B∗ B c + jd ( c + jd ) ( c – jd ) c +d ( ac + bd ) ( bc – ad ) = ---------------------.28 ------------------------------------------------2 2 25 25 B 25 4 + j3 (4 + j3)(4 – j3) 4 +3 3. Example 3.+ j ---------------------2 2 2 2 c +d c +d (3.72).= ------------------------------------.7 Division of Phasors When performing division of phasors. and B = 4 + j3 .= --. then. if A = a + jb and B = c + jd .96 + j 0.= ------------. Find A ⋅ A∗ Solution: A ⋅ A∗ = ( 3 + j5 ) ( 3 – j5 ) = 3 + 5 = 9 + 25 = 34 2 2 3.6 It is given that A = 3 + j5 .⋅ -----. Find A ⁄ B Solution: Using the procedure of (3.7 It is given that A = 3 + j4 . we get 73 + j4 (3 + j4)(4 – j3) 24 + j 7 A = ------------. Using this procedure. Thus.= 12 – j 9 + j 16 + 12 = ---------------.= 24 + j ----. This procedure is called rationalization of the quotient. and it is done by multiplying the denominator by its conjugate.72) In (3. we see that the quotient is easily separated into a real and an imaginary part.= -------------------------------------.73) 3−13 .8 Exponential and Polar Forms of Phasors The relations e and Numerical Analysis Using MATLAB® and Excel®.= 0. we multiplied both the numerator and denominator by the conjugate of the denominator to eliminate the j operator from the denominator of the quotient.Division of Phasors Example 3. Third Edition Copyright © Orchard Publications jθ = cos θ + j sin θ (3.

75) This expression represents a phasor. we get a + b = ( C cos θ ) + ( C sin θ ) = C ( cos θ + sin θ ) = C Then.⎞ or ⎝a⎠ j ⎛ tan ⎝ b⎞ -a⎠ (3. to convert a phasor from rectangular to exponential form. we use the expression a + jb = a +b e 2 2 –1 (3.73) by the real positive constant C we get: Ce jθ – jθ = cos θ – j sin θ (3.81) 3−14 Numerical Analysis Using MATLAB® and Excel®.75) and (3.77) or C = a +b 2 2 (3. say a + jb . we get a = C cos θ and b = C sin θ Squaring and adding the expressions in (3.76) Equating real and imaginary parts in (3.78) Also.77) b = C sin θ = tan θ -. we use the expressions Ce Ce jθ = C cos θ + j C sin θ = C cos θ – j C sin θ – jθ (3.80) To convert a phasor from exponential to rectangular form. C = a +b 2 2 2 2 2 2 2 2 2 2 2 (3. Multiplying (3.74) = C cos θ + j C sin θ (3. Third Edition Copyright © Orchard Publications . from (3.77). and thus Ce jθ = a + jb (3.79) Therefore.76).Chapter 3 Sinusoids and Phasors e are known as the Euler’s identities.--------------a C cos θ –1 b θ = tan ⎛ -.

8. we will verify the results with the following MATLAB co-ordinate transformation functions: [theta.y) − transforms from Cartesian to polar co−ordinates. In Examples 3. Ce jθ = C ∠θ (3. thetax=angle(x)*180/pi. disp(thetax) 5 53. We must remember that the phase angle θ is always measured with respect to the positive real axis.1° 3 Re Figure 3. 4 Im 5 53. and the right side is the polar form.8.1 ° = 5 ∠53.8 and 3. 3 + j4 = 3 +4 ⋅e 2 2 j tan ( 4 ⁄ 3 ) – = 5e j53.9 below. Third Edition Copyright © Orchard Publications 3−15 . The real and imaginary components of this phasor are shown in Figure 3.y] = pol2cart(theta.r) − transforms from polar to Cartesian co−ordinates Example 3. disp(magx). magx=abs(x).82) is the exponential form.1° Check with MATLAB: x=3+j*4. – 1 + j2 c. 3 + j4 b. and rotates in the counterclockwise direction. that is.Exponential and Polar Forms of Phasors The polar form is essentially the same as the exponential form but the notation is different.1301 or Numerical Analysis Using MATLAB® and Excel®. 4 – j3 Solution: a.82) where the left side of (3. The components of 3 + j4 Then.8 Convert the following phasors to exponential and polar forms: a.r] = cart2pol(x. [x. – 2 – j d.

y = 4.9. magy=abs(y). – 1 + j2 = 2 Figure 3.6° = 2. thetay=angle(y)*180/pi. The components of – 1 + j2 1 +2 e 2 j tan ( 2 ⁄ – 1 ) – = 5e j116.6° Check with MATLAB: y=−1+j*2.6 ° = 5 ∠116.Chapter 3 Sinusoids and Phasors x = 3.236 ∠116.8 (a) b. A brief introduction to Simulink is provided in Appendix B. 3−16 Numerical Analysis Using MATLAB® and Excel®.4° −1 116. ISBN 09744239-7-1. disp(thetay) * The reader who is not familiar with Simulink may skip this model and all others without loss of continuity.y). Simulink model for Example 3.9273 r = 5 deg = 53. The K value for the Gain block has been specified as 180 ⁄ π to convert radians into degrees. [theta. Im 5 63. For an introduction to Simulink. Third Edition Copyright © Orchard Publications . The real and imaginary components of this phasor are shown in Figure 3.1301 We can also verify the result with Simulink®* as shown in the model of Figure 3. deg = theta*180/pi theta = 0.r] = cart2pol(x.10. disp(magy). Figure 3.9. please refer to “Introduction to Simulink with Engineering Applications”.10.6° Re 2 Then.

236 ∠– 153. deg = theta*180/pi theta = 2.6° = 5e j ( – 153.12.2361 116.Exponential and Polar Forms of Phasors 2. Third Edition Copyright © Orchard Publications 3−17 .6 ° = Im 5 ∠206. disp(thetav) Numerical Analysis Using MATLAB® and Excel®.4°(Measured Clockwise) −1 Figure 3. magv=abs(v). –2 –j 1 = 2 +1 ⋅e 2 2 j tan ( – 1 ⁄ – 2 ) – = 5e j206.6° 5 Re −153.5651 Check with the Simulink model of Figure 3.2361 deg = 116.5651 or x = −1.8 (b) c.11: Figure 3. thetav=angle(v)*180/pi.4 ° 206. The real and imaginary components of this phasor are shown in Figure 3.6° −2 26.12. The components of – 2 – j Check with MATLAB: v=−2−j*1. Then. disp(magv). Simulink model for Example 3.0344 r = 2. y = 2.y).r] = cart2pol(x.4 ) ° = 2. [theta.11.

1° −36.1 ° = 5 ∠323. disp(thetaw) 3−18 Numerical Analysis Using MATLAB® and Excel®.y).Chapter 3 Sinusoids and Phasors 2.4349 Check with the Simulink model of Figure 3. Third Edition Copyright © Orchard Publications . Im 323. disp(magw).4349 or x = −2.8 (c) d.9 ° = 5 ∠– 36. deg = theta*180/pi theta = -2.14.9° −3 5 4 Re Then. y = −1.2361 -153.1° = 5e – j36. thetaw=angle(w)*180/pi. 4 –j 3 = 4 +3 ⋅e 2 2 Figure 3. The real and imaginary components of this phasor are shown in Figure 3.13: Figure 3.14.9 ° Check with MATLAB: w=4−j*3. [theta.6779 r = 2. Simulink model for Example 3.2361 deg = -153. The components of 4 – j3 j tan ( – 3 ⁄ 4 ) – = 5e j323.13.r] = cart2pol(x. magw=abs(w).

15: Figure 3. Since each j rotates a vector by 90° counterclockwise.y). Therefore. – 2 ∠30° = 2 ∠( 30° + 180° ) = 2 ∠210° = 2 ∠– 150° The components of this phasor are shown in Figure 3. then – 2 ∠30° is the same as 2 ∠30° rotated counterclockwise by 180° .9 Express the phasor – 2 ∠30° in exponential and in rectangular forms. deg = theta*180/pi theta = -0.Exponential and Polar Forms of Phasors 5 -36. Third Edition Copyright © Orchard Publications 3−19 .6435 r = 5 deg = -36.15.8699 or x = 4.r] = cart2pol(x. [theta.16.8699 Check with the Simulink model of Figure 3. Simulink model for Example 3. y = −3.8 (d) Example 3. Numerical Analysis Using MATLAB® and Excel®. Solution: We recall that – 1 = j 2 .

Third Edition Copyright © Orchard Publications .Chapter 3 Sinusoids and Phasors Im 210° −1.17: Figure 3.7578 y = -0. that is.9541 Check with the Simulink model of Figure 3.17. AB = MN ∠( θ + φ ) = M e jθ Ne jφ = MN e j(θ + φ ) (3.r) x = -1. we multiply the magnitudes and we add the phase angles. [x.y] = pol2cart(theta*180/pi.16. theta = 30/pi.5 ) = – 1. Simulink model for Example 3. however. To multiply two phasors in exponential (or polar) form.73 – j Check with MATLAB: r = −2.9 Note: The rectangular form is most useful when we add or subtract phasors. The components of 2 ∠– 150° = 2 ( cos 150° – j sin 150° ) = 2 ( – 0. the exponential and polar forms are most convenient when we multiply or divide phasors. if A = M ∠θ and B = N ∠φ then.73 30° 2 Re −150°(Measured Clockwise) −1 Then.866 – j0. 2 ∠– 150 ° = 2e – j 150 ° Figure 3.83) 3−20 Numerical Analysis Using MATLAB® and Excel®.

It can be done faster with just MATLAB.Exponential and Polar Forms of Phasors Example 3. r2=7. Simulink model for Example 3. Third Edition Copyright © Orchard Publications 3−21 .83 ∠– 44.22. The intent here is to introduce relevant Simulink blocks for more complicated models. deg2=−118.58.7° ) = 90.7° ) = 90.3° by B = 7.58 ∠74.7° ) ] = 90.10 Multiply A = 12.7.10 * It would certainly be a waste of time to use Simulink for such an application. deg=deg1+deg2 r = 90.4 ° and multiplication in exponential form yields AB = ( 12.8276 deg = -44.83e j ( 74.7° Solution: Multiplication in polar form yields AB = ( 12.3.58e j74.22e – j 118.4000 Check with the Simulink model of Figure 3.58 × 7.22 ) ∠[ 74.3° ) ( 7.18*: Figure 3.22 ∠– 118. Numerical Analysis Using MATLAB® and Excel®.83e – j 44. r=r1*r2. deg1=74.4° Check with MATLAB: r1=12.18.3° – 118.3° + ( – 118.

= ---.74e --– j118.74e j193° = 1.7° Solution: Division in polar form yields A 12.3. deg2=−118.11 Divide A = 12.= ---------------------------------. that is.e N jφ N B Ne jθ (3.74e – j 167° Check with MATLAB: r1=12. r=r1/r2.22.7° = 1.7.74 ∠– 167 ° B 7. deg1=74.= 1.3° --.3° j74. if then.58e A = ---------------------------.Chapter 3 Sinusoids and Phasors To divide one phasor by another when both are expressed in exponential or polar form.∠( θ – φ ) = -----------.22e j74.22 ∠– 118.19*: * Same comment as on the footnote of the previous page.3° – ( – 118.74 ∠193° = 1.7° Division in exponential form yields 12.7° B 7. r2=7.58.3° by B = 7.= ---. deg=deg1−deg2 r = 1.74 ∠[ 74.7424 deg = 193 Check with the Simulink model of Figure 3. Third Edition Copyright © Orchard Publications .58 ∠74. we divide the magnitude of the dividend by the magnitude of the divisor.58 ∠74.= 1. A = M ∠θ and B = N ∠φ M j(θ – φ ) M Me A --.7° ) ] = 1.22 ∠– 118. and we subtract the phase angle of the divisor from the phase angle of the dividend.84) Example 3. 3−22 Numerical Analysis Using MATLAB® and Excel®.3° e j118.

19.Exponential and Polar Forms of Phasors Figure 3. Third Edition Copyright © Orchard Publications 3−23 .11 Numerical Analysis Using MATLAB® and Excel®. Simulink model for Example 3.

denoted as Hz . if y = cos x . or both as sine functions. is used to express the number of cycles per second. • Two (or more) sinusoids can be out-of-phase by a phase angle other than 90° . • The notation cos –1y or arc cos y is used to denote an angle whose cosine is y . • It is customary to express the phase angle in degrees rather than in radians in a sinusoidal function. • The period T of an alternating current or voltage is the smallest value of time which separates recurring values of the alternating waveform. we say that the cosine and sine functions are out-of-phase by 90° . These are called Inverse Trigonometric Functions. • The cosine function leads (is ahead of) the sine function by π ⁄ 2 radians or 90° . these are of the same frequency since two sinusoids of different frequencies can never be in phase. or there is a phase angle of 90° between the cosine and sine functions. 3−24 Numerical Analysis Using MATLAB® and Excel®. and should also be written with positive amplitudes. • A negative amplitude implies 180° phase shift. • The angular velocity ω is commonly known as angular or radian frequency and ωT = 2π • The term frequency in Hertz. • It is important to remember that when we say that one sinusoid leads or lags another sinusoid. whose radius is r units in length.9 Summary • An alternating current (or voltage) alternates between positive and negative values at regularly recurring intervals of time. • The radian is a circular angle subtended by an arc equal in length to the radius of the circle. • A phasor is a rotating vector expressed as a complex number where j is an operator that rotates a vector by 90° in a counterclockwise direction. Alternately.Chapter 3 Sinusoids and Phasors 3. Third Edition Copyright © Orchard Publications . these must first be written either both as cosine functions. f = 1 ⁄ T . For example. The frequency is denoted by the letter f and in terms of the period T . The circumference of a circle is 2πr . Thus. and the sine function lags (is behind) the cosine function by π ⁄ 2 radians or 90° . then x = cos – 1y . we write v ( t ) = 100 sin ( 2000πt – π ⁄ 6 ) as v ( t ) = 100 sin ( 2000πt – 30° ) • When two sinusoids are to be compared in terms of their phase difference. • Sine and cosine waveforms and these are referred to as sinusoids. The frequency f is often referred to as the cyclic frequency to distinguish it from the radian frequency ω.

and an imaginary component equal to the sum of the imaginary components.Summary • Two phasors A and B where A = a + jb and B = c + jd . we use the expressions Ce Ce jθ = C cos θ + j C sin θ = C cos θ – j C sin θ – jθ • The polar form is essentially the same as the exponential form but the notation is different. Numerical Analysis Using MATLAB® and Excel®. Third Edition Copyright © Orchard Publications 3−25 . if A = a + jb and B = c + jd . then A ⋅ B = ac + jad + jbc – b d = ( ac – bd ) + j ( ad + bc ) • The conjugate of a phasor. it is desirable to obtain the quotient separated into a real part and an imaginary part. If A = a + jb and B = c + jd . is another phasor with the same real component. if A = a + jb and B = c + jd . Thus. are equal if and only if their real parts are equal and also their imaginary parts are equal. we change the signs of the components of the subtrahend and we perform addition. and rotates in the counterclockwise direction. however. Thus.= ------------. the exponential and polar forms are most convenient when we multiply or divide phasors. denoted as A∗ . This is achieved by multiplying the denominator by its conjugate. we use the expression a + jb = a +b e 2 2 j ⎛ tan ⎝ –1 b⎞ -a⎠ • To convert a phasor from exponential to rectangular form. Ce jθ = C ∠θ and it is important to remember that the phase angle θ is always measured with respect to the positive real axis. • The rectangular form is most useful when we add or subtract phasors. then A + B = ( a + c ) + j ( b + d ) and A – B = ( a – c ) + j ( b – d ) • Phasors are multiplied using the rules of elementary algebra.= ------------------------------------.( bc – ad ) ) A --. • When performing division of phasors. and with an imaginary component of opposite sign.= ---------------------.+ j ---------------------2 2 2 2 2 2 B c + jd ( c + jd ) ( c – jd ) c +d c +d c +d • The relations e ties. For subtraction.= ----------------------------------------------------. A = B if and only if a = c and b = d. Thus. a + jb ( a + jb ) ( c – jd ) ( ac + bd ) + j ( bc – ad ) ( ac + bd . jθ = cos θ + j sin θ and e – jθ = cos θ – j sin θ are known as the Euler’s identi- • To convert a phasor from rectangular to exponential form. then. Thus. then A∗ = a – j b . that is. if A = a + jb . • The sum of two phasors has a real component equal to the sum of the real components.

Chapter 3 Sinusoids and Phasors • To multiply two phasors in exponential (or polar) form. if then. if A = M ∠θ and B = N ∠φ then. that is. we divide the magnitude of the dividend by the magnitude of the divisor.∠( θ – φ ) = -----------.e N jφ N B Ne jθ 3−26 Numerical Analysis Using MATLAB® and Excel®. AB = MN ∠( θ + φ ) = M e jθ Ne jφ = MN e j(θ + φ ) • To divide one phasor by another when both are expressed in exponential or polar form. and we subtract the phase angle of the divisor from the phase angle of the dividend.= ---. A = M ∠θ and B = N ∠φ M j(θ – φ) M Me A --. Third Edition Copyright © Orchard Publications .= ---. we multiply the magnitudes and we add the phase angles. that is.

a. ( – 3 + j5 ) – ( 1 + j6 ) c.b. Third Edition Copyright © Orchard Publications 3−27 . Perform the following operations.b. ( 2 – j3 ) – ( 2 – j3 )∗ d. and check your answers with MATLAB. ---------------------. ( 3 – j2 ) ⋅ ( 3 – j2 )∗ e. 6 jθ n jnθ or n re jθ = n re jθ ⁄ n . Any phasor A can be expressed as A = a + jb = r ( cos θ + j sin θ ) = re jθ Using the identities ( re ) = r e a. 22 + j6 8 + j6 120 ( 3 – j2 ) a. 4 100 2 ( 1 – j ) Check your answers with MATLAB 4.10 Exercises 1. ------------------– j30° 5 ∠– 150° – 2e j60° Check your answers with MATLAB 9 – j4 6. -------------. ----------------. ( 2 – j4 ) ⋅ ( 3 + j5 ) ⋅ ( 3 – j2 ) ⋅ ( – 2 – j3 ) 2. 5. ( 2 – j4 ) ⋅ ( 3 + j5 ) f. Perform the following operations. Find the real and imaginary components of -----------------– 5 + jx Numerical Analysis Using MATLAB® and Excel®.d.c. ---------------. ( 3 – j2 ) ⋅ ( – 2 – j3 ) g. compute: 12 + j5 b. Compute the exponential and polar forms of – 8 + j3 9 + j5a. ( 2 – j4 ) + ( 3 + j4 ) b. Compute the rectangular form of 4 ∠30° e a. and check your answers with MATLAB. --------------------3 + j2 –3–j 4 – j10 ( 3 – j2 )∗ 3. ---------------.b. -----------------– 4 – j2 – 2 + j4 Check your answers with MATLAB.Exercises 3.

(2−4j)*(3+5j)..0000i ans = -3.Chapter 3 Sinusoids and Phasors 3.0600e+002i 3−28 Numerical Analysis Using MATLAB® and Excel®.2200e+002 . (2−3j)−(2+3j). ( 2 – j4 ) + ( 3 + j4 ) = 5 + 0 = 5 b. (−3+5j)−(1+6j). (3−2j)*(3+2j).5.0000 . ( 2 – j3 ) – ( 2 – j3 )∗ = ( 2 – j3 ) – ( 2 + j3 ) = 0 – j6 d.. a. (3−2j)*(−2−3j).. ( – 3 + j5 ) – ( 1 + j6 ) = – 4 – j c.0000 .0000i ans = 0 .2.1.0000 .0000i ans = -12. g.6. ( 2 – j4 ) ⋅ ( 3 + j5 ) = 6 + j10 – j12 + 20 = 26 – j2 f. ( 3 – j2 ) ⋅ ( 3 – j2 )∗ = ( 3 – j2 ) ⋅ ( 3 + j2 ) = 9 + j6 – j6 + 4 = 13 e. (2−4j)*(3+5j)*(3−2j)*(−2−3j) ( 3 – j2 ) ⋅ ( – 2 – j3 ) = – 6 – j9 + j4 – 6 = – 12 – j5 ( 2 – j4 ) ⋅ ( 3 + j5 ) ⋅ ( 3 – j2 ) ⋅ ( – 2 – j3 ) = ( 6 + j10 – j12 + 20 ) ⋅ ( – 6 – j9 + j4 – 6 ) = ( 26 – j2 ) ⋅ ( – 12 – j5 ) = – 312 – j130 + j24 – 10 = – 322 – j106 ans = 5 ans = -4. Third Edition Copyright © Orchard Publications .0000i ans = 13 ans = 26. Check with MATLAB: (2−4j)+(3+4j).11 Solutions to End−of−Chapter Exercises 1.1.

12/13i 3.⋅ -------------.1i ans = 120/29 + 300/29i ans = 5/13 . 120/(4−10j). (3−2j)/(3+2j) ans = 6 .= ---------------.6883 – j0.= 78 – j26 = 6 – j2 2 2 3 + j2 3 + j2 3 – j2 3 +2 13 8 + j6 – 3 + j – 24 + j8 – j18 – 6 8 + j6 b.0658 + j sin 0.= -------------. a. ---------------.0658 ) = 1. --------------------.4485 × 1.= 9 – j6 – j6 – 4 = 5 – j12 = ----.= – 30 – j10 = – 3 – j ----------------------2 2 –3–j –3–j –3+j 3 +1 10 300 120 . 22 + j6 3 – j2 66 – j44 + j18 + 12 ---------------------------------a. (100*sqrt(2)*(1−j))^(1/4) Numerical Analysis Using MATLAB® and Excel®.2i ans = -3 .3948 ⁄ 6 = 13 1⁄6 ⋅e j0. 6 12 + j5 = 6 13e j0.⋅ ------------.= -----------------.395 = 6 13 ⋅ e j0.= -------------------------------------------. 22 + j6 = ---------------. 4 100 2 ( 1 – j ) = 4 100 2 ⋅ 2e – jπ ⁄ 4 = ( 100 2 ⋅ 2e – jπ ⁄ 4 1 ⁄ 4 ) = ( 100 2 ) 1⁄4 ⋅ 2 1 ⁄ 4 – jπ ⁄ 16 e = ( 3.= 120 + j ⋅ -----------------------------------------2 4 – j10 4 – j10 4 + j10 4 + 10 116 116 29 29 ( 3 – j2 ) ( 3 – j2 ) 12 ( 3 – j2 ) 5-----------------------------------------------d.Solutions to End−of−Chapter Exercises 2.– j ⋅ ----2 2 ( 3 – j2 )∗ ( 3 + j2 ) ( 3 – j2 ) 3 +2 13 13 13 Check with MATLAB: 22+6j)/(3+2j).4 + j10 1200 120 -------c.= ---------------------------------------------.5334 ( cos 0. (8+6j)/(−3−j).0905 ) ( cos ( π ⁄ 16 ) – j sin ( π ⁄ 16 ) ) = 3.0658 = 1.⋅ -----------------. Third Edition Copyright © Orchard Publications 3−29 .7337 Check with MATLAB: (12+5j)^(1/6).53 + j0.10 b.⋅ ---------------.= 4802 + j1200 = 480 + j ⋅ ----------. -------------.

abs(x).9105e j42. y=(−8+3j)/(−2+4j).5 ( cos 90° + j sin 90° ) = – 0.0981 9 + j5 9 +5 ⋅e 106 ⋅ e ----------------.7337i 4. 4 ∠30° ---------------------.= – 0.9105 ans = 42.8744° = 1.6883 .= ---------------------------------.5104 ans = 1. – j 3.5301 + 0.0. – 2 + j4 2 +4 ⋅e = 1. ------------------.7483 – 8 + j3 8 +3 ⋅e -----------------.5081 ° –1 = 2.5071 = 2.3022 ans = -177.= 1.1071 2 2 j tan ( 4 ⁄ – 2 ) 20 e b.= --------. angle(x)*180/pi.3022e – j177. Third Edition Copyright © Orchard Publications . angle(y)*180/pi ans = 2.5e – j30° – 2e j60° j90° = – 0.8744° Check with MATLAB: x=(9+5j)/(−4−2j).6052 – 4 – j2 2 2 j tan ( – 2 ⁄ – 4 ) 20 ⋅ e 4 +2 ⋅e 2 2 j tan ( 5 ⁄ 9 ) –1 j0.1008i ans = 3.5 3−30 Numerical Analysis Using MATLAB® and Excel®.5081 ° – j0. a..⋅ ----------------.3588 2 2 j tan ( 3 ⁄ – 8 ) 73 e j0.= ------------------------------------------------------.3022 ∠– 177.8789 5.9105e –1 – j1. a.Chapter 3 Sinusoids and Phasors ans = 1. abs(y)..= 2..9105 ∠42.8 5 ∠– 150° e b.= ----------------------------------------------------.5 ( 0 + j ) = – j0.3022e –1 j3.= ( 4 ⁄ 5 ) ∠180° = – 0.

9 – j4 – 5 – j x – 45 – j9x + j20 – 4x 9 – j4 – 4x – 45 – 9x + 20 -----------------...⋅ -------------.0.+ j ---------------------2 2 2 2 – 5 + jx – 5 – j x – 5 + jx x + 25 5 +x x + 25 Numerical Analysis Using MATLAB® and Excel®.0000 .= -----------------. Third Edition Copyright © Orchard Publications 3−31 ..0.= -------------------------------------------------. exp(pi*j/3)/(−2*exp(−pi*j/6)) ans = -0.5000i 6.= ---------------------.Solutions to End−of−Chapter Exercises Check with MATLAB: 4*(cos(pi/6)+sin(pi/6)*j)/(5*(cos(−5*pi/6)+sin(−5*pi/6)*j)).0000i ans = -0.8000 .

…. 4. a22. and reference to more advance topics in matrix theory. Numerical Analysis Using MATLAB® and Excel®. a 33. the elements a 11. if a matrix has five rows and five columns. it is said to be a square matrix of order 5. Thus. Thus. Third Edition Copyright © Orchard Publications 4−1 . we say that the matrix elements a 11.Chapter 4 Matrices and Determinants T his chapter is an introduction to matrices and matrix operations. Alternately. and j indicates the column in which each element is positioned. and Gauss’s elimination method are introduced.1 Matrix Definition A matrix is a rectangular array of numbers such as those shown below. Cramer’s rule. 2 3 7 1 –1 5 or 1 3 1 –2 1 –5 4 –7 6 In general form. the matrix is said to be a square matrix of order m (or n ). a nn are called the main diagonal elements. a matrix A is denoted as a 11 a 12 a 13 … a 1n a 21 a 22 a 23 … a 2n A = a 31 a 32 a 33 … a 3n … … … … … a m1 a m2 a m3 … a mn (4. In a square matrix. These are denoted with a dagger ( † ) and may be skipped. Determinants. a 43 indicates the element positioned in the fourth row and third column. Some definitions and examples are not applicable to subsequent material presented in this text. a 33. but are included for subject continuity. are located on the main diagonal. a nn . a 22.1) The numbers a ij are the elements of the matrix where the index i indicates the row. If m = n . …. A matrix of m rows and n columns is said to be of m × n order matrix.

A+B % Define matrices A and B % Add A and B * Henceforth. that is. 2 . if and only if a ij = b ij i = 1. 2. If A = a ij and B = b ij are conformable for addition (subtraction).1 Compute A + B and A – B given that A = 1 2 3 and B = 2 3 0 0 1 4 –1 2 5 Solution: A+B = 1+2 0–1 2+3 1+2 3+0 = 3 5 4+5 –1 3 3 9 and A–B = 1–2 0+1 2 – 3 3 – 0 = –1 –1 3 1–2 4–5 1 –1 –1 Check with MATLAB: A=[1 2 3. B=[2 3 0. where each element of C is the sum (difference) of the corresponding elements of A and B . Third Edition Copyright © Orchard Publications . is called a zero matrix. These are discussed in matrix theory textbooks. −1 2 5]. 3 . that is. …. m j = 1 .3) Example 4.Chapter 4 Matrices and Determinants † The sum of the diagonal elements of a square matrix A is called the trace* of A . n (4. A = B . † A matrix in which every element is zero. 4−2 Numerical Analysis Using MATLAB® and Excel®. if they are of the same order m × n. their sum (difference) will be another matrix C with the same order as A and B . …. all paragraphs and topics preceded by a dagger ( † ) may be skipped.2) Two matrices are said to be conformable for addition (subtraction). 0 1 4]. C = A ± B = [ a ij ± b ij ] (4. 4. 3.2 Matrix Operations Two matrices A = a ij and B = b ij are equal.

is the multiplication of every element of A by k . k2=(−3 + 2*j). A=[1 −2. Example 4. k1*A % Define scalars k1 and k2 % Define matrix A % Multiply matrix A by constant k1 ans = 5 10 k2*A -10 15 %Multiply matrix A by constant k2 Numerical Analysis Using MATLAB® and Excel®. and not [ k ] which is a 1 × 1 matrix. then multiplication of a matrix A by the scalar k . k 2 ⋅ A = ( – 3 + j2 ) × 1 – 2 = ( – 3 + j2 ) × 1 ( – 3 + j2 ) × ( – 2 ) = – 3 + j2 2 3 ( – 3 + j2 ) × 2 ( – 3 + j2 ) × 3 – 6 + j4 6 – j4 – 9 + j6 Check with MATLAB: k1=5. 2 3]. Third Edition Copyright © Orchard Publications 4−3 .2 Multiply the matrix A = 1 –2 2 3 by (a) k 1 = 5 and (b) k 2 = – 3 + j2 Solution: a.Matrix Operations ans = 3 -1 A−B 5 3 3 9 % Subtract B from A ans = -1 1 -1 -1 3 -1 If k is any scalar (a positive or negative number). k 1 ⋅ A = 5 × 1 – 2 = 5 × 1 5 × ( – 2 ) = 5 – 10 2 3 5×2 5×3 10 15 b.

0000i -6.0000i 6. then. Shows that A and B are conformable for multiplication A m×p B p×n Indicates the dimension of the product A ⋅ B For the product B ⋅ A we have: Here. B and A are not conformable for multiplication B p×n A m×p For matrix multiplication. only when the number of columns of matrix A is equal to the number of rows of matrix B .0000i -9. to obtain the product A ⋅ B . the product A ⋅ B (but not B ⋅ A ) is conformable for multiplication only if A is an m × p and matrix B is an p × n matrix.0000+ 4.3 Given that 1 C = 2 3 4 and D = – 1 2 compute the products C ⋅ D and D ⋅ C Solution: The dimensions of matrices C and D are respectively 1 × 3 3 × 1 .0000i Two matrices A and B are said to be conformable for multiplication A ⋅ B in that order. That is.0000+ 2.4. Third Edition Copyright © Orchard Publications . A convenient way to determine if two matrices are conformable for multiplication is to write the dimensions of the two matrices side−by−side as shown below.0000+ 6. Example 4. we add these products. therefore the product C ⋅ D is 4−4 Numerical Analysis Using MATLAB® and Excel®.0000. the operation is row by column. we multiply each element of a row of A by the corresponding element of a column of B . The product A ⋅ B will then be an m × n matrix.Chapter 4 Matrices and Determinants ans = -3. Thus.

and will result in a 1 × 1 . 2]. an equivalent operation exists. However. a 11 a 12 a 13 … a 1n A = 0 a 22 a 23 … a 2n 0 0 … … … … … 0 … … 0 0 0 … a mn (4. D=[1. C*D % Define matrices C and D % Multiply C by D ans = 7 D*C % Multiply D by C ans = 2 -2 4 3 -3 6 4 -4 8 Division of one matrix by another. the product D ⋅ C is also feasible. is not defined.Special Forms of Matrices feasible. that is. Third Edition Copyright © Orchard Publications 4−5 . 1 D ⋅ C = –1 2 3 4 = 2 (1) ⋅ (2) (1) ⋅ (3) (1) ⋅ (4) ( –1 ) ⋅ ( 2 ) ( –1 ) ⋅ ( 3 ) ( –1 ) ⋅ ( 4 ) (2) ⋅ (2) (2) ⋅ (3) (2) ⋅ (4) 2 3 4 = –2 –3 –4 4 6 8 Check with MATLAB: C=[2 3 4]. −1. 1 C ⋅ D = 2 3 4 –1 = ( 2 ) ⋅ ( 1 ) + ( 3 ) ⋅ ( –1 ) + ( 4 ) ⋅ ( 2 ) = 7 2 The dimensions for D and C are respectively 3 × 1 1 × 3 and therefore.4) Numerical Analysis Using MATLAB® and Excel®. and it will become apparent later in this chapter. when we discuss the inverse of a matrix. Multiplication of these will produce a 3 × 3 matrix as follows. The matrix A below is an upper triangular matrix. 4.3 Special Forms of Matrices † A square matrix is said to be upper triangular when all the elements below the diagonal are zero.

6) † A diagonal matrix is called a scalar matrix. when all the elements above the diagonal are zero. The matrix C below is a diagonal matrix. if a 11 = a 22 = a 33 = … = a nn = k where k is a scalar. not all elements below the diagonal need to be non−zero. Shown below are 2 × 2 . 4 D = 0 0 0 0 4 0 0 0 0 4 0 0 0 0 4 (4. For applications.5) In a lower triangular matrix. refer to Chapter 14. if all elements are zero. a 11 B = 0 0 … 0 0 … 0 a 21 a 22 … … … 0 0 … … … … 0 a m1 a m2 a m3 … a mn (4. † A square matrix is said to be diagonal. and 4 × 4 identity matrices.Chapter 4 Matrices and Determinants In an upper triangular matrix. The matrix D below is a scalar matrix with k = 4 . refer to Chapter 14. For example. except those in the diagonal. 4−6 Numerical Analysis Using MATLAB® and Excel®. not all elements above the diagonal need to be non−zero.8) The MATLAB eye(n) function displays an n × n identity matrix. For applications. The matrix B below is a lower triangular matrix. † A square matrix is said to be lower triangular. a 11 0 C = 0 0 0 0 … 0 0 a 22 0 … 0 0 … 0 0 0 0 … 0 0 0 … a mn (4. Third Edition Copyright © Orchard Publications .7) A scalar matrix with k = 1 . is called an identity matrix I . 3 × 3 . 1 0 0 0 1 0 0 0 1 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 1 0 0 1 (4.

For example. −2 1 −5. the eye(size(A)) function. denoted as A . 4 5 6] % Define matrix A A = 1 4 2 5 3 6 Numerical Analysis Using MATLAB® and Excel®. produces an identity matrix whose size is the same as matrix A . is the matrix that is obtained when the rows and columns of matrix A are interchanged. A=[1 2 3.9) In MATLAB we use the apostrophe (′) symbol to denote and obtain the transpose of a matrix. For example. if 1 1 2 3 then A T = 2 4 5 6 3 4 5 6 A= (4. for the above example. Thus.Special Forms of Matrices eye(4)% Display a 4 by 4 identity matrix ans = 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 Likewise. Third Edition Copyright © Orchard Publications 4−7 . let A be defined as A=[1 3 1. eye(size(A)) 3 1 -7 1 -5 6 displays ans = 1 0 0 0 1 0 0 0 1 T † The transpose of a matrix A . 4 −7 6] % Define matrix A A = 1 -2 4 then.

3 2−3j] % Define and display matrix A A = 1.0000 + 2.0000 + 3. is called skew−symmetric. computes the conjugate of a matrix A . computes the complex conjugate of any complex number.2.1. Using MATLAB with the matrix A defined as above.0000 . that is. conj(x). An example of a symmetric matrix is shown below.0000i % Compute and display the conjugate of A conj_A = 1. 1 2 3 2 4 –5 3 –5 6 1 2 3 2 4 –5 = A 3 –5 6 A = A = T (4. is called the conjugate of A .3.0000i 3. The first.0000i 3. 0 2 –3 A = –2 0 –4 3 4 0 0 –2 A = 2 0 –3 –4 T 3 4 = –A 0 4−8 Numerical Analysis Using MATLAB® and Excel®. An example is shown below. Third Edition Copyright © Orchard Publications .0000i T † A square matrix A such that A = – A .0000i 2.0000 . the transpose of a matrix A is the same as A .0000 conj_A=conj(A) 0 + 1.0000i 2. conj(A). A = 1 + j2 3 j 2 – j3 A∗ = 1 – j2 3 –j 2 + j3 † MATLAB has two built−in functions which compute the complex conjugate of a number.Chapter 4 Matrices and Determinants A'% Display the transpose of A ans = 1 2 3 4 5 6 T † A symmetric matrix A . For example. the matrix obtained from A by replacing each element by its conjugate.0000 0 . is one such that A = A .10) † If a matrix A has complex numbers as elements. we obtain A = [1+2j j. and the second. and it is denoted as A∗ .

matrix A above is Hermitian. is called skew−Hermitian. matrix A above is skew symmetric.12) (4. † A square matrix A such that A j A = –1–j –2 1–j 3j j T∗ = – A . 4. –1–j 3j j –2 –j T* j A = 1+j 0 2 –1+j – 3j –j –2 –j = –A 0 j 2 T j A = 1–j 2 0 Therefore.Determinants Therefore. that is.13) Numerical Analysis Using MATLAB® and Excel®. Third Edition Copyright © Orchard Publications 4−9 . For example.4 Determinants Let matrix A be defined as the square matrix a 11 a 12 a 13 … a 1n a 21 a 22 a 23 … a 2n A = a a a …a 31 32 33 3n … … … … … a n1 a n2 a n3 … a nn (4. † A square matrix A such that A 1 A = 1+j 2 1–j 3 –j T∗ = A . A = a 11 a 12 a 21 a 22 (4. the determinant of A .11) then. 1+j 3 j 2 1 T* A = 1–j –j 0 2 1+j 3 j 2 –j = A 0 2 1 T A = 1–j j 0 2 Therefore. is defined as detA = a 11 a 22 a 33 …a nn + a 12 a 23 a 34 …a n 1 + a 13 a 24 a 35 …a n 2 + … – a n 1 …a 22 a 13 … – a n 2 …a 23 a 14 – a n 3 …a 24 a 15 – … The determinant of a square matrix of order n is referred to as determinant of order n . Let A be a determinant of order 2 . matrix A above is skew−Hermitian. is called Hermitian. denoted as detA . For example.

It is highly recommended that this % function file is created in MATLAB's Editor Window. * For an example using this block.4 Given that A = 1 2 and B = 2 – 1 3 4 2 0 compute detA and detB . 3 4]. please refer to Introduction to Simulink with Engineering Applications.14) Example 4.* The MATLAB user−defined function file below can be used to compute the determinant of a 2 × 2 matrix. detA = a 11 a 22 – a 21 a 12 (4. this function is not included in the MATLAB Run−Time Function Library List that is used with the Simulink Embedded MATLAB Function block.m in the current Work Directory. ISBN 0−9744239−7−1. % This file computes the determinant of a 2x2 matrix % It must be saved as function (user defined) file % det2x2. % function y=det2x2(A). Page 16−3. 4−10 Numerical Analysis Using MATLAB® and Excel®. Make sure % that his directory is added to MATLAB's search % path accessed from the Editor Window as File>Set Path> % Add Folder. det(A) ans = -2 det(B) % Compute the determinant of B ans = 2 While MATLAB has the built−in function det(A) for computing the determinant of a matrix A. Third Edition Copyright © Orchard Publications . Solution: detA = 1 ⋅ 4 – 3 ⋅ 2 = 4 – 6 = – 2 detA = 2 ⋅ 0 – 2 ⋅ ( – 1 ) = 0 – ( – 2 ) = 2 % Define matrices A and B % Compute the determinant of A Check with MATLAB: A=[1 2.Chapter 4 Matrices and Determinants Then. B=[2 −1. 2 0].

Example 4. these will be defined shortly.5 Compute detA and detB given that 2 A = 1 2 3 5 0 1 1 0 2 –3 –4 0 –2 0 –5 –6 and B = 1 Solution: Numerical Analysis Using MATLAB® and Excel®.16) A convenient method to evaluate the determinant of order 3. we must first compute the cofactors. is to write the first two columns to the right of the 3 × 3 matrix.16) above..2)*A(2. Third Edition Copyright © Orchard Publications 4−11 .1). and add the products formed by the diagonals from upper left to lower right. Let A be a matrix of order 3.15) then.] and then % type det2x2(A) at the command prompt.2)−A(1. a 11 a 12 a 13 a 11 a 12 a 21 a 22 a 23 a 21 a 22 a 31 a 32 a 33 a 31 a 32 − + This method works only with second and third order determinants. When this is done properly.. % % To run this program. define the 2x2 matrix in % MATLAB's Command Window as A=[. To evaluate higher order determinants. then subtract the products formed by the diagonals from lower left to upper right as shown on the diagram of the next page. that is.Determinants y=A(1..1)*A(2. detA is found from detA = a 11 a 22 a 33 + a 12 a 23 a 31 + a 11 a 22 a 33 – a 11 a 22 a 33 – a 11 a 22 a 33 – a 11 a 22 a 33 (4. a 11 a 12 a 13 A = a 21 a 22 a 23 a 31 a 32 a 33 (4. we obtain (4.

.2)*A(3. % % To run this program.3)*A(2.3)*A(1.1)*A(1.1)+A(1.. det(B) % Define matrix B and compute detB % Define matrix A and compute detA ans = -18 The MATLAB user−defined function file below can be used to compute the determinant of a 3 × 3 matrix. define the 3x3 matrix in % MATLAB's Command Window as A=[. Third Edition Copyright © Orchard Publications . It is highly recommended that this % function file is created in MATLAB's Editor Window. detA = ( 2 × 0 × 0 ) + ( 3 × 1 × 1 ) + ( 5 × 1 × 1 ) – ( 2 × 0 × 5 ) – ( 1 × 1 × 2 ) – ( 0 × 1 × 3 ) = 11 – 2 = 9 2 –3 –4 2 –3 detB = 1 0 – 2 1 – 2 0 –5 –6 2 –6 detB = [ 2 × 0 × ( – 6 ) ] + [ ( – 3 ) × ( – 2 ) × 0 ] + [ ( – 4 ) × 1 × ( – 5 ) ] – [ 0 × 0 × ( – 4 ) ] – [ ( – 5 ) × ( – 2 ) × 2 ] – [ ( – 6 ) × 1 × ( – 3 ) ] = 20 – 38 = – 18 or Check with MATLAB: A=[2 3 5. Make sure % that his directory is added to MATLAB's search % path accessed from the Editor Window as File>Set Path> % Add Folder.3)+A(1.1)*A(3.2)*A(2.m in the current Work Directory.] and then % type det3x3(A) at the command prompt.1)*A(2.1)−A(3. 4−12 Numerical Analysis Using MATLAB® and Excel®. 2 1 0].3)*A(2.3)*A(3. 1 0 1. 0 −5 −6].2)*A(2..3)−A(3. −A(3. y=A(1. 1 0 −2.. % function y=det3x3(A).Chapter 4 Matrices and Determinants 2 detA = 1 2 3 5 2 3 0 1 1 0 1 0 2 1 or Likewise.1)*A(2.2). % This file computes the determinant of a 3x3 matrix % It must be saved as function (user defined) file % det3x3.2).. det(A) ans = 9 B=[2 −3 −4.2)*A(1.

M 23 . The signed minor ( – 1 ) i+j M ij is called the cofactor of a ij and it is denoted as α ij . M 22 . α 12 and α 13 .18) compute the minors M 11 . and it is denoted as M ij .17) If we remove the elements of its ith row.Minors and Cofactors 4. M 33 and cofactors Numerical Analysis Using MATLAB® and Excel®. M 13 and the cofactors α 11 . Third Edition Copyright © Orchard Publications 4−13 . a 11 a 12 a 13 … a 1n a 21 a 22 a 23 … a 2n A = a a a … a 31 32 33 3n … … … … … a n1 a n2 a n3 … a nn (4.5 Minors and Cofactors Let matrix A be defined as the square matrix of order n as shown below. the determinant of the remaining n – 1 square matrix is called the minor of determinant A . Example 4.6 Given that a 11 a 12 a 13 A = a 21 a 22 a 23 a 31 a 32 a 33 (4. M 31 . M 32 . Solution: M 11 = a 22 a 23 a 32 a 33 M 12 . and jth column. M 12 = a 21 a 23 a 31 a 33 M 11 = a 21 a 22 a 31 a 32 and α 11 = ( – 1 ) 1+1 M 11 = M 11 α 12 = ( – 1 ) 1+2 M 12 = – M 12 α 13 = M 13 = ( – 1 ) 1+3 M 13 The remaining minors M 21 .

the value of the determinant of A is the sum of the products obtained by multiplying each element of any row or any column by its cofactor. and α 33 are defined similarly.21) α 22 = ( – 1 ) 2+2 α 23 = ( – 1 ) 2+3 (4.22) α 31 = ( – 1 ) 3+1 2 – 3 = – 8.24) It is useful to remember that the signs of the cofactors follow the pattern + − − + + − − + + − + − − + + − − + + − + − + − + that is. Let A be a square matrix of any size.19) compute its cofactors. 4−14 Numerical Analysis Using MATLAB® and Excel®.23) α 33 = ( – 1 ) 3+3 1 2 = –8 2 –4 (4. α 22. Example 4.7 Given that 1 2 –3 A = 2 –4 2 –1 2 –6 (4. the cofactors on the diagonals have the same sign as their minors. –4 2 α 32 = ( – 1 ) 3+2 1 – 3 = –8 2 2 (4. Solution: α 11 = ( – 1 ) 1+1 – 4 2 = 20 2 –6 1+3 α 12 = ( – 1 ) 1+2 2 2 = 10 –1 –6 2+1 (4. Third Edition Copyright © Orchard Publications . α 23.20) α 13 = ( – 1 ) 2 –4 = 0 –1 2 1 –3 = –9 –1 –6 α 21 = ( – 1 ) 2 –3 = 6 2 –6 1 2 = –4 –1 2 (4.Chapter 4 Matrices and Determinants α 21. α 31. α 32.

A = 1 2 –3 2 –4 2 –1 2 –6 (4. −1 2 −6].9 Compute the value of the determinant Numerical Analysis Using MATLAB® and Excel®.26) +a 31 a 22 a 23 a 24 – a 41 a 22 a 23 a 24 Determinants of order five or higher can be evaluated similarly.8 Compute the determinant of A using the elements of the first row. We must use the above procedure to find the determinant of a matrix A of order 4 or higher. a fourth−order determinant can first be expressed as the sum of the products of the elements of its first row by its cofactor as shown below. Thus. Example 4. a 11 a 12 a 13 a 14 A = a 21 a 22 a 23 a 24 a 31 a 32 a 33 a 34 a 41 a 42 a 43 a 44 a 22 a 23 a 24 a 42 a 43 a 44 a 12 a 13 a 14 a 42 a 43 a 44 a 12 a 13 a 14 = a 11 a 32 a 33 a 34 – a 21 a 32 a 33 a 34 a 42 a 43 a 44 a 12 a 13 a 14 a 32 a 33 a 34 (4.25) Solution: detA = 1 – 4 2 – 2 2 2 – 3 2 – 4 = 1 × 20 – 2 × ( – 10 ) – 3 × 0 = 40 2 –6 –1 –6 –1 2 Check with MATLAB: A=[1 2 −3. 2 −4 2. Third Edition Copyright © Orchard Publications 4−15 .Minors and Cofactors Example 4. det(A) % Define matrix A and compute detA ans = 40 The MATLAB user−defined function file below can be used to compute the determinant of a 4 × 4 matrix.

m in the current Work Directory.Chapter 4 Matrices and Determinants 2 –1 0 A = –1 1 0 4 0 3 –3 0 0 –3 –1 –2 1 (4. [ c ] = 0 . we will multiply each element of the first column by its cofactor. Make sure % that his directory is added to MATLAB's search % path accessed from the Editor Window as File>Set Path> % Add Folder. −3 0 0 1].5 or Example 4.8. delta = det(A) delta = -33 The MATLAB user−defined function file below can be used to compute the determinant of a n × n matrix. % and det3x3(A) for a 3x3 matrix. % The following statement defines the size of the matrix A [n. It is highly recommended that this % function file is created in MATLAB's Editor Window. [ b ] = – 3 . [ d ] = – 36 detA = [ a ] + [ b ] + [ c ] + [ d ] = 6 – 3 + 0 – 36 = – 33 We can verify our answer with MATLAB as follows: A=[ 2 −1 0 −3. 4 0 3 −2. % This file computes the determinant of a nxn matrix % It must be saved as function (user defined) file % detnxn. Third Edition Copyright © Orchard Publications ⎧ ⎪ ⎪ ⎨ ⎪ ⎪ ⎩ ⎧ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎩ . −1 1 0 −1. Then. 1 0 –1 A=2 0 3 – 2 0 0 1 [a] –1 0 –3 –( –1 ) 0 3 –2 0 0 1 [b] –1 0 – 3 +4 1 0 – 1 0 0 1 [c] –1 0 – 3 –( –3 ) 1 0 – 1 0 3 –2 [d] ⎧ ⎪ ⎪ ⎨ ⎪ ⎪ ⎩ ⎧ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎩ Next. % function y=detnxn(A). % The following statement initializes y y=0. using the procedure of Example 4. we find and thus [ a ] = 6 .n]=size(A). % MATLAB allows us to use the user-defined functions to be recursively % called on themselves so we can call det2x2(A) for a 2x2 matrix.27) Solution: Using the above procedure. 4−16 Numerical Analysis Using MATLAB® and Excel®.

i)*detnxn(A(2:n. For example.] and then % type detnxn(A) at the command prompt. Third Edition Copyright © Orchard Publications 4−17 . An example of this is the determinant of the cofactor [ c ] above. Some useful properties of determinants are given below.29) Here. Check with MATLAB: A=[2 4 1..28) then. 1 2 1]. return end % if n==3 y=det3x3(A). detA = 2 3 1 4 6 2 1 1 1 2 3 1 4 6 = 12 + 4 + 6 – 6 – 4 – 12 = 0 2 (4. if 2 A = 3 1 4 6 2 1 1 1 (4. end % % To run this program. return end % For 4x4 or higher order matrices we use the following: % (We can define n and matrix A in Command Window for i=1:n y=y+(−1)^(i+1)*A(1. Property 2: If all the elements of one row or column are m times the corresponding elements of another row or column. the determinant is zero..Minors and Cofactors if n==2 y=det2x2(A).. define the nxn matrix in % MATLAB's Command Window as A=[. the determinant is zero. [1:(i−1) (i+1):n])). Property 1: If all elements of one row or one column are zero. detA is zero because the second column in A is 2 times the first column. det(A) ans = 0 Numerical Analysis Using MATLAB® and Excel®. 3 6 1.

y .30) is a homogeneous set of equations.10 Use Cramer’s rule to find v 1. then. and v 3 if 2v 1 – 5 – v 2 + 3v 3 = 0 – 2v 3 – 3v 2 – 4v 1 = 8 v 2 + 3v 1 – 4 – v 3 = 0 (4. 4. if A = B = C = 0 .Chapter 4 Matrices and Determinants Property 3: If two rows or two columns of a matrix are identical. x = y = z = 0 also. and C. and D 3 are all zero as we found in Property 1 above. 4−18 Numerical Analysis Using MATLAB® and Excel®. for the coefficients of the desired unknown. Then. that is. Third Edition Copyright © Orchard Publications . B. Cramer’s rule applies to systems of two or more equations. This follows from Property 2 with m = 1 .30) and let Δ a 11 a 12 a 13 = a 21 a 22 a 23 a 31 a 32 a 33 D1 = A a 11 a 13 B a 21 a 23 C a 31 a 33 D2 = a 11 A a 13 a 21 B a 23 a 31 C a 33 D3 = a 11 a 12 A a 21 a 22 B a 31 a 32 C Cramer’s rule states that the unknowns x . and z can be found from the relations D1 x = ----- Δ D2 y = ----- Δ D3 z = ----- Δ (4.6 Cramer’s Rule Let us consider the systems of the three equations below a 11 x + a 12 y + a 13 z = A a 21 x + a 22 y + a 23 z = B a 31 x + a 32 y + a 33 z = C (4. Example 4. v 2. If (4. We observe that the numerators of (4. D 1.32) and verify your answers with MATLAB.31) provided that the determinant Δ (delta) is not zero. D 2. the determinant is zero.31) are determinants that are formed from Δ by the substitution of the known values A.

= – -------.= ----. format rat % Express answers in ratio form B=[2 −1 3. % Compute the determinant of D1 d2=[2 5 3. % The elements of D3 detd3=det(d3). % Compute he determinant of D3 Numerical Analysis Using MATLAB® and Excel®. by Cramer’s rule. Δ = 2 –1 3 –4 –3 –2 3 1 –1 5 –1 3 8 –3 –2 4 1 –1 2 –4 3 5 3 8 –2 4 –1 5 8 4 2 –1 – 4 – 3 = 6 + 6 – 12 + 27 + 4 + 4 = 35 3 1 5 –1 8 – 3 = 15 + 8 + 24 + 36 + 10 – 8 = 85 4 1 2 –4 3 5 8 = – 16 – 30 – 48 – 72 + 16 – 20 = – 170 4 D1 = D2 = D3 = 2 –1 –4 –3 3 1 2 –1 – 4 – 3 = – 24 – 24 – 20 + 45 – 16 – 16 = – 55 3 1 Therefore. % The following script will compute and display the values of v1.31) we obtain D1 85 17 x 1 = ----. Third Edition Copyright © Orchard Publications 4−19 . 8 −3 −2.34) We will verify with MATLAB as follows. 3 1 −1]. % The elements of the determinant D delta=det(B). 3 4 −1]. v2 and v3. 3 1 4]. −4 −3 −2.33) Now. −4 8 −2. −4 −3 8. % The elements of D1 detd1=det(d1).Cramer’s Rule Solution: Rearranging the unknowns v . we obtain 2v 1 – v 2 + 3v 3 = 5 – 4v 1 – 3v 2 – 2v 3 = 8 3v 1 + v 2 – v 3 = 4 (4. and transferring known values to the right side. using (4.= ----35 7 Δ D2 170 34 x 2 = ----. % The elements of D2 detd2=det(d2). % Compute the determinant of D2 d3=[2 −1 5.= – ----.= – ----35 7 Δ D3 55 11 x 3 = ----. 4 1 −1].= – ----35 7 Δ (4. % Compute the determinant D of B d1=[5 −1 3.

disp(v3).7 Gaussian Elimination Method We can find the unknowns in a system of two or more equations also by the Gaussian elimination method.disp(v2). v3=detd3/delta. Example 4. Then. disp('v3='). disp('v2='). % disp('v1='). and v 3 of 2v 1 – v 2 + 3v 3 = 5 – 4v 1 – 3v 2 – 2v 3 = 8 3v 1 + v 2 – v 3 = 4 (4. v2=detd2/delta.11 Use the Gaussian elimination method to find v 1. Third Edition Copyright © Orchard Publications . substitution of the two values found can be made into an equation with three unknowns from which we can find the value of the third unknown. the objective is to eliminate one unknown at a time. by substitution to another equation with two unknowns. we can find the second unknown. This procedure is repeated until all unknowns are found. 4−20 Numerical Analysis Using MATLAB® and Excel®.34) 4. % Compute the value of v1 % Compute the value of v2 % Compute the value of v3 % Display the value of v1 % Display the value of v2 % Display the value of v3 v1= 17/7 v2= -34/7 v3= -11/7 These are the same values as in (4. This method is best illustrated with the following example which consists of the same equations as the previous example.Chapter 4 Matrices and Determinants v1=detd1/delta. With this method.35) Solution: As a first step. This can be done by multiplying the terms of any of the equations of the system by a number such that we can add (or subtract) this equation to another equation in the system so that one of the unknowns will be eliminated. we add the first equation of (4. Subsequently.disp(v1). v 2.35) with the third to eliminate the unknown v 2 and we obtain the following equation.

– ----.36) we obtain --------5v 1 + 2 ⋅ ⎛ – 11 ⎞ = 9 or v 1 = 17 ⎝ 7 ⎠ 7 (4. However. is defined as the n square matrix shown on the next page. By substitution of (4. Then the adjoint of A . we can find the last unknown v 2 from any of the three equations of (4.– ----.40) These are the same values as those we found in Example 4.38) Now. we can find the unknown v 1 from either (4. as in Example 4. Third Edition Copyright © Orchard Publications 4−21 . The Gaussian elimination is further discussed in Chapter 14 in conjunction with the LU factorization method. it becomes impractical if the coefficients are large or fractional numbers.8 The Adjoint of a Matrix Let us assume that A is an n square matrix and α ij is the cofactor of a ij .35) by 3.38) into (4. denoted as adjA . Numerical Analysis Using MATLAB® and Excel®.= – ----7 7 7 7 (4. By substitution into the first equation we obtain 34 33 35 34 v 2 = 2v 1 + 3v 3 – 5 = ----.39) Finally. Then.37) Subtraction of (4.36) yields 11 7v 3 = – 11 or v 3 = – ----7 (4. 5v 1 – 5v 3 = 20 (4.11. we multiply the third equation of (4.37).The Adjoint of a Matrix 5v 1 + 2v 3 = 9 (4.36) Next.36) or (4.10.37) from (4. we obtain the following equation.35). The Gaussian elimination method works well if the coefficients of the unknowns are small integers. and we add it with the second to eliminate v 2 . 4.

12 Compute adjA given that 1 2 3 A = 1 3 4 1 4 3 (4. This topic is discussed in Appendix C.Chapter 4 Matrices and Determinants α 11 α 21 α 31 … α n1 α 12 α 22 α 32 … α n2 adjA = α α α … α 13 23 33 n3 … … … … … α 1n α 2n α 3n … α nn (4.13 Given that 4−22 Numerical Analysis Using MATLAB® and Excel®. are the elements of the ith column (row) of adjA . that is.42) Solution: 3 4 adjA = – 1 1 1 1 4 3 4 3 3 4 – 2 4 1 1 3 3 3 3 2 3 3 4 – 2 3 3 4 = –7 6 –1 1 0 –1 1 –2 1 – 1 2 1 4 1 2 1 3 4. if the determinant of that matrix is very small. if detA ≠ 0 . Example 4. Third Edition Copyright © Orchard Publications . A is called non−singular. If an n square matrix A is nearly singular. Example 4. the matrix is said to be ill−conditioned.9 Singular and Non−Singular Matrices An n square matrix A is called singular if detA = 0 .41) We observe that the cofactors of the elements of the ith row (column) of A .

detA = 9 + 8 + 12 – 9 – 16 – 6 = – 2 . Numerical Analysis Using MATLAB® and Excel®. Solution: detA = 1 2 3 2 3 3 4 5 7 1 2 2 3 = 21 + 24 + 30 – 27 – 20 – 28 = 0 3 5 Therefore.The Inverse of a Matrix 1 A = 2 3 2 3 3 4 5 7 (4. we can compute its inverse from the relation A –1 1 = ----------. and since this is a non−zero value.14 Given that 1 2 3 A = 1 3 4 1 4 3 (4.43) determine whether this matrix is singular or non−singular. find A–1 Solution: Here.44). B is called the inverse of A . that is. that is. A is called the inverse of B .45) compute its inverse.adjA detA (4. denoted as B = A–1 . it is possible to compute the inverse of A using (4. 4. A = B –1 If a matrix A is non−singular.10 The Inverse of a Matrix If A and B are n square matrices such that AB = BA = I . matrix A is singular. Third Edition Copyright © Orchard Publications 4−23 . and likewise.12. where I is the identity matrix.44) Example 4. From Example 4.

2 – 3 = 2 –2 4 detA –1 2 and 4−24 Numerical Analysis Using MATLAB® and Excel®. A –1 –7 6 –1 3.5000 -0. 1 4 3].5000 -0.47) Example 4.0000 0 1.5000 -0. invA=inv(A) % Define matrix A and compute its inverse A = 1 1 1 invA = 3.5000 0.0000 0.47) for A = 4 2 3 2 2 –3 –2 4 Proof: detA = 8 – 6 = 2 and adjA = Then.5 1 1 = ----------.5 0 0.Chapter 4 Matrices and Determinants adjA = –7 6 –1 1 0 –1 1 –2 1 Then.5 –2 detA 1 –2 1 – 0.adjA = -.1 0 – 1 = – 0. A –1 1 1 1 –3 ⁄ 2 = ----------.5 1 – 0.adjA = ----.5 – 3 0.15 Prove the validity of (4. that is.5000 Multiplication of a matrix A by its inverse A–1 produces the identity matrix I .5 (4.5000 2 3 4 3 4 3 -3.46) Check with MATLAB: A=[1 2 3. AA –1 = I or A A = I –1 (4. Third Edition Copyright © Orchard Publications . 1 3 4.

16 Given the system of equations ⎧ 2x 1 + 3x 2 + x 3 = 9 ⎫ ⎪ ⎪ ⎨ x 1 + 2x 2 + 3x 3 = 6 ⎬ ⎪ ⎪ ⎩ 3x 1 + x 2 + 2x 3 = 8 ⎭ (4. we can use (4. and x 3 using the inverse matrix method.Solution of Simultaneous Equations with Matrices AA –1 = 4 2 3 1 –3 ⁄ 2 = 4 – 3 2 –1 2 2–2 –6+6 = 1 –3+4 0 0 = I 1 4.48) where A and B are matrices whose elements are known. Solution: In matrix form.49) (4.52) Then. We assume that A and X are conformable for multiplication.48) by A–1 yields: A AX = A B = IX = A B –1 –1 –1 (4. B = 6 2 3 1 2 8 x3 (4.11 Solution of Simultaneous Equations with Matrices Consider the relation AX = B (4. x 2.50) to solve any set of simultaneous equations that have solutions. and X is a matrix (a column vector) whose elements are the unknowns. the given set of equations is AX = B where 2 A= 1 3 x1 3 1 9 . Third Edition Copyright © Orchard Publications 4−25 . We will refer to this method as the inverse matrix method of solution of simultaneous equations.50) or X=A B –1 Therefore. Numerical Analysis Using MATLAB® and Excel®. Multiplication of both sides of (4. Example 4. X = x2 .51) compute the unknowns x 1.

this function is not included in the MATLAB Run−Time Function Library List that is used with the Simulink Embedded MATLAB Function block. we could use the MATLAB inv(A) function. it is easier to use the matrix left division operation X = A \ B . we find the determinant detA . A –1 1 –5 7 1 1 = ----------.94 1 –5 7 9 35 35 ⁄ 18 1 1 X = x 2 = ----. this is MATLAB’s solution of example. For this % Observe that B is a column vector –1 A=[2 3 1.7 1 – 5 6 = ----. where matrix X is the same size as matrix B .53) we obtain the solution as follows. Third Edition Copyright © Orchard Publications . 3 1 2].55) To verify our results. A B for the matrix equation A ⋅ X = B . The MATLAB user− defined function file below can be used to compute the determinant of a 2 × 2 matrix. 1 2 3.28 –5 7 1 8 5 x3 x1 (4.2778 As stated earlier. A userdefined function to compute the inverse of an n × n is presented in Chapter 14. 1 –5 7 detA = 18 and adjA = 7 1 – 5 –5 7 1 Therefore.29 = 29 ⁄ 18 = 1. and the adjoint adjA . 1.54) Next. X = 1. while MATLAB has the built−in function det(A) for computing the determinant of a matrix A.6111 0.61 18 18 5 ⁄ 18 0.Chapter 4 Matrices and Determinants X = A B –1 (4.53) –1 or x1 x2 x3 2 = 1 3 3 1 2 3 1 2 9 6 8 (4. However. and multiply A–1 by B . B=[9 6 8]'. X=A \ B 4−26 Numerical Analysis Using MATLAB® and Excel®.7 1 – 5 18 detA –5 7 1 and by (4.adjA = ----.9444 1.

V = 0 0 – 9 15 0 I1 and I = I 2 I3 The next step is to find R–1 . and I 3 . the matrix equation is RI = V or I = R–1 V . R –1 219 135 81 1 1 = ----------. Third Edition Copyright © Orchard Publications 4−27 .57) Therefore. where 10 – 9 0 100 R = – 9 20 – 9 . I 2. we find the determinant and the adjoint of R .Solution of Simultaneous Equations with Matrices Example 4. For this example.adjR = -------.1.17 For the electric circuit of Figure 4. Circuit for Example 4. the mesh equations are 1Ω 2Ω 9Ω 2Ω 9Ω 4Ω − I 1 V = 100 v + I2 I3 Figure 4.56) Use the inverse matrix method to compute the values of the currents I 1. adjR = 135 150 90 81 90 119 (4.135 150 90 975 detR 81 90 119 and Numerical Analysis Using MATLAB® and Excel®.17 10I 1 – 9I 2 = 100 0 0 – 9I 1 + 20I 2 – 9I 3 = – 9I 2 + 15I 3 = (4.adjR detR (4. Solution: For this example. we find that 219 135 81 detR = 975.58) Then. This is found from the relation R –1 1 = ----------.1.

V=[100 0 0]'. A B C D E F G H 1 Spreadsheet for Matrix Inversion and Matrix Multiplication 2 3 10 -9 0 100 4 R= -9 20 -9 V= 0 5 0 -9 15 0 6 7 0. R(2.1)=10. R(3. R(3. The procedure is as follows: 1. the above script could also have been written as: R(1.85 975 975 81 90 119 0 81 8. Third Edition Copyright © Orchard Publications .846 8 0.46 1 100 I = I 2 = -------.138 0. Accordingly. Solution of Example 4.135 150 90 0 = -------.2)=20.2.2)=−9.225 0.3)=15. R(2. I=R\V I = 22. I=R\V I = 22.3) since it is zero.Chapter 4 Matrices and Determinants 219 135 81 100 219 22. % No need to make entry for A(1.154 0.8462 8.462 -1 R = 0.122 8.17 with a spreadsheet 4−28 Numerical Analysis Using MATLAB® and Excel®. We start with a blank spreadsheet and in a block of cells. 0 −9 15].2. Then. V=[100 0 0]'. we can use Microsoft Excel’s MINVERSE (Matrix Inversion) and MMULT (Matrix Multiplication) functions. say B3:D5. to obtain the values of the three currents in Example 4.092 I= 13.083 22.1)=−9.138 0. R(1. For instance. we enter the elements of matrix R as shown in Figure 4. −9 20 −9.31 I3 I1 Check with MATLAB: R=[10 −9 0.3)=−9.4615 13.135 = 13.4615 13.083 0. R(2. we enter the elements of matrix V in G3:G5.3077 Spreadsheets also have the capability of solving simultaneous equations using the inverse matrix method.2)=−9.092 0.17.3077 9 10 Figure 4.8462 8.3077 We can also use subscripts to address the individual elements of the matrix.

+ -----------------.59) and the voltages V 1 and V 2 can be computed from the nodal equations V 1 – 170 ∠0° V 1 – V 2 V 1 – 0 ------------------------------.3. we highlight them.60) (4.+ -------------. As before. Example 4.= 0 85 100 j200 V 2 – 170 ∠0° V 2 – V 1 V 2 – 0 ------------------------------.+ -------------. Circuit for Example 4.3. that is. We format this block for number display with three decimal places. The values of I then appear in G7:G9. Next. we compute and display the inverse of R. and express the current I x in both rectangular and polar forms by first simplifying like Numerical Analysis Using MATLAB® and Excel®. the current I X can be found from the relation VS + 85 Ω V1 j200 Ω R1 C IX R2 −j100 Ω V2 50 Ω − 170∠0° R3 = 100 Ω L Figure 4.= 0 – j100 100 50 (4. Third Edition Copyright © Orchard Publications 4−29 .+ -----------------. With this range highlighted and making sure that the cell marker is in B7.61) Compute. we type the formula =MININVERSE(B3:D5) and we press the Crtl−Shift−Enter keys simultaneously. we type the formula =MMULT(B7:D9. We observe that R–1 appears in these cells.18 V1 – V2 I X = -----------------R3 (4.18 For the phasor circuit of Figure 4. Now. and with the cell marker positioned in G7. We choose B7:D9 for the elements of this inverted matrix. 3.G3:G5) and we press the Crtl−Shift−Enter keys simultaneously. R–1 . we choose the block of cells G7:G9 for the values of the current I .Solution of Simultaneous Equations with Matrices 2.

63). we find I X from R3=100.61).01V 2 = 2 – 0. 1. we obtain ( 0. V=Y\I.005 – 0. % Display V1 and V2 with dash line underneath fprintf('%9. For the polar form we use magIX=abs(IX) magIX = % Compute the magnitude of IX 4−30 ⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ ⎧ ⎨ ⎩ Numerical Analysis Using MATLAB® and Excel®. and to do all other computations.03 + j0.01 V 1 + ( 0. % Insert a line disp(' V1 V2'). IX=(V(1)−V(2))/R3 % Compute the value of IX IX = 0.416 Next.905 53.01j].V(1).62) in matrix form as 0.60) and (4. we write (4.01. The script is shown below.3f %9.01 0. and find V fprintf('\n').0218 – j0.7 Next. and then writing the above relations in matrix form as YV = I .01 0. Third Edition Copyright © Orchard Publications . Y=[0.03 + j0.7 ⎧ ⎨ ⎩ I (4.Chapter 4 Matrices and Determinants terms.005 )V 1 – 0. I=[2. Simplifying and rearranging the nodal equations of (4. % Define Y.3f\n'.62) where the matrices Y . and I = current .01 )V 2 = j1. where Y = admit tan ce . disp(' ------------------').01 – 0.03+0. collecting. −0.01 Y V1 V2 V = 2 j1. I.0218 – j0. V .0.V(2)) % Display values of V1 and V2 in tabular form fprintf('\n')% Insert another line V1 V2 -----------------104. We will use MATLAB to compute the voltages V 1 and V 2 . Solution: The elements of the Y matrix are the coefficients of V 1 and V 2 .0218−0.005j −0. V = voltage .5149 . and I are as indicatedin (4.7j].0590i and this is the rectangular form of I X .

5326 Therefore. Third Edition Copyright © Orchard Publications 4−31 . Numerical Analysis Using MATLAB® and Excel®.Solution of Simultaneous Equations with Matrices 0. in polar form I X = 0.5183 thetaIX=angle(IX)*180/pi % Compute angle theta in degrees thetaIX = -6.53° Spreadsheets have limited capabilities with complex numbers. and thus we cannot use them to compute matrices that include complex numbers in their elements.518 ∠– 6.

The product A ⋅ B will then be an m × n matrix. then. …. 2 . n • Two matrices are said to be conformable for addition (subtraction). • Two matrices A = a ij and B = b ij are equal. C = A ± B = [ a ij ± b ij ] • If k is any scalar (a positive or negative number). • Two matrices A and B are said to be conformable for multiplication A ⋅ B in that order. the product A ⋅ B (but not B ⋅ A ) is conformable for multiplication only if A is an m × p and matrix B is an p × n matrix. 2. Thus. their sum (dif- ference) will be another matrix C with the same order as A and B . • For matrix multiplication. is the multiplication of every element of A by k . A matrix of m rows and n columns is said to be of m × n order matrix. that is. i. and j indicates the column in which each element is positioned. If m = n .Chapter 4 Matrices and Determinants 4. 3. and not [ k ] which is a 1 × 1 matrix. the matrix is said to be a square matrix of order m. 3 . If A = a ij and B = b ij are conformable for addition (subtraction).. then multiplication of a matrix A by the scalar k . Third Edition Copyright © Orchard Publications . the operation is row by column. if and only if a ij = b ij i = 1. to obtain the product A ⋅ B .12 Summary • A matrix is a rectangular array of numbers whose general form is a 11 a 12 a 13 … a 1n a 21 a 22 a 23 … a 2n A = a 31 a 32 a 33 … a 3n … … … … … a m1 a m2 a m3 … a mn The numbers a ij are the elements of the matrix where the index i indicates the row. A = B . • Division of one matrix by another. where each element of C is the sum (difference) of the corresponding elements of A and B .e. 4−32 Numerical Analysis Using MATLAB® and Excel®. is not defined. only when the number of columns of matrix A is equal to the number of rows of matrix B . we multiply each element of a row of A by the corresponding element of a column of B . we add these products. That is. …. if they are of the same order m × n . m j = 1 .

Numerical Analysis Using MATLAB® and Excel®. the value of the determinant of A is the sum of the products obtained by multiplying each element of any row or any column by its cofactor. and it is denoted as M ij . is called an identity matrix I . • Let A be a square matrix of any size. and jth column. denoted as A . is the matrix that is obtained when the rows and columns of matrix A are interchanged. A scalar matrix with k = 1 . • The transpose of a matrix A . We must use this procedure to find the determinant of a matrix A of order 4 or higher. the determinant of the remaining n – 1 square matrix is called the minor of determinant A . Third Edition Copyright © Orchard Publications 4−33 . • The MATLAB eye(n) function displays an n × n identity matrix and the eye(size(A)) function displays an identity matrix whose size is the same as matrix A . • The determinant of a square matrix A where a 11 a 12 a 13 … a 1n a 21 a 22 a 23 … a 2n A = a a a …a 31 32 33 3n … … … … … a n1 a n2 a n3 … a nn T is denoted as detA and it is defined as detA = a 11 a 22 a 33 …a nn + a 12 a 23 a 34 …a n 1 + a 13 a 24 a 35 …a n 2 + … – a n 1 …a 22 a 13 … – a n 2 …a 23 a 14 – a n 3 …a 24 a 15 – … • If from a matrix A be defined as a 11 a 12 a 13 … a 1n a 21 a 22 a 23 … a 2n A = a a a …a 31 32 33 3n … … … … … a n1 a n2 a n3 … a nn we remove the elements of its ith row. • The signed minor ( – 1 ) i+j M ij is called the cofactor of a ij and it is denoted as α ij .Summary • A scalar matrix is a square matrix where a 11 = a 22 = a 33 = … = a nn = k and k is a scalar.

the determinant is zero. the determinant is zero. is defined as the n square matrix below. c. y . the adjoint of A . denoted as adjA . This procedure is repeated until all unknowns are found. • We can find the unknowns in a system of two or more equations also by the Gaussian elimination method. Subsequently. and z can be found from the relations D1 x = ----- Δ D2 y = ----- Δ D3 z = ----- Δ provided that the determinant Δ (delta) is not zero. With this method. the determinant is zero. This can be done by multiplying the terms of any of the equations of the system by a number such that we can add (or subtract) this equation to another equation in the system so that one of the unknowns will be eliminated. b. Third Edition Copyright © Orchard Publications . • Cramer’s rule states that if a system of equations is defined as a 11 x + a 12 y + a 13 z = A a 21 x + a 22 y + a 23 z = B a 31 x + a 32 y + a 33 z = C and we let Δ a 11 a 12 a 13 = a 21 a 22 a 23 a 31 a 32 a 33 D1 = A a 11 a 13 B a 21 a 23 C a 31 a 33 D2 = a 11 A a 13 a 21 B a 23 a 31 C a 33 D3 = a 11 a 12 A a 21 a 22 B a 31 a 32 C the unknowns x . If two rows or two columns of a matrix are identical. Then. 4−34 Numerical Analysis Using MATLAB® and Excel®. If all elements of one row or one column are zero.Chapter 4 Matrices and Determinants • Some useful properties of determinants are: a. substitution of the two values found can be made into an equation with three unknowns from which we can find the value of the third unknown. we can find the second unknown. by substitution to another equation with two unknowns. If all the elements of one row or column are m times the corresponding elements of another row or column. • If A is an n square matrix and α ij is the cofactor of a ij . the objective is to eliminate one unknown at a time.

Numerical Analysis Using MATLAB® and Excel®. We refer to this method as the inverse matrix method of solution of simultaneous equations. we can use the relation X=A –1 B to solve any set of simultaneous equations that have solutions.Summary α 11 α 21 α 31 … α n1 α 12 α 22 α 32 … α n2 adjA = α α α … α 13 23 33 n3 … … … … … α 1n α 2n α 3n … α nn • An n square matrix A is called singular if detA = 0 . that is. where I is the identity matrix. –1 • We can use Microsoft Excel’s MINVERSE (Matrix Inversion) and MMULT (Matrix Multiplication) functions. A is called the inverse of B . to solve any set of simultaneous equations that have solutions. where matrix X is the same size as matrix B . • If A and B are n square matrices such that AB = BA = I . this is MATLAB’s solution of A B for the matrix equation A ⋅ X = B . denoted as B = A –1 . A = B –1 • If a matrix A is non-singular.adjA detA • Multiplication of a matrix A by its inverse A –1 produces the identity matrix I . X is a matrix (a column vector) whose elements are the unknowns and A and X are conformable for multiplication. B is called the inverse of A . we can compute its inverse from the relation A –1 1 = ----------. A is called non-singular. that is. and likewise. • The matrix left division operation is defined as X = A \ B . if detA ≠ 0 . However. AA –1 = I or A A = I –1 • If A and B are matrices whose elements are known. Third Edition Copyright © Orchard Publications 4−35 . we cannot use them to compute matrices that include complex numbers in their elements.

A ⋅ C c. detB c. Third Edition Copyright © Orchard Publications . C + D e. Solve the following system of equations using Cramer’s rule.Chapter 4 Matrices and Determinants 4. – x 1 + 2x 2 – 3x 3 + 5x 4 = 14 x 1 + 3x 2 + 2x 3 – x 4 = 9 3x 1 – 3 x 2 + 2x 3 + 4x 4 = 19 4x 1 + 2x 2 + 5x 3 + x 4 = 27 7. A + C c. A + B b. Perform the following computations. detA b. C ⋅ D e. det ( A ⋅ C ) 4. if possible. C ⋅ A g. Verify your answers with Excel or MATLAB. A – C g. a. if possible.13 Exercises For Exercises 1 through 3 below. Verify your answers with Excel or MATLAB. if possible. B – D h. · a. x 1 – 2x 2 + x 3 = – 4 – 2x 1 + 3x 2 + x 3 = 9 3x 1 + 4x 2 – 5x 3 = 0 5. 6. B ⋅ A f. detD e. B . Solve the following system of equations using the inverse matrix method. Use the MATLAB det(A) function to find the unknowns of the system of equations below. C – D 2. x1 –3 1 3 4 3 1 –2 ⋅ x2 = –2 0 2 3 5 x3 4−36 Numerical Analysis Using MATLAB® and Excel®. D ⋅ A h. A – B f. D ⋅ C 3. Verify your answers with Excel or MATLAB. detC d. Perform the following computations. the matrices A . Verify your answers with Excel or MATLAB. Verify your answers with Excel or MATLAB. a. B ⋅ D d. Perform the following computations. A ⋅ B b. C and D are defined as: 1 –1 –4 A = 5 7 –2 3 –5 6 5 9 –3 B = –2 8 2 7 –4 6 4 6 C= – 3 8 5 –2 D = 1 –2 3 –3 6 –4 1. B + D d. det ( A ⋅ B ) f. Repeat Exercise 4 using the Gaussian elimination method.

Numerical Analysis Using MATLAB® and Excel®.Exercises 8. Third Edition Copyright © Orchard Publications 4−37 . Use Excel to find the unknowns for the system 2 2 –1 2 4 3 –4 1 3 –4 –2 2 x1 1 –2 x2 10 3 ⋅ = x3 – 14 2 7 1 x4 Verify your answers with the MATLAB left division operation.

a. Third Edition Copyright © Orchard Publications . C + D not conformable for addition e. B – D not conformable for subtraction h. A + C not conformable for addition c. A ⋅ C = 17 109 -37 -29 -13 17 1 × 6 + ( –1 ) × 8 + ( –4 ) × ( –2 ) 5 × 6 + 7 × 8 + ( –2 ) × ( –2 ) 3 × 6 + ( –5 ) × 8 + 6 × ( –2 ) 1 × 4 + ( –1 ) × ( –3 ) + ( –4 ) × 5 5 × 4 + 7 × ( –3 ) + ( –2 ) × 5 3 × 4 + ( –5 ) × ( –3 ) + 6 × 5 – 13 6 = – 11 90 57 – 34 c. 3 −5 6]. C – D not conformable for subtraction 2. A + B = 5 – 2 6 8 –7 1+5 –1+9 –4–3 = 3 15 0 7+8 –2+2 10 – 9 12 3+7 –5–4 6+6 b. A⋅B = 1 × 5 + ( –1 ) × ( –2 ) + ( –4 ) × 7 5 × 5 + 7 × ( –2 ) + ( –2 ) × 7 3 × 5 + ( –5 ) × ( –2 ) + 6 × 7 – 21 17 – 29 – 3 109 – 13 67 – 37 17 1 × 9 + ( –1 ) × 8 + ( –4 ) × ( –4 ) 5 × 9 + 7 × 8 + ( –2 ) × ( –4 ) 3 × 9 + ( –5 ) × 8 + 6 × ( –4 ) 1 × ( –3 ) + ( –1 ) × 2 + ( –4 ) × 6 5 × ( –3 ) + 7 × 2 + ( –2 ) × 6 3 × ( –3 ) + ( –5 ) × 2 + 6 × 6 a. −2 8 2. 7 −4 6]. ⋅ D = ( – 3 ) × 1 + 8 × ( – 3 ) 4 × ( –2 ) + 6 × 6 4 × 3 + 6 × ( –4 ) ( –3 ) × ( –2 ) + 8 × 6 ( –3 ) × 3 + 8 × ( –4 ) 5 × 1 + ( –2 ) × ( –3 ) 5 × ( –2 ) + ( –2 ) × 6 5 × 3 + ( –2 ) × ( –4 ) 4 × 1 + 6 × ( –3 ) – 14 28 – 12 = – 27 54 – 41 11 – 22 23 4−38 Numerical Analysis Using MATLAB® and Excel®. A – B = 5 + 2 3–7 1–5 –1–9 7–8 –5+4 – 4 – 10 – 1 –4+3 = 7 –1 –4 –2–2 –4 –1 0 6–6 f.Chapter 4 Matrices and Determinants 4. = Check with MATLAB: A=[1 −1 −4. B + D not conformable for addition d.14 Solutions to End−of−Chapter Exercises 1. B ⋅ D not conformable for multiplication d. A*B ans = -21 -3 67 b. 5 7 −2. A – C not conformable for subtraction g. B=[5 9 −3.

Third Edition Copyright © Orchard Publications 4−39 . det ( A ⋅ C ) does not exist because detC does not exist Numerical Analysis Using MATLAB® and Excel®.Solutions to End−of−Chapter Exercises 5 × 1 + 9 × 5 + ( –3 ) × 3 B ⋅ A = 5 × ( –1 ) + 9 × 7 + ( –3 ) × ( –5 ) 5 × ( –4 ) + 9 × ( –2 ) + ( –3 ) × 6 41 73 – 56 = 44 48 4 5 – 65 16 ( –2 ) × 1 + 8 × 5 + 2 × 3 ( –2 ) × ( –1 ) + 8 × 7 + 2 × ( –5 ) ( –2 ) × ( –4 ) + 8 × ( –2 ) + 2 × 6 7 × 1 + ( –4 ) × 5 + 6 × 3 7 × ( –1 ) + ( –4 ) × 7 + 6 × ( –5 ) 7 × ( –4 ) + ( –4 ) × ( –2 ) + 6 × 6 e. detC does not exist. = h. matrix must be square d. a. det ( A ⋅ B ) = 252 × 658 = 165816 f. detD does not exist. f. 1 –1 –4 1 –1 detA = 5 7 – 2 5 7 3 –5 6 3 –5 = 1 × 7 × 6 + ( –1 ) × ( –2 ) × 3 + ( –4 ) × 5 × ( –5 ) – [ 3 × 7 × ( –4 ) + ( –5 ) × ( –2 ) × 1 + 6 × 5 × ( –1 ) ] = 42 + 6 + 100 – ( – 84 ) – 10 – ( – 30 ) = 252 5 9 –3 5 9 detB = – 2 8 2 – 2 8 7 –4 6 7 –4 = 5 × 8 × 6 + 9 × 2 × 7 + ( –3 ) × ( –2 ) × ( –4 ) – [ 7 × 8 × ( –3 ) + ( –4 ) × 2 × 5 + 6 × ( –2 ) × 9 ] = 240 + 126 – 24 – ( – 168 ) + 40 – ( – 108 ) = 658 b. D ⋅ C = 3. matrix must be square e. c. C ⋅ A not conformable for multiplication D⋅A = 1 × 1 + ( –2 ) × 5 + 3 × 3 1 × ( –1 ) + ( –2 ) × 7 + 3 × ( –5 ) 1 × ( –4 ) + ( –2 ) × ( –2 ) + 3 × 6 ( –3 ) × 1 + 6 × 5 + ( –4 ) × 3 ( –3 ) × ( –1 ) + 6 × 7 + ( –4 ) × ( –5 ) ( –3 ) × ( –4 ) + 6 × ( –2 ) + ( –4 ) × 6 0 – 30 18 15 65 – 24 1 × 4 + ( –2 ) × ( –3 ) + 3 × 5 ( –3 ) × 4 + 6 × ( –3 ) + ( –4 ) × 5 1 × 6 + ( –2 ) × 8 + 3 × ( –2 ) = 25 – 16 ( –3 ) × 6 + 6 × 8 + ( –4 ) × ( –2 ) – 50 38 g. et ( A ⋅ B ) = detA ⋅ detBand from parts (a) and (b).

= -------. 1 –2 1 1 –2 Δ = –2 3 1 –2 3 3 4 –5 3 4 = 1 × 3 × ( –5 ) + ( –2 ) × 1 × 3 + 1 × ( –2 ) × 4 – [ 3 × 3 × 1 + 4 × 1 × 1 + ( –5 ) × ( –2 ) × ( –2 ) ] = – 15 – 6 – 8 – 9 – 4 + 20 = – 22 –4 –2 1 4 –2 9 3 1 9 3 0 4 –5 0 4 D1 = = –4 × 3 × ( –5 ) + ( – 2 ) × 1 × 0 + 1 × 9 × 4 – [ 0 × 3 × 1 + 4 × 1 × 4 + ( –5 ) × 9 × ( –2 ) ] = 60 + 0 + 36 – 0 + 16 – 90 = 22 1 –4 1 1 –4 D2 = –2 9 1 –2 9 3 0 –5 3 0 = 1 × 9 × ( –5 ) + ( –4 ) × 1 × 3 + 1 × ( –2 ) × 0 – [ 3 × 9 × 1 + 0 × 1 × 1 + ( –5 ) × ( –2 ) × ( –4 ) ] = – 45 – 12 – 0 – 27 – 0 + 40 = – 44 1 –2 –4 1 –2 D3 = –2 3 9 –2 3 3 4 0 3 4 = 1 × 3 × 0 + ( –2 ) × 9 × 3 + ( –4 ) × ( –2 ) × 4 – [ 3 × 3 × ( –4 ) + 4 × 9 × 1 + 0 × ( –2 ) × ( –2 ) ] = 0 – 54 + 32 + 36 – 36 – 0 = – 22 D1 22 x 1 = -----.= -------.= 2 Δ – 22 D3 – 22 x 3 = -----. x 1 – 2x 2 + x 3 = – 4 (1) – 2x 1 + 3x 2 + x 3 = 9 (2) 3x 1 + 4x 2 – 5x 3 = 0 (3) Multiplication of (1) by 2 yields 2x 1 – 4x 2 + 2x 3 = – 8 (4) Addition of (2) and (4) yields – x 2 + 3x 3 = 1 (5) 4−40 Numerical Analysis Using MATLAB® and Excel®.= 1 Δ – 22 5.= – 1 Δ – 22 D2 – 44 x 2 = -----.Chapter 4 Matrices and Determinants 4.= -------. Third Edition Copyright © Orchard Publications .

x4=det(D4)/det(Delta) x1=1 x2=2 x3=3 x4=4 Numerical Analysis Using MATLAB® and Excel®. 1 9 2 −1. 3 −3 2 4... 4 2 27 1]. 4 2 5 1]. 9 3 2 −1. D3=[−1 2 14 5. 3 −3 2 19. x1=det(D1)/det(Delta). x3=det(D3)/det(Delta). Delta=[−1 2 −3 5. 4 2 5 27].Solutions to End−of−Chapter Exercises Multiplication of (1) by – 3 yields – 3 x 1 + 6x 2 – 3x 3 = 12 (6) Addition of (3) and (6) yields 10x 2 – 8x 3 = 12 (7) Multiplication of (5) by 10 yields – 10x 2 + 30x 3 = 10 (8) Addition of (7) and (8) yields 22x 3 = 22 (9) or x 3 = 1 (10) Substitution of (10) into (7) yields 10x 2 – 8 = 12 (11) or x 2 = 2 (12) and substitution of (10) and (12) into (1) yields x 1 – 4 + 1 = – 4 (13) or x 1 = – 1 (14) 6. 3 −3 19 4. 1 3 2 9. D1=[14 2 −3 5. 1 3 9 −1. 1 3 2 −1. D4=[−1 2 −3 14. 27 2 5 1]. 19 −3 2 4.. 4 27 5 1]. 3 19 2 4. D2=[−1 14 −3 5. Third Edition Copyright © Orchard Publications 4−41 . x2=det(D2)/det(Delta).

50 1.⋅ – 19 – 3 14 = 19 ⁄ 18 3 ⁄ 18 – 14 ⁄ 18 detA – 18 – 7 ⁄ 18 – 3 ⁄ 18 8 ⁄ 18 7 3 –8 – 11 ⁄ 18 3 ⁄ 18 10 ⁄ 18 – 3 19 ⁄ 18 3 ⁄ 18 – 14 ⁄ 18 – 2 = – 7 ⁄ 18 – 3 ⁄ 18 8 ⁄ 18 0 33 ⁄ 18 – 6 ⁄ 18 + 0 – 57 ⁄ 18 – 6 ⁄ 18 + 0 = 21 ⁄ 18 + 6 ⁄ 18 + 0 27 ⁄ 18 – 63 ⁄ 18 = 27 ⁄ 18 A x1 X = x2 = x3 1.50 – 3. 1 3 4 1 3 detA = 3 1 – 2 3 1 2 3 5 2 3 = 1 × 1 × 5 + 3 × ( –2 ) × 2 + 4 × 3 × 3 – [ 2 × 1 × 4 + 3 × ( – 2 ) × 1 + 5 × 3 × 3 ] = 5 – 12 + 36 – 8 + 6 – 45 = – 18 11 – 3 – 10 adjA = – 19 – 3 14 7 3 –8 –1 11 – 3 – 10 – 11 ⁄ 18 3 ⁄ 18 10 ⁄ 18 1 1 = ----------.50 4−42 Numerical Analysis Using MATLAB® and Excel®.Chapter 4 Matrices and Determinants 7. Third Edition Copyright © Orchard Publications .⋅ adjA = -------.

0000 Numerical Analysis Using MATLAB® and Excel®. B=[1 10 −14 7]'. −1 3 −4 2. Third Edition Copyright © Orchard Publications 4−43 .Solutions to End−of−Chapter Exercises 8. A=[2 4 3 −2. A\B ans = -11.0000 9.5000 12. 2 −4 1 3.5000 1. 2 −2 2 1].

1 Simple Differential Equations In this section we present two simple examples to show the importance of differential equations in engineering applications. State Variables.3) we obtain v C ( t ) = It + k where k represents the constants of integration of both sides.2) By substitution of (5. Solution: It is given that the current.1) where i C ( t ) is the current through the capacitor. 5. Numerical Analysis Using MATLAB® and Excel®.Chapter 5 Differential Equations. and State Equations T his chapter is a review of ordinary differential equations and an introduction to state variables and state equations.3) (5. Example 5. Solutions of differential equations with numerical methods is discussed in Chapter 9. and the constant C is the capacitance in farads (F).2) into (5. Third Edition Copyright © Orchard Publications 5−1 . For this example C = 1 F and the capacitor is being charged by a constant current I .4) Integrating both sides of (5. dv C = Idt (5. is constant.1) we obtain dv C -------. i C ( t ) = I = cons tan t (5.= I dt and by separation of the variables. Find the voltage v C across this capacitor as a function of time given that the voltage at some reference time t = 0 is V 0 . v C ( t ) is the voltage across the capacitor. that is.1 The current and voltage in a capacitor are related by dv C i C ( t ) = C -------dt (5. as a function of time.

**Chapter 5 Differential Equations, State Variables, and State Equations
**

We can find the value of the constant k by making use of the initial condition, i.e., at t = 0 , v C = V 0 and (5.4) then becomes

V0 = 0 + k

(5.5) (5.6)

**or k = V 0 , and by substitution into (5.4),
**

v C ( t ) = It + V 0

This example shows that when a capacitor is charged with a constant current, a linear voltage is produced across the terminals of the capacitor. Example 5.2 Find the current i L ( t ) through an inductor whose slope at the coordinate ( t, i L ) is cos t and the current i L passes through the point ( π ⁄ 2 ,1 ) . Solution: We are given that

di L ------- = cos t dt

(5.7) (5.8) (5.9)

**By separating the variables we obtain
**

di L = cos tdt

**and integrating both sides we obtain
**

i L ( t ) = sin t + k

where k represents the constants of integration of both sides. We find the value of the constant k by making use of the initial condition. For this example, ω = 1 and thus at ωt = t = π ⁄ 2 , i L = 1 . With these values (5.9) becomes

-1 = sin π + k 2

(5.10)

**or k = 0 , and by substitution into (5.9),
**

i L ( t ) = sin t

(5.11)

5.2 Classification

Differential equations are classified by: 1. Type - Ordinary or Partial

5−2

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

Classification

2. Order - The highest order derivative which is included in the differential equation 3. Degree - The exponent of the highest power of the highest order derivative after the differential equation has been cleared of any fractions or radicals in the dependent variable and its derivatives For example, the differential equation

2 ⎛ d 4 y⎞ ⎛ d 3 y⎞ ⎛ d 2 y⎞ dy 8 y ------- ⎟ + 5 ⎜ ------- ⎟ + 6 ⎜ ------- ⎟ + 3 ⎛ ----- ⎞ + -------------- = ye – 2x ⎜ 4 ⎝ dx⎠ 3 ⎝ dx ⎠ ⎝ dx 3⎠ ⎝ dx 2⎠ x +1 2 4 6

is an ordinary differential equation of order 4 and degree 2 . If the dependent variable y is a function of only a single variable x , that is, if y = f ( x) , the differential equation which relates y and x is said to be an ordinary differential equation and it is abbreviated as ODE. The differential equation

2 dy d y ------- + 3 ----- + 2 = 5 cos 4t 2 dt dt

**is an ODE with constant coefficients. The differential equation
**

dy d y 2 2 x 2 ------- + x ----- + ( x – n ) = 0 2 dt dt

2

is an ODE with variable coefficients. If the dependent variable y is a function of two or more variables such as y = f ( x, t ) , where x and t are independent variables, the differential equation that relates y , x , and t is said to be a partial differential equation and it is abbreviated as PDE. An example of a partial differential equation is the well-known one-dimensional wave equation shown below.

2 2∂ y ∂ y ------- = a ------2 2 ∂x ∂t 2

Most engineering problems are solved with ordinary differential equations with constant coefficients; however, partial differential equations provide often quick solutions to some practical applications as illustrated with the following three examples.

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

5−3

**Chapter 5 Differential Equations, State Variables, and State Equations
**

Example 5.3 The equivalent resistance R T of three resistors R 1 , R 2 , and R 3 in parallel is obtained from

1 1 1 1 ------ = ----- + ----- + ----RT R1 R2 R3

Given that initially R 1 = 5 Ω , R 2 = 20 Ω , and R 3 = 4 Ω , compute the change in R T if R 2 is increased by 10 % and R 3 is decreased by 5 % while R 1 does not change. Solution: The initial value of the equivalent resistance is R T = 5 || 20 || 4 = 2 Ω . Now, we treat R 2 and R 3 as constants and differentiating R T with respect to R 1 we obtain

RT 2 ∂R T 1 1 ∂R T – ------ --------- = – ----- or --------- = ⎛ ------ ⎞ 2 2 ∂R 1 ⎝ R 1 ⎠ R1 R T ∂R 1

Similarly,

RT 2 ∂R T ⎛ R T ⎞ 2 ∂R T --------- = -----and --------- = ⎛ ------ ⎞ ∂R 2 ⎝ R 2 ⎠ ∂R 3 ⎝ R 3 ⎠

**and the total differential dRT is
**

RT 2 ∂R T ∂R T ∂R T RT 2 RT 2 R T = --------- dR 1 + --------- dR 2 + --------- dR 3 = ⎛ ------ ⎞ dR 1 + ⎛ ------ ⎞ dR 2 + ⎛ ------ ⎞ dR ⎝ R1 ⎠ ⎝ R2 ⎠ ⎝ R3 ⎠ ∂R 1 ∂R 2 ∂R 3

**By substitution of the given numerical values we obtain
**

2- 2 2 2 2 2 dR T = ⎛ -- ⎞ ( 0 ) + ⎛ ----- ⎞ ( 2 ) + ⎛ -- ⎞ ( – 0.2 ) = 0.02 – 0.05 = – 0.03 ⎝ 20 ⎠ ⎝4⎠ ⎝5 ⎠

Therefore, the eequivalent resistance decreases by 3 % . Example 5.4 In a series RC electric circuit that is excited by a sinusoidal voltage, the magnitude of the impedance Z is computed from Z =

R + X C . Initially, R = 4 Ω and X C = 3 Ω . Find the change

2 2

in the impedance Z if the resistance R is increased by 0.25 Ω ( 6.25 % ) and the capacitive reactance X C is decreased by 0.125 Ω ( – 4.167% ).

5−4

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

Classification

Solution:

∂Z ∂Z We will first find the partial derivatives ------ and ---------- ; then we compute the change in impedance ∂R ∂X C

**from the total differential dZ . Thus,
**

XC ∂Z R ∂Z ------ = --------------------------- and --------- = --------------------------∂R ∂X C 2 2 2 2 R + XC R + XC

and

R dR + X C dX C ∂Z ∂Z dZ = ------ dR + ---------- dX C = --------------------------------------∂X C ∂R 2 2 R + XC

**and by substitution of the given values
**

1 – 0.375 4 ( 0.25 ) + 3 ( – 0.125 ) dZ = ---------------------------------------------------- = ------------------------- = 0.125 5 2 2 4 +3

**Therefore, if R increases by 6.25 % and X C decreases by 4.167% , the impedance Z increases by
**

4.167% .

Example 5.5 A light bulb is rated at 120 volts and 75 watts. If the voltage decreases by 5 volts and the resistance of the bulb is increased by 8 Ω , by how much will the power change? Solution: At V = 120 volts and P = 75 watts, the bulb resistance is

--------------R = V - = 120 - = 192 Ω P 75

2 2

and since

V∂P V∂P -----P = ----- then ------ = 2V and ------ = – ----2 ∂R R ∂V R R

2 2

and the total differential is

2 2 2 ( 120 ) ∂P ∂P 2V 120 VdP = ------ dV + ------ dR = ------ dV – ----- dR = ---------------- ( – 5 ) – ---------- ( 8 ) = – 9.375 2 2 192 ∂V ∂R R R 192

That is, the power will decrease by 9.375 watts.

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

5−5

**Chapter 5 Differential Equations, State Variables, and State Equations 5.3 Solutions of Ordinary Differential Equations (ODE)
**

A function y = f ( x) is a solution of a differential equation if the latter is satisfied when y and its derivatives are replaced throughout by f ( x) and its corresponding derivatives. Also, the initial conditions must be satisfied. For example a solution of the differential equation

d y ------- + y = 0 2 dx

2

is

y = k 1 sin x + k 2 cos x

since y and its second derivative satisfy the given differential equation. Any linear, time-invariant system can be described by an ODE which has the form

d y d y dy a n -------- + a n – 1 --------------- + … + a 1 ----- + a 0 y n n–1 dt dt dt d x d x dx b m --------- + b m – 1 ---------------- + … + b 1 ----- + b 0 x m n–1 dt = dt dt Excitation ( Forcing ) Function x ( t ) NON – HOMOGENEOUS DIFFERENTIAL EQUATION

m m–1 n n–1

(5.12)

If the excitation in (B12) is not zero, that is, if x ( t ) ≠ 0 , the ODE is called a non-homogeneous ODE. If x ( t ) = 0 , it reduces to:

d y d y dy a n -------- + a n – 1 --------------- + … + a 1 ----- + a 0 y = 0 n n–1 dt dt dt HOMOGENEOUS DIFFERENTIAL EQUATION

n n–1

⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩

(5.13)

The differential equation of (5.13) above is called a homogeneous ODE and has n different linearly independent solutions denoted as y 1 ( t ), y 2 ( t ), y 3 ( t ), …, y n ( t ) . We will now prove that the most general solution of (5.13) is:

yH ( t ) = k1 y1 ( t ) + k2 y2 ( t ) + k3 y3 ( t ) + … + kn yn ( t )

(5.14)

where the subscript H on the left side is used to emphasize that this is the form of the solution of the homogeneous ODE and k 1, k 2, k 3, …, k n are arbitrary constants.

5−6

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

**Solutions of Ordinary Differential Equations (ODE)
**

Proof: Let us assume that y 1 ( t ) is a solution of (5.13); then by substitution,

d y1 d y1 dy 1 a n ----------- + a n – 1 ----------------- + … + a 1 ------- + a 0 y 1 = 0 n n–1 dt dt dt

n n–1

(5.15)

**A solution of the form k 1 y 1 ( t ) will also satisfy (5.13) since
**

d d d a n ------ ( k 1 y 1 ) + a n – 1 ------------- ( k 1 y 1 ) + … + a 1 ---- ( k 1 y 1 ) + a 0 ( k 1 y 1 ) n n–1 dt dt dt d y1 dy 1 ⎛ d y1 ⎞ = k 1 ⎜ a n ----------- + a n – 1 ----------------- + … + a 1 ------- + a 0 y 1⎟ = 0 n n–1 dt ⎝ dt ⎠ dt

n n–1 n n–1

(5.16)

**If y = y 1 ( t ) and y = y 2 ( t ) are any two solutions, then y = y 1 ( t ) + y 2 ( t ) will also be a solution since
**

d y1 d y1 dy 1 a n ----------- + a n – 1 ----------------- + … + a 1 ------- + a 0 y 1 = 0 n n–1 dt dt dt

n n–1

and

d y2 d y2 d y2 a n ----------- + a n – 1 ----------------- + … + a 1 --------- + a 0 y 2 = 0 n n–1 dt dt dt

n n–1

Therefore,

dd da n ------ ( y 1 + y 2 ) + a n – 1 ------------- ( y 1 + y 2 ) + … + a 1 ---- ( y 1 + y 2 ) + a 0 ( y 1 + y 2 ) n n–1 dt dt dt n n–1 d d d = a n ------ y 1 + a n – 1 ------------- y 1 + … + a 1 ---- y 1 + a 0 y 1 n n–1 dt dt dt n n–1 d d d + a n ------ y 2 + a n – 1 ------------- y 2 + … + a 1 ---- y 2 + a 0 y 2 = 0 n n–1 dt dt dt

n n–1

(5.17)

In general, if

y = k 1 y 1 ( t ), k 2 y 1 ( t ), k 3 y 3 ( t ), …, k n y n ( t )

**are the n solutions of the homogeneous ODE of (5.13), the linear combination
**

y = k1 y1 ( t ) + k2 y1 ( t ) + k3 y3 ( t ) + … + kn yn ( t )

is also a solution. In our subsequent discussion, the solution of the homogeneous ODE, i.e., the complementary

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

5−7

**Chapter 5 Differential Equations, State Variables, and State Equations
**

solution, will be referred to as the natural response, and will be denoted as y N ( t ) or simply y N . The particular solution of a non-homogeneous ODE will be referred to as the forced response, and will be denoted as y F ( t ) or simply y F . Accordingly, we express the total solution of the non-homogeneous ODE of (5.12) as:

y(t) = y

Natural Response

+y

Forced Response

= yN + yF

(5.18)

The natural response y N contains arbitrary constants and these can be evaluated from the given initial conditions. The forced response y F , however, contains no arbitrary constants. It is imperative to remember that the arbitrary constants of the natural response must be evaluated from the total response.

**5.4 Solution of the Homogeneous ODE
**

Let the solutions of the homogeneous ODE

d y d y dy a n -------- + a n – 1 --------------- + … + a 1 ----- + a 0 y = 0 n n–1 dt dt dt

n n–1

(5.19)

**be of the form
**

y = ke

st

(5.20)

**Then, by substitution of (5.20) into (5.19) we obtain
**

a n ks e + a n – 1 ks

n st n – 1 st

e + … + a 1 kse + a 0 ke

st

st

st

= 0

or

( an s + an – 1 s

n n–1

+ … + a 1 s + a 0 ) ke

= 0

(5.21)

**We observe that (5.21) can be satisfied when
**

( an s + an – 1 s

n n–1

+ … + a 1 s + a 0 ) = 0 or k = 0

or s = – ∞

(5.22)

but the only meaningful solution is the quantity enclosed in parentheses since the latter two yield trivial (meaningless) solutions. We, therefore, accept the expression inside the parentheses as the only meaningful solution and this is referred to as the characteristic (auxiliary) equation, that is,

⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ an s + an – 1 s

n n–1

+ … + a1 s + a0 = 0

(5.23)

Characteristic Equation

Since the characteristic equation is an algebraic equation of an nth-power polynomial, its solutions are s 1, s 2, s 3, …, s n , and thus the solutions of the homogeneous ODE are:

5−8

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

**Solution of the Homogeneous ODE
**

y 1 = k 1 e , y 2 = k 2 e , y 3 = k 3 e , …, y n = k n e

s1 t s2 t s3 t sn t

(5.24)

Case I − Distinct Roots If the roots of the characteristic equation are distinct (different from each another), the n solutions of (5.23) are independent and the most general solution is:

yN = k1 e

s1 t

+ k2 e

s2 t

+ … + kn e

sn t

(5.25)

FOR DISTINCT ROOTS

Case II − Repeated Roots If two or more roots of the characteristic equation are repeated (same roots), then some of the terms of (5.24) are not independent and therefore (5.25) does not represent the most general solution. If, for example, s 1 = s 2 , then, k1 e

s1 t s t

s1 t

+ k2 e

s2 t

= k1 e

s1 t

+ k2 e

s1 t

= ( k 1 + k 2 )e

s1 t

= k3 e

s1 t

and we see that one term of (5.25) is lost. In this case, we express one of the terms of (5.25), say as k 2 te 1 . These two represent two independent solutions and therefore the most general solution has the form:

k2 e y N = ( k 1 + k 2 t )e

s1 t

+ k3 e

s3 t

+ … + kn e

sn t

(5.26)

**If there are m equal roots the most general solution has the form:
**

yN = ( k1 + k2 t + … + km t

m–1

)e

s1 t

+ kn – i e

s2 t

+ … + kn e

sn t

(5.27)

FOR M EQUAL ROOTS

Case III − Complex Roots If the characteristic equation contains complex roots, these occur as complex conjugate pairs. Thus, if one root is s 1 = – α + jβ where α and β are real numbers, then another root is

s 1 = – α – j β . Then, k1 e

s1 t

+ k2 e

s2 t

= k1 e = e = e = e

– αt + jβt

+ k2 e

– αt – j βt

= e

– αt

( k1 e

jβt

+ k2 e

– j βt

)

– αt – αt – αt

**( k 1 cos βt + jk 1 sin β t + k 2 cos βt – jk 2 sin β t ) [ ( k 1 + k 2 ) cos βt + j ( k 1 – k 2 ) sin β t ] ( k 3 cos βt + k 4 sin β t ) = e
**

– αt

k 5 cos ( βt + ϕ )

FOR TWO COMPLEX CONJUGATE ROOTS

(5.28)

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

5−9

**Chapter 5 Differential Equations, State Variables, and State Equations
**

If (5.28) is to be a real function of time, the constants k 1 and k 2 must be complex conjugates. The other constants k 3 , k 4 , k 5 , and the phase angle ϕ are real constants. The forced response can be found by a. The Method of Undetermined Coefficients or b. The Method of Variation of Parameters We will study the Method of Undetermined Coefficients first.

**5.5 Using the Method of Undetermined Coefficients for the Forced Response
**

For simplicity, we will only consider ODEs of order 2 . Higher order ODEs are discussed in differential equations textbooks. Consider the non-homogeneous ODE

a dy

2 2

d + b ---- y + cy = f ( x ) dt dt

(5.29)

where a , b , and c are real constants. We have learned that the total (complete) solution consists of the summation of the natural and forced responses. For the natural response, if y 1 and y 2 are any two solutions of (5.29), the linear combination

y 3 = k 1 y 1 + k 2 y 2 , where k 1 and k 2 are arbitrary constants, is also a solution, that is, if we know

the two solutions, we can obtain the most general solution by forming the linear combination of y 1 and y 2 . To be certain that there exist no other solutions, we examine the Wronskian Determinant defined below.

y1 y2 d d W ( y 1, y 2 ) ≡ d = y1 ----- y 2 – y 2 ----- y1 ≠ 0 d dx dx ----- y1 ----- y 2 dx dx WRONSKIAN DETERMINANT

(5.30)

If (5.30) is true, we can be assured that all solutions of (5.29) are indeed the linear combination of y 1 and y 2 . The forced response is obtained by observation of the right side of the given ODE as it is illustrated by the examples that follow.

5−10

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

**Using the Method of Undetermined Coefficients for the Forced Response
**

Example 5.6 Find the total solution of the ODE

dy + 4 ----- + 3y = 0 dt dt d y

2 2

(5.31)

subject to the initial conditions y ( 0 ) = 3 and y' ( 0 ) = 4 where y' = dy ⁄ dt Solution: This is a homogeneous ODE and its total solution is just the natural response found from the characteristic equation s + 4s + 3 = 0 whose roots are s 1 = – 1 and s 2 = – 3 . The total response is:

y ( t ) = yN ( t ) = k1 e + k2 e

–t – 3t 2

(5.32)

**The constants k 1 and k 2 are evaluated from the given initial conditions. For this example,
**

y ( 0 ) = 3 = k1 e + k2 e

0 0

or

k1 + k2 = 3

(5.33)

–t – 3t t=0

Also,

dy y' ( 0 ) = 4 = ----dt = – k 1 e – 3k 2 e

t=0

or

– k 1 – 3k 2 = 4

(5.34)

**Simultaneous solution of (5.33) and (5.34) yields k 1 = 6.5 and k 2 = – 3.5 . By substitution into (5.32), we obtain
**

y ( t ) = y N ( t ) = 6.5e – 3.5e

–t – 3t

(5.35)

**Check with MATLAB:
**

y=dsolve('D2y+4*Dy+3*y=0', 'y(0)=3', 'Dy(0)=4')

**y = (-7/2*exp(-3*t)*exp(t)+13/2)/exp(t)
**

pretty(y)

- 7/2 exp(-3 t) exp(t) + 13/2 ----------------------------exp(t) The function y = f ( t ) , of relation (5.35), shown in Figure 5.1, was plotted with the use of the MATLAB script Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

5−11

**Chapter 5 Differential Equations, State Variables, and State Equations
**

y=dsolve('D2y+4*Dy+3*y=0', 'y(0)=3', 'Dy(0)=4'); ezplot(y,[0 5])

3.5 3 2.5 2 1.5 1 0.5 0 13/2 exp(-t)-7/2 exp(-3 t)

0

1

2 t

3

4

5

Figure 5.1. Plot for the function y = f ( t ) of Example 5.6.

**Example 5.7 Find the total solution of the ODE
**

dy – 2t + 4 ----- + 3y = 3e dt dt dy

2 2

(5.36)

subject to the initial conditions y ( 0 ) = 1 and y' ( 0 ) = – 1 Solution: The left side of (5.36) is the same as that of Example 5.6.Therefore,

yN ( t ) = k1 e + k2 e

–t – 3t

(5.37)

(We must remember that the constants k 1 and k 2 must be evaluated from the total response). To find the forced response, we assume a solution of the form

y F = Ae

– 2t

(5.38)

We can find out whether our assumption is correct by substituting (5.38) into the given ODE of (5.36). Then,

4Ae

– 2t

– 8Ae

– 2t

+ 3Ae

– 2t

= 3e

– 2t

(5.39)

from which A = – 3 and the total solution is

5−12

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

**Using the Method of Undetermined Coefficients for the Forced Response
**

y ( t ) = yN + yF = k1 e + k2 e

–t – 3t

–3 e

– 2t

(5.40)

**The constants k 1 and k 2 are evaluated from the given initial conditions. For this example,
**

y ( 0 ) = 1 = k 1 e + k 2 e – 3e

0 0 0

or

k1 + k2 = 4

(5.41)

–t – 3t

Also,

dy y' ( 0 ) = – 1 = ----dt = – k 1 e – 3k 2 e

t=0

+ 6e

– 2t t=0

or

– k 1 – 3k 2 = – 7

(5.42)

**Simultaneous solution of (5.41) and (5.42) yields k 1 = 2.5 and k 2 = 1.5 . By substitution into (5.40), we obtain
**

y ( t ) = y N + y F = 2.5e + 1.5e

–t – 3t

–3 e

– 2t

(5.43)

**Check with MATLAB:
**

y=dsolve('D2y+4*Dy+3*y=3*exp(−2*t)', 'y(0)=1', 'Dy(0)=−1')

**y = (-3*exp(-2*t)*exp(t)+3/2*exp(-3*t)*exp(t)+5/2)/exp(t)
**

pretty(y)

-3 exp(-2 t) exp(t) + 3/2 exp(-3 t) exp(t) + 5/2 -----------------------------------------------exp(t) The plot is shown in Figure 5.2 was produced with the MATLAB script

y=dsolve('D2y+4*Dy+3*y=3*exp(−2*t)', 'y(0)=1', 'Dy(0)=−1'); ezplot(y,[0 8])

**Example 5.8 Find the total solution of the ODE
**

dy + 6 ----- + 9y = 0 dt dt dy

2 2

(5.44)

subject to the initial conditions y ( 0 ) = – 1 and y' ( 0 ) = 1

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

5−13

46) – 3t Also.46) and (5.1 0 0 1 2 t 3 4 5 5/2 exp(-t)+3/2 exp(-3 t)-3 exp(-2 t) Figure 5. Thus. State Variables. Plot for the function y = f ( t ) of Example 5.45) Next.3 0.9 0.45).47) From (5. 'y(0)=−1'.5 0. By substitution into (5. y ( 0 ) = – 1 = k 1 e + k 2 ( 0 )e 0 0 or k1 = –1 (5.4 0.7 0. ----y' ( 0 ) = 1 = dy dt = – 3k 1 e t=0 + k2 e – 3t – 3k 2 te – 3t t=0 or – 3k 1 + k 2 = 1 (5. Check with MATLAB: y ( t ) = –e – 3t – 2te – 3t (5.6 0.Chapter 5 Differential Equations.48) y=dsolve('D2y+6*Dy+9*y=0'. and State Equations 1 0. we evaluate the constants k 1 and k 2 from the given initial conditions. Solution: This is a homogeneous ODE and therefore its total solution is just the natural response found from the characteristic equation s + 6s + 9 = 0 whose roots are s 1 = s 2 = – 3 (repeated roots).47) we obtain k 1 = – 1 and k 2 = – 2 .7.2 0. Third Edition Copyright © Orchard Publications . For this example.8 0.2. 'Dy(0)=1') 5−14 Numerical Analysis Using MATLAB® and Excel®. the total response is y ( t ) = yN = k1 e – 3t 2 + k 2 te – 3t (5.

6 -0.2 -0.7 -0.1 -0. ezplot(y. Example 5.49) are s 1 = – 2 and s 2 = – 3 and thus the natural response has the form yN = k1 e – 2t + k2 e – 3t (5.5 1 1.9 -1 0 0.5 -0. we find the forced response by assuming a solution of the form y F = Ae – 2t (5. Numerical Analysis Using MATLAB® and Excel®.3 was produced with the MATLAB script y=dsolve('D2y+6*Dy+9*y=0'.3. By inspection.49) Solution: No initial conditions are given.51) We can find out whether our assumption is correct by substitution of (5.50) Next. 'y(0)=−1'. 'Dy(0)=1'). therefore. Plot for the function y = f ( t ) of Example 5.51) into the given ODE of (5.+ 6y = 3e dt dt dy 2 2 (5. Then.Using the Method of Undetermined Coefficients for the Forced Response y = -exp(-3*t)-2*exp(-3*t)*t The plot shown in Figure 5.5 3 -exp(-3 t)-2 exp(-3 t) t Figure 5. Third Edition Copyright © Orchard Publications 5−15 .4 -0.3 -0.9 Find the total solution of the ODE dy – 2t + 5 ----.8 -0. the roots of the characteristic equation of (5.49).5 t 2 2.8. we will express the solution in terms of the constants k 1 and k 2 .[0 3]) 0 -0.

Third Edition Copyright © Orchard Publications . the total response is y ( t ) = yN + yF = k1 e – 2t + k2 e – 3t + 3te – 2t (5.53) that is. and State Equations 4Ae – 2t – 10Ae – 2t + 6Ae – 2t = 3e – 2t (5. to find the forced response and we assume a solution of the form 5−16 Numerical Analysis Using MATLAB® and Excel®. we will express solution in terms of the constants k 1 and k 2 . Therefore.51) by t in order to eliminate the duplication of terms in the total response. we multiply (5.49) and equating like terms.Chapter 5 Differential Equations.49) has the same form as one of the terms of the natural response of (5. by substitution of (5. The problem here is that the right side of the given ODE of (5.55) Solution: No initial conditions are given. Therefore. we assume that the forced response has the form y F = Ate – 2t (5.52) but the sum of the three terms on the left side of (5.56) Next. To work around this problem. Then. namely the term k 1 e – 2t . that is.10 Find the total solution of the ODE d 2y + 5 dy + 6y = 4 cos 5t ----2 dt dt (5.52) is zero whereas the right side can never be zero unless we let t → ∞ and this produces a meaningless result. State Variables.50). We observe that the left side of (5.55) is the same of that of Example 5. Example 5. therefore.53) into (5. we find that A = 3 .9. the natural response is the same.54) Check with MATLAB: y=dsolve('D2y+5*Dy+6*y=3*exp(−2*t)') y = -3*exp(-2*t)+3*t*exp(-2*t)+C1*exp(-3*t)+C2*exp(-2*t) We observe that the first and last terms of the displayed expression above have the same form and thus they can be combined to form a single term C3*exp(-2*t). it has the form yN = k1 e – 2t + k2 e – 3t (5.

This inconsistency is a result of our failure to recognize that the derivatives of A cos 5t produce new terms of the form B sin 5t and these terms must be included in the forced response.57) We can find out whether our assumption is correct by substitution of the assumed solution of (5. x where n is a positive integer. we let y F = k 3 sin 5 t + k 4 cos 5t (5.55) we obtain – 25 k 3 sin 5t – 25k 4 cos 5 t + 25k 3 cos 5 t – 25k 4 sin 5 t + 6k 3 sin 5t + 6k 4 cos 5 t = 4 cos 5 t Collecting like terms and equating sine and cosine terms. – 25A cos 5t – 25A sin 5 t + 6A cos 5t = – 19A cos 5t – 25A sin 5 t = 4 cos 5t but this relation is invalid since by equating cosine and sine terms.57) into the given ODE of (5.1 summarizes the forms of the forced response for a second order Numerical Analysis Using MATLAB® and Excel®. Then.56).59) We use MATLAB to solve (5. the total solution is y ( t ) = yN + yF ( t ) = k1 e – 2t + k2 e – 3t – 38 50 + -------. y=simple(y) y = -38/493*cos(5*t)+50/493*sin(5*t)+C1*exp(-3*t)+C2*exp(-2*t) In most engineering problems the right side of the non−homogeneous ODE consists of elementary functions such as k (constant). 25*x−19*y−4) k3 = 50/493 k4 = -38/493 Therefore.Using the Method of Undetermined Coefficients for the Forced Response y F = A cos 5t (5.sin 5t + -------. e . cos kx .58) and by substitution into (5.60) Check with MATLAB: y=dsolve('D2y+5*Dy+6*y=4*cos(5*t)'). we obtain the following set of equations 19k 3 + 25k 4 = 0 25k 3 – 19 k 4 = 4 (5.59) format rat. we find that A = – 4 ⁄ 19 and also A = 0 . Accordingly. Third Edition Copyright © Orchard Publications n kx 5−17 . Table 5. [k3 k4]=solve(19*x+25*y. sin kx . and linear combinations of these.cos 5t 493 493 (5.

This term with n = 1 .+ 4y = te – e dt dt (5. and α = 0 .11 Find the total solution of the ODE d 2y 2 dy – 2t – 2t + 4 ----. TABLE 5. the most general form of the forced response y F ( t ) is the linear combination of these terms. s 1 = s 2 = – 2 . r = – 2 .+ cy = f ( t ) 2 dt dt f (t) k (constant) k t ( n = positive integer) ke rt n 2 Form of Forced Response y F ( t ) K (constant) K0 t + K1 t Ke rt n n–1 + … + Kn – 1 t + Kn ( r =real or complex) k cos αt or k sin αt ( α =constant) K 1 cosαt + K 2 sin αt k t e cos αt or k t e sin α t n rt n rt ( K0 t + K1 t n n n–1 + … + K n – 1 t + K n )e cos αt + … + K n – 1 t + K n )e sin αt rt rt + ( K0 t + K1 t n–1 We must remember that if f ( t ) is the sum of several terms.1 Form of the forced response for 2nd order differential equations d y dy Forced Response of the ODE a ------.Chapter 5 Differential Equations.61) Solution: No initial conditions are given.+ b ----. Example 5. if a term in y F ( t ) is a duplicate of a term in the natural response y N ( t ) . we refer to the table of the previous page and from the last row we choose the term k t e cos αt . and thus the natural response has the form yN = k1 e –2 t + k 2 te –2 t (5. that is. n rt 5−18 Numerical Analysis Using MATLAB® and Excel®.62) To find the forced response (particular solution). and State Equations ODE with constant coefficients. State Variables. Third Edition Copyright © Orchard Publications . therefore we will express solution in terms of the constants k 1 and k 2 . The roots of the characteristic equation are equal. Also. we must multiply y F ( t ) by the lowest power of t that will eliminate the duplication.

that is 2 ( 3k 3 t + k 4 )e – 2t = te – 2t –e – 2t (5. f=simple(f) % Form and simplify the left side of the given ODE f = 2*(3*k3*t+k4)*exp(-2*t) Finally. we obtain the total solution y ( t ) = k1 e –2 t + k 2 te –2 t 1 3 –2 t 1 2 –2 t + -.Using the Method of Undetermined Coefficients for the Forced Response reduces to kte –2 t .61) and equating with the right side. f1=diff(f0). Therefore the forced response will have the form y F = ( k 3 t + k 4 )e –2 t (5.61). therefore.2).64) Now.64) into the given ODE of (5. Third Edition Copyright © Orchard Publications 5−19 . f2=simple(f2) % Compute and simplify second derivative f2 = 2*exp(-2*t)*(3*k3*t+k4-6*k3*t^2-4*k4*t+2*k3*t^3+2*k4*t^2) f=f2+4*f1+4*f0. we need to evaluate the constants k 3 and k 4 . we equate f above with the right side of the given ODE. syms t k3 k4 f0=(k3*t^3+k4*t^2)*exp(−2*t).t e – -.62) by t to obtain a suitable form for the forced response which now is y F = ( k 3 t + k 4 t )e 3 2 –2 t – 2t – 2t (5. f1=simple(f1) % Define symbolic variables % Forced response (5. we multiply (5.64) and combining the forced response with the natural response. By substitution of these values into (5.64) % Compute and simplify first derivative f1 = -t*exp(-2*t)*(-3*k3*t-2*k4+2*k3*t^2+2*k4*t) f2=diff(f0. We use MATLAB do the computations as shown below.t e 6 2 (5.65) and we find k 3 = 1 ⁄ 6 and k 4 = – 1 ⁄ 2 .66) We verify this solution with MATLAB as follows: z=dsolve('D2y+4*Dy+4*y=t*exp(−2*t)−exp(−2*t)') z = 1/6*exp(-2*t)*t^3-1/2*exp(-2*t)*t^2 +C1*exp(-2*t)+C2*t*exp(-2*t) Numerical Analysis Using MATLAB® and Excel®. This is done by substituting (5.63) 2 But the terms e and te are also present in (5.

6 Using the Method of Variation of Parameters for the Forced Response In certain non−homogeneous ODEs.= f ( t ) dt dt dt dt (5. the right side f ( t ) cannot be determined by the method of undetermined coefficients.70) Simultaneous solution of (5. and State Equations 5.69) (5.+ α ( t ) ----.71) in terms of the constants k 1 and k 2 by the a. then. method of variation of parameters Solution: With either method. we must first find the natural response. This method will work with all linear equations including those with variable coefficients such as dy d2 y ------.Chapter 5 Differential Equations.+ β ( t )y = f ( t ) 2 dt dt (5.67) will yield the total solution.68) (5. Third Edition Copyright © Orchard Publications .⋅ ------.12 Find the total solution of d 2 y + 4 dy + 3y = 12 ----------2 dt dt (5. The method of variation of parameters replaces the constants k 1 and k 2 by two variables u 1 and u 2 that satisfy the following three relations: y = u1 y1 + u2 y2 du1 du2 -----.y2 = 0 dt dt du1 dy1 du 2 dy 2 ------.67) provided that the general form of the natural response is known.68) and (5.y1 + ------. method of undetermined coefficients b.+ ------. which when substituted into (5. integration of these will produce u 1 and u 2 .69) will yield the values of du1 ⁄ dt and du 2 ⁄ dt . The characteristic equation yields 5−20 Numerical Analysis Using MATLAB® and Excel®. Our discussion will be restricted to second order ODEs with constant coefficients. For these ODEs we must use the method of variation of parameters.⋅ ------. Example 5. State Variables.

( – 3e – 3t ) = 12 dt dt (5.= -----------------------------------------.y 1 + ------. the natural response is yN = k1 e + k2 e –t –3 t (5.= ------------------------------. Third Edition Copyright © Orchard Publications 5−21 .76) and from (5.Using the Method of Variation of Parameters for the Forced Response the roots s 1 = – 1 and s 2 = – 3 . du1 dy1 du 2 dy 2 ------.= f ( t ) dt dt dt dt or du 1 du 2 ------.77) Next. from (5.78) –e –t – 3e – 3t and Numerical Analysis Using MATLAB® and Excel®.75) Also.y 2 = 0 dt dt or du 1 – t du 2 – 3t ------.72) a. Using the method of undetermined coefficients we let y F = k 3 (a constant).( – e – t ) + ------. the total solution is y = u1 y1 + u2 y2 or y = u1 e + u2 e –t – 3t (5.e + ------.72) and we let the solutions y 1 and y 2 be represented as y1 = e –t and y 2 = e – 3t (5. With the method of variation of parameters we begin with the natural response found above as (5.70). by substitution into (5.⋅ ------. du 2 du 1 ------.+ ------. we find du1 ⁄ dt and du 2 ⁄ dt by Cramer’s rule as follows: 0 e – 3t – 3t – 3t – 3t du 1 t 12 – 3e – 12e – 12e ------.e = 0 dt dt (5.68). Therefore. Then.71) we obtain k 3 = 4 and thus the total solution is y ( t ) = yN + yF = k1 e + k2 e –t –3 t +4 (5.74) Then by (5.73) b.= 6e –t – 3t – 4t – 4t – 4t dt e – 3e + e – 2e e (5.69).⋅ ------.= ----------------.

80) (5.84) (5. Check with MATLAB: y=dsolve('D2y+4*Dy+3*y=12') y = (4*exp(t)+C1*exp(-3*t)*exp(t)+C2)/exp(t) Example 5.85) We let y 1 = cos 2t and y 2 = sin 2t Then.81) = ( 6e + k 1 )e + ( – 2 e + k 2 )e – 3t = 6 + k1 e – 2 + k2 e = k1 e + k2 e –t – 3t +4 We observe that the last expression in (5.= – 6 e – 4t – 4t dt – 2e – 2e (5.83) (5. by (5. integration of (5.81) is the same as (5.82) in terms of the constants k 1 and k 2 by any method.+ 4y = tan 2t 2 dt 2 (5. The characteristic equation is s + 4 = 0 from which s = ± j2 and thus the natural response is yN = k1 e j2t 2 + k2 e – j 2t (5. we will use the method of variation of parameters.78) and (5. Solution: This ODE cannot be solved by the method of undetermined coefficients.75) yields u 1 = 6 e dt = 6e + k 1 y = u1 e + u2 e –t –t – 3t t ∫ t t u 2 = – 6 e dt = – 2 e + k 2 –t 3t – 3t ∫ 3t 3t (5.= -------------------------------.13 Find the total solution of d y ------. and State Equations e –t 0 –t –t du 2 3t 12e –e 12------.79) and substitution into (5. Third Edition Copyright © Orchard Publications .= -------------.Chapter 5 Differential Equations.79) Now. State Variables.73) of part (a).68) the solution is y = u1 y 1 + u 2 y 2 = u1 cos 2t + u 2 sin 2t 5−22 Numerical Analysis Using MATLAB® and Excel®. therefore.

85) yields 1 sin 2t sin 2t 1 u 1 = – -.y1 + ------.69).= f ( t ) = ------. from (5.= -------------------------------------------------.ln ( sec 2t + tan 2t ) + k 1 4 2 cos 2t 4 ∫ 2 (5.= ------------------------------------------.70).dt = -----------.= -----------2 2 dt (5.= ----------------2 2 cos 2t sin 2t 2 cos 2t dt 2 cos 2t + 2 sin 2t – 2 sin 2t 2 cos 2t 0 sin 2t 2 (5.( – 2 sin 2t ) + ------. du 2 du1 dy1 du 2 dy 2 du1 ------.cos 2t ln ( sec 2t + tan 2t ) + k 1 cos 2t + k 2 sin 2t 4 (5.⋅ ------.89) and substitution into (5.88) and (5.-------------.( 2 cos 2t ) = tan 2t dt dt dt dt dt dt (5.⋅ ------.86) and from (5.91) 1 cos 2t u 2 = -.= -----------------------------------------------------.90) (5.– -.cos 2t + ------.89) Now. Third Edition Copyright © Orchard Publications 5−23 . du1 du 2 ------.Using the Method of Variation of Parameters for the Forced Response Also.– -.4 4 4 1 = – -.88) and cos 2t 0 du 2 sin 2t – 2 sin 2t tan 2t ------. integration of (5.92) Check with MATLAB: y=dsolve('D2y+4*y=tan(2*t)') y = -1/4*cos(2*t)*log((1+sin(2*t))/cos(2*t))+C1*cos(2*t)+C2*sin(2*t) Numerical Analysis Using MATLAB® and Excel®.+ ------. we find du1 ⁄ dt and du 2 ⁄ dt by Cramer’s rule as follows: sin 2t – -------------2 du1 tan 2t 2 cos 2t cos 2t – sin 2t ------.+ k 2 2 4 ∫ sin 2t cos 2t 1 sin 2t cos 2t y = u1 y 1 + u 2 y 2 = -------------------------.cos 2t ln ( sec 2t + tan 2t ) + k 1 cos 2t – -------------------------.+ k 2 sin 2t .sin 2t = 0 dt dt (5.87) Next.sin 2t dt = – ------------.y 2 = 0 dt dt or du 2 du1 ------.

i + -.R di .= – --.---. However. our discussion will be limited to linear. Third Edition Copyright © Orchard Publications . The state variable method is best illustrated through several examples presented in this chapter.98) (5.+ a 0 y ( t ) = x ( t ) dt (5.Ri + L ---. time−invariant systems.94) Differentiating both sides and dividing by L we obtain 2 1 1d .-----.= x 1 dt dt Then. The state variable method offers the advantage that it can also be used with non−linear and time−varying systems.14 A system is described by the integro−differential equation di 1 .99) and dx 1 di · x 2 = ---.---.95) or 2 1 R di 1 d t . State Variables. these will be discussed in Chapter 9.i = -.jωe 2 L L dt LC dt jωt (5. or simply state equations.+ ------.96) Next.jωe 2 L L dt LC dt jωt (5. The resulting first−order differential equations are called state space equations.– ------. and State Equations 5. State equations can also be solved with numerical methods such as Taylor series and Runge− Kutta methods.97) (5. Example 5.Chapter 5 Differential Equations.7 Expressing Differential Equations in State Equation Form A first order differential equation with constant coefficients has the form dy a 1 ----. An nth−order differential equation can be resolved to n first−order simultaneous differential equations with a set of auxiliary variables called state variables. we define two state variables x 1 and x 2 such that x1 = i (5.= ------. 2 2 · x 2 = d i ⁄ dt 5−24 Numerical Analysis Using MATLAB® and Excel®.------t + --.+ --dt C t ∫ – ∞ i dt = e jωt (5.93) In a second order differential equation the highest order is a second derivative.

4.101) in a compact form as · x = Ax + bu (5. · x2 A = 0 1 – -----LC 1 . Thus.96) through (5. Third Edition Copyright © Orchard Publications 5−25 .x 2 – -----. the state representation of a system can be described by the pair of the of the state space equations · x = Ax + bu y = Cx + du (5.1. · x + A d u b + Σ ∫ dt x C + Σ + y Figure 5.99).x 1 + -.105) Numerical Analysis Using MATLAB® and Excel®. Therefore. we write the state equations of (5.103) The output y ( t ) is expressed by the state equation y = Cx + du (5. and d is a column vector.Expressing Differential Equations in State Equation Form · where x k denotes the derivative of the state variable x k . and u = any input -. b= 1 .105) can be realized with the block diagram of Figure 5. Block diagram for the realization of the state equations of (5.100) as 0 · x1 = 1 · – -----x2 LC 1 x 0 1 + 1 u R – -.101) We usually write (5.jωe L LC L (5.j ω e jωt L L (5.105) The state space equations of (5.104) where C is another matrix.x 2 -. we obtain the state equations · x1 = x2 1 1 jωt R · x 2 = – -.100) It is convenient and customary to express the state equations in matrix form.j ω e jωt L (5. From (5. –R -L x = x1 x2 .102) 0 where · x · x = 1.

= x 4 = – a 0 x 1 – a 1 x 2 – a 2 x 3 – a 3 x 4 + u ( t ) 4 dt 4 (5. Third Edition Copyright © Orchard Publications .110) where 5−26 Numerical Analysis Using MATLAB® and Excel®. (5. we must define four state variables that will be used with the resulting four first−order state equations.109) is written as · x = Ax + bu (5. and x 4 . and State Equations We will learn how to solve the matrix equations of (5.108) and in matrix form · x1 · x2 · x3 · x4 0 0 = 0 –a0 1 0 0 –a1 0 1 0 –a2 0 0 1 –a3 x1 0 x2 + 0 u(t) x3 0 1 x4 (5.105) in the subsequent sections. Example 5. We denote the state variables as x 1. x 2.106) where y ( t ) is the output and u ( t ) is any input.109) In compact form.107) We observe that · x1 = x2 · x2 = x3 · x3 = x4 d y · --------.106) as a set of state equations. x 3 .Chapter 5 Differential Equations. Express (5. State Variables. therefore. Solution: The differential equation of (5.15 A fourth−order system is described by the differential equation 4 d y + a d y + a d y + a dy + a y ( t ) = u ( t ) --------3 -------2 ------1 ----0 3 2 4 dt dt dt dt 3 2 (5.106) is of fourth−order. and we relate them to the terms of the given differential equation as x1 = y ( t ) dy x 2 = ----dt d y x 3 = -------2 dt 2 d y x 4 = -------3 dt 3 (5.

and u = u ( t ) 5. and the initial condition. Next.⎨ e dt dt ⎩ ∫t e 0 t – ατ β u ( τ ) dτ ⎬ ⎫ ⎭ Numerical Analysis Using MATLAB® and Excel®.114) reduces to x ( t 0 ) = x 0 and thus the initial condition is satisfied. we differentiate (5. 0 b= 0. 0 1 · x= . is denoted as x0 = x ( t0 ) (5.( e x 0 ) + ---.112).113) satisfies the initial condition of (5.Solution of Single State Equations · x1 · x2 · x3 · x4 0 0 A= 0 –a0 1 0 0 –a1 0 1 0 –a2 0 0 . Then. 1 –a3 x1 x= x2 x3 x4 .115) The second term of (5.114) The first term in the right side of (5. we must show that (5.113) Proof: First. x ( t0 ) = e α ( t0 – t0 ) x0 + e αt ∫t t0 0 e –α τ β u ( τ ) dτ (5.⎧ αt · x ( t ) = ---.α ( t – t0 ) d. Therefore. This is done by substitution of t = t 0 in (5.114) reduces to x 0 since e α ( t0 – t0 ) x0 = e x0 = x0 0 (5.111) is x(t) = e α ( t – t0 ) x0 + e αt ∫t e 0 t – ατ β u ( τ ) dτ (5.111). β . To prove this. if non−zero. (5.112) We will now prove that the solution of the first state equation in (5.111) where α . we must prove that (5. k 1 .8 Solution of Single State Equations Let us consider the state equations · x = αx + βu y = k1 x + k2 u (5. and k 2 are scalar constants.113) satisfies also the first equation in (5.113).114) is zero since the upper and lower limits of integration are the same.113) with respect to t and we obtain d. Third Edition Copyright © Orchard Publications 5−27 .

and we will prove that the solution of the matrix differential equation · (5.122) x ( t0 ) = x0 is obtained from the relation At 5−28 Numerical Analysis Using MATLAB® and Excel®. · x = αx + βu and this is the same as the first equation of (5.120) where for two or more simultaneous differential equations A and C are 2 × 2 or higher order matrices.113).118) with initial condition x0 = x ( t0 ) is obtained from the relation x( t) = e α ( t – t0 ) x0 + e αt ∫t e 0 t –α τ β u ( τ ) dτ (5. In summary. State Variables.111). The second equation of (5.119) 5. and b and d are column vectors with two or more rows. Therefore. if α and β are scalar constants. the solution of · x = αx + βu (5.117) (5.116) We observe that the bracketed terms of (5. and State Equations α ( t – t0 ) αt · x(t) = αe x0 + α e α ( t – t0 ) αt ∫t ∫t t 0 t 0 e e – ατ β u ( τ ) dτ + e [ e αt – ατ βu(τ)] τ = t βu(t) = α e x0 + e t 0 – ατ β u ( τ ) dτ + e e αt – αt α ( t – t0 ) · x(t)= α e x0 + ∫t e α(t – τ) β u ( τ ) dτ + β u ( t ) (5.121) x = Ax + bu with initial conditions (5.111) is an algebraic equation whose coefficients are scalar constants. In this section we will introduce the state transition matrix e .116) are the same as the right side of the assumed solution of (5. Third Edition Copyright © Orchard Publications .Chapter 5 Differential Equations.9 The State Transition Matrix Let us again consider the state equations pair · x = Ax + bu y = Cx + du (5.

123) Proof: Let A be any n × n matrix whose elements are constants. that is.128) where we have used (5. From (5.34).125) for the initial condition.125) Differentiation of (5. Third Edition Copyright © Orchard Publications 5−29 . d ---.120). another n × n matrix denoted as ϕ ( t ) .126) and by comparison with (5.A t 3! 2! n! (5. we find that ϕ(0) = e A0 = I + A0 + … = I (5.124).123) is the solution of the first equation of (5.e = 0 + A ⋅ 1 + A t + … = A + A t + … dt (5. we must prove that it satisfies both the initial condition and the matrix differential equation.127) To prove that (5.124) with respect to t yields d At 2 2 ϕ' ( t ) = ---. is said to be the state transition matrix of (5.e At = Ae At dt Numerical Analysis Using MATLAB® and Excel®.124) we obtain d ---. if it is related to the matrix A as the matrix power series ϕ(t) ≡ e At 1 22 1 33 1 nn = I + At + ---.120) is also satisfied. To prove that the first equation of (5.A t + ---.127). The integral is zero since the upper and lower limits of integration are the same. Then. The initial condition is satisfied from the relation x ( t0 ) = e A ( t0 – t0 ) x0 + e At 0 ∫t t0 0 e –A τ bu ( τ ) dτ = e A0 x 0 + 0 = Ix 0 = x 0 (5.124) where I is the n × n identity matrix.A t + … + ---.e At = Ae At dt (5.The State Transition Matrix x(t) = e A ( t – t0 ) x0 + e At ∫t e 0 t –A τ bu ( τ ) dτ (5. we differentiate the assumed solution x( t) = e A ( t – t0 ) x0 + e At ∫t e 0 t –A τ bu ( τ ) dτ with respect to t and we use (5.

and I be the n × n identity matrix. At 5. the solution of second or higher order systems using the state variable method. Third Edition Copyright © Orchard Publications .132). and integration of (5. if A is an n × n matrix whose elements are constants. Thus.10 Computation of the State Transition Matrix e i = 1. By definition. State Variables. n ≥ 2 . and b is a column vector with n elements. Evaluation of the state transition matrix e is based on the Cayley−Hamilton theorem.130) with initial condition x0 = x ( t0 ) (5. A ( t – t0 ) At · x ( t ) = Ae x 0 + Ae t ∫t e 0 –A τ bu ( τ ) dτ + e e At – A t bu ( t ) or A ( t – t0 ) At · x0 + e x(t) = A e ∫t e 0 t –A τ bu ( τ ) dτ + e e At – A t bu ( t ) (5. The roots of the polynomial of (5.Chapter 5 Differential Equations. n of A are the roots of the nth order polynomial det [ A – λI ] = 0 At Let A be an n × n matrix.133) We recall that expansion of a determinant produces a polynomial.129) as x ( t ) . This theorem states that a matrix can be expressed as an ( n – 1 )th degree polynomial in terms of the matrix A as e At At = a0 I + a1 A + a2 A + … + an – 1 A 2 n–1 (5. (5. 2.129) We recognize the bracketed terms in (5.129) reduces to · x ( t ) = Ax + bu In summary. the expression (5. and State Equations Then. or complex numbers. the eigenvalues λ i . entails the computation of the state transition matrix e .132) Therefore. and the last term as bu ( t ) .134) 5−30 Numerical Analysis Using MATLAB® and Excel®. ….131) is x(t) = e A ( t – t0 ) x0 + e At ∫t e 0 t –A τ bu ( τ ) dτ (5. the solution of · x ( t ) = Ax + bu (5.133) can be real (unequal or equal).

that is. we must find the coefficients a i of (5. Third Edition Copyright © Orchard Publications 5−31 . We accept (5. that is. if all eigenvalues of a given matrix A are distinct. The proof can be found in Linear Algebra and Matrix Theory textbooks. Since A is a 2 × 2 matrix. Since the coefficients a i are functions of the eigenvalues λ . we must consider the following cases: Case I: Distinct Eigenvalues (Real or Complex) If λ 1 ≠ λ 2 ≠ λ 3 ≠ … ≠ λ n .134) without proving it. we only need to consider the first two terms of that relation.Computation of the State Transition Matrix where the coefficients a i are functions of the eigenvalues λ . = (– 2 – λ)(– 1 – λ) = 0 (λ + 1)(λ + 2) = 0 λ 1 = – 1 and λ 2 = – 2 (5.134).16 Compute the state transition matrix e Solution: We must first find the eigenvalues λ of the given matrix A . Numerical Analysis Using MATLAB® and Excel®.135) = e λn t Example 5. These are found from the expansion of For this example. det [ A – λI ] = 0 At given that A = – 2 1 0 –1 ⎧ ⎫ 1 det [ A – λI ] = det ⎨ – 2 1 – λ 1 0 ⎬ = det – 2 – λ = 0 0 1 ⎭ 0 –1–λ ⎩ 0 –1 or Therefore. the coefficients a i are found from the simultaneous solution of the following system of equations: a0 + a1 λ1 + a2 λ1 + … + an – 1 λ1 a0 + a1 λ2 + a2 λ2 + … + an – 1 λ2 … a0 + a1 λn + a2 λn + … + an – 1 λn 2 n–1 2 2 n–1 n–1 = e = e λ1 t λ2 t (5.136) Next.

if it is a 3 × 3 matrix. a0 + a1 λ1 = e a0 + a1 λ2 = e λ1 t λ2 t or a0 + a1 ( –1 ) = e a0 + a1 ( –2 ) = e –t – 2t (5. 5−32 Numerical Analysis Using MATLAB® and Excel®. and so on.141) If A matrix is a 3 × 3 matrix. If the eigenvalues are distinct. it will yield three eigenvalues. it will yield two eigenvalues. we use only the first 2 terms of the right side of the state transition matrix e At = a0 I + a1 A + a2 A + … + an – 1 A 2 n–1 (5. we use the first 3 terms.137).135). otherwise we refer to Case II.138) Simultaneous solution of (5. 2.Chapter 5 Differential Equations. we compute the state transition matrix e procedure: At for a given matrix A using the following 1. We can write [ A – λI ] at once by subtracting λ from each of the main diagonal elements of A . If the dimension of A is a 2 × 2 matrix. and State Equations e At = a0 I + a1 A (5. State Variables. For this example.139) and by substitution into (5.140) 0 In summary. and so on. Third Edition Copyright © Orchard Publications . we perform steps 2 through 4 below.137) The coefficients a 0 and a 1 are found from (5. If the dimension of A is a 2 × 2 matrix.138) yields a 0 = 2e – e a1 = e – e –t –t – 2t – 2t (5. We find the eigenvalues λ from det [ A – λI ] = 0 . e At = ( 2e – e –t – 2t ) 1 0 0 1 –t + (e – e –t – 2t ) –2 1 0 –1 or e At = e – 2t e –e e –t – 2t (5.

We obtain [ A – λI ] at once.143) and expansion of this determinant yields the polynomial λ – 6λ + 11λ – 6 = 0 3 2 (5. and we simplify. r(1)).141). fprintf('lambda2 = %5..2f \t'. r=roots(p).2f \t'. r(2)). 4.144). fprintf('lambda1 = %5. fprintf('lambda3 = %5. We obtain the a i coefficients from a0 + a1 λ1 + a2 λ1 + … + an – 1 λ1 a0 + a1 λ2 + a2 λ2 + … + an – 1 λ2 … a0 + a1 λn + a2 λn + … + an – 1 λn 2 n–1 2 2 n–1 n–1 λ1 t λ2 t = e = e = e λn t We use as many equations as the number of the eigenvalues. and we solve for the coefficients ai . p=[1 −6 11 −6].Computation of the State Transition Matrix 3.17 Compute the state transition matrix e At given that 7 –5 4 –1 8 –3 5 A = 0 2 (5.144) We will use MATLAB roots(p) function to obtain the roots of (5. Example 5.142) Solution: 1. fprintf(' \n'). Third Edition Copyright © Orchard Publications 5−33 .00 lambda3 = 1.00 and thus the eigenvalues are Numerical Analysis Using MATLAB® and Excel®. by subtracting λ from each of the main diagonal elements of A . We substitute the a i coefficients into the state transition matrix of (5.2f'.00 lambda2 = 2. det [ A – λI ] = det 5–λ 0 2 7 –5 = 0 4–λ –1 8 –3–λ (5... Then. We first compute the eigenvalues from det [ A – λI ] = 0 . r(3)) lambda1 = 3.

exp(2*t). syms t.. We also use MATLAB to perform the substitution into the state transition matrix. 1 2 4..e – e + -. a1 = −5/2*exp(t)−3/2*exp(3*t)+4*exp(2*t). Third Edition Copyright © Orchard Publications .148) 4.. B=sym('[1 1 1. 0 4 −1. and a 2 from a0 + a1 λ1 + a2 λ1 = e a0 + a1 λ2 + a2 λ2 = e a0 + a1 λ3 + a2 λ3 = e 2 2 2 λ1 t λ2 t λ3 t or a0 + a1 + a2 = e a 0 + 2a 1 + 4a 2 = e a 0 + 3a 1 + 9a 2 = e t 2t 3t (5. 1 3 9]').. and to perform the matrix multiplications. e At = a0 I + a1 A + a2 A 2 (5.147). and State Equations λ1 = 1 λ2 = 2 λ3 = 3 (5. We obtain the coefficients a 0.. disp('a2 = '). Since A is a 3 × 3 matrix..134).e 2 2 (5.e 2 2 1 t 2t 1 3t a 2 = -. disp('a1 = ').. disp(a(1)).. b=sym('[exp(t). disp(a(2)). fprintf(' \n').e + 4e – -. eAt=a0*eye(3)+a1*A+a2*A^2 5−34 Numerical Analysis Using MATLAB® and Excel®. A = [5 7 −5. disp(a(3)) a0 = 3*exp(t)-3*exp(2*t)+exp(3*t) a1 = -5/2*exp(t)+4*exp(2*t)-3/2*exp(3*t) a2 = 1/2*exp(t)-exp(2*t)+1/2*exp(3*t) Thus. a=B\b.145) 2. exp(3*t)]'). that is. a 1.Chapter 5 Differential Equations. a2 = 1/2*exp(t)+1/2*exp(3*t)−exp(2*t). disp('a0 = '). a 0 = 3e – 3e + e t 2t 3t 5 t 2t 3 3t a 1 = – -.147) We will use the following MATLAB script for the solution of (5.. The script is shown below. a0 = 3*exp(t)+exp(3*t)−3*exp(2*t). 2 8 −3]. we need to use the first 3 terms of (5. State Variables.146) 3.

we will assume that the polynomial of det [ A – λI ] = 0 (5. and m of these roots are equal.Computation of the State Transition Matrix eAt = [-2*exp(t)+2*exp(2*t)+exp(3*t).-3*exp(t)+5*exp(2*t)-exp(3*t).-9*exp(t)+10*exp(2*t)-exp(3*t).-6*exp(t)+5*exp(2*t)+exp(3*t). λn (5.18 Compute the state transition matrix e At given that A = –1 0 2 –1 Solution: 1. λm + 1 . Numerical Analysis Using MATLAB® and Excel®. For this example. 2*exp(t)-3*exp(2*t)+exp(3*t)] [-3*exp(t)+4*exp(2*t)-exp(3*t).150) The coefficients a i of the state transition matrix e At = a0 I + a1 A + a2 A + … + an – 1 A 2 n–1 (5. In other words. Example 5.152) below. the roots are λ1 = λ2 = λ3 … = λm . 6*exp(t)-6*exp(2*t)+exp(3*t)] Thus.149) has n roots. We first find the eigenvalues λ of the matrix A and these are found from the polynomial of det [ A – λI ] = 0 . e At – 2e + 2e + e = – e + 2e – e t 2t t 2t t 2t 3t – 6 e + 5e + e – 3e + 5e – e t 2t t 2t 3t t 2t 3t 4e – 3e – e 2e – 3e + e 6e – 6e + e t 2t t 2t t 2t 3t 3t 3t 3t 3t – 3e + 4e – e – 9e + 10e – e 3t Case II: Multiple Eigenvalues In this case. 4*exp(t)-3*exp(2*t)-exp(3*t)] [-exp(t)+2*exp(2*t)-exp(3*t). Third Edition Copyright © Orchard Publications 5−35 .151) are found from the simultaneous solution of the system of equations of (5.

Third Edition Copyright © Orchard Publications . State Variables. 2.e 1 dλ 1 dλ 1 or a0 + a1 λ1 = e λ1 t λ1 t a 1 = te and by substitution with λ 1 = λ 2 = – 1 . For this example. e At = a0 I + a1 A (5.e 1 1 m–1 m–1 dλ 1 dλ 1 a0 + a1 λm + 1 + a2 λm + 1 + … + an – 1 λm + 1 = e a 0 + a 1 λn + a 2 λ n + … + a n – 1 λ n 2 n–1 2 n–1 m–1 m–1 2 2 (5. and State Equations a0 + a1 λ1 + a2 λ1 + … + an – 1 λ1 2 n–1 = e λ1 t d d λt -------. we only need the first two terms of the state transition matrix.( a 0 + a 1 λ 1 ) = -------. Since A is a 2 × 2 matrix.152).( a 0 + a 1 λ 1 + a 2 λ 2 + … + a n – 1 λ n – 1 ) = ------.e 1 1 1 dλ 1 dλ 1 d .λt d -------2 ( a 0 + a 1 λ 1 + a 2 λ 2 + … + a n – 1 λ n – 1 ) = -------2 e 1 1 1 dλ 1 dλ 1 … d d -------------. that is. We find a 0 and a 1 from (5.Chapter 5 Differential Equations.153) 3.152) λ1 t λ m + 1t … = e λn t 0 = 0 det [ A – λI ] = det – 1 – λ 2 –1–λ = (– 1 – λ)(– 1 – λ) = 0 = (λ + 1) = 0 λ1 = λ2 = –1 2 and thus.λ t -------. we obtain 5−36 Numerical Analysis Using MATLAB® and Excel®.( a 0 + a 1 λ 1 + a 2 λ 2 + … + a n – 1 λ n – 1 ) = -------------. a0 + a1 λ1 = e λ1 t dd.

0000 2.153). 2 −1].154) into (5. and we will briefly discuss eigenvectors on the next section. 2 8 −3]. lambda=eig(B) lambda = 1. We will first use MATLAB to verify the values of the eigenvalues found in Examples 5.16: A= [−2 1. 0 −1]. 0 4 −1.16 through 5. lambda=eig(A) lambda = -2 -1 For Example 5. we obtain e At –t –t = ( e + te ) 1 0 0 0 + te – t – 1 0 = e At = e –t –t 2 –1 1 2te e –t (5.17: B = [5 7 −5.0000 For Example 5. we invoke the help eig command.0000 3. By substitution of (5. To find out how it is used.154) 4. lambda=eig(C) lambda = -1 -1 Numerical Analysis Using MATLAB® and Excel®.18. Third Edition Copyright © Orchard Publications 5−37 .155) We can use the MATLAB eig(x) function to find the eigenvalues of an n × n matrix.18: C = [−1 0.Computation of the State Transition Matrix a0 – a1 = e –t –t a 1 = te Simultaneous solution of the last two equations yields a 0 = e + te a 1 = te –t –t –t (5. For Example 5.

161) As we know. and State Equations 5. We can express this relation in matrix form as a 11 a 12 … a 1n x 1 a 21 a 22 … a 2n x 2 … … … … … a n1 a n2 … a nn x n = λ x1 x2 … xn (5. that is. Third Edition Copyright © Orchard Publications .11 Eigenvectors Consider the relation AX = λX (5. the roots λ of the characteristic equation are the eigenvalues of the matrix A . 5−38 Numerical Analysis Using MATLAB® and Excel®.156) where A is an n × n matrix.160) results in a polynomial equation of degree n in λ . and corresponding to each eigenvalue λ .160) Expansion of the determinant of (5. and λ is a scalar number. We can express (5.159) will have non−trivial solutions if and only if its determinant is zero*.157) as or ( a 11 – λ )x 1 a 21 x 1 … an 1 x1 ( A – λI )X = 0 a 12 x 2 … an2 x2 … a1 n xn a2 n xn = 0 (5.Chapter 5 Differential Equations.159) The equations of (5.158) to be a non−zero vector and the product ( A – λI )X to be zero.158) ( a 22 – λ )x 2 … … … … ( a nn – λ )x n (5. there is a non−trivial solution of the column vector X .. if ( a 11 – λ ) det a 21 … an 1 a 12 … an2 … a1n a2n = 0 ( a 22 – λ ) … … … … ( a nn – λ ) (5. *. and it is called the characteristic equation. i.e. This is because we want the vector X in (5. X is a column vector. State Variables.157) We write (5.160) in a compact form as det ( A – λI ) = 0 (5.

Numerical Analysis Using MATLAB® and Excel®. Solution: a. they are normalized to unit length. T T X ≠ 0 . In many engineering applications the unit eigenvectors are chosen such that X ⋅ X = I where X is the transpose of the eigenvector X . Example 5. Form a set of unit eigenvectors using the eigenvectors of part (b). so that the sum of the squares of their components is equal to unity. The example which follows. This is the same matrix as in Example 5. its eigenvalues. This vector X is called eigenvector. there is a different eigenvector for each Two vectors X and Y are said to be orthogonal if their inner (dot) product is zero.17. Find eigenvectors corresponding to each eigenvalue of A c. and I is the identity matrix. This is done by dividing each component of the eigenvector by the square root of the sum of the squares of their components. Eigenvectors are generally expressed as unit eigenvectors. and eigenvectors. A set of eigenvectors constitutes an orthonormal basis if the set is normalized (expressed as unit eigenvectors) and these vector are mutually orthogonal. illustrates the relationships between a matrix A . where we found the eigenvalues to be λ1 = 1 λ2 = 2 AX = λX x1 X = x2 x3 λ3 = 3 b. it is discussed in Chapter 14.Eigenvectors eigenvalue. Third Edition Copyright © Orchard Publications 5−39 . Find the eigenvalues of A b. that is. An orthonormal basis can be formed with the Gram− Schmidt Orthogonalization Procedure.19 Given the matrix 5 A = 0 2 7 –5 4 –1 8 –3 a. Obviously. We begin with and we let Then.

we obtain x 1 = x 2 . or MATLAB.164) For λ = 1 . an eigenvector for λ = 1 is x1 Xλ = 1 = x2 = x3 2x 2 x2 3x 2 2 2 = x2 1 = 1 3 3 (5. and solve x 1 and x 3 in terms of x 2 . Therefore.165) By Crame’s rule. and x 3 are scalars.Chapter 5 Differential Equations.167). we obtain ( 5 – λ )x 1 0 2x 1 7x 2 ( 4 – λ )x 2 8x 2 – 5x 3 –x3 – ( 3 – λ )x 3 0 = 0 0 (5.162) or 5x 1 0 2x 1 (5. is known. Similarly. and State Equations 5 0 2 x1 7 –5 x1 4 –1 x2 = λ x2 8 –3 x3 x3 7x 2 – 5x 3 4x 2 –x3 8x 2 – 3x 3 λx 1 = λx 2 λx 3 (5. Then. an eigenvector for λ = 2 is 5−40 Numerical Analysis Using MATLAB® and Excel®. (5. Third Edition Copyright © Orchard Publications . we obtain x 1 = 2x 2 . we can assume that one of these. we obtain the indeterminate values x1 = 0 ⁄ 0 x2 = 0 ⁄ 0 x3 = 0 ⁄ 0 (5.163) Equating corresponding rows and rearranging. and x 3 = 2x 2 . and x 3 = 3x 2 . x 2. for λ = 2 .164) reduces to 4x 1 + 7x 2 – 5x 3 = 0 3x 2 – x 3 = 0 2x 1 + 8x 2 – 4x 3 = 0 (5. Then.167) since any eigenvector is a scalar multiple of the last vector in (5. say x 2 . State Variables.166) Since the unknowns x 1.

2 1 3 4 1 9 ⎛ --------. Numerical Analysis Using MATLAB® and Excel®. and x 3 = x 2 .+ ----.170).+ ----.171) and the same is true for the other two unit eigenvectors in (5.= 1 ⎝ 14 ⎠ ⎝ 14 ⎠ ⎝ 14 ⎠ 14 14 14 (5.169) c.⎞ 2 = ----.⎞ 2 + ⎛ --------.⎞ 2 + ⎛ --------. an eigenvector for λ = 3 is x1 Xλ = 3 = x2 = x3 –x2 x2 x2 = x2 –1 –1 = 1 1 1 1 (5. Then.170) We observe that for the first unit eigenvector the sum of the squares is unity. These are: 2 +1 +3 = 1 +1 +2 = ( –1 ) + 1 + 1 = 2 2 2 2 2 2 2 2 2 14 6 3 The unit eigenvectors are 2--------14 1Unit X λ = 1 = --------14 3--------14 1-----6 1Unit X λ = 2 = -----6 2-----6 –1 -----3 1Unit X λ = 3 = -----3 1-----3 (5. We find the unit eigenvectors by dividing the components of each vector by the square root of the sum of the squares of the components. that is.168) Finally. we obtain x 1 = – x 2 .Eigenvectors x1 Xλ = 2 = x2 = x3 x2 x2 2x 2 1 1 = x2 1 = 1 2 2 (5. Third Edition Copyright © Orchard Publications 5−41 . for λ = 3 .

t ) . • The ODE d y d y dy d x d x dx a n -------. and t is said to be a partial differential equation and it is abbreviated as PDE.+ b m – 1 ---------------. where x and t are independent variables. • Generally. if y = f ( x) . • The most general solution of an homogeneous ODE is the linear combination yH ( t ) = k1 y1 ( t ) + k2 y2 ( t ) + k3 y3 ( t ) + … + kn yn ( t ) where the subscript H is used to denote homogeneous and k 1. and State Equations 5.+ … + a 1 ----. the differential equation that relates y . k n are arbitrary constants. 5−42 Numerical Analysis Using MATLAB® and Excel®. The total solution of the non−homogeneous ODE is the summation of the natural and forces responses. State Variables. known as forcing function. Also. in engineering the solution of the homogeneous ODE. and is denoted as y F ( t ) or simply y F .12 Summary • Differential equations are classified by: Type − Ordinary or Partial Order − The highest order derivative which is included in the differential equation Degree − The exponent of the highest power of the highest order derivative after the differential equation has been cleared of any fractions or radicals in the dependent variable and its derivatives • If the dependent variable y is a function of only a single variable x . k 3. is referred to as the natural response. and is denoted as y N ( t ) or simply y N . k 2. that is. Third Edition Copyright © Orchard Publications . that is. also known as the complementary solution. the initial conditions must be satisfied.+ a 0 y = b m --------. If the forcing function is zero. the differential equation which relates y and x is said to be an ordinary differential equation and it is abbreviated as ODE. • A function y = f ( x) is a solution of a differential equation if the latter is satisfied when y and its derivatives are replaced throughout by f ( x) and its corresponding derivatives. the differential equation is referred to as homogeneous differential equation.+ … + b 1 ----. is not zero. The particular solution of a non−homogeneous ODE is be referred to as the forced response.+ b 0 x n n–1 m n–1 dt dt dt dt dt dt n n–1 m m–1 is a non−homogeneous differential equation if the right side. x . …. • If the dependent variable y is a function of two or more variables such as y = f ( x.Chapter 5 Differential Equations.+ a n – 1 --------------.

the n solutions of the natural response are independent and the most general solution is: yN = k1 e s1 t + k2 e s2 t + … + kn e sn t • If the solution of the characteristic equation contains m equal roots. s n are the solutions of the characteristic equation an s + an – 1 s n n–1 + … + a1 s + a0 = 0 and a n.Summary y( t) = y Natural Response +y Forced Response = yN + yF The natural response y N contains arbitrary constants and these can be evaluated from the given initial conditions. a 0 are the constant coefficients of the ODE • If the roots of the characteristic equation are distinct. It is imperative to remember that the arbitrary constants of the natural response must be evaluated from the total response. …. for two complex conjugate roots we evaluate the constants from the expressions k1 e s1 t + k2 e s2 t = e – αt ( k 3 cos βt + k 4 sin β t ) = e – αt k 5 cos ( βt + ϕ ) • The forced response of a non−homogeneous ODE can be found by the method of undetermined coefficients or the method of variation of parameters. s 2. Then. then another root is s 2 = – α – j β . a n – 1.1. the right side f ( t ) cannot be determined by the method of undetermined coefficients. y 2 = k 2 e . a 1. the forced response is a function similar to the right side of the non−homogeneous ODE. …. • For an nth order homogeneous differential equation the solutions are y 1 = k 1 e . y n = k n e s1 t s2 t s3 t sn t where s 1. For these ODEs we must use the method of variation of parameNumerical Analysis Using MATLAB® and Excel®. Third Edition Copyright © Orchard Publications 5−43 . • In certain non−homogeneous ODEs. however. …. The form of the forced response for second order non−homogeneous ODEs is given in Table 5. Thus. contains no arbitrary constants. The forced response y F . these occur as complex conjugate pairs. • With the method of undetermined coefficients. y 3 = k 3 e . if one root is s 1 = – α + jβ where α and β are real numbers. the most general solution has the form: yN = ( k1 + k2 t + … + km t m–1 )e s1 t + kn – i e s2 t + … + kn e sn t • If the characteristic equation contains complex roots.

• For second order ODEs with constant coefficients. and the initial condition.Chapter 5 Differential Equations.y1 + ------.⋅ ------. Third Edition Copyright © Orchard Publications . and State Equations ters. • The state representation of a system can be described by the pair of the of the state space equations · x = Ax + bu y = Cx + du • In a system of state equations of the form · x = αx + βu y = k1 x + k2 u where α .= f ( t ) dt dt dt dt Simultaneous solution of last two expressions above will yield the values of du1 ⁄ dt and du 2 ⁄ dt .⋅ ------. integration of these will produce u 1 and u 2 . then. This method will work with all linear equations including those with variable coefficients provided that the general form of the natural response is known. if non−zero is denoted as x 0 = x ( t 0 ) . The resulting first−order differential equations are called state space equations. State Variables. k 1 . and k 2 are scalar constants.+ ------. • An nth−order differential equation can be resolved to n first−order simultaneous differential equations with a set of auxiliary variables called state variables. β . or simply state equations. the solution of the first state equation above is x( t) = e α ( t – t0 ) t x0 + e αt ∫t e 0 – ατ β u ( τ ) dτ • In a system of state equations of the form · x = Ax + bu y = Cx + du 5−44 Numerical Analysis Using MATLAB® and Excel®. which when substituted into the first will yield the total solution. the method of variation of parameters replaces the constants k 1 and k 2 by two variables u 1 and u 2 that satisfy the following three relations: y = u1 y1 + u2 y2 du2 du1 -----.y2 = 0 dt dt du1 dy1 du 2 dy 2 ------.

the solution of the matrix · differential equation x = Ax + bu with initial conditions x ( t 0 ) = x 0 is obtained from the relation x( t) = e A ( t – t0 ) x0 + e At ∫t e 0 t –A τ bu ( τ ) dτ where the state transition matrix e ϕ(t) ≡ e At At is defined as the matrix power series 1 22 1 33 1 nn = I + At + ---. i = 1. if all eigenvalues of a given matrix A are distinct. This theorem states that a matrix can be expressed as an ( n – 1 )th degree polynomial in terms of the matrix A as e At At = a0 I + a1 A + a2 A + … + an – 1 A 2 n–1 where the coefficients a i are functions of the eigenvalues λ . • If A is an n × n matrix. that is. and b and d are column vectors with two or more rows. λ m + 1 . • If λ 1 ≠ λ 2 ≠ λ 3 ≠ … ≠ λ n . that is. λ n . the eigenvalues λ i . n of A are the roots of the nth order polynomial det [ A – λI ] = 0 • Evaluation of the state transition matrix e is based on the Cayley−Hamilton theorem. the coefficients a i of the state transition matrix e At = a0 I + a1 A + a2 A + … + an – 1 A 2 n–1 Numerical Analysis Using MATLAB® and Excel®. if λ 1 = λ 2 = λ 3 … = λ m . and I be the n × n identity matrix.A t 2! n! 3! and I is the n × n identity matrix.Summary where for two or more simultaneous differential equations A and C are 2 × 2 or higher order matrices. the coefficients a i are found from the simultaneous solution of the following system of equations: a0 + a1 λ1 + a2 λ1 + … + an – 1 λ1 a0 + a1 λ2 + a2 λ2 + … + an – 1 λ2 … a0 + a1 λn + a2 λn + … + an – 1 λn 2 n–1 2 2 n–1 n–1 λ1 t λ2 t = e = e = e λn t • If the polynomial of det [ A – λI ] = 0 has n roots. ….A t + ---. and m of these roots are equal. Third Edition Copyright © Orchard Publications 5−45 . 2.A t + … + ---.

e 1 1 1 dλ 1 dλ 1 d λt d -------2 ( a 0 + a 1 λ 1 + a 2 λ 2 + … + a n – 1 λ n – 1 ) = -------2 e 1 1 1 dλ 1 dλ 1 … λ t d d -------------. the vector X is called eigenvector.( a 0 + a 1 λ 1 + a 2 λ 2 + … + a n – 1 λ n – 1 ) = ------. that is. and λ is a scalar number. State Variables. 5−46 Numerical Analysis Using MATLAB® and Excel®.( a 0 + a 1 λ 1 + a 2 λ 2 + … + a n – 1 λ n – 1 ) = -------------. Obviously.Chapter 5 Differential Equations. This is done by dividing each component of the eigenvector by the square root of the sum of the squares of their components. a0 + a1 λ1 + a2 λ1 + … + an – 1 λ1 2 n–1 = e λ1 t dd.λ t -------. X is a non−zero column vector. there is a different eigenvector for each eigenvalue.e 1 1 1 m–1 m–1 dλ 1 dλ 1 m–1 m–1 2 2 a0 + a1 λm + 1 + a2 λm + 1 + … + an – 1 λm + 1 = e a 0 + a 1 λn + a 2 λ n + … + a n – 1 λ n 2 n–1 2 n–1 λ m + 1t … = e λn t • We can use the MATLAB eig(x) function to find the eigenvalues of an n × n matrix. they are normalized to unit length. • If A is an n × n matrix. Third Edition Copyright © Orchard Publications . and State Equations are found from the simultaneous solution of the system of equations below. so that the sum of the squares of their components is equal to unity. Eigenvectors are generally expressed as unit eigenvectors.

and C below.( cos 2t + 1 ) 2 dt dt 2 4. d y + y = sec t ------2 dt 2 5 Express the integro−differential equation below as a matrix of state equations where k 1.13 Exercises Solve the following ODEs by any method and verify your answers with MATLAB. Express the matrix of the state equations below as a single differential equation. and k 3 are constants. Numerical Analysis Using MATLAB® and Excel®. d y + 4 ----. -------. dy 1.+ 3y = t – 1 ------2 dt 2 2 dt dy –t 2.+ k 2 v + k 1 2 dt dt ∫0 v dt t = sin 3t + cos 3t 6. and let x(y) = y(t) . 2 dv dv ------.+ k 3 ----.+ 3y = 4e ------2 dt dt 2 dy 1 2 2 d y 3.+ y = cos t Hint: Use cos t = -. · x1 · x2 · x3 · x4 x1 0 1 0 0 0 x = 0 0 1 0 ⋅ 2 + 0 u(t) x3 0 0 0 1 0 –1 –2 –3 –4 1 x4 7. Third Edition Copyright © Orchard Publications 5−47 .Exercises 5. d y + 4 ----. B . Compute the eigenvalues of the matrices A . k 2. A = 1 2 3 –1 B = a 0 –a b 0 1 0 C = 0 0 1 – 6 – 11 – 6 Hint: One of the eigenvalues of matrix C is – 1 .+ 2 ----.

State Variables. Third Edition Copyright © Orchard Publications .Chapter 5 Differential Equations. and State Equations 8. Compute e At given that A = 0 1 0 0 0 1 – 6 – 11 – 6 5−48 Numerical Analysis Using MATLAB® and Excel®.

For the forced response. Third Edition Copyright © Orchard Publications 5−49 . By substitution of this assumed solution into the given ODE and using MATLAB to find the first and second derivatives we obtain: y = k1 e + k2 e –t – 3t + k 3 te –t Numerical Analysis Using MATLAB® and Excel®. y = k1 e + k2 e –t – 3t 1 . For the forced response. Thus y N = k 1 e + k 2 e –t – 3t 2 .1 and we assume a solution of the form y F = k 3 t + k 4 and the total solution is y = k1 e + k2 e –t – 3t + k3 t + k4 The first and second derivatives of y are dy ⁄ dt = – k 1 e – 3k 2 e 2 2 –t –t – 3t + k3 – 3t d y ⁄ dt = k 1 e + 9k 2 e and by substitution into the given ODE k 1 e + 9k 2 e –t – 3t + 4 ( – k 1 e – 3k 2 e –t – 3t + k3 ) + 3 ( k1 e + k2 e –t – 3t + k3 t + k4 ) = t – 1 Equating like terms we obtain 4k 3 + 3k 3 t + 3k 4 = t – 1 3k 3 t = t 4k 3 + 3k 4 = – 1 and simultaneous solution of the last two yields k 3 = 1 ⁄ 3 and k 4 = – 7 ⁄ 9 .-+ -.14 Solutions to End−of−Chapter Exercises 1. we refer to Table 5. The characteristic equation of the homogeneous part is s + 4s + 3 = 0 from which s 1 = – 1 and s 2 = – 3 . Therefore.1 and we assume a solution of –t –t –t the form y F = k 3 te where we multiplied e by t to avoid the duplication with k 1 e . we refer to Table 5.Solutions to End−of−Chapter Exercises 5.t – 7 3 9 Check with MATLAB: y=dsolve('D2y+4*Dy+3*y=t−1’). y=simple(y) y = -7/9+1/3*t+C1/exp(t)+C2/exp(t)^3 2. The characteristic equation of the homogeneous part is the same as for Exercise 1 and thus yN = k1 e + k2 e –t – 3t .

2). f1=simple(y1) % Define symbolic variables % Assumed form of total solution % Compute and simplify first derivative f1 = -k3*exp(-t)*(-1+t) Thus.Chapter 5 Differential Equations. y1=diff(y0). and State Equations We will use MATLAB to find the first and second derivatives of this expression. The characteristic equation yields two equal roots s 1 = s 2 = – 1 and thus the natural response has the form y N = k 1 e + k 2 te –t –t For the forced response we assume a solution of the form y F = k 3 cos 2t + k 4 sin 2t + k5 5−50 Numerical Analysis Using MATLAB® and Excel®. f2=simple(y2) –t –t % Compute and simplify second derivative f2 = k3*exp(-t)*(-2+t) and the second derivative of y is d yF ⁄ dt = – 2k 3 e + k 3 te f=y2+4*y1+3*y0. State Variables. Third Edition Copyright © Orchard Publications . Therefore. syms t k3 y0=k3*t*exp(−t). y = k1 e + k2 e –t + 2te –t Check with MATLAB: y=dsolve('D2y+4*Dy+3*y=4*exp(−t)'). 3. y=simple(y) 2*t/exp(t)-1/exp(t)+C1/exp(t)+C2/exp(t)^3 We observe that the second and third terms of the displayed expression above have the same form and thus they can be combined to form a single term C3/exp(t). the first derivative of y F is dyF ⁄ dt = k 3 e – k 3 te y2=diff(y0. f=simple(f) 2 2 –t –t % Form and simplify the left side of the given ODE f = 2*k3/exp(t) and by substitution into the given ODE 2k 3 e –t = 4e – 3t –t or k 3 = 2 .

the first derivative of y F is dyF ⁄ dt = – 2k 3 sin 2t + 2k 4 cos 2t y2=diff(y0.2). Therefore.– -------------------------------------2 50 Check with MATLAB: Numerical Analysis Using MATLAB® and Excel®. f=simple(f) % Form and simplify the left side of the given ODE 2 2 f = -3*k3*cos(2*t)-3*k4*sin(2*t)-4*k3*sin(2*t)+4*k4*cos(2*t)+k5 Simplifying this expression and equating with the right side of the given ODE we obtain: cos 2t -( – 3k 3 + 4k 4 ) cos 2t – ( 4k 3 + 3k 4 ) sin 2t + k 5 = ------------. the forced response is y F = ( – 3 ⁄ 50 ) cos 2t + ( 4 ⁄ 50 ) sin 2t + 1 ⁄ 2 and the total response is –t – t 1 3 cos 2t – 4 sin 2t y = k 1 e + k 2 te + -. y1=diff(y0). f2=simple(y2) % Compute and simplify second derivative f2 = -4*k3*cos(2*t)-4*k4*sin(2*t) and the second derivative of y is d yF ⁄ dt = – 4k 3 cos 2t – 4k 4 sin 2t f=y2+2*y1+y0.Solutions to End−of−Chapter Exercises We will use MATLAB to find the first and second derivatives of this expression.+ 1 2 2 Equating like terms and solving for the k terms we obtain – 3k 3 + 4k 4 = 1 ⁄ 2 – 4k 3 – 3k 4 = 0 k5 = 1 ⁄ 2 Simultaneous solution of the first two equations above yields k 3 = – 3 ⁄ 50 and k 4 = 4 ⁄ 50 . f1=simple(y1) % Define symbolic variables % Assumed form of total solution % Compute and simplify first derivative f1 = -2*k3*sin(2*t)+2*k4*cos(2*t) Thus. Third Edition Copyright © Orchard Publications 5−51 . syms t k1 k2 k3 k4 k5 y0=k3*cos(2*t)+k4*sin(2*t)+k5.

we will use the method of variation of parameters.= f ( t ) = ------.Chapter 5 Differential Equations.⋅ ------. to assume a solution for the forced response of this ODE.69).68) the solution is y = u1 y 1 + u 2 y 2 = u1 cos t + u 2 sin t (1) Also.= ---------------------------------------.⋅ ------.= – tan t (2) 2 2 cos t sin t dt 1 cos t + sin t – sin t cos t 0 sin t and cos t 0 du 2 – sin t sec t 1 ------.= -----------------------------. du1 du 2 ------. Third Edition Copyright © Orchard Publications .cos t + ------. The characteristic equation is s + 1 = 0 from which s = ± j and thus the natural response is yN = k1 e + k2 e jt –j t 2 We let y 1 = cos t and y 2 = sin t Then.+ ------.( – sin t ) + ------.= -----------. du1 dy1 du 2 dy 2 du 2 du1 ------.70).y 2 = 0 dt dt or du 2 du1 ------.= ---------------------------------------. and State Equations y=dsolve('D2y+2*Dy+y=cos(2*t)/2+1/2').= -. f=simple(y) f = -3/50*cos(2*t)+2/25*sin(2*t)+1/2+C1*exp(-t)+C2*exp(-t)*t 4.( cos t ) = sec t dt dt dt dt dt dt Next. It is very difficult. from (5. if not impossible.sin t = 0 dt dt and from (5. State Variables.= 1 dt 1 1 (3) Integration of (2) and (3) above and substitution into (1) yields u1 = ∫ ( – tan t ) dt = – ( – ln cos t ) + k 1 = ln cos t + k 1 u2 = ∫ dt = t + k2 5−52 Numerical Analysis Using MATLAB® and Excel®. Therefore.y1 + ------. by (5. we find du1 ⁄ dt and du 2 ⁄ dt by Cramer’s rule as follows: sin t – --------du1 sec t cos t cos t – tan t ------.

+ k dv + k v = 3 cos 3t – 3 sin 3t = 3 ( cos 3t – sin 3t ) ------3 ------2 ----1 2 3 dt dt dt 2 or 3 dv dv dv ------.+ k dv .= – k 3 ------. Third Edition Copyright © Orchard Publications 5−53 .= x 3 3 dt 2 · dv ------. the state equations are · x1 = x2 · x2 = x3 · x 3 = – k 1 x 1 – k 2 x 2 – k 3 x 3 + 3 ( cos 3t – sin 3t ) and in matrix form · x1 x1 0 1 0 0 · = 0 0 1 ⋅ x + 0 ⋅ 3 ( cos 3t – sin 3t ) x2 2 –k1 –k2 –k3 · 1 x3 x3 Numerical Analysis Using MATLAB® and Excel®. Differentiating the given integro−differential equation with respect to t we obtain 3 dv .– k 2 ----. f=simple(y) f = sin(t)*t+log(cos(t))*cos(t)+C1*sin(t)+C2*cos(t) 5.Solutions to End−of−Chapter Exercises y = u1 y 1 + u 2 y 2 = ( ln cos t + k 1 ) cos t + ( t + k 2 ) sin t = k 1 cos t + k 2 sin t + t sin t + cos t ( ln cos t ) Check with MATLAB: y=dsolve('D2y+y=sec(t)').= x 3 = x 2 2 dt Then. and by substitution into (1) · x 3 = – k 1 x 1 – k 2 x 2 – k 3 x 3 + 3 ( cos 3t – sin 3t ) Thus.– k 1 v + 3 ( cos 3t – sin 3t ) (1) 2 3 dt dt dt 2 We let v = x1 · dv ----.= x 2 = x 1 dt 3 · dv ------.

+ 3 ------. Matrix A is the same as Matrix C in Exercise 7. Then. and thus λ 1 = a λ2 = b c. λ2 = –2 λ1 = –3 a.= λ + 5λ + 6 ⇒ ( λ + 1 ) ( λ + 2 ) ( λ + 3 ) = 0 (λ + 1) 3 2 2 3 2 and thus λ 1 = – 1 8. a.Chapter 5 Differential Equations. Expansion of the given matrix yields · x1 = x2 · x2 = x3 · x3 = x2 3 · x 4 = – x 1 – 2x 2 – 3x 3 – 4x 4 + u ( t ) 2 Letting x = y we obtain 4 dy dy dy dy ------. λ1 = –1 λ2 = –2 λ1 = –3 5−54 Numerical Analysis Using MATLAB® and Excel®. A = 1 2 3 –1 ⎛ ⎞ 2 det ( A – λI ) = det ⎜ 1 2 – λ 1 0 ⎟ = det 1 – λ = 0 ⎝ 3 –1 0 1⎠ 3 –1 –λ 2 2 ( 1 – λ ) ( – 1 – λ ) – 6 = 0 . and thus λ 1 = 7 λ2 = – 7 b. – 1 – λ + λ + λ – 6 = 0 .+ 2 ----. 2 λ + 6λ + 11λ + 6 --------------------------------------------.+ y = u ( t ) 3 2 4 dt dt dt dt 7. State Variables.+ 4 ------. Third Edition Copyright © Orchard Publications . Then. and State Equations 6. B = a 0 –a b ⎛ ⎞ det ( B – λI ) = det ⎜ a 0 – λ 1 0 ⎟ = det a – λ 0 = 0 ⎝ –a b 0 1⎠ –a b – λ ( a – λ ) ( b – λ ) = 0 . 0 1 0 C = 0 0 1 – 6 – 11 – 6 ⎛ 0 1 0 1 0 0⎞ ⎜ ⎟ det ( C – λI ) = det ⎜ 0 0 1 –λ 0 1 0 ⎟ ⎜ ⎟ 0 0 1⎠ ⎝ – 6 – 11 – 6 = det –λ 1 0 0 –λ 1 =0 – 6 – 11 – 6 – λ λ ( – 6 – λ ) – 6 – ( – 11 ) ( – λ ) = λ + 6λ + 11λ + 6 = 0 and it is given that λ 1 = – 1 . λ = 7 .

1/2*exp(-t)-4*exp(-2*t)+9/2*exp(3*t)] Then. we compute e At of (1) with the following MATLAB code: syms t. disp('a1 = '). eAt=a0*eye(3)+a1*A+a2*A^2 eAt = [3*exp(-t)-3*exp(-2*t)+exp(-3*t). disp(x(3)) a0 = 3*exp(-t)-3*exp(-2*t)+exp(-3*t) a1 = 5/2*exp(-t)-4*exp(-2*t)+3/2*exp(-3*t) a2 = 1/2*exp(-t)-exp(-2*t)+1/2*exp(-3*t) Thus. -1/2*exp(-t)+2*exp(-2*t)-3/2*exp(-3*t)] [3*exp(-t)-12*exp(-2*t)+9*exp(-3*t). fprintf(' \n')... disp('a0 = ').5e – 4e a 2 = 0. a2=1/2*exp(−t)−exp(−2*t)+1/2*exp(−3*t). -5/2*exp(-t)+8*exp(2*t)-9/2*exp(-3*t).Solutions to End−of−Chapter Exercises and since A is a 3 × 3 matrix the state transition matrix is e At = a0 I + a1 A + a2 A 2 2 2 λ1 t λ2 t λ3 t 2 (1) –t – 2t – 3t Then. 0 0 1. Third Edition Copyright © Orchard Publications 5−55 ... a 0 = 3e – 3e –t –t –t – 2t + 3e – 3t – 3t a 1 = 2. fprintf(' \n'). disp(x(1)). A=[1 −1 1..5e – 3t Now. −6 −11 −6].5e – e – 2t + 1. 5/2*exp(-t)-16*exp(2*t)+27/2*exp(-3*t). a1=5/2*exp(−t)−4*exp(−2*t)+3/2*exp(−3*t). 1 −3 9]. 5/2*exp(-t)-4*exp(-2*t)+3/ 2*exp(-3*t)... 1/2*exp(-t)-exp(-2*t)+1/2*exp(-3*t)] [-3*exp(-t)+6*exp(-2*t)-3*exp(-3*t).5e – 2t + 0. x=A\a. 1 −2 4. Numerical Analysis Using MATLAB® and Excel®. exp(−2*t).. A=[0 1 0. disp('a2 = ')... exp(−3*t)]'). disp(x(2)). a=sym('[exp(−t). a0 + a1 λ1 + a2 λ1 = e a0 + a1 λ2 + a2 λ2 = e a0 + a1 λ3 + a2 λ3 = e ⇒ a0 – a1 + a2 = e ⇒ a 0 – 2a 1 + 4a 2 = e ⇒ a 0 – 3a 1 + 9a 2 = e syms t.. a0=3*exp(−t)−3*exp(−2*t)+exp(−3*t)..

State Variables.5e – 16e –t – 2t – 4.5e – 3t 5−56 Numerical Analysis Using MATLAB® and Excel®.5e – 4e – 2t – 1. Third Edition Copyright © Orchard Publications .5 e + 8e 2.Chapter 5 Differential Equations.5e – 2t + 4. and State Equations 3e – 3e 3e – 12e –t –t – 2t e At +e – 3t 2.5e – 0.5e – 3t – 3t – 3t 0.5e – 4e –t –t – 2t + 1.5e – e –t –t –t – 2t + 0.5e + 9e – 3t – 2t + 13.5e – 3t – 3t = – 3 e – t + 6e – 2t – 3e –3t – 2t – 2.5 e + 2e 0.

2) where the first term a 0 ⁄ 2 is a constant. or current i ( t ) . sinusoids whose frequencies are multiples of a fundamental frequency (or first harmonic).a 0 + a 1 cos ωt + a 2 cos 2ωt + a 3 cos 3ωt + a 4 cos 4ωt + … 2 + b 1 sin ωt + b 2 sin 2ωt + b 3 sin 3ωt + b 4 sin 4ωt + … (6. and represents the DC (average) component of f ( t ) . Numerical Analysis Using MATLAB® and Excel®. and so on. We begin with the definition of sinusoids that are harmonically related and the procedure for determining the coefficients of the trigonometric form of the series.1 Wave Analysis The French mathematician Fourier found that any periodic waveform. a series of sinusoids with frequencies 1 MHz . a third harmonic of 3 MHz . The alternate trigonometric and the exponential forms are also presented. 6. the terms with the coefficients a 2 and b 2 together. We conclude with a discussion on power series expansion with the Taylor and Maclaurin series. if f ( t ) represents some voltage v ( t ) . The terms with the coefficients a 1 and b 1 together. For example. represent the second harmonic component 2ω . 3 MHz .1) or 1 f ( t ) = -. that is. the term a 0 ⁄ 2 is the average value of v ( t ) or i ( t ) . and so on.Chapter 6 Fourier. 2 MHz . i. represent the fundamental frequency component ω *. a second harmonic of 2 MHz . and so on. we discuss the different types of symmetry and how they can be used to predict the terms that may be present. Taylor. Third Edition Copyright © Orchard Publications 6−1 . can be expressed as a series of harmonically related sinusoids. * We recall that k 1 cos ωt + k2 sin ωt = k cos ( ωt + θ ) where θ is a constant. contains the fundamental frequency of 1 MHz . and Maclaurin Series T his chapter is an introduction to Fourier and power series. Several examples are presented to illustrate the approach. Likewise. a waveform that repeats itself after some time. any periodic waveform f ( t ) can be expressed as 1 f ( t ) = -. Then. In general..e. Thus.a 0 + 2 n=1 ∑ ( a cos nωt + b sin nωt ) n n ∞ (6.

1. the sum of two or more sinusoids of different frequencies produce a waveform that is not a sinusoid as shown in Figure 6. Summation of a fundamental.5) 2π cos m t dt = 0 ( sin mt ) ( cos nt ) dt = 0 The integrals of (6. Taylor.4) (6. and so on. Let us consider the functions sin mt and cos m t where m and n are any integers.5) is also is zero since 2π 2π sin mt dt = 0 (6. the product of the sine and cosine functions under the integral evaluated from 0 to 2π is zero. This will be shown shortly. Then. second and third harmonic 6.2 Evaluation of the Coefficients Evaluations of a i and b i coefficients of (6.1. the second harmonic. 2nd Harmonic Total 3 2 1 0 -1 -2 -3 0 2 4 6 8 10 12 Fundamental 3rd Harmonic Figure 6. The inte- 6−2 Numerical Analysis Using MATLAB® and Excel®. must produce the waveform f ( t ) . it follows that the sum of the DC .4) are zero since the net area over the 0 to 2π area is zero. Generally.3) and (6.1) is not a difficult task because the sine and cosine are orthogonal functions. and for convenience.3) (6. that is. and Maclaurin Series Since any periodic waveform f ( t ) can be expressed as a Fourier series. we have assumed that ω = 1 .Chapter 6 Fourier. Third Edition Copyright © Orchard Publications . the fundamental. ∫0 ∫0 ∫0 gral of (6.

if m and n are different integers. where we observe that the net shaded area above and below the time axis is zero.3. Third Edition Copyright © Orchard Publications 6−3 . Graphical proof of ∫0 2π ( sin mt ) ( cos nt ) dt = 0 Moreover.2. sin x sin x ⋅ cos x cos x Figure 6.6) 1 ( sin x ) ( sin y ) = -. where m = 2 and n = 3 .2. then.Evaluation of the Coefficients 1 sin x cos y = -.[ cos ( x – y ) – cos ( x – y ) ] 2 The integral of (6.[ sin ( x + y ) + sin ( x – y ) ] 2 This is also obvious from the plot of Figure 6. We observe that the net shaded area above and below the time axis is zero.6) can also be confirmed graphically as shown in Figure 6. sin 2x sin 3x sin 2x ⋅ sin 3x Numerical Analysis Using MATLAB® and Excel®. ∫0 since 2π ( sin mt ) ( sin nt ) dt = 0 (6.

Graphical proof of ∫0 2π ( cos m t ) ( cos nt ) dt = 0 for m = 2 and n = 3 However.Chapter 6 Fourier.4.9) can also be seen to be true graphically with the plots of Figures 6.4. ∫0 since 2π ( cos m t ) ( cos nt ) dt = 0 (6. Taylor. and Maclaurin Series Figure 6.7) can also be confirmed graphically as shown in Figure 6.8) ∫0 2π ( cos m t ) dt = π 2 (6.9) The integrals of (6. if m and n are different integers.7). We observe that the net shaded area above and below the time axis is zero. It was stated earlier that the sine and cosine functions are orthogonal* to each other. ∫0 and 2π ( sin mt ) dt = π 2 (6. cos 3x cos 2x cos 2x ⋅ cos 3x Figure 6. Third Edition Copyright © Orchard Publications . then. Graphical proof of ∫0 2π ( sin mt ) ( sin nt ) dt = 0 for m = 2 and n = 3 Also.8) and (6. then. m = n . where m = 2 and n = 3 .[ cos ( x + y ) + cos ( x – y ) ] 2 The integral of (6.5 and 6.6) and (6. if in (6.3.6. The simpli* We will discuss orthogonal functions in Chapter 14 6−4 Numerical Analysis Using MATLAB® and Excel®.7) 1 ( cos x ) ( cos y ) = -.

1). 1 f ( t ) sin 2t = -. we let ω = 1 .Evaluation of the Coefficients fication obtained by application of the orthogonality properties of the sine and cosine functions. Graphical proof of ∫0 2π ( cos m t ) dt = π 2 In (6. say b 2 .5. for simplicity. Graphical proof of ∫0 2π ( sin mt ) dt = π 2 cos x cos x 2 Figure 6. Third Edition Copyright © Orchard Publications 6−5 . 1 f ( t ) = -. becomes apparent in the discussion that follows.a 0 sin 2t + a 1 cos t sin 2t + a 2 cos 2t sin 2t + a 3 cos 3t sin 2t + a 4 cos 4t sin 2t + … 2 b 1 sin t sin 2t + b 2 ( sin 2t ) + b 3 sin 3t sin 2t + b 4 sin 4t sin 2t + … 2 Numerical Analysis Using MATLAB® and Excel®. Then. we multiply both sides of (6. sin x sin x 2 Figure 6. Then.a 0 + a 1 cos t + a 2 cos 2t + a 3 cos 3t + a 4 cos 4t + … 2 + b 1 sin t + b 2 sin 2t + b 3 sin 3t + b 4 sin 4t + … (6.6.10) To evaluate any coefficient.10) by sin 2t .

we multiply both sides of the above expression by dt . (6. The remaining coefficients can be evaluated similarly.13) 1 b n = -π ∫0 2π f ( t ) sin nt dt (6.14) 6−6 Numerical Analysis Using MATLAB® and Excel®. ∫0 2π 1 f ( t ) sin 2t dt = -.a 0 = ----2 2π 1 a n = -π 2π ∫0 2π f ( t ) dt (6. Taylor. and b n are found from the following relations. and Maclaurin Series Next.11) reduces to ∫0 or 2π f ( t ) sin 2t dt = b 2 ∫0 2π ( sin 2t ) dt = b 2 π 2 1 b 2 = -π ∫0 2π f ( t ) sin 2t dt and thus we can evaluate this integral for any given function f ( t ) . and we integrate over the period 0 to 2π .Chapter 6 Fourier. Then. The coefficients a 0 .11) sin 3t sin 2t dt 2π sin 4t sin 2t dt + … We observe that every term on the right side of (6. a n .6) and (6. Therefore.a 0 2 ∫0 2π sin 2t dt + a 1 ∫0 2π cos t sin 2t dt + a 2 ∫0 2π cos 2t sin 2t dt + a3 + b1 + b4 ∫0 2π cos 3t sin 2t dt + a 4 sin t sin 2t dt + b 2 ∫0 2π cos 4t sin 2t dt + … 2 ∫0 ∫0 2π ∫0 2π ( sin 2t ) dt + b 3 ∫0 2π (6. Third Edition Copyright © Orchard Publications .11) except the term b2 ∫0 2π ( sin 2t ) dt 2 is zero as we found in (6. 1 1-.7).12) ∫0 f ( t ) cos nt dt (6.

These are: 1. polynomials with even exponents only. Fortunately. In other words. it is possible to predict which terms will be present in the trigonometric Fourier series. Some waveforms have cosine terms only. and odd functions have nothing to do with odd harmonics. all the a i coefficients including a 0 . if in an even function we replace t with – t . if it is an even function. Even symmetry − If a waveform has even symmetry. Odd symmetry − If a waveform has odd symmetry. the function f ( t ) does not change.6. Numerical Analysis Using MATLAB® and Excel®. and sine terms all present. In other words.t. the most common waveforms used in science and engineering. and a 0 may or may not be zero. 2. and with or without constants. * We will discuss power series later in this chapter. by observing whether or not the given waveform possesses some kind of symmetry.+ … 2! 4! 6! 2 4 6 Other examples of even functions are shown in Figure 6. Third Edition Copyright © Orchard Publications 6−7 . only odd (odd cosine and odd sine) harmonics will be present.12) yields the average ( DC ) value of f ( t ) . the cosine function is an even function because it can be written as the power series* t. if f ( t ) is an odd function.+ ---. We will discuss three types of symmetry that can be used to facilitate the computation of the trigonometric Fourier series form. the series will consist of cosine terms only.7. Half−wave symmetry − If a waveform has half−wave symmetry (to be defined shortly). 3.Symmetry The integral of (6. will be zero. while others have sine terms only. 6. For instance.3 Symmetry With a few exceptions such as the waveform of Example 6.15) that is. if f ( t ) is an even function. do not have the average. cosine. In other words. We will now define even and odd functions and we should remember that even functions have nothing to do with even harmonics. Thus. the series will consist of sine terms only. that is. are even functions. all even (even cosine and even sine) harmonics will be zero.– ---. f ( –t ) = f ( t ) (6. if it is an odd function. that is.tcos t = 1 – ---. all the b i coefficients will be zero. Still other waveforms have or have not DC components. A function f ( t ) is an even function of time if the following relation holds.

Henceforth. 2 The product of two even or two odd functions is an even function. Examples of even functions A function f ( t ) is an odd function of time if the following relation holds. the reverse is not always true.16) that is. For instance.8.17) Numerical Analysis Using MATLAB® and Excel®. polynomials with odd exponents only. Thus. f ( 0 ) = 0 . we obtain the negative of the function f ( t ) . Taylor. if in an odd function we replace t with – t . ∫– T 6−8 T f e ( t ) dt = 2 ∫0 T f e ( t ) dt (6. f e ( t ) and f o ( t ) will be used to represent even and odd functions of time respectively. that is.7. Thus. f(t) mt 0 f(t) t3 t 0 f(t) t 0 t Figure 6. Third Edition Copyright © Orchard Publications . we will denote an even function with the subscript e .+ ---.+ … 3! 5! 7! 3 5 7 Other examples of odd functions are shown in Figure 6.7. is an odd function. and no constants are odd functions. and the product of an even function times an odd function.– ---. if f ( 0 ) = 0 . Examples of odd functions We observe that for odd functions. However. the sine function is an odd function because it can be written as the power series t t t sin t = t – ---.Chapter 6 Fourier.8. and an odd function with the subscript o . and Maclaurin Series f(t) t2 0 f(t) k t 0 t2 + k t f(t) 0 t Figure 6. An example of this is the function f ( t ) = t in Figure 6. –f ( –t ) = f ( t ) (6. we should not conclude that f ( t ) is an odd function. Also.

but inverted. the shape of the negative half−cycle of the waveform is the same as that of the positive half−cycle.16) with (6. 1.21) that is. we recall that any periodic function with period T .17).22) that is.[ f ( t ) + f ( – t ) ] 2 (6.18) A function f ( t ) that is neither even nor odd can be expressed as 1 f e ( t ) = -.9. we get f ( t ) = fe ( t ) + fo ( t ) (6. is expressed as f(t) = f(t + T) (6. and therefore. Square waveform For the waveform of Figure 6.[ f ( t ) – f ( – t ) ] 2 (6. Third Edition Copyright © Orchard Publications 6−9 .20) By addition of (6.19) or as 1 f o ( t ) = -. the function with value f ( t ) at any time t . a 0 = 0 . has half−wave symmetry if –f ( t + T ⁄ 2 ) = f ( t ) (6. A periodic waveform with period T . any function of time can be expressed as the sum of an even and an odd function.23) that is.Symmetry and T ∫– T f o ( t ) dt = 0 (6. We will test the waveforms of Figures 6.9 through 6. the average value over one period T is zero. It is also an odd function and has half − wave symmetry since – f ( – t ) = f ( t ) and –f ( t + T ⁄ 2 ) = f ( t ) . will have the same value again at a later time t + T. Numerical Analysis Using MATLAB® and Excel®. To understand half−wave symmetry.13 for any of the three types of symmetry.

Square waveform with ordinate axis shifted If we shift the ordinate axis π ⁄ 2 radians to the right. Taylor.Chapter 6 Fourier. Also.9.10. a 0 = 0 . the average value over one period T is zero and therefore. 2. such as f ( a ) and f ( b ) . as shown in Figure 6.9. It is also an odd function because – f ( – t ) = f ( t ) . but has no half−wave symmetry 6−10 Numerical Analysis Using MATLAB® and Excel®. we see that the square waveform now becomes an even function and has half−wave symmetry since f ( – t ) = f ( t ) and – f ( t + T ⁄ 2 ) = f ( t ) . T A −π/2 π/2 2π π −2π −π 0 ωt T/2 −A T/2 Figure 6.10. a 0 = 0 . 3. Square waveform test for symmetry An easy method to test for half−wave symmetry is to choose any half−period T ⁄ 2 length on the time axis as shown in Figure 6. Sawtooth waveform For the sawtooth waveform of Figure 6. if the ordinate axis is shifted by any other value other than an odd multiple of π ⁄ 2 . these will always be equal but will have opposite signs as we slide the half−period T ⁄ 2 length to the left or to the right on the time axis at non−zero values of f ( t ) . Third Edition Copyright © Orchard Publications . the waveform will have neither odd nor even symmetry. and observe the values of f ( t ) at the left and right points on the time axis. and Maclaurin Series T A T/2 f(b) π 2π 0 f(a) T/2 ωt −A Figure 6. If there is half−wave symmetry.11. Square waveform with ordinate shifted by π ⁄ 2 Obviously.

and third harmonics for half−wave symmetry. when separated by T ⁄ 2 . second.13 shows a fundamental. is chosen as the half period of the period of the fundamental frequency. and third harmonic of a typical sinewave where the half period T ⁄ 2 .11. The third harmonic has half−wave symmetry since the c and – c values. The second harmonic has no half−wave symmetry because the ordinates b on the left and b on the right. Moreover. a 0 = 0 . Second and Third Harmonics of a Sinusoid Figure 6. This is necessary in order to test the fundamental. Triangular waveform For this triangular waveform of Figure 6. Also. Third Edition Copyright © Orchard Publications 6−11 . although are equal. second. Triangular waveform test for symmetry 5. the average value over one period T is zero and therefore. The fundamental has half−wave symmetry since the a and – a values. It is also an odd function since – f ( – t ) = f ( t ) . Sawtooth waveform test for symmetry 4. are equal and opposite. there are not opposite in sign. all three waveforms have zero average value unless the abscissa axis is shifted up or down. These waveforms can be either odd or even depending on the position of the ordinate.12.Symmetry since – f ( t + T ⁄ 2 ) ≠ f ( t ) T A −2π −π 0 T/2 π T/2 2π ωt −A Figure 6. it has half−wave symmetry because – f ( t + T ⁄ 2 ) = f ( t ) T A −2π −π T/2 0 π T/2 2π ωt −A Figure 6. Numerical Analysis Using MATLAB® and Excel®. when separated by T ⁄ 2 are equal and opposite.12. Fundamental.

second. and third harmonic test for symmetry In the expressions of the integrals in (6. Example 6. if the waveform has half− wave symmetry and is also an odd or an even function. Fundamental. 6−12 Numerical Analysis Using MATLAB® and Excel®. Page 6−6. This point will be illustrated in Example 6. or has half−wave symmetry. However.5 -1 c a b −a 0 b T/2 (for fundamental) 2 4 T/2 6 (for 2nd harmonic) 8 T/2 10 (for 3rd harmonic) −c 12 Figure 6. it is important to remember that when using these shortcuts. Third Edition Copyright © Orchard Publications . Taylor. and also that the product of two odd functions results also in an even function. that is.2.14. we can compute the non−zero coefficients a n and b n by integrating from 0 to π only.5 0 -0.12) through (6.13. The proof is based on the fact that.4 Waveforms in Trigonometric Form of Fourier Series We will now derive the trigonometric Fourier series of the most common periodic waveforms. we can choose the limits of integration as – π to +π . if the given waveform is an odd function. 6. one period T . or an even function. and Maclaurin Series 3rd harmonic 2nd harmonic Fundamental 1 0. we must evaluate the coefficients a n and b n for the integer values of n that will result in non−zero coefficients. we can choose the limits of integration from 0 to π ⁄ 2 and multiply the integral by 4 . the product of two even functions is another even function. Moreover.14). Also. Of course. and multiply the integral by 2 . the limits of integration for the coefficients a n and b n are given as 0 to 2π .Chapter 6 Fourier.1 Compute the trigonometric Fourier series of the square waveform of Figure 6.

Also.26) yields Numerical Analysis Using MATLAB® and Excel®. all a i coefficients are zero.( – cos n t 0 + cos nt nπ 2π ) π AA= ----. but if we attempt to verify this using (6.14). However. this waveform is an odd function.25) The b i coefficients are found from (6. we will compute all coefficients to verify this. Then.( sin nπ – 0 – sin n2π + sin nπ ) = ----.24).12). Third Edition Copyright © Orchard Publications 6−13 .( 2 sin nπ – sin n2π ) nπ nπ (6. that is.24) and since n is an integer (positive or negative) or zero.( 1 – 2 cos nπ + cos 2nπ ) nπ nπ (6. since the square waveform has odd symmetry. only odd harmonics will be present since this waveform has half−wave symmetry. we will get the indeterminate form 0 ⁄ 0 .26) For n = even .( – cos nπ + 1 + cos 2nπ – cos nπ ) = ----. Square waveform for Example 6. the average ( DC ) value is zero.( π – 0 – 2π + π ) = 0 π (6.1 Solution: The trigonometric series will consist of sine terms only because. Also. 1 b n = -π ∫0 2π 1 f ( t ) sin nt dt = -π ∫0 π A sin nt dt + ∫π 2π Aπ ( – A ) sin nt dt = ----.Waveforms in Trigonometric Form of Fourier Series T A π 0 −A 2π ωt Figure 6.24) are zero and therefore. we will evaluate a 0 directly from (6. Moreover. as we already know. we will assume that ω = 1 . the terms inside the parentheses on the second line of (6. by inspection.14. for brevity. as expected.( sin nt 0 – sin nt nπ 2π ) π A A = ----. (6. The a i coefficients are found from 1 a n = -π ∫0 2π 1 f ( t ) cos nt dt = -π ∫0 π A cos nt dt + ∫π 2π A π ( – A ) cos nt dt = ----. 1 a 0 = -π ∫0 π A dt + ∫π 2π A ( – A ) dt = --. To work around this problem.

( 1 – 2 + 1 ) = 0 nπ as expected.( 1 + 2 + 1 ) = 4A nπ nπ and thus 4A b 1 = -----π 4A b 3 = -----3π 4A b 5 = -----5π and so on. Example 6.27) It was stated above that. Solution: Since the waveform is an odd function and has half−wave symmetry. 1 b n = 4 -π ∫0 π⁄2 4A f ( t ) sin nt dt = -----. if the given waveform has half−wave symmetry. and multiply the integral by 4 . Therefore. and Maclaurin Series A b n = ----. we are only concerned with the odd b n coefficients. Third Edition Copyright © Orchard Publications . (6. Taylor. since the square waveform has half−wave symmetry.28) For n = odd . For n = odd .+ 1⎞ ⎠ 2 nπ ⎝ (6.⎛ sin ωt + -.( – cos n t nπ π⁄2 0 π 4A ) = -----. (6.sin nωt n (6.sin 3ωt + -.Chapter 6 Fourier. We will apply this property to the following example. we can integrate from 0 to π ⁄ 2 .⎛ – cos n -.28) becomes 6−14 Numerical Analysis Using MATLAB® and Excel®.2 Compute the trigonometric Fourier series of the square waveform of Example 1 by integrating from 0 to π ⁄ 2 . Then. and it is also an odd or an even function. the trigonometric Fourier series for the square waveform with odd symmetry is 4A 1 1 4A f ( t ) = -----.sin 5ωt + …⎞ = -----⎠ π 5 3 π ⎝ n = odd ∑ 1 -. and multiplying the result by 4 .21) reduces to A-----b n = ----.

3 Since the waveform has half−wave symmetry and is an even function. 5. 9. and thus the series is the same as in Example 1. Third Edition Copyright © Orchard Publications 6−15 . and so on. except that the ordinate has been shifted to the right by π ⁄ 2 radians.Waveforms in Trigonometric Form of Fourier Series 4A -----b n = -----. it still has half−wave symmetry. For n = odd .30) becomes Numerical Analysis Using MATLAB® and Excel®. and has become an even function. Example 6. and thus all even harmonics are zero as expected. Therefore.( – 0 + 1 ) = 4A nπ nπ (6.⎛ sin n -. Waveform for Example 6. will alternate between +1 and – 1 depending π 2 on the odd integer assigned to n .( sin nt nπ π⁄2 ) 0 π 4A = -----. the average ( DC ) value is zero. 13 .15. Thus.30) We observe that for n = even .. and multiply the integral by 4 . all a n coefficients are zero.31) For n = 1. the trigonometric Fourier series will consist of odd cosine terms only.⎞ 2⎠ nπ ⎝ (6. Solution: This is the same waveform as in Example 6. A T π/2 0 π 3π / 2 2π ωt −A Figure 6. by inspection.30) that sin n -. (6. we observe from (6. 4A a n = ± -----nπ (6.29) as before.3 Compute the trigonometric Fourier series of the square waveform of Figure 6. it will suffice to integrate from 0 to π ⁄ 2 .1. Also. However.15. The a n coefficients are found from 1 a n = 4 -π ∫0 π⁄2 4 f ( t ) cos nt dt = -π ∫0 π⁄2 4A A cos nt dt = -----.

cos ωτ – -. and afterwards we want to recompute the series with reference to a different ordinate. 6−16 Numerical Analysis Using MATLAB® and Excel®.sin 3 ⎛ ωτ + -. 15 . Third Edition Copyright © Orchard Publications .⎞ + -. we can use the result of Example 6.⎞ + -.⎞ + … ⎝ π 2⎠ 2⎠ 5 ⎝ 2⎠ 3 ⎝ (6.27).sin ⎛ ωτ + -. 7..34) and using the identities sin ( x + π ⁄ 2 ) = cos x . we let ωt = ωτ + π ⁄ 2 .⎞ + … ⎝ ⎝ ⎠ 5 ⎝ ⎠ 3 π 2⎠ 2 2 4A 1 1 π 5π 3π = -----. and so on.cos 5ωt – …⎞ = -----⎠ 5 π π ⎝ 3 n = odd ∑ ( –1 ) (n – 1 ) ---------------2 1 -. 11.sin ⎛ ωτ + -.sin 5ωt + …⎞ ⎝ ⎠ π 3 5 (6. it becomes – 4A a n = ---------nπ Then. that is.sin 5 ⎛ ωτ + -.35) and this is the same as (6.1. the trigonometric Fourier series for the square waveform with even symmetry is 1 4A 4A -f ( t ) = -----. relation (6. if we compute the trigonometric Fourier series with reference to one ordinate.3 is the same as of Example 6.1. Taylor.⎛ cos ω t – 1 cos 3ωt + -. but shifted to the right by π ⁄ 2 radians.cos 5ωτ – … 5 3 π (6.sin 3ωt + -.sin ⎛ 3ωτ + ----. Therefore.33) and substitute ωt with ωt + π ⁄ 2 . we can use the above procedure to save time. and so on. i. With this substitution.cos n ωt n (6. and Maclaurin Series 4A a n = -----nπ and for n = 3. sin ( x + 3π ⁄ 2 ) = – cos x .⎛ sin ωt + -. 4A 1 1 f ( t ) = -----.e.⎞ + -.Chapter 6 Fourier.34) as 1 1 4A f ( τ ) = -----.⎞ + -.32) Alternate Solution: Since the waveform of Example 6.33) becomes 1 1 4A π π π f ( τ ) = -----. we rewrite (6.sin ⎛ 5ωτ + ----.cos 3ωτ + -.

we only need to evaluate the b n coefficients.cos ax 2 a a (6. we can choose the limits from – π to +π . As before. it contains sine terms only with both odd and even harmonics. therefore.Waveforms in Trigonometric Form of Fourier Series Example 6.t π –π < t < π Better yet.16.36) Numerical Analysis Using MATLAB® and Excel®.t π ⎪ f(t) = ⎨ ⎪ A t – 2A --⎩π 0<t<π π < t < 2π However. By inspection. we can integrate from 0 to π . and multiply the integral by 2 . this is what we will do. and thus we will only need one integration since A f ( t ) = --. Sawtooth waveform Solution: This waveform is an odd function but has no half−wave symmetry. If we choose the limits of integration from 0 to 2π we will need to perform two integrations since A ⎧ --. since the waveform is an odd function. we will assume that ω = 1 . T A − 2π −π π 0 −A 2π ωt Figure 6. the DC component is zero. From tables of integrals.sin a x – -. Accordingly. Third Edition Copyright © Orchard Publications 6−17 . ∫ x sin ax dx Then.4 Compute the trigonometric Fourier series of the sawtooth waveform of Figure 6. 1 x = ---.16.

⎛ ---. If n = odd . and will multiply the 6−18 Numerical Analysis Using MATLAB® and Excel®. cos nπ = – 1 .5 Solution: This waveform is an odd function and has half−wave symmetry. we only need to evaluate the b n coefficients.sin 2ωt + -. and Maclaurin Series 2 b n = -π ∫0 π A --.t sin nt dt = 2A -----π π2 ∫0 π t 2A 1t sin nt dt = -----. Then.( nπ ) = -----n2 π2 nπ that is. 2A 2A b n = ---------. the odd harmonics have positive coefficients. Accordingly.17.⎛ sin ωt – -. Third Edition Copyright © Orchard Publications .37) 2A = ---------.sin nωt n (6. the trigonometric Fourier series for the sawtooth waveform with odd symmetry is 2A 1 1 1 2A f ( t ) = -----.Chapter 6 Fourier. then. sin nπ = 0 and cos nπ = 1 .38) Example 6. sin nπ = 0 . Triangular waveform for Example 6.sin nt – -. the even harmonics have negative coefficients. A −2π −π −A T ωt 0 π/2 π 2π Figure 6. (6.( sin nt – nt cos nt ) n2 π2 π 0 2A = ---------. Then.17.cos nt⎞ ⎠ n π2 ⎝ n2 π 0 (6. the trigonometric Fourier series will contain sine terms only with odd harmonics.( sin nπ – nπ cos nπ ) n2 π2 We observe that: 1. 2. We will choose the limits of integration from 0 to π ⁄ 2 .5 Find the trigonometric Fourier series of the triangular waveform of Figure 6. Taylor.sin 4ωt + …⎞ = -----⎠ π 4 3 2 π ⎝ ∑ ( –1 ) n–1 1 -.( – nπ ) = – -----n2 π2 nπ that is.37) reduces to 2A 2A b n = ---------. Assume ω = 1 .sin 3ωt – -. Thus. If n = even .

and we observe that: π cos n -.39) π⁄2 0 ∫0 π⁄2 2A -----.42) Express f ( t ) as a trigonometric Fourier series.( sin nt – nt cos nt ) 2 2 n π π π π 8A = ---------. 7.6 A half−wave rectification waveform is defined as ⎧ sin ωt f (t) = ⎨ ⎩ 0 0 < ωt < π π < ωt < 2π (6.⎛ ---.41) 2 n Example 6.sin 5ωt – ----.= ⎨ 2 8A ⎪ – 1 for n = 3. Third Edition Copyright © Orchard Publications 6−19 . b = ---------n ⎪ 2 2 n π ⎪ π sin n -.cos n -. ∫ x sin ax dx Then.Waveforms in Trigonometric Form of Fourier Series integral by 4 .sin n ωt (6.sin nt – -.t sin nt dt = 8A -----2 π π π⁄2 0 ∫0 t 8A 1t sin nt dt = -----. By inspection. 11. From tables of integrals. Assume ω = 1 .= 0 2 For odd integers of n . the sine term yields 8A⎧ 1 for n = 1.⎛ sin n -. the DC component is zero.40) We are only interested in the odd integers of n .cos nt⎞ 2 ⎝ 2 ⎠ n n π 8A = ---------. b = – ---------n ⎪ 2 2 n π ⎩ Thus. 4 b n = -π 1x = ---2 sin a x – -. 5.⎛ sin ω t – -.sin 7ωt + …⎞ = -----2 ⎝ 2 ⎠ 49 25 9 π n = odd π ∑ (n – 1) ---------------2 1 ---.sin 3ωt + ----.cos ax a a π⁄2 (6. … then.– n -. Solution: The waveform for this example is shown in Figure 6. 9. Numerical Analysis Using MATLAB® and Excel®.⎞ 2 2 ⎝ 2⎠ 2 2 n π (6. the trigonometric Fourier series for the triangular waveform with odd symmetry is 8A 1 1 8A 1 ( –1 ) f ( t ) = -----.18. … then.

--------------------------.⎛ -----------------------.⎧ cos nπ cos nπ 2 ⎫ A ⎧ – cos nπ – cos nπ 2 ⎫ a n = – ----.⎨ – -.– ------------.⎨ -----------------.⎨ -------------.+ --------------------------π ⎩ 2 1+n 1–n ⎫ ⎬ 0⎭ A ⎧ cos ( π – nπ ) cos ( π + nπ ) 1 1 ⎫ = – ----. Taylor. we expect all terms to be present.⎨ ---------------------------.+ ----------. and from tables of integrals A a n = --π ∫0 π A sin t cos nt dt + --π ∫π 2π 0 cos nt dt ∫ ( sin mx ) ( cos nx ) dx Then.43).6 By inspection.44) 6−20 Numerical Analysis Using MATLAB® and Excel®.+ ------------.– ----------.Chapter 6 Fourier. A.⎛ --------------------------------------------------------------------------------------. by substitution into (6.43) Using the trigonometric identities and we obtain and cos ( x – y ) = cos x cos y + sin xsiny cos ( x + y ) = cos x cos y – sin x sin y cos ( π – nπ ) = cos π cos nπ + sin π sin nπ = – cos nπ cos ( π + nπ ) = cos π cos nπ – sin π sin nπ = – cos nπ Then. The a n coefficients are found from 1 a n = -π ∫0 2π f ( t ) cos nt dt For this example.⎬ 2 2 2π ⎩ 1 – n 1+n 2π ⎩ 1 – n 1+n 1–n ⎭ 1–n ⎭ A cos nπ + 1 A cos nπ + n cos nπ + cos nπ – n cos nπ 2 = ----.+ -------------. and Maclaurin Series –2 π –π 0 π 2π 3π Figure 6. 2 2 cos ( m – n )x cos ( m + n )x = – -----------------------------.( m ≠ n ) 2( m + n) 2(m – n) π A ⎧ 1 cos ( 1 – n )t cos ( 1 + n )t a n = --.18. Therefore.⎬ 2π ⎩ 1–n 1+n 1–n n+1 ⎭ (6.+ ----------------------------. the average is a non−zero value.– -----------------------------. and the waveform has neither odd nor even symmetry.⎞ n ≠ 1 2 2⎠ π ⎝ ( 1 – n2 ) ⎠ 2π ⎝ 1–n 1–n (6. Third Edition Copyright © Orchard Publications .⎬ = ----. f ( t ) for Example 6.+ -----------------.+ -------------⎞ = --.

for n = even . Now. we get a 0 = 2A ⁄ π Therefore.Waveforms in Trigonometric Form of Fourier Series Next.50) (6. we get 2AA ( cos 4π + 1 ) a 4 = --------------------------------.a 0 = A --2 π (6. First. and thus.= 0 2 π(1 – 3 ) (6. therefore. 1 b n = A -π ∫0 2π A f ( t ) sin nt dt = --π ∫0 π A sin t sin nt dt + --π ∫π 2π 0 sin nt dt Numerical Analysis Using MATLAB® and Excel®. from (6.45) We cannot use (6.46) From (6.= – -------2 35π π(1 – 6 ) 2AA ( cos 8π + 1 ) a 8 = --------------------------------.44) to obtain the value of a 1 .47) (6.44) with n = 2. 3. we will evaluate the integral A a 1 = --π ∫0 sin t cos t dt 12 = ----.= – -------2 15π π(1 – 4 ) 2AA ( cos 6π + 1 ) a 6 = --------------------------------. the DC value is 1 -. … . However.( sin ax ) 2a = 0 0 π From tables of integrals. we get A cos 2π + 1 -----a 2 = --. we need to evaluate the b n coefficients.48) We see that for odd integers of n . except a 1 .⎛ -----------------------.= – -------2 63π π(1 – 8 ) (6. By substitution of n = 0 .49) (6. we can evaluate all the a n coefficients.⎞ = – 2A π ⎝ ( 1 – 22 ) ⎠ 3π A ( cos 3π + 1 ) a 3 = --------------------------------. For this example.51) and so on. ∫ ( sin ax ) ( cos ax ) dx A 2 a 1 = ----. 5. Third Edition Copyright © Orchard Publications 6−21 . we will evaluate a 0 to obtain the DC value.44). 4.( sin t ) 2π π (6. a n = 0 .

– ------------.b n = --.– -----------π 2 4 π 0 A π sin 2π A . 8 0 .+ --.– ----------------------------.⋅ -. Full−wave rectified waveform with even symmetry 6−22 Numerical Analysis Using MATLAB® and Excel®.– -------------------------π 2⎩ 1+n 1–n ⎫ ⎬ 0⎭ A sin ( 1 – n )π sin ( 1 + n )π = ----.45) and (6.47) through (6.– 0 + 0 = 0 ( n ≠ 1 ) 2π 1–n 1+n that is.19.------------. 2 2 sin ( m – n )x sin ( m + n )x = ----------------------------. We will find b 1 by direct substitution into (6.+ ------------. 5 0 . 1 0 – 2π 0 2 –π 4 6 0 8 π 1 0 1 2 2π Figure 6. ∫ ( sin mx ) ( sin nx ) dx Therefore.= --.+ ------------. all the b n coefficients. 9 0 . A b 1 = --π ∫0 π A t sin 2t 2 . 2 0 .-. 6 A 0 . we find that the trigonometric Fourier series for the half−wave rectification waveform with no symmetry is A cos 2t cos 4t cos 6t cos 8t A A f ( t ) = --.Chapter 6 Fourier.= --π 2 4 2 (6.( m ≠ n ) 2(m + n) 2( m – n) π A 1 ⎧ sin ( 1 – n )t sin ( 1 + n )t .⎨ -------------------------.14) for n = 1 .52).--------------------------. are zero.53) Example 6.( sin t ) dt = --. 4 0 . Thus.7 A full−wave rectification waveform is defined as f ( t ) = A sin ωt (6. Taylor.54) Express f ( t ) as a trigonometric Fourier series. and Maclaurin Series and from tables of integrals.-. Solution: The waveform is shown in Figure 6. except b 1 .52) Combining (6. Third Edition Copyright © Orchard Publications .+ … . 3 0 .π 2 63 3 15 35 π (6.sin t – --.– ---------------------------.19 where the ordinate was arbitrarily chosen as shown.+ ------------. 1 0 . 7 0 . Assume ω = 1 .

The a n coefficients are found from 1 a n = -π ∫0 2π f ( t ) cos nt dt where for this example.-------------------------------------.Waveforms in Trigonometric Form of Fourier Series By inspection.56).⎬ π⎩ n–1 n+1 n–1 n+1 ⎭ A 1 – cos ( nπ + π ) cos ( nπ – π ) – 1 = --.⋅ -.56) simplifies to A – 2 + ( n – 1 ) cos nπ – ( n + 1 ) cos nπ A .n ≠ 1 2 π(n – 1) (6. from (6.( m ≠ n ) 2(m + n) 2(n – m) cos ( x – y ) = cos ( y – x ) = cos x cos y + sin xsiny π 2A 1 ⎧ cos ( n – 1 )t cos ( n + 1 )t . Third Edition Copyright © Orchard Publications 6−23 .57).– ----------.– ----------.1 + cos nπ – 1 + cos nπ = --.55) as 2 2 cos ( m – n )x cos ( m + n )x = -----------------------------. except a 1 . we will evaluate a 0 to Numerical Analysis Using MATLAB® and Excel®. (6.56) To simplify the last expression in (6. Therefore.------------------------------------------------------------------------------------2 π π n–1 n+1 n –1 – 2A ( cos nπ + 1 ) = ---------------------------------------.-----------------------. ∫ ( sin mx ) ( cos nx ) dx Since we express (6.+ ------------------------------------π n+1 n–1 (6. We choose the period of the input sinusoid so that the output will be expressed in terms of the fundamental frequency.– --------------------------π 2⎩ n–1 n+1 ⎫ ⎬ 0⎭ A ⎧ cos ( n – 1 )π cos ( n + 1 )π 1 1 ⎫ = --.⎨ --------------------------. we make use of the trigonometric identities and cos ( nπ + π ) = cos nπ cos π – sin nπsinπ = – cos nπ cos ( nπ – π ) = cos nπ cos π + sin nπsinπ = – cos nπ Then. we observe that the waveform has even symmetry.– -----------------------------. 1 a n = -π ∫ 2A A sin t cos nt dt = -----π –π π ∫0 π sin t cos nt dt (6. First.– ----------------------------. We also choose the limits of integration as – π and +π .57) Now. we can evaluate all the a n coefficients.-----------------------a n = --.a n = -----. we expect only cosine terms to be present.⎨ ---------------------------.55) and from tables of integrals. the average is a non−zero value.

the trigonometric form of the Fourier series for the full−wave rectification waveform with even symmetry is 6−24 Numerical Analysis Using MATLAB® and Excel®.60) through (6.61) (6.Chapter 6 Fourier.63) and so on. Third Edition Copyright © Orchard Publications .+ ----------------.+ ----------------.( sin t ) 2π π = 0 0 (6.+ ----------------.64) Therefore. the DC value is 1 -.58) and (6. a n = 0 . By substitution of n = 0 .63) we get ⎫ 4A ⎧ cos 2ωt cos 4ωt cos 6ωt cos 8ωt -----f ( t ) = 2A – -----.58) From (6.59) For n = even .= – -------2 15π π(4 – 1) – 2A ( cos 6π + 1 ) 4A a 6 = ---------------------------------------. from (6.⎨ ----------------.57) we observe that for all n = odd . other than n = 1 .a 0 = 2A -----2 π (6.( sin ax ) 2a From tables of integrals.62) (6.57) we get -----a 2 = – 2A ( cos 2π + 1 ) = – 4A ---------------------------------------2 3π π(2 – 1) 4A– 2A ( cos 4π + 1 ) a 4 = ---------------------------------------. Then. Taylor. To obtain the value of a 1 . ∫ ( sin ax ) ( cos ax ) dx and thus. we get -----a 0 = 4A π Therefore.+ … ⎬ π ⎩ 3 63 35 15 π ⎭ (6.= – -------2 35π π(6 – 1) – 2A ( cos 8π + 1 ) 4A a 8 = ---------------------------------------. we must evaluate the integral 1 a 1 = -π π ∫0 sin t cos t dt 1 2 = ----.60) (6. and Maclaurin Series obtain the DC value.= – -------2 63π π(8 – 1) (6. combining the terms of (6. 12 a 1 = ----.

This is because we chose the period to be from – π and +π .65) shows that there is no component of the fundamental frequency. either cosine or sine.66) If a given waveform does not have any kind of symmetry. 2. the period is defined as the shortest period of repetition. … . 4. and the sum is combined to a single term.a 0 + a 1 cos ωt + a 2 cos 2ωt + a 3 cos 3ωt + a 4 cos 4ωt + … 2 + b 1 sin ωt + b 2 sin 2ωt + b 3 sin 3ωt + b 4 sin 4ωt + … (6.20. ϕn cn = bn an + bn bn bn sin θ n = --------------------. we rewrite (6.= ----cn an + bn an bn ϕ n = atan ----θ n = atan ----bn an Figure 6.cos nωt 2 (n – 1) n = 2.= ----cn an + bn cn θn an an an cos θ n = --------------------.Alternate Forms of the Trigonometric Fourier Series 2A 4A f ( t ) = -----. we still need to compute the a n and b n coefficients separately. Third Edition Copyright © Orchard Publications 6−25 . and for n = 1. it is very unlikely that a Fourier series will consist of even harmonic terms only.– -----π π 1 -----------------.5 Alternate Forms of the Trigonometric Fourier Series We recall that the trigonometric Fourier series is expressed as 1 f ( t ) = -. 6.20 for the derivation of the alternate forms. Derivation of the alternate form of the trigonometric Fourier series cos θ n = sin ϕ n We assume ω = 1 .66) as Numerical Analysis Using MATLAB® and Excel®. … ∑ ∞ (6. 3. Generally. In any waveform where the period is chosen appropriately. We use the triangle shown in Figure 6. However. 6.65) This series of (6. it may be advantageous of using the alternate form of the trigonometric Fourier series where the cosine and sine terms of the same frequency are grouped together.

a 0 + c 1 ⎛ 2 ⎝ ⎠ sin ( t + ϕ 1 ) ⎧ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎩ c2 ⎛ ⎝ ⎧ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎩ sin ( 2t + ϕ 2 ) ⎧ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎩ sin ( nt + ϕ n ) an sin ϕ 2 cos 2t + cos ϕ 2 sin 2t ⎞ + … + cn ⎛ ⎠ ⎝ sin ϕ n cos nt + cos ϕ n sin nt ⎞ ⎠ and.69) Example 6.21. Since it is customary to use the cosine function in the time domain to phasor transformation. in general.Chapter 6 Fourier. we get 1 f ( t ) = -.sin nt⎞ ⎝ cn ⎠ cn 1 = -.a 0 + 2 ⎧ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎩ cos ( nt – θ n ) ∞ cos θ n cos nt + sin θ n sin nt ⎞ ⎠ n=1 ∑ 1 c n cos ( nωt – θ n ) = -.68) The series of (6.68) can be expressed as phasors. for ω ≠ 1 .67) and (6.sin t⎞ + c 2 ⎛ ---. 6−26 Numerical Analysis Using MATLAB® and Excel®.a 0 + 2 n=1 ∑ ∞ 1 c n sin ( nωt + ϕ n ) = -.a 0 + c 1 ⎛ ---. Third Edition Copyright © Orchard Publications . we get 1 f ( t ) = -. where ω ≠ 1 . and Maclaurin Series a1 a2 b1 b2 1 f ( t ) = -.a 0 + 2 n=1 ∑ cn sin ⎛ nωt + atan ----n⎞ ⎝ b⎠ ∞ (6.8 Find the first 5 terms of the alternate form of the trigonometric Fourier series for the waveform of Figure 6. sin ϕ 1 cos t + cos ϕ 1 sin t 1 ⎞ f ( t ) = -. in general.a 0 + ⎝ ⎠ 2 an n=1 ∑ cn ∠– atan ----na ∞ bn (6.63) below. we choose to use the transformation of (6.a 0 + 2 n=1 ∑ cn cos ⎛ nωt – atan ----n-⎞ ⎝ a ⎠ ∞ (6. 1 -.sin 2t⎞ + … ⎝ c1 ⎠ ⎝ c2 ⎠ 2 c1 c2 bn an + c n ⎛ ---.cos t + ---.⎞ ⇔ -.a 0 + c 1 ⎛ 2 ⎝ + cn ⎛ ⎝ ⎧ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎩ cos ( t – θ 1 ) ⎧ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎩ cos ( 2t – θ 2 ) bn cos θ 1 cos t + sin θ 1 sin t ⎞ + c2 ⎛ ⎠ ⎝ cos θ 2 cos 2t + sin θ 2 sin 2t ⎞ +… ⎠ and. Taylor.cos nt + ---.67) Similarly.cos 2t + ---.a 0 + 2 n=1 ∑ ∞ bn 1 c n cos ⎛ nωt – atan ---.

Alternate Forms of the Trigonometric Fourier Series f( t) 3 2 1 ω = 1 t π/2 π 3π/2 2π Solution: Figure 6. by inspection the DC value is not zero. Then.sin n -.+ ----. the DC value.21.sin n2π – ----. 1 a n = -π ∫0 π⁄2 1 ( 3 ) cos nt dt + -π ∫ 2π 3 ( 1 ) cos nt dt = ----. we expect both cosine and sine functions with odd and even terms present.70) We observe that for n = even .71) and (6.sin nt nπ π⁄2 π⁄2 0 1 + ----.sin n -.a 0 = ----2 2π 1 = ----2π ∫0 π⁄2 1 ( 3 ) dt + ----2π ∫ π ⁄ 2 ( 1 ) dt 2π 1 = ----. We will compute the a n and b n coefficients.73) The b n coefficients are Numerical Analysis Using MATLAB® and Excel®.= ----.8 The given waveform has no symmetry. Also. Third Edition Copyright © Orchard Publications 6−27 . For n = odd . 2 a 1 = -π 2a 3 = – ----3π (6.sin nt nπ 2π π⁄2 π 2 π 1 π 1 3 = ----.sin n -2 2 nπ nπ 2 nπ nπ (6. Waveform for Example 6. a n = 0 .63).( π + 2π ) = 3 ------⎝2 ⎠ 2π 2 2 (6.( 3t 2π π⁄2 0 +t 2π ) π⁄2 1 ⎛ 3π + 2π – π⎞ = ----. thus.72) The DC value is 1 1 -. and we will combine them to get an expression in the form of (6.

⎞ ∠– 45° ⎝ π⎠ ⎝ π⎠ 2 2 Similarly.∠– 90° ⇔ -.= an a n + b n ∠– θ n = a n – jb n Thus.a 0 + 2 where n=1 ∑ ∞ bn 1 c n cos ⎛ nωt – atan ---.83). we get: 2 2 a 1 – jb 1 = -.= -.73) and (6. 3. Third Edition Copyright © Orchard Publications .= --------.77) (6.76) (6.∠– 45° ⇔ --------.cos nt nπ π⁄2 π⁄2 0 –1 + ----.+ ----.= ----.– j -.∠– 90° ⇔ ----.= π π = 2 2 ⎛ -. 1 -.⎞ + ⎛ -.82) and 111a 4 – jb 4 = 0 – j ----. for n = 1.cos n -.81) (6.83) Combining the terms of (6.– j ----. 2.a 0 + ⎝ ⎠ 2 an bn 2 2 a n + b n ∠– atan ---.+ ----. and Maclaurin Series 1 b n = -π ∫0 π⁄2 1 ( 3 ) sin nt dt + -π ∫ 2π –3 ( 1 ) sin nt dt = ----.= ----.79) ∞ bn bn c n ∠– atan ---.75) (6.74) Then.cos ( 2ωt – 90° ) π π π 2 2 2 2 2 2 a 3 – jb 3 = – ----. b1 = 2 ⁄ π b2 = 1 ⁄ π b 3 = 2 ⁄ 3π b 4 = 1 ⁄ 2π (6.cos n2π + ----.80) through (6.cos ( 3ωt – 135° ) 3π 3π 3π 3π (6.a 2 – jb 2 = 0 – j -. Taylor.∠– 135° ⇔ --------. we find that the alternate form of the trigonometric Fourier series representing the waveform of this example is 6−28 Numerical Analysis Using MATLAB® and Excel®.cos ( 4ωt – 90° ) 2π 2π 2π (6.Chapter 6 Fourier.69).cos ( ωt – 45° ) π π 2 π (6. and 4 .∠– 45° = --------.cos nt nπ 2π π⁄2 1π 1π 3 –1 –3 2= ----.⎞ ⇔ -.( 3 – cos n2π ) = ----2 nπ nπ 2 nπ nπ nπ nπ (6.cos n -.78) From (6.= an 2 2 n=1 ∑ cn ∠– atan ----na (6.80) 1 1 1 . 2 2 2 2 8 ----.

a 0 2 (6.89) (6.88) The terms of (6.– ---.a 0 + a 1 ⎛ --------------------------.cos ( 4ωt – 90° ) + … 2 3 ] (6.90) (6.⎞ = -.( a n + jb n ) ⎠ 2 2⎝ j bn 1 1 C n = -.84) 6. we get a b 2 – j2ωt ⎛ a 1 b 1 ⎞ – jωt 1 a 1 b 1 jωt a b 2 j2ωt f ( t ) = … + ⎛ ---2 – ---.⎛ a n – ---. Moreover.⎞ + b 2 ⎛ ------------------------------. The advantage of the exponential form is that we only need to perform one integration rather than two.⎞ + ⎝ ⎠ ⎝ ⎠ 2 2 2 jωt – jωt j2ωt – j2ωt e –e e –e … + b 1 ⎛ -------------------------.⎞ + … ⎝ ⎠ ⎝ ⎠ j2 j2 jωt – jωt j2ωt – j2ωt (6.6 The Exponential Form of the Fourier Series The Fourier series are often expressed in exponential form.a 0 + ⎛ ---.⎞ e + ---.The Exponential Form of the Fourier Series 1 -.88) in parentheses are usually denoted as bn 1 1 C – n = -. The exponential form is derived from the trigonometric form by substitution of +e --------------------------cos ωt = e 2 jωt – jωt (6. one for the a n .+ ----⎞ e + ⎛ ---2 + ----⎞ e ⎝ 2 j2 ⎠ ⎝ 2 j2 ⎠ ⎝ 2 j2⎠ ⎝ 2 j2⎠ 2 (6.⎞ + a 2 ⎛ -------------------------------.87) and grouping terms with same exponents.cos ( 3ωt – 135° ) + -.( a n – j b n ) ⎠ 2 2⎝ j 1 C 0 = -.91) Numerical Analysis Using MATLAB® and Excel®.⎛ a n + ---.e + -.86) into f ( t ) . in most cases the integration is simpler. Thus.⎞ = -. 1 e +e e +e f ( t ) = -.85) and e –e sin ωt = -------------------------j2 jωt – jωt (6.f ( t ) = 3 + -2 π [2 2 cos ( ωt – 45° ) + cos ( 2ωt – 90° ) 1 2 2 + --------. and another for the b n coefficients in the trigonometric form of the series. Third Edition Copyright © Orchard Publications 6−29 .

as we did in the derivation of the a n and b n 2π 2π coefficients of the trigonometric form. ∫0 2π f ( t )e – jnt dt = … + + + ∫0 C –2 e C0 e C2 e – j2t – jnt e dt + 2π ∫0 ∫0 2π ∫0 C–1 e e dt – jt – jnt dt (6.92) by e – jnωt and integrating over one period.95) or 1 C n = -T T f ( t )e – jnωt d( ωt ) (6. with ω = 1 . and Maclaurin Series Then.( a n – jb n + a n + jb n ) 2 6−30 Numerical Analysis Using MATLAB® and Excel®. 1 C n = ----2π ∫0 ∫0 2π f ( t )e – jnωt d( ωt ) (6. are complex and occur in complex conjugate pairs. in general.90).96) We can derive the trigonometric Fourier series from the exponential series by addition and subtraction of the exponential form coefficients C n and C –n . from (6.Chapter 6 Fourier.93) We can derive a general expression for the complex coefficients C n . Third Edition Copyright © Orchard Publications . except C 0 . by multiplying both sides of (6.88) is written as f ( t ) = … + C –2 e – j2ωt + C –1 e – jωt + C0 + C1 e jωt + C2 e j2ωt +… (6. (6. Thus.94) – jnt dt + 2π ∫0 C1 e e jt – jnt 2π j2t – jnt e dt + … + ∫0 ∫0 Cn e jnt – jnt e dt We observe that all the integrals on the right side of (6. that is. C – n = C n∗ (6. 1 C n + C – n = -.97) are zero except the last. for ω ≠ 1 . Then. ∫0 or 2π f ( t )e – jnt dt = ∫0 2π Cn e jnt – jnt e dt = 2π C n dt = 2πC n 1C n = ----2π ∫0 2π f ( t )e – jnt dt and.92) We must remember that the C i coefficients.89) and (6. Therefore. Taylor.

⎛ a n + ---. both C –n and C n are also zero for n = even . C n = 0 for n = even We recall from the trigonometric Fourier series that if there is half−wave symmetry. 2.101) Numerical Analysis Using MATLAB® and Excel®.100) and (6. 5.⎞ = -. all even harmonics are zero. If there is no symmetry. C –n = C n∗ always This can be seen in (6.100) and (6. If there is half−wave symmetry.The Exponential Form of the Fourier Series or an = Cn + C–n (6. and thus.⎛ a n – ---. f ( t ) is complex. For odd functions. Therefore.89) and (6.⎞ = -. the b n coefficients in (6.100) and (6.100) and bn 1 1 C n = -.( a n – jb n – a n – j b n ) 2 or bn = j ( Cn – C–n ) (6. Therefore. both C –n and C n are imaginary. the a n coefficients in (6. For even functions. in (6.98) Similarly.90) that bn 1 1 C – n = -. 1 C n – C – n = -.( a n – j b n ) ⎝ ⎠ 2 2 j (6. Third Edition Copyright © Orchard Publications 6−31 . all coefficients C i are real We recall from (6.97) (6.101) Since even functions have no sine terms.( a n + jb n ) ⎝ ⎠ 2 2 j (6.101) the coefficients a n and b n are both zero for n = even . 3.101) are zero.100) and (6.101) are zero. 4. all coefficients C i are imaginary Since odd functions have no cosine terms.99) Symmetry in Exponential Series 1. both C –n and C n are real. Therefore.

we get 1C n = ----2π ∫0 2π f ( t )e – jnt 1dt = ----2π 2π ∫0 π Ae – jnt 1dt + ----2π ∫π 2π –A e – jnt dt and for n = 0 .22.( 1 – e +e –e ) jn 2jπn 2π – jn 2 A A – jn2π – jnπ . 1 C 0 = ----2π ∫0 π Ae dt + –0 as expected. then.------.9 Compute the exponential Fourier series for the square waveform of Figure 6.------. it is an odd function. C n = 0 for n = even . 1C n = ----2π ∫π A –0 ( – A )e dt = ----. Assume that ω = 1 . has half− wave symmetry. n = even Cn 2 A A .( e – 1 ) + ---. T A π 0 2π ωt −A Figure 6. Therefore.( e 2jπn 2jπn (6.1.( e 2jπn 2jπn (6. Using (6.102) For n = even .22 below.( e –e ) = ----------.Chapter 6 Fourier.( 1 – 1 ) 2 = 0 = ----------. and C 0 = 0 . e – jnπ = 1 . Waveform for Example 6. Third Edition Copyright © Orchard Publications .( 1 + e – 2e ) = ----------. the C n coefficients will be imaginary. Taylor.e – jn 2π π A – jn2π – jnπ A 1 A –jnπ – jnπ – jn2π – jnπ = ----.95) with ω = 1 .– jnπ – 1 ) = ----------.e 2π – jn π 0 – A – jnt + ------.–jnπ – 1 ) = ----------.9 Solution: This is the same waveform as in Example 6. and as we know.103) 6−32 Numerical Analysis Using MATLAB® and Excel®. For n ≠ 0 . and its DC component is zero. and Maclaurin Series Example 6.– jnt dt = ----.A.( π – 2π + π ) = 0 2π ∫0 π Ae – jnt dt + ∫π 2π –A e – jnt 1.

92). and so on. To prove that (6. 2 A . Figure 6. bn 4/π 0 1 3 5 7 9 nωt Figure 6.e n jnωt (6.( – 1 – 1 ) = ----------. Numerical Analysis Using MATLAB® and Excel®.105) The minus (−) sign of the first two terms within the parentheses results from the fact that C –n = C n∗ . We observe that f ( t ) is purely imaginary. e – jnπ = – 1 . Therefore.( e – 1 ) = ----------.e –e +e + -. we group the two terms inside the parentheses of (6. since the waveform is an odd function. f ( t ) = … + C–2 e – j2ωt + C–1 e – jωt + C0 + C1 e jωt + C2 e j2ωt +… we obtain the exponential Fourier series for the square waveform with odd symmetry as 1 – j3ωt – jωt 2A jωt 1 f ( t ) = -----. that is.e 3 jπ ⎝ 3 j3ωt ⎞ ⎠ 2A = -----jπ n = odd ∑ 1 -. Line spectrum for square waveform of Example 6. it is useful to plot the amplitudes of the harmonics on a frequency scale that shows the first (fundamental frequency) harmonic.– jnπ AA2 2 2A = ----------.1. this will produce the fundamental frequency sin ωt .1 * An instrument that displays the spectral lines of a waveform.105) and (6.Line Spectra as expected.7 Line Spectra When the Fourier series are known.23 shows the line spectrum of the square waveform of Example 6.104) Using (6.23.( – 2 ) = ------2jπn 2jπn 2jπn jπn n = odd Cn (6. Then. as expected. we group the two terms for which n = 3 . Third Edition Copyright © Orchard Publications 6−33 . For n = odd . and this will produce the third harmonic sin 3ωt . For instance. since C 3 = 2A ⁄ j3π .105) for which n = 1 . it follows that C –3 = C 3∗ = – 2A ⁄ j3π .22) are the same. 6. and the higher harmonics times the amplitude of the fundamental.⎛ … – -. Such a plot is known as line spectrum and shows the spectral lines that would be displayed by a spectrum analyzer*.

A T T /κ −2π −π −π/κ 0 π/κ π 2π ωt Solution: This recurrent rectangular pulse is used extensively in digital communications systems. is k times the pulse duration.10 Compute the exponential Fourier series for the waveform of Figure 6. For this example. Example 6. the pulse duration is T ⁄ k .6. Line spectrum for half−wave rectifier of Example 6. Thus. Taylor.25.106) we get 6−34 Numerical Analysis Using MATLAB® and Excel®.106) The value of the average ( DC component) is found by letting n = 0 . and Maclaurin Series Figure 6. As shown in Figure 6. Third Edition Copyright © Orchard Publications . and plot its line spectra. the recurrence interval (period) T .Chapter 6 Fourier. A/ 2 A/π DC 2 4 6 8 nωt 0 1 Figure 6. from (6. To determine how faithfully such pulses will be transmitted.25. Waveform for Example 6.24. Assume ω = 1 . Then. In other words.11 ∫– π π Ae – jnt Adt = ----2π ∫– π ⁄ k e π⁄k – jnt dt (6.24 shows the line spectrum for the half−wave rectification waveform of Example 6.25. it is necessary to know the frequency components. k is the ratio of the pulse repetition time to the duration of each pulse.6 The line spectra of other waveforms can be easily constructed from the Fourier series. the components of the exponential Fourier series are found from 1C n = ----2π Figure 6.

⋅ ------------------------k nπ ⁄ k (6.28.= --.109) has the sin x ⁄ x form.– jnt C n = -------------.26 through 6.26.4 1.4 1./(5.2 1 0.8 0./(10. k = 5 and k = 10 respectively by using the MATLAB scripts below.*x).t 2π π⁄k –π ⁄ k A.Line Spectra AC 0 = ----.106) yields A . and the line spectrum is shown in Figures 6.2 0 -0.= ----.*x)'.*x)'.108) f(t) = ∑ n = –∞ ∞ A sin ( nπ ⁄ k ) --. Third Edition Copyright © Orchard Publications 6−35 .109) for k = 2 Numerical Analysis Using MATLAB® and Excel®.2]) 1. fplot('sin(2.π π A = ----.4 0. Line spectrum of (6.[−4 4 −0.⋅ ------------------------nπ k nπ ⁄ k (6.⋅ sin ⎛ ----.e – jn2 π π⁄k –π ⁄ k nπ AA.2 -0.[−4 4 −0. integration of (6.⎞ = --. sin ( nπ ⁄ k ) A sin ( nπ ⁄ k ) = A ------------------------.+ -.[−4 4 −0.4 1.2]) fplot('sin(10.*x).*x).2π ⎝ k k ⎠ k (6.⎞ ⎝ k⎠ nπ nπ j2 jnπ ⁄ k – jnπ ⁄ k and thus.109) The relation of (6. for k = 2 .107) For the values for n ≠ 0 .⎛ -.⋅ -----------------------------------.e –e = ----./(2.2]) fplot('sin(5.*x)'.4 -4 -3 -2 -1 0 1 2 3 4 K=2 Figure 6.6 0.

the lines get closer together while the lines are further apart as k gets smaller.2 -0. Line spectrum of (6.4 0.6 0. Examples are meteorological or economic quantities 6−36 Numerical Analysis Using MATLAB® and Excel®.27.2 1 0.6 0.4 -4 -3 -2 -1 0 1 2 3 4 K=10 Figure 6. it is necessary to construct the Fourier expansion of a function based on observed values instead of an analytic expression.Chapter 6 Fourier. Line spectrum of (6. as k gets larger.8 Numerical Evaluation of Fourier Coefficients Quite often.2 -0. Taylor. Third Edition Copyright © Orchard Publications .2 1 0.2 0 -0.109) for k = 5 1. 6.4 0. and Maclaurin Series 1.8 0.28.4 -4 -3 -2 -1 0 1 2 3 4 K=5 Figure 6.8 0.112) for k = 10 The spectral lines are separated by the distance 1 ⁄ k and thus.2 0 -0.

and Figure 6. Next. Obviously. such as Microsoft Excel. Even though we may already know the Fourier series from analytical methods. a week. If the time axis is in degrees.. and so on. Then. and we divide by π to obtain the coefficients a 1 and b 1 . we enter the values of sin x and y sin x in Columns F and G respectively.29.1459. Third Edition Copyright © Orchard Publications 6−37 . the more pulses we use. we form the sums of y cos x and y sin x . we need to evaluate the integral(s) using numerical integration. y sin 2x . were we have divided it into small pulses of width Δx . We enter these in Column B and we denote them as x . we can choose Δx to be 2.31 is a partial table showing the computation of the coefficients of a clipped sine waveform. f(x) x Figure 6. using a spreadsheet. will work for both the waveforms that have an analytical solution and those that do not. The procedure presented here. Similarly. Figure 6. a month or even a year. The complete tables extend to the seventh harmonic to the right and to 360° down. Numerical Analysis Using MATLAB® and Excel®. and enter these values in Column A of the spreadsheet. we can divide the period 0° to 360° in 2. we multiply degrees by π (3.Numerical Evaluation of Fourier Coefficients whose period may be a day. In Columns D and E we enter the values of cos x and the product y cos x respectively. y sin 3x .30 is a partial table showing the computation of the coefficients of the square waveform. we can use this procedure to check our results. In Column C we enter the corresponding values of y = f ( x ) as measured from the waveform.5° and it is convenient to start at the zero point of the waveform.29.5° intervals. To compute the coefficients of the higher order harmonics. we multiply these by Δx . y cos 3x . In these situations. the better the approximation. we form the products y cos 2x . Waveform whose analytical expression is unknown Since the arguments of the sine and the cosine are in radians.. and we enter these in subsequent columns of the spreadsheet.) and divide by 180 to perform the conversion. Consider the waveform of f ( x ) shown in Figure 6.

5 10.423 0.000 0.000 0.131 0.131 0.000 0.423 0.0 22.866 0.349 0.737 0.966 0.500 ysin3x 0.643 0.537 0.707 0.383 0.500 -0.500 0.000 0.000 0.707 0.5 35.000 1.707 0.906 0.991 0.044 0.000 0.0 47.611 0.174 0.524 0.000 0.966 0.044 0.000 -0.500 0.000 0.0 27.574 0.174 0.423 0.866 0.000 0.259 0.000 0.5 0.000 1.866 0.866 sin3x 0.996 0.500 0.609 0.383 0.643 0.906 0.0 -0.044 0.991 0.262 0. and Maclaurin Series Figure 6.707 0.044 0.273 0.707 0.793 0.643 0.766 0.985 0.866 0.131 -0.000 1.254 0.0 37.742 0.5*a0 0.424 0.342 0.216 0.5 Square waveform f(t)=4(sinwt/p+sin3wt/3p+sin5wt/5p+ ….000 0.423 0.906 0.000 0.131 0.000 1.000 0.259 0.819 0.676 0.000 0.000 0.609 0.000 1.000 -0.707 -0.999 0.707 0.996 0.707 0.906 0.574 0.500 0.436 0.000 -0.000 1.087 0.940 0.819 0.087 -0.383 -0.000 0.000 1.044 0.609 0.766 0.991 0.609 -0.985 0.383 0.044 0.342 0.707 0.000 1.643 0.462 0.180 0.044 0.174 0.991 1.087 0.766 1.131 -0.991 0. Taylor.301 0.044 0.698 0.966 0.707 0.087 0.866 0.793 0.044 0.301 0.887 0.537 0.044 0.866 0.785 0.793 0.940 0.500 0.5 25.000 1.044 0.985 0.866 0.500 x(deg) x(rad) Numerical Analysis Using MATLAB® and Excel®.643 0.793 0.259 0.924 0.766 0.259 0.737 0.574 0.175 0.819 0.000 1.0 2.044 0.654 0.087 -0.087 0.954 0.966 0.000 0.707 -0.574 0.0 42.000 0.0 0.966 0.873 .843 0.991 0.940 0.5 15.044 0.866 0.000 1.480 0.500 -0.393 0.999 0. Third Edition Copyright © Orchard Publications 0.342 0.924 0.793 0.500 0.766 0.000 0.0 17.985 cos3x ycos3x 1.766 0.0 6.174 0.000 1.500 0.000 1.866 0.174 0.000 1.000 0.259 -0.707 0. Numerical computation of the coefficients of the square waveform (partial listing) y=f(x) 0.843 0.996 0.342 0.793 -0.819 0.500 0.924 0.383 0.044 1.766 0.000 1.000 0.000 0.259 0.940 0.707 0.087 0.609 -0.259 0.0 2.985 cosx ycosx sinx ysinx cos2x ycox2x sin2x ysin2x 0.216 0.000 0.000 1.000 0.643 0.924 0.0 4.966 0.044 0.5 45.044 0.044 0.000 0.966 0.996 0.676 0.5 5.737 0.044 0.0 0.500 0.924 0.0 32.707 0.966 0.383 -0.000 1.866 0.643 0.500 0.996 0.383 0.829 0.044 0.906 0.996 1.866 0.940 0.0 7.793 -0.676 0.044 0.0 DC= a1= a2= a3= a4= a5= a6= a7= b1= b2= b3= b4= b5= b6= b7= 1.259 0.609 0.342 0.991 0.044 0.087 0.996 0.976 0.0 -1.259 0.462 0.567 0.991 1.924 0.5 30.259 0.574 0.887 0.000 -0.609 0.0 12.643 0.676 0.342 0.174 0.383 0.131 0.996 1.707 0.259 -0.954 0.6−38 Analytical: 1.819 0.609 0.000 1.737 0.819 0.000 0.174 0.30.793 0.793 0.985 0.5 50.924 0.174 0.259 0.) Numerical: Average= 1.5 40.044 0.5 20.218 0.000 1.976 0.5 -1.940 0.866 0.966 0.131 0.000 1.000 0.423 0.423 0.985 0.793 0.000 0.087 0.966 0.609 0.000 Chapter 6 Fourier.000 0.906 0.305 0.0 8.131 0.609 0.985 0.966 0.131 0.5 0.707 0.966 0.866 0.766 0.574 0.044 0.924 0.

866 0.259 0.216 0.171 0.183 0.766 0.000 b3= 0.643 0.737 0.271 0.211 0.174 0.5 45.000 -0.453 0.500 0.023 0.006 0.498 0.002 0.000 0.924 0.087 0.996 1.131 0.044 0.174 0.5 20.087 0.196 0.567 0.940 0.000 -0.397 -0.287 0.000 0.383 -0.006 0.5*a0 ysin2x 0.000 0.301 0.866 0.004 0.996 0.500 0.179 0.5 50.643 0.966 0.342 0.410 0.574 0.338 0.321 0.0 17.087 0.500 ysin3x 0.422 0.996 0.378 0.906 0.043 0.609 0.017 0.609 -0.423 0.383 0.819 0.121 0.262 0.873 6−39 .022 0.000 0.500 0.500 0.087 0.423 0.383 0.000 0.436 0.324 0.000 0.0 -0.0 27.191 -0.000 0.737 0.966 0.31.009 0.0 0.065 -0.146 0.183 0. Numerical: f(t)=unknown 1.239 0.044 0.342 0.397 0.000 0.000 0.044 0.793 0.996 0.458 0.354 0.793 0.087 -0.0 6.213 0.000 b5= 0.924 0.966 0.500 0.321 0.676 0.044 0.524 0.574 0.131 0.0 42.000 0.966 0.017 0.5 30.259 0.010 0.470 0.766 0.0 22.259 0.000 0.304 0.259 0.183 0.008 0.129 -0.126 0.480 0.5 -1.5 5.906 0.018 0.354 0.109 0.020 0.496 0.174 0.015 0.924 0.991 0.296 0.008 0.0 4.000 0.000 -0.000 b7= -0.5 2.0 7.044 0.0 12.999 0.574 0.171 0.369 0.5 1.059 0.087 0.000 0.022 0.991 1.843 0.966 0.462 0.369 0.272 0.500 0.011 0.022 0.866 0.087 0.218 0.483 0.819 0.044 -0.707 0.985 0.492 0. Third Edition Copyright © Orchard Publications y=f(x) cosx 1.793 0.985 0.084 0.216 0.698 0.500 0.866 0.433 0.086 0.171 0.131 0.433 0.5 25.305 0.5 35.087 0.224 0.000 0.000 0.000 0.0 DC= a1= a2= a3= a4= a5= a6= a7= b1= 0.609 0.500 0.707 0.013 0.766 0.262 0.163 0.000 0.349 0.0 0.707 0.250 0.707 0.991 0.067 0.129 0.043 0.887 0.304 0.131 -0.304 -0.000 0.433 0.866 0.707 0.383 0.000 0.087 0.000 0.676 0.537 0.991 0.090 0.393 0.643 0.0 0.492 cos3x ycos3x 1.906 0.742 0.150 0.000 0.250 Figure 6.022 0.0 37.500 0.500 0.866 0.966 0.500 0.500 0.0 47.015 0.707 0.220 0.265 0.423 0.483 0.985 ycosx sinx ysinx cos2x ycox2x sin2x 0.246 0.354 -0.271 0.462 0.866 0.5 10.175 0.028 b6= 0.954 0.000 -0.5 15. 5π/6 etc.5 40. Numerical computation of the coefficients of a clipped sine waveform (partial listing) x(deg) x(rad) Numerical Evaluation of Fourier Coefficients 0.397 0.321 0.383 1.338 0.707 -0.500 0.342 0.462 0.022 0.793 0.383 0.410 0.211 0.940 0.707 0.000 0.138 b4= 0.259 0.408 0.146 0.940 0.087 0.500 0.785 0.172 0.129 0.259 -0.000 Numerical Analysis Using MATLAB® and Excel®.250 0.609 0.060 0.924 0.022 0.609 0.131 0.131 0.611 0.287 0.174 0.0 32.654 0.030 0.354 0.433 sin3x 0.498 0.500 -0.022 0.819 0.500 0.269 0.034 0.423 0.004 0.129 0.354 0.0 2.985 0.250 -0.172 0.091 0.383 0.609 b2= 0.250 0.Analytical: Sine wave clipped at π/6.5 0.050 0.000 0.0 8.174 0.002 0.022 0.766 0.643 0.301 0.000 0.500 0.0 -1.117 0.047 0.383 0.976 0.829 0.342 0.354 0.132 0.022 0.259 0.793 -0.

+ ( kv ) .+ … 3! 5! 7! x x x cos x = 1 – ---.+ ---. Example 6.110) Some familiar power series expansions for real values of x are x x x x e = 1 + x + ---.116) Substitution of (6.11 If the applied voltage v is small (no greater than 5 volts).+ -----------. 6−40 Numerical Analysis Using MATLAB® and Excel®.+ -----------.– ---.Chapter 6 Fourier.+ … 2! 3! 4! 2 3 4 (6.-----------. Taylor.114) suggests the power series expansion of (6.115) Express the current i in (6. and the input voltage is a sinusoid.113) x x x sin x = x – ---.+ ---.111).( kv ) .+ … – 1⎞ = a ⎛ kv + ( kv ) . we rewrite (6.114) where a and k are arbitrary constants.114) as ( kv ) .115) into (6. Accordingly.– ---.+ ---. and Maclaurin Series 6.( kv ) -----------. that is.116) yields.+ …⎞ ⎝ ⎠ ⎝ ⎠ 2! 3! 4! 2! 3! 4! 2 3 4 2 3 4 kv (6.111) (6.+ … 2! 4! 6! 2 4 6 3 5 7 The following example illustrates the fact that a power series expansion can lead us to a Fourier Series. Third Edition Copyright © Orchard Publications .114) as a power series. Solution: The term e inside the parentheses of (6.112) (6. the current i in a semiconductor diode can be approximated by the relation i = a(e kv – 1) (6.+ ---.+ ( kv ) . v = V max cos ω t (6.9 Power Series Expansion of Functions A power series has the form ∞ ∑ ak x k=0 k = a0 + a1 x + a2 x + … 2 (6.-----------i = a ⎛ 1 + kv + -----------.

+ ------------------------------.121) Relation (6.120) If x 0 = 0 . that is.+ …⎞ ⎝ ⎠ 2! 3! 4! 2 3 4 (6. We observe that it consists of a constant term.cos 3x 4 4 1 4 3 1 cos x = -.cos 2x .110) with an = f (n) ( 0 ) ⁄ n! . where i ( v ) represents some experimental data for the current−voltage ( i – v ) characteristics of a semiconductor diode operating at the 0 ≤ v ≤ 5 volts region.+ -. Numerical Analysis Using MATLAB® and Excel®.8 8 2 (6.x + … + ---------------.121) is known as Maclaurin series. (6. To appreciate the usefulness and application of the Taylor series.+ -. 6.118) into (6.x 2! n! (n) (6.117) and after simplification.cos x + -.10 Taylor and Maclaurin Series A function f ( x ) which possesses all derivatives up to order n at a point x = x 0 can be expanded in a Taylor series as f'' ( x 0 ) f ( x0 ) 2 n f ( x ) = f ( x 0 ) + f' ( x 0 ) ( x – x 0 ) + ------------.cos 2x + -. and has the form of power series of (6.119) is the trigonometric series form of the Fourier series.2 2 1 3 3 cos x = -. and terms of all harmonic frequencies. we obtain a series of the following form: i = A 0 + A 1 cos ωt + A 2 cos 2ωt + A 3 cos 3ωt + A 4 cos 4ωt + … (6.120) reduces to f (0) n f'' ( 0 ) 2 f ( x ) = f ( 0 ) + f' ( 0 )x + ----------.119) We recall that the series of (6.118) Then. Third Edition Copyright © Orchard Publications 6−41 .cos 4x .32. higher frequencies which are multiples of the fundamental frequency.( x – x 0 ) 2! n! (n) (6. we will consider the plot of Figure 6.117) This expression can be simplified with the use of the following trigonometric identities: 2 1 1 cos x = -.( x – x 0 ) + … + ------------------. substitution of (6.+ ------------------------------.Taylor and Maclaurin Series ( kV p cos ω t ) ( kV p cos ω t ) ( kV p cos ω t ) i = a ⎛kV max cos ω t + ------------------------------. a term of the fundamental frequency.

We assume that the first n derivatives of the function i ( v ) exist at this point. we are seeking a constant that it will be the best approximation to the given curve in the vicinity of point P .Chapter 6 Fourier.33. suppose that we want to approximate the function i ( v ) by a power series. Therefore. the horizontal line i 0 passes through point P . i 0 ) shown in Figure 6. Third Edition Copyright © Orchard Publications .32. We begin by referring to the power series of (6.33. Taylor. in the neighborhood of some arbitrary point P ( v 0. and we denote this first approximation as a 0 shown in Figure 6. and Maclaurin Series i i(v) 0 v Figure 6.34. i 0 ) i0 0 v0 v Figure 6. i i(v) P ( v 0. Current-voltage (i-v) characteristics for a typical semiconductor diode Now. Approximation of the function i ( v ) by a power series 6−42 Numerical Analysis Using MATLAB® and Excel®. Obviously.110). where we observe that the first term on the right side is a constant.

we express the desired power series as f ( v ) = a0 + a1 ( v – v0 ) + a2 ( v – v 0 ) + a3 ( v – v0 ) + a4 ( v – v0 ) + … 2 3 4 (6.35. But since we want the power series to be a good approximation to the given function for some distance on either side of point P . i 0 ) i0 a0 0 v0 Figure 6. This will be accomplished if the linear term has the same slope as the given function as shown in Figure 6. i i( v) P ( v 0. First approximation of i ( v ) v The next term in the power series is the linear term a 1 x . Second approximation of i ( v ) It is evident that the slope of i ( v ) at v 0 is i' ( v 0 ) = a 1 and therefore.Taylor and Maclaurin Series i i(v) P ( v 0. Accordingly. we are interested in the difference v – v 0 . we want the linear term a 0 + a 1 ( v – v 0 ) to be the best approximation to the function i ( v ) in the vicinity of point P .34. i 0 ) i0 a0 + a0 ( v – v0 ) a0 a0 ( v – v0 ) v 0 v0 Figure 6. Third Edition Copyright © Orchard Publications 6−43 .122) Now. Numerical Analysis Using MATLAB® and Excel®. we seek a linear term of the form a 0 + a 1 x .35. Thus. the linear term a 0 + a 1 ( v – v 0 ) can be expressed as i ( v 0 ) + i' ( v 0 ) ( v – v 0 ) .

that is. 2a 2 = i'' ( v 0 ) or a 2 = i'' ( v 0 ) ⁄ 2 . fourth. When this is done. by a Taylor series as illustrated by the following example. 6−44 Numerical Analysis Using MATLAB® and Excel®.36.122).( v – v 0 ) + … 2! 3! (6.Chapter 6 Fourier.( v – v 0 ) + -------------. Taylor.125) and since we are asked to compute the first three terms. The remaining coefficients a 3.( x – a ) + -----------. and Maclaurin Series The third term in (6. Third approximation of i ( v ) Then.36. i'' ( v 0 ) 2 i''' ( v 0 ) 3 i ( v ) = i ( v 0 ) + i' ( v 0 ) ( v – v 0 ) + ------------. we obtain the following Taylor series. i 0 ) a0 + a1 ( v – v0 ) a0 a1 ( v – v0 ) a2 ( v – v0 ) i0 2 0 v0 v Figure 6. a 4.12 Compute the first three terms of the Taylor series expansion for the function y = f ( x ) = tan x (6. i i( v) a0 + a1 ( v – v0 ) + a2 ( v – v0 ) 2 2 P ( v 0. a 2 ( v – v 0 ) is a quadratic and therefore.122) are found by matching the third.123) We can also describe any function that has an analytical expression. and so on of (6.( x – a ) + … 3! 2! (6. Third Edition Copyright © Orchard Publications . we choose a 2 such that it matches the second derivative of the function i ( v ) in the vicinity of point P as shown in Figure 6. and higher order derivatives of the given function with these coefficients.124) at a = π ⁄ 4 . fifth. Solution: The Taylor series expansion about point a is given by f'' ( a ) 2 f''' ( a ) 3 f n ( x ) = f ( a ) + f' ( a ) ( x – a ) + ----------. we must find the first and second derivatives of f ( x ) = tan x . a 5 . Example 6.

129) and by substitution into (6.126).126) Next. 2 dz ----. using ----.sec x = sec x ⋅ tan x . n produces the first n terms in the series.130) We can also obtain a Taylor series expansion with the MATLAB taylor(f. pretty(z) 2 3 4 1 + 2x .128) Now. Then.sec x = 2 sec x ( sec x ⋅ tan x ) = 2 sec x ⋅ tan x dx dx (6. and a defines the Taylor approximation about point a .a) function where f is a symbolic expression. Third Edition Copyright © Orchard Publications 6−45 . using the trigonometric identity sec x = tan x + 1 2 2 (6.1/4 pi) + 10/3(x .(x . so f ′ ( x ) = sec x . we get dx d 2 dz ----.pi/4).⎞ = tan ⎛ -.= 2 sec x ----. A detailed description can be displayed with the help taylor command.1/4 pi) + --. at point a = π ⁄ 4 we have: π π f ( a ) = f ⎛ -. x=sym('x').(x .Taylor and Maclaurin Series 2 2 From math tables. For example.127) into (6.⎞ = 2 ( 1 + 1 )1 = 4 ⎝4⎠ π 2 (6.n. so we let z = sec x .⎞ = 1 + 1 = 2 ⎝4⎠ π f'' ( a ) = f'' ⎛ -.1/4 pi) 64 5 244 6 2176 7 + -. y=tan(x). z=taylor(y. we get.1/4 pi) + ---.= f'' ( x ) = 2 ( tan x + 1 ) tan x dx (6.125). the following MATLAB script computes the first 8 terms of the Taylor series expansion of y = f ( x ) = tan x about a = π ⁄ 4 . 2 --fn ( x ) = 1 + 2 ⎛ x – π ⎞ + 2 ⎛ x – π ⎞ + … ⎝ ⎠ ⎝ ⎠ 4 4 (6. d tan x = sec x .127) and by substitution of (6.1/4 pi) 15 45 315 Numerical Analysis Using MATLAB® and Excel®.⎞ = 1 ⎝4⎠ ⎝4⎠ f' ( a ) = f' ⎛ -.1/4 pi) + 8/3(x . To find f'' ( x ) we need to find the first dx d 2 2 derivative of sec x .(x .1/2 pi + 2(x .8.

6−46 Numerical Analysis Using MATLAB® and Excel®. and V T is the thermal voltage which depends on the temperature. t=sym('t').134) where I D is the DC (average) component of the current.+ … 2! 3! 2 3 (6.133) MATLAB displays the same result.x n! 2! t t y = f (t) = e t (6.13 Express the function in a Maclaurin’s series.x + … + ---------------.132). Solution: A Maclaurin’s series has the form of (6. Taylor. that is. Since all derivatives are e . we can use the following Maclaurin’s series. fn=taylor(exp(t)). t t f n ( t ) = 1 + t + ---.131) (n) (6. Solution: Since the voltage is small and varies about v D = 0 . the instantaneous current i D and voltage v D are related as iD ( vD ) = ID e v D ⁄ nV T (6.+ ---. we have f ( t ) = e and thus f ( 0 ) = 1 .n f'' ( 0 . f' ( 0 ) = f'' ( 0 ) = f''' ( 0 ) = … = 1 and therefore.2 ) f ( x ) = f ( 0 ) + f' ( 0 )x + ----------. Third Edition Copyright © Orchard Publications .132) For this function. and Maclaurin Series Example 6. then.14 In a semiconductor diode D .Chapter 6 Fourier. pretty(fn) 2 3 4 5 1 + t + 1/2 t + 1/6 t + 1/24 t + 1/120 t Example 6. and its value at room temperature is approximately 25 mV . the constant n has a value between 1 and 2 depending on the material and physical structure of the diode. ) f (0. Expand this relation into a power series that can be used to compute the current when the voltage is small and varies about v D = 0 .

2 1i D ( v D ) = I D ⎜1 + --------.⋅ I D e dv D nV T nV T v D ⁄ nV T 1 d 1 i'' D ( v D ) = -----------.136) To compute the second and third terms of (6.138) Then.134). (6.138) into (6.v D + --------------.i D = -----------.135). we must find the first and second derivatives of (6. and (6.v D + … 2! 3! (6.137) (6.136). Then.Taylor and Maclaurin Series i'' D ( 0 ) 2 i''' D ( 0 ) 3 i D ( v D ) = i D ( 0 ) + i' D ( 0 )v D + -------------.v D + -----------.137).139) Numerical Analysis Using MATLAB® and Excel®. These are: v D ⁄ nV T d 1 1 and i' D ( 0 ) = --------.⋅ I D 2 2 2 2 2 n VT n VT d vD 2 (6.134).135) The first term i D ( 0 ) on the right side of (6. Third Edition Copyright © Orchard Publications 6−47 . by substitution of (6.135) is found by letting v D = 0 in (6.v D + …⎟ 2 2 ⎝ nV T ⎠ n VT (6.135) we get ⎛ ⎞ 1 . iD ( 0 ) = ID (6.⋅ I D e and i''D ( 0 ) = -----------.i D = --------.⋅ I D i' D ( v D ) = -------.

represent the fundamental frequency component ω . a n . all even (even cosine and even sine) harmonics will be zero. The coefficients a 0 . has half−wave symmetry if –f ( t + T ⁄ 2 ) = f ( t ) that is.11 Summary • Any periodic waveform f ( t ) can be expressed as 1 f ( t ) = -. if it is an even function. If a waveform has half−wave symmetry only odd (odd cosine and odd sine) harmonics will be present.a 0 + 2 ∞ n=1 ∑ ( a cos nωt + b sin nωt ) n n where the first term a 0 ⁄ 2 is a constant. the series will consist of sine terms only. • If a waveform has even symmetry.a 0 = ----2 2π 1 a n = -π ∫0 2π f ( t ) dt ∫0 2π f ( t ) cos nt dt 1 b n = -π ∫0 2π f ( t ) sin nt dt • If a waveform has odd symmetry. Third Edition Copyright © Orchard Publications .sin nωt n 6−48 Numerical Analysis Using MATLAB® and Excel®. • The trigonometric Fourier series for the square waveform with odd symmetry is 4A 1 1 4A f ( t ) = -----.Chapter 6 Fourier. Taylor. Likewise. the terms with the coefficients a 2 and b 2 together.sin 5ωt + …⎞ = -----⎠ π 5 3 π ⎝ n = odd ∑ 1 -. Even functions are those for which f( –t ) = f( t ) • A periodic waveform with period T . The terms with the coefficients a 1 and b 1 together.⎛ sin ωt + -. and Maclaurin Series 6. In other words. the series will consist of cosine terms only. Odd functions are those for which – f ( – t ) = f ( t ) . that is. and represents the DC (average) component of f ( t ) . and b n are found from the following relations: 1 1 -. the shape of the negative half−cycle of the waveform is the same as that of the positive half−cycle. and a 0 may or may not be zero.sin 3ωt + -. and so on. but inverted. represent the second harmonic component 2ω . if it is an odd function. that is.

( a n + jb n ) 2 2⎝ j⎠ bn 1 1 C n = -. Third Edition Copyright © Orchard Publications 6−49 .sin t – --.f ( t ) = A + --.⎛ cos ω t – -.⎞ = -.sin 2ωt + -.⎞ = -.cos n ωt n • The trigonometric Fourier series for the sawtooth waveform with odd symmetry is 2A 1 1 1 2A f ( t ) = -----.+ ------------.sin 3ωt + ----.+ ------------.------------.⎛ sin ωt – -. 4.sin 5ωt – ----. are complex. that is.Summary • The trigonometric Fourier series for the square waveform with even symmetry is 4A 1 4A 1 f ( t ) = -----.– -----π π 1 -----------------.sin 3ωt – -.a 0 2 • The C i coefficients.( a n – j b n ) ⎠ 2 2⎝ j 1 C 0 = -.+ … π 2 63 35 15 3 π • The trigonometric Fourier series for the full−wave rectification waveform with even symmetry is 2A 4A f ( t ) = -----.cos nωt 2 (n – 1) n = 2.⎛ sin ω t – -.⎛ a n + ---.+ ------------. … ∑ ∞ • The Fourier series are often expressed in exponential form as f ( t ) = … + C –2 e – j2 ω t + C –1 e –j ω t + C0 + C1 e jωt + C2 e j2 ω t +… where the C i coefficients are related to the trigonometric form coefficients as bn 1 1 C – n = -.sin 4ωt + …⎞ = -----⎝ ⎠ π 4 3 2 π ∑ ( –1 ) n–1 1 -. except C 0 .cos 3ωt + -. C – n = C n∗ Numerical Analysis Using MATLAB® and Excel®. and appear as complex conjugate pairs.sin n ωt 2 n • The trigonometric Fourier series for the half−wave rectification waveform with no symmetry is A cos 2t cos 4t cos 6t cos 8t A --.sin 7ωt + …⎞ = -----( –1 ) 2 ⎝ 2 ⎠ 49 25 9 π n = odd π ∑ (n – 1) ---------------2 1 ---.⎛ a n – ---.sin nωt n • The trigonometric Fourier series for the triangular waveform with odd symmetry is 8A 1 8A 1 1 f ( t ) = -----.cos 5ωt – …⎞ = -----⎠ π 5 π ⎝ 3 n = odd ∑ ( –1 ) (n – 1) ---------------2 1 -. 6.

**Chapter 6 Fourier, Taylor, and Maclaurin Series
**

• In general, for ω ≠ 1 , 1 C n = -T

∫0

T

f ( t )e

– jn ω t

1 d( ωt ) = ----2π

∫0

2π

f ( t )e

– jn ω t

d( ωt )

• We can derive the trigonometric Fourier series from the exponential series from the relations a n = Cn + C–n

and

b n = j ( Cn – C–n )

• For even functions, all coefficients C i are real • For odd functions, all coefficients C i are imaginary • If there is half−wave symmetry, C n = 0 for n = even • C –n = C n∗ always • A line spectrum is a plot that shows the amplitudes of the harmonics on a frequency scale. • The frequency components of a recurrent rectangular pulse follow a sin x ⁄ x form. • We can evaluate the Fourier coefficients of a function based on observed values instead of an

analytic expression using numerical evaluations with the aid of a spreadsheet.

• A power series has the form

∑ ak x k=0

in a Taylor series as

∞

k

= a0 + a1 x + a2 x + …

2

**• A function f ( x ) that possesses all derivatives up to order n at a point x = x 0 can be expanded f'' ( x 0 ) f ( x0 ) 2 n f ( x ) = f ( x 0 ) + f' ( x 0 ) ( x – x 0 ) + ------------- ( x – x 0 ) + … + ------------------- ( x – x 0 ) 2! n!
**

(n)

**If x 0 = 0 , the series above reduces to
**

f (0) n f'' ( 0 ) 2 f ( x ) = f ( 0 ) + f' ( 0 )x + ----------- x + … + ---------------- x 2! n!

(n)

**and this relation is known as Maclaurin series
**

• We can also obtain a Taylor series expansion with the MATLAB taylor(f,n,a) function where f is a symbolic expression, n produces the first n terms in the series, and a defines the Taylor approximation about point a .

6−50

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

Exercises 6.12 Exercises

1. Compute the first 5 components of the trigonometric Fourier series for the waveform below. Assume ω = 1 .

f( t) A

0

ωt

**2. Compute the first 5 components of the trigonometric Fourier series for the waveform below. Assume ω = 1 .
**

f( t) A 0 ωt

**3. Compute the first 5 components of the exponential Fourier series for the waveform below. Assume ω = 1 .
**

f(t)

A

0

ωt

**4. Compute the first 5 components of the exponential Fourier series for the waveform below. Assume ω = 1 .
**

f( t) A⁄2 0 –A ⁄ 2 ωt

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

6−51

**Chapter 6 Fourier, Taylor, and Maclaurin Series
**

5. Compute the first 5 components of the exponential Fourier series for the waveform below. Assume ω = 1 .

f(t)

A

0

ωt

**6. Compute the first 5 components of the exponential Fourier series for the waveform below. Assume ω = 1 .
**

f(t)

A ωt

0 −A Figure 6.37. Waveform for Exercise 6

**7. Compute the first 4 terms of the Maclaurin series for each of the following functions. a. f ( x ) = e
**

–x

b. f ( x ) = sin x

c. f ( x ) = sinh x

Confirm your answers with MATLAB. 8. Compute the first 4 terms of the Taylor series for each of the following functions.

1 a. f ( x ) = -- about a = – 1 x -b. f ( x ) = sin x about a = – π 4

**Confirm your answers with MATLAB. 9. In a non−linear device, the voltage and current are related as
**

v 1.5 i ( v ) = k ⎛ 1 + --- ⎞ ⎝ V⎠

where k is a constant and V is the DC component of the instantaneous voltage v . Expand this function into a power series that can be used to compute the current i , when the voltage v is small, and varies about v = 0 .

6−52

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

**Solutions to End−of−Chapter Exercises 6.13 Solutions to End−of−Chapter Exercises
**

1.

f(t) A A --- t π

– 2π

–π

0

π

2π

ωt

This is an even function; therefore, the series consists of cosine terms only. There is no half− wave symmetry and the average ( DC component) is not zero. We will integrate from 0 to π and multiply by 2 . Then,

2 a n = -π

∫0

π

From tables of integrals,

A --- t cos nt dt = 2A -----2 π π

∫0 t cos nt dt

π

(1)

∫ x cos ax dx

and thus (1) becomes

t 2A 1 - a n = ------ ⎛ ---- cos nt + -- sin nt ⎞ 2⎝ 2 ⎠ n π n

x 1= ---- cos ax + -- sin a x 2 a a

π

0

t 2A 1 1 - = ------ ⎛ ---- cos nπ + -- sin ntπ – ---- – 0 ⎞ 2⎝ 2 ⎠ 2 n π n n

**and since sin ntπ = 0 for all integer n ,
**

2A 2A 1 1 - a n = ------ ⎛ ---- cos nπ – ---- ⎞ = ---------- ( cos nπ – 1 ) (2) 2⎝ 2 2⎠ 2 2 π n n n π

**We cannot evaluate the average ( 1 ⁄ 2 ) ⁄ a 0 from (2); we must use (1). Then, for n = 0 ,
**

1 2A -- a 0 = -------2 2 2π

∫0

π

A t - t dt = ---- ⋅ --2 π 2

2 π 0

A π - = ---- ⋅ ---2 π 2

2

or

( 1 ⁄ 2 ) ⁄ a0 = A ⁄ 2

**We observe from (2) that for n = even , a n = even = 0 . Then,
**

– 4A 4A– 4A -----for n = 1, a 1 = – 4A, for n = 3, a 3 = ---------- , for n = 5, a5 = – ---------- , for n = 7, a 3 = ---------2 2 2 2 2 2 2 π 3 π 5 π 7 π

and so on. Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

6−53

**Chapter 6 Fourier, Taylor, and Maclaurin Series
**

Therefore,

4A 111 4A 1 --f ( t ) = -- a 0 – ------ ⎛ cos t + -- cos 3t + ----- cos 5t + ----- cos 7t + …⎞ = A – -----2⎝ ⎠ π 49 25 9 2 2 π

∑

n = odd

∞

1---- cos nt 2 n

2.

f( t) A 2A ------ t π 0 π⁄2 π 3π ⁄ 2 π

ωt

This is an even function; therefore, the series consists of cosine terms only. There is no half− wave symmetry and the average ( DC component) is not zero.

1 Area 2 × [ ( A ⁄ 2 ) ⋅ ( π ⁄ 2 ) ] + Aπ 3A ⋅ ( π ⁄ 2 ) 3A Average = -- a 0 = ---------------- = --------------------------------------------------------------- = -------------------------- = -----2 Period 2π 2π 4 2 a n = -π

∫0

π⁄2

2A 2 ------ t cos nt dt + -π π

and with

∫π ⁄ 2 A cos nt dt

π

(1)

∫ x cos ax dx

(1) simplifies to

**1x 1= ---- cos ax + -- sin a x = ---- ( cos ax + ax sin ax ) 2 2 a a a
**

π⁄2

0

4A 1 - a n = ------ ---- ( cos nt + nt sin nt ) 2 2 π n

2A + ------ sin nt nπ

π π⁄2

2A nπ 4A - ----= ---------- ⎛ cos nπ + ----- sin nπ – 1 – 0⎞ + ------ ⎛ sin nπ – sin nπ⎞ --------⎠ nπ ⎝ 2 2⎝ 2 2⎠ 2 2 n π

**and since sin ntπ = 0 for all integer n ,
**

4A 4A nπ 2A nπ 4A 2A nπ nπ a n = ---------- cos ----- + ------ sin ----- – ---------- – ------ sin ----- = ---------- ⎛ cos ----- – 1⎞ ⎠ 2 2 2 2 2⎝ nπ 2 2 2 2 nπ 2 n π n π n π 4A 4A 4A 2A for n = 1, a 1 = ------ ( 0 – 1 ) = – ------ , for n = 2, a 2 = -------- ( – 1 – 1 ) = – -----2 2 2 2 π 4π π π 4A – 4A 4A for n = 3, a 3 = -------- ( 0 – 1 ) = – -------- , for n = 4, a 4 = ---------- ( 1 – 1 ) = 0 2 2 2 2 9π 9π 7 π

We observe that the fourth harmonic and all its multiples are zero. Therefore,

6−54

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

**Solutions to End−of−Chapter Exercises
**

1 4A 1 -----f ( t ) = 3A – ------ ⎛ cos t + -- cos 2t + -- cos 3t + …⎞ 2⎝ ⎠ 9 2 4 π f(t)

A

3.

0

π

2π

ωt

This is neither an even nor an odd function and has no half−wave symmetry; therefore, the series consists of both cosine and sine terms. The average ( DC component) is not zero. Then,

1 C n = ----2π

∫0

π

2π

f ( t )e

– jn ω t

d( ωt )

and with ω = 1

1C n = ----2π

∫0

2π

f ( t )e

– jnt

1dt = ----2π

∫0

Ae

– jnt

dt +

∫π

π

0

2π

0e

– jnt

Adt = ----2π

∫0 e

π

– jnt

dt

The DC value is

A C 0 = ----2π

∫0

π

A 0 e dt = ----- t 2π

π

0

A = --2

For n ≠ 0

A C n = ----2π

∫0

π

e

– jnt

A – jnt dt = -------------- e – j2 nπ

A – jn π = ----------- ( 1 – e ) j2nπ

Recalling that

e

– jn π

= cos nπ – j sin nπ

– jn π

for n = even , e

– jn π

= 1 and for n = odd , e

= – 1 . Then,

A C n = even = ----------- ( 1 – 1 ) = 0 j2nπ

and

A A C n = odd = ----------- [ 1 – ( – 1 ) ] = ------j2nπ jnπ

**By substitution into the expression
**

f ( t ) = … + C –2 e

– j2 ω t

+ C –1 e

–j ω t

+ C0 + C1 e

jωt

+ C2 e

j2 ω t

+…

we find that Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

6−55

**Chapter 6 Fourier, Taylor, and Maclaurin Series
**

A 1 –j3 ω t –j ω t j ω t 1 j3 ω t --f ( t ) = A + ---- ⎛ … – -- e –e + e + -- e + …⎞ ⎠ 3 3 2 jπ ⎝

The minus (−) sign of the first two terms within the parentheses results from the fact that C – n = C n∗ . For instance, since C 1 = 2A ⁄ jπ , it follows that C –1 = C 1∗ = – 2A ⁄ jπ . We observe that f ( t ) is complex, as expected, since there is no symmetry. 4.

f( t) A⁄2 0 –A ⁄ 2 ωt

**This is the same waveform as in Exercise 3 where the DC component has been removed. Then,
**

1 – j3 ω t – j ω t A j ω t 1 j3 ω t f ( t ) = ---- ⎛ … – -- e –e + e + -- e + …⎞ ⎠ 3 jπ ⎝ 3

It is also the same waveform as in Example 6.9, Page 6−32, except that the amplitude is halved. This waveform is an odd function and thus the expression for f ( t ) is imaginary. 5.

f(t)

A

–π –π ⁄ 2

0 π⁄2

π

ωt

This is the same waveform as in Exercise 3 where the vertical axis has been shifted to make the waveform an even function. Therefore, for this waveform C n is real. Then,

1C n = ----2π

∫– π

π

f ( t )e

– jnt

Adt = ----2π

∫– π ⁄ 2 e

π⁄2

– jnt

dt

The DC value is

AC 0 = ----- t 2π

π⁄2

–π ⁄ 2

A- π π A = ----- ⎛ -- + -- ⎞ = --- 2π ⎝ 2 2⎠ 2

For n ≠ 0

6−56

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

**Solutions to End−of−Chapter Exercises
**

AC n = ----2π

∫– π ⁄ 2

π⁄2

e

– jnt

A - – jnt dt = -------------- e – j2 nπ

π⁄2

–π ⁄ 2

A - – jn π ⁄ 2 jn π ⁄ 2 = -------------- ( e –e ) – j2 nπ

– jn π ⁄ 2

–e A A e A - ----- jn π ⁄ 2 – e – jn π ⁄ 2 ) = ----- ⎛ ------------------------------------- ⎞ = ----- sin nπ = ----------- ( e ⎠ nπ j2 nπ ⎝ j2nπ 2

jn π ⁄ 2

**and we observe that for n = even , C n = 0 For n = odd , C n alternates in plus (+) and minus (−) signs, that is,
**

AC n = ----- if n = 1, 5, 9, … nπ A C n = – ----- if n = 3, 7, 11, … nπ

Thus,

A f ( t ) = --- + 2

n = odd

∑

A⎛ ± ----- e jn ω t⎞ ⎝ nπ ⎠

where the plus (+) sign is used with n = 1, 5, 9, … and the minus (−) sign is used with n = 3, 7, 11, … . We can express f ( t ) in a more compact form as

A f ( t ) = --- + 2

6.

n = odd

∑

( –1 )

(n – 1) ⁄ 2

A ----- e jn ω t nπ

f(t) –π ⁄ 2 –π −A

A

2A ------ t – 1 π π

ωt

π⁄2 0

**We will find the exponential form coefficients C n from
**

1C n = ----2π

From tables of integrals

∫–π f ( t )e

ax

π

– jnt

dt

∫

Then,

e xe dx = ------ ( ax – 1 ) 2 a

ax

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

6−57

**Chapter 6 Fourier, Taylor, and Maclaurin Series
**

1C n = ----2π

∫

⎛ – 2A t – 1⎞ e – jnt dt + -----⎝ π ⎠ –π

0

∫0 ⎛ ------- t – 1⎞ e ⎝ π ⎠

π

2A

– jnt

dt

**Integrating and rearranging terms we get
**

jn π – jn π 2A e e 1- 4A 4A –e –e +e --------------------------- ⎞ – ------ ⋅ --------------------------C n = ----- – -------- + -------- ⎛ nπ ⋅ --------------------------- + e ⎠ n 2π n 2 π n 2 π ⎝ j2 j2 2 jn π – jn π jn π – jn π

4A nπ = -------------- ⎛ – 1 + nπ sin nπ + cos nπ – ----- sin nπ⎞ 2 2⎝ ⎠ 2 2n π

**and since sin nπ = 0 for all integer n ,
**

2A C n = ---------- ( cos nπ – 1 ) 2 2 n π –4 A For n = even , C n = 0 and for n = odd , cos nπ = – 1 , and C n = ---------2 2 n π

**Also, by inspection, the DC component C 0 = 0 . Then,
**

1 – j3 ω t –j ω t j ω t 1 j3 ω t 4A +e + e + -- e + …⎞ f ( t ) = – ------ ⎛ … + -- e ⎠ 2⎝ 9 9 π

The coefficients of the terms e

– j3 ω t

and e

–j ω t

are positive because all coefficients of C n are

real. This is to be expected since f ( t ) is an even function. It also has half−wave symmetry and thus C n = 0 for n = even as we’ve found. 7.

f (0) n f'' ( 0 ) 2 f ( x ) = f ( 0 ) + f' ( 0 )x + ----------- x + … + ---------------- x 2! n!

(n)

a. f ( x ) = e , f ( 0 ) = 1 , f' ( x ) = – e , f' ( 0 ) = – 1 , f'' ( x ) = e , f'' ( 0 ) = 1 , f''' ( x ) = – e f''' ( 0 ) = – 1 , and so on. Therefore,

x x f n ( x ) = 1 – x + ---- – ---- + … 2! 3!

2 3

–x

–x

–x

–x

,

**MATLAB displays the same result.
**

x=sym('x'); fn=taylor(exp(−x)); pretty(fn) 2 1 - x + 1/2 x - 1/6 x 3 + 1/24 x 4 - 1/120 x 5

6−58

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

**Solutions to End−of−Chapter Exercises
**

b. f ( x ) = sin x ,

f (0) = 0 , f' ( x ) = cos x , f' ( 0 ) = 1 , f'' ( x ) = – sin x , f'' ( 0 ) = 0 , f''' ( x ) = – cos x , f''' ( 0 ) = – 1 , and so on. Therefore, x x x f n ( x ) = x – ---- + ---- – ---- + … 3! 5! 7!

3 5 7

**MATLAB displays the same result.
**

x=sym('x'); fn=taylor(sin(x)); pretty(fn) 3 x - 1/6 x + 1/120 x f'' ( x ) = sin hx , f'' ( 0 ) = 0 , 5

c. f ( x ) = sin hx ,

**f ( 0 ) = 0 , f' ( x ) = cos hx , f' ( 0 ) = 1 , f''' ( x ) = cos hx , f''' ( 0 ) = 1 , and so on. Therefore, x x x f n ( x ) = x + ---- + ---- + ---- + … 3! 5! 7!
**

3 5 7

**MATLAB displays the same result.
**

x=sym('x'); fn=taylor(sinh(x)); pretty(fn) 3 x + 1/6 x + 1/120 x 5

8.

f'' ( a ) 2 f''' ( a ) 3 f n ( x ) = f ( a ) + f' ( a ) ( x – a ) + ----------- ( x – a ) + ------------ ( x – a ) + … 2! 3!

a. f ( x ) = 1 ⁄ x , f ( a ) = f ( – 1 ) = – 1 , f' ( x ) = – 1 ⁄ x , f' ( a ) = f' ( – 1 ) = – 1 , f'' ( x ) = 2 ⁄ x ,

f'' ( a ) = f'' ( – 1 ) = – 2 , f''' ( x ) = – 6 ⁄ x , f''' ( a ) = f''' ( – 1 ) = – 6 , and so on. Therefore, fn ( x ) = –1 –( x + 1 ) – ( x + 1 ) – ( x + 1 ) + …

2 3 4

2

3

or

fn ( x ) = –2 –x – ( x + 1 ) – ( x + 1 ) + …

2

3

**MATLAB displays the same result.
**

x=sym('x'); y=1/x; z=taylor(y,4,−1); pretty(z) 2 -2 - x - (x + 1) - (x + 1) f' ( x ) = cos x , f' ( a ) = f' ( – π ⁄ 4 ) = 2 ⁄2, 3

b. f ( x ) = sin x ,

f ( a ) = f ( –π ⁄ 4 ) = – 2 ⁄ 2 ,

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

6−59

**Chapter 6 Fourier, Taylor, and Maclaurin Series
**

f'' ( x ) = – sin x , f'' ( a ) = f'' ( – π ⁄ 4 ) = 2 ⁄ 2, f''' ( x ) = – cos x ,

f''' ( a ) = f''' ( – π ⁄ 4 ) = – 2 ⁄ 2 , and so on. Therefore, f n ( x ) = – 2 ⁄ 2 + ( 2 ⁄ 2 ) ( x + π ⁄ 4 ) + ( 2 ⁄ 4 ) ( x + π ⁄ 4 ) – ( 2 ⁄ 12 ) ( x + π ⁄ 4 ) + …

2 3

**MATLAB displays the same result.
**

x=sym('x'); y=sin(x); z=taylor(y,4,−pi/4); pretty(z) 1/2 - 1/2 2 1/2 - 1/12 2 (x + 1/4 pi) v 1.5 i ( v ) = k ⎛ 1 + --- ⎞ ⎝ V⎠ + 1/2 2 1/2 (x + 1/4 pi) + 1/4 2 3 1/2 (x + 1/4 pi) 2

9.

**The Taylor series for this relation is
**

i'' ( v 0 ) 2 i''' ( v 0 ) 3 i ( v ) = i ( v 0 ) + i' ( v 0 ) ( v – v 0 ) + ------------- ( v – v 0 ) + -------------- ( v – v 0 ) + … 2! 3!

Since the voltage v is small, and varies about v = 0 , we expand this relation about v = 0 and the series reduces to the Maclaurin series below.

i'' ( 0 ) 2 i ( v ) = i ( 0 ) + i' ( 0 )v + ----------- v + … 2!

(1)

**By substitution of v = 0 into the given relation we get
**

i(0) = k

**The first and second derivatives of i are
**

3k v 1⁄2 i' ( v ) = ------ ⎛ 1 + --- ⎞ 2V ⎝ V⎠ 3k v –1 ⁄ 2 i'' ( v ) = --------- ⎛ 1 + --- ⎞ 2⎝ V⎠ 4V 3k i' ( 0 ) = -----2V 3k i'' ( 0 ) = --------2 4V

**and by substitution into (1)
**

3k 2 3 2 3k 3 i ( v ) = k + ------ v + --------- v + … = k ⎛ 1 + ------ v + --------- v + …⎞ 2 2 ⎝ ⎠ 2V 2V 8V 8V

6−60

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

**Solutions to End−of−Chapter Exercises
**

MATLAB displays the same result.

x=sym('x'); i=sym(‘i’); v=sym(‘v’); k=sym(‘k’); V=sym(‘V’);... i=k*(1+v/V)^1.5; z=taylor(i,4,0); pretty(z) 2 3 k v k v k v k + 3/2 --- + 3/8 ---- - 1/16 ---V 2 3 V V

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

6−61

x 2. x j ) = ---------------------------xi – xj (7. data smoothing. x n – 1. x j ) – f ( x j. Then. T 7. x k ) f ( x i. and numerical integration. x j. the first divided difference is defined as: f ( xi ) – f ( xj ) f ( x i. x k ) = --------------------------------------------xi – xk (7. x n be some values of x in the interval x 0 ≤ x ≤ x n . ….1.Chapter 7 Finite Differences and Interpolation his chapter begins with finite differences and interpolation which is one of its most important applications.1) Likewise. x 1. It is customary to show the independent variable x . within this interval. the second divided difference is defined as: f ( x i. numerical differentiation. Third Edition Copyright © Orchard Publications 7−1 . Finite Differences form the basis of numerical analysis as applied to other numerical methods such as curve fitting. These applications are discussed in this and the next three chapters.2) Numerical Analysis Using MATLAB® and Excel®.1 Divided Differences Consider the continuous function y = f ( x ) and let x 0. TABLE 7. and its corre- sponding values of y = f ( x ) in tabular form as in Table 7.1 The variable x and y = f ( x ) in tabular form x x0 x1 x2 f (x) f ( x0 ) f ( x1 ) f ( x2 ) … xn – 1 xn … f ( xn – 1 ) f ( xn ) Let x i and x j be any two. not necessarily consecutive values of x .

= 14 3–7 Likewise. TABLE 7.1) as 1 – 27 -------------.3 with six columns. 1. x 3 ) Example 7. x 1 ) x1 f ( x1 ) f ( x 1. 3. the second value on the first divided difference is found from (7. 4. x 3 ) f ( x 0. x 2. 2. x 2. (7. x 1. the second the values of f ( x ) . Third Edition Copyright © Orchard Publications . The divided differences are indicated in a difference table where each difference is placed between the values of the column immediately to the left of it as shown in Table 7.2).2.Chapter 7 Finite Differences and Interpolation The third. For instance. x 2 ) f ( x 0. The first column contains the given values of x .2 Conventional presentation of divided differences x x0 f (x) f ( x0 ) f ( x 0. These differences are computed from (7.= 13 1–3 and third value on the second divided difference is found from (7. and the first through the fourth divided differences. and other relations for higher order divided differences. x 2 ) x2 f ( x2 ) f ( x 2.2) as 37 – 93 ----------------. and so on divided differences. the values of f ( x ) corresponding to y = f ( x ) = x .1 Form a difference table showing the values of x given as 0.1). 3 Solution: We construct Table 7. 7. x 1. x 3 ) x3 f ( x3 ) f ( x 1. fourth. and 9 . for the third divided difference we have 7−2 Numerical Analysis Using MATLAB® and Excel®. and the third through the sixth contain the values of the first through the fourth divided differences. are defined similarly.

These values are referred to as just the differences of the function. the denominators are all the same. the differences are sets of consecutive values. the second differences are Δ f k = Δ ( Δf k ) = Δf k + 1 – Δf k 2 (7. all subsequent differences will be equal to zero. – 1.– 2. if the values of the nth divided difference are the same. as in the fifth column (third divided differences for this example). Moreover. … (7. 0. 2.5) Numerical Analysis Using MATLAB® and Excel®. therefore. the values of x in a table are equally spaced.3) We can now express the first differences in terms of the difference operator Δ as Δf k = f k + 1 – f k (7. Third Edition Copyright © Orchard Publications 7−3 . In this case.1 Function x 0 1 3 4 7 9 Divided Differences 3 f( x) = x 0 First 1 Second Third Fourth 1 13 27 37 64 93 343 193 729 4 1 8 1 14 1 20 0 0 4–8 = 1 ----------0–4 and for the fourth 1–1 ----------. they can be omitted. 1. If the constant difference between successive values of x is h . the typical value of x k is x k = x 0 + kh for k = … .= 0 0–4 We observe that.Divided Differences TABLE 7.3 Divided differences for Example 7. In most cases.4) Likewise.

3. that is. 4.4 Divided differences table in terms of the difference operator Δ Function x f First Second Differences Third Fourth … x0 f0 Δf 0 x1 f1 Δf 1 Δ f0 Δ f0 Δ f1 Δf 2 Δ f1 Δ f2 Δf 3 2 3 2 3 2 x2 f2 Δ f0 4 x3 f3 x4 … xn f4 fn Example 7. 3 7−4 Numerical Analysis Using MATLAB® and Excel®. 6. Δ ( Δ fk ) = Δ m n m+n fk (7. 7 and 8 .7) We construct the difference table in terms of the difference operator Δ as shown in Table 7. Third Edition Copyright © Orchard Publications .2 Construct a difference table showing the values of x given as 1. for positive integer values of n Δ fk = Δ ( Δ n n–1 fk ) = Δ n–1 fk + 1 – Δ n–1 fk (7.6) The difference operator Δ obeys the law of exponents. TABLE 7. and the first through the fourth differences.Chapter 7 Finite Differences and Interpolation and. 5. 2. the values of f ( x ) corresponding to y = f ( x ) = x . in general.4.

5.9) For k = 0 . as expected.5 Difference table for Example 7.f k + n – 2 + … + ( – 1 ) nf k + 1 + ( – 1 ) f k 2! (7.10) Numerical Analysis Using MATLAB® and Excel®. n = 1.Divided Differences Solution: Following the same procedure as in the previous example.2 Function k 1 2 3 4 5 6 7 8 xk 1 2 3 4 5 6 7 8 fk 1 7 8 19 27 37 64 61 125 91 216 127 343 169 512 4 Differences Δf k Δ fk 2 Δ fk 3 Δ fk 4 … 12 6 18 6 24 6 30 6 36 6 42 0 0 0 0 We observe that the fourth differences Δ f k are zero. 2. Using the binomial expansion we can show that n! ⎛ n⎞ = -------------------⎝ j⎠ j! ( n – j )! (7.8) n n–1 n n(n – 1) Δ f k = f k + n – nf k + n – 1 + ------------------. we construct Table 7. TABLE 7. relation (7. 3 and 4 . Third Edition Copyright © Orchard Publications 7−5 .9) reduces to Δf 0 = f 2 – f 1 Δ f 0 = f 2 – 2f 1 + f 0 Δ f 0 = f 3 – 3f 2 + 3f 1 – f 0 Δ f 0 = f 4 – 4f 3 + 6f 2 – 4 f 1 + f 0 4 3 2 (7.

14) we obtain Δp n ( x ) = 1x a 1 + 2a 2 ( x ) 0 (1) + 3a 3 ( x ) (2) + … + na n ( x ) (n – 1) (7. we take the first difference of p n ( x ) in (7. the nth differences of a polynomial of degree n are constant. is the difference quotient whose limit defines the derivative of a continuous function that is defined as Δx → 0 lim f ( x 1 + Δx ) – f ( x 1 ) Δy -----. we can express that polynomial as pn ( x ) = a0 + a1 ( x ) (1) + a2 ( x ) (2) + … + an ( x ) (n) (7. we find that 7−6 Numerical Analysis Using MATLAB® and Excel®.15) are very similar to differentiation of x and x .2 Factorial Polynomials The factorial polynomials are defined as and (x) (n) = x ( x – 1 ) ( x – 2 )… ( x – n + 1 ) 1 = --------------------------------------------------------( x – 1 ) ( x – 2 )… ( x + n ) n –n (7.12) and (7. relation (7.13) (x) –( n ) These expressions resemble the power functions x and x in elementary algebra.15) n –n –( n ) = –n ( x ) –( n – 1 ) We observe that (7.16) and now our task is to compute the coefficients a k . Using (7.13) we obtain and Δ(x) Δ(x) (n) = n(x) (n – 1) (7.16) reduces to a0 = pn ( 0 ) (7.17) To compute the coefficient a 1 .10). Then.14) (7.= lim ------------------------------------------Δx Δx Δx → 0 (7. 7. Occasionally.16).18) and letting x = 0 . it is desirable to express a polynomial p n ( x ) as a factorial polynomial. For x = 0 . in analogy with Maclaurin power series.11) As with derivatives.Chapter 7 Finite Differences and Interpolation It is interesting to observe that the first difference in (7.12) (7. Using the difference operator Δ with (7. Third Edition Copyright © Orchard Publications .14) and (7.

16).25) 3.25) by ( x – 2 ) to obtain a quotient q 2 ( x ) and a remainder r 2 which turns out to be the constant term a 2 . we obtain p n ( x ) = r 0 + x [ r 1 + ( x – 1 )q 1 ( x ) ] = r 0 + r 1 ( x ) (1) + x ( x – 1 )q 1 ( x ) (7. we express (7.24) By substitution of (7. With this method we perform the following steps: 1.for j = 0.27) Numerical Analysis Using MATLAB® and Excel®.24) into (7. (7. we obtain pn ( x ) = r0 + r1 ( x ) = r0 + r1 ( x ) (1) (1) + x ( x – 1 ) [ r 2 + ( x – 2 )q 2 ( x ) ] + r2 ( x ) (2) + x ( x – 1 ) ( x – 2 )q 2 ( x ) (7.25). and using the form of relation (7.23) 2.20) and for x = 0 . We divide q 1 ( x ) in (7. We divide q 0 ( x ) in (7. and thus q 1 ( x ) = r 2 + ( x – 2 )q 2 ( x ) (7. ….23) by ( x – 1 ) to obtain a quotient q 1 ( x ) and a remainder r 1 which turns out to be the constant term a 1 . 2.26) By substitution of (7. q 0 ( x ) = r 1 + ( x – 1 )q 1 ( x ) (7.26) into (7.21) In general.16) as p n ( x ) = r 0 + xq 0 ( x ) (7. 1. Then.23).= ------------------2⋅1 2! Δ pn ( 0 ) a j = -----------------.Factorial Polynomials a 1 = Δp n ( 0 ) (7. Then.22) Factorial polynomials provide an easier method of constructing a difference table. n j! j 2 (7. We divide p n ( x ) in (7. Third Edition Copyright © Orchard Publications 7−7 .19) (n – 2) Differencing again we obtain Δ p n ( x ) = 2 ⋅ 1a 2 + 3 ⋅ 2a 3 ( x ) 2 (1) + … + n ( n – 1 )a n ( x ) 2 (7.16) by x to obtain a quotient q 0 ( x ) and a remainder r 0 which turns out to be the constant term a 0 . Δ pn ( 0 ) Δ pn ( 0 ) a 2 = ------------------.

31) as a factorial polynomial. [q3. x % Computation of first quotient and remainder % Coefficients of second divisor.22). we will use the MATLAB deconv(p.r0]=deconv(px. r 1.d1) d2=[1 −2]. construct the difference table with h = 1 . i. i. x−1 % Computation of second quotient and remainder % Coefficients of third divisor. we must evaluate the remainders r 0.e. we will use (7. [q2. x−2 % Computation of third quotient and remainder % Coefficients of fourth divisor. The general form of a factorial polynomial is pn ( x ) = r0 + r1 ( x ) (1) + r2 ( x ) (2) + … + rn – 1 ( x ) j (n – 1) + rn ( x ) (n) (7.d2) d3=[1 −3].d0) d1=[1 −1]. The MATLAB script is as follows: px=[1 −5 0 3 4].16) and (7. i.r2]=deconv(q1.e.28) and from (7.30) or Δ p n ( 0 ) = j!r j j Example 7. but for convenience. we obtain a new quotient whose degree is one less than preceding quotient and therefore. Solution: Since the highest power of the given polynomial p ( x ) is 4 .r4]=deconv(q3. Third Edition Copyright © Orchard Publications . then.r1]=deconv(q0.d4) % Coefficients of given polynomial % Coefficients of first divisor.d3) d4=[1 −4].29) (7. d0=[1 0]. Δ pn ( 0 ) r j = a j = -----------------j! (7. r 3 and r 4 . r 2. Then. x−3 % Computation of fourth quotient and remainder % Coefficients of fifth (last) divisor. We can compute the remainders by long division. [q0.28) to determine p n ( x ) .r3]=deconv(q2.Chapter 7 Finite Differences and Interpolation Continuing with the above procedure.e. [q1.e.q) function which divides the polynomial p by q.3 Express the algebraic polynomial p ( x ) = x – 5x + 3x + 4 4 3 (7. i.e. x−4 % Computation of fifth (last) quotient and remainder q0 = 7−8 Numerical Analysis Using MATLAB® and Excel®. i. the process of finding new quotients and remainders terminates after ( n + 1 ) steps. [q4.

32).31). with reference to (7. For this example. This can be easily done with the MATLAB collect(‘s_expr’) function. We will now compute the leading entries for the difference table using (7. Then.30) and (7. the factorial polynomial is pn ( x ) = 4 – ( x ) (1) (2) (3) (4) -5 0 -4 0 -2 0 0 0 -4 0 3 0 4 -1 -8 1 – 8(x) + (x) + (x) (7. by expansion of the factorials using (7. Third Edition Copyright © Orchard Publications 7−9 .32) We can verify that (7. the MATLAB script is syms x. Numerical Analysis Using MATLAB® and Excel®.12).Factorial Polynomials 1 r0 = 0 q1 = 1 r1 = 0 q2 = 1 r2 = 0 q3 = 1 r3 = 0 q4 = 0 r4 = 1 Therefore. px=collect((x*(x−1)*(x−2)*(x−3))+(x*(x−1)*(x−2))−(8*x*(x−1))−x+4) px = x^4-5*x^3+3*x+4 We observe that this is the same algebraic polynomial as in (7. where ‘s_expr’ is a symbolic expression.31).32) is the same polynomial as (7.28).

8. Third Edition Copyright © Orchard Publications . We enter the values of (7.7.6 Leading entries of (7.Chapter 7 Finite Differences and Interpolation Δ p ( 0 ) = 0! ⋅ 4 = 4 Δ p ( 0 ) = 1! ⋅ ( – 1 ) = – 1 Δ p ( 0 ) = 2! ⋅ ( – 8 ) = – 16 Δ p ( 0 ) = 3! ⋅ 1 = 6 Δ p ( 0 ) = 4! ⋅ 1 = 24 Δ p ( 0 ) = 5! ⋅ 0 = 0 5 4 3 2 1 0 (7. We obtain the next set of values by crisscross addition as shown in Table 7.33) 1. 2. 3.33) in table form x p(x) 4 Δ −1 Δ 2 Δ 3 Δ 4 Δ 5 −16 6 24 0 TABLE 7.6.7 Crisscross addition to find second set of values x p(x) 4 Δ Δ 2 Δ 3 Δ 4 Δ 5 −1 3 −16 −17 −10 30 24 6 24 0 7−10 Numerical Analysis Using MATLAB® and Excel®. The second crisscross addition extends the difference table as shown in Table 7.33) in the appropriate spaces as shown in Table 7. TABLE 7.

Factorial Polynomials TABLE 7. TABLE 7.9 Complete difference table for Example 7. This is shown in Table 7. Third Edition Copyright © Orchard Publications 7−11 .9. Continuation of this procedure produces the complete difference table.8 Second crisscross addition to find third set of values x p(x) 4 Δ Δ 2 Δ 3 Δ 4 Δ 5 −1 3 −16 −17 6 −14 −27 −10 30 20 54 24 0 24 4.3 x p(x) 4 Δ Δ 2 Δ 3 Δ 4 Δ 5 −1 3 −16 −17 6 −14 −27 −41 −7 −48 67 19 −10 30 20 54 74 24 0 24 Numerical Analysis Using MATLAB® and Excel®.

by (7. its antidifference is (x) –1 (1) (x) (x) (x) Δ p n ( x ) = -----------.+ -----------.+ 4 ( x ) + C 3 5 4 2 (5) (4) (3) (2) (7. These are discussed below.36) Then. It is computed from Δ (x) –1 –1 (n) (x) = -------------------(n + 1) (n + 1) (7.– 8 ------------. in finite differences we have the definite sum of p n ( x ) which for the interval a ≤ x ≤ a + ( n – 1 )h is denoted as α + ( n – 1 )h x=α ∑ p n ( x ) = p n ( α ) + p n ( α + h ) + p n ( α + 2h ) + … + p n [ α + ( n – 1 )h ] (7. we can find the antidifference of a factorial polynomial. Before we present an example.35) Solution: This is the same algebraic polynomial as that of the previous example. In analogy with definite integrals for continuous functions. Third Edition Copyright © Orchard Publications .4 Compute the antidifference of the algebraic polynomial p ( x ) = x – 5x + 3x + 4 4 3 (7. We denote the antidifference as Δ p n ( x ) .37) where C is an arbitrary constant.– ------------. in analogy with the fundamental theorem of integral calculus which states that 7−12 Numerical Analysis Using MATLAB® and Excel®.34) Example 7. we can integrate or antidifferentiate to find the function. By a similar method.38) Also.3 Antidifferences We recall from elementary calculus that when we know the first derivative of a function. where we found that the corresponding factorial polynomial is pn ( x ) = 4 – ( x ) (1) – 8(x) (2) + (x) (3) + (x) (4) (7. we need to review the definite sum and the fundamental theorem of sum calculus. Antidifferences are very useful in finding sums of series.Chapter 7 Finite Differences and Interpolation 7.34).

Antidifferences

∫a f ( x ) dx

α + ( n – 1 )h

b

= f(b) – f(a )

(7.39)

we have the fundamental theorem of sum calculus which states that

∑ x=α

pn ( x ) = Δ pn ( x )

–1

α + nh α

(7.40)

Example 7.5 Derive a simple expression, in closed form, that computes the sum of the cubes of the first n odd integers. Solution: An odd number can be expressed as 2m – 1 , and thus its cube is ( 2m – 1 ) . To use (7.40), we must express this term as a factorial polynomial. Recalling from (7.12) that

(x)

(n)

3

= x ( x – 1 ) ( x – 2 )… ( x – n + 1 )

(7.41)

**and using the MATLAB expand(f) function where f is a symbolic expression, we execute
**

syms m; f = (2*m−1)^3; expand(f)

**and we obtain ans = 8*m^3-12*m^2+6*m-1 Thus
**

p ( m ) = ( 2m – 1 ) = 8m – 12m + 6m – 1

3 3 2

(7.42)

**Following the procedure of Example 7.3, we find p n ( m ) with MATLAB as
**

pm=[8 −12 6 −1]; d0=[1 0]; [q0,r0]=deconv(pm,d0) d1=[1 −1]; [q1,r1]=deconv(q0,d1) d2=[1 −2]; [q2,r2]=deconv(q1,d2) d3=[1 −3]; [q3,r3]=deconv(q2,d3)

q0 = 8 -12 6

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

7−13

**Chapter 7 Finite Differences and Interpolation
**

r0 = 0 q1 = 8 r1 = 0 q2 = 8 r2 = 0 q3 = 0 r3 = 8 Therefore,

pn ( m ) = 8 ( m )

(3)

0 -4 0

0

-1

2

12

+ 12 ( m )

(2)

+ 2(m)

(1)

–1

(7.43)

**Taking the antidifference of (7.43) we obtain
**

–1 (1) Δ p n ( m ) = 8 ( m ) + 12 ( m ) + 2 ( m ) – ( m ) ------------------ --------------------- -----------------4 3 2 (4) (3) (2)

(7.44)

= 2(m)

(4)

+ 4(m)

(3)

+ (m)

(2)

– (m)

(1)

and with (7.40)

∑

Since

cubes = 2 ( m )

(4)

+ 4(m)

(3)

+ (m)

(2)

– (m)

(1)

n+1 m=1

= 2 ( n + 1 )n ( n – 1 ) ( n – 2 ) + 4 ( n + 1 )n ( n – 1 ) + ( n + 1 )n – ( n + 1 ) –2 ( 1 )

(4)

(7.45)

–4 ( 1 )

(4) (3) (2) (1)

(3)

– (1)

(2)

+ (1)

(1)

(1) (1) (1) (1)

= 1(1 – 1)(1 – 2 )(1 – 3 ) = 0 = 1(1 – 1)(1 – 2 ) = 0 = 1(1 – 1) = 0 = 1

(7.46)

relation (7.45) reduces to

∑ cubes

7−14

= 2 ( n + 1 )n ( n – 1 ) ( n – 2 ) + 4 ( n + 1 )n ( n – 1 ) + ( n + 1 )n – ( n + 1 ) + 1

(7.47)

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

**Newton’s Divided Difference Interpolation Method
**

and this can be simplified with the MATLAB collect(f) function as follows.

syms n; sum=collect(2*(n+1)*n*(n−1)*(n−2)+4*(n+1)*n*(n−1)+(n+1)*n−(n+1)+1)

sum = 2*n^4-n^2 that is,

∑ cubes

= 2n – n = n ( 2n – 1 )

4

2

2

2

(7.48)

We can verify that this is the correct expression by considering the first 4 odd integers 1, 3, 5, and 7 . The sum of their cubes is

1 + 27 + 125 + 343 = 496

**This is verified with (7.48) since
**

n ( 2n – 1 ) = 4 ( 2 ⋅ 4 – 1 ) = 16 ⋅ 31 = 496

2 2 2 2

One important application of finite differences is interpolation. Newton’s divided−difference interpolation method, Lagrange’s interpolation method, Gregory−Newton forward, and Gregory− Newton backward interpolation methods are discussed in Sections 7.4 through 7.7 below. We will use spreadsheets to facilitate the computations. Interpolation using MATLAB is discussed in Section 7.8 below.

**7.4 Newton’s Divided Difference Interpolation Method
**

This method, has the advantage that the values x 0, x 1, x 2, …, x n need not be equally spaced, or taken in consecutive order. It uses the formula

f ( x ) = f ( x 0 ) + ( x – x 0 ) f ( x 0, x 1 ) + ( x – x 0 ) ( x – x 1 ) f ( x 0, x 1, x 2 ) + ( x – x 0 ) ( x – x 1 ) ( x – x 2 ) f ( x 0, x 1, x 2, x 3 )

(7.49)

where f ( x 0, x 1 ) , f ( x 0, x 1, x 2 ) , and f ( x 0, x 1, x 2, x 3 ) are the first, second, and third divided differences respectively. Example 7.6 Use Newton’s divided−difference method to compute f ( 2 ) from the experimental data shown in Table 7.10.

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

7−15

**Chapter 7 Finite Differences and Interpolation
**

TABLE 7.10 Data for Example 7.6 x y = f (x) −1.0 3.0 0.0 −2.0 0.5 −0.375 1.0 3.0 2.5 16.125 3.0 19.0

Solution: We must compute the first, second, and third divided differences as required by (7.49). The first divided differences are:

– 2.000 – 3.000 ------------------------------------ = – 5.000 0 – ( – 1.0 ) – 0.375 – ( – 2.000 ) = 3.250 -------------------------------------------0.5 – 0.0 3.000 – ( – 0.375 ) ---------------------------------------- = 6.750 1.0 – 0.5 16.125 – 3.000 ----------------------------------- = 8.750 2.5 – 1.0 19.000 – 16.125 = 5.750 -------------------------------------3.0 – 2.5

(7.50)

**The second divided differences are:
**

3.250 – ( – 5.000 ) = 5.500 ---------------------------------------0.5 – ( – 1.0 ) 6.750 – 3.250 -------------------------------- = 3.500 1.0 – 0.0 8.750 – 6.750 = 1.000 -------------------------------2.5 – 0.5 5.750 – 8.750 -------------------------------- = – 1.500 3.0 – 1.0

(7.51)

**and the third divided differences are:
**

3.500 – 5.500 -------------------------------- = – 1.000 1.0 – ( – 1.0 ) 1.000 – 3.500 -------------------------------- = – 1.000 2.5 – 0.0 – 1.500 – 1.000 ------------------------------------ = – 1.000 3.0 – 0.5

(7.52)

With these values, we construct the difference Table 7.11.

7−16

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

**Lagrange’s Interpolation Method
**

TABLE 7.11 Difference table for Example 7.6

1st Divided Difference 2nd Divided Difference 3rd Divided Difference

x −1.0 0.0 0.5 1.0 2.5 3.0

f (x) 3.000 −2.000

f ( x 0, x 1 ) −5.000

f ( x 0, x 1, x 2 )

f ( x 0, x 1, x 2, x 3 )

5.500 3.250 −1.000 3.500 6.750 −1.000 1.000 8.750 −1.000 −1.500 5.750

−0.375 3.000 16.125 19.000

Now, we have all the data that we need to find f ( 2 ) . We start with x 0 = 0.00 ,* and for x in (7.49), we use x = 2 . Then,

f ( 2 ) = – 2.0 + ( 2 – 0 ) ( 3.250 ) + ( 2 – 0 ) ( 2 – 0.5 ) ( 3.500 ) + ( 2 – 0 ) ( 2 – 0.5 ) ( 2 – 1 ) ( – 1.000 ) = – 2.0 + 6.5 + 10.5 – 3 = 12

This, and other interpolation problems, can also be solved with a spreadsheet. The Excel spreadsheet for this example is shown in Figure 7.1.

**7.5 Lagrange’s Interpolation Method
**

Lagrange’s interpolation method uses the formula

( x – x 1 ) ( x – x 2 )… ( x – x n ) ( x – x 0 ) ( x – x 2 )… ( x – x n ) f ( x ) = ----------------------------------------------------------------------- f ( x 0 ) + ----------------------------------------------------------------------- f ( x 1 ) ( x 0 – x 1 ) ( x 0 – x 2 )… ( x 0 – x n ) ( x 1 – x 0 ) ( x 1 – x 2 )… ( x 1 – x n ) ( x – x 0 ) ( x – x 1 )… ( x – x n – 1 ) + ----------------------------------------------------------------------------- f ( x n ) ( x n – x 0 ) ( x n – x 2 )… ( x n – x n – 1 )

(7.53)

and, like Newton’s divided difference method, has the advantage that the values x 0, x 1, x 2, …, x n need not be equally spaced or taken in consecutive order.

* We chose this as our starting value so that f ( 2 ) will be between f ( 1 ) and f ( 2.5 )

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

7−17

Chapter 7 Finite Differences and Interpolation

Interpolation with Newton's Divided Difference Formula f(x) = f(x0)+(x-x0)f(x0,x1)+(x-x0)(x-x1)f(x0,x1,x2)+(x-x0)(x-x1)(x-x2)f(x0,x1,x2,x3) In this example, w e w ant to evaluate f(x) at x= 2 1st divided difference x -1.00 f(x) 3.000 -5.000 0.00 -2.000 3.250 0.50 -0.375 6.750 1.00 3.000 8.750 2.50 16.125 5.750 3.00 19.000 We use the above formula w ith starting value x0.00 f(2)=B12+(E3-E18)*C13+(E3-E18)*(E3-A14)*D14+(E3-E18)*(E3-A14)*(E3-A16)*E15 or f(2)= 12.00 The plot below verifies that our answ er is correct -1.000 3.000

20 15

2nd divided 3rd divided difference f(x0, x1, x2) difference f(x0,x1,x2,x3)

f(x0, x1)

5.500 -1.000 3.500 -1.000 1.000 -1.000 -1.500

0.000 -2.000 f(x) 0.500 -0.375 1.000 3.000 2.500 16.125 3.000 19.000

10 5 0 -5 -1.0 0.0 1.0 2.0 3.0

x

Figure 7.1. Spreadsheet for Example 7.6

Example 7.7 Repeat Example 7.6 using Lagrange’s interpolation formula. Solution: All computations appear in the spreadsheet of Figure 7.2 where we have used relation (7.53).

7−18

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

**Gregory−Newton Forward Interpolation Method
**

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 A B C

2 f(x) 3.000 -2.000 -0.375 3.000 16.125 19.000 x-x1 2.000 x0-x1 -1.000 x-x0 3.000 x1-x0 1.000 x-x0 3.000 x2-x0 1.500 x-x0 3.000 x3-x0 2.000 x-x0 3.000 x4-x0 3.500 x-x0 3.000 x5-x0 4.000 x-x2 1.500 x0-x2 -1.500 x-x2 1.500 x1-x2 -0.500 x-x1 2.000 x2-x1 0.500 x-x1 2.000 x3-x1 1.000 x-x1 2.000 x4-x1 2.500 x-x1 2.000 x5-x1 3.000 x-x3 1.000 x0-x3 -2.000 x-x3 1.000 x1-x3 -1.000 x-x3 1.000 x2-x3 -0.500 x-x2 1.500 x3-x2 0.500 x-x2 1.500 x4-x2 2.000 x-x2 1.500 x5-x2 2.500 x-x4 -0.500 x0-x4 -3.500 x-x4 -0.500 x1-x4 -2.500 x-x4 -0.500 x2-x4 -2.000 x-x4 -0.500 x3-x4 -1.500 x-x3 1.000 x4-x3 1.500 x-x3 1.000 x5-x3 2.000 x-x5 -1.000 x0-x5 -4.000 x-x5 -1.000 x1-x5 -3.000 x-x5 -1.000 x2-x5 -2.500 x-x5 -1.000 x3-x5 -2.000 x-x5 -1.000 x4-x5 -0.500 x-x4 -0.500 x5-x4 0.500 f(x0) 3.000

D

E

F

G

H

I

J

Numer. Partial Prods 4.500

K

Denom. Partial Prods

L

Division of Partial Prods

Lagrange's Interpolation Method Interpol. at x= x -1.00 0.00 0.50 1.00 2.50 3.00

x0 x1 x2 x3 x4 x5

-0.107 -42.000 f(x1) -2.000 -4.500 -1.200 3.750 f(x2) -0.375 -1.125 0.600 -1.875 f(x3) 3.000 13.500 4.500 3.000 f(x4) 16.125 -145.125 11.057 -13.125 f(x5) 19.000 -85.500 -2.850 30.000 f(2)= Sum= 12

Figure 7.2. Spreadsheet for Example 7.7

**7.6 Gregory−Newton Forward Interpolation Method
**

This method uses the formula

r(r – 1) 2 r( r – 1 )( r – 2) 3 f ( x ) = f 0 + rΔf 0 + ----------------- Δ f 0 + --------------------------------- Δ f 0 + … 2! 3!

2 3

(7.54)

where f 0 is the first value of the data set, Δf 0 , Δ f 0 , and Δ f 0 are the first, second, and third forward* differences respectively. The variable r is the difference between an unknown point x and a known point x 1 divided by the interval h , that is,

( x – x1 ) r = -----------------h

(7.55)

* This is an expression to indicate that we use the differences in a forward sequence, that is, the first entries on the columns

where the differences appear.

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

7−19

**Chapter 7 Finite Differences and Interpolation
**

The formula of (7.54) is valid only when the values x 0, x 1, x 2, …, x n are equally spaced with interval h . It is used to interpolate values near the smaller values of x , that is, the values near the beginning of the given data set. The formula that we will study on the next section, is used to interpolate values near the larger values of x , that is, the values near the end of the given data set. Example 7.8 Use the Gregory−Newton forward interpolation formula to compute f ( 1.03 ) from the following data.

TABLE 7.12 Table for Example 7.8 x y = f (x) 1.00 1.000000 1.05 1.257625 1.10 1.531000 1.15 1.820875 1.20 2.128000 1.25 2.453125

Solution: We enter the given x and f ( x ) values in a difference table; then, we compute the first, second, and third differences. These are not divided differences and therefore, we simply subtract the second value of f ( x ) from the first, the third from the second, and so on, as shown in Table 7.13. For this example,

f 0 = f ( 1.00 ) = 1.000000 h = x 1 – x 0 = 1.05 – 1.00 = 0.05 x – x1 1.03 – 1.00 r = ------------- = -------------------------- = 0.60 h 0.05

(7.56)

**and with the values shown in Table 7.13 and using (7.54), we obtain
**

( 0.60 ) ⋅ ( 0.60 – 1 ) f ( 1.03 ) = 1.000000 + ( 0.60 ) ⋅ ( 0.257625 ) + ------------------------------------------2! ( 0.60 ) ⋅ ( 0.60 – 1 ) ( 0.60 – 2 ) + -------------------------------------------------------------------- ⋅ ( 0.000750 ) = 1.152727 3!

(7.57)

The spreadsheet of Figure 7.3 shows the layout and computations for this example.

7−20

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

**Gregory−Newton Backward Interpolation Method
**

TABLE 7.13 Difference table for Example 7.8 1st Difference x 1.00 1.05 1.10 1.15 1.20 1.25 f (x) 1.000000 1.257625 0.273375 1.531000 0.289875 1.820875 0.307125 2.128000 0.325125 2.453125 0.018000 0.017250 0.000750 0.016500 0.000750 f ( x 0, x 1 ) 0.257625 0.015750 0.000750 2nd Difference f ( x 0, x 1, x 2 ) 3rd Difference f ( x 0, x 1, x 2, x 3 )

**7.7 Gregory−Newton Backward Interpolation Method
**

This method uses the formula

r(r + 1) 2 r(r + 1 )(r + 2 ) 3 f ( x ) = f 0 + rΔf – 1 + ----------------- Δ f – 2 + ---------------------------------- Δ f – 3 + … 2! 3!

2 3

(7.58)

where f 0 is the first value of the data set, Δf –1 , Δ f –2 , and Δ f –3 are the first, second and third backward differences, and

( x – x1 ) r = -----------------h

**Expression (7.58) is valid only when the values x 0, x 1, x 2, …, x n are equally spaced with interval
**

h . It is used to interpolate values near the end of the data set, that is, the larger values of x .

Backward interpolation is an expression to indicate that we use the differences in a backward sequence, that is, the last entries on the columns where the differences appear. Example 7.9 Use the Gregory−Newton backward interpolation formula to compute f ( 1.18 ) from the data set of Table 7.14.

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

7−21

**Chapter 7 Finite Differences and Interpolation
**

Gregory-Newton Forward Interpolation Method See expressions (7.54) and (7.55) Interpolate f(x) at x= x f(x) Δf 0.257625 1.05 1.257625 0.273375 1.10 1.531000 0.289875 1.15 1.820875 0.307125 1.20 2.128000 0.325125 1.25 2.453125 h= A10-A8= 0.05 0.6 r= (D5-A8)/C20= 0.018000 0.017250 0.000750 0.016500 0.000750 0.015750 0.000750 1.03 Δ 2f Δ 3f

1.00 1.000000

f(1.12)= B8+F20*C9+(F20*(F20-1)*D10)/FACT(2)+(F20*(F20-1)*(F20-2)*E11)/FACT(3) = 1.152727 1.00 1.000000 1.05 1.257625 1.10 1.531000 1.15 1.820875 1.20 2.128000 1.25 2.453125 3.00 2.50 2.00 1.50 1.00 0.50 0.00 1.00 1.05 1.10 1.15 1.20 1.25

Figure 7.3. Spreadsheet for Example 7.8 TABLE 7.14 Data for Example 7.9 x y = f (x) 1.00 1.000000 1.05 1.257625 1.10 1.531000 1.15 1.820875 1.20 2.128000 1.25 2.453125

7−22

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

**Gregory−Newton Backward Interpolation Method
**

Solution: We arbitrarily choose f 0 = 2.128000 as our starting point since f ( 1.18 ) lies between f ( 1.15 ) and

f ( 1.20 ) . Then, h = 1.20 – 1.15 = 0.05

and

r = ( x – x 1 ) ⁄ h = ( 1.18 – 1.20 ) ⁄ 0.05 = – 0.4

Now, by (7.58) we have:

( – 0.4 ) ( – 0.4 + 1 ) f ( 1.18 ) = 2.128 + ( – 0.4 ) ( 0.307125 ) + ----------------------------------------- ( 0.01725 ) 2! ( – 0.4 ) ( – 0.4 + 1 ) ( – 0.4 + 2 ) + -------------------------------------------------------------------- ( 0.00075 ) = 2.003032 3!

The computations were made with the spreadsheet of Figure 7.4. If the increments in x values are small, we can use the Excel VLOOKUP function to perform interpolation. The syntax of this function is as follows.

VLOOKUP(lookup_value, table_array, col_index_num, range lookup)

where:

lookup_value is the value being searched in the first column of the lookup table table_array are the columns forming a rectangular range or array col_index_num is the column where the answer will be found range lookup is a logical value (TRUE or FALSE) that specifies whether we require VLOOKUP to find an exact or an approximate match. If TRUE is omitted, an approximate match is returned.

In other words, if an exact match is not found, the next largest value that is less than the lookup_value is returned. If FALSE is specified, VLOOKUP will attempt to find an exact match, and if one is not found, the error value #N/A will be returned.

A sample spreadsheet is shown in Figure 7.5 where the values of x extend from −5 to +5 volts. Only a partial table is shown.

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

7−23

**Chapter 7 Finite Differences and Interpolation
**

Gregory-New ton Backw ard Interpolation Method See form ula 7.58 Interpolate f(x) at x= 1.18 x 1.00 1.05 1.10 1.15 1.20 f(x) 1.000000 0.257625 1.257625 0.273375 1.531000 0.289875 1.820875 0.307125 2.128000 0.05 r= (C5-A16)/C18= -0.4 0.017250 0.016500 0.000750 0.015750 0.000750 Δf Δ 2f Δ 3f

h= A16-A14=

f(1.18)= B16+F18*C15+(F18*(F18+1)*D14)/FACT(2)+(F18*(F18+1)*(F18+2)*E13)/FACT(3) = 1.00 1.05 1.10 1.15 1.20 2.003032 2.500 1.000000 2.000 1.257625 1.531000 1.820875 2.128000 1.500 1.000 0.500 0.000 1.00

1.05

1.10

1.15

1.20

Figure 7.4. Spreadsheet for Example 7.9

**7.8 Interpolation with MATLAB
**

MATLAB has several functions that perform interpolation of data. We will study the following: 1. interp1(x,y,xi) performs one dimensional interpolation where x and y are related as y = f(x) and xi is some value for which we want to find y(xi) by linear interpolation, i.e., “table lookup”. This command will search the x vector to find two consecutive entries between which the desired value falls. It then performs linear interpolation to find the corresponding value of y. To obtain a correct result, the components of the x vector must be monotonic, that is, either in ascending or descending order.

7−24

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

0316 -1.2.TRUE) = =VLOOKUP(-1. Third Edition Copyright © Orchard Publications 7−25 .700 -0.Interpolation with MATLAB V I 0.0267 -1.A2:B282.950 -0.0313 -1.000 -0.z.0326 -1.y.825 -0.725 -0.8250 -0.0323 -1.5500 4.xi.05 -2 -1 0 1 2 3 4 5 -2. this requires equidistant values of x 3. interp2(x.00 -0.1484323 Figure 7.925 -0.0277 -1.8375.10 0. 4.0309 -1.625 -0.0299 -1. The spline method does not apply to two dimensional interpolation.xi) but performs two dimensional interpolation.y.’method’) but performs two dimensional interpolation.z.0292 -1.030927 -0.0305803 -0.yi.2) = =VLOOKUP(-1.y.’method’) performs the same operation as interp1(x.1496775 =VLOOKUP(4.775 -0.A2:B282.800 -0. interp2(x.y.2) = 0. We will illustrate the applications of these functions with the examples below.0309266 -1.575 -0.875 -0.900 -0.xi.0285 -1. The default is linear.0288 -1.5.1484323 0.A2:B282.xi) where the string method allows us to specify one of the methods listed below.xi. this is the default interpolation spline − cubic spline interpolation.0320 -1. Using the Excel VLOOKUP function for interpolation 2.8500 -0.5535.0302 -1.5750 B 0. interp1(x.FALSE) = A 264 265 4. this does also extrapolation cubic − cubic interpolation.0274 -1.05 0.030927 #N/A =VLOOKUP(-1.0306 -1.xi.’method’) is similar to interp1(x.yi) is similar to interp1(x.y.525 -0.2.850 -0.750 -0.0330 -1.8375.20 0.650 -0.A2:B282.0263 A 8 9 B -1.0295 -1.15 0.8375.550 -0.975 -0.675 -0.y.0281 -1. Numerical Analysis Using MATLAB® and Excel®. nearest − nearest neighbor interpolation linear − linear interpolation.0270 -1.600 -0.

Third Edition Copyright © Orchard Publications . % Heading of the two−column matrix fprintf(' \n').0262 -0. grid. and using linear interpolation compute i when v = 1.4f \n'. voltage (volts) % for a diode whose v−i characteristics are i=0.m % It computes the values of current (in milliamps) vs.10 The i – v (current−voltage) relation of a non−linear electrical device is given by i ( t ) = 0. number_of_values) command.2759 milliamps -0. % Linear (default) interpolation fprintf('current (in milliamps) @ v=1.0297 -0.1.265 volts.2 .5172 -1.2v ( t ) – 1) (7.1(exp(0.a).xi) command to interpolate at the desired value. disp(v_a). we will use the MATLAB linspace(first_value. Then.265 is %2.7586 -1. % This script is for Example_7_10.* (exp(0.1 ( e 0. 30). % Specify 30 intervals in the −2<=v<=5 interval a=0. ma) The data and the value obtained by interpolation are shown below. % v=linspace(−2. plot i versus v in this range.1. 5. % Display values of volts and amps below the heading ma=interp1(v. ylabel('current (milliamps)').0225 7−26 Numerical Analysis Using MATLAB® and Excel®.y.59) where v is in volts and i in milliamperes. Solution: We are required to use 30 data points within the given range. % Define "v_a" as a two−column matrix to display volts % and amperes side−by−side.265). xlabel('voltage (volts)'). % We use "a" for current instead of "i" to avoid conflict % with imaginary numbers v_a=[v. the corresponding values in milliamperes. and plots the data for this range.a]'. accordingly. The script below produces 30 values in volts.2v)−1). Compute i for 30 data points of v within the interval – ( 2 ≤ v ≤ 5 ) . % We can use the MATLAB function 'interp1' to linearly interpolate % the value of milliamps for any value of v within the specified interval.a.0000 -1. title('volt−ampere characteristics for a junction diode'). we use the interp1(x.* v)−1).0330 -0. last_value.Chapter 7 Finite Differences and Interpolation Example 7. fprintf(' volts milliamps \n'). volts -2. plot(v.

0000 -0.Interpolation with MATLAB -1.0014 0.7931 -0.0256 0.4138 0.3103 -0.0939 0.0847 0.0345 -0.1035 0.0318 0.6.1034 2.7931 4.0760 0.5517 -0.1352 0.0140 0.0086 0.8276 3.1135 0.0060 -0. Numerical Analysis Using MATLAB® and Excel®.1724 0.3448 2.0345 4.1468 0.0035 0.0147 -0.0451 0.0104 -0.0196 0.0598 0.265 is 0.0383 0.1241 0.1590 0.6207 1.1379 1.2759 4. Third Edition Copyright © Orchard Publications 7−27 .0677 0.0690 3.5172 4.7586 5.0690 0.5517 3.0523 0.1718 current (in milliamps) @ v=1.3103 3.6552 0.0187 -0.0288 The plot for this example is shown in Figure 7.8621 2.8966 1.5862 2.3793 1.

2*pi. 'linear'). Then.2 volt-ampere characteristics for a junction diode 0. % The label 'linear' on the right side of the above statement % could be have been omitted since the default is linear % cubic_int=interp1(x.120). 'cubic'). cubic.y). Solution: The script below plots (7.xi.^ 5). 3π ⁄ 5 .15 current (milliamps) 0.y.Chapter 7 Finite Differences and Interpolation 0. xlabel('x'). % We need these two y=(cos(x)) . use the MATLAB interp1(x. Plot for Example 7.60) in the interval 0 ≤ x ≤ 2π with 120 inte rme diate values. and spline interpolation methods. Third Edition Copyright © Orchard Publications . and spline methods.6.’method’) function to interpolate at π ⁄ 8 . title('y=cos^5(x)').11 Plot the function y = f ( x ) = cos 5x (7. grid. % plot(x.05 -2 0 2 voltage (volts) 4 6 Figure 7. Compare the values obtained with the linear. ylabel('y'). with the analytical values. and 3π ⁄ 7 . % statements for the plot % analytic=(cos([pi/8 pi/4 3*pi/5 3*pi/7]').[pi/8 pi/4 3*pi/5 3*pi/7]'.y. for comparison with the linear.y.[pi/8 pi/4 3*pi/5 3*pi/7]'. cubic.10 Example 7.60) and produces the values of analytical values.05 0 -0. π ⁄ 4 .1 0. 7−28 Numerical Analysis Using MATLAB® and Excel®.^ 5. % linear_int=interp1(x. % This is the script for Example_7_11 % x=linspace(0.

fprintf('%8. fprintf('%8.2)=error2.*100 .y. % 4th column of matrix fprintf(' \n').5f\n'./ analytic. % 3rd column of matrix y(:. % % The statements below compute the percent error for the three % interpolation methods as compared with the exact (analytic) values % error1=(linear_int−analytic). % Insert line disp('The percent errors for each interpolation method are:') fprintf(' \n').'spline'). Third Edition Copyright © Orchard Publications 7−29 .5f\t %8./ analytic.1)=error1. % 2nd column of matrix y(:.5f\t %8./ analytic.3). % 1st column of matrix z(:. error2=(cubic_int−analytic).Interpolation with MATLAB % spline_int=interp1(x.3)=error3. % 1st column of matrix y(:. The plot for the function of this example is shown in Figure 7.2)=linear_int. % 3rd column of matrix % fprintf(' \n').3)=cubic_int.5f\n'. fprintf('Linear Int \t Cubic Int \t Spline Int \n') fprintf(' \n'). % 2nd column of matrix z(:.5f\t %8.1)=analytic.y') fprintf(' \n'). error3=(spline_int−analytic).4).*100 . % Construct a 4 x 3 matrix of zeros z(:.5f\t %8. % Insert line fprintf('Analytic \t Linear Int \t Cubic Int \t Spline Int \n') fprintf(' \n'). Numerical Analysis Using MATLAB® and Excel®.z') fprintf(' \n').*100 .4)=spline_int.5f\t %8. % y=zeros(4.% Construct a 4 x 4 matrix of zeros y(:.[pi/8 pi/4 3*pi/5 3*pi/7]'.7. % z=zeros(4.

7 in Chapter 1 whose script and plot are shown below.6 -0.00062 Cubic Int 0. Solution: % The file is Example_7_12.7.17678 -0.00281 0.00002 0. Third Edition Copyright © Orchard Publications .00296 0.67310 0.00012 -0.17718 -0.09681 13.01027 -0.Chapter 7 Finite Differences and Interpolation 1 0.00055 The percent errors for each interpolation method are: Linear Int -0.6 0.17678 -0.64706 Spline Int 0.00282 0.4 -0.00055 Linear Int 0.12 For the impedance example of Section 1. Plot the function of Example 7.05211 0.8 0. Analytic 0.67274 0.11 The analytical and interpolated values are shown below for comparison. use the spline method of interpolation to find the magnitude of the impedance at ω = 792 rad ⁄ s .17678 -0.8 -1 y y=cos 5(x) 0 1 2 3 x 4 5 6 Figure 7.00282 0. Cubic Int 0.27678 Example 7.22707 5.00011 -0.40465 -0.2 0 -0.00054 Spline Int 0.67311 0.07445 7−30 Numerical Analysis Using MATLAB® and Excel®.67310 0.00184 -0.4 0.m % It calculates and plots the impedance Z(w) versus radian frequency w.2 -0.

2)=(10+(10. z(:..611 51.088 73.2).182 40.032 187. xlabel('w in rads/sec').. Radian Frequency 1200 1000 800 |Z| in Ohms 600 400 200 0 2 10 10 w in rads/sec 3 10 4 Figure 7. Radian Frequency').056 1014. Plot for the function of Example 7.751 120. 42..*(0.^4−j. 469.339 52.792. fprintf('%2. 2400 2500 2600 2700 2800 2900 3000].0f\t %10.845].. semilogx(w.182 436.122./w). grid.^6.052 145..8.12 Numerical Analysis Using MATLAB® and Excel®..938..481.abs(z)') % w=[300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500..603 81.3f Ohms \n'.1)=w'..717 46..513 62.'spline').z.432 38.3f\n'. 103. 58.1./(10+j.437 222.111 90.. Third Edition Copyright © Orchard Publications 7−31 .z)./w)))'.830 266.*10.789 71.^5. % zi=interp1(w.*w−10. 1600 1700 1800 1900 2000 2100 2200 2300.588 67.286 44.Interpolation with MATLAB % % Use the following five statements to obtain |Z| versus radian frequency w w=300:100:3000.468 48. Magnitude of Impedance vs..353. z(:.665 140.. z=[39.. fprintf(' \n') fprintf('Magnitude of Z at w=792 rad/s is %6. title('Magnitude of Impedance vs. z=zeros(28. zi) fprintf(' \n') The plot for the function of this example is shown in Figure 7.104 97.240 54.8. ylabel('|Z| in Ohms').

13 Generate the plot of the function sin R Z = ----------R (7. and z .0) to x−y points Z=sin(R). y 0 ) = ( 0. we can form a function z = f ( x.Y. This function is the equivalent of the function y = sin x ⁄ x in two dimensions. 0 ) . R is a matrix that contains the distances from the origin to each point in the pair of [ X. y . [X. Example 7.Y]=meshgrid(x. 7−32 Numerical Analysis Using MATLAB® and Excel®. % Create X and Y matrices R=sqrt(X. Y ] matrices that form a rectangular grid of points in the x – y plane. X and Y comprise a pair of matrices representing a rectangular grid of points in the x – y plane.^ 2). Using these points. Then. For convenience.y). 2. % y must have same number of points as x [X. % Compute distances from origin (0.034 Ohms Two−dimensional plots were briefly discussed in Chapter 1. mesh(Z) − Plots the values in the matrix Z as height values above a rectangular grid. Third Edition Copyright © Orchard Publications .Y]=meshgrid(x. % eps prevents division by zero mesh(X. Here. % Define interval in increments of pi/24 y=x. is R = X +Y 2 2 (7. % Generate mesh plot for Z=sin(R)/R xlabel('x').Z).^ 2 + Y.Chapter 7 Finite Differences and Interpolation MATLAB interpolates the impedance at ω = 792 rad ⁄ s and displays the following message: Magnitude of Z at w=792 rad/s is 217.y) − Generates interpolation arrays which contain all combinations of the x and y points which we specify. Y ] matrices. 1.61) in three dimensions x . and the plot in the intervals – 2π ≤ x ≤ 2π and – 2π ≤ y ≤ 2π . ylabel('y'). and connects adjacent points to form a mesh surface. we will review the following commands which can be used for two−dimensional interpolation. y ) where z is a matrix. we write and execute the following MATLAB script. % This is the script for Example_7_13 x=−2*pi: pi/24: 2*pi. Solution: The matrix R that contains the distances from the origin to each point in the pair of [ X. zlabel('z').62) We let the origin be at ( x 0./ (R+eps).

^3−3. y 0 ) = ( 0.25 y=x. % y must have same number of points as x [X.9. Solution: We let the origin be at ( x 0.^3+Y.Y. Then. and z .14 Generate the plot of the function z = x + y – 3xy 3 3 (7.Z). we write and execute the following script. 0 ) . % Generate mesh plot xlabel('x').Interpolation with MATLAB title('Plot for the Three−dimensional sin(R) / R Function') The plot for the function of this example is shown in Figure 7.13 Example 7. % Define interval in increments of 0. −1. fprintf(' \n') Numerical Analysis Using MATLAB® and Excel®.9.Y]=meshgrid(x. Plot for Example 7. z_int=interp2(X.2.25: 10. mesh(X. and the plot in the intervals – 10 ≤ x ≤ 10 and – 10 ≤ y ≤ 10 . Figure 7. % Create X and Y matrices Z=X.63) in three dimensions x .*Y.Y.'cubic').*X.Z. % This is the script for Example_7_14 x=−10: 0. ylabel('y'). y . Third Edition Copyright © Orchard Publications 7−33 .14'). zlabel('z').y). Use the cubic method to interpolate the value of z at x = – 1 and y = 2 . title('Plot for the Function of Example 7.

12 500.13 500. Figure 7.10. Plot for Example 7.17 500.02 500.12 500.02 500.18 500.11 500.21 500.01 500.17 500.11 500.17 500. Third Edition Copyright © Orchard Publications .18 500.17 500.13 500.12 500.17 500.09 500.15 500.03 500.08 500.z_int) fprintf(' \n') The plot for the function of this example is shown in Figure 7.15 500.19 500.17 This rectangular land parcel is 175 meters wide and 275 meters deep.17 500.12 500.05 500. 500.12 500.11 500.14 500.18 500.09 500.16 500.2f \n'.13 500.12 500.15 A land surveyor measured and recorded the data below for a rectangular undeveloped land which lies approximately 500 meters above sea level.14 500.10.16 500.09 500.17 500.09 500.11 500.14 500.13 500.17 500.18 500.13 500.14 500.14 500.01 500.14 Interpolated Value of z at x = -1 and y = 2 is z = 13.08 500.14 500.02 500.12 500.01 500.13 500.09 500.13 500.14 500.05 500.14 500.15 500.08 500.10 500.12 500.09 500.15 500.11 500. 7−34 Numerical Analysis Using MATLAB® and Excel®.Chapter 7 Finite Differences and Interpolation fprintf('Interpolated Value of z at x = −1 and y = 2 is z = %4.19 500.15 500.13 500.11 500.14 500.12 500.16 500.05 500.15 500. The measurements shown above were made at points 25 meters apart.02 500.08 500.14 500.14 500.10 500.15 500.08 500.15 500.14 500.09 500.09 500.11 500.15 500.13 500.15 500.00 Example 7.11 500.12 500.09 500.16 500.

17 500.13 500.14 500.17 500.5: 275.12 500.14 500.02. % mesh(x. zlabel('Height. 500. % Returns a row vector containing the size of xi where the % first element denotes the number of rows and the second is the number % of columns.11.01 500. 500. Denoting the width as the x – axis .13 500.12 500.14.z).14 500.19 500.16 500. % This script is for Example_7_15 % x=0: 25: 175. ylabel('y−axis.01 500.02 500. b. title('Parcel map') % The pause command below stops execution of the program for 10 seconds % so that we can see the mesh plot pause(10).11 500. 500. and y = 177 m.01 500.'cubic'). disp('Interpolated z is:').16 500.12.15 500.z. m').15 500. plot the given data to form a rectangular grid. Interpolate the value of z at x = 108 m. 500.09 500. Solution: The MATLAB script and plot are shown below and explanations are provided with comment statements.09 500.18 500.14 500. % Make y−axis finer disp('size(yi)'). 500.y. the depth as the y – axis and the height as the z – axis .21 500.03 500.12 500. 500. Compute the maximum height and its location on the x – y plane.14 500.y.14 500.15 500.09 500.15.10.13 500.y).5: 175.12 500. m').Interpolation with MATLAB a. % Forms grid of all combinations of xi and yi Numerical Analysis Using MATLAB® and Excel®. size(xi) yi=0: 2. box off xlabel('x−axis.17 500.yyi]=meshgrid(xi.09 500.17 500.17]. Third Edition Copyright © Orchard Publications 7−35 . xi=0: 2.05 500.14 500.15 500.17 500.02 500.11 500.13 500.17 500. meters above sea level'). % x−axis varies across the rows of z y=0: 25: 275.11 500.177.12 500. grid off.05 500.09.13 500.16 500. c. size(xi) = 1 71 disp('size(xi)').12. size(yi) [xxi.12 500.02 500.yy]=meshgrid(x.13 500.108.13 500.15 500.18 500.10 500. % y−axis varies down the columns of z z=[500.15. 500.08 500.12 500.15.17 500. 500.15 500.14 500.11 500.19 500.14 500.18 500. 500.15 500. z_int [xx.09 500.17 500.13 500. axis([0 175 0 275 500 502]). z_int=interp2(x.09 500.13 500.16 500.08 500.yi).08 500.08 500.05 500.12 500.11 500.14 500.11. 500. 500. % Make x−axis finer % size(xi). Here.18 500.14 500.08 500.09 500.11 500.09 500.

'cubic').yy]=meshgrid(x. % Cubic interpolation − interpolates % all combinations of xxi and yyi and constructs the matrix zzi mesh(xxi. [xx. size(zzi) zzi=interp2(x.zzi). title('Map of Rectangular Land Parcel') hold off. and zeros when they are not equal % [m. box off xlabel('x−axis. Observe that size(max(zzi)) = 1 71 % and size(max(max(zzi))) = 1 1 zmax=max(max(zzi)) % Estimates the peak of the terrain % The 'find' function returns the subscripts where a relational expression % is true.n]=find(zmax==zzi) %m= 7−36 Numerical Analysis Using MATLAB® and Excel®.y). Likewise max(zzi) returns a row vector which contains the % maxima of the columns in zzi. a32=5. grid off. disp('size(yyi)'). % Grid with original data plot3(xx. size(yyi). % max(x) returns the largest element of vector x % max(A) returns a row vector which contains the maxima of the columns % in matrix A. size(yyi).xxi. Third Edition Copyright © Orchard Publications .'*k').Chapter 7 Finite Differences and Interpolation % size(xxi) = size(yyi) = size(zzi) = 111 71 disp('size(xxi)'). m').z. ylabel('y−axis. −2 5 6].y.z.yyi. 2 3 −4. disp('size(zzi)'). zlabel('Height. a31 a32 a33] or % A=[−1 0 3. size(zzi) size(xxi). % A=[a11 a12 a13.yyi. meters above sea level').j]=find(A>2) % returns %i= % % 2 % 3 % 1 % 3 % % %j= % % 2 % 2 % 3 % 3 % That is. For Example. a13=3 and a33=6 % satisfy the condition A>2 % The == operator compares two variables and returns ones when they % are equal. m'). axis([0 175 0 275 500 503]). a21 a22 a23.yy. % Plot smoothed data hold on. the elements a22=3. % [i. size(xxi).

xxi.'cubic').2000 % Row 65. j is the column index.Interpolation with MATLAB % % 65 % %n= % % 36 % % that is. zmax is located at zzi = Z(65)(36) % % the x−cordinate is found from xmax=xi(n) % xmax = % % 1. and x−axis % varies across the rows of z and y−axis varies down the columns of z Interpolated z is: z_int = 500.y. size(xi) = 1 71 % and the y−coordinate is found from ymax=yi(m) % ymax = % % 3. Third Edition Copyright © Orchard Publications 7−37 .yyi.z.1492 size(xi) ans = 1 size(yi) 71 ans = 1 size(xxi) 111 ans = 111 size(yyi) 71 ans = 111 71 zzi=interp2(x. size(yi) = 1 111 % Remember that i is the row index.7500 % Column 36. % Cubic interpolation − interpolates % all combinations of xxi and yyi and constructs the matrix zzi Numerical Analysis Using MATLAB® and Excel®.

Figure 7. Third Edition Copyright © Orchard Publications .11.15 7−38 Numerical Analysis Using MATLAB® and Excel®.11.Chapter 7 Finite Differences and Interpolation size(zzi) ans = 111 zmax = 500. The interpolated value of z at x = 108 m and y = 177 m is z = 500. The plot is shown in Figure 7.2108 m = 111 n= 9 xmax = 20 ymax = 275 71 zmax=max(max(zzi)) % Estimates the peak of the terrain These values indicate that z max = 500.192 .21 where the x and y coordinates are x = 20 and y = 275 . Plot for Example 7.

• If the values of x are equally spaced and the denominators are all the same. – 1. the second divided difference is defined as: f ( x i. 2. not necessarily consecutive values of x . 3 and 4 . in general. n = 1.– 2. and so on divided differences are defined similarly. the above relation reduces to Numerical Analysis Using MATLAB® and Excel®. x j ) = ---------------------------xi – xj where x i and x j are any two. x j ) – f ( x j. x k ) f ( x i.f k + n – 2 + … + ( – 1 ) nf k + 1 + ( – 1 ) f k 2! For k = 0 .Summary 7. these values are referred to as the differences of the function. 0. the second differences are expressed as Δ f k = Δ ( Δf k ) = Δf k + 1 – Δf k 2 and. • If the constant difference between successive values of x is h . Third Edition Copyright © Orchard Publications 7−39 . the typical value of x k is x k = x 0 + kh for k = … . 1.9 Summary • The first divided difference is defined as: f ( xi ) – f ( xj ) f ( x i. … • We can now express the first differences are usually expressed in terms of the difference operator Δ as Δf k = f k + 1 – f k • Likewise. x j. within an interval. x k ) = --------------------------------------------xi – xk and the third. 2. fourth. • Likewise. for positive integer values of n Δ fk = Δ ( Δ n n–1 fk ) = Δ n–1 fk + 1 – Δ n–1 fk • The difference operator Δ obeys the law of exponents which states that Δ ( Δ fk ) = Δ n m n m+n fk • The nth differences Δ f k are found from the relation n n–1 n n(n – 1) Δ f k = f k + n – nf k + n – 1 + ------------------.

Then. The procedure is as follows: 1. n j! j • Factorial polynomials provide an easier method of constructing a difference table. Third Edition Copyright © Orchard Publications . Then. the nth differences of a polynomial of degree n are constant. in analogy with Maclaurin power series. the factorial polynomial reduces to p n ( x ) = r 0 + xq 0 ( x ) 7−40 Numerical Analysis Using MATLAB® and Excel®. we can express that polynomial as pn ( x ) = a0 + a1 ( x ) (1) –n + a2 ( x ) (2) + … + an ( x ) (n) where Δ pn ( 0 ) a j = -----------------. We divide p n ( x ) by x to obtain a quotient q 0 ( x ) and a remainder r 0 which turns out to be the constant term a 0 . • The factorial polynomials are defined as and (x) (n) = x ( x – 1 ) ( x – 2 )… ( x – n + 1 ) 1 = --------------------------------------------------------( x – 1 ) ( x – 2 )… ( x + n ) (x) –( n ) Using the difference operator Δ with the above relations we obtain and Δ(x) Δ(x) (n) = n(x) (n – 1) –( n ) = –n ( x ) n –( n – 1 ) These are very similar to differentiation of x and x . 1. …. 2. • We can express any algebraic polynomial f n ( x ) as a factorial polynomial p n ( x ) .Chapter 7 Finite Differences and Interpolation Δf 0 = f 2 – f 1 Δ f 0 = f 2 – 2f 1 + f 0 Δ f 0 = f 3 – 3f 2 + 3f 1 – f 0 Δ f 0 = f 4 – 4f 3 + 6f 2 – 4 f 1 + f 0 4 3 2 • As with derivatives.for j = 0.

Third Edition Copyright © Orchard Publications 7−41 . q 0 ( x ) = r 1 + ( x – 1 )q 1 ( x ) and by substitution we obtain p n ( x ) = r 0 + x [ r 1 + ( x – 1 )q 1 ( x ) ] = r 0 + r 1 ( x ) (1) + x ( x – 1 )q 1 ( x ) 3. and thus q 1 ( x ) = r 2 + ( x – 2 )q 2 ( x ) and by substitution we obtain pn ( x ) = r0 + r1 ( x ) = r0 + r1 ( x ) (1) (1) + x ( x – 1 ) [ r 2 + ( x – 2 )q 2 ( x ) ] + r2 ( x ) (2) (2) + x ( x – 1 ) ( x – 2 )q 2 ( x ) (n – 1) and in general. • The definite sum of p n ( x ) for the interval a ≤ x ≤ a + ( n – 1 )h is α + ( n – 1 )h x=α ∑ p n ( x ) = p n ( α ) + p n ( α + h ) + p n ( α + 2h ) + … + p n [ α + ( n – 1 )h ] • In analogy with the fundamental theorem of integral calculus which states that ∫a f ( x ) d x b = f( b) – f(a ) Numerical Analysis Using MATLAB® and Excel®. to obtain a quotient q 1 ( x ) and a remainder r 1 which turns out to be the constant term a 1 . It is denoted as Δ p n ( x ) . We divide q 0 ( x ) by ( x – 1 ) .Summary 2. We divide q 1 ( x ) by ( x – 2 ) . to obtain a quotient q 2 ( x ) and a remainder r 2 which turns out to be the constant term a 2 . Then. and it is computed from Δ (x) –1 –1 (n) (x) = -------------------(n + 1) (n + 1) • Antidifferences are very useful in finding sums of series. pn ( x ) = r0 + r1 ( x ) (1) + r2 ( x ) + … + rn – 1 ( x ) j + rn ( x ) (n) where Δ pn ( 0 ) r j = a j = -----------------j! • The antidifference of a factorial polynomial is analogous to integration in elementary calculus.

x n are equally spaced with interval h . ….f ( x 0 ) + ----------------------------------------------------------------------. and f ( x 0. the values near the beginning of the given data set. x 2. Δ f 0 . x 2. • The Gregory−Newton Forward Interpolation method uses the formula r(r – 1) 2 r( r – 1)( r – 2) 3 f ( x ) = f 0 + rΔf 0 + ----------------. x 1. like Newton’s divided difference method. x 1. Δf 0 . or taken in consecutive order. x 1. and third forward differences respectively. x n need not be equally spaced or taken in consecutive order. x 1. x 2. x 1.Δ f 0 + … 2! 3! where f 0 is the first value of the data set. x n need not be equally spaced. The variable r is the difference between an unknown point x and a known point x 1 divided by the interval h . Third Edition Copyright © Orchard Publications . and Δ f 0 are the first. 7−42 Numerical Analysis Using MATLAB® and Excel®. second. This method has the advantage that the values x 0. …. x 2.Chapter 7 Finite Differences and Interpolation we have the fundamental theorem of sum calculus which states that α + ( n – 1 )h x=α ∑ pn ( x ) = Δ pn ( x ) –1 α + nh α • One important application of finite differences is interpolation. f ( x 0. x 1. x 1. It is used to interpolate values near the smaller values of x . hence the name forward interpolation. that is. ( x – x1 ) r = -----------------h 2 3 This formula is valid only when the values x 0. x 2 ) + ( x – x 0 ) ( x – x 1 ) ( x – x 2 ) f ( x 0. x 2. second. x 1 ) + ( x – x 0 ) ( x – x 1 ) f ( x 0. …. • Newton’s Divided Difference Interpolation Method uses the formula f ( x ) = f ( x 0 ) + ( x – x 0 ) f ( x 0.Δ f 0 + --------------------------------. • Lagrange’s Interpolation Method uses the formula ( x – x 1 ) ( x – x 2 )… ( x – x n ) ( x – x 0 ) ( x – x 2 )… ( x – x n ) f ( x ) = ----------------------------------------------------------------------. that is. has the advantage that the values x 0. x 3 ) where f ( x 0.f ( x n ) ( x n – x 0 ) ( x n – x 2 )… ( x n – x n – 1 ) and. x 1 ) . x 3 ) are the first. x 2 ) .f ( x 1 ) ( x 0 – x 1 ) ( x 0 – x 2 )… ( x 0 – x n ) ( x 1 – x 0 ) ( x 1 – x 2 )… ( x 1 – x n ) ( x – x 0 ) ( x – x 1 )… ( x – x n – 1 ) + ----------------------------------------------------------------------------. and third divided differences respectively.

Backward interpolation is an expression to indicate that we use the differences in a backward sequence. spline (cubic spline interpolation which does also extrapolation).xi. Third Edition Copyright © Orchard Publications 7−43 .y. that is.y. ….3 ) f ( x ) = f 0 + rΔf –1 + ----------------. the larger values of x . x n are equally spaced with interval h . we can use the Excel VLOOKUP function to perform interpolation.yi) which is similar to interp1(x.2 ) r(r + 1 )(r + 2. the default interpolation). x 1. linear (linear interpolation.y. second and third backward differences. that is. interp1(x. and ( x – x1 ) r = -----------------h 2 3 This formula is valid only when the values x 0.xi. x 2.z. and interp2(x.Δ f – 2 + ---------------------------------. • If the increments in x values are small. cubic (cubic interpolation which requires equidistant values of x ).y.Summary • The Gregory−Newton Backward Interpolation method uses the formula r(r + 1. • We can perform interpolation to verify our results with the MATLAB functions interp1(x.xi) but performs two dimensional interpolation. the last entries on the columns where the differences appear. Δ f –2 .Δ f – 3 + … 2! 3! where f 0 is the first value of the data set.’method’) where method allows us to specify nearest (nearest neighbor interpolation). Δf –1 .xi). and Δ f –3 are the first. It is used to interpolate values near the end of the data set. Numerical Analysis Using MATLAB® and Excel®.

2 55.416 56 7.141 52 7. b.323 2.9 x 50 7.0 3. to compute: a.112 1.3 ) b. Third Edition Copyright © Orchard Publications .95 ) x y=f(x) 1.211 53 7. Use the data of the table below and the appropriate (forward or backward) Gregory−Newton formula.054 1. f ( 1.2 1.1 1.10 Exercises 1. 50.636 1. and then construct the difference table with h = 1 .483 x 3.5 1.071 51 7.8 2.280 54 7. f ( x ) = x – 2x + 4x – x + 6 5 4 3 2.219 1. f ( 1. Use the data of the table below and Newton’s divided difference formula to compute: a.7 2.Chapter 7 Finite Differences and Interpolation 7. Express the given polynomial f ( x ) below as a factorial polynomial p ( x ) .011 7−44 Numerical Analysis Using MATLAB® and Excel®.348 55 7. calculate the leading differences.

r 1. [q0. Third Edition Copyright © Orchard Publications 7−45 .e.e. x−5 % Computation of sixth (last) quotient and remainder q0 = 1 r0 = 0 q1 = 1 r1 = 0 q2 = 1 r2 = 0 0 0 13 1 5 0 0 0 2 -1 3 3 0 0 0 0 6 -2 4 0 -1 Numerical Analysis Using MATLAB® and Excel®. we will use the MATLAB deconv(p. i. x−2 % Computation of third quotient and remainder % Coefficients of fourth divisor.r3]=deconv(q2. i.11 Solutions to End−of−Chapter Exercises 1. x−3 % Computation of fourth quotient and remainder % Coefficients of fifth divisor.r0]=deconv(px. x % Computation of first quotient and remainder % Coefficients of second divisor. i. [q3. [q4.r2]=deconv(q1. pn ( x ) = r0 + r1 ( x ) (1) + r2 ( x ) (2) + … + rn – 1 ( x ) (n – 1) + rn ( x ) (n) We can compute the remainders by long division but.d2) d3=[1 −3]. r 2. i. repeated below.d4) d5=[1 −5].d0) d1=[1 −1].e. then.e. f ( x ) = x – 2x + 4x – x + 6 5 4 3 The highest power of the given polynomial f ( x ) is 5 . The MATLAB script is as follows: px=[1 −2 4 0 −1 6].d1) d2=[1 −2]. we will use (7.q) function which divides the polynomial p by q.e. we must evaluate the remainders r 0.e. i. d0=[1 0]. i. r 4 and r 5 .28). x−4 % Computation of fifth quotient and remainder % Coefficients of sixth (last) divisor. [q5. for convenience.r4]=deconv(q3. [q2.r5]=deconv(q4. x−1 % Computation of second quotient and remainder % Coefficients of third divisor. [q1.d5) % Coefficients of given polynomial % Coefficients of first divisor. to determine p n ( x ) .d3) d4=[1 −4]. r 3.r1]=deconv(q0.Solutions to End−of−Chapter Exercises 7.

e. the factorial polynomial is pn ( x ) = 6 + 2 ( x ) (1) 4 0 17 8 + 13 ( x ) (2) + 17 ( x ) (3) + 8(x) (4) + (x) (5) We will verify that p n ( x ) above is the same polynomial as the given f n ( x ) by expansion of the factorials using (7.28)..12). (13*x*(x−1))+2*x+6)) px = x^5-2*x^4+4*x^3-x+6 We observe that this is the same algebraic polynomial as f ( x ) . Third Edition Copyright © Orchard Publications .e. 7−46 Numerical Analysis Using MATLAB® and Excel®.Chapter 7 Finite Differences and Interpolation q3 = 1 r3 = 0 q4 = 1 r4 = 0 q5 = 0 r5 = 1 Therefore. i.. syms x. (x) (n) = x ( x – 1 ) ( x – 2 )… ( x – n + 1 ) with the MATLAB collect(‘s_expr’) function. with reference to (7. i. We will now compute the leading entries for the difference table using (7..30). Δ p n ( 0 ) = j!r j and p n ( x ) above Δ p ( 0 ) = 0! ⋅ 6 = 6 Δ p ( 0 ) = 3! ⋅ 17 = 102 Δ p ( 0 ) = 6! ⋅ 0 = 0 6 3 0 j Δ p ( 0 ) = 1! ⋅ 2 = 2 4 1 Δ p ( 0 ) = 2! ⋅ 13 = 26 Δ p ( 0 ) = 5! ⋅ 1 = 120 5 2 Δ p ( 0 ) = 4! ⋅ 8 = 192 We enter these values in the appropriate spaces as shown on the table below. px=collect((x*(x−1)*(x−2)*(x−3)*(x-4)+(8*x*(x−1)*(x−2)*(x-3))+(17*x*(x−1)*(x−2))+.

416 56 7.2 should be in the interval 50 ≤ x ≤ 51 .Solutions to End−of−Chapter Exercises x p( x) 6 Δ 2 Δ 2 Δ 3 Δ 4 Δ 5 Δ 6 26 102 192 120 0 We obtain the remaining set of values by crisscross addition as shown on the table below. that is.141 52 7. the first entries on the columns where the differences appear. Third Edition Copyright © Orchard Publications 7−47 . and so on.211 53 7. We enter the given x and f ( x ) values in a difference table.280 54 7.348 55 7.483 x a. These are not divided differences and therefore. we compute the first. x 50 7. as shown below. x p( x) 6 Δ 2 Δ 2 Δ 3 Δ 4 Δ 5 Δ 6 8 28 36 156 192 578 770 1606 2376 3672 6048 26 102 128 294 422 606 1028 1038 2066 432 312 120 192 120 0 2. second. and third differences. the third from the second.071 51 7. We will use the differences in a forward sequence. This is because the value of 50. we simply subtract the second value of f ( x ) from the first. Numerical Analysis Using MATLAB® and Excel®. then.

Check with MATLAB: x =[50 51 52 53 54 55 56].20 ) ⋅ ( 0. disp(spline_interp) spline interpolation yields f(50.348 7.141 0.000 0.071 h = x 1 – x 0 = 51 – 50 = 1 x – x1 -------------------------r = ------------.001 ) = 7.20 ) ⋅ ( 0.000 -0.000 0.Chapter 7 Finite Differences and Interpolation 1st Difference x 50 51 52 53 54 55 56 f (x) 7. fprintf('\n').416 7. x 1 ) 0.'spline').483 f 0 = f ( 50 ) = 7.54).071 7.0 = 0.069 0.2) = 7.20 – 1 ) ( 0.211 7.20 1 h and with these values.070 -0.071 ) + ------------------------------------------2! ( 0.2 – 50. spline_interp=interp1(x. x 2 ) 3rd Difference f ( x 0.2]'..= 50.085 3! The spreadsheet below shows the layout and computations for Part (a).280 7. Third Edition Copyright © Orchard Publications .070 7. x 2.[50.20 – 2 ) + -------------------------------------------------------------------.348 7..001 -0.001 0.068 -0. fprintf('spline interpolation yields f(50..fx.2) = \n').2 ) = 7.068 0.071 + ( 0. fx=[7.001 -0. x 1.416 7.⋅ ( 0.067 7.001 0.141 7. x 3 ) f ( x 0.001 0.071 2nd Difference f ( x 0.0849 7−48 Numerical Analysis Using MATLAB® and Excel®.211 7.20 ) ⋅ ( 7. using (7. we obtain ( 0.483].001 0.20 – 1 ) f ( 50. x 1.280 7.

071 0.068 7. Third Edition Copyright © Orchard Publications 7−49 .0 55.001 0.085 50 51 52 53 54 55 56 7.483 r= (D5-A8)/C22= 0.348 0.002 -0.00 50 51 52 53 54 55 56 b.0 53.54) and (7.3) = 7.0 52.2 -0.30 7.000 -0.483 7.211 7.2 Δ 2f Δ 3f Δ 4f Δ 5f Δ 6f h= A10-A8= 1.Solutions to End−of−Chapter Exercises Gregory-New ton Forw ard Interpolation Method for Exercise 7. we will use the backward interpolation formula r(r + 1.001 0.071 7.141 7.0 56.280 7.070 7.2)= round(B8+F22*C9+(F20*(F20-1)*D10)/FACT(2)+(F20*(F20-1)*(F20-2)*E11)/FACT(3).068 7.20 7.001 0. Since the value of 55.348 7.003 0.141 0.001 -0.001 0.Δ f – 2 + ---------------------------------.211 0.001 -0.50 7.001 -0.10 7.3 ) f ( x ) = f 0 + rΔf – 1 + ----------------.2(a) See expressions (7.000 -0.003 Δf 50.280 0.Δ f – 3 + … 2! 3! Numerical Analysis Using MATLAB® and Excel®.416 0.2 ) r(r + 1 )(r + 2.001 0.000 0.067 7.070 7.069 7.40 7.0 f(x) 7.416 7.0 54.00 f(50.9 is very close to the last value in the given range.0 51.55) Interpolate f(x) at x= x 50.000 -0.

and ( x – x1 ) r = -----------------h 2 3 We arbitrarily choose f 0 = 7.1 ) ( – 0.1 + 1 ) f ( 55.= 1.636 – 1.9) = \n').090 1.476 3! Check with MATLAB: x =[50 51 52 53 54 55 56].440 – 2.8 – 1.280 7.070 ) + ----------------------------------------.440 -------------------------------2.= 2.2 – 1.636 1.011 a.( 0.000 1.8 2.9 – 56.= 2. Then.323 = 3. and Δ f –3 are the first.1 1.. fprintf('spline interpolation yields f(55.1 + 1 ) ( – 0. fx=[7.7 – 1. Δf –1 .054 – 1.483].58) we have: ( – 0.323 – 2.390 -------------------------------.8 The second divided differences are: 1.0 3.690 -------------------------------.1 Now.9) = 7.8 – 1.9 ) lies between f ( 55 ) and f ( 56 ) .323 2.001 ) = 7.348 7. x y=f(x) 1.416 7.= 1.fx.483 as our starting point since f ( 55.[55.2 2.090 -------------------------------. Δ f –2 .. fprintf('\n').= 2.7 2.7 7−50 Numerical Analysis Using MATLAB® and Excel®.2 3.400 1..054 = 2.4764 3.= 1.9 ) = 7.1 ) ( 0. The first divided differences are: 1.690 – 2.0 ) ⁄ 1 = – 0.2 1.211 7. by (7. disp(spline_interp) spline interpolation yields f(55.390 – 1.112 -------------------------------.219 -------------------------------.5 1.( – 0.'spline').070 1.5 – 1.5 2.011 – 2.5 2. second and third backward differences.9]'.0 – 1.800 -------------------------------1.1 2.5 – 1.690 -------------------------------1.500 2.219 1. Third Edition Copyright © Orchard Publications .054 1.0 – 1.090 – 1.7 3. spline_interp=interp1(x.1 ) ( – 0. h = 56 – 55 = 1 and r = ( x – x 1 ) ⁄ h = ( 55.7 – 1.112 1.636 -------------------------------.1 1.070 = 0.071 7.141 7.483 + ( – 0.Chapter 7 Finite Differences and Interpolation where f 0 is the first value of the data set.219 – 1.390 1.000 ) 2! ( – 0.1 + 2 ) + -------------------------------------------------------------------.

x 1. x 1.090 1.49).3 – 1.800 1.1 and for x in (7.8 ) ( 2.054 2.95 .8 2.000 1.3 ) .0 – 1.011 To find f ( 1.000 = 1.1 ) ( 1.Solutions to End−of−Chapter Exercises and the third divided differences are: 1. x 1.4 ) + ( 1. we use x = 1.95 – 1.7 ) ( 1 ) = 3.172 – 0.3 – 1.350 b.1 2. x 3 ) f ( 1.5 ) + ( 1.8 – 1.818 The spreadsheet below verifies our calculated values. x 2 ) + ( x – x 0 ) ( x – x 1 ) ( x – x 2 ) f ( x 0.2 1.000 2.2 2.3 ) = 1.95 – 1. Then.011 – 0.000 – 1.95 – 1. x 2.400 2. x 2 ) f ( x 0. x 2.5 1.390 1.3 – 1. we construct the difference table below.1 1.800 = 1.004 = 1. x 2.112 + 0.95 ) = 3.002 = 2. Numerical Analysis Using MATLAB® and Excel®.214 + 0.5 ) ( 1 ) = 1.028 – 0.44 ) + ( 1.323 3. f ( x ) = f ( x 0 ) + ( x – x 0 ) f ( x 0. 1st Divided Difference 2nd Divided Difference 3rd Divided Difference x 1.95 – 2 ) ( 1. Then.112 + ( 1.000 -------------------------------2.500 – 2.3 – 1.3 .000 -------------------------------1.400 = 1.2 ) ( 1.95 ) we start with x 0 = 2.440 1. x 3 ) 0.49).0 f (x) 1.000 2.7 1.1 ) ( 1. we use x = 1.000 2.011 + ( 1.070 f ( x 0. We start with x 0 = 1.019 – 0.690 1.5 With these values. x 2 ) + ( x – x 0 ) ( x – x 1 ) ( x – x 2 ) f ( x 0. Third Edition Copyright © Orchard Publications 7−51 .219 f ( x 0.3 – 1. x 1.000 -------------------------------1. f ( x ) = f ( x 0 ) + ( x – x 0 ) f ( x 0.8 ) ( 1.7 – 1.400 – 0. x 1. x 3 ) f ( 1.0 and for x in (7. x 1 ) + ( x – x 0 ) ( x – x 1 ) f ( x 0.95 – 2 ) ( 3.3 – 1.To find f ( 1.95 – 2 ) ( 1.1 ) ( 1. x 1 ) + ( x – x 0 ) ( x – x 1 ) f ( x 0. x 1 ) 1.2 ) ( 1.07 ) + ( 1. x 1.636 2.500 3.112 1.

2 1.000 0.5 1.323 3.7 1.090 2.3): 1.[1.00 f(x) 2.00 1.3 1.219 1.000 1.690 2.95): \n\n').2 1.00 1.95): 2.636 2..219 1.8 2.8 2.7 1.7 1.800 1.440 3.1 1. fx=[ 1.000 2nd divided 3rd divided difference difference f(x0.011].323 3.000 1. disp(spline_interp) spline interpolation value of f(1.x2.054 2...fx.323 3.500 2..Chapter 7 Finite Differences and Interpolation 1st divided difference f(x0.3): \n\n').95]'.x3) x 1.fx. fprintf('spline interpolation value of f(1.3]'.5 1.'spline').1 1.112 1. fprintf('spline interpolation value of f(1.636 2.x1. x2) f(x0.8 2... spline_interp=interp1(x.50 3.054 2. fprintf('\n').112 1.0 1.1 f(x) 1.112 1. disp(spline_interp) spline interpolation value of f(1.50 1. Third Edition Copyright © Orchard Publications .2 1. x1) 1.50 2.400 1. fprintf('\n').219 1.070 1.054 2.0 Check with MATLAB: x=[ 1.3380 spline_interp=interp1(x.636 2.8 2 1.0 1.5 x 1.5 1.011 3.'spline').0].011 2.390 1. x1.8184 7−52 Numerical Analysis Using MATLAB® and Excel®.[1.

for the data of Figure 8. For example.Chapter 8 Linear and Parabolic Regression T his chapter is an introduction to regression and procedures for finding the best curve to fit a set of data. and regression with power series approximations. This process is called regression. Parabolic line approximation In finding the best line. we normally assume that the data.1.1 Curve Fitting Curve fitting is the process of finding equations to approximate straight lines and curves that best fit given sets of data. For Figure 8.1) y = mx + b y x Figure 8.2) x Figure 8. Straight line approximation.2.2.1. We will discuss linear and parabolic regression. shown by the small circles in Figures 8. Third Edition Copyright © Orchard Publications 8−1 . we can use the equation of a straight line. 8. (8. Numerical Analysis Using MATLAB® and Excel®. We will illustrate their application with several examples. represent the independent variable x . and our task is to find the dependent variable y . we can use the equation for the quadratic or parabolic curve of the form y = ax + bx + c y 2 (8. that is.1 and 8.2.

( x n.5). we compute the coefficients m (slope) and b (y-intercept) of the straight line equation y = mx + b (8. we have observed the values y 1. 8. and assume that corresponding to the values x 1. We denote the straight line equations passing through these points as y 1 = mx 1 + b y 2 = mx 2 + b y 3 = mx 3 + b … y n = mx n + b (8. we call it a least-squares parabola. ( x 2. If it is a parabola. The best fitting straight line or curve has the property that d 1 + d 2 + … + d 3 = minimum 2 2 2 (8. y 2. Obviously. we can find more than one straight line or curve to fit a set of given data. and so on. Financial analysts in large corporations use regression to forecast future costs. etc. We derive the values of m and b . x n . and we have observed that the points ( x 1. y 3 ). but we interested in finding the most suitable. Thus. Let the distance of data point x 1 from the line be denoted as d 1 . Third Edition Copyright © Orchard Publications . the slope m and y-intercept b are the same in all equations since we have assumed that all points lie close to one straight line.Chapter 8 Linear and Parabolic Regression Regression can be linear (straight line) or curved (quadratic. …. cubic.3) is called a least squares line. the distance of data point x 2 from the same line as d 2 . a straight line that satisfies (8.3) and it is referred to as the least-squares curve. …. y 3.4) such that the sum of the squares of the errors will be minimum. x 2.2 Linear Regression We perform linear regression with the method of least squares. x 3. we will not obtain valid values for all points if we solve just two 8−2 Numerical Analysis Using MATLAB® and Excel®. Investment corporations use regression analysis to compare a portfolio’s past performance versus index figures. Now. However. y n ) approximate a straight line.) and it is not restricted to engineering applications. y 2 ). and the Census Bureau use it for population forecasting. as follows: Let x and y be two related variables. ( x 3. that will make the equation of the straight line to best fit the observed data. With this method.5) In (8. we need to determine the values of the unknowns m and b from all n equations. y n . let us suppose that we have plotted the values of y versus the corresponding values of x . …. y 1 ).

∑ squares = – 2x 1 [ y 1 – ( mx 1 + b ) ] – 2x 2 [ y 2 – ( mx 2 + b ) ] – … – 2x n [ y n – ( mx n + b ) ] = 0 (8. This difference could be positive or negative. will be minimum.8) we obtain ∑ squares will have its minimum * A linear system of independent equations that has more equations than unknowns is said to be overdetermined and no exact solution exists. Collecting like terms.8) The second derivatives of (8. Third Edition Copyright © Orchard Publications 8−3 .6) we must equate to zero its two partial derivatives with respect to m and b .8) are positive and thus value.6) ∑ squares ∂ ∂m is a function of two variables m and b . Let the sum of the squares of the errors be ∑ squares Since = [ y 1 – ( mx 1 + b ) ] + [ y 2 – ( mx 2 + b ) ] + … + [ y n – ( mx n + b ) ] 2 2 2 (8. Then.Linear Regression equations with two unknowns. Numerical Analysis Using MATLAB® and Excel®. Likewise.7) and ∂ ∂b ∑ squares = – 2 [ y 1 – ( mx 1 + b ) ] – 2 [ y 2 – ( mx 2 + b ) ] – … – 2 [ y n – ( mx n + b ) ] = 0 (8. we find the sum of the squared distances between observed points and the points on the straight line. some may cancel each other and give us an erroneous sum of the distances. depending on the position of the observed value. For this reason. * The error (difference) between the observed value y 1 .7) and (8. and simplifying (8. a system that has more unknowns than equations is said to be underdetermined and these systems have infinite solutions. and the corresponding values on the straight line. is y 1 – ( mx 1 + b ) . to minimize (8. if we were to combine positive and negative values. this method is referred to as the method of least squares. and the value at the point on the straight line. This will be achieved if we use the magnitudes (absolute values) of the distances.7) and (8. Accordingly. and the value that lies on the straight line. The straight line that we choose must be a straight line such that the distances between the observed values. On the contrary. the error between the observed value y 2 and the value that lies on the straight line is y 2 – ( mx 2 + b ) and so on.

3 80 51.1 90 54 100 56.9) simultaneously by Cramer’s rule. we use the spreadsheet of Figure 8. With Cramer’s rule. The temperature increments and the observed resistance values are shown in Table 8-1. Compute the straight line equation that best fits the observed data.11) Example 8.Chapter 8 Linear and Parabolic Regression ( Σx )m + ( Σx )b = Σxy ( Σx )m + nb = Σy 2 (8.1 .6 10 31. or with MATLAB using matrices. TABLE 8. m and b are computed from D1 m = ----Δ D2 b = ----Δ Σxy Σx Σy n 2 (8.7 R (Ω) Solution: There are 11 sets of data and thus n = 11 .1 Data for Example 8.5 70 48.6 60 45.10) where Δ = Σx Σx Σx n 2 D1 = D2 = Σx Σxy Σx Σy (8.0 30 37 40 40 50 42. the resistance R in Ω (denoted as y in the equations above) increases with an increase in temperature T in °C ( denoted as x ) . 8−4 Numerical Analysis Using MATLAB® and Excel®. Third Edition Copyright © Orchard Publications .9) where Σx = sum of the numbers x Σy = sum of the numbers y Σxy = sum of the numbers of the product xy Σx = sum of the numbers x squared n = number of data x 2 We can solve the equations of (8.0 20 34.1 In a typical resistor. or with Excel. For convenience.3 where we enter the given values and we perform the computations using spreadsheet formulas.Resistance versus Temperature T ( C) ° x y 0 27.

7 467.0 0 20 40 60 80 100 Temperature 550 = 11 m=D1/Δ = 0.0 34.1 x (0C) 0 10 20 30 40 50 60 70 80 90 100 550 Σx 2 y(Ω ) 27. we compute the sums so they can be used with (8. we enter the x (temperature) values in Column A. All work is shown on the spreadsheet of Figure 8.10) and (8.6 31.0 40.123 26559 = 467.3. Spreadsheet for Example 8.0 30. Third Edition Copyright © Orchard Publications 8−5 . and y (the measured resistance corresponding to each temperature value) in Column B.0 20.3.0 42.12) We can use Excel’s Add Trendline feature to produce quick answers to regression problems.8 Σx x2 0 100 400 900 1600 2500 3600 4900 6400 8100 10000 38500 xy 0 310 680 1110 1600 2130 2730 3381 4088 4860 5670 26559 38500 = 550 26559 = 467.288x + 28.0 56.Linear Regression Spreadsheet for Example 8. The values of m and b are shown in cells I20 and I24 respectively.5 48.123 (8.11). Numerical Analysis Using MATLAB® and Excel®. the straight line equation that best fits the given data is y = mx + b = 0. Thus.8 38500 = 550 Resistance versus Tem perature 60.0 Resistance 50.0 40.0 37.6 45.3 51.288 550 = 11 b=D2/Δ = 28. We will illustrate the procedure with the following example.8 3402850 34859 121000 Σx Σ xy Σy Σx 2 n Σx n Σ xy Σy Σx Figure 8.1 Accordingly.1 54. Columns C and D show the x 2 and xy products. Then.

6 45.6 31. We observe now that the plot area is white. and we click OK.2 To produce the plot of Figure 8. we type Resistance (Ohms). we choose Plot Area from the taskbar below the main taskbar. we click on Next>Next> Next>Finish. we type Temperature (degrees Celsius). We click on the Titles tab. we type Straight line for Example 8.4.Chapter 8 Linear and Parabolic Regression Example 8. We click on the Scale tab. and we click on the small (with the hand) box.2 Repeat Example 8. we change the minimum value from 0 to 20. We click on the Chart Wizard icon. We click on the Series 1 block inside the Chart box. we click on the small (with the hand) box. Next. we change the Major Unit to 10.1 54. We click anywhere on the y-axis to select it. we click on the Major gridlines box to deselect it.0 42. 3.0 40. Third Edition Copyright © Orchard Publications Resistance (Ohms) . On the Chart sub-types options. The displayed chart types appear on the Standard Types tab. We click anywhere on the x-axis to select it. we change the maximum value from 150 to 100. We click on the Scale tab. and on the Value Y-axis.5 48. We click on Chart on the main taskbar. To change the plot area from gray to white. and on the Chart title box. we click on the top (scatter) sub-type.7 60 50 40 30 20 0 20 40 60 80 100 Temperature (degrees Celsius) Straight line for Example 8. Solution: We first enter the given data in columns A and B as shown on the spreadsheet of Figure 8. Under the Value (Y) axis. we click anywhere on the perimeter of the Chart area. Then.0 37.1 using Excel’s Add Trendline feature.2 Figure 8. and we press the Delete key to delete it.0 56. 8−6 Numerical Analysis Using MATLAB® and Excel®.2. on the Patterns tab we click on the white box (below the selected gray box).3 51. we perform the following steps: 1. and we click on OK. 2. x (0C) 0 10 20 30 40 50 60 70 80 90 100 y(Ω ) 27. We click on XY (Scatter) Type. and on the Gridlines tab.4. and observe six square handles (small black squares) around it. Plot of the straight line for Example 8.0 34.4. on the Value X-axis. and we observe that the plot appears next to the data. and we click on OK. and we click on the small (with the hand) box.

we can stretch or shrink the width of the plot to the left or to the right.13) where the coefficients a. Numerical Analysis Using MATLAB® and Excel®. we click on Add Trendline. Similarly. We now observe that the points on the plot have been connected by a straight line. Example 8. To make the plot more presentable. we click anywhere on the perimeter of the Chart area. and move it downwards with the mouse.3 Find the least−squares parabola for the data shown in Table 8. We can also stretch (or shrink) the height of the Chart area by placing the cursor near the center handle of the lower side of the graph. It provides a wealth of information for statisticians.4. and we click on Chart above the main taskbar. This feature will be illustrated in Example 8. Third Edition Copyright © Orchard Publications 8−7 . We can also use Excel to compute and display the equation of the straight line. we move it upwards by moving the mouse in that direction. and we click on OK. and when the two-directional arrow appears. and contains several terms used in probability and statistics.3 Parabolic Regression We find the least-squares parabola that fits a set of sample points with y = ax + b + c 2 (8. We place the cursor near the center handle of the upper side of the graph. by placing the cursor near the center handle of the left or right side of the Chart area. On the pull-down menu. 5. The Data Analysis Toolpack in Excel includes the Regression Analysis tool which performs linear regression using the least squares method. and we observe the six handles around it.2. 8. b. On the Type tab. We click anywhere on the perimeter of the Chart area to select it.Parabolic Regression 4.14) where n = number of data points. and c are found from ( Σx )a + ( Σx )b + nc = Σy ( Σx )a + ( Σx )b + ( Σx )c = Σxy ( Σx )a + ( Σx )b + ( Σx )c = Σx y 4 3 2 2 3 2 2 (8. we click on the first (Linear).

5 73.31 17 Σx= Σy= Σx2= Σx3= Σx4= Σxy= Σx2y= 18 79.49a + 4004.1 3. we compute the values shown in Columns C through G.15b + 79.6 6.2 1.98 3336.2 Data for Example 8.7 7.1 6.50a + 530.35 21.48 4 1.16 282.8 3.49 185.08 38.6 5.40 6.98 8 4.37 (8.1c = 437. b.14).44 18.96 636.93 11 6.6 9.29 6 3.15c = 2698.01 117.91 2541.6 4.48 37.73 2. as shown in Figure 8.3 7.14 1677.Chapter 8 Linear and Parabolic Regression TABLE 8.2 3.37 Figure 8.5 5.4 84.4 Solution: We construct the spreadsheet of Figure 8.2 4.79 92.15a + 79.9 7.70 17. The procedure was presented in Chapter 4.4 4.25 3.51 341.3 x y x y 1.6 4.76 438.65 576.7 5.38 5. and c .0 9.2 32.15) with matrix inversion and multiplication.65 11.9 7.19 1055. A B C 2 D 3 E 4 F G 2 x x x xy 1 x y xy 2 1.2 4.49 437.4 28.04 233.69 7163.72 44.50 10.79 5 2.72 2698.1 7.8 2.64 778. 530.93 31.96 262.1 87.68 26.1 6.83 10.04 941.48 3 1.7 7.24 182.1 6.8 4004.06 5470.15) We solve the equations of (8.24 5.3 7.5 5.8 7.8 2. Spreadsheet for Example 8. Third Edition Copyright © Orchard Publications .5.76 17.6.5 4.27 7 4.82 15 9.46 259.7 7.76 294.7 96.22 38. The sum values are shown in Row 18.06 7.2 3.09 148.1 7. and from these we form the coefficients of the unknown a. and from the data of Columns A and B.4 6.88 31.48 9 5.7 6.39 134.17 46.58 14 8.50b + 530.15 4004.9 1.72 32331.5.60 41.88 789.05 39.5 5.42 45.28 287.22 207.61 29.4 6.3 7.6 2.8 5.9 40.71 13 7.49 79.1 57.1 2.70 67.6 50.5 1.1b + 15c = 87.86 332.78 16 9.58 45.44 1.50 32331.6 24.8 530.62 12 7.3 18.6 5.3 By substitution into (8.8 5.6 6.3 9.3 7.41 357. 8−8 Numerical Analysis Using MATLAB® and Excel®.70 332.0 8.19 9223.87 10 5.07 5.

78 Figure 8.1 0.204 4. Solution: We enter the given data on the spreadsheet of Figure 8.7.344 3.15 79.72 4 2 Σx y= 2698. for brevity. x 0.3 0. Third Edition Copyright © Orchard Publications y 8−9 .385 0.4 The voltages (volts) shown on Table 8.20 -1 A = 8 -0.160 3.49 4004.520 4.78 2 The plot for this parabola is shown in Figure 8.3 1.181 -0.972 3.0 1.50 530. Use Excel’s Add Trendline feature to derive a polynomial that best approximates the given data.3 were applied across the terminal of a non−linear device and the current ma (milliamps) values were observed and recorded.8 where.032 -0.78 8 7 6 5 4 3 2 1 0 0 1 2 3 4 5 x 6 7 8 9 10 2 Figure 8.964 5.94 9 0.5 0.20x + 1.00 87.15 5 6 7 0.524 3.20x +1.7 0.6.0 0.9 1.80 3 Σxy= A= 4004.040 4.5 y 2.15) Therefore. only a partial list of Numerical Analysis Using MATLAB® and Excel®.3 Example 8. Parabola for Example 8.2 1.7.15 79.2 0.700 3.820 4.10 437.780 2.240 y = −0.37 32331.1 1.104 5.94x + 2.979 -0.4 1.Parabolic Regression A B C D E F G 1 Matrix Inversion and Matrix Multiplication for Example 8.016 0.3 2 Σy= 530.872 4.10 15.50 530.672 4.385 0.002 a= -0.032 c= 2.016 b= 1.94x+2. the least−squares parabola is y = – 0.8 0.6 0.364 4.4 0. Spreadsheet for the solution of the equations of (8.

0177 R2 = 0.42 1.18 0.75 0.23 4.01 3.75 1.50 0. it is discussed in probability and statistics textbooks.0. and on OK. we clicked on Display equation on chart.05 3.72 1.25 0.14 0. However.4 Experimental Data Volts ma Volts ma 0. we clicked on Display R squared value on chart.08 0.85 1.25 0.1275x .9997 1 2 3 4 5 Figure 8. TABLE 8. Here.8.91 2.60 0.75 0.01 0.* 2 * It is also discussed in Mathematics for Business. This is the Pearson correlation coefficient R .0.00 0.0403x 2 + 0.75 0. On the Options tab.25 2.75 1.03 0.Chapter 8 Linear and Parabolic Regression the given data is shown.50 0.08 3.25 1.0182x 3 .75 1.11 0.75 0. we need to enter all data in Columns A and B.2. 8−10 Numerical Analysis Using MATLAB® and Excel®.34 0.00 0.11 4.28 5.39 2.00 2.3 Data for Example 8.50 0. but we clicked on the polynomial order 3 on the Add trendline Type tab.50 1.00 1.00 0.34 Volts Amps 2.03 3. Science. The quantity R is a measure of the goodness of fit for a straight line or.50 0. as in this example.25 0. to obtain the plot.75 2.00 1.25 1. Third Edition Copyright © Orchard Publications .00 1.00 1.23 0.00 0.00 0.25 0.14 4. Plot for the data of Example 8.25 1.50 2.18 4. and Technology.63 2.18 1.00 2.00 0.00 0.50 0.4 Following the steps of Example 8. we create the plot shown next to the data.25 0.00 0 y = 0.05 0.50 0.25 0. for parabolic regression.50 1.75 0.42 0.75 0. the smooth curve was chosen from the Add trendline feature.00 0.50 0.50 1.28 0. ISBN 0-9709511-0-8.

Parabolic Regression The correlation coefficient can vary from 0 to 1. Thus. it computes the best quadratic polynomial approximation and returns the coefficients of this polynomial.6 31 34 37 40 42. hold % Hold current plot so we can add other data p=polyfit(T. Third Edition Copyright © Orchard Publications 8−11 .6 45.y. the result of Example 8. quadratic. % Establishes desired x and y axes limits plot(T. ylabel('R'). Use n = 1 to compute the best straight line approximation. MATLAB returns the value of the polynomial at point x. Let p denote the polynomial (linear. n=1'). and so on.x) command to evaluate the best line approximation p in the 0 ≤ T ≤ 100 range in ten degrees increments. title('R (Ohms) vs T (deg Celsius.5 % T= [ 0 10 20 30 40 50 60 70 80 90 100].x) function to evaluate the polynomial. there is a nearly perfect fit between the cubic polynomial and the experimental data. % This is the script for Example 8. xlabel('T').1). that is. and returns the coefficients m and b .R.5 48. When R ≈ 0 . % Display experimental (given) points with asterisk % and smoothed data with blue line grid. and compute the percent error (difference between the given values and the polynomial values). MATLAB computes the best straight line approximation. If x is a scalar. If n = 2 . Use also the polyval(p.3 51. or higher order) approximation that is computed with the MATLAB polyfit(x. % y−axis data axis([−10 110 20 60]).y. Example 8. % x−axis data R=[27.R. Solution: The following MATLAB script will do the computations and plot the data. Suppose we want to evaluate the polynomial p at one or more points. Plot resistance R versus temperature T in the range – 10 ≤ T ≤ 110 °C . Thus. and n is the degree of the polynomial.1 54 56.n) function.n) function.'*b'). cubic. that is. We can use the polyval(p. regression is performed with the polyfit(x. When R ≈ 1 . where x and y are the coordinates of the data points.4 indicates that there is a strong relationship between the variables x and y .1 using the MATLAB’s polyfit(x. it computes the best cubic polynomial approximation. if n = 1 .y. Likewise. there is no relationship between the dependent y and independent x variables. With MATLAB. the polynomial is evaluated for all values of the vector x. if n = 3 . there is a nearly perfect relationship between these variables. If x is a row vector. % Fits a first degree polynomial (straight line since n =1) and returns % the coefficients m and b of the straight line equation y = mx + b 2 2 Numerical Analysis Using MATLAB® and Excel®.n) command.7]. linear regression.5 Repeat Example 8.

*100.a). Third Edition Copyright © Orchard Publications .9.q) The plot for the data of this example is shown in Figure 8. plot(a. disp(p). Plot for Example 8. R_smoothed=polyval(p.9.2881 28.T) % Compute and display the values of the fitted % polynomial at same points as given % (experimental) values of R R_exper = R % Display the experimental values of R for comparison % The statement below computes the percent error between % the fitted polynomial and the experimental data disp('% Error at points of given values is:') % The percent error is computed with the following statement error=(R_smoothed−R_exper). 60 R (Ohms) vs T (deg Celsius.5 MATLAB also displays the following data: Coefficients m and b are: 0. n=1 55 50 45 R 40 35 30 25 0 10 20 30 40 50 T 60 70 80 90 100 Figure 8.Chapter 8 Linear and Parabolic Regression % Define range to plot the polynomial % Compute p for each value of a % Plot the polynomial % Display the coefficients m and b fprintf('\n') % Insert line disp('Coefficients m and b are:'). format bank % Two decimal place display will be sufficient disp('Smoothed R values evaluated from straight line are:'). fprintf('\n')./R_exper format short % Return to default format a=0: 10: 100.1227 Smoothed R values evaluated from straight line are: R_smoothed = 8−12 Numerical Analysis Using MATLAB® and Excel®. q=polyval(p.

2)=R_exper'.00 34. % 2nd column of matrix y(:.0 40.Parabolic Regression Columns 1 through 5 28.34 Columns 6 through 10 -0.3 51.70 % Error at points of given values is: error = Columns 1 through 5 1.63 0.0 56.53 45.12 31. % x−axis data R=[27. R_exper = R. % y−axis data p=polyfit(T. error=(R_smoothed−R_exper).0 42.0 34.1 54.29 Column 11 56.88 0.05 37./R_exper.0 37.60 45.4)=abs(error)'. % Insert line fprintf('Temp \t Exper R\t Smoothed R \t |Error| \n') fprintf(' \n').00 -0.4f\t %5.y') 36.4).4f\t %5.88 Columns 6 through 10 42.09 Numerical Analysis Using MATLAB® and Excel®.17 -0.65 54.10 40.41 48.6 45.R.00 51. The following MATLAB script will do that and will display the error in absolute values. % Insert line fprintf('%3.00 54.60 31.17 39.77 51.T).00 Columns 6 through 10 42. y=zeros(11. % Construct an 11 x 4 matrix of zeros y(:.4f\n'. R_smoothed=polyval(p.3)=R_smoothed'.20 -0. T= [0 10 20 30 40 50 60 70 80 90 100].7].00 33. % 1st column of matrix y(:. Third Edition Copyright © Orchard Publications 8−13 .41 We can make the displayed data more presentable by displaying the values in four columns.02 Column 11 0.01 -0.*100.1)=T'.1).6 31. % 3rd column of matrix y(:.0f\t %5.89 0.93 R_exper = Columns 1 through 5 27.5 48.50 48.30 Column 11 56. % 4th column of matrix fprintf(' \n').14 -0.

Chapter 8 Linear and Parabolic Regression fprintf(' \n').4.23 4.50 0.75 0.18 1.0000 34.11 4.25 0.50 0.72 1.6 Experimental Data Volts ma Volts ma 0. TABLE 8.05 3.3000 51.50 0.60 0. Temp 0 10 20 30 40 50 60 70 80 90 100 Exper R Smoothed R |Error| 27.00 1.50 0.25 0.00 0.4 Regression with Power Series Approximations In cases where the observed data deviate significantly from the points of a straight line.1227 31. Third Edition Copyright © Orchard Publications . were applied across the terminal of a non−linear device.00 0. MATLAB displays the following where the error is in percent.0000 37.0000 42.75 0.7000 28.34 Solution: We begin by plotting the given data and we draw a smooth curve as shown in spreadsheet of Figure 8.8939 0.18 4.2891 51.00 0. 8−14 Numerical Analysis Using MATLAB® and Excel®.0036 33.75 1.0117 0.4 Data for Example 8.75 0.4082 48.9318 1.01 3.0226 0. Use the power series method to derive a polynomial that best approximates the given data.75 0.10.0943 0.00 1. Example 8.0509 56.25 0.25 1.63 2.42 0. % Insert line When this script is executed.4089 8.3396 0.25 0.28 5.08 3.50 0. we can draw a smooth curve and compute the coefficients of a power series by approximating the derivatives di ⁄ dv with finite differences Δi ⁄ Δv . and the current ma (milliamps) values were observed and recorded.6339 0.5273 45.00 1.6464 42.6 The voltages (volts) shown in Table 8.1000 54.50 1.6000 31.1700 54.00 0.0000 40.0000 56. The following example illustrates the procedure.91 2.8841 0.85 1.14 4.1370 0.7655 39.03 3.2018 0.8845 36.00 2.1707 0.6000 45.5000 48.39 2.

63 2.40 2.60 0.50 2.14 0.10 we read the voltmeter reading and the corresponding smoothed ma readings and enter the values in Table 8.75 0. Third Edition Copyright © Orchard Publications 8−15 .00 0.60 0.6 Using the plot of Figure 8.49 0.25 1.00 0.00 0.5.20 0.50 0.14 4.11 2.50 1.33 −0.75 5.04 3.00 2.75 2.40 1.10.18 0.34 0.50 0.20 1.75 0.40 0.01 1. 2.25 4.50 1.18 4.50 0.27 5.72 1.00 4.02 Figure 8.80 0.08 0.50 0.00 3.01 0. we compute Δ i ⁄ Δ v for i = 1.5 Data for the first derivative Smoothed Data for Computation of Δi / Δv Volts ma Volts ma 0.23 0.20 0 1 2 3 v 4 5 6 Smoothed Experimental Data From this plot.28 0.41 Next.09 3. …20 To facilitate the computations.Regression with Power Series Approximations Experim ental volts 0.25 0. In cell E4 we enter the formula =(B5-B4)/(A5-A4) and we copy it down to E5:E23.11.00 0.85 1.25 0.60 1.80 1.89 2.75 1.25 0.75 1. TABLE 8.25 2.25 3.42 0.00 ma 0.75 4. i | v =0 = i(0) = −0.50 3.01 3. we enter these values in the spreadsheet of Figure 8.50 4.25 1.39 1.11 4.00 0.75 0. Numerical Analysis Using MATLAB® and Excel®.50 0.91 ma 0.85 1.00 0.05 0.02 2.22 4.25 0.00 1.06 3.00 0.00 1.03 0.00 1.50 0.20 1.60 0.18 1.72 0.00 1.00 1.75 0. Spreadsheet for Example 8.00 -0.25 0.75 3.63 1.

11 0.00 0.40 1.49 0.25-0. all three evaluated at v = 0 and are read from the plots.85 1.04v + 0.′′ 2 i ( v ) = i ( 0 ) + i′ ( 0 ) + ---.50 3.25 2.12 are shown in Figure 8. Spreadsheet for computation of Δi ⁄ Δv in Example 8. Third Edition Copyright © Orchard Publications .25 1.01-(-0.75 3. Finally. for this example we will consider only the first three terms of the polynomial whose coefficients i.92 1.64 0.11.63))/(0.25-0.16 0. Therefore. However.24 0.50 2.04 … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … 1.20 0.06 0. the polynomial that 1.04-(0. we plot the computed values of Δi ⁄ Δv versus v and again we smooth the data as shown in the spreadsheet of Figure 8.50 1.80 0.02))/(0.13.72 0. and from these we compute Δ i ⁄ Δ v .89 C D E Computed Δi / Δv Δi1 / Δv1= (0.48 0.76 0.25 3.33 0.50 4.08 )v 2! = – 0.5 ( – 0.01 1.12 0.00)= Δi2 / Δv2= (0.25-0.41 0.50 0.02 0.52 0.14.12.6 Next.02 2 2 2 2 2 2 2 (8. The smoothed values of the plot of Figure 8.25 0. we plot Δ i ⁄ Δ v versus volts and again we smooth the data as shown in Figure 8.75 1.12 0.75 5.08 0.12 0.18 0.i ( 0 ) + … = – 0.04 0.32 0.09 0.25 4.12v + 0.16) 8−16 Numerical Analysis Using MATLAB® and Excel®. Δi ⁄ Δv and best fits the given data is Δi ⁄ Δv .27 0.00 B Smoothed ma -0.60 0.12 0.00)= 0.14 0.00 3.00 4.00 1. Following the same procedure we can find higher order derivatives.89-(1.00 2.22 0.20 1.01))/(0.75 2.44 0.02 + 0.00)= Figure 8.Chapter 8 Linear and Parabolic Regression A 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Volts 0.01 0.63 Δi20 / Δv20= (1.32 0.12v – 0.75 4.08 0.16 0.

24 0.49 0.11 0.0 0.75 1.n) function.32 0.92 1.08 0.41 0.12 0.04 Smoothed Δ i / Δ v 1.12 0.8 0.00 0.89 1.y. Numerical Analysis Using MATLAB® and Excel®. where n ≥ 2 .4 using the MATLAB polyfit(x.25 0.40 1.01 0. Solution: With MATLAB.75 3.01 0.27 0.80 0.2 0.6 0. Plot to obtain smoothed data for Δi ⁄ Δv in Example 8.00 4.25 4.7 Repeat Example 8.44 0.25 3.60 0.85 1.2 1.50 0.75 5.63 1.14 0.06 0.75 4.Regression with Power Series Approximations Com puted Volts 0. In this example we will use n = 3 as we did with Excel.6 Example 8.52 0.22 0.00 2.09 0. The MATLAB script below computes the smoothed line and produces the plot shown on Figure 8.00 Δi / Δv 0. Third Edition Copyright © Orchard Publications 8−17 .n) function.18 0.16 0.32 0.12 0.12 0.20 0.33 0.04 0.48 0.0 0 1 2 volts 3 4 5 1.50 4.76 0.00 1.00 3.02 0.25 2. (Δi / Δv) | v=0 = i'(0) = 0.25 1.y.15.75 2. higher degree polynomial regression is also performed with the polyfit(x.50 3.16 0.64 0.20 Figure 8.4 From this plot.08 0.50 1.50 2.12.12 Δi / Δv -0.72 0. Use n = 3 to compute the best cubic polynomial approximation.

5 3. grid % Indicate data points with + and straight line in red % hold % hold current plot so we can add other data disp('Polynomial coefficients in descending order are: ') % p=polyfit(v..11-0.5 4.05 0.14 0.73 0.13. % y−axis data axis([−1 6 −1 2]).44 0.Chapter 8 Linear and Parabolic Regression A 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 B C D E Computed 2 2 Δi / Δv 2 1 2 Δi 2 Smoothed Δi/Δv Volts 0.00)= 1.00 4.ma.18 1.04 -0.64 0.12 0.20 0.91].03-0.25-0.75 3.75 3.25 2.50 2.11 0.03 Δi220 / Δv220= Figure 8.00)= (0.29 0. % Establishes desired x and y axes limits plot(v.39 1.12 0.36 0.08 0.ma.00 2.75 5.00 2 2 0.24 0.03 0.42 0..25:5.25 3.35 0.'+r').44 … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … … (1.6 v=[0 0.00 0.75 1 1.92 Δi / Δv21= / 2 Δv 2= (0. 0.56 0..50 4.25 0.50 0.25-0.75 2.92)/(0.72 0.18 0.16 0.04 0.36 0.3) % Fits a third degree polynomial to % the data and returns the coefficients % of the polynomial (cubic equation for % this example since n=3) a=0:0.11 0. % Define range to plot the polynomial q=polyval(p.5 1.50 3.11 0.40 0.25 0.50 0.28 0.63 1.04 0.01 0.00 3.00 0.50 1.a).12 0.75 4 4.28..25 1.60 0.75 5].25 3.10 0.24 0.32 0.75 2 2.04 0. % x−axis data ma=[0 0.08 0.85 1.10-011)/(0.17 0.24 0.25 4. 3.25 1.12 0..20 0.34 0..25 2.10 0. Spreadsheet for computation of Δi ⁄ Δv in Example 8.00 1.32 0.00 1.75 4. Third Edition Copyright © Orchard Publications .00)= -0.23 0.14 0.5 0.12)/(0.48 0.5 2.25 4.75 1.41 0.25-0.% Calculate p at each value of a % continued on the next page 8−18 Numerical Analysis Using MATLAB® and Excel®.82 0.

04 0.48 0.08 0. % 4th column of matrix fprintf(' \n')..50 Δ i2 / Δ v2 -0.12 0.0 -0. % 1st column of matrix y(:.24 0.04 0.35 0.73 Figure 8.12 0. % 2nd column of matrix y(:.25 4.1 0 3 From this plot.56 0. title('milliamps vs volts. % y=zeros(21.92 4 5 0.16 0.25 2.2 0. ylabel('ma') % Plot the polynomial % Display actual.29 0.20 0. % Calculate the values of the fitted polynomial ma_exper = ma.4 0.64 0.v)./(ma_exper+eps).q).75 1.00 0.36 0.24 0.2)=ma_exper'.50 1.00 2.11 0.6 % plot(a..00 3.10 0.14 0.04 0.12 0.1)=v'.5 0.14.50 3.75 2.50 0.44 2 2 Smoothed Δ i / Δ v 0.75 4.24 0.3)=ma_smooth'.08 v 2 2 0.00 4. Third Edition Copyright © Orchard Publications 8−19 .75 3.04 -0.17 0.28 0.3 0.12 0..4). (Δi 21 Δv2) | v=0 =2 / i''(0) = − 0. smoothed and % error values ma_smooth=polyval(p.10 0.1 0.40 0.20 0.4)=abs(error)'.32 0.41 0.00 0. % The following statement computes the percent error between the % smoothed polynomial and the experimental (given) data error=(ma_smooth−ma_exper).32 0. Plot to obtain smoothed data of Δi ⁄ Δv in Example 8. n=3').*100. % Insert line % continued on the next page Numerical Analysis Using MATLAB® and Excel®. % 3rd column of matrix y(:. % Construct a 21 x 4 matrix of zeros y(:.25 3.25 0.36 0. xlabel('v').50 2.82 0.00 1.11 0.25 1.Regression with Power Series Approximations Com puted Volts 0.

58785 1.75 3.14441 0.22191 0.05797 1.34000 0.50 1.1275 -0.50 2.03828 0.59657 |%Error| 7955257388080461.0177 volts 0. fprintf('%3.14852 0.57123 8−20 Numerical Analysis Using MATLAB® and Excel®.08000 0.69226 3.17324 0.08775 0.40716 0.Chapter 8 Linear and Parabolic Regression 2 milliamps vs volts.00 0.93513 3.51632 2.25 1.01000 0.60000 -0.5 5 Figure 8.01197 0.5 4 4.25 Exper ma Smoothed ma 0.2f\t %7.61614 25.5 0 -0.7 fprintf('volts \t Exper ma\t Smoothed ma \t |%%Error| \n').y') fprintf(' \n').25 0.00 3.5 v 3 3.03000 0.50000 0.18000 0.16677 3. n=3 1.00000 19.00000 0.5 1 1.5 0 0. Polynomial coefficients in descending order are: p = 0.05000 0. Third Edition Copyright © Orchard Publications .11000 0.27275 0.25 2.42000 0.00 2.75 2.14000 0.5f\t %7.5 1 ma 0.75 1. Plot for Example 8.5f\t %7. MATLAB computes and displays the following data.5f\n'.15.74402 27.50 0.33393 0.78451 3.49413 0.28000 0.06298 0.17970 0.0182 -0.95052 9.5 2 2.00 1.0403 0.23000 0.01766 0. fprintf(' \n').11433 0.

19511 1.16.37399 1. y ( 2 ). y ( 16 ) c.89005 0.91000 0. y ( 8 ).54911 1. seventh. y ( 12 ).18000 1.7).00000 1.28020 0.75362 0. plot the function y = sin x ⁄ x (8. Example 8.. y ( 6 ). b..17) is shown in Figure 8.25 4. % of x and y with fifth. We added eps to avoid division by zero.50 3.85000 1.[0 16 −0. y ( 14 ). p9=polyfit(x.09538 1.39000 1. compute y ( 0 ). Third Edition Copyright © Orchard Publications 8−21 .53040 0.y. Solution: a.5 1]). y=sin(x).50 4.9)..Regression with Power Series Approximations 3.72000 0.17) in the interval 0 ≤ x ≤ 16 radians. x=0:2:16. % and ninth degree polynomials % continued on the next page Numerical Analysis Using MATLAB® and Excel®. plot y versus x for these values and use the MATLAB polyfit(x.75 4.85467 1.04436 We will conclude this chapter with one more example to illustrate the uses of the MATLAB polyfit(x.75 5. y ( 10 ).40048 1. y ( 4 ).01374 1.n) and polyval(p.63155 1.y. We use the script below to evaluate y at the specified points./(x+eps) p7=polyfit(x.n) and polyval(p.y.y.63000 1.8 Use MATLAB to a. grid. The fplot function below plots y = sin x ⁄ x . b.71618 0. title('(sinx)/x curve for x > 0') The plot for the function of (8.00 4./(x+eps)'.00 0.x) functions.x) functions to find a suitable polynomial that best fits the x and y data. fplot('sin(x).

1: 16. hold off format short e % Exponential short format disp('The coefficients of 5th order polynomial in descending order are:') p5_coef=polyfit(x.'−−'.1. % The following two statements establish coordinates for three legends % in x and y directions to indicate degree of polynomials x_ref=[2 5. '9th degree polynomial').y_ref. % The following are text legends for each curve text(5.5.'−.y_ref−0. text(5. '5th degree polynomial'). p5=polyfit(x.1892 −0.8 c.4546 −0.y. grid.p9_pol. '7th degree polynomial').'. % Fits the polynomial to the data x_span=0: 0.x.y.'−. p9_pol=polyval(p9.p7_pol. % The following are line legends for each curve plot(x_ref. x_span). Plot for Example 8.0466 0.'.3. x=[0 2 4 6 8 10 12 14 16]. x_span).1.'*'). p7_pol=polyval(p7.0. y=[1 0.Chapter 8 Linear and Parabolic Regression 1 (sinx)/x curve for x > 0 0.'−'). x_span.5) disp('The coefficients of 7th order polynomial in descending order are:') p7_coef=polyfit(x.0544 −0.9) format short % We could just type format only since it is the default 8−22 Numerical Analysis Using MATLAB® and Excel®. % Compute the polynomials for this range of x values.5 0 2 4 6 8 10 12 14 16 Figure 8. x_span).1237 −0.0447 0.3].16.x_ref.5.0180]. Third Edition Copyright © Orchard Publications .5.p5_pol. y_ref=[1.y.0708 −0.9. % Specifies values for x−axis p5_pol=polyval(p5.x_ref. x_span.4. hold on.'−'.'−−'.y. The script for finding a suitable polynomial is listed below. plot(x_span. text(5.3].y.5).7) disp('The coefficients of 9th order polynomial in descending order are:') p9_coef=polyfit(x.1.2.y_ref−0.5 0 -0.3.1.

4 0.0293e+000 The coefficients of 7th order polynomial in descending order are: p7_coef = Columns 1 through 6 2.4572e-002 -1.4 5th degree polynomial 1.1644e-003 -9. 7th.Regression with Power Series Approximations The 5th. Polynomials for Example 8.9437e-002 1.4318e-004 -4. Third Edition Copyright © Orchard Publications 8−23 .4 0 2 4 6 8 10 12 14 16 Figure 8.5825e-003 1.2092e-002 3.7650e-003 Columns 7 through 10 -8.6672e-004 -5.17. 7th.5865e-006 6.1923e-006 9.5032e-004 -9.17.2 0 -0.0000e+000 Numerical Analysis Using MATLAB® and Excel®.0444e-008 1.2 7th degree polynomial 1 9th degree polynomial 0.8340e-005 4. The coefficients of 5th order polynomial in descending order are: p5_coef = 6.2508e-001 9.8 The coefficients of the 5th.6529e-001 -1.3560e-001 Columns 7 through 8 7.6 0.6483e-006 -1.9965e-001 p9_coef = Columns 1 through 6 -1.2 -0. 1.0057e-001 0 4. and 9th order polynomials are shown below. and 9th order polynomials are shown in Figure 8.8 0.9165e-001 The coefficients of 9th order polynomial in descending order are: -4.0067e-002 -1.

Σy = sum of the numbers y Σxy = sum of the numbers of the product xy . etc. and c are found from ( Σx )a + ( Σx )b + nc = Σy ( Σx )a + ( Σx )b + ( Σx )c = Σxy ( Σx )a + ( Σx )b + ( Σx )c = Σx y 4 3 2 2 3 2 2 where n = number of data points. cubic. With this method. we compute the coefficients m (slope) and b (y-intercept) of the straight line equation y = mx + b such that the sum of the squares of the errors will be minimum.5 Summary • Curve fitting is the process of finding equations to approximate straight lines and curves that best fit given sets of data. b. The values of m and b can be found from the relations ( Σx )m + ( Σx )b = Σxy 2 2 2 2 where ( Σx )m + nb = Σy Σx = sum of the numbers x . Σx = sum of the numbers x squared n = number of data x 2 The values of m and b are computed from D1 m = ----Δ Σx Σx Σx n 2 where Δ = D2 b = ----Δ Σxy Σx Σy n Σx Σxy Σx Σy 2 2 D1 = D2 = • We find the least−squares parabola that fits a set of sample points with y = ax + b + c where the coefficients a. 8−24 Numerical Analysis Using MATLAB® and Excel®.) • The best fitting straight line or curve has the property that d 1 + d 2 + … + d 3 = minimum and it is referred to as the least−squares curve. A straight line that satisfies this property is called a least squares line. we call it a least−squares parabola. • Regression is the process of finding the dependent variable y from some data of the independent variable x . Third Edition Copyright © Orchard Publications . • We perform linear regression with the method of least squares. Regression can be linear (straight line) or curved (quadratic.Chapter 8 Linear and Parabolic Regression 8. If it is a parabola.

if n = 1 . it computes the best quadratic polynomial approximation and returns the coefficients of this polynomial. and so on. regression is performed with the polyfit(x. If n = 2 . Thus. it computes the best cubic polynomial approximation. linear regression. Third Edition Copyright © Orchard Publications 8−25 . MATLAB computes the best straight line approximation. where x and y are the coordinates of the data points. we can draw a smooth curve and compute the coefficients of a power series by approximating the derivatives dy ⁄ dx with finite differences Δy ⁄ Δx . Numerical Analysis Using MATLAB® and Excel®. and n is the degree of the polynomial.y. • In cases where the observed data deviate significantly from the points of a straight line.Summary • With MATLAB.n) command. Likewise. and returns the coefficients m and b . if n = 3 . that is.

3. Consider the system of equations below derived from some experimental data. 2x + y = – 1 x – 3y = – 4 x + 4y = 3 3x – 2y = – 6 – x + 2y = 3 x + 3y = 2 Using the relations (8. The sales during these years are shown on the next page.60 160 0. Third Edition Copyright © Orchard Publications .55 140 0. 2.70 180 0.1 to compute the straight line equation that best fits the given data. measurements yielded the following sets of values: millivolts milliamps 100 0. Repeat Exercise 2 above using Excel’s Trendline feature.n) and polyval(p. 8−26 Numerical Analysis Using MATLAB® and Excel®.80 200 0.45 120 0.Chapter 8 Linear and Parabolic Regression 8. 5.85 Use the procedure of Example 8.x) functions. In a non−linear device. 4. Repeat Exercise 2 above using the MATLAB’s polyfit(x.10) and (8. find the values of x and y that best fit this system of equations. A sales manager wishes to forecast sales for the next three years for a company that has been in business for the past 15 years.6 Exercises 1.y.11).

745 19.547.837 207.Exercises Year 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Sales $9.y.216 112. Repeat Exercise 5 above using the MATLAB polyfit(x.048.369 72.149.687 28.548 13. 6.983 176.321 148.574 130.873.642.n) and polyval(p.784 54.x) functions. Third Edition Copyright © Orchard Publications 8−27 . Numerical Analysis Using MATLAB® and Excel®.456.127 39. compute the sales for the next three years.147.456.163.456.352 204.545.453.147.547.782 89.632 206.678. choose an appropriate polynomial to smooth the given data and using the polynomial found. You may round the sales to the nearest thousand.987 Using Excel’s Trendline feature.

1 2 2 2 a b c a ab b ac bc 3 4 5 2 1 -1 4 2 1 -2 -1 6 1 -3 -4 1 -3 9 -4 12 7 1 4 3 1 4 16 3 12 8 3 -2 -6 9 -6 4 -18 12 9 -1 2 3 1 -2 4 -3 6 10 1 3 2 1 3 9 2 6 11 7 5 -3 17 -2 43 -22 47 12 Σ 13 2 Σa Σab 14 17 -2 15 Δ = = 727 2 Σab Σb 16 -2 43 x=D1/Δ= -1.039 8−28 Numerical Analysis Using MATLAB® and Excel®. x = – 1. Third Edition Copyright © Orchard Publications . We construct the spreadsheet below by entering the given values and computing the values from the formulas given.172 17 Σac Σab 18 -22 -2 19 D1 = = -852 2 Σbc Σb 20 47 43 y=D2/Δ= 1.172 and y = 1.7 Solutions to End−of−Chapter Exercises 1. A B C D E F G H I J 1 Spreadsheet for Exercise 8.039 21 2 Σa Σac 22 17 -22 23 D2 = = 755 Σab Σbc 24 -2 47 Thus.Chapter 8 Linear and Parabolic Regression 8.

We construct the spreadsheet below by entering the given values and computing the values from the given formulas.004x + 0.Solutions to End−of−Chapter Exercises 2.85 40000 3. y = mx + b = 0.00 Milliamps 0.0 142000 900 900 = 6 m=D1/Δ = 0.0 2000 171 42000 Σx Σ xy Σy 2 Σx Σx Thus.004 900 = 6 b=D2/Δ = 0.0476 Numerical Analysis Using MATLAB® and Excel®. Spreadsheet for Exercise 8.80 32400 0. Third Edition Copyright © Orchard Publications 8−29 .55 14400 0.0476 621 = 4.70 25600 0.60 19600 0.2 x (mV) y(mA) 100 120 140 160 180 200 900 x2 xy 45 66 84 112 144 170 621 Milliam ps versus Millivolts 1.60 0.40 100 0.80 0.95 142000 120 140 160 180 200 Millivolts Σx 2 Σx = n Σx = n Σ xy = Σy 142000 900 621 4.45 10000 0.

axis([100 200 0 1]).60 0.ma. plot(mv. % Compute p for each value of a plot(a.55 0. n=1'). hold % Hold current plot so we can add other data p=polyfit(mv.55 14400 0. xlabel('mv'). ma=[0. Trendline for Exercise 8.0 142000 900 900 = 6 m=D1/Δ = 0. we obtain the trendline shown below.80 0.45 0.40 100 0.'*b').3 x (mV) y(mA) 100 120 140 160 180 200 900 x2 xy 45 66 84 112 144 170 621 Milliam ps versus Millivolts 1.1).004 900 = 6 b=D2/Δ = 0.85 40000 3.2.q) % Plot the polynomial % Display the coefficients m and b fprintf('\n') % Insert line mv= [100 120 140 160 180 200].60 19600 0.00 Milliamps 0.a). ylabel('ma').80 32400 0.0476 621 = 4. 8−30 Numerical Analysis Using MATLAB® and Excel®. Third Edition Copyright © Orchard Publications . title('ma (milliamps) vs mv (millivolts.0 2000 171 42000 Σx Σ xy Σy Σx 2 Σx 4.70 0. % x-axis data % y-axis data % Establishes desired x and y axes limits % Display experimental (given) points with % asterisk and smoothed data with blue line grid. Following the procedure of Example 8.70 25600 0.45 10000 0. % Define range to plot the polynomial q=polyval(p.Chapter 8 Linear and Parabolic Regression 3.60 0. % Fits a first degree polynomial (straight line since n =1) and returns % the coefficients m and b of the straight line equation y = mx + b a=0: 10: 200.95 142000 120 140 160 180 200 Millivolts Σx 2 Σx = n Σx = n Σ xy = Σy 142000 900 621 4.ma.85].80 0.

5108 2.4405 1.5362 0.7000 0. % Display the experimental values of ma for comparison % The statement below computes the percent error between % the fitted polynomial and the experimental data % disp('% Error at points of given values is:').3 0.4). % The percent error is computed with the following statement error=(ma_smoothed-ma_exper).8000 0.1)=mv'. format bank % Two decimal place display will be sufficient ma_smoothed=polyval(p. % 1st column of matrix y(:.2 0.0f\t %5.2)=ma_exper'. disp(p). Third Edition Copyright © Orchard Publications 8−31 .4006 ma (milliamps) vs mv (millivolts.0582 2. % Insert line fprintf('mv \t Exper ma\t Smoothed ma \t |Error| percent \n') fprintf(' \n').1361 2.8500 0. % 4th column of matrix fprintf(' \n').4 0.1 0 0 20 40 60 80 100 mv 120 140 160 180 200 0.7 0.4500 0. n=1 Exper ma Smoothed ma 0.mv).y') fprintf(' \n').3)=ma_smoothed'. % 3rd column of matrix y(:. fprintf('\n').0041 mv 100 120 140 160 180 200 0.Solutions to End−of−Chapter Exercises disp('Coefficients m and b are:'). % Construct an 6 x 4 matrix of zeros y(:.4f\n'.9365 0.7805 0.5 0.6176 0. % 2nd column of matrix y(:.4f\t %5.5500 0. format short % Return to default format y=zeros(6./ma_exper.9 0.*100.8 0. % Compute the values of the fitted polynomial at % same points as given (experimental) values of ma ma_exper = ma.4548 0. % Insert line Coefficients m and b are: 0.4)=abs(error)'.6 ma 0.0476 |Error| percent 1. % Insert line fprintf('%3.6990 0.4f\t %5.8619 Numerical Analysis Using MATLAB® and Excel®.6000 0.

p=polyfit(year. ylabel('Yearly Sales').2E+06x 2 + 1E+07x . % y-axis data plot(year.2E+06 R2 = 0. 8−32 Numerical Analysis Using MATLAB® and Excel®..4). These results indicate that non-linear interpolation is. % Compute p for each value of a plot(a. % x-axis data sales=[9149548 13048745 19147687 28873127 39163784 . 15).Chapter 8 Linear and Parabolic Regression 5.q) % Plot the polynomial % Display coefficients ofpolynomial fprintf('\n') % Insert line disp('Coefficients are:'). title('Yearly Sales vs Years... disp(p).a). n=4'). unreliable. year= [1 2 3 4 5 6 7 8 9 10 11 12 13 14 15]. in most cases. 15. Following the procedure of Example 8. fprintf('\n')..'*b'). 54545369 72456782 89547216 112642574 130456321 . % Fits a first degree polynomial (n=4) and returns % the coefficients of the polynomial a=linspace(0. % Display experimental (given) points with % asterisk and smoothed data with blue line hold % Hold current plot so we can add other data grid. xlabel('Years').sales. Third Edition Copyright © Orchard Publications . y 16 = – 17797x + 436354x – 2 × 10 x + 10 x – 2 × 10 y 17 = – 17797x + 436354x – 2 × 10 x + 10 x – 2 × 10 y 18 = – 17797x + 436354x – 2 × 10 x + 10 x – 2 × 10 4 3 6 2 7 4 3 6 2 7 4 3 6 2 7 6 x = 16 6 x = 17 6 x = 18 = 266961792 = 247383965 = 206558656 6. we choose Polynomial 4 and we obtain the trendline shown below.9966 258000000 208000000 158000000 108000000 58000000 8000000 0 5 10 15 20 The sales for the next 3 years are from the equation above produced by Excel. 148678983 176453837 207547632 206147352 204456987]. We will compare these values with the results of Exercise 6. % Define range to plot the polynomial q=polyval(p.4. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 9149548 13048745 19147687 28873127 39163784 54545369 72456782 89547216 112642574 130456321 148678983 176453837 207547632 206147352 y = -17797x 4 + 436354x 3 .sales.

year).63 Numerical Analysis Using MATLAB® and Excel®. fprintf('year\t Exper sales\t Smoothed sales \t |Error| percent \n') fprintf(' \n').4).5 Yearly Sales 1 0.63 1.0436 -0.67 11.1641 -0.5 x 10 8 0. n=4 0 5 Years 10 15 year Exper sales Smoothed sales 1 2 3 4 9149548 13048745 19147687 28873127 7258461 14529217 21374599 29344934 |Error| percent 20. fprintf('%2.4)=abs(error)'.1)=year'.5 2 1.5 0 -0.0f\t %5.0f\t %9. % Display the experimental values of ma for comparison % The statement below computes the percent error between % the fitted polynomial and the experimental data % The percent error is computed with the following statement error=(sales_smoothed-sales_exper).2415 Yearly Sales vs Years.0f\t %9.*100.0018 2. sales_smoothed=polyval(p.3)=sales_smoothed'. Coefficients are: 1.2)=sales_exper'. % 2nd column of matrix y(:.35 11. % 3rd column of matrix y(:./sales_exper.y') fprintf(' \n'). % Construct an 15 x 4 matrix of zeros y(:.0e+007 * -0. % 1st column of matrix y(:.2f\n'. Third Edition Copyright © Orchard Publications 8−33 .Solutions to End−of−Chapter Exercises % Compute the values of the fitted polynomial at % same points as given (experimental) values of ma sales_exper = sales. y=zeros(15. % 4th column of matrix fprintf(' \n').2386 1.

x=17.56 5.415 × 10 4 4 5 3 6 2 7 6 and from it we find the values of y (the yearly sales) as follows: x=16.96 2.27 5.71 1. y18=−1.7923e+008 y17 = 1.13 1.7243e+007 These values vary significantly from those of Exercise 5.22 1.8 × 10 x + 4.1641*10^7*x−2.415*10^6.4462e+008 y18 = 8. we form the polynomial y = – 1. non-linear interpolation especially for polynomials of fourth degree and higher give inaccurate results. We should remember that the equations produced by both Excel and MATLAB represent the equations that best fit the experimental values.36*10^5*x^3−2. 8−34 Numerical Analysis Using MATLAB® and Excel®.1641*10^7*x−2.8*10^4*x^4+4.30 From the coefficients produced by MATLAB.386*10^6*x^2+1.63 1.8*10^4*x^4+4.415*10^6.36*10^5*x^3−2. y17. Third Edition Copyright © Orchard Publications .36 × 10 x – 2.386*10^6*x^2+1.1641 × 10 x – 2.415*10^6. For extrapolation. linear regression gives the best approximations.34 4.386 × 10 x + 1. y16=−1. y17=−1. y16.386*10^6*x^2+1.53 0. As stated above. shown on the previous page.02 3. y18 y16 = 1.36*10^5*x^3−2.Chapter 8 Linear and Parabolic Regression 5 6 7 8 9 10 11 12 13 14 15 39163784 54545369 72456782 89547216 112642574 130456321 148678983 176453837 207547632 206147352 204456987 39563426 52726163 69102111 88533118 110433913 133792104 157168183 178695519 196080363 206601848 207111986 1. x=18.8*10^4*x^4+4.1641*10^7*x−2.

y''1 h 2 + ---. x 1 = 0.y''' 0 h 1 + ---. close to x 1 . For another value x 2 > x 1 . if x 1 > a is a value close to a .y 1( 4 ) h 2 + … 1 2! 3! 4! (9.y''0 h 1 + ---.3) In general. Adams’.y''' i h i + 1 + ---. Such approximations are necessary when no exact solution can be found. then. x 3 = 0. we obtain: 1 1 1 2 3 4 y 1 = y 0 + y' 0 h 1 + ---. Letting h 1 = x – a in (9.1). to minimize the error f ( x 1 ) – y 1 we need to keep h 1 sufficiently small. 1112 3 4 y 2 = y 1 + y'1 h 2 + ---. (9. 1 1 1 2 3 4 y i + 1 = y i + y' i h i + 1 + ---.y''' h 2 + ---.5) correct to four decimal places for values x 0 = 0.4) Example 9. and Numerical Analysis Using MATLAB® and Excel®.1) Now. The Taylor.( x – a ) n! 2! (n) (9. 9. x 4 = 0.2) Obviously. we can find the approximate value y 1 of f ( x 1 ) by using the first k + 1 terms in the Taylor expansion of f ( x 1 ) about x = a . Runge−Kutta.( x – a ) + … + ----------------.Chapter 9 Solution of Differential Equations by Numerical Methods T his chapter is an introduction to several methods that can be used to obtain approximate solutions of differential equations. we repeat the procedure with h 2 = x 2 – x 1 .y'' i h i + 1 + ---. and Milne’s methods are discussed.y i( 4 ) h i + 1 + … 2! 3! 4! (9.3 .0 .5 with the initial condition y ( 0 ) = 1 .1 Taylor Series Method We recall from Chapter 6 that the Taylor series expansion about point a is n f'' ( a ) f (a) 2 y n = f ( x ) = f ( a ) + f' ( a ) ( x – a ) + ----------. x 2 = 0.4 .y 0( 4 ) h 1 + … 2! 3! 4! (9. Third Edition Copyright © Orchard Publications 9−1 .1 .2 .1 Use the Taylor series method to obtain a solution of y' = – xy x 5 = 0.

0001 -0.2 0.01 0.7) = ( – x + 3x ) ( – xy ) + ( – 3x + 3 )y = ( x – 6x + 3 )y 3 2 4 2 We use the subscript i to express them as y' i = – x i y i y'' i = ( x i – 1 )y i y''' i = ( – x i + 3x i )y i yi (4) 3 2 (9.1 0.6) for i = 0.0081 -0.000004y i (4) (9.1 .299 2.91 0.0 -1. y i + 1 = y i + 0.9401 7 0.0 .3 0. h = x 1 – x 0 = 0. Using the values of the coefficients of y i in (9. x 2 = 0.4681 9 0.00 0. and x 4 = 0.4 -0.16 0.0000 0.1y' i + 0.136 2.04 0.0640 0.0256 -0.7616 8 0. Third Edition Copyright © Orchard Publications .Chapter 9 Solution of Differential Equations by Numerical Methods Solution: For this example.8) = ( x i – 6x i + 3 )y i 4 2 where x i represents x 0 = 0.00 0.1 and by substitution into (9.4).3 -0.8).99 0.0010 0.000 3. 2.0270 0.0656 10 0. we construct the spreadsheet of Figure 9. The first through the fourth derivatives of (9.1 -0.000167y''' i + 0.873 2.4 .1 9−2 Numerical Analysis Using MATLAB® and Excel®.2 . 1.1.3 .4 Figure 9.0000 6 0.1 – 0. Spreadsheet for Example 9.0016 -0.84 1.1.09 0. x 3 = 0.0 0.96 0.592 2. 3.0080 0. A B C D E F G H 1 Differential Equation is y' = −xy 2 Numerical solution by Taylor method follows 3 2 3 4 2 3 4 2 xi xi xi -xi xi -1 -xi +3xi xi -6xi +3 xi 4 5 0.2 -0. and 4 .0 = 0.0000 0.5) are: y' = – xy y'' = – xy' – y = – x ( – xy ) – y = ( x – 1 )y y''' = ( x – 1 )y' + 2xy = ( x – 1 ) ( – xy )2xy = ( – x + 3x )y y (4) 2 2 2 3 (9.005y'' i + 0. x 1 = 0.

99 y 1 y''2 = ( x 2 – 1 ) y 2 = – 0.9) through (9.12) By substitution of (9. we obtain: Numerical Analysis Using MATLAB® and Excel®.2y 1 y' 3 = – x 3 y 3 = – 0.6).8).1y 1 y' 2 = – x 2 y 2 = – 0.12) into (9.4681y 3 3 y ( 4 ) = ( x 4 – 6x 4 + 3 ) y 4 = 2.Taylor Series Method The values in E6:E10.136y 4 4 y ( 4 ) = ( x 0 – 6x 0 + 3 ) y 0 = 3y 0 0 y ( 4 ) = ( x 1 – 6x 1 + 3 ) y 1 = 2.7616y 2 2 y ( 4 ) = ( x 3 – 6x 3 + 3 ) y 3 = 2. G6:G10.299y 1 1 y''' = ( – x 2 + 3x 2 ) y 2 = 0. and H6:H10 of the spreadsheet of Figure 9.0656y 4 4 4 2 4 2 4 2 4 2 4 2 3 3 3 3 3 2 2 2 2 2 (9. and we obtain the following relations: y' 0 = – x 0 y 0 = – 0y 0 = 0 y' 1 = – x 1 y 1 = – 0.91 y 3 y'' 4 = ( x 4 – 1 ) y 4 = – 0.4y 1 y''0 = ( x 0 – 1 ) y 0 = – y 0 y''1 = ( x 1 – 1 ) y 1 = – 0.3y 1 y' 4 = – x 4 y 4 = – 0.9401y 1 1 y ( 4 ) = ( x 2 – 6x 2 + 3 ) y 2 = 2. and using the given initial condition y 0 = 1 .84 y 1 y''' = ( – x 0 + 3x 0 ) y 0 = 0 0 y''' = ( – x 1 + 3x 1 ) y 1 = 0.10) (9.873y 3 3 y''' = ( – x 4 + 3x 4 ) y 4 = 1. F6:F10.11) (9.96 y 2 y''3 = ( x 3 – 1 ) y 3 = – 0.1. are now substituted into (9.9) (9. Third Edition Copyright © Orchard Publications 9−3 .592y 2 2 y''' = ( – x 3 + 3x 3 ) y 3 = 0.

1y'3 + 0.99501 ) = 0.000004 ( 3 ) = 1 – 0.00495 + 0.00019 + 0.00005 + 0.x 2 or y = e (9.16) (9.000167y''' + 0.005y''3 + 0.005y''1 + 0.14) (4) (9.99501 (9.0048 + 0.88249 ----The differential equation dy = – xy of this example can be solved analytically as follows: dx dy = – xdx ----y (4) (4) (9.0042 + 0.000004y 1 1 = ( 1 – 0.980194 ) = 0.005y''4 + 0. Third Edition Copyright © Orchard Publications .13) Similarly.000004y 4 4 = ( 1 – 0.005 + 0.0001 + 0.000012 = 0.923107 ) = 0.005 ( – 1 ) + 0.17) ∫ -----y dy = – x dx ∫ 1 2 ln y = – -.01 – 0.923107 y 5 = y 4 + 0.005y''2 + 0.00015 + 0.03 – 0. y 2 = y 1 + 0.00001 )y 2 = 0.1y'2 + 0.1y'1 + 0.000167y''' + 0.00001 )y 4 = 0.15) (4) (9.00001 )y 3 = 0.000004y 2 2 = ( 1 – 0.000167y''' + 0.000004y 0 0 (4) = 1 + 0.04 – 0.1 ( 0 ) + 0.Chapter 9 Solution of Differential Equations by Numerical Methods y 1 = y 0 + 0.1y'0 + 0.005y''0 + 0.955993 ) = 0.980194 y 3 = y 2 + 0.955993 y 4 = y 3 + 0.00167 ( 0 ) + 0.9656 ( 0. 1 ln 1 = – -.18) 9−4 Numerical Analysis Using MATLAB® and Excel®.00455 + 0.98511 ( 0.97531 ( 0.1y'4 + 0.00001 )y 1 = 0.000004y 3 3 = ( 1 – 0.( 0 ) + C 2 C = 0 –x ⁄ 2 1 2 ln y = – -.x + C 2 and with the initial condition y = 1 when x = 0 .02 – 0.95600 ( 0.000167y''' + 0.000167y''' + 0.

It differs from the Taylor series method in that we use values of the first derivative of f ( x. y n ) k 2 = hf ( x n + h .8825 and we observe that this value is in close agreement with the value of (9.y i h + … 2! 3! (9.20) 9. We can verify the analytical solution of Example 9.'y(0)=1'. For a Runge−Kutta method of order 2.( k 1 + k 2 ) 2 For Runge-Kutta Method of Order 2 (9. can be extended to apply to a second order differential equation y'' = f ( x. we can use order 3 or order 4. we need to apply the additional formula 1 1 (4) 3 2 y' i + 1 = y' i + y''i h + ---. y ) at several points instead of the values of successive derivatives at a single point.2 Runge−Kutta Method The Runge−Kutta method is the most widely used method of solving differential equations with numerical methods.1 with MATLAB’s dsolve(s) function using the following script: syms x y z z=dsolve('Dy=−x*y'.y'''i h + ---.125 = 0.'x') z = exp(-1/2*x^2) The procedure used in this example. the following formulas are applicable. y n + h ) 1 y n + 1 = y n + -. The applicable formulas are as follows: Numerical Analysis Using MATLAB® and Excel®. y' ) (9. y. k 1 = hf ( x n .Runge−Kutta Method For x 5 = 0.21) When higher accuracy is desired.5 (9.18) yields y = e – 0.19) In this case.17). Third Edition Copyright © Orchard Publications 9−5 .

( m 1 + 2m 2 + 2m 3 + m 4 ) 6 For Runge-Kutta Method of Order 4 (9.2 ) ] = 0.21).328 2 2 and 1 1 y 1 = y 0 + -.264 2 2 (9. we begin with x = 0 .2 + ( 1 + 0. we use (9. 3. y n + ------⎞ ⎝ 2 2⎠ m 4 = hf ( x n + h .⎞ = l 2 ⎝ 2 2⎠ m2 h m 3 = hf ⎛ x n + -.( l 1 + 4l 2 + l 3 ) 6 For Runge-Kutta Method of Order 3 m 1 = hf ( x n .2 k 2 = hf ( x n + h .328 ) = 1.2 ( 0 + 1 ) = 0.Chapter 9 Solution of Differential Equations by Numerical Methods l 1 = hf ( x n . and y = 1 . and 4 Runge−Kutta methods with h = 0. y n ) = l 1 = k 1 m1 -m 2 = hf ⎛ x n + h . y n + 2l 2 – l 1 ) 1 y n + 1 = y n + -. Solution: a. Third Edition Copyright © Orchard Publications . Then. y n + m 3 ) 1 y n + 1 = y n + -. y n + h ) = 0.( 0. y n ) = 0. k 1 = hf ( x n . Use order 2.22) (9.2 .2 + 0. For order 2. y n ) = k 1 l1 h l 2 = hf ⎛ x n + -. y n + ---⎞ ⎝ 2 2⎠ l 3 = hf ( x n + h .25) 9−6 Numerical Analysis Using MATLAB® and Excel®. Since we are given that y ( 0 ) = 1 ..2 [ 0 + 0.2 Compute the approximate value of y at x = 0.24) given the initial condition y ( 0 ) = 1 ..2 from the solution y ( x ) of the differential equation y' = x + y 2 (9.( k 1 + k 2 ) = 1 + -.23) Example 9. y n + -----.

the pair of 3rd−order formulas* are: * Third and fourth order formulas can also be used. y. y' ) (9. y n ) = l 1 = k 1 = 0.2 [ ( 0 + 0. Then.276 m 4 = hf ( x n + h .262 + 0. we use (9.. Third Edition Copyright © Orchard Publications 9−7 .29) The Runge−Kutta method can also be used for second order differential equations of the form y'' = f ( x.2 2 l1 1 1 h l 2 = hf ⎛ x n + --.366 and 1 y 1 = y 0 + -.2⎞ = 0. l 1 = hf ( x n .2 m1 h m 2 = hf ⎛ x n + -.391 2 and 1 1 y 1 = y 0 + -.( m 1 + 2m 2 + 2m 3 + m 4 ) 6 1 = 1 + -.Runge−Kutta Method b. but these will not be discussed in this text. They can be found in differential equations and advanced mathematics texts.262 2 m 3 = hf ⎛ x n + -..262 + 2 × 0. y n ) = k 1 = 0.2 0 + -----.2 [ 0 + 0. For order 3. y n + m 3 ) = 0.+ ⎛ 1 + -----------.2⎞ + ⎛ 1 + -.2 ) + ( 1 + 2 × 0. y n + --.27) c.⋅ 0. we use (9.22).⋅ 0. y n + ------⎞ = l 2 = 0.2 ⎛ 0 + -. y n + 2l 2 – l 1 ) = 0.⎞ = 0. m 1 = hf ( x n .30) For second order differential equations.( 0.262 ⎝ 2 2⎠ m2 h 0.⎞ ⎝ 2 2⎠ 2 ⎝ 2 ⎠ 2 (9.2 0.2 ) ] = 0.276 + 0.23).274 6 (9.262 – 0.2 + 4 × 0.276 ) ] = 0. For order 4.2 + ( 1 + 0.366 ) = 1.26) l 3 = hf ( x n + h .391 ) = 1.28) = 0. Then.( 0. Numerical Analysis Using MATLAB® and Excel®.( l 1 + 4l 2 + l 3 ) = 1 + -.2 + 2 × 0.273 6 6 (9. y n + ------⎞ = 0.262 ⎠ ⎝ ⎠ ⎝ ⎝ 2 2 2 2⎠ (9.

32) with the initial conditions y ( 0 ) = 1 .( l' 1 + 4l' 2 + l' 3 ) 6 For Runge-Kutta Method of Order 3 2nd Order Differential Equation (9. y 0 = 0 . y n + --. y' ( 0 ) = – 1 .31) we obtain: 9−8 Numerical Analysis Using MATLAB® and Excel®.2 . We will use h = x 1 – x 0 = 0.33) and using (9. y' 0 = – 1 and we are seeking the values of y 1 and y' 1 at x 1 = 0. y n + 2l 2 – l 1.Chapter 9 Solution of Differential Equations by Numerical Methods l 1 = hy' n l' 1 = hf ( x n . y' n + 2l' 2 – l' 1 ) 1 y n + 1 = y n + -.⎞ ⎝ 2⎠ l1 l' 1 h l' 2 = hf ⎛ x n + --. y n .2 . y' n ) l' 1 l 2 = h ⎛ y' n + ---.3 Given the 2nd order non−linear differential equation y'' – 2y = 0 3 (9.31) Example 9. compute the approximate values of y and y' at x = 0. y' n + ----⎞ ⎝ 2 2 2⎠ l 3 = h ( y' n + 2l' 2 – l' 1 ) l' 3 = hf ( x n + h. Third Edition Copyright © Orchard Publications . Use h = 0.2 ..( l 1 + 4l 2 + l 3 ) 6 1 y' n + 1 = y' n + -.2 . We rewrite the given equation as y'' = 2y = 0 ⋅ x + 2y 3 3 (9. Solution: We are given the values of x 0 = 0 .

y0). y' 0 ) = 0.36) Numerical Analysis Using MATLAB® and Excel®.2 ( 1.2726 3 3 3 3 3 (9.2 ) ] + 0 } = 0.4 Use the MATLAB ode23 function to find the analytical solution of the second order nonlinear equation y'' – 2y = 0 3 (9.31).4 l' 1 -----l 2 = h ⎛ y' 0 + ---.4 ) = – 0.32 + 0. ode23.( l' 1 + 4l' 2 + l' 3 ) = – 1 + -. y 0 + --.16 ) – 0. f.( – 0.16 ⎝ ⎝ 2⎠ 2⎠ l1 l' 1 h – 0. Example 9.tspan. a set of x values and the corresponding set of y values that represent points of the function y = f ( x ) .2726 ) = – 0. we will use the ode23 function in our subsequent discussion.88 ) ] = 0. y 0 + 2l 2 – l 1. Both have the same syntax.6935 6 6 (9. y0. we obtain: 1 1 y 1 = y 0 + -. y' 0 + 2l' 2 – l' 1 ) = 0.2916 – 0. This function produces two outputs.2 { 0 + 2 [ 1 + 2 ( – 0. uses second and third−order Runge−Kutta methods.2916 ) + 0. tspan. ode45. The first.1 ) ] = 0.2 ⎛ – 1 + 0. therefore. y 0.2 ( 0 + 2 × 1 + 0 ) = 0.2916 l 3 = h ( y' 0 + 2l' 2 – l' 1 ) = 0.⎞ = 0.35) MATLAB has two functions for computing numerical solutions of Ordinary Differential Equations (ODE).2 ( – 1 + 2 × 0.1634 ) = 0.2 [ 2 ( 1 – 0.⎞ = 0. represents the initial condition or boundary point that is needed to determine a unique solution.4 + 4 ( 0.( l 1 + 4l 2 + l 3 ) = 1 + -.2 [ 2 ( 0.2 3 l' 2 = hf ⎛ x 0 + --.Runge−Kutta Method l 1 = hy' 0 = 0.458 ) = 0. Third Edition Copyright © Orchard Publications 9−9 . defines the desired time span of the interval over which we want to evaluate the function y = f ( x ) .2 ( – 1 ) = – 0.1634 l' 3 = hf ( x 0 + h.2 0 + 2 ⎛ 1 + --------.⎞ + 0 ⎝ ⎠ ⎝ 2 2 2 2 ⎠ = 0. The third argument. is the name of the user defined MATLAB function.2 ) ] = 0.( 0.2 l' 1 = hf ( x 0. The first argument. The second.2 + 4 ( 0.34) By substitution into the last two formulas of (9. The second..4⎞ = – 0.16 ) – ( – 0. The syntax for ode23 is ode23(‘f’.2 [ 2 ( 1 – 0. in single quotation marks. uses fourth and fifth−order Runge−Kutta methods. y' 0 + ---.8328 6 6 1 1 y' 1 = y' 0 + -.

3.y(0)=1. % Use 2nd and 3rd Order Runge−Kutta % Plot numeric values with the statements below plot(x.Dy(0)=−1'.2 ) and y' ( 0. compare favorably with those that we found in Example 9. The script is shown below.. d2y=[y(2).. 9−10 Numerical Analysis Using MATLAB® and Excel®. We observe that the values at points y ( 0. function d2y=fex9_4(x. 'Ob−−') title('Numerical Solution for Differential Equation of Example 9.y]=ode23('fex9_4'. yprime (lower curve)').1). Then.2 ) . although few methods are available for special cases.4').'x') MATLAB displays the following message: Warning: Explicit solution could not be found.2 ) and y' ( 0. by first writing a user defined m−file which we denote as fex9_4. y0). plot the numerical solution using the function ode23 for the tspan interval 0 ≤ x ≤ 1 .3. This warning indicates that MATLAB could not find a closed−form solution for this non−linear differential equation. grid The plots for y and y' are shown in Figure 9.. we write and execute the script below to obtain the plots for y and y ′ .2 ) . % Output must be a column This file is saved as fex9_4. y(:. non−linear differential equations cannot be solved analytically. Next. These can be found in differential equations textbooks. % Interval over which we want to evaluate y=f(x) y0=[1.2*y(1)^3]. % Given initial conditions [x. x. Solution: If we attempt to find the analytical solution with the following MATLAB script syms x y y=dsolve('D2y=2*y^3.Chapter 9 Solution of Differential Equations by Numerical Methods with the initial conditions y ( 0 ) = 1 and y' ( 0 ) = – 1 . Compare values with those of Example 9.−1]. tspan. xlabel('x'). tspan=[0 1]. Third Edition Copyright © Orchard Publications .y). ylabel('y (upper curve). y(:. The numerical solution for this non−linear differential equation is obtained and plotted with the following script. in general.2). This is because. at points y ( 0. '+r−'.2.

compute and plot the numerical solution using the command ode23 along with points of the analytical solution.1/3 log(x) . the analytical solution of (9. Third Edition Copyright © Orchard Publications 9−11 . Plot for Example 9.37) with the given initial conditions is found with MATLAB as follows: syms x y y=dsolve('x^2*D2y−x*Dy−3*y=x^2*log(x).5 Use MATLAB to find the analytical solution of x y'' – x y' – 3y = x ln x 2 2 (9. Dy(1)=0.Runge−Kutta Method Figure 9.2/9) x and therefore. 'x') y = 1/9*(−3*x^3*log(x)−2*x^3−7)/x y=simple(y) y = (−1/3*log(x)−2/9)*x^2−7/9/x pretty(y) 2 (. Solution: The analytical solution of (9.4 Example 9.7/9 1/x Numerical Analysis Using MATLAB® and Excel®. Then.2.37) is .37) with the initial conditions y ( 1 ) = – 1 and y' ( 1 ) = 0 . to verify the accuracy of the numerical solution. y(1)=−1'.

x.. % This is the first derivative of (9. x. we create and save a user defined m−file. 'O'. yprime (line with O)'). % output must be a column The following MATLAB script computes and plots the numerical solution values for the interval 1 ≤ x ≤ 4 and compares these with the actual values obtained from the analytical solutions..5').. fex9_5. then y(1)'=y(2) % % and y(2)'=y(2)/x^2+3*y(1)/x^2+log(x) % d2y=[y(2).. y(:. tspan=[1 4].*x). '−').38) anal_yprime=((−2.*log(x)−7. ylabel('y (line with +).y)./9). Third Edition Copyright © Orchard Publications .3. % analytic solution of the 2nd order differential equation of (9. % Interval over which we want to evaluate y=f(x) y0=[−1.. y(:. % This is the.38) % Plot numeric and analytic values with the statements below plot(x.3. % Use 2nd and 3rd Order Runge−Kutta anal_y=((−1. '−'.ln x – -. y(2)/x+3*y(1)/x^2+log(x)]. % Given initial conditions [x. logx=lnx % % we let y(1) = y and y(2)=y'. y'=1st der. tspan./(9. '+'. grid The numeric and analytical solutions are shown in Figure 9.5 % x^2*y''−x*y'−3*y=x^2*log(x) where y''=2nd der.2).⎞ x 2 – ----⎝ 3 9⎠ 9x (9. Plot for Example 9./3)./9)./3)./(9. title('Numeric and Analytic Solutions of Differential Equation of Example 9. function d2y=fex9_5(x.y]=ode23('fex9_5'..0].. % Produces the derivatives of Example 9.Chapter 9 Solution of Differential Equations by Numerical Methods 1 2 7 y = ⎛ – -.1). y0). xlabel('x').^2−7. anal_y.*x+7.^2).38) Next..*log(x)−2. anal_yprime. Figure 9.*x.*x.5 9−12 Numerical Analysis Using MATLAB® and Excel®. x.

2. y ) . Obviously. the step from y n to y n + 1 is performed by a formula expressed in terms of differences of f ( x.Adams’ Method 9. Solution: The spreadsheet of Figure 9. y 4 . Numerical Analysis Using MATLAB® and Excel®. compute the approximate values of y for x = 0.39) where h = xn + 1 – xn f n = ( x n.1.4 and 0. y 3.Δ f n + -. y 2. it is necessary to have several (4 or more) approximate values of y ( x ) in addition to the given initial condition y ( 0 ) .6 correct to three decimal places using Adams’ method.Δf n + ----.4 shows the results of the computations of y 1. Third Edition Copyright © Orchard Publications 9−13 .5 by the third − order Runge−Kutta method.40) with the initial condition y ( 0 ) = 1 . find the value of y corresponding to x = 0. These values can be found by other methods such as the Taylor series or Runge−Kutta methods. and y5 using the third−order Runge−Kutta method as in Example 9.3. Example 9. 0.6 Given the differential equation y' = 2y + x (9. Adams’ method uses the formula 1 5 2 3 3 y n + 1 = y n + h f n + -.3 Adams’ Method In this method. 0.Δ f n + … 2 12 8 (9. 0. Then.2. y n ) Δf n = f n – f n – 1 Δ f n = Δf n – Δf n – 1 2 and so on. to form a table of differences.

yn+0.5619 h*f(xn+h.2859 h*f(xn+h.5*h)+2*(1+0. Third Edition Copyright © Orchard Publications .5*L(1)) h*[(0+0.5148 2.2553 h*f(xn+0.1000 0.5063 h*f(xn+0.3286 1.5*L(1))]= 0.3 0.yn+2*L(2)-L(1)) h*[(0+h)+2*(1+2*L(2)-L(1))]= 0.yn+2*L(2)-L(1)) h*[(0+h)+2*(1+2*L(2)-L(1))]= 0.5146 y(2)= y(1) +(L(1) + 4*L(2) + L(3))/6 h= x(2)-x(1)= x(0.yn+0.yn) h*(0.2000 x5= h*f(xn+0.5*L(1))]= 0.2+2*1.1000 1.4055 h*f(xn+0.4 0.2267 0.yn+2*L(2)-L(1)) h*[(0+h)+2*(1+2*L(2)-L(1))]= 0.5*h)+2*(1+0.3000 1.3229 h*f(xn+0.5*L(1))]= 0. Spreadsheet for Example 9.1)= Next value x(0) + h y(1)= From previous computation L(1)= L(2)= L(3)= h*f(xn.4124 1.5*L(1)) h*[(0+0.3+2*1.3)= Next value x(0) + 3*h y(3)= From previous computation L(1)= L(2)= L(3)= 39 40 41 42 43 44 45 46 47 48 h*f(xn.5*h)+2*(1+0.6 9−14 Numerical Analysis Using MATLAB® and Excel®.6398 2.2)= Next value x(0) + 2*h y(2)= From previous computation L(1)= L(2)= L(3)= h*f(xn.1000 0.yn+0.0 0.yn+2*L(2)-L(1)) h*[(0+h)+2*(1+2*L(2)-L(1))]= 0.5*h.4000 2.Chapter 9 Solution of Differential Equations by Numerical Methods A 1 2 3 B C D E F Differential Equation y' = x + 2y Numerical solution by Runga-Kutte method follows h= x(0)= y(0)= x0= x1= x2= x3= x4= h*f(xn.5*h)+2*(1+0.5*h)+2*(1+0.5*h.2 0.3313 0.8969 y(5)= y(2) +(L(1) + 4*L(2) + L(3))/6 Figure 9.2250 h*f(xn+h.yn+0.1000 0.4510 h*f(xn+h.5*L(1))]= 0.yn) h*(0+2*1)= 0.1000 0.5*L(1)) h*[(0+0.4)= Next value x(0) + 4*h y(4)= From previous computation L(1)= L(2)= L(3)= 49 50 51 52 53 h*f(xn.5*L(1)) h*[(0+0.yn) h*(0.2600 x(1)-x(0)= (given) Initial condition (given) 0.1+2*1.5*h.8773 0.5*h.0000 1.3+2*1.5146 0.8773)= 0.8773)= 0.2267)= 0.yn+2*L(2)-L(1)) h*[(0+h)+2*(1+2*L(2)-L(1))]= 0.3313 y(4)= y(2) +(L(1) + 4*L(2) + L(3))/6 h= x(4)-x(3)= x(0.1000 0.2000 1.5 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 L(1)= L(2)= L(3)= y(1)= y(0) +(L(1) + 4*L(2) + L(3))/6 h= x(2)-x(1)= x(0.5146)= 0.yn+0.8773 y(3)= y(2) +(L(1) + 4*L(2) + L(3))/6 h= x(3)-x(2)= x(0.4.5*h.1 0.0000 1.3602 h*f(xn+h.5*L(1)) h*[(0+0.5*L(1))]= 0.yn) h*(0.2267 0.yn) h*(0.

5146 1.5 yn 1.39) 3 51 y 6 = 2. y n + 1 ) (9.42) and 1 Y n + 1 = y n – 1 + -.1496 0.0546 1.8969 + 0.1 6. we accept Y n + 1 as the best approximation. Numerical Analysis Using MATLAB® and Excel®.43) serves as a check for the value y n + 1 = f ( x n + 1.8773 2.43) respectively.( 1.3 0.599 8 12 2 (9.42) and (9. Adams’ method can also be applied to second order differential equations of the form y'' = f ( x.41) As with the other methods. we compute the following values to be used in Adams’ formula of (9.2638 + -.2938 0.0000 0.2232 0.4 0.h [ 2f n – f n – 1 + 2f n – 2 ] 3 (9. y' ) with initial conditions y ( x 0 ) = y 0 and y'( x 0 ) = y' .5534 2.h [ f n + 1 + 4f n + f n – 1 ] 3 (9.0272 Δfn Δ 2 fn Δ 3 fn and by substitution into (7.1826 0. These are shown below.0000 1.2232 ) + -.6758 3. we must reduce the interval h .0406 ) = 3.0330 0.1 0.( 0. xn 0.8254 4. y.2267 1.( 0.2312 6.44) If y n + 1 and Y n + 1 in (9. It uses the predictor−corrector pair 4 y n + 1 = y n – 3 + -.0 0.43) The corrector formula of (9.0626 1.0406 0.0080 5. Third Edition Copyright © Orchard Publications 9−15 . If they differ significantly.2 0.1224 0. 0 9.5534 0.39).Milne’s Method Next.2292 0.4 Milne’s Method Milne’s method also requires prior knowledge of several values of y .2312 ) + ----.3313 2. do not differ considerably.8969 fn=xn+2yn 2.

1 Table for Example 9.6 for the differential equation y' = 2y + x (9. and from Example 9.599 . y' ) with initial conditions y ( x 0 ) = y 0 and y'( x 0 ) = y'0 .2 0.8969 fn=xn+2yn 3.0626 ) = 3.1 ( 2 × 6.6 .( 0.2938 + 5.5146 + -.4 0.5146 1. 9−16 Numerical Analysis Using MATLAB® and Excel®.1 ( 7.7984 and 1 Y 6 = y 4 + -. Milne’s method can also be extended to second order differential equations of the form y'' = f ( x.6 where we found the following values: TABLE 9.0. this is found from f 6 = x 6 + 2y 6 where x 6 = 0.45) with the initial condition y ( 0 ) = 1 . f 6 = x 6 + 2y 6 = 0.7 n 2 3 4 5 xn 0.7 Use Milne’s method to find the value of y corresponding to x = 0.2938 and using the predictor formula we find 4 y 6 = y 2 + -.1 ( f 6 + 4f 5 + f 4 ) 3 1 = 2. we must find the value of f 6 .6 + 2 × 3.43).47) We see from (9.0626 + 2 × 4.7984 + 4 × 6. y.3313 + -.0626 6.× 0.46) and (9.0546 ) = 3.599 3 (9. Third Edition Copyright © Orchard Publications .1 ) [ 2f 5 – f 4 + 2f 3 ] 3 4 = 1. Then.3 0.2292 4.2938 – 5.3313 2.5 yn 1.5 y 6 = 3.599 3 (9.46) Before we use the corrector formula of (9. Solution: This is the same differential equation as in Example 9.0546 5.× 0.Chapter 9 Solution of Differential Equations by Numerical Methods Example 9.47) that the predictor−corrector pair is in very close agreement.599 = 7.8773 2.

y n + h ) 1 y n + 1 = y n + -. We can use this series method to obtain approximate solutions of differential equations with the relation 1112 3 4 y i + 1 = y i + y' i h i + 1 + ---. • The Taylor series method can also be extended to apply to a second order differential equation y'' = f ( x. y n + 2l 2 – l 1 ) • For a Runge−Kutta method of order 4 we use the relations m 1 = hf ( x n . y n + ------⎞ ⎝ 2 2⎠ m1 h m 2 = hf ⎛ x n + -.y''' i h i + 1 + ---. y' ) using the relation 1 1 (4) 3 2 y' i + 1 = y' i + y''i h + ---.. y. y n + -----. it is the most widely used method of solving differential equations using numerical methods.( m 1 + 2m 2 + 2m 3 + m 4 ) 6 • The Runge−Kutta method can also be used for second order differential equations of the form Numerical Analysis Using MATLAB® and Excel®.. y ) at several points. y n + m 3 ) 1 y n + 1 = y n + -. y n ) = l 1 = k 1 m2 h m 3 = hf ⎛ x n + -. • For a Runge−Kutta method of order 3 we use the relations l 1 = hf ( x n . Third Edition Copyright © Orchard Publications 9−17 .( k 1 + k 2 ) 2 provided that h is sufficiently small such as h = 0.y'' i h i + 1 + ---.y i h + … 2! 3! • The Runge−Kutta method uses values of the first derivative of f ( x. • For a Runge−Kutta method of order 2 we use the relations k 1 = hf ( x n . y n + ---⎞ ⎝ 2 2⎠ 1 y n + 1 = y n + -.y'''i h + ---.5 Summary • The Taylor series method uses values of successive derivatives at a single point. y n ) k 2 = hf ( x n + h .y i( 4 ) h i + 1 + … 2! 3! 4! provided that h is sufficiently small such as h = 0.1 .Summary 9.1 ..( l 1 + 4l 2 + l 3 ) 6 l 3 = hf ( x n + h .⎞ = l 2 ⎝ 2 2⎠ m 4 = hf ( x n + h . y n ) = k 1 l1 h l 2 = hf ⎛ x n + -.

Both have the same syntax. The first. To use this method. 9−18 Numerical Analysis Using MATLAB® and Excel®. • Adams’ method provides the transition from y n to y n + 1 and the step is performed by a formula expressed in terms of differences of f ( x. y ) . y n + 2l 2 – l 1. in single quotation marks. represents the initial condition or boundary point that is needed to determine a unique solution. y n . the pair of 3rd−order relations are: l 1 = hy' n l' 1 l 2 = h ⎛ y' n + ---.Chapter 9 Solution of Differential Equations by Numerical Methods y'' = f ( x.( l 1 + 4l 2 + l 3 ) y' n + 1 = y' n + -. The third argument. defines the desired time span of the interval over which we want to evaluate the function y = f ( x ) . y' n ) l1 l' 1 h l' 2 = hf ⎛ x n + --. a set of x values and the corresponding set of y values that represent points of the function y = f ( x ) . These values can be found by other methods such as the Taylor series or Runge−Kutta methods. ode23.Δ f n + -. The second.( l' 1 + 4l' 2 + l' 3 ) 6 6 Third and fourth order formulas can also be used but they were not be discussed in this text. They can be found in differential equations texts. y' n + ----⎞ ⎝ 2 2 2⎠ l' 3 = hf ( x n + h. Third Edition Copyright © Orchard Publications . y n ) Δf n = f n – f n – 1 Δ f n = Δf n – Δf n – 1 2 and so on. f.tspan.Δf n + ----. uses second and third−order Runge−Kutta methods. y0. y. it is necessary to have several (4 or more) approximate values of y ( x ) in addition to the given initial condition y ( 0 ) . y' ) • For second order differential equations.⎞ ⎝ 2⎠ l 3 = h ( y' n + 2l' 2 – l' 1 ) yn + 1 l' 1 = hf ( x n . y' n + 2l' 2 – l' 1 ) 1 1 = y n + -. The first argument. y n + --.Δ f n + … 2 12 8 where h = xn + 1 – xn f n = ( x n. • MATLAB has two functions for computing numerical solutions of Ordinary Differential Equations (ODE). tspan.y0). ode45.. is the name of the user defined MATLAB function. • The syntax for ode23 is ode23(‘f’. This method uses the formula 1 5 2 3 3 y n + 1 = y n + h f n + -. uses fourth and fifth−order Runge−Kutta methods. This function produces two outputs. The second.

It can be found in differential equations texts. Numerical Analysis Using MATLAB® and Excel®. y' ) with initial conditions y ( x 0 ) = y 0 and y' ( x 0 ) = y'0 . we accept Y n + 1 as the best approximation. The procedure for this method was not discussed. we must reduce the interval h . It uses the predictor−corrector pair 4 y n + 1 = y n – 3 + -. Third Edition Copyright © Orchard Publications 9−19 . y. y n + 1 ) If y n + 1 and Y n + 1 do not differ considerably. Milne’s method can also be extended to second order differential equations of the form y'' = f ( x. If they differ significantly.Summary • Milne’s method also requires prior knowledge of several values of y .h [ f n + 1 + 4f n + f n – 1 ] 3 where y n + 1 is the predictor formula and Y n + 1 is the corrector formula. The corrector formula serves as a check for the value y n + 1 = f ( x n + 1.h [ 2f n – f n – 1 + 2f n – 2 ] 3 and 1 Y n + 1 = y n – 1 + -.

Then. 6.2 correct to four decimal places using the third−order Runge− Kutta method.Chapter 9 Solution of Differential Equations by Numerical Methods 9.5 . 2. Given the differential equation y' = x – y 2 with the initial condition y ( 0 ) = 1 and x 0 = 0. Use MATLAB to find the analytical solution of y' = f ( x ) = 3x 2 with the initial condition y ( 2 ) = 0. 5. 9−20 Numerical Analysis Using MATLAB® and Excel®.707 using the command ode23 for the interval 0 ≤ x ≤ 10 . and x 0 = 0. Use the MATLAB ode23 function to verify the analytical solution of Example 9. 4.6 Exercises 1. 3.2 given that y ( 0 ) = 1 .1 and x 0 + 0.0 correct to four decimal places. Construct a spreadsheet for the numerical solutions of Example 9. It is suggested that a spreadsheet is used to do all computations. Use MATLAB to plot the numerical solution of the non−linear differential equation y' = – y + 0.1. Third Edition Copyright © Orchard Publications .0 find the values of y corresponding to the values of x 0 + 0.2. y' ( 0 ) = – 1 . compute and plot the numerical solution using the MATLAB function ode23 along with points of the analytical solution to verify the accuracy of the numerical solution for the interval 2 ≤ x ≤ 4 . using the third−order Runge−Kutta method. Given the differential equation y'' + y' = xy compute the approximate values of y and y' at x 0 + 0.1 and x 0 + 0.2 sin x 3 with the initial condition y ( 0 ) = 0. It is suggested that a spreadsheet is used to do all computations.

Numerical Analysis Using MATLAB® and Excel®. xlabel('x').−1]. y0)..7 Solutions to End−of−Chapter Exercises 1..1'). We write and save the following function file: function dy = func_exer9_1(x. % Given initial conditions [x. 'Ob−−') title('Numeric Solution of Differential Equation of Exercise 9. Third Edition Copyright © Orchard Publications 9−21 . % Use 2nd and 3rd Order Runge−Kutta % Plot numeric values with the statements below plot(x. we write and execute the MATLAB script below. Next. yprime (lower curve)').Solutions to End−of−Chapter Exercises 9. y(:.1). tspan=[0 3]. % Interval over which we want to evaluate y=f(x) y0=[1. x..2). grid The plot below shows the function y = f ( x ) and its derivative dy ⁄ dx . '+r−'. ylabel('y (upper curve).y) dy = −x*y. y(:.y]=ode23('func_exer9_1'. tspan.

% Use 2nd and 3rd Order Runge−Kutta % Plot numeric values with the statements below plot(x.5*h.3280 1. tspan.2000 0.3655 1. The analytical solution is found with syms x y y=dsolve('Dy=3*x^2. '+r−') title('Numeric Solution of Differential Equation of Exercise 9.5*m(2)) 19 m(3)= h*f(xn+h.2000 0.5*L(1)) 13 L(2)= h*f(xn+h.3'). Third Edition Copyright © Orchard Publications ..yn+0. The MATLAB script for the numerical solution is as follows: tspan=[2 4]. grid 9−22 Numerical Analysis Using MATLAB® and Excel®.5*h)+(1+0.5*h)+(1+0. Dy=3*x^2. it follows that y(0) = 7.5(k(1)+k(2)) 10 y(1)= 11 h*f(xn.3906 1. % Interval over which we want to evaluate y=f(x) y0=7.0000 1. ylabel('y'). xlabel('x').2620 0.y(2)=0.yn+0.yn) 12 L(1)= h*f(xn+0.5*m(1)) 18 m(2)= h*f(xn+0.yn+0. y0).5*(D11+D12) h*(0+1^2)= h*[(0+0..yn) 8 k(2)= h*f(xn+h.y]=ode23('fexer9_3'.'x') and MATLAB displays y = x^3-15/2 Next.5.0000 h*(0+1^2)= h*(0+0.yn) 17 m(1)= h*f(xn+0.2735 h*(0+1^2)= h*[(0+0.yn+2*L(2)-L(1)) 14 L(3)= y(0) +(L(1) + 4*L(2) + L(3))/6 15 y(1)= 16 h*f(xn.5.2+(1+0.2620 0.5*h.2000 0.5*m(2))^2]= h*[(0+h)+(1+m(3))^2]= 3.5*L(1)^2)]= h*[(0+h)+(1+2*L(2)-L(1))^2]= 0. y..2640 0.yn+m(3)) 20 m(4)= 21 y(1)= y(0) +(1/6)*(m(1) + 2*m(2)+2*m(3) + m(4)) C D 0.5*h)+(1+0.Chapter 9 Solution of Differential Equations by Numerical Methods 2. % Initial condition: Since y=x^3−15/2 and y(2) = 0.y). A B 1 Differential Equation y' = x + y2 2 Numerical solution by Runga-Kutte method follows 3 h= (given) 4 x(0)= (given) 5 y(0)= Initial condition (given) 6 7 k(1)= h*f(xn.2758 0.2731 0.5*m(1))^2]= h*[(0+0.5'. we write and save the following statements as function file fexer9_3 function Dy=fexer9_3(x.5 [x.5*h.2000 0.2)^2)= y(0)+0.yn+h) 9 y(0)+0.

x..x0). x0=[0.tspan...num_x]=ode23('fexer4'. We write and save the following statements as function file fexer4 function Dy=fexer4(x. xlabel('x').num_x.num_x.y).. The MATLAB script and the plot for the numerical solution are as follows: tspan=[0 10]. title('Numeric solution of non−linear differential equation dy/dx=−x^3+0. '−').'+'. [x... Third Edition Copyright © Orchard Publications 9−23 .707].2sinx'). ylabel('y=f(x)'). grid 5.Solutions to End−of−Chapter Exercises 4. Dy=−y^3+0. Numerical Analysis Using MATLAB® and Excel®. The spreadsheet is shown on the following two pages. plot(x.2*sin(x).

9052+2*L(2)-L(1))]= 21 22 y(2)= y(1) +(L(1) + 4*L(2) + L(3))/6 23 24 h= x(2)-x(1)= 0.1 0.0787 0.5*h.5*L(1))]= L(2)= h*f(xn+0.8773)= L(1)= h*f(xn.yn+0.5146)= L(1)= h*f(xn.2)= Next value x(0) + 2*h 0.2000 26 y(2)= From previous computation 0.2+2*1.5*h)^2-(1+0.3+2*1.5*h)^2-(0.1000 -0.5*L(1)) h*[(0+0.yn+0.5*L(1))]= L(2)= h*f(xn+0.y 2 Numerical solution by Runga-Kutte method follows 3 4 h= x(1)-x(0)= 0.2000 35 x(0.1000 5 x(0)= (given) 0.9052 17 18 h*(0.1000 16 y(1)= From previous computation 0.9052)= L(1)= h*f(xn.5*L(1))]= L(2)= h*f(xn+0.0 0.yn+2*L(2)-L(1)) 41 42 y(4)= y(2) +(L(1) + 4*L(2) + L(3))/6 43 D E F x0= x1= x2= 0.yn+0.2 -0.9052 -0.0838 -0.8213 27 28 h*(0.2405 1.5*h)+2*(1+0.5*h.5*L(1)) 10 h*[(0+h)^2-(1+2*L(2)-L(1))]= L(3)= h*f(xn+h.Chapter 9 Solution of Differential Equations by Numerical Methods A B C 1 Differential Equation is y' = x2 .0000 6 y(0)= Initial condition (given) 1.yn+2*L(2)-L(1)) h*[(0.3)= Next value x(0) + 3*h -0.6000 36 y(3)= From previous computation 1.1000 15 x(0.0000 7 8 h*(0-1)= L(1)= h*f(xn.1981 0. Third Edition Copyright © Orchard Publications .1)= Next value x(0) + h 0.1843 0.0305 37 38 h*(0.yn) 29 h*[(0+0.5*L(1)) 40 h*[(0+h)+2*(1+2*L(2)-L(1))]= L(3)= h*f(xn+h.yn) 9 h*[(0+0.9052+0.5*L(1)) 30 h*[(0+h)+2*(1+2*L(2)-L(1))]= L(3)= h*f(xn+h.2138 -0.5*h.yn+2*L(2)-L(1)) 11 12 y(1)= y(0) +(L(1) + 4*L(2) + L(3))/6 13 14 h= x(2)-x(1)= 0.8213 0.yn) 19 L(2)= h*f(xn+0.1000 25 x(0.yn+0.1+h)^2-(0.2077 0.0305 -0.8063 continued on next page 9−24 Numerical Analysis Using MATLAB® and Excel®.yn) 39 h*[(0+0.5*h)+2*(1+0.0948 -0.0901 0.0895 -0.5*h.1^2-0.2922 -0.yn+2*L(2)-L(1)) 31 32 y(3)= y(2) +(L(1) + 4*L(2) + L(3))/6 33 34 h= x(3)-x(2)= -0.5*L(1))]= 20 L(3)= h*f(xn+h.

L(1). y(0) + 0.0950 0.9050 -0.0000 -0.5*h.8063 6.0000 0.0000 1. Third Edition Copyright © Orchard Publications 9−25 .0992 0.5*h. y'(0) + 0.yn+0.3+2*1.0000 0.0000 0. y'(0))= 11 L(2)= h*(y'(0) + 0.5*L(1))= 13 L(3)= h*(y'(0) + 2*L'(2) .8063 h*(0.5*L'(1))= 12 L'(2)= h*f(x(0) + 0.1000 0.L'(1))= 15 16 y(1)= y(0) +(L(1) + 4*L(2) + L(3))/6 17 y'(1)= y'(0) + (L'(1) + 4*L'(2) + L'(3))/6= 18 19 h= x(1)-x(0)= 20 x(0.1)= Next value x(0) + h 21 y(1)= From previous computation 22 y'(1)= From previous computation 23 C 0.0000 -1.0901 0.Solutions to End−of−Chapter Exercises A 44 45 46 47 48 49 50 51 52 53 h= x(0. y(0).5*L(1).L'(1))= 14 L'(3)= h*f(x(0) + h.4)= y(3)= L(1)= L(2)= L(3)= y(4)= B x(4)-x(3)= Next value x(0) + 4*h From previous computation h*f(xn.9003 continued on next page Numerical Analysis Using MATLAB® and Excel®. y(0) + 2*L(2) . y'(0) + 2*L'(2) .5*h)+2*(1+0.1000 -0.9050 -0.0000 0.8773)= h*[(0+0.5*L(1))]= h*[(0+h)+2*(1+2*L(2)-L(1))]= D E F 0.0000 0.yn+2*L(2)-L(1)) y(2) +(L(1) + 4*L(2) + L(3))/6 C 0.0998 -0.1000 0.1000 0.5*L(1)) h*f(xn+h. A B 1 Differential Equation is y''+y'=xy or y''=xy-y' 2 Numerical solution by Runga-Kutte method follows 3 4 h= x(1)-x(0)= 5 x(0)= Initial condition (given) 6 y(0)= Initial condition (given) 7 y'(0)= Initial condition (given) 8 9 L(1)= h*y'(0)= 10 L'(1)= h*f(x(0).1000 0.yn) h*f(xn+0.9003 0.

y(0).0949 -0.0803 0.5*L'(1))= h*(y'(1) + 2*L'(2) . y'(0) + 0. y'(1) + 0.5*L(1).5*L(1).0991 -0.5*h. y'(1) + 2*L'(2) .0851 0.L(1).1000 0.L'(1))= h*f(x(0) + h.5*h. y'(0))= h*(y'(0) + 0. y(0) + 0.L(1).2)= y(2)= y'(2)= L(1)= L'(1)= L(2)= L'(2)= L(3)= L'(3)= y(3)= y'(3)= B h*y'(1)= h*f(x(1). y(1). y(0) + 2*L(2) .0966 -0.Chapter 9 Solution of Differential Equations by Numerical Methods 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 A L(1)= L'(1)= L(2)= L'(2)= L(3)= L'(3)= y(2)= y'(2)= h= x(0. y'0 + 2*L'(2) .0709 0.7074 9−26 Numerical Analysis Using MATLAB® and Excel®.L'(1))= y(0) +(L(1) + 4*L(2) + L(3))/6 y'(0) + (L'(1) + 4*L'(2) + L'(3))/6= x(2)-x(1)= Next value x(0) + 2*h From previous computation From previous computation h*y'(0)= h*f(x(0).L'(1))= h*f(x(1) + h.5*L'(1))= h*f(x(1) + 0.0754 0.8199 -0. Third Edition Copyright © Orchard Publications .5*L'(1))= h*f(x(0) + 0.8023 0.0968 0.8199 -0.5*L(1))= h*(y'(0) + 2*L'(2) .7444 -0. y(1) + 0.0802 0.8023 -0.0934 0.2000 0.0900 0.0980 -0.L'(1))= y(0) +(L(1) + 4*L(2) + L(3))/6 y'(0) + (L'(1) + 4*L'(2) + L'(3))/6= C -0. y'(1))= h*(y'(1) + 0. y(1) + 2*L(2) .

and so on.1 The Trapezoidal Rule Consider the function y = f ( x ) for the interval a ≤ x ≤ b .. yn – 1 xn – 1 yn x b x2 Figure 10. Third Edition Copyright © Orchard Publications 10−1 .1. f (x) P0 y0 0 a P1 P2 Pn Pn – 1 y1 x1 y2 . and Simpson’s rule that computes a function f ( x ) with a set of quadratic functions. from x 1 to x 2 .Chapter 10 Integration by Numerical Methods T his chapter is an introduction to numerical methods for integrating functions which are very difficult or impossible to integrate using analytical means...1. the number of points between x 0 = a and x n = b is ---------- n x 1 = a + Δx. Integration by the trapezoidal rule To evaluate the definite integral ∫a f ( x ) dx .. shown in Figure 10. The total area is ∫a b f ( x ) dx = ∫a x1 f ( x ) dx + ∫x x2 1 f ( x ) dx + … + ∫x b n f ( x ) dx = n–1 k=1 ∑ ∫x xk k–1 f ( x ) dx The integral over the first subinterval. We will discuss the trapezoidal rule that computes a function f ( x ) with a set of linear functions. and finally from x n – 1 to b . x n – 1 = a + ( n – 1 )Δx . we divide the interval a ≤ x ≤ b b into n subintervals each of length Δx = b – a .. 10. ….. Then... Therefore.. the integral from a to b is the sum of the integrals from a to x 1 .. x 2 = a + 2Δx. can now be approximated by the area of the trapezoid Numerical Analysis Using MATLAB® and Excel®.

( y 1 + y 2 )Δx + … + -. Then.Chapter 10 Integration by Numerical Methods 1 aP0 P 1 x 1 that is equal to -.3 3 3 (10. Third Edition Copyright © Orchard Publications .= 2.y n⎞ Δx ⎝2 2 ⎠ Trapezoidal Rule (10.2) Compare with the exact value.( y n – 1 + y n )Δx 2 2 2 or 1 1 T = ⎛ -.( y 1 + y 2 )Δx . Solution: The exact value of this integral is ∫1 2 x x dx = ---3 2 3 2 1 8 1 7 = -.1 Using the trapezoidal rule with n = 4 . and so on. the trapezoidal approximation becomes 2 1 1 1 T = -.33333 . and compute the percent error.= -.( y 0 + y 1 )Δx plus the area of the trapezoid x 1 P 1 P 2 x 2 that is equal to 2 1 -.1) Example 10.y 0 + y 1 + y 2 + … + y n – 1 + -.( y 0 + y 1 )Δx + -. estimate the value of the definite integral ∫1 x d x 2 2 (10.– -. y = f (x) = x 2 10−2 Numerical Analysis Using MATLAB® and Excel®.3) For the trapezoidal rule approximation we have x0 = a = 1 xn = b = 2 n = 4 --------------------Δx = b – a = 2 – 1 = 1 4 4 n Then.

= -------. compute the percent error when n = 5 and n = 10 Solution: The exact value is found from ∫1 -.4).n) where y is the integral with respect to x. Example 10.= 2. Third Edition Copyright © Orchard Publications 10−3 .33333 % Error = -------------------------------------------.+ ----.34375 ⎝ 2 16 16 16 16 2 16 ⎠ 4 16 4 32 (10.× -.+ -.y) to approximate the value of the integral ∫1 -.5) The MATLAB function trapz(x.× ----. Numerical Analysis Using MATLAB® and Excel®.34375 – 2.6) and by comparison with the exact value.× 100 = 0.3) and (10.4) From (10.0000 = 0.6931 For the approximation using the trapezoidal rule.× ----.⎞ × -.dx x 2 1 = ln x 2 1 = ln 2 – ln 1 = 0.= ----. approximates the integral of a function y = f ( x ) using the trapezoidal rule.+ ----.y. we find that the percent error is 2.⎞ = 25 ----⎝4 ⎠ 16 6 2 y 2 = f ( x 2 ) = ⎛ -.T = ⎛ -.6931 – 0. 150 1 75 1 1 16 25 36 49 1 64 .2 Use the MATLAB function trapz(x.45 % 2.The Trapezoidal Rule x0 = a = 1 x 1 = a + Δx = 5 -4 -x 2 = a + 2Δx = 6 4 -x 3 = a + 3Δx = 7 4 x4 = b = 2 2 ----y 0 = f ( x 0 ) = 1 = 16 16 5 2 y 1 = f ( x 1 ) = ⎛ -..⎞ ⎝ 4⎠ 2 = 64 ----16 and by substitution into (10.1).33333 (10. we let x 5 represent the row vector with n = 5 .⎞ = 49 ----⎝ 4⎠ 16 8 y 4 = f ( x 4 ) = ⎛ -.⎞ = 36 ----⎝ 4⎠ 16 7 2 y 3 = f ( x 3 ) = ⎛ -. and n (optional) performs integration along dimension n .+ ----.dx x 2 1 (10.

× 100 = 0.7) where τ is a dummy variable of integration. 2. The MATLAB script for this example is t=linspace(0.56 % 0.6931 % Error = -------------------------------------.y) MATLAB displays the following result.y5).2.6931 and the percent error when Δx = 1 ⁄ 10 is used is 0. y5=1. area10=trapz(x10./x10.10).6939 – 0. Use the MATLAB trapz(x.6931 Example 10. y=exp(−t. Third Edition Copyright © Orchard Publications .6931 % Error = -------------------------------------. Solution: We use the same procedure as in the previous example.× 100 = 0. area5=trapz(x5.y) function to find the area under this integral with n = 10 when the upper limit of integration is t = 2 .∫ e–τ dτ π 0 10−4 Numerical Analysis Using MATLAB® and Excel®.^2).2. that is. The corresponding values of y are denoted as y 5 and y 10 . Δx = 1 ⁄ 5 and Δx = 1 ⁄ 10 respectively.12 % 0.6970 – 0.6939 The percent error when Δx = 1 ⁄ 5 is used is 0.6970 area10 = 0.Chapter 10 Integration by Numerical Methods and x 10 the vector with n = 10 . x5=linspace(1. and the areas under the curve as area5 and area10 respectively. is called the error function* and it is used extensively in communications theory.u 2 * The formal definition of the error function is erf ( u ) = -----.y10) area5 = 0./x5.10). y10=1.5). We use the following MATLAB script. area=trapz(t. x10=linspace(1.2.3 The integral f( t) = ∫0 t e –τ 2 dτ (10.

2 sin 3t – 1) (10.4 The i – v (current−voltage) relation of a non−linear electrical device is given by i ( t ) = 0. The energy W ( t 0. The MATLAB script below will compute and plot the power. Third Edition Copyright © Orchard Publications 10−5 .10) When this script is executed. For this example.100). p=v. it returns a warning.b) function where f is a symbolic expression. we will find and sketch the power and energy by the trapezoidal rule using the MATLAB trapz(x. t=sym('t'). t 1 ) dissipated in this device from t 0 = 0 to t 1 = 10 s. v=sin(3.0. By any means.*(exp(0.2v ( t ) – 1) (10.2. MATLAB returns the following message when integration is attempted with the symbolic expression of (10. MATLAB displays the following message.1 ∫0 10 s sin 3t ( e 0.y) function.*t). t=linspace(0. The instantaneous power is p ( t ) = v ( t )i ( t ) = 0.8) where v ( t ) = sin 3t .1 sin 3t ( e 0. and a and b are the lower and upper limits of integration respectively. we choose n = 100 . t 1 ) = ∫t t1 0 p ( t ) dt = 0. We can try the MATLAB int(f.10.10) An analytical solution of the last integral is possible using integration by parts. The instantaneous power p ( t ) b.8818 Example 10. i=0.a. so that Δx = 1 ⁄ 100 .1*sin(3*t)*(exp(0.2*sin(3*t))−1). that is.9) b.2 sin 3t – 1 ) dt (10. The energy is the integral of the instantaneous power. Next.The Trapezoidal Rule area = 0. find a.1 ( e 0. When MATLAB cannot find a solution. W ( t 0.*v)−1).*i. Solution: a. For this example. but it is not easy. Numerical Analysis Using MATLAB® and Excel®.1. s=int(0. Warning: Explicit integral could not be found.10).

2 Simpson’s Rule The trapezoidal and Simpson’s rules are special cases of the Newton−Cote rules which use higher degree functions for numerical integration. energy=trapz(t. xlabel('seconds'). title('Energy vs Time').2.p)..025 Power vs Time 0. and thus we expect the energy to be constant.1013 joule. '+'). 0. xlabel('seconds'). ylabel('joules') energy = 0.4 The plot of Figure 10. The energy is shown in Figure 10.01 0.energy..3. plot(t.2. Plot for the power variation in Example 10. title('Power vs Time').02 0.4 be represented by the parabola y = αx + βx + γ 2 (10. grid.1013 Thus. The MATLAB script below computes and plots the energy.11) 10−6 Numerical Analysis Using MATLAB® and Excel®.2 shows that the power is uniform for all time. grid.005 0 0 1 2 3 4 5 seconds 6 7 8 9 10 Figure 10. the value of the energy is 0. ylabel('watts') The power varies in a uniform fashion as shown by the plot of Figure 10. Third Edition Copyright © Orchard Publications . 10..015 watts 0.Chapter 10 Integration by Numerical Methods plot(t. Let the curve of Figure 10.p).

5 joules 0 -0.3.+ γ x 3 2 2 3 2 αx 3 βx 2 h –h 3 αh . ( 0.– γ h⎞ = 2 α h . y 0 ) . and ( h. y2 ) .4.βh-------.5 -1 0 1 2 3 4 5 seconds 6 7 8 9 10 Figure 10. y 1 ) .Simpson’s Rule 1.4 y y0 y1 y2 x −h 0 h Figure 10. Third Edition Copyright © Orchard Publications 10−7 .5 Energy vs Time 1 0.+ ------.+ ------.h ( 2 α h + 6 γ ) 3 The curve passes through the three points ( – h. Simpson’s rule of integration The area under this curve for the interval – h ≤ x ≤ h is Area h –h = ∫– h ( α x 3 h 2 + β x + γ ) dx = -------.+ ------. Then. Plot for the energy of Example 10. by (10.11) we have: Numerical Analysis Using MATLAB® and Excel®.β h ----------= -------.12) 1 3 = -.+ γ h – ⎛ – α h .+ 2 γ h ⎝ 3 ⎠ 3 2 2 3 (10.

14) (10. By substitution of (b) of (10. γ and express (10. we can apply (10.15) yields 2 α h = y 0 – 2y 1 + y 2 2 (10. Thus.14) with (10. A B C y0 y1 h a y2 y3 y4 yn – 1 yn b Figure 10. This is done with the following procedure.5. y 1 and y 2 .12) we obtain Area h –h 1 1 3 = -.h ( 2 α h + 6 γ ) = -.h [ ( y 0 – 2y 1 + y 2 ) + 6y 1 ] 3 3 Area h –h (10. Simpson’s rule of integration by successive segments From Figure 10.5.16) and by substitution into (10. we observe that each segment of width 2h of the curve can be approximated by a parabola through its ends and its midpoint.13) into (a) and (c) and rearranging we obtain α h – β h = y0 – y1 α h + β h = y2 – y1 2 2 (10.15) Addition of (10.Chapter 10 Integration by Numerical Methods y0 = α h – β h + γ y1 = γ y2 = α h + β h + γ 2 2 (a) (b) (c) (10. Third Edition Copyright © Orchard Publications .12) in terms of h .13) We can now evaluate the coefficients α.h ( y 0 + 4y 1 + y 2 ) 3 (10.5. β. y 0 .18) to successive segments of any curve y = f ( x ) in the interval a ≤ x ≤ b as shown on the curve of Figure 10. the area under segment AB is 10−8 Numerical Analysis Using MATLAB® and Excel®.17) or 1 = -.18) Now.

21) Since each segment has width 2h . Numerical Analysis Using MATLAB® and Excel®. we construct the following table using the spreadsheet of Figure 10. When the areas under each segment are added. syms x Area=int(1/x.h ( y 0 + 4y 1 + y 2 ) 3 (10. the area under segment BC is Area BC 1 = -.2) Area = log(2) We recall that log(x) in MATLAB is the natural logarithm.19) Likewise.6. we obtain 1 Area = -.b) function where f is a symbolic expression.dx x 2 1 (10.20) and so on. and a and b are the lower and upper limits of integration respectively. compute the approximate value of ∫1 -.6931 .1. to apply Simpson’s rule of numerical integration.a. the number n of subdivisions must be even. for convenience. To use Simpson’s rule.Simpson’s Rule Area AB 1 = -.21) is found from b–a h = ---------n n = even (10.22) Example 10. We can also find the analytical value with MATLAB’s int(f.5 Using Simpson’s rule with 4 subdivisions ( n = 4 ) . For this example.23) Solution: This is the same integral as that of Example 10. The value of h for (10. This restriction does not apply to the trapezoidal rule of numerical integration.2 where we found that the analytical value of this definite integral is ln = 0. Third Edition Copyright © Orchard Publications 10−9 .h ( y 2 + 4y 3 + y 4 ) 3 (10.h ( y 0 + 4y 1 + 2y 2 + 4y 3 + 2y 4 + … + 2y n – 2 + 4y n – 1 + y n ) 3 Simpson′s Rule of Numerical Integration (10.

MATLAB has two quadrature functions for performing numerical integration.31905 0.5 2 ∫ (1/x)dx evaluated from a = 1 to b = 2 with n = 4 3 Numerical integration by Simpson's method follows a= 1 4 Given b= 2 5 n= 4 6 0.57143 4 x4=b= 17 2. while the quad8 function uses the so−called Newton−Cotes 8−panel rule.28571 0.Chapter 10 Integration by Numerical Methods A B C D 1 Example 10. The description of these can be seen by typing help quad or help quad8.23) By comparison of the numerical with the exact value. q=quad(‘f’. Third Edition Copyright © Orchard Publications .00000 1 x1=a+h= 11 1.50000 1 19 Sum of Products = 20 Area = (h/3)*(Sum of Products) = (1/12)*8.00000 y0=1/x0= 10 1.69325 Figure 10.2500 7 Then. Both functions have the same syntax. h = (b-a)/n = Multiplier 8 x0=a= 9 1. that is.66667 2 x3=a+3h= 15 1.31905 = E Products 1.80000 4 x2=a+2h= 13 1.a. and a and b are the lower and upper limits of –3 integration respectively.20000 1.00000 y4=1/x4= 18 0. The quad function uses an adaptive form of Simpson’s rule.b. The quad8 function detects and handles irregularities more efficiently. Both of these functions use adaptive quadrature methods.50000 8. 10−10 Numerical Analysis Using MATLAB® and Excel®. it is understood to be the standard tolerance of 10 . The string ‘f’ is the name of a user defined function. we observe that the error is very small when Simpson’s method is applied.25000 y1=1/x1= 12 0.tol). If tol is omitted. When such irregularities occur.00000 3. and integrate to a relative error tol which we must specify.50000 y2=1/x2= 14 0. the quad and quad8.6. MATLAB displays a warning message but still provides an answer. Spreadsheet for numerical integration of (10. this means that these methods can handle irregularities such as singularities.75000 y3=1/x3= 16 0.33333 2.

2. syms x.tol) function where tol = 10 d.a.a. Obtain the value of this integral with the q=quad8(‘f’. With the specified tolerance.a. First.m as shown below.0.b.10^−10) y_tol = 0. function y = errorfcn1(x) y = exp(−x.2) % Define symbolic variable x and integrate – 10 – 10 pretty(y) y = 1/2*erf(2)*pi^(1/2) 1/2 1/2 erf(2) pi erf is an acronym for the error function and we can obtain its definition with help erf b. We name it errorfcn1. we write and execute the following MATLAB script.a.0.^2). Obtain the value of this integral with the q=quad(‘f’. Obtain the value of this integral with the q=quad8(‘f’. Use MATLAB’s symbolic int function to obtain the value of this integral b.88211275610253 c.2) We obtain the answer in standard tolerance form as y_std = 0. We will use format long to display the values with 15 digits. Obtain the value of this integral with the q=quad(‘f’.tol) function where tol = 10 Solution: a.Simpson’s Rule Example 10. we need to create and save a function m−file.88208139076242 Numerical Analysis Using MATLAB® and Excel®. y=int(exp(−x^2). format long y_std=quad('errorfcn1'.b.24) a. Third Edition Copyright © Orchard Publications 10−11 .b) function c.0.b) function e. With this file saved as errorfcn1.m. the script and the answer are as follows: y_tol=quad('errorfcn1'.6 Given the definite integral y= f(x ) = ∫0 2 e –x 2 dx (10.

% fprintf(' \n'). With the standard tolerance.8 ). 10−12 Numerical Analysis Using MATLAB® and Excel®. both quad and quad8 produce the same result.2.26) where a and b are non−negative values.b) where a and b are non−negative. b=input('Enter second point "b" (non−negative): ').7 Using the quad and quad8 functions with standard tolerance. evaluate the integral y = f(x) = b – 10 tolerance.2. Substitution of the values of the given values of a and b will be included in the MATLAB script below. The script is then saved as Example_10_7. The sqrt function in a built−in function and therefore. Use the fprintf function to display first the analytical values.2) y_std8 = 0.25) at ( ( a. and (3.0. We will include the input function in the script. we obtain y = ∫a b x 1⁄2 x dx = --------3⁄2 3⁄2 b a 2 3⁄2 3⁄2 = -. ∫a x dx (10. 0. b ) = ( 0. ( 1. we need not write a user defined m−file.Chapter 10 Integration by Numerical Methods d.88208139076242 We observe that with the 10 Example 10. y_std8=quad8('errorfcn1'. % Insert line a=input('Enter first point "a" (non−negative): '). With the specified tolerance. then. Third Edition Copyright © Orchard Publications .88208139076194 e. 2.3 ) ). the numerical values produced by the quad and quad8 functions for each set of data.4.10^−10) y_tol8 = 0. % This script displays the approximations obtained with the quad and quad8 functions % with the analytical results for the integration of the square root of x over the % interval (a. Solution: Evaluating the given integral.0.( b –a ) 3 (10.8) . y_tol8=quad8('errorfcn1'.

2 Enter second point "b" (non-negative): 0.. Third Edition Copyright © Orchard Publications 10−13 .b). Example_10_7 Enter first point "a" (non-negative): 0. kq=quad('sqrt'. that is.4 Enter second point "b" (non-negative): 2. fprintf(' \n').^(1.a..417396 Example_10_7 0. quad8: 0.kq8)..221080 Enter first point "a" (non-negative): 3 Enter second point "b" (non-negative): 8 Analytical: 11.^(1.k..620825 11..b).620843 Numerical quad.8 Analytical: 0. quad8: 1. fprintf(' \n') % Insert two lines Now. we execute this saved file by typing its name.221080 Example_10_7 1. quad8: 11.620843 Numerical Analysis Using MATLAB® and Excel®. fprintf(' \n'). quad8: %f %f \n'. kq8=quad8('sqrt'.5))..a.5)−a.3 Analytical: 1.kq. % Insert line fprintf(' Analytical: %f \n Numerical quad.Simpson’s Rule k=2/3.417399 Enter first point "a" (non-negative): 1.417399 Numerical quad.*(b.221080 Numerical quad.

• MATLAB has two quadrature functions for performing numerical integration. –3 10−14 Numerical Analysis Using MATLAB® and Excel®. while the quad8 function uses the so−called Newton−Cotes 8− panel rule. Third Edition Copyright © Orchard Publications . it is understood to be the standard tolerance of 10 . • We can evaluate a definite integral using the expression ∫a f ( x ) dx b with Simpson’s rule of numerical integration 1 Area = -. and a and b are the lower and upper limits of integration respectively. and n (optional) performs integration along dimension n . approximates the integral of a function y = f ( x ) using the trapezoidal rule.h ( y 0 + 4y 1 + 2y 2 + 4y 3 + 2y 4 + … + 2y n – 2 + 4y n – 1 + y n ) 3 where the number n of subdivisions must be even. • We can perform numerical integration with the MATLAB function int(f.3 Summary • We can evaluate a definite integral ∫ f ( x ) dx with the trapezoidal approximation a b 1 1 T = ⎛ -. If tol is omitted.tol). The number n of n subdivisions can be even or odd. Both functions have the same syntax. The quad function uses an adaptive form of Simpson’s rule. q=quad(‘f’.a.b) function where f is a symbolic expression. and a and b are the lower and upper limits of integration respectively. • The trapezoidal and Simpson’s rules are special cases of the Newton−Cote rules which use higher degree functions for numerical integration.n) where y is the integral with respect to x.y n ⎞ Δx ⎝2 2 ⎠ ---------by dividing interval a ≤ x ≤ b into n subintervals each of length Δx = b – a . and integrate to a relative error tol which we must specify. that is.y.b. The string ‘f’ is the name of a user defined function. Both of these functions use adaptive quadrature methods. • The MATLAB function trapz(x. The quad8 function detects and handles irregularities more efficiently. the quad and quad8.a.Chapter 10 Integration by Numerical Methods 10.y 0 + y 1 + y 2 + … + y n – 1 + -.

∫ x dx 0 2 n = 4 b. Use Simpson’s rule to approximate the following definite integrals and compare your results with the analytical values. Verify your answers with the MATLAB quad(‘f’.y. a. ∫ x dx 4 0 2 n = 4 1d.Exercises 10. Third Edition Copyright © Orchard Publications 10−15 .dx 2 1 2 x n = 4 2.n) function. ∫ ---.a.4 Exercises 1. ∫ x dx 3 0 2 n = 4 c. a.dx 2 0 1 x +1 n = 4 Numerical Analysis Using MATLAB® and Excel®. ∫ sin x dx 0 π n = 4 1 c. ∫ x dx 2 0 2 n = 4 b.b) function. Use the trapezoidal approximation to compute the values the following definite integrals and compare your results with the analytical values. Verify your answers with the MATLAB trapz(x. ∫ -------------.

2.y) area = 2 b. Third Edition Copyright © Orchard Publications .Chapter 10 Integration by Numerical Methods 10. area=trapz(x.+ 1 + -.y n ⎞ Δx ⎝2 2 ⎠ a. y=x. The exact value is 2 4 2 0 ∫0 x x dx = ---4 3 = 4 For the trapezoidal rule approximation we have 10−16 Numerical Analysis Using MATLAB® and Excel®.4).× 0 + -.+ -.5 Solution to End−of−Chapter Exercises 1.= 4 × -.T = ⎛ -. The exact value is ∫0 2 x x dx = ---2 2 2 0 = 2 For the trapezoidal rule approximation we have x0 = a = 0 xn = b = 2 n = 4 --------------------Δx = b – a = 2 – 0 = 1 2 4 n y = f (x) = x x0 = a = 0 -x 1 = a + Δx = 1 2 x 2 = a + 2Δx = 1 3 x 3 = a + 3Δx = -2 x4 = b = 2 y0 = f ( x0 ) = 0 y1 = f ( x1 ) = 1 -2 y2 = f ( x2 ) = 1 3 y 3 = f ( x 3 ) = -2 y4 = f ( x4 ) = 2 1 1 1 1 3 1 . 1 1 T = ⎛ -.× 2 ⎞ × -.y 0 + y 1 + y 2 + … + y n – 1 + -.= 2 ⎝2 ⎠ 2 2 2 2 2 x=linspace(0.

y) area = 4. c. area=trapz(x.= ----------. y=x. Third Edition Copyright © Orchard Publications 10−17 .= ----------. The exact value is 2 5 2 0 ∫0 4 xx dx = ---5 = 32 = 6.= -n 4 2 y = f (x) = x 4 Numerical Analysis Using MATLAB® and Excel®.Solution to End−of−Chapter Exercises x0 = a = 0 xn = b = 2 n = 4 b–a 2–0 1 Δx = ---------.^3.4).= ⎛ 5 + -.= -n 4 2 y = f (x) = x x0 = a = 0 1 x 1 = a + Δx = -2 x 2 = a + 2Δx = 1 -x 3 = a + 3Δx = 3 2 x4 = b = 2 3 y0 = f ( x0 ) = 0 1 y 1 = f ( x 1 ) = -8 y2 = f ( x2 ) = 1 y 3 = f ( x 3 ) = 27 ----8 y4 = f ( x4 ) = 8 1 1 1 7 1 27 1 .25 ⎝ ⎠ 2 ⎝2 ⎠ 2 2 8 8 2 x=linspace(0.= 4.T = ⎛ -.+ -.4444 The deviations from the exact value are due to the small number of divisions n we chose.× 0 + -.× 8 ⎞ × -.2.4 ----5 For the trapezoidal rule approximation we have x0 = a = 0 xn = b = 2 n = 4 b–a 2–0 1 Δx = ---------.⎞ × -.+ 1 + ----.

Chapter 10 Integration by Numerical Methods x0 = a = 0 1 x 1 = a + Δx = -2 x 2 = a + 2Δx = 1 -x 3 = a + 3Δx = 3 2 x4 = b = 2 y0 = f ( x0 ) = 0 1 y 1 = f ( x 1 ) = ----16 y2 = f ( x2 ) = 1 y 3 = f ( x 3 ) = 81 ----8 y 4 = f ( x 4 ) = 16 1 1 1 41 1 81 1 .5720 d.----.= ----------. Third Edition Copyright © Orchard Publications .⎞ × -.y) area = 7.= 7.2.× 0 + ----.× 16 ⎞ × -.⎞ × -.× 1 + 16 + 4 + 16 + -.4)..⎞ × -.+ -.T = ⎛ -.0625 ⎝ ⎠ 2 ⎝2 ⎠ 2 8 16 16 2 x=linspace(0.+ 1 + ----.5090 ⎝ 1918 ⎠ 4 ⎝2 2 4⎠ 4 25 9 49 10−18 Numerical Analysis Using MATLAB® and Excel®.^4.= ⎛ 9 + ----.-.. area=trapz(x.dx = – 1 -2 x x 2 1 1 = -2 For the trapezoidal rule approximation we have x0 = a = 1 xn = b = 2 n = 4 b–a 2–1 1 Δx = ---------.= 0.× -.T = ⎛ -. The exact value is ∫1 2 1 ---. y=x.= ⎛ ----------.= -n 4 4 y = f (x) = 1 ⁄ x x0 = a = 1 -x 1 = a + Δx = 5 4 3 x 2 = a + 2Δx = -2 -x 3 = a + 3Δx = 7 4 x4 = b = 2 2 y0 = f ( x0 ) = 1 y 1 = f ( x 1 ) = 16 ----25 4 y 2 = f ( x 2 ) = -9 y 3 = f ( x 3 ) = 16 ----49 1 y 4 = f ( x 4 ) = -4 3905 1 1 1 1 1 ----.

B C D E Exercise 10.00000 Sum of Products = Area = (h/3)*(Sum of Products) = (1/12)*8. We write and execute the following MATLAB script: y_std=quad('exer_10_2_a'.m as shown below.0000 Products We create and save a function m−file.4).= 2. h = (b-a)/n = x0=a= y0=x02= x1=a+h= y1=x12= x2=a+2h= y2=x22= x3=a+3h= y3=x32= x4=b= y4=x42= 0 2 4 0.00000 0.Solution to End−of−Chapter Exercises x=linspace(1. 1 Area = -. The exact value is ∫0 A 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 2 x x dx = ---3 2 3 2 0 8 = -. We name it exer_10_2_a. y=1.5158 2. area=trapz(x. function y = exer_10_2_a(x) y = x.2.0000 16.a ∫ x 2 dx evaluated from a = 0 to b = 2 with n = 4 Numerical integration by Simpson's method follows Given a= b= n= Then.y) area = 0.h ( y 0 + 4y 1 + 2y 2 + 4y 3 + 2y 4 + … + 2y n – 2 + 4y n – 1 + y n ) 3 a. Third Edition Copyright © Orchard Publications 10−19 .0000 1 0.0./x.00000 1.6667 3 To use Simpson’s rule we construct the following table using a spreadsheet.0000 2 2.6667 Numerical Analysis Using MATLAB® and Excel®.0000 4 1.00000 4.^2.5000 Multiplier 0.31905 = 1 4.50000 2.2) y_std = 2.^2.0000 2.2.6667 4 9.25000 1.50000 0.00000 1.00000 0.25000 2.

70711 1. 10−20 Numerical Analysis Using MATLAB® and Excel®.70711 3.78540 0. h = (b-a)/n = x0=a= y0=sinx0= x1=a+h= y1=sinx1= x2=a+2h= y2=sinx2= x3=a+3h= y3=sinx3= x4=b= y4=sinx4= 0 3.8284 1 0.7854 Multiplier 0.0000 4 2.0000 Products We create and save a function m−file. The exact value is π ∫0 sin x dx A 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 B = – cos x π 0 = –( – 1 – 1 ) = 2 To use Simpson’s rule we construct the following table using a spreadsheet.0046 4 2.7854 4 To use Simpson’s rule we construct the following table using a spreadsheet.14159 4 0.b ∫ sinxdx evaluated from a = 0 to b = π with n = 4 Numerical integration by Simpson's method follows Given a= b= n= Then. function y = exer_10_2_b(x) y = sin(x).= 0.14159 0. We name it exer_10_2_b. C D E Exercise 10.00000 0.dx = tan – 1x 2 x +1 1 0 π = -.0000 c.00000 Sum of Products = Area = (h/3)*(Sum of Products) = (1/12)*8.Chapter 10 Integration by Numerical Methods b. Third Edition Copyright © Orchard Publications .0.2.m as shown below.00000 2.35619 0.0000 7. The exact value is ∫0 1 1 -------------.57080 1.pi) y_std = 2.6569 2. We write and execute the following MATLAB script: y_std=quad('exer_10_2_b'.8284 2 2.31905 = 1 0.00000 0.

50000 0.00000 1.7647 1 1. function y = exer_10_2_c(x) y = 1.80000 0.7854 4 2.2500 Multiplier 0.5000 9.2. Third Edition Copyright © Orchard Publications 10−21 .25000 0.4247 0.Solution to End−of−Chapter Exercises A 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 B C D E Exercise 10.75000 0.94118 0.m as shown below. We name it exer_10_2_c.64000 1.1) y_std = 0.c ∫ (1/(x 2 +1))dx evaluated from a = 0 to b = 1 with n = 4 Numerical integration by Simpson's method follows Given a= b= n= Then.0000 Products We create and save a function m−file.^2+1).31905 = 1 0.00000 0.6000 4 3.5600 2 1.7854 Numerical Analysis Using MATLAB® and Excel®.50000 Sum of Products = Area = (h/3)*(Sum of Products) = (1/12)*8. We write and execute the following MATLAB script: y_std=quad('exer_10_2_c'.0.00000 0. h = (b-a)/n = x0=a= y0=1/(x02+1)= x1=a+h= y1=1/(x12+1)= x2=a+2h= y2=1/(x22+1)= x3=a+3h= y3=1/(x32+1)= x4=b= y4=1/(x42+1)= 0 1 4 0./(x.

i.2) is written as Ef ( x ) = f ( x + 1 ) = f x + 1 (11. A difference equation is a specific type of recurrence relation. and the subscript k is normally omitted. and other sciences. and Applications The difference equations discussed in this chapter. The E operator increases the argument of a function by one interval h . in (11.e. h = 1 . Numerical Analysis Using MATLAB® and Excel®. Third Edition Copyright © Orchard Publications 11−1 . the difference operator E is Ef ( x k ) = f ( x k + h ) = f ( x k + h ) (11. and a practical example in electric circuit theory is given at the end of this chapter.3). and this type is discussed in this chapter.2 Definition.1 Introduction In mathematics.Chapter 11 Difference Equations T his chapter is an introduction to difference equations based on finite differences. physics. The general form of a linear. In terms of the interval h .2) The interval h is usually unity. a recurrence relation is an equation which defines a sequence recursively: each term of the sequence is defined as a function of the preceding terms. (11. and r is a positive integer that denotes the order of the difference equation.1) where a k represents a constant coefficient and E is an operator similar to the D operator in ordinary differential equations. Difference equations as used with discrete type systems. The discussion is limited to linear difference equations with constant coefficients.3) If. Solutions. that is. Thus. we obtain the second order operator E 2 . 11. mathematics. The Fibonacci numbers are defined. are used in numerous applications such as engineering.. are discussed in Appendix A. we increase the argument of f by another unit. constant coefficient difference equation has the form ( ar E + ar – 1 E r r–1 + … + a 1 E + a 0 )y = φ ( x ) (11. 11.

we add the particular solution Y ( x ) to it to obtain the total solution.7) If y 1 ( x ) and y 2 ( x ) are any two solutions of (11. we first find the solution of the homogeneous difference equation.Chapter 11 Difference Equations E f ( x ) = E [ Ef ( x ) ] = Ef ( x + 1 ) = f ( x + 2 ) = f x + 2 2 (11. any other solution of (11. then the complete solution is y = k1 y1 ( x ) + k2 y2 ( x ) + Y ( x ) (11. Third Edition Copyright © Orchard Publications .5) and in general.8) then. the equation is referred to as a homogeneous difference equation. it is a non−homogeneous difference equation. that is. the linear combination k 1 y 1 ( x ) + k 2 y 1 ( x ) is also a solution. 11−2 Numerical Analysis Using MATLAB® and Excel®. y 2 ( x ) ] = y1 ( x ) Ey 1 ( x ) y2 ( x ) Ey 2 ( x ) ≠0 (11.3) is a linear combination of terms such as kx . If. Moreover. in (11.1).6) and if the right side is zero. cos kx . if.10) where ϕ ( x ) ≠ 0 . We find Y ( x ) by the Method of Undetermined Coefficients. if C [ y 1 ( x ). if Y ( x ) is any solution of (11. and x . it reduces to ( a2 E + a1 E + a0 ) y = 0 2 (11.7). is non−zero. ϕ ( x ) = 0 . analogous to the Wronskian determinant in ordinary differential equations. if the Casorati determinant. Also. we let r = 2 .9) where k 1 and k 2 are constants. then. For the non−homogeneous difference equation ( a 2 E + a 1 E + a 0 )y = φ ( x ) 2 (11. we obtain the second order difference equation ( a2 E + a1 E + a0 ) y = φ ( x ) 2 (11.4) (11. in (11. and if ϕ ( x ) ≠ 0 .10).11) As with ordinary differential equations. where k is a non−zero constant and n is a non−negative integer. E f ( x ) = f ( x + r ) = fx + r n r As with ordinary differential equations.1).7) can be expressed as y3 ( x ) = k1 y1 ( x ) + k2 y2 ( x ) (11. the right side of (11.

As with algebraic quadratic equations.13) can be real and unequal.Definition. Third Edition Copyright © Orchard Publications 11−3 . compute y 5 = y ( 5 ) . real and equal.13) and this is the characteristic equation of a second order difference equation. Then. and Applications We have assumed that the coefficients a i in (11.12) By substitution into (11. or complex conjugates depending on whether the discriminant a 1 – 4a 2 a 0 is positive. Solutions. in analogy with the solution of the differential equation of the form y = ke . we obtain a2 M x+2 + a1 M x+1 + a0 M x = 0 (11. or negative.1. the roots of (11. TABLE 11.7). then. Solution: The characteristic equation of (11.1 Find the solution of the difference equation ( E – 6 E + 8 )y = 0 2 (11.1 Roots of the characteristic equation in difference equations Characteristic equation a 2 M + a 1 M + a 0 = 0 of ( a 2 E + a 1 E + a 0 )y = 0 2 2 2 Roots M 1 and M 2 Real and Unequal M1 ≠ M2 Real and Equal M1 = M2 Discriminant a 1 – 4a 2 a 0 > 0 2 General Solution x x y = k1 M1 + k2 M2 k 1 and k 2 cons tan ts a 1 – 4a 2 a 0 = 0 2 x x y = k 1 M 1 + k 2 xM 2 k 1 and k 2 cons tan ts Complex Conjugates M1 = α + j β M2 = α –j β a 1 – 4a 2 a 0 < 0 2 x y = r ( C 1 cos θx + C 2 sin θx ) r = α +β 2 2 –1 β θ = tan --α Example 11.14) with initial conditions y 0 = y ( 0 ) = 3 and y 1 = y ( 1 ) = 2 .14) is Numerical Analysis Using MATLAB® and Excel®. These cases are summarized in Table 11. zero.10) are constants. for the homogeneous difference equation we assume a solution of the form y = M x ax (11. and recalling that Ef ( x ) = f ( x + 1 ) .

20) is M + 2M + 4 = 0 2 (11.18) yields k 1 = 5 and k 2 = – 2 .2 Find the solution of the difference equation ( E + 2E + 4 ) y = 0 2 (11. we obtain y 5 = 5 ⋅ 2 – 2 ⋅ 4 = 5 × 32 – 2 × 1024 = – 1888 5 5 Example 11. Then. Therefore.20) Solution: The characteristic equation of (11.19) For x = 5 . Third Edition Copyright © Orchard Publications . we let x = 1 . the solution is –1 11−4 Numerical Analysis Using MATLAB® and Excel®.21) ( –1 ) + ( 3 ) = 2 2 2 and its roots are M 1 = – 1 + j 3 and M 2 = – 1 – j 3 .16) To make use of the first initial condition. From Table 11.16) becomes y1 = 2 = k1 2 + k2 4 1 1 or 2k 1 + 4k 2 = 2 (11. r = and θ = tan ( 3 ⁄ ( – 1 ) ) = 2π ⁄ 3 . Thus. the solution is yx = 5 ⋅ 2 – 2 ⋅ 4 x x (11.17) and (11.16) becomes y0 = 3 = k1 2 + k2 4 0 0 or k1 + k2 = 3 (11. (11.18) Simultaneous solution of (11.15) and its roots are M 1 = 2 and M 2 = 4 . Therefore. with reference to Table 11. Then. we let x = 0 .1.1.Chapter 11 Difference Equations M –6 M + 8 = 0 2 (11.17) For the second initial condition. we obtain the solution yx = y ( x ) = k1 2 + k2 4 x x (11. (11.

Third Edition Copyright © Orchard Publications 11−5 . and we apply the operator E .Definition. we combine the particular solution with the solution of the homogeneous equation shown in (11.x⎞ ⎝ 3 3 ⎠ (11. For the particular solution.22) The constants C 1 and C 2 can be evaluated from the initial conditions.23) The characteristic equation of (11.3 Find the solution of the difference equation Solution: ( E –5 E + 6 ) y = x + 2 2 x (11. φ ( x ) .24) Numerical Analysis Using MATLAB® and Excel®. Solutions.x + C 2 sin ----. TABLE 11. this choice must be multiplied by x until there is no duplication of terms.10).2 Form of the particular solution for a non−homogeneous difference equation Non−homogeneous difference equation ( a 2 E + a 1 E + a 0 )y = φ ( x ) φ (x) α (constant) α x (k = positive integer) αk x k k 2 Form of Particular Solution Y ( x ) A (constant) Ak x + Ak – 1 x Ak x k–1 + … + A1 x + A0 α cos mx or α sin m x α x l cos mx or α x l sin m x k x k x A 1 cos mx + A 2 sin mx ( Akx + Ak – 1 x k k k–1 + … + A 1 x + A 0 ) l cos mx + … + B 1 x + B 0 ) l sin m x x x + ( Bk x + Bk – 1 x k–1 Example 11.23) is M –5 M + 6 = 0 2 (11. Table 11. we start with a linear combination of all the terms of the right side. For non−homogeneous difference equations of the form of (11. and Applications x 2π 2π y = 2 ⎛ C 1 cos ----. If any of the terms in the initial choice duplicates a term in the solution of the homogeneous equation. that is.2 shows the form of the particular solution for different terms of φ ( x ) .11).

**Chapter 11 Difference Equations
**

and its roots are M 1 = 2 and M 2 = 3 . From Table 11.1, the solution Y H of the homogeneous difference equation is

YH = k1 2 + k2 3

x x

(11.25)

x x x,

For the particular solution we refer Table 11.2. For the first term x of the right side of (11.23), we use the term A 1 x + A 0 , or Ax + B . For the second term 2 , we obtain A2 or C2 particular solution has the form

Y P = Ax + B + C2

x x

and thus, the

(11.26)

But the term C2 in (11.26), is also a term in (11.25). Therefore, to eliminate the duplication, we multiply the term C2 by x . Thus, the correct form of the particular solution is

Y P = Ax + B + Cx2

x x

(11.27)

To evaluate the constants

A,

**B , and C , we substitute (11.27) into (11.23). Then,
**

x+2

[A(x + 2) + B + C(x + 2) ⋅ 2

m+n

] – 5[A(x + 1) + B + C(x + 1) ⋅ 2 ] x x + 6 [ Ax + B + Cx2 ] = x + 2

m n

x+1

(11.28)

Using the law of exponents W

= W

**× W , simplifying, and equating like terms, we obtain
**

x

2Ax + ( – 3A + 2B ) – 2C2

= x+2

x

(11.29)

**Relation (11.29) will be true if
**

2A = 1 – 3A + 2B = 0 B = 0.75 – 2C = 1 C = – 0.5

or

A = 0.5

**By substitution into (11.28), we obtain the particular solution
**

Y P = 0.5x + 0.75 – 0.5 x2

x

(11.30)

**Therefore, the total solution is the sum of (11.25) and (11.30), that is,
**

y total = Y H + Y P = k 1 2 + k 2 3 + 0.5x + 0.75 – 0.5 x2

x x x

(11.31)

11−6

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

**Fibonacci Numbers 11.3 Fibonacci Numbers
**

The Fibonacci numbers are solutions of the difference equation

yx + 2 = yx + 1 + yx

(11.32)

that is, in a series of numbers, each number after the second, is the sum of the two preceding numbers. Example 11.4 Given that y 0 = 0 and y 1 = 1 , compute the first 12 Fibonacci numbers. Solution: For x = 0, 1, 2, 3 and so on, we obtain the Fibonacci numbers

1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, 233, …

We will conclude this chapter with an application to electric circuit analysis. Example 11.5 For the electric network of Figure 11.1, derive an expression for the voltage V x at each point P x where x = 0, 1, 2, …, n , given that the voltage V 0 at point P 0 is known.

P0 R P1

+

P2 R V1 2R

+

P3 R 2R

+

Pn – 3

+

Pn – 2

+

Pn – 1 R

+

Pn R

+

+

V0

−

2R

−

−

V2

−

V3

2R

−

R V n – 3 2R

−

V n – 2 2R

−

Vn – 1

Vn

−

. Solution:

Figure 11.1. Electric network for Example 11.5

We need to derive a difference equation that relates the unknown voltage V x to the known voltage V 0 . We start by drawing part of the circuit as shown in Figure 11.2, and we denote the voltages and currents as indicated.

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

11−7

**Chapter 11 Difference Equations
**

Px

+

Px + 1

R 2R

+

Px + 2

R

+

2R

−

Vx

−

V x + 1 2R

−

Vx + 2

Figure 11.2. Part of the circuit of Figure 11.1

**By application of Kirchoff’s Current Law (KCL) at node P x + 1 of Figure 11.2, we obtain
**

Vx + 1 – Vx Vx + 1 Vx + 1 – Vx + 2 -------------------------- + ------------- + -------------------------------- = 0 R R 2R

(11.33)

**and after simplification,
**

2 --- ( V x + 2 – 2.5 V x + 1 + V x ) = 0 R

**Of course, the term 2 ⁄ R cannot be zero. Therefore, we must have
**

V x + 2 – 2.5 V x + 1 + V x = 0

(11.34)

Relation (11.34) is valid for all points except P 1 and P n – 1 * as shown in Figure 11.1; therefore, we must find the current relations at these two points. Also, by application of Kirchoff’s current law (KCL) at node P 1 of Figure 11.1, we obtain

V1 – V0 V1 V1 – V2 ------------------ + ------ + ------------------ = 0 R R 2R

**and after simplification,
**

V 2 – 2.5V 1 + V 0 = 0

(11.35)

**Likewise, at node P n – 1 of Figure 11.1, we obtain
**

Vn – 1 – Vn – 2 Vn – 1 Vn – 1 – Vn -------------------------------- + ------------ + ------------------------- = 0 2R R R

**Observing that V n = 0 , and simplifying, we obtain
**

2.5V n – 1 – V n – 2 = 0

(11.36)

* The voltages at nodes P 0 and P n are V 0 and Vn respectively.

11−8

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

Fibonacci Numbers

Relation (11.35) is a difference equation of the form

( E – 2.5E + 1 ) y = 0

2

**where y = V x . Its characteristic equation is
**

M – 2.5M + 1 = 0

2

(11.37)

**The roots of the characteristic equation of (11.37) are M 1 = 0.5 and M 2 = 2 . Thus, the solution is
**

y = V x = k 1 ( 0.5 ) + k 2 ( 2 )

x x

(11.38)

The constant coefficients k 1 and k 2 in (11.38), are found by substitution of this relation into (11.35) and (11.36). Thus, from (11.37) and (11.38), we obtain

k 1 ( 0.5 ) + k 2 ( 2 ) – 2.5 ( k 1 ( 0.5 ) + k 2 ( 2 ) ) + V 0 = 0

2 2 1 1

or

0.25k 1 + 4k 2 – 1.25k1 – 5k 2 + V 0 = 0

or

k1 + k2 = V0

(11.39)

**Likewise, from (11.38) and (11.36) we obtain
**

n – 1⎞ n–2 1 n–1 1 n–2 2.5 ⎛ k 1 ⎛ -- ⎞ + k2 ( 2 ) – k 1 ⎛ -- ⎞ – k2 ( 2 ) = 0 ⎝ ⎝2⎠ ⎠ ⎝2⎠

or

k1 2.5k 1 n–1 n–2 ------------ + 2.5k 2 ( 2 ) – ----------- – k 2 ( 2 ) = 0 n–2 n–1 2 2

or

2 ( 2.5 )k 1 2.5k 2 ( 2 ) 4k 1 k 2 ( 2 ) -------------------- + ----------------------- – ------- – --------------- = 0 n n 2 4 2 2 k1 n ----- + k 2 ( 2 ) = 0 n 2

n n

or

(11.40)

**Simultaneous solution of (11.39) and (11.40) yields
**

2 k 1 = ---------------- V 0 2n 2 –1

2n

–1 k 2 = ---------------- V 0 2n 2 –1

(11.41)

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

11−9

**Chapter 11 Difference Equations
**

Finally, substitution of (11.41) into (11.38) yields a solution of the difference equation in terms of V 0 , that is,

x 2 1 x –1 y = V x = ---------------- V 0 ⎛ -- ⎞ + ---------------- V 0 ( 2 ) ⎝2⎠ 2n 2n 2 –1 2 –1

2n

or

⎛ 2 2n x⎞ V 0 y = V x = ⎜ ------- – 2 ⎟ ---------------⎝ 2x ⎠ 22 n – 1

(11.42)

**We observe that when x = 0 ,
**

2n V0 2 y = V x = ⎛ ------- – 1⎞ ---------------- = V 0 ⎝ 1 ⎠ 2n 2 –1

and when x = n ,

⎛ 2n V0 n⎞ V 0 n n ------y = V x = ⎜ 2 - – 2 ⎟ ---------------- = ( 2 – 2 ) ---------------- = 0 2n 2n n ⎝2 ⎠2 – 1 2 –1

11−10

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

Summary 11.4 Summary

• The general form of a linear, constant coefficient difference equation has the form

( ar E + ar – 1 E

r r–1

+ … + a 1 E + a 0 )y = φ ( x )

where a k represents a constant coefficient and E is an operator similar to the D operator in ordinary differential equations. As with ordinary differential equations, the right side is a linear combination of terms such as kx , cos kx , and x , where k is a non−zero constant and n is a non−negative integer. If ϕ ( x ) = 0 , the equation is referred to as a homogeneous difference equation, and if ϕ ( x ) ≠ 0 , it is a non−homogeneous difference equation. • The difference operator E is

Ef ( x k ) = f ( x k + h ) = f ( x k + h )

n

**The interval h is usually unity, i.e., h = 1 , and the subscript k is normally omitted. Thus, (11.3) is written as
**

Ef ( x ) = f ( x + 1 ) = f x + 1

and in general,

E f ( x ) = f ( x + r ) = fx + r

r

• If y 1 ( x ) and y 2 ( x ) are any two solutions of a homogeneous difference equation, the linear combination k 1 y 1 ( x ) + k 2 y 1 ( x ) , where k 1 and k 2 are constants, is also a solution. • If the Casorati determinant, analogous to the Wronskian determinant in ordinary differential equations, is non−zero, that is, if

C [ y 1 ( x ), y 2 ( x ) ] = y1 ( x ) Ey 1 ( x ) y2 ( x ) Ey 2 ( x ) ≠0

**then, any other solution of the homogeneous difference equation can be expressed as
**

y3 ( x ) = k1 y1 ( x ) + k2 y2 ( x )

**• For the non−homogeneous difference equation
**

( a 2 E + a 1 E + a 0 )y = φ ( x )

2

**where ϕ ( x ) ≠ 0 , if Y ( x ) is any solution of (11.11), then the complete solution is
**

y = k1 y1 ( x ) + k2 y2 ( x ) + Y ( x )

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

11−11

**Chapter 11 Difference Equations
**

As with ordinary differential equations, we first find the solution of the homogeneous difference equation; then, we add the particular solution Y ( x ) to it to obtain the total solution. We find Y ( x ) by the Method of Undetermined Coefficients. • In analogy with the solution of the differential equation of the form y = ke , for the homogeneous difference equation, we assume a solution of the form

y = M

x ax

**• Since Ef ( x ) = f ( x + 1 ) , the characteristic equation of a second order difference equation is
**

a2 M

x+2

+ a1 M

x+1

+ a0 M

x

= 0

and as with algebraic quadratic equations, the roots can be real and unequal, real and equal, or complex conjugates depending on whether the discriminant a 1 – 4a 2 a 0 is positive, zero, or negative. These cases are summarized in Table 11.1. • For non−homogeneous difference equations we combine the particular solution with the solution of the homogeneous equation. For the particular solution, we start with a linear combination of all the terms of the right side, that is, φ ( x ) , and we apply the operator E . If any of the terms in the initial choice duplicates a term in the solution of the homogeneous equation, this choice must be multiplied by x until there is no duplication of terms. The form of the particular solution for different terms of φ ( x ) is shown in Table 11.2. • The Fibonacci numbers are solutions of the difference equation

yx + 2 = yx + 1 + yx

2

that is, in a series of numbers, each number after the second, is the sum of the two preceding numbers.

11−12

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

Exercises 11.5 Exercises

Find the total solution of the following difference equations. 1. ( E + 7E + 12 ) y = 0 2. ( E + 2E + 2 ) y = 0 3. ( E – E – 6 ) y = x + 3 4. ( E + 1 ) y = sin x

2 2 2 2

x

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

11−13

**Chapter 11 Difference Equations 11.6 Solutions to End−of−Chapter Exercises
**

1. The characteristic equation is

( E + 7E + 12 ) y = 0 M + 7M + 12 = 0

2 2

**and its roots are M 1 = – 3 and M 2 = – 4 . Therefore, with reference to Table 11.1, we obtain the solution
**

yx = y ( x ) = k1 ( –3 ) + k2 ( –4 )

x x

(1)

The constants k 1 and k 2 can be evaluated from the initial conditions. Since they were not given, let us assume that y 0 = y ( 0 ) = 1 and y 1 = y ( 1 ) = 2 . To make use of the first initial condition, we let x = 0 . Then, (1) becomes

y0 = 1 = k1 ( –3 ) + k2 ( –4 )

0 0

or

k 1 + k 2 = 1 (2)

**For the second initial condition, we let x = 1 . Then, (1) becomes
**

y1 = 2 = k1 ( –3 ) + k2 ( –4 )

1 1

or

– 3 k 1 – 4k 2 = 2 (3)

**Simultaneous solution of (2) and (3) yields k 1 = 6 and k 2 = – 5 . Thus, the solution is
**

yx = y ( x ) = 6 × ( – 3 ) –5 × ( –4 )

x x

**2. The characteristic equation is
**

( E + 2E + 2 ) y = 0 M + 2M + 2 = 0

2 2

**and its roots are M 1 = – 1 + j and M 2 = – 1 – j . From Table 11.1, r =
**

θ = tan 1 ⁄ ( – 1 ) = – π ⁄ 4 . Therefore, the solution is y = 2 ( C 1 cos ( – π ⁄ 4 )x + C 2 sin ( – π ⁄ 4 )x )

x

–1

( –1 ) + ( 1 ) =

2

2

2 and

11−14

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

**Solutions to End−of−Chapter Exercises
**

The constants C 1 and C 2 can be evaluated from the initial conditions. For this exercise, they were not given. 3. The characteristic equation is

(E – E – 6) y = x + 3 M –M – 6 = 0

2 2

x

**and its roots are M 1 = – 2 and M 2 = 3 From Table 11.1, the solution Y H of the homogeneous difference equation is
**

Y H = k 1 ( – 2 ) + k 2 3 (1)

x x

For the particular solution we refer Table 11.2. For the first term x of the right side of the given equation we use the term A 1 x + A 0 , or Ax + B . For the second term 3 , we obtain A3 or C3

x, x x

**and thus, the particular solution has the form
**

Y P = Ax + B + C3

x x x

But the term C3 is also a term in the given equation. Therefore, to eliminate the duplication, we multiply the term C3 by x . Thus, the correct form of the particular solution is

Y P = Ax + B + Cx3

x

(2)

To evaluate the constants equation. Then,

A,

**B , and C , we substitute the last expression above into the given
**

x+2 x+1

[A(x + 2) + B + C(x + 2) ⋅ 3

m+n

] –[ A ( x + 1 ) + B + C ( x + 1 ) ⋅ 3 ] x x – 6 [ Ax + B + Cx3 ] = x + 3

m

Using the law of exponents W obtain

= W

× W , simplifying, and equating like terms, we

n

– 6 Ax + ( A – 6 B ) + 15C3

x

= x+3

x

**This relation will be true if
**

– 6A = 1 A –6 B = 0 B = 1 ⁄ 36 15C = 1 C = 1 ⁄ 15

or

A = –1 ⁄ 6

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

11−15

**Chapter 11 Difference Equations
**

By substitution into (2), we obtain the particular solution

Y P = ( – 1 ⁄ 6 )x + 1 ⁄ 36 + ( 1 ⁄ 15 )x2

x

(3)

**Therefore, the total solution is the sum of (1) and (3), that is,
**

y total = Y H + Y P = k 1 ( – 2 ) + k 2 3 + ( – 1 ⁄ 6 )x + 1 ⁄ 36 + ( 1 ⁄ 15 )x2

x x x

**4. The characteristic equation is
**

( E + 1 ) y = sin x (1) M +1 = 0

2 2

**and its roots are M 1 = j and M 2 = – j From Table 11.1, r = of the solution is
**

( 1 ) = 1 and θ = tan 1 ⁄ 0 = π ⁄ 2 . Therefore, the homogeneous part Y H = C 1 cos ( π ⁄ 2 )x + C 2 sin ( π ⁄ 2 )x (2)

2 –1

For the particular solution we refer Table 11.2 where we find that the solution has the form A 1 cos mx + A 2 sin mx , and for this exercise

Y P = A cos x + B sin x

Since the cosine and sine terms appear in the complimentary solution, we multiply the terms of the particular solution by x and we obtain

Y P = Ax cos x + Bx sin x (3)

To evaluate the constants we obtain Using the trig identities

A,

B , and C , we substitute the last expression above into (1) and

A ( x + 2 ) cos ( x + 2 ) + B ( x + 2 ) sin ( x + 2 ) + Ax cos x + Bx sin x = sin x

cos ( a + b ) = cos acos b – sin asin b sin ( a + b ) = sin acos b – sin b cos a

**expanding, rearranging, equating like terms, and combining the complimentary and particular solutions we obtain
**

π π sin x + sin ( x – 2 ) y = C 1 cos -- x + C 2 sin -- x + ----------------------------------------2 2 2 ( 1 + cos 2 )

11−16

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

Chapter 12

Partial Fraction Expansion

his chapter is an introduction to partial fraction expansion methods. In elementary algebra we learned how to combine fractions over a common denominator. Partial fraction expansion is the reverse process and splits a rational expression into a sum of fractions having simpler denominators.

T

**12.1 Partial Fraction Expansion
**

The partial fraction expansion method is used extensively in integration and in finding the inverses of the Laplace, Fourier, and Z transforms. This method allows us to decompose a rational polynomial into smaller rational polynomials with simpler denominators, from which we can easily recognize their integrals or inverse transformations. In the subsequent discussion we will discuss the partial fraction expansion method and we will illustrate with several examples. We will also use the MATLAB residue(r,p,k) function which returns the residues (coefficients) r of a partial fraction expansion, the poles p and the direct terms k. There are no direct terms if the highest power of the numerator is less than that of the denominator. Let

----------F(s) = N(s) D(s)

(12.1)

**where N ( s ) and D ( s ) are polynomials and thus (12.1) can be expressed as
**

bm s + bm – 1 s + bm – 2 s + … + b1 s + b0 N(s) F ( s ) = ----------- = ------------------------------------------------------------------------------------------------------------------n n–1 n–2 D(s) an s + an – 1 s + an – 2 s + … + a1 s + a0

m m–1 m–2

(12.2)

The coefficients a k and b k for k = 0, 1, 2, …, n are real numbers and, for the present discussion, we have assumed that the highest power of N ( s ) is less than the highest power of D ( s ) , i.e., m < n . In this case, F ( s ) is a proper rational function. If m ≥ n , F ( s ) is an improper rational function. It is very convenient to make the coefficient a n of s in (12.2) unity; to do this, we rewrite it as

1 ---- ( b m s m + b m – 1 s m – 1 + b m – 2 s m – 2 + … + b 1 s + b 0 ) an N(s) F ( s ) = ----------- = -----------------------------------------------------------------------------------------------------------------------------a1 D(s) a0 n an – 1 n – 1 an – 2 n – 2 + ---------- s + … + ---- s + ---s + ---------- s an an an an

n

(12.3)

Numerical Analysis Using MATLAB® and Excel®, Third Edition Copyright © Orchard Publications

12−1

Case I: Distinct Poles If all the poles p 1.= --------------. ….4) where p k is distinct from all other poles.+ ----------------. that is. we can factor the denominator of F ( s ) in the form N(s) F ( s ) = ------------------------------------------------------------------------------------------------( s – p1 ) ⋅ ( s – p2 ) ⋅ ( s – p3 ) ⋅ … ⋅ ( s – pn ) (12.8) 12−2 Numerical Analysis Using MATLAB® and Excel®.4) as r2 r3 rn r1 F ( s ) = ----------------.6) r k = lim ( s – p k )F ( s ) = ( s – p k )F ( s ) s → pk s = pk Example 12. in most engineering applications we are interested in the nature of the poles. we multiply both sides of (12.7) Solution: r2 r1 3s + 2 3s + 2 F 1 ( s ) = ------------------------.+ ----------------. We will consider the nature of the poles for each case. the partial fraction expansion method allows us to express (12.5) by ( s – p k ) . then.+ … + ----------------( s – p1 ) ( s – p2 ) ( s – p3 ) ( s – pn ) (12.7) below. we let s → p k . Then.Chapter 12 Partial Fraction Expansion The roots of the numerator are called the zeros of F ( s ) .+ --------------2 (s + 1) (s + 2) (s + 1)(s + 2) s + 3s + 2 3s + 2 r 1 = lim ( s + 1 )F ( s ) = --------------(s + 2) s → –1 = –1 s = –1 (12. r 2. or combinations of real and complex conjugates. (12. p n of F ( s ) are distinct (different from each another). or repeated. p 3.3) can be real and distinct. To evaluate the residue r k . r n are the residues of F ( s ) . The zeros and poles of (12. and are found by letting N ( s ) = 0 in (12. However. ….= -------------------------------. or complex conjugates. The roots of the denominator are called the poles of F ( s ) and are found by letting D(s) = 0 . Third Edition Copyright © Orchard Publications . 3s + 2 F 1 ( s ) = ------------------------2 s + 3s + 2 (12. r 3.5) where r 1.3) . p 2.1 Use partial fraction expansion to simplify F 1 ( s ) of (12.

and k vectors. These expressions are * Of course. Numerical Analysis Using MATLAB® and Excel®. p. 2]. Ds = [1. For this example. The vector k is referred to as the direct term. this function is used with symbolic expressions. 2]. and direct terms respectively.10) below. 3s 2 + 2s + 5 F 2 ( s ) = -----------------------------------------------s 3 + 12s 2 + 44s + 48 (12.9) We can us the MATLAB residue(r. However. by substitution into (12. 3. The first value of the vector r is associated with the first value of the vector p.10) Solution: First. we obtain 4 3s + 2 –1 F 1 ( s ) = ------------------------. MATLAB displays the r.= --------------.2 Use partial fraction expansion to simplify F 2 ( s ) of (12. Ds) r = 4 -1 p = -2 -1 k = [] where we have denoted Ns and Ds as two vectors that contain the numerator and denominator coefficients of F 1 ( s ) . the direct term k is empty.Partial Fraction Expansion 3s + 2 r 2 = lim ( s + 2 )F ( s ) = --------------(s + 1) s → –2 = 4 s = –2 Therefore. Example 12. poles.p.+ --------------2 (s + 1) (s + 2) s + 3s + 2 (12. p.k) function to verify our answers with the following script: Ns = [3. we observe that the highest power of the denominator is s 2 whereas the highest power of the numerator is s and therefore. [r. the second value of r is associated with the second value of p. The factor(s) function is a good alternative. Third Edition Copyright © Orchard Publications 12−3 . we can use the roots(p) function. we will use the MATLAB function factor(s) to express the denominator polynomial of F 2 ( s ) in factored form. these represent the residues.8). and so on. k] = residue(Ns. and it is always empty (has no value) whenever F ( s ) is a proper rational function.* This function returns an expression that contains the prime factors of a polynomial.

For example.Chapter 12 Partial Fraction Expansion explained below. which we have used before. Third Edition Copyright © Orchard Publications . a1 and k2 .dx x 1 MATLAB contains the so-called Symbolic Math Toolbox. like roots(p).2 and using MATLAB we have: syms s. on the other hand. Some examples of symbolic expressions are given below.+ --------------2 3 (s + 2)(s + 4)(s + 6) (s + 2) (s + 4) (s + 6) s + 12s + 44s + 48 Next. can manipulate mathematical expressions without using actual numbers.= --------------. s . z . are display numeric expressions. y .+ --------------.( 3t – 4t + 5t + 8 ) 2 dt 2 u = ∫ -. 2 2 r1 r2 r3 3s + 2s + 5 3s + 2s + 5 F 2 ( s ) = -----------------------------------------------. sin x 2 e – αt d.+ --------------2 3 (s + 2) (s + 4) (s + 6) s + 12s + 44s + 48 2 12−4 Numerical Analysis Using MATLAB® and Excel®. syms x y z a1 k2 defines the symbolic variables x .= --------------. s = sym (‘s’) creates the symbolic variable s . we find the residues r 1 .= ------------------------------------------------. Returning to Example 12. For example. The functions.3 2 y = -----. factor(den) ans = (s+4)*(s+2)*(s+6) and thus. that is. our interest is in using the factor(s) to express the denominator of (12. we can use the syms function to define one or more symbolic variables with a single statement. These are 3s + 2s + 5 r 1 = -------------------------------( s + 4 )( s + 6) 2 s = –2 9 = -8 3s + 2s + 5 r 2 = -------------------------------(s + 2)(s + 6) 2 s = –4 37 =– ----4 3s + 2s + 5 r 3 = -------------------------------(s + 2)(s + 4) 2 s = –6 89 = ----8 Therefore. For the present. 9⁄8 – 37 ⁄ 4 89 ⁄ 8 3s + 2s + 5 F 2 ( s ) = -----------------------------------------------. This is done with the sym function.10) as a product of simple factors. we must create a symbolic variable x . t etc.+ --------------. they are discussed in detail in MATLAB User’s Manual. This is a collection of tools (functions) which are used in solving symbolic expressions. Before using symbolic expressions. r 2 . den=s^3+12*s^2+44*s+48. Alternately. they produce numerical results. and r 3 . y . Symbolic expressions.

we express the denominator in factored form to identify the poles of F 3 ( s ) . s+3 F 3 ( s ) = -----------------------------------------3 s + 5s 2 + 12s + 8 (12. roots_p=roots(p) roots_p = -2.= --------------. the poles of a proper rational function F ( s ) are complex.0000i Then. factor(s^3 + 5*s^2 + 12*s + 8) we obtain ans = (s+1)*(s^2+4*s+8) Since the factor(s) function did not factor the quadratic term*.11) below.11) Solution: As a first step.0000-2.0000+2. Numerical Analysis Using MATLAB® and Excel®. Thus if p k is a complex pole.0000i -2.12) and the residues are * For some undocumented reason. and since complex poles occur in complex conjugate pairs.Partial Fraction Expansion Case II: Complex Poles Quite often. we will use the roots(p) function to find its roots by treating it as a polynomial. Example 12.3 Use partial fraction expansion to simplify F 3 ( s ) of (12. Third Edition Copyright © Orchard Publications 12−5 .= -----------------------------------------------------------------------2 3 ( s + 1 ) ( s + 2 + j2 ) ( s + 2 – j2 ) s + 5s + 12s + 8 r1 r2 r3 s+3 = -----------------------------------------. the number of complex poles is even. The partial fraction expansion method can also be used in this case. as illustrated by the following example. the factor(s) function does not seem to work with complex numbers. p=[1 4 8].+ -----------------------2 3 ( s + 1 ) ( s + 2 + j2 ) ( s + 2 – j 2 ) s + 5s + 12s + 8 (12. s+3 s+3 F 3 ( s ) = -----------------------------------------.+ --------------------------. then its complex conjugate p k∗ is also a pole. Using the MATLAB script syms s.

Third Edition Copyright © Orchard Publications .+ j -----⎞ = – -.– j ----⎝ 5 20⎠ 5 20 and this is always true since complex roots occur in conjugate pairs.14) Case III: Multiple (Repeated) Poles In this case.= ----------------– 8 – j4 ( – 1 + j2 ) ( j4 ) 3) ( 1 – j2 ) .---------------------.13) We can express (12. has a multiplicity m .+ ----------------------------------. by substitution into (12.F 3 ( s ) = --------------. say p 1 . Then.⋅ -----------------------------( s + 2 ) 5 ( s 2 + 4s + 8 ) (12.13) in a different form if we want to eliminate the complex presentation.---------------------. we obtain 2⁄5 – 1 ⁄ 5 + j3 ⁄ 20 – 1 ⁄ 5 – j3 ⁄ 20 F 3 ( s ) = --------------. the partial fraction expansion of (12. This is done by combining the last two terms on the right side of (12.15) and denoting the m residues corresponding to multiple pole p 1 as r 11.+ ---------------------------------(s + 2) ( s + 2 + j2 ) ( s + 2 –j 2 ) (12.= ( – 16 ) – j 12 = – 1 – j ----------------------------( – 8 – j4 ) ( – 8 + j4 ) 5 20 80 Of course. F ( s ) has simple poles but one of the poles.( – 8 + j4 = ---------------------.= – -.15) can be expressed as 12−6 Numerical Analysis Using MATLAB® and Excel®. … r 1m .= -----------------( – 1 – j2 ) ( – j4 ) – 8 + j4 s = –1 s+3 r 2 = ----------------------------------------( s + 1 ) ( s + 2 –j 2 ) s = – 2 – j2 3 ( 1 – j2 ) ( – 8 – j4 ) – 16 + j12 1 = ---------------------. r 12.– -. Then. the last evaluation was not necessary since r 3 = r 2∗ or 3 * 3 1 1 r 3 = ⎛ – -.+ j ----( – 8 + j4 ) ( – 8 – j4 ) 80 5 20 s+3 r 3 = -------------------------------------------( s + 1 ) ( s + 2 + j2 ) s = – 2 + j2 1 – j2 1 – j2 = --------------------------------.= -----------------------.13) to form a single term and now is written as 2⁄5 1 ( 2s + 1 ) .12).Chapter 12 Partial Fraction Expansion s+3 r 1 = ------------------------2 s + 4s + 8 2 = -5 1 – j2 1 – j2 = -----------------------------------. N(s) F ( s ) = ------------------------------------------------------------------------------------------m ( s – p 1 ) ( s – p 2 )… ( s – p n – 1 ) ( s – p n ) (12.

+ ---------------------------. then.18) r3 rn r2 m + lim ( s – p 1 ) ⎛ ----------------. that is. then.⎞ ⎝(s – p ) (s – p ) ( s – p n )⎠ 2 3 (12. … p n we proceed as before. we let s → p 1 . we multiply both sides of (12. To find the second residue r 12 of the second repeated pole p 1 . we obtain s → p1 lim ( s – p 1 ) F ( s ) 2 m–1 s → p1 m = r 11 + lim [ ( s – p 1 )r 12 + ( s – p 1 ) r 13 + … + ( s – p 1 ) r 1m ] (12. that is.20) To find the third residue r 13 of the repeated pole p 1 . r k = lim ( s – p k )F ( s ) = ( s – p k )F ( s ) s → pk s = pk m To find the first residue r 11 of the repeated pole.+ ----------------.17) Next. m d r 12 = lim ---.16) For the simple poles p 1.21) Numerical Analysis Using MATLAB® and Excel®.17).+ … + ----------------m m–1 m–2 ( s – p1 ) ( s – p1 ) ( s – p1 ) ( s – p1 ) r3 rn r2 + ----------------.18) with respect to s .Partial Fraction Expansion r 11 r 12 r 13 r 1m F ( s ) = --------------------. Then. p 2. m d r 13 = lim ------. we differentiate (12.[ ( s – p 1 ) F ( s ) ] s → p 1 ds (12.⎞ ⎝(s – p ) (s – p ) ( s – p n )⎠ s → p1 2 3 or r 11 = lim ( s – p 1 ) F ( s ) s → p1 m (12. Third Edition Copyright © Orchard Publications 12−7 .+ ---------------------------.+ ----------------. ( s – p 1 ) F ( s ) = r 11 + ( s – p 1 )r 12 + ( s – p 1 ) r 13 + … + ( s – p 1 ) m 2 m–1 r 1m r2 r3 rn m + ( s – p 1 ) ⎛ ----------------.18) twice with respect to s .+ ----------------. taking the limit as s → p 1 on both sides of (12.[ ( s – p1 ) F ( s ) ] s → p 1 ds 2 2 (12.+ … + ----------------.16) by ( s – p 1 ) . that is.+ ----------------( s – p2 ) ( s – p3 ) ( s – pn ) (12. we let s → p 1 .19) and thus (12.+ … + ----------------.19) yields the residue of the first repeated pole. we first differentiate the relation of (12.

for repeated poles the residue r 1k can be derived from the relation ( s – p 1 ) F ( s ) = r 11 + r 12 ( s – p 1 ) + r 13 ( s – p 1 ) + … m 2 (12.= --------------.26) The residues are s+3 r 1 = -----------------2 (s + 1) s + 3r 21 = --------------(s + 2) = 1 s = –2 = 2 s = –1 d. by substitution into (12.26).⎛ ---------. Third Edition Copyright © Orchard Publications .+ --------------2 ( s + 2 ) ( s + 1 )2 ( s + 1 ) (s + 2)(s + 1) 12−8 Numerical Analysis Using MATLAB® and Excel®. s+3 1 2 –1 F 4 ( s ) = ----------------------------------. s+3 F 4 ( s ) = ----------------------------------2 (s + 2)(s + 1) (12.25) in partial fraction expansion form is r 21 r1 r 22 s+3 F 4 ( s ) = ----------------------------------.24) Example 12.------------.+ -----------------.+ -----------------.= --------------.22) whose ( m – 1 )th derivative of both sides is d m ( k – 1 )!r 1k = lim ------------.23) or d m 1 r 1k = lim -----------------.+ --------------2 2 (s + 2) (s + 1) (s + 1) (s + 2)(s + 1) (12.25) Solution: We observe that there is a pole of multiplicity 2 at s = – 1 and thus.4 Use partial fraction expansion to simplify F 4 ( s ) of (12.25) below. (12.Chapter 12 Partial Fraction Expansion This process is continued until all residues of the repeated poles have been found.⎞ ds ⎝ s + 2 ⎠ s = –1 (s + 2) – (s + 3) = -------------------------------------2 (s + 2) = –1 s = –1 Then.[ ( s – p 1 ) F ( s ) ] s → p 1( k – 1 )! ds k – 1 k–1 (12.s + 3 r 22 = ---.[ ( s – p 1 ) F ( s ) ] k–1 s → p 1 ds k–1 (12. In general.

% Coefficients of (s + 2) term in D(s) Ds=conv(d1. we will use the expand(s) function. % Multiplies polynomials d1 and d2 to express denominator D(s) as polynomial [r. To check our answers with MATLAB. it is true for any value of s.p.0000 2. % Coefficients of the numerator N(s) d1 = [1 2 1].d2).5 Use partial fraction expansion to simplify F 5 ( s ) of (12. Its description can be displayed with the help expand command. *We must remember that (2.Ds) r = 1.27) below.0000 -1. Like the factor(s) function. expand(s) is used with symbolic expressions. Numerical Analysis Using MATLAB® and Excel®. s+3 ----------------------------------2 (s + 1) (s + 2) 1 = --------------(s + 2) 2 + -----------------2 (s + 1) s=0 r 22 + --------------(s + 1) s=0 s=0 s=0 or 3 ⁄ 2 = 1 ⁄ 2 + 2 + r 22 from which r 22 = – 1 as before.0000 p = -2. the residue r 22 could be found by substitution of the already known values of r 1 and r 21 into (12. and letting s = 0 *.0000 -1.Partial Fraction Expansion Instead of differentiation.0000 k = [] Example 12.k]=residue(Ns.0000 -1.45) is an identity. Third Edition Copyright © Orchard Publications 12−9 .26). % Coefficients of (s + 1)^2 = s^2 + 2*s + 1 term in D(s) d2 = [0 1 2]. and as such. Check with MATLAB: syms s expand((s + 1)^2) % Create symbolic variable s % Express it as a polynomial ans = s^2+2*s+1 Ns = [1 3]. that is.

the residue r 21 if found by evaluating F 5 ( s ) at s = – 2 .⎞ 2 ds ⎝ ( s + 2 ) 3 ⎠ s = –1 1 (s + 2) – 3(s + 2) (s + 4) = -.⎛ s + 3 s + 1⎞ r 12 = ---.⎛ -----------------.⎟ 2! ds 2 ⎝ ( s + 2 ) 2 ⎠ 2 1 d d ⎛ s + 3 s + 1⎞ .30) s = –1 ( s + 2 ) ( 2s + 3 ) – 2 ( s + 2 ) ( s + 3 s + 1 ) = --------------------------------------------------------------------------------------------4 (s + 2) s+4 = -----------------.⎜ -------------------------⎟ 2 ds ds ⎝ ( s + 2 ) 2 ⎠ 3 2 s = –1 s = –1 1 d.+ --------------.------.Chapter 12 Partial Fraction Expansion s2 + 3 s + 1 F 5 ( s ) = ------------------------------------( s + 1 )3 ( s + 2 )2 (12.---.---.29) The residue r 12 is found by first taking the first derivative of F 5 ( s ) .= 3 3 (s + 2) The residue r 13 is found by taking the second derivative of F 5 ( s ) and evaluating it at s = – 1 .+ -----------------.⎜ ------------------------.---. Then. Thus.⎜ ------------------------. Then.r 13 = ---.⎛ ---------------------------------. 12−10 Numerical Analysis Using MATLAB® and Excel®. Third Edition Copyright © Orchard Publications . and the residue r 22 is found by first taking the first derivative of F 5 ( s ) and evaluating it at s = – 2 .+ -----------------. 2 d.⎟ ds ⎝ ( s + 2 ) 2 ⎠ 2 s = –1 2 (12.+ --------------3 2 ( s + 1 ) ( s + 2 )2 ( s + 2 ) (s + 1) (s + 1) (12.31) 1 s + 2 – 3s – 12 = -. Therefore. 2 2 1 d ⎛ s + 3 s + 1⎞ . and a pole of multiplicity 2 at s = – 2 .28) We find the residue r 11 by evaluating F 5 ( s ) at as s = – 1 r 11 = s + 3 s + 1 ------------------------2 (s + 2) 2 = –1 s = –1 (12.--------------------------------------------------------------6 2 (s + 2) –s–5 = -----------------4 (s + 2) = –4 s = –1 (12..= -. in partial fraction expansion form r 21 r 11 r 12 r 13 r 22 F 5 ( s ) = -----------------.⎞ 4 ⎠ 2⎝ (s + 2) s = –1 Similarly. and evaluating it at s = – 1 .s + 4 = -.27) Solution: We observe that there is a pole of multiplicity 3 at s = – 1 .

.e.0000 -4.Ds) r = 4. i.0000 Numerical Analysis Using MATLAB® and Excel®. we introduce the collect(s) function that we can use to multiply two or more symbolic expressions to obtain the result in a polynomial form. polynomial coefficients only.⎟ ds ⎝ ( s + 1 ) 3 ⎠ 3 s = –2 2 2 = ( s + 1 ) ( 2s + 3 ) – 3 ( s + 1 ) ( s + 3 s + 1 ) -------------------------------------------------------------------------------------------------6 (s + 1) ( s + 1 ) ( 2s + 3 ) – 3 ( s + 3 s + 1 ) r 22 = ---------------------------------------------------------------------------4 (s + 1) 2 s = –2 2 s = –2 – s – 4s = -------------------4 (s + 1) = 4 s = –2 By substitution of these residues into (12.+ --------------.k]=residue(Ns.+ --------------3 2 ( s + 1 ) ( s + 2 )2 ( s + 2 ) (s + 1) (s + 1) (12. [r.+ -----------------.0000 3.Partial Fraction Expansion s + 3s + 1 r 21 = ------------------------3 (s + 1) 2 = 1 s = –2 2 d ⎛ s + 3 s + 1⎞ r 22 = ---. We must remember that the conv(p.+ -----------------.q) function is used with numeric expressions.0000 -1. Before we do this. The MATLAB script for this example is as follows. Its description can be displayed with the help collect command. Ds=[1 7 19 25 16 4]. Third Edition Copyright © Orchard Publications 12−11 .⎜ ------------------------. syms s.32) We will now verify the values of these residues with MATLAB. % We must first define the variable s in symbolic form % The function "collect" below multiplies (s+1)^3 by (s+2)^2 Ds=collect(((s+1)^3)*((s+2)^2)) Ds = s^5+7*s^4+19*s^3+25*s^2+16*s+4 % We now use this result to express the denominator D(s) as a % polynomial so we can use its coefficients with the "residue" function % Ns=[1 3 1].28). we obtain F 5 ( s ) in partial fraction expansion as –1 3 –4 1 4 F 5 ( s ) = -----------------.p.0000 1.

i. the highest power m of the numerator is less than the highest power n of the denominator.0000 -1. Example 12. m > n and thus we need to express F 6 ( s ) in the form of (12.+ s + 1 s+1 s+1 2 (12. we must divide the numeraror N ( s ) by the denominator D ( s ) to obtain an expression of the form F ( s ) = k0 + k1 s + k2 s + … + km – n s 2 m–n N(s) + ----------D(s) (12.Chapter 12 Partial Fraction Expansion p = -2. Ds = [1 1]. m < n .e. Third Edition Copyright © Orchard Publications 12−12 . p.0000 -1. and before we apply the partial fraction expansion..33) so that m < n . [r.34) Solution: In (12. Ds) r = 1 p = -1 k = 1 1 Numerical Analysis Using MATLAB® and Excel®.33). 1 s + 2s + 2 F 6 ( s ) = ------------------------. was based on the condition that F ( s ) is a proper rational function.34) below in partial expansion form.35) Check with MATLAB: Ns = [1 2 2].6 Express F 6 ( s ) of (12.0000 -1. By long division. k] = residue(Ns.= ---------.0000 -2.34). s + 2s + 2 F 6 ( s ) = ------------------------s+1 2 (12.0000 k = [] Case for m ≥ n Our discussion thus far. F ( s ) is an improper rational function. that is. If m ≥ n .

we assign the sum of n partial fractions as shown below. r2 s + k2 r1 s + k1 rn s + kn -------------------------. We arrange the terms of both sides in decreasing powers of s .37) both sides equal.7 Express F 7 ( s ) of (12.+ ----------------. We set the given F ( s ) equal to the sum of these partial fractions. Then. We solve the resulting equations for the residues. To this factor. we can express F ( s ) as rm r1 r2 ----------F ( s ) = N ( s ) = ---------. 4. If not. We multiply each term of the right side by the appropriate factor to make the denominators of (12.+ … -----------------2 D(s) s – a (s – a) ( s – a )m n m (12. 3. Numerical Analysis Using MATLAB® and Excel®.54). We also assume that the denominator D ( s ) can be expressed as a product of real linear and quadratic factors. 5. 6.38) below as a sum of partial fractions using the equating the numerators procedure. we perform the following steps: 1.+ --------------------------------. Now. We equate the coefficients of corresponding powers of s.2 Alternate Method of Partial Fraction Expansion The partial fraction expansion method can also be performed by the equating the numerators procedure thereby making the denominators of both sides the same. We assume that the degree on the numerator N ( s ) is less than the degree of the denominator. We repeat Step 1 for each of the distinct linear and quadratic factors of D ( s ) . and then equating the numerators. let s 2 + αs + β be a quadratic factor of D ( s ) and suppose that ( s 2 + αs + β ) is the highest power of this factor that divides F ( s ) . If these assumptions prevail.36) Next.+ … + --------------------------------2 2 n 2 s + αs + β ( s 2 + αs + β ) ( s + αs + β ) 2. 7. we first perform a long division and then work with the quotient and the remainder as before. Third Edition Copyright © Orchard Publications 12−13 . Example 12. 12. we let s – a be a linear factor of D ( s ) and we suppose that ( s – a ) is the highest power of s – a that divides D ( s ) .Alternate Method of Partial Fraction Expansion The direct terms k = [ 1 1 ] are the coefficients of the s term and the constant in (2.

43) and by substitution into the first equation of (12. therefore.43) and (12.42).+ ----------------.45) and using the fourth equation of (12.42.41) is an identity in s . – 2s + 4 = ( r 1 + r 22 )s 3 + ( – 2r 1 + A – r 22 + r 21 )s 2 + ( r 1 – 2A + r 22 ) s + ( A – r 22 + r 21 ) (12. Accordingly. we obtain 4 = 2r 1 or r 1 = 2 (12. the coefficients of each power of s on the left and right sides are equal.39) By Step 4.+ --------------2 2 2 2 (s – 1) (s + 1)(s – 1) (s + 1) (s – 1) (12.44) Next.41) Relation (12.42). r 21 r1 s + A r 22 – 2s + 4 F 7 ( s ) = ------------------------------------.42) Subtracting the second equation from the fourth in (12. yields – 2 = 2 – 2A – 2 or A = 1 (12. 6.40) and by Steps 5. we obtain: 4 = 1 + 2 + r 21 or r 21 = 1 (12.+ ----------------. we obtain 0 = 2 + r 22 or r 22 = – 2 (12.– --------------2 2 2 2 (s – 1) (s + 1)(s – 1) (s + 1) (s – 1) 12−14 Numerical Analysis Using MATLAB® and Excel®.38) Solution: By Steps 1 through 3 above.= -----------------. – 2s + 4 = ( r 1 s + A ) ( s – 1 ) 2 + r 21 ( s + 1 ) + r 22 ( s – 1 ) ( s + 1 ) 2 2 (12. we obtain 0 = r 1 + r 22 0 = – 2r 1 + A – r 22 + r 21 – 2 = r 1 – 2A + r 22 4 = A – r 22 + r 21 (12.= -----------------.46) Therefore F 7 ( s ) in partial fraction expansion form becomes 1 – 2s + 4 2s + 1 2 F 7 ( s ) = ------------------------------------.42). Third Edition Copyright © Orchard Publications .44) into the third equation of (12. substitution of (12. by equating like powers of s .Chapter 12 Partial Fraction Expansion – 2s + 4 F 7 ( s ) = ------------------------------------2 2 (s + 1)(s – 1) (12. and 7.

50) s = –1 To compute r 2 and r 3 . that is. To find the third residue r 3 .52) With r 1 = 2 ⁄ 5 . by substitution into (12. Third Edition Copyright © Orchard Publications 12−15 .= ---------. we equate the constant terms of (12.k) function.+ ------------------------2 2 s + 1 s + 4s + 8 ( s + 1 ) ( s + 4s + 8 ) (12.49). Numerical Analysis Using MATLAB® and Excel®. Equating the coefficient of s 2 on the left side.Alternate Method of Partial Fraction Expansion Example 12.52) yields r 2 = – 2 ⁄ 5 . using the residue(r. we use the equating the numerators procedure and we obtain ( s + 3 ) = r 1 ( s + 4s + 8 ) + ( r 2 s + r 3 ) ( s + 1 ) 2 (12.– -. s+3 F 7 ( s ) = ----------------------------------------------2 ( s + 1 ) ( s + 4s + 8 ) (12. we obtain r 3 = – 1 ⁄ 5 . We will conclude the partial fraction expansion topic with a few more examples. 3 = 8r 1 + r 3 .47) below. we first find the residue of the linear factor as s+3 r 1 = ------------------------2 + 4s + 8 s 2 = -5 (12.48) and in partial fraction expansion form.49) As in Example 12. with the coefficients of s 2 on the right side of (12. where we found that the denominator can be expressed in factored form of a linear and a quadratic factor.8 Use the equating the numerators procedure to obtain the partial fraction expansion of F 8 ( s ) in (12. we obtain 2 ⁄ 5 1 ( 2s + 1 ) .p.3.47) Solution: This is the same rational function as that of Example 12. that is. r2 s + r3 r1 s+3 F 7 ( s ) = ----------------------------------------------.-----------------------------( s + 2 ) 5 ( s 2 + 4s + 8 ) as before. s+3 F 8 ( s ) = -----------------------------------------2 3 s + 5s + 12s + 8 (12.51).51) Since r 1 is already known.F 7 ( s ) = --------------. and with r 1 = 2 ⁄ 5 . which is zero. The remaining steps are the same as in Example 12.3. (12. Then. we only need two equations in r 2 and r 3 .51).3. we obtain 0 = r1 + r2 (12.

Then.= ---------. [r.+ ------------s + p1 s + p2 (12.9 Use the residue(r.+ ---------s + p1 s + p2 s+2 s+1 (12. r 2 . it can be written as 12−16 Numerical Analysis Using MATLAB® and Excel®.and r 3 be the residues of F 10 ( s ) . Third Edition Copyright © Orchard Publications .54) The MATLAB script for this example is as follows: num=[0 8 2].p. % The semicolon suppress the display of the row vector typed % and zero is typed to make the numerator have same number % of elements as the denominator.56) Solution: Let p 1 .10 Use the residue(r.55) Example 12.56) below.p. s+3 F 10 ( s ) = ----------------------------------------------2 ( s + 1 ) ( s + 4s + 8 ) (12.Chapter 12 Partial Fraction Expansion Example 12. and p 3 be the poles (the denominator roots) and r 1 . F 9 ( s ) can be written as r2 r1 F 9 ( s ) = ------------.p. but recommended den=[1 3 2].+ ------------.53) Solution: Let p 1 and p 2 be the poles (the denominator roots) and r 1 and r 2 be the residues. not necessary.k) function to compute the poles and residues of the function 8s + 2 F 9 ( s ) = ------------------------2 s + 3s + 2 (12. p 2 . F 9 ( s ) in partial fraction expansion form is written as r2 r1 14 –6 F 9 ( s ) = ------------. Then.den) r = 14 -6 p = -2 -1 k = [] Therefore.k]=residue(num.k) function to compute the poles and residues of F 10 ( s ) in (12.

– 0.3].b) to multiply the two factors of the denominator of (12.0000i Therefore.0000+ -2. c.2 – 0.15j + -----------------------------.57) The poles and the residues can be found with the statement [r.den) function. we can reduce a rational polynomial to simple terms of a polynomial.k]=residue(num. [r.+ ------------s + p1 s + p2 s + p3 (12. whereas semicolons will suppress the display.15j 0.59) Numerical Analysis Using MATLAB® and Excel®. b=[1 4 8]. Third Edition Copyright © Orchard Publications 12−17 .2 + 0.2000-0. Then.p.b).1500i 2.2000+ 0. Example 12. num=[1.+ ------------.+ ---------s+1 s + 2 + 2j s + 2 – 2j s + p1 s + p2 s + p3 (12.11 Use the deconv(num.58) By repeated use of the deconv(num. den=c.+ ------------.den) c = 1 r = -0.0000-1. a=[1 1].1500i 0.+ ------------. we need to express the denominator as a polynomial.5x – 1 3 2 (12. Before we use this statement. den).den) function to express the following rational polynomial as a polynomial with four terms.4F 10 ( s ) = ------------.4000 p = -2. We will use the function conv(a. c=conv(a.0000 k = [] 5 12 8 0. x + 2x + 1 f 1 ( x ) = ----------------------------0. We also recall that commas will display the results.p. We recall that we can write two or more statements on one line if we separate them by commas or semicolons. F 10 ( s ) in partial fraction expansion form is r2 r3 r1 ----------------------------.56).Alternate Method of Partial Fraction Expansion r1 r2 r3 F 10 ( s ) = ------------.0000i 2.= – 0.k]=residue(num. where the last term is a rational polynomial whose order of the numerator is less than that of the denominator as illustrated by the following example.

12−18 Numerical Analysis Using MATLAB® and Excel®.3x + 4. Page 1−27.5x ) 3 2 5 2 (12.61) we must use the function fzero(‘function’. such as the function f 2 ( x ) defined as 3x + 7x + 9 0.den) q = 2 r = 0 0 0 17 2 17 f 1 ( x ) = 2x + 8x + 16 + ------------------0. den=[0 0 0.5x – 1 8 16 Therefore. [q.+ --------------------------------------------------.5x + 6.11x + 2. f 1 ( x ) can now be written as (12.r]=deconv(num. and x0 is a required initial value. where function is a pre−defined string.60) It is important to remember that the function roots(p) is used with polynomials only.75 ( 12x + 2x + 13x + 25 ) ( 23x + 16x + 7. We can approximate this value by first plotting f 2 ( x ) to find out where it crosses the x −axis. Third Edition Copyright © Orchard Publications . If we want to find the zeros of any function.Chapter 12 Partial Fraction Expansion Solution: num=[1 2 0 1].5 −1].x0).35 1 f 2 ( x ) = ------------------------------------------------------------. This is discussed in Chapter 1.+ ----------------------------6 4 2 6 3 4.

p 2. • If all the poles p 1.e. n are real numbers. Third Edition Copyright © Orchard Publications 12−19 . and are found by letting N ( s ) = 0 . • If the function bm s + bm – 1 s + bm – 2 s + … + b1 s + b0 N(s) F ( s ) = ----------.= -----------------------------------------------------------------------------------------------------------------------------a1 D(s) a0 n an – 1 n – 1 an – 2 n – 2 s + ---------.s + … + ---.( b s m + b m – 1 s m – 1 + b m – 2 s m – 2 + … + b 1 s + b 0 ) an m N(s) F ( s ) = ----------. …. In most engineering applications we are interested in the nature of the poles.s + ---an an an an to make a n unity.= ------------------------------------------------------------------------------------------------------------------n n–1 n–2 D(s) + an – 2 s + … + a1 s + a0 an s + an – 1 s m m–1 m–2 where the coefficients a k and b k for k = 0. we rewrite this function as 1---. If m ≥ n . • The roots of the numerator are called the zeros of F ( s ) . 2.3 Summary • The function + bm – 2 s + … + b1 s + b0 bm s + bm – 1 s N(s) F ( s ) = ----------. or repeated. p n of F ( s ) are distinct we can factor the denominator of F ( s ) in the form Numerical Analysis Using MATLAB® and Excel®. If F ( s ) is an improper rational function we divide the numeraror N ( s ) by the denominator D ( s ) to obtain an expression of the form F ( s ) = k0 + k1 s + k2 s + … + km – n s 2 m–n + N(s) ----------D(s) so that m < n . …. F ( s ) is an improper rational function.. • The zeros and poles can be real and distinct.= ------------------------------------------------------------------------------------------------------------------n n–1 n–2 D(s) + an – 2 s + … + a1 s + a0 an s + an – 1 s m m–1 m–2 is a proper rational function where a n is a non−zero integer other than unity.Summary 12. is a proper rational function if the highest power of the numerator N ( s ) is less than the highest power of of the denominator D ( s ) . 1. or complex conjugates. • Partial fraction expansion applies only to proper rational functions.s + ---------. or combinations of real and complex conjugates. and the roots of the denominator are called the poles of F ( s ) and are found by letting D(s) = 0 . m < n . p 3. i.

• The partial fraction expansion can also be used if the poles are complex. r k = lim ( s – p k )F ( s ) = ( s – p k )F ( s ) s → pk s = pk • We can use the MATLAB residue(r. Third Edition Copyright © Orchard Publications .+ ----------------. we let s → p k . r 12. Then.+ ----------------. To evaluate the residue r k . then. their associated poles.k) function to verify our answers. the partial fraction expansion method allows us to write the above expression as r2 r3 rn r1 F ( s ) = ----------------. that is. if p k is a complex pole. then its complex conjugate p k∗ is also a pole. • If a rational function F ( s ) has simple poles but one of the poles.+ ----------------. the function is expressed as N(s) F ( s ) = ------------------------------------------------------------------------------------------m ( s – p 1 ) ( s – p 2 )… ( s – p n – 1 ) ( s – p n ) and denoting the m residues corresponding to multiple pole p 1 as r 11.5) by ( s – p k ) .+ ---------------------------.+ … + ----------------m m–1 m–2 ( s – p1 ) ( s – p1 ) ( s – p1 ) ( s – p1 ) r3 rn r2 + ----------------. r 2.+ ----------------( s – p2 ) ( s – p3 ) ( s – pn ) • If a rational function F ( s ) has simple poles but one of the poles.p. For proper rational functions there is no direct term.+ … + ----------------( s – p1 ) ( s – p 2 ) ( s – p3 ) ( s – pn ) where r 1. say p 1 . has a multiplicity m . the partial fraction expansion can be expressed as r 12 r 13 r 1m r 11 F ( s ) = --------------------. has a multiplicity m . say p 1 . and a direct term.Chapter 12 Partial Fraction Expansion N(s) F ( s ) = ------------------------------------------------------------------------------------------------( s – p1 ) ⋅ ( s – p 2 ) ⋅ ( s – p 3 ) ⋅ … ⋅ ( s – p n ) where p k is distinct from all other poles.+ ---------------------------. …. This function returns the residues. r n are the residues of F ( s ) . for the simple poles we use the same procedure as for distinct poles. Since complex poles occur in conjugate pairs. r 3. … r 1m . The first residue of a repeated pole is found from r 11 = lim ( s – p 1 ) F ( s ) s → p1 m The second repeated pole is found from 12−20 Numerical Analysis Using MATLAB® and Excel®. we multiply both sides of (12.

[ ( s – p 1 ) F ( s ) ] s → p 1 ds 2 2 and this process is continued until all residues of the repeated poles have been found. Numerical Analysis Using MATLAB® and Excel®.Summary m dr 12 = lim ---. We assume that the denominator D ( s ) can be expressed as a product of real linear and quadratic factors. and then equating the numerators. • With the alternate method of partial fraction expansion we use the equating the numerators procedure thereby making the denominators of both sides the same. Third Edition Copyright © Orchard Publications 12−21 .[ ( s – p 1 ) F ( s ) ] s → p 1 ds the third from m d r 13 = lim ------.

4 Exercises Perform partial fraction expansion for the following. Use MATLAB to simplify and to verify your results. ------------------------2 s + 2s + 1 s(s + 1) 1 6. ---------------------2 s –2 s – 3 2 s 5. ------------------------2 s + 4s – 5 s 3. ----------------------------------2 1 8. ---------2 1 –s 1 2. ----------------------------2 s(s + s + 1) (s + 1)(s + 1 ) 12−22 Numerical Analysis Using MATLAB® and Excel®. --------------------2 1 7. 11.Chapter 12 Partial Fraction Expansion 12. Third Edition Copyright © Orchard Publications . ---------------------2 s –2 s – 3 5s – 3 4.

−5]. Third Edition Copyright © Orchard Publications 12−23 .+ ---------2 2 (s + 1)(s – 1) s+1 s–1 s –1 1 –s –1 r 1 = ---------s+1 = –1 ⁄ 2 s=1 –1 r 2 = ---------s–1 = 1⁄2 s = –1 Then.+ ---------2 s–1 s+5 (s – 1)(s + 5) s + 4s – 5 1r 1 = ---------s+5 = 1⁄6 s=1 1r 2 = ---------s–1 = –1 ⁄ 6 s = –5 Then. 1 ⁄ 6 1 ⁄ 61 ------------------------. Ds = [1. 0. [r. 1]. –1 1⁄2 1⁄2 -----------. −1].= ---------. 0.= ---------. 4.– ---------2 s–1 s+5 s + 4s – 5 format rat.= ---------. Ds) r = 0.= – ---------.+ ---------2 s+1 s–1 s –1 Ns = [0. p. −1]. Ns = [0. 0. Ds = [1.= -------------------------------. Ds) r = -1/6 1/6 p = -5 1 k = [] Numerical Analysis Using MATLAB® and Excel®.5000 p = -1 1 k = [] 2.5000 -0.= -----------.5 Solutions to End−of−Chapter Exercises 1. k] = residue(Ns.= -------------------------------. p. k] = residue(Ns.Solutions to End−of−Chapter Exercises 12. r1 r2 –1 –1 1---------. [r. r2 r1 1 1 ------------------------.

= ---------. k] = residue(Ns. r1 r2 s s ---------------------. Ds = [1.= ---------. p. 0].Chapter 12 Partial Fraction Expansion 3. k] = residue(Ns.+ ---------2 s+1 s–3 s –2 s – 3 Ns = [0. Ds = [1. 1. 5. −3]. [r. s 1⁄4 3⁄4 ---------------------.= -------------------------------. Third Edition Copyright © Orchard Publications . r2 r1 5s – 3 5s – 3 ---------------------.= ---------.+ ---------2 ( s + 1 )( s – 3) s+1 s–3 s –2 s – 3 s r 1 = ---------s–3 = 1⁄4 s = –1 s r 2 = ---------s+1 = 3⁄4 s=3 Then. −2. −2. p. Ds) r = 3 2 p = 3 -1 k = [] 12−24 Numerical Analysis Using MATLAB® and Excel®.= -------------------------------. Ds) r = 3/4 1/4 p = 3 -1 k = [] 4. −3]. [r. −3].+ ---------2 s+1 s–3 ( s + 1 )( s – 3) s –2 s – 3 5s – 3 r 1 = ------------s–3 = 2 s = –1 5s – 3 r 2 = ------------s+1 = 3 s=3 Then.= ---------.+ ---------2 s+1 s–3 s –2 s – 3 format rat. Ns = [0. 325s – 3 ---------------------.

= --. Ds) 2 r = -2 1 p = -1 -1 k = 1 6. Ds = [1. we must divide the numeraror N ( s ) by the denominator D ( s ) to obtain an expression of the form F ( s ) = k0 + k1 s + k2 s + … + km – n s 2 m–n + N(s) ----------D(s) We could perform long division but we will use the MATLAB deconv(num. 0.Solutions to End−of−Chapter Exercises 5. 1]. r 21 r1 r 22 1 -------------------.+ -----------------. Third Edition Copyright © Orchard Publications 12−25 .= 1 – -----------------. 2.= -----------------.+ --------------2 2 2 2 (s + 1) s + 2s + 1 (s + 1) s + 2s + 1 (s + 1) -2 -1 r1 r2 –2 s – 1 -----------------.den) function to express the following rational polynomial as a polynomial with four terms. [q. [r.+ --------------2 2 (s + 1) (s + 1) (s + 1) r2 = –2 –2 s – 1 = r1 + r2 ( s + 1 ) r1 = 1 r1 + r2 = –1 Then. den=[1 2 1].r]=deconv(num.+ --------------2 s ( s + 1 )2 ( s + 1 ) s(s + 1) Numerical Analysis Using MATLAB® and Excel®.= 1 + -----------------.= 1 + ------------------------. k] = residue(Ns. 0].+ --------------2 2 2 (s + 1) s + 2s + 1 s + 2s + 1 (s + 1) Ns = [1. This is an improper rational function.= 1 + ------------------------. and before we apply the partial fraction expansion.den) q = 1 r = 0 and thus 2 r1 r2 s –2 s – 1 2s + 1 ------------------------.= 1 – -----------------. p. s –2 –2 s – 1 1 ------------------------. num=[1 0 0 ].

Third Edition Copyright © Orchard Publications . p. k] = residue(Ns.= ----------------------------------. eq2.⎛ -. 1]. 0.Chapter 12 Partial Fraction Expansion 1r 1 = ---------s+1 = 1 s=0 -r 21 = 1 s = –1 s = –1 d. 1. Ds) r = -1 -1 1 p = -1 -1 0 k = [] 7. [r.= -.+ -------------------------------------. Ds = [1. 0].= ---------.1 r 22 = ---.+ ----------------.⎞ ds ⎝ s ⎠ s = –1 1= – --2 s = –1 s = –1 Then. eq3) r2 + r3 = 0 r1 + r3 = 1 2 2 eq1 = r1+r2 eq2 = 12−26 Numerical Analysis Using MATLAB® and Excel®. 2. 0.+ --------------2 s ( s + 1 )2 ( s + 1 ) s(s + 1) syms s.+ -----------------.2 2 2 2 s + 1 s2 + 1 (s + 1)(s + 1) (s + 1 )(s + 1) (s + 1)(s + 1 ) 2 Equating numerators and like terms we obtain 1 = r1 s + r1 + r2 s + r2 s + r3 s + r3 r1 + r2 = 0 syms r1 r2 r3 eq1=r1+r2−0 eq2=r2+r3−0 eq3=r1+r3−1 S=solve(eq1. r1 r1 ( s + 1 ) r s + r3 ( r2 s + r3 ) ( s + 1 ) 1 ----------------------------------. 1 1 –1 –1 -------------------. expand(s*(s+1)^2) ans = s^3+2*s^2+s Ns = [0.

Thus. 1 1 ⁄ 2 ( – 1 ⁄ 2 )s + 1 ⁄ 2 ----------------------------------. [r. eq2.r2 ans = -1/2 S. k] = residue(Ns. Ds) r = 1/2 -1/4 . r 2 = – 1 ⁄ 2 .. 0.1/4i -1/4 + 1/4i p = -1 -1/6004799503160662 + 1i -1/6004799503160662 . 0. eq3.r1 ans = 1/2 S. 1].. Third Edition Copyright © Orchard Publications 12−27 . p.= ---------.1i k = [] Numerical Analysis Using MATLAB® and Excel®.r3 ans = 1/2 The statement S=solve(eq1.+ -----------------------------------2 2 s+1 (s + 1 )(s + 1 ) s +1 syms s. Ds = [1. 1. 1.Solutions to End−of−Chapter Exercises r2+r3 eq3 = r1+r3-1 S = r1: [1x1 sym] r2: [1x1 sym] r3: [1x1 sym] S.eqN) returns the solutions in the structure S whose named fields hold hold the solution for each variable. r 1 = 1 ⁄ 2 . Then. expand((s+1)*(s^2+1)) ans = s^3+s+s^2+1 Ns = [0. . 1]. and r 3 = 1 ⁄ 2 .

+ ---------------------.= ------------------------------. r 2 = – 1 . 8. 1. Ds) r = -1/2 -1/2 1 p = -1/2 + 1170/1351i -1/2 . can make arbitrarily large changes in the resulting poles and residues. Problem formulations making use of state-space or zero-pole representations are preferable. p.= 1 + ---------------------2 s s2 + s + 1 (s + 1 )(s + 1 ) syms s. Ds = [1. and r 3 = 1 . 0. 1].+ ----------------------------2 2 2 2 s s +s+1 s(s + s + 1) s(s + s + 1) s(s + s + 1) 2 Equating numerators and like terms we obtain 1 = r1 s + r1 s + r1 + r2 s + r3 s r1 + r2 = 0 r2 + r3 = 0 r1 = 1 2 2 By inspection. the partial fraction expansion of a ratio of polynomials represents an ill-posed problem. then small changes in the data. these values are inconsistent with those we’ve found. [r. The MATLAB help residue command displays the following: Warning: Numerically. 0.= --. Then. k] = residue(Ns.1170/1351i 0 k = [] As in Exercise 7. including roundoff errors. ( r 2 s + r 3 )s r2 s + r3 r1 r1 ( s + s + 1 ) 1 ----------------------------. A(s). If the denominator polynomial. –s+3 1 -----------------------------------. is near a polynomial with multiple roots. r 1 = 1 . Third Edition Copyright © Orchard Publications . expand(s*(s^2+s+1)) ans = s^3+s^2+s Ns = [0. 1.Chapter 12 Partial Fraction Expansion These values are inconsistent with those we’ve found. + 390/1351i 390/1351i 12−28 Numerical Analysis Using MATLAB® and Excel®. 0].

2) We will evaluate the integral of (13. ∫ f ( x ) dx where the limits of integration a or b or both are infinite b. with (13. 13. we write (13. It is defined as Γ(n) = ∫0 x ∞ n – 1 –x e dx (13.1 The Gamma Function The gamma function. We will derive the basic properties of the gamma function and its relation to the well known factorial function n! = n ( n – 1 ) ( n – 2 )…3 ⋅ 2 ⋅ 1 (13. a a b b Numerical Analysis Using MATLAB® and Excel®.Chapter 13 The Gamma and Beta Functions and Distributions T his chapter is an introduction to the gamma and beta functions and their distributions used with many applications in science and engineering.5) we obtain –x xdu = – e dx and v = ---n n Then.1) by performing integration by parts using the relation ∫ u dv Letting u = e –x = uv – v du and dv = x n–1 ∫ (13.1) as * Improper integrals are two types and these are: a.3). denoted as Γ ( n ) .4) (13. and in the computation of certain integrals. Third Edition Copyright © Orchard Publications 13−1 . ∫ f ( x ) dx where f ( x ) becomes infinite at a value x between the lower and upper limits of integration inclusive. They are also used in probability.3) (13.1) and this improper* integral converges (approaches a limit) for all n > 0 . is also known as generalized factorial function.

. that is.= 0 ------------------------------------------------------------------m–n x x→∞ x e n–m x m n d m–1 m–1 nx n–1 (13.f ( x ) ⁄ ----. This can be proved with L’ Hôpital’s rule* by differentiating both numerator and denominator m times. L’Hôpig( a) 0 x → a g(x) ddtal’s rule states that if f ( a ) = g ( a ) = 0 .Chapter 13 The Gamma and Beta Functions and Distributions x e Γ ( n ) = -----------n n –x ∞ 1 + -n x=0 ∫0 x e ∞ n –x dx (13.6) vanishes at the lower limit. the first term on the right side of (13. dx dx x→a ddf(x) lim ---------. and we wish to find this limit.= -.10) or nΓ ( n ) = Γ ( n + 1 ) (13. for some value of x . where m ≥ n .6) With the condition that n > 0 .g ( x ) as x approaches a exists. results in the indeterminate form g( x) f(a) 0 f(x ) ---------. then.= lim ⎛ ----. we observe that ) -------------------Γ(n) = Γ(n + 1n (13. the ratio of two functions.f ( x ) ⁄ ----.1) we have Γ(n) = ∫0 ∞ x n – 1 –x 1 e dx = -n ∫0 x e n –x dx (13. such as ---------.= … x m–1 n x → ∞ ne x x → ∞ dm x→∞ d x ne ne m m–1 dx dx n ( n – 1 ) ( n – 2 )… ( n – m + 1 )x = lim -----------------------------------------------------------------------------------x x→∞ ne ) = lim ( n – 1 ) ( n – 2 )… ( n – m + 1 . we consider the limit lim ---------. To work around this problem..Then. for x = 0 . (13.= lim ------. if it exists. d n –x n x→∞ m x e dx dx x lim -----------. It also vanishes at the upper limit as x → ∞ .9) By comparing the two integrals of (13.= lim ----------------------------------.= lim -----------------.8) and with (13.9)..g ( x )⎞ ⎠ dx g(x) x → a ⎝ dx 13−2 Numerical Analysis Using MATLAB® and Excel®.11) f(x) * Quite often. Third Edition Copyright © Orchard Publications .6) reduces to 1 Γ ( n ) = -n ∫0 x e ∞ n –x dx ∞ (13. and if the limit ----. say a .7) Therefore.

but if we substitute the numbers 0. 2. – 2. as defined in that relation. we have derived the important relation.5 ) in terms of Γ ( 0.11) for n > 0 . whereas the factorial n! is defined only for zero (recall that 0! = 1 ) and positive integer values. the gamma function exists (is continuous) everywhere except at 0 and negative integer numbers.13) From the recurring relation of (13. axis([−4 4 −6 6]).5 . n=−4: 0. title('The Gamma Function'). For instance. xlabel('n'). Third Edition Copyright © Orchard Publications 13−3 . Numerical Analysis Using MATLAB® and Excel®.14) and in general Γ ( 4 ) = 3 ⋅ Γ ( 3 ) = 3 ⋅ 2 = 3! Γ ( n + 1 ) = n! for n = 1.15) The formula of (13. This is done with the script below which produces the plot shown in Figure 13. From (13.The Gamma Function It is convenient to use (13. – 1.5 ) . and for all negative fractional. We must remember that. we see that Γ ( n ) becomes infinite as n → 0 . we can find Γ ( – 0.12) Thus. ylabel('Gamma(n)') Figure 13. (13. that is. it establishes the relationship between the Γ ( n ) function and the factorial n! . 3. Stated in other words. and so on. but not negative integer values.11). we obtain values which are not consistent with the definition of the Γ ( n ) function.05: 4.1 shows the plot of the function Γ ( n ) versus n .15) is a very useful relation.g). … (13.11). the Γ ( n ) function is defined for all positive integers and positive fractional values. – 1. when n = – 0.1. – 3 . g=gamma(n). plot(n. Γ(1) = 1 (13.1) yields Γ(1) = ∞ –x –x ∞ 0 ∫0 e dx = – e = 1 (13.10) for n < 0 . – 3 and so on in (13. – 2. and (13.10). For n = 1 . we obtain Γ(2) = 1 ⋅ Γ(1) = 1 Γ ( 3 ) = 2 ⋅ Γ ( 2 ) = 2 ⋅ 1 = 2! (13. grid. We can use MATLAB’s gamma(n) function to plot Γ ( n ) versus n .

5 + 1 ) Γ ( 0.5 13−4 Numerical Analysis Using MATLAB® and Excel®.6 ) ( 1. and from math tables Therefore. but we can use (13. From (13.6 )Γ ( 1.6 ) = ( 2. Of course.Chapter 13 The Gamma and Beta Functions and Distributions 6 The Gamma Function 4 2 Gamma(n) 0 -2 -4 -6 -4 -3 -2 -1 0 n 1 2 3 4 Figure 13.717 Then. we can use MATLAB to find any valid values of n .5 ) Γ ( 0.6 ) Γ ( 1.8953 Γ ( 3.10) or (13. Γ ( – 0.11) Then.10) Γ(n + 1) Γ ( n ) = -------------------n a. b.6 ) = 2. Γ ( 3.5 ) = -----------------------. Γ ( 1.6 ) ( 1. From (13.5 ) Γ ( n + 1 ) = nΓ ( n ) Γ ( 3.= --------------0.6 ) = ( 2.6 ) b.11) to compute values outside this range.1. Γ ( 0.6Γ ( 2.8953 ) = 3. Example 13. Third Edition Copyright © Orchard Publications . Plot of the gamma function Numerical values of Γ ( n ) for 1 ≤ n ≤ 2 .5 0.6 ) = 0.6 ) ( 0.5 ) c.1 Compute: Solution: a. can be found in math tables.

5 ) – 0.5 – 0.5 ) = 0.5 gammaln(x) 1.5) a = 3.= – 2Γ ( 0.10) Γ(n + 1) Γ ( n ) = -------------------n Γ ( 1. Therefore. Using Excel to find Γ ( n ) Numerical Analysis Using MATLAB® and Excel®.2. it does not produce an answer if n is negative as shown in Figure 13. Third Edition Copyright © Orchard Publications 13−5 . however.2.544 We can verify these answers with MATLAB as follows: a=gamma(3. c.7725 #NUM! Figure 13.5 -0. b=gamma(0.The Gamma Function and from math tables Therefore.772 Then. we can use the EXP(GAMMALN(n)) function to evaluate Γ ( n ) at some positive value of n .8862 ) = 1. c=gamma(−0.5 ) = Γ ( – 0.5 . From (13. ) ) ------------------------------------------Γ ( – 0. But because it first computes the natural log.5 + 1 . Γ ( – 0.8862 Γ ( 0.= Γ ( 0. x 3.5 and using the result of (b).5724 #NUM! exp(gammaln(x))= gamma(x) 3.772 ) = – 3.5 ) = ( 2 ) ( 0.7725 c = -3.3129 0.5 ) = ( – 2 ) ( 1.5449 Excel does not have a function which evaluates Γ ( n ) directly.6 0.6).7170 1. have the GAMMALN(x) function. It does.7170 b = 1.5).5 ) = – 2 Γ ( 0.

21) and since Γ ( 1 ) = 1 .1).20) By n repeated substitutions.18) Next.19) into (13.18).2 Prove that when n is a positive integer.24) Then.17) (13.23) Example 13. Γ ( n + 1 ) = nΓ ( n ) Γ ( n ) = ( n – 1 )Γ ( n – 1 ) (13. we have or Γ ( n ) = ( n – 1 ) ( n – 2 ) ( n – 3 )…1 Γ ( n ) = ( n – 1 )! (13.16) is true. Γ(n) = ∫0 x ∞ n – 1 –x e dx ∞ (13.Chapter 13 The Gamma and Beta Functions and Distributions Example 13. the relation Γ ( n ) = ( n – 1 )! (13. we obtain Γ ( n ) = ( n – 1 ) ( n – 2 ) ( n – 3 )…1Γ ( 1 ) (13.11).25) 13−6 Numerical Analysis Using MATLAB® and Excel®.5 – 1 – x e dx = ∫0 x – 0. replacing n with n – 1 on the left side of (13. Third Edition Copyright © Orchard Publications .⎞ = ⎝2⎠ ∫0 ∞ x 0. we obtain Γ ( n – 1 ) = ( n – 2 )Γ ( n – 2 ) (13.19) Substitution of (13.3 Use the definition of the Γ ( n ) function to compute the exact value of Γ ( 1 ⁄ 2 ) Solution: From (13. Then.18) yields Γ ( n ) = ( n – 1 ) ( n – 2 )Γ ( n – 2 ) (13. 1 Γ ⎛ -. Proof: From (13.22) (13.5 – x e dx (13.

⎞ ⎝2⎠ 2 = 4 ∫0 ∞ e –x 2 dx ∫0 ∞ e –y dy = 4 ∫0 ∫0 e ∞ ∞ –( x + y ) 2 2 dx dy (13.26) Next.⎞ = 2 ⎝2⎠ 1 Γ ⎛ -.31) 2. from differential calculus 2 2 d d 2 ----.= 2y dy or dx = 2ydy By substitution of the last three relations into (13.30) and by recalling that: 1.25).e u = e u ----.The Gamma Function Letting x = y2 we obtain dx ----. we define Γ ( 1 ⁄ 2 ) as a function of both x and y .33) We observe that as x → ∞ and y → ∞ .u 2 = 2ue u du du (13. we let 1 Γ ⎛ -. ∫ρ ρ2 1 ρe –ρ 2 1 –ρ 2 dρ = – -. that is. we convert (13.27) (13.29) to polar coordinates by making the substitution ρ 2 = x +y 2 2 (13.5 ) – y e 2 2ydy = 2 ∫0 ∞ y ye –1 –y 2 dy = 2 ∫0 e ∞ –y 2 dy (13.⎞ = ⎝2⎠ ∫0 ∞ y 2 ( – 0.28) yields 1 Γ ⎛ -. we obtain 1 Γ ⎛ -.e 2 (13. Numerical Analysis Using MATLAB® and Excel®.⎞ = 2 ⎝2⎠ ∫0 ∫0 2 ∞ e –x 2 dx (13. Third Edition Copyright © Orchard Publications 13−7 .28) ∞ e –y 2 dy Multiplication of (13.32) Then.29) Now. the total area of a region is found by either one of the double integrals A = ∫ ∫ dx dy = ∫ ∫ r dr dθ (13.27) by (13.

34) into (13.5 ) –2 π Γ ( – 1.5 ) Γ (n + 1) Γ ( n ) = --------------------.= --------.5 ) b .5 ) = -----------------------------. Γ ( – 0.Chapter 13 The Gamma and Beta Functions and Distributions ρ → ∞ and θ → π ⁄ 2 (13.= ------------------.5 ) π Γ ( – 0.= -.33) and (13. for n = – 2.= – ----.30).= ---------.π – 2.5 ) = n π b.5 a.5 + 1 ) Γ ( – 1. Third Edition Copyright © Orchard Publications . 4 Γ ( – 1.= – 2 π – 0. (13.36) -Γ ( n )Γ ⎛ n + 1 ⎞ = ⎝ 2 ⎠ πΓ ( 2n ) (13.5 .⎞ = ⎝2⎠ π (13. for n = – 1.4 Compute: Solution: Using the relations we obtain: a.37) for any n ≠ negative integer 13−8 Numerical Analysis Using MATLAB® and Excel®.5 15 – 2.= ------------------.5 . Γ ( – 1.π 3 – 1.5 c.5 ) = -----------------------------. for n = – 0.5 – 1.= ------------.5 – 2. Γ ( – 2.34) Substitution of (13.5 Other interesting relations involving the Γ ( n ) function are: π Γ ( n )Γ ( 1 – n ) = ------------sin nπ for 0 < n < 1 2 2n – 1 (13.⎞ ⎝2⎠ 2 = –2 ∫0 π⁄2 ⎛ e–ρ ⎝ 2 ⎞ dθ = – 2 ρ = 0⎠ ∞ ∫0 π⁄2 ( 0 – 1 ) dθ = 2 ∫0 π⁄2 dθ = 2 θ π⁄2 0 = π and thus.5 – 1.29) yields 1 Γ ⎛ -.5 ) Γ ( – 2.and Γ ( 0.5 .5 – 0.π 3 8 Γ ( – 2.5 ) c.5 ) = ---------------. Γ ( 0.5 + 1 ) Γ ( – 0.35) Example 13. 4 -. we have obtained the exact value 1 Γ ⎛ -.

⎞ ⎝2⎠ 2 = π Therefore.36) to prove that 1 Γ ⎛ -. 1 Γ ⎛ -.⎞ Γ ⎛ -.5 Use (13. the factorial n! can be approximated as n! ≈ 2πnn e n –n (13.– ------------------.+ ------------. If n is a positive integer.⎞ = ⎝2⎠ π Example 13.⎞ = ---------⎝ 3⎠ ⎝ ⎠ 3 sin π -3 or Numerical Analysis Using MATLAB® and Excel®.38) Relation (13. we obtain 1 1 π Γ ⎛ -.The Gamma Function Γ ( n + 1 ) = n! = ⎫ n –n ⎧ 1 1139 571 2πnn e ⎨ 1 + -------.+ … ⎬ 12n 288n 2 51840n 3 2488320n 4 ⎩ ⎭ (13.⎞ Γ ⎛ -.⎞ Γ ⎛ 1 – -.– ------------------------.38) is referred to as Stirling’s asymptotic series for the Γ ( n ) function.⎞ = ---------⎝ 2 ⎠ ⎝ 2⎠ ⎠ ⎝2⎠ ⎝ π 2 sin -2 or 1 Γ ⎛ -.39) Example 13.⎞ = ⎝2⎠ π Proof: 1 1 1 π-Γ ⎛ -.36).⎞ ⎝3⎠ ⎝3⎠ Solution: Using (13.⎞ Γ ⎛ 1 – 1 ⎞ = Γ ⎛ -. Third Edition Copyright © Orchard Publications 13−9 .6 Compute the product 1 2 Γ ⎛ -.

To find the exact value.⎞ = -----------------.⎞ ⎝2⎠ Solution: 2 3–1 3 3 1 Γ ⎛ -. we use the relation Γ ( n + 1 ) = n! and the MATLAB gamma(n) function.= ----------. gamma(50+1) ans = 3. =FACT(50) and Excel displays 3. we type and execute the expression sqrt(2*pi*50)*50^50*exp(−50) ans = 3.Chapter 13 The Gamma and Beta Functions and Distributions 1 2 2π 2 3π π Γ ⎛ -.⎞ = .7 Use (13.= -----⎝2⎠ 4Γ ( 2 ) 4⋅1 2 Example 13.0363e+064 This is an approximation. that is.= ------------⎝ 3⎠ ⎝ 3⎠ 3 3⁄2 3 Example 13.⎞ Γ ⎛ -. Third Edition Copyright © Orchard Publications .= -----. that is.0414e+064 We can check this answer with the Excel FACT(n) function.+ -.04141E+64 13−10 Numerical Analysis Using MATLAB® and Excel®.39) to compute 50! Solution: 50! ≈ 2 π × 50 × 50 50 ×e – 50 We use MATLAB as a calculator.⎞ Γ ⎛ -.⎝ 2 ⎠ ⎝2 2 ⎠ 3 2 2 Γ ⎛ -.37) to find 3 Γ ⎛ -. Then.⎞ Γ ( 2 ) = ⎝2⎠ 3 πΓ ⎛ 2 ⋅ -.⎞ ⎝ 2⎠ πΓ ( 3 ) 3 2! π πΓ(3) π Γ ⎛ -.⎞ = ------------.8 Use (13.

Some examples follow. then.The Gamma Function The Γ ( n ) function is very useful in integrating some improper integrals. It is initially at rest. and by substitution.= ----64 64 64 8 Example 13. Solution: Let the center of attraction 0 be the point zero on the x −axis. Assuming that the particle moves towards the center of the positively charged side. and then moves towards the positively charged side with a force inversely proportional to its distance from it. evaluate the integrals a.= ----. as indicated in Figure 13.9 Using the definition of the Γ ( n ) function. a. Example 13.10 A negatively charged particle is α meters apart from the positively charged side of an electric field. Numerical Analysis Using MATLAB® and Excel®. ∞ n – 1 –x e dx = Γ ( n ) ∫0 b. Third Edition Copyright © Orchard Publications 13−11 .= -------. ∞ ∞ x e dx = Γ ( 5 ) = 4! = 24 4 –x Let 2x = y . 4 –x ∫0 ∞ x e 5 – 2x dx ∫0 x Then. derive an expression for the time required the negatively charged particle to reach 0 in terms of the distance α and its mass m . ∫0 x e 5 – 2x dx = ∫0 ∞ 5 1 ⎛ y ⎞ e – y dy = --------6 ⎝2⎠ 2 2 ∫0 ∞ y e dy 5 –y Γ ( 6 ) 5! 120 15 = ----------. dx = dy ⁄ 2 . Solution: By definition.3. ∫0 ∞ x e dx b. considered to be the center of attraction 0 .

44).42) into (13. Then. dx ----.40) yields dv -mv ----.40) where m = mass of particle x = distance (varies with time) k = positive constant of proportionality and the minus (−) sign indicates that the distance x decreases as time t increases.Chapter 13 The Gamma and Beta Functions and Distributions movement of particle 0 Figure 13. the particle is assumed to be located on the x −axis at point x = α . 13−12 Numerical Analysis Using MATLAB® and Excel®. Third Edition Copyright © Orchard Publications . and moves towards the origin at x = 0 . Then.10 α x By Newton’s law.= v dt 2 dv dv dx dx dv .45) where C represents the constants of integration of both sides. Sketch for Example 13. Let the velocity of the particle be v .3. At t = 0 .41) and (13.= v ----2 dx dx dt dt dt (13.44) Integrating both sides of (13.( dx ) x (13. and it is evaluated from the initial condition that v = 0 when x = α .------. we obtain mv --------.= – -2 x dt 2 (13.= – k ln x + C 2 2 (13. dx k m ------.= ----.42) Substitution of (13.= ----.43) or (13.----.= – k dx x k mvdv = – -.

T = ∫ m dτ = – ----2k 0 t ∫α ------------------------ln ( α ⁄ x ) y 0 dx (13.46) and by substitution into (13.51).----m x dt (13.⎞ = ∞ ⎝ x⎠ x→0 ⎝ x⎠ –y (13. We denote this time as T .52) or Also.51) as Numerical Analysis Using MATLAB® and Excel®. it is found from the relation below.45).50) We are interested in the time required for the particle to reach the origin 0 .51). x = 0 . then e ⎝x⎠ α α = -x –y (13. that is. mv α --------.51) To simplify (13.⎞ .47) Solving for v and taking the square root of both sides we obtain 2k α dx .49) for dt we obtain dx mdt = – ----.48) Since x decreases as t increases. are being replaced with 0 and ∞ respectively. we let y = ln ⎛ -.53) x→α the lower and upper limits of integration in (13. Therefore. and at τ = t .⎞ = 0 and lim ln ⎛ -. noting that the integration on the right side is with respect to the distance x where at t = 0 .α .-----------------------2k ln ( α ⁄ x ) (13. Third Edition Copyright © Orchard Publications 13−13 .v = ----. we express (13. x = α .= ± ----.The Gamma Function C = k ln α (13. Then.ln -m x dt 2 (13.= k ln α – k ln x = k ln -2 x 2 (13. we choose the negative sign. since x = α e . dx = – 2k ln -----. and dx = – α e dy α α lim ln ⎛ -.49) Solving (13.

Next. with these substitutions. we obtain du = 2xdx and dv = 2ydy .Chapter 13 The Gamma and Beta Functions and Distributions m T = – ----2k ∫0 ∞ – α e dy m -------------------. we obtain 1 m πm m T = α Γ ⎛ -.= α ------⎝ 2 ⎠ 2k 2k 2k (13.54) Example 13. letting u = x 2 and v = y 2 .⎞ ----. Third Edition Copyright © Orchard Publications .56) by (13. relation (13.11 Evaluate the integrals ∫0 Solution: π⁄2 cos θ dθ and n ∫0 π⁄2 sin θ dθ n (13.59) to polar coordinates by letting x = ρ cos θ and y = ρ sin θ Then.56) Also.58) it written as Γ ( m )Γ ( n ) = ∫0 ∞ x 2 (m – 1) 2xe –x 2 dx ∫0 e ∞ y 2 2 (n – 1) 2ye –y 2 dy = 4 ∫0 ∞ x 2m – 2 xe –x 2 dx ∫0 y ∞ 2n – 2 ye –y 2 dy = 4 ∫0 ∫ 0 ∞ ∞ x 2m – 1 2n – 1 – ( x + y ) y 2 (13.= α ----2k y –y ∫0 y ∞ –1 ⁄ 2 –y e dy Finally. Then. Γ(n) = ∫0 x ∞ ∞ n – 1 –x e dx (13.57) For m > 0 and n > 0 .55) From the definition of the Γ ( n ) function. 13−14 Numerical Analysis Using MATLAB® and Excel®.57) yields Γ ( m )Γ ( n ) = ∫0 ∞ u m – 1 –u e du ∫0 v ∞ n – 1 –v e dv (13.59) dx dy Next.58) where u and v are dummy variables of integration. Γ(m) = ∫0 x m – 1 –x e dx (13. multiplication of (13. using the definition of the Γ ( n ) function. we convert (13.= α π ----.

⎞ ⎝ 2 ⎠ π n ---------------------. m = ( n + 1 ) .⎝2 ⎠ ⎝ 2 2⎠ (13. in (13.-. n+1 Γ ⎛ ----------..⎞ Γ ⎛ -. that is.⎞ Γ ⎛ ----------.-----⎛ n +1⎞ 2 ⎛ n+1+1⎞ Γ -2Γ ----------.-----⎛ m +1⎞ 2 ⎛ 1+m+1⎞ Γ --2Γ -. then.⎞ ⎝ 2 ⎠ π ⎝ 2 ⎠ ⎝2⎠ n cos θ dθ = -----------------------------------------. Therefore.62). we let ρ2 = w .⎞ ⎝ 2 ⎠ π ⎝2 ⎠ ⎝ 2 ⎠ m .63) If.61) 2m – 1 2n – 1 θ dθ ⋅ Γ ( m + n ) Rearranging (13.65) Numerical Analysis Using MATLAB® and Excel®.sin θ dθ = ⎛ n + 1⎞ 2 Γ -⎝2 ⎠ ∫0 π⁄2 cos θ dθ = n ∫0 π⁄2 n > –1 (13. we obtain the special case of (13. we replace 2m – 1 with 0 and 2n – 1 with m .60) 2 ρ dρ e To simplify (13.= -------------------------.64) We observe that (13.sin θ dθ = --------------------------------------------. we obtain the integral of the sin θ function as n ∫0 π⁄2 m+1 m+1 1 Γ ⎛ ------------.62) as 2 π⁄2 ∫0 n+1 n+1 1 Γ ⎛ ----------. n = -.63) and (13.-----.60).64) are equal since m and n can be interchanged. and 1 2n – 1 with 0 . Third Edition Copyright © Orchard Publications 13−15 . that is.⎞ Γ ⎛ -.⎞ Γ ⎛ ------------.60) is written as Γ ( m )Γ ( n ) = 2 = 2 ∫0 ∫0 π⁄2 cos cos 2m – 1 θ ⋅ sin θ ⋅ sin 2n – 1 θ dθ ∫0 w ∞ m + n – 1 –w e dw π⁄2 (13.= ------------------------. Then. dw = 2 ρ d ρ and thus relation (13.62) ---------------and this expression can be simplified by replacing 2m – 1 with n .The Gamma Function Γ ( m )Γ ( n ) = 4 = 2 ∫0 ∫0 ∫0 π⁄2 π⁄2 ∞ ( ρ cos θ ) 2m – 1 2m – 1 ( ρ sin θ ) θ dθ 2n – 1 – ρ 2 e ρ dρ dθ 2 cos θ ⋅ sin 2n – 1 ∫0 ρ ∞ 2m + 2n – 2 – ρ (13.61) we obtain ∫0 π⁄2 cos 2m – 1 θ ⋅ sin 2n – 1 θ dθ = ------------------------- Γ ( m )Γ ( n ) 2Γ ( m + n ) 2 (13.------------⎝ 2 ⎠ ⎝2 2 ⎠ (13.

The pdf for the gamma distribution. it is referred to as Erlang distribution.2 2.0 β^n 8.8 1. Third Edition Copyright © Orchard Publications .1084 0. ISBN 0−9709511− 0−8.0082 0. 13.65) are known as Wallis’s formulas.0 f(x) 0. x 0.10 0. β ) = ------------------------.Chapter 13 The Gamma and Beta Functions and Distributions The relations of (13. n.4 shows the probability density function (pdf) of the gamma distribution for n = 3 and β = 2 .alpha.4 n 3. Science.66) A detailed discussion of this probability distribution is beyond the scope of this book.6 1.20 0.0608 0.cumulative) where: x = value at which the distribution is to be evaluated alpha = the parameter n in (13. and Technology.1007 0. n.2 The Gamma Distribution One of the most common probability distributions* is the gamma distribution which is defined as n – 1 –x ⁄ β e x f ( x.0920 0.0719 0.0000 0.15 f(x) 0.0 0. β > 0 n β Γ(n) (13.4 0.2 0.05 0.0023 0.beta.0 1.0167 0.0379 Probability Density Function of the gamma distribution for n = 3 and β = 2 0.6 0.0268 0.2 1. 13−16 Numerical Analysis Using MATLAB® and Excel®. it will suffice to say that it is used in reliability and queuing theory.0 β 2. When n is a positive integer.4.66) beta = the parameter β in (13. Figure 13. We can evaluate the gamma distribution with the Excel GAMMADIST function whose syntax is GAMMADIST(x.66) * Several probability distributions are presented in Mathematics for Business.0 Γ(n) 2.0823 0.4 1.00 0 2 4 6 x 8 10 12 Figure 13.0 2.0494 0.x > 0.8 2.

0925 We observe that these values are consistent with the plot of Figure 13.67) where m > 0 and n > 0 .FALSE) returns 0.13 Prove that B ( m. 13. n = 3 .4. denoted as B ( m. GAMMADIST returns the cumulative distribution function (cdf). then. Numerical Analysis Using MATLAB® and Excel®. n ) . and if FALSE. x = 4 . y → 0 . * Several probability density functions are also presented on the text mentioned on the footnote of the previous page.The Beta Function cumulative = a TRUE / FALSE logical value. that is. Then. Third Edition Copyright © Orchard Publications 13−17 . it returns the probability density function* (pdf). and β = 2 b. Example 13.2. we specify the FALSE condition. x = 7 . n ) = B ( n.3. y → 1 and as x → 1 .1353 =GAMMADIST(7. b.FALSE) returns 0. m ) (13. n ) = ∫0 x 1 m–1 (1 – x) n–1 dx (13. We observe that as x → 0 . if TRUE.68) Proof: Let x = 1 – y . and β = 2 Solution: Since we are interested in the probability density function (pdf) values. Example 13. the pdf of the gamma distribution if: a. =GAMMADIST(4. is defined as B ( m.3 The Beta Function The beta function. dx = – dy .3. n = 3 . a.2. Therefore.12 Use Excel’s GAMMADIST function to evaluate f ( x ) .

n ) function is also useful in evaluating certain integrals as illustrated by the following examples. The B ( m.70). θ → 0 and as x → 1 .15 Prove that Γ ( m )Γ ( n ) B ( m. n ) = 2 ∫0 π⁄2 cos 2m – 1 θ ⋅ sin 2n – 1 θ dθ (13. Then. Third Edition Copyright © Orchard Publications . n ) = = ∫0 ∫0 1 x m–1 (1 – x) y n–1 dx = – dy = ∫1 ( 1 – y ) n–1 0 m–1 [1 – (1 – y)] n–1 dy 1 (1 – y) m–1 n–1 ∫0 y 1 (1 – y) m–1 dy = B ( n.70) ∫0 π⁄2 2m – 1 θ ) ( cos 2m – 1 θ ) dθ Example 13. We observe that as x → 0 .71) Proof: The proof is evident from (13. B ( m. θ → π ⁄ 2 . dx = 2 sin θ cos θ d θ . Example 13. 13−18 Numerical Analysis Using MATLAB® and Excel®. n ) = ∫0 x ∫0 1 m–1 (1 – x) 2 n–1 dx 2 n–1 = 2 = 2 π⁄2 ( sin θ ) ( sin m–1 ( cos θ ) sin θ cos θ dθ (13.69) Proof: We let x = sin 2θ .62) and (13.14 Prove that B ( m.Chapter 13 The Gamma and Beta Functions and Distributions B ( m. then. m ) and thus (13.68) is proved. n ) = -----------------------Γ(m + n) (13.

format rat. Example 13.The Beta Function Example 13. However. For this example.72) Solution: By definition B ( m.= -------Γ(9) 8! 40320 40320 280 (13.00 n= 4 Γ(n) exp(gammaln(n)) 6. Third Edition Copyright © Orchard Publications 13−19 .= --------.5.71) we obtain Γ ( 5 )Γ ( 4 ) 4!3! 24 × 6 144 1 B ( 5.16 Evaluate the integral ∫0 x ( 1 – x ) dx 4 3 1 (13. ∫0 x ( 1 – x ) d x 4 3 1 = B ( 5.71) for its computation as shown in Figure 13. Computation of the beta function with Excel.00 Γ(m+n) exp(gammaln(m+n)) 40320.17 Evaluate the integral ∫ x --------------.n) function.5. n ) = ∫0 x 1 m–1 (1 – x) n–1 dx and thus for this example. n ) function directly.4) z = 1/280 Excel does not have a function that computes the B ( m.73) We can also use MATLAB’s beta(m. % display answer in rational format z=beta(5.74) Numerical Analysis Using MATLAB® and Excel®.= -------------. we can use (13. p g Γ(m) exp(gammaln(m)) 24.n)= Γ(m) x Γ(n) / Γ(m+n) 1/280 m= 5 Figure 13.dx 0 2–x 2 2 (13.= -------------. 4 ) Using (13.00 Beta(m. 4 ) = ----------------------.

Then.6. n ) m–1 n–1 (13.⋅ x Γ ( m )Γ ( n ) x < 0 < 1. n ) = -------------------------------------.dx = 4 2 ⋅ 2! ⋅ π = 64 2 -----------------------------------------15 15 π ⁄ 8 0 2–x 2 2 (13. (13. n ) = -----------------------. and dx = 2dv .76) we obtain ∫ x --------------.Chapter 13 The Gamma and Beta Functions and Distributions Solution: Let x = 2v . m. etc. driving habits.dv = 4 2 1–v 2 ∫0 v ( 1 – v ) 2 1 –1 ⁄ 2 dv 1 Γ ( 3 )Γ ( 1 ⁄ 2 ) = 4 2 ⋅ B ⎛ 3. from (13.74) becomes 2 2 ∫0 where Γ ( 3 ) = 2! 1 4v 8 -----------------. eating times and places. (13.75) Γ(1⁄2) = π Γ ( 7 ⁄ 2 ) = ( 7 ⁄ 2 – 1 )Γ ( 7 ⁄ 2 – 1 ) = ( 5 ⁄ 2 ) Γ ( 5 ⁄ 2 ) = ( 5 ⁄ 2 ) ( 5 ⁄ 2 – 1 )Γ ( 5 ⁄ 2 – 1 ) = 5 ⁄ 2 ⋅ 3 ⁄ 2 ⋅ ( 3 ⁄ 2 – 1 )Γ ( 3 ⁄ 2 – 1 ) = ( 15 ⁄ 8 )Γ ( 1 ⁄ 2 ) = 15 π ⁄ 8 (13.77) 13. v → 0 . m.x < 0 < 1.74).⎞ = 4 2 -----------------------------⎝ 2⎠ Γ(7 ⁄ 2) (13. then x = 4v . m. a detailed discussion of the beta probability distribution is beyond the scope of this book. -.75) and (13.78) A plot of the beta probability density function (pdf) for m = 3 and n = 2 .4 The Beta Distribution The beta distribution is defined as x (1 – x) f ( x.m – 1 ( 1 – x )n – 1 f ( x. Using (13. m. n > 0 B ( m. As with the gamma probability distribution. n > 0 (13. Third Edition Copyright © Orchard Publications .76) Then. it will suffice to say that it is used in computing variations in percentages of samples such as the percentage of the time in a day people spent at work. We observe that as x → 0 . is shown in Figure 13. and as x → 2 .71) we can express the beta distribution as Γ(m + n) .79) 13−20 Numerical Analysis Using MATLAB® and Excel®. v → 1 .2 dv = -----2 2 – 2v 2 ∫0 1 v --------------.

02 0.7560 0.0 Γ(m) 2.0 0. Third Edition Copyright © Orchard Publications 13−21 .4530 0.06 0.0064 for m = 3 and n = 20.6003 0.7200 0.8800 0.0324 0.0000 0.0 0.0144 2.20 0.9600 0.0184 0. is zero.79) beta = the parameter n in (13.0784 0.0036 of the Beta Distribution 0.2580 0.A.7000 0.0047 0.0676 0.04 0.16 0.79) A = the lower bound to the interval of x B = the upper bound to the interval of x From the plot of Figure 13.14 0.1156 Figure 13.alpha.4 0.12 0.6 1.24 0.30 0.0707 0.0406 0.0900 0.8200 0.1080 0.0000 Numerical Analysis Using MATLAB® and Excel®.n) 1.m.6774 0.3.The Beta Distribution x 0.3840 0. f ( x.9800 0.1) which returns 1.m.0 0.00 0.8 0.4 x 0.1024 0. The pdf of the beta distribution We can evaluate the beta cumulative distribution function (cdf) with Excels’s BETADIST function whose syntax is BETADIST(x. However.0004 0.0 0.08 0.0576 0.9156 Probability Density Function 0.2.8600 0.8356 0.1521 0.0 n 2. we see that when x = 1 . This value can be verified by =BETADIST(1.32 0.0400 0.beta.6.0.7800 0.9000 0.0 Γ(m+n) 24.0 x(m-1) 0.0196 0.5253 0.10 0. the cumulative distribution (the area under the curve) at this point is 100% or unity since this is the upper limit of the x −range.0484 0.2 0.8400 0.3188 0.8 f(x.6.0016 (1-x)(n-1) 1.9200 0.0 Γ(n) 1.6 0.0256 0.7600 0.28 0.2023 0.22 0.6800 0.34 m 3.18 0.9400 0.7400 1.2 0.6600 f(x.0000 0.n) 0.B) where: x = value between A and B at which the distribution is to be evaluated alpha = the parameter m in (13.26 0.0100 0.8000 0. n ) which represents the probability density function.0000 0. m.

It is defined as Γ(n) = ∫0 x ∞ n – 1 –x e dx • It is convenient to use the relation Γ(n) = Γ(n + 1) -------------------n for n < 0 and the relation nΓ ( n ) = Γ ( n + 1 ) for n > 0 . • The Γ ( n ) function and the factorial n! are related as Γ ( n + 1 ) = n! for n = 1. … • We can use MATLAB’s gamma(n) function to obtain values of Γ ( n ) . the 13−22 Numerical Analysis Using MATLAB® and Excel®. • The Γ ( n ) function is defined for all positive integers and positive fractional values. denoted as Γ ( n ) . and for all negative fractional. we can use the relation • Other useful relations are shown below.+ ------------. 2. is also known as generalized factorial function. Γ ( n ) = ( n – 1 )! Γ(1⁄2 ) = π π Γ ( n )Γ ( 1 – n ) = ------------sin nπ for 0 < n < 1 2 2n – 1 Γ(n) π = πΓ ( 2n ) • The relation Γ ( n + 1 ) = n! = for any n ≠ negative integer ⎫ n –n ⎧ 1 1139 571 2πnn e ⎨ 1 + -------. • To evaluate Γ ( n ) when n is a positive integer. 3. but not negative integer values. Third Edition Copyright © Orchard Publications .Chapter 13 The Gamma and Beta Functions and Distributions 13. • We can use the EXP(GAMMALN(n)) function to evaluate Γ ( n ) at some positive value of n .5 Summary • The gamma function.– ------------------.+ … ⎬ 12n 288n 2 51840n 3 2488320n 4 ⎩ ⎭ is referred to as Stirling’s asymptotic series for the Γ ( n ) function.– ------------------------. If n is a positive integer.

n. n ) = ------------------------Γ(m + n) • The beta B ( m. n ) function is also useful in evaluating certain integrals. • The beta distribution is defined as (1 – x) x f ( x. n > 0 B ( m.sin θ dθ = ---------------------. n ) where m > 0 and n > 0 is defined as B ( m.n) function to evaluate the beta B ( m.Summary factorial n! can be approximated as n! ≈ 2πnn e n –n • The Γ ( n ) function is very useful in integrating some improper integrals. • We can use MATLAB’s beta(m. m. B ( m. n.x < 0 < 1. m. n ) and gamma function Γ ( n ) are related by Γ ( m )Γ ( n ) . B ( m. β ) = ------------------------. n ) m–1 n–1 Numerical Analysis Using MATLAB® and Excel®. n ) = -------------------------------------. • The relations n+1 Γ ⎛ ----------. β > 0 n β Γ(n) • The beta function. • The gamma distribution which is defined as n – 1 –x ⁄ β x e f ( x. Third Edition Copyright © Orchard Publications 13−23 .-----⎛ n + 1⎞ 2 Γ -⎝2 ⎠ ∫0 π⁄2 cos θ dθ = n ∫0 π⁄2 n > –1 are known as Wallis’s formulas. n ) function.x > 0. n ) = ∫0 x 1 m–1 (1 – x) n–1 dx • The beta function B ( m.⎞ ⎝ 2 ⎠ π n .

Third Edition Copyright © Orchard Publications . compute B ( m.6 Exercises 1. n ) Verify your answer with MATLAB and Excel 3. Evaluate the following integrals a. b. c. ∫0 ∞ e –x 3 dx 3 ∫0 ∞ xe –x dx ∫0 -----------------4 1–x 1 dx d. e.5 and n = – 1. ∫0 π⁄2 tan θ dθ dx ∫0 --------------------2 3x – x 3 13−24 Numerical Analysis Using MATLAB® and Excel®. Given that m = 10 and n = 8 . compute Γ(m + n) -----------------------Γ ( m )Γ ( n ) Verify your answer with MATLAB and Excel 2.25 . Given that m = 2.Chapter 13 The Gamma and Beta Functions and Distributions 13.

7 Solutions to End−of−Chapter Exercises 1.= 5. Then.= ---------------------------------------.= ------------------------17! Γ(m + n) Γ ( 18 ) 9⋅8⋅7⋅6⋅5⋅4⋅3⋅2⋅7⋅6⋅5⋅4⋅3⋅2 = -----------------------------------------------------------------------------------------------------------------------------------------17 ⋅ 16 ⋅ 15 ⋅ 14 ⋅ 13 ⋅ 12 ⋅ 11 ⋅ 10 ⋅ 9 ⋅ 8 ⋅ 7 ⋅ 6 ⋅ 5 ⋅ 4 ⋅ 3 ⋅ 2 –6 7⋅6⋅5⋅4⋅3⋅2 5040 = -----------------------------------------------------------------------------.5 + ( – 1.25 ) Γ ( 2.5 )Γ ( – 1.5. gamma(m+n)/(gamma(m)*gamma(n)) ans = 0.25.1419 × 10 17 ⋅ 16 ⋅ 15 ⋅ 14 ⋅ 13 ⋅ 12 ⋅ 11 ⋅ 10 980179200 Check with MATLAB: beta(10.1419e-006 Numerical Analysis Using MATLAB® and Excel®.= -------------------------------------.6256 and Γ ( 3 ⁄ 4 ) = 1.6256 5Γ ( m + n )-----------------------.⋅ --------------------------. Γ ( m )Γ ( n ) Γ ( 10 ) ⋅ Γ ( 8 ) ( 9! ) × ( 7! ) B ( m.2254 Γ ( m )Γ ( n ) Check with MATLAB: m=2. Third Edition Copyright © Orchard Publications 13−25 .= -------------------------. by substitution into (1) we obtain: 3.5 )Γ ( – 1. 2.8) ans = 5. -------------------By repeated use of the relations nΓ ( n ) = Γ ( n + 1 ) for n > 1 and Γ ( n ) = Γ ( n + 1 ) for n < 1 .= --------------------------------------------------------------------------------Γ ( –5 ⁄ 4 + 1 ) 3 ⁄ 2 ⋅ Γ ( 3 ⁄ 2 ) ⋅ ( –4 ⁄ 5 ) ⋅ Γ ( –1 ⁄ 4 ) Γ ( 3 ⁄ 2 + 1 ) ⋅ -----------------------------(1) ( –5 ⁄ 4 ) 5 Γ(1 ⁄ 4) 1 ⁄ 4 ⋅ Γ(1 ⁄ 4) = --------------------------------------------------------------------------------------------. where we find that Γ ( 1 ⁄ 4 ) = 3.⋅ -----------------------------Γ ( 3 ⁄ 4 ) 48 π ⋅ Γ ( 3 ⁄ 4 ) 3 ⁄ 2 ⋅ 1 ⁄ 2 ⋅ Γ ( 1 ⁄ 2 ) ⋅ ( – 4 ⁄ 5 ) ⋅ -----------------–1 ⁄ 4 There are no exact values for Γ ( 1 ⁄ 4 ) and Γ ( 3 ⁄ 4 ) .= -----------------------------. therefore.1739 We cannot check the answer with Excel because it cannot compute negative values.= ----. n ) = -----------------------.25 ) Γ ( 5 ⁄ 2 ) ⋅ Γ ( –5 ⁄ 4 ) Γ(1 ⁄ 4 + 1) 1 ⁄ 4 ⋅ Γ( 1 ⁄ 4 ) = --------------------------------------------------------------.= 0.= ----.2254 . we obtain their approximate values from tables.= -------------------------------------------Γ ( m )Γ ( n ) Γ ( 2.25 ) ) Γ ( 1.1739 48 π ⋅ 1.Solutions to End−of−Chapter Exercises 13. n we obtain Γ(m + n) Γ ( 2.25 ) Γ(5 ⁄ 4) -----------------------. n=−1.

4 ∞ xe –x 3 dx = x 2 – 1 –x e dx = Γ ( 2 ) = 1! = 1 ( –3 ⁄ 4 ) 1 We let x = y or x = y 1⁄4 1 .62). ∫0 π⁄2 tan θ dθ = ∫0 π⁄2 sin θ 1 ⁄ 2 ⎛ ----------.⋅ ----------------------------------------. 1 1 dx ----------------. Γ ( 3 ⁄ 4 ) = -----------------Γ(1 ⁄ 4) and by substitution into (1) ∫0 -----------------4 1–x 1 dx 1 Γ(1 ⁄ 4) ⋅ π 1 Γ(1 ⁄ 4) ⋅ Γ(1 ⁄ 2) = -.⋅ -------------------------------4 2π ⁄ Γ ( 1 ⁄ 4 ) 4 Γ(3 ⁄ 4) 2 π 2 {Γ(1 ⁄ 4)} = ------------. dx = ( 1 ⁄ 4 )y dy and thus ( –1 ⁄ 2 ) ∫0 Also.⋅ { Γ ( 1 ⁄ 4 ) } = --------------------------4 2π 4 2π d. Let x = y . ∫0 13−26 π⁄2 cos 2m – 1 θ ⋅ sin 2n – 1 θ dθ = ------------------------- Γ ( m )Γ ( n ) 2Γ ( m + n ) Numerical Analysis Using MATLAB® and Excel®. dx = -.= -4 4 1–x 1 = -4 ∫0 1 --------------.Chapter 13 The Gamma and Beta Functions and Distributions 3.⋅ y ( – 3 ⁄ 4 ) dy = 1 -4 1–y 1⁄2–1 ∫0 ( 1 – y ) ⋅y ( –3 ⁄ 4 ) dy ∫0 ( 1 – y ) 1 ⋅y 1⁄4–1 1 Γ(1 ⁄ 4) ⋅ Γ(1 ⁄ 2) dy = -. so 3 dx = ∫0 b. then x = y 3 1⁄3 1 ( –2 ⁄ 3 ) .⎞ 3 3 ⎝3⎠ 3 ∫0 c. a. Third Edition Copyright © Orchard Publications .= ------------sin ( 3π ⁄ 4 ) 2⁄2 Γ(3 ⁄ 4) ⋅ Γ(1 ⁄ 4) = 2π 2π . Then.Γ ⎛ -.= -.⋅ y dy .⋅ y dy = -. ∞ e –x 3 ∫0 e ∫0 ∞ ∞ –y 1 1⁄3–1 1 1 ⋅ -.⎞ dθ = ⎝ cos θ ⎠ ∫0 π⁄2 ( sin θ ) 1⁄2 ( cos θ ) –1 ⁄ 2 ( dθ ) From (13.⋅ ----------------------------------------4 Γ(3 ⁄ 4) (1) or π π Γ ( 3 ⁄ 4 ) ⋅ Γ ( 1 – 3 ⁄ 4 ) = ------------------------.

= x(3 – x) –1 ⁄ 2 ∫0 x –1 ⁄ 2 –1 ⁄ 2 ⋅ (3 – x) –1 ⁄ 2 dx (1) = ( 3y ) = ( 3 ⁄ 3 )y Let x = 3y .= ---------2Γ ( 3 ⁄ 4 + 1 ⁄ 4 ) 2Γ ( 1 ) 2 3 3 ∫0 3 dx --------------------.= π ⎝2 2⎠ Γ(1 ⁄ 2 + 1 ⁄ 2) Γ(1) 2 3x – x Numerical Analysis Using MATLAB® and Excel®. and the integral of (1) becomes 3 -----.= -------------. x when x = 3 .--------------------. y = 1 .( 1 – y ) dy 3 –1 ⁄ 2 (2) ∫0 y –1 ⁄ 2 (1 – y) dy Recalling that B ( m. n – 1 = – 1 ⁄ 2 . y = 0 and ∫0 1 y –1 ⁄ 2 ⋅ ( 3 – 3y ) –1 ⁄ 2 dy = = 3 1 ∫0 y 1 –1 ⁄ 2 –1 ⁄ 2 1 ⋅ -----.Then.= --------------------------.⎞ = ----------------------------------------. π⁄2 Γ ( 3 ⁄ 4 )Γ ( 1 ⁄ 4 ) 2π 2π tan θ dθ = ------------------------------------.= 2 3x – x ∫0 dx ----------------------.Solutions to End−of−Chapter Exercises Letting 2m – 1 = 1 ⁄ 2 and 2n – 1 = – 1 ⁄ 2 we obtain m = 3 ⁄ 4 and n = 1 ⁄ 4 .⋅ 3 3 –1 ⁄ 2 . When x = 0 . n ) = ∫0 x 1 m–1 (1 – x) n–1 Γ(m) ⋅ Γ(n) dx = ----------------------------Γ(m + n) it follows that m – 1 = – 1 ⁄ 2 . m = 1 ⁄ 2 and thus ∫0 3 2 1 1 Γ(1 ⁄ 2) ⋅ Γ(1 ⁄ 2) {Γ(1 ⁄ 2)} dx . Third Edition Copyright © Orchard Publications 13−27 . then dx = 3dy . m = 1 ⁄ 2 . -.= B ⎛ --. ∫0 e.

and we conclude with the factorization methods LU.1. orthogonal vectors. QR. Figure 14. Orthogonal lines Orthogonality applies also to curves.1. We begin with orthogonal lines and functions. Third Edition Copyright © Orchard Publications 14−1 . Orthogonal curves Numerical Analysis Using MATLAB® and Excel®. and Singular Value Decomposition. Cholesky. From analytic geometry and elementary calculus we know that two lines are orthogonal if the product of their slopes is equal to minus one. Figure 14.2 shows the angle between two curves C 1 and C 2 . 14. y slope = m 1 slope = m 2 m1 ⋅ m2 = –1 x Figure 14.Chapter 14 Orthogonal Functions and Matrix Factorizations T his chapter is an introduction to orthogonal functions. Mutually orthogonal systems of curves and vectors are of particular importance in physical problems.1 Orthogonal Functions Orthogonal functions are those which are perpendicular to each other. orthogonal trajectories.2. This is shown in Figure 14.

3) and (14. L 1 and L 2 are orthogonal if m 2 = – 1 ⁄ m 1 .1 Prove that every curve of the family xy = a a≠0 (14.Chapter 14 Orthogonal Functions and Matrix Factorizations By definition. the slope is xdy + ydx = 0 or dy y ----. Other orthogonal functions are the cos x and sin x functions as we’ve learned in Chapter 6.2.2) Proof: At a point P ( x. in Figure 14.1). Example 14. y ) on any curve of (14. then.2) the slope is 2xdx – 2ydy = 0 or dy = x -----y dx (14. The two families of these curves are shown in Figure 14. If m 1 and m 2 are the slopes of these two lines. Third Edition Copyright © Orchard Publications . the curves of (14. the angle between the curves C 1 and C 2 is the angle β between their tangent lines L 1 and L 2 .4) we see that these two curves are orthogonal since their slopes are negative reciprocals of each other.3) On any curve of (14. The cases where x = 0 or y = 0 cannot occur because we defined a ≠ 0 and b ≠ 0 .3.1) is orthogonal to every curve of the family x –y = b 2 2 b≠0 (14.2 Orthogonal Trajectories Two families of curves with the property that each member of either family cuts every member of the other family at right angles are said to be orthogonal trajectories of each other.1). Thus. 14. 14−2 Numerical Analysis Using MATLAB® and Excel®.4) From (14.= – -dx x (14.2) are orthogonal trajectories of the curves of (14.

Orthogonal Trajectories Figure 14. Orthogonal trajectories Example 14.6) From (14. Third Edition Copyright © Orchard Publications 14−3 .9) Numerical Analysis Using MATLAB® and Excel®.8) or 2ydy + xdx = 0 2 y dy + x dx = 0 2 y . c = y ⁄ x and thus we rewrite (14.3.x = ----2 dx x x 2 (14. the slope of the orthogonal family we are seeking must be xdy = – --------2y dx (14.5).7) Therefore.2 Find the orthogonal trajectories of the family of parabolas y = cx 2 Solution: The slope of (14.= k ( cons tan t ) 2 2 2 ∫ ∫ x + 2y = C ( cons tan t ) 2 2 (14.5) dy = 2cx ----dx (14.x2 ---.5) is c≠0 (14.+ ---.= 2 ---.6) as y dy 2y ----.

9) represents a family of ellipses and the trajectories are shown in Figure14. Third Edition Copyright © Orchard Publications .3 Orthogonal Vectors Let X = [ x 1 x 2 x 3 … x n ] and Y = [ y 1 y 2 y 3 … y n ] be two vectors of the same length.10) Example 14. Example 14.Chapter 14 Orthogonal Functions and Matrix Factorizations Relation (14. Their inner (dot) product is defined as X ⋅ Y = x 1 y 1 + x 2 y 2 + x 3 y 3 + … + x n y n (a scalar ) (14. 14−4 Numerical Analysis Using MATLAB® and Excel®.4 Test the vectors X 1 = [ 1 1 1 ] and X 2 = [ 1 – 2 1] for orthogonality. Orthogonal trajectories for families of parabolas and ellipses. 14.3 Given that X = [ 1 1 1 ] and Y = [ 2 1 2 ] find the dot product X ⋅ Y Solution: X ⋅ Y = (1) ⋅ (2) + (1) ⋅ (1) + (1) ⋅ (2) = 5 Definition:Two vectors X 1 and X 2 are said to be orthogonal if their dot product is zero.4. Figure 14.4.

-.-. This process is called normalization. T Numerical Analysis Using MATLAB® and Excel®.= -. the matrix A is orthogonal and A⋅A = I T (14.5 Given that X = [2 4 4] compute the unit vector U X .11) where A is the transpose of A and I is the identity matrix. 1 2 2 2 4 4 . we compute the magnitude X . Third Edition Copyright © Orchard Publications 14−5 . If the column (or row) vectors of a square matrix A are mutually orthogonal unit vectors. With any vector X ≠ 0 we may associate a unique unit vector U which is obtained by dividing each component of X by each magnitude X defined as X = x1 + x2 + … + xn 2 2 2 where x i represents an element of the vector X ..Orthogonal Vectors Solution: X1 ⋅ X2 = ( 1 ) ⋅ ( 1 ) + ( 1 ) ⋅ ( –2 ) + ( 1 ) ⋅ ( 1 ) = 0 Therefore. X = 2 +4 +4 = 6 2 2 2 To compute the unit vector U X we divide each element of X by the magnitude X .U X = -. Thus..-2 3 3 6 6 6 A basis that consists of mutually orthogonal vectors is referred to as an orthogonal basis. Example 14.-. For this example. the vectors X 1 and X 2 are orthogonal to each other. If these vectors are also unit vectors. Solution: First.. the basis is called orthonormal basis.

11).13) Equating like terms we obtain z1 + z2 = 1 2 2 2 2 –z1 + z2 = 0 2 2 2 From the second equation we obtain z 1 = z 2 and by substitution into the first we obtain 2z 1 = 1 or * It is strongly suggested that the reader reviews the definitions of eigenvalues and eigenvectors in Chapter 5 at this time. with these eigenvalues we can obtain an infinite number of eigenvectors.12) or z1 + z2 2 –z1 2 2 –z1 + z2 2 z1 + 2 z2 + 2 z2 = 1 0 0 1 (14. we find the eigenvalues of the matrix A from the relation det ( A – λI ) = 0 where for this example ⎛ ⎞ det ⎜ 1 ⁄ 2 1 ⁄ 4 – λ 1 0 ⎟ = 0 ⎝ 1⁄2 1⁄2 0 1 ⎠ det 1 ⁄ 2 – λ 1⁄2 2 1⁄4 1⁄2–λ = 0 λ – λ + 3 ⁄ 16 = 0 from which λ 1 = 1 ⁄ 4 and λ 2 = 3 ⁄ 4 and as we’ve learned in Chapter 5.6 Given that A = 1⁄2 1⁄4 1⁄2 1⁄2 find an orthonormal set of eigenvectors* and verify that the result satisfies (14.Chapter 14 Orthogonal Functions and Matrix Factorizations Example 14. To find a 2 × 2 square matrix Z such that we begin with z1 z2 –z1 z2 ⋅ Z⋅Z = I z1 –z1 z2 z2 2 2 T = 1 0 0 1 (14. Third Edition Copyright © Orchard Publications . Solution: First. 14−6 Numerical Analysis Using MATLAB® and Excel®.

…Y m using the following relations.⋅ Y m – 1 – … – -----------------. 14.The Gram-Schmidt Orthogonalization Procedure z1 = z2 = 1 ⁄ ( ± 2 ) This result indicates that we can choose either 1 ⁄ 2 or 1 ⁄ ( – 2 ) for the values of z 1 and z 2 .⋅ Y 2 – ----------------.We choose the value 1 ⁄ 2 and then the first (left most) matrix in (14.4 The Gram-Schmidt Orthogonalization Procedure Let X 1.14) below are the inner (dot) products and if X = [ x 1 x 2 x 3 … x n ] and Y = [ y 1 y 2 y 3 … y n ] are two vectors of the same length their dot product is defined as X ⋅ Y = x 1 y 1 + x 2 y 2 + x 3 y 3 + … + x n y n (a scalar ) . We must remember that the products in (14.⋅ Y 1 Ym – 1 ⋅ Ym – 1 Y1 ⋅ Y1 (14.14) Also. X 2.⋅ Y 1 Y1 ⋅ Y1 Y1 ⋅ X3 Y2 ⋅ X3 Y 3 = X 3 – ----------------. Thus in the second equation in (14. A simpler procedure is the GramSchmidt orthogonalization procedure which is discussed on the next section. Third Edition Copyright © Orchard Publications 14−7 .12) is Z = 1⁄ 2 1⁄ 2 1 ⁄ (– 2) 1 ⁄ 2 and as a check.14) the dot products on the numerator and denominator must be found first and the result must be from the dot product of it and Y 1 Y1 = X1 Y1 ⋅ X2 Y 2 = X 2 – ----------------.⋅ Y 1 Y2 ⋅ Y2 Y1 ⋅ Y1 … Ym – 1 ⋅ Xm Y1 ⋅ Xm Y m = X m – -------------------------------. Y 2. …X m be some column vectors. We can find an orthogonal basis Y 1. 1 ⁄ 2 ⋅ 1 ⁄ 2 1 ⁄ (– 2) = 1 0 0 1 1 ⁄ (– 2) 1 ⁄ 2 1⁄ 2 1⁄ 2 1⁄ 2 The computations for finding orthonormal sets of eigenvectors for larger size ( 3 × 3 or higher) matrices using the above procedure becomes quite involved. the unit vectors Numerical Analysis Using MATLAB® and Excel®.

15) are mutually orthogonal and form an orthonormal basis.⋅ Y 2 – ----------------.14) Y1 = X1 = [ 1 1 1 ] T T T T Y1 ⋅ X2 T 0 T Y 2 = X 2 – ----------------. we have: 1⁄ 3 A = 1⁄ 3 1⁄ 3 T 1⁄ 6 –2 ⁄ 6 1⁄ 6 –1 ⁄ 2 0 1⁄ 2 We can verify that A ⋅ A = I with the MATLAB script below. I=A*A' 14−8 Numerical Analysis Using MATLAB® and Excel®. m (14. X 2 = [ 1 – 2 1 ] . A=[1/sqrt(3) 1/sqrt(6) −1/sqrt(2).7 Given that X 1 = [ 1 1 1 ] . and X 3 = [ 1 2 3 ] .15) Y1 T U 1 = -------.⋅ Y 1 = [ 1 – 2 1 ] 3 Y1 ⋅ Y1 Y1 ⋅ X3 Y2 ⋅ X3 6 T 0 T Y 3 = X 3 – ----------------. 1/sqrt(3) −2/sqrt(6) 0.[ 1 1 1 ] 3 Y2 ⋅ Y2 Y1 ⋅ Y1 6 = [ 1 2 3 ] – [ 2 2 2 ] = [ –1 0 1 ] T T T and from (14.⋅ Y 1 = [ 1 – 2 1 ] – -. In our subsequent discussion the column vectors will be denoted as row vectors transposed. Example 14. 1/sqrt(3) 1/sqrt(6) 1/sqrt(2)].= [ 1 ⁄ 3 1 ⁄ 3 1 ⁄ 3 ] Y1 Y2 T U 2 = -------. Third Edition Copyright © Orchard Publications . … .Chapter 14 Orthogonal Functions and Matrix Factorizations Yi U i = ------Yi i = 1.⋅ Y 2 – -. find an orthonormal basis.⋅ Y 1 = [ 1 2 3 ] – -. 2.= [ 1 ⁄ 6 – 2 ⁄ 6 1 ⁄ 6 ] Y2 Y3 T U 3 = -------.= [ – 1 ⁄ 2 0 1 ⁄ 2 ] Y3 and denoting the matrix whose elements are the unit vectors as A . Solution: From (14.

9153 0. we may find different values depending on the process being used. B=[1 1 1.0000 0.0000 0.0000 We can also use the MATLAB function orth(A) to produce an orthonormal basis as shown below. L 11 0 0 x1 x3 b1 b3 L 21 L 22 0 L 31 L 32 L 33 ⋅ x2 = b2 The unknowns are found from Numerical Analysis Using MATLAB® and Excel®. 1 2 3]. I=C*C' T I = 1.0000 1.0000 -0.0000 0.0000 1.The LU Factorization I = 1. Third Edition Copyright © Orchard Publications 14−9 .0000 -0.4027 0. the vectors produced by MATLAB also satisfy the condition C ⋅ C = I . Consider the following 3 × 3 lower triangular case.0000 14.0000 0 0.0000 0 0. As shown below.0000 -0. 1 −2 1.0000 0. In Chapter 4 we saw how the method of Gaussian elimination proceeds by systematically removing the unknowns from a system of linear equations. C=orth(B) C = -0. The reason for this difference is that the orthogonalization process is not unique. that is. The LU factorization method decomposes a matrix A into a lower triangular matrix L and an upper triangular matrix U so that A = L ⋅ U . computers use the so-called matrix factorization methods to decompose a matrix A into a product of other smaller matrices.5 The LU Factorization In matrix computations.0000 -0.0000 0 1.0000 0 1.0000 -0.9153 0.0000 1.4027 We observe that the vectors of the C matrix produced by MATLAB are different from those we derived with the Gram-Schmidt orthogonalization procedure.0000 -0.

5 v 3 = – 3. we begin by multiplying the first equation by – 2 and subtracting it from the second equation. With these two reductions we obtain 2v 1 – v 2 + 3v 3 = 5 – 5v 2 + 4v 3 = 18 2. to solve U 11 U 12 U 13 0 0 U 33 x1 x3 b1 0 U 22 U 23 ⋅ x 2 = b 2 b3 (14. Likewise.17) we start from the bottom to the top as shown below.5v 2 – 5.Chapter 14 Orthogonal Functions and Matrix Factorizations x 1 = b 1 ⁄ L 11 x 2 = ( b 2 – L 21 x 1 ) ⁄ L 22 x 3 = ( b 3 – L 31 x 1 – L 31 x 2 ) ⁄ L 33 (14.20) 14−10 Numerical Analysis Using MATLAB® and Excel®. Example 14.5 (14.18) is referred to as backward substitution. Thus. This removes v 1 from the second equation. x 3 = b 3 ⁄ U 33 x 2 = ( b 2 – U 23 x 3 ) ⁄ U 22 x 1 = ( b 1 – U 12 x 2 – U 13 x 3 ) ⁄ U 11 (14.19) To find the three unknowns. the unknowns are written in reverse order.18) provided that U 11 ⋅ U 22 ⋅ U 33 ≠ 0 . we multiply the first equation by 3 ⁄ 2 and we subtract it from the third equation.16) is referred to as forward substitution. The substitution order in (14.16) provided that L 11 ⋅ L 22 ⋅ L 33 ≠ 0 . The substitution order in (14. 2v 1 – v 2 + 3v 3 = 5 – 4v 1 – 3v 2 – 2v 3 = 8 3v 1 + v 2 – v 3 = 4 (14.8 Let us review the example given in Chapter 4 which consists of the following equations. Third Edition Copyright © Orchard Publications . For the upper triangular case.

To find the elements of matrix L we use MATLAB to multiply matrix A by the inverse of matrix U . U 11 U 12 U 13 0 U 22 U 23 0 0 U 33 2 –1 3 = 0 –5 4 0 0 – 3.20).22) where the first matrix on the left side is the lower triangular matrix L and the second is the upper triangular matrix U . A=[2 −1 3.5 Now. 0 0 −3. Third Edition Copyright © Orchard Publications 14−11 .20) by – 1 ⁄ 2 and we subtract it from the third equation of (14. −4 −3 −2.16) and (14. The elements of matrix U are the coefficients of v 1 . v 2 . and v 3 in (14.18) to find the lower and upper triangular matrices of our example where 2 –1 3 A = –4 –3 –2 3 1 –1 We want to find L ij and U ij such that L 11 0 0 ⋅ U 11 U 12 U 13 2 –1 3 0 U 22 U 23 = A = – 4 – 3 – 2 3 1 –1 0 0 U 33 L 21 L 22 0 L 31 L 32 L 33 (14.20) and we obtain the system of equations below. 3 1 −1]. 0 −5 4.21). 2v 1 – v 2 + 3v 3 = 5 – 5v 2 + 4v 3 = 18 – 3. L=A*inv(U) L = Numerical Analysis Using MATLAB® and Excel®.5 (14.5 v 3 = 5.21) We see that the eliminations have transformed the given square system into an equivalent upper triangular system that gives the same solution which is obtained as follows: v 3 = – 11 ⁄ 7 v 2 = ( 18 – 4v 3 ) ⁄ ( – 5 ) = – 34 ⁄ 7 v 1 = ( 5 + v 2 – 3v 3 ) ⁄ 2 = 17 ⁄ 7 The elements of the upper triangular matrix U are the coefficients of the unknowns in (14.5]. U=[2 −1 3.The LU Factorization Next we multiply the second equation of (14. Thus. Thus. let us use the relations of (14.

and L 32 = – 1 ⁄ 2 . −2 1 0.5]. U=[2 −1 3.U]=lu(A) to decompose the matrix 2 –3 1 A = –1 5 –2 3 –8 4 14−12 Numerical Analysis Using MATLAB® and Excel®. These values are the multipliers that we’ve used in the elimination process in succession. 2 –1 3 1 0 = –2 –4 –3 –2 1 3 1 –1 3 ⁄ 2 –1 ⁄ 2 0 2 –1 3 ⋅ 0 –5 4 0 1 0 0 – 3. we found the elements of the lower triangular matrix L by first computing the inverse of the upper triangular matrix U and performing the matrix multiplication but not L = U ⋅ A . Third Edition Copyright © Orchard Publications . 0 0 −3.5 Check with MATLAB: L=[1 0 0. A=L*U A = 2 -4 3 -1 -3 1 3 -2 -1 In the example above. L 31 = 3 ⁄ 5 .9 Use the MATLAB function [L. 0 −5 4.Chapter 14 Orthogonal Functions and Matrix Factorizations 1 -2 3/2 0 1 -1/2 0 0 1 Therefore. Was this necessary? The answer is no. the lower triangular matrix has the form L = A⋅U 1 0 0 L = L 21 1 0 L 31 L 32 1 –1 –1 and in our example we found that the values of the subdiagonal elements are L 21 = – 2 . the matrix A has been decomposed to a lower triangular matrix L and an upper matrix U as shown below. 3/2 −1/2 1]. Example 14. For a square matrix where none of the diagonal elements are zero.

3 −8 4]. 1 0 L' = – 1 ⁄ 3 1 2⁄3 1 0 0 1 (14. When a matrix lacks structure we say that it is permuted.U] = lu(X) stores an upper triangular matrix in U and a "psychologically lower triangular matrix" (i.U]=lu(A) L = 2/3 -1/3 1 U = 3 0 0 -8 7/3 0 4 -2/3 -1 1 1 0 1 0 0 We observe that while the upper triangular matrix U has the proper structure. To put it in the given form we need to make the same interchanges in rows as with the lower triangular matrix. A=[2 −3 1. U=[3 −8 4. Only part of the display is shown below.e. Let us now use MATLAB to see if L' ⋅ U = A . help lu LU LU factorization. Third Edition Copyright © Orchard Publications 14−13 . To find out how MATLAB performs LU factorization.23) The new matrix L' has now the proper structure. let us invoke the help lu command. we must interchange the first and third rows. To put L in the proper structure. let us interchange the first and third rows. 2/3 1 1].The LU Factorization into a lower and an upper triangular. [L. −1/3 1 0. a product Numerical Analysis Using MATLAB® and Excel®. −1 5 −2. A1=L1*U A1 = 3 -1 2 -8 5 -3 4 -2 1 We observe that matrix A is now permuted. the lower triangular matrix L lacks structure. that is. L1=[1 0 0. Then. [L. 0 0 −1]. Solution: format rat. 0 7/3 −2/3.

and permutation matrix P so that P*X = L*U.P] = lu(X) returns unit lower triangular matrix L. we obtain the permutation matrix P below.9. for example. [L. Third Edition Copyright © Orchard Publications .U. Solution: This is the same matrix as in Example 14. so that X = L*U.P]=lu(A) L = 1 -1/3 2/3 0 1 1 0 0 1 14−14 Numerical Analysis Using MATLAB® and Excel®. 3 −8 4].U. X can be rectangular.U.P]=lu(A) to decompose the matrix 2 –3 1 A = –1 5 –2 3 –8 4 into a lower and an upper triangular and show that P ⋅ A = L ⋅ U .9. the identity matrix 1 0 0 I = 0 1 0 0 0 1 If we interchange the first and third rows of the identity matrix I above. A=[2 −3 1. Example 14. [L. upper triangular matrix U. The permutation matrix P is an identity matrix that is permuted so that the rows of this matrix indicate the interchanges. Consider.24) and matrix P indicates the same interchanges as with the lower triangular matrix in Example 14.Chapter 14 Orthogonal Functions and Matrix Factorizations of lower triangular and permutation matrices) in L. −1 5 −2. Thus.10 Use the MATLAB function [L. 0 0 1 P = 0 1 0 1 0 0 (14.

].. Third Edition Copyright © Orchard Publications 14−15 . LU=L*U PA = 3 -1 2 LU = 3 -1 2 -8 5 -3 4 -2 1 -8 5 -3 4 -2 1 We observe that P ⋅ A = L ⋅ U with the first and second rows interchanged when compare with the given matrix A . % j = {row # to be interchanged with row i}. X(j. i = {first row # to be interchanged}. Also. The user−defined function ExchRows below. X(i. % This file is saved as ExchRows.24).j) % temp = X(i..The LU Factorization U = 3 0 0 P = 0 0 1 0 1 0 1 0 0 -8 7/3 0 4 -2/3 -1 We observe that the lower triangular matrix has now the proper structure and the P matrix displayed by MATLAB is the same as in (14. The MATLAB matrix left division operator x = A\b uses the L ⋅ U factorization approach.:) = temp. and Numerical Analysis Using MATLAB® and Excel®. PA=P*A. define the matrix or vector % X and the indices i and j in MATLAB's Command Window % as X=[.i.:)..m % To run this program.:).:) = X(j. % The function ExchRows interchanges rows i and j % of a matrix or vector X % function X = ExchRows(X. interchanges rows i and j j of a vector or matrix X.

Third Edition Copyright © Orchard Publications . 4 −9 −8 −1]. let us explain the use of MATLAB’s built-in function max(v) where v is a row or a column vector. Example 14. ExchRows(X. As an example.m user−defined function above to interchange rows 1 and 3. j = 3.Chapter 14 Orthogonal Functions and Matrix Factorizations % then type ExchRows(X. First. and for matrices is a row vector containing the maximum element from each column.i.i.11 Given that the matrix X is defined as –2 5 X = –3 –6 7 –5 4 –9 –4 8 3 –8 9 1 2 –1 use the ExchRows.m]=max(abs(v)) 14−16 Numerical Analysis Using MATLAB® and Excel®.j)at the command prompt. −3 −6 8 1. max(v) ans = 7 [Amax.m]=max(v) Amax = 7 m = 5 [Amax. Solution: At the MATLAB command prompt we enter X=[−2 5 −4 9. i = 1. 7 −5 3 2.j) and MATLAB outputs X = 7 -3 -2 4 -5 -6 5 -9 3 8 -4 -8 2 1 9 -1 The user−defined function GaussElimPivot below. let v=[2 −1 3 −5 7 −9 −12]'. performs Gauss elimination with row pivoting.

% A is a matrix that contains the coefficients of % the system of equations.k).1:n))).2) > 1. end if m ~= k b = ExchRows(b. z = zeros(n. b=b'.m). and b % is a column vector that contains the known values % on the right hand side. m = m + k − 1. A = ExchRows(A.m] = max(abs(A(k:n.k+1:n) − alpha*A(k. error('Matrix is singular'). function x = GaussElimPivot(A. end % The statements below exchange rows if required for k = 1:n−1 [Amag. Third Edition Copyright © Orchard Publications 14−17 . z = ExchRows(z.m). end end end % Back substitution phase for k = n:−1:1 b(k) = (b(k) − A(k. end n = length(b). A(i.k) ~= 0 alpha = A(i.k).k+1:n)*b(k+1:n))/A(k.m).k+1:n) b(i) = b(i) − alpha*b(k). z(i) = max(abs(A(i. end % Elimination steps for i = k+1:n if A(i. if Amag < eps.k. % Set up scale factor array for i = 1:n.The LU Factorization Amax = 12 m = 7 % This user−defined function file solves A*x=b by % the Gauss elimination with row pivoting method.k+1:n) = A(i.k. end Numerical Analysis Using MATLAB® and Excel®./z(k:n)).b) if size(b. x is a column vector that % will display the computed unknown values.1).k.k)).k)/A(k.

5714 -1.0000 2.0000 A = 7.0000 -5.0000 4.2857 -4.8571 9.0000 -8. Solution: At the MATLAB command prompt we enter A=[−2 5 −4 9.1429 3.5714 -6.9298 -9.0000 9.1429 3. Third Edition Copyright © Orchard Publications .1429 14−18 Numerical Analysis Using MATLAB® and Excel®.0000 1.7143 2. −3 −6 8 1.0000 -8.0000 1.0000 A = 7.0000 1.0000 4.1429 2.0000 -2.0000 3.1429 3.0000 9.1429 5.2857 0.5714 -6.5714 -9. GaussElimPivot(A.1429 -5.5714 -2.0000 -2.8571 9.0000 -9.0000 -3.0000 -5.0000 -2.Chapter 14 Orthogonal Functions and Matrix Factorizations % Enter the values of A and b at the MATLAB's % command window and type GaussElimPivot(A. x Example 14.b).0000 -3.0000 2.0000 -3.0000 4.7143 3. b = 2 .2857 -3.3860 -2.0000 3.0000 -8.0000 9.2857 -3.0000 -1.1429 3.1429 -9.0000 9.0000 -8.0000 A = 7.0000 -5.0000 1. b = [−3 2 8 5]'. x and MATLAB outputs the following: A = 7. x = x2 x3 2 8 –1 5 x4 use the GaussElimPivot user−defined function above to compute the values of the vector x.0000 -2. 4 −9 −8 −1].8571 9.0000 -8.1429 -8.b). 7 −5 3 2.0000 4.0000 -3.12 Given that –2 5 A = –3 –6 7 –5 4 –9 –4 8 3 –8 x1 9 –3 1 .8571 10.

7193 0.0000 -3.1176 -0.1429 -6.5714 -1./z(k:n)).0000 -2.1429 3.1).k. if Amag < eps error('Matrix is singular') end if m ~= k z = ExchRows(z.0000 1.9298 0.3860 -3.7193 3.1889 -0.0000 x = 0. end % Exchange rows if necessary for k = 1:n−1 [Amag. returns A = L*U % and the row permutation vector permut n = size(A. performs LU decomposition.2857 -16.1373 Check with MATLAB’s left division: x=b\A x = 0.0980 0.k.0000 4. m = m + k − 1.m] = max(abs(A(k:n.9298 -16.0000 -8.7451 -5.1).0980 3.0000 9. function [A.5439 2.0000 -8.1429 3.m).0000 1. Third Edition Copyright © Orchard Publications 14−19 .m).1429 -5.1373 The user−defined function LUdecomp below.k)).0000 -3.8571 10.8571 -3.1176 2. permut = (1:n)'. for i = 1:n. A = ExchRows(A.2857 0. z(i) = max(abs(A(i. z = zeros(n.0000 4.5714 -6.permut] = LUdecomp(A) % LU decomposition of matrix A.1:n))). Numerical Analysis Using MATLAB® and Excel®.0000 9.5439 10.0000 -2.0000 A = 7.The LU Factorization A = 7.7451 -1. and returns matrix A as A=L*U and the row permutation vector permut.

4386 3. end % Elimination pass for i = k+1:n if A(i.k+1:n).4286 0.k)/A(k.2857 -16. end end end Example 14.0556 2. A(i.7193 -0.permut] = LUdecomp(A) A = 7.Chapter 14 Orthogonal Functions and Matrix Factorizations permut = ExchRows(permut.1889 14−20 Numerical Analysis Using MATLAB® and Excel®.k+1:n) − alpha*A(k.k) = alpha. [A.5714 -0.0000 1. Third Edition Copyright © Orchard Publications .8571 -3. Solution: At the MATLAB command prompt we enter A=[−2 5 −4 9.m).1429 0.7544 -0. −3 −6 8 1.13 Given that –2 5 A = –3 –6 7 –5 4 –9 –4 8 3 –8 9 1 2 –1 use the LUdecomp user−defined function above to decompose matrix A and show how the given rows were permuted. 4 −9 −8 −1].5439 10.k).k)~=0 alpha = A(i.0000 -0.2857 permut = 3 2 4 1 Check: -5.k+1:n) = A(i.0000 9. 7 −5 3 2. A(i.k.0000 -8.

0000 9.4490 0 0 0 0 0 1 0 0 1.0000 -8.0816 -0.7193 -0.The LU Factorization [L.0000 9.0000 1. b = b'.0556 2. % if size(b) > 1.0761 -5.0000 -8.9796 -3.2857 -5.5439 10.0000 2. Numerical Analysis Using MATLAB® and Excel®.0000 -0. is saved as LUsolPivot.0000 0. x = b.0612 0.permut) % % The six statements below rearrange vector b and % stores it in vector x.7544 -0.m and will be used in the user−defined function matInvert that follows. % In this user−defined function matrix A and column % vector b are entered in MATLAB's command window % and “permut” holds the row permutation data.1).1429 0.0000 0 0 0 P= 1 0 0 0 L*U 0 1. x(i)= b(permut(i)).4286 0.5714 -0.0000 -0.8571 -3. % function x = LUsolPivot(A.6612 0 0 0 0 1.1889 The user−defined function LUsolPivot listed below.1376 0. end n = size(A.U. for i = 1:n.0000 1.2857 -16.4694 -15.0417 3.4347 10.P]=lu(A) L = 1. Third Edition Copyright © Orchard Publications 14−21 .0000 -0.2632 0 1 0 0 0 0 1 0 ans = 7.0408 U = 7.4386 3.b.

Third Edition Copyright © Orchard Publications .permut).i).i) = LUsolPivot(A.Ainv(:. 14−22 Numerical Analysis Using MATLAB® and Excel®.k). % Ainv = eye(n). % The statement below performs LU decomposition % using the user−defined function LUdecomp(A) saved % previously % [A.14 Invert the matrix A below with the user−defined function matInvert. % for i = 1:n Ainv(:.permut] = LUdecomp(A). end The user−defined function matInvert below. end for k = n:−1:1 x(k) = (x(k) − A(k. and store results in Ainv % replacing the corresponding vector using the % user−defined function LUsolPivot saved previously.Chapter 14 Orthogonal Functions and Matrix Factorizations end % % The next six statements perform forward and % backward substitution % for k = 2:n x(k) = x(k)− A(k. end Example 14. % Creates identity matrix of size n.1:k−1)*x(1:k−1). % Assigns the size of A to n. inverts matrix A with LU decomposition. % This user−defined function inverts a matrix A % defined in MATLAB's command prompt using LU % decomposition % function Ainv = matInvert(A) % n = size(A.k+1:n)*x(k+1:n))/A(k. % % The last three statements solve for each vector % on the right side.1).

1174 0.0389 0. It is a special case of LU factorization and requires fewer computations than the LU factorization method of the previous section. there exists an upper triangular matrix G with positive diagonal elements such that G ⋅G = A T (14. 4 −9 −8 −1].0201 -0.0416 -0. Let us invoke the MATLAB help chol command to see how MATLAB performs this factorization.0447 0. Third Edition Copyright © Orchard Publications 14−23 .0078 0.1174 0. Numerical Analysis Using MATLAB® and Excel®.0208 0.0201 -0.0416 -0.0610 0.0683 -0. ans = -0. −3 −6 8 1. Under those conditions.0055 14. 7 −5 3 2.0078 0.0242 -0.0389 0.0447 0. A = A .The Cholesky Factorization –2 5 A = –3 –6 7 –5 4 –9 –4 8 3 –8 9 1 2 –1 Solution: In MATLAB’s command prompt we enter A=[−2 5 −4 9.0055 Check with MATLAB’s built−in inv(A) function.0783 -0.26) is referred to as Cholesky factorization.25) for every x ≠ 0 and A is symmetric. Ainv = matInvert(A) and MATLAB returns Ainv = -0.0683 -0.0719 0.0242 -0.0783 -0. that is.0369 0.0719 0.26) Relation (14.0981 inv(A) -0.0208 0.0610 0.0369 0.0013 -0.0013 -0.6 The Cholesky Factorization A matrix is said to be positive definite if x ⋅A⋅x>0 T T (14.0981 -0.

Example m = 5. K > 0 is above the main diagonal and K < 0 is below the main diagonal. If X is positive definite.0) and puts V on the main diagonal.p] = CHOL(X). then p is a positive integer. CHOL(X) uses only the diagonal and upper triangle of X. [R. an error message is printed.N) or EYE([M. help eye EYE Identity matrix. We will consider an example using the Cholesky factorization after we review the MATLAB functions eye(n) and diag(v. then R = CHOL(X) produces an upper triangular R so that R'*R = X.-1) produces a tridiagonal matrix of order 2*m+1. The lower triangular is assumed to be the (complex conjugate) transpose of the upper.Chapter 14 Orthogonal Functions and Matrix Factorizations CHOL Cholesky factorization. DIAG(V) is the same as DIAG(V. But if X is not positive definite.N]) is an M-by-N matrix with 1's on the diagonal and zeros elsewhere. See also ONES. DIAG(X. If X is not positive definite.K) when V is a vector with N components is a square matrix of order N+ABS(K) with the elements of V on the K-th diagonal.1). with two output arguments. never produces an error message. RANDN. help diag DIAG Diagonal matrices and diagonals of a matrix. 14−24 Numerical Analysis Using MATLAB® and Excel®. See also SPDIAGS.1) + diag(ones(2*m.1). RAND. EYE(N) is the N-by-N identity matrix. K = 0 is the main diagonal. TRIL.k) as defined by MATLAB. EYE(M.K) when X is a matrix is a column vector formed from the elements of the K-th diagonal of X. then p is 0 and R is the same as above. DIAG(DIAG(X)) is a diagonal matrix. DIAG(X) is the main diagonal of X. Third Edition Copyright © Orchard Publications . TRIU. DIAG(V. EYE(SIZE(A)) is the same size as A. If X is positive definite. diag(-m:m) + diag(ones(2*m. ZEROS.

00 0 0 -0. Numerical Analysis Using MATLAB® and Excel®.R]=qr(A) produces an n × n matrix whose columns form an orthonormal or unitary* matrix Q and an upper triangular matrix R of the same size as matrix A .19 0 0 0 -1.11 2.15 Use MATLAB to compute the Cholesky factorization of matrix A as defined below.50 0 0 0 -0. B=[−0.00].50 0 -0. 1).75 5. G=chol(A). The MATLAB function [Q.24 0 0 0 0 A1 = 5.00 -0. A=5*eye(5)+diag(B. Third Edition Copyright © Orchard Publications 14−25 .00 We observe that A1 = A .00 -0.25 0 0 0 -0.75 0 0 0 G = 2.00 5.75 5. that is.25 −0.00 0 0 0 -0.7 The QR Factorization The QR factorization decomposes a matrix A into the product of an orthonormal matrix and an upper triangular matrix.50 5.22 2.25 5.45 2.25 0 -0.27) * An n × n matrix A is called unitary if ( A∗ ) = A T –1 where A∗ is the complex conjugate matrix of A . A1=G'*G Solution: Execution of the MATLAB script above displays the following: A = 5.34 2.23 0 0 0 -0.00 -0.75 −1.50 5.00 5. In other words. A can be factored as A = QR (14.The QR Factorization Example 14. the matrix product G ⋅ G = A is satisfied.00 -0.00 0 0 -0.75 0 T 0 0 -0.00 -1.22 0 0 0 -0.00 -1. 14. format bank.00 -0.21 0 0 0 -0.50 −0.25 5. −1)+diag(B.00 0 0 0 -1.

[Q.Chapter 14 Orthogonal Functions and Matrix Factorizations Then.R]=qr(A) is described as follows: help qr QR Orthogonal-triangular decomposition. 14−26 Numerical Analysis Using MATLAB® and Excel®. The column permutation E is chosen so that abs(diag(R)) is decreasing. an upper triangular R and a unitary Q so that A*E = Q*R. Note that R = chol(A'*A).0) and [C.29) The MATLAB [Q.28) and multiplying both sides of (14.B. Q.R. By itself.R. returns C = Q'*B. a system described by Ax = b becomes QRx = b (14.R] = QR(A.0) produces an "economy size" decomposition in which E is a permutation vector.B). [Q. where B has as many rows as A.R] = QR(A. QR(A) is the output of LAPACK'S DGEQRF or ZGEQRF routine. The column permutation E is chosen so that abs(diag(R)) is decreasing. then only the first n columns of Q are computed. The full version of QR does not return C. [Q. C. TRIU(QR(A)) is R.A*x R\(R'\(A'*r)) x + e. but Q is often nearly full.0) produce economy size results.R] = QR(A.E] = QR(A. R = QR(A) returns only R. If A is m-by-n with m > n.E] = QR(A) produces a permutation matrix E.R] = QR(A) produces an upper triangular matrix R of the same dimension as A and a unitary matrix Q so that A = Q*R.E). Third Edition Copyright © Orchard Publications . R = QR(A. [Q.R] = QR(A) returns both Q and R. The least squares approximate solution to A*x = b can be found with the Q-less QR decomposition and one step of iterative refinement: x r e x = = = = R\(R'\(A'*b)) b . so that Q*R = A(:.28) by Q ⋅ Q = I we obtain Rx = Q b T T (14.0) produces the "economy size" decomposition.

28 1.R]=qr(A) and x = R\Q b .00 -0.27 -0.00 Let us verify that the matrix Q is unitary.15 9.35 -3.58 0.80 R = -3.R]=qr(A).00 0.58 0.16 Given that 2 –3 1 2 A = – 1 5 – 2 and b = 4 3 –8 4 5 solve Ax = b using the MATLAB function [Q.00 1.00 1.24 0 -0.53 0. Third Edition Copyright © Orchard Publications 14−27 .00 Check=A\b -0.77 0.14 4. x=R\Q'*b T Q = -0.00 T –1 or Q ⋅ Q = I .43 7.00 -0.The QR Factorization Example 14. 3 −8 4].00 -0.74 0 0 x = 4.62 -0. b=[2 4 5]'.00 0.58 Check = 4. [Q.58 -4. Of course. since the elements are real numbers. the complex conjugate of Q is also Q and thus we only need to show that Q = Q Q*Q' ans = 1.00 -0. Solution: A=[2 −3 1. T Numerical Analysis Using MATLAB® and Excel®.54 0.14 4.43 7. −1 5 −2.

S . and unitary matrices U and V so that A = U⋅S⋅V T (14.30) The matrices U . of the same dimension as A and with nonnegative diagonal elements in decreasing order. b=[2 4 5]'. This function returns the vector X that minimizes norm(A*X−b) subject to X ≥ 0 provided that the elements of A and b are real numbers. then pinv(A)=inv(A). Of course.b). Third Edition Copyright © Orchard Publications .0000 Underdetermined systems have infinite solutions and MATLAB selects one but no warning message is displayed.4286 7. X=lsqnonneg(A. For example. In an overdetermined system. The function pinv(A) displays the pseudoobtaininverse of a m × n (non−square) matrix A . As we’ve learned in Chapter 4. if A is square.1429 4. 14. V]=svd(A).Chapter 14 Orthogonal Functions and Matrix Factorizations QR factorization is normally used to solve overdetermined systems. and V . −1 5 −2. A=[2 −3 1. so we select the vector X which produces the minimum error. the MATLAB function inv(A) produces the inverse of the square matrix A and an error message is displayed if A is not a square matrix.17 Decompose the matrix * We defined overdetermined and underdetermined systems in Chapter 8 14−28 Numerical Analysis Using MATLAB® and Excel®. Example 14. MATLAB does this with either the left division operator ( \ ) or with the non−negative least−squares function lsqnonneg(A. there is no vector X which can satisfy the entire system of equations. decomposed from a given matrix A . S. 3 −8 4].8 Singular Value Decomposition The Singular Value Decomposition (SVD) method decomposes a matrix A into a diagonal matrix S . can be found with the MATLAB function [U. systems with more equations than unknowns as in applications where we need to find the least square distance in linear regression.* that is.b) returns X = 4.

8050 -0.0000 1.4605 0 0 V = -0.Singular Value Decomposition 2 –3 1 A = –1 5 –2 3 –8 4 into two unitary matrices and a diagonal matrix with non−negative elements. S.8632 -0.5184 0. −1 5 −2.0000 -0.0000 -0. A=[2 −3 1. U*U' ans = 1.0000 -0.0000 1.0000 -0.0000 -0. Solution: We will use the MATLAB [U.0000 -0. 3 −8 4].3116 0.0000 ans = 1.5675 0 0 0.0000 -0.4731 -0.8929 As expected.4420 0.3972 -0.0863 0.V]=svd(A) U = -0.8228 S = 11.2122 -0.1782 0 -0.5924 -0. [U. V]=svd(A) function. We also verify that the matrices U and V are unitary as shown below.5028 0. Third Edition Copyright © Orchard Publications 14−29 .3150 0.0000 Numerical Analysis Using MATLAB® and Excel®.0000 0.0000 1.0000 0.0000 0.0000 1.0325 0 1.S.0000 V*V' -0.6521 0.9463 -0.2440 0.0000 0. the diagonal elements of the triangular S matrix are non−negative and in decreasing values.

• The magnitude of a vector X . Under those conditions. that is. A = A . The MATLAB function [Q. the basis is called orthonormal basis. T T T T T 14−30 Numerical Analysis Using MATLAB® and Excel®. • The LU factorization method decomposes a matrix A into a lower triangular matrix L and an upper triangular matrix U so that A = L ⋅ U . these vector are said to be orthogonal to each other. The MATLAB function [L. This process is referred to as the Cholesky factorization. • We can find an orthonormal set of eigenvectors in a 2 × 2 matrix easily from the eigenvalues but the computations for finding orthonormal sets of eigenvectors for larger size ( 3 × 3 or higher) matrices using the above procedure becomes quite involved. A simpler procedure is the Gram−Schmidt orthogonalization procedure which we will discuss on the next section. denoted as X . the matrix A is said to be orthogonal and A ⋅ A = I where A is the transpose of A and I is the identity matrix. • Two families of curves with the property that each member of either family cuts every member of the other family at right angles are said to be orthogonal trajectories of each other.U]=lu(A) decomposes the matrix A into a lower triangular matrix L and an upper triangular matrix U . If these vectors are also unit vectors.9 Summary • Orthogonal functions are those which are perpendicular to each other. • The QR factorization decomposes a matrix A into the product of an orthonormal matrix and an upper triangular matrix. • If the column (or row) vectors of a square matrix A are mutually orthogonal unit vectors. there exists an upper triangular matrix G with positive diagonal elements such that G ⋅ G = A . • The inner (dot) product of two vectors X = [ x 1 x 2 x 3 … x n ] and Y = [ y 1 y 2 y 3 … y n ] is a scalar defined as X ⋅ Y = x 1 y 1 + x 2 y 2 + x 3 y 3 + … + x n y n • If the dot product of two vectors X 1 and X 2 is zero.Chapter 14 Orthogonal Functions and Matrix Factorizations 14.R]=qr(A) produces an n × n matrix whose columns form an orthonormal or unitary matrix Q and an upper triangular matrix R of the same size as matrix A . • A basis that consists of mutually orthogonal vectors is referred to as an orthogonal basis. • A matrix is said to be positive definite if x ⋅ A ⋅ x > 0 for every x ≠ 0 and A is symmetric. A unique 2 2 2 unit vector U is obtained by dividing each component of X by the magnitude X and this process is referred to as normalization. Third Edition Copyright © Orchard Publications . is defined as X = x 1 + x 2 + … + x n .

can be found with the MATLAB function [U. V]=svd(A). S. The matrices U . and unitary matrices U and V so that A = U ⋅ S ⋅ V . T Numerical Analysis Using MATLAB® and Excel®. S .Summary • The Singular Value Decomposition (SVD) method decomposes a matrix A into a diagonal matrix S . Third Edition Copyright © Orchard Publications 14−31 . decomposed from a given matrix A . of the same dimension as A and with nonnegative diagonal elements in decreasing order. and V .

4.10 Exercises 1.20 4. 7.00 Then. use the MATLAB Cholesky factorization function to obtain the matrix G and verify that G ⋅ G = A .k) to define and display the matrix A shown below. Given the vectors X 1 = [ 2 – 1 ] and X 2 = [ 1 – 3 ] . 14−32 Numerical Analysis Using MATLAB® and Excel®. Use a suitable function to verify your results. Using the MATLAB functions eye(n) and diag(v. Show that the curve x + 3y = k 1 and the curve 3y = k 2 x where k 1 and k 2 are constants. 2.00 – 1. Verify your answers with MATLAB.00 4. 5.00 – 0.8 to decompose the system of equations x 1 + 2x 2 + 3x 3 = 14 – 2x 1 + 3x 2 + 2x 3 = 10 5x 1 – 8 x 2 + 6x 3 = 7 T T 2 2 2 2 3 into an upper triangular matrix U and a lower triangular matrix L . are orthogonal to each other. 6. Use the appropriate MATLAB function to decompose the system of equations A = 1 0 –1 3 and b = 5 0 1 2 –2 –3 4 9 T into an upper triangular matrix R of the same dimension as A and a unitary matrix Q so that Q ⋅ R = A . use the Gram−Schmidt orthogonalization procedure to find two vectors Y 1 and Y 2 to form an orthonormal basis.00 0 – 1. Use MATLAB to find another set of an orthonormal basis with the vectors given in Exercise 3.80 0 A = – 0.20 0 0 – 1. Third Edition Copyright © Orchard Publications .00 0 0 – 1.Chapter 14 Orthogonal Functions and Matrix Factorizations 14. Use the Gaussian elimination method as in Example 14.80 4. 4. Find the orthogonal trajectories of the curves of the family 2x + y = kx 3.

Use the appropriate MATLAB function to decompose the matrix A given as A = 1 0 –1 0 1 2 –2 –3 4 into a diagonal matrix S of the same dimension as A and with non−negative diagonal elements in decreasing order and unitary matrices U and V so that U ⋅ S ⋅ V = A . Third Edition Copyright © Orchard Publications 14−33 .Exercises 8. T Numerical Analysis Using MATLAB® and Excel®.

= 0 dx or 1 x dy .y 2 = ----.( 2x 2 ) + ----.Chapter 14 Orthogonal Functions and Matrix Factorizations 14.= k dx and solving for dy ⁄ dx .= k x 2 (4) ----------------2 dx 3 2 From (2).⋅ -. x + 3y = k 1 (1) 3y = k 2 x 3 2 2 (2) Implicit differentiation of (1) yields dy 2x + 6y ----. 2x + y = kx (1) 2 2 Implicit differentiation of (1) yields d d d ----.⋅ x 2 = 3 y (5) -----3 x dx x We observe that (5) is the negative reciprocal of (3) and thus the given curves are orthogonal to each other.----.( kx ) dx dx dx dy 4x + 2y ----. Third Edition Copyright © Orchard Publications .(2) dx 2y From (1).11 Solutions to End−of−Chapter Exercises 1.= -------------.(3) 3 y dx Differentiation of (2) yields dy = 3k 2 x . 3y k 2 = ----3 x and by substitution into (4) we obtain 3y dy = ----. dy k – 4x ----.= – -. 2. 14−34 Numerical Analysis Using MATLAB® and Excel®.

we need to find the curves whose slopes are given by the negative reciprocal of (4).= -----------------. dy = udx + xdu and by substitution into (6) ( 2x – u x ) ( udx + xdu ) = 2x udx 2 2 2 2 Division of both sides of the above by x yields ( 2 – u ) ( udx + xdu ) = 2udx 2 2 Collecting like terms and simplifying we obtain x ( 2 – u )du = u dx Separating the variables we obtain 2 (2 – u ) du dx ----. we need to find the family of the curves of 2xy dy = ------------------.Solutions to End−of−Chapter Exercises 2x + y k = ------------------.du 3 x u u and by integrating these terms we find –1 ln x + ln u = ----.(5) ----2 2 dx 2x – y We rewrite (5) as ( 2x – y )dy = 2xydx (6) 2 2 and we let y = ux .(4) x – 2x + y ----dx 2y 2xy 2 2 Now.du = ---.– 4x 2 2 dy = -------------------------------.= -----------------------.+ ----. that is.du – ----3 3 u x u u 2 2 3 or 2 dx du ----.(3) x 2 2 and by substitution into (2) 2x + y ------------------.= ---.+ C 2 u By substitution of u = y ⁄ x we obtain Numerical Analysis Using MATLAB® and Excel®. Third Edition Copyright © Orchard Publications 14−35 . Then.

8507 I=B*B' -0.= [ 2 ⁄ 5 Y1 Y2 U 2 = -------. we have: A = T 2⁄ 5 –1 ⁄ 5 –1 ⁄ 5 –2 ⁄ 5 We verify that A ⋅ A = I as shown below. 1 −3].⋅ [ 2 – 1 ] T T Y1 ⋅ Y1 [ 2 –1 ] ⋅ [ 2 –1 ] T [2 + 3] T T 5 T = [ 1 – 3 ] – ---------------.⋅ [ 2 – 1 ] = [ 1 – 3 ] – -.⋅ [ 2 – 1 ] = [4 + 1] 5 = [ 1 –3 ] – [ 2 –1 ] = [ –1 2 ] T T T and from (14. From (14. 2⁄ 5 –1 ⁄ 5 –1 ⁄ 5 ⋅ 2 ⁄ 5 –2 ⁄ 5 –1 ⁄ 5 –1 ⁄ 5 = –2 ⁄ 5 4⁄5+1⁄5 –2 ⁄ 5 + 2 ⁄ 5 –2 ⁄ 5 + 2 ⁄ 5 = 1 0 1⁄5+4⁄5 0 1 4.5257 14−36 Numerical Analysis Using MATLAB® and Excel®.15) Y1 U 1 = -------.⋅ C⎞ = ln Cy = – x -------⎝ x ⎠ 2 x y and thus the family of curves orthogonal to the given family is x = – y ln Cy 2 2 3. B=orth(A) B = -0.8507 0.= [ – 1 ⁄ 5 Y2 – 1 ⁄ 5] T –2 ⁄ 5 ] T and denoting the matrix whose elements are the unit vectors as A .5257 -0. Third Edition Copyright © Orchard Publications .14) Y1 = X1 = [ 2 –1 ] T T T Y1 ⋅ X2 T [ 2 –1 ] ⋅ [ 1 –3 ] T Y 2 = X 2 – ----------------.Chapter 14 Orthogonal Functions and Matrix Factorizations 2 y ln x + ln y + ln C = ln ⎛ x ⋅ -. A=[2 −1.⋅ Y 1 = [ 1 – 3 ] – -----------------------------------------------.

and – 18 ⁄ 7 . x 1 + 2x 2 + 3x 3 = 14 7x 2 + 8x 3 = 38 – 18 x 2 – 9 x 3 = – 63 (2) Next. we multiply the second equation in (2) by – 18 ⁄ 7 and we subtract it from the third in (2). L 31 . 1.0000 -0. x 3 = 243 ⁄ 81 = 3 x 2 = ( 38 – 8x 3 ) ⁄ 7 = 2 x 1 = ( 14 – 3x 3 – 2x 2 ) = 1 The multipliers that we’ve used are – 2 .Solutions to End−of−Chapter Exercises I = 5.0000 -0. Likewise. and L 32 respectively.0000 1. we multiply the first equation by 5 and we subtract it from the third in (1). Therefore. These are the elements L 21 . Third Edition Copyright © Orchard Publications 14−37 . Then. after simplification x 1 + 2x 2 + 3x 3 = 14 7x 2 + 8x 3 = 38 ( 81 ⁄ 7 )x 3 = 243 ⁄ 7 (3) Thus.0000 x 1 + 2x 2 + 3x 3 = 14 – 2x 1 + 3x 2 + 2x 3 = 10 (1) 5x 1 – 8 x 2 + 6x 3 = 7 Multiplying the first equation of (1) by – 2 and subtracting it from the second in (1) we obtain the second equation in (2) and thus x 1 is eliminated. Then. 5 . the lower triangular matrix is 1 0 L = –2 1 5 – 18 ⁄ 7 0 0 1 The elements of the upper triangular matrix are the coefficients of the unknowns in (3) and thus U = 1 2 3 0 7 8 0 0 81 ⁄ 7 Numerical Analysis Using MATLAB® and Excel®.

[Q.00 -0.Chapter 14 Orthogonal Functions and Matrix Factorizations Now.00 0 0 0 A1 = 4.00 Q = -0.6733 0 4. x=R\Q'*b. A=[1 0 −1.80 4. 5 −18/7 1]. A=L*U A = 1 -2 5 6.1381 0.8944 R = -2. we use MATLAB to verify that L ⋅ U = A L=[1 0 0.4781 2.8018 0. −1)+diag(B. G=chol(A). Third Edition Copyright © Orchard Publications . 0 0 81/7].2].5976 0.00 -1.96 0 0 -0. QQT=Q*Q'.62 1. A1=G'*G A = 4. Check=A\b -0.0 −1. 0 1 2.8 −1. 0 7 8. 2 3 -8 3 2 6 format bank.20 4.00 -1. A=4*eye(4)+diag(B.00 4.0249 -0. −2 1 0.93 0 0 -1. b=[3 5 9]'.51 1.80 0 0 G = 2. −2 -3 4].5345 0.7171 -0.4472 0 0.20 4.00 -1. B=[−0.2361 0 0 -2.00 0 0 -1.3586 0. 1).80 0 0 7.6833 -1.00 -0.00 0 -0.2673 14−38 Numerical Analysis Using MATLAB® and Excel®.00 -1.20 0 0 -1.00 4.40 1.20 0 -0.80 4.90 0 0 -1. U=[1 2 3.00 0 0 -0.R]=qr(A).

7922 UUT = 1.1574 0 0 0.0000 Numerical Analysis Using MATLAB® and Excel®.7500 -2.0000 0.1345 0.7500 -2.5000 3.V]=svd(A).0000 0.0000 -0.0000 1.3795 -0. [U.2493 -0.2464 0.9480 -0. VVT=V*V' 0.0000 0.0000 VVT = 1.0000 -0.3184 -0.0000 -0. A=[1 0 −1.8187 0. UUT=U*U'.2076 0.0000 0.0000 -0.0000 0.0000 1.7000 -0.0000 U = -0.9577 S = 5.0000 1.0000 0.5583 0.0000 0.9154 0.0413 0 0.0000 1.7500 8.0000 0.0000 0.Solutions to End−of−Chapter Exercises QQT = 1.4779 0.0000 1.0000 0.S.0000 -0.1977 0.0000 -0.7500 Check = 6. 0 1 2.0000 0.5000 3.5985 0 0 V = -0.0000 1. Third Edition Copyright © Orchard Publications 14−39 . −2 −3 4].2410 0 2.9556 -0.0000 x = 6.2093 0.

1 The Bessel Function The Bessel functions. cannot be solved in terms of familiar functions as those which we encountered in ordinary differential equations with constant coefficients. such as (15. Applying the method of Frobenius to (15. these are discussed in advanced mathematics textbooks. and telephone equations. acoustics. and Chebyshev Functions T his chapter is an introduction to some very interesting functions. Then. are used in engineering. Bessel functions are solutions of the differential equation x2 dy dx 2 2 +x dy 2 2 + ( x – n )y = 0 dx (15. thermodynamics.Chapter 15 Bessel. aeronautics. the form of the general solution of (15. we will accept the solutions without proof. in the electrical engineering field. They are solutions of differential equations with variable coefficients and. or even a complex number. fractional.2) This series is referred to as Bessel function of order n where n is any positive real number or zero.1) depends on the value of n . It is beyond the scope of this book to derive the infinite series which are approximations to the solutions of these differential equations. under certain conditions. we obtain the infinite power series Jn ( x ) = ∞ k=0 ∑ x k ( – 1 ) ⋅ ⎛ -. These are special functions that find wide applications in science and engineering. 15. The usual procedure is to derive solutions in the form of infinite series. and the most common are the Method of Frobenius and the Method of Picard. we obtain the relation Numerical Analysis Using MATLAB® and Excel®. we replace n with – n . they are used in frequency modulation.2). Legendre. positive or negative integer. Third Edition Copyright © Orchard Publications 15−1 .1). Therefore. Differential equations with variable coefficients.1) where n can be any number. theory of elasticity and others. transmission lines. denoted as J n ( x ) . For instance. satisfy the orthogonality principle.⎞ ⎝2⎠ n + 2k 1 ⋅ --------------------------------------k! ⋅ Γ ( n + k + 1 ) n≥0 (15.1). If in (15.

– ----------------------.– ----------------------. 2. and compare with the series of J 1 ( x ) in (15. … (15.6) reduces to the following series: x x x x J 0 ( x ) = 1 – --------------------. Γ ( n + k + 1 ) = ( n + k )! (15. 1 and 2 .8).+ ------------------------. and odd when n is odd.⎨ 1 – ----------------------------------.– ----------------------. J n ( x ) is an even function of x .– --------------------. For the special case where n is a positive integer or zero. Legendre.– … 2 2 4 2 6 2 8 2 2 ⋅ ( 1! ) 2 ⋅ ( 2! ) 2 ⋅ ( 3! ) 2 ⋅ ( 4! ) x x x x x J 1 ( x ) = -.+ ----------------------------------------------------n 2 4 2 ⋅ n! ⎩ 2 ⋅ 1! ⋅ ( n + 1 ) 2 ⋅ 2! ⋅ ( n + 1 ) ( n + 2 ) 6 ⎫ x – ----------------------------------------------------------------------. if we multiply the series for J 1 ( x ) by x and differentiate it.4) and (15.+ ----------------------. Third Edition Copyright © Orchard Publications .3) and the function J –n ( x ) is referred to as the Bessel function of negative order n .2) reduces to Jn ( x ) = or k=0 ∑ ∞ x n + 2k 1 k ⋅ --------------------------( – 1 ) ⋅ ⎛ -. we will find that 15−2 Numerical Analysis Using MATLAB® and Excel®. we see that d J ( x ) = –J1 ( x ) dx 0 (15.+ ----------------------.+ --------------------.– … 3 5 7 9 2 2 ⋅ 1! ⋅ 2! 2 ⋅ 2! ⋅ 3! 2 ⋅ 3! ⋅ 4! 2 ⋅ 4! ⋅ 5! x x x x x J 2 ( x ) = -------------. that when n is zero or even. 1.9). (15.⎞ ⋅ ------------------------------------------⎝2⎠ k! ⋅ Γ ( – n + k + 1 ) (15.7) through (15.– … 2 4 6 8 10 2 ⋅ 2! 2 ⋅ 1! ⋅ 3! 2 ⋅ 2! ⋅ 4! 2 ⋅ 3! ⋅ 5! 2 ⋅ 4! ⋅ 6! 2 4 6 8 10 3 5 7 9 2 4 6 8 (15. and Chebyshev Functions J–n ( x ) = k=0 ∑ ( –1 ) ∞ k x – n + 2k 1 ⋅ ⎛ -.7).9) We observe from (15.5) n 2 4 x ⎧ x x J n ( x ) = -------------.– ----------------------.Chapter 15 Bessel. If we differentiate the series of J 0 ( x ) in (15.6) For n = 0.7) (15.+ ----------------------.+ --------------------.+ … ⎬ 6 ⎭ 2 ⋅ 3! ⋅ ( n + 1 ) ( n + 2 ) ( n + 3 ) (15.10) Also.⎞ ⎝2⎠ k! ⋅ ( n + k )! n = 0.8) (15.

3391 c.The Bessel Function d { xJ 1 ( x ) } = xJ 0 ( x ) dx (15.– ----. besselj(1.2239 ans = 0. 1 1 1 1 1 J 2 ( 1 ) = -.2240 192 4 36 576 4 64 64 × 36 256 × 576 or from math tables.1149 Numerical Analysis Using MATLAB® and Excel®. From (15. J 2 ( 1 ) = 0.= 0.1-. From (15.----J 0 ( 2 ) = 1 – 4 + 16 – ----------------.1) ans = 0.x) function or the Excel BESSELJ(x.+ ----------.1 Compute.= 0. correct to four decimal places. the values of a.+ ----------------------. 19683 -. From (15.----. we obtain besselj(0.2187J 1 ( 3 ) = 3 – 27 + 243 – -------------.7). J 2 ( 1 ) Solution: a. 431256 64 -. J 0 ( 2 ) = 0. We can use the MATLAB besselj(n.= 0.2239 b.+ ----------------------.= -------.9). J 1 ( 3 ) c. J 0 ( 2 ) b.= 1 – ----.+ -------------------.1149 . With MATLAB.2).8). J 1 ( 3 ) = 0. Third Edition Copyright © Orchard Publications 15−3 .3391 ans = 0.1152 8 96 1536 184320 17694720 or from math tables.3400 2 16 384 18432 1474560 or from math tables. besselj(2.+ -------.3).11) Example 15.n) function for the above computations.– ----------------.-------.

1 and 15. 15−4 Numerical Analysis Using MATLAB® and Excel®.x). respectively. z = besselj(2. J 1 ( x ) and J 2 ( x ) using MATLAB We can also use Excel to plot these series as shown in Figure 15. represent the carrier.Chapter 15 Bessel.5 J2(x) Jn(x) 0 -0. J 2 ( β ) and so on.1.85.x. besselj(2. 'J2(x)') The plots for J 0 ( x ) . The functions J 0 ( β ) .3391 besselj(1. text(4. and Chebyshev Functions and with Excel.1)= 0. Plots of J 0 ( x ) . all J n ( x ) are infinite series and thus. tables of some of the roots of J 0 ( x ) and J 1 ( x ) are shown in math tables. * In Frequency Modulation (FM). ylabel('Jn(x)'). grid. 0.0)= 0. plot(x. title('Bessel Functions of the First Kind'). w = besselj(1.00.1.w. J 1 ( x ) . first sideband.2.v. and J 2 ( x ) series.5 0 1 2 3 4 5 x 6 7 8 9 10 Figure 15.35.05: 10.2239 besselj(3. 'J1(x)'). text(0. 'J0(x)'). 0.1149 The MATLAB script below plots J 0 ( x ) .x).x).* 1 J0(x) Bessel Functions of the First Kind J1(x) 0. v = besselj(0.2)= 0. x = 0. The definition of a Bessel function of the first kind will be explained shortly. text(2.z).20.00: 0. 0.95. The x −axis crossings in the plot of Figures 15. Fortunately. it is a very difficult and tedious task to compute all roots of these series. Legendre.25.60. xlabel('x'). and J 2 ( x ) . J 1 ( x ) and J 2 ( x ) are shown in Figure 15. However. x is denoted as β and it is called modulation index. J 1 ( x ) . second sideband etc. Third Edition Copyright © Orchard Publications . J 1 ( β ) .x.2 show the first few roots of the J 0 ( x ) .

such as 1 ⁄ 2 .1723 0.5 Jn(x) J2(x) 0.0078 0. we obtain J1 ⁄ 2 ( x ) = 2----.0671 Bessel Functions of the First Kind 1.9944 0.0000 0.1 which shows the first 5 positive roots of these equations.9844 0.0000 0.9120 0.9776 0. for example. 5 ⁄ 2 . then J n ( x ) can be expressed in a finite form of sines and cosines.9900 0. This is observed on Table 15. the so−called half−order Bessel functions J 1 ⁄ 2 ( x ) and J ( – 1 ⁄ 2 ) ( x ) . the difference between consecutive roots approaches the value of π.5 0 2 4 x 6 8 10 Figure 15. For instance.1483 0.55 0.2647 0.5201 of J 0 ( x ) .10 0.25 0.0748 0.0050 0.0112 0.2. If we let n = 1 ⁄ 2 in (15.65 0.4048 and 5.45 0.3081 0.0152 0. and the differences between consecutive roots.05 0. They have no complex roots 2.0000 0. As the roots become larger and larger. that is.2867 0. J 0 ( x ) and J 1 ( x ) . Third Edition Copyright © Orchard Publications 15−5 . we observe that the first root 3. and so on. Between two consecutive roots of one of these equations lies one and only one root of the other equation.9385 0.0250 0.0510 0.3492 J2(x) 0. Plots of J 0 ( x ) .0306 0.0 -0.20 0. J 1 ( x ) and J 2 ( x ) using Excel The equations J 0 ( x ) = 0 and J 1 ( x ) = 0 exhibit some interesting characteristics.70 0.0012 0. the roots of these equations separate each other.8812 0. Each has an infinite number of distinct real roots 3. 3 ⁄ 2 .1960 0.2423 0.00 0.60 0.0249 0.9258 0. The most noteworthy are: 1.9696 0.9500 0. these series behave like the cos x and sin x functions.15 0.2194 0.2).8642 J1(x) 0.9604 0.8971 0.50 0.0197 0.The Bessel Function Plot of Bessel Function Jn(x) for n = 0. are almost periodic with period almost 2π .0588 0.0437 0.8317 of J 1 ( x ) lies between the roots 2.30 0.0369 0.0995 0. In other words.12) Numerical Analysis Using MATLAB® and Excel®.40 0.9975 0.9994 0. 4.1240 0.35 0. If n is half of an odd integer.3290 0.0 J0(x) J1(x) 0.0028 0.0499 0.75 J0(x) 1. Consider.sin x πx (15. that is. 1 and 2 x 0.0003 0.

1469 10. J 1 ⁄ 2 ⎛ -.pi/4). correct to four decimal places. besselj(−0.1 The first few roots of J 0 ( x ) and J 1 ( x ) J0 ( x ) = 0 Roots 2.1839 Differences … … … … Likewise. Using (15.1735 3.5.= -.= 0.⎞ = ⎝4⎠ 2 1 2 2 -----------------. π J 1 ⁄ 2 ⎛ -.⎞ = ⎝4⎠ 2 1 2 42 ----------------.8 ⋅ -----. and Chebyshev Functions TABLE 15.cos ( π ⁄ 4 ) = -.sin ( π ⁄ 4 ) = -.= 0.cos x πx (15. Third Edition Copyright © Orchard Publications .5201 3.= -.0156 3.2 Compute. the values of a.= ----. J ( –1 ⁄ 2 ) ⎛ -.1153 5.1394 14.⎞ ⎝ ⎠ π 4 Solution: a. if we let n = 1 ⁄ 2 in (15.Chapter 15 Bessel.⎞ ⎝ ⎠ π 4 b.7915 3.9309 16.6366 2 π π 2π π(π ⁄ 4) b.6537 3.13).3).5.4706 13.1336 8.1579 J1 ( x ) = 0 Differences Roots 3.1502 7. we obtain J( –1 ⁄ 2 ) ( x ) = 2 ----.6366 2 π π π(π ⁄ 4) Check with MATLAB: besselj(0. π J ( – 1 ⁄ 2 ) ⎛ -.3237 3.8 ⋅ -----.4048 3. Legendre.8317 3. Using (15.13) Example 15.1378 11.12).pi/4) 15−6 Numerical Analysis Using MATLAB® and Excel®.

and thus their solutions are called Modified Bessel functions. if we let k = n + m in the second summation. If n is an integer. we will prove that J–n ( x ) = ( – 1 ) Jn ( x ) for n = 1. or Hankel functions.3). … n (15. Other Bessel functions. then. and will be explained in the next paragraph. can be obtained by replacing n with – n in (15.6366 The Bessel functions which we have discussed thus far. – 1. the first summation in the above relation is zero for k = 0. that the numbers n = 0.= ( – 1 ) n J n ( x ) Γ ( n + m + 1 )! ⋅ m! m and thus (15. certain differential equations resemble the Bessel equation. 1. 3. … yield infinite values in Γ ( n ) . – 2. These are additional solutions of the Bessel’s equation. Also. J –n ( x ) = = k=0 n–1 k=0 ∑ ∑ ∞ ( –1 ) ⋅ ( x ⁄ 2 ) -----------------------------------------------k! ⋅ Γ ( – n + k + 1 ) k k – n + 2k ( –1 ) ⋅ ( x ⁄ 2 ) -----------------------------------------------. or Weber functions. a Bessel function J –n ( x ) for n > 0 . 2.14) Proof: From (15. Also. As mentioned earlier.14) has been proved. 2.2). ….The Bessel Function ans = 0. Numerical Analysis Using MATLAB® and Excel®. or Neumann functions. we see that m=0 ∑ ∞ ⋅ (x ⁄ 2 ) ( –1 ) ------------------------------------------------------------------( n + m )! ⋅ Γ ( m + 1 ) = ( –1 ) n n+m – n + 2n + 2m m=0 ∑ ∞ ⎛x⎞ -⎝2⎠ n + 2m ( –1 ) -------------------------------------------.15) Now. n – 1 . we recall from Chapter 13.+ k! ⋅ Γ ( – n + k + 1 ) – n + 2k k=n ∑ ∞ ( –1 ) ⋅ ( x ⁄ 2 ) -----------------------------------------------k! ⋅ Γ ( – n + k + 1 ) k – n + 2k (15.5). Third Edition Copyright © Orchard Publications 15−7 . are referred to as Bessel functions of the first kind. denoted as Y n ( x ) and referred to as Bessel functions of the second kind.6366 ans = 0. after simplification and comparison with (15.

15−8 Numerical Analysis Using MATLAB® and Excel®. y represents the general solution of (15. (15. the general solution of the Bessel equation is y = AJ n ( x ) + BJ – n ( x ) n ≠ 0.13) into (15.16) is the general solution. The following example illustrates the fact that when n is not an integer or zero. and Chebyshev Functions It is shown in advanced mathematics textbooks that. and for this reason. therefore.+ C 2 ----------x x (15. (15.14).sin x + B ----.18).12) and (15.16) is not the general solution.1). Example 15.16) For n = 1. if n is not an integer.19) and letting C 1 = A 2 ⁄ π and C 2 = B 2 ⁄ π .17) We will show that the general solution of (15. 3. … and so on.3 Find the general solution of Bessel’s equation of order 1 ⁄ 2 .cos x πx πx (15. Legendre.18) By substitution of (15. Third Edition Copyright © Orchard Publications .Chapter 15 Bessel. 3. for this case. 2. we obtain 2 2 y = A ----. that is.y = 0 dx ⎝ 4⎠ (15. 1.20) are linearly independent. for this case. the functions J n ( x ) and J –n ( x ) are not linearly independent as we have seen in (15.17) is y = AJ 1 ⁄ 2 ( x ) + BJ –1 ⁄ 2 ( x ) (15.17). we obtain x 2 dy dx 2 2 +x dy ⎛ 2 1 ⎞ + x – -. … (15. J n ( x ) and J –n ( x ) are linearly independent. 2. the Bessel functions of the second kind are introduced to obtain the general solution.20) Since the two terms on the right side of (15. relation (15.19) can be written as sin x cos x y = C 1 ---------. these two series produce only one solution. Solution: By the substitution n = 1 ⁄ 2 in (15.

and examples for each can be found by invoking help bessel for MATLAB. if a and b are distinct roots of J 0 ( x ) = 0 .t – 1 ⎞ -2⎝ t⎠ 1 = 0 (15. The descriptions.25) * Two functions constitute an orthogonal system. they constitute an orthogonal system. ∫0 π ⎧ πJ n ( x ) n = even cos nφ ⋅ cos ( x sin φ ) dφ = ⎨ n = odd ⎩ 0 n = even ⎧0 sin n φ ⋅ sin ( x sin φ ) dφ = ⎨ ⎩ πJ n ( x ) n = odd ∫0 π (15. They are also orthog- = ∑ Jn ( x )t n = –∞ ∞ n (15. Also.22) is referred to as the generating function for Bessel functions of the first kind of integer order. and others. then. The function e x⎛ -. syntax. Third Edition Copyright © Orchard Publications 15−9 . Numerical Analysis Using MATLAB® and Excel®. can be evaluated at specified values either with MATLAB or Excel. whereas 2k – 1 in the second. More detailed discussion and proofs can be found in advanced mathematics textbooks. ∑ Jn ( x ) n = –∞ 2 ∞ = 1 ∞ (15. One very important property of the Bessel’s functions is that within certain limits. Using this function. J 0 ( a ) = 0 and J 0 ( b ) = 0 .24) ∑1 J2k – 1 ( x ) sin ( 2k – 1 ) φ k= where the subscript 2k denotes that the first relation is valid for even values of k . third kind.23) cos ( x sin φ ) = J 0 ( x ) + 2 sin ( x sin φ ) = 2 ∞ k=1 ∑ J2k ( x ) cos 2kφ (15.21) and we say that J 0 ( ax ) and J 0 ( bx ) are orthogonal in the interval 0 ≤ x ≤ 1 . when the average of their cross product is zero within some specified limits. Some of these are given below without proof. we can obtain several interesting properties for integer values of n . ∫0 xJ0 ( ax )J0 ( bx ) dx onal with the variable x . indicates that the second relation is valid for odd values of k . and help for Excel.The Bessel Function The Bessel functions of the second kind.* For instance.

Chapter 15 Bessel.4 ) y 1 = a 0 1 – ------------------.27) is y = y 1 + y 2 or n ( n + 1 ) 2 ( n – 2 )n ( n + 1 ) ( n + 3 . n is a constant. Applying the method of Frobenius.26) appear in frequency modulation. For example. the average power is shown to be 1 2 P ave = -. Then. while y 2 is an odd function. n ( n + 1 ) 2 ( n – 2 )n ( n + 1 ) ( n + 3 .27) known as Legendre’s equation.x + -----------------------------------------------------. we obtain two independent solutions y 1 and y 2 as follows. and if it is zero or a positive integer.x – … 2! 4! ( n – 1 )( n + 2) 3 ( n – 3)( n – 1)( n + 2)( n + 4 ) 5 y 2 = a 1 x – --------------------------------. Here.2 Legendre Functions Another second−order differential equation with variable coefficients is the equation dy 2 d y ( 1 – x ) --------. Third Edition Copyright © Orchard Publications .30) and this series is absolutely convergent* for x > 1 .A C 2 and with (15. as in the Bessel equation. the general solution of (15.x + -------------------------------------------------------------------. and Chebyshev Functions and 1 J n ( x ) = -π π ∫0 cos ( nφ – x sin φ ) dφ ∑ Jn ( β ) n = –∞ 2 ∞ (15.+ n ( n + 1 )y = 0 2 dx dx 2 (15.23) through (15.A C 2 15.29) where a 0 and a 1 are constants.– 2x ----.x + -------------------------------------------------------------------.23).x – … 3! 5! (15.27) has polynomial solutions of special interest.x – … 3! 5! (15.28) (15. it reduces to 1 2 P ave = -. then (15.26) Relations (15.x – … 2! 4! (n – 1)(n + 2) 3 (n – 3)(n – 1)(n + 2)(n + 4) 5 + a 1 x – --------------------------------. 15−10 Numerical Analysis Using MATLAB® and Excel®. We observe that y 1 is an even function of x .x + -----------------------------------------------------. Legendre.4 ) y = a 0 1 – ------------------.

x + … 3! 2 ⋅ 4 ⋅ 6 ⋅ … ⋅ (n – 1) for n = odd integer (15.1 – ------------------.x n – 2k n 2 k! ( n – k )! ( n – 2k )! k (n – 1) n where N = -.33) From (15.for n = odd 2 2 (15.x – --------------------------------. The Legendre polynomials are defined as Pn ( x ) = ( –1 ) n⁄2 n(n + 1.32) can be combined to a single relation as Pn ( x ) = k=0 ∑ N ( – 1 ) ⋅ ( 2n – 2k )! --------------------------------------------------. whenever n is zero or a positive integer. * Assume that the infinite series ∑ un ( x0 ) n=1 ∞ = u 1 ( x 0 ) + u 2 ( x 0 ) + … converges. Therefore. The even and odd functions of (15. Third Edition Copyright © Orchard Publications 15−11 . If.Legendre Functions The parameter n is usually a positive integer.2 ) 1 ⋅ 3 ⋅ 5 ⋅ … ⋅ (n – 1) ⋅ ------------------------------------------------. when we replace the terms of this series by their absolute value. the first term on the right side of (15.30) contains only a finite number of terms. they will not be discussed in this text. reaches a limit. this series is said to be absolutely convergent. Accordingly. if it is odd.x + … 2! 2⋅4⋅6⋅…⋅n (15. The infinite series solution Q n ( x ) is referred to as Legendre functions of the second kind. or (15.e. i.31) and (15.31) and (15.for n = even and N = ---------------.33). ∑ un ( x0 ) n=1 ∞ = u 1 ( x 0 ) + u 2 ( x 0 ) + … also Numerical Analysis Using MATLAB® and Excel®. and an infinite series solution which is denoted as Q n ( x ) . we find that the resulting series converges. If n is zero. These become infinite as x → ± 1 and their applications to science and engineering problems are very limited. we obtain the following first 6 Legendre polynomials.32) and these are also referred to as surface zonal harmonics. the second term contains only a finite number of terms. the general solution of Legendre’s equation contains a polynomial solution which is denoted as P n ( x ) ..31) for n = 0 or n = even integer Pn ( x ) = ( –1 ) (n – 1) ⁄ 2 (n – 1)(n + 2) 3 1⋅3⋅5⋅…⋅n ⋅ ------------------------------------------------. or an even positive integer.32).

Another important identity involving Legendre polynomials. of equal magnitude but of opposite sign or polarity.x n – 2k k! ( n – k )! ( n – 2k )! n k (15. From the binomial theorem.( 35x – 30x + 3 ) 8 P1 ( x ) = x 1 3 P 3 ( x ) = -. (x – 1) = 2 n k=0 n ∑ ( – 1 ) ⋅ n! --------------------------.38) We will illustrate the use of the Legendre polynomials with the following example. * A dipole is a pair of electric charges or magnetic poles.⋅ -------. separated by a small distance.( x 2 – 1 ) = n dx n k=0 ∑ N ( – 1 ) ⋅ n! ( 2n – 2k )! ----------------------.33). a dipole is an antenna. and offers another method of expressing the Legendre polynomials.36) and differentiation of (15.x 2n – 2k k! ⋅ ( n – k )! k (15.37) Now.34) The relation n n 1 d . by comparison with (15. 15−12 Numerical Analysis Using MATLAB® and Excel®. Example 15.( x – 1 ) n n 2 ⋅ n! dx (15.35) is known as Rodrigues’ formula.2 P n ( x ) = -------------.( 5x – 3x ) 2 1 5 3 P 5 ( x ) = -.Chapter 15 Bessel. consisting of two equal rods extending outward in a straight line. We prove (15.35) is proved. we recognize (15. Alternately. Third Edition Copyright © Orchard Publications .⋅ -----------------------.36) with respect to x n times yields n d ------. and the angle which point P makes with the center of the dipole.( 63x – 70x + 15x ) 8 (15.35) as follows.( 3x – 1 ) 2 1 4 2 P 4 ( x ) = -.= P 0 ( x ) + P 1 ( x )t + P 2 ( x )t + … + P n ( x )t = 2 1 – 2xt + t ∑ Pn ( x )t n=0 ∞ n (15. usually fed from the center. is the generating function for Legendre polynomials which is defined as 2 n 1 ------------------------------. Legendre.4 Find the potential difference (voltage) v at a point P developed by a nearby dipole* in terms of the distance between the point P and the dipole. and Chebyshev Functions P0 ( x ) = 1 1 2 P 2 ( x ) = -.37) as 2 ⋅ n! ⋅ P n ( x ) and thus (15.

⎝r r ⎠ 2 1 (15.43) or * Permittivity is a measure of the ability of a material to resist the formation of an electric field within it. we will denote the quantity 1 ⁄ ( 4πε 0 ) as k and thus we rewrite (15.= r d 2d cos θ ----.4πε 0 ⎝ r 2 r 1 ⎠ (15. we need to express r 1 and r 2 in terms or d and r.40) Next.⎛ --. Figure for Example 15. Third Edition Copyright © Orchard Publications 15−13 .⎞ = kq ( r 2 – r 1 ) .– --.39) as –1 –1 1 1 V P = kq ⎛ --. we obtain r2 --. By the law of cosines.4 Let the potential at point P be V P . Numerical Analysis Using MATLAB® and Excel®. P r1 r θ −q d 0 d q r2 Figure 15.39) where ε 0 is the permittivity* of the vacuum. From electromagnetic field textbooks we find that q .3.⎞ .+ 1 – -----------------2 r r 2 (15.Legendre Functions Solution: Let the charges q and – q of the dipole be a distance 2d apart with the origin 0 as the midpoint as shown in Figure 15. For simplicity.– --.42) and d + r – 2 dr cos θ Dividing both sides of (15. r1 = d + r – 2 dr cos ( 180° – θ ) = r2 = 2 2 2 2 d + r + 2dr cos θ 2 2 (15.1 1 V P = ----------.41) (15.42) by r .3.

and y = d ⁄ r . This requirement is satisfied since x and y.44) we obtain ⎛ d2 2d cos θ⎞ ⎜ ----.⎟ 2 r ⎠ ⎝ r –1 ⁄ 2 = ( 1 – 2xy + y ) 2 –1 ⁄ 2 = ∑ Pn ( x )y n=0 ∞ n (15. Then. P n ( – cos θ ) = – P n ( cos θ ) and the odd powers in (15. thus.49) and for r » d .40).Chapter 15 Bessel.+ 1 – ----------------.48).P 1 ( cos θ ) ⎛ -. all even powers vanish. we simply replace x with – x in (15. and Chebyshev Functions 1⎛ d 2 –1 2d cos θ⎞ r 2 = -.49) can be approximated as d 2kdq 2kq V P ≈ -------.⎞ = -----------.45) holds only if x < 1 and y < 1 .⎜ ----. (15.47) Since x = cos θ . and therefore.45) We recall that (15. But when n is odd. We let x = cos θ .⎞ ⎝r ⎠ (15. and y = d ⁄ r . and thus ( 1 + 2xy + y ) 2 –1 ⁄ 2 –1 = ∑ Pn ( –x )y n=0 n n ∞ n (15. to a Legendre polynomial. We do this by expressing these terms in the form of the generating function of (15. Legendre. To find a similar expression for r 1 . we can express (15.48) However.⎞ ⎝r ⎠ (15. kq V P = ----r ∞ n=0 ∑ d 2n + 1 2P 2n + 1 ( cos θ ) ⎛ -.45) and (15. by substitution into (15. we obtain kq V P = ----r ∑ [ Pn ( x ) y n=0 ∞ – P n ( – x )y ] (15.+ 1 – ----------------.⎟ r ⎝ r2 r ⎠ –1 ⁄ 2 (15. Third Edition Copyright © Orchard Publications .cos θ * 2 ⎝r ⎠ r r (15.44) In all practical applications. we assume that r > d .50) 15−14 Numerical Analysis Using MATLAB® and Excel®.48) are duplicated.47) as kq V P = ----r ∑ n=0 ∞ d n [ P n ( cos θ ) – P n ( – cos θ ) ] ⎛ -.46) into (15. are both less than unity.38). the point P is sufficiently far from the origin. as defined.45). then. P n ( – cos θ ) = P n ( cos θ ) .46) By substitution of (15. if n is even in (15. Now.44). we want to relate the terms inside the parentheses of (15.

r.50) can.52) is the same as (15.53) * Here.( r 2 – r 1 ) = -----------.⎞ = -----------. It is denoted with the letter p . of course. the distances r 1 .52) We observe that (15. the negative and positive charges look like a single point charge.2 ⎝r r ⎠ r1 ⋅ r2 r 2 1 (15.cos θ . = cos θ .– --. Then.50). and using (15. P r1 × r2 = r 2 r2 r 2 – r 1 = 2d cos θ r r1 θ θ −q d 0 d q Figure 15.40) we obtain kq 2kqd 1 1 V P = kq ⎛ --. Third Edition Copyright © Orchard Publications 15−15 . we have used the identity P1 ( cos θ ) form of the Legendre polynomials.Legendre Functions The term 2dq is the magnitude of the so−called dipole moment.4. This will be seen shortly in (15.51) The relation of (15.4. and r 2 can be approximated by parallel lines as shown in Figure 15.57) when we discuss the trigonometric Numerical Analysis Using MATLAB® and Excel®. It is a vector directed from the negative charge towards the positive charge. is 2n + 1 B n = -------------2 ∫–1 f ( x )Pn ( x ) dx 1 (15. p = 2qd (15. be derived without the use of Legendre polynomials as follows: For r » d . Derivation of the voltage developed by a dipole Another interesting relation that can be used to find the Legendre polynomial series of a function f ( x ) for x < 1 . that is.

then. we will substitute these into (15. Waveform for Example 15.( 5x – 3x ) dx + -2 2 –1 0 ∫0 1 ⋅ -. We will also use (15.x 2 0 –1 0 1 + -.( 3x 2 1 – 1 ) dx = 0 ∫ 1 7 3 ( – 1 ) ⋅ -. Legendre. For this example.54).34) for the polynomials of P n ( x ) . and Chebyshev Functions Then.4 4 2 5 B 2 = -2 7 B 3 = -2 9 B 4 = -2 ∫ 1 5 2 ( – 1 ) ⋅ -.( 35x 8 1 – 30x + 3 ) = 0 2 15−16 Numerical Analysis Using MATLAB® and Excel®. Example 15. 1 B 0 = -2 3 B 1 = -2 ∫ ∫ 1 ( – 1 ) ⋅ 1 ⋅ dx + -2 –1 0 0 ∫0 1 ⋅ 1 ⋅ d x ∫0 1 1 1 = – -.( 35x – 30x + 3 ) dx + -8 2 –1 0 ∫0 1 ⋅ -.Chapter 15 Bessel.53). a function f ( x ) can be expanded as f ( x ) = B0 P0 ( x ) + B1 P1 ( x ) + B2 P2 ( x ) + … + Bn Pn ( x ) = ∑0 Bn Pn ( x ) n= ∞ (15.5 Compute the Legendre polynomial series representing the waveform of Figure 15.5.-.2 2 1 0 3 ( – 1 ) ⋅ x ⋅ dx + -2 –1 3 2 1 ⋅ x ⋅ dx = – -. Third Edition Copyright © Orchard Publications .( 5x 2 1 1 3 -– 3x ) dx = – 7 8 4 ∫ 1 9 4 2 ( – 1 ) ⋅ -.x –1 4 1 2 3 3 3 = -.x 2 1 0 = –1+1 = 0 -. f (x) 1 −1 0 −1 1 x Figure 15.5 Solution: We will first compute the coefficients B n from (15.5.x 4 1 3 2 + -.54) The example below illustrates how this relation is being used.+ -.( 3x – 1 ) dx + -2 2 –1 0 ∫0 1 ⋅ -.= -.

2 cos ( n – 2 ) θ 2 ⋅ 4 ⋅ … ⋅ 2n 2 2 ⋅ 4 ⋅ … ⋅ ( 2n – 2 ) 1 ⋅ 3 1 ⋅ 3 ⋅ … ⋅ ( 2n – 5 ) + --------. P 0 cos θ = 1 P 1 cos θ = cos θ 3 cos 2 θ + 1 P 2 cos θ = --------------------------4 5 cos 3 θ + 3 cos θ P 3 cos θ = --------------------------------------8 35 cos 4 θ + 20 cos 2 θ + 9 P 4 cos θ = ---------------------------------------------------------64 ------------------------------------------------------------------------P 5 cos θ = 63 cos 5 θ + 35 cos 3 θ + 30 cos θ 128 (15.x 64 128 128 (15.P 1 ( x ) – -. Third Edition Copyright © Orchard Publications 15−17 ..( 63x – 70x + 15x ) 8 2 16 8 2 7113 3 5 3 = -.⋅ -------------------------------------------.⋅ -.⋅ -------------------------------------------.4. satisfy the orthogonality principle when m ≠ n as indicated by the following integral. Therefore.( 63x 5 – 70x 3 + 15x ) dx + ----8 2 –1 0 ∫0 -.34) we obtain 3 7 11 f ( x ) = -. Also.P 5 ( x ) 2 8 16 7 1 11 1 3 3 5 3 . In many applications.x + -------.= -.55) We observe that the waveform of f ( x ) is an odd function and.( 5x – 3x ) + ----.⋅ -. the algebraic form of the Legendre polynomials is usually the most useful. using (15. Numerical Analysis Using MATLAB® and Excel®.2 cos ( n – 4 ) θ + … 2 ⋅ 4 2 ⋅ 4 ⋅ … ⋅ ( 2n – 4 ) (15.⋅ ( 63x – 70x + 15x ) 16 128 2 525 3 693 5 525 = -------. It is shown in advanced mathematics textbooks that 1 ⋅ 3 ⋅ … ⋅ ( 2n – 1 ) 1 1 ⋅ 3 ⋅ … ⋅ ( 2n – 3 ) P n cos θ = -------------------------------------------. However. there are times when we want to express the polynomials in terms trigonometric functions.x – -------.54) and (15.( 63x 8 1 1 5 3 ----– 70x + 15x ) = 11 16 and so on.( 5x – 3x ) + -------.x – ----.56) From (15. as we found above.P 3 ( x ) + ----. the trigonometric forms are most convenient with the cylindrical and spherical coordinate systems. as we did in Example 15.57) The Legendre polynomials in algebraic form.x – -.Legendre Functions – 11 B 5 = -------2 ∫ 1 11 -.2 cos n θ + -.56) we obtain the first 6 Legendre polynomials in trigonometric form listed below. its expansion contains only odd Legendre polynomials.

the Legendre polynomials in trigonometric form satisfy the orthogonality principle when m ≠ n as indicated by the following integral. There is another class of Legendre functions which are solutions of the differential equation 2 dy 2 d y m ( 1 – x ) --------. and Chebyshev Functions ⎧ 0 ⎪ P m ( x )P n ( x ) dx = ⎨ 2 –1 ⎪ -------------⎩ 2n + 1 1 ∫ m≠n m = n (15. Third Edition Copyright © Orchard Publications .60) and this is referred to as the associated Legendre differential equation. the polynomials P n ( x ) .+ n ( n + 1 ) – ------------.35).60) reduces to (15.P n ( x ) m dx d ⋅ --------. We observe that if m = 0 . ∫0 π ⎧ 0 ⎪ P m ( cos θ )P n ( cos θ ) sin θ dθ = ⎨ 2 ⎪ -------------⎩ 2n + 1 m≠n m = n (15. The general solution of (15.61) where C 1 and C 2 are arbitrary constants.63) Relations (15. that is.Chapter 15 Bessel. Legendre.63) are also known as spherical harmonics.– 2x ----.59) We must remember that all the Legendre polynomials we have discussed thus far are referred to as surface zonal harmonics.y = 0 2 2 dx dx 1–x 2 (15.62) and Qn ( x ) = ( –1 ) ( 1 – x ) m m 2 m⁄2 (15.Q n ( x ) m dx m m (15.58) Similarly. The functions P n ( x ) and Q n ( x ) are referred to as associated Legendre functions of the first and second kind respectively. These are evaluated from Pn ( x ) = ( –1 ) ( 1 – x ) m m 2 m⁄2 d ⋅ --------.62) and (15. (15. m 15−18 Numerical Analysis Using MATLAB® and Excel®. We will restrict our subsequent discussion to the associated Legendre functions of the first kind.60) is y = C1 Pn ( x ) + C2 Qn ( x ) m m m m (15.27). and math tables include values of these as computed from Rodrigues’ formula of (15.

Pn ( x ) = ( –1 ) ( 1 – x ) m m 2 m⁄2 1 x = 0. second.x) function computes the associated Legendre functions of the first kind of degree n and order m = 0.-------------------. that is.5 c.5 As stated above.⎛ 3x – 1⎞ dx ⎝ 2 ⎠ 2 x = 0. for the first. Example 15. a.5) The values for m = 0. ….P 2 ( x ) dx = –( 1 – x ) x = 0. Here. n = 2 and thus MATLAB will return a matrix whose rows correspond to the values of m = 0 . P 3 ( x ) x = – 0.62).62). n evaluated for each element of x .⎞ ⎝ 2 ⎠ 2 = – 0.125 x = 0. we obtain P2 ( x ) 1 x = 0.6 Find the following associated Legendre functions and evaluate as indicated. and m = 2 . and third rows respectively.P n ( x ) m dx m and the appropriate relations of (15.5 ( 3x ) = – 1. and order m = 0. the MATLAB legendre(n. Check with MATLAB: disp('The values for m = 0. we use the relation (15. n evaluated for each element of x .2990 For m = 0 in (15. m = 1 .x) function that computes the associated Legendre functions of the first kind of degree n .Legendre Functions At present. 2. P 2 ( x ) Solution: For this example.34).5 2 1⁄2 = ( –1 ) ( 1 – x ) = –( 1 – x ) 2 1⁄2 1 d----. 2.0. a. Excel does not have any functions related to Legendre polynomials. ….25 d ⋅ --------. m = 1 and m = 2 are:'). 1. P2 ( x ) 1 x = 0. Third Edition Copyright © Orchard Publications 15−19 .5 3x – 1 = ⎛ ---------------.5 2 1⁄2 d. For this example. MATLAB provides the legendre(n. m = 1 and m = 2 are: Numerical Analysis Using MATLAB® and Excel®. legendre(2.5 2 3 b. P 2 ( x ) x = 0. 1.

and thus MATLAB will display a matrix whose rows correspond to the values of m = 0 .5).25 2 3⁄2 d ------. y(:.0.y') c.6250 -9.5 2 3 x = – 0. for the first. second.5 = ( –1 ) ( 1 – x ) 2 d ------.y') m 0 1 2 b.3248 -5. m 2 3 15−20 Numerical Analysis Using MATLAB® and Excel®.5).⎞ dx ⎝ 2 ⎠ 2 = ( 1 – x ) ( 15x ) x = – 0. and m = 3 . m=0:2.2).1250 -1.0f\t %7.2990 2.2500 or more elegantly.⎛ --------------------⎞ 2⎝ 2 ⎠ dx 2 x = – 0. y(:. then P n ( x ) = 0 .5 5x – 3x 2 d = ( 1 – x ) ------.0f\t %7. Check with MATLAB: m=0:3.4f \n'. it follows that P 2 ( x ) = 0 .25 and since the third derivative of 3x – 1 is zero. Third Edition Copyright © Orchard Publications . y(:.4f \n'.1250 -1. y=zeros(4. Legendre.P 3 ( x ) 2 dx 2 x = – 0. fprintf('%2. fprintf('m\t Legendre \n').2)=legendre(3. fprintf('%2. fprintf('m\t Legendre \n').( 3x 2 – 1 ) 3 dx 3 x = 0.25 = ( –1 ) ( 1 – x ) 3 2 3⁄2 d ------. m = 2 .7428 P2 ( x ) 3 x = 0. In general. if m > n .2).1)=m'. n = 3 . Legendre -0. fprintf('\n').Chapter 15 Bessel. y=zeros(3.2500 2 2⁄2 P3 ( x ) 2 x = – 0.6250 Here.P 2 ( x ) 3 dx 3 = –( 1 – x ) x = 0.5 = – 5. m 0 1 2 3 Legendre 0.4375 -0. y(:.1)=m'. m = 1 .⎛ ------------------. fprintf('\n').2990 2.−0.2)=legendre(2.5 2 d 15x – 3 = ( 1 – x ) ----. third and fourth rows respectively. and Chebyshev Functions ans = -0.

we obtain L0 ( x ) = e L1 ( x ) = e L2 ( x ) = e x d dx 0 0 (x e ) = e e 0 –x x –x = e = 1 0 x –x x –x –x d ( xe ) = e ( e – xe ) = 1 – x dx x d dx 2 2 3 ( x e ) = e e ( 2 – 4x + x ) = 2 – 4x + x 2 –x x –x 2 2 (15. are the Laguerre polynomials L n ( x ) . L 2 ( x ) d. Third Edition Copyright © Orchard Publications 15−21 .Laguerre Polynomials 15.66) Compute the Laguerre polynomials a.3 Laguerre Polynomials Another class of polynomials that satisfy the orthogonality principle.% Differentiate y twice with respect to x Numerical Analysis Using MATLAB® and Excel®.67) x 3 –x 2 3 L 3 ( x ) = e d ( x e ) = 6 – 18x + 9x – x 3 dx The differentiation of the last two polynomials in (15.2).65). L 1 ( x ) c.64) These polynomials are computed with the Rodrigues’ formula x n –x Ln ( x ) = e d ( x e ) n dx n (15. these are solutions of the differential equation x dy dx 2 2 + (1 – x) dy + ny = 0 dx (15. L 0 ( x ) b.67) was performed with MATLAB as follows: syms x y z y=x^2*exp(−x).65) The orthogonality principle for these polynomials states that ∫0 Example 15. L 3 ( x ) Solution: Using Rodrigues’s formula of (15.7 ∞ e L m ( x )L n ( x ) dx = 0 –x (15. z=diff(y.

y=x^3*exp(−x). Legendre. We will restrict our discussion to the T n ( x ) polynomials. simplify(L2x) ans = 2-4*x+x^2 syms x y z w . we will avoid notation C k ( x ) for the Chebyshev polynomials. these are denoted as y = U n ( x ) .2). and Chebyshev Functions L2x=exp(x)*z.Chapter 15 Bessel. (a) 15−22 Numerical Analysis Using MATLAB® and Excel®.* The solutions of (15.68) are referred to as Chebyshev polynomials of the first kind and are denoted as y = T n ( x ) . simplify(L3x) ans = 6-18*x+9*x^2-x^3 15. that is. within this range. * Some books use the notation Ck ( x ) for these polynomials. For x > 1 they increase or decrease more rapidly than any other polynomial of order n . z=diff(y. They exhibit equiripple amplitute characteristics over the range – 1 ≤ x ≤ 1 .% Differentiate y twice % we cannot differentiate three times at once w=diff(z).+ n y = 0 2 dx dx 2 (15. However. Third Edition Copyright © Orchard Publications .% Differentiate one more time L3x=exp(x)*w. another class of orthogonal functions known as Genenbauer or Ultraspherical functions use the notation Cn ( x ) and for this reason. We will plot some of these later in this section. Two interesting properties of the T n ( x ) polynomials are: 1.– 3x ----.68) and dy 2 d y ( 1 – x ) --------.4 Chebyshev Polynomials The Chebyshev polynomials are solutions of the differential equations dy 2 d y 2 ( 1 – x ) --------. 2. they oscillate with the same ripple. Both kinds comprise a set of orthogonal functions.+ n ( n + 2 )y = 0 2 dx dx 2 (15.69) The solutions of (15.69) are the Chebyshev polynomials of the second kind. This property is the basis for the Chebyshev approximation in the design of Chebyshev type electric filters.– x ----.

73).70) can be expressed as a polynomial.2. we obtain T n + 1 ( y ) + T n – 1 ( y ) = 2 cos ny cos y = 2T n ( y )x = 2xT n ( y ) (15. A good reference is the Handbook of Mathematical Functions.2 Chebyshev polynomials expressed in powers of x n 0 1 2 3 4 5 6 Tn ( x ) 1 x 2x – 1 4x – 3x 8x – 8x + 1 16x – 20x + 5x 32x – 48x + 18x – 1 6 4 2 5 3 4 2 3 2 To show that the relation of (15.74) Similarly. Dover Publications.73) and T n ( y ) = cos ny Next. TABLE 15.70) or (15. Third Edition Copyright © Orchard Publications 15−23 . we let x = cos y (15.75). we replace n with n + 1 and we obtain T n + 1 ( y ) = cos ( n + 1 )y = cos ny cos y – sin ny sin y (15. we add (15.71). They can also be derived from the following relations.73) and (15.Chebyshev Polynomials These polynomials are tabulated in reference books which contain mathematical functions. we can express T n ( x ) as polynomials in powers of x . we obtain T n – 1 ( y ) = cos ( n – 1 )y = cos ny cos y + sin ny sin y (15.74) with (15.72). T n ( x ) = cos ( n cos x ) T n ( x ) = cos h ( n cos h x ) –1 –1 for x ≤ 1 for x > 1 (15.75) Now.71) Using (15.70) (15. replacing n with n – 1 . and making use of (15. in (15.72) (15. Some are shown in Table 15.76) Numerical Analysis Using MATLAB® and Excel®.

77).73) yields T0 ( y ) = T0 ( x ) = 1 (15. and Chebyshev Functions or T n + 1 ( y ) = 2xT n ( y ) – T n – 1 ( y ) Then. Third Edition Copyright © Orchard Publications . from (15.81) –1 Alternately. 4 and so on. we can prove the first 3 entries of Table 15. we use the recurrence formula of (15. 3.78) For n = 1 .70) by letting y = cos x . T 3 ( x ) = 2xT 2 ( x ) – T 1 ( x ) = 4x – 2x – x = 4x – 3x 3 (15. T 2 ( x ) = 2xT 1 ( x ) – T 0 ( x ) = 2x – 1 2 (15.77) The polynomials in Table 15. when n = 2 . for n = 0 .72).80) 3 and when n = 3 .Chapter 15 Bessel. T 2 ( x ) = cos ( 2 ⋅ cos x ) = cos ( 2 ⋅ y ) = cos 2y = 2 cos y – 1 ⎧ ⎪ ⎪ ⎨ ⎪ ⎪ ⎩ ⎧ ⎪ ⎪ ⎨ ⎪ ⎪ ⎩ = 2 cos ( cos x ) – 1 = 2 2 –1 –1 2 cos ( cos x ) x 2 –1 ⋅ cos ( cos x ) x –1 –1 or T 2 ( x ) = 2x – 1 Relation (15. can now be verified by using a combination of the above relations. we can replace y with x to obtain T n + 1 ( x ) = 2xT n ( x ) – T n – 1 ( x ) Recurrence Relation (15. (15.79) To derive the algebraic expressions corresponding to n = 2.70) as follows: We recall that 15−24 Numerical Analysis Using MATLAB® and Excel®. T 1 ( x ) = cos ( 1 ⋅ cos x ) = cos ( 1 ⋅ y ) = cos y = x –1 and for n = 2 . Thus. For instance. Legendre.73) and (15. we obtain T1 ( y ) = T1 ( x ) = x (15.2. T 0 ( x ) = cos ( 0 ⋅ cos x ) = cos ( 0 ⋅ y ) = 1 –1 For n = 1 . Thus.2 with (15. for n = 0 .71) can be derived from (15.

title('Chebyshev Polynomials of the First Kind'). Tnx3=cos(3*acos(x)). Tnx0=cos(0*acos(x)).2:0.. x.2 1.70).84) (15. Tnx5=cos(5*acos(x)). jα = cosh v . xlabel('x'). x. Tnx4.Chebyshev Polynomials e +e cos α = ----------------------2 jα –j α (15.83) (15.70) and (15. Tnx1=cos(1*acos(x)). plot(x. cos x = – j cosh x * –1 –1 By substitution into (15.. and that cosh ( – t ) = cosh t . if n = 3 . making use of (15. x. Tnx5. ylabel('Tn(x)') % We could have used the gtext function to label the curves but it is easier with the Figure text % tool * Let cos α = cosh jα = v . x.01:1.. Tnx4=cos(4*acos(x)). expand(cosh(3*acosh(x))) ans = 4*x^3-3*x ans = 4*x^3-3*x The MATLAB script below plots the T n ( x ) for n = 0 through n = 6 . then α = cos v . Tnx2.85) follows. axis([−1. grid. syms x. For example.5]).85). Tnx1. we obtain T n ( x ) = cos [ n ( – j cosh x ) ] = cos ( – jn cosh x ) = cosh ( jnj cosh x ) = cosh [ j ( – jn cosh x ) ] = cosh ( n cosh x ) –1 –1 –1 –1 –1 and this is the same as (15. Tnx0.2 −1..5 1. x. Tnx6).2. Third Edition Copyright © Orchard Publications 15−25 .82) and e +e cos h α = ------------------2 α –α (15. % Chebyshev polynomials % x=−1. expand(cos(3*acos(x))). Tnx6=cos(6*acos(x)). We can also use MATLAB to convert (15. x.71) to polynomials. Tnx3. –1 –1 –1 –1 Numerical Analysis Using MATLAB® and Excel®.71). Tnx2=cos(2*acos(x)). cos α = cos hj α and when x > 1 .85) Then. j cos v = cosh v and (15.

among other applications. For example. if we want to design a second order Chebyshev low−pass filter. ISBN 0−9744239−9−8. are used in the design of electric filters.6 shows the plot of the Chebyshev polynomials of the first kind T n ( x ) for n = 0 through n = 6 .5 T5 -1 T4 T6 T2 T3 -1.8 -0.* The filters are described in terms of rational polynomials that approximate the behavior of ideal filters.2 0 x 0. 15−26 Numerical Analysis Using MATLAB® and Excel®.86) becomes α A ( ω ) = ----------------------------------------------------2 2 1 + ε [ 2 ( ω ⁄ ωC ) – 1 ] 2 (15.5 -1 -0. Orchard Publications. Chebyshev polynomials.6 -0. Third Edition Copyright © Orchard Publications .8 1 Figure 15.87) * For a thorough discussion on the design of analog and digital filters refer to Signals and systems with MATLAB Applications. and Chebyshev Functions Figure 15.4 0.5 Tn(x) 0 -0. Legendre. Plot of Chebyshev polynomials with MATLAB As mentioned earlier. and α and ε are other parameters that are used to specify the order and type of the electric filter.6 0.6. we use the Chebyshev polynomial T 2 ( x ) = 2x – 1 2 and (15. ω C is the cutoff frequency.86) where ω is the operating radian frequency. 1.5 Chebyshev Polynomials of the First Kind 1 T0 T1 0.Chapter 15 Bessel.2 0.4 -0. The basic Chebyshev low−pass filter approximation is defined as α A ( ω ) = -----------------------------------------2 2 1 + ε T n ( ω ⁄ ωC ) 2 (15.

– … 2 2 4 2 6 2 8 2 2 ⋅ ( 1! ) 2 ⋅ ( 2! ) 2 ⋅ ( 3! ) 2 ⋅ ( 4! ) x x x x x J 1 ( x ) = -. fractional.+ ----------------------.– … 2 4 6 8 10 2 ⋅ 2! 2 ⋅ 1! ⋅ 3! 2 ⋅ 2! ⋅ 4! 2 ⋅ 3! ⋅ 5! 2 ⋅ 4! ⋅ 6! 2 4 6 8 10 3 5 7 9 2 4 6 8 • Two more useful relations are d J ( x ) = –J ( x ) 1 dx 0 Numerical Analysis Using MATLAB® and Excel®.– --------------------.– ----------------------. and the most common are the Method of Frobenius and the Method of Picard. or even a complex number.+ --------------------.– … 2 2 3 ⋅ 1! ⋅ 2! 2 5 ⋅ 2! ⋅ 3! 2 7 ⋅ 3! ⋅ 4! 2 9 ⋅ 4! ⋅ 5! x x x x x J 2 ( x ) = -------------. Third Edition Copyright © Orchard Publications 15−27 . • The series Jn ( x ) = k=0 ∑ ∞ x n + 2k 1 k ⋅ --------------------------------------( – 1 ) ⋅ ⎛ -. • Bessel functions are solutions of the differential equation x2 dy dx 2 2 +x dy 2 2 + ( x – n )y = 0 dx where n can be any number.– ----------------------. The general solution depends on the value of n .+ ------------------------. 1 and 2 the series reduce to x x x x J 0 ( x ) = 1 – --------------------.+ --------------------.– ----------------------.⎞ ⋅ ------------------------------------------⎝2⎠ k! ⋅ Γ ( – n + k + 1 ) is referred to as the Bessel function of negative order n . positive or negative integer.+ ----------------------. • For n = 0.Summary 15. • The series J –n ( x ) = k=0 ∑ ( –1 ) ∞ k x – n + 2k 1 ⋅ ⎛ -.5 Summary • Differential equations with variable coefficients cannot be solved in terms of familiar functions as those which we encountered in ordinary differential equations with constant coefficients.+ ----------------------.– ----------------------.⎞ ⎝2⎠ k! ⋅ Γ ( n + k + 1 ) n≥0 where n is any positive real number or zero is referred to as Bessel function of order n . The usual procedure is to derive solutions in the form of infinite series.

then. and thus their solutions are called Modified Bessel functions.n) function. Legendre. Also.+ n ( n + 1 )y = 0 2 dx dx 2 2 where n is a constant. denoted as Y n ( x ) and referred to as Bessel functions of the second kind. ∫0 xJ0 ( ax )J0 ( bx ) dx • The differential equation 1 = 0 and thus we say that J 0 ( ax ) and J 0 ( bx ) are orthogonal in the interval 0 ≤ x ≤ 1 . for this case. certain differential equations resemble the Bessel equation. They also can be found in math table books. dy d y ( 1 – x ) --------. and Chebyshev Functions d { xJ 1 ( x ) } = xJ 0 ( x ) dx • Values of J n ( x ) can be calculated using the appropriate series given above.sin x and J ( – 1 ⁄ 2 ) ( x ) = πx 2 ----. 2. or Neumann functions exist. other Bessel functions. the general solution of the Bessel equation is y = AJ n ( x ) + BJ – n ( x ) n ≠ 0. J 0 ( a ) = 0 and J 0 ( b ) = 0 . or Hankel functions. denoted as Q n ( x ) . • The infinite series solution of the Legendre functions. is referred to as Legendre functions of the second kind. • If n is not an integer.x) function or the Excel BESSELJ(x. J n ( x ) and J –n ( x ) are linearly independent. is known as Legendre’s equation. 3. • Besides the above functions known as Bessel functions of the first kind.cos x πx are known as half−order Bessel functions. … • If a and b are distinct roots of J 0 ( x ) = 0 . Third Edition Copyright © Orchard Publications . and can also be found with the MATLAB besselj(n. 1. or Weber functions.Chapter 15 Bessel.– 2x ----. 15−28 Numerical Analysis Using MATLAB® and Excel®. • The Bessel functions J1 ⁄ 2 ( x ) = 2 ----.

is 2n + 1 B n = -------------2 ∫–1 f ( x )Pn ( x ) dx ∑ Bn Pn ( x ) n=0 ∞ 1 and with this relation we can find a polynomial f ( x ) defined as f ( x ) = B0 P0 ( x ) + B1 P1 ( x ) + B2 P2 ( x ) + … + Bn Pn ( x ) = • The trigonometric form of the Legendre polynomials is 1 ⋅ 3 ⋅ … ⋅ ( 2n – 1 ) 1 1 ⋅ 3 ⋅ … ⋅ ( 2n – 3 ) P n cos θ = -------------------------------------------.x n – 2k n 2 k! ( n – k )! ( n – 2k )! k (n – 1) n where N = -. Numerical Analysis Using MATLAB® and Excel®.2 cos n θ + -. and offers another method of expressing the Legendre polynomials.( x – 1 ) n n 2 ⋅ n! dx is known as Rodrigues’ formula.Summary • The Legendre polynomials are defined as Pn ( x ) = k=0 ∑ N ( – 1 ) ⋅ ( 2n – 2k )! --------------------------------------------------.2 P n ( x ) = -------------.( 5x – 3x ) 2 1 5 3 P 5 ( x ) = -.d .2 cos ( n – 4 ) θ + … 2 ⋅ 4 2 ⋅ 4 ⋅ … ⋅ ( 2n – 4 ) and the first 6 Legendre polynomials in trigonometric form listed below.for n = even and N = ---------------. • The Legendre polynomial series of a function f ( x ) for x < 1 .2 cos ( n – 2 ) θ 2 ⋅ 4 ⋅ … ⋅ 2n 2 2 ⋅ 4 ⋅ … ⋅ ( 2n – 2 ) 1 ⋅ 3 1 ⋅ 3 ⋅ … ⋅ ( 2n – 5 ) + --------.⋅ -------------------------------------------.( 35x – 30x + 3 ) 8 P1 ( x ) = x 1 3 P 3 ( x ) = -.( 63x – 70x + 15x ) 8 • The relation n n 1 .( 3x – 1 ) 2 1 4 2 P 4 ( x ) = -. Third Edition Copyright © Orchard Publications 15−29 .⋅ -------.for n = odd 2 2 and the first 6 Legendre polynomials are P0 ( x ) = 1 1 2 P 2 ( x ) = -.⋅ -------------------------------------------.

Legendre. and Chebyshev Functions P 0 cos θ = 1 P 1 cos θ = cos θ 3 cos 2 θ + 1 P 2 cos θ = --------------------------4 5 cos 3 θ + 3 cos θ P 3 cos θ = --------------------------------------8 35 cos 4 θ + 20 cos 2 θ + 9 P 4 cos θ = ---------------------------------------------------------64 ------------------------------------------------------------------------P 5 cos θ = 63 cos 5 θ + 35 cos 3 θ + 30 cos θ 128 • The Legendre polynomials in algebraic form. The functions P n ( x ) and Q n ( x ) are referred to as associated Legendre functions of the first and second kind respectively. satisfy the orthogonality principle when m ≠ n as indicated by the integral ∫ ⎧ 0 ⎪ P m ( x )P n ( x ) dx = ⎨ 2 –1 ⎪ -------------⎩ 2n + 1 1 m≠n m = n • The Legendre polynomials in trigonometric form satisfy the orthogonality principle when m ≠ n as indicated by the integral ∫0 π ⎧ 0 ⎪ P m ( cos θ )P n ( cos θ ) sin θ dθ = ⎨ 2 ⎪ -------------2n + 1 ⎩ m≠n m = n • The differential equation 2 dy 2 d y m ( 1 – x ) --------.y = 0 2 2 dx dx 1–x 2 is referred to as the associated Legendre differential equation. Third Edition Copyright © Orchard Publications .P n ( x ) m dx m 15−30 Numerical Analysis Using MATLAB® and Excel®. also known as spherical harmonics.– 2x ----. These functions. The general solution of this equation is y = C1 Pn ( x ) + C2 Qn ( x ) m m where C 1 and C 2 are arbitrary constants.+ n ( n + 1 ) – ------------. are evaluated from the relations Pn ( x ) = ( –1 ) ( 1 – x ) m m 2 m⁄2 m m d ⋅ --------.Chapter 15 Bessel.

Third Edition Copyright © Orchard Publications 15−31 .Q n ( x ) m dx m • The MATLAB legendre(n. they oscillate with the same ripple as shown in Figure 15. …. • The T n ( x ) polynomials are derived from the relations T n ( x ) = cos ( n cos x ) T n ( x ) = cos h ( n cos h x ) –1 –1 for x ≤ 1 for x > 1 These polynomials exhibit equiripple amplitute characteristics over the range – 1 ≤ x ≤ 1 .x) function computes the associated Legendre functions of the first kind of degree n . These polynomials are satisfy the orthogonality principle.Summary Qn ( x ) = ( –1 ) ( 1 – x ) m m 2 m⁄2 d ⋅ --------.+ n ( n + 2 )y = 0 2 dx dx 2 The solutions of the first differential equation are referred to as Chebyshev polynomials of the first kind and are denoted as y = T n ( x ) . that is. and order m = 0. 2. Numerical Analysis Using MATLAB® and Excel®. 1.– 3x ----. They are computed with the Rodrigues’ formula x n –x Ln ( x ) = e d ( x e ) n dx n • The Chebyshev polynomials are solutions of the differential equations dy 2 d y 2 ( 1 – x ) --------.6.+ n y = 0 2 dx dx 2 and dy 2 d y ( 1 – x ) --------. • The solutions of the differential equation x dy dx 2 2 + (1 – x) dy + ny = 0 dx are known as Laguerre polynomials and are denoted as L n ( x ) . these are denoted as y = U n ( x ) .– x ----. This property is the basis for the Chebyshev approximation in the design of Chebyshev type electric filters. Both kinds comprise a set of orthogonal functions. within this range. n evaluated for each element of x . The solutions of the second are the Chebyshev polynomials of the second kind.

e. J 1 ⁄ 2 ( π ⁄ 6 ) d.Chapter 15 Bessel.25 3. J 0 ( 3 ) b. J 1 ( 2 ) c.75 c. P 1 ( x ) d.. P 2 ( x ) e. i. use MATLAB or Excel to plot g ( x ) and compare with f ( x ) . a.5 m b.5 x = – 0. Use the appropriate series of the Bessel functions J n ( x ) to compute the following values using the first 4 terms of the series and check your answers with MATLAB or Excel. P 2 ( x ) 3 x = 0. P 3 ( x ) 2 x = 0.6 Exercises 1. and Chebyshev Functions 15. Third Edition Copyright © Orchard Publications . P 0 ( x ) through P 4 ( x ) will be sufficient.5 x = 0. The first 5 terms of P n ( x ) . and check your answers with MATLAB. Use the appropriate Legendre polynomials P n ( x ) or P n ( x ) and Rodrigues’s formulas to compute the following. f (x) 1 −1 1 x 15−32 Numerical Analysis Using MATLAB® and Excel®. Legendre. Compute the Legendre polynomial g ( x ) representing the waveform f ( x ) of the figure below. P 3 ( x ) f.25 x = 0. P 1 ( x ) 2 x = 0. Then. J –1 ⁄ 2 ( π ⁄ 3 ) 2. a.

5.= -----.= -----.+ -------.⋅ -.3898 We observe that the first value returned by MATLAB above is significantly different from that we obtained from the series.pi/3) ans = -0. J1 ⁄ 2 ( x ) = 2 ----.= 0.5764 16 384 18432 c.5767 ans = 0.sin ( π ⁄ 6 ) = --------.+ ---------------------.3008 4 64 2304 b.5513 π 2 2π π π(π ⁄ 6) 6 1 6 2 .+ ----.2).= – 0.+ --------------------.cos ( π ⁄ 3 ) = -----.– -------------. x x x x J 1 ( x ) = -.– --------------------2 2 4 2 6 2 2 ⋅ ( 1! ) 2 ⋅ ( 2! ) 2 ⋅ ( 3! ) 2 4 6 x=3 9 81 729 = 1 – -.7 Solutions to End−of−Chapter Exercises 1. x x x J 0 ( x ) = 1 – --------------------.cos x πx = x = π⁄3 Check with MATLAB: besselj(0. J–1 ⁄ 2 ( x ) = 2----.⋅ -.2601 ans = 0. besselj(0.pi/6). Third Edition Copyright © Orchard Publications 15−33 .3).sin x πx = x = π⁄6 12 1 2 3 3 2 . a.5.3898 π 2 2π π(π ⁄ 3) d.5513 ans = 0.– ------------------------------2 22 ⋅ 4 22 ⋅ 42 ⋅ 6 22 ⋅ 42 ⋅ 62 ⋅ 8 3 5 7 x=2 8 32 128 = 1 – ----.– -----------.-----------------.= --------.– ----------. besselj(−0.= 0. besselj(1.= 0.Solutions to End−of−Chapter Exercises 15. Numerical Analysis Using MATLAB® and Excel®.2563 and this is much closer to the value obtained with MATLAB. This is because our computation was based on the first 4 terms of the series. Had we taken also the fifth term our answer would have been – 0.-----------------.

P3 ( x ) 2 x = 0.75 c. P0 ( x ) = 1 1 2 P 2 ( x ) = -.( x ) 2 dx 2 = 0 x = 0. P2 ( x ) x = 0.P n ( x ) m dx m a.( 5x – 3x ) ⎬ 2 2 dx ⎩ ⎭ x = 0.= ( 1 – x ) ------.( 3x – 1 ) 2 1 4 2 P 4 ( x ) = -.5 m = 0 This is because m > n .4688 ( 30x ) x = 0. P3 ( x ) x = 0.5 We recall that if m > n . We will use the relations of (15.75 1 2 = -.⋅ ----.34) and (15. P1 ( x ) x = 0.( 3x – 1 ) 2 = 0. Legendre.25 2 ⎫ 2 d ⎧1 3 . They are repeated below for convenience.Chapter 15 Bessel.P 1 ( x ) 2 dx x = 0.( 15x – 3 ) 2 dx = 0.25 ) – 3 × 0. Third Edition Copyright © Orchard Publications .25 2 2 3 d.25 ) = – 0.5 = ( –1 ) ( 1 – x ) 2 2 2⁄2 d ------.25 1 3 = -.5 = x = 0.P 3 ( x ) 2 dx 2 x = 0.75 ) – 1 ) = 0.⎨ -.25 x = 0.P 2 ( x ) = ------. e.25 = ( –1 ) ( 1 – x ) 2 2 2⁄2 d ------.5160 15−34 Numerical Analysis Using MATLAB® and Excel®.5 ( 3 × ( 0.( 5x – 3x ) 2 1 5 3 P 5 ( x ) = -.5 d = ( 1 – x ) ⋅ ------.( 3x 2 – 1 ) ⎬ = 0 .⎨ -.( 35x – 30x + 3 ) 8 m Pn ( x ) m P1 ( x ) = x 1 3 P 3 ( x ) = -. and Chebyshev Functions 2. In other words.5 ( 5 × ( 0.25 = 3.5 2 b.62).3 3 2 dx dx ⎩ ⎭ f.( 5x – 3x ) 2 = 0.( 63x – 70x + 15x ) 8 2 m⁄2 = ( –1 ) ( 1 – x ) d ⋅ --------.25 ) ⋅ -.25 1 d 2 2 = ( 1 – 0. then P n ( x ) = 0 .3438 x = 0. 3 3 ⎫ d ⎧1 d ------.3359 x = 0. 2 P 1 ( x ) x = 0. P2 ( x ) 3 x = – 0.

2).2).0. m=0:2. 1.−0.75).0. fprintf('%2.0f\t %7.2)=legendre(2. fprintf('\n'). m > n .1)=m'. fprintf('m\t Legendre \n').8660 b.2).5156 3 -13.1)=m'.4f \n'. y=zeros(2. y=zeros(3. m=0:1. fprintf('\n'). y=zeros(2.5000 1 -0..4f \n'. the legendre(n. m=0:2.5000 1 -0.0f\t %7. y(:.5).6160 d. y(:.9985 2 3. y(:.4f \n'.y') As in (d) above m > n and MATLAB returns Numerical Analysis Using MATLAB® and Excel®. fprintf('\n'). y(:. y=zeros(3. fprintf('%2.2)=legendre(2. y(:.0.1)=m'. fprintf('m\t Legendre \n'). For this example. y(:. y(:.5).4f \n'.0.25).Solutions to End−of−Chapter Exercises Check with MATLAB: a.y') m Legendre 0 0. . evaluated for each element of x. that is.3359 1 0.0f\t %7.y') m Legendre 0 -0.8660 Here. y=zeros(4. fprintf('m\t Legendre \n').. fprintf('\n').x) function computes the associated Legendre functions of degree n and order m = 0.5).. n.2)=legendre(1. fprintf('m\t Legendre \n').2)=legendre(1.1)=m'.0f\t %7.1)=m'. fprintf('\n'). m=0:1.4f \n'.2)=legendre(3. y(:. m=0:3. fprintf('%2. fprintf('%2.y') m Legendre 0 0.3125 c. For this reason we’ve used m=0:1. y(:.2). m = 2 and n = 1 and the statement m=0:2 is not accepted.0f\t %7. Third Edition Copyright © Orchard Publications 15−35 . e.2). y(:. fprintf('m\t Legendre \n'). fprintf('%2.4882 2 1.3438 1 -1.y') m Legendre 0 0.

2). fprintf('\n').⋅ ---2 3 1 = -4 = 1 -2 1 0 1 3 1 0 ∫0 1 1 5 2 x ⋅ -.⋅ ---2 2 3 xx ⋅ P 1 ( x ) ⋅ dx = -.( 5x – 3x ) dx = -2 4 ∫0 5 3 7 .– ---. m Legendre 0 -0. Legendre.2990 2 2.9985 2 3.y') 3. m=0:3.– 3x .0.3359 1 0.5156 3 -13.Chapter 15 Bessel.⎞ 16 ⎝ 6 4 2 ⎠ 3 = – ----32 15−36 Numerical Analysis Using MATLAB® and Excel®.⎞ 4⎝ 4 2 ⎠ 3 4 2 5 = ----16 1 ∫0 1 1 7 3 x ⋅ -.⎛ 5x . y(:.– ---------.⎞ 4⎝ 5 3 ⎠ 1 = 0 0 9 B 4 = -2 ∫0 1 1 9 4 2 x ⋅ -. Third Edition Copyright © Orchard Publications .6160 2n + 1 B n = -------------2 ∫–1 f ( x )Pn ( x ) dx ∫0 f ( x )Pn ( x ) dx 2 1 0 1 1 For this exercise f ( x ) = 0 for – 1 ≤ x ≤ 0 and thus 2n + 1 B n = -------------2 Then. 1 B 0 = -2 3 B 1 = -2 5 B 2 = -2 7 B 3 = -2 ∫0 ∫0 1 1 x .⎛ ------.------. y(:.1250 1 1.x ⋅ P 0 ( x ) ⋅ dx = -.+ ------.25). y=zeros(4.4f \n'.( 3x – 1 ) dx = -2 4 ∫0 1 1 5 3x 4 x 2 ( 3x – x ) dx = -. and Chebyshev Functions m Legendre 0 -0. fprintf('m\t Legendre \n').1)=m'.2)=legendre(3.( 35x – 30x + 3 ) dx = ----8 16 1 0 ∫0 ( 35x 5 – 30x + 3x ) dx 3 9 35x 6 30x 4 3x 2 = ----. fprintf('%2.2500 f.0f\t %7.------( 5x – 3x ) dx = -.⎛ ---------.

P 0 ( x ) + -.Solutions to End−of−Chapter Exercises and by substitution into f ( x ) = B0 P0 ( x ) + B1 P1 ( x ) + B2 P2 ( x ) + … + Bn Pn ( x ) = we obtain n=0 ∑ Bn Pn ( x ) ∞ 1 5 3 1 f ( x ) = -. Third Edition Copyright © Orchard Publications 15−37 .= -. Also. plot(x. xlabel(‘x’).4 0.9 1 Numerical Analysis Using MATLAB® and Excel®.⋅ -.( 3x – 1 ) – ----.01:1.= -.P 2 ( x ) + 0 ⋅ P 3 ( x ) – ----.*x.7 0.– -------4 2 32 32 256 256 256 2 4 2 or 12 4 f ( x ) = -------.7 0.5 x 0. f ( 1 ) = 248 ⁄ 256 and this value is close to unity.^4).*x.1 0 0 0.*x+210.⋅ x + ----.3 0.P 1 ( x ) + ----.6 f(x) 0.9 0. These values are close to zero. grid 1 0.^2−105.5 0.2 0.3 0./256. fx=(15+128.8 0.– ----. ylabel(‘f(x)’).8 0. x=0:0.( 35x – 30x + 3 ) 16 2 2 32 8 4 1 x 15x 5 105x 90x 9 ..+ -.+ ---------.P 4 ( x ) 2 16 32 4 5 1 1 3 1 1 2 4 2 .⋅ 1 + -.fx).2 0.6 0. We plot f ( x ) with the MATLAB script below.( 15 + 128x + 210x – 105x ) 256 we note that f ( – 1 ) = – 8 ⁄ 256 and f ( 0 ) = 15 ⁄ 256 .4 0.– ------------.1 0.⋅ -.+ ---------.

All other programs are considered non−linear. These variables may or may not be independent of each another. Example 16. which is dependent on a finite number of variables.1 Data for Example 16. TABLE 16. Numerical Analysis Using MATLAB® and Excel®. its product mix at times of high consumer demand. The material types.1.00 per 1000 Due to limited supplies of silicon. and we will illustrate these with some simple but practical examples. ABC Semiconductor can only buy 450 parts of Material A . Thus..00 per unit RAMs (1000s) 2 10 $20. This corporation needs to know what combination of μPs and RAMs will maximize the overall profit. dynamic programming. are linearly related. A and B .Chapter 16 Optimization Methods T his chapter introduces three methods for maximizing or minimizing some function in order to achieve the optimum solution.i.e. Our intent here is to introduce these methods with the basic ideas. is subject to limited supplies. 16. We will discuss linear programming. Third Edition Copyright © Orchard Publications 16−1 .1 The ABC Semiconductor Corporation produces microprocessors ( μPs ) and memory ( RAM ) chips. and in most cases are subject to certain conditions or limitations referred to as constraints.1 Linear Programming In linear* programming we seek to maximize or minimize a particular quantity. These methods are topics discussed in detail in a branch of mathematics called operations research and it is concerned with financial and engineering economic problems. referred to as the objective. required to manufacture the μPs and RAMs and the profits for each are shown in Table 16. phosphorus and boron. * A linear program is one in which the variables form a linear combination. and network analysis.1 Parts of Material Types μPs Semiconductor Material A Semiconductor Material B Profit 3 5 $25. and 1000 parts of Material B .

and x and y must be integers. the corporation is confronted with the following constraints: 3x + 2y ≤ 450 5x + 10y ≤ 1000 We now can state the problem as subject to the constraints Maximize z = 25 × μP + 20 × RAMs 3x + 2y ≤ 450 5x + 10y ≤ 1000 (16.2) Two additional constraints are x ≥ 0 . and with Material B 5 μPs and 10 RAMs . there are only two variables. b . x and y . therefore. We will solve this example graphically. Plot of constraint lines for Example 16. y ) satisfying all constrains. a graphical solution is possible.1. For this example.Chapter 16 Optimization Methods Solution: Since with Material A we can produce 3 μPs and 2 RAMs .* y 250 200 150 100 50 a 5x + 10y = 1000 c 50 100 150 200 250 x 3x + 2y = 450 Isoprofit line b Figure 16.1 where the cross−hatched area indicates the feasible region.1) (16. * The feasible region is the area which includes all points ( x. Third Edition Copyright © Orchard Publications . 16−2 Numerical Analysis Using MATLAB® and Excel®.1 The equation of the straight line of the maximum profit is referred to as isoprofit line. and c . y ≥ 0 . This line will pass through one of the three corners denoted as a . The x and y intercept corresponding to the above equations is shown in the plot of Figure 16.

then. and the highest isoprofit line passes through point b . that is. and y = 37 . that is.3) We can express this equation in y = mx + b form. z max = 25 × 125 + 20 × 37 = $3865 (16. y = 0 . we obtain the points 3865 x = ----------.x + ----. Then. 25 C y = – ----.1 are found by simultaneous solution of 3x + 2y = 450 5x + 10y = 1000 (16. so we accept the values x = 125 .1).5) Using MATLAB for the solution of (16. 5*x+10*y−1000) x = 125 y = 75/2 Of course. The coordinates of point b in Figure 16.4) where k is the y −intercept. these values must be integers. by substitution into (16.Linear Programming The isoprofit line that we are interested is described by the equation z = 25 × μP + 20 × RAM = cons tan t = 25x + 20y = C (16.7) by first letting x = 0 .1.= 154.25x + k 20 20 (16.= – 1. Numerical Analysis Using MATLAB® and Excel®.6) and the isoprofit line can be drawn from the equation 25x + 20y = 3865 (16.5) we obtain syms x y [x y]=solve(3*x+2*y−450.6 25 and ----------y = 3865 = 193. Then.25 20 This is shown as a dotted line on the plot of Figure 16. Third Edition Copyright © Orchard Publications 16−3 . are parallel to each another. Therefore. all possible isoprofit lines have the same slope.

The procedure is included in the spreadsheet of Figure 16. The alternative paths are represented by the line segments ab . OK . and the choice of alternative paths when leaving a given state. 5. 16−4 Numerical Analysis Using MATLAB® and Excel®. we must resort to algebraic methods such as the simplex method. This and other methods are described in operations research textbooks. Add B12>=0. Add B13>=0.Chapter 16 Optimization Methods It was possible to solve this problem graphically because it is relatively simple. Use Add-Ins if necessary to add it. and so on. We can find the optimum solution to this type of problems with Excel’s Solver feature.1 with Excel’s solver 16. we cannot obtain the solution by graphical methods and therefore.2 Dynamic Programming Dynamic Programming is based on R. Add B13=Integer.3 represents a line graph. Add B17<=450. Enter zeros in B12 and B13 2. Solve x(μPs)= 124 y(RAMs)= 38 Maximum Profit= Semiconductor Material A= Semiconductor Material B= $3. On the Solver Parameters screen enter the following: Set Target Cell: B15 Equal to: Max By Changing Cells: B12:B13 Click on Add and enter Constraints: B12=Integer. Bellman’s Principle of Optimality which states that: An optimum policy has the property that whatever the initial state and the initial decisions are. is called a decision.2. Figure 16.Maximum Profit for Example 16.1 1.2 A 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 B C D E F Optimization . In B15 enter =25*B12+20*B13 3. ac . From the Tools drop menu select Solver . Add B18<=1000. where the nodes a through h represent the states. bd . the remaining decisions must constitute an optimum policy with regard to the state resulting from the first decision. In most cases. Third Edition Copyright © Orchard Publications . Spreadsheet for solution of Example 16. In B17 enter =3*B12+2*B13 and in B18 =5*B12+10*B13 4. however.860 448 1000 Note: Contents of A12:A18 are typed-in for information only Figure 16.

( J ce + J eh ) } (16. Third Edition Copyright © Orchard Publications 16−5 .10) and (16. Numerical Analysis Using MATLAB® and Excel®. ( J ac + g c ) } (16. for the path a .8) The costs for the other possible paths are defined similarly. ( J bg + J gh ) } (16. let us suppose that the initial state is a . if the entire path from a to h is to be optimum (minimum in this case).11) The optimum path policy of (16. the path from state c to h must be optimum. and the initial decision has been made to go to state b . in this case.10) Likewise. Line graph for a typical dynamic programming example We assume that all segments are directed from left to right.Dynamic Programming c e d a b h f g Figure 16.9) can now be expressed in terms of (16. that is. there is a cost associated with each segment. we say that the optimum path policy for this line graph is J min = min { ( J ac + J cf + J fh ).9) Now. and each has a value assigned to it which we will refer to as the cost. g c = min { ( J cf + J fh ). Accordingly. For example. Let the minimum cost from state b to h be denoted as g b . c . ( J ac + J ce + J eh ). ( J ab + J bd + J dh ). and it is usually denoted with the letter J . the cost is J ah = J ac + J cf + J fh (16.12) This relation indicates that to obtain the minimum cost we must minimize: 1.11) as g a = J min = min { ( J ab + g b ). g b = min { ( J bd + J dh ). if the initial decision is to go from state a to c . For the line graph of Figure 16. costs J ab and J ac . f . the path from b to h must be selected optimally. and h .3. the objective is to go from state a to state h with minimum cost. Then. Then. The part which is related to the present decision.3. Thus. ( J ab + J bg + J gh ) } (16.

15) The final decision is at state a and thus g a = min { ( 5 + g b ). ( 6 + 8 ) } = 14 (b → e) g c = min { ( 4 + g e ).4.4. ( 5 + g k ) } = min { ( 3 + 5 ). g b = min { ( 9 + g d ). we make the first decision at state e .Chapter 16 Optimization Methods 2. Therefore. The part which represents the minimum value of all future costs starting with the state which results from the first decision.2 Find the minimum cost route from state a to state m for the line graph of Figure 16.2 8 6 e 5 k 4 3 6 h 5 m Solution: We observe that at states h . ( 5 + 4 ) } = 8 (e → h → m) (16.14) (16. d . d 9 b 5 a 3 c 4 6 f Figure 16. The line segments are directed from left to right and the costs are indicated beside each line segment. ( 3 + g c ) } = min { ( 5 + 14 ). g e = min { ( 3 + g h ).13) Next. Third Edition Copyright © Orchard Publications .16) 16−6 Numerical Analysis Using MATLAB® and Excel®. we make decisions at states b and c . ( 6 + g f ) } = min { ( 4 + 8 ). k . Then. ( 3 + 12 ) } = 15 (a → c) (16. ( 6 + g e ) } = min { ( 9 + 6 ). Example 16. and f have no alternative paths since the lines are directed from left to right. ( 6 + 8 ) } = 12 (c → e) (16. Line graph for Example 16.

node A represents an airport in New York City and nodes B through L several airports throughout Europe and Asia. Line graph showing the minimum cost for Example 16.3 L 3 5 4 3 4 K 6 4 J 2 H 3 Numerical Analysis Using MATLAB® and Excel®. or L at the shortest possible time. Which airport should he choose ( H . All flights originate at A and fly eastward. and in how many hours after departure from A will he reach his destination? D 4 B 8 A 6 3 C 5 7 4 2 2 E 4 3 F 7 Figure 16. The encircled numbers represent waiting times in hours at each airport.6. J .3 On the line graph of Figure 16. as shown in Figure 16.6. or L ) to minimize his total travel time. Third Edition Copyright © Orchard Publications 16−7 . also in hours.Dynamic Programming Therefore.5. and the numbers beside the line segments indicated the flight times.5 d 9 b 5 a 3 c 4 6 f Figure 16. the minimum cost is 15 and it is achieved through path a → c → e → h → m . The numbers in squares show the hours he must travel by an automobile to reach his destination. K .2 8 6 e 5 k 4 3 6 h 5 m Example 16. K . J . Line graph for Example 16. A salesman must leave New York City and be in one of the airports H .

g K = 4 . E . ( 5 + 13 ) } = 3 + 16 = 19 C→E (16. ( 2 + 9 ) } = 4 + 11 = 15 B→E g C = 3 + min { ( 7 + g E ). ( 4 + 4 ) } = 4 + 5 = 9 E→J g F = 3 + min { ( 6 + g K ).3 16−8 Numerical Analysis Using MATLAB® and Excel®. ( 6 + 19 ) } = 23 A→B (16. the minimum cost (minimum time from departure to arrival at destination) is 23 hours and it is achieved through path A → B → E → J .7. and g L = 3 (16.21) (16. g J = 2 . ( 4 + g J ) } = 2 + min { ( 5 + 3 ).7. ( 2 + g E ) } = 4 + min { ( 4 + 8 ). ( 5 + g F ) } = 3 + min { ( 7 + 9 ). where we find g A = min { ( 8 + g B ). g D = 2 + min { ( 5 + g H ).19) (16. ( 7 + g L ) } = 3 + min { ( 6 + 4 ). as shown in Figure 16. Line graph showing the minimum cost for Example 16. ( 4 + gK ) } = 4 + min { ( 3 + 2 ). ( 6 + g C ) } = min { ( 8 + 15 ). 5 4 3 E 4 3 F 7 L 3 4 K 6 4 J 2 H 3 D 4 B 8 A 6 3 C 5 7 4 2 2 Figure 16. and F .23) Therefore.20) The next decisions are made at B and C where we find that g B = 4 + min { ( 4 + g D ). ( 7 + 3 ) } = 3 + 10 = 13 F → K or F → L (16.Chapter 16 Optimization Methods Solution: The hours that the salesman must travel by automobile to reach his destination are g H = 3 .17) The first decisions are made at D . ( 4 + 2 ) } = 2 + 6 = 8 D→J g E = 4 + min { ( 3 + g J ).22) The final decision is made at A .18) (16. Then. Third Edition Copyright © Orchard Publications .

we express the profits in millions.000 and the company may allocate all the money to just one product or split it between these three products.000 $6.000.000.000 $1. subject to the constraint that the amount invested does not exceed four million dollars.000.000 $7. In other words is.000. B and C. Investments in each product are assumed to be multiples of $1. Third Edition Copyright © Orchard Publications 16−9 .000. high−technology company.000.2.25) subject to the constraint Numerical Analysis Using MATLAB® and Excel®.000 $2. has $4.000 $4. TABLE 16.000 $3.000 $6. The expected profits are shown in Table 16.4 A start−up. TABLE 16.000.000 Return on Investment How should the money be allocated so that company will realize the maximum profit? Solution: This problem can also be solved with linear programming methods but we will use the so called tabular form of solution.000 $2.2 x p(x) pA ( x ) pB ( x ) pC ( x ) 0 0 0 0 1 2 1 1 2 5 3 4 3 6 6 5 4 7 7 8 Our objective is to maximize the total profit z that represents the sum of the profits from each product.000 $5.000. when x units of dollars are invested in it.000 to invest in three different products A.000 $3.000.000.000.000.3 Modified Table 16. For simplicity.000.000 $7. and we enter these in Table 16.000. Let pi ( x ) i = A.000.000 $8. we want to maximize z = p A ( x ) + p B ( x ) + p C ( x ) (16.24) denote the profits in millions from product i .000.3.000 $4.000 $1. C (16. B.2 Amounts invested and return on investment for each product Investments Amount Invested 0 Product A Product B Product C 0 0 0 $1.000.000 $5.Dynamic Programming Example 16.000.

do not have 4 units to invest in Product C. p C ( 4 ) } = max { 0. p C ( 3 ).Chapter 16 Optimization Methods xA + xB + xC ≤ 4 (16. The computations are done in three stages. and u represents the number of money units.29) (16. we have three or less. j = C . Also. v C ( 4 ) = max { p C ( 0 ). The possibilities that we allocate 0 or 1 or 2 or 3 or 4 units (millions) to Product C. and u = 4 .32) with decision dC ( 2 ) = 4 With only one unit left to invest. We start by allocating units (millions) to Product C (Stage C). are the amounts to be invested in products A. p C ( 2 ). from Table 16.. 4. 4. B and C respectively. p C ( 3 ) } = max { 0. 1. p C ( 1 ).27) (16. p C ( 1 ) } = max { 0. and x C . 8 } = 8 (16. one per product.30) with decision dC ( 3 ) = 5 If we have only two units left. x B . and C for Product C.33) 16−10 Numerical Analysis Using MATLAB® and Excel®. and the corresponding returns are. p C ( 1 ). 1. p C ( 2 ) } = max { 0. the optimum value v C ( 3 ) is found from v C ( 3 ) = max { p C ( 0 ). by a previous decision. i.e. Third Edition Copyright © Orchard Publications . 1.. At Stage C. we obtain the maximum from v C ( 2 ) = max { p C ( 0 ). The next possibility is that one unit was invested in either Product A or Product B. i. 4 millions assumed to be allocated to Product C. we have v C ( 1 ) = max { p C ( 0 ). 4 } = 4 (16.31) (16.3. but since we do not know what units were allocated to the previous products A and B. B for Product B. p C ( 2 ). If we invest the remaining three units in Product C. we must consider all possibilities. 5. In this case. where the subscript j indicates the number or stage assigned to the product. and we invest them in Product C.26) where x A . A for Product A. We let v j ( u ) denote the value of the optimum profit that can be achieved.e.28) with decision dC ( 4 ) = 8 that is. 1 } = 1 (16. we let d j ( u ) be the decision that is being made to achieve the optimum value from v j ( u ) . the maximum appears in the fourth position since the left most is the zero position. p C ( 1 ). 5 } = 5 (16.

p B ( 3 ) + v C ( 4 – 3 ). 3. TABLE 16. Inserting the appropriate values.39) with decision dB ( 4 ) = 0 since the maximum value is the zero position term.35) (16. p B ( 4 ) + v C ( 4 – 4 ) } (16. we construct Table 16. we obtain v B ( 4 ) = max { 0 + 8. again we must consider all possibilities. we have v B ( 4 ) = max { p B ( 0 ) + v C ( 4 – 0 ). 3 + 4.36) with decision dC ( 0 ) = 0 With these values. Third Edition Copyright © Orchard Publications 16−11 . 1 and 0 .4 Optimum profit and decisions made for Stage C u 0 Stage C Stage B Stage A vC ( u ) dC ( u ) vB ( u ) dB ( u ) vA ( u ) dA ( u ) 0 0 … … … … 1 1 1 … … … … 2 4 2 … … … … 3 5 3 … … … … 4 8 4 … … … … Next. p B ( 1 ) + v C ( 4 – 1 ). v C ( 0 ) = max { p C ( 0 ) } = max { 0 } = 0 (16. 7 + 0 } = 8 (16. and since we do not know what units were allocated to Product A (Stage A).4. 2. Numerical Analysis Using MATLAB® and Excel®. 1 + 5. and so on.Dynamic Programming with decision dC ( 1 ) = 1 Finally.38) (16. it is possible that all four units were invested in Product C. it is possible that 3 units were invested in Product C.34) (16. With j = B and u = 4. we consider Stage B. with no units left to invest in Product C. p B ( 2 ) +v C ( 4 – 2 ).37) This expression says that if zero units were invested in Product B. or if one unit was invested in Product B. 6 + 1.

47) (16.Chapter 16 Optimization Methods Using a similar reasoning. 3 + 0 } = 4 with decision dB ( 2 ) = 0 v B ( 1 ) = max { p B ( 0 ) + v C ( 1 – 0 ).45) (16.40) or with decision (16.52) or v B ( 0 ) = max { 0 + 0 } = 0 with decision dB ( 0 ) = 0 Next.5. p B ( 1 ) + v C ( 2 – 1 ). 1 + 4. we update the previous table to include the Stage B values. 1 + 1.48) (16. v B ( 2 ) = max { p B ( 0 ) + v C ( 2 – 0 ).42) (16. p B ( 3 ) + v C ( 3 – 3 ) } v B ( 3 ) = max { 0 + 5. we have v B ( 3 ) = max { p B ( 0 ) + v C ( 3 – 0 ). 3 + 1. Also. These are shown in Table 16. or dB ( 1 ) = 1 if we consider the first position term.49) dB ( 3 ) = 3 Also.50) (16. 16−12 Numerical Analysis Using MATLAB® and Excel®. v B ( 0 ) = max { p B ( 0 ) + v C ( 0 – 0 ) } (16.46) (16.41) (16.44) (16. p B ( 2 ) +v C ( 3 – 2 ). 6 + 0 } = 6 (16. p B ( 1 ) + v C ( 1 – 1 ) } or v B ( 1 ) = max { 0 + 1. p B ( 1 ) + v C ( 3 – 1 ). Third Edition Copyright © Orchard Publications . 1 + 0 } = 1 with decision dB ( 1 ) = 0 if we consider the zero position term.51) (16. p B ( 2 )+v C ( 2 – 2 ) } or v B ( 2 ) = max { 0 + 4.43) (16.

this tells us that we should allocate 2 units to Product A . 6 + 1.000.6 indicates that the maximum profit is realized with v A ( 4 ) = 9 .000.000. Third Edition Copyright © Orchard Publications 16−13 . 5 + 4. p A ( 1 ) + v B ( 4 – 1 ). Numerical Analysis Using MATLAB® and Excel®. all possibilities have been considered. Therefore.000. Table 16. all 4 units can be allocated to the Product A or some of these can be allocated to Products B and C. with j = A and u = 4 v A ( 4 ) = max { p A ( 0 ) + v B ( 4 – 0 ).000. 7 + 0 } = 9 (16.53) or v A ( 4 ) = max { 0 + 8. and the given table shows that 2 units ($2. 2 + 6. 9 units. v A ( 4 ) considers all possibilities. p A ( 2 ) +v B ( 4 – 2 ).5 Updated table to include Stage B values u 0 Stage C Stage B Stage A vC ( u ) dC ( u ) vB ( u ) dB ( u ) vA ( u ) dA ( u ) * 1 1 1 1 1 … … 2 4 2 4 0 … … 3 5 3 6 3 … … 4 8 4 8 0 … … 0 0 0 0 … … Finally.54) (16. and this is always the case since in deriving v A ( 4 ) and d A ( 4 ) . * Since this is the first stage. that is. To determine the investment allocations to achieve this profit. p A ( 4 ) + v B ( 4 – 4 ) } (16.000) invested in this product will return $5. and thus the maximum profit is $9. p A ( 3 ) + v B ( 4 – 3 ).6. we start with d A ( 4 ) = 2 . The only entries are in the last column.Dynamic Programming TABLE 16.55) with decision dA ( 4 ) = 2 We complete the table by entering the values of Stage A in the last two rows as shown in Table 16.

000.000. A typical network 16−14 Numerical Analysis Using MATLAB® and Excel®. we consider the decision at Stage B. The decision at Stage C yields d C ( 4 – 0 – 2 ) = d C ( 2 ) . we see that d 2 ( 2 ) = 0 . and by reference to the Table 16. Figure 16. and from Table 16. as defined here.3 Network Analysis A network. BD . we should allocate: 1. zero units to Product B to earn $0 3.Chapter 16 Optimization Methods TABLE 16.000. Thus. to obtain the maximum profit of $9. C .8 is a network with 5 nodes A .000. BE . B .6. and CD . and 6 branches AB . B E A D C Figure 16.000.6.6 Updated table to include values for all stages u 0 Stage C Stage B Stage A vC ( u ) dC ( u ) vB ( u ) dB ( u ) vA ( u ) dA ( u ) 0 0 0 0 … … 1 1 1 1 1 … … 2 4 2 4 0 … … 3 5 3 6 3 … … 4 8 4 8 0 9 2 We now have two units left to invest in Products B and C. d C ( 2 ) = 2 .000 16. This tells us that we should not invest any units in Product B if only two units are left. is a set of points referred to as nodes and a set of lines referred to as branches.000 2.000. D and E . To find out where we should invest these units. In summary. AC . we have d B ( 4 – 2 ) = d B ( 2 ) . two units to Product C to earn $4. This indicates that we should invest the remaining two units to Product C where we can obtain a return of $4. Third Edition Copyright © Orchard Publications . AD . two units to Product A to earn $5. Since two out of the four units have already been invested.8.

For example. or two−way. the network of Figure 16. one−way. Example 16. the network shown in Figure 16. B E A D C Figure 16. if there is a path (branch) connecting each pair of nodes.9 is also connected.8 is connected. In such problems.8. A network which is connected The network of Figure 16. the branches BD and BE in Figure 16. B E A C D Figure 16. the branches indicate possible paths. the network of Figure 16. A network which is not connected A tree is a connected network which has n branches and n + 1 nodes. Thus.10. that is. and the numbers beside the Numerical Analysis Using MATLAB® and Excel®. If no direction is assigned.10 is not connected since the branch CD is removed. A tree network Network analysis is a method that is used to solve minimum span problems. However. they are considered to be two−way.Network Analysis Branches can be either directed (or oriented). A network is said to be connected.11 is a tree network.12 represents a network for a project that requires telephone cable be installed to link 7 towns.5 Figure 16.9. and the sum of the costs (shortest total distance) is a minimum. if they have a direction assigned to them. we seek to find a tree which contains all nodes. are directed but the others are not. B E A D Figure 16. The towns are the nodes.11. Thus. Third Edition Copyright © Orchard Publications 16−15 .

We accept this branch as the first branch of the minimum span tree and we draw a solid line from Node A to Node B as shown in Figure 16. the least amount of telephone cable required to link each town.12. B 3 A 4 C 5 4 10 5 D 5 F 8 6 7 2 E 5 G Figure 16. B 3 A 4 C 5 4 10 5 D 5 F 8 6 7 2 E 5 G Figure 16. Network of Example 16.14. we redraw the given network with dotted lines as shown in Figure 16. and AC . i.13.Chapter 16 Optimization Methods branches..5 with first connection 16−16 Numerical Analysis Using MATLAB® and Excel®. show the distance (not to scale) between towns in kilometers.13.56) we find that branch AB is the shortest. Third Edition Copyright © Orchard Publications .e. AB . B 3 A 4 C 5 4 10 5 8 E 6 7 5 F 2 5 G D Figure 16. Network of Example 16. and we arbitrarily choose A as the starting node.5 with no connections We observe that there are 3 branches associated with node A . AD . or from the expression min { AB = 3. AC = 4 } = AB = 3 (16. Find the minimal spanning tree. By inspection. AD = 5.14. that is.5 Solution: For convenience. Network for Example 16.

58) and we add branch CD to the network shown in Figure 16.15. B 3 A 4 C 4 10 8 D 5 F E 6 7 2 5 G Figure 16. The dotted lines AD and BD have been removed since we no longer need to consider branch AD and BD . Network of Example 16. BD = 5.57) and thus. We find that the minimum of these is min { AD = 5. BE = 8. B and C . Numerical Analysis Using MATLAB® and Excel®. we will not have a tree network. AC = 4. otherwise. considering all branches associated with Nodes B .Network Analysis Next. CD = 4. CF = 10 } = CD = 4 (16.5 with the second connection We continue by considering all branches associated with Nodes A . Network of Example 16. C .15. DF = 5.16. and we find that the shortest is min { AD = 5. DE = 6. we consider all branches associated with Nodes A and B .17.5 with the third connection Next. Third Edition Copyright © Orchard Publications 16−17 . BD = 5.16.59) and the network now is connected as shown in Figure 16. B 3 A 4 C 5 4 10 5 8 E 6 7 5 F 2 5 G D Figure 16. and D and we find that the shortest is min { BE = 8. because Nodes B and D are already connected. BE = 8 } = AC = 4 (16. AC is connected to the network as shown in Figure 16. DG = 7. CF = 10 } = DF = 5 (16.

We find that min { EG = 5.5 with the fourth connection Continuing. we obtain min { BE = 8. DE = 6.Chapter 16 Optimization Methods B 3 A 4 C 4 8 6 D 5 F 7 2 E 5 G Figure 16. Third Edition Copyright © Orchard Publications . Network of Example 16. FG = 2 } = FG = 2 (16.18 B 3 A 4 C 4 D 5 F 6 7 2 E 5 G Figure 16.17. Network of Example 16. 16−18 Numerical Analysis Using MATLAB® and Excel®.19. B 3 A 4 C 4 D 5 F 2 6 G E Figure 16. DG = 7 } = DE = 6 (16. Network of Example 16.19.18. DG = 7.61) and the complete minimum span tree is shown in Figure 16.5 with all connections Figure 16.19 shows that the minimum distance is 3 + 4 + 4 + 6 + 5 + 2 = 24 kilometers.5 with the fifth connection The last step is to determine the shortest branch to Node G .60) and the network is connected as shown in Figure 16.

as defined in this chapter. and in most cases are subject to certain conditions or limitations referred to as constraints. • Network analysis is a method that is used to solve minimum span problems. Numerical Analysis Using MATLAB® and Excel®. • Dynamic Programming is based on R. • A network. In such problems.Summary 16. referred to as the objective.4 Summary • Linear programming is a procedure we follow to maximize or minimize a particular quantity. Third Edition Copyright © Orchard Publications 16−19 . and the sum of the costs (shortest total distance) is a minimum. • A tree is a connected network which has n branches and n + 1 nodes. which is dependent on a finite number of variables. the remaining decisions must constitute an optimum policy with regard to the state resulting from the first decision. we seek to find a tree which contains all nodes. is a set of points referred to as nodes and a set of lines referred to as branches. These variables may or may not be independent of each another. Bellman’s Principle of Optimality which states that an optimum policy has the property that whatever the initial state and the initial decisions are.

5 Exercises 1. d 8 b 3 a 5 c 4 6 f 7 5 e 4 k 5 3 5 h 4 m 3.Chapter 16 Optimization Methods 16. Third Edition Copyright © Orchard Publications . How many barrels of each grade of oil should he buy so that after mixing the two grades can minimize his cost while at the same time meeting EPA’s requirement? Solve this problem graphically and check your answers with Excel’s Solver. The line segments are directed from left to right and the costs are indicated beside each line segment. A large oil distributor can buy Grade A oil which contains 7% lead for $25.00 per barrel from one oil refinery company. Repeat Example 16.3 for the line graph shown below. The Environmental Protection Agency (EPA) requires that all oil sold must not contain more than 10% lead.00 per barrel from another oil refinery company. He can also buy Grade B oil which contains 15% lead for $20. 2. Use dynamic programming to find the minimum cost route from state a to state m for the line graph shown below. D 2 B 3 A 5 2 C 1 7 5 4 1 E 3 4 F 6 L 4 5 K 3 6 4 4 J 2 H 2 7 16−20 Numerical Analysis Using MATLAB® and Excel®.

and ignore travel time from customer to customer. The third row (zero hours) indicates the commission that he will receive if he just calls instead of visiting them. He will earn the commissions shown on the table below for various visiting times. Visit Time (Hours) 0 1 2 3 4 Customer 1 $20 $45 $65 $75 $83 2 $40 $45 $57 $61 $69 3 $40 $52 $62 $71 $78 4 $80 $91 $95 $97 $98 5. B 30 40 A 50 20 C 40 D 50 F 50 90 70 E Numerical Analysis Using MATLAB® and Excel®. A salesman has 4 hours available to visit 4 of his customers. Third Edition Copyright © Orchard Publications 16−21 .5 for the network shown below. Consider only integer number of visiting hours. Compute the optimal allocation of time that he should spent with his customers so that he will maximize the sum of his commissions. Repeat Example 16.Exercises 4.

10 ⁄ 0.15y ≤ 0. x + y = 1 (3) Moreover. y = 0.Chapter 16 Optimization Methods 16. Let x be the number of barrels of Grade A oil and y be the number of barrels of Grade B oil.15 = 2 ⁄ 3 and if y = 0 . z = 25x + 20y (1) We want to minimize (1) because it represents a cost. if x = 0 . we must have 0.07x + 0.15y = 0.10 ⁄ 0. not a profit. (3). Thus from (2). from (3).6 Solutions to End−of−Chapter Exercises 1.00y or.00x + $20. for simplicity. To determine the feasible region we plot (2) and (3) where the x and y crossings are found by first setting x = 0 and then y = 0 . The objective is to minimize z = $25. Each barrel to be sold must not contain more than 10 % lead and since Grade A contains 7 % and Grade B 15 % .10 (2) The oil of Grade A and Grade B used in each barrel to be sold must be equal to unity. Thus. Third Edition Copyright © Orchard Publications .07 = 10 ⁄ 7 . y 1 Isoprofit line x+y = 1 2⁄3 a 1⁄3 b c 1⁄7 2⁄7 3⁄7 4⁄7 5⁄7 6⁄7 1 0. x and y cannot be negative numbers. Likewise. x = 1 . and if y = 0 .10 8⁄7 9 ⁄ 7 10 ⁄ 7 x The isoprofit line passes through point b and its coordinates are found by simultaneous solu- 16−22 Numerical Analysis Using MATLAB® and Excel®. if x = 0 . therefore x≥0 y ≥ 0 (4) The problem then can be stated as: Minimize (1) subject to the constraints of (2). and (4). y = 1 . x = 0.07x + 0.

+ 20 × -. the distributor should buy Grade A oil at the ratio x = 5 ⁄ 8 and Grade B at the ratio y = 3 ⁄ 8 and by substitution into (1) 5 3 z = 25 × -. 5. On the Solver Parameters screen enter the following: Set Target Cell: B15 Equal to: Min By Changing Cells: B12:B13 Click on Add and enter Constraints: B12>=0.125 -------8 8 8 and this represents his cost per barrel.10.Solutions to End−of−Chapter Exercises tion of (2) and (3).125 and the y −intercept is found by setting x in the equation above to zero and we find that y – intercept = 23.1 1. Solve Grade A= 0. OK . Enter zeros in B12 and B13 2. Add B13>=0.10.000001 Note: Contents of A12:A18 are typed-in for information only Numerical Analysis Using MATLAB® and Excel®. [x.= 185 = $23.15*B13 and in B18 =B12+B13 4. Third Edition Copyright © Orchard Publications 16−23 .07*B12+0.10 Grade A + Grade B= 1. we use the following MATLAB script: syms x y. From the Tools drop menu select Solver . Add B17<=0.1563 Check with Excel’s Solver: A 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 B C D E F Optimization . For convenience.125 Lead Content= 0. Use Add-Ins if necessary to add it.y]=solve(0.125 ⁄ 20 = 1. Add B18=1.374999 Minimum Cost= $23.Minimum Cost for Exercise 16. x+y−1) x = 5/8 y = 3/8 Therefore.15*y−0. The isoprofit line is z = 25x + 20y = 23.625002 Grade B= 0. In B15 enter =25*B12+20*B13 3.07*x+0. In B17 enter =0.

Last stage: g H = 2 . the best route from state e to state m passes through state h . 6 + g f } = min { 4 + 7. g b = min { 8 + g d. g J = 2 . Third Edition Copyright © Orchard Publications . D 2 B 3 A 5 2 C 1 7 5 4 1 E 3 4 F 6 L 4 5 K 3 6 4 4 J 2 H 2 7 The numbers in circles represent the waiting time at these nodes. 5 + 11 } = 15 Thus. 5 + 7 } = 12 and g c = min { 4 + g e. 5 + g e } = min { 8 + 5. 6 + 7 } = 11 Finally. Next.Chapter 16 Optimization Methods 2. the best (shortest) path is a → b → e → h → m 3. 4 + 5 } = 7 Therefore. g K = 3 . state e is the first node where a decision must be made and thus g e = min { 3 + g h. and g L = 4 16−24 Numerical Analysis Using MATLAB® and Excel®. d 8 b 3 a 5 c 4 6 f 7 5 e 4 k 5 3 5 h 4 m Since the segments are directed from left to right. g a = min { 3 + g b. 4 + g k } = min { 3 + 4. 5 + g c } = min { 3 + 12.

and ignore travel time from customer to customer. Consider only integer number of visiting hours. Compute the optimal allocation of time that he should spend with his customers so that he will maximize the sum of his commissions. Third Edition Copyright © Orchard Publications 16−25 . A 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 B C D E F G H Exercise 16. 1 $20 $45 $65 $75 $83 Customer 2 $40 $45 $57 $61 $69 3 $40 $52 $62 $71 $78 4 $80 $91 $95 $97 $98 continued on the next page Numerical Analysis Using MATLAB® and Excel®. 4 + g J } = 1 + 6 = 7 g E = 3 + min { 4 + g J.Solution A salesman has 4 hours available to visit of his customers. 1 + g F } = 2 + 15 = 17 from B to D from C to F Initial stage: g A = 0 + min { 3 + g B. 5 + g C } = 17 from A to B Therefore. minimum path is from A → B → D → J and the numerical minimum cost is 17 .4 . Visit Time (Hours) 0 1 2 3 4 Solution We will follow the same method as in Example 16.4 It is convenient to rearrange the table as shown below.Solutions to End−of−Chapter Exercises Next stage to the left: g D = 1 + min { 6 + g H. 5 + g K } = 3 + 6 = 9 g F = 4 + min { 7 + g K. 4. 6 + g L } = 4 + 10 = 14 from D to J from E to J from F to K or F to L Next to initial stage: g B = 5 + min { 2 + g D. He will earn the commissions shown on the table below for various visiting times. The third row (zero hours) indicates the commission that he will receive if he just calls instead of visiting them. 4 + g E } = 5 + 9 = 14 g C = 2 + min { 7 + g E.

C25+C38.D26) m4(1)= max(B26. f3(4)+m4(4-4)] MAX(B25+F38.E26) m4(2)= max(B26. we compute the values of m3(u) and d3(u) m3(4)=max[f3(0)+m4(4-0).E26.C25+D38.C25+E38.E25+B38) m3(2)=max[f3(0)+m4(2-0). u 0 m4(u) d4(u) m3(u) d3(u) m2(u) d2(u) m1(u) d1(u) Next. f3(2)+m4(4-2).C26. f3(1)+m4(1-1)] MAX(B25+C38.C26.F25+B38) m3(3)=max[f3(0)+m4(3-0). f3(1)+m4(2-1).C26) m4(1)= max(B26) The values of m4(u) and d4(u) are entered in the table below.Chapter 16 Optimization Methods A 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 f(x) \ x f1(x) f2(x) f3(x) f4(x) B 0 20 40 40 80 C 1 45 45 52 91 D u 2 65 57 62 95 E 3 75 61 71 97 F 4 83 69 78 98 G H I m4(4)= max(B26.F26) m4(3)= max(B26. f3(3)+m4(4-3). 16−26 Numerical Analysis Using MATLAB® and Excel®. f3(1)+m4(3-1).D25+C38. Third Edition Copyright © Orchard Publications .D26.D25+B38) m3(1)=max[f3(0)+m4(1-0).C25+B38) m3(0)=max[f3(0)+m4(0-0)] MAX(B25+B38) 80 0 120 0 160 0 1 91 1 132 1 172 0 2 95 2 143 1 183 0 3 97 3 153 2 193 0 4 98 4 162 3 202 0 248 2 98 with d4(4)= 97 with d4(3)= 95 with d4(2)= 91 with d4(1)= 80 with d4(0)= 4 3 2 1 0 162 with d3(4)= 3 153 with d3(3)= 2 143 with d3(2)= 1 132 with d3(1)= 1 120 with d3(0)= 0 continued on the next page. f3(2)+m4(2-2)] MAX(B25+D38. f3(1)+m4(4-1).C26.E25+C38. f3(2)+m4(3-2).D26. f3(3)+m4(3-3)] MAX(B25+E38.D25+D38.

Rows 42 and 43 Stage 1 is the last stage and there is only one state associated with it. Rows 40 and 41 Similarly. f2(3)+m3(4-3). f1(4)+m2(4-4)] MAX(B23+F42. f2(4)+m3(4-4)] MAX(B24+F40. f2(1)+m3(2-1). the salesman should spend 2 hours with Customer #1 (d1(4)=2).C24+E40. 1 hour = Total = $65 $40 $52 $91 $248 202 with d2(4)= 0 193 with d2(3)= 0 183 with d2(2)= 0 172 with d2(1)= 0 160 with d2(0)= 0 248 with d1(4)= 2 Numerical Analysis Using MATLAB® and Excel®.C24+B40) m2(0)=max[f2(0)+m3(0-0)] MAX(B24+B40) These values are added to the table above. f1(2)+m2(4-2).E24+C40. Check: Customer #1. f2(2)+m3(2-2)] MAX(B24+D40. 0 hours with Customer #2 (d2(4)=0). u=4.E24+B40) m2(2)=max[f2(0)+m3(2-0). f1(1)+m2(4-1). f2(1)+m3(3-1).C24+D40.D24+D40. 2 hours = Customer #2. and thus m1(4)=max[f1(0)+m2(4-0). f2(2)+m3(3-2). to achieve the maximum sum of commissions. f2(2)+m3(4-2). we compute the values of m2(u) and d2(u) m2(4)=max[f2(0)+m3(4-0). 1 hour = Customer #4.Solutions to End−of−Chapter Exercises A 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 B C D E F G H I These values are now added to the table above. and for the remaining 2 hours he should spend 1 hour with Customer #3 and 1 hour with Customer #4.D23+D42. 0 hours = Customer #3.F25+B40) m2(3)=max[f2(0)+m3(3-0).C24+C40. f2(3)+m3(3-3)] MAX(B24+E40.D24+B40) m2(1)=max[f2(0)+m3(1-0).D24+C40. f2(1)+m3(1-1)] MAX(B24+C40. f1(3)+m2(4-3).E23+C42.F23+B42) These two values are the last entries into the table in Cells F44 and F45 The table (Rows 36 through 45) indicates that. Third Edition Copyright © Orchard Publications 16−27 . f2(1)+m3(4-1).C23+E42.

C → F ] = min [ 40. 50. A → C ] = min [ 30. 40 ] = 40 min [ B → E. B 30 40 A 50 20 C 40 D 50 F 50 90 70 E min [ A → B. C → F ] = min [ 50. C → F ] = min [ 30. 20 ] = 20 A→C A→B choose B → D C→F B→E min [ A → B. 50. 50. D → E. Third Edition Copyright © Orchard Publications . D → E. F → E ] = min [ 50. 40 ] = 30 min [ B → D. C → D. 40 ] = 40 min [ B → E. 50. 90.Chapter 16 Optimization Methods 5. C → D. 90. 70 ] = 50 B 30 40 A 20 D 50 E C 40 F and thus the minimum distance is 20 + 30 + 40 + 40 + 50 = 180 kilometers 16−28 Numerical Analysis Using MATLAB® and Excel®. D → F. B → E.

* For an introduction and applications of the Z-transform please refer to Signals and Systems with MATLAB Computing and Simulink Modeling. Third Edition. Obtain the forced response (particular solution) y P ( n ) in terms of an arbitrary real constant k 3 . The right side of relation (A. While recursive methods yield a desired result. where a 1 and a 2 are also real constants. they do not provide a closed−form solution. that is.2 Method of Undetermined Coefficients A second−order difference equation has the form y ( n ) + a1 y ( n – 1 ) + a2 ( n – 2 ) = f ( n ) (A. ISBN 0-9744239-9-8.1) where a 1 and a 2 are constants and the right side is some function of n . The general (closed-form) solution of relation (A.1) is referred to as the forcing function. If a closed−form solution is desired. and this method is similar to the classical method of solving linear differential equations with constant coefficients. This method is described in the next section. yC ( n ) = k1 a1 + k2 a2 n n (A.Appendix A Difference Equations in Discrete−Time Systems T his appendix is a treatment of linear difference equations with constant coefficients and it is confined to first− and second−order difference equations and their solution. Third Edition Copyright © Orchard Publications A−1 . A. The three steps are as follows: 1. each term of which is determined by application of a formula to preceding terms. An example of a recursive method is Newton’s method† for solving non−linear equations. The solution of a difference equation is often obtained by recursive methods. Obtain the natural response (complementary solution) y C ( n ) in terms of two arbitrary real constants k 1 and k 2 . we can solve difference equations using the Method of Undetermined Coefficients.* A. a recursion is an expression. Higher−order difference equations of this type and their solution is facilitated with the Z−transform. This difference equation expresses the output y ( n ) at time n as the linear combination of two previous outputs y ( n – 1 ) and y ( n – 2 ) . Numerical Analysis Using MATLAB® and Excel®.1) is the same as that used for solving second−order differential equations. † Newton’s method is discussed in Chapter 2. such as a polynomial.1 Recursive Method for Solving Difference Equations In mathematics.2) 2.

we must remember that if f ( n ) is the sum of several terms. As in the case with the solutions of ordinary differential equations with constant coefficients. 3.4) using the given initial conditions. Also. Solve for k 1 and k 2 in (A. It is important to remember that the constants k 1 and k 2 must be evaluated from the total solution of (A. not from the complementary solution y C ( n ) .1. we must multiply y P ( n ) by the lowest power of n that will eliminate the duplication. yP ( n ) = k3 a3 n (A. y ( n ) = yC ( n ) + yP ( n ) = k1 a1 + k2 a2 + yP ( n ) n n (A.4).Appendix A Difference Equations in Discrete-Time Systems that is.5) subject to the initial conditions y ( – 2 ) = 25 and y ( – 1 ) = 6 .1 Find the total solution for the second−order difference equation 1 5 –n y ( n ) – -. * Ordinary differential equations with constant coefficients are discussed in Chapter 5.* TABLE A. Example A. It is best to illustrate the Method of Undetermined Coefficients via examples.1 Forms of the particular solution for different forms of the forcing function Form of forcing function Form of particular solutiona Constant an − a is a constant ab ±n k k − a constant k 0 + k 1 n + k 2 n + … + k k n − k i is constant Expression proportional to b k 1 cos ( nω ) + k 2 sin ( nω ) ±n 2 k − a and b are constants acos ( nω ) or asin ( nω ) a. A−2 Numerical Analysis Using MATLAB® and Excel®. that is.3) where the right side of (A. Third Edition Copyright © Orchard Publications .y ( n – 1 ) + -.y ( n – 2 ) = 5 6 6 n≥0 (A. if a term in y P ( n ) is a duplicate of a term in the complementary solution y C ( n ) . Add the natural response (complementary solution) y C ( n ) and the forced response (particular solution) y P ( n ) to obtain the total solution.3) is chosen with reference to Table A.4) 4. the most general form of the particular solution y P ( n ) is the linear combination of these terms.

⎞ 5 = 1 ⎝6⎠ ⎝6⎠ or k 3 = 1 and thus yP ( n ) = 5 –n (A.6) we obtain 1 n 1 n –n –n y C ( n ) = k 1 ⎛ -.13) The total solution is the addition of (A.8) Division of (A.y ( n – 1 ) + -. that is.7) Substitution of y ( n ) = a n into (A. 5 –( n – 1 ) 1 –( n – 2 ) –n –n + k 3 ⎛ -.a 6 6 (A.⎞ 5 = 5 k 3 5 – k 3 ⎛ -.9) are 1 a 1 = -2 1 a 2 = -3 and by substitution into (A.11) 2. Since the forcing function is 5 –n .⎞ 5 + ⎛ -.5) is 5 1 y ( n ) – -.6) The homogeneous equation of (A. we assume that the particular solution is yP ( n ) = k3 5 –n (A.Method of Undetermined Coefficients Solution: 1. y ( n ) = yC ( n ) + yP ( n ) = k1 2 + k2 3 + 5 –n –n –n (A. Third Edition Copyright © Orchard Publications A−3 .⎞ = k 1 2 + k 2 3 ⎝2⎠ ⎝3⎠ (A.8) by a n – 2 yields 5 2 -a – -.a = 0 a – -. We assume that the complementary solution y C ( n ) has the form yC ( n ) = k1 a1 + k2 a2 n n (A.13).y ( n – 2 ) = 0 6 6 n≥0 (A.12) and by substitution into (A.⎞ 5 ⎝6⎠ ⎝6⎠ Division of both sides by 5 –n yields 1 2 5 k 3 1 – ⎛ -.5).9) (A.a + 1 = 0 6 6 (A.⎞ + k 2 ⎛ -.11) and (A.14) Numerical Analysis Using MATLAB® and Excel®.7) yields 5 n–1 1 n–2 n + -.10) The roots of (A.

14) reduces to 4k 1 + 9k 2 = 0 y ( –1 ) = k1 2 + k2 3 + 5 = 6 1 1 1 (A.8 1.2 0 0 1 2 3 4 5 6 7 8 9 10 Figure A.^(−n).Appendix A Difference Equations in Discrete-Time Systems To evaluate the constants k 1 and k 2 we use the given initial conditions.5.⎞ 2 + ⎛ – -.6 1. s y ( – 2 ) = 25 and y ( – 1 ) = 6 .⎞ 3 + 5 ⎝ 3⎠ ⎝2⎠ n≥0 (A. i.yn).16) yields 3 k 1 = -2 2 k 2 = – -3 (A.1.1.15) and (A.8 0.^(−n)−(2.e.1 A−4 Numerical Analysis Using MATLAB® and Excel®.4 1./3). Third Edition Copyright © Orchard Publications . grid The plot is shown in Figure A..15) from which 2k 1 + 3k 2 = 1 (A.18) To plot this difference equation for the interval 0 ≤ n ≤ 10 . stem(n. Plot for the difference equation of Example A.6 0. (A. we use the following MATLAB script: n=0:1:10. 2 1. For n = – 2 .4 0.*2.*3. yn=1.14) we obtain the total solution as 3 –n –n 2 –n y ( n ) = y C ( n ) + y P ( n ) = ⎛ -. (A.2 1 0.14) reduces to y ( – 2 ) = k 1 2 + k 2 3 + 5 = 25 2 2 2 from which For n = – 1 .17) and by substitution into (A.16) Simultaneous solution of (A.^(−n)+5.

Method of Undetermined Coefficients Example A. we would assume that the particular solution is yP ( n ) = k3 + k4 3 –n (A.22) by a n – 2 yields 3 2 1 a – -.22) Division of (A.1.21) Substitution of y ( n ) = a n into (A.19) is 1 3 y ( n ) – -.21) yields 3 n–1 1 n–2 n a – -.20) we obtain 1 n n –n y C ( n ) = k 1 ⎛ -.19) subject to the initial conditions y ( – 2 ) = 0 and y ( – 1 ) = 2 Solution: 1.26) However. Since the forcing function is 1 + 3 –n .24) The roots of (A.25) and (A.26) contain common terms.y ( n – 2 ) = 1 + 3 2 2 n≥0 (A.= 0 2 2 (A.a + -.23) are -a1 = 1 2 a2 = 1 and by substitution into (A. we observe that both relations (A. in accordance with the first and third rows of Table A.y ( n – 2 ) = 0 2 2 n≥0 (A.23) (A.2 Find the total solution for the second−order difference equation 1 3 –n y ( n ) – -. Third Edition Copyright © Orchard Publications A−5 . To avoid the duplication.19) we obtain Numerical Analysis Using MATLAB® and Excel®.27) into (A. the constants k 2 and k 3 .y ( n – 1 ) + -. that is. We assume that the complementary solution y C ( n ) has the form yC ( n ) = k1 a1 + k2 a2 n n (A.25) 2.a + -. we choose the particular solution as yP ( n ) = k3 n + k4 3 –n (A.a = 0 2 2 (A.⎞ + k 2 ( 1 ) = k 1 2 + k 2 ⎝2⎠ (A.20) The homogeneous equation of (A.27) and by substitution of (A.y ( n – 1 ) + -.

28). (A.30) and (A. that is.29) To evaluate the constants k 1 and k 2 we use the given initial conditions.⎞ k 3 n + ⎛ -.29) we obtain the total solution as A−6 Numerical Analysis Using MATLAB® and Excel®.⎞ k 3 ( n – 1 ) – ⎛ -.⎞ k 4 3 + -. Third Edition Copyright © Orchard Publications .28) The total solution is the addition of (A.29) reduces to y ( –1 ) = k1 2 + k2 – 2 + 3 = 2 1 1 from which 2k 1 + k 2 = 1 (A.30) For n = – 1 .Appendix A Difference Equations in Discrete-Time Systems 3 3 1 1 –n –( n – 1 ) –( n – 2 ) –n k 3 n + k 4 3 – ⎛ -.32) and by substitution into (A.⎞ k 3 – k 3 = 1 + 3 ⎝2 ⎠ Equating like terms. y P ( n ) = 2n + 3 –n (A. s y ( – 2 ) = 0 and y ( – 1 ) = 2 .k 3 ( n – 2 ) + ⎛ -. y ( n ) = y C ( n ) + y P ( n ) = k 1 2 + k 2 + 2n + 3 –n –n (A. we obtain ⎛ 3 ⎞k – k = 1 -3 ⎝2 ⎠ 3 k4 3 –n = 3 –n and after simplification.29) reduces to y ( –2 ) = k1 2 + k2 – 4 + 9 = 0 2 from which 4k 1 + k 2 = – 5 (A.⎞ k 4 3 + -.⎞ k 4 3 = 1 + 3 ⎝2 ⎠ ⎝2⎠ ⎝2 ⎠ ⎝2 ⎠ 2 3 –n –n k 4 3 + ⎛ -.31) Simultaneous solution of (A.31) yields k1 = –3 k2 = 7 (A.⎞ k 3 – ⎛ -.27). i.⎞ k 4 3 = 1+3 ⎝2 ⎠ ⎝2⎠ ⎝2⎠ 2 3 1 9 3 9 –n –n –n –n k 3 n + k 4 3 – ⎛ -. k3 = 2 k4 = 1 By substitution into (A. For n = – 2 ..e.25) and (A. (A.k 3 n – k 3 + ⎛ -.

yn=(−3).2.35) The homogeneous equation of (A.Method of Undetermined Coefficients y ( n ) = y C ( n ) + y P ( n ) = ( – 3 )2 + 7 + 2n + 3 –n –n n≥0 (A.^(−n).9y ( n – 1 ) = 0 n≥0 (A.9a n n–1 = 0 (A.37) Numerical Analysis Using MATLAB® and Excel®.*2.^(−n)+7+2. stem(n.*n+3. Plot for the difference equation of Example A. we use the following MATLAB script: n=0:1:10.33) To plot this difference equation for the interval 0 ≤ n ≤ 10 . We assume that the complementary solution y C ( n ) has the form yC ( n ) = k1 a n (A.2 Example A. grid 30 25 20 15 10 5 0 0 1 2 3 4 5 6 7 8 9 10 Figure A.5 + ( 0.yn). Third Edition Copyright © Orchard Publications A−7 .9 ) n–1 n≥0 (A.35) yields a – 0.9y ( n – 1 ) = 0.34) subject to the initial condition y ( – 1 ) = 5 Solution: 1.36) Substitution of y ( n ) = a n into (A.34) is y ( n ) – 0.3 Find the total solution for the first-order difference equation y ( n ) – 0.

n ( 0.9 ) n (A.39) 2.5 + ( 0.5 + ( 0.9 ) 0.39) and (A.39) and (A. that is.9 = 0 a = 0.9 ) + 5 9 (A.9 ) + ----.9 (A.9 ) 0.9 ) n and k 3 ( 0.34) we obtain k 2 + k 3 n ( 0.9 ) 9 (A.5 + ( 0.9 ) n n n –1 n –1 = 0.41) into (A.1.43) A−8 Numerical Analysis Using MATLAB® and Excel®.9 ) n (A. the constants k 1 ( 0.9 ) –1 n Equating like terms.9 ) n–1 + 0. To avoid the duplication.n ( 0.9 0.42) The total solution is the addition of (A.9 ) (n – 1) = 0. Since the forcing function is 0.1k 2 + k 3 n ( 0.9 ) – k 3 n ( 0.9 ) n n n (n – 1) n (n – 1) = 0. we obtain 0. we would assume that the particular solution is y P ( n ) = k 2 + k 3 ( 0.9 ) n–1 0.9 ) n – 1 . we choose the particular solution as y P ( n ) = k 2 + k 3 n ( 0.9 ) – 0.9 ) ( 0.9 ) ( 0. k2 = 5 ----k 3 = 10 9 By substitution into (A. 10 n n y ( n ) = y C ( n ) + y P ( n ) = k 1 ( 0.9k 3 ( 0. that is.9k 2 – 0.9k 3 n ( 0.1k 2 + k 3 n ( 0.9k 3 ( n – 1 ) ( 0.9k 3 ( 0.37) by a n – 1 yields a – 0.9 ) n (A.1k 2 = 0.9 + 0.5 + ( 0.9 ) n –1 n and after simplification.38) and by substitution into (A. in accordance with the first and third rows of Table A.9 ) = ( 0.9k 3 n ( 0.9 ) = 0.41).40) contain common terms.41) and by substitution of (A.42). we observe that both relations (A.Appendix A Difference Equations in Discrete-Time Systems Division of (A.5 + ( 0.1k 2 + k 3 n ( 0.5 k 3 ( 0. Third Edition Copyright © Orchard Publications .40) However.9 ) n .9 ) – 0.9 ) + k 3 ( 0. 10 n y P ( n ) = 5 + ----.9 ) – 0.5 + ( 0.35) we obtain y C ( n ) = k 1 ( 0.9 ) n–1 n–1 = 0.9 ) 0.

3.9 ) +5 n–1 +5 y ( n ) = ( n + 1 ) ( 0. yn=(n+1). y ( – 1 ) = 5 .yn). Plot for the difference equation of Example A.9). we use the following MATLAB script: n=0:1:10. Third Edition Copyright © Orchard Publications A−9 .81y ( n – 2 ) = 2 –n n≥0 (A.45) To plot this difference equation for the interval 0 ≤ n ≤ 10 . For n = – 1 .46) Numerical Analysis Using MATLAB® and Excel®. grid 10 9 8 7 6 5 4 3 2 1 0 0 1 2 3 4 5 6 7 8 9 10 Figure A.k 1 – 100 = 0 -------9 81 from which 10 k 1 = ----9 (A.9 ) n–1 n≥0 (A.44) and by substitution into (A.43) reduces to 10 –1 –1 y ( – 1 ) = k 1 ( 0..9 ) + 5 = 5 9 10 ----. stem(n. i.^(n-1)+5.( – 1 ) ( 0.3 Example A.9 ) n–1 + n ( 0.43) we obtain the total solution as y ( n ) = ( 0.e.9 ) + ----. (A.*(0.Method of Undetermined Coefficients To evaluate the constant k 1 we use the given initial condition.4 Find the total solution for the second−order difference equation y ( n ) – 1.8y ( n – 1 ) + 0.

8y ( n – 1 ) + 0.8 )2 + ( 0.9 ) + k 2 n ( 0.81 = 0 2 (A.53) –( n – 2 ) and by substitution into (A. 1.64 16 2 A−10 Numerical Analysis Using MATLAB® and Excel®.6 + 3.47) The homogeneous equation of (A.= ----0.46).8 )2 –n –( n – 1 ) + k 3 ( 0. Since the forcing function is 2 –n . a 1 = a 2 = 0.48) yields a – 1.51) The roots of (A.46) is y ( n ) – 1. that is. Third Edition Copyright © Orchard Publications . k 3 2 – k 3 ( 1.48) Substitution of y ( n ) = a n into (A. We assume that the complementary solution y C ( n ) has the form yC ( n ) = k1 a1 + k2 a2 n n (A.81 )2 ] = 1 k 3 [ 1 – 3.50) are repeated roots.8a n n–1 + 0.49) Division of (A. we assume that the particular solution is yP ( n ) = k3 2 –n (A.81y ( n – 2 ) = 0 n≥0 (A.52) 2.81a n–2 = 0 (A.9 ) n n (A.8a + 0.Appendix A Difference Equations in Discrete-Time Systems subject to the initial conditions y ( – 2 ) = 25 and y ( – 1 ) = 6 Solution: No initial conditions are given and thus we will express the solution in terms of the unknown constants.50) (A. we accept the complementary solution to be of the form y C ( n ) = k 1 ( 0.81 )2 = 2 –n Division of both sides by 2 –n yields k 3 [ 1 – ( 1.9 and as in the case of ordinary differential equations.24 ] = 1 1 25 k 3 = --------.49) by a n – 2 yields a – 1.

⎞ ⎝2⎠ n≥0 (A.57).9 ) n .9 ) + k 2 n ( 0.9 ) n but this is of the same form as the second term on the right side of (A. But this is of the same form as the first term on the right side of (A. the proper choice would be y P ( n ) = k 3 n ( 0. 25 –n n n y ( n ) = y C ( n ) + y P ( n ) = k 1 ( 0.57) Row 3 in Table A.59) Solution: From Row 4 in Table A.9 ) n n≥0 (A.57). the particular solution has the form nπ nπ y P ( n ) = k 1 sin ⎛ ----.Method of Undetermined Coefficients and thus 25 –n y P ( n ) = ⎛ ----. Therefore.9 ) n instead of 2 –n where we found that the complementary solution is y C ( n ) = k 1 ( 0.9 ) n n (A.4 where the forcing function is ( 0.60) Numerical Analysis Using MATLAB® and Excel®.5y ( n – 1 ) = sin ⎛ ----. that is. Third Edition Copyright © Orchard Publications A−11 .58) Example A.55) Example A.9 ) 2 n (A.54).6 Find the particular solution for the first-order difference equation nπ y ( n ) – 0.5 For the second−order difference equation y ( n ) – 1.1 indicates that a good choice for the particular solution would be k 3 ( 0.1 we see that for a sinusoidal forcing function.81y ( n – 2 ) = ( 0.56) what would be the appropriate choice for the particular solution? Solution: This is the same difference equation as that of Example A.⎞ + k 2 cos ⎛ ----.⎞ ⎝ 2⎠ ⎝ 2⎠ (A.⎞ 2 ⎝ 16 ⎠ (A.9 ) + k 2 n ( 0.⎞ 2 ⎝ 16 ⎠ (A.54) The total solution is the addition of (A.9 ) + ⎛ ----. The next choice would be a term of the form k 3 n ( 0.8y ( n – 1 ) + 0.52) and (A.

64) and simultaneous solution of (A.60) into (A.⎞ – 2 cos ⎛ ----.5k 1 + k 2 = 0 (A.⎞ ⎝ 2⎠ ⎝ 2⎠ ⎝2⎠ 2 2 nπ nπ nπ nπ π nπ π .⎞ – 0.59) nπ nπ ( n – 1 )π ( n – 1 )π nπ k 1 sin ⎛ ----.63) (A.⎞ = sin ⎛ ----.⎞ – 0.⎞ ⎝ 2⎠ ⎝ 2⎠ ⎝ 2⎠ ⎝ 2⎠ ⎝2⎠ (A.5k 1 sin ------------------. the particular solution of (A. sin ⎛ θ – π ⎞ = – cos θ -⎝ 2⎠ π cos ⎛ θ – -.= sin ⎛ ----.5k 2 cos ------------------. Third Edition Copyright © Orchard Publications .– -.⎝ 2 ⎠ 2 2 and by substitution into (A.62) Equating like terms.– -.⎞ ⎝ 2⎠ ⎝ 2⎠ 5 5 (A.= sin ⎛ ----. nπ nπ π sin ----.= – cos ⎛ ----.⎞ + k 2 cos ⎛ ----.⎞ – 0..5k 1 sin ----.⎞ .⎞ + 0.65) A−12 Numerical Analysis Using MATLAB® and Excel®.– 0.– -.– -.⎞ .⎞ = sin θ ⎝ 2⎠ Then.61) From trigonometry.⎞ + k 2 cos ⎛ ----.5k 2 = 1 0.k 1 sin ⎛ ----.⎝ 2 ⎠ 2 2 nπ nπ π cos ----.5k 1 cos ⎛ ----.Appendix A Difference Equations in Discrete-Time Systems and by substitution of (A.– 0.61) nπ nπ nπ nπ nπ k 1 sin ⎛ ----.-y P ( n ) = -.5k 2 cos ----.64) yields 4 k 1 = -5 2 k 2 = – -5 Therefore.sin ⎛ ----.59) is nπ nπ 4 .⎞ ⎝2⎠ ⎝ 2⎠ ⎝ 2⎠ 2 2 2 2 (A.= sin ⎛ ----.5k 2 sin ⎛ ----. we obtain k 1 – 0.⎞ + k 2 cos ⎛ ----.63) and (A.

We invoke Simulink from within MATLAB. B. and for this purpose. and revise our models in either environment at any point. We will compute v c ( t ) . and v c ( 0 − ) = 0.1) and by Kirchoff’s voltage law (KVL).+ v C = u 0 ( t ) dt (B. This author feels that we can best introduce Simulink with a few examples.1. Circuit for Example B.1. Some familiarity with MATLAB is essential in understanding Simulink. We will introduce Simulink with a few illustrated examples. and thus we can analyze.1 For this example. the initial conditions are i L ( 0 − ) = 0 . di L Ri L + L -----.5 V .Appendix B Introduction to Simulink® T his appendix is a brief introduction to Simulink. dv C i = i L = i C = C -------dt (B. it is highly recommended that the novice to MATLAB reader reviews Chapter 1 which serves as an introduction to MATLAB.1 Simulink and its Relation to MATLAB The MATLAB® and Simulink® environments are integrated into one entity.1 For the circuit of Figure B.2) yields Numerical Analysis Using MATLAB® and Excel®.2) Substitution of (B. Example B.1) into (B. Third Edition Copyright © Orchard Publications B−1 . simulate. R L 1⁄4 H C 1Ω + vC ( t ) − + − i(t) 4⁄3 F vs ( t ) = u0 ( t ) Figure B.

for comparison.+ 4 -------. the natural response is the sum of the terms k 1 e be –s1 t –s t –s t – at and k 2 e –s2 t .5) To appreciate Simulink’s capabilities.3) Substituting the values of the circuit constants and rearranging we obtain: 1 d v C 4 dv C -. the total solution will –s1 t –s2 t v c ( t ) = natural response + forced response = v cn ( t ) + v cf ( t ) = k 1 e + k2 e + v cf ( t ) (B.+ 4 -------.3 dt 2 3 dt dv C d vC ---------. Appendix B for a thorough discussion.-------.4) t>0 (B.5) is a second−order.+ -. and thus the complete solution will consist of the sum of the forced response and the natural response. B−2 Numerical Analysis Using MATLAB® and Excel®. three different methods of obtaining the solution are presented. non−homogeneous differential equation with constant coefficients. and the solution using Simulink follows. decaying exponentials of the form ke where k and a are constants. First Method − Assumed Solution Equation (B. Third Edition Copyright © Orchard Publications .6) The values of s 1 and s 2 are the roots of the characteristic equation * Please refer to Circuit Analysis II with MATLAB Applications.---------. It can be shown* that if k 1 e 1 and k 2 e 2 where k 1 and k 2 are constants and s 1 and s 2 are the roots of the characteristic equation of the homogeneous part of the given differential equation. are possible candidates since their derivatives have the same form but alternate in sign. ISBN 0−9709511−5−9.Introduction to Simulink® d vC dv C RC -------.+ v C = u 0 ( t ) 2 dt dt 2 (B. Therefore. the solution cannot contain sinusoidal functions (sine and cosine) since the derivatives of these are also sinusoids.+ LC ---------.+ v C = u 0 ( t ) .+ 3v C = 3 2 dt dt 2 2 (B..+ 3v C = 3u 0 ( t ) 2 dt dt 2 dv C d vC ---------. It is obvious that the solution of this equation cannot be a constant since the derivatives of a constant are zero and thus the equation is not satisfied. However. Also.

= --dt dt C dv C -------dt iL ( 0 ) 0 = ----------. Third Edition Copyright © Orchard Publications B−3 . and dv C dv C i L i L = i C = C -------.13) t=0 Next. 2 dv C d vC ---------.8).13).12) Also.e..= --.5 0 0 (B.14) By equating the right sides of (B.14) we obtain Numerical Analysis Using MATLAB® and Excel®.7) yields of s 1 = – 1 and s 2 = – 3 and with these values (B. we differentiate (B. First.9) is a constant. and equate it with (B. Thus. we evaluate it at t = 0 .= 0 C C (B.+ 4 -------.6) is written as vc ( t ) = k1 e + k2 e –t –3 t + v cf ( t ) (B. i.11) The constants k 1 and k 2 will be evaluated from the initial conditions.+ 3v C = 3 2 dt dt t>0 (B.9) we obtain 0 + 0 + 3k 3 = 3 or v Cf = k 3 = 1 (B.5 k 1 + k 2 = – 0.. dv C -------dt = – k 1 – 3k 2 t=0 (B.7) Solution of (B.11) at t = 0 .Simulink and its Relation to MATLAB s + 4s + 3 = 0 2 (B. -------. using v C ( 0 ) = 0.5 V and evaluating (B. the forced response will also be a constant and we denote it as v Cf = k 3 . By substitution into (B.13) and (B. yields the total solution as v C ( t ) = v Cn ( t ) + v Cf = k 1 e + k 2 e –t –3 t +1 (B.10) Substitution of this value into (B.11).9) Since the right side of (B.5). we obtain v C ( 0 ) = k 1 e + k 2 e + 1 = 0.8) The forced component v cf ( t ) is found from (B.

⎛ -.e – 3 e ⎞ = e – e A -⎠ dt 3⎝ 4 4 (B.i = i L = i C = C --------. y1=diff(y0) % Define symbolic variable t % The total solution y(t)..2.17) Second Method − Using the Laplace Transformation The transformed circuit is shown in Figure B.1 B−4 Numerical Analysis Using MATLAB® and Excel®. By substitution into (B.75 e + 0.16) Check with MATLAB: syms t y0=−0.15). gives k 1 = – 0. R 1 L 0.Introduction to Simulink® – k 1 – 3k 2 = 0 (B.2.12) and (B.25*exp(−3*t)+1.25 . the solution has been verified by MATLAB.25e –t –3 t + 1 )u 0 ( t ) (B. we obtain the total solution as v C ( t ) = ( – 0.2) % The second derivative of y(t) y2 = -3/4*exp(-t)+9/4*exp(-3*t) y=y2+4*y1+3*y0 % Summation of y and its derivatives y = 3 Thus. Third Edition Copyright © Orchard Publications . we find the expression for the current as dv C 4 3 –t – 3t – t – 3t . for our example. vc(t) % The first derivative of y(t) y1 = 3/4*exp(-t)-3/4*exp(-3*t) y2=diff(y0.16).5 ⁄ s + V (0) C − Figure B.15) Simultaneous solution of (B. Using the expression for v C ( t ) in (B.25s C 3 ⁄ 4s + VC ( s ) − Vs ( s ) = 1 ⁄ s + − I(s) 0. Transformed Circuit for Example B.8).75 and k 2 = 0.75*exp(−t)+0.= -.

† Partial fraction expansion is discussed in Chapter 12. Signals and Systems with MATLAB Computing and Simulinl Modeling. ISBN 0−9709511−2−4. u ( t ) denotes any input.+ --------------s (s + 1) (s + 3) s(s + 1)(s + 3) 2 Taking the Inverse Laplace transform‡ we find that v C ( t ) = 1 – 0. ‡ For an introduction to Laplace Transform and Inverse Laplace Transform.5s + 2s + 31.5s + 2s + 3 r 3 = --------------------------------s(s + 1) 2 2 2 = 1 s=0 = – 0.= --. Numerical Analysis Using MATLAB® and Excel®.Simulink and its Relation to MATLAB By the voltage division* expression. please refer to Chapters 2 and 3.+ --------------s (s + 1) (s + 3) s(s + 1 )( s + 3) (B.18) 0.5s + 2s + 3----------------------------------.25e –t – 3t Third Method − Using State Variables di L Ri L + L -----.= 1 + --------------.+ v C = u 0 ( t ) ** dt * For derivation of the voltage division and current division expressions.25s + 3 ⁄ 4s ) ⎝ s s s ⎠ s s(s + 1)(s + 3) s ( s + 4s + 3 ) Using partial fraction expansion. ISBN 0-9744239-9-8. this text. we will denote the Unit Step Function as u0 ( t ) . ISBN 09744239-9-8. For a detailed discussion on State−Space and State Variables Analysis.25 0.75e + 0.75 s = –1 = 0.5s + 2s + 3-.----------V C ( s ) = --------------------------------------------.5⎞ + 0. please refer to Chapter 5. in State−Space and State Variables Analysis. Third Edition Copyright © Orchard Publications B−5 .5s + 2s + 3 r 1 = --------------------------------(s + 1)(s + 3) --------------------------------r 2 = 0.5s + 2s + 3 s(s + 3) 0.⋅ ⎛ 1 – 0.5 = -------------------------------.-----. For distinction.5 = ----------------------------------2 ( 1 + 0.– 0. 2 3 ⁄ 4s 0.25 s = –3 and by substitution into (B.18) 0. Signals and Systems with MATLAB Computing and Simulinl Modeling.-.† we let 2 r2 r3 r1 0.+ --------------.5 .+ 0.75V C ( s ) = ----------------------------------. ** Usually. please refer to Circuit Analysis I with MATLAB Applications.

21) Also. di L · x 1 = -----. Signals and Systems with MATLAB Computing and Simulinl Modeling. or dv C i L = C -------dt dv C 4· · x 1 = i L = C -------. and (B. ISBN 0-9744239-9-8. Third Edition Copyright © Orchard Publications .x 1 4 and in matrix form. B−6 Numerical Analysis Using MATLAB® and Excel®.19).= – 4i L – 4v C + 4 dt (B.22) Therefore. we define the state variables x 1 = i L and x 2 = v C . x = dx ⁄ dt .19) Next. · x x1 = –4 –4 1 + 4 u0 ( t ) ·2 3 ⁄ 4 0 x2 0 x (B. and thus. we obtain the state equations · x 1 = – 4x 1 – 4x 2 + 4 3 · x 2 = -.23) is given in Chapter 5.20) and dv C · x 2 = -------dt (B.23) yields · · * The notation x (x dot) is often used to denote the first derivative of the function x .4 dt or di L -----.x 1 4 (B. Then.* dt (B. that is. (B. from (B.23) Solution† of (B.x 2 3 dt 3 · x 2 = -.22). † The detailed solution of (B.20).= Cx 2 = -. we obtain di 1 ------.L = ( – 1 )i L – v C + 1 .Introduction to Simulink® By substitution of given values and rearranging.

75e + 0. the left side is referred to as the Tree Pane and displays all Simulink libraries installed.25e –t – 3t Then.= – 4 -------.1 as 2 dv C d vC ---------. Numerical Analysis Using MATLAB® and Excel®. Make sure that Simulink is installed in your system. The Simulink icon Upon execution of the Simulink command.3.3. It appears on the top bar on MATLAB’s Command prompt.– 3v C + 3u 0 ( t ) 2 dt dt (B. The right side is referred to as the Contents Pane and displays the blocks that reside in the library currently selected in the Tree Pane.Simulink and its Relation to MATLAB x1 x2 = e –e –t –t – 3t – 3t 1 – 0.5.25e –t (B.4. Let us express the differential equation of Example B. Third Edition Copyright © Orchard Publications B−7 . We will use Simulink to draw a similar block diagram.4.75 e + 0.1 with Simulink To run Simulink. we must first invoke MATLAB. x1 = iL = e –e (B. In the MATLAB Command prompt.26) A block diagram representing relation (B. we can click on the Simulink icon shown in Figure B. all Simulink block diagrams will be referred to as models. In Figure B. the Commonly Used Blocks appear as shown in Figure B. Figure B.26) above is shown in Figure B.24) – 3t and x 2 = v C = 1 – 0. we type: simulink Alternately.* * Henceforth.25) Modeling the Differential Equation of Example B.

The Simulink Library Browser d vC ---------2 dt 2 u0 ( t ) 3 Σ ∫ dt −4 dv C -------dt ∫ dt vC −3 Figure B. On the Simulink Library Browser.6. we click on the leftmost icon shown as a blank page on the top title bar. Block diagram for equation (B.Introduction to Simulink® Figure B.26) using Simulink. Third Edition Copyright © Orchard Publications . A new model window named untitled will appear as shown in Figure B.5.26) To model the differential equation (B. B−8 Numerical Analysis Using MATLAB® and Excel®. we perform the following steps: 1.4.

and on the right side we scroll down until we see the unit step function shown as Step. With the Equation_1_26 model window and the Simulink Library Browser both visible. We save file Equation_1_26 using the File drop menu on the Equation_1_26 model window (right side of Figure B. If the Equation_1_26 model window is no longer visible. we click on the Sources appearing on the left side list. The new model window will now be shown as Equation_1_26.8). and draw a straight line to connect the two Numerical Analysis Using MATLAB® and Excel®.8. 3. The gain block in Simulink is under Commonly Used Blocks (first item under Simulink on the Simulink Library Browser).6.6 is the model window where we enter our blocks to form a block diagram. Model window for Equation_1_26.5. The window of Figure B. This is done from the File drop menu of Figure B.mdl. Third Edition Copyright © Orchard Publications B−9 . We select it. we observe that we need to connect an amplifier with Gain 3 to the unit step function block. The triangle on the right side of the unit step function block and the > symbols on the left and right sides of the gain block are connection points. The Untitled model window in Simulink. Figure B.Simulink and its Relation to MATLAB Figure B. With reference to block diagram of Figure B. and we drag it into the Equation_1_26 model window which now appears as shown in Figure B. See Figure B.7. We save this as model file name Equation_1_26. Simulink will add the extension . See Figure B. and all saved files will have this appearance.6 where we choose Save as and name the file as Equation_1_26.8. We point the mouse close to the connection point of the unit step function until is shows as a cross hair. We choose the gain block and we drag it to the right of the unit step function.7.mdl file 2. See Figure B. it can be recalled by clicking on the white page icon on the top bar of the Simulink Library Browser.8. 4.

See Figure B.* We double−click on the gain block and on the Function Block Parameters.9.8. and left−clicking on the destination block. Third Edition Copyright © Orchard Publications . B−10 Numerical Analysis Using MATLAB® and Excel®.Introduction to Simulink® blocks. Dragging the unit step function into File Equation_1_26 Figure B.9. we change the gain from 1 to 3. then holding down the Ctrl key. Figure B. File Equation_1_26 with added Step and Gain blocks * An easy method to interconnect two Simulink blocks is by clicking on the source block to select it.

and we drag it into the Equation_1_26 model window.11. and we copy and paste it twice. we specify 3 inputs. we add the Scope block which is found in the Commonly Used Blocks on the Simulink Library Browser. From the Commonly Used Blocks of the Simulink Library Browser. we need to add a thee−input adder. We double click it. and we label these as Gain 2 and Gain 3. we double−click on these gain blocks and on the Function Block Parameters window.12. we choose the Integrator block.12. We flip the pasted Gain blocks by using the Flip Block command from the Format drop menu. Figure B.10. Next. We click on the text “Integrator” under the first integrator block. Finally. we drag it into the Equation_1_26 model window.10. Figure B. we change the gains from to −4 and −3 as shown in Figure B.11. To complete the block diagram. We select it. and on the Function Block Parameters window which appears. we change the text “Integrator 1” under the second Integrator to “Integrator 2” as shown in Figure B. File Equation_1_26 complete block diagram Numerical Analysis Using MATLAB® and Excel®. We repeat this step and to add a second Integrator block. and we change it to Integrator 1. we click on the Gain block. Then. File Equation_1_26 with added gain block 6. and we connect it to the output of the Add block. File Equation_1_26 with the addition of two integrators 7. The adder block appears on the right side of the Simulink Library Browser under Math Operations. Third Edition Copyright © Orchard Publications B−11 .Simulink and its Relation to MATLAB 5. We then connect the output of the of the gain block to the first input of the adder block as shown in Figure B. Figure B.

Figure B. The waveform for the function v C ( t ) for Example B. This is done by specifying the simulation time to be 10 seconds on the Configuration Parameters from the Simulation drop menu. and then clicking on the Autoscale icon.13. Third Edition Copyright © Orchard Publications . is to use State− Space block from Continuous in the Simulink Library Browser.14.1. we double click on the Scope block.1 with the State−Space block.1 Another easier method to obtain and display the output v C ( t ) for Example B. 9. Figure B. To see the output waveform.13.Introduction to Simulink® 8. We also need to specify the simulation time. and v c ( 0 ) = 0. as shown in Figure B. We can start the simulation on Start from the Simulation drop menu or by clicking on the icon. The initial conditions i L ( 0 − ) = C ( dv C ⁄ dt ) t=0 = 0 .5 V are entered by − double clicking the Integrator blocks and entering the values 0 for the first integrator. B−12 Numerical Analysis Using MATLAB® and Excel®. and 0. we obtain the waveform shown in Figure B.5 for the second integrator. Obtaining the function v C ( t ) for Example B.14.

15. From Equation B. reside in the MATLAB Workspace.23.Simulink and its Relation to MATLAB The simout To Workspace block shown in Figure B. Numerical Analysis Using MATLAB® and Excel®. · x1 x = –4 –4 1 + 4 u0 ( t ) ·2 3 ⁄ 4 0 x2 0 x The output equation is or y = Cx + du y = [0 1] x1 x2 + [ 0 ]u We double−click on the State−Space block.15. This gives us the ability to delete or modify selected variables.14 writes its input to the workspace. The Function block parameters for the State−Space block. We issue the command who to see those variables. The data and variables created in the MATLAB Command window. Page B−6. This block writes its output to an array or structure that has the name specified by the block's Variable name parameter. and in the Functions Block Parameters window we enter the constants shown in Figure B. Figure B. Third Edition Copyright © Orchard Publications B−13 .

2 A fourth−order network is described by the differential equation 4 d y + a d y + a d y + a dy + a y ( t ) = u ( t ) --------3 -------2 ------1 ----0 3 2 4 dt dt dt dt 3 2 (B.5.16.1 with the State−Space block. and the initial conditions are y ( 0 ) = y' ( 0 ) = y'' ( 0 ) = y''' ( 0 ) = 0 . x2=0. Third Edition Copyright © Orchard Publications . Example B.27) where y ( t ) is the output representing the voltage or current of the network. We will express (B.27) as a set of state equations B−14 Numerical Analysis Using MATLAB® and Excel®. and u ( t ) is any input. The waveform for the function v C ( t ) for Example B. to start the simulation we click clicking on the icon. we double click on the Scope block.Introduction to Simulink® The initials conditions [ x1 x2 ]' are specified in MATLAB’s Command prompt as x1=0. Figure B. we form. and then clicking on the Autoscale obtain the waveform shown in Figure B.16. a. and to see the output waveicon. As before. The state−space block is the best choice when we need to display the output waveform of three or more variables as illustrated by the following example.

28) subject to the initial conditions y ( 0 ) = y' ( 0 ) = y'' ( 0 ) = y''' ( 0 ) = 0 . The differential equation of (B.31) and in matrix form · x1 · x2 · x3 · x4 0 0 = 0 –a0 1 0 0 –a1 0 1 0 –a2 0 0 1 –a3 0 x2 + 0 u(t) x3 0 1 x4 x1 (B. It is known that the solution of the differential equation 4 d y d y ------. 1. and x 4 .= x 4 = – a 0 x 1 – a 1 x 2 – a 2 x 3 – a 3 x 4 + u ( t ) 4 dt 4 (B. and using Simulink. (B.Simulink and its Relation to MATLAB b.32) is written as · x = Ax + bu (B.+ y ( t ) = sin t 2 4 dt dt 2 (B.+ 2 ------. therefore.125 [ ( 3 – t ) – 3t cos t ] 2 (B. Third Edition Copyright © Orchard Publications B−15 . we will display the waveform of the output y ( t ) . x 2. and a 0 so that the new set of the state equations will represent the differential equation of (B. the output is where y = Cx + du Numerical Analysis Using MATLAB® and Excel®. x 3 .33) (B. we must define four state variables that will be used with the four first−order state equations.30) We observe that · x1 = x2 · x2 = x3 · x3 = x4 d y · --------.32) In compact form.28).28) is of fourth−order. a 1. We denote the state variables as x 1. we will select appropriate values for the coefficients a 3.34) Also. has the solution y ( t ) = 0. and we relate them to the terms of the given differential equation as x1 = y ( t ) dy x 2 = ----dt d y x 3 = -------2 dt 2 d y x 4 = -------3 dt 3 (B. a 2.29) In our set of state equations.

38) Since the output is defined as in matrix form it is expressed as B−16 Numerical Analysis Using MATLAB® and Excel®.36) 2.35) and since the output is defined as relation (B.37) where · x1 · x2 · x3 · x4 0 0 A= 0 –a0 1 0 0 0 0 1 0 –2 0 0 . 0 1 · x= . 0 b= 0.27) will be reduced to the differential equation of (B. the differential equation of (B.34) is expressed as x1 y = [1 0 0 0] ⋅ x2 x3 x4 + [ 0 ]u ( t ) (B. 1 –a3 x1 x= x2 x3 x4 y ( t ) = x1 .28) if we let a3 = 0 a2 = 2 a1 = 0 a0 = 1 u ( t ) = sin t and thus the differential equation of (B.Introduction to Simulink® · x1 · x2 · x3 · x4 0 0 A= 0 –a0 1 0 0 –a1 0 1 0 –a2 0 0 . 0 1 · x= . 0 b= 0. and u = u ( t ) (B.28) can be expressed in state−space form as · x1 · x2 · x3 · x4 0 0 = 0 –a0 1 0 0 0 0 1 0 –2 0 0 1 0 0 + 0 sin t x3 0 1 x4 x2 x1 (B. and u = sin t (B. Third Edition Copyright © Orchard Publications . By inspection. 1 0 x1 x= x2 x3 x4 y ( t ) = x1 .

we drag the Signal Generator block on the Example_1_2 model window.2 We now double−click on the Signal Generator block and we enter the following in the Function Block Parameters: Wave form: sine Time (t): Use simulation time Amplitude: 1 Frequency: 2 Units: Hertz Next. 0 0 1 0. we click and drag the State−Space block from the Continuous on Simulink Library Browser. we start Simulink by clicking on the Simulink icon. 0 0 0 1. we double−click on the state−space block and we enter the following parameter values in the Function Block Parameters: A: [0 1 0 0. −a0 −a1 −a2 −a3] B: [0 0 0 1]’ C: [1 0 0 0] D: [0] Initial conditions: x0 Numerical Analysis Using MATLAB® and Excel®. We also add the Display block found under Sinks on the Simulink Library Browser.Simulink and its Relation to MATLAB x1 y = [1 0 0 0] ⋅ x2 x3 x4 + [ 0 ] sin t (B.39) We invoke MATLAB. Block diagram for Example B. Figure B. and we save this model as Example_1_2. Third Edition Copyright © Orchard Publications B−17 . We connect these four blocks and the complete block diagram is as shown in Figure B.17. On the Simulink Library Browser we select Sources.17. on the Simulink Library Browser we click on the Create a new model (blank page icon on the left of the top bar). and we click and drag the Scope block from the Commonly Used Blocks on the Simulink Library Browser.

z 2 .18. we will create a model that will produce the simultaneous solution of three equations with three unknowns. x0=[0 0 0 0]’. a1=0. a3=0. To see the output waveform. then clicking on the Autoscale icon. we switch to the MATLAB Command prompt and we type the following: >> a0=1.2 The Display block in Figure B.18. Figure B. we obtain the waveform shown in Figure B. a2=2.Introduction to Simulink® Absolute tolerance: auto Now.40) B−18 Numerical Analysis Using MATLAB® and Excel®.2 using four Integrator blocks by this approach would have been more time consuming. Waveform for Example B. and we start the simulation by clicking on the icon. and z 3 in the system of the equations a1 z1 + a2 z2 + a3 z3 + k1 = 0 a4 z1 + a5 z2 + a6 z3 + k2 = 0 a7 z1 + a8 z2 + a9 z3 + k3 = 0 (B.3 Using Algebraic Constraint blocks found in the Math Operations library. Display blocks found in the Sinks library. we double click on the Scope block.2 have clearly illustrated that the State−Space is indeed a powerful block. Third Edition Copyright © Orchard Publications .1 and B. We change the Simulation Stop time to 25 . The model will display the values for the unknowns z 1 . We could have obtained the solution of Example B.17 shows the value at the end of the simulation stop time. and Gain blocks found in the Commonly Used Blocks library. Examples B. Example B.

a8=1.These values are shown in the Display blocks of Figure B. Model for Example B. k1=−8. a3=−1. a4=1. we observe the values of the unknowns as z 1 = 2 . the Initial guess parameter is zero.19. a6=4. and z 3 = 5 .3 Next. a7=−6. By default. a9=2. z 2 = – 3 . This enables us to specify algebraic equations for index 1 differential/algebraic systems (DAEs). The Algebraic Constraint block constrains the input signal f ( z ) to zero and outputs an algebraic state z ..19. Numerical Analysis Using MATLAB® and Excel®. After clicking on the simulation icon.. k2=−7.. k3=5. We can improve the efficiency of the algebraic loop solver by providing an Initial guess for the algebraic state z that is close to the solution value.Simulink and its Relation to MATLAB The model is shown in Figure B.19. The block outputs the value necessary to produce a zero at the input. we go to MATLAB’s Command prompt and we enter the following values: a1=2. Third Edition Copyright © Orchard Publications B−19 . The output must affect the input through some feedback path. a2=−3. a5=5. Figure B.

k1=14. a8=4.20. and z 3 for the values specified in MATLAB’s Command prompt.19 represented as a subsystem The Display blocks in Figure B.20 show the values of z 1 . We can double−click on the Subsystem block to see its contents. ISBN 0−9744239−7−1. Chapter 2. a4=11. Then. a7=−8. a2=−1. B−20 Numerical Analysis Using MATLAB® and Excel®. to be shown as a single Subsystem block. The model of Figure B. z 2 . These blocks are described in Section 2. † The contents of the Subsystem block are not lost. It is highly recommended that the reader becomes familiar with the block libraries found in the Simulink Library Browser.Introduction to Simulink® An outstanding feature in Simulink is the representation of a large model consisting of many blocks and lines. k2=−6. the reader can follow the steps delineated in The MathWorks Simulink User’s Manual to run the Demo Models beginning with the thermo model. we choose Create Subsystem from the Edit menu.* For instance.20† where in MATLAB’s Command prompt we have entered: a1=5. Section 2. The Subsystem block replaces the inputs and outputs of the model with Inport and Outport blocks. Introduction to Simulink with Engineering Applications. ISBN 0−9744239−7−1.1.1.. a9=15. Figure B. we can group all blocks and lines in the model of Figure B. Introduction to Simulink with Engineering Applications.. the reader with no prior knowledge of Simulink. a5=6. This model can be seen by typing thermo in the MATLAB Command prompt. should be ready to learn Simulink’s additional capabilities. a6=9. k3=9.. and this model will be shown as in Figure B. B. a3=4. Page 2−2. * The Subsystem block is described in detail in Chapter 2. Third Edition Copyright © Orchard Publications . Page 2−2.2 Simulink Demos At this time.19 except the display blocks.

g.0). other than the zero vector. An example is the two−dimensional Euclidean space denoted as R . −3 −6 8 1. compute the Euclidean norm of the matrix A . The Euclidean norm of a matrix A .. Example C. The Euclidean norm assigns to each vector the length of its arrow.1) and it is computed with the MATLAB function norm(A). we enter A=[−2 5 −4 9. 7 −5 3 2. norm(A) and MATLAB outputs Numerical Analysis Using MATLAB® and Excel®.1 The Norm of a Matrix A norm is a function which assigns a positive length or size to all vectors in a vector space. defined as –2 5 A = –3 –6 7 –5 4 –9 –4 8 3 –8 9 1 2 –1 Solution: At the MATLAB command prompt. (2. We will introduce a reference against which the determinant can be measured to classify a matrix as a well− or ill−conditioned.1 Using the MATLAB function norm(A). is defined as A = 2 ∑ ∑ Aij 2 i=1j=1 n n (C. The elements of the Euclidean vector space (e. C.5)) are usually drawn as arrows in a two−dimensional cartesian coordinate system starting at the origin (0. denoted as A . 4 −9 −8 −1].Appendix C Ill−Conditioned Matrices T his appendix supplements Chapters 4 and 14 with concerns when the determinant of the coefficient matrix is small. Third Edition Copyright © Orchard Publications C−1 .

−3 −6 8 1. 7 −5 3 2. if det ( A ) = 0 . compute the condition number of the matrix A defined as –2 5 A = –3 –6 7 –5 4 –9 –4 8 3 –8 9 1 2 –1 Solution: At the MATLAB command prompt.Appendix C Ill−Conditioned Matrices ans = 14.2 Using the MATLAB function cond(A). the inverse of A is undefined. cond(A) and MATLAB outputs ans = 2. Page 4−22.1) above.5539 C. Third Edition Copyright © Orchard Publications . 4 −9 −8 −1]. The condition number of a matrix A is computed with the MATLAB function cond(A). We recall from Chapter 4 that if the determinant of a square matrix A is singular. Please refer to Chapter 4.2) where A is the norm of the matrix A defined in relation (C.2 Condition Number of a Matrix The condition number of a matrix A is defined as k( A) = A ⋅ A –1 (C. we enter A=[−2 5 −4 9. that is. Example C. Matrices with condition number close to unity are said to be well−conditioned matrices.3724 This condition number is relatively close to unity and thus we classify matrix A as a well-conditioned matrix. and those with very large condition number are said to be ill−conditioned matrices. C−2 Numerical Analysis Using MATLAB® and Excel®.

Hilbert Matrices Now.. A unit fraction is the reciprocal of this integer.. …. … are unit fractions.3 Hilbert Matrices Let n be a positive integer.-. Accordingly. A Hilbert matrix is a matrix with unit fraction elements B ij = 1 ⁄ ( i + j – 1 ) (C.-. . we classify such a matrix as ill−conditioned.. let us consider that the coefficient matrix* is very small.. a mn are the coefficients of the system.-.5 6 1 1 -.. a m1 x 1 + a m2 x 2 + …+a mn x n = b m where x 1. The coefficient matrix is the m × n matrix with the coefficient a ij as the (i. C. that is.-. Third Edition Copyright © Orchard Publications C−3 .8 9 (C.. 1 ⁄ 3.3) Shown below is an example of the 5 × 5 Hilbert matrix.7 8 1 1 -. a 12.6 7 1 1 -..-. almost singular. 1 ⁄ n .. 1 ⁄ 2.e.. . i. 1 ⁄ 1.4 5 1 1 -. Thus.. x 2.. x n are the unknowns and the numbers a 11. . 1 -1 1 -2 1 -3 1 -4 1 -5 1 -2 1 -3 1 -4 1 -5 1 -6 1 -3 1 -4 1 -5 1 -6 1 -7 1 1 -. a system with m linear equations and n unknowns can be written as a 11 x 1 + a 12 x 2 + …+a 1n x n = b 1 a 21 x 1 + a 22 x 2 + …+a 2n x n = b 2 . . ….j)-th entry: a 11 a 12 … a 1n a 21 a 22 … a 2n … … … … a m1 a m1 a m1 a m1 Numerical Analysis Using MATLAB® and Excel®.4) * In general.

187 0. Third Edition Copyright © Orchard Publications . Thus.3 Compute the determinant and the condition number of the 6 × 6 Hilbert matrix using MATLAB.728 0.Appendix C Ill−Conditioned Matrices MATLAB’s function hilb(n) displays the Hilbert n × n matrix.256 Compute the values of the vector x .585 0.464 0. seven decimal places might be lost.256 C−4 Numerical Analysis Using MATLAB® and Excel®. Solution: Here. we are asked to find the values of x 1 and x 2 of the linear system 0.4 Let Ax = b where A = 0.3673e-018 This is indeed a very small number and for all practical purposes this matrix is singular. cond(hilb(6)) ans = 1.379 ⋅ x 1 = 0. we enter det(hilb(6)) and MATLAB outputs ans = 5.4951e+007 This is a large number and if the coefficient matrix is multiplied by this number. Solution: At the MATLAB command prompt.728 0.585 0. We can find the condition number of a matrix A with the cond(A) MATLAB function.379 and b = 0. Let us consider another example. Example C. Example C.187 x2 0. for the 6 × 6 Hilbert matrix.464 0.

Hilbert Matrices Using MATLAB. 0.728 0.585 0.464].4341 3.585 0.378.464].378. b=[0. Example C.256]'. Numerical Analysis Using MATLAB® and Excel®.728 0.187 0. we conclude that this system of equations is ill-conditioned and the solution is invalid. i.9428 Check: A=[0.6265 Therefore. and we use the left division operation. determinant = det(A) determinant = -0.0171 but these are not the given values of the vector b .8852]'.4 above should serve as a reminder that when we solve systems of equations using matrices.8852 b = 2.9428 1. we define A and b . A=[0. 0.. so let us check the determinant and the condition number of the matrix A . Third Edition Copyright © Orchard Publications C−5 .0037 condition=cond(A) condition = 328. b=A*x 1. x=b\A x = 2.e. x=[2. we should check the determinants and the condition number to predict possible floating point and roundoff errors.

References and Suggestions for Further Study A. ISBN 0−9709511−2−4 3. They are available from The MathWorks. ISBN 0-8493-0626-4 . Mathematics for Business. Using MATLAB 3. Science. Handbook of Mathematical Functions. Signals and Systems with MATLAB Computing and Simulink Modeling. 1. ISBN 0-9744239-9-8 6. 3 Apple Hill Drive. ISBN-13: 978−1−934404−01−2 2. and Technology with MATLAB and Excel Computations. ISBN 0−9709511−5−9 4. CRC Standard Mathematical Tables. ISBN 0-4866127-2-4 7. Statistics Toolbox B.com. are listed below. are highly recommended for further study. ISBN 0-9744239-7-1 5. 1. Other references indicated in footnotes throughout this text. MA. Using MATLAB Graphics 4. Getting Started with MATLAB 2. Circuit Analysis II with MATLAB Applications.mathworks. Natick. Financial Toolbox 5. Third Edition. Introduction to Simulink with Engineering Applications. www. Circuit Analysis I with MATLAB Applications. Third Edition. The following publications by The MathWorks. 01760.

7-8 default color in MATLAB 1-14 default line in MATLAB 1-15 default marker in MATLAB 1-14 default values in MATLAB 1-11 degree of differential equation 5-3 demo in MATLAB 1-2 determinant .x) MATLAB function 15-3 BESSELJ(x. 11-1 differences . 1-22 B backward substitution. 8-4 cubic interpolarion in MATLAB 7-25 cubic spline interpolarion in MATLAB 7-25 curve fitting 8-1 curved regression 8-7.see matrix collect(s) MATLAB function 7-15. 8-14 cycle 3-2 cyclic frequency 3-3 D Data Analysis Toolpack in Excel 8-7 data points in MATLAB 1-14 decibel 12 decimal format 2-21 deconv(p. 12-11 column vector 1-19 command screen in MATLAB 1-1 command window in MATLAB 1-1 commas in MATLAB 1-7 comment line in MATLAB 1-2 Commonly Used Blocks in Simulink B-7 complementary solution A-1 complex conjugate 1-4 complex numbers 1-2. 1-21 elements of a matrix .see matrix configuration parameters in Simulink B-12 congugate of a matrix .see matrix eps in MATLAB 1-22.see matrix diagonal of a matrix .n) Excel function 15-3 beta distribution 13-20 beta function 13-17 beta(m. 1-27 equating the numerators procedure in partial fraction expansion 12-13 Erlang distribution 13-16 error function 10-4 error in method of least squares 8-3 Euclidean norm C-1 Euler’s identities 3-14 even functions 6-7.q) MATLAB function 1-6. 3-11 complex poles 12-5 complex roots of characteristic equation 5-9 cond(A) MATLAB function C-2 condition number of a matrix .k) MATLAB function 14-24 diagonal elements of a matrix . 1-2 eigenvalues 5-30 eigenvectors 5-39 element-by-element division in MATLAB 1-21 element-by-element exponentiation in MATLAB 1-21 element-by-element multiplication in MATLAB 1-19.see matrix diff(s) MATLAB function 2-6 difference equations A-1 difference operator 7-4. 6-31 even symmetry 6-7 exit MATLAB command 2.see matrix algebraic constrain block in Simulink B-18 algebraic form of the Legendre polynomials 15-17 alternate form of the trigonometric Fourier series 6-25 alternating current 3-1 angle(z) MATLAB function 1-24 angular velocity 3-2 antidifference 7-12 approximations with Excel 2-7 associated Legendre differential equation 15-18 associated Legendre functions of the first and second kind 15-18 AutoFill Excel feature 2-9.Index Symbols % (percent) symbol 1-2 %d MATLAB symbol 2-5 %e MATLAB symbol 2-21 %s MATLAB symbol 2-21 %u MATLAB symbol 2-21 A abs(z) MATLAB function 1-24 absolute cell in Excel 2-19 absolutely convergent 15-10 Adams’ method 9-13 Add Trendline Excel feature 8-6 adjoint of a matrix . 14-24 eye(size(A)) MATLAB function 4-7 F factor(p) MATLAB function 12-3 factorial polynomials 7-6 Fibonacci numbers 11-7 figure window in MATLAB 1-14 finite differences 7-1 first divided difference 7-1 first harmonic 6-1 fixed point format 2-21 Flip Block command in Simulink B-11 forced response 5-8 forcing function A-1 format MATLAB command 1-31 format specifiers in MATLAB 2-21 IN1 . 8-6 Chebyshev polynomials 15-22 Chebyshev polynomials of the first kind 15-22 Chebyshev polynomials of the second kind 15-22 Cholesky factorization 14-23 classification of differential equations 5-2 clc MATLAB command 1-2 clear MATLAB command 1-2 coefficient matrix . 14-10 Bessel function of negative order n 15-2 Bessel function of n 15-1 Bessel functions 15-1 Bessel functions of the first kind 15-7 Bessel functions of the second kind 15-7 besselj(n. 1-27 Editor/Debugger in MATLAB 1-1.see matrix cofactors .n) MATLAB function 13-19 BETADIST Excel function 13-21 bisection method for root approximation 2-19 box MATLAB command 1-12 C Casorati’s determinant 11-2 Cayley-Hamilton theorem 5-30 characteristic (auxiliary) equation 5-8 characteristic equation of a second order difference equation 11-3 chart type in Excel 2-10 Chart Wizard Excel feature 2-10.see matrix diag(v.see finite differences dipole 15-12 direct terms in MATLAB 12-1 directed network 16-15 disp(x) MATLAB function 2-5 Display block in Simulink B-18 display formats in MATLAB 1-31 distinct poles 12-2 distinct roots of characteristic equation 5-9 divided differences 7-1 division of complex numbers 3-13 dot division operator in MATLAB 1-21 dot exponentation operator in MATLAB 1-21 dot multiplication operator in MATLAB 1-21 dsolve(s) MATLAB function 9-5 dynamic programming 16-4 E Editor window in MATLAB 1-1. 2-15 autoscale icon in Simulink B-12 axis MATLAB command 1-17. 12-9 exponential form of complex numbers 3-14 exponential form of the Fourier series 6-29 eye(n) MATLAB function 4-6.see matrix conj(A) MATLAB function 4-8 conj(x) MATLAB function 4-8 conjugate of a complex number 3-12 connected network 16-15 constraints in linear programming 16-1 Contents Pane in Simulink B-7 conv(a. 33 EXP(GAMMALN(n)) Excel function 13-5 expand(s) MATLAB function 7-13.b) MATLAB function 1-6 Cramer’s rule 4-18.

z) MATLAB command 1-17 mesh(Z) MATLAB function 7-32 meshgrid(x.see matrix ill-conditioned matrix .see matrix Math Operations Library in Simulink B-11 MATLAB Demos 1-2 matrix (matrices) 4-1 adjoint of 4-22 coefficient C-3 cofactors of 4-11 condition number of C-2 conformable for addition 4-2 conformable for multiplication 4-4 conformable for subtraction 4-2 congugate of 4-8 determinant of 4-9 diagonal elements of 4-1 diagonal of 4-6 elements of 4-1 Hermitian 4-9 Hilbert C-3 identity 4-6 ill-conditioned 4-22. 14-9 Genenbauer 15-22 general solution of a differential equation 5-6 generalized factorial function 13-1 generating function for Bessel functions of the first kind of integer order 15-9 generating function for Legendre polynomials 15-12 Goal Seek Excel feature 2-16 Gram-Schmidt orthogonalization procedure 5-39. A-1 nearest neighbor interpolarion 7-25 network analysis 16-14 Neumann functions 15-7 .see matrix matrix left division in MATLAB 4-26 matrix multiplication in MATLAB 1-19.xi. 11-2.b) MATLAB function 10-5 interp1(x.forward substitution 14-10 Fourier analysis 3-2 Fourier series 6-1 fplot MATLAB command 1-27 fprintf MATLAB command 7-29 frequency 3-2 Frequency Modulation (FM) 15-4 frequency response 1-11 full-wave rectifier with even symmetry 6-24 function block parameters in Simulink B-10 function files in MATLAB 1-26. 6-31 help in MATLAB 1-2. 6-22 half-wave symmetry 6-7.y.see matrix lsqnonneg MATLAB function 14-28 LU factorization method 14-9 lu(A) MATLAB function 14-12.see matrix imag(z) MATLAB function 1-24 imaginary axis 3-10 imaginary number 3-10 IN2 improper integral 13-1 improper rational function 12-1 increments between points in MATLAB 1-14 in-phase sinusoids 3-3 input(‘string’) MATLAB command 2-2 int(f. 8-3 least squares line 8-2 least-squares curve 8-2 least-squares parabola 8-2.xi) MATLAB function 7-24 interp1(x. 1-27 Milne’s method 9-15 minimum span problems 16-15 minor of determinant . 1-29 fzero(‘function’.xi. 1-15 Hermitian matrix .x0) MATLAB function 12-18 fzero(f. 13-10 GAMMADIST Excel function 13-16 GAMMALN Excel function 13-5 Gaussian elimination method 4-20.x) MATLAB function 15-19 Legendre’s equation 15-10 lims = MATLAB function 1-27 line graph 16-4 line spectrum 6-33 linear factor 1-9 linear interpolarion 7-25 linear programming 16-1 linear regression 8-2 linspace MATLAB command 1-14 ln (natural logarithm) 1-12 log (common logarithm) 1-12 log(x) MATLAB function 1-12 log10(x) MATLAB function 1-12 log2(x) MATLAB function 1-12 loglog MATLAB command 1-12.z. 15-25 H half-wave rectifier 6-19. 1-27.y.see matrix hilb(n) MATLAB function C-4 Hilbert matrix . 8-7 Legendre functions 15-10 Legendre functions of the second kind 15-11 Legendre polynomials 15-11 Legendre polynomials in trigonometric form 15-18 legendre(n.’method’) MATLAB function 7-25 interp2(x. 4-29 modified Bessel functions 15-7 modulation index 15-4 multiple poles 12-6 multiplication of complex numbers 3-12 N NaN in MATLAB 1-27 natural response 5-8.a. A-1 method of variation of parameters 5-20 m-file in MATLAB 1-1 1-26.x) MATLAB function 2-7 G Gain block in Simulink B-18 gamma distribution 13-16 gamma function 13-1 gamma(n) MATLAB function 13-3.z. 8-3 method of Picard 15-1 method of undetermined coefficients 5-10. C-2 inverse of 4-23 lower triangular 4-6 main diagonal elements of 4-1 main diagonal of 4-1 minor of determinant 4-13 non-singular 4-22 scalar 4-6 singular 4-22 size of 4-7 skew-Hermitian 4-9 skew-symmetric 4-8 square 4-1 symmetric 4-8 trace of 4-2 transpose of 4-7 unitary 14-25 upper triangular 4-5 well-conditioned C-2 zero matrix .y. 2-4 fundamental frequency 6-1 fundamental theorem of integral calculus 7-12 fundamental theorem of sum calculus 7-13 fzero MATLAB function 1-27.yi) MATLAB function 7-25 interp2(x.yi.see matrix isoprofit line 16-2 J j operator 3-11 L L’ Hôpital’s rule 13-2 lagging sinudoid 3-3 Lagrange’s interpolation method 7-17 Laguerre polynomials 15-21 leading sinusoid 3-3 least squares 8-2.y) MATLAB command 1-17 method of Frobenius 15-1 method of least squares 8-2. 1-20 mesh(x.see matrix homogeneous difference equation 11-2 Hz (Hertz) 3-3 I identity matrix . 7-6 main diagonal elements .’method’) MATLAB function 7-25 interpolation 7-1 interpolation with MATLAB 7-24 interval halving 2-19 inverse of a matrix .y. 14-7 Gregory-Newton Backward Interpolation Formula 7-21 Gregory-Newton Forward Interpolation Formula 7-19 grid MATLAB command 1-11 gtext MATLAB command 1-14. 14-14 M Maclaurin power series 6-41.xi.see matrix MINVERSE Excel function 4-28 MMULT Excel function 4-28. 34 lower triangular matrix .y.see matrix main diagonal of a matrix .

15-18 roots of polynomials 1-3 roots(p) MATLAB function 1-3. 10-5 Tree Pane in Simulink B-7 Trendline Excel feature 8-9 triangular waveform 6-11. 6-19 trigonometric Fourier series 6-1 trigonometric relations 3-5 two-dimensional plots 7-32 type of a diferential equation 5-2 U ultraspherical functions 15-22 undetermined system 8-3 unit fraction C-3 unitary matrix . 1-8 rotating vector 3-5 round(n) MATLAB function 1-24 row vector 1-3.p. 8-11 power series 6-40 proper rational function 12-1 Q QR factorization 14-25 qr(A) MATLAB function 14-25 quad MATLAB function 10-10 quad8 MATLAB function 10-10 quadratic curve 8-1 quadratic factor 1-9 quit MATLAB command 1-2 R radian frequency 3-2 rational polynomial 1-8 rationalization of the quotient 3-13 real axis 3-10 real number 3-11 real(z) MATLAB function 1-24 recursion A-1 recursive method A-1 regression 8-1 regression analysis 8-7 relative cell in Excel 2-19 repeated poles 5-9.x) MATLAB function 1-5. 14-4 orthonormal basis 14-5 out-of-phase sinusoids 3-3 overdetermined system 8-3 P parabolic curve 8-1 partial differential equation (PDE) 5-3 partial fraction expansion 12-1 PDE (Partial Differential Equation) 5-3 Pearson correlation coefficient 8-10 period 3-2. 6-14.see matrix Singular Value Decomposition 14-28 Sinks library B-18 sinusoids 3-2 size of a matrix .r) MATLAB function 1-24 poles 12-2 poly MATLAB function 1-4 polyder MATLAB function 1-7 polyfit(x.y. 6-31 T Taylor series 5-24.see matrix trapezoidal rule 10-1 trapz(x.see matrix symmetry 6-7. 1-15 plot3 MATLAB command 1-16 polar form of complex numbers 3-15 polar plot in MATLAB 1-24 polar(theta. 12.see matrix upper triangular matrix .k) MATLAB function 12-1 revolutions per second 3-5 Rodrigues’ formula 15-12. 6-18 scalar matrix . 3-3 periodic waveform 3-2 phasor 3-2 plot area in Excel 6 plot MATLAB command 1-9.see matrix square waveform 6-9. 6-41.see matrix skew-symmetric matrix .see matrix skew-Hermitian matrix . 1-19 Runge-Kutta method 5-24. 9-5 S sawtooth waveform 6-10. 6-44 Taylor series expansion method 9-1 text MATLAB command 1-14.see matrix norm C-1 norm(A) MATLAB function C-1 numeric expressions in MATLAB 12-4 numerical evaluation of Fourier coefficients 6-36 O odd functions 6-8.see matrix solution of the homogeneous ODE 5-8 solutions of ODEs 5-6 space equations 5-24 spectrum analyzer 33 spherical harmonics 15-18 sprintf MATLAB command 2-5 sqrt MATLAB function 10-12 square matrix .Newton’s divided difference interpolation method 7-15 Newton-Cotes 8-panel rule 10-10 non-homogeneous difference equation 11-2 non-homogeneous ODE 5-6 non-singular matrix . 14-1. 6-14. 6-49 start simulation in Simulink B-12 state equations 5-24 state transition matrix 5-28 state variables 5-24 State-Space block in Simulink B-12 Stirling’s asymptotic series for the G(n) function 13-9 string in MATLAB 1-17 subplot MATLAB command 1-18 surface zonal harmonics 15-11 svd(A) MATLAB function 14-28 sym. 6-16. 6-31 odd symmetry 6-7 ODE (Ordinary Differential Equation) 5-3 ode23 MATLAB function 9-9 ode45 MATLAB function 9-9 one-dimensional wave equation 5-3 optimum path policy 16-5 order of a differential equation 5-3 ordinary differential equation 5-3 oriented network 16-15 orthogonal basis 14-5 orthogonal functions 6-2.6 residue(r. 1-17 third harmonic 6-1 title(‘string’) in MATLAB 1-12 trace of a matrix .y) MATLAB function 10-3. 14-2 orthogonal system 15-9 orthogonal trajectories 14-2 orthogonal unit vectors 14-5 orthogonal vectors 5-39.see matrix transpose of a matrix .see matrix zeros 12-2 zlabel MATLAB command 1-17 IN3 .see matrix V VLOOKUP Excel function 7-23 W Wallis’s formulas 13-16 Weber functions 15-7 well-conditioned matrix .see matrix Scope block in Simulink B-12 script file in MATLAB 1-26 second divided difference 7-1 second harmonic 6-1 semicolons in MATLAB 1-7 semilogx MATLAB command 1-12 semilogy MATLAB command 1-12 simple differential equations 5-1 simplex method 16-4 Simpson’s rule 10-6 Simulation drop menu in Simulink B-12 simulation start icon in Simulink B-12 Simulink icon B-7 Simulink Library Browser B-8 singular matrix .n) MATLAB function 8-11 polynomial construction from known roots in MATLAB 1-4 polyval(p. 1-12.see matrix while end in MATLAB 2-4 Wronskian determinant 5-10. 11-2 X xlabel MATLAB command 1-12 XY (Scatter) in Excel 8-6 Y ylabel MATLAB command 1-12 Z zero matrix . 6-48. syms MATLAB symbolic expressions 12-4 symbolic expressions in MATLAB 12-3 Symbolic Math Toolbox in MATLAB 12-4 symmetric matrix .