307 23 12MB
English Pages 182 [197] Year 2020
T ensor A nalysis for E ngineers Second Edition
Tensor-Analysis-for-Engineers_2E_CH00_FM.indd 1
10/3/2020 10:17:02 AM
LICENSE, DISCLAIMER OF LIABILITY, AND LIMITED WARRANTY By purchasing or using this book (the “Work”), you agree that this license grants permission to use the contents contained herein, but does not give you the right of ownership to any of the textual content in the book or ownership to any of the information or products contained in it. This license does not permit uploading of the Work onto the Internet or on a network (of any kind) without the written consent of the Publisher. Duplication or dissemination of any text, code, simulations, images, etc. contained herein is limited to and subject to licensing terms for the respective products, and permission must be obtained from the Publisher or the owner of the content, etc., in order to reproduce or network any portion of the textual material (in any media) that is contained in the Work. Mercury Learning and Information (“MLI” or “the Publisher”) and anyone involved in the creation, writing, or production of the companion disc, accompanying algorithms, code, or computer programs (“the software”), and any accompanying Web site or software of the Work, cannot and do not warrant the performance or results that might be obtained by using the contents of the Work. The author, developers, and the Publisher have used their best efforts to insure the accuracy and functionality of the textual material and/or programs contained in this package; we, however, make no warranty of any kind, express or implied, regarding the performance of these contents or programs. The Work is sold “as is” without warranty (except for defective materials used in manufacturing the book or due to faulty workmanship). The author, developers, and the publisher of any accompanying content, and anyone involved in the composition, production, and manufacturing of this work will not be liable for damages of any kind arising out of the use of (or the inability to use) the algorithms, source code, computer programs, or textual material contained in this publication. This includes, but is not limited to, loss of revenue or profit, or other incidental, physical, or consequential damages arising out of the use of this Work. The sole remedy in the event of a claim of any kind is expressly limited to replacement of the book, and only at the discretion of the Publisher. The use of “implied warranty” and certain “exclusions” vary from state to state, and might not apply to the purchaser of this product.
Tensor-Analysis-for-Engineers_2E_CH00_FM.indd 2
10/3/2020 10:17:02 AM
T ensor A nalysis for E ngineers Transformations-Mathematics-Applications Second Edition
Mehrzad Tabatabaian, PhD, PEng
MERCURY LEARNING AND INFORMATION Dulles, Virginia Boston, Massachusetts New Delhi
Tensor-Analysis-for-Engineers_2E_CH00_FM.indd 3
10/3/2020 10:17:04 AM
Copyright © 2021 by Mercury Learning and Information LLC. All rights reserved. This publication, portions of it, or any accompanying software may not be reproduced in any way, stored in a retrieval system of any type, or transmitted by any means, media, electronic display or mechanical display, including, but not limited to, photocopy, recording, Internet postings, or scanning, without prior permission in writing from the publisher. Publisher: David Pallai Mercury Learning and Information 22841 Quicksilver Drive Dulles, VA 20166 [email protected] www.merclearning.com (800) 232-0223 M. Tabatabaian. Tensor Analysis for Engineers: Transformations-Mathematics-Applications, Second Edition. ISBN: 978-1-68392-601-6 The publisher recognizes and respects all marks used by companies, manufacturers, and developers as a means to distinguish their products. All brand names and product names mentioned in this book are trademarks or service marks of their respective companies. Any omission or misuse (of any kind) of service marks or trademarks, etc. is not an attempt to infringe on the property of others. Library of Congress Control Number: 2020946631 202122321 This book is printed on acid-free paper in the United States of America. Our titles are available for adoption, license, or bulk purchase by institutions, corporations, etc. For additional information, please contact the Customer Service Dept. at 800-232-0223(toll free). All of our titles are available in digital format at www.academiccourseware.com and other digital vendors. The sole obligation of Mercury Learning and Information to the purchaser is to replace the book, based on defective materials or faulty workmanship, but not based on the operation or functionality of the product.
Tensor-Analysis-for-Engineers_2E_CH00_FM.indd 4
10/3/2020 10:17:04 AM
To my teachers and mentors for their invaluable transfer of knowledge and direction.
Tensor-Analysis-for-Engineers_2E_CH00_FM.indd 5
10/3/2020 10:17:04 AM
Tensor-Analysis-for-Engineers_2E_CH00_FM.indd 6
10/3/2020 10:17:04 AM
CONTENTS Prefacexi About the Author xiii Chapter 1:
Introduction 1.1 Index Notation—The Einstein Summation Convention
1 2
Chapter 2:
Coordinate Systems Definition
3
Chapter 3:
Basis Vectors and Scale Factors
5
Chapter 4:
Contravariant Components and Transformations
9
Chapter 5:
Covariant Components and Transformations
13
Chapter 6:
Physical Components and Transformations
15
Chapter 7:
Tensors—Mixed and Metric
17
Chapter 8:
Metric Tensor Operation on Tensor Indices 8.1 Example: Cylindrical Coordinate Systems 8.2 Example: Spherical Coordinate Systems
21 23 25
Chapter 9:
Dot and Cross Products of Tensors 9.1 Determinant of an N × N Matrix Using Permutation Symbols
29
Gradient Vector Operator—Christoffel Symbols 10.1 Covariant Derivatives of Vectors—Christoffel Symbols of the 2nd Kind 10.2 Contravariant Derivatives of Vectors 10.3 Covariant Derivatives of a Mixed Tensor 10.4 Christoffel Symbol Relations and Properties—1st and 2nd Kinds
35
Chapter 10:
Tensor-Analysis-for-Engineers_2E_CH00_FM.indd 7
34
35 39 40 41
10/3/2020 10:17:04 AM
viii • Contents Chapter 11:
Derivative Forms—Curl, Divergence, Laplacian 11.1 Curl Operations on Tensors 11.2 Physical Components of the Curl of Tensors—3D Orthogonal Systems 11.3 Divergence Operation on Tensors 11.4 Laplacian Operations on Tensors 11.5 Biharmonic Operations on Tensors 11.6 Physical Components of the Laplacian of a Vector—3D Orthogonal Systems
51 51 54 55 57 58 59
Chapter 12:
Cartesian Tensor Transformation—Rotations 65 12.1 Rotation Matrix 67 12.2 Equivalent Single Rotation: Eigenvalues and Eigenvectors67
Chapter 13:
Coordinate Independent Governing Equations 75 13.1 The Acceleration Vector—Contravariant Components 76 13.2 The Acceleration Vector—Physical Components 78 13.3 The Acceleration Vector in Orthogonal Systems—Physical Components 79 13.4 Substantial Time Derivatives of Tensors 82 13.5 Conservation Equations—Coordinate Independent Forms 85
Chapter 14: Collection of Relations for Selected Coordinate Systems89 14.1 Cartesian Coordinate System 89 14.2 Cylindrical Coordinate Systems 91 14.3 Spherical Coordinate Systems 93 14.4 Parabolic Coordinate Systems 96 14.5 Orthogonal Curvilinear Coordinate Systems 98 Chapter 15: Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix 15.1 Active and Passive Rotations 15.2 Euler Angles
Tensor-Analysis-for-Engineers_2E_CH00_FM.indd 8
103 104 107
10/3/2020 10:17:04 AM
Contents • ix
Chapter 16:
15.3 Categorizing Euler Angles
113
15.4 Gimbal Lock-Euler Angles Limitation
121
15.5 Quaternions-Applications for Rigid Body Rotation
123
15.6 From a Given Quaternion to Rotation Matrix
135
15.7 From a Given Rotation Matrix to Quaternion
138
15.8 From Euler Angles to a Quaternion
139
15.9 Putting It All Together
140
Worked-out Examples
143
16.1 Example: Einstein Summation Conventions
143
16.2 Example: Conversion from Vector to Index Notations144 16.3 Example: Oblique Rectilinear Coordinate Systems
145
16.4 Example: Quantities Related to Parabolic Coordinate System
149
16.5 Example: Quantities Related to Bi-Polar Coordinate Systems
152
16.6 Example: Application of Contravariant Metric Tensors155 16.7 Example: Dot and Cross Products in Cylindrical and Spherical Coordinates
155
16.8 Example: Relation between Jacobian and Metric Tensor Determinants
157
16.9 Example: Determinant of Metric Tensors Using Displacement Vectors
158
16.10 Example: Determinant of a 4 × 4 Matrix Using Permutation Symbols
159
16.11 Example: Time Derivatives of the Jacobian
159
16.12 Example: Covariant Derivatives of a Constant Vector
160
16.13 Example: Covariant Derivatives of Physical Components of a Vector
161
16.14 Example: Continuity Equations in Several Coordinate Systems
161
16.15 Example: 4D Spherical Coordinate Systems
162
Tensor-Analysis-for-Engineers_2E_CH00_FM.indd 9
10/3/2020 10:17:04 AM
x • Contents 16.16 Example: Complex Double Dot-Cross Product Expressions165 16.17 Example: Covariant Derivatives of Metric Tensors 166 16.18 Example: Active Rotation Using Single-Axis and Quaternions Methods 166 16.19 Example: Passive Rotation Using Single-Axis and Quaternions Methods 168 16.20 Example: Successive Rotations Using Quaternions Method 170 Chapter 17:
Exercise Problems
References Index
Tensor-Analysis-for-Engineers_2E_CH00_FM.indd 10
171 177 179
10/3/2020 10:17:04 AM
PREFACE This is the second edition of the published textbook: Tensor Analysis for Engineers. In this edition, we expand the content on the rigid body rotation and Cartesian tensors by including Euler angles and quaternions methods. In addition to the rotation matrix method, presented in the first edition and included in this edition, we collect all three methods in this volume of the textbook. In this edition, the quaternions and their algebraic calculation rules are presented. We also discuss the active and passive rotations and present several worked-out examples using the Euler angles and quaternions methods applications and their interrelations. The problem of gimbal lock is also analyzed and presented with detailed worked out examples. Additional references have been included in the second edition. In engineering and science, physical quantities are often represented by mathematical functions, namely tensors. Examples include temperature, pressure, force, mechanical stress, electric/magnetic fields, velocity, enthalpy, entropy, etc. In turn, tensors are categorized based on their rank, i.e. rank zero, one, and so forth. The so-called scalar quantities (e.g., temperature) are tensors of rank zero. Likewise, velocity and force are tensors of rank one and mechanical stress and gradient of velocity are tensors of rank two. In Euclidean space, which could be of dimension N = 3, 4, ..., we can define several coordinate systems for our calculation and measurement of physical quantities. For example, in a 3D space, we can have Cartesian, cylindrical, and spherical coordinate systems. In general, we prefer defining a coordinate system whose coordinate surfaces (where one of the coordinate variables is invariant or remains constant) match to the physical problem geometry at hand. This enables us to easily define the boundary conditions of the physical problem to the related governing equations, written in terms of the selected coordinate system. This action requires transformation of the tensor quantities and their related derivatives (e.g., gradient,
Tensor-Analysis-for-Engineers_2E_CH00_FM.indd 11
10/3/2020 10:17:04 AM
xii • Preface curl, divergence) from Cartesian to the selected coordinate system or vice versa. The topic of tensor analysis (also referred to as “tensor calculus” or “Ricci’s calculus” since originally developed by Ricci, 1835–1925, [1], [2]), is mainly engaged with the definition of tensor-like quantities and their transformation among coordinate systems and others. The topic provides a set of mathematical tools which enables users to perform transformation and calculations of tensors for any well-defined coordinate systems in a systematic way—it is a “machine.” The merit of tensor analysis is to provide a systematic mathematical formulation to derive the general form of the governing equations for arbitrary coordinate systems. In this book, we aim to provide engineers and applied scientists the tools and techniques of tensor analysis for applications in practical problem solving and analysis activities. The geometry is limited to the Euclidean space/ geometry, where the Pythagorean Theorem applies, with well-defined Cartesian coordinate systems as the reference. We discuss quantities defined in curvilinear coordinate systems, like cylindrical, spherical, parabolic, etc., and present several examples and coordinates sketches with related calculations. In addition, we listed several worked-out examples for helping the readers with mastering the topics provided in the prior sections. A list of exercises is provided for further practice for readers.
Tensor-Analysis-for-Engineers_2E_CH00_FM.indd 12
Mehrzad Tabatabaian, PhD, PEng Vancouver, BC September 30, 2020
10/3/2020 10:17:04 AM
ABOUT THE AUTHOR Dr. Mehrzad Tabatabaian is a faculty member at the Mechanical Engineering Department, School of Energy at BCIT. He has several years of teaching and industry experiences. In addition to teaching courses in mechanical engineering, he performs research on renewable energy systems and modelling. Dr. Tabatabaian is currently Chair of the BCIT School of Energy Research Committee. He has published several papers in scientific journals and conferences, and he has written textbooks on multiphysics and turbulent flow modelling, advanced thermodynamics, tensor analysis, and direct energy conversion. He holds several registered patents in the energy field. Recently, Dr. Tabatabaian volunteered to help establish the Energy Efficiency and Renewable Energy Division (EERED), a new division at Engineers and Geoscientists British Columbia (EGBC). Mehrzad Tabatabaian received his BEng from Sharif University of Technology and advanced degrees from McGill University (MEng and PhD). He has been an active academic, professor, and engineer in leading alternative energy, oil, and gas industries. Mehrzad has also a Leadership Certificate from the University of Alberta and holds an EGBC P.Eng. license.
Tensor-Analysis-for-Engineers_2E_CH00_FM.indd 13
10/3/2020 10:17:04 AM
Tensor-Analysis-for-Engineers_2E_CH00_FM.indd 14
10/3/2020 10:17:04 AM
CHAPTER
1
INTRODUCTION Physical quantities can be represented mathematically by t ensors. In further sections of this book we will define tensors more rigorously; however, for the introduction we will use this definition. An example of a tensor-like quantity is the temperature in a room (which could be a function of space and time) expressed as a scalar, a tensor of rank zero. Wind velocity is another example, which can be defined when we know both its magnitude and speed—a scalar quantity—and its direction. We define velocity as a vector or a t ensor of rank one. Scalars and vectors are familiar quantities to us and we encounter them in our daily life. However, there are quantities, or tensors of rank two, three, or higher that are normally dealt with in technical engineering computations. Examples include mechanical stress in a continuum, like the wall of a pressure vessel—a tensor of rank two—the modulus of elasticity or viscosity in a fluid—tensors of rank four—and so on. Engineers and scientists calculate and analyze tensor quantities, including their derivatives, using the laws of physics, mostly in the form of governing equations related to the phenomena. These laws must be expressed in an objective form as governing equations, and not subjected to the coordinate system considered. For example, the amount of internal stress in a continuum should not depend on what coordinate system is used for calculations. Sometimes tensors and their involved derivatives in a study must be transformed from one coordinate system to another. Therefore, to satisfy these technical/engineering needs, a mathematical “machine” is required to perform these operations accurately and systematically between arbitrary coordinate systems. Furthermore, the communication of technical computations requires precise definitions for tensors to guarantee a reliable level of standardization, i.e., identifying true tensor quantities from apparently
Tensor-Analysis-for-Engineers_2E_CH01.indd 1
10/3/2020 10:18:04 AM
2 • Tensor Analysis for Engineers, 2E tensor-like or non-tensor ones. This machine is called tensor analysis [2], [3], [4], [5], [6], [7]. The subject of tensor analysis has two major parts: a) definitions and properties of tensors including their calculus, and b) rules of transformation of tensor quantities among different coordinate systems. For example, consider again the wind velocity vector. By using tensor analysis, we can show that this quantity is a true tensor and transform it from a Cartesian coordinate system to a spherical one, for example. A major outcome of tensor analysis is having general relations for gradient-like operations in arbitrary coordinate systems, including gradient, curl, divergence, Laplacian, etc. which appear in many governing equations in engineering and science. Using tensor notation and definitions, we can write theses governing equations in explicit coordinate- independent forms.
1.1 INDEX NOTATION—THE EINSTEIN SUMMATION CONVENTION Writing expressions containing tensors could become cumbersome, especially when higher ranked tensors and higher dimensional space are involved. For example, we usually use hatted arrow symbol for vectors (like A) and boldfont symbols for second rank tensors (like A). But this approach is very limited for expressing, for example, the modulus of elasticity, a 4th ranked tensor. Another limitation shows up when writing the components of a tensor in N -dimension space. For example, in 3D we write vector A in a Cartesian coor 3 dinate system as A = ∑Ai Ei = A1 E1 + A2 E2 + A3 E3 , where Ai is the compo 1 nent and Ei the unit vector. Taking this approach for N -dimension space and carrying the summation symbol (i.e., Σ) is cumbersome and seems unnecessary. The Einstein summation convention allows us to dispose of the summation symbol, if we carry summation operation for repeated indices in product-type expressions (unless otherwise specified). Using this approach, we can write A = Ai Ei and a tensor of rank 4, for example, as T = Tijkl Ei E j Ek El , with all indices range for 1 to N. In practice, however, we even ignore writing the unit vectors and just use the component representing the original t ensor; hence this method is also referred to as component or index notation, or umber A ≡ Ai and T ≡ Tijkl . In addition, we represent a tensor’s rank by the n of free (i.e., not repeated) indices. We use these definitions and conventions throughout this book.
Tensor-Analysis-for-Engineers_2E_CH01.indd 2
10/3/2020 10:18:08 AM
CHAPTER
2
COORDINATE SYSTEMS DEFINITION For measuring and calculating physical quantities associated with geometrical points in space we require coordinate systems for reference. These systems (for example, in a 3D space) are composed of surfaces that mutually intersect to specify a geometrical point. For reference, an ideal system of coordinates called a Cartesian system is defined, in Euclidean space, such that it composes of three flat planes. These planes are considered by default to be mutually perpendicular to each other to form an orthogonal Cartesian system. In general, the orthogonality is not required to form a coordinate system—this kind of coordinate system is an oblique or slanted system. A Cartesian coordinate system, although ideal, is central to engineering and scientific calculations, since it is used as the reference compared to other coordinate systems. It also has properties, as will be shown in further sections, that enable us to calculate the values of tensors [3]. For practical purposes, sometimes it is more convenient to consider curved planes instead of flat ones in a coordinate system. For example, for cases when a cylindrical or a spherical water or an oil tank is involved, we prefer that all or some of coordinate surfaces match the geometry of the tank. For a cylindrical coordinate system, we consider the Cartesian system again but replace one of its planes with a cylinder, whereas for a spherical system we replace two flat planes with a sphere and a cone.
Tensor-Analysis-for-Engineers_2E_CH02.indd 3
10/3/2020 10:19:06 AM
4 • Tensor Analysis for Engineers, 2E
FIGURE 2.1 Sketches of Cartesian, Cylindrical, and Spherical coordinate systems.
Many other coordinate systems are defined/used in practice such as parabolic, bi-spherical, etc. Figure 2.1 shows some examples of common coordinate systems. Animations of the coordinates shown in Figure 2.1 are provided in the accompanying files. For organizing the future usage of symbols for arbitrary coordinate systems we define coordinate variables with superscripted letters, for reasons that will be explained in further sections. For Cartesian systems we use yi and for other arbitrary systems we use x i —note that i is merely an index, not a raised power. For example, in a 3D space we have y1 ≡ X , y2 ≡ Y, and y3 ≡ Z where ( X , Y, Z ) are common notations for Cartesian coordinates. In general, for N dimensional systems we have
yi ≡ ( y1 , y2 , y3 ,, yN ) Cartesian coordinates i 2.1 1 2 3 N x ≡ ( x , x , x ,, x ) arbitrary coordinates In this book, we limit the geometry to that known as Euclidean geometry [2], [4], [5]. Therefore, curved space geometry (i.e., Riemann geometry), space-time, and discussions of General Relativity are not included. However, curvilinear coordinate systems and oblique non-orthogonal systems are covered, where applicable, and defined with reference to Euclidean geometry/space.
Tensor-Analysis-for-Engineers_2E_CH02.indd 4
10/3/2020 10:19:10 AM
CHAPTER
3
BASIS VECTORS AND SCALE FACTORS For the measurement of quantities we need to define metrics or scales in whatever coordinate system we use. These scales are, usually, vectors defined at a point (such as the origin) and are tangent to the coordinate surfaces at that point. These vectors are called basis vectors. For Cartesian system yi we define basis vector Ei (see Figure 3.1). Note that subscript i (i.e., i = 1,2,, N) is merely an index and the reason why we used it as a subscript for basis vectors willbe explained in a further section. Now, considering an incremental vector ds , from point P to neighboring point P′ , we define the direc tion of the basis vector as moving from P to P′ or Ei = yi P → yi + dyi P′ . For example, E1 = y1 P → y1 + dy1 P′. Now we define the magnitude of Ei such that the magnitude of distance from P to P′ is given as ds ( i ) = dyi Ei 3.1
Also, we can get the distance = PP′ ds = ( i ) dyi in a Cartesian coordinate system. Therefore, the basis vectors magnitude is unity, or Ei = 1 . In other words, the basis vectors in Cartesian system have unit lengths and are well-known unit vectors (usually represented by i , j, k ). This is more than a trivial result and as shown in further sections, enables us to calculate/ measure quantities in other coordinate systems with reference to the Cartesian system. The directed distance vector ds is given as N i = ds ∑ = dy Ei dyi Ei 3.2
Tensor-Analysis-for-Engineers_2E_CH03.indd 5
i =1
10/3/2020 10:19:31 AM
6 • Tensor Analysis for Engineers, 2E
The magnitude of this vector is the square root of the dot-product of ds with itself, ds = ds ⋅ ds = dyi dy j Ei ⋅ E j 3.3
FIGURE 3.1 A Cartesian system with unit vectors and an incremental vector ds.
Assuming the Pythagorean theorem holds in Euclidean space we have = ds
= dyi dyi
∑ ( dy )
i 2
3.4
i
From Equations 3.3 and 3.4, we can conclude that 1 for i = j 3.5 Ei ⋅ E j = 0 for i ≠ j
Which also shows that the basis/unit vectors in Cartesian system yi are mutually perpendicular or the system is orthogonal, as defined.
Now we assume an arbitrary system x i, which may be neither rectilinear nor orthogonal (see Figure 3.2). In this system the distance between two points—say P and P′—is the same when considering the vector ds . In other words, the vector components in system x i compared to those in system yi could change along with the basis vector corresponding to system x i such that the vector itself remains the same, or as an invariant. This requirement is simply a statement of independence of physical quantities (and the laws of nature) regardless of the coordinate system we consider for calculation and analysis. Therefore, we have i = ds dx = ei dyi Ei 3.6
Tensor-Analysis-for-Engineers_2E_CH03.indd 6
10/3/2020 10:19:33 AM
Basis Vectors and Scale Factors • 7
where ei is the basis vector corresponding to system x i. In general the basis vector ei can vary, both in magnitude and/or direction, from point to point in space. Also, the dx i may be dimensionless, like the angle coordinate in a polar coordinate system.
FIGURE 3.2 Sketches of a curvilinear coordinate system and basis vectors1.
The magnitude of basis vector ei is the scale factor hi, or
hi = ei 3.7
Note that hi is unity for a Cartesian coordinate system and may be d ifferent from unity in general curvilinear systems. We will derive formulae for the calculation of hi in an arbitrary coordinate system in further sections. The unit vector e ( i ) can be defined as ei e ( i) = ( no summation on i ) 3.8 hi Obviously, in a Cartesian system the unit vectors and the basis vectors are identical, since hi is unity.
Common copyright (cc), https://commons.m.wikimedia.org/wiki/File:Vector_1-form. svg#mw-jump-to-license 1
Tensor-Analysis-for-Engineers_2E_CH03.indd 7
10/3/2020 10:19:34 AM
Tensor-Analysis-for-Engineers_2E_CH03.indd 8
10/3/2020 10:19:34 AM
CHAPTER
4
CONTRAVARIANT COMPONENTS AND TRANSFORMATIONS For transformations between systems x i and Cartesian yi , we must have the functional relationships between their coordinate variables. For example, x i = Fi ( y1 , y2 ,, yN )4.1
gives N number of functions transforming yi to x i system. Inversely [4], we can transform from x i to yi system using function Gi given as
yi = Gi ( x1 , x2 ,, x N ) 4.2
For example, for a 2D polar coordinate system, ( r ,θ ) ≡ ( x1 , x2 ) with reference to Cartesian system ( X , Y ) ≡ ( y1 , y2 ) we have 1 1 2 X = r cosθ y = x cos ( x ) ≡ 2 . Or inversely, 1 2 Y = r sin θ y = x sin ( x ) 2 2 1 r x ( y1 ) + ( y2 ) = X2 + Y2 = ≡ . y2 −1 Y θ = tan x2 = tan −1 1 X y
Tensor-Analysis-for-Engineers_2E_CH04.indd 9
10/3/2020 10:19:53 AM
10 • Tensor Analysis for Engineers, 2E
Now by taking partial derivatives of Gi (see Equation 4.2) we can relate i the differentials dyi , in a Cartesian system, to dx in an arbitrary coordinate system as dyi =
∂yi j dx 4.3 ∂x j
1 For example, after expanding Equation 4.3 for i = 1 we get, dy = 1 dy =
∂y1 1 ∂y1 2 ∂y1 N ∂yi . The transformation coefficient can be dx + dx + + dx ∂x j ∂x1 ∂x2 ∂x N ∂y1 ∂y1 1 ∂x ∂x N ∂yi written in matrix form as j = . The determinant of the ∂x N ∂yN ∂y ∂x N ∂x1 transformation coefficient matrix is defined as the Jacobian of the transformation, or
=
∂ ( y1 , y2 ,, yN ) = ∂ ( x1 , x2 ,, x N )
∂y1 ∂x1 ∂yN ∂x1
∂y1 1 ∂y1 2 dx + 2 dx + ∂x1 ∂x
∂y1 ∂x N 4.4
∂yN ∂x N
The Jacobian can be interpreted as the density of the space. In other words, let’s say that we have = 5 for a given system x i . This means that we have packed 5 units of Cartesian space into a volume in x i space through transformation from Cartesian to the given system. The smaller the Jacobian, the smaller the space density would be and vice versa. Similarly, we can use function Fi (see Equation 4.1) to have dx i =
∂x i dy j4.5 ∂y j
It can be shown [4], [2], that the determinant of transformation coefficient ∂x i is the inverse of , or ∂y j
Tensor-Analysis-for-Engineers_2E_CH04.indd 10
10/3/2020 10:19:55 AM
Contravariant Components and Transformations • 11
∂ ( x1 , x2 ,, x N ) = = ∂ ( y1 , y2 ,, yN ) −1
∂x1 ∂y1 ∂x ∂y1
N
∂x1 ∂yN
4.6
∂x ∂yN N
Equations 4.3 and 4.5 show a pattern for the transformation of differentials dyi and dx i, respectively. That is, the partial derivatives of the c orresponding coordinates appear in the numerator of the transformation coefficient. Nevertheless, one can ask: does this pattern maintain for general system-tosystem transformation? The short answer is “yes” [7]. Hence, for arbitrary systems x i and x′i we have
= dx i
∂x i ∂x′i j j i ′ ′ = dx and dx dx 4.7 ∂x′ j ∂x j
Any quantity, say A i, that transforms according to Equation 4.7 is defined as a contravariant type, with the standard notation of having the index i as a superscript. Therefore, transformation A i A′i reads
= A′i
∂x′i j ∂x j i j = A and A A′ 4.8 ∂x j ∂x′i
Obviously, performing the related calculations requires the functional relations between the two systems, i.e., x i = func ( x′1 , x′2 ,, x′N ) or x′i = func ( x1 , x2 ,, x N ) . This in turn requires having the Cartesian system as a reference for calculating the values of A j or A′i, since x i and x′i are arbii trary. For example, transforming A directly from a spherical to a cylindrical system requires having the functional relations between these coordinates with reference to the Cartesian system as the main reference. The contravariant component of a vector has a geometrical meaning as well. To show this, we consider a rectilinear non-orthogonal system ( x1 , x2 ) in 2D, as shown in Figure 4.1. The components of vector A can be obtained by drawing parallel lines to the coordinates x1 and x2 to find contravariant components A1 and A2, respectively.
Tensor-Analysis-for-Engineers_2E_CH04.indd 11
10/3/2020 10:19:57 AM
12 • Tensor Analysis for Engineers, 2E
FIGURE 4.1 Contravariant components of a vector in an oblique coordinate system.
With reference to Figure 4.1, we can also find another set of components, A1 and A2 of the same vector A by drawing perpendicular lines to the coordinates x1 and x2 . This is shown in Figure 4.2. Obviously, components A1 and A2 are different in magnitude from the contravariant components. We define A1 and A2 as covariant components. In the next section we define the transformation rule for covariant quantities.
FIGURE 4.2 Covariant components of a vector in an oblique coordinate system.
We can conclude from these definitions that contravariant and covariant components of a vector in a Cartesian system are identical and there is no distinction between them.
Tensor-Analysis-for-Engineers_2E_CH04.indd 12
10/3/2020 10:19:58 AM
CHAPTER
5
COVARIANT COMPONENTS AND TRANSFORMATIONS
We use the standard notation of writing the index i as a subscript for covariant quantities. We consider basis vector ei as a covariant quantity. We can rewrite Equation 3.6, considering two arbitrary systems x j and x′k and the fact that ds is coordinate-system independent or invariant, as ′k e′k dx j e j 5.1 = ds dx = ∂x j After substituting for dx j , using Equation 4.7, we get dx′k e′k = k dx′k e j or ∂x′ ∂x j dx′k e′k − k e j = 0 5.2 ∂x′
Since dx′k is arbitrary (i.e., the relation is valid for any choice of system = x′k and selections 0 , for all values of i = 1,2,, N ) of dx′k i ≠ 0 and dx′k ≠ i = therefore the expression in the bracket must be equal to zero, or we have
Tensor-Analysis-for-Engineers_2E_CH05.indd 13
∂x j e′k = k e j 5.3 ∂x′
10/3/2020 10:20:15 AM
14 • Tensor Analysis for Engineers, 2E
All quantities, say Ai , that transform according to Equation 5.3 defined as covariant type, with the standard notation of writing the index i as a subscript. Therefore, transformation Ai A′j reads
= Ai
∂x′ j ∂x i ′ ′ = A and A Ai 5.4 j j ∂x i ∂x′ j
See Figure 4.2 for the geometrical interpretation of the covariant c omponent of a vector.
Tensor-Analysis-for-Engineers_2E_CH05.indd 14
10/3/2020 10:20:15 AM
CHAPTER
6
PHYSICAL COMPONENTS AND TRANSFORMATIONS
Having defined the contravariant and covariant quantities, like those of the components of a vector, one can ask this question: which one of these two types of components is the actual vector’s components in magnitude? The short answer is “none”! However, note that the combination of the contravariant or covariant components and their corresponding basis vectors (contravariant basis vectors will be defined in further section) gives the magnitude of the vector correctly. To find the actual magnitude of the vector, we should find the unit vectors and then the components of the vector corresponding to these unit vectors, or the physical components. In other words, the physical components of a vector (or in general, a tensor) are scalars whose physical dimensions and magnitudes are those of the components vectors tangent to the coordinates’ lines. Therefore, for vector A , if we designate A ( i ) as its physical component and the corresponding unit vector as e ( i ) we can write, then i = A A= ei A ( i ) e ( i ) 6.1 Now, using Equation 3.8 and substituting for ei i A e ( i ) hi = A ( i ) e ( i ) . Therefore,
we can write
A ( i ) = A i hi , ( no summation on i ) 6.2
Or A (1 ) = A1 h1 , A ( 2 ) = A2 h2 , and so forth. The scale factor hi is the parameter that turns the contravariant component into the physical component, when multiplied by it. Again, we can conclude that in a Cartesian system the physical and contravariant/covariant components are identical, since hi = 1 .
Tensor-Analysis-for-Engineers_2E_CH06.indd 15
10/3/2020 10:20:31 AM
16 • Tensor Analysis for Engineers, 2E Up to this point we have defined contravariant, covariant, and physical components of a vector along with covariant and physical/unit basis vectors. In principle, we can represent a vector by any combination of its components with the corresponding basis vectors. This requires having the contravariant basis vector definition. We will define this quantity and how it is transformed in a further section. First, we must have the definition of a tensor, as well as a specific tensor called a metric tensor.
Tensor-Analysis-for-Engineers_2E_CH06.indd 16
10/3/2020 10:20:32 AM
CHAPTER
7
TENSORS—MIXED AND METRIC In previous sections, we defined vectors or tensors of rank one. Higher order tensors could be defined based on similar definitions. For example, a tensor of rank two requires more than one free index, like mechanical stress or strain tensors. In general, we have the contravariant components written as A ij for tensor A . We avoid placing double-arrows over the symbol for tensors, for simplicity and generality. Since, in principle, we can have tensors of rank N, designating them with N number of hatted arrows is not a practical exercise. The invariant quantity is written as A = A ij ei e j 7.1
Since A is invariant, its value remains the same regardless of the coordinate system used. We transform A from an arbitrary coordinate system x′i ij to another system, say x i , or A′ e′i e′j = A km ek em . Now, using Equation 5.3, we transform the covariant basis vectors ek and em to system x′i . Therefore, ∂x′i ∂x′ j we have A km ek em = A km k m e′i e′j . Using these last two relations, we obtain ∂x ∂x j i ′ ′ ∂ x ∂ x 0 . Therefore, since basis vectors are non-zero, e′i e′j A′ij − k m A km = ∂x ∂x we have
Tensor-Analysis-for-Engineers_2E_CH07.indd 17
A′ij =
∂x′i ∂x′ j km A 7.2 ∂x k ∂x m
10/3/2020 10:20:47 AM
18 • Tensor Analysis for Engineers, 2E Quantities that transform according to Equation 7.2 are defined as doubly contravariant tensors, of rank two. Similarly, we can define mixed tensors, for example of rank two, as A′ji =
∂x′i ∂x m k Am 7.3 ∂x k ∂x′ j
The rule for transformation of tensors is like those given for vectors, i.e., the contravariant component of the tensor transforms like a contravariant vector and the covariant component like a covariant vector. Also, we can expand the definition to a doubly covariant tensor, like A′ij =
∂x k ∂x m Akm 7.4 ∂x′i ∂x′ j
Equations 7.2–7.4 are useful relations defining second rank tensors and their transformations. Expanding this rule, we can write the transformation of a N -rank tensor, for example contravariant components, as A′ijN =
∂x′i ∂x′ j ∂x′N kmM A 7.5 ∂x k ∂x m ∂x M
In an arbitrary curvilinear system, the Pythagorean Theorem applies but does not appear in the same form as it does in flat Cartesian systems. This is mainly due to the fact that basis vectors in an arbitrary system change in magnitude and/or direction, from point to point. Therefore, measuring the distance ds between two points on a curved surface, like a sphere, for example, requires scaling the coordinates by multiplying them with some scalar coefficients. These coefficients are the components of a tensor, called a metric tensor, associated with the coordinate system selected. Recalling Equation 3.6, we can find the square of the magnitude of differential distance by dot- ds product operation, or ds2 = ds ⋅ ds = ( dx i ei ) ⋅ dx j e j . The result is
(
)
= ds2 dx i dx j ei ⋅ e j 7.6
The term in the bracket, on the R.H.S of Equation 7.6, is a scalar (i.e., a t ensor of rank zero) resulting from a dot-product operation on the basis vectors. Hence, we can write it as = ei ⋅ e j ei e= hi h j cos α j cos α 7.7
(
Tensor-Analysis-for-Engineers_2E_CH07.indd 18
)
10/3/2020 10:20:48 AM
Tensors—Mixed and Metric • 19
where α is the angle between the tangents to x i and x j axes of the c oordinate system at the selected point in space. Geometrically speaking, Equation 7.7 is the multiplication of the magnitude of a basis vector with the projection of the other one along the direction of the original one. Therefore, this quantity could be used to identify whether a system (or at least the basis vectors selected) is orthogonal or not. Also, it could be a metric for measuring the distance over a curved surface. It has more properties and applications, which we will encounter in future sections; hence, it is designated by a s ymbol and a name, i.e., metric tensor gij gij= ei ⋅ e j 7.8
It is easily seen from Equation 7.8 that a metric tensor is symmetric since order in a dot-product is irrelevant (i.e., commutative). Also, if gij = 0 for i ≠ j then the coordinates of the system x i are orthogonal. For example, the metric tensor in a 3D orthogonal system can be represented by a 3 × 3 matrix containing null off-diagonal elements, g11 gij = 0 0
0 0 , orthogonal system 7.9 g33
0 g22 0
It would be useful to apply metric tensor definitions and expand Equation 7.6 using Equations 7.7 and 7.9 for an orthogonal 3D system, which gives ds2 = h12 ( dx1 ) + h22 ( dx2 ) + h32 ( dx3 ) 7.10 2
2
2
For example, in a Cartesian system we recover the familiar form of 2 2 2 2 the Pythagorean theorem (i.e., ds = ( dX ) + ( dY ) + ( dZ ) ), since g11 = g22 = g33 = Ei ⋅ Ei = 1. Similarly, for a spherical coordinate system ( r ,ϕ ,θ ) no sum on i
2 2 2 g= 1 , h= g= r 2 , and h= g= r 2 sin 2 ϕ (see Example 8.2 we have h= 1 11 2 22 3 33 2 2 for scale factors). Therefore, we obtain ds = ( dr ) + r 2 ( dϕ )2 + r 2 sin 2 ϕ ( dθ )2 .
So far, we have not shown that gij is a tensor. To do this, we transform the metric tensor to another arbitrary coordinate system, say x′i . We then can ∂x k ∂x m ∂x k ∂x m ′ ⋅ e e g = gkm ; write g′ij = e′i ⋅ e′j = i . Therefore, we have ( ) m ij k ∂x′ ∂x′ j ∂x′i ∂x′ j = gkm
that is, when compared to Equation 7.4 we can conclude that gkm is a doubly covariant tensor.
Tensor-Analysis-for-Engineers_2E_CH07.indd 19
10/3/2020 10:20:51 AM
20 • Tensor Analysis for Engineers, 2E As mentioned in the previous sections, to calculate or measure tensorlike quantities we use a Cartesian system as the reference. To evaluate gij we use Equation 7.8 and transform the covariant basis vector ei to the k m ∂y ∂y Cartesian system, yi . Therefore, gij = ei ⋅ e j = i Ek ⋅ Em . The quanj ∂x ∂x tity Ek ⋅ Em , or the dot-product of the basis/unit vectors in the Cartesian system, is either one or zero, due to orthogonality of the coordinates. Hence, 1, when k = m . Using this property, we can write Ek ⋅ Em = 0, when k ≠ m
(
gij =
)
∂yk ∂yk 7.11 ∂x i ∂x j
Using Equation 7.11, we can readily conclude that the metric tensor in a Cartesian system is the familiar Kronecker delta, δ= Ei ⋅ E j . For 3D ij 1 0 0 Cartesian coordinates, we receive a diagonal unity matrix δ ij = 0 1 0 . 0 0 1 It can be shown ([3], [4]) that the determinant of a metric tensor is equal to the square of the Jacobian of the transformation matrix, or = g g= 2 7.12 ij
Therefore, in an orthogonal system xi , having gij = 0 for i ≠ j , we receive
= g g= h12 h12 hN2 , orthogonal system 7.13 11 g22 g NN gii e= hi2 (no sum on i ) (see Equation 7.7). From Note that= i ei Equations 7.12 and 7.13 we can conclude that the Jacobian for an orthogonal system is equal to the products of the scale factors, or
= h1 h2 hN , orthogonal system7.14
For example, in a spherical coordinate system the Jacobian is = h= r 2 sin ϕ (see Example 8.2 for scale factors). spherical r hϕ hθ
Tensor-Analysis-for-Engineers_2E_CH07.indd 20
10/3/2020 10:20:53 AM
CHAPTER
8
METRIC TENSOR OPERATION ON TENSOR INDICES An important and useful application of metric tensor is that it can be used to lower and raise indices of a tensor. Therefore, we can change a contravariant component of a tensor to a covariant one, or vice versa, by multiplying the appropriate metric tensor to the tensor at hand. For example, let’s define the covariant component of a vector (i.e., a tensor j of rank one) as Ai = A gij , where A j is the given contravariant compoj nent. Now, we show that the quantity A gij is transformed like a covariant quantity, according to Equation 5.4. Considering systems x i and x′i , ∂x′ j k ∂x m ∂x n ∂x′ j ∂x n ∂x m k we can write A′ j g′ij = = A g A gmn . k i j mn j k i ∂x ∂x′ ∂x′ ∂x ∂x′ ∂x′ 1 , n = k ∂x′ j ∂x n ∂x n = = δ kn , where δ kn = is j k k ∂x ∂x′ ∂x 0, n ≠ k a mixed second rank tensor (a Kronecker delta). Therefore, we have ∂x′ j ∂x n ∂x m k ∂x m n k ∂x m n δ = A g = A g ( k ) mn ∂x′i ( A gmn ) . The expression in k mn j i ∂x′i ∂x ∂x′ ∂x′ n
Note that the expression
=A
the last bracket is actually Am, according to our definition. Finally, we have ∂x m ′ j g′ij = A′i A= Am , which clearly shows that the quantity A′ j g′ij is trans∂x′i formed like a covariant component (see Equation 5.4). Therefore, we have, also using the symmetry property of a metric tensor or gij = g ji ,
Tensor-Analysis-for-Engineers_2E_CH08.indd 21
j = Ai A= gij A j g ji 8.1
10/3/2020 10:21:18 AM
22 • Tensor Analysis for Engineers, 2E
Using Equation 8.1, we can write the expression for a contravariant basis vector, or ei = e j gij 8.2
which shows that the metric tensor lowers the contravariant index, whilethe dummy index is dropped out. This will enable us to write the vector A in terms of covariant or contravariant basis vectors, as i = A A= ei A i gij= ej A i gij = e j A j e j 8.3
Note that by using physical components and corresponding unit vectors, we can write j i = A A= ei A= A ( k ) e ( k ) 8.4 je
(
) (
)
So far, we have defined the doubly covariant metric tensor gij . Similarly, the doubly contravariant metric tensor is defined g ij = e i ⋅ e j . Therefore, we can write ij i = A A= ei A j g= ei A j e j Also, the combination of the two types of metric tensor gives the mixed one, as g ij g jk = δ ki 8.5
ij Equation 8.5 can be derived as follow: we have= A i g= A j g ij g jk A k . In this
(
=Aj
)
0 , but expression, we write A = δ A , therefor we have A δ − g g jk = ij k since A is arbitrary the expression in the bracket is zero. Hence g g jk = δ ki. i
i k
k
k
i k
ij
With reference to Figure 3.2, Figure 4.1, Figure 4.2, and Equation 8.3 we can sketch the vector A in terms of its covariant/contravariant components with their corresponding basis vectors, as shown in Figure 8.1 (see Example 15.3). Users may also want to watch a related video at https://www.youtube.com/watch?v=CliW7kSxxWU.
Tensor-Analysis-for-Engineers_2E_CH08.indd 22
10/5/2020 11:42:13 AM
Metric Tensor Operation on Tensor Indices • 23
FIGURE 8.1 Sketch for covariant and contravariant components of a vector and basis vectors in a non-orthogonal coordinate system xi and alternate system ξ i .
In the following subsections, we present two examples demonstrating related calculations for cylindrical and spherical coordinate systems.
8.1 EXAMPLE: CYLINDRICAL COORDINATE SYSTEMS Consider a cylindrical polar coordinate system ( x1 , x2 , x3 ) ≡ ( r ,θ , z ) where r is the radial distance, θ the azimuth angle, and z the elevation from X − Y plane, w.r.t. the Cartesian coordinates ( y1 , y2 , y3 ) ≡ ( X , Y, Z ) , as shown in Figure 8.2. The functional relations corresponding to cylindrical and Cartesian systems are:
FIGURE 8.2 Cylindrical coordinate system.
= cosθ y1 x1 sin x2 X r= 2 = r sin θ ≡ y= x1 sin x2. Find the inverse relations (i.e., x i = Fi ( y1 , y2 ,, yN ) Y Z z= y3 x 3 = x i = Fi ( y1 , y2 ,, yN ) , the basis vectors ei and e i for the cylindrical coordinate system in terms of the Cartesian unit vectors. Also find the scale factors, unit vectors, and metric tensors gij and g ij and line element magnitude ds .
Tensor-Analysis-for-Engineers_2E_CH08.indd 23
10/3/2020 10:21:22 AM
24 • Tensor Analysis for Engineers, 2E Solution: The functions Fi can be obtained by solving for X , Y, Z , using the functional relation, given
1 1 2 2 2 x y y 2 2 = + ( ) ( ) r X +Y = y1 2 −1 Y . To find the covariant basis vectors, x tan −1 2 = θ tan ≡ = X y z = Z x3 = Y 3 j 1 2 3 ∂y ∂y ∂y ∂y we use ei = i E j . Therefore, we receive e1 = 1 E1 + 1 E2 + 1 E3 = cosθ E1 + sin θ E2 ∂x ∂x ∂x ∂x 1 2 3 1 2 3 ∂y ∂y ∂y ∂y ∂y ∂y . Similarly, , e = E + E + E = − r sin θ E + r cos θ E E E E cos θ E sin θ E + + = + 2 1 2 3 1 2 1 2 3 1 2 2 2 2 ∂x ∂x ∂x ∂x1 ∂x1 ∂x1 1 2 3 ∂y ∂y ∂y and e3 = 3 E1 + 3 E2 + 3 E3 = E3 . In terms of coordinates notation, we ∂x ∂x ∂x have er cosθ E1 + sin θ E2 = − r sin θ E1 + r cosθ E2 eθ = ez = E3 ∂x i For calculating contravariant basis vectors, we receive e i = j E j. After ∂y expansion and using the functional relations, we get e r cosθ E1 + sin θ E2 = sin θ cosθ θ − E1 + E2 e = r r z e = E3 The scale factors are the magnitudes of the covariant basis vectors. Hence, h= h= er ⋅ er = 1 , h= h= eθ ⋅ e= r , and h= h= 1 . The unit vectors 1 r 2 θ θ 3 z e e ( i ) = i , are hi e ( r ) cosθ E1 + sin θ E2 = − sin θ E1 + cosθ E2 e (θ ) = e ( z ) = E3
Tensor-Analysis-for-Engineers_2E_CH08.indd 24
10/3/2020 10:21:24 AM
Metric Tensor Operation on Tensor Indices • 25
Note that unit vectors e ( r ) and e (θ ) in cylindrical coordinate system change with location and are not constant vectors. 2 2 2 h= 1 , gθθ= h= r 2 , g= h= 1 , and The metric tensor is gij = ei ⋅ e j , or g= rr r θ zz z the off-diagonal elements of the metric tensor are null, hence the coordinate 1 0 0 system is orthogonal, or gij = 0 r 2 0 . The Jacobian= is h= r. r hθ hz 0 0 1 The contravariant metric tensor can be calculated using g ij = e i ⋅ e j , or 1 0 0 ij g = 0 r −2 0 , which is equal to the inverse of covariant metric tensor. 0 0 1
Another way for calculating the contravariant basis vectors is by using the 2 ij 11 33 22 e 1 g= e1 e1 = e 3 g= e3 e3 . relation e i = g e j , or= , e g= e2 r −2 e2 , and= 2 2 2 2 For the line element we have, ds = ( hi dx i ) = ( dr ) + ( rdθ ) + ( dz ) .
8.2 EXAMPLE: SPHERICAL COORDINATE SYSTEMS Consider a spherical polar coordinate system ( x1 , x2 , x3 ) ≡ ( r ,ϕ ,θ ) where r is the radial distance, ϕ the polar/meridian angle, and θ the azimuthal angle w.r.t. the Cartesian coordinates ( y1 , y2 , y3 ) ≡ ( X , Y, Z ) , as shown in Figure 8.3. The functional relations corresponding to spherical and Cartesian systems are:
FIGURE 8.3 Spherical coordinate system.
Tensor-Analysis-for-Engineers_2E_CH08.indd 25
10/3/2020 10:21:25 AM
26 • Tensor Analysis for Engineers, 2E
= sin ϕ cosθ y1 x1 sin x2 cos x3 X r= = ≡ y2 x1 sin x2 sin x3 . Find the inverse relations (i.e., Y r sin ϕ sin θ = 1 2 3 = cos ϕ Z r= y x cos x x i = Fi ( y1 , y2 ,, yN ) , the basis vectors ei and e i for the spherical coordinate system in terms of the Cartesian unit vectors. Also find the scale factorsand unit vectors and metric tensors gij and g ij and line element magnitude ds . Solution: The functions Fi can be obtained by solving for X , Y, Z , using corresponding func 2 2 2 x1 = y1 ) + ( y2 ) + ( y3 ) ( 2 2 2 r= X + Y + Z y3 Z 2 −1 −1 = = ≡ x cos tional relations, or ϕ cos 2 2 2 2 2 2 X +Y +Z ( y1 ) + ( y2 ) + ( y3 ) Y θ = tan −1 y1 X x3 = tan −1 2 y
.
∂y To find the covariant basis vectors, we use ei = i E j. Therefore, we get ∂x 1 2 3 ∂y ∂y ∂y e1 = 1 E1 + 1 E2 + 1 E3 = sin x2 cos x3 E1 + sin x3 sin x2 E2 + cos x2 E3 Similarly, ∂x ∂x ∂x 1 2 3 ∂y ∂y ∂y e2 = 2 E1 + 2 E2 + 2 E3 = x1 cos x2 cos x3 E1 + x1 sin x3 cos x2 E2 − x1 sin x2 E3 , and ∂x ∂x ∂x 1 2 ∂y ∂y3 ∂y e3 = E + E + E3 = − x1 sin x2 sin x3 E1 + x1 cos x3 sin x2 E2 . In terms of coor1 2 3 3 3 ∂x ∂x ∂x dinate variables, we have er = sin ϕ cosθ E1 + sin ϕ sin θ E2 + cos ϕ E3 eϕ = r cos ϕ cosθ E1 + r cos ϕ sin θ E2 − r sin ϕ E3 − r sin ϕ sin θ E1 + r sin ϕ cosθ E2 eθ = j
Tensor-Analysis-for-Engineers_2E_CH08.indd 26
10/3/2020 10:21:27 AM
Metric Tensor Operation on Tensor Indices • 27
The scale factors are the magnitudes of the basis vectors. Hence h= h= er ⋅ er = 1 , h= h= eθ ⋅ e= r sin ϕ . h= eϕ ⋅ e= r , and h= 1 r 3 θ θ 2 ϕ ϕ e The unit vectors e ( i ) = i , are hi e ( r ) = sin ϕ cosθ E1 + sin ϕ sin θ E2 + cos ϕ E3 e (ϕ ) = cos ϕ cosθ E1 + cos ϕ sin θ E2 − sin ϕ E3 − sin θ E1 + cosθ E2 e (θ ) = Note that unit vectors in spherical coordinate system change with location 2 h= 1, and are not constant vectors. The metric tensor is gij = ei ⋅ e j , or g= rr r 2 2 2 2 2 gϕϕ= h= r , gθθ= h= r sin ϕ , and the off-diagonal elements of ϕ θ the metric tensor are null, hence the coordinate system is orthogonal, 1 0 0 = r 2 sin ϕ . The g ii = hi−2 , is h= gij = 0 r 2 0 . The Jacobian r hϕ hθ 2 0 0 ( r sin ϕ ) 1 0 0 or g ij = 0 r −2 0 , then we can calculate the contravariant basis −2 0 0 ( r sin ϕ ) 1 1 vectors, as e j = g ij ei , or e 1 = e1 , e 2 = 2 e2 , and e 3 = 2 2 e3 which yields r r sin ϕ
e r = sin ϕ cosθ E1 + sin ϕ sin θ E2 + cos ϕ E3 ϕ cos ϕ cosθ cos ϕ sin θ sin ϕ E1 + E2 − E3 e = r r r sin θ cosθ θ e E E2 = − + 1 r sin ϕ r sin ϕ For the line element we have, ds =
Tensor-Analysis-for-Engineers_2E_CH08.indd 27
( h dx )
i 2
i
=
( dr )2 + ( rdϕ )2 + ( r sin ϕ dθ )2 .
10/3/2020 10:21:29 AM
Tensor-Analysis-for-Engineers_2E_CH08.indd 28
10/3/2020 10:21:29 AM
CHAPTER
9
DOT AND CROSS PRODUCTS OF TENSORS We often must multiply tensors by each other. We consider two vectors for discussion here, without losing generality. When we multiply two scalars (i.e., tensors of rank zero) we just deal with arithmetic multiplication. But a vector has a direction, in addition to its magnitude, that should be taken into consideration for multiplication operations. In principle we can have two vectors just multiply by each other like AB, which is a tensor of rank two, called a dyadic product. Or, multiply the vectors’ magnitudes and form a new vector with the resulting magnitude directed perpendicular to the plane containing the original vectors, the cross-product A × B, which is a tensor of rank one. Or, multiply the vectors’ magnitudes and reduce the rank of the product by projecting one vector on the other one, the dot-product A ⋅ B , which is a tensor of rank zero. Obviously, for higher than rank-one tensors we would have a greater number of combinations as the result of their products. In general, compared to the rank of original tensors, dyadic product increases the rank, cross-product keeps the rank, and dot-product decreases the rank, i.e., the result is a tensor quantity of higher, the same, or lower rank (by one level/ rank) w.r.t to the original quantities, respectively. We start with dot-product operations. Let’s consider A and B ; two vectors ⋅ B ( A i ei ) ⋅ B j e j , using the dot-product of these two vectors is given by A = the contravariant components of the vectors. We expand this expression, with Equations 7.7 and 7.8, to receive
(
Tensor-Analysis-for-Engineers_2E_CH09.indd 29
)
j A= ⋅ B A i B= gij A i B j hi h j cos α 9.1
10/3/2020 10:22:00 AM
30 • Tensor Analysis for Engineers, 2E
where α is the angle between the two vectors. Further, using Equation 8.1 we receive A⋅B = A i B=i A j B j 9.2 Similarly, we could start with using the covariant components of the vector, but the conclusive results would be the same as the expression given by Equation 9.2. Readers may want to do this as an exercise. From Equation 9.2 we can conclude that the order of multiplication does not change the result of the dot-product, or A i Bi = Bi A i . In an orthogonal Cartesian system, either Equation 9.1 or 9.2 would result in A ⋅ B= A i B jδ ij= A1 B1 + A2 B2 + + A N BN= A1 B1 + A2 B2 + + AN BN . The second equality results from the fact that in orthogonal Cartesian systems there is no distinction between contravariant and covariant components. An example of the dot-product quantity of two vectors is the mechanical work— the dot product of force and distance vectors. The cross-product operation can be established with a more general formulation using the permutation symbol (also referred to as the LeviCivita symbol) defined as +1 even permutation ei1 i2 iN = −1 odd permutation 9.3 0 repeated index
For example, for N = 6 we receive ei1 i2 i6. The permutation is considered even if an even number of interchanges of indices put the indices in arithmetic order (i.e., 123456 ), and is considered odd if an odd number of interchanges of indices put the indices in order. For any case with a repeated index the permutation symbol is zero. As an example, let’s consider e342165 . Examining the indices, we find out that 4 interchanges (i.e. ( 3 ↔1 )
(4↔2)
(4↔3)
(6 ↔ 5)
342165 → 142365 → 124365 → 123465 → 123456 ) put them in order; note that the sequence of interchanges is irrelevant. Therefore e342165 = 1 . Similarly, e362145 = −1 , since 5 interchanges put the indices in order. If any of the indices is repeated then it is zero, for example e= e= 0, etc. We 344165 142165 purposefully called the permutation symbol a symbol and not a tensor, for reasons that will be explained later in this section. Another way of identifying the even or odd number of permutations is to write down the given indices as a set and connect them with the equivalent number set but in arithmetic
Tensor-Analysis-for-Engineers_2E_CH09.indd 30
10/3/2020 10:22:02 AM
Dot and Cross Products of Tensors • 31
order. Then the number of points at the cross-section of the lines connecting the same numbers is the order of permutation. This method is shown in the following figure for e342165. The connecting lines intersect at six points; hence the permutation is even. 3
4
2
1
6
5
1
2
3
4
5
6
Now considering two vectors A and B in a Cartesian system the cross product of these two vectors is another vector C , given as (we propose this expression, for now at least)
C i = eijk A j Bk = − eijk B j A k , Cartesian 9.4 A× B
B× A
for the i -component of C . This relation clearly shows that the order of multiplication matters for cross-product operation. For example, in a 3D Cartesian system we have A = ( A1 , A2 , A3 ) and B = ( B1 , B2 , B3 ) and we find their cross-product using Equation 9.4, C= ( C1 , C2 , C3 )= A × B as, = C1 A2 B3 − A3 B2 = C3 A1 B2 − A2 B1 . , C2 A3 B1 − A1 B3 , and=
Note that in Equation 9.4, we have written the relation for the contravari ant component of the resulting vector C. As we mentioned previously, in a Cartesian system contravariant or covariant components need not be considered; hence we usually place all indices as subscripts. But in an arbitrary system we should make sure that the resulting C i is transformed like a contravariant component/quantity. To show this, we set guidelines to form a tensor expression in an arbitrary coordinate system, x i as follows: a. Identify the rank of the expression representing the desired quantity in terms of a tensor; i.e., zero is a scalar, one is a vector, two or more a tensor of the corresponding rank. b. Identify and decide if the expression should be in contravariant or covariant form and place the indices appropriately (i.e., as subscript for covariant and superscript for contravariant). c. Form the expression based on guidelines (a) and (b).
Tensor-Analysis-for-Engineers_2E_CH09.indd 31
10/3/2020 10:22:04 AM
32 • Tensor Analysis for Engineers, 2E
d. Transform the expression from arbitrary system x i to another system, ∂x i ∂x′i say x′i (or vice versa) using the linear operators j or according to ∂x′ j ∂x the combination rule (see Sections 4 and 5). e. Show that the resulting expression recovers the related familiar form in Cartesian coordinates. Usually, starting from an expression form in a Cartesian system (as listed in part (e), above) gives a reasonable and usually the correct form to start with. Let’s apply these guidelines to the expression given by Equation 9.4. By observing the expression, we conclude that the contravariant components A j and Bk are correct forms since they transform like contravariant quantities. However, for the permutation symbol we need to make sure that it is a tensor quantity, i.e., it transforms like a tensor. It can be shown that the permutation symbol does not transform like a tensor unless a necessary adjustment is implemented [2], [4], [7] by multiplying it by the Jacobian of the transformation (see Equation 4.4). Now, we define the covariant and contravariant permutation tensors by multiplying Jacobian and inverse of Jacobian to the permutation symbol, respectively, given as
i1 i2 iN = ei1 i2 iN 9.5 i1 i2 iN = −1 e i1 i2 iN Therefore, the general expression for the vector C, for example, is Ci = eijk A j Bk and C i = −1 eijk A j Bk . Using permutation tensor, we will get j k Ci = eijk A j Bk covariant component = ijk A B 9.6 i ijk A j Bk −1 eijk A j Bk contravariant component = C =
Note that in Equation 9.6, for the covariant (contravariant) component of the resulting vector C , we used the combination of covariant (contravariant) permutation tensor and (covariant) components of the two contravariant contributing vectors A and B. j To show that, for example, ijk A Bk transforms like a covariant quantity, we write down its transformation between two arbitrary systems, or l ∂x l ∂x m ∂x n p ∂x′ j q ∂x′k p q ∂x ′ A′ j B′k lmn i δ mδ n = A = B = A B ijk lmn j p q k i p q ′ ′ ′ ′ ∂x ∂x ∂x ∂x ∂x ∂x
∂x l . Similarly, for the contravariant form (i.e., the second ∂x′i expression in Equation 9.6) it can be shown that it does indeed transform like a contravariant quantity. Readers may want to do this as an exercise.
(
lmn
A m Bn )
Tensor-Analysis-for-Engineers_2E_CH09.indd 32
10/3/2020 10:22:06 AM
Dot and Cross Products of Tensors • 33
The cross-product of two vectors can be written in terms of their physical components as well. This can be done using, for example, the contravariant component when multiplied by the corresponding scale factor. Therefore, we can write for an orthogonal system (using Equations 9.5, 9.6, and 11.8) hi eijk hi eijk h j hk A ( j ) B ( k ) , or hiC i = hi ijk A j Bk= , or C ( i ) = A j Bk C( i )
eijk
C ( i) =
hi h j hk A ( j ) B ( k ) , orthogonal 9.7
For orthogonal coordinate systems, using = h1 h2 h3 , we receive
( A × B) (1) =C (1) = A ( 2 ) B ( 3 ) − A ( 3 ) B ( 2 ) ( A × B) ( 2 ) =C ( 2 ) = A ( 3 ) B (1) − A (1) B ( 3 ) 9.8 ( A × B) ( 3 ) =C ( 3 ) = A (1) B ( 2 ) − A ( 2 ) B (1)
Expanding on these results, in a N -dimensional space, the generalized crossproduct of N − 1 vectors reads, using Equation 9.6,
Ci1 = i1 i2 iN A i2 Bi3 Z iN 9.9 i1 i i i i i i C = 1 2 N A 2 B 3 Z N
where now vector C is perpendicular to the hyperplane formed by the N − 1 vectors A through Z .
Immediately, using Equation 9.9 we can conclude that in an arbitrary system the unit volume d is, [2], d = dx1 dx2 dx N 9.10
For example, in the spherical system ( r , ϕ , θ ) we have d = r 2 sin ϕ drdϕ dθ . Another useful extension is for the orthogonal systems (i.e., gi ≠ j = 0), for which we get= gij ∏ = gii g11 g22 gNN . But g11 = e1 ⋅ e1 = h12 , g22 = e2 ⋅ e2 = h22 and so forth. Hence, gij = h12 h22 hN2 . Using the relationship gij = 2 , we finally reach at
=
Tensor-Analysis-for-Engineers_2E_CH09.indd 33
h1 h2 hN orthogonal system 9.11
i
i
And hence = d
= h ∏
= h dx ( h dx )( h dx )( h ∏ i
i
i
1
1
2
2
N
dx N ) orthogonal system 9.12
10/3/2020 10:22:08 AM
34 • Tensor Analysis for Engineers, 2E
9.1 DETERMINANT OF AN PERMUTATION SYMBOLS
MATRIX USING
Using the permutation symbol eijk (see Equation 9.3), we can write the eijk M i1 M j 2 Mk 3, determinant of a 3 × 3 matrix= [ M ] , as M e= ijk M1 i M2 j M3 k given by expansion based on row or column, respectively, and M ij represent the elements of the matrix [2], [4]. To expand on these results, we consider [ M ] to be an N × N matrix, (N ≥ 3). Therefore, we can write its determinant as = M e= ei1 i2 iN M i1 1 M i2 2 M iN N 9.13 i1 i2 iN M1 i1 M2 i2 M NiN See the application example in Section 16.10.
Tensor-Analysis-for-Engineers_2E_CH09.indd 34
10/3/2020 10:22:09 AM
CHAPTER
10
GRADIENT VECTOR OPERATOR—CHRISTOFFEL SYMBOLS In engineering and science, most if not all governing equations (e.g., equilibrium, Navier-Stokes, Maxwell’s equations) contain terms that involve derivatives of tensor quantities. These equations are mathematical models of related quantity transports, like momentum, mass, energy, etc. Different forms of derivatives result from the gradient operator, which is a vector that takes the derivative of and performs dyadic, dot-product, cross-product, etc. operations on the involved tensors. Depending on the rank of the tensors under the gradient operation, the result could be complex expressions. In the previous sections, we have defined and derived necessary formulae to tackle the derivation of gradient vector transformation and consequently find the general forms of related derivatives appearing in the governing equations.
10.1 COVARIANT DERIVATIVES OF VECTORS— CHRISTOFFEL SYMBOLS OF THE 2ND KIND ∂ The covariant component of the gradient vector ∇ is defined as ∇ i = i , ∂x ∂ or we can write the full vector as ∇ = e i i = e i∇ i . To show that ∇ i trans∂x forms like a covariant component between two arbitrary systems, we can ∂ ∂x j ∂ ∂x j ′i write ∇= = = ∇ j . This expression clearly shows that ∇′i is a ∂x′i ∂x′i ∂x j ∂x′i covariant quantity. The transformation from x′i to x i is obvious.
Tensor-Analysis-for-Engineers_2E_CH10.indd 35
10/3/2020 10:22:35 AM
36 • Tensor Analysis for Engineers, 2E When a gradient operates on a scalar, or tensor, of rank zero like temperature, pressure, or concentration, we can write it as ∇ iψ and the corresponding ∂x j transformation is ∇′iψ ′ = ∇′iψ = i ∇ jψ , using the fact that scalar ψ is ∂x′ invariant (or ψ = ψ ′ ) and independent of the coordinate systems selected. Now, assuming that the gradient operates on a vector, say , which is also A an invariant quantity, the result is ∇A, or a tensor of rank two. The covariant component is then ∇ i A, which is a vector by itself. Expanding, we get ∇i A = ∇i A je j = e j ∇ i A j + A j ∇ i e j 10.1
(
)
(
)
(
)
The first term in the R.H.S of Equation 10.1 is the straight-forward derivative of the component A j . But the second term involves a gradient of the basis vector e j , which is not necessarily zero in an arbitrary system since the basis vector’s magnitude and/or direction may change from point to point. Note that in a Cartesian system, ∇ i E j = 0 , i.e., the second term in Equation 10.1, vanishes. Therefore, we can interpret ∇ i e j as a measure of the curvature of the arbitrary coordinate surface x i = constant (e.g., in a spherical coordinate system X 2 + Y 2 + Z 2 = r 2 is the surface of the sphere). The physi cal meaning of ∇ i e j is this that it represents the rate of change of e j with i respect to coordinate variable x in the direction of basis vector ek . This k quantity is represented by Γ kij (some authors use instead) and is the ij Christoffel symbol of the second kind [8]. It may be useful to pause at this point and make sure that the physical/ geometrical meaning of the Christoffel symbol is well understood. To help with this understanding, we consider a simple 2D polar coordinate system, ( r,θ ) ≡ ( x1 , x2 ) as shown in Figure 10.1.
FIGURE 10.1 2D Polar coordinate system.
Tensor-Analysis-for-Engineers_2E_CH10.indd 36
10/3/2020 10:22:38 AM
Gradient Vector Operator—Christoffel Symbols • 37
As shown, the covariant basis vectors er and eθ change from point to point in space, e.g., points A and A’ , in general. Let’s consider, for example ∂e ∇θ eθ =θ which represents the change for basis vector eθ with respect to ∂θ coordinate variable θ . This derivative is a vector and has two components in the ( r,θ ) coordinate system. Therefore, we can write it as a linear combina tion of the basis vectors in polar coordinate system er and eθ , or ∂eθ ∇θ eθ = = α er + β eθ 10.2 ∂θ
Similar expressions can be written down for other possible derivatives (i.e., ∇θ er , ∇ r eθ , ∇ r er ). In our example (see Equation 10.2), the coefficients α and β are the corresponding Christoffel symbols, also referred to as connec∂eθ tion coefficients. α is the component of the vector in the direction of er , ∂ θ ∂eθ r or Γθθ and β is the component of the vector in the direction of eθ , or ∂θ Γθθθ . Using these definitions, we can write all possible derivatives with their corresponding Christoffel symbols as ∂eθ r θ ∂θ = Γθθ er + Γθθ eθ ∂er = Γ r e + Γθ e rθ r rθ θ ∂θ 10.3 ∂eθ = Γθr r er + Γθθ r eθ ∂r ∂e r = Γ rrr er + Γθrr eθ ∂r The values of Christoffel symbols can be calculated using the functional relations between the polar and Cartesian systems. For this purpose, we need the relations for the basis vectors. Using Equation 5.3, we have ∂X ∂Y ∂X ∂Y = = er E1 + E2 and eθ E1 + E2 , where Ei are unit vectors in ∂r ∂r ∂θ ∂θ Cartesian system. Performing the calculations, using the functional relation between the polar and Cartesian systems, as X = r cosθ and Y = r sin θ , we receive er cosθ E1 + sin θ E2 = 10.4 − r sin θ E1 + r cosθ E2 eθ =
Tensor-Analysis-for-Engineers_2E_CH10.indd 37
10/3/2020 10:22:43 AM
38 • Tensor Analysis for Engineers, 2E ∂eθ Using Equations 10.4 we receive = − r cosθ E1 − r sin θ E2 = − rer , ∂θ ∂er ∂e 1 ∂eθ 1 = − sin θ E1 + cosθ E2 = eθ , = − sin θ E1 + cosθ E2 = eθ , and r = 0 . ∂θ r ∂r r ∂r Finally, comparing these results with Equation 10.3, we find the Christoffel 1 r symbols values for polar coordinate system as Γθθ = − r , Γθrθ = Γθθ r = , and r Γθθθ =Γθrr =Γθr r =Γ rrθ =Γ rrr =0 . Table 10.1 gives the summary of these results and the matrix form of these symbols for polar coordinate systems. TABLE 10.1 Christoffel symbols of the 2nd kind for a 2D polar coordinate system ( r ,θ ) .
Symbol Meaning r eθ change w.r.t θ Γθθ eθ change w.r.t r Γθr r er change w.r.t r Γ rrr Γθθθ eθ change w.r.t θ er change w.r.t θ Γθrθ Γθrr
in er direction in er direction in er direction in eθ direction in eθ direction
er change w.r.t r in eθ direction
Value −r
Symmetry −
0
= Γ
0
−
0
−
1/r 0
r rθ
Matrix form Γr Γ r = rrr Γθ r
θ
θ θr
= Γ
Γ rr Γθ = θ
Γθ r
Γ rrθ r Γθθ
Γθrθ θ Γθθ
−
The number of independent Christoffel symbols can be obtained using N 2 ( N + 1) , for an N -dimensional system. For example, in a 2D polar coor2 4×3 dinate system ( r,θ ) with N = 2 , we receive = 6 symbols, among them 2 only two are non-zero. Note that Christoffel symbols are symmetric with respect to the lower indices (see the next section). Now we continue with our general formulation and discussion, after establishing the physical/geometrical meaning of Christoffel symbols of 2nd kind with this example. Rewriting ∇ i e j using the definition of Christoffel symbol (see Equation 10.3), we have
Tensor-Analysis-for-Engineers_2E_CH10.indd 38
∇ie j = Γ kij ek 10.5
10/3/2020 10:22:48 AM
Gradient Vector Operator—Christoffel Symbols • 39
Plugging back into Equation 10.1, we receive ∇ i A = e j ∇ i A j + A j Γ kij ek . In this relation, for the last term k indices, since they are we interchange j and dummy indices. Then, ∇ i A = e j ∇ i A j + A k Γ ikj e=j e j ∇ i A j + Γ ikj A k . The expression in the bracket is a tensor, since both ∇ i A and e j are tensors. We j define the expression in the bracket as A, i where the comma in the subscript indicates differentiation and index i represents the covariant component of gradient vector. Hence, ∂A j = A, ji + Γ ikj A k 10.6 i ∂x
(
( (
)
)
)
Equation 10.6 is the covariant derivative of the contravariant component of vector A . Or ∇i A = A, ji e j 10.7
10.2 CONTRAVARIANT DERIVATIVES OF VECTORS At this point in our discussion, it seems logical to seek the contravariant component of gradient vector, ∇ i . We can write this operator as ∇ i = g ij ∇ j using a metric tensor (see Section 8). Therefore, we can raise the derivative index by multiplying the contravariant metric tensor to the expression, or A,k ⋅i = g ij A ,kj 10.8
Equation 10.8 is the contravariant derivative of the contravariant component of vector A . Or ∇i A = A,k ⋅i ek 10.9 From the start we could use the covariant component of vector A , to find the corresponding covariant and contravariant gradient components as well (see Equation 10.1). To this end, we can write the covariant derivative of the covariant component of vector A , as ∇i A = ∇i A je j = e j ∇ i A j + A j ∇ i e j 10.10
(
)
(
)
(
)
The first term on the R.H.S of Equation 10.10 is straight-forward, the derivative of the component A j . The second term involves the gradient of
Tensor-Analysis-for-Engineers_2E_CH10.indd 39
10/3/2020 10:22:50 AM
40 • Tensor Analysis for Engineers, 2E
the contravariant basis vector e j , for which we requirea new expression. To ∂e j e, ji + Γ ikj e k . Due to the find this quantity, we use Equation 10.6 to write = i ∂x fact that e, ji is a tensor (this can be shown by transformation of the terms involved between arbitrary systems, like A, ji ) and its value in a Cartesian system is null, hence it is equal to zero in any arbitrary coordinate system. j ∂e j Therefore, = e, i + Γ ikj = e k 0 , or ∂x i j ∂e j ∇ie = = −Γ ikj e k 10.11 i ∂x j Substituting for ∇ i e from Equation 10.11, into Equation 10.10, we receive ∇ i A j e j = e j ∇ i A j − A j Γ ikj e k . In this relation for the last term we interchange j and k indices, since they are dummy indices. Then, after factoring out e j , we get ∇ i A j e j = e j ∇ i A j − Ak Γ kij . We define the expression in the bracket as A j,i where the comma in subscript indicates differentiation and i represents the covariant component of gradient vector. Hence
(
)
(
)
(
(
)
)
(
A= j, i
)
∂A j
− Γ kij Ak 10.12 ∂x i Equation 10.12 is the covariant derivative of the covariant component of vector A . Or ∇i A = A j, i e j 10.13
10.3 COVARIANT DERIVATIVES OF A MIXED TENSOR Comparing Equations 10.6 and 10.12, we observe that for each index a term involving a Christoffel symbol is added to the expression on the R.H.S and when a covariant component of a vector is involved in a derivative operation, a minus sign is multiplied to the corresponding Christoffel symbol, contrary to a plus sign for when the contravariant component is involved. This rule can be used to extend the derivative calculation of a tensor of higher ranks. To show this, we consider a mixed tensor of third rank, for example Akij with the invariant A = Akij ei e j e k . The covariant derivative can be written as
∂Akij j i m + Γ nm Akjm + Γ nm Akim − Γ nk Amij 10.14 ∂x n Therefore, ∇ n A = Akij, n ei e j e k which is the component n of a tensor of rank four (i.e., ∇A = Akij, n ei e j e k e n ).
Tensor-Analysis-for-Engineers_2E_CH10.indd 40
ij A= k, n
10/3/2020 10:22:52 AM
Gradient Vector Operator—Christoffel Symbols • 41
10.4 CHRISTOFFEL SYMBOL RELATIONS AND PROPERTIES—1ST AND 2ND KINDS The definition of a Christoffel symbol of the first kind, properties, and some relations for calculation of Christoffel symbols are discussed in this section. The properties are: 1. Symmetry of the lower indices: The Christoffel symbol of the second kind is symmetric w.r.t lower indices, or Γ kij =Γ kji 10.15
To show this, we use Equation 10.5, after changing index k → n , and perform a dot product operation on it by contravariant basis vector e k. Or e k ⋅ ∇ i e j = Γ nij e k ⋅ en = Γ kij. Therefore, we can write δ nk
Γ kij = e k ⋅ ∇ i e j 10.16
(
)
We can use Equation 10.16 to calculate Christoffel symbol Γ kij in Cartesian coordinate system variables yi. After expanding the R.H.S expression by appropriate transformation and using chain rule, we will receive ∂x k ∂yn ∂ ∂yp ∂x k ∂yn ∂ ∂yp m e k= Ep (E ⋅ Ep ) = ⋅ ∇ i e j m Em ⋅ i n = j j m i n ∂y ∂x ∂y ∂x ∂y ∂x ∂y ∂x δm
(
)
p
p ∂x k ∂yn ∂x q ∂ 2 y ∂x q . Where, in the last expression we inserted ( = 1, ∂yp ∂x i ∂yn ∂x q ∂x j ∂x q
δ iq
or using chain rule). Finally, we receive p ∂x k ∂ 2 y Γ kij = p i j 10.17 ∂y ∂x ∂x
Equation 10.17 clearly shows the symmetry of Γ kij for its lower indices, since ∂ 2 yp ∂ 2 yp the order of differentiation is irrelevant, i.e., i j = j i . ∂x ∂x ∂x ∂x 2. Christoffel symbol is not a tensor: Because its value in a Cartesian system is zero; hence, if it transforms as a tensor it should be zero in any arbitrary system as well. But we know that this is not the case (see Equations 10.10 and 10.17).
Tensor-Analysis-for-Engineers_2E_CH10.indd 41
10/3/2020 10:22:54 AM
42 • Tensor Analysis for Engineers, 2E 3. Covariant derivative of the metric tensor: This quantity leads to a useful formula for calculating the value of a Christoffel symbol. First, we find the covariant derivative of the metric tensor, or ∇ i g jk =∇ i e j ⋅ ek =e j ⋅ ( ∇ i ek ) + ∇ i e j ⋅ ek . Now, using Equation 10.5, we can write ∇ i ek = Γ nik en and ∇ i e j = Γ nij en . Therefore, n n ∇ i g jk = Γ ik (e j ⋅ en ) + Γ ij (en ⋅ ek ) and finally we receive
(
)
(
gnk
g jn
)
∇ i g jk =
∂g jk
= Γ nik g jn + Γ nij gnk 10.18 ∂x i Note that by comparing Equation 10.18 with the rule of covariant derivatives of tensors (see Equation 10.14) we can conclude that g jk, i = 0, a reasonable result considering that a metric tensor in a Cartesian system is a constant/ unity and hence its derivative is equal to zero in an arbitrary system. 4. Christoffel symbol of the first kind: So far, we have defined the Christoffel symbol of the second kind. We now derive formulas that can be used for calculating its value as well as defining a Christoffel symbol of the first kind. To do this, we manipulate Equation 10.18 by permuting the indices (i.e., i → j, j → k, and k → i ), twice in sequence. Hence we ∂g get two alternative but equivalent relations as kij = Γ nji g jn + Γ njk gni and ∂x ∂gij n n = Γ kj gin + Γ ki gnj . Now we subtract Equation 10.18 from the sum of ∂x k the latter two relations, using along the way the symmetry of both metric tensor and Christoffel symbols. The result reads Γ nij gnk =
1 ∂g jk ∂gik ∂gij + j − k 10.19 ∂x ∂x 2 ∂x i
In Equation 10.19, the quantity Γ nij gnk is the Christoffel symbol of the first kind, also written as Γ ij, k ≡ ij, k. We can use Γ nij gnk to find a relation for calculation of the Christoffel symbol of the second kind in terms of metric tensor. To do this we multiply it by g mk to get Γ nij gnk g mk =δ nm Γ nij =Γ mij . Therefore, after multiplying Equation 10.19 by g mk, we get = Γ mij
g mk ∂g jk ∂gik ∂gij + j − k 10.20 ∂x ∂x 2 ∂x i
Equation 10.20 states that the Christoffel symbol of the second kind is equal to the first kind multiplied by the doubly contravariant metric tensor related to the coordinate system in use. Observing Equation 10.19, a symmetry shows up. That is, the derivative is w.r.t. the coordinates indices and for each term in the bracket the remaining indices are used for the metric tensor.
Tensor-Analysis-for-Engineers_2E_CH10.indd 42
10/3/2020 10:22:56 AM
Gradient Vector Operator—Christoffel Symbols • 43
5. For an orthogonal coordinate system we can simplify Equation 10.20, using the property of gij and g ij being diagonal tensors or matrices ij (see Equation 7.9). In other words, we can write g= g= 0 for i ≠ j . ij Therefore, examining the terms on the R.H.S of Equation 10.20, we can ∂gij ∂g conclude that k = 0 and ikj ≠ 0 only when i = k . Also g mk ≠ 0 when ∂x ∂x ∂g jk m = k and ≠ 0 , since j cannot be equal to k . By implementing ∂x i these relations among the indices (i.e., i= k= m and i ≠ j ) and rewriting Equation 10.20, we receive g ii ∂gii Γ iij = , 2 ∂x j
no summmation on i 10.21
Equation 10.20, for i= j ≠ k = m reduces to Γ kii =−
g kk ∂gii , 2 ∂x k
no summmation on i and k 10.22
For orthogonal coordinate systems, using Equation 7.13, we substitute for 2 hi−2 ∂ ( hi ) i 2 ii −2 , or gii = hi and g = hi in Equation 10.21 to obtain Γ ij = 2 ∂x j 1 ∂hi Γ iij = , no summmation on i, ( orthogonal ) 10.23 hi ∂x j Simply, by letting i = j in Equation 10.23, we receive 1 ∂hi Γ iii = , hi ∂x i
no summmation on i, ( orthogonal ) 10.24
2 hk−2 ∂ ( hi ) Similarly, Equation 10.22 yields Γ =− , or 2 ∂x k k ii
Γ kii =−
hi ∂hi , hk2 ∂x k
no summmation on i and k, ( orthogonal ) 10.25
6. Another useful relation can be derived for the value of Γ iin, which is ei change w.r.t x n in the direction of ei. We use the fact that the derivative of the permutation tensor, or ,nijk is zero, since its value in the Cartesian coordinates is null. Using Equation 10.14, we can
Tensor-Analysis-for-Engineers_2E_CH10.indd 43
10/3/2020 10:22:59 AM
44 • Tensor Analysis for Engineers, 2E
∂ ijk j ijm i + Γ nm mjk + Γ nm imk + Γ knm = 0 . To simplify the ∂x n derivation, without losing generality, we select=i 1,=j 2,= k 3. Hence, ∂ 123 m 23 1 2 1m3 3 12 m we get = − ( Γ nm + Γ nm + Γ nm ) . Collecting non∂x n 123 ∂ zero terms, we have = − ( Γ1n1 123 + Γ 2n2 123 + Γ 3n3 123 ) = − 123 Γ nii , n ∂x summation applies on i . Recalling that 123 = 1 / and Γ nii = Γ ini , we ∂ 1 1 get = − Γ iin (recall is the Jacobian). Performing the n ∂x differentiation and rearranging terms, we receive write = ,nijk
1 ∂ Γ iin = n 10.26 ∂x
Note that Equation 10.26 works for arbitrary coordinate systems and it recovers Equation 10.23 when the system is orthogonal. 10.4.1. Example: Christoffel symbols for cylindrical and spherical coordinate systems In this example, we calculate and derive the relations for Christoffel symbols of the second kind, for cylindrical and spherical coordinate systems. The results are also expressed in terms of unit vectors (i.e., the physical components) for each coordinate system. Readers should note that the results depend on the order of coordinate axes, defined. In other words, attention should be given when reading comparable results from other references in relation to their defined corresponding coordinates axes. Cylindrical coordinates From the results of Example 8.1, for a cylindrical coordinate system ( r ,θ , z ) (note the order of the coordinates defined here, see Figure 8.2) we can 2 2 write the non-zero terms of the metric tensor, as g= h= 1 , gθθ= h= r 2, rr r θ 2 rr and g= h= 1 . Therefore, the contravariant metric tensor reads g = 1, zz z gθθ = r −2 , and g zz = 1 . Using Equations 10.20 and 10.25 and considering that indices i, j, k, m are various permutations of cylindrical variables r,θ , z we receive the Christoffel symbol of the second kind as
Tensor-Analysis-for-Engineers_2E_CH10.indd 44
10/3/2020 10:23:01 AM
Gradient Vector Operator—Christoffel Symbols • 45
Γ rrr r r Γ = Γθ r Γ rzr Γθ θ θrr Γ = Γθ r Γθzr Γz z zrr Γ = Γθ r Γ zzr
r Γθθ
Γ rrθ r Γθθ Γ rzθ
Γθrθ θ Γθθ
Γ rrz 0 0 0 Γθr z = 0 − r 0 Γ rzz 0 0 0
Γθzθ
Γθrz 0 1 / r 0 Γθθ z = 1 / r 0 0 10.27 Γθzz 0 0 0
Γ rzθ z Γθθ Γ zzθ
Γ rzz 0 0 0 Γθz z = 0 0 0 Γ zzz 0 0 0
h ∂ ( hθ ) 1 ∂ ( hθ ) 1 ∂r ∂r 2 r 1 / r and Γθθ For example, Γ12 = Γ 221 = Γθrθ = = = = − θ2 = −r = −r hθ ∂r r ∂r ∂r hr ∂r h ∂ ( hθ ) ∂r = − θ2 = −r = − r . Recall that the symbols, given by Equation 10.27 are defined hr ∂r ∂r ∂er based on the covariant basis vectors, ei (see Equation 10.5), or = (1 / r ) eθ ∂θ Γθr θ ∂eθ and = ( − r ) er . It is useful to write the Christoffel symbols in terms of ∂θ r Γθθ
the physical components of the vectors involved. This is done by using ∂ ( e (θ ) ) ∂ ( re (θ ) ) Equation 3.8 (i.e. ei = hi e ( i ) ), or = −e ( r ) , = − re ( r ) , hence ∂θ ∂θ r or the Christoffel symbol based on unit vector reads Γˆ θθ = −1 (hatted to distinguish them from the relation given in Equation 10.27). Similarly, for ∂ ( e ( r )) 1 ∂ ( e ( r )) ∂er = (1 / r ) eθ we can write = ( re (θ ) ) , hence = e (θ ), or ∂θ ∂θ r ∂θ Γθr θ 1. the Christoffel symbol based on unit vector reads Γˆ θ = rθ
r We could also calculate the Christoffel symbols Γˆ θrθ and Γˆ θθ , using the unit vectors or the physical components, as given in Example 8.1. Writing cosθ − sin θ these relations in vector form, we have e ( r ) = sin θ , e (θ ) = cosθ , 0 0
Tensor-Analysis-for-Engineers_2E_CH10.indd 45
10/3/2020 10:23:04 AM
46 • Tensor Analysis for Engineers, 2E
0 ∂e ( r ) ∂e (θ ) ∂e ( z ) and e ( z ) = 0 . Therefore, the derivatives are; = = = 0 , ∂r ∂r ∂r 1 − sin θ − cosθ ∂e ( r ) ∂e (θ ) ∂e ( z ) ∂e ( r ) ∂e (θ ) ∂e ( z ) = = cosθ e (θ ) , = −e ( r ) , = 0, = = = 0 − sin θ = ∂θ ∂θ ∂θ ∂z ∂z ∂z 0 0 ∂e ( r ) ∂e (θ ) ∂e ( z ) = = = 0. Therefore, the non-zero Christoffel symbols calculated ∂z ∂z ∂z r based on the unit vectors are Γˆ θrθ = 1 and Γˆ θθ = −1. The final results can be summarized as
Γˆ rrr Γˆ r = Γˆ θr r ˆr Γ zr Γˆ θrr ˆθ ˆθ Γ = Γθ r ˆθ Γ zr Γˆ rrz ˆz ˆz Γ = Γθ r ˆz Γ zr
Γˆ rrθ Γˆ r
θθ
Γˆ rzθ Γˆ θrθ Γˆ θ
θθ
Γˆ θzθ Γˆ rzθ Γˆ z
θθ
Γˆ zzθ
Γˆ rrz 0 Γˆ θr z = 0 Γˆ rzz 0 Γˆ θrz 0 Γˆ θθ z = 1 Γˆ θzz 0
0 0 −1 0 0 0 1 0 0 0 10.28 0 0
Γˆ rzz 0 0 0 Γˆ θz z = 0 0 0 Γˆ zzz 0 0 0
Spherical coordinates From the results of Example 8.2, for a spherical system ( r ,ϕ ,θ ) , (note the order of the coordinates defined here, see Figure 8.3) we can write 2 2 the non-zero terms of the metric tensor as g= h= 1 , gϕϕ= h= r 2 , and rr r ϕ 2 2 2 gθθ= h= r sin ϕ . Therefore, the contravariant metric tensor reads g rr = 1 , θ ϕϕ 2 g = 1 / r , and gθθ = 1 / ( r 2 sin 2 ϕ ) . Using Equations 10.20 and 10.25 and considering that indices i, j, k, m are various permutations of spherical variables r,ϕ ,θ we get the Christoffel symbol of the second kind as
Tensor-Analysis-for-Engineers_2E_CH10.indd 46
10/3/2020 10:23:07 AM
Gradient Vector Operator—Christoffel Symbols • 47
Γ rrr r r Γ = Γϕ r Γθr r Γϕ ϕ ϕrr Γ = Γϕ r Γθϕr Γθrr Γθ = Γθϕ r Γθθ r
Γ rrϕ r Γϕϕ r Γθϕ
Γϕrϕ Γϕϕϕ ϕ Γθϕ Γθrϕ Γθϕϕ Γθθϕ
ϕ For example, Γ 233 = Γθθ = −
Recall
that
the
Γ rrθ 0 0 0 r Γϕθ = 0 − r 0 r 2 Γθθ 0 0 − r sin ϕ 0 Γϕrθ 0 1 / r 0 0 10.29 Γϕϕθ = 1 / r ϕ sin 2ϕ Γθθ 0 0 − 2 Γθrθ 0 0 1/r θ Γϕθ = 0 0 cot ϕ Γθθθ 1 / r cot ϕ 0 h3 ∂h3 r sin ϕ ∂ ( r sin ϕ ) sin 2ϕ . = − 2 = − sin ϕ cos ϕ = − 2 2 ∂ϕ h2 ∂x r 2
symbols are defined based on the covari ∂eθ sin 2ϕ ant basis vectors, ei (see Equation 10.5), or = − eϕ . ∂θ 2 We can write this ∂ ( r sin ϕ e (θ ) ) as = ∂θ
ϕ Γθθ
relation in terms of the physical components, ∂ ( e (θ ) ) sin 2ϕ = ( − cos ϕ ) e (ϕ ) . or − re (ϕ ) , ∂θ 2 ϕ Γˆ θθ
Similarly, the other non-zero symbols can be written as; hϕ ∂hϕ ∂ ( r sin ϕ ) h ∂h ∂r r r = − θ2 θ = − r sin ϕ = − r sin 2 ϕ, Γϕϕ = − 2 = −r = − r , Γθθ hr ∂r hr ∂r ∂r ∂r 1 ∂hϕ 1 ∂r 1 1 ∂hθ 1 ∂ ( r sin ϕ ) 1 Γθθ r = = =, Γϕϕ r = Γϕrϕ = = = , Γθrθ = hϕ ∂r r ∂r r hθ ∂r r sin ϕ r ∂r ∂ r sin ϕ ( ) ∂ h 1 1 θ θ and Γθϕθ = physical Γθϕ = = = cot ϕ . In terms of hθ ∂ϕ r sin ϕ ∂ϕ ∂eϕ components of the corresponding vectors, we have = ( − r ) er or ∂ϕ Γϕr ϕ ∂ ( re(ϕ ) ) ∂ e r θ ˆ = − re ( r ) = − e ( r ), hence Γϕϕ = −1 . Also, = ( − r sin 2 ϕ ) er ∂ϕ ∂θ r Γθθ
Tensor-Analysis-for-Engineers_2E_CH10.indd 47
10/3/2020 10:23:10 AM
48 • Tensor Analysis for Engineers, 2E ∂ ( r sin ϕ e(θ ) ) r or = − sin ϕ. = − r sin 2 ϕ e ( r ) = − sin ϕ ) e ( r ), hence Γˆ θθ ( ∂θ r ˆ Γθθ ∂e ( r ) ∂er / r ) re (ϕ ) e (ϕ ) , hence Γˆ ϕrϕ = 1. Also, Also, or = (1 / r ) eϕ= (1= ∂ϕ ∂ϕ Γϕrϕ ∂e ( r ) ∂er == 1 / r ) eθ or / r ) ( r sin ϕ e (θ ) ) sin ϕ e (θ ) , hence Γˆ θrθ = sin ϕ. ( (1= ∂θ ∂θ θ Γˆ θrθ Γrθ ∂ ( re (ϕ ) ) ∂eϕ ϕ ( r sin ϕ e (θ ) ) cos Also, = ( cot ϕ )= eθ or cot = ϕ e (θ ) , hence ∂θ ∂ θ Γˆ θϕθ Γθϕθ ˆΓθ = cos ϕ . ϕθ The Christoffel symbols can be obtained directly using the unit vectors as well. From the results of Example 8.1, writing the physical components relasin ϕ cosθ cos ϕ cosθ tions in vector form, we have e ( r ) = sin ϕ sin θ , e (ϕ ) = cos ϕ sin θ , and cos ϕ − sin ϕ − sin θ ∂e ( r ) ∂e (ϕ ) ∂e (θ ) e (θ ) = cosθ . Therefore, the derivatives are; = = = 0 , ∂r ∂r ∂r 0 − sin ϕ cosθ cos ϕ cosθ ∂e (ϕ ) ∂e (θ ) ∂e ( r ) = −e ( r ) , = 0, = = cos ϕ sin θ e (ϕ ) , − sin ϕ sin θ = ∂ϕ ∂ϕ ∂ϕ − cos ϕ − sin ϕ − sin ϕ sin θ − cos ϕ sin θ ∂e ( r ) ∂e (ϕ ) = = sin ϕ cosθ e (θ ) sin ϕ= , cos ϕ cosθ e (θ ) cos ϕ , = ∂θ ∂θ 0 0 − cosθ ∂e (θ ) = − e ( r ) sin ϕ − e (ϕ ) cos ϕ . Therefore, the non-zero − sin θ = ∂θ 0 Christoffel symbols calculated based on the unit vectors are
Tensor-Analysis-for-Engineers_2E_CH10.indd 48
10/3/2020 10:23:13 AM
Gradient Vector Operator—Christoffel Symbols • 49
Γˆ rrr Γˆ rrϕ Γˆ rrθ 0 0 0 r r r r ˆ ˆ ˆ ˆ = Γ Γϕ r Γϕϕ Γ= 0 −1 0 ϕθ ˆr ˆr ˆr Γθ r Γθϕ Γθθ 0 0 − sin ϕ Γˆ ϕrr Γˆ ϕrϕ Γˆ ϕrθ 0 1 0 ˆϕ ˆϕ ϕ ϕ ˆ ˆ = Γϕ r Γϕϕ Γϕθ 1 0 0 = Γ 10.30 ˆϕ ϕ ϕ ˆ ˆ − ϕ 0 0 cos Γθ r Γθϕ Γθθ Γˆ θrr Γˆ θrϕ Γˆ θrθ 0 0 sin ϕ ˆθ ˆθ θ θ ˆ ˆ 0 cos ϕ Γ = Γϕ r Γϕϕ Γϕθ = 0 ˆθ ˆθ ˆθ 0 Γθ r Γθϕ Γθθ sin ϕ cos ϕ Readers can use the Christoffel symbols of the 2nd kind calculated in this example and further calculate the Christoffel symbols of the 1st kind using Equation 10.19.
Tensor-Analysis-for-Engineers_2E_CH10.indd 49
10/3/2020 10:23:13 AM
Tensor-Analysis-for-Engineers_2E_CH10.indd 50
10/3/2020 10:23:13 AM
CHAPTER
11
DERIVATIVE FORMS—CURL, DIVERGENCE, LAPLACIAN In the governing equations for physical phenomena we usually have terms which contain various forms of gradient operator, including the gradient vector itself—for example, when the gradient vector operates as cross product or dot product with another tensor quantity. The cross-product is called the curl, ∇ × A and the dot-product is the divergence, ∇ ⋅ A. For example, in fluid mechanics the curl of the velocity vector is the vorticity vector and divergence of velocity vector vanishes, for incompressible fluids. As mentioned, one of the objectives of tensor analysis is to provide a tool to write down the governing equations in coordinate-independent forms while the covariancy or contravariancy of tensor quantities are correctly implemented in these equations. In the following sections, we derive the coordinate-independent relations for curl, divergence, Laplacian, and biharmonic operators of tensors for an arbitrary coordinate system.
11.1 CURL OPERATIONS ON TENSORS To form the curl operator expression, we first consider it for a vector in 3D coordinate space and then extend it to higher order dimension. In Section 9, we gave a list of guidelines for forming expressions for t ensors. Following these guidelines and using Equation 9.6, we can write the covariant component of the curl of a vector as Ci = ∇ × A = ijk ∇ j A k . We
(
Tensor-Analysis-for-Engineers_2E_CH11.indd 51
)
i
10/3/2020 10:24:00 AM
52 • Tensor Analysis for Engineers, 2E
know that ijk is a tensor (see Section 9). But the expression ∇ j A k does not transform like a tensor, due to extra terms appearing when transformed to another arbitrary coordinate system [4], [7]. However, if we use the results obtained in Section 10 and replace ∇ j A k with the tensor A ,k ⋅ j (i.e., the contravariant derivative of the contravariant component) then we get the proper expression as (see Equation 10.8) Ci = ijk A k, ⋅ j 11.1
Similarly, the contravariant component of the curl is written as, using Equations 9.6 and 10.12, C i = ijk Ak, j 11.2
Both expressions given by Equations 11.1 and 11.2 are tensors, since terms are involved are tensors as well as both reduce to the Cartesian forms when ∂A written in the Cartesian coordinate system (i.e., Ci = eijk kj because the ∂y covariant and contravariant components are identical and Jacobian is unity). To have a more practical and explicit relation for the curl, we write Equation ∂A ijk kj − ijk Γ njk An examining the last term, C i ijk A= 11.2 in detail or = k, j ∂x on the R.H.S, using the symmetry property of the Christoffel symbol (i.e., Γ njk = Γ kjn ) and the anti-symmetry of the permutation symbol (i.e., ijk = − ikj) yields, ijk Γ njk An = − ikj Γ nkj An . But after interchanging j ↔ k , dummy indices, only in the term on the R.H.S of the latter expression we receive − ikj Γ nkj An = − ijk Γ njk An. After comparing this result with the original expression (i.e., − ijk Γ njk An = ijk Γ njk An) we can conclude that ijk Γ njk An = 0. Therefore, Equation 11.2 can be written as Ci =
(∇× A)
i
=
eijk ∂Ak 11.3 ∂x j
The curl vector is then written as C = C i ei .
Similarly, we write Equation 11.1 in detail, also using Equation 10.8, or ∂A k k Ci = ijk A ,k ⋅ j = ijk g jn n + ijk g jnΓ nm A m . Examining the last term on the ∂x R.H.S, it turns out that it doesn’t vanish. Therefore, we have
Tensor-Analysis-for-Engineers_2E_CH11.indd 52
10/3/2020 10:24:03 AM
Derivative Forms—Curl, Divergence, Laplacian • 53
Ci =
(∇× A)
i
∂A k k = g jn eijk n + Γ nm A m 11.4 ∂x
Note that the curl operation result is a tensor of the same rank as the original quantities. As in our example C is a vector, like ∇ and A . Now, we extend the discussion and find the relation for tensors of higher rank. For simplicity of writing the expressions, without losing generality, we consider a mixed tensor of second rank A ij . The invariant quantity is A = A ij ei e j and the curl reads ∇ × A = ∇ × A ij ei e j . The gradient operator just differentiates all the terms in front of it but the vector part of ∇ can form a cross-product with either ei or e j, both are possible and legitimate operations. One can form, then two forms of the curl of the quantity A . In general, we can have N number of forms for the curl of a tensor of order N. For our example here, we have two forms:
(
)
1. We consider the case for which the cross-product operation occurring with ei . Therefore, the contravariant component of A ij should be involved with the permutation tensor indices. write, using Equation 11.1, We can the component of curl of A as ∇ × ei A ij e j . Careful attention should be given to pick up the right indices to form the curl expression. Since we picked the ei for cross-product operation with gradient, the covariant permutation tensor should be used, say kmn . Therefore, the index m corresponds to the contravariant differentiation and index n summed up with the contravariant component of the tensor A . The index k , is then the covariant component of the result or the curl tensor. The rank of the result should be the same as the original quantity A ij (i.e., rank two), hence the free covariant index j remains intact and we end up with a doubly-covariant tensor, or
(
)
∂A nj = Ckj kmn = A nj ,m⋅ g mp ekmn p + Γ npq A qj 11.5 ∂x
2. We consider the case for which the cross-product operation occurs with e j . Therefore, the covariant component of A ij should be involved with the permutation tensor. We can write, using Equation 11.1, the compo nent of curl of A as ∇ × e j A ij ei . Careful attentions should be given to pick up the right indices to form the curl expression. Since we picked the e j for cross-product operation with gradient, the contravariant
(
Tensor-Analysis-for-Engineers_2E_CH11.indd 53
)
10/3/2020 10:24:06 AM
54 • Tensor Analysis for Engineers, 2E
permutation tensor should be used, say kmn . Therefore, the index m corresponds to the covariant differentiation and index n is summed up with the covariant component of the tensor A . The index k , is then the contravariant component of the result or the curl. But the rank of the result should be the same as the original quantity A ij (i.e., rank two), hence the free contravariant index i remains intact and we end up with a doubly contravariant tensor, or kmn i = C ki = An,m
ekmn ∂Ani 11.6 ∂x m
A similar operation can be used to write down the curl of a tensor of rank N, which has N possible outcomes.
11.2 PHYSICAL COMPONENTS OF THE CURL OF TENSORS—3D ORTHOGONAL SYSTEMS In many applications, we consider orthogonal systems (i.e., gij = 0 , i ≠ j ). In this section, we derive expressions for the physical components of the curl of a vector A in 3D orthogonal systems. Recalling from previous sections i i A A= ei A= A ( i ) e ( i ) and hi = ei . (see Sections 3 and 8), we can write= ie j δ i j, for orthogonal systems we have ei ⋅ e i = Using ei ⋅ e = 1 which means that the covariant basis vector ei and contravariant basis vector e i both point in the same direction. For example, e 1 and e1 are along the same line and point in the same direction. Therefore, e i = 1 / hi 11.7
Substituting ei = hi e ( i ) in A ( i ) e ( i ) = A i ei and e i = e ( i ) / hi in A ( i ) e ( i ) = Ai e i gives i A= A i / hi ( i ) h= iA
no sum on i 11.8
1 2 3 Or A= A1 / h1 , A= A2 / h2 , A= A3 / h3 . (1) h= ( 2 ) h= ( 3 ) h= 1A 2A 3A
Tensor-Analysis-for-Engineers_2E_CH11.indd 54
10/3/2020 10:24:09 AM
Derivative Forms—Curl, Divergence, Laplacian • 55
Now we would like to write the physical components of the curl of vector A . To i eijk ∂Ak i do this we multiply Equation 11.3 by hi , or ∇ × A = hi h= hi iC . ∂x j But hiC i = C ( i ) and Ak = hk A ( k ) (see Equation 11.8). Therefore, we will get the physical component C ( i ) of the curl of vector A in terms of its physical component A ( k ) , as
(
C ( i ) = hi
eijk ∂ ( hk A ( k ) )
∂x j
)
no sum on i, orthogonal 11.9
Or, using = h1 h2 h3 for orthogonal systems, we have
(1 ) ( ∇ × A )=
(
1 ∂ ( h3 A ( 3 ) ) ∂ ( h2 A ( 2 ) ) − h2 h3 ∂x2 ∂x3
∂ ( h1 A (1 ) ) ∂ ( h3 A ( 3 ) ) − 11.10 ∂x3 ∂x1 1 ∂ ( h2 A ( 2 ) ) ∂ ( h1 A (1 ) ) ∇× A = − ( 3 ) C= (3) h1 h2 ∂x1 ∂x2
(∇× A= )(2)
C= (1 ) C= (2)
1 h1 h3
)
It can be concluded that for a Cartesian system we recover the familiar ∂Az ∂Ay ∂AX ∂AZ expressions, or ∇ × A = C= , ∇ × A = C= − − , 1 2 1 2 ∂X ∂Z ∂Y ∂Z ∂AY ∂AX h= h= 1. and ∇ × A = C= − 1 2 3 3 , since h= 3 ∂Y ∂X
(
(
)
(
)
)
11.3 DIVERGENCE OPERATION ON TENSORS To form the divergence operator expression, we first consider it for a vector in 3D coordinate space and then extend the results to higher order dimension. In Section 9, we gave a list of guidelines for forming expressions. Following these guidelines and using Equation 9.2, we can write the divergence (using the covariant component of the gradient) as
Tensor-Analysis-for-Engineers_2E_CH11.indd 55
∇ ⋅ A = A ,ii 11.11
10/3/2020 10:24:11 AM
56 • Tensor Analysis for Engineers, 2E which reduces to proper expression in Cartesian coordinates and shows that the divergence of a vector results in a scalar, or in general divergence operation reduces the rank of the tensor, on which it is operating, by one. To further simplify Equation 11.11, we write it explicitly (see Equation 10.6) as ∂A i i A = + Γ ini A n and plug in for Γ iin from Equation 10.26 (after interchanging ,i i ∂x ∂A i A i ∂ dummy indices i ↔ n), to get = + . After substituting for A,i i A ,ii ∂x i ∂x i into Equation 11.11 and performing some manipulations, we receive 1 ∂ ( A i ) ∇⋅A = 11.12 ∂x i
In addition, Equation 11.11 can be written inij terms of the covariant comij ponent of the vector using g , or ∇ ⋅ A = g A j , after expanding gives ,i ij ij = ∇ ⋅ A g ij A j + A= g g A or after interchanging i ↔ j and using j j, i ,i , i
( )
(
( )
)
=0
the symmetry of metric tensor, we receive ∇ ⋅ A = g ij Ai, j 11.13
Now we extend the discussion by considering a tensor of higher rank, for example and without losing generality a tensor of second rank A = A ij ei e j. The divergence of A is ∇ ⋅ A = ∇ ⋅ A ij ei e j in which we have choices between ei or e j for the gradient vector performing a dot-product operation. Therefore, we have two cases for this example: 1. ∇ dot products with ei: In this case we get ∇ ⋅ A = ∇ ⋅ ei A ij e j . Note that the gradient differentiates all terms in front of it but as a vector onlydotted with ei . In other words, the contravariant index i is involved in the dot-product operation. Using Equation 11.11 we can write ∇ ⋅ A = A ij, i which is a tensor of rank one, as expected since divergence reduces the rank of A ij by one. But using Equation 11.12, we have ∂A ij i i + Γ im A= A mj − Γ mij Ami in which the first two terms on the R.H.S. j, i ∂x i i ∂A ij 1 ∂ A j i m gives , using Equation 11.12. Finally, we get + Γ im A j = ∂x i ∂x i the result as
(
)
(
(
Tensor-Analysis-for-Engineers_2E_CH11.indd 56
)
)
(
)
i 1 ∂ A j i j ∇ ⋅ A = ∇ ⋅ ei A j e = − Γ mij Ami 11.14 ∂x i
(
)
10/3/2020 10:24:14 AM
Derivative Forms—Curl, Divergence, Laplacian • 57
2. ∇ dot products with e j : In this case we get ∇ ⋅ A = ∇ ⋅ e j A ij ei . Note that the gradient differentiates all terms in front of it but as a vector only dotted with e j . In other words, the index j is involved in the dot- product operation. Using Equation 11.11 we can write ∇ ⋅ A = g kj Aki , j which is a tensor of rank one, as expected since divergence reduces the rank of A ij by one. Therefore, we have
(
)
∇ ⋅ A = ∇ ⋅ e j A ij ei = Aki , . k 11.15
(
)
Using similar operations, we can write the expression for divergence of a tensor of rank N . Both relations given by Equations 11.14 and 11.15 reduce to familiar forms for divergence in a Cartesian coordinate system. For ∂Ax ∂Ay ∂Az example, for divergence of vector A we get ∇ ⋅= . A A= + + i, i ∂x ∂y ∂z
11.4 LAPLACIAN OPERATIONS ON TENSORS A Laplacian operator is the result of a gradient operator forming a dot-product with itself, or divergence of gradient, ∇ 2 = ∇ ⋅ ∇ . Performing Laplacian on i a scalar ψ and using Equation 11.12, we can write ∇ 2ψ = ∇ ⋅ ∇ψ = ( ∇ψ ), i , i ∂ψ 1 ∂ ( ∇ ψ ) 2 g ij j . Therefore, or ∇ ψ = where ∇ iψ = i ∂x ∂x
(
)
(
)
1 ∂ ∂ψ ∇ 2ψ = i g ij j 11.16 ∂x ∂x
Since ∇ 2 is a tensor of rank zero, then ∇ 2ψ is a scalar, as well. In general, the rank is determined by the quantity that Laplacian is operating on. We the discussion to find the Laplacian of a vector, for e xample extend A = A i ei . Since for Laplacian the dot-product operation is performed between the two gradients involved, we don’t have many choices, as was the case for the divergence operator, and hence the formulation is more definite in terms of the results. Using Equations 11.13 and 11.16, we can write the contravariant component of the gradient of vector A as g nm A,in and the components of the Laplacian as g nm A,in . Note that index m indicates the dot,m product operation between the contravariant component of the gradient of A and the covariant derivative of the second gradient involved. Expanding nm i i nm the last expression, we receive g nm A = g,nm A,inm . But g,m = 0 , ,n m A, n + g
(
)
(
Tensor-Analysis-for-Engineers_2E_CH11.indd 57
)
,m
10/3/2020 10:24:17 AM
58 • Tensor Analysis for Engineers, 2E since it is a tensor and its value is zero in the Cartesian coordinate system, hence it should be zero for an arbitrary system, as well. Therefore, we have
A ) (g ( ∇= 2
i
nm
i A= g nm A,inm 11.17 , n ), m
The term A,inm which is the second covariant derivative is a new term, so ∂ ( A,in ) j i i i far. By expanding this term, we get= = + Γ mj A,nm ( A A,nj − Γ mn A,i j. , n ), m m ∂x ∂ ( Ai ) i i i k But = A,n A + Γ nk A and after substituting and rearranging similar ∂x n terms, we receive = A,inm
i k i ∂2 Ai k ∂Γ nk i ∂A i ∂A + A + Γ + Γ nk mj ∂x m ∂x n ∂x m ∂x m ∂x j 11.18 i j ∂A j j i k i k − Γ mn j + Γ mjΓ nk A − Γ mnΓ jk A ∂x
Similar expressions can be written for higher order tensors, using hints from Equation 11.18 for writing the terms involving Christoffel symbols A with appropriate signs. The higher the rank of the more complicated the 2 related expression for ∇ A becomes. For an orthogonal coordinate system, Equation 11.17 simplifies, since g nm = 0 for m ≠ n , or (for N = 3 ) ∇ 2 A =i g11 A,i11 + g 22 A,i22 + g 33 A,i33 ( orthogonal )
(
)
11.5 BIHARMONIC OPERATIONS ON TENSORS A biharmonic operator is the result of a Laplacian operator operating on itself, or ∇ 4 =∇ 2 ( ∇ 2 ) . Performing the operation on a scalar ψ and using Equation 11.12, we can write
1 ∂ ∂ 1 ∂ ij ∂ψ ∇ 4ψ = k g kl l g 11.19 ∂x ∂x i ∂x j ∂x ∇ 2ψ
Tensor-Analysis-for-Engineers_2E_CH11.indd 58
10/3/2020 10:24:19 AM
Derivative Forms—Curl, Divergence, Laplacian • 59
When a biharmonic operates on a vector, thei result is a vector as well. For example, for A = A i ei , we can write ∇ 4 A = B ei. Since the result is a vector we can use Equation 11.17 to write Bi = g lm g jk A,i jklm 11.20
The quantity A,i jklm is a lengthy expression and can be written in detail [2], [4], [7] using hints taken from Equation 11.18. In principle, one can extend the 2 2 2 ∇ ∇ discussion to find operators like ∇ 2 n =∇ . In practice we encounter n times
mostly up to the level of the biharmonic operator in governing equations.
11.6 PHYSICAL COMPONENTS OF THE LAPLACIAN OF A VECTOR—3D ORTHOGONAL SYSTEMS In this section, we derive expressions for physical components of the Laplacian of a vector. Relations for physical components are useful since they are the magnitude of the components using unit vectors as the scale of measurement (see Section 6). We start with Equation 11.18 and write it for a vector, like A = A ( i ) e ( i ) 1 ∂ ∂A or ∇ 2 A = i g ij j . In an orthogonal system, we have g ij = 0 (when ∂x ∂x ii 2 i ≠ j ) and g = 1 / hi . Therefore, after substitution into the expression, we 1 ∂ ∂ A ( i ) e ( i ) 2 will have ∇ A = i 2 . Performing the differentiations ∂x hi ∂x i along with using related relations for Christoffel symbols and after some manipulations, we get the physical j -component as [7] ∇2 A
( j)
1 ∂h j ∂A ( i ) 1 ∂hi ∂A ( i ) = ∇2 A ( j ) + 2 2 i − j h j hi2 ∂x j ∂x i hi h j ∂x ∂x A ( i ) ∂h j ∂h j A ( i ) ∂hi ∂hi A ( j ) ∂h j ∂h j A ( k ) ∂hi ∂hi + + − 2 2 i i + 11.21 j j i i 3 3 h j hi ∂x ∂x hi h j ∂x ∂x hk h j hi2 ∂x k ∂x j hi h j ∂x ∂x A ( i ) ∂ ∂h j A ( i ) ∂ ∂h i + − j j i i 2 2 ∂x hi h j ∂x ∂x h j hi ∂x
Tensor-Analysis-for-Engineers_2E_CH11.indd 59
10/3/2020 10:24:20 AM
60 • Tensor Analysis for Engineers, 2E
In Equation 11.21, summation is done on indices i and k but not on index j, since it is the free index. To show the application of Equation 11.21, we present two examples to calculate the components of Laplacian of a vector in cylindrical and spherical coordinate systems. Note that the Laplacian of a vector is a vector as well, which has three components. 11.6.1 Example: Physical components of the Laplacian of a vector—cylindrical systems
Consider a cylindrical coordinate system ( x1 , x2 , x3 ) ≡ ( r ,θ , z ) where r is the radial distance, θ the azimuthal angle, and z the vertical coordinate, w.r.t. the Cartesian coordinates ( y1 , y2 , y3 ) ≡ ( X , Y, Z ). The functional relations are X = r cosθ Y = r sin θ . Find the Laplacian expressions for a scalar and the physical Z = z components of a vector in this system. Let A (1 ) = Ar , A ( 2 ) = Aθ, and A ( 3 ) = Az . Solution:
∂y To find the covariant basis vectors, we use ei = i E j . Therefore, we get ∂x ∂X ∂Y ∂Z ∂X ∂Y e1 = E1 + E2 + E3 = cosθ E1 + sin θ E2. Similarly, e2 = E1 + E2 + ∂r ∂r ∂r ∂θ ∂θ ∂Z ∂X ∂Y ∂Z E3 = − r sin θ E1 + r cosθ E2. And e3 = E1 + E2 + E3 = E3. Therefore, ∂θ ∂z ∂z ∂z we have er cosθ E1 + sin θ E2 = − r sin θ E1 + r cosθ E2 eθ = ez = E3 The scale factors are the magnitudes of the basis vectors. Hence h1 = hr = er ⋅ er = 1, h2 = hθ = eθ ⋅ eθ = r , and h3 = hz = ez ⋅ ez = 1 . The e unit vectors e ( i ) = i , are hi e ( r ) cosθ E1 + sin θ E2 = − sin θ E1 + cosθ E2 e (θ ) = e (ϕ ) = E3
Tensor-Analysis-for-Engineers_2E_CH11.indd 60
j
10/3/2020 10:24:23 AM
Derivative Forms—Curl, Divergence, Laplacian • 61
0 0 1 2 The Jacobin = is h= r and g = h , or g = 0 1 / r 0 . Now r hθ hz 0 0 1 1 ∂ ∂ψ using Equation 11.16 for orthogonal system i.e. ∇ 2ψ = i g ii i , ∂x ∂x ij since = g 0 for i ≠ j , and substituting for corresponding values we will get ii
−2 i
ij
1 ∂ ∂ψ 1 ∂ 2ψ ∂ 2ψ . In other words, the Laplacian operator + r + r ∂r ∂r r 2 ∂θ 2 ∂z2 in cylindrical coordinate system reads
= ∇ 2ψ
= ∇2
1 ∂ ∂ 1 ∂ 2 ∂ 2 11.22 r + + r ∂r ∂r r 2 ∂θ 2 ∂z2
Similarly, using Equation 11.21 we get the physical components of the Laplacian for the vector A, as A 2 ∂A 2 ∇ 2 Ar − 2 θ − 2r ∇ A r = r ∂θ r A ∂ A 2 2 r ∇ 2 Aθ + 2 − 2θ 11.23 ∇ A θ = r ∂θ r 2 2 ∇ A = ∇ Az z
Where the operator ∇ 2 is given by Equation 11.22 and Ar, Aθ , Az represent the physical components of A. 11.6.2 Example: Physical components of the Laplacian of a vector—spherical systems
Consider a spherical coordinate system ( x1 , x2 , x3 ) ≡ ( r , ϕ , θ ) where r is the radial distance, ϕ the polar and θ the azimuthal angle, w.r.t. the Cartesian coordinates ( y1 , y2 , y3 ) ≡ ( X , Y, Z ) . The functional relations are X = r sin ϕ cosθ Y = r sin ϕ sin θ . Find the Laplacian expressions for a scalar and the Z = r cos ϕ physical components of a vector in this system. Let A (1 ) = Ar, A ( 2 ) = Aϕ, and A ( 3 ) = Aθ.
Tensor-Analysis-for-Engineers_2E_CH11.indd 61
10/3/2020 10:24:25 AM
62 • Tensor Analysis for Engineers, 2E Solution:
j ∂y To find the covariant basis vectors, we use ei = i E j . Therefore, we get ∂x ∂X ∂Y ∂Z e1 = E1 + E2 + E3 = sin ϕ cosθ E1 + sin ϕ sin θ E2 + cos ϕ E3 . Similarly, ∂r ∂r ∂r ∂Y ∂Z ∂X e2 = E1 + E2 + E3 = r cos ϕ cosθ E1 + r cos ϕ sin θ E2 − r sin ϕ E3 . And ∂ϕ ∂ϕ ∂ϕ ∂X ∂Y ∂Z − r sin ϕ sin θ E1 + r sin ϕ cosθ E2 . Therefore, we e3 = E1 + E2 + E3 = ∂θ ∂θ ∂θ have er = sin ϕ cosθ E1 + sin ϕ sin θ E2 + cos ϕ E3 eϕ = r cos ϕ cosθ E1 + r cos ϕ sin θ E2 − r sin ϕ E3 e r ϕ θ E r ϕ θ E sin sin sin cos = − + 1 2 θ
The scale factors are the magnitudes of the basis vectors. Hence h1 = hr = er ⋅ er = 1 , h2 = hϕ = eϕ ⋅ eϕ = r , and h3 = hθ = eθ ⋅ eθ = r sin ϕ . e The unit vectors e ( i ) = i , are hi e ( r ) = sin ϕ cosθ E1 + sin ϕ sin θ E2 + cos ϕ E3 e (ϕ ) = cos ϕ cosθ E1 + cos ϕ sin θ E2 − sin ϕ E3 − sin θ E1 + cosθ E2 e (ϕ ) = 1 0 0 r 2 sin ϕ and g ii = hi−2 , or g ij = 0 1 / r 2 The Jacobin= is h= 0 r hϕ hθ 2 2 1 0 1 / ( r sin ϕ ) 0 0 0 g ij = 0 1 / r 2 0 . Now using Equation 11.16 for orthogonal 2 2 0 1 / ( r sin ϕ ) 0 1 ∂ ∂ψ system i.e. ∇ 2ψ = i g ii i , since = g ij 0 for i ≠ j, and substituting ∂x ∂x 1 ∂ 2 ∂ψ 1 ∂ for corresponding values we will = get ∇ 2ψ r + 2 2 r ∂r ∂r r sin ϕ ∂ϕ ∂ψ 1 ∂ 2ψ . In other words, the Laplacian operator in sin ϕ + ∂ϕ r 2 sin 2 ϕ ∂θ 2 spherical coordinate system reads
Tensor-Analysis-for-Engineers_2E_CH11.indd 62
10/3/2020 10:24:28 AM
Derivative Forms—Curl, Divergence, Laplacian • 63
= ∇2
1 ∂ 2 ∂ 1 ∂ ∂ 1 ∂2 11.24 r + 2 sin ϕ + 2 2 2 r ∂r ∂r r sin ϕ ∂ϕ ∂ϕ r sin ϕ ∂θ 2
Similarly, using Equation11.21 we receive the physical components of the Laplacian for the vector A, as
2 2A 2 ∂(sin ϕ Aϕ ) 2 ∂ Aθ ∇ 2 Ar − 2 r − 2 − 2 ∇ A r = ∂ϕ r r sin ϕ r sin ϕ ∂θ A 2 2 ∂A 2 cos ϕ ∂Aθ 11.25 ∇ 2 Aϕ − 2 ϕ 2 + 2 r − 2 2 ∇ A ϕ = ∂ ∂ r r r ϕ ϕ ϕ θ sin sin 2 A 2 ∂Ar 2 cos ϕ ∂ Aϕ ∇ 2 Aθ − 2 θ 2 + 2 + 2 2 ∇ A θ = r sin ϕ r sin ϕ ∂θ r sin ϕ ∂θ Where the operator ∇ 2 is given by Equation 11.24 and Ar , Aϕ , Aθ represent the physical components of A. In the next section, we focus on relations pertinent to tensor transformations between merely Cartesian coordinate systems.
Tensor-Analysis-for-Engineers_2E_CH11.indd 63
10/3/2020 10:24:29 AM
Tensor-Analysis-for-Engineers_2E_CH11.indd 64
10/3/2020 10:24:29 AM
CHAPTER
12
CARTESIAN TENSOR TRANSFORMATION— ROTATIONS When transformation of tensors is performed from one Cartesian coordinate system to another one, relations derived in previous sections take simpler forms, called Cartesian tensors. As mentioned previously, a Cartesian system consists of three mutually perpendicular flat surfaces for which; the basis vectors are unit vectors, there is no distinction between covariant and contravariant components, and the components of a tensor are the physical components. Therefore, we use subscript indices for all types of components, regardless. We will consider two Cartesian systems, yi and y′i with a common origin— otherwise we can always redefine them to have a common origin with transforming, say, the origin of y′i to that of yi . Following the discussion from Section 3, we can write the differential displacement ds in these systems, as
= ds dy = dy′j E′j 12.1 i Ei
where Ei and E′j are the corresponding unit vectors. However, we ∂y have dyi = i dy′j , and after substituting in Equation 12.1 we get ∂y′j
Tensor-Analysis-for-Engineers_2E_CH12.indd 65
10/3/2020 10:25:12 AM
66 • Tensor Analysis for Engineers, 2E
∂yi ∂yi Ei = E′j or, 0 . Since dy′j is arbitrary, we conclude Ei − E′j dy′j = ∂y′j ∂y′j by i ↔ j , we have
∂y j E′i = E j 12.2 ∂y′i Now performing dot-product on both sides of Equation 12.2 with Ek gives, ∂y j ∂yk . Hence, Ek ⋅= E′i E j ⋅= Ek ∂y′i ∂y′i δ jk
∂y Ek ⋅ Ei′ =k 12.3 ∂y′i ∂y This means that the cosine of the angle between yk and y′i is equal to k , ∂y′i called the cosine direction. ∂y′ Similarly, we can write Ek = i E′i , and after dot-product with E′j we receive ∂yk ′ ∂ y ′ ∂yi j ′j . Hence, interchanging i ↔ j, we receive E′j ⋅ E= E′i ⋅ E= k ∂yk ∂yk δ ij
∂y′ E′i ⋅ Ek =i 12.4 ∂yk ∂y′ This means that the cosine of the angle between yk and y′i is equal to i . ∂yk Therefore, comparing Equations 12.3 and 12.4 and considering the commutativity of the dot-product operation, we will have ∂y′i ∂yk 12.5 = ∂yk ∂y′i
∂y j ∂y′j ∂yi ∂y′i ∂y′j = dy′j and = E′i = Ej E j , but ∂y′j ∂yi ∂y′i ∂y j only in Cartesian coordinate systems. In other words, these relations prove that there is no distinction between covariant and contravariant components in a Cartesian system (see Equations 4.8 and 5.4, for comparison).
Equation 12.5 yields dyi =
Tensor-Analysis-for-Engineers_2E_CH12.indd 66
10/3/2020 10:25:14 AM
Cartesian Tensor Transformation—Rotations • 67
12.1 ROTATION MATRIX Now if we assume that coordinate system y′i is obtained by rotating coordi nates y j at an angle θ about an axis parallel to a unit vector n = ( n1 , n2 , n3 ) , ∂yi′ = Rij is a function of θ and n and it can be shown then the quantity that [9], it reads as
∂y j
(
)
Rij = ni n j + δ ij − ni n j cosθ + eijk nk sin θ 12.6
Equation 12.6 is Rodrigues’ rotation formula, or the 3D rotation matrix. Readers should note that the Right-Hand-Rule convention applies to the rotation of coordinates with respect to the positive direction of unit vector n when using Equation 12.6. For example, for n = ( 0,0,1 ) and θ = π / 2 , Rij gives the transformation of x − y plane about the positive direction of z -axis for an angle of 90° . ∂y′i = dy′i = dy j Rij dy j which yields, y′i = Rij y j. Note that Now we integrate ∂y j the constant of integration is zero due to having a common origin for both systems, and Rij is a constant for given n and θ. Similarly, we can rotate y′i at an angle (− θ ) about the same unit vector n to recover the original −1 ′ R ji y′j, by having yi R= y j coordinates. Therefore, we can write= ij y j −1 Rij being equal to the inverse or transpose of Rij [4]. Further examining Equation 12.6, we can conclude that Rij for a given angle θ is equal to R ji, or its transpose, calculated for angle ( −θ ). This is the result of the property of the last term on the R.H.S of Equation 12.6 when i ↔ j to get R ji , − e jik nk sin θ =− e jik nk sin( θ ). The remaining terms are not or eijk nk sin θ = affected by this index-change operation.
12.2 EQUIVALENT SINGLE ROTATION: EIGENVALUES AND EIGENVECTORS It would be useful to write the matrix form of the Cartesian coordinate rotations, as well. For example (in a 3D space with N = 3), y′i = Rij y j can be y1′ R11 R12 R13 y1 written as y2′ = R21 R22 R23 y2 , and after using Equation 12.6, for y′ R 3 31 R32 R33 y3 n = ( n1 , n2 , n3 ) and rotation angle θ, we get
Tensor-Analysis-for-Engineers_2E_CH12.indd 67
10/3/2020 10:25:20 AM
68 • Tensor Analysis for Engineers, 2E
y1′ y2′ = y′ 3 cosθ + n1 n1 (1 − cosθ ) n1 n2 (1 − cosθ ) − n3 sin θ n1 n3 (1 − cosθ ) + n2 sin θ
n1 n2 (1 − cosθ ) + n3 sin θ cosθ + n2 n2 (1 − cosθ ) n2 n3 (1 − cosθ ) − n1 sin θ
n1 n3 (1 − cosθ ) − n2 sin θ y1 n2 n3 (1 − cosθ ) + n1 sin θ y2 cosθ + n3 n3 (1 − cosθ ) y3
12.7
From Equation 12.7 (or 12.6) we can conclude that the trace of the rotation matrix (i.e., the sum of diagonal elements, Rii) is equal to 1 + 2 cosθ, (note 2 that ni ni = n = n12 + n22 + n32 = 1, since n is a unit vector), or Rii − 1 12.8 2
θ = cos−1
Equation 12.8 is a useful relation for finding rotation angle θ, once Rij is known/given. However, readers may ask: if Rij is known, how can one find the corresponding unit vector parallel to the axis of rotation? This q uestion is all the more significant considering that we can have a sequence of rotations to arrive at a final desired orientation of the coordinate axes. For example, if we rotate the original system through a sequence of rotations and arrive at the final system orientation, i.e., yi → yit 1 → → yitn → y′i , sequence
then we can apply Equation 12.7 for each sequence and calculate the final rotation matrix. That is {y′i } = Rijtn Rijtn−1 Rijt 1 {yi }. Note that for each Rij
rotation in the sequence, the corresponding rotation matrix is pre-multiplied to the previous ones, and the normal vector parallel to the rotation axis is defined based on the current coordinate system at hand in the sequence (see Example 12.2.1) Now, if we want to replace all intermediate rotations with just one equivalent rotation, Equation 12.8 gives the value of the angle for the desired equivalent single rotation. However, to find the axis of rotation parallel to the unit vector for the equivalent single rotation we need a mathematical procedure/ tool. This tool can be obtained by finding the real eigenvector of the given rotation matrix since, through all rotations, only the eigenvectors remain in the same direction [2], [4]. The mathematical procedure is as follows:
Tensor-Analysis-for-Engineers_2E_CH12.indd 68
10/3/2020 10:25:22 AM
Cartesian Tensor Transformation—Rotations • 69
We consider a normal unit vector ni perpendicular to a plane and write the components of the rotation tensor on this plane, or Rij n j . This quantity is a vector whose components are the elements of the rotation tensor on the plane considered. We identify this quantity as Rij n j = Ti . In general, the direction of the vector Ti does not necessarily coincide with the normal vector to the plane, as shown in Figure 12.1.
FIGURE 12.1 Rotation matrix component T , about an arbitrary plane with normal n.
In other words, we could demand a specific plane and calculate its corresponding normal direction such that vector Ti ends up in the same direction as the normal to the plane considered. Therefore, we can write Ti ni = Rij n j ni. This quantity, which is the projection of Ti in the direction of ni, is a quadratic function of vector ni. To calculate its extremum value with the constraint that ni ni = 1 (which means that having the normal vector as a unit vector), we use the Lagrange multiplier method for extremizing the scalar quantity = Rij n j ni − λ ( ni ni − 1 ), where the constant λ is the Lagrange ∂ multiplier, or the eigenvalue of Rij . This is achieved by letting = 0, ∂ni ∂ = 0 recovers the constraint, ni ni = 1) which leads to (note that the ∂λ
(R
In
matrix
form,
ij
)
− λδ ij n j = 0 12.9
Equation
12.9
can
be
written
as
R12 R13 n1 R11 − λ R22 − λ ε 23 n2 = 0. This system of equations has a non R21 R31 R32 R33 − λ n3 trivial (i.e., non-zero) solution, for ni, if the determinant of the matrix is zero, or R11 − λ R12 R13 det Rij −= λδ ij R21 R22 − λ = ε 23 0 12.10
(
)
R31
Tensor-Analysis-for-Engineers_2E_CH12.indd 69
R32
R33 − λ
10/3/2020 10:25:24 AM
70 • Tensor Analysis for Engineers, 2E
By evaluating the determinant, we get a cubic equation in terms of λ (which has three roots, or eigenvalues) as Rii R jj − Rij R ji 0 12.11 λ − Rij = 2 3
λ 3 − ( Rii ) λ 2 +
1
2
Equation 12.11 is known as the characteristic equation for matrix Rij, where 1 is the first principal scalar invariant, which is equal to the trace of Rij; or 1 = Rii = λ1 + λ2 + λ3 , 2 is the second principal scalar invariant, which is 2 = λ1 λ2 + λ2 λ3 + λ3 λ1 , or the sum of the diagonal minors; and 3 is the third principal scalar invariant, which is equal to the determinant of Rij, or = 3 e= λ1 λ2 λ3 . ijk R1 i R2 j R3 k Equation 12.11 contains several important properties and pieces of information about the system’s rotation matrix. Some examples related to the discussion here include:
For symmetric tensors with real components there exist three real value answers for λ (or λ1 , λ2 , and λ3). This is not the case for Rij, since it is not symmetric, and we get only one real value λ, [10]. The answers for λ are the eigenvalues (or principal values) of Rij. The corresponding ni, calculated and normalized for each eigenvalue, are the eigenvectors (or principal directions) of Rij. A given eigenvector when multiplied by a constant real number (positive or negative) it gives a new vector, but all have the same direction parallel to the principal direction.
The method of finding equivalent rotation is very useful in practice for designing the motion of machine parts in robotic applications, such as a robot arm’s motion. 12.2.1 Example: Equivalent single rotation to sequential rotations of a Cartesian system Having a Cartesian system labeled as yi = ( y1 , y2 , y3 ) and its transformation to another system labeled as y′i = y′1 , y′2 , y′3 such that y′1 coincides with y2 , y2′ with y3 , and y3′ with y1 , as shown in Figure 12.2, find the overall equiva lent single rotation matrix for y′i = Rij y j , the axis of rotation direction n and the angle of rotation θ.
(
Tensor-Analysis-for-Engineers_2E_CH12.indd 70
)
10/3/2020 10:25:27 AM
Cartesian Tensor Transformation—Rotations • 71
FIGURE 12.2 Sequential rotations of a Cartesian coordinate system (left) and the equivalent single rotation (right).
Solution: We consider two sequential rotations: 1) rotation about y3 -axis with normal vector ( 0,0,1 ) with an angle of 90° (R.H.R. applies) to get the intermediate transformation ( y1 , y2 , y3 ) → ( y1′ , y2t , y3 ) , and 2) rotation about new y′1 with normal vector (1,0,0 ) with an angle of 90° (R.H.R. applies) to get the second transformation ( y1′ , y2t , y3 ) → ( y1′ , y′2 , y′3 ) (see Figure 12.2). Therefore, using Equation 12.7 and the data given for the first transformation with y1′ 0 1 0 y1 n = ( 0,0,1 ) we have y2t = −1 0 0 y2 . Similarly, for the second y 0 0 1 y 3 3 y1′ 1 0 0 y1′ rotation with n = (1,0,0 ) , we have y2′ = 0 0 1 y2t . Note that y′ 0 −1 0 y 3 3 the second rotation is performed on the current system at hand, which is the result of the first rotation. Therefore, after substitution we can write y1′ 1 0 0 0 1 0 y1 y1′ 0 1 0 y1 = y2′ 0 0 1 −1 0 0 y2 , or y2′ = 0 0 1 y2 . The final y′ 0 −1 0 0 0 1 y y′ 1 0 0 y 3 3 3 3
0 1 0 rotation matrix is then Rij = 0 0 1 . Note that each rotation matrix 1 0 0
Tensor-Analysis-for-Engineers_2E_CH12.indd 71
10/3/2020 10:25:29 AM
72 • Tensor Analysis for Engineers, 2E pre-multiplies the previous one in the sequence of rotations. Having the final rotation matrix, we can calculate the single rotation angle. Using Equation 12.8, 2π −1 R ii − 1 −1 0 − 1 we= get θ cos= . The corresponding = cos 2 3 2 rotation axis is the eigenvector associated with the real eigenvalue for 0 −λ 1 0 1 0 Rij = 0 0 1 . We can write Rij − λδ ij = 0 −λ 1 =−λ 3 + 1 =0 . 1 0 0 1 0 −λ
Hence, we get the characteristic equation as ( λ − 1 ) ( λ 2 + λ + 1) = 0 with its only real root being λ = 1 . Note that we could use Equation 12.11 instead for calculating the eigenvalues. For λ = 1 , we get the eigenvector by solving 0 1 0 n1 n1 n2 = n1 0 0 1 n2 = (1 ) n2 , or n3 = n2 . This system is not determinate but n = n n 1 0 0 n3 3 1 3 gives n= n= n3. However, the constraint for having n as a unit vector 1 2 3 gives n12 + n22 + n32 = 1, or 3 n12 = 1. Therefore, n= and finally n= n= 1 2 3 3 3 3 3 , , we have the equivalent rotation axis or the unit vector n = . 3 3 3 In conclusion, we can say that instead of two intermediate sequential rotations, as mentioned above, we can rotate the original yi = ( y1 , y2 , y3 ) at an angle of 120°, using right-hand-rule convention, about an axis parallel 3 3 3 , , to the unit vector (see Figure 12.2). Readers should note 3 3 3 that for calculating the rotation matrix using Equation 12.7, values of the components of n should be those of the unit vector. However, to obtain the direction of the rotation axis we can multiply the components of the unit vector by a constant—for this example the multiplier is 3 . As shown in Figure 12.2, the result is a vector, or (1,1,1), which is a vector along the same direction as the unit vector n . This is consistent with the property of the eigenvalues. We can examine these results by using Equation 12.6 to 3 3 3 , , calculate the single rotation matrix, using = θ 120° and n = , 3 3 3
Tensor-Analysis-for-Engineers_2E_CH12.indd 72
10/3/2020 10:25:32 AM
Cartesian Tensor Transformation—Rotations • 73
2 2 3 3 1 1 or R11 = − = 0, similarly R= R= 0. + 1 − cos (120° ) = 22 33 3 3 3 3 2 2 3 3 3 Also, we get= R12 e123 = ° ) 1. + δ cos (120° ) + sin (120 12 − 3 3 = 0 3 =1
Similarly,
we
receive
R= R= 1, 23 31
and
the
remaining
elements
R= R= R= 0. These results are identical with the one obtained using 21 32 13 sequential rotations.
Tensor-Analysis-for-Engineers_2E_CH12.indd 73
10/3/2020 10:25:34 AM
Tensor-Analysis-for-Engineers_2E_CH12.indd 74
10/3/2020 10:25:34 AM
CHAPTER
13
COORDINATE INDEPENDENT GOVERNING EQUATIONS Reliable mathematical models, also referred to as governing equations, are important tools for engineering analysis. A reliable and validated mathematical model of a physical phenomenon is a set of algebraic relations among quantities and their various derivatives, such as Newton’s 2nd law of motion, equilibrium equations for momentum flux, Fourier’s law of heat flux, Fick’s law of mass flux, Navier-Stokes equations for flow of fluids, Maxwell’s electromagnetic equations, etc. These governing equations, along with some fundamental principles (like the 2nd law of thermodynamics, conservations of energy, mass, electric charge, etc.) form the foundation of engineering science and its applications. The quantities related to any physical phenomenon can be represented by tensors of different ranks stated in the relevant governing equations, such as force and acceleration vectors and mass of a body as a scalar, in Newton’s 2nd law of motion; gradient of temperature in Fourier’s law; divergence of velocity vector in fluid flow, etc. Sometimes, for analysis and design purposes, we need to have the relevant governing equations written in coordinate systems other than the Cartesian system. For example, for analyzing mechanical stresses in the wall of a cylindrical pressure vessel we prefer to choose the cylindrical polar coordinate system that is a natural fit to the shape of the vessel. This requirement has encouraged scientists and engineers to define various coordinate systems suitable for solving the governing equations in practice. In other words, and in the context of tensor analysis, we would like to use the relations derived and discussed in the
Tensor-Analysis-for-Engineers_2E_CH13.indd 75
10/3/2020 10:26:18 AM
76 • Tensor Analysis for Engineers, 2E previous sections to write down the terms involved in well-known governing equations in engineering in a form that is general enough for application in an arbitrary but well-defined coordinate system. In this section, we derive some new relations mostly used in engineering, in addition to those discussed in the previous sections. We hope that this helps readers in writing down similar coordinate independent terms involved in equations of their choice for their applications.
13.1 THE ACCELERATION VECTOR—CONTRAVARIANT COMPONENTS Newton’s 2nd law is a mathematical model for the motion mechanics of physical objects—specifically, it is the balance of applied forces and the rate of change of momentum, or mass times acceleration. Velocity is the time rate of change of displacement vector ds (see Section 2). Therefore, in an ds i arbitrary coordinate system x , we can write the velocity vector v as v = dt for the corresponding time increment of dt . Using Equation 3.6, we have ds dx i . Since ds is a vector, hence v is a vector and we can write it in = v = ei dt dt terms of its contravariant component vi , or
dx i i = v v= ei ei 13.1 dt Equation 13.1 clearly gives the contravariant component of the velocity dx i vector as vi = . Acceleration vector a is the time derivative of the velocity dt vector, or dv = a = a i ei 13.2 dt i dv d ( v ei ) dvi i dei = ei +v Using Equation 13.1, we can write= . To relate dt dt dt dt d dx j ∂ the time derivative to the space gradient we use chain rule , = dt dt ∂x j
hence
Tensor-Analysis-for-Engineers_2E_CH13.indd 76
vj
10/3/2020 10:26:21 AM
Coordinate Independent Governing Equations • 77
d ∂ = v j j 13.3 dt ∂x dv j ∂vi i j ∂ei Substituting into Equation 13.2, we get = . + v v a = ei v j j dt ∂x ∂x ∂ei ∂vi But j = Γ kij ek , using Equation 10.5. Therefore,= a ei v j j + vi v j Γ kij ek and ∂x ∂x in the last term we interchange dummy indices i ↔ k and use the symmetry property of the Christoffel symbol to obtain ∂vi = a v j j + vk v j Γ ijk ei 13.4 ∂x
Comparing Equations 13.2 and 13.4, we get the contravariant component of the acceleration vector as ∂vi = a i v j j + vkΓ ijk 13.5 ∂x
∂vi + vkΓ ijk , after using Equation 10.6, to ∂x j receive the compact form of Equation 13.5, as In addition, we can write v,i j =
a i = v j v,i j 13.6
Equation 13.6 recovers the familiar expression for acceleration in Cartesian ∂vi coordinates (i.e., a i = v j j ). ∂x We can also write the covariant component of the acceleration vector using a metric tensor, or ak = gik a i (see Section 8) for an arbitrary coordinate system. Now, we can write the Newton’s 2nd law in a general coordinate independent form. We choose to use the contravariant component of the acceleration vector; hence, for compatibility we must use the contravariant component of the total force vector, F = F i ei, as well. Therefore, we can write
Tensor-Analysis-for-Engineers_2E_CH13.indd 77
= F i mv j
∂vi j i + mvk v= Γ jk mv j v,i j 13.7 j ∂x
10/3/2020 10:26:27 AM
78 • Tensor Analysis for Engineers, 2E For example, in a 3D coordinate system we get three equations of motions for each dimension, with summation on indices j and k only, as 1 1 j ∂v = + mvk v j Γ1jk F mv j ∂ x 2 2 j ∂v + mvk v j Γ 2jk 13.8 F = mv j ∂x 3 3 j ∂v = + mvk v j Γ 3jk F mv j ∂x
Note that in a Cartesian system the last term in Equation 13.7, associated with the Christoffel symbol, is zero and we recover the familiar form of Newton’s equation. But when we write this equation in a curvilinear coordinate system, the term associated with the Christoffel symbol is not necessarily equal to zero. This extra term acts like an inertia force and affects the path of moving objects. For example, in a spherical coordinate system, like Earth’s geometry, ∂vi mv j j . The left-hand-side we can write Equation 13.7 as F i − mvk v j Γ ijk = ∂x ∂vi can be treated as a new total force, or F i = F i − mvk v j Γ ijk = mv j j . The ∂x k j i i term mv v Γ jk is the Coriolis force, Fc = Fci mvk v j Γ ijk 13.9
For example, from the point of view of an observer at the north pole of the Earth an object with an initial velocity along a meridian towards the equator will shift to the right, due to the inertia force resulting from the Coriolis force, while to an observer at the South Pole it will shift to the left. Or we can say that a force equivalent to the negative of the Coriolis force is acting on the object.
13.2 THE ACCELERATION VECTOR—PHYSICAL COMPONENTS Physical components of acceleration vector can be obtained using Equation 6.2, or a ( i ) = hi a i. Substituting for a i , Equation 13.5 yields ∂vi dvi a ( i ) = hi v j j + hi v j vk Γ ijk = hi + hi v j vk Γ ijk. In the latter expression, ∂ x dt =
dvi dt
Tensor-Analysis-for-Engineers_2E_CH13.indd 78
10/3/2020 10:26:29 AM
Coordinate Independent Governing Equations • 79
we write the velocity components in terms of their physical components d ( v ( i ) / hi ) v( j) v( k) i + hi Γ jk . But the first as well to receive a ( i )= hi dt h j hk term can be expanded to get hi
d ( v ( i ) / hi ) dv ( i ) d (1 / hi ) . = + hi v ( i ) dt dt dt
dv ( i ) = v ( i ) and use Equation 13.3 to write dt v ( i ) v ( j ) ∂ ( hi ) d (1 / hi ) d ( hi ) 1 = − v( i) = − . Now substituting into hi v ( i ) ∂x j dt hi dt hi h j
Now, we define
the relation for a ( i ), we obtain
v ( i ) v ( j ) ∂hi hi v ( j ) v ( k ) i + Γ jk 13.10 a ( i) = v ( i ) − hi h j ∂x j h j hk
Equation 13.10 gives the physical components of the acceleration vector in curvilinear coordinate systems. Note that for Equation 13.10 summation applies only on indices j and k , but not i.
13.3 THE ACCELERATION VECTOR IN ORTHOGONAL SYSTEMS—PHYSICAL COMPONENTS A useful application of Equation 13.10 is for orthogonal coordinate systems. As ij g= 0 for mentioned previously, for an orthogonal system we should have g= ij i ≠ j and we can use relations given by Equations 10.23–10.25 for Christoffel symbols. Therefore, expanding Equation 13.10, for example for i = 1, we v (1 ) − receive a (1 ) =
v (1 ) v ( j ) ∂h1 h1 h j
∂x
j
+
h1 v ( j ) v ( k ) h j hk
v (1 ) − summations on j and k yields a (1 ) =
Γ1jk , which after writing the
v (1 ) v (1 ) ∂h1 v ( 2 ) ∂h1 v ( 3 ) ∂h1 + + h1 h1 ∂x1 h2 ∂x2 h3 ∂x3 sum on j
h1 v (1 ) v (1 ) 1 v ( 2 ) 1 v ( 3 ) 1 h1 v ( 2 ) v (1 ) 1 v ( 2 ) 1 v ( 3 ) 1 + Γ11 + Γ12 + Γ13 + Γ 21 + Γ 22 + Γ 23 + h1 h1 h h3 h2 h1 h2 h3 2 sum on k , j =1 sum on k , j = 2
Tensor-Analysis-for-Engineers_2E_CH13.indd 79
10/3/2020 10:26:32 AM
80 • Tensor Analysis for Engineers, 2E
h1 v ( 3 ) v (1 ) 1 v ( 2 ) 1 v ( 3 ) 1 But from Equations 10.23-10.25, we Γ 31 + Γ 32 + Γ 33 . h3 h1 h h3 2 sum on k , j = 3
1 ∂h 1 ∂h1 1 1 ∂h1 1 h ∂h receive Γ111 = 11 , Γ112 = , Γ13 = , Γ 22 = Γ121 = Γ131 = − 2 2 12 , 2 3 h1 ∂x h1 ∂x h1 ∂x ( h1 ) ∂x h ∂h and Γ133 = Γ132 =. 0 Substituting for the corresponding − 3 2 13 and Γ123 = ∂ x h ( 1) v (1 ) v ( 2 ) ∂h1 v (1 ) + + Christoffel symbols and simplifying, we receive a (1 ) = h1 h2 ∂x2 v (1 ) v ( 3 ) ∂h1 v ( 2 ) v ( 2 ) ∂h2 v ( 3 ) v ( 3 ) ∂h3 . By similar operation we can − − h1 h3 ∂x3 h1 h2 ∂x1 h1 h3 ∂x1 calculate the second and third components, i.e., a ( 2 ) and a ( 3 ) . Equation 13.11 lists all the physical components of acceleration vector in a 3D orthogonal coordinate system, as
v (1 ) v ( 2 ) ∂h1 v (1 ) v ( 3 ) ∂h1 v ( 2 ) v ( 2 ) ∂h2 v ( 3 ) v ( 3 ) ∂h3 + − − v (1 ) + a (1 ) = 2 3 1 ∂ ∂ ∂ h h x h h x h h x h1 h3 ∂x1 1 2 1 3 1 2 v (1 ) v ( 2 ) ∂h2 v ( 2 ) v ( 3 ) ∂h2 v (1 ) v (1 ) ∂h1 v ( 3 ) v ( 3 ) ∂h3 13.11 + − − v ( 2 ) + a ( 2 ) = 1 3 2 2 ∂ ∂ ∂ ∂ h h x h h x h h x h h x 1 2 2 3 1 2 2 3 v (1 ) v ( 3 ) ∂h3 v ( 2 ) v ( 3 ) ∂h3 v (1 ) v (1 ) ∂h1 v ( 2 ) v ( 2 ) ∂h2 v ( 3 ) + + − − a ( 3 ) = h1 h3 ∂x1 h2 h3 ∂x2 h1 h3 ∂x3 h2 h3 ∂x3 Relations given by Equation 13.11 can be written in index form as, summation on index j only, v ( i ) v ( j ) ∂hi v ( j ) v ( j ) ∂h j 13.12 − a ( i) = v ( i ) + ∂x i hi h j ∂x j hi h j
In the next section, we use Equation 13.12 to calculate an acceleration vector’s physical components in cylindrical and spherical coordinate systems. For a Cartesian system, Equation 13.12 recovers the familiar relation (i.e., d ( v( i)) ). a ( i) = dt
Tensor-Analysis-for-Engineers_2E_CH13.indd 80
10/3/2020 10:26:34 AM
Coordinate Independent Governing Equations • 81
13.3.1 Example: Acceleration vector physical components in cylindrical and spherical coordinate systems From the results obtained in Examples 8.1 and 8.2, we have the scale factors for cylindrical and spherical coordinates as ( hr , hθ , hz ) = (1, r ,1 )
( h , h , h ) = (1, r, r sin ϕ ), respectively. Recall that the cylindrical coordinates are ( x , x , x ) ≡ ( r ,θ , z ) and those of a spherical system are ( x , x , x ) ≡ ( r,ϕ ,θ ) . Now, using Equation 13.11 we can write the physiand
r
ϕ
θ
1
1
2
2
3
3
cal components of the acceleration vector for cylindrical coordinates as v(2) v(2) v (1 ) v ( 2 ) a= , a= , and a ( 3 ) = v ( 3 ) . In terms (1) v (1) − ( 2 ) v ( 2 ) + r r vθ2 vv v − vθ + r θ , and of coordinate notation, we can write it as a= , a= r r θ r r az = v z . Note that all components are the physical components of the corresponding vectors. Therefore, the acceleration vector is written as v2 a = v r − θ r
vr vθ e ( r ) + vθ + r
e (θ ) + v z e ( z )
Note that e ( i ) are unit vectors in cylindrical coordinates. We can also write E the acceleration vector in terms of Cartesian unit vectors i, by substitut e ( r ) cosθ E1 + sin θ E2, e (θ ) = − sin θ E1 + cosθ E2, and e ( z ) = E3 (see ing for= Example 8.1), or
v2 vv a = v r − θ cosθ − vθ + r θ r r
sin θ E1
v2 vv + v r − θ sin θ + vθ + r θ cosθ E2 + v z E3 r r
13.13
In terms of coordinates themselves, we have a = r − rθ2 cosθ − rθ + 2 rθ sin θ E1 2 + r − rθ sin θ + rθ + 2 rθ cosθ E2 + zE3
(
)
(
(
)
)
)
(
)
)
Similarly, for a spherical coordinate system, using Equation 13.11, we can write the physical components of the acceleration vector for spherical vr vϕ vθ2 cos ϕ v2 v2 coordinates as a (1 ) = ar = v r − ϕ − θ , a ( 2 ) = , and − aϕ = vϕ + r r sin ϕ r r
Tensor-Analysis-for-Engineers_2E_CH13.indd 81
10/3/2020 10:26:38 AM
82 • Tensor Analysis for Engineers, 2E vr vθ vθ vϕ cos ϕ . Note that all components are the physical + r r sin ϕ components of the corresponding vectors. Therefore, the acceleration vector is a(3) = aθ = vθ +
vϕ2 + vθ2 a = v r − r
vr vϕ vθ2 cos ϕ vr vθ vθ vϕ cos ϕ − + e ( r ) + vϕ + e (ϕ ) + vθ + e (θ ) r r sin ϕ r r sin ϕ
Note that e ( i ) are unit vectors in spherical coordinates. In terms of the coordinates themselves, we have a = r − rϕ 2 − r sin 2 ϕθ2 e ( r ) + rϕ + 2 rϕ − r sin ϕ cos ϕθ2 e (ϕ ) cos ϕ e (θ ) . + rθsin ϕ + 2 rθ sin ϕ + 2 rϕθ
We can also write the acceleration vector in terms of Cartesian unit vectors Ei, by substituting for e ( r ) = sin ϕ cosθ E1 + sin ϕ sin θ E2 + cos ϕ E3, − sin θ E1 + cosθ E2 e (ϕ ) = cos ϕ cosθ E1 + cos ϕ sin θ E2 − sin ϕ E3, and e (θ ) = (see Example 8.1), or vϕ2 v2 vr vϕ vθ2 cos ϕ vr vθ vθ vϕ cos ϕ a = v r − − θ sin ϕ cosθ + vϕ + − + cos ϕ cosθ − vθ + sin θ E1 r r r r sin ϕ r r sin ϕ 13.14 vr vϕ vθ2 cos ϕ vr vθ vθ vϕ cos ϕ − + sin ϕ sin θ + vϕ + cos ϕ sin θ + vθ + cosθ E2 r sin ϕ r r sin ϕ r + ( cos ϕ − sin ϕ ) E3
vϕ2 v2 + v r − − θ r r
Equations 13.13 and 13.14 are useful relations for calculating acceleration vectors in cylindrical and spherical coordinate systems, respectively. Similar calculations can be performed for arbitrary orthogonal curvilinear systems.
13.4 SUBSTANTIAL TIME DERIVATIVES OF TENSORS Another form of derivative of tensors appearing in governing equations is the substantial derivative—also referred to as the total or convective time derivative. For a quantity like A —a scalar, vector, or tensor—that varies with space and time, we would like to collect all its derivatives. Let’s assume that we have a body of fluid moving in a fixed coordinate system, for e xample, a Cartesian system. We consider two scenarios: a) we “ride” the fluid and
Tensor-Analysis-for-Engineers_2E_CH13.indd 82
10/3/2020 10:26:39 AM
Coordinate Independent Governing Equations • 83
move with it and register its variation with respect to time, and b) we stay at a fixed location in space and register the variation of the fluid passing by with respect to time. In case (a) the changes are due to any variation of the fluid w.r.t time, since we move with it and relative space change is absent/null. In other words, the change w.r.t space is implicitly included. But in case (b) we may have changes w.r.t both time and space for the fluid in motion.
FIGURE 13.1 Local coordinates attached to a moving fluid body in a fixed Cartesian coordinate system.
Let’s consider function A = A ( yi , t ) , where A is the tensor quantity, yi is the fixed Cartesian coordinate system, and t is time (see Figure 13.1). Here, we δ yi assume that the ratio = υ i applies, or the change of space coordinates of δt the fluid w.r.t time is the same as the local fluid velocity. Now the total time DA derivative is given by the following = limit, Dt
∂A ∂A ∂yi lim + i , or δ yi , δ t → 0 ∂t ∂y ∂t
DA ∂A ∂A = + υ i i 13.15 Dt ∂t ∂y An immediate extension is to write the total derivative for A = A ij ei e j , a i mixed tensor of the second rank, in an arbitrary system x , or i j ∂ A ij ei e j DA ∂ A j ei e k 13.16 = +υ Dt ∂t ∂x k
Tensor-Analysis-for-Engineers_2E_CH13.indd 83
(
)
(
)
10/3/2020 10:26:40 AM
84 • Tensor Analysis for Engineers, 2E Using Equation 10.14, we can write the second term on the right-hand-side ∂ A ij ei e j ∂A ij k k i j = υ A j, k ei e . The first term is just equal to as υ , if the coor∂ xk ∂t dinates x i are also fixed. Note that if the x i system is moving, we should consider the changes of the basis vectors with reference to time as well (see Equation 13.16). Therefore, we get the components (mixed contravariant/ covariant) of the substantial time derivative of A, as
(
)
( )
i i DA ∂ A j + υ k A ij,k 13.17 = ∂t Dt j
Note that A= i j, k
∂A ij
i + Γ km A mj − Γ nkj Ani .
∂x A specific case is to let A be equal to velocity itself, which results in the total derivative of vi , as k
i
i Dυ ∂υ = + υ jυ,ij 13.18 ∂t Dt
In orthogonal coordinate systems, we can write the physical components the convective time derivative of a vector as DA ∂A ( i ) v ( j ) ∂A ( i ) υ ( i ) A ( j ) ∂hi + ( i) = + ∂t h j ∂x j hi h j ∂x j Dt 13.19 υ ( j ) A ( j ) ∂h j , no sum on i − ∂x i hi h j Using Equation 13.19 we can write the total time derivative of vectors like velocity or acceleration. The following example demonstrates these calculations. 13.4.1 Example: Substantial time derivation of acceleration vectors—physical components From previous examples, we have ( hr , hθ , hz ) = (1, r ,1 ) for cylindrical coordinate systems and ( hr , hϕ , hθ ) = (1, r , r sin ϕ ) for spherical coordinate systems. We can write the physical components of the total derivative of acceleration vector a = ( ar , aθ , az ) , in cylindrical coordinates, by using Equation 13.19, as
Tensor-Analysis-for-Engineers_2E_CH13.indd 84
10/3/2020 10:26:41 AM
Coordinate Independent Governing Equations • 85
Da ∂ar a ∂ar υθ ∂ar ∂a + + vz r − vθ θ = + vr r ∂θ r ∂r ∂z Dt r ∂t Da ∂aθ ∂aθ υθ ∂aθ ∂a υ a + + υ z θ + θ r 13.20 = + vr r ∂θ r ∂t ∂r ∂z Dt θ Da ∂az ∂az υθ ∂az ∂a + + vz z = + vr ∂r r ∂θ ∂z Dt z ∂t
Similarly, for spherical coordinates we receive the physical components of the total derivative of acceleration vector as
Da ∂ar aϕ υ ∂ar a ∂ar υϕ ∂ar + + θ − vϕ − vθ θ = + vr ∂r r ∂ϕ r sin ϕ ∂θ r r Dt r ∂t Da ∂aϕ ∂aϕ υϕ ∂aϕ υ ∂aϕ υϕ ar υθ aθ cos ϕ + vr + + θ + − 13.21 = ∂t ∂r r ∂ϕ r sin ϕ ∂θ r r sin ϕ Dt ϕ Da υ ∂aθ υθ ar υθ aϕ cos ϕ ∂aθ ∂a υϕ ∂aθ + vr θ + + θ + + = ∂t ∂r r ∂ϕ r sin ϕ ∂θ r r sin ϕ Dt θ Note that in these relations (i.e., Equations 13.20 and 13.21) all components are the physical components of the corresponding vectors.
13.5 CONSERVATION EQUATIONS—COORDINATE INDEPENDENT FORMS A fundamental equation for the transformation of a tensor quantity per unit volume, Ψ in an arbitrary control volume with a velocity field V, can be obtained using the Gauss divergence theorem and the law of the principle of conservation [11]. An integral equation defining the time rate and divergence of convective transformation balanced with the rate of production/destruction Ω of the quantity leads to the differential relation given by Equation 13.22, as ∂Ψ + ( Ψ V i ), i =Ω 13.22 ∂t
The second term can be written as ( Ψ V i ), i= V i Ψ , i + Ψ V i, i since covariant differentiation follows the differentiation product rule. Therefore, L.H.S
Tensor-Analysis-for-Engineers_2E_CH13.indd 85
10/3/2020 10:26:42 AM
86 • Tensor Analysis for Engineers, 2E
of Equation 13.22 can be written as
∂Ψ DΨ i + V i Ψ , i + Ψ V= + Ψ V i, i . In ,i t Dt ∂ =
vector form, this equation reads
DΨ Dt
DΨ DΨ + Ψ∇ ⋅ V = + Ψ V i, i = Ω 13.23 Dt Dt
Equation 13.22 (or equally Equation 13.23) can be considered the equation of motion for the control volume or, in general, the medium that the quantity Ψ transports in it. The source term is associated with the quantity in question: for example, the mass, momentum, energy, electric charge, etc. Now we consider mass conservation, or let Ψ =ρ where ρ is mass per unit volume or density of the quantity in motion. After substitution in Dρ + ρ V i, i = Ω. If mass source is zero, then we get Equation 13.23, we get Dt the continuity equation, as Dρ + ρ V i, i = 0 13.24 Dt
Note that
Dρ ∂ρ = + V i ρ, i . Some specific cases can be observed: Dt ∂t
ρ = ρ ( t ) only, i.e., we have an inhomogeneous quantity whose density changes with time. In this case the continuity equation reads as ∂ρ ∂ρ + V i ( ρ, i ) + ρ V i , i = 0 , or + ρ V i, i = 0. ∂t ∂t =0
ρ = ρ ( x i ) only, i.e., we have an inhomogeneous quantity whose density
changes from location to location in the continuum but remains constant with time at any given location. In this case the continuity equation reads i as ∂ρ + V i ρ + ρ V i = 0 , or ( ρ V ) = 0 . ,i ,i ∂t =0
ρ is a constant, i.e., we have a homogeneous quantity with constant
density at all locations and times in the continuum. In this case the Dρ i 0 , or V i, i = 0. continuity equation reads as + ρ V ,i = Dt =0
Tensor-Analysis-for-Engineers_2E_CH13.indd 86
10/3/2020 10:26:44 AM
Coordinate Independent Governing Equations • 87
Now we consider momentum conservation, or let Ψ =ρ V , i.e., momentum per unit volume of the quantity in motion. In this case Ψ is a vector quantity. Substituting in Equation 13.22, the i th contravariant component reads
∂ ( ρV i )
∂t
(
+ ρ V iV j
)
, j
= Ω i 13.25
Equation 13.25 is Newton’s 2nd law written in the coordinate independent form for a medium, for example a fluid moving in the continuum. The L.H.S is the inertia force and the R.H.S is the applied force on the material, per unit volume. Considering the material as a fluid, then the applied force could be, in general, a combination of hydrostatic, viscous, gravitational, electromagnetic, etc. forces.
Tensor-Analysis-for-Engineers_2E_CH13.indd 87
10/3/2020 10:26:45 AM
Tensor-Analysis-for-Engineers_2E_CH13.indd 88
10/3/2020 10:26:45 AM
CHAPTER
14
COLLECTION OF RELATIONS FOR SELECTED COORDINATE SYSTEMS In this section, we provide a list of some commonly used coordinate systems as well as relations for an arbitrary orthogonal curvilinear coordinate system. We categorize the content based on the coordinate systems. All coordinates are considered in a 3D Euclidean space. Cartesian coordinate systems are fixed references consisting of three flat planes. The curvilinear coordinates may include one or more curved coordinate surfaces.
14.1 CARTESIAN COORDINATE SYSTEM We use ( e (1 ) , e ( 2 ) , e ( 3 ) ) = Ex , Ey , Ez , ( A (1 ) , A ( 2 ) , A ( 3 ) ) = Ax , Ay , Az symbols for physical components. Note that in a Cartesian system, all contravariant, covariant, and physical components of a tensor quantity are identical, hence all indices are shown as subscripts.
(
Tensor-Analysis-for-Engineers_2E_CH14.indd 89
)
(
)
10/3/2020 10:27:04 AM
90 • Tensor Analysis for Engineers, 2E
TABLE 14.1 Relations for tensors and their related derivatives in Cartesian coordinate systems.
Coordinates Basis/Unit vectors, Ei
( x, y, z ) ≡ ( X, Y, Z )
Scale factors, hi
( E , E , E ) ≡ ( i , j, k ), orthogonal ( h , h , h ) = (1,1,1)
Metric tensors, gij
g= g= g= 1 , the rest are null xx yy zz
Jacobian
=1
x
y
x
y
z
z
d𝕍 = dxdydz 2 2 2 Line element and magnitude dS = dxEx + dyEy + dzEz dS = ( dx ) + ( dy ) + ( dz ) A = Ax , Ay , Az = Ax Ex + Ay Ey + Az Ez Vector components Unit volume
(
)
Christoffel symbols
null A ⋅= B Ax Bx + Ay By + Az Bz Dot-product (two vectors) A ×= B Ay Bz − Az By Ex + ( Az Bx − Ax Bz ) Ey Cross-product (two vectors) + Ax By − Ay Bx Ez
(
)
(
)
Gradient vector
∂ ∂ ∂ = ∇ Ex + Ey + Ez ∂x ∂y ∂z
Gradient of a scalar, Ψ
∂Ψ ∂Ψ ∂Ψ = ∇Ψ Ex + Ey + Ez ∂x ∂y ∂z
Gradient of a vector, A
∂Ax ∂x ∂Ax ∇A = ∂y ∂A x ∂z
∂Ay ∂x ∂Ay ∂y ∂Ay ∂z
∂Az ∂x ∂Az ∂y ∂Az ∂z
Curl of a vector, A
∂Az ∂Ay ∂Ax ∂Az ∇= ×A − − Ex + ∂z ∂x ∂z ∂y
Divergence of a vector, A
∂Ax ∂Ay ∂Az ∇ ⋅= A + + ∂x ∂y ∂z
∂Ay ∂Ax − Ez Ey + ∂y ∂x
(continued)
Tensor-Analysis-for-Engineers_2E_CH14.indd 90
10/3/2020 10:27:06 AM
Collection of Relations for Selected Coordinate Systems • 91
∂2 Ψ ∂2 Ψ ∂2 Ψ ∇ 2 Ψ = ∇ ⋅ ∇Ψ = + 2 + 2 ∂x2 ∂y ∂z 2 ∇ A=
(
Laplacian of a scalar, Ψ
)
2 2 2 ∂ 2 Ax ∂ 2 Ax ∂ 2 Ax ∂ Ay ∂ Ay ∂ Ay + + + + E Ey 2 + x ∂x2 ∂y2 ∂z2 ∂y2 ∂z2 ∂x ∂ 2 A ∂ 2 Az ∂ 2 Az + 2z + + 2 Ez ∂y2 ∂z ∂x
Laplacian of a vector, A
∂4 ∂4 ∂4 ∇ 4 = 4 + 4 + 4 ∂y ∂z ∂x
Biharmonic operator
∂4 ∂4 ∂4 + 2 2 2 + 2 2 + 2 2 ∂x ∂y ∂y ∂z ∂x ∂z
14.2 CYLINDRICAL COORDINATE SYSTEMS We use ( e (1 ) , e ( 2 ) , e ( 3 ) ) = ( e ( r= A (1 ) , A ( 2 ) , A ( 3 ) ) ) , e (θ ) , e ( z ) ) , A (= ( A ( r ) , A (θ ) , A ( z) ) to designate the unit vectors e ( i ) and physical components of vector A. Most of the expressions are written in terms of physical components. For Christoffel symbols, see Equation 10.28. TABLE 14.2 Relations for quantities and their related derivatives in cylindrical coordinate systems.
Coordinates (orthogonal)
( x , x , x ) ≡ ( r,θ, z ) , θ is the azimuth angle 1
2
3
Coordinate functions
x = r cos θ
y = r sin θ
z=z
Coordinate surfaces
x 2 + y2 = r 2, cylinders
y / x = tan θ, planes
z = constant, planes
Basis vectors, ei, covariant
cos θ − r sin θ 0 ( er , eθ , ez ) = sin θ , r cosθ , 0 0 0 1
Basis vectors, e i, contravariant
cos θ − sin θ / r 0 r θ z ( e , e , e ) = sin θ , cosθ / r , 0 1 0 0
Scale factors, hi
( hr , hθ , hz ) = (1, r,1) (continued)
Tensor-Analysis-for-Engineers_2E_CH14.indd 91
10/3/2020 10:27:10 AM
92 • Tensor Analysis for Engineers, 2E
Unit vectors, physical components
cos θ − sin θ 0 = e r , e θ , e z ( ( ) ( ) ( ) ) sin θ , cosθ , 0 0 0 1
Metric tensors, gij and g ij
g= g= ( g rr )= ( g zz )= 1, g= rr zz θθ null-system is orthogonal
Jacobian
=r
Unit volume
d𝕍 = rdrdθ dz
Line element and magnitude
dS =dre ( r ) + rdθ e (θ ) + dze ( z )
−1
−1
( g )= θθ
−1
dS =
( dr )2 + ( rdθ )2 + ( dz )2
A = A ( r ) e ( r ) + A (θ ) e (θ ) + A ( z ) e ( z )
Vector components Christoffel symbols, based on Unit vectors
r Γˆ θθ = −1 , Γˆ θrθ = Γˆ θθ r =, 1 the rest are null
Christoffel symbols, based on covariant Basis vectors
r Γθθ = − r , Γθrθ = Γθθ r = 1 / r , the rest are null
Dot-product (two vectors) Cross-product (two vectors)
r 2 , the rest are
= A ⋅ B A ( r ) B ( r ) + A (θ ) B (θ ) + A ( z ) B ( z ) = A × B A (θ ) B ( z ) − A ( z ) B (θ ) e ( r ) + A ( z ) B ( r ) − A ( r ) B ( z ) e (θ ) + A ( r ) B (θ ) − A (θ ) B ( r ) e ( z )
Gradient vector
∂ 1 ∂ ∂ = ∇ er + eθ + ez ∂r r ∂θ ∂z
Gradient of a scalar, Ψ
∂Ψ 1 ∂Ψ ∂Ψ = ∇Ψ er + eθ + ez ∂r r ∂θ ∂z
Gradient of a vector, A
∂A ( r ) ∂r ∂A ( r ) A (θ ) = ∇A − r r ∂θ ∂A ( r ) ∂z
∂A (θ ) ∂r ∂A (θ ) A ( r ) + r ∂θ r ∂A (θ ) ∂z
∂A ( z ) ∂r ∂A ( z ) r ∂θ ∂A ( z ) ∂z
(continued)
Tensor-Analysis-for-Engineers_2E_CH14.indd 92
10/3/2020 10:27:13 AM
Collection of Relations for Selected Coordinate Systems • 93
∂A ( z ) ∂A (θ ) ∂A ( r ) ∂A ( z ) = ∇×A − − e (r) + e (θ ) r θ z ∂ ∂ ∂r ∂z ∂ ( rA (θ ) ) ∂A ( r ) + − e ( z) r ∂r r ∂θ
Curl of a vector, A
Divergence of a vector, A
∂ ( rA ( r ) ) ∂A (θ ) ∂A ( z ) = ∇⋅A + + ∂z r ∂r r ∂θ
Laplacian operator
1 ∂ ∂ 1 ∂2 ∂2 ∇2 = ∇ ⋅ ∇ = r + + r ∂r ∂r r 2 ∂θ 2 ∂z2
Laplacian of a scalar, Ψ
1 ∂ ∂Ψ 1 ∂ 2 Ψ ∂ 2 Ψ ∇ 2 Ψ = ∇ ⋅ ∇Ψ = + r + r ∂r ∂r r 2 ∂θ 2 ∂z2
Laplacian of a vector, A
Biharmonic operator
( ) (
)
2 2 ∂A (θ ) A ( r ) ∇ 2 A =∇ − 2 e (r) A(r) − 2 r ∂θ r 2 ∂A ( r ) A (θ ) + ∇ 2 A (θ ) + 2 − 2 e (θ ) + ∇ 2 A ( z ) e ( z ) r ∂ θ r 1 ∂ ∂ 1 ∂ ∂ 1 ∂2 ∇4 = r + r r ∂r ∂r r ∂r ∂r r 2 ∂θ 2
1 ∂2 ∂4 + 4 2 2 r ∂θ ∂z
14.3 SPHERICAL COORDINATE SYSTEMS We use ( e (1 ) , e ( 2 ) , e ( 3 ) ) = ( er , eϕ , eθ ), ( A (1 ) , A ( 2 ) , A ( 3 ) ) = ( A ( r ) , A (ϕ ) , A (θ ) ) to designate the unit vectors e ( i ) and physical components of vector A. Most of the expressions are written in terms of physical components. For Christoffel symbols, see Equation 10.28. TABLE 14.3 Relations for quantities and their related derivatives in spherical coordinate systems.
Coordinates (orthogonal) Coordinate functions Coordinate surfaces
( r,ϕ ,θ ), θ azimuth and ϕ polar angle x = r sin ϕ cos θ x2 + y2 + z2 = r 2, spheres
y = r sin ϕ sin θ
(x
z = r cos ϕ
+ y2 ) / z 2 = tan 2 ϕ, y / x = tan θ, planes cones 2
(continued)
Tensor-Analysis-for-Engineers_2E_CH14.indd 93
10/3/2020 10:27:17 AM
94 • Tensor Analysis for Engineers, 2E
Basis vectors, ei , covariant
sin ϕ cos θ r cos ϕ cos θ − r sin ϕ sin θ = e , e , e ( r ϕ θ ) sin ϕ sin θ , r cos ϕ sin θ , r sin ϕ cosθ cos ϕ − r sin ϕ 0
Basis vectors, e i , contravariant
sin ϕ cos θ cos ϕ cos θ − sin θ 1 r ϕ θ 1 ( e , e , e ) = sin ϕ sin θ , r cos ϕ sin θ , r sin ϕ cosθ 0 cos ϕ − sin ϕ
Scale factors, hi
( h , h , h ) = (1, r, r sin ϕ )
Unit vectors, physical components
sin ϕ cos θ cos ϕ cos θ − sin θ = e r , e ϕ , e θ ( ) ( ) ( ) ( ) sin ϕ sin θ , cos ϕ sin θ , cosθ cos ϕ − sin ϕ 0
r
= grr
Metric tensors, gij ij and g
ϕ
θ
( g= ) rr −1
1,= gϕϕ
= r 2 sin ϕ
Unit volume
d𝕍 = r 2 sin ϕ drdϕ dθ dS dre ( r ) + rdϕ e (ϕ ) = + r sin ϕ dθ e (θ )
Line element and magnitude
r 2= , gθθ
dS =
(g = ) ( r sin ϕ ) , the rest θθ
−1
2
( dr )2 + ( rdϕ )2 + ( r sin ϕ dθ )2
A = A ( r ) e ( r ) + A (ϕ ) e (ϕ ) + A (θ ) e (θ )
Vector components Christoffel symbols, based on covariant Basis vectors
r Γϕϕ = −r
Christoffel symbols, based on Unit vectors
r Γϕϕ = −1
Cross-product (two vectors)
ϕϕ −1
are null-system is orthogonal
Jacobian
Dot-product (two vectors)
(g = )
Γϕrϕ = Γθrθ
− r sin 2 ϕ
ϕ Γθθ = − sin ϕ cos ϕ
r Γθθ = − sin ϕ
ϕ Γθθ = − cos ϕ
Γϕrϕ = 1,
r Γθθ =
=1/r
sin ϕ Γθrθ =
Γθϕθ = cot ϕ Γθϕθ = cos ϕ
= A ⋅ B A ( r ) B ( r ) + A (ϕ ) B (ϕ ) + A (θ ) B (θ )
= A × B A (ϕ ) B (θ ) − A (θ ) B (ϕ ) e ( r ) + A (θ ) B ( r ) − A ( r ) B (θ ) e (ϕ ) + A ( r ) B (ϕ ) − A (ϕ ) B ( r ) e (θ )
(continued)
Tensor-Analysis-for-Engineers_2E_CH14.indd 94
10/3/2020 10:27:20 AM
Collection of Relations for Selected Coordinate Systems • 95
∂ 1 ∂ 1 ∂ = ∇ er + eϕ + eθ ∂ ∂ r ϕ r sin ϕ θ ∂r
Gradient vector
∂Ψ 1 ∂Ψ 1 ∂Ψ er + = ∇Ψ eϕ + eθ ∂r r ∂ϕ r sin ϕ ∂θ
Gradient of a scalar, Ψ
∇A = ∂A ( r ) ∂r ∂A ( r ) A (ϕ ) − r ∂ϕ r 1 ∂A ( r ) A (θ ) − r r sin ϕ ∂θ
Gradient of a vector, A
= ∇×A
∂A (ϕ ) ∂r ∂A (ϕ ) A ( r ) + r ∂ϕ r 1 ∂A (ϕ ) A (θ ) − cot ϕ r sin ϕ ∂θ r
∂A (θ ) ∂r ∂A (θ ) r ∂ϕ 1 ∂A (θ ) A ( r ) A (ϕ ) + + cot ϕ r sin ϕ ∂θ r r
1 ∂ ( A (θ ) sin ϕ ) ∂A (ϕ ) − e (r) ∂ϕ ∂θ r sin ϕ 1 1 ∂A ( r ) ∂ ( rA(θ ) + − e (ϕ ) r sin ϕ ∂θ ∂r 1 ∂ ( rA (ϕ ) ) ∂A ( r ) + − e (θ ) r ∂r ∂ϕ
Curl
1 ∂ ( r 2 A ( r ) ) 1 ∂ ( A (ϕ ) sin ϕ ) 1 ∂A (θ ) Divergence of a= ∇⋅A + + 2 vector, A ∂r ∂ϕ r r sin ϕ r sin ϕ ∂θ
Divergence of a rank tensor, A
∂ ( r 2 A ( rr ) ) 1 ∂ ( A ( rϕ ) sin ϕ ) A (ϕϕ ) + A (θθ ) 1 ∂A ( rθ ) e (r) = ∇⋅ A + − + 2 ∂ ∂ sin sin ϕ ϕ ϕ ∂θ r r r r r r ∂ ( r 2 A ( rϕ ) ) 1 ∂ ( A (ϕϕ ) sin ϕ ) A ( rϕ ) 1 ∂A (ϕθ ) A (θθ ) 2nd cot ϕ e (ϕ ) + + + + − 2 r r r sin r r r sin r ϕ ϕ ϕ θ ∂ ∂ ∂ 2 ∂ ( r A ( rθ ) ) ∂A (ϕθ ) 1 ∂A (θθ ) A ( rθ ) 2 A (ϕθ ) cot ϕ e (θ ) + + + + + r 2 ∂r r ∂ϕ r sin ϕ ∂θ r r
Laplacian operator
∂ ∂ ∂2 1 ∂ 2 ∂ 1 1 ∇2 = ∇ ⋅ ∇ = 2 r + 2 sin ϕ + 2 2 r ∂r ∂r r sin ϕ ∂ϕ ∂ϕ r sin ϕ ∂θ 2
Laplacian of a scalar, Ψ
1 ∂ 2 ∂Ψ 1 ∂ ∂Ψ ∇ 2 Ψ = ∇ ⋅ ∇Ψ = 2 r + sin ϕ r ∂r ∂r r 2 sin ϕ ∂ϕ ∂ϕ 1 ∂2 Ψ + 2 2 r sin ϕ ∂θ 2
( ) (
)
(continued)
Tensor-Analysis-for-Engineers_2E_CH14.indd 95
10/3/2020 10:27:22 AM
96 • Tensor Analysis for Engineers, 2E
Laplacian of a vector, A
2Ar 2 sin A 2 A 2 A 2 A r 2 2 er 2 r r sin r sin A 2 A r 2 cos A 2 A 2 2 2 2 2 e r r r sin sin A 2 A r 2 coss A 2 A 2 2 2 2 2 e r sin r sin r sin
Biharmonic operator, ∇ 4
1 ∂ 2 ∂ 1 ∂ 2 ∂ ∇4 = 2 r r r ∂r ∂r r 2 ∂r ∂r ∂ ∂ 1 ∂ ∂ 1 + 2 sin ϕ 2 sin ϕ r sin ϕ ∂ϕ ∂ϕ r sin ϕ ∂ϕ ∂ϕ 1 1 ∂2 ∂2 + 2 2 r sin ϕ ∂θ 2 r 2 sin 2 ϕ ∂θ 2
14.4 PARABOLIC COORDINATE SYSTEMS We use ( e (1 ) , e ( 2 ) , e ( 3 ) ) = ( e (ξ ) , e (η ) , e (θ ) ), ( A (1 ) , A ( 2 ) , A ( 3 ) ) = ( A (ξ ) , A (η ) , A (θ ) ) to designate the unit vectors e ( i ) and physical components of vector A. Most of the expressions are written in terms of physical components. For Christoffel symbols, see Example 15.4. TABLE 14.4 Relations for quantities and their related derivatives in parabolic coordinate systems.
Coordinates (orthogonal) Coordinate functions
(ξ ,η ,θ ) x = ξη cos θ
x 2 + y2 =
Coordinate surfaces
Basis vectors, eicovariant
y = ξη sin θ
= z
(ξ
2
− η2 ) / 2
x 2 + y2 =
−2ξ 2 ( z − ξ 2 / 2 ) ,
−2η 2 ( z − η 2 / 2 ) ,
paraboloids
paraboloids
y / x = tan θ , planes
η cos θ ξ cos θ −ξη sin θ ( eξ , eη , eθ ) = η sin θ , ξ sin θ , ξη cosθ ξ −η 0
(continued)
Tensor-Analysis-for-Engineers_2E_CH14.indd 96
10/3/2020 10:27:29 AM
Collection of Relations for Selected Coordinate Systems • 97
Basis vectors, e i, contravariant
η cos θ 1 ξ η θ ( e , e , e ) = ξ 2 + η 2 η sin θ , ξ 2 +1 η 2 ξ
Scale factors, hi
(h , h ,h ) = (ξ ξ
η
θ
2
+ η 2 , ξ 2 + η 2 , ξη
ξ cos θ − sin θ 1 ξ sin θ , cos θ −η ξη 0
)
( e (ξ ) , e (η ) , e (θ ) ) = Unit vectors, physical components
1 2 2 ξ +η
Metric tensors, gij and g ij
gξξ = ξ 2 η 2 , gηη = ξ 2 η 2 , gθθ = (ξη ) , ( gξξ ) =+ ( gηη ) =+ ( gθθ ) = the rest are null
Basis vectors, e i , contravariant
η cos θ ξ cos θ − sin θ 1 1 ξ η θ 1 ( e , e , e ) = ξ 2 + η 2 η sin θ , ξ 2 + η 2 ξ sin θ , ξη cosθ ξ −η 0
Jacobian
η cos θ 1 η sin θ , 2 2 ξ +η ξ
ξ cos θ − sin θ ξ sin θ , cos θ −η 0
−1
−1
−1
2
= ξη (ξ 2 + η 2 )
Unit volume
= d𝕍 ξη (ξ 2 + η 2 ) dξ dη dθ dS =
Line element and magnitude
dS = ξ 2 + η 2 dξ e (ξ ) + dη e (η ) + ξη dθ e (θ )
Vector components
A = A (ξ ) e (ξ ) + A (η ) e (η ) + A (θ ) e (θ )
(ξ
2
+ η 2 ) ( dξ ) + (ξ 2 + η 2 ) ( dη ) + (ξη dθ ) 2
2
2
θ Γθξθ = Γθξ ξ Γξξξ = −Γηη
ξ Γξξη = Γηξ
=1/ξ
ξ Christoffel symbols, θ θ Γθθ = Γηθθ = Γηθ = Γθη η = Γηξη = Γηξ = Γηηη based on covariant 2 2 2 2 2 2 −ξη / (ξ + η ) −ηξ / (ξ + η ) = 1 / η Basis vectors = ξ / (ξ 2 += η 2 ) η / (ξ 2 + η 2 ) θ θ Γηθ = Γθη
= 1 /η
Dot-product (two vectors)
= A ⋅ B A (ξ ) B (ξ ) + A (η ) B (η ) + A (θ ) B (θ )
(continued)
Tensor-Analysis-for-Engineers_2E_CH14.indd 97
10/3/2020 10:27:32 AM
98 • Tensor Analysis for Engineers, 2E
Cross-product (two vectors)
= A × B A (η ) B (θ ) − A (θ ) B (η ) e (ξ ) + A (θ ) B (ξ ) − A (ξ ) B (θ ) e (η ) + A (ξ ) B (η ) − A (η ) B (ξ ) e (θ )
∂ ∂ 1 ∂ 1 1 eξ + eη + ∇ Gradient= vector eϕ 2 2 2 2 ξη ∂ϕ ξ + η ∂r ξ + η ∂η 1 ∂Ψ 1 ∂Ψ 1 ∂Ψ Gradient = of a eξ + e + e ∇Ψ 2 2 2 2 η ξη ∂ϕ ϕ scalar, Ψ ξ + η ∂r ξ + η ∂η ∇2 = ∇ ⋅ ∇ =
( )
Laplacian operator
∂ 1 2 ξ (ξ + η ) ∂ξ 2
+
1
(ξη )2
∇ 2 Ψ = ∇ ⋅ ∇Ψ = Laplacian of a scalar, Ψ
Laplacian of a vector, A
(
)
∂ ∂ ∂ 1 ξ + η 2 2 ∂ξ η (ξ + η ) ∂η ∂η
∂2 ∂θ 2
∂ ∂Ψ ∂ ∂Ψ 1 1 ξ + η 2 2 2 ξ (ξ + η ) ∂ξ ∂ξ η (ξ + η ) ∂η ∂η 2
+
1
(ξη )2
∂2 Ψ ∂θ 2
∇2 A = , use Equation 11.21.
For this coordinate system, we didn’t list all relations, since they are lengthy. But readers can use the relations given in the next section, as they are applicable to all orthogonal coordinate systems.
14.5 ORTHOGONAL CURVILINEAR COORDINATE SYSTEMS We use ( x1 , x2 , x3 ) ≡ ( x1 , x2 , x3 ) symbols for presentation. Note that all components listed in Table 14.5 are given as physical components. The functional relations for coordinate variables vs. those of Cartesian ones are required for calculating the values of quantities listed in Table 14.5.
Tensor-Analysis-for-Engineers_2E_CH14.indd 98
10/3/2020 10:27:33 AM
Collection of Relations for Selected Coordinate Systems • 99
TABLE 14.5 Relations for quantities and their related derivatives in curvilinear orthogonal coordinate systems.
( x1 , x2 , x3 ) ≡ ( x1 , x2 , x3 )
Coordinates (orthogonal)
Basis vectors ei , covariant
( e1 , e2 , e3 )
Scale factors, hi
( h1 , h2 , h3 )
Unit vectors e ( i ) , physical components
( e (1) , e ( 2 ) , e ( 3 ) ) = ( e1 / h1 , e2 / h2 , e3 / h3 )
−1 Metric tensors, gij = g11 (= g11 ) ij and g rest are null
, g22 ( h1 )= 2
Basis vectors e i contravariant
( e , e , e ) = ( e g
Jacobian
= h1 h2 h3
Unit volume
d𝕍 = dx1 dx2 dx3
1
2
3
1
11
g ) g ) (h = ) , g (= ( h ) , the (=
Christoffel symbols Dot-product (two vectors)
2
33
3
dS =
( h1 dx1 )
2
+ ( h2 dx2 ) + ( h3 dx3 ) 2
2
A = A (1 ) e (1 ) + A ( 2 ) e ( 2 ) + A ( 3 ) e ( 3 )
(use Equations 10.20-10.25) = A ⋅ B A (1 ) B (1 ) + A ( 2 ) B ( 2 ) + A ( 3 ) B ( 3 ) = A × B A ( 2 ) B ( 3 ) − A ( 3 ) B ( 2 ) e (1 ) + A ( 3 ) B (1 ) − A (1 ) B ( 3 ) e ( 2 ) + A (1 ) B ( 2 ) − A ( 2 ) B (1 ) e ( 3 )
Cross-product (two vectors)
∇ Gradient vector= Gradient of a scalar, Ψ
33 −1
2
2
, e2 g 22 , e3 g 33 )
dS h1 dx1 e (1 ) + h2 dx2 e ( 2 ) Line element and= magnitude + h3 dx3 e ( 3 ) Vector components
22 −1
= ∇Ψ
1 ∂ 1 ∂ 1 ∂ e (1 ) + e (2) + e (3) h1 ∂x1 h2 ∂x2 h3 ∂x3 1 ∂Ψ 1 ∂Ψ 1 ∂Ψ e (1 ) + e (2) + e (3) h1 ∂x1 h2 ∂x2 h3 ∂x3
(continued)
Tensor-Analysis-for-Engineers_2E_CH14.indd 99
10/3/2020 10:27:36 AM
100 • Tensor Analysis for Engineers, 2E
Gradient of a vector, A
∂A i k i j ∇ A =∇ j A e j = ∇ j ( A i ei ) e j = j + A Γ kj ei e , see Example 15.13. ∂x
(
= ∇×A Curl of a vector, A
)
1 h2 h3 + +
∂ ( h3 A ( 3 ) ) ∂ ( h2 A ( 2 ) ) − e (1 ) ∂x2 ∂x3
1 ∂ ( h1 A (1 ) ) ∂ ( h3 A ( 3 ) ) − e (2) ∂x3 ∂x1 h1 h3 1 ∂ ( h2 A ( 2 ) ) ∂ ( h1 A (1 ) ) − e (3) ∂x1 ∂x2 h1 h2
1 ∂ ( h2 h3 A (1 ) ) ∂ ( h1 h3 A ( 2 ) ) ∂ ( h1 h2 A ( 3 ) ) Divergence of a = ∇ ⋅A + + vector, A h1 h2 h3 ∂x1 ∂x2 ∂x3
Divergence of a 2nd rank tensor, A
1 ∂ ( h2 h3 A (11 ) ) ∂ ( h1 h3 A ( 21 ) ) ∂ ( h1 h2 A ( 31 ) ) + + ∂x1 ∂x2 ∂x3 h1 h2 h3 e (1 ) ∇ ⋅ A = A A A A 12 31 22 33 ( ) ( ) ( ) ( ) h ∂ h h h ∂ ∂ ∂ 3 1 1 2 + + − − h1 h2 ∂x2 h1 h3 ∂x3 h1 h2 ∂x1 h1 h3 ∂x1 1 ∂ ( h2 h3 A (12 ) ) ∂ ( h1 h3 A ( 22 ) ) ∂ ( h1 h2 A ( 32 ) ) + + ∂x1 ∂x2 ∂x3 h1 h2 h3 e (2) + A A A A 23 12 33 11 ( ) ( ) ( ) ( ) h ∂ h ∂ h ∂ h ∂ 3 2 2 1 + − − + h2 h3 ∂x3 h1 h2 ∂x1 h2 h3 ∂x2 h1 h2 ∂x2 1 ∂ ( h2 h3 A (13 ) ) ∂ ( h1 h3 A ( 23 ) ) ∂ ( h1 h2 A ( 33 ) ) + + ∂x1 ∂x2 ∂x3 h1 h2 h3 e (3) + 31 23 11 22 A A A A ( ) ( ) ( ) ( ) h h ∂ ∂ h h ∂ ∂ 3 3 1 2 + + − − h1 h3 ∂x1 h2 h3 ∂x2 h3 h1 ∂x3 h2 h3 ∂x3
∇ 2 Ψ = ∇ ⋅ ∇Ψ Laplacian of a scalar, Ψ =
(
)
1 ∂ h2 h3 ∂Ψ ∂ h1 h3 ∂Ψ ∂ h1 h2 ∂Ψ + + h1 h2 h3 ∂x1 h1 ∂x1 ∂x2 h2 ∂x2 ∂x3 h3 ∂x3
(continued)
Tensor-Analysis-for-Engineers_2E_CH14.indd 100
10/3/2020 10:27:37 AM
Collection of Relations for Selected Coordinate Systems • 101
∇2 A =
∂ h2 ∂ ( h1 A1 ) ∂ ( h3 A3 ) − ∂x1 1 ∂ 1 ∂x3 h1 h3 ∂x3 ∇⋅A + e1 h2 h3 ∂ h ∂ ( h2 A2 ) ∂ ( h1 A1 ) h1 ∂x1 3 − − ∂x h h ∂x ∂x2 2 1 2 1 ∂ h3 ∂ ( h2 A2 ) ∂ ( h1 A1 ) − ∂x2 1 ∂ 1 ∂x1 h1 h2 ∂x1 + ∇ ⋅A + e2 h1 h3 ∂ h ∂ ( h3 A3 ) ∂ ( h2 A2 ) h2 ∂x2 1 − − ∂x3 ∂x3 h2 h3 ∂x2 ∂ h1 ∂ ( h3 A3 ) ∂ ( h2 A2 ) − ∂x3 1 ∂ 1 ∂x2 h2 h3 ∂x2 + ∇ ⋅A + e3 h2 h1 ∂ h ∂ ( h1 A1 ) ∂ ( h3 A3 ) h3 ∂x3 2 − − ∂x1 ∂x1 h1 h3 ∂x3
(
Laplacian of a vector, A
)
(
)
(
)
1 ∂ ∂ 1 ∂ ∂ ij g kl l g ∇4 = j k i ∂x ∂x ∂x ∂x ∇2 Biharmonic operator, ∇ 4
∂ h h ∂ 1 ∂ h2 h3 ∂ 2 3 ∂x1 h1 ∂x1 h1 h2 h3 ∂x1 h1 ∂x1 ∂ h1 h3 ∂ 1 ∂ h1 h3 ∂ 1 = + h1 h2 h3 ∂x2 h2 ∂x2 h1 h2 h3 ∂x2 h2 ∂x2 ∂ h1 h2 ∂ + ∂ h1 h2 ∂ 1 ∂x3 h3 ∂x3 h1 h2 h3 ∂x3 h3 ∂x3
Modern engineering software tools provide facilities to include curvilinear coordinate systems for computer modeling. For example, COMSOL Multiphysics® has a curvilinear coordinate interface for defining object orientation in an arbitrary system [12].
Tensor-Analysis-for-Engineers_2E_CH14.indd 101
10/3/2020 10:27:38 AM
Tensor-Analysis-for-Engineers_2E_CH14.indd 102
10/3/2020 10:27:38 AM
CHAPTER
15
RIGID BODY ROTATION: EULER ANGLES, QUATERNIONS, AND ROTATION MATRIX In Chapter 12, we discussed Cartesian tensors and rotation of coordinate system using Rodrigues’ formula (see Equation 12.6). For practical applications, mainly rigid body rotation, in this chapter we would like to expand on this topic in relation to Euler angles and quaternions methods. A rigid body transformation can be composed of two parts: a) translation and b) rotation. The translation is usually referred to a vector connecting the origin of the coordinate system to the centre of mass. Therefore, we can always refer the rotation to the instantaneous center of mass and focus on the rotation for calculation and analysis of rigid body motion. In general, there are at least eight methods to represent rotation [13]. Among these three inter-related methods commonly used for calculating rigid body orientation after it goes through possible rotations in a 3D space. These are: 1) rotation matrix, 2) single axis-angle, and 3) quaternions. The first two methods are discussed in Chapter 12 and expanded in this chapter with the inclusion of Euler angles method. The third method is more robust and with less limitations compared to the other methods in terms of practical applications and is new to this second edition. When dealing with rotation of a rigid body, several confusions may arise in terms of definition of rotation matrix and rotation variables. To avoid these ambiguities, identifying two definitions is critical. First, the assumed
Tensor-Analysis-for-Engineers_2E_CH15.indd 103
10/3/2020 10:28:34 AM
104 • Tensor Analysis for Engineers, 2E transformation should be clear by the way the rotation matrix (see Equation 12.7) is meant to be, i.e., it is meant to relate the coordinates obtained after rotation to the original ones or vice versa. Second, how the rotation itself is performed, i.e., is it that the coordinates are rotated and quantities, like a vector, is kept fixed and we seek to find its new coordinates with reference to the rotated ones or is it that the quantities, like a vector, are rotated and the coordinates are kept fixed and we would like to calculate the coordinates of the rotated vector with reference to the fix-kept coordinates. In this book, we resolve these confusing issues by following the convention described in Chapter 12, i.e., rotation matrix relates the known/given coordinates prior to the rotation to those resulted from the rotation through the rotation matrix ( y′i = Rij y j ). Therefore, we assign a passive convention, see next section. Assuming this definition then the second definition- so-called active and passive rotations [14]- is discussed in next section. Readers should note that some authors/references may have different definition assumed (sometimes implicitly) for the rotation matrix (e.g., y j = R ji y′i , or using the transpose of the rotation matrix considered in this book). Also, readers should note that an important fundamental property of all methods mentioned above is that the rotation procedure reduces to the rotation of one vector about another vector. This can be shown by finding the equivalent single axis and angle for any combinations of rotations [14], see Chapter 12. Examples of practical applications of the methods mentioned above are: rigid body rotation and orientation, computer graphics and animation, aircraft maneuver control, robot arm control, signal processing, quantum mechanics, special and general relativity-among others, [15], [16].
15.1 ACTIVE AND PASSIVE ROTATIONS Following Section 12.1, let’s assume, without losing generality, a 2D coor-
dinate system ( y1 , y2 ) ≡ ( x, y ) being transformed to ( y1′ , y2′ ) ≡ ( x′, y′ ) through rotation about the axis y3 ≡ z with unit vector n ≡ ni = ( 0,0,1) by an angle θ , according to R.H.R. Therefore, using the rotation matrix Rij (see x′ cosθ sin θ x Equation 12.7) we have = and z′ = z. Readers should y′ − sin θ cosθ y
Tensor-Analysis-for-Engineers_2E_CH15.indd 104
10/3/2020 10:28:35 AM
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 105
note that the rotation matrix given here relates the rotated coordinates to the existing/current ones. Now if we have a vector V = Vx , Vy given in x − y system and making angle ϕ with positive x axis then its components in x′ − y′ V cosθ sin θ Vx system, in terms of Vx and Vy , are given as x′ = . This Vy′ − sin θ cosθ Vy
(
)
type of transformation for which the vector is kept fixed and the coordinate system rotates is called passive rotation. We can also perform the opposite operation by rotating the vector and keep the original coordinate system fixed. Therefore, we keep the system x − y fixed and rotate the vector V by an angle equal to −θ about the same rotation axis (i.e., ni = ( 0,0,1 ) ) to get the vector V′. This type of transformation for which the vector is rotated, and the coordinate system is kept fixed is called active rotation. See Figure Vx′ 15.1. The components of the rotated vector are V ′ = ′ and are equivalent Vy V to x′ in their magnitudes. In other words, passive rotation through an Vy′ angle θ is equivalent to corresponding active rotation through an angle equal to −θ . For 3D rotations, we can expand the relationship between passive and active rotations by using the properties of rotation matrix, see Section 12.1, or Rij (= n,θ ) R ji ( n, −θ )
15.1
Equation 15.1 states that the rotation matrix for a given angle and axis of rotation is equal to its transpose when the rotation angle is in opposite direction but about the same rotation axis. As mentioned in this book, we consider Rij, as given by Equation 12.7, for passive rotations and use its transpose (or inverse) for active rotations. 15.1.1 Example: Active and Passive Rotations-Numerical Example Vector = V = Vx , Vy ( 3,4 ) is given in Cartesian coordinate system ( x, y ), as
(
)
shown in Figure 15.1. Show that passive and active rotations provide equivalent answers. The absolute value of angle of rotation is θ = 30° and rotation is about positive z-axis.
Tensor-Analysis-for-Engineers_2E_CH15.indd 105
10/3/2020 10:28:37 AM
106 • Tensor Analysis for Engineers, 2E Solution:
The magnitude of the vector V is V = 3 2 + 4 2 = 5 and it makes an angle 4 = ϕ tan −1 = 53.13° with the x-axis. For passive rotation, we keep the vec3 tor V fixed and rotate the coordinate system ( x, y ) by 30° in the counter clockwise direction to get the resulted coordinate system ( x′, y′ ) . Therefore, the cos30 sin 30 3 /2 rotation matrix is = − sin 30 cos30 −1/2
1/2 . The components of the 3 /2
3 /2 1/2 3 4.598 V vector V in the ( x′, y′ ) system are x′ = = . V 4 1.964 ′ y 3 /2 passive −1/2 For active rotation, we keep the coordinate system ( x, y ) fixed and rotate the vector by 30° in the clockwise direction (or −30° ) to receive the vec ° tor V ′ = Vx′, Vy′ that makes an angle ϕ=′ 53.13° − 30= 23.13° with
(
)
= = the x-axis. Therefore, the components are Vx′ 5cos ( 23.13 ) 4.598 V′ 4.598 = = and Vy′ 5sin ( 23.13 ) 1.964 , or x = . Note that Vy′ 1.964 active
the rotation matrix follows the relation given in Equation 15.1, or T cos30 sin 30 cos30 − sin 30 ° 30 Rij ( 30° ) = = R − = ) sin 30 cos30 . ji ( − sin 30 cos30 y
y
V
y
V x V
Passive
x
Active
x
FIGURE 15.1 Passive (left) and active (right) transformations for a 2D coordinate system.
Tensor-Analysis-for-Engineers_2E_CH15.indd 106
10/3/2020 10:28:42 AM
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 107
15.2 EULER ANGLES We expand the discussion presented in the previous section by considering 3D rotation of a rigid body, for example an airplane. With reference to Chapter 12, we consider the following. Let’s assume the airplane maneuvers in the sky, we can then relate its orientation, focusing on its rotation only, at any given time to its original orientation in a step-wise manner in two ways; 1) by relating its successive rotations to a fixed global Cartesian coordinate system or 2) by relating its orientations to its instantaneous body-attached coordinate system resulted from successive rotations. The latter is so-called Intrinsic, and the former is Extrinsic method. In this book, we use the intrinsic approach, [17], [18]. The extrinsic approach is more common for computer graphics applications. The concept of describing orientation of a rigid body by a limited number of successive rotations goes back to Euler. His theorem, the Euler’s rotation theorem [18], [19], [14] from a practical application point of view, states that any orientation of a rigid body can be calculated through a single rotation about a specified axis, with reference to its original orientation. This relates to what we showed and used, single axis-angle method, in Chapter 12, see Example 12.2.1. Further, Euler showed that the final orientation of a rigid body can be obtained through at most three successive rotations, so called Euler angles, [18], [20], [21]. The intrinsic method, as mentioned above, is mostly used for successive rotations. There exist several definitions for the body-attached coordinate system in terms of axis orientations. Here we use the convention mostly used in engineering by defining positive x-axis (roll) direction towards the front/cockpit, y-axis (pitch) towards the right wing, and z-axis (yaw) downward towards the ground along with positive rotations about each axis. See Figure 15.2, for which R.H.R. applies. Two immediate successive rotations around the same axis are not counted as two separate/independent rotations since we can equally have a single rotation equal to the algebraic sum of the two and hence the three-axis rotation reduces to two-axis rotation. Therefore, in 3D space, having ( x, y, z) coordinates, we can have 12 ( = 3 × 2 × 2) combinations of Euler angles as shown in Table 15.1.
Tensor-Analysis-for-Engineers_2E_CH15.indd 107
10/3/2020 10:28:42 AM
108 • Tensor Analysis for Engineers, 2E
FIGURE 15.2 Body attached coordinates and corresponding Euler angles. TABLE 15.1 Possible combination of Euler angles for rigid body rotations.
Roll group
Pitch group
Yaw group
xyz
xzy
yxz
yzx
zxy
zyx
(roll-pitchyaw)
(roll-yawpitch)
(pitch-rollyaw)
(pitch-yawroll)
(yaw-rollpitch)
(yaw-pitchroll)
xyx
xzx (roll-yawroll)
yxy
yzy
(pitch-rollpitch)
(pitch-yawpitch)
zxz (yaw-rollyaw)
(yaw-pitchyaw)
(roll-pitchroll)
zyz
Few points should be noted about the order of executing Euler angles. As previously explained the rotation matrix related to each successive rotation should be pre-multiplied by the previous ones. Considering the order of rotations as the reference. This is a direct consequence of our definition for rotation matrix that transforms previous axes to the current rotated ones. Conversely, if one defines the transpose of the rotation matrix given by Equation 12.7 (i.e., transforming current/rotated coordinates to the original ones) then the sequence of rotations through Euler angles and their multiplications of their corresponding rotation matrices to generate the final rotation matrix should be performed in the same order. For example, according to our definition, see Equation 12.7, considering the combination yxz the order of rotation
Tensor-Analysis-for-Engineers_2E_CH15.indd 108
10/3/2020 10:28:44 AM
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 109
is pitch first, followed by roll and having yaw as the last rotation, or (pitch roll-yaw). With this definition, yxz should be read in backward order, yxz to get the correct final rotation matrix resulted from multiplication of rotation matrices combination, see Section 12.2. Therefore, the related rotation T
T
T
matrices multiplication order is Rθ3 Rθ2 Rθ1 = Rθ1 Rθ2 Rθ3 , where θ1 is the first angle of rotation (for example pitch), θ 2 second (for example roll), and θ 3 third (for example yaw). The resulted rotation matrix reads
[ R ] = Rθ
3
Rθ2 Rθ1
15.2
In this book, we use the convention of index-order from 1 to 3 for successive rotations, i.e., first rotation expressed as θ1 and so on. Once the rotation matrix [ R ] is given, it is possible to calculate the corresponding Euler angles which are 12 sets, or mutations of 3 Euler angles corresponding to the rotation matrix, see Table 15.1. As discussed in Chapter 12, we can use rotation matrix [ R ] to calculate the single axis-angle rotation equivalent to three successive rotations by Euler angles. The single angle is given by Equation 12.8, repeated here for convenience, or Rii − 1 R11 + R22 + R33 1 = − . Recall that Rii is the trace of matrix [ R ]. cosθ = 2 2 2 The single axis of rotation is the eigenvalue of [ R ], refer to Example 12.2.1. Another method, and may be a quicker one, for calculating the components of a unit vector along the rotation axis, represented by the unit vector n ≡ nk, can be obtained by manipulating the diagonal elements of [ R ]. For exam2 ple, considering element R (1,1 ) = cosθ + ( n1 ) (1 − cosθ ), see Equation 12.7. R −1 R −1 2 After substituting for cosθ we receive R (1,1 ) = ii + ( n1 ) 1 − ii . 2 2 Factoring out n1 gives, after some manipulation, n1 =
2 R (1,1 ) − Rii + 1 . 3 − Rii
Therefore, for component k of the vector n = ( n1 , n2 , n3 ) , parallel to the rotation axis, we receive
Tensor-Analysis-for-Engineers_2E_CH15.indd 109
10/3/2020 10:28:46 AM
110 • Tensor Analysis for Engineers, 2E
nk =
2 R ( k, k ) − Rii + 1 = , k 1,2,3 3 − Rii
15.3
As an example, using data given in Example 12.2.1, we receive the trace of the rotation matrix, Rii = 0 and R ( k, k ) = 0. Using Equation 15.3, we receive 1 nk = 1/ 3 , or n = ± (1,1,1) . This is identical to the results obtained in 3 Example 12.2.1. 15.2.1 Example: Euler Angles-Numerical Example An airplane undergoes a zyz (yaw-pitch-yaw) maneuver. The first yaw angle is θ1 = 30, pitch angle θ 2 = 45, and the final yaw angle is θ 3 = 15. Calculate its final orientation in the sky with reference to the earth fixed coordinates and the equivalent single axis-angle rotation. Solution: We calculate the rotation matrices, using Equation 12.6 and multiply them using Equation 15.2. First yaw rotation: the rotation is about z-axis with ni = ( 0,0,1 ) and θ1 = 30 3 1 0 cos30 sin 30 0 2 2 which gives Rθ1 = − = sin 30 cos30 0 . 1 3 − 0 0 0 1 2 2 0 1 0 Pitch rotation: this rotation is about current y-axis (resulted from the previous rotation) with ni = ( 0,1,0 ) and θ 2 = 45 which gives cos 45 0 − sin 45 Rθ2 = 0 1 0 = sin 45 0 cos 45
Tensor-Analysis-for-Engineers_2E_CH15.indd 110
2 2 0 2 2
0 − 1 0
2 2 0 . 2 2
10/3/2020 10:28:48 AM
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 111
Last yaw rotation: this rotation is about current z-axis (resulted from the immediate previous rotation) with ni = ( 0,0,1 ) and θ 3 = 15 which gives 3 +1 3 −1 0 cos15 sin15 0 2 2 2 2 Rθ3 = sin15 cos15 0 − = . 3 −1 3 +1 − 0 0 0 1 2 2 2 2 0 0 1 For final rotation matrix, we first calculate 6 2 2 1 2 3 4 0 4 0 − 2 2 2 2 1 3 Rθ2 Rθ1 = 0 1 0 1 = − 3 2 0 2 − 2 2 2 2 0 6 2 0 1 2 2 0 4 4 to pre-multiply Rθ3 to this result, or 3 +1 2 2 [ R ] = 3 − 1 − 2 2 0
Tensor-Analysis-for-Engineers_2E_CH15.indd 111
6 3 −1 0 4 2 2 1 3 +1 − 0 2 2 2 6 0 1 4
3+ 2 + 3 − 6 8 −3 − 2 + 3 − 6 = 8 6 4
2 2 3 0 − 2 2 2 Rθ2 Rθ1 = 0 1 0 1 − 2 2 2 2 − 0 2 2 2 0 0 . The final operation is 2 2
2 4 3 2 2 4
−
1 2 3 2 0
2 2 0 2 2
1+3 2 + 3 − 6 8 1+3 2 − 3 + 6 8 2 4
1+ 3 4 −1 + 3 4 2 2
−
10/3/2020 10:28:49 AM
0 = 0 1
112 • Tensor Analysis for Engineers, 2E
0.4621 0.5657 −0.6830 ≅ −0.6415 0.7450 0.1830 . 0.6124 0.3536 0.7071
The corresponding single rotation angle is, using Equation 12.8, 2 1 = − = 62.8°. The elements of the unit vector along the single rotaθ cos−1 2 4 2 × 0.4621 − 1.9142 + 1 tion axis are obtained using 15.3, or n= = 9.6 × 10 −2, 1 3 − 1.9142 2 × 0.7071 − 1.9142 + 1 +1 2 × 0.745 − 1.9142= = 0.6786 0.7282, and n3 = 3 − 1.9142 3 − 1.9142 0.6786. Answers are rounded. In other words, we can obtain the final orientation of the airplane by rotating it through an angle of 62.8° about an axis in the direction of nk = ( 0.096, 0.7282, 0.6786 ) with reference to the original coordinate system ( x, y, z ). The successive rotations are shown in Figure 15.3.
= n2 2 × 0.7071 − 1.9142 + 1 3 − 1.9142
FIGURE 15.3 Successive rotation of an airplane, Euler angels left-to-right, yaw=30, pitch=45, yaw=15 degrees.
Tensor-Analysis-for-Engineers_2E_CH15.indd 112
10/3/2020 10:28:50 AM
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 113
It seems useful to refer the readers to some animation tools for Euler angles. The following Apps, available through the following links, could be used for demonstrating the Euler angles. “Controlling Airplane Flight” [22], “Euler Angles for Space Shuttle” [23], and “Euler Angles” [24].
15.3 CATEGORIZING EULER ANGLES There are several ways possible to categorize Euler angles. With reference to Table 15.1, we observe that there are two sets of Euler angles combination. One set is those six combinations that contain repeated rotations (1st and 3rd) about similar (but not the same) axes ( xyx, xzx, yxy, yzy, zxz, zyz), the socalled proper Euler angles and another set that has different axis of rotations for all three rotations (xyz, xzy, yxz, yzx, zxy, zyx), the so-called Tait-Bryan angles. The latter is commonly used in engineering, specifically in aerospace discipline [18], [19]. It is useful to collect all Euler angles types and their corresponding rotation matrices, as shown in Table 15.2 and Table 15.3. The corresponding single-axis rotation matrix of each case can be calculated as well, like those in Section 15.2.1.
Tensor-Analysis-for-Engineers_2E_CH15.indd 113
10/3/2020 10:28:51 AM
Tensor-Analysis-for-Engineers_2E_CH15.indd 114
yaw
roll
pitch
pitch
yaw
yaw
yzy
zxz
zyz
yaw
roll
roll
xzx
pitch
pitch
roll
xyx
θ2
yxy
θ1
yaw
yaw
pitch
pitch
roll
roll
θ3
0 0 1 0 cos sin ( ) ( ) 0 − sin ( ) cos ( )
Rotation matrixstep
Proper Euler Angles
Successive rotation-angles
Roll: ni = (1,0,0 )
unit vectorrotation axis cos ( ) 0 − sin ( ) 1 0 0 sin ( ) 0 cos ( )
= xyx
= xzx
= yxy
= yzy
= zxz
= zyz
0 0 cos (θ 2 ) sin (θ 2 ) 0 1 0 0 1 0 cos θ sin θ sin θ cos θ 0 0 cos θ sin − ( ) ( ) ( ) (θ1 ) ( ) ( ) 3 3 2 2 1 0 − sin (θ 3 ) cos (θ 3 ) 0 0 1 0 − sin (θ1 ) cos (θ1 ) 0 0 cos (θ1 ) 0 − sin (θ1 ) cos (θ 3 ) 0 − sin (θ 3 ) 1 0 1 0 0 cos θ sin 1 0 ( ) (θ2 ) 0 2 sin (θ 3 ) 0 cos (θ 3 ) 0 − sin (θ 2 ) cos (θ 2 ) sin (θ1 ) 0 cos (θ1 ) cos (θ 3 ) 0 − sin (θ 3 ) cos (θ 2 ) sin (θ 2 ) 0 cos (θ1 ) 0 − sin (θ1 ) 1 0 1 0 0 − sin (θ 2 ) cos (θ 2 ) 0 0 0 0 1 sin (θ1 ) 0 cos (θ1 ) sin (θ 3 ) 0 cos (θ 3 ) 0 0 cos (θ1 ) sin (θ1 ) 0 cos (θ 3 ) sin (θ 3 ) 0 1 sin θ cos θ 0 0 cos θ sin − ( ) (θ2 ) − sin (θ1 ) cos (θ1 ) 0 ( ) ( ) 3 3 2 0 0 1 0 − sin (θ 2 ) cos (θ 2 ) 0 0 1 cos (θ 3 ) sin (θ 3 ) 0 cos (θ 2 ) 0 − sin (θ 2 ) cos (θ1 ) sin (θ1 ) 0 1 0 − sin (θ 3 ) cos (θ 3 ) 0 0 − sin (θ1 ) cos (θ1 ) 0 0 0 1 sin (θ 2 ) 0 cos (θ 2 ) 0 0 1
Rotation matrix
cos ( ) sin ( ) 0 − sin ( ) cos ( ) 0 0 0 1
Yaw: ni = ( 0,0,1 )
0 0 cos (θ 2 ) 0 − sin (θ 2 ) 1 0 0 1 0 cos θ sin θ 0 1 0 0 cos θ sin ( ) (θ1 ) ( ) ( ) 3 3 1 0 − sin (θ 3 ) cos (θ 3 ) sin (θ 2 ) 0 cos (θ 2 ) 0 − sin (θ1 ) cos (θ1 )
Matrices multiplication for successive rotations
Pitch: ni = ( 0,1,0 )
TABLE 15.2 Proper Euler angles and rotation matrices.
114 • Tensor Analysis for Engineers, 2E
10/3/2020 10:28:53 AM
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 115
where,
xyx
cos (θ 2 ) sin (θ1 ) sin (θ 2 ) −cos (θ1 ) sin (θ 2 ) cos (θ 3 ) cos (θ1 ) cos (θ 3 ) sin (θ1 ) sin (θ 2 ) sin (θ 3 ) − sin (θ 3 ) cos (θ 2 ) sin (θ1 ) + sin (θ 3 ) cos (θ 2 ) cos (θ1 ) = − sin (θ 3 ) cos (θ1 ) cos (θ 3 ) cos (θ 2 ) cos (θ1 ) cos (θ 3 ) sin (θ 2 ) − cos (θ 3 ) cos (θ 2 ) sin (θ1 ) − sin (θ 3 ) sin (θ1 )
Conversely; for a given single-axis rotation matrix the corresponding 1 − R 2 (1,1 ) R (1,2 ) −1 −1 , and Euler angles are: θ1 = tan , θ 2 = tan R 1,1 R 1,3 − ( ) ( ) R 2,1 ( ) θ 3 = tan −1 R ( 3,1 )
xzx
cos (θ 2 ) cos (θ1 ) sin (θ 2 ) sin (θ1 ) sin (θ 2 ) − sin (θ1 ) sin (θ 3 ) cos (θ1 ) sin (θ 3 ) − sin (θ 2 ) cos (θ 3 ) = + cos (θ 3 ) cos (θ 2 ) cos (θ1 ) + sin (θ1 ) cos (θ 2 ) cos (θ 3 ) − sin (θ1 ) cos (θ 3 ) cos (θ1 ) cos (θ 3 ) sin (θ 2 ) sin (θ 3 ) − − cos θ cos θ sin θ sin θ cos θ sin θ ( ) ( ) ( ) ( ) ( ) ( ) 1 2 3 3 2 1
Conversely; for a given single-axis rotation matrix the corresponding 1 − R 2 (1,1 ) R (1,3 ) −1 , and Euler angles are: θ1 = tan −1 , tan θ = 2 R (1,1 ) R (1,2 ) R ( 3,1 ) θ 3 = tan −1 − R ( 2,1 )
yxy
cos (θ 3 ) cos (θ1 ) sin (θ 2 ) sin (θ 3 ) − cos (θ 3 ) sin (θ1 ) − cos (θ1 ) cos (θ 2 ) sin (θ 3 ) − sin (θ 3 ) cos (θ 2 ) sin (θ1 ) = sin (θ1 ) sin (θ 2 ) cos (θ 2 ) cos (θ1 ) sin (θ 2 ) sin (θ 3 ) cos (θ1 ) − sin (θ 2 ) cos (θ 3 ) cos (θ 3 ) cos (θ 2 ) cos (θ1 ) + cos (θ ) cos (θ ) sin (θ ) − sin (θ 3 ) sin (θ1 ) 3 2 1
Conversely; for a given single-axis rotation matrix the corresponding 1 − R 2 ( 2,2 ) R ( 2,1 ) −1 , and tan θ = Euler angles are: θ1 = tan −1 , 2 R ( 2,2 ) R ( 2,3 ) R (1,2 ) θ 3 = tan −1 − R ( 3,2 )
Tensor-Analysis-for-Engineers_2E_CH15.indd 115
10/3/2020 10:28:55 AM
116 • Tensor Analysis for Engineers, 2E
yzy
− sin (θ 3 ) sin (θ1 ) sin (θ 2 ) cos (θ 3 ) − sin (θ 3 ) cos (θ1 ) − cos (θ 3 ) cos (θ 2 ) sin (θ1 ) + cos (θ 3 ) cos (θ 2 ) cos (θ1 ) = − cos (θ1 ) sin (θ 2 ) cos (θ 2 ) sin (θ1 ) sin (θ 2 ) sin (θ1 ) cos (θ 3 ) sin (θ 2 ) sin (θ 3 ) − sin (θ 3 ) cos (θ 2 ) sin (θ1 ) + cos (θ ) cos (θ ) sin (θ ) + cos (θ 3 ) cos (θ1 ) 1 2 3
Conversely; for a given single-axis rotation matrix the corresponding 1 − R 2 ( 2,2 ) R ( 2,3 ) −1 −1 , and Euler angles are: θ1 = tan , θ 2 = tan R ( 2,2 ) − R ( 2,1 ) R ( 3,2 ) θ 3 = tan −1 R (1,2 )
zxz
cos (θ 3 ) cos (θ1 ) cos (θ 3 ) sin (θ1 ) sin (θ 3 ) sin (θ 2 ) − sin (θ 3 ) cos (θ 2 ) sin (θ1 ) + sin (θ 3 ) cos (θ 2 ) cos (θ1 ) = − sin (θ 3 ) cos (θ1 ) cos (θ 3 ) cos (θ 2 ) cos (θ1 ) cos (θ 3 ) sin (θ 2 ) − sin (θ 3 ) sin (θ1 ) − sin (θ1 ) cos (θ 2 ) cos (θ 3 ) sin (θ1 ) sin (θ 2 ) − cos (θ1 ) sin (θ 2 ) cos (θ 2 )
Conversely; for a given single-axis rotation matrix the corresponding 1 − R 2 ( 3,3 ) R ( 3,1 ) −1 , and Euler angles are: θ1 = tan −1 , tan θ = 2 R ( 3,3 ) − R ( 3,2 ) R (1,3 ) θ 3 = tan −1 R ( 2,3 )
zyz
− sin (θ 3 ) sin (θ1 ) sin (θ 3 ) cos (θ1 ) − cos (θ 3 ) sin (θ 2 ) + cos (θ 3 ) cos (θ 2 ) cos (θ1 ) + sin (θ1 ) cos (θ 2 ) cos (θ 3 ) = − sin (θ1 ) cos (θ 3 ) − sin (θ 3 ) cos (θ 2 ) sin (θ1 ) sin (θ 3 ) sin (θ 2 ) + cos (θ 3 ) cos (θ1 ) − sin (θ 3 ) cos (θ 2 ) cos (θ1 ) cos (θ1 ) sin (θ 2 ) sin (θ1 ) sin (θ 2 ) cos (θ 2 )
Conversely; for a given single-axis rotation matrix the corresponding 1 − R 2 ( 3,3 ) R ( 3,2 ) −1 , and Euler angles are: θ1 = tan −1 , tan θ = 2 R ( 3,3 ) R ( 3,1 ) R ( 2,3 ) θ 3 = tan −1 − R (1,3 ) Tensor-Analysis-for-Engineers_2E_CH15.indd 116
10/3/2020 10:28:57 AM
zyx
zxy
yzx
yxz
xzy
xyz
yaw
yaw
pitch
pitch
roll
roll
θ1
pitch
roll
yaw
roll
yaw
pitch
θ2
roll
pitch
roll
yaw
pitch
yaw
θ3
= xyz
= xzy
= yxz
= yzx
= zxy
= zyx
0 0 cos (θ 3 ) 0 − sin (θ 3 ) cos (θ 2 ) sin (θ 2 ) 0 1 0 1 0 sin θ cos θ 0 0 cos θ sin − ( ) ( ) ( ) (θ1 ) 2 2 1 sin (θ 3 ) 0 cos (θ 3 ) 0 0 1 0 − sin (θ1 ) cos (θ1 ) 0 0 cos (θ1 ) 0 − sin (θ1 ) cos (θ 3 ) sin (θ 3 ) 0 1 sin θ cos θ 0 0 cos θ sin 1 0 − ( ) (θ2 ) 0 ( ) ( ) 3 3 2 0 0 1 0 − sin (θ 2 ) cos (θ 2 ) sin (θ1 ) 0 cos (θ1 ) 0 0 cos (θ 2 ) sin (θ 2 ) 0 cos (θ1 ) 0 − sin (θ1 ) 1 0 cos θ sin 1 0 ( 3) (θ3 ) − sin (θ2 ) cos (θ2 ) 0 0 0 − sin (θ 3 ) cos (θ 3 ) 0 0 1 sin (θ1 ) 0 cos (θ1 ) 0 0 cos (θ1 ) sin (θ1 ) 0 cos (θ 3 ) 0 − sin (θ 3 ) 1 1 0 0 0 cos (θ 2 ) sin (θ 2 ) − sin (θ1 ) cos (θ1 ) 0 sin (θ 3 ) 0 cos (θ 3 ) 0 − sin (θ 2 ) cos (θ 2 ) 0 0 1 0 0 cos (θ 2 ) 0 − sin (θ 2 ) cos (θ1 ) sin (θ1 ) 0 1 0 cos θ sin 1 0 ( 3) (θ3 ) 0 − sin (θ1 ) cos (θ1 ) 0 0 − sin (θ 3 ) cos (θ 3 ) sin (θ 2 ) 0 cos (θ 2 ) 0 0 1
Rotation matrix
cos ( ) sin ( ) 0 − sin ( ) cos ( ) 0 0 0 1
Yaw: ni = ( 0,0,1 )
0 0 cos (θ 3 ) sin ( 3 ) 0 cos (θ 2 ) 0 − sin (θ 2 ) 1 sin θ cos θ 0 0 1 0 0 cos θ sin − ( ) (θ1 ) ( ) ( ) 3 3 1 0 0 1 sin (θ 2 ) 0 cos (θ 2 ) 0 − sin (θ1 ) cos (θ1 )
Matrices multiplication for successive rotations
cos ( ) 0 − sin ( ) 1 0 0 sin ( ) 0 cos ( )
0 0 1 0 cos sin ( ) ( ) 0 − sin ( ) cos ( )
Rotation matrixstep
Tait-Bryan Angles
Tensor-Analysis-for-Engineers_2E_CH15.indd 117
Successive rotation-angles
Pitch: ni = ( 0,1,0 )
Roll: ni = (1,0,0 )
unit vectorrotation axis
TABLE 15.3 Tait-Bryan angles and rotation matrices.
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 117
10/3/2020 10:29:00 AM
118 • Tensor Analysis for Engineers, 2E where,
xyz
cos (θ1 ) sin (θ 3 ) sin (θ 3 ) sin (θ1 ) cos (θ 2 ) cos (θ 3 ) + cos (θ 3 ) sin (θ 2 ) sin (θ1 ) − cos (θ 3 ) sin (θ 2 ) cos (θ1 ) = − cos (θ 2 ) sin (θ 3 ) − sin (θ 3 ) sin (θ 2 ) sin (θ1 ) cos (θ 3 ) sin (θ1 ) + cos (θ 3 ) cos (θ1 ) + sin (θ 3 ) sin (θ 2 ) cos (θ1 ) sin (θ 2 ) − cos (θ 2 ) sin (θ1 ) cos (θ 2 ) cos (θ1 )
Conversely; for a given single-axis rotation matrix the correspond R ( 3,1 ) − R ( 3,2 ) −1 , ing Euler angles are: θ1 = tan −1 , θ 2 = tan 1 − R 2 ( 3,1 ) R ( 3,3 ) − R ( 2,1 ) θ 3 = tan −1 R (1,1 )
xzy
sin (θ 3 ) sin (θ1 ) − sin (θ 3 ) cos (θ1 ) cos (θ 3 ) cos (θ 2 ) + cos (θ 3 ) sin (θ 2 ) cos (θ1 ) + cos (θ 3 ) sin (θ 2 ) sin (θ1 ) = − sin (θ 2 ) cos (θ1 ) cos (θ 2 ) sin (θ1 ) cos (θ 2 ) − cos (θ 3 ) sin (θ1 ) sin (θ 3 ) sin (θ 2 ) sin (θ1 ) sin (θ 3 ) cos (θ 2 ) + cos (θ1 ) sin (θ 2 ) sin (θ 3 ) + cos (θ 3 ) cos (θ1 )
Conversely; for a given single-axis rotation matrix the correspond − R ( 2,1 ) R ( 2,3 ) −1 , θ = tan ing Euler angles are: θ1 = tan −1 , 2 1 − R 2 ( 2,1 ) R ( 2,2 ) R ( 3,1 ) θ 3 = tan −1 R (1,1 )
yxz
cos (θ 3 ) cos (θ1 ) cos (θ 2 ) sin (θ 3 ) − cos (θ 3 ) sin (θ1 ) + sin (θ 3 ) sin (θ 2 ) cos (θ1 ) + sin (θ 3 ) sin (θ 2 ) sin (θ1 ) = − sin (θ 3 ) cos (θ1 ) cos (θ 2 ) cos (θ 3 ) sin (θ 3 ) sin (θ1 ) + cos (θ 3 ) sin (θ 2 ) cos (θ1 ) + cos (θ 3 ) sin (θ 2 ) sin (θ1 ) sin (θ1 ) cos (θ 2 ) cos (θ1 ) cos (θ 2 ) − sin (θ 2 )
Tensor-Analysis-for-Engineers_2E_CH15.indd 118
10/3/2020 10:29:02 AM
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 119
Conversely; for a given single-axis rotation matrix the corresponding − R ( 3,2 ) R ( 3,1 ) −1 , and Euler angles are: θ1 = tan −1 , θ 2 = tan 1 − R 2 ( 3,2 ) R ( 3,3 ) R (1,2 ) θ 3 = tan −1 R ( 2,2 )
yzx
cos (θ1 ) cos (θ 2 ) sin (θ 2 ) − sin (θ1 ) cos (θ 2 ) sin (θ 3 ) sin (θ1 ) cos (θ 3 ) cos (θ 2 ) sin (θ 3 ) cos (θ1 ) = − cos (θ 3 ) cos (θ 2 ) cos (θ1 ) + cos (θ 3 ) sin (θ 2 ) sin (θ1 ) cos (θ 3 ) sin (θ1 ) − sin (θ 3 ) cos (θ 2 ) − sin (θ 3 ) sin (θ 2 ) sin (θ1 ) + cos (θ 3 ) cos (θ1 ) + cos (θ1 ) sin (θ 2 ) sin (θ 3 )
Conversely; for a given single-axis rotation matrix the corresponding R (1,2 ) − R (1,3 ) −1 , and Euler angles are: θ1 = tan −1 , θ 2 = tan 1 − R 2 (1,2 ) R (1,1 ) − R ( 3,2 ) θ 3 = tan −1 R ( 2,2 )
zxy
cos (θ1 ) cos (θ 3 ) cos (θ 3 ) sin (θ1 ) − cos (θ 2 ) sin (θ 3 ) − sin (θ1 ) sin (θ 2 ) sin (θ 3 ) + sin (θ 3 ) sin (θ 2 ) cos (θ1 ) = − sin (θ1 ) cos (θ 2 ) cos (θ1 ) cos (θ 2 ) sin (θ 2 ) sin (θ 3 ) cos (θ1 ) − cos (θ 3 ) sin (θ 2 ) cos (θ1 ) cos (θ 2 ) cos (θ 3 ) + sin (θ ) sin (θ ) cos (θ ) + sin (θ1 ) sin (θ 3 ) 1 2 3
Conversely; for a given single-axis rotation matrix the corresponding R ( 2,3 ) − R ( 2,1 ) −1 , and Euler angles are: θ1 = tan −1 , θ 2 = tan 1 − R 2 ( 2,3 ) R ( 2,2 ) −R (1,3 ) θ 3 = tan −1 R ( 3,3 )
Tensor-Analysis-for-Engineers_2E_CH15.indd 119
10/3/2020 10:29:03 AM
120 • Tensor Analysis for Engineers, 2E
zyx
cos (θ1 ) cos (θ 2 ) sin (θ1 ) cos (θ 2 ) − sin (θ 2 ) − cos (θ 3 ) sin (θ1 ) cos (θ 3 ) cos (θ1 ) cos (θ 2 ) sin (θ 3 ) = + cos (θ1 ) sin (θ 2 ) sin (θ 3 ) + sin (θ 3 ) sin (θ 2 ) sin (θ1 ) sin (θ 3 ) sin (θ1 ) − sin (θ 3 ) cos (θ1 ) cos (θ 2 ) cos (θ 3 ) + cos (θ ) sin (θ ) cos (θ ) + sin (θ ) sin (θ ) cos (θ ) 3 2 1 1 2 3
Conversely; for a given single-axis rotation matrix the corresponding − R (1,3 ) − R (1,2 ) −1 , and Euler angles are: θ1 = tan −1 , θ 2 = tan 1 − R 2 (1,3 ) R (1,1 ) R ( 2,3 ) θ 3 = tan −1 R ( 3,3 ) 15.3.1 Example: Euler Angles- Yaw-Pitch-Yaw, Numerical example Repeat the example given in Section 15.2.1 using relation zyz, given in Table 15.2. Solution: using Table 15.2, we have θ1 = 30 for first yaw rotation, θ 2 = 45 for pitch rotation, and θ1 = 15 for third yaw rotation. Substituting in relation for zyz gives
zyz
− sin (15 ) sin ( 30 ) sin (15 ) cos ( 30 ) − cos (15 ) sin ( 45 ) + cos (15 ) cos ( 45 ) cos ( 30 ) + sin ( 30 ) cos ( 45 ) cos (15 ) = − sin ( 30 ) cos (15 ) − sin (15 ) cos ( 45 ) sin ( 30 ) sin (15 ) sin ( 45 ) + cos (15 ) cos ( 30 ) − sin (15 ) cos ( 30 ) cos ( 45 ) cos 30 sin 45 sin 30 sin 45 cos 45 ( ) ( ) ( ) ( ) ( ) 0.4621 0.5657 −0.6830 −0.6415 0.7450 0.1830 . These results are identical with those 0.6124 0.3536 0.7071 from the example given in Section 15.2.1 Therefore, x′ y′ = z′
0.4621 0.5657 −0.6830 x , where ( x′ , y′, z′) are coordinates −0.6415 0.7450 0.1830 y 0.6124 0.3536 0.7071 z
Tensor-Analysis-for-Engineers_2E_CH15.indd 120
10/3/2020 10:29:05 AM
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 121
of a fixed given point in space with initial coordinates ( x, y, z). In other words, the initially body-attached axes ( x, y, z) orient themselves to the ( x′ , y′, z′) after the airplane goes through the given successive yaw-pitch-yaw rotations.
15.4 GIMBAL LOCK-EULER ANGLES LIMITATION For a wide range of applications, the Euler angles method provide us a useful tool for calculating a rigid body orientation. However, for certain values of rotation angles the process breaks down and ran into a mathematical ”singularity” situation. This is usually referred to gimbal lock. This happens when the rigid body loses one of its three rotational degrees of freedom and rotation of two axes coincide when the rotation about the third remaining π axis is some integer multiple of . To see the mathematical explanation 2 and implication of gimbal lock, we use one of the cases from Table 15.3, for example roll-pitch-yaw with final rotation matrix xyz, repeated here for convenience.
xyz
cos (θ1 ) sin (θ 3 ) sin (θ 3 ) sin (θ1 ) cos (θ 2 ) cos (θ 3 ) + cos (θ 3 ) sin (θ 2 ) sin (θ1 ) − cos (θ 3 ) sin (θ 2 ) cos (θ1 ) = − cos (θ 2 ) sin (θ 3 ) − sin (θ 3 ) sin (θ 2 ) sin (θ1 ) cos (θ 3 ) sin (θ1 ) + cos (θ 3 ) cos (θ1 ) + sin (θ 3 ) sin (θ 2 ) cos (θ1 ) − cos (θ 2 ) sin (θ1 ) sin (θ 2 ) cos (θ 2 ) cos (θ1 )
Let the pitch angle be θ 2 = metric simplifications, xyz
π 2
. Therefore, we receive, after some trigono-
0 sin (θ1 + θ 3 ) −cos (θ1 + θ 3 ) =0 cos (θ1 + θ 3 ) sin (θ1 + θ 3 ) . Writing the transformed coordi1 0 0
x′ 0 sin ( β ) −cos ( β ) x nates, we receive y′ = 0 cos ( β ) sin ( β ) y . Where β= θ1 + θ 3. z 0 0 z′ 1
Tensor-Analysis-for-Engineers_2E_CH15.indd 121
10/3/2020 10:29:07 AM
122 • Tensor Analysis for Engineers, 2E Writing the system of equations = gives x′ y sin ( β ) − zcos= ( β ), y′ ycos ( β ) + z sin ( β ) = y′ ycos ( β ) + z sin ( β ), and z′ = x . Therefore, the rotated z-axis and x-axis are coincided, or so-called locked. Obviously, the roll angle θ1 and yaw angle θ 3 can take on several different values but their changes are indistinguishable since angle β changes value if either of θ1 or θ 3 changes. When pitch angle is equal to
π
(or in general, some odd integer multiple of
π
) then the rigid body 2 2 degrees of freedom is reduced from three to two, since z′ = x and yaw and roll rotations provide the same result in terms of the rigid body orientation. Similarly, other rotation matrices as listed in Table 15.2 and Table 15.3 can be used to show their corresponding gimbal lock orientation, [25]. It seems useful to follow the successive rotations step-by-step as well, for helping with understanding the gimbal lock. The first rotation is roll by 0 0 x x1 1 angle θ1 which results in orientation y1 = 0 cos ( θ1 ) sin (θ1 ) y. z1 0 − sin (θ1 ) cos (θ1 ) z Obviously, these leaves x-axis intact, x1 = x and rotates y − z plane about x-axis. The second rotation is pitch by angle θ 2 which results in orientation x2 cos (θ 2 ) 0 − sin (θ 2 ) x1 . Obviously, these leaves y -axis intact, 1 0 1 y2 = 0 y1 z sin (θ ) 0 cos (θ ) z 2 2 2 1
π
y2 = y1 and rotates x1 − z1 plane about y1-axis. Now for the value of θ 2 = 2 x2 0 0 −1 x1 we receive y2 = 0 1 0 y1 , or x2 = − z1, y2 = y1, and z= x= x. 2 1 z 1 0 0 z 2 1 The latter relation shows that original x-axis is coinciding with z2 axis, or so-called gimbal-locked. The last rotation is yaw by angle θ 3 which results x3 cos (θ 3 ) sin ( θ 3 ) 0 x2 in orientation y3 = − sin (θ 3 ) cos (θ 3 ) 0 y2 . Obviously, these leaves 0 0 1 z2 z3
z= x= x and rotates x2 − y2 plane about z2-axis. z3-axis intact, z= 3 2 1
Tensor-Analysis-for-Engineers_2E_CH15.indd 122
10/3/2020 10:29:11 AM
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 123
Performing the matrix multiplication using all three matrices, produces 0 0 x x′ x3 cos (θ 3 ) sin ( θ 3 ) 0 0 0 −1 1 ′ y = y3 = − sin (θ 3 ) cos (θ 3 ) 0 0 1 0 0 cos ( θ1 ) sin (θ1 ) y , z′ z 0 0 1 1 0 0 0 − sin (θ1 ) cos (θ1 ) z 3 x′ 0 sin (θ1 + θ 3 ) −cos (θ1 + θ 3 ) x or y′ =0 cos (θ1 + θ 3 ) sin (θ1 + θ 3 ) y, or again showing the loss of z′ 1 z 0 0 one degrees of freedom out of three as z′ = x , or gimbal lock. To recover the lost degree of freedom due to gimbal lock situation we can perform manual maneuver, if/when possible to rotate the rigid body as smoothly as possible out of the gimbal lock position. This approach is not practical, and the control system should be programmed such that close to values of rotation angle that creates gimbal lock a warning sent out and users become aware of it. As shown in the animation video, cited above/below, we can use other more robust methods for removing gimbal lock. This method makes use of quaternions. We discuss the mathematics of quaternions and their applications in rigid body rotation and matrix transformation, in the following sections. Readers interested in discussion about gimbal lock can consult with references cited [14], [26], [18]. Readers may also find it helpful to consult with the following lecture series, specifically those ones related to rigid body kinematics [27]. Also using Wolfram Alpha, readers can visualize a given quaternion, along with some related computations. An example for animation of a gimbal lock position, which could be helpful to visualize this phenomenon, can be found at references [28] and [29].
15.5 QUATERNIONS-APPLICATIONS FOR RIGID BODY ROTATION A quaternion is defined as a 4D complex number. Mathematically, it is represented with a combination of a scalar and a vector-like complex number. Its development started when William R. Hamilton was visioning and working on extending the complex numbers to more than 2D space, a real and
Tensor-Analysis-for-Engineers_2E_CH15.indd 123
10/3/2020 10:29:12 AM
124 • Tensor Analysis for Engineers, 2E a complex numbers dimension. He first struggled adding one dimension to the complex number (i.e., defining it in 3D space) and ran into difficulties for performing basic algebraic operations, like multiplications, etc. For example, the product of two imaginary numbers ij is not defined in 3D space when having one real and two imaginary dimensions. Then he realized that adding an additional dimension to the complex numbers and defining it in 3D space would solve the problem, hence Quaternions are defined as complex numbers in 4D space, [30]. This newly discovered mathematical quantity, the quaternions, found to be very useful in several fields of science/ physics and engineering such as kinematics of rigid body orientation, robotics, aerospace, computer graphics/animation and gaming software, relativity (General and Special), quantum mechanics. In this section, we discuss their definition, algebraic operations, and their applications for rigid body rotation [18], [21]. 15.5.1 Definition-Quaternion A quaternion can be defined as the combination of a scalar and a vector, as q = ( q0 , q ) . The vector-like part is in principle defined as a 3D complex number when using their multiplication relation defined as given by Equation 15.4. It is well-known that in 1843 Hamilton, excited with his spark discovery, engraved this expression as a message on Broome Bridge in Dublin, Ireland [31], [32]. i2 = j2 = k2 = ijk = −1
15.4
Therefore, we can write a quaternion q as q =q0 + iq1 + jq2 + kq3 ≡ ( q0 , q1 , q2 , q3 )
15.5
where, q0 is a real number and remaining terms are complex numbers. For example, 2 − i + 3 j + 2 k. Using Equation 15.5, we can write complex num= 0,1,0,0 ) , j ( 0,0,1,0 = bers i, j, k as quaternions, or i (= ) , and k ( 0,0,0,1). Note that the similarity between ( i, j, k), here defined as complex num bers with those of unit vectors ( e1 , e2 , e3 ) ≡ i, j, k that justifies the vectorlike representation of a quaternion and hence to some extent simplifies the mathematical expressions resulted from algebraic manipulation of quaternions. The relations for the product of these complex quantities follows the cross-product of unit vectors, except for those of identical unit vectors
(
Tensor-Analysis-for-Engineers_2E_CH15.indd 124
)
10/3/2020 10:29:13 AM
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 125
1, etc.). Please note i2 = ii is an 0 and i ⋅ i = (e.g., i2 = −1, whereas i × i = expression representing arithmetic multiplication of a complex number i by itself, hence manipulation of quaternions is relatively easier than those of vectors.
All possible combinations of complex numbers products are listed in Table 15.4, with clockwise multiplication considered to be positive. From these relations we can conclude that these relations are not commutative, for − ij, etc. example, ij ≠ ji = TABLE 15.4 Hamilton relations, quaternions multiplication rule.
k
i
i = ( 0,1,0,0 )
j = ( 0,0,1,0 )
k = ( 0,0,0,1 )
i
i2 = −1
ij = k
ik = − j
j
ji = − k
j2 = −1
jk = i
k
ki = j
kj = − i
k2 = −1
+ j
In terms of geometrical visualization, it is difficult to vision a 4D space since we are accustomed to visioning 3D space intuitively. It may help to imagine the surface of a hypersphere as representation of a surface defined in 4D space, like a 2D surface defined in a 3D space [21]. Readers may find it useful to watch the video clip cited in reference [33], demonstrating some animations of the geometry of a quaternion in terms of hyperspheres. 15.5.1.1 Example: Derivation of Hamilton Quaternion Relations Using Equation 15.4, derive the identities given in Table 15.4. Recall that quaternions are not commutative. Solution: We start with ijk = −1 and multiply both sides by i to receive i2 jk = − i . −1 Hence, jk = i. Similarly, kj = − i. Now multiply both sides by j to receive ji jk = − j. Hence, − i j2 k = − j or ik = − j. Similarly, ki = j. − ij −1
Tensor-Analysis-for-Engineers_2E_CH15.indd 125
10/3/2020 10:29:15 AM
126 • Tensor Analysis for Engineers, 2E Now multiply both sides by k to receive ki k = k or jk = − k. Hence, i kj − ik − ij k2 = k. Hence, ij = k. Similarly, ji = − k.
− jk
−1
15.5.2 Quaternions-Basic Algebraic Operations and Properties We consider two quaternions, q =q0 + iq1 + jq2 + kq3 and
p = p0 + ip1 + jp2 + kp3
p = p0 + ip1 + jp2 + kp3 . The summation/subtraction operation on these quantities q ± p= ( q0 ± p0 ) + i(q1 ± p1 ) + j(q2 ± p2 ) + k ( q3 ± p3 is straight forward and produces p= ( q0 ± p0 ) + i(q1 ± p1 ) + j(q2 ± p2 ) + k ( q3 ± p3 ) . Obviously, for more than two quaternions it can be extended to include as many as involved in the operation. However, the multiplication operation is not as straight forward. As a matter of fact, it was this operation that held Hamilton back for a while before he realized that a 4D space is required to properly define quaternions. The multiplication operation is as follow: qp = ( q0 + iq1 + jq2 + kq3 )( p0 + ip1 + jp2 + kp3 ) . Performing the term-by-term multiplications and using the relations given in Table 15.4, after some rearranging, we receive qp = q0 p0 − ( q1 p1 + q2 p2 + q3 p3 ) + q0 ( ip1 + jp2 + kp3 ) + p0 ( iq1 + jq2 + kq3 )
+ i ( q2 p3 − q3 p2 ) + j ( q3 p1 − q1 p3 ) + k ( q1 p2 − q2 p1 ) 15.6
This is a long expression and some simplifications is justified. We can make use of properties of the vector-like part of the quaternion, q = iq1 + jq2 + kq3 and p = ip1 + jp2 + kp3, and their dot and cross products when the complex numbers i, j, k are interpreted as unit vectors. Or q ⋅ p= q1 p1 + q2 p2 + q3 p3 × p i ( q2 p3 − q3 p2 ) + j ( q3 p1 − q1 p3 ) + k ( q1 p2 − q2 p1 ) . Substituting for and q= these expressions back into multiplication result (Equation 15.6), we receive qp = ( q0 p0 − q ⋅ p ) + ( q0 p + p0 q + q × p ) 15.7 scalar
vector − complex number
Equation 15.7 clearly demonstrates that qp ≠ pq, since q × p ≠ p × q.
The author’s caution is to note the dual interpretations of complex numbers i, j, k as unit vectors when the complex part of a quaternion is treated as 2 0 for vec1 or i × i = a vector (for example i = −1 but we treat as i ⋅ i = tor operation). Readers should note that simply the results of the product operations are the same, not necessarily the term-by-term operation, hence Equation 15.7 provides us with the same answer as that of Equation 15.6.
Tensor-Analysis-for-Engineers_2E_CH15.indd 126
10/3/2020 10:29:19 AM
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 127
However, Equation 15.7 is presented in more compact memorable form of the quaternions’ product expression. It is useful to define conjugate of the quaternion, q∗ = ( q0 − q ) = q0 − ( iq1 + jq2 +
q∗ = ( q0 − q ) = q0 − ( iq1 + jq2 + kq3 ). We can use Equation 15.7, to show that 2 qq∗ = q∗ q = q02 + q = q02 + q12 + q22 + q32 . This quantity is equal to the square of the norm or magnitude of the quaternion q , or
q=
qq∗ =
q02 + q12 + q22 + q32 = q∗ 15.8
q =Q0 + iQ1 + jQ2 + kQ3 is defined as a quaternion q with magnitude/norm equal to unity, with having QQ∗ = 1. Therefore, for a given unit quaternion its inverse and conjugate are equal, Q −1 = Q∗ . Also, 2 since q∗ q = q , see Equation 15.8, we can define the inverse of the quaterq∗ nion as q−1 = 2 . Therefore, the division operation of two quaternions can q be performed. This, however, has two forms since the quaternions are noncommutative. For example, the ratio q (assuming p ≠ 0) can be written as p −1 −1 qp or p q. Users should define which quotient is desired for their related calculations. A unit quaternion, Q =
15.5.2.1 Quaternions Algebraic Operations-Numerical Example Given q = 3 + i − 2 j + k and p = 2 − i + 2 j + 3 k , compute q + p, qp, their corresponding conjugate, norm, inverse, and unit quaternion. Solution: For the prodq + p =( 3 + i − 2 j + k ) + ( 2 − i + 2 j + 3 k ) =5 + 4 k . uct we calculate the dot and cross products of the correspond ing vectors, or and q ⋅ p =(1, −2,1 ) ⋅ ( −1,2,3 ) =−1 − 4 + 3 =−2 q × p =emnl qn pl =−8 i − 4 j. Therefore, using Equation 15.7, we have
qp =[ 6 + 2 ] + −3 i + 6 j + 9 k + 2 i − 4 j + 2 k − 8 i − 4 j = 8 − 9 i − 2 j + 11k . The conjugates are q∗ = 3 − ( i − 2 j + k ) and p∗ = 2 − ( − i + 2 j + 3 k ) . The p = 4 + 1 + 4 + 9= 3 2. norm and q = 9 + 1 + 4 + 1 = 15
Tensor-Analysis-for-Engineers_2E_CH15.indd 127
10/3/2020 10:29:22 AM
128 • Tensor Analysis for Engineers, 2E
The
inverse
q∗ 3 − i + 2 j − k 1 1 2 1 = − i+ q−1 = 2 = j− k 15 5 15 15 15 q
and
p∗ 2 + i − 2 j − 3 k 1 1 1 1 p−1 = 2 = = + i − j − k. The unit quaternions are, 18 9 18 9 6 p 3+ i−2 j+k 2 − i + 2 j + 3k 15 15 2 15 15 and 2 2 2 Q= i− j+ k = + = − P= i+ j+ 5 15 15 15 3 6 3 15 3 2 2 − i + 2 j + 3k 2 2 2 2 P= = − i+ j+ k. 3 6 3 2 3 2 Having the algebraic rules defined for quaternions, in the following sections we present their application for coordinates system and rigid body rotation. Since quaternions are basically an extension to the 2D complex numbers, first we discuss the relation between 2D complex numbers and rotation matrix followed by similar relations for 4D quaternions. 15.5.3 Complex Numbers and Rotation Matrix Complex numbers can be used to represent rotation of a rigid body when Euler formula is used. Euler formula relates an exponential complex number/variable to periodic sinusoidal functions [25], or = e iα cos α + i sin α 15.9
where i= −1 . For example, in a 2D space, we can define a complex variable z= z ( x, y )= x + iy representing a vector drawn from coordinates’ ori y gin to the point z on the x − y plane and making an angle = tan −1 , x phase/argument, with the positive x-axis and having the magnitude z = zz∗ = ( x + iy )( x − iy ) . Writing down the components, we get x = z cos ϕ and y = z sin ϕ . Or z = z ( cosϕ + i sinϕ ) = z e iϕ . Now, we operiθ = z e iϕ z e i(ϕ +θ ) . Note that the magniate the unit vector e iθ on z, to get z′ e= tude of the original vector is preserved, but its phase angle changes to (ϕ + θ ) . Expanding this expression, using Equation 15.9, we receive the components x cosθ − y sin θ + i ( x sin θ + y cosθ ) , of transformed vector z′ = ( x + iy) eiθ = x′ cosθ − sin θ x as (omitting i) = . This is identical to active rotation y′ sin θ cosθ y operation discussed in Section 15.1. Similarly, operating e− iθ on z gives the equivalent of passive rotation operation.
Tensor-Analysis-for-Engineers_2E_CH15.indd 128
10/3/2020 10:29:25 AM
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 129
15.5.4 Quaternions and Rotation To extend the application of complex numbers to 3D rotations, we make use of the process/method mentioned in the previous section and include quaternions in the discussion. First, we note that an arbitrary vector, V = iV1 + jV2 + kV3 =( V1 , V2 , V3 ) can be considered as a quaternion whose real part is zero, i.e., pure quaternions. Similarly, a real quaternion is a quaternion whose vector part is null with non-zero real part. Now, we consider a unit 2 quaternion Q = Q0 + Q. Therefore, Q = Q02 + Q = Q02 + Q12 + Q22 + Q32 = 1 which implies that we can define an angle β such that
Q0 = cos β 15.10 Q = sin β Q Q n = We also define a unit vector as = using the vector Q Q12 + Q22 + Q32 part (i.e., vector Q ) of the unit quaternion Q. Therefore, we can write the unit quaternion as = Q cos β + n sin β . Now, we can write as an extension to Euler formula, see Equation 15.9, the so-called De Moivre’s formula for quaternion Q, [26] nβ = Q e= cos β + n sin β 15.11
Note that n is a 3D imaginary number treated as a vector-like quantity calculated from the imaginary part of the unit quaternion. Using Equation 15.10, Q12 + Q22 + Q32 and since n = ( n1 , n2 , n3 ) is defined we can write β = tan −1 Q0 2 1. as a unit vector we have n1 + n22 + n32 =
Recall that the rotation matrix/operation preserves the magnitude of the vector which goes under rotation. In other words, magnitude of a vector under rotation is invariant (see Sections 12.2 and 15.5.3). Therefore, for using quaternions in rotation operation we look for an operator which is a combination of relevant quaternion and maintains the magnitude of the vector on which it operates. The quaternion operation on a vector is done in a two-step process, using a unit quaternion and its inverse/conjugate to make sure that the result is a vector with preserved magnitude. For this purpose, ∗ −1 we use the unit quaternion Q and form the combination QVQ = QVQ to transform vector V to V ′ by rotation, or
(
Tensor-Analysis-for-Engineers_2E_CH15.indd 129
)
10/3/2020 10:29:28 AM
130 • Tensor Analysis for Engineers, 2E ∗ V ′ QVQ enβ Ve− nβ 15.12 = =
As mentioned, the reason that we use unit quaternion and its inverse/conjugate is simply because the combination QV , as much as it may intuitively seem the correct combination, it doesn’t result in a new vector and nor maintaining the magnitude of V , under rotation operation. This can be seen using Equation 15.7 Q0 , Q1 , Q2 , Q3 ) ( 0, V1 , V2 , V3 ) and writing V as= a pure quaternion. Or QV (= −Q ⋅ V + Q0 V + Q × V , which is clearly a quaternion quantity and not a vector due to having a scalar part (i.e. Q ⋅ V ). Now, after expanding Equation 15.12, using Equation 15.11, we have QVQ∗ =( cos β + n sin β ) V ( co = V cos β + nV sin β ( cos β − n sin β ) = QVQ∗ =( cos β + n sin β ) V ( cos β − n sin β ) = V cos2 β − nVn sin 2 β + nV − Vn sin β cos β . But using vector identities, we can write nVn = V − 2 n ⋅ V n and nV − Vn =2 n ×V . After substituting we
( ) ( )
(
)
receive, in terms of double angle 2 β , ∗ QVQ = V cos2 β + (1 − cos2 β ) n ⋅ V n + n × V sin 2 β
(
) (
)
Therefore, we can write Equation 15.12 in index notation as (note that in Equation 15.13 subscripts i, j, k are simply indices, ranging from 1 to 3 and should not be confused with the imaginary numbers)
Vi′ Vi cos2 β + (1 − cos2 β ) ni n j V j + eijk n j Vk sin 2 β 15.13 = Writing Equation 15.13 in matrix form, let θ = 2 β , produces
V1′ cos θ + n1 n1 (1 − cos θ ) V ′ = n n (1 − cos θ ) + n3 sin θ 2 1 2 V3′ n1 n3 (1 − cos θ ) − n2 sin θ
n1 n2 (1 − cos θ ) − n3 sin θ cos θ + n2 n2 (1 − cos θ ) n2 n3 (1 − cos θ ) + n1 sin θ
n1 n3 (1 − cos θ ) + n2 sin θ V1 n2 n3 (1 − cos θ ) − n1 sin θ V2 cos θ + n3 n3 (1 − cos θ ) V3
15.14 This is identical to using Rodrigues equation when rotation angle is measured in the opposite direction, or the transpose of rotation matrix given in Equation 12.7. In other words, the operation performed according to Equations 15.12 or 15.14 gives active rotation of vector V through angle θ = 2 β about unit vector n. Therefore, the quaternions should be written as half-angle of the rotation angle (i.e., β = θ /2), Or
Tensor-Analysis-for-Engineers_2E_CH15.indd 130
10/3/2020 10:29:32 AM
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 131
Q θ rotation (2= = β )quaternion 2 tan −1 Q0 Q n = 2 Q1 + Q22 + Q32
15.15
Readers may recall that for active rotation we keep the coordinate system fixed and rotate the vector, see Section 15.1. From the above-mentioned derivation, we can conclude the following theorem (note that Q −1 = Q∗): = Q0 + Q, vector For any given vector V and unit quaternion Q QVQ −1 is equivalent to the active rotation of the vector V through Q about an axis parallel to Q or equivaan angle θ = 2 tan −1 Q0 Q lently the unit vector n = Q
Similarly, it can be shown that the combination Q∗ VQ = e− nβ Venβ 15.16 is equivalent to a passive rotation of vector V . Two points worth mentioning, about quaternions and rotations. 1) It takes two quaternions, the unit quaternion itself and its inverse/conjugate, to perform a rotation for a rigid body. Each quaternion carries half of the total angle of rotation, i.e., β =
θ
. 2) Further transformation follows the same 2 rule. For example, the vector Q∗ VQ can be transformed again using unit quaternion P representing a second rotation, or
∗ P ∗ Q∗ VQ P = ( QP ) V ( QP ) = W ∗ VW 15.17 nd
(
)
st
1 rotation
Tensor-Analysis-for-Engineers_2E_CH15.indd 131
2
rotation
10/3/2020 10:29:34 AM
132 • Tensor Analysis for Engineers, 2E As seen from Equation 15.17, we can combine the related two quaternions as their product W = QP and perform the operation on the desired vector as a single rotation. The first rotation corresponds to Q and the second one to P. Similarly, for an active rotation we can write the following relation for two successive rotations, ∗ P QVQ∗ P ∗ = ( PQ ) V ( PQ ) = V ∗ 15.18 nd
(
)
2
st
1 rotation
rotation
Example: Show that following theorem is correct. = Q0 + Q , comFor any given vector V and unit quaternion Q ponents of vector Q −1 VQ are equivalent to those of vector V in the rotated coordinates (i.e., passive rotation) by an angle Q θ = 2 tan−1 about an axis parallel to the vector Q or equiv Q0 Q alently unit vector n = Q Solution: Note that quaternions should be written as half of the total rotation angle,
θ
, Q−1 VQ =( cos β − n sin β ) V ( cos β + n sin β ) == ( V cos β − nV sin β ) ( cos β + n sin β ) = 2 = V cos2 β − nVn sin 2 β − nV − Vn sin β cos β . However, sin β ) ( cos β + n sin β ) = ( V cos β − nV using vector identities, we can write nVn = V − 2 n ⋅ V n and nV − Vn = 2 n × V ∗ nV − Vn = 2 n × V . Substituting, we obtain, in terms of double angle 2 β , Q= VQ V cos2 β + (1 − cos2 β ∗ Q= VQ V cos2 β + (1 − cos2 β ) n ⋅ V n − n × V sin 2 β . Or in index notation form or β =
(
(
) (
(
)
)
)
we have (note that in Equations 15.13 and 15.19 subscripts i, j, k are simply indices range from 1 to 3 and should not be confused with the imaginary numbers)
= Vx′i Vi cos2 β + (1 − cos2 β ) ni n j V j − eijk n j Vk sin 2 β 15.19 Writing Equation 15.19 in matrix form, let θ = 2 β , produces
Tensor-Analysis-for-Engineers_2E_CH15.indd 132
10/3/2020 10:29:39 AM
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 133
Vx′ cos θ + n1 n1 (1 − cos θ ) Vy′ = n1 n2 (1 − cos θ ) − n3 sin θ Vz′ n1 n3 (1 − cos θ ) + n2 sin θ
n1 n2 (1 − cos θ ) + n3 sin θ cos θ + n2 n2 (1 − cos θ ) n2 n3 (1 − cos θ ) − n1 sin θ
n1 n3 (1 − cos θ ) − n2 sin θ V1 n2 n3 (1 − cos θ ) + n1 sin θ V2 cos θ + n3 n3 (1 − cos θ ) V3
15.20 This is identical to the rotation matrix given in Equation 12.7. In other words, the operation given by Q −1 VQ gives the components of vector V when coordinates are rotated through angle θ = 2 β about unit vector n , see Equation 15.15. Readers may recall that for passive rotation we keep the vector fixed and rotate the coordinate system, see Section 15.1. Example: Rotation using quaternions-numerical example 2π Consider the rotation of vector V = i= (1,0,0 ) by an angle θ = about an 3 axis in the direction r = (1,1,1 ). Find the transformed vector assuming active and passive rotations using following methods: a. Rodrigues’ formula b. quaternions Solution:
r 1 1 1 , , = . r 3 3 3 The transformation matrix is, using Equations 15.13 or 12.7, for active
a. The unit vector related to the rotation axis is = n
transformation (vector rotates and coordinates are fixed) gives (note that 2π
when 12.7) θ2π= − 1 for active 2π rotation, 1 2π plugged 1 2in π Equation 1 2π 1 2π sin sin 1 − cos − 1 − cos + cos 3 + 33 1 − cos 3 3 3 3 3 3 3 3 3 1 V1′ 2π 1 2π 2π 1 2π 1 2π 1 2π 1 sin cos sin + 1 − cos 0 V2′ = 1 − cos + 1 − cos − 3 3 3 3 3 3 3 3 3 3 V ′ 3 0 3 1 2π 1 2π 1 2π 1 2π 2π 1 2π + 1 − cos cos 1 − cos +2π sin 3 11 − cos 3 2π− sin 2π 3 2π − 13 sin 3 1 13 − cos 2π 3+ 13 sin 2π 3 + 1 − cos 3 1 13 − cos cos 3 3 3 3 3 3 3 3 3 3 3 1 V1′ 2π 1 2π 2π 1 2π 1 2π 1 2π 1 sin cos + 1 − cos sin V2′ = 1 − cos 0 + 1 − cos − 3 3 3 3 3 3 3 3 3 3 V ′ 3 0 3 1 2π 1 2π 1 2π 1 2π 2π 1 2π 1 cos sin 1 cos sin cos 1 cos − − − + + − 3 3 3 3 3 3 3 3 3 3 3
Tensor-Analysis-for-Engineers_2E_CH15.indd 133
10/3/2020 10:29:41 AM
134 • Tensor Analysis for Engineers, 2E Or V1′ 0 0 1 1 0 1 0 0 0 1 . Hence active rotation of vector i gives vector = V2′ = V ′ 0 1 0 0 0 3 j. Similarly, the passive rotation (coordinates rotate and vector is fixed) gives 2π (note 2that passive rotation, when plugged in Equation 12.7) 2π π 1θ= 3 for 2π 1 2π 1 2π 1 2π 1 cos 1 cos sin 1 cos sin + 1 − cos − + − −
3 3 3 3 3 3 3 3 3 3 3 Vx′ 2π 1 2π 2π 1 2π 1 2π 1 2π 1 sin cos sin = + 1 − cos Vy′ = 1 − cos − 1 − cos + 3 3 3 3 3 3 3 3 3 3 3 Vz′ 1 2π 1 2π 1 2π 1 2π 2π 1 2π 1 − cos sin 1 − cos sin cos − + 1 − cos + 2π 3 1 2π 3 1 3 3 2 π 1 3 2π 3 1 3 2π 3 1 3 2π 3 3 sin sin 1 − cos + 1 − cos − cos 3 + 3 1 − cos 3 3 3 3 3 3 3 3 3 Vx′ 2π 1 2π 2π 1 2π 1 2π 1 2π 1 sin cos sin = + 1 − cos Vy′ = 1 − cos − 1 − cos + 3 3 3 3 3 3 3 3 3 3 3 Vz′ 1 2π 1 2π 1 2π 1 2π 2π 1 2π sin sin cos + 1 − cos 1 − cos − 3 1 − cos 3 + 3 3 3 3 3 3 3 3 3
0 1 0 1 0 . Hence passive rotation of vector i gives vector k . 0 0 1 0 = 0 1 0 0 0 1
b. When quaternions are used for rotation note that half-angle π should be used = /2 β θ= . The unit quaternion reads, 3 π π 1 i j k 3 1 i j k using Equation 15.11, Q = + cos + n sin = + + = + + + 3 3 2 3 2 2 2 2 3 3 2 i j k j k 3 1 i j k . Hence, Q = 1 and Q i = + + . For active 0 + + = + + + 2 2 2 2 2 2 2 2 3 3 2 3 ∗ 1 i j k 1 i j k QVQ = + + + ( i ) − − − = rotation, we have 2 2 2 2 2 2 2 2 1 i k j 1 i j k . Similarly, for passive rotation j − 2 + 2 − 2 + 2 2 − 2 − 2 − 2 = 1 i j k 1 i j k we have Q∗ VQ = − − − ( i ) + + + = k. Readers can 2 2 2 2 2 2 2 2 instead use Equations 15.14 and 15.20, respectively.
Tensor-Analysis-for-Engineers_2E_CH15.indd 134
10/3/2020 10:29:44 AM
2
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 135
15.6 FROM A GIVEN QUATERNION TO ROTATION MATRIX In this section, we would like to derive the rotation matrix from a given quaternion related to a rigid body rotation. For a given unit quaternion, we can use Equations 15.14 in conjunction with Equation 15.15 for active rotation and Equation 15.20 for passive rotation for calculating the equivalent rotation matrices. However, equivalently we can directly calculate the rotation matrix by implementing quaternions algebraic product rules on QVQ∗, for example, for a given vector V = iV1 + jV2 + kV3. For a given unit quaternion, we have Q = Q0 + Q = Q0 + iQ1 + jQ2 + kQ3 along with Q = Q02 + Q12 + Q22 + Q32 = 1. Therefore, the combination QVQ∗ can be written as follow. QVQ∗ = ( Q0 + iQ1 + jQ2 + kQ3 ) ( iV1 + jV2 + kV3 )( Q0 − iQ1 − jQ2 − kQ3 ) . Performing the multiplication on the first two brackets, produces QVQ∗ = − ( Q1 V1 + Q2 V2 + Q3 V3 ) + i ( Q0 V1 + Q2 V3 − Q3 V2 ) + j ( Q0 V2 − Q1 V3 + Q3 V1 ) + k ( Q0 V3 + Q1 V2 −
) + j ( Q0 V2 − Q1 V3 + Q3 V1 ) + k ( Q0 V3 + Q1 V2 − Q2 V1 ) ( Q0 − iQ1 −
jQ2 − kQ3 ) . Performing the multiplica-
tion on the last bracket produces ∗ QVQ = i V1 ( Q02 + Q12 − Q22 − Q32 ) + 2 V2 ( Q1Q2 − Q0 Q3 ) + 2 V3 ( Q0 Q2 + Q1Q3 ) + j 2 V1 ( Q0 Q3 + Q1Q2 ) + V2 ( Q02 − Q12 + Q22 − Q32 ) + 2 V3 ( Q2 Q3 − Q0 Q1 ) +
k 2 V1 ( Q1Q3 − Q0 Q2 ) + 2 V2 ( Q0 Q1 + Q2 Q3 ) + V3 ( Q02 − Q12 − Q22 + Q32 ) . Now
1 we can substitute for −Q22 − Q32 = Q02 + Q12 − 1, using Q02 + Q12 + Q22 + Q32 =
−Q12 − Q32 = Q02 + Q22 − 1, and −Q12 − Q22 = Q02 + Q32 − 1, to obtain the rotation matrix as: (1st row is the coefficients of i, 2nd row those of j, and 3rd row those of k) Q02 + Q12 − 0.5 Q1Q2 − Q0 Q3 Q0 Q2 + Q1Q3 = [ Ractive ] 2 Q0Q3 + Q1Q2 Q02 + Q22 − 0.5 Q2Q3 − Q0Q1 15.21 Q1Q3 − Q0 Q2 Q0 Q1 + Q2 Q3 Q02 + Q32 − 0.5
Or V1′ V1 V2′ = [ Ractive ] V2 V ′ V 3 3
Tensor-Analysis-for-Engineers_2E_CH15.indd 135
10/3/2020 10:29:47 AM
136 • Tensor Analysis for Engineers, 2E Similarly, we can calculate the passive rotation matrix using Q∗ VQ, to receive
Q02 + Q12 − 0.5 Q0 Q3 + Q1Q2 Q1Q3 − Q0 Q2 Rpassive 2 Q1Q2 − Q0 Q3 Q02 + Q22 − 0.5 Q0 Q1 + Q2 Q3 15.22 = Q0 Q2 + Q1Q3 Q2 Q3 − Q0 Q1 Q02 + Q32 − 0.5
Or Vx′ V1 Vy′ = Rpassive V2 V V z′ 3 Reader may note that Rpassive = [ Ractive ] , as expected. Readers should also note that elements of the above-mentioned matrices are calculated using related unit quaternions. T
Example: Calculating rotation matrix from a given quaternion-numerical example The quaternion q = 3 + i − 2 j + k is given. Find the equivalent rotation matrix for both active and passive rotation scenarios. For each scenario calculate the equivalent single axis and angle of rotation and compare the results. Solution: We first check if the given quaternion is a unit quaternion by calculating its norm/magnitude. Hence for q = 3 + i − 2 j + k we have q = 9 + 1 + 4 + 1 = 15 ≠ 1. Therefore, we need to calculate the corre1 1 −2 1 3 sponding unit quaternion = Q + k) , , , ( 3 + i − 2 j= . 15 15 15 15 15 3 Now using Equation 15.21 for active rotation and plugging in for Q0 = , 15 −2 1 1 , Q2 = , and Q3 = we receive Q1 = 15 15 15
Tensor-Analysis-for-Engineers_2E_CH15.indd 136
10/3/2020 10:29:49 AM
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 137
2 2 1 1 2 3 6 1 1 9 − − − + 15 + 15 − 2 − 15 − 15 3 3 3 15 15 2 11 2 3 2 9 4 1 2 3 − . = − + − − −= [ Ractive ] 2 15 15 3 15 15 15 15 2 15 15 1 14 2 1 6 3 2 9 1 1 + − + − 3 15 15 15 15 2 15 15 15 15 The single axis of rotation is the vector corresponding to the vector part of the Q 1 −2 1 −1 unit quaternion, or Q = , , and the angle is 2 tan Q = 15 15 15 0 6 ° 2 tan −1 = 78.46 , see Equation 15.15. Note that the corresponding 3 unit vector is 1 , −2 , 1 Q 15 15 15 1 −2 1 . , , = = = n active 1 4 1 Q 6 6 6 + + 15 15 15 Similarly, for passive rotation, using Equation 15.22, we find 2 14 1 3 15 15 2 11 2 Rpassive = − 3 15 15 . We use the resulted rotation matrix to calcu 2 2 1 − − 3 3 3 21 − 1 ° late the single angle of rotation, or cos 15 = 78.46 , see Equation 2 12.8. The axis of rotation is the vector part of the conjugate unit −1
quaternion
Tensor-Analysis-for-Engineers_2E_CH15.indd 137
= Q∗
1 −1 2 −1 3 − k) , , , ( 3 − i + 2 j= , 15 15 15 15 15
or
10/3/2020 10:29:51 AM
e
138 • Tensor Analysis for Engineers, 2E
2 −1 −1 , , ∗ −1 2 −1 15 15 −1 2 −1 Q = , , 15= with the unit = vector n , , = n act passive 15 15 15 1 4 1 6 6 6 + + 2 −1 −1 15 15 15 , , 15 15 −1 2 −1 15= , , = n active . 1 4 1 6 6 6 + + 15 15 15
15.7 FROM A GIVEN ROTATION MATRIX TO QUATERNION In this section, we would like to derive the equivalent quaternion from a given rotation matrix related to a rigid body rotation. Having the rotation matrix, we can use the relations obtained in the previous section to obtain the corresponding quaternion. Let us consider a passive rotation case, given by Equation 15.22. Note that the trace for both active and passive rotation matrices are equal in their values. The trace of the rotation matrix is Rtrace = 2 ( 3Q02 + Q12 + Q22 + Q32 − 1.5= ) 2 2Q02 + Q02 + Q12 + Q22 + Q32 − 1.5 = 4Q02 − 1. Therefore, we have the real 1 part of the quaternion as Q0 = ±
1 + Rtrace 2
15.23
Knowing the real part, we can calculate the elements of complex part Q using the diagonal elements of rotation matrix. Therefore, 2 R (1,1 ) − Rtrace + 1 R (1,1 ) − 2Q02 + 1 R (1,1 ) = 2Q02 + 2Q12 − 1 or Q1 = ± = . 2 2 Similarly, q2 = ±
2 R ( 2,2 ) − Rtrace + 1
, and q3 = ±
2 R ( 3,3 ) − Rtrace + 1
2 2 Therefore, component m of the complex part of the quaternion reads
.
2 R ( m, m ) − Rtrace + 1 , m= 1,2,3 15.24 Qm = ± 2
Tensor-Analysis-for-Engineers_2E_CH15.indd 138
10/3/2020 10:29:53 AM
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 139
From Equations 15.23 and 15.24, it is shown that for a given rotation matrix we get two unit quaternions that describes the rotation, or Q = Q0 + Q1 + Q2 + Q3 −Q = −Q0 − Q1 − Q2 − Q3 . Here, Q represents the unit quaternion and P = for a positive (R.H.R.) rotation about the axis of rotation and P represents the rotation about the same axis but in opposite direction, [27]. Example: Calculating quaternion from a given rotation matrix-numerical example 0 0 1 The rotation matrix R = 1 0 0 is given. Find the corresponding quaternions. 0 1 0 Solution: Using Equation 15.23, we receive the real part of the corresponding unit 1+0 1 ± = ± . The vector part can be calculated using quaternion as Q0 = 2 2 0 −0 +1 1 Q2 = Q3 = ± = ± . Therefore the quaterEquation 15.24, or Q1 = 2 2 1 1 1 1 1 1 1 1 − qˆ = nions are Q = , , , and Q = − ,− ,− ,− . 2 2 2 2 2 2 2 2
15.8 FROM EULER ANGLES TO A QUATERNION In this section, we would like to derive the quaternion from given Euler angles related to a rigid body rotation. Using the relations derived in the previous section, we can find corresponding quaternions for Euler angles. For example, for 0 0 1 a roll rotation by angle θ roll about x-axis, we have 0 cos ( θ roll ) sin (θ roll ) , 0 − sin (θ roll ) cos ( θ roll ) see Table 15.2, as the rotation matrix. Therefore, using Equation 15.23, 1 + 1 + 2 cos (θ roll ) 2 − 2 cos (θ roll ) − 1 + 1 θ θ = = cos roll , Q1 Q0 = = sin roll , 2 2 2 2 2 cosθ roll − 2 cos (θ roll ) − 1 + 1 = = = 0. This offers the correQ and Q 2 3 2 sponding unit quaternion equivalent to the roll rotation as
Tensor-Analysis-for-Engineers_2E_CH15.indd 139
10/3/2020 10:29:55 AM
140 • Tensor Analysis for Engineers, 2E
= Qθ roll cos
θ roll 2
+ i sin
θ roll 15.25 2
Similarly, the unit quaternions for pitch and yaw rotations are = Qθ pitch cos
θ pitch 2
+ j sin
θ pitch 2
15.26
and = Qθ yaw cos
θ yaw 2
+ k sin
θ yaw 2
15.27
Similarly, we can obtain 12 quaternions related to 12 Euler angles combinations. This can be achieved by equating the elements of each rotation matrix given in Table 15.2 and Table 15.3 with Equation 15.22. The result would be nine equations, which can be solved for corresponding quaternion real and imaginary parts [27]. However, in practice it is more efficient to calculate the final rotation matrix and extract the corresponding quaternions, as described in Section 15.7. Another method of extracting the quaternion related to a given set of Euler angles (without calculating the single-axis rotation matrix) is to use Equation 15.17 in conjunction with Equations 15.25 through 15.27.
15.9 PUTTING IT ALL TOGETHER It seems useful to collect all related methods and their interrelations, see Table 15.5. In this table, we list the related equations for calculating parameters of a desired method from those of a given method, directly. For example, S→E indicates the direct relations for obtaining Euler angles from a given rotation matrix. Readers should note that it is always possible to calculate the rotation matrix using one of the three methods (i.e., Euler’s angle, quaternions, single axis-angle) and then use it for calculating the parameters related to another method.
Tensor-Analysis-for-Engineers_2E_CH15.indd 140
10/3/2020 10:29:59 AM
Tensor-Analysis-for-Engineers_2E_CH15.indd 141
procedure S→R→E Eq. 12.7 S→R→Q
S-single axis-angle
S→E
S→R
S→Q
E→Q
E→R
E→S
E-Euler angles
E→R→Q
Table 15- & Table 15-3.
E→R→S
procedure
TABLE 15.5 Rigid body rotation direct relations and methods.
Q→E
Q→R
Q→S
Q→R→E
Eq. 15.21 &15.22
Eq. 15.15
Q-quaternions procedure
R→Q
R→E
R→S
R-rotation matrix
Eq. 15.23, 15.24 & 15.24
Table 15- & Table 15-3.
Eq. 12.8, 15.3
procedure
Rigid Body Rotation: Euler Angles, Quaternions, and Rotation Matrix • 141
10/3/2020 10:29:59 AM
Tensor-Analysis-for-Engineers_2E_CH15.indd 142
10/3/2020 10:29:59 AM
CHAPTER
16
WORKED-OUT EXAMPLES In this section, we present several worked-out examples related to the topics covered in the previous sections.
16.1 EXAMPLE: EINSTEIN SUMMATION CONVENTIONS Write out the expanded expression in full detail for N = 3 dimensions for, ∂x′ j = A dx i ei − i e′ j . ∂x Solution:
∂x′ j Summation applies on both indices i and j , hence A = dx i ei − i e′ j = ∂x j 1 2 3 ∂x′ ∂x′ ∂x′ ∂x′ dx i ei − dx i i e′ j = dx1 e1 + dx2 e2 + dx3 e3 − dx1 1 e1′ + 1 e2′ + 1 e3′ − ∂x ∂x ∂x ∂x ∂x′1 ∂x′2 ∂x′3 ∂x′1 ∂x′2 ∂x′3 dx2 2 e1′ + 2 e2′ + 2 e3′ − dx3 3 e1′ + 3 e2′ + 3 e3′ . ∂x ∂x ∂x ∂x ∂x ∂x 1 2 ∂x′ ∂x′ ∂x′3 Re-arranging the terms, we receive A dx1 e1 − 1 e1′ − 1 e2′ − 1 e3′ + = ∂x ∂x ∂x ∂x′1 ∂x′2 ∂x′3 ∂x′1 ∂x′2 ∂x′3 dx2 e2 − 2 e1′ − 2 e2′ − 2 e3′ + dx3 e3 − 3 e1′ − 3 e2′ − 3 e3′ . ∂x ∂x ∂x ∂x ∂x ∂x
Tensor-Analysis-for-Engineers_2E_CH16.indd 143
10/3/2020 10:31:57 AM
144 • Tensor Analysis for Engineers, 2E
16.2 EXAMPLE: CONVERSION FROM VECTOR TO INDEX NOTATIONS Write out the following expressions written in vector form in index notation and expand for 3D coordinates. 2 a. ∇ ⋅ A b. A ⋅ ∇ 2 A c. A ⋅ ∇ ∇ ⋅ A
(
)
(
)
( (
))
d. ∇ ⋅ A ( ∇ 2 Φ ) e. ∇ 2 A ⋅ ( ∇Φ ) f. A ⋅ ∇ ( ∇ 2 Φ )
(
)
( (
) )
Solution: 2 2 k a. ∇ ⋅ A= A j j A = A 1 + A22 + A 33 ) ′ ′ k ( ′1 ′ ′ 2 = A A k= Ak ′⋅ jj A1 A1′ ⋅ jj + A2 A2 ′ ⋅ jj + A3= A3 ′⋅ jj A1 A1 ′⋅ 11 + A1 ′⋅ 22 + A1′ ⋅33 b. A ⋅ ∇
(
)
(
)
(
)
(
(
+ A2 A1′ ⋅ 11 + A1′ ⋅22 + A1′ ⋅33 + A3 A1′ ⋅11 + A1′ 2⋅2 + A1′ ⋅33
(
)
j k 3 k ⋅A A= A′kj A1 A′kk1 + A2 A′kk 2 + A= A′k 3 A1 A1′11 + A2′21 + A3′31 c. A ⋅ ∇ ∇=
( (
))
(
) (
+ A2 A1′11 + A2′21 + A3′31 + A3 A1′11 + A2′21 + A3′31
)
)
)
k k k k 1 2 3 d. ∇ ⋅ A ( ∇ 2 Φ= ) A′j j ( ∇ kΦ )= ′ ⋅ A′1 ( ∇ k Φ ) ′ ⋅ + A′2 ( ∇ k Φ )′ ⋅ + A′3 ( ∇ k Φ )= ′⋅ 1 2 3 1 2 3 A ′11 ( ∇1 Φ )′ ⋅ + ( ∇ 2 Φ ) ′⋅ + ( ∇ 3 Φ )′ ⋅ + A ′22 ( ∇1 Φ ) ′⋅ + ( ∇ 2 Φ ) ′⋅ + ( ∇ 3 Φ ) ′⋅ +
(
)
1 2 3 A ′33 ( ∇1 Φ ) ′⋅ + ( ∇ 2 Φ ) ′⋅ + ( ∇ 3 Φ ) ′⋅
Tensor-Analysis-for-Engineers_2E_CH16.indd 144
10/3/2020 10:31:58 AM
Worked-out Examples • 145
e. ∇ 2 A ⋅ ( ∇Φ = )
(
) (A
= =
k
( ∇ k Φ ) )′ j ⋅ j
( A (∇ Φ ) + A (∇ Φ ) + A (∇ Φ ))′ ( A (∇ Φ ) + A (∇ Φ ) + A (∇ Φ ))′ + ( A (∇ Φ ) + A (∇ Φ ) + A (∇ Φ )) ′ + ( A (∇ Φ ) + A (∇ Φ ) + A (∇ Φ )) ′ 1
2
1
1
2
1
1 1
2
1
3
2
2
1
3
2
2 2
3 3
j
3
j⋅
3
1 1⋅
3
2 2⋅
3
3 3⋅
k f. A ⋅ ∇ ( ∇ 2 Φ )= A j ( ∇ k Φ ) ⋅ j
(
)
′
= A1 ( ∇ k Φ )
k + A 3 ( ∇ k Φ ) ⋅3 ′⋅ 2 ′ 3 1 1 2 = A ( ∇1 Φ ) ⋅ 1 + ( ∇ 2 Φ ) ⋅ 1 + ( ∇ 3 Φ ) ⋅1 ′ ′ ′ 1 2 3 2 + A ( ∇1 Φ ) ⋅2 + ( ∇ 2 Φ ) ⋅2 + ( ∇ 3 Φ ) ⋅2 ′ ′ ′ 1 2 3 + A3 ( ∇1 Φ ) ⋅ 3 + ( ∇ 2 Φ ) ⋅ 3 + ( ∇ 3 Φ ) ⋅3 ′ ′ ′ k
′⋅ 1
+ A2 ( ∇ k Φ )
k
16.3 EXAMPLE: OBLIQUE RECTILINEAR COORDINATE SYSTEMS Consider a 2D oblique/slanted coordinate system, x i in which the coordinate axes are not at right angles. With reference to the Cartesian coordinate sys tem yi, as shown in Figure 16.1, for a given vector A find: a. Covariant and contravariant components; Ai and A i b. Covariant and contravariant basis vectors; ei and e i with sketch on the coordinate system i i c. Show that = A A= ei A= Ac1 E1 + Ac 2 E2 ; where E1 and E2 are ie Cartesian unit vectors d. Verify that e i ⋅ e j = δ ji and sketch all basis vectors e. Metric tensors f. Unit vectors ei / ei , or scale factor, and e i / e i
Tensor-Analysis-for-Engineers_2E_CH16.indd 145
10/5/2020 11:27:26 AM
146 • Tensor Analysis for Engineers, 2E
FIGURE 16.1 Oblique coordinate system x i and Cartesian coordinate system y i.
Solution: Assume axes x1 and x2 make angles α and β with Cartesian axis y1 , respectively. Therefore, we can write the following functional relations between the oblique and Cartesian systems: y1 x1 cos α + x2 cos β = , or inversely 2 1 2 = y x sin α + x sin β
1 − csc (α − β ) y1 sin β − y2 cos β x = 2 csc (α − β ) y1 sin α − y2 cos α x =
a. Covariant components of A are obtained by drawing perpendicular
lines to the coordinate axes. We have Ai =
∂y j ∂y1 A or A1 Ac1 + cj, = ∂x i ∂x1
∂y2 ∂y1 ∂y2 = A cos α A + sin α A . Similarly, A A Ac 2 = cos β Ac1 + = + 2 1 2 c c c 2 c1 ∂x1 ∂x2 ∂x2 sin β Ac 2 . The contravariant components of A are obtained by d rawing parallel lines to the coordinate axes. We have A i =
∂x i Acj , or A1 = j ∂y
∂x1 ∂x1 A Ac 2 = + − sin β csc (α − β ) Ac1 + cos β csc (α − β ) Ac 2. Similarly, c1 ∂y1 ∂y2
∂x2 ∂x2 A Ac 2 = sin α csc (α − β ) Ac1 − cos α csc (α − β ) Ac 2 . Note + c 1 ∂y1 ∂y2 that for the Cartesian system both covariant and contravariant components are identical. A2 =
Tensor-Analysis-for-Engineers_2E_CH16.indd 146
10/3/2020 10:32:03 AM
Worked-out Examples • 147
∂y b. The covariant basis vectors for oblique system are ei = i E j , ∂x 1 2 1 ∂ y ∂ y ∂y or e1 = 1 E1 + 1 E2 = cos α E1 + sin α E2. Similarly,= e2 E1 + ∂x ∂x ∂x2 j
∂y2 ∂x i = E2 cos β E1 + sin β E2. The contravariant basis vectors are e i = j E j, 2 ∂y ∂x
∂x1 ∂x1 e1 = E E sin β csc α β E cos β csc α β E + = − − + − ( ) ( ) 1 2 1 2. ∂y1 ∂y2 ∂x2 ∂x2 Similarly, e 2 = E1 + 2 E2 = sin α csc (α − β ) E1 − cos α csc (α − β ) E2 . 1 ∂y ∂y Note that for the Cartesian system both covariant and contravariant basis vectors are identical and are unit vectors, i.e. Ei = 1 . c. A i ei = A1 e1 + A2 e2 = [− sin β csc (α − β ) Ac1 + cos β csc (α − β ) Ac 2 ][cos α E1 + sin α E2 ] + [sin α csc (α − β ) Ac1 − cos α csc (α − β ) Ac 2 ][cos β E1 + sin β E2 ]. After some manipulations and rearranging the terms, we find A i ei = i 1 2 Ac1 E1 + Ac 2 E2 . Similarly, Ai e = A1 e + A2 e = [cos α Ac1 + sin α Ac 2 ] [− sin β csc (α − β ) E1 + cos β csc (α − β ) E2 ] + [cos β Ac1 + sin β Ac 2 ][sin α csc (α − β ) E1 − cos α csc (α − β ) E2 ]. After some manipulations and rearranging the terms, we find= Ai e i Ac1 E1 + Ac 2 E2. or
d. Dot-product of covariant and contravariant basis vectors can cos α sin α sin α cos α now be written as e1 ⋅ e 2 = e 2 ⋅ e1 = − = 0 and sin (α − β ) sin (α − β ) cos β sin β sin β cos β 1 e2 ⋅ e 1 = e ⋅ e2 = − + = 0 . These results consin (α − β ) sin (α − β ) firm that e1 ⊥ e 2 and e2 ⊥ e 1 , as shown in Figure 16.2. Also, we can cos β sin α cos α sin β sin α cos β write e1 ⋅ e 1 = e2 ⋅ e 2 − − + = 1 and = sin (α − β ) sin (α − β ) sin (α − β ) sin β cos α δ ji . = 1. These results verify that e i ⋅ e j = sin (α − β )
Tensor-Analysis-for-Engineers_2E_CH16.indd 147
10/3/2020 10:32:06 AM
148 • Tensor Analysis for Engineers, 2E
FIGURE 16.2 Covariant and contravariant basis vectors in oblique coordinate systems.
e. The
doubly
covariant
e1 ⋅ e1 e1 ⋅ e2 metric = tensor gij = e2 ⋅ e1 e2 ⋅ e2
1 cos (α − β ) . We can calculate doubly contravariant metric 1 cos (α − β ) e1 ⋅ e1 e1 ⋅ e 2 ij = g 2 1 = tensor by inverting the gij , or using csc 2 (α − β ) 2 2 ⋅ ⋅ e e e e 1 − cos (α − β ) i . These results confirm that x is a non− cos α − β 1 ( ) orthogonal system, since off-diagonal elements of metric tensors are not equal to zero for all values of angles α and β with the condition that
π
(α − β ) ≠ ( 2 k + 1) . where, k is an integer.
2 e1 f. Unit vectors along the covariant basis vectors are; = cos α E1 + sin α E2 e1 e2 and = cos β E1 + sin β E2. The unit vectors along the contravariant basis e2 e1 e1 e2 e2 = − sin β E + cos β E vectors are; 1 = and = = 1 2 e e2 e1 ⋅ e1 e2 ⋅ e2 sin α E1 − cos α E2 .
Tensor-Analysis-for-Engineers_2E_CH16.indd 148
10/3/2020 10:32:08 AM
Worked-out Examples • 149
16.4 EXAMPLE: QUANTITIES RELATED TO PARABOLIC COORDINATE SYSTEM Consider a parabolic coordinate system ( x1 , x2 , x3 ) ≡ (ξ ,η ,θ ) given in terms of Cartesian ( y1 , y2 , y3 ) ≡ ( X , Y, Z ), as shown in Figure 16.3. The functional relations are
FIGURE 16.3 Parabolic coordinate system.
X = ξη cosθ . Find the basis vectors e1, e2, e3 for the parabolic coordi Y = ξη sin θ Z (ξ 2 − η 2 ) / 2 = nate system in terms of the Cartesian unit vectors. Also find the scale factors, unit vectors, metric tensors (covariant and contravariant), Jacobian, volume element, and Christoffel symbols of the 2nd kind. Solution:
j ∂y The covariant basis vector ( eξ , eη , eθ ) reads, using ek = k E j ∂x eξ = η cosθ E1 + η sin θ E2 + ξ E3 eη = ξ cosθ E1 + ξ sin θ E2 − η E3 −ξη sin θ E1 + ξη cosθ E2 eθ =
Tensor-Analysis-for-Engineers_2E_CH16.indd 149
10/3/2020 10:32:09 AM
150 • Tensor Analysis for Engineers, 2E e= ξ 2 + η 2 , h= e= The scale factors are h= ξ ξ η η The unit vectors are, using e ( k ) = ek / hk ,
ξ 2 + η 2 , and hθ = ξη.
e (ξ ) = η cosθ E1 + η sin θ E2 + ξ E3 / ξ 2 + η 2 2 2 e (η ) = ξ cosθ E1 + ξ sin θ E2 − η E3 / ξ + η − sin θ E1 + cosθ E2 e (θ ) =
( (
Covariant
metric
) )
tensor
reads,
using
gij= ei ⋅ e j ,
ξ 2 + η 2 0 0 2 2 = gij 0 ξ +η 0 . Therefore, the parabolic coordinate sys 0 (ξη )2 0 tem is an orthogonal one, since metric tensor is a diagonal matrix with null off-diagonal elements. The contravariant metric tensor is the inverse (ξ 2 + η 2 )−1 0 0 −1 ij 2 2 −1 = g = gij 0 0 . of the covariant one, or (ξ + η ) −2 0 0 (ξη ) g Jacobian is the square root of the determinant of ij , or the product of all scale
( )
factors,=
= gij ξη (ξ 2 + η 2 ) . The volume element is d d d d
g mk ∂g jk ∂gik ∂gij + j − k ∂x ∂x 2 ∂x i g11 ∂g11 ξ = Γ111 = to calculate the Christoffel symbols. Therefore, , 2 2 ∂ξ ξ + η 2
2 2 d d d . We use Equation 10.20,= or Γ mij
= Γ112
g11 ∂g11 η = , 2 2 ∂η ξ + η 2
2 = Γ12
g 22 ∂g22 ξ = , 2 2 ∂ξ ξ + η 2
Tensor-Analysis-for-Engineers_2E_CH16.indd 150
∂g11 η − ∂η =− ξ 2 + η 2 , 33 g ∂g33 1 3 2 3 3 Γ13 = , Γ13 = Γ11 = Γ12 =, 0 = 2 ∂ξ ξ Γ113 = 0,
g 22 2 Γ11 = 2
10/3/2020 10:32:13 AM
Worked-out Examples • 151
∂g22 ξ − 2 , − ∂ξ = ξ +η2 g 22 ∂g22 η = Γ 222 = , 2 2 ∂η ξ + η 2 g11 Γ122 = 2
= Γ 323
∂g33 ξη 2 − = − , ∂ξ ξ 2 +η2 g 22 ∂g ηξ 2 Γ 233 = − 33 = − 2 , ξ +η2 2 ∂η g11 Γ133 = 2
Γ123 = 0, Γ 223 = 0,
g 33 ∂g33 1 3 = , Γ 33 = 0. In matrix form, we have 2 ∂η η
Γξξξ ξ Γξ = Γηξ ξ Γθξ
Γηξξ η Γη = Γηξ Γηθξ
Γξξη ξ Γηη ξ Γθη
Γηξη η Γηη Γηθη
ξ 2 ξ +η2 ξ Γξθ η ξ Γηθ = ξ 2 + η 2 ξ Γθθ 0 η − 2 ξ +η2 Γηξθ ξ η Γηθ = ξ 2 +η2 Γηθθ 0
Γθξξ θ Γθ = Γηξ Γθθξ
Tensor-Analysis-for-Engineers_2E_CH16.indd 151
Γθξη θ Γηη Γθθη
η ξ +η ξ − 2 ξ +η2 2
2
0
ξ ξ +η η 2 ξ +η2
0 θ Γξθ θ Γηθ = 0 Γθθθ 1 ξ
2
2
0
0 0 1
η
0 ξη 2 − 2 ξ +η2 0
0 2 ηξ − 2 ξ +η2 0
1 ξ 1 η 0
10/3/2020 10:32:14 AM
152 • Tensor Analysis for Engineers, 2E
16.5 EXAMPLE: QUANTITIES RELATED TO BI-POLAR COORDINATE SYSTEMS Consider a bi-polar coordinate system ( x1 , x2 , x3 ) ≡ (ξ ,η , z ) given in terms
of Cartesian ( y1 , y2 , y3 ) ≡ ( X , Y, Z ) , as shown in Figure 16.4. The functional a sinhη X = coshη − cos ξ a sin ξ relations are Y = . Find the basis vectors e1 , e2 , e3 for the coshη − cos ξ Z = z bi-polar coordinate system in terms of the Cartesian unit vectors. Also find the scale factors, unit vectors, metric tensors (covariant and contravariant), Jacobian, volume element, and Christoffel symbols of the 2nd kind.
FIGURE 16.4 Bipolar coordinates: ξ and η iso-surfaces (Foci are located at (−1, 0) and (1, 0)1.
(https://commons.wikimedia.org/wiki/File:Bipolar_isosurfaces.png)
1
Tensor-Analysis-for-Engineers_2E_CH16.indd 152
10/3/2020 10:32:15 AM
Worked-out Examples • 153
Solution:
j ∂y The covariant basis vector ( eξ , eη , ez ) reads, using ek = k E j ∂x
− a sinhη sin ξ a ( coshη cos ξ − 1) E1 + E2 = eξ 2 2 cosh cos cosh cos η − ξ η − ξ ( ) ( ) a (1 − coshη cos ξ ) a sinhη sin ξ E1 − E2 = eη 2 ( coshη − cos ξ )2 ( coshη − cos ξ ) ez = E3 e= The scale factors, after simplification, are h= ξ ξ a coshη − cos ξ
a , h= e= η coshη − cos ξ η , and hz = 1. The unit vectors are, using e ( k ) = ek / hk,
− sinhη sin ξ coshη cos ξ − 1 e ξ E1 + E2 = ( ) coshη − cos ξ coshη − cos ξ 1 − coshη cos ξ sinhη sin ξ = E1 − E2 e (η ) coshη − cos ξ coshη − cos ξ e ( z ) = E 3 The
covariant
a2 2 ( coshη − cos ξ ) 0 gij = 0
metric
tensor 0 a2
( coshη − cos ξ )2 0
reads,
using
gij= ei ⋅ e j ,
0 0 . Therefore, the parabolic 1
coordinate system is an orthogonal one, since metric tensor is a diagonal
Tensor-Analysis-for-Engineers_2E_CH16.indd 153
10/3/2020 10:32:16 AM
154 • Tensor Analysis for Engineers, 2E matrix with null off-diagonal elements. The contravariant metric tensor is the inverse of the covariant one, or ( coshη − cos ξ )2 0 0 2 a 2 −1 ( coshη − cos ξ ) . Jacobian is the g ij = gij = 0 0 a2 0 0 1 square root of the determinant of gij , or the product of all scale factors, a2 dξ dη dθ . The volume element= is d = = = gij ( coshη − cos ξ )2
( )
a2
( coshη − cos ξ )2
dξ dη dz. Similarly, as given in Example 16.5, the Christoffel
symbols can be calculated as
Γξξξ ξ Γξ = Γηξ Γξzξ
Γηξξ η Γη = Γηξ Γηzξ
Γξξη ξ Γηη Γξzη
Γηξη η Γηη Γηzη
sin ξ cos ξ − coshη Γξξ z sinhη Γηξ z = cos ξ − coshη Γξzz 0
sinhη cos ξ − coshη − sin ξ cos ξ − coshη 0
0 0 0
− sinhη cos ξ − coshη Γηξ z sin ξ Γηη z = cos ξ − coshη Γηzz 0
sin ξ cos ξ − coshη sinhη cos ξ − coshη 0
0 0 0
Γ z =[ 0 ]
Tensor-Analysis-for-Engineers_2E_CH16.indd 154
10/3/2020 10:32:17 AM
Worked-out Examples • 155
16.6 EXAMPLE: APPLICATION OF CONTRAVARIANT METRIC TENSORS Using the property of contravariant metric tensors, show that the following relation holds, g ilε ljk A j Bk = ε ijk A j Bk. Solution: Knowing that metric tensors replace repeated indices with the free index while raising (contravariant metric tensor) or lowering (covariant metric tensor) them. We can write the expression in the L.H.S of the above relation as, g ilε ljk A j Bk = jki A j Bk. But we can write A j = g jn An and Bk = g km Bm , substituting, gives g ilε ljk A j Bk = jn jn km i km i in km inm jk g= An g Bm jk g= g An Bm = AB An Bm , after changing k g n m dummy indices, i.e., n → j and m → k, we get inm An Bm = ijk A j Bk. Therefore, the given relation holds.
16.7 EXAMPLE: DOT AND CROSS PRODUCTS IN CYLINDRICAL AND SPHERICAL COORDINATES Given that A ( i ) and B ( i ) are the physical components of vectors A and B, respectively, write down the expressions for physical components of the cross-product, A × B, and dot-product A ⋅ B in terms of A ( i ) and B ( i ), for the following coordinate systems (assume a 3D space, N = 3 ):
a. Cylindrical polar ( x1 , x2 , x3 ) ≡ ( r ,θ , z ) b. Spherical polar ( x1 , x2 , x3 ) ≡ ( r ,ϕ ,θ ) Solution:
The dot-product and cross-product relations are A ⋅ B = A i Bi, C i = ijk A j Bk (contravariant component). We use the relations between covariant/ contravariant and physical components, or Ai = hi A ( i ) and A i = A ( i ) / hi .
a. For cylindrical coordinates, we have the scale factors as ( h1 , h2 , h3 ) = (1, r,1) and the Jacobian is = r (see Section 11.6.1). We = B1 h= B (= 1 ) , B2 h= rA ( 2= can write B(3). ) , B3 h= 1 B (1 ) 2B(2) 3B(3)
Tensor-Analysis-for-Engineers_2E_CH16.indd 155
10/3/2020 10:32:20 AM
156 • Tensor Analysis for Engineers, 2E A(2) A(2) A (1 ) A(3) ,= A2 = = A (1 ) , = A3 = A ( 3 ) . h2 r h1 h3 Therefore, the dot-product reads A⋅B = A i B=i A1 B1 + A2 B2 + A3 B=3 A (1 ) B (1 ) + A ( 2 ) B ( 2 ) + A ( 3 ) B ( 3 ) , a scalar quantity. The physical component of the cross-product i hi ijk A j Bk, no-summation on index i. Expanding is C= ( i ) h= iC e1 jk 1 the expression gives C= h1 1 jk A= A= (1) h= 1C j Bk j Bk 1 A ( 2 ) B ( 3 ) − A ( 3 ) B ( 2 ), and C ( 2 ) = h2 A ( 2 ) h3 B ( 3 ) − h3 A ( 3 ) h2 B ( 2 ) = r e2 jk h2 C2 = r 2 jk A j Bk = r A j Bk = h3 A ( 3 ) h1 B (1 ) − h1 A (1 ) h3 B ( 3 ) = e3 jk 3 jk A ( 3 ) B (1 ) − A (1 ) B ( 3 ) , and C (= 3 ) = h3 C3 h= A j Bk A j Bk = 3 1 A (1 ) B ( 2 ) − A ( 2 ) B (1 ) . h1 A (1 ) h2 B ( 2 ) − h2 A ( 2 ) h1 B (1 ) = r
Similarly, = A1
b. For spherical coordinates, we have the scale factors as ( h1 , h2 , h3 ) = (1, r, r sin ϕ ) and the Jacobian is = r 2 sin ϕ , see section 11.6.1. We = B1 h= B (= 1 ) , B2 h= rA ( 2= can write B(3). ) , B3 h= 1 B (1 ) 2B(2) 3B(3) A (1 ) A(2) A(2) A(3) ,= = A (1 ) , = A2 = A3 = A ( 3 ) . h1 h2 r h3 Therefore, the dot-product reads A⋅B = A i B=i A1 B1 + A2 B2 + A3 B= A (1 ) B (1 ) + A ( 2 ) B ( 2 ) + A ( 3 ) B ( 3 ) , 3 a scalar quantity.
Similarly, = A1
i The physical component of the cross-product is C= hi ijk A j Bk , ( i ) h= iC 1 no-summation on index i. Expanding the expression gives C= (1) h= 1C e1 jk 1 h1 1 jk A j Bk = A j Bk = h2 A ( 2 ) h3 B ( 3 ) − h3 A ( 3 ) h2 B ( 2 ) = 2 r sin ϕ
Tensor-Analysis-for-Engineers_2E_CH16.indd 156
10/3/2020 10:32:23 AM
Worked-out Examples • 157
A (2) B(3) − A (3) B(2),
and
2 C= r 2 jk A= r ( 2 ) h= 2C j Bk
e2 jk
A= j Bk
A ( 3 ) B (1 ) − A (1 ) B ( 3 ) , and C ( 3 ) = h3 A ( 3 ) h1 B (1 ) − h1 A (1 ) h3 B ( 3 ) = e3 jk 1 h3 C3 = h3 3 jk A j Bk = r sin ϕ A j Bk = h1 A (1 ) h2 B ( 2 ) − h2 A ( 2 ) h1 B (1 ) = r A (1 ) B ( 2 ) − A ( 2 ) B (1 ) .
16.8 EXAMPLE: RELATION BETWEEN JACOBIAN AND METRIC TENSOR DETERMINANTS Show that determinant g of metric tensor gij is equal to the square of Jacobian , in a coordinate system, when measured against Cartesian system. Solution: Let’s consider two arbitrary system x i and x′i . We can write the metric ten∂x k ∂x l gkl , since it is a doubly covariant tensor. ∂x′i ∂x′ j ∂x k ∂x k ∂x l g Taking determinant of the both sides gives g′ij = . But kl ∂x′i ∂x′i ∂x′ j
sor transformation as g′ij =
∂x l is the Jacobian of transformation, hence g′ij = 2 gkl . Rewriting j ∂x′ g′ the expression and using gkl = g and g′ij = g′ , we get g′ = 2 g , or = . g
or
Now if the original x i system is the Cartesian system, then gcartesian = 1 and we receive g′ = 2.
Tensor-Analysis-for-Engineers_2E_CH16_3pp.indd 157
9/30/2020 1:01:29 PM
158 • Tensor Analysis for Engineers, 2E
16.9 EXAMPLE: DETERMINANT OF METRIC TENSORS USING DISPLACEMENT VECTORS For an arbitrary coordinate system x i ( , , ) the magnitude of displace ment vector is given by ds2 = 5 dξ 2 + 3 dη 2 + 7 dζ 2 − 4 dξ dη − 8 dξ dζ + 2 dη dζ . Find g and g ij (i.e., the determinant of covariant metric tensor and the contravariant metric tensor). Solution: We have ds2 = gij dx i dx j, expanding for the given coordinate system and using gij = g ji, gives ds2 = g11 dξ 2 + g22 dη 2 + g33 dζ 2 + 2 g12 dξ dη + 2 g13 dξ dζ + 2g23 dη dζ . Therefore, after comparison we receive g11 = 5 , g22 = 3 , g33 = 7 , g12 = −2 , g13 = −4 , and g23 = 1 . The matrix form of covariant metric tensor 5 −2 −4 is gij = −2 3 1 . Determinant of gij = g =5 ( 21 − 1 ) + 2 ( −14 + 4 ) − −4 1 7 4 ( −2 + 12 ) =40. Note that the given system is not orthogonal, since off-
diagonal elements of gij are not zero. For contravariant metric tensor, we find 5 −2 −4 5 −2 −4 ij the inverse of gij = −2 3 1 , or g = −2 3 1 −4 1 7 −4 1 7
ij This is calculated using the formula g=
−1
1 1 1 2 4 4 1 19 3 . = 4 40 40 1 3 11 4 40 40
Cij cofactor of gij , for = g determinante of gij
3 1 5 −2 − −4 1 C23 C11 1 7 21 − 1 −5 + 8 3 23 11 example; g= , = = = = = = 20, g= 40 40 40 40 40 40 40 etc.
Tensor-Analysis-for-Engineers_2E_CH16.indd 158
10/3/2020 10:32:30 AM
Worked-out Examples • 159
16.10 EXAMPLE: DETERMINANT OF A 4 × 4 MATRIX USING PERMUTATION SYMBOLS 5 8 15 21 4 10 13 11 , using Equation 9.13. Calculate determinant of matrix M = 3 6 9 12 2 4 7 1 Solution: We expand based on row, hence M = eijkl M1 i M2 j M3 k M4 l. Expanding we get M = e1234M11M22M33M44 + e1243M11M22M34M43 + e1342M11M23M34M42 + e1324M11M23M32M44 + e1423M11M24M32M43 + e1432M11M24M33M42 + e2134M12M21M33M44 + e2143M12M21M34M43 + e2341M12M23M34M41 + e2314M12M23M31M44 + e2413M12M24M31M43 + e2431M12M24M33M41 + e3124M13M21M32M44 + e3142M13M21M34M42 + e3241M13M22M34M41 + e3214M13M22M31M44 + e3412M13M24M31M42 + e3421M13M24M32M41 + e4123M14M21M32M43 + e4132M14M21M33M42 + e4231M14M22M33M41 + e4213M14M22M31M43 + e4312M14M23M31M42 + e4321M14M23M32M41. Substituting for the values of the matrix elements and the permutation symbols (i.e., +1 for even and −1 for odd number of interchanges for indices), we receive M = 18. Note that the number of terms out of expansion is equal to 4! = 4 × 3 × 2 = 24, or in general N ! for an N × N matrix.
16.11 EXAMPLE: TIME DERIVATIVES OF THE JACOBIAN 1 ∂ Show that the time derivative of the Jacobian reads = ∇ ⋅ V, where V ∂t is the velocity vector. Solution: Considering the transformation from coordinate system x i to that of x′i, we can write the Jacobian as the determinant of the transformation, using ∂x1 ∂x2 ∂x3 . Taking the time derivative, we have ∂x′k ∂x′n ∂x′l ∂ ∂x1 ∂x2 ∂x3 ∂x1 ∂ ∂x2 ∂x3 ∂x1 ∂x2 ∂ ∂x3 ∂ = ε knl k + k n l+ k . l n n l ∂t ∂x′ ∂t ∂x′ ∂x′ ∂x′ ∂x′ ∂t ∂x′ ∂x′ ∂x′ ∂x′ ∂t
Equation 4.4, or = ε knl
Tensor-Analysis-for-Engineers_2E_CH16.indd 159
10/3/2020 10:32:32 AM
160 • Tensor Analysis for Engineers, 2E
But we can write
∂x1 ∂x2 ∂x3 = V 1, = V 2, = V 3, i.e., contravariant components ∂t ∂t ∂t
of velocity vector. Hence, using chain-rule or ε
knl
∂x i ∂x i ∂x j we can write ∂ = = j n n ∂x′ ∂x ∂x′ ∂t
∂V 1 ∂x i ∂x2 ∂x j ∂x3 ∂x p ∂x1 ∂x j ∂V 2 ∂x i ∂x3 ∂x p ∂x1 ∂x j ∂x2 ∂x i ∂V 3 ∂x p . + j i ′k j ′n p ′l + j ′k ′l ∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x i ∂x′n ∂x p ∂x′l ∂x ∂x′k ∂x i ∂x′n ∂x p ∂x ≠ 0; i = 1 & j = 2 & p= 3 ≠ 0; j = 1 & i = 2 & p= 3 ≠ 0; j = 1 & i = 2 & p = 3
Therefore, after rearranging the terms, in each group, we receive ∂ ∂V 1 knl ∂x1 ∂x2 ∂x3 ∂V 2 knl ∂x1 ∂x2 ∂x3 ∂V 3 knl ∂x1 ∂x2 ∂x3 = 1 ε + ε + ε . ∂t ∂x ∂x′k ∂x′n ∂x′l ∂x2 ∂x′k ∂x′n ∂x′l ∂x3 ∂x′k ∂x′n ∂x′l
But the terms in each bracket is the Jacobian, and hence we receive
1 ∂V i ∂V 1 ∂V 2 ∂V 3 ∂ = 1 + 2 + 3 = i = ∇ ⋅ V , ∂x ∂t ∂x ∂x ∂x
which
∇⋅V
1 ∂ yields = ∇ ⋅ V. For divergence of a vector, see Equation 11.12. ∂t
16.12 EXAMPLE: COVARIANT DERIVATIVES OF A CONSTANT VECTOR Show that for a constant vector C, the covariant derivative C ,i j is not necessarily zero. Find the expression for its derivative.
Solution:
∂ We have C = C i ei ; writing its covariant derivative, we receive C i ei ) e j . j ( ∂x i ∂ ∂C i ∂ The covariant component then reads as j ( C i ei ) = e = j ei + C j ( i) ∂x ∂ x ∂ x ∂C i ∂C i i k k i Γ= e ei j + C Γ kj . Now, since C is a constant, the first j ei + C ijk ∂x ∂x i↔ k C′ ij
term in the bracket vanishes but the second term does not. Therefore, we i receive C′= C k Γ kji . j
Tensor-Analysis-for-Engineers_2E_CH16.indd 160
10/3/2020 10:32:33 AM
Worked-out Examples • 161
16.13 EXAMPLE: COVARIANT DERIVATIVES OF PHYSICAL COMPONENTS OF A VECTOR i A A= ei A ( i ) e ( i ), write down the expression for A ( i ) j. For the vector= ′ Use the of the vector A in terms its physical components, covariant derivative or ∇ j A = ∇ j ( A ( i) e ( i)) = e ( i) A ( i) j . ′ Solution: The covariant derivative, in terms of contravariant component of the ∂A i k i ∇ j ( A i ei ) = vector is ∇ j A = j + A Γ kj ei . Substituting for contravariant ∂x i components A = A ( i ) / hi and covariant basis vector ei = hi e ( i ) , we get ∇ j A = A ( i) ∂ h i + A ( k ) Γ i h e (= i) kj ∂x j i hk
∂ ( A ( i ) ) A ( i ) ∂ ( hi ) h i − + A ( k ) Γ kji e ( i ) . j j hi ∂x hk ∂x
∂ ( A ( i ) ) A ( i ) ∂ ( hi ) hi A ( i) − + A ( k ) Γ kji =∇ j A ( i ) − ∇ j hi + Therefore, A ( i ), j = j j ∂x hi ∂x hk hi
hi A ( k ) Γ kji . Note that the Christoffel symbol could be written in terms of hk physical components/ Unit vectors, i.e., Γ kji could be written in terms of e ( i ) , designated by Γˆ i (see Example 10.4.1). kj
16.14 EXAMPLE: CONTINUITY EQUATIONS IN SEVERAL COORDINATE SYSTEMS The continuity for a fluid with density ρ and velocity V i is given as ∂ρ 1 ∂ + 0. Write this equation in tensor notation for N = 3, and (ρ V i ) = ∂t ∂x i express the continuity equation in: a. Cartesian coordinates ( x, y, z) with velocity physical components given as V = (Vx , Vy , Vz ) .
Tensor-Analysis-for-Engineers_2E_CH16.indd 161
10/3/2020 10:32:36 AM
162 • Tensor Analysis for Engineers, 2E b. Cylindrical coordinates (r ,θ , z) with velocity physical components given as V = (Vr , Vθ , Vz ). c. Spherical coordinates (r ,ϕ ,θ ) with velocity physical components given as V = (Vr , Vϕ , Vθ ) . Solution:
∂ρ 1 ∂ ∂ ∂ + 1 ( ρ V 1 ) + 2 ( ρ V 2 ) + 3 ( ρ V 3 ) = 0. ∂t ∂x ∂x ∂x a. Jacobian for a Cartesian system is unity and physical and contravariant components are identical, hence the continuity equation reads, Summation on i gives,
∂ρ ∂ ∂ ∂ + ( ρ Vx ) + ρ Vy + ( ρ Vz ) = 0. ∂t ∂x ∂y ∂z
(
)
b. Jacobian for cylindrical systems is = r and V 1 = Vr , V 2 = Vθ / r , V 3 = Vz . Hence, the continuity reads, after some manipulations, ∂ρ 1 ∂ 1 ∂ ∂ 0. + ( r ρ Vr ) + ( ρ Vθ ) + ( ρ Vz ) = r ∂θ ∂t r ∂r ∂z c. Jacobian for spherical systems is = r 2 sin ϕ and V 1 = Vr, V 2 = Vϕ / r, V 3 = Vθ / ( r sin ϕ ). Hence, the continuity reads, after some manipulations,
∂ρ ∂ ∂ 1 ∂ 2 r sin ϕρ Vr ) + r sin ϕρ Vϕ ) + 0. + 2 ( r ρ Vθ ) = ( ( ∂t r sin ϕ ∂r ∂ϕ ∂θ
16.15 EXAMPLE: 4D SPHERICAL COORDINATE SYSTEMS A 4D spherical system x i ≡ (r,ψ ,θ ,ϕ ) with two polar angles ψ and θ range from 0 to π and azimuth angle ϕ from 0 to 2π , is related to the 4D Cartesian coordinates yi ≡ ( x, y, z, w) as x = r sinψ sin θ cos ϕ y = r sinψ sin θ sin ϕ z = r sinψ cosθ w = r cosψ a. Find the Jacobian for the spherical system and volume of a 4D sphere in this coordinate system. b. Find the expression for the Laplacian of a scalar in the spherical coordinates and write out it in detail.
Tensor-Analysis-for-Engineers_2E_CH16.indd 162
10/3/2020 10:32:40 AM
Worked-out Examples • 163
Solution:
∂ ( x, y, z, w ) = 1. Jacobian reads ∂= ( r,ψ ,θ ,ϕ )
sinψ sin θ cos ϕ sinψ sin θ sin ϕ sinψ cosθ cosψ
r cosψ sin θ cos ϕ r cosψ sin θ sin ϕ r cosψ cosθ − r sinψ
∂x ∂r ∂y ∂r ∂z ∂r ∂w ∂r
∂x ∂x ∂x ∂ψ ∂θ ∂ϕ ∂y ∂y ∂y ∂ψ ∂θ ∂ϕ = ∂z ∂z ∂z ∂ψ ∂θ ∂ϕ ∂w ∂w ∂w ∂ψ ∂θ ∂ϕ
r sinψ sin θ cos ϕ r sinψ cosθ sin ϕ − r sinψ sin θ
− r sinψ sin θ sin ϕ r sinψ sin θ cos ϕ 0 0
0
=
− r sin ψ sin θ . 3
The
2
differential
volume
is
d = drdψ dθ dϕ,
with
0 ≤θ &ψ ≤π
and 0 ≤ ϕ ≤ 2π . Volume is then a quadruplet integral of d , or π
π
2π
π
1 − cos2ϕ π 2 R 4 4 . = = d ϕ π R ∫ ∫ 2 2 0 0 0 0 0 1 2. The expression for the Laplacian of scalar S reads ∇ 2 S = ∇ i g ij ∇ j S . This coordinate system is orthogonal, since gij = 0 for i ≠ j. For example g12 = e1 ⋅ e2 = r sinψ cosψ sin 2 θ cos2 θ − r sinψ cosψ sin 2 θ sin 2 ϕ + r sinψ cosψ 1 = hi2 . For calculating cos2 θ − r sinψ cosψ = 0 . Therefore, g= ii g ii gii= ei ⋅ ei , need the covariant basis vectors, or R
3 2 = ∫ ( −r ) ∫ sin ψ dψ ∫ sinθ dθ
(
)
= e1 sinψ sin θ cos ϕ E1 + sinψ sin θ sin ϕ E2 + sinψ cosθ E3 + cosψ E4 e2 r cosψ sin θ cos ϕ E1 + r cosψ sin θ sin ϕ E2 + r cosψ cosθ E3 − r sinψ E4 = e3 = r sinψ cosθ cos ϕ E1 + r sinψ cosθ sin ϕ E2 − r sinψ sin θ E3 − r sinψ sin θ sin ϕ E1 + r sinψ sin θ cos ϕ E2 e4 =
Tensor-Analysis-for-Engineers_2E_CH16.indd 163
10/3/2020 10:32:41 AM
164 • Tensor Analysis for Engineers, 2E
which yields, g11 = 1, g22 = r 2 , g33 = r 2 sin 2 ψ , g44 = r 2 sin 2 ψ sin 2 θ . Now, ∇2S =
1 1 1 1 ∇ i g ij ∇ j S = ∇ 2 S = ∇ i ( g ii∇ iS ) = ∇1 ( g11∇1S ) + ∇ 2 ( g 22 ∇ 2 S ) +
(
)
1 1 ∇ 3 ( g 33 ∇ 3 S ) + ∇ 4 ( g 44 ∇ 4 S ) . Each term reads, after substitution 1 −1 ∂ 3 2 ∂S 3 ∂S ∂ 2 S ∇1 ( g11∇1S )= 3 2 + − r sin ψ sin θ = r sin ψ sin θ ∂r ∂r r ∂r ∂r 2 −1 ∂ −2 3 2 ∂S 1 = ∇ 2 ( g 22 ∇ 2 S ) − r r sin ψ sin θ 3 2 ∂ψ r sin ψ sin θ ∂ψ ∂S ∂ 2 S 1 = 2 cotψ + ∂ψ ∂ψ 2 r 1 −1 ∂ − r 3 sin 2 ψ sin θ ∂S ∇ 3 ( g 33 ∇ 3 S ) = r 3 sin 2 ψ sin θ ∂θ r 2 sin 2 ψ ∂θ 1 ∂S ∂ 2 S = 2 2 cot θ + ∂θ ∂θ 2 r sin ψ 1 −1 ∂ − r 3 sin 2 ψ sin θ ∂S ∇ 4 ( g 44 ∇ 4 S ) = r 3 sin 2 ψ sin θ ∂ϕ r 2 sin 2 ψ sin 2 θ ∂ϕ ∂ 2S 1 = 2 2 r sin ψ sin 2 θ ∂ϕ 2 Hence, ∇ 2 S=
3 ∂S ∂ 2 S 1 ∂S ∂ 2 S 1 ∂S ∂ 2 S + 2 + 2 cotψ + + 2 2 cot θ + + 2 r ∂r ∂r r ∂ψ ∂ψ r sin ψ ∂θ ∂θ 2
∂ 2S . r 2 sin 2 ψ sin 2 θ ∂ϕ 2 1
Tensor-Analysis-for-Engineers_2E_CH16.indd 164
10/3/2020 10:32:43 AM
Worked-out Examples • 165
16.16 EXAMPLE: COMPLEX DOUBLE DOT-CROSS PRODUCT EXPRESSIONS × × The expression ∇ A ∇ A is given. The notations mean that the two gradient ×
×
vectors (marked by × ) are cross product to each other as well as the two vectors A, then the resulted vectors of the cross-product operations are dotted. Write the expression in tensor notation and simplify the result. Solution: The two cross-product results are dotted; hence we require the contravariant component of the cross-product of the gradient’s and the covariant component of the cross-product of the vectors A . Therefore, we can write × × ∇ ε ijk ε ilm A l A′mk . Expanding the resulted expression yields, A ∇ A= ′j × ×
(
)
ε ijkε ilm ( A l A′mk ) ′ j = δ l jδ mk ( A l A′ mk ) ′ j − δ mj δ lk ( A l A′ mk ) ′ j = ( Al A′kk) ′ l − ( Ak A′ kj ) ′ j. × × Performing the covariant derivative operations gives ∇ A A ll A kk + A ∇ = ′ ′ × ×
A l A kkl − A kj A kj − A k A jkj = A l l A kk − A k j A j k . The first term of the result can ′ ′ ′ ′ ′ ′ ′ ′ k→ l, j→ k 2 be written in vector notation as A l l A kk = ∇ ⋅ A . Therefore, we finally have ′ ′ × × 2 j k ∇ A ∇ A = ∇ ⋅ A − A ′ j A ′k . Using the method given in this example, apply × × it to the following expressions:
(
(
)
)
× × ijk l = A∇ ε ilm A l Ak m⋅ δ l jδ mk A l Ak m⋅ j − δ mj δ lk A = Ak m⋅ a. ∇ A ε = ′ ′ ′ ′ ′j j ′ × ×
(
( A A ′ )′ j
k
k ⋅
j
)
(
)
) ( A A ′ )′ −
(
j
k k ⋅
j
= A j j Ak ⋅k + A j Ak ⋅ kj − A k j Ak ⋅ j − A k Ak ⋅ jj . In vector notation, the ′ ′ ′ ′ ′ ′ × ×
(
)
resulted expression can be written as ∇ A∇ A = (∇ ⋅ A ) + A ⋅ ∇ (∇ ⋅ A) − × × A k j Ak ⋅ j − A ⋅ ∇ 2 A . ′ ′
(
Tensor-Analysis-for-Engineers_2E_CH16.indd 165
)
2
10/3/2020 10:32:46 AM
166 • Tensor Analysis for Engineers, 2E
l l j b. ∇∇ A A ε ijkε ilm ( A m A= δ l jδ mk − δ mj δ lk ( A m A= ( Ak Ak )′⋅ j − k) ⋅j k) ⋅j = ′ ′ × × 2 k A j Ak ⋅ j = ∇ 2 A2 − 2 A ⋅ ∇ ∇ ⋅ A − A j⋅k Ak j − ∇ ⋅ A . ′ ′ ′ ×
(
×
(
( (
)
)
))
(
)
Note that vector-type notation is not sufficient to express some of the terms in these expressions—another reason to use index/tensor notation for its generality.
16.17 EXAMPLE: COVARIANT DERIVATIVES OF METRIC TENSORS Show that the derivative of metric tensors (i.e., gij and g ij ) is equal to zero. Solution: Since metric tensor is a constant in the Cartesian coordinate system, hence its derivative is null and will be the same for any coordinate system. ∂gij To show this, we use Equation 10.14, or gij = − Γ nik gnj − Γ njk gni . Now k ′ ∂x k using Equation 10.19 we can expand the last two terms on the R.H.S, or ∂gij 1 ∂gij ∂gkj ∂g 1 ∂gij ∂g ∂gkj gij = k − k + i − kij − k + kij − i . The terms on ′k ∂x ∂x ∂x 2 ∂x ∂x ∂x 2 ∂x
( )
( )
( )
the R.H.S cancel out and the expression is equal to zero, hence gij = 0 , ′k and similarly g ij = 0. ′k
( )
16.18 EXAMPLE: ACTIVE ROTATION USING SINGLE-AXIS AND QUATERNIONS METHODS Consider the vector OA = (1, −1, 2 ) connecting the origin to point A in the x , y , z coordinate system. Now we rotate this vector by an angle of 60° ( ) 1 3 about an axis in the direction of unit vector n = 0, , . Calculate the 2 2 resulted vector OA′ , using a) rotation matrix with Equation 12.7, and b)
Tensor-Analysis-for-Engineers_2E_CH16.indd 166
10/3/2020 10:32:48 AM
Worked-out Examples • 167
quaternion method with Equation 15.12. Note that the rotation angle here is given for an active rotation (i.e., vector rotates but coordinate system is kept fix). Solution:
a. Since the vector OA is rotated by 60 degrees, active rotation, then we use θ = −60° in Equation 12.7. Therefore, for n1 = 0, 1 2 3 n2 = 1 / 2, and n3 = 3 / 2 we receive R = 4 3 − 4 The rotated vector component in the ( x, y, z ) is 1 2 3 4 3 − 4
3 4 5 8 3 8
−
3 4 1 3 −1= 8 2 7 8
3 4 5 8 3 8
−
1 3 3 + + 10 + 4 3 2 4 2 3 5 3 1 − + = 1+ 2 3 , 4 8 4 8 14 − 3 3 3 3 7 − + − 8 4 4
3 4 3 . 8 7 8
given by
or
10 + 4 3 1 + 2 3 14 − 3 3 OA′ = , , . 8 8 8 b. Using the active quaternion operator, see Equation 15.12, we need to have the corresponding quaternion based on half of the rotation angle, 1 3 3 1 3 60° + j+ k sin 30 = k. Now, we = 30°, Q = cos30 + j + 2 2 2 4 4 2 n
∗ 3 1 3 1 3 3 + j+ k ( i − j + 2k) − j− k= calculate OA=′ QOAQ= 4 4 2 4 2 4
Tensor-Analysis-for-Engineers_2E_CH16.indd 167
10/3/2020 10:32:51 AM
168 • Tensor Analysis for Engineers, 2E
1 3 1 3 3 3 10 + 4 3 1+2 3 14 − 3 3 1 + + + 2k) j − − 3 k ( i − j= i+ j+ k − i− 4 2 2 4 4 4 8 8 8 10 + 4 3 1 + 2 3 14 − 3 3 10 + 4 3 1+2 3 14 − 3 3 Or OA ′ = , , . k i j 2 k i j k . − = + + + ) ( 8 8 8 8 8 8 Alternatively, we can use Equation 15.14. Readers may want to use Wolfram Alpha (www.wolframalpha.com) to visualize the quaternion as a rotation operator. After accessing the wolfram Web site, enter the phrase “quaternions” in the space provided and use the format provided similar to those examples given on the page.
16.19 EXAMPLE: PASSIVE ROTATION USING SINGLE-AXIS AND QUATERNIONS METHODS Repeat Example 16.18 for a passive rotation of an angle 45˚. Having passive rotation we keep vector OA fixed and rotate coordinates ( x, y, z ). Repeating = the data here, for convenience, OA
1 3 (1, −1, 2 ) , n = 0, , . 2 2
Solution:
a. Since the vector OA is rotated by 45 degrees, passive rotation, then we use θ = 45° in Equation 12.7. Therefore, for n1 = 0, n2 = 1 / 2 , and 2 2 6 n3 = 3 / 2 we receive R = − 4 2 4
6 4 2+3 2 8 2 3− 6 8
components in the rotated coordinates
Tensor-Analysis-for-Engineers_2E_CH16.indd 168
2 4 2 3− 6 . The vector 8 6+ 2 8 −
( x′, y′, z′)
is given by
10/3/2020 10:32:53 AM
Worked-out Examples • 169
2 2 6 − 4 2 4
6 4 2+3 2 8 2 3− 6 8
2 4 1 2 3 − 6 −1 = − 8 2 6 + 2 8 −
2 6 2 − − 2 4 2 −2 6 2 + 3 2 2 3 − 6 1 − + = −2 − 3 2 + 4 4 8 4 8 12 + 4 2 − 2 2 2 3 − 6 6+ 2 − + 4 8 4
2 6 2 − − 2 4 2 −2 6 −2 6 −2 − 3 2 + 4 3 − 4 6 12 + 4 2 − 2 3 + 6 2 + 3 2 2 3 − 6 1 , = − + − = −2 − 3 2 + 4 3 − 4 6 , or OA′ 8 , 8 8 4 8 4 8 2 2 3 − 6 6+ 2 12 + 4 2 − 2 3 + 6 − + 8 4 4 −2 6 −2 − 3 2 + 4 3 − 4 6 12 + 4 2 − 2 3 + 6 , , ≅ ( −0.61237, −1.13905,2.0803 ). 8 8 8
∗
b. The quaternion related to this rotation, using half of the rotation angle, is Q = cos
1 3 π + j+ k sin = 8 2 2 8
π
2+ 2 2− 2 6−3 2 + j+ k. 2 4 4
n
Now,
we
calculate
2+ 2 2− 2 6−3 2 − OAQ j− k ( i − j + 2k) 2 4 4
2+ 2 2− 2 6−3 2 ∗ = (OA )′ Q= − OAQ j− k( i − 2 4 4 2+ 2 2− 2 6−3 2 + j+ k ≅. 2 4 4
( 0.92388 − 0.191342 j − 0.331414 k )( i − j + 2 k )( 0.92389 + 0.191342 j + 0.331414k ) = ( 0.47149 + 0.20979 i − 1.2553 j + 2.03912 k ) ( 0.92389 + 0.191342 j + 0.331414 k ) = −0.61237 i − 1.13905 j + 2.0803 k. Or (OA )′ = (OA )′ =
( −0.61237, −1.13
( −0.61237, −1.13907,2.08032 ) . Alternatively, we can use Equation 15.20.
Tensor-Analysis-for-Engineers_2E_CH16.indd 169
10/3/2020 10:32:55 AM
∗
170 • Tensor Analysis for Engineers, 2E
16.20 EXAMPLE: SUCCESSIVE ROTATIONS USING QUATERNIONS METHOD Use data given in Example 16.18 and rotate the resulted vector, actively in that example through a successive rotation of an angle 45°. For this operation use the quaternion method. Solution: From Example 16.18 we have the resulted vector 10 + 4 3 1 + 2 3 14 − 3 3 1 3 OA′ = , , and the unit vector n = 0, , 8 8 8 2 2 along the axis of rotation. The follow up rotation quaternion is related to rotation angle of 45° = P = cos
1 3 π + j+ k sin = 8 2 2 8
π
π
. Hence, using half-angle we have 4 2+ 2 2− 2 6−3 2 + j+ k . Now, 2 4 4
∗ 2 + 2 2− 2 ′P = = + OA″ POA 2 4 ∗ 2 + 2 2− 2 6 − 3 2 10 + 4 3 1+2 3 14 − 3 3 2 + 2 2− 2 ′P = = + OA″ POA j+ k i+ j+ k − 2 4 4 8 8 8 2 4 1+2 3 14 − 3 3 2 + 2 2− 2 6−3 2 j+ k − j− k ≅ 1.5436 i + 1.8708 j + 0.3425 k. 8 8 2 4 4 We can also use the original vector OA with the dual-quaternion as the product of the two corresponding quaternions, see Equation ∗ 15.18, or ( PQ ) OA ( PQ ) . Note that the q belongs to the first rotausing
n
Equation
15.18
we
can
write
tion and p (60° rotation) to the second one (45° rotation). But
2+ 2 2− 2 6 − 3 2 3 1 3 ˆˆ = + + j+ pq j+ k k ≅ 0.60876 + 0.39668 j + 0.68706 k 2 4 2 4 4 4 0.60876 + 0.39668 j + 0.68706 k . Therefore, its conjugate ( pq ˆ ˆ ) =0.60876 − 0.39668 j − 0.68706 k ∗ ˆ ˆ ) OA ( pq ˆ ˆ ) ( 0.60876 + 0.39668 j + 0.68706 k ) =0.60876 − 0.39668 j − 0.68706 k . Now, we can = pq write OA″ (= ∗
( 0.60876 + 0.39668 j + 0.68706 k )
( i − j + 2 k )( 0.60876 − 0.39668 j − 0.68706 k ) =
1.5436 i + 1.8708 j + 0.3425 k
1.5436 i + 1.8708 j + 0.3425 k. As seen the two methods provide us with identical results.
Tensor-Analysis-for-Engineers_2E_CH16.indd 170
10/3/2020 10:32:59 AM
CHAPTER
17
EXERCISE PROBLEMS 1. Write out the following expression in full detail for N = 3 dimensions using Einstein summation convention: a. A = A ij ei e j A i B j gij b. A ⋅ B = 1 ∂ ∂ψ c. ∇ 2ψ = i g ij j ∂x ∂x
d. C ( i ) = hi e. Akij, n
eijk ∂ ( hk A ( k ) )
∂x j
Akij j i nm Akjm nm Akim mnk Amij x n
2. By rotating bipolar coordinates about X-axis (the axis where foci are located on) we obtain the bi-spherical coordinate system, ( x1 , x2 , x3 ) ≡ (ξ ,η ,ϕ ) a sin ξ cos ϕ X = coshη − cos ξ a sin ξ sin ϕ 1 2 3 given in terms of Cartesian ( y , y , y ) ≡ ( X , Y, Z ), as Y = . coshη − cos ξ a sinhη Z = coshη − cos ξ Find the basis vectors e1 , e2, e3 for the bi-spherical coordinate system in terms of the Cartesian unit vectors. Also find the scale factors, unit vectors, metric tensors (covariant and contravariant), Jacobian, volume element, and Christoffel symbols of the 1st kind.
Tensor-Analysis-for-Engineers_2E_CH17.indd 171
10/3/2020 10:33:50 AM
172 • Tensor Analysis for Engineers, 2E
FIGURE 17.1 Bi-spherical coordinate system.
3. By rotating bipolar coordinates about Y-axis (the axis perpendicular to the line connecting the foci) we obtain a toroidal coordinate system,
( x , x , x ) ≡ (ξ ,η ,ϕ ) given in terms of Cartesian ( y , y , y ) ≡ ( X, Y, Z ) , 1
2
3
1
2
3
a sinhη cos ϕ X = coshη − cos ξ a sinhη sin ϕ as Y = . Find the basis vectors e1, e2 , e3 for the toroidal − cosh η cos ξ a sin ξ Z = coshη − cos ξ coordinate system in terms of the Cartesian unit vectors. Also find the scale factors, unit vectors, metric tensors (covariant and contravariant), Jacobian, volume element, and Christoffel symbols of the 2nd kind.
FIGURE 17.2 Toroidal coordinate system.
Tensor-Analysis-for-Engineers_2E_CH17.indd 172
10/3/2020 10:33:50 AM
Exercise Problems • 173
4. Having displacement vector r = yi Ei defined in Cartesian coordinate system yi, express r in terms of the contravariant and physical components in cylindrical, spherical, bi-polar, bi-spherical, and toroidal coordinate systems. Use the definitions of these systems as given in previous examples and exercises.
= gii 1= / g ii hi2 5. Show that for an orthogonal coordinate system the relation holds, i.e., the diagonal elements of covariant and contravariant metric tensors. Note, no sum on index i. 6. For an orthogonal coordinate system, show that the following relations holds for Christoffel symbols of the 2nd kind (all indices are set to be different): Γ kij = 0, Γ iij =−
hi ∂hi i 1 ∂hi i 1 ∂hi , Γ ij = , Γ ii = . hi ∂x j hi ∂x i h2j ∂x j
7. Find the relations for Christoffel symbols of the 1st kind for cylindrical and spherical coordinate systems. Use results from the example given in Section 10.4.1. 8. Show that covariant differentiation obeys the same rule as does ordinary
(
)
j differentiation, when operating on products, e.g., A i B = A′ik B j + A i B′ kj . ′k
9. Write down the expression for the covariant derivative of a tensor of rank ij five, Akm , using hints from Section 10.14. 10. Show that metric tensors behave like a constant under a covariant dif-
(
)
(
= A′ik gij and Ai g ij
)
= Ai′ k g ij . 11. Show that the recursive relation for the curl of a vector A reads A ( n ) = −∇ 2 A ( n − 2 ), for n > 2. where n is the number of curl operations × ∇ × × ∇ × A (note, A ( 0 ) = A ). and A ( n ) = ∇
ferentiation operation, i.e., A i gij
′k
′k
n times
12. Using the recursive relation for the curl of a vector given in Exercise 11, show that the recursive relation for the biharmonic operator reads A ( n) = ∇ 4 A ( n − 4 ) , for n > 4 . 13. Using the results of Exercises 11 and 12, show that the recursive rela tions for higher order operator ∇ 2 n reads A ( 2 n + 1 ) =− ( 1 ) n ∇ 2 n A (1 ) and A ( 2 n ) = −∇ 2 n− 2 A ( 2 ) .
Tensor-Analysis-for-Engineers_2E_CH17.indd 173
10/3/2020 10:33:52 AM
174 • Tensor Analysis for Engineers, 2E 14. For the 4D spherical coordinate system from Example 15.15, calculate the Christoffel symbols: Γ rrr r Γ =
Γ rrψ r Γψψ
Γ rrθ r Γψθ
r Γθθ
Γθrr θ Γ =
Γθrψ θ Γψψ
Γθrθ θ Γψθ
Γθθθ
Γψrr Γ rrϕ r Γψϕ , Γψ = r Γθϕ r Γϕϕ
Γψrψ ψ Γψψ
Γψrθ ψ Γψθ
Γψθθ
Γθrϕ Γϕrr θ Γψϕ , ϕ θ Γ = Γθϕ θ Γϕϕ
Γϕrψ ϕ Γψψ
Γϕrθ ϕ Γψθ
ϕ Γθθ
Γψrϕ ψ Γψϕ , Γψθϕ Γψϕϕ Γϕrϕ ϕ Γψϕ . ϕ Γθϕ Γϕϕϕ
15. Show that the following expressions simplify as given (see Example 15.16): × × k j jk 2 A = ∇ A A A + A ⋅ ∇ A − A A − A ⋅ ∇ ∇⋅A a. ∇ ′ k j k j ′ ′ ⋅ ′ ⋅ × × × × j 2 k b. ∇ A ∇∇ Φ = ∇ ⋅ A ( ∇ Φ ) − A′ j ( ∇ k Φ )′ ⋅
(
×
(
×
( (
)
))
)
× × 2 jk 2 2 c. ∇∇ Φ =∇ A ⋅ ( ∇Φ ) − ∇ ∇ ⋅ A ⋅ ( ∇Φ ) − A′ ⋅ ( ∇ k Φ ) − ∇ ⋅ A ( ∇ Φ ) − A ⋅ ∇ ( ∇ Φ ) A∇ × × jk 2 ⋅ A ⋅ ( ∇Φ ) − A′ ⋅ ( ∇ k Φ ) − ∇ ⋅ A ( ∇ Φ ) − A ⋅ ∇ ( ∇ 2 Φ )
(
)
(
)
(
)
(
)
(
)
(
)
16. Write the below expressions in index notation: a. V ∇ ⋅ A + ∇ ⋅ V A b. A ⋅ B × C c. A ⋅ ∇ ∇ ⋅ A
(
)
(
(
)
)
( (
))
17. Show that the following relations holds:
δ li δ mi a. ε ijk ε lmn =δ l j
δ lk
δ ni
ijk k ijk = δ mj δ nk − δ njδ mk , c. ε ε ijn =2δ n δ mj δ nj , b. ε ε imn δ mk δ nk
d. ε ijk ε ijk =6
Tensor-Analysis-for-Engineers_2E_CH17.indd 174
10/3/2020 10:33:54 AM
)
Exercise Problems • 175
18. Using Equation 9.13, write the determinant for a 5 × 5 matrix. 19. Use Equation 13.19 to find physical components of the convective time derivative of a velocity vector, and express the results for (3D) (hint: see Example 15.14): a. Cartesian coordinates b. Cylindrical coordinates c. Spherical coordinates 20. Show that the non-zero Christoffel symbols of the 1st kind for cylindrical and spherical coordinates read: a. Cylindrical coordinates: Γ rθ ,θ = r and Γθθ ,r = −r . r2 − r sin 2 ϕ , Γθθ ,ϕ = − sin 2ϕ , − r , Γθθ , r = b. Spherical coordinates: Γϕϕ ,r = 2 r2 2 Γ rϕ ,ϕ = r and Γ rθ ,θ = r sin ϕ , Γϕθ ,θ =sin 2ϕ 2 21. Using cross-product expression for two vectors, show that ( is the Jacobian): a. ijk =
1 mnl gmi gnj glk
1 ijk mnl 2 b. = gmi gnj gkl 6 22. Calculate the single equivalent rotation angle and rotation axis for transforming Cartesian system ( y1 , y2 , y3 ) to ( y′1 , y′2 , y′3 ) such that y′1 coincides with y1, y′2 coincides with y3 , and y′3 coincides with y1 .
23. Find the rotation matrix for transforming ( y1 , y2 , y3 ) : a. Rotate about (0, 0, 1) for 45° b. Then, rotate about (1, 0, 0) for 60° c. Then, rotate about (0, 1, 0) for 75°
d. Find the final rotation matrix and equivalent rotation angle e. Find equivalent axis of rotation
Tensor-Analysis-for-Engineers_2E_CH17.indd 175
10/3/2020 10:33:56 AM
176 • Tensor Analysis for Engineers, 2E ˆ Q + Q and 24. Show the following relations hold for unit quaternion Q = 0 arbitrary vector V . ˆ ˆ ∗ = Q2 − Q 2 V + 2 Q ⋅ V Q + 2Q Q × V , active rotation a. QVQ 0 0 b. Qˆ ∗ VQˆ =
)
(
( 2Q
2 0
(
)
(
)
− 1 ) V + 2 V ⋅ Q Q + 2Q0 V × Q , passive rotation
(
)
(
)
25. Derive The rotation matrix given by Equation 15.22, using unit quater nion Qˆ and arbitrary vector V . 26. Using Example 16.18 rotation matrix result, calculate the corresponding Euler angles. 27. Using Example 16.19 rotation matrix result, calculate the corresponding Euler angles. 28. Using the quaternions method as shown in Example 16.20, calculate the = (1, −1,2 ) after two successive rotations. The components of vector OA 1 3 first rotation is by an angle of 80° about the unit vector n = 0, , 2 2 and the second rotation is by an angle of 50° about the unit vector
3 1 n= , ,0 . Perform the calculations for both active and passive 2 2 rotation scenarios. 29. For Exercise 28, calculate the corresponding rotation matrix using the product of the quaternions involved. Also using the resulted rotation matrix calculate the equivalent axis and angles of single rotation. 30. A pilot maneuvers its airplane through a pitch-roll-yaw series or rotations. Using the relation for the corresponding rotation matrix yxz analyze the gimbal lock orientation when roll angle is equal to 3π . 2 31. Show that quaternion operation for active rotation ( QVQ∗ ) preserves dot-products and triple products, hence it is a rotation. 32. Repeat the Exercise 16.20 considering the rotation as a passive one (i.e., coordinates are rotated).
Tensor-Analysis-for-Engineers_2E_CH17.indd 176
10/3/2020 10:33:57 AM
REFERENCES 1. G. Ricci and T. Levi-Civita. “Méthodes de calcul differéntiel absolu et leurs applications,” Math. Ann., vol. 54, 1901. 2. Y. I. Dimitrienko, Tensor Analysis and Nonlinear Tensor Functions, Springer, 2002. 3. D. Leigh, Nonlinear Continuum Mechanics: An Introduction to Continuum Physics and Mathematical Theory of the Nonlinear Mechanical Behavior of Materials, McGraw-Hill, 1968. 4. I. Sokolnikoff, Tensor Analysis: Theory and Applications, John Wiley & Sons, 1951. 5. P. Grinfeld, Introduction to Tensor Analysis and the Calculus of Moving Surfaces, Springer, 2010. 6. R. J. Kee, M. E. Coltrin, and P. Glarborg, Chemically Reactive Flow: Theory & Practice, John Wiley & Sons, 2033. 7. G. G. Bach, Tensor Analysis and Applications (lecture notes), Montreal, Canada: Mech. Eng., McGill University, 1979. 8. E. Christoffel, “Über die Transformation der homogenen Differentialausdrücke zweiten Grades,” Journal für die reine und angewandte Mathematik, vol. B, pp. 46–70, 1869. 9. O. Rodrigues, “Des lois géometriques qui regissent les déplacements d’ un systéme solide dans l’ espace, et de la variation des coordonnées provenant de ces déplacement considérées indépendant des causes qui peuvent les produire,” J. Math. Pures Appl., vol. 5, pp. 380–440, 1840. 10. E. Kreyszig, Advanced Engineering Mathematics, 10th ed., John Wiley & Sons, 2011. 11. W. Z.U.A., Fluid Dynamics: Theoretical and Computational Approaches, 3rd ed., CRC Press, 2005. 12. “Anisotropic Heat Transfer through Woven Carbon Fibers,” COMSOL, Application ID: 16709. 13. B. K. Horn, “Some Notes on Unit Quaternions and Rotation,” 2001. 14. D. Koks, Explorations in Mathematical Physics: The Concepts Behind an Elegant Language, Springer, 2006.
Tensor-Analysis-for-Engineers_2E_CH18_Ref.indd 177
10/3/2020 10:34:13 AM
178 • Tensor Analysis for Engineers, 2E
15. Erik B. Dam, M. Koch, and M. Lillholm, “Quaternion, Interpolation and Animation,” Department of Computer Science, University of Copenhagen, Denmark, 1998. 16. S. Oldenburger, “Applications of Quaternions. Written project of the course “Problem Solving Techniques in Applied Computer Science,” Department of Computer Science, Iowa State University, 2005. 17. H. Goldstein, C. Poole, and J. Safko, Classical Mechanics, 3rd Ed., Addison Wesley, 2000. 18. J. Kuipers, Quaternions and Rotation: A Primer with Applications to Orbits, Aerospace and Virtual Reality, Princeton University Press, 1999. 19. R. Brannon, Rotation, Reflection, and Frame Changes: Orthogonal tensors in computational engineering mechanics, IOP Publishing Ltd, 2018. 20. W. Heard, Rigid Body Mechanics: Mathematics, Physics and Applications, WILEYVCH Verlag GmbH & Co, 2005. 21. J. Vince, Quaternions for Computer Graphics, Hardcover, Springer, 2011. 22. E. Zeleny, “Controling Airplane Flight,” Wolfram Demonstration Project, 2011. 23. S. Blinder, “Euler Angles for Space Shuttle,” Wolfram Demonstrations Project, 2011. 24. F. W. Strauch, “Euler Angles,” Wolfram Demonstrations Project, 2011. 25. D. Hosier, “Avoiding Gimbal Lock in a Trajectory Simulation,” U.S. Army Armament Research Development and Engineering Center, New Hersey, 2018. 26. D. Hoag, “Apollo Guidance and Navigation: Consideration of Apollo IMU Gimbal Lock,” Apollo Lunar Surface Journal, Cambridge, MIT, 1963. 27. H. Schaub, “Spacecraft Dynamics & Control,” University of Colorado Boulder, 2019. 28. GuerrillaCG, “Euler (gimbal lock) Explained,” 2009. 29. “Choose Rotation Order and Gimbal Lock | Blender 2.8x – Rigging Tutorial,” 2019. 30. W. R. Hamilton, “On quaternions or a new system of imaginaries in algebra,” Philosophical Magazine, link to David R. Wilkins collection at Trinity College, Dublin, Dublin. 31. A. J. Hanson, “Quaternion Application,” School of Informatics and Computing , Indiana University, 2012. 32. L.C. Bruno and L.W. Baker, “Math and mathematicians : the history of math discoveries around the world,” Detroit, Mich. : U X L, 1999. 33. YouTube, “Visualizing 4D Geometry – A Journey Into the 4th Dimension [Part 2],” 2017. 34. E. W. Weisstein, “Euler Formula From MathWorld—A Wolfram Web Resource”. 35. E. Cho, “De Moivre’s Formula for Quaternions,” Appl. Math. Lett, vol. 11, no. 6, pp. 33–35, 1998. 36. D. Henderson, “Euler Angles, Quaternions, A140 Transformation Matrices for Space Shuttle Analysis,” NASA-McDonnell Douglas Technical Services, 1977.
Tensor-Analysis-for-Engineers_2E_CH18_Ref.indd 178
10/3/2020 10:34:13 AM
INDEX A acceleration vector contravariant components of, 76–78 physical components in, 78–79 cylindrical coordinate systems, 81–82 orthogonal systems, 79–82 spherical coordinate systems, 81–82 substantial time derivation, 84–85 active rotations, 104–105 definition, 105 example, 105–107 arbitrary coordinate systems, 4, 31
B basis vectors, 5–7 biharmonic operations on tensors, 58–59
C Cartesian coordinate system, 3–4, 89–91 equivalent single rotation to sequential rotations of, 70–73 fluid in motion, 83 Cartesian tensors, 65–66 equivalent single rotation, 67–73 rotation matrix, 67 characteristic equation, 70 Christoffel symbols, 52, 58, 59, 77–80 of 2nd kind, 35–39 for cylindrical coordinates, 44–46 for 2D polar coordinate system, 38 relations and properties, 41–44 for spherical coordinates, 46–49 of 1st kind, relations and properties, 41–44
Tensor-Analysis-for-Engineers_2E_CH19_Index.indd 179
complex numbers, 128 component notation, 2 connection coefficients, 37 conservation equations, coordinate independent forms, 85–87 contravariant components, 31, 33 of acceleration vector, 76–78 of curl, 52 and transformations, 9–12 contravariant derivatives of vectors, 39–40 contravariant metric tensor, 22 contravariant permutation tensors, 32 coordinate systems Cartesian, 89–91 cylindrical, 91–93 definition, 3–4 orthogonal curvilinear, 98–101 parabolic, 96–98 spherical, 93–96 covariant components, and transformations, 13–14 covariant derivatives of mixed tensor, 40 of vectors, Christoffel symbols of the 2nd kind, 35–39 covariant/contravariant basis vectors, 22 covariant permutation tensors, 32 cross-product, 29–33, 53 curl operations on tensors, 51–54 cylindrical coordinate systems, 3, 91–93 acceleration vector physical components in, 81–82 Christoffel symbols for, 44–46 example, 23–25
10/3/2020 11:05:07 AM
180 • Index
metric tensor, 23–25 physical components of Laplacian of a vector in, 60–61
D determinant of a 3 × 3 matrix using permutation symbols, 34 divergence operation on tensors, 55–57 dot-product, 29–33, 57. See also cross-product
E eigenvalues and eigenvectors, 67–73 Einstein summation convention, 2 equivalent single rotation, 67–73 Euler angles, 107–110 categorizing, 113–120 derive the quaternion from given, 139–140 example, 110–113, 120–121 for rigid body rotations, 108 exercises, 171–176
F first principal scalar invariant, 70
G Gauss divergence theorem, 85 gimbal lock, 121 gimbal lock-Euler angles limitation, 121–123
H Hamilton quaternion relations, derivation of, 125–126
I index notation, 2 invariant, 6
J Jacobian and metric tensor determinants, relation between, 157 time derivatives of , 159–160
L Lagrange multiplier method, 69 Laplacian of a vector, physical components of, 59–60
Tensor-Analysis-for-Engineers_2E_CH19_Index.indd 180
cylindrical systems, 60–61 spherical systems, 61–63 Laplacian operations on tensors, 57–58 Levi Civita symbol, 30
M metric tensor, 18–22 cylindrical coordinate systems, 23–25 spherical coordinate systems, 25–27 mixed tensor, 18 covariant derivatives of, 40
N Newton’s 2nd law, 74, 76–78, 87
O orthogonal Cartesian system, 3, 30 orthogonal coordinate systems, 33 acceleration vector physical components in, 79–82 curvilinear, 98–101
P parabolic coordinate systems, 96–98 passive rotations, 104–105, 105–107 definition, 105 example, 105–107 permutation symbols, determinant of a 3 × 3 matrix using, 34 physical components of acceleration vector in, 78–79 cylindrical coordinate systems, 81–82 orthogonal systems, 79–82 spherical coordinate systems, 81–82 substantial time derivation, 84–85 curl of tensors-3D orthogonal systems, 54–55 of Laplacian of a vector cylindrical systems, 60–61 spherical systems, 61–63 and transformations, 15–16 proper Euler angles, 113, 114 Pythagorean theorem, 6, 18, 19
Q quaternion algebraic operations and properties, 126–127
10/3/2020 11:05:07 AM
Index • 181
example, 127–128 and rotation, 129–132 definition, 123–125 derive the rotation matrix from a given, 135–138 quaternions multiplication rule, 125
R right-hand-rule convention, 67, 72 rigid body rotations Euler angles for, 108 quaternions-applications for, 123–134 direct relations and methods, 141 Rodrigues’ rotation formula, 67 rotation matrix, 67, 114, 117, 128 calculating, 72 derive the equivalent quaternion from, 138–139 rotation, quaternion and, 129–132 example, 133–134
S scale factor, 7 second principal scalar invariant, 70 spherical coordinate systems, 3, 93–96 acceleration vector physical components in, 81–82 for Christoffel symbols, 46–49 example, 25–27 metric tensor, 25–27 physical components of Laplacian of a vector in, 61–63 substantial time derivatives of tensors, 82–85
T Tait-Bryan angles, 113, 117 third principal scalar invariant, 70 total/convective time derivative, 82 transformation coefficient, 10, 11 transformations, 15–16 contravariant components and, 9–12 covariant components and, 13–14 physical components and, 15–16
Tensor-Analysis-for-Engineers_2E_CH19_Index.indd 181
U unit vector, 6, 7
W worked-out examples active rotation using single-axis and quaternions methods, 166–168 application of contravariant metric tensors, 155 complex double dot-cross product expressions, 165–166 continuity equations in several coordinate systems, 160–161 conversion from vector to index notations, 144–145 covariant derivatives of a constant vector, 160 covariant derivatives of metric tensors, 166 covariant derivatives of physical components of a vector, 161 determinant of a 4 × 4 matrix using permutation symbols, 159 determinant of metric tensors using displacement vectors, 158 dot and cross products in cylindrical and spherical coordinates, 155–156 Einstein summation convention, 143 4D spherical coordinate systems, 162–165 oblique rectilinear coordinate systems, 145–148 passive rotation using single-axis and quaternions methods, 168–169 quantities related to bi-polar coordinate systems, 152–154 quantities related to parabolic coordinate system, 149–151 relation between Jacobian and metric tensor determinants, 157 successive rotations using quaternions method, 170 time derivatives of the Jacobian, 159–160
10/3/2020 11:05:08 AM
Tensor-Analysis-for-Engineers_2E_CH19_Index.indd 182
10/3/2020 11:05:08 AM